Three common programs are used in conducting phase II clinival trials, and each will be presented in its own page. This page presents an introduction and summary only.
When a potentially useful new drug is identified in the laboratory, its introduction to clinical use in the human involves four distinctive phases, as follows.
- Phase I : This is when the drug is given in graduated doses to human subjects, and its efficacies and toxicity studied. The
main aims of Phase I study is to identify those drugs which are therapeutically effective at a dosage that is acceptable
in terms of toxicity and side effects.
- Phase II : This is to test a drug found to be therapeutically useful from Phase I, to compare its efficacy against existing
outcomes, with or without current standard treatment. The main aims of phase II study is to identify those drugs that
potentially can produce a better outcome than those existing currently
- Phase III : This is a major controlled trial comparing the new drug against the current standard treatment, to produce
evidence that it is better and should replace the current treatment
- Phase IV : also called post-marketing trial, is a long term study to identify uncommon and unexpected adverse effects
of the new treatment.
This page will discuss phase II studies only.
Phase II studies or trials are essentially preliminary studies, to filter out and reject a large number of therapeutically
active drugs that are no better than existing treatments, and to accept the few that are potentially better for further studies.
The main focus is therefore to make as good a decision as possible with as small a sample size as possible, but postpone the
"Proof" to the more expensive phase III controlled trials.
In phase II studies, therefore, the terms accept and reject refer to decisions whether to take the drug to phase III studies, and not referring to rejection of the null hypothesis
StatsToDo offers 3 common procedures for phase II trials. Detailed descriptions and how to use them are
discussed individually in their respective pages, but they are briefly compared in this page.
The Fleming's procedure
The procedure is presented in
SSizFleming.php. It is an adaptation of a quality control procedure commonly used in industry. The user defines two levels of successes, the existing one, and the expected improved level the new treatment should achieve. The procedure then calculate the sample size required to test the new treatment. After the required number of cases are studied, the proportion of success is compared with that required, to determine whether the expected improvement has been met, thus whether to accept the new drug for further study and trial (phase III).
The Gehan's procedure
The procedure is presented in
SSizGehan.php. It minimizes the sample size required for a decision, by separating the decision to reject or accept the new treatment in two stages.
As most drugs entering a phase II study is expected to fail and be rejected, the first stage is designed to reject those with success rates lower than the acceptable level, with as few cases as possible. The second stage is designed to ensure that the acceptable level of success is achieved before the drug is accepted for further testing and trial (phase III).
The Simon's procedure
This is described in
SSizSimon.php. It is similar to Gehan's procedure conceptually, but includes an ability to truncate the study at an earlier stage if the results are obvious. This procedure is therefore the most cost effective one in terms of small sample size.
Comparison of the 3 procedures
The following example can be used to compare the three procedures.
We have an aggressive cancer where the one year survival rate is 10% (p0=0.1). A new chemotherapeutic drug is up for a phase II trial, and we decide that if the 1 year survival rate can be improved to 40% (p1=0.4), then a full scale phase III controlled trial would be worthwhile (acceptance). Otherwise the new drug is not worth pursuing and should be abandoned (rejection).
For all 3 models, we will use the probability of Type I Error (α) of 0.05, and a power (1-β) of 0.8, meaning
the probability of wrongly accepting the treatment is 5%, and the probability of wrongly rejecting the treatment is 20%.
- Fleming's procedure. We require 10 cases. At the end of which, if 4 or more survived more than a year, we will accept
the new treatment for phase III trials. Otherwise we will reject the new treatment
- Gehan's procedure. We require 4 cases in stage 1.
- If no one survived more than a year, we terminate the study and reject the new treatment
- If 1 case survived more than a year, we require 96 cases in stage 2 and a total of 40 survivals for more than a year
(39 + 1 from stage 1) to accept the treatment for phase III study. Short of this the treatment is rejected.
- If 2 cases survived more than a year, we require 70 cases in stage 2 and a total of 30 survivals for more than a year
(28 + 2 from stage 1) to accept the treatment for phase III study. Short of this the treatment is rejected.
- If 3 cases survived more than a year, we require 22 cases in stage 2 and a total of 11 survivals for more than a year
(8 + 3 from stage 1) to accept the treatment for phase III study. Short of this the treatment is rejected.
- If all 4 survived more than a year, we terminate the study and accept the treatment for for phase III study
- Simon's procedure. We will use the optimal model as we are more interested in rejecting those with results below expectation.
We require 4 cases in stage 1.
- If no one survive more than a year at the end of 4 cases, we terminate the study and reject the new treatment
- With the very first case that survive more than a year, we enter phase 2, with a total sample size of 15 (14 +
the success in stage 1)
- If at any time we reached 4 survivals (3 more on top of the one from stage 1) for more than a year, we terminate
the study and accept the treatment for phase III study
- If, at the end of 15 cases, there are less than 4 that survived more than a year, we reject the treatment from
further studies
From these descriptions, we can compare the 3 models
- Fleming's procedure is the simplest to follow. A decision to accept the treatment can be made anytime with 4 successes,
but a decision to reject the treatment can only be made after all 10 cases are observed.
- Gehan's procedure potentially allows the decision to be made at the end of the first stage, if all 4 cases are
successes or failures. If the results are less obvious however, the combined sample size from the two stages is very much
greater than Fleming's, both in accepting and rejecting the treatment.
- Simon's procedure is the most efficient of all. A decision can be made to reject if no success in the first stage after 4 cases, and to accept the treatment any time after 4 successes are encountered. Otherwise, it requires a maximum of 15 cases to
reject the treatment, slightly more than that used by Fleming's.
References
Fleming's Procedure
- Machin D, Campbell M, Fayers, P, Pinol A (1997) Sample Size Tables for
Clinical Studies. Second Ed. Blackwell Science IBSN 0-86542-870-0 p. 254-255
Gehan's Procedure
- Gehan E A (1961) The determination of the number of patients required in a preliminary and a follow-up trial of a
new chemotherapeutic agent Journal of Chronic Diseases. 13:4, Pages 346-353
<
- Machin D, Campbell M, Fayers, P, Pinol A (1997) Sample Size Tables for
Clinical Studies. Second Ed. Blackwell Science IBSN 0-86542-870-0 p. 255-256
Simon's Procedure
- Simon R (1989) Optimal two-stage designs for phase II clinical trials.
Control Clin Trials 10:1-10
- Machin D, Campbell M, Fayers, P, Pinol A (1997) Sample Size Tables for
Clinical Studies. Second Ed. Blackwell Science IBSN 0-86542-870-0 p. 256-257