Feasibility study of FDG PET as an indicator of early response to aromatase inhibitors and trastuzumab in a heterogeneous group of breast cancer patients

Breast cancer is a common, life-threatening malignancy [1]; its biology may be driven by estrogen receptor (ER) expression and/or by overexpression or amplification of the epidermal growth factor family receptor HER2. Endocrine therapies such as aromatase inhibitors (AI) and targeted monoclonal antibodies such as trastuzumab (T; Herceptin®, Genentech, South San Francisco, CA, USA) have dramatically improved breast cancer outcomes for patients with tumors bearing the appropriate marker for response: ER for AIs and HER2 for T. However, not all tumors bearing these receptors respond to targeted therapy. Furthermore, in the metastatic setting, initial response to targeted therapy is commonly followed by eventual disease progression and resistance to therapy. Additional predictive markers of treatment sensitivity and resistance could allow rapid identification of effective treatments and avoidance of futile therapy.

Ki-67, an immunohistochemical assay of tumor proliferation [2], has been demonstrated to be a biomarker of early response to chemotherapy [3] and endocrine therapy [4, 5]. In a study of 158 patients undergoing endocrine therapy (anastrozole or tamoxifen or anastrozole + tamoxifen), higher Ki-67 after 2 weeks of therapy predicted shorter recurrence-free survival, with a hazard ratio of 1.95 (95% CI 1.23 to 3.07) for 2.7-fold higher Ki-67, controlling for the effects of tumor size and quantitative ER level [4]. Data regarding the predictive utility of Ki-67 and other markers for HER2-targeted therapy are more limited. In vitro studies have suggested that the HER2-targeting tyrosine kinase inhibitor lapatinib has direct effects on pathways involving proliferation [6], supporting Ki-67 as a candidate biomarker for early response to HER2-targeted therapy. However, other evidence suggests that the primary mechanism of the HER2-targeted monoclonal antibody trastuzumab is pro-apoptotic [6–8]. Further study is needed to understand the mechanisms of HER2-targeted therapy and to identify markers of early response [9].

Although the evidence is strong for Ki-67 as an early response biomarker for endocrine therapy, serial tissue biopsies are not always practical, and not all biopsies yield evaluable tumor tissue. Noninvasive imaging techniques which evaluate the entire tumor burden can overcome these obstacles and offer serial evaluation of tumor metabolic activity in vivo. 2-Deoxy-2-¹⁸F]fluoro-d-glucose positron emission tomography (FDG PET) holds promise as a marker of tumor glycolysis that may parallel Ki-67 [10, 11] as an indirect measure of cell proliferation [12]. Serial FDG PET can serve as a pharmacodynamic marker of response to chemotherapy in several tumors, including breast [13–15]. FDG PET is widely available and has been shown to predict response to breast cancer-targeted therapy as well as chemotherapy. For example, an agonist response to endocrine therapy or introduction of estradiol results in a ‘metabolic flare’ of FDG uptake that predicts clinical response to endocrine therapy [16–18]. FDG PET may also show an early response to the HER2-targeted agent lapatinib [14]. Based upon these data, we hypothesized that reductions in estrogen levels induced by AI therapy would result in an early decline in FDG PET standardized uptake value (SUV). This pilot study also explored whether a similar response would be seen in HER2-targeted therapy with T.

To determine whether change in FDG PET SUV at 2 weeks into therapy concurs with the 2-week proliferative index (Ki-67), we paired imaging and tissue assays in two groups: patients with ER + disease initiating AI therapy and patients with HER2+ disease initiating T therapy. We hypothesized that change in FDG SUV would parallel post-therapy Ki-67.

FDG SUV demonstrated substantial decline over a 2-week course of AI therapy. This is consistent with the reasoning that if hormone-dependent breast cancers respond to estradiol challenge with an increase in FDG uptake [16–18], then tumors dependent upon the ER pathway for growth should display a decline in FDG uptake as AI therapy lowers estradiol levels. These early declines in FDG uptake corresponded to low post-therapy proliferation (Ki-67), which predicts early response to endocrine therapy [4]. For all 11 patients in the AI group who showed an SUV decline of 20% or greater, the Ki-67 at follow-up was ≤5%; for all three AI patients who did not attain SUV reductions of 20% or greater, the Ki-67 at follow-up was >5%. This supports the feasibility of serial FDG SUV as an early indicator of response to endocrine therapy.

