This study was performed in accordance with PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines (table 1 of the supplementary data). We searched the PubMed/Medline (National Library of Medicine, United States) and EMBASE (Elsevier, Netherlands) databases from inception to October 3, 2023, to identify longitudinal studies comparing the incidence of mortality in adult ambulatory patients with CCC and OC. No language restrictions were applied. We used the following search terms: Chagas cardiomyopathy, Chagas disease, Trypanosoma cruzi, American trypanosomiasis, mortality, and outcomes, among others. The complete search strategy is described in the supplementary data.

We included clinical trials and all observational studies (eg, cross-sectional, cohort, and case-control studies), except for case reports and case-series. We also excluded systematic reviews and meta-analyses. Studies that assessed adult patients with CCC and compared the mortality risk with those with OC were considered. The follow-up of included studies also had to be performed in an outpatient setting. We excluded studies in the pediatric population or animals.

The selection process was carried out using Rayyan.6 Two independent reviewers screened the titles and abstracts, while considering the selection criteria. Any disagreements were resolved by a third reviewer. After this stage, full texts were reviewed to determine whether each study fulfilled the selection criteria. Relevant data from included studies was extracted using an Excel form. For studies reporting only medians and ranges (interquartile range, range, and maximum-minimum values), these values were converted into means and standard deviations using the method explained by Hozo, et al.7

The study quality was independently evaluated by 2 authors employing the Newcastle-Ottawa Scale. In instances where a consensus was elusive, a third author arbitrated to reach a resolution. The quality of each study was analyzed on a 10-point scale, stratified into “low risk of bias” for 9 to 10 points, “medium risk of bias” for 6 to 8 points, and “high risk of bias” for scores less than 6.

The summary measures for continuous variables are reported as the mean±standard deviation, while categorical variables are expressed as proportions. Participants were classified into 4 groups, initially: a) chronic Chagas cardiomyopathy (CCC); b) nonischemic cardiomyopathy (NICM), and c) ischemic cardiomyopathy (ICM). In addition, a fourth non-Chagas cardiomyopathy (non-CC) group was included for studies that did not characterize individuals in the comparator population (without Chagas cardiomyopathy) with respect to their etiology. Data synthesis was conducted by employing random effect models using the inverse variance method to estimate the pooled size effects, while the Paule-Mandel estimator was used to account for random error. In addition, the Hartung-Knapp adjustment was applied. Pooled unadjusted risk ratios (RR) were calculated from the mortality information reported, while adjusted hazard ratios (HR) were directly extracted from the multivariate models of the studies. Heterogeneity was assessed using the I2 statistic. Value higher than 75% were considered to indicate high heterogeneity, those between 25% and 75% as moderate, and those less than 25% as low. Meta-regression analysis was performed in contrasts with more than 10 studies to assess variables that might explain heterogeneity in the observed associations. A meta-regression analysis was also performed in contrasts with more than 10 studies to assess variables that might explain the heterogeneity in the observed associations, while a sensitivity analysis comparing studies published before and after 2016 was performed to explore the potential influence of the use of the more recently introduced angiotensin receptor-neprilysin inhibitors. Finally, publication bias was assessed using funnel plots and Egger's test. If potential publication bias was identified, the trim-and-fill method was employed using the metafor package to calculate an adjusted effect size, accounting for the presence of this bias. Statistical significance was set at a P<.05. All analyses were conducted using R Statistical Software (v4.2.3; R Core Team 2023) and the meta and metafor packages.

Out of 2134 screened studies, 37 met the selection criteria and were included in the meta-analysis (figure 1 and table 1).3,8–43 These studies were performed predominantly in Brazil (n=32, 86.5%) and were primarily prospective cohorts (64.9%). A total of 17 949 patients (4258 with CCC and 13 691 with OC) were analyzed. CCC patients were younger (mean difference, −2.35 years; 95% confidence interval [95%CI], −4.05 to −.65 years; I2, 93%) and less frequently male (odds ratio [OR], 0.74; 95%CI, 0.63-0.87; I2, 40%) compared with patients with OC, with no significant differences in the proportion of patients in New York Heart Association class III-IV status (OR, 0.88; 95%CI, 0.70-1.09; I2, 44.3%) or left ventricular ejection fraction (mean difference, 0.77%; 95%CI, −0.79% to 2.33%; I2, 90%). The mean follow-up times in the assessed studies ranged from 100 days to 1970 days (median, 758 days). Finally, most of the studies (86.5%) were classified as having a low risk of bias (figure 2A,B).

Figure 1.

Central illustration. Flowchart of included studies.

(0.8MB).

Figure 2.

