Aromatase inhibition 2013: clinical state of the art and questions that remain to be solved

Following their successful implementation for the treatment of metastatic breast cancer, the ‘third-generation’ aromatase inhibitors (anastrozole, letrozole, and exemestane) have now become standard adjuvant endocrine treatment for postmenopausal estrogen receptor-positive breast cancers. These drugs are characterized by potent aromatase inhibition, causing >98% inhibition of estrogen synthesis in vivo. A recent meta-analysis found no difference in anti-tumor efficacy between these three compounds. As of today, aromatase inhibitor monotherapy and sequential treatment using tamoxifen followed by an aromatase inhibitor for a total of 5 years are considered equipotent treatment options. However, current trials are addressing the potential benefit of extending treatment duration beyond 5 years. Regarding side effects, aromatase inhibitors are not found associated with enhanced risk of cardiovascular disease, and enhanced bone loss is prevented by adding bisphosphonates in concert for those at danger of developing osteoporosis. However, arthralgia and carpal tunnel syndrome preclude drug administration among a few patients. While recent findings have questioned the use of aromatase inhibitors among overweight and, in particular, obese patients, this problem seems to focus on premenopausal patients treated with an aromatase inhibitor and an LH-RH analog in concert, questioning the efficacy of LH-RH analogs rather than aromatase inhibitors among overweight patients. Finally, recent findings revealing a benefit from adding the mTOR inhibitor everolimus to endocrine treatment indicate targeted therapy against defined growth factor pathways to be a way forward, by reversing acquired resistance to endocrine therapy.


Introduction
The history of endocrine therapy for advanced breast cancer started more than a decade ago with the seminal discovery by George Beatson (1896) that ovarian ablation may cause tumor regression in premenopausal women. While ovarian estrogen synthesis ceases at the menopause, postmenopausal women still have plasma estrogen levels present at low concentration. Previously believed to occur by adrenal glandular synthesis, it later became clear that the adrenals are contributors of circulating androgens, subsequently converted into estrogens in different body compartments (Lønning et al. 1990). The idea that estrogen ablation might also work in postmenopausal women triggered implementation of adrenalectomy as well as hypophysectomy in the 1950s (Luft et al. 1952, Huggins & Dao 1953, Fracchia et al. 1967, 1971.
The fact that adrenalectomy as well as hypophysectomy was an effective antitumor therapy albeit at a cost of high morbidity motivated trials on 'medical adrenalectomy', leading to testing of glucocorticoids (Kofman et al. 1958, Lemon 1959, as well as adrenal enzyme inhibitors like ketoconazole . While the response to these drugs was inferior to surgical adrenalectomy/ hypophysectomy, these attempts, by chance, paved the way for aminoglutethimide, subsequently leading to implementation of aromatase inhibition for breast cancer therapy.
The details of how aminoglutethimide was introduced for breast cancer therapy has been outlined elsewhere (Lønning & Kvinnsland 1988). It was originally developed as an unsuccessful antiepileptic compound causing significant adrenocortical toxicity. Following an initial clinical observation revealing efficacy of aminoglutethimide in a breast cancer patient by Ralph Cash et al. (1967), Richard Santen et al. (1974) systematically implemented aminoglutethimide in concert with glucocorticoid substitution in an attempt to achieve an effective medical adrenalectomy. Their studies demonstrated clinical efficacy of aminoglutethimide for the treatment of postmenopausal breast cancer . In addition, their systematic endocrine studies revealed surprising findings with respect to its mechanism of action. Contrary to expectations, they found adrenal androgen synthesis to be preserved (Samojlik et al. 1980) despite profound suppression of plasma estrogen levels (Santen 1981). Thus, in a seminal study, they confirmed aminoglutethimide to act as an aromatase inhibitor in vivo (Santen et al. 1978), introducing aromatase inhibition as a novel concept in breast cancer therapy. Studies conducted several years later revealed aminoglutethimide, in addition, to enhance estrogen metabolism (Lønning et al. 1987(Lønning et al. , 1989a. In parallel, Harry and Angela Brodie experimentally worked on androstenedione derivatives for aromatase inhibition (Brodie et al. 1977(Brodie et al. , 1983, leading to the first pilot trial revealing anti-tumor efficacy of 4-hydroxyandrostenedione in metastatic breast cancer (Coombes et al. 1984).
Following development of aminoglutethimide and 4-hydroxyandrostenedione (known as lentaron), other so-called second-generation aromatase inhibitors, including fadrozole, were developed; for a detailed review of clinical studies evaluating first-and second-generation aromatase inhibitors for metastatic breast cancer, the readers are referred to a previous review (Lønning 2004). None of these compounds are in clinical use any longer. In brief, while some of these compounds, like fadrozole and 4-hydroxyandrostenedione, were associated with less side effects compared with standard treatment regimens at that time, anti-tumor effects were not superior to the effect of aminoglutethimide or other contemporary regimens like tamoxifen and progestins administered at high pharmacological doses (see Lønning (2004) for references). However, the lessons learned from the clinical use of these compounds, in concert with translational research evaluating their endocrine effects, provide important information to our understanding of key principles related to treatment with aromatase inhibitors.

