Search

Abstract Purpose Real world evidence is crucial to understanding the diffusion of new oncologic therapies, monitoring cancer outcomes, and detecting unexpected toxicities. In practice, real world evidence is challenging to collect rapidly and comprehensively, often requiring expensive and time‐consuming manual case‐finding and annotation of clinical text. In this Review, we summarise recent developments in the use of artificial intelligence to collect and analyze real world evidence in oncology. Methods We performed a narrative review of the major current trends and recent literature in artificial intelligence applications in oncology. Results Artificial intelligence (AI) approaches are increasingly used to efficiently phenotype patients and tumors at large scale. These tools also may provide novel biological insights and improve risk prediction through multimodal integration of radiographic, pathological, and genomic datasets. Custom language processing pipelines and large language models hold great promise for clinical prediction and phenotyping. Conclusions Despite rapid advances, continued progress in computation, generalizability, interpretability, and reliability as well as prospective validation are needed to integrate AI approaches into routine clinical care and real‐time monitoring of novel therapies.

Abstract Endocrine therapy (ET) is an effective first-line therapy for women with estrogen receptor-positive (ER + ) breast cancers. While both ionizing radiation (RT) and ET are used for the treatment of women with ER+ breast cancer, the most effective sequencing of therapy and the effect of ET on tumor radiosensitization remains unclear. Here we sought to understand the effects of inhibiting estrogen receptor (ER) signaling in combination with RT in multiple preclinical ER+ breast cancer models. Clonogenic survival assays were performed using variable pre- and post-treatment conditions to assess radiosensitization with estradiol, estrogen deprivation, tamoxifen, fulvestrant, or AZD9496 in ER+ breast cancer cell lines. Estrogen stimulation was radioprotective (radiation enhancement ratios [rER]: 0.51–0.82). Conversely, when given one hour prior to RT, ER inhibition or estrogen depletion radiosensitized ER+ MCF-7 and T47D cells (tamoxifen rER: 1.50–1.60, fulvestrant rER: 1.76–2.81, AZD9496 rER: 1.33–1.48, estrogen depletion rER: 1.47–1.51). Combination treatment resulted in an increase in double-strand DNA (dsDNA) breaks as a result of inhibition of non-homologous end joining-mediated dsDNA break repair with no effect on homologous recombination. Treatment with tamoxifen or fulvestrant in combination with RT also increased the number of senescent cells but did not affect apoptosis or cell cycle distribution. Using an MCF-7 xenograft model, concurrent treatment with tamoxifen and RT was synergistic and resulted in a significant decrease in tumor volume and a delay in time to tumor doubling without significant toxicity. These findings provide preclinical evidence that concurrent treatment with ET and RT may be an effective radiosensitization strategy.

Standard radiation therapy (RT) does not reliably provide locoregional control for women with multinode-positive breast cancer and triple-negative breast cancer (TNBC). We hypothesized that CDK4/6 inhibition (CDK4/6i) would increase the radiosensitivity not only of estrogen receptor–positive (ER+) cells, but also of TNBC that expresses retinoblastoma (RB) protein. We found that CDK4/6i radiosensitized RB WT TNBC (n = 4, radiation enhancement ratio [rER]: 1.49–2.22) but failed to radiosensitize RB-null TNBC (n = 3, rER: 0.84–1.00). RB expression predicted response to CDK4/6i + RT (R2 = 0.84), and radiosensitization was lost in ER+/TNBC cells (rER: 0.88–1.13) after RB1 knockdown in isogenic and nonisogenic models. CDK4/6i suppressed homologous recombination (HR) in RB WT cells but not in RB-null cells or isogenic models of RB1 loss; HR competency was rescued with RB reexpression. Radiosensitization was independent of nonhomologous end joining and the known effects of CDK4/6i on cell cycle arrest. Mechanistically, RB and RAD51 interact in vitro to promote HR repair. CDK4/6i produced RB-dependent radiosensitization in TNBC xenografts but not in isogenic RB1-null xenografts. Our data provide the preclinical rationale for a clinical trial expanding the use of CDK4/6i + RT to difficult-to-control RB-intact breast cancers (including TNBC) and nominate RB status as a predictive biomarker of therapeutic efficacy.

