Background/Purpose:  Pulmonary complications contribute substantially to systemic sclerosis (SSc) associated morbidity and are the most frequent disease-related cause of death. We explore predictors of clinically significant pulmonary fibrosis (csPF) and pulmonary hypertension (PH) in a large single-centre cohort of unselected SSc patients.

Methods: The study cohort consisted of consecutive patients with disease onset between 1995 and 1999. Multivariable Cox regression was used for model building. Continuous variables were categorised and risk points, based on the β values associated with the predictor variables were calculated. Internal validation of the model was done in an independent second cohort of consecutive SSc subjects with disease onset between 2000 and 2003.

Results: A total of 398 SSc patients, 146 with diffuse cutaneous (dc)SSc and 252 with limited cutaneous (lc)SSc, formed the predictor model derivation cohort, while 279 (130 dcSSc, 149 lcSSc) were used for model validation. Frequency of csPF was very similar in the derivation cohort (22% of lcSSc,42% of dcSSc) and the validation cohort (21% of lcSSc and 40% of dcSSc).  Cumulative frequency of PH was slightly lower in the validation cohort (9% in lcSSc and 10% in dcSSc) compared to the derivation cohort (17% of lcSSc and 12%of dcSSc), potentially reflecting a shorter follow-up period.

In a multivariable analysis, significant positive predictors of csPFwere dcSSc subset (HR 1.77, p=0.027), age at onset (HR 1.02, p=0.031) and anti-topoisomerase I antibody (ATA), which was time-dependent, therefore we used its interaction term with disease duration in years (HR 1.16, p=0.001). Significant negative predictors were forced vital capacity (FVC) (HR 0.97, p=0.003), carbon monoxide diffusing capacity (DLCO) (HR 0.97, p=0.006), and anti-centromere antibodies (ACA)(HR 0.19, p=0.003). The final model using categorised variables and risk points is shown in Table1. The AUC for the model using risk points was 0.81. Risk score was calculated for all patients from the validation cohort and the AUC was 0.82.

For the PH prediction model the variables that demonstrated significant association in a multivariable analysis were age (HR 1.03, p=0.026), DLCO (HR 0.94, p<0.001), creatinine (HR 1.004, p=0.026), ATA (HR 0.41, p=0.04), anti-RNA polymerase antibody (HR 3.23, p=0.023), anti-U3RNP antibody (HR 3.92, p=0.017) and presence of renal crisis (SRC) within the first 5 years of disease (HR 0.004, p=0.009). In addition, there was one significant interaction in the model between SRC and DLCO (HR 1.08, p=0.019). Details of the model using categorical variables and risk points are presented in Table 1. The model using risk scores had AUC 0.79 which was replicated in the validation cohort.

Conclusion: We present predictive models that could be used as clinical tools in unselected SSc cases for patient risk stratification and could facilitate cohort enrichment for event driven studies.

Table 1. Multivariable analysis with categorical variables and risk score points for survival,  pulmonary fibrosis and  pulmonary hypertension
	Variables	β	HR	95.0% CI for HR		p-value	Points
Pulmonary fibrosis	DcSSc	0.526	1.692	1.052	2.721	0.03	1
	Age>55 years	0.306	1.358	0.866	2.127	0.182	1
	FVC 65-80%	0.542	1.719	1.005	2.939	0.048	1
	FVC<65%	1.157	3.18	1.765	5.727	<0.001	2
	DLCO≤55%	1.108	3.027	1.752	5.231	<0.001	2
	ACA	-1.728	0.178	0.061	0.519	0.002	-3
	ATAxT (5 years)	0.707	2.028	1.291	3.186	0.002	1
Pulmonary hypertension	Age>55years	0.565	1.759	0.991	3.124	0.054	1
	DLCO 55-65%	1.247	3.481	1.161	10.442	0.026	1
	DLCO<55%	2.492	12.089	4.591	31.83	<0.001	2
	Creatinine>85μmol/L	0.43	1.538	0.805	2.939	0.192	0
	ATA	-0.926	0.396	0.172	0.915	0.03	-1
	ARA	0.681	1.975	0.839	4.649	0.119	1
	AFA	1.155	3.175	1.192	8.454	0.021	1
	SRC5y	1.541	4.668	0.847	25.715	0.077	2
	SRC5y*DLCO 55-65%	-1.526	0.217	0.015	3.097	0.26	-2
	SRC5y*DLCO<55%	-2.754	0.064	0.006	0.674	0.022	-3

---

« Back to 2013 ACR/ARHP Annual Meeting

ACR Meeting Abstracts - https://acrabstracts.org/abstract/prediction-of-pulmonary-complications-in-systemic-sclerosis-model-development-and-validation/