Drinking pattern and time lag of alcohol consumption with colorectal cancer risk in US men and women

Abstract

Background

Association between light to moderate alcohol consumption and colorectal cancer (CRC) incidence remains understudied, especially regarding drinking pattern, beverage type, and temporal aspects.

Methods

Hazard ratios (HRs) and 95% confidence intervals (CIs) for time to CRC diagnosis were estimated among 137 710 participants. Estimates based on remote (eg, >10 years before follow-up) and recent (eg, the preceding 10 years before follow-up) alcohol intake, using different cutoffs (eg, 8, 10, 12 years) and mutual adjustment, enabled separating independent effects and investigating time lag of alcohol–CRC association.

Results

In total, 3599 CRC cases were documented over 3 decades. Light to moderate drinking was associated with an increased CRC risk only in men: HR (95% CI) for 5-14.9 and 15-29.9 vs 0 g/day of alcohol intake was 1.19 (1.01 to 1.41) and 1.38 (1.13 to 1.67). In women, that for 0.1-4.9 and 5-14.9 vs 0 g/day of alcohol was 1.07 (0.96 to 1.20) and 1.05 (0.91 to 1.20). Drinkers with both high drinking frequency and daily intake had the highest CRC risk, suggesting total alcohol intake was the critical factor. We estimated the time lag between alcohol consumption and CRC occurrence to be 8 to 12 years. Former drinkers did not experience a significant reduction in CRC risk even after 10 years of quitting or reducing consumption.

Conclusions

Based on 2 cohorts of health professionals, our findings suggest that the increased risk of CRC associated with alcohol intake is driven mainly by total quantity and remote intake. Former drinkers did not experience an immediate reduction in CRC risk after quitting or reducing consumption.

Introduction

Alcohol is classified as a Group 1 carcinogen by the International Agency for Research on Cancer, indicating sufficient evidence of its carcinogenicity in humans.¹ Similarly, World Cancer Research Fund/American Institute for Cancer Research recommends limiting alcohol consumption for cancer prevention.² Colorectal cancer (CRC) ranks as the third common incident cancer and accounts for 8% of all new cancer diagnoses in the United States in 2023.³ Studies consistently show that alcohol consumption above a moderate level (1 drink/15 g per day for women and 2 drinks/30 g per day for men) increases the risk of CRC.^4-6 However, the relationship between light to moderate consumption and CRC remains unclear, with previous meta-analyses or pooling projects reporting associations ranging from inverse to positive.^7-9 Furthermore, few studies have comprehensively examined the quantity, frequency, and beverage type of alcohol consumption with CRC risk.

Other important questions include the time lag between alcohol consumption and CRC occurrence, as well as how quitting or reducing consumption affects subsequent CRC risk. Investigating these questions has been limited by the availability of repeated diet assessment and appropriate methods. For example, some studies omitted within-person changes by modeling “lifetime average intake,” which prevents the analysis of quitting or reducing consumption.^10,11 Other studies have modeled trajectories of alcohol intake within specific populations and provided insight on CRC risk comparisons between different trajectories,^12,13 but focusing less on the induction period or time lag between alcohol and CRC. Lagged analysis was previously used in studies based on the Nurses’ Health Study (NHS) and the Health Professionals Follow-up Study (HPFS).^14,15 For example, an 8-year lag would be relating alcohol intake in 1980 to cases that occurred between 1988 and 1992. However, the high correlation of alcohol intake over time (eg, in NHS and HPFS, correlation coefficient = 0.7 in women and 0.5 in men for intakes 10 years apart) makes this type of analysis less informative. Therefore, we will focus on an alternative strategy.¹⁶ It sets a hard cutoff for “remote” and “recent” intake (eg, intake within the last 10 years before follow-up as recent and earlier intake as remote), includes both in the model for mutual adjustment, and thus enables identifying independent effects of alcohol intake at different time periods, further assisting in time lag estimation.

In the current study, we aimed to provide new evidence regarding the long-standing controversy of light to moderate alcohol consumption with CRC incidence while addressing the gap in the literature concerning drinking pattern, beverage type, and temporal aspects of alcohol intake in relation to CRC risk, using data from the NHS and HPFS, 2 large prospective cohorts in the United States.

Methods

Study population

NHS was initiated in 1976 and recruited 121 700 female registered nurses aged 30-55 years.¹⁷ HPFS was initiated in 1986 and recruited 51 529 male health professionals aged 40-75 years.¹⁸ Participants were followed biennially through mailed questionnaires to collect information on lifestyle, medical history, and disease diagnoses. The study protocols were approved by the institutional review boards of Brigham and Women’s Hospital and Harvard T.H. Chan School of Public Health, and those of participating registries (as required). All participants provided written consent to retrieve medical records.

