Impact of type 2 diabetes mellitus on interstitial lung disease risk in rheumatoid arthritis

Abstract

BACKGROUND

This study investigates the impact of type 2 diabetes mellitus (T2DM) on the risk of interstitial lung disease (ILD) and its subtypes in patients with rheumatoid arthritis (RA). RA is often complicated by ILD. T2DM has systemic proinflammatory effects, but its impact on RA-related ILD is unclear. This research aims to elucidate the interplay between these conditions to inform clinical management and patient care strategies.

AIM

To determine if RA patients with T2DM have a higher occurrence of ILD compared to RA patients without T2DM.

METHODS

We conducted a retrospective cohort study using the 2019-2020 National Inpatient Sample. Adult RA patients with and without T2DM were identified via International Classification of Diseases, 10^th Revision (ICD-10) codes. Propensity score matching (1:1) balanced 15+ confounders. Logistic regression assessed the association of T2DM with ILD (overall and by subtype) and secondary outcomes (acute respiratory distress syndrome, pneumothorax, pleural effusion, pulmonary hypertension). Missing data were excluded. ILD subtypes were included based on ICD-10 codes and case count.

RESULTS

Among 199380 RA inpatients, ILD was more common in those with T2DM (2.25%) vs without (1.11%). After matching (n = 121046), ILD remained higher in RA + T2DM [odds ratio (OR) = 2.02, 95%CI: 1.84-2.22, P < 0.001], with an absolute risk increase of about 1.14%. T2DM was associated with higher odds of ILD subtypes including usual interstitial pneumonia (OR = 3.20) and non-specific interstitial pneumonia (OR = 3.50). Other subtypes showed elevated ORs; eosinophilic pneumonia showed an inverse association (OR = 0.23). PAH and pneumothorax were also more common in RA + T2DM (OR = 1.40 and 1.85, respectively). Acute respiratory distress syndrome and pleural effusion rates did not differ by T2DM status. Rare subtype findings should be interpreted cautiously.

CONCLUSION

T2DM increases ILD risk in RA and is linked to higher rates of pulmonary hypertension and pneumothorax, suggesting a role in exacerbating RA-related lung complications.

Key Words: Rheumatoid arthritis; Diabetes mellitus; Interstitial lung disease; Pulmonary hypertension; Pulmonary fibrosis

Core Tip: In this large inpatient study, we found that rheumatoid arthritis (RA) patients with type 2 diabetes have about twice the odds of developing interstitial lung disease compared to those without diabetes. This novel association remained significant after adjusting for numerous confounders. Diabetic RA patients also showed more pulmonary hypertension and pneumothorax. These results highlight the “diabetic lung” as a potential new consideration in RA care. Clinicians should be aware that co-morbid diabetes might heighten pulmonary risks in RA, although prospective studies are needed to confirm if controlling diabetes can improve lung outcomes.

INTRODUCTION

Rheumatoid arthritis (RA) is a common rheumatic disease, affecting around 0.5%-1% of the adult population worldwide[1]. It is an autoimmune and inflammatory disease that causes swelling, pain, and eventual damage to the joints. With the increasing incidence of RA, physicians are now seeing increasing numbers of extra-articular complications.

Pulmonary complications, including RA-related interstitial lung disease (ILD), contribute significantly to mortality and morbidity, with a lifetime risk of around 10%[2]. RA's most common ILD pattern is the usual interstitial pneumonia (UIP) pattern on high-resolution computed tomography (CT)[3]. However, non-specific interstitial pneumonia (NSIP) also occurs and may overlap with the UIP pattern, rendering the differentiation between both entities hard, especially in the setting of idiopathic pulmonary fibrosis (IPF)[4]. Subclinical or asymptomatic ILD is being detected more with increased use of CT scanning. Up to 30% of early RA patients may show signs of ILD[5]. Known risk factors for ILD include male gender, older age, smoking, elevated lactate dehydrogenase, high rheumatoid factor levels, and genetic factors[6,7]. However, endocrine and metabolic disorders haven’t been investigated as a risk factor for ILD in the RA population.

