Goss, PE

549 Background: MA.17 evaluated letrozole (LET) or placebo (PLAC) after 5 years of tamoxifen (Tam) and showed [median follow-up 30 months (mos)] significant improvement in disease-free survival (DFS) for LET [hazard ratio (HR) 0.57, p = 0.00008]. The trial was unblinded and PLAC patients (pts) were offered LET. An ITT analysis of all outcomes, before and after unblinding, based on the original randomization was performed. Methods: A stratified log-rank test was used to compare DFS, distant (D)DFS, overall survival (OS) and incidence of contra-lateral breast cancer (CBC). The Cox regression model used baseline stratification variables and two prespecified factors, menopausal status at the start of Tam and time on Tam. Subgroup analyses for DFS and OS were performed for the two prespecified subsets. All p-values were two-sided Results: 5187 pts were randomized at baseline and, at unblinding, 1655 of 2268 PLAC pts accepted LET. At median follow-up of 54 mos (range,16–86) 363 recurrences or CBC’s (144 LET and 219 PLAC) occurred; 118 LET and 176 PLAC pts had recurrent disease and 26 LET and 43 PLAC pts had CBC. 4 year DFS was 94.3% (LET) and 91.4% (PLAC) (HR 0.64; 95% CI, 0.52 - 0.79; p = 0.00002). Corresponding 4 year DDFS was 96.2% and 94.9% (HR 0.76; 0.58–0.99; p = 0.041). 4 year OS was 95.0% (LET) and 95.1% (PLAC) (HR 1.00; 0.78–1.28; p = 0.99). LET was equally effective in node +ve and -ve pts (i.e., similar HRs) in DFS. OS was not significantly different for LET and PLAC in any subgroup. The annual rate of CBC was 0.29% LET (0.18–0.40) and 0.47% PLAC (0.34–0.61); HR 0.61 (0.38–0.98) p = 0.037. 255 pts had died as of the data cut-off (128 LET and 127 PLAC). Conclusions: In this ITT analysis, pts originally randomized to LET within 3 months of stopping Tam did better than PLAC pts in DFS, DDFS and CBC, despite 73% of PLAC pts crossing to LET after unblinding. This highlights the strong beneficial effect of extended adjuvant therapy with LET. [Table: see text]

Abstract Hepatotoxicity is associated with small molecule tyrosine kinase inhibitors (TKI) in use for the treatment of a variety of cancers. Retrospective studies have identified and confirmed that specific Class II Human Leukocyte Antigen (HLA) alleles are strongly associated with ALT elevation in women treated with the TKI lapatinib for breast cancer. This study aimed to further evaluate and validate the role of the specified HLA alleles as predictors of elevated ALT in a pre-defined analysis of a large, randomized, double-blind, placebo-controlled study of lapatinib monotherapy in early stage HER2 positive breast cancer, the TEACH study (Tykerb Evaluation After Chemotherapy, EGF105485). This prospectively defined pharmacogenetic study compared the frequency of hepatobiliary adverse events between pre-specified Major Histocompatibility Complex (MHC) genetic variants, including the HLA alleles DQA1*02:01 and DRB1*07:01. The primary focus was on elevated ALT, as well as rare cases of concurrent ALT (>3x ULN) and bilirubin (>2x ULN) elevation, which represent possible Hy's Law cases and a high risk of acute liver failure, among 1194 patients randomized to lapatinib treatment from whom pharmacogenetic data was available. This study prospectively validated prior reports of the association of the specified MHC variants with elevated ALT among women treated with lapatinib. The strongest effects were observed for carriers of the HLA alleles DQA1*02:01 and DRB1*07:01, with odds ratios of 20 (95% CI: 8–40) between cases (n = 34) and controls (n = 807–808). These two HLA alleles are highly correlated, inherited together in most individuals and are consistent with a single genetic association. The overall risk of patients having an ALT (>3xULN) elevation was 3.0% and 0.7% during treatment with lapatinib and placebo respectively. Carriers of either HLA allele had a 12% chance (positive predictive value) of having an elevated ALT (>3xULN), in contrast to a 0.9% risk (negative predictive value, 99.1%) for non-carriers of the specified HLA alleles. These associations were maintained for higher ALT elevation thresholds and for cases of concurrent ALT and TBL elevation, consistent with possible Hy's Law cases. These results strongly support the role of Class II HLA-modulated immune mechanisms in lapatinib-induced hepatotoxicity. Our results validate the large strength of association of the HLA alleles DRB1*07:01 and DQA1*02:01 with hepatotoxicity and provide the possibility of managing the risk of hepatotoxicity in women receiving lapatinib for early or late stage HER2 positive breast cancer. This association with immune mechanisms may have implications for toxicities with other TKIs in current use in cancer patients. Citation Information: Cancer Res 2012;72(24 Suppl):Abstract nr PD10-05.

