Factors predictive of response to hormone therapy in breast cancer

Transcription

1 Tumori, 94: , 2008 Factors predictive of response to hormone therapy in breast cancer Francesca Rastelli and Sergio Crispino Istituto Toscano Tumori, Dipartimento Oncologico USL7, Siena, Italy ABSTRACT Aims and background. Approximately half of metastatic breast cancers expressing estrogen and/or progesterone receptors responds to endocrine therapy, and postoperative adjuvant endocrine therapy provides about a 50% reduction in the development of recurrent disease. A number of publications have focused on the correlation of biomarkers, in particular estrogen and progesterone receptors and HER-2/neu status as well as different gene profiles, multigene assays and genetic polymorphisms with response to hormone therapy. The purpose of this article is to review the literature to identify biological markers predictive of response to tamoxifen and aromatase inhibitors. Methods. A computerized literature search through Medline and ASCO abstract databases was performed, applying the words endocrine therapy and predictive markers and each of the following: early and metastatic breast cancer, estrogen receptors, progesterone receptors, HER2/neu, multigene assays, polymorphisms. The last search was updated in June In the examined literature, biological markers were retrospectively assayed to establish whether such variables were predictive for endocrine therapy efficacy. Results. The role of estrogen receptor content as a predictor of response to endocrine treatment was confirmed: benefit from endocrine treatment was directly proportional to estrogen receptor levels. Progesterone receptor status was only a strong time-dependent prognostic value, and it has not yet been validated as a predictive factor of tamoxifen efficacy. Retrospective clinical data from upfront and sequential studies of aromatase inhibitors were discordant regarding the degree of benefit of these drugs over tamoxifen according to progesterone receptor status. HER-2 positivity was associated with a significantly greater risk of endocrine therapy failure in metastatic and neoadjuvant settings. The current generation of genomic assays for tamoxifen sensitivity all contain a combination of prognostic information that it is difficult to integrate into clinical practice. Conclusions. Available clinical data are inconclusive to support preferential use of aromatase inhibitors over tamoxifen in progesterone-receptor-negative and HER-2- positive tumors, but it was also clear that lower estrogen receptors, lower progesterone receptors, and positive HER-2 are associated with lower responsiveness to any type of endocrine therapy. Tumors overexpressing HER-2 are endocrine resistant and they require the blockage of the HER-2 pathway in addition to estrogen deprivation. Recent molecular studies have shown that endocrine responsiveness is to a large extent influenced by estrogen-receptor-related pathways. In the future, the key to the correct tailoring of hormone therapy will probably be the ability to subtype estrogenreceptor-positive breast cancer. Introduction Breast cancer is classified as one of the hormone-dependent tumors. Endocrine therapy, such as antiestrogen therapy and estrogen ablation, is the treatment of Key words: aromatase inhibitors, breast cancer, c-erbb2, estrogenprogesterone receptors, gene expression profile, tamoxifen. Correspondence to: Francesca Rastelli, c/o UO Oncologia, USL 7, Ospedale Alta Val D Elsa, Campostaggia (Siena), Italy. Tel /4429; fax ; francesca.rastelli@libero.it Received August 13, 2007; accepted November 21, 2007.

2 HORMONE THERAPY PREDICTIVE FACTORS IN BREAST CANCER 371 choice for patients with breast cancer expressing estrogen receptors (ER) and/or progesterone receptors (PR). The clinical usefulness of endocrine therapy has been proven in the prevention, adjuvant, neo-adjuvant and metastatic settings. Endocrine sensitivity, assessed by the expression of ER and/or PR, has long been the only recognized and validated predictive factor to guide therapeutic decisions. The Oxford Overview on adjuvant therapies for breast cancer clearly showed that the benefit of endocrine treatment is limited to patients with tumors expressing ER and/or PR 1. The degree of endocrine sensitivity is currently quantified by immunohistochemistry (IHC) and may be expressed as the percentage of cells stained after treatment with anti-er/pr antibodies. According to the conclusions reached at the Consensus Conference held by the National Cancer Institute in the year 2000, all tumors with at least 10% of cells expressing ER/PR are considered as potentially endocrine sensitive 2. The use of ErbB2 to guide decisions regarding endocrine therapy has not been completely clarified, but it has entered into clinical practice as a justification for estrogen deprivation therapy versus treatment with an antiestrogen with agonistic potential. Several retrospective studies have suggested that HER-2+ tumors may be less sensitive to endocrine treatments and, namely, to tamoxifen. This finding has been attributed to the existence of a cross talk between the ER and HER-2 metabolic pathways 3 and to the detection of lower absolute levels of ER when HER-2 is amplified/over-expressed 4. The aim of this article to review the most recent publications, reviews and phase II /III clinical trials, focused on the predictive value of single (ER, PR, HER-2) and combined markers with response to an endocrine agent. The combined markers examined are represented by ER pathway and its cross talk with HER-2, a tyrosine kinase growth factor receptor. The principal object of this review is publications concerning tamoxifen in different settings of treatment and recent trials in metastatic and adjuvant settings with aromatase inhibitors (letrozole, anastrozole, and exemestane). Moreover, a genotypic and pharmacogenetic evaluation of the tumor, although with preliminary data, has been carried out to explain the limits of these promising applications in clinical practice. The article is structured on the analysis of single endocrine markers, followed by the analysis of endocrine cellular pathways corroborated by clinical data reporting their