Increased mortality rate in patients with Takayasu arteritis is largely driven by early disease activity and damage: analysis of two cohorts

OBJECTIVE: To evaluate mortality risk and pre-treatment predictors of mortality in Takayasu arteritis (TAK). METHODS: From two cohorts, outcomes at last visit (survival/mortality) were recorded. Standardized mortality ratios (SMR) were computed (indirect standardization to population death rates from India and Turkiye). Cox proportional hazards regression was used to estimate predictors of mortality [hazard ratios (HR) with 95%CI]. The final multivariable-adjusted models included significant predictors of mortality from initial analyses based on demographic features, disease activity, damage (VDI), and clinical features or arterial involvement. Random forest (RF) and gradient boosted machines (GBM) machine learning tree models were used to understand the relative contributions of covariates in the multivariable-adjusted regression models to mortality risk. RESULTS: Among 529 patients with TAK [India, n = 310, Turkiye, n = 219, mean age at onset 29.12 years, 416 females], 54 deaths were recorded (29 out-of-hospital, 29/54 from cardiovascular disease, 10/54 from infections). Survival was 97.73%, 96.55%, 90.24% and 82.37% at 1, 2, 5, and 10 years, respectively. TAK was associated with increased mortality rate [SMR for India 35.88 (95%CI 24.41-50.94) and Turkiye 27.61 (95%CI 17.53-41.41)]. Higher baseline disease activity by ITAS2010 (HR 1.06), damage (HR 1.30), impaired renal function (HR 2.21), or abdominal aorta involvement (HR 2.03) significantly predicted mortality risk with good model discrimination accuracy (Harrell's C-statistic 0.772). Both RF and GBM models identified baseline activity and damage scores as major drivers of mortality. New-onset vascular events were largely driven by baseline ITAS2010. CONCLUSION: Increased mortality rate in TAK is largely driven by disease activity and early damage.