Association between work physical activity, dietary factors, and risk of rheumatoid arthritis

Abstract

Objective:

We aimed to determine the association of physical activity and dietary factors on RA risk.

Methods:

This case-control study within the Mayo Clinic Biobank matched incident RA cases (two codes plus disease-modifying anti-rheumatic drug, PPV 95%) to controls 1:3 on age, sex, and recruitment year/location. A baseline questionnaire assessed activity and dietary exposures. Logistic regression models calculated adjusted odds ratios (aOR) with 95% confidence intervals (CI) of RA for each of 45 activity/dietary exposures.

Results:

We identified 212 incident RA cases and 636 controls (mean age 64, 70% female). Active work physical activity was associated with elevated risk of RA (aOR 3.00, 95% CI 1.58-5.69 vs. sedentary); leisure activity was not (aOR 0.96, 95% CI 0.64-1.42 sedentary vs. active). Three or more servings high-fat food and 5+ servings fruits/vegetables daily showed non-significant associations with RA (aOR 1.22, 95% CI 0.74-2.00 vs. 0-1 time; aOR 0.75, 95% CI 0.51-1.11 vs. 0-3 times), especially in sensitivity analyses with at least five years between questionnaire and RA (aOR 1.80, 95% CI 0.69-4.71; aOR 0.54, 95% CI 0.27-1.08). Alcohol binging was not associated with RA risk (aOR 1.28, 95% CI 0.56-2.96). Finally, sensitivity (versus primary) analyses showed a nonsignificant increase in RA risk for most vitamins and supplements.

Conclusion:

Active work physical activity and some nutritional profiles (increased high-fat, reduced fruit/vegetable consumption) may be associated with increased risk of RA. Confirmatory studies are needed.

Introduction

Nearly 1% of the world population has rheumatoid arthritis (RA) (1), a debilitating joint disease associated with premature mortality (2). Early and sustained medication use is associated with the best outcomes (3). However, patients and their family members frequently ask what lifestyle modifications they can make to reduce disease activity and risk as well. Smoking (4-6) and obesity (6-8) represent the main known modifiable risk factors for RA risk and disease activity. However, modifying these risk factors is difficult, unappealing, or not applicable for many patients.

Physical activity may represent an additional modifiable risk factor for RA that could apply to a broader subset of patients. Encouragingly, increased leisure physical activity has been associated with decreased risk of RA in two studies (9, 10). However, work physical activity has only been studied in one cross-sectional study (11). Furthermore, the interaction between leisure and work physical activity on RA risk remains unstudied.

While data from most previous studies agree that a “healthier diet” overall may be protective for RA (12-15), several specific dietary factors need to be studied in greater detail (16). For example, many studies show that modest alcohol consumption is associated with reduced risk of RA (6, 17), whereas some recent studies have shown no association (18-20), or a harmful association (21). Dose-response analysis (i.e., including binge alcohol consumption) could be helpful to reconcile these differences. In addition, despite the association between obesity and RA (6-8), high-fat diet has not yet been studied as a risk factor for RA. Many vitamins and supplements such as B vitamins also remain unstudied despite evidence that certain supplements like vitamin D (22) or omega-3 fatty acids (23) may be associated with reduced RA risk.

Finally, the interactions between dietary factors, physical activity, and body mass index (BMI) remain relatively unstudied as well. Studying their interactions is important, as these represent the three main modifiable lifestyle risk factors for patients. For example, one study suggested that the association between healthy diet and RA may occur through mediation by BMI (13).

To address these three gaps, we aimed to determine (1) the effect of work and leisure physical activity on incident RA, (2) the effect of specific dietary factors including alcohol binging, high-fat diet, and certain vitamins/supplements on risk of incident RA, and (3) presence of pairwise interactions between dietary factors, physical activity, and BMI on RA risk. We hypothesized that decreased physical activity through both leisure and work and especially in combination (i.e., synergistic interaction) would be associated with increased risk of RA, as would alcohol binging and high-fat diet but not vitamins/supplements aside from Vitamin D or omega-3 fatty acids given their general lack of association. In addition, we hypothesized that reduced physical activity is especially associated with RA in obese individuals.

