Chapter  100:  Criteria to Determine Disability Related to Multiple Sclerosis

Purpose

The purpose of this project, nominated by the Social Security Administration (SSA) and contracted through the Agency for Healthcare Research and Quality (AHRQ), was to determine whether current medical knowledge supports the SSA's stated policies regarding MS. In January 2003, the Duke Evidence-based Practice Center began work on this 13-month task to review evidence from the medical literature for use in updating SSA's listing of impairments for multiple sclerosis (MS) and for revising its disability policy (if indicated).

Background

The Social Security Administration runs the world's largest disability program and processes more than 3.5 million claims each year. Multiple sclerosis is the third most common neurological diagnosis cited as the cause for disability. SSA uses the most stringent criteria of any disability program in the world to define disability.¹

Knowledge of the terms used in the SSA disability evaluation process, components of that process, and Medical Listing criteria related to MS is critical to the reader's understanding of this report. To assist in the preparation of the report, SSA provided explanations of terms and processes as currently defined by SSA regulations and rulings. The terms cited below, as well as other terms and processes used by SSA for disability determination, are defined and described in the SSA publication, Disability Evaluation Under Social Security 2003. ²

The statutory definition of “Disability” is: The inability to engage in any substantial gainful activity by reason of a medically determinable physical or mental impairment(s) which can be expected to result in death or which has lasted or can be expected to last for a continuous period of not less than 12 months. This definition differs from the clinically used definition of the World Health Organization's International Classification of Impairments, Disabilities and Handicaps (1980),³ which defines disability as “any restriction or lack of ability to perform an activity in a manner or within the range considered normal for a human being.” While much of the medical literature uses the latter, broader definition, the reader must be aware that the goals of this report relate to the statutory definition.

The following terms are defined by current (2003) SSA regulations:

“Claimant” is anyone who has filed a disability claim.

“Substantial Gainful Activity” is the ability to earn an average of $800 per month.

“Medically Determinable Impairment” is a physical or mental impairment that results from anatomical, physiological, or psychological abnormalities which can be shown by medically acceptable clinical and laboratory diagnostic techniques.

“Evidentiary Requirements” for disability determination are described by SSA regulation. An acceptable medical source must report signs, symptoms, and laboratory findings diagnostic of an impairment. Although a claimant's reported signs and symptoms are not sufficient to meet the evidentiary requirements for establishing the presence of a medically determinable impairment, all available evidence including the claimant's report of symptoms is used to evaluate the impact of any documented impairment(s) on the claimant's ability to carry out work tasks.

“Severe Impairment” is defined by the agency as any “impairment that more than minimally limits the claimant's ability to do basic work activities.”

The regulations include a Listing of Impairments for each body system that define disability. Often referred to as the “medical listings,” this list allows quick disability determinations to be made on the basis of medical criteria alone. The SSA publication, Disability Evaluation Under Social Security 2003, ² under the neurological category of impairments, includes Listing 11.09.

11.09 Multiple Sclerosis with:

1. Disorganization of motor function as described in 11.04; or

2. Visual or mental impairment as described under the criteria in 2.02, 2.03, 2.04, or 12.02: or

3. Significant, reproducible fatigue of motor function with substantial muscle weakness on repetitive activity, demonstrated on physical examination, resulting from neurological dysfunction in areas of the central nervous system known to be pathologically involved by the multiple sclerosis process.

All pages pertinent to the Medical Listing for multiple sclerosis, including the imbedded references to sections 2, 11, and 12, are found in Appendix A.

“Residual Functional Capacity” is assessed when a claimant is determined to have a “severe” impairment that does not meet or equal the intent of the medical listings. Physical capacity (lifting, carrying, walking, standing, sitting, pedaling, etc.) and mental capacity (cognitive and behavioral, thought processing, concentration, pace, behavior) are assessed in determining residual functional capacity.

In order to adjudicate claims by individuals with MS for disability benefits, SSA must determine whether the claims file includes information from an acceptable medical source that documents the signs, symptoms, and laboratory findings that are diagnostic of a physical or mental impairment. SSA adjudicators also determine whether the impairment would be expected to more than minimally interfere with the claimant's capacity to carry out basic work activities for at least 12 consecutive months or end in death. If a severe impairment is identified, the adjudicator determines whether the medical findings meet or equal an impairment in the medical listings. If the documented impairment does not meet or equal a listed impairment, the adjudicator must determine the claimant's residual functional capacity and consider vocational factors prior to making a final disability determination.

Research Questions

This evidence report covers five major topic areas framed within seven research questions, all of which are targeted to the adult population with MS. Our primary goal was to identify, review, and evaluate the published literature to answer the research questions; our secondary goal was to identify areas where no evidence exists or where the evidence has important limitations and then describe the type of data that would be needed to more fully address the question.

The questions are listed below by topic area, along with a brief description of our analytical approach, including interventions and outcomes of interest.

Reliability of MS Diagnostic Criteria

Question 1a. What is the reliability of new McDonald criteria (incorporating supplementary information from radiologic and laboratory studies including magnetic resonance imaging [MRI], visual evoked potential [VEP], and cerebrospinal fluid [CSF] analyses) compared with long-term follow-up diagnosis of clinically definite MS according to the Poser criteria?

The major difference between the Poser criteria⁴ and the new McDonald criteria⁵ is the addition of MRI findings. Our approach to this question was to identify studies in two categories: (1) those that specifically compared the new McDonald criteria with the reference standard of long-term diagnosis of clinically definite MS according to Poser criteria; and (2) those that provided data on the accuracy of various MRI techniques, CSF, and VEP (paraclinical diagnostic techniques incorporated into the criteria) with regard to the diagnosis of MS and thus supported their use as a supplement to clinical diagnosis.

In reporting results, the focus was on both relative measures (e.g., Hazard ratios) and absolute rates (e.g., percentages of patients with or without positive CSF who met Poser criteria at long-term follow up), with a primary focus on the latter.

Question 1b. What is the inter-rater reliability of diagnosis of MS according to Poser or McDonald criteria among neurologists or between neurologists and non-neurologist physicians?

The relevant diagnostic criteria were the Poser and new McDonald criteria plus any other clinical, laboratory, neurological exam, MRI, CSF, VEP, or other data supporting the MS diagnosis. Results had to describe data on agreement or disagreement on the MS diagnosis between evaluating physicians. Agreement statistics could include kappa scores, sensitivity and specificity rates, or other data of the type that could be used in completing a two-by-two table.

Predictors of Physical and Mental Impairments at 12 Months

Question 2. What clinical indicators, including particularly time-course of impairments, predict physical or mental impairment at 12 months?

There were four main categories of clinical predictors of particular interest to the analysis: (1) clinical characteristics such as exacerbation rates, disease type, age at disease onset, sex, degree of remission after relapse, and type and number of neurological symptoms; (2) imaging studies, particularly MRI; (3) laboratory test results such as apolipoprotein E (APOE) ε4 allele and intrathecal immunoglobulin-M (IgM) synthesis; and (4) self-reported health status using validated scales.

The evaluation of studies for this question was limited to those with a time course of 12 months (SSA's statutory limit), a timeframe which treating physicians would not ordinarily consider an important decision point. For this disease, the course has typically been studied over time horizons of many years.

Effect of Treatment and Symptom Management on Disease Course

Question 3a. Among patients with MS, do current disease-modifying treatments result in long-term improvements in physical or mental outcomes compared to placebo or usual care?

Interventions of interest for this question were all current (2003) disease-modifying immunomodulatory treatments (interferons and glatiramer acetate) and immunosuppressive treatments (e.g., azathioprine, mitoxantrone, cyclophosphamide, intravenous [IV] immunoglobulin-G [IgG]).

Outcomes of interest were physical functioning (primarily Expanded Disability Status Scale [EDSS] scores), proportion of patients with “improvement,” relapse frequency, cognitive functioning, quality of life, and adverse events.

Question 3b. Among patients with MS, do treatments aimed at symptom management result in improvements in physical or mental outcomes compared to usual care?

The effectiveness of symptomatic therapies for spasticity, rehabilitation, urinary management, fatigue, depression, and cognitive impairment was evaluated. Relevant outcomes were analyzed within six categories: (1) symptom-specific functional status or quality-of-life outcomes; (2) physical functioning (primarily EDSS); (3) cognitive functioning; (4) work or employment outcomes; (5) generic quality-of-life outcomes; and (6) adverse events.

The analysis of studies relevant to Questions 3a and 3b was complicated by issues of definition, particularly for outcomes reporting “improvement,” “long-term improvement,” and “relapse rates.” Our reporting of the results and subsequent analysis are presented within SSA's regulatory definition of “disability,” which considers physical or mental impairments that can be expected to result in death or which have lasted or can be expected to last for a continuous period of not less than 12 months.

Association of Clinical Findings with Work Ability

Question 4. Among individuals with MS, what physical, mental, laboratory, or radiographic findings have been associated with inability to work?

The phrasing of this question predetermined the outcome of interest as ability to work. Findings reported as absolute and relative measures of physical and mental/cognition function and laboratory and radiographic testing related to work activity were assessed.

Environmental Factors and Work Ability

Question 5. Among individuals with MS, how does elevated temperature or other environmental factors impair the capacity to work?

This question was interpreted as the association of workplace environmental conditions and demands (specifically, ambient temperature, individual's body temperature, or exposure to heat or cold) on the ability of an individual with MS to work. Relative and absolute measures of association were assessed.

Limitations of Report

In requesting this evidence report, SSA sought evidence from the medical and scientific literature to determine whether current medical knowledge supports SSA's stated policies regarding MS. Seven specific questions were framed within five topic areas. The information compiled in this report may enable SSA, for example: (1) to improve consistency of disability claims by applying more objective criteria, but only if the criteria are valid; (2) to change the population eligible for disability through a change in the diagnostic criteria for MS; (3) to influence changes in treatment that might reduce the number of people permanently disabled by MS; and (4) beyond motor and cognitive impairments, to consider how other significant symptoms, such as fatigue and urinary urgency, may be incorporated into considerations of disability status.

We believe the evidence presented in this report could also be used as the basis for a consensus conference of multidisciplinary experts on Listing of Impairments for MS that would employ formal consensus methods to update the current listing, as well as possibly expanding the disability process to include sociocultural factors that impinge upon desiring, seeking, finding, acquiring, and sustaining a job.

Literature Search and Review

Abstract and Full-text Screening Criteria

The seven specified questions spanned several topic areas and produced a considerably large and varied literature, which complicated the screening process. For each question, we developed fairly detailed inclusion/exclusion criteria. The titles and abstracts of the 1487 articles were reviewed against these criteria by at least two of five clinical investigators (“title-and-abstract screening” stage). Where no abstract was available, the title, source, and keywords were screened. At this stage, articles were included if requested by one investigator. The full text of each article passing the title-and-abstract screening was obtained for further review.

At the “full-text review” stage, each article was independently evaluated by two investigators, who forwarded their decisions to the task order manager for recording and comparison. If indicated, reviewers were asked to reconcile differences of opinion and return a reconciled final decision. If reviewers had difficulty reaching agreement, or submitted indecisive codes, the principal investigator was the arbiter.

We developed detailed screening instructions for the title-and-abstract screening (Appendix C) and additional decision rules for the full-text screening (Appendix D). For the full-text screening, we also produced a summary decision sheet, on which screeners recorded their include/exclude decision, research question assignment, and specific exclusion criterion (if appropriate) for each article.

Table 1. Results of title-and-abstract screening and (more...)

Table 1. Results of title-and-abstract screening and full-text article review

Articles identified:	1487
Abstracts:
Included	739
Excluded	748
Full-text articles:
Included	168
Excluded	571

Table 2. Included full-text articles by research question (more...)

Table 2. Included full-text articles by research question

Question 1a (diagnostic reliability of McDonald criteria)	13
Question 1b (inter-rater reliability of diagnosis with McDonald and Poser criteria)	2
Question 2 (predictors of physical and mental impairments at 12 months)	12
Question 3a (disease-modifying therapies and long-term improvement)	51
Question 3b (symptom management and improvement)	68
Question 4 (association of clinical findings with work ability)	22
Question 5 (environmental factors and work ability)	1
Total	168*

*

Note: One article was included for both Question 4 and Question 5

Summaries of the results of the title-and-abstract screening and full-text review are provided in Tables 1 and 2.

Diagnostic Reliability of McDonald Criteria

Introduction

This section addresses results for Question 1a: What is the reliability of new McDonald criteria (incorporating supplementary information from radiologic and laboratory studies including magnetic resonance imaging [MRI], visual evoked potential [VEP], and cerebrospinal fluid [CSF] analyses) compared with long-term follow-up diagnosis of clinically definite multiple sclerosis (MS) according to the Poser criteria?

The diagnosis of MS has traditionally been based on clinical history and examination, and on the absence of an alternative diagnosis. Paraclinical tests such as imaging techniques,⁹ evoked potentials,¹⁰ and CSF analyses have been used to aid diagnosis, but their precise role in formal diagnostic criteria for MS has been a source of debate. With the advent of MRI, which provides much more detailed images of the brain and spinal cord than were possible with computed tomography (CT), debate over the role of paraclinical tests in the diagnosis of MS has been renewed.

Of the various diagnostic criteria proposed for MS, the most widely used have been those by Poser et al. (1983).⁴ The Poser criteria focus on objective examination evidence of abnormalities in at least two separate areas of the central nervous system, with historical information to suggest at least two periods of MS involvement over time. The criteria may be summarized as follows:⁴

• Clinically definite MS:

  • - 2 attacks and clinical evidence of 2 separate lesions; or

  • - 2 attacks; clinical evidence of 1 lesion and paraclinical evidence of another, separate lesion.

• Laboratory-supported definite MS:

  • - 2 attacks; clinical or paraclinical evidence of 1 lesion; and CSF immunologic abnormalities; or

  • - 1 attack; clinical evidence of 2 separate lesions; and CSF immunologic abnormalities; or

  • - 1 attack; clinical evidence of 1 lesion and paraclinical evidence of another, separate lesion; and CSF immunologic abnormalities.

• Clinically probable MS:

  • - 2 attacks and clinical evidence of 1 lesion; or

  • - 1 attack and clinical evidence of 2 separate lesions; or

  • - 1 attack; clinical evidence of 1 lesion and paraclinical evidence of another, separate lesion.

• Laboratory-supported probable MS:

  • - 2 attacks and CSF immunologic abnormalities.

In light of the growing potential of paraclinical tests - particularly MRI - to contribute to the diagnosis of MS, the International Panel on the Diagnosis of Multiple Sclerosis met in July 2000 to reassess existing diagnostic criteria. The panel recommended changes to the criteria that (1) integrated MRI into the overall diagnostic scheme and (2) provided for the diagnosis of primary progressive disease.⁵ The new criteria (here referred to as the McDonald criteria)⁵ were designed to be used by practicing physicians and adapted, as necessary, for clinical trials. They dropped the Poser categories of “clinically definite,” “laboratory-supported definite,” “clinically probable,” and “laboratory-supported probable” MS and proposed that the disease be diagnosed using the following criteria:⁵

• 2 or more attacks; objective clinical evidence of 2 or more lesions

  • - No additional data needed

• 2 or more attacks; objective clinical evidence of 1 lesion; plus

  • - Dissemination in space, demonstrated by:

    • MRI, or

    • 2 or more MRI-detected lesions consistent with MS plus positive CSF, or

    • Further clinical attack implicating a different site.

• 1 attack; objective clinical evidence of 2 or more lesions; plus

  • - Dissemination in time (demonstrated by MRI).

• 1 attack; objective clinical evidence of 1 lesion (monosymptomatic presentation, clinically isolated syndrome); plus

  • - Dissemination in space, demonstrated by:

    • MRI, or

    • 2 or more MRI-detected lesions consistent with MS plus positive CSF, and

  • - Dissemination in time, demonstrated by:

    • MRI or

    • Second clinical attack.

• Insidious neurological progression suggestive of MS; plus

  • - Positive CSF, and

  • - Dissemination in space, demonstrated by:

    • 9 or more T2 lesions in brain, or

    • 2 or more lesions in spinal cord, or

    • 4–8 brain lesions plus 1 spinal cord lesion, or

    • abnormal VEP associated with 4–8 brain lesions, or

    • abnormal VEP with fewer than 4 brain lesions plus 1 spinal cord lesion; and

  • - Dissemination in time, demonstrated by:

    • MRI, or

    • Continued progression for 1 year.

The publication of the McDonald criteria has renewed the discussion of the value of incorporating paraclinical testing, particularly MRI, in the diagnosis of MS. The Social Security Administration (SSA), aware of the difficulties brought on by improper or inadequate diagnosis, has sought evidence from the literature regarding the value of the various diagnostic criteria. Question 1a seeks specific information regarding the reliability of the McDonald criteria.

In this review, we sought to identify specifically those studies that evaluated whether patients diagnosed with MS according to the McDonald criteria would later meet the Poser criteria for clinically definite MS. We also sought articles that, although not specifically using the McDonald criteria, may have evaluated components of the McDonald criteria against later diagnosis of MS according to the Poser criteria.

There exists a significant literature regarding the diagnostic utility of newer MRI techniques, such as magnetization transfer or various measures of atrophy,^11–21 but we did not include such articles as they are not a part of currently used diagnostic criteria in general or the McDonald criteria specifically.

Results

Table 3. Data supporting the validity of MRI components (more...)

