Alcohol Consumption and Risk of Incident Rheumatoid Arthritis in Women: a Prospective Study

Abstract

Objective

To evaluate the association of alcohol consumption with risk of rheumatoid arthritis (RA) in two large prospective cohorts, the Nurses’ Health Study (NHS) and Nurses’ Health Study II (NHS II).

Methods

The NHS established in 1976 that enrolled 121,701 US female registered nurses. The NHSII began in 1989, enrolling 116,430 female nurses. Lifestyle and environmental exposures have been collected through biennial questionnaires. Alcohol consumption was assessed with a food frequency questionnaire completed every 4 years. The incident RA cases were identified using the connective tissue disease screening questionnaire and a medical record review. Separate Cox proportional hazards models were used to estimate hazard ratios (HRs) after adjusting for potential confounders in NHS and NHSII. Pooled HR from two cohorts was estimated using a DerSimonian and Laird random-effect model.

Results

Among 1.90 million years of person-time from 1980 to 2008, 580 incident RA cases were diagnosed in NHS, and among 1.78 million years of person-time from 1989 to 2009, 323 incident RA cases were diagnosed in NHSII. Compared to no use, the pooled multivariable adjusted HR for alcohol use of 5.0–9.9 grams/day (g/d) was 0.78 (95% CI, 0.61–1.00). For seropositive RA cases, the association appeared stronger (HR= 0.69, 95% CI: 0.50–0.95). In addition, Women who drank beer 2–4 times a week had a 31% decreased risk compared with women who never drank beer.

Conclusion

We found a modest association between long-term moderate alcohol drinking and reduced risk of RA. Future studies are needed to confirm our findings in other populations.

Introduction

Rheumatoid arthritis (RA) is an autoimmune disease that causes a destructive, inflammatory arthritis and affects approximately 1% of the adult population.(1, 2) The prevalence of RA is 3 times higher in women than in men.(3) The cause of RA is unknown, but it is considered to be a multifactorial disease, resulting from the interaction of both genetic and environmental factors. (1) Epidemiologic research has produced convincing data for strong environmental risk factors, including cigarette smoking, (4, 5) exogenous hormone use (6) and female reproductive factors. (7, 8)

Alcohol use has been associated with reduced risk of RA in previous case-control studies. (9),(10–12) These findings have generated substantial public health interest and concern. However, as these case-control studies ascertained alcohol exposure retrospectively, there is the potential for recall bias and reverse causation bias. A recent prospective cohort study also found that moderate consumption of alcohol was associated with reduced risk of RA, but was limited by a small number of cases (197 cases) and a short period of follow-up (from 2003 to 2009).(13) We studied the relationship between alcohol consumption and the risk of RA in two large, prospective cohorts of women with information on alcoholic beverage consumption, as well as important potential confounders such as cigarette smoking and reproductive factors, collected on an ongoing basis for up to 26 years. This approach permitted use of long-term exposure data and resulted in a more accurate estimate of exposure than in previous studies.

Materials and Methods

Population for Analysis

The Nurses’ Health Study (NHS) was established in 1976 and enrolled 121,701 US female registered nurses, ages 30–55 years. NHSII is a prospective cohort study that began in 1989, enrolling 116,430 female nurses aged 25–42 years. Based on self-report, more than 90% of women in both cohorts are Caucasian, reflecting the ethnicity of women entering the nursing profession during the recruitment years. Follow-up rate has been extremely high with only 5.0% of person-time lost to follow-up.(14) In both cohorts, all women completed an initial questionnaire and have been followed biennially to update exposures, lifestyle, health practices and disease diagnoses. We routinely search the National Death Index every 2 years for nonresponders.(15) All aspects of this study were approved by the Partners HealthCare Institutional Review Board.

