Therasse identifies several pitfalls in Phase I studies which make the results difficult to apply:

• vague definition of MTD and DLT for Phase II investigation

• use of an objective tumor response as an endpoint although not relevant in Phase I protocols

• selecting an inappropriate dose for Phase II investigation

Phase I studies might be improved by (1) better preclinical models for selecting the starting dose, (2) new methods to expose fewer patients to the experimental drug, and (3) random assignment of patients in the dose titration.

Currently, there are at least 10 experimental designs available for conducting Phase II studies. Each design has a specific endpoint and should be utilized under specific conditions. Two designs are used regularly: the Gehan method for early Phase II (J Chron Dis 1961;13:346-353) and that of Simon (Controlled Clinical Trials 1989;10:1-10) and Fleming (Biometrics 1982;38: 143-151) for late Phase II.

In Therasse's opinion, common pitfalls in Phase II include the use of:

• response rate (RR) as a surrogate for therapeutic benefit

• feasibility studies to evaluate therapeutic benefit

• vague and nonstandard definitions of tumor response criteria for evaluation

Future studies could avoid investigator bias by randomly assigning patients to test or control treatment.

The study protocol requires patient randomization to assigned treatment(s) and balanced groups in multicenter evaluations. Special attention to uncontrolled variables, such as time, is necessary to assure the validity of the observed therapeutic effect. There are basically three protocol designs for Phase III studies: (1) parallel group (A vs B), (2) crossover (A followed by B and B by A), and (3) factorial using two or more treatments given alone or combined.

Endpoints in a Phase III trial often include:

• time to tumor progression

• duration of survival

• symptom control

• quality of life

• response rate, complete or partial

Therasse recommends measuring the time of an event from the date of randomization to treatment.

Common pitfalls in Phase III trials merit consideration:

• inadequate sample size

• too many or vague endpoints

• data torture by numerous subgroup statistical analyses

Phase III intergroup trials have special requirements. These include (1) setting a common study objective, (2) agreeing on a common methodology, and (3) having an independent group in charge of quality control and data management.

Therasse envisions future Phase III protocols incorporating interim analyses and early stopping rules. An independent data monitoring committee is needed for trials of more than 500 patients and recruitment periods over 4 years in duration.