At the population level we are interested in the number of spared events or cured patients by the treatment over a given time-period， i.e. the Public Health Impact of the treatment. We should draw this value from the available data， i.e. those collected by clinical trials.

How do we measure efficacy of a new treatment? Our estimate is a quantity of efficacy expressed by indicator， either an absolute benefit， a relative risk or an odds ratio. This quantity is measured by randomised clinical trials. The indicators are all time-cumulative， which means that their value is time-dependent for a given treatment， a given disease and a given therapeutic objective: they are valued at a given time.

Where does our knowledge come from?  As said above， it comes from the randomised clinical trials， and more precisely from the studied population. All the information we got on new therapy efficacy relies on what has been observed on the studied population.

Thus the studied population is the key component of the process. However， one should realise that there is a problem with the studied population: it is representative of no other population but itself. Further， as Heraclite stated it nicely， “the water of the river is never the same at the same place”. Thus it is irrelevant to used its characteristics to describe the therapy target population. Hence we should manipulate the available data.

How can we manipulate the data? If we can predict for a given patient or a group the value of Rc -Rt， which is the benefit we are interested in and the efficacy value that matters for the patients， and also for the community， we can make. Rt is the risk of event (the event we want to prevent， i.e. the therapeutic objective) in treated subjects and Rc is the same risk of event if the subjects were untreated.

Actually， it is possible to know Rc - Rt. We can predict Rc (e.g. the Framingham risk score). And we can estimate the relation between this risk and Rt， thank to the effect model theory. There are empirical evidence as well as theoretical considerations that show the reality of the effect model. The effect model stands for a therapy， a therapeutic objective and a disease. Empirical evidence came from deep exploration of a series of well documented examples， e.g. ACEI and CV mortality in heart failure patients， betabloking agents and total mortality in post MI patients， class Ic antiarrhythmics and sudden death in post MI patients， antihypertensive therapies， aspirin in the prevention of stroke and other cardiovascular events，…

Theoretical considerations rely on simulation of the efficacy of therapies based on a numerical model of the mechanism of action of a pharmacophore in the body. The model accounts for all the steps， i.e. absorption， distribution and metabolism， drug-receptor interaction， post-drug-receptor transduction， homeostatic retroaction， alteration of the disease process leading to a change in the chance of the clinical event to occur.

Thus the relation between Rt and Rc exists. It hows that the efficacy， hence the public health impact， depends on the population since the patients characteristics determine Rc. With a precise enough knowledge on the effect model of a therapy it becomes possible to predict the Public Health Impact.

The ingredients for such computation are: the clinical trial data， including the description of the studied population; epidemiological data from the population of interest; the therapy effect model; the threshold “s” which depends on either the rate and severity of side effects or the resources the community accepts to make available for treating the corresponding disease. However， attention should be paid to the time dependence.

In conclusion， it is possible to derive from the studied population data a prediction of the public health impact of the new therapy. This would be a great help for the clearance decision.  Also， the derivation has the same basis as the transposability issue. Transposability is the answer to the following question: how to transpose results obtained in a Caucasian studied population in oder to treat an Asian population?