issue {randomizeR}

Assessment criteria for clinical trials

Description

Summarizes the criteria for the assessment of randomization procedures.

Details

Randomization in clinical trials is supposed to control certain properties in clinical trials. In the randomizeR package, these properties are called issues. It is crucial to decide which of the issues is relevant in the present clinical trial, because a randomization procedure that manages well one issue might behave very badly for another. The issues include

• Selection bias can occur if future treatment allocations are predictable due to restricted randomization and unmasking of past treatment assignments. The influence of selection bias on the test decision is represented by the selBias class. The measure for the predictability of a randomization procedure is implemented in the corGuess class representing the expected number of correct guesses.

• Chronological bias can occur if a time trend is present in the data. Time trends occur due to learning curves, relaxed inclusion/ exclusion criteria or new co-medication. Chronological bias is represented by the chronBias class.

• Additive combination of chronological and selection bias may occur if a time trend and selection bias are present in the data. The combined bias is represented by the combineBias class.

• Balance is important in order to ensure proper power estimation properties of the treatments. However, a high degree of balance favors selection bias. Depending on the clinical context, a randomization procedure should be chosen that admits a suitable imbalance. Imbalance bias is represented by the imbal class. The power loss due to imbalance can be assessed directly via the setPower class

See Also

Representation of randomization procedures: randPar

Generation of randomization sequences: genSeq

Assessment of randomization sequences: assess

Comparison of randomization sequences: compare

Other issues: chronBias, combineBias(), corGuess, imbal, selBias, setPower()