Table 3. Data supporting the validity of MRI components of the McDonald criteria

Study	Diagnosis	Time to diagnosis of CDMS	Para-clinical test(s)	Findings associated with CDMS	Proportion developing CDMS	SN	SP
Barkhof et al. 199722	CIS	Median follow up (with range): 9 mo (1–48 mo) for patients with CDMS (n = 33); 39 mo (23–96 mo) for patients without CDMS (n = 41)	MRI	No. of T2 lesions ≥ 9	24/30 (80%)	73%	80%
				No. of abnormal MRI criteria:*
				0	16%
				1	11%
				2	54%
				3	75%
				4	87%
				Specific MRI criteria:
				Final model		82%	78%
				Paty et al. 198835		88%	54%
				Fazekas et al. 198836		88%	54%
Brex et al. 200123	CIS	1-year follow up (n = 68)	MRI	T2 lesions	16/48 (33%)	89%	36%
				Enhancing lesions	11/21 (52%)	61%	80%
				T2 lesion at baseline and new T2 lesion	15/27 (56%)	83%	76%
				Enhancing lesion at baseline and new enhancing lesion	7/10 (70%)	39%	94%
CHAMPS Study Group 200224	CIS	18-month follow up (n = 190)	MRI	No. of T2 lesions:
				2–4	20%
				5–8	15%
				9–12	33%
				13–21	33%
				22–34	26%
				> 35	20%
				No. of enhancing lesions:
				0	23%
				1	33%
				> 1	43%
				> 2	52%
Comi et al. 200125	CIS	2 years (n = 241)	MRI	No. of T2 lesions:	NR (OR, 3.64; 95% CI, 1.3 to 10.2; p = 0.014)
				0–8
				> 8
Filippi et al. 199428	CIS	Mean follow up, 63 months (n = 89)	MRI	Initial MRI normal	2/32 (6%)
				Initial lesion load > 1.23 cm3	19/21 (90%)
				Initial MRI abnormal, but lesion load < 1.23 cm3	17/31 (55%)
Morrissey et al. 199330	CIS	5-year follow up (n = 89)	MRI	No. of lesions:
				0	2 (6%)
				1	1 (17%)
				2–3	12 (67%)
				4–10	12 (92%)
				> 11	16 (80%)
Optic Neuritis Study Group 200132	Optic Neuritis	5-year follow up (n = 388)	MRI	No. of lesions:
				0	16%
				1	44%
				2	26%
				> 3	51%
O'Riordan et al. 199831	CIS	10-year follow up (n = 81)	MRI	No. of asymptomatic lesions at baseline MRI:
				0	3/27 (11%)
				1	1/3 (33%)
				2–3	14/16 (87%)
				4–10	13/15 (87%)
				> 10	17/20 (85%)
Sastre-Garriga et al. 200333	CIS (brainstem syndrome)	12-month follow up (n = 51)	MRI	Abnormal MRI	17/46 (37%)	94%	42%
			CSF	Abnormal CSF-OCB	11/25 (44%)	100%	42%
			EPs	Abnormal EPs	10/29 (34%)
				Specific MRI criteria:
				Paty et al. 198835	16/32 (50%)	89%	52%
				Fazekas et al. 198836	16/33 (48%)	89%	48%
				Barkhof et al. 199722	14/27 (52%)	78%	61%

*

Abnormal MRI criteria: ≥ 1 gadolinium-enhancing lesions, ≥ 1 juxtacortical lesions, ≥ 3 periventricular lesions, and ≥ 1 infratentorial lesions.

Abbreviations: CDMS = clinically definite multiple sclerosis; CI = confidence interval; CIS = clinically isolated syndrome; CSF = cerebrospinal fluid; EPs = evoked potentials; mo = month(s); MRI = magnetic resonance imaging; No. = Number; OCB = oligoclonal bands; OR = odds ratio; SN = sensitivity; SP = specificity

Table 4. Studies validating the McDonald criteria

Table 4. Studies validating the McDonald criteria

Study	Diagnosis	Time from onset of symptoms	Time to diagnosis of CDMS	Sensitivity	Specificity	PPV	NPV	Accuracy
Dalton et al. 200227	CIS	3 months (n = 79)	1 year	0.73	0.87	0.58	0.93	0.84
		1 year (n = 50)	3 years	0.94	0.83	0.77	0.96	0.87
Tintoré et al. 200334	CIS	1 year (n = 86)	49 months (mean follow up)	0.74	0.86	NR	NR	0.80

Abbreviations: CDMS = clinically definite multiple sclerosis; CIS = clinically isolated syndrome; NPV = negative predictive value (predictive value of a negative test result); NR = not reported; PPV = positive predictive value (predictive value of a positive test result)

Using the Poser criteria as the diagnostic gold standard in this way is problematic, mainly because the limited duration of any clinical study means that some patients may be classified as non-diseased at the conclusion of a study who later go on to meet the Poser criteria for MS. Once the Poser criteria are defined as the gold standard, other tested criteria can at best correlate highly with Poser diagnosis. It is possible that the McDonald criteria might perform better than the Poser criteria (in terms of sensitivity, ability to diagnose MS early, and/or association with prognosis) if both could be compared to a (hypothetical) true gold standard. The high level of agreement of McDonald criteria with later Poser criteria diagnosis observed in the studies reviewed here is not inconsistent with this possibility.

Inter-rater Reliability of Diagnosis with McDonald and Poser Criteria

Results

Two studies met our inclusion criteria (see Evidence Table 1b in Appendix F).^{37, 38} Both examined inter-rater reliability of neurologist-physicians in diagnosing MS according to the Poser criteria. One of the studies³⁸ also examined inter-rater reliability for diagnosing MS according to the McDonald criteria. We found no data examining the inter-rater reliability among other clinicians.

Ford et al. (1996)³⁷ examined the inter-rater reliability of the Poser criteria. Overall, there was substantial agreement between the two observers in classifying MS (kappa = 0.65; 95 percent confidence interval [CI], 0.52 to 0.78). There was disagreement as to whether a patient had one or more “attacks” of MS. Agreement was substantial for clinical evidence of separate lesions and was almost perfect for both paraclinical evidence and laboratory support (all kappa values > 0.90). The primary disagreement was in defining “attacks” of MS, and the authors appropriately noted that this might be due to the retrospective nature of the evaluation.

The primary shortcomings of this study were that only two evaluators were compared, and diagnoses were based on retrospective analysis of data from clinical records. This process is significantly different from prospective diagnosis. Prospective diagnosis allows the clinician access to detailed information that may be inadequately recorded in the medical record, such as lesion location, intensity of enhancement, and overall appearance of MRI changes. Likewise the details of clinical exacerbations may be communicated to an examining physician making prospective diagnoses, but may be inadequately or incompletely recorded in the medical record. In general, retrospective diagnosis is frequently based on relatively less complete information as recorded in the medical record.

Table 5. Kappa statistics for multiple raters using (more...)

Table 5. Kappa statistics for multiple raters using Poser and McDonald criteria

Poser criteria	McDonald criteria
Diagnosis of MS (all categories): 0.57	Diagnosis of MS (all categories): 0.57
Clinically definite MS: 0.39	MS: 0.57
Clinically probable MS: 0.37	Possible MS: 0.49
Dissemination in time: 0.69	-
Dissemination in space: 0.46	-

Abbreviation: MS = multiple sclerosis

Zipoli et al. (2003)³⁸ specifically addressed Question 1b by examining the inter-rater reliability of both the Poser and McDonald criteria (see Table 5). In this study, four neurologists assessed all patients consecutively admitted for diagnosis to a university department of neurology from September 2001 through June 2002 who had been followed for at least 6 months. The mean follow up was 12.7 months. The study evaluated 41 MS patients, of whom 15 had relapsing-remitting MS, two had secondary progressive MS, five had primary progressive MS, and 19 had clinically isolated syndromes. Three additional patients who were not diagnosed with MS were included in the evaluation. Data were abstracted from medical records onto standardized forms by a non-evaluating neurologist. No diagnoses were recorded. The four evaluating neurologists were stated to have similar clinical experience in MS diagnosis and management. The study does not report any specific training of these neurologists with regard to the McDonald criteria or MRI analysis. The four evaluating neurologists reviewed the standardized forms and all MRI scans independently without discussing their findings. Eighteen patients had follow-up MRI scans in addition to baseline scans.

In this study, the primary difficulty in the McDonald criteria appeared to be decreased agreement in MRI interpretation, specifically in those patients with high lesion loads. The authors commented that this study utilized neurologist evaluators rather than neuroradiologists. Previous studies have correlated level of radiographic training with agreement in interpretation.^{39, 40} Judging dissemination in time was of particular difficulty in those patients with clinically isolated symptoms. The authors suggested that neuroradiologists be encouraged to interpret scans in MS patients with the McDonald MRI criteria in mind, providing specific information regarding lesion location and timing.

Predictors of Physical and Mental Impairments at 12 Months

Introduction

Figure 1. Temporal relationship of pre-disability evaluation (more...)

   Figure 1. Temporal relationship of pre-disability evaluation and occurrence of work disability

The experiences of three individuals over a 9- to 24-month time frame are depicted. Variability in the point at which an individual becomes unable to work is shown by the variable position of the arrow labeled “work disability.” Variability in the time of assessment of prognostic data is depicted by variation in the position of the left-pointing arrow. Variability in the duration of follow up is depicted by variation in the position of the right-pointing arrow. Abbreviations: EDSS = Expanded Disability Status Scale; MS = multiple sclerosis.

This section examines the evidence pertaining to Question 2: What clinical indicators, including particularly time-course of impairments, predict physical or mental impairment at 12 months? The notion here is that an individual diagnosed with MS but not yet work disabled could, within the follow-up period of disability evaluation, become work disabled. In formulating a response to this question, we considered three crucial study features: the assessment of possible prognostic features, the period of follow up, and the assessment of a valid measure of physical or mental impairment. The assessment of possible prognostic features had to occur between the time of diagnosis of MS and the occurrence of work disability; in most studies this assessment occurred at or near study entry, but this was not necessary. The 12-month timeframe specified in this question was the major limiting criterion in the number of studies that could be included for evaluation; to expand the pool of studies, we included any evaluations in which follow up occurred within a window of 9 to 24 months from study entry (or measurement of the purported prognostic feature). Thus, a study that related prognostic data available at one time with outcome data at another time 9 to 24 months later would be eligible; furthermore, a study might provide such data over multiple different time frames (see Figure 1). In all cases, the available outcome measure was the Disability Status Scale (DSS) or the Expanded Disability Status Scale (EDSS). For purposes of judging a level approximately associated with work disability, we focused on the transition from an EDSS score of less than 6 to greater than 6. Since deficits associated with MS unrelated to acute exacerbations are not expected to regress, we did not include studies in which patients were already severely disabled.

Disease-modifying Therapies and Long-term Improvement

Results

Fifty-one publications^53–103 describing 34 separate trials, met our inclusion criteria (see Evidence Table 3a in Appendix F). The treatments evaluated included interferon beta, glatiramer acetate, mitoxantrone, glucocorticoids, intravenous (IV) immunoglobulin-G (IgG), azathioprine, cyclophosphamide, plasma exchange, methotrexate, cladribine, cyclosporine, and combinations of these therapies. Outcome measures were primarily group or mean changes in rates of relapses, changes in rates of progression on the EDSS, and occasionally changes in some measures of cognition or quality of life. These studies are adequate to answer issues regarding the efficacy of therapy, but do not provide information regarding individual patient improvement.

Table 6. Relapse rate outcomes from placebo-controlled (more...)

Table 6. Relapse rate outcomes from placebo-controlled RCTs of disease-modifying therapies

Study	Therapy	Relapse rate parameter	Active	Placebo	Difference	P-value
Achiron et al. 199853	IV IgG	ARR:
		Baseline	1.85	1.55	0.30	0.34
		Year 1	0.75	1.80	1.05	0.0002
		Year 2	0.42	1.42	1.0	0.0009
Fazekas et al. 1997a63	IV IgG	ARR:
Fazekas et al. 1997b64		Year 1	0.49	1.30	0.81	0.011
		Year 2	0.42	0.83	0.41	0.006
Cohen et al. 200259	IFNβ-1a	ARR	0.20	0.30	0.10	0.008
Jacobs et al. 199676	IFNβ-1a	ARR	0.67	0.82	0.15	0.04
Rudick et al. 199797
Fischer et al. 200065
Jacobs et al. 200075
Rudick et al. 200096
SPECTRIMS Study Group 2001 100	IFNβ-1a	ARR	0.50	0.71	0.21	0.001
European Study Group on Interferon beta-1b in Secondary Progressive MS 199862	IFNβ-1b	ARR	0.44	0.64	0.20	0.0002
IFNB Multiple Sclerosis Study Group 199372	IFNβ-1b	ARR:
IFNB Study Group 199573		Year 1	0.96	1.44	0.48
IFNB Study Group 199674		Year 2	0.85	1.18	0.33
Pliskin et al. 199692		Year 3	0.66	0.92	0.26
		Year 4	0.67	0.88	0.21
		Year 5	0.57	0.81	0.24
PRISMS Study Group 199893	IFNβ-1a	Relapse rate, not annualized, ~ 2 years	1.73	2.56	0.83
Liu et al. 199982
Liu et al. 200284
Patten et al. 200189
Bastianello et al. 199454	Mitoxantrone	Mean RR	0.54	1.67	1.13	0.014
Millefiorini et al. 199786	Mitoxantrone	No. of relapses	0.89	2.62	1.73	0.0002
Johnson et al. 1995 78	Glatiramer	ARR	0.59	0.84	0.25	NR
Weinstein et al. 1999 103
Liu et al. 2000 83
Johnson et al. 199877
Goodkin et al. 199168	Azathioprine	ARR:
		Year 1	0.74	1.17	0.43	0.16
		Year 2	0.30	0.79	0.49	0.05
Romine et al. 199995	Cladribine	Relapse rates, annualized (?)	0.77	1.67	0.90	NR

Abbreviations: ARR = annual relapse rate; IFNβ-1a = interferon β-1a; IFNβ-1b = interferon β-1b; IV IgG = intravenous immunoglobulin-G; No. = Number; NR = not reported; RCTs = randomized controlled trials; RR = relapse rate

Relapse rates. Data on relapse rate outcome measures are summarized in Table 6.^{53, 54, 59, 62–65, 68, 72–78, 82–84, 86, 89, 92, 93, 95–97, 100, 103} In most of these studies the baseline mean annual relapse rate was approximately one relapse per year. The effect of most of the therapies studied was a mean reduction in the range of 0.3 relapses per year; higher reductions (up to one relapse per year) were reported after treatment with IV IgG when baseline relapse rates were in the range of 1.5 to 2 relapses per year. We found only one study that reported reductions in individual patient relapse rates. Kappos et al. (2001)⁷⁹ reported that treatment with interferon beta increased the proportion of patients who were relapse-free or had decreased relapse rates from 45 percent (placebo) to 53.1 percent (treatment). In summary, the effect of treatment with any of the currently used therapies is a mean reduction in relapses of less than one relapse per year. The benefit may be higher in patients with higher baseline relapse rates, but we did not find data stratified in this manner.

Table 7. Improvements in physical function (EDSS) in (more...)

Table 7. Improvements in physical function (EDSS) in placebo-controlled RCTs of disease-modifying therapies

Study	Therapy	Definition(s) of improvement (change in EDSS)	Active	Placebo	Difference
Achiron et al. 199853	IV IgG	-1.0 point over 2 years	31%	14%	17%
Fazekas et al. 1997a63	IV IgG	-1.0 point over 2 years	31%	14%	16%
Fazekas et al. 1997b64
Cohen et al. 200259	IFNβ-1a	Improvement not defined	7.5%	7.3%	0.2% (NS)
Jacobs et al. 199676	IFNβ-1a	-1.0 point (not sustained)	19.3%	11.5%	8.2%
Rudick et al. 199797		-0.5 point (not sustained)	15.7%	11.5%	4.2%
Fischer et al. 200065		-1.0 point (sustained)	18.2%	8.9%	9.7%
Jacobs et al. 200075		-0.5 point (sustained)	25.5%	16.1%	9.4%
Rudick et al. 200096
Patti et al. 199990	nIFNβ	-0.5 or -1.0 point:
		Relapsing-remitting	52%	3%	49%
		Secondary progressive	40%	5%	35%
Johnson et al. 1995 78	Glatiramer	-1.0 point:
Weinstein et al. 1999 103		2-year trial	24.8%	15.2%	9.6%
Liu et al. 2000 83		Extension trial	27.2%	15.0%	12.2%
Johnson et al. 199877
Bornstein et al. 198755	Glatiramer	-1.0 point over 2 years	20.0%	8.7%	11.4%
		-2.0 points over 2 years	12.0	0	12.0%
		-3.0 points over 2 years	0	4.4%	-4.4%
Ghezzi et al. 198967	Azathioprine	Improvement not defined:
		Relapsing	16%	0%	16% (NS)
		Relapsing-progressive	5%	7%	-2% (NS)
Goodkin et al. 199168	Azathioprine	Improvement on EDSS or 9HPT	22.2%	20%	2.2%
van de Wyngaert et al. 2001102	Mitoxantrone	Improvement not defined	35%	22%	13% (NS)
Canadian Cooperative Multiple Sclerosis Study Group 199158	Cyclo-phosphamide	-1.0 point sustained for:
		1 year	6%	2%	4%
		2 year	6%	0%	6%
		Final	4%	2%	2%
Canadian Cooperative Multiple Sclerosis Study Group 199158	Plasma exchange	-1.0 point sustained for:
		1 year	8%	2	6%
		2 year	3%	0%	3%
		Final	2%	2%	0
Khatri et al. 198580	Plasma exchange	> 1.0 point at 11 months:
		≥ 3 points	10%	0	10%
		2 points	13%	3%	10%
		1 point	13%	14%	-1%

Abbreviations: 9-HPT = 9-Hole Peg Test; EDSS = Expanded Disability Status Scale; IFNβ-1a = interferon β-1a; IV IgG = intravenous immunoglobulin-G; nIFNβ = natural interferon-β; NS = not statistically significant; RCTs = randomized controlled trials

Physical function (EDSS). Studies reporting data on individual patient improvement in EDSS scores are summarized in Table 7.^{53, 55, 58, 59, 63–65, 67, 68, 75–78, 80, 83, 90, 96, 97, 102, 103} Most of these studies provided a definition (and in some cases more than one definition) of improvement, usually corresponding to at least a one-point decrease in EDSS score. Changes of half a point (0.5) in EDSS may not be reliable due to uncertainty regarding this degree of change and uncertain clinical significance. The proportion of patients meeting varied definitions of improvement in placebo groups ranged from 0 to 22 percent; the differences in proportions of patients improving ranged from -1 to 49 percent, although most fell into the range of 0 to 12 percent. Higher EDSS change thresholds (-2 points or -3 points) resulted in lower proportions of patients meeting improvement definitions in one study.⁵⁵

Improvement in cognitive function. Two included studies evaluated the efficacy of therapy with regard to cognitive function.^{76, 78} Both studies demonstrated benefit of therapy over placebo but did not present data with regard to the quantitative improvement seen in individual patients.

