How to Read a Paper''' In 10 minutes

1
How to Read a Paper... In 10 minutes

• Adam Hardie, ST3
• Hot Topic
• September 2009

2
Learning outcomes

• By the end of the session you should
• Grasp EBM basics and define it
• Be able to formulate a structured clinical
  question
• Appreciate the hierarchy of evidence
• Understand the systematic assessment of research
  articles using a structured approach
• Learn (hopefully something) about basic
  statistics

3
Background The Source

• Trisha Greenhalgh, 2006
• Professor of Primary Care, London
• Basics of EBM
• Excellent book with good references

4
EBM Definition

• The use of mathematical estimates of the risk of
  benefit and harm, derived from high-quality
  research on population samples, to inform
  clinical decision making in the diagnosis,
  investigation or management of individual
  patients

5
The cynics definition

• The argument... that no health-related action
  should ever be taken by a doctor, nurse or
  manager of health services, or a politician,
  unless and until the results of several large and
  expensive research trials have appeared in print
  and been approved by a committee of experts

6
Common pitfalls in medical decision making

• Decision-making by anecdote
• Human nature to some extent
• Trials of n1
• Decision making by press-cutting
• Basing decisions on abstracts alone
• Consensus statements
• Experts consensus the old guidelines
• Decision by cost minimisation
• Cheap interventions are not necessarily the best
• Converse is also true e.g. Cancer drugs

7
Clinical questions

• PICO
• Population
• Intervention
• Comparison
• Outcome
• In a 72 year old woman with osteoarthritis of the
  knee, can COX-2 inhibitor use decrease the risk
  of GI bleeding compared with other NSAIDs?
• In patients who have had a stroke, does repeated
  encouragement about exercise result in a
  sustained improvement in their activity levels
  compared to standard therapy?

8
The Hierarchy of Evidence
9
Research paper trashing

• 33- 99 of research in the journals is
  methodologically flawed
• Reject a paper before reading the results
• Three preliminary questions
• What is the research question and is the study
  needed? Often lacking.
• Is there a null hypothesis?
• Has the question already been answered?
• What was the research design?
• Was it appropriate to the research question?
• If it was not an RCT, should it have been?

10
Assessing methodological quality to bin or not
to bin?

• 5 further questions
• Was the study original?
• A piece of the jigsaw
• More powerful? More robust methodology?
• Who is it about?
• Will it affect your patients?
• Is the design sensible?
• What is the intervention?
• What endpoints were used?
• Was bias avoided?
• Were basic statistical questions addressed?
• Size of the sample
• Duration of follow up
• Completeness of follow up drop out rate

11
Statistical analysis

• The average human has one breast and one
  testicle.  Des McHale
• Initial considerations
• Are the two groups similar
• Is the data parametric or non-parametric?
• Have appropriate tests been used?

12
(No Transcript)
13
Other statistical considerations

• Correlation and regression
• Correlation does a relationship exist?
• Regression is there an equation whereby x can
  be calculated from y?
• One- and two-tailed tests
• One tailed tests deal with the extremes at one
  end of the scale only

14
(No Transcript)
15
Diagnostic tests

• Sensitivity a/ac
• Specificity d/bd
• Positive predictive value a/ab
• Negative predictive value d/cd
• Likelihood ratio Sensitivity/(1-specificity)

16
Meta-analyses

• Forrest plots

17
Meta-analyses

• Funnel plot
• Publication bias

18
Questions and comments