User login
Elisha Hughes, PhD, director of research biostatistics at Myriad Genetics (which funded the study), and colleagues combined a risk model containing 149 single-nucleotide polymorphisms (SNPs), of which just over one-third were related to genetic ancestry, with the Tyrer-Cuzick (TC) breast cancer risk model.
The resulting combined risk score, which was developed in a cohort of over 145,000 women and validated in another group of almost 69,000 women, was not only well calibrated, but also able to reclassify just over 17% of women into a different risk group versus the clinical model.
The research (abstract P2-11-21) was presented at the San Antonio Breast Cancer Symposium on Dec. 8.
“This is the first breast cancer risk model based on a polygenic score, the 149-SNP PRS, that incorporates genetically determined ancestral composition and is validated for diverse ancestries,” the team reported.
The combined model substantially improved risk stratification over TC alone and may “lead to enhanced breast cancer risk reduction strategies, such as increased surveillance and use of preventive medications,” the researchers reported.
Breast cancer has a substantial genetic component that can “inform risk prediction and personalized preventive measures.” However, polygenic risk scores are largely derived from studies of women of European descent and tend to have poor performance in non-European ancestries.
Combined score substantially improved risk stratification over TC alone
The research team developed a polygenic risk score based on 149 SNPs for women of diverse backgrounds who did not have pathologic variants in breast cancer susceptibility genes, and included 56 ancestry-informative variants with 93 BC-associated variants. They combined the 149-SNP polygenic risk score with the TC risk model to create a combined risk score that was developed in a cohort of 145,786 women who were unaffected by breast cancer, following a fixed-stratified model to avoid double counting between confounded factors.
Of the women included in the cohort, 69.1% were of European descent, while 10.2% were Hispanic, 10.0% Black/African, 1.9% Asian, and 8.8% all other groups.
An independent cohort of 68,803 women of a similar ethnic distribution was then used to evaluate the calibration of the combined risk score against the TC risk model alone, and to examine the relative contributions of the 149-SNP PRS, family history, and other clinical factors.
The results showed that, overall, the combined risk score was well calibrated across ancestries and percentiles of risks, and the absolute lifetime risks were similar to those derived from the TC risk model alone. The only exception was Hispanic carriers of a protective Amerindian SNP who had a lower score on the combined risk score than the TC model.
Using an ANOVA model, the team found that family history contributed 48% to the lifetime risk of breast cancer, while the 149-SNP PRS contributed 35% and other factors 17%. Family history was weakly, but significantly correlated with the 149-SNP PRS.
Determining the impact of adding the 149-SNP PRS to the TC risk model on risk classification, the team showed that across all ancestries, 17.3% of women were reclassified by the combined risk score versus the TC model alone, with 10.8% having their lifetime risk increased to high risk and 29.1% having their risk decreased by the combined model to low risk.
The largest reclassifications were seen for women of European descent, while the smallest were for Black/African women.
Study may have ‘cracked the code’
“What’s exciting is that I think we kind-of ‘cracked the code’ to some extent of how to do this across diseases for all ancestries,” Thomas P. Slavin, MD, chief medical officer at Myriad Genetics, said in an interview. “The adaptation for breast cancer risk stratification and the new panel [is] for breast cancer across all ancestries, but what we developed is something that could be used across diabetes, or colon cancer, or anything.”
He explained that they realized that “for each one of these little hot spots” in the SNPs, “that make one person different from another, you really need to find out where in the world that originated from. So, if you have genetic ancestry on an individual, you can say this spot in the genome has more of an African ancestry to it, or a European ancestry, and then you can weight it appropriately by the population.”
Dr. Slavin said that standard PRSs that simply add up SNPs are “pretty good” and “add a lot” to risk stratification, “but to fine-tune it a little bit and make the best risk model, you really do need to bring in clinical and family history factors.”
