A Predictive Nomogram for Early Death in Pheochromocytoma and Paraganglioma.

BACKGROUND

Pheochromocytoma (PHEO) and paraganglioma (PGL) are relatively rare neuroendocrine tumors. The factors affecting patients with early death remain poorly defined. We aimed to study the demographic and clinicopathologic pattern and to develop and validate a prediction model for PHEO/PGL patients with early death.

METHODS

Data of 800 participants were collected from the Surveillance Epidemiology and End Results (SEER) database as a construction cohort, while data of 340 participants were selected as a validation cohort. Risk factors considered included the year of diagnosis, age at diagnosis, gender, marital status, race, insurance status, tumor type, primary location, laterality, the presence of distant metastasis. Univariate and multivariate logistic regressions were performed to determine the risk factors. R software was used to generate the nomogram. Calibration ability, discrimination ability, and decision curve analysis were analyzed in both construction and validation cohorts.

RESULTS

PHEO and PGL patients accounted for 54.3% (N=434) and 45.7% (N=366), respectively. More than half of tumors (N=401, 50.1%) occurred in the adrenal gland, while 16.9% (N=135) were in aortic/carotid bodies. For the entire cohort, the median overall survival (OS) was 116.0 (95% CI: 101.5-130.5) months. The multivariate analysis revealed that older age ( age between 31 and 60: OR=2.03, 95% CI: 1.03-4.03, ; age older than 60: OR=5.46, 95% CI: 2.68-11.12, ), female gender ( OR=0.59, 95% CI: 0.41-0.87, ), tumor located in aortic/carotid bodies ( OR=0.49, 95% CI: 0.27-0.87, ) and the presence of distant metastasis ( OR=4.80, 95% CI: 3.18-7.23, ) were independent risk factors of early death. The predictive nomogram included variables: age at diagnosis, gender, primary tumor location, and distant metastasis. The model had satisfactory discrimination and calibration performance: Harrell's C statistics of the prediction model were 0.733 in the construction cohort and 0.716 in the validation cohort. The calibration analysis showed acceptable coherence between predicted probabilities and observed probabilities.

CONCLUSIONS

We developed and validated a predictive nomogram utilizing data from the SEER database with satisfactory discrimination and calibration capability which can be used for early death prediction for PHEO/PGL patients.