Chapter  97:  Pharmacological Treatment of Dementia

The Questions

Given the range of pharmacological agents that have been used to treat dementia, evaluation of all of these interventions in a systematic review (which afforded a consistent methodology) should serve as a meaningful contribution in this area. The purpose of this systematic review is to answer the following questions:

1. Does pharmacotherapy for dementia syndromes improve cognitive symptoms and outcomes?

2. Does pharmacotherapy delay cognitive deterioration or delay disease onset of dementia syndromes?

3. Are certain drugs, including alternative medicines (non-pharmaceutical), more effective than others?

4. Do certain patient populations benefit more from pharmacotherapy than others?

5. What is the evidence base for the treatment of VaD?

This review considers different dementia populations (not just AD) and subjects from both community and institutional settings. The interventions were limited to pharmacological agents (including nutriceuticals), and these were not restricted to those that have received official approval in North America. The studies eligible in this systematic review were restricted to parallel RCTs, but the study outcomes were not limited to specific types.

The review will serve to evaluate the quality of the evidence and identify important gaps in the literature. Future recommendations will serve the dementia research community specifically. This evidence report will support the American College of Physicians (ACP) and the American Academy of Family Physicians (AAFP) in developing “best practices” and practice guidelines for the evidence-based treatment of dementia for providers, patients and the public.

Refinement of Questions

The first step during the topic assessment and refinement process was to organize a teleconference with the partner organization, the Task Order Officer (TOO), invited topic experts, and the McMaster team in order to define the magnitude of the topic addressed and to refine/clarify the preliminary research questions for this evidence report. It was agreed that this evidence report would focus on addressing the efficacy of pharmacotherapies for dementia syndromes. Regular teleconferences were held with the TOO, the partner, and technical experts throughout the data refinement and extraction phase.

Search Strategy

Table 1. Databases searched for relevant RCTs

Table 1. Databases searched for relevant RCTs

Database searched	Search date	Period searched
Cochrane Central	February 3, 2003	1st Quarter 2003
MEDLINE® & PreMedline®	February 4, 2003	1998 to 2003 week 4
EMBASE	February 6, 2003	1998 to 2003 week 5
AMED	March 4, 2003	1985 to 2003 February
CINAHL®	March 5, 2003	1982 to February 2003 week 3
Ageline	March 6, 2003	1978 to 2002 December
PsycINFO	March 7, 2003	1967 to 2002 December

Expert opinion was sought on the most efficient search strategies to minimize noise in the collection of citations. Some of the medical subject headings (MeSH) used to select RCTs yielded a large number of non-RCT literature due to misclassification of the study design terms. For example, in previous indexing, terms like “longitudinal study” or “comparative study” were applied to RCTs; conversely, the MeSH terms “random” or “randomized” in the title or abstract were not consistently used. However, some recent methodological work has suggested that more specific search term approaches can be used, which increases the sensitivity and specificity of the search results.⁴⁰ The Cochrane Central Trial Registry contains correctly re-classified RCT/Controlled Clinical Trial (CCT) trials that were misclassified in MEDLINE® and EMBASE from 1966 to 1998. All published RCTs to 1998 are contained within this database. Hence we commenced our search with the Cochrane Central Trial Registry database. For this reason, MEDLINE® and EMBASE were searched from 1998 forward for relevant studies, and all the other databases from their inception.

Specific drug names and manufacturer brands were considered as potential search terms. However, the local research team was in agreement that listing specific drug names would bias the yield to include only those pharmacological agents searched and would not capture newer drug therapies. Thus the recommendation was to not restrict the search to known pharmacological agents but to include whatever agents were in the literature.

In addition to the electronic databases, the bibliographies of retrieved papers were retrieved. Any citations recommended by the local research team, the TEP, or the peer reviewers were retrieved and screened.

Eligibility Criteria

Inclusion. Studies were included that contained the following criteria:

1. Age: Studies involving dementia patients who were 18 years or older in age

2. Diagnostic model used to determine dementia: The diagnosis of dementia using any of these criteria:

   1. ICD 9 or 10.^{41, 42}

   2. DSM III, III-R, and IV.^{43, 44, 45}

   3. NINCDS.⁹

   4. NINCDS-ADRDA⁹ or NINCDS-AIREN.⁴⁶

3. Diagnostic criteria used to determine cognitive impairment (pre-dementia): In the case of not yet diagnosed dementia, specific diagnostic categories were accepted for the following:

   1. mild cognitive impairment (MCI)..⁴⁷

   2. cognitive impairment not dementia (CIND).⁴⁸

   3. cognitive loss no dementia (CLoND).⁴⁹

4. Disease classifications for dementia: These included AD, senile dementia of the Alzheimer's type (SDAT), Lewy body disease, VaD, multi-infarct dementia (MID), AIDS/HIV dementia, Parkinson's disease dementia (PDD), progressive supranuclear palsy (PSP), mixed diagnosis dementia, encephalopathy, Mesulam syndrome, progressive non-fluent aphasia, Binswanger disease, subcortical leukoencephalopathy, circumscribed lobar brain atrophy, Pick disease, amyloid beta-protein (not Down's syndrome or trisomy), cerebral amyloid angiopathy, neurofibrillary tangles, threads, senile plaques, corticobasil ganglionic degeneration, cerebral autosomal dominant ischemia with subcortical leukoencephalopathy (CADISIL), Huntington's disease with dementia, hydrocephalus (for additional terms used in the search strategy, see Appendix A).

5. Severity classification: This was accepted in whichever classification system the studies specified. The majority of studies specified threshold criteria using the MMSE as follows: mild > 22, moderate 14 – 21, and 10 – 14 as severe. Many studies used the definition of mild to moderate as a range from 10 to 26 based on criteria established by Folstein et al.⁵⁰ Some studies specified a category (i.e. mild to moderate) but did not report the baseline MMSE values for the groups compared.

Some studies specified two categories (mild to moderate) and (moderate to severe) based on the DSM-III-R criteria. Cambridge Examination for Mental Disorders in the Elderly (CAMDEX) specifies levels of severity (minimal, mild, moderate, severe). Similarly, some studies reported a category of severity without stating which method was used. In these instances, the category of severity specified was accepted as reported by the study authors.

Exclusion. Studies that had populations with any of the characteristics listed below were excluded.

1. Dementia disease classification: i) alcohol caused dementia/ Korsakoff's syndrome, ii) Creutzfeldt-Jakob syndrome, c) spongiform encephalopathy, iii) hypothyroidism, iv) vitamin B12 deficiency, v) neurosyphilis.

2. Dementia diagnosed using only Lowb, Hachinski (specific for VaD) criteria.⁵¹

3. All organically caused dementias which includes “Delirium, Dementia, Amnesic Disorders, and Cognitive Disorder Otherwise Specified. The predominant disturbance is a clinically significant deficit in cognition that represents a significant change from a previous level of functioning. For each disorder in this section, the etiology is either a general medical condition (although the specific general medical condition may not be identifiable) or a substance (i.e., a drug of abuse, medication, or toxin), or a combination of these factors.”⁴³

4. Temporary dementia (e.g. side effect of anesthesia) classified as follows: Delirium: a delirium is characterized by a disturbance of consciousness and a change in cognition that develop over a short period of time. The disorders included in the “Delirium” section are listed according to presumed etiology: delirium due to a general medical condition, substance-induced delirium (i.e. due to a drug of abuse, a medication, or toxin exposure), delirium due to multiple etiologies, or delirium not otherwise specified (if the etiology is indeterminate).

5. Normal or healthy volunteers: studies that deal with healthy people (i.e. prevention is limited to people who have any form of the above); volunteer study population

6. General population of elderly persons.

7. Study subjects selected for depression (some patients may have dementia but not all) and where there is no stratified analysis by disease subgroup (i.e. the dementia subjects).

Study Design. Eligible studies included parallel design RCTs only. Although crossover trials are suitable for chronic diseases, they may be prone to period effects or period by treatment interactions. Period effects are systematic changes in the outcome that apply to all patients due to temporal changes in the disease or to the measurement instrument. Period by treatment interactions occur when the efficacy of the intervention varies by period. Additionally, a carry-over effect may occur if there is not an adequate washout period. Apart from the weaknesses of this design, some limitations arise when considering the potential for meta-analytic analyses. Traditionally, first period data from a crossover trial are abstracted and can be potentially combined with parallel trials for analyses of a pooled estimate; the reporting of the study results (positive or negative) would also be based on this first period data. In a preliminary phase of the review, several crossover trials were examined. Many did not report first period data, which precluded any potential for combining with parallel trials; many trials also did not undertake statistical tests during the first experience, thus making it difficult to report the direction and significance of the first period findings. Finally, because this systematic review was considering a variety of drug interventions administered over differing time intervals, period effects might be an important source of bias. For all these reasons, the decision was made to exclude crossover trials from this systematic review.

Language of Publication. Studies published in the English language were eligible. The scope and resources of this review did not permit translation of studies published in other languages.

Sample Size. No sample size restrictions were applied.

Treatment Interventions. Drug interventions were eligible in the following manner:

1. Pharmacological agents: all types of pharmacological treatment were considered in this review, including food supplements (as defined by the FDA). Government approval was not a requirement, and as such, off-label use of drugs (i.e. drugs approved for other conditions but used in the treatment of dementia) were eligible for this review.

2. Dose: all doses and dosing schedules and any mode of administration (oral, subdermal, transdermal, intravenous, suppository, or intra-muscular injection) were considered.

3. Treatment period: the period of treatment must equal or exceed 1 day.

4. Follow-up length: Any duration of follow-up was eligible. Different drugs require different time periods to show an effect. For example, antidepressant and anti-psychotic medications may take a month or more to be effective. Some dementia drugs take a minimum of 2 months. For interventions such as vitamin E or Ginkgo biloba, the time to effect is not well established. Thus, an absolute limit to the minimum number of months of follow-up could not be applied to all potential interventions. It was anticipated that many studies with some of the most recent pharmacological agents (i.e. donepezil) would have a minimum follow-up of 24 weeks.

Study Outcomes. No specific set of commonly accepted outcomes that define efficacy or “clinical relevance” were applicable to all the pharmacological interventions that have been used to treat dementia. The literature evaluating outcome measures in dementia trials would suggest that most instruments have significant limitations or at least more data are required to establish the required properties for acceptability of the scales. Since none of the outcomes used in dementia trials have been accepted as standards (no consensus), the selection of the most appropriate or clinically relevant outcome is purely arbitrary. Similarly, establishing a rationale to exclude studies based on the specific type of outcome measure would be arbitrary. For this reason, no exclusion criteria based on outcome measures were used as eligibility criteria for this study; rather the domains of interest for inclusion have been identified.

Studies with the following outcomes were included:

• General cognitive function (e.g., ADAS-cog).

• Specific cognitive function (e.g., Weschler Memory Tests).

• Global clinical assessment (e.g., CIBIC).

• Behavior/mood (disturbances characterized by agitation, wandering, sleep cycle disturbance, depression, obsessive compulsive activities) (e.g., Behavioral Pathology in Alzheimer's Disease Rating Scale (BEHAVE_AD)).

• Quality of life/ADL (e.g., Instrumental Activities of Daily Living (IADL)).

• Effects on primary caregiver (also referred to as caregiver burden).

• Safety as measured by the incidence of adverse effects (e.g., particularly serious adverse events).

• Acceptability of treatment as measured by withdrawal rate from trial due to side effects of the medication. (e.g., dropouts due to adverse events).

• Mortality.

• Dependency or Rate of Institutionalization/ or continued residence in own home.

• Use of services.

Studies with the following outcomes were excluded as follows:

• Studies which reported only biological/ physiological outcomes, such as plasma levels, changes on functional imaging, or electroencephalography (EEG) activity, were noted but not assessed as efficacy measures.

• Outcomes reported in the trials should reflect changes in the person with dementia. If the study population did not all have dementia, only data subgrouped for dementia was examined.

Minimum quality threshold score for eligibility.

Exclusion part I: Pre-Jadad score. Studies were also screened to determine a minimum threshold for quality, sometimes described as “fatal flaws” in the trial design. Specifically, all studies had to include at least some mention of the term “randomization” or “withdrawal(s)” in the text of the paper. Trials that did not at least mention these components were excluded, as they possessed a fatal flaw.

Exclusion part II: Post-Jadad score. The methodological quality of the primary studies was assessed using the modified Jadad scale for RCTs⁵² (Appendix B). The reliability of this modified scale was shown to be high, as measured by the intraclass correlation coefficient (ICC = 0.90).⁵² Each study was evaluated by two reviewers, and the level of agreement was determined statistically. The first three items on the scale rate elements that have been shown to bias meta-analytic results. These include randomization, blinding, and withdrawal. If these items alone are considered, the maximum score is 5. Any study that did not score 3 or more on the scale was excluded from the review. Therefore, this review abstracts detailed data only from studies that achieved moderate to high ratings on the quality scale.

Evaluating the methodological quality of studies and rating the strength of the evidence.

Quality of the RCT. The methodological quality of the primary studies was assessed using the modified Jadad scale for RCTs.⁵²

Quality of reporting adverse events. The potential for risk, or adverse events, was an important component to consider with respect to efficacy. The Jadad scale for quality does not take into account factors associated with adequate collection and reporting of adverse events as detailed by Ioannidis and Lau (2002).³⁵ Therefore, a summary checklist was developed to determine the potential quality in the collection and reporting of adverse events (Appendix B). This score was used to evaluate the relative quality of the adverse events reported.

Data Collection and Reliability of Study Selection

During the identification phase, two independent reviewers evaluated the title and abstract for eligibility; those meeting the criteria were retrieved as well as those that reported insufficient information to determine eligibility. Two independent reviewers examined the full text of these articles (passing from the title and abstract phase). All studies meeting eligibility criteria were reviewed to assess quality and abstracted according to predetermined criteria. The articles were grouped according to the pharmacological agent used in the intervention.

A team of study assistants was trained in the eligibility criteria for the purposes of this systematic review. Standardized forms and a guide explaining the criteria were developed from previous templates (Appendix B). Two reviewers were used for the identification, selection, validity, and abstraction phases of the systematic review. Disagreements were resolved by consensus. The reviewers were experienced EPC staff with post-graduate training in research methods. The reviewers and abstractors would consult with more senior members of the TEP for content expertise or methods-related issues.

Summarizing Results: Descriptive and Analytic Approaches

It was expected that studies of the pharmacological agents used in the management of dementia would be quite diverse with respect to the intended therapeutic effect. For these studies, evidence and summary tables (Appendix C) were constructed to describe the more salient characteristics of the included studies.

Meta-analysis

Statistical meta-analysis was not appropriate for all outcomes or interventions. Before calculating a pooled effect measure, the reasonableness of pooling was assessed on clinical and biological grounds, in terms of clinical homogeneity. Tabular summaries of key characteristics, participants, interventions, and outcomes were considered. A priori, it was decided that pooled estimates would be undertaken for studies with the same pharmacological intervention and the same outcome measure and that a minimum of three studies was necessary for pooling for a specific outcome^*. Consideration was given to the similarity of study populations when selecting studies to be included in the pooled estimates. Although many studies evaluated multiple outcomes, data necessary for meta-analysis were not provided in all eligible trials. When sufficient data were provided to estimate the weighted mean difference (WMD), then a meta-analysis was undertaken. WMD was selected as the pooled estimate (versus the standardized mean difference) because the outcome measures did not differ between studies eligible for pooled estimates. For WMD, the difference between the treated and control groups are weighted by the inverse of the variance.

Analysis was undertaken in RevMan 4.2 (Review Manager, Cochrane Collaboration, 2003), and the random-effects model was used to conduct our analyses. In cases where heterogeneity existed, the results of the random-effects model only were considered for interpretation of the results of the pooled estimate. RevMan 4.2 automatically tests the homogeneity of the results of the individual studies for each comparison of dichotomous or continuous data. Tests of homogeneity are formal statistical analyses for examining whether the observed variation in study results is compatible with the variation expected by chance alone. The more significant the results of the test (the smaller the p-value), the more likely that the observed differences were due to unknown factors likely not controlled for in the study. Sensitivity analysis or meta-regression was not undertaken to assess the extent to which the methodological quality of studies, population characteristics, dose, etc., accounted for variation in the primary outcome.

Power Analyses

Power analyses were conducted for select pharmacological interventions reporting non-significant findings for all primary outcomes reported in the paper. In addition, if the trial reported the outcomes of MMSE, ADAS-cog, or the CIBIC+, the power for these was also estimated. It was assumed that the desired level of significance was set to alpha equal to 0.05. Adequate power was defined as at least 80% power.

Peer Review Process

A list of potential peer reviewers was created at the outset of the study. During the course of the project, additional names were added to this list by the McMaster Center and Agency for Healthcare Research and Quality (AHRQ). In May 2003, the individuals on the list were approached by the McMaster team and asked if they would act as peer reviewers of this evidence report. A total of 26 experts agreed and received a copy of the draft report and a copy of the “Structured Format for Referee's Comments” (Appendix D). A list of the reviewers' names and their affiliation is provided in Appendix D. In addition, a criticism editor, Dr. Patricia Huston, who is external to the McMaster EPC, was asked to review the draft report and synthesize the peer review comments. The report from the criticism editor was then used to prioritize the incorporation of peer review comments into the final version of this evidence report.

Interpretation of the Results in the Overall Summary Tables (OST) for Individual Studies

Table 9. Guide to Overall Summary Tables - Outcome (more...)