Montserrat García-Closas, MD, DrPH, deputy director of the cancer epidemiology and genetics for the National Cancer Institute, said the study is of interest, but “does not give information on how ancestry was considered in the models used to derive the scores.” She also cautioned that the method used in the study to calibrate the model seems “to mean a comparison of scores, rather than comparing the observed and expected risk in prospective cohorts by ancestry groups. This would be a way to estimate bias in risk prediction by ancestry.”
Nevertheless, Dr. García-Closas said the degree of risk reclassification seen with the combined risk score is as expected and pointed to recent work by her and her colleagues in which they tested an integrated model incorporating classical risk factors and a 313-variant PRS to predict breast-cancer risk and achieved similar results.
Several study authors disclosed ties with Myriad Genetics, as well as AstraZeneca, Bristol Myers Squibb, Clovis Oncology, Helix BioPharma, Konica Minolta, Ambry Genetics, Invitae, Stryker, GAIL, Phenogen Sciences, Novartis, Pfizer, CancerIQ, Tempus, 54gene, Color Genetics, Roche/Genentech, ImpediMed, Prelude Therapeutics, BD, Agendia, Targeted Medical Education, Cerebrotech Medical Systems, Integra LifeSciences, Puma Biotechnology, GeneDX/BioReference, Change Health Care, Research to Practice, Clinical Care Options, Physician Education Resource, and Daiichi Sankyo.
The headline for this article was updated on 1/6/22.
Elisha Hughes, PhD, director of research biostatistics at Myriad Genetics (which funded the study), and colleagues combined a risk model containing 149 single-nucleotide polymorphisms (SNPs), of which just over one-third were related to genetic ancestry, with the Tyrer-Cuzick (TC) breast cancer risk model.
The resulting combined risk score, which was developed in a cohort of over 145,000 women and validated in another group of almost 69,000 women, was not only well calibrated, but also able to reclassify just over 17% of women into a different risk group versus the clinical model.
The research (abstract P2-11-21) was presented at the San Antonio Breast Cancer Symposium on Dec. 8.
“This is the first breast cancer risk model based on a polygenic score, the 149-SNP PRS, that incorporates genetically determined ancestral composition and is validated for diverse ancestries,” the team reported.
The combined model substantially improved risk stratification over TC alone and may “lead to enhanced breast cancer risk reduction strategies, such as increased surveillance and use of preventive medications,” the researchers reported.
Breast cancer has a substantial genetic component that can “inform risk prediction and personalized preventive measures.” However, polygenic risk scores are largely derived from studies of women of European descent and tend to have poor performance in non-European ancestries.
Combined score substantially improved risk stratification over TC alone
The research team developed a polygenic risk score based on 149 SNPs for women of diverse backgrounds who did not have pathologic variants in breast cancer susceptibility genes, and included 56 ancestry-informative variants with 93 BC-associated variants. They combined the 149-SNP polygenic risk score with the TC risk model to create a combined risk score that was developed in a cohort of 145,786 women who were unaffected by breast cancer, following a fixed-stratified model to avoid double counting between confounded factors.
Of the women included in the cohort, 69.1% were of European descent, while 10.2% were Hispanic, 10.0% Black/African, 1.9% Asian, and 8.8% all other groups.
An independent cohort of 68,803 women of a similar ethnic distribution was then used to evaluate the calibration of the combined risk score against the TC risk model alone, and to examine the relative contributions of the 149-SNP PRS, family history, and other clinical factors.
The results showed that, overall, the combined risk score was well calibrated across ancestries and percentiles of risks, and the absolute lifetime risks were similar to those derived from the TC risk model alone. The only exception was Hispanic carriers of a protective Amerindian SNP who had a lower score on the combined risk score than the TC model.
Using an ANOVA model, the team found that family history contributed 48% to the lifetime risk of breast cancer, while the 149-SNP PRS contributed 35% and other factors 17%. Family history was weakly, but significantly correlated with the 149-SNP PRS.
