We identified 21,857 records in the literature search and excluded 4,797 duplicates (Figure 1). Of the remaining 17,060, we excluded 16,805 by reviewing the title, abstract or both. Of 255 remaining records, 228 were excluded by reviewing the body text. We included 27 studies in our qualitative assessment, (Supplementary Table 2). All studies were retrospective: 11 compared PWID vs non-PWID, and 16 only reported data for PWID. The majority of studies were from the US (15), followed by Germany (3), Canada (2), Spain (2), Sweden (2), England (1), Italy (1), and Switzerland (1). We recreated eIPD from 13 studies that published Kaplan-Meier curves using the Guyot algorithm and created Kaplan-Meier curves from five studies that published an IPD table (Carrell, Frater, Hubbell, Mammana, and Shetty) and used these data for our primary outcome analysis. Three studies provided survival curves for time-to-reoperation and were used for our secondary outcome analysis. Maximum follow-up ranged from 52 days to 29 years. Data was reported for 926 PWID and 1,822 non-PWID. Patients in both PWID and non-PWID groups were majority male, 68.2% and 69.1%, respectively. PWID were younger than non-PWID (mean age, 34.9 years, 95% CI 32.4-37.7, vs. 51.4 years, 95% CI 46.9-56.3, p<0.001). Table 1 describes the characteristics for the included studies.
In total, 2,141 microbiologic pathogens were reported, 738 in PWID from 24 cohorts, and 1,403 in non-PWID from nine cohorts. Staphylococcus aureus (43.0% vs. 24.7%, p=0.001) was more common in PWID than in non-PWID. Streptococci (29.4% vs. 16.7%, p<0.001), coagulase-negative Staphylococci (13.4% vs. 3.7%, p<0.01), Enterococci (12.1% vs. 7.3%, p<0.01), and culture negative endocarditis (9.6% vs. 5.8%, p=0.01) were more common in non-PWID than in PWID. Summarized microbiologic data are presented in Table 2, and by study microbiologic data are available in Table 3 of the Supplementary Appendix.
The number and type of affected valves were reported by all studies. Two papers (Thalme and Asgerisson) combined valve-related data of those in whom surgery was and was not performed and were not included in aggregated totals. Embolic events were reported for PWID in 10 studies, and non-PWID in six studies. Valve data are presented in Table 2. A total of 2,874 valves were included: 922 in PWID and 1,952 in non-PWID. Although tricuspid surgical procedures (33.5% vs 9.6%) were more common in PWID compared to non-PWID, this difference was not significantly different (p=0.53). Mitral (36.8% vs. 25.6%) and aortic (53.1% vs. 39.7%) procedures were more common in non-PWID compared to PWID, but again, were not significantly different between PWID and non-PWID (p=0.67 and p=0.85, respectively). Prosthetic valve endocarditis was more common in non-PWID than PWID (30.2 vs. 7.9%, p<0.01). Valve data is broken out by study in Table 4 of the Supplementary Appendix.
Outcomes in PWID and non-PWID
In one-step random effects meta-analysis of mortality with eIPD from 13 studies and IPD for 5 studies, we included data for 649 PWID and 1,578 non-PWID. In PWID, survival was 94.3%, 81.0%, 62.1% and 56.6% at 30-days, one-, five-, and ten-years, respectively. In non-PWID, survival was 96.4%, 85.0%, 70.3%, and 63.4% at 30-days, one-, five-, and ten-years, respectively (Supplementary Table 5). In the mixed effects Cox Proportional Hazards model, the hazard ratio (HR) for PWID was 1.47 (95% CI 1.05-2.05, p=0.02) compared to non-PWID thus non-PWID survive significantly longer than PWID after valvular surgery for IE. Pooled survival curves are presented in Figure 2. Survival curves by study are presented in Supplementary Figure 1. Survival curves for reoperation were reported by three studies. We estimated IPD for 183 PWID and 986 non-PWID. PWID had a higher hazard of reoperation than non-PWID (HR 2.37, 95% CI, 1.25-4.50, p<0.01). In PWID, median survival to reoperation was 78.1 months (Supplementary Figure 3). Funnel plots to visually assess for publication bias were reported in Supplementary Figures 2 and 4, for the mortality and reoperation outcomes, respectively.