Table 9. Guide to Overall Summary Tables - Outcome Measures Classified by Domain

General cognitive function measure	Specific cognitive function measure		Global Assessment	Behavior/Mood	Quality of Life/ADL/Function	Caregiver Burden	Other
ADAS-Cog (also ADAS-11)	ACPT	R-AVL	ADAS	ABID	ABS	CATS	CAUST
AMTS	Babcock Story recall	RM	ADCS-CGIC	ABSR	ADCS-ADL	CSS	SAS
BCRS	Barbizet Visuospatial	RPM	ADS	ACES	ADFACS	CSI	AIMS
CamCOG	BLM	Rey Memory Test	AGS-E	ADAS-Non-cog	ADL	SCB	BARS/ BAS
CASI	BNT	Set test	Bf-S	AFBS	ADL-C		ERP
CETM	BSRT	Snodgrass Picture Naming Task	BGP	BDI	ADL-PDS		ESRS
IQCODE	BSV	SRT-DR	Blessed-D/ BDRS	BEHAVE-AD	BI		Finger Tapping
MCPT	BVR	SWFIT	CAPE	BPRS	Dependency Scale		Test
MMSE	CCASSS	SWFT	CDR-NH	BRMS	DAD		SAS
MMMSE	Category Fluency	SKT	CDR-SB	BRDS	FAST		UPDR
SMMSE	CDT	TK	CGAE	CERAD-BRSD	FIM
CMMSE	CNTB	TMT	CGI	CMAI	IADL
MQ	Controlled Challenge Word Association	WMS (MQ)	CGIC	CS or CSDD	IDDD
RMT	COWAT	WMS-RR	CGRS	DSCS	NAA
RVM	CVLT	ZVT	CIBIC	DSS	NAI
SIB	Digit Span Test		CICIC+	Facial Behavior	OARS-ADL
SMQ	DSST		DBDS	GS	PDS
SMST	EFR		DMR	HAM-A	PSMS
TP, TPAT	FCMT		DRS	HAM-D	PSQI
WAIS	FIGT		EIS	HDRS	QoL
	FOM		FCCA	HDS-R	QoL-P
	GAGS		FRS	IPSC-E	QoL-C
	Grooved Pegboard Test		GERRI	LPRS	SF-36
	Letter Cancellation		GBS	MAACL-R	SIP
	Letter Fluency		GDS	MADRS	Time to functional decline
	LMT		GIS	MOSES
	LNNB		GPI-E	NAB
	MAE		HDS	NMS
	MEMT		HIS	NOSGER-IADL
	MNLT		MAC-F	NPI-NH
	NCT		NOSGER	NSL
	NDT		NOSIE	OAS
	NLT		NPI (NPI-4, NPI10)	PANSS-EC
	NMIC		PDRS	POMS
	NST		PGIR	RMBPC
	OLT		Plutchik CGS	RPT
	OMDR		RAGS	SBI
			RGRS	SHGRT
			SCAG	SRT
			Stockton GRS	VHB
			TSI
			VRGI

For each of the outcomes reported by a study, four interpretations of the results were possible. The four options for interpretation are as follows:

SC = significant change. Demonstrated by statistical significance (p ≤ 0.05) for the primary outcomes from an ITT analysis comparing treatment and placebo groups, or comparing differences among dose groups.

NS = not significant. The corollary of SC indicating no statistical significance.

MX = mixed results. Primary outcomes within the same domain show opposite or inconclusive statistical significance; for example, in the general cognition domain, half the studies show significant change and the other half show no significance).

NR = not reported. Outcome was collected but not statistically evaluated or not reported in the publication.

NT = not tested. No outcomes in this domain were tested.

Secondary outcome results were reported in the absence of any primary outcome data (for the domain of interest) and were demarcated with a (2°) in the OST. Similarly, analyses other than ITT were denoted with an asterisk (*) in the OST. If the report describes my subgroup analyses, the word SUBGROUP appears in the “other” column.

Adverse events were not always clearly described in many studies. A priori, we selected 5 generic symptoms (nausea, dizziness, agitation, eating disorder, and diarrhea) and selected to detail the ranges amongst studies for both placebo and treatment groups for these symptoms. The percent of withdrawals for both groups due to adverse events was reported. Adverse events reported to be statistically significant are highlighted for the reader. The details in addition to the quality score rating will assist the reader in evaluating the potential for harm.

Statistical Analysis

Power Analyses and Measures of Effect for combined studies. Power analyses (PW) were for individual trials for select pharmacological interventions (donepezil, galantamine, tacrine, rivastigmine, memantine, estrogen, carnitine, ginkgo biloba, selegiline) for all primary outcomes. In addition, if the trial reported the outcomes of the MMSE, ADAS-cog, or CIBIC+, power was also estimated (for individual trials of pharmacological interventions that had a minimum of three trials with a common outcome). Quantitative meta-analyses were undertaken in interventions that had a minimum of three trials using the same outcome scales and which provided sufficient data to permit calculation of effect sizes (as an Odds Ratio (OR), Relative Risk (RR) and Weighted mean difference (WMD)). The random-effects model results are presented to the reader.

Quantitative and Descriptive Analyses

Results of cholinergic neurotransmitter modifying agents (CNMA)

Table 2. List of Cholinergic neurotransmitter modifying (more...)

Table 2. List of Cholinergic neurotransmitter modifying agents and the number of studies vs. placebo for each of these. Asterisk (*) indicates report of a drug vs. drug trial [comparator drug(s) in brackets]

Drug	Number of studies vs. placebo
Antagonic Stress *[Meclofenoxate] *[Nicergoline]	0**
Carnitine	6
Donepezil *[Vitamin E]	10*
Eptastigmine	2
Galantamine	6
Huperzine-A *[Tablet Capsule]	1*
Linopirdine	2
Mexofenoxate *[Antagonic stress]	0*
Metrifonate	9
Nicergoline*[Antagonic Stress]	4*
Physostigmine	4
Posatirelin *[Citicoline]	4*
Rivastigmine	6
Sabeluzole	1
Tacrine *[Idebenone] *[Silymarin]	6**
Velnacrine	3

Number of studies. A total of six studies^{54, 55, 56, 57, 58, 59} evaluating carnitine were included in this review. Four of the reports were published from 1990 to 1992^{56, 57, 58, 59} while the remaining two, both by same authors, were published in 1996 and 2000.^{55, 54}

Design/methodology. A total of 925 subjects were evaluated in these six studies comparing carnitine and placebo. The range of study sample sizes was from 30 to 431 subjects. Quality scores (out of 8 points) ranged from moderate⁵⁷ (5) to high^{58, 59} (7), and all of the studies were partially or totally funded by industry.

Populations. All trials were conducted on AD patients, and all but one study used the NINCDS criteria for diagnosis. None of the trials reported including patients with severe dementia; all were classified as mild to moderate.

One trial⁵⁸ had a mix of community and institutional patients, and one study reported using a community sample.⁵⁶ The mean age of the samples ranged from 59⁵⁴ to 79 years,⁵⁹ with the majority reporting mean age greater than 70 years. One study⁵⁶ did not report mean age. Three studies specified the baseline MMSE^{54, 55, 57} (range 16.1 to 20.6) and one trial specified the modified MMSE⁵⁶ (mean 35) demonstrating no differences between placebo and treatment groups.

Intervention. The dose varied from 2 to 3 grams per day, and treatment duration was either 24 weeks^{56, 57, 59} or 52 weeks.^{54, 55, 58} One study⁵⁷ did not report the dose used. No titration period was used for this drug in any of the studies.

Primary outcomes. All of the studies measured cognition as a main outcome; half of the trials also measured outcomes in the behavior/mood domain.^{58, 54, 55} All but one of the studies⁵⁸ used a quality of life/ADL or functional status measure. Only one study⁵⁵ evaluated caregiver burden. Of the four studies^{59, 58, 54, 55} reporting the findings from a global measure, only one⁵⁵ used the Caregiver-rated Global Impression of Change (CGIC).

Analysis. Half of the studies reported ITT analyses^{58, 54, 55} and the remaining trials reported OC results.^{57, 59, 56} The ability to combine results was limited with only three studies having the common outcomes of ADAS-cog, modified MMSE (mMMSE), and Clinical Dementia Rating (CDR).

Figure 4. Weighted Mean Difference (WMD) from the Random (more...)

   Figure 4. Weighted Mean Difference (WMD) from the Random Effects Model (Random) for the MMSE comparing carnitine versus placebo

Figure 4

Number of studies. A total of 11 studies^{60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70} evaluating donepezil were eligible for this systematic review. One study ⁷⁰ compared donepezil to vitamin E rather than placebo. All were published within the last 6 years (n = 1, 1996), (n = 2, 1998), (n = 1, 1999), (n = 5, 2001), (n = 2, 2002). Three of these studies^{66, 67, 69} were undertaken by the same research group at different time periods and had unrelated cohorts of patients.

Design/methodology. A total of 3239 subjects (range of study sample size, 30 – 893) were included in these trials. The modified Jadad scale quality scores ranged from 5⁶¹ to 8.^{63, 64} All studies were funded by industry sponsors with the exception of a single trial⁷⁰ that did not specify their source of support.

Populations. All but one study⁶⁰ used the NINCDS criteria to diagnose dementia. Eight studies included only AD patients, one study included only VaD,⁶⁸ and the remaining two included Parkinson's disease dementia (PDD⁶⁵) and AD patients with cardiovascular disease.⁶⁴ A single trial included subjects with Down's syndrome and AD.⁶⁰ The severity of the dementia patients was described as mild to moderate in five studies,^{65, 66, 60, 62, 70} mild to moderately severe in two studies,^{69, 67} probable in two studies,^{61, 68} and moderate to severe in two studies.^{63, 64}

Some studies specified that the dementia patients were recruited from the community,^{68, 63, 60} one study from institutional setting,⁶⁴ and the remaining did not specify the living arrangements. Mean ages of the study subjects ranged from 54 to 85.7 years with most studies representing ages in the upper to mid 70s.

Six studies ^{61, 62, 65, 66, 67, 69} specified the race of the subjects, and of these, the overwhelming sample was Caucasian (range from 92 – 100%). All but one study⁶⁸ specified the proportion of men recruited, the range being from 18 – 46%. Four of these studies presented some results stratified by gender,⁶² age,⁶⁴ APOE genotype,⁶² baseline MMSE,^{63, 64} patients with Down's syndrome,⁶⁰ and the use of psychoactive drugs.⁶³ Three studies specified the baseline MMSE^{61, 64, 70} demonstrating no differences between placebo and treatment group, and the mean values varied from 14 to 16.

Intervention. Five studies evaluated a 10 mg dose given once daily,^{60, 61, 62, 63, 66} two studies 5 mg daily,^{64, 69} and four studies compared 5 mg and 10 mg dose groups.^{65, 68, 67, 70} Titration periods observed included 7 days^{65, 67} and 4 weeks,^{70, 68} these were not specified in the remaining studies. The total duration of the drug (including titration) varied from 12,⁶⁵ 15,⁶⁷ 23/24,^{60, 63, 64, 65, 66, 68} and 54/56^{61, 62} weeks.

Primary outcomes. Specific cognitive tests and caregiver burden were not evaluated in these studies. Nine studies used the MMSE, and six studies the ADAS-cog.

Analysis. All but one of the studies⁶⁰ comparing donepezil to placebo used ITT analysis. The study using OC analysis showed no statistical difference between treatments. It had the smallest sample size of 30 subjects and was underpowered (PW = 0.16) for the behavioral measure used, NPI.

Figure 5. Weighted Mean Difference (WMD) from the Random (more...)

   Figure 5. Weighted Mean Difference (WMD) from the Random Effects Model (Random) for the MMSE comparing donepezil and placebo

Figure 6. Weighted Mean Difference (WMD) from the Random (more...)

   Figure 6. Weighted Mean Difference (WMD) from the Random Effects Model (Random) for the ADAS-cog comparing donepezil versus placebo

Figure 5

Figure 5

Figure 6

Figure 7. Weighted Mean Difference (WMD) from the Random (more...)

   Figure 7. Weighted Mean Difference (WMD) from the Random Effects Model (Random) for the CIBC+ (continuous data) comparing donepezil versus placebo

Figure 8. Relative Risk (RR) from the Random Effects (more...)

   Figure 8. Relative Risk (RR) from the Random Effects Model (Random) for the CIBIC+ (dichotomous data probability of improving) for a 5 mg dose of donepezil

Figure 9. Relative Risk (RR) from the Fixed Effect (more...)

   Figure 9. Relative Risk (RR) from the Fixed Effect Model (fixed) for the CIBIC+ (dichotomous data [improved versus not]) for a 10 mg dose of donepezil

Figure 10. Weighted Mean Difference (WMD) from the (more...)

   Figure 10. Weighted Mean Difference (WMD) from the Random Effects Model (Random) for the Clinical Dementia Rating (CDR) comparing donepezil versus placebo

Figure 7

Figure 8

Figure 9

Figure 10

Five of the eight studies^{60, 61, 63, 64, 65, 66, 67, 69} measuring Activities of Daily Living (ADL) found a significant difference in the various outcomes used to assess ADL, but none of these could be combined into a summary estimate. It should be noted that the majority of trials selected these ADL variables as secondary outcomes. Behavior outcomes were not significant or showed mixed results for the three studies that evaluated this domain but these lacked sufficient power. Only one study collected caregiver stress and health service utilization outcomes⁶³ but did not report these data.

Number of studies. Six studies of galantamine^{71, 72, 73, 74, 75, 76} were included in this review. All compared galantamine with placebo and were published between 2000 and 2002 (from six different authors).

Design/methodology. The sample sizes for subjects ranged from 285⁷³ to 978⁷⁴ with 3530 subjects evaluated in total. All quality scores were high, with either 7 or 8 on the Jadad scale. Funding sources for these studies varied; one study did not report funding source,⁷⁴ one was funded by a non-industry source,⁷⁵ one was partially funded by industry,⁷¹ and three were funded by industry.^{72, 73}

Populations. All but one of the studies⁷¹ included AD patients only, and this single study mixed VaD and AD patients. All subjects in these studies were classified as mild to moderate.

One study specified that the subjects were from the community.⁷³ All studies included from 36 – 53% male subjects, and the mean age ranged from 72.2 to 76.8 years. Three studies^{71, 74, 76} specified race, and with the largest proportions being white (range from 91.5 – 99.9%). Two studies evaluated subgroups, based on baseline MMSE⁷⁵ and APOE genotype.^{75, 76}

Intervention. All studies had a titration period, starting at 4 mg per day^{71, 73} or 8 mg per day.^{72, 74, 75, 76} Four studies increased the dose weekly,^{71, 72, 75, 76} and one study increased every 2 to 3 days.⁷³ All studies had a treatment dose of 24 mg per day. Three studies^{73, 75, 76} included other doses ranging from 16 mg per day to 36 mg per day. Trials were undertaken for a minimum of 3 and maximum of 6 months.

Primary outcomes. All domains were measured except for specific cognitive tests and caregiver burden. All studies used the ADAS-cog and CIBIC or CGIC measures as primary outcomes. None reported baseline mean MMSE values.

Analysis. All but one⁷¹ of the studies reported ITT analysis, and the results of this study did not differ from the others.

Figure 11. Weighted Mean Difference (WMD) from the (more...)

   Figure 11. Weighted Mean Difference (WMD) from the Random Effects Model (Random) for the ADAS-cog comparing galantamine at 24 mg dose versus placebo

Figure 12. Weighted Mean Difference (WMD) from the (more...)

   Figure 12. Weighted Mean Difference (WMD) from the Random Effects Model (Random) for the ADAS-cog comparing galantamine at 32 mg dose versus placebo

Figures 11

12

Figure 11

Figure 12

Figure 13. Relative Risk (RR) from the Random Effects (more...)

   Figure 13. Relative Risk (RR) from the Random Effects Model (Random) for the CIBIC comparing galantamine at 24 mg dose versus placebo

Figure 14. Relative Risk (RR) from the Fixed Effects (more...)

   Figure 14. Relative Risk (RR) from the Fixed Effects Mode Fixed l for CIBIC comparing galantamine at 32 mg dose versus placebo

Figure 15. Weighted Mean Difference (WMD) from the (more...)

   Figure 15. Weighted Mean Difference (WMD) from the Random Effects Model (Random) for the DAD comparing galantamine at 24 mg dose versus placebo

FIgure 16. Weighted Mean Difference (WMD) from the (more...)

   FIgure 16. Weighted Mean Difference (WMD) from the Random Effects Model (Random) for the DAD comparing galantamine at 32 mg dose versus placebo

Figures 13

14

Figures 15

16

Number of studies. Nine studies^{78, 79, 80, 81, 82, 83, 84, 85, 86} were eligible for this systematic review. Studies were published from 1996 to 1999, and all studies compared metrifonate to placebo.

Design/methodology. Overall, 2759 subjects were evaluated. Sample sizes randomized ranged from 27 to 605 subjects. Quality of the studies ranged from 5⁸⁰ to 8,⁸¹ with the majority of studies earning 6 points.^{82, 83, 79, 85, 78} Two studies were non-industry-funded^{78, 79} and one did not specify the source of funding,⁸⁰ while the other six studies were partially or wholly funded by industry.

Populations. The subjects in all included trials were classified as having mild to moderate AD. Not all trials specified the source of recruitment or the racial composition of subjects. Three studies specified a community sample,^{86, 79, 85} and three trials reported the racial composition l,^{83, 86, 84} which was greater than 90% white in all cases. Mean age for all of the studies ranged from 71.4 to 75.0 years, with one study not reporting the mean age.⁸⁰

Intervention. All but one study⁸⁶ reported the loading dose, which varied from 0.5 mg per kg to 5.0 mg per kg. Following this initial loading period, the maintenance dose varied from 0.65 mg per kg to 4 mg per kg and 50 mg per day. The duration of the study treatments varied from 21 days to 26 weeks.

Primary outcomes. All outcome domains were evaluated with the exception of caregiver burden. ADAS-cog, CIBIC+, and MMSE were most frequently used as outcomes.

Analysis. Four trials reported OC analyses^{78, 79, 80, 85} and the remaining reported ITT analyses.

