Table 2.
GRADE classification of main outcomes considering the different studies that contributed to the compiled effect estimate
| Outcome | No. of participants (studies) | Risk of bias1 | Inconsistency2 | Indirectness | Imprecision3 | Other considerations | Quality of the evidence (GRADE) | Events in acyclovir group | Events in control group | Relative risk (95% CI) | Anticipated absolute effects | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Risk without acyclovir treatment | Risk difference with acyclovir treatment (95% CI) | |||||||||||
| Hospital all-cause mortality | 831 (8) | Serious | Not serious | Not serious | Not serious | All plausible residual confounding would reduce the demonstrated effect |
⨁⨁◯◯ Low |
189/465 (40.6%) | 193/366 (52.7%) | 0.74 (0.64; 0.85) | 527 per 1.000 | 137 fewer per 1.000 (from 190 fewer to 79 fewer) |
| 30-day all-cause mortality | 633 (3) | Serious | Not serious | Not serious | Serious | None |
⨁◯◯◯ Very low |
120/361 (33.2%) | 112/272 (41.2%) | 0.75 (0.59; 0.94) | 412 per 1.000 | 103 fewer per 1.000 (from 169 fewer to 25 fewer) |
| ICU all-cause mortality | 629 (4) | Serious | Not serious | Not serious | Serious | None |
⨁◯◯◯ Very low |
141/368 (38.3%) | 112/261 (42.9%) | 0.73 (0.51; 1.05) | 429 per 1.000 | 116 fewer per 1.000 (from 210 fewer to 21 more) |
CI confidence interval, ICU intensive care unit
1Risk of bias was high in all but one study due to the non-randomized study design
2Inconsistency (heterogeneity) was judged to be not serious when heterogeneity was low or moderate. Each issue judged as bearing a serious potential impact on the assessed features and rated as having a serious risk to the quality of evidence was downgraded by one level and, in the case of risk of bias, by two levels due to the high risk of bias
3Imprecision was assessed calculating the optimal information size (OIS) (α = 0.05; β = 0.1 and power 90%)
Link: https://gradepro.org