Different Patterns of Conditional Survival of Breast Cancer Patients by Age and Histologic Types: Evidence from the Korean Nationwide Registry

Breast cancer is the most common cancer in women worldwide with 1.7 million new cases in 2012 (1), the second most common cancer among females in Korea, and has increased in incidence constantly throughout the 1999–2015 (2). The 5-year and 10-year relative survival (RS) of Korean breast cancer patients was reported as 91.2% and 84.8%, respectively (3). However, these RS do not provide relevant information on the current prognosis of patients who have already survived over a period of time. In addition, RS is relatively pessimistic because they are based on the survival estimates for the patients who die during the first few years. With the high increase in the number of breast cancer survivors, we need more up-to-date and suitable survival data.

Conditional relative survival (CRS) describes the survival of patients who have already survived for a certain period of time after diagnosis (4). It could provide more useful information for personal health-related planning to cancer survivors and for optimal surveillance to oncologists (5). In previous reports on CRS among breast cancer patients (6–15), the 5-year CRS had increased from 85% to 88% at diagnosis to 92% to 95% after 5 years (8, 11). However, most of the studies did not report on detailed factors such as age at diagnosis, stage, and histologic type.

In this study, we estimated the 5-year CRS of Korean breast cancer patients for each subsequent year survived, using the population-based Korea Central Cancer Registry (KCCR) database, and evaluated the excess mortality of breast cancer patients compared with the general population. In addition, we also computed the 5-year CRS by sex, age, stage, histologic type, and time after diagnosis.

This study included Korean breast cancer patients [with International Classification of Diseases, 10th revision (ICD-10) code: C50 (16)], who were diagnosed at age 20 to 79 years during 2002–2013 and registered in the KCCR, a population-based national cancer registry that includes information on more than 98% of patients with newly diagnosed cancer in Korea (2, 17). Mortality data from 2002 to 2015 were obtained from Statistics Korea, which covered 100% of the Korean population. We merged breast cancer cases with mortality data using the 13-digit unique personal identification number, which all Koreans possess. Passive follow-up was done with the date of last follow-up as December 31, 2015. Data for the general population for the same period were also collected from the resident registration population in Statistics Korea (http://kosis.kr).

Patients were classified by sex (male, female), age (<40, 40–49, 50–59, 60–69, and ≥70 years), and year of diagnosis (2002–2005, 2006–2009, and 2010–2013). In addition, we categorized the patients by histologic type according to the classification of the World Health Organization (16). The stages at diagnosis, available from 2005 in KCCR, were classified using the Surveillance, Epidemiology, and End Results (SEER) staging system (18).

Survival duration was calculated as the difference between date of diagnosis and either date of death, loss to follow-up, or date of follow-up termination, whichever came first.

We defined RS as the observed survival among breast cancer patients divided by the expected survival of a comparable group from the general population with the same sex, age, and year of death. The expected survival was calculated using the Ederer II method, with the life table provided by Statistics Korea.

We estimated RS using flexible parametric models that were implemented using the Stata command stpm2. Modeling of RS involved fitting the models on the log cumulative excess hazard scale, which is an extension of the flexible parametric models for survival analysis that were developed by Royston and Parmar (19), as we previously described (20). H(t) denotes the total cumulative hazard, H₀(t) denotes the expected cumulative hazard, and |{\rm{\Lambda }}( t )$| denotes the excess cumulative hazard, i.e., H(t) – H₀(t). Then the log cumulative excess hazard is defined as |\ln ( {{\rm{\Lambda }}( t )} ){\rm{\ }} = {\rm{\ }}\ln ( {{{\rm{\Lambda }}_0}( t )} ) + x\beta .$| The log baseline cumulative excess hazard, |\ln ( {{{\rm{\Lambda }}_0}( t )} )$|⁠, is modeled using restricted cubic splines with knots located at k, the vector of knot position, i.e., |\ln ( {{{\rm{\Lambda }}_0}( t )} )\ = {\rm{\ s}}( {\ln ( {\rm{t}} ){\rm{|}}\Upsilon },{\rm{\ }}{k_0}} )$|⁠. Finally, we modeled the log cumulative excess hazard scale with time-dependent effect, as mentioned below:

where |{\rm{s}}( {\ln ( t ){\rm{|}}\Upsilon } )$| denotes the restricted cubic spline function of ln|{\rm{\ }}t$| with knots located at |k$|⁠, D denotes number of time-dependent effects, and |x$| covariates. In our model, we used 5 knots for restricted cubic splines and 3 df for the time-dependent effect of age. For covariates, we used sex, age at diagnosis, stage at diagnosis, histologic type, and year of diagnosis. We extrapolated 10-year RS and 5-year CRS for breast cancer patients diagnosed from 2010 to 2013 based on this model.

