Alcohol Consumption and the Risk of Colorectal Adenoma
Alcohol Consumption and the Risk of Colorectal Adenoma
Figure 1 showed the selection process for the studies involved in this meta-analysis. The initial search yielded 455 studies. After screening of the titles and abstracts, 254 of these were excluded due to irrelevant topics. Following the evaluation of the full text, a total of 25 studies were included in our study. Two cohort studies and 23 case–control studies concerning CRA incidence and alcohol intake published between 1990 and 2013 were identified, among which nine studies were from Asia (all were conducted in Japan), five from Europe, and 11 from the US, as shown in Table 1 and Table 2 . In addition, six studies reported RRs for colon adenoma, and three for rectum. In terms of size, there were only two studies presented the data. Meanwhile, the data concerning risk level of adenoma were available in one study. Considering the limited number, the size and risk level of CRA were not conducted in the analyses.
(Enlarge Image)
Figure 1.
Flow chart of the study selection process.
The pooled random effects RRs for comparison with non-/occasional drinkers were as follows: all drinkers, 1.16 (95% CI 1.10–1.22); light drinkers, 1.07 (95% CI 1.02–1.13); moderate drinkers, 1.23 (95% CI 1.15–1.32); heavy drinkers, 1.37 (95% CI 1.26–1.49) and past drinkers, 1.13 (95% CI 1.05–1.22) ( Table S1 ).
Figure 2 and Table S1 presented RRs for CRA risk and alcohol intake from 23 case–control and two cohort studies. In terms of type of studies, the summary RRs of CRA were 1.17 (95% CI 1.11–1.22) in case–control studies and 1.08 (95% CI 0.88–1.33) in cohort studies respectively, in comparison between all drinkers and non-/occasional drinkers. And there was a statistically significant heterogeneity among studies (I = 43.70%, P = 0.011).
(Enlarge Image)
Figure 2.
Pooled risk estimates for colorectal adenoma incidence for all drinkers vs. non-/occasional drinkers according to type of studies (case–control and cohort studies).
Figure 3 indicated RRs estimated for CRA incidence in female (1.03, 95% CI 0.95–1.10) and male (1.19, 95% CI 1.07–1.32) individually, in the comparison between all drinkers and nondrinkers or occasional drinkers (I = 24.00%, P = 0.208). In addition, it had been no significant difference in CRA risk between female and male among light (I = 0.00%, P = 0.711), moderate (I = 30.90%, P = 0.152) and heavy (I = 0.00%, P = 0.474) drinkers, compared with non-/occasional drinkers ( Table S1 ).
(Enlarge Image)
Figure 3.
Pooled risk estimates for colorectal adenoma incidence for all drinkers vs. non-/occasional drinkers according to gender (male and female).
Figure 4 and Table S1 showed summarised RRs for CRA risk stratified by geographical region: Asia (1.20, 95% CI 1.12–1.28), Europe (1.24, 95% CI 1.12–1.36) and the US (1.12, 95% CI 1.05–1.20) respectively. Apparently, the risk in European studies was higher than them in the US and Asia. And the difference existed in the pooled analysis of all drinkers (I = 43.70%, P = 0.011), light drinkers (I = 33.00%, P = 0.064) and moderate drinkers (I = 52.10%, P = 0.004), compared with non-/occasional drinkers.
(Enlarge Image)
Figure 4.
Pooled risk estimates for colorectal adenoma incidence for all drinkers vs. non-/occasional drinkers according to geographical region (Asia, Europe and USA).
In Figure 5, RRs evaluated for CRA risk in colon and rectum were 1.18 (95% CI 1.08–1.30) and 1.42 (95% CI 1.03–1.96) respectively with no significant heterogeneity (I = 0.00%, P = 0.911). In addition, there was no significant difference in CRA risk between colon and rectum among light (I = 0.00%, P = 0.589), moderate (I = 0.00%, P = 0.873), heavy (I = 0.00%, P = 0.535) and past (I = 0.00%, P = 0.877) drinkers, compared with non-/occasional drinkers ( Table S1 ).
(Enlarge Image)
Figure 5.
Pooled risk estimates for colorectal adenoma incidence for all drinkers vs. non-/occasional drinkers according to site of adenoma (colon and rectum).
Detailed evaluation of publication bias suggested a statistically significant evidence of publication bias in the analysis of all drinkers vs. non-/occasional drinkers (Begg's Test, P = 0.030) (Figure S1). In addition, the shape of the funnel plots for the studies on the light alcohol category and CRA risk seemed symmetrical, and Begg's adjusted rank correlation test showed no statistical evidence of publication bias (P = 0.154) (Figure S2). Furthermore, according to Figure S3, the publication bias of moderate category is unlikely (Begg's test, P = 0.144).
