SUBJECTS AND METHODS

STATISTICAL ANALYSIS

Differences in subjects and tumor characteristics between the various breast cancer subtypes were analyzed using analysis of variance for continuous variables and chi-square test for categorical variables. The Kaplan-Meier product limit method was used to estimate the overall and disease-free survival. Ninety-five percent confidence intervals (CI) for the percentage surviving at a particular time were estimated using the logit transformation. Overall survival was measured from the date of diagnosis to the date of death from any cause. Disease-free survival was measured from the date of first definitive treatment to the date of first relapse or death from any cause. Survival times were censored at the dates of last contact for subjects who were lost to follow-up. Each breast cancer subtype was compared with the most common reference group of ER/PR+, Her2− subtype.

Cox-proportional hazard model was used to estimate the hazard ratios and 95% CI for overall and disease-free survival between the breast cancer subtypes adjusting for age, lymph node status and disease stage. S-plus statistical software was used for survival analysis and SAS 9.1 (SAS Institute, Cary, NC, USA) was used for all other analyses. A P-value of <0.05 was considered statistically significant.

RESULTS

Final analysis included 1134 invasive breast cancer subjects identified in the Marshfield Clinic/St. Joseph’s Hospital Cancer Registry from January 1, 1998 to June 30, 2005. The mean age of all subjects was 62.7 years (SD, 13.8; range, 27.9 to 95.8 years). Baseline characteristics of subjects including tumor subtype are presented in table 2⇓. Of 1134 subjects, 116 (10.2%) were ER/PR+, Her2+, 781 (68.9%) were ER/PR+, Her2−, 85 (7.5%) were ER/PR−, Her2+, and the remaining 152 (13.4%) were classified as triple negative (table 2⇓).

Differences in baseline characteristics between the four subtypes are presented in table 3⇓. Subjects with ER/PR+, Her2− subtype were more likely to be older (P<0.001), have early stage breast cancer (P<0.001), present with small tumor (P<0.001) and have a well/moderately differentiated histological grade (P<0.001). They were less likely to be lymph node positive (P<0.001), have a lobular tumor type (P<0.001), and be treated with chemotherapy (P<0.001).

Ninety-nine subjects (8.7%) had recurrence and 140 (12.3%) died during the follow-up period. The estimated median follow-up period for all subjects was 4.8 years (range, 3.2 months to 9.4 years). The 5-year overall survival for all subjects was 87.8% (95%CI, 85.4–89.9) and the disease-free survival was 83.1% (95% CI, 80.5–85.5). The 5-year overall and disease-free survival by tumor subtype, ER/PR status and Her2 status is presented in table 4⇓. Of the 99 subjects with recurrence, 45 (45.5%) had local recurrence and the remaining had recurrences in bone (39.4%), liver (22.2%), lung (15.1%), mediastinal lymph nodes (10.1%), brain (7.1%) and other sites (11.1%) (table 5⇓). The age adjusted odds ratio for tumor characteristics associated with various subtype are presented in table 6⇓.

The Kaplan-Meier curve for overall and disease-free survival by tumor subtype is shown in figure 1⇓. In the Cox regression analysis, subjects with the triple negative subtype, ER/PR−, Her2− had the worst overall survival (hazard ratio, 1.8; 95% CI, 1.0–3.0) and worst disease-free survival (hazard ratio, 1.5; 95% CI, 0.8–3.0) when compared with subjects with ER/PR+, Her2− subtype (table 7⇓). Adjustment was made for age, stage, histological grade, chemotherapy treatment and lymph node status (table 7⇓).

Among the 781 subjects with ER/PR+, Her2− subtype, 524 received no chemotherapy and 257 received chemotherapy. Those who received chemotherapy had significantly better overall (hazard ratio, 0.45; 95% CI, 0.26–0.77; P = 0.004) and disease-free (hazard ratio, 54; 95% CI, 0.34–0.86; P = 0.009) survival benefits when compared with subjects who did not receive chemotherapy. No significant differences in overall and disease-free survival benefits were observed in other subtypes. This may be due to the small number of subjects in the other subtype groups.

