Statistical model error

Abstract

I thoroughly enjoyed reading a recently published study entitled “A systematic review and meta-analysis of murine models of uremic cardiomyopathy” published in Kidney International.1Soppert J. Frisch J. Wirth J. et al.A systematic review and meta-analysis of murine models of uremic cardiomyopathy.Kidney Int. 2022; 101: 256-273Abstract Full Text Full Text PDF PubMed Scopus (7) Google Scholar In this meta-analysis, the authors did not consider the significance of homogeneity of their numerical data in forest plots demonstrated in Figures 6a,b and 7b. As described by Higgins et al.,2Higgins J.P. Thompson S.G. Deeks J.J. Altman D.G. Measuring inconsistency in meta-analyses.BMJ. 2003; 327: 557-560Crossref PubMed Scopus (40900) Google Scholar data sets with I2 equal to 0%–25% are considered to have a very low heterogeneity, and data sets with I2 equal to 25%–50% are deemed to have low heterogeneity. The appropriate model to analyze data sets with low heterogeneity is the fixed-effects model, whereas the random-effects model is adopted when the data are deemed to be moderately to highly heterogeneous.3Dettori J.R. Norvell D.C. Chapman J.R. Fixed-effect vs random-effects models for meta-analysis: 3 points to consider.Global Spine J. 2022; 12: 1624-1626Crossref Scopus (8) Google Scholar In the analysis demonstrated in Figures 6a,b and 7b, I2 is equal to 0%, 36%, and 0%, respectively. However, despite the low heterogeneity of the data, the authors have used the random-effects model. The same error occurred another time at a different level of analysis in Figure 7b, when the authors compared all studies from different intervals. Even though I2 was equal to 32%, the random-effects model was used, which makes the results of this analysis less reliable, as the error repeatedly has occurred in the same assay. In conclusion, the authors must use the fixed-effects model for the analysis demonstrated in Figures 6a,b and 7b. A systematic review and meta-analysis of murine models of uremic cardiomyopathyKidney InternationalVol. 101Issue 2PreviewChronic kidney disease (CKD) triggers the risk of developing uremic cardiomyopathy as characterized by cardiac hypertrophy, fibrosis and functional impairment. Traditionally, animal studies are used to reveal the underlying pathological mechanism, although variable CKD models, mouse strains and readouts may reveal diverse results. Here, we systematically reviewed 88 studies and performed meta-analyses of 52 to support finding suitable animal models for future experimental studies on pathological kidney-heart crosstalk during uremic cardiomyopathy. Full-Text PDF Open AccessThe authors replyKidney InternationalVol. 103Issue 6PreviewPatients with chronic kidney disease (CKD) suffer from uremic cardiomyopathy, and animal models could help to understand pathologic kidney-heart crosstalk. Thus, we performed meta-analyses of the effect of CKD on different cardiac outcomes in mice.1 Ahmadi-Hadad2 raised the concern that we might have chosen an incorrect statistical model in some of our analyses and favored the use of a fixed-effects model over the random-effects model for data sets with low heterogeneity. The author referred to Figures 6a,b and 7b of our paper;1 however, it needs to be noted that Figure 6a and b of our paper1 have an overall data heterogeneity of 83% and 72%, respectively, with for each only 1 subgroup presenting with low heterogeneity. Full-Text PDF