Risk factors and prediction of second primary cancer in primary female non-metastatic breast cancer survivors

Introduction

Female breast cancer (BC) is the most common cancer in women. In the US, BC showed the highest incidence among female cancers regardless of race or ethnicity, and the mortality was second only to lung and bronchus cancer between 2012 and 2016 [1]. The American Cancer Society reported that the incidence of female BC increased slightly from 2006 to 2015 by 0.4% annually, and more than 3.5 million women had a history of invasive breast cancer by 2016 [2]. By contrast, the mortality of BC has dropped significantly by 40% from 1989 to 2016 [3], and this is attributed to improvements in early detection, diagnosis, and treatments.

With the increasing number of BC survivors, long-term negative outcomes, such as developing a second primary cancer (SPC), have become a new public health concern. Evidence shows that BC survivors have a higher risk for subsequent cancers than the general population [4–7], particularly during the first 10 years after initial diagnosis [4]. Age [8], race [9], hormone receptor status [10], and lifestyle have been reported to be related to the development of SPC among female BC survivors.

This study aimed to investigate the risk factors of developing second primary BC and non-BC among BC survivors, with emphasis on prediction of the individual probabilities of SPCs conditioned on the patient’s characteristics. In most previous studies, the risk of SPCs in BC survivors was either calculated with standard incidence ratios (SIRs) [6, 11–13], or with hazard ratio in traditional survival analysis [10, 13, 14]. The former is a typical univariate analysis, and focused on the overall risk to the general population, while the latter neglects the impact of death, which precludes the occurrence of SPCs in BC survivors, leading to an upward bias. In our study, the multivariate competing-risk model was used, in which death was considered as a competing event. Furthermore, the competing-risk model-based nomogram was provided as the tool for predicting SPC risks in female BC survivors for clinical convenience.

Results

The characteristics of the study population are listed in Table 1. Of the 208,474 FBC patients identified, 6,242 (3.0%) developed second primary BC, and 12,350 (5.9%) developed second primary non-BC. There were almost 80 sites developing SPCs in addition to breast. The top-10 most commonly diagnosed tumor sites of second non-BC were lung and bronchus, corpus uteri, melanoma of the skin, thyroid, pancreas, ovary, kidney, urinary bladder, NHL-Nodal and cecum (Supplementary Figure 1). The median and maximum follow-up time were 7.4 and 12 years, respectively.

CIFs of second primary BC and second primary non-BC are shown in Figures 1 and 2. Regarding death as a competing event, the 3-, 5- and 10-year cumulative incidences of second BC were 0.9%, 1.6% and 4.4%, respectively; the 3-, 5- and 10-year CIFs of second non-BC were 2.3%, 3.9%, and 7.8%, respectively. Female BC patients diagnosed at older age had higher incidence rate of developing SPCs (P < 0.001 for both types of SPCs). Race, nodal status, tumor grade, tumor stage, and treatment were also significantly associated with cumulative incidences of both second primary BC and non-BC.

Results for the selected variables in Fine-Gray models are listed in Table 2. Compared with young patients (age < 50), patients aged over 70 years had 47% (sHR=1.47, 95%CI: 1.36 - 1.58) excessive risk of second BC, while the risk of developing second non-BC in patients aged between 50-59, 60-69 and over 70 years were 1.57 times (sHR = 1.59, 95%CI: 1.48 - 1.67), 2.56 times (sHR = 2.59, 95%CI: 2.42 - 2.72) and 4.65 times (sHR = 4.70, 95%CI: 4.39 - 4.93) higher, respectively. Black women had the highest risk (sHR=1.47, 95%CI: 1.36 - 1.58) for second BC compared with white women and Asian/Pacific Islander, and Asian or Pacific Islanders were less likely to develop second non-BC (sHR=0.84, 95%CI: 0.78 - 0.91). Tumor size over 2 cm (sHR = 1.06, 95%CI = 1.01 - 1.12), negative HR (sHR=1.40, 95%CI: 1.32 - 1.49), mixed histology (sHR = 1.16, 95%CI = 1.07 - 1.26), lumpectomy alone (sHR=1.67, 95%CI: 1.55 - 1.81), and surgeries combined with radiotherapy were significantly associated with increased risk of second BC. Tumor size over 2 cm (sHR = 1.09, 95%CI = 1.06 - 1.14) was also related to the risk of second non-BC.

