The search results as well as the decisions made during the eligibility process are displayed in a PRISMA flowchart (Fig. 1). Search strategies yielded 5.527 unique references. After completing the screening, we identified a total of 25 studies –corresponding to 27 articles- involving 771.468 participants (490.082 exposed to VKA and 281.386 to dabigatran) that were eligible for the review (supplementary table III) [15–41].
A total of 142 studies were excluded due to several reasons. See supplementary table IV for details. Additionally, we excluded thirteen studies that reported overlapped data with other included studies (supplementary table V).
Description of the studies
A summary description of included studies is provided in supplementary table III. All the studies reported findings from a cohort study design. Twenty-three studies were retrospective [15–35, 37–39, 41] and two studies were prospective [36, 40]. The source of the data was from an administrative database in thirteen studies [15, 16, 19, 21–23, 25, 30, 34–40] and three used a commercial database [20, 29, 41]. Instead, in nine studies [17, 18, 24, 26–28, 31–33], data were extracted from a national healthcare database.
The studies used different approaches to control for confounding, being the propensity score the method used more frequently in seventeen studies [15–17, 19, 22, 23, 25–27, 30, 31, 33–39, 41]. On the other hand, eight studies performed adjusted analyses according to a Cox analysis [18, 20, 21, 24, 26, 28, 32, 40]. Twenty-five articles provided data about our primary outcome/s, eighteen of which also reported on secondary outcome/s. Two articles [28, 34] provided data only on secondary outcome/s.
Two of the studies (with 9,232 participants) [32, 36] had as an intervention dabigatran only at a dose of 110 mg; fourteen studies (with 154,243 participants) [18–20, 22, 23, 27–31, 34, 35, 39–41] had only dabigatran 150 mg, and six studies [15–17, 21, 25, 26, 33] had both doses (with 38,612 participants at dose of 110mg and 28,098 at dose of 150 mg). Moreover, there were three studies (with 51,201 participants) [24, 37, 38] that provided only unspecified doses (but not 75 mg).
With regard to the comparison group, nineteen of the studies (with 400,060 participants) used warfarin [15, 16, 19–24, 26, 27, 29, 30, 32–37, 39–41] two studies (with 47,477 participants) [25, 38] used phenprocoumon, and in four of the studies (with 47,545 participants) [17, 18, 28, 31] it was not specified which VKA was used.
As for the follow-up, since the vast majority were retrospective analyses, the duration was different for each patient, ranging from 6 months to 3 years on average.
Risk of bias of included studies
All the included studies in the review had an overall moderate risk of bias mainly due to potential confounding (baseline). The risk of bias was summarised at supplementary table VI.
Effects of the intervention
Two effectiveness outcomes were pre-specified as primary outcomes: ischemic stroke and the composite of ischemic stroke/systemic embolism. Dabigatran (either at a dose of 150 mg or 110 mg) may reduce slightly the hazard (or instantaneous risk) to develop an ischemic stroke compared to VKA (8% lower risk at any particular time during the study period as indicated by a HR 0.92, 95% CI 0.84 to 1.02; 19 comparison groups; I2 65) (Fig. 2a). Of note, there was a subgroup effect suggesting the results are affected by the dose. Thus, the reduction observed with dabigatran at a dose of 150 mg was statistically significant (15% lower risk, HR 0.85, 95% CI 0.74 to 0.98; 11 comparisons; I2 61), whereas no difference was found with dabigatran at a dose of 110 mg compared to VKA (HR 0.97, 95% CI 0.85 to 1.12; 6 comparisons; I2 55). Regarding to the composite of ischemic stroke/systemic embolism, dabigatran (either at a dose of 150 mg or 110 mg) may reduce the hazard to develop this outcome compared to VKA (15% lower risk as indicated by a HR 0.85, 95% CI 0.71 to 1.02; 6 comparisons; I2 52) (Fig. 2b). No subgroup effect related with the dose was observed for this outcome. With regard to myocardial infarction, dabigatran 150 mg may reduce slightly, despite non significantly, the hazard to develop this outcome compared to VKA (HR 0.86, 95% CI 0.71 to 1.04; 10 comparisons; I2 49), whereas dabigatran 110 mg did not modify the risk of myocardial infarction compared to VKA (HR 1.02, 95% CI 0.83 to 1.25; 5 comparisons; I2 27) (Fig. 2c). Globally, dabigatran also reduced the hazard of all-cause mortality compared to VKA (25% lower risk as indicated by a HR 0.75, 95% CI 0.67 to 0.85; 15 comparisons; I2 90). Of note, there was a subgroup effect suggesting the comparisons were affected by the dose, with a greater effect observed with dabigatran at a dose of 150 mg (34% lower mortality, HR 0.66, 95% CI 0.58 to 0.74; I2 71), in relation with dabigatran at a dose of 110 mg (HR 0.91, 95% CI 0.79 to 1.04; I2 79) (Fig. 2d).
Regarding the safety outcomes, dabigatran (either at a dose of 150 mg or 110 mg) reduced the hazard to develop a major bleeding compared to VKA (26% lower risk at any particular time during the study period as indicated by a HR 0.74, 95% CI 0.66 to 0.83; 19 comparisons; I2 80) (Fig. 3a) as well as the hazard of an intracranial bleeding (55% lower risk at any particular time during the study period as indicated by a HR 0.45, 95% CI 0.39 to 0.52; 17 comparisons; I2 24) (Fig. 3b). No subgroup effect related with the dose was observed for either of these outcomes. In addition, among the secondary outcomes, dabigatran (either at a dose of 150 mg or 110 mg) may reduce the hazard to develop a fatal bleeding compared to VKA (24% lower risk as indicated by a HR 0.76, 95% CI 0.60 to 0.95; I2 0) (Fig. 3c), based on five comparison groups corresponding to only 3 studies. As for gastrointestinal bleeding, dabigatran (either at a dose of 150 mg or 110 mg) may increase the hazard to develop a gastrointestinal bleeding compared to VKA, although this increase was not significant (HR 1.08, 95% CI 0.99 to 1.18; 23 comparisons; I2 62) (Fig. 3d). No subgroup effect related with the dose was observed. A summary of the effectiveness and safety results is showed in the Table 1. Other secondary outcomes such as systemic embolism and pulmonary embolism are showed in the supplementary material (e-figure I).
Table 1
Effectiveness and safety summary of the results.
| Primary outcomes | Secondary outcomes |
| Ischaemic stroke | Ischaemic stroke/SE | Major bleeding | Intracranial bleeding | Fatal bleeding | GI bleeding | Systemic embolism | Myocardial infarction | Pulmonary embolism | Mortality |
| efficacy | efficacy | safety | safety | safety | safety | efficacy | efficacy | efficacy | |
Num. of studies | 21 | 6 | 20 | 19 | 5 | 26 | 2 | 16 | 1 | 15 |
Globally | Limit ↓ | Limit ↓ | Y ↓ | Y ↓ | Y ↓ | Y ↑ | Y ↓ | Limit | N | Y ↓ |
150 mg | Y ↓ | N | Y ↓ | Y ↓ | N | N | N | Limit | N | Y ↓ |
110 mg | N | Limit ↓ | N | Y ↓ | N | N | Y ↓ | N | NA | Limit |
GI: gastrointestinal; Limit: no difference but close to statistical significance; N: no difference; NA: not assessed; SE: systemic embolism; Y: statistically significant difference; ↓ risk reduction; ↑ increased risk |