Endocrine principles of aromatase inhibition
In postmenopausal women, estrogens are synthesized in most of the body compartments, including the liver, muscle, connective tissue, and skin . While one single aromatase gene exists, the gene contains at least ten different promoters (Bulun et al. 2003), with different promoters and ligands regulating estrogen synthesis across different tissue types (Agarwal et al. 1996, Clyne et al. 2004, Mendelson et al. 2005, Zhou et al. 2005. Notably, these promoters play a different role in benign vs malignant breast tissue; although the 1.4 promoter is the main activator in normal breast tissue, promoters II, 1.3, and 1.7 have been shown to play a role in addition to 1.4 in breast cancer tissue (Bulun et al. 2003). However, proteins coded for by the different promoters are similar. The aromatase is able to convert testosterone into estradiol (E 2 ) and androstenedione into estrone (E 1 ). While circulating androstenedione as well as testosterone in postmenopausal women is considered of adrenal origin, the ovary seems to provide a minor, albeit significant, contribution of circulating testosterone , Sluijmer et al. 1995, Couzinet et al. 2001. These circulating androgens are taken up by the different tissue compartments for subsequent aromatization.
Circulating androstenedione levels (4-5 nM) exceed the circulating levels of testosterone (0.5-1 nM) by a factor of about 5-10 (Geisler et al. 1995); in addition, the aromatase enzyme has a four-to fivefold higher affinity for androstenedione compared with testosterone (Lønning et al. 1990). Thus, aromatization of androstenedione into E 1 is the major pathway of estrogen synthesis in postmenopausal women. While E 1 is inactive by itself with respect to stimulating estrogen receptor activation, it is easily converted to E 2 by multiple dehydrogenases (Haynes et al. 2010).
Considering plasma estrogen levels in postmenopausal women, plasma E 1 (50-70 pM) exceeds E 2 (12-20 pM) concentrations by a factor of 4-5. In addition, postmenopausal women reveal plasma concentrations of E 1 sulfate (E 1 S) of about 4-600 pM (Geisler et al. 2008b). To be biologically active, E 1 S has to be deconjugated prior to conversion into E 2 .

Measuring in vivo aromatase inhibition
A major problem in evaluating the biochemical efficacy of aromatase inhibitors in vivo relates to technical difficulties measuring estrogen concentrations in the low concentration range. To achieve robust methods to assess in vivo aromatase inhibition and compare efficacy of different aromatase inhibitors, in collaboration with Mitch Dowsett's group, we developed a sensitive method for in vivo aromatization assessment based on combined 3 H-androstenedione and 14 C-E 1 injections, followed by measurement of the isotope ratio in urinary estrogen metabolites (Lønning et al. 1989b. A formal assessment of this method revealed a sensitivity indicating an ability to detect aromatase inhibition of O99.1% in the majority of patients (Dowsett et al. 1995).
Using this method, we systematically classified different aromatase inhibitors , 1994, 1995, Geisler et al. 1996b based on their efficacy in inhibiting total body in vivo aromatization (Table 1). The findings provide three key messages; first, while it has been unclear whether the three-dimensional structure of the aromatase protein allows combined binding of a non-steroidal and a steroidal (4-hydroxyandrostenedione or exemestane) substrate-pocket binding compound, we found that adding aminoglutethimide to 4-hydroxyestrone enhanced aromatase inhibition, a finding consistent with data on plasma estrogen values with the same combined regimen (Geisler et al. 1996a). Secondly, there is a difference between 'third-generation' aromatase inhibitors and previous compounds regarding in vivo efficacy. Notably, each of the three so-called third-generation inhibitors in current use for breast cancer treatment (exemestane, anastrozole, and letrozole) causes on average O98% inhibition in individual patients. In contrast, the firstand second-generation inhibitors cause aromatase inhibition of !90%. Thirdly, this difference seems to be translated into clinically meaningful effects, as the third-generation inhibitors, in contrast to the first/ second-generation compounds, have revealed clinical superiority compared with other endocrine treatment regimens (see below).

Evaluating plasma estrogen levels in patients on treatment with aromatase inhibitors
While in vivo tracer injections provide the 'gold standard' when measuring in vivo aromatization and the endocrine efficacy of different aromatase inhibitors, the method is time-and source-consuming and may be applied on a limited number of patients for research purposes only. It may sometimes be necessary to determine on-treatment plasma estrogen levels in relation to treatment with different aromatase inhibitors as part of quality control.
While in vivo tracer studies indicate that thirdgeneration aromatase inhibitors may inhibit total body estrogen synthesis by O98%, there are several studies reporting plasma estrogen levels to be sustained at 20-40% of pretreatment levels on therapy. As for most of these studies, clearly the assays applied did not have the sensitivity required for such a low concentration analysis. Taking normal plasma levels of E 1 and E 2 into account, assays with a sensitivity limit of 5-7 and 1-2 pM respectively is needed to detect O90% suppression of plasma hormone levels during treatment with an aromatase inhibitor. In addition, for steroidal compounds like exemestane, there may be cross-contamination by the drug itself or some of its metabolites, requesting preanalytical sample purification by chromatographic methods (Johannessen et al. 1997). Misinterpretations  (1998) due to technical problems carry a potential hazard; one example includes the use of local estrogen application for gynecological complications. Such local application leads to a modest but significant increase in plasma estrogen levels (Kendall et al. 2006, Wills et al. 2012) that may easily be overlooked with the use of less sensitive analytical methods, leading to potential erroneous conclusions regarding its safety in patients on treatment with an aromatase inhibitor. Notably, while there is no international standardization regarding measurement of estrogens in the low concentration range, a few laboratories around the world over the years have put much effort into developing highly sensitive and specific assays for that purpose. Thus, the group headed by Mitch Dowsett in London (Dowsett et al. 1987, Dixon et al. 2008) as well as our own group (Lønning & Ekse 1995, Geisler et al. 2008a) have developed such highly sensitive assays. Using these methods, we detected plasma estrogen suppression of O90% with the third-generation aromatase inhibitors (Geisler et al. 1996b, 2008b, Dixon et al. 2008, a finding consistent with the results obtained from tracer studies.

