Esserman, Laura

Abstract Background Immune checkpoint inhibitors can result in serious, long-lasting immune-related adverse events (irAEs). Early identification of symptoms predictive of irAEs could enhance monitoring and timely intervention. This study assessed whether symptoms within the first 8 weeks of treatment could predict subsequent development of immune-related adrenal insufficiency(AI) or hypothyroidism. Methods This retrospective cohort study analyzed prospectively collected data from the I-SPY2 trial, a phase 2 platform trial for high-risk stage II/III breast cancer across 30 U.S. sites. The cohort included 482 women treated with experimental immunotherapy agents concurrent with weekly paclitaxel neoadjuvant chemotherapy. The primary outcomes were grade ≥ 1 hypothyroidism or AI, adjudicated by an independent safety group, up to 1-year post-treatment. Symptoms and irAEs were assessed using the Common Terminology Criteria for Adverse Events. Symptom burden was quantified as area under the curve (AUC) based on symptom grade and duration. Predictive modeling was performed using logistic regression and ROC analysis; symptom enrichment between cases and controls was evaluated using Fisher’s exact tests. Results Among 482 participants, 107 (22.2%) developed irAEs, with hypothyroidism (n = 61, 12.7%) occurring more frequently than AI (n = 38, 7.9%) at medians of 99 and 105 days from treatment initiation, respectively. Symptom enrichment analysis identified early predictive symptoms. Fatigue (17.2% vs 6.8%, p = 0.011) and rash (20.7% vs 7.8%, p = 0.0037) were predictive of hypothyroidism, while diarrhea (45.9% vs 31%, p = 0.048), constipation (5.4% vs 0.2%, p = 0.018), and taste changes (5.4% vs 0.5%, p = 0.034) were associated with AI. A predictive model demonstrated moderate performance (AUC 0.65 for AI, p < 0.0001; AUC 0.61 for hypothyroidism, p = 0.012). Model accuracy in an external validation cohort was 72.8% for AI and 74.7% for hypothyroidism. Conclusions This study presents a predictive framework to identify patients at risk for adrenal insufficiency and hypothyroidism as irAEs, enabling personalized care and proactive intervention to improve treatment outcomes and safety.

Importance Molecular analyses of biospecimens collected from study participants are essential for identifying biomarkers that can tailor treatments to specific subsets of patients who are most likely to benefit. Sharing of data and biospecimens from clinical trials enables personalized, patient-centric use of cancer therapies and accelerates the development of new treatments. Objective To describe obstacles to sharing data and biospecimens and to propose strategies to enhance access and collaboration. Evidence Review This is a Special Communication authored by 53 academic investigators and patient representatives from the breast cancer community with extensive experience in conducting clinical and translational research. The article also evaluates the impact of biomarker research on specifying responsive subpopulations in the 29 registrational clinical trials that have led to approval of a new drug for treatment of breast cancer between 2017 and 2024. Findings Clinical trial participants are increasingly asked to provide tissue and/or body fluid biospecimens for biomarker research that is typically controlled by the sponsoring pharmaceutical company, but published biomarker studies are rare. Among 29 breast cancer registrational studies reported in the past 8 years, none resulted in biomarker research that restricted a drug’s approved indication. Herein, strategies to maximize the value of clinical data and biospecimens contributed by participants are proposed, thereby supporting the shared goals of the pharmaceutical industry and academia to improve patient care. These strategies include (1) establishing coleadership structures involving academia and patients in clinical trial design and conduct, (2) ensuring that informed consent forms state that data and biospecimens will be shared with academia for future research, (3) requiring the sharing of clinical data as a condition for regulatory approval, and (4) enabling access to biospecimens and translational research data for independent studies on biomarkers that may indicate drug efficacy and toxicity. Conclusions and Relevance Data and biospecimen sharing from registrational trials has been suboptimal. Improving clinical data, biospecimens, and biospecimens’ related data sharing requires concrete actions and a multidimensional stakeholder approach to accelerate the impact of clinical cancer research on the quality of patient care.