predictive value. The reports of clinical data are subdivided by different settings: adjuvant, metastatic and neoadjuvant. The latter part of this review is oriented on data concerning different gene expression profiles and pharmacogenetic evaluation. Estrogen and progesterone receptors ER and PR were the first predictive biomarkers recommended for routine clinical use in breast cancer by the Tu- mor Marker Panel of the American Society of Clinical Oncology 5. In practice, their primary use is to distinguish patients who have little or no chance of benefiting from endocrine therapy from those who do have some reasonable chance. The justification for this endorsement was based on many studies conducted over the past two decades, involving patients in randomized clinical trials which showed that these tests were sufficiently sensitive, specific, and reproducible to reliably identify subsets of patients with significantly different risks for recurrence, survival, or treatment response. ER exist as two isoforms, ER alpha and ER beta, that are encoded by two different genes. ER beta might be of importance to predict resistance to endocrine treatment 6, as studies with clinical samples suggest that ER alpha-expressing breast cancer with low ER beta content tends to be tamoxifen resistant 7. PR is expressed as two different isoforms, PR-A and PR-B proteins, resulting from a single gene 8. The presence of PR might be a better indicator of hormone dependence than ER, because ER may be present but not functional in some patients. The presence of ER is considered to be a prerequisite for PR positivity, and ER-/PR+ tumors could be the result of a false-negative ER assay. PR is an ER-regulated gene, although this interaction can be modified by growth factor cross talk, e.g., by HER-2 or epidermal growth factor receptor (EGFR) amplification/overexpression 9. It has been hypothesized that HER-2 overexpression may interact with some of the metabolic pathways triggered by the activation of the ER. In fact, lower mean levels of PR were found in HER-2+ versus HER-2- tumors 10. It has also been reported that the P13K-Akt-mTOR pathway can actually reduce the expression of PR at the transcriptional level by activating the AP-1 site in the PR gene promoter 11. Yet, international consensus on how to study hormonal receptors has not been reached. Compared with ligand-binding assays, IHC is easier to perform, less expensive, applicable to a wider variety of samples and more sensitive and specific in the identification of heterogeneous ER and PR expression due to components with dissimilar degrees of differentiation in a single tumor. The cutoff of ER positivity by IHC was based on the best predictive value of ER status for disease-free survival (DFS) 12. DFS survival curves for all possible IHC scores show that higher scores are related to better DFS (Figure 1). On the basis of these results, tumors were defined as ER-positive if their total Allred-IHC score was greater than 2 (corresponding to as few as 1% to 10% weakly positive cells) and ER-negative if their score was 0 or 2 (less than 1% of tumor cell nuclei immunostained). This observation validates the concern expressed by Allred and Moshin 13 that a 10% cutoff for ER could exclude patients unnecessarily from the benefits of endocrine therapy. Many hospital and commercial laboratories use diverse methodologies, and most have arbitrarily chosen 10% or even 20% positive tumor cells as their cutoff for defining ER positivity, potentially

3 372 F RASTELLI, S CRISPINO DFS probablity Best cutpoint: IHC score >2 (P <0.0001) Time (months) IHC score (% patients) 8 (5.8% 7 (19.8%) 5 (17.4%) 6 (23:4%) 4 (11.7%) 3 (5.1%) 2 (2.1%) 0 (14.7%) Er positive Er negative Figure 1 - Univariate DFS curves for all possible total immunohistochemistry (IHC) scores in patients receiving any adjuvant endocrine therapy (almost always tamoxifen). An IHC score G 2 was the optimal cut point for predicting significantly improved outcome (P = ), and this value was used to define ER positivity. It is noteworthy that higher scores were related to better disease-free survival (DFS) in ER+ breast cancer patients. (Modified from Harvey JM, J Clin Oncol, ) denying a substantial number of patients the benefits of adjuvant hormone therapy. HER-2 HER-2 is over-expressed in approximately 25% of breast cancers. Recent studies have shown that 45-50% of HER-2-overexpressing breast cancers are also ER positive 14. HER-2 (c-erbb-2/neu) is a tyrosine kinase growth factor receptor located on the cell membrane of normal and tumor cells. Activation of the receptor induces a phosphorylation cascade in cytoplasmic kinases, which result in the activation of nuclear transcription proteins and cellular growth. HER2/neu has been detected in the serum of 20% to 50% of patients with metastatic breast cancer and is correlated with overexpression of HER-2 in tissue. In most studies, 80% to 100% of patients with HER-2/extracellular domain (ECD) detectable in the serum are 2/3+ by IHC or fluorescence in situ hybridization positive 15. Elevated serum levels of HER-2 are defined as an ECD concentration of greater than 15 ng/ml, but some studies considered positive HER-2 serum levels greater than 30 ng/ml 16. The functional significance of ECD shedding has not been determined, but in vitro data suggest that deletion of the extracellular carboxyl terminus of the molecule enhances the signaling activity and transforming ability of the NH3 terminally truncated receptor, p95her-2. In fact the cleavage of the ECD of HER2/neu receptor leads to increased phosphorylation of the intracellular tyrosine kinase 17. Interactions between HER-2 and ER pathways HER-2 expression in human breast cancer is downregulated by estrogens 18. Conversely, overexpression of HER-2 promotes estrogen-independent growth and is associated with resistance to tamoxifen in vitro and in animal models, possibly by promoting phosphorylation of receptors. In preclinical models, transfection of HER- 2 into MCF-7 (ER-positive, tamoxifen sensitive) cell lines was shown to induce resistance to tamoxifen 19. Overexpression of HER-2 in tamoxifen-sensitive MCF-7 breast cancer cells results in mitogen-activated protein kinase (MAPK) hyperactivity and resistance to antiestrogen. MAPK hyperactivity promotes increased association of the ER with coactivators and decreased association with corepressors, thus favoring gene transcription. Since both aberrant EGFR and HER-2 signals can hyperactivate MAPK, overexpression of these receptors can lead to antiestrogen resistance by MAPK-dependent mechanisms. Resistant MCF-7 cells exhibited markedly elevated levels of EGFR and HER-2; they also exhibited increased levels of activated MAPK. Tamoxifen resistance may be explained by the HER1/HER-2 heterodimerization and p21ras/raf pathway activation that leads to MEK/MAPK phosphorylation, driving cells to proliferation by proliferation-genes (c-jun/fos) or through phosphorylation of ER homodimers. In the latter case, removing the ligand, ER becomes monomeric and cannot be phosphorylated. This is a possible biological explanation for the efficacy of the aromatase inhibitors in HER1/2-positive breast cancer. Gene profiles In a recently published paper 20 with the aim of evaluating concordance among several gene-expressionbased predictors for breast cancer, including the Amsterdam 70-gene, Rotterdam 76-gene, and OncotypeDX 21-gene signatures, the authors reported that these gene-expression signatures did predict for outcome and were comparable in their ability to predict an increased risk of relapse. The first such assay to publish results was the Amsterdam 70-gene profile 21, which attempted to distinguish patients with operable breast cancer who are likely to relapse from those not likely to relapse and included ER-positive, ER-negative, lymph node-positive, and lymph node-negative disease. The assay, Mamma Print, is undergoing prospective validation as a part of the MINDACT Trial (micro array in node-negative disease may avoid chemotherapy) 22 in Europe. It is thus a validation trial, which will assess the clinical utility of these micro array-based tools compared with best standard practice, as well as with Adjuvant! Online. The Rotterdam 76-gene signature 23 has also been externally validated in patients who received no treatment for ERpositive or ER-negative, lymph node-negative disease. Both of these micro array-based tools generate binary results i.e., good or bad prognosis for a given case. A third signature that has been extensively evaluated is the 21-gene Oncotype DX assay 24. The assay includes 16

4 HORMONE THERAPY PREDICTIVE FACTORS IN BREAST CANCER 373 tumor-related genes and 5 reference genes that generate a quantitative recurrence score, expressed as a number from 1 to 100. The ER-related group of genes consists of ER, PR, Bcl2 and SCUBE2. The score is correlated to a specific probability of distant recurrence at 10 years. In patients who have a mid-range score, which can account for anywhere from 40% to 60% of all patients tested, the score is not necessarily informative because we do not know for sure whether chemotherapy will be useful in reducing the risk of recurrence in that group. This is the question that the TAILORx (Trial Assigning Individualized Options for Treatment(Rx)) trial is designed to define. The trial therefore uses a not inferior design to determine that hormonal therapy alone is not inferior to chemohormonal therapy. Single nucleotide polymorphisms Single nucleotide polymorphisms are alterations of DNA sequence situated at well-defined positions along gene sequences, which occur at higher frequency than casual mutations. In terms of protein function, single nucleotide polymorphisms can disrupt or silence the normal activity of a gene or the respective encoded protein, thus resulting in an abnormal drug response (cancer resistance or patient toxicity) 25. Tamoxifen and its metabolites display large interindividual variation with profound implications for breast cancer outcomes. Tamoxifen is hydroxylated to the potent metabolites, 4- hydroxytamoxifen and endoxifen, by various cytochrome P450 (CYP450) genes including CYP2C9 and CYP2D6. CYP2D6 genetic variants have been correlated with reduced plasma concentrations of endoxifen (the most active form of tamoxifen). Aromatase inhibitors are the most effective suppressors of endogenous estrogen production, but they behave differently in terms of their metabolic pathways, they affect the P450 enzyme differently and therefore they are probably likely to have different collateral effects. Aromatase protein (CYP450-19) catalyzes the formation of aromatic C18 estrogens from C19 androgens, a critical reaction for estrogen biosynthesis. Material and methods A computerized literature search through Medline and ASCO abstract databases was performed, applying the words endocrine therapy and predictive markers and each of the following: early and metastatic breast cancer, ER, PR, HER2/neu, multigene assays, and polymorphisms. The last search was updated in June In the examined literature, biological markers were retrospectively assayed to establish, through a multivariate test of tamoxifen or aromatase inhibitors/covariate interactions, whether these variables predicted for endocrine therapy efficacy. Definition of a predictive marker: a predictive marker is correlated with response to a specific therapy in an independent manner with respect to prognosis. Predictive markers identify patient subgroups likely to benefit from specific therapies but permit selection of a therapeutic approach, where the likelihood of response is linked only to the predictive marker. Estimates of endocrine therapy effect were expressed as risk ratio and hazard ratio for death of hormonal drug over no treatment, in a metastatic (or neoadjuvant) and in an adjuvant setting, respectively. Single markers and clinical data concerning their predictive value are reported separately. Results We report separately literature search results concerning predictive values of single markers like ER, PR, HER- 2, then of combined markers, such as ER/PR/HER-2. Finally, clinical data about the predictive value of gene profile, two genes ratio and genetic polymorphisms are reported. Other biological markers have been investigated in randomized trials, such as S-phase, ploidy, EGFR, prolactin receptor and tumor angiogenesis (count of micro vessels within