A Combination of Healthy Lifestyle Behaviors Reduces Risk of Incident Systemic Lupus Erythematosus

Abstract

Objective:

While the association between individual factors related to lifestyle and the risk of SLE has been previously evaluated, it is unclear how the combination of these factors might affect the risk of incident SLE. We prospectively evaluated whether a combination of healthy lifestyle factors is associated with lower risk of incident SLE and its subtypes (dsDNA positive and negative).

Methods:

We included 185,962 women with 203 incident SLE cases (96 anti-dsDNA positive,107 anti-dsDNA negative) during 4,649,477 person-years of follow-up from the Nurses’ Health Study (NHS) and NHSII cohorts. The Healthy Lifestyle Index Score (HLIS) was calculated at baseline and approximately every 2 years in follow-up using five factors: alcohol consumption, body mass index, smoking, diet, and exercise. A time-varying Cox hazards regression model estimated the adjusted hazard ratios (HR) for SLE risk. The partial population attributable risk (PAR%) of SLE development was calculated.

Results:

A higher HLIS was associated with a lower risk of SLE overall (HR 0.81 [95%CI 0.71–0.94]) and dsDNA positive SLE (HR 0.78 [95%CI 0.63–0.95]). Women with four or more healthy factors had the lowest SLE risk overall (HR 0.42 [95%CI 0.25–0.70]) and dsDNA positive SLE (HR 0.35 [95%CI 0.17–0.75]) compared to women with less than or equal to one healthy behavior. The PAR% for SLE from adhering to four or more healthy behaviors was 47.7% [95%CI 23.1–66.6].

Conclusion:

Nearly half of SLE risk, a disease where significant evidence of genetic involvement has been established, might be reduced with adherence to modifiable healthy lifestyle behaviors.

Introduction

Systemic lupus erythematosus (SLE) is a chronic multisystem inflammatory autoimmune condition that remains a leading cause of death among young women (1). It was recently estimated that 200,000 people in the U.S. have SLE by classification criteria, with prevalence nine times higher among females than males (2). Although the mortality risk in SLE patients has decreased substantially over the past few decades, many will still experience progressive and/or complicated disease, underscoring the public health impact of SLE and the importance of disease prevention research given that few effective medical therapies are available.

A complex interplay between genetic factors and environmental exposures are thought to ultimately lead to autoimmunity in SLE. About 5–12% of subjects with a first degree relative with SLE will develop SLE in their lifetime, whereas SLE will develop in up to 90% of persons with a congenital deficiency of the complement component C4 (3). Environmental exposures such as ultraviolet light (4), medications (5), infectious agents (6), silica (7), cigarette smoke (8, 9), alcohol (9–12), hormonal factors (13), and obesity (14) have been hypothesized to be associated with SLE development, although the strength of these associations varies. Potential biological mechanisms linking environmental exposures and SLE risk include increased systemic inflammation, oxidative stress, upregulation in inflammatory cytokines, and epigenetic modifications (15). A strong association between smoking and risk of developing SLE that is specifically positive for anti-double stranded DNA (dsDNA) antibodies, a more severe subtype of SLE, also support the possibility that smoking may have a direct role in anti-dsDNA production (8).

In previous SLE studies, individual modifiable lifestyle factors were examined within the Nurses’ Health Study (NHS), Nurses’ Health Study II (NHSII) (8, 10, 14, 16, 17) and other cohorts (9, 11, 12, 18, 19). Individual healthy lifestyle factors have also been shown to reduce the risk of developing other autoimmune diseases such as rheumatoid arthritis (RA) (20, 21) and multiple sclerosis (22). However, people generally follow a pattern of lifestyle behaviors as these factors are often correlated with each other. Studies have reported that having a relatively high healthy lifestyle index score (HLIS) (a favorable combination of meeting recommended guidelines for diet, exercise, smoking, alcohol consumption, and body weight) was associated with reduced risk of developing various diseases including cardiovascular disease (23), stroke (24), sudden cardiac death (25), diabetes (26), and cancer (27, 28), as well as an increased overall life expectancy (29). This has also been implicated in autoimmune diseases such as RA (30). These studies suggest that adopting a healthier lifestyle could reduce chronic disease risk including autoimmune conditions; therefore, primary prevention through lifestyle interventions should be promoted.