271 Background: Many studies have shown that fatigue is associated with poor outcomes such as quality of life and physical function in breast cancer survivors.Early treatment of symptoms that result in persistent fatigue may be a strategy to mitigate long-term fatigue.As a result, we studied symptoms and developed risk estimates for patients that are most likely to develop fatigue by pre-surgery and follow up among patients treated on the I-SPY 2 neoadjuvant trial. I-SPY 2 is a phase-II adaptive trial for high-risk early-stage breast cancer patients. Methods: In this retrospective cohort study of 306 women undergoing neoadjuvant chemotherapy +/- targeted therapy, patients complete electronic surveys of symptoms via 34 PRO-CTCAE items and QOL via 9 PROMIS instruments. Fatigue is collected as part of the PROMIS survey and scored using Health Measures. Study timepoints are screening (SC), weekly study treatment (ST), 12 weeks (12w), pre-surgery (PS) and at 1, 6, 12 and 24 months (FU1-FU4) following surgery. Correlation between symptoms and fatigue at equal timepoints was quantified using spearman correlations and BH-adjusted p-values. For symptom odds analysis, symptoms were binarized such that those that had an average grade of mild or above (severity), rarely or above (frequency) or a little bit or above (interference) between ST weeks 1-6 were considered present. Clinically significant (CS) increase in fatigue was defined as an increase >=0.5 standard deviation (SD) in mean from baseline. Logistic regression was used to determine odds ratios of factors associated with fatigue. Results: Baseline PROMIS fatigue mean score was 47.14 (SD 9.31) with a median of 47.45. A CS increase in fatigue from baseline was seen in 55% of patients at 12w, and 61% of patients pre-surgery. Of the patients that had a CS increase in fatigue at any point prior to surgery, 45.7% had persistent fatigue through FU1-4. Symptoms associated with Fatigue at 12w were shortness of breath, headache, and nausea (P< 0.03). At PS, fatigue was significantly correlated with GI symptoms, neuropathy, and shortness of breath (P< 0.03). Risk estimates for patients that filled out symptom surveys at early timepoints (weeks 1-6) and also had fatigue were evaluated. Patients with early GU (genitourinary) symptoms had higher risk of fatigue at IR (OR=9.33, p=0.02, 88.9% in cases vs 33.3% in controls). Patients with early GI symptoms at PS (OR=8.56, p=0.03, 90.1% vs 50%), or FU2-4 (OR= 5.88 and 7.36, p= 0.04 and 0.02, 80% vs 39.1% and 81.8% vs 36.3%) also had a higher risk of developing fatigue. Conclusions: A CS increase in fatigue during treatment and follow up is seen in the majority of patients and is associated with a large number of persistent symptoms.

11113 Background: Patient-reported outcomes (PROs) are valuable for understanding treatment tolerability and toxicity. I-SPY2 is a phase-II neoadjuvant clinical trial for breast cancer with electronic PROs and quality of life (QOL) assessments. We sought to understand how well the GP5 self-reported tolerability item reflected toxicity burden and QOL during treatment and follow-up in I-SPY 2 participants. Methods: Our study included 306 patients who completed the FACIT Item GP5 and answered PROMIS surveys at two or more timepoints. Analysis timepoints included screening (SC), inter-regimen (IR), pre-surgery (PS), 1-month post-surgery (FU1), and 6-months post-surgery (FU2). Symptoms were assessed weekly using PRO-CTCAE and CTCAEv4.0. QOL was evaluated using PROMIS. The GP5: “I am bothered by side effects of treatment,” was used to estimate tolerability, where worse tolerability is a higher score. Odds ratios, Spearman correlation, and regularized multi-variate regression were used to evaluate the associations between individual symptoms and adverse events with GP5 scores. Results: Demographic information was available for 243 of the 306 (70.2%) participants (median age=47,73.7% white). In the 306, tolerability scores worsened over study treatment, with a maximum mean score of 1.71 at Cycle 11. Tolerability improved after treatment cessation at FU1(1.09). Mean tolerability at week 6 was 1.53. The only clinical factor that was associated with early change in tolerability (week 6 to PS) was age >50 (mean score ranged from 1.55-1.70 vs 1.29-1.45 (age<50), p<0.05). PRO-CTCAE and GP5 correlative analysis suggested that higher grade abdominal pain, decreased appetite, dizziness and blurry vision significantly associated with higher GP5 scores at multiple timepoints (p<0.05). Symptoms not associated with tolerability were nail loss, acne, hair loss, and rash (p>0.05). Additionally, the total number of low-grade (mild to moderate) symptoms a patient experienced during study treatment was significantly correlated to poor tolerability (p<0.00001). Among patients with CTCAE data, 9% had adrenal insufficiency, and higher GP5 was found across all study timepoints in those with (compared to without) adrenal insufficiency (mean GP5 = [1.7-1.9] vs [1.3-1.5] respectively). Similarly, among patients with neuropathy (14%), higher GP5 scores were reported across all study timepoints (mean GP5 = [1.8-2.3] vs [1.3-1.5] respectively). Higher GP5 scores during treatment were correlated with higher fatigue scores at PS and FU1 (p<0.02 and p<0.001) and lower physical functioning at PS (p<0.01). Conclusions: TheGP5 measure is associated with self-reported symptoms, higher grade adverse events and quality of life scores. Furthermore, greater accumulation of lower grade symptoms can lead to poor tolerability. The single measure GP5 composite score reflects multiple components of treatment toxicity.