The baseline was defined as the year when alcohol intake data were first collected (1980 for NHS and 1986 for HPFS). We excluded participants with a baseline history of cancer (n = 3278) or ulcerative colitis (n = 1397), incomplete food frequency questionnaire (FFQ) or implausible energy intake (n = 29 502), or missing baseline alcohol intake data (n = 254). These criteria resulted in an analytic population of 137 710 participants (90 325 from NHS and 47 385 from HPFS).

Assessment of alcohol intake

Alcohol intake was assessed via a validated FFQ nearly every 4 years since baseline.^19,20 In each FFQ, participants were asked how often they consumed 1 standard serving over the previous year of each alcoholic beverage (eg, liquor, beer, red wine, white wine). Alcohol intake, in grams per day, was calculated for each beverage as daily servings multiplied by the ethanol content for 1 serving of each alcoholic beverage; total alcohol intake was calculated as the sum across all alcoholic beverages. The FFQ-measured alcohol intakes have been validated against multiple diet records (correlation coefficients r = 0.80, 0.81, and 0.83 for liquor, beer, and wine in women, and 0.86, 0.76, and 0.70 in men) and showed the expected linear association with HDL-cholesterol level in blood samples.^20,21 In addition to the amount, participants were also asked the frequency of alcohol use in a typical week in 1986, 1988, 1996, 2000, and 2004 in NHS and in 1986, 1988, 1998, 2002, 2004, 2006, 2008, and 2012 in HPFS. This drinking days per week measure correlated highly with that from diet records as well (r = 0.79).²⁰

Ascertainment of CRC cases

Incident CRC cases were first identified by self-report through the main biennial questionnaire. Unreported lethal CRC cases were then identified through next of kin, cancer registries, death certificates, or the National Death Index. After permission was granted, cohort investigators reviewed medical records and pathology reports to ascertain all CRC diagnoses and extract information on anatomic location.

Statistical analyses

Participants were followed from baseline until the occurrence of CRC, death, or the end of follow-up (June 2018 for NHS and January 2018 for HPFS), whichever came first.

Using the Wald test for the interaction term between sex and alcohol intake, we did not detect statistically significant heterogeneity by sex (except for drinking frequency; statistical significance at P < .05, 2-sided), and thus we performed analyses in each cohort and then in the pooled population to increase power. Cox proportional hazards models with time-varying exposures and covariates were used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) of incident CRC with alcohol intake (cumulative average), drinking frequency, and each alcoholic beverage. Cox models with a duplication method for competing risks were performed to examine alcohol and CRC by subsite (proximal colon, distal colon, and rectal cancer).²² All models were stratified jointly by age in months and calendar time in 2-year groups (and cohort in the pooled population) and were adjusted for race, body mass index (BMI), physical activity, pack-years of smoking, regular use of aspirin or other nonsteroidal anti-inflammatory drugs (NSAIDs), family history of CRC, history of endoscopy (colonoscopy or sigmoidoscopy), Alternative Healthy Eating Index (AHEI) 2010²³ without the alcohol component, intakes of energy, coffee, dietary fiber, total calcium, and total folate, multivitamin use, and menopausal status and hormone use (among women). We tested linearity using the quadratic term of continuous alcohol intakes and drinking days. We tested nonlinearity using restricted cubic splines and likelihood ratio tests comparing the model with and without the cubic spline term.

Considering that illness or underlying cancer symptoms could have influenced the baseline alcohol consumption, we conducted 2 sensitivity analyses to assess the impact. First, we alternatively used the lowest alcohol intake group (0.1-4.9 g/day) instead of nondrinkers as the reference group. Second, we repeated the analyses excluding former regular drinkers, defined as those who reported a significant reduction in alcohol consumption over the past 10 years. Considering the strong association of red meat and processed meat with CRC, even though red meat was already accounted for as a component in AHEI 2010, we further adjusted for red meat and processed meat intake above AHEI and checked the robustness of our main findings.

We conducted lagged analyses assuming years of 4-, 8-, 12-, 16-, and 20-year lag for the association of total alcohol intake with CRC risk. We also applied the alternative strategy to estimate time lag: we categorized alcohol intake into “remote periods (eg, >10 years before follow-up periods)” and “recent periods (eg, the preceding 10 years before follow-up periods)” (Figure 1) and included both remote and recent intake in one model to identify their independent effects after mutual adjustment. Following the same logic, we defined the recent periods as the preceding 6, 8, 12, 14, 16, 18, and 20 years before follow-up periods, estimating the time lag between alcohol and CRC through this series of analyses. Finally, we examined the association between joint categories of remote and recent alcohol intake and the risk of CRC, seeking insights into how changes in long-term alcohol consumption influence the risk. SAS version 9.4 (SAS Institute) was used for all analyses.

