Association Between Alcohol Consumption and Risk of

0 downloads 0 Views 693KB Size Report
Objective: To conduct a meta-analysis summarizing the risk of cardiovascular disease (CVD) and all-cause mortality (ACM) in relation to alcohol consumption in ...
ORIGINAL ARTICLE

Association Between Alcohol Consumption and Risk of Cardiovascular Disease and All-Cause Mortality in Patients With Hypertension: A Meta-Analysis of Prospective Cohort Studies Chao Huang, BS; Jian Zhan, MPH; Yu-Jian Liu, MPH; De-Jia Li, PhD; Su-Qing Wang, PhD; and Qi-Qiang He, PhD Abstract Objective: To conduct a meta-analysis summarizing the risk of cardiovascular disease (CVD) and all-cause mortality (ACM) in relation to alcohol consumption in patients with hypertension, focusing on clarifying dose-response associations. Patients and Methods: PubMed and EMBASE were searched for eligible prospective cohort studies from December 3, 1949, through January 18, 2014. The semi-parameter method and dose-response analysis were used. Results: Nine studies (11 cohorts) were included in the meta-analysis. Compared with the lowest alcohol level (abstainers/occasional drinkers), the pooled relative risk (RR) was 0.72 (95% CI, 0.68-0.77) for the third highest category (median, 10 g/d), 0.81 (95% CI, 0.71-0.93) for the second highest category (median, 20 g/d), and 0.60 (95% CI, 0.54-0.67) for the highest category (median, 30 g/d). A J-shaped relationship between alcohol use and ACM was observed, and the nadir (RR, 0.82; 95% CI, 0.76-0.88) was found to be at a dose of 8 to 10 g of alcohol consumption per day. Conclusion: Findings of this meta-analysis suggest that low-to-moderate alcohol consumption was inversely significantly associated with the risk of CVD and ACM in patients with hypertension. ª 2014 Mayo Foundation for Medical Education and Research

A

lcohol drinking is very common all over the world. Excessive alcohol consumption has a deleterious effect on health. It has been estimated that alcohol is responsible for 3.8% of global deaths and 4.6% of global disability-adjusted life-years worldwide.1 However, several studies have suggested that regular light-moderate alcohol intake is inversely associated with cardiovascular disease (CVD) and all-cause mortality (ACM) in the healthy population.2-5 Similarly, patients with hypertension also seem to benefit from low levels of alcohol intake. 6,7 The latest US hypertension guideline advises that alcohol intake should not exceed 2 drinks per day among men and 1 drink per day among women.8 Recommendations from the European Society of Hypertension and the European Society of Cardiology also suggest moderate alcohol consumption in subjects with hypertension.9 Although there is evidence that alcohol elevates blood pressure in a dose-dependent

n

Mayo Clin Proc. 2014;89(9):1201-1210

way5,10 and the relationship between stroke and alcohol consumption is J-shaped,11 guidelines about alcohol consumption are most likely to be grounded on the consensus of experts rather than substantial scientific evidence, because no evidence-based studies were cited in any of them.8,9 It remains unclear whether patients with hypertension should be advised to drink small amounts or cut down or abstain completely from alcohol. Considering that patients with hypertension are at a higher risk of cardiovascular events than healthy people, we, therefore, conducted a meta-analysis to clarify the quantitative associations between alcohol consumption and the risk of CVD and ACM among populations with hypertension.

From the School of Public Health (C.H., J.Z., Y.-J.L., D.-J.L, S.-Q.W., Q.-Q.H.) and Global Health Institute (D.-J.L, S.-Q.W., Q.-Q.H.), Wuhan University, Wuhan, P. R. China.

METHODS Search Strategy and Inclusion Criteria PubMed and EMBASE were searched for prospective cohort studies that evaluated the

Mayo Clin Proc. n September 2014;89(9):1201-1210 n http://dx.doi.org/10.1016/j.mayocp.2014.05.014 www.mayoclinicproceedings.org n ª 2014 Mayo Foundation for Medical Education and Research

