Asthma, Chronic Obstructive Pulmonary Disease, and Subsequent Risk for Incident Rheumatoid Arthritis among Women: A Prospective Cohort Study

Abstract

Objectives:

Inflamed airways are hypothesized to contribute to rheumatoid arthritis (RA) pathogenesis due to RA-related autoantibody production, and smoking is the strongest environmental RA risk factor. However, the role of chronic airway diseases in RA development is unclear. We investigated whether asthma or COPD were associated with RA.

Methods:

We performed a prospective cohort study of 205,153 women in the Nurses’ Health Study (NHS, 1988-2014) and NHSII (1991-2015). Exposures were self-reported physician-diagnosed asthma or COPD confirmed by validated supplemental questionnaires. Outcomes were incident RA confirmed by medical record review by 2 rheumatologists. Covariates (including smoking pack-years/status) were assessed via biennial questionnaires. Multivariable hazard ratios (HRs) and 95% confidence intervals (CIs) for RA were estimated using Cox regression.

Results:

We identified 15,148 women with confirmed asthma, 3,573 with confirmed COPD, and 1,060 incident RA cases during 4,384,471 person-years of follow-up in NHS and NHSII. Asthma was associated with increased RA risk (HR 1.53, 95%CI 1.24,1.88) compared to no asthma/COPD after adjusting for covariates including smoking pack-years/status. Asthma remained associated with increased RA risk among never-smokers only (HR 1.53, 95%CI 1.14,2.05). COPD was also associated with increased RA risk (HR 1.89, 95%CI 1.31,2.75). The association of COPD with RA was most pronounced in the subgroup of ever-smokers aged >55 years (HR 2.20, 95%CI 1.38,3.51).

Conclusions:

Asthma and COPD were each associated with increased risk for incident RA, independent of smoking status/intensity and other potential confounders. These results provide support for the hypothesis that chronic airway inflammation may be crucial in RA pathogenesis.

INTRODUCTION

Patients with rheumatoid arthritis (RA) have increased respiratory morbidity and mortality(1,2). Pulmonary inflammation has been implicated in RA pathogenesis(3–6). Whether diseases of chronic airway inflammation increase risk of developing RA, however, is unclear.

Asthma is a common disease characterized by chronic airway inflammation(7). Prior studies investigating an association between asthma and RA risk(8–16) were limited by small sample size(8,9), lack of adjustment for smoking (an established RA risk factor)(8,12,13,15), and inability to measure RA phenotypes characterized by autoantibodies(9,11,13,15). Chronic obstructive pulmonary disease (COPD) is characterized by chronic inflammation and narrowing of small airways, and smoking is a proven major risk factor(17). While RA has been shown to increase risk of subsequent COPD(18–21), to our knowledge no prior prospective cohort studies have examined COPD as a risk factor for incident RA.

We investigated the associations between asthma, COPD and incident RA using two large prospective cohorts, the Nurses’ Health Study (NHS) and NHSII. We hypothesized that asthma and COPD would each increase risk of incident RA, independent of smoking.

METHODS

Study population and design

We performed a prospective cohort study by pooling two Nurses’ Health Studies, prospective cohort studies of female registered nurses in the United States. The NHS began in 1976 and enrolled 121,700 nurses aged 30-55 years; NHSII began in 1989 and enrolled 116,429 nurses aged 25-42 years. Participants completed baseline and biennial questionnaires detailing lifestyle, health behaviors, medications, and diseases. Both cohorts have >90% follow-up response rates and only 5% of person-time has been lost to follow-up(22).

Participants who reported RA or other connective tissue disease (CTD) at study baseline, had missing data related to smoking pack-years at baseline, or did not return any follow-up questionnaire after study baseline were excluded. For the asthma analysis, we also excluded participants with self-reported COPD at baseline. For the COPD analysis, we excluded participants 35 years old or younger who reported COPD as in previous studies(23), since COPD is rarely diagnosed prior to 35 years of age(24). Flow diagrams of the analyzed study populations for both the asthma and COPD analyses are presented in Supplemental Figure 1 and Supplemental Figure 2, respectively. All participants provided informed consent and the study protocol was approved by the institutional review boards of the Brigham and Women’s Hospital and Harvard T.H. Chan School of Public Health.

Exposure variables: asthma and COPD

Asthma.