Determining the impact of adding the 149-SNP PRS to the TC risk model on risk classification, the team showed that across all ancestries, 17.3% of women were reclassified by the combined risk score versus the TC model alone, with 10.8% having their lifetime risk increased to high risk and 29.1% having their risk decreased by the combined model to low risk.
The largest reclassifications were seen for women of European descent, while the smallest were for Black/African women.
Study may have ‘cracked the code’
“What’s exciting is that I think we kind-of ‘cracked the code’ to some extent of how to do this across diseases for all ancestries,” Thomas P. Slavin, MD, chief medical officer at Myriad Genetics, said in an interview. “The adaptation for breast cancer risk stratification and the new panel [is] for breast cancer across all ancestries, but what we developed is something that could be used across diabetes, or colon cancer, or anything.”
He explained that they realized that “for each one of these little hot spots” in the SNPs, “that make one person different from another, you really need to find out where in the world that originated from. So, if you have genetic ancestry on an individual, you can say this spot in the genome has more of an African ancestry to it, or a European ancestry, and then you can weight it appropriately by the population.”
Dr. Slavin said that standard PRSs that simply add up SNPs are “pretty good” and “add a lot” to risk stratification, “but to fine-tune it a little bit and make the best risk model, you really do need to bring in clinical and family history factors.”
Montserrat García-Closas, MD, DrPH, deputy director of the cancer epidemiology and genetics for the National Cancer Institute, said the study is of interest, but “does not give information on how ancestry was considered in the models used to derive the scores.” She also cautioned that the method used in the study to calibrate the model seems “to mean a comparison of scores, rather than comparing the observed and expected risk in prospective cohorts by ancestry groups. This would be a way to estimate bias in risk prediction by ancestry.”
Nevertheless, Dr. García-Closas said the degree of risk reclassification seen with the combined risk score is as expected and pointed to recent work by her and her colleagues in which they tested an integrated model incorporating classical risk factors and a 313-variant PRS to predict breast-cancer risk and achieved similar results.
Several study authors disclosed ties with Myriad Genetics, as well as AstraZeneca, Bristol Myers Squibb, Clovis Oncology, Helix BioPharma, Konica Minolta, Ambry Genetics, Invitae, Stryker, GAIL, Phenogen Sciences, Novartis, Pfizer, CancerIQ, Tempus, 54gene, Color Genetics, Roche/Genentech, ImpediMed, Prelude Therapeutics, BD, Agendia, Targeted Medical Education, Cerebrotech Medical Systems, Integra LifeSciences, Puma Biotechnology, GeneDX/BioReference, Change Health Care, Research to Practice, Clinical Care Options, Physician Education Resource, and Daiichi Sankyo.
The headline for this article was updated on 1/6/22.
Elisha Hughes, PhD, director of research biostatistics at Myriad Genetics (which funded the study), and colleagues combined a risk model containing 149 single-nucleotide polymorphisms (SNPs), of which just over one-third were related to genetic ancestry, with the Tyrer-Cuzick (TC) breast cancer risk model.
The resulting combined risk score, which was developed in a cohort of over 145,000 women and validated in another group of almost 69,000 women, was not only well calibrated, but also able to reclassify just over 17% of women into a different risk group versus the clinical model.
The research (abstract P2-11-21) was presented at the San Antonio Breast Cancer Symposium on Dec. 8.
“This is the first breast cancer risk model based on a polygenic score, the 149-SNP PRS, that incorporates genetically determined ancestral composition and is validated for diverse ancestries,” the team reported.
The combined model substantially improved risk stratification over TC alone and may “lead to enhanced breast cancer risk reduction strategies, such as increased surveillance and use of preventive medications,” the researchers reported.
Breast cancer has a substantial genetic component that can “inform risk prediction and personalized preventive measures.” However, polygenic risk scores are largely derived from studies of women of European descent and tend to have poor performance in non-European ancestries.
