A total of 1,886 citations were identified by the search, and 81 potentially eligible articles were retrieved in full text (Fig. 1). Overall, 57 trials were included in the review (Supplementary 2). We excluded from the statistical analysis one study that did not report any data on the pre-defined selected outcomes [31], and six studies [32,33,34,35,36,37] with a sample size < 100 participants. The remaining 50 studies (n = 165,533 patients) were included in our quantitative analysis. Supplementary 2 summarizes the characteristics of the 57 included trials, published between 1978 and 2020. Twenty-eight of 57 (49 %) studies randomized patients with ischemic stroke or TIA as index event, 25 (44 %) with ischemic stroke, and four (7 %) with TIA only. Thirty-nine (68 %) trials included only patients with non-cardioembolic strokes. The included studies assessed the effect of 21 different antiplatelet drugs or combinations. The median follow-up period was 18 months (range, 7 days – 4 years). Overall, 30 (53 %) trials were rated as low risk of bias, 18 (32 %) as moderate, and 9 (16 %) as high risk (Supplementary 3). We also identified four ongoing studies (Supplementary 4).
Figure 2 and Supplementary 5 shows the network geometry for each primary and secondary outcome, respectively. Results from pairwise meta-analysis for each primary and secondary outcome are reported in Supplementary 6. Figure 3 shows the estimates of primary and secondary outcomes of each antiplatelet/combination of antiplatelet drugs against placebo/no treatment from the NMA, with the corresponding confidence in the evidence. The league tables including results from the NMA for each treatment compared to other treatments are reported in the Supplementary excel file.
Thirty-seven RCTs (147,742 participants) addressed all stroke outcome comparing an antiplatelet drug or combination versus placebo/no treatment or another antiplatelet drug/combination. Moderate to high-confidence in evidence indicated that cilostazol, clopidogrel alone and in combination with aspirin, ticagrelor alone and in combination with aspirin, dipyridamole in combination with aspirin, ticlopidine and aspirin both ≤ 150 and > 150 mg/day were significantly associated with the greatest benefits when compared with placebo/no treatment (Fig. 3). OR estimates ranged from 0.51 (cilostazol) to 0.84 (aspirin > 150 mg/day), corresponding to ARD from 3.6 to 1.1 % fewer strokes. When compared with aspirin ≤ 150 mg/day, only clopidogrel (with high confidence), clopidogrel plus aspirin and cilostazol (with moderate confidence) significantly reduced the risk of all strokes (OR range from 0.65 to 0.79, corresponding to an ARD from 2.7 to 1.6 % fewer strokes (Supplementary 7). The common standard deviation heterogeneity estimate was 0.11.
Forty-two RCTs (154,016 participants) addressed all-cause mortality. Only aspirin > 150 mg/day significantly reduced all-cause mortality compared to placebo/no treatment (OR 0.86, 95 % CI 0.76 to 0.97; ARD 0.9 %, 95 % CI 1.5–0.2 % fewer, moderate confidence). Moderate to high-confidence in evidence indicated that ticlopidine and dipyridamole in combination with aspirin could reduce all-cause mortality, with OR estimates below 0.9 versus placebo/no treatment, although not significant (Fig. 3). No drug was significantly more effective than aspirin ≤ 150 mg/day. The common standard deviation heterogeneity estimate was 0.07.
Thirty-nine RCTs (148,414 participants) addressed ischemic stroke. Moderate to high-confidence in evidence indicated that clopidogrel alone and in combination with aspirin, cilostazol, ticagrelor alone and in combination with aspirin, dipyridamole in combination with aspirin, ticlopidine and aspirin both ≤ 150 and > 150 mg/day were significantly associated with the greatest benefits compared with placebo/no treatment (Fig. 3). OR estimates ranged from 0.54 (clopidogrel in combination with aspirin) to 0.77 (aspirin > 150 mg/day), corresponding to ARD from 2.0 to 1.0 % fewer ischemic stroke. In comparison with aspirin ≤ 150 mg/day, we found high confidence in the benefit for clopidogrel in combination with aspirin and moderate confidence for cilostazol and ticagrelor in combination with aspirin (OR range from 0.73 to 0.78, corresponding to ARD from 1.9 to 1.5 % fewer ischemic stroke) (Supplementary 7). The common standard deviation heterogeneity estimate was 0.09.
Thirty-nine RCTs (146,202 participants) addressed cardiovascular events. Moderate to high-confidence in evidence indicated that clopidogrel alone and in combination with aspirin, cilostazol, dipyridamole in combination with aspirin, ticagrelor, ticlopidine and aspirin both ≤ 150 and > 150 mg/day were significantly associated with the greatest benefits versus placebo/no treatment (Fig. 3). OR estimates ranged from 0.64 (clopidogrel in combination with aspirin) to 0.84 (aspirin ≤ 150 mg/day), corresponding to ARD from 3.3 to 1.4 % fewer cardiovascular events. In comparison with aspirin ≤ 150 mg/day, clopidogrel and dipyridamole in combination with aspirin (with high confidence), and cilostazol (with moderate confidence) were significantly more beneficial (OR range from 0.76 to 0.82, corresponding to ARD from 2.3 to 1.7 % fewer cardiovascular event), while indobufen was more harmful (OR 1.86, 95 % CI 1.09 to 3.17, ARD 7.4 %, 95 % CI 0.8–16.5 % more cardiovascular events; high confidence) (Supplementary 7). The common standard deviation heterogeneity estimate was 0.03.