We included participants in NHS and NHSII cohorts who had alcohol data at baseline (1980 in NHS and 1989 in NHSII). We excluded prevalent cases of RA before 1982 in NHS and 1991 in NHSII. We censored all women who reported psoriasis, psoriatic arthritis and connective tissue diseases, in which the diagnosis was not subsequently confirmed as RA at self-reported date. Women lost to follow-up were censored at their last response to questionnaires because incident cases could not be identified. Additionally, cancer or its treatments may alter immune system, and the secondary symptoms (such as joint pain from paraneoplastic syndromes) may change the behavior of alcohol consumption. We therefore excluded participants with a history of cancer at baseline and censored incident cancer cases at self-reported date in the analysis. The final group included 82,472 women followed from 1980 to 2008 in NHS and 110,737 women followed from 1989 to 2009 in NHSII.

Alcohol Consumption

In NHS, alcohol consumption was assessed within a semi-quantitative food frequency questionnaire (FFQ) including separate items for beer, wine and liquor in 1980, 1984, 1986, 1990, 1994, 1998, 2002 and 2006. In NHS II, alcohol data was collected in 1989 questionnaire and FFQs in 1991, 1995, 1999, 2003 and 2007. We specified standard portions as a glass, bottle, or can of beer; a 4-oz glass of wine; and a shot of liquor. For each beverage, participants were asked to estimate their average consumption over the past year. The estimated alcohol content of each beverage was 13.2 g per bottle or can of beer, 10.8 g per glass of wine, and 15.1 g per standard drink of liquor. Total alcohol intake was recorded as the sum of these 3 beverages. The reproducibility and validity of the assessment of alcohol intake were evaluated among 173 Boston-area participants who completed written one-week dietary records every three months for a year during which time they weighed or measured all their food and drinks. The correlation of alcohol intake on the questionnaire with alcohol intake on the dietary records was 0.9. (16) We measured the alcohol consumption by amount of alcohol (gram) per day and number of drinks per week on each FFQ. We used cumulative average estimates of consumption since they are more representative for regular alcohol exposure than one-time measure to minimize the inaccuracy of exposure information. (17) Cumulative average intake was calculated by averaging repeated alcohol measures from baseline up to 4–6 years before diagnosis for RA cases. As it is possible that early symptoms of RA may alter the behavior of alcohol use, we excluded the most recent FFQ (lagged by one cycle) in calculation of the cumulative average alcohol consumption to reduce the possibility of recall bias and reverse causation bias. For example, for RA incidence during 1998 to 2000 time period, cumulative average alcohol use was obtained by averaging the daily consumption reported from baseline to 1994 FFQ (excluding the 1998 alcohol measure). According to the Dietary Guidelines for Americans, drinking in moderation for women is typically defined as having up to 1 standard drink per day (equal to 10–15 grams of pure alcohol). This definition is referring to the amount consumed on any single day and is not intended as an average over several days. (18) We therefore referred the moderate alcohol consumption as <10g/d on average in the analysis. We coded the alcohol consumption as an ordinal variable: none, <5.0 g/d, 5.0–9.9 g/d, ≥ 10 g/d.

RA Case Identification

RA case identification was a two-stage procedure. The connective tissue disease (CTD) screening questionnaire (CSQ) was first mailed to self-reports of a physician diagnosis of RA. Medical records of those who screen positive were requested and reviewed independently by 2 board-certified rheumatologists to confirm RA according to American College of Rheumatology classification criteria.(19) Women included in these analyses had definite, confirmed incident RA, documented serologic status (rheumatoid factor (RF) and/or anti–cyclic citrullinated peptide (CCP)) and date of diagnosis and symptom onset from the medical records.

Study Covariates

All information about potential confounders was self-reported on the mailed questionnaires administered every 2 years since 1976 in NHS and since 1989 in NHS II and updated through follow-up. We selected covariates based on prior studies in NHS cohorts (5, 7) for RA risk factors, or other studies showing association between diet or lifestyle and alcohol intake. Cigarette smoking is a strong environmental risk factor for RA (5) and highly correlated with alcohol intake. We adjusted for time-varying pack-years of smoking (product of years of smoking and packs of cigarettes per day) in analyses. Updated body mass index was computed for each 2-year time interval by using most recent weight in kilograms divided by height in meters squared. Recreational physical activity was measured biennially with a validated questionnaire asking about the average time spent on 10 common activities. The information was summed and calculated as weekly energy expenditure in metabolic equivalent hours weighting each activity by its intensity level.(20) Reproductive covariates included parity and duration of breastfeeding (nulliparous, parous and no breastfeeding, parous and 1–12 months breastfeeding, parous and >12 months breastfeeding), oral contraceptive use, menopausal status and post-menopausal hormone use. (7) Age at menarche was reported in baseline questionnaire. The total calories intake was measured through FFQ. As a proxy of socioeconomic status, we included the 2000 U.S. Census tract median income for the nurses’ residences.