Abstract Background: ASP is hypothesized to decrease the risk of breast cancer (BRCA) and BRCA recurrences. Thus we performed an exploratory analysis of ASP and CC use within the NCIC CTG MA.27 adjuvant trial comparing exemestane (E;N=3789) to anastrozole (A; N=3787) with a factorial second randomization to CC or placebo (P). Neither A or E was superior in breast cancer outcomes. Baseline low-dose ASP use is an accepted surrogate of cardiovascular risk factors and was used as a stratification factor across all 4 arms. Randomization to CC-P was discontinued after 18 months (n = 1622) due to concerns of cardiac toxicity. Methods: Patients taking >81mg of ASP daily at baseline were ineligible for randomization and use of >81mg of ASP daily was not allowed. Women enrolled during CC randomization were included in the comparison of E and A, stratified by whether they had been randomized to CC [yes, no;N=1622] and concomitant low-dose ASP [≤81 mg/day (yes, no); N=2209]. Other stratification factors included: lymph-nodes (negative, positive, or unknown); prior adjuvant chemotherapy (yes, no). The primary endpoint, event-free-survival (EFS), was defined as time from randomization to time of locoregional or distant disease recurrence, new primary breast cancer, or death from any cause. Secondary endpoints included overall survival (OS) defined as time from randomization to time of death from any cause and distant disease-free-survival (DDFS), defined as time from randomization to time of distant disease recurrence. Univariate (uni) assessment of CC and ASP use was assessed with stratified log-rank test, adjusting for lymph-node status and adjuvant chemotherapy and applied by intention-to-treat. Exploratory multivariate (multi) analyses (N = 1622) had forced inclusion of treatment and used step-wise forward stratified Cox modeling to examine the effects of CC, ASP use and baseline patient characteristics on outcomes; a factor was added with Wald test statistic p ≤ 0.05. Results: At median follow-up of 4.1 years, 186/1622 (11%) patients had an EFS event; 125 (8%) had died from any cause, and 80 (5%) had distant BRCA relapse. CC did not have significant uni association with outcomes: EFS p-value=0.92; OS p-value p = 0.56; DDFS p-value=0.55. ASP use was associated with worse EFS [p = 0.006, HR 1.48 (95% CI 1.12–1.96)], worse OS [p = 0.0002, HR 1.87 (95% CI (1.35–2.61)], and non-significant difference in DDFS (p = 0.72). CC had no multi association with EFS, OS or DDFS. ASP use had no multi association with EFS and DDFS (p > 0.05). However, ASP use had multi prognostic association with worse OS [p = 0.01; HR 1.67 (95% CI 1.13–2.49)]. Conclusions: Users of either CC or low dose ASP had similar DDFS to non-users in MA.27. As expected, low-dose ASP users (with presumptive cardiovascular risk factors) had worse OS than non-users. Inadequate numbers of patients randomized to CC and lack of randomization to aspirin leaves inadequate evidence from MA.27 to determine whether anti-inflammatories influence breast cancer outcomes. Citation Information: Cancer Res 2012;72(24 Suppl):Abstract nr P2-13-02.

