Identification of novel biomarkers for the prediction of subclinical coronary artery atherosclerosis in patients with rheumatoid arthritis: an exploratory analysis

BACKGROUND

Cardiovascular (CV) risk estimation calculators for the general population underperform in patients with rheumatoid arthritis (RA). The purpose of this study was to identify relevant protein biomarkers that could be added to traditional CV risk calculators to improve the capacity of coronary artery calcification (CAC) prediction in individuals with RA. In a second step, we quantify the improvement of this prediction of CAC when these circulating biomarkers are added to standard risk scores.

METHODS

A panel of 141 serum and plasma proteins, which represent a broad base of both CV and RA biology, were evaluated and prioritized as candidate biomarkers. Of these, 39 proteins were selected and measured by commercial ELISA or quantitative mass spectroscopy in 561 individuals with RA in whom a measure of CAC and frozen sera were available. The patients were randomly split 50:50 into a training/validation cohort. Discrimination (using area under the receiver operator characteristic curves) and re-classification (through net reclassification improvement and integrated discrimination improvement calculation) analyses were performed first in the training cohort and replicated in the validation cohort, to estimate the increase in prediction accuracy for CAC using the ACA/AHA (American College of Cardiology and the American Heart Association) score with, compared to without, addition of these circulating biomarkers.

RESULTS

The model containing ACC/AHA score plus cytokines (osteopontin, cartilage glycoprotein-39, cystatin C, and chemokine (C-C motif) ligand 18) and plus quantitative mass spectroscopy biomarkers (serpin D1, paraoxonase, and clusterin) had a statistically significant positive net reclassifications index and integrated discrimination improvement for the prediction of CAC, using ACC/AHA score without any biomarkers as the reference category. These results were confirmed in the validation cohort.

CONCLUSION

In this exploratory analysis, the addition of several circulating CV and RA biomarkers to a standard CV risk calculator yielded significant improvements in discrimination and reclassification for the presence of CAC in individuals with RA.

Accessing a New Dimension in TP53 Biology: Multiplex Long Amplicon Digital PCR to Specifically Detect and Quantitate Individual TP53 Transcripts

TP53, the most commonly-mutated gene in cancer, undergoes complex alternative splicing. Different TP53 transcripts play different biological roles, both in normal function and in the progression of diseases such as cancer. The study of TP53’s alternative RNA splice forms and their use as clinical biomarkers has been hampered by limited specificity and quantitative accuracy of current methods. TP53 RNA splice variants differ at both 5’ and 3’ ends, but because they have a common central region of 618 bp, the individual TP53 transcripts are impossible to specifically detect and precisely quantitate using standard PCR-based methods or short-read RNA sequencing. Therefore, we devised multiplex probe-based long amplicon droplet digital PCR (ddPCR) assays, which for the first time allow precise end-to-end quantitation of the seven major TP53 transcripts, with amplicons ranging from 0.85 to 1.85 kb. Multiple modifications to standard ddPCR assay procedures were required to enable specific co-amplification of these long transcripts and to overcome issues with secondary structure. Using these assays, we show that several TP53 transcripts are co-expressed in breast cancers, and illustrate the potential for this method to identify novel TP53 transcripts in tumour cells. This capability will facilitate a new level of biological and clinical understanding of the alternatively-spliced TP53 isoforms.

Evofosfamide for the treatment of human papillomavirus-negative head and neck squamous cell carcinoma