Diabetes mellitus (DM), a prevalent metabolic condition, has been increasingly recognized for its systemic effects beyond glucose metabolism[8]. Research indicates that the risk of type 2 DM (T2DM) is increased in RA by approximately 50%-70% after accounting for risk factors such as obesity[9]. Proinflammatory cytokines secreted in RA, including tumor necrosis factor-alpha and interleukin-6, induce insulin resistance by inhibiting insulin signaling pathways and impaired glucose metabolism[10-12]. Additionally, glucocorticoids, which are frequently prescribed for RA management, can worsen insulin sensitivity and increase the risk of diabetes, particularly with long-term use[13].

The relationship between DM and pulmonary outcomes is complex and varies across studies. A study of IPF patients found higher rates of DM, with 17.8% of IPF patients having concurrent DM. IPF patients with DM exhibited more severe radiological and clinical manifestations[14]. In contrast, a separate large prospective study of acute respiratory distress syndrome (ARDS) found a negative correlation between DM and lung disease. In this study, DM was associated with a 24% lower risk of developing ARDS compared to nondiabetics, even after adjusting for potential confounding factors such as medications[15]. Considering this conflicting evidence regarding the general population, our study, the first of its kind, aims to clarify the impact of DM on pulmonary outcomes by focusing on the unique RA cohort. This approach is taken due to RA's distinct clinical profile, marked by chronic inflammation and the common use of glucocorticoids. These factors potentially affect both DM and lung health in ways not seen in the general population.

The primary objective of this study was to compare the prevalence of ILD in RA patients with vs. without T2DM. Secondary objectives included evaluating differences in ILD subtypes (UIP, NSIP, etc.) and other pulmonary complications (ARDS), pneumothorax, pleural effusion, and pulmonary arterial hypertension (PAH) between the groups. By leveraging the United States National Inpatient Sample (NIS)-a large hospitalization database-and applying robust matching methods, we aimed to isolate the association of T2DM with ILD in RA. We also discuss potential mechanisms (e.g., chronic hyperglycemia-induced lung changes) and highlight that our findings are associative and hypothesis-generating, not proof of causation.

MATERIALS AND METHODS

We performed a retrospective cohort study to explore the association of T2DM with ILD and its subclasses, as well as secondary pulmonary outcomes such as ARDS, pneumothorax, pleural effusion, and PAH in patients with RA. Hospitalization outcomes, including length of stay (LOS), were also examined. The study utilized the NIS database for 2019 and 2020, which is recognized as the largest publicly available all-payer inpatient healthcare database in the United States. This study was exempt from institutional review board approval because the NIS contains de-identified patient data.

We identified adult inpatients (≥ 18 years) with RA by querying records for International Classification of Diseases, 10^th Revision (ICD-10) diagnosis codes M05.00-M06.9 (RA categories) as either the principal or a secondary diagnosis during the admission. Patients were included if RA was present in any diagnosis field, not only primary, to capture all RA patients hospitalized for any reason. We excluded patients with type 1 diabetes (E10.x codes) to focus on T2DM and avoid conflating different diabetes types. We also excluded those with other connective tissue diseases, history of certain ILD-causing exposures (e.g., radiation therapy), or conditions strongly associated with ILD (e.g., sarcoidosis, listed in Supplementary Table as exclusion codes). These exclusions were intended to reduce confounding from other ILD etiologies (Supplementary Table 1).

The study population was then stratified into two cohorts: RA with T2DM (RA+DM) and RA without T2DM (RA-DM). T2DM status was determined by presence of ICD-10 codes E11.0-E11.9 in any diagnosis field. We did not have data on duration of diabetes or glycemic control in NIS. Baseline characteristics and comorbidities were compared between RA+DM and RA-DM groups before matching.

To control for confounding, we considered a broad range of covariates known or suspected to influence ILD or to differ by DM status. These included age, sex, race, median household income by ZIP code, insurance type, hospital characteristics, and the following comorbidities: Hypertension, smoking history, obesity, chronic obstructive pulmonary disease (COPD), asthma, heart failure, ischemic heart disease, atrial fibrillation, chronic kidney disease (CKD), GERD (gastroesophageal reflux), alcohol abuse, and others (full list in Table 1). We also included steroid use (ICD-10 code Z79.52 for long-term glucocorticoid therapy) and immunosuppressant use (Z79.5-) as binary covariates, acknowledging that we lacked dosage/duration details. These variables were chosen a priori based on clinical relevance to RA or ILD risk.