Beginning with the 1988 (NHS) and 1991 (NHSII) questionnaires, participants were asked to report physician diagnosis of asthma. Positive responders were sent a previously validated supplemental respiratory questionnaire with detailed questions regarding asthma symptoms, medications, and diagnostic testing(25). The supplemental respiratory questionnaire categorized reported asthma according to the following case definitions: case definition 1 (“possible” asthma) was considered confirmed if the participant reiterated a physician diagnosis of asthma and reported using an asthma medication since diagnosis; case definition 2 (“probable” asthma) was met if the participant fulfilled case definition 1 criteria and reported use of a long-term preventive asthma medication in the past year; and case definition 3 (“definite” asthma) was met if all preceding criteria were met and participant reported physician diagnosis of asthma was made within one month of symptom onset. Camargo and colleagues validated case definition 2 within a random sample of 100 women in 1998 with high accuracy compared to the gold standard of presence of asthma by medical record review from a physician(25); we considered asthma per case definition 2 or higher (“probable” or “definite”) as confirmed asthma in our analyses. Participants who self-reported asthma but either failed to return the supplemental respiratory questionnaire or were disconfirmed per the respiratory questionnaire (did not meet criteria for case definition 2 or higher) were censored at time of initial self-report. Asthma status was time-updated during study follow-up.

COPD.

Participants self-reported physician diagnosis of emphysema or chronic bronchitis biennially starting in 1988 (NHS) and 1999 (NHSII), which was confirmed with a validated supplemental respiratory questionnaire(26). The supplemental respiratory questionnaire classified participants as “possible” COPD if they answered affirmatively to having physician-diagnosed chronic bronchitis or emphysema or COPD; “probable” COPD if criteria for “possible” case were met and the participant reported having a diagnostic test at diagnosis such as pulmonary function testing, chest radiograph, or chest computed tomography scan; or “definite” COPD if criteria for “possible” case were met and the participant reported having pulmonary function testing within the past year demonstrating forced expiratory volume in 1 second (FEV₁) less than 80% predicted or FEV₁/FVC (forced vital capacity) less than 70%. Barr and colleagues validated these definitions in a cohort of 422 women finding a positive predictive value of 88% for “probable” COPD against the gold standard of medical record review by a physician(26). We considered a nurse who self-reported COPD to have confirmed COPD if the criteria for probable or definite case was met. Participants who self-reported COPD but did not return the respiratory questionnaire or were disconfirmed per the respiratory questionnaire (did not meet criteria for probable or definite case) were censored at time of report. If a participant self-reported asthma (but asthma diagnosis was not validated by questionnaire) prior to validated COPD diagnosis, she was included as an exposed individual in the COPD analysis. COPD status was time-updated during study follow-up.

Non-exposed group: no asthma or COPD.

For each analysis, subjects contributed person-time to the non-exposed group until they self-reported asthma or COPD; if they were confirmed on validated supplemental questionnaires as asthma/COPD, they contributed person-time to that exposed group thereafter. If they reported asthma/COPD but did not return or were not validated by the supplemental questionnaire, they were censored and no longer contributed person-time to that analysis. Therefore, the non-exposed group never reported asthma or COPD up to each cycle considered in all analyses.

Outcome: Incident RA

Participants who self-reported a new diagnosis of RA were mailed the CTD Screening Questionnaire (CSQ)(27). Medical records of participants with positive CSQ were obtained and reviewed independently by two rheumatologists to identify RA cases meeting the 1987 American College of Rheumatology (ACR)(28) or 2010 ACR/European League Against Rheumatism RA classification criteria(29). Date of RA diagnosis and clinical laboratory results of rheumatoid factor (RF) and anti-cyclic citrullinated peptide antibodies (CCP) were collected from medical records. An RA case was determined to be seropositive if RF or CCP were above the upper limit of normal of the laboratory assay documented.

Covariates

We selected covariates as potential confounders associated with asthma, COPD, and RA based on prior literature(30,31,40–46,32–39), and all covariates were time-updated. Sociodemographic covariates included age, race, geographic region, and household income (categorized by quartile of US Census tract-based median household income at ZIP code level). Potential reproductive confounders were parity/total breastfeeding duration, menopausal status, and postmenopausal hormone (PMH) use. We used a combined “parity/total breastfeeding duration” variable categorized as: nulliparous, parous/0-<1 month, parous/1-11 months, or parous/≥12 months, and a combined variable for menopausal status and PMH use: premenopausal, postmenopausal/never, or postmenopausal/ever. We categorized body mass index (BMI) as: <25.0, 25.0 to <30.0, or ≥30.0 kg/m². Physical activity was categorized as <3 MET-hours/week vs. ≥3 MET-hours/week (47). Dietary intake, including alcohol consumption, was assessed by a semi-quantitative food frequency questionnaire, the Alternative Healthy Eating Index, and categorized in quartiles (48). Considering healthcare utilization as a potential confounder, we assessed whether the participant had a physical examination in the past two years on each questionnaire.