1201

MAYO CLINIC PROCEEDINGS

association between alcohol consumption and the risk of ACM or CVD in patients with hypertension from December 3, 1949, to January 18, 2014. The search terms included the following key words: “alcohol OR ethanol OR beer OR wine OR spirits,” “stroke OR cerebral infarction OR brain infarction OR coronary heart disease OR intracerebral hemorrhage OR brain hemorrhage OR ([cerebrovascular OR cardiovascular OR CVD OR coronary OR stroke] AND [disease OR incidence OR survival OR morbidity OR fatal OR fatality OR mortality OR mortalities OR death]) OR allcause mortality OR total mortality,” and “hypertension OR hypertensive OR hypertensives OR blood pressure.” We limited the search to studies on humans and published in English. Contents of bibliographies from retrieved articles including original articles, reviews, and other article types were searched for additional studies of interest. We included studies in this meta-analysis if they met the following criteria: (1) prospective cohort studies; (2) the exposure was alcohol consumption, including alcohol, wine, beer, and spirits; (3) the outcomes were ACM and CVD events, the latter including mortality and incidence of CVD, coronary heart disease (CHD), stroke, and heart failure; and (4) quantitative data were available on the relationship between alcohol consumption and the risk of ACM or CVD in patients with hypertension, including risk estimates with 95% CIs or sufficient data to calculate these statistics. In cases of duplicate publication by the same authors, the study with the longest follow-up was included. Supplemental Figure 1 (available online at http://www.mayoclinicproceedings.org) shows a flowchart of the publication screening in this article. Data Extraction and Quality Assessment Two investigators (H.C. and Z.J.) independently performed the literature search and extracted all data; any disagreements were resolved by consensus. The following information was recorded from each study: authors, country and study designation, sex, outcome, number of cases and participants/personyears of follow-up, multivariate-adjusted risk estimates (relative risks [RRs] or hazard ratios, but hereinafter formalized with the generic term RRs12,13) and corresponding 95% CIs, 1202

Mayo Clin Proc.

n

alcohol consumption categories, follow-up period, age range, and confounders. If the reference group was not the lowest alcohol consumption category or the multivariate-adjusted risk estimates were not available in the study, unadjusted crude risk estimates were calculated using original data. The study quality was evaluated using an assessment scale for observational studies modified from previous studies.14 The total score ranges from 0 to 10 points, with a higher score indicating higher quality. Statistical Analyses Because the studies reported alcohol consumption in diverse measuring units, we transferred the exposure data into a uniform measurement of grams of ethanol per day (g/d) for the purpose of pooling estimates across studies. If a study did not provide data on the ethanol content of a drink or a unit, the average alcohol consumption was assumed to be 12 g per drink in studies conducted in the United States, 10 g per unit in Europe,15 and 11.5 g per drink in Japan.16 The median dose of each alcohol category was extracted and recalculated if not ready-provided for the purpose. For closed exposure interval, the median of the range was computed. If the lower bound for the lowest category was not provided, half of the upper bound of that category was taken as the median dose. For the highest open-ended exposure levels, the corresponding dose was assigned as the summation of the lower limit and threequarters of the preceding category range.17 The semi-parametric method used in this analysis was modified from Ding et al.18 Four alcohol consumption groups were created, namely, the lowest, the third highest (median alcohol consumption, 10 g/d; range, 5.1-15 g/d), the second highest (median alcohol consumption, 20 g/d; range, 15.1-25 g/d), and the highest (median alcohol consumption, 30 g/d; range, 25.1-65 g/d). Abstainers/occasional drinkers, the lowest category, were considered as the reference group. For each included study, the lowest alcohol consumption categories were arranged to the lowest groups, and the highest to the highest. For other exposure categories in the study, they were arranged to either the second or the third highest group, depending on whether the median consumptions fell into the

September 2014;89(9):1201-1210

n

http://dx.doi.org/10.1016/j.mayocp.2014.05.014 www.mayoclinicproceedings.org

ALCOHOL AND CARDIOVASCULAR RISK IN HYPERTENSION

range of the second or the third highest group in the meta-analysis. At first, a random effects model was used. If no significant heterogeneity was found between studies (I2 index .05), it was replaced with a fixed effects model. We applied generalized least-squares regression models19 based on the maximum likelihood method to perform a dose-response meta-analysis. Before estimating the dose-response model, a 2-stage fixed effects generalized linear model was used at first.19 If the P value for heterogeneity was less than .05, it was changed to a random effects generalized linear model.18 Furthermore, we evaluated the nonlinearity trend of alcohol consumption and CVD and ACM. A fixed/ random effects restricted cubic spline model with 4 knots (5%, 35%, 65%, and 95% of exposure levels) was used.17 In sensitivity analysis, each study was eliminated in turn from the pooled analysis to assess its effect on pooled results. Stratified analysis was carried out by sex, source of participants, region, publication year, alcohol consumption reporting methods, quality score, and CVD subtype (stroke and CHD). To examine the potentiality for publication bias, we assessed using both Begg funnel plots20 and the Egger tests.21 All statistical analyses were carried out by using STATA version 11.0 (STATA Corp) in this meta-analysis. RESULTS Characteristics of Included Studies After a prudent process of electronic search and identification, 9 relevant studies (11 cohorts)16,22-29 were included in this metaanalysis, with a total number of 394,840 participants. The quality scores ranged from 4.5 to 8.5, and the median was 6.5. Three studies were conducted in Europe,22,26,29 1 in Japan,16 and 5 in the United States.23-25,27,28 In 2 studies,16,29 unadjusted RRs were recalculated taking the nondrinkers/occasional drinkers as the reference group. The end points involved in our analysis included death due to all causes, overall CVD, CHD, stroke, and myocardial infarction and the incidence of overall CVD, CHD, stroke, myocardial infarction, angina pectoris, heart failure, as well as acute coronary syndrome. Detailed characteristics of the studies are listed in the Supplemental Table (available online at http:// www.mayoclinicproceedings.org). Mayo Clin Proc. n September 2014;89(9):1201-1210 www.mayoclinicproceedings.org

