Back to Insights

The second half of that title is the more useful half.

In January 2026, FDA investigators published a retrospective analysis of 22 randomized trials in metastatic non-small cell lung cancer finding tumor growth rate inversely associated with overall and progression-free survival. The authors concluded it may provide an additional or supplementary approach to assessing potential clinical activity.

Those two sentences are the whole of what can be responsibly said about the finding, and they are worth reading carefully, because a good deal of what is now being said about tumor growth rate goes further than they do.

What the Analysis Is

Tumor growth rate modeling treats change in tumor burden as continuous parameters fitted across time points rather than as a category assigned at each assessment. Rather than sorting a patient into complete response, partial response, stable disease or progressive disease, it estimates the underlying rates of growth and decay in the measured disease.

The method is not new and it is not proprietary to anyone. Growth and growth-inhibition modeling is established practice in pharmacometrics and model-informed drug development, the literature is public, and reference implementations are open source. What is new is the source and the scale of the analysis — a regulator's own investigators, on pooled data from randomized trials.

What It Does Not Establish

Four things, each of which is being blurred somewhere in the market right now.

Not policy An analysis authored by agency investigators is not agency policy, and nothing about its publication makes tumor growth rate validated, endorsed, accepted, backed or qualified by the FDA. Any vendor using those constructions is telling you something about their claim discipline generally.
Not an endpoint The paper is explicit that the threshold for a meaningful change in growth rate has not been defined. Without a defined threshold there is no decision rule, and without a decision rule there is no endpoint — there is an exploratory measure that can support an argument.
Not a replacement It does not replace RECIST. It complements it. Regulators still expect the RECIST call, and a sponsor who arrives with growth rate instead of a response assessment has not strengthened their position.
Not general It does not generalize beyond what was studied. The evidence base here is metastatic non-small cell lung cancer. That is a large and important population, and it is not every population.

Why It Matters for Early-Phase Sponsors Anyway

Because of what the authors say about where response rate is least informative.

The analysis notes that an endpoint such as objective response rate may not fully capture a drug's effect on tumor growth rate, and that this is particularly relevant where a therapeutic agent is not expected to produce a high response rate — a later-line setting, or an agent with a primarily cytostatic effect.

Read that as a qualification criterion rather than a talking point. If your asset works primarily by holding disease stable, or your trial is later-line, or a single-arm study is carrying your efficacy argument, the endpoint you are being judged on is describing a different kind of drug. Not occasionally, at the margin, in an unlucky readout. Structurally, at every readout.

That is a materially different problem from the near miss everyone talks about, and it does not go away with a larger cohort.

The Data Requirement Is Lower Than Most Sponsors Assume

This is the finding with the most practical consequence and the least attention.

The FDA analysis included every patient with at least two tumor measurements. Not a long longitudinal series. A baseline and a follow-up.

Qualifier One qualifier has to travel with that number, and any source that gives it to you without the qualifier is overselling. Two measurements is the threshold to attempt a fit, not a guarantee of getting one. A meaningful minority of patients produce no growth rate at all — some fit no model, and some fit pure decay, which is arguably the best news in the dataset and still yields no growth constant. A design that treats a missing growth rate as a failure has misunderstood the output.

One Argument From the Commentary Worth Carrying Forward

An invited commentary on the analysis makes a point that goes further than the analysis itself: summarizing disease as a single aggregate burden implicitly assumes survival is driven by the average lesion, when in practice individual aggressive lesions drive outcomes.

That is an argument for reporting each lesion's own trajectory alongside the aggregate, and it is not a technology argument. The reader already measured every target lesion at every time point. The aggregation is what discarded the information.

Where This Should Be Delivered

The method is public, published and implementable by anyone. It is nobody's moat, including ours. So the question that actually matters to a sponsor is not who has the mathematics. It is where the output lands.

Today this analysis is typically a separate workstream, run by a different team, after the read is complete — or a whitepaper with a download form on it. That is not a criticism of anyone doing the work. It is a consequence of the analysis living outside the read environment.

Inside the read, alongside the RECIST call, in the same audit trail, signed by the same radiologist, at the time point where it would change a decision, is a different thing from the same numbers arriving six weeks later in a slide deck.

Response analytics — continuous depth of response, distance to the partial response threshold, and lesion-level trajectories — is in development at GenPhase and is being built into the read output alongside the RECIST call.

Sources: Malinou JN, Fan J, Cheng J, Gong Y, Shen Y-L, Larkins E. An FDA analysis of the association of tumor growth rate, overall survival, and progression-free survival in patients with metastatic non-small cell lung cancer. The Oncologist 2026;31(3):oyag009. Goldmacher G. Tumor growth rate as an intermediate trial endpoint. The Oncologist 2026;31(6):oyag149.