Materials and Methods

Study design and population

This study took place within the Mayo Clinic Biobank, which includes over 56,000 participants and their electronic health record data (24). Mayo Clinic Biobank recruited its participants mainly from Mayo Clinic primary care locations in Minnesota between 2009 and 2015. Approximately 29% of those invited chose to participate and completed a baseline questionnaire. Of those, 77% completed the follow-up questionnaire sent approximately four years later. This case-control study matched each RA case to three controls based on age (within 5 years), sex, recruitment year, recruitment location (Minnesota, Wisconsin, or Florida), and distance from recruitment location (within 500 miles). We defined index date as the date of RA diagnosis as estimated by the date of second RA diagnosis code, or matched date for controls. This study received Institutional Review Board approval (17-010806) and complies with the Declaration of Helsinki.

Rheumatoid arthritis

We identified RA cases using a rules-based algorithm combining two diagnosis codes (ICD-9 714.0 or 714.9 or ICD-10 M05x or M06x) at least 30 days apart with use of a disease-modifying anti-rheumatic drug (DMARD). This definition had a positive predictive value of 95% in a manual verification substudy of 100 RA cases (25). For this study, we included only the subset of RA cases with no RA self-report on the baseline questionnaire at biobank enrollment and who met the criteria for RA based on two diagnosis codes and a DMARD at least 90 days after the most recent questionnaire (i.e., incident RA). Controls included patients without self-reported RA and no diagnosis codes for RA.

Activity exposures

We identified all study exposures primarily using self-report on the baseline questionnaire. However, participants missing an item on the baseline questionnaire had their response from the follow-up questionnaire “filled in” for that item, provided the follow-up questionnaire still occurred before index date of RA or matched date.

Both the baseline and follow-up questionnaires assessed leisure physical activity using the Godin Leisure-Time Exercise Questionnaire (<14=sedentary, 14-23=moderate, 24+=active [ref]) (26, 27). The baseline questionnaire assessed work physical activity by asking participants, “For the job (includes homemaking) you held the longest, approximately how much of the time were you engaged in each of the following physical activities: sitting, standing, walking, light manual labor, heavy manual labor,” with answer choices including none, a little, some, most, or all of the time. We assigned weight to each of these responses and totaled the score per Table 1, with prespecified groupings including 0-1 (which we designated as “sedentary”) [ref], 2-5 (which we designated as “moderate”), and 6+ (which we designated as “active”). Note we changed the reference group from the active to the sedentary group for work activity during analysis so that the reference group would align with the lowest RA risk.

Table 1.

Work physical activity score calculation

Job activity	Little	Some	Most	All
Sitting	0	0	0	0
Standing	0	1	2	3
Walking	1	3	5	6
Light labor	2	4	7	8
Heavy labor	3	5	8	10

Dietary exposures

The questionnaires also asked the following dietary questions, with items present only on the follow-up questionnaire marked by (*): times a day eating high-fat food such as red meat, fried food, whole milk, regular cheese, ice cream, baked goods, or regular salad dressing (0-1[ref],2,3+), *times a day eating red meat over the last 2 years (don’t eat-1[ref],2+), *times a day eating fish over the last 2 years (don’t eat-1[ref], 2+), *times a day eating poultry over the last 2 years (Don’t eat-1[ref],2+), servings a day of fruit (0-1[ref],2,3+), servings a day of vegetables (0-1[ref],2,3+), servings a day of milk, dairy products, or calcium supplements (0-1[ref],2-3,4+), servings of diet soft drinks per day (0[ref],1-2,3+), servings of regular soft drinks per day (0[ref],1+), cups of caffeinated coffee (total cups*percent caffeinated) (0[ref], <1 per month to 1 cups per day, 2+ cups per day), cups of decaffeinated coffee (total cups*percent decaf) (0[ref], <1 per month to 6 cups per week, 1+ cups per day), *cups of caffeinated tea (0[ref], <1 per month to 6 cups per week, 1+ cups per day), *cups of decaffeinated tea (0[ref], <1 per month to 6 cups per week, 1+ cups per day), drinks of alcohol in the last 12 months (never to <1 per month[ref], twice a month to 5 times per week, 6+ times per week), drinks per session of alcohol (0-2[ref], 3-4, and within alcohol consumers, frequency of alcohol binge (6+ drinks on one occasion) (never[ref], ever).