Evofosfamide (TH-302) is a clinical-stage hypoxia-activated prodrug of a DNA-crosslinking nitrogen mustard that has potential utility for human papillomavirus (HPV) negative head and neck squamous cell carcinoma (HNSCC), in which tumor hypoxia limits treatment outcome. We report the preclinical efficacy, target engagement, preliminary predictive biomarkers and initial clinical activity of evofosfamide for HPV-negative HNSCC. Evofosfamide was assessed in 22 genomically characterized cell lines and 7 cell line-derived xenograft (CDX), patient-derived xenograft (PDX), orthotopic, and syngeneic tumor models. Biomarker analysis used RNA sequencing, whole-exome sequencing, and whole-genome CRISPR knockout screens. Five advanced/metastatic HNSCC patients received evofosfamide monotherapy (480 mg/m2 qw × 3 each month) in a phase 2 study. Evofosfamide was potent and highly selective for hypoxic HNSCC cells. Proliferative rate was a predominant evofosfamide sensitivity determinant and a proliferation metagene correlated with activity in CDX models. Evofosfamide showed efficacy as monotherapy and with radiotherapy in PDX models, augmented CTLA-4 blockade in syngeneic tumors, and reduced hypoxia in nodes disseminated from an orthotopic model. Of 5 advanced HNSCC patients treated with evofosfamide, 2 showed partial responses while 3 had stable disease. In conclusion, evofosfamide shows promising efficacy in aggressive HPV-negative HNSCC, with predictive biomarkers in development to support further clinical evaluation in this indication.

A steroid metabolizing gene variant in a polyfactorial model improves risk prediction in a high incidence breast cancer population

Background

We have combined functional gene polymorphisms with clinical factors to improve prediction and understanding of sporadic breast cancer risk, particularly within a high incidence Caucasian population.

Methods

A polyfactorial risk model (PFRM) was built from both clinical data and functional single nucleotide polymorphism (SNP) gene candidates using multivariate logistic regression analysis on data from 5022 US Caucasian females (1671 breast cancer cases, 3351 controls), validated in an independent set of 1193 women (400 cases, 793 controls), and reassessed in a unique high incidence breast cancer population (165 cases, 173 controls) from Marin County, CA.

Results

The optimized PFRM consisted of 22 SNPs (19 genes, 6 regulating steroid metabolism) and 5 clinical risk factors, and its 5-year and lifetime risk prediction performance proved significantly superior (~ 2-fold) over the Gail model (Breast Cancer Risk Assessment Tool, BCRAT), whether assessed by odds (OR) or positive likelihood (PLR) ratios over increasing model risk levels. Improved performance of the PFRM in high risk Marin women was due in part to genotype enrichment by a CYP11B2 (-344T/C) variant.

Conclusions and general significance

Since the optimized PFRM consistently outperformed BCRAT in all Caucasian study populations, it represents an improved personalized risk assessment tool. The finding of higher Marin County risk linked to a CYP11B2 aldosterone synthase SNP associated with essential hypertension offers a new genetic clue to sporadic breast cancer predisposition.

•

A polyfactorial breast cancer risk assessment model (PFRM) was built and validated.

•

The optimized PFRM incorporates both genetic (22 SNPs/19 genes) and clinical risk factors.

•

The PFRM was further validated in a high risk USA/Marin breast cancer population.

•

This PFRM consistently performed significantly better than the BCRAT (Gail model).

•

A functional aldosterone synthase SNP in PFRM improved predictive performance in Marin.

Validation of the power model of ovarian nongrowing follicle depletion associated with aging in women

OBJECTIVE

To validate recently proposed models of ovarian nongrowing follicle (NGF) decay associated with aging within the context of an independent data set.

DESIGN

Prospective investigation.

SETTING

Academic medical center.

PATIENT(S)

Normal appearing ovaries collected from 52 women (age 28-51 years) undergoing oophorectomy for benign gynecologic indications.

INTERVENTION(S)

Determining ovarian NGF counts with systematic random sampling rules and a validated fractionator/optical disector technique. The goodness-of-fit of predicted NGF counts based on the power and double Gaussian models and those observed in the validation set was assessed with the calculation of the Akaike information criterion and R(2) values.

MAIN OUTCOME MEASURE(S)

The goodness-of-fit between observed and expected ovarian NGF counts.

RESULT(S)

The power model was an excellent fit to the observed data. The average difference between the observed and expected NGF count was 0.161 (95% CI, -0.058, 0.327). In the present study population, the power model was a superior fit to the observed data compared with the double Gaussian model.

CONCLUSION(S)

This prospective investigation with an independent set of ovarian NGF counts validates the power model as an excellent characterization of the ovarian NGF decline associated with aging.

