NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Ross SD, Estok R, Chopra S, et al. Management of Newly Diagnosed Patients with Epilepsy: A Systematic Review of the Literature. Rockville (MD): Agency for Healthcare Research and Quality (US); 2001 Sep. (Evidence Reports/Technology Assessments, No. 39.)

  • This publication is provided for historical reference only and the information may be out of date.

This publication is provided for historical reference only and the information may be out of date.

Cover of Management of Newly Diagnosed Patients with Epilepsy

Management of Newly Diagnosed Patients with Epilepsy: A Systematic Review of the Literature.

Show details


In general, MetaWorks investigators used systematic review methods derived from the evolving science of review research (Mulrow and Oxman, 1997; Mulrow, Cook, and Davidoff, 1997; Sacks, Berrier, Reitman, et al., 1987). These methods were generally applied according to standard operating procedures at MetaWorks and displayed in Figure 1.

Figure 1. MetaWorks systematic review process diagram.


Figure 1. MetaWorks systematic review process diagram.

A "causal pathway" was developed to focus the systematic review (Figures 2 and 3). This pathway depicts an assumed hierarchical structure of health care services for diagnosis, treatment, and monitoring of patients with a first diagnosis of epilepsy. It shows assumptions about the traffic of patients through the service hierarchy. Evidence in support of both the structure and flow of patients as depicted in the causal pathway was sought in the systematic review.

Figure 2. Causal Pathway A.


Figure 2. Causal Pathway A.

Figure 3. Causal Pathway B.


Figure 3. Causal Pathway B.

The review followed a prospective protocol that was developed a priori and shared with the nominating partner (CDC), a technical experts panel (TEP) (with representation from consumer groups and professional groups: neurology, epileptology, primary care, and nursing); and the Task Order Officer at the Agency for Healthcare Research and Quality (AHRQ). The protocol outlined the methods to be used for the literature search, study eligibility criteria, data elements for extraction, and methodologic strategies to minimize bias and maximize precision during the process of data extraction and synthesis.

Literature Search

The published literature was searched from 1980 to the present. The original search cutoff date was April 15, 1999, and the final search cutoff date was December 1, 1999. The retrieval cut-off date was January 14, 2000. The search was originally run in June 1999 back to 1990 for (1) explode epilepsy and all treatment and human, and (2) explode epilepsy and diagnosis and human. Since the project was conducted in two distinct phases, in the second phase the search was updated for 1999 in MEDLINE using the following sequence: (1) explode epilepsy, (2) human, (3) 1999, (4) diagnostic use (Medical Subject Heading, MeSH), (5) diagnosis, (6) therapy (MeSH), (7) therapeutic use, and (8) 4 or 5 or 6 or 7. Review articles for the manual search were identified in MEDLINE by searching back to 1996 by (1) explode epilepsy, (2) human, (3) diagnosis (MeSH), and (4) review (TW). Bibliographies of these reviews were perused for other likely citations. In addition, the 1999 Current Contents CD-ROM was searched as: (1) epilepsy (KW), (2) diagnosis (KW), (3) 1 and 2, (4) epilepsy (KW), (5) treatment (KW), (6) 4 and 5, (7) epilepsy (KW), (8) monitoring (KW), and (9) 7 and 8. The Cochrane Library of Systematic Reviews was searched by first checking the Database of Systematic Reviews for epilepsy (KW) and, second, checking the Database of Abstracts of Reviews of Effectiveness for epilepsy (KW). Lastly, Internet searches were performed on December 2, 1999, by checking the following Web sites for studies, guidelines, authors, and special interest groups: Dr.,,,,, and and Epilepsy Foundation ( All citations and abstracts were printed and screened at MetaWorks for any mention of any exclusionary criteria below (Level I screening).

Exclusion Criteria

Exclusion criteria comprised the following:

  • Reviews and meta-analyses, letters, and case reports.
  • Non-English language studies.
  • Animal studies.
  • Pharmacodynamic/pharmacokinetic studies.

Studies were rejected for the following reasons: (a) not newly diagnosed patients; (b) outcomes not extractable for newly diagnosed patients; (c) reviews; (d) case studies; (e) other languages; (f) studies of patients with secondary seizures, i.e., seizures not idiopathic or cryptogenic. These included studies selecting for patients with seizures secondary to tumors, trauma, infection, stroke, chronic medical or neurologic disease, psychogenic or pseudoseizures, and febrile convulsions. It should be noted that in some studies, information about secondary causes of seizures was not supplied or was incomplete, so we could not be sure that all included studies consisted completely of patients with primary epilepsy, but the intent was to develop a database as consistent as possible in this regard, given the vagaries of reporting in the literature.

All studies passing Level I screening were retrieved for a second screening (Level II) applying the following inclusion criteria.

