With an appropriate post hoc statistical test we can reduce the...

With an appropriate post hoc statistical test we can reduce the risk of type I error (this error is inflated in multiple comparison between several subgroups) and also obtain acceptable statistical power, which of course depends on the sample size and data variance in subgroups. Brief, choosing an appropriate multi comparison test and having high number of data points in subgroups and low variance make the post hoc statistical analysis scientifically valid.

I found below a good summary of what we need to consider in post hoc analysis. In our case showing cells are more efficient in treatment of patients with higher inflammation (e.g, higher IL-6) is highly plausible and correlated with the proposed mechanism of action.

“Reporting of treatment effects on subgroups of patients are common in the medical literature. Although knowing how well a new treatment works in patients with a specific biomarker or a specific combination of disease stage and histology are important for the patients and their clinicians to make informed treatment decisions, it is important to be cautious when interpreting the results of subgroup analyses. Common challenges with subgroup analyses include poor definitions, low statistical power, and inflated type I error due to multiple hypotheses testing. The decision to conduct a subgroup analysis should depend biological justification or on evidence of treatment heterogeneity from existing preliminary data. When a large number of unplanned subgroup analyses are conducted, the treatment may spuriously appear to be effective in one or more subgroups. One should always assess the validity of subgroup analysis results based on biological plausibility, sample size for the subgroup, proper type I error control and power.”