If there is no improvement in a population health outcome after the introduction of a medicine we need to consider a number of possible explanations for failures to find an association [59].
First, a drug may not have been used by a large enough proportion of the patient population to produce a detectable improvement in the population health outcome. Estimates may be obtained from epidemiological modelling of the likely effects of medicine use on population health outcomes for (i) varying percentages of patient coverage and (ii) varying estimates of the amount by which the efficacy observed in clinical trials is reduced in routine clinical use.
Second, we need to consider whether our study has had sufficient statistical power to detect improvements in health outcomes of the size that we expect (given the probable efficacy of the medicine and its likely patient coverage). Statistical power will depend upon both the size of the expected impact on population health outcomes and the rate of increase in medicine use. It may be difficult to detect changes in population health outcomes if there is a slow incremental increase in medicine use. It may be easiest to detect an effect if there is a steep, substantial and sustained increase in medicine use for a condition for which no effective treatment previously existed.
Third, if a large proportion of eligible patients in the population are receiving the medicine we also need evidence that the drug is being prescribed to patients who will benefit from it and that these patients are taking the drug in the required dosages with the required frequency.
Fourth, we need to consider the possible impact of countervailing factors on population health outcomes. These may include worsening trends in risk factors that offset any population health benefits or the effects of other medications.
Answers to all your Biology Questions