Forest plots comparing mortality risk between CCC and heart failure of other etiologies. A: unadjusted RR for CCC vs NICM, ICM, and non-CC. B: adjusted HR derived from multivariable Cox proportional-hazards models for CCC vs NICM, ICM, and non-CC. 95%CI, 95% confidence interval; CCC, chronic Chagas cardiomyopathy; CM; cardiomyopathy; HR, hazard ratio; ICM, ischemic cardiomyopathy; LVEF, left ventricular ejection fraction; NICM, nonischemic cardiomyopathy; non-CC, nonchagasic cardiomyopathy; NOS, Newcastle-Ottawa Scale; RR, risk ratio. The bibliographic references cited in the figure correspond to: Issa et al.,3 Areosa et al.,8 Ayub-Ferreira et al.,9 Barbosa et al.,11 Barbosa et al.,10 Bertolino et al.,12 Bestetti et al.,13 Bestetti et al.,14 Bradfield et al.,15 Braga et al.,16 Cardoso et al.,17 Cerqueira-Silva et al.,18 de Albuquerque et al.,19 De Campos Lopes et al.,20 de Melo et al.,21 Oliveira Jr et al.,22 Dubner et al.,23 Echeverría et al.,24 Femenía et al.,25 Freitas et al.,26 Freitas et al.,27 Heringer-Walther et al.,28 Martinelli et al.,29 Nadruz et al.,30 Nakazone et al.,31 Nunes et al.,32 Nunes et al.,33 Ochiai et al.,34 Olivera et al.,35 Passos et al.,36 Pereira et al.,37 Rassi et al.,38 Shen et al.,39Silva et al.,40Traina et al.,41Vilas-Boas et al.,42 Ferreira et al.43

(0.99MB).

In the unadjusted analyses, CCC patients had a significantly higher risk of mortality during follow-up compared with patients with NICM (RR, 1.44; 95%CI, 1.21-1.71; I2, 65%; 12 studies), ICM (RR, 1.34; 95%CI; 1.11-1.63; I2, 65%; 10 studies), and non-CC (RR, 1.42; 95%CI, 1.30-1.55; I2, 37%; 22 studies) (figure 2A). Twenty studies provided adjusted effect measures for the risk of mortality, mainly including age, sex, New York Heart Association class, left ventricular ejection fraction, and HF medications as adjustment covariates. The meta-analysis of adjusted measures reported a significantly higher mortality risk in the CCC group compared with the NICM (HR, 2.04; 95%CI, 1.60-2.60; I2, 47%; 8 studies) and the non-CC (HR 2.26; 95%CI, 1.65-3.10; I2, 71%; 11 studies) groups, while no significant effect was observed when patients with CCC and ICM were compared (HR, 1.72; 95%CI, 0.80-3.66; I2, 69%; 4 studies) (figure 2B). Of note, some studies reported information for more than one comparator group (ICM /NICM/non-CC), and consequently the total number of contrasts performed was higher than the number of articles included.

The results of meta-regression analysis revealed that none of the evaluated study characteristics (age, sex, New York Heart Association class, left ventricular ejection fraction, publication year, and follow-up time) was a significant source of heterogeneity (P>.05) (table 2 of the supplementary data). Moreover, a sensitivity analysis by publication year (before 2016 and after 2016) showed no significant differences in the effects between groups among the different contrasts (table 3 of the supplementary data). At the same time, Egger's test did not suggest publication bias in most of the performed analyses (P>.05), except for the unadjusted contrast between CCC and non-CC patients, which showed a potential bias toward reporting of larger effects (P=.024) (Figure 1 of the supplementary data). Nonetheless, after implementing the trim-and-fill method for this contrast, the adjusted estimates showed a consistently higher risk of mortality in the CCC group (RR, 1.39; 95%CI, 1.26-1.52; I2, 44%) (Figure 2 of the supplementary data).

Despite its comprehensiveness, our study has several limitations. First, the moderate to high heterogeneity among the included studies in some comparisons, including various populations, methodologies, and clinical settings, could have introduced potential biases. Although difficult to determine, the observed interstudy heterogeneity may be due to multiple factors. These factors include the year of study publication. Although the year of publication was not significantly associated with the impact of CCC diagnosis on mortality in the meta-regression analyses, the relatively low sensitivity of this approach does not allow us to rule out the presence of a significant effect. This is because the diagnostic and therapeutic approaches to HF varied significantly between the publication of the first included study (1997) and the last study (2023). Other important factors include the type and stage of cardiac involvement, since some studies did not clarify the etiologies of the comparator groups or whether the individuals included had a diagnosis of HF or were in earlier stages of cardiomyopathy, as well as the mean follow-up time, which, although not significant in the meta-regression analysis, could also have influenced the differences observed. Finally, although most of the studies were performed in Brazilian populations, 7 studies (19%) were conducted in other countries, mainly Colombia and the United States. Differences in the distribution of Trypanosoma cruzi discrete typing units and other variables could have also influenced the observed results.

Second, we identified a potential publication bias in the studies reporting unadjusted risks for CCC vs non-CC groups. However, a significant effect was still observed after we adjusted by this type of bias using a trim-and-fill method. Moreover, the potential confounders included in the multivariate-adjusted models varied significantly among studies, as different covariate selection approaches were used. Nevertheless, the absence of significant results in the meta-regression analyses supports the accuracy of our results. Furthermore, the absence of patient-level data restricted our ability to conduct more refined subgroup analyses or assess the influence of individual patient characteristics on outcomes.

Of note, patients with CCC represent a vulnerable population, with limited access to health services and, therefore, to HF medications and therapies that reduce mortality, such as neurohormonal blockade and implantable cardioverter-defibrillators, which may influence their survival. However, we were unable to include data on socioeconomic status, access to HF therapies, or adherence to these drugs in our analyses, representing an important limitation.

Finally, assessment of all-cause mortality allowed the inclusion of a larger number of studies, not discriminating between the different causes of mortality (HF, sudden cardiac death, stroke, among others).