Alcohol intake and associated risk of major cardiovascular outcomes in women compared with men: a systematic review and meta-analysis of prospective observational studies

This review was conducted and was reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis Statement issued in 2009 [11]. Any prospective observational study that evaluated the association between alcohol intake and subsequent major cardiovascular outcomes or total mortality risk in men and women was eligible for inclusion in our study, and no restrictions were placed on language or publication status. Relevant studies were identified using the following procedure:The literature search, data extraction, and quality assessment were independently undertaken by two investigators (YHZ and CZ) using a standardized approach. Any inconsistencies between these investigators were identified by the primary investigator (JH) and resolved by consensus. We restricted our study to prospective observational studies that were less likely to be subject to confounding variables or bias compared to traditional case control studies [12]. A study would be eligible for inclusion in our meta-analysis if the following criteria were met: (1) the study was a prospective observational study (prospective cohort study or nested case control study); (2) the study investigated the association between alcohol intake and the risk of major cardiovascular outcomes or total mortality in men and women separately; and (3) the authors reported effect estimate (risk ratio [RR], odds ratio [OR], or hazard ratio [HR]) and 95 % confidence intervals (CIs) on cardiovascular outcomes (coronary disease, total mortality, cardiac death, stroke, and ischemic stroke) for comparisons of different dosage of alcohol intake with the lowest alcohol intake or non-drinking.

The information collected the included group’s name, country, study design, sample size for men and women, age at baseline, percentage of sample size for different alcohol intake categories, follow-up duration, and covariates in the fully adjusted model. We also extracted the effect estimate and its 95 % CIs. For studies that reported several multivariable adjusted RRs, we selected the effect estimate that was maximally adjusted for potential confounders.

The Newcastle-Ottawa Scale (NOS) has been partially validated for evaluating the quality of observational studies, was employed to evaluate methodological quality [13, 14]. The NOS is based on the following three subscales: selection (four items), comparability (one item), and outcome (three items). A “star system” (range, 0–9) has been developed for assessment [13]. The data extraction and quality assessment were conducted independently by two authors (YHZ and FL). Referring to the original studies, information was examined and adjudicated independently by an additional author (JH).

We examined the relationship between alcohol intake and the risk of major cardiovascular outcomes or total mortality in women compared with men based on the effect estimate (RR, OR, or HR) and its 95 % CI in each study. For every study, sex-specific RRs and 95 % CIs were used to estimate the female-to-male ratio of RRs (relative risk ratio [RRR]) and its 95 % CIs [15]. First, we used a random effects model to calculate summary RRs and 95 % CIs for different exposure categories versus the lowest alcohol intake in men and women separately. Next, both fixed-effect and random-effect models were used to evaluate the pooled RRR for the comparison of different exposure categories versus the lowest alcohol intake in women compared with men; the results from the random-effect model which assumed that the true underlying effect varied among included trials, were presented here [16, 17].

Heterogeneity between studies was investigated using the Q statistic, and we considered P-values of < 0.10 as indication of significant heterogeneity [18‐20]. Subgroup analyses were conducted for coronary disease or total mortality based on the country, sample size, physical activity, serum cholesterol, hypertension, diabetes, follow-up duration, and the study quality.

We also performed a sensitivity analysis by removing a specific study from the meta-analysis [21]. Several methods were used to check for potential publication bias. Visual inspections of funnel plots for coronary disease and total mortality were conducted. The Egger [22] and Begg tests [23] were also used to statistically assess publication bias for coronary disease and total mortality. All reported P-values were 2-sided and P-values of <0.05 were considered statistically significant for all included studies. Statistical analyses were performed using STATA software (version 12.0; Stata Corporation, College Station, TX, USA).

The results of the study selection process were shown in Fig. 1. We identified 2,567 articles in our initial electronic search, of which 2,436 were excluded because they were duplicate or irrelevant articles. A total of 131 potentially eligible studies were selected. After detailed evaluations, 23 prospective studies including 18 cohorts were selected for the final meta-analysis [8‐10, 24‐43]. A manual search of the reference lists within these studies did not yield any new eligible studies. The general characteristics of the included studies were presented in Tables 1 and Additional file 2: Table S1.

Of the 18 included cohorts, reporting data on 489,696 individuals, 16 cohorts (20 studies) had been examined using a prospective cohort study design [8‐10, 24‐37, 39, 42, 43], and the remaining two cohorts (three studies) had been examined by using a prospective nested case control study design [38, 40, 41]. The follow-up period for participants was 5.0–20.0 years, and 1,620–114,928 individuals were included in each study. A total of six cohorts (seven studies) were conducted in the US [10, 24, 26, 30, 31, 42, 43], ten (13 studies) were conducted in Europe [8, 25, 27‐29, 32, 33, 35, 37‐41], and two (three studies) were conducted in other countries [9, 34, 36]. Study quality was assessed using the NOS. Overall, two cohorts had a score of 9 [29, 34‐36], four cohorts had a score of 8 [9, 24, 32, 33, 42, 43], seven cohorts had a score of 7 [8, 10, 27, 28, 31, 37, 39], three cohorts had a score of 6 [25, 30, 38], and the remaining two cohorts had a score of 5 [26, 40, 41].

