Predicting BRCA1 and BRCA2 gene mutation carriers: comparison of LAMBDA, BRCAPRO, Myriad II, and modified Couch models

An advanced machine learning method for simultaneous breast cancer risk prediction and risk ranking in Chinese population: A prospective cohort and modeling study

BACKGROUND

Breast cancer (BC) risk-stratification tools for Asian women that are highly accurate and can provide improved interpretation ability are lacking. We aimed to develop risk-stratification models to predict long- and short-term BC risk among Chinese women and to simultaneously rank potential non-experimental risk factors.

METHODS

The Breast Cancer Cohort Study in Chinese Women, a large ongoing prospective dynamic cohort study, includes 122,058 women aged 25-70 years old from the eastern part of China. We developed multiple machine-learning risk prediction models using parametric models (penalized logistic regression, bootstrap, and ensemble learning), which were the short-term ensemble penalized logistic regression (EPLR) risk prediction model and the ensemble penalized long-term (EPLT) risk prediction model to estimate BC risk. The models were assessed based on calibration and discrimination, and following this assessment, they were externally validated in new study participants from 2017 to 2020.

RESULTS

The AUC values of the short-term EPLR risk prediction model were 0.800 for the internal validation and 0.751 for the external validation set. For the long-term EPLT risk prediction model, the area under the receiver operating characteristic curve was 0.692 and 0.760 in internal and external validations, respectively. The net reclassification improvement index of the EPLT relative to the Gail and the Han Chinese Breast Cancer Prediction Model (HCBCP) models for external validation was 0.193 and 0.233, respectively, indicating that the EPLT model has higher classification accuracy.

CONCLUSIONS

We developed the EPLR and EPLT models to screen populations with a high risk of developing BC. These can serve as useful tools to aid in risk-stratified screening and BC prevention.

Utilization of breast cancer risk prediction models by cancer genetic counselors in clinical practice predominantly in the United States

Risk assessment in cancer genetic counseling is essential in identifying individuals at high risk for developing breast cancer to recommend appropriate screening and management options. Historically, many breast cancer risk prediction models were developed to calculate an individual's risk to develop breast cancer or to carry a pathogenic variant in the BRCA1 or BRCA2 genes. However, how or when genetic counselors use these models in clinical settings is currently unknown. We explored genetic counselors' breast cancer risk model usage patterns including frequency of use, reasons for using or not using models, and change in usage since the adoption of multi-gene panel testing. An online survey was developed and sent to members of the National Society of Genetic Counselors; board-certified genetic counselors whose practice included cancer genetic counseling were eligible to participate in the study. The response rate was estimated at 23% (243/1,058), and respondents were predominantly working in the United States. The results showed that 93% of all respondents use at least one breast cancer risk prediction model in their clinical practice. Among the six risk models selected for the study, the Tyrer-Cuzick (IBIS) model was used most frequently (95%), and the BOADICEA model was used least (40%). Determining increased or decreased surveillance and breast MRI eligibility were the two most common reasons for most model usage, while time consumption and difficulty in navigation were the two most common reasons for not using models. This study provides insight into perceived benefits and limitations of risk models in clinical use in the United States, which may be useful information for software developers, genetic counseling program curriculum developers, and currently practicing cancer genetic counselors.

Assessing Risk of Breast Cancer: A Review of Risk Prediction Models

Accurate and individualized breast cancer risk assessment can be used to guide personalized screening and prevention recommendations. Existing risk prediction models use genetic and nongenetic risk factors to provide an estimate of a woman's breast cancer risk and/or the likelihood that she has a BRCA1 or BRCA2 mutation. Each model is best suited for specific clinical scenarios and may have limited applicability in certain types of patients. For example, the Breast Cancer Risk Assessment Tool, which identifies women who would benefit from chemoprevention, is readily accessible and user-friendly but cannot be used in women under 35 years of age or those with prior breast cancer or lobular carcinoma in situ. Emerging research on deep learning-based artificial intelligence (AI) models suggests that mammographic images contain risk indicators that could be used to strengthen existing risk prediction models. This article reviews breast cancer risk factors, describes the appropriate use, strengths, and limitations of each risk prediction model, and discusses the emerging role of AI for risk assessment.

Estimating CDKN2A mutation carrier probability among global familial melanoma cases using GenoMELPREDICT

BACKGROUND

Although rare in the general population, highly penetrant germline mutations in CDKN2A are responsible for 5%-40% of melanoma cases reported in melanoma-prone families. We sought to determine whether MELPREDICT was generalizable to a global series of families with melanoma and whether performance improvements can be achieved.