BackgroundAromatase inhibitors (AI) reduce recurrence and death in patients with early-stage hormone receptor-positive (HR +) breast cancer. Treatment-related toxicities, including AI-induced musculoskeletal symptoms (AIMSS), are common and may lead to early AI discontinuation. The objective of this study was to replicate previously reported associations for candidate germline genetic polymorphisms with AIMSS.MethodsWomen with stage 0-III HR + breast cancer initiating adjuvant AI were enrolled in a prospective clinic-based observational cohort. AIMSS were assessed by patient-reported outcomes (PRO) including the PROMIS pain interference and physical function measures plus the FACT-ES joint pain question at baseline and after 3 and 6 months. For the primary analysis, AIMSS were defined as ≥ 4-point increase in the pain interference T-score from baseline. Secondary AIMSS endpoints were defined as ≥ 4-point decrease in the physical function T-score from baseline and as ≥ 1-point increase on the FACT-ES joint pain question from baseline. The primary hypothesis was that TCL1A rs11849538 would be associated with AIMSS. Twelve other germline variants in CYP19A1, VDR, PIRC66, OPG, ESR1, CYP27B1, CYP17A1, and RANKL were also analyzed assuming a dominant genetic effect and prespecified direction of effect on AIMSS using univariate logistic regression with an unadjusted α = 0.05. Significant univariate associations in the expected direction were adjusted for age, race, body mass index (BMI), prior taxane, and the type of AI using multivariable logistic regression.ResultsA total of 143 participants with PRO and genetic data were included in this analysis, most of whom were treated with anastrozole (78%) or letrozole (20%). On primary analysis, participants carrying TCL1A rs11849538 were not more likely to develop AIMSS (odds ratio = 1.29, 95% confidence interval: 0.55–3.07, p = 0.56). In the statistically uncorrected secondary analysis, OPG rs2073618 was associated with AIMSS defined by worsening on the FACT-ES joint pain question (OR = 3.33, p = 0.004), and this association maintained significance after covariate adjustment (OR = 3.98, p = 0.003).ConclusionCarriers of OPG rs2073618 may be at increased risk of AIMSS. If confirmed in other cohorts, OPG genotyping can be used to identify individuals with HR + early breast cancer in whom alternate endocrine therapy or interventions to enhance symptom detection and implement strategies to reduce musculoskeletal symptoms may be needed.

Background Immune checkpoint inhibition is effective in several cancers. Expression of programmed death-ligand 1 (PD-L1) on circulating tumor or immune effector cells could provide insights into selection of patients for immune checkpoint inhibition. Methods Whole blood was collected at serial timepoints from metastatic breast cancer patients and healthy donors for circulating tumor cell (CTC) and platelet PD-L1 analysis with a phycoerythrin-labeled anti-human PD-L1 monoclonal antibody (Biolegend clone 29E.2A3) using the CellSearch® assay. CTC PD-L1 was considered positive if detected on at least 1% of the cells; platelet PD-L1 was considered positive if ≥100 platelets per CellSearch frame expressed PD-L1. Results A total of 207 specimens from 124 metastatic breast cancer patients were collected. 52/124 (42%) samples at timepoint-1 (at or close to time of progressive disease) had ≥5 CTC/7.5ml whole blood. Of those, 21 (40%) had positive CTC PD-L1. In addition, platelet PD-L1 expression was observed in 35/124 (28%) at timepoint-1. Platelet PD-L1 was not detected in more than 70 specimens from 12 healthy donors. Platelet PD-L1 was associated with ≥5 CTC/7.5ml whole blood (p = 0.0002), less likely in patients with higher red blood cell counts (OR = 0.72, pConclusion PD-L1 expression was found in metastatic breast cancer patients on both CTC and platelets in an independent fashion. Inter-patient platelet PD-L1 expression was highly heterogeneous suggesting that it is a biological event associated with cancer in some but not all patients. Taken together, our data suggest that CTC and platelet PD-L1 expression could play a role in predicting which patients should receive immune checkpoint inhibition and as a pharmacodynamics biomarker during treatment.

