Forget proof of concept – Michael Herschel


The pros and cons of a broad phase II testing are provided. Broad phase II means several hundred patients, and includes the option of a phase II nested into the phase III trial. The question of whether it is possible to make an informed “go or no-go” decision is examined as is the impact on costs, safety of study participants.

In this, the second of “Herschel’s Heresies”, I would like to burn some of the textbook wisdom that: clinical phase II should to be done in small, highly selected numbers of patients (phase IIa), leading to phase (IIb and) III, the real pivotal trials.

The textbooks tell us that phase I in healthy volunteers is followed by small well-controlled clinical trials in a few hundred patients at most, which serve to show that the drug works – at least against placebo. Phase II, beyond that, is the phase where surrogate endpoints are to be used to allow for small sample size.

We shall therefore look at the assumptions underlying this approach and the actual conduct of it, being aware that some companies are already seeking a cautious distance from this dogma.

The assumptions are few
1. If one selects a highly homogeneous population, the variability is small in general and for the primary endpoint in particular.
2. Small focused trials allow us to make a clear decision whether to progress into phase III trials.
3. Selected populations give a high likelihood of few adverse events occurring thereby not negatively influencing the risk-benefit profile at an early stage of development.
4. Ethics Committees are more likely to accept small trials in the face of lack of knowledge of a drug rather than large ones.
5. As many of phase II compounds do not make it into phase III, large phase II trials are a waste of time and money.

Assumption 1)
I have selected an overview by Collins on diuretics in pregnancy. The sample sizes in the trials ranged from 78 to 2706 patients. There is no correlation between the sample size and the variability. Small samples may be more homogeneous, but they need not be. Similar conclusions can be drawn from a meta-analysis where intranasal steroids were compared to oral antihistaminics.
In addition, the reduction of variability greatly reduces information that may be relevant for phase III. For instance, the relationship between initial blood pressure and the effect of drugs becomes clearer if a wide range is included at the start.

Assumption 2)
The decision, after phase II, is whether to progress into phase III or stop clinical development altogether. For the latter decision there is a high post-decisional regret, thus the likelihood to “kill” a “good” drug must be very low. Whether this is achieved today cannot be said, however, there is anecdotal evidence that several companies have revitalised “dead” drugs and made them winners. Known examples are BI’s meloxicam and Bayer’s nifedipine. On the other hand, few compounds are killed after phase II altogether, except if they show no superiority over placebo in a trial where an active drug is also part of the treatment arm. About half of the drugs that are stopped are terminated after phase II because of lack of efficacy, the remainder showed safety problems or formulation difficulties.
The decision to go into phase III carries the risk of not having a drug but just a compound that showed a “proof of concept”. When taking a sample of compounds that died at the end of phase III or were not marketed, the following reasons were given:
– Lack of efficacy
– Safety problems
– Lack of superiority over gold standard

It is obvious that such information is desirable before beginning phase III. However, currently the phase II rarely allows a reliable decision regarding phase III.

Assumption 3)
The dogma of the selected population that does reduce the risk of side effects has never been questioned. From clinical testing of abacavir, for instance it can be said that both the frequency and body system profile was similar in phase II and III. However, larger samples increase the likelihood that the observed incidence is “real”.

Assumption 4)
Ethics Committees do not wish to give carte blanche. Therefore, less restricted populations would be more acceptable if the information on such adverse event profiles, be it directly or via Data Safety and Monitoring Boards, is fast and comprehensive. Funnelling such information via Internet or other means so that decisions by Ethics Committees are facilitated are a prerequisite.

Assumption 5)
The data on clinical trial cost is not too reliable. A small calculation may underscore this point: if a phase III program costs 50 million Euro, and if 50% of such decisions could be made by the end of phase II, then a company bringing four compounds into phase III each year could save 100 million Euro per year. On the other hand, of the 10 compounds it brings into phase II, additional costs could come up by enlarging phase II, i.e. on the basis of total cost per patient of 10,000 Euro and 300 additional patients, this would add 3 million Euro per compound, and thus allow a net gain of 100 minus 30, i.e. 70 million Euro.
Having discussed the cons of a broad phase II, let me briefly summarise the pros:

1. Recruitment of patients will be easier.If the criteria of eligibility are simplified and exclusion criteria are scrapped, the inclusion of patients will be faster. The likelihood finding patients will be higher, however, the problem of including women has to be solved by providing sufficient toxicology data.

2. The type of patients will be closer to realityThis issue follows from the first. While this may disappoint stock exchange analysts who like “great” phase II data, it helps to make a more realistic drug profile planning for the subsequent phases.

3. Safety problems may be known earlierSome of the safety problems may come up much earlier, such as features that may differentiate the new compound from the marketed competitors.

A broad phase II may not be a panacea for all clinical development. When a drug is developed for a disease where no drug is available yet, a small phase II versus placebo may be all that is needed. But in all other cases, where a dose-response-relationship needs be established early and reliably, the current policy tends to lead to the usual phase IIb before phase III is started, and, therefore, to a loss of time and money. It seems that a compromise is the phase II that leads almost automatically into phase III which is sometimes is called “nested”.

But why, then, do we still teach newcomers that “phase II is done in small samples of patients” ?.

(C) 2008 – Alle Rechte vorbehalten

Schreibe einen Kommentar

Deine E-Mail-Adresse wird nicht veröffentlicht. Erforderliche Felder sind mit * markiert.