594 Background: Recent analyses in HER2+ and triple negative (TN) breast cancer (BC) treated with neoadjuvant systemic therapy (NAST) suggest that extent of disease at presentation (tumor [T] size and nodal [N] involvement) influences outcomes independently of response to NAST, even in the context of pathologic complete response (pCR). Given this, we evaluated the association of clinical T and N status prior to NAST with long-term outcomes in I-SPY2. Methods: We identified patients (pts) with TN and HER2+ BC who received NAST on the I-SPY2 trial with available clinical T size and N status prior to NAST, pathologic response after NAST, and follow up for recurrence and survival. We conducted univariable and multivariable (MV) Cox proportional models for Event Free (EFS), Distant Recurrence Free (DRFS), and Overall Survival (OS) in TN and HER2+ BC according to response after NAST, adjusting for residual cancer burden (RCB) category. Results: Of 1,170 pts with TN or HER2+ BC, 1,013 were eligible (TN: 623 and HER2+: 390), after 157 pts were excluded due to missing critical variables. Median follow up was 4.4 years (range 0.3-10.2). Median age was 48 for TN and 49 for HER2+ BC. Following NAST, pCR was achieved in 277 (45%) pts with TN and 197 (51%) pts with HER2+ BC. Among pts with pCR, separate MV analyses for TN and HER2+ BC showed no independent association between clinical T and N category with EFS, DRFS or OS (p values all > 0.05). However, among pts with residual disease, clinical T category (but not clinical N status) prior to NAST retained independent prognostic information in both TN and HER2+ BC in a MV model adjusting for age, RCB class, and hormone receptor status (for HER2+ BC). Compared to cT1-2, cT3-4 stage at diagnosis was associated with worse EFS, DRFS, and OS in TNBC, and with worse EFS, DRFS (but not OS) in HER2+ BC (Table). Additionally, hormone receptor status was independently associated with outcomes in HER2+ BC with residual disease (data not shown). The 5y EFS for cT1-2 vs cT3-4 according to RCB class in TNBC were: RCB0: 94% vs 95%, RCB1: 90% vs 65%, RCB2: 74% vs 44% and RCB3: 45% vs 20%. The 5y EFS for cT1-2 vs cT3-4 according to RCB class in HER2+ were: RCB0: 95% vs 88%, RCB1: 89% vs 84%, RCB2: 85% vs 64% and RCB3: 56% vs 50%. Conclusions: Tumor size (but not N involvement) prior to NAST was independently associated with outcomes in TN and HER2+ BC with residual disease, independently of RCB class. TN and HER2+ BC achieving pCR exhibited favorable outcomes regardless of extent of disease prior to NAST. These data are consistent with prior findings of the Neobioscore contributing only in the context of residual disease. Clinical trial information: NCT01042379 . [Table: see text]

Abstract Background: Disseminated tumor cells (DTCs) in bone marrow detected after treatment may represent occult residual disease. We enumerated DTCs after neoadjuvant chemotherapy (NACT) in patients (pts) diagnosed with high-risk early stage breast cancer and examined the relationship of these cells with response and survival. Methods: I-SPY SURMOUNT is a sub-study of the I-SPY 2 TRIAL (NCT01042379). Pts enrolled on I-SPY 2, who signed consent for this sub-study, had bone marrow aspirates (BMA) collected after NACT at the time of surgery. DTCs were isolated and enumerated from BMA using immunomagnetic enrichment/flow cytometry (IE/FC). DTCs were defined as EPCAM-positive and CD45-negative nucleated cells. Samples were considered positive using a predetermined threshold of &gt;4 DTCs per mL (Magbanua et al, unpublished data). Pathologic response was assessed using the residual cancer burden (RCB) method at local sites, and pts underwent standard adjuvant therapy if indicated and follow up for recurrence events and death. Relationship of DTCs with clinicopathologic variables was examined using Chi-squared test. Group means were compared using t tests. The log-rank test was used to compare survival curves. Results: A total of 73 patients were enrolled, 51 of whom had successful DTC assessment. The median DTC per mL was 4 (interquartile range 1.2-11.6). 