Improvement in quality-of-life measures. Two included studies (described in three publications) evaluated the efficacy of therapy with regard to quality-of-life measures.^{59, 66, 79} Both studies demonstrated benefit of therapy over placebo in the groups studied, but did not present data with regard to the quantitative improvement seen in individual patients.

Symptom Management and Improvement

Results

We included a total of 68 articles describing 65 separate studies (see Evidence Table 3b in Appendix F). By topic, these included: for spasticity, 35 articles^104–138 describing 32 separate studies; for rehabilitation, 10 articles/studies;^139–148 for depression, eight articles/studies;^149–156 for fatigue seven articles/studies;^157–163 for voiding dysfunction, seven articles/studies;^164–170 and for cognitive problems, one article/study.¹⁷¹

Spasticity. Among the 32 studies of specific treatments aimed at management of spasticity were randomized controlled trials (RCTs) of oral medications, intramuscular medications, and intrathecal medications, as well as physical treatments (magnetic stimulation and electrical neuromuscular stimulation). Trials of drug treatments were most common and considered the following agents: tizanidine, baclofen, diazepam, gabapentin, dantrolene, threonine, botulinum toxin, cannabinoids/delta-9-tetrahydrocannabinol (delta-9-THC), progabide, and amantadine. Baclofen was the subject of six comparisons with placebo,^{105, 109, 117, 124, 129, 130} six comparisons with tizanidine,^{104, 108, 113, 122, 126, 134, 136} two comparisons with diazepam,^{106, 110} and one with stretching exercises.¹⁰⁵ In addition, one trial compared intrathecal baclofen with placebo.¹²⁵

Tizanidine was studied in two placebo comparisons,^{133, 137} six comparisons with baclofen (see above), and one with diazepam.¹²⁷

Dantrolene was studied in one placebo comparison¹¹¹ and one comparison with diazepam.^{131, 132} We found two studies each of gabapentin,^{107, 121} progabide,^{119, 120, 128} botulinum toxin,^{114, 135} and threonine^{112, 116} versus placebo. Delta-9-THC was tested in two trials^{115, 138} against both placebo and cannabis sativa plant extract. Electrical¹¹⁸ and magnetic¹²³ neuromuscular stimulation were each compared to sham stimulation.

Study quality. All 32 trials randomly allocated patients to treatment groups, but the methods of randomization were clearly described for only 5/32 (16 percent). Most failed to provide enough information to determine whether there had been adequate concealment of the allocation schedule (25/32; 78 percent). In the seven studies that reported enough information for evaluation, there was adequate concealment in four, and inadequate concealment in three.

Thirty of 32 trials (94 percent) were described as double-blind; patients were described as blinded in 29 trials, and in one trial¹¹⁸ it was unclear whether patients were blinded. Treating investigators were blinded in 28 trials; outcome assessors were blinded in 30 trials, but it was unclear whether they were blinded in another study.¹⁰⁵ Most studies (27/32; 84 percent) reported the number of withdrawals in each treatment group.

Twenty of the 32 studies were of crossover design; the remaining trials were of parallel-group design. Among the crossover trials, we assessed several dimensions of quality unique to the crossover design. The presence of period effects or carry-over effects was not discussed in over half the studies (11/20; 55 percent). One trial reported having a period or carry-over effect and appropriately analyzed first-period data only.¹²¹ Most crossover trials (17/20; 85 percent) included a washout period. Thirteen (65 percent) failed to report clearly the number of patients assigned to each treatment sequence. In six studies it was unclear whether patients who did not complete both treatment periods were excluded; two studies appeared to use an inappropriate unpaired analysis retaining patients who did not complete the crossover in the final analysis.

Study populations. Only 11 of the studies describe the study population in terms of EDSS score at baseline; mean baseline EDSS ranged from 4.7 to 7.4 in studies reporting means. The two studies of botulinum toxin^{114, 135} appeared to include the most severely impaired patients, with median EDSS scores above 7.5. Other study populations included mostly subjects with EDSS scores in the range of 5.0 to 7.0.

Studies used a variety of outcome measures, many of which were not validated. Of the placebo-controlled trials, 12 of 20 (60 percent) used the Ashworth Scale or Modified Ashworth Scale. However, this scale was often combined with other measures¹³⁵ or summed across sides and muscle groups to create outcome variables and statistical testing that have not been validated.

Table 8. Randomized controlled trials of symptomatic (more...)

Table 8. Randomized controlled trials of symptomatic treatments for spasticity

Treatment	Study	Baseline EDSS	No. of patients started (completed)	Outcomes/Results
Baclofen PO (to 80 mg/d) vs.Placebo	Feldman et al. 1978109	NR	33	Spasm frequency (p < 0.05)
Crossover				Resistance to movement (p < 0.05)
				Clonus (knee) (p < 0.001)
				Ambulation (p = NS)
				Transfer activity (p = NS)
				Spastic limb pain (p = NS)
				Use of spastic limb (p = NS)
				Functional assessment (Barthel) (p = NS)
Baclofen PO (to 80 mg/d) vs. Placebo	Sachais et al. 1977129	NR	166 (106)	Flexor spasm severity (p < 0.02)
				Ankle clonus (p = NS)
				ADL (p = NS)
				Overall disability (p = NS)
Baclofen PO (to 45 mg/d) vs. Placebo	Ørsnes et al. 2000124	5.0 (median)	14	Voluntary power (0–5 scale) (p = NS)
Crossover		3.5–6.0 (range)		Ashworth Scale (p = 0.33)
				Tendon reflexes (p = 0.14)
				EDSS & Ambulation Index (p = NS)
				NRS (p = NS)
				MSIS (p = NS)
Baclofen PO (to 80 mg/d) vs. Placebo	Levine et al. 1977117	NR	19 (18)	Ashworth Scale (p = NS)
Baclofen PO (to 60 mg/d) vs.Placebo	Sawa et al. 1979130	NR	21 (18)	Muscle tone (p = NR)
Crossover
Baclofen PO (to 20 mg/d) vs. Stretching exercises + baclofen vs. Stretching exercises vs.Placebo	Brar et al. 1991105	NR	38 (30)	Spasticity: Cybex flexion scores (p < 0.05; baclofen and combination vs. placebo)
Crossover				Ashworth Scale (p = 0.1; combination vs. placebo)
				Self-rated questionnaire of functional abilities (p = NR)
Baclofen intrathecal vs. Placebo	Penn et al. 1989125	NR	20 (10 MS)	Ashworth Scale (p < 0.0001)
Crossover				Spasm score (p < 0.0005)
Tizanidine PO (to 36 mg/d) vs. Placebo	Smith et al. 1994133	NR	257 (159)	Ashworth Scale (p = 0.46)
				Spasm response (p = 0.03)
				Clonus response (p = NS)
Tizanidine PO (to 36 mg/d) vs. Placebo	UK Tizanidine Trial Group 1994137	NR	187 (155)	Ashworth Scale (p < 0.004)
				Spasm frequency (p = NS)
				Muscle strength (p = NS)
				EDSS (p = NS)
				Tendon reflexes
				Intermediate motor skills (p = NS)
				Upper limb function (p = NS)
				Impact on physical therapy (p = NS)
				Impact on nursing care (p = 0.09)
Gabapentin PO 2700 mg vs.Placebo	Cutter et al. 2000107	6.0–9.0 (range)	22 (21)	Spasm frequency scale (p = 0.0001)
Crossover				Spasm severity scale (p = NS)
				Interference with function scale (p = 0.02)
				Painful spasm scale (p = NS)
				Global assessment scale (p = 0.003)
				Modified Ashworth Scale (p = 0.0005)
				EDSS (p = NS)
Gabapentin PO 400 mg vs. Placebo	Mueller et al. 1997121	NR	15	Visual Faces Scale rating of pain and spasticity (p = 0.008)
Crossover				Ashworth Scale (p = 0.007)
				Clonus (p = 0.1)
				Reflex withdrawal to pain (p = NS)
				EDSS (p = 0.03)
Dantrolene PO (to 350 mg/d) vs. Placebo	Gambi et al. 1983111	NR	24 (22)	Hip flexor movement (p = NS)
Crossover				Spasticity scale (p < 0.05)
				Muscular strength (p = NS)
				Clonus (p = NS)
				Knee and ankle tendon reflexes scale (p = NS)
Delta-9-THC Vs. Cannabis sativa extract vs. Placebo	Zajicek et al. 2003138	NR	630 (611)	Muscle tone (p=0.4)
				Pain (p=0.003; favoring active treatments)
				Subjective spasticity (p=0.01)
				10-meter walk (p=0.015; favoring delta-9-THC)
Delta-9-THC vs. Cannabis sativa extract vs. Placebo	Killestein et al. 2002115	6.2 ± 1.2 (mean ± SD)	16	MSFC score (p = 0.09; favoring placebo)
Crossover				9-HPT (p = 0.02; favoring placebo)
				EDSS (p = NS)
				Muscle tone (p = NS)
				SF-36 mental health subscale (p = 0.02) and psychological status domain (p = 0.02)
				Other parameters (p = NS)
Progabide PO (to 45 mg/kg/d) vs. Placebo	Rudick et al. 1987128	6.3 ± 1.7 (mean ± SD)	32 (25)	Ashworth Scale (p < 0.01)
Crossover				EDSS (p = NS)
				8-meter walk (p = 0.62)
				Spasm frequency (p = 0.28)
				Reflex scores (p = 0.20)
				Strength (p = 0.77)
				Various functional tasks (p = NS)
Progabide PO (14.3–32.7 mg/kg/d) vs. Placebo	Mondrup et al. 1984119120	NR	17 (14)	Spasticity (0–4 scale) (p < 0.01)
Crossover				Spasms (p < 0.01)
				Reflex response (p < 0.01, patellar; p = NS, Achilles)
				Flexor reflexes (p = NS)
				Muscle strength (p = NS)
				Spasm frequency (p < 0.05)
				Spasm pain (p = NS)
Botulinum toxin IM vs. Placebo	Hyman et al. 2000114	7.5 (median [estimated])	74 (60)	Hip abduction (p = NS)
				Modified Ashworth Scale (p = NS)
				Spasm frequency (p = NS)
				Hygiene assessment (p = NR)
Botulinum toxin IM vs. Placebo	Snow et al. 1990135	8.0–9.5 (range)	10 (9)	Spasticity score (Ashworth Scale + spasm frequency) (p = 0.009)
Crossover				Hygiene score (p = 0.02)
Threonine PO (7.5 g/d) vs. Placebo	Hauser et al. 1992112	4.7 ± 1.5 (mean ± SD)	26 (21)	Spasticity (clinician scale) (p = 0.04)
Crossover				Spasticity (patient scale) (p = 0.18)
Threonine PO 6 g/d vs. Placebo	Lee et al. 1993116	7.4 (mean)	41 (33)	Spasticity score (based on Ashworth) (p = NR)
Crossover		2–9 (range)		Barthel Index (p = NS)
				EDSS (p = NS)
Electrical neuromuscular stimulation vs. Sham stimulation	Livesley 1992118	NR	40 (39)	Spasticity, active movement, and function (p = NS)
				Subjective evaluation (p = NR)
Magnetic stimulation vs. Sham stimulation	Nielsen et al. 1996123	NR	38	Ashworth Scale (p = NR)
				Tendon reflexes (p = NR)
				Patient's self-score for ease of daily activities (p = NS)
Tizanidine PO (to 32 mg/d) vs. Baclofen PO (to 80 mg/d)	Bass et al. 1988104	NR	66 (48)	Spasticity (6-point ordinal scale) (p = NS)
Crossover	Rice 1989126			Strength (7-point ordinal scale) (p = NS)
				EDSS (p = NS)
				Pedersen functional disability scale (p = NR)
				Overall evaluation (patient, clinician) (p = NR)
Tizanidine PO (to 36 mg/d) vs. Baclofen PO (to 80 mg/d)	Smolenski et al. 1981134	NR	21	Muscle strength (p = NR)
				Ashworth Scale (p = NR)
				Spasms scale (p = NR)
Tizanidine PO (to 24 mg/d) vs. Baclofen PO (to 60 mg/d)	Eyssette et al. 1988108	NR	100	Flexor spasms (p = NS)
				Muscle tone (p = NS)
Tizanidine PO (to 24 mg/d) vs. Baclofen PO (to 60 mg/d)	Hoog-straten et al. 1988113	6.1 ± 0.8 (mean ± SD)	16 (11)	Ashworth Scale (p = NS)
Crossover				EDSS (p = NS)
				Functional systems (p = NS)
				Isometric strength (p = NS)
Tizanidine PO (to 16 mg/d) vs. Baclofen PO (to 40 mg/d)	Newman et al. 1982122	NR	36 (26)	Ashworth Scale (p = NS)
Crossover				EDSS (p = NS)
				Pedersen functional disability scale (p = NS)
Tizanidine PO (to 35 mg/d) vs. Baclofen PO (to 90 mg/d)	Stien et al. 1987136	NR	40 (38)	Ashworth Scale (p = NS)
				EDSS (p = NS)
				Pedersen functional disability scale (p = NS)
				Overall evaluation (p = NS)
Tizanidine PO (to 18 mg/d) vs. Diazepam PO (to 22.5 mg/d)	Rinne 1980127	NR	30 (26)	Ashworth Scale (p = NS)
Baclofen PO (30–60 mg/d) vs. Diazepam PO (15–30 mg/d)	Cartlidge et al. 1974106	NR	40 (34)	Ashworth Scale (p = NS)
Crossover				Subjective impression (p = NS)
Baclofen PO (30–120 mg/d) vs. Diazepam PO (10–40 mg/d)	From et al. 1975110	NR	17	Ashworth Scale (p = NR)
Crossover				Spasms (p = NS)
				Clonus (p = NS)
				Urine retention, incontinence (p = NR)
				Walking (p = NR)
Dantrolene PO (25 or 75 mg, 4 times per day) vs. Diazepam PO (to 2 or 5 mg, 4 times per day)	Schmidt et al. 1975131132	5.5 (mean)	46 (42)	10- item exam evaluating spasticity, clonus, hyperreflexia, muscle stiffness, crampinga (p = NR)
Crossover				Subjective evaluation (p = NR)

Abbreviations: 9-HPT = 9-Hole Peg Test; ADL = activities of daily living; d = day(s); EDSS = Expanded Disability Status Scale; g = gram(s); IM = intramuscular; kg = kilogram(s); mg = milligram(s); MRD = Minimal Record of Disability; MS = multiple sclerosis; MSFC = Multiple Sclerosis Functional Composite; MSIS = MS-Impairment Scale; No. = Number; NR = not reported; NRS = Neurologic Rating Scale; NS = not statistically significant; PO = per os (by mouth); SD = standard deviation; SF-36 = Medical Outcomes Study 36-Item Short-Form Health Survey; THC = tetrahydrocannabinol

Findings. Trial design and results are summarized in Table 8. Assessments of muscle tone were not consistently positive for most of the drug treatments, with the exception of two studies each of gabapentin^{107, 121} and progabide.^{119, 120, 128} Results of functional assessments were largely inconclusive; the only positive findings on functional assessments were associated with delta-9-THC; deterioration in 9-Hole Peg Test (9-HPT) and Multiple Sclerosis Functional Composite (MSFC) was associated with delta-9-THC in one small study¹¹⁵ while a much larger study found improvements in 10-meter walk time¹³⁸ with delta-9-THC compared to placebo. Both studies reported these changes in the absence of a significant effect on objective measures of muscle tone. However, beneficial effects in the latter study for mobility are not easily ascribable to the drug's psychoactive properties alone.

In contrast to oral treatments, which were of uncertain effectiveness, intrathecal baclofen treatment showed a profound effect on muscle tone and spasms in one trial;¹²⁵ however, the patients in this study were selected based on response to a pre-trial intrathecal dose of baclofen.

None of the trials comparing active drug treatments showed statistically significant differences in efficacy between agents, although there were some differences in tolerability.

Rehabilitation. Among the 10 studies we classified as rehabilitation were RCTs of a variety of interventions delivered in a variety of settings. Probably the most comprehensive was a study of home-based management, which included not only traditional rehabilitation services but also nursing, education, psychological support, and social services.¹⁴⁴ Several studies examined more typical rehabilitation interventions delivered in inpatient or outpatient settings.^{139, 140, 142, 145} One study examined supervised exercise,¹⁴³ and one physiotherapy.¹⁴⁸ Three studies used interventions that were probably less intense than traditional rehabilitations described as professionally guided self-care,¹⁴¹ wellness intervention (education in health behaviors and lifestyle change),¹⁴⁶ and symptom management and adjustment (education and behavioral therapy).¹⁴⁷

Only one study used an active treatment comparison group, comparing inpatient rehabilitation to outpatient rehabilitation.¹³⁹ Two other studies used control interventions presumed to be ineffective: a home exercise program¹⁴⁵ and self-exercise treatment.¹⁴²

Study quality. All 10 trials randomly allocated patients to treatment groups, but the methods of randomization were clearly described for only five (50 percent). Most failed to provide enough information to determine whether there had been adequate concealment of the allocation schedule (six; 60 percent). In the four studies that reported enough information for evaluation, there was adequate concealment in three and inadequate concealment in one.