Abstract P2-14-05: Sporadic breast cancer risk assessment in women without family history and biopsy using the OncoVue® polyfactorial model

Background: The majority of breast cancer cases are considered to be sporadic and occur in women without a strong family history. Accurate estimation of a woman's risk of developing breast cancer is essential for guiding decisions regarding surveillance and prevention. The Breast Cancer Risk Assessment Tool (BCRAT) uses only clinical factors and has been shown to be of marginal value for individual risk counseling. Even though the BCRAT risks for women without a biopsy and no first degree relative with breast cancer are low, a significant fraction of cases occur in these women. OncoVue® is a logistic regression model developed from individual genetic and personal factor data from a large case-control study. In three independent validation populations, including a blinded validation, OncoVue has been shown to significantly outperform BCRAT. Here we compare OncoVue to the BCRAT in stratifying risk in women without a first degree relative with breast cancer that have never had a breast biopsy.

Materials and Methods: Absolute risks for Caucasian participants ranging in age from 35 to 89 were analyzed for a case-control study conducted in six distinct geographic regions of the United States. This analysis focused on participants that had never had a biopsy and reported no first degree relatives with breast cancer (cases prior to diagnosis/controls at time of enrollment). DNAs from a total of 2729 women (842 cases and 1887 controls) were genotyped for 22 SNP variants and genotype information was combined with clinical risk information to calculate 5-year and lifetime risks. Clinical factor information alone was used to calculate BCRAT risk scores for comparison to OncoVue.

Results: The 5-year risks for OncoVue ranged from 0.10 to 5.0% (50-fold) compared to 0.20 to 3.0% (15-fold) for BCRAT risks. OncoVue lifetime risks ranged from 2.4 to 25.0% (11-fold) compared to 4 to 16% (4-fold) for BCRAT risks. The OncoVue model 5-year risk placed 11% of this sub-population of controls at ≥ 2.0% risk compared to the BCRAT which placed only 3.5% at this risk. The BCRAT did not place anyone at a lifetime risk over 16% while OncoVue placed 10% of this sub-population of controls above 16%. The Odds Ratio (OR) was determined at increasing model risk output levels. Over comparable ranges, OncoVue 5-year risks exhibited statistically significant ORs while BCRAT ORs did not reach significance. OncoVue associations for lifetime risk reach high levels. For example, an OR of 5.0 (p = 0.03) was observed at a risk level ≥ 23%.

Conclusion: The OncoVue model stratifies risk over a much wider range than the BCRAT in this group of women at low risk by traditional methods. In the 5-year risk frame OncoVue exhibited improved performance in estimating individual risk compared to the BCRAT. Due to the lack of any elevated lifetime risks generated by the BCRAT, a direct comparison of lifetime risks for OncoVue to the BCRAT was not possible. The OncoVue model which uses both genetics and clinical factors significantly improves individualized breast cancer risk estimation. These results demonstrate the importance of the genetic component of the OncoVue model in accurately estimating individual risk of sporadic breast cancer.

Citation Information: Cancer Res 2013;73(24 Suppl): Abstract nr P2-14-05.

Abstract P4-12-05: Enrichment of aldosterone synthase (CYP11B2) gene C-allele carriers in women at high risk in the OncoVue® polyfactorial risk model

Background: Models for accurately assessing breast cancer risk are utilized for guiding clinical decisions regarding early detection and prevention. We developed a polyfactorial risk model (PFRM; OncoVue) from analysis of a large case-control database including both functional gene polymorphism and clinical factor data. A multivariate logistic regression based statistical protocol was used to derive a single risk model consisting of 22 single nucleotide polymorphisms (SNPs) in 19 genes and 5 traditional risk factors. Several studies have shown that the PFRM is significantly improved in individualized risk estimation compared to the Breast Cancer Risk Assessment Tool (BCRAT) or Gail model. Here we have examined the genotype frequencies of women placed in the highest decile of risk by the PFRM compared to both the overall control population and to women in the control population placed at the lowest decile of risk.

