Table 4.
GRADE table for adverse events, aspirin vs. control
| Certainty assessment | No. of events/patients (% experienced) | Effect of aspirin vs. control | Confidence in efficacy statement | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| No. of studies | Study design | Risk of bias | Inconsistency | Indirectness | Imprecision | Other | Aspirin | No aspirin | Relative (95% CI) | Absolute (95% CI) | Efficacy statement | |
| Adverse events—any | ||||||||||||
| 1 | RCT | Not seriousa | Not serious | Seriousb | Not serious | None | 82/156 (52.6%) | 57/154 (37.0%) | OR 1.89 (1.20 to 2.97) | 156 more per 1000 (from 43 more to 266 more) | Aspirin increases risk of adverse events |
⨁⨁⨁◯ Moderate |
| Adverse events—serious | ||||||||||||
| 1 | RCT | Seriousa | Not serious | Seriousb | Very seriousc | None | 63/156 (40.4%) | 52/154 (33.8%) | OR 1.33 (0.84 to 2.11) | 66 more per 1000 (from 38 fewer to 181 more) | Aspirin has no significant effect on the risk of serious adverse events |
⨁◯◯◯ Very low |
| Adverse events—severe bleeding | ||||||||||||
| 1 | RCT | Not seriousa | not serious | Seriousb | Not serious | None | 13/156 (8.3%) | 2/154 (1.3%) | OR 6.91 (1.53 to 31.15) | 70 more per 1000 (from 7 more to 278 more) | Aspirin increases risk for severe bleeding |
⨁⨁⨁◯ Moderate |
| Adverse events—intracranial haemorrhage | ||||||||||||
| 3d | Non-RCT | Seriousd | Not serious | Seriouse | Not applicablef | None |
Pooled data from trials: aspirin group 7/221 (3.2%), control group 0/212 (0%) OR 14.86 (0.83 to 250.43) 30 more per 1000 (0.2 less to 353 more)g Registry data (after adjustments and competing risk correction): compared to no dementia/no aspirin—aspirin group HR 2.22 (1.07–4.62), control group HR 2.02 (1.10–3.72) Est. numbers of ICH—aspirin 9 (4–19)/3476 person-years, control group 11 (6–20)/4452 person-years Est. absolute difference—1 more per 1000 people (15 less to 20 more) |
Aspirin has no statistically significant effect on the risk of intracranial haemorrhage |
⨁◯◯◯ Very low |
|||
CI, confidence interval; OR, odds ratio
aCochrane risk of bias tool highlighted high levels of deviation from intended treatment (decreases confidence in null findings but not positive findings as intention-to-treat analysis used)
bIncludes only those with Alzheimer’s disease and without high vascular risk
cConfidence interval crosses lines of clinical importance on both benefit and harm
dIncludes pooled randomised and non-randomised data (Thoonsen et al.), primarily randomised from AD2000 (risk of bias felt possible from deviation from intended treatment) and registry study (Lee et al., observational but low formal risk of bias)
eIncludes only those with Alzheimer’s disease
fStudies are not combined; however, it is noted that estimates from one study cross the line of clinically significant harm, and the other study crosses the lines of both significant harm and significant benefit
gEffect rare and calculation of absolute effect confidence intervals are based on estimates of background rate based on 1:1000 over 2 years
fAs calculated by Cochrane Review Manager