Normal breast tissue and intratumor estrogen levels
It has been challenged to what extent findings related to plasma estrogen suppression and total body aromatization may reflect alterations at the tumor tissue level. There are two main reasons for such concerns: first, it has been known for decades that estrogen levels are higher in tissue compared with plasma (van Landeghem et al. 1985). Secondly, immunostaining has revealed aromatase protein expression in breast cancer tissue .
If the estrogen levels are elevated in normal as well as cancerous breast tissue due to local synthesis of estrogen, this would have significant implications in breast cancer prevention as well as therapy. First, it opens the possibility that for some tumors, lack of response could be due to inefficient tissue estrogen suppression, not detected by total body tracer studies or plasma estrogen measurement. Secondly, the fact that local aromatization is regulated by tissue-specific promoters and ligands raises the possibility of 'targeted' or local estrogen synthesis inhibition (Fig. 1). Such therapeutic or preventive strategies may offer great advantages, omitting unwanted side effects from systemic estrogen deprivation.
To address the topic of local estrogen production, we measured the levels of breast cancer, normal tissue, and plasma estrogens, and correlated hormone levels with the expression of hormone-modulating enzymes, including the different steroid dehydrogenases as well as aromatase and sulphokinase/sulphatase levels , Haynes et al. 2010). The results have been discussed in detail elsewhere (Lønning et al. 2011). In brief, no correlation between intratumor estrogen levels and intratumor aromatase expression levels was found. Rather, normal breast and breast cancer tissue estrogen levels in general reflect plasma estrogen levels due to rapid equilibrium between the compartments (Fig. 1). The reason why tissue E 2 and E 1 levels exceed plasma concentration probably reflects the lipophilicity of these compounds; in contrast, plasma levels of E 1 S exceeds tissue concentration. In addition, there seems to be an intra-tumor conversion of E 1 into E 2 by local dehydrogenases. Finally, a significant amount of elevated intratumor E 2 reflects an intratumor pool of estrogen receptor-bound hormone.
Consistent with these findings, studies by William Miller et al. (1998) in Edinburgh as well as our own team (Geisler et al. , 2008b have documented third-generation compounds like anastrozole and letrozole to consistently suppress intratumor estrogen levels as well. So far, there is no evidence indicating that local tumor estrogen synthesis may be a cause of therapy failure in patients on treatment with a third-generation aromatase inhibitor. Local vs total body aromatization as a source for estrogen. There is an extensive exchange of estrone (E 1 ) and estradiol (E 2 ) between the plasma and the breast and breast cancer tissue due to total body aromatization, which overrules the local aromatization in the breast. Local administration of an aromatase inhibitor is therefore not a rational strategy. A, androstenedione; E 2 -ER, estradiol bound to estrogen receptor; E 1 S, estrone sulfate.

Aromatase inhibitors in the adjuvant setting
The findings from major studies comparing thirdgeneration aromatase inhibitors with tamoxifen for adjuvant treatment (Kaufmann et al. 2007, Forbes et al. 2008, van de Velde et al. 2011, Bliss et al. 2012, Dubsky et al. 2012, Boccardo et al. 2013) are summarized in Table 2. These studies evaluated two treatment approaches, aromatase inhibitor monotherapy or sequential treatment, where 2-3 years of tamoxifen is followed by an aromatase inhibitor. In addition, one study (BIG 1-98) also included a fourth arm; patients randomized to 2 years of letrozole, followed by 3 years with tamoxifen (Mouridsen et al. 2009). The rationale for the sequential approach was based on the findings from studies in the metastatic setting, revealing lack of crossresistance between tamoxifen and aromatase inhibitors; thus, the idea was that switching from tamoxifen to an aromatase inhibitor during 5 years of adjuvant therapy may prevent acquired resistance from developing. Notably, a combined meta-analysis of these data ) did not reveal superiority for any of the three compounds (anastrozole, letrozole, or exemestane) compared with any of the two others.
The key findings, based on this meta-analysis, are summarized in the following. Regarding the two major strategies (aromatase inhibitor monotherapy for 5 years after surgery, alternatively, tamoxifen for 2-3 years to be followed by an aromatase inhibitor for 3-2 years for a total duration of 5 years), each strategy revealed superiority compared with tamoxifen monotherapy in preventing recurrence. Among 9856 patients allocated to monotherapy with either tamoxifen or an aromatase inhibitor, following a mean duration of follow-up of 5.8 years, aromatase inhibitor monotherapy decreased relapse rate from 12.6% (for tamoxifen) to 9.6% with an aromatase inhibitor. As for survival, there was a non-significant improvement related to aromatase inhibition, but the follow-up is still too short to fully assess this end-point. Taking sequential treatment, analyzing patients from the time of randomization between continuing tamoxifen and switching to an aromatase inhibitor, the recurrence rate was reduced from 8.1 to 5% at 3 years of follow-up from randomization with a significant 0.7% reduction in breast cancer mortality among those patients receiving an aromatase inhibitor. Based on the studies presented, aromatase inhibitors have now become standard adjuvant endocrine therapy for postmenopausal breast cancer patients. However, the data summarized above raised the question of whether aromatase inhibitor monotherapy, or sequential treatment, is the optimal strategy. In the four-arm BIG 1-98 study (Table 2), 1548 patients were randomized to tamoxifen for 2 years followed by letrozole for 3 years while 2563 patients had letrozole monotherapy (Mouridsen et al. 2009). At a median follow-up of 71 months from randomization, disease-free as well as overall survival were non-significantly inferior in the crossover compared with the monotherapy arm (hazard ratio (HR) of 1.05 and 1.13 respectively). Interestingly, a benefit was observed among node-positive but not among node-negative patients. In the TEAM study, tamoxifen for 2.5-3 years followed by exemestane for a total treatment duration of 5 years was compared with exemestane monotherapy (van de Velde et al. 2011). With a total of 9766 patients analyzed on an intention-to-treat basis and with a median follow-up of 5.1 years, no difference in disease-free survival between patients in the two arms was recorded.
