1 of 11

What and How we do as a pharmaceutical statistician

Sabrina Wan

Merck & Co., Inc.

Conference on Advances in Statistical and Computational Methods for Analysis of Biomedical, Genetic, and Omics Data

March, 2023

​

​

Public

2 of 11

Outline

  1. Drug Development Process
  2. Common considerations in late phase drug development: statistical, clinical, regulatory, operational, etc.
  3. Key elements for a pharmaceutical statistician’s job

2

Public

3 of 11

3

​

​

​

​

Drug Development Process

Research & Discovery Stage

Pre-clinical

Phase 1

Phase 2

Phase 3

Phase 4

Activities

  • Develop scientific rationale based on disease etiology;
  • Conduct laboratory research to test candidate compounds
  • Perform additional laboratory research and testing in animals
  • Obtain information on dosing and toxicity levels
  • 20 to 100 healthy participants or with disease
  • To evaluate initial safety and dosage

Clinical Development

  • Up to hundreds of participants with the disease / condition
  • To evaluate efficacy (signal finding/dose ranging) and side effects
  • Hundreds to thousands of participants with disease / condition
  •  To show treatment effect with statistical significance
  •  Continue monitoring of adverse reactions

Investigational New Drug (IND) application to FDA

End of Phase 2 Meeting with FDA

NDA/BLA Application

  • After FDA approval, post- marketing monitoring for safety

Public

4 of 11

Some Considerations in Late Phase Drug Development (1)

  1. What is the unmet medical need? Disease landscape defines patient population.
  2. What are the current available therapies? Use SOC as control.
  3. What data evidence do we have? Literature or in-house data
  4. What measure to use to quantify treatment effect? Endpoint definition
  5. What is the study objective? Proof of concept, signal finding, confirmatory
  6. What is the regulatory pathway? Consider FDA’s expedited programs
  7. Fixed-sample design or adaptive design? Budget/timeline/feasibility/regulatory considerations
  8. Is blinding needed/feasible? Double-blind/single-blind/open-label
  9. How is randomization done? Stratified or adaptive randomization
  10. What is the multiplicity control strategy? Strong control of Family-wise Error Rate (FWER)

​

​

​

Public

5 of 11

Some Considerations in Late Phase Drug Development (2)

11. What is the most appropriate statistical model/analysis method?

12. What is the estimand and how missing data/intercurrent events are handled?

13. How to establish contribution of component for combination therapy?

14. What is the strategy for integrated analysis of efficacy and safety?

15. How consistent the results are within important subgroups?

​

​

​

Public

6 of 11

An oncology Study: KN426

A P3 open-label study comparing Pembrolizumab plus Axitinib versus Sunitinib for Advanced Renal-Cell Carcinoma1

​

Study Design Schema can be found at

https://www.urotoday.com/conference-highlights/asco-gu-2019/asco-gu-2019-kidney-cancer/112886-asco-gu-2019-phase-iii-keynote-426-study-pembrolizumab-plus-axitinib-versus-sunitinib-as-first-line-therapy-for-locally-advanced-or-metastatic-renal-cell-carcinoma.html

​

​

​

​

​

​

​

1https://www.nejm.org/doi/full/10.1056/NEJMoa1816714

​

​

​

​

​

​

Public

7 of 11

KN426 Considerations (1)

  1. What is the unmet medical need? 1L RCC patients
  2. What are the current available therapies? At the time of study design, TKI is the SOC.
  3. What data evidence do we have? Promising anti-tumor activity noted based on Phase 1b data1
  4. What measure to use to quantify treatment effect and safety? OS, PFS and ORR; AEs
  5. What is the study objective? Confirmatory study
  6. What is the regulatory pathway? FDA granted breakthrough status, potential full approval1

​

​

Public

8 of 11

KN426 Considerations (2)

7. Fixed-sample design or adaptive design? Group sequential design to allow stop trial earlier for efficacy or futility

8. Is blinding needed/feasible? Open-label due to different dosing schedules, in-house blinding

9. How is randomization done? Stratified randomization as limited # of strata (prognostic factors)

10. What is the multiplicity control strategy? Graphical approach in group sequential design

11. What is the most appropriate statistical model/analysis method? Log rank test, cox proportional model (under proportional hazard assumption) and KM curve estimate

12. What is the estimand and how intercurrent events are handled? Depend on the endpoint, apply different strategies to handle intercurrent events. *

13. How to establish contribution of component for combination therapy? Cross trial comparison to monotherapy**

*https://www.fda.gov/media/148473/download

**https://aacrjournals.org/clincancerres/article/26/24/6406/82943/Regulatory-Considerations-for-Contribution-of

​

​

​

Public

9 of 11

A non-oncology Program

VRAYLAR: one P2 and two P3 double-blind placebo-controlled study comparing VRAYLAR vs placebo to support indication of treatment of schizophrenia and a randomized withdrawal trial to support as maintenance treatment*

1. What is the strategy for integrated analysis of efficacy (ISE) and safety (ISS)?

Pooling of similar doses for ISE and all doses/similar doses for ISS

2. How consistent the results are within important subgroups?

Subgroup analysis and post hoc analysis adjusted by important covariates

​

​

​

*https://www.rxabbvie.com/pdf/vraylar_pi.pdf

​

​

Public

10 of 11

Key Elements

​

​

Appropriate Statistical Techniques

Good understanding of Disease and Regulatory Landscape

Proactive Interactions with cross-functional teams and regulatory agency

Improved Efficiency in Drug Development

Public

11 of 11

Thank you!

​

​

​

​

​

​

​

It takes a team to escape from the dinosaur!

11

Public