In this paper we report a failed attempt to validate the Barbare's hypothesis  that tamoxifen might be effective in patients with better prognosis and confirm the previous evidence that there is no survival advantage for treating HCC patients with tamoxifen in addition to supportive therapy.
Because of the broad eligibility criteria applied in the CLIP-1 trial, we had already tested possible survival differences within subgroups defined by locoregional treatment . Because of the strict correlation between prognosis and locoregional treatment, the present lack of significance was expected.
We also tested the same hypothesis in subgroups defined according to the CLIP score, that currently is the most widely accepted and validated prognostic score for HCC [11, 12]. CLIP scores 0 and 1 definitely identified patients without major hepatic insufficiency better than Okuda score (Figures 1a and 2a), but once again no evidence of tamoxifen effectiveness was found.
Our results substantiate the criticism against subgroup analyses, even more urgently when they are not previously planned. In multiple subgroup analyses both more false negative results, due to the smaller size of the groups, and more false positive results, due to multiple comparisons, are expected. Exploring data in search for differences possibly depending on certain baseline characteristics may be acceptable on its own, but subgroup findings must not be over-interpreted . Although the power is usually limited, only statistical tests of interaction should be performed rather than subgroup-specific tests. Anyway, the interpretation strictly depends on whether findings are biologically plausible, how many analyses are done, whether they are planned ahead and how much strong is the statistical evidence, still being aware that most subgroup claims tend to exaggerate the truth [13, 14].
Often, it is enough to correct for multiple testing so that results are no more statistically significant and conclusions not sounded; as would have been the case in the trial recently reported by Barbare et al .
Although Barbare et al. discussed their findings must be regarded with great caution, their conclusion was that new trials are warranted in the specific population without major hepatic insufficiency. That conclusion brought about this report, but we failed to substantiate it.