Functional annotation of the 2q35 breast cancer risk locus implicates a structural variant in influencing activity of a long-range enhancer element.

A combination of genetic and functional approaches has identified three independent breast cancer risk loci at 2q35. A recent fine-scale mapping analysis to refine these associations resulted in 1 (signal 1), 5 (signal 2), and 42 (signal 3) credible causal variants at these loci. We used publicly available in silico DNase I and ChIP-seq data with in vitro reporter gene and CRISPR assays to annotate signals 2 and 3. We identified putative regulatory elements that enhanced cell-type-specific transcription from the IGFBP5 promoter at both signals (30- to 40-fold increased expression by the putative regulatory element at signal 2, 2- to 3-fold by the putative regulatory element at signal 3). We further identified one of the five credible causal variants at signal 2, a 1.4 kb deletion (esv3594306), as the likely causal variant; the deletion allele of this variant was associated with an average additional increase in IGFBP5 expression of 1.3-fold (MCF-7) and 2.2-fold (T-47D). We propose a model in which the deletion allele of esv3594306 juxtaposes two transcription factor binding regions (annotated by estrogen receptor alpha ChIP-seq peaks) to generate a single extended regulatory element. This regulatory element increases cell-type-specific expression of the tumor suppressor gene IGFBP5 and, thereby, reduces risk of estrogen receptor-positive breast cancer (odds ratio = 0.77, 95% CI 0.74-0.81, p = 3.1 × 10-31).

A likelihood ratio approach for utilizing case-control data in the clinical classification of rare sequence variants: application to BRCA1 and BRCA2.

A large number of variants identified through clinical genetic testing in disease susceptibility genes, are of uncertain significance (VUS). Following the recommendations of the American College of Medical Genetics and Genomics (ACMG) and Association for Molecular Pathology (AMP), the frequency in case-control datasets (PS4 criterion), can inform their interpretation. We present a novel case-control likelihood ratio-based method that incorporates gene-specific age-related penetrance. We demonstrate the utility of this method in the analysis of simulated and real datasets. In the analyses of simulated data, the likelihood ratio method was more powerful compared to other methods. Likelihood ratios were calculated for a case-control dataset of BRCA1 and BRCA2 variants from the Breast Cancer Association Consortium (BCAC), and compared with logistic regression results. A larger number of variants reached evidence in favor of pathogenicity, and a substantial number of variants had evidence against pathogenicity - findings that would not have been reached using other case-control analysis methods. Our novel method provides greater power to classify rare variants compared to classical case-control methods. As an initiative from the ENIGMA Analytical Working Group, we provide user-friendly scripts and pre-formatted excel calculators for implementation of the method for rare variants in BRCA1, BRCA2 and other high-risk genes with known penetrance.

Breast Cancer Polygenic Risk Score and Contralateral Breast Cancer Risk.

Previous research has shown that polygenic risk scores (PRSs) can be used to stratify women according to their risk of developing primary invasive breast cancer. This study aimed to evaluate the association between a recently validated PRS of 313 germline variants (PRS313) and contralateral breast cancer (CBC) risk. We included 56,068 women of European ancestry diagnosed with first invasive breast cancer from 1990 onward with follow-up from the Breast Cancer Association Consortium. Metachronous CBC risk (N = 1,027) according to the distribution of PRS313 was quantified using Cox regression analyses. We assessed PRS313 interaction with age at first diagnosis, family history, morphology, ER status, PR status, and HER2 status, and (neo)adjuvant therapy. In studies of Asian women, with limited follow-up, CBC risk associated with PRS313 was assessed using logistic regression for 340 women with CBC compared with 12,133 women with unilateral breast cancer. Higher PRS313 was associated with increased CBC risk: hazard ratio per standard deviation (SD) = 1.25 (95%CI = 1.18-1.33) for Europeans, and an OR per SD = 1.15 (95%CI = 1.02-1.29) for Asians. The absolute lifetime risks of CBC, accounting for death as competing risk, were 12.4% for European women at the 10th percentile and 20.5% at the 90th percentile of PRS313. We found no evidence of confounding by or interaction with individual characteristics, characteristics of the primary tumor, or treatment. The C-index for the PRS313 alone was 0.563 (95%CI = 0.547-0.586). In conclusion, PRS313 is an independent factor associated with CBC risk and can be incorporated into CBC risk prediction models to help improve stratification and optimize surveillance and treatment strategies.