24/51 (47%) were DTC-positive. Clinical characteristics by DTC status are shown in the table. DTC-positive pts were significantly younger (p=0.02) and had larger pretreatment tumors (longest diameter by magnetic resonance imaging) compared to DTC-negative pts (p=0.032). DTCs were not associated with receptor subtype. Thirty pts (41%) achieved a pathologic complete response (pCR). DTCs were not associated with pCR (p= 0.166); however, DTC-positive patients were significantly more likely to have residual cancer (RCB-II/III) after NACT compared to DTC-negative patients (OR 3.3, p=0.037). Median follow up of this cohort was 2.8 years (range: 0.9-4.8). Interim survival analysis showed that DTCs were not significantly correlated with EFS (p=0.6) or DRFS (p=0.41). Conclusions: Detection of DTCs at surgery after NACT is significantly more common in young patients, those with larger tumors, and those with residual disease at surgery. While these associations suggest higher risk for later recurrence, larger studies and longer follow up are necessary to determine if DTCs add prognostic value over pathologic evaluation alone for pts receiving NACT. Citation Format: Mark Jesus M Magbanua, Laura van 't Veer, Amy Clark, A. Jo Chien, Judy Boughey, Heather Han, Anne Wallace, Heather Beckwith, Minetta Liu, Christina Yau, E. Paul Wileyto, Lamorna Brown Swigart, Jane Perlmutter, Lauren Bayne, Shannon Deluca, Stephanie Yee, Erica Carpenter, Laura Esserman, John Park, Lewis Chodosh, Angela DeMichele. Outcomes associated with disseminated tumor cells at surgery after neoadjuvant chemotherapy in high-risk early stage breast cancer: The I-SPY SURMOUNT study [abstract]. In: Proceedings of the 2020 San Antonio Breast Cancer Virtual Symposium; 2020 Dec 8-11; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2021;81(4 Suppl):Abstract nr PS2-07.

Abstract Background: Achieving a pathologic complete response (pCR) has been shown on the patient level to predict excellent long-term event-free survival outcomes. Residual cancer burden (RCB) quantifies the extent of residual disease for patients who did not achieve pCR. We have previously observed in the I-SPY 2 TRIAL that while metastatic events outside the central nervous system (CNS) were dramatically reduced in the setting of pCR, the incidence of CNS metastasis remained similar across RCB classes, raising the possibility that these CNS events may be independent of response in the breast. In this study, we evaluate the type and sites of recurrences by RCB in a large pooled dataset, which allows for analysis within subtype, to validate these findings. Methods: 5161 patients pooled across 12 institutions/trials with available RCB and event-free survival (EFS) data were included in this analysis. EFS was calculated as the interval between treatment initiation, and locoregional recurrence, distant recurrence or death from any cause; patients without event are censored at time of last follow-up. The median follow-up is 4.6 years. We summarized the EFS event type, further sub-dividing the distant recurrence events (DR) by their site of relapse (CNS-only, CNS and other sites, Non-CNS). We used a competing risk (Fine-Gray) model to assess which of these site-specific relapses differ between RCB classes and estimated the cumulative incidence of CNS-only and non-CNS events at 5 years. Analyses were performed across the entire study population and within HR/HER2 defined subtypes. Results: Among the 5161 subjects, there were 1164 EFS events, including 92 (7.9%) local recurrences (without distant recurrence and/or death) and 1072 distant recurrence-free survival (DRFS) events. Among the DRFS events, 158 patients died without a distant recurrence. 