Figure 1.

Remote and recent alcohol intake on risk of colorectal cancer: timeline of analyses (10 years as the cutoff). Remote periods are indicated by the darker color, and the subsequent 10-year periods are indicated by the lighter color. In the Nurses’ Health Study (NHS), the first remote period was 1980-1982, the first recent period was 1982-1992, and thus the follow-up for this first cycle was from 1992 to 2018. In the Health Professionals Follow-up Study (HPFS), the first remote period was 1986-1988, the first recent period was 1988-1998, and thus the follow-up for this first cycle was from 1998 to 2018.

Results

We documented 2139 incident CRC cases in NHS and 1460 cases in HPFS. The median follow-up was 38 years in NHS and 31 years in HPFS. At baseline, 32% of women and 24% of men were nondrinkers. The average alcohol intake among drinkers was 9.5 g/day in women and 14.7 g/day (about 1 drink) in men (Figure S1). Compared with nondrinkers, alcohol drinkers tended to be more physically active and regular users of NSAIDs, but they also had a higher level of smoking and lower intake of fiber and calcium (and folate in heavy drinkers), despite the overall higher AHEI score (Table 1).

Table 1.

Baseline characteristics according to total alcohol intake in the NHS (women) and HPFS (men).^a

	Total alcohol intake (g/day)
Characteristic	0	0.1-4.9	5-14.9	15-29.9	≥30
Women (1980), n	28 789	30 368	20 220	6062	4886
Age, years	46.5 (7.3)	45.7 (7.3)	46.5 (7.1)	46.9 (6.9)	47.7 (6.8)
Total alcohol intake, g/day	0 (0)	1.9 (1.2)	9.8 (3.0)	20.7 (4.6)	41.8 (13.2)
From liquor, g/day	0 (0)	0.6 (0.7)	3.9 (4.3)	8.1 (5.5)	20.6 (18.4)
From beer, g/day	0 (0)	0.2 (0.5)	1.6 (3.0)	3.0 (4.2)	11.0 (17.6)
From wine,b g/day	0 (0)	1.1 (1.2)	4.3 (4.0)	9.6 (8.6)	10.3 (14.3)
Days drinking in a week	NA	2.1 (1.3)	3.6 (2.0)	5.1 (1.9)	5.8 (1.7)
AHEI score excluding alcohol	43.4 (9.5)	43.9 (9.2)	44.2 (9.0)	44.5 (8.9)	42.9 (8.9)
Total energy intake, kcal/day	1563 (517)	1537 (495)	1550 (482)	1622 (484)	1751 (505)
Dietary fiber intake, g/day	14.6 (6.4)	14.3 (5.8)	13.7 (5.5)	13.0 (5.3)	10.9 (4.9)
Total calcium intake, mg/day	753 (344)	751 (312)	723 (289)	683 (268)	597 (264)
Total folate intake, µg/day	367 (293)	369 (278)	365 (254)	365 (257)	344 (251)
Multivitamin use, %	33	34	35	36	36
BMI, kg/m2	25.3 (5.1)	24.5 (4.4)	23.6 (3.7)	23.3 (3.5)	23.5 (3.7)
Physical activity, MET-h/week	12.1 (15.8)	13.7 (18.5)	15.0 (19.5)	15.4 (18.3)	13.1 (15.7)
Never smoker, %	57	44	33	27	19
Current smoker, %	23	27	33	35	53
Pack-years among ever smokers	9.0 (15.3)	10.1 (14.5)	13.2 (16.2)	14.8 (16.7)	23.2 (21.1)
Family history of colorectal cancer, %	18	18	19	18	18
Regular aspirin and NSAIDs use,c %	25	26	28	30	30
Past colonoscopy/sigmoidoscopy, %	10	10	10	10	11
Postmenopausal, %	46	42	43	45	50
Men (1986), n	11 207	11 372	13 090	6043	5673
Age, years	54.4 (9.9)	53.3 (9.9)	53.7 (9.7)	53.6 (9.4)	55.4 (9.5)
Total alcohol intake, g/day	0 (0)	2.2 (1.2)	9.6 (3.0)	20.0 (4.1)	46.0 (16.0)
From liquor, g/day	0 (0)	0.6 (0.7)	3.4 (3.8)	7.5 (5.2)	22.7 (19.8)
From beer, g/day	0 (0)	0.6 (0.6)	3.2 (3.4)	5.7 (4.5)	15.4 (18.1)
From wine, g/day	0 (0)	1.1 (1.0)	3.0 (3.2)	6.8 (6.1)	7.9 (12.8)
From red wine, g/day	0 (0)	0.4 (0.6)	1.1 (1.8)	2.7 (3.6)	3.2 (7.5)
From white wine, g/day	0 (0)	0.7 (0.8)	1.9 (2.5)	4.2 (4.9)	4.7 (9.1)
Days drinking in a week	NA	1.6 (0.6)	3.1 (1.7)	5.0 (1.6)	6.3 (1.1)
AHEI score excluding alcohol	46.8 (11.3)	47.8 (11.0)	47.6 (10.8)	47.4 (10.4)	44.5 (10.3)
Total energy intake, kcal/day	1922 (630)	1920 (615)	1956 (600)	2074 (599)	2215 (609)
Dietary fiber intake, g/day	22.5 (7.9)	22.1 (7.3)	21.1 (6.5)	20.2 (5.9)	17.0 (5.7)
Total calcium intake, mg/day	970 (477)	925 (431)	895 (408)	843 (370)	767 (360)
Total folate intake, µg/day	486 (292)	488 (280)	485 (272)	483 (265)	442 (251)
Multivitamin use, %	41	40	42	44	43
BMI, kg/m2	25.7 (3.6)	25.6 (3.3)	25.4 (3.2)	25.3 (3.0)	25.6 (3.2)
Physical activity, MET-h/week	18.4 (24.6)	19.8 (24.1)	22.0 (24.7)	22.9 (25.1)	20.5 (24.9)
Never smoker, %	62	54	46	39	27
Current smoker, %	7	8	9	10	20
Pack-years among ever smokers	9.9 (17.7)	10.6 (16.9)	12.6 (17.7)	14.3 (17.9)	21.8 (22.3)
Family history of colorectal cancer, %	14	15	15	15	14
Regular aspirin and NSAIDs use,c %	30	30	33	36	38
Past colonoscopy/sigmoidoscopy, %	23	27	27	27	26