Competing-risk nomograms for based on the Fine-Gray model were constructed to predict the 5- and 10-year probabilities of second primary BC (Figure 3). Each selected variable is listed separately in the figure, with a corresponding point assigned to a given magnitude of the variable according to the point scale at the top of image. Then for every patient, the total points by summing up points of all variables were matched to the total points scale, which correspond to the 5- and 10-year predictive probabilities at the bottom of image. For example, a 65-year-old black woman was diagnosed with small (< 2 cm), HR-negative, lobular, and localized breast tumor and treated with lumpectomy alone, then she would have 3.8% and 10.4% probabilities of developing a second primary BC after initial diagnosis at 5 years and 10 years, respectively. Calibration curves for 5- and 10-year probabilities of second primary BC indicated good calibration with high correlation between predicted probabilities and observed probability of second BC (Figure 4). The c-index was 0.61 with a moderate discrimination power. DCA was applied to evaluate clinical usefulness of the prediction nomogram. Figure 5 compares the net benefit of the predictive model to those in two hypothetic scenarios: screening all BC survivors and screening none. Within a wide range, the clinical net benefit of the risk model used in 5- and 10-year second BC predictions was larger than that under the assumption of not using the model or screening all patients (Figure 5A and 5B).

Discussion

To our knowledge, the current study is the largest one to predict the risk of second primary BC and compare the risk factors for second BC and non-BC in female BC survivors. According to our population-based findings, the overall 3-, 5-, and 10-year cumulative incidences for second primary BC were 0.9%, 1.6% and 4.4% respectively, and for second primary non-BC were 2.3%, 3.9%, and 7.8%, respectively. Competing-risk events occur frequently in oncology studies. For example, 5.1% of patients died before developing an SPC in this study. Given that censoring these patients will lead to an overestimated probability of SPCs, the subdistribution competing-risk model was used. For clinical convenience, a competing-risk nomogram was proposed, and the clinical 5- or 10-year risk of developing second BC could be estimated. Identifying subpopulations and individuals with distinct risk of SPCs using the proposed risk model can greatly benefit patients through decision making, because a patient with high risk of SPCs might be asked to consider more subsequent screening, changing adverse living habits or choosing optimal treatment for the current BC to avoid increasing risk from the side effect. Ideally, for those with high risk of developing SPCs, active and regular surveillance could benefit them from early prevention, early detection and early treatment, to improve their quality of life. High-risk BC patients would be identified according to our proposed nomogram, and only for those with high-risk patients, active strategy could be considered. In practice, the implementation of follow-up surveillance strategy not only depends on the risk of subsequent SPC, but also on other factors, such as patient’s financial arrangement, medical insurance and access to healthcare.

Results from the multivariate competing-risk models indicated that age of initial diagnosis and race were significantly related to the risks of both second primary BC and non-BC. Risk differences of age at diagnosis have been extensively examined but remains debatable. Some studies showed that young patients (< 50 years) had greater predisposition to SPCs than the old patients (≥ 50 years) [4, 6, 11, 12, 15]. However, these findings concluded from comparisons between every age group of general population and of the study population without considering other factors simultaneously, thereby SIR cannot comprehensively measure the effect of age. Instead, multivariate Cox proportional hazard regression or its extensions would be more appropriate to identify risk differences based on multiple factors. Our research found that an older age was associated with higher risk of second primary BC and non-BC in BC patients, which is consistent with other multivariate studies [8, 10, 13, 14]. It is well documented that advanced age is among the most important risk factors for many specific cancer types due to the functional decline in immune system, increased susceptibility to carcinogen, reduced DNA repair, and long-term effect of harmful lifestyle. In addition, that older patients tended to develop more SPCs may also affected by less systemic therapy due to their poor physical conditions. Systemic therapy, including chemotherapy, hormonal therapy and immunotherapy, works throughout the entire body for treating or preventing metastasis, and it may also inhibit the development of new tumors before their diagnosis.

Race is also an important demographic predictor for SPC risk. The incidence and risk of second BC was higher among black women than white women and Asian/Pacific Islanders. Other studies also confirmed this racial disparity in the risk of second primary BC [9, 16, 17], which might be explained by the combination of genetic or biologic, clinical and socioeconomic factors. For example, some genetic differences in BRCA1, BRCA2 and p53 mutations between African Americans and Caucasians in the US have been shown to be associated with different pathological characteristics and tumor prognosis [18]. Moreover, a lower socioeconomic status in any race tends to cause worse tumor pathological characteristics, inappropriate treatments and less healthcare utilization before and after first primary BC [19].