METHODS

In total, 2116 familial melanoma cases were ascertained by the international GenoMEL Consortium. We recapitulated the MELPREDICT model within our data (GenoMELPREDICT) to assess performance improvements by adding phenotypic risk factors and history of pancreatic cancer. We report areas under the curve (AUC) with 95% confidence intervals (CIs) along with net reclassification indices (NRIs) as performance metrics.

RESULTS

MELPREDICT performed well (AUC 0.752, 95% CI 0.730-0.775), and GenoMELPREDICT performance was similar (AUC 0.748, 95% CI 0.726-0.771). Adding a reported history of pancreatic cancer yielded discriminatory improvement (P < .0001) in GenoMELPREDICT (AUC 0.772, 95% CI 0.750-0.793, NRI 0.40). Including phenotypic risk factors did not improve performance.

CONCLUSION

The MELPREDICT model functioned well in a global data set of familial melanoma cases. Adding pancreatic cancer history improved model prediction. GenoMELPREDICT is a simple tool for predicting CDKN2A mutational status among melanoma patients from melanoma-prone families and can aid in directing these patients to receive genetic testing or cancer risk counseling.

Prediction of BRCA1 germ-line mutation status in patients with breast cancer using histoprognosis grade, MS110, Lys27H3, vimentin, and KI67

Family structure, lack of reliable information, cost, and delay are usual concerns when deciding to perform BRCA analyses. Testing breast cancer tissues with four antibodies (MS110, lys27H3, vimentin, and KI67) in addition to grade evaluation enabled us to rapidly select patients for genetic testing identification. We constituted an initial breast cancer tissue microarray, considered as a learning set, comprising 27 BRCA1 and 81 sporadic tumors. A second independent validation set of 28 BRCA1 tumors was matched to 28 sporadic tumors using the same original conditions. We investigated morphological parameters and 21 markers by immunohistochemistry. A logistic regression model was used to select the minimal number of markers providing the best model to predict BRCA1 status. The model was applied to the validation set to estimate specificity and sensibility. In the initial set, univariate analyses identified 11 markers significantly associated with BRCA1 status. Then, the best multivariate model comprised only grade 3, MS110, Lys27H3, vimentin, and KI67. When applied to the validation set, BRCA1 tumors were correctly classified with a sensitivity of 83% and a specificity of 81%. The performance of this model was superior when compared to other profiles. This study offers a new rapid and cost-effective method for the prescreening of patients at high risk of being BRCA1 mutation carriers, to guide genetic testing, and finally to provide appropriate preventive measures, advice, and treatments including targeted therapy to patients and their families.

Validation of three BRCA1/2 mutation-carrier probability models Myriad, BRCAPRO and BOADICEA in a population-based series of 183 German families

Many studies have evaluated the performance of risk assessment models for BRCA1/2 mutation carrier probabilities in different populations, but to our knowledge very few studies have been conducted in the German population so far. In the recent study, we validated the performance of three risk calculation models by names BRCAPRO, Myriad and BOADICEA in 183 German families who had undergone molecular testing of mutations in BRCA1 and BRCA2 with an indication based on clinical criteria regarding their family history of cancer. The sensitivity and specificity at the conventional threshold of 10% as well as for a threshold of 20% were evaluated. The ability to discriminate between carriers and non-carriers was judged by the area under the receiver operating characteristics curve. We further focused on the performance characteristic of these models in patients carrying large genomic rearrangements as a subtype of mutations which is currently gaining increasing importance. BRCAPRO and BOADICEA performed almost equally well in our patient population, but we found a lack of agreement to Myriad. The results obtained from this study were consistent with previously published results from other population and racial/ethnic groups. We suggest using model specific decision thresholds instead of the recommended universal value of 10%. We further suggest integrating the CaGene5 software package, which includes BRCAPRO and Myriad, in the genetic counselling of German families with suspected inherited breast and ovarian cancer because of the good performance of BRCAPRO and the substantial ease of use of this software.

