BackgroundPrognostic information is essential for decision-making in breast cancer management. In recent years, trials and clinical practice have emphasized genomic prognostication tools, despite clinicopathological methods being more affordable and accessible. PREDICT v3 is one such tool with promising results across cohorts. Advances in machine learning (ML), transfer learning, and ensemble methods provide opportunities to enhance these approaches, especially where missing data and model assumptions differ across diverse populations. ObjectiveThis study evaluates the potential to improve survival prognostication in breast cancer. More precisely, we compare de novo ML, transfer learning from the pretrained prognostication model PREDICT v3, and a stacked ensemble approach. MethodsData from the MA.27 trial (NCT00066573) were used for model training, with external validation on data from the Tamoxifen Exemestane Adjuvant Multinational trial (NCT00279448 and NCT00032136) and a US Surveillance, Epidemiology, and End Results cohort. Transfer learning was applied by re-estimating the parameters of the pretrained prognostic tool PREDICT v3. De novo ML included random survival forests and extreme gradient boosting, and the ensemble was implemented using weighted linear stacking of model predictions. Internal and external validation was assessed in terms of the integrated calibration index and discrimination. Shapley Additive Explanations values were used to explain model predictions and decision-curve analysis to facilitate the interpretation of performance differences. ResultsTransfer learning, de novo random survival forest, and the stacked ensemble improved calibration in MA.27 over the pretrained model (integrated calibration index reduced from 0.042 in PREDICT v3 to ≤0.007) while discrimination remained comparable (AUROC increased from 0.738 in PREDICT v3 to 0.744-0.799). In decision-curve analysis, these approaches demonstrated consistently positive net benefit across clinically relevant thresholds, while PREDICT v3 lost net benefit beyond 7.5% predicted risk. Invalid PREDICT v3 predictions were observed in 23.8% to 25.8% of MA.27 individuals due to missing information. In contrast, ML models and the stacked ensemble predicted survival despite missing data. Across all models, patient age, nodal status, pathological grading, and tumor size had the highest Shapley Additive Explanations values, indicating their importance for survival prognostication. External validation in the US Surveillance, Epidemiology, and End Results cohort confirmed the benefits of transfer learning, RSF, and ensemble in terms of calibration while maintaining discrimination at comparable levels. In contrast, generalizability was limited in the Tamoxifen Exemestane Adjuvant Multinational trial, a cohort with a substantially different distribution of clinicopathological characteristics. ConclusionsThis study demonstrates that transfer learning, de novo RSF, and a stacked ensemble can improve prognostication compared with the pretrained PREDICT v3, particularly in the presence of missing or uncertain inputs. Transportability may be limited in cohorts with different clinicopathological profiles, requiring local validation before clinical deployment. Ultimately, better survival estimation can provide more meaningful guidance in breast cancer care. Trial RegistrationClinicalTrials.gov NCT00066573; https://clinicaltrials.gov/study/NCT00066573, NCT00279448; https://clinicaltrials.gov/study/NCT00279448, NCT00032136; https://clinicaltrials.gov/study/NCT00032136

Statistically standardized estrogen receptor (ER) and progesterone receptor (PgR) differentiated prognosis. Here we examined statistically standardized human epidermal growth receptor 2 (HER2). CCTG MA.27 (NCT00066573) was an adjuvant phase III trial of exemestane versus anastrozole in postmenopausal women with ER + and/or PgR + tumors. We centrally quantitated machine-image immunohistochemical HER2, defined American Society of Clinical Oncology (ASCO)/College of American Pathologists (CAP) dual-probe FISH HER2/CEP17 categories, determined ultra-low HER2 (IHC 0 with (0,10

Background/Objectives: Insomnia is associated with immune function. This study evaluated the association between insomnia and febrile neutropenia in women treated with adjuvant chemotherapy for breast cancer. Methods: This secondary analysis used data from the Canadian Cancer Trial Group MA.21 trial, which compared three chemotherapy regimens (CEF, EC/T dose-dense, or AC/T) in 2104 women with high-risk locoregional breast cancer. A total of 1731 patients completed the EORTC QLQ-C30 questionnaire. We compared "insomnia patients" with patients considered "good sleepers" based on the sleep item of this questionnaire. The primary endpoint was the occurrence of febrile neutropenia. Secondary endpoints were the occurrences of leucopenia and infection. Chemotherapy dose reduction was added as a secondary outcome in an unplanned analysis. Results: Patients with insomnia (16.3%) had a significantly higher rate of febrile neutropenia than good sleepers (12.2%; p = 0.01). After controlling for various confounders, the contribution of insomnia in explaining febrile neutropenia remained statistically significant (OR 1.45, 95% CI 1.07-1.97, p = 0.02). Chemotherapy dose reductions were significantly more frequent in patients with insomnia (30.6%) than in good sleepers (21.8%; p < 0.0001). The relationship remained significant in the multivariate analysis (OR 1.67, 95% CI 1.30-2.15, p < 0.0001). Conclusions: In the MA21 trial, insomnia was associated with febrile neutropenia. Furthermore, chemotherapy dose reductions were more common in women with insomnia. These results suggest that the role of insomnia in potential cancer outcomes needs to be confirmed in other studies, given the possible implication of dose reductions on the prognosis of women receiving chemotherapy for breast cancer.

