Based on data including 693 patients from the SEER database, we constructed a nomogram model to predict the 5-year and 8-year OS for patients with OBC. Six variables were selected by clinical significance to construct and validate the capability of the model, which could provide the basis for future clinical decisions. Measured by range along nomogram scales, age was the most important prognostic factor, followed by tumor stage, ER-status, chemotherapy, radiotherapy and surgery. To our knowledge, this was the first and largest population-based nomogram model to predict the prognosis of patients with OBC.
In our study, prognosis measurement was estimated by OS, which was a common and objective index for patients with OBC. In univariate Cox regression analysis, we found that age at diagnosis, tumor laterality, AJCC-TNM stage, surgery, radiotherapy, chemotherapy, ER status and PR status were significantly correlated with OS. To decrease the estimation bias and further confirm the independent prognosis factors on OS for patients with OBC, the multivariate Cox regression analysis was conducted. After adjusting for demographic, clinicopathological and therapeutic variables, we found that age, AJCC-TNM stage, radiotherapy, chemotherapy, ER status and surgery were still significantly associated with OS. Based on univariate and multivariate Cox regression analysis, six variables were selected to construct the prediction model.
Notably, we found that treatment patterns were the essential prognosis factors in our model. The nomogram model indicated that surgical operation played an important role to prolong the survival interval for patients with OBC, and mastectomy recipient had better OS than BCS recipient. Meanwhile, surgery strategies for OBC had been not reached consensus. Some previous research suggested mastectomy could provide the most effective local treatment for OBC patients[10, 30–32], while these findings were inconsistent with the recent studies that the application of BCS had similar OS outcome, even better than mastectomy [33–36]. This paradoxical results might be caused by selection bias, because the rate of BCS recepient was only 11.5%, which was much less than other two groups. Furthermore, radiotherapy and chemotherapy were significantly associated with better OS, which were consistent with previous studies[37–42].
And our nomogram also demonstrated that the effect of chemotherapy might exceed that of radiotherapy.
In addition, ER-positive status played an important role to prolong the survival interval for patients with OBC, which was consistent with previous studies[6, 43, 44]. The underlying reason was that the majority of ER-positive breast cancer are luminal A and luminal B, which had achieved better prognosis[45–47]. Due to luminal A and luminal B breast cancer were sensitive to endocrine therapy, and this treatment was an important part of the comprehensive therapy of hormone receptor positive breast cancer and its efficacy had been widely accepted[48-50]. But PR status was not correlated with OS in our study, we considered that might be caused by retrospective bias, because the percent of PR-negative BC (54.5%) was 1.5 times than the percent of PR-positive BC (33.3%) in our study.
In conclusion, the nomogram model could be applied to predict the survival outcome of various cancers, which integrated clinical and demographic factors to evaluate the risk of specific diseases[18, 19, 21, 23]. In traditional, the AJCC-TNM stage system was the primary choice to predict prognosis and make clinical decision. But patients at the same stage usually had different prognosis, because AJCC-TNM stage system had not considered various variables comprehensively, such as clinical characteristics, treatment method and sociodemographic characteristics. Therefore, we compared the nomogram, which involves more variables, with the conventional AJCC-TNM stage system. The the NRI value and IDI value of the nomogram versus the TNM stage system suggested that the nomogram prediction model had better predictive capability than the TNM stage system alone. Furthermore, DCA curves demonstrated that our model forecasted survival outcome with better clinical valus and utility than the conventional staging system. In the validation cohort, the results could also be replicated favorably. In conclusion, our nomogram could provide accurate and individual prediction of OS in patients with OBC.
Although the nomogram performed well, our subject indeed had some limitations, as shown below: Firstly, the nomogram based on retrospective study, which could not prove causation and result in selection bias. Secondly, we were unable to exclude the impact of potential confounders, such as family history, comorbidities, health status, patient anxiety, BRCA gene status, which were not included in SEER database. Lastly, P value < 0.05 was used to possess the statistics sense, and no adjustment was made for multiple analysis; the chance of falsely rejecting a null hypothesis may exceed 0.05. Multicenter clinical validation was also needed to evaluate the external utility of our nomogram.