n

Semi-parameter Analysis Between Alcohol Consumption and Risk of CVD Figure 1 comprises 3 forest plots presenting summarized estimates for the risk of CVD relative to the 3 alcohol categories, respectively. Compared with the lowest category of alcohol drinking (median, 0 g/d), the pooled RR for overall CVD outcome was 0.72 (95% CI, 0.68-0.77) for the third highest (median, 10 g/d), 0.81 (95% CI, 0.71-0.93) for the second highest (median, 20 g/d), and 0.60 (95% CI, 0.54-0.67) for the highest category (median, 30 g/d). Nonsignificant heterogeneity (P>.05) was found in these analyses for each category of alcohol consumption. In sensitivity analysis (Supplemental Figure 2 [available online at http://www.mayoclinicproceedings.org]), nonexclusion of the studies did significantly change the pooled estimates.

Stratified Analysis for CVD Pooled estimates of CVD for each alcohol category are stratified by confounders and presented in a combined forest plot in Figure 2. Significant variances obtained from random effects models were found at the third highest alcohol category in population-based studies (RR, 0.67; 95% CI, 0.54-0.82; P for heterogeneity¼.02) vs occupation-based studies (RR, 0.72; 95% CI, 0.68-0.78; P for heterogeneity¼.59); in studies of Europe (RR, 0.69; 95% CI, 0.54-0.87; P for heterogeneity¼0.02) vs those of Japan (RR, 0.6; 95% CI, 0.42-0.84; P for heterogeneity¼.12) and the United States (RR, 0.72; 95% CI, 0.68-0.78; P for heterogeneity¼.59); and in studies with the method of ascertainment being interviewer-administrated (RR, 0.62; 95% CI, 0.47-0.82; P for heterogeneity¼.02) vs self-administrated (RR, 0.73; 95% CI, 0.680.78; P for heterogeneity¼.49). In contradiction to other subgroups and overall RRs, the pooled RR for the third highest group was 0.82 (95% CI, 0.68-0.99) in females, higher than that for the second highest group, 0.70 (95% CI, 0.54-0.91). In addition, lower RRs were observed in studies published later after the year 2000 and administrated by professional interviewers regardless of any alcohol category. For all that, results from the stratified analysis remained consistent with the overall categorized analysis.

http://dx.doi.org/10.1016/j.mayocp.2014.05.014

1203

MAYO CLINIC PROCEEDINGS

Third highest level vs lowest level Author

Year

Gender

Outcome

Palmer Palmer Palmer Higashiyama Higashiyama Higashiyama Beulens Beulens Beulens Beulens Beulens Beulens Beulens Britton Britton Britton Malinskj Bos Bos Djousssé Hansen Hansen

1995 1995 1995 2013 2013 2013 2007 2007 2007 2007 2007 2007 2007 2009 2009 2009 2004 2010 2010 2008 2011 2011

M F M M,F M,F M M M M M M M M M M M M F F M M F

CHD mortality CHD mortality Stroke mortality CVD CHD Stroke CVD mortality CVD mortality MI mortality MI mortality MI MI Stroke CHD Angina pectoris MI CVD mortality CHD Stroke Heart failure ACS ACS

RR (95% CI) 0.75 (0.54, 1.04) 1.03 (0.70, 1.52) 0.56 (0.33, 0.94) 0.61 (0.38, 0.97) 0.38 (0.20, 0.73) 1.09 (0.52, 2.44) 0.84 (0.57, 1.25) 0.78 (0.53, 1.14) 0.85 (0.55, 1.33) 0.45 (0.28, 0.74) 0.80 (0.54, 1.20) 0.86 (0.59, 1.26) 0.86 (0.57, 1.30) 0.72 (0.63, 0.83) 0.71 (0.62, 0.83) 0.78 (0.64, 0.97) 0.61 (0.49, 0.77) 0.92 (0.68, 1.24) 0.65 (0.44, 0.95) 0.72 (0.57, 0.91) 0.41 (0.27, 0.63) 0.54 (0.30, 0.96)

Overall (I2=28.9%, P=.102)

Median dose 7.9 5.7 7.9 11.5 11.5 11.5 7.5 12.5 7.5 12.5 7.5 12.5 5 10.3 10.3 10.3 5.9 7.15 5.35 10.3 14.3 14.3