Combined score substantially improved risk stratification over TC alone
The research team developed a polygenic risk score based on 149 SNPs for women of diverse backgrounds who did not have pathologic variants in breast cancer susceptibility genes, and included 56 ancestry-informative variants with 93 BC-associated variants. They combined the 149-SNP polygenic risk score with the TC risk model to create a combined risk score that was developed in a cohort of 145,786 women who were unaffected by breast cancer, following a fixed-stratified model to avoid double counting between confounded factors.
Of the women included in the cohort, 69.1% were of European descent, while 10.2% were Hispanic, 10.0% Black/African, 1.9% Asian, and 8.8% all other groups.
An independent cohort of 68,803 women of a similar ethnic distribution was then used to evaluate the calibration of the combined risk score against the TC risk model alone, and to examine the relative contributions of the 149-SNP PRS, family history, and other clinical factors.
The results showed that, overall, the combined risk score was well calibrated across ancestries and percentiles of risks, and the absolute lifetime risks were similar to those derived from the TC risk model alone. The only exception was Hispanic carriers of a protective Amerindian SNP who had a lower score on the combined risk score than the TC model.
Using an ANOVA model, the team found that family history contributed 48% to the lifetime risk of breast cancer, while the 149-SNP PRS contributed 35% and other factors 17%. Family history was weakly, but significantly correlated with the 149-SNP PRS.
Determining the impact of adding the 149-SNP PRS to the TC risk model on risk classification, the team showed that across all ancestries, 17.3% of women were reclassified by the combined risk score versus the TC model alone, with 10.8% having their lifetime risk increased to high risk and 29.1% having their risk decreased by the combined model to low risk.
The largest reclassifications were seen for women of European descent, while the smallest were for Black/African women.
Study may have ‘cracked the code’
“What’s exciting is that I think we kind-of ‘cracked the code’ to some extent of how to do this across diseases for all ancestries,” Thomas P. Slavin, MD, chief medical officer at Myriad Genetics, said in an interview. “The adaptation for breast cancer risk stratification and the new panel [is] for breast cancer across all ancestries, but what we developed is something that could be used across diabetes, or colon cancer, or anything.”
He explained that they realized that “for each one of these little hot spots” in the SNPs, “that make one person different from another, you really need to find out where in the world that originated from. So, if you have genetic ancestry on an individual, you can say this spot in the genome has more of an African ancestry to it, or a European ancestry, and then you can weight it appropriately by the population.”
Dr. Slavin said that standard PRSs that simply add up SNPs are “pretty good” and “add a lot” to risk stratification, “but to fine-tune it a little bit and make the best risk model, you really do need to bring in clinical and family history factors.”
Montserrat García-Closas, MD, DrPH, deputy director of the cancer epidemiology and genetics for the National Cancer Institute, said the study is of interest, but “does not give information on how ancestry was considered in the models used to derive the scores.” She also cautioned that the method used in the study to calibrate the model seems “to mean a comparison of scores, rather than comparing the observed and expected risk in prospective cohorts by ancestry groups. This would be a way to estimate bias in risk prediction by ancestry.”
Nevertheless, Dr. García-Closas said the degree of risk reclassification seen with the combined risk score is as expected and pointed to recent work by her and her colleagues in which they tested an integrated model incorporating classical risk factors and a 313-variant PRS to predict breast-cancer risk and achieved similar results.
Several study authors disclosed ties with Myriad Genetics, as well as AstraZeneca, Bristol Myers Squibb, Clovis Oncology, Helix BioPharma, Konica Minolta, Ambry Genetics, Invitae, Stryker, GAIL, Phenogen Sciences, Novartis, Pfizer, CancerIQ, Tempus, 54gene, Color Genetics, Roche/Genentech, ImpediMed, Prelude Therapeutics, BD, Agendia, Targeted Medical Education, Cerebrotech Medical Systems, Integra LifeSciences, Puma Biotechnology, GeneDX/BioReference, Change Health Care, Research to Practice, Clinical Care Options, Physician Education Resource, and Daiichi Sankyo.
The headline for this article was updated on 1/6/22.
FROM SABCS 2021