Services

01 — Evidence for decisions

HTA & market access

Reimbursement decisions turn on a small number of choices made long before submission: which comparator, which population, which outcome, and what counts as adequate evidence for each. We help you make those choices deliberately.

The EU HTA Regulation has changed the shape of the problem. Since 12 January 2025, new oncology medicines and advanced therapy medicinal products go through a Joint Clinical Assessment at EU level, with orphan medicines and then all new medicines following in subsequent stages. The JCA does not replace national decisions — it front-loads them, and it multiplies the number of PICO combinations a dossier has to be able to answer. Planning for that after the trial has read out is expensive; planning for it during protocol design is not.

For medtech and diagnostics the assessment landscape is different again — smaller evidence bases, more heterogeneous comparators, and national bodies that judge clinical and organisational value together. The method has to fit the assessor, not the other way round.

Where we usually add most

• Deciding what evidence to generate before the pivotal study locks its design
• Turning a heterogeneous set of national PICOs into a manageable analysis plan
• Making the case for real-world evidence where a head-to-head trial does not exist
• Writing or reviewing the clinical-effectiveness argument so it reads the way assessors read
• Preparing the team for the questions the assessment body will actually ask

02 — Studies in real data

Real-world evidence studies

A real-world study is only as good as the question it was designed around. We start from the decision that has to be made, define the estimand explicitly, and only then go looking at what the data can support.

The Nordic setting is unusually strong for this: nationwide registers with near-complete coverage, a personal identifier that links across health, prescription and socioeconomic data, and decades of follow-up. It is also unusually easy to get wrong — coverage and coding practice change over time, linkage introduces its own selection, and access runs through secure environments with real constraints on what you can do and export.

Method choice follows the question. Target trial emulation where the question is causal and the temptation to compare prevalent users is strong. Propensity or overlap weighting where confounders are measured. External or synthetic control construction where the trial had no comparator arm. Descriptive epidemiology where that is honestly all the data can carry — and saying so is part of the job.

Where we usually add most

• Feasibility: can this data source answer this question, at what precision
• Designing away the biases that get studies rejected — immortal time, prevalent user, informative censoring
• Pre-specified protocols and analysis plans that survive external scrutiny
• Turning a completed analysis into a report a committee can follow

03 — The analysis itself

Biostatistics & data science

Sometimes the need is not a whole study, it is a statistician: someone senior to write the analysis plan, do the modelling, and be able to defend both.

We work mostly in R, with analysis organised so it can be re-run from raw data to final table by someone else — which matters more than it sounds, because it is what makes a result checkable a year later when a reviewer asks. Where a client works in SAS or Stata we can read and review it; where a client works inside a controlled environment, we work inside it under their rules.

Methods we use routinely

• Survival and time-to-event analysis, including competing risks and time-varying exposure
• Longitudinal and mixed-effects models for repeated measures
• Causal inference: propensity scores, weighting, G-methods, instrumental variables where warranted
• Missing data: multiple imputation, and sensitivity to what the mechanism might really be
• Sample size, power and precision for both trials and observational studies
• Bayesian methods where the borrowing of information is the point
• Prediction and diagnostic model development, validation and calibration

04 — Capability, not dependency

Training & courses

Teaching is a large part of what TruePositive does, and it is deliberately the part designed to make clients need us less.

Courses are hands-on and built around the participants' own setting — their data, their assessment body, their tooling — rather than a generic curriculum. They run in-house for a single organisation, or as open sessions for mixed audiences of clinical researchers, statisticians and market access teams.

See the course catalogue →