Statistical Analysis

All analyses were conducted separately in NHS and NHSII. The primary analysis was to assess the influence of alcohol consumption on the incidence of RA. Descriptive statistics such as frequency tables for categorical variables and minimum, maximum, mean and standard deviation for continuous variables were used to summarize the data as well as detect outliers and missing values. Baseline age-adjusted characteristics across categories of alcohol consumption were compared using Kruskal-Wallis tests for continuous variables and Chi-square tests for categorical variables.

For each woman, person-years of follow-up were calculated from the date of return of the baseline questionnaire to the end of follow-up defined as RA diagnosis, censorship due to reported psoriasis or rheumatic disease not confirmed as RA, censorship due to cancer, death, or end of the study, whichever came first. We used Cox proportional hazards models to assess associations after controlling for time-varying covariates including age, smoking, body mass index, physical activity, parity and breastfeeding, menopausal status, oral contraceptive use, total calorie intake and median census-tract household income as well as age at menarche. The multivariable-adjusted hazard ratios (HR) approximating relative risk (RR) for associations between alcohol consumption and RA were calculated along with their 95% confidence intervals (CIs). The possible non-linear association was assessed using a restricted cubic spline method. (21) The interaction terms between alcohol consumption (<5 vs ≥5 g/d) and smoking (<20 vs ≥20 pack-years) were evaluated in the adjusted models using likelihood ratio tests. Analyses were repeated stratified by RA serologic status (RF and/or CCP positive) in both cohorts to evaluate whether the association between alcohol consumption and RA differed by serological status of RA cases. Pooled HR estimates of the two cohorts were calculated using a random effects model, and heterogeneity was assessed using the DerSimonian-Laird method.(22)

To examine the effects of separate alcoholic beverages, we performed a secondary analysis in which we fitted a Cox proportional hazards model that included cumulative average consumption of beer, wine and liquor as well as the potential covariates listed above. We measured the frequency of each alcoholic beverage consumption using ordinal variables (none, ≤1, 2 – 4, ≥5 drinks/week). We also performed sensitivity analyses using the updating alcohol consumption (again excluding the most recent FFQ) rather than the cumulative average measures in time-varying Cox proportional hazards models. Data analyses were performed using SAS 9.2 (SAS Institute, Inc, Cary, North Carolina) with a two-sided significance level of 0.05.

Results

In this study with the exposure measured up to 4–6 years prior to RA onset, 580 incident RA cases were diagnosed among 1.90 million years of person-time from 1980 to 2008 in NHS, and 323 incident RA cases were diagnosed among 1.78 million years of person-time from 1989 to 2009 in NHSII. The age-adjusted incidence rates of RA were 28, 33, 23 and 34/100,000 person-years in NHS, and 21, 18, 17, and 11/100,000 person-years in NHSII across increasing levels of alcohol consumption. Characteristics of the women participating at baseline (1980 in NHS, 1989 in NHSII) are shown in Table 1 according to levels of alcohol intake. Women with higher alcohol intake were more likely to be smokers, oral contraceptive users, have higher income, lower BMI and higher level of physical activities, and less likely to be parous and ever breastfeed more than 12 months. In NHS, 44% of participants were postmenopausal at baseline, while only 2.3% of women in NHSII. The proportion of postmenopausal women who use hormones currently was higher with increased levels of alcohol consumption.

Table 1.