DISCUSSION

This study confirmed breast cancer as a multifaceted disease comprised of distinct biological subtypes with diverse natural history which are increasingly recognized as presenting a varied spectrum of clinical, pathologic and molecular features with different prognostic and therapeutic implications.³ Our results reveal statistically significant differences in clinical and pathologic features and outcomes between subtypes. Using the most common subtype (ER/PR+, Her2−) as a reference, the triple negative subtype (ER/PR−, Her2−) had the worst overall survival (hazard ratio, 1.8; 95% CI, 1.06–3.2), and worst disease-free survival (hazard ratio, 1.5; 95% CI, 0.8–3.0). These expected survival differences have meaningful implications for communication and care decisions between providers and their cancer patients. This subclassification should, however, be complemented with the many other important traditional prognostic variables for the individual such as age, tumor size, lymph node status, comorbidity, and adjuvant therapy.

IHC-based classification of both ER/PR and Her2 status provides prognostic and therapeutic information not achievable from either alone. Prior classifications separating breast cancer into one of two categories based on ER expression alone is less discriminatory in terms of prognosis, and the additional subclassification based on Her2 expression provides enhanced and important therapeutic guidance. Breast cancer has also sometimes been dichotomized into triple negativity or other.¹³ This classification is informative but simplistic and may be misleading by grouping the ER/PR−, Her2+ with ER/PR+, Her2+ and ER/PR+, Her2−. This was borne out in our results, where the ER/PR+, Her2+ had statistically equivalent survival to the referent ER/PR+, Her2− subtype, and in practice, both types have better prognostic and therapeutic connotations. However, the ER/PR−, Her2+ point estimates were more similar to the triple negative values. Also, recent studies have suggested that within the ER/PR+ subtypes, the clinical and pathologic response to chemotherapy varies with the ER/PR+, Her2+ subtype defined by both hormone receptor and Her2 expression showing better response to chemotherapy.¹³ ER/PR+, Her2+ tumors virtually always have a high recurrence score.¹⁴ Recently it was shown in a retrospective analysis that ER/PR+, Her2− tumor may benefit less from taxanes in the adjuvant setting.¹⁵

We have classified breast cancer using IHC into 4 global subtypes out of the 8 possible subtypes commonly used by other authors.¹⁶ We believe this classification is practical, simple, informative, clinically useful, and quite discriminative between the subtypes. The other four groups will emerge if we differentiate based on PR expression (ER+/PR+ vs. ER+/PR− tumors).

The independent prognostive and predictive role of PR expression irrespective of ER has been a subject of great controversy as demonstrated by the report from the ATAC (Arimidex, Tamoxifen, Alone or in Combination) adjuvant trial, a large worldwide trial comparing the efficacy of tamoxifen with that of the aromatase inhibitor anastrazole, showing overall that patients with ER+/PR+ tumors had a lower recurrence rate than those with ER+/PR− tumors.¹⁷ The observation from the same study that patients with ER+/PR− tumors respond nearly as well to anastrozole as those with ER+/PR+ tumors suggests that the ER signaling pathway is functional in many ER+/PR− tumors, consistent with the well-known fact that the PR gene is regulated by the estrogen pathway.¹⁷ Also, the relative resistance by ER+/PR− tumors was not observed in the BIG 1–98 trial which is the largest study of an aromatase inhibitor as up-front adjuvant therapy for early breast cancer.¹⁸ Studies that have been classified as using more than 4 subtypes are plagued by these controversies and those inherent in small sample size and multiplicity of variables.^19,20

Recent publications have shown that newer molecular classification of breast cancer also has important prognostic value. Subtyping breast cancer using microarrays for gene expression analysis is the best way to perform such molecular classification. However, most archived clinical specimens are not amenable to such analysis. Moreover, until recently when Oncotype DX and MammaPrint became commercially available, such assays were limited to research laboratories and, therefore, were not optimally available for commonplace clinical practice. The IHC-based classification systems are still useful in clinical practice, especially when fresh tissue is not available, and has been shown to correlate well with intrinsic classification using gene expression microarrays: ER/ PR+, Her2+ with Luminal B; ER/PR+, Her2− with Luminal A; ER/PR−, Her2+ (ER−/Her2+) and ER/PR−, Her2− with triple negative/basal-like tumors.^3,13 It is worth noting that the reliability of the ER/PR and Her2 testing is imperfect. There is substantial intralaboratory and interlaboratory variation in ER results because fixation, antigen retrieval, and staining methods may differ among laboratories.^21–23 Substantial discordance among Her2 results generated in different laboratories from the same specimen has also been reported.^12,21,24 For this classification to be more helpful, ongoing efforts¹² should also be directed at standardization of current testing and development of more reliable and reproducible testing for ER/PR and Her2/ neu expression.^12,21–24