Because of the nature of the interventions, none of these studies was able to blind patients or therapists; however, four (40 percent) masked outcome assessors. Most studies (seven; 70 percent) reported the number of withdrawals in each treatment group. All but one of the studies was parallel-group in design; one study of home physiotherapy used a crossover design.¹⁴⁸

Study populations. Most of the studies (seven; 70 percent) described the study population in terms of EDSS score at baseline. Two studies had populations with notably lower mean EDSS scores of 3.4: the single study of supervised exercise¹⁴³ and one of the more educational/behavioral intervention studies,¹⁴⁷ which notably had a broad range of EDSS scores, from 0 to 9. In other studies, the populations were more impaired on average, with mean EDSS scores of 5.9 to 6.2 and medians of 5.5 to 6.5.

Outcome measures used in these studies included measures of impairment (e.g., EDSS), disability (e.g., Functional Independence Measure [FIM]), and handicap (e.g., SF-36), and not only related to individual functions, but often used comprehensive measures of overall disability and handicap.

Table 9. Randomized controlled trials of rehabilitation (more...)

Table 9. Randomized controlled trials of rehabilitation interventions

Treatment	Study	Baseline EDSS	No. of patients started (completed)	Outcomes/Results
Inpatient rehab (3 mo) vs. Outpatient rehab	Francabandera et al. 1988139	NR	84 (73)	ISS (p < 0.05)
				Need for home assistance (p = 0.17)
Inpatient rehab (20 d) vs. Wait-list control	Freeman et al. 1997140	6.5 (median)	70 (66)	EDSS (p = NS)
		5–9 (range)		FIM motor score (p < 0.001)
				LHS (p = 0.01)
Inpatient rehab (3 wk) vs. Home exercise program	Solari et al. 1999145	5.5 (median)	50 (45)	FIM (p < 0.01)
		3–7 (range)		EDSS (p = NS)
				SF-36 (p = NS)
Outpatient rehab (6 wk) vs. Self-exercise treatment	Patti et al. 2002142	6.2 (mean)	111 (106)	Fatigue Impact Scale (p < 0.001)
		4–8 (range)		EDSS (p = NS)
				SET (p < 0.001)
				BDI (p < 0.001)
				SF-36 (p < 0.05 on all subscales)
Home PT (8 wk) vs. Outpatient PT vs. No PT	Wiles et al. 2001148	6.0 (median)	42 (40)	Rivermead mobility index (p < 0.001, each active treatment vs. control)
Crossover		4–6.5 (range)		Balance time (p = 0.004; p = 0.001)
				Walk A (p < 0.003; p = 0.002)
				9-HPT (p = 0.01; p = 0.08)
				Global mobility (p < 0.001; p < 0.001)
Supervised exercise (15 wk) vs. No treatment	Petajan et al. 1996143	3.4 ± 1.0 (mean ± SD)	54 (46)	FSS (p = NS)
				SIP physical dimension subscale (p < 0.05)
				EDSS (p = NS)
				ISS (p = NS)
				VO2max (p < 0.01)
				POMS (p = NS)
Home-based management (rehab, nursing, education, psychological support, social services) vs. Usual care	Pozzilli et al. 2002144	5.9 ± 1.5 (mean ± SD)	201 (188)	FSS (p = NS)
				FIM (p = NS)
				SF-36 (p < 0.05 for bodily pain, general health, social function and emotional role subscales, and physical and mental component scores)
				EDSS (p = NS)
				MMSE (p = NS)
				STAXI (p = NS)
				STAI (p = NS)
				CDQ (p = 0.11)
Professionally guided self-care vs. No treatment	O'Hara et al. 2002141	NR	183 (169)	SDDR subscale O (p = 0.6)
				SDDR subscale E (p = 0.04)
				SF-36 (p < 0.05 on mental health and vitality subscales)
Education in health behaviors and lifestyle change vs. Usual care	Stuifbergen et al. 2003146	NR	142 (113)	SF-36 (p < 0.05 for bodily pain and mental health subscales)
				Self-efficacy (p < 0.01)
				Barriers scale (p = NS)
				PRQ (p = NS)
				HPLP-II (p < 0.01)
Multifaceted outpatient intervention (educational, behavioral) vs. Usual care	Wassem et al. 2003147	3.4 (mean)	27 (16)	Fatigue (p = 0.09)
		0–9 (range)		Sleep disturbance (p = 0.07)
				Pain (p = NS)
				Sum of symptom severity (p = 0.03)
				SEAB (p = 0.55)
				PAIS-SR (p = 0.72)

Abbreviations: 9-HPT = 9-Hole Peg Test; BDI = Beck Depression Inventory; CDQ = Clinical Depression Questionnaire; d = day(s); EDSS = Expanded Disability Status Scale; FIM = Functional Independence Measure; FSS = Fatigue Severity Scale; HPLP-II = Health Promoting Lifestyle Profile-II; ISS = Incapacity Status Scale; LHS = London Handicap Scale; MMSE = Mini Mental State Examination; mo = month(s); No. = Number; NR = not reported; NS = not statistically significant; PAIS-SR = Psychosocial Adjustment to Illness Scale-Self-Report; POMS = Profile of Mood States; PRQ = Personal Resources Questionnaire; PT = physiotherapy; SD = standard deviation; SDDR = Standard Day Dependency Record; SEAB = Self-Efficacy for Adjustment Behaviors Scale; SET = Tempelaar Social Experience Checklist; SF-36 = Medical Outcomes Study 36-Item Short-Form Health Survey; SIP = Sickness Impact Profile; STAI = State-Trait Anxiety Inventory; STAXI = State-Trait Anger Expression Inventory; VO_2max = maximum oxygen consumption with exercise; wk = wk(s)

Findings. Physiotherapy interventions failed to influence impairments as measured by EDSS (see Table 9). These interventions were, however, associated with measurable changes in functional status. Improvements in health (handicap) were observed in the SF-36 and several other measures.

Depression. Among the eight studies of treatments aimed at depression were RCTs of a variety of interventions delivered in a variety of settings. Two placebo-controlled studies used drug treatments either alone¹⁵⁵ or with a psychotherapy co-intervention.¹⁵⁶ One study tested psychotherapy,¹⁴⁹ and five trials used behavioral therapy, described as cognitive-behavioral therapy^{150, 151, 153, 154} or cognitive remediation.¹⁵²

The two drug studies were placebo-controlled and were the only studies to employ blinding of patients and investigators. The remaining studies used control groups that were passive (wait-list^{150, 151} or no treatment^{152, 154}) or active (current events discussion group,¹⁴⁹ support group¹⁵³); one study used both an active and passive control group.¹⁴⁹

Study quality. All eight trials randomly allocated patients to treatment groups, but the methods of randomization were clearly described for only two (25 percent). Six trials (75 percent) failed to provide enough information to determine whether there had been adequate concealment of the allocation schedule; the other two trials did not have adequate concealment of allocation. Because of the nature of the interventions, none of the behavioral therapy studies was able to blind patients or therapists; however, neither did any of these studies mask outcome assessors. Both studies of drug treatments were double-blind. Half of the studies failed to report the number of withdrawals in each treatment group. All but one of the studies were parallel-group in design; one study of amitriptyline used a crossover design.¹⁵⁵ Although this study did not report the number of patients in each sequence, it did employ a 1-week washout period and stated that there was no period or carry-over effect.

Study populations. Among the eight trials, the populations studied were quite variable with regard to the presence of depression or depressive symptoms. Some studies selected patients with MS without regard to the presence of a diagnosis of depression or depressive symptoms;^{149, 150, 152} however, most studies either required a formal diagnosis of a depressive disorder^{153, 156} or required a certain score on a depression scale.^{151, 154} One study specifically included patients based on the particular complaint of episodes of involuntary laughing or weeping.¹⁵⁵

Table 10. Randomized controlled trials of treatments (more...)

Table 10. Randomized controlled trials of treatments for depression

Treatment	Study	Baseline EDSS	No. of patients started (completed)	Outcomes/Results
Psychotherapy vs. Active control (current events discussion group) vs. Passive control (no treatment)	Crawford et al. 1985149	NR	32	MMPI Depression-30 scale (p = 0.025)
				IECS (p = 0.005)
				ASQ (p = NS)
				SES (p = NS)
CBT vs. Wait-list control	Larcombe et al. 1984151	NR	21 (19)	BDI (p < 0.01)
				HRSD (p < 0.01)
				Significant-Other Rating Scale (p < 0.01)
				Worst mood (p < 0.05)
				Best mood (p = NS)
				Average mood (p = NS)
CBT + relaxation vs. Wait-list control	Foley et al. 1987150	6 (mean)	41 (36)	BDI (p < 0.05)
		1–8 (range)		STAI-S (p < 0.05)
				STAI-T (p = NS)
				Hassles Scale (p < 0.05)
				PFC (p < 0.05)
Cognitive remediation (CBT + memory aids) vs. Active control (support group)	Mendoza et al. 2001152	NR	20	BDI (p = NS)
CBT vs. Support group (SEG) vs. Sertraline (200 mg/d)	Mohr et al. 2001153	2.4 (mean)	63 (52)	BDI CBT vs. SEG (p = 0.003)
		0–8 (range)		Sertraline vs. SEG (p = 0.05)
				CBT vs. sertraline (p = NS)
				HRSD CBT vs. SEG (p = 0.02)
				Sertraline vs. SEG (p = 0.45)
				CBT vs. sertraline (p = 0.13)
CBT (telephone-administered) vs. Control (usual care)	Mohr et al. 2000154	NR	32 (23)	POMS Depression-Dejection scale
				(p < 0.003, completers; p < 0.01, ITT population)
Amitriptyline vs. Placebo	Schiffer et al. 1985155	NR	17 (12)	Improvement in number of episodes of pathological laughing or crying; BDI; HRSD (p = 0.011)
Desipramine + psychotherapy vs. Placebo + psychotherapy	Schiffer et al. 1990156	4.6 (mean)	32 (28)	Clinical improvement in psychosocial function (p = 0.05)
				BDI (p = 0.16)
				HRSD (p = 0.02)

Abbreviations: ASQ = Anxiety Scale Questionnaire; BDI = Beck Depression Inventory; CBT = cognitive behavioral therapy; d = day; EDSS = Expanded Disability Status Scale; HRSD = Hamilton Rating Scale for Depression; IECS = Internal-External Control Scale; ITT = intention-to-treat; mg = milligram(s); MMPI = Minnesota Multiphasic Personality Inventory; No. = Number; NR = not reported; NS = not statistically significant; PFC = Problem-Focused Coping score from Ways of Coping Checklist; SEG = supportive-expressive group therapy; SES = Self-Esteem Scale; PFC = Problem-Focused Coping score from the Ways of Coping Checklist; POMS = Profile of Mood States; STAI-S = State-Trait Anxiety Inventory-State; STAI-T = State-Trait Anxiety Inventory-Trait

Findings. The included trials are summarized in Table 10. All but one of the studies reported statistically significant improvement in mood outcomes; Mendoza et al. (2001)¹⁵² failed to achieve statistical significance for changes in the Beck Depression Inventory (BDI), but the results did show a strong trend in favor of cognitive remediation. Treatment effects were similarly significant for studies of patients with depression diagnoses or those meeting minimum depression scale scores.

The studies were consistent in finding that untreated control groups^{149–152, 154} showed no improvement over time. In contrast, a control group of patients receiving psychotherapy as a co-intervention did show improvement.¹⁵⁶ Two studies that used active controls (current events discussion group¹⁴⁹ or support group¹⁵²) found marginal improvements in some outcomes, but still allowed detection of treatment effects.

Fatigue. Among the seven studies of specific treatments aimed at management of fatigue associated with MS were RCTs of oral and topical medications. Amantadine was the subject of four placebo comparisons.^{157–159 161} Pemoline was evaluated in three placebo comparisons,^{159, 161, 163} two of which were three-arm studies including both amantadine and pemoline.^{159, 161} Two more placebo comparisons evaluated 4-aminopyridine¹⁶² and transdermal histamine/caffeine.¹⁶⁰

Study quality. All seven trials randomly allocated patients to treatment groups, but the methods of randomization were clearly described for only one.¹⁵⁷ None provided enough information to determine whether there had been adequate concealment of the allocation schedule. All of the studies were described as double-blind, and all described the number of dropouts or withdrawals from each treatment group.

Four of the trials used a crossover design.^{157, 158, 162, 163} One reported a period effect and appropriately analyzed first-period data;¹⁵⁷ the other three crossover studies tested for and reported no period or carry-over effects. One of the trials did not use a washout period.¹⁶² Two reported the number of patients in each sequence.

Study populations. All studies selected patients based on complaints of persistent fatigue; some studies required either a threshold score on a fatigue measurement scale or a run-in period during which fatigue symptoms were demonstrated to be stable. Mean EDSS scores at baseline were reported for six of the seven studies (86 percent). Two studies had populations with higher EDSS scores than the others: Rossini et al. (2001),¹⁶² with a mean EDSS score of 6.2; and Gillson et al. (2002),¹⁶⁰ with entry criteria requiring EDSS between 5.0 and 6.5. The other studies had less impaired populations on average, with mean EDSS scores ranging from 2.6 to 4.3.

Table 11. Randomized controlled trials of treatments (more...)

Table 11. Randomized controlled trials of treatments for fatigue

Treatment	Study	Baseline EDSS (Mean ± SD)	No. of patients started (completed)	Outcomes/Results
Amantadine (100 mg twice per day × 3 wk) vs. Placebo	Canadian MS Research Group 1987157	4.3 ± 1.9	115 (109)	Change in VAS fatigue score (p = NS)
				Most affected activity VAS (p < 0.05)
				ADL total score (p = 0.09)
Amantadine (100 mg twice per day × 4 wk) vs. Placebo	Cohen et al. 1989158	4.0 ± 1.4	29 (22)	Fatigue score (p = 0.58)
Amantadine (100 mg twice per day × 6 wk) vs. Pemoline vs. Placebo	Krupp et al. 1995161	2.9 ± 0.9	119 (93)	MS-FS A vs. placebo (p = 0.04)
				P vs. placebo (p = 0.4)
				FSS A vs. placebo (p = NS)
				P vs. placebo (p = 0.8)
Amantadine (100 mg twice per day × 6 wk) vs. Pemoline vs. Placebo	Geisler et al. 1996159	2.6 ± 0.7	45	SDMT A vs. placebo (p < 0.05)
				P vs. placebo (p = NR)
				Digit Span (p = NS)
				Selective Reminding Test (p = NS)
				Benton Visual Retention Test (p = NS)
				Finger Tapping Test (p = NS)
Pemoline vs. Placebo	Weinshenker et al. 1992163	3.6 ± 2.0	46 (41)	EDSS (p = NS)
				Fatigue VAS (p = NS)
				Fatigue 4-point scale (p = 0.06)
Transdermal histamine/caffeine vs. Placebo	Gillson et al. 2002160	NR	29 (26)	MFIS (p < 0.02)
				25-foot timed walk (p =NS)
				9-HPT (p = NS)
				PASAT (p = NS)
4-aminopyridine vs. Placebo	Rossini et al. 2001162	6.2 ± 0.8	54 (49)	FSS (p = 0.19)
				EDSS (p = NS)

Abbreviations: 9-HPT = 9-Hole Peg Test; ADL = activities of daily living; EDSS = Extended Disability Status Scale; FSS = Fatigue Severity Scale; MFIS = Modified Fatigue Impact Scale; mg = milligram(s); MS = multiple sclerosis; MS-FS = MS-Specific Fatigue Scale; No. = Number; NR = not reported; NS = not statistically significant; PASAT = Paced Auditory Serial Addition Test; SD = standard deviation; SDMT = Symbol Digit Modalities Test; VAS = visual analog scale; wk = week(s)

Findings. Included trials are summarized in Table 11. Four studies comparing amantadine with placebo found statistically significant results on a few outcome measures, and statistical trends on others, consistent with a treatment effect. The two largest studies^{157, 161} found statistically significant treatment effects on the MS-Specific Fatigue Scale (MS-FS) or visual analog fatigue scores. Two of three studies comparing pemoline to placebo also reported statistically significant treatment effects.^{161, 163} One small study reported a treatment effect on the Modified Fatigue Impact Scale (MFIS) associated with transdermal histamine/caffeine, but no effects on functional measures (25-foot timed walk, 9-HPT) or cognitive performance.¹⁶⁰ The single trial of 4-aminopyridine included in our review failed to show a treatment effect.¹⁶²

Voiding dysfunction. Among the seven studies of treatments for voiding dysfunction were five placebo-controlled trials of desmopressin nasal spray,^{164–167, 170} one trial of pelvic floor rehabilitation,¹⁶⁹ and one trial of abdominal vibration or pressure.¹⁶⁸

Study quality. All seven trials randomly allocated patients to treatment groups, but none clearly described the methods of randomization. None provided enough information to determine whether there had been adequate concealment of the allocation schedule. All of the desmopressin studies were described as double-blind. Blinding was not possible for the vibration and rehabilitation studies. All but one¹⁶⁷ of the studies described the number of dropouts or withdrawals from each treatment group.

Six of the seven studies were of crossover design; one study of amitriptyline used a parallel-group design.¹⁶⁹ Only one of the crossover trials incorporated a washout period,¹⁶⁸ and only one tested for carry-over or period effect.¹⁷⁰

Study populations. Only two studies characterized populations in terms of EDSS, with means of 4.4 and 6.7. All studies recruited patients with MS and some kind of voiding dysfunction such as nocturia, urinary frequency, or incontinence.

Table 12. Randomized controlled trials of treatments (more...)

