Concordance in histological and biological parameters between first and second primary breast cancers

Authors


Abstract

BACKGROUND:

Women with breast cancer are more likely to have a second breast cancer than women in the general population are to have a primary cancer. However, the biological relationship between primary and second breast cancers is not clear.

METHODS:

A total of 30,617 patients diagnosed with bilateral breast cancers between 1990 and 2007 were identified through 17 cancer registries of the Surveillance, Epidemiology, and End Results program. Logistic regression with odds ratios (ORs) and 95% confidence intervals (CIs) was used to model strength of association in hormone receptor status, grade, and histology between 2 cancers.

RESULTS:

There was a strong association in estrogen receptor status between 2 bilateral tumors (OR, 7.64; 95% CI, 7.00-8.35). The strength of association in estrogen receptor status depended on the time interval between the first and second tumors and age at diagnosis. The OR was 25.9 for synchronous tumors (within 1 month) and 3.69 for metachronous tumors separated by ≥10 years. The strength of association was stronger in patients whose first cancer was diagnosed before age 50 (OR, 11.7) versus after age 50 (OR, 5.71). A similar pattern was observed for progesterone receptor, grade, and histological type, but with relatively weaker association.

CONCLUSIONS:

The strong concordance in hormone receptor status of primary and second breast cancers suggests that 2 breast cancers arise in a common milieu and that tumor subtypes are predetermined in the early stage of breast carcinogenesis. Cancer 2011. © 2010 American Cancer Society.

Women with breast cancer have a more than 2-fold higher risk of developing a second breast cancer than the risk women in the general population have of developing a primary breast cancer.1, 2 However, the biological relationship between the 2 breast cancers is not well understood. It is unclear whether the second cancer represents an independent second primary tumor or a sequential event of a primary tumor. Even if the 2 cancers develop from different origins, they may not truly be independent, because they are subjected to similar hormonal, environmental, and genetic influences. Therefore, understanding the relationship between the 2 tumors has implications in both cancer treatment and understanding of breast carcinogenesis.

Previous studies have evaluated the concordance in hormone receptors between primary and contralateral breast cancers occurring in the same patient.3-11 Many, but not all, studies found a positive association in the level (or status) of estrogen receptor (ER) and progesterone receptor (PR) between the first primary and contralateral breast cancers. However, the degree of association has not been defined. Several factors may affect this association, including time interval between 2 cancers and tamoxifen treatment. Coradini et al8 reported that there was a positive correlation in levels of ER and PR between primary and contralateral breast cancers, and that the correlation for ER levels, but not PR levels, was higher in synchronous than in metachronous breast cancer. Gong et al9 found that the concordance in PR status was higher in synchronous than in metachronous cancers. Swain et al10 showed that ER status of primary breast cancer was predictive of ER status of contralateral breast cancer among patients who did not receive tamoxifen, but the association was not statistically significant in patients receiving tamoxifen. In contrast, Arpino et al6 showed that the ER and PR status of the primary breast cancer did not predict hormone receptor status of contralateral breast cancer among patients treated with or without tamoxifen. The inconsistencies in these studies are probably due to limited sample size, because the sample sizes ranged from several dozen to several hundred.

Using the large dataset from the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) registries, we conducted this study to provide a more solid answer regarding the relationship between hormone receptor status of primary and second breast cancers. Specifically, the goals of the study were to 1) assess the association in morphological and pathological features (histological type and tumor grade) and molecular markers (ER and PR) between primary and second breast cancers; 2) examine the factors that modify these associations; and 3) compare the proportions of ER and PR positivity between primary and second breast cancers.

MATERIALS AND METHODS

Since 1990, hormone receptor status has been reported to the National Cancer Institute's SEER database. To obtain precise estimates of association and sufficient power for subgroup analysis, we gathered data from all 17 SEER cancer registries, which currently represent 26% of the United States population.12 Using a unique identifier assigned to each patient, we identified 33,587 patients who had diagnoses of bilateral breast cancers between 1990 and 2007. The majority of these patients had 2 diagnoses of breast cancer, and 856 (2.5%) had 3 or more diagnoses of breast cancer. For simplicity of the analysis, we only included the first 2 diagnoses for patients who had more than 2 diagnoses. To avoid the potential metastases to the contralateral breast in patients with stage 4 disease, we excluded 3219 patients who had distant metastasis in either of the 2 cancers and 651 patients who had missing data in metastasis status. The remaining 30,617 bilateral breast cancer patients were included in the analysis.

