Risk of progression of interstitial pneumonia with autoimmune features to a systemic autoimmune rheumatic disease

Abstract

Objective

The aim of this study was to determine the risk of developing a systemic autoimmune rheumatic disease (ARD) after an initial diagnosis of interstitial pneumonia with autoimmune features (IPAF).

Methods

We performed a retrospective cohort study of patients with interstitial lung disease (ILD) who were evaluated at Columbia University Irving Medical Center from 2009 to 2017. We divided patients with idiopathic ILD into two groups: those who met IPAF criteria and those who did not meet IPAF criteria at initial ILD diagnosis. We examined the association between IPAF and diagnosis of ARD during the follow-up period using a multivariable-adjusted logistic regression model.

Results

Of the 697 patients with ILD who were screened, 174 met inclusion criteria (50 met IPAF criteria and 124 did not). During a median follow-up period of 5.2 years, 16% (8/50) of subjects with IPAF were diagnosed with an ARD compared with 1.6% (2/124) of subjects without IPAF (P = 0.001). Adjusting for age, sex, smoking status and use of immunosuppressive therapy, the odds of progressing to an ARD were 14 times higher in subjects with IPAF than in those without IPAF (odds ratio 14.18, 95% CI 1.44–138.95, P = 0.02).

Conclusion

The presence of IPAF confers an increased risk of developing an ARD. Patients with IPAF should therefore be followed closely for the development of an ARD.

Rheumatology key messages

• Interstitial pneumonia with autoimmune features confers an increased risk of developing a systemic autoimmune rheumatic disease.

• Patients with interstitial pneumonia with autoimmune features should be monitored by rheumatologists for development of systemic autoimmune rheumatic diseases.

Introduction

Systemic autoimmune rheumatic diseases (ARDs) are a relatively common cause of interstitial lung disease (ILD) [1], and ILD may be the initial manifestation of an ARD [2]. In a retrospective cohort study of 50 patients presenting to an ILD clinic for evaluation, 50% were subsequently diagnosed with an ARD [3]. Therefore, it is important to identify risk factors for progression to an ARD among patients with ILD as this could allow for early detection and treatment of ARDs.

In 2015, the American Thoracic Society (ATS) and European Respiratory Society (ERS) proposed the term ‘interstitial pneumonia with autoimmune features’ (IPAF) to describe the 15–25% of ILD patients who have some autoimmune features but do not meet classification criteria for an ARD [4–6]. To meet classification criteria for IPAF, a patient must have at least one feature from at least two of the following three domains: clinical, serological and morphological [6].

Data on progression of IPAF to ARDs are limited and conflicting. Scire et al. found that 42% of individuals with anti-synthetase antibodies who initially met IPAF criteria ultimately developed myositis, RA or a myositis–RA overlap syndrome [7]. Ito et al. found that 12.2% of patients who met the serological and morphological domain of IPAF were later diagnosed with an ARD during a median follow-up period of 4.5 years [8]. However, Chartrand et al. found that none of 56 IPAF patients progressed to an ARD over a 4-year follow-up period [9]. Thus, it remains unclear whether the risk of progression to an ARD is greater in those with IPAF than in those with idiopathic ILD without autoimmune features. The primary aim of this study was to determine the risk of developing an ARD after an initial diagnosis of IPAF. Our secondary aim was to identify risk factors for the development of an ARD among patients with idiopathic ILD. We hypothesized that adults with IPAF would have a greater risk of developing an ARD than those with idiopathic ILD who did not meet classification criteria for IPAF.

Methods

Participants

We performed a retrospective cohort study of patients with ILD who were evaluated at Columbia University Irving Medical Center (CUIMC) between 1 January 2009 and 1 January 2017. Data were extracted from the electronic medical record. Patients were screened if they had an ILD ICD-9 or ICD-10 code (516.3, 516.32, 516.35, 515, J84.9, J84.111, J84.113, J84.10, J84.112, J84.116, J84.2, J84.117) documented in at least six different clinic visits. Patients were included if they were at least 18 years of age at the time of ILD diagnosis; had at least one follow-up clinic visit at least 6 months after initial pulmonary evaluation at CUIMC; and had available data on rheumatological serologies, imaging pattern on high resolution computed tomography (HRCT) scan of the chest (as per clinician and radiologist impression), pulmonary function tests and work up for known causes of ILD at initial ILD diagnosis. Patients were excluded if they had ILD due to a known cause (e.g. hypersensitivity pneumonitis, established ARD, sarcoidosis, familial pulmonary fibrosis). This study was approved by the Institutional Review Board at CUIMC (no. AAAR244).