Prognostic information is essential for decision-making in breast cancer management. Recently trials have predominantly focused on genomic prognostication tools, even though clinicopathological prognostication is less costly and more widely accessible. Machine learning (ML), transfer learning and ensemble integration offer opportunities to build robust prognostication frameworks. We evaluate this potential to improve survival prognostication in breast cancer by comparing de-novo ML, transfer learning from a pre-trained prognostic tool and ensemble integration. Data from the MA.27 trial was used for model training, with external validation on the TEAM trial and a SEER cohort. Transfer learning was applied by fine-tuning the pre-trained prognostic tool PREDICT v3, de-novo ML included Random Survival Forests and Extreme Gradient Boosting, and ensemble integration was realized through a weighted sum of model predictions. Transfer learning, de-novo RSF, and ensemble integration improved calibration in MA.27 over the pre-trained model (ICI reduced from 0.042 in PREDICT v3 to <=0.007) while discrimination remained comparable (AUC increased from 0.738 in PREDICT v3 to 0.744-0.799). Invalid PREDICT v3 predictions were observed in 23.8-25.8

BACKGROUND:Given the intensive resources required to conduct economic analyses in clinical trials, a key need is identifying scalable measures of costs. Contact days-days with health care contact outside the home-may represent such a practical measure. METHODS:We conducted a secondary analysis of a trial that evaluated two pre-transplant chemotherapy regimens for lymphoma. We used trial resource use and patient-reported data to calculate contact days, direct costs, and indirect costs such as lost productivity. We assessed the association between the number of contact days and cost outcomes using linear regression models and Pearson correlation coefficients. RESULTS:Contact days were moderately correlated with direct costs (r = .47, $762/ contact day, P < .0001), and strongly correlated with direct costs in the DHAP arm (r = 0.60, $727/ contact day, P < .0001). Contact days were very weakly correlated with pooled indirect costs (r 0.19, $60/ contact day, P = .0003). Among the 3 indirect cost outcomes, the relationship was strongest with paid caregiving hours (r = 0.33, 1.8 hours/ contact day, P < .0001) and weakest for unpaid hours provided by informal care partners (r = .06, .7 hours/ contact day, P = .247). Results were robust when zeroing out costs of hospitalization in the arm receiving inpatient chemotherapy and when evaluating indirect costs among patients working full-time. CONCLUSIONS:Contact days have the potential as a surrogate measure of direct health system costs, which deserves further exploration. The weak correlation with indirect cost outcomes suggests that the extent of true patient and care partner burdens extends beyond just the number of contact days.Trial registration: ClinicalTrials.gov Identifier: NCT00078949.