the tumor), but none of these seems to predict the efficacy of endocrine therapy 4. ER The role of ER content as a predictor of response to endocrine treatment in early breast cancer patients has been studied and consistently recognized. Results from the Oxford overview 26 indicated that women with ERpositive tumors derived significant benefit from 5 years of tamoxifen in reducing the odds of recurrence and death (Table 1). In a retrospective analysis of neoadjuvant tamoxifen versus letrozole 3, there was a significant difference in response rates between ER+ and ER- tumors (P = ) based on a linear logistic model (Figure 2). It seems that the benefit was directly proportional to the level of ER, with patients with high tumor ER levels deriving the greatest benefit from therapy. However, quantitative analysis from the Trans-ATAC trial 27 showed no evidence of a greater relative benefit from aromatase inhibitors (in this case anastrozole) in postmenopausal women with high ER levels tumors than in patients with low ER levels. The results of a recent study 28 confirmed that ER status, quantified by the 21-gene Oncotype DX assay, was primarily predictive of tamoxifen benefit and not significantly associated with prognosis in untreated patients (Table 1). Whether these results may be related more to a real biologic effect than to problems with PR measurement still needs to be clarified. PR It was first hypothesized that PR might provide additional information to more accurately predict which pa-

5 374 F RASTELLI, S CRISPINO Table 1 - Predictive markers of response to tamoxifen: ER and PR status Reference End point Type of study/setting Laboratory analysis Marker Tam P (trend) Oxford Metanalysis Recurrence rate Metanalysis-adjuvant Not centralized ER+ vs ER- < studies analysis IHC ER+PR+ vs ER+PR- Ns Baehner FL 28 DFS Retrospective-NSABP-14 Multigene assay ER quantitative (Adj tamoxifen vs Placebo) expression PR quantitative 0.64 expression Bardou V-J 31 DFS/OS Retrospective-Program project Centralized analysis ER+PR+ vs Database (Adj tamoxifen) Dextran-coated method ER+PR- (pos 5 fmol/mg) Tovey S 34 3-year early relapse Adj tamoxifen IHC ER+PR+ vs ER+PR ER, estrogen receptor; PR, progesterone receptor; IHC, immunohistochemistry; Ns, not statistically significant; DFS, disease-free survival; OS, overall survival; Adj, Adjuvant. tients will respond to hormonal therapy because the presence of PR should serve as an indicator of a functionally intact estrogen response pathway. In a metastatic setting, patients with advanced breast cancer have a higher response rate to endocrine therapy when they have ER+/PR+ tumors than patients with ER+/PR- tumors 29. In a neoadjuvant setting, Ellis et al. 3 reported that the relationship between PR level and hormonal response is not linear (Figure 2). In fact, maximal response rates to aromatase inhibitors (and tamoxifen) occurred at intermediate levels of expression of PR, not at the highest levels of expression: high, as well as low PR expression scores were associated with a lower chance of responding than intermediate scores (Table 2). The degree of statistical significance associated with the inverse V-shaped relationship between PR expression levels and response to letrozole (P = ) suggests that new hypotheses and models for the predictive relationship between PR and the effectiveness of aromatase inhibitors should be considered. Perhaps PRrich tumors are associated with levels of aromatase activity or hypersensitivity to estrogens sufficiently high enough to blunt the efficacy of estrogen deprivation therapy with aromatase inhibitors. ER- PR+ cases are uncommon but responsive to letrozole therapy (50% risk ratio), consistent with the hormone sensitivity of this rare tumor phenotype. Moreover, Dowsett et al. 30 showed that, even in an adjuvant setting, measurement of PR status in ER- tumors defines a group of patients that benefits from tamoxifen but that would be excluded from antiestrogen therapy on the basis of ER status alone. % of cases in each category Total % of cases in each Allred category Response % tamoxifen letrozole Response % tamoxifen letrozole A ER allred score B PgR allred score Figure 2 - A) Clinical response rate versus ER Allred score for letrozole and tamoxifen. The P value for a linear logistic model was for letrozole and for tamoxifen. B) Clinical response rate vs PR Allred score for letrozole and tamoxifen. Logistic models were linear assuming a peak response rate associated with a score of 5 for letrozole and 4 for tamoxifen. The P values for these models were and , respectively.

6 HORMONE THERAPY PREDICTIVE FACTORS IN BREAST CANCER 375 Table 2 - ER, PR, HER2, EGFR status as a single or combined factors. Interaction tests (P trend) with tamoxifen and aromatase inhibitors Marker Laboratory analysis Tam response rates P trend Aromatase inhibitor P response rates ER+ vs ER- -IHC pos 10% 40 vs 11% vs 19% nuclear cell staining -ER Allred score vs vs for a linear ( 10% of cells) vs % for a linear 45-65% logistic model (10-90% of cells) logistic model PR+ vs PR- -IHC pos 10% 43 vs 28% vs 41% nuclear cell staining -PR Allred score 18-60% vs 45-25% for an inverse vs 80-45% 0,0015 for an inverse 0-4 vs 5-8 V-shaped model V-shaped model ER+PR+ vs ER+PR- IHC 41 vs 39% vs 53% ER+HER2+ vs ER+HER2- IHC 17 vs 40% vs 53% ER+EGFR+ vs ER+EGFR- IHC 0 vs 37% vs 56% ER+EGFR/HER2+ vs IHC 21 vs 42% vs 54% ER+EGFR/HER2- ER, estrogen receptor; PR, progesterone receptor; EGFR, epidermal growth factor receptor; IHC, immunohistochemistry. (Data from Ellis JM et al, JCO ) The latest meta-analysis of all tamoxifen trials from the Early Breast Cancer Trialists Collaborative Group (EBCTCG) 26 showed that being PR positive did not make a difference with respect to response to tamoxifen therapy: relative reduction in relapse is similar in ER+/PR+ and ER+/PR- tumors (37% and 32%) (Table 1). PR expression quantification by the Oncotype DX assay 28 confirmed that the level of PR expression was primarily prognostic in untreated patients and not predictive of a tamoxifen benefit (Table 1). For this reason, ER and PR seem to have a different role in the biology of ER-positive breast cancer (predictive and prognostic, respectively). Only