(Enlarge Image)
Figure S1.
Funnel plot of studies evaluating the association between all drinkers vs. non-/occasional drinkers (Begg's Test, P = 0.030) and colorectal adenoma risk.
(Enlarge Image)
Figure S2.
Funnel plot of studies evaluating the association between light drinkers vs. non-/occasional drinkers (Begg's Test, P = 0.154) and colorectal adenoma risk.
(Enlarge Image)
Figure S3.
Funnel plot of studies evaluating the association between moderate drinkers vs. non-/occasional drinkers (Begg's Test, P = 0.144) and colorectal adenoma risk.
A sensitivity analysis was performed to investigate the effect of each study on the overall by excluding one at a time and then calculating the summarised RRs for the remainder. There was no change in the direction of the influence when any one study was excluded. For example, when the study by Cho et al. (the study that carried the most weight) was excluded from the analysis, the result of pooled RR of all drinkers remained similar (1.17, 95% CI 1.12–1.22), compared with non-/occasional drinkers. Meanwhile, the summarised RR for light drinkers (1.08, 95% CI 1.03–1.14) and moderate drinkers (1.25, 95% CI 1.16–1.34) also share the same result. This analysis supported the stability of the positive association between alcohol consumption and CRA risk.
Among the second-order fractional polynomial random effects models, the best explanatory variables are dose and quadratic dose. The relationship between alcohol intake and CRA risk was ln (RR) = 0.001292*dose + 0.000035*dose^2. Compared with non-/occasional alcohol consumption, the fractional polynomial model estimates of RR were 1.02 (95% CI, 0.89–1.16), 1.06 (95% CI, 0.92–1.20), 1.16 (95% CI, 1.02–1.33) and 1.61 (95% CI, 1.42–1.84) for 10, 25, 50 and 100 g/day of alcohol intake respectively (Figure 6).
(Enlarge Image)
Figure 6.
Relative risk function and the corresponding 95% confidence interval, describing the best-fitting dose–response association of alcohol drinking (in g/day) and colorectal adenoma incidence.
Results
Study Characteristics
Figure 1 showed the selection process for the studies involved in this meta-analysis. The initial search yielded 455 studies. After screening of the titles and abstracts, 254 of these were excluded due to irrelevant topics. Following the evaluation of the full text, a total of 25 studies were included in our study. Two cohort studies and 23 case–control studies concerning CRA incidence and alcohol intake published between 1990 and 2013 were identified, among which nine studies were from Asia (all were conducted in Japan), five from Europe, and 11 from the US, as shown in Table 1 and Table 2 . In addition, six studies reported RRs for colon adenoma, and three for rectum. In terms of size, there were only two studies presented the data. Meanwhile, the data concerning risk level of adenoma were available in one study. Considering the limited number, the size and risk level of CRA were not conducted in the analyses.
(Enlarge Image)
Figure 1.
Flow chart of the study selection process.
In General
The pooled random effects RRs for comparison with non-/occasional drinkers were as follows: all drinkers, 1.16 (95% CI 1.10–1.22); light drinkers, 1.07 (95% CI 1.02–1.13); moderate drinkers, 1.23 (95% CI 1.15–1.32); heavy drinkers, 1.37 (95% CI 1.26–1.49) and past drinkers, 1.13 (95% CI 1.05–1.22) ( Table S1 ).
Type of the Studies
Figure 2 and Table S1 presented RRs for CRA risk and alcohol intake from 23 case–control and two cohort studies. In terms of type of studies, the summary RRs of CRA were 1.17 (95% CI 1.11–1.22) in case–control studies and 1.08 (95% CI 0.88–1.33) in cohort studies respectively, in comparison between all drinkers and non-/occasional drinkers. And there was a statistically significant heterogeneity among studies (I = 43.70%, P = 0.011).
(Enlarge Image)
Figure 2.
Pooled risk estimates for colorectal adenoma incidence for all drinkers vs. non-/occasional drinkers according to type of studies (case–control and cohort studies).
Gender
Figure 3 indicated RRs estimated for CRA incidence in female (1.03, 95% CI 0.95–1.10) and male (1.19, 95% CI 1.07–1.32) individually, in the comparison between all drinkers and nondrinkers or occasional drinkers (I = 24.00%, P = 0.208). In addition, it had been no significant difference in CRA risk between female and male among light (I = 0.00%, P = 0.711), moderate (I = 30.90%, P = 0.152) and heavy (I = 0.00%, P = 0.474) drinkers, compared with non-/occasional drinkers ( Table S1 ).