Common variants in breast cancer risk loci predispose to distinct tumor subtypes.

BACKGROUND: Genome-wide association studies (GWAS) have identified multiple common breast cancer susceptibility variants. Many of these variants have differential associations by estrogen receptor (ER) status, but how these variants relate with other tumor features and intrinsic molecular subtypes is unclear. METHODS: Among 106,571 invasive breast cancer cases and 95,762 controls of European ancestry with data on 173 breast cancer variants identified in previous GWAS, we used novel two-stage polytomous logistic regression models to evaluate variants in relation to multiple tumor features (ER, progesterone receptor (PR), human epidermal growth factor receptor 2 (HER2) and grade) adjusting for each other, and to intrinsic-like subtypes. RESULTS: Eighty-five of 173 variants were associated with at least one tumor feature (false discovery rate < 5%), most commonly ER and grade, followed by PR and HER2. Models for intrinsic-like subtypes found nearly all of these variants (83 of 85) associated at p < 0.05 with risk for at least one luminal-like subtype, and approximately half (41 of 85) of the variants were associated with risk of at least one non-luminal subtype, including 32 variants associated with triple-negative (TN) disease. Ten variants were associated with risk of all subtypes in different magnitude. Five variants were associated with risk of luminal A-like and TN subtypes in opposite directions. CONCLUSION: This report demonstrates a high level of complexity in the etiology heterogeneity of breast cancer susceptibility variants and can inform investigations of subtype-specific risk prediction.

Physical activity, sedentary time and breast cancer risk: a Mendelian randomisation study.

OBJECTIVES: Physical inactivity and sedentary behaviour are associated with higher breast cancer risk in observational studies, but ascribing causality is difficult. Mendelian randomisation (MR) assesses causality by simulating randomised trial groups using genotype. We assessed whether lifelong physical activity or sedentary time, assessed using genotype, may be causally associated with breast cancer risk overall, pre/post-menopause, and by case-groups defined by tumour characteristics. METHODS: We performed two-sample inverse-variance-weighted MR using individual-level Breast Cancer Association Consortium case-control data from 130 957 European-ancestry women (69 838 invasive cases), and published UK Biobank data (n=91 105-377 234). Genetic instruments were single nucleotide polymorphisms (SNPs) associated in UK Biobank with wrist-worn accelerometer-measured overall physical activity (nsnps=5) or sedentary time (nsnps=6), or accelerometer-measured (nsnps=1) or self-reported (nsnps=5) vigorous physical activity. RESULTS: Greater genetically-predicted overall activity was associated with lower breast cancer overall risk (OR=0.59; 95% confidence interval (CI) 0.42 to 0.83 per-standard deviation (SD;~8 milligravities acceleration)) and for most case-groups. Genetically-predicted vigorous activity was associated with lower risk of pre/perimenopausal breast cancer (OR=0.62; 95% CI 0.45 to 0.87,≥3 vs. 0 self-reported days/week), with consistent estimates for most case-groups. Greater genetically-predicted sedentary time was associated with higher hormone-receptor-negative tumour risk (OR=1.77; 95% CI 1.07 to 2.92 per-SD (~7% time spent sedentary)), with elevated estimates for most case-groups. Results were robust to sensitivity analyses examining pleiotropy (including weighted-median-MR, MR-Egger). CONCLUSION: Our study provides strong evidence that greater overall physical activity, greater vigorous activity, and lower sedentary time are likely to reduce breast cancer risk. More widespread adoption of active lifestyles may reduce the burden from the most common cancer in women.

Association of germline genetic variants with breast cancer-specific survival in patient subgroups defined by clinic-pathological variables related to tumor biology and type of systemic treatment.