Statistical analysis

We compared baseline characteristics between patients with and without IPAF using Kruskal–Wallis, Fisher’s exact and χ² tests, as appropriate. Our primary outcome was diagnosis of an ARD by a rheumatologist in the follow-up period. Via medical record review, we also determined which of the ARD diagnoses would meet the following sets of classification criteria: 2010 ACR/EULAR RA classification criteria [10], 2013 ACR/EULAR classification criteria for SSc [11], 2017 EULAR/ACR classification criteria for adult and juvenile idiopathic inflammatory myopathies [12], and Chapel Hill Consensus Conference criteria for systemic vasculitides [13].

To examine the association between IPAF at initial ILD diagnosis and rheumatologist diagnosis of an ARD in the follow-up period, we modelled IPAF as the independent binary variable of interest in logistic regression models where diagnosis of an ARD in the follow-up period was the dependent binary variable. We initially estimated odds ratios (ORs) for diagnosis of an ARD in the follow-up period for adults with IPAF compared with those without IPAF in a logistic regression model without adjustment for additional independent variables. We then constructed a multivariable logistic regression model adjusting for age, sex, smoking status and use of immunosuppressive therapy. We also estimated hazard ratios for ARD diagnosis for patients with vs without IPAF in a Cox proportional hazards model without adjustment for additional independent variables. We then constructed a multivariable Cox model adjusting for age, sex, smoking status and use of immunosuppressive therapy, and generated a Kaplan–Meier curve. In exploratory analyses, we examined the association of (i) the IPAF clinical domain, (ii) the IPAF serological domain, (iii) the IPAF morphological domain, and (iv) female sex with progression to an ARD [6]. We calculated the area under the receiver operating characteristic curve (AUC) to estimate which combination of covariates best predicted progression to an ARD among patients with idiopathic ILD. All analyses were performed using STATA version 13.1 (StataCorp, College Station, TX, USA).

Results

Of the 697 patients screened, 419 met inclusion criteria. Of these, 245 (58%) had ILD due to a known cause and were excluded, while 174 (42%) patients had idiopathic ILD at baseline. Of the 174 subjects with idiopathic ILD, 50 (29%) met ATS/ERS classification criteria for IPAF at initial ILD diagnosis and 124 (71%) did not (Supplementary Fig. S1, available at Rheumatology online).

Compared with subjects without IPAF, those with IPAF were younger (median age 56 vs 66 years, P < 0.001) and a greater proportion (i) were female (60% vs 36%, P < 0.001); (ii) were non-smokers (48% vs 30%, P = 0.03); (iii) had clinical features of autoimmunity (38% vs 0, P < 0.001); (iv) had positive autoantibodies (98% vs 25%, P < 0.001); (v) had a non-specific interstitial pneumonia pattern on HRCT (82% vs 15%, P < 0.001) or biopsy (73% vs 12%, P < 0.001); and (vi) had pulmonary hypertension defined by right-sided heart catheterization (14% vs 4%, P = 0.02) at initial ILD diagnosis (Tables 1–3). A smaller proportion of subjects with IPAF than without IPAF had a usual interstitial pneumonia (UIP) pattern on HRCT (18% vs 75%, P < 0.001) or biopsy (20% vs 71%, P < 0.001). The median [interquartile range (IQR)] duration of observation was similar for both groups [6.3 (4.3–8.9) years for IPAF vs 5.8 (3.0–8.5) years for non-IPAF, P = 0.74]. During the observation period, a greater proportion of subjects with IPAF than those without underwent rheumatological evaluation (78% vs 19%, P < 0.001) and received immunosuppressive therapy (96% vs 52%, P < 0.001).

Table 1.