Conditional survival was defined as the probability of surviving an additional y years based on the condition that a patient had already survived x years. Thus, given that a cancer patient has already survived x years, the conditional survival for another y years can be expressed as: Conditional survival (y | x) = S(x + y)/S(x), where S(x) is the RS at time x. For example, to compute the 5-year CS for a breast patient who has already survived 2 years, the survival at 5 + 2 years, S(7), is divided by the survival at 2 years, S(2). In this study, based on survival data obtained during the follow-up period, the 5-year RS was calculated by each year survived up to 10 years after breast cancer diagnosis, conditional on being alive at the beginning of that year (5-year CRS).

For the evaluation of excess mortality of cancer patients compared with the general population, we classified excess mortality as follows: substantial, little/small, and hardly any for 5-year CRS <90%, 90% to 95%, and >95%, respectively (13). A 5-year CRS over 95% meant that there was no excess mortality, and the survival of breast cancer patients was then similar to that of the general population. Statistical analyses were performed using SAS 9.3 software (SAS Institute) and Stata 13 (StataCorp LP).

In total, 145,083 breast cancer patients were included in the analysis. Table 1 shows the baseline characteristics and number of patients available for CRS analysis after 2 and 5 years. Most of the patients were women (99.6%). Regarding age distribution, the patients ages 40 to 49 years at diagnosis were the most frequent (38.1%), followed by those ages 50 to 59 years at diagnosis (26.8%). Localized disease, regional metastasis, and distant metastatic disease accounted for 43.9%, 28%, and 13.1% of cases, respectively. Regarding histologic type, invasive carcinoma of no special type (NST) accounted for 89.6% of cases; lobular carcinoma, mucinous carcinoma, and metaplastic carcinoma had proportions of 3.2%, 1.7%, and 0.8%, respectively.

During the follow-up period, 19,764 patients died (13.6%). The 5-year RS was 90.8%, and the 10-year RS was 85.7% (Table 2; Fig. 1A). The patients ages 40 to 49 years at diagnosis showed the most favorable RS (93.0% for 5-year RS, 88.7% for 10-year RS). However, very young (age <40 years) and very old (age ≥ 70 years) patients had relatively poor prognosis (5-year RS of 89.7% and 80.4%, respectively). The patients with localized disease showed 97.8% 5-year RS and 95.7% 10-year RS. According to histologic type, NST showed 90.6% and 85.3% 5-year and 10-year RS, respectively; 93.1% and 86.2% for lobular type, and 82.0% and 78.0% for the metaplastic type.

Figure 1B shows the 5-year CRS for each additional year after diagnosis. The 5-year CRS was 91.0% for patients who have survived the first year and steadily increased to 94.3% after 5 years of survival, and 96% after 10 years of survival. Women had better 5-year CRS than men who have survived the 5 years (94.3% of women vs. 75.5% of men; Table 2). Very young (age <40 years) and very old (age ≥70 years) patients had worse 5-year CRS for patients who have survived the 5 years (92.2% in those aged <40 years, 92.6% in those aged ≥70 years vs. 95.4% in the 40–49 years, 94.3% in the 50–59 years, and 93.7% in the 60–69 years groups, respectively; Table 2; Fig. 2A). The 5-year CRS of localized disease patients was consistently over 97% from the time at diagnosis. In regional and distant metastatic disease, it increased from 89.8% at diagnosis to 92.1% after 5 years of survival and from 38.2% to 49.1% (Table 2).

