Ineffective or effective? Interpreting the effect of CBT on fatigue in ME/CFS. A short analysis of a meta-analysis

Mark Vink^1*Friso Vink-Niese²

• ¹Independent researcher, Family and insurance physician, Netherlands
• ²Independent researcher, Osnabrück, Germany

The meta-analysis by Kolala et al. (2025a) selected 12 studies and was set up to examine "whether a non-protocol based CBT is effective in a population with CFS that is not defined Discussion 1. Registration According to the meta-analysis' registration with PROSPERO, the international prospective register of systematic reviews, the objectives of the review are to find out what the efficacy of CBT is "on quality of life and level of fatigue in patients diagnosed with Chronic Fatigue Syndrome both immediately after the delivery of the intervention, as well as long term" (PROSPERO, 2023). There is nothing about determining the efficacy of individual face-to-face or self-directed CBT separately. Consequently, these are post-hoc analyses and deviations from their protocol. Why the meta-analysis does not mention that or why these were done, is unclear. Five studies using protocol based CBT with a time contingent increase in physical activity, were used to claim that individual face-to-face CBT is effective for fatigue. However, they, like all the other studies in the meta-analysis, should have been excluded because they did not examine non-protocol based CBT which the meta-analysis said it would examine.As noted by the meta-analysis, Prins et al. (2001) used the CDC criteria, apart from requiring four of eight additional symptoms to be present. Consequently, they used the Oxford criteria. According to Deale's long-term follow-up article (2001), that study used those too. Both studies should therefore have been excluded. Jason et al. (2007) concluded that there were few differential results among the 4 non-pharmacologic interventions (CBT, cognitive therapy with pacing, aerobic treatment and relaxation). They also noted that their sample was not representative for CFS because many patients were working and their study was small. 25% (82/328) of the selected patients refused to take part in Gotaas et al. (2023) affecting the generalisability of its results because people who think they will not benefit or are made worse by the treatment, are usually the ones who refuse to take part. Not adequately treated were 24.4% (19/78) in the 16-week CBT group and 20.5% (16/78) in the eight-week interpersonal oriented CBT group. In Prins et al. (2001), 40.9% dropped out from the CBT group. Usually, people who do not improve and/or are negatively affected by treatment are the ones who drop out. According to the 5 and 20 rule, studies are highly biased if 20% or more drop out or are inadequately treated (Heneghan et al., 2017). Consequently both Gotaas et al. (2023) and Prins et al. (2001) were highly biased because of that. The bias in both studies increased further because Gotaas et al. (2023) did not publish its objective test, VO2max, and Prins et al. (2001) did not publish two of them, objective cognitive function and the actometer, and they claimed that the absence of effect on hours worked was because patients did not have time to find work, even though they had at least six months to do so. The other two objective outcomes were published years later and showed that CBT does not lead to objective improvement (Knoop et al., 2007a;Wiborg et al., 2010).Van der Schaaf et al. (2024, was a study of female patients only. The CIS-fatigue score after CBT was 33.6 which is only slightly better than the score for severely fatigued (⩾35). Moreover, a score of 30.5 is high, patients with sarcoidosis in clinical remission score 17.2 (Korenromp et al., 2011) and the healthy controls in Van der Schaaf et al. (2024) scored 16.9. In Prins et al. (2001) the scores after CBT were 40.5 (eight months) and 39.5 (14 months). The other three studies did not use this scale.The Sickness Impact Profile-8 (SIP8) score after CBT in Van der Schaaf et al. ( 2024) was 943.8, whereas with a score of ⩾700 one is severely disabled according to the study itself and the score of healthy women is 65.5 (Knoop et al., 2007b). After CBT in Prins et al. (2001) it was 1220 (eight months) and 1170 (14 months). The other three studies did not use this scale. Three studies were used to claim that self-directed CBT is effective for physical functioning. Janse et al. (2018) noted that "the fact that our study did not select on the level of physical functioning will make it more difficult to find an effect of iCBT on physical functioning" (p. 116). Knoop et al. (2008) did not use a physical functioning entry requirement either and Tummers et al. ( 2012) noted that "the inclusion criterion that a patient must have disabilities at the level of physical and/or social functioning [scoring 70 or less] meant that not all patients experienced disabilities in physical and social functioning" (p. 2208). Consequently, all three studies did not have a physical function entry requirement and therefore should not have been used to claim efficacy on physical functioning. Moreover, after treatment the score was only 65.4 in Tummers et al. (2012). The SIP8 score after treatment in Knoop et al. (2008) was 1079, in Janse et al. (2018) it was 867.8 (protocol driven feedback iCBT) and 885 (feedback on the demand iCBT) and Tummers et al. (2012) did not use that.Additionally, the three studies were done by the same Institute, i.e. The Dutch Knowledge Centre of Chronic Fatigue (NKCV), whose director (Knoop) was involved in all three studies. Systematic reviews and meta-analyses should encompass multiple studies from multiple centres to establish robust evidence (Drucker et al., 2016). Whereas the 3 studies were from only one centre instead of from multiple centres. Finally, according to the NKCV, there is a bad correlation between subjective and objective measures of physical functioning. Most CBT studies do not use objective outcomes, but 3 studies by