Demographic, clinical, and treatment characteristics of patients with idiopathic ILD

Characteristic	Overall cohort	IPAF	Non-IPAF	P-value
	(n = 174)	(n = 50)	(n = 124)
Age, median (IQR), years	63 (56–72)	56 (47–64)	66 (59–73)	<0.001
Female sex, n (%)	74 (43)	30 (60)	44 (36)	0.003
Ever smoker, n (%)	111 (64)	25 (50)	86 (69)	0.03
BMI, median (IQR), kg/m2	28 (25–31)	28 (25–31)	27 (25–30)	0.38
FVC, mean (s.d.), % predicted	67 (20)	63 (20)	68 (19)	0.12
FEV1, mean (s.d.), % predicted	70 (20)	65 (20)	72 (20)	0.054
DLCO, mean (s.d.), % predicted	41 (14)	37 (11)	42 (14)	0.006
IPAF clinical domain, n (%)	19 (11)	19 (38)	0	<0.001
Mechanic’s hands, n (%)	2 (1)	2 (4)	0	0.14
Distal digital tip ulceration, n (%)	1 (0.6)	1 (2)	0	0.31
Inflammatory arthritis, n (%)	3 (1.7)	3 (6)	0	0.03
Palmar telangiectasias, n (%)	1 (0.6)	1 (2)	0	0.49
Raynaud's phenomenon, n (%)	15 (8.6)	15 (30)	0	<0.001
Unexplained digital oedema, n (%)	0	0	0	—
Unexplained fixed rash on the digital extensor surfaces, n (%)	1 (0.6)	1 (2)	0	0.31
Rheumatological evaluation during the observation period, n (%)	62 (36)	39 (78)	23 (19)	<0.001
Immunosuppression during the observation perioda, n (%)	112 (64)	48 (96)	64 (52)	<0.001
Immunosuppression within first year of ILD diagnosis, n (%)	78 (45)	41 (82)	37 (30)	0.002

^aImmunosuppressive medications used: steroids, mycophenolate mofetil, azathioprine, cyclophosphamide, calcineurin inhibitors. DLCO: diffusing capacity for carbon monoxide; FEV1: forced expiratory volume in 1 s; FVC: forced vital capacity; IPAF: interstitial pneumonia with autoimmune features; IQR, interquartile range.

Table 2.

Baseline rheumatological serologies in patients with idiopathic ILD

Rheumatological serologies	Overall cohort	IPAF	Non-IPAF	P-value
	(n = 174)	(n = 50)	(n = 124)
IPAF serological domain	80 (46)	49 (98)	31 (25)	<0.001
ANA positivea	50 (29)	31 (62)	19 (15)	<0.001
RF ≥2× upper limit of normal	15 (9)	7 (14)	8 (7)	0.11
Anti-CCP positive	7 (6) (n = 112)	4 (11) (n = 37)	3 (4) (n = 75)	0.09
Anti-dsDNA positive	5 (4)	2 (5)	3 (4)	0.63
	(n = 118)	(n = 39)	(n = 79)
Anti-Ro (SS-A) positive	20 (12)	18 (36)	2 (2)	<0.001
	(n = 167)	(n = 50)	(n = 117)
Anti-La (SSB) positive	2 (1.1)	1 (2)	1 (0.8)	0.49
	(n = 167)	(n = 50)	(n = 117)
Anti-RNP positive	9 (5)	7 (14)	2 (2)	0.003
	(n = 167)	(n = 50)	(n = 117)
Anti-Smith positive	3 (1.7)	2 (4)	1 (0.8)	0.19
	(n = 167)	(n = 50)	(n = 117)
Anti-Scl-70 positive	1 (0.6)	1 (2)	0	0.30
	(n = 167)	(n = 50)	(n = 117)
Anti-tRNA synthetase positiveb	9 (5.2)	9 (18)	0	<0.001
	(n = 167)	(n = 50)	(n = 117)
Anti-PM-Scl positive	0	0	0	—
	(n = 3)	(n = 2)	(n = 1)
Anti-MDA-5 positive	0	0	0	—
	(n = 3)	(n = 2)	(n = 1)
Other serologiesc
Anti-RNA polymerase III positive	0	0	0	—
	(n = 53)	(n = 20)	(n = 33)
Anti-centromere positive	1 (0.9)	0	1 (1.4)	0.99
	(n = 106)	(n = 36)	(n = 70)
Anti-MPO positive	4 (3.5)	3 (8)	1 (1)	0.07
	(n = 112)	(n = 37)	(n = 75)
Anti-PR3 positive	2 (2)	1 (3)	1 (1)	0.49
	(n = 112)	(n = 37)	(n = 75)

Data presented as n (%).

