Checklist for Biostatistical Review of Research Protocols

Hypotheses/Objectives

• Clearly articulated and consistent with aims, both the null and alternative hypothesis needs to be specified.

General design

• General study design (observational, RCT, pragmatic etc) will effectively answer the question
• General study design is appropriate for current state of knowledge

Population and sample

• Population from which the sample will be drawn is representative of the population of interest (e.g. disease severity, time point during disease progression), or the study is correlational / comparative in a way that is likely to cancel out any shifts in distributions from population to sample
• Inclusion and exclusion criteria appropriate for state of knowledge
• Screening and enrollment processes do not introduce bias

Variables and measurements

• Choice of measurements, especially the response variable, are justified and consistent with the hypothesis/objectives
• Measurements are of maximum feasible resolution with no unecessary categorization
• Continuous responses, or ordinal responses with several well-populated levels, are used wherever possible
• If a binary response variable is used, it should be justified and the effect on power and precision acknowledged
• Algorithms used to derive variables or score outcome assessments are appropriate

Treatment assignment

• Method for assigning treatments (e.g. randomization) minimizes bias or potential confounding factors are carefully thought out.

Data integrity

• Data collection methods are sufficient to result in accurate measurement of primary variable(s)
• Data management uses appropriate tools (e.g. REDCap)
• Plans for monitoring or interim data checks are appropriate

Data handling rules

• Range and consistency checks are included
• Handling of missing data is addressed

Statistical environment

• Statistical collaborator is engaged
• Software is acceptable (not Excel)
• Plans for reproducibility are stated

Analysis plan

• Statistical approach is consistent with hypothesis and objectives
• A plan for describing the dataset is given
• Unit of analysis is clearly described for each analysis
• Analysis populations clearly described (intent to treat set, per protocol set, full analysis set etc)
• Analytical approach is justified, including consideration of assumptions and alternative approaches
• Approach minimizes unnecessary exposure of participants to risk

Sample size justification

• Alpha and 1-beta given for all statistical tests and justified based on state of current knowledge
• Tests are identified as one or two-tailed and justification provided
• Justification is based on clinically important differences or on precison (margin of error) with a margin of error that is acceptable to advance learning, or at least the PI has acknowledged the limited learning from a poorly powered study and can justify the risk given the limited benefit.