Age-standardized characteristics of Nurses’ Health Study (NHS) participants in 1980 and Nurses’ Health Study II (NHSII) participants in 1989 according to daily alcohol consumption

	Alcohol intake (grams/day) in NHS				Alcohol intake (grams/day) in NHSII
	None	0.1–4.9	5.0–9.9	≥10.0	None	0.1–4.9	5.0–9.9	≥10.0
	No. of participants	26416	27772	8961	19320	41890	46633	11873	10341
Age in years, mean (SD)	50.6(7.2)	50.6(7.2)	50.6(7.2)	50.7(7.1)	34.4(4.6)	34.2(4.7)	34.2(4.8)	34.4(4.7)
Census tract median family income $ (x1000), mean (SD)a	59.1(22.5)	65.0(25.0)	67.2(26.8)	70.8(29.1)	56.7(20.1)	61.5(22.6)	63.9(24.3)	63.2(24.8)
Pack year smoking, %
Never	0.59	0.46	0.38	0.28	74.4	65.4	53.7	42.7
<20	0.21	0.31	0.35	0.36	20.0	28.9	39.6	46.5
20+	0.20	0.23	0.27	0.36	5.6	5.6	6.7	10.9
Body mass index (kg/m2), mean (SD)	25.3(5.1)	24.5(4.4)	23.7(3.8)	23.3(3.5)	25.4(5.9)	24.5(5.2)	23.5(4.2)	23.7(4.3)
Physical activity (total METs), mean (SD)	12.1(17.9)	14.2(20.9)	15.7(21.2)	15.6(21.5)	18.5(25.1)	21.3(27.4)	24.1(29.8)	25.4(31.7)
Menarche before age 12, %	24.3	23.5	22.2	21.2	25.8	24.5	22.1	21.9
Oral contraceptive use, %	46.2	49.0	51.0	54.1	79.4	84.6	88.1	88.5
Parous, %	93.0	93.2	92.3	91.1	76.2	69.3	60.9	54.8
Breastfeeding ≥ 12 months %	19.0	17.4	16.7	16.9	27.7	23.8	20.5	17.6
Postmenopausal (PM), %	45.0	43.4	43.1	43.1	2.7	2.2	1.7	1.9
Current hormone use in PM %	17.2	17.8	18.3	19.5	82.1	84.4	80.6	82.1

^aMedian census tract family income was measured in 1986 for NHS and 1989 for NHSII. The values were based on the 2000 census data and adjusted by Consumer Price Index.

The multivariable adjusted HRs of developing RA are shown in Table 2 according to levels of alcohol intake. The pooled multivariable adjusted HR for alcohol use of 5.0–9.9 g/d compared to no use was 0.78 (95% CI, 0.61–1.00). We found a non-significant 25% risk decrease for alcohol intake of 5.0–9.9 g/d in NHS (HR=0.75, 95%CI: 0.54–1.03, p=0.077), but a significant inverse association for alcohol use of ≥10.0 g/d in NHSII (HR= 0.48, 95% CI: 0.29–0.82, p=0.007). We also conducted separate analyses for seropositive and seronegative RA cases (Table 2). Among 580 incident RA cases in NHS, 365 were seropositive and 215 were seronegative, and among 323 incident RA cases in NHSII, 213 were seropositive and 110 were seronegative RA cases. For seropositive RA cases, the pooled multivariable adjusted HR for alcohol use of 5.0–9.9 g/d compared to no use was 0.69 (95% CI, 0.50–0.95), while for seronegative RA cases, no significant associations were found.

Table 2.

Hazard ratios (HR) for incident RA, and incident seropositive or seronegative RA phenotypes, according to daily alcohol consumption in Nurses’ Health Study (NHS) and Nurses’ Health Study II (NHSII)