Inclusion Criteria

Level II inclusion criteria were as follows:

  • Study designs: observational (prospective, retrospective, and cross sectional), or interventional (randomized controlled trials [RCTs], nonrandomized controlled trials [nRCTs], and uncontrolled case series [UCS]).
  • At least 10 patients, adults or children, with a first seizure, a first presentation with epilepsy or seizures, or a first diagnosis of epilepsy.
  • Studies addressing any of the following diagnostic interventions: history; neuropsychological assessment; physical examination; imaging with CT, MRI, or PET scans; EEG (standard, video, invasive, ambulatory); lumbar puncture; or blood test.
  • Studies addressing any of the following monitoring interventions: EEG (standard, video, invasive, or ambulatory), laboratory tests (hematology/biochemistry) and drug assays (standard or sophisticated), and clinical/pharmacologic expertise.
  • Studies addressing any treatment intervention: clinical/pharmacologic expertise, counseling/psychosocial, surgery (resective or vagus), physical therapy/occupational therapy (PT/OT), speech language, investigational study, and education.
  • Studies reported in English only.

The most difficult aspect of screening papers at both levels was establishing the presence of the population of interest, i.e., newly diagnosed patients. This term was defined to include patients presenting with a first seizure, patients presenting for a first diagnosis, and patients with a new diagnosis just prior to study entry. It did not include studies in patients with a long-time, established diagnosis of epilepsy. Studies of patients with a misdiagnosis, or uncertain diagnosis, were eligible as long as the patients were diagnosed with recent onset of seizures. Making this distinction in retrospective observational studies was the most difficult. Such studies could be included as long as information about patients at the time they were newly diagnosed (i.e., time zero) was included in the observation window and extractable for that group.

The second most difficult challenge in screening studies for inclusion was establishing whether data were extractable for the population of interest. In many cases where a series of patients attending a clinic, for instance, were studied, those with a new diagnosis were not distinguished from those with established epilepsy in the reporting of interventions and/or outcomes. This difficulty was most pronounced in study designs that were observational as opposed to interventional.

Since study selection is a critical component of any systematic review, the Level II screen required the agreement of two reviewers for each rejected paper. In cases of disagreement, a third reviewer adjudicated. The list of eligible studies was also subsequently shared with the project TEP for review and comment prior to locking the database.

Rating the Evidence

All eligible studies were rated for internal and external validity at the time of data extraction. A customized method was used, combining two established methods: (1) the Jadad method (Jadad, Moore, Carroll, et al., 1996) which can only be used for RCTs, and (2) the Level of Evidence (LoE) method (Cook, Guyatt, Laupacis, et al., 1992) which can be applied to all study designs. Combining the two scoring instruments permits the assessment of all studies with a single unified scoring system. These scoring methods assess features of study design, execution, and reporting. An Evidence Score was computed for each study by dividing the Jadad score by the Level of Evidence score. For studies with no Jadad score, i.e., all nRCT studies, a default score of 1 was assigned and then divided by the score. Preliminary reports of RCTs were not scored. Possible scores therefore ranged from 0.3 to 5.0 (Jadad score 1 to 5 divided by LoE 1 to 3), with higher scores suggesting higher validity of evidence. Evidence scores may be used as categorical or continuous variables in subsequent sensitivity analyses (e.g., multivariate regression analyses) or by exploring the impact of outliers on the results. They are also important to consider in interpretations of the reliability and significance of the evidence available in support of any conclusions. Evidence scores are, however, arbitrary and should not be overinterpreted. For instance, a well-done observational study may make a more valid contribution to evidence than a poorly done RCT, although their scores will not necessarily reflect this.

Data Extraction and Database Development

Key data from each eligible study were extracted by one researcher recording data from original reports onto a unique data extraction form (DEF) and agreed to by a second researcher checking all DEF fields (both filled and blank) against the original report. Differences were resolved prior to data entry. DEFs were designed in advance and pilot tested on a small sample of eligible studies. The pilot test allowed for necessary edits to the DEF to be made prior to broad implementation on all studies.

Key data elements sought for extraction from each study included characteristics of studies, patients, and interventions. Only clearly reported aggregate results were extracted from studies. Results that were only reported for individual patients and results that would require extrapolations from graphs or derivations from figures or tables were not captured.