Abstract Background: Exemestane, a steroidal aromatase inhibitor, reduced the incidence of invasive breast cancers by 65% among 4560 post-menopausal randomized to exemestane or placebo for 5 years on MAP.3. Differences in quality of life (QOL) were judged to be minimal, but only summary information was reported. Purpose: To provide more detailed information about effects of exemestane on menopause-specific and health-related qualities of life. Method: Participation in quality of life assessment was an eligibility criterion. Menopause-specific and health-related qualities of life were assessed using the MENQOL (4 scales; physical, vasomotor, psychosocial, sexual) and SF-36 (8 scales; physical health, role function — physical, bodily pain, general health, vitality, social function, role function — emotional, mental health, and 2 summary scales) instruments, respectively at baseline, 6 months and then yearly after randomization. Compliance with QOL questionnaire completion at each follow-up visit ranged from 93–98%, and did not differ by group. Change scores for each MENQOL and SF-36 scale, calculated for each assessment time relative to baseline, were compared using the Wilcoxon Rank-Sum test. Summary scores were used to summarize the QOL scores observed at each time point for each SF-36 dimension and overall mental (MCS) and physical component summaries (PCS) and MENQOL domains. Clinically important worsening of MENQOL change scores was defined as an increase of ≥0.5/8 points. SF-36 change scores were considered worsened if scores decreased by ≥ 5 points from baseline. Results: Both groups were balanced on scores for MENQOL and SF-36 at baseline. Median follow-up was 35 months and the proportion of women who stopped study medication early for toxicity reasons was 15% in the exemestane arm and 11% in the placebo arm. There was a statistically significant difference in change scores for vasomotor symptoms among women on exemestane during the first 4 years (p-values <0.01), compared to placebo. However, no between-group differences in vasomotor change met the criterion for clinical importance. Women on exemestane had statistically poorer sexual functioning (mean change = −0.02, SD=1.37) compared to placebo (mean change = −0.12, SD=1.32) during the first 6 months on study (p-value = 0.03) but the differences were not statistically significant thereafter or clinically important at any time. Among the 8 SF-36 scales, only bodily pain was statistically different between exemestane and placebo for the first 24 months on study medication (p-value <0.01), but no between-group difference in change scores exceeded 5 points. Overall SF-36 PCS and MCS assessing changes in overall physical and mental health-related QOL did not differ significantly by group at any assessment. Conclusion: Our assessment that early differences in vasomotor symptoms and pain were probably not clinically important is supported by the observation of no between-group differences when overall physical and mental health-related QOL changes were compared. Exemestane does not appear to have a major negative impact on the quality of life among these women. Citation Information: Cancer Res 2011;71(24 Suppl):Abstract nr S6-1.

Abstract Background: Based on recent breast cancer literature, we hypothesized that there could be substantive differences in apparent efficacy estimates using a log-normal (LN) survival model rather than with standard Kaplan-Meier (K-M) or Cox model methods. While both Cox and LN survival analyses offer greater specification by individual patient characteristics, the LN model may more robustly estimate survival under model misspecification. Methods: We recently pooled data for 5 NCIC CTG primary breast cancer trials: MA.5, MA.12, MA.14, MA.21, and MA.27. The total patient count for patients who received at least 1 dose of trial therapy is 11,253. Compilation included definition of STEEP endpoints (C Hudis, JCO, 2008) and standardized factor categorizations. The primary endpoint is Breast Cancer Free Interval (BCFI) defined as the time from randomization until recurrence: first local invasive or DCIS; regional, or distant; contralateral invasive or DCIS; or death from breast cancer. We found substantive evidence of non-proportionality for 7 factors compiled for the meta-analyses. In this work, we fit multivariate Cox and LN models with these 7 factors, lymph node status and pathologic T status. We then compare BCFI efficacy estimates for patient and tumour characteristics at 1-, 3-, and 5-years obtained with K-M, Cox, and LN models. Results: There was evidence that the Cox assumption of proportional hazards was violated for 7 factors: age, menopausal status, hormone receptor status, anthracycline use, chemotherapy use, race, and ECOG performance status. Differences between models were intrinsically affected by timing and extent of non-proportionality; there was no consistent pattern. In particular, investigations to date indicate efficacy estimates with absolute differences between K-M, Cox and LN estimates which varied by time of assessment: at 1-year 0.0 to 6.7%, at 3-years 0.4 to 18.6%, and at 5-years 0.2 to 17.0%. BCFI estimates with the K-M were inconsistently closer to those with the LN or Cox model: for K-M to Cox at 1-year 0.4 to 5.2%, at 3-years 0.4 to 15%, at 5-years 0.4 to 14.3%; for K-M to LN at 1-year 0.0 to 6.7%, at 3-years 0.5 to 18.6%, at 5-years 0.2 to 17.0%; for Cox to LN at 1-year 0.8 to 1.8%, at 3-years 1.9 to 6.0%, at 5-years 0.6 to 5.7%. K-M and Cox models have step-wise adjustments at events for K-M and Cox, rather than smooth modeling with the LN. Discussion: Even with reasonably large population subgroups, there were substantive differences in apparent survival (0.0 to 18.6%) between K-M, Cox and LN model types. The magnitude of differences in survival estimates was large enough to be clinically relevant and warrant further consideration as we evaluate new therapies and prognostic/predictive factors. We will be statistically investigating framework robustness under differing levels of model misspecification. Citation Information: Cancer Res 2011;71(24 Suppl):Abstract nr P5-14-01.