Given the associations between active and passive smoking with risk of COPD(49–51), asthma(52,53), and RA(54–57), adjusting for these was an important aspect of our analysis. On the baseline questionnaire, participants reported smoking status (never/past/current) and age at which they started smoking. Current smokers reported the number of cigarettes typically smoked per day, and past smokers provided the age at which they stopped smoking and the number of cigarettes smoked per day before quitting. On subsequent questionnaires, women reported smoking status and intensity (1–14, 15–24, ≥25 cigarettes/day). Smoking pack-years were derived by multiplying packs of cigarettes smoked per day (20 cigarettes per pack) with number of years smoked. We used smoking pack-years and smoking pack-years squared as continuous variables in our model, to include both a linear and quadratic term to account for the impact of smoking intensity on RA risk. We also adjusted for smoking status (never/past/current). All smoking variables were time-updated. To address passive smoking, participants were asked whether parents smoked in the house when participant was growing up (yes/no) and whether she lived with a smoker >1 year (ever/never).

Statistical analysis

We performed separate analyses for the co-primary exposures of asthma and COPD, each compared to participants without reported asthma or COPD. We pooled individual level data from the NHS and NHSII for statistical efficiency. We reported descriptive statistics for covariates at the baseline of this analysis (NHS 1988, NHSII 1991) in three groups: asthma and no COPD, COPD (with or without asthma), and no asthma or COPD.

Person-years of follow-up for each participant accrued from the date of return of the study baseline questionnaire (NHS 1988, NHSII 1991 for asthma; NHS 1988, NHSII 1999 for COPD) to the date of censoring, whichever came first: RA diagnosis (outcome), reported other CTD not confirmed as RA, death, loss to follow-up (date of last questionnaire submitted), or end of follow-up for this analysis (June 1, 2014 for NHS and June 1, 2015 for NHSII). For the asthma analysis, we also censored at date of self-reported COPD diagnosis. For the COPD analysis, we included participants who self-reported asthma prior to COPD that was not confirmed by supplemental questionnaire, with the rationale that self-reported asthma prior to confirmed COPD diagnosis likely represented COPD.

We used Cox proportional hazards models to test for the association between the exposure (asthma or COPD) for RA risk. Base models were adjusted for age, cohort, and questionnaire cycle (each cohort pooled by similar calendar times; e.g., the 1988 cycle in the NHS was pooled with the 1989 cycle in the NHSII). The multivariable model was additionally adjusted for the covariates discussed above. Given the possibility of collinearity among certain covariates (such as smoking status and pack-years), we initially considered partial models that adjusted for smoking status and continuous smoking pack-years separately. The point estimates were unchanged in the partial models compared to the final model that considered them together, with little evidence for collinearity between the categorical smoking status variable and the continuous smoking pack-years variable. Therefore, we reported the final model that included both of these terms since tightly controlling for possible confounding by smoking (especially for the COPD analysis) was a priority of the study. Since smoking is known to be strongly related to COPD, we expected relatively few women with COPD to be non-smokers. Therefore, we performed a subgroup analysis among ever-smokers, again using partial models for smoking status (past/current) and continuous pack-years prior to including both in the final model, all producing nearly identical point estimates for RA risk without evidence for collinearity.

We further investigated the association between asthma and RA risk by analyzing additional subgroups, including among never smokers only, and asthma at study baseline (proxy of childhood-onset since the precise age of onset prior to baseline was unavailable) or incident during follow-up (proxy of adult-onset). For COPD, we also investigated RA risk among participants who were ever-smokers and >55 years old since the prevalence of COPD is highest in this demographic. Finally, we analyzed COPD and RA risk among women with confirmed COPD who never self-reported asthma.

We tested the proportional hazards assumption by including an interaction term between time after baseline and the RA outcomes and verified no statistically significant interactions in all analyses. Two-sided p<0.05 was considered statistically significant. Analyses were performed using SAS v.9.4.

RESULTS

Sample size, asthma/COPD exposures, and RA outcomes

After baseline exclusions, there were a total of 196,409 participants included in the asthma analysis and 205,153 participants included in the COPD analysis. We identified 15,148 women with confirmed asthma, 3,573 women with confirmed COPD, and 1,060 incident RA cases (63% seropositive) during a total of 4,384,471 person-years of follow-up (median 23.9 [IQR 18.3-24.5] years for asthma analysis; median 24.0 [IQR 20.0-24.5] years for COPD analysis).

Characteristics of participants

Table 1 displays pooled baseline characteristics of the NHS and NHSII study participants categorized by exposure (asthma without COPD, COPD, and no asthma or COPD). Women in the COPD group were older with mean age of 52.7 years (compared to 42.5 in the asthma group, and 44.4 in the no asthma/COPD group). Those in the COPD group were also more likely to be postmenopausal (70.3% in the COPD group compared to 30.4% in asthma, and 34.6% in the no asthma/COPD group). Pooled baseline characteristics of the NHS and NHSII study participants included in the asthma analysis are presented in Supplemental Table 1.

Table 1.

Pooled baseline characteristics of study sample in 1988 in the Nurses’ Health Study and 1991 in the Nurses’ Health Study II (n=205,153).