914 experienced distant recurrences, including 90 (9.8%) with CNS-only, 145 (15.9%) with CNS and other sites, 664 (72.6%) with non-CNS distant recurrence; 15 (1.6%) patients had missing recurrence site information. Table 1 summarizes the cumulative incidence of CNS-only and non-CNS recurrence at 5 years and the proportion of CNS-only recurrences among DR events by RCB class overall and within each HR/HER2 subtypes. The incidence of CNS-only recurrences was low and similar across RCB classes. In contrast, the incidence of non-CNS recurrences increases with increasing RCB. As a result, CNS-only recurrences are proportionally higher within the RCB-0 and RCB-I than in the RCB-II and RCB-III groups, largely because of the low DR event rate and relative low frequency of non-CNS recurrence events within the RCB-0 and RCB-I classes. Overall, 27% of the recurrences in the setting of pCR (RCB-0) are due to CNS-only recurrences.Conclusions: Consistent with previous studies, our large pooled analysis confirmed that CNS-only recurrences are uncommon but appear similar across RCB groups, independent of response, suggesting that the CNS is a treatment sanctuary site. In contrast, non-CNS recurrence rates increase as RCB increases. These findings suggest that inclusion of CNS-only recurrences as an outcome event may impact the association between neoadjuvant therapy response and long-term outcomes in the context of current therapies. Novel therapies that cross the blood brain barrier will be needed to impact CNS recurrence rates. Table 1: Cumulative Incidence of CNS Only and non-CNS Distant Recurrences at 5 years and proportion of CNS-only events among DR eventsRCB Class0IIIIIIpOverall (5161)N16766622017806Cum. Inc. CNS Only2%2%2%1%0.627Cum. Inc. Non-CNS3%6%16%27%&lt;0.001# CNS-Only / # DR events (%)26/96 (27%)14/74 (19%)39/443 (9%)11/301 (4%)HR-HER2- (1774)N770212590202Cum. Inc. CNS Only2%3%2%4%0.298Cum. Inc. Non-CNS4%11%19%42%&lt;0.001# CNS-Only / # DR events (%)13/50 (26%)6/32 (19%)13/148 (9%)8/111 (7%)HR-HER2+ (572)N3766710029Cum. Inc. CNS Only1%5%5%0%0.022Cum. Inc. Non-CNS2%5%18%38%&lt;0.001# CNS-Only / # DR events (%)4/17 (24%)3/10 (30%)6/31 (19%)0/13 (0%)HR+HER2+ (858)N31317229182Cum. Inc. CNS Only1%1%2%0%0.37Cum. Inc. Non-CNS2%3%15%26%&lt;0.001# CNS-Only / # DR events (%)3/10 (30%)2/16 (12%)7/68 (10%)0/29 (0%)HR+HER2- (1957)N2172111036493Cum. Inc. CNS Only3%2%1%0.2%0.087Cum. Inc. Non-CNS5%4%13%20%&lt;0.001# CNS-Only / # DR events (%)6/19 (32%)3/16 (19%)13/196 (7%)3/148 (2%) Citation Format: Sonal Shad, Marieke van der Noordaa, Marie Osdoit, Diane de Croze, Anne-Sophie Hamy, Marick Lae, Fabien Reyal, Miguel Martin, María Del Monte-Millán, Sara López-Tarruella, I-SPY 2 TRIAL Consortium, Judy C Boughey, Matthew P Goetz, Tanya Hoskin, Rebekah Gould, Vicente Valero, Gabe Sonke, Tessa G Steenbruggen, Maartje van Seijen, Jelle Wesseling, John Bartlett, Stephen Edge, Mi-Ok Kim, Jean Abraham, Carlos Caldas, Helena Earl, Elena Provenzano, Stephen-John Sammut, David Cameron, Ashley Graham, Peter Hall, Lorna Mackintosh, Fan Fang, Andrew K Godwin, Kelsey Schwensen, Priyanka Sharma, Angela DeMichele, Janet Dunn, Louise Hiller, Larry Hayward, Jeremy Thomas, Kimberly Cole, Lajos Pusztai, Laura Van't Veer, Fraser Symmans, Laura Esserman, Christina Yau. Site of recurrence after neoadjuvant therapy: A multi-center pooled analysis [abstract]. In: Proceedings of the 2020 San Antonio Breast Cancer Virtual Symposium; 2020 Dec 8-11; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2021;81(4 Suppl):Abstract nr PD13-02.