^aANA ≥1: 320 titre, diffuse, speckled, homogeneous patterns or ANA nucleolar pattern (any titre) or ANA centromere pattern (any titre).

^bIncludes anti-histidyl-tRNA synthetase (anti-Jo-1), anti-glycyl-tRNA synthetase (anti-EJ), anti-threonyl-tRNA synthetase (anti-PL-7) and anti-alanyl-tRNA synthetase (anti-PL-12) antibodies.

^cSerologies not included in the IPAF serological domain. dsDNA: double stranded DNA; IPAF: interstitial pneumonia with autoimmune features; MDA-5: melanoma differentiation-associated protein; PR3: proteinase 3.

Table 3.

Baseline morphological characteristics of patients with idiopathic ILD

Morphological characteristics	Overall cohort	IPAF	Non-IPAF	P-value
	(n = 174)	(n = 50)	(n = 124)
IPAF morphological domain	74 (43)	46 (92)	28 (23)	<0.001
HRCT pattern
NSIP	59 (34)	41 (82)	18 (15)	<0.001
OP	7 (4.7)	3 (6)	4 (3.2)	0.41
NSIP with OP overlap	3 (1.7)	3 (6)	0	0.02
LIP	0	0	0	–
Biopsy pattern
NSIP	37 (34)	29 (73)	8 (12)	<0.001
	(n = 109)	(n = 40)	(n = 69)
OP	13 (11.9)	7 (17.5)	6 (8.7)	0.17
	(n = 109)	(n = 40)	(n = 69)
NSIP with OP overlap	5 (4.6)	4 (10)	1 (1.4)	0.06
	(n = 109)	(n = 40)	(n = 69)
LIP	0	0	0	–
	(n = 109)	(n = 40)	(n = 69)
Histopathological feature
Interstitial lymphoid aggregates with germinal centers	3 (2.8)	3 (7.5)	0	0.02
	(n = 109)	(n = 40)	(n = 69)
Diffuse lymphoplasmacytic infiltration (with or without lymphoid follicles)	7 (6.4)	6 (15.0)	1 (1.4)	0.01
	(n = 109)	(n = 40)	(n = 69)
Unexplained pleural effusion or thickening	2 (1.1)	1 (2)	1 (0.8)	0.49
Unexplained pericardial effusion or thickening	0	0	0	–
Unexplained intrinsic airways disease	5 (2.9)	2 (4)	3 (2.4)	0.63
Unexplained pulmonary vasculopathy	5 (2.8)	4 (8)	1 (0.8)	0.02
HRCT pattern: UIP	102 (59)	9 (18)	93 (75)	<0.001
Biopsy pattern: UIP	57 (52)	8 (20)	49 (71)	<0.001
	(n = 109)	(n = 40)	(n = 69)
Pulmonary hypertensiona	12 (7)	7 (14)	5 (4)	0.02

Data presented as n (%). ^aDiagnosed by right-sided heart catheterization. HRCT: high resolution computed tomography; IPAF: interstitial pneumonia with autoimmune features; LIP: lymphocytic interstitial pneumonia; NSIP: non-specific interstitial pneumonia; OP: organizing pneumonia; UIP: usual interstitial pneumonia.

