Systematic review an overview and posing the question

1
Systematic review an overview and posing the
question

• Dr. Will Whiteley
• CSO Clinical Academic Training Fellow
• Division of Clinical Neurosciences, University of
  Edinburgh

2
Why review the literature systematically?

• We often need to make sense of multiple studies
  addressing the same question
• We can either pick those studies that most agree
  with our preconceptions or make an unbiased
  assessment of as many of the studies as possible
• The unbiased approach is known as systematic
  review, and when done well can lead to the most
  reliable synthesis of the data.

3
Expert opinion
Cumulative meta-analysis
RCTs - clear benefit
Experts follow
The use of thrombolysis in myocardial infarction
4
Proteus effect
Early extreme contradictory estimates may appear
in published research The Proteus phenomenon in
molecular genetics research and randomized trials
John P.A. Ioannidisa, Thomas A. Trikalinosa
5
Systematic review aims to

• answer a specific question,
• reduce bias in the selection and inclusion of
  studies,
• appraise the biases of the included studies
• to summarise them objectively
• commonly require more time and people than
  narrative reviews

6
Systematic review a medical novelty?
astronomy
Imperial army medicine
quantum physics
psychology, education, criminology, agriculture,
.
7
Sir R.A. Fisher
Karl Pearson
To call in the statistician after the experiment
is done may be no more than asking him to perform
a post-mortem examination he may be able to say
what the experiment died of
8
Advantages and disadvantages of systematic review
to your research projects

• A systematic review of your research question
  will take longer than a simple narrative review
• The results of your systematic review (even if
  in includes a meta-analysis) may be wrong

65
70
Discrepancies between meta analysis and
subsequent randomised controlled trials N Engl J
Med 1997337536-42.
9
However, a systematic review is more likely to be
published and cited than a narrative one!
10
Systematic review overview of process

• Clear question to answer, and pre-specified
  selection criteria
• Clearly defined methods for data extraction to
  remove as much bias as possible from the data
  selection process
• Comprehensive search of the published (and if
  time available unpublished) literature to
  ensure negative studies are found which might
  not be published in the top tier journals
• Duplication of data extraction and study
  selection
• Summary estimation if sufficient high quality
  studies are found

11
Posing the research question
In neurologically asymptomatic people, what is
the prevalence of imaging findings on high
resolution brain MR images?
e.g.

• Patients
• Intervention
• Control
• Outcome
• Decide what type and quality of studies you will
  accept this will be different for an established
  treatment to a new biomarker of disease
• A good question will help to define your search
  strategy and study selection

In patients within one week after ischaemic
stroke, what is the prognostic value of IL6 for
disability or death?
In patients within 48 hours of subarachnoid
haemorrhage, what is the value of oral
simvastatin compared to control for prevention of
radiological vasospasm?
12
Data extraction

• Decide prior to looking at the literature what
  data you will need to answer your question
• What is likely to contribute to study related
  heterogeneity
• Design a form either paper or in the database
  program of your choice
• Make sure it is attractive and easy to fill in
• Go through a few iterations of design
• Data extract in duplicate (choose a helpful
  colleague)

13
Measuring bias
Most study designs have potential for bias that
can be minimised by taking appropriate
precautions Quality is a measure of this
bias Set a quality threshold for inclusion
e.g. prospective cohort studies
only or Investigate effect of quality on outcome
Quality measures are available for a number of
study designs Prospective observational studies
MOOSE Diagnostic studies STARD Randomised
controlled trials QUORUM Genetic
associations HuGE reviews etc. etc.
14
In summary

• Systematic review aims to reduce bias in
  assessing literature
• Select you question carefully make sure it is
  answerable
• Pre-specify your methods of study selection and
  data-extraction
• Perform your search strategy .