Assessment of correlations among all favourable and unfavourable effects for rosiglitazone/metformin, and for metformin alone, in Type 2 diabetes.

Assessment of correlations among all favourable and unfavourable effects for rosiglitazone/metformin, and for metformin alone, in Type 2 diabetes.

None

None

Recent research on the quantitative modelling of the benefit-risk balance of drugs now makes it possible formally to incorporate uncertainty into the analysis. To date, most quantitative models use statistical summaries of the favourable and unfavourable effects of a drug compared to a placebo. However, many of these summaries do not use all the patient-level data. For example, a statistical mean gives an indication of an average patient's response to a drug and a confidence interval indicates uncertainty about the mean, but those summaries do not provide sufficient information to account fully for patient variability.

Attempts to deal with this loss of information in quantitative benefit-risk models has led to reconstructing patient-level data by adding assumptions to the statistical summaries, which enables the benefit-risk balance to be calculated on the full range of patient data by applying Monte-Carlo analysis. This method enables many different combinations of patient responses to be weighted by how often they occur. The results show that the additional information improves the certainty with which benefit-risk decisions can be made.

Unfortunately, the method looks at all possible combinations of effects without regard to the possibility that some combinations may never occur. Thus, the Monte Carlo studies may give misleading results. We simply don't know, because benefit-risk models to date have not been based on the full patient-level data; those studies focused on whether quantitative modelling was possible and desirable. Research now gives a positive 'yes' in answer to both.

The next step in improving the quantitative models, ensuring they are as accurate as possible given available data, is to examine the correlations among all the favourable and unfavourable effects in any given study. We expect that knowing the inter-correlations would cause the Monte Carlo analysis to give even lower probabilities to combinations of low-probability effects, thereby increasing the certainty that is inherent in patient-level data.

Thus, the aim of the research is to compare the benefit-risk balance of this combination of drugs when statistical summaries are used in the quantitative model to describe the effects as compared with the results when patient-level probability distributions describe the effects.

Under the direction of Professor Phillips, who is experienced in applying multi-criteria decision analysis (MCDA) to benefit-risk modelling of drugs, a post-graduate student on the MSc in Management Science at the LSE will carry out the research, helped by any members of staff who are experts in statistics and probabilistic simulation. The student's MSc project report will provide the necessary information for a publishable paper.

The value of this research is that it will be a first step in improving the explicit, quantitative modelling of the benefit-risk of drugs, to the advantage of all decision makers.

[{ "PostingID": 2444, "Title": "GSK-712753/007", "Description": "A randomized, double-blind trial to evaluate the efficacy and safety of fixed dose rosiglitazone/metformin combination therapy compared to both rosiglitazone and metformin monotherapies in drug naive type 2 diabetes mellitus subjects

Medicine: rosiglitazone/metformin, Condition: Diabetes Mellitus, Type 2, Phase: 3, Clinical Study ID: 712753/007, Sponsor: GSK" }]

The publication citation will be added after the research is published.