The joint impact of multiple healthy behaviors and maintaining healthy body weight on the prevention of SLE development has not been assessed. Furthermore, as some risk factors have been associated with anti-dsDNA positivity, studying specific SLE subtypes may provide insight into potential mechanisms of disease pathogenesis. Therefore, in this study, we prospectively evaluated whether a healthy lifestyle as measured by HLIS in the NHS and NHSII was associated with a lower risk of incident SLE and its subtypes stratified by dsDNA positivity.

Patient and Methods

Study design and population

The NHS and NHSII cohorts were established in 1976 and 1989, respectively (31). A total of 121,700 married female registered nurses, aged 30–55 years from 11 U.S. states, were enrolled in NHS; 116,430 married female registered nurses, aged 25–42 years from 14 U.S. states, were enrolled in NHSII. Although the participants had slightly higher socioeconomic status compared to the general population and were mostly White (97%), the nurses’ health knowledge and commitment to the research provided high-quality data. Furthermore, follow-up rates in these longitudinal cohort studies have been high with only ~5% of person-time lost to follow-up (32). Since NHSII is a younger cohort compared to NHS, new information on exposures in adolescence and early adult life was obtained and included more details on oral contraceptive use and other reproductive risk factors. Questionnaires including assessments of a range of lifestyle factors, health-related behaviors, and the development of new diseases including SLE and other outcomes were mailed to and completed by participants at baseline and then biennially in follow-up. A comprehensive self-administered validated Food Frequency Questionnaire (FFQ) with over 130 items was mailed about every four years starting in 1984 in NHS and 1991 in NHSII. Past validation studies demonstrate that lifestyle behaviors such as dietary patterns and BMI are stable over time (33–35).

The current analysis excludes participants with prevalent SLE or other connective tissue disease (CTD) at baseline (n=7,177 in NHS and n=1,371 in NHSII). We included 96,240 women in NHS (followed 1986–2016) and 105,460 women in NHSII (followed 1991–2017). This study was approved by the Partners’ HealthCare Institutional Review Board.

SLE identification

Incident SLE in NHS cohorts was identified in two stages (36). Participants self-reported newly diagnosed diseases biennially on their follow-up questionnaires. For any new report of SLE, the participant was asked to complete the validated CTD Screening Questionnaire, which included 13 questions concerning symptoms of SLE (36). Medical records for participants who screened positive were requested and independently reviewed by two board-certified rheumatologists to confirm whether patients fulfilled the Updated American College of Rheumatology 1997 SLE Classification Criteria (37). Methods for SLE case identification and validation have previously been reported (13, 36). The reviewers were blinded to the questionnaire exposure data. Participants with prevalent SLE or other CTD at enrolment were excluded. Participants were censored during follow-up upon self-report of a non-SLE CTD, or upon SLE self-report not confirmed by medical record review. Anti-dsDNA status at SLE diagnosis was determined by medical record review. Laboratory tests including the anti-dsDNA were performed using standard assays at participating sites. The secondary outcomes were dsDNA positive and negative SLE subtypes.

Healthy Living Index Score

Healthy lifestyle index score (HLIS) was calculated at baseline and approximately every two years in follow-up. The HLIS ranges from 0 (least healthy) to 5 (most healthy), computed as a sum of healthy factors. The components of the HLIS were based on five traditional lifestyle factors that were identified in prior NHS and NHSII HLIS studies (23, 26). Following the same HLIS scoring system and definitions used in these studies, we applied a binary score for each factor where participants were assigned one point for each of the following low-risk behaviors: never smokers and past smokers (quit >4 years ago), not being overweight or obese (body mass index [BMI] <25 kg/m²), drinking alcohol in moderation (5g/day or higher) as opposed to low alcohol consumption (high-risk behavior), healthy diet (highest 40th percentile of the Alternative Healthy Eating Index [AHEI]) (38), and regular exercise (at least 19 metabolic equivalent [MET] hours per week, corresponding to at least 30 minutes of brisk walking every day). As with prior HLIS studies, subjects who were underweight (BMI <18.5 kg/m²) or drank more 30g of alcohol per day were not excluded from the study. BMI was calculated using updated self-reported height and weight and treated as a dichotomous variable. Self-reported weight (r=0.97) and alcohol intake (r=0.9) in NHS has been previously validated as highly accurate (34, 39).