Table 1 Baseline characteristics before matching, n (%)/mean ± SD.

Baseline characteristics	RA+DM (n = 60523)	RA-DM (n = 138848)	P value
Age	69.06 (11.56)	68.25 (14.41)	< 0.001
Gender
Female	42031 (69.45)	102277 (73.66)	< 0.001
Race
White	40840 (67.48)	108904 (78.43)	< 0.001
Black	10059 (16.62)	14765 (10.63)	< 0.001
Hispanic	6368 (10.52)	9481 (6.83)	< 0.001
Asian	1375 (2.27)	2718 (1.96)	< 0.001
Native	1196 (1.98)	1999 (1.44)	< 0.001
Other	685 (1.13)	990 (0.71)	< 0.001
Comorbidities
Gastro-esophageal reflux disease	21363 (35.3)	46663 (33.61)	< 0.001
Chronic obstructive pulmonary disease	7849 (12.97)	16699 (12.03)	< 0.001
Asthma	6271 (10.36)	12672 (9.13)	< 0.001
Smoking	5780 (9.55)	15831 (11.4)	< 0.001
Heart failure	29061 (48.02)	38053 (27.4)	< 0.001
Chronic kidney disease	20373 (33.66)	23650 (17.03)	< 0.001
Atrial fibrillation	9868 (16.3)	19183 (13.81)	< 0.001
Ventricular tachycardia	1065 (1.76)	1996 (1.44)	< 0.001
Hypertension	53568 (88.51)	99210 (71.45)	< 0.001
Obesity	19604 (32.39)	25059 (18.05)	< 0.001
Alcohol use	1132 (1.87)	4936 (3.55)	< 0.001
Steroids use	5767 (9.53)	15274 (11.0)	< 0.001
Coronary artery disease	21901 (36.19)	31360 (22.58)	< 0.001

RA: Rheumatoid arthritis; DM: Diabetes mellitus.

Initial assessment of confounders revealed a significant imbalance, prompting the use of propensity score matching to create comparable cohorts (Table 1). Logistic regression was employed to calculate propensity scores, followed by 1:1 matching using the nearest neighbors' ball tree algorithm without replacement. Following matching, multivariate logistic regression was conducted to evaluate the association of T2DM with primary and secondary outcomes while adjusting for residual confounding factors. Statistical significance was defined as a P value < 0.05.

An independent t-test was used to compare the characteristics of the two groups for continuous parametric variables. The Mann-Whitney U test was employed for continuous nonparametric variables. To assess the differences in categorical variables, the χ² and Fisher exact tests were utilized. All statistical analyses were performed using R version 4.1.2 (R Foundation for Statistical Computing, Vienna, Austria). Importantly, any missing data on key variables (< 5% for race, < 1% for others) led to those records being dropped (complete-case approach), which still left > 99% of cases.

RESULTS

From the 13798325 patients sampled in the NIS 2019 and 2020 cohorts, we identified 214453 patients with RA. After applying exclusion criteria, our final pre-match cohort included 199380 patients.

The RA+DM group had a slightly higher mean age (69.06 years with a standard deviation of 11.56) than the RA-DM group (68.25 years with a standard deviation of 14.41). However, this difference was not statistically significant. Females were more prevalent in both groups, with significantly more females in the RA-DM group compared to the RA+DM group. Both samples consisted of predominantly white subjects, with significantly more white subjects in the RA-DM group than in the RA+DM group (Table 1). The RA+DM group had a significantly higher prevalence of most comorbidities compared to the RA-DM group, including hypertension (HTN), CKD, COPD, and asthma (Table 1).

Using propensity score matching to account for baseline confounders, our analysis resulted in a matched sample of 121046 patients. The success of the matching process was confirmed by the standard mean differences for all confounders being less than 0.01. Additionally, the P-values associated with these confounders were non-significant, further validating the efficacy of the matching (Table 2).

Table 2 Baseline characteristics after matching, n (%)/mean ± SD.

