Journal of Biostatistics and Epidemiology (Sep 2023)

Survival Prognostic Factors of Male Breast Cancer Using Appropriate Survival Analysis for Small Sample Size: Three Center Experience

  • solmaz taheri,
  • Mohamad Esmaeil Akbari,
  • hossein bonakchi,
  • ahmadreza baghestani

DOI
https://doi.org/10.18502/jbe.v9i3.15451
Journal volume & issue
Vol. 9, no. 3

Abstract

Read online

Introduction: Breast cancer in men is a rare disease that has been increasing in recent decades. Identifying factors influencing the survival rate of these patients is particularly important considering the small sample size. The aim of this study was to present the results of the conventional Cox- LASSO method and compare it with the newer refined generalized log-rank (RGLR) method for analyzing survival data with a small sample size. Methods: Available information related to men with breast cancer referred to 3 treatment centers in the country (Iran) between 2012 and 2020 were reviewed. Cox-LASSO and RGLR models were fitted on the data. The analyzes were done using R.4.1.2 software and the significance level of 0.05 was considered. Results: About 60% of the conflicts are reported on the left side. About 53% of men have been diagnosed at a low stage. The tumor size of 75% of the patients was between 2 and 4.3. Most patients have received modified radical mastectomy (MRM) treatment and adjuvant radiotherapy. 80% of patients had received chemotherapy and most had received anthracycline-taxane base. According to Akaike's criterion, RGLR model (AIC=289.32) was better than Cox-LASSO (AIC=314.76) model. Results of RGLR model indicated that, age (p-value= 0.038, HR >50 vs <50 = 6.75, 95% CI: 2.70–17.30), left laterality (p-value = 0.019, HR left vs right = 3.45, 95% CI: 1.48–8.02), larger tumor size (p-value=0.033, HR T2 vs T1 = 3.70, 95% CI: 2.92–6.68; HR T3 vs T1 = 4.34, 95% CI: 3.17–5.95), higher tumor grades (p-value<0.001, HR grade 2 or 3 vs grade1 = 8.67, 95% CI: 5.10–14.71), are influential factors decreasing male breast cancer patient’s survival. Conclusion: Although the results of the two existing models in the field of small sample size survival analysis (Cox-LASSO and RGLR) are close to each other, the RGLR model has performed better than the Cox-LASSO. With smaller AIC and SE of parameter estimation, RGLR model was choose compared to Cox-LASSO model.

Keywords