Hyperprogressive disease (HPD) is characterized by an acceleration of tumor growth in a subset of patients due to receiving immune-oncology (IO) therapy. Patients, policy makers and treating physicians should be made aware of the potential HPD related to a specific IO-therapy. Multiple methods based on radiological criteria comparing the pre-treatment and post-treatment tumor growth have been proposed in the literature to identify HPD cases. In the absence of a consensus regarding which methods to use, we compared the different methods and proposed two simple combinations of the existing methods to estimate the potential incidence of HPD related to the received IO therapy. Data from 305 patients across multiple centres (Gustave Roussy, Vall d'Hebron Institute of Oncology, START Madrid-CIOCC and Institute of Cancer Research Marsden) were pooled. 163 patients had pre-baseline study visit disease assessment available and progressive disease (PD) as per RECIST version 1.1 at the first post-baseline assessment on target lesions exclusively. The presence of HPD according to the ‘tumor growth rate’ (TGR), ‘tumor growth kinetics’ (TGK) and ‘tumor growth difference’ (TGD) method was analyzed in this subsample of 163 patients. In addition, a conservative method requiring all three methods to declare HPD (“intersection method”) and a sensitive method requiring any of the three methods to declare HPD (“union method”) was also used. The three most common primary malignancies in the pooled data were lung cancer (117 patients, 38.36%), colorectal cancer (32 patients, 10.49%) and melanoma (30 patients, 9.84%). Median age was 59 (IQR=18), 176 patients (57.70%) were male and 129 patients were female (42.30%). The TGR method identified 53 patients (17.38%) with HPD, the TGK method 62 patients (20.33%) and the TGD method 41 patients (13.44%). The pairwise agreement in HPD cases identified across methods was estimated using Cohen’s Kappa. The Kappa statistics were 0.88, 0.71 and 0.70 for the pairwise concordance between TGR-TGK, TGR-TGD, and TGK-TGD, respectively. HPD was declared by all three methods in 37 patients (12.13%), i.e. the intersection method. The union approach identified HPD in 62 patients (20.33%) and was identical to the TGK method, making it the most sensitive method of the three in this analysis. HPD assessment methods comparing tumor growth acceleration before and after receiving IO-therapy provide insight in the dynamics of the target lesions. Our analysis suggests that the TGR, TGK and TGD methods are concordant. Rather than favouring one method over the other, we propose to combine the existing methods into a sensitive and restrictive method. The sensitive method can serve as an upper bound of HPD incidence, and the restrictive as a lower bound. These boundaries can inform patients, policy makers and treating physicians of the potential HPD related to a specific IO-therapy. Luc Boone, Roberto Ferrara, Giuseppe Lo Russo, Penelope Bradbury, Lesley Seymour, Stephane Champiat, Christophe Le Tourneau, Anna Minchom, Ruth Plummer, Larry Schwartz, Bingshu Chen, Elena Garralda Cabanas, Scott Laurie, Saskia Litière, Jan Bogaerts, Emiliano Calvo. Comparing and combining existing radiological criteria for hyperprogressive disease in patients receiving immune-oncology therapy: Building towards a sensitive and conservative method to assess the presence of hyperprogressive disease [abstract]. In: Proceedings of the AACR-NCI-EORTC International Conference on Molecular Targets and Cancer Therapeutics; 2025 Oct 22-26; Boston, MA. Philadelphia (PA): AACR; Mol Cancer Ther 2025;24(10 Suppl):Abstract nr C072.

Background: We proposed adjunctive statistical standardization of quantitated ER and PgR to improve inter-laboratory comparability of biomarker results and therapeutic management of breast cancer. Adjunctive statistical standardization of quantitated HER2 is used here; we also examined the effects on outcome of ultra-low HER2 and very low statistically standardized HER2. Methods: We utilized CCTG MA.27 (NCT00066573), an adjuvant phase III trial of exemestane versus anastrozole in postmenopausal women with ER+ and/or PgR+ tumors. IHC HER2 HSCORE and % positivity (%+) were centrally assessed by machine image quantitation, and statistically standardized to mean of 0, standard deviationn (SD) of 1 followig Box-Cox variance stabilization transformations of 1.) natural logarithm (ln with addition of 0.1 to 0 HSCOREs and 0 %+), 2. square root. Additionally, centrally assessed FISH HER2 and CEP17 values were used to define ASCO/CAP categorizatiion. Post hoc, the effects of ultra-low HER2 , IHC 0 with (0,10%] 1+ stain, were examined. The primary endpoint was distant disease-free survival (DDFS) at the longest trial follow-up of median 4.1 years. Survival was described with Kaplan-Meier plots and tested with the univariate Wilcoxon (Peto-Prentice) test statistic. We examined cut-points at standard deviations about mean of 0 (<-1; (-1,0]; (0,1]; >1). Cox multivariant regressions were adjusted for age, T and N stage, grade, lymphovascular invasion, treatment, baseline patient demographics, ER and PgR; 2-sided Wald tests had nominal significance if p<0.05. Results: Of 7576 women accrued to MA.27, 2900 women had ER, 2726 had PgR, and 2680 had HER2 results; 2325 had all three biomarkers for multivariant investigations. Twenty-five women received received herceptin with only one experiencng a DDFS event. ASCO/CAP categorization significantly differentiated univariate DDFS (p=0.01). Image analysis identified 57% of IHC 0 to have ultra-low HER2. Five-year DDFS for IHC 0 without stain was 92% [95% CI (90,95); N=864] which was similar to that for ultra-low HER2 of 96% [95% CI (94,97); N=1143]. Statistical standardization did not significantly differentiate univariate DDFS (p=0.08-0.27). DDFS for ln standardized values <-1.0 (HSCORE, or %+ <0.1) was similar to that with standardized values >1.0 (HSCORE >19, or %+ >14): for HSCORE <1.0, 5-year DDFS was 92% [95% CI (85.98); N=88] vs for >1.0, 92% [95% CI (89,95); N=577]; for %+, 5-year DDFS was 91% [95% CI (85,98); N=102] vs >1.0, 92% [95% CI (89,95); N=613].In multivariant assessments with ASCO/CAP guideline and statistically standardized data, both ER (p=0.65-0.94) and HER2 (p=0.20-0.97) were not significantly associated with the DDFS primary endpoint in models with PgR; while higher PgR had significantly better DDFS (p<.003) in models with ER and HER2. Conclusions: ASCO/CAP HER2 guidelines significantly differentiated univariate DDFS although not values of IHC 0 and ultra-low HER2, with <10% weak stain. Statistical standardization did not differentiate univariate DDFS. Image quantitation identified very small numbers of 1+/2+/3+ intensity stained nuclei. DDFS was similar for any intensity of low ln(HER2) stain (<1 SD below the mean) compared to any intensity of higher HER2 stain (>1 SD above the mean), although we offer caution in assessment of ultra-low, or very low, HER2 stain due to the dynamic range of the HER2 assay. Neither ASCO/CAP nor standardized HER2 had multivariant significance in these hormone receptor rich patient tumors. The adjunctive statistical standardization of ER, PgR, and HER2 performed here is similar to that mandated for clinical practice by the World Health Organization for BMD. Citation Format: Judith-Anne Chapman, Jane Bayani, Sandip SenGupta, John M.S. Bartlett, Tammy Piper, Mary Anne Quintayo, Shakeel Virk, Paul E. Goss, James N. Ingle, Matthew J. Ellis, George W. Sledge, G. Thomas Budd, Manuela Rabaglio, Rafat H. Ansari, Richard Tozer, David P. D'Souza, Haji Chalchal, Silvana Spadafora, Vered Stearns, Edith A. Perez, Karen A. Gelmon, Timothy J. Whelan, Catherine Elliott, Lois E. Shepherd, Bingshu E. Chen, Karen J. Taylor. Adjunctive statistical standardization of quantitated adjuvant HER2 and very low statistically standardized HER2 in CCTG MA.27 [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2024; 2024 Dec 10-13; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2025;31(12 Suppl):Abstract nr P1-07-06.

