Lancet psychiatry correspondence, by Michael Sharpe, Kimberly A Goldsmith, Anthony L Johnson, Trudie Chalder, Jane Walker, Peter D White, 18 January 2016 Authors’ reply:
The original PACE trial Article reported that the rehabilitative treatments of cognitive behaviour therapy (CBT) and graded exercise therapy (GET) produced a greater improvement in fatigue and physical function for patients with chronic fatigue syndrome than did adaptive pacing therapy (APT) or standard medical care alone, when measured at the trial final outcome 1 year after randomisation.
We used a standard research definition of chronic fatigue syndrome (Oxford Criteria) for entry to the trial; we found similar results in the subgroup analyses for other definitions of chronic fatigue syndrome and of myalgic encephalomyelitis. Self-rated measures of fatigue and physical function were chosen as primary outcomes because chronic fatigue syndrome is diagnosed by self-report (in the absence of an alternative diagnosis) and we therefore considered these to be the best way to measure improvement.
The trial was rigorously conducted and closely monitored. As our collaborators in the research, trial participants received regular newsletters approved by the research ethics committee, which covered a variety of relevant topics. As noted, one of these newsletters included positive quotes from participants. Since these participants were from all four treatment arms (which were not named) these quotes were neither evidence of lack of equipoise nor a source of bias, as suggested by James Coyne and Keith Laws.
Our more recent Article reporting long-term patient outcomes found that improvement was maintained for those originally allocated to CBT or GET. While recovery for all participants receiving these treatments has never been claimed, a mean improvement of approximately 20 points on the medical outcomes study short-form 36 (SF-36) physical functioning scale from trial entry is, contrary to Frank Twisk’s comments about the lack of treatment effect, clearly clinically meaningful. Nor is its value to patients negated by Coyne and Laws’ reference to mean SF-36 scores from selected cross-sectional studies of patients with other illnesses.
Participants allocated to APT or standard medical care alone in the trial showed poorer outcomes at 12 months. However, the outcomes reported by those who received these treatments in the trial and participated in the follow-up study had improved to the extent that, in a comparison of long-term outcomes by original trial treatment allocation, we found few differences.
Although this finding might mean that all PACE treatments have similar long-term outcomes, we cannot draw this conclusion. This is because, as explained in the Article, the follow-up study was not, as suggested by Coyne and Laws, a continuation of the trial (with or without crossover), but rather a naturalistic follow-up after trial completion.
During this follow-up period, more than half of participants originally allocated to standard medical care alone and to APT sought at least some treatment with CBT or GET. Consequently, the long-term outcome in these participants is not of standard medical care alone or APT alone, but of these therapies supplemented in many cases by CBT or GET.
Furthermore, we know that the seeking of such additional therapy was not, as suggested, merely a response to getting only standard medical care, because participants who had been allocated to APT (which offered the same contact time as CBT and GET and was rated as at least as logical by participants) also frequently sought it.
The statistical analysis accounted for missing data using linear mixed effects models and made the missing-at-random assumption. These are widely accepted approaches, given that missing data assumptions cannot be properly tested; the missing-at-random assumption was acknowledged as a potential limitation in the Article.
The rationale for including the missing data predictors was stated in the Article and makes the missing-at-random assumption more plausible. The covariates were either stratification factors or chosen as significant predictors in a model with lack of data as the dependent variable.
As highlighted by Charles Shepherd, the findings of the PACE trial and follow-up study differ from those of a patient association membership survey, which reported that CBT and GET were unhelpful and even harmful. Although of interest, these survey findings do not negate the results of the trial. There are many possible reasons for the difference between the survey and our results.
First, inclusion in the trial was based on a diagnostic assessment for chronic fatigue syndrome, whereas we cannot be confident of the diagnoses of survey participants identified simply by being a member of a patient association. The treatments given in the trial were rigorously monitored for quality, whereas we know little about the treatment reported on in the survey and do know that in clinical practice treatment may be mislabelled or poorly delivered. The trial had a good follow-up rate, whereas we know that surveys with an unknown sample and response rate, as is the case with this survey, are subject to bias.
Finally, we respectfully point out that disagreeing with the findings of a research study does not necessarily mean that the study is flawed. The marked antipathy of some toward this study, apparent in the correspondence and elsewhere, is perhaps related in part to the view, expressed by Shepherd, that the findings support “a simplistic and flawed model of psychosocial causation”. They do not. They do, however, offer practical hope of useful improvement, a finding reinforced from the routine monitoring of the outcomes of the many patients who attend clinics that offer these and similar treatments.
PDW has done voluntary and paid consultancy word for the UK Government and a reinsurance company. TC has received royalties from Sheldon Press and Constable and Robinson. MS has received royalties from Oxford University Press. KAG and JW declare no competing interests.
1. White, PD, Goldsmith, KA, and Johnson, AL. Comparison of adaptive pacing therapy, cognitive behaviour therapy, graded exercise therapy, and specialist medical care for chronic fatigue syndrome (PACE): a randomised trial. Lancet. 2011; 377: 823–836 Summary | Full Text | Full Text PDF | PubMed | Scopus (234)
2. Sharpe, M, Goldsmith, KA, Johnson, AL et al. Rehabilitative treatment for chronic fatigue syndrome: long-term follow up from the PACE trial. Lancet Psychiatry. 2015; 2: 1067–1074
Summary | Full Text | Full Text PDF
3. ME Association. ME/CFS illness management survey results. “No decisions about me without me”. http://www.meassociation.org.uk/wp-content/uploads/2015-ME-Association-Illness-Management-Report-No-decisions-about-me-without-me-30.05.15.pdf. ((accessed Jan 5, 2016).)
4. Brimmer, DJ, Maloney, E, Devlin, R et al. A pilot registry of unexplained fatiguing illnesses and chronic fatigue syndrome. BMC Res Notes. 2013; 6: 1–11
CrossRef | Scopus (0)
5. Gladwell, PW, Pheby, D, Rodriguez, T, and Poland, F. Use of an online survey to explore positive and negative outcomes of rehabilitation for people with CFS/ME. Disabil Rehabil. 2014; 36: 387–394
CrossRef | Scopus (4)
6. Crawley, E, Collin, SM, White, PD, Rimes, K, Sterne, JAC, and May, MT. Treatment outcome in adults with chronic fatigue syndrome: a prospective study in England based on the CFS/ME National Outcomes Database. QJM. 2013; 106: 555–565