Core concept
Evidence-based practice combines:
- the best available research evidence
- clinical expertise
- patient values and circumstances
Evidence does not replace judgement. It improves the information on which judgement is based.
Start with the clinical question
A useful structure is PICO:
- Patient or population
- Intervention
- Comparator
- Outcome
A precise question determines the most appropriate study design.
Study designs
Randomised controlled trial
Best suited to many intervention questions because randomisation reduces confounding.
Limitations can include:
- poor recruitment
- lack of blinding
- crossover
- loss to follow-up
- limited generalisability
- insufficient power
- difficulty standardising surgical expertise
Cohort study
Starts with exposure or treatment groups and follows outcomes. Useful for prognosis, risk factors and uncommon exposures.
Case-control study
Starts with outcome status and looks back for prior exposures. Efficient for uncommon outcomes but vulnerable to selection and recall bias.
Case series
Describes outcomes in a group without a control group. Useful for early experience and hypothesis generation but weak for causal inference.
Systematic review and meta-analysis
A systematic review uses a reproducible method to identify and appraise relevant studies. Meta-analysis statistically combines compatible results.
A meta-analysis is only as credible as:
- the quality of included studies
- the appropriateness of combining them
- handling of heterogeneity and bias
Hierarchy is not enough
A poorly conducted randomised trial may provide less reliable evidence than a well-designed observational study.
Appraise:
- risk of bias
- precision
- consistency
- directness
- follow-up
- outcome selection
- applicability to your patient
Common biases
- selection bias
- performance bias
- detection bias
- attrition bias
- reporting bias
- confounding in observational studies
Internal versus external validity
Internal validity asks whether the study's result is credible for the participants studied.
External validity asks whether the result applies to other patients, settings and surgeons.
Surgical evidence
Orthopaedic trials face particular issues:
- learning curve
- surgeon expertise
- implant evolution
- difficulty blinding
- rehabilitation differences
- long time to important outcomes such as revision
Viva approach
Do not simply recite a level-of-evidence pyramid. Explain why a particular design is appropriate for the clinical question and identify the main threats to validity.
Framing the clinical question
Evidence-based practice starts with a focused question. A useful structure is:
- patient or population
- intervention
- comparator
- outcome
The question determines which study design is most suitable. Therapy questions are often best addressed by randomised trials, prognosis by longitudinal cohorts and diagnostic accuracy by studies comparing an index test with an appropriate reference standard.
Hierarchy and study quality
A hierarchy of evidence is useful but should not replace appraisal. A poorly performed randomised trial can be less reliable than a high-quality observational study for some questions.
Assess:
- selection bias
- allocation and concealment
- blinding where feasible
- completeness of follow-up
- outcome definition
- confounding
- selective reporting
- applicability to the patient in front of you
Randomisation
Randomisation reduces systematic differences between groups by distributing known and unknown confounders by chance. Allocation concealment prevents foreknowledge of the next assignment and is distinct from blinding.
Blinding reduces differential behaviour or outcome assessment after allocation, but it may be impossible for many surgical interventions.
Effect measures
For binary outcomes, understand:
- absolute risk
- relative risk
- absolute risk reduction
- number needed to treat
- odds ratio
Relative measures can make a modest absolute benefit appear dramatic when baseline risk is low. Clinically useful interpretation therefore includes both relative and absolute effect.
Confidence intervals
A confidence interval indicates the precision of an estimate. Wide intervals imply uncertainty. Clinical interpretation should consider whether the interval includes effects that are meaningfully beneficial or harmful, not simply whether a p value crosses 0.05.
Statistical significance versus clinical importance
A statistically significant difference can be too small to matter to patients. Conversely, a clinically important difference may fail to achieve statistical significance in an underpowered study.
Patient-reported outcome measures should be interpreted with concepts such as minimal clinically important difference where appropriate.
Systematic reviews
A good systematic review uses an explicit search strategy, eligibility criteria, quality assessment and reproducible synthesis. Meta-analysis is a statistical technique and is not automatically appropriate when studies are too heterogeneous.
Heterogeneity can arise from:
- patient populations
- interventions
- follow-up
- outcome definitions
- study quality
Guidelines and shared decisions
Evidence must be combined with:
- clinical expertise
- patient values
- comorbidity and risk
- resource context
Guidelines support decisions but do not eliminate individual judgement.
FRCS synthesis
In a viva, do not stop at “level 1 evidence”. Explain whether the study was well designed, whether the effect is precise and clinically meaningful, and whether the study population resembles the patient being treated.