Are we removing the reasons drugs fail from our experiments?

In experiments, we standardise how cells are grown, the dose of a stimulus and the time of measurement. That makes drug effects easier to compare.

But if differences in patients' responses arise from surrounding inflammation or the state of their tissues, could standardising the conditions also remove the very differences we want to understand?

In the placental study covered in this issue, developmental effects were not uniform even among fetuses growing in the same mother. Affected fetuses also had changes in their placentas. Differences emerged between fetuses that could not be explained simply by their sharing the same mother. Original paper ↗

We can ask a similar question about drug evaluation using patient-derived cells. Cells from different patients have different properties. But placing them under the same culture conditions does not necessarily preserve the conditions that made those patients respond differently to a drug in the body.

The distinction here is between reproducible measurement and uniform responses. We need to minimise variation in the procedure; we do not need every sample to respond in the same way. If a difference appears repeatedly, it becomes something to investigate in its own right.

One possible future experiment would be to test drugs in fibrosis models from several patients while varying the conditions of inflammatory stimulation. Even if the average effects were similar, a candidate with a small effect in many samples would call for a different development strategy from one with a large effect in a subset. For the latter, identifying the conditions associated with a response could help researchers decide which patients to include in the next trial.

This would also change how non-responding samples are treated. Was there a problem with the culture or measurement, or did the experiment capture a genuine difference in patient response? Without distinguishing the two, an important clue could be discarded as an experimental failure.

Business value could emerge from accumulating the knowledge needed to make that distinction. Researchers would link responses in experimental models with patients' treatment outcomes, then test whether the relationship holds in other patients. At that point, a service that measures drug effects could expand into one that helps decide which patient populations a candidate should be developed for.

The useful result would be not only that a drug worked, but the conditions under which it stopped working. Could the ability to explain that boundary be what makes the data valuable for the next development decision?