When reading the article by Şahın et al.(1) (Effects of a home-based pulmonary rehabilitation program with and without telecoaching on health-related outcomes in COVID-19 survivors: a randomized controlled clinical study) published in this issue of the Jornal Brasileiro de Pneumologia, we identified elements that could have explored the results better with great clinical implications.
 
Starting from the central question of the research, the results in Table 3(1) showed that there were no major effects on the investigated outcomes between the groups. However, there was an exclusive effect of time, in which case it would be applied to the two groups indifferently, or there would be an effect of time-group interaction, in which case one of the groups would have a different behavior over time.
 
Let’s exemplify: The FVC reveals only the major effect of time, when both groups increased their indicator, but in a large magnitude (Cohen’s d > 0.8), which was not highlighted by the researchers. The same occurs with the six-minute walk distance outcome; the study group has a d = 2.30 and the control group has a d = 2.07. This shows the great clinical effect of pulmonary rehabilitation on FVC in these individuals.
 
The modified Medical Research Council scale outcome has a time-group interaction that needs to be analyzed first. It was observed that the study group evolved better over time than did the control group, with a magnitude of d = 4.51 in the intragroup analysis and d = 2.10 in the intergroup analysis, that is, telecoaching clinically enhanced this outcome. This also occurred with the social aspects when comparing the study and the control groups (d = 5.88 vs. d = 2.14), a clinical effect almost two times greater (d = 1.83). The isolated interpretation of the partial eta only allows measuring the explanatory power of the built model and not the specific effects of the factors that Cohen’s d allows for balanced groups.(1)
 
These interesting findings reveal inconsistencies identified in the measures of the standard deviations of the groups presented in Table 3 and the distribution of the groups in Figure 2 in the study by Şahın et al.(1) Table 3(1) shows that the standard deviations of the study and control groups were the same both before and after the intervention for almost all outcomes.
 
This is minimally odd for interventions when individual variability follows distinct progressions. In figure 2,(1) on the other hand, the outcomes six-minute walk distance, modified Medical Research Council scale score, and perceived dyspnea and fatigue reveal distinct variability, which may lead to the invalidation of the application of factorial ANOVA.(2) It is suggested that the authors make explicit the real variability of the outcomes in order to obtain accurate values for the measures of clinical effect.
 
Finally, we recommend a data analysis using a generalizable mixed model in order to minimize independence biases of residues of repeated measures and the heterogeneity of variance that are apparent in the published results.(3)
 
REFERENCES
 
1.            Şahın H, Naz İ, Karadeniz G, Süneçlı O, Polat G, Ediboğlu O. Effects of a home-based pulmonary rehabilitation program with and without telecoaching on health-related outcomes in COVID-19 survivors: a randomized controlled clinical study. J Bras Pneumol. 2023;49(1):e20220107. https://doi.org/10.36416/1806-3756/e20220107
2.            Lakens D. Calculating and reporting effect sizes to facilitate cumulative science: a practical primer for t-tests and ANOVAs. Front Psychol. 2013;4:863. https://doi.org/10.3389/fpsyg.2013.00863
3.            Guimarães LSP, Hirakata VN. Use of the Generalized Estimating Equation Model in longitudinal data analysis. Rev HCPA. 2012;32(4):503-11.