In the present study, 4 pathological characteristics were compared between the first primary and second primary breast cancers. These characteristics include histological type, tumor grade, ER status, and PR status. According to ICD-O-3 histology codes, histological type was grouped into 9 categories: ductal (8500, 8523), lobular (8520, 8522, 8524), comedocarcinoma (8501), mucinous (8480, 8481), papillary (8503, 8504, 8260, 8050, 8051), tubular (8211), inflammatory (8530), medullary (8510, 8512, 8513, 8514), and others. We also further collapsed histological type into ductal, lobular, and others. Tumor grade in SEER was categorized into 3 levels: well-differentiated, moderately differentiated, and poorly differentiated or undifferentiated. ER and PR status were recorded as positive versus negative.

We explored the following variables as potential factors that modify the concordance in pathological markers: time interval between the first and second primary breast cancers (<1 month was considered synchronous), age at first diagnosis of breast cancer, year at first diagnosis of breast cancer, stage of first breast cancer (SEER stage: in situ, localized, and regional), surgery for the first breast cancer (no, lumpectomy, mastectomy), radiation therapy for the first breast cancer, and race (white, black, Asian/Pacific Islander, Native American).

Logistic regression with odds ratios (ORs) and 95% confidence intervals (CIs) was used to assess the strength of association in hormone receptor status between 2 cancers. An OR >1 indicated a positive association. Specifically, the binary ER status of the second cancer was the dependent variable, and ER status of the first cancer was the independent variable. To examine the factors that modify the associations, a main effect of a modifier (eg, age group) and an interaction of the modifier and ER status of the first cancer were added in the model as independent variables. For synchronous cases, 1 cancer was randomly chosen as the first cancer. Because the OR is a symmetric metric for the strength of association, the order of the first and second cancers was inconsequential.

The strength of association in histological type and tumor grade can also be measured using the OR, but there are multiple ORs because they are nominal or ordinal variables. Using the 3-category histological type as an example, we can model ductal versus lobular and other types, lobular versus ductal and other types, and lobular and ductal versus other types in 3 separate binary logistic regressions, yielding 3 separate ORs. Because a single, global OR quantifying the overall strength of an association is desired, we used a generalized estimating equation model with a logit link to estimate the overall associations between histological type and between tumor grade.13 This model essentially combined ORs from separate binary logistic regression models listed above. To calculate standard errors, we specified independent working correlation and used the robust (sandwich) variance estimator.

Conditional logistic regression was used to examine the change in proportion of hormone receptor positivity. Only metachronous cases were included in the analysis. We modeled the proportion of positive hormone receptor status as the function of time interval between the first primary and second breast cancer. ORs and 95% CIs were calculated from the model, with OR<1 indicating that the hormone receptor positive proportion decreases. Statistical analysis was conducted using Stata 11.1 software (StataCorp, College Station, TX).

RESULTS

Of 30,617 patients who had bilateral breast cancers between 1990 and 2007, almost all were female (99.7%). The race distribution was 85.0% white, 8.0% black, 6.6% Asian/Pacific Islander, and 0.3% Native American. The average age was 60.5 years at first breast cancer diagnosis and 63.1 years at second diagnosis. For 8787 (28.7%) patients, the tumors presented synchronously (within 1 month of each other). Among metachronous cases, the median time interval between 2 cancers was 31 months (interquartile range, 6-70 months). After the first breast cancer, 51.2% of patients underwent mastectomy, 47.5% underwent lumpectomy, and 41.4% received radiation therapy.

