ABC | Volume 111, Nº5, November 2018

Original Article Rattanawong et al PR interval and chronic resynchronization therapy Arq Bras Cardiol. 2018; 111(5):710-719 PR prolongation in comparison with normal PR interval is associated with outcomes in CRT-dependent HF patients by assessing all-cause mortality, HF hospitalization rate, and composite outcome as our interest. Method Search strategy Two investigators (NP and TR) independently searched for published studies indexed in MEDLINE and EMBASE databases from inception to January 2017 using a search strategy that included the terms “PR interval” and “cardiac resynchronization therapy” described in supplementary document 1. Only English language publications were included. A manual search for additional pertinent studies and review articles using references from retrieved articles was also made. Inclusion criteria The eligibility criteria included the following: (1) Cohort study (prospective or retrospective) reporting incident of all-cause mortality, HF hospitalization, or composite outcome, after the CRT and the corresponding index date for controls. (2) Relative risk, hazard ratio, incidence ratio, or standardized incidence ratio with 95% confidence intervals or sufficient raw data for the calculation were provided. (3) Participants without PR prolongation were used as controls. Study eligibility was independently determined by two investigators (NP and TR) and differences were resolved by mutual consensus. A Newcastle-Ottawa quality assessment scale was used to evaluate each study in three domains: recruitment and selection of the participants, similarity and comparability between the groups, and ascertainment of the outcome of interest among cohort studies. 11 Data extraction A standardized data collection form was used to obtain the following information from each study: title of study, name of first author, year of study, year of publication, country of origin, number of participants, demographic data of participants, method used to identify cases and controls, method used to diagnose the outcomes of interest (all-cause mortality, HF hospitalization rate and composite outcome), and average duration of follow-up with confounders that were adjusted and adjusted effect estimates with 95% confidence interval and covariates that were adjusted in the multivariable analysis. To ensure accuracy, all investigators independently performed this data extraction process. Any data discrepancy was resolved by referring back to the original articles. Statistical analysis We performed a meta-analysis of the included cohort studies using a random-effects model. The extracted studies were excluded from the analysis if they did not present an outcome in each intervention group or did not have enough information required for continuous data comparison. We pooled the point estimates from each study using the generic inverse-variance method of Der Simonian and Laird. 12 The heterogeneity of effect size estimates across these studies was quantified using the I 2 statistic. The I 2 statistic ranges in value from 0 to 100% (I 2  <  25%, low heterogeneity; I 2  =  25%–50%, moderate heterogeneity; and I 2  >  50%, substantial heterogeneity). 13 A sensitivity analysis was performed to assess the influence of the individual studies on the overall results by omitting one study at a time. Meta-regression was performed to explore source of heterogeneity. Publication bias was assessed using funnel plot and Egger’s regression test 14 (p  <  0.05 was considered significant). All data analyses were performed using the Stata/SE 14.1 software from StataCorp LP. Results Description of the included studies Our search strategy yielded 580 potentially relevant articles (82 articles from EMBASE and 498 articles from MEDLINE). After exclusion of 204 duplicated articles, 376 underwent title and abstract review. Three hundred and seventy articles were excluded at this stage since they were not cohort studies, did not report the outcome of interest (incidence of death/ HF hospitalization) or were not conducted in patients with CRT, leaving six for full-length article reviews. Therefore, six retrospective cohort studies with 17,432 normal PR and 4,278 prolonged PR patients were included in this meta-analysis. Figure 1 outlines the search and literature review process. The clinical characteristics and summary of the included studies are described in Table 1. Quality assessment of the included studies Newcastle-Ottawa scales of the included studies are described in Table 2. The Newcastle-Ottawa scale uses a star system (0 to 9) to evaluate the included studies on three domains: selection, comparability, and outcomes. Higher scores represent higher study quality. Intra-study risks of bias of the included studies are also described in Table 3. Meta-analysis results Six studies 2,4,7,8,15,16 from January 1991 to May 2017 were included in this meta-analysis. All-cause mortality rate is available in four studies 2,4,7,16 that involved 17,432 normal PR and 4,278 prolonged PR. All four studies revealed an increased death rate among patients with prolonged PR interval but with of the four achieving statistical significance. The pooled analysis demonstrates a statistically significant increased risk of all-cause mortality in patients with prolonged PR interval compared to participants without prolonged PR interval with the pooled risk ratio of 1.34 (95 % confidence interval: 1.08‑1.67, p < 0.01). The statistical heterogeneity was substantial with I 2 of 57.0%. Forest plot of this meta-analysis is shown in Figure 2A. 711