Expert Input

This evidence report was produced by a multidisciplinary group that included: Cynthia Mulrow, MD, MSc, a general internist, Gilbert Ramirez, DrPH, a research methodologist, and John Cornell, PhD, a psychometrician. Martha Harris, AHIP, MLS, MA, a medical librarian, conducted literature searches and maintained the reference databases. Christine Aguilar, MD, MPH, a physician research associate, and Jodi Sapp, RN, a nurse research associate, abstracted information from the articles; David Mullins developed electronic databases for the project and provided data entry/data validation support. Drs. Ramirez and Cornell calculated effect size statistics for comparing study outcomes. Drs. Mulrow, Ramirez and Aguilar coordinated the project administratively. Karen Stamm edited and formatted the report, and coordinated communication with technical experts and peer reviewers. Drs. Ramirez, Cornell, and Mulrow wrote the report.

An international expert panel provided input regarding: the plan and format for the evidence report; interpretation and presentation of findings; critique of drafts; and future research priorities. Members of the panel from the United States included: Dedra Buchwald, MD, University of Washington-Seattle; Mark Demitrack, MD, Eli Lilly and Company; Sudhir Gupta, MD, PhD, University of California-Irvine; Leonard A. Jason, Ph.D., DePaul University; James F. Jones, MD, National Jewish Medical and Research Center; Nancy G. Klimas, MD, University of Miami; Kathy Rabin, JD, a CFS patient; John H. Renner, MD, National Council for Reliable Health Information (until his death in September 2000); Joan Shaver, PhD, RN, University of Illinois-Chicago; Eng M. Tan, MD, Scripps Research Institute; and Vicki C. Walker, BA, Research and Public Policy Project Manager, The CFIDS Association of America. Non-U.S. members of the technical expert panel included: Susan Abbey, MD, University of Toronto, Canada; Peter Campion, PhD, FRCGP, MRCP (UK), DCCH, University of Hull, United Kingdom; Robert H. Loblay, MD, University of Sydney, Australia; and Peter D. White, MD, St. Bartholomew's and the Royal London School of Medicine and Dentistry, Queen Mary and Westfield College, London, United Kingdom. Ex-officio members included: David Morens, MD, National Institute for Allergy and Infectious Diseases, and William Reeves, MD, MSPH, Centers for Disease Control and Prevention. In addition, the report's penultimate draft was peer-reviewed by 15 individuals selected for their expertise in CFS (see Appendix A for a full list of external peer reviewers).

Questions Addressed in Evidence Report

The agency that nominated the topic initially suggested a broad-ranging list of more than 20 questions for our report to address. We enlisted the help of the technical expert panel to determine which questions we could realistically address, given the enormity of the data, and the limits on time and resources. Using a consensus Delphi process, the technical experts narrowed our initial scope to the following highest priority questions:

1. What are the existing case definitions for CFS?
2. Which case definitions, if any, have been substantiated and/or validated with reliably discriminating constellations of symptoms in adults?
3. What are the prevalence and natural history of CFS in adults?
4. Do controlled studies in adults show that particular therapies improve clinical symptoms of CFS, when compared to placebo, no therapy, or each other?

Literature Search and Selection Methods

Selection Process

Figure 1 depicts the selection process. Of 5,221 identified records from the various electronic databases and sources, 1,111 were eliminated because they were duplicate citations cited in more than one database (i.e., same authors, title, journal source information). Two reviewers independently screened the titles and abstracts of the remaining 4,110 records to select articles that were relevant to evidence questions one through four. Of the 4,110 records that were screened, 518 were considered potentially relevant to the specified evidence report questions. Of 124 that addressed case definitions for CFS, 41 met selection criteria. Of 83 articles that addressed prevalence or natural history, 20 met the selection criteria. Of 311 articles that addressed therapies, 38 were controlled trials that met selection criteria.

Figure

Figure 1. Literature search results flow chart.

Data Abstraction Process

Two reviewers (physician, psychometrician, methodologist and/or nurse) independently abstracted data from the selected studies. None of the abstractors were blinded either to study title or to author's names. Disagreements in abstractions were uncommon (less than 1% of items) and were resolved by consensus. No formal reliability testing was done.

We critically appraised the quality or rigor of studies by assessing several parameters. For studies addressing prevalence, we assessed: the method and time course of ascertainment, the sampling method, whether the sample was representative, whether an established case definition was used, and the response rate. For studies addressing natural history, we assessed: sampling methods, whether the sample was representative, whether an established case definition was used, methods of followup, and followup rates. For studies addressing substantiation of case definitions, we assessed: incorporation and selection biases, whether the study sample was representative, how the diagnosis was verified and by whom, whether compared groups were equivalent with respect to key sociodemographic or clinical variables that could affect the manifestations being studied, and whether groups were matched on such confounding variables. For studies addressing therapy, we assessed: whether the controlled study was randomized, the adequacy of randomization (method and concealment of assignment); whether the trial was single or double-blind; numbers of dropouts; and appropriateness of analyses methods.

Data Synthesis Process

We synthesized data descriptively, emphasizing methodological characteristics of the studies, such as sources of populations enrolled, CFS case definitions used to select study participants, sample sizes, adequacy of randomization process, interventions and comparisons. We examined relationships between clinical outcomes, participant characteristics, and methodological characteristics in evidence tables and graphical summaries, such as forest plots. We grouped outcomes according to a list that had been developed specifically to assess symptoms in patients with CFS.²⁰

To help compare outcomes across studies, we computed standardized mean differences between treatment and comparison group scores as a measure of effect size for each study. These estimates were adjusted for between-group differences at baseline and for small sample bias.²¹ We adjusted for baseline differences by calculating an “effect size” at baseline; by definition, it should be zero if study groups were well matched. When we found a non-zero “effect size” at baseline, we adjusted outcome effect sizes were by subtracting the baseline effect size.