Triple-negative breast cancer (TNBC) is the most aggressive molecular subtype of breast cancer, accounting for 10-20% of all cases with a poor prognosis (1–7). Statistically, the postoperative recurrence rate of this disease was as high as 25% (5), the 5-year overall survival (OS) for regional tumors was about 64%-72% (1, 2, 4, 7–9), and the median survival time after metastasis was only 13.3 months (5). However, as valuable as these data may be at initial diagnosis, they may be inaccurate for survivors. Traditional survival analysis can only use the patient’s status at the initial diagnosis as an assessment landmark and never consider how long the patient has survived (10, 11). Such static evaluations were unfortunate for patients who survived for several years, as no improvement in survival was observed during long-term follow-up. Previous studies demonstrated that the prognosis of long-term survivors improved dynamically over time (10–13). With improved treatment strategies, survivors achieved long-term survival and might prefer to know accurate real-time prognostic information during ongoing follow-up.

More recently, results from the M.D. Anderson Cancer Center suggested that the risk of recurrence and death in non-metastatic TNBC was strongly time-dependent, but no further studies (7). Conditional survival (CS) analysis conveyed the dynamics of survival rates over time which could provide a real-time updated estimate of survival. It was defined as the probability that a patient would survive for a further y years after surviving for x years after diagnosis (10, 11). Currently, CS analysis has been widely used in many cancers (e.g., colorectal cancer (12, 13), stomach cancer (14), liver cancer (15), esophageal cancer (16), nasopharyngeal cancer (17), lymphoma (18), pancreatic cancer (19), etc.) to assist physicians in optimizing clinical decisions and significantly reduce the psychological distress of survivors. However, to our knowledge, although many TNBC prognostic assessment tools have been developed (9, 20–27), they have never focused on prognostic changes over time. In addition, the common nomograms, while allowing for individualized prediction, have never provided dynamic prognostic information over survival time.

Considering the significant differences in clinicopathological characteristics, treatment strategies, and prognosis between metastatic TNBC and stage I-III TNBC, a subset of patients with non-metastatic TNBC was selected for our study. The aim of this study was to reveal the change in survival of non-metastatic TNBC patients over time and to develop the first CS-nomogram to provide clinicians and patients with personalized and real-time updated dynamic prognostic information.

This study was the first to analyze conditional survival patterns in non-metastatic TNBC patients and found that real-time OS in survivors improved dynamically over time. Meanwhile, this improvement was non-linear and slow in the 2nd year after diagnosis. To enable individualized and real-time prognostic assessment, we developed the CS-nomogram, which provided survivors with real-time updated on survival information.

Non-metastatic TNBC is the most malignant subtype of breast cancer but is benefiting from advances in treatment, with patients achieving long-term survival (1–6, 28–30). However, when survivors hope for accurate information about their current prognosis at follow-up, traditional survival analysis may not allow for real-time response (2–7, 28, 31). CS provides a way to quantify the improvement of a survivor’s prognosis over time, making it easier for clinicians and patients to understand. In our study, the survival of non-metastatic TNBC patients did improve over time, and these valuable data helped to reduce patient anxiety levels and improve adherence to cancer follow-up. Second, according to the guideline recommendation, the routine follow-up period for breast cancer was five years. For non-metastatic TNBC patients, it may be seven years, after which survivors’ 10-year real-time OS remained above 95%, meaning they were no different from the general population (32, 33). In addition, we found that patients’ 1-year CS (the probability of surviving several years after diagnosis and then living for one year) decreased in the second year (from CS(1|0) =96.1% to CS(1|1) =93.2%, Figure 2B ). This result corresponded to a high mortality rate in 2nd year in the annual hazard curve and validated the M.D. Anderson Cancer Center’s conclusion that non-metastatic TNBC might have a higher risk of recurrence and death in the first three years of follow-up (7). The exact cause of the substantial increase in mortality in patients in the 2nd year after diagnosis remained unclear. One explanation was that the intensity of supportive therapy received by patients decreased from the first year onwards, leading to poor survival in the next year. According to the “natural selection” effect of the Zamboni et al. residual lifetime analysis (13), the death of high-risk non-metastatic TNBC patients might promote natural selection in low-risk patients, resulting in a more favorable prognosis for survivors. Therefore, more frequent follow-up over two years was necessary for non-metastatic TNBC, and the follow-up strategy should incorporate the results of CS analysis into the final decision.

Follow-up is a dynamic process, and it is necessary to develop a dynamic assessment tool (14). However, all currently published TNBC nomograms never provided real-time updated prognostic information over time for survivors, for which we developed a novel CS-nomogram. This model may be of more interest to long-term survivors than the traditional nomogram, especially for those with poor initial survival. With both dynamic assessment and personalized prediction, the CS-nomogram provided real-time updated prognostic information over survival time. For example, a patient has a 10-year OS of 35% at the initial diagnosis. After 5 years of follow-up, the CS model can tell her, “Your 10-year survival rate has increased by 40%”, which will significantly strengthen her confidence in fighting cancer. Clinicians can adjust the patient-appropriate monitoring plan in real-time based on the changing risk. The application of the CS-nomogram cannot be limited to non-metastatic TNBC but to all current high-burden diseases. This method will be one of the effective strategies to optimize clinical decision-making to clarify patient prognosis and save healthcare resources.

The CS analysis was calculated as a subset of survivors in the data, which was why we used the SEER database, because the data from surviving patients decreased over time (10, 11). Regarding the performance of the CS-nomogram, the calibration curves nearly overlapping with the ideal curve, the stable AUC curves over 10 years, and the DCA that showed a high net benefit were all evidence of its strength. Thus, this novel model provided prognostic information consistent with real-time follow-up and had considerable clinical utility.

Our study had several limitations. First, as a retrospective study, it was inevitably biased; second, the SEER database was missing variables such as lymphatic vascular invasion, Ki-67, chemotherapy cycles and drugs, and radiotherapy doses, which might limit some of our analyses. Third, our study was unsuitable for patients with metastatic TNBC because their treatment was inconsistent with stage I-III, which may affect the model performance. Fourth, external validation was challenging because of the large amount of data needed for CS analysis. Fifth, the CS nomogram needed to be updated within a few years with the advent of new therapies.

The CS analysis in this study clarified the dynamic and non-constant improvement in survival over time in non-metastatic TNBC patients. For real-time and personalized prognostic prediction, we developed a novel CS-nomogram. This model provided accurate prognostic information to assist in optimizing clinical decisions and significantly reduced the psychological stress of survivors through real-time prognosis updates. However, future studies need prospective data and more predictors to validate and update the model to make it more generalizable.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Ethical review and approval was not required for the study on human participants in accordance with the local legislation and institutional requirements. The ethics committee waived the requirement of written informed consent for participation.

XM and YG: Designed the study. XM, YC, and XC: Performed the study and analyzed the data. XM: Wrote the manuscript. XM and YG: Provided the expert consultations and clinical suggestions. XC and YC: Conceived of the study, participated in its design, and coordination. YC, XC, and YG: Helped to draft the manuscript. All authors contributed to the article and approved the submitted version.