Researchers performed a retrospective analysis to identify risk factors and subsequently develop a predictive scoring model for pulmonary embolisms in patients with lung cancer. The initial analyses identified 7 parameters that would inform the nomogram scoring system.
In the report, published in Clinical and Applied Thrombosis/Hemostasis, the study’s lead author, Ning Zhu, wrote that the team’s scoring system demonstrated “good accuracy and discrimination” for distinguishing between patients with lung cancer who did or did not have a pulmonary embolism, and was also effective for predicting risk of pulmonary embolism in patients with lung cancer.
Nomogram Risk Scoring for Pulmonary Embolism in Lung Cancer
The authors’ retrospective analysis included clinical and laboratory data of 900 patients with lung cancer, both with and without pulmonary embolisms. The cohort was divided 7:3 into internal development and validation groups and researchers used logistic regression models to construct a diagnostic model of the nomogram scoring system.
The 7 variables used to develop the predictive model were:
- Stage III-IV disease
- Indwelling central venous catheter
- Serum albumin levels
- Serum hemoglobin levels
- Serum D-dimer levels
In the development cohort, authors calculated the diagnostic model had an area under the receiver operating characteristic curve (AUC) of 0.918 (95% CI, 0.893-0.943), and sensitivity and specificity of 88.5% and 80.5%, respectively, for identifying patients with pulmonary embolism.
In the internal validation group, investigators reported the model had an AUC of 0.921 (95% CI, 0.884-0.953) with a sensitivity of 90.5% and specificity of 80.4%. Additionally, in an external validation cohort of 108 patients, the model reportedly had an AUC, sensitivity, and specificity of 0.929 (95% CI, 0.875-0.983), 85.0%, and 87.5%, respectively.
While the authors acknowledged their retrospective analysis-based nomogram scoring system requires validation in larger, prospective studies, they nonetheless suggested their predictive model can help clinicians screen patients with lung cancer for pulmonary embolism.