		NHS			NHSII
	Alcohol intake (grams/day)	No. of Cases	Person-Years	Multivariable HR (95% CI) a	No. of Cases	Person-Years	Multivariable HR (95% CI) a	Pooled HR (95% CI) ┼
All RA	None	128	464191	1.00(Referent)	106	514368	1.00(Referent)	1.00(Referent)
	0.1–4.9	257	786222	1.12(0.90,1.40)	163	895462	0.84(0.65,1.09)	0.98(0.74,1.30)
	5.0–9.9	58	251238	0.75(0.54,1.03)	37	217793	0.83(0.56,1.23)	0.78(0.61,1.00)
	≥10.0	137	398213	1.06(0.82,1.38)	17	155223	0.48(0.29,0.82)	0.74(0.34,1.60) b
	p trend			0.735			0.010	0.289
Seropositive RA	None	82	425591	1.00(Referent)	63	500447	1.00(Referent)	1.00(Referent)
	0.1–4.9	169	732847	1.08(0.82,1.42)	113	876803	0.95(0.69,1.31)	1.02(0.83,1.26)
	5.0–9.9	34	235105	0.61(0.40,0.92)	23	213713	0.83(0.51,1.37)	0.69(0.50,0.95)
	≥10.0	80	370691	0.87(0.63,1.21)	14	152505	0.64(0.35,1.16)	0.81(0.61,1.08)
	p trend			0.096			0.138	0.028
Seronegative RA	None	46	462769	1.00(Referent)	43	511712	1.00(Referent)	1.00(Referent)
	0.1–4.9	88	783445	1.20(0.82,1.76)	50	892333	0.68(0.44,1.03)	0.91(0.52,1.60) b
	5.0–9.9	24	250458	1.02(0.60,1.72)	14	217117	0.85(0.45,1.60)	0.95(0.63,1.42)
	≥10.0	57	396995	1.47(0.95,2.26)	3	154778	0.23(0.07,0.75)	0.63(0.10,3.87) b
	p trend			0.132			0.019	0.677

^aAdjusted for age (months), community median income (quartile), smoking (pack-years), body mass index (<25, 25–<30, ≥30 kg/m²), physical activity (0–3, 3–9, 9–18, 18–27, ≥27 METs/week), menarche age (<12, 12, >12 years), parity/breastfeeding (nulliparous, parous/no breastfeeding, parous/1–12 months breastfeeding, parous/>12 months breastfeeding), menopausal status, post-menopausal hormone use (pre-menopausal, post-menopausal with never use, current use and past use) and total calories intake.

^bp value for test of heterogeneity between the 2 cohorts < 0.05. Risk estimates were pooled by using a DerSimonian and Laird random-effects model

In sensitivity analysis using the simple updating alcohol measures rather than cumulative average alcohol consumption, the association was attenuated and no significant association was observed (the pooled HR for alcohol use of 5.0–9.9 grams/day was 0.98, 95% CI: 0.77–1.24). When we repeated the analysis including all the cancer cases, the results were highly consistent. Furthermore, we did not observe non-linear relationships of alcohol consumption with the risk of all RA, seropositive and seronegative RA. However, we may not have enough power to examine the effect of heavy alcohol intake on risk of RA due to a limited number of incident RA cases, especially in NHSII.

To examine the potential effect modification by smoking status, we included smoking by alcohol interactions in the adjusted models. No significant interactions were found between alcohol consumption and smoking status for all RA as well as for seropositive and seronegative RA cases in both cohorts (all P values >0.05).

Among individual alcoholic beverages, the significant association with the risk of RA was found in beer after controlling for the other two alcoholic beverages and covariates listed above (Table 3). Compared with women who never drank beer, the pooled multivariable HR for 2–4 drinks per week was 0.69 (95% CI 0.50–0.95). In contrast, wine or liquor consumption of 2–4 drinks a week was associated with a non-significant reduced risk of RA after controlling for the other two alcoholic beverages and potential confounding factors in both cohorts.

Table 3.

Hazard ratios (HR) for incident RA according to individual alcoholic beverage intake in Nurses’ Health Study (NHS) and Nurses’ Health Study II (NHSII)

