Alcohol and Breast Cancer: Update 2012
Alcohol and Breast Cancer: Update 2012
An association between alcohol and breast cancer was first suggested in the early 1980s by case–control studies. A network of case–control studies from several areas of North America (Rosenberg et al., 1982) reported a relative risk (RR) of 1.9 (95% confidence interval, CI, 1.5–2.4) for ever vs. never-drinkers. A case–control study from Italy, where alcohol consumption in women was frequent and particularly high at that time (Talamini et al., 1984; Bosetti et al., 2007), gave a corresponding RR of 2.5 (95% CI, 1.7–3.7).
More than 100 epidemiological studies on alcohol consumption and female breast cancer were published afterwards, and a positive association is now established. Two recent International Agency for Research on Cancer (IARC) Monographs considered the effect of alcohol drinking in cancer aetiology, and concluded that female breast cancer is causally related to alcohol consumption (Secretan et al., 2009; IARC, 2010). Several aspects of alcohol consumption on breast cancer risk are, however, still under discussion. Among these are the effects of low amount of drinking (i.e. up to one drink/day) and the risk of breast cancer defined by estrogen receptor (ER) and progesterone receptor (PR) status.
In this section, we review the epidemiological evidence on alcohol drinking—with particular focus on different levels of consumption—and the risk of breast cancer. We also present a meta-analysis of data on light alcohol drinking.
It is particularly important to quantify the relation between light alcohol drinking and breast cancer risk. In fact, it is still unclear whether there is any threshold in intake below which no effect of alcohol on breast cancer is evident.
We performed a literature search in MEDLINE, ISI Web of Science (Science Citation Index Expanded) and EMBASE for epidemiological studies published prior to November 2011. We also reviewed references cited by the relevant retrieved articles. Articles were included in the meta-analysis only if they satisfied the following criteria: (a) case–control or cohort studies published as original articles; (b) studies that reported findings expressed as odds ratio, RR or hazard ratio (or reporting sufficient data to compute them) for light drinkers (≤12.5 g/day ethanol; ≤1 drink/day) vs. non-drinkers; (c) studies that reported standard errors or CIs of the risk estimates, or provided sufficient data to calculate them. We thus computed a pooled RR of breast cancer for light drinkers vs. non-drinkers, using random-effects models (DerSimonian and Laird, 1986). Statistical heterogeneity among studies was evaluated using I, which is the proportion of total variation contributed by between-study variance (Higgins and Thompson, 2002). Subgroup analyses and meta-regression models were carried out to investigate potential sources of between-study heterogeneity. We used a χ statistics to test for differences of summary estimates among subgroups (Greenland, 1987).
A total of 3431 papers were retrieved from the literature search. Of these, 113 papers reporting breast cancer risk estimates for light drinkers were included in the meta-analysis. The complete reference list is reported in Appendix.
The analysis included 44,552 cases in the reference category of non-drinkers (40,899 incident cases and 3653 deaths) and 77,539 cases in the light drinkers' category (76,303 incident cases and 1236 deaths). These numbers are slightly underestimated, since for six studies the number of exposed and/or non-exposed cases was not reported. Case–control was the most common study design (64% of studies); 51% of the included studies were from North America, 38% from Europe, 6% from Asia and 10% from other regions or from more than one region; 36% of the reported estimates were adjusted for the main risk factors (age, family history, parity, menopausal status, oral contraceptive/hormonal replacement therapy use), while 16% of the estimates included occasional drinkers in the reference category.
Figure 1 shows the RR of breast cancer in light drinkers vs. non-drinkers in each of the 113 included studies. We found substantial heterogeneity among single study estimates (I = 64%). The random-effect summary RR was 1.04 (95% CI, 1.02–1.07). We did not find evidence of heterogeneity in pooled estimates by design (P = 0.93) and area in which the study was carried out (P = 0.71). Results did not appreciably change from those of the overall analysis when considering only estimates adjusted for the main risk factors (pooled RR = 1.03, 95% CI, 1.00–1.07), as well as including only estimates not considering occasional drinkers in the reference category of non-drinkers (pooled RR = 1.04, 95% CI, 1.01–1.07).
(Enlarge Image)
Figure 1.
RRs of breast cancer for light drinkers vs. non-drinkers. Squares indicate study-specific RRs. Horizontal lines indicate the 95% CIs. Diamond indicates pooled RR with its corresponding 95% CI.
