Continuous data are presented as the mean±SD. Categorical variables were grouped and compared using the χ2 test or Fisher’s exact test. Continuous variables were compared using Student’s t-test. Univariate and multivariate survival curves were generated using the Kaplan–Meier method, and the differences between curves were analyzed using logrank tests. Univariate and multivariate Cox proportional hazard models were used to identify factors that are associated with cancer-specific survival. The cancer-specific survival in the SEER database was defined as the cause of death. It means that patients were dead because of breast cancer, not other diseases. The prognostic factors were selected as nomogram parameters. Akaike information criterion (AIC) and Harrell’s C statistic were used to estimate the proportion of correct predictions and relative discriminative abilities. To adjust the comparisons and avoid distortions from bias in retrospective trials, a propensity score analysis was used.5,6 All statistical tests were two sided, and P values<0.05 were considered to be statistically significant. Statistical analyses were performed using SPSS 13.0 and R software version 3.3.0 (http://www.rproject.org) with the “SEERaBomb”, “rms”, “MatchIt”, “PSAgraphics” and “AICcmodavg” packages.
A total of 16998 patients with breast cancer who were included in the SEER database from 2010 to 2013 fulfilled the study criteria. Among these patients, 13007 underwent surgery. The demographic and pathological information for all patients are presented in Table 1. These data represent all patients, including those with and without surgery. Overall survival and cancer-specific survival were significantly better in patients who underwent surgery than in those who did not (P<0.001; Figure S1A and S1B). In addition, overall survival and cancer-specific survival were significantly better in patients who received radiotherapy than in those who did not (P<0.001; Figure S1C and S1D).
(To view a larger version of Table 1, click here.)
(To view a larger version of Figure 1, click here.)
For patients who underwent a surgical procedure (including partial mastectomy, subcutaneous mastectomy, simply total mastectomy, modified radical mastectomy, radical mastectomy and extended radical mastectomy) with exact pathological information, the detailed demographic and pathological information are presented in Table 2. In patients older than 80 years, the patient’s age at diagnosis was a factor that affected overall survival (P<0.001; Figure 1). The risk of a cancer-specific modality dramatically increased as patients get older. The cancer-specific 3-year survival rate was 79% in patients older than 95 years and 94% in those between 80 and 85 years. A univariate analysis of overall survival indicated that patient race, the size of the breast cancer and its histological grade, the status of distant metastasis, the surgical approach, the use of radiotherapy, estrogen receptor (ER) positivity, progesterone receptor (PR) positivity, human epidermal growth factor receptor-2 (HER-2) positivity and the status of metastatic axillary lymph nodes were prognostic factors (Figure 1). After a Cox proportional hazard analysis was performed, the first nine of these factors were found to be independent prognostic factors. These independent prognostic factors were therefore included in the nomogram model. In the nomogram estimation system, each factor from the multivariate Cox proportional hazard regression model was attributed a weighted point value that implied its contribution to a prognosis of survival. We found that elderly breast cancer patients with higher scores had a worse cancer-specific prognosis than was observed in those with lower scores. The final nomogram model that was developed to predict cancer-specific survival in elderly breast cancer patients is shown in Figure 2A.
(To view a larger version of Table 2, click here.)
(To view a larger version of Figure 2, click here.)