BACKGROUND: Given the high heterogeneity among breast tumors, associations between common germline genetic variants and survival that may exist within specific subgroups could go undetected in an unstratified set of breast cancer patients. METHODS: We performed genome-wide association analyses within 15 subgroups of breast cancer patients based on prognostic factors, including hormone receptors, tumor grade, age, and type of systemic treatment. Analyses were based on 91,686 female patients of European ancestry from the Breast Cancer Association Consortium, including 7531 breast cancer-specific deaths over a median follow-up of 8.1 years. Cox regression was used to assess associations of common germline variants with 15-year and 5-year breast cancer-specific survival. We assessed the probability of these associations being true positives via the Bayesian false discovery probability (BFDP < 0.15). RESULTS: Evidence of associations with breast cancer-specific survival was observed in three patient subgroups, with variant rs5934618 in patients with grade 3 tumors (15-year-hazard ratio (HR) [95% confidence interval (CI)] 1.32 [1.20, 1.45], P = 1.4E-08, BFDP = 0.01, per G allele); variant rs4679741 in patients with ER-positive tumors treated with endocrine therapy (15-year-HR [95% CI] 1.18 [1.11, 1.26], P = 1.6E-07, BFDP = 0.09, per G allele); variants rs1106333 (15-year-HR [95% CI] 1.68 [1.39,2.03], P = 5.6E-08, BFDP = 0.12, per A allele) and rs78754389 (5-year-HR [95% CI] 1.79 [1.46,2.20], P = 1.7E-08, BFDP = 0.07, per A allele), in patients with ER-negative tumors treated with chemotherapy. CONCLUSIONS: We found evidence of four loci associated with breast cancer-specific survival within three patient subgroups. There was limited evidence for the existence of associations in other patient subgroups. However, the power for many subgroups is limited due to the low number of events. Even so, our results suggest that the impact of common germline genetic variants on breast cancer-specific survival might be limited.

Mendelian randomisation study of smoking exposure in relation to breast cancer risk.

BACKGROUND: Despite a modest association between tobacco smoking and breast cancer risk reported by recent epidemiological studies, it is still equivocal whether smoking is causally related to breast cancer risk. METHODS: We applied Mendelian randomisation (MR) to evaluate a potential causal effect of cigarette smoking on breast cancer risk. Both individual-level data as well as summary statistics for 164 single-nucleotide polymorphisms (SNPs) reported in genome-wide association studies of lifetime smoking index (LSI) or cigarette per day (CPD) were used to obtain MR effect estimates. Data from 108,420 invasive breast cancer cases and 87,681 controls were used for the LSI analysis and for the CPD analysis conducted among ever-smokers from 26,147 cancer cases and 26,072 controls. Sensitivity analyses were conducted to address pleiotropy. RESULTS: Genetically predicted LSI was associated with increased breast cancer risk (OR 1.18 per SD, 95% CI: 1.07-1.30, P = 0.11 × 10-2), but there was no evidence of association for genetically predicted CPD (OR 1.02, 95% CI: 0.78-1.19, P = 0.85). The sensitivity analyses yielded similar results and showed no strong evidence of pleiotropic effect. CONCLUSION: Our MR study provides supportive evidence for a potential causal association with breast cancer risk for lifetime smoking exposure but not cigarettes per day among smokers.

MiR-21 Is Required for the Epithelial-Mesenchymal Transition in MDA-MB-231 Breast Cancer Cells.

Breast cancer (BCa) is one of the leading health problems among women. Although significant achievements have led to advanced therapeutic success with targeted therapy options, more efforts are required for different subtypes of tumors and according to genomic, transcriptomic, and proteomic alterations. This study underlines the role of microRNA-21 (miR-21) in metastatic MDA-MB-231 breast cancer cells. Following the knockout of miR-21 from MDA-MB-231 cells, which have the highest miR-21 expression levels compared to MCF-7 and SK-BR-3 BCa cells, a decrease in epithelial-mesenchymal transition (EMT) via downregulation of mesenchymal markers was observed. Wnt-11 was a critical target for miR-21, and the Wnt-11 related signaling axis was altered in the stable miR-21 knockout cells. miR-21 expression was associated with a significant increase in mesenchymal markers in MDA-MB-231 BCa cells. Furthermore, the release of extracellular vesicles (EVs) was significantly reduced in the miR-21 KO cells, alongside a significant reduction in relative miR-21 export in EV cargo, compared with control cells. We conclude that miR-21 is a leading factor involved in mesenchymal transition in MDA-MB-231 BCa. Future therapeutic strategies could focus on its role in the treatment of metastatic breast cancer.