The HLIS was computed using those with complete component data. Several methods, defined a priori, were used for handling missing data. Subjects with missing item(s) in every questionnaire cycle were excluded. Individual mean imputation was used for alcohol consumption and exercise (using the cumulative averaged METS). For missing BMI, the subject’s previous BMI was carried forward twice and then individual mean imputation was used for further missing values. AHEI was updated approximately every 4 years; for missing AHEI, the subject’s previous AHEI was carried forward once, and then individual mean imputation was used for further missing values. Smoking data were carried forward if necessary; subjects with missing data in every cycle were excluded.

Time-varying covariates

Demographic and clinical data were updated on biennial questionnaires. Race was treated as binary (White vs. non-White). Household median income for each U.S. Census-tract as a marker of socioeconomic status was categorized by quartiles. We also included potential reproductive covariates associated with incident SLE as confounders, including oral contraceptive use (never vs. ever), age at menarche (≤10 years vs. >10 years), menopausal status (pre-menopausal, post-menopausal/never used post-menopausal hormones, and post-menopausal/ever used post-menopausal hormones) (13).

Statistical methods

Time-varying Cox regression models estimated the multivariable-adjusted hazard ratio (HRs [95% confidence intervals]) of incident SLE associated with the number of healthy lifestyle factors, adjusting for potential confounders, both overall and by dsDNA subtype. Confounders were chosen a priori based on existing literature. We calculated the p-trend from the continuous HLIS model. P-values <0.05 were considered statistically significant. Due to small number of SLE cases in the extreme categories of HLIS, four categories of the HLIS were created by collapsing HLIS=0/1 and HLIS=4/5 into single categories. We calculated the population attributable risk percent (PAR%) (40, 41), an estimate of the percentage of incident SLE cases in this population during follow-up that would not have occurred if all participants had at least four healthy lifestyle behaviors, assuming a causal relation.

We performed a sensitivity analysis to calculate a SLE-specific HLIS based on a priori- established SLE risk factors alone (10, 14) including smoking, BMI, and alcohol consumption, but not exercise or dietary quality as the impact of regular exercise on SLE has not been examined and an association between dietary quality and SLE risk has not been found (16, 17). An additional 80 SLE cases were included in this analysis because they developed SLE during the eight years prior to the first diet questionnaire. Participants were assigned one point for each of the following low-risk behaviors: never smokers and past smokers (quit > 4 years ago), normal BMI (18–24.9 kg/m2, excluding underweight participants as done in a prior SLE NHS/NHSII study due to insufficient numbers (14)), and drinking alcohol in moderation (5g/day or higher). Additional sensitivity analyses included: 1) redefining smoking to never smokers vs. ever smokers; 2) redefining moderate alcohol consumption to >=5g/day to <30g/day to ensure that this association was not driven by outliers; and 3) excluding incident cases occurring within 12 months of HLIS assessment (e.g., lag analysis), to address potential reverse causation whereby underlying disease might affect lifestyle factors. Analyses were conducted using SAS 9.4 (SAS Institute, Cary NC).

Results

Characteristics of Participants

During 4,649,477 person-years of follow-up, there were 203 incident SLE cases (96 anti-dsDNA positive and 107 anti-dsDNA negative), diagnosed at a median duration of 10.8 years (range 0.3–38.4 years) after enrolment. Table 1 summarizes the study baseline and disease characteristics of the participants. The mean HLIS was 1.4 (standard deviation (SD) 1.0).

Table 1.

Baseline characteristics in Nurses’ Health Study (NHS) (1986) and NHSII (1991) by traditional Healthy Living Index Score (HLIS) (n=185,962)