Table 12. Randomized controlled trials of treatments for voiding dysfunction

Treatment	Study	Baseline EDSS	No. of patients started (completed)	Outcomes/Results
Desmopressin nasal spray (20 mcg/d) vs. Placebo nasal spray	Fredrikson 1996164	NR	27 (22)	No. of voidings in 6 hr (p < 0.05)
				No. of voidings in 24 hr (p = NS)
Desmopressin nasal spray (20 mcg/d) vs. Placebo nasal spray	Hilton et al. 1983165	NR	16	Subjective benefit in nocturia (p < 0.01)
				Daytime urinary frequency (p = NS)
				Nighttime urinary frequency (p < 0.001)
Desmopressin nasal spray (20 mcg/d) vs. Placebo nasal spray	Hoverd et al. 1998166	NR	28 (24)	Daytime urinary frequency (p = 0.008)
				Nighttime urinary frequency (p = 0.26)
				Urine volume in 6 hr (p = 0.006)
				Urine volume in 24 hr (p = 0.052)
Desmopressin nasal spray vs. Placebo nasal spray	Kinn et al. 1990167	NR	13 (12)	No. of voidings in 6 hr (p < 0.05)
				No. of voidings in 24 hr (p = NS)
				Urine volume in 6 hr (p < 0.05)
Desmopressin nasal spray (10 mcg/d) vs. Placebo nasal spray	Valiquette et al. 1996170	6.7 (mean) 2.5–8.5 (range)	17 (11)	Nights with nocturia (p < 0.01)
				Incontinence (p = 0.08)
				Frequency of nocturia (p < 0.01)
				Uninterrupted sleep hours (p < 0.01)
Abdominal vibration vs. Abdominal pressure vs. No treatment	Prasad et al. 2003168	NR	30 (28)	No. of patients with no incontinence in 72 hr (p = NR)
				No. of voidings in 72 hr (p = NR)
				Post-void residual volume:
				Vibration vs. placebo (p = 0.002)
				Vibration vs. pressure (p = 0.059)
Pelvic floor rehabilitation (biofeedback + exercise) vs. No treatment	Vahtera et al. 1997169	4.4 (mean) 1–6.5 (range)	80	Incontinence (p < 0.05)
				Nocturia (p < 0.05)
				Frequency of UTIs (p = NS)
				Urinary symptom-related handicap (p < 0.05)

Abbreviations: d = day(s); EDSS = Extended Disability Status Scale; hr = hour(s); mcg = microgram(s); No. = Number; NR = not reported; NS = not statistically significant; UTIs = urinary tract infections

Findings. Included studies are summarized in Table 12. The five studies of desmopressin nasal spray all demonstrated statistically significant reductions in short term urinary frequency (number of voidings) during 6-hour periods after dosing. Most studies dosed at bedtime to control nocturia, but one study used daytime dosing to control daytime urinary frequency.¹⁶⁶ At longer time intervals (24 hours) the treatments failed to show significant effects. The number of episodes of incontinence showed a statistical trend in one study,¹⁷⁰ but this outcome was not reported in the other studies of desmopressin.

Among the two nonpharmacological treatments, a program of pelvic floor rehabilitation (biofeedback and exercise) was effective in reducing incontinence episodes and nocturia;¹⁶⁹ this study also showed a measurable reduction in urinary symptom-related handicap (p < 0.05). A small study of the use of a handheld abdominal vibrator at time of voiding succeeded in reducing post-void residual urine volumes and also appeared to reduce incontinence, but not urinary frequency.¹⁶⁸ Interpretation of the results of this study was limited by poor reporting of the data and statistical analysis (no statistical tests were reported for several outcomes).

Cognition. Studies aimed at treatment of cognitive deficits in MS are challenging to characterize. We did not identify a single study that (1) selected a study population with demonstrable cognitive deficits, (2) delivered a treatment (drug, behavioral, or psychological) aimed at improving cognitive performance, and (3) measured effects of treatment using tests of cognitive performance. However, one trial met at least two of these criteria and merits discussion here.¹⁷¹ In addition, some of the studies aimed at treatment of depression assessed cognitive performance at baseline or as secondary outcomes.¹⁵² Some of the rehabilitation trials evaluated mood-related outcomes in addition to functional outcomes, but none evaluated cognitive performance as an outcome measure. Finally, two of the studies aimed at treatment of fatigue also evaluated effects on cognitive function.^{159, 160}

Study quality. Lincoln et al. (2002)¹⁷¹ describe both the method of randomization and adequate concealment of allocation. Because of the nature of the intervention, blinding of patients and therapists was not possible; however, outcome assessors were masked. The study reported the number of withdrawals and dropouts in each treatment group.

Study populations. No studies selected patients based on screening for cognitive deficits at study entry. Lincoln et al. (2002)¹⁷¹ enrolled MS patients without regard to cognitive performance for their study of a cognitive intervention; however, 78 percent of the study population enrolled scored greater than 1 (the recommended cut off) on the mental disability question from Guy's Neurological Disability Scale (GNDS), and 95 percent either reported cognitive problems or showed significant impairment (score below cut off) on the Brief Repeatable Battery (BRB-N). Thus, while the proportion of patients with cognitive deficits may be elevated in this study due to selection biases, the prevalence of measurable cognitive deficits in other MS trials is likely also elevated.

Findings. Lincoln et al. (2002)¹⁷¹ found no differences between the cognitive rehabilitation, assessment, and screening groups in any of the outcomes measured at 4 months (shortly after the end of the 6-week intervention) or at 8 months. Although the prevalence of cognitive deficits at baseline was relatively high, a large number of patients had self-reported cognitive deficits that were not detectible on GNDS or BRB-N, which may have hampered the ability to detect treatment effects. Furthermore, the population was heterogeneous with regard to their cognitive performance, which at best results in increased variance, making it more difficult to show statistical significance to differences between groups.

Geisler et al.(1996)¹⁵⁹ evaluated the effect of amantadine and pemoline on fatigue and cognition. Although the Symbol Digit Modalities Test (SDMT), a measure of attention and visuomotor search, showed a drug treatment effect (with the amantadine-treated group showing the greatest improvement), all other neurobehavioral outcomes failed to demonstrate a treatment effect. The authors acknowledge that the study did not have sufficient statistical power to detect small- to medium-sized treatment effects. But also, the study found no statistically significant correlation between fatigue symptoms and cognitive deficits either at baseline or post-treatment.

Gillson et al. (2002)¹⁶⁰ reported results from the Paced Auditory Serial Additions Test (PASAT) in a trial of transdermal histamine and caffeine in a population of MS patients with fatigue. No significant treatment effect was detected.

Association of Clinical Findings with Work Ability

Introduction

This section summarizes findings related to Question 4: Among individuals with MS, what physical, mental, laboratory, or radiographic findings have been associated with inability to work?

Aspects surrounding ability to work are multifactorial in nature involving individual skill level, education level, interest in working, financial needs, domestic responsibilities, transportation needs, and other factors such as social norms and expectations. Functional limitations (physical and/or mental) associated with MS are presumably the primary determinants of work capacity. Adverse, intermittent, and poorly measurable symptoms such as fatigue and pain compound the task of determining if an individual with MS is able to work. Employers' willingness or ability to provide workplace accommodations for workers who have limitations also impacts whether or not an individual with MS can work.

Table 13. Included studies by type of measure used (more...)

Table 13. Included studies by type of measure used and type of function considered (physical, mental, other)

Type of measure used	Type(s) of function considered	Study
Hyllested criteria, etc.	Physical and mental	Rozin et al. 1975191
		Rozin et al. 1982190
Expanded Disability Status Scale (EDSS) and Disability Status Scale (DSS)	Physical	Canadian Burden of Illness Study Group 1998a174
		Grima et al. 2000180
		Hammond et al. 1996183
		LaRocca et al. 1982187
		Miller et al. 2000188
		Verdier-Taillefer et al. 1995193
Mobility aids	Physical	Dyck et al. 2000176
		Kornblith et al. 1986186
Job characteristics	Physical, mental, symptoms	Dyck et al. 2000176
		Grønning et al. 1990181
		LaRocca et al. 1982187
		Rozin et al. 1975191
		Scheinberg et al. 1980192
		Verdier-Taillefer et al. 1995193
Cognitive function	Mental/cognitive and physical	Beatty et al. 1995172
		Beukelman et al. 1985173
		Edgley et al. 1991177
		Genevie et al. 1987179
		Rao et al. 1991189
Disease subtype	-	Jacobs et al. 1999184
		Verdier-Taillefer et al 1995193
Self-report	Physical, mental, symptoms	Dyck et al. 2000176
		Edgley et al. 1991177
		Freal et al. 1984178
		Gulick et al. 1989182
		Scheinberg et al. 1980192

Several different types of physical and cognitive function measures (e.g., EDSS, disease subtype, mobility aids) were employed within and across studies to determine work ability among individuals with MS. For the purpose of this review we have grouped the summary of study findings according to these types of measures, which are listed in Table 13. Some studies used various types of measures and therefore are listed and discussed in more than one section. Categorizing studies or specific tests into physical function versus mental/cognitive function for this review was somewhat challenging since some tools measure both types of function. For example, the EDSS focuses on the functioning of numerous systems, including mental functioning, but is weighted more heavily towards testing ambulation. For the purpose of this review, tools like this are categorized according to the dominant area of function tested, resulting in the categorization of EDSS under “physical function.” Tools testing multiple types of function are discussed in greater detail below.

Results

A total of 22 articles^172–193 representing 20 research studies were included in this review (see Evidence Table 4 in Appendix F). The majority of studies (n = 15) employed tests to measure physical function, while fewer studies (n = 5) measured cognitive or mental function (Table 13). To date, no studies have been ascertained that used laboratory or radiographic measures to determine work ability among individuals with MS.

Most of the studies (n = 18) were cross-sectional in design, where work status or work ability was measured at a single point in time, primarily at the time of study enrollment. The other two studies were retrospective in design, including one case-control study¹⁹³ and one retrospective cohort study.¹⁸¹ No prospective studies, where changes in physical and/or cognitive function over time were considered in relation to changes in work status or ability, were identified for this review.

Twenty papers describing 18 studies used work status (yes/no or full-time, part-time, unemployed, retired, housewife) as a proxy measure for work ability among individuals with MS.^{172–189, 192, 193} The remaining two^{190, 191} attempted to incorporate several aspects of work ability as a study outcome: the Hyllested criteria. The remainder of this section is organized by types of measurement tools used to examine inability to work: Hyllested criteria, EDSS/DSS, cognitive measures, use of mobility aids, MS disease subtype, job type/characteristics, and self-report.

Table 14. Studies by Rozin et al.190, 191 measuring (more...)

Table 14. Studies by Rozin et al.190, 191 measuring ability to work among individuals with MS

Rozin et al. 1975191	Rozin et al. 1982190 (Series I and II combined)
1) Study participants (n = 172) initially grouped:	1) Study participants (n = 299) initially grouped:
n = 41 - Group A: Completely handicapped with no rehabilitation potential	n = 71 - Group A: A Completely handicapped with no rehabilitation potential
n = 37 - Group B: Potential for vocational rehabilitation, but unemployed or currently employed, but needs rehabilitation services for continuation of employment	n = 53 - Group B: Potential for vocational rehabilitation, but unemployed or currently employed, but needs rehabilitation services for continuation of employment
n = 94 - Group C: Currently working, holding previous jobs, or changed jobs without intervention of rehabilitation	n = 175 - Group C: Currently working, holding previous jobs, or changed jobs without intervention of rehabilitation
2) Type of MS disability by group:	2) Type of MS disability by group:
No disability:	No disability:
NR - Group A	NR - Group A
NR - Group B	3% - Group B
50% - Group C	29% - Group C
Physical disability due to MS:	Physical disability due to MS:
39% - Group A	59% - Group A
81% - Group B	75% - Group B
41% - Group C	61% - Group C
Physical and mental disability due to MS:	Physical and mental disability due to MS:
56% - Group A	30% - Group A
19% - Group B	11% - Group B
3% - Group C	6% - Group C
Mental disability due to MS:	Mental disability due to MS:
NR - Group A	1% - Group A
NR - Group B	2% - Group B
1% - Group C	1% - Group C
Other causes of disability not connected with MS:	Other causes of disability not connected with MS:
5% - Group A	7% - Group A
NR - Group B	2% - Group B
5% - Group C	1% - Group C
	MS and other causes of disability:
	3% - Group A
	7% - Group B
	2% - Group C
3) Hyllested criteria of disability:	3) Hyllested criteria of disability:
Group A (no rehabilitation potential) (n = 41)	Group A (no rehabilitation potential) (n = 71)
0% - Mild (0–2)	15% - Mild (0–2)
0% - Moderate (3–4)	38% - Moderate (3–4)
100% - Severe (5–6)	46% - Severe (5–6)
Group B (vocational rehabilitation needed among unemployed and employed) (n = 37)	Group B (vocational rehabilitation needed among unemployed and employed) (n = 53)
0% - Mild (0–2)	36% - Mild (0–2)
57% - Moderate (3–4)	51% - Moderate (3–4)
43% - Severe (5–6)	13% - Severe (5–6)
Group C (working) (n = 94)	Group C (working) (n = 175)
70% - Mild (0–2)	74% - Mild (0–2)
30% - Moderate (3–4)	25% - Moderate (3–4)
0% - Severe (5–6)	0.6% - Severe (5–6)

Abbreviations: MS = multiple sclerosis; NR = not reported

Hyllested criteria. As indicated above, only two^{190, 191} of the 20 included studies sought to determine ability to work among individuals with MS beyond the measurement of work status. Findings from these studies are reported separately (Table 14) from other studies that employed similar methods of measuring physical and mental functioning. Two cross-sectional studies conducted in Israel by Rozin et al. (1975¹⁹¹ and 1982¹⁹⁰) used similar methods for determining work ability among selected groups of individuals with MS ages 17 to 50 years. Interviews were conducted by social workers in the study participants' homes where demographic and occupational information was collected, as well as information about desire to be trained and employed. Subjects were evaluated by neurologists to determine degree of disability using a scale similar to the EDSS (described below) called the Hyllested criteria, which ranges from 0 (no functional disability, no residual signs) to 6 (bedridden, incontinent, requires constant supervision), with the mid-level score of 3 defined as moderate disability with work impairment sufficient to require a lighter job. Using these data, study participants were categorized into one of three groups:

(1) Group A - completely handicapped with no rehabilitation potential; (2) Group B - potential for vocational rehabilitation, but unemployed or currently employed, but needs rehabilitation services for continuation of employment; or (3) Group C - currently working without need of rehabilitation intervention. All Group B patients underwent additional tests to evaluate their functioning potential by a rehabilitation physician, occupational therapist, and psychologist. Study participants were also categorized into types of disability (physical, mental, or both); however, it is unclear how researchers measured mental disability. Furthermore, it is unclear if physical disability was determined strictly on the basis of the Hyllested criteria or if additional information was used. (The earlier study by Rozin et al. [1975]¹⁹¹ addresses additional aspects of employment by examining disability type and level by job type, discussed in greater detail under “Job type/characteristics,” below.)

Group A participants who were handicapped without rehabilitation potential were more likely to be disabled due to a decrease in physical or physical and mental function, with few being disabled strictly due to a decrease in mental function. Both studies^{190, 191} observed that about half of Group C (fully employed without need of vocational rehabilitation) were physically disabled and ranked as being either mild to moderately disabled on the Hyllested scale. Subjects in Group B (those who would benefit from vocational rehabilitation) were more likely to have moderate to severe physical disability. Although these patients had significant physical limitations, they were still considered to be individuals who would benefit from vocational rehabilitation and capable of working.

The obvious advantages of these studies include the consideration of work ability beyond work status, as well as the examination of both physical and mental function among the same sample population with regard to work ability, which most other studies did not report. However, a limitation of these methods is that researchers used current work status to determine ability to work, which could possibly bias the outcome of their evaluation. If a study participant was not working and expressed no desire to work, but was actually capable of performing a job, they may have been classified as someone who is not a candidate for vocational rehabilitation. Additional limitations include the omission of detailed information about how mental function was measured. These studies were also limited by small sample sizes, and data collected about education and marital status were not included as possible confounders in multivariate analyses. Because these were cross-sectional studies, the levels of physical and mental functioning were measured at the time of the study and not at the time when study participants ceased employment. The timing between impaired function and inability to work was not established.

Table 15. Cross-sectional studies examining current (more...)

Table 15. Cross-sectional studies examining current employment status and EDSS level among individuals with MS

Canadian Burden of Illness Study Group 1998a174	Grima et al. 2000180	Miller et al. 2000188
EDSS ≤ 2.5 (n = 62):	EDSS 1–2 (n = 78):	EDSS 0–3.0 (n = 113):
37% - Full-time	44% - Full-time	42.0% - Full-time
13% - Part-time	14% - Part-time	20.5% - Part-time
29% - Unemployed	15% - Not working due to MS	37.5% - None
13% - Other	27% - Not working, other reasons
EDSS 3–6: (n =68)	EDSS 3–6: (n = 75)	EDSS 3.5–6.5 (n = 131)
28% - Full-time	15% - Full-time	15.4% - Full-time
10% - Part-time	8% - Part-time	10.0% - Part-time
44% - Unemployed	51% - Not working due to MS	74.6% - None
18% - Other	13% - Not working, other reasons
	4% - NR
EDSS ≥ 6.5: (n = 68)	-	EDSS 7.0–8.5 (n = 56)
4% - Full-time		8.9% - Full-time
6% - Part-time		5.4% - Part-time
57% - Unemployed		85.7% - None
32% - Other

Abbreviations: EDSS = Expanded Disability Status Scale; MS = multiple sclerosis; NR = not reported

EDSS/DSS. Three cross-sectional studies (Table 15)^{174, 180, 188} and one case-control study¹⁹³ used the EDSS, and two cross-sectional studies^{183, 187} used the DSS to assess ability to work among individuals with MS. The EDSS (and its earlier version, the DSS) is a clinical tool commonly used for rating neurological impairment in individuals with MS.¹⁹⁴ Clinicians determine a patient's EDSS level by first assigning a separate grade for eight functional systems including pyramidal, cerebellar, bowel and bladder, cerebral, brain stem, sensory, visual, and other functions. A composite of grades is then used to determine an individual's EDSS score ranging from 0 (normal neurological exam) to 10.0 (death due to MS).¹⁹⁴ The level of function for each of the eight systems is considered for EDSS score; however, assignment of a level is superseded by an individual's ability to ambulate (e.g., free from mobility aids vs. need for mobility aids), possibly giving more weight to ambulation than the other seven functional systems. For example, individuals with MS who are able to walk without ambulatory aid would receive a score of 0 through 4.5, whereas a need for constant bilateral assistant (e.g., canes, crutches or braces) would predetermine an individual to receive a score of 6.5. Although mental function is factored into the EDSS scoring system as one of eight systems, it is not considered independently of ambulation at any EDSS level.

