How to Read a Clinical Paper
Giovanni L. Pappagallo
To read a clinical paper…
HOW
was planned
(Internal Validity)
Material and Methods Section
HOW
was
conducted
(External Validity)
Results
Section
Research Question − Design − Endpoint − Statistics
Internal Validity
F.I.N.E.R.
A strategy for assessing
the value of a reserach question
FEASIBLE INTERESTING
NOVEL
ETHICAL
RELEVANT
Research Question − Design − Endpoint − Statistics
Internal Validity
Research Question − Design − Endpoint − Statistics
Internal Validity
Population (P)
Interventions (I) or Exposures (E) Outcomes (O)
Centre for Evidence Based Medicine, Oxford, UK www.cebm.net
Research Question − Design − Endpoint − Statistics
Internal Validity
Population (P)
Interventions (I) or Exposures (E) Outcomes (O)
Centre for Evidence Based Medicine, Oxford, UK www.cebm.net
Research Question − Design − Endpoint − Statistics
Internal Validity
Population (P)
Interventions (I) or Exposures (E) Outcomes (O)
Centre for Evidence Based Medicine, Oxford, UK www.cebm.net
Research Question − Design − Endpoint − Statistics
Internal Validity
Research Question − Design − Endpoint − Statistics
Internal Validity
The Iniparib Story
Phase II (NEJM) Phase III (J Clin Oncol)
Research Question − Design − Endpoint − Statistics
Internal Validity
Population (P)
Interventions (I) or Exposures (E) Outcomes (O)
Centre for Evidence Based Medicine, Oxford, UK www.cebm.net
Research Question − Design − Endpoint − Statistics
Internal Validity
Randomization minimizes allocation bias, balancing both known and unknown prognostic factors in the assignment of treatments.
Research Question − Design − Endpoint − Statistics
Internal Validity
Research Question − Design − Endpoint − Statistics
Internal Validity
Research Question − Design − Endpoint − Statistics
Internal Validity
M. Di Maio, this afternoon
Research Question − Design − Endpoint − Statistics
Internal Validity
Research Question − Design − Endpoint − Statistics
Internal Validity
Do control statistical errors within conventional limits (5% for type I error, and 20% for type II error, respectively) by sample size calculation…
Confidence − Directness − Relevance
External Validity
Confidence − Directness − Relevance
External Validity
Sources of bias in clinical trials
Confidence − Directness − Relevance
External Validity
Confidence − Directness − Relevance
External Validity
Confidence − Directness − Relevance
External Validity
Subject at low risk of developing CRC
Consumption of red meat
NO consumption of red meat
Diagnosis of CRC
No analysis by risk (of CRC) group = indirectness for the
present clinical question
Confidence − Directness − Relevance
External Validity
• Statistical Significance
“Is an observed difference likely to be real”
dependent on the magnitude of the number of patients and/or the magnitude of the difference NOT on whether the difference is meaningful for patients
• Clinical Significance
“Is an observed difference likely to be meaningful for patients”
dependent on the magnitude of the difference NOT the number of patients
M. Di Maio, this afternoon
How to Read a Clinical Paper
Giovanni L. Pappagallo