	NHS			NHSII
No. of drinks/week	No. of Cases	Person-Years	Multivariable HR (95% CI) a	No. of Cases	Person-Years	Multivariable HR (95% CI) a	Pooled HR b (95% CI) ┼
Beer
None	405	1273199	1.00(Referent)	260	1279459	1.00(Referent)	1.00(Referent)
≤1	79	249701	1.01(0.78,1.30)	42	296816	0.81(0.58,1.14)	0.93(0.76,1.14)
2–4	30	116850	0.80(0.55,1.18)	13	148435	0.49(0.28,0.87)	0.69(0.50,0.95)
≥5	11	56864	0.56(0.31,1.02)	6	36796	0.87(0.38,1.99)	0.65(0.40,1.06)
Wine
None	205	679822	1.00(Referent)	190	941032	1.00(Referent)	1.00(Referent)
≤1	163	497384	1.08(0.87,1.35)	83	503271	0.96(0.73,1.29)	1.03(0.87,1.23)
2–4	89	334238	0.89(0.67,1.17)	39	245156	0.95(0.65,1.39)	0.91(0.73,1.13)
≥5	67	188056	1.17(0.86,1.58)	9	81147	0.64(0.32,1.27)	1.06 (0.80,1.40)
Liquor
None	297	953348	1.00(Referent)	256	1335607	1.00(Referent)	1.00(Referent)
≤1	115	388045	0.92(0.73,1.16)	46	301617	0.89(0.64,1.24)	0.91(0.76,1.10)
2–4	60	202862	0.88(0.65,1.18)	16	95236	0.95(0.56,1.60)	0.90(0.69,1.16)
≥5	53	153757	0.97(0.71,1.32)	4	26954	0.75(0.28,2.04)	0.95(0.70,1.27)

^aAdjusted for the other two alcoholic beverages and age (months), community median income (quartile), smoking (pack-years), body mass index (<25, 25–<30, ≥30 kg/m²), physical activity (0–3, 3–9, 9–18, 18–27, ≥27 METs/week), menarche age (<12, 12, >12 years), parity/breastfeeding (nulliparous, parous/no breastfeeding, parous/1–12 months breastfeeding, parous/>12 months breastfeeding), menopausal status, post-menopausal hormone use (pre-menopausal, post-menopausal with never use, current use and past use), and total calories intake.

^bAll p values for test of heterogeneity between the 2 cohorts >0.05. Risk estimates were pooled by using a DerSimonian and Laird random-effects model.

Discussion

In this large prospective cohort study, we observed a modest association between moderate alcohol intake and risk of developing RA. Women who drank 5.0–9.9 g/d alcohol (equivalent to 3–5 standard drinks per week) had a 22% reduced risk of developing RA, and a 31% reduced risk of developing seropositive RA, compared to women who did not drink any alcohol. Consistent with the primary analysis using total alcohol intake in grams per day, beer but not wine and liquor intake was associated with reduced risk of RA.

An indication that alcohol consumption may decrease the risk for RA and RA progression has come from several case-control studies.(9, 12, 23) Consistent with findings for stronger associations between environmental factors and seropositive RA, (5, 7) we found a significant association of moderate alcohol consumption and seropositive RA. Similarly, in a Danish study, individuals who consumed alcohol had an overall lower risk of developing ACPA-positive RA compared to those who did not.(9) A dose-dependent effect was demonstrated in a subsequent study of two studies, Epidemiologic Investigation of Rheumatoid Arthritis (EIRA) and the Danish Case–Control Study of Rheumatoid Arthritis (CACORA). Those with the highest consumption (≥5 drinks or 80 g alcohol per week) had a decreased risk of RA on the order of 40–50% compared with those with low-to-no consumption (<0.5 g alcohol per week). However, case-control studies ascertain alcohol exposure retrospectively raising the potential of recall bias. Early RA symptoms and some medication use (e.g. nonsteroidal anti-inflammatory drugs, NSAIDs or methotrexate) may change the behavior of alcohol use and induce reverse causation in case-control studies. One published prospective study found no association between alcohol consumption and risk of RA, (24) while a recent cohort study found that moderate consumption of alcohol was associated with reduced risk of RA, but was also limited by a small number of incident cases (197 cases) and a short period of follow-up. (13) Moreover, they did not stratify by serologic status in the analysis and used only two alcohol assessments which may lead to misclassification. Using cumulative average intake to represent long-term alcohol exposure over a time period as long as 26 years, our prospective analyses confirmed an inverse association of moderate alcohol consumption with RA risk. We used lagged analysis to eliminate the potential for recall bias and reverse causation bias. To our knowledge, this is the largest prospective study to date including over 900 incident RA cases evaluated RA risk in relation to alcohol consumption using repeated alcohol measures prior to RA diagnosis.