one study by Bardou et al. 31 demonstrated in univariate and multivariate analysis that PR status, when accurately measured, is an independent predictive factor for benefit from adjuvant endocrine therapy (Table 1). The study indicated that patients with ER+/PR- tumors may want to more seriously consider the addition of chemotherapy given the reduced benefit expected from hormone therapy alone. Differently from all studies examined in the Oxford meta-analysis, in the study by Bardou et al., 31 all assays were performed at central laboratories with strict quality control. It should be remembered that PR assays have been known to have more variation and to be less accurate than those for ER. In fact, PR is displayed more heterogeneously in tissue and may therefore be more difficult to measure accurately in small specimens. The predictive value of PR is still debated and often confused with its time-dependent prognostic value, which has been repeatedly reported in the literature 32. In a recent published study on about 1000 patients who received different endocrine treatments (tamoxifen, Gn-RH, aromatase inhibitors) 33, lack of PR expression (P = 0.003) showed a stronger prognostic association with shorter DFS than HER-2 (P = 0.05). As in laboratory models, the absence of PR expression in ER+ tumors has been interpreted as a surrogate marker for HER-2 activation. ER+/PR- tumors express higher levels of HER-1 and HER-2 and display more aggressive features than ER+/PR+ tumors 9. This hypothesis has been questioned by a recent study by Tovey et al. 34, which show that ER+/PR-/HER-2+ cases were significantly more likely to relapse early during the first three years on tamoxifen (P = 0.017). The study did not include a no-treatment arm, which would be required in order to ascertain the existence of a true endocrine resistance of ER+/PR-/HER-2+ tumors. The Trans-ATAC trial 27 collected tumor blocks from the ATAC trial 35 to try to retrospectively identify molecular characteristics of tumors that indicate differential benefit from anastrozole and tamoxifen. The ATAC trial examined the role of anastrozole versus tamoxifen as adjuvant endocrine therapy for 5880 postmenopausal women with breast cancer. Subsequent CRF-based retrospective analysis 36 examined patient response to anastrozole by ER, PR and HER-2 status, with some suggestion of a particular benefit from anastrozole for ER+/PR- tumors (P <0.001) not seen in the ER+/PR+ group (P = 0.07) (Table 3). In contrast with this earlier CRF-based hypothesis, recent data from TransATAC tissue analysis indicated that the benefit of anastrozole over tamoxifen was similar for ER+/PR+ and ER+/PRtumors (Table 3). It is important to remember that block

7 376 F RASTELLI, S CRISPINO collection in the TransATAC trial was incomplete (of 5880 patients, tissue was sufficient for analysis in only 1856 tumor specimens) and that the 1856 patients were mostly English because of the availability of specimens. In the BIG1-98 trial 37, comparing upfront letrozole versus tamoxifen as adjuvant hormonal therapy, the degree of benefit from letrozole compared with tamoxifen suggested only a slight difference in the magnitude of the letrozole effect according to PR status. Based on the centrally reviewed receptor values, the benefit of letrozole over tamoxifen was numerically larger for the ER+/PR+ compared with the ER+/PR- cohort but without a statistically significant difference (Table 3). The current report from MA (randomization of patients who had completed 5 years of tamoxifen to 5 years of letrozole compared with placebo) showed that the ER+PgR+ cohort benefits more from letrozole following tamoxifen than the ER+/PR- cohort. In this study of extended hormone therapy after 5-years of tamoxifen, it is important to note that ER+/PR+ tumors constituted 73% of the study population, whereas ER+/PR- tumors constituted only 12%. Patient population selection is evidenced by the high proportion of ER+/PR+ tumors (73%) compared to the ATAC (62%) or BIG1-98 study (63%) (Table 3). In the IES study 39, comparing 5 years of tamoxifen to a sequential approach where patients free of disease switched to exemestane after 2-3 years on tamoxifen, there was apparent difference in relapse in ER+/PR+ and ER+/PR- tumors (Table 3). The lack of significance of PR status when patients switch to aromatase ihibitors after 2-3 years on tamoxifen was confirmed by two large studies of sequential therapy with anastrozole, ABCSG Trial 8 and ARNO (Table 3). HER-2 The predictive value of HER-2 status for endocrine sensitivity is highly debated. At this time, the use of HER-2 status to select endocrine therapy for early breast cancer is not recommended 41. In the adjuvant setting, there are only small, non-randomized studies, with both heterogeneous patient populations and methods for measuring HER-2 status. Five historical studies that addressed HER-2 as a predictive factor for hormone Table 3 - ER+PR+ versus ER+PR- subgroups: comparison of hazard ratios (HR) for DFS and RR from adjuvant and neoadjuvant clinical trials of aromatase inhibitors and tamoxifen Reference Drug/end point Laboratory analysis Subgroups (% patients) HR / RR P TransATAC 27 AI/DFS Centralized (C) ER+PR+ HR 0.72 Ns 1856 tumor specimens percentage of cell staining ER+PR- HR 0.68 (positive>10%) on whole sections CRF ATAC 36 AI/DFS Not centralized (NC) percentage of cell ER+PR+ HR 0.84 P < Case Record staining (positive >10%) (61%) Form (CRF) ER+PR- HR 0.43 (15%) BIG-98 Trial 37 AI/DFS Centrally reviewed receptor values ER+PR+ HR 0.84 Ns (63%) ER+PR- HR 0.83 (20%) MA 17 Trial 38 AI/DFS Not centralized ER+PR+ HR 0.49 P = 0.02 subset analysis (73%) ER+PR- HR 1.21 (12%) IES study 39 AI/DFS Not centralized ER+PR+ HR 0.66 Ns subset analysis (55%) ER+PR- HR 0.58 (15%) ABCSG 8 /ARNO AI/DFS Not centralized ER+PR+ HR 0.66 Ns subset analysis ER+PR- HR 0.42 Ellis JM, JCO AI/RR Subset analysis ER+PR+ RR 64% Ns ER+PR- RR 53% De Placido S, Tam/DFS Subset analysis ER+PR+ HR 0.65 Ns Clin Canc Res ER+PR- HR 0.73 AI, aromatase inhibitor; Tam, tamoxifen; DFS, disease-free survival; ER, estrogen receptor; PR, progesterone receptor; HR, hazard ratio; RR, response rate; Ns, Not statistically significant.

