LAUSR

We read with great interest the article by Xie and colleagues regarding the impact of Janus kinase inhibitors (Jakinibs) on the risk of cardiovascular events in patients with rheumatoid arthritis (RA). This systematic review and metaanalysis revealed no significant change in cardiovascular risks for Jakinibtreated patients with RA in a shortterm perspective study. However, some methodological issues must be addressed. First, Cochran Q is the weighted sum of squares on a standardised scale, and when it is reported with low p values, it indicates heterogeneity. However, care must be taken when interpreting the χ test results because it has low power in the common situation of a metaanalysis, especially when studies have a small sample size or few cases. While a statistically significant result may indicate a problem with heterogeneity, a nonsignificant result must not be considered evidence of no heterogeneity. Although no definitive cutoff value has been established at which no data pooling should be performed, a p value of 0.10, rather than the conventional level of 0.05, is usually used to determine statistical significance. Second, two statistical models can be used for metaanalyses: the fixedeffects model and the randomeffects model. The selection of the appropriate model is important to ensure that various statistics are estimated correctly. The choice of the metaanalysis method should be explained, and a rationale for choosing the statistical model should also be stated. The fixedeffects model is generally used in the absence of heterogeneity in a metaanalysis that includes a large number of studies, preferably with large sample sizes. If heterogeneity is a concern, the randomeffects model has been advocated for data pooling. A high heterogeneity was detected among the three kinds of Jakinibs for venous thromboembolism events (I=63.5%) and all cardiovascular events (I=71.7%). Considering the substantial and considerable heterogeneity, the randomeffects model would be the more appropriate choice in those cases. A randomeffects metaanalysis may be used to incorporate heterogeneity among studies. However, this is not a substitute for a thorough investigation of heterogeneity. Exploration of possible causes of heterogeneity is always advisable. Thus, we believe that the findings of this study should be interpreted with consideration of the aforementioned methodological concerns.