Understanding evidence

Does this study actually answer my question?

A strong study can answer a different question from the one you need to decide.

With a rare cancer, much of the evidence may come from adjacent diseases or a broader mix of patients. That evidence can still matter. The important step is to make the transfer explicit.

Check the match before the headline

  1. Who was studied? Cancer site, tumor type, stage, biomarkers, and prior treatment.
  2. When was treatment given? Before surgery, after surgery, or for disease that cannot be removed?
  3. What was compared? Adding a drug to a combination does not test replacing the combination with that drug.
  4. What was measured? Tumor response, recurrence, disease-free survival, overall survival, and quality of life answer different questions.
  5. How was the study designed? Randomization and an appropriate comparison help separate treatment effects from differences between the patients who receive treatment.

Two tempting shortcuts

“The new drug helped, so the older drugs are unnecessary.”

That conclusion needs a study that actually tests leaving out the older drugs. In ATOMIC, both groups received chemotherapy. The trial tested the addition of atezolizumab in resected stage III dMMR colon cancer. It cannot isolate what atezolizumab alone would do after appendix surgery. [1]

“The result was not statistically significant, so there is no benefit.”

A study can be too small or too imprecise to distinguish a useful effect from no effect. Look at the estimate and its confidence interval. The reverse also matters: a statistically significant association in an observational study does not by itself establish that treatment caused the difference.

Ask for the absolute difference

A relative effect is not the same as the number of people helped. Ask for outcomes over the same period in both groups, the uncertainty around those estimates, and the harms. Do not interpret a hazard ratio as a simple percentage-point reduction in your personal risk.

The question behind the number

“Out of 100 people sufficiently similar to me, how many might benefit from this option, over what period, and how uncertain is that estimate?”

Separate evidence strength from case fit

A well-run colon-cancer trial may give a reliable answer for its population while still leaving uncertainty for appendix cancer. A small appendix cohort may be closer to the diagnosis but more vulnerable to bias.

I want to know both: how trustworthy is the result, and how much work is required to apply it to my case? Population statistics provide context; they cannot determine one person’s outcome. [2]

Sources & scope

  1. Atezolizumab plus FOLFOX for Stage III Mismatch Repair–Deficient Colon Cancer ↗ATOMIC investigators · New England Journal of Medicine. Randomized postoperative colon-cancer trial. Not an appendix-cancer or immunotherapy-alone trial.
  2. Understanding cancer prognosis ↗National Cancer Institute. Why group statistics cannot predict one person’s outcome.

Research links checked November 2026. Patient-written interpretation; not a clinical review.

Keep readingCompare pathology and doctor perspectives