Increased rates of locoregional recurrence (LR) have been observed in triple negative breast cancer (TNBC) despite multimodality therapy, including radiation (RT). Recent data suggest inhibiting the androgen receptor (AR) may be an effective radiosensitizing strategy, and AR is expressed in 15–35% of TNBC tumors. The aim of this study was to determine whether seviteronel (INO-464), a novel CYP17 lyase inhibitor and AR antagonist, is able to radiosensitize AR-positive (AR+) TNBC models. In cell viability assays, seviteronel and enzalutamide exhibited limited effect as a single agent (IC50 > 10 μM). Using clonogenic survival assays, however, AR knockdown and AR inhibition with seviteronel were effective at radiosensitizing cells with radiation enhancement ratios of 1.20–1.89 in models of TNBC with high AR expression. AR-negative (AR−) models, regardless of their estrogen receptor expression, were not radiosensitized with seviteronel treatment at concentrations up to 5 μM. Radiosensitization of AR+ TNBC models was at least partially dependent on impaired dsDNA break repair with significant delays in repair at 6, 16, and 24 h as measured by immunofluorescent staining of γH2AX foci. Similar effects were observed in an in vivo AR+ TNBC xenograft model where there was a significant reduction in tumor volume and a delay to tumor doubling and tripling times in mice treated with seviteronel and radiation. Following combination treatment with seviteronel and radiation, increased binding of AR occurred at DNA damage response genes, including genes involved both in homologous recombination and non-homologous end joining. This trend was not observed with combination treatment of enzalutamide and RT, suggesting that seviteronel may have a different mechanism of radiosensitization compared to other AR inhibitors. Enzalutamide and seviteronel treatment also had different effects on AR and AR target genes as measured by immunoblot and qPCR. These results implicate AR as a mediator of radioresistance in AR+ TNBC models and support the use of seviteronel as a radiosensitizing agent in AR+ TNBC.

LncRNAs have been associated with cancer. Here, the authors carry out a systematic review of lncRNAs in breast cancer and show that DSCAM-AS1is highly expressed in oestrogen receptor positive tumours and enhances cancer through an interaction with hnRNPL; and is also associated with tamoxifen resistance.

Fulvestrant is a dose dependent selective estrogen receptor (ER) down‐regulator (SERD) used in ER‐positive metastatic breast cancer (MBC). Nearly all patients develop resistance. We performed molecular analysis of circulating tumor cells (CTC) to gain insight into fulvestrant resistance. Preclinical studies were performed with cultured breast cancer cells spiked into human blood and analyzed on the CellSearch® system. Clinical data are limited to a subset of patients with ER‐positive MBC from a previously reported pilot trial whose disease was progressing on fulvestrant (N = 7) or aromatase inhibitors (AIs) (N = 10). CTCs were enumerated and phenotyped for ER and B‐cell lymphoma (BCL2) using the CellSearch® CXC kit. In preclinical modeling, tamoxifen and AIs resulted in stabilized ER expression, whereas fulvestrant eliminated it. Five of seven patients progressing on fulvestrant had ≥5CTC/7.5 ml WB. Two of these five, treated with 500 mg/month fulvestrant, had no detectable CTC‐expression of ER and BCL2 (an ER regulated gene). Three patients had heterogeneous CTC‐ER and BCL2 expression indicating incomplete degradation of the ER target by fulvestrant. Two of these patients received 250 mg/month whereas the third patient received 500 mg/month fulvestrant. Her cancer harbored a mutation (Y537S) in the estrogen receptor alpha gene (ESR1). All seven ER positive patients progressing on AIs had heterogeneous CTC‐ER expression. These results suggest heterogeneous mechanisms of resistance to fulvestrant, including insufficient dosage, ESR1 mutation, or conversion to dependence on non‐ER pathways. CTC enumeration, phenotyping, and genotyping might identify patients who would benefit from fulvestrant dose escalation versus switching to alternative therapies.