(Enlarge Image)
Figure 3.
Pooled risk estimates for colorectal adenoma incidence for all drinkers vs. non-/occasional drinkers according to gender (male and female).
Geographical Region
Figure 4 and Table S1 showed summarised RRs for CRA risk stratified by geographical region: Asia (1.20, 95% CI 1.12–1.28), Europe (1.24, 95% CI 1.12–1.36) and the US (1.12, 95% CI 1.05–1.20) respectively. Apparently, the risk in European studies was higher than them in the US and Asia. And the difference existed in the pooled analysis of all drinkers (I = 43.70%, P = 0.011), light drinkers (I = 33.00%, P = 0.064) and moderate drinkers (I = 52.10%, P = 0.004), compared with non-/occasional drinkers.
(Enlarge Image)
Figure 4.
Pooled risk estimates for colorectal adenoma incidence for all drinkers vs. non-/occasional drinkers according to geographical region (Asia, Europe and USA).
Site of Adenoma
In Figure 5, RRs evaluated for CRA risk in colon and rectum were 1.18 (95% CI 1.08–1.30) and 1.42 (95% CI 1.03–1.96) respectively with no significant heterogeneity (I = 0.00%, P = 0.911). In addition, there was no significant difference in CRA risk between colon and rectum among light (I = 0.00%, P = 0.589), moderate (I = 0.00%, P = 0.873), heavy (I = 0.00%, P = 0.535) and past (I = 0.00%, P = 0.877) drinkers, compared with non-/occasional drinkers ( Table S1 ).
(Enlarge Image)
Figure 5.
Pooled risk estimates for colorectal adenoma incidence for all drinkers vs. non-/occasional drinkers according to site of adenoma (colon and rectum).
Publication Bias
Detailed evaluation of publication bias suggested a statistically significant evidence of publication bias in the analysis of all drinkers vs. non-/occasional drinkers (Begg's Test, P = 0.030) (Figure S1). In addition, the shape of the funnel plots for the studies on the light alcohol category and CRA risk seemed symmetrical, and Begg's adjusted rank correlation test showed no statistical evidence of publication bias (P = 0.154) (Figure S2). Furthermore, according to Figure S3, the publication bias of moderate category is unlikely (Begg's test, P = 0.144).
(Enlarge Image)
Figure S1.
Funnel plot of studies evaluating the association between all drinkers vs. non-/occasional drinkers (Begg's Test, P = 0.030) and colorectal adenoma risk.
(Enlarge Image)
Figure S2.
Funnel plot of studies evaluating the association between light drinkers vs. non-/occasional drinkers (Begg's Test, P = 0.154) and colorectal adenoma risk.
(Enlarge Image)
Figure S3.
Funnel plot of studies evaluating the association between moderate drinkers vs. non-/occasional drinkers (Begg's Test, P = 0.144) and colorectal adenoma risk.
Sensitivity Analyses
A sensitivity analysis was performed to investigate the effect of each study on the overall by excluding one at a time and then calculating the summarised RRs for the remainder. There was no change in the direction of the influence when any one study was excluded. For example, when the study by Cho et al. (the study that carried the most weight) was excluded from the analysis, the result of pooled RR of all drinkers remained similar (1.17, 95% CI 1.12–1.22), compared with non-/occasional drinkers. Meanwhile, the summarised RR for light drinkers (1.08, 95% CI 1.03–1.14) and moderate drinkers (1.25, 95% CI 1.16–1.34) also share the same result. This analysis supported the stability of the positive association between alcohol consumption and CRA risk.
Dose–Response Meta-analysis
Among the second-order fractional polynomial random effects models, the best explanatory variables are dose and quadratic dose. The relationship between alcohol intake and CRA risk was ln (RR) = 0.001292*dose + 0.000035*dose^2. Compared with non-/occasional alcohol consumption, the fractional polynomial model estimates of RR were 1.02 (95% CI, 0.89–1.16), 1.06 (95% CI, 0.92–1.20), 1.16 (95% CI, 1.02–1.33) and 1.61 (95% CI, 1.42–1.84) for 10, 25, 50 and 100 g/day of alcohol intake respectively (Figure 6).
(Enlarge Image)
Figure 6.
Relative risk function and the corresponding 95% confidence interval, describing the best-fitting dose–response association of alcohol drinking (in g/day) and colorectal adenoma incidence.