The baseline questionnaire also recorded use of one of 27 vitamins or supplements regularly (defined as 2 times a week for at least 3 months). Vitamins included multivitamins, prenatal vitamin, vitamin A, B vitamins, vitamin C, vitamin D, vitamin E, beta carotene, calcium, folate, iron, selenium, zinc. Supplements included 5-hydroxytryprophan (HTP), acidophilus, bee pollen or royal jelly, chondroitin, coenzyme Q10 (CoQ10), dehydroepiandrosterone (DHEA), fiber supplement, fish oil/omega fatty acids, glucosamine, melatonin, progesterone cream, S-adenosyl-L-methionine (SAM-e), xanadrine, and other.

BMI and other covariates

We obtained BMI (<25 [ref], 25-<30, 30+ kg/m²) and age (continuous) from the electronic health record at the index of RA diagnosis or matched date. We selected the following additional covariates from the baseline biobank questionnaire based on their established association with RA: sex (male vs. female), race/ethnicity (non-Hispanic white vs. other), education (bachelor’s degree or higher vs. less than bachelor’s), smoking status (never, past, current), asthma (yes vs. no/blank), and parity (continuous).

Statistical analysis

We presented continuous variables as mean (± standard deviation) and discrete variables as frequency (percentage) based on the number with data available. We used Wilcoxon’s rank sum test to test continuous and ordinal variables and Pearson’s chi-squared for nominal variables. For Aims 1 and 2, we used logistic regression models with each physical activity or dietary item as the exposure variable to calculate odds ratios (OR) for RA. Models adjusted for covariates age, sex, race/ethnicity, education, smoking status, asthma, and parity. For men, parity was mapped as zero. Missing values for asthma were recorded as “no.” Women missing information for number of live births and number of children breastfed had the class imputed as the lowest value, and an indicator for those with missing values was included. Participants missing exposure data were excluded for that analysis. Firth’s bias-reduced penalized likelihood was used for models with convergence issues to get finite parameter estimates. For Aim 3, we assessed interactions on the additive scale using the attributable proportion due to interaction (AP) and relative excess risk due to interaction (RERI) when compared to the healthiest reference group (i.e., normal BMI, high physical activity, etc.) (28). We also tested for multiplicative interactions because they can also be biologically plausible depending on the scenario (29).

For sensitivity analyses, we performed all the above analyses restricted to the subset of participants whose exposure data was obtained five or more years before index date of RA or matched date, along with their corresponding controls. We considered such results potentially more biologically relevant than the primary analyses given evidence that RA likely originates many years before clinical onset (30). However, the timing of when dietary exposures are most impactful for RA is uncertain. Furthermore, we were not sure at the time of study protocol completion whether this subset would be sufficiently large for a standalone analysis, and thus included it as a sensitivity analysis. Throughout the study, we used a significance threshold of two-sided alpha=0.05 and calculated 95% confidence intervals (CI). All analyses were pre-specified in a study protocol unless otherwise noted. Analyses were conducted using R version 4.1.2 (R Foundation for Statistical Computing, Vienna, Austria) and SAS version 9.4 (SAS Institute, Cary, NC).

Results

Participant characteristics

We identified 212 incident RA cases and 636 matched controls (mean age 64, 70% female). RA cases tended to have less education and more obesity than controls (Table 2). Most (97%) participants were recruited from Minnesota, with 2% from Wisconsin, and <1% from Florida. Both cases and controls completed their baseline questionnaires a mean of 52 months (standard deviation [SD] 31) before index date of RA diagnosis or matched date (p=0.96). Before index date of RA diagnosis (or matched date), 67 (32%) cases and 202 (32%) controls completed the follow-up questionnaire (p=0.97), with completion on average 30 months (SD 22) and 29 months (SD 22) prior to index date, respectively (p=0.86).

Table 2.

Demographic characteristics of 212 incident RA cases from Mayo Clinic Biobank and matched controls

