High sensitivity and negative predictive value of the DETECT algorithm for an early diagnosis of pulmonary arterial hypertension in systemic sclerosis: application in a single center

Eighty-three patients with SSc who had undergone at least one RHC were studied retrospectively. The RHC was indicated due to suspicion of PAH after undergoing an annual complete pulmonary function test and annual echocardiography as routine monitoring of SSc. The reasons for carrying out the RHC in our population were the presence of right ventricular systolic pressure (RVSP) >36 mmHg plus a forced vital capacity/diffusing capacity for carbon monoxide (FVC/DLCO) ratio >1.6 in 35 patients, 34 patients had RVSP >36 mmHg, 8 patients had a FVC/DLCO ratio >1.6 and 6 patients suffered exclusively from progressive unexplained dyspnoea. All patients fulfilled LeRoy’s SSc diagnostic criteria [13], and 80 patients also met the American College of Rheumatology/European League Against Rheumatism (ACR/EULAR) 2013 classification criteria [14]. The duration of disease was more than 3 years from the first non-Raynaud’s phenomenon symptom in all cases. For validation of the DETECT algorithm, four patients were excluded as they presented with a FVC lower than the 40% predicted value, and another patient was excluded who did not meet the inclusion criteria due to DLCO ≥60%.

Three different scenarios were conducted for the comparison of the DETECT algorithm with the ESC/ERS 2009 guidelines: First, 35 patients diagnosed with PAH and 28 with no PH after undergoing RHC were selected for the application of the DETECT algorithm (see Fig. 1). Second, these 63 patients were used to explore the capacity of both screening tests to identify mPAP ≥21 mmHg (both borderline mPAP and PAH). Third, with neither inclusion nor exclusion criteria 52 patients diagnosed with PH and 31 with no PH were selected to investigate the detection of any sort of PH.

Data from the first available RHC of each patient were revised. PH was defined as mPAP ≥25 mmHg in RHC, and was classified in the following groups: WHO group 1 or PAH was defined if pulmonary artery wedge pressure (PAWP) was ≤15 mmHg and pulmonary vascular resistance (PVR) was >3 Wood units in RHC [15]; WHO group 2 or PH due to left heart disease, if PAWP was >15 mmHg; WHO group 3 or PH due to interstitial lung disease (ILD) if PAWP was ≤15 mmHg and predicted FVC <60% or if the extent of interstitial disease was identified as moderate-severe on high-resolution computed tomography (HRCT) [12]. Borderline mPAP (BoPAP) was considered if mPAP was between 21 and 24 mmHg, with PAWP ≤15 mmHg and if there was no ILD or the FVC was >60% [16].

Based on the extent of skin involvement, patients were divided into those with diffuse cutaneous SSc (dcSSc) if thickening of the skin was distal and proximal to the elbows or knees, limited cutaneous SSc (lcSSc) if the affected skin did not exceed this limit proximally and sine scleroderma systemic sclerosis (ssSSc) if no skin thickening was present [17]. The disease onset was described as the date of the first symptom attributable to SSc, including Raynaud’s phenomenon (RP), although the time from the first symptom excluding RP also was analysed. All data were collected at the time of RHC.

The presence of telangiectasias, past or current digital ulcers (DU) and past history of scleroderma renal crisis (SRC) was recorded. ILD was defined as radiological evidence of interstitial disease on HRCT. Cardiac involvement was defined as current or past history of pericardial effusion, left ventricular ejection fraction (LVEF) <50% and ischaemic heart disease with no cardiovascular risk factor or conduction abnormalities.

Antinuclear antibodies (ANA) were evaluated by indirect immunofluorescence (IIF) assay using HEp-cell line 2. Anticentromere antibodies (ACA) were described by IIF or commercial line blot assay to centromere proteins A and B (EUROLINE Systemic Sclerosis (Nucleoli) Profile (IgG), Euroimmun, Germany). Anti-topoisomerase I antibodies were determined by enzyme-linked immunosorbent assay.

