Interview: clinical trial statistician
Knowledge framework
Book: Statistical design, monitoring and analysis of clinical trials
Behavior
Statistics
Study design
Example questions
- why use OR instead of RR in case control studies
Survival analysis
(datasets: survival::pbc)
use ggsurvfit
Hypothesis testing
Longitudinal data
(datasets: mmrm::bcva_data, mmrm::fev_data)
Regression
Example questions
- what is confounding, how can you adjust for it
Missing data
Meta analysis
Programming
- How would you implement survival analysis in R? Walk me through the process of creating Kaplan-Meier curves and performing Cox proportional hazards analysis.
- Can you explain how you would handle missing data using the mice package in R? Provide an example.
- What packages would you use for meta-analysis in R? Can you demonstrate how to perform a meta-analysis using the meta or metafor package?
- How would you simulate a clinical trial in R to calculate power? What tools or packages would you use for this purpose?
- Can you provide an example of how to perform bootstrapping to assess the robustness of a clinical trial result?
Background
Statistician’s role
Workflow: design -> data collection, monitoring -> analysis
Design
- define primary and secondary endpoints, need to be clearly measurable (e.g. progression free survival, overall survival, tumor response)
- randomization and blinding
- calculate sample size
- design study design (phase 1, 2, 3), sometimes adaptive design, Bayesian approach
Analysis
- survival (usually primary endpoint)
- censored data handling
- efficacy and safety: tumor shrinkage; QoL, side effects
- side effects frequencies
- subgroup analysis (need to pay attention to the risks of over-interpreting data)
- biomarker to predict treatment response
Reporting
Regulatory
Regulatory and Compliance:
- Can you explain the ICH E9 guidelines, and why they are important in clinical trials?
- How would you prepare the statistical section of a regulatory submission for the FDA or EMA?
- What are the CONSORT guidelines, and how would you ensure compliance with them in reporting trial results?
Data Monitoring and Reporting:
- What is the role of a Data Safety Monitoring Board (DSMB), and how would you collaborate with them as a statistician?
- How do you ensure data integrity in the case of interim analyses or adaptive trial designs?
Oncology
- In oncology, surrogate endpoints like progression-free survival (PFS) and overall response rate (ORR) are often used. How would you handle the analysis of these surrogate endpoints?
- Can you explain the RECIST criteria used for assessing tumor response in clinical trials?
- How do you design a basket trial or umbrella trial in oncology, and what statistical challenges do these designs present?
Hospital vs Pharmaceuticals
Hospital
Topics of interest: long-term patient outcomes, QoL, comparative effectiveness of treatments, rare conditions that are less commercially viable. E.g. new surgical technique; compare different treatment regiments for patients with specific biomarkers
Design: more flexible and exploratory, novel designs; smaller, single centered trials; early-stage, proof-of-concept
Pharma
Topics of interest: strongly focused on meeting requirements of regulation, focus on predefined endpoints like efficacy and safety that align with drug approval processes. E.g. new cancer drug compared to placebo or existing treatment
Design: rigid design aligned with regulatory requirements; large, multicenter; RCT with precise inclusion/exclusion criteria