Another person and I put this together fairly quickly. Hopefully there are no errors but probably best to check yourself if doing something formal or ask a few people. Primary outcome measures – Primary efficacy measures: Not reported. Refusal to do so. Mean Chalder Fatigue Questionnaire (Likert scoring) scores reported instead. Not reported. Refusal to do so. Mean SF-36 physical function scores reported instead. Not reported. Refusal to do so. Total score for pre-specified primary outcomes: 0 out of 3, or 0 out of 7 if counting individual parts. Secondary outcome measures – Secondary efficacy measures Elevated to primary outcome. Percentages have been reported in score bands of 1+2, 3+4+5, 6+7. Neither of these have been reported in any detail, but the SSMC doctor-rated scores were used to substitute missing patient-rated CGI data in the recovery criteria. No. All parts of the recovery criteria were changed, some drastically, there's now overlap between baseline entry criteria and recovery criteria. Yes. Yes. Yes but only overall value, not each dimension. Yes, but after minimal results this measure is now being downplayed as unreliable and not an "objective outcome measure of physical capacity". Only in graph form, and reported 4 years after the first paper was published. Null result challenges the rationale and some reported benefits of CBT and GET (suggests patients not increasing activity as presumed). A request for summary figures was rejected as 'vexatious'. Sort of: Borg/%max HR reached in Figure 2 of mediation analysis. Details reported, but after null results the investigators denied its usefulness as a "more objective measure of function". Was (sort of), always in violation of CDC criteria (symptoms counted over 1 week not 6 months). Likert scale data (that means not binary) was not given. Instead all that was given was present/absent data and only for two symptoms: "Poor concentration or memory" and "Postexertional malaise". Also, a composite yes/no score for the 9 CDC symptoms ["Chronic fatigue syndrome symptom count"] is given. No. Yes. Other outcomes The above does not include other outcomes. Under assumptions they state: That would have been related to the now abandoned primary outcomes. It certainly isn't possible now when the improvement rates were usually so high in all groups that it was impossible to get two to three times (would reach 100%). One of the two adverse outcomes was changed to a drop of 20 points over two followup measurements instead of one. But measurements only were taken at 0, 12, 24, 52 weeks, so two consecutive measurements may be several months and relapses lasting months would not be detected (but may be detected by other safety outcomes?). Still waiting on a paper on predictors. In the statistical analysis plan (which came out after the cost effectiveness paper was published), the PACE Trial investigators said they would: What they actually did was They also wrongly claimed: http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0040808 PACE Trial protocol said: Both of these were changed (for CGI, it became 6 and 7) There is no mention both of these were changed: One person: The serious adverse events and serious adverse reactions were reported, but the non-serious adverse reactions remain totally unreported (despite a dedicated paper on adverse effects). Non-serious adverse events reported and similar between groups but aggregated together (they were graded by severity but this hasn't been reported so there's a possibility that GET had more severe effects). As other definitions of adverse effects are quite strict, the true story of safety might be hidden in this data. Relapses significant to patients lasting up to a month could occur without being classed as serious. Post-hoc additions: Normal range in fatigue and physical function and both for individuals; Clinically useful difference in fatigue and physical function and both for individuals. Correctly reported as non-pre-specified? Normal range in fatigue and physical function: Yes (post-hoc). Individual level clinically useful difference: Yes (post-hoc). Recovery criteria: No. Claimed it was "pre-specified" but evidence suggests changes after data unblinding and knowledge of components to recovery criteria. Details of changes to mediation analysis unclear, seems to be an exploratory analysis. ---- This is largely based on the protocol. In the statistical analysis plan, it was said other things would be reported but we didn't have time at this stage to cross check it.