8 HORMONE THERAPY PREDICTIVE FACTORS IN BREAST CANCER 377 therapy showed lower response rates or reduced survival in hormonally treated patients whose tumors overexpressed HER-2, reaching statistical significance in three of the five studies 42,45,46. Of note, in these cited five studies, hormone therapy was represented by tamoxifen (three studies), megace (two studies), and droloxifene (one study). It is possible that these findings are due to the failure to exclude ER-negative tumors (which likely would be over-represented in the Her-2 positive group). In three retrospective subgroup analyses published between 2000 and , HER-2 status was not associated with a significant difference in 5- or 10-year DFS. A subgroup analysis of the CALGB 8541 study 50 showed that tamoxifen effect on DFS and overall survival was independent of HER-2 status in node-positive early breast cancer treated with tamoxifen and concomitant chemotherapy (CAF). In homogeneously tamoxifentreated patients, Arpino et al. 9 showed that HER-2 is an independent predictor of tamoxifen resistance. This was confirmed by a recent retrospective analysis of the National Atlantic Treaty Organisation and Cancer Research Campaign adjuvant trials, where HER-2-positive patients failed to benefit from tamoxifen treatment. In the update of the GUN1 study 4, Carlomagno et al. 46 showed that the previously reported detrimental effect of tamoxifen in HER-2 positive tumors was no longer observed after 15 years of follow-up. However, the study confirmed a statistically significant difference in hazard ratios for death patients with tamoxifen-treated ER+/HER-2+ vs ER+/HER-2- tumors (P = 0.04) 4 (Table 4). In the adjuvant setting, the centralized tissue analysis of the TransATAC trial 27 did not confirm in HER-2+ tumors a greater relative benefit from aromatase inhibitors than tamoxifen (P = ns) (Table 4). A neoadjuvant study that randomized treatment between letrozole and tamoxifen offered a context in which this hypothesis could be tested prospectively 3. Differences in response rates between letrozole and tamoxifen were most marked for EGFR+ and/or HER-2+ ER-positive tumors (88% vs 21%, respectively; P = ) (Table 5). This suggest that EGFR and HER-2 signaling is ER-ligand dependent and that the growth-promoting effects of these tyrosine-kinase receptors on ER+ tumors can be inhibited by estrogen deprivation therapy. At this point, it is clear that with these discordant data from retrospective analyses of small groups of patients, only a study with a very large sample size or a meta-analysis in the adjuvant setting may be helpful to clarify the role of HER-2 as predictive marker. Lipton et al. 51 reported retrospective serum analyses from a trial of first line comparing letrozole versus tamoxifen for metastatic breast cancer. Overall, response rate and time to treatment failure were reduced in patients with increased HER-2 levels, without any difference in outcomes measured for patients treated with letrozole or tamoxifen. In advanced breast cancer, the presence of an elevated ECD-HER-2 correlated with a lower response to both regimens (aromatase inhibitors and tamoxifen). However, there was a statistically significant improvement in time to disease progression in patients with shedding ECD treated with letrozole versus tamoxifen, again suggesting that aromatase inhibitors may exhibit some advantage in HER-2 positive tumors 52. In a metastatic setting, a large meta-analysis of nonrandomized studies 53 reported that tissue-her-2 positivity is associated with a significantly greater risk of endocrine therapy failure, regardless of the agent tested. It is unclear whether this reflects HER-2 s value as a prognostic marker, a predictive marker, or both. The question remains whether one class of hormone agents may be more effective in patients with HER-2+ disease. Available clinical data are inconclusive to support the preferential use of aromatase inhibitors over tamoxifen in postmenopausal women with ER+/HER-2+ metastatic breast cancer 54. Table 4 - ER+HER2- versus ER+HER2+ subgroups: HRs for DFS from adjuvant clinical trials of tamoxifen and aromatase inhibitors Reference Adjuvant Subgroups HR P therapy of pts (95% CI) De Placido S, Tamoxifen ER+/HER2- HR CCR vs ER+/HER2+ HR 1.09 ER+/EGFR- HR vs ER+/EGFR+ HR 0.76 TransATAC- Tamoxifen ER+HER2- HR 0.92 Ns Dowsett or anastrozole vs ER+HER2+ HR 0.66 M-BCRT up-front HRs, hazard ratios; DFS, disease-free survival; Ns, not statistically significant. Table 5 - Difference between letrozole and tamoxifen in different subgroups of patients Subgroups End point Letrozole Tamoxifen OR (odds ratio) P difference ER+PR+ Response rate 64% 41% ER+EGFR+ and/or HER2+ Response rate 88% 21% ER+EGFR- and/or HER2- Response rates 54% 42% Modified from Ellis JM, JCO

9 378 F RASTELLI, S CRISPINO Combined markers: ER+/PR+/HER-2+ The evidence of the cross talk between the two different signaling pathways suggests that combining treatments that target these different pathways may improve outcome over monotherapy. The randomized, controlled, open-label, multicenter, phase III TAnDEM study 55 was performed to evaluate the efficacy and safety of trastuzumab plus anastrozole compared to anastrozole alone in postmenopausal women with HER-2-positive and ER+ and/or PR+ metastatic breast cancer. Combination of herceptin with anastrozole in this subset of triple-positive metastatic breast cancer led to doubling of progression-free survival (7.7 vs 3.8 months) and significant improvements in clinical benefit rate, time to progression (4.8 vs 2.4 months) and overall risk ratio compared to treatment with anastrozole alone. There was a trend to a longer overall survival in the combination arm: 28.5 vs 23.9 mo (P = 0.325) despite crossover of more than half of the patients in the anastrozole arm to receive herceptin upon progression. Overall survival, in patients without liver metastasis, was significantly improved for the combined arm, with a median of 41.9 months versus 32 months (P = ). This is the first trial to show in a randomized setting that HER-2 positivity can be used to guide biologic treatment with trastuzumab in combination with an aromatase inhibitor in metastatic disease. Studies to explore the use of this approach in adjuvant treatment are overdue. Gene profile The Oncotype DX assay was initially evaluated in patients with ER-positive, lymph-node-negative breast cancer enrolled in the NSABP trial B-14 who had received 5 years of tamoxifen and who had been followed for at least 10 years. The Oncotype DX recurrence score was predictive for distant relapse, and in a multivariate model, it outperformed other established prognostic clinical factors, such as tumor size and grade. Recurrence score was predictive for relapse when evaluated according to risk categories. For example, the risk of distant recurrence at 10 years for patients with a recurrence score of <18, 18-30, or >30 was 7%, 14%, and 31%, respectively. These groups were therefore designated low, intermediate, and high risk. Approximately one half of the patients were low risk by this definition, 25% were intermediate risk, and 25% were high risk. These investigators also found that the score correlated as a continuous variable. In other words, with a specific recurrence score, one could plot a specific risk of relapse. They then studied archival tissue from