As for the BIG 1-98 study, another interesting comparison was made between patients treated with letrozole upfront for 2 years followed by tamoxifen for 3 years (nZ1540) vs letrozole monotherapy (nZ1546). Here again, no difference in outcome between patients in the two treatment arms was recorded.
Taken together, this evidence advocates the use of aromatase inhibitors in the adjuvant treatment of postmenopausal women. However, so far, there are no strong scientific arguments in favor of either sequential or monotherapy compared with the alternative treatment strategy. Considering the cost per quality-adjusted life year gained related to each strategy, notably this depends on the relapse risk but, in addition, patient age at diagnosis (Lønning 2006).

Duration of adjuvant therapy
While several studies have reported no additional benefit from extending tamoxifen adjuvant therapy beyond 5 years (Fisher et al. 1996, Tormey et al. 1996, Stewart et al. 2001, recent data from the large ATLAS study (Davies et al. 2012) found 10 years of tamoxifen treatment to be superior compared with 5 years of therapy. Most interestingly, while the relative ratio of recurrence was 0.90 between the two treatment arms between 5 and 9 years from diagnosis, it dropped to 0.75 after 10C years. The fact that the benefit was delayed may explain why it was overlooked in previous studies. Furthermore, it emphasizes the importance of long-term follow-up in studies evaluating benefit from endocrine treatment.
Notably, three phase III studies have evaluated the effect of adding an aromatase inhibitor following 5 years of tamoxifen. In the MA.17 trial, patients completing 5 years of tamoxifen treatment were randomized to letrozole vs placebo ; the study had to be terminated early (median follow-up of 30 months) and the patients were unblinded to treatment arm due to the extent of benefit (HR for relapse reduced to 0.58 by letrozole treatment). The results further lead to the termination of the NSABP B-33 trial (Mamounas et al. 2008) comparing exemestane with placebo, with a median follow-up time of 30 months, where a HR in favor of exemestane treatment of 0.68 was recorded. However, the effect was not statistically significant (PZ0.07) due to the limited number of patients (nZ1598) enrolled prior to early termination. Finally, the open-labeled Austrian ABCSG-6a evaluated the benefit of adding anastrozole for 3 years following 5 years of tamoxifen. At a median follow-up of 62.3 months, a benefit for extended therapy with the aromatase inhibitor was recorded (Jakesz et al. 2007).
While today current practice implements the use of aromatase inhibitors at an early stage during the sequence (upfront or after 2-3 years), the results from the ATLAS trial, together with the findings from studies on the extended use of aromatase inhibitors, challenge the concept of 5 years on endocrine therapy as the optimal duration. Currently, there are several studies comparing extended vs 5 years of endocrine therapy with aromatase inhibitor regimens (Table 3).

Aromatase inhibitors in the neoadjuvant setting
Pre-surgical systemic therapy offers the benefit of downstaging tumors to allow more limited surgery ). In addition, it offers a unique setting to evaluate potential predictive factors (Lønning 2003) as well as changes in molecular parameters , Lønning & Knappskog 2013 in response to drug therapy. Several studies (Eiermann et al. 2001, Semiglazov et al. 2005, Cataliotti et al. 2006) have compared third-generation aromatase inhibitors to tamoxifen as primary medical therapy. Whereas letrozole (Eiermann et al. 2001) revealed superiority over tamoxifen, no statistically significant benefit of anastrozole compared with tamoxifen was observed in the two trials performed with this compound , Cataliotti et al. 2006. Regarding exemestane, one study revealed a benefit in response rate compared with tamoxifen (Semiglazov et al. 2005), but the number of patients was too small for statistical comparison.
During the last few years, the proliferation marker Ki67 has become an important surrogate marker for response to endocrine therapy in the neoadjuvant setting. Comparing the percentage of Ki67 C proliferating cells before and after 2 weeks of endocrine therapy will indicate those patients with ERC breast cancer that are likely to respond with tumor regression and furthermore predict their long-term outcome (Dowsett et al. , 2007. Patients with a substantial drop in Ki67 have been shown repeatedly to achieve the best response to such treatment (Dowsett et al. 2011a). A further benefit of Ki67 measurement is the early identification of patients with treatment failure, as increasing Ki67 will later translate into clinical tumor progression (Dowsett et al. 2011b). In the PeriOperative Endocrine Therapy for Individualizing Care (POETIC) trial, this knowledge is expanded upon where patients with primary breast cancer are biopsied before and after 2 weeks on either a non-steroidal aromatase inhibitor or no treatment to identify novel biomarkers for response (Dowsett et al. 2011b).

Aromatase inhibitors in metastatic disease
The role of aromatase inhibitors in metastatic disease has been reviewed elsewhere (Lønning 2004). However, the picture has changed in recent years, since today most patients with metastatic, ER-positive breast cancer have already experienced progression on adjuvant aromatase inhibition.
