Why can a prognostic biomarker such as mitotic count, KIT immunolabeling pattern, and Kiupel grade appear predictive in univariable analysis?

• A. Some adjunctive-treated low-grade MCTs that appeared to have a good response to treatment may not have progressed regardless of whether treatment was given
• B. Prognostic biomarkers always predict outcomes in any analysis
• C. They reflect tumor size alone
• D. They measure the mitotic index but do not inform treatment decisions

Answer: (A)

The main idea is that a biomarker can look predictive in a simple, single-variable analysis even when its apparent effect is really driven by treatment choices or other confounding factors. For prognostic markers in canine mast cell tumors—mitotic count, KIT labeling pattern, and Kiupel grade—these features are linked to tumor biology and help predict outcomes in the overall, untreated population. But when you analyze one factor at a time, some tumors with these biomarker features receive adjunctive therapy and appear to do well, not necessarily because the biomarker itself changed the course, but because the treatment (or the natural history of those tumors) influenced the result. In other words, the good outcome might occur regardless of treatment for a subset of tumors, making the biomarker seem predictive in univariable analysis even though it isn’t independently prognostic once you adjust for treatment and other factors. That’s exactly what the option describes.

To truly know if a biomarker independently affects prognosis, you need analyses that account for treatment and other variables (multivariable models or randomized designs).