Predicting BRCA1 and BRCA2 gene mutation carriers: comparison of PENN II model to previous study

A number of models have been developed to predict the probability that a person carries a detectable germline mutation in the BRCA1 or BRCA2 genes. Their relative performance in a clinical setting is variable. To compare the performance characteristics of a web-based BRCA1/BRCA2 gene mutation prediction model: the PENNII model ( www.afcri.upenn.edu/itacc/penn2 ), with studies done previously at our institution using four other models including LAMBDA, BRCAPRO, modified PENNI (Couch) tables, and Myriad II tables collated by Myriad Genetics Laboratories. Proband and family cancer history data were analyzed from 285 probands from unique families (27 Ashkenazi Jewish; 277 female) seen for genetic risk assessment in a multispecialty tertiary care group practice. All probands had clinical testing for BR.CA1 and BRCA2 mutations conducted in the same single commercial laboratory. The performance for PENNII results were assessed by the area under the receiver operating characteristic curve (AUC) of sensitivity versus 1-specificity, as a measure of ranking. The AUCs of the PENNII model were higher for predicting BRCA1 than for BRCA2 (81 versus 72%). The overall AUC was 78.7%. PENN II model for BRCA1/2 prediction performed well in this population with higher AUC compared with our experience using four other models. The ease of use of the PENNII model is compatible with busy clinical practices.

Efficiency of BRCAPRO and Myriad II mutation probability thresholds versus cancer history criteria alone for BRCA1/2 mutation detection

Considerable differences exist amongst countries in the mutation probability methods and thresholds used to select patients for BRCA1/2 genetic screening. In order to assess the added value of mutation probability methods, we have retrospectively calculated the BRCAPRO and Myriad II probabilities in 306 probands who had previously been selected for DNA-analysis according to criteria based on familial history of cancer. DNA-analysis identified 52 mutations (16.9%) and 11 unclassified variants (UVs, 3.6%). Compared to cancer history, a threshold ≥10% with BRCAPRO or with Myriad II excluded about 40% of the patients from analysis, including four with a mutation and probabilities <10% with both programs. All four probands had a BRCA2 mutation. BRCAPRO and Myriad II showed similar specificity at 10% threshold, overall BRCAPRO was more sensitive than Myriad II for the detection of mutations. Only two of the probands with an UV had probabilities >20% with BRCAPRO and Myriad II. In summary, BRCAPRO and Myriad II are more efficient than cancer history alone to exclude patients without a mutation. BRCAPRO performs better for the detection of BRCA1 mutations than of BRCA2 mutations. The Myriad II scores provided no additional information than the BRCAPRO scores alone for the detection of patients with a mutation. The use of thresholds excluded from analysis the majority of patients carrying an UV.

The application of genetics and genomics to cancer prevention

Advances in technology have accelerated the translation of genetics and genomics into the arena of cancer prevention. This provides unique opportunities to individualize cancer risk prediction so early intervention can either modify risk or allow for early diagnosis thereby potentially decreasing the morbidity and mortality of cancer and containing costs. While the full potential of these genetic/genomic discoveries have yet to be realized, many have clear clinical relevance such as the value of family history and/or tumor profiling to identify those who may harbor a mutation in a cancer susceptibility gene and are therefore candidates for genetic testing. Here, we provide an overview of the scope of genetic and genomic influences on cancer risk assessment and the entire spectrum of cancer prevention.

Screening for genomic rearrangements in BRCA1 and BRCA2 genes in Czech high-risk breast/ovarian cancer patients: high proportion of population specific alterations in BRCA1 gene

Large genomic rearrangements (LGR) represent substantial proportion of pathogenic mutations in the BRCA1 gene, whereas the frequency of rearrangements in the BRCA2 gene is low in many populations. We screened for LGRs in BRCA1 and BRCA2 genes by multiplex ligation-dependent probe amplification (MLPA) in 521 unrelated patients negative for BRCA1/2 point mutations selected from 655 Czech high-risk breast and/or ovarian cancer patients. Besides long range PCR, a chromosome 17-specific oligonucleotide-based array comparative genomic hybridization (aCGH) was used for accurate location of deletions. We identified 14 patients carrying 8 different LGRs in BRCA1 that accounted for 12.3% of all pathogenic BRCA1 mutations. No LGRs were detected in the BRCA2 gene. In a subgroup of 239 patients from high-risk families, we found 12 LGRs (5.0%), whereas two LGRs were revealed in a subgroup of 282 non-familial cancer cases (0.7%). Five LGRs (deletion of exons 1-17, 5-10, 13-19, 18-22 and 21-24) were novel; two LGRs (deletion of exons 5-14 and 21-22) belong to the already described Czech-specific mutations; one LGR (deletion of exons 1-2) was reported from several countries. The deletions of exons 1-17 and 5-14, identified each in four families, represented Czech founder mutations. The present study indicates that screening for LGRs in BRCA1 should include patients from breast or ovarian cancer families as well as high-risk patients with non-familial cancer, in particular cases with early-onset breast or ovarian cancer. On the contrary, our analyses do not support the need to screen for LGRs in the BRCA2 gene. Implementation of chromosome-specific aCGH could markedly facilitate the design of primers for amplification and sequence analysis of junction fragments, especially in deletions overlapping gene boundaries.