Abstract Background: Pembrolizumab (Pembro), an anti-PD-1 immune checkpoint inhibitor, has been approved for the treatment of a variety of cancers. Pembro was recently evaluated in HER2- breast cancer patients in the neoadjuvant I-SPY 2 TRIAL and graduated in the triple negative (TN), HR+HER2-, and HER2- signatures. TN breast cancers tend to have high numbers of immune infiltrates, including T cells and tumor associated macrophages (TAMs). We utilized multiplex multispectral imaging to evaluate the presence and spatial proximity of various immune cell populations as specific predictors of response to Pembro. Methods: Pre-treatment biopsies were available from 54 patients. FFPE sections were immunostained using Opal fluorescence reagent kits on a fully automated Ventana Discovery platform and imaged with a Vectra automated imaging system. Image analyses were performed with a variety of software packages. Immune biomarkers included CD3, CD8, FoxP3, CD68, PD-1, PD-L1, and cytokeratins. Cell densities were determined per area of tissue. Spatial analyses were performed to quantitate proximity of cell types. We used logistic regression with the likelihood ratio test to evaluate associations between pathologic complete response (pCR) with immune cell counts, ratios, and spatial proximity measurements. This analysis was also performed adjusting for HR status, and within receptor subsets, sample size permitting. Our statistics are descriptive rather than inferential and do not adjust for multiple hypothesis testing or multiplicities of other biomarkers outside this study. Results: Densities of CD3+ T cells ranged from 193-5155 cells/mm2, with CD8+ cytotoxic T cells (Tc) ranging from 103 to 3774 cells/mm2 and FoxP3+ Tregs from 0-416 cells/mm2. Densities of PD-1+ T cells and PD-1+ Tc cells ranged from 0-1341 cells/mm2 and 0-698 cells/mm2, respectively. Macrophage cell densities ranged from 84-2644 cells/mm2. PD-L1+ tumor cell densities ranged from 0-2981 cells/mm2. High PD-L1 expression (&gt;=10% of tumor cells positive for PD-L1) was observed in 12% of the cases. Overall T cell density was positively associated with pathological complete response (pCR), as were Treg cell density and PD-1+ T cell density. The ratio of macrophages to Tc cells was negatively associated with response. Finally, the spatial distribution of CD3+ T cells in proximity to cancer cells correlated positively with pCR. Conclusion: Cell density, immune cell ratios and spatial resolution provide insight to response and resistance. We confirmed the observation that T cell density predicts pCR, but more specifically that Treg and PD1+ T cell densities increase the chance of pCR. An increase in macrophages relative to cytotoxic T cells predicts resistance. The closer the T cells are to the tumor, the better the response. These results demonstrate the utility of characterizing the immune microenvironment, in the context of therapy with immuno-oncology agents, not only to predict response, but to gain insight and learn now to potentiate agents and combinations. Citation Format: Michael J. Campbell, Christina Yau, Jennifer Bolen, Scott Vandenberg, Clifford Hoyt, Lamorna Brown-Swigart, Gillian Hirst, Rita Nanda, Minetta Liu, Smita Asare, Laura van'tVeer, Douglas Yee, Angie DeMichele, Don Berry, Laura Esserman, I SPY 2 TRIAL Investigators. Analysis of immune cell infiltrates as predictors of response to the checkpoint inhibitor pembrolizumab in the neoadjuvant I-SPY 2 TRIAL [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr CT003.