Decision rules for extractors were developed during the pilot testing of the DEF and subsequently as questions arose during the data extraction process. The main rules are reviewed here to assist in interpretation of the resulting database. With regard to study level data, it should be noted that the total number of patients is distinct from the total number of healthy controls. If studies had a control group that was not randomized, either concurrent or historical, the design was indicated as an nRCT. All study durations were converted to months. Studies with a one-test, one-time intervention were listed as cross-sectional (XS). If a patient population was not clearly stated as all adult, or all children, the default selection was both. Age 16 was chosen as the age cutoff point above which patients were judged to be adult. Quality-of-life (QoL) results were indicated as present only if a validated instrument or a reproducible scale such as a visual analog scale (VAS) was used. Costs results were indicated as present if any results were reported in currency terms. We sought information as to the physical location of interventions (e.g., home or clinic), particularly if different from the location of the study investigators (typically academic hospital settings). It was evident early on that the study investigators' level or field of training was rarely identifiable, so we decided to show an academic affiliation if the study affiliations included a medical school or university setting. Investigators affiliated with a department of neurology were labeled as neurologists. Most studies showed investigators with multiple affiliations. The usual care provider, although of interest, was almost never reported.

As for the patient inclusion criteria, if a number of seizures was specified and within a certain time period previous to study entry, this was captured. If only patients with so-called unprovoked seizures were included, this was noted. As for patient exclusion criteria, if specific secondary causes of seizures, such as tumors or infections, were excluded, these were noted. Only the total numbers of patients were captured, since isolated comparisons of particular treatment or intervention results were not the objective of this project. Summary data tables, therefore, do not display results by treatment group, but overall by study. Seizure types at study entry were frequently reported without an explicit epilepsy syndromic classification or vice versa. For this reason, these originally separate fields were combined to capture both. A family history of febrile convulsions was not taken as equivalent to a family history of epilepsy. Prior emergency medical treatment to suppress seizures did not constitute prior AED use, which was intended to capture those receiving AEDs on a long-term basis.

For data capture of information about diagnostic, monitoring, and/or treatment interventions, any measures reported as required per the study protocol were captured, and those not part of the study protocol were not captured. Often, however, this distinction was unclear, and we chose to err on the side of over-capturing such data. The number of patients subjected to any intervention was captured in reference to the total number of patients entering the study, without regard for the number of study dropouts, which was often not explicitly reported. The timing and/or sequence of interventions was captured when explicitly reported, but this was rarely available. For this information, extractors made broad categorizations as prestudy, at study entry, throughout the study, or at some specific followup time or interval. For diagnostic interventions, a "gold standard" was always sought, but rarely found. When reported, the gold standard was typically a general clinical assessment rather than a specific test result, such as a particular EEG finding. With regard to so-called standard EEGs, we discovered that authors used different definitions, with regard to the inclusion of a sleep phase, photic stimulation, the number of channels, and montages used.

For monitoring and treatment interventions, seizure outcomes were preferentially captured as the number of patients achieving freedom from seizures for a specified period of time (remission rate) or the number of patients having a seizure during a certain followup period (recurrence rate). For side effects information, only the number of patients withdrawing because of intervention-related side effects was captured, as representing the most fundamental intervention-related side effect. Similarly, for compliance information, only the number of patients withdrawn because of noncompliance was captured. More detailed information regarding the nature and sequelae of side effects and compliance was not captured, as this was outside the scope of this review. If as a result of the intervention, a change in diagnosis or AED (drug, not dose) was made, this was sought (but rarely found), with the number of patients so affected. As for information on the use and consequences of measuring drug levels, these were categorized as standard unless plasma concentrations of metabolites or levels in other body fluids, e.g., cerebral spinal fluid (CSF), were measured, in which case they were categorized as sophisticated. Lastly, the outcomes and outcomes formats used were so disparate that we frequently had to resort to an "Other" data field to capture the main results.

Data were entered from the data extraction forms into MetaHub, a relational database of clinical trials. At the time each DEF was entered, 100 percent of data elements were checked back against the originals. Prior to locking the database, a 20 percent random sample of data in the completed database was checked by the quality control (QC) group at MetaWorks against the DEFs. Error rates in excess of 2 percent of QC-checked data would trigger a 100 percent recheck of all data elements in the database.

Statistical Methods

No statistical analyses were planned beyond basic descriptive statistics used to summarize data.

Peer Review

A group of 19 peer reviewers was assembled to review the draft Final Report describing this project. The peer reviewers were drawn from consumer groups and professional organizations (American Academy of Neurology, Epilepsy Foundation, and American Epilepsy Society), the nominating partner (CDC), and the AHRQ. The reviewers represented several medical specialties (neurology, nursing, primary care). All reviewers were asked to complete a list of questions about the format and content of the report (see Appendix L), and to also provide any text comments. All reviewer comments were shared with the AHRQ. The peer reviewers comments were considered and, wherever feasible and within the scope of this project, incorporated into the Final Report. Comments were ultimately received from 14 of the 19 reviewers who were invited to comment.

PubReader format: click here to try


  • PubReader
  • Print View
  • Cite this Page

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...