In our current analysis we have not considered the semi-quantitative information from IHC in terms of ER/PR or Her2 levels of expression on clinical outcomes largely because we do not have adequate sample for such analysis. We believe such subsetting within the subtypes may be unreliable, with regards to the message highlighted in this study, due to inadequate sample size.

This study of a predominantly Caucasian population reports the distribution of subtypes as different from that seen in a predominantly African American population where ER/PR−, Her2− is more prevalent (39% premenopausal versus 14% postmenopausal African American women versus 16% non-African American women of all ages).³ Also of note, 80% of our subjects are ER+ accounting in part for the overall 5-year survival of 87.8% (95% CI, 85.4–89.9) and 5-year disease-free survival of 83.1% (95% CI, 80.5–85.5). A large percentage of our patients demonstrate favorable features such as small tumor size (<2 cm; 71%), negative nodal status (61%), and low to moderate histologic grade (59%).

An investigation of all subgroups showed benefit from chemotherapy, but after controlling for age, tumor size, and lymph node status, the sample size was not sufficient to make a strong assertion except in the ER/PR+, Her2− subgroup. In the ER/PR+, Her2− subgroup (781 subjects), 524 patients did not receive chemotherapy and 257 patients did receive chemotherapy. Chemotherapy conferred overall and disease-free survival advantages (P = 0.003 and P = 0.009, respectively).

This study supports other studies^3,13,25 which have shown both the triple negative and Her2+/ER− subtypes to have poorer clinical, pathologic and molecular prognoses. The triple negative group has the worst overall and disease-free survival. For the triple negative group the disease-free survival curve tends to plateau after the second and third year, but the curve continues downward for the Her2+/ER−. In 2008, tumor-expressed proteins, such as ER and Her2, play an increasingly important role in determining breast cancer treatment. For example, women with ER+ breast cancer typically receive endocrine therapy (tamoxifen or aromatase inhibitors) and women with Her2+ breast cancer may receive anti-Her2 (trastuzumab [Herceptin] and lapatinib [Tykerb]). It is reasonable to assume that trastuzumab and lapatinib will shift the Her2+ curve upward (Her2+/ER−). We lack targeted therapies for triple negative breast cancer and this continues to direct the focus of ongoing research.^26–28

Despite the enormous effort and funding channeled towards molecular diagnostics, there is still relevance for IHC, especially when performed by inexperienced centers. Although molecular arrays have been around for approximately a decade, new therapeutic target proteins are not being identified, thus the predictive value of the assays are relatively global or limited to known targets such as ER/PR protein or the Her2 gene. Also, despite multiple and different gene sets used in most of the molecular testing, there is significant agreement in the outcome predictions for individual patients by these tests,¹⁴ suggesting that they are probably tracking a common set of biologic phenotypes which are heavily weighted toward ER/PR and Her2 gene pathways. Finally, the superiority of molecular technology over IHC testing is theoretical and based on the premise that molecular technology provides quantization and reproducibility. This presumptive theory is the basis for some ongoing studies but is yet to be proven.

CONCLUSIONS

Our study showed the triple negative subtype (ER/PR−, Her2−) has the worst overall and disease-free survival compared to the other subtypes. Further confirmatory studies are necessary to refine IHC classification. We support IHC classification as a clinical tool as ER/PR and Her2 testing is widely available at a reasonable cost, is a clinically-used, therapeutically informative classification of breast cancer based on immunophenotype/biologic phenotypes, and is prognostic as well as somewhat predictive. Additional ongoing efforts¹² should be directed at standardization of current testing methods and development of more reliable and reproducible testing.

ACKNOWLEDGMENTS

The authors thank Marshfield Clinic Research Foundation for its support through the assistance of Anne Nikolai and Alice Stargardt in the preparation of this manuscript. We also appreciate the editorial comments of Dr. William Hocking.