a

Values are means (SDs) for continuous variables and percentages for categorical variables. All variables were standardized by age.

b

Red wine and white wine were asked separately in NHS since 1984.

c

Regular user was defined as ≥2 tablets of aspirin (325 mg/tablet) or other NSAIDs per week.

Abbreviations: AHEI = Alternative Healthy Eating Index; BMI = body mass index; HPFS = Health Professionals Follow-up Study; MET = metabolic equivalent of task; NA = not applicable; NHS = Nurses' Health Study; NSAID = nonsteroidal anti-inflammatory drug. 

Light to moderate drinking was associated with an increased CRC risk only in men (Table 2 and Figure S2): HR (95% CI) for 5-14.9 and 15-29.9 vs 0 g/day of alcohol intake was 1.19 (1.01 to 1.41) and 1.38 (1.13 to 1.67). In women, that for 0.1-4.9 and 5-14.9 vs 0 g/day of alcohol was 1.07 (0.96 to 1.20) and 1.05 (0.91 to 1.20). Per 15 g/day estimates seem to suggest a stronger association for men compared with women, and for distal colon (in men) and rectal cancer (both sexes) compared with proximal colon cancer, but the interaction by sex and by subsite did not reach statistical significance (P_interaction = .19 for sex, P_interaction = .19 for subsites in women and 0.22 in men). We also did not observe a significant interaction by age at diagnosis (Table S1). The positive association between total alcohol intake and CRC persisted when alternatively using 0.1-4.9 g/day as the reference group, when former regular drinkers were excluded, when red meat and processed meat intakes were additionally adjusted, and when applying up to a 12-year lag period (Tables S2 and S3).

Table 2.

Association between total alcohol intake and risk of colorectal cancer.

