OBJECTIVES To evaluate the treatment effect of tafamidis on overall survival (OS) using data from the long-term extension (LTE) of ATTR-ACT, accounting for treatment switching by placebo patients in the LTE. During ATTR-ACT, a 30-month trial, patients with transthyretin amyloid cardiomyopathy receiving tafamidis had improved survival versus placebo (hazard ratio: 0.70 [95% CI 0.51, 0.96]). The hazards were divergent during the trial; therefore, this point estimate of treatment effect may underestimate survival benefit. The LTE provided tafamidis to placebo patients after 30 months in ATTR-ACT; results accounting for switching are needed to estimate treatment effect.

METHODS : Parametric survival models were fitted to the ATTR-ACT placebo arm to allow extrapolated prediction of OS beyond treatment switching; selection was based upon information criteria. The ratios of cumulative hazards from these models and the non-parametric estimate of cumulative hazard of the ATTR-ACT LTE tafamidis 80mg arm yield average hazard ratios to the evaluation time. These estimates will be compared to the treatment effect within ATTR-ACT.

RESULTS : Potential models of placebo OS were the Weibull, Gompertz, Gamma and Generalised Gamma, which predicted 60-month survival of 17.6% (95% CI 7.9%, 27.7%); 2.3% (0.0%, 16.3%); 23.0% (14.3%, 31.2%); 8.3% (0.0%, 56.5%) respectively. These models demonstrated monotonically increasing hazard for placebo, suggesting the observed tafamidis effect in ATTR-ACT may be an underestimate.

CONCLUSIONS : Extrapolative models of placebo OS, when compared to observed on treatment data, allow for further learnings of the long-term treatment effect of tafamidis, likely correcting the underestimation inherent in ITT analysis due to switch from placebo to active treatment.