Baseline characteristics	RA+DM (n = 60523)	RA-DM (n = 60523)	P value
Age	69.06 (11.56)	69.1 (11.60)	0.986
Gender
Female	42031 (69.45)	41985 (69.37)	0.774
Race
White	40840 (67.48)	40910 (67.59)	0.667
Black	10059 (16.62)	9997 (16.51)	0.632
Hispanic	6368 (10.52)	6391 (10.55)	0.830
Asian	1375 (2.27)	1384 (2.28)	0.862
Native	1196 (1.98)	1149 (1.89)	0.327
Other	685 (1.13)	692 (1.14)	0.850
Comorbidities
Gastro-esophageal reflux disease	21363 (35.30)	21363 (35.30)	1.000
Chronic obstructive pulmonary disease	7849 (12.97)	7833 (12.94)	0.898
Asthma	6271 (10.36)	6268 (10.36)	0.985
Smoking	5780 (9.55)	5761 (9.52)	0.860
Heart failure	29061 (48.02)	29049 (48.00)	0.950
Chronic kidney disease	20373 (33.66)	20348 (33.62)	0.884
Atrial fibrillation	9868 (16.30)	9863 (16.30)	0.975
Ventricular tachycardia	1065 (1.76)	1059 (1.75)	0.913
Hypertension	53568 (88.51)	53525 (88.44)	0.705
Obesity	19604 (32.39)	19580 (32.35)	0.888
Alcohol use	1132 (1.87)	1129 (1.87)	0.966
Steroids use	5767 (9.53)	5760 (9.52)	0.953
Coronary artery disease	21901 (36.19)	21890 (36.17)	0.952

RA: Rheumatoid arthritis; DM: Diabetes mellitus.

The least common ILD variants, including desquamative interstitial pneumonia (n = 3), lymphoid interstitial pneumonia (n = 4), alveolar proteinosis (n = 5), pulmonary langerhans cell histiocytosis (n = 6), and lymphangioleiomyomatosis (n = 8), were not included in the post-match analysis due to their low prevalence.

In the matched sample, the prevalence of ILD in the RA+DM group was significantly higher, with 1360 patients (2.25%) compared to 672 patients (1.11%) in the RA-DM group, yielding an odds ratio (OR) of 2.023 (95%CI: 1.843-2.219, P < 0.001). Subgroup analysis revealed that pulmonary fibrosis, including UIP, had an OR of 3.200 (95%CI: 2.800-3.700, P < 0.001), indicating a strong association with T2DM. Similarly, NSIP was significantly associated with T2DM, with an OR of 3.500 (95%CI: 2.900-4.100, P < 0.001). These ORs are high, but it should be noted that baseline prevalence of UIP/NSIP in RA-DM was low (about 0.3% or less).

Other ILD subtypes, such as respiratory bronchiolitis ILD and Cryptogenic Organizing Pneumonia (COP), also displayed significant associations with T2DM, with ORs of 1.636 (95%CI: 0.531-5.036, P = 0.002) and 96.040 (95%CI: 0.860-1072, P = 0.06), respectively (Figure 1). However, the association for COP did not reach statistical significance due to the small sample size. Additionally, acute interstitial pneumonitis (AIP) had an OR of 2.201 (95%CI: 0.764-6.335, P < 0.001). Conversely, eosinophilic lung disease showed an inverse association with DM, with an OR of 0.232 (95%CI: 0.112-0.415, P < 0.001) (Tables 3 and 4).

Figure 1 Forest plot representing the odds ratio and confidence interval of primary outcomes. NSIP: Non-specific interstitial pneumonia; UIP: Usual interstitial pneumonia; ILD: Interstitial lung disease.

Table 3 Primary outcomes after propensity matching, n (%).