As detailed in Table 15, a lower frequency of employment was consistently observed in groups with higher EDSS levels in all three cross-sectional studies. Unemployment among study participants with an EDSS ranging from 3 to 6 was reported to be approximately 42 percent,¹⁷⁴ 52 percent,¹⁸⁰ and 72 percent,¹⁸⁸ respectively, while employment among lower EDSS levels (≤ 2.5) was 37 percent, 42 percent, and 51 percent, respectively. Unemployment was most common among individuals with higher levels of EDSS (≥ 6.5). A case-control study¹⁹³ observed that mean EDSS levels were significantly different between unemployed cases (0 = 5.4; standard deviation [SD] = 0.1) and employed controls (0 = 4.5; SD = 0.1) with MS (p = 0.01). This is only a 1-point difference on the EDSS scale, but 4.5 and 5.4 straddle the scale's demarcation of work ability, with 4.5 defined as “able to work a full day” and 5.5 defined as “disability severe enough to preclude full daily activities.”¹⁹⁴ These findings may reflect the timing of the neurological exam to assess EDSS, which was conducted at the outset of the study and not at the point when employment ceased. Physicians may not have been blinded to study participants' work status at the time of the exam, possibly biasing their evaluation.

Hammond et al. (1996)¹⁸³ conducted a large (n = 2099) cross-sectional study in Australia and reported that after adjusting for age, men with moderate DSS levels (4–6) were almost three times more likely to be unemployed (prevalence ratio [PR], 2.7; 95 percent CI, 2.1 to 3.6), and women were four times more likely to be unemployed (PR, 4.0; 95 percent CI, 2.7 to 5.8) compared to men and women (respectively) with lower DSS levels (0–3). Men and women with severe DSS (7–9) were also more likely to be unemployed (men PR, 17.9; 95 percent CI, 7.5 to 41.5; women PR, 24.6; 95 percent CI, 8.0 to 76.1) when compared to this same group. The second study¹⁸⁷ observed that a 1-point increase in DSS was associated with a seven percent decrease in the likelihood of being employed, and being male increased the probability of employment by 11 percent after controlling for numerous factors such as age, sex, education, marital status, and parenthood.

Findings from these studies suggest that individuals with higher EDSS/DSS levels are more likely to report not working. The three cross-sectional studies that examined EDSS had small sample sizes such that adjustment of prevalence ratios for other aspects associated with work ability was not possible. Extrapolation of these findings is limited because they focus only on a single dimension of work ability.

Studies by Larocca et al. (1982)¹⁸⁷ and Hammond et al. (1996)¹⁸³ included multivariate analyses where adjusted estimates were reported; however, no measures of cognitive impairment or job characteristics and responsibilities were considered. Again, these studies were cross-sectional, and the assessment of EDSS during enrollment in the study failed to establish the timing between impaired physical function and inability to work.

From a more global perspective, a semantic issue with EDSS deserves mention. Level 5.5 denotes disability severe enough to preclude full daily activities. It is not clear to what degree clinicians equate this EDSS-based activity preclusion with being incapable of working, without exploring other aspects of work ability such as cognitive function and employer accommodations. Conversely, cognitive impairment sufficient to impair work capacity would not typically be reflected in the EDSS score.¹⁸⁹

Table 16. Cognitive function and work status among (more...)

Table 16. Cognitive function and work status among individuals with MS

Study, number of subjects	Cognitive function measures	Primary findings
Beatty et al. 1995172	Categories:	Significant differences between workers and non-workers were observed on all measures except Digit Span, LOT, and WCST-% Perseverative Response
N = 102	Verbal ability (SILS Vocabulary Test)	49% of the variance in employment status was explained by walking ability, age, two measures of memory, and one test of verbal fluency
38 employed	Naming (Boston Naming Test)	Partial R2
64 retired	Visuospatial (Perception-LOT)	Ambulation index: 0.25
	Attention/Concentration (Digit Span)	STM (short term memory): 0.13
	Information processing (Speed, SDMT, FAS)	SRT (delay recall/memory): 0.04
	Category fluency	Age (29–62 years): 0.03
	Memory (STM-Correct, NMT-Delay, SRT-Total, SRT-Delay Recall, SRT-Delay Recognition)	FAS (verbal ability/letter fluency): 0.03
	Problem solving/abstraction (SILS-Abstraction, SILS-Conceptual Quotient, WCST-Categories, WCST-% Perseverative Responses)
	Ambulatory Index (score 1–6) (highly correlated with EDSS: r = 0.96)
	Beck Depression Index
Edgley et al. 1991177	PDQ score - sum of 4 subscales:	PDQ score (mean [SD]):
N = 602	1. attention/concentration	Unemployed: 1.6 (0.7)
	2. planning/organizing	Employed: 1.4 (0.7)
	3. retrospective memory	p < 0.001
	4. prospective memory	Mobility assistance (mean [SD]):
	(each subscale ranked 0–4:	Unemployed: 3.1 (1.2)
	0 = never	Employed: 2.2 (1.0)
	1 = rarely	p < 0.001
	2 = sometimes	“A significant multivariate main effect for employment status was obtained. Compared to individuals who were employed, unemployed individuals had more mobility problems (indicated above), obtained higher scores on the self-report PDQ (indicated above), had fewer years of education. Occupational level, number of people living at home and illness duration did not impact employment status.”
	3 = often
	4 = almost always)
	Mobility assistance:
	1. no ambulatory problems
	2. a bit unsteady
	3. need cane/brace
	4. wheelchair
	5. can't walk
Rao et al. 1991189	A battery of 36 tests were used to evaluate the following:	From the Environmental Status Scale (ESS), one of seven domains - Actual Work Status - was compared to cognitive impairment (yes/no)
N = 100	Dementia Screen (MMS)	Mean score of ESS scale (range 0–4) for Actual Work Status was lower (approximately 1.8) for cognitively impaired versus intact (approximately 2.8; p < 0.01)
52 intact	Verbal Intelligence (WAIS-R)
48 impaired	Memory Immediate
	Memory Recent
	Memory Remote
	Abstract Reasoning
	Attention/Concentration
	Language
	Visuospatial Perception
	100 MS patients were grouped as being either intact or impaired

Abbreviations: EDSS = Expanded Disability Status Scale; ESS = Environmental Status Scale; FAS = letter fluency test (saying as many words as possible that begin with F, A, and S, 60 seconds each); LOT = Line Orientation Test; MMS = Mini-Mental State; MS = multiple sclerosis; N = number of subjects; NMT = New Map Test; PDQ = Perceived Deficit Questionnaire; SD = standard deviation; SDMT = Symbol Digit Modalities Test; SILS = Shipley Institute of Living Scale; SRT = Selective Reminding Test; STM = Short Term Memory Test; WCST = Wisconsin Card Sorting Test; WAIS-R = Wechsler Adult Intelligence Scale-Revised

Cognitive measures. Three studies primarily examined cognitive function and work status among patients with MS (Table 16). Two of these^{172, 189} administered a battery of cognitive tests, while the third¹⁷⁷ collected data on cognitive function (attention/concentration, planning/organizing, retrospective and prospective memory) and ambulatory assistance through a self-report survey. The former studies examined a broad spectrum of function including verbal skills, memory, visuospatial perception, problem solving, and attention and concentration. In addition to these tests, Beatty et al. (1995)¹⁷² also administered the Ambulation Index (which is highly correlated with the EDSS, r = 0.96). Ambulation, short-term memory, delay recall, age, and verbal ability were found to explain 49 percent of the variance in employment status. Patients who were still working attained significantly higher scores on most of the individual measures of cognitive performance and were impaired on significantly fewer cognitive domains. Rao et al. (1991)¹⁸⁹ reported that cognitively impaired patients were also less likely to be employed compared to individuals who were cognitively intact, but information on which specific cognitive tests (or impairments) were associated with employment was not reported, and level of physical function was not considered in the analyses. Self-perceived cognitive deficit and need for mobility assistance were also associated with unemployment, as were fewer years of education and age.¹⁷⁷ However, occupational level (socioeconomic index), number of people living at home, and duration of illness did not impact employment status. A study by Genevie et al. (1987)¹⁷⁹ also considered self-reported cognitive function in combination with physical function and other symptoms, but details of the types of cognitive limitations were not described (see Evidence Table 4 in Appendix F for details of study limitations).

One additional study focused on self-reported expressive communication disorder, and study participants were asked if the communication disorder interfered with employment.¹⁷³ Employment status among those with self-reported communication problems was less compared to the entire study sample. Methodological problems (described in Evidence Table 4 in Appendix F) prevented further interpretation of this study.

Common sense suggests that impaired cognitive function has the potential to seriously impact work ability. However, the three studies described in Table 16 do not provide the evidence needed to determine the type and/or level of cognitive impairment when an individual with MS is no longer able to work. Unlike the studies by Rao et al. (1991)¹⁸⁹ and Edgely et al. (1991),¹⁷⁷ Beatty et al. (1995)¹⁷² provided far more detail about the specific cognitive tests that were associated with not working. In addition, these researchers also considered level of ambulation in combination with cognitive function and demographic characteristics. Since only a global measure of variance was provided it is difficult to interpret the strength of association between each of these domains (cognitive function, level of ambulation, demographic characteristics) and work ability. The 1991 study by Rao et al.¹⁸⁹ did provide details about the types of tests that were administered, but used a global measure of “intact versus not intact” to examine work ability. The method of self-report of cognitive function used by Edgley et al. (1991)¹⁷⁷ has limitations in that someone with impairment may not be able to objectively measure their own level of cognitive function. Finally, the temporal relationship between cognitive impairment and cessation of work among study participants was not captured in these cross-sectional studies.

Mobility aids. The number and type of mobility aids study participants used was measured in two studies^{176, 186} as a proxy measure for degree of disability. Kornblith et al. (1986)¹⁸⁶ developed a three-level Mobility Dysfunction Index (MDI) ranging from no assistance needed (Level 1), to any combination of cane, walker, leg brace, etc. (Level 2), to use of a wheel chair for more than half the time in- or outdoors (Level 3). A 1-point increase in MDI decreased the probability of males working by 24.3 percent, while it decreased the likelihood employment for females by 15.4 percent, leading investigators to conclude that mobility was a major determinant of employment, while age and duration of disease were minor. Dyke et al. (2000)¹⁷⁶ considered the number of mobility aids used and reported that only 20 percent of the variance in employment among a sample of women was accounted for by the number of mobility aids used, age, and education.

Use of certain mobility aids (e.g., wheelchair) can certainly provide a measure of degree of physical disability, as well as indicate the possible level to which the disease progression has hindered physical function, but Dyke et al. (2000)¹⁷⁶ considered only the number of aids used. A limitation of using the number of mobility aids to measure degree of disability is that it most likely is not sensitive enough to detect changes in other aspects of disease status and mental function, and it certainly does not capture job requirements. Desk jobs that require only sitting may enable someone who uses a wheelchair, but is not cognitively impaired, to continue working.

Disease subtype. Although there is great variability in the course of MS, three subtypes of disease are generally recognized: (1) relapsing-remitting; (2) primary progressive; and (3) secondary progressive.¹⁹⁵ The terms of these subtypes have changed over time due to refinements made within each classification, which are reflected in the different terms used in the following studies. One cross-sectional study¹⁸⁴ and one case-control study¹⁹³ compared MS patients' current work status with disease subtype. Both studies reported a higher frequency of employment among study participants with relapsing-remitting compared to primary progressive¹⁸⁴ and relapsing-progressive MS.¹⁹³ These findings are consistent with the greater degree of disability typically noted among individuals with progressive MS;¹⁹⁶ however, the analyses for both of these studies were crude and did not consider other factors associated with ability to work, except for Jacobs et al. (1999),¹⁸⁴ which attempted to control for age by restricting analyses to individuals less than 60 years of age. Furthermore, disease subtype was not measured until enrollment into the study. The disadvantage of using disease subtype for determining work ability is that the range of cognitive and/or physical function within each classification can vary tremendously. Furthermore, these studies do not provide the needed information for the measurement of physical and/or cognitive function that results in cessation of employment.

Table 17. Job type/characteristics and current work (more...)

Table 17. Job type/characteristics and current work status among individuals with MS

Study	Findings
Grønning et al. 1990181	Univariate analyses of time to unemployment:
	Non-remittent MS vs. remittent (p < 0.001)
	Heavy vs. light work (p < 0.01)
	Male vs. female (p < 0.05)
	Age > 30 at onset (p < 0.01)
	Multivariate analyses, when disease subtype was not considered, occupation (heavy work) and age (> 30 years) were predictive of early unemployment
Hammond et al. 1996183	“Authors noted that trade and farm workers were less likely to be in paid employment than professional or clerical workers as their level of disability increased.” (Researchers provided no data to support this statement.)
LaRocca et al. 1982187	84% of variability in employment status was unexplained by age, sex, education, marital status, occupation, and parenthood
	However, variability in employment status was explained by factors such as premorbid personality, coping style, characteristics of the workplace, and social support systems. Authors suggested that these findings contribute to the probability of a patient with MS staying at work. (Researchers provided no data to support this statement.)
Scheinberg et al. 1980192	Job category of currently employed subjects (n = 51):
	35.3% - Clerical
	23.5% - Professional
	13.7% - Semi-professional
	13.7% - Skilled labor
	7.8% - Managerial
	2.0% - Unskilled labor
	3.9% - Other
Verdier-Taillefer et al. 1995193	Job characteristics and odds of unemployment (odds ratio [95% CI]):
	Desk job - 0.3 (0.1 to 0.5)
	Sitting position - 0.3 (0.1 to 0.7)
	Possibility of obtaining specific arrangements - 0.4 (0.2 to 0.8)
	Travel time > 30 minutes - 1.7 (0.9 to 3.2)
	Daily work > 8 hours - 2.6 (1.2 to 5.7)
	Accessibility problems - 1.9 (0.9 to 4.0)
	Work requirements and odds of unemployment (odds ratio [95% CI]):
	Close attention - 0.9 (0.4 to 1.8)
	Good memory - 0.7 (0.3 to 1.5)
	Physical strength - 7.6 (3.2 to 18.2)
	Manual precision - 3.1 (1.6 to 6.3)
	Rigid work schedule - 2.2 (1.1 to 4.6)
	Decision making - 1.7 (0.7 to 3.4)
	Frequent moves - 2.5 (1.3 to 4.9)
Rozin et al. 1975191	Study participants were initially grouped into A, B, or C (described below), followed by examination of changes in work status/job type
	Changes in work status are from onset of MS to time of study in 1971; work type by predetermined work groups
	Group A (n = 41):	Group B (n = 37):
	Completely handicapped with no rehabilitation potential	Potential for vocational rehabilitation, but unemployed or currently employed, but needs rehabilitation services for continuation of employment
	Unskilled labor:	Unskilled labor:
	18% - onset of MS	28% - onset of MS
	0% - at time of study	3% - at time of study
	Skilled, semiskilled, service:	Skilled, semiskilled, service:
	27% - onset of MS	31% - onset of MS
	0% - at time of study	3% - at time of study
	Clerical, professional, student:	Clerical, professional, student:
	37% - onset of MS	31% - onset of MS
	0% - at time of study	8% - at time of study
	Housewives:	Housewives:
	2% - onset of MS	5% - onset of MS
	0% - at time of study	8% - at time of study
	Not working:	Not working:
	6% - onset of MS	5% - onset of MS
	100% - at time of study	65% - at time of study
	Group C (n = 94):	Authors note that “of the 131 clients with working potential (groups B and C), only 18% stopped working because of MS”
	Currently working, holding previous jobs, or changed jobs without intervention of rehabilitation
	Unskilled labor:
	22% - onset of MS
	8% - at time of study
	Skilled, semiskilled, service:
	18% - onset of MS
	17% - at time of study
	Clerical, professional, student:
	40% - onset of MS
	37% - at time of study
	Housewives:
	12% - onset of MS
	38% - at time of study
	Not working:
	8% - onset of MS
	0% - at time of study

Abbreviations: CI = confidence interval; MS = multiple sclerosis

Job type/characteristics. Six studies (Table 17) either focused primarily on job type or work characteristics^{181, 192, 193} or included information about occupation as a secondary aim in their study.^{183, 187, 191} Although the purpose of this review was to summarize information about measurements of physical and/or mental function among individuals with MS associated with inability to work, examining job requirements provides an indirect measure of the physical and/or mental levels of function needed to sustain employment.