During the median follow-up period of 5.2 years, 8 of 50 (16%) subjects with IPAF were diagnosed with an ARD compared with 2 of 124 (1.6%) subjects without IPAF (P = 0.001; Table 4). A description of these patients’ initial presentation as well as clinical features that arose in the follow-up period is included in Supplementary Table S1, available at Rheumatology online. Of the eight patients with IPAF who were later diagnosed with an ARD, two were diagnosed with RA, three with systemic sclerosis (SSc), one with PM and two with ANCA-associated vasculitis. Both patients without IPAF who were later diagnosed with an ARD developed RA. Even though diagnosis of an ARD was made by the patient’s rheumatologist based upon clinical judgment—not necessarily relying on classification criteria—review of medical records reveals that both ANCA-associated vasculitis diagnoses would meet the Chapel Hill Consensus Conference definition [13], all four RA diagnoses would meet the 2010 ACR/EULAR RA classification criteria [10], the one PM diagnosis would be classified as definite myositis according to the 2017 EULAR/ACR classification criteria for adult and juvenile idiopathic inflammatory myopathies [12], and two out of the three SSc diagnoses would meet the 2013 ACR/EULAR classification criteria for SSc [11]. The one SSc patient who did not meet the 2013 ACR/EULAR classification criteria for SSc (case 9 in Supplementary Table S1, available at Rheumatology online) presented with oesophageal dysmotility and small intestinal bacterial overgrowth as well as Raynaud’s phenomenon, and had abnormal nailfold capillaries on nailfold capillaroscopy; these manifestations, combined with the presence of ILD, led to a multi-disciplinary consensus diagnosis of SSc.

Table 4.

Association between IPAF and diagnosis of ARD in the follow-up period

	IPAF (n = 50)	Non-IPAF (n = 124)	P-value
Diagnosis of ARD in the follow-up period, n (%)	8 (16)	2 (1.6)	0.001
Type of ARD	2 RA, 3 SSc, 1 PM, 2 AAV	2 RA	—
Total duration of observation, median (IQR), years	6.3 (4.3–8.9)	5.8 (3–8.5)	0.74
Time from ILD diagnosis to ARD diagnosis, median (IQR), years	3.4 (1.8–5.3)	7.8 (6.2–9.4)	0.19
Time from first rheumatological evaluation to ARD diagnosis, median (IQR), years	0.9 (0.4–4.2)	2 (1–3)	0.51
Odds ratio, median (IQR)
Unadjusted	11.61 (2.37–56.90)	1	0.002
Adjusteda	14.18 (1.44–138.95)	1	0.02

^aAdjusted for age, sex, smoking status and use of immunosuppression. AAV: ANCA-associated vasculitis; ARD: systemic autoimmune rheumatic disease; IPAF: interstitial pneumonia with autoimmune features; SSc: systemic sclerosis.

The multivariable-adjusted odds of progressing to an ARD were 14-fold higher in patients with IPAF than in those without IPAF (OR 14.18, 95% CI 1.44–138.95, P = 0.02; Table 4), adjusting for age, sex, smoking status, and use of immunosuppressive therapy. In a multivariable Cox regression analysis, the adjusted hazard of progressing to an ARD was >20-fold higher in IPAF patients than in those without IPAF (hazard ratio 20.23, 95% CI 1.90–214.70, P = 0.01; Fig. 1). In an exploratory analysis, the IPAF clinical domain was not statistically significantly associated with progression to an ARD in univariate (OR 3.96, 95% CI 0.93–16.85, P = 0.06) or multivariable analyses (OR 3.27, 95% CI 0.62–17.21, P = 0.16). The IPAF serological domain was statistically significantly associated with progression to an ARD in univariate analysis (OR 11.22, 95% CI 1.30–90.60, P = 0.02), and trended toward significance in multivariable analysis (OR 9.60, 95% CI 0.99–92.78, P = 0.051). The IPAF morphological domain was not statistically significantly associated with progression to an ARD in univariate (OR 2.60, 95% CI 0.71–9.58, P = 0.15) or multivariable analyses (OR 1.04, 95% CI 0.23–4.70, P = 0.95). The multivariable-adjusted odds of progressing to an ARD were 9-fold higher in females with idiopathic ILD than in males (OR 9.40, 95% CI 1.12–79.22, P = 0.04), adjusting for age, IPAF status, smoking status and use of immunosuppression. The best predictors of progression to an ARD were (i) the fully adjusted model (IPAF, age, sex, smoking status, use of immunosuppressive therapy; AUC 0.87, 95% CI 0.80–0.94) or (ii) a combination of female sex and the IPAF serological domain (AUC 0.83, 95% CI 0.74–0.92; Table 5). The AUCs of these models were not statistically different from one another (P = 0.40), indicating that the two models were statistically similar in the prediction of progression to an ARD.

