TABLE 3.
Evaluation of bias and heterogeneity of 34 primary comparisons evaluating systematic corticosteroid use on COVID-19.
| Publication | Outcomes | No. of studies | I2 (95% CI)* | p heterogeneity* | Egger’s p value † | SD largest | SD fixed | SD random | Observed ‡ | E largest ‡ | E fixed ‡ | E random ‡ | p value § |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Budhathoki et al. (2020) (Nepal) | Mortality | 14 | 88.8% (89.1–94.8%) | p < 0.001 | 0.815 | 43.38 | 11.4 | 42.72 | 9 | 13.99 | 14 | 6.426 | — |
| Discharge rate | 9 | 62.0% (21.0–81.0%) | p = 0.007 | 0.771 | 1.136 | 1.332 | 3.424 | 4 | 8.751 | 8.991 | 6.5 | — | |
| Hospitalization | 4 | 80.0% (48.0–93.0%) | p = 0.002 | 0.772 | 329.58 | 5.802 | 22.935 | 1 | 0.304 | 1.357 | 0.29 | — | |
| Clinical improvement | 9 | 76.0% (55.0–88.0%) | p < 0.001 | 0.317 | 11.443 | 1.032 | 1.032 | 8 | 3.681 | 9 | 9 | — | |
| MV | 6 | 88.0% (77.0–93.0%) | p < 0.001 | 0.3 | 200.381 | 5.137 | 18.347 | 3 | 4.572 | 4.128 | 0.756 | — | |
| Hospital stay | 4 | 70.1% (14.0–89.6%) | p = 0.018 | 0.661 | 21.043 | 20.911 | 43.154 | 4 | 4 | 4 | 3.996 | — | |
| Negative conversion | 4 | 14.4% (0–86.9%) | p = 0.320 | 0.4781 | 19.317 | 17.046 | 20.578 | 2 | 3.876 | 3.976 | 3.872 | — | |
| Cano et al. (2020) (America) | Mortality | 27 | 89.0% (86.0–92.0%) | p < 0.001 | 0.03897 | 4.294 | 5.024 | 28.17 | 12 | 22.0914 | 21.978 | 20.614 | — |
| Cheng et al. (2020) (China) | Clinical improvement | 1 | — | — | — | — | — | — | — | — | — | — | — |
| Mortality | 6 | 96.0% (93.0%-0.97) | p < 0.001 | 0.5223 | 43.38 | 10.015 | 36.474 | 4 | 5.994 | 5.994 | 0.63 | — | |
| VCT | 5 | 91.4% (82.9–95.7%) | p < 0.001 | 0.9977 | 17.533 | 5.653 | 22.372 | 2 | 0.395 | 4.995 | 1.84 | — | |
| MV | 1 | — | — | — | — | — | — | — | — | ||||
| Hospital stay | 3 | 98.7% (97.7–99.2%) | p < 0.001 | 0.369 | 1.673 | 6.386 | 25.631 | 3 | 3 | 3 | 1.617 | — | |
| ICU stay | 1 | — | — | — | — | — | — | — | — | ||||
| Lu et al. (2020) (China) | Mortality | 4 | 91.0% (80.0–96.0%) | p < 0.001 | 0.6228 | 50.001 | 7.756 | 35.043 | 2 | 2.557 | 3.88 | 0.748 | — |
| Duration of fever | 1 | — | — | — | — | — | — | — | — | — | — | — | |
| Lung inflammation absorption | 1 | — | — | — | — | — | — | — | — | — | — | — | |
| Hospital stay | 2 | 0 | p = 0.603 | — | 10.201 | 9.796 | 9.796 | 2 | 1.8 | 1.998 | 1.998 | 0.112 | |
| Pasin et al. (2020) (multiple nations) | Mortality | 5 | 0 (0–69.6%) | p = 0.603 | 0.3308 | 3.476 | 4.475 | 4.475 | 1 | 3.99 | 3.389 | 3.389 | — |
| MV | 3 | 0 (0–86.3%) | p = 0.469 | 0.6173 | 6.384 | 4.41 | 3.643 | 0 | 2.802 | 2.826 | 0.995 | — | |
| Pei et al. (2020) (China) | Mortality | 5 | 61.9% (0–85.7%) | p = 0.033 | 0.1141 | 15.413 | 13.004 | 20.838 | 4 | 2.518 | 4.914 | 3.301 | — |
| Sarkar et al. (2020) (India) | Mortality | 12 | 96.3% (94.8–97.4%) | p < 0.001 | 0.3671 | 4.294 | 5.763 | 52.191 | 9 | 9.818 | 12 | 6.226 | — |
| Hospital stay | 6 | 92.5% (86.5–95.9%) | p < 0.001 | 0.515 | 21.043 | 15.334 | 32.134 | 4 | 6 | 5.994 | 3.761 | — | |
| VCT | 2 | 0 | p = 0.517 | — | 19.729 | 12.194 | 12.194 | 0 | 0.726 | 0.8 | 0.8 | — | |
| Siemieniuk et al. (2020) (multiple nations) | Mortality | 1 | — | — | — | — | — | — | — | — | — | — | — |
| MV | 1 | — | — | — | — | — | — | — | — | — | — | — | |
| Hospital stay | 1 | — | — | — | — | — | — | — | — | — | — | — | |
| Sterne et al. (2020) (multiple nations) | Mortality | 7 | 30.9% (0–70.5%) | p = 0.192 | 0.3292 | 2.59 | 2.211 | 3.263 | 1 | 6.972 | 6.937 | 5.363 | — |
| AEs | 6 | 17.4% (39.3–88.4%) | p = 0.691 | 0.4461 | 3.837 | 2.557 | 5.586 | 0 | 4.416 | 3.401 | 1.584 | — | |
| Tlayjeh et al. (2020) (Arabia) | Mortality | 10 | 82.2% (70.7–89.2%) | p < 0.001 | 0.884 | 3.734 | 4.159 | 13.37 | 7 | 9.803 | 9.493 | 1.898 | — |
| MV | 3 | 74.5% (15.1–92.3%) | p = 0.020 | 0.498 | — | 5.231 | 11.431 | 2 | — | 1.588 | 2.362 | — | |
| van Paassen et al. (2020) (Netherlands) | Mortality | 22 | 60.6% (39.0–74.5%) | p < 0.001 | 0.29 | — | 4.558 | 11.546 | 7 | — | 20.966 | 4.422 | — |
| MV | 7 | 10.9% (0–74.0%) | p = 0.346 | 0.591 | — | 2.658 | 2.892 | 1 | — | 6.709 | 6.495 | — | |
| Ye et al. (2020) (multiple nations) | Mortality | 2 | 0 | p = 0.768 | — | — | 19.911 | 19.911 | 2 | — | 0.6266 | 0.6266 | 0.099 |
Abbreviations: AEs, adverse events; 95% CI, 95% confidence interval; ICU, intensive care unit; MV, mechanical ventilation; NA, not available; SD, standard difference; VCT, virus clearance time.
I2 of the heterogeneity test, p value for the Q test.
From Egger’s regression asymmetry test.
Observed and expected number of significant studies using effects from the largest study (most with the smallest SE), fixed-effects results, random-effects results of each included meta-analysis as plausible effect size.
Two-sided p value of the excess significance test on intended outcomes. Hospital stay in the study of Lu et al. and mortality in the study of Ye et al. had O > E results, and the p value was calculated.