Table 1 cross-tabulates ER status of the first and second cancers according to time interval between 2 cancers, age, and stage of the first cancer. There was a very strong association in ER status (OR, 7.64; 95% CI, 7.00-8.35) between the primary and contralateral breast cancer. This association corresponds to 81% overall agreement: for patients in whom the first cancer was ER-positive, the chance that the second cancer was ER-positive was 87.5%, compared with 47.8% for patients in whom the first cancer was ER-negative. The strength of association in ER status depended on the time interval between the first and second tumors (P < .0001). As illustrated in Figure 1, the strength of association decreased as the time interval between 2 tumors increased: the OR was 25.9 for synchronous tumors and 3.69 for metachronous cancers separated by ≥10 years. Interestingly, the strength of association in ER status between bilateral cancers did not vary after 12 months (P = 0.34). The association in ER status also depended on age at diagnosis in patients with bilateral breast cancers (P < .0001): the OR was 11.7 and 5.71 in patients in whom the first breast cancer was diagnosed before and after 50 years of age, respectively. This modification effect by age at diagnosis remained significant after adjusting for time interval between 2 cancers (P < .0001). We also examined whether year at first diagnosis, tumor stage, surgery, radiation therapy, and race modify the association in ER status and found that none of these factors was a significant effect modifier after adjustment for age and time interval.

Figure 1.

The association of hormone receptor status between 2 bilateral breast cancers is shown.

Table 1. Association of Estrogen Receptor Status in Patients With 2 Primary Breast Cancers
 ER Status of First CancerER Status of Contralateral Cancer
Negative, No. (%)Positive, No. (%)OR (95% CI)
  1. ER indicates estrogen receptor; OR, odds ratio; CI, confidence interval.

All patientsNegative1608 (52.2)1474 (47.8) 
 Positive1589 (12.5)11133 (87.5)7.64 (7.00-8.35)
Time interval
 <1 moNegative420 (60.7)272 (39.3) 
 Positive262 (5.6)4387 (94.4)25.9 (21.2-31.5)
 1-5 moNegative190 (56.9)144 (43.1) 
 Positive105 (5.2)1913 (94.3)24.0 (17.9-32.2)
 6-11 moNegative63 (56.2)49 (43.8) 
 Positive87 (17.3)416 (82.7)6.15 (3.96-9.54)
 12-59 moNegative512 (50.0)511 (50.0) 
 Positive646 (22.1)2273 (77.9)3.53 (3.03-4.10)
 ≥60 moNegative423 (45.9)498 (54.1) 
 Positive489 (18.6)2144 (81.4)3.72 (3.17-4.38)
 P for interaction   <.0001
Age at first cancer
 <50 yNegative751 (66.9)372 (33.1) 
 Positive350 (14.7)2030 (85.3)11.7 (9.90-13.9)
 ≥50 yNegative857 (43.7)1102 (56.3) 
 Positive1239 (12.0)9103 (88.0)5.71 (5.13-6.36)
 P for interaction   <.0001
Stage of first cancer
 In situNegative115 (48.9)120 (51.1) 
 Positive106 (9.4)1016 (90.6)9.19 (6.64-12.7)
 LocalizedNegative952 (51.3)905 (48.7) 
 Positive1006 (12.3)7177 (87.7)7.50 (6.71-8.40)
 RegionalNegative541 (54.6)449 (45.4) 
 Positive477 (14.0)2940 (86.0)7.43 (6.34-8.70)
 P for interaction   .49

There was also a strong association in PR status between 2 tumors (OR, 4.22; 95% CI, 3.92-4.55) for contralateral pairs (Table 2). The strength of association decreased as the time interval between 2 tumors increased (P < .0001) (Fig. 1): the OR was 12.3 for synchronous cases and 1.87 for metachronous cancers separated by 10 years or longer. The strength of association in PR status between bilateral cancers did not vary after 12 months (P = 0.28). The association in PR status between 2 bilateral tumors was stronger among patients aged <50 years versus patients aged ≥50 years (P < .0001). These results paralleled those seen for ER status, except that the magnitude of associations was weaker for PR status.

Table 2. Association of Progesterone Receptor Status in Patients With 2 Primary Breast Cancers
 PR Status of First CancerPR Status of Contralateral Cancer
Negative, No. (%)Positive, No. (%)OR (95% CI)
  1. PR indicates progesterone receptor; OR, odds ratio; CI, confidence interval.