	HLIS Category
Characteristic	0–1 (n=40715)	2 (n=62188)	3 (n=51110)	4–5 (n=31949)
Sociodemographic
Mean age, years (SD)	44.81 (9.71)	43.46 (10.02)	43 (10.03)	42.67 (9.93)
White, %	93.13	93.20	92.60	93.61
Census-tract median household income by zip code <$60,000, %	30.27	25.66	21.30	16.40
Components of the HLIS
Mean BMI, kg/m2 (SD)	28.94 (5.72)	25.01 (5)	23.3 (3.7)	22.06 (2.27)
BMI <25 kg/m 2 , %	17.30	60.10	79.78	95.01
Mean alcohol consumption, g/day (SD)	2.07 (5.29)	3.89 (8.04)	5.35 (8.32)	8.13 (8.73)
Alcohol >= 5 g/day, %	5.86	18.63	33.70	61.34
Never or past smokers (quit >4 years), %	54.25	78.43	86.95	94.45
Regular exercise (>= 19 MET-hrs/week), %	3.58	17.10	46.43	82.96
Highest 40 percentile of the AHEI, %	5.57	25.74	53.14	85.35
Medications and reproductive factors
Oral contraceptive use, %	67.65	67.77	68.40	70.59
Age at menarche <=10 years, %	9.26	6.71	6.15	5.72
Pre-menopausal, %	66.60	68.02	68.80	69.62
Post-menopausal, never used post-menopausal hormones, %	16.87	14.81	13.50	12.13
Post-menopausal, ever used post-menopausal hormones, %	13.87	14.66	15.20	15.92

Abbreviations: AHEI, Alternative Healthy Eating Index; BMI, body mass index; HLIS, Healthy Living Index Score; hrs, hours; MET, metabolic equivalent; SD, standard deviation

Supplemental Table 1 shows the characteristics of the 203 incident SLE cases. The mean age of diagnosis was 52.1 (SD 11.4) with most patients of self-reported White race. Over 98% of SLE cases had a positive anti-nuclear antibody test (ANA) and 47.3% had positive anti-dsDNA at time of diagnosis.

Incident SLE Risk Using the Traditional HLIS

In multivariable models, a higher continuous traditional HLIS was associated with a lower SLE risk overall (HR 0.81 [95%CI 0.71–0.94]) and lower dsDNA positive SLE risk (HR 0.78 [95%CI 0.63–0.95]) per unit increase (Table 2). The risk of dsDNA negative SLE was also reduced per unit increase in HLIS (HR 0.85 [95%CI 0.70–1.03]), although this was not statistically significant. Women in the highest category of HLIS with at least four healthy behaviors had the lowest risks of developing SLE overall (HR 0.42 [95%CI 0.25–0.70]) and dsDNA positive SLE (HR 0.35 [95%CI 0.17–0.75]) compared to women with one or no healthy behaviors. The PAR% for SLE risk was 47.7% [95%CI 23.1–66.6%] assuming the entire population had adhered to at least four healthy lifestyle behaviors.

Table 2.

Hazard ratios (95%CIs) for incident SLE risk among women in Nurses’ Health Study (1986–2016) and Nurses’ Health Study II (1991–2017) in relation to the traditional Healthy Lifestyle Index Score¹ defined as drinking alcohol in moderation (>=5g/day), maintaining a healthy body weight (body mass index <25 kg/m2), never or past smoking (quit more than 4 years ago), healthy diet (highest 40th percentile of the Alternative Healthy Eating Index (AHEI), and regular exercise (at least 19 metabolic equivalents (MET) hours per week

	Overall SLE (n=203)			dsDNA Positive SLE (n=96)			dsDNA Negative SLE (n=107)
Number of Healthy Behaviors	Cases/ person-years	MV HR (95%CI)2	p-trend	Cases/ person-years	MV HR (95%CI)2	p-trend	Cases/ person-years	MV HR (95%CI)2	p-trend
0–1	58/ 1079480	1.00 (ref)	0.004	30/ 1079160	1.00 (ref)	0.016	28/ 1079143	1.00 (ref)	0.10
2	67/ 1490963	0.81 (0.57–1.15)		30/ 1490581	0.72 (0.43–1.20)		37/ 1490576	0.92 (0.56–1.51)
3	58/ 1220655	0.85 (0.59–1.23)		27/ 1220331	0.78 (0.46–1.32)		31/ 1220310	0.94 (0.56–1.57)
4–5	20/ 858380	0.42 (0.25–0.70)		9/ 858223	0.35 (0.17–0.75)		11/ 858229	0.50 (0.24–1.01)
Per unit increase in behavior	203/ 4649477	0.81 (0.71–0.94)		96/ 4648294	0.78 (0.63–0.95)		107/ 4648257	0.85 (0.70–1.03)

Abbreviations: CI, confidence interval; HR, hazard ratio; MV, multivariable; PY, person-years; ref, reference.

