frankstat — Biostatistics & Clinical Trial Methodology

Consultant. Trainer. DMC Statistician.

Independent biostatistical expertise for biotech, pharma and CROs — from study design to data monitoring.

20 years in clinical drug development 60-person global team led 50+ methodology publications
Consulting
Rigorous statistical support across the full clinical development lifecycle.

Statistical & Methodological Consulting

I help biotech, pharma and CRO teams make sound methodological decisions — from first-in-human through confirmatory trials. Whether you need a second opinion on a design, a defensible analysis plan, an independent review of your CRO's work, or hands-on support through a submission, I bring the perspective of someone who has worked these problems from inside a major pharmaceutical statistics function.

For teams without in-house biostatistics I also act as fractional lead statistician — the senior voice in design discussions, regulatory interactions and CRO meetings.

  • Study design & sample size planning
  • Regulatory interactions (FDA, EMA)
  • Adaptive & dose-finding designs
  • Interim analysis planning
  • Fractional lead statistician
  • SAP development & review
  • Randomisation strategy
  • CRO oversight & output review
Training
Turning statistical depth into skills your team can use.

Training, Masterclasses & Coaching

I design and deliver in-house and open masterclasses for clinical statisticians and cross-functional teams — each built around real trial problems rather than textbook abstractions. Beyond scheduled courses, I offer one-to-one coaching for statisticians moving toward leadership, and tailored sessions for teams strengthening a specific methodological area.

For cross-functional and leadership audiences I offer sessions that build statistical literacy without the mathematics — enough to challenge a design, read an interim report, or brief a board.

For statisticians
  • Statistical Leadership in Drug Development
  • Adaptive Designs
  • Dose Finding
  • Bayesian Methods in Clinical Trials
  • Clinical Trial Randomisation
  • 1:1 leadership coaching
For clinical, regulatory & leadership teams
  • Statistics for Decision-Makers
  • Clinical Trial Statistics for Non-Statisticians
  • Reading and Challenging Interim Results
  • Tailored in-house sessions

Delivered in English or German · on-site or remote.

Data Monitoring
Serving as the statistical voting member of Data Monitoring Committees.

DMC Member (Statistician)

I serve as the independent statistician on Data Monitoring Committees — the voting member who reviews and interprets unblinded interim analyses and contributes to the committee's recommendations on trial continuation, modification, or termination. My focus is on critically evaluating the evidence presented to the DMC, not on producing it.

I keep a clear separation from the reporting statistician who prepares the outputs and from the sponsor team, preserving the independence that makes a DMC credible.

  • Independent statistical voting member
  • Review of unblinded interim data
  • Safety & efficacy interpretation
  • Continue / modify / stop recommendations
  • Firewall separation from reporting & sponsor
  • No conflicting engagements
How we work together
Three engagement models, one point of contact.

Engagement models

Every engagement starts with a short, no-cost conversation about the problem. From there, we agree on one of three formats:

Project — A defined scope with a defined outcome: a design review, a sample-size and simulation package, an SAP review, or regulatory-meeting preparation. Fixed price or capped effort, agreed up front.

Retainer — A monthly time budget for teams that need a senior statistician on call: fractional lead statistician for a biotech without in-house biostatistics, or standing methodological back-up for a pharma statistics group.

Mandate — DMC statistical membership or a training and coaching programme, priced per meeting or per course day, including preparation.

CROs engage me as a subcontracted expert — under their own client relationship where preferred. NDA-ready, available at short notice, and used to working inside sponsor and CRO processes alike. All work is delivered in English or German.

No fixed price list — rates follow scope and depend on urgency and travel. Ask for a quote.

About

Two decades of statistical leadership in drug development.

I am Frank Fleischer, a biostatistician with 20 years in clinical drug development. Until 2026 I was Global Head of Therapeutic Area & Methodology Statistics at Boehringer Ingelheim, leading a group of up to 60 statistical and data-science experts across Europe, the US, and China, with oversight of randomisation, unblinding, and DMC support. Since 2026 I work independently under the frankstat brand.

My background pairs deep methodological work with the leadership of large statistical functions. I hold a Dr. rer. nat. in stochastics (Ulm University) and an MSc in mathematics (University of Wisconsin–Milwaukee), have authored over 50 publications, and have taught Clinical Trials as a guest lecturer at Ulm University since 2015. Away from the desk, I hold a chess candidate-master title — the same habit of reasoning several moves ahead under uncertainty that statistics rewards.

