Performance of the DETECT Algorithm and Cluster Analysis in Screening for Systemic Sclerosis Pulmonary Hypertension Groups

Ryan Osgueritchian¹, Hoda Mombeini¹, Vivek Jani², Adrianne Woods³, Steven Hsu¹, Matthew Lammi⁴, Paul Hassoun⁴, Fredrick Wigley⁵, Laura Hummers⁶, Stephen Mathai⁷, Ami Shah⁸ and Monica Mukherjee¹, ¹Johns Hopkins University Division of Cardiology, Baltimore, MD, ²Johns Hopkins University School of Medicine, Baltimore, MD, ³Johns Hopkins University Division of Rheumatology, Baltimore, MD, ⁴Johns Hopkins University Division of Pulmonary and Critical Care Medicine, Baltimore, MD, ⁵Johns Hopkins University, Division of Rheumatology, Baltimore, MD, Baltimore, MD, ⁶Johns Hopkins University, Division of Rheumatology, Baltimore, MD, Ellicott City, MD, ⁷Johns Hopkins University Division of Pulmonary and Critical Care Medicine,, Blatimore, ⁸Division of Rheumatology, Johns Hopkins University, Ellicott City, MD

Meeting: ACR Convergence 2024

Keywords: Cardiovascular, Imaging, prognostic factors, risk assessment, Scleroderma

Session Information

Date: Monday, November 18, 2024

Title: Systemic Sclerosis & Related Disorders – Clinical Poster III

Session Type: Poster Session C

Session Time: 10:30AM-12:30PM

Background/Purpose: Pulmonary hypertension (PH) portends poor outcomes in Systemic Sclerosis (SSc). Recent efforts for early identification and intervention in SSc patients with pulmonary arterial hypertension (PAH) with algorithms like DETECT have drastically improved outcomes in this population. However, such algorithms have failed in other WHO (World Health Organization) PH subgroups. This study aims to evaluate the effectiveness of the DETECT algorithm and Cluster analysis in predicting Group 2 (PH due to left heart disease) and Group 3 (PH due to chronic lung disease).

Methods: We included a well-characterized cohort of SSc patients from our center with clinical, laboratory, and echocardiographic data near the right heart catheterization date. Scores were generated for each step of the DETECT algorithm when applicable. Then, k-means cluster analysis was conducted using clinical, laboratory, and echocardiographic variables. Table 2. Cluster characteristics were compared through the Kruskal-Wallis test. Clustering was decided according to the elbow plot. ROC analysis across all three PH groups was then conducted for step 1 of the DETECT algorithm, using an empirically optimal cutoff. For the cluster model, ROC analysis considered inclusion in cluster 2 as a positive screening test, while inclusion in cluster 1 was considered negative.

Results: Our cohort consisted of 435 SSc patients, 183 with PAH, 68 with Group 2 PH, 109 with Group 3 PH, and 73 without PH: mean age 60 ± 12 years, 82% female, 72% White, 76.8% limited SSc. Table 1. Across PH Groups, AUC ranged from 0.68 – 0.81, with SSc- PAH having the greatest AUC. Table 3. The DETECT algorithm’s performance in Group 2 SSc PH (AUC = 0.64) and Group 3 SSc PH (AUC = 0.68) was considerably less reliable than in SSc PAH (AUC = 0.81). Variables chosen for cluster analysis are all listed in Table 2. The clustering model did not perform adequately in predicting PAH (AUC = 0.69) or Group 3 PH (AUC = 0.59); however, it performed better than DETECT in screening for Group 2 PH (AUC = 0.73). Table 3.

Conclusion: The DETECT algorithm performs poorly in screening for SSc Group 2 and Group 3 PH. Patient clustering provides improved performance in predicting Group 2 PH but was not effective in predicting PAH or Group 3 PH.

Table 1 Detailing Demographic, Laboratory, Clinical, and Echocardiographic Data of the Study Population

Table 2 Comparing Demographic, Laboratory, Clinical, and Echocardiographic Data Between Clusters 1 and 2

Table 3 AUC for Step 1 of the DETECT algorithm and the Clustering Model for Screening each Systemic Sclerosis PH Group

Disclosures: R. Osgueritchian: None; H. Mombeini: None; V. Jani: None; A. Woods: None; S. Hsu: None; M. Lammi: None; P. Hassoun: None; F. Wigley: None; L. Hummers: AbbVie/Abbott, 2, AstraZeneca, 5, Biotest, 1, 2, Boehringer-Ingelheim, 2, 5, Cumberland, 5, GlaxoSmithKlein(GSK), 5, Kadmon, 5, Medpace, 5, Merck/MSD, 5, Mitsubishi Tanabe, 5, prometheus, 5; S. Mathai: None; A. Shah: Arena Pharmaceuticals, 5, Kadmon, 5, Medpace LLC, 5; M. Mukherjee: None.

To cite this abstract in AMA style:

Osgueritchian R, Mombeini H, Jani V, Woods A, Hsu S, Lammi M, Hassoun P, Wigley F, Hummers L, Mathai S, Shah A, Mukherjee M. Performance of the DETECT Algorithm and Cluster Analysis in Screening for Systemic Sclerosis Pulmonary Hypertension Groups [abstract]. Arthritis Rheumatol. 2024; 76 (suppl 9). https://acrabstracts.org/abstract/performance-of-the-detect-algorithm-and-cluster-analysis-in-screening-for-systemic-sclerosis-pulmonary-hypertension-groups/. Accessed .

« Back to ACR Convergence 2024

ACR Meeting Abstracts - https://acrabstracts.org/abstract/performance-of-the-detect-algorithm-and-cluster-analysis-in-screening-for-systemic-sclerosis-pulmonary-hypertension-groups/