LBA1005 Background: Standard first line therapy in ER+/ HER2 negative endocrine sensitive metastatic breast cancer (MBC) is a CDK4/6 inhibitor plus AI. Upon progression the majority of patients will receive further endocrine based therapy. Alterations in PI3K/AKT pathway genes are a known mechanism of endocrine resistance – either de novo or acquired. The MA.40 trial evaluated the efficacy and safety of the AKT inhibitor ipatasertib versus placebo plus fulvestrant in the metastatic breast cancer (MBC) setting immediately post progression on 1 st line CDK4/6 inhibitor and AI. Methods: This phase III, randomized, double-blind trial enrolled pre/peri/postmenopausal women and men with ER+/HER-2 negative MBC. Patients were randomly assigned 1:1 to receive ipatasertib plus fulvestrant versus (vs) placebo plus fulvestrant. Stratification factors included: AKT pathway altered ( PIK3CA , AKT1 , and/or PTEN genomic alteration(s)) vs wild-type/unknown and endocrine resistance (primary vs secondary). Primary objective: To compare investigator assessed PFS (RECIST 1.1) between treatment arms in the ITT population. Pre-specified secondary analysis: PFS in the AKT pathway altered cohort using a hierarchical procedure. The FoundationOne Liquid cfDNA NGS assay was utilized to assess genomic alterations in the AKT pathway for stratification. Results: 250 participants (females 247/males 3) were enrolled from Canada, Australia and New Zealand between January 2021 and May 2024. Baseline characteristics between arms were balanced. 44.4% of the study population had AKT pathway alteration(s) per cfDNA assay. Median follow-up 15.2 months(mo); proportion remaining on protocol treatment at time of analysis 21.0% ipatasertib vs 11.3% placebo arm. Median PFS ITT ipatasertib vs placebo arms were: 5.32 mo (95% CI: 3.58 to 5.62 mo) vs 1.94 mo (95% CI: 1.84 to 3.22 ) [HR 0.61, 95% CI: 0.46 to 0.81; p= 0.0007] and in the AKT pathway altered cohort: 5.45 mo (95% CI: 3.55 to 11.01 ) vs 1.91 mo (95% CI: 1.77 to 3.48) [HR = 0.47, 95% CI: 0.31 to 0.72; p= 0.0005]. Grade 3 or higher adverse events (AE) ipatasertib vs placebo arms (%):37.1 vs 27.4. Grade 3 or higher non haematological treatment related AE > 1% ipatasertib vs placebo arms: diarrhea (16% vs 0%); fatigue (3% vs 0%); vomiting (2% vs 0), rash (2% vs 0%). Treatment discontinuation due to AEs ipatasertib vs placebo arms: 6.5% vs 0.8%. Conclusions: Ipatasertib plus fulvestrant significantly prolongs PFS compared to placebo/fulvestrant in patients with hormone receptor–positive MBC post progression on 1 st line CDK 4/6 inhibitor and AI. Follow-up and additional analyses continue. Supported by Hoffmann-La Roche Ltd, CCS grant #707213. Clinical trial information: NCT04650581 .