	Asthma, no COPD	COPD	No asthma or COPD
n*	6,250	1,004	197,899
Cohort, %
NHS	37.9	73.9	46.3
NHSII	62.1	26.1	53.7
Mean age, years (SD)	42.5 (9.5)	52.7 (10.2)	44.4 (10.9)
White race, %	94.2	95.1	93.0
Mean body mass index, kg/m2 (SD)	25.8 (5.8)	26.4 (6.2)	25.0 (5.1)
US Geographic region, %
West	25.7	21.7	23.0
Midwest	26.5	22.3	27.0
Mid-Atlantic	33.3	37.9	35.0
New England	9.5	11.0	9.0
Southeast	5.0	7.2	6.0
Median household income, %
Quartile 1 – lowest income	22.3	30.1	26.7
Quartile 2	24.5	25.6	24.3
Quartile 3	25.7	24.7	24.4
Quartile 4 – highest income	27.5	19.6	24.6
Mean smoking pack-years (SD)	5.3 (10.1)	24.3 (24.5)	7.5 (13.6)
Ever-smokers only	13.6 (12.2)	34.0 (22.5)	17.6 (16.0)
Smoking status, %
Never	61.5	28.5	57.1
Past	31.2	39.5	28.0
Current	7.3	32.0	14.5
Ever lived with smoker, %	49.4	70.8	44.0
Parents smoked in house when growing up, %	64.4	66.8	51.8
Cumulative average physical activity <3 MET-hours/week, %	17.6	27.1	15.1
Cumulative average Alternate Healthy Eating Index, %
Quartile 1 – lowest quality diet	23.0	26.2	19.7
Quartile 2	23.7	20.8	19.8
Quartile 3	22.9	21.9	19.9
Quartile 4 – highest quality diet	25.1	21.8	19.9
Parity/breastfeeding, %
Nulliparous	19.9	11.8	15.1
Parous/<1 month	22.7	38.3	23.3
Parous/1-11 months	26.9	30.5	24.0
Parous/≥12 months	27.1	14.9	20.5
Menopausal status/postmenopausal hormone use, %
Premenopausal	69.6	29.7	65.4
Postmenopausal/never	13.3	29.3	18.1
Postmenopausal/ever	17.1	41.0	16.5

Missing data are not shown.

^*We identified a total of 15,148 women with confirmed asthma and a total of 3,573 women had confirmed COPD by the end of study follow-up for these analyses.

Asthma and RA risk

Compared to women without asthma or COPD, the multivariable-adjusted HR for developing RA was 1.53 (95%CI 1.24,1.88) among women with asthma (Table 2). Asthma was associated with both seropositive and seronegative RA, and HRs for seropositive versus seronegative RA risk were not significantly different (p for heterogeneity 0.45).

Table 2.

Hazard ratios for incident rheumatoid arthritis (overall and by serologic phenotype) by time-updated asthma compared to women without asthma or COPD in the Nurses’ Health Studies (n=196,409).

	No asthma or COPD HR (95%CI)	Asthma HR (95%CI)
Outcome: All RA
Cases/person-years	874/3,841,747	100/265,359
Age-adjusted model	1.00 (Ref)	1.67 (1.35, 2.05)
Multivariable model*	1.00 (Ref)	1.53 (1.24, 1.88)
Outcome: Seropositive RA
Cases/person-years	562/3,834,291	60/264,341
Age-adjusted model	1.00 (Ref)	1.51 (1.15, 1.97)
Multivariable model*	1.00 (Ref)	1.42 (1.08, 1.86)
Outcome: Seronegative RA
Cases/person-years	312/3,833,574	40/264,329
Age-adjusted model	1.00 (Ref)	1.98 (1.42, 2.76)
Multivariable model*	1.00 (Ref)	1.75 (1.25, 2.45)

^*Multivariable model was adjusted for age, questionnaire period, cohort, US geographic region (West, Midwest, Mid-Atlantic, New England, Southeast), median household income (quartile), smoking pack-years (continuous), smoking status (never/past/current), cumulative average physical activity (<3, ≥3 MET-hours/week), parity/breastfeeding in months (nulliparous, parous/<1 month, parous/1-11 months, parous/≥12 months), menopausal status/postmenopausal hormone use (premenopausal, postmenopausal/never, postmenopausal/ever), cumulative average Alternate Healthy Eating Index (quartile), body mass index category (<25.0, 25.0 to <30.0, ≥30.0 kg/m²), parents smoked in house when growing up (yes/no), lived with smoker (ever/never), and physical exam in last two years (yes/no).

CI, confidence interval; COPD, chronic obstructive pulmonary disease; HR, hazard ratio; RA, rheumatoid arthritis.

