This study adhered to human experimentation guidelines of the U.S. Department of Health and Human Services and complies with the Helsinki Declaration. All participants were volunteers who gave informed consent.
Study design
The survey used a stratified two-stage cluster design similar to that described in our previous study of fatiguing illnesses in the general population of Wichita, Kansas [1]. Briefly, the first stage of sampling selected a primary sampling unit (PSU) in each stratum, and the second stage drew a sample of telephone numbers for each PSU. Phase I of data collection screened households for individuals identified as fatigued for ≥ 1 month. Phase II comprised detailed telephone interviews with individuals identified as fatigued and with a random sample of persons identified as non-fatigued. Phase III was a clinical evaluation of chronically fatigued subjects who, based on telephone interviews, had no exclusionary medical or psychiatric conditions and who met fatigue and symptom criteria of CFS (CFS-like).
Sampling strategy
To examine regional and metropolitan differences, we constructed strata from statistical areas defined by the US Census. Each of the four US Census regions (Northeast, Midwest, South, and West) was further stratified into metropolitan statistical areas (MSAs) and non-MSA counties. The U.S. Office of Management and Budget defines several categories of MSA, according to specific standards. In general terms an MSA has a core area containing a sizable population, together with adjacent communities that have a high degree of economic and social integration with that core. From each of the eight strata, we randomly selected a primary sampling unit (PSU), either an MSA or a non-MSA county (as appropriate). Because only one PSU was selected from each stratum, the results do not provide estimates of within-stratum variability. The eight PSUs selected were Buffalo-Niagara Falls, New York (Northeast urban), Chicago, Illinois (Midwest urban), Baton Rouge, Louisiana (South urban), Oakland, California (West urban), Franklin County, Pennsylvania (Northeast rural), Ripley County, Indiana (Midwest rural), Monroe County, Georgia (South rural), and Chaves County, New Mexico (West rural). For each PSU, 1,800 telephone numbers were randomly selected using the GENESYS Sampling System (Marketing Systems Group, Fort Washington, PA) and identifiable non-working and business numbers were removed. Advance letters, explaining the study, were sent to all households whose telephone number could be matched to a mailing address.
Phase I – telephone screening interviews
In each selected household, we screened an adult household informant who was at least 18 years old. The nature of the study was explained to the informant, who was requested to consent verbally before proceeding with the interview. The informant enumerated individual household members and reported on their age, sex, race, and fatigue status. Informants were asked whether any household members were currently suffering from severe fatigue, extreme tiredness, or exhaustion that had lasted 1 month or longer.
Phase II – detailed telephone interviews
Household residents older than 18 years identified with fatigue ≥ 1 month were contacted for detailed interviews; parents were interviewed on behalf of adolescents (12 to 17 years old). We also conducted detailed interviews on a random sample of non-fatigued adults 18 to 69 years old selected in a two-step process. First, households were randomly selected (with a probability of 0.25) to supply a non-fatigued person for a detailed interview. Second, if the household enumeration identified at least one age-eligible non-fatigued person in the household, a non-fatigued person was selected at random. Only one non-fatigued person could be selected from a household. A household in which a fatigued person was selected was eligible to have a non-fatigued person selected as well. However, in this study no household had both a fatigued and non-fatigued resident selected for interview.
As with the screening interview, the nature of the study was explained and subjects provided verbal informed consent before proceeding with the interview. The detailed interview included questions on fatigue (characteristics and duration), symptoms (occurrence, nature, and duration), demographics, and medical/psychiatric history. It also included the 12-item Short Form Health Survey® (QualityMetric, Inc., Lincoln, RI), which is designed to measure health-related quality of life.
Case definitions
CFS is an illness defined by symptoms and associated disability and by excluding medical diseases or psychiatric conditions that could explain them. There are no confirmatory physical signs or characteristic laboratory abnormalities. The current international CFS case definition [3] defines CFS as clinically evaluated, medically or psychiatrically unexplained, persistent or relapsing fatigue of at least 6 months duration that is not the result of ongoing exertion, is not substantially alleviated by rest, and results in substantial reduction in previous levels of occupational, educational, social, or personal activities. The fatigue must be accompanied by at least 4 of 8 symptoms that must have persisted or recurred during at least 6 consecutive months and cannot have predated the fatigue. Case defining symptoms include: 1) unusual post-exertional malaise of more than a day's duration following previously tolerated levels of mental or physical exertion; 2) unrefreshing sleep; 3) impaired short-term memory or concentration with substantial reduction in occupational, educational, social, or personal activities; 4) headaches of a new type, pattern, or severity; 5) muscle pain; 6) multi-joint pain without swelling or redness; 7) sore throat; 8) tender cervical/axillary lymph nodes. The current CFS case definition relies entirely on self-reported symptoms and disability and specifies no standard measures. Recently an International Chronic Fatigue Syndrome Study Group has recommended resolutions for the major ambiguities [4].
This study relied on telephone interview to classify subjects. We used the CDC detailed telephone interview questionnaire to identify and characterize fatigue, accompanying symptoms, and exclusionary conditions. This questionnaire (available from the authors on request) included minor modifications of one used over 4 years of surveillance in Wichita [1]. Individuals reporting fatigue lasting at least 6 months were classified as having chronic fatigue. We classified chronically fatigued respondents as having a CFS-like illness if they had no medical or psychiatric exclusions identified during the interview and reported chronic fatigue that was not alleviated by rest and was accompanied at least 4 of the 8 CFS defining symptoms. Whether fatigue substantially interfered with work, educational, or personal activities was not assessed in the detailed telephone interviews. Classification as CFS requires a complete physical examination to accurately detect medical or psychiatric conditions that subjects may not recount on interview. Fatigue categories were analyzed as discrete groups (chronic fatigue and CFS-like), and each subject was counted only in the most restrictive category that applied.
Statistical analyses
Prevalence estimates were based on weighted data. Each household received a base sampling weight that reflected the probability of selection of the PSU and selection of the household telephone number within the PSU. The final household sampling weight incorporated adjustments for multiple residential lines, non-response, and households without telephones. For non-fatigued individuals who completed detailed interviews, the person-level weight was determined by the household weight and the person's probability of selection. For fatigued individuals, all of whom were selected with certainty, the person-level weight for the detailed interview equaled the household weight. Weighted prevalences and weighted Pearson chi-squared and Wald statistics were calculated by using STATA 7.0 (Stata Corporation, College Station, TX). Confidence intervals for prevalences were constructed by using a logit transformation (hence the lower endpoint was always greater than 0).