With the exception of a single study, quality scores for reporting adverse events were greater than 3 and generally well reported. However, only one trial ⁸³ tested for differences between groups and found nausea and vomiting, diarrhea, and muscle and joint disorder to be significantly different. The range of frequencies of the a priori symptoms of interest are as follows: 1) nausea and vomiting (placebo = 3 – 14%, all doses = 2 – 50%), 2) dizziness (placebo = 1%, all doses = 3 – 4%), 3) diarrhea (placebo = 4 – 14%, all doses = 11–19%), 4) agitation (placebo = 2 – 14%, all doses = 8 – 33%), and none reported eating disorder as an adverse event. Withdrawal rates due to adverse outcomes varied from 0 – 9% for placebo and 0 – 12% for treatment groups. Some studies indicated arrhythmia^{80, 82, 83, 85} and hypotension^{82, 85} and hematological abnormality⁸² but did not test for differences between groups. The majority of studies reported that laboratory tests including liver function and hematology were within normal limits. Overall, it was difficult to determine which types of adverse events reported had the potential to cause serious harm. This is some concern as metrifonate is no longer used as a therapy for dementia due to its potential for serious adverse events that include: respiratory paralysis, bradychardia, severe leg cramps and dyspepsia.⁸⁷

Number of studies. Four studies^{88, 89, 90, 91} compared the effect of nicergoline to placebo, and one study,⁹² published in 1994, compared nicergoline to antagonic-stress. The four placebo comparison studies were published in 1995,⁹¹ 1997,^{89, 90} and 2001.⁸⁸

Design/methodology. Sample sizes in the controlled studies varied from 108 to 346, with the total number of subjects included totaling 705. The drug versus drug study had only 62 subjects. The placebo trials all had quality scores of 6 points, while the non-placebo trial had a quality score of 5. Funding sources were reported only in two studies,^{88, 90} and both were industry-funded.

Populations. These studies had a very mixed population of dementia patients. Two included AD only,^{88, 92} one trial MID only,⁸⁹ one trial included both senile dementia of Alzheimer type (SDAT) and MID,⁹¹ and one trial included PDD, VaD, and mixed dementia.⁹⁰ All subjects had mild to moderate dementia. Studies included 38 – 55% male subjects; one study⁹¹ did not report the gender proportions. Mean age of subjects ranged from 69.3 to 73.7 years with one study not reporting⁹¹ this value. One study⁹¹ compared SDAT patients to MID patients.

Intervention. All trials versus placebo used 60 mg per day, but duration varied from 2 months,⁹¹ 6 months,^{88, 89, 92} and 12 months.⁹⁰

Primary outcomes. Caregiver burden was the only domain not evaluated by at least one of the studies. Three trials^{89, 90, 91} specified baseline MMSE, and this varied from 20 to 22.

Analysis. Two of the trials reported OC analyses^{90, 91} and two reported ITT analyses^{89, 88}; the trial comparing nicergoline to antagonic-stress presented OC analysis only.

Quality scores for reporting adverse events varied from 2 to 5 for these four trials, and none tested for differences between groups. Withdrawal due to adverse events varied from 0 – 8% for placebo and 0 to 9% for the treatment group. The trial with the lowest quality reporting score ⁸⁹ reported the most number of different events (up to 23 event types). With the exception of headache, which was reported in all four trials, it was difficult to determine which types of adverse events most characterized exposure to this pharmacological agent. The range of frequencies of the a priori symptoms of interest are as follows: 1) nausea (placebo = 3%, all doses = 3%), 2) dizziness (placebo = 1 – 2%, all doses = 0% or not reported), 3) diarrhea (placebo = 2 – 6%, all doses = 2 – 4%, 4) agitation (placebo = 5%, all doses = not reported), and none reported eating disorder as an adverse event.

Number of studies. Four studies were eligible for our review,^{94, 95, 96, 97} all comparing physostigmine to placebo only. The studies were from 1996, 1999, and 2000 with two of the studies being by the same author.^{95, 96}

Design/methodology. Sample size ranged from 176⁹⁷ to 475⁹⁶ with an overall total of 1198 subjects. Quality scores were 5,⁹⁶ 6,^{94, 95} and 7⁹⁷ out of 8 possible points. Two studies were industry-funded,^{97, 96} one was partially funded by industry,⁹⁵ and one did not report funding source.⁹⁴

Populations. All subjects had a diagnosis of mild to moderate or probable AD. Only one study⁹⁶ reported that all subjects were drawn from the community. Mean age ranged from 68.6 to 73.4 years, and the proportion of male subjects varied from 39.8 – 63%.

Intervention. Treatment schedules varied across the studies. One study⁹⁴ used a patch (30 and 60 mg), one trial used 30 mg (15 mg twice daily),⁹⁷ one trial had a washout and titration every 3 weeks to 30 or 36 mg per day,⁹⁶ and another trial titrated weekly to 15 mg twice daily.⁹⁵ Duration of treatment ranged from 6 weeks⁹⁵ to 24 weeks.^{94, 96}

Primary outcomes. All studies used the ADAS-cog as a primary outcome, and none reported caregiver burden or behavior/mood measures. No studies reported baseline MMSE scores.

Analysis. All but one trial⁹⁴ used ITT analysis.

The quality scores for reporting adverse events were generally low, scoring 1 or 2 out of 5. Withdrawal rates due to adverse events varied from 1 – 5% for placebo and 12 – 55%in the treatment group, with one study⁹⁷ not reporting rates. The high withdrawal rates were in studies with sample sizes that varied from181 to 475 subjects. A single study⁹⁷ tested for differences between groups, and found that dizziness, tremor, weight loss, asthenia (varying from 6 – 22% for all doses), confusion, delirium, and respiratory problems were significantly different. The cluster of reported types of adverse events suggests that gastrointestinal problems (abdominal pain, diarrhea) (placebo = 1 – 9%, all doses = 13 – 28%), nausea and vomiting (placebo = 1 – 9%, all doses = 9 – 75%) and eating disorder (placebo = 2 – 6%, all doses = 5 – 16%) were most frequently reported. Dizziness (placebo = 4 – 13%, all doses = 11 – 38%) and agitation (placebo = 6 – 16%, all doses = 4 – 8%) were also reported. No events deemed serious enough for hospitalization were reported.

Number of studies. Four studies^{98, 99, 100, 101} compared posatirelin to placebo, and one of these ¹⁰¹ also compared it to citicoline. One study was published in each of the years from 1995 to 1998, and all studies were conducted in Italy. Two studies were by the same author.^{100, 101}

Design/methodology. Populations randomized in the studies varied from 136¹⁰⁰ to 360⁹⁹ with a total of 931 in all trials. Quality scores ranged from 5⁹⁸ to 7¹⁰⁰ out of a possible 8 points. Three of the four studies did not report the source of their funding, but one trial¹⁰⁰ reported partial funding by industry.

Populations. No two studies included exactly the same populations; one trial had only AD,¹⁰¹ one trial had only VaD,¹⁰⁰ one trial had mixed AD and VaD,⁹⁹ and another trial had mixed AD, VaD, and PDD.⁹⁸ This latter trial⁹⁸ compared populations in a subgroup analysis of AD versus VaD. All studies evaluated populations with mild to moderate disease.

The mean age of the subjects ranged from 69.4¹⁰⁰ to 78.8⁹⁸ with the percentage of male subjects varying from 34¹⁰¹ to 66%.¹⁰⁰ One study¹⁰⁰ included a dementia population who also had hypertension.

Intervention. A dose of 10 mg per day was used in all studies, and treatment interval varied from 3 to 4 months.

Primary outcomes. General cognitive function was evaluated in all studies using the intellectual impairment Gottfries-Bråne-Steen (GBS) subscale and the MMSE was used in one trial.¹⁰¹ Specific cognitive function was evaluated in one trial.¹⁰⁰ Quality of life/ADL was evaluated with the ADL subscale of the GBS, and behavior/mood was evaluated with the emotional impairment subscale of the GBS. GBS total score was assumed to be a measure of quality of life/ADL rather than global assessment. None of the trials reported baseline MMSE scores.

Analysis. Two of the studies^{100, 101} used OC analysis to report outcomes.

Quality scores for reporting adverse events varied from 2 to 4. Withdrawal rates due to adverse events ranged from 0 – 3%in placebo and 0 – 4%in the treatment group. One trial¹⁰⁰ did not report the rate of withdrawal. None of the studies tested for significant differences between groups. All studies reported the presence of agitation, and at least three studies reported arrhythmia, nausea/vomiting, headache, rash/skin disorder, and sleep disorder; there is no evidence to suggest that these differed between placebo and treatment group. The range of frequencies of the a priori symptoms of interest are as follows: 1) nausea (placebo = 3%, all doses = 1 – 4%), 2) dizziness (placebo = not reported, all doses = 1%), 3) diarrhea (placebo = 2%, all doses = 2%), 4) agitation (placebo = 1 – 5%, all doses = 1 – 5%), and none reported eating disorder as an adverse event. No serious adverse events were reported.

Number of studies. Six studies were eligible for this review, all comparing rivastigmine to placebo. Studies were published in 1998,^{102, 103} 1999,^{104, 105} 2000,¹⁰⁶ and 2001.¹⁰⁷

Design/methodology. Six studies evaluated 2071 subjects in total, with studies ranging from 27¹⁰⁷ to 725¹⁰⁴ subjects. The quality of studies varied from 5 to 8, with three studies^{106, 104, 103} scoring 8 points and one¹⁰⁷ earning 5 points. All studies were funded by industry sponsors.

Populations. Four studies were evaluated in AD patients,^{107, 105, 104, 103} one trial¹⁰² dementia of the Alzheimer's type (DAT), and one study¹⁰⁶ Lewy body dementia subjects for mild to moderately severe subjects. One study¹⁰³ reported a subgroup analysis by vascular risk. One study reported a community sample in their trial.¹⁰⁴ Mean age for the studies ranged from 69.4¹⁰² to 75.9 years.¹⁰⁷ Two studies^{107, 105} did not report the ratio of male subjects in their study, and the other four varied from 39 – 56%. One trial¹⁰⁴ reported co-morbidity of diabetes, hypertension and arthritis; one trial¹⁰³ reported concurrent medication use for cardiovascular, gastrointestinal and analgesic aids.

Intervention. Doses for rivastigmine varied from 1 mg¹⁰³ to 12 mg,^{106, 105, 104} and treatment duration varied from 14¹⁰⁴ to 26^{107, 103} weeks. All studies titrated the dose of drug over a period ranging from 2 weeks¹⁰⁶ to 12 weeks.¹⁰⁷

Primary outcomes. The ADAS-cog and CIBIC+ were evaluated in half the studies. Baseline MMSE was reported in two trials,^{103, 106} and the mean scores varied from 18 to 20. Specific cognitive function, behavior/mood, and quality of life/ADL were infrequently evaluated, and caregiver burden was not evaluated in the trials.

Analysis. Trials were evenly divided between ITT analyses^{106, 104, 103} and OC.^{107, 105, 102}

Figure 17. Weighted Mean Difference (WMD) from the (more...)

   Figure 17. Weighted Mean Difference (WMD) from the Random Effects Model (Random) for the ADAS-cog comparing rivastigmine versus placebo

Figure 17

Two trials evaluated specific cognitive tests; one trial¹⁰⁶ showed statistical significance for the CCASSS but not for the MMSE (general cognitive test). Similarly, another trial¹⁰⁵ found the Weschler Logical memory (instant recall) to be significant but the ADAS-cog (general cognitive outcome) was borderline significant.

Figure 18. Relative Risk (RR) from the Random Effects (more...)

   Figure 18. Relative Risk (RR) from the Random Effects Model (Random) for the CIBIC+ comparing rivastigmine versus placebo

Figure 18

Figure 18

Quality scores for reporting adverse events varied from 2 to 5. Withdrawal rates due to adverse events ranged from 4 – 11% in the placebo and 11 – 27% in the treatment group. One trial¹⁰⁷ did not report the withdrawal rates or the types of adverse events observed. Two trials^{103, 104} demonstrated a dose response; however, one of these trials¹⁰⁴ showed significant differences for nausea and vomiting only, and the other trial¹⁰³ showed significant for all the adverse events reported. With respect to the types of adverse events, the majority of studies reported dizziness, nausea and vomiting, eating disorder/weight loss, and headache. It should be noted that one study¹⁰⁵ allowed intentional prescribed anti-emetic drugs to increase the tolerance of subjects taking rivastigmine. The range of frequencies of the a priori symptoms of interest are as follows: 1) nausea (placebo = 3 – 10%, all doses = 8 – 58%), 2) dizziness (placebo = 0 – 7%, all doses = 6 – 20 %), 3) diarrhea (placebo = 2 – 9%, all doses = 7 – 17%), 4) eating disorder (placebo = 4 – 8%, all doses = 4 – 19%), and 5) agitation was not reported. No serious adverse events were reported.

Number of studies. Eight studies evaluating tacrine were eligible for this review; tacrine was compared to placebo in six trials^{108, 109, 110, 111, 112, 113} (lecithin was assumed to be like placebo) and to other drugs in two trials.^{114, 26} One study¹¹⁴ compared two arms with tacrine, one with silymarin added and one placebo arm. The other non-placebo controlled trial compared tacrine with idebenone.²⁶ These drug versus drug trials were published in 1999 and 2002. The placebo controlled trials were published 1991,¹¹¹ 1994,^{108, 109, 113} 1996,¹¹⁰ and 1999.¹¹²

Design/methodology. The placebo studies evaluated 994 patients in total with a range of 13¹¹¹ to 663¹⁰⁸ subjects per study. Quality scores out of 8 points were evenly distributed with two studies each having scores of 5, 6 and 7. Both drug versus drug trials had scores of 7. One study¹⁰⁹ was not funded by industry, and one trial¹¹⁴ did not report its source of funding; the other six studies had at least partial, if not full, industry support.

Populations. All studies included AD subjects; one trial²⁶ also included PDD patients. Subjects in all studies had mild to moderate or probable disease. Five of the studies^{113, 109, 111, 114, 26} reported that their sample was from the community. The mean age ranged from 68¹¹⁰ to 75 years,¹¹³ and the percentage of male subjects in the studies varied from 13¹¹⁰ – 54%.¹¹³ One study reported race with 100% white subjects.²⁶

Primary outcomes. All six of our identified domains were evaluated by at least one trial. All trials measured cognition; however, sufficient data to permit pooled analyses could not be adequately abstracted from all these studies that had similar outcomes. Baseline MMSE varied from 14 to 18 in the fours trials^{112, 110, 109, 114} that reported this.

Analysis. Five studies^{108, 112, 113, 114, 26} reported ITT analysis and three did not.^{109, 111, 110}

Results and interpretation. Of the six placebo-controlled studies, only one trial¹⁰⁸ showed statistical significance for general cognitive function as measured by the ADAS-cog (ES = -0.268). Three doses (80 mg,120 mg,160 mg) were compared in this trial, and the 120 and 160 mg per day were shown to be statistically significant (approximately a mean change of 2 points on the ADAS-cog). One trial¹¹² showed mixed results for the two outcomes used (CASI and MMSE) to evaluate general cognitive function; this trial was underpowered for both these outcomes (PW = 0.22 and 0.26, respectively). Three trials^{109, 110, 111} found no statistical differences between treatment and placebo but had small sample sizes, ranging from 12 to 32 subjects, and were likely underpowered (insufficient reporting to estimate power). A fourth trial¹¹³ also found no statistical difference (p = 0.55) for general cognitive function, but the study duration was 12 weeks. It should be noted that this study used an 80 mg dose, which was shown to have no benefit relative to higher doses of 120 and 160 mg.¹⁰⁸ A single trial¹⁰⁹ of small sample size evaluated specific cognitive tests and did not show statistical differences.

Three studies evaluated global assessment, and two ^{113, 108} found statistical significance; the trial showing no benefit¹¹² also showed inconclusive findings for general cognitive function as well (PW = 0.05 for the CGIC). Four trials^{108, 109, 110, 113} evaluated behavior/mood and showed no difference between groups. Two trials^{109, 111} with small sample sizes measured quality of life/ADL and showed no significant changes; lack of sufficient power cannot be ruled out. There were not enough similar outcomes reported to complete a pooled analysis for tacrine.

The quality scores for reporting adverse events varied from 1 to 3. The proportion of subjects withdrawing due to adverse events ranged from 0 – 12% for placebo and 0 – 55% in the treatment group. The higher rates of withdrawal were associated with higher doses. Elevated alanine transaminase (ALT) or hepatic abnormality (placebo = 4 – 13%, all doses = 7 – 67%) was reported in six studies, suggesting the potential for serious liver damage. None of these trials tested for differences between treatment and placebo with respect to adverse events. Five of the studies reported nausea and vomiting (placebo = 0 – 9%, all doses = 9 – 37%); gastrointestinal problems and dizziness (placebo = 0 – 16%, all doses = 4 – 14%) was also noted in several studies. Frequencies of other a priori symptoms of interest are as follows: 1) agitation (placebo = 5 – 12%, all doses = 5 – 9%), and 2) diarrhea (placebo = 0 – 13%, all doses = 4 – 18%).

Number of studies. Three studies evaluated velnacrine versus placebo, and these were published in 1991,¹¹⁵ 1995,¹¹⁶ and 1996.¹¹⁷

Design/methodology. A total of 774 subjects were studied with sample sizes ranging from 16¹¹⁵ to 449.¹¹⁶ Quality scores out of a possible 8 points varied from scores of 6^{117, 115} or 7.¹¹⁶ All studies were sponsored by industry.

Populations. The characteristics of the populations all included probable AD subjects. The mean age of the participants ranged from 70.5 to 72.8 years and the percentage of male subjects ranged from 31 – 41%. Location of recruitment was not specified.

Intervention. The doses given for this drug overlapped between the studies, but they were on different schedules (once, twice, or three times per day). None of the studies had a titration period. One study¹¹⁷ compared four doses (three daily doses of 10 mg, 25 mg, 50 mg, and 75 mg). One study¹¹⁶ compared doses of 150 mg or 225 mg per day. The other study used a dose of 100 mg twice daily and had the smallest sample size.

