The phase 3 Avastin in Melanoma (AVAST-M; ISRCTN81261306) trial showed that administering adjuvant bevacizumab (Avastin®) to patients with resected melanoma did not improve overall survival and yielded only a small improvement in disease-free survival.1,2 (Roche Pharmaceuticals, the maker of the drug, provided free bevacizumab during the trial.)
Now the authors of a new study analyzing data from the same trial are claiming that the drug may improve disease-free and overall survival in younger, but not older, patients because of certain aging-related changes in angiogenesis.3 The authors also suggested that the drug may not work in older patients because vascular endothelial growth factor (VEGF) — the target of bevacizumab — decreases with age, while secreted frizzled-related protein 2 (sFRP2) takes over as the primary driver of angiogenesis.
However, experts not involved in the study pointed out that the findings about potential interaction between age and treatment were not statistically significant, and that a separate clinical trial would be needed to determine whether administering the drug could be beneficial in younger patients with melanoma. (“Interaction” is what occurs when the effect of a treatment is significantly different between different subgroups.4 One example of interaction would be if a medication were significantly more effective in women than in men.)
“They did not find a significant interaction between age and treatment on the outcome of overall survival, and they did not find a significant interaction between age and treatment on the outcome of disease-free survival,” Kristin Sainani, PhD, associate professor of epidemiology and population health at Stanford University in California, who was not involved in the study, wrote in an email to Cancer Therapy Advisor. “So, in both cases, you cannot say that the effect of treatment differs significantly by age group.”
Dr Sainani thinks that the authors committed what she called a “common statistical fallacy,” which occurs when investigators attempt to compare statistical significance between groups instead of comparing effect sizes.4 The fallacy centers on the incorrect belief that showing that a treatment’s effect is significant in one subgroup but not in another is enough to prove an interaction. The correct assumption is that an investigator needs to prove that a significant difference exists in the effects between different subgroups.
While perusing the paper, Dr Sainani spotted a P value of .32 for heterogeneity between 3 different age groups lurking at the bottom of a figure. “Heterogeneity” is a different term for interaction, Dr Sainani explained.
“The test of heterogeneity starts with the null hypothesis that the effects in all 3 groups are equal,” she said. “A P value of .32 means that we do not have evidence to reject this null hypothesis. In other words, the differences between the groups are within the range that could easily arise by chance.”
The number 1 step that should be taken in an analysis like this, she said, is to first establish whether interaction, or heterogeneity, is significant. Only if it is indeed significant — and in this case, it was not — should researchers proceed to testing any within-group effects.