Abstract Background: Triple negative breast cancer (TNBC), generally considered aggressive with few options beyond standard chemotherapy, is currently experiencing a period of optimism, with multiple trials showing increased efficacy of platinum containing regimens (with and without PARP-inhibition), and more recently, immunotherapy. In I-SPY 2, veliparib/carboplatin (VC) and the immune checkpoint inhibitor pembrolizumab (Pembro) graduated in the TN subtype. As part of the I-SPY 2 biomarker program, we previously identified the signature PARPi7, a measure of DNA repair deficiency (DRD), as predicting response to VC. As well, immune signatures, including one representing dendritic cells, predicted response to Pembro. Here we investigate the overlap between these biomarkers to identify patient subgroups more likely to respond to immunotherapy or platinum-based therapy and estimate response rates to VC and Pembro in patients positive for one, both, or neither biomarker. Methods: 153 TNBC patients (Control: 85; VC: 39; Pembro: 29) were considered in this analysis. Continuous DNA repair deficiency and immune signatures assayed at the pre-treatment time point were evaluated as published (Wolf, 2017; Danaher, 2017). To identify optimal dichotomizing thresholds, 2-fold cross-validation was repeated 500 times. Bayesian logistic regression was used to estimate pCR rates by arm for patient subsets defined by biomarker combinations. Our study is exploratory with no claims for generalizability; and this analysis does not adjust for multiplicities. Results: Using the optimal dichotomizing cutpoints, 54% of TNBC patients are classified as DRD+, and 67% Immune+. As expected, DRD+ patients have a high estimated pCR rate to VC (75%); and Immune+ patients have a high estimated pCR rate to Pembro (82%). Combining these biomarkers, 40% of TNBC are positive for both biomarkers, 40% for only one biomarker (26% Immune+/DRD-, 14% Immune-/DRD+), and 20% for neither. The Immune+/DRD+ TN subset predicted sensitive to both VC and Pembro has a higher estimated pCR rate to both VC and Pembro relative to control (estimated pCR rates: VC: 75%, Pembro: 73% vs. control: 18%). In contrast, the Immune+/DRD- group, predicted to be sensitive to Pembro, has the highest pCR rate to Pembro (VC: 48%, Pembro: 83% vs. control 21%), whereas the Immune-/DRD+ group is most responsive to VC (VC: 63%, Pembro: 41% vs. control 29%). For the 20% of Immune-/DRD- TNBC patients, their estimated response rates are lower than the biomarker-positive patients (VC: 27%, Pembro: 43% vs. control: 19%). Conclusion: Molecular phenotypes capturing immune activation and DNA repair deficiency may predict response of TNBC and help prioritize selection of platinum or immunotherapy containing regimens. Citation Format: Denise M. Wolf, Christina Yau, Julia Wulfkuhle, Emanuel Petricoin, Lamorna Brown-Swigart, Gillian Hirst, Smita Asare, I-SPY 2 Consortium, Douglas Yee, Angela DeMichele, Hope Rugo, Olufunmilayo Olopade, Rita Nanda, Minetta Liu, Laura Esserman, Laura van t' Veer. Integration of DNA repair deficiency and immune biomarkers to predict which early-stage triple-negative breast cancer patients are likely to respond to platinum-containing regimens vs. immunotherapy: The neoadjuvant I-SPY 2 trial [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 2679.

506 Background: Pembro is an anti-PD-1 antibody with single agent activity in HER2– metastatic BC. I-SPY 2 is a multicenter, phase 2 platform trial which evaluates novel neoadjuvant therapies; the primary endpoint is pathological complete response (pCR, ypT0/Tis ypN0). We report current efficacy results, with final results at ASCO. Methods: Patients (pts) with invasive BC ≥2.5 cm by exam or ≥2 cm by imaging are assigned weekly paclitaxel x 12 (control) +/- an experimental agent, followed by doxorubicin/cyclophosphamide x 4. Combinations of hormone-receptor (HR), HER2, & MammaPrint (MP) status define the 8 signatures studied. MP low HR+ BC is excluded. Adaptive randomization is based on each arm’s Bayesian probability of superiority over control. Graduation by signature is based on an arm’s Bayesian predictive probability of a successful 1:1 randomized phase 3 trial with a pCR endpoint. We provide raw & Bayesian estimated pCR rates adjusted for covariates, time effects over the course of the trial, & serial MRI modeling for pts not yet assessed for pCR surgically. Results: 69 pts were randomized to pembro (HER2- subsets only) from Dec 2015 until it graduated in Nov 2016. 46 pts have undergone surgery (table); the other 23 have on-therapy MRI assessments. In 29 HR–/HER2– (TNBC) pts, pembro increased raw & estimated pCR rates by >50% & 40%, respectively; in 40 HR+/HER– pts, it did so by 13% and 21%. 5 pts had immune-related grade 3 adverse events (AEs); 1 hypophysitis & 4 adrenal insufficiency. 4 pts presented after completion of AC (149-179 d after starting pembro); 1 presented prior to AC (37 d after starting pembro). 7 pts had grade 1-2 thyroid abnormalities. Conclusion: Pembro added to standard therapy improved pCR rates in all HER2- BCs that meet I-SPY 2 eligibility, especially in TNBC. Immune-mediated AEs were observed; pt follow up is ongoing. Clinical trial information: NCT01042379. [Table: see text]