Primary outcomes. General cognitive function was evaluated in all studies with the ADAS-cog, and none specified baseline MMSE values. A variety of outcomes were used to evaluate global assessment. At least one of these trials evaluated the other outcomes domains

Analysis. Only one of the trials¹¹⁶ used an ITT analysis.

All studies included assessment of global functioning, for which two^{117, 116} found significant differences, and one¹¹⁵ had mixed results. Behavior/mood was evaluated in only one study¹¹⁷ as a secondary outcome with an OC analysis, and no significant effect was found. Similarly, quality of life/ADL was measured in two studies as secondary outcomes, which produced opposite results (not significant¹¹⁷ and significant¹¹⁶). One of the trials¹¹⁶ measured effects on caregiver burden as a secondary outcome and found a significant effect. There were not enough similar outcomes reported to complete a pooled analysis for velnacrine.

Quality scores for reporting adverse events were 3 for all studies. Withdrawal rates varied from 0 – 22% for the placebo group and 5 – 33% for the treatment group. None of the studies reported a dose response. None of the studies tested for statistical differences between the placebo and treatment groups. Two studies reported aberrant hematology and hepatic abnormality^{116, 117}; for these two studies the rate of occurrence were 2 – 21% for placebo, and 32 – 40% for all doses. All studies reported diarrhea and nausea and vomiting. The range of frequencies of the a priori symptoms of interest are as follows: 1) nausea (placebo = 0 – 4%, all doses = 3 – 8%), 2) dizziness (placebo = 3%, all doses = 0 – 8%), 3) diarrhea (placebo = 3%, all doses = 2 – 33%), 4) agitation (placebo = 4%, all doses = 1 – 4%), and 5) eating disorder (placebo = 1%, all doses 2 – 4%).

The remaining agents classified as cholinergic neurotransmitter modifying agents were grouped according to the number of studies for the purposes of presentation:

Cholinergic pharmacological agents that had two trials eligible for this review and were compared to placebo.

Cholinergic pharmacological agents with only one trial eligible for this systematic review.

Results of non-cholinergic neurotransmitter/neuropeptide modifying agents (NCNMA)

Table 3. List of Non-cholinergic neurotransmitter/neuropeptide (more...)

Table 3. List of Non-cholinergic neurotransmitter/neuropeptide modifying agents and the number of studies vs. placebo for each of these. Asterisk (*) indicates report of a drug vs. drug trial [comparator drug(s) in brackets]

Drug	Number of studies vs. placebo
Alaproclate	1
Alprazolam *[Lorazepam]	0*
Anapsos	1
BMY (Nootropic)	1
Carbamazepine	2
Citalopram *[Mianserin] *[Perphenazine]	2**
Diphenhydramine *[Haloperidol, Oxazepam]	0*
Divalproex	2
Fluoxetine *[Haloperidol] *[Amitriptyline]	2**
Fluvoxamine	1
Haloperidol **[Risperidone] *[Loxapine] *[ Diphenhydramine Oxazepam ] *[Fluoxetine] *[ Tiapride] *[ Trazodone]	4*******
Imipramine *[Paroxetine]	1*
Lisuride	1
Lorazepam *[Alprazolam] *[Olanzapine]	1**
Loxapine *[Haloperidol] *[Thioridazine]	1**
Lu25-109	1
Maprotiline	1
Melperone *[Tiapride]	0*
Memantine	3
Mianserin *[Citalopram]	0*
Minaprine	1
Moclobemide	1
Naftidrofuryl	1
Olanzapine *[Lorazepam]	2*
Oxazepam *[Diphenhydramine Haloperidol]	0*
Paroxetine *[Imipramine]	0*
Perphenazine *[Citalopram]	1*
Phosphatidylserine	2
Risperidone **[Haloperidol]	1**
Selegiline *[Vitamin E]	7*
Sertraline	2
Thioridazine *[Loxapine]	1*
Tiapride *[Haloperidol] *[Melperone]	1**
Trazodone *[Haloperidol] *[5′-MTHF]	1**
Xanomeline	1

Number of studies. Five studies^{124, 125, 126, 127, 128} evaluating haloperidol relative to placebo as well as another drug were included in this review. Three additional trials^{129, 130, 131} (from 1990, 1993, and 2001) compared haloperidol to another drug and did not include a placebo group (these are detailed in question 3). One study was published in each of 1982, 1997, and 1999; two were from 2000.

Design/methodology. Sample sizes for the placebo-controlled studies were generally small with samples of 15,¹²⁸ 64,¹²⁶ 149,¹²⁵ 344,¹²⁷ and 306¹²⁴ for an overall total sample size of 622 subjects. All but one of the studies had a quality score of 6 out of a possible 8 points; the other study¹²⁷ had a score of 7 points. One study¹²⁴ did not indicate a funding source, three studies^{125, 126, 127} indicated some industry funding, although none showed total industry funding, and one study¹²⁸ had no industry funding.

Populations. Populations evaluated in the studies included three with only mild to moderate or probable AD,^{124, 125, 128} one with PDD and MID,¹²⁶ and one with PDD, VaD, and mixed dementia¹²⁷ (which reported subgroup information about VaD versus all subjects). Two placebo studies¹²⁶ reported the presence of subjects with severe disease. Two trials^{126, 127} studied institutionalized patients while one¹²⁸ looked at community subjects. Ages in the studies ranged from a mean of 72.7 to 81.0 years, and 33 – 49% of subjects were male.

Intervention. Haloperidol doses ranged from 3 mg to 20 mg per day for a treatment period of 3 weeks,¹²⁴ 6 weeks,¹²⁸ 10 weeks,¹²⁶ 12 weeks¹²⁷ or 16 weeks.¹²⁵ The other drugs that haloperidol was compared to included fluoxetine,¹²⁸ loxapine,¹²⁶ risperidone,¹²⁷ tiapride,¹²⁴ trazodone & BMT¹²⁵ in the placebo controlled studies; loxapine,¹²⁹ risperidone,¹³¹ oxazepam & diphenhydramine¹³⁰ were evaluated in the head to head comparisons.

Primary outcomes. All studies evaluated behavioral outcomes, and at least one study evaluated the effect of haloperidol in each of the other domains included in this review with the exception of specific cognitive function. None of the studies reported baseline MMSE values.

Analysis. Two studies^{124, 127} reported ITT analysis and three^{128, 126, 125} did not.

The quality scores for reporting adverse events varied from 1 to 5. Only three of five studies reported withdrawal rates; the proportion of subjects withdrawing due to adverse events ranged from 5 – 17% for placebo and 17 – 33% in the treatment group. One trial¹²⁴ showed a dose response effect, but the study only lasted for 3 weeks. Three trials tested for differences between treatment and placebo with respect to extrapyramidal symptoms (placebo = 17 – 32%, all doses = 34 – 97%), and two^{124, 125} found significant differences. One study¹²⁵ found significant differences between groups for balance-related problems. Although reported by only two trials, the range of frequencies of the a priori symptoms of interest are as follows: 1) nausea (placebo = 3%, all doses = not reported), and 2) dizziness (placebo = 24%, all doses = 21%), 3) no frequencies were reported for agitation, diarrhea, or eating disorder.

Number of studies. Three studies comparing memantine to placebo were eligible for review. One study was published in 1999¹³² and the other two^{133, 134} in 2002.

Design/methodology. Sample sizes ranged from 166 to 579 for a total population evaluated of 1066 subjects. Two studies^{132, 133} earned 6 points out of 8 for the quality score while the other¹³⁴ earned 7 points. One report¹³² did not indicate the source of funding, and the other two had industry support or funding.

Populations. Two studies^{133, 134} included VaD patients only, one of which¹³⁴ analyzed subgroups based on MMSE, type of VaD, and gender. The other study¹³² included VaD, DAT and PDD patients and did subgroup analysis comparing VaD to DAT and grouping for care dependence. One trial¹³² included patients with severe disease and was the only study to report that all of their subjects were institutionalized. One study¹³⁴ included only community subjects and the other study¹³³ did not report source of patients. Study subjects had a mean age of 71.2,¹³² 76.4,¹³³ and 77.4¹³⁴ years, and 42 – 53% were male.

Intervention. Two studies^{133, 134} had a 4 week titration period with a final dose of 20 mg per day for the remaining 24 week study duration. The third study used a 2-week titration period with a final dose of 10 mg per day for the remaining 10 weeks of the study.

Primary outcomes. All studies evaluated global function. The ADAS-cog was evaluated in two studies^{134, 133} and showed smaller changes of decline relative to placebo by approximately 1.5 points. All studies measured global function with the CGI-C but did not provide variance data to permit the calculation of the pooled estimates. Although, all trials measured MMSE, none reported baseline values for this outcome. Only one trial¹³² evaluated behavior/mood and quality of life/ADL. No study evaluated specific cognitive function or caregiver burden.

Analysis. All studies performed ITT analysis.

The quality scores for reporting adverse events varied from 3 to 4. Only two of three studies reported withdrawal rates; the proportion of subjects withdrawing due to adverse events ranged from 7 – 13% for placebo and 9 – 12% in the treatment group. One trial¹³³ tested for differences between treatment and placebo, and none of the comparisons were significant. The range of frequencies of the a priori symptoms of interest are as follows: 1) nausea (placebo = 3%, all does = 5%), 2) dizziness (placebo = 3 – 8%, all doses = 6 – 11%), 3) diarrhea (placebo = 4%, all doses = 4%), 4) agitation (placebo = 7 – 8%, all doses = 4 – 5%), and none reported eating disorder as an adverse event.

Number of studies. Six studies evaluated the effect of selegiline compared to placebo.^{135, 136, 137, 138, 139, 140} A single study¹³⁵ compared selegiline to vitamin E, placebo and the combination of selegiline plus vitamin E . The studies were published in 1991,¹³⁶ 1992,¹³⁷ 1993,¹³⁸ 1997,¹³⁵ 1998,¹⁴⁰ and 1999.¹³⁹

Design/methodology. Sample sizes ranged from 10¹³⁷ to 341¹³⁵ with a total population evaluated of 733 subjects. Study quality scores were 5,^{140, 135, 137} 6,^{139, 138} and 7.¹³⁶ Three trials^{139, 137, 136} did not report the source of funding, and the other three^{140, 135, 138} had some industry support.

Populations. Studies included patients with mild to moderate PDD, DAT, and AD. Two subgroup analyses based on the results of the clock drawing test¹³⁹ and the GDS result¹³⁶ were reported. One study reported that the included patients were institutionalized.¹³⁹ Mean age of the subjects in the trials ranged from 68.6 to 83.0 years and all had male subjects (29 – 74%).

Intervention. All trials used the same dose, 10 mg per day, with three of the trials^{135, 138, 137} giving the drug in two 5 mg doses. One trial¹⁴⁰ reported a titration period of 7 days. The duration of the trials varied with treatment times of 2 months,¹³⁷ 3 months,¹³⁶ 6 months,^{139, 140} 15 months,¹³⁸ and 24 months.¹³⁵

Primary outcomes. Quality of life/ADL and caregiver burden were not evaluated in any of the studies.

Analysis. Two studies^{135, 140} carried out ITT analysis.

There is some evidence that shows that selegiline and selegiline combined with vitamin E, increases the time to important functional decline milestones¹³⁵ using time to event in the survival analysis. The results of this study showed that the vitamin E, selegiline, and combined groups were statistically different (i.e., declined less) from the placebo group in analyses that included baseline MMSE score as a covariate (not significant when excluded). The median survival was 230 days (vitamin E), 215 days (selegiline), and 145 days (combined group). Moreover, the vitamin E group showed a statistically significant difference for the endpoint of institutionalization, and the other treatment groups did not. Thus, the findings of this study suggest that selegiline and vitamin E may delay clinically important deterioration in patients with moderately severe AD; this delay varied from 20 to 32 weeks. It should be noted that this study evaluated subjects over a 2 year period, the longest of any dementia trial; moreover, the population was moderate to severe with respect to severity.

The quality scores for reporting adverse events varied from 0 to 3. The proportion of subjects withdrawing due to adverse events ranged from 0 – 4% for placebo and 0 – 9% in the treatment group. Two trials^{137, 138} did not report any adverse events. Only one trial¹³⁵ tested for differences between the treatment and placebo groups and showed that balance (worse) and falls were significantly different between groups (particularly the group with selegiline combined with vitamin E (22%) versus placebo (5%)). However, when adjusted for multiple comparisons, these were no longer significant. The range of frequencies of the a priori symptoms of interest are as follows: 1) nausea (placebo = 2%, all doses = 0%), 2) dizziness (placebo = 2 – 20%, all doses = 0 – 30%), and 3) agitation (placebo = 4 – 16%, all doses = 4 – 23%); no trial diarrhea or reported eating disorder as an adverse event.

Ten non-cholinergic neurotransmitter neuropeptide modifying agents versus placebo were studied in only two included trials:

Results of other agents

Table 4. List of Other pharmacological agents and the (more...)

Table 4. List of Other pharmacological agents and the number of studies vs. placebo for each of these. Asterisk (*) indicates report of a drug vs. drug trial [comparator drug(s) in brackets]

Drug	Number of studies vs. placebo
Aniracetam	1
5′-MTHF *[Trazodone]	0*
Amitriptyline *[Fluoxetine]	0*
Ateroid	1
Buflomedil	1
Cerebrolysin	6
Choro-San	1
Choto-San	1
Citicoline *[Posatirelin] *[Sulphomucopolysaccharides]	0**
Cyclandelate	2
Denbufylline	1
Desferrioxamine	1
Diclofenac	1
Ergokryptine (CMB 36-733)	1
Ergokryptine (Dek)	1
Estrogens	5
Ginkgo Biloba	3
Glycosaminoglycan Polysulfate	1
Guanfacine	1
Hydergine *[Pyritinol]	1*
Hydroxychloroquine	1
Idebenone *[Tacrine]	4*
Indomethacin	1
Misoprostol *[Diclofenac]	0*
Monosialotetrahexosylganglioside (GM-1)	1
N-Acetylcysteine	1
Nimesulide	1
Nimodipine	2
Nizatidine	1
Nootropic	1
ORG 2766	2
Oxiracetam	5
Pentoxifylline *[Sulodexide]	3*
Piracetam	1
Prednisone	1
Propentofylline	4
Pyritinol *[Hydergine]	0*
Silymarin + Tacrine *[Placebo + Tacrine]	0*
Simvastatin	1
Sulphomucopolysaccharides *[Citicoline]	0*
Sulodexide *[Pentoxifylline]	0*
Thiamine	1
Vasopressin (DDAVP)	1
Vincamine	1
Vitamin E *[Donepezil] *[Selegiline]	1**
Xantinolnicotinate	1

Number of studies. Six included studies ^{168, 169, 170, 171, 172, 173} compared cerebrolysin to placebo. One report was from 1994,¹⁷² one from 1999,¹⁷¹ two from 2000,^{169, 170} one from 2001,¹⁶⁸ and one from 2002.¹⁷³

Design/methodology. The sample size in the studies ranged from 53¹⁶⁹ to 192¹⁷³ with a total of 819 subjects. The quality of studies varied from scoring 6¹⁷¹ to 8^{168, 173} points out of a possible 8 points. One study¹⁷² did not indicate the source of funding, one trial had non-industry funding,¹⁷³ and the four remaining trials were funded by industry.^{168, 169, 170, 171}

Populations. All but one of the six studies included AD patients; one study¹⁷¹ evaluated patients who had mild to moderate VaD. Mean ages of the subjects in the studies ranged from 69.7¹⁷¹ to 74.1 years.¹⁷³ The proportion of males in the trials varied from 34 – 69%, and only one trial¹⁷³ specified the proportion of Caucasians.

Intervention. All of the studies used the same dose of cerebrolysin, 30 ml per day, for 5 days per week. One trial¹⁷² was for 28 days, four studies^{169, 170, 171} lasted 4 weeks, one trial¹⁶⁸ lasted 16 weeks, and one trial¹⁷³ lasted 24 weeks.

Primary outcomes. Most studies evaluated general cognitive function and three trials^{168, 169, 173} used the ADAS-cog. Baseline MMSE was reported in a single trial¹⁷³ with a score of 21. All studies evaluated global function, and at least two studies evaluated one outcome in each of the remaining domains with the exception of caregiver burden.

Analysis. All but one¹⁷³ of the studies used ITT analysis.

Figure 19. Weighted Mean Difference (WMD) from the (more...)

   Figure 19. Weighted Mean Difference (WMD) from the Random Effects Model (Random) for the ADAS-cog comparing cerebrolysin versus placebo

Figure 19

Figure 20. Odd Ratio (OR) from the Random Effects Model (more...)

   Figure 20. Odd Ratio (OR) from the Random Effects Model (Random) for the CGI comparing cerebrolysin versus placebo

Figure 20

Two^{169, 172} of the six trials scored 5 out of 5 on our quality scale for rating adverse events, yet they did not report any adverse events. Two studies^{173, 168} scored 4, and the other two trials scored 3¹⁷¹ and 2.¹⁷⁰ All the studies with scores equals to 4 or less tested for statistical differences in adverse events between placebo and treatment groups. Withdrawals due to adverse events were not reported in one study,¹⁷⁰ and were 1% in two studies.^{173, 168} Three studies^{169, 172, 171} reported no withdrawals. A significant difference between treatment and control group was reported in one study¹⁷³ for weight change, anxiety, and headache. The range of frequencies of the a priori symptoms of interest are as follows: 1) nausea (placebo = 10 – 24%, all doses = 3 – 21%), 2) dizziness (placebo = 0 – 12, all doses = 1 – 8%), and 3) agitation (placebo = 1%, all doses = 0%), and none reported diarrhea or eating disorder as an adverse event.

Number of studies. Five studies^{174, 175, 176, 177, 178} evaluated estrogens for dementia patients: one published in 1999,¹⁷⁸ three in 2000,^{176, 177, 175} and one in 2001.¹⁷⁴ None compared estrogens to another drug.