While previous studies revealed the superiority of aromatase inhibitors compared with tamoxifen (Mouridsen et al. 2001), the fact that most patients with ERC tumors that relapse today have received an aromatase inhibitor for adjuvant therapy changes the scenario. As for patients relapsing say O1 year following termination of adjuvant therapy, re-implementation of the aromatase inhibitor may be a reasonable choice. In contrast, patients relapsing on treatment or shortly after terminating adjuvant therapy need alternative treatment options. Note that tamoxifen (Mouridsen et al. 2003) as well as fulvestrant (Chia et al. 2008) may have antitumor effects in patients where aromatase inhibitors fail, and the steroidal compound exemestane has been shown effective among patients becoming resistant to a non-steroidal aromatase inhibitor (Lønning et al. 2000). Interestingly, a randomized trial demonstrated a similar efficacy of fulvestrant and exemestane among patients where anastrozole or letrozole fail (Chia et al. 2008). Notably, most previous studies (Howell et al. 2002, 2004, Osborne et al. 2002 administered fulvestrant at a dose of 250 mg injections. While a later study revealed benefit from fulvestrant 500 mg injections compared with anastrozole in first line (Robertson et al. 2012), the data on fulvestrant 500 mg are not ample. Therefore, the data are insufficient to conclude that fulvestrant is superior to aromatase inhibitors, both in the first-and second-line setting. However, fulvestrant is generally well tolerated in this patient population, and the adherence to therapy is ensured by the depot i.m. injections in patients where poor compliance could be a problem.
Currently, there is no general consensus regarding optimal sequencing of endocrine therapy for metastatic breast cancer. However, for patients with metastatic ERC breast cancer, it remains important to extend endocrine treatment for as long as their disease responds, prior to implementing chemotherapy.

Side effects of aromatase inhibitors
Estrogens play a key role in many physiological processes other than reproduction. Thus, aromatase knock-out mice reveal multiple metabolic defects (Jones et al. 2001), and aromatase deficiency due to germline mutations causes Table 3 Ongoing studies comparing extended vs 5 years of endocrine therapy with aromatase inhibitor regimens. a The study compares an additional 5 years of letrozole/placebo after completing 5 years of tamoxifen and 5 years of letrozole for patients in the original MA.17 study, or letrozole/placebo years 5-10 for patients that previously got an aromatase inhibitor years 1-5.
osteopenia as well as metabolic disturbances in both genders (Morishima et al. 1995, Rochira & Carani 2009). Thus, a major concern with respect to aromatase inhibition in early breast cancer has been the enhanced bone loss as well as disturbances in lipid metabolism, which could increase the risk of cardiovascular diseases. Osteoporosis is a major health threat to the aging female population in most countries. Osteoporotic fractures are associated with a significant morbidity and excess mortality (Johnell et al. 2004). The lifetime risk for a hip fracture among European and USA Caucasian females is in the range of 15-20%. In some countries, like in Scandinavia, it may exceed 25% (Kanis et al. 2002). It is now well established that all aromatase inhibitors moderately enhance bone loss. However, most studies have addressed the effect of aromatase inhibitors on bone loss in phase III studies comparing efficacy and side effects to tamoxifen , Eastell et al. 2008, the second expressing anabolic effects on bone metabolism in postmenopausal women (Powles et al. 1996). The effects of exemestane ) as well as letrozole (Perez et al. 2006), however, on bone metabolism have also been compared with placebo, revealing a moderate loss in bone density. Notably, while ongoing treatment with an aromatase inhibitor is associated with increased bone fracture rate (Coates et al. 2007, Coombes et al. 2007, Eastell et al. 2008 in comparison with tamoxifen, any detrimental effects of aromatase inhibitors on bone metabolism disappear upon terminating the drug (Geisler et al. 2006, Eastell et al. 2008. With the encouraging results from the Austrian Breast Cancer Group, revealing that zoledronic acid may completely prevent aggravated bone loss, even among premenopausal women exposed to ovarian ablation and anastrozole in concert (Gnant et al. 2008), detrimental effects on bone metabolism may be fully preventable.
