Reply to Tleyjeh

Clinical Infectious Diseases, vol. 72, pp. e1155–e1156

Abstract

TO THE EDITOR—We thank Professor Tleyjeh for his letter [1] and we appreciate the opportunity to address the concerns that may have risen regarding our paper [2]. Tleyjeh expressed concerns regarding the use of unadjusted data in the meta-analysis. We discussed this issue in our paper and also provided the individual adjusted estimates of the studies that reported them. All these adjusted estimates confirmed the lower mortality for the Tocilizumab (TCZ) group: hazard ratio, 0.035 [3]; hazard ratio, 0.38 [4]; hazard ratio, 0.58 [5]; and odds ratio, 0.78 [6]. The pooled estimate of these adjusted analyses is .39 (95% confidence interval [CI] .21–.76), which is similar to the pooled estimate of crude data and does not change the conclusion: Risk Ratio (RR), .27 (95% CI, .12–.59). Tleyjeh expressed concerns regarding inclusion of studies in the meta-analysis that, in his view, are at critical risk of bias. Indeed, we did not exclude studies based on risk-of-bias assessment. This approach allowed us to explore the impact of risk of bias on the results in sensitivity analyses and also gave us sufficient studies to explore other possible sources of heterogeneity (eg, co-medication). As indicated in our paper, the results remained the same when restricted to studies with the highest quality. Similarly, when we re-do the analysis for the studies Tleyjeh indicated to be of moderate risk of bias [4–6], the pooled RR is .25 (95% CI, .063–1.02), which is very similar to the RR of .27 (95% CI, .12–.59) of the full analysis. It should also be mentioned that the reliability is low for such risk-of-bias assessment tools, so different reviewers may classify studies differently, which could be problematic when studies are excluded based on their presumed risk of bias [7]. Tleyjeh expressed concerns regarding a possible effect of “immortal time bias” on the pooled estimate. We agree that “immortal time bias” is an important potential source of bias. Therefore, we re-evaluated the included papers for differences in duration of symptoms or start of mechanical ventilation [5] between the TCZ and control group. This difference could serve as a landmark [8]. Patients in the control group who died before the landmark, as is visible in the Kaplan-Meier and other figures, were excluded from the analysis. Comparing the results from the landmark method with the non–landmark method on the same set of studies [3–5, 9] gave similar results and did not change the conclusion: a pooled RR of .12 (95% CI, .05–.27) for the landmark method and a pooled RR of .094 (95% CI, .03–.31) for the non–landmark method. Although there are apparent limitations to this method, it does provide some insight regarding this bias. The analyses of our rapid review are based on the literature until the end of June 2020 and indicated that patients with coronavirus disease 2019 (COVID-19) may benefit from TCZ treatment based on observational studies and that this effect appears to be modified by co-medication. Acknowledging the limitations of observational studies, as we did in our paper, the published results of randomized controlled trials are awaited to provide more definitive answers as to whether or not TCZ could benefit patients with COVID-19. Potential conflicts of interest. The author: No reported conflicts of interest. The author has submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

Authors 3

1. Jishnu Malgie Aachen Department of Orthopedics, Leiden University Medical Center

   Leiden University Medical Center

   Affiliation as printed

   Department of Orthopedics, Leiden University Medical Center, Leiden, The Netherlands

2. Jan W. Schoones Aachen Walaeus Library

   Leiden University · Leiden University Medical Center

   Affiliation as printed

   Walaeus Library, Leiden University Medical Centre, Leiden, The Netherlands

3. Bart G. C. W. Pijls corresponding Aachen Department of Orthopedics, Leiden University Medical Center

   Leiden University Medical Center

   Affiliation as printed

   Department of Orthopedics, Leiden University Medical Center, Leiden, The Netherlands