Design/methodology. The number of subjects included in the studies ranged from 15¹⁷⁸ to 120 subjects¹⁷⁷ with a total of 247 patients. Quality of the studies ranged from 5¹⁷⁴ to 8¹⁷⁸ points out of a possible 8 points. All studies were partially or fully funded by industry.

Populations. Four of the studies^{176, 177, 175, 174} included patients with mild to moderate AD, and one study¹⁷⁸ included moderate to severe dementia patients who were all institutionalized. Only one of the studies¹⁷⁸ included male subjects. Mean age ranged from 71.8¹⁷⁵ to 80.0 years¹⁷⁴ in the AD studies, and it was 83.8 in the dementia study.¹⁷⁸

Intervention. One of the studies with AD patients used 0.10 mg per day¹⁷⁴ for 8 weeks, and the others used 1.25 mg per day for 12 weeks,¹⁷⁵ 16 weeks,¹⁷⁶ and 52 weeks.¹⁷⁷ The study¹⁷⁸ including subjects with severe disease used 2.5 mg per day for 4 weeks.

Primary outcomes. At least one study evaluated each of the included domains with the exception of caregiver burden.

Analysis. Two of the studies^{177, 175} performed ITT analysis and the other three used OC analysis.

One¹⁷⁸ of the five trials scored 5 out of 5 on our quality scale for rating adverse events, and surprisingly, this same trial did not report any adverse event. Two trials^{176, 177} scored 3; one trial¹⁷⁵ scored 2, and one¹⁷⁴ scored 1. This latter study reported adverse events, but did not test for significant differences between groups. Withdrawal rates due to adverse events ranged from 0 – 5% for placebo and 0 –14% for the treatment group. The most frequently reported adverse event was vaginal bleeding,^{175, 177, 176} and a single trial¹⁷⁵ reported a significant difference between placebo and treatment group for vaginal bleeding. It was not clear from the descriptions provided in the study if they had ascertained whether vaginal bleeding was present prior to the trial commencement. Nausea was the single a priori symptom of interest that was reported and by a single trial; frequencies varied from 0% for the placebo group and 4% for the treatment group.

Number of studies. Three studies^{179, 180, 181} evaluating Ginkgo biloba were eligible to be included in this review. All of the studies compared the drug to placebo only. One of the studies was reported in 1996¹⁸¹ and two were reported in 1997.^{179, 180}

Design/methodology. The studies included evaluated 20 subjects,¹⁸⁰ 216 subjects,¹⁸¹ and 327 subjects¹⁷⁹ (totaling 563 subjects). Two of the reports^{179, 181} scored 8 quality points out of a possible 8 points, and the other¹⁸⁰ earned 6 points. One study did not indicate the funding source,¹⁸⁰ and the other two had industry funding.

Populations. All of the studies included a mix of dementia diagnoses as follows: 1) mild to moderately severe AD and MID,¹⁷⁹ 2) mild to moderate DAT and PDD¹⁸⁰, and 3) mild to moderate DAT and MID¹⁸¹ in community dwelling patients. Two of the studies reported subgroup analysis, one comparing diagnoses and comparing effects based on baseline MMSE score¹⁷⁹ and the other based on diagnosis.¹⁸¹ The patients in these trials had mean ages of 64.6,¹⁸⁰ 69.0,¹⁷⁹ and 69.6 years.¹⁸¹

Intervention. Two of the trials gave 240 mg per day for 3 months¹⁸⁰ and 6 months,¹⁸¹ and the other trial¹⁷⁹ gave 40 mg three times daily for 12 months.

Primary outcomes. None of the studies reported on quality of life/ADL or caregiver burden.

Analysis. All of the trials used an ITT analysis.

One¹⁸⁰ of the three trials scored 5 out of 5 on our quality scale for rating adverse events. One study¹⁸¹ scored 4, and one trial¹⁷⁹ scored 3. Two studies^{181, 180} had no withdrawals due to adverse events, and one trial¹⁷⁹ had a withdrawal rate of 6% for both placebo and treatment groups. Two studies^{179, 180} did not report any adverse event. One study¹⁸¹ reported a statistically significant difference between the treatment and the placebo group for skin disorders. The same study reported gastrointestinal and headache adverse effects, but did not test for statistical differences between the placebo and the treatment group. None of the trials reported any of the a priori symptoms of interest.

Number of studies. Four studies^{182, 183, 184, 185} were included in this review that evaluated idebenone versus placebo, and one study²⁶ compared idebenone to tacrine but not to placebo. The placebo trials were published in 1992,¹⁸⁴ 1994,¹⁸² 1997,¹⁸⁵ and 1998,¹⁸³ the tacrine trial was published in 2002 by the same author as a previous placebo trial.¹⁸³

Design/methodology. Sample sizes in the studies ranged from 92¹⁸² to 450 subjects¹⁸³ with a total of 950 patients in the placebo-controlled studies. The study comparing idebenone with tacrine included 203 subjects, but a large number withdrew, and only 44 completed the trial. One of the trials¹⁸⁵ earned 5 points out of a possible 8 points on the quality scale, two of the trials earned 6 points,^{182, 183} and one earned 7 points.¹⁸⁴ None of the placebo studies reported their funding source. The tacrine study earned 7 points on the quality scale and was partially funded by industry.

Populations. The studies included patients with AD, MID, PDD, and DAT. Two of the trials^{182, 183} reported that the subjects had mild to moderately severe disease and the remainder reported mild to moderate disease. Two of the studies reported subgroup analysis based on disease severity.^{185, 183} Mean ages in the studies ranged from 69.9¹⁸³ to 73.6 years.¹⁸⁴

Intervention. Dosing schemes were 30 or 90 mg per day for 6 months,¹⁸⁵ 30 mg three times per day for 3 months,¹⁸² 45 mg twice daily for 4 months,¹⁸⁴ and 120 mg three times per day for 12 months.¹⁸³ The tacrine trial used 360 mg per day for 14 months.

Primary outcomes. Caregiver burden was the only domain in this review that was not evaluated in at least one of the studies.

Analysis. Two of the studies^{183, 185} used ITT analysis while the other two used OC analysis. The tacrine trial used ITT analysis.

Quality scores for reporting adverse events varied from 1 to 5. Rates of withdrawal due to adverse events varied from 0 – 5% for the placebo group and 0 – 5% in the treatment group; a single trial¹⁸³ did not report withdrawal rates. Two trials^{183, 185} tested for statistical differences between groups and found no differences. Although no clear pattern emerges, three studies identified at least one balance-related adverse event across studies. The range of frequencies of the a priori symptoms of interest are as follows: 1) nausea (placebo = 2%, all doses = 2 – 11%), 2) dizziness (placebo = not reported, all doses = 2%), and 3) not reported for diarrhea, agitation, or eating disorder as an adverse event.

Number of studies. Five trials^{186, 187, 188, 189, 190} included in this review evaluated oxiracetam versus placebo. The studies were published in 1988,¹⁸⁸ 1989,¹⁸⁷ and 1992.^{190, 189, 186}

Design/methodology. A total of 554 patients were included in the studies, ranging from 30 patients¹⁸⁸ to 289 patients.¹⁸⁷ Four of the studies earned 6 points out of a possible 8 points on the quality scale, and the other study¹⁸⁹ earned 4 points. Two of the studies^{189, 187} did not report the source of their funding, and the other three trials had partial industry funding.

Populations. The trials included a mixture of diagnoses, including AD, PDD, mixed dementia, and MID, and none of the studies reported severe disease. One of the studies¹⁸⁷ performed subgroup analysis based on diagnosis, comparing MID to PDD. The mean age of the subjects included in the trials ranged from 62.0¹⁸⁸ to 73.8 years.¹⁸⁹

Intervention. All of the trials used a dose of 800 mg twice daily, for a duration of 12^{186, 187} to 26 weeks.¹⁸⁹

Primary outcomes. At least one trial evaluating oxiracetam evaluated one of the outcome domains examined in this review with the exception of caregiver burden. A single trial¹⁸⁹ reported baseline MMSE at 22 for both placebo and treatment groups.

Analysis. None of the trials used ITT analysis.

The quality scores for reporting adverse events varied from 2 to 5. The proportion of withdrawals due to adverse events varied form 0 – 9% for the placebo group and 0 – 6% for the treatment group. No clear pattern for adverse events is evident, but three of the five studies reported gastrointestinal related problems, primarily associated with abdominal pain. Although, only single trials evaluated the range of frequencies of the a priori symptoms of interest are as follows: 1) dizziness (placebo = not reported, all doses = 11%), and 2) agitation (placebo = 1%, all doses = not reported); no trial reported nausea, eating disorder, or diarrhea as an adverse event.

Number of studies. Three trials^{191, 192, 193} in this review evaluated pentoxifylline versus placebo. One trial,¹⁹⁴ published in 1997, compared pentoxifylline to sulodexide rather than placebo. The placebo trials were published in 1987,¹⁹³ 1992,¹⁹² and 1996.¹⁹¹

Design/methodology. The studies included 36 patients,¹⁹³ 64 patients¹⁹² and 289 patients.¹⁹¹ The sulodexide trial included 93 patients. All placebo trials had 6 points out of a possible 8 points on the quality scale and had partial or full industry funding. The sulodexide trial earned 5 points on the quality scale and did not report the source of funding.

Populations. The three placebo-controlled trials included patients with mild to moderate MID, and one trial¹⁹³ also included PDD patients. The sulodexide trial had only patients with mild to moderate VaD. Subgroup analysis was performed in two trials, looking at MID versus PDD diagnosis¹⁹³ and grouping by vascular change versus discrete stroke.¹⁹² The mean age of the studies ranged from 69.7¹⁹¹ to 77.0 years.¹⁹³

Intervention. All of the studies gave 1200 mg per day of pentoxifylline; one study gave the drug once a day for 9 months,¹⁹¹ one study gave 400 mg three times per day for 9 months,¹⁹² and one gave 400 mg three times per day for 3 months.¹⁹³ The sulodexide study gave the drug once a day for 6 months.

Primary outcomes. At least one trial evaluated one of the outcome domains examined in this review with the exception of caregiver burden.

Analysis. A single¹⁹¹ trial used an ITT analysis.

The quality scores for reporting adverse events were generally low, varying from 1 to 3. Withdrawal rates due to adverse events varied from 0 – 25% in the placebo group and 0 – 22% in the treatment group. The two studies that reported adverse events indicated the presence of gastrointestinal disturbances, including abdominal pain or nausea and vomiting (placebo = 7% and all doses = 14%). None of the trials reported dizziness, agitation, eating disorder or diarrhea.

Number of studies. Four studies^{195, 196, 197, 198} in this review evaluated propentofylline versus placebo. The first trial was published in 1990.¹⁹⁸ Two, by the same author, were published in 1996¹⁹⁶ and 1998.¹⁹⁷ One was published in 1997.¹⁹⁵

Design/methodology. The number of subjects in the studies ranged from 30 subjects^{197, 196} to 260 subjects,¹⁹⁵ with a total of 510 subjects. Three of the studies^{197, 195, 196} earned 5 points out of a possible 8 points on the quality scale, and the other study¹⁹⁸ earned 6 points. Only one study indicated the source of funding for the trial,¹⁹⁶ and it was industry-supported.

Populations. The trials included subjects with mild to moderate AD only,¹⁹⁷ VaD only,¹⁹⁶ mild dementia only,¹⁹⁸ and mild to moderate combined AD and VaD.¹⁹⁵ Two trials presented subgroup analysis: one for AD versus VaD,¹⁹⁵ and one based on MMSE baseline score. The mean age in the studies ranged from 64.8 ¹⁹⁷ to 72.4 years.¹⁹⁵

Intervention. All four studies gave 300 mg three times a day for 3 months with the exception of one trial¹⁹⁵ which had a duration of 12 months.

Primary outcomes. At least one trial evaluated one outcome in each of the domains examined in this review with the exception of caregiver burden. Baseline MMSE was reported in two trials^{197, 198} and varied from 20 and 21 for both placebo and treatment groups.

Analysis. One of the studies¹⁹⁵ used an ITT analysis.

Figure 21. Weighted Mean Difference (WMD) from the (more...)

   Figure 21. Weighted Mean Difference (WMD) from the Random Effects Model (Random) for the MMSE change score comparing propentofylline versus placebo

Figure 22. Weighted Mean Difference (WMD) from the (more...)

   Figure 22. Weighted Mean Difference (WMD) from the Random Effects Model (Random) for the DSST change score comparing propentofylline versus placebo

Figure 21

Figure 22

The quality scores for reporting adverse events varied from 1 to 4. The percentage of withdrawals varied from 0 – 13% for the placebo group and 0 – 12% for the treatment group. None of the trials tested for differences between groups. Three of the trials^{195, 197, 198} reported gastrointestinal events that included abdominal pain, constipation, and nausea and vomiting (placebo = 2%, all doses = 7%). Dizziness (placebo = 3 – 5%, all doses = 1 – 6%) was the only other a priori symptom of interest.

Interventions with two studies included for review. Six other agents were compared with placebo in only two included trials:

Delay of Onset of Dementia

The concept of “delay onset” was operationalized to imply delay in conversion from a cognitive disturbance state, classified as MCI, CLOND or CIND, to a true dementia state. No studies with this population met the final eligibility criteria, although four trials^{231, 232, 233, 234} advanced to the full text screening stage. The lack of studies eligible for evaluation in this systematic review points to a gap in the literature for pharmacological interventions (attempting to demonstrate a delay in disease onset) in MCI-type populations.

Delay of Progression

In general, very few studies evaluated patients who were classified as “severe”. Five studies^{126, 208, 129, 178, 132} had moderate to severe groups of dementia patients, and only one trial reported all three levels¹⁶³ of the disease spectrum. The interventions evaluated in these trials were estrogen, haloperidol, glycosaminoglycan polysulfate, memantine, and naftidrofuryl. This suggests that there is a bias in the trials eligible in this systematic review towards evaluating mild to moderate disease; this in turn reflects the underlying assumption that the less severe groups are most likely to benefit from drug trials. Since so few studies have evaluated the more severe groups, this assumption may require some empirical justification. Therefore, delay in progression has not been considered in severe patients.

The selected studies used two approaches for showing “delaying disease progression”. The first method for evaluating the potential for a drug to delay disease progression used longer-term follow-up; survival analyses (time to a relevant event) were then used to show differences between the two groups. The second design approach used withdrawal from treatment for a period and continued monitoring of the treatment and placebo groups (to demonstrate a deviation of the treatment group from the natural history as represented by the placebo group). Such designs have been termed withdrawal, active-extension, randomized withdrawal, randomized start, and staggered start.^{235, 236, 19, 21} From our 186 included studies, we then further selected a subgroup of papers that had the potential to demonstrate delay in disease progression through the use of one of these two designs. Therefore, any eligible trial that employed a survival analysis or a two-period approach, where the pharmacological agent was withdrawn during one of the periods, was selected for further evaluation to answer this question.

Survival Analyses

Two studies^{135, 61} using survival analyses were identified. In a 2-year study¹³⁵ that compared placebo to three other groups (selegiline, selegiline with vitamin E, and vitamin E), time to the development of significant dementia milestones (death, institutionalization, loss of ability to perform ADL, or score on scale indicating severe dementia) was used as the time to event in the survival analysis. The results of this study showed that the vitamin E, selegiline, and combined groups were statistically different (i.e. declined less) from the placebo group in analyses that included baseline MMSE score as a covariate (not significant when excluded). The median survival was 230 days (vitamin E), 215 days (selegiline), and 145 days (combined group). Moreover, the vitamin E group showed a statistically significant difference for the endpoint of institutionalization, and the other treatment groups did not. There were no statistical differences between groups with respect to adverse events. Thus, the findings of this study suggest that selegiline and vitamin E may delay clinically important deterioration in patients with moderately severe AD; this delay varied from 20 to 32 weeks. The second study⁶¹ used survival analyses to evaluated the time to the development of severe functional impairments in a comparison of placebo and donepezil with a follow-up of 54 weeks. The results of the Kaplan-Meier analysis showed a mean number of days to significant functional decline of 252 days for placebo and 357 days for the donepezil group (mean difference of 100 days). The treatment group was 38% less likely to decline over a 1-year period. Both these studies demonstrated some delay in disease progress varying from 100 to 230 days for these three different pharmacological agents.

Staggered Withdrawal

Delay in disease progression can also be evaluated using a “time to return to baseline” following withdrawal of treatment. Similarly, staggering the start of the treatment parallels the staggered withdrawal and can be used to evaluate disease progression. In this design approach, the time to return to baseline is compared to the placebo group, which represents the natural course of the disease. Of the studies that were eligible for this research question used a classic withdrawal design (withdrawal in period II after the intervention was administered); none of these studies were able to maintain double blinding after the withdrawal of the intervention. Justification for the selection of the length of the washout or follow-up period was not consistently provided (which possibly reflects the lack of a priori aim to show delay in progression).

Table 5. Studies that withdrew the treatment agent (more...)

Table 5. Studies that withdrew the treatment agent but maintained at least single blinding

Study	Drug	Schedule	Result
Ruether 2001	Cerebrolysin	4w drug + 8w washout + 4w drug + 12w washout	All patients:
			ADAS-noncog, maintained difference from placebo
			NAI returned to baseline
			Subgroup MMSE<20:
			ADAS Noncog, CGI, ADAS-cog, SKT maintained difference from placebo
Nyth 1990	Citalopram	4w drug + 8 w open drug + 4w new random drug	NR
Rogers 1996	Donepezil	12w drug + 2w SB Pl washout	5 mg maintained effect, 3 mg no maintenance of effect for ADAS-cog (NS)
Rogers 1998b	Donepezil	24w drug + 6w SB and placebo washout	Return to placebo levels for ADAS-cog, MMSE, CIBIC (all NS)
Wilcock 2002	Memantine	2w SB and placebo + 28w drug + 2w SB placebo washout	NR
McKeith 2000	Rivastigmine	20w drug + 3w rest	Return to placebo levels for NPI and computerized cognitive assessment (NS)
Antuono 1995	Velnacrine	2w SB placebo + 24w drug + 6w SB placebo washout	Return to placebo levels for the ADAS-cog but SC for CGIC remained for washout
Bodick 1997	Xanomeline	24w drug + 4 w SB placebo	SC at week 24 with CNTB
			No differences vs. placebo at w4 of washout

Table 6. Studies that withdrew treatment and did not (more...)

