Risk factors for the development of premature ventricular complex-induced cardiomyopathy: a systematic review and meta-analysis

Selected studies

A total of 1567 studies were identified and 1540 were excluded. There were 65 full-text publications reviewed, of which 39 were excluded: 31 studies were based on the same cohorts (mostly representing abstracts of otherwise available complete studies) and 8 studies did not provide risk factors of interest or appropriate statistics. This resulted in 26 studies included in the present systematic review and meta-analysis [5, 7,8,9,10,11,12, 28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46] (Fig. 1).

Fig. 1figure 1

Study selection chart flow

Baseline study characteristics are presented in Table 1. The included studies reported data on patients treated between 1989 and 2019. They consisted of 9 prospective and 17 retrospective studies. One of the retrospective studies was a re-analysis of a register (the California Health Care Cost and Utilization Project (CHCCUP)) evaluating 16,757,903 patients that was qualitatively analyzed but was eventually excluded from the meta-analysis because of the bias caused by its extreme weight. The 25 other studies provided a total of 6738 patients.

Table 1 Study baseline characteristics

Further details regarding inclusion and exclusion criteria for each study and definitions of both PVC-CM and PVCs are presented in Supplemental Table 3. Fifteen of 26 (57.7%) studies provided a definition of PVC-CM: the CMP was mostly defined as an LVEF < 50% and 9/26 (34.6%) studies took a time component into account (e.g., normalization or increase in the EF over time). The requirement for LVEF improvement in the PVC-CM definition varied from 10 to 15% in these studies.

Baseline patient characteristics

Often, several groups were analyzed in each study, which did not always report data for the overall cohort. The analyzed groups are presented in Table 2. In summary, the overall patient population was rather young (weighted mean age of 50.2 years old, 55.0 years old when excluding data from the predominant CHCCUP study) and with a weighted mean PVC burden of 16.5% (not reported in the CHCCUP study). The weighted mean percentage of women in the overall analyzed dataset was 57.6%, which decreased to 44.2% when excluding data from the CHCCUP. In a significant proportion of the studies and reported groups, there was no described attempt to assess for the presence of underlying structural heart disease or this detail was not reported (8/26 studies, Supplemental Tables 4 and 5).

Table 2 Baseline characteristics of the patient groups in the 26 selected studiesAssessment of outcomes

Most of the studies assessed the presence of PVC-CM (17/26), the recovery of LVEF after PVC-CM 4/27 (defined as a binary variable), or the worsening of LVEF suspected to be due to PVC-CM 2/27 (also defined as a binary variable). We conducted a pooled analysis for these three outcomes, as these are solely different ways to define a PVC-CM. Studies reporting continuous LVEF change over time (3/26) were rare (Table 3).

Table 3 Derived models in the different studies and recorded outcomes and risk factorsAssessed risk factors

Table 4 presents the occurrence of all risk factors throughout the selected studies and the occurrence of reporting which were suitable for quantitative analysis (≥ 3 occurrences in multivariable model assessing a binary change in LV function).

Table 4 Candidate risk factors proposed in the 26 studies and their relative occurrence (either overall or in multivariable models assessing a binary change in LVEF—either an improvement, worsening in EF, or the development of a PVC-CMP—suitable for quantitative summary analysis)

Supplemental table 6 presents the risk factors analyzed by each study. The exact definitions of each risk factor, as provided by the individual studies, are presented in the supplemental.

PVC burden was the most commonly analyzed risk factor (24/26 studies, 20/26 studies for quantitative summary), followed by sex (13/26), PVC origin (11/26), PVC and morphology (10/26), and PVC and QRS duration (each in 8/26 studies). Only few other risk factors (age, coupling interval, non-sustained VTs, interpolation, and the presence of symptoms) were investigated in ≥ 3 studies and suitable for quantitative summary. Further investigated risk factors were baseline LVEF, coupling interval, polymorphic PVCs, and outflow tract origins. These risk factors did not appear often enough (< 3 appearances) or were differently defined, hence not suitable for quantitative summary.

Quantitative associations of risk factors with PVC-CM

When summarized quantitatively, age (OR 1.02 per increase in year of age, 95% CI [1.01, 1.02]), the presence of symptoms (OR 0.18, 95% CI [0.05, 0.64]), non-sustained VTs (OR 3.01, 95% CI [1.39, 6.50]), LV origin (OR 2.20, 95% CI [1.14, 4.23]), epicardial origin (OR 4.72, 95% CI [1.81, 12.34]), the presence of interpolation (OR 4.93, 95% CI [1.66, 14.69]), PVC burden (OR 1.06 per percent increase in burden, 95% CI [1.04, 1.08]), and PVC duration (OR 1.05 per ms increase in QRS-PVC duration [1.004; 1.096]) were all significantly associated with PVC-CM (Figs. 2, 3, 4, 5, 6, 7, 8, and 9). Coupling interval, polymorphic PVCs, outflow tract origin, sex, and QRS duration did not display a significant association (Supplemental Fig. 1).

