Parametric Models for Survival Analysis of Childhood Cancer Patients’ Data

The rate of childhood cancer death has dropped steadily over the past 50 years. The pediatric cancer risk has remained under investigation.

This study aims at investigating the associated factors with the survival of pediatric patients with retinoblastoma, sarcoma, brain tumor, and leukemia cancer.

The cohort study of 1879 children with retinoblastoma, sarcoma, brain tumor, and leukemia aged < 1, 1 - 5, 6 - 10, 11 - 15, and > 15 years in Mahak Hospital and Rehabilitation Complex from 2007 to 2016 were enrolled in the study. Median survival time was reported for each cancer. Parametric survival models including Gompertz, Weibull, lognormal, and log-logistic models were fitted. Then, the model with almost minimum Akaike information criterion (AIC) was chosen. The hazard ratio (HR) and the analysis were performed by R3.5.1.

Totally, 270 (14.37%) patients with Retinoblastoma, 667 (35.5%) with leukemia, 625 (33.26%) with a Brain tumor, and 317 (16.87%) with Sarcoma were included in this study; 815 (43.37%) patients were female. Gompertz’s model was chosen to fit the data due to the minimum AIC. The associated factors with the survival of childhood cancers were as follows: age < 1 year, parental relation, locoregional relapse and chemotherapy alone (HR: 7.63, 1.56, 4.61, 1.12) in leukemia, other nationalities, metastasis or metastasis and loco-regional relapse and chemotherapy alone (HR = 3.74, 5.75, 2.12) in retinoblastoma, loco-regional relapse and metastasis (HR = 2.40, 3.71) in brain tumor, other ages except for 5 - 10 years, parental relation, chemotherapy alone, and metastasis (HR = 33.3, 1.80, 3.57, 3.8) in sarcoma.

Age, parental familial relationships, combination therapy, and metastasis of primary cancer were the risk factors for survival of children with 4 common cancers of leukemia, retinoblastoma, brain tumors, and sarcoma, using the Gompertz model.