Clinical Research Statistics for Non-Statisticians
Download as PPTX, PDF5 likes2,818 views
Jennifer Marcello is a senior biostatistician with over 5 years of experience in statistical planning, analysis, and reporting for clinical trials, particularly in oncology and various other diseases. The document outlines key aspects of clinical trial planning, including study flow, sample size determination, randomization methods, and the importance of statistical analysis plans. It emphasizes the need for bias reduction, proper data capture, and ongoing evaluations throughout the clinical trial process.
![Planning Your Trial –
Blinding/Masking
Advantages
• Decrease bias
• Participant response
not influenced by
knowledge of
treatment
• [DB] Investigator
preconception does
not matter
Disadvantages
• Patient consent
• Another layer of
complexity
• [TB]Patient safety
• Can the study really
be blinded?](https://image.slidesharecdn.com/statisticswebinarslideshare-160524143322/85/Clinical-Research-Statistics-for-Non-Statisticians-9-320.jpg)
More Related Content
Similar to Clinical Research Statistics for Non-Statisticians (20)
Clinical Research Statistics for Non-Statisticians
- 2. JenniferMarcello, SeniorBiostatistician • Jennifer has experience in statistical planning, analysis, and reporting for Phase 1, 2 and 3 clinical trials. Her research experience includes over 5 years of experience working on clinical trials in oncology including brain, skin, lung, breast, head and neck, and colorectal cancers. In addition to oncology, she has experience in clinical trials for palliative care, hepatitis, HIV, Alzheimer’s disease, anti-infectives, and pain. Jennifer is trained in writing detailed statistical analysis plans, performing sample size computation, preparing statistical analysis specifications of analysis databases, developing summary displays including summary tables for integrated safety and efficacy data, utilizing SAS® software for programming and analysis of clinical data, and providing ongoing safety evaluations for data monitoring committees. She also has experience in analyzing quality of life data, nutrition data and other patient reported outcomes.
- 3. Agenda Clinical Trial Study Flow Planning Your Trial Sample Size and Power Data Capture Randomization Statistical Analysis Plan Interim Analyses Database Lock Final Analyses CSR When do I need a statistician?
- 4. Clinical Trial Study Flow CSR Displays Analysis Datasets SAP CRF Randomization Scheme Protocol Study Planning
- 5. Planning Your Trial What is our goal? What data do we collect? How do we test them?
- 6. Planning Your Trial - Example • OA of the kneeIndication: • Show our product is better than placeboGoal: • Pain by VAS on 50 foot walk test, multiple collection times Data to collect:
- 7. Statistical Review- Example • OA of the knee unilateral, bilateral, age?Population: • ContinuousType of data: • Repeated Measures Number of time points: • LS Means Difference based on Repeated Measures Population Average ModelTest: • Unilateral vs. Bilateral, Missing DataSensitivity:
- 8. Planning Your Trial – Blinding/Masking Single Blinding • The participant doesn’t know to which intervention they have been assigned. Double Blinding • The participant and the investigator don’t know to which intervention the participant has been assigned. Triple Blinding • The participant, investigator, and monitoring committee do not know to which intervention the participant is assigned.
- 9. Planning Your Trial – Blinding/Masking Advantages • Decrease bias • Participant response not influenced by knowledge of treatment • [DB] Investigator preconception does not matter Disadvantages • Patient consent • Another layer of complexity • [TB]Patient safety • Can the study really be blinded?
