Post on 02-Jan-2016
Systematic Reviews
By Jonathan Tsun & Ilona Blee
What is a systematic review? Define it!
Results from similar randomized trials are brought together (“synthesized”) to help identify which forms of health care work, which don’t work, and which are harmful.
This combination of randomized trials needs to be done in a systematic way – hence it is called a systematic review
The methods used include steps to minimize bias in the process – this is done by including all of the relevant studies
What are the TWO main types of bias?
1. Selection
2. Information
What are the limitations of systematic reviews?
“Synthesis” of results from different studies can oversimplify important distinctions
Different conclusions can be reached depending on the wording of the “review question”
For example compare these two questions: to be able to know whether we can predict a problem and intervene early vs. to identify the best ways of diagnosing/identifying problems
Reviews can make it difficult for doctors to apply the results of the review to every situation they come across over-generalisation, medicine isn’t an exact science!
Findings from systematic reviews aren’t always consistent with other findings of large-scale high quality trials
What is a meta-analysis?
When the results of individual studies are combined to produce an overall statistic.
This can be done without a systematic review by combining results from more than one trial
What is the disadvantage of this?
Without the systematic review, there will be bias present if there is any from that study
This can lead to clinically misleading results!!!
Forest plot – what is it? A graphical representation of a meta analysis,
illustrating the effect sizes of individual studies and the meta-analysis as a whole.
What is “effect size”? Quantitative measure of the strength of a
phenomenon (e.g. correlation between 2 variables, etc)
It will have a list of the studies included in the meta analysis
What is the line at x=0 called? What does it indicate? Line of No Effect
If a study’s confidence interval line crosses the Line of No Effect, it demonstrates that at the given level of confidence, the study’s effect size does not differ from no effect. (AKA the study’s findings are not significant)
What does the size of the green square indicate? It is a graphical representation of the weight of that
particular study in the meta-analysis.
What do the lines sticking out from the green boxes indicate? They are graphical representations of the size of confidence
intervals.
What does the diamond represent? The effect size of the meta-analysis as a whole
What do the 2 horizontal points of the diamond indicate? What does it mean if it crosses x=0? The 2 horizontal points indicate the confidence interval for the
combined data
If it crosses the line of no effect... Answer is below in the notes.
Funnel Plot – what are they/ what do they do?
They assess whether the results of a systematic review have been influenced by “publication bias”
Publication bias occurs when the publication of research results depends on their nature and direction
eg. Positive results are more likely to be published than negative studies and can mean systematic reviews end up biased towards positive results
If the plot is symmetric, this means that there is probably no publication bias
If the plot is asymmetric, this means that publication bias is likely
Symmetrical = no publication bias!
Asymmetrical= publication bias likely
If we were to critically appraise a study, what
would we ask? (3)1. Is the study valid?
2. Are the results reliable?
3. Can I generalise from this study to my workplace/own practice? OR will the results help locally?
Why is critical appraisal important?
Ensures that relevant articles contribute towards patient care
Tries to ensure that bias is minimised
Ensures that the results from a study can be used in real life
An example of evidence-based medicine!
CURB-65
This is something you should have covered in CBL and comes up in exams/ on the wards etc.
What does it stand for?
Confusion
Urea levels of >7mmol/L
Respiratory rate >30 breaths/min
Blood pressure <90mmHg systolic over <60mmHg diastolic
Age >65 years
You get a point for each of these
What does PRISMA stand for?
Preferred Reporting Items for Systematic reviews & Meta-Analyses
In other words, helps authors report the results of a systematic review
More at: http://prisma-statement.org/
Using systematic reviews!
Where do we search for systematic reviews?
1. PICO! Population/patient problem, intervention, comparator, outcome
2. Cochrane library
3. Medline
What is PICO used for?
Asking a “good” question
What are the steps for using PICO/how do we use PICO?
How to explain this ..
Don’t forget to learn your equations and calculations!
Weighted mean difference: Mean(g1) – Mean(g2)
Standardised mean difference: Mean(g1) – Mean(g2)
pooled sd
Relative risk:
Number needed to treat (NNT):
Risk Ratio:
Be able to interpret risk ratio quite quickly
E.g. RR = 2
Means the intervention outcome is twice as likely compared to the control outcome
E.g. RR = 0.5
Means the intervention outcome is half as likely compared to the control outcome
What is bias? Any systematic error in an epidemiological study
which either underestimates or overestimates the parameter of interest because of a deficiency in the design or execution of the study.
Selection Information
Sampling bias Recall bias
Allocation bias Observer/interviewer/reporting bias
Responder bias Social desirability bias
Attrition bias
Publication bias
Healthy worker effect
Hierarchy of evidence!
1. Systematic reviews – always at the top over anything ever because of the way it minimises bias!
Meta-analyses
Non meta-analytic reviews
2. Experimental studies
Randomised controlled trials
Controlled trials
3.Observational studies
Cohort studies
Case-controlled studies
4.Descriptive studies
Cross-sectional designs
What order to the study designs go in?
Match the question with the type of evidence needed!
Question Evidence
How effective is this preventative intervention?
RCT
How effective is this therapy? RCT
How effective is this screening/diagnostic tool?
RCT
Is this a causative factor for this disease?
Cohort or case-controlled studies
What is the prognosis for this? Cohort or case-controlled studies
How do patients feel about their condition?
Qualitative experimental studies
How common is this condition? Cross-sectional studies
How cost-effective is this intervention?
RCT
The end! Any questions?