¹Healthy lifestyle index score (HLIS) as a summed score from 0–1 (one or no healthy behaviors) to 4–5 (4 or more healthy behaviors) or continuous score (per 1 unit of behavior change)

²Multivariable (MV)-adjusted hazard ratios (HR) from time-varying Cox proportional hazards models, adjusted for age (months), questionnaire cycle, cohort, non-white race, census-tract median household income (quartiles), ever oral contraceptive use, age at menarche <=10, menopausal status (pre-menopausal vs. post-menopausal never/ever used post-menopausal hormones)

Incident SLE Risk Using the SLE-Specific HLIS

When using only previously established SLE-specific healthy lifestyle behaviors (alcohol consumption, BMI and cigarette smoking), there were 283 incident SLE cases (120 anti-dsDNA positive and 163 anti-dsDNA negative) in 5,815,211 person-years of follow-up. Their patient characteristics (Supplemental Table 2 and 3) were similar to the participants in the primary analysis. The overall incidence rate for SLE (4.9 cases per 100,000 person-years) was similar to the primary analysis cohort.

In multivariable models, a higher continuous SLE-specific HLIS was associated with a lower SLE risk overall (HR 0.75 [95%CI 0.64–0.88]), dsDNA positive subtype (HR 0.67 [95%CI 0.53–0.85]), and dsDNA negative subtype (HR 0.82 [95%CI 0.67–1.00]) per unit increase (Table 3). Those in the highest category of HLIS (three healthy behaviors) had a reduced SLE risk overall (HR 0.45 [95%CI 0.24–0.85]) compared to women with no healthy behaviors. For dsDNA positive subtype, those with two healthy behaviors (HR 0.46 [95%CI 0.23–0.93]) or three healthy behaviors (HR 0.39 [95%CI 0.16–0.94]) had reduced SLE risk. Unlike the traditional HLIS, for the SLE-specific HLIS, a larger reduction in SLE risk can be observed for every increase in HLIS category. The SLE-specific HLIS PAR% for SLE risk was 43.4% [95%CI: 17.4–63.8%] assuming the entire population had adhered to three healthy lifestyle behaviors.

Table 3.

Hazard ratios (95% CIs) for incident SLE risk among women in Nurses’ Health Study (1976–2016) and Nurses’ Health Study II (1989–2017) in relation to the SLE-specific Healthy Lifestyle Index Score¹ defined as drinking alcohol in moderation (>=5g/day), maintaining a healthy body weight (body mass index 18.5–24.9 kg/m2), and never or past smoking (quit more than 4 years ago)

	Overall SLE (n=283)			dsDNA Positive SLE (n=120)			dsDNA Negative SLE (n=163)
Number of Healthy Behaviors	Cases/ person-years	MV HR (95%CI)2	p-trend	Cases/ person-years	MV HR (95%CI)2	p-trend	Cases/ person-years	MV HR (95%CI)2	p-trend
0	19/ 269502	1.00 (ref)	0.0003	10/ 269391	1.00 (ref)	0.001	9/ 269379	1.00 (ref)	0.048
1	133/ 2337444	1.00 (0.61–1.62)		62/ 2336573	0.80 (0.41–1.58)		71/ 2336593	1.22 (0.60–2.45)
2	110/ 2444354	0.73 (0.45–1.19)		38/ 2443370	0.46 (0.23–0.93)		72/ 2443709	1.05 (0.52–2.12)
3	21/ 763911	0.45 (0.24–0.85)		10/ 763669	0.39 (0.16–0.94)		11/ 763684	0.53 (0.22–1.28)
Per unit increase in behavior	283/ 5815211	0.75 (0.64–0.88)		120/ 5813003	0.67 (0.53–0.85)		163/ 5813366	0.82 (0.67–1.00)

Abbreviations: CI, confidence interval; HR, hazard ratio; MV, multivariable; PY, person-years; ref, reference.

