An introduction to statistical methods for health technology assessment: a review
White S J, Ashby D, Brown P J
Record ID 32000008233
The aims of this report were to: - document recommended practice in relevant and related areas - document current practice critically - map current methodological research - identify areas relevant to health technology assessment where statistical methodology is either inadequate or not being employed to full advantage, and - identify suitable areas for further research.
Authors' results and conclusions: Statistical training The review of textbooks from MSc courses showed that students are being offered courses in statistical theory and methods, design of experiments, linear models and generalised linear models, survival analysis, repeated measures, spatial statistics, multivariate methods, multilevel models, distribution-free statistics, Bayesian inference and methods, measurement errors, computational statistics, clinical trials and epidemiology. This represents a wider range than any one person can learn in a year. Much is relevant to health technology assessment, but the links are not yet very explicit, and there are no directly relevant textbooks recommended. Statistical guidelines Statistical guidelines have been developed in areas relevant to health technology assessment, in particular drug regulation, and in systematic reviews of randomised trials, through the Cochrane Collaboration. The linchpin technology in both of these areas is the randomised controlled trial. However, they mostly emphasise principles and ways of working rather than detail, with only meta-analysis covered in depth. Publications A review of the papers potentially relevant to health technology assessment, published in statistical journals in 199495 yielded 505 papers. These were predominantly about new methodology rather than discussion or review papers, mainly used classical rather than Bayesian approaches, and largely used re-analysed or simulated data, rather than primary analyses. Most related to preclinical or clinical trials rather than other kinds of studies. Study designs Much of the statistical literature on study designs that relate to health technology assessment comes from clinical trials; there are relatively few publications that cover the more complex experimental designs, meta-analysis or studies of drug safety. Within the medical literature, of the more complex experimental designs, (bio-)equivalence and crossover studies can be identified but not other designs in large numbers. Methods for analysis Of the statistical literature relevant to health technology assessment on analysis of follow-up studies, both survival and longitudinal data feature regularly, and repeated measures (non-longitudinal) occur less often. Within the medical literature, survival analysis is extremely common, particularly proportional hazards and Cox regression. Much of this work is in the context of cancer and heart disease. Identifying longitudinal studies in the medical literature is straightforward, and they cover a range of conditions, but identifying the use of longitudinal methods of analysis is much harder. Needs of the HTA programme In health technology assessment the question 'Does the technology work?' is most easily answered using standard statistical methods. 'For whom?' raises statistical questions of subgroup analysis and interactions, and wider questions of generalisability. 'At what cost?' raises questions of identification and measurement of costs, with appropriate handling of associated uncertainty. 'How does it compare with alternatives?' brings a need for more formal decision analysis, and revisiting work on complex experimental design.
Authors' recomendations: The NHS R&D programme could consider training strategies for continuing professional development of statisticians, for example by allocating a fund to allow attendance of relevant courses. The NHS R&D programme could consider commissioning induction courses for statisticians working in health technology assessment. The purpose of such courses would partly be to give an introduction to health technology assessment and associated disciplines, and partly to focus on reinforcing statistical methods particularly pertinent to health technology assessment. Researchers in health technology assessment could avail themselves of existing guidelines, specifically those in drug regulation, and, if involved in meta-analysis, those of the Cochrane Collaboration. The NHS R&D programme could sponsor workshops to bring together statisticians and others who have been working in health technology assessment to identify and develop future statistical issues in health technology assessment. Case studies are needed on decision making under uncertainty using established Bayesian methodology to integrate health outcomes with wider costs. Established statistical methodology on design of experiments is potentially relevant to complex questions in health technology assessment. The development of specimen protocols explicitly using such methodology could be commissioned.
Authors' methods: Review
Project Status: Completed
URL for project: http://www.hta.ac.uk/938
Year Published: 2000
English language abstract: An English language summary is available
Publication Type: Not Assigned
Country: England, United Kingdom
- Technology Assessment, Biomedical
Organisation Name: NIHR Health Technology Assessment programme
Contact Address: NIHR Journals Library, National Institute for Health and Care Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK
Contact Name: email@example.com
Contact Email: firstname.lastname@example.org
Copyright: 2009 Queen's Printer and Controller of HMSO
This is a bibliographic record of a published health technology assessment from a member of INAHTA or other HTA producer. No evaluation of the quality of this assessment has been made for the HTA database.