Similar starting averages concealed different tumour-size groups in a reanalysis of a mouse pancreatic-cancer experiment. A treatment contrast appeared among animals with smaller initial tumours that was unclear in the pooled results. The peer-reviewed study also examined public experiments, highlighting baseline distributions as a factor in interpreting laboratory outcomes. Its retrospective subgroup findings concern mouse models and do not establish treatment benefit in patients.
A reanalysis of a pancreatic-cancer experiment in mice found that comparable starting averages concealed clusters of smaller and larger tumours. The peer-reviewed study examined a 28-day experiment using human tumour cells grown in mice. At the outset, tumour volumes spanned roughly 80–200 cubic millimetres. Only starting measurements determined the clusters; outcomes after treatment played no part in that classification.[1]
Among mice with smaller starting tumours, antibody-treated tumours stayed near baseline while controls grew to about three times their starting size. Mice with larger initial tumours did not show the same clear contrast, and pooling the animals obscured the smaller group's result. This retrospective subgroup comparison does not independently establish antibody efficacy. It concerns growth inhibition in a laboratory mouse model, with no demonstrated clinical benefit in people.[1]
The authors classified 43 of 62 eligible public experiments as having heterogeneous baseline distributions. In 11 of 19 experiments allowing treatment-response comparisons, pooled and stratified assessments disagreed for at least one arm. These datasets had different designs, rather than replicating one drug effect. The paper disclosed author relationships with Comed Therapeutics and proposed defining baseline strata prospectively, with independent prospective validation needed to assess that approach.[1]
