A Lognormal Mixture cure model with an application in eye survival of children with Retinoblastoma

Background and Retinoblastoma is one of the most common intraocular tumors. In survival analysis where some patients have not experienced the event of interest, we prefer to use cure models. Since many of patients among those diagnosed with retinoblastoma will not require enucleation surgery, these cases assumed to be cured. In order to investigate the factors affecting eye survival, mixture cure models were utilized in present study. Methods & Materials: In this prospective study, we investigated and analyzed the eye survival of 170 pediatric patients suffering from retinoblastoma in Rasool-e-Akram Hospital during 2002 to 2007. Factors affecting eye survival were analyzed using a cure model with lognormal survival function and logit link function. There were statistically significant effects on eye survival for both tumor stage and chemotherapy in mixture cure model (P<0.001, (P<0.001), respectively. Cure rates among children diagnosed with non-advanced tumor stage, for patients who had been treated with chemotherapeutic regimens and patients who had never received chemotherapy obtained 50% and 90%, respectively. Among children diagnosed with advanced tumor stage, cure rates obtained 3% and 24% (chemotherapy vs. no treatment). Cure rates for mentioned groups were 67%, 90%, 0%, and 25% respectively using Kaplan-Meier estimator. In case of patients with long-term of survival applying a cure model is appropriate. Since in eye survival analysis among patients suffering from retinoblastoma. the assumption that all members of society maybe at risk was not considered here, Cox model can be replaced with Lognormal Mixture cure model