Alcohol Consumption by Beverage Type and Risk of Breast Cancer

A Dose-Response Meta-Analysis of Prospective Cohort Studies

Qiuyu Sun; Weihong Xie; Yanli Wang; Feifei Chong; Mengmeng Song; Tiandong Li; Linping Xu; Chunhua Song


Alcohol Alcohol. 2020;55(3):246-253. 

In This Article


Systematic Review

For this meta-analysis, 53 articles that discussed alcohol consumption and breast cancer incidence were identified between 1992 and 2017. Firstly screen the article through abstract, when an abstract meet the study needs, then browse full-text, after full-text evaluation, eight articles were excluded, including five articles (Adami et al., 1988; Berstad et al., 2008; Zhang and Holman, 2011; Chandran et al., 2013; Baglia et al., 2017) are case-control study, and three articles (Jain et al., 2000; Feigelson et al., 2001; Dam et al., 2016) outcome is mortality. Finally, 22 articles were included for analysis (Figure 1): study characteristics including year of publication, country, exposure categories, consumption categories, cases, adjusted RR or HR, alcohol assessment, and adjusted variables are summarized in Table 1. All included studies were considered as high or moderate quality based on the NOS (Supplementary Table S3).

All of the 22 included studies were cohort studies. In Japan (Lin et al., 2005), all participants were followed up for all-cause mortality; because there was no national cancer registry system, subjects were followed up for cancer incidence only in areas where cancer registry systems existed. Two articles (Park et al., 2014; Romieu et al., 2015) presented risk estimates from different study populations and were considered as one study in the analysis. All studies were adjusted for at least one lifestyle or risk factor in addition to age, sex and either education or socioeconomic status. Detailed information on alcohol consumption were all evaluated by questionnaire (n = 22).

Publication Bias and Heterogeneity

Egger's test for small study effects demonstrated that no small studies remained unpublished (P = 0.090) (Supplementary Figure S1). Judging from I 2 statistics (Figure 2a), there has been considerable heterogeneity (I-squared = 64.7%, P = 0.000); however, when assessing heterogeneity in subgroup analyses (Figure 2b), the heterogeneity is not substantial. From sensitivity analysis (Supplementary Figure S2), the first study (Chen et al., 2011) had the largest impact on the total combined effect amount, in which the combined effect amount after removal was RRs = 1.14 (95%CI = 1.11–1.17) from the original value to the RRs = 1.16 (95%CI = 1.13–1.20). Other studies had little or no impact on the combined effect amount.

Figure 2.

I 2 test and stratified analysis. a) Forest plot of summary relative risks (RRs) of breast cancer for alcohol consumption. The pooled RR for breast cancer was 1.17 (95%CI = 1.11–1.24, I 2 = 64.7%, P heterogeneity = 0.000). b) Subgroup analysis; there is no heterogeneity.

Risk Estimates From Classical Meta-analysis

Table 1 summarizes both unadjusted and adjusted estimates for consumption status under different receptor status from the classical meta-analysis. The adjusted RRs for current drinkers compared with never drinkers were 1.40 (95%CI = 1.30–1.51) in ER+/PR+ status, ER+/PR- patients had a 1.39 (95%CI = 1.12–1.71)-fold increased risk, and the RRs = 1.21 (95%CI = 1.02–1.43) were lowest in ER-/PR- patients. The observed RRs remained comparable to the unadjusted estimates.

Risk Estimates From Dose-response Meta-analysis

Dose-response curves estimated from studies reporting on total alcohol (n = 22), wine (n = 5), beer (n = 4) and spirits (n = 4) intake and breast cancer risk are depicted in Figure 3a and c–e; the relationship between total alcohol and postmenopausal breast cancer is presented in Figure 3b. The shape of both the intensity and pack-day curves is showing a rapid increase of breast cancer risk in total alcohol and wine, in which statistical tests for nonlinearity showed that it is linear (P > 0.05 at all investigated knots). For total alcohol and breast cancer, there was a 1.2-fold (95%CI = 1.17–1.27) higher risk of breast cancer per 20 g/d higher intake, while 1.19-fold (95%CI = 1.08–1.30) risk for wine. And for postmenopausal women, it was a 1.23-fold (95%CI = 1.18–1.29) higher risk each additional 20 g intake. However, for beer and spirits, there wasn't a statistically significant linear trend.

Figure 3.

Pooled log-linear associations estimated the risks of breast cancer and postmenopausal breast cancer in relation to total alcohol, beer, wine and spirits consumption. Solid lines correspond to pooled relative risks, and dashed lines correspond to the 95% confidence intervals of the pooled relative risks. a) Total alcohol intake with breast cancer risk and b) postmenopausal breast cancer risk. c) Beer, d) wine and e) spirits intake with breast cancer risk.

Attributable Risk Percentage (ARP)

By combining with the pooled risk estimates per continent from this meta-analysis, ARP was calculated (Table 2). Breast cancer attributable to drinking was 13.04% of drinkers in North America and 11.50% in Asia. However, the ARP in Europe seems higher (20.63%) compared with both North America and Asia.