	Total alcohol intake (g/day)
	0	0.1-4.9	5-14.9	15-29.9	≥30	Ptrend	Per 15 g/day
Colorectal cancer
Women
No. of cases	446	949	447	200	97
Multivariable HR (95% CI)	Referent	1.07 (0.96 to 1.20)	1.05 (0.91 to 1.20)	1.13 (0.95 to 1.34)	1.26 (1.00 to 1.59)	.08	1.06 (0.99 to 1.14)
Men
No. of cases	230	436	402	229	163
Multivariable HR (95% CI)	Referent	1.04 (0.88 to 1.22)	1.19 (1.01 to 1.41)	1.38 (1.13 to 1.67)	1.66 (1.33 to 2.07)	<.001	1.17 (1.10 to 1.24)
Pooled
Multivariable HR (95% CI)	Referent	1.06 (0.97 to 1.17)	1.11 (1.00 to 1.23)	1.24 (1.09 to 1.41)	1.45 (1.24 to 1.69)	<.001	1.12 (1.07 to 1.17)
Proximal colon cancer
Women
No. of cases	189	481	216	86	46
Multivariable HR (95% CI)	Referent	1.22 (1.03 to 1.44)	1.15 (0.94 to 1.40)	1.08 (0.83 to 1.41)	1.41 (1.00 to 1.98)	.68	1.02 (0.92 to 1.13)
Men
No. of cases	82	151	134	81	54
Multivariable HR (95% CI)	Referent	0.97 (0.74 to 1.27)	1.02 (0.77 to 1.35)	1.20 (0.87 to 1.66)	1.31 (0.90 to 1.90)	.07	1.10 (0.99 to 1.21)
Pooled
Multivariable HR (95% CI)	Referent	1.14 (0.99 to 1.32)	1.11 (0.94 to 1.31)	1.17 (0.95 to 1.43)	1.41 (1.10 to 1.80)	.08	1.07 (0.99 to 1.15)
Distal colon cancer
Women
No. of cases	129	257	123	58	23
Multivariable HR (95% CI)	Referent	1.08 (0.87 to 1.33)	1.06 (0.82 to 1.36)	1.23 (0.89 to 1.69)	1.04 (0.65 to 1.65)	.87	1.01 (0.89 to 1.15)
Men
No. of cases	59	119	107	68	53
Multivariable HR (95% CI)	Referent	1.17 (0.85 to 1.60)	1.32 (0.95 to 1.83)	1.71 (1.18 to 2.47)	2.17 (1.45 to 3.26)	<.001	1.24 (1.12 to 1.37)
Pooled
Multivariable HR (95% CI)	Referent	1.11 (0.93 to 1.32)	1.15 (0.94 to 1.40)	1.41 (1.11 to 1.79)	1.55 (1.16 to 2.06)	.002	1.14 (1.05 to 1.23)
Rectal cancer
Women
No. of cases	112	179	94	45	24
Multivariable HR (95% CI)	Referent	0.82 (0.64 to 1.04)	0.87 (0.66 to 1.16)	1.01 (0.71 to 1.46)	1.21 (0.76 to 1.93)	.02	1.18 (1.03 to 1.35)
Men
No. of cases	48	86	87	40	35
Multivariable HR (95% CI)	Referent	0.99 (0.69 to 1.41)	1.26 (0.88 to 1.82)	1.17 (0.75 to 1.83)	1.81 (1.12 to 2.92)	.004	1.20 (1.06 to 1.37)
Pooled
Multivariable HR (95% CI)	Referent	0.86 (0.71 to 1.05)	1.00 (0.80 to 1.24)	1.03 (0.78 to 1.36)	1.39 (1.01 to 1.92)	<.001	1.18 (1.08 to 1.29)

The multivariable models were stratified jointly by age in months and calendar time in 2-year groups (and cohort in the pooled population), and additionally adjusted for race (White, non-White); body mass index (continuous); physical activity (continuous); pack-years of smoking (continuous); regular use of aspirin or other nonsteroidal anti-inflammatory drugs (yes, no); family history of colorectal cancer (yes, no); past history of colonoscopy or sigmoidoscopy (yes, no); Alternative Healthy Eating Index 2010 score excluding alcohol (continuous); intakes of energy, coffee, dietary fiber, total calcium, and total folate (all continuous); and multivitamin use (yes, no), as well as hormone use (premenopausal, never, past, current users) among women. Cumulative averages of total alcohol intake and dietary covariates were used to better represent long-term diet and reduce measurement errors. CI = confidence interval; HR = hazard ratio.

In women, drinking frequency was not associated with CRC risk, whereas in men, drinking 5-7 days per week was associated with a higher CRC risk than drinking 1-2 days per week (P_interaction = .01 for sex) (Table S4 and Figure S3). Each 1-day increment was associated with a 5% higher risk of overall CRC among male drinkers. Daily alcohol intake increased the risk among frequent drinkers (female drinkers who consumed ≥3 days and male drinkers who consumed ≥5 days per week) (Figure 2 and Table S5). The risk tended to increase with the combination of more frequent consumption and high daily intake, suggesting that overall quantity consumed was the critical factor.

Figure 2.

Dose-response relationship between daily dose and risk of colorectal cancer (CRC) by drinking frequency strata. Multivariable (MV) models were adjusted for the same set of covariates as in Table 2. CI = confidence interval; HR = hazard ratio.

The positive association between total alcohol intake and CRC was driven mainly by the consumption of liquor and beer rather than wine (Table S6). In the pooled population, 15 g/day increment of liquor and beer was associated with a 12% and 22% higher risk of CRC, respectively, whereas the associations for wine were nonsignificant. The estimates remained similar when drinking frequency and socioeconomic status were additionally adjusted (Table S7).