The role of germline genetics in adjuvant aromatase inhibitor (AI) treatment efficacy in ER-positive breast cancer is poorly understood. We employed a two-stage candidate gene approach to examine associations between survival endpoints and common germline variants in 753 endocrine resistance-related genes. For a discovery cohort, we screened the Breast Cancer Association Consortium database (n ≥ 90,000 cases) and retrieved 2789 AI-treated patients. Cox model-based analysis revealed 125 variants associated with overall, distant relapse-free, and relapse-free survival (p-value ≤ 1E-04). In validation analysis using five independent cohorts (n = 8857), none of the six selected candidates representing major linkage blocks at CELA2B/CASP9, NR1I2/GSK3B, LRP1B, and MIR143HG (CARMN) were validated. We discuss potential reasons for the failed validation and replication of published findings, including study/treatment heterogeneity and other limitations inherent to genomic treatment outcome studies. For the future, we envision prospective longitudinal studies with sufficiently long follow-up and endpoints that reflect the dynamic nature of endocrine resistance.

BackgroundInsufficient patient accrual is a major challenge in clinical trials and can result in underpowered studies, as well as exposing study participants to toxicity and additional costs, with limited scientific benefit. Real-world data can provide external controls, but insufficient accrual affects all arms of a study, not just controls. Studies that used generative models to simulate more patients were limited in the accrual scenarios considered, replicability criteria, number of generative models, and number of clinical trials evaluated. ObjectiveThis study aimed to perform a comprehensive evaluation on the extent generative models can be used to simulate additional patients to compensate for insufficient accrual in clinical trials. MethodsWe performed a retrospective analysis using 10 datasets from 9 fully accrued, completed, and published cancer trials. For each trial, we removed the latest recruited patients (from 10% to 50%), trained a generative model on the remaining patients, and simulated additional patients to replace the removed ones using the generative model to augment the available data. We then replicated the published analysis on this augmented dataset to determine if the findings remained the same. Four different generative models were evaluated: sequential synthesis with decision trees, Bayesian network, generative adversarial network, and a variational autoencoder. These generative models were compared to sampling with replacement (ie, bootstrap) as a simple alternative. Replication of the published analyses used 4 metrics: decision agreement, estimate agreement, standardized difference, and CI overlap. ResultsSequential synthesis performed well on the 4 replication metrics for the removal of up to 40% of the last recruited patients (decision agreement: 88% to 100% across datasets, estimate agreement: 100%, cannot reject standardized difference null hypothesis: 100%, and CI overlap: 0.8-0.92). Sampling with replacement was the next most effective approach, with decision agreement varying from 78% to 89% across all datasets. There was no evidence of a monotonic relationship in the estimated effect size with recruitment order across these studies. This suggests that patients recruited earlier in a trial were not systematically different than those recruited later, at least partially explaining why generative models trained on early data can effectively simulate patients recruited later in a trial. The fidelity of the generated data relative to the training data on the Hellinger distance was high in all cases. ConclusionsFor an oncology study with insufficient accrual with as few as 60% of target recruitment, sequential synthesis can enable the simulation of the full dataset had the study continued accruing patients and can be an alternative to drawing conclusions from an underpowered study. These results provide evidence demonstrating the potential for generative models to rescue poorly accruing clinical trials, but additional studies are needed to confirm these findings and to generalize them for other diseases.