We examined the relationship between asthma and RA risk stratified by never- and ever-smoking (Table 3). Among never-smokers only, asthma was associated with all RA (HR 1.53, 95%CI 1.14,2.05) and seronegative RA (HR 1.90, 95%CI 1.22,2.96) but not with seropositive RA (HR 1.32, 95%CI 0.88,1.96), compared to women without asthma or COPD. Among ever-smokers only, asthma had HRs for all RA of 1.49 (95%CI 1.10,2.02), for seropositive RA of 1.50 (95%CI 1.04,2.18), and for seronegative RA of 1.48 (95%CI 0.87,2.50).

Table 3.

Hazard ratios for incident rheumatoid arthritis (overall and by serologic phenotype) by time-updated asthma compared to women without asthma or COPD in the Nurses’ Health Studies, stratified by never smoking (n=110,872) or ever smoking (n=85,537).

Never-smokers only	No asthma or COPD HR (95%CI)	Asthma HR (95%CI)	Ever-smokers only	No asthma or COPD HR (95%CI)	Asthma HR (95%CI)
Outcome: All RA			Outcome: All RA
Cases/person-years	421/2,218,758	52/159,472	Cases/person-years	453/1,622,988	48/105,887
Age-adjusted model	1.00 (Ref)	1.69 (1.27, 2.27)	Age-adjusted model	1.00 (Ref)	1.59 (1.18, 2.15)
Multivariable model*	1.00 (Ref)	1.53 (1.14, 2.05)	Multivariable model**	1.00 (Ref)	1.49 (1.10, 2.02)
Outcome: Seropositive RA			Outcome: Seropositive RA
Cases/person-years	268/2,214,956	28/158,816	Cases/person-years	294/1,619,336	32/105,525
Age-adjusted model	1.00 (Ref)	1.40 (0.95, 2.08)	Age-adjusted model	1.00 (Ref)	1.57 (1.08, 2.27)
Multivariable model*	1.00 (Ref)	1.32 (0.88, 1.96)	Multivariable model**	1.00 (Ref)	1.50 (1.04, 2.18)
Outcome: Seronegative RA			Outcome: Seronegative RA
Cases/person-years	153/2,214,780	24/158,973	Cases/person-years	159/1,618,793	16/105,356
Age-adjusted model	1.00 (Ref)	2.23 (1.44, 3.46)	Age-adjusted model	1.00 (Ref)	1.64 (0.97, 2.75)
Multivariable model*	1.00 (Ref)	1.90 (1.22, 2.96)	Multivariable model**	1.00 (Ref)	1.48 (0.87, 2.50)

^*Multivariable model was adjusted for age, questionnaire period, cohort, US geographic region (West, Midwest, Mid-Atlantic, New England, Southeast), median household income (quartile), cumulative average physical activity (<3, ≥3 MET-hours/week), parity/breastfeeding in months (nulliparous, parous/<1 month, parous/1-11 months, parous/≥12 months), menopausal status/postmenopausal hormone use (premenopausal, postmenopausal/never, postmenopausal/ever), cumulative average Alternate Healthy Eating Index (quartile), body mass index category (<25.0, 25.0 to <30.0, ≥30.0 kg/m²), parents smoked in house when growing up (yes/no), lived with smoker (ever/never), and physical exam in last two years (yes/no).

^**The ever smoker analysis was additionally adjusted for smoking pack-years (continuous) and smoking status (never/past/current).

CI, confidence interval; COPD, chronic obstructive pulmonary disease; HR, hazard ratio; RA, rheumatoid arthritis.

Table 4 shows the associations of prevalent asthma at study baseline (proxy for childhood-onset) and incident asthma during follow-up (proxy for adult-onset) with RA, compared to reference group of women without asthma or COPD. Overall RA risk was significantly increased in both prevalent asthma (HR 1.46, 95%CI 1.06,2.01) and incident asthma (HR 1.61, 95%CI 1.23,2.09).

Table 4.

Hazard ratios for incident rheumatoid arthritis (overall and by serologic phenotype) according to prevalent asthma at study baseline or incident asthma during follow-up, each compared to women without asthma or COPD in the Nurses’ Health Studies (n=196,409).

