Prediction of mortality in acute pulmonary embolism in cancer-associated thrombosis (MAUPE-C): derivation and validation of a multivariable model.

Abstract

Optimal risk stratification of patients with cancer and pulmonary embolism (PE) remains unclear. We constructed a clinical prediction rule (CPR) named 'MAUPE-C' to identify patients with low 30days mortality. The study retrospectively developed and internally validated a CPR for 30days mortality in a cohort of patients with cancer and PE (both suspected and unsuspected). Candidate variables were chosen based on the EPIPHANY study, which categorized patients into 3 groups based on symptoms, signs, suspicion and patient setting at PE diagnosis. The performance of 'MAUPE-C' was compared to RIETE and sPESI scores. Univariate analysis confirmed that the presence of symptoms, signs, suspicion and inpatient diagnosis were associated with 30days mortality. Multivariable logistic regression analysis led to the exclusion of symptoms as predictive variable. 'MAUPE-C' was developed by assigning weights to risk factors related to the β coefficient, yielding a score range of 0 to 4.5. After receiver operating characteristic (ROC) curve analysis, a cutoff point was established at ≤ 1. Prognostic accuracy was good with an area under the curve (AUC) of 0.77 (95% CI 0.71-0.82), outperforming RIETE and sPESI scores in this cohort (AUC of 0.64 [95% CI 0.57-0.71] and 0.57 [95% CI 0.49-0.65], respectively). Forty-five per cent of patients were classified as low risk and experienced a 2.79% 30days mortality. MAUPE-C has good prognostic accuracy in identifying patients at low risk of 30days mortality. This CPR could help physicians select patients for early discharge.