Estrogen signaling plays an essential role in breast cancer progression, and estrogen receptor (ER) status has long been a marker of hormone responsiveness. However, ER status alone has been an incomplete predictor of endocrine therapy, as some ER+ tumors, nevertheless, have poor prognosis. Here we sought to use expression profiling of ER+ breast cancer cells to screen for a robust estrogen-regulated gene signature that may serve as a better indicator of cancer outcome. We identified 532 estrogen-induced genes and further developed a 73-gene signature that best separated a training set of 286 primary breast carcinomas into prognostic subtypes by stepwise cross-validation. Notably, this signature predicts clinical outcome in over 10 patient cohorts as well as their respective ER+ subcohorts. Further, this signature separates patients who have received endocrine therapy into two prognostic subgroups, suggesting its specificity as a measure of estrogen signaling, and thus hormone sensitivity. The 73-gene signature also provides additional predictive value for patient survival, independent of other clinical parameters, and outperforms other previously reported molecular outcome signatures. Taken together, these data demonstrate the power of using cell culture systems to screen for robust gene signatures of clinical relevance.

BackgroundImmune checkpoint inhibition is effective in several cancers. Expression of programmed death-ligand 1 (PD-L1) on circulating tumor or immune effector cells could provide insights into selection of patients for immune checkpoint inhibition.MethodsWhole blood was collected at serial timepoints from metastatic breast cancer patients and healthy donors for circulating tumor cell (CTC) and platelet PD-L1 analysis with a phycoerythrin-labeled anti-human PD-L1 monoclonal antibody (Biolegend clone 29E.2A3) using the CellSearch® assay. CTC PD-L1 was considered positive if detected on at least 1% of the cells; platelet PD-L1 was considered positive if ≥100 platelets per CellSearch frame expressed PD-L1.ResultsA total of 207 specimens from 124 metastatic breast cancer patients were collected. 52/124 (42%) samples at timepoint-1 (at or close to time of progressive disease) had ≥5 CTC/7.5ml whole blood. Of those, 21 (40%) had positive CTC PD-L1. In addition, platelet PD-L1 expression was observed in 35/124 (28%) at timepoint-1. Platelet PD-L1 was not detected in more than 70 specimens from 12 healthy donors. Platelet PD-L1 was associated with ≥5 CTC/7.5ml whole blood (p = 0.0002), less likely in patients with higher red blood cell counts (OR = 0.72, pConclusionPD-L1 expression was found in metastatic breast cancer patients on both CTC and platelets in an independent fashion. Inter-patient platelet PD-L1 expression was highly heterogeneous suggesting that it is a biological event associated with cancer in some but not all patients. Taken together, our data suggest that CTC and platelet PD-L1 expression could play a role in predicting which patients should receive immune checkpoint inhibition and as a pharmacodynamics biomarker during treatment.

Whole genome amplification (WGA) technologies can be used to amplify genomic DNA when only small amounts of DNA are available. The Multiple Displacement Amplification Phi polymerase based amplification has been shown to accurately amplify DNA for a variety of genotyping assays; however, it has not been tested for genotyping many of the clinically relevant genes important for pharmacogenetic studies, such as the cytochrome P450 genes, that are typically difficult to genotype due to multiple pseudogenes, copy number variations, and high similarity to other related genes. We evaluated whole genome amplified samples for Taqman™ genotyping of SNPs in a variety of pharmacogenetic genes. In 24 DNA samples from the Coriell human diversity panel, the call rates and concordance between amplified (~200-fold amplification) and unamplified samples was 100% for two SNPs in CYP2D6 and one in ESR1. In samples from a breast cancer clinical trial (Trial 1), we compared the genotyping results in samples before and after WGA for four SNPs in CYP2D6, one SNP in CYP2C19, one SNP in CYP19A1, two SNPs in ESR1, and two SNPs in ESR2. The concordance rates were all >97%. Finally, we compared the allele frequencies of 143 SNPs determined in Trial 1 (whole genome amplified DNA) to the allele frequencies determined in unamplified DNA samples from a separate trial (Trial 2) that enrolled a similar population. The call rates and allele frequencies between the two trials were 98% and 99.7%, respectively. We conclude that the whole genome amplified DNA is suitable for Taqman™ genotyping for a wide variety of pharmacogenetically relevant SNPs.