Table 6. Studies that withdrew treatment and did not specify if blinding for washout or extension was maintained

Study	Drug	Schedule	Result
Dehlin 1985	Alaproclate	2w placebo + 4w drug + 2w placebo	SC for GBS intellectual subscale at w4 of treatment
			No significant difference at w2 of washout
Cutler 1993	BMY 21,502	12w drug + 4w placebo washout	Treatment showed no significant change and follow-up showed no change
Amaducci 1988	Phosphatidylserine	3m drug + 21 m follow-up	SC remained for severe disease patients, not moderate
Raskind 1997	Metrifonate	26w drug + 8w follow-up	NR
Parnetti 1995	Posatirelin	90d IM + 30d follow-up placebo	“Maintained positive effect” but specific numbers not reported
Agid 1998	Rivastigmine	10w drug + 2w placebo washout	NR

Head to Head Comparisons

A total of 26^{125, 152, 129, 237, 130, 238, 239, 145, 150, 124, 114, 131, 127, 240, 241, 135, 242, 101, 243, 70, 194, 128, 244, 26, 92, 245} studies compared efficacy of two or more pharmacological agents relative to each other. In general, few drugs showed statistically significant differences relative to each other. Those that did include the following (drug performing better is listed first):

1. Sulphomucopolysaccharides versus CDP-choline²³⁸ - Significant differences were seen in favor of sulphomucopolysaccharides in measures of behavior and global assessment in 30 institutionalized patients with mild to moderate MID.

2. Donepezil and vitamin E⁷⁰ - Significant differences were seen in favor of donepezil in general cognitive function in 54 patients with mild AD.

3. Antagonic stress versus nicergoline⁹² - Significant differences were seen in favor of antagonic stress in cognition as well as a global assessments in 62 subjects with mild to moderate AD.

4. Antagonic stress versus meclofenate²⁴² - Significant differences were seen in favor of antagonic stress in measures of cognition and global assessment in 63 patients with mild to moderate AD.

5. Posatirelin versus citicoline¹⁰¹ - Significant differences were seen in favor of posatirelin in general cognitive measure and mood in 222 community living patients with mild to moderate AD.

6. Pyritinol versus hydergine²⁴³ - A significant difference was found in favor of pyritinol in a global assessment measure in 102 Hispanic patients with mild to moderate AD.

7. Idebenone²⁶ versus tacrine-Mixed results were observed; the Efficacy Index Score showing a significant benefit over tacrine, while the global assessment showed no difference in 203 AD patients, 44 of whom completed the study.

Figure 23. Weighted Mean Difference (WMD) from the (more...)

   Figure 23. Weighted Mean Difference (WMD) from the Random Effects Model (Random) for the ADAS-cog comparing donepezil versus placebo

Figure 24. Weighted Mean Difference (WMD) from the (more...)

   Figure 24. Weighted Mean Difference (WMD) from the Random Effects Model (Random) for the ADAS-cog comparing galantamine versus placebo

Figure 25. Weighted Mean Difference (WMD) from the (more...)

   Figure 25. Weighted Mean Difference (WMD) from the Random Effects Model (Random) for the ADAS-cog comparing rivastigmine versus placebo

Figure 26. Weighted Mean Difference (WMD) from the (more...)

   Figure 26. Weighted Mean Difference (WMD) from the Fixed Effects Model (Fixed) for the ADAS-cog comparing tacrine versus placebo

Figure 27. Relative Risk (RR) from the Random Effects (more...)

   Figure 27. Relative Risk (RR) from the Random Effects Model (Random) for the CIBIC comparing donepezil versus placebo

Figure 28. Relative Risk (RR) from the Random Effects (more...)

   Figure 28. Relative Risk (RR) from the Random Effects Model (Random) for the CIBIC comparing galantamine versus placebo

Figure 29. Relative comparison of effect sizes for (more...)

   Figure 29. Relative comparison of effect sizes for studies using the CIBIC rivastigmine versus placebo

Figure 30. Relative comparison of effect sizes for (more...)

   Figure 30. Relative comparison of effect sizes for studies using the CIBIC comparing tacrine versus placebo

Figures 23

30

Summary of the Systematic Review Results

A total of 97 interventions in 186 studies were eligible for evaluation in this systematic review and were distributed as follows:

• A total of 16 different cholinergic neurotransmitter modifying pharmacological agents in 72 studies.

• A total of 35 non-cholinergic neurotransmitter/neuropeptide modifying agents in 61 studies.

• A total of 46 other pharmacological agents in 76 studies^*.

• two studies compared two NCNMAs with an OTHER.

• two studies compared a CNMA with an OTHER.

• one study compared a CNMA with two OTHERS.

• two studies compared an NCNMA with an OTHER.

The evidence for all these pharmacological agents was presented in great detail in Chapter 3 and in Evidence Tables of Key Study Characteristics, Tables of Study Results, and Tables of Study Adverse Events contained in Appendix C. Conclusions regarding those pharmacological agents that had a minimum of three trials are summarized here. The summary of the pharmacological agents that had fewer than three trials can be found in Chapter 3.

Summary of Cholinergic Neurotransmitter Modifying Agents

Carnitine. Six trials evaluated carnitine in 925 subjects with mild to moderate severity, recruited predominately from the community at doses of 2 to 3 g for either 24 or 52 weeks. Evidence of benefit is conflicting for the domains of cognition. Most studies were not statistically significant and the lack of sufficient power may have been an important factor. Similarly, no significant differences were found in the domains of global assessment, behavior/mood, and quality of life/ADL; power could not be evaluated for the majority of these outcomes.

Four of the six studies scored 3 for quality on reporting adverse events. Withdrawal rates due to adverse events varied from 0 – 3% (excluding results from one outlier trial²⁴⁸), and gastrointestinal symptoms were the most frequently reported types of adverse events. The percent of subjects reporting the a priori symptoms of interest across all studies were as follows: 1) nausea (placebo = 6 – 14%, all doses carnitine = 28%), and 2) agitation (placebo = 6%, all doses carnitine = 7%). Dizziness, diarrhea, or eating disorder were not reported by any study. No serious adverse events requiring hospitalization and associated with carnitine were reported.

Donepezil. Ten trials in 3239 subjects evaluated the efficacy of donepezil compared to placebo, and one trial compared donepezil to a group given vitamin E. The majority of studies (n = 8) evaluated AD patients, for which half were recruited from the community (other studies did not specify). The subjects had predominately mild to moderate disease and doses of 5 or 10 mg were used with varying duration from 12 to 56 weeks.

There is consistent evidence of benefit in the domains of general cognitive function and global assessment. The combined effect sizes for the ADAS-cog and the CIBIC were estimated. Evidence is inconsistent for a dose response in these domains based on the three studies that evaluated two different doses (5 and 10 mg); the benefit was of similar magnitude for both dose groups for global assessment outcomes. Similarly, two of the three studies that evaluated behavior/mood outcomes (NPI) showed no statistically significant changes relative to placebo; these trials lacked sufficient power to detect a difference. There is some evidence of benefit in ADL outcomes, although this outcome domain was evaluated with a variety of instruments. Caregiver burden outcomes was evaluated in a single study that did not report the findings for this domain.

Adverse events quality scores were 3 or greater for the majority of studies (n=7). Four trials provided evidence of a dose response for adverse events. One study showed a statistical difference for balance-related problems and asthenia (neurological fatigue) between placebo and treatment groups. Withdrawal due to adverse events ranged from 0 – 18% for treatment groups and 0 – 11% for placebo. Four out of six studies testing differences between groups were statistically significant for diarrhea (placebo = 3 – 21%, all doses donepezil = 0 – 38%), nausea and vomiting (placebo = 4 – 9%, all doses donepezil = 4 – 25%). The other a priori symptom reported was agitation and frequencies for placebo varied from 0 – 8% and for all doses from 3 – 19%; but these were not shown to be statistically different.

Galantamine. Six trials in 3530 subjects evaluated the efficacy of galantamine compared to placebo. Doses of 24 and 32 mg were evaluated in half of these studies. Five studies evaluated AD patients and there was limited information regarding whether the subjects were from the general community or institutional settings. All studies recruited subjects with mild to moderate disease and the drug was administered with varying duration of 3 or 6 months.

Evidence of benefit is consistent in the domains of general cognitive function, global assessment, quality of life/ADL. Two of the three studies that evaluated, behavior/mood found statistically significant differences. A small dose effect was evident in the ADL domain when comparing the pooled estimates of the DAD; no dose effect was observed for outcomes in the global assessment domain, and dose effect could not be evaluated for the general cognition domain. The caregiver burden domain was not evaluated in any trials.

Five of the six trials scored 3 out of 5 on our quality scale for rating adverse events. Withdrawal rates due to adverse events ranged from 4 – 9% for placebo and 8 – 27% for the treatment group. One study showed a dose response for adverse events. Although, most trials did not report testing for differences between groups, two trials reported a statistically significant difference in weight loss with the treatment group having more than the placebo group. The most common types of adverse events reported were gastrointestinal symptoms (nausea and vomiting, diarrhea), eating disorders/weight loss, and dizziness. The range of frequencies of the a priori symptoms of interest are as follows: 1) nausea and vomiting (placebo = 3 – 13%, all doses = 6 – 44%), 2) dizziness (placebo = 3 – 11%, all doses = 4 – 19%), 3) diarrhea (placebo = 2 – 10%, all doses = 4 – 19%), 4) agitation (placebo = 1 – 9%, all doses = 6 – 15%), and 5) eating disorder (placebo = 0 – 6%, all doses = 4 – 20%).

Metrifonate. Nine studies compared metrifonate to placebo in 2759 subjects with mild to moderate AD (likely from community settings as the majority of studies did not specify this). Metrifonate dosages evaluated varied from 50 to 80 mg, and study duration ranged from 21 days to 26 weeks duration.

All but one study showed metrifonate to have a consistent positive effect on measures of general cognitive function; none of the studies evaluated any specific cognitive function measures. The effects on global assessment were less consistent, but suggested a positive effect in four of the eight studies that reported this outcome. Evidence for effect in the domains of behavior/mood and quality of life/ADL were not significant in the majority of studies that evaluated these domains, however these were primarily evaluated as secondary outcomes and likely lacked sufficient power.

With the exception of a single study, quality scores for reporting adverse events were greater than 3. However, only one trial⁸³tested for differences between groups and found nausea and vomiting, diarrhea, and muscle and joint disorder to be significantly different. The range of frequencies of the a priori symptoms of interest are as follows: 1) nausea and vomiting (placebo = 3 – 14%, all doses = 2 – 50%), 2) dizziness (placebo = 1%, all doses = 3 – 4%), 3) diarrhea (placebo = 4 – 14%, all doses = 11 – 19%), 4) agitation (placebo = 2 – 14%, all doses = 8 – 33%), and none reported eating disorder as an adverse event. Withdrawal due to adverse events varied from 0 – 9% for placebo and 0 – 12% for the treatment group. Overall, it was difficult to determine which types of adverse events reported had the potential to cause serious harm. This is noteworthy as metrifonate has been withdrawn from use in North America, and Bayer has suspended Phase III trials,⁸⁷ because some patients in clinical trials have experienced serious muscle weakness . This decision was based on the results of an experimental study showing risk of respiratory paralysis with the use of metrifonate. Other adverse events of concern included severe leg cramps, dyspepsia, and bradycardia. None of the studies we reviewed indicated that if present, these events differed significantly between groups. It is not clear if this inconsistency is a function of the methods used to collect and report adverse events or a limitation of RCTs as a source of detecting serious adverse events when incidence is low.

Nicergoline. Four trials in 705 subjects compared nicergoline to placebo and one trial compared it to antagonic-stress in mixed populations that included AD, MID, PDD, VaD, mixed dementia, and SDAT, which were classified as mild to moderate in severity.

All placebo-controlled trials found a positive effect for general cognitive outcomes, but half the results were based on OC analyses. The evidence was mixed for benefit in the domain of global assessments. No significant differences were found for behavior/mood, and quality of life/ADL outcomes, but these were evaluated in few studies and as secondary outcomes (suggesting that sufficient power was an issue).

Quality scores for reporting adverse events varied from 2 to 5 for these four trials, and none tested for differences between groups. Withdrawal due to adverse events varied from 0 – 8% for placebo and 0 – 9% for the treatment group. With the exception of headache, which was reported in all four trials, it was difficult to determine which types of adverse events most characterized exposure to this pharmacological agent. The range of frequencies of the a priori symptoms of interest are as follows: 1) nausea (placebo = 3%, all doses = 3%), 2) dizziness (placebo = 1–2%, all doses = 0% or not reported), 3) diarrhea (placebo = 2 – 6%, all doses = 2 – 4%), 4) agitation (placebo = 5%, all doses = not reported), and none reported eating disorder as an adverse event.

Physostigmine. Four studies of 1198 subjects with mild to moderate AD evaluated physostigmine administered in patch and oral form (30 to 60 mg dose) for study duration varying from 6 to 24 weeks. All subjects were recruited from the community.

There is evidence that physostigmine has a statistically significant effect on general cognitive function, as three of the four studies showed improvement. Evidence for an effect on global function was mixed with no consistent benefit. Similarly, for quality of life/ADL outcomes, all three studies that evaluated this domain were not statistically significant but these were secondary outcomes and may reflect a lack of power. Behavior/mood and caregiver burden were not tested in these trials.

The quality scores for reporting adverse events were generally low, scoring 1 or 2 out of 5. Withdrawal rates due to adverse events varied from 1 – 5% for placebo and 12 – 55% in the treatment group, with one study not reporting rates. The high withdrawal rates were in studies with sample sizes that varied from181 to 475 subjects. A single study tested for differences between groups, and found that dizziness, tremor, weight loss, asthenia, confusion, delirium, and respiratory problems were significantly different. The cluster of reported types of adverse events suggests that gastrointestinal problems (abdominal pain, diarrhea) (placebo = 1 – 9%, all doses = 13 – 28%), nausea and vomiting (placebo = 1 – 9%, all doses = 9 – 75%) and eating disorder (placebo = 2 – 6%, all doses = 5 – 16%) were most frequently reported. Dizziness (placebo = 4 – 13%, all doses = 11 – 38%) and agitation (placebo = 6 – 16%, all doses = 4 – 8%) were also reported.

Posatirelin. Four trials evaluated posatirelin in 931 subjects in a variety of mild to moderate dementia populations (AD, PDD, VaD) using a dose of 10 mg per day over 3 months duration.

Three of the four trials showed statistical significance for general cognitive function and quality of life/ADL (as measured by GBS subscales for these domains). The evidence remains inconsistent for benefit in global assessment (evaluated in only one trial) and behavior/mood (mixed results). Caregiver burden and specific cognitive function were not evaluated in any trial.

Quality scores for reporting adverse events varied from 2 to 4. Withdrawal rates due to adverse events ranged from 0 – 3% in placebo and 0 – 4% in the treatment group. None of the studies tested for significant differences between groups. All studies reported the presence of agitation, and at least three studies reported arrhythmia, nausea/vomiting, headache, rash/skin disorder, and sleep disorder. The range of frequencies of the a priori symptoms of interest are as follows: 1) nausea (placebo = 3%, all doses = 1 – 4%), 2) dizziness (placebo = not reported, all doses = 1%), 3) diarrhea (placebo = 2%, all doses = 2%), 4) agitation (placebo = 1 – 5%, all doses = 1 – 5%), and none reported eating disorder as an adverse event.

Rivastigmine. Six studies evaluated 2071 subjects and three of these studies were limited to AD patients only. Doses for rivastigmine varied from 1 to 12 mg, and treatment ranged from 14 to 26 weeks and only one study specified a community sample.

The evidence shows that general cognitive function improves with rivastigmine at a dose of 12 mg, but there is mixed results for efficacy at lower doses. Two trials also evaluated specific cognitive function but the results were not consistent within studies (between general and specific measures) and between studies for these domains. There is consistent evidence of benefit for the outcome of global assessment but the dosage at which this is significant varies highly between studies. In the domains of behavior/mood and quality of life/ADL, the findings were not statistically significant nor consistent; most of these analyses were not based on intention to treat analysis and lack of sufficient power cannot be ruled out. Caregiver burden outcomes were not evaluated by any trial.

Quality scores for reporting adverse events varied from 2 to 5. Withdrawal rates due to adverse events ranged from 4 – 11% in the placebo and 11 – 27% in the treatment group. Two trials demonstrated a dose response; however, one of these trials showed significant differences for nausea and vomiting only, and the other trial showed significant difference for all the adverse events reported. The majority of studies reported dizziness, nausea and vomiting, eating disorder/weight loss, and headache. It should be noted that one study allowed intentional prescribed anti-emetic drugs to increase the tolerance of subjects taking rivastigmine. The range of frequencies of the a priori symptoms of interest are as follows: 1) nausea (placebo = 3 – 10%, all doses = 8 – 58%), 2) dizziness (placebo = 0 – 7%, all doses = 6 – 20 %), 3) diarrhea (placebo = 2 – 9%, all doses = 7 – 17%), 4) eating disorder (placebo = 4 – 8%, all doses = 4 – 19%), and 5) agitation was not reported.

Tacrine. Six studies^{108, 109, 110, 111, 112, 113} evaluated tacrine in 994 subjects predominately with mild to moderate AD at doses of 80 to 160 mg lasting either 12 – 13 or 3 – 36 weeks in duration. Two other studies^{114, 26} involving 425 patients were non-placebo controlled studies.