We observed the association of remote alcohol intake with CRC risk was the strongest, whereas that of recent intake was the weakest and close to 1 when using a cutoff as 8-12 years for definition of remote and recent intake, suggesting the time lag between drinking alcohol and change in CRC risk was around this period (Figure 3 and Table S8). The time lag was similarly observed in women and men (Figure S4). It might vary across subsites, though: 10-12 years for proximal colon cancer, 8 years for rectal cancer (strong remote and weak recent intake association), and perhaps less than 6 years for distal colon cancer (recent intake association increasing from the beginning) (Figure S5).

Figure 3.

Association of remote or recent alcohol intake (≥30 vs 0 g/day) with risk of colorectal cancer (CRC). X indicates the cutoff for remote and recent intake in years. Models were multivariable adjusted for the same set of covariates as in Table 2, and mutually adjusted for remote and recent intake (continuous). CI = confidence interval; HR = hazard ratio.

Finally, when participants were jointly categorized according to remote and recent alcohol intake using a 10-year cutoff (midpoint of the proposed time lag), the majority maintained a stable intake level, with a portion quitting or reducing their alcohol consumption, whereas a small number began drinking later in life. We found the risk of CRC was influenced mainly by the remote intake; people who drank 15+ g/day in the past exhibited the highest risk increase on average (Table 3). Drinkers did not experience a significant reduction in CRC risk after quitting or reducing consumption. For example, those who previously drank 0.1-14.9 g/day but had quit for 10 years before follow-up had a HR (95% CI) of 1.07 (0.90 to 1.26) compared with nondrinkers, and 0.94 (0.82 to 1.09) compared with who continued drinking 0.1-14.9 g/day. Using various cutoffs (Table S9), people who previously drank 0.1-14.9 g/day did not experience a significant reduction of CRC risk compared with those who kept drinking until they had quit for at least 14 years, 0.79 (0.65 to 0.96), and those who previously drank 15+ g/day did not see a significant risk reduction even after 18 years of quitting. On the contrary, individuals who began drinking alcohol in the recent periods experienced an increased risk of CRC. These findings should be interpreted cautiously with the limited number of individuals in some categories.

Table 3.

Joint association of remote and recent alcohol intake with risk of colorectal cancer in the pooled population (cutoff = 10 years).

Remote to recent	Cases	Multivariable HR (95% CI)	Multivariable HR (95% CI)	Multivariable HR (95% CI)
Nondrinkers all the time	337	Referent	0.89 (0.78 to 1.01)	0.68 (0.58 to 0.81)
Nondrinkers to 0.1-14.9 g/day	73	1.29 (0.99 to 1.66)	1.14 (0.90 to 1.45)	0.88 (0.67 to 1.14)
Nondrinkers to 15+ g/day	3	2.94 (0.94 to 9.20)	2.60 (0.83 to 8.12)	2.01 (0.64 to 6.29)
0.1-14.9 g/day to nondrinkers	235	1.07 (0.90 to 1.26)	0.94 (0.82 to 1.09)	0.73 (0.61 to 0.88)
0.1-14.9 g/day all the time	997	1.13 (0.99 to 1.28)	Referent	0.77 (0.67 to 0.89)
0.1-14.9 g/day to 15+ g/day	69	1.06 (0.81 to 1.39)	0.94 (0.74 to 1.20)	0.73 (0.56 to 0.95)
15+ g/day to nondrinkers	16	1.38 (0.83 to 2.30)	1.23 (0.74 to 2.01)	0.95 (0.57 to 1.57)
15+ g/day to 0.1-14.9 g/day	120	1.27 (1.02 to 1.57)	1.12 (0.93 to 1.36)	0.87 (0.70 to 1.07)
15+ g/day all the time	291	1.46 (1.23 to 1.74)	1.30 (1.13 to 1.49)	Referent

Multivariable models were adjusted for the same set of covariates as in Table 2. Alcohol intake was categorized into “remote (eg, >10 years before follow-up)” and “recent (eg, the preceding 10 years before follow-up).” With the 10-year cutoff, the “0.1-14.9 g/day to nondrinkers” joint category would refer to participants who previously drank 0.1-14.9 g/day of alcohol but had quit drinking in the recent 10 years before follow-up. The 3 columns of multivariable hazard ratio (HR) (95% confidence interval [CI]) were referenced to nondrinkers throughout the study period, those who consistently consumed 0.1-14.9 g/day, and those who consistently consumed 15+ g/day, respectively.