Outcomes	RA+DM (n = 50423)	RA-DM (n = 50423)	P value
Interstitial lung diseases	1360 (2.25)	672 (1.11)	< 0.001
Usual interstitial pneumonia/pulmonary fibrosis	840 (1.67)	227 (0.45)	< 0.001
Non-specific interstitial pneumonia	494 (0.98)	133 (0.26)	< 0.001
Eosinophilic lung disease	18 (0.04)	78 (0.15)	< 0.001
Respiratory bronchiolitis interstitial lung disease	8 (0.02)	5 (0.01)	< 0.001
Cryptogenic organizing pneumonia	17 (0.03)	1 (< 0.01)	0.057
Acute interstitial pneumonitis	11 (0.02)	5 (0.01)	< 0.001

RA: Rheumatoid arthritis; DM: Diabetes mellitus.

Table 4 Primary and secondary outcomes odds ratio after multivariate regression.

Outcomes	Odds ratio	P value	95%CI
Interstitial lung diseases	2.203	< 0.001	1.843-2.219
Usual interstitial pneumonia/pulmonary fibrosis	3.200	< 0.001	2.800-3.700
Non-specific interstitial pneumonia	3.500	< 0.001	2.900-4.100
Eosinophilic lung disease	0.232	< 0.001	0.112-0.415
Respiratory bronchiolitis interstitial lung disease	1.636	0.002	0.531-5.036
Cryptogenic organizing pneumonia	96.040	0.06	0.860-1072
Acute interstitial pneumonitis	2.201	< 0.001	0.764-6.335
Acute respiratory distress syndrome	1.116	0.260	0.921-1.353
Pneumothorax	1.848	< 0.001	1.465-2.332
Pleural effusion	1.028	0.49	0.959-1.101
Pulmonary arterial hypertension	1.397	< 0.001	1.331-1.467

As for the secondary outcomes, Pneumothorax was significantly more prevalent in the RA+DM group (203 patients, 0.40%) compared to the RA-DM group (110 patients, 0.22%), with an OR of 1.848 (95%CI: 1.465-2.332, P < 0.001). PAH also had a higher prevalence in the RA+DM group (4141 patients, 8.21%) compared to the RA-DM group (2964 patients, 5.88%), with an OR of 1.397 (95%CI: 1.331-1.467, P < 0.001). In contrast, ARDS and pleural effusion did not show significant differences between the groups. ARDS was observed in 221 patients (0.44%) in the RA+DM group and 198 patients (0.39%) in the RA-DM group, with an OR of 1.116 (95%CI: 0.921-1.353, P = 0.260). Pleural effusion was noted in 1694 patients (3.36%) in the RA+DM group and 1648 patients (3.27%) in the RA-DM group, with an OR of 1.028 (95%CI: 0.959-1.101, P = 0.49). (Tables 4 and 5). Lastly, the linear regression analysis also demonstrated a significantly increased LOS in the RA+DM group by an average of 0.57 days (95%CI: 0.508-0.626, P < 0.001).

Table 5 Secondary outcomes after propensity matching, n (%).

Outcomes	RA+DM (n = 50423)	RA-DM (n = 50423)	P value
Acute respiratory distress syndrome	221 (0.44)	198 (0.39)	0.260
Pneumothorax	203 (0.40)	110 (0.22)	< 0.001
Pleural effusion	1694 (3.36)	1648 (3.27)	0.49
Pulmonary arterial hypertension	4141 (8.21)	2964 (5.88)	< 0.001

RA: Rheumatoid arthritis; DM: Diabetes mellitus.

DISCUSSION

This study investigated the relationship between T2DM and pulmonary outcomes-mainly ILD and its subtypes-in patients with RA. Our analysis revealed a notable correlation: RA patients with concurrent T2DM exhibited a significantly higher likelihood of developing ILD compared to non-diabetic RA patients. In our propensity-matched sample, 2.25% of RA+DM patients had ILD vs 1.11% of RA-DM patients, translating to an OR of about 2.0. The risk increase was especially pronounced for fibrotic ILD patterns: Patients with RA and DM had roughly threefold higher odds of UIP and NSIP patterns than those without DM. These findings suggest that T2DM, a common comorbidity in RA, may play an exacerbating role in ILD development or progression in this population.

Moreover, the results indicated an association of DM with increased incidences of PAH, pneumothorax, and AIP in RA patients. The large sample size provided by the NIS database lends credibility to our findings and suggests they may be applicable to the broader RA patient population. As one of the first studies to examine this relationship with such a comprehensive data set, it contributes to a more nuanced understanding of the interplay between RA, DM, and ILD. Furthermore, the matching process allowed us to compare the RA+DM and RA-DM groups on an equal footing by accounting for more than 15 confounding factors within the baseline characteristics and comorbidities.