Another major concern has been with respect to detrimental effects of estrogen suppression on lipid metabolism (Engan et al. 1995) and homocysteine levels (Anker et al. 1995) that could lead to an increased risk of cardiovascular disease. As for the latter, contrary to previous claims, recent evidence suggest that plasma homocysteine may not be a major risk factor with respect to cardiovascular disease after all (Bonaa et al. 2006). For decades, estrogen replacement therapy was believed to protect against cardiovascular events in postmenopausal women. However, while hormone replacement therapy slightly elevates HDL-cholesterol levels, this effect seems not to translate into a reduced risk of cardiovascular disease (Trial 1995, Hulley et al. 1998, Alexander et al. 2001, Manson et al. 2003, Anderson et al. 2004. Considering the effects of aromatase inhibitors on plasma lipid levels, studies conducted on non-fasting subjects as well as studies on patients with metastatic disease, often suffering from metabolic disturbances, are subject to multiple confounding variables (see Lønning & Geisler (2008) for discussion and references). In the two studies evaluating the effects of an aromatase inhibitor vs placebo in early disease, both exemestane  as well as letrozole (Wasan et al. 2005) had minor effects on plasma lipid levels. Considering cardiovascular events in the phase III trials comparing aromatase inhibitors to tamoxifen (Table 4), there is no substantial evidence suggesting detrimental effects of aromatase inhibitors with respect to cardiovascular morbidity and mortality in early breast cancer (Jakesz et al. 2005, Boccardo et al. 2006, Kaufmann et al. 2007, Forbes et al. 2008, Colleoni et al. 2011, van de Velde et al. 2011, Bliss et al. 2012, Dubsky et al. 2012. A third type of side effects now receiving more attention is musculoskeletal: joint pain and stiffness, including carpal tunnel syndrome (Morales et al. 2007, Nishihori et al. 2008, Dizdar et al. 2009, Sestak et al. 2009, Mieog et al. 2012. While most patients have moderate disturbances, there is evidence that probably 20% of the patient population do not adhere to prescribed therapy with aromatase inhibitors (Partridge et al. 2008), and musculoskeletal and joint pain may be responsible for at least 50% of these withdrawals (Dent et al. 2007). Notably, Belgian investigators reported synovial deposits detectable by magnetic resonance imaging (MRI) scans among patients suffering from tendon and joint pain (Morales et al. 2008). For these patients, tamoxifen, or probably fulvestrant, may be considered as alternative treatment options. Interestingly, it seems that certain single nucleotide polymorphisms are associated with the musculoskeletal side effects of aromatase inhibitors, related to the expression of interleukin 17 receptor A . Such analysis might allow us to identify upfront patients in the future who will not tolerate aromatase inhibitors and should have another endocrine treatment.

Aromatase inhibitors and obesity
Obesity is associated with significantly elevated risk of breast cancer (Morimoto et al. 2002, Key et al. 2003) as well as a poor prognosis among postmenopausal breast cancer patients (Protani et al. 2010, Kwan et al. 2012. While the mechanisms are incompletely understood, the fact that obesity has been associated with elevated levels of E 2 in postmenopausal women (Meldrum et al. 1981, Poortman et al. 1981, as well as recent findings indicating that obesity may not influence outcome in triple-negative breast cancers (Dawood et al. 2012), indicate that elevated E 2 levels may be (at least partly) responsible for these effects. However, others (Niraula et al. 2012) have argued that obesity may confer a poor prognosis, independent of estrogen receptor levels and menopausal status.
Recent data have thrown concern over the efficacy of aromatase inhibitors among obese individuals. The Austrian ABCSG-12 trial randomized premenopausal breast cancer patients in a 2!2 trial design to either treatment with goserelin plus tamoxifenC/Kzoledronic acid or goserelin plus anastrozoleC/Kzoledronic acid. Recently, Pfeiler et al. (2011) reported overweight (BMI O25) to be associated with an enhanced relapse rate within the group of patients treated with anastrozole. In contrast, no detrimental effect of obesity was observed among patients treated with tamoxifen. Moreover, overweight individuals treated with anastrozole had a 50% increased risk of a relapse but a threefold increased risk of death compared with overweight patients on tamoxifen treatment. Analyzing the ATAC study, Sestak et al. (2010) found women with a BMI O35 to have a poor prognosis compared with lean women, independent of treatment arm (anastrozole or tamoxifen). However, there was a non-significant trend indicating a reduced benefit of anastrozole compared with tamoxifen among obese individuals. In contrast, analyzing data from the BIG 1-98 study, Ewertz et al. (2012) found the benefit of letrozole compared with tamoxifen to be independent of BMI value. The question of whether obese patients on treatment with aromatase inhibitors express elevated plasma estrogen levels compared with individuals with a normal BMI is a current issue of controversy (Diorio et al. 2012. Taken together, data at this stage do not advocate that aromatase inhibitors should be avoided among overweight patients. Importantly, there are several potential explanations to the effects observed in the Austrian trial. If the problem is failure on goserelin among obese individuals, this may be predicted to have little (if any) effect on patients treated with tamoxifen in concert, while it could be detrimental to the effect of anastrozole; it is well known that aromatase inhibitors may not prevent ovarian estrogen synthesis among preand perimenopausal women . At this stage, it may be wise to do regular plasma hormone assessment among overweight and obese premenopausal women having goserelin treatment independent of whether they receive concomitant treatment with an aromatase inhibitor or not. As for postmenopausal women receiving aromatase inhibitor monotherapy, notably, the detrimental effect of body weight observed on anastrozole treatment related to patients with a BMI O35. Further, the fact that no detrimental effect of obesity was recorded for patients treated with letrozole should be underlined. In vivo studies demonstrated letrozole to be significantly more potent than anastrozole in inhibiting total body in vivo aromatization (Geisler et al. 2002) and suppressing breast cancer tissue estrogen levels (Geisler et al. 2008b). In summary, while there may be some uncertainty related to the use of anastrozole among obese patients, data so far (at least with respect to letrozole) seem reassuring.

Molecular markers predicting benefit to aromatase inhibitors compared with tamoxifen?