- 10. Study Populations Who do you want in your study? • Inclusion & Exclusion criteria Ensure that statistical inference can be made to targeted market population • Safety Analysis Set • Full analysis sets (ITT Population) • Per Protocol • Depending on draft guidance, Clinically evaluable
- 11. Sample Size and Power • TOO SMALL • NOT ADEQUATE TO ADDRESS QUESTIONS • TOO LARGE • WASTED TIME, RESOURCES, AND MONEY • POTENTIALLY EXPOSE PTS TO INEFFECTIVE TRT 2 Basic Approaches • Power of a hypothesis test (most common) • Precision
- 12. Sample Size and Power • Sample size calculation depends on: • Planned analysis method/ hypothesis • Clinically significant difference/ effect size • α:Type I error • β:Type II error • σ: standard deviation • Other considerations • Cost • Expected dropout rate
- 13. Sample Size and Power Standard Deviation Power Desired Acceptable Error Clinically Significant Difference Power Desired Cost
- 14. Hypothesis Testing Study comparing Drug X to placebo in lowering pain due to osteoarthritis of the knee. • The null hypothesis, H0, is the hypothesis to be tested. H0: μdrug = μplacebo • The alternative hypothesis, Ha, is the hypothesis which contradicts the null hypothesis. Ha: μdrug ≠ μplacebo
- 15. Possible Outcomes of Hypothesis Tests Correct decision Type II error (β) Type I error (α) Correct decision True State of Nature H0 true H0 false Decision RejectH0FailtorejectH0 False negative False positive (H0 = null hypothesis)
- 16. SMOKE ALARM SYSTEM Correct decision Type II error (β) Type I error (α) Correct decision No Fire Fire AlarmNoAlarm False negative False positive
- 17. Sample Size and Power The lower the allowable error, the bigger the sample size REMEMBER:
- 18. Sample Size and Power The higher the power, the bigger the sample size REMEMBER:
- 19. Sample Size and Power The bigger the standard deviation, the bigger the sample size REMEMBER:
- 20. Sample Size and Power The bigger the clinically significant difference, the smaller the sample size REMEMBER:
- 21. Sample Size and Power All differences can be “statistically significant” if you have enough subjects, power only for your clinically significant difference! REMEMBER:
- 22. Statistical Significance • Informally, a p-value is the probability under a specified statistical model that a statistical summary of the data (for example, the sample mean difference between two compared groups) would be equal to or more extreme than its observed value. Ronald L. Wasserstein & Nicole A. Lazar (2016): The ASA's statement on p-values: context, process, and purpose, The American Statistician, DOI: 10.1080/00031305.2016.1154108
- 23. Statistical Significance • If p-value > α, we fail to reject the null hypothesis, and the result is considered statistically insignificant • If p-value ≤ α, we reject the null hypothesis, and the result is considered statistically significant
- 24. Type I Error Rate Control • Multiple looks (Unmasked/Unblinded interim analyses) • Multiple comparisons (More than one primary hypothesis/endpoint)
- 25. Data Capture CRF design is integral to capturing the data you need for a successful analysis. Statisticians need to participate in CRF design to make sure assessments align with analyses! It’s VERY difficult to go back and obtain data after the fact! Will this study be part of a submission? • CDASH (CRF=Case Report Form)
- 26. Data Capture – Missing Data • Potential source of bias • Minimize through protocol design • Consult guidance, literature, and your statistician for candidate methods for analysis • Define and justify the proposed method • Communicate with client and the internal team See: O’Neill, R and Temple, R. “The Prevention and Treatment of Missing Data in Clinical Trials: An FDA Perspective on the Importance of Dealing With It.” Clin Pharmacol Ther. 2012 Mar, 91(3); 550-4.