In a case-control study conducted by Verdier-Talliefer et al. (1995)¹⁹³ several job characteristics were examined for their relationship with unemployment among MS patients. After adjusting for age, sex, type of disease, and level of education, an elevated odds of unemployment was observed among study participants whose jobs required physical strength (odds ratio [OR], 7.6; 95 percent CI, 3.2 to 18.2), manual precision (OR 3.1; 95 percent CI, 1.6 to 6.3), and frequent moves (OR, 2.5; 95 percent CI, 1.3 to 4.9). Furthermore, the odds of unemployment decreased when the job was a “desk job” (OR, 0.3; 95 percent CI, 0.1 to 0.5), or one that required sitting (OR, 0.3; 95 percent CI, 0.1 to 0.7). When all of the demographic and job characteristics were considered together in a multivariate model, work in the public sector was protective against unemployment (OR, 0.4; p < 0.05), and work requiring physical strength increased the odds of unemployment (OR, 4.5; p < 0.001). Analyses stratified by sex revealed that factors associated with unemployment for men involved a rigid work schedule (OR, 17.1; p < 0.01), while for women unemployment was strongly associated with work requiring physical strength (OR, 4.5; p < 0.05). These findings are consistent with those of Scheinberg et al. (1980),¹⁹² who observed that currently employed (n = 51) study participants were more likely to hold jobs that were clerical (35.3 percent) or professional (37.2 percent) as opposed to skilled (13.7 percent) or unskilled (2.0 percent) labor. An early study by Rozin et al. (1975)¹⁹¹ categorized study participants into groups according to their level of function and ability to work, described in greater detail above. This study crudely assessed changes in employment from the time of diagnosis to the time of the study. Study participants included in Group B (unemployed, but had the potential for vocational rehabilitation or employed, but needed rehabilitation services to continue employment) were more likely to remain in clerical and professional type jobs, compared to those in skilled and unskilled labor. Furthermore, those in Group C (currently working without need of rehabilitation intervention) were able to remain in the workplace, although they shifted employment from labor-intensive jobs to clerical or professional. Authors note that of the 131 clients with working potential, only 18 percent indicated that they stopped work because of MS, but provided no additional information about why study participants left work. Again, the limitations of this study are its small sample size and the lack of consideration of additional factors that influence ability to work. Grønning et al. (1990)¹⁸¹ reported consistent findings that heavy work was predictive of early unemployment; however, this study had serious limitations with regard to how jobs were categorized into heavy versus light work (see Evidence Table 4 in Appendix F).

Two studies did not focus primarily on job characteristics in the analyses, but did provide commentary about it. LaRocca et al. (1982)¹⁸⁷ reported that a significant portion of the variance in employment status was unexplained by typical demographic characteristics such as age, education, and occupation, but was explained by more subjective measures of workplace characteristics, social support, and coping style. Unfortunately, additional information about workplace characteristics was not provided. Dyck et al. 2000¹⁷⁶ commented that 17 percent of the women in their study reported that they quit work because they were unable to negotiate reduced work hours with their managers.

From what we know about the possible physical and cognitive limitations associated with MS, as well as the resulting fatigue and other symptoms, it is not surprising that unemployment is more common among individuals whose jobs required physical exertion. The strength of the case-control study by Verdier-Talliefer et al. (1995)¹⁹³ is that it considered numerous working conditions that increased or decreased the odds of unemployment. The remaining studies provided descriptive information that was parallel with findings reported by Verdier-Talliefer et al. (1995).¹⁹³ Unfortunately, none of these studies systematically examined whether employers' willingness or ability to provide workplace accommodations or flexible work schedules fostered continued employment.

Table 18. Self-report on why individuals with MS continued (more...)

Table 18. Self-report on why individuals with MS continued or terminated employment

Study	Findings
Dyck et al. 2000176	Factors contributing to maintaining employment - 44% of currently employed women were limited in the kind and amount of work they could do because of MS including:
	NR - fatigue “most common”
	16% - difficulty with standing and stairs
	15% - walking
	12% - writing
	11% - memory/concentration
	17% no longer working indicated “inability to negotiate reduced work hours” with their manager as reason for quitting work
Edgley et al. 1991177	Study participants who indicated that they quit working because of MS symptoms were asked an open-ended question about types of symptoms (n = 313; 78% of sample)
	41% - Ambulation problems
	39% - Fatigue
	12% - Memory problems
	10% - Emotional problems
	12% - Visual difficulties
	6% - Coordination problems
	2% - Pain
	1% - Incontinence
	22% left for reasons unrelated to MS. Women (26%) were significantly more likely than men (11%) to cite reasons unrelated to MS as the primary cause of unemployment (X2 = 9.3, P < 0.01)
Gulick et al. 1989182	Ranked comparison of conditions/situations that impede work performance (data on housewives and retired participants not described here):
	Fatigue:	Tremors:
	Employed: 50%	Employed: NR
	Unemployed: 25%	Unemployed: 10%
	Walking:	Use of wheelchair:
	Employed: 12%	Employed: NR
	Unemployed: NR	Unemployed: 10%
	Standing:	Restricted mobility:
	Employed: 8%	Employed: NR
	Unemployed: 12%	Unemployed: 9%
	Numbness:	Stiffness:
	Employed: 8%	Employed: 5%
	Unemployed: 5%	Unemployed: NR
Scheinberg et al. 1980192	Among those having left employment, the most common reason for leaving among multiple reasons given by 182 subjects (categories not mutually exclusive):
	52.7% - Physical difficulty
	15.9% - Visual difficulty
	12.1% - Transportation difficulty
	9.3% - Fatigue
	1.3% - Emotional difficulty
	37.4% - Other (mainly marriage and/or pregnancy)

Abbreviations: MS = multiple sclerosis; NR = not reported

Self-report. Several studies provided descriptive information about conditions or situations that influenced individuals with MS to cease employment (Table 18). Physical difficulty, ambulation problems, visual difficulties, emotional problems, and fatigue were reasons for ceasing employment among participants in two studies^{177, 192} who indicated that they left work because of MS. In addition, both studies reported that a significant percentage of women (37.4 percent and 26 percent, respectively) indicated leaving for reasons other than MS, including marriage and/or pregnancy. Among individuals who remained at work, fatigue was reported as the most common symptom impeding work performance or restricting the work that could be done in two cross-sectional studies.^{176, 182} An additional study not included in Table 18 ¹⁷⁸ reported that 10 percent (n = 30) of study participants indicated that they quit work because of fatigue. Although these findings do not involve specific clinical tests to determine the presence or absence of cognitive or physical impairment, qualitative data like these are useful for shaping quantitative data analyses, as well as shaping future research.

Environmental Factors and Work Ability

Results

A single research study met the inclusion criteria for this question (see Evidence Table 5 in Appendix F).¹⁸² Other candidate articles were typically excluded due to a lack of data regarding work capacity, work status, or disability status variables. In considering some of the excluded interventional studies aimed at short-term reduction of individual body temperature,^{199, 201–203} it was clear that an unacceptable degree of extrapolation would be required to relate the conclusions of these studies to work demands or circumstances of likely relevance to SSA.

Table 19. Conditions/situations impeding or enhancing (more...)

Table 19. Conditions/situations impeding or enhancing performance of work/chores

	Employed outside home	Homemaker	Unemployed	Retired
Conditions/situations impeding performance of work/chores (% of respondents citing)
Fatigue	50%	51%	25%	25%
Balance	10%	16%	11%	19%
Vision	12%	9%	11%	7%
High temperature	7%	6%	-	-
Conditions/situations enhancing performance of work/chores (% of respondents citing)
Intermittent rest	21%	22%	12%	19%
Assistance with tasks	18%	22%	22%	11%
Adaptive aids	16%	18%	12%	26%
Self-pacing	13%	11%	12%	13%
Cool temperature	-	8%	-	6%
Positive attitude	6%	5%	5%	6%

The single included article,¹⁸² a questionnaire survey, also has significant limitations in its applicability to SSA-relevant issues. It does, however, suggest that some MS patients perceive temperature as a factor that can either impede (high temperature) or enhance (cool temperature) work performance. This cross-sectional study divided 508 respondents with MS into four self-reported work status categories. Table 19 summarizes subjects' responses to open-ended questions regarding impediments to and enhancers of work/chore performance. Independent raters coded the actual responses into several categories. Table 19 includes only temperature-related conditions/situations and those endorsed by all four work groups. Overall, 53 work-impeding categories were identified, with 22 of those being endorsed by at least five percent of respondents. For work-enhancement, 27 total categories were identified, with 17 being endorsed by at least five percent of respondents.

Discussion of Evidence

Reliability of Criteria for Diagnosing MS (Questions 1a and 1b)

The recently proposed McDonald criteria for diagnosing MS are well supported by two types of evidence. First, two studies show that between 73 and 94 percent of patients presenting with clinically isolated syndrome (CIS) who go on to develop clinically definite MS (CDMS) over 1–4 years of follow up could be diagnosed with MS according to the McDonald criteria. Furthermore, the specificity of these criteria is reasonably high, ranging from 83 to 87 percent. Second, many studies support the magnetic resonance imaging (MRI) component of the McDonald criteria, by showing a strong and consistent association between the number of T2 lesions on MRI and the subsequent development of CDMS among patients with CIS or optic neuritis. Thus the McDonald criteria appear to have substantial evidence for validity and offer the obvious potential advantage of resulting in an earlier diagnosis of MS than the Poser criteria permit.

The McDonald criteria have been criticized for their complexity in comparison with previous criteria; however, we found data that demonstrate that the McDonald criteria yield a good overall diagnostic reliability, at least as good as the previous Poser criteria. However, widespread adoption of the new criteria could result in deterioration of this reliability. Barkhof et al. (1997)³⁹ demonstrated that among neurologists inter-rater reliability of MRI diagnosis significantly improves with increased level of training. Non-neurologists are unlikely to be able to achieve the same level of MRI agreement and are therefore unlikely to be able to maintain this level of agreement with the McDonald criteria as a whole. Further research on the inter-rater reliability of these criteria in broader clinical settings would be helpful to determine the quality of MS diagnosis.

While these data may be sufficient to secure a place in clinical practice for the McDonald criteria, certain difficulties arise in applying these criteria retrospectively from medical record review for the process of determining disability. At the present time most patients have not been diagnosed according to the specific application of the McDonald criteria. Therefore, they may have medical records that do not clearly delineate the nature and timing of their specific MRI changes in a manner that conforms to the McDonald criteria.

Predictors of Physical and Mental Impairments at 12 Months (Question 2)

The ability to predict future course of MS has been an active area of MS research; however, most studies examining disease course do so over quite long time periods of 5 to 20 years. We found a paucity of data describing changes in neurological or other impairments over 9–24 months, which we used to approximate the 12-month time horizon dictated by statutory requirements. Clinical characteristics have been the best studied, with four reports providing evidence for this review. While clinical features do not individually provide reliable guidance on prognosis, multivariate predictive models based on relatively easy-to-obtain features may have better performance. However, the reliability and validity of these predictive models has not been evaluated and thus their value for predicting disability has yet to be determined.

In contrast to their value in predicting development of CDMS among patient with CIS, imaging studies do not appear to provide especially useful prognostic data among patients with MS. The absence of lesions on sequential MRI studies is associated with a lower probability of an exacerbation in the ensuing month, and long-term prediction of health outcomes needed in the context of disability assessments has not been shown to be possible. Somewhat more promising strategies for predicting outcome of MS patients is the use of laboratory markers, such as apolipoprotein E (APOE) ε4 allele, interleukin-2 (IL-2) levels, or intrathecal immunoglobulin-M (IgM) synthesis, although the current level of evidence is best considered preliminary. These reports may provide the rationale for further validation studies, but are not standard practice and their practical impact is thus limited.

A single study of quality-of-life measures as predictors of long-term outcome was suggestive, especially considering the persistence of the association after adjusting for clinical characteristics. In addition to the uncertain generalizability of these results, such measures could have substantial reliability problems in the context of disability assessment.

Notably, no study was identified that examined the relationship between various factors and subsequent mental impairment.

Disease-modifying Therapies and Long-term Improvement (Question 3a)

Most of the data presented suggest that few patients improve on therapy. Those few who do improve generally do so only in the range of 1.0 point on the Expanded Disability Status Scale (EDSS). We found no data regarding improvement in work ability and no data that would correlate a 1.0-point improvement in EDSS with improvement in work ability. The significance of a 1.0-point EDSS improvement varies depending on baseline EDSS score (because the scale is non-linear), but the improvement data available are not generally stratified according to baseline EDSS score. With regard to work ability, the significance of the available data on clinical improvement is unclear. We found no data that quantified individual patient improvement with regard to cognitive function or quality-of-life measures.

The data examined in this evidence review do not support the conclusion that the current therapies are likely to result in substantial improvement in a significant proportion of patients with MS. This finding is consistent with expert opinion and demonstrated by the inherent design of current clinical trials, that is, the use of lack of decline in EDSS scores as the primary outcome measure. The present state of therapy is generally regarded as allowing for modest reduction in progression of MS — particularly in the relapsing-remitting patient population — but is not generally expected to result in significant long-term improvement. The rare exception is most likely in patients with relapsing disease who are progressing rapidly and undergo aggressive immunosuppressive therapy.

In general, the studies reviewed were not designed to answer the question we have asked. Indeed, individual patient improvement is not a common expectation of these trials. The authors believe that, despite the relative lack of data, the conclusions from the data examined do reflect the state of the current therapies. We believe that the data available document that individual patient improvement is an uncommon result of the current therapies. Further studies of currently available therapies would be unlikely to yield different conclusions; however, the recent trend of combining treatments in MS could yield different results. Combinations of currently available therapies and new therapies now under investigation may result in greater potential for individual patient improvement in neurological status.

Symptom Management and Improvement (Question 3b)

Treatment aimed at alleviation of symptomatic manifestations of MS, rather than the underlying disease, could have an important role in maximizing functioning among people with MS. Among the six areas we investigated, the degree of impairments and the effectiveness of the treatments varied. We review the conclusions and discuss the limitations and implications for further research by sub-topic below.

Spasticity. Although drugs such as baclofen, diazepam, dantrolene, and tizanidine are often used to reduce spasticity in MS, the research evidence for a beneficial therapeutic effect is inconsistent. This may be due, in part, to measurement issues. Spasticity is a difficult parameter to measure; too much muscle tone interferes with function due to spams and rigidity (resistance to movement); too little muscle tone can also interfere with function due to weakness. In some patients, a certain degree of elevated muscle tone in certain muscle groups can be desirable.

We had additional difficulty in estimating the clinical relevance of improvements that were reported, even when those changes were statistically significant. Many studies that dichotomized patients into “improved” versus “unimproved” failed to provide a definition or threshold of what changes represented “improvement.” Nonetheless, the relatively high baseline EDSS scores of patients enrolled in spasticity trials and unimpressive results of treatment suggest that anti-spasticity treatment is unlikely to have a clinically important impact on patients' functional status, and is, thus, unlikely to impact disability determination per se. Furthermore, all of these drugs are limited by poor tolerability at therapeutic doses. Our findings were consistent with another recent systematic review.²⁰⁴ Better measurement tools may be required in order to confirm the clinical impression that widely used anti-spasticity drugs such as baclofen, tizanidine, and dantrolene are more effective than placebo. Given current measurement techniques, it is not surprising that active-treatment comparison studies fail to show clinically important differences among these drugs.

Rehabilitation. Physiotherapy interventions failed to influence impairments as measured by EDSS. These interventions were, however, associated with measurable changes in functional status. Improvements in health (handicap) were observed in the Medical Outcomes Study 36-Item Short-Form Health Survey (SF-36) and several other measures. It is interesting that the historically first randomized controlled trial (RCT) of rehabilitation was not designed to assess the effectiveness of the intervention, but to evaluate whether less costly outpatient rehabilitation was as effective as inpatient rehabilitation.¹³⁹ Only more recently have trials been conducted that were designed to assess the value of rehabilitation compared to controls. Except for one study of supervised exercise,¹⁴³ the interventions employed in these studies were multifaceted, and it is difficult to attribute beneficial effects to particular components of the interventions. However, three other trials focused on non-physiotherapy interventions, which may be part of a rehabilitation program.^{141, 146, 147} Two of these studies found changes in health measures (SF-36) not unlike those seen in physiotherapy-based rehabilitation interventions; however, these studies did not include measures of impairments or function.

Depression. Depression treatments, including psychotherapy, behavioral therapy, and certain drug therapies, can lead to measurable improvements in mood. There are fewer data linking treatment of depression to improvements in other symptoms (such as fatigue or cognitive impairments) or other outcomes (such as functional status or quality of life). The changes demonstrated in instruments designed to measure depression were not small but, still, the link to improved functional status and, further, to ability to work was not demonstrated in these studies per se. To do so would require extrapolation from studies of treatment of depression in non-MS populations.

Fatigue. Amantadine appears to have some ability to alleviate fatigue in MS, as demonstrated in statistically significant differences in some outcomes in several trials; however, the clinical significance of these effects is likely small. Pemoline has been less often studied and shows results suggesting some effect. There is little support for the efficacy of 4-aminopyridine. Modafinil has shown promising results in phase-II trials,²⁰⁵ but has not yet been evaluated in a double-blind RCT. Measurement of fatigue is limited by a definition that spans several domains, leading to difficulty with validation. Further research on new pharmacological therapies (such as modafinil) and development of additional data on the validity of instruments for fatigue measurement and their sensitivity to change would be helpful directions for future research.

Voiding dysfunction. Desmopressin was highly effective at reducing urine volume and also consistently effective at reducing urinary frequency. This was demonstrated to translate into improvements in uninterrupted sleep hours and in fewer episodes of incontinence. Physical treatments, including both pelvic floor rehabilitation and use of a handheld vibrator during micturition, were also shown to reduce urinary symptoms compared with control. Only studies of pelvic floor rehabilitation measured impact on symptom-related handicap. These studies showed clear improvements in symptoms, but provided less clear data on how improvements in urinary symptoms impact other areas of health, and no data on how these symptomatic improvements might impact work ability.

Many interventions commonly used for urinary disorders in MS have not been studied in randomized controlled trials of MS patients. Commonly used interventions for which no RCTs have been performed among MS patients include anticholinergic and antimuscarinic drugs, behavior modification, and intermittent or indwelling urinary catheterization. Data supporting their use comes from trials in other populations or from case reports/series in MS.

Cognitive dysfunction. None of the treatments studied has had a consistent measurable effect on cognitive performance in MS; however, the question has been little studied and indirectly studied, in the sense that most data on cognitive effects are inferred from studies aimed at treatment of fatigue or depression. One study suggested that fatigue symptoms do not correlate with cognitive impairment, although they do correlate with symptoms of depression.¹⁵⁹ Future studies would benefit from more precise delineation of study population based on screening for cognitive performance deficits within a relatively narrow and defined range; this would likely improve the chances of finding a treatment effect and also make clearer the population for whom the results would be applicable.

Association of Clinical Findings with Work Ability (Question 4)

Findings and limitations. There is a significant gap between what is included in the literature and the research evidence using objective measures for determining ability to work. Although objective physical and cognitive measures have been developed, their application in the occupational literature is sparse. Furthermore, assessment of how symptoms such as pain and fatigue impact work ability was essentially absent.