Characteristic	Incident RA (N = 212)	Controls (N = 636)	p-value
Age at index, years (mean ±SD)	64 (±14)	64 (±14)	n/a*
Female sex, n (%)	148 (70)	444 (70)	n/a*
Recruitment year (mean ±SD)	2011.2 (±1.7)	2011.2 (±1.7)	n/a*
Race White, non-Hispanic, n (%)	206 (97)	606 (95)	0.24
Education bachelor's degree or higher, n (%)	85 (41)	325 (52)	0.006
BMI at index, kg/m2			0.020
<25	51 (24)	190 (30)
25-<30	67 (32)	220 (35)
30+	94 (44)	225 (35)
Smoking status at questionnaire, n (%)			0.56
Never	114 (53)	364 (57)
Past	87 (41)	235 (37)
Current	11 (5.2)	37 (5.8)
Asthma, n (%)	34 (16)	78 (12)	0.15
Parity (mean ±SD)	2.4 (±1.4)	2.6 (±1.2)	0.061
Number children breastfed, n (%)			0.93
Missing	88	195
None	42 (34)	122 (34)
1-2	55 (44)	150 (42)
3-5	26 (21)	80 (22)
6+	1 (0.8)	5 (1.4)

BMI = body mass index, kg = kilograms, m = meters, n/a = not applicable, RA = rheumatoid arthritis, SD = standard deviation

^*Matched factor

Work and leisure physical activity

Most participants reported active leisure and work physical activity (Table 3). After adjusting for covariates, leisure physical activity was not associated with RA risk. However, an active or moderate level of work physical activity was associated with an increased risk of RA compared to sedentary work activity. In particular, active work physical activity was associated with a 3-fold increased risk of RA (Table 3). In sensitivity analyses using the subset of 72 (34%) RA cases with at least five years between baseline questionnaire and index date of RA diagnosis and their controls, the association between active or moderate work physical activity and RA remained significant (Table 3).

Table 3.

Association between leisure and work physical activity and incident RA

Physical Activity	Incident RA (N = 212)	Controls (N = 636)	Adjusted* OR (95% CI)	Sensitivity Analysis ** OR (95% CI)
Godin Leisure-Time score
24+ (active)	117 (56)	383 (61)	[ref]	[ref]
14-23 (moderate)	42 (20)	94 (15)	1.43 (0.94,2.19)	1.55 (0.77,3.12)
<14 (sedentary)	50 (23)	147 (24)	0.96 (0.64,1.42)	0.60 (0.26,1.37)
Work physical activity score
6+ (active)	151 (75)	405 (66)	3.00 (1.58,5.69)	6.76 (1.44,31.6)
2-5 (moderate)	38 (19)	111 (18)	2.83 (1.39,5.77)	5.23 (1.19,22.9)
0-1 (sedentary)	12 (6)	99 (16)	[ref]	[ref]

CI = confidence interval, OR = odds ratio, RA = rheumatoid arthritis

^*Adjusting for age, sex, race/ethnicity, education, smoking status, asthma, parity. bold values indicate statistical significance (p<0.05).

^**Including only the 72 (34%) cases and their 216 controls with at least 5 years before survey and RA diagnosis

Due to the strong observed association between work activity and RA, we performed post hoc adjustment for available occupations with both high work physical activity and inhalant exposure (active-duty military, construction, farming, manufacturing, mining, waste management versus all others). Such occupations were reported in 11 (5.2%) RA cases and 32 (5.0%) controls. Adjusting for them did not change results (OR 2.84 95% CI 1.39-5.79 for moderate work activity and OR 3.04 95% CI 1.60-5.78 for active work activity). Next, because of the small size of the sedentary work physical activity group, we performed additional post hoc analyses recategorizing the groups using score cutoffs of 0-5 (50 cases, 120 controls), 6-12 (79 cases, 228 controls), and 13+ (72 cases, 177 controls). Compared to the lowest work activity group, the odds of RA in the middle work activity group was 1.42 (95% CI 0.94-2.15) and in the highest work activity group was 1.67 (95% CI 1.08-2.57). In sensitivity analyses, the odds of RA were 1.34 (95% CI 0.69-2.59) for the middle work activity group and 1.04 (95% CI 0.49-2.19) in the highest work activity group.

Dietary factors

After adjusting for confounders in the full study population, dietary factors including alcohol use and binging were not statistically associated with RA risk (Table 4). However, three dietary factors (high-fat food, fruit, and vegetables) showed a non-significant dose-response association in the sensitivity analysis using the subset of RA cases with at least five years between baseline questionnaire and index date of RA diagnosis and their corresponding controls (Table 4). Because both fruits and vegetables showed a similar inverse relationship with RA, we performed a post hoc analysis combining their intake together. In the full cohort this was not associated with RA (OR 1.01, 95% CI 0.67-1.53 and OR 0.75, 95% CI 0.51-1.11 for 3-4 and 5+ servings compared to 0-3). In the sensitivity cohort, however, consuming 3-4 or 5+ fruits and vegetables was associated with lower odds of RA compared to only 0-3 per day (OR 0.45, 95% CI 0.21-0.97 and OR 0.54, 95% CI 0.27-1.08 respectively). Moderate milk/dairy/calcium intake also showed a statistically significant protective association with RA. However, the association was not dose-dependent (Table 4).