	No asthma or COPD HR (95%CI)	Prevalent asthma at baseline HR (95%CI)
Outcome: All RA
Cases/person-years	855/3,792,384	40/115,491
Age-adjusted model	1.00 (Ref)	1.60 (1.16, 2.20)
Multivariable model*	1.00 (Ref)	1.46 (1.06, 2.01)
Outcome: Seropositive RA
Cases/person-years	548/3,785,152	23/115,007
Age-adjusted model	1.00 (Ref)	1.37 (0.90, 2.09)
Multivariable model*	1.00 (Ref)	1.29 (0.85, 1.97)
Outcome: Seronegative RA
Cases/person-years	307/3,784,460	17/115,063
Age-adjusted model	1.00 (Ref)	2.05 (1.25, 3.36)
Multivariable model*	1.00 (Ref)	1.79 (1.09, 2.94)
	No asthma or COPD HR (95%CI)	Incident asthma during follow-up HR (95%CI)
Outcome: All RA
Cases/person-years	874/3,841,695	60/148,348
Age-adjusted model	1.00 (Ref)	1.76 (1.35, 2.29)
Multivariable model*	1.00 (Ref)	1.61 (1.23, 2.09)
Outcome: Seropositive RA
Cases/person-years	562/3,834,240	37/147,814
Age-adjusted model	1.00 (Ref)	1.65 (1.18, 2.31)
Multivariable model*	1.00 (Ref)	1.56 (1.11, 2.18)
Outcome: Seronegative RA
Cases/person-years	312/3,833,522	23/147,767
Age-adjusted model	1.00 (Ref)	1.96 (1.28, 3.01)
Multivariable model*	1.00 (Ref)	1.73 (1.13, 2.66)

^*Multivariable model was adjusted for age, questionnaire period, cohort, US geographic region (West, Midwest, Mid-Atlantic, New England, Southeast), median household income (quartile), smoking pack-years (continuous), smoking status (never/past/current), cumulative average physical activity (<3, ≥3 MET-hours/week), parity/breastfeeding in months (nulliparous, parous/<1 month, parous/1-11 months, parous/≥12 months), menopausal status/postmenopausal hormone use (premenopausal, postmenopausal/never, postmenopausal/ever), cumulative average Alternate Healthy Eating Index (quartile), body mass index category (<25.0, 25.0 to <30.0, ≥30.0 kg/m²), parents smoked in house when growing up (yes/no), lived with smoker (ever/never), and physical exam in last two years (yes/no).

CI, confidence interval; COPD, chronic obstructive pulmonary disease; HR, hazard ratio; RA, rheumatoid arthritis.

COPD and RA risk

Compared to women without asthma or COPD, the multivariable-adjusted HR for developing RA was 1.89 (95%CI 1.31,2.75) among women with COPD (Table 5). COPD significantly increased risk for seropositive RA (HR 2.07, 95%CI 1.31,3.25) but not seronegative RA (HR 1.59, 95%CI 0.83,3.05). Among the subgroup of ever-smokers aged >55 years, there was a stronger association between COPD and seropositive RA (HR 2.85, 95%CI 1.63,4.99; Table 6). Among women with confirmed COPD who never self-reported asthma, COPD was associated with RA (HR 2.57, 95%CI 1.51,4.39).

Table 5.

Hazard ratios for incident rheumatoid arthritis (overall and by serologic phenotype) by time-updated COPD compared to women without asthma or COPD in the Nurses’ Health Studies (n=205,153).

	No asthma or COPD HR (95%CI)	COPD HR (95%CI)
Outcome: All RA
Cases/person-years	1,029/4,337,186	31/47,285
Age-adjusted model	1.00 (Ref)	2.39 (1.66, 3.43)
Multivariable model*	1.00 (Ref)	1.89 (1.31, 2.75)
Outcome: Seropositive RA
Cases/person-years	642/4,328,257	21/47,134
Age-adjusted model	1.00 (Ref)	2.69 (1.73, 4.18)
Multivariable model*	1.00 (Ref)	2.07 (1.31, 3.25)
Outcome: Seronegative RA
Cases/person-years	387/4,327,740	10/47,121
Age-adjusted model	1.00 (Ref)	1.93 (1.02, 3.64)
Multivariable model*	1.00 (Ref)	1.59 (0.83, 3.05)

^*Multivariable model was adjusted for age, questionnaire period, cohort, US geographic region (West, Midwest, Mid-Atlantic, New England, Southeast), median household income (quartile), smoking pack-years (continuous), smoking pack-years-squared (continuous), smoking status (never/past/current), cumulative average physical activity (<3, ≥3 MET-hours/week), parity/breastfeeding in months (nulliparous, parous/<1 month, parous/1-11 months, parous/≥12 months), menopausal status/postmenopausal hormone use (premenopausal, postmenopausal/never, postmenopausal/ever), cumulative average Alternate Healthy Eating Index (quartile), body mass index category (<25.0, 25.0 to <30.0, ≥30.0 kg/m²), parents smoked in house when growing up (yes/no), lived with smoker (ever/never), and physical exam in last two years (yes/no).

CI, confidence interval; COPD, chronic obstructive pulmonary disease; HR, hazard ratio; RA, rheumatoid arthritis.

Table 6.

Hazard ratios for incident rheumatoid arthritis (overall and by serologic phenotype) by time-updated COPD compared to women without asthma or COPD in the Nurses’ Health Studies among ever smokers aged >55 years (n=21,525).