Abstract Background Exemestane significantly reduced invasive and preinvasive breast cancers in postmenopausal women at increased risk for breast cancer in the NCIC CTG MAP3 trial with no serious toxicities, including excess fractures or osteoporosis. Purpose: To provide additional information on the effect of exemestane on bone loss in women at high risk for breast cancer, within a subset of women participating on the NCIC CTG MAP.3B study. The primary hypothesis is that exemestane does not induce clinically significant bone loss in postmenopausal women at increased risk of developing breast cancer at 2 years. The primary objective of this companion study is to examine the effect of exemestane on lumbar spine and total hip BMD by DEXA at 2 years in women participating in the MAP3 trial. Methods: The MAP.3B bone sub-study registered women from the main MAP. 3 trial from May 2008 to March 2010. Eligible women had to have an acceptable quality BMD scan by DEXA taken within 12 months prior to randomization to MAP.3. A BMD T-score >-2.0 SD (i.e. better than 2 standard deviations below the average peak BMD of a young adult woman) was established as the study population cutoff. A questionnaire including information on height, falls, fractures, lifestyle information including physical activity, tobacco and alcohol use was completed at baseline, 12 months, 24 months and at last visit. Fasting serum for bone biomarkers was collected at 12 months and total hip and L1-L4 (postero-anterior) spine BMD were measured 2 years after randomization on the same Lunar or Hologic scanner. The primary objective was to determine differences in hip and spine BMD at 2 years. Secondary outcomes include number of skeletal fractures and development of osteoporosis 2 years after randomization and changes in bone biomarkers at 1 year after randomization. For the analysis of the primary endpoints, the upper limit of a one sided 95% confidence interval for the difference in mean percentage changes between placebo and exemestane will be calculated for the BMD by DEXA at each site. We will conclude that exemestane does not induce significant bone loss in postmenopausal women at increased risk of developing breast cancer at 2 years when the upper limit is less than 3% for both sites. Similar confidence interval approach will be used to analyze the secondary endpoints. Results: Between May 2008 and March 2010, 238 postmenopausal women were recruited. Median age was 61.8 years, and the majority of women were Caucasian (91%), with approximately 20% of the participants reporting a recent fall (within past 12 months) and another 13% reporting a recent fracture prior to randomization. We will report results from the primary as well as the secondary endpoints at the SABCS meeting. Citation Information: Cancer Res 2011;71(24 Suppl):Abstract nr P4-11-13.

Abstract Background Aromatase inhibitors (AIs) are standard adjuvant therapy for postmenopausal women with hormone receptor-positive breast cancer. We previously reported that racial/ethnic minority women receiving an AI in NCIC CTG MA.17 experienced fewer hot flashes, fatigue, and arthralgias than Caucasians (Moy B et al. Ann Oncol 2006;17:1637). Here we examined whether race affected clinical outcomes in the MA.27 AI trial comparing exemestane (E) with anastrozole (A). Methods: Fisher's exact test was used to compare observed side effects (0 vs grades 1–5) between minority and Caucasian women. Adverse events (AE), including menopausal symptoms, were assessed according to the Common Terminology Criteria of the National Cancer Institute (version 3.0). ITT univariate test of race effects on overall survival (OS) was determined with a stratified log-rank test, while multivariate testing was with stratified Cox regression. Results: Among 7312 for whom race was known, distribution was: Caucasian (n=6939; 95%); black (n=235; 3%); Asian/Native Hawaiian/Pacific Islander (n=93); 1%; American Indian (n=39; 1%); and mixed race (n=6). Among women treated with E, minorities reported a significantly lower incidence of hot flashes (45% vs. 56%; p=0.003) and fatigue (34% vs. 46%; p=0.001) compared to Caucasians. Similarly with A, minority women reported significantly fewer hot flashes (47% vs. 58%; p=0.002) and lower cholesterol (12% vs. 18%; p=0.01); however, they reported more headaches than Caucasians (16% vs. 10%; p=0.01). Caucasian women were more likely to discontinue therapy due to AE or side effects (32% vs. 24%). There was a significant OS interaction (p=0.02) between race and treatment with minority women on E having fewer deaths than those on A. Caucasian women had HR of E to A of 0.98 (95% CI 0.81−1.20), p=0.85, while minorities had HR of E to A of 0.72 (95% CI 0.33−1.58), p=0.41. Conclusions: Minority women tolerated AIs better and were also more compliant than Caucasians, supporting our previous findings. A significant interaction between race and type of AI therapy for OS was seen, favoring exemestane in minority women. Replication of these findings in larger cohorts is necessary. Since race/ethnicity may serve as a surrogate for genetic differences, our findings suggest pharmacogenomic differences which require further research. Citation Information: Cancer Res 2011;71(24 Suppl):Abstract nr P2-14-02.

