Predictive Modeling to Assess Pretest Probability of Transthyretin Gene Variants Based on Demographic Information

Background: Transthyretin amyloid cardiomyopathy (ATTR-CM) is a morbid condition, though recent advances in diagnosis and therapy stand to change its natural history. Patients' TTR genotype may guide family screening as more treatments and preventive strategies become available. An efficient, intuitive means of determining pretest genetic risk may better inform patients/clinicians when pursuing genetic testing. Methods: This is a cohort study of 767 consecutive patients diagnosed with ATTR-CM who underwent genetic testing. Classification and regression trees (CART) analysis created a decision tree assessing likelihood of carrying a pathologic TTR gene variant. Age, sex, and race were used as independent variables. Logistic regression was also performed to model probability of pathologic TTR genotype. The primary outcome was the decision tree's accuracy in 2 separate institutions' ATTR-CM registry. Results: In our study cohort, 208 patients (27.1%) had ATTRv. Race has served most efficiently as the root node followed by age and sex in a CART algorithm, and showed 88.2% accuracy (75.3% sensitivity, 93.9% specificity) in the validation cohort. The odds of having a TTR gene variant were greater in Black patients compared with non-Black patients (OR, 34.6 [95% CI, 20.5-58.3]; P<0.001). Non-Black patients with ATTR-CM aged 69 years and older had <4% risk of having a predisposing mutation. Conclusions: This CART algorithm incorporating age, sex, and race was able to determine which patients with ATTR-CM have pathogenic TTR mutations with high specificity. Non-Black patients diagnosed at age 69 years or older with ATTR-CM have a low likelihood to have ATTRv.