Table 4.

Association between preceding dietary factors and incident RA

	Number (%)
Dietary Factor	Incident RA (N = 212)	Controls (N = 636)	Adjusted** OR (95% CI)	Sensitivity Analysis ***OR (95% CI)
High-fat food, per day
0-1 time	110 (52)	333 (52)	[ref]	[ref]
2 times	72 (34)	232 (37)	0.95 (0.67,1.34)	1.45 (0.81,2.59)
3+ times	28 (13)	70 (11)	1.22 (0.74,2.00)	1.80 (0.69,4.71)
Red meat, per day*
0-1 time	57 (85)	160 (80)	[ref]	[ref]
2+ times	10 (15)	41 (20)	0.61 (0.28,1.34)	0.63 (0.26,1.51)
Fish, 2+ times per day*	4 (6)	22 (11)	0.52 (0.17,1.61)	0.53 (0.14,1.97)
Poultry, per day*
0-1 time	51 (77)	143 (71)	[ref]	[ref]
2+ times	15 (22)	58 (29)	0.68 (0.35,1.36)	0.68 (0.32,1.47)
Fruit, per day
0-1 time	72 (34)	187 (29)	[ref]	[ref]
2 times	80 (38)	243 (38)	0.89 (0.61,1.30)	0.52 (0.25,1.06)
3+ times	59 (28)	206 (32)	0.79 (0.52,1.20)	0.60 (0.28,1.27)
Vegetables, per day
0-1 time	52 (25)	157 (25)	[ref]	[ref]
2 times	98 (46)	252 (40)	1.16 (0.77,1.73)	0.65 (0.31,1.36)
3+ times	61 (29)	227 (36)	0.81 (0.52,1.27)	0.54 (0.25,1.15)
Milk/dairy/calcium, per day
0-1 time	74 (35)	187 (29)	[ref]	[ref]
2-3 times	105 (50)	392 (62)	0.65 (0.45,0.93)	0.78 (0.41,1.49)
4+ times	32 (15)	57 (9)	1.39 (0.82,2.36)	1.46 (0.61,3.47)
Diet soft drinks, per day
None	131 (62)	398 (63)	[ref]	[ref]
1-2 times	64 (31)	189 (30)	1.05 (0.74,1.49)	2.00 (1.09,3.65)
3+ times	15 (7)	49 (8)	0.93 (0.49,1.75)	0.49 (0.10,2.47)
Regular soft drinks, 1+ per day	18 (9)	75 (12)	0.67 (0.38,1.16)	0.58 (0.22,1.53)
Caffeinated coffee
None	49 (23)	149 (24)	[ref]	[ref]
<2 cups per day	45 (21)	191 (30)	0.72 (0.45,1.16)	0.55 (0.22,1.35)
2+ cups per day	118 (56)	295 (47)	1.24 (0.81,1.88)	1.06 (0.50,2.25)
Decaffeinated coffee
None	163 (78)	464 (75)	[ref]	[ref]
<1 cup per day	20 (10)	83 (13)	0.71 (0.42,1.21)	0.42 (0.15,1.17)
1+ cups per day	27 (13)	75 (12)	1.03 (0.63,1.70)	0.70 (0.31,1.60)
Caffeinated tea*
None	25 (46)	52 (35)	[ref]	[ref]
<1 cup per day	20 (36)	81 (55)	0.55 (0.26,1.14)	0.47 (0.21,1.06)
1+ cups per day	10 (18)	15 (10)	1.33 (0.49,3.61)	0.87 (0.24,3.13)
Decaffeinated tea*
None	33 (62)	85 (62)	[ref]	[ref]
<1 cup per day	18 (34)	40 (29)	1.33 (0.63,2.82)	1.05 (0.46,2.42)
1+ cups per day	2 (3.8)	12 (9)	0.47 (0.10,2.29)	0.62 (0.12,3.36)
Alcoholic beverages
<1 per month	92 (43)	254 (40)	[ref]	[ref]
2/month to 5/week	100 (47)	322 (51)	0.93 (0.66,1.31)	0.75 (0.41,1.38)
6+ times per week	20 (9)	60 (9)	0.99 (0.55,1.78)	1.00 (0.37,2.70)
Alcohol beverages per session 3+
Rarely drinks	92 (44)	254 (40)	[ref]	[ref]
0-2 per day	95 (46)	315 (50)	0.90 (0.64,1.28)	0.78 (0.42,1.43)
3+ per day	22 (11)	59 (9)	1.14 (0.64,2.04)	0.91 (0.28,2.94)
Alcohol binge frequency
never	169 (83)	533 (86)	[ref]	[ref]
ever	35 (17)	90 (14)	1.17 (0.73,1.85)	0.98 (0.41,2.34)

