PHRM3031 - Systematic Reviews and Meta-Analysis

Question 1

bias

types

Answer 1

refers to any systematic error is estimating the effect of drug, exposure or risk factor on specified outcome

selection, information, publication, industry, p values

Question 2

systematic reviews

Answer 2

• adopts a scientific approach to combining and/or contrasting the results of research in a particular area
• the overall structure of a good systematic review is similar to any good clinical study; clear research questions, clear and detailed study plan including how the results will be analysed, well- justified inclusion and exclusion criteria
• systematic reviews that use statistical approaches to combine data from the original studies are meta-analyses

Question 3

narrative reviews

Answer 3

-contrast systematic reviews
-involves an expert critically reviewing the literature in on specific topic
-narrative reviews often make a case for a particular interpretation of the data
usually discuss and incorporate different types of evidence

Question 4

8 key components of meta-analysis

Answer 4

1.clear research question
2.replicable literature search strategy
3.explicit (and well-justified) inclusion and exclusion criteria
4.formal assessment of each study for risk of bias
5.description of individual study results
6.explanatory analysis of difference between studies and study results
7.aggregated results providing results are suitably homogenous
8.summary of limitations
(7 is waht really makes it a meta-analysis)

Question 5

why do we need meta-analyses?

Answer 5

many important effects and harms are modest in size

• many (most) clinical studies are not large enough to detect small but important effects of harms
• since most trials are powered for treatment benefits, this is particular problem for treatment harms and assessing the effects of a treatment in subgroups

Question 6

limitations and challenges of meta-analyses

3

Answer 6

• rely on big assumptions
  1. the reviewed studies are of high quality and without systematic bias
  3. the studies reviewed are comprehensive (non publication bias)
  3. the effect the meta-analysis is estimating is constant across the studies, i.e the studies are homogenous

Question 7

homogeneity assumption

Answer 7

• clinical trials are not repeated, different trials are conducted in different population and in patients with differing level of baseline risks
• typically we know the homogeneity assumption is false
• the question is whether the degree of heterogeneity is large enough to effect the observed result of the meta-analysis

Question 8

meta-analyses vs randomised trials

Answer 8

• well-conducted meta-anaylses have the biggest impact when they combine a series of small randomised trials that enrol similar patients, use similar methods and report similar results
• due to the limitations of meta-analysis, a single well conducted large randomised trial will likely provide better evidence than a meta-analysis of many small trials