The primary objective of this study was to implement recommendations of the International CFS Study Group  and define CFS on the basis of scores from standardized and validated instruments that assess the major dimensions of the illness as specified by the 1994 CFS case definition . Functional impairment, fatigue and an accompanying symptom complex characterize CFS. We defined functional impairment as scores ≤ 70 on the physical function or ≤ 50 on the role physical or ≤ 75 on the social function or ≤ 66 on the role emotional subscales of the SF-36. We defined severe fatigue as scores ≥ 13 on the general fatigue or ≥ 10 on the reduced activity subscales of the MFI. Finally, we defined the accompanying symptom complex as reporting the occurrence of ≥ 4 of 8 symptoms and scoring ≥ 25 on the Symptom Inventory Case Definition subscale. One could debate our choice of specific subscales from the SF-36 and MFI and the specific cut-off values we chose on the SF-36, MFI and Symptom Inventory. However, these instruments have been validated, have been used extensively in studies of CFS and other illnesses, and are known to be reproducible. In contrast, most studies of CFS merely note that they used the 1994 case definition and they do not generally specify how disability, fatigue and symptom occurrence were elucidated. Thus, it is difficult to assess the validity of their diagnostic criteria and essentially impossible to compare results between studies critically.
This study showed scant stability of CFS over time, when diagnosed by the usual algorithm (based on patients' subjective responses to direct questions as to whether they feel fatigued, if they perceive their fatigue causes substantial reduction in daily activities, and whether at least 4 case defining symptoms are present). There was poor correlation between illness classification during surveillance (recruitment classification) and classification by the same criteria during the clinical study. While this might reflect fluctuation in illness over time, illness categories (CFS, ISF, Remission, non-fatigued) defined by this surveillance classification scheme were not consistent with respect to overall illness severity. In contrast, when we defined illness categories (i.e. CFS, ISF, Not Ill) according to validated clinically empirical instruments, we found an extremely strong relationship between diagnostic category and every measure of severity. It is not surprising that scores on the 4 SF-36 and 2 MFI scales used in classification were significantly correlated with diagnosis, but the 4 SF-36 and 3 MFI scales not used in classification showed similar significant associations with diagnostic category. Although correlated, the dimensions of impairment and fatigue defined by these instruments represent statistically independent factors. Finally, while only 13% of patients who met the 1994 case definition criteria during surveillance met those same criteria in the present study, 40% fulfilled CFS criteria of the clinically empirical definition. Thus, the clinically empirical case definition may be less affected by the day-to-day fluctuation of the illness and rather reflect the underlying chronic illness process.
Defining CFS in this empirical manner will improve the precision of case ascertainment in research studies, it will provide a standard reproducible means of following the clinical course over time, and it will help to clarify the extent to which patients from different referral clinics are similar (or dissimilar). Finally, this strategy can be used in primary care settings and will give health care providers a standard and reproducible method for diagnosing CFS. The SF-36, MFI and Symptom Inventory are relatively short forms that are completed by patients: scoring is straightforward and can be accomplished by clinic staff. The MFI and Symptom Inventory are in the public domain but the SF-36 is sold under license.
Use of this diagnostic strategy also provides health care providers with objective measures of the disability associated with CFS. The SF-36 measures functional impairment in 8 distinct dimensions. Normative values have been rigorously derived and documented in each dimension for healthy individuals and in a wide variety of disease states [11, 17, 18]. For example, in this study subjects with empirically defined CFS had lower scores on all scales, except physical function and general health, than those of patients with congestive heart failure. In addition to documenting impairment, scores on individual components of the SF-36 and MFI can be used to identify specific aspects of the illness for specific interventions and changes in scores over time can be used to monitor response to therapeutic interventions.
In addition, there is a certain mathematical appeal to the new criteria. Principal components analysis of the matrix of 14 scores (8 SF-36, 5 MFI, and the Symptom Inventory Case Definition subscale) for those enrolled in this study showed that their symptoms define a roughly 3-dimensional space (rather than 14) and the first 3 principal components accounted for 76% of the total variation among individuals. These 3 components roughly align with the most significant scales of the SF-36, MFI and Symptom Inventory, all of which are part of the 3 inequalities we used to empirically classify illness. As a consequence, the 3 components cleanly separate the empirically classified groups of CFS, ISF and Not Ill in 3-dimensional space, whereas this separation could not be achieved with the older classification scheme
Finally, our findings highlight the importance of continued evaluation of persons with CFS. Clinical management of CFS poses a difficult challenge for health care providers, patients and patients' families/associates because of the absence of physical signs and because illness severity and symptoms vary dramatically over time. Not all new symptoms or worsening of severity should be attributed to CFS a priori. Thirty-two (13%) of the study participants had serious medical or psychiatric diseases diagnosed for the first time during this study and another 18 (7%) were newly discovered to have melancholic depression, thus supporting inclusion of this disease in the exclusionary criteria. Those 49 subjects had been repeatedly evaluated during the Wichita surveillance study and had no prior evidence of such conditions.