Therefore, this meta-analysis reported a modest but significant association between light drinking and breast cancer. The estimate was based on the results of more than one hundred studies. Women drink less than men (Gronbaek et al., 1994) and therefore low and moderate intakes are usually investigated more frequently and more in detail in women than in men, though the bias due to under reporting of even moderate alcohol consumption may be more relevant for women than for men (Allen et al., 2009). Since several populations show a high prevalence of light drinkers among women, even the small increase in risk we reported—in the order of 5%—represents a major public health issue in terms of breast cancers attributable to alcohol consumption.
High levels (i.e. ≥3 drinks/day) of alcohol consumption were associated with increased risk of breast cancer in the largest available studies (Hamajima et al. 2002; Allen et al., 2009; Chen et al., 2011). In the collaborative reanalysis of 53 epidemiological studies on breast cancer (Hamajima et al., 2002), the RRs were 1.32 (95% CI, 1.19–1.45) for consumption of 35–44 g/day and 1.46 (95% CI, 1.33–1.61) for consumption of ≥45 g/day of alcohol, when compared with non-drinkers. The association was similar in never- and ever-smokers. In the Million Women Study (Allen et al., 2009), the RR was 1.29 (95% CI, 1.23–1.35) for the highest level of consumption considered, i.e. ≥21 g/day, after adjustment for smoking and several other covariates. A 51% (95% CI, 35–70%) increase in risk emerged for drinkers of ≥30 g/day vs. non-drinkers in the Nurses' Health Study (Chen et al., 2011). Further, a case–control study of over 2500 women with breast cancer conducted in Italy, in a population characterized by relatively high alcohol drinking, reported an RR of 1.41 (95% CI, 1.17–1.71) for consumption >27 g/day when compared with abstainers (Ferraroni et al. 1998). Therefore, these results consistently indicate a 40–50% elevated risk of breast cancer in women consuming three or more alcohol drinks/day.
With reference to the dose–risk relation, the collaborative reanalysis on alcohol and breast cancer (Hamajima et al., 2002) found that the RR of breast cancer increased by 7.1% (95% CI, 5.5–8.7%) for each additional 10 g/day of alcohol intake. In the Million Women Study (Allen et al., 2009), the corresponding increase in alcohol consumption was associated to a 12% (95% CI, 9–14%) increased risk of breast cancer. A meta-analysis based on 49 studies and over 44,000 cases gave a dose–risk function which approached a RR of 1.5 at 40 g/day, and further increased for highest levels of intake (Fig. 2) (Bagnardi et al., 2001). A subsequent meta-analysis of about 100 epidemiological studies provided results according to the degree of control for confounding and the overall quality of the studies identified (Key et al., 2006). The increase in risk for additional 10 g/day of alcohol varied between 10 and 13%, according to the inclusion criteria used. Further, the trend in risk was highly significant (P < 0.001). There is, therefore, consistent evidence for a positive dose–risk relation between alcohol drinking and breast cancer.
(Enlarge Image)
Figure 2.
Dose-risk function between alcohol consumption and breast cancer (extracted from Bagnardi et al., 2001).
Since alcohol consumption might affect the risk of breast cancer through hormone-related mechanisms [such as increased estrogen and androgen levels (Sarkar et al., 2001; Singletary and Gapstur 2001) or increased levels of plasma insulin-like growth factor produced by the liver following alcohol drinking (Yu and Berkel, 1999)], several studies examined the association with breast cancer defined by ER and PR status (Deandrea et al., 2008; Suzuki et al., 2008; Lew et al., 2009; Chen et al., 2011). A meta-analysis summarized information on this issue, including 20 epidemiological studies (i.e. 4 cohort and 16 case–control studies) published up to 2007 (Suzuki et al., 2008). The meta-analysis reported an increased risk of 27% (95% CI, 17–38%) of all ER+ and of 14% (95% CI, 3–26%) of all ER-breast cancers for the highest vs. lowest level of alcohol drinking.
When the data were analyzed according to combined ER and PR status, the corresponding summary RRs were 1.22 for ER +/PR+ (95% CI, 1.11–1.34, based on 15 studies and over 11,000 cases of breast cancer), 1.28 for ER +/PR−(95% CI, 1.07–1.53, 11 studies, ~1900 cases), 1.31 for ER−/PR + (95% CI, 0.99–1.74, 8 studies, 580 cases) and 1.10 for ER−/PR− (95% CI, 0.98–1.24, 15 studies, ~4000 cases). Subsequent investigations further supported a stronger association between heavy alcohol consumption and ER + (and particularly ER +/PR+) breast cancers (Deandrea et al., 2008; Lew et al., 2009; Chen et al., 2011).
