I have answered Yes to the second question because I think it will enhance the debate around this approach to personalizing medicine, not because this analysis is likely to have any direct clinical relevance. The particular example is unfortunate because type 2 diabetes is a risk state where the effect of lowering blood glucose is often unconnected to any long term benefit or harm, especially at this low target HbA1c. Identifying individuals who show a glucose response over a limited period is not the same as identifying individuals who would get long-term benefit or harm. For this, there is no surrogate measure which can be depended on.