All patientsNegative2421 (56.7)1848 (43.3) 
 Positive2554 (23.7)8230 (76.3)4.22 (3.92-4.55)
Time interval
 <1 moNegative724 (61.6)452 (38.4) 
 Positive458 (11.5)3511 (88.5)12.3 (10.5-14.3)
 1-5 moNegative316 (56.0)248 (44.0) 
 Positive211 (12.4)1494 (87.6)9.02 (7.24-11.2)
 6-11 moNegative98 (56.0)77 (44.0) 
 Positive140 (34.4)267 (65.6)2.43 (1.69-3.49)
 12-59 moNegative729 (56.4)541 (42.6) 
 Positive937 (38.2)1515 (61.8)2.18 (1.90-2.50)
 ≥60 moNegative454 (51.1)530 (48.9) 
 Positive808 (35.9)1443 (64.1)1.87 (1.61-2.16)
 P for interaction   <.0001
Age at first cancer
 <50 yNegative813 (68.9)367 (31.1) 
 Positive517 (23.8)1652 (76.2)7.08 (6.04-8.29)
 ≥50 yNegative1608 (52.1)1481 (47.9) 
 Positive2037 (23.6)6578 (76.4)3.51 (3.22-3.82)
 P for interaction   <.0001
Stage of first cancer
 In situNegative186 (58.3)133 (41.7) 
 Positive146 (15.5)798 (84.5)7.64 (5.75-10.2)
 LocalizedNegative1443 (54.6)1202 (45.4) 
 Positive1650 (23.8)5293 (76.2)3.85 (3.50-4.23)
 RegionalNegative792 (60.7)513 (39.3) 
 Positive758 (26.2)2139 (73.8)4.36 (3.79-5.00)
 P for interaction   <.0001

There was a significant concordance in tumor grade between the first and second breast cancers (global OR, 2.80; 95% CI, 2.65-2.95) for bilateral pairs (Table 3). The association in grade also depended on the time interval between 2 tumors (Fig. 2, P < .0001). There was a clear trend of global OR decreasing as the time interval between 2 tumors increased (P < .0001) but stabilized after 12 months (global OR, ≈2.3; P = 0.46). Similar to results for hormone receptors, the association in tumor grade between 2 bilateral tumors was stronger among patients aged <50 years than among patients aged ≥50 years (P < .0001). The strength of association in tumor grade between 2 bilateral tumors was stronger among patients with metastatic disease than among other patients.

Figure 2.

The association of tumor grade and histological type between 2 bilateral breast cancers is shown.

Table 3. Association of Tumor Grade in Patients With 2 Primary Breast Cancers
 Grade of First CancerGrade of Contralateral Cancer
I, No. (%)II, No. (%)III, No. (%)GOR (95% CI)
  1. GOR indicates global odds ratio; CI, confidence interval

All patientsI1650 (40.4)1633 (40.0)803 (19.7) 
 II1926 (22.8)4331 (51.2)2203 (26.0) 
 III991 (14.7)2273 (33.8)3456 (51.4)2.80 (2.65-2.95)
Time interval
 <1 moI655 (45.2)577 (39.8)217 (15.0) 
 II544 (20.1)1624 (60.0)538 (19.9) 
 III208 (12.2)531 (31.2)963 (56.6)3.48 (3.21-3.76)
 1-5 moI368 (56.5)217 (33.3)66 (10.1) 
 II414 (29.2)777 (54.7)229 (16.1) 
 III181 (19.7)356 (38.8)381 (41.5)3.12 (2.81-3.47)
 6-11 moI58 (33.9)74 (43.3)39 (22.8) 
 II111 (28.2)178 (45.2)105 (26.6) 
 III73 (23.3)102 (32.6)138 (44.1)1.96 (1.61-2.38)
 12-59 moI339 (32.0)450 (42.5)269 (25.4) 
 II487 (21.5)1011 (44.7)762 (33.7) 
 III294 (13.9)693 (32.8)1124 (53.2)2.44 (2.24-2.66)
 ≥60 moI230 (30.4)315 (41.6)212 (28.0) 
 II370 (22.0)741 (44.1)569 (33.9) 
 III235 (14.0)591 (35.3)850 (50.7)2.23 (2.02-2.45)
 P for interaction    <.0001
Age at first cancer
 <50 yI222 (36.4)258 (42.3)130 (21.3) 
 II308 (18.9)821 (50.3)503 (30.8) 
 III205 (9.4)618 (28.2)1369 (62.5)3.60 (3.27-3.96)
 ≥50 yI1428 (41.1)1375 (39.6)673 (19.4) 
 II1618 (23.7)3510 (51.4)1700 (24.9) 
 III786 (17.4)1655 (36.6)2087 (46.1)2.44 (2.31-2.59)
 P for interaction    <.0001
Stage of first cancer
 In situI204 (40.1)195 (38.3)110 (21.6) 
 II273 (22.3)647 (52.9)302 (24.7) 
 III206 (16.9)478 (39.2)534 (43.8)2.36 (2.11-2.64)
 LocalizedI1163 (39.3)1210 (40.9)584 (19.7) 
 II1155 (22.8)2564 (50.7)1341 (26.5) 
 III457 (14.0)1064 (32.6)1740 (53.4)2.82 (2.65-3.00)
 RegionalI283 (45.6)228 (36.8)109 (17.6) 
 II498 (22.9)1120 (51.4)560 (25.7) 
 III328 (14.6)731 (32.6)1182 (52.7)3.05 (2.79-3.33)
 P for interaction    .0008