My therapeutic-area experience is deepest in oncology and immunology, with broad exposure across other indications. I work in R for analysis and simulation, and use Claude as an AI assistant for drafting, review and course development — never with client data unless explicitly agreed, and always with the statistician's judgement kept firmly in the loop.

Selected methodological contributions
  • Inventor of Bayesian MCPMod for incorporating external information into dose finding
  • Established BLRM-EWOC as the standard for oncology dose-finding trials at Boehringer Ingelheim
  • Established MCPMod as the standard for non-oncology dose finding at Boehringer Ingelheim
  • First to implement Bayesian borrowing in a clinical trial at Boehringer Ingelheim
  • Lead author of quantitative Go / No-Go frameworks for early-phase decisions
Dr. rer. nat., stochastics · Ulm MSc mathematics · UW–Milwaukee 50+ publications · ResearchGate ↗ Guest lecturer · Ulm University
Case studies

What an engagement looks like in practice.

Two anonymised examples of typical consulting work. Settings and numbers have been altered and the figures are illustrative; the methods and the reasoning are real.

Case study 01
Sharing information between cohorts without forcing them to be alike.

Interim decisions in a multi-cohort basket trial

The question

A team planned a randomised trial of one compound across six patient cohorts with a time-to-event primary endpoint. Each cohort was too small to stand alone, yet the team needed a credible interim rule to decide which cohorts to continue.

Approach, outcome & figure
The approach
  • Compared cohort-by-cohort analysis with Bayesian hierarchical borrowing (EXNEX), which shares information between similar cohorts without forcing them to be alike.
  • Incorporated a small earlier study as a robust, deliberately down-weighted prior, reflecting its methodological limitations.
  • Simulated the operating characteristics of candidate interim rules under optimistic, null and mixed scenarios, holding the false-go rate constant to keep the options comparable.
  • Translated the results into explanatory slides for a non-statistical audience.
The outcome

The team received a pre-specified interim decision framework with transparent error rates, a justified degree of borrowing with sensitivity analyses, and a clear view of how much the earlier data could reasonably contribute.

stand-alone with borrowing (EXNEX) 0.25 0.5 1 2 Cohort A Cohort B Cohort C Cohort D Cohort E Cohort F Hazard ratio (log scale) · values < 1 favour treatment
Illustrative interim data, 35 events per cohort: point estimates with 95% intervals. Borrowing narrows the intervals for the similar cohorts A–E, while the outlying cohort F is only partly pulled towards the others.
Case study 02
Power answers “what if we are right?” Assurance asks how likely that is.

Go / no-go criteria and assurance for a small signal-finding study

The question

A team planned an early study with two dose arms of 18 patients each and a response-rate endpoint. A response rate of 30% was considered uninteresting, 50% was the target. The decision rule: go if at least one arm shows 9 or more responders. At the target rate this gives an 83% chance of a go — but the target was a hope, not a fact. How likely was a go, realistically?

Approach, outcome & figure
The approach
  • Calculated exact go probabilities for the rule across the range of true response rates: 12% at the uninteresting rate, 83% at the target.
  • Expressed the existing evidence as a prior distribution for the true response rate — centred on 45%, carrying the weight of about 11 patients.
  • Averaged the go probability over this prior. The resulting assurance was 60%, well below the 83% the team had in mind.
  • Decomposed the result: the prior gave a 63% chance that the compound is truly active (40% or higher). If it is, the study says go in 83% of cases; if the true rate is 30% or lower, in only 4%.
The outcome

The team went into the study with a realistic expectation of success and could explain it to governance. They also saw that the gap between 83% and 60% reflects uncertainty about the compound rather than a weakness of the design — and what a go, once observed, would and would not mean.

0% 25% 50% 75% 100% 10% 20% 30% 40% 50% 60% 70% 80% 12% at 30% 83% at 50% Assurance 60% Probability of go Prior for the true rate True response rate
Probability of a go decision (at least one of two arms with 9 or more responders out of 18) against the true response rate, assumed equal in both arms. Averaged over the prior, the probability of a go is 60%.
Contact

Let's discuss your trial, your team, or your committee.

Tell me a little about the problem — a design under review, a course for your statisticians, or a DMC seat to fill — and I'll come back with how I can help.

Based in the Ulm area, southern Germany Available for business travel Courses in EN & DE