Prediction of outcome after diagnosis of metachronous contralateral breast cancer

Within their lifetime, 2-20% of breast cancer patients develop a new tumour in their contralateral breast [1‐3]. These contralateral breast cancers (CBC) are called synchronous if the second tumour (BC2) develops within a short time interval from the first tumour (BC1), and metachronous if the time interval between tumours is longer. In line with several previous studies, we define metachronous tumours as CBC diagnosed at least three months after BC1 [3‐5]. However, a clear cut-off time is not defined in the literature. CBC is today treated as a new primary tumour (two individual tumours), but the biological relationship between BC1 and BC2, and the impact of a second primary tumour on prognosis is debated [4, 6‐18]. Previous studies indicate that prognosis after CBC could be associated with age, time interval between BC1 and BC2, mode of detection of BC2, and adjuvant treatment for BC1 [4, 15‐17, 19]. However, despite women with a history of breast cancer have a high lifetime risk of developing CBC, the annual risk remains at a relative low level of 0.5-1%. A long follow-up time is hence needed in order to obtain a large cohort of patients with CBC.

For this study data was abstracted from individual charts for all patients diagnosed with metachronous CBC in the Southern Healthcare Region of Sweden (a region with 1.7 million inhabitants) from 1977 to 2007. This gave us a unique cohort, including more than 700 patients from multiple medical centres, providing information on patient and tumour characteristics, treatment, and outcome. The aims of this study were to examine prognosis after CBC in relation to time interval between BC1 and BC2, mode of detection of BC2, and treatment for BC1.

Using a large population-based cohort, including patient, tumour, and treatment information, we have studied prognosis after CBC in relation to time interval to, and mode of detection of BC2. A short time interval to BC2 was proven to be a significant negative prognostic factor, supporting earlier studies [4, 15‐17]. However sub-group analysis indicates that this effect is seen only for younger women. Data additionally indicate that women diagnosed by routine follow-up examination, have a significantly lower risk of developing metastases at a later stage.

In line with our data, previous studies have found a short time interval to BC2 to be associated with an impaired prognosis [4, 15‐17]. This study includes extended individual data from a large population-based cohort and strongly validate that a short time interval to BC2 is a negative prognostic factor. Time interval to BC2 was analysed both as a continuous and as a dichotomised variable with a cut-point of three years, and yielded independent prognostic information by both methods. Furthermore, previous results have differed when comparing prognosis after synchronous, metachronous, and unilateral breast cancer [10]. However, there have been indications that synchronous breast cancer (defined as CBC diagnosed within 3-12 months after BC1) might result in a worsened prognosis compared to metachronous breast cancer [10, 11, 18].

The underlying cause for the impaired prognosis observed for CBC diagnosed within a short time interval from BC1 is unclear. One potential explanation could be that these tumours more often represent a metastatic spread of BC1. Comparisons of genetic alternations in bilateral breast cancer have shown that although most CBCs represent a new primary tumour, contralateral spreading from BC1 does occur [6‐9]. A short time interval to BC2 has also been found to correlate with increased genetic and morphological similarities between bilateral tumours [21, 22]. This could suggest a higher prevalence of metastatic spread, but could also be a result of these tumours having developed in a similar biological environment. Imyanitov et al. found the highest correlation between tumours in women who developed both tumours while premenopausal [21]. If this correlation results from a higher percentage of contralateral metastatic spread in these patients, it could partially explain the different effects of time interval to BC2 observed for the different age categories in our study, where most patients (95%) diagnosed when under 50 years of age were premenopausal. However, Imyanitov et al. observed the lowest correlation between bilateral tumours separated by menopause, making hormonal environment at the time of development another likely cause to similarities vs. differences observed [21].

Another potential explanation could be that CBC diagnosed during, or soon after, adjuvant treatment has developed resistance to treatment and a more aggressive phenotype. The incidence of CBC in Sweden has decreased since the early 1980s [4]. This observation is potentially resulting from an increased use of adjuvant treatment, reducing the risk of CBC [2, 3, 23‐25]. However, during the same time period mortality increased for those women who did develop CBC [4], which could reflect a treatment-escape phenomenon once therapy has failed to prevent a second tumour. CBC developed after tamoxifen treatment is also ER-negative to a larger extent [2, 26‐28], indicating that treatment given for BC1 affect the biology of BC2. If younger patients have received more adjuvant treatment, this might hence be one explanation to why the time interval between tumours is of greater importance in younger patients. In the present cohort we found support for this theory in that BC1 in younger patients was more often treated with radiotherapy and chemotherapy, while older women received more endocrine therapy (data not shown).