Table 4 cross-tabulates 3-category histological type between the primary and second breast cancer. There was a significant concordance in histological type (global OR, 3.04; 95% CI, 2.91-3.18). The association in histological type depended on the time interval between 2 tumors (P < .0001) (Fig. 2): it was higher in tumors separated within 6 months but was stable after 12 months (global OR, ≈2.3; P = 0.39). Furthermore, the association in histological type was stronger among patients aged <50 years than among patients aged ≥50 years (P = .02). The strength of association in histological type between 2 bilateral tumors was stronger among patients with metastatic disease at the first diagnosis than among other patients. The results for 9-category histological type were similar (results not shown).

Table 4. Association of Histological Type in Patients With 2 Primary Breast Cancers, SEER 1990-2007
 Histology of First CancerHistology of Contralateral Cancer
Ductal, No. (%)Lobular, No. (%)Other, No. (%)GOR (95% CI)
  1. GOR indicates global odds ratio; CI, confidence interval.

All patientsDuctal13368 (69.6)3261 (17.0)2578 (13.4) 
 Lobular2729 (39.9)3430 (50.1)686 (10.0) 
 Other2696 (59.1)798 (17.5)1071 (23.5)3.04 (2.91-3.18)
Time interval
 <1 moDuctal4013 (74.9)772 (14.4)573 (10.7) 
 Lobular828 (36.5)1255 (55.3)186 (8.2) 
 Other593 (51.1)187 (16.1)380 (32.8)5.07 (4.66-5.51)
 1-5 moDuctal1751 (60.5)705 (24.4)436 (15.1) 
 Lobular486 (27.6)1104 (62.8)168 (9.6) 
 Other273 (46.8)156 (26.8)154 (26.4)3.23 (2.90-3.60)
 6-11 moDuctal577 (65.1)170 (19.2)139 (15.7) 
 Lobular147 (39.3)182 (48.7)45 (12.0) 
 Other134 (55.8)51 (21.3)55 (22.9)2.40 (1.99-2.89)
 12-59 moDuctal3808 (68.0)919 (16.4)874 (15.6) 
 Lobular698 (48.7)558 (38.9)178 (12.4) 
 Other891 (63.6)205 (14.6)305 (21.8)2.22 (2.06-2.41)
 ≥60 moDuctal3219 (72.0)695 (15.5)556 (12.4) 
 Lobular570 (56.4)331 (32.8)109 (10.8) 
 Other805 (68.2)199 (16.9)177 (15.0)2.31 (2.12-2.52)
 P for interaction    <.0001
Age at first cancer
 <50 yDuctal3249 (69.8)835 (17.9)571 (12.3) 
 Lobular566 (35.4)899 (56.3)132 (8.3) 
 Other645 (59.9)204 (18.9)228 (21.2)3.33 (3.05-3.63)
 ≥50 yDuctal10119 (69.5)2426 (16.7)2007 (13.8) 
 Lobular2163 (41.2)2531 (48.2)554 (10.6) 
 Other2051 (58.8)594 (17.0)843 (24.2)2.96 (2.82-3.11)
 P for interaction    .02
Stage of first cancer
 In situDuctal2272 (67.4)649 (19.2)452 (13.4) 
 Lobular577 (36.7)859 (54.7)135 (8.6) 
 Other1053 (58.2)339 (18.7)417 (23.1)2.82 (2.59-3.08)
 LocalizedDuctal7722 (69.9)1831 (16.6)1498 (13.6) 
 Lobular1410 (43.1)1497 (45.7)366 (11.2) 
 Other1327 (59.9)370 (16.7)518 (23.4)2.86 (2.69-3.03)
 RegionalDuctal3374 (70.5)781 (16.3)628 (13.1) 
 Lobular742 (37.1)1074 (53.7)185 (9.2) 
 Other316 (58.4)89 (16.5)136 (25.1)3.79 (3.46-4.15)
 P for interaction    <.0001

