Guest post by Thomas Zwingers, Head of Statistical Consultancy
An interesting overview on the average costs of clinical trial in different phase of drug development was given by Sertkaya et.al. in a recent article in Clinical trials 2016, Vol13(2). According to their survey, the average cost of a Phase I trial ranges from $1.4 million USD in the therapeutic area of pain to $6.6 million USD in the area of immunomodulation. For Phase II trials they report an average of $7 million USD in cardiovascular research to $19.6 million USD in hematology, and for Phase III trials the average cost ranges from $11.5 million USD in dermatology to $52.9 million USD in pain research.
They point out that for a Phase II or Phase III trial the percentage of the total costs for “per-patient costs” is approx. 44%, the “per-site costs” is 50% and the “per-study costs” is 6%.
It is quite obvious that the number of patients in a trial is the most important cost factor. The sample size also influences the number of sites to be included in a trial. In a traditional drug development process, the costs of each phase will accumulate into a total cost.
Also worth considering is the time of the development process. In a traditional approach, each phase is planned separately after the results of the previous phase are available. This individual set up takes considerable time and prolongs the time for the drug to reach market.
Modern biometric methodology provides tools capable of reducing the total number of patients needed and shortening the total duration of the development process. These methods are summarized under the label “Seamless Designs”.
The main idea of these designs is to use the data of one development phase also in the next phase. This is displayed in the following figure.
The savings in patient numbers and time is obvious.
The “seamless” principle is already very often used when combining Phase I, looking for the “maximum tolerated dose”, and a consecutive “Proof of Principle” study.
But why are these designs not always applied to combine Phase II and Phase III in the development processes?
The main limitation is related to the primary endpoint of these development phases, efficacy and the fact that a decision has to be made, at a certain time point, to continue with the next phase.
In many indications efficacy is measured only after a long time, e.g. in cancer it is “survival time”, which might need years to evaluate or in COPD where “time to first exacerbation” is a common endpoint. By the time there is enough information on the endpoint, either events or number of patients who have been observed long enough to calculate the endpoint, usually the number of patients planned in the protocol is already recruited and no selection can be made or recruitment has to stop until a selection can be made. In this situation simulations on the recruitment rate and time to endpoint can help to decide whether a seamless design is feasible or not.
Another solution could be to use a surrogate endpoint for the selection, instead of the anticipated primary endpoint in the final analysis. The problem with this scenario is that only a small number of endpoints are validated as “surrogate endpoints”. The use of non-validated endpoints in the selection process increases the risk for the pharmaceutical company to make a wrong decision and consequently the power in the final analysis decreases.
Seamless study designs have the opportunity to decrease the number of patients needed and decrease the time of the drug development process. Whenever possible they should be applied in the planning of the drug development process, but thorough investigations and simulations should support the decision whether to use or not to use a seamless design.