What 273 studies agree on

A single preclinical nanomedicine study is close to unfalsifiable. Its model, dose, schedule and endpoint are its own whereas the next study might change all four. In consequence, the literature accumulates results that cannot be stacked. Therefore, the only way to properly ask whether an idea works is to pool its repetitions and accept that pooling costs precision.

A 2025 meta-analysis in Nature Nanotechnology utilized the pooling approach for combination nanomedicine: two drugs delivered by nanocarriers rather than one. Seven hundred and forty-two manuscripts were screened down to 273 preclinical studies and the funnel is as much the method as the statistics that follow.

The effect sizes

Against the three obvious comparators, combination nanotherapy reduced tumour growth by a further 42.6 percent versus a single free drug, 30.0 percent versus single-drug nanotherapy, and 29.1 percent versus a free-drug combination. Survival moved with it: 16 percent of combination studies achieved full cohort survival, against 2 percent for single-drug nanotherapy and none at all for either free-drug regimen.

The finding easiest to skim past is the one a formulator would act on. Putting both drugs inside the same particle beat giving two separately loaded particles by a further 19 percent, measured as a paired comparison within studies that ran both arms. Same drugs, same carriers, same experimental setting; the only variable is whether the two travel together. That turned out to be a formulation decision with a noticeable effect size attached to it.

Interestingly, the article also identified a subgroup where the difference is categorical rather than incremental. In drug-resistant tumour models, combination nanotherapy was the only regimen to produce a statistically significant response at all.

The negative results are the most informative part

Two findings run against expectation, and they are the ones I am inclined to trust most - as emergent properties, nobody set out to demonstrate them.

PEGylation, near-universal practice in the field, showed no statistically significant added value for combination nanotherapy outcomes. The one approved product in this class is itself deliberately non-PEGylated.

Secondly, active targeting outperformed passive targeting specifically for combination nanotherapy, a distinction that did not appear for single-drug formulations at all.

A meta-analysis that confirmed what the field already believed might have confirmed its own selection criteria instead. These two did not.

What anchors it clinically

Vyxeos, a liposomal formulation of daunorubicin and cytarabine at a fixed 5:1 ratio, is the field’s only approved multi-drug nanomedicine. In a Phase III trial against standard free-drug therapy, it raised median overall survival from six months to ten while at the same time enabling a lower cumulative drug dose. The approval is narrower than the disease: acute myeloid leukaemia with myelodysplasia-related changes, and therapy-related AML. One clinical case cannot validate 273 preclinical ones, but it points into the same direction from an entirely different kind of evidence.

The authors are careful about what they cannot fix. Publication bias is acknowledged rather than argued away, with the observation that it appears to affect single-drug and multi-drug literature symmetrically, so the comparison between them survives better than any absolute number does. It should also be considered that the preclinical models metabolise faster and clear nanoparticles sooner than humans do, so the margins measured there are not expected to transfer one to one.

What the analysis really demonstrated is the value of pooling in a field that rewards the spectacular single result. An individual study would not necessarily convince anyone that co-encapsulation beats co-administration by about a fifth. Two hundred and seventy-three of them, compared within experiments rather than across them, make a claim that a formulation scientist can plan around, which is a different and more useful kind of finding than a headline.