	No asthma or COPD HR (95%CI)	COPD HR (95%CI)
Outcome: All RA
Cases/person-years	295/928,014	21/29,365
Age-adjusted model	1.00 (Ref)	2.26 (1.44, 3.55)
Multivariable model*	1.00 (Ref)	2.20 (1.38, 3.51)
Outcome: Seropositive RA
Cases/person-years	176/926,338	15/29,271
Age-adjusted model	1.00 (Ref)	2.80 (1.64, 4.80)
Multivariable model*	1.00 (Ref)	2.85 (1.63, 4.99)
Outcome: Seronegative RA
Cases/person-years	119/926,271	6/29,279
Age-adjusted model	1.00 (Ref)	1.52 (0.66, 3.50)
Multivariable model*	1.00 (Ref)	1.40 (0.59, 3.29)

^*Multivariable model was adjusted for age, questionnaire period, cohort, US geographic region (West, Midwest, Mid-Atlantic, New England, Southeast), median household income (quartile), smoking pack-years (continuous), smoking pack-years-squared (continuous), smoking status (past/current), cumulative average physical activity (<3, ≥3 MET-hours/week), parity/breastfeeding in months (nulliparous, parous/<1 month, parous/1-11 months, parous/≥12 months), menopausal status/postmenopausal hormone use (premenopausal, postmenopausal/never, postmenopausal/ever), cumulative average Alternate Healthy Eating Index (quartile), body mass index category (<25.0, 25.0 to <30.0, ≥30.0 kg/m²), parents smoked in house when growing up (yes/no), lived with smoker (ever/never), and physical exam in last two years (yes/no).

CI, confidence interval; COPD, chronic obstructive pulmonary disease; HR, hazard ratio; RA, rheumatoid arthritis.

DISCUSSION

In this large prospective cohort study with lengthy follow-up, asthma was associated with a greater than 50% increase in the risk of subsequent RA compared to no asthma/COPD, independent of potential confounders, most notably smoking status and duration/intensity. COPD conferred a nearly 90% increased risk of developing RA compared to no asthma/COPD in this cohort after multivariable adjustment including adjustment for smoking, with a greater than two-fold increased risk of RA among older smokers. These findings identify asthma and COPD as risk factors for the development of rheumatoid arthritis, and to our knowledge, this is the first prospective study to examine asthma or COPD as RA risk factors. These findings help identify asthma and COPD patients as an at-risk population for purposes of research into RA development, and provide helpful information for clinicians caring for asthma/COPD patients about elevated RA risk in these groups.

While one retrospective study of Israeli soldiers found an inverse association between asthma and RA(12), the majority of the preexisting literature suggests asthma increases RA risk(8–11,13,15,16). Sheen et al(9) identified asthma via medical record review and RA outcomes using International Classification of Disease codes in a population-based case-control study, finding asthma had odds ratio (OR) 1.73 (95%CI 1.03,2.92) for RA compared to matched controls. While this study adjusted for several factors including smoking status and had high diagnostic accuracy for asthma, there was no adjustment for smoking duration/intensity, and sample size limited ability to examine RA serologic status. Kronzer and colleagues(10) identified RA cases in a biobank population using a rules-based algorithm, finding that self-reported asthma had increased OR 1.28 (95%CI 1.04,1.67) for RA compared to matched controls. Similar to Sheen et al(9), the authors adjusted for smoking status but not duration/intensity, and the results of this clinically sampled population may lack generalizability. While our study similarly finds a positive association between asthma and RA risk, our finding adds to the literature by nature of the prospective, detailed data collection with time-updated adjustment for multiple covariates including smoking status as well as smoking duration/intensity. Furthermore, we utilized time-updated exposure and outcome assessment using confirmed, not self-reported, diagnoses of both asthma and RA with serologic phenotype. Our large sample size enabled us to phenotype asthma exposures (according to never/ever smoking status, and incident/prevalent disease) and RA outcomes (according to serologic status).

Prior literature has demonstrated a positive association between RA and risk of developing COPD(18–21). However, we identified only two studies examining COPD as an RA risk factor, neither of which were prospective. A Swedish nested case-control study(58) examined COPD (per GOLD Stage from spirometry performed for research purposes) as a risk factor for RA, and ORs were non-significant, likely due to small number of COPD exposures. In the previously discussed case-control study by Sheen et al(9), COPD was not associated with RA as an unadjusted baseline variable. Our findings are the first to demonstrate COPD as a risk factor for seropositive RA after adjustment for smoking. While our primary analysis for COPD allowed women to also self-report asthma, the secondary analysis that only considered confirmed COPD without asthma self-report had an even stronger association of COPD with increased RA risk. Therefore, it is unlikely that the COPD results were explained by participants who may have also had pre-existing or concurrent asthma.

