Common Problems in Writing Statistical Plan of Clinical Trial Protocol Liying XU CCTER CUHK

The importance of statistical planning International Conference on Harmonization (ICH) E9 Guideline on Statistical Principles for Clinical Trials

Statistical quality Any high quality trial will include a detailed analysis plan as part of (or an appendix to) the protocol. Statistical quality –Professional competence –Professional responsibility

Pre-specification of the analysis Statistical section of the protocol should include all the principal features of the proposed confirmatory analysis of the primary variable(s) and the way in which anticipated analysis problems will be handled.

Factors Affecting Statistical Methods Used The nature of the variables The number of treatment compared The experimental design Additional factors taken into account for analysis (e.g. baseline)

Failure to define the analysis sets Full analysis set The set of subjects that is as close as possible to the ideal implied by the ITT principle. It is derived from the set of all randomized subjects by minimal and justified elimination of subjects.

Intent to Treat Principle (ITT) The patients data should be analyzed in their assigned treatment group after they have been randomized regardless the treatment they are actual received. Fundamental point –Excluding participants or observed outcomes from analysis and sub-grouping on the basis of outcome or response variables can lead to biased results of unknown magnitude or direction.

Criteria to exclude randomized subjects from full analysis set 1.Failure to satisfy major entry criteria 2.Failure to take at least one does of trial medication 3.The lack of any data post randomization

Per Protocol Set ‘valid cases’, ‘efficacy sample’ or the ‘evaluable subjects’. The set of data generated by the subsets who complied with the protocol sufficiently to ensure that these data would be likely to exhibit the effect of treatment, according to the underlying scientific model.

Criteria of defining Per Protocol Set The completion of a certain pre-specified minimal exposure to the treatment regimen The availability of measurements of the primary variable(s) The absence of any major protocol violations including the violation of entry criteria

An Example: Protocol criteria for patients included in evaluable and ITT analysis Patients who complete all of the visits without violation or major deviations and are at least 80% compliant in taking medication, will be analyzed in the per protocol analysis. All patients taking at least one does of study medication will be included in the intention to treat analysis.

Testing the baseline imbalance This is a common procedure which has no justification in statistical theory Baseline imbalance can not justify the integrity of randomization process. Randomization does not guarantee the balanced of baseline Baseline will be adjusted in the analysis.

Fail to specify the policy on missing values and outliers Imputation techniques to compensate for missing data: –Carry forward of the last observation –Complex mathematical models –Defer detailed policy on irregularity until the blind review of the data at the end of the trial

Blind review The checking and assessment of data during the period of time between trial completion (the last observation on the last subject) and the breaking of the blind, for the purpose of finalizing the planned analysis.

Failure to specify data transformation Transformation (e.g. square root, logarithm) should be specified in the protocol and a rational provided. Especially for the primary variable(s).

Failure to define other derived variables Change from baseline Percentage change from baseline AUC of repeated measures Ratio of two different variables

Excessive emphasis on p-values Confidence Intervals are much more informative Justification for one sided test Type I error Statistical model and the assumptions underlying such models Parametric and non parametric

Failure to consider the adjustment of multiplicity Significance Confidence levels

Multiplicity and Method to Reduce Multiplicity

Inappropriate (or insufficient) use of covariate information Using change from baseline rather than fitting baseline as a covariate. –The inference based on the covariance adjustment is generally more precise than that based on the change adjustment ( Patel (1983,1986) Kenward and Jones (1987) To adjust the main analysis for covariates measured after randomization

Covariant Definition –Efficacy variables or treatment responses are often influenced by or related to factors other than treatment.

Covariant Adjustment Randomization can not guarantee the comparability or the balance of all the covariates especially in smaller studies. In order to obtain a valid and more precise inference of the treatment effect, it is necessary to adjust for covariates that are statistically correlated with the clinical endpoints.

Possible Covariates (Confounding factors, Prognostic factors, risk factors) Demographic –age, gender and race Patient characteristics –disease severity, concomitant medication and medical history Centre in a multicentre study

Data Type and Adjustment Procedure

Failure to model outcomes adequately Treating ordered categorical data in a way that ignores the ordering

Relationship Between Frequency of Caesarian Section(CS) and Maternal Shoe Size

Snoring Behavior in Relation to Presence or Absence of Heart Disease

Fail to reflect the trial structure Carry out a multicentre trial and not fitting centre the centre effect.