While only about 10% of ER-positive tumors overexpress HER-2 (Sørlie et al. 2001, Penault-Llorca et al. 2009), notably about 50% of all HER-2-amplified tumors are positive for ER expression, although at low or moderate levels (Untch et al. 2008, Penault-Llorca et al. 2009). Comparing letrozole with tamoxifen as pre-surgical therapy, Ellis et al. (2001) reported a particular superiority for letrozole over tamoxifen in tumors overexpressing either HER-1 or HER-2. Moreover, they reported letrozole to provide a particular benefit compared with tamoxifen for patients with tumors revealing a moderate Allred ER score. However, these findings have not been reproduced in the phase III adjuvant studies. Thus, data from the TransATAC, comparing anastrozole with tamoxifen , as well as the BIG 1-98, comparing letrozole with tamoxifen (Rasmussen et al. 2008), revealed a higher relapse rate for patients with HER-2-positive vs HER-2-negative tumors in the aromatase inhibitor as well as in the tamoxifen-treated arm. The relative benefit for the aromatase inhibitor over tamoxifen, however, was similar in both patient groups. Similarly, separating patients into quartiles based on ER expression status, Dowsett et al. (2008), found the relative benefit from anastrozole over tamoxifen to be independent of ER expression status. Notably, adding either trastuzumab (Kaufman et al. 2009) or lapatinib (Johnston et al. 2009) to treatment with an aromatase inhibitor in patients with ERC/HER-2C metastatic breast cancer improves time to progression. Whether this relates to reversal of endocrine resistance or, simply, to different treatment options administered in concert is not known. However, experimental studies have revealed cross talk between HER-2 and ER signaling mediated via both the PI3K-Akt-mTOR and the Ras-Raf-MEK-MAPK pathways (Campbell et al. 2001, Knowlden et al. 2003, Jordan et al. 2004, Jelovac et al. 2005. Interestingly, the study by Johnston et al. (2009) also included patients with ER-positive tumors harboring a normal HER-2 status. While no benefit for lapatinib was recorded among HER-2-negative tumors on an intentionto-treat basis, a pre-defined analysis revealed superiority for lapatinib in a subgroup of HER-2 non-amplified tumors with an early relapse on tamoxifen. Potential interactions between the HER-2 and ER pathways should be further examined. Accordingly, we found that primary treatment with aromatase inhibitors may increase tumor HER-2 levels in non-amplified tumors (Flageng et al. 2009). Currently, the GCC 0901 study examines the effect of adding the mTOR inhibitor everolimus to patients progressing on letrozole and lapatinib in concert (ClinicalTrials.gov #NCT01499160). However, in HER-2 overexpressing breast cancer, PI3K signaling inhibition leads to increased HER-2-mediated ERK activation, pointing to yet another important growth-promoting signaling axis, the Ras-Raf-MEK-ERK pathway (Fig. 2), and the potential need for adding for instance MEK inhibitors in certain patient subgroups (Serra et al. 2011). Resistance toward treatment with aromatase inhibitors: current observations related to targeted therapy As mentioned above, there is increasing evidence indicating that cross talk between the estrogen receptor pathway and several other growth-controlling pathways may cause resistance to aromatase inhibitors (Miller et al. 2011, Sabnis & Brodie 2011. Accordingly, a large number of clinical trials are currently conducted to assess whether various signal transduction inhibitors can augment the efficacy of aromatase inhibitors in postmenopausal patients with breast cancer. These include inhibitors of the PI3K-Akt-mTOR and Ras-Raf-MEK-MAPK pathways, insulin-like growth factor 1 (IGF1) receptor (IGF1R), gamma secretase/Notch, cyclin-dependent kinase 4/6 (CDK4/6), histone deacetylase (HDAC), and Src/Abl a.o. (Fig. 2), as elaborated on elsewhere (Fedele et al. 2012). For instance, adding the CDK4/6 inhibitor PD0332991 to letrozole in the first-line treatment of patients with ERC metastatic breast cancer increased PFS from 7.5 to 26.1 months in a phase II study with 165 patients (Finn et al. 2012).
The PI3K-Akt-mTOR signaling pathway has come up as a major resistance pathway in endocrine resistance, including resistance to aromatase inhibitors (Miller et al. 2011, Villarreal-Garza et al. 2012. Activating mutation in PIK3CA, the gene encoding the p100a subunit of the PI3K protein, is present in 28-47% of ER-positive breast cancers, and is the single gene most frequently mutated in this disease (Miller et al. 2011). Currently, numerous studies are conducted where drugs targeting PI3K, Akt, and/or mTOR are tested in combination with aromatase inhibitors (Baselga 2011). Recently, two studies revealed that the mTOR inhibitor everolimus significantly improves time to progression in patients with ER-positive metastatic breast cancer undergoing endocrine therapy. The BOLERO-2 study randomized a total of 742 patients developing resistance to a non-steroidal aromatase inhibitor at a 2:1 ratio between exemestane plus everolimus vs exemestane monotherapy; here, median time to progression was extended from 4.1 to 10.6 months (Baselga et al. 2012). The effect seems not to be limited to the use of aromatase inhibitors; in a smaller study, Bachelot et al. (2012) found everolimus to improve time to progression among patients treated with tamoxifen. While this study indicated a benefit among patients with acquired resistance only, the number of patients was too low to allow for any definite conclusion.
However, there are results at variance. In a recent study, Wolff et al. (2013) found only a non-significant and modest benefit from adding another mTOR inhibitor temsirolimus to letrozole in aromatase inhibitor-naïve patients, independent of previous adjuvant tamoxifen therapy. While these results, like the findings by Bachelot et al. (2012), may indicate an effect of mTOR inhibition on acquired but not primary drug resistance across different compounds, more data are needed to draw any final conclusion. This issue will be addressed further in the BOLERO-4 phase II study, where 200 patients with metastatic breast cancer will be treated with everolimus and letrozole in the first-line setting (NCT01698918).
Furthermore, it should be recalled that the PI3K-Akt-mTOR pathway is involved in chemotherapy resistance as well (Lønning & Knappskog 2013). Additionally, mTOR inhibitors have revealed antitumor efficacy across other tumor forms, indicating that they exhibit anti-tumor efficacy by themselves, not only as adjuvants to other cancer therapies (Motzer et al. 2010, Yao et al. 2010).