- 27. Clinical Data Standards • Clinical Data Interchange Standards Consortium (CDISC) – Clinical Data Acquisitions Standard Harmonization (CDASH) -> data collection – Study Data Tabulation Model (SDTM) -> ‘raw’ data – Analysis Dataset Model (ADaM) -> analysis ready data www.cdisc.org
- 28. Randomization Reasons • Reduction of bias • Sound statistical basis for evaluation • Produces treatment groups in which the distributions of prognostic factors, known and unknown, are similar
- 29. Randomization - Types Simple Randomization Like flipping a coin Pro: Easy! Con: You could randomize everyone to the same group
- 30. Randomization - Types Permuted Block Randomization Randomized by block Pro: Balance across intervention arms Con: If you know the block size (and it’s small), you may be able to guess the next treatment Block 1 2 3 4 5 6 7 8 9 Treatments ABC CBA CAB BCA ACB ACB ABC CAB BCA
- 31. Randomization - Types Stratified Randomization If a key factor may affect how an intervention works in a particular group, stratify by that factor. Can combine this method with permuted block for balance: Permuted block stratified by baseline pain: Moderate pain: AABB ABAB BBAA Severe pain: ABAB BBAA BABA 3 blocks of size 4 3 blocks of size 4
- 32. Statistical Analysis Plan • What data will we use? • Which participants will be included? • Exactly how will we analyze? • Factors affecting analysis? Gets down to the nuts and bolts of the statistical analyses
- 33. Statistical Analysis Plans • Descriptive Statistics • T – test and Non- parametric Test (Wilcoxon Test) • ANOVA and ANCOVA • Linear Regression • Linear Mixed Models Continuous Outcomes • Descriptive Statistics • χ2 / Fisher’s Exact Test • CMH test, Odds Ratio, Relative Risk • McNemar’s, Agreement (Kappa) • Logistic Regression • Poisson Models Categorical Outcomes
- 34. Statistical Analysis Plans • Kaplan Meier • Log Rank Test • Survival Rates • Poisson Models • Cox Proportional Hazard Models Survival Analysis • Pattern Mixture Models • Missing Data Imputation Methods • LOCF • BOCF • Multiple Imputation Sensitivity Analyses
- 35. Interim Analyses • Is an interim analysis planned? • What is the purpose of the interim analysis? • Interim analysis timing and frequency • Is an unblinded interim team needed? • What is the data cleaning process for the interim analysis? • How does this affect α?
- 36. Interim Analyses - IDMC/DSMB • SafetyPurpose • Based on outcome and safety concerns Timing and Frequency • Possible, not always necessaryUnmasked Team • Interim database locks, snapshotsData Cleaning • No efficacy dataAffected α?
- 37. Interim Analyses – Sample Size Recalculation • Ensure necessary sample size based on SD assumptionsPurpose • Usually just once Timing and Frequency • Not necessaryUnmasked Team • Interim database locksData Cleaning • Not if performed in a pooled SD adjustmentAffected α?
- 38. Interim Analyses – Stopping Rules • EfficacyPurpose • Based on primary outcomeTiming and Frequency • RequiredUnmasked Team • Interim database locksData Cleaning • Yes, if study continuesAffected α?
- 39. Interim Analyses – Adaptive Designs • EfficacyPurpose • Based on primary outcome Timing and Frequency • Usually, based on arms involvedUnmasked Team • Interim database locksData Cleaning • Yes, but not usually well controlled studies (Phase I or II)Affected α? Example: Pruning
- 40. Database Lock and Unmasking • All analysis plans should be complete • Per-Protocol population selection • Statistician sign off • Data quality • Missing data? • Unmasking
- 41. Final Analyses • Hypothesis Testing • Primary/Secondary Outcomes • Safety Reporting • Missing Data?
- 42. Clinical Study Report • Statistical Reporting • Primary Endpoint Discussion • What does it all mean??
- 43. Summary • Determining trial objectives and corresponding endpoints, primary and secondary, is important initial step. • The type of trial should be aligned with sponsor’s clinical plan. • Determining the sample size early is very important to the projected cost for running the trial. • Statistical parts of the protocol serve as starting point for all remaining activities. • Emphasis on design and statistical principles protects the study from bias by specifying the analysis methods a priori.
- 44. References • ICH-E3: Structure and Content of CSRs • ICH-E6:Good Clinical Practice: Consolidated Guidance • ICH-E9: Statistical Principles for Clinical Trials • ICH-E10: Choice of Control Group and Related Issues in Clinical Trials • FDA Guidance for Industry: Various Indications • National Academy of Science Missing Data Guidance • “Statistical Reasoning in Medicine: The Intuitive P-Value Primer” – Lemuel A. Moyé • “Fundamentals of Clinical Trials” - Lawrence M. Friedman and Curt D. Furberg