anti-interleukin-6 - versus potential COVID-19 treatments - for COVID-19 severe or critically pdf   xlsx method abbreviations

Outcome Relative effect 95%CI LoD Trt. better when I2 k (RCT/OBS) Bayesian probability Overall ROB Publication bias Degree of certainty Endpoint importance Published MA

efficacy endpoints 00

death D28 1.20 [0.81, 1.78]< 146%8 studies (8/-)17.7 %some concernnot evaluable moderatecrucial-
deaths 0.98 [0.76, 1.28]< 139%13 studies (13/-)55.0 %some concernlow moderatecrucial-
deaths (time to event analysis only) 1.23 [0.86, 1.77]< 137%4 studies (4/-)12.5 %some concernnot evaluable moderatecrucial-
clinical deterioration 0.56 [0.26, 1.20]< 10%1 study (1/-)93.1 %NAnot evaluable important-
clinical improvement 1.36 [1.05, 1.76]> 155%4 studies (4/-)99.0 %some concernnot evaluable moderateimportant-
clinical improvement (14-day) 1.86 [1.20, 2.87]> 10%1 study (1/-)99.7 %NAnot evaluable important-
clinical improvement (28-day) 1.19 [0.81, 1.75]> 10%1 study (1/-)81.0 %NAnot evaluable important-
clinical improvement (time to event analysis only) 1.09 [0.87, 1.35]> 10%2 studies (2/-)76.9 %some concernnot evaluable moderateimportant-
death or ventilation 0.96 [0.65, 1.43]< 126%3 studies (3/-)57.8 %some concernnot evaluable moderateimportant-
hospital discharge 0.95 [0.79, 1.16]> 10%2 studies (2/-)31.3 %some concernnot evaluable moderateimportant-
mechanical ventilation 0.87 [0.47, 1.61]< 132%3 studies (3/-)66.9 %some concernnot evaluable moderateimportant-
ICU admission 1.04 [0.17, 6.33]< 175%2 studies (2/-)48.5 %some concernnot evaluable moderatenon important-
recovery 0.64 [0.14, 2.92]> 10%1 study (1/-)28.3 %NAnot evaluable non important-

safety endpoints 00

serious adverse events 0.99 [0.77, 1.28]< 116%6 studies (6/-)52.7 %some concernserious moderateimportant-
adverse events 1.18 [0.95, 1.48]< 10%5 studies (5/-)7.1 %some concernnot evaluable moderatenon important-

LoD: level of statistical demonstration: Statistically conclusive: statistically significant with a strict control of overall risk of type 1 error (statistically demonstrated), does not take into account the risk of bias; suggested: nominally statistically significant but without a strict control of overall risk of type 1 error; inconclusive: not nominally statistically significant; safety concerns;
Bayesian probability: Bayesian posterior probability of treatment effect (computed with a noninformative prior); ROB: risk of bias; k: number of studies; published MA: number of published meta-analysis on the same topic; degree of certainty adapted from GRADE. Trt. better when: indicates when the relative treatment effect shows that the studied treatment is better than control.