Genetic predisposition to in situ and invasive lobular carcinoma of the breast.

Invasive lobular breast cancer (ILC) accounts for 10-15% of all invasive breast carcinomas. It is generally ER positive (ER+) and often associated with lobular carcinoma in situ (LCIS). Genome-wide association studies have identified more than 70 common polymorphisms that predispose to breast cancer, but these studies included predominantly ductal (IDC) carcinomas. To identify novel common polymorphisms that predispose to ILC and LCIS, we pooled data from 6,023 cases (5,622 ILC, 401 pure LCIS) and 34,271 controls from 36 studies genotyped using the iCOGS chip. Six novel SNPs most strongly associated with ILC/LCIS in the pooled analysis were genotyped in a further 516 lobular cases (482 ILC, 36 LCIS) and 1,467 controls. These analyses identified a lobular-specific SNP at 7q34 (rs11977670, OR (95%CI) for ILC = 1.13 (1.09-1.18), P = 6.0 × 10(-10); P-het for ILC vs IDC ER+ tumors = 1.8 × 10(-4)). Of the 75 known breast cancer polymorphisms that were genotyped, 56 were associated with ILC and 15 with LCIS at P<0.05. Two SNPs showed significantly stronger associations for ILC than LCIS (rs2981579/10q26/FGFR2, P-het = 0.04 and rs889312/5q11/MAP3K1, P-het = 0.03); and two showed stronger associations for LCIS than ILC (rs6678914/1q32/LGR6, P-het = 0.001 and rs1752911/6q14, P-het = 0.04). In addition, seven of the 75 known loci showed significant differences between ER+ tumors with IDC and ILC histology, three of these showing stronger associations for ILC (rs11249433/1p11, rs2981579/10q26/FGFR2 and rs10995190/10q21/ZNF365) and four associated only with IDC (5p12/rs10941679; rs2588809/14q24/RAD51L1, rs6472903/8q21 and rs1550623/2q31/CDCA7). In conclusion, we have identified one novel lobular breast cancer specific predisposition polymorphism at 7q34, and shown for the first time that common breast cancer polymorphisms predispose to LCIS. We have shown that many of the ER+ breast cancer predisposition loci also predispose to ILC, although there is some heterogeneity between ER+ lobular and ER+ IDC tumors. These data provide evidence for overlapping, but distinct etiological pathways within ER+ breast cancer between morphological subtypes.

Cadherin-5: a biomarker for metastatic breast cancer with optimum efficacy in oestrogen receptor-positive breast cancers with vascular invasion.

BACKGROUND: A glycoproteomic study has previously shown cadherin-5 (CDH5) to be a serological marker of metastatic breast cancer when both protein levels and glycosylation status were assessed. In this study we aimed to further validate the utility of CDH5 as a biomarker for breast cancer progression. METHODS: A nested case-control study of serum samples from breast cancer patients, of which n=52 had developed a distant metastatic recurrence within 5 years post-diagnosis and n=60 had remained recurrence-free. ELISAs were used to quantify patient serum CDH5 levels and assess glycosylation by Helix pomatia agglutinin (HPA) binding. Clinicopathological, treatment and lifestyle factors associated with metastasis and elevated biomarker levels were identified. RESULTS: Elevated CDH5 levels (P=0.028) and ratios of CDH5:HPA binding (P=0.007) distinguished patients with metastatic disease from those that remained metastasis-free. Multivariate analysis showed that the association between CDH5:HPA ratio and the formation of distant metastases was driven by patients with oestrogen receptor (ER+) positive cancer with vascular invasion (VI+). CONCLUSIONS: CDH5 levels and the CDH5 glycosylation represent biomarker tests that distinguish patients with metastatic breast cancer from those that remain metastasis-free. The test reached optimal sensitivity and specificity in ER-positive cancers with vascular invasion.

Association of the CHEK2 c.1100delC variant, radiotherapy, and systemic treatment with contralateral breast cancer risk and breast cancer-specific survival.