Fig. 1.

Kaplan-Meier estimates for progression to ARD

Kaplan–Meier curves for the risk of progression to an ARD among patients with idiopathic ILD (IPAF and non-IPAF). Risk estimates are adjusted for sex, age, smoking status and use of immunosuppression. ARD: systemic autoimmune rheumatic disease; HR: hazard ratio; IPAF: interstitial pneumonia with autoimmune features.

Table 5.

Performance characteristics for baseline variables to predict progression to ARD

Variables	AUC (95% CI)	P-valuea	P-valueb
IPAF	0.77 (0.64–0.90)	—	0.07
IPAF, sex, age, smoker, use of immunosuppression (fully adjusted model)	0.87 (0.80–0.94)	0.07	—
Clinical domain	0.60 (0.45–0.75)	0.04	0.002
Serological domain	0.73 (0.62–0.83)	0.39	0.02
Morphological domain	0.62 (0.45–0.75)	0.01	0.001
Sex	0.75 (0.65–0.86)	0.83	0.03
Sex and serological domain	0.83 (0.74–0.92)	0.42	0.40
Clinical and serological domain	0.75 (0.62–0.88)	0.28	0.10
Serological and morphological domain	0.75 (0.61–0.89)	0.78	0.13

^a P-values for the comparison of AUCs of IPAF to a given variable or set of variables.

^b P-values for the comparison of AUCs of the fully adjusted model to a given variable or set of variables. ARD: systemic autoimmune rheumatic disease; AUC: area under the curve; IPAF: interstitial pneumonia with autoimmune features.

Discussion

We found that patients with idiopathic ILD who met classification criteria for IPAF had 14-fold higher odds of progression to an ARD compared with patients with idiopathic ILD who did not meet IPAF criteria, even when controlling for important potential confounders such as age, sex, smoking status and use of immunosuppressive therapy. In an exploratory analysis, we also found that among patients with idiopathic ILD, female sex was associated with progression to an ARD, and the IPAF serological domain trended towards significance. In fact, the combination of female sex and the IPAF serological domain predicted progression to an ARD just as well as the fully adjusted model. However, other autoimmune features, such as the IPAF clinical or morphological domain, were not associated with progression to an ARD. The proportion of patients progressing to ARD among patients who met the serological and morphological domain of the IPAF criteria in the study of Ito et al. was similar to in our study [8]. However, the study of Ito et al. did not include a control group for comparison. Chartrand et al. reported that evolution to ARD may have been halted in their cohort because all of the patients were on immunosuppression [9]. The results of our study suggest that IPAF criteria could be used to identify ILD patients at high risk of developing an ARD. Previous studies have shown the benefit of early immunomodulatory treatment in preventing or delaying progression from undifferentiated arthritis to RA [14]. Therefore, an important question is whether immunomodulatory medication could similarly prevent or delay progression to an ARD among patients with IPAF.

The prevalence of the various ARDs in the IPAF group was similar to one another, which is quite different from the general population, in which RA is the most prevalent ARD. This is likely due to the fact that ILD has differing prevalences among different ARDs. For example, up to 90% of SSc patients exhibit some evidence of ILD on HRCT and 40–75% of them have clinically significant ILD [15, 16]. In contrast, a much lower percentage of RA patients (33%) have radiographic ILD and only 10% have clinically significant ILD [17, 18]. Castelino et al. also observed a similar incidence of various ARDs in their cohort of 50 ILD patients [3]. Interestingly, two of the ten patients who progressed to an ARD in our study developed ANCA-associated vasculitis, but ANCAs are not included in the serological domain of the IPAF criteria. If the IPAF criteria ever undergo a revision, we suggest including ANCAs in the serological domain.