Epidemiology of Alcohol and Breast Cancer
Introduction
An association between alcohol and breast cancer was first suggested in the early 1980s by case–control studies. A network of case–control studies from several areas of North America (Rosenberg et al., 1982) reported a relative risk (RR) of 1.9 (95% confidence interval, CI, 1.5–2.4) for ever vs. never-drinkers. A case–control study from Italy, where alcohol consumption in women was frequent and particularly high at that time (Talamini et al., 1984; Bosetti et al., 2007), gave a corresponding RR of 2.5 (95% CI, 1.7–3.7).
More than 100 epidemiological studies on alcohol consumption and female breast cancer were published afterwards, and a positive association is now established. Two recent International Agency for Research on Cancer (IARC) Monographs considered the effect of alcohol drinking in cancer aetiology, and concluded that female breast cancer is causally related to alcohol consumption (Secretan et al., 2009; IARC, 2010). Several aspects of alcohol consumption on breast cancer risk are, however, still under discussion. Among these are the effects of low amount of drinking (i.e. up to one drink/day) and the risk of breast cancer defined by estrogen receptor (ER) and progesterone receptor (PR) status.
In this section, we review the epidemiological evidence on alcohol drinking—with particular focus on different levels of consumption—and the risk of breast cancer. We also present a meta-analysis of data on light alcohol drinking.
Light Alcohol Drinking
It is particularly important to quantify the relation between light alcohol drinking and breast cancer risk. In fact, it is still unclear whether there is any threshold in intake below which no effect of alcohol on breast cancer is evident.
We performed a literature search in MEDLINE, ISI Web of Science (Science Citation Index Expanded) and EMBASE for epidemiological studies published prior to November 2011. We also reviewed references cited by the relevant retrieved articles. Articles were included in the meta-analysis only if they satisfied the following criteria: (a) case–control or cohort studies published as original articles; (b) studies that reported findings expressed as odds ratio, RR or hazard ratio (or reporting sufficient data to compute them) for light drinkers (≤12.5 g/day ethanol; ≤1 drink/day) vs. non-drinkers; (c) studies that reported standard errors or CIs of the risk estimates, or provided sufficient data to calculate them. We thus computed a pooled RR of breast cancer for light drinkers vs. non-drinkers, using random-effects models (DerSimonian and Laird, 1986). Statistical heterogeneity among studies was evaluated using I, which is the proportion of total variation contributed by between-study variance (Higgins and Thompson, 2002). Subgroup analyses and meta-regression models were carried out to investigate potential sources of between-study heterogeneity. We used a χ statistics to test for differences of summary estimates among subgroups (Greenland, 1987).
A total of 3431 papers were retrieved from the literature search. Of these, 113 papers reporting breast cancer risk estimates for light drinkers were included in the meta-analysis. The complete reference list is reported in Appendix.
The analysis included 44,552 cases in the reference category of non-drinkers (40,899 incident cases and 3653 deaths) and 77,539 cases in the light drinkers' category (76,303 incident cases and 1236 deaths). These numbers are slightly underestimated, since for six studies the number of exposed and/or non-exposed cases was not reported. Case–control was the most common study design (64% of studies); 51% of the included studies were from North America, 38% from Europe, 6% from Asia and 10% from other regions or from more than one region; 36% of the reported estimates were adjusted for the main risk factors (age, family history, parity, menopausal status, oral contraceptive/hormonal replacement therapy use), while 16% of the estimates included occasional drinkers in the reference category.
Figure 1 shows the RR of breast cancer in light drinkers vs. non-drinkers in each of the 113 included studies. We found substantial heterogeneity among single study estimates (I = 64%). The random-effect summary RR was 1.04 (95% CI, 1.02–1.07). We did not find evidence of heterogeneity in pooled estimates by design (P = 0.93) and area in which the study was carried out (P = 0.71). Results did not appreciably change from those of the overall analysis when considering only estimates adjusted for the main risk factors (pooled RR = 1.03, 95% CI, 1.00–1.07), as well as including only estimates not considering occasional drinkers in the reference category of non-drinkers (pooled RR = 1.04, 95% CI, 1.01–1.07).
(Enlarge Image)
Figure 1.
RRs of breast cancer for light drinkers vs. non-drinkers. Squares indicate study-specific RRs. Horizontal lines indicate the 95% CIs. Diamond indicates pooled RR with its corresponding 95% CI.