Figure 3 presents the number of patients whose hormone receptor status changed from first to second breast cancer. The comparison of the number of patients whose hormone receptor status changing from positive to negative and the number changing from negative to positive is equivalent to the comparison of hormone receptor proportion between 2 time points. Both the proportions of ER and PR positivity were the same between 2 cancers separated by <1 year. ER positivity decreased by 21% (OR 511/646, 0.79; 95% CI, 0.70-0.89) in second contralateral breast cancers occurring 12-59 months after the first tumors and was the same between the first tumors and tumors occurring after 60 months. PR positivity decreased by 42% (OR 541/937, 0.58; 95% CI, 0.52-0.64) in second contralateral breast cancers occurring 12-59 months after the first tumors, and this reduction remained significant even for contralateral cancers occurring after 120 months.

Figure 3.

The number of patients whose estrogen receptor (ER) and progesterone receptor (PR) status changed between 2 bilateral breast cancers is shown. The numbers above the bars represent odds ratios for hormone receptor positivity (an odds ratio <1 represents a reduction). *P < .05.

DISCUSSION

This study investigated a large cohort of patients diagnosed with 2 breast cancers within a national database and demonstrated a strong concordance in both ER and PR status between primary and second breast cancers. The results are consistent with those reported in hospital-based series3, 4, 7-10 and hereditary breast cancer patients.5 Weitzel et al5 showed that the ORs for ER status between primary and contralateral breast cancers were 6.4 for BRCA1 carriers and 9.5 for BRCA2 carriers, similar to the OR of 7.64 for bilateral breast cancers in our study. Our study also confirmed previous findings that the concordance in ER was stronger in synchronous cancers than in metachronous cancers.3, 5, 8, 11 Due to the large sample size, this study depicts the relationship between the degree of concordance in ER status and the time interval between 2 tumors: the association in ER status between 2 tumors decreased as the time interval between 2 tumors increased. Although this association leveled off after 12 months, it was still highly significant even for 2 tumors separated by 10 years. Our observation of a continuous relationship suggests that there is no biological cutoff point for synchronous breast cancer; however, we consider 12 months an appropriate diagnostic period for a synchronous breast cancer. It is noteworthy that previous studies have used different diagnostic periods for a synchronous breast cancer (0-24 months).3, 5, 8, 11 We also demonstrated that the strength of association in ER status was stronger among patients aged <50 years. Very similar patterns were observed for PR status, but the strength of association for PR status was weaker than for ER status overall and in each stratum, reflecting that PR is an ER-regulated gene and that its production depends on estrogen and ER.14

This study also found significant concordance in tumor grade and histological type between the primary cancer and second breast cancer. However, the strength of association for these 2 morphological characteristics was weaker than that for hormone receptors. Similar to hormone receptor, time interval between 2 tumors and age at diagnosis modified the association in tumor grade.

