core guide chapter 4 - meta analysis Flashcards
defining the question
begin with clearly defined qn
idea of what info source you will use
searching for resources
use bibliographic database
specific search terms eg mesh
combine with Boolean operators - AND, OR, NOT
limit by year/language etc
refining the research
use details in qn to guide the search strategy
keeping records
accurate bibliographic details, search histories, critique details, key info helps you find things again quickly
reference manager, endnote, mendley useful online
selecting research
depends on the purpose of the article
if want quick ans - l;imit to reviews
if want long answer- read all
inclusion and exclusion criteria
appraise
read the papers see if
relevant
valid
applicable
apply research
only if valid - discuss reasons why not applying it with patient etc
reference databases
organise info on all relevant info on the field
medline Embase and psychinfo - all include biomedical info and some other books and publications
advisable to search more than 1 database
citation data
citation - reference to another article or source of information
most influential articles are cited the most
information on ISI web of science and google scholher
conference abstracts
published first at conferences may appear in official conference proceedings
accessed through Conference proceedings index 0 through Web of Knowledge
open access resources
publically funded research is in open access
eg pubmed
evidence based medicine resources
see if already been reviewed by a reputable group
Cochrane library - systematic reviews, clinical trials health technology assessments, economic evaluations
BMJ clinical evidence - publish systematic reviews
TRIP database
bibliographic software
manage all referenced evidence by enabling you to store it in your own database or library
manual cataloguing of references relating to research or topics
automatic collection and organisation of references from bibliographic databases, library catalogues etc
quick searches for a particular reference
search and retrieval of bibliographical subsets
print/save a list of references
integration with word processing software to insert and format citations and bibliographies
format for data sharing and other packages
format according to a particular style
save full text articles and access them anywhere online
list of bibliographic resources
endnote Mendeley reference manager refworks Zotero
things to consider in a systematic review
address important qn
thorough search
methodological quality assessed and studies weighted
appropriate to combine
publication bias assessed
results interpreted correctly and with a association with the larger picture
diagnositic test accuracy studies
aim defined
used gold standard test
appropriate spectrum of participants
disease status established
methods for test described in enough detail
work up bias avoided
observer bias avoided
CI for sensitivity specificity, P/NPV
could results be influenced by the gold standard test
study findings placed in wider context of other diagnostic tests
relevant to clinical practice.
qualitive research studies
qn
method appropriate
researchers perspective influence the results
recruitment strategy appropriate for the aims
ethics
transcribed and analysed in a rigorous manner
conclusions justified findings transferablle to other clinical settings
useful insight for clinical practice
steps in a systematic review
define qn search literature appraise studies extract data synthesise data report and apply the findings
what should a systematic review be based on
protocol that is produced before:
qn
description of the search strategy - including databases and search items to be used - methods for search eg contacting authors
details of who will perform the search
method for assessing the quality of the search
description of the data to be extracted
method for synthesising the results - qualitive or narrative report of the findings , tables summarising the populations, quantitive synthesis through meta analysis
plan for reporting and applying the results
reporting a systematic review
methods of PRIMSA
describe methods
resukts should be in a flow chart
details of included papers in flow charts
what is a meta-analysis
statistical analysis of a collection of studies
collected through a systematic review
statistical techniques allow quantitive results to be taken by combining results from all individual studies
why conduct a meta-analysis
individual studies have inconsistent findings
insufficient to ans a research qn or provide evidence to base practice
advantages of meta analysis
identifies consistant patterns include more people identify heterogeneity between studies explore publication bias provide evidence for making decisions
extraction …
off main result from each study with probability that this is due to chance
checking ….
whether appropriate to pool
calculation
of summary result as a weighted average across the studies
weight takes into account the varience of the study reflects the size of the study
greater weight to more informative studies
presentation
forrest plots
graphical representation from each study combined with a meta-analysis - combined meta analysis result
forest plot
overall estimate is shown at the bottom - diamond
centre respond to summary effect esitimate
width of diamond = confidence interval
clinical check if it is appropriate to pool results
judgement of whether reports are reflecting the same thing and therefore that the results are meaningful and helpful
not appropriate if differences in study participants, interventions or outcomes that suggest a different underlying research qn
statistical reasons for not pooling
assessed to see if significant variation with respect to populations, exposures, clinical settings, and designs used
Galbrith plot
fascilitate the examination of heterogenety
regression line through origen represent the pooled estimate with 95% CI - little heterogeneity 95% studies should fall in the lines
vertical spread describes the extent of herterogenity and the outliers
calculating a pooled result
method depends on whether statistical heterogeneity
no heterogeneity - fixed effects model can be used
heterogeneity can be allowed for by doing the random effects model to pool estimates
fixed model
no evidence of heterogeneity
study evaluates common treatment/exposure effect
assumes single true/fixed underlying effect
random effects model
effects different in different studies
distribution of effects from which the study estimates the means of different effects
sub groups can be pooled in sub analysis
what to watch out for in a metanalysis
false impression of consistency across individual study results
systematic variations
appropriate to pool?
possibility all studies have same systematic error
publication bias
exploring publication bias
effect may be over or underestimated caused when only a subset of information is available funnel plots link between study size and estimate symmetrical = no pub bias
