Key Takeaways

  • 1

    Start with the primary outcome

  • 2
    Look at effect size, not just p-values
  • 3

    Use the confidence interval to judge precision

  • 4

    Prioritise patient-important surgical outcomes

  • 5
    Ask if the result is practice-changing
Pearl Information
General Intro to Research Research Methods Research Pearls

A 4-question Approach for Surgical Research

1. What was the main outcome?

Before looking at the p value, ask:

  • What was the primary outcome?
  • Was it patient-important?
  • Was it clinically meaningful?

In surgery, prioritise:

  • Mortality
  • Major complications
  • Re-operation
  • Readmission
  • Leak/SSI/conversion

→ Pearl: If the outcome doesn’t matter, the result doesn’t either.

2. How big was the effect?

Look at the effect size, not just whether it was “significant.”

Focus on:

  • Risk ratio / Odds ratio / Hazard ratio (RR / OR / HR) 
  • Absolute risk reduction
  • Mean difference

Example:

A “30% relative reduction” sounds impressive…

But if complications fall from 10% → 7%, that’s only a 3% absolute reduction.

Pearl → Big relative effects can hide small absolute benefits.

3. How certain is the result?

Look at the 95% confidence interval (CI)

A CI tells you:

  • how precise the estimate is
  • whether the result is compatible with:
    • benefit
    • no effect
    • harm

Quick rule:

  • Narrow CI = more precise
  • Wide CI = more uncertainty
  • If the CI crosses:

    • 1 for RR / OR / HR
    • 0 for mean difference

Then the result is usually not statistically significant.

Pearl → A “positive” paper with a wide CI still deserves scepticism.

4. Would this actually change practice?

Ask the important clinical questions:

  • Is the benefit big enough to matter?
  • Are the outcomes patient-important?
  • Is it reproducible outside one expert centre?
  • Was the study powered for major complications, or just surrogates?
  • Would I trust this in real surgical practice?

A paper can improve:

  • LOS
  • operative time
  • “technical success”

…but still not improve outcomes that matter.

Bottom line: Statistical significance ≠ clinical significance