Abstract Background: The non-steroidal reversible aromatase inhibitor (AI), anastrozole (A) is approved as adjuvant monotherapy in hormone receptor (HR) positive early breast cancer (EBC). The steroidal irreversible AI exemestane (E) is approved in sequence after an initial 2-3 years of tamoxifen. MA27 tested the hypothesis that exemestane would have greater efficacy and better end-organ safety than anastrozole for several reasons: E is an irreversible and more potent AI than A; unlike A, E does not induce intra-tumoral aromatase and; through its mild androgenic activity E may exert a second anti-tumor effect and a more favorable bone and lipid metabolism profile than A. Initially the study also tested the role of a cyclooxygenase-2 inhibitor, celecoxib, used in combination with an AI. However, celecoxib was discontinued early in MA27 following identification of cardiovascular toxicity with COX-2 inhibitors. Methods: Postmenopausal women with hormone receptor positive primary breast cancer were randomized after adequate local treatment to 5 years of adjuvant A (1mg/day) or E (25mg/day) with or without celecoxib/placebo (400mg/plac twice daily for 3 years). The primary objective is the comparison of event free survival (EFS) between women treated with A or E. Secondary objectives include overall survival (OS), distant recurrence (DDFS), incidence of contralateral breast cancer (CBC) and safety. Quality of Life in a subset of women has been evaluated, and a very extensive tissue bank created. Stratification included: prior chemotherapy; lymph node status, celecoxib (during use), aspirin use (during randomization to celecoxib), and trastuzumab use (since November 2005). The primary endpoint (EFS) will look for an improvement from 87.5% to 89.9%; HR 0.80, 2-sided 5% level and 80% power, with 644 events required. Results: 7576 women were randomized between June 2003 and July 2008. Celecoxib/placebo, was discontinued 2 years into the trial in December 2004 after 1635 patients had received celecoxib. The number of events required for final analysis was reached April 2010. Patient characteristics were balanced for age (median 64.1yrs); race; ECOG PS; mastectomy; TNM staging and; prior adjuvant chemotherapy. EFS, OS, DDFS, and CBC as well as clinical fractures, cardiovascular events and adverse events will be presented by arm in the final analysis. Conclusions: NCIC CTG MA.27 is the first definitive EBC trial comparing a non-steroidal and steroidal AI as initial adjuvant therapy. Exemestane is not currently approved as initial adjuvant treatment and superiority or equivalence will lead to its inclusion as a first-line option for EBC in postmenopausal HR positive breast cancer. Efficacy, clinical fractures, cardiovascular events and AI related symptoms will help to define the cost-benefit ratios of these two AIs. Citation Information: Cancer Res 2010;70(24 Suppl):Abstract nr S1-1.