A single trial¹⁰⁸ was found to show benefit for general cognitive function with a small effect and this was based on a series of related publications. The five trials showing no benefit for general cognitive function comprised small sample sizes and much shorter study duration. Overall, the evidence for benefit for general cognitive function is limited to this single trial. There is evidence for benefit in global function from two of the three trials that evaluated this domain. Changes in behavior/mood, quality of life/ADL domains, specific cognitive function, and caregiver burden were all not significant, but lack of sufficient power cannot be ruled out.

The quality scores for reporting adverse events varied from 1 to 3. The proportion of subjects withdrawing due to adverse events ranged from 0 – 12% for placebo and 0 – 55% in the treatment group. The higher rates of withdrawal were associated with higher doses. Elevated alanine transaminase (ALT) or hepatic abnormality (placebo = 4 – 13%, all doses tacrine = 7 – 67%) was reported in six studies, raising concerns for the potential for serious liver damage. None of these trials tested for differences between treatment and placebo with respect to adverse events. Five of the studies reported nausea and vomiting (placebo = 0 – 9%, all doses = 9 – 37%); gastrointestinal problems; dizziness (placebo = 0 – 16%, all doses = 4 – 14%) was also noted in several studies. Frequencies of other a priori symptoms of interest are as follows: 1) agitation (placebo = 5 – 12%, all doses = 5 – 9%), and 2) diarrhea (placebo = 0 – 13%, all doses = 4 – 18%). There is evidence for the potential for serious adverse events associated with liver function in six trials.

Velnacrine. Three studies evaluated the effects of velnacrine in 774 AD patients with a diagnosis of AD. The doses that were shown to effect significant changes were 75 mg twice daily and 225 mg daily in studies with a 15 and 24 week duration. Location of recruitment was not specified.

Statistically significant effects were observed for general cognitive function, and global assessment in the two studies with sample sizes over 300 subjects. Behavior/mood and caregiver burden showed some benefit in one trial¹¹⁶ at the highest dose only. Quality of life/ADL was tested as a secondary outcome and showed mixed findings.

Quality scores for reporting adverse events were 3 for all studies. Withdrawal rates varied from 0 – 22% for the placebo group and 5 – 33% for the treatment group. None of the studies reported a dose response. None of the studies tested for statistical differences between the placebo and treatment groups. Two studies reported aberrant hematology and hepatic abnormality;^{116, 117} for these two studies the rates of occurrence were 2 – 21% for placebo, and 32 – 40% for all doses. All studies reported diarrhea and nausea and vomiting. The range of frequencies of the a priori symptoms of interest are as follows: 1) nausea (placebo = 0 – 4%, all doses = 3 – 8%), 2) dizziness (placebo = 3%, all doses = 0 – 8%), 3) diarrhea (placebo = 3%, all doses = 2 – 33%), 4) agitation (placebo = 4%, all doses = 1 – 4%), and 5) eating disorder (placebo = 1%, all doses 2 – 4%). The potential for serious liver effects was not well specified in these trials.

Summary of Non-cholinergic Neurotransmitter/Neuropeptide Modifying Agents

Haloperidol. Five studies evaluated the effect of haloperidol relative to placebo in a total of 622 subjects with mild to moderate disease and included AD patients^{124, 125, 128} and mixed populations (MID/VaD/ PDD).^{126, 127} One trial¹²⁸ had only 15 patients, and one trial¹²⁴ lasted only three weeks. Two studies recruited subjects from institutions; one from the community and two did not specify.

Mixed results were observed for improvement in global assessment. In three of the trials there was benefit in the domain of behavior/mood which reached statistical significance. Two trials evaluated caregiver burden and found no statistically significant differences but lack of sufficient power cannot be ruled out. Few studies evaluated outcomes in quality of life/ADL. Haloperidol did not affect general cognitive function in two trials and was not evaluated in the other studies.

The quality scores for reporting adverse events varied from 1 to 5 and only three of five studies reported withdrawal rates; the proportion of subjects withdrawing due to adverse events ranged from 5% to 17% for placebo and 17 – 33% in the treatment group. One trial¹²⁴ showed a dose-response effect but the study only lasted for three weeks. Three trials tested for differences between treatment and placebo with respect to extra pyramidal symptoms (placebo = 17 – 32%, all does = 34 – 97%), and two found statistically significant differences.^{124, 125} One study¹²⁵ found significant differences between groups for balance-related problems. Although reported by only two trials, the range of frequencies of the a priori symptoms of interest are as follows: 1) nausea (placebo = 3%, all doses = not reported, and 2) dizziness (placebo = 24%, all doses = 21%). No frequencies were reported for agitation, diarrhea, or eating disorder.

Memantine. Three trials evaluated memantine in 1066 patients, primarily with VaD, with 10 or 20 mg doses lasting 12 or 28 weeks. Disease severity was moderate to severe in a single study¹³² and mild to moderate in the remaining two studies^{133, 134}. One study included patients that were institutionalized, one from the community and the third did not specify.

There is consistent evidence of benefit for general cognitive function in the two studies that evaluated this domain. The findings for global assessment are mixed. The sole trial that evaluated mixed dementia populations (including some VaD) with moderate to severe dementia found significant differences for global function, behavior/mood, and quality of life/ADL outcomes, but did not evaluate general cognitive function. It should be noted that this trial with mixed populations used half the dose of memantine for half the study duration in patients with greater disease severity, and had approximately half the sample size of the other two trials evaluated in this systematic review. Despite a lower dose, a smaller number of more severely affected patients, and a shorter duration, a statistically significant difference was found.

The quality scores for reporting adverse events varied from 3 to 4. Only two of three studies reported withdrawal rates; the proportion of subjects withdrawing due to adverse events ranged from 7% to 13% for placebo and 9 – 12% in the treatment group. A single trial tested for differences between treatment and placebo, and none of the comparisons were significant. The range of frequencies of the a priori symptoms of interest are as follows: 1) nausea (placebo = 3%, all doses = 5%), 2) dizziness (placebo = 3 – 8%, all doses = 6 – 11%), 3) diarrhea (placebo = 4%, all doses = 4%), 4) agitation (placebo = 7 – 8%, all doses = 4 – 5%), and none reported eating disorder as an adverse event.

Selegiline. Six trials^{135, 136, 249, 138, 139, 140} evaluated selegiline in 733 patients with AD, PDD, and DAT with 10 mg per day and a study duration of 60 days or 2 years. Only one study reported that patients were from institutional settings.

All but one trial that evaluated general cognition showed no statistically significant changes. A single trial found statistical improvements in specific cognitive tests (Sternberg Memory tests); this trial also showed statistically significant improvements in global assessment and behavior/mood. Only this trial, which had the highest quality score (7), showed consistently positive findings across domains tested. Three of the five trials that evaluated part or all of these domains had very small sample sizes and were likely underpowered, possibly accounting for the inconsistent findings. There is some evidence that selegiline and selegiline combined with vitamin E, increases the time to important functional decline milestones; this is based on a single study.

The quality scores for reporting adverse events varied from 0 to 3. The proportion of subjects withdrawing due to adverse events ranged from 0 – 4% for placebo and 0 – 9% in the treatment group. Only one trial¹³⁵ tested for differences between the treatment and placebo groups and showed that balance and falls were significantly different (worse) between groups (22% for the group with selegiline combined with vitamin E versus 5% in the placebo). When adjusted for multiple comparisons, these were no longer significant. The range of frequencies of the a priori symptoms of interest are as follows: 1) nausea (placebo = 2%, all doses = 0%), 2) dizziness (placebo = 2 – 20%, all doses = 0 – 30%), and 3) agitation (placebo = 4 – 16%, all doses = 4 – 23%); no trial reported diarrhea or eating disorder as an adverse event.

Summary of Other Pharmacological Agents

Cerebrolysin. Six studies evaluated the effect of cerebrolysin in a total of 819 subjects. All but one of the trials¹⁷¹ included only AD patients with mild to moderate disease. All of the studies used the same dose of cerebrolysin, 30 ml per day for 5 days per week for 4 to 24 weeks. Location of recruitment was not specified.

Cerebrolysin showed a statistically significant effect on cognition in four out of five studies. Although, a pooled estimate for the ADAS-cog was calculated, the model was positive for heterogeneity and the overall estimate was not significant. The results for specific cognitive tests for the three trials that evaluated this domain were inconsistent. Global assessment measures showed a significant effect in five of the trials. This model was also positive for heterogeneity but significant for an overall effect. Two out of three studies showed an effect for behavior/mood, and none of the six studies showed an effect on quality of life/ADL. No study measured caregiver burden.

Two of the six trials scored 5 out of 5 on our quality scale for rating adverse events, yet they did not report any adverse events. Two studies scored 4, and the other two trials scored 3 and 2. All the studies with scores equals to 4 or less tested for statistical differences in adverse events between placebo and treatment groups. Withdrawals due to adverse events were not reported in one study¹⁷⁰ and were 1% in two studies^{173, 168} and none withdrew in three studies.^{169, 172, 173} One study reported significant differences between treatment and control group¹⁷³ for weight change, anxiety, and headache. The range of frequencies of the a priori symptoms of interest are as follows: 1) nausea (placebo = 10 – 24%, all doses = 3 – 21%), 2) dizziness (placebo = 0 – 12, all doses = 1 – 8%), and 3) agitation (placebo = 1%, all doses = 0%), and none reported diarrhea or eating disorder as an adverse event.

Estrogen. Five studies evaluated estrogens for dementia in 247 patients with primarily mild to moderate AD, with the exception of one study¹⁷⁸ that included moderate to severe dementia patients who were all institutionalized. One of the studies with AD patients provided 0.10 mg per day¹⁷⁴ by skin patch for 8 weeks and the others used 1.25 mg per day for 12 to 52 weeks.¹⁷⁷ The study including severe subjects used 2.5 mg per day for 4 weeks.¹⁷⁸

Three trials evaluated general cognitive function and all showed non-significant findings; two of these trials lacked sufficient power for the ADAS-cog. Similarly, two trials evaluated specific cognitive function but results were mixed. Most of the outcomes evaluated in the domains of global assessment, behavior/mood, and quality of life/ADL were secondary outcomes and none showed significance; lack of power could be a factor in these trials.

One of the five trials scored 5 out of 5 on our quality scale for rating adverse events, and surprisingly, this same trial did not report any adverse event. Withdrawal rates due to adverse events ranged from 0 – 5% for placebo and 0 –14% for the treatment group. The most frequently reported adverse event was vaginal bleeding and a single trial reported a significant difference between placebo and treatment group for this symptom. It was not clear from the descriptions provided in the study if they had ascertained whether vaginal bleeding was present prior to the trial commencement. Nausea was the single a priori symptom of interest that was reported and by a single trial; frequencies varied from 0% for the placebo group and 4% for the treatment group.

Ginkgo biloba. Three trials evaluated Ginkgo biloba, 120 to 240 mg per day for 3 to 12 months, in a total of 563 subjects with mixed dementias of mild to moderate severity.

The largest trial¹⁷⁹ had the longest treatment interval but the lowest daily dosage and reported a significant effect for general cognitive function but had mixed findings for global assessment. A second large trial¹⁸¹ found positive changes for neuropsychological tests, global assessment, and behavior outcomes with double the dosage of the previously described trial and half the treatment interval. In this same RCT, clinical efficacy was assessed by using a responder analysis, with therapy response being defined as response in at least two of the three variables: CGI (global function), SKT (special cognitive function), and NAB (ADL). A single trial evaluated behavior/mood and was not significant. No trial evaluated caregiver burden or quality of life/ADL.

All three trials scored 3 or greater on the quality scale for rating adverse events. Two studies had no withdrawals due to adverse events, and one trial had a withdrawal rate of 6% for both placebo and treatment groups. Two studies did not report any adverse event. One study reported a statistically significant difference between the treatment and the placebo group for skin disorders. The same study reported gastrointestinal and headache adverse effects, but did not test for statistical differences between the placebo and the treatment group. None of the trials reported the presence of the a priori symptoms of interest.

Idebenone. Four studies^{185, 183, 182, 184} evaluated the drug idebenone in 1153 subjects of mixed dementia populations of mild to moderate severity; one of these trials²⁶ evaluated idebenone relative to tacrine. Doses varied from 30 mg per day to 360 mg per day, and the treatment interval ranged from 90 days to 60 weeks.

There was evidence of benefit for general cognitive function and global assessment. Several studies evaluated behavior/mood and quality of life/ADL and these outcomes were found to be significantly different. None of the trials evaluated caregiver burden.

Quality scores for reporting adverse events varied from 1 to 5. Rates of withdrawal due to adverse events varied from 0 – 5% for the placebo group and 0 – 5% in the treatment group; a single trial¹⁸³ did not report withdrawal rates. Two trials^{183, 185} tested for statistical differences between groups and found no differences. Although no clear pattern emerges, three studies identified at least one balance-related adverse event most consistently reported across studies. The range of frequencies of the a priori symptoms of interest are as follows: 1) nausea (placebo = 2%, all doses = 2 – 11%), 2) dizziness (placebo = not reported, all doses = 2%), and 3) not reported for diarrhea, agitation, or eating disorder as an adverse event.

Oxiracetam. Five studies^{186, 187, 188, 189, 190} evaluated oxiracetam in 554 subjects with different dementia syndromes of mild to moderate severity. All analyses were observed cases and not ITT. All studies used 1600 mg daily, with one exception¹⁸⁹ where the dose ranged between 1600 – 2400 mg per day. The treatment interval ranged from 90 days to 26 weeks.

All outcomes shown to be positive for this drug were based on observed case evaluation. The two trials that evaluate general cognitive function showed benefit. The findings for specific cognitive function were mixed. A single trial evaluated global assessment and showed statistically significant change. Behavior/mood, and quality of life/ADL outcomes showed mixed results. No study evaluated caregiver burden.

The quality scores for reporting adverse events varied from 2 to 5. The proportion of withdrawals due to adverse events varied form 0 – 9% for the placebo group and 0 – 6% for the treatment group. No clear pattern for adverse events is evident, but three of the five studies reported gastrointestinal related problems, primarily associated with abdominal pain. Although, only single trials evaluated the range of frequencies of the a priori symptoms of interest are as follows: 1) dizziness (placebo = not reported, all doses = 11%), and 2) agitation (placebo = 1%, all doses = not reported); no trial reported nausea, eating disorder, or diarrhea as an adverse event.

Pentoxifylline. Three placebo-controlled studies^{193, 192, 191} evaluated pentoxifylline and one study¹⁹⁴ compared pentoxifylline to sulodexide, with a total of 482 subjects with predominately MID. The total dose administered in all studies was 1200 mg per day but varied from 400 mg three times per day to 1200 mg once per day. The treatment intervals ranged from 12 to 36 weeks.

All three placebo trials showed non-significant findings for any primary outcome evaluated on all subjects in the study. It should be noted that two of these trials had very small sample sizes (n = 38, n =28) that were evaluated in the OC analyses; this suggests that the trials lacked sufficient power to evaluate multiple outcomes. The remaining trial had a large sample size (n = 289) and employed an ITT analysis; all primary outcomes evaluated were not significant.

The quality scores for reporting adverse events were generally low, varying from 1 to 3. Withdrawal rates due to adverse events varied from 0 – 25% in the placebo group and 0 – 22% in the treatment group. The two studies that reported adverse events indicated the presence of gastrointestinal disturbances, including abdominal pain or nausea and vomiting (placebo = 7% and all doses = 14%). None of the trials reported dizziness, agitation, eating disorder or diarrhea.

Propentofylline. Four trials^{197, 196, 250, 198} using propentofylline in 510 patients with AD and VaD were included. A dose of 900 mg per day was consistent across all studies, and the treatment duration ranged from 3 to 12 months.

The two studies with small sample sizes (n = 30) showed no significant results for any outcome evaluated but lack of power cannot be ruled out. There were two trials that found benefit in general cognitive function based on the MMSE. The results for specific cognitive function as measured by the DSST were mixed, as were those for global assessment. Behavior/ mood outcomes were evaluated by a single trial¹⁹⁵ and shown to be significant; this same trial evaluated quality of life/ADL and showed no significant difference. No trial evaluated caregiver burden.

The quality scores for reporting adverse events varied from 1 to 4. The percentage of withdrawals varied from 0 – 13% for the placebo group and 0 – 12% for the treatment group. None of the trials tested for differences between groups. Three of the trials^{195, 197, 198} reported gastrointestinal events that included abdominal pain, constipation, and nausea and vomiting (placebo = 2%, all doses = 7%). Dizziness (placebo = 3 – 5%, all doses = 1 – 6%) was the only other a priori symptom of interest.

Methodological Issues and Limitations in Assessing Efficacy of Dementia Agents

Definition of clinically significant or meaningful difference. The stance undertaken in this review has been cautious with regards to interpreting “clinically significant” differences within and across studies. This systematic review has highlighted some of the concerns expressed in the literature on pharmacological efficacy research in dementia. Ultimately, clinical significance is a complex issue, and its definition can vary across individuals and groups of individuals. Wherever possible, attempts were made to identify the magnitude of differences in the studies and the limitations of the data from some of these primary studies.

In drug development programs, an ordered series of trials are undertaken: dose tolerance (phase I), dose finding (phase II), dose efficacy (Phase III), and post-marketing (phase IV). However, due to the pressures on pharmaceutical companies to develop drugs quickly and cost-efficiently, a drug may move into the next phase of development before evidence of the previous phase is known.²⁵¹ Even when phase III trials are carried out in an adequate manner, the interpretation of the efficacy results is hampered by multiple p-values, disagreement over the need for multiplicity corrections, and the potential for conflicting evidence from trials of different sizes.²⁵¹ Some of these difficulties can be minimized by measuring a single primary efficacy variable at one point in time and using a p-value of less than 0.025 (one-tailed, as the aim is for the statistical test to determine if the drug performs better than the placebo or low dose).²⁵² This presumes that good dose-response data exist, identifying a single dose level as the best candidate for further evaluation. Lastly, interpreting differences on the basis of statistical significance has long been recognized as problematic. Clinically meaningful change reflects a different level of “significance” and often requires consensus among experts within the field for these criteria.

