---

ABSTRACT Background and Aim: The Cox proportional hazard model is the standard approach for analyzing survival data in many cases. One restriction of this method, however, is that it assumes that the log of the hazard function relates to the covariates through a linear function. As a consequence, it fails to estimate efficiently the effect of the non-linear terms. But we can estimate the survival rate by using spline functions. Our goal is to investigate the appropriateness of an alternative method, the so-called single index model, for estimation of the survival rate of the patients with acute myocardial infarction. Material and Methods: This was a descriptive analytical cohort study which included 650 subjects with acute myocardial infarction. The patients were followed up for one year to ensure survival or detect death events. Data were recorded in a pre-defined check list. In this study the relationship between the log of the hazard function and covariates were considered unknown. We estimated the coefficients of the model by using the polynomial spline and penalized partial likelihood. Data analysis was carried out by using R version 2.12 software and significant levels were considered 0.05. Results: We found the Cox model with unknown link function to have larger log likelihood than the standard Cox model. The effects of estimated parameters in both models were relatively different. Effects of diabetes and arrhythmia in Cox model with unknown link function were significant (P<0.05). In standard Cox model unlike Cox model with unknown link function the age was significant (P<0.05). Conclusion: Considering the results of this study Cox model with unknown link function could estimate the effect of factors such as diabetes and arrhythmia in the survival of the patients, in addition to the effects of streptokinase and ejection fraction. Key words: Survival analysis, Single index model, Cox proportional hazard model, Myocardial infarction. Received: May 10, 2012 Accepted: Jun 26, 2012

---

Background and Aim: This study aimed to analyze the factors affecting time and experience of relapse in the patients with Hodgkin's lymphoma, using cure fraction.
Material and Methods: This retrospective study included all the patients diagnosed as Hodgkin's lymphoma in the Center for oncology and hematology in Shafa Hospital in Ahwaz City from 2002 to 2012. We used survival analysis and cure fraction models In order to answer the question, "why are some people in the study don’t experience recurrence?". We used mixture and nonmixture cure models using Weibull, exponential, log-normal, log-logistic and gamma distributions, and a logistic model for estimation of  the proportion of safe individuals; and Cureregr8 instruction for study of its associated factors. STATA13 software was used for data analysis. Akaike information criterion (AIC) was used to compare the performance of these models.
Results:  Parametric model of Weibull distribution had the lowest AIC (804.3171).  Factors affecting long-term survival of the patients with Hodgkin's lymphoma, were hemoglobin > 10.5 (P-value = 0.018) and stage of the disease (P-value = 0.032). Factors affecting short-term survival of the patients included age >30 years (P-value = 0.001), involvement of the groin site (P-value = 0.010) Stage of disease (P-value < 0.001) and bone marrow involvement (P-value = 0.003).
Conclusion: When the study population includes two susceptible and non-susceptible (safe or healed) subgroups in regard to recurrence of Hodgkin's lymphoma, use of cure models for separate evaluation of the variables associated with the long and short term survival and cure rates is appropriate.
 

---

Background and Aim: Breast cancer is the most important cause of cancer death in women. The purpose of this study was to evaluate the effect of Estrogen Receptor (ER), Human Epidermal Growth Receptor (HER2) and other factors on post-surgical survival of the patients with breast cancer using Bayesian approach for parametric proportional hazards model.
Materials and Methods: This was a retrospective study. Data of 165 breast cancer patients who had undergone surgery at Ahvaz Healing Diagnostic Center from 2004 to 2014 were recorded in a data collection form. The variables of age, tumor size, number of lymph nodes involved, cancer grade, ER status and HER2 status were evaluated. Survival time was calculated from the date of surgery to the date of death or study end date (September 2015), in months. In the Bayesian approach in parametric survival analysis models with proportional hazards, the lateral distribution of parameters was estimated using MCMC method. Also, we evaluated efficiency of the models using the deviance information criterion. All data analysis steps were performed by using Stata15 software. Significance coefficients of the model were determined using the 95% credible interval.
Results: The mean and standard deviation of age were 46.40 and 9.94 years, respectively. Deviance information criterion for Weibull parametric model was lower than those of other parametric models. Based on the Bayesian estimation of the Weibull's proportional hazards parametric model, tumor size (HR = 1.40), the number of involved lymph nodes (HR = 1.016), Ki67 status (HR = 1.115), tumor grade (HR = 1.022), HER2 status (HR = 1.760) and ER status (HR = 1.381) had a positive effect on risk of death. Age had a negative effect on risk of death (HR=0.978).
Conclusion: Based on the Bayesian proportional hazards Weibull model, tumor size, the number of involved lymph nodes, Ki67, tumor's grade, HER2 and ER had a positive effect on the risk of death.