Polygenic Risk for Cardiometabolic and Cardiovascular Disease in a Multiethnic Cohort of Breast Cancer Survivors

Figures

• 

  Figure 1. Manhattan Plot of PGS With Incident CMD and CVD Risk in the European Population A Manhattan-style plot in which each point represents a test of the association between polygenic scores (PGS) and cardiometabolic disorder (CMD) or cardiovascular disease (CVD) events. The x-axis represents the 12 CMD (A) and CVD (B) events, and the y-axis is the −log10(P) value of the association test. The red horizontal line depicts the Bonferroni significance threshold (P ≤ 1.54 × 10−4). The size of the circles above each line is proportional to the HR of the association in the European population. Significant associations are labeled with the PGS trait tested. BMI = body mass index; Hb A1c = glycated hemoglobin; HF = heart failure; LDL = low-density lipoprotein; VTE = venous thromboembolism.

• 

  Figure 2. Liability R2 Value of the Best Performing PGS Across Ancestry Groups For 8 CMD and CVD outcomes which significant associations with PGS were identified, liability R2 was calculated and plotted as a bar plot with 95% CIs across ancestry groups. The numbers to the right of the bars show R2, and those that explained a significant proportion of the variance in the outcome (P < 0.05) are denoted with asterisks. In each group, the best performing PGS was chosen. (A) Outcomes with which multiple PGS were associated. (B) Outcomes with which only 1 PGS was associated. AFR = African; ALL = all ancestries; AMR = Amerindian; EAS = East Asian; EUR = European; other abbreviations as in Figure 1.

• 

  Figure 3. Predictive Performance of Multi-PGS in the European Population For the 5 CMD and CVD outcomes with which multiple PGS were significantly associated, a multi-PGS was calculated using PRSmix+. The predictive performance (liability R2) of the combined PRS and single constituent PGS are shown in bar plots along with 95% CIs. The liability R2 value is shown to the right of each bar. Abbreviations as in Figure 1.

• 

  Figure 4. Incremental Discrimination Performance for PGS Added to Base Models With Clinical Covariates in the European Population Areas under the receiver-operating characteristic curve (AUCs) for the top PGS from the base models with clinical covariates only (green) are plotted in bar plots with 95% CIs and compared with those from the models with the best performing single PGS and clinical covariates (blue) and the models with the combined PGS and clinical covariates (red) in the European (EUR) population. We show results for outcomes with higher performing PGS, which showed an R2 value >3% in the EUR population and for which a multi-PGS model was available. P values of the pairwise DeLong test are shown. Abbreviations as in Figure 1.

• 

  Figure 5. PGS Discrimination Performance by Cardiotoxic Breast Cancer Treatment The AUCs for the top-performing PGS with clinical covariates are plotted in bar plots with 95% CIs (red) and compared with those from the models with clinical covariates alone (blue) for individuals who received and did not receive cardiotoxic treatment separately. Only results with significantly different AUCs between treatment and no treatment (P < 0.05) are shown. Abbreviations as in Figure 1 and 4.

• 

  Central Illustration. Polygenic Scores and Cardiovascular and Cardiometabolic Diseases in Women With Breast Cancer Women with breast cancer are at increased risk for cardiovascular disease (CVD) and cardiometabolic disorder (CMD) because of cancer treatment–related factors. Polygenic scores (PGS) function to aggregate the nonzero effects of many genetic variants associated with a trait, but most CVD PGS are developed in non–breast cancer populations that do not face the treatment-related risk factors for CVD. In a large multiethnic patient population, we computed 27 PGS for CVD-related traits and examined their associations with 12 incident CMD and CVD events. In the plots, the red horizontal lines depict the Bonferroni significance threshold (P ≤ 1.54 × 10−4). The size of the circle above each line are proportional to the HR of the association in the EUR population. We found that PGS were strongly associated with risk for incident CMD and CVD in women with histories of breast cancer, but their prediction performance needs improvement to be clinically useful. BMI = body mass index; Hb A1c = glycated hemoglobin; HF = heart failure; LDL = low-density lipoprotein; VTE = venous thromboembolism.