Prostate cancer (PCa) is the second most common cancer among men worldwide [1
]. In 2018, worldwide, approximately 1.3 million individuals were diagnosed with PCa with an incidence rate of 29.3 cases per 100,000 persons per year, and 360,000 patients died of PCa with mortality rate of 7.6 deaths per 100,000 persons per year [2
]. Yet, only a few risk factors for prostate cancer are considered established, including body fatness and adult attained height [3
]. Alcohol is widely consumed and is known to be a major risk factor for several types of cancers, including oropharynx, larynx, esophagus, liver, colon, rectum, and breast cancer [4
]. Yet interestingly, it is still under debate whether alcohol consumption is associated with the risk of PCa or not.
Associations between alcohol intake and PCa risk have been studied by various researchers. Several studies found positive associations [5
], others found inverse associations [8
], and some found no associations [10
]. By alcohol type, liquor intake was positively associated with total PCa risk but wine and beer intakes were not in a cohort study [6
]. On the contrary, a case-control study on total PCa found an inverse association with red wine intake but no association with beer and liquor intakes [13
]. By PCa type, total alcohol intake was inversely associated with fatal PCa but not with advanced PCa in one study [14
], while it was associated with a decreased risk of both aggressive and fatal PCa in another study [8
With these inconsistent results, several meta-analyses have been conducted, including the latest one published in 2016 [15
]. This latest meta-analysis suggested that alcohol intake adversely influences prostate cancer outcomes in a dose-response manner. However, the study did not explore the shape of the dose-response relationship and did not account for heterogeneity by types of alcoholic beverage and clinical heterogeneity of PCa. In addition, more studies were published thereafter, including a large cohort study based on 47,568 participants and 869 cases [9
]. To provide further insights regarding the effect of alcohol intake on prostate carcinogenesis, we conducted linear and non-linear dose-response meta-analyses by types of alcoholic beverages (wine, beer, liquor) and PCa (non-aggressive, aggressive).
In this dose-response meta-analysis of alcohol intake and PCa risk by types of alcohol and PCa, consistency of evidence for a positive association between the primary and secondary analyses was more pronounced with non-aggressive PCa than with aggressive PCa. For non-aggressive PCa, by alcohol type, the risk increased linearly with liquor (approximately 4% per every increase of 14 g/day intake) and non-linearly with beer (approximately 3–5% significantly increased risk observed across 9–32 g/day of drinking). Wine was not significantly associated with the risk of non-aggressive PCa. For aggressive PCa, non-linear relationship was indicated for all types of alcohol in the sensitivity analysis that was restricted to studies that provided results for both non-aggressive and aggressive PCa. With liquor intake, approximately 3–17% increased risk was observed across 2–37 g/day of drinking; and with wine intake, approximately 35–77% increased risk was observed across 50–67 g/day of drinking. Interestingly, with beer, approximately 4–21% decreased risk was observed across 2–32 g/day of drinking.
The latest meta-analysis in 2016 found a trend of increasing PCa adverse outcomes with increasing total alcohol intake [15
]. On the contrary, this updated meta-analysis observed no association between total alcohol intake and PCa outcomes. These inconsistent findings may be explained by multiple methodological factors. First, while our study included only cohort studies, the previous study included case-control studies as well, which are susceptible to recall bias or selection bias. Second, the previous meta-analysis used a fixed-effects model while we employed a random-effects model to better account for potential heterogeneity. When we reran the linear meta-analysis using a fixed-effects model, a significant positive association emerged between total alcohol intake and non-aggressive PCa (RR = 1.02, 95% CI = 1.01–1.03). Third, outcome of the previous meta-analysis total PCa including mortality and/or morbidity from PCa while our study analyzed non-aggressive PCa and aggressive PCa separately. In view of the heterogenous associations we found by type of alcohol and PCa, alcohol appears to play a multifaceted role in the development and progression of PCa.
Although our meta-analysis examined the effect of pre-diagnosis alcohol consumption on PCa risk, analysis on the effect of change in dose of drinking from pre-diagnosis to post-diagnosis of PCa on survival can offer further insight into the role of alcohol across the entire prostate carcinogenesis.
According to a cohort study in Canada [36
], high drinking in both pre- and post-diagnosis was associated with an approximately two-fold increased risk of PCa-specific mortality compared to no drinking in both time windows. Interestingly, any level of pre-diagnosis alcohol consumption, regardless of whether individuals stop or reduced drinking post-diagnosis, increased the risk of PCa-specific mortality. Therefore, this study, along with our findings, suggests overall harmful effects of alcohol in prostate carcinogenesis.