Breast cancer (BC) patients with a germline CHEK2 c.1100delC variant have an increased risk of contralateral BC (CBC) and worse BC-specific survival (BCSS) compared to non-carriers. We aimed to assess the associations of CHEK2 c.1100delC, radiotherapy, and systemic treatment with CBC risk and BCSS. Analyses were based on 82,701 women diagnosed with invasive BC including 963 CHEK2 c.1100delC carriers; median follow-up was 9.1 years. Differential associations of treatment by CHEK2 c.1100delC status were tested by including interaction terms in a multivariable Cox regression model. A multi-state model was used for further insight into the relation between CHEK2 c.1100delC status, treatment, CBC risk and death. There was no evidence for differential associations of therapy with CBC risk by CHEK2 c.1100delC status The strongest association with reduced CBC risk was observed for the combination of chemotherapy and endocrine therapy [HR(95%CI): 0.66 (0.55-0.78)]. No association was observed with radiotherapy. Results from the multi-state model showed shorter BCSS for CHEK2 c.1100delC carriers versus non-carriers also after accounting for CBC occurrence [HR(95%CI) :1.30 (1.09-1.56)]. In conclusion, systemic therapy was associated with reduced CBC risk irrespective of CHEK2 c.1100delC status. Moreover, CHEK2 c.1100delC carriers had shorter BCSS, which appears not to be fully explained by their CBC risk. (Main MS: 3201 words).

Association of the CHEK2 c.1100delC variant, radiotherapy, and systemic treatment with contralateral breast cancer risk and breast cancer-specific survival.

BACKGROUND: Breast cancer (BC) patients with a germline CHEK2 c.1100delC variant have an increased risk of contralateral BC (CBC) and worse BC-specific survival (BCSS) compared to non-carriers. AIM: To assessed the associations of CHEK2 c.1100delC, radiotherapy, and systemic treatment with CBC risk and BCSS. METHODS: Analyses were based on 82,701 women diagnosed with a first primary invasive BC including 963 CHEK2 c.1100delC carriers; median follow-up was 9.1 years. Differential associations with treatment by CHEK2 c.1100delC status were tested by including interaction terms in a multivariable Cox regression model. A multi-state model was used for further insight into the relation between CHEK2 c.1100delC status, treatment, CBC risk and death. RESULTS: There was no evidence for differential associations of therapy with CBC risk by CHEK2 c.1100delC status. The strongest association with reduced CBC risk was observed for the combination of chemotherapy and endocrine therapy [HR (95% CI): 0.66 (0.55-0.78)]. No association was observed with radiotherapy. Results from the multi-state model showed shorter BCSS for CHEK2 c.1100delC carriers versus non-carriers also after accounting for CBC occurrence [HR (95% CI): 1.30 (1.09-1.56)]. CONCLUSION: Systemic therapy was associated with reduced CBC risk irrespective of CHEK2 c.1100delC status. Moreover, CHEK2 c.1100delC carriers had shorter BCSS, which appears not to be fully explained by their CBC risk.

Aggregation tests identify new gene associations with breast cancer in populations with diverse ancestry.

BACKGROUND: Low-frequency variants play an important role in breast cancer (BC) susceptibility. Gene-based methods can increase power by combining multiple variants in the same gene and help identify target genes. METHODS: We evaluated the potential of gene-based aggregation in the Breast Cancer Association Consortium cohorts including 83,471 cases and 59,199 controls. Low-frequency variants were aggregated for individual genes' coding and regulatory regions. Association results in European ancestry samples were compared to single-marker association results in the same cohort. Gene-based associations were also combined in meta-analysis across individuals with European, Asian, African, and Latin American and Hispanic ancestry. RESULTS: In European ancestry samples, 14 genes were significantly associated (q < 0.05) with BC. Of those, two genes, FMNL3 (P = 6.11 × 10-6) and AC058822.1 (P = 1.47 × 10-4), represent new associations. High FMNL3 expression has previously been linked to poor prognosis in several other cancers. Meta-analysis of samples with diverse ancestry discovered further associations including established candidate genes ESR1 and CBLB. Furthermore, literature review and database query found further support for a biologically plausible link with cancer for genes CBLB, FMNL3, FGFR2, LSP1, MAP3K1, and SRGAP2C. CONCLUSIONS: Using extended gene-based aggregation tests including coding and regulatory variation, we report identification of plausible target genes for previously identified single-marker associations with BC as well as the discovery of novel genes implicated in BC development. Including multi ancestral cohorts in this study enabled the identification of otherwise missed disease associations as ESR1 (P = 1.31 × 10-5), demonstrating the importance of diversifying study cohorts.