Validation of predictive models for germline mutations in DNA mismatch repair genes in colorectal cancer

Lynch syndrome is defined by the presence of germline mutations in mismatch repair (MMR) genes. Several models have been recently devised that predict mutation carrier status (Myriad Genetics, Wijnen, Barnetson, PREMM and MMRpro models). Families at moderate-high risk for harboring a Lynch-associated mutation, referred to the BC Cancer Agency (BCCA) Hereditary Cancer Program (HCP), underwent mutation analysis, immunohistochemistry and/or microsatellite testing. Seventy-two tested cases were included. Twenty-five patients were mutation positive (34.7%) and 47 were mutation negative (65.3%). Nineteen of 43 patients who were both microsatellite stable and normal on immunohistochemistry for MLH1 and MSH2 were also genotyped for mutations in these genes; all 19 were negative for MMR gene mutations. Model-derived probabilities of harboring a MMR gene mutation in the proband were calculated and compared to observed results. The area under the ROC curves were 0.75 (95%CI; 0.63-0.87), 0.86 (0.7-0.96), 0.89 (0.82-0.97), 0.89 (0.81-0.98) and 0.93 (0.86-0.99) for the Myriad, Barnetson, Wijnen, MMRpro and PREMM models, respectively. The Amsterdam II criteria had a sensitivity and specificity of 0.76 and 0.74, respectively, in this cohort. The PREMM model demonstrated the best performance for predicting carrier status based on the positive likelihood ratios at the >10%, >20% and >30% probability thresholds. In this referred cohort, the PREMM model had the most favorable concordance index and predictive performance for carrier status based on the positive LR. These prediction models (PREMM, MMRPro and Wijnen) may soon replace the Amsterdam II and revised Bethesda criteria as a prescreening tool for Lynch mutations.

A screening and prevention programme serving an ethnically diverse population of women at high risk of developing breast and/or ovarian cancer

INTRODUCTION

We describe a screening and prevention programme primarily targeting under-served minority women at high risk of breast and/or ovarian cancer. Women attending this Bellevue Hospital Center (BHC) Clinic were either self-referred from a variety of special outreach programmes or referred internally by medical professionals caring for relatives or friends. Our objective was to delineate referral sources and preliminary risk-assessment findings in relation to demographic features in this population.

METHODS

Following a detailed family and personal history intake and physical examination, each woman on her initial visit is categorized into a low (standard) risk, high-risk or indeterminate-risk group. Women found to be at high risk of developing breast and/or ovarian cancers are referred for further testing, additional screening measures, or participation in chemoprevention trials. All other women are counselled concerning follow-up and lifestyle issues.

RESULT

Between 2003 and 2007, 171 women for whom complete information was obtained were analysed. Thirty-four of the women were Caucasians (19.8%) and 137 (80.2%) were ethnically diverse minority women. Sixty-two (36.2%) were found to be at high risk with a median age of 42 years. The majority of the high-risk women were referred to the clinic by medical professionals (58%), most of whom were from within the BHC health care system. In fact, one-fourth of the referrals were women who carried a diagnosis of cancer, mostly arising in the breast, and who were concerned with risks to other family members. Trends in genetic testing results indicate fewer mutations among high-risk Asians than among other ethnicities.

CONCLUSION

Accurate risk assessments and implementation of screening and prevention measures have been challenging during the first few years of operation. Nevertheless, the need for providing consultation from internal referrals and the potential for genetic and psychosocial research in an ethnically diverse population are powerful incentives for continuing to evolve these services.

Genetic testing for cancer susceptibility

Genetic testing for mutations in genes associated with an inherited predisposition to cancer is rapidly moving outside specialty genetic services and into mainstream health care. Surgeons, as front-line providers of cancer care, are uniquely positioned to identify those who may benefit from genetic testing and institute changes to their health care management based on those results. This article provides an overview of the critical elements of the process of genetic testing for cancer susceptibility.