In this study chemotherapy given for BC1 was an independent negative prognostic factor, though this was not seen if chemotherapy had been given for BC2. Adjuvant endocrine treatment after BC1, however, had no effect on prognosis after BC2. Radiotherapy and endocrine treatment were positive prognostic factors if given after BC2, but not if given after BC1. Similar results have been seen in a study by Hartman et al and this could be interpreted as an indication that a CBC diagnosed after prior chemotherapy is more aggressive [4]. However, choice of adjuvant therapy is strongly dependent on tumour stage and biology, and patients selected to receive chemotherapy are those with the worst predicted prognosis. The survival analysis is adjusted for several prognostic factors including tumour stage and all treatment given. There are however still other factors, affecting prognosis and choice of therapy (hormone receptor status, histological grade, etc.) which are not provided for all patients in the present study, including patients for several decades. Hence, although these results are interesting and in line with those found by Hartman et al [4], they need to be interpreted with caution and further studies are required. The effect on prognosis by adjuvant endocrine treatment for BC1 is not fully elucidated since the cohort was not stratified according to hormone receptor status. Future studies including biomarker analysis of BC1 and BC2 can hopefully shed some light on this issue and, more importantly, on the effect of adjuvant treatment after BC1 on tumour biology of BC2.

Although early detection of distant metastases does not affect survival or quality of life [29, 30], recent studies suggest that early detection of local recurrences or CBC does improve prognosis [19, 31, 32]. However, results are not unanimous [33‐35]. Previous studies are retrospective, leading to problems with lead-time (earlier detection results in a longer follow-up until event, even if the disease progression is the same) and length-time bias (slower growing tumours will be more easily detected, since they are detectable over a longer time period). To avoid lead-time bias we have looked at the risk of metastasis, using logistic regression, instead of DDFS with regard to mode of detection. Length-time bias may still be a problem in our study, although previous studies have not shown this to be a major source of concern [32, 36‐38]. We found that significantly fewer patients diagnosed with CBC within follow-up examinations subsequently developed metastases, which is probably to a large extent due to the fact that the tumours were discovered at an earlier stage. However, mode of detection remained a significant prognostic factor even after adjusting for tumour size, node status, and treatment for both tumours. This finding is in line with data from unilateral breast cancer, where diagnosis by screening mammography has been shown to be an independent prognostic factor even after adjustment for disease stage [39‐42].

Presuming that surveillance after breast cancer diagnosis is effective, how long should it continue? In Sweden, a follow-up time of ten years was advocated until the mid-1990s. Today, clinical surveillance is often reduced to five years or less, followed by mammographic surveillance every 24 months within a screening program. In this study we found mode of detection to be associated with risk of metastasis even when BC2 was diagnosed more than ten years after BC1, suggesting that a long follow-up time could be of value.

The present study was based on a unique cohort, including over 700 patients with CBC, with information on patient and tumour characteristics, treatment, and outcome. Despite the large selection of patients and information included in this study, some potential sources of bias should be considered. For example, subgroup analysis will include fewer events. Inclusion in the cohort was based on data from the Swedish Cancer Register which is nationwide. However, some cases of CBC might still be missing or misclassified, and 150 charts were never found. Although data from the patients' individual charts is probably more reliable than register data, the quality of clinical notes and pathological records varied over time and between physicians. Due to the low annual incidence of contralateral breast cancer, a long period of follow-up was needed to receive a large cohort. Hence, standard surgical methods, routine histopathological analysis and adjuvant treatment have changed during the study period. The patients were included in the cohort based on diagnose date of the CBC. Hence, for women initially treated before 1977, there could be a selection of patients with a longer time interval to BC2. When adjusting our analyses for diagnose date of BC1 the multivariate analysis of time interval to BC2 including all patients and treatment was no longer significant. Otherwise significance in all other analysis remained constant.

Patients with CBC are currently treated according to the tumour biology of BC1 and BC2 individually, and it is controversial whether prognosis after CBC is worse than after unilateral breast cancer [4, 10‐18]. Our study was based on clinical data from individual patients and results indicate that the time interval to BC2, especially for patients younger than 50 years, is a strong prognostic factor. Additionally, mode of detection was proven to be closely related to the risk of developing metastases. By taking time interval to BC2 and mode of detection into account when diagnosing patients with CBC, this study seems to have identified patients with a poor prognosis. Indeed, among patients diagnosed with symptomatic CBC within three years from BC1, more than 50% later developed metastases. Further translational studies of the tumour biology of BC2 in relation to time interval, adjuvant treatment after BC1 and mode of detection are needed to confirm and explain these results.