An interesting extension of the combined mTOR inhibitor/endocrine therapy trials in breast cancer is what to do when the treatment fails. In the BELLE-3 trial, a pan-PI3K inhibitor BKM120 or placebo is combined with fulvestrant in patients with advanced breast cancer who have progressed on an mTOR inhibitor plus endocrine therapy (NCT01633060). This trial will address whether upstream inhibition of PI3K is worthwhile in patients where an mTOR inhibitor fails. Moreover, rapamycin analogs (rapalogs), which are the kind of mTOR inhibiting drugs currently used in the clinic, mainly inhibit mTORC1 but not the mTORC2 complex (Jacinto et al. 2004, Villarreal-Garza et al. 2012, and the consequence of this is ongoing Akt activation by mTORC2 during rapalog administration (Fig. 2). Second-generation mTOR inhibitors, like PP242 and CC-223, inhibit both mTORC1 and mTORC2 (Janes et al. 2010, Weigelt et al. 2011; however, the potential benefit from this additional effect remains to be evaluated. Another important question is whether to keep the mTOR inhibitor beyond progression. In a planned phase II study by the German Breast Group, patients who exhibit progression of their disease on exemestane and everolimus change their endocrine treatment whereas everolimus is kept as the backbone (NCT01773460).
Another possible resistance mechanism has also been identified where aromatase inhibitors fail due to plateletderived growth factor receptor (PDGFR)/Abl signaling upregulation (Weigel et al. 2013). Src is a downstream hub for various signaling pathways, like EGFR and HER-2, and a potential signaling link between non-genomic ER and HER-2 via p130Cas, which might be involved in resistance to endocrine therapy (Cabodi et al. 2004, Fox et al. 2009, Vallabhaneni et al. 2011. Src, Abl, and PDGFR can be inhibited by the oral small molecule inhibitor dasatinib, which is already registered for use in chronic myeloid leukemia with BCR-ABL mutation, resistant to imatinib. Combination clinical trials of dasatinib and aromatase inhibitors are currently ongoing. Another interesting observation regarding resistance to aromatase inhibitors relates to the in vitro phenomenon of 'estrogen hypersensitivity'. Breast cancer cells that have grown under long-term estrogen deprivation (LTED) become extremely sensitive to estrogen (Masamura et al. 1995. Whereas estrogen at high concentrations normally stimulates cell growth, it becomes cytotoxic in LTED cells . While the exact mechanism causing LTED has not been fully elucidated, upregulation of the ERa, in addition to the PI3K-mTOR, and MAPK pathways has been shown to occur in LTED cells. Taking the concept of 'estrogen hypersensitivity' to the clinic, we demonstrated that estrogen in high doses can be used therapeutically in ER-positive breast cancer with acquired resistance to aromatase inhibitors ), a finding subsequently confirmed by Ellis et al. (2009). Future aspects on aromatase inhibition and issues that remain to be settled While aromatase inhibitors have become the standard therapy for ERC postmenopausal breast cancer, several issues remain to be settled. A current issue relates to the optimal duration of therapy. The next decade will address the question of how long aromatase inhibitors should be administered to derive the maximum benefit in the adjuvant setting. In this respect, the initial results of the ongoing MA.17 extension trial and NSABP B-42 study are expected in 2015. This primarily relates to therapeutic efficacy, but long-term toxicity is another important aspect in the adjuvant setting. Whereas the issue of toxicity has been thoroughly addressed with respect to 5 years of treatment, potential side effects related to extended therapy need to be carefully monitored.
A key issue with respect to treatment with aromatase inhibitors, like most other cancer compounds (Lønning & Knappskog 2013), relates to the design of targeted strategies to prevent drug resistance in vivo. Currently, studies, like the POETIC trial, have been designed to specifically address this issue. We have an emerging understanding of which signaling pathways are involved in endocrine resistance and an increasing number of signal transduction inhibitors to target these pathways. However, the challenge will be which drugs to pick, how to combine them, and how they will be tolerated with respect to side effects (for the patient) and economic cost (for society).
Clearly, large phase III trials cannot be used when dozens of deregulated signaling pathways are to be targeted simultaneously. Importantly, new technology, such as 'next-generation deep sequencing', is becoming increasingly efficient and affordable, which, in the near future, will allow an increasing number of research laboratories to conduct complete exome sequencing of individual tumors. Such extensive understanding of each tumor's genomic profile, merged with biological evidence linking disturbances in defined growth factor pathways to endocrine resistance, has the potential to optimize endocrine therapy dramatically in the future.
In studies, such as the ongoing SHIVA study in France (NCT01771458), the molecular profile of each individual cancer is used to design targeted therapy, as opposed to conventional therapy, to compare the old and new school in oncology. The therapeutic potential of such strategies will be answered shortly.
Declaration of interest P E Lønning has received speakers honoraria and has been compensated for participating at Advisory Boards for Pfizer, Inc., Astra-Zeneca and Novartis, and for writing conference proceedings for Novartis. H P Eikesdal has received speakers' honoraria and has been compensated for participating at Advisory Boards for GlaxoSmithKline and Bristol Myers Squibb and has been compensated for writing conference proceedings for Novartis. Both authors are active participants in the Belle-3 study conducted by Novartis (P E Lønning is a member of the International Steering Committee).

Funding
The work providing the basis for this review has been supported by generous grants from Helse Vest, the Norwegian Cancer Society, and Bergen Medical Research Foundation.