Table 3: Summary of main Phase III clinical trial designs.

DesignsAdvantagesDisadvantages

Historical control
(literature and existing databases or medical charts)
Allows ethical consideration.
Increase enrollment as patients are assured of receiving new therapy.
Shorter time and less cost.
Good for initial testing of new treatments, when disease diagnosis is clearly established, prognosis is well known, or disease is highly fatal.
Vulnerable to bias.
Disease rate and mortality rate have changed over time.
No comparison to control group data.
Literature controls particularly poor.
Tends to exaggerate the value of a new treatment.

Concurrent control, not randomizedEliminates time trends.
Data of comparable quality.
Randomization does not interfere with treatment selection.
Easier to select a group to receive the intervention and select the controls matching key characteristics.
Reduced cost, relative simplicity, investigator and participant acceptance.
Intervention and control groups may not be comparable because of selection bias and different treatment groups are not comparable.
Difficult to prove comparability because of the need for information on all important prognostic factors and matching several factors is impractical.
Uncertainty about unknown or unmeasured factors exists even for large studies.
Covariance analysis not adequate.

Randomized clinical trials (RCT)Considered to be “gold standard”.
Removes potential bias in group allocation.
Randomization and concurrent control produce comparable groups.
Guarantees the validity of statistical tests and valid comparison.
General use.
Subjects may not represent general patient population.
Increased sample size and cost.
Acceptability of randomization process.
Administrative complexity.

Sequential RCT designContinues to randomize subjects until null hypothesis is either rejected or “accepted.”
Good for acute response, paired subjects, and continuous testing.
Good for one-time dichotomous decisions such as regulatory approval, and so forth.
Multiple testing inflates type I error.
Inhibits adaptation due to the requirement of prespecifying all possible study outcomes.

Bayesian RCT designDynamic learning adaptive feature.
Incorporates external evidence.
Add new interventions and drop less effective ones without restarting trial.
Improves timeliness and clinical relevance of trial results.
Lowest sample size and cost.
May be criticized as too subjective, not well planned, or too complicated.