a group of patients enrolled in NSABP trial B In the trial, patients who had ERpositive, lymph-node-negative breast cancer were randomized to receive tamoxifen for 5 years or tamoxifen plus chemotherapy, which consisted either of cyclophosphamide, methotrexate, and 5-fluorouracil or methotrexate and 5-fluorouracil. In the study, the addition of chemotherapy significantly reduced the risk of recurrence in patients who had a high recurrence score, but not in those who had a low or intermediate recurrence score. Genomic assays [two-gene (HOXB13/IL17BR) ratio] In the original study by Ma et al. 57, frozen samples of ER-positive breast cancer after adjuvant tamoxifen therapy were examined by a gene micro array. In the analysis, a two-gene (HOXB13/IL17BR) ratio, derived from a gene micro array in frozen samples, was independently associated with distant relapse-free survival. Jansen et al. 58 tested the ratio of the same two genes in 112 frozen primary ER-positive breast cancer samples treated with first-line tamoxifen therapy and reported a significant association with objective response (P = 0.04). A more recent article by the same authors 59 provided evidence that high HOXB13-to-IL17BR expression levels are associated with tamoxifen failure (odds ratio; P = 0.001) and a shorter progression-free survival (P = 0.001). Gene expression evaluation from these reported studies is not comparable and we do not know whether the two-gene ratio test has a true predictive value in tamoxifen-treated breast cancer patients. Genetic polymorphisms Relatively little work has been done to identify genetic variants that might predict response/toxicity to tamoxifen or aromatase inhibitors 60. Previous studies have shown that genetic variation in CYP2D6 is associated with a higher risk of relapse in tamoxifen-treated patients 61. Tamoxifen seems to be less active in poor metabolizers (CYP2D6*4/*4), but no extensive data are available for metastatic breast cancer. ER alpha Pvu II genotypes were shown to influence tamoxifen levels in adjuvant breast cancer treatment: only one study reporting this relationship has been published, and no data are available regarding aromatase inhibitors 62. Selected CYP19 polymorphisms have been investigated in paraffin-embedded breast cancer samples for their possible association with the efficacy of aromatase inhibitors. Albeit no correlation with age, HER-2 expression, Ca 15.3 levels, ER or PR was found, the authors reported that the presence of single nucleotide polymorphisms on the 3 untranslated region of the CYP19 aromatase gene was associated with improved aromatase inhibitors efficacy and could help to select patients 63. The GIM 5 project, led by Gruppo Italiano Mammella, is an ongoing Italian multicentric study designed with the aim to establish a possible correlation of gene CYP19 with letrozole efficacy in postmenopausal early breast cancer patients after tamoxifen. In the current literature, numerous studies have focused on CYP19 gene polymorphisms, but the clinical significance of most of those polymorphisms remains undefined.

10 HORMONE THERAPY PREDICTIVE FACTORS IN BREAST CANCER 379 Discussion The EBCTCG overview analysis 26 found a correlation between ER status and response to adjuvant tamoxifen, but no further benefit was associated with positive PR status. The principal evidence deriving from recent 28 and historical studies is that higher ER scores are related to better outcome in ER+ breast cancer patients (Figure 1). The increased sensitivity of IHC for ER has lead to dichotomization of ER status ( any-or-none method), but this may result in loss of valuable clinical information 12. It still has to be clarified whether the lack of a difference in risk reduction between ER+/PR+ and ER+/PR- subgroups in the Oxford meta-analysis 1 is the result of inaccuracies in measuring PR content in different laboratories around the world rather than lack of a clinical or biologic significance of the receptor. However, PR positivity was a positive prognostic factor regardless of therapy. PR expression quantification by the Oncotype DX assay 27 confirmed that PR gene expression levels are primarily prognostic in untreated patients and not predictive of a tamoxifen benefit. Based on a small study of neoadjuvant therapy 3 in which aromatase inhibitors performed better than tamoxifen in a growth-factor-rich environment, subsequent studies in an adjuvant setting explored the hypothesis that PR status could be used to select a cohort of patients with ER+ disease who would benefit most from adjuvant aromatase inhibitors. Unfortunately, data from upfront studies, ATAC 36 and BIG , are discordant. The controversy arises when sequential studies (i.e., MA ) are compared with those upfront. In the MA.17 study, the ER+PR+ subgroup seemed more responsive to extended letrozole therapy (Table 3). However, in the trial there was a selection bias because only tamoxifen-responsive tumors were randomized and ER+/PR+ patients constituted 73% of the whole population study. The ER+/PR- subgroup represented only 12% of the tumors, because most of them has recurred during tamoxifen treatment and before randomization (P = 0.02; Table 3). Similarly, in early switch trials (after 2-3 years of tamoxifen) (IES, ABCSG 8, ARNO 95) 39,40, the lack of significance of PR-negative status has been attributed to the fact that most recurrences in ER+/PR- tumors actually occurred before the switch. At present, there are too many conflicting data from upfront aromatase inhibitor trials 36,37 and selection bias in early and late-switch studies to rule out a possible predictive role of PR status. For this reason, we fully support the recommendation of Goss et al. 38 : caution against using these results for clinical decision-making. PR status has only a strong time-dependent prognostic value and it has not yet been validated as a predictive factor to drive a specific endocrine-therapy choice. Caution is needed before concluding that PR loss simply represents an epiphenomenon of HER-2 overexpression or that these markers provide comparable predictive information. A possible mechanism is that enhanced EGF signaling reduces PR levels and that these ER+ tumor cells respond less completely to antiestrogen treatment than ER+/PR+ tumor cells. These concepts have recently been confirmed and extended with reference to human HER 2-neu 64 and insulin-like growth factor receptor signaling 65. Experimental studies have also demonstrated that the activation of both RAS/RAF/MEK/MAPK and PI3K/AKT/ mtor, the two major intracellular signaling pathways of HER-2 growth factor receptor, leads to phosphorylation and activation of ER. An indirect evidence of the AKT activity in estrogen-resistant breast cancer cells is that the blockage of AKT by mtor (mammalian target of rapamycin) inhibition restores endocrine sensitivity. Furthermore, a negative regulator of AKT a tumor suppressor phosphate and tensin homologue deleted in chromosome 10 is frequently inactivated in breast cancer progressing during tamoxifen treatment. The lesson from the TransATAC analysis 27 is that none of the individual values or combinations of values for ER, PR or HER-2/neu could aid in the selection of tamoxifen vs anastrozole. Several data suggest a causal association between overexpression and/or aberrant activity of the HER- 2/neu signaling pathway and