the NKCV that did use them, showed that CBT does not lead to objective improvement of physical functioning (Wiborg et al., 2010). Kolala et al. (2025a) excluded studies that used the Oxford criteria because PEM, the main characteristic of ME/CFS, is not compulsory for diagnosis. They did not exclude studies that used the Fukuda criteria, even though PEM is only an optional criterion according to those criteria (Fukuda et al., 1994). Consequently, those studies should have been excluded as well. Just like the 2 studies that used the Oxford criteria but were erroneously not excluded. A basic principle of a properly conducted study is that patients in the control group receive the same number of sessions, attention and expectation raised about efficacy of the control treatment as patients in the therapy group so that the only difference between those groups is the therapy (Cuijpers et al., 2016). Because one then knows that an improvement is down to the treatment and not caused by all sorts of undefined differences between the groups.All three studies that were used to claim efficacy on physical functioning used a waiting list control group. Two of the 5 studies that were used to claim efficacy on fatigue (Van der Schaaf et al., 2024 andGotaas et al., 2023) used the waiting list and Prins et al. (2001) used natural course. Only Deale et al. (1997) and Jason et al. (2007) used an active control group (relaxation). According to a study by psychology professor Cuijpers (2016) entitled: How to prove that your therapy is effective, even when it is not: a guideline, one can get false-positive results for an intervention by using subjective primary outcomes and an inactive control group in non-blinded studies. Moreover, a systematic review by Fordham et al. (2021), which concluded that CBT is effective for a range of diseases, including ME/CFS, also noted that if studies use an active instead of a passive control group, that the effect size is trivial (0.09). The meta-analysis concluded that "there was inconsistent reporting of adverse effects, however no serious adverse effects were reported" (Kolala et al., 2025a, p. 01). The latter means that no one had to be admitted to hospital or died as a consequence of the therapy which is not surprising because patients do not complain about this. Núñez et al. (2011) did report about adverse effects and concluded that treatment led to "worse SF-36 physical function and bodily pain scores" but also that the treatment was "ineffective…and may in fact be harmful " (p. 381,388). The meta-analysis uses the Cochrane Risk of Bias Tool (RoB). Of the 8 studies used to declare CBT effective for fatigue and physical functioning, only the study by Tummers et al. ( 2012) was deemed to be high risk of bias. There was some concern about risk of bias in selection of reported outcome in the study by Prins et al. (2001). Yet as we have seen, they did not publish 2 objective outcome measures that showed that CBT did not lead to objective improvement, and they used a fallacy to ignore the null effect on the number of hours worked.As we have seen in Gotaas et al. (2023), a substantial number of participants did not adhere to the treatment. In Janse et al. (supplement, 2018), 81% (65/80, feedback on demand) and 84% (67/80, protocol-driven feedback), respectively, did not adhere to treatment in the two treatment groups. Consequently, all three studies should have been labelled as having high risk of bias. Moreover, all studies were non-blinded by definition that relied on subjective outcomes, in studies where patients were instructed to interpret their symptoms differently, i.e. as symptoms of normal life instead of a disease. Consequently, one does not know if a change in scores in subjective outcomes is down to the treatment or patients simply interpreting questions at the end of treatment differently than at the beginning of the study.The RoB notes the following about this. "Review authors will need to judge whether it is likely that participants' reporting of the outcome was influenced by knowledge of intervention received, in which case risk of bias is considered to be high". It also gives the following example: "level of pain reported at the end of a course of acupuncture, in a study comparing acupuncture with no treatment, is likely to be affected by knowledge of the intervention received" (Higgins et al., 2019, p. 51). Consequently, all trials in the meta-analysis had a high risk of bias. This short analysis shows that there are many issues with the meta-analysis by Kolala et al. (2025a) and the studies in it. The meta-analysis was set up to examine whether non-protocol based CBT was effective for ME/CFS, yet none of the included studies studied that. Moreover, in none of the included studies was PEM, the main characteristic of the disease, required for diagnosis so that it's unclear how many patients in the studies actually suffered from the disease under investigation.Moreover, is a small improvement on subjective outcomes, i.e. fatigue and physical functioning, in non-blinded studies that use badly designed (non-active) control groups, when quality of life doesn't improve, adherence to treatment is bad, dropout is high, studies use selective reporting of objective outcome measures, there is no objective improvement, patients are still very ill and severely disabled, and more patients are unable to work and reliant on illness benefits than before treatment, as found by an extensive review of the literature (Vink and Vink-Niese, 2019), really indicative of efficacy of a treatment and the need for booster sessions? Or does it highlight its inefficacy and the fact that it's not the right treatment for ME/CFS? Author contributions: Conceptualization, M.V.; methodology, M.V. and F.V.-N.; validation, M.V. and F.V.-N.; writing-original draft preparation, M.V.; writing-review and editing, M.V. and F.V.-N.; and supervision, M.V. and F.V.-N. All authors have read and agreed to the published version of the manuscript.