¹Healthy lifestyle index score (HLIS) as a summed score from 0 (no healthy behaviors) to 3 (all healthy behaviors) or continuous score (per 1 unit of behavior change)

²Multivariable (MV)-adjusted hazard ratios (HR) from time-varying Cox proportional hazards models, adjusted for age (months), questionnaire cycle, cohort, non-white race, census-tract median household income (quartiles), ever oral contraceptive use, age at menarche <=10, menopausal status (pre-menopausal vs. post-menopausal never/ever used post-menopausal hormones)

Additional Sensitivity Analysis

The results remained similar to the primary analysis when we changed the classification of cigarette smoking to never smokers vs. ever smokers (Supplementary Table 4), where a higher continuous traditional HLIS continued to be associated with a lower SLE risk overall (HR 0.79 [95%CI 0.68–0.92]) and for the dsDNA positive subtype (HR 0.74 [95%CI 0.59–0.93]). After excluding participants who consumed ≥30 grams /day of alcohol (Supplementary Table 5), a higher continuous traditional HLIS remained associated with a lower SLE risk overall (HR 0.76 [95%CI 0.65–0.89]) and associated with a lower risk of dsDNA positive SLE (HR 0.70 [95%CI 0.55–0.89]), and dsDNA negative SLE (HR 0.81 [95%CI 0.66–0.99]). After adding the time lag for exercise and diet (Table 4), we had 178 SLE incident cases remaining. However, we found similar results to our primary analysis that did not include the time lag. A higher continuous traditional HLIS was still associated with a lower SLE risk overall (HR 0.78 [95%CI 0.67–0.91]) and lower dsDNA positive SLE risk (HR 0.75 [95%CI 0.60–0.93]) per unit increase. Women in the highest category of HLIS with at least four healthy behaviors had the lowest risks of developing SLE overall (HR 0.40 [95%CI 0.23–0.69]), dsDNA positive SLE (HR 0.34 [95%CI 0.15–0.74]), and dsDNA negative SLE (HR 0.47 [95%CI 0.23–0.99]) compared to women with one or no healthy behaviors.

Table 4.

Sensitivity analysis with time lag for diet and regular exercise for hazard ratios (95%CIs) for incident SLE risk among women in Nurses’ Health Study (1986–2016) and Nurses’ Health Study II (1991–2017) in relation to the traditional Healthy Lifestyle Index Score¹ defined as drinking alcohol in moderation (>=5g/day), maintaining a healthy body weight (body mass index <25 kg/m2), never or past smoking (quit more than 4 years ago), healthy diet (highest 40th percentile of the Alternative Healthy Eating Index (AHEI)), and regular exercise (at least 19 metabolic equivalents (MET) hours per week

	Overall SLE (n=178)			dsDNA Positive SLE (n=89)			dsDNA Negative SLE (n=89)
Number of Healthy Behaviors	Cases/ person-years	MV HR (95%CI)2	p-trend	Cases/ person-years	MV HR (95%CI)2	p-trend	Cases/ person-years	MV HR (95%CI)2	p-trend
0–1	56/ 988626	1.00 (ref)	0.002	28/ 988374	1.00 (ref)	0.008	28/ 988345	1.00 (ref)	0.068
2	57/ 1365758	0.72 (0.50–1.04)		30/ 1365452	0.77 (0.46–1.29)		27/ 1365449	0.68 (0.40–1.16)
3	47/ 1112595	0.72 (0.49–1.07)		23/ 1112347	0.70 (0.40–1.24)		24/ 1112296	0.74 (0.43–1.30)
4–5	18/ 773959	0.40 (0.23–0.69)		8/ 773830	0.34 (0.15–0.74)		10/ 773838	0.47 (0.23–0.99)
Per unit increase in behavior	178/ 4240937	0.78 (0.67–0.91)		89/ 4240003	0.75 (0.60–0.93)		89/ 4239927	0.82 (0.66–1.01)

Abbreviations: CI, confidence interval; HR, hazard ratio; MV, multivariable; PY, person-years; ref, reference.

