Table 2.
Average treatment effect estimates and probabilities of effects
| Outcome | Effect estimates | Probability of effects with 12 mg dexamethasone | |||||||
|---|---|---|---|---|---|---|---|---|---|
| Dexamethasone 12 mg | Dexamethasone 6 mg | Relative difference | Absolute difference | Any benefit | Any harm | Clinically important benefit | Clinically important harm | No clinically important difference | |
| Primary analyses using weakly informative priors | |||||||||
| Days alive without life support at day 28 | Mean: 17.8 (16.7–19) days | Mean: 16.5 (15.4–17.6) days | IRR: 1.08 (0.98–1.18) |
MD: 1.3 (−0.3 to 2.9) days |
94.2% | 5.8% | 63.9% | 0.3% | 35.9% |
| Serious adverse reactions at day 28 | Prob.: 11.4% (8.9–14.1%) | Prob.: 13.3% (10.7–16.2%) | RR: 0.85 (0.63–1.16) |
RD: −1.9% (−5.7 to 1.9%) |
84.1% | 15.9% | 48.4% | 2.1% | 49.5% |
| Mortality at day 28 | Prob.: 27.5% (24–31.2%) | Prob.: 31.8% (28.1–35.6%) | RR: 0.87 (0.73–1.03) |
RD: −4.3% (−9.4 to 0.9%) |
94.8% | 5.2% | 80.7% | 0.9% | 18.5% |
| Mortality at day 90 | Prob.: 32.5% (28.8–36.3%) | Prob.: 37.1% (33.2–41.0%) | RR: 0.88 (0.75–1.02) |
RD: −4.6% (−10 to 0.9%) |
95.1% | 4.9% | 82.3% | 0.8% | 16.9% |
| Days alive without life support at day 90 | Mean: 59.3 (54.6–64.2) days | Mean: 55.7 (51.1–60.6) days | IRR: 1.06 (0.95–1.2) |
MD: 3.6 (−3.1 to 10.2) days |
85% | 15% | 77.2% | 9.2% | 13.6% |
| Days alive and out of hospital at day 90 | Mean: 44.1 (40.9–47.3) days | Mean: 40.2 (37–43.5) days | IRR: 1.10 (0.99–1.22) |
MD: 3.9 (−0.6 to 8.4) days |
95.7% | 4.3% | 89.7% | 1.5% | 8.8% |
| Pre-specified sensitivity analyses using sceptic priors | |||||||||
| Days alive without life support at day 28 | Mean: 17.8 (16.7–18.9) days | Mean: 16.5 (15.5–17.6) days | IRR: 1.07 (0.99–1.17) |
MD: 1.2 (−0.2 to 2.7) days |
94.9% | 5.1% | 61.8% | 0.2% | 38% |
| Serious adverse reactions at day 28 | Prob.: 12% (9.9–14.3%) | Prob.: 12.7% (10.5–15.1%) | RR: 0.95 (0.79–1.14) |
RD: −0.7% (−3 to 1.6%) |
72.3% | 27.7% | 13.2% | 1% | 85.8% |
| Mortality at day 28 | Prob.: 28.6% (25.5% to 31.8%) | Prob.: 30.7% (27.6% to 34%) | RR: 0.93 (0.82–1.05) |
RD: −2.1% (−5.7 to 1.5%) |
87.2% | 12.8% | 52.3% | 1.2% | 46.4% |
| Mortality at day 90 | Prob.: 33.6% (30.3–37%) | Prob.: 36% (32.6–39.4%) | RR: 0.93 (0.83–1.05) |
RD: −2.4% (−6.3 to 1.5%) |
88.3% | 11.7% | 57.3% | 1.4% | 41.3% |
| Days alive without life support at day 90 | Mean: 59 (54.6–63.7) days | Mean: 56 (51.6–60.7) days | IRR: 1.05 (0.95–1.17) |
MD: 3 (−3 to 9.1) days |
83.6% | 16.4% | 74.4% | 9.7% | 16% |
| Days alive and out of hospital at day 90 | Mean: 43.1 (40.3–46.1) days | Mean: 41.1 (38.2–44.1) days | IRR: 1.05 (0.96–1.14) |
MD: 2 (−1.6 to 5.7) days |
86.2% | 13.8% | 71.1% | 5.1% | 23.8% |
Analyses conducted in all patients with available outcome data (Table 1). All analyses were adjusted for the stratification variables, and effect sizes are presented as average treatment effects as outlined in the methods section, summarised using median posterior values as point estimates and percentile-based 95% credible intervals (CrIs). Data from additional post hoc sensitivity analyses and results estimated for reference patients are presented in Tables S2–S7 in the ESM
Any benefit is the probability of a MD > 0 days (IRR > 1) or a RD < 0 percentage points (RR < 1); any harm is the probability of a MD < 0 days (IRR < 1) or a RD > 0 percentage points (RR > 1); no clinically important difference is the probability of an absolute MD < 1 days or an absolute RD < 2 percentage points; clinically important benefit/harm are probabilities of effect sizes larger than no clinically important difference in either direction. All definitions of clinically important effect sizes were pre-specified in the protocol [10]
ESM electronic supplementary material, IRR incidence rate ratio (> 1 favours 12 mg); MD mean difference (> 0 favours 12 mg), prob. Probability, RD risk difference in percentage points (< 0 favours 12 mg), RR relative risk (< 1 favours 12 mg)