Albeit the biological mechanism underlying a harmful effect of alcohol intake on prostate carcinogenesis is not fully elucidated, several possible mechanisms have been suggested. First, acetaldehyde, the first metabolite of ethanol, promotes oxidative stress, which damages DNA directly or indirectly by producing DNA adducts, all of which contribute to PCa carcinogenesis [37
]. Second, prostatitis is known to increase the risk of PCa [38
] and alcohol enhances inflammation. Furthermore, chronic inflammation is known to create an immunosuppressive environment that negate antitumor immunity [39
]. This provides an advantage for tumor formation and progression [41
]. In our study, an adverse effect of alcohol intake was most evident when ingested as liquor, with a positive association emerging starting from a low dose across the wide range and manifesting with both non-aggressive and aggressive PCa. On the contrary, when alcohol was ingested as wine, a significant positive association with aggressive PCa arose particularly at high doses (50–67 g/day of drinking). Lack of a positive association at lower doses might suggest that our body can tolerate a certain amount of alcohol but exceeding this can cause harm. Alternatively, it might be in part explained by anti-cancer effect of polyphenols in wine [42
], which may outweigh the carcinogenic effect of alcohol. As antioxidants, polyphenols repair oxidative DNA damages [43
], and reduce reactive oxygen species (ROS) thereby decreasing prostatic tissue exposure to alcohol-generated ROS [44
]. Polyphenol may also mitigate alcohol-induced inflammation by modulating activities of proinflammatory enzymes and inflammatory cells [45
]. Such potential cancer-preventive benefits might be able to counteract the adverse effect of alcohol at lower intakes of wine, but not at higher intakes.
Interestingly, an inverse association was suggested between beer intake and aggressive PCa. The possible biological mechanism might be related to sex hormone level. An experiment revealed that a repeated ingestion of alcohol reduces testosterone level in normal men [46
]. Particularly beer, containing barley and hops as main ingredients, contains phytoestrogen and polysaccharides that induce prolactin increase [47
]. Although the content of phytoestrogen in beer is low, its concentration can be 10-fold increased by human intestinal microbiome [49
]. In addition, phytoestrogens are known to be converted into biologically active derivatives by intestinal microbiota through de-glycosylation and metabolization [50
]. Although the mechanism of the polysaccharide to increase prolactin is not illuminated, several studies found the same result [48
]. These compounds increase estrogen level which ultimately decrease testosterone level. Furthermore, a recent meta-analysis has proven that low concentration of circulating free testosterone is associated with reduced risk of PCa [53
]. The author suggested that low level of circulating testosterone reduces androgen receptor signaling, which leads to lowered risk of PCa.
By types of PCa, positive association was more consistently observed with non-aggressive PCa than with aggressive PCa. This could be due to methodological bias. Men who consume great quantity of alcoholic beverages might be aware of their higher risk to various diseases. As a result, they could undergo medical screening more frequently than moderate drinkers, which helps them to discover asymptomatic or latent PCa. Thus, heavy drinkers may have had more chance to be diagnosed with non-aggressive cancer. Further studies should account for screening practices and detection bias when examining alcohol intake and PCa.
This study has several strengths. First, by conducting meta-analyses by types of alcohol and PCa, we accounted for potential heterogeneity in the physiological mechanisms of alcoholic beverages and etiology of PCa. Indeed, we observed differential associations between alcohol intake and PCa, which would have been masked if analyzed altogether. Second, by conducting dose-response meta-analysis, we identified the shape of the relationships and specifically quantified the amount of alcohol associated with a risk level. Third, we only included cohort studies, which are less prone to biases such as recall bias and selection bias compared to case-control studies. Lastly, by conducting sensitivity analysis restricted to only studies that provided results for both non-aggressive PCa and aggressive PCa, we were able to examine etiologic heterogeneity in the associations with alcohol, after controlling for the undue effect of the number of studies on statistical power.
Yet, our study does have limitations. First, like any other meta-analyses, the validity of our results is influenced by methodological limitations of each original study included. For instance, Prostate-Specific Antigen (PSA) screening may serve as an important confounder for four studies out of 12 studies did not control for the confounding. However, in our subgroup analyses, the results remained consistent regardless of adjustment for confounding by PSA screening (data not shown). It is also notable that 10 out of 11 included studies used non-drinkers as the reference group. Considering that non-drinkers could include not only lifelong abstainers but also individuals quitted drinking due to underlying diseases, reverse causation could have biased our meta-analysis results such as distorting the strength or shape of dose-response relationships. Future studies on alcohol intake and PCa are advised to use moderate drinking as the reference category. Measurement errors in alcohol intake might have compromised the validity of our results. However, as we included only cohort studies, the errors are likely to be random errors that generally attenuates effect sizes.
Second, as we examined associations between alcohol intake and PCa by type of alcoholic beverages and stage of PCa at diagnosis, the number of studies included in each meta-analysis was limited and thus, no extensive subgroup analyses were performed with respect to other potential modifiers. Yet, heterogeneity as estimated by I2
values was generally low. Third, analyses by type of PCa led to exclude cohort studies with total PCa endpoint. However, when a meta-analysis of total PCa comparing highest versus lowest intake of total alcohol intake was performed among 11 studies included in our study, our result (RR = 1.11, 95% CI = 1.01–1.22) was consistent with the result of previous equivalent meta-analysis (RR = 1.08, 95% CI = 1.04–1.12) [15
], which shows representativeness of studies included in our meta-analysis. Finally, statistical significance of non-linearity of aggressive PCa with beer and liquor was sensitive to inclusion of a study [9
], whose population consisted of health-conscious health professionals with lower level of drinking compared to other study populations included. In addition, because studies included in the meta-analysis contributed less data toward the upper end of alcohol intakes observed, we cannot rule out the possibility that stronger associations observed at lower intakes of beer and liquor than at higher intakes might be in part driven by a few unstable data at higher doses.