Demographics, cardiovascular risk factors, anthropometrical data, complete pulmonary function test (PFT), 6-minute walking distance (6MWD), WHO-functional class, laboratory data including serum urate, N-terminal pro-brain natriuretic peptide (NT-proBNP) and right axis deviation in electrocardiography (ECG) were registered. Echocardiography included the determination of left ventricular ejection fraction (LVEF), right atrium (RA) area, tricuspid annular plane systolic excursion (TAPSE), tricuspid regurgitation velocity (TRV) and RVSP.

RHC details consisted of mPAP, PAWP, right atrial pressure (RAP), transpulmonary pressure gradient (TPG), cardiac output (CO), pulmonary vascular resistance (PVR), systemic vascular resistance (SVR), mixed venous oxygen saturation (SvO2) measured in pulmonary artery and arterial oxygen saturation (SaO2) measured in arterial blood gas analysis.

The application of the DETECT algorithm was conducted in two steps through the website PAH risk calculator (http://​detect-pah.​com) [18], the first step for referring the patient for echocardiography and the second for carrying out RHC [12]. The indication of RHC following ESC/ERS 2009 guidelines was considered if: (1) TRV was >3.4 m/s; or (2) TRV was >2.8 to ≤3.4 m/s, with symptoms (undue dyspnoea or fatigue, chest pain, near syncope or signs of right heart failure); or (3) TRV was ≤2.8 m/s, with symptoms and additional echocardiographic variables suggestive of PH.

Qualitative data were expressed as the mean and standard deviation (SD) after checking the data using the normal distribution test, and non-normally distributed qualitative variables were described as the median and interquartile range (IQR). To assess whether there were statistically significant differences, Student’s t test or the Mann-Whitney U test were used according to the result of the normal distribution test. Categorical variables were analysed by the chi-square test and Fisher’s exact test. The 95% confidence intervals (CIs) were calculated for sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV). A p value <0.05 was considered significant. Statistical analysis was conducted using SPSS 20.0 for Windows (SPSS Inc., Chicago, IL, USA).

Comparing the indications for RHC in 83 patients, the proportions of patients diagnosed with PH were 24/35 patients (68.5%) with RVSP >36 mmHg plus FVC/DLCO ratio >1.6, 23/34 patients (67.6%) with RVSP >36 mmHg, 4/8 patients (50.0%) with FVC/DLCO ratio >1.6 and 1/6 patients (16.6%) with progressive unexplained dyspnoea. Two the four patients who were excluded had PH due to ILD. Following the RHC results in the 78 patients who fulfilled the inclusion criteria and did not meet the exclusion criteria, PH was identified in 50 patients (64.1%) and PH was ruled out in the other 28 subjects (35.8%), although 14 of them had BoPAP with mPAP between 21 and 24 mmHg (Fig. 1). According to PH classification, 35 patients (44.8%) were diagnosed with PAH, 6 (7.6%) had PH due to left heart disease and 9 (11.5%) had PH due to ILD. Sixty-three patients with PAH and without PH were selected for DETECT algorithm analysis.

The baseline characteristics of the selected patients with SSc are summarized in Table 1. There were 59 women and 4 men, the mean ± SD age at RHC was 62.4 ± 11.6 years. Based on skin involvement, lcSSc was the most frequent cutaneous subtype in 45 patients (71.4%), 10 patients (15.9%) had dcSSc and 8 patients (12.7%) had ssSSc. The mean time from the first SSc symptom including RP was 18.6 ± 12.3 years, while for the time from first non-RP it was 10.5 ± 8.9 years. Almost the whole study population (n = 61 (96.8%)) fulfilled the ACR/EULAR 2013 classification criteria. There were no statistical differences in the presence of telangiectasias, past or current DU, history of SRC, ILD or cardiac involvement, when comparing the two groups. Although there were no statistical differences in positivity to antinuclear antibodies or the observation of a nucleolar pattern in the indirect immunofluorescence, patients with PAH more often had ACA (n = 23, 65.7% vs n = 11, 39.3%, p = 0.03), but lower positivity against anti-topoisomerase I antibodies (n = 3, 8.6% vs n = 9, 32.1%, p = 0.01). Arterial hypertension was the most frequent cardiovascular risk factor in the whole group, being identified in 27 patients (42.9%), followed by dyslipidaemia in 15 patients (23.8%); however, there were no patients with diabetes mellitus. No statistical differences were identified in the anthropometrical variables. The patients with PAH received less immunosuppressant therapy than the control group, and less treatment with calcium channel blockers. However, there were no statistical differences in the use of specific vasodilator therapy, for which the indication was RP in all cases.