A total of nine cohorts (11 studies) reported an association between alcohol intake and the risk of coronary disease [8‐10, 27, 29, 32, 33, 38‐41]. The summary RR of the associated between alcohol intake and coronary disease in men and women, were separately listed in Table 2. The pooled RRR (female to male) of low alcohol intake (<15 g/day) versus the lowest alcohol or no alcohol intake was 1.01 (95 % CI: 0.84–1.21; P = 0.947; Table 2 and Additional file 2: Figure S1), with no evidence of heterogeneity among included studies . Furthermore, the pooled RRR (female to male) was 0.96 (95 % CI: 0.75–1.23; P = 0.772; Table 2 and Additional file 2: Figure S2) for moderate alcohol intake (15–30 g/day). There was a significant heterogeneity among the included studies (I² = 40.7 %; P = 0.096). Finally, the pooled RRR (female to male) was reduced by 10 % (RRR, 0.90; 95 % CI: 0.66–1.22; P = 0.503; with moderate heterogeneity; Table 2 and Additional file 2: Figure S3) for heavy alcohol intake (>30 g/day), but this reduction was not statistically significant.

A total of 10 cohorts (14 studies) reported an association between alcohol intake and the risk of total mortality [8, 10, 25‐29, 34‐36, 40‐43]. The summary RR of the associated between alcohol intake and total mortality in men and women, were separately listed in Table 2. The pooled RRR (female to male) for moderate alcohol intake and the risk of total mortality was statistically significantly increased (RRR, 1.10; 95 % CI: 1.00–1.21; P = 0.047; Table 2 and Additional file 2: Figure S5). Although the summary RRR (female to male) increased, there was no significant association between low (RRR, 1.07; 95 % CI: 0.98–1.17; P = 0.143; Table 2 and Additional file 2: Figure S4) or heavy alcohol intake (RRR, 1.09; 95 % CI: 0.99–1.21; P = 0.084; Table 2 and Additional file 2: Figure S6) and the risk of total mortality in women compared with men.

The breakdown for the number of cohorts available for each outcome were four (seven studies), five (seven studies), and three (five studies) for cardiac death [25, 31, 34‐36, 42, 43], stroke [10, 30, 34‐38], and ischemic stroke [24, 30, 34‐36] respectively. These associations in men and women separately were shown in Table 2. The summary RRRs (female to male) of low alcohol intake were 0.93, 0.99, and 0.94 for cardiac death (RRR, 0.93; 95 % CI: 0.83–1.04; P = 0.216; Table 2 and Additional file 2: Figure S8), stroke (RRR, 0.99; 95 % CI: 0.83–1.16; P = 0.864; Table 2 and Additional file 2: Figure S9), and ischemic stroke (RRR, 0.94; 95 % CI: 0.74–1.20; P = 0.633; Table 2 and Additional file 2: Figure S7) respectively. Similarly, the summary RRRs (female to male) of moderate alcohol intake were 0.99, 0.90, and 0.88 for cardiac death (RRR, 0.99; 95 % CI: 0.87–1.14; P = 0.934; Table 2 and Additional file 2: Figure S10), stroke (RRR, 0.90; 95 % CI: 0.74–1.10; P = 0.299; Table 2 and Additional file 2: Figure S11), and ischemic stroke (RRR, 0.88; 95 % CI: 0.66–1.16; P = 0.366; Table 2 and Additional file 2: Figure S12) respectively. Finally, the summary RRRs (female to male) of low alcohol intake were 1.44, 1.35, and 1.04 for cardiac death (RRR, 1.14; 95 % CI: 0.99–1.32; P = 0.075; Table 2 and Additional file 2: Figure S13), stroke (RRR, 1.35; 95 % CI: 0.77–2.35; P = 0.292; Table 2 and Additional file 2: Figure S14), and ischemic stroke (RRR,1.04; 95 % CI: 0.80–1.36; P = 0.762; Table 2 and Additional file 2: Figure S15) respectively.

Sensitivity analyses indicated that exclusion of any individual study did not significantly alter the results (data not shown). Heterogeneity testing for the analysis showed P >0.10 for coronary disease and total mortality. We concluded that heterogeneity was not significant in the overall analysis, which suggested that most variation was attributable to chance alone. Subgroup analyses were also conducted for coronary disease and total mortality to evaluate the effect of alcohol intake in women compared with men in specific subpopulations. The summary RRR (female to male) was significantly increased for the association between moderate alcohol intake and the risk of coronary disease for studies conducted in the US. Furthermore, the summary RRR (female to male) was significantly increased for the association between heavy alcohol intake and subsequent total mortality risk for studies conducted in the US. The study was not adjusted for physical activity, diabetes, or low NOS score (Table 3).

Review of the funnel plots could not rule out the potential for publication bias for the risk of coronary disease, and total mortality. The Egger [22] and Begg test [23] results showed no evidence of publication bias for the risk of coronary disease (low [Additional file 2: Figure S16], moderate [Additional file 2: Figure S17] and heavy alcohol intake [Additional file 2: Figure S18]), total mortality (low [Additional file 2: Figure S16], moderate [Additional file 2: Figure S17] and heavy alcohol intake [Additional file 2: Figure S18]) in women compared with men.