534 Background: An ongoing constant risk of recurrence out to 20 years is well established in hormone receptor positive breast cancer (BC) less so in other BC subtypes. This study aims to describe patters of early (≤ 5 years of BC diagnosis) and late recurrence (> 5 years of BC diagnosis) across immunohistochemically defined BC subtypes - luminal (ER/PgR+HER2-), triple negative (TN: ER/PgR/HER2-) and HER2+ (any ER/PgR) in CCTGMA.32 (NCT01101438) which investigated metformin vs placebo in patients enrolled 2010-2013. Methods: 3649 patients with high-risk non-metastatic BC were enrolled and followed for first locoregional and distant recurrence, new primary cancers and death. Annual rates of these events were calculated in each BC subtype and averaged for early (years 0-5) and late (after 5 years) post randomization. Results: In luminal (n = 2104), TN (n = 925), HER2+ (n = 620) BC the median follow-ups were 96.2 (range 0.2 to 120.7), 94.5 (0.03 to 120.5), 95.2 months (0.03 to 119.8), respectively. Patterns of events varied across subtypes and early vs late. In luminal BC, the early vs late annual invasive cancer event rates (ICERs) was 3.04 vs. 2.31 % (late rate 0.76 of early rate). The annual early vs late rates of distant recurrence (DR) were 2.33 vs 1.72% (late rate 0.74 of early rate). Bone was the most common site of DR both early and late. In the TN BC, the early vs late annual ICERs were 4.6 and 1.21% (late rate 0.35 of early rate). Annual early vs late DR rates were 3.09 vs. 0.20 % (late rate 0.28 of early rate). Visceral metastases (lung, liver, CNS) were most common early. In HER2+, early vs late annual ICERs were 2.93 vs 1.47% (late rate 0.50 of early rate). Annual early vs late DR rates were 2.25 vs 0.71% (late rate 0.32 of early rate). Bone and visceral metastases were common early. CNS was rare after 5 years in all BC subtypes. Second primary cancers (new BC and non-primary BC) were frequent across BC subtypes, with no fall-off over time; they were responsible for the majority of late events in TN and HER2+ BC. Conclusions: In luminal BC, risk of late ICER remains high (annual rate about three-quarters of early rate), while risk of late events was lower in TN and HER2+BC (late rates one quarter to one-third of early rates). Risk of second primary cancers did not decrease over time, and second primaries were the most frequent late events in TN and HER2+BC. Clinical trial information: NCT01101438 . Luminal TN HER2+ Annual event rate (%) Annual event rate (%) Annual event rate (%) Year 0-5 Year 5+ Year 0-5 Year 5+ Year 0-5 Year 5+ Any Invasive Cancer Event 3.04 2.31 4.60 1.21 2.93 1.47 Locoregional Event 0.50 0.29 1.15 0.26 0.64 0.15 Distant Recurrence* 2.08 1.29 3.09 0.20 2.03 0.57 Sites of First Metastasis: Bone 1.40 0.88 1.13 0.10 0.65 0.42 Lung 0.59 0.56 1.73 0.20 0.83 0.07 Liver 0.71 0.56 0.73 0.00 0.50 0.21 CNS 0.18 0.06 0.65 0.00 0.61 0.00 Second Primary Cancer** 0.66 0.92 0.96 0.90 0.60 0.74 *Including distant recurrence after a local regional events. **Non-breast cancer and new breast cancer events.

Power electronic devices are widely used in renewable power systems, and the characteristics of partial discharge (PD) signals under repetitive impulse excitation are vastly different from those observed under power frequency excitation, causing many pattern recognition methods to fail. This paper proposes a improved identification method based on representation learning that can directly process PD sequences received by ultra-high frequency (UHF) sensors. First, a representation learning architecture based on the Transformer Encoder module is proposed and the model is pretrained in an unsupervised manner to extract features from the raw input discharge sequences. Then, the extracted features are processed using the ridge regression classifier to achieve end-to-end PD source identification. Finally, the effectiveness of the proposed model is validated through an experimental study, achieving an improved accuracy of 98.6%, which is better than classical PD identification deep learning models under repetitive impulse voltage. Furthermore, the model exhibits strong robustness against noise interference and sampling rates. The effectiveness of the model was also tested on an inverter-fed motor stator device. To further validate its practical applicability, it is essential to collect diverse PD data from various electrical equipment.

PURPOSE ASCO/College of American Pathologists guidelines recommend reporting estrogen receptor (ER) and progesterone receptor (PgR) as positive with (1%-100%) staining. Statistically standardized quantitated positivity could indicate differential associations of positivity with breast cancer outcomes. METHODS MA.27 (ClinicalTrials.gov identifier: NCT00066573 ) was a phase III adjuvant trial of exemestane versus anastrozole in postmenopausal women with early-stage breast cancer. Immunochemistry ER and PgR HSCORE and % positivity (%+) were centrally assessed by machine image quantitation and statistically standardized to mean 0 and standard deviation (SD) 1 after Box-Cox variance stabilization transformations of square for ER; for PgR, (1) natural logarithm (0.1 added to 0 HSCOREs and 0%+) and (2) square root. Our primary end point was MA.27 distant disease-free survival (DDFS) at a median 4.1-year follow-up, and secondary end point was event-free survival (EFS). Univariate survival with cut points at SDs about a mean of 0 (≤–1; (–1, 0]; (0, 1]; >1) was described with Kaplan-Meier plots and examined with Wilcoxon (Peto-Prentice) test statistic. Adjusted Cox multivariable regressions had two-sided Wald tests and nominal significance P < .05. RESULTS Of 7,576 women accrued, 3,048 women's tumors had machine-quantitated image analysis results: 2,900 (95%) for ER, 2,726 (89%) for PgR, and 2,582 (85% of 3,048) with both ER and PgR. Higher statistically standardized ER and PgR HSCORE and %+ were associated with better univariate DDFS and EFS ( P < .001). In multivariable assessments, ER HSCORE and %+ were not significantly associated ( P = .52-.88) with DDFS in models with PgR, whereas higher PgR HSCORE and %+ were significantly associated with better DDFS ( P = .001) in models with ER. CONCLUSION Adjunctive statistical standardization differentiated quantitated levels of ER and PgR. Patients with higher ER- and PgR-standardized units had superior DDFS compared with those with HSCOREs and %+ ≤–1.

