González-Arribas et al. present an important prospective application of the DETECT algorithm for pulmonary hypertension (PH) screening in systemic sclerosis (SSc), addressing a pragmatic question: how the tool performs in a real-world cohort with predominantly limited cutaneous disease and relatively preserved diffusion capacity for carbon monoxide (DLCO).1 The authors applied the two-step DETECT protocol, performed transthoracic echocardiography (TTE) when Step 1 thresholds were exceeded and referred for right heart catheterization (RHC) when Step 2 indicated, collecting clinical, laboratory, pulmonary function and echocardiographic data to identify correlates of PH.

In methods, the cohort (n=85) was well phenotyped with routine variables used by DETECT (telangiectasia, anti-centromere antibody [ACA], serum uric acid, forced vital capacity [FVC], DLCO, ECG axis, TRV, and right atrial area). Multivariable logistic regression identified tricuspid regurgitation velocity (TRV) as a strong independent predictor (OR ∼11.6) and DLCO (%) as inversely associated with PH. The authors report 11 confirmed PH cases (12.9%) and note detection even among patients with DLCO>60%.

The results underscore two clinically relevant messages. First, TRV when obtainable retains high predictive value and should remain central to echocardiographic risk stratification in SSc. Second, a substantial fraction of patients with only mildly reduced or preserved DLCO harbored PH, suggesting that exclusive reliance on DLCO thresholds may miss early vascular disease.

There are several considerations that, if addressed, would strengthen the translational impact of these findings. The multivariable model shows very wide confidence intervals for TRV and extreme OR estimates that raise concern for model instability. Reporting events-per-variable, variance inflation factors or other collinearity diagnostics, and providing internal validation (bootstrap or penalized regression) would clarify model robustness and mitigate overfitting risk.2 The authors note exclusion of certain correlated predictors but do not detail alternative modeling approaches (dimension reduction or shrinkage) that preserve information while controlling collinearity. Converting continuous predictors such as TRV and DLCO into clinically actionable thresholds and reporting calibration (e.g., calibration plots, Hosmer–Lemeshow) and decision-analytic measures (net reclassification improvement, decision curves) would help clinicians translate statistical associations into screening decisions.3

Operationally, greater emphasis on echocardiographic acquisition and interpretation standards would be valuable: specify whether TRV measurements were centrally adjudicated, whether sonographers/readers were blinded to clinical data, and how non-measurable TRV values were handled—issues that materially affect real-world yield. Finally, exploring integrated prediction that combines routinely available biomarkers with imaging (for example, NT-proBNP together with TRV and DLCO) using modern prediction techniques could improve specificity while preserving DETECT's high sensitivity.

In summary, González-Arribas et al. provide practical evidence that the DETECT pathway identifies PH even among SSc patients with relatively preserved DLCO and that TRV remains a potent predictor. Addressing model performance, measurement rigor, and clinically oriented decision metrics in subsequent work will accelerate safe, cost-effective implementation of structured screening in varied SSc populations.