The demographic characteristics observed in this study, including the predominance of white female subjects in the RA population, are consistent with findings from prior studies on RA. These trends have been reported in various epidemiological studies, reinforcing the notion that RA predominantly affects individuals of certain demographic backgrounds[16]. The prevalence of ILD and its subtypes in the RA population in our study correlated with the range of 4% to 8%, documented in the literature[3,4].

Moreover, we found that the RA+DM group had a higher prevalence of comorbidities before matching, such as HTN, CKD, asthma, and COPD, which is consistent with the broader understanding of the increased burden of comorbidities in patients with DM, which aligns with previous research and highlight the complexity of these patients[17]. The higher incidence of comorbidities could potentially influence the incidence, management, and severity of ILD and other pulmonary outcomes such as pulmonary hypertension[18]. However, the impact of these comorbidities was potentially nullified after matching.

Our study results suggest that individuals who had both RA and DM are twice as likely to develop any ILD, as well as three times more likely to develop pulmonary fibrosis, whether UIP or NSIP patterns. Numerous studies have indicated a link between type 2 diabetes and the development of restrictive lung diseases, mostly IPF[19-23]. Studies also showed that the prevalence of diabetes is higher in patients with pulmonary fibrosis compared to the general population or those with other lung diseases[24,25]. Interestingly, The SAKU study, involving 1028 diabetic individuals aged 40-69, found that those with poor glycemic control (HbA1c ≥ 8.0%) had a 2.42 times higher likelihood of restrictive lung impairment compared to those with good control (HbA1c < 6.9%), a statistically significant finding (P = 0.002)[26]. However, the subtype of restrictive lung disease the patients developed was not mentioned.

Multiple intertwined mechanisms, such as microangiopathy, chronic inflammation, and non-enzymatic glycation in diabetes, can be postulated to worsen RA’s already harmful effects on the lungs, potentially explaining our results[17]. NSIP often manifests with diffuse, uniform inflammation/fibrosis. The systemic inflammation and oxidative stress caused by T2DM (due to hyperglycemia) may exacerbate lung inflammation, leading to the diffuse alveolar damage seen in NSIP[4]. Additionally, diabetes-related microangiopathy-small vessel damage akin to that in diabetic nephropathy and retinopathy-could occur in lung microvasculature. This could thicken alveolar walls and pulmonary capillaries, impair gas exchange, and promote the uniform fibrotic response typical of NSIP. There is some support for this: Studies of diabetic lung tissue show capillary basement membrane thickening and reduced capillary counts which parallel changes in NSIP pathology[27,28]. Moreover, reduced pulmonary capillary blood flow and consequent hypoxia can trigger pro-fibrotic pathways.

UIP, a histopathological pattern seen in IPF and most commonly in RA, is characterized by heterogeneous lung fibrosis with areas of normal lung interspersed with fibrotic regions and honeycombing. Several diabetes-related mechanisms align more closely with the pathophysiology of UIP/IPF. The autonomic neuropathy and impaired respiratory muscle function in diabetes reduce lung volumes and can accelerate restrictive lung changes, contributing to the patchy fibrotic changes observed in UIP. Hyperglycemia-induced stiffening of the lung parenchyma and pleura through the glycation of proteins such as collagen and elastin can lead to the structural lung changes characteristic of UIP[28].

This stiffening can also explain the significant increase in the incidence of pneumothorax, as the altered lung mechanics make the lung parenchyma more susceptible to rupture. Additionally, hyperglycemia-driven oxidative stress and inflammation further exacerbate fibroproliferative responses, promoting the focal fibrotic lesions in UIP. Studies have shown that diabetic patients, particularly those with poor glycemic control, have a higher prevalence of IPF, suggesting a link between metabolic dysregulation and the development of UIP[21-23]. Since this previous finding wasn’t explored previously in the literature, it is recommended that a correlation between the conditions be established and the underlying pathophysiology of such a connection explored.

