Building Defensible RWE for Payer Submissions

Real-world evidence has moved from a supplementary data source to a central pillar of payer coverage decisions. Formulary committees, pharmacy benefit managers, and health technology assessment bodies now routinely require RWE packages that demonstrate clinical and economic value in real-world populations — not just in the controlled conditions of a randomized trial.

The problem is that not all RWE is created equal. Payer reviewers have become increasingly sophisticated about the methodological limitations of observational research. A poorly designed RWE study — one that does not adequately control for confounding, uses an inappropriate comparator, or makes unsupported assumptions about generalizability — will not just fail to support coverage. It will actively undermine the credibility of the evidence package.

This post describes what separates RWE packages that succeed from those that get rejected.

The most common mistake in RWE development is starting with the available data and working backward to a research question. The right approach is the opposite: start with the coverage decision you are trying to support, define the evidence requirements for that decision, and then identify the data sources that can generate that evidence.

This sounds obvious, but it has significant practical implications. It means engaging with payer medical directors early — before the study is designed — to understand what evidence they need and what methodological standards they will apply. It means designing the study to answer the payer's question, not the question that is easiest to answer with available data.

In comparative effectiveness research, the choice of comparator is the most consequential methodological decision. The comparator should reflect the actual treatment alternative that payers are evaluating — not the comparator that makes your therapy look best.

Payer reviewers are acutely aware of comparator manipulation. A study that compares a new therapy to an outdated standard of care, or to a suboptimal dose of a competitor, will be dismissed regardless of how sophisticated the statistical methods are. The comparator must be clinically appropriate and reflect current treatment practice in the population of interest.

Propensity score methods, inverse probability weighting, and instrumental variable analysis are powerful tools for controlling confounding in observational data. But methodological sophistication is not the same as methodological credibility.

Payer reviewers want to understand what confounders were controlled for, why those confounders were selected, and what residual confounding might remain. A study that uses a complex weighting approach without clearly explaining the confounders it addresses — and the ones it cannot address — will raise more questions than it answers.

The standard we apply at Bracy Analytics is transparency over sophistication. Use the simplest method that adequately controls for the relevant confounders. Document every methodological decision. Report sensitivity analyses that test the robustness of the results to different assumptions. Make it easy for reviewers to understand what the study can and cannot conclude.

Randomized trials are criticized for their limited generalizability to real-world populations. RWE studies face the opposite criticism: they reflect the real world, but which real world? The population in a claims database may differ systematically from the population a payer covers.

Addressing the generalizability question requires explicit characterization of the study population — who is included, who is excluded, and how the study population compares to the payer's covered population. It may require stratified analyses that show results are consistent across relevant subgroups. It may require sensitivity analyses that test the impact of different inclusion/exclusion criteria.

This is not a methodological nicety. It is a core requirement for a credible payer evidence package.

Health economic models — cost-effectiveness analyses, budget impact models — are only as credible as their assumptions. Payer reviewers will scrutinize every assumption: the time horizon, the discount rate, the utility values, the cost inputs, the treatment effect estimates.

Every assumption should be sourced from published literature or primary data, with clear documentation of the source and the rationale for the choice. Assumptions that favor the therapy under review should be accompanied by sensitivity analyses that test the impact of more conservative assumptions.

A model that produces favorable results only under optimistic assumptions is not a credible evidence package. A model that produces favorable results under a range of assumptions — including conservative ones — is.

Building defensible RWE for payer submissions requires a combination of methodological rigor, clinical knowledge, and strategic thinking about what payers need to make coverage decisions. It is not a purely technical exercise.

The organizations that do this well engage payers early, design studies to answer payer questions, apply transparent and appropriate methods, and document every decision with the rigor of a regulatory submission. The result is evidence that payers can trust — and that supports the coverage decisions that make therapies accessible to the patients who need them.