CI = confidence interval, N/A = not applicable due to low sample size, OR = odds ratio, RA = rheumatoid arthritis

^*from follow-up survey only, including 67 RA cases and 202 controls

^**adjusting for age, sex, race/ethnicity, education, smoking status, asthma, parity. Bold values indicate statistical significance (p<0.05).

^***including only the 72 (34%) cases and their 216 controls with at least 5 years before survey and RA diagnosis

Most participants reported consuming at least one vitamin (92% RA cases and 96% of controls) and at least one supplement (62% of RA cases and 59% of controls) (Table 5). After adjustment for covariates, none of the vitamins or supplements were associated with RA risk except beta carotene and folate (Table 5), nor were number of vitamins (OR 0.56 95% CI 0.26-1.21 for 3+ compared to 0) or number of supplements (OR 1.23, 95% CI 0.68-2.24 for 3+ compared to 0). However, in sensitivity analyses of the subset of RA cases with at least five years between baseline questionnaire and index date of RA diagnosis, the point estimates for every vitamin and supplement increased, with vitamin A, folate, zinc, and acidophilus showing associations with increased RA risk (Table 5).

Table 5.

Association between preceding vitamins/supplements and incident RA

	Number (%)
Item	Incident RA (N = 212)	Controls (N = 636)	Adjusted* OR (95% CI)	Sensitivity Analysis **OR (95% CI)
Any vitamin/mineral	159 (92)	482 (96)	0.53 (0.26,1.10)	1.05 (0.19,5.68)
Multivitamin	122 (71)	382 (76)	0.83 (0.56,1.23)	1.84 (0.78,4.33)
Prenatal vitamin	6 (3.5)	16 (3.2)	0.75 (0.26, 2.12)	1.91 (0.31, 11.9)
Vitamin A	5 (2.9)	5 (1.0)	2.75 (0.76, 9.95)	17.3 (1.52, 2382)
B Vitamins	31 (18)	78 (16)	1.03 (0.64,1.66)	1.71 (0.74,3.93)
Vitamin C	38 (22)	93 (19)	1.23 (0.80,1.89)	1.51 (0.72,3.19)
Vitamin D	70 (41)	182 (36)	1.20 (0.83,1.74)	1.76 (0.89,3.48)
Vitamin E	20 (12)	37 (7)	1.55 (0.86, 2.80)	2.17 (0.84, 5.60)
Beta carotene	4 (2.3)	2 (0.4)	6.11 (1.05, 35.8)	N/A
Calcium	87 (50)	251 (50)	1.09 (0.74,1.61)	1.52 (0.78,2.95)
Folate	12 (7)	13 (2.6)	2.82 (1.24, 6.44)	5.20 (1.35, 20.0)
Iron	20 (12)	43 (9)	1.31 (0.73, 2.33)	1.69 (0.65, 4.41)
Selenium	5 (2.9)	10 (2.0)	1.57 (0.52, 4.79)	4.85 (0.86, 27.5)
Zinc	12 (6.9)	21 (4.2)	1.59 (0.75, 3.34)	8.13 (1.82, 36.3)
Any supplement	107 (62)	294 (59)	1.14 (0.79,1.65)	1.54 (0.80,2.97)
5-hydroxytryprophan	1 (0.6)	3 (0.6)	N/A	N/A
Acidophilus	11 (6.4)	17 (3.4)	1.71 (0.76, 4.86)	4.31 (1.04, 17.8)
Bee pollen or royal jelly	0 (0)	1 (0.2)	N/A	N/A
Chondroitin	15 (8.7)	54 (10.8)	0.80 (0.43,.48)	0.93 (0.35,2.52)
Coenzyme Q10	13 (7.5)	31 (6.2)	1.22 (0.62, 2.43)	1.51 (0.49, 4.68)
Dehydroepiandrosterone	1 (0.6)	7 (1.4)	0.45 (0.05, 3.72)	2.17 (0.18, 26.3)
Fiber supplement	27 (16)	61 (12)	1.22 (0.73,2.02)	1.95 (0.76,5.04)
Fish oil/omega fatty acids	75 (43)	209 (42)	1.08 (0.76,1.56)	1.53 (0.80,2.92)
Glucosamine	35 (20)	85 (17)	1.30 (0.83,2.05)	1.44 (0.66,3.14)
Melatonin	9 (5)	29 (6)	0.83 (0.38, 1.83)	0.14 (0.00, 1.22)
Progesterone cream	1 (0.6)	1 (0.2)	N/A	N/A
SAM-e	1 (0.6)	2 (0.4)	N/A	N/A
Xanadrine	0 (0)	0 (0)	N/A	N/A
Other	3 (1.7)	14 (2.8)	0.61 (0.17, 2.20)	1.10 (0.27, 4.49)