In our analysis of secondary outcomes, we revealed an increasing incidence of PAH in the RA+DM group. PAH is associated with connective tissue disorders, influenced by factors such as autoimmunity, vascular changes, and chronic inflammation. In the context of RA, PAH may be induced through mechanisms including chronic pulmonary embolism, vasculopathy, and pulmonary fibrosis[29]. Additionally, DM has been identified not only as a factor that increases the incidence and severity of PAH in patients with chronic lung disease but also as an independent factor contributing to the development of PAH[30]. Understanding these associations provides valuable insights into the complex interplay between RA, DM, and PAH.

It must be stressed, however, that these mechanistic explanations are hypotheses. We did not directly investigate the biological pathways in this administrative data study. The mechanisms by which T2DM could exacerbate ILD are speculative and drawn from prior literature on “diabetic lung” changes[28]. Our findings do not prove causation; it is possible that RA patients who develop ILD have some underlying predisposition that also makes them more likely to develop T2DM (for instance, a shared genetic or environmental factor, or RA severity leading to steroid use causing DM). Also, chronic RA inflammation could contribute to developing T2DM (through cytokine-induced insulin resistance), highlighting a bidirectional relationship. Untangling cause and effect requires longitudinal studies.

There are several limitations to acknowledge. First, this was an observational study using hospital discharge data, and causality cannot be established. We report associations: T2DM is associated with higher ILD odds in RA, but we cannot confirm T2DM causes ILD. Unmeasured factors (e.g., RA duration, specific medications, or unrecorded pulmonary risk factors) could confound results. We did not have data on RA disease activity (like DAS28 scores) or seropositivity, which are risk factors for ILD. We also lacked granular data on RA treatments such as methotrexate or biologics. This is important because methotrexate, for example, can cause pneumonitis or affect ILD risk, and its use might differ between diabetics and non-diabetics. Without these variables, residual confounding is possible. We did include a proxy for immunosuppressant use (Z79.52, long-term steroid, and Z79.5, immunosuppressant therapy), but these are crude indicators. Similarly, we did not have metrics of glycemic control - a patient with well-controlled DM might have different outcomes than one with poor control. Prior studies suggest poor glycemic control correlates with worse lung function; we could not explore that, but it’s an important question for future research. Second, misclassification bias is a concern. We relied on ICD-10 codes to identify RA, T2DM, ILD, and subtypes. Coding errors or variability can occur. RA diagnoses in the NIS are likely accurate for established RA, but we could not verify them clinically. ILD subtyping by ICD-10 has limitations; for example, distinguishing UIP vs NSIP via codes may not always reflect biopsy or high-resolution CT (HRCT) results. We mitigated this by examining broad patterns and excluding extremely rare codes, but the validity of our ILD subtype data depends on coding quality. Rare outcomes like COP had imprecise estimates. The enormous OR for COP is not meaningful given only a handful of cases (likely coding that as a distinct entity is inconsistent). Also, some ILD cases might not have been coded if subclinical. It’s worth noting that we excluded any co-diagnoses that could indicate alternative ILD causes, to hone in on RA-associated ILD, but some misclassification might remain (e.g., some “idiopathic” pulmonary fibrosis cases in our data might not truly be RA-ILD but rather coexistent IPF). Third, the NIS captures only hospitalized patients. This inherently biases toward more severe disease manifestations. RA patients with mild ILD managed outpatient would be missed. Thus, our incidence figures for ILD in RA are not generalizable to all RA patients, but rather apply to those who were hospitalized (which often implies significant illness). The generalizability is further limited by the fact that the NIS is United States-based and our matched sample was predominantly white (67%-78% in groups) and female (about 71% overall). RA-ILD epidemiology can differ in other ethnic groups; for example, some studies suggest RA-ILD might be more prevalent in male or older populations.

Our study’s majority-female cohort reflects RA demographics but had fewer males than some RA-ILD case series, possibly because women with RA are more likely to be hospitalized for other comorbidities including DM. Thus, caution is needed in extrapolating to, say, male RA patients or non-white populations. We noted a significant proportion of Black patients in RA+DM (16.6%) vs RA-DM (10.6%) pre-match, consistent with higher DM rates among Black Americans. We matched on race, but some residual differences could still subtly influence results (e.g., if Black patients have different ILD risk). However, given ILD is often more common in male smokers (and Black RA patients in the United States may have lower smoking rates than whites), it’s hard to guess direction of any bias there.