Regarding respiratory parameters, the PAH group had higher FVC percent predicted/DLCO percent predicted than controls (2.0 ± 0.7% vs 1.5 ± 0.4%, p < 0.01) and shorter 6MWD (232.8 ± 76.7 m vs 322.4 ± 105.0 m, p = 0.01) (Table 2). Advanced WHO functional classes III/IV were more common in patients with PAH (n = 15 (42.9%)), than in the group with no PH (n = 2 (7.1%)) (p < 0.01). There were higher NT-proBNP and serum urate levels and higher prevalence of right axis deviation on ECG in the PAH group. In relation to echocardiographic data, patients with PAH had a greater RA area; moreover, they presented with higher TRV and RVSP than patients with no PH. Therefore, there were statistically significant differences between PAH and without PH patients in seven out of eight variables considered in DETECT algorithm.

RHC in the PAH group was characterized by superior mean PAP (Table 3). Specifically, the median (IQR) of mPAP was 42.0 (33.0 − 50.0) mmHg in patients with PAH, while it was 20.5 (17.0 to 23.0) mmHg in the group with no PH (p < 0.001). Furthermore, TPG was increased in patients with PAH, but CO was lower in those patients. The PAH group had higher pulmonary vascular resistance, with lower SvO2 and SaO2 during the RHC, compared with the group with no PH.

The DETECT algorithm steps are represented in Table 4. After step 1 of the algorithm 35 patients in the PAH group (100%) would have been referred to echocardiography compared to 24 in the group without PH (85.7%) (p = 0.03). Following the echocardiography results, RHC would have been recommended in all 35 patients in the PAH group (100%), while this was only the case in 16 subjects (57.1%) in patients with no PH (p < 0.001). However, based on the ESC/ERS 2009 guidelines, 32 patients with PAH (91.4%) and 4 patients in the group without PH (14.3%) would have been referred for RHC. The RHC referral ratio was higher using DETECT than it was using the ESC/ERS guidelines. The main reasons for not referring these patients for RHC using the ESC/ERS 2009 guidelines would have been that 12 patients had TRV >2.8 to ≤3.4 m/s but with no symptoms, 5 patients presented with TRV ≤2.8 m/s with symptoms but with no additional echocardiographic signs of PH, and another 10 patients had TRV ≤2.8 m/s with no symptoms and no additional echocardiographic signs of PH. There would have been no missed diagnosis of PAH with DETECT, but using the ESC/ERS guidelines it would have reached 8.5% (Table 5). The sensitivity (95% CI) of the DETECT algorithm was 100% (90.1–100) compared to sensitivity of 91.4% (77.6–97.0) using the ESC/ERS 2009 guidelines. The specificity was lower with DETECT at 42.9% (26.5–60.9) whereas it was 85.7% (68.2–94.3) using the ESC/ERS 2009 guidelines. No statistical differences were found in PPV, which was 68.6% (55.0–79.7) using DETECT compared to 88.9% (74.7–95.6) using the ESC/ERS 2009 guidelines. However, NPV reached 100% (75.7–100) using the DETECT algorithm, while using the ESC/ERS 2009 guidelines it was 88.9% (71.9–96.1).

As a strategy to identify mPAP ≥21 mmHg (both BoPAP and PAH), the DETECT algorithm had superior sensitivity of 93.9% (83.5–97.9) and NPV of 75.0% (46.8–91.1) compared to ESC/ERS guidelines for which the sensitivity and NPV was 69.4% (55.5–80.5) and 44.4% (27.6–62.7), respectively. Furthermore, DETECT missed 6.1% of diagnoses of mPAP ≥21, whereas using the ESC/ERS 2009 guidelines the percentage was increased to 30.6%. In fact, analysing only patients with BoPAP, there were 3 out of 14 (21.1%) missed diagnoses using DETECT, whereas there were 12 out of 14 (85.7%) missed using the ESC/ERS 2009 guidelines. However, specificity and PPV were slightly lower using the DETECT algorithm, at 64.3% and 90.2% compared to 85.7% and 94.4%, respectively, using the ESC/ERS guidelines.