A little learning: reflections on 10 years of NICE technology appraisals.

Within just a few days of joining NICE, I realised the scale of the task. I thought that grounding in biological science, pharmaceutical R+D and the actual production of submissions to NICE had equipped me with relevant knowledge and skills. How naïve; this grounding was merely the starting point of what was needed. The realisation crystallised for me over initial briefings on the ongoing appraisal of drugs for multiple sclerosis. I had joined NICE at the point that it was concluded that the health economic models produced for the appraisal, from pharmaceutical companies and from the academic assessment team, did not help the appraisal committee decision makers reach conclusions on the drugs with sufficient certainty. The decision: NICE would commission the development of another health economic model. The commission went to the School of Health and Related Research in Sheffield UK (ScHARR). The first major task after my appointment was to work with the ScHARR health economists to produce an economic model that would provide robust estimates of the cost-effectiveness of the multiple sclerosis (MS) drugs. That could not be too difficult. This was a world-leading research group who had experience in the therapeutic area. I was wrong. I had forgotten that my new environment was not one where the only people scrutinising the science were a critical bunch of like-minded scientists. I was now in a world of transparent, inclusive and interactive health policy decision making. In its forming principles, NICE committed itself to working in a way that no other NHS health policy organisation before it had worked. The words transparent and inclusive are simple words. Implementing the principle of engagement that lies behind them is not simple at all. I was about to find out how difficult it could get. People wanted to know, in detail, what was wrong with the current models. These stakeholders had submitted the existing models and thought they had produced robust submissions, some with input of leading academic modellers. Why could ScHARR do it any better than they had? How could they interact with the research group? Could they submit data? Could they scrutinise this output? We provided a mechanism for