A mechanistic basis for a role of alcohol in the etiology of RA has not been well developed. Alcohol is thought to have effects on both the hormonal and immunologic systems, although the current knowledge is incomplete and often conflicting. For example, in a study of healthy premenopausal women, the alcohol use as low as approximately one drink per day caused an increase in serum estradiol concentrations of 27 to 38%. (25) Grimaldi and colleagues have demonstrated that estrogens play an important role in B cell maturation, selection, and activation, and potentially the breakdown of immune tolerance.(26) From an immunologic perspective, several reports have suggested that alcohol intake influences systemic inflammation and inflammatory arthritis (27, 28). Alcohol has been shown to diminish the response to immunogens in animals as well as in humans, and to suppress significantly the synthesis of proinflammatory cytokines and chemokines, such as tumor necrosis factor (TNF)α, interleukin (IL)-6, and IL-8 both in vivo and in vitro in alveolar macrophages and human blood monocytes.(29) Notably, addition of alcohol to the drinking water of mice was recently shown to reduce clinical signs of arthritis as well as joint destruction. The beneficial effects may be mediated by up-regulation of testosterone production, which in turn inhibits the Nuclear Expression of Transcription (NF-kB) activation, leading to decreased cytokine/chemokine production and decreased chemotactic activity of leukocytes.(27) Using NHSII, a previous study including 473 non-RA women reported that women who consumed 1–2 drinks/drinking day had significantly lower sTNFR1 and sTNFR2 levels as well as lower CRP and IL-6 levels compared to non-drinkers, (30) We recently found that the moderate alcohol consumption was associated with reduction in plasma biomarkers of inflammation, including sTNFR2 and IL-6, among women with pre-RA where blood was collected up to 12 years prior to first symptoms of RA suggesting an effect of alcohol during the asymptomatic phases.(31) Therefore, moderate alcohol drinking may reduce inflammation in both non-RA population and preclinical RA cases. This suggests that alcohol consumption may prevent or delay the process of RA development.

We observed that moderate consumption of beer, but not wine and liquor was associated with a decreased risk of RA. Similar trends without statistical significance were also found in wine and liquor intake. It may be due to reduced statistical power by stratified analyses. In sensitivity analysis using the updating alcohol consumption other than cumulative average measures, the association was attenuated. This suggests that the cumulative effect of long-term moderate alcohol exposure, not short-term alcohol consumption may reduce the risk of developing RA.

The strengths of this study include the large number of incident RA cases, length of follow-up, and detailed prospective and updated assessments of alcohol consumption across different age periods affording the most comprehensive evaluation of the effect of alcohol consumption throughout a woman’s adult life. We studied alcohol consumption and risk of RA in a lagged analysis by excluding the most recent alcohol measure. It is likely that immune changes and early symptoms occur prior to RA diagnosis. Lagged analyses address the timing of exposure of alcohol prior to the development of RA to reduce the potential for reverse causation. Our study has some limitations however. Firstly, given our strict definition of RA, we may have excluded possible RA cases, introducing selection bias. Secondly, as an observational study, women were not randomly assigned to alcohol use. However, it is unlikely that such a long-term randomized trial will ever be performed. Thirdly, misclassification of exposures in this study may result from self-reported FFQ. In our study, subjects were asked to relate alcohol intake to the previous year in general, which can reflect relatively long-term consumption and decrease the effect of short-term variability. Moreover, we have obtained repeated measurements of alcohol consumption during follow-up and utilized cumulative average estimates to reduce the within-subject variation. Fourthly, participants in NHS and NHSII were from different generations of women with heterogeneous characteristics, although consistent results were found from two cohorts. Finally, we may have limited power to study the effect of heavy drinking (such as >15, >20 grams per day), and possible non-linear relationships.

In summary, our study provides a comprehensive assessment of the relationship between alcohol intake and RA risk in terms of timing, quantity, and types of alcohol in two large prospective cohorts with detailed information on RA risk factors. We found a modest association between long-term moderate alcohol drinking over multiple decades and reduced risk of RA. Our findings have implications for RA prevention and could have large potential public health implications.