Warwick Evidence,Warwick Medical School,University of Warwick.
Int J Technol Assess Health Care. 2019 Jan;35(2):160-167. doi: 10.1017/S0266462319000175.
Before an intervention is publicly funded within the United Kingdom, the cost-effectiveness is assessed by the National Institute of Health and Care Excellence (NICE). The efficacy of an intervention across the patients' lifetime is often influential of the cost-effectiveness analyses, but is associated with large uncertainties. We reviewed committee documents containing company submissions and evidence review group (ERG) reports to establish the methods used when extrapolating survival data, whether these adhered to NICE Technical Support Document (TSD) 14, and how uncertainty was addressed.
A systematic search was completed on the NHS Evidence Search webpage limited to single technology appraisals of cancer interventions published in 2017, with information obtained from the NICE Web site.
Twenty-eight appraisals were identified, covering twenty-two interventions across eighteen diseases. Every economic model used parametric curves to model survival. All submissions used goodness-of-fit statistics and plausibility of extrapolations when selecting a parametric curve. Twenty-five submissions considered alternate parametric curves in scenario analyses. Six submissions reported including the parameters of the survival curves in the probabilistic sensitivity analysis. ERGs agreed with the company's choice of parametric curve in nine appraisals, and agreed with all major survival-related assumptions in two appraisals.
TSD 14 on survival extrapolation was followed in all appraisals. Despite this, the choice of parametric curve remains subjective. Recent developments in Bayesian approaches to extrapolation are not implemented. More precise guidance on the selection of curves and modelling of uncertainty may reduce subjectivity, accelerating the appraisal process.
在英国，某项干预措施获得公共资金之前，国家卫生与保健卓越研究所（NICE）会对其成本效益进行评估。干预措施对患者终生的疗效通常会影响成本效益分析，但这与较大的不确定性相关。我们查阅了包含公司提交材料和证据审查组（ERG）报告的委员会文件，以确定在推断生存数据时使用的方法，这些方法是否符合 NICE 技术支持文件（TSD）14，以及如何处理不确定性。
在 NHS Evidence Search 网页上进行了系统搜索，该搜索仅限于 2017 年发表的癌症干预措施的单一技术评估，信息来自 NICE 网站。
确定了 28 项评估，涵盖了 18 种疾病的 22 种干预措施。每个经济模型都使用参数曲线来对生存数据进行建模。所有提交的材料都使用拟合优度统计和推断的合理性来选择参数曲线。25 份提交的材料在情景分析中考虑了替代参数曲线。6 份提交的材料报告在概率敏感性分析中包含了生存曲线的参数。在 9 项评估中，ERG 同意公司对参数曲线的选择，在 2 项评估中同意所有与生存相关的主要假设。
所有评估都遵循了 TSD 14 关于生存推断的规定。尽管如此，参数曲线的选择仍然具有主观性。最近在推断方面贝叶斯方法的发展并未得到实施。关于曲线选择和不确定性建模的更精确指南可能会减少主观性，从而加速评估过程。