Fig. 2figure 2

Random effects model showing the overall effect of age on the risk of developing PVC-CM. TE, estimate of treatment effect; seTE, standard error of treatment estimate; OR, odds ratio; CI, confidence interval

Fig. 3figure 3

Random effects model showing the overall effect of overall PVC burden on the risk of developing PVC-CM. TE, estimate of treatment effect; seTE, standard error of treatment estimate; OR, odds ratio; CI, confidence interval

Fig. 4figure 4

Random effects model showing the overall effect of epicardial origin of the PVC on the risk of the developing PVC-CM. TE, estimate of treatment effect; seTE, standard error of treatment estimate; OR, odds ratio; CI, confidence interval

Fig. 5figure 5

Random effects model showing the overall effect of interpolated PVCs on the risk for PVC-CM. TE, estimate of treatment effect; seTE, standard error of treatment estimate; OR, odds ratio; CI, confidence interval

Fig. 6figure 6

Random effects model showing the overall effect of left ventricular origin of the PVC on the risk of the developing PVC-CM. TE, estimate of treatment effect; seTE, standard error of treatment estimate; OR, odds ratio; CI, confidence interval

Fig. 7figure 7

Random effects model showing the overall effect of non-sustained ventricular tachycardia on the risk for PVC-CM. TE, estimate of treatment effect; seTE, standard error of treatment estimate; OR, odds ratio; CI, confidence interval

Fig. 8figure 8

Random effects model showing the overall effect of symptoms on the risk for PVC-CM. TE, estimate of treatment effect; seTE, standard error of treatment estimate; OR, odds ratio; CI, confidence interval

Fig. 9figure 9

Random effects model showing the effect of PVC duration (per ms increase in QRS PVC duration) on the risk for PVC-CM. TE, estimate of treatment effect; seTE, standard error of treatment estimate; OR, odds ratio; CI, confidence interval

Dose–response analysis of PVC burden

In the dose–response analysis encompassing 7 studies reporting PVC burden at different cutoffs, there was a highly significant association between increase in PVC burden and an exponential increase in risk for PVC-CM (at 10% PVC burden, beta-coefficient 1.54 [1.3, 1.8], at 20% PVC burden beta-coefficient 1.5 [1.7, 3.6], at 30% PVC burden beta-coefficient 4 [2.3, 7], Fig. 10). A univariate Cochran Q test for residual heterogeneity was highly significant, with an I2 statistic of 89.7%.

Fig. 10figure 10

Dose–response plot of PVC burden and association with PVC-CMP. Based on 7 studies reporting PVC burden with a cutoff, a dose–response analysis was conducted. The black line represents the predicted increase in PVC-CMP risk associated with an increase in PVC burden in %. The gray ribbon represents the confidence interval of the prediction

Modification of the risk associated with PVC burden through meta-regression

When assessing the risk modification associated with the publication year or with study quality, older studies and studies with higher quality were associated with a non-significant trend in increased risk for the development of PVC-CM with a growing PVC burden.

The PVC-CM risk associated with PVC burden decreased of 0.28% (− 0.28%, 95% CI [− 1.02%, 0.46%], P = 0.462, Supplemental Fig. 2) with each increase in publication year, meaning that studies published in 2020 displayed a non-significant 2.8% lower risk association of PVC-CM with PVC burden as compared with the studies published in 2010.

Inversely, the PVC-CM risk associated with PVC burden increased of 0.09% (95% CI [− 0.13%, 0.31%], P = 0.413, Supplemental Fig. 3) with each increase in quality point of the summed QUIPS tool, meaning that studies with a low risk of bias (in mean 45 points in the summed QUIPS tool) presented a 2.7% higher risk association of PVC-CM with PVC burden as compared with the studies with high risk of bias (in mean 15 points in the summed QUIPS tool).

Publication bias

On funnel plot analysis of PVC burden, study distribution was mildly asymmetric (Fig. 11) but the Egger test did not suggest any publication bias (P = 0.07).

Fig. 11figure 11

Assessment of publication bias using a contour-enhanced funnel plot. The contour-enhanced funnel plot represents the different studies reporting estimated for the association between PVC burden (continuous increase in %) and assess the risk for publication bias. The 7 studies reporting a cutoff of PVC burden were summarized beforehand as the “dose–response analysis.” The dotted line represents the overall estimate using all available studies and the dashed line represents a classical funnel plot with the expected distribution of the studies if no publication bias is present. The contour-enhanced funnel plot is centered at 0 (i.e., the value under the null hypothesis of no relationship) and various levels of statistical significance are indicated by the shaded region. The white region corresponds to non-significant P values. Highly significant P values appear in the light gray region

Quality assessment

As presented in Fig. 12, all of the studies presented with at least a moderate risk of bias. The uncontrolled risk of confounding appeared as the most problematic throughout all recorded studies.

Fig. 12figure 12

Assessment of study quality. Evaluation of study quality according to the QUIPS tool. Five domains of bias (participation, attrition, prognostic factor measurement, outcome measurement, confounding and statistical analysis and reporting) are represented with the associated risk of bias (high in red, moderate in yellow, and low in green). The overall column represents the mean risk of bias from the 6 domains

留言 (0)

沒有登入
gif