There is increasing interest in the lung and airway mucosa as a site for RA pathogenesis. RA-specific autoantibodies are increased in the sputum of unaffected first-degree relatives of RA patients prior to detected elevation in the serum(59). In newly diagnosed RA patients, lymphoid aggregates are present near airways and interstitium(5,6). These findings provided the biologic underpinning of our hypotheses that asthma and COPD would increase RA risk. We might have also expected stronger associations between asthma/COPD and seropositive RA compared to seronegative RA. While we did observe this in COPD, there was not a statistically significant difference in the HRs for seropositive and seronegative RA in our primary asthma analysis. Prevalent asthma at study baseline was significantly positively associated with overall and seronegative RA, but not seropositive RA. We were unable to investigate RF+ RA and CCP+ RA separately as we relied on clinical testing and the CCP test was not available in the early years of follow-up. It is unclear whether these discrepancies in risk according to RA serologic status have biological and clinical meaning, or whether the smaller numbers of exposed cases in certain analyses affected statistical power.

Our study has several key strengths. We used a validated method to identify women with asthma and COPD throughout follow-up based on self-report and then confirmed on a supplemental respiratory questionnaire. The reference group had never reported asthma or COPD on every main questionnaire. Since many women reported asthma/COPD during the study, manual medical record review to confirm presence or absence of these diseases would have been infeasible. Therefore, our methods allowed us to compare confirmed asthma/COPD to women who never reported these diseases throughout the lengthy follow-up in these large prospective cohorts. We have a high level of confidence in RA outcomes and date of RA diagnosis, which were agreed upon by two rheumatologists according to classification criteria. The NHS and NHSII are large cohorts with detailed prospectively collected data, which permitted adjustment for multiple time-updated covariates, most importantly smoking (both smoking status and duration/intensity). Our findings were robust to sensitivity analysis regarding smoking (particularly the positive association between asthma and RA among never-smokers only).

The limitations of this study include that the Nurses’ Health Studies included only women, the majority white, limiting generalizability. While we used validated methods to identify confirmed asthma/COPD, the supplemental questionnaire was sent only once for the purpose of establishing presence of these diseases. The presence of these diseases has acceptable validity when compared to medical records, but it is still possible that there was some misclassification since both of the separate questionnaires were by self-report. We mitigated the possibility of misclassification by not analyzing ambiguous cases that reported once but not again or were not deemed to have high likelihood of truly having asthma or COPD. Furthermore, inclusion of women without true asthma or COPD in these analyses would be expected to bias results to the null, and among ever-smokers aged >55 years (a group at higher risk for COPD), we observed the strongest association with RA risk. Further, reports have quantified similar proportions of non-smokers with COPD as we observed, perhaps from air pollution, occupational exposures, or other inhalants(60). Since data on clinical characteristics of asthma/COPD were only collected once, we were unable to analyze how the disease course of asthma/COPD may impact RA risk. We also had limited ability to analyze asthma and COPD according to disease severity, imaging/pulmonary function test abnormalities, medication use, or (in the case of asthma) atopy. Future studies should investigate whether characteristics of asthma/COPD based on subtype, severity, medication use, and flares are associated with RA risk and RA-related autoantibody production.

As expected, women with confirmed COPD were older so were more likely to be postmenopausal. It is possible that they may have had other differences related to hormonal/reproductive factors that could also impact RA risk. Our multivariable analyses therefore included the following hormonal/reproductive factors: parity, breastfeeding duration, menopausal status, and postmenopausal hormone use. We further mitigated the possibility that the COPD results may have been explained by hormonal differences by performing a secondary analysis for COPD among only older women. This showed that COPD remained strongly associated with seropositive RA so is unlikely that differences in menopause explained our results. While we measured and adjusted for several reproductive/hormonal factors, it is always possible that residual unmeasured confounding remains. Regarding the strong association between COPD and subsequent RA risk among ever-smokers aged >55 years, it is possible that hormonal factors may particularly impact RA risk in this sub-group; we mitigated this possibility by adjusting for menopausal status/post-menopausal hormone use as well as parity/breastfeeding. We chose the cut-off point of age 55 years to classify women as younger or older in that particular analysis based on prior literature(61) as well as to reflect the menopausal transition and since COPD is known to be an older-onset disease, though we acknowledge the possibility that the selection of a different age cut-off may have yielded different results.

Finally, inherent to any observational study is the possibility of unmeasured confounding, particularly related to healthcare utilization.

In this large prospective cohort study of women, asthma and COPD were risk factors for subsequent rheumatoid arthritis after adjustment for multiple potential confounders including smoking status and duration/intensity. These novel findings further implicate chronic airway inflammation in the pathogenesis of RA, and they identify populations at-risk for RA for the purposes of research as well as informing clinical care. Providers caring for patients with asthma or COPD should be aware of increased RA risk in these populations and have a low threshold to evaluate for RA in asthma or COPD patients with inflammatory joint symptoms.