Discussion

In 2 large prospective cohorts, we observed a higher risk of CRC in light to moderate alcohol drinkers compared with nondrinkers in men but not in women. Drinkers with both high drinking frequency and daily intake had the highest CRC risk, suggesting total alcohol intake was the critical factor. The positive association between alcohol consumption and CRC was likely driven by liquor and beer rather than wine. Notably, overall CRC risk was influenced mainly by alcohol intake 8-12 years before follow-up, whereas recent intake had less impact. Former drinkers did not experience a significant reduction in CRC risk even after 10 years of quitting or reducing consumption.

There is substantial heterogeneity in previous literature regarding the association between light to moderate alcohol consumption and CRC risk. A meta-analysis of 16 cohort studies found a 4% higher risk among light drinkers (≤1 drink/day) and 10% higher risk among moderate drinkers (1-2 drinks/day), with slightly larger magnitude in men than women (light vs nondrinking: HR (95% CI) = 1.02 (0.98 to 1.06) for women and 1.06 (1.01 to 1.11) for men; moderate vs nondrinking: 1.04 (0.95 to 1.13) for women and 1.19 (1.06 to 1.35) for men).⁸ Our findings generally aligned with these results, although our study population was mainly White, whereas 4 of their 16 included studies were conducted in Asia. There has not been direct evidence of alcohol metabolism-related gene polymorphism modifying the association. For example, a 0.5 million-sized cohort study in China found the gene polymorphism modified associations between alcohol consumption and esophageal cancer, but not for CRC.²⁴ A meta-analysis of 16 case-control studies in North America and Europe reported an 8% lower risk of CRC among who drank <2 drinks/day than nondrinkers,⁹ but they focused on current instead of lifetime consumption and included former drinkers in the reference group. A pooled analysis of 8 cohorts in similar regions reported a null association for <2 drinks/day.⁷ The maximum follow-up of included cohorts ranged from 6 to 16 years, with the latest follow-up ending in 1998. In contrast, our study provides data for extended follow-up, which might be necessary to observe the effect of alcohol on CRC risk. The larger magnitude of alcohol–CRC association in men compared with women in our studies also aligns with previous literature. Several factors have been proposed to explain this difference, including generally lower alcohol intake, larger benefits of alcohol consumption on insulin and inflammation in women, as well as the protective effect of estrogen on CRC.^21,25

Few studies have examined drinking pattern and CRC risk. One particularly comprehensive study used the Korean National Health Insurance System database and investigated both drinking frequency and daily alcohol intake.²⁶ With 0.3 million gastrointestinal cancer cases, the study similarly found frequency was associated with an increase in CRC risk among men but not women. Additionally, their joint analyses suggested that daily intake was most impactful among more frequent drinkers (5-7 days per week) for overall gastrointestinal cancer, implying the overall quantity consumed is most important. Regarding the type of alcoholic beverages, an earlier European study found an increase in CRC risk only for beer but not for wine consumption.¹⁰ They did not find an association for liquor, possibly due to its low level of intake (population mean <2 g/day). The first metabolite of ethanol, acetaldehyde, is carcinogenic. Acetaldehyde present in epithelial cells may facilitate cancer development through influencing genetic abnormality (eg, mutation, impaired DNA synthesis due to folate depletion), epigenetic dysregulation (eg, DNA hypomethylation), and tumor microenvironment (eg, inflammation, oxidative stress).⁶ Polyphenols are known for anti-inflammatory and antioxidative potential.²⁷ Variations in polyphenol levels (wine > beer > liquor) might contribute to the differences across alcoholic beverages, because polyphenols counteract the carcinogenic effects of ethanol.²⁸

Our data implies limited benefits for CRC prevention within 10 years after quitting or reducing alcohol consumption. A case-control study found the duration of cessation was inversely associated with CRC risk, and former drinkers experienced a significant reduction in risk only after 15 years after quitting.²⁹ Previous analyses in our cohorts revealed that individuals who reduced their alcohol intake to a low level (<15 g/day) in mid-adulthood after high intake in early adulthood had a similar CRC risk to those who maintained a high alcohol intake.³⁰ They still had a 35% higher risk than nondrinkers in both life stages. A trajectory analysis similarly showed a nonsignificant reduction in CRC risk by decreasing alcohol intake from early/mid-adulthood to late adulthood, whereas men who increased their alcohol intake from moderate to high levels during adulthood experienced an elevation in CRC risk.¹²