0.72 (0.68, 0.77) .25

.5

1

2

4

A Second highest level vs lowest level Author

Year

Gender

Outcome

RR (95% CI)

Median dose

Palmer Palmer Beulens Beulens Beulens Beulens Bos Bos et al Hansen Hansen

1996 1996 2007 2007 2007 2007 2010 2010 2011 2011

M M M M M M F F M F

CHD mortality Stroke mortality CVD mortality MI mortality MI Stroke CHD Stroke ACS ACS

1.05 (0.72, 1.51) 0.62 (0.32, 1.21) 0.86 (0.58, 1.26) 0.79 (0.50, 1.25) 0.69 (0.46, 1.03) 1.40 (0.93, 2.12) 0.72 (0.52, 1.01) 0.76 (0.47, 1.22) 0.59 (0.38, 0.93) 0.49 (0.21, 1.04)

22.9 22.9 22.5 22.5 22.5 20 15 15 24.3 24.3

Overall (I2=35.7%, P=.123)

0.81 (0.71, 0.93) .25

.5

1

4

2

B Highest level vs lowest level Author

Year

Gender

Outcome

RR (95% CI)

Median dose

Palmer Palmer Palmer Higashiyama Higashiyama Higashiyama Beulens Beulens Beulens Beulens Britton Britton Britton Malinskj Bos Bos Djousssé Hansen Hansen

1995 1995 1995 2013 2013 2013 2007 2007 2007 2007 2009 2009 2009 2004 2010 2010 2008 2011 2011

M F M M,F M,F M,F M M M M M M M M F F M M F

CHD mortality CHD mortality Stroke mortality CVD CHD Stroke CVD mortality MI mortality MI Stroke CHD Angina pectoris MI CVD mortality CHD Stroke Heart failure ACS ACS

0.65 (0.44, 0.96) 0.85 (0.45, 1.61) 0.55 (0.28, 1.05) 0.50 (0.26, 0.93) 0.27 (0.09, 0.69) 1.00 (0.38, 2.59) 0.86 (0.45, 1.63) 0.42 (0.17, 1.04) 0.37 (0.14, 0.97) 0.91 (0.53, 1.57) 0.62 (0.45, 0.84) 0.62 (0.44, 0.86) 0.57 (0.35, 0.95) 0.56 (0.44, 0.71) 0.65 (0.46, 0.90) 0.81 (0.51, 1.29) 0.38 (0.20, 0.72) 0.40 (0.27, 0.62) 0.71 (0.36, 1.38)

40.7 16.4 40.7 63.25 63.3 63.3 65 65 65 45 16.25 16.3 16.3 18.3 27.5 27.5 16.3 46.5 36.5

Overall (I2=1.5%, P=.438)

0.60 (0.54, 0.67) .25

.5

1

2

4

C FIGURE 1. A, Risk estimates and 95% CI for the association between the third highest level (median consumption, 10 g/d) of alcohol consumption and the risk of CVD compared with the lowest level (median consumption, 0 g/d) of alcohol consumption in patients with hypertension. B, Risk estimates and 95% CI for the association between the second highest level (median consumption, 20 g/d) of alcohol consumption and the risk of CVD compared with the lowest level (median consumption, 0 g/d) of alcohol consumption in patients with hypertension. C, Risk estimates and 95% CI for the association between the highest level (median consumption, 30 g/d) of alcohol consumption and the risk of CVD compared with the lowest level (median consumption, 0 g/d) of alcohol consumption in patients with hypertension. ACS ¼ acute coronary syndrome; F ¼ female; M ¼ male; MI ¼ myocardial infarction.

1204

Mayo Clin Proc.

n

September 2014;89(9):1201-1210

n

http://dx.doi.org/10.1016/j.mayocp.2014.05.014 www.mayoclinicproceedings.org

ALCOHOL AND CARDIOVASCULAR RISK IN HYPERTENSION

Dose-Response Analysis Between Alcohol Consumption and the Risk of CVD In the dose-response analysis, a nonlinear inverse association was observed between alcohol consumption and the risk of CVD (P for nonlinearity .05; Figure 3, B) and CHD (P for nonlinearity >.05; Figure 3, C). Alcohol Consumption and the Risk of ACM The pooled estimates between categorized alcohol intake and the risk of ACM are shown in Figure 4. Compared with the lowest alcohol level (median, 0 g/d), the pooled RR for ACM was 0.83 (95% CI, 0.77-0.90) for the third highest (median, 10 g/d), 0.80 (95% CI, 0.68-0.95) for the second highest (median, 20 g/d), and 1.01 (95% CI, 0.89-1.14) for the highest category (median, 30 g/d). Although a significant heterogeneity (P