Materials and Methods: Analyses were performed on 1671 breast cancer cases and 3351 controls that enrolled in a case-control study conducted in six regions of the United States. DNAs were genotyped for the 22 genetic polymorphisms and combined with clinical risk factor information to calculate PFRM risk estimates for the individual participants. The PFRM lifetime scores were used to stratify the upper and lower deciles of risk. The genotype frequencies for these stratified risk groups were determined and compared to the overall control population and to each other (upper vs. lower decile). Odds Ratios (OR) were determined for the CYP11B2 genotypes because they exhibited the largest differences in frequency among the groups.

Results: For CYP11B2 comparison of the genotype frequencies in the highest decile risk group to the overall genotype frequencies of controls identified significant enrichment of the C-allele carriers in the highest risk group. The OR for C-allele carriers (C/C+C/T) compared to the most common T/T genotype was 3.0 (1.9, 4.8; 95% C.I.) with a p-value <0.0001. Similarly for the upper decile risk group compared to the lower decile risk group the association of C-allele carriers was even stronger with OR = 11.0 (6.6,18.4; p <0.0001). In the upper decile of risk 86% of the individuals were C-allele carriers compared to only 38% in the lower decile of risk.

Conclusions: Our previous studies have shown an age dependent association of the CYP11B2 C/C and C/T genotypes with breast cancer risk. The enrichment of these genotypes in women in the upper decile risk group as determined by the PFRM further reinforces the role of this genotype in influencing breast cancer risk and suggests that intervention in the aldosterone synthase pathway be explored as a method of breast cancer prevention.

Citation Information: Cancer Res 2012;72(24 Suppl):Abstract nr P4-12-05.

Abstract P6-09-04: Accurate Identification of Women at High Risk of Breast Cancer Using OncoVue®

Background: Accurate estimation of individualized risk of developing breast cancer is critical to making informed clinical decisions regarding surveillance and prevention. OncoVue is a logistic regression model developed from individual genetic and personal factor data from a large case-control study. In three independent validation populations, including a blinded validation, OncoVue has been shown to significantly outperform both the Gail model and composite risk scores produced by combining GWAS SNP risks with the Gail model. Here we have examined the application of the OncoVue risk score to identifying women that exceed a lifetime risk of 20% and would be candidates for surveillance by MRI. Materials and Methods: Risk scores were analyzed for participants ranging in age from 35 to 89 for a subset of participants that had enrolled in a larger case-control study conducted in six distinct geographic regions of the United States. As in previous studies, the original assignment of the samples into a discovery set of 4768 Caucasian women (1592 cases and 3176 controls) and two independent validation sets was used. The first validation set consisted of 1137 Caucasian women (376 cases and 761 controls) while the second consisted of 494 African American women (149 cases and 345 controls). DNAs were genotyped for 22 SNP variants and genotype information was combined with personal factors information to calculate the OncoVue risk scores for the individual participants. Personal factor information was used to calculate Gail model risk scores. Results: For both models, positive likelihood ratios (PLR) were calculated as the proportion of patients with breast cancer with an elevated risk estimate (≥20%) divided by the proportion of disease-free individuals with an elevated risk estimate. In both the discovery and validation sets, OncoVue exhibited a 1.6- to 1.8-fold improvement compared to the GM in more accurately assigning elevated risk estimates to breast cancer cases rather than controls. At higher risk thresholds, the fold improvement increasedand exceeded 2.5 in some sample sets. Another measure of clinical utility of a risk assessment test is the ability to correctly place more breast cancer cases at elevated risk. This was examined with a normalized analysis of the number of breast cancer cases assigned elevated risk (≥20%) estimates by OncoVue compared to the Gail model. In these analyses, the percent improvement in number of cases placed at elevated risk by OncoVue ranged from 20 to 33%. Conclusion: OncoVue exhibited significantly improved performance, compared to the Gail model alone, in estimating lifetime individual risk and accurately assigning elevated risk to cases. The improved performance of OncoVue was similar to that observed in previous studies using other risk frames and thresholds. Together these studies demonstrate the improved ability of OncoVue to produce more accurate individualized breast cancer risk estimation.

Citation Information: Cancer Res 2010;70(24 Suppl):Abstract nr P6-09-04.