The Canadian Cancer Trials Group (CCTG) LY.17 is an ongoing multi-arm randomized phase II trial evaluating novel salvage therapies compared with R-GDP (rituximab, gemcitabine, dexamethasone and cisplatin) in autologous stem cell transplantation (ASCT)-eligible patients with relapsed/refractory diffuse large B-cell lymphoma (RR-DLBCL). This component of the LY.17 trial evaluated a dose-intensive chemotherapy approach using a single cycle of inpatient R-DICEP (rituximab, dose-intensive cyclophosphamide, etoposide and cisplatin) to achieve both lymphoma response and stem cell mobilization, shortening time to ASCT. This report is the result of the protocol-specified second interim analysis of the 67 patients who were randomized to either 1 cycle of R-DICEP or to 3 cycles of R-GDP. The overall response rate (ORR) was 65.6% for R-DICEP and 48.6% for R-GDP. The ASCT rate was 71.9% versus 54.3%, and 1-year progression-free survival rate was 42% versus 32%, respectively, for R-DICEP versus R-GDP. Although the improvement in ORR for R-DICEP versus R-GDP exceeded the pre-specified 10% threshold to proceed to full accrual of 64 patients/arm, higher rates of grade 3-5 toxicities, and the need for hospitalization led to the decision to stop this arm of the study. CCTG LY.17 will continue to evaluate different salvage regimens that incorporate novel agents.

The Cox regression model or accelerated failure time regression models are often used for describing the relationship between survival outcomes and potential explanatory variables. These models assume the studied covariates are connected to the survival time or its distribution or their transformations through a function of a linear regression form. In this article, we propose nonparametric, nonlinear algorithms (deepAFT methods) based on deep artificial neural networks to model survival outcome data in the broad distribution family of accelerated failure time models. The proposed methods predict survival outcomes directly and tackle the problem of censoring via an imputation algorithm as well as re-weighting and transformation techniques based on the inverse probabilities of censoring. Through extensive simulation studies, we confirm that the proposed deepAFT methods achieve accurate predictions. They outperform the existing regression models in prediction accuracy, while being flexible and robust in modeling covariate effects of various nonlinear forms. Their prediction performance is comparable to other established deep learning methods such as deepSurv and random survival forest methods. Even though the direct output is the expected survival time, the proposed AFT methods also provide predictions for distributional functions such as the cumulative hazard and survival functions without additional learning efforts. For situations where the popular Cox regression model may not be appropriate, the deepAFT methods provide useful and effective alternatives, as shown in simulations, and demonstrated in applications to a lymphoma clinical trial study.

Abstract Background: Adjuvant breast cancer therapy is informed by whether a tumour is positive or negative for the biomarkers ER, PgR, and HER2, often without regard to level of positivity. Quantitation has been proposed to improve therapeutic management. Adjunctive statistical standardization has been proposed to improve inter-laboratory comparability of biomarkers results. Methods: This primary report utilized adjunctive statistical standardization of machine-quantitated image analysis biomarker assessments. CCTG MA.27 (NCT00066573) is an adjuvant phase III trial of exemestane versus anastrozole in postmenopausal women with ER+ and/or PgR+ tumours. IHC ER, PgR, and HER2 were centrally assessed, with FISH (HER2;HER2/CEP17) determinations for equivocal IHC HER2. HSCOREs were statistically standardized to a mean of 0, standard deviation of 1 following Box-Cox variance stabilization transformations of square for ER and natural logarithm for PgR (0.1 was added to 0 HSCOREs). The primary endpoint was STEEP distant disease-free survival (DDFS) at the longest trial follow-up of median 4.1 years. Survival was described with Kaplan-Meier plots. The univariate Wilcoxon (Peto-Prentice) test statistic was used with usual designation of negative/positive (0; >0), and standardized cut-points at standard deviations about mean of 0(<-1; (-1,0]; (0,1]; >1). Cox multivariate regressions adjusted for age, T and N stage, grade, lymphovascular invasion, treatment, and baseline patient demographics, utilized likelihood ratio tests. Nominal significance was p=0.05. Results: Of the 7576 women accrued, 3048 had machine-quantitated image analysis results: 2900 (95%) for ER; 2726 (89%) for PgR. Only 8 women were ASCO/CAP ER- (HSCORE 0); PgR HSCORE was 0 for 533. Statistically standardized units differentiated DDFS ER levels (p< 0.001) and PgR levels (p< 0.001). In adjusted multivariate analyses, higher ER HSCORE was associated with better DDFS (p=0.05) with weak evidence of an association (p=0.11) for standardized HSCORE, and no significant association (respectively, p=0.28, p=0.54) in models with PgR. Higher PgR was associated with better DDFS (p=0.001) in all multivariate assessments, including those with ER. Conclusions: DDFS was superior for patients with higher ER and PgR standardized units compared with those with HSCOREs <-1. Adjunctive statistical standardization, similar to that mandated for clinical practice by the World Health Organization for BMD, should improve inter-laboratory comparability of biomarker results for similar patient populations. Biomarker N DDFS DDFS 5-year (%) 95% CI ER total 2900 ER <-1 506 86 (82, 91) ER (-1, 0] 934 94 (92, 96) ER ( 0, 1] 919 94 (92, 96) ER >1 541 96 (93, 98) PgR total 2726 PgR <-1 734 89 (86, 92) PgR (-1, 0] 439 92 (89, 95) PgR ( 0, 1] 967 95 (93, 96) PgR >1.0 586 98 (97,100) Citation Format: Judy-Anne Chapman, Jane Bayani, Sandip SenGupta, John MS Bartlett, Tammy Piper, Mary Anne Quintayo, Shakeel Virk, Paul Goss, James Ingle, Matthew Ellis, George Sledge Jr, George Budd, Manuela Rabaglio, Rafat Ansari, Richard Tozer, David D'Souza, Haji Chalchal, Silvana Spadafora, Vered Stearns, Edith A. Perez, Karen Gelmon, Timothy Whelan, Catherine Elliott, Lois Shepherd, Bingshu Chen, Karen Taylor. Adjunctive statistical standardization of quantitated machine image analysis of Estrogen and Progesterone Receptors: CCTG MA.27 trial [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PO1-27-10.