¹Healthy lifestyle index score (HLIS) as a summed score from 0–1 (one or no healthy behaviors) to 4–5 (4 or more healthy behaviors) or continuous score (per 1 unit of behavior change)

²Multivariable (MV)-adjusted hazard ratios (HR) from time-varying Cox proportional hazards models, adjusted for age (months), questionnaire cycle, cohort, non-white race, census-tract median household income (quartiles), ever oral contraceptive use, age at menarche <=10, menopausal status (pre-menopausal vs. post-menopausal never/ever used post-menopausal hormones)

Discussion

To our knowledge, this is the first study to prospectively evaluate the association between an overall healthy lifestyle and SLE risk. In order to evaluate the temporal relationship between a healthy lifestyle and SLE risk, we required a dataset that was sufficiently large in scope, collected accurate longitudinal data with minimal loss to follow-up in a well-characterized prospective cohort. We believe the NHS and NHSII cohorts are among the only data sources in the U.S. that meet these criteria, notwithstanding limitations regarding the demographic makeup of the cohort recruitment. Therefore, we applied the well-established HLIS to these two large, nationwide cohorts of female nurses followed for up to 40 years with detailed and updated covariable data from biennial questionnaires and over 5.8 million person-years of follow-up. We found that adherence to multiple healthy behaviors was associated with a lower risk of SLE development overall (19% reduction for each additional healthy behavior) and risk of dsDNA positive SLE (22% reduction for each additional healthy behavior). Those with the highest number of healthy behaviors had the lowest risk of SLE overall and anti-dsDNA positive SLE compared to those with the least healthy lifestyle. Strikingly, we found that the risk of SLE in participants who are likely already at risk due to inherited genetic risk factors, potentially could be reduced by nearly half with adherence to healthy and modifiable lifestyle behaviors.

Prior studies, including those outside of NHS and NHSII, have evaluated the impact of lifestyle behaviors on the risk of SLE development but considering each factor individually, finding variable associations. Cigarette smoking (8, 9, 18) and obesity (14) have been associated with increased SLE risk, while an inverse association was found with alcohol consumption (9–12) and no association was identified in relation to various dietary patterns (16, 17). In prior NHS and NHSII studies, current smokers had a higher risk of anti-dsDNA positive SLE than never smokers (HR 1.86 [95%CI 1.14–13.04]) (8), obese women had an 85% higher risk of SLE compared to women with a normal BMI (HR 1.85, 95%CI 1.17–2.91) (14), and moderate alcohol intake (≥5 grams/day or >0.5 drinks/day) was associated with a decreased risk of incident SLE (HR 0.61 [95%CI 0.41–0.89]) (10). However, no association was found between SLE risk and various diets including the AHEI (16, 17). On the other hand, the role of physical activity in risk for developing SLE has not been elucidated. The impact of exercise on the immune system including higher levels of T-regulatory cells and a shift in the Th1/Th2 balance has been shown to be protective against other autoimmune diseases such as RA (20) and suggests that it may also influence SLE risk. Despite varying strengths of association with SLE risk, our study showed that in combination, these modifiable risk factors significantly reduce SLE risk.

The approach of studying multiple lifestyle factors together rather than individually provides a more pragmatic and holistic understanding of how lifestyle factors can affect risk of disease events. Recognizing whether individuals are meeting recommended guidelines across multiple lifestyle factors, the HLIS tool has also been applied to examine its impact on other chronic diseases (23, 24, 26–28). For instance, women who adhered to all five healthy lifestyle factors had an 80% lower incidence of coronary events (23) and 91% lower incidence of type 2 diabetes (26). Future studies using the HLIS tool should examine how multiple healthy lifestyle behaviors can reduce the risk of other autoimmune diseases such as psoriatic arthritis. It is also important to recognize that although the evidence thus far points to encouraging patients to adhere to as many healthy behaviors as possible for the greatest benefit, there are many institutional and structural factors that contribute to one’s ability to adhere or achieve a healthy lifestyle. The many social determinants of health, including effects of poverty, pollution, toxins, stress, and institutional and structural racism, have disproportionate impacts on non-White groups in the US, the same groups that have the highest incidence and severity of SLE. These factors likely negatively affect ability to achieve a healthy lifestyle and were not examined in the current study, but deserve to be in future research. Therefore, even though these results suggest that SLE may be, to some degree, preventable, the barriers to achieving a healthy lifestyle and thus modifying one’s SLE risk are unlikely to be equally distributed in the population. Future studies should examine how to improve adherence to lifestyle interventions and also addressing the factors that prevent or limit a person’s ability to be able to meet these healthy goals among people at risk of SLE from socio-demographically and diverse backgrounds.