Identification, cloning, and characterization of two N-acetylgalactosamine-binding lectins from the albumen gland of Helix pomatia.

Helix pomatia agglutinin (HPA), the lectin from the albumen gland of the Roman snail, has been used in histochemical studies relating glycosylation changes to the metastatic potential of solid tumors. To facilitate the use of HPA in a clinical (diagnostic) setting, detailed analysis of the lectin, including cloning and recombinant production of HPA, is required. A combination of isoelectric focusing, amino acid sequence analysis, and cloning revealed two polypeptides in native HPA preparations (HPAI and HPAII), both consistent with GalNAc-binding lectins of the H-type family. Pairwise sequence alignment showed that HPAI and HPAII share 54% sequence identity whereas molecular modeling using SWISS-MODEL suggests they are likely to adopt similar tertiary structure. The inherent heterogeneity of native HPA highlighted the need for production of functional recombinant protein; this was addressed by preparing His-thioredoxin-tagged fusion products in Escherichia coli Rosetta-gami B (DE3) cells. The recombinant lectins agglutinated human blood group A erythrocytes whereas their oligosaccharide specificity, evaluated using glycan microarrays, showed that they predominantly bind glycans with terminal α-GalNAc residues. Surface plasmon resonance with immobilized GalNAc-BSA confirmed that recombinant HPAI and HPAII bind strongly with this ligand (K(d) = 0.60 nm and 2.00 nm, respectively) with a somewhat higher affinity to native HPA (K(d) = 7.67 nm). Recombinant HPAII also bound the breast cancer cells of breast cancer tissue specimens in a manner similar to native lectin. The recombinant HPA described here shows important potential for future studies of cancer cell glycosylation and as a reagent for cancer prognostication.

The lectin Helix pomatia agglutinin recognizes O-GlcNAc containing glycoproteins in human breast cancer.

There has been considerable interest in understanding the epitopes that bind the lectin Helix pomatia agglutinin (HPA) in breast cancer as the lectin has been shown to identify glycosylation changes associated with the development of metastatic disease. HPA has previously been shown to recognize aberrant O-linked α-N-acetylgalactosamine (GalNAcα)/mucin glycosylation in cancer, including exposed Tn epitopes. However, recent glycan-array analysis reported that diverse epitopes are also recognized by the lectin, e.g. consortium for functional glycomics (CFG) data: GalNAcα1,3Gal; ß-GalNAc; GlcNAcß1,4Gal. The intriguing observations from the CFG array led to this study, in which HPA-binding epitopes were localized and characterized in an in vitro model of breast cancer metastasis. HMT3522 (benign disease), BT474 (primary cancer) and T47D/MCF7 (metastatic cancer) cells were assessed in confocal microscopy-based co-localization studies and a glycoproteomic analysis based on 2-dimensional electrophoresis (2DE), western blotting and mass spectrometry was adopted. HPA binding correlated with levels of integrin α6, transcription factors heterogeneous nuclear ribonuclear protein (HnRNP) H1, HnRNP D-like, HnRNP A2/B1 as well as heat shock protein 27 (Hsp27), glial fibrillary acidic protein and enolase 1 (ENO1). These glycoproteins were non-detectable in the non-metastatic breast cancer cell lines. The recognition of HnRNPs, Hsp27 and ENO1 by HPA correlated with O-GlcNAcylation of these proteins. Integrin α6 was the most abundant HPA glycoprotein in the breast cancer cells with a metastatic phenotype; this concurred with previous findings in colorectal cancer. This is the first report in which HPA has been shown to bind O-GlcNAcylated transcription factors. This class of proteins represents a new means by which HPA differentiates cancer cells with an aggressive metastatic phenotype.