The study period coincided partly with the coronavirus disease 2019 (COVID-19) pandemic (2020). We did not specifically exclude or adjust for COVID-related ILD or ARDS. It is possible some ILD diagnoses in 2020 were influenced by post-COVID pulmonary fibrosis. We suspect this impact is minimal on our results for a few reasons: (1) We required an RA diagnosis in the admission, which is less common in pure COVID pneumonia admissions; and (2) Our ILD codes were specific, and likely clinicians would distinguish acute COVID pneumonitis from chronic ILD. Nevertheless, we acknowledge that the pandemic could have altered hospital utilization and coding. For example, some RA patients with respiratory issues might have been coded differently or avoided hospitals in early 2020. We mention this as a limitation and an area to interpret carefully.

Despite these limitations, our study has notable strengths. The use of a very large nationwide database increases statistical power and diversity of the sample. Our propensity score matching rigorously controlled for many confounders, making the RA+DM vs RA-DM groups as comparable as possible given observed data. The primary findings remained robust in sensitivity analyses. We also examined a range of outcomes to build a consistent story (DM associated not just with ILD but also with related complications like PH and pneumothorax, which triangulates with the ILD finding). To our knowledge, this is the first report quantifying the link between T2DM and ILD in RA patients. It opens the door for further research into metabolic interventions in RA-ILD.

Clinical implications

If our findings are confirmed, they suggest that RA patients with T2DM might warrant closer monitoring for ILD. For instance, clinicians managing RA patients who also have diabetes might consider a lower threshold for pulmonary evaluation (e.g., pulmonary function tests or HRCT scans) when patients present with respiratory symptoms, even mild ones. This is speculative, but given ILD’s impact on RA outcomes (RA-ILD significantly shortens survival), an awareness that T2DM could be a risk factor might encourage more vigilant screening in that subgroup. Additionally, our results raise the question of whether better diabetes management could improve lung outcomes in RA. Optimal glycemic control might hypothetically slow fibrotic processes (since high glucose can drive AGEs and oxidative stress).

There is some evidence from non-RA studies that diabetes medications (like metformin or pioglitazone) have anti-fibrotic effects in lung models[31,32]. Future prospective studies or clinical trials could explore if treating insulin resistance (with drugs or lifestyle) in RA patients reduces ILD incidence or progression. We caution that such recommendations are premature until longitudinal data are available. Other medications, including PPAR-γ agonist rosiglitazone, may protect against asthma[33]. GLP-1 agonists could reduce COPD exacerbation and mortality, and SGLT-2 inhibitors interfere with pulmonary artery smooth muscle cells[33]. However, significant unanswered questions remain regarding the positive or negative impact of diabetes treatments on lung function. Clinical trials are needed to probe whether DM pharmacotherapies could slow RA-ILD progression, as well as whether patients would benefit from early and more aggressive intervention strategies such as the use of IPD and PAH-specific therapies.

Additionally, a more proactive approach to pulmonary screening for this patient subgroup is needed. Implementing more frequent and detailed assessments, such as high-resolution CT scans, could be crucial in early identification and potential mitigation of disease progression. Additionally, more studies are needed to determine whether a more proactive approach to screening for both ILD and PAH would benefit this subgroup.

CONCLUSION

In conclusion, the autoimmune and inflammatory nature of RA, along with the frequent use of steroids, elevate the risk of developing T2DM. The literature currently lacks a comprehensive understanding of how DM in RA patients influences the occurrence of ILD, a common complication of RA. Our retrospective cohort study, following confounder matching, revealed a twofold increase in the risk of ILD in RA patients with T2DM. Additionally, it demonstrated an elevated risk of pulmonary fibrosis, PAH, and pneumothorax. Further research is needed to unravel the mechanisms through which DM amplifies the risk of these pulmonary complications in RA patients, raising questions about whether stringent blood sugar control or specific medical interventions can prevent early pulmonary involvement.