In epidemiologic research we draw conclusions from the body of existing work and never from a single study. There are criteria that aid in the judgment of causality called the Bradford-Hill Criteria,²⁰⁶ which include strength of association, consistency of findings, temporality, dose-response relationship, biological plausibility, coherence, and specificity. After applying several of these criteria to the pool of information in this review we conclude that the research findings presented here are insufficient to demonstrate that a causal relationship between specific physical and/or cognitive measures and work ability has been established. The reported findings did display some consistency across studies. For example, individuals who had higher EDSS levels or low cognitive function were more likely to report not working. However, the strength of association across these studies was not clearly demonstrated, as most studies reported frequencies or crude estimates of association. Several studies consisted of small sample sizes, which hindered researchers from calculating risk estimates that were adjusted for potential biases such as age, education, level of employer assistance, job type, and desire to work. In addition, most studies considered only physical function or cognitive function, when both can hinder employment. A dose-response relationship of selected functional measures and degree of work capacity was not established. Because the majority of studies were cross-sectional, a temporal relationship between impaired function and inability to work was not established. Impaired function may not have occurred until after the study participants had ceased employment.

Although the bulk of these studies are descriptive in nature, they are useful for generating hypotheses for future studies to examine the causal relationship between impaired function and work ability. Some patterns that are noteworthy and should be considered when designing future studies to examine the risk of inability to work related to impaired function and adverse symptoms are as follows:

• The study outcome of work ability that extended beyond the definition of work status was informative for determining possible vocational rehabilitation potential for moderately to severely disabled individuals.

• As EDSS level increased, the frequency of reporting “not working” also increased. This same pattern was observed for the Hyllested criteria. However, there was considerable overlap between ability to work and EDSS level. Degree of disability or impaired physical function did not solely determine work ability.

• Level of cognitive impairment resulting in work inability was not adequately determined; however, the combination of both physical and cognitive measures proved necessary when assessing ability to work.

• Use of mobility aids and disease subtype as proxy measures for function provided very little information that was useful. Extrapolation of these definitions has the potential to lead to various interpretations, resulting in limited application of findings.

• Individuals with jobs that required less physical exertion were more likely to report remaining employed. These findings supported studies that measured both physical and cognitive function directly.

• The impact of fatigue on work ability was captured only through self-report. Direct measures of fatigue did not appear in the included studies.

In conclusion to the findings presented here it is important to discuss a significant limitation of observational research as it relates to determining if an individual with MS is able to work. Population-based epidemiologic research is useful for determining patterns, trends, and causal relationships between exposures and disease outcomes among groups of individuals, but in prior studies^{207, 208} researchers have demonstrated that risk factors or combinations of risk factors included in a statistical model serve as poor screening tools at the individual level. In the case of MS, future epidemiologic studies may indicate that certain physical or cognitive function tests are strongly associated with inability to work; however, extrapolation of these findings for the purpose of predicting work ability at the individual is not possible.

Future research. Future research about work ability among individuals with MS can shed a great deal of light on factors that foster or hinder employment. Work ability involves numerous medical and non-medical factors that have been discussed in great detail. In light of the gaps in the current literature, it would be advantageous to design future research endeavors to simultaneously address the following domains:

1. Objective clinical data: Collect data on clinical measures, such as disease subtype, that the Social Security Administration (SSA) typically uses when determining disability among individuals with MS. Data collection should be tailored to gather information from medical records that SSA deems relevant to its process.

2. Physical/cognitive/symptom measures: Objective measures of physical and cognitive function, as well as symptoms such as fatigue and pain. Consider the tools that have been well established and tested for measuring these types of impairments. As previously noted, the use of well-established tools in the occupational literature has been sparse with regard to MS.

3. Work ability measures: The measurement of work ability should be extended beyond the definition of work status to include examination of skill level, education level, career interests, willingness to work, and vocational rehabilitation potential.

4. Occupational requirements/employer accommodations: Assess job responsibilities and employer's willingness to provide accommodations in tandem with work ability measures. Current occupational requirements may not match with current level of function.

5. Demographics: Age, sex, education level, marital status.

6. Family responsibilities/support: Family demographics, income, responsibilities, number of children, and number of elderly parents who need care.

7. Subjective reports from individuals with MS: Qualitative information about hindrances and enhancers of work, including characteristics specific to the job or workplace and symptoms related to MS.

8. Subjective reports from employers: Qualitative information about how employers are able to provide accommodations and support for workers who have significant and/or intermittent fatigue, pain, and other symptoms related to MS. Gather information about how employers accommodate workers who also have physical and/or cognitive limitations, as well as information about conditions or situations where they cannot provide accommodations.

We suggest that a series of case-control studies be conducted, as well as a prospective cohort. For both these study designs, we suggest that resources already available through MS research centers across the country be used. A recent report by the Institute of Medicine²⁰⁹ describes certain limitations of research based in MS centers, but indicates that multisite research of this sort is valuable for conducting epidemiologic studies requiring a large number of patients with MS. Conducting a series of studies at several centers in the US could provide consistent evidence across numerous populations and geographic regions. Furthermore, data needed about employment history and disease progression may already be collected for established patients enrolled in ongoing studies. It is recommended that SSA collaborate with the National Multiple Sclerosis Society when selecting sites for this research.

Although it would be ideal to conduct a prospective cohort study for the purpose of establishing a temporal relationship between impaired function and work inability, this could take considerable time and money. We recommend initially a series of case-control studies where cases (defined as individuals with MS who are no longer employed) and controls (defined as individuals with MS who are still employed) are compared on the numerous domains described above. This should provide SSA with more detailed information about the types of clinical measures it currently deems relevant for determining work ability. Sample size was an issue in almost every study included in the review. Sample sizes should be large enough to detect true differences between groups, especially when considering cognitive function. Furthermore, the sample size should be large enough so that subtype of disease can be appropriately considered through stratified analyses. The pattern of disease progression varies between disease subtypes in ways that could influence employment. For example, patients with relapsing-remitting MS may not be able to remain steadily employed because of the erratic and unpredictable nature of disease exacerbations, whereas individuals with primary progressive MS may have a more predictable disease course that does not intermittently interfere with work. Although there is significant overlap of symptoms between the current disease subtypes, ability to work may vary considerably between them.

The case-control studies should be advantageous for examining the timing or sequence of clinical information and physical and cognitive function testing prior to cessation of employment. Demographic information would most likely be available as well. However, information surrounding employment issues, triggers for leaving work, obstacles or enhancers to work, and employers' willingness to help may not be available. In order to capture this information, especially around the time that an individual with MS decides to leave work, we recommend that researchers conduct a prospective cohort study. Incident cases of MS must be captured and followed over time so that changes in physical and cognitive function can be examined in relation to ability to work. Ideally, incident or fairly recent cases where the individual is still employed should be captured for follow up. As previously indicated, it would be ideal for researchers to build on prospective studies that are already ongoing. Information needed for this study could be added to what is already being collected for other studies.

The qualitative data collected through focus groups, personal interviews, and surveys have proven invaluable for guiding quantitative analyses. Information gathered from individuals with MS, as well as their employers, could provide a great deal of information about the types of accommodations that are useful and effective. Furthermore, companies may struggle with providing accommodations, especially with an individual who has frequent relapses.

Several study limitations have been highlighted in this review, as well as patterns of findings, which should be considered when developing future research. Future studies that consider the multiple factors influencing work can lead to generating unbiased risk estimates of inability to work among individuals with MS. In the development phase of these studies, it is imperative that SSA be involved in determining the types of data to be collected. Data collection should reflect the information that SSA deems important when making decisions about ability to work.

Findings in the context of other literature on work and MS. Several of the studies included in this review reported findings or trends about their study populations that reflect what is already known or established in the literature about issues surrounding employment independent of MS and vice-versa.

The onset of MS occurs between ages 15 to 50 years for approximately 90 percent of cases.²¹⁰ Thus, MS strikes individuals during peak years of education, training, and employment. Labor market analyses by the Bureau of Labor Statistics²¹¹ reported that, in general, higher levels of education enhance labor force participation, and severely disabled individuals are more likely to participate in the labor force if they have a college degree (57 percent) compared to disabled individuals with less than 4 years of high school (17.3 percent), high school completion (31.2 percent), and some college (39.1 percent).²¹¹

The incidence of disease is twice as high among women as among men.¹⁹⁵ MS impacts women not only during peak years of employment, but also during peak reproductive and childbearing years. Attachment to the workforce among first-time pregnant women in the US is influenced by several factors including age, education, years of work experience, and whether or not they were employed and established in their careers prior to the onset of pregnancy.²¹² The domestic responsibilities that follow pregnancy influence employment as well: compared to unmarried women without children, both married mothers and single mothers commit far fewer hours to the workforce (although the differences have declined significantly over the past two decades).²¹³ Some of the studies included here concurred with these findings in that women reported leaving work for reasons unrelated to MS, but related to domestic responsibilities.¹⁹¹ It can be especially challenging to distinguish between the impact of MS and the responsibilities of childrearing among women when examining ability to work. Some studies recognized these sex differences by controlling for them in their analyses.^{183, 186, 187}

Symptoms associated with MS vary between individuals, but can include fatigue, ataxia, dementia, optic neuritis, bladder urgency and incontinence, spasticity, pain, and sexual dysfunction.¹⁹⁵ As reported above, fatigue was the most common reason individuals with MS reported ceasing employment;¹⁷⁷ however, fatigue, like pain, is very subjective and difficult to measure (see discussions of Question 3b in this and previous chapters).

The importance of employer involvement in providing accommodations for disabled individuals was nationally endorsed in 1990 with the enactment of the Americans with Disabilities Act (ADA). In an earlier study by Mitchell (1981)²¹⁴ two-thirds of male postal workers with MS received work accommodations with respect to their MS related work capacity, which enabled them to remain at work. With the ADA in place for more than a decade, and MS representing the third most common neurological diagnosis that SSA receives,¹ the paucity of information in the literature about how employers have enabled individuals with MS to remain at work exists is unfortunate.

Environmental Factors and Work Ability (Question 5)

Current clinical wisdom about thermal sensitivity in some MS patients is based not on large controlled research studies but rather on a combination of many decades of clinical observations, case reports, small trials of thermal challenge outcomes, and reports of beneficial therapeutic effects of body cooling. The present report, while not directly addressing the question, provides no basis for rejecting such clinical observations that excessive heat can have an adverse temporary effect on the well-being and symptoms of some MS patients.

With regard to work impairment, limitation, or disability related to temperature conditions, we found remarkably little research that met our inclusion criteria. This should probably not be surprising. The difficulties of conducting population-based research on work capacity in patients with MS are reflected in the discussion of Question 4. Given this reality, researching temperature as an independent determinant of functional capacity would be particularly challenging.

There remain important questions about what proportion of MS patient populations may be thermo-sensitive, to what degree, and why?^{197, 200} As to temperature sensitivity significant enough to impede work, this report found only one includable report.¹⁸² This study has significant limitations in its generalizabililty to a Social Security Disability Insurance (SSDI) applicant pool due to a number of issues, including the fact that the assessment of thermal sensitivity was completely subjective. The low percentage of respondents citing thermal factors in this study suggests that exposure to heat is not among major perceived critical job demands.

Of interest, there is little or no report of thermal factors in other studies that might have identified perceptions of workplace temperature¹⁹³ or self-reported thermal sensitivity^{177, 185, 192} as variables that impede work or affect employment status.

It is possible that most MS research has focused at a more primary symptom level (e.g., affective lability, numbness, speech, fatigue, cognition, ambulation, vision, incontinence) without addressing more subtle factors, such as temperature, that may exacerbate symptoms. Forced-choice questionnaires, for example, may not include an option to report thermal sensitivity.¹⁷⁹

We conclude that answers to Question 5 remain mostly unknown. The evidence provides no basis for generalizations such as maximum appropriate working temperature levels unique to MS patient populations. The one included report confirmed that some MS patients perceive heat to impede their work capacity. It is not likely that medical data in SSDI application files in the current era will include objective diagnostic test results identifying MS patients who respond adversely to heat challenges.²¹⁵ However, subjective patient reports may describe such associations with or without clinician comment or correlation with objective clinical status measures. Although not necessarily founded on randomized controlled trial data, current clinical impression seems to hold that ambient and/or exercise-induced body temperature effects may bear a relationship to MS symptom status in some patients, perhaps more so than is thought to be the case for chronic disease states in general.

Knowledge Gaps and Future Research Recommendations

This report has identified several gaps in current knowledge that can direct future research. Regarding diagnosis of MS, the available studies evaluating the validity and reliability of the recent McDonald criteria,⁵ though few in number, are strong and consistent. The evidence reviewed suggests that diagnosis of MS using MRI data as implemented in the McDonald criteria has good validity compared with ultimate clinical diagnosis; further research is needed on the prognosis of patients diagnosed with MS using McDonald criteria, especially with regard to the ability of clinical or radiographic features to predict clinical outcomes.

Regarding prognostic studies, most studies report prognosis over long periods of time, over which MS shows inexorable progression. Shorter-term studies demonstrate a high degree of variability, and few factors seem to predict near-term prognosis. Studies of prognosis are needed that focus on the early disease course (an inception cohort) defined by McDonald criteria; such studies may be more fruitful at identifying prognostic factors than previous research has been. Data of this sort may be most efficiently gathered in the course of disease-modifying treatment trials, which are also a high priority in this population.

Regarding treatment issues, while differences in relapse rates and EDSS scores between control groups and groups receiving disease-modifying agents have been modest, reductions in disease activity as measured by MRI have been marked; the long-term significance of this reduction is not yet clear. Whether the failure to show differences in clinical outcomes is a limitation of the EDSS, the efficacy of the experimental agents, or the follow-up time is uncertain. Further treatment studies are needed that (1) target earlier disease (as diagnosed by McDonald criteria) and (2) are large enough and long enough to correlate MRI response with clinical response. Furthermore, while disease-modifying drug treatment trials have focused on physical function outcomes, this choice of outcome measure may preclude evaluating whether these treatments prevent or slow the development of cortical atrophy, which may correlate better with disability in MS.

Regarding symptomatic treatment issues, of the six symptomatic areas we explored, several were well covered. Spasticity drug treatments have been well studied, although the clinical relevance of the major outcome measure used in these studies, the Ashworth Scale, has been questioned. Further research on spasticity should be directed at evaluating functional status outcomes rather than muscle tone outcomes; and while drug treatments may be important, greater attention should be directed toward physical therapy, rehabilitation, and behavioral approaches instead of, or in addition to, drug treatments.

Treatment of fatigue has also focused on drug treatments, which have been largely unsuccessful. A new drug agent, modafinil, is currently under study, and may prove more useful; however, further research in non-drug approaches instead of, or in addition to, drug treatment may be necessary.

Mental and psychological functioning in MS has rarely been the target in intervention trials, although mood and cognitive disorders have been shown to be prevalent in MS. Further attention is warranted toward measurement of these impairments with the goal of developing outcome measures for disease-modifying drug treatment trials or symptom-directed intervention trials.

The major clinical interventions in urinary management in MS were little studied in MS patients specifically; it is unclear whether further research on groups of subjects with MS would be fruitful.

Regarding ability to work, this report highlights significant evidence and information gaps concerning:

• patterns of MS patient reports regarding functional limitations;

• information commonly collected in medical encounters with MS patients (and therefore available to SSA);

• knowledge about the impact on performance of specific work tasks of commonly objectified parameters such as coordination, strength, and vision, and especially of factors such as fatigue or cognitive dysfunction, which are either difficult to measure or are less commonly assessed in detail; and

• effective research methods for categorizing job or task demands in such a way as to isolate those demands that are likely to be “critical” for an SSDI applicant with MS.

In the context of these gaps, it may be productive to pursue research approaches that simultaneously address four domains:

• subjective reports (this domain is not sufficient alone for SSDI determination purposes);

• objective clinical data (ideally of the sort commonly encountered in medical records);

• in-depth objective measures (which may be available and not widely applied clinically, but which may be used with subsets of subjects to explore correlation with other domains); and

• work status measures (ideally longitudinal, with stratifications based on work demands).

Such an approach may apply to thermal sensitivity as well, with some additional specification and focus. Parallel assessment of concomitant ambient temperature, physical exertion, and core body temperature would address key relevant physiological exposure factors. Outcome measures could include the domains outlined above, for example:

• self-perceived well-being and level of symptoms such as fatigue;

• clinical parameters such as walking speed or muscle strength;

• in-depth measures such as potentially associated biomarkers or physiological parameters;^{199, 200} and

• work status measures, including absenteeism and disability benefits use.

Obviously, considerable caution would be required for subject safety related to heat exposure.^{216, 217} Such a research approach would be relatively complex and expensive, but might provide relevant information in the arenas of basic science and clinical care, and to a broad range of agencies, employers, and insurers dealing with MS work capacity issues.

Generating research results of practical relevance to SSA would, in some ways, be an even more demanding goal. The range of potential job situations in question for SSDI determinations always has the potential to extend beyond the claimant's own or previous occupation to a wide range of substantial gainful employment possibilities.²¹⁸ Heat exposure may well be an MS-related critical job demand in the context of various degrees of outdoor physical labor in warm climates. However, it seems likely that less physically demanding work in temperature-controlled job environments would frequently be the relevant job capacity circumstance by which SSDI applications were ultimately determined. Therefore, it may be necessary to use a narrower range of exposure and outcome variables in order to address the actual questions that might arise in the SSDI process. Such restrictions may limit a study's power to demonstrate associations or effects. The likely value to SSA of such an effort may well be influenced by an awareness of the frequency of and the particular circumstances surrounding SSDI determination situations in which thermal sensitivity is a critical determinant of the process outcome. Historical SSDI information may shed light on those particular permutations of ambient workplace temperature and physical demands that have commonly represented a critical or significant question in SSDI determinations.