Development and Optimization of a Machine-Learning Prediction Model for Acute Desquamation After Breast Radiation Therapy in the Multicenter REQUITE Cohort.

Purpose: Some patients with breast cancer treated by surgery and radiation therapy experience clinically significant toxicity, which may adversely affect cosmesis and quality of life. There is a paucity of validated clinical prediction models for radiation toxicity. We used machine learning (ML) algorithms to develop and optimise a clinical prediction model for acute breast desquamation after whole breast external beam radiation therapy in the prospective multicenter REQUITE cohort study. Methods and Materials: Using demographic and treatment-related features (m = 122) from patients (n = 2058) at 26 centers, we trained 8 ML algorithms with 10-fold cross-validation in a 50:50 random-split data set with class stratification to predict acute breast desquamation. Based on performance in the validation data set, the logistic model tree, random forest, and naïve Bayes models were taken forward to cost-sensitive learning optimisation. Results: One hundred and ninety-two patients experienced acute desquamation. Resampling and cost-sensitive learning optimisation facilitated an improvement in classification performance. Based on maximising sensitivity (true positives), the "hero" model was the cost-sensitive random forest algorithm with a false-negative: false-positive misclassification penalty of 90:1 containing m = 114 predictive features. Model sensitivity and specificity were 0.77 and 0.66, respectively, with an area under the curve of 0.77 in the validation cohort. Conclusions: ML algorithms with resampling and cost-sensitive learning generated clinically valid prediction models for acute desquamation using patient demographic and treatment features. Further external validation and inclusion of genomic markers in ML prediction models are worthwhile, to identify patients at increased risk of toxicity who may benefit from supportive intervention or even a change in treatment plan.

Incorporating progesterone receptor expression into the PREDICT breast prognostic model.

BACKGROUND: Predict Breast (www.predict.nhs.uk) is an online prognostication and treatment benefit tool for early invasive breast cancer. The aim of this study was to incorporate the prognostic effect of progesterone receptor (PR) status into a new version of PREDICT and to compare its performance to the current version (2.2). METHOD: The prognostic effect of PR status was based on the analysis of data from 45,088 European patients with breast cancer from 49 studies in the Breast Cancer Association Consortium. Cox proportional hazard models were used to estimate the hazard ratio for PR status. Data from a New Zealand study of 11,365 patients with early invasive breast cancer were used for external validation. Model calibration and discrimination were used to test the model performance. RESULTS: Having a PR-positive tumour was associated with a 23% and 28% lower risk of dying from breast cancer for women with oestrogen receptor (ER)-negative and ER-positive breast cancer, respectively. The area under the ROC curve increased with the addition of PR status from 0.807 to 0.809 for patients with ER-negative tumours (p = 0.023) and from 0.898 to 0.902 for patients with ER-positive tumours (p = 2.3 × 10-6) in the New Zealand cohort. Model calibration was modest with 940 observed deaths compared to 1151 predicted. CONCLUSION: The inclusion of the prognostic effect of PR status to PREDICT Breast has led to an improvement of model performance and more accurate absolute treatment benefit predictions for individual patients. Further studies should determine whether the baseline hazard function requires recalibration.

Clinical relevance assessment of animal preclinical research (RAA) tool: development and explanation.

BACKGROUND: Only a small proportion of preclinical research (research performed in animal models prior to clinical trials in humans) translates into clinical benefit in humans. Possible reasons for the lack of translation of the results observed in preclinical research into human clinical benefit include the design, conduct, and reporting of preclinical studies. There is currently no formal domain-based assessment of the clinical relevance of preclinical research. To address this issue, we have developed a tool for the assessment of the clinical relevance of preclinical studies, with the intention of assessing the likelihood that therapeutic preclinical findings can be translated into improvement in the management of human diseases. METHODS: We searched the EQUATOR network for guidelines that describe the design, conduct, and reporting of preclinical research. We searched the references of these guidelines to identify further relevant publications and developed a set of domains and signalling questions. We then conducted a modified Delphi-consensus to refine and develop the tool. The Delphi panel members included specialists in evidence-based (preclinical) medicine specialists, methodologists, preclinical animal researchers, a veterinarian, and clinical researchers. A total of 20 Delphi-panel members completed the first round and 17 members from five countries completed all three rounds. RESULTS: This tool has eight domains (construct validity, external validity, risk of bias, experimental design and data analysis plan, reproducibility and replicability of methods and results in the same model, research integrity, and research transparency) and a total of 28 signalling questions and provides a framework for researchers, journal editors, grant funders, and regulatory authorities to assess the potential clinical relevance of preclinical animal research. CONCLUSION: We have developed a tool to assess the clinical relevance of preclinical studies. This tool is currently being piloted.

