Getting the quantitative question right before the trial answers a different one.
Six places a quantitative choice decides a programme
Estimands under ICH E9(R1)Defining what your trial is actually estimating, and aligning the estimator with it. I co-authored three of the papers the industry now uses to put E9(R1) into practice.
Probability of successAssessing whether a development programme should proceed to a pivotal trial, and making that assessment something governance can act on rather than argue about.
Quantitative benefit–riskStructured assessment of a compound’s profile, including Bayesian methods that incorporate prior data properly rather than informally.
Innovative and adaptive designsPlatform trials, master protocols, decentralised and hybrid elements — what they buy you, what they cost, and when the added complexity is not worth it.
Biosimilar development strategyComparability programmes and Phase III design, from five published biosimilar studies across filgrastim and pegfilgrastim.
Review and second opinionReading a protocol, a statistical analysis plan or a submission package with the eye of someone who has defended such documents to regulators.
I co-authored the three papers the industry uses for estimands
Estimands are the clearest case. When ICH E9(R1) landed, the guidance said what an estimand was but not how to choose one. A working group wrote the papers that closed that gap: Choosing Estimands, Defining Efficacy Estimands and Aligning Estimators With Estimands. I am an author on all three.
I also chair the Bayesian Scientific Working Group of the ASA Biopharmaceutical Section: over 250 statisticians, academics and regulatory scientists across eight subteams, elected through three successive offices since 2017. And I have taught on the UCSF–Stanford CERSI course on Bayesian thinking in clinical research.
Underneath that is fifteen years of doing the work at Eli Lilly, AstraZeneca, UCB, Sandoz and Novartis, across neuroscience, oncology, immunology, cardiovascular, psychiatry and rare disease, through Phase II, Phase III and regulatory submission.
This fits four situations, and three it does not
A good fit
You have no senior statistician in house and decisions are waiting on one
You are designing a pivotal trial and want the estimand settled properly first
Your CRO’s analysis plan needs independent review before you commit
You need to defend a quantitative choice to a regulator or a board
Not a fit
You need a full statistical function staffed and run
You want programming and data management delivery
You want analysis shaped toward a conclusion already chosen
Choosing Estimands in Clinical Trials: Putting the ICH E9(R1) Into PracticeRatitch, B., et al., incl. Singh, P. (2019). Therapeutic Innovation & Regulatory Science. DOI
A New Comprehensive Approach to Assess the Probability of Success of Development Programs Before Pivotal TrialsHampson, L. V., et al., incl. Singh, P. (2021). Clinical Pharmacology & Therapeutics, 111(5). DOI
Incorporating Prior Data in Quantitative Benefit–Risk Assessments: Case Study of a Bayesian MethodDharmarajan, S., et al., incl. Singh, P. (2024). Therapeutic Innovation & Regulatory Science, 58(3). DOI
Advancing innovative clinical trials to efficiently deliver medicines to patientsBeckman, R. A., Natanegara, F., Singh, P., et al. (2022). Nature Reviews Drug Discovery, 21(8). DOI