Esophageal Cancer Using Weibull or Log-Normal Cure Models 

  Ghadimi MR¹ (PhD Candidate), Rasouli M¹ (PhD Candidate), Mahmoodi M² * (PhD), Mohammad K¹ (PhD), Zeraati H¹ (PhD)

  ¹ Department of Epidemiology and Biostatistics, School of Public Health,

  Tehran University of Medical Sciences, Tehran, Iran

  ² Department of Epidemiology and Biostatistics, School of Public Health,

  Tehran University of Medical Sciences & National Institute of Health Research, Tehran, Iran

  Received 26 Oct 2010, Accepted: 10 May 2011

  Abstract

  Introduction: Esophageal cancer is one of the eight common cancers in the world and the sixth cause of cancer death. The standard survival analysis, Cox proportional hazard model, is based on the assumption that all individuals in study experience the desired event however, due to the nature of information and type of event, this assumption would be unrealistic. Naive use of Cox regression analysis can be misleading. In this study, we aimed to determine impact of risk factors and demographic factors on survival and cure fraction of patients with esophageal cancer using cure parametric models.

  Methods: Data were collected from the cancer registry of Babol affiliated to Tehran University of Medical Sciences, School of Public Health. A total of 359 patients were recruited to the study and were followed for a period of 15 years.

  Results: Results showed that 62.7% of patients were men and 37.3% were women. The mean age was 60.0 year for men and 55.3 years for women. Estimated survival rates in one, three, and five years following diagnosis were 23%, 15%, and 13%, respectively. The median of survival time was 8.9 months. To assess the impact of personal factors on survival of patients, we used the mixture cure parametric (weibull and Log-normal). In both models, family history of patients was statistically significant.

  Conclusion: The results of this study suggest that screening of individuals who have family history of cancer (especially in men) can be an effective factor in reducing the risk of death in patients with esophageal cancer. Preventive programs and monitoring should be greatly considered. Since proportional hazards assumption is rejected in this study, log normal Mixture Cure Model with Logit link function could be used instead of Cox and Weibull models in survival analysis of patients with esophageal cancer.

  Key words: Survival analysis, Esophageal cancer, Cure mixture models, Log normal model, AIC criterion

  Hakim Research Journal 2011 14(1): 41- 49.

  * Corresponding Author: Department of Epidemiology and Biostatistics, School of Public Health, Tehran University of Medical Sciences, Tehran, Iran. P.O.Box: 14155-6446. Tel: +98- 21- 88989123, Fax: +98- 21- 6462267,
E-mail: mahmoodim@tums.ac.ir