CI = confidence interval, N/A = not applicable due to sample size, OR = odds ratio, RA = rheumatoid arthritis

^*adjusting for age, sex, race/ethnicity, education, smoking status, asthma, parity. bold values indicate statistical significance (p<0.05).

^**including only the 72 (34%) cases and their 216 controls with at least 5 years before survey and RA diagnosis

Interaction between activity, dietary factors, and BMI

BMI was associated with a nonsignificant increase in risk of RA (OR 1.09, 95% CI 0.71-1.67 for BMI 25-<30 kg/m² and OR 1.40, 95% CI 0.93-2.11 for BMI 30+ kg/m²). There were no significant interactions between activity, dietary, factors, and BMI. The nonsignificant interaction between work activity and BMI (AP 0.21, 95% CI −0.10-0.50 and multiplicative OR 1.24 95% CI 0.90-1.69) became stronger but was still not significant in the sensitivity analysis where only participants with at least five years between the baseline questionnaire and index date of RA diagnosis were included (AP 1.36, 95% CI −0.27-2.99 and multiplicative OR 1.62, 95% CI 0.93-2.82). Leisure activity and work activity also showed no interaction (AP 0.07, 95% CI −0.13-0.26 and multiplicative OR 1.14, 95% CI 0.83-1.44).

Finally, 29 (13%) of RA cases and 96 (15%) of controls were missing any demographic variable (before imputation), whereas 58 (27%) of RA cases and 177 (28%) of controls were missing any of the 41 baseline exposure items. Characteristics of participants missing data at baseline did not differ substantially from those not missing data except that they tended to be younger and (among women) have slightly fewer children (see Supplementary Table S1).

Conclusions

This study of incident RA with extensive preceding activity and dietary assessment found that active work physical activity was associated with an increased risk of RA. Another novel finding was that high-fat diet was associated with slightly increased risk of RA, whereas increasing fruit and vegetable intake was associated with slightly lower RA risk. Alcohol and most vitamin and supplement use had minimal effect on RA risk, though there was a trend towards certain vitamins or supplements increasing RA risk especially with more distant use. These findings highlight additional potential lifestyle interventions for reducing RA risk.

The first key finding from this study was that active work physical was associated with increased risk of RA. Work physical activity has only been studied in one other study that found no association (11), though its cross-sectional design after RA diagnosis is concerning for recall bias. This direction of association was the opposite of our original hypothesis. Therefore it may be spurious due to residual confounding by socioeconomic status or other unknown factors. However, this study did adjust for education, which is highly correlated to socioeconomic status for predicting RA risk (31). The observed association is unlikely to be from reverse causation since early RA symptoms would make people less likely to have active jobs. However, the 3- to 6-fold association we observed was likely artificially magnified by the small numbers in the reference group. When redistributing groups, the dose-dependent association was still present but smaller in the primary analysis and was nonsignificant in the sensitivity analysis. Alternatively, it may be a true association, as bone and joint trauma is known to increase the risk of psoriatic arthritis (32). Therefore, further studies investigating this association between active work physical activity and RA are needed, especially related to type of occupation and with more complete adjustment for socioeconomic factors.