Thirty-one RCTs (130,274 participants) addressed hemorrhagic stroke. No drug was significantly more effective than placebo/no treatment (Fig. 3). High-confidence in evidence indicated that cilostazol significantly reduced hemorrhagic stroke in comparison with aspirin ≤ 150 mg/day (OR 0.42 95 % CI 0.19 to 0.92, ARD 0.2 %, 95 % CI 0.3–0 % fewer hemorrhagic strokes) (Supplementary 7). Conversely, vorapaxar (with high confidence) and ticagrelor in combination with aspirin (with moderate confidence) were significantly associated with the greatest harms when compared with placebo/no treatment. OR estimates ranged from 2.85 (vorapaxar vs. placebo added to standard antiplatelet therapy) to 6.02 (ticagrelor in combination with aspirin), corresponding to ARD from 1.4 to 3.7 % more hemorrhagic strokes) (Fig. 3). When compared with aspirin ≤ 150 mg/day, ticagrelor in combination with aspirin was significantly associated with the greatest harms (OR 4.98, 95 % CI 1.08 to 23.01, ARD 1.6 %, 95 % CI 0–8.4 % more hemorrhagic strokes, with moderate confidence) (Supplementary 7). The common standard deviation heterogeneity estimate was 0.10.
Twenty-seven RCTs (106,309 participants) addressed intracranial hemorrhage. No drug was more beneficial than placebo/no treatment (Fig. 3), while cilostazol (OR 0.30 95 % CI 0.16 to 0.59, ARD 0.5 % 95 % CI 0.7–0.3 % fewer intracranial hemorrhage, high confidence) and clopidogrel (OR 0.62 95 % CI 0.39 to 0.96, ARD 0.3 % 95 % CI 0.5–0 % fewer intracranial hemorrhages, moderate confidence) were significantly more beneficial than aspirin ≤ 150 mg/day (Supplementary 7). Moderate to high-confidence in evidence indicated that vorapaxar, terutroban, aspirin > 150 mg/day and ticagrelor in combination with aspirin were significantly associated with the greatest harms versus placebo/no treatment (Fig. 3). OR estimates ranged between 2.54 (vorapaxar vs. placebo added to standard antiplatelet therapy) to 7.02 (ticagrelor in combination with aspirin), corresponding to ARD from 1.1 to 4.2 % more intracranial hemorrhages. In comparison with aspirin ≤ 150 mg/day, only ticagrelor in combination with aspirin was significantly more harmful (OR 3.32 95 % CI 1.33 to 8.28, ARD 1.8 % 95 % CI 0.3–5.3 % more intracranial hemorrhages; high confidence) (Supplementary 7). The common standard deviation heterogeneity estimate was almost null.
Forty RCTs (146,826 participants) addressed major bleeding. No drug significantly reduced the risk of major bleeding when compared with placebo/no treatment (Fig. 3). High-confidence in evidence indicated that dipyridamole, cilostazol and triflusal significantly reduced the risk of the outcome in comparison with aspirin ≤ 150 mg/day. OR estimates ranged between 0.25 (dipyridamole) to 0.37 (cilostazol and triflusal), corresponding to ARD from 1.0 to 0.9 % fewer events of major bleeding (Supplementary 7). Moderate to high-confidence in evidence indicated that dipyridamole in combination with aspirin, aspirin both ≤ 150 and > 150 mg/day, terutroban, vorapaxar, clopidogrel in combination with aspirin and ticagrelor in combination with aspirin were significantly associated with the greatest harms versus placebo/no treatment. OR estimates ranged from 1.67 (clopidogrel in combination with aspirin) to 7.94 (ticagrelor in combination with aspirin), corresponding to ARD from 0.8 to 7.5 % more events of major bleeding (Fig. 3). In comparison with aspirin ≤ 150 mg/day, clopidogrel in combination with aspirin (OR 1.65 95 % CI 1.15 to 2.37, ARD 0.9 % more 95 % CI 0.2–1.8 % more; high confidence) and ticagrelor in combination with aspirin (OR 3.99 95 % CI 1.56 to 10.22, ARD 3.9 % more 95 % CI 0.8–11.2 % more events of major bleeding; high confidence) were significantly associated with the greatest harms (Supplementary 7). The common standard deviation heterogeneity estimate was 0.23.
We did not perform NMA for ischemic stroke or TIA outcome as it was assessed only by 11 RCTs (17,342 participants) and only two comparisons had at least two studies. This resulted in a disconnected network (see the network plot in Supplementary 5). Results from standard pairwise meta-analysis are available in Supplementary 6.
SUCRA, absolute probability to be the best treatment, and the mean rank are reported in Supplementary 8. Results from the assessment of incoherence are showed in Supplementary 9.
We compared the magnitude of the heterogeneity estimates with the empirical distributions of heterogeneity values derived by Turner [27] for objective (all-cause mortality only) and semi-objective outcomes, for comparisons of treatments versus placebo. Our heterogeneity estimates, except for major bleeding, were below the median of the empirical distributions of heterogeneity values. Therefore, given the small heterogeneity and little incoherence for all outcomes except those for major bleeding, subgroup analyses (stroke subtypes; time from first ischemic event to randomization; and treatment duration) were performed for the major bleeding outcome only (see results in Supplementary excel file). Subgroup analyses for age and gender were not performed because of scarce available data. Results from the sensitivity analyses for primary outcomes are also reported in Supplementary excel file. The comparison-adjusted funnel plots for each outcome are reported in Supplementary 10.