A major strength of our cohorts is the repeated assessment of diet, which allows our study to investigate the time lag and changes in alcohol consumption in relation to CRC risk. Importantly, by employing an analytic strategy that sets a hard threshold for remote and recent intake and enables mutual adjustment, we could separate associations over time periods despite the correlation in alcohol consumption habits. Through a series of analyses setting different thresholds and observing changes in association estimates for remote and recent intake, we were able to approximate the time lag. The collection of detailed data on drinking frequency and beverage type provided us the opportunity to contribute new insights to this less investigated area. Several limitations should also be considered. First, the observational nature of our study prevents the establishment of causal relationships. However, the FFQ-measured diet has been validated against diet records, comprehensive collection of covariates reduced the impact of confounding, and sensitivity analyses suggested little evidence of reverse causation. Second, the majority of our study population consisted of White health professionals with generally healthy lifestyles and moderate alcohol consumption. However, a previous evaluation has demonstrated that exposure-disease findings from our cohorts had good generalizability to different populations.³¹ Third, the number of CRC cases in the remote-recent joint analyses was limited in some categories, and the analyses for CRC subsites accounting for competing risks were also underpowered. Therefore, these results should be interpreted cautiously and warrant replication.

Our findings indicate higher alcohol consumption, especially in the form of liquor or beer compared with wine, was associated with an increased risk of CRC. The time lag between alcohol consumption and CRC occurrence might be around 10 years, and former drinkers did not experience an immediate reduction in risk after quitting or reducing consumption.

Acknowledgments

We would like to thank the participants and staff of the NHS and HPFS for their valuable contributions. We would also like to acknowledge the contribution to this study from central cancer registries supported through the Centers for Disease Control and Prevention’s National Program of Cancer Registries (NPCR) and/or the National Cancer Institute’s Surveillance, Epidemiology, and End Results (SEER) Program. Central registries may also be supported by state agencies, universities, and cancer centers. Participating central cancer registries include the following: Alabama, Alaska, Arizona, Arkansas, California, Colorado, Connecticut, Delaware, Florida, Georgia, Hawaii, Idaho, Indiana, Iowa, Kentucky, Louisiana, Massachusetts, Maine, Maryland, Michigan, Mississippi, Montana, Nebraska, Nevada, New Hampshire, New Jersey, New Mexico, New York, North Carolina, North Dakota, Ohio, Oklahoma, Oregon, Pennsylvania, Puerto Rico, Rhode Island, Seattle SEER Registry, South Carolina, Tennessee, Texas, Utah, Virginia, West Virginia, and Wyoming.

Author contributions

Xinyi Li, PhD (Conceptualization; Formal analysis; Methodology; Writing—original draft), Jinhee Hur, PhD (Conceptualization; Funding acquisition; Validation; Writing—review & editing), Yin Zhang, MS (Methodology; Writing—review & editing), Mingyang Song, MD, ScD (Funding acquisition; Methodology; Writing—review & editing), Stephanie A. Smith-Warner, PhD (Writing—review & editing), Liming Liang, PhD (Writing—review & editing), Kenneth J. Mukamal, MD (Writing—review & editing), Eric B. Rimm, ScD (Writing—review & editing), and Edward L. Giovannucci, MD, ScD (Conceptualization; Funding acquisition; Supervision; Writing—review & editing).

Supplementary material

Supplementary material is available at JNCI: Journal of the National Cancer Institute online.

Funding

The Nurses’ Health Study is supported by NIH grants UM1 CA186107 and R01 CA49449. The Health Professionals Follow-up Study is supported by NIH grant U01 CA167552. This work was in addition supported by American Cancer Society Clinical Research Professor grant CRP-23-1014041 (to ELG), NIH grant R00 CA215314 (to MS), American Cancer Society Mentored Research Scholar grant MRSG-17-220-01-NEC (to MS), a research grant from the Ottogi Ham Taiho Foundation (to JH), and the National Research Foundation of Korea grant RS-2024-00349274 funded by the Ministry of Science and ICT (to JH). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. The authors assume full responsibility for the analyses and interpretation of these data. The funding sources played no role in the study design, data collection, data analysis, and interpretation of results, or the decisions made in preparation and submission of the article.

Conflicts of interest

All authors have completed the ICMJE uniform disclosure form at www.icmje.org/disclosure-of-interest/ and declare support from the National Institutes of Health, American Cancer Society, Ottogi Ham Taiho Foundation, and National Research Foundation of Korea for the submitted work; no financial relationships with any organizations that might have an interest in the submitted work in the previous 3 years; no other relationships or activities that could appear to have influenced the submitted work.

Data availability

Data described in the article, code book, and analytic code will be made available upon request pending approval by the Channing Division of Network Medicine at Brigham and Women’s Hospital and Harvard Medical School. Further information including the procedures to obtain and access data from the Nurses’ Health Study and the Health Professionals Follow-up Study is described at https://www.nurseshealthstudy.org/researchers (contact e-mail: [email protected]) and https://sites.sph.harvard.edu/hpfs/for-collaborators/.