Regarding dietary factors, this study found that a high-fat diet may be associated with increased risk of RA, whereas fruit and vegetable intake may be associated with reduced risk. The trend towards high-fat food increasing risk of RA especially the sensitivity analysis of more distant dietary exposures is novel to this study. One prior study showed a nonsignificant association between increased “fat intake” and RA, supporting the validity of our finding (20). The current study also showed that increasing fruits and vegetable consumption by only two servings per day was associated with half the risk of RA. Prior studies found no statistically significant association of fruits and vegetables, though all showed a protective trend despite not adjusting for socioeconomic status (16, 18, 33, 34). Further studies examining fruit and vegetable consumption and RA with more complete adjustment for socioeconomic status, as discussed above, might be informative. The lack of association between alcohol binging and RA is novel to this study and confirms more recent literature showing no association (19, 20), or even a harmful association (21) between alcohol and RA. The observed association with modest intake of milk/dairy/calcium with RA risk seems likely to be spurious after considering the small sample size, lack of dose response association, and discordance with previous literature on this topic (35, 36). Overall, the results with high-fat, fruit, and vegetable consumption support growing literature that a “healthy” diet is protective for RA (12-15) while giving clinicians and patients more granularity as to specific steps they can take to achieve that.

We observed minimal associations between vitamin or supplement use and RA risk. This included no significant association with B vitamins, multivitamins, and fiber supplements, which were all new to this study. The lack of association with vitamin C (37) and calcium (36) also corroborates prior studies. The lack of association of omega 3 fatty acids and vitamin D conflicts with a recent randomized controlled trial that showed protective associations of each in older adults, though it was nonsignificant for omega 3 fatty acids (22). Interestingly, however, this study’s sensitivity analyses suggested that more distant vitamin and supplement use was associated with a nonsignificant yet noticeable increase in RA risk across all vitamins and supplements studied. In particular, the statistically significant associations with vitamin A, beta carotene, folate, zinc, and acidophilus were all novel to this study, though the number of exposed individuals was very small. This observation raises the question of whether duration and/or timing of supplement use may explain some historically discordant results such as that of vitamin D use, where studies of more recent exposure show a protective effect (22, 38), whereas those of more distant exposure show no effect (39, 40). Given the large societal magnitude of vitamin and supplement use, future studies should replicate these findings and explore the relationship between vitamin and supplement duration and timing on RA risk.

Strengths of this study included its use of incident RA, adjustment for many covariates, and ability to study 45 different activity and dietary factors. There are several important limitations to consider. First, small sample size likely limited our power to detect potentially novel associations such as interactions between activity, diet, and BMI as well as presumed RA protective factors such as leisure activity (41), tea (42), or fish (33) and presumed RA risk factors such as coffee (43) and soda (44). Second, the biobank’s use of a convenience sample for selection creates potential selection bias and limits generalizability, though this was somewhat mitigated by its recruitment in primary care divisions and at multiple locations. Third, misclassification of exposures is possible given reliance on self-report at one time point. Also, questionnaires did not ask participants about vegetarian status, which has been shown to be an important threshold for reducing other diseases including cardiovascular disease (45) and cancer (46). Fourth, unmeasured confounders are possible such as socioeconomic status (31), total calories and sodium intake (47), the gut microbiome (48), smoking pack-years (rather than status), inhalant exposures not captured by available occupations, or other health-seeking behaviors such as those that would prompt vitamin/supplement use. Fifth, we did not have sufficient data to study results by serostatus. Sixth, it is possible some of the findings may have been spurious considering that multiple associations were measured. Therefore, these findings require confirmation. Finally, reverse causation is possible given possible misclassification of date of RA diagnosis and since patients tend to alter diet following diagnosis of RA (49). However, there was a 4.3-year interval between exposure assessment and RA.

In summary, active work-related physical activity may increase the risk of RA, though further study on this point is needed. Weak evidence suggested that other modifiable risk factors such as a high-fat diet and decreased fruit and vegetable intake might be associated with increased risk of RA. We found little evidence that alcohol, vitamin, or supplement use are protective against RA, and some may be associated with increased risk of RA. These findings help inform patient-physician discussions regarding modifiable lifestyle interventions that can impact RA.