A data science approach for early-stage prediction of Patient's susceptibility to acute side effects of advanced radiotherapy.

The prediction by classification of side effects incidence in a given medical treatment is a common challenge in medical research. Machine Learning (ML) methods are widely used in the areas of risk prediction and classification. The primary objective of such algorithms is to use several features to predict dichotomous responses (e.g., disease positive/negative). Similar to statistical inference modelling, ML modelling is subject to the class imbalance problem and is affected by the majority class, increasing the false-negative rate. In this study, seventy-nine ML models were built and evaluated to classify approximately 2000 participants from 26 hospitals in eight different countries into two groups of radiotherapy (RT) side effects incidence based on recorded observations from the international study of RT related toxicity "REQUITE". We also examined the effect of sampling techniques and cost-sensitive learning methods on the models when dealing with class imbalance. The combinations of such techniques used had a significant impact on the classification. They resulted in an improvement in incidence status prediction by shifting classifiers' attention to the minority group. The best classification model for RT acute toxicity prediction was identified based on domain experts' success criteria. The Area Under Receiver Operator Characteristic curve of the models tested with an isolated dataset ranged from 0.50 to 0.77. The scale of improved results is promising and will guide further development of models to predict RT acute toxicities. One model was optimised and found to be beneficial to identify patients who are at risk of developing acute RT early-stage toxicities as a result of undergoing breast RT ensuring relevant treatment interventions can be appropriately targeted. The design of the approach presented in this paper resulted in producing a preclinical-valid prediction model. The study was developed by a multi-disciplinary collaboration of data scientists, medical physicists, oncologists and surgeons in the UK Radiotherapy Machine Learning Network.

Breast Cancer Risk Factors and Survival by Tumor Subtype: Pooled Analyses from the Breast Cancer Association Consortium.

BACKGROUND: It is not known whether modifiable lifestyle factors that predict survival after invasive breast cancer differ by subtype. METHODS: We analyzed data for 121,435 women diagnosed with breast cancer from 67 studies in the Breast Cancer Association Consortium with 16,890 deaths (8,554 breast cancer specific) over 10 years. Cox regression was used to estimate associations between risk factors and 10-year all-cause mortality and breast cancer-specific mortality overall, by estrogen receptor (ER) status, and by intrinsic-like subtype. RESULTS: There was no evidence of heterogeneous associations between risk factors and mortality by subtype (P adj > 0.30). The strongest associations were between all-cause mortality and BMI ≥30 versus 18.5-25 kg/m2 [HR (95% confidence interval (CI), 1.19 (1.06-1.34)]; current versus never smoking [1.37 (1.27-1.47)], high versus low physical activity [0.43 (0.21-0.86)], age ≥30 years versus <20 years at first pregnancy [0.79 (0.72-0.86)]; >0-<5 years versus ≥10 years since last full-term birth [1.31 (1.11-1.55)]; ever versus never use of oral contraceptives [0.91 (0.87-0.96)]; ever versus never use of menopausal hormone therapy, including current estrogen-progestin therapy [0.61 (0.54-0.69)]. Similar associations with breast cancer mortality were weaker; for example, 1.11 (1.02-1.21) for current versus never smoking. CONCLUSIONS: We confirm associations between modifiable lifestyle factors and 10-year all-cause mortality. There was no strong evidence that associations differed by ER status or intrinsic-like subtype. IMPACT: Given the large dataset and lack of evidence that associations between modifiable risk factors and 10-year mortality differed by subtype, these associations could be cautiously used in prognostication models to inform patient-centered care.