antiestrogen resistance in human breast cancer. Independently of the type of hormone therapy used, breast cancer cells tend to become resistant to anti-estrogen therapy over time, progressing from an estrogen-sensitive to an estrogen-independent phenotype. Several mechanisms have been proposed to explain the development of this resistant phenotype. These include: reduction or loss of ER (alpha) expression (a rare occurrence), altered expression of co-regulators/co-repressors, cross talk between EGFR/HER-2 and ER, and metabolism of hormonal agents (CYP2D6 variants for tamoxifen). Co-factors, like AIB1/SRC-3, provide a mode for achieving specificity in target gene expression and provide for the ligand- and tissue-selective activities of the ER. Elevated expression of AIB1 in breast and ovarian cancers is coupled with hyperactivated growth factor kinase pathways (i.e., HER-2/neu) which target AIB1, thereby enhancing their activities on ER 66. Signal transduction inhibitors may be effective in hormonal-resistant breast cancer. Other than trastuzumab, another targeted approach is represented by the adenosine three phosphate-competitive small molecules that bind to the ATP site in the receptors catalytic domain and inhibit signal transduction. Recent studies have shown that using small tyrosine kinase inhibitors of ErbB1 (gefitinib, erlotinib) and a dual EbB1- ErbB2-TK inhibitor (lapatinib), it is possible to enhance the action of selective ER modulators in hormone-resistant human breast cancer both in vitro and in vivo. The tyrosine kinase inhibitor ZD1839 (gefitinib) has shown promising results in inhibiting tamoxifen-resist-

11 380 F RASTELLI, S CRISPINO ant breast cancer cells not only expressing HER-1 but also HER The key to the correct individualization of endocrine therapy will probably be the ability to subtype ER+ breast cancer. It would therefore be useful for gene expression profiling studies to clearly define two distinct classes of ER+ breast cancer, the clinically indolent luminal type A and the more aggressive luminal type B 68. Three distinct subtypes of breast cancer (triple-negative, HER-2 positive, and high-level ER-positive) have been recognized clinically 69 but only recently have been found to have distinct genotypes by Perou et al. 70 (basallike, HER-2 positive and luminal type). The luminal type of breast cancer generally has high levels of ER expression and nonexistent levels of HER-2. Another approach to ameliorate the definition of ER+ breast cancer is the quantification of mrna expression of different genes by reverse-transcriptase polymerase chain reaction in breast cancer tissue. This method, in the adjuvant setting, has been adopted by Paik et al. 24 and resulted in the well-known recurrence score algorithm. This Oncotype DX gene profile based on the evaluation of 16 tumor-related genes has confirmed its predictive role for establish patients who benefit from tamoxifen treatment only, instead of chemo-hormone therapy. The implication here is that this particular molecular diagnostic test is not necessarily just a more accurate way of evaluating the expression of specific individual genes such as, for example, ER or HER-2. Rather, all of the genes evaluated in the score contribute to the risk of relapse as well as to endocrine treatment resistance. In fact, in case of a high recurrence score and endocrine resistance, low expression of ER-related genes is accompanied with overexpression of proliferation genes, probably due to the increased growth factor receptor signaling. Thus far, these molecular diagnostic tests have been shown to provide prognostic information. Now that numerous molecular markers have been shown to be prognostic, the next challenge is going to be to integrate those markers in the most cost-efficient way into clinical practice. In general, it is difficult to dissect endocrine sensitivity from inherent prognosis because ER expression and the state of tumor differentiation are related. In fact, the current generation of genomic assays for tamoxifen sensitivity all contain a combination of predictive and prognostic information. Conclusions In clinical practice, there is evidence that a substantial portion of ER-positive patients fail to respond to hormone therapy. Additional markers are therefore needed to improve the ability to predict response. The latest meta-analysis of all tamoxifen trials 26 showed that being PR positive did not make a difference with respect to response to tamoxifen therapy: relative reduction in relapse is similar in ER+/PR+ and ER+/PR- tumors (37% and 32%, respectively). The level of PR expression was primarily prognostic and not predictive of a tamoxifen benefit. Overall, aromatase inhibitors proved more effective than tamoxifen regardless of ER level, PR level, or HER-2 value. There is no clear clinical evidence to support the use of HER-2 evaluation to predict endocrine therapy sensitivity. However, it remains very important to remember the biology of HER-2-positive tumors: HER-2 overexpression may interfere with the inhibitory effect of tamoxifen on estrogen-related gene transcription. The more appropriate way to avoid the agonistic activity of the drug in HER-2+ cells may thus be estrogen deprivation through use of aromatase inhibitors. It was also clear that tumors over expressing HER-2 will require a targeted therapy in addition to hormones. Recent molecular studies 23 have shown that endocrine responsiveness is to a large extent influenced by ER-related pathways, like proliferation index and growth factor receptor activation, which need to be analyzed in order to obtain a satisfying predictive assessment in an individual patient. The major question actually is about hormone receptor status, whether one should consider it as a static characteristic isolated from other pathways or as a part of a dynamic network involving proliferation genes and growth factor receptor signaling. The ER/PR testing strategy will move towards a more functional approach, with the assessment of cross talk HER family members and the analysis of cytoplasmic protein localized downstream from membrane receptors (like G-protein RAS, mtor), which should be integrated in routine histopathologic assessment. A spectrum of differentiation among ER-positive breast cancers is illustrated by the genomic tests that combine the measurements of gene expression related to ER, HER-2, and proliferation. An alternative strategy is to measure the transcriptional output from ER activity to assay endocrine sensitivity. Epigenetic and proteomic strategies are also being developed. Time will eventually tell which of these strategies were successfully translated into clinically robust tests. References 1. Early Breast Cancer Trialists Collaborative Group (EBCTCG). The Oxford Overview. Effects of chemotherapy and hormonal therapy for early breast cancer on recurrence and 15-year survival: an overview of the randomised trials. Lancet, 365: , The National Cancer Institutes of Health Consensus Development Conference: adjuvant therapy for breast cancer. Bethesda, Maryland, USA, November 1-3, 2000, proceedings. J Natl Cancer Inst Monogr, 30: 1-152, Ellis MJ, Coop A, Singh B, Mauriac L, Llombert-Cussac A, Jänicke F, Miller WR, Evans DB, Dugan M, Brady C, Quebe- Fehling E, Borgs M: Letrozole is more effective neoadjuvant endocrine therapy than Tamoxifen for ErbB-1- and/or