When we evaluated the 5-year CRS by stage in each age group, the patients with localized disease had more than 95% 5-year CRS in all age groups (Fig. 3; Supplementary Table S1). However, for the patients with regional or distant metastatic breast cancer, there was a different pattern of 5-year CRS for each age group. Although the 5-year CRS did not change in the very young breast cancer patient (age <40 years) with distant metastatic disease (from 44.2% at diagnosis to 44.7% at 5 years after diagnosis), it increased from 26.9% at diagnosis to 63.6% after 5 years of survival in the very old women (age ≥70 years; Fig. 3A and E). In the women with regional metastatic disease, the 5-year CRS of in the very young breast cancer patient did not change (from 89.0% to 90.6%). However, the very old women showed increase of 5-year CRS from 77.3% to 91.0%.

Regarding histologic type, the 5-year CRS for patients with NST, the most common type, steadily increased from 90.6% at diagnosis to 94.1% at 5 years after diagnosis (Table 2). Although patients with lobular carcinoma showed better 5-year RS at diagnosis (93.1%), their 5-year CRS did not change with each year survived and was worse than that for NST after 3 years. Patients with metaplastic carcinoma showed worse 5-year and 10-year RS than patients with other types. However, their 5-year CRS increased and became better than those with NST and lobular carcinoma after 4 years (94.0% in metaplastic carcinoma, 93.5% in NST, and 92.4% in lobular carcinoma, respectively, Fig. 2B).

Although little excess mortality (5-year CRS: 90%–95%) persisted until 5 years after diagnosis among breast cancer patients, hardly any excess mortality (5-year CRS >95%) was seen beyond 7 years after diagnosis (Fig. 1B). Hardly any excess mortality was found among the patients ages 40 to 49 years after having survived for 5 years (Table 2). Little excess mortality was observed among very young patients (age <40 years) and very old patients (age ≥70 years) after 3 years and 4 years, respectively. The patients with localized disease had no excess mortality from the time of diagnosis.

Patients with mucinous and medullary breast cancer showed no excess mortality from the time of diagnosis. The patients with NST and lobular type persistently had little excess mortality and did not reach 95% of CRS until 5 years after diagnosis. Excess mortality among the patients with metaplastic carcinoma was substantial at diagnosis, but it became small after 3 years; hardly any excess mortality was observed after 5 years (Table 2).

This study showed that the 5-year CRS of breast cancer patients increased with each year survived, and there was no excess mortality after 7 years. In addition, we also identified different patterns of CRS by sex, age, stage, and histologic type among breast cancer patients.

Previous studies also reported CRS for all cancer and breast cancer patients (6–13). Table 3 summarized the CRS results of breast cancer patients from this current study and other eight studies, which have been published since 2010. In most of the studies, the 5-year CRS increased among all breast cancer patients and was more than 95% at 5 years after diagnosis in the localized stage. These tendencies were consistent with our results. In the results for excessive mortality, our study showed more favorable prognosis than other studies. Because these results could be explained by the differences of year of diagnosis and follow-up of studies, the result of our study could give the more recent and relevant information.

We found that CRS was clearly different among the age groups. Women ages 40 to 49 years, who had the highest incidence, had the most favorable CRS during follow-up and little excess mortality after 5 years. However, the 5-year CRS was worse among very young (age <40 years) and very old (age ≥70 years) breast cancer patients, for whom excess mortality was substantial at 3 and 4 years, respectively. Although the 5-year CRS increased every year in all age groups, differences in the CRS among age groups persisted over time. We also demonstrated the obviously different patterns of 5-year CRS change by age in the women with regional and distant metastatic disease. These findings could be explained by that tumor characteristics and patterns of treatments would be different causes by patients' age (10, 13). Previous studies have reported that young breast cancer patients have aggressive biological characteristics (21, 22), including higher proliferation, higher tumor grade, and estrogen receptor (ER) negativity, which resulted in worse outcomes than those in case of other age groups (23). By contrast, the worse prognosis of very old breast cancer patients could be explained by the higher rate of comorbidities and the low treatment intensity (24, 25).