Therefore, this meta-analysis reported a modest but significant association between light drinking and breast cancer. The estimate was based on the results of more than one hundred studies. Women drink less than men (Gronbaek et al., 1994) and therefore low and moderate intakes are usually investigated more frequently and more in detail in women than in men, though the bias due to under reporting of even moderate alcohol consumption may be more relevant for women than for men (Allen et al., 2009). Since several populations show a high prevalence of light drinkers among women, even the small increase in risk we reported—in the order of 5%—represents a major public health issue in terms of breast cancers attributable to alcohol consumption.
Heavy Alcohol Drinking
High levels (i.e. ≥3 drinks/day) of alcohol consumption were associated with increased risk of breast cancer in the largest available studies (Hamajima et al. 2002; Allen et al., 2009; Chen et al., 2011). In the collaborative reanalysis of 53 epidemiological studies on breast cancer (Hamajima et al., 2002), the RRs were 1.32 (95% CI, 1.19–1.45) for consumption of 35–44 g/day and 1.46 (95% CI, 1.33–1.61) for consumption of ≥45 g/day of alcohol, when compared with non-drinkers. The association was similar in never- and ever-smokers. In the Million Women Study (Allen et al., 2009), the RR was 1.29 (95% CI, 1.23–1.35) for the highest level of consumption considered, i.e. ≥21 g/day, after adjustment for smoking and several other covariates. A 51% (95% CI, 35–70%) increase in risk emerged for drinkers of ≥30 g/day vs. non-drinkers in the Nurses' Health Study (Chen et al., 2011). Further, a case–control study of over 2500 women with breast cancer conducted in Italy, in a population characterized by relatively high alcohol drinking, reported an RR of 1.41 (95% CI, 1.17–1.71) for consumption >27 g/day when compared with abstainers (Ferraroni et al. 1998). Therefore, these results consistently indicate a 40–50% elevated risk of breast cancer in women consuming three or more alcohol drinks/day.
Dose–risk Relation
With reference to the dose–risk relation, the collaborative reanalysis on alcohol and breast cancer (Hamajima et al., 2002) found that the RR of breast cancer increased by 7.1% (95% CI, 5.5–8.7%) for each additional 10 g/day of alcohol intake. In the Million Women Study (Allen et al., 2009), the corresponding increase in alcohol consumption was associated to a 12% (95% CI, 9–14%) increased risk of breast cancer. A meta-analysis based on 49 studies and over 44,000 cases gave a dose–risk function which approached a RR of 1.5 at 40 g/day, and further increased for highest levels of intake (Fig. 2) (Bagnardi et al., 2001). A subsequent meta-analysis of about 100 epidemiological studies provided results according to the degree of control for confounding and the overall quality of the studies identified (Key et al., 2006). The increase in risk for additional 10 g/day of alcohol varied between 10 and 13%, according to the inclusion criteria used. Further, the trend in risk was highly significant (P < 0.001). There is, therefore, consistent evidence for a positive dose–risk relation between alcohol drinking and breast cancer.
(Enlarge Image)
Figure 2.
Dose-risk function between alcohol consumption and breast cancer (extracted from Bagnardi et al., 2001).
Breast Cancer Defined by Estrogen and PR Status
Since alcohol consumption might affect the risk of breast cancer through hormone-related mechanisms [such as increased estrogen and androgen levels (Sarkar et al., 2001; Singletary and Gapstur 2001) or increased levels of plasma insulin-like growth factor produced by the liver following alcohol drinking (Yu and Berkel, 1999)], several studies examined the association with breast cancer defined by ER and PR status (Deandrea et al., 2008; Suzuki et al., 2008; Lew et al., 2009; Chen et al., 2011). A meta-analysis summarized information on this issue, including 20 epidemiological studies (i.e. 4 cohort and 16 case–control studies) published up to 2007 (Suzuki et al., 2008). The meta-analysis reported an increased risk of 27% (95% CI, 17–38%) of all ER+ and of 14% (95% CI, 3–26%) of all ER-breast cancers for the highest vs. lowest level of alcohol drinking.
When the data were analyzed according to combined ER and PR status, the corresponding summary RRs were 1.22 for ER +/PR+ (95% CI, 1.11–1.34, based on 15 studies and over 11,000 cases of breast cancer), 1.28 for ER +/PR−(95% CI, 1.07–1.53, 11 studies, ~1900 cases), 1.31 for ER−/PR + (95% CI, 0.99–1.74, 8 studies, 580 cases) and 1.10 for ER−/PR− (95% CI, 0.98–1.24, 15 studies, ~4000 cases). Subsequent investigations further supported a stronger association between heavy alcohol consumption and ER + (and particularly ER +/PR+) breast cancers (Deandrea et al., 2008; Lew et al., 2009; Chen et al., 2011).
