Patients and methods
With institutional review board (IRB) approval, we retrospectively reviewed the medical records of twenty-eight women who developed 39 in-situ or invasive breast cancers. These women were treated for HL between 1959 and 1999; twenty-four patients were treated at Massachusetts General Hospital, while 4 patients were treated elsewhere. All were treated for their BC at Massachusetts General Hospital between 1981 and 2005.
The original surgical pathology reports, medical and Tumor Registry records were reviewed. The details of HL treatment were reviewed [treatment modality, radiotherapy (RT) machine, RT dose, RT field, and CT regimens], as well as the mode of presentation of the index breast cancers. Pathological characteristics of breast cancers as well as tumor location within the breast were recorded. We evaluated the pathological type, T-stage, and axillary nodal status of the first tumor in patients who had bilateral disease. Treatment details of breast cancers were also collected including surgical procedure, adjuvant RT and/or systemic treatment.
Hazards estimate for metachronous bilateral BC was calculated as the number of cancers during the follow-up period divided by the total number of women-years at risk in that interval [
10]. The median follow-up after the first BC was 63.4 months (range, 8.9 to 301.7 months) with a total of 186 patient years (149 patient years after exclusion of patients with synchronous BC).
To address the treatment as well as the outcome of the index BC occurring after HL as compared to primary BC, we conducted a case-control analysis for patients with invasive tumors. We excluded from our case-control analysis all women with ductal carcinoma in-situ (DCIS) (3 patients), patients where less than 3 matches could be found in our database (2 patients), and patients with some information missing (2 patients). For each patient of the remaining 21 patients, 3 patients with BC and no history of HL were randomly selected from our database. The cases were matched for five criteria: age (within 5 years), year of diagnosis (within 5 years), tumor size, nodal positivity (0, 1 to 3, > 3) and estrogen receptors status (positive versus negative). If the exact match was not available, we relaxed the selection criteria on only four attempting to choose a comparison patient with less favorable prognostic feature (e.g. larger tumor size, etc).
As a result, 21 patients with BC after HL were compared to a group of 63 patients with primary BC. The median follow-up in the 21 patients was 62.3 months (range, 8.9 to 301.7 months) and 71.9 months (range, 3.8 to 292 months) in the control group. For patients with synchronous bilateral disease, we matched the tumor with the worst pathological features and for those with metachronous disease we matched the first BC. Both groups were compared for histological features, treatment, and outcome, including disease-free and overall survival. Exact Fisher's test was used to assess differences between the study group and the comparison group in the distribution of prognostic variables and treatment approaches. Survival curves for study and comparison groups were estimated using the Kaplan-Meier method [
11].
Discussion
The median interval between diagnosis of HL and development of BC was 16.1 years which was similar to the intervals reported in series from Stanford (17 years) [
13], Memorial Sloan Kettering Cancer Center (15 years) [
10] and Mayo Clinic (19.9 years) [
6]. Despite the long interval to develop radiation-induced BC, the median age of BC diagnosis in this cohort was 45.3 years. This concurs with results from other studies reporting that BC after treatment of HL occurs at a relatively younger age compared to primary BC [
6,
10,
13].
The majority of index breast cancers (60.7%) in our series were discovered during screening mammography. This number might be an underestimation since we were not able to determine the date of the last mammogram in individuals presenting with clinically apparent BC. Yahalom et al [
10] and Dershaw et al [
14] also reported the success of mammography in detecting 80–90% of breast cancers among their cohorts. It is quite possible these studies included patients who were better screened, or more compliant with screening. Dershaw's group compared this technique with physical examination, which revealed fewer than 40% of the tumors. However, breast self-examination or clinical examination was reported by other studies as the prevalent method of detection [
3,
13,
15]. While our study was not designed to evaluate screening, the high rate of Tis and T1 cancers found, and the high rate of mammographically detected cancers in our and other studies, highlights the importance of intensive screening.
In patients with primary BC, the reported incidence of bilateral disease is variable, ranging from 4% to 21%, the majority of cancers being metachronous [
16,
17]. Eleven of our 28 patients (39.2%) developed bilateral BC; four of whom had synchronous tumors (14.2%) and seven had metachronous tumors (25%). This rate is significantly higher than that reported in the general population and also higher than those reported by Bhatia et al (29%) [
1] and Yahalom et al (22%) [
10]. Of note, our cohort has longer median follow-up compared to these two reports (5 years versus 3 years for each). Furthermore, it should be noted that women who develop BC at young age are at an increased risk to develop contralateral disease [
18] as this may reflect more years of follow-up and smaller risk of death from other causes [
19]. In our series, the average annual hazards rate for metachronous bilateral BC (3.2%) was higher than that of primary BC (0.5 to 1%) [
20‐
24] and also higher than that reported in other studies for BC after HL (1.36% to 2.6%) [
10,
13]. Whether this higher rate of bilaterality warrants surgical prophylaxis is an open question. As second tumors seem to be detected quite early with vigilant mammographic screening, and as Magnetic Resonance Imaging (MRI) screening may allow more complete early detection, the role of prophylaxis remains a personal choice.
The majority of breast cancers (64.1%) in our patients were laterally located within the breast, with the upper outer quadrant being the most frequent location (59%). This concurs with results from other reports [
3,
6,
13,
25,
26] and is also similar to the incidence of upper quadrant tumors in primary BC (61–65%) [
27,
28]. In a study of doses delivered to the breast during mantle irradiation, unshielded upper outer quadrant appears to receive higher radiation doses compared to tissue beneath the lung block [
29]. Of interest, some authors have reported a higher incidence of medially located tumors for patients who develop BC after HL [
1,
10,
15]. Apparently, radiation-induced breast cancers following treatment for HL may occur anywhere in the breast. This might be inferred from the study reporting large dose gradient (3–42 Gy) across the breast following typical mantle treatment with a midline dose of 40 Gy [
30]. There is convincing evidence for a strongly linear radiation dose response in the lower dose range (up to 5 or 10 Gy) [
31‐
36]. Therefore, low doses of radiation delivered incidentally to any of the breast quadrants appear to be of concern. This was also confirmed in the setting of RT for BC; Stovall et al reported that women < 40 years of age who received > 1 Gy of absorbed dose to the specific quadrant of the contralateral breast had a 2.5-fold greater risk for contralateral BC than unexposed women [
37].
