Assessment of recovery status in chronic fatigue syndrome using normative data
Adamowicz et al. have reviewed criteria previously employed to define recovery in chronic fatigue syndrome (CFS). They suggested such criteria have generally lacked stringency and consistency between studies and recommended future research should require “normalization of symptoms and functioning”.
Options regarding how “normalization of symptoms and functioning” might be operationalized for CFS cohorts are explored.
A diagnosis of CFS excludes many chronic disabling illnesses present in the general population, and CFS cohorts can almost exclusively consist of people of working age; therefore, it is suggested that thresholds for recovery should not be based on population samples which include a significant proportion of sick, disabled or elderly individuals. It is highlighted how a widely used measure in CFS research, the SF-36 physical function subscale, is not normally distributed. This is discussed in relation to how recovery was defined for a large intervention trial, the PACE trial, using a method that assumes a normal distribution. Summary data on population samples are also given, and alternative methods to assess recovery are proposed.
The “normalization of symptoms and function” holds promise as a means of defining recovery from CFS at the current time. However, care is required regarding how such requirements are operationalized, otherwise recovery rates may be overstated, and perpetuate the confusion and controversy noted by Adamowicz et al.
KeywordsChronic fatigue syndrome CFS Recovery Operational definition Normative data
Adamowicz et al.  discussed significant shortcomings in published definitions of ‘recovery’ from chronic fatigue syndrome (CFS), suggesting that more stringent and comprehensive criteria are generally needed and should include “normalization of symptoms and functioning.” Examining this in more detail reveals further important issues to consider when establishing thresholds for normal or recovery.
One important aspect is the selection of an appropriate set of population scores for comparison. CFS remains a diagnosis of exclusion, with a patient required to be unaffected by any medical or psychiatric condition that could explain the symptom(s) or disability . Exclusionary conditions include chronic disabling illnesses that are common in the general population, thus it seems reasonable to suggest that general adult population norms which include these illnesses are unsuitable to accurately assess recovery from CFS. An ideal approach to define recovery would be to use data from a demographically representative comparison group, which after a thorough clinical assessment, do not satisfy either the CFS criteria or the exclusions for the diagnosis. A similar approach was taken in a study by Goodwin et al.  though not in the context of recovery and relied upon self-report rather than clinical assessment. However, such a detailed approach may not always be possible, and the nearest equivalent group is likely to be a “healthy population”, which has previously been used to assess recovery from CFS .
CFS cohorts, particularly in intervention studies, tend to include very few patients who are elderly. For example, in the large PACE trial, only 21 out of the 640 adults (3 %) were aged 60 or more years , with a mean (±SD) age of 38 (±12) years at baseline . In such a scenario, working age population norms are preferable to those based on general population samples which include a significant proportion of elderly individuals. If further precision is necessary, the population scores could be subgrouped into age brackets to better match the age distribution of study participants.
Adamowicz et al.’s review included studies which used thresholds such as the mean ±1 SD of population data to define recovery. However, this simple method assumes that the population data follows a normal distribution, which is often not the case for health status scales. It is questionable to apply this method to the physical function (PF) subscale of the SF-36 health survey commonly used in CFS research, as it is not normally distributed, is highly left-skewed, and most scores are clumped toward the ceiling of the scale [7, 8]. A previous study on recovery from CFS had used the mean ±1 SD of healthy population data to define recovery for a range of outcome measures and acknowledged the unsuitability of this method when the assumption of normality has been violated, but did not account for it when establishing a recovery threshold for SF-36 PF because “we do not know the exact distribution of the SF-36 scores” .
The PACE trial offers an example of how such issues can cause normative data to be misinterpreted or misapplied when defining recovery from CFS. The investigators deviated from the trial protocol by lowering the SF-36 PF threshold for recovery from ≥85 to ≥60 points out of 100. The reported justification for this change incorrectly asserted that a threshold of ≥85 “would mean that approximately half the general working age population would fall outside the normal range” . The survey data they reference is available in the UK National Archive but shows that the median (IQR) score for the working age subgroup (16–64 years) of the general population is 100 (90–100) while only 17.6 % scored under 85 . The mean ± SD value of 84 ± 24 that White et al. used to calculate a threshold of ≥60 is for the overall general population aged 16–95 years, but even for this group the median score is 95 and still only 28.4 % scored under 85. Furthermore, if one restricts the sample to the working age population who did not rate themselves as having a long-term health problem that limited their daily activities, which could be described as a healthy population cohort, the mean ± SD is 95.0 ± 10.2 and the median (IQR) is 100 (95–100), with 61.4 % scoring the maximum of 100 and 92.3 % scoring 85–100. Contrast such scores with the threshold of ≥60, obtained using the formula mean −1 SD on general population data, that was used to define normal physical functioning (a criterion for recovery in the trial) . A score of 60 requires the reporting of multiple limitations (four to eight out of ten items depending on severity) and was low enough to overlap with the trial entry criteria for “significant disability” (which was operationalized as SF-36 PF ≤ 65 points) .