We also observed increased second primary BC risk in BC subtypes classified by clinical tumor characteristics, such as tumor size, HR status, tumor histology and SEER stage. HR-negative BC patients had a higher risk of SPC, which is consistent with previous studies [10, 20, 21]. However, it remains unclear whether the correlation between HR status of the initial BC and risk of second BC reflects an underlying genetic predisposition or exposure to hormone or other risk factors. A review concluded that BRCA1 mutation carriers were more likely to have triple-negative (ER-, PR- and HER2-negative) BC [22], and these women also had increased risk of developing SPC [23]. In addition, previous studies showed that the distribution of HR status in the female BC patients might be caused by racial or ethnic disparities [24, 25]. Specifically, triple-negative patients were more likely to be non-Hispanic black and Hispanic, while white women had the highest incidence of HR-positive BC subtype. As we discussed previously, black women suffered higher risk of second primary BC than white and Asian or Pacific Islanders; hence, the interaction between HR status and race may partly explain the higher risk of second BC. In summary, it is important to emphasize that in addition to the poor prognosis of HR-negative BC patients, they may also have an elevated risk of developing second primary BC, which would in turn lead to a poor survival. Other studies also examined the association between human epidermal growth factor receptor 2 (HER2) status and the risk of SPC which still remains further research. One previous study reported that those with HER2-overexpressing (ER-/HER2+) and triple-negative BC had elevated risks of developing second primary contralateral breast cancer [21]. A recent study found those with HER2- BC and without adjuvant trastuzumab therapy had higher risk of contralateral breast cancer [26]. While another study claimed that HER2 status did not seem to be a marker of risk for a SPC [27]. For second primary non-BC, it was reported that positive HER2 status increased the risk of digestive system and thyroid tumors [28]. Other pathological features, such as large tumor size, reflected an advanced tumor stage. BC patients diagnosed with advanced tumor might have lived for a longer survival time after tumor cell appeared than patients diagnosed at early stage, so they could have higher risk of recurrence or SPCs. Besides, the misclassification of SPC and metastasis also might happen.

In the current study, the effect of surgical type and radiation treatment were investigated. Lumpectomy plus radiotherapy is one of the common treatment strategies and is likely to be as effective as mastectomy. Although radiotherapy was regarded as an important risk factor for SPCs, it showed no significant effect on non-BC in the final model and less adverse effect than lumpectomy alone on developing second BC. It was reported by other studies that radiotherapy combined with surgery did not increase the risk of second primary BC or non-BC compared to surgery-alone treatment [29], and that radiotherapy after surgery reduced the long-term risk of recurrence [30]. What is more, the progress in technology can provide better radiation quality and dose conformality to target, hence better preserve the adjacent normal tissue and reduce the risk of radiotherapy-related SPCs like lung cancer and leukemia. The decreasing adverse effect of radiotherapy has been reported in serval previous studies [31, 32]. Patients treated with lumpectomy alone had the highest risk of second BC, which may imply the drawback of this surgery type that BC would come back if tumor cells were not removed completely.

The current study had strengths with respect to its methodology and model ability of identifying high-risk SPC patients. Limitations mainly came from the absence of some important variables in the SEER database, such as detailed treatment information including chemotherapy, radicality of breast-conserving surgery and so on, and factors that influence cancer incidence including smoking status, sociodemographic status and genetic factors. A moderate C-index may also result from the absence of detailed information. Besides, the quality of collected data may differ among registry sites over time. Future work will focus on external validation with other populations and perfecting our proposed model by including more covariates.

Conclusions

Important risk factors for second primary BC after initial primary BC diagnosis included age at diagnosis, black race, large tumor size, negative HR status, mixed tumor histology, localized tumor and receiving lumpectomy and radiotherapy. While the risk of second primary non-BC was less relevant to initial breast tumor characteristics and treatment. The proposed nomogram was clinically beneficial to identify patients at high risk of developing SPCs, so they may benefit from further efforts targeted at improving access to better healthcare and SPC prevention.

Materials and Methods

Supplementary Materials

Supplementary Figures