The prevalence of frailty in clinical trials of lymphoma is unknown. We conducted a secondary analysis of the phase III LY.12 trial in which patients with relapsed aggressive non-Hodgkin lymphoma were randomized to different salvage regimens before autologous stem cell transplant. The primary objective was to construct a lymphoma clinical trials-specific frailty index (LyFI) using previously described methods. The secondary objective was to describe the association of frailty withover all and event-free survival (OS, EFS). The LyFI was constructed using 619 patients, and11% (N = 70) were classified as frail. Frailty was associated with EFS (HR 1.94, 95%CI 1.53-2.46) and OS (HR 2.01, 95%CI 1.57-2.58) in univariable analysis, but was only significant as a continuous (not binary) variable in multivariable analysis controlling for prognostic score, suggesting limitations of a FI in this trial population. Future work could validate the FI using clinical assessments and/or apply it to an older trial population.

In the LY.17 randomized phase II clinical trial, adults with relapsed and refractory diffuse large B-cell lymphoma treated with ibrutinib-R-GDP (IR-GDP) for up to three cycles had more documented bacterial and fungal infections, without improvement in overall response, compared with R-GDP. CR, complete response; DLBCL, diffuse large B-cell lymphoma; PD, progressive disease; PR, partial response; R/R, relapsed/refractory; SD, stable disease.

567 Background: We proposed adjunctive statistical standardization of quantitated ER and PgR to improve inter-laboratory comparability of biomarker results and therapeutic management of breast cancer. Methods: We utilized CCTG MA.27 (NCT00066573), an adjuvant phase III trial of exemestane versus anastrozole in postmenopausal women with ER+ and/or PgR+ tumours. IHC ER and PgR HSCORE and % positivity (%+) were centrally assessed by machine image quantitation, and each statistically standardized to mean of 0, standard deviation of 1 following Box-Cox variance stabilization transformations of square for ER; for PgR, 1.) natural logarithm (0.1 added to 0 HSCOREs and 0 %+), 2.) square root. The primary endpoint was STEEP distant disease-free survival (DDFS) at the longest trial follow-up of median 4.1 years; a secondary endpoint was event-free survival (EFS). Survival was described with Kaplan-Meier plots and tested with the univariate Wilcoxon (Peto-Prentice) test statistic. We examined cut-points at standard deviations about mean of 0 (<-1; (-1,0]; (0,1]; >1) and explored single cut-points. Cox multivariate regressions were adjusted for age, T and N stage, grade, lymphovascular invasion, treatment, and baseline patient demographics; 2-sided Wald tests had nominal significance if p<0.05. Results: Of the 7576 women accrued, 3048 women had machine-quantitated image analysis results: 2900 (95%) for ER; 2726 (89%) for PgR. Statistically standardized HSCORE and %+ units differentiated both univariate DDFS and EFS; DDFS was significantly different by ER levels (p<0.001) and PgR levels (p<0.001). In multivariable assessments, ER HSCORE and %+ were not significantly associated (p=0.52-0.88) with the DDFS primary endpoint in models with PgR, while higher PgR HSCORE and %+ had significantly better DDFS (p=.001, in all instances) in models with ER. Conclusions: DDFS was superior for patients with higher ER and PgR standardized units compared with those with HSCOREs and %+ <-1. The adjunctive statistical standardization of ER and PgR performed here is similar to that mandated for clinical practice by the World Health Organization for BMD. [Table: see text]