With regard to the SF-36 PF, a high percentage of the population score the ceiling value of 100 points, e.g., 43.8, 39.0 and 38.7 % for general population samples in England, South Australia and the USA, respectively, so data on the numbers achieving such a score can be useful for top-box analysis [10, 11, 12, 13]. In the general population sample from England with SF-36 PF scores (n = 2,047), 24.4 % were aged 65 or more years, and 22.0 % reported a chronic disabling health problem. For the healthy working age subgroup (n = 1,298) with a mean ± SD age of 38.3 ± 13.0 years (typical for CFS studies), 61.4 % scored the maximum of 100, 67.1 % reported no limitations for vigorous activities, and 93.8 % reported no limitations for moderate activities . Similar rates should be expected from study participants of working age who have completely recovered from CFS.
An alternative approach to population norms is to use a similar control group of people who do not have CFS. This was done in a follow-up study of patients diagnosed with CFS 25 years previously . Patients were divided up into those who considered themselves to have CFS and those that did not; the latter were called the “remit” group. Interestingly, on 21 out of 23 outcomes examining long-term health, symptoms and disability, those classed as remitted from CFS showed significantly more impairments compared to the healthy controls, seriously questioning their remission status.
This highlights another method to verify recovery status: statistical testing at the group level. A group of patients may be within some normative threshold (such as mean −1 SD) on some scale but should not be classed as completely recovered if the cohort still demonstrates significantly worse scores than an appropriate comparison group. If researchers do not wish to perform such an analysis, a range of summary data for any ‘recovered’ groups should still be reported, e.g., means, standard deviations, medians, modes, ranges and interquartile ranges for the relevant measures, accompanied with histograms or frequency tables if possible. Choosing a precise threshold may be difficult at this time and as pointed out by Adamowicz et al. , there is a lack of consensus on how recovery should be defined or interpreted, which has generated confusion and controversy. Making such additional information available would therefore allow researchers, clinicians and patients to better estimate the functional status of study participants classified as recovered and help to minimize the potential for misunderstanding.
Conflict of interest
- 1.Adamowicz, J. L., Caikauskaite, I., Friedberg, F. (2014). Defining recovery in chronic fatigue syndrome: a critical review. Quality of Life Research, 23(9), 2407–2416.Google Scholar
- 3.Goodwin, L., White, P. D., Hotopf, M., Stansfeld, S. A., & Clark, C. (2011). Psychopathology and physical activity as predictors of chronic fatigue syndrome in the 1958 British birth cohort: A replication study of the 1946 and 1970 birth cohorts. Annals of Epidemiology, 21(5), 343–350.CrossRefPubMedCentralPubMedGoogle Scholar
- 5.Couch, I. C. (2013). Re: PACE trial authors’ reply to letter by Kindlon. BMJ December 2 http://www.bmj.com/content/347/bmj.f5963/rr/675527
- 6.White, P. D., Goldsmith, K. A., Johnson, A. L., Potts, L., Walwyn, R., DeCesare, J. C., et al. (2011). Comparison of adaptive pacing therapy, cognitive behavior therapy, graded exercise therapy, and specialist medical care for chronic fatigue syndrome (PACE): A randomized trial. The Lancet, 377(9768), 823–836.CrossRefGoogle Scholar
- 7.Bowling, A., Bond, M., Jenkinson, C., & Lamping, D. L. (1999). Short Form 36 (SF-36) Health Survey Questionnaire : Which normative data should be used ? Comparisons between the norms provided by the Omnibus Survey in Britain, the Health Survey for England and the Oxford Healthy Life Survey. Journal of Public Health Medicine, 21, 255–270.CrossRefPubMedGoogle Scholar
- 9.White, P. D., Goldsmith, K., Johnson, A. L., Chalder, T., Sharpe, M., & PACE Trial Management Group. (2013). Recovery from chronic fatigue syndrome after treatments given in the PACE trial. Psychological Medicine, 31, 1–9.Google Scholar
- 10.Office of Population Censuses and Surveys. Social Survey Division, OPCS Omnibus Survey, November 1992 [computer file]. Colchester, Essex: UK Data Archive [distributor], September 1997. SN: 3660, doi:10.5255/UKDA-SN-3660-1
- 11.Behavioural Epidemiology Unit. (1995). South Australian population norms for the Short Form 36 (SF-36) Health Status Questionnaire. Adelaide, South Australian Health Commission. http://health.adelaide.edu.au/pros/docs/reports/general/generalhealthandwellbeing_south_australian_population_norms.pdf
- 12.Ware, J. E., Jr., & Kosinski, M. (2001). SF-36 physical & mental health summary scales: A manual for users of version 1 (2nd ed.). RI Quality Metric Inc: Lincoln.Google Scholar