We found that surgery and radiotherapy had no effect on the association between the 4 histopathological characteristics examined in this study. A previous study reported that tamoxifen treatment attenuated the association in ER status between primary and contralateral breast cancer.10 Endocrine therapy was not recorded in the SEER database, so this study can only provide a population-based estimate on hormone receptor concordance in the era of tamoxifen. In the United States, the proportion of adjuvant tamoxifen use in breast cancer patients rapidly increased between 1987 and 1991 and reached approximately 40%-60% in 1999, and patients with ER-positive tumors were more likely to use tamoxifen than patients with ER-negative tumors.15, 16 In the present study, we found that the odds of ER or PR positivity decreased by approximately 21%-42% when comparing a breast cancer occurring 12-59 months after the first cancer and the primary cancer. These findings may be explained by tamoxifen use in cancer patients, because tamoxifen decreases the risk of ER-positive breast cancer but not ER-negative cancer.17, 18

The findings from this epidemiological study may enhance our understanding of breast carcinogenesis, particularly regarding the debate about a monoclonal versus multicentric origin of cancer. It is unclear whether the second cancer is an independent second primary tumor or the sequential event of a primary tumor. We found there is a clear gradient that the association between all 4 pathological features decreased as the time interval between 2 tumors increased within 12 months but leveled off after 12 months, suggesting that biological closeness of 2 tumors is a function of time. We postulate that intramammary spread of a single primary cancer may exist for 2 bilateral tumors that manifest clinically within 12 months of each other (ie, synchronous cases).

It is conceivable that 2 cancers developing from different origins are similar. Theoretically, 2 cancers in the same individual are subjected to the same hormonal, environmental, and genetic influences. Empirically, we demonstrated that a strong concordance in hormone receptor, grade, or histological type exists in subpopulations with less contamination from metastasic disease, as in patients with in situ contralateral cancer. Breast cancer develops from an accumulation of a series of mutations or epigenetic alterations, likely in mammary stem/progenitor cells, because these cells have slow-dividing, long-living, and self-renewing properties.19 Several breast cancer molecular subtypes have been proposed based on gene expression profiling, with the ESR gene being a cornerstones of the molecular classification.20 However, it is unclear whether these subtypes have distinct cancer stem cells of their own. Our data support the model that ER-positive and ER-negative tumors arise from different stem/progenitor cells.21 The present study observed that 2 bilateral cancers occurring 10 years apart are still likely to have the same tumor features, and 2 bilateral cancers occurring in patients aged <50 years have stronger similarity than bilateral cancers occurring in patients aged >50 years. These findings suggest that tumor subtypes as indicated by ER/PR status are established in the early steps of breast carcinogenesis. This predisposition of breast cancer subtypes may be determined by genetic susceptibility and environmental risk factors exposed in early life. For example, previous studies have shown that BRCA1-related breast cancer is mainly basal-like breast cancer22 and that low-penetrant genetic variants are ER-specific.23

To our knowledge, this is the largest study to investigate the concordance in cancer characteristics in patients with 2 breast cancers. The large sample size not only reconciled the inconsistencies of previous studies but also facilitated a subgroup analysis to answer several questions with certainty. Furthermore, because the patients were from population-based registries, the study findings are applicable to the general population. There are several limitations to our study, however. First, hormone receptor assays were not standardized across clinics in SEER, although a recent study showed that ER measurement is reliable for SEER registries.24 Nevertheless, the misclassification should be to reduce the strength of true association so we would observe stronger association if standardized assay were available. Second, primary breast cancer rather than recurrent or metastatic breast cancers were presumably recorded in SEER, but it is still possible that some metastatic cancers made it into the registries.

In conclusion, we observed strong associations between hormone receptor, tumor grade, and histological type between 2 bilateral breast cancers in the same patients, supporting the notion that 2 tumors arise in a common milieu and that tumor subtypes are predetermined in the early stage of breast carcinogenesis. These findings have important implications in the prevention and management of second breast cancers. For a specific individual with initial breast cancer, the risk factors for her second breast cancer may be the same as for the first one, thus a similar systemic management may be appropriate for the second cancer. We also observed a reduction of ER/PR-positive proportion for second breast cancer compared with the first one, which may be explained by hormonal therapy for the initial cancer.

CONFLICT OF INTEREST DISCLOSURES

Supported in part by the National Cancer Institute at the National Institutes of Health (grants R03 CA132143-01A1 and P50 CA125183).

Ancillary