ARDs evolve from an asymptomatic or minimally symptomatic phase, during which specific autoantibodies are present in the serum, to a clinically overt phase that manifests with symptoms and signs that allow for diagnosis to be made. For example, autoantibodies have been shown to be present many years prior to the onset of clinical manifestations of both SLE and RA [19, 20]. During this preclinical phase, autoantibodies are detected initially against a single autoantigen but over time against additional autoantigens (i.e. ‘epitope spread’). Clinical symptoms develop anywhere from 1 to 10 years after the appearance of the first autoantibody [19–21]. In patients with Raynaud’s phenomenon, the presence of SSc-specific autoantibodies has been associated with a significantly higher risk of progression to SSc [22]. The presence of both SSc-specific autoantibodies and abnormal findings on nailfold capillaroscopy was associated with an even greater risk of progression to SSc. Our findings are consistent with the above studies and suggest that the presence of a combination of autoimmune features in ILD patients confers a significantly higher risk of progression to an ARD.

IPAF shares common clinical features with the systemic autoimmune rheumatic disease associated-interstitial lung diseases (ARD-ILDs), which are complex, heterogeneous diseases. Similar to traditional ARD-ILD, patients with IPAF may present with inflammatory forms of ILD (e.g. organizing pneumonia, non-specific interstitial pneumonia) or more fibrotic disease (e.g. UIP). These radiographic and histopathological patterns can be associated with differing treatment responses and prognosis. Oldham et al. showed that IPAF patients with UIP had a similar prognosis to that of patients with idiopathic pulmonary fibrosis, which was poorer than that of IPAF patients without UIP [23]. While clinical features may differ among patients with different forms of ARDs or IPAF, treatment of ILD in both settings is tailored to a patient’s underlying radiographic and histopathologic ILD subtype. Both rheumatologists and pulmonologists should be involved in the care of patients with ARD-ILDs and IPAF. Even though clinicians from these specialties approach these diseases from different perspectives, we strive to achieve consensus on diagnosis and treatment by sharing our expertise during multi-disciplinary discussions. Such multi-disciplinary consensus is critical to the proper classification of ARD-ILDs and IPAF.

There were some limitations of our study. First, it was retrospective. However, as part of our inclusion criteria, all subjects had ANA and RF testing and their HRCT pattern available. Ninety-seven percent of the subjects had results of anti-SSA, anti-SSB, anti-RNP, anti-Smith, anti-Scl-70 and anti-Jo-1 testing. Therefore, we were able to assess uniformly for the serological and morphological domains of the IPAF criteria. Second, it was conducted at a single tertiary referral centre for ILD, which may limit generalizability. Third, a small number of ARD cases were observed in the follow-up period, which may be due to the fact that the majority of IPAF patients received immunosuppression to treat their lung disease. However, as this limitation would have biased our results toward the null hypothesis, the true association between IPAF and ARD unbiased by ILD treatment is likely even larger than the large effect we detected. Fourth, a greater proportion of patients with IPAF underwent rheumatological evaluation than those without IPAF. However, the median observation time was similar between both groups, permitting equal opportunity for referral for rheumatological evaluation if an ARD was suspected. In addition, the higher frequency of rheumatological evaluation among IPAF patients may be due to the fact that these patients developed an ARD at a higher frequency than the controls.

Our study has several strengths. Patients who are evaluated at the CUIMC ILD Program undergo a comprehensive evaluation, which includes uniform testing of a broad range of autoantibodies. Thus, we were able to collect detailed information on the clinical phenotypes of these patients, including rheumatological serologies, HRCT imaging/lung biopsy patterns and pulmonary function tests. Because we screened charts of patients with an ILD ICD-9 or ICD-10 code documented in at least six different clinic visits, we were able to increase the specificity for ILD diagnosis and establish a cohort of idiopathic ILD patients with long-term follow-up at our institution.

In summary, the presence of IPAF confers an increased risk of developing an ARD. Patients with IPAF should therefore be followed by rheumatologists for the development of an ARD. In particular, women with idiopathic ILD and a positive autoantibody may be at increased risk for the development of an ARD, and should be monitored closely. Future work should aim to study the natural history of IPAF prospectively, as IPAF may represent a ‘window of opportunity’ for early diagnosis and management, and perhaps even prevention of ARDs.

Supplementary data

Supplementary data are available at Rheumatology online.