Abstract Background: The MA.17 study showed that extended adjuvant endocrine therapy with letrozole (LET) after completing 5 years of tamoxifen (TAM) markedly reduced the risk of recurrence in women with ER+ early stage breast cancer and improved overall survival in women presenting with node +ve disease. The HOXB 13:IL17BR gene expression ratio (signature) has been shown to predict outcome in breast cancer patients treated with adjuvant tamoxifen monotherapy and provides additional information beyond that from known positive (ER and PR) and negative (Her-1 and Her-2) predictors of responsiveness to tamoxifen in node-ve women. We report a case control evaluation of the Breast Cancer Index (BCI; bioTheranostics, Inc.), which combines the HOXB13 and IL17BR twogene and the molecular grade index (MGI) gene expression signatures, with respect to distinguishing which patients are at risk of late recurrences and who would respond to extended endocrine therapy with LET. The prognostic and predictive utility of quantitative immunofluorescence of ER, PR, Her-2, tumor aromatase, COX-2, GATA3 and Nat1 in the TAM-PLACEBO and the TAM-LET cohorts will also be evaluated and compared to results derived by standard immunohistochemistry. Methods: FFPE tumor blocks were collected from patients who experienced a breast cancer recurrence up to unblinding of MA.17. Controls were matched 2:1 for age, tumor size, lymph node status, and prior chemotherapy, and were all disease free for longer than cases. All cases were reviewed for standard histopathology by two independent pathologists. RNA was extracted, amplified, converted to cDNA and subjected to RT-PCR with primers and probes to HOXB13, IL17BR, BUB1A, CENPA, NEK2, RACGAP1 and RRM2. ER, PR HER1, HER2, COX2, Aromatase, GATA3 and NAT1 will be analyzed by routine IHC techniques and by immunoflourescent Automated Quantitative Analysis (AQuA). Results: 105 cases and 210 matched controls are available for evaluation. All sections are under review and tissue microarrays have been performed on all cases and controls. Detailed results on the BCI and ER, PR, Her-2 will be available at the SABCS. Discussion: MA.17 has shown that extended adjuvant endocrine therapy after tamoxifen is effective at preventing disease recurrence given for an additional 5 years. Numerous clinical trials are exploring whether extending AIs will show this benefit, and there is an increasing need to improve the therapeutic index by distinguishing those at risk from those who are not. It is also important to determine which patients will benefit from the therapy and which will recur without benefit. The latter patients could be triaged to clinical trials of novel therapies to overcome endocrine resistance. This study will help to define these issues and pave the way for more effective selection of specific patients for adjuvant endocrine strategies. Citation Information: Cancer Res 2010;70(24 Suppl):Abstract nr P3-10-26.

Abstract Background: Mammographic density, the variation in fat, epithelial and stromal tissues seen on screening mammography, is a strong risk factor for breast cancer and can be modified by hormonal agents. Changes in density from tamoxifen or postmenopausal hormone (PMH) use are associated with risk, suggesting that density may be a surrogate marker of therapeutic efficacy. Aromatase inhibitors (AIs) are given as adjuvant therapy in hormone receptor positive postmenopausal breast cancer and are known to decrease levels of estrone and estradiol in both serum and breast tissue. Our goal here was to examine the influence of AIs on mammographic density in women with early breast cancer. Methods: We conducted a case-control study of postmenopausal breast cancer patients initiating adjuvant AI therapy (anastrozole or exemestane) on protocols NCIC CTG MA27, NCCTG N063I and MC (Mayo Clinic) 0532. Eligibility included; an intact contralateral breast with no prior surgery; a screening mammogram within twelve months before AI initiation and at 9-15 months on therapy; no prior endocrine therapy and informed consent. Controls were sampled from the Mayo Mammography Health Study, a cohort of 19,924 receiving screening mammography at the Mayo Clinic, and matched to cases on age, prior PMH use, baseline body mass index (BMI) and interval between mammograms. Pre-treatment and on-study mammograms for cases (corresponding mammograms for controls) were digitized. Change in percent density was estimated on the craniocaudal view of the non-cancerous breast using two methods: a subjective assessment of change by an expert radiologist (within 5%; 5-10% increase, 10-25% increase, 25%+ increase, 5-10% decrease, 10-25% decrease and 25%+ decrease) and a quantitative assessment of absolute change using a computer-assisted thresholding program (Cumulus). Analyses compared magnitude of change in density by both the subjective and quantitative methods between cases and matched controls. Results: 574 pairs were eligible for analyses (MA27-505 cases; N063I-12 cases; MC0532-57 cases). Characteristics of the two groups are shown in the table below. Using either density estimation method, there was a greater decrease in density among women on AI therapy vs. matched controls. In 33% (95% CI: 29-37%) of pairs, there was at least a one greater category decrease for the case relative to her control by subjective estimation. In 14% (95% CI: 11-18%) of the pairs, there was at least a 5% greater decrease for the case relative to her control by quantitative estimation. Data will be available according to AI class (non-steroidal versus steroidal) in November. Conclusions: In the largest report to date to examine the influence of AI therapy on mammographic density, we provide evidence that AI is associated with decreases in density in a small subgroup of women. We are currently examining factors that influence these AI-associated decreases in density and whether these differences are unique to one class of AI. (Supported in part by NIH grants P50CA116201, U01GM61388, U10CA77202, U10CA25224) Citation Information: Cancer Res 2010;70(24 Suppl):Abstract nr P2-09-03.