For trastuzumab, 6 of 12 patients showed substantial decreases in FDG PET SUV, including 2 of 3 patients with serial biopsy, but none of the three 2-week biopsies had Ki-67 expression ≤5%. Although numbers are too small to draw conclusions, a less consistent pattern of change is perhaps not surprising. Earlier studies have suggested that T may mediate apoptotic pathways [7], rather than resulting in an early change in proliferation. In clinical practice and in most studies, T is given in combination with chemotherapy. Data emerging from recent neoadjuvant clinical trials suggest that HER2+ disease can be effectively targeted with additional agents aimed at HER2 such as lapatinib [26] and pertuzumab [27]. Further study is needed before drawing any conclusions about the efficacy of FDG PET as an early indicator of response to trastuzumab or other HER2-targeted therapies.

We did not see correspondence between change in FDG uptake and change in Ki-67 with AI therapy. Although post-therapy Ki-67 may be a better biomarker than the change in Ki-67 (in the context of predicting breast cancer recurrence) [4], other possible contributing factors are attributable to the preliminary nature of our pilot study. Of the 17 baseline tissue samples available for pre-post comparisons, 11 were archival (8 AI 3 T), with intervening therapies including cytotoxic chemotherapy, endocrine therapy, and HER2-targeted therapy. Small sample size and heterogeneity of disease characteristics and tissue biopsy sites may also confound our comparisons of imaging parameters to the change in Ki-67.

Another limitation of our study was the timing of imaging and biopsy. Our intentions were to assess the same lesion by imaging and biopsy and to schedule FDG PET scans before biopsy. However, clinical scheduling (diagnostic biopsy preceding PET scans) and feasibility concerns (i.e., imaging of lesions near a high-contrast organ or biopsy of an inaccessible tumor) had to be accommodated. We did not find evidence that post-biopsy inflammation strongly affected SUV. For three AI patients with baseline FDG PET within 21 days after biopsy of the index lesion (6, 13, 17 days), the percent changes in SUV were −28%, −7%, and −3%. For the three AI patients with post-therapy FDG PET within 21 days after biopsy (3, 5, 7 days), the percent changes in SUV were −14%, −47%, and −60%. If post-biopsy SUV were inflated, the opposite trend (greater decrease when the baseline scan occurred post-biopsy, lesser decrease when the post-therapy scan occurred post-biopsy) would be expected. Extension of run-in periods may be considered to avoid having post-therapy scans within a healing period after the baseline biopsy.

Our choice of a 20% threshold as attributable to therapy was motivated by our hypothesis that short-term exposure would produce a pharmacodynamic effect detectable by FDG PET and beyond the limits of agreement in test-retest studies. This is a different task than the estimation of clinical response, for which alternative measures such as PERCIST 1.0 have been suggested. Of note, 8 of the 20 AI group patients with early response by FDG PET SUV had decreases between 20% and 30%, and 4 of those had absolute change in SUV of less than 1.2. Alternative measures of FDG uptake may provide more robust evidence of a drug response [25]; however, such analysis was beyond the scope of this pilot study.

This pilot study supports the use of FDG PET to examine early response to endocrine therapy and exposed challenges in collecting research biopsies in metastatic lesions. Early response patterns to AI therapy seen by imaging and biopsy were not apparent in response to T therapy, supporting that the findings for AI therapy were not due to technical issues in repeat imaging and/or biopsy. A further study in patients with de novo early stage disease, using paired serial imaging and tissue sampling to investigate early response to preoperative run-in targeted therapy, may determine whether imaging can replace tissue biomarkers. Our pilot data support the potential utility of molecular imaging with FDG and other tracers as an in vivo assay to measure pharmacodynamic effects of new anticancer drug therapies and combinations.