Lifestyle behaviors acting in combination to influence the risk of SLE and potentially produce stronger effects together than individually provides some insights into the pathogenesis of SLE. It is plausible that environmental exposures may work synergistically via common biological pathways to influence risk of SLE, including induction of oxidative stress, damage to endogenous proteins and DNA, autoantibody production, and upregulation of pro-inflammatory cytokines to induce epigenetic changes, resulting in altered gene expression affecting immune homeostasis. Both exposure to toxic components of cigarette smoke and obesity are known to cause oxidative stress resulting in elevated intracellular levels of reactive oxygen species that can damage DNA forming immunogenic DNA adducts, which may promote dsDNA antibody production (42, 43). This is keeping with our study where we found that a healthier lifestyle was strongly associated with a lower risk of developing anti-dsDNA positive SLE. This relationship was less clear with anti-dsDNA negative SLE. The smoke by-products themselves could also augment autoreactive B cells in the native repertoire while autoantibody production in obese animal models and human subjects has been demonstrated (44). Cigarette smoking has also been shown to induce pulmonary ANA and increase SLE-associated pro-inflammatory cytokine B-lymphocyte stimulator (BLyS) expression in the lungs of exposed mice (45). In a recent NHS and NHSII study, elevated BLyS and lower IL-10 levels (an anti-inflammatory cytokine) were found among current smokers, particularly among women who were ANA positive (46). Smoking has also been shown to increase pro-inflammatory cytokines including TNF-α and interleukin-6 (47) which play important roles in the modulation of insulin resistance. Adipose tissue, in particular visceral fat, secretes pro-inflammatory adipocyte-derived cytokines. Obese individuals have higher levels of C-reactive protein, soluble TNF-α receptor 2, IL-6 than non-obese individuals (48). Alcohol consumption, on the other hand, contains several compounds (e.g., ethanol and anti-oxidants) that can potentially counteract the changes induced by smoking and obesity, including diminishing cellular responses to immunogens, suppressing synthesis of immunoglobulins and influence pro-inflammatory cytokines (TNF-α, IL-6, IL-8, interferon gamma) and inhibit key enzymes in DNA synthesis (49). In vitro and animal studies may further elucidate the biological mechanisms and pathways by which joint risk factors may play a role in the etiology of SLE.

The strengths of our study include using data from well-described cohorts with up to 5.8 million person-years of prospective follow-up. There were detailed data on potential time-varying confounders to reduce the within-subject variation, minimize inaccuracy of exposure data, and decrease the potential for reverse causation and recall biases. The strengths of the HLIS include strong face validity and ease of application at the individual level or for public health purposes.

We also recognize some important limitations. The study population was predominantly White, and all participants were female nurses; their lifestyle habits, such as dietary intake, likely differ from that of other ethnic populations. Therefore, it will be important to validate our findings in other more diverse SLE cohorts. Also, there was a relatively small number of incident SLE cases limiting the ability to examine more extreme categories. As there may be a long prodromal period in the development SLE, an extra lag period between the exposure and the outcome windows was added to address potential reverse causation. As a result, a large number of cases were not analyzed, limiting statistical power particularly in the analysis of exercise. The overall incidence rate for SLE in our study is lower than recent reported incidence rates of SLE among women in the U.S. likely as a result of our stringent case definition of SLE (50). This may have been due to exclusion of early onset SLE before age 35 in the NHS and before age 25 in the NHSII, exclusion of possible SLE cases that later became definite cases and exclusion of patients prior to administration of the diet questionnaire in the NHS (but not in the NHSII), which was administered approximately eight years after cohort inception. Also due to the higher mean age of the cohorts and inclusion of potentially less severe SLE cases, this study should also be reexamined in a cohort of younger women and those with more severe disease. Finally, there is a potential for misclassification as self-reported questionnaires were used and there may be unknown confounders not accounted for in the analysis.

In conclusion, we found an inverse association between a combination of healthy lifestyle behaviors and SLE risk, associated with nearly half of the population attributable risk. Our findings have implications for SLE prevention and the promotion of multiple lifestyle behaviors to derive the greatest benefit. We also provided further insight into the pathogenesis of SLE as a greater than expected proportion of SLE risk may be attributable to modifiable lifestyle factors.