The evidence for tafamidis to treat ATTR-CM comes from a multicentre, randomised, double-blind, phase III study (ATTR-ACT) which evaluated the efficacy and safety of the salt formulation, tafamidis meglumine, versus placebo in 441 patients with ATTR-CM. Tafamidis meglumine (pooled dose groups) significantly reduced the risk of all-cause mortality and cardiovascular-related hospitalisation compared with placebo over the 30-month study period. There were also significantly less declines in 6MWT and KCCQ-OS score in tafamidis meglumine treated patients compared with placebo-treated patients. Other secondary outcomes also favoured tafamidis meglumine over placebo. The submitting company presented a cost-utility analysis evaluating the use of tafamidis within its full licensed indication. The analysis compared tafamidis with best supportive care, which comprised a variety of therapeutic classes including diuretics, anticoagulants, angiotensinconverting enzyme (ACE) and renin-angiotensin-aldosterone system (RAAS) inhibitors and beta-blockers. Clinical experts stated that best supportive care represents the most appropriate comparator. Key weaknesses: • The Committee noted the lack of subgroup analysis in the economic model to explore the impact on the cost-effectiveness separately for wild-type and hereditary disease patients. While it was acknowledged this analysis would likely introduce additional uncertainty through the use of data from post-hoc subgroup analyses, it would have been helpful to explore this given the potential for the outcomes and clinical management of these patients to differ. • A stopping rule has been applied which assumes all tafamidis-treated patients will discontinue treatment at NYHA Class IV (as a surrogate of end-stage heart failure). This rule was not applied within the clinical study nor is it stipulated within the marketing authorisation. Given the severity of the condition, and the observation that some patients reaching the NYHA class IV state subsequently transitioned to earlier states, it is unclear whether discontinuation of tafamidis would be appropriate in patients progressing to NYHA class IV. It was also highlighted by clinical experts that this could be very hard to implement in clinical practice as the treatment is well tolerated and patients can fluctuate between health states in a short amount of time. A scenario analysis excluding the stopping rule resulted in an increase from the base case. • The approach to survival extrapolation is highly uncertain, and the models selected in the company base case appear overly optimistic. The use of a log-normal distribution for tafamidis and a generalised gamma for BSC, does not appear justified based on the assessment of statistical goodness-of-fit and may rely too heavily on the heavily censored latter stages of the observed Kaplan-Meier data. Numerous alternative distributions appear equally plausible and fall within the 95% confidence intervals of the observed data. Statistical advice received by the SMC stated that the approach relied upon many assumptions without accounting for the combined error in the modelling, and that numerous distributions could be equally plausible. The Weibull distribution was deemed to be equally well-fitting by NDC, while still enabling the independent extrapolation of survival for each treatment arm. This was felt to be a plausible alternative approach to the base case estimate and resulted in significantly higher ICERs. • Similarly, the approach to estimating treatment duration is uncertain and is likely to underestimate the plausible treatment duration with tafamidis. In fact, the use of the exponential function (best fitting by Bayesian Information Criterion [BIC]) results in the shortest duration of all approaches tested, while the gompertz model (best fitting by Akaike Information Criterion and second-best by BIC) results in the longest duration. The use of the lognormal function (the second best-fitting and second-longest estimates) highlights a significant sensitivity to this approach. It is also worth noting that, in contrast to the estimation of overall survival, treatment duration has not been risk-adjusted by NYHA class. This may add an additional degree of bias. • Following treatment discontinuation, the clinical benefit of tafamidis is assumed to continue across the time horizon. As a result, patients may continue to move into less severe health states during the extrapolation phase, and achieve the same degree of survival benefits, despite no longer accruing the costs of active treatment. This assumption seems highly implausible; it would be more appropriate for patients to wane towards best supportive care probabilities from the point of discontinuation. This alternative approach results in higher ICERs, due to a lower proportion of patients moving into earlier health states over time. • The submitting company has adjusted the medicine acquisition costs on the relative dose intensity and assumed that the remaining are not lost to wastage. The submitting company was asked to provide scenario analysis with a relative dose intensity of 100% to reflect the potential for wastage, resulting in a moderate increase in the ICER. • The use of a single extrapolation matrix derived from the total 30 month study observation period assumes no change in the effectiveness of tafamidis over time. In addition, a gradual plateauing in transitions between classes was observed in the latter stages of the study period. The use of this ongoing treatment effect, derived based on the early study periods where more rapid improvements in NYHA class were observed, is likely to overestimate the effectiveness of tafamidis. A scenario using transition matrices based on observed data at month 18, 24 and 30 was deemed more appropriate. • Utilities were not adjusted for increasing age and may therefore bias the result of the economic model in favour of tafamidis. A revised scenario was provided which applied age adjustment in the extrapolation period, resulting in a moderate increase in the ICER. The provided analysis is believed to be an underestimate of the full effect of age-adjusted utilities, as the estimates used by the company for NYHA class I and II result in higher utilities than a healthy matched population. • The assumption of a differential treatment effect on utilities, as well as assuming that patients will benefit from movement to earlier NYHA classes, creates the possibility of double counting the utility benefits of tafamidis. In addition, the treatment specific utilities do not reflect utility for patients who have discontinued tafamidis. The use of treatment independent calculation of health effects would seem preferable to avoid this potential issue and result in an increased ICER. • A number of alternative approaches described above were considered by NDC to be more methodologically appropriate and equally valid to the company’s preferred approach. When combined, these resulted in a significant increase in the ICER. The ICER increased further still when the stopping rule was removed. The Committee also considered the benefits of tafamidis in the context of the SMC decision modifiers that can be applied when encountering high cost-effectiveness ratios and agreed that a number of the criteria were satisfied: a substantial improvement in life expectancy in the patient population targeted in the submission; a substantial improvement in quality of life; and the absence of other treatments of proven benefit. In addition, as tafamidis is an orphan medicine, SMC can accept greater uncertainty in the economic case. After considering all the available evidence and the output from the PACE process, and after application of the appropriate SMC modifiers, the Committee was unable to accept tafamidis for use in NHSScotland.