Issues of diagnosis and severity. Three methodological issues related to population classifications have limited the inferences that can be garnered from this systematic review. The first issue concerns the classification models used for diagnosing dementia; they are not interchangeable among the various types of dementia and the “pre-clinical” forms of slight cognitive impairment. Moreover, there are still concerns about the accuracy of these criteria. For example, in the American Academy of Neurology's (AAN) recent evidence-based review of dementia case definitions, none met the AAN's highest evidence standard. A clinical diagnosis of AD is only 28% specific after age 79 years. Similarly, no dementia screening measure is accurate enough to be recommended by the American Society of Internal Medicine.²⁵³ The AAN specifically faulted the emphasis on memory function in dementia case definitions.²⁵⁴ Yet tests like the ADAS-cog emphasize memory loss at the expense of other cognitive domains, especially executive control function,³¹ and many anti-dementia treatment strategies target neurotransmitters and structures (like acetylcholine and the hippocampus), which mediate memory test performance.

A second consideration in defining populations of dementia patients concerns determination of severity level. The MMSE, although frequently used, may not best capture severity. Many studies were observed to define the severity (mild, moderate, severe) of dementia populations based on the MMSE. For example, a range from 10 to 26 has been used to define a mild to moderate severity level.⁵⁰ Given that the maximum and minimum instrument scores are 0 and 30, this suggests that the extreme ends of the spectrum, particularly the “severe” end (i.e. <10), represent a very narrow proportion of patients. These two broad categories (mild to moderate and severe) may not actually reflect the cognitive and functional differences in a clinically meaningful manner. The MMSE does not address issues of executive control function (as required by the DSM-IV dementia case definition), which is known to be a good predictor of functional status. From a research perspective, a better classification reflecting disease severity may be an important factor for stratification and determining the efficacy of pharmacological interventions.

Outcome issues. The studies evaluated in our review used 181 different outcomes across seven domains. This raises the issue of which of these outcomes are considered by clinicians to be most “clinically relevant”. Let us assume that the most clinically relevant outcomes for all the drug interventions for dementia are the ADAS-cog and the MMSE because they are very commonly reported in studies.

In this dementia review, numerous studies did not measure outcomes evaluating cognition, as the intended effect of the drug was not always in the domain of cognition (e.g. neuroleptics for behavior control). Moreover, a large number of the studies that used important clinical cognition outcomes, such as the MMSE, did so only to establish baseline severity, or they used it as a secondary outcome. This presents us with some difficulty in the consistency of reporting on this limited set of “clinically relevant” outcomes. There is also the issue of which domain (i.e. cognitive function versus ADL versus behavior) is the most clinically relevant. The FDA guidelines suggest cognition and global assessment; the EMEA guidelines suggest the addition of an ADL or quality of life/ADL measure as being most clinically relevant. Thus, some consensus work needs to be done among experts in the field to determine the most clinically relevant outcomes and domains. For example, the choice of most clinically relevant outcome may depend upon type and stage of dementia (e.g. for mild AD, neuropsychological outcomes may be are the most important domain while for severe AD, behavior may be the most relevant outcome), which may challenge the achievement of consensus.

To our knowledge, no specific set of outcomes that define “clinical relevance” applies to all the drug interventions we evaluated. The FDA has recommended that “dual efficacy” of dementia drug interventions be established by significant change in both a psychological measure and a global change measure. The outcomes measuring these attributes within these two domains were not specified. However, there was a general trend for using the outcomes ADAS-cog and CIBIC+ to capture these two attributes when evaluating drugs for AD populations.

Ideally, all outcomes should have demonstrated acceptable psychometric properties, such as reliability, validity (construct), and responsiveness. We did not a priori evaluate the properties of outcomes reported in the eligible studies. In some cases, these outcomes were developed in non-English languages but the original study was reported in English. In considering the psychometric properties of some of the outcome instruments used, the attribute of responsiveness is critical, and some have suggested that this has not been adequately evaluated in many outcome measures.^{33, 30, 255}

We might envision a clinically relevant pharmacological treatment as one that has made a real difference, where the change is both relevant and important to the patient or to clinicians. This fundamentally shows the difference between clinically significant (relevant and important) versus statistically significant (associated with probabilities), where the latter determines that the results are not due to chance. Moreover, a clinically important change will vary depending on whether importance is defined from the patient or clinician perspective.

Five different levels of responsiveness (ability to detect change) of outcome measures have been defined:²⁵⁶ 1) Minimal change potentially detectable (essentially an attribute of the scoring method of the outcome), 2) Minimal change actually detectable beyond measurement error of the instrument (also defined as Minimum Detectable Change (MDI) or Reliability Change Index (RCI), which includes the Standard Error of the Measurement (SEM)), 3) Observed change (often reported as the standardized response mean (SRM) or effect size (ES). 4) Observed change in those estimated to have improved; the key to understanding change in this instance is that an external standard is used to determine whom has improved (often reported as comparison between groups that have improved versus those who have not; the improved group can be defined by either patient and/or clinician or a combination), and 5) Observed change in those estimated to have important improvement (often reported as the minimal clinically important difference and can be determined by the patient or clinician, or a combination of both).

Consider the ADAS-cog and the CIBIC+: The minimal change detectable is 1/70 = 0.0143 for the ADAS-cog and 1/7 = 0.143 for the CIBIC+, suggesting that the ADAS-cog can detect smaller increments of change relative to the CIBIC+. Thus different instruments have differing sensitivities to detecting change. There is scant literature on the responsiveness of outcome measures as defined in number 4 above, observed change in those that have improved, or as in number 5 above, observed change in those estimated to have important improvement. Thus, we have identified a significant gap in the literature with regard to estimating clinically important changes. Much greater consideration of issues of responsiveness should be given in future research in efficacy trials of pharmacological agents. Greater understanding of clinically important change suggests that some of our current judgments of efficacy are limited as these important differences need to be established.

Analysis issues. The inability to estimate the power of a study to detect a difference presented significant limitations in interpreting those studies that showed no significant differences. Similarly, the lack of sufficient data for estimating effect size limited the ability to show the magnitude of the change. It is recommended that future trials evaluating the efficacy of pharmacological agents adhere to the CONSORT guidelines in order to provide sufficient data to estimate power and effect size for all relevant outcomes.

Although the difficulty of maintaining adherence to long-term drug interventions among dementia patients is acknowledged, the ITT analysis should continue to be the analysis of choice in trials. Ideally, both ITT and OC analyses should be presented. If both suggested the same conclusion, confidence in the study results would be increased.

Problems with funding/ sponsorship exclusively from drug companies. The sponsorship of studies by for-profit organizations has led to bias towards the publishing of positive results.²⁵⁷ These findings suggest that there are powerful disincentives for pharmaceutical companies to publish negative trials. This is contrary to what academic-based, non-industry funded trials show, where the publication of negative trials are more likely.

A recent evaluation of FDA databases for antidepressant drugs²⁵⁸ in the US, suggested that less than half of antidepressant trials were negative, which does not correspond to the published literature. In this systematic review, no attempts were made to contact industry for unpublished trials, which introduces the possibility of a bias associated with not reporting negative trials. Additionally, we did not contact authors who did not specify funding sources for their studies. Future research on the efficacy of pharmacological agents to treat dementia should indicate all sources of funding and who undertook the study analyses.

Adverse events. In this systematic review, the type and frequency of adverse events associated with the use of a drug intervention were scrutinized and reported to a greater extent than previous reviews of anti-dementia drugs. Attempts were made to weigh the potential for harm against the benefits when determining the efficacy of pharmacological interventions. Empirical evidence across diverse medical fields indicates that reporting of safety information (including milder adverse events) receives much less attention than the positive efficacy outcomes.³⁵ Thus, it was recognized that an evaluation of the benefits of anti-dementia pharmacological agents alone may present a biased view of the efficacy of the intervention.

The ability to capture and evaluate adverse events proved to be difficult for several reasons. For example, although metrifonate had good evidence of positive effects on cognitive function, it was banned from use due to the risk of respiratory paralysis. The description of serious adverse events in the trials we evaluated did not capture this type of event, nor did different studies identify “serious events” in a consistent manner. This points to several fundamental limitations. The first of these relates to the limitation associated with the RCT design itself, which is less likely than the longitudinal cohort study designs to capture serious adverse events that are rare. Secondly, many trials were of relatively short duration and captured “idealized” dementia populations. Many of these trials were from pre-marketing studies contracted by pharmaceutical companies in carefully controlled research settings. Dementia patients seen in practice may have more complex medical illnesses and are at greater risk for potential side effects. In addition, drugs used in “polypharmacy” have even greater potential for pharmacological interactions. Furthermore, practitioners may prescribe these pharmacological agents for wider indications than originally intended, or may not refrain from withholding the drug from certain high-risk subgroups, leading to increased risk of adverse events. Thus, published rates of adverse events in well-controlled trials may underestimate true rates seen in practice.

Thirdly, by their nature, some adverse events are not easily anticipated, and therefore are not screened for in some trials. Adverse events may be hard to predict or anticipate but can be captured only if a trial protocol was designed to measure these events. This problem is compounded by the lack of consistency in what constitutes “serious” events or how the severity of the typical events is rated. A limited number of standardized instruments exist to capture these events reliably, but the overwhelming majority of studies in this systematic review did not use these instruments. Furthermore, capturing information from individuals with cognitive decline can create problems; the validity of the self-report instrument, even if completed by the caregiver, can be problematic. More research on the reliable collection of adverse events in dementia populations (with compromised cognition) may be required.

A fourth consideration concerns the issue of off-label use of pharmacological agents. Given that only four drugs are currently approved by the FDA for the treatment of dementia, the other 97 interventions evaluated in this review are classified as “off label use” but many are not approved by the FDA and not, therefore, available. For some of these off-label medications the potential mechanism of action on the disease process has not been fully established (if even considered), yet they have been applied to dementia populations. This off-label use of pharmacological agents may present further difficulties in evaluating adverse events.

Summary of Systematic Review Results

Few studies evaluated delay of onset or delay in disease progression. A definite gap for evaluating disease onset (as defined by the selection of populations at risk such as MCI populations) has been identified in this review.

Conversely, the need for good evaluation of disease progression in trials was also identified. In general, few studies evaluated subjects in more severe state of the disease. This suggests that a bias exists towards evaluating mild to moderate disease in the trials eligible in this systematic review. This in turn reflects the underlying assumption that the less severe groups are most likely to benefit from drug trials. Since so few studies have evaluated the more severe groups, this assumption may require some empirical justification in future research. Those studies that evaluated severe patients showed some potential for benefit. Future research in this area may require some consensus regarding the classification of severity levels.

Three studies evaluating cerebrolysin,¹⁶⁸ selegiline and vitamin E,¹³⁵ and donepezil⁶¹ have shown significant effects in delaying disease progress in mild to moderate^{61, 168} and moderately severe disease in patients with AD. This delay in progress was expressed in terms of delay in days to primary event^{135, 61} or statistical differences between placebo at a specified time interval.¹⁶⁸ Although these two trials coincidentally evaluated dementia patients over the longest time interval, it did not withdraw the drug at the end of the study. Theoretically, conclusive evidence of disease delay would be demonstrated if the treatment groups did not return to the level of the placebo. Thus, distinguishing between symptomatic and disease modifying effects is not possible unless the drug is withdrawn and the treatment group(s) are observed for these changes.

When studies attempted to evaluate disease progression, long-term (1 year or greater) trials continued in an “open-label fashion”, where blinding was no longer maintained. This limits the confidence that bias did not affect the subsequent changes in the outcomes. It was observed that increasing levels of dropout (for a variety of reasons) also plagued these open-label phases of evaluation. From a practical perspective, maintaining adherence in longer-term trials in dementia patients are challenging,¹⁹ particularly for those in the placebo arm or for those interventions that have a high proportion of adverse events.

A number of trial designs have been proposed to capture delay in disease progression versus symptomatic treatment. Some of these trial designs include withdrawal of treatment, active-extension, randomized withdrawal, randomized start, and staggered start designs.^{235, 236, 19, 21} One important aspect of these designs is the selection of an adequate washout period or an adequate follow-up period. In addition, longer evaluation with survival analyses may be a good strategy to evaluate delay in disease progress for some drugs. One advantage of this design is the selection of clinically relevant milestones (functional changes over time), which was utilized in two studies ^{63, 136}; the selection of such events may merit greater consideration in future trials evaluating delay. A more critical analysis of the staggered/start/stagger withdrawal design in comparison to the survival analysis would be helpful. Also, one could provide a more extensive analysis of the data on propentofylline and vitamin E,¹³⁶ which represent the most extensive efforts to use the stagger/start/stagger withdrawal and survivor analysis approaches, respectively. Future research seeking to establish efficacy should clearly specify if symptomatic treatment or delay in progression is the therapeutic aim. This is important for determining specifically if efficacy is considered with respect to these two aims. Accordingly, a design that can establish this aim should be selected.

Methodological Issues

Figure 31. Delay of symptomatic treatment effects

   Figure 31. Delay of symptomatic treatment effects

Figure 31

Figure 31

Figure 32. Dealy in disease progression treatment effects (more...)

   Figure 32. Dealy in disease progression treatment effects

Figure 32

Figure 32

31

32

Figure 31

Summary of Systematic Review Results

What may be most relevant to clinicians are head to head comparisons of the cholinergic modifying neurotransmitter pharmacological agents, particularly those currently approved for the treatment of dementia (tacrine, rivastigmine, galantamine, donepezil) in the United States. The evidence for each of these drugs has been extensively detailed, and the relative merits and handicaps of each were outlined in chapter 3. Relative effectiveness as demonstrated by effect sizes for the ADAS-cog and the CIBIC were also shown in chapter 3. Although, the psychometric properties of these two outcomes are well accepted, comparison across the populations in these pooled estimates may not lend themselves to direct comparison across these four different specific drugs. Thus, inferences about the relative effectiveness of these four medications specific for the treatment of dementia should be made cautiously as head to head comparisons were not undertaken.

Relative efficacy must be evaluated in direct comparison trials

From a methodological perspective, addressing the question of being “more effective” requires head to head comparisons of pharmacological interventions.

An evaluation of the trials that undertook direct head to head comparison of two distinct pharmacological agents was limited because only seven trials were identified. Although, these trials may have shown some relative benefit of one drug versus another, the clinical relevance of these particular agents is limited as none of the drugs currently approved by the FDA specifically for the treatment of dementia is represented in these eligible studies. Moreover, these studies are essentially limited to single trials and are not sufficiently strong to base recommendations on the relative effectiveness of drugs. Head to head comparison studies are beginning to appear in abstract form only and a significant gap in the literature has been identified.

Summary of Systematic Review Results

In general, very few trials examined the efficacy of dementia drugs across different populations or population characteristics. From the 13 studies that reported stratified analyses, eight different variables were identified, which included age, gender, APOE genotype, disease type, disease severity (as determined by MMSE/ ADAS-cog threshold levels), treatment center, care dependence, and presence of depression. Additionally, three trials were identified that evaluated efficacy in 1) patients with Down's syndrome and dementia, 2) different ethnicities as a function of treatment center in a multicenter trial, and 3) depressed patients. Given the relatively small number of trials evaluating these variables within different populations and different pharmacological interventions, the findings of this review are limited with respect to these patient variables. These reflect merely what has been reported in the literature rather than variables of importance with respect to efficacy of pharmacological therapies. A significant gap in the literature has been identified.

Representativeness of populations in the drug trials

The study population characteristics were detailed for the trials evaluated. A recent study,²² suggests that many “real world” dementia patients in Ontario would not have met the eligibility criteria for participation in several of the cholinesterase inhibitor studies. This study highlights an important limitation of the pharmacological literature in that dementia patients recruited are not representative of the general dementia population. Additionally, clinicians and researchers should note that when a when a drug is approved for use, it is for a specific indication and a specific patient population. Evidence for one type of patient population may not necessarily be applied to another population. This is critical information to have when establishing clinical practice guidelines.

Summary of Systematic Review Results

A total of 20 pharmacological interventions in 29 studies^{211, 220, 238, 171, 200, 199, 146, 68, 181, 184, 133, 134, 132, 161, 89, 91, 93, 247, 187, 191, 192, 194, 193, 100, 98, 196, 195, 245, 217} were applied specifically to VaD classified dementias. The majority of these pharmacological interventions (n = 14) were represented by single trials, these interventions included ateroid, buflomedil, cerebrolysin, sulphomucopolysaccharides (CDP choline), citalopram, donepezil, Ginkgo biloba, idebenone, minaprine, nimodipine, oxiracetam, 5-THF (trazodone), vincamine, and xantinolnicotinate. Six interventions had more than a single trial, and these included Choto-san (n = 2), memantine (n = 3), nicergoline(n = 2), pentoxifylline (n = 4), posatirelin (n = 2), and propentofylline (n = 2). In general, when the drug interventions were shown to be effective, it was in the domains of cognitive function (both general and specific) and global assessment. Other domains were less frequently evaluated. Several trials attempted to test for differences between VaD groups and other dementia types.

Diagnosis Classification of VaD

Erkinjuntti et al (1997)¹ compared six commonly used classification schemes (DSM-III, DSM-III-R, DSM-IV, ICD-9, ICD-10 and the CAMDEX) and demonstrated that the prevalence of dementia can differ by a factor of 10 depending on the diagnostic criteria used. Two other studies have demonstrated that the prevalence of VaD varies with the classification system; therefore these criteria for diagnosis are not interchangeable.^{10, 11}

There is controversy about the validity of the clinical classification of VaD, as autopsy confirmation often does not substantiate the clinical diagnosis.^{12, 13} The majority of dementias were actually AD with co-existing VaD and PDD lesions.¹⁴ In contrast, the clinical accuracy of AD diagnosis is relatively high.⁷ Future research in vascular dementia should attempt to better distinguish this subgroup.