Male breast cancer patients showed unfavorable prognosis and continued to have a substantial excess mortality even until 5 year after diagnosis. It could be explained that male breast cancer patients were older than female patients and had more advanced stage at diagnosis (Supplementary Table S2). Hong and colleagues reported that male breast cancer patients had older age at diagnosis, lesser use of treatment, and poor prognosis with a 62% overall survival (26). However, male breast cancer patients had similar survival as female breast cancer patients in an age-, stage-matched study using the Korean Breast Cancer Society Database (27). Because male breast cancer patients showed continuously unfavorable CRS, clinicians should take care of them and try to identify ways of improving their survival.

Our study identified different patterns of 5-year CRS by histologic type in breast cancer. Among them, lobular carcinoma was the second most common type of breast cancer, with largely ER-positive, HER2-negative, and low-to-intermediate tumor grade (28), and had a favorable prognosis (29, 30). On the contrary, metaplastic carcinoma has been known as a clinically aggressive subtype of breast cancer with triple-negative phenotype and unfavorable prognosis (31, 32). This study showed that the 5-year CRS of lobular carcinoma and metaplastic carcinoma was reversed during the follow-up period, although CRS at the diagnosis of lobular carcinoma was better than that of metaplastic carcinoma. In addition, patients with metaplastic carcinoma showed hardly any excess mortality at 5 years after the diagnosis during the follow-up period. It could be explained that lobular carcinoma had a tendency for late recurrence (33), and metaplastic carcinoma showed poor prognosis because of early recurrence and death (32, 34, 35). These findings might be evidence that different surveillance methods and durations of endocrine treatment have to be tailored according to the histologic type among breast cancer patients.

The strengths of this study lie in the analysis of a population-based, nationwide cancer registry and our estimation of CRS according to important prognostic factors, including sex, age, stage at diagnosis, and histologic type. To the best of our knowledge, this is the first nationwide study on CRS according to age, stage, and each histologic type of breast cancer.

This study had some limitation in that we could not take into account tumor characteristics (hormone receptor status and HER2 status) because they were not available in the KCCR. However, we evaluated CRS by histologic type. This could provide us with some information regarding CRS according to tumor characteristics because it is well known that mucinous carcinoma and lobular carcinoma tend to be hormone receptor positive and HER2 negative, whereas metaplastic carcinoma is predominantly hormone receptor negative and HER2 negative. In this study, treatment options for breast cancer were not considered due to the lack of relevant information. Because the treatment could affect cancer-related survival, second cancer incidence, and other life-threatening complications, we need to perform more detailed analyses considering treatment options in future. This study population's follow-up period was not enough to get the 5-year CRS of 10-year survivors. So, we adapted parametric survival model called flexible survival model to estimate CRS in this study. Parametric survival modeling has advantages for prediction, extrapolation, quantification, and complex model such as time-dependents effects. So we could estimate CRS for the 5-year survivals of breast patients diagnosed in 2010–2013. However, these results should be interpreted with caution as the estimates in this study are model-based; therefore, these results depended on the parametric assumptions.

The 5-year CRS for breast cancer patients generally increased with each year survived, and they had hardly any excess mortality after 7 years. In addition, we demonstrated the different CRS patterns according to the age of breast cancer patients and histologic type. These results may be useful to patients in planning their remaining lives as well as to aid clinicians in planning breast cancer surveillance.

No potential conflicts of interest were disclosed.

Conception and design: S.-Y. Jung, K.-W. Jung, Y.A. Kim, E.S. Lee

Development of methodology: S.-Y. Jung, K.-W. Jung, J. Ha, Y.A. Kim

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K.-W. Jung, J. Ha, Y.-J. Won, E.S. Lee

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S.-Y. Jung, K.-W. Jung, J. Ha, Y. Kwon, E.S. Lee

Writing, review, and/or revision of the manuscript: S.-Y. Jung, K.-W. Jung, J. Ha, Y.-J. Won, Y.A. Kim, Y. Kwon, S.-Y. Kong, E.S. Lee

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Ha, E.S. Lee

Study supervision: S.-Y. Jung, K.-W. Jung, E.S. Lee

This work was supported by a grant from the National Research and Development Program for Cancer Control, Ministry of Health and Welfare, Republic of Korea (no. 1520240), and the National Cancer Center (NCC-1610202).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.