We evaluated the pathological type, T-stage, and axillary nodal status of the index BC in patients who had bilateral disease, as the contralateral tumors are often detected at an early stage due to intensive screening. The incidence of axillary nodes involvement in our series was 31.8%, which was similar to the 31% [
10] and 27% [
13] reported by others, and also similar to T-stage adjusted rate in primary BC [
38]. On the other hand, Cutuli al al [
39] reported higher incidence of axillary nodes involvement (62%) among their series. Data from the current study and other studies [
10,
13] reported that the histological features of BC after HL are similar to those of primary BC. Sanna et al [
40] reported the same findings with the exception of the proliferation index that showed higher rates in BC among the lymphoma group as compared to the group of primary BC.
Based on concerns about possible severe consequences arising after a high total cumulative dose to the breast, several authors [
6,
10,
13] have suggested mastectomy as the treatment of choice for BC after HL. Our case-control analysis showed that mastectomy was the predominant surgery among the lymphoma group (86%) as compared to the control group (32%). The history of previous thoracic irradiation appeared to be the reason of high rate of mastectomy in the lymphoma group, particularly if we take into account that the majority of patients had early-stage tumors. According to the difference in the surgical management, the use of adjuvant RT was significantly different between the two groups. The two patients who received RT following mastectomy did not show any radiation-related complications at their last follow-up. Nevertheless, due to paucity of data in the setting of BC after HL, the decision of RT following mastectomy should be individualized with careful outweighing of benefits and potential toxicity for each patient.
Only two patients had adjuvant RT following conservative surgery with excellent cosmetic outcome. Similarly, two studies [
41,
42] reported good to excellent cosmetic results, with follow-up of 30 and 46 months respectively, in 14 patients treated by lumpectomy and whole breast RT to doses of 46 to 50 Gy with 10 to 15 Gy boost to the tumor bed. Recently, Intra et al [
43] presented intraoperative electron beam RT following lumpectomy as an option to avoid mastectomy in six BC patients previously irradiated for HL, but the follow-up, 30 months, is still relatively short to judge the treatment outcome. On the other hand, Wolden et al [
13] reported severe soft tissue necrosis 6 years after lumpectomy and radiation (the patient was treated with tangents to 45.6 Gy and a boost of 15 Gy to the upper inner quadrant); the breast irradiation fields overlapped the prior mantle field in some regions. Overall, the small number of patients treated by a second radiation does not allow making solid conclusions, but the use of RT, and especially partial breast irradiation, warrants further investigation, particularly for women refusing mastectomy.
In the adjuvant setting, the case-control analysis showed that anthracycline-based regimens were less frequently used among the cohort of BC after HL compared to patients with primary BC. It should be noted that patients from both groups were treated at the time when the standard adjuvant CT for BC was 5'flurouracil and cyclophosphamide with either anthracylines or methotrexate. With respect to disease-free and overall survival, figures
1 and
2, respectively, show overlap of the confidence intervals indicating no significant difference between the study and the control groups at 5 and 10 years. The lack of statistical significance in presence of absolute difference of 17% in 10-year disease-free survival could be explained by the small number of patients. Furthermore, precisely because the confidence intervals on the curves are big, one should not take this difference at the face value that is the real difference could be 0% or 17% in the opposite direction. Therefore, based on our data, we could not reject the null hypothesis of similar disease-free and overall survival for the group of BC after HL and that of primary BC.
Similiary, Yahalom et al [
10] reported that the prognosis of patients with BC after HL was strongly dependent on their axillay nodal status with the survival data similar to survival information of patients with primary BC. Two other studies [
13,
39] reported the dependence of disease-free survival for BC after HL on the disease stage exactly like the primary BC.
On the other hand, Hancock et al [
3] reported that survival of BC that occurred in previously irradiated HL patients tends to be slightly lower than expected for BC in the general population. Sanna et el [
40] also reported that patients with BC after HL experienced significantly lower disease-free and overall survival; they attributed these findings to the reduced use of anthracyclines in the adjuvant treatment and/or genetic damage by previous therapies and ultimately treatment resistance. We don't think that our patients with BC after HL were undertreated in terms of adjuvant therapy. The reduced use of postmastectomy RT could be explained by the fact that the majority of this group has early-stage breast cancers with negative or less than three positive axillary lymph nodes that would have been eligible for breast-conserving surgery (as shown in the control group) despite the prior irradiation that made mastectomy the preferred option of treatment. In terms of adjuvant systemic therapy, our patients with BC after HL were treated with CT and/or hormone therapy as indicated. Although we could not exclude possible resistance to CT due to prior therapy as suggested by Sanna et al [
40], it might be difficult to determine if this resistance is limited to certain types of CT rather than others especially if we consider that anthracyclines were given only for two patients among the group of BC after HL. Collectively, it seems more reasonable, as shown by our data, that the survival in patients with BC after HL is rather linked to the known independent prognostic factors e.g. lymph node status and tumor size same as primary BC.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
MAA, KSH and AGT were involved in the initial study conception and draft writing. MAA suggested the design of the case-control analysis. AN and SIG were involved in the statistical analysis. All authors read and approved the final manuscript.