By David Tuller, DrPH
*This is a crowdfunding month at UC Berkeley. If you’d like to support my work, the link is here.
Last year, eClinicalMedicine, a journal in the Lancet stable, published an article called “Effects of therapeutic interventions on long COVID: a meta-analysis of randomized controlled trials.” The paper reported that, according to “high-certainty evidence,” exercise training could improve health and “should be prioritized”. These conclusions were preposterous, for multiple reasons. (I wrote about it here and here.)
I wrote a letter to the editor, which has just been published. The editing process was relatively easy. I was asked to remove some words that were considered a bit excessive for academic discourse. For example, I’d written that the claimed findings “cannot be taken seriously,” and the word “preposterous” somehow found its way into the letter. No matter. The changes were easy enough to make.
The journal has also published a very unsatisfactory response and formal corrigendum from the meta-analysis authors. I have included the text of both below.
**********
My letter to eClinicalMedicine
Comments regarding “Effects of therapeutic interventions on long COVID: a meta-analysis of randomized controlled trials”
David Tuller
A recent article, “Effects of therapeutic interventions on long COVID: a meta-analysis of randomized controlled trials,” reported “high-certainty evidence” that exercise training can improve health and “should be prioritized”.1 I would like to raise some concerns regarding this study that may impact the conclusions drawn by the authors.
First, the dozens of trials in the meta-analysis feature heterogeneous definitions of Long COVID. In some cases, it appears that trial participants did not even have Long COVID. For example, a Brazilian study included “subjects with coronavirus disease 2019 in the acute phase”.2 Pooling together trials with heterogeneous populations, and including results from participants who do not meet definitions of Long COVID, makes it challenging to interpret the meta-analysis results.
Second, the authors themselves determined that the data used in the meta-analysis must be interpreted with caution. They found that, according to Cochrane’s Risk of Bias Tool, only three of the included trials were at “low risk” of bias, with 33–the majority–at “high risk” of bias. Their analysis of quality of evidence for the many outcomes, per the Grading of Recommendations Assessment, Development and Evaluation (GRADE) framework,3 yielded similar results, with “27% of the evidence … rated as very low, 57% as low, 11% as moderate, and 5% as high.”
Given these acknowledged limitations, the confident assertion that exercise training is effective and “should be prioritized” is unwarranted, as is the statement that these findings constitute “high-certainty evidence.”
References
1. Tan, C. ∙ Meng, J. ∙ Dai, X. ∙ et al.
Effects of therapeutic interventions on long COVID: a meta-analysis of randomized controlled trials
eClinicalMedicine. 2025; 87, 103412
2. Rodríguez-Blanco, C. ∙ Bernal-Utrera, C. ∙ Anarte-Lazo, E. ∙ et al.
Breathing exercises versus strength exercises through telerehabilitation in coronavirus disease 2019 patients in the acute phase: a randomized controlled trial
Clin Rehabil. 2022; 36(4):486-497
3.Guyatt, G.H. ∙ Oxman, A.D. ∙ Vist, G.E. ∙ et al.
GRADE: an emerging consensus on rating quality of evidence and strength of recommendations
BMJ. 2008; 336(7650):924-926
**********
The authors’ corrigendum
Corrigendum to ‘Effects of therapeutic interventions on long COVID: a meta-analysis of randomized controlled trials’
Tan et al
Description of the error and impact on results and conclusion
After the publication of this article, a post-publication review revealed that, due to an oversight during the literature screening process, one reference which did not meet the inclusion criteria was erroneously included in the study.
We sincerely regret this error and have taken corrective measures. The non-compliant reference, “Breathing exercises versus strength exercises through telerehabilitation in coronavirus disease 2019 patients in the acute phase: a randomized controlled trial”,1 has been removed, and the meta-analysis has been re-executed accordingly. A total of five meta-analyses in exercise training incorporated this erroneous article. Among them, the four outcomes—6MWT, 30sSTS, VAFS, and MBDS—exhibited no change in either the direction of the pooled effect size or its statistical significance. A change in the direction of the pooled effect size was observed for VE (L/min), although this remained non-significant. These adjustments do not affect the robustness of the article’s overall conclusions, nor do they alter its academic value or clinical relevance.
(After this, the corregendum includes revised calculations in the sections and tables impacted by the removal of the trial.)
**********
This response is inadequate and unsatisfactory. The study I mentioned in my letter was an example of a trial included in the meta-analysis in which the participants did not necessary have Long COVID. However, it wasn’t the only such trial included in the meta-analysis, so their reanalysis is incomplete.
Moreover, the authors have reaffirmed “the robustness” of the article’s conclusions while failing to address a key fact—that most of the trial results were of low quality and/or at high risk of bias. Contrary to what these investigators appear to believe, combining lots of problematic data does not yield trustworthy or robust results but serves mainly to amplify the problems in the data.