Evidenced-based Dentistry Flashcards
What are the advantages of a sytematic review compared to a single study
- Saves readers time
- Provides reliable evidence
- Resolves inconsistencies – single studies brings all these conflicting studies together
- Identifies gaps
- Identifies where questions have been fully answered and it is unethical to conduct anymore studies
- Explores differences between studies – identifying factors that influence the effect of an intervention
What is the importance of systematic reviews
- Reduce large quantities of info into manageable portions, makes it easy for professionals to access research
- Formulate policy and develop guidelines - guidelines are based on these
- Efficient use of recourses – studies have already been done
- Increased power and precision - small studies alone don’t mean much but pulled together, it gives greater power in estimates
- Limit bias and improve accuracy
What is a process of systematic review
- Authors – two or more authors, one is a topic expert and the other methodological
- Study protocol – in advance set out what they plan to do methodologically
- Create a specific question using PICO
- Search strategy
o Using multiple electronic databases
o Include published and unpublished literature
o Ensure search is both comprehensive and repeatable - Inclusion/exclusion
o Specific
o Agreed in advance - Critical appraisal
o Thorough
o Assess for risk of bias - Synthesis
o Can be qualitative (narrative) or quantitative (meta-analysis)
What are the key characteristics of a systematic review
o well formulated question
o comprehensive data search
o unbiased selection and abstraction process
o assessment of papers
o synthesis of data
How do we produce a well formulated question
- Use PICO
PICO
What is meant by participants
who is the review looking at
What is meant by interventions
PICO
What is the px being exposed to
What is the comparison
PICO
- control group
- what will the intervention be compared to
What is outcome
PICO
which outcome will tell you which intervention is most effective
What do we define the inclusion/exclusion criteria based on
the question
PICO
What should we consider using when doing the comprehensive data search
o Electronic Databases
o Reference lists
o Hand searching
o English language/non-english language
speak to librarians for help
What are the different reporting bias
positive results are more likely to be
o Published – publication bias
o Published rapidly – time lag bias
o Published in English – language bias
o Cited by others – citation bias
How do we conduct an unbiased selection and abstraction process
- Select relevant papers
- Data extraction to a predefined data extraction form
- Process should be conducted independently by at least 2 reviewers
- Clear description of reasons for exclusion
- Adequate description of included studies
- Details of studies funding source
How do we assess the papers
- How well the studies have been designed and conducted – methodology
- Process should be conducted independently by at least 2 reviewers
- Results of the assessment should be reflected in the analysis
What can the assessment of papers be used as
o as a threshold for inclusion of studies
o as a possible explanation for differences in results between trials
o in sensitivity analyses
o as weights in statistical analysis of results
What are the 2 main quality assessment tools in the assesment of papers
composite scales
component approach
What is composite scales
assign numerical value to individual items to provide overall estimate of quality
What is a component approach
assess relevant methodological aspects individually e.g randomisation, blinding, dropouts
Why is the component approach preferred to the composite scale
more specific in the downfalls of the studies
don’t know the specifics of what made the study unreliable in the composite scale
What does the risk of bias assessment consist of
done as part of the assessment of papers
o determines extent to which results of study can be believed
o direction of bias: causes overestimation or underestimation of effect
o magnitude of bias
What aspects of methodology does a risk of bias assessment for RCT consist of
Random sequence generation (selection bias)
Allocation concealment (selection bias)
Blinding (outcome assessors)
Incomplete outcome data (attrition bias)
Selective outcome reporting (reporting bias)
Other bias
What is a variation in risk of bias be an explanation for
heterogeity between results of different studies
significant risk of bias in included studies should give rise to cautious conclusions in a systematic review
What are the two types of data synthesis
qualitative - narrative
quantitative - metanylsis
When is meta-analysis inappropriate
when not enough data or heterogeneity exists
What is meta-analysis
- This is the process of using statistical methods to combine the results of different studies
What are the aims of meta-analysis
- Aim is to integrate findings, pool the data and identify overall trends in results
o Calculates a treatment effect based on pooled data from a group of studies
o Estimates a common tx effect across studies
o Improves precision of a point estimate by using all available data
What are the advantages of meta-analysis
o increase in power
o improvement in precision
o ability to answer questions not posed by individual studies
What are the disadvantages of meta-analysis
potential to mislead, need to be careful when it is undertaken
When should meta-analysis be undertaken
o minimal differences in characteristics across studies
o same outcome measure
o data in each study is available
What are the 2 different types of data
dichotomous
continous
What is dichotomous data
1 out of 2 forms
e.g death or no death
What are the choices of summary statistics for dichomotous data
- odds ratio
- risk ratio
- percent risk reduction or relative risk reduction
- risk difference or absolute risk reduction
- number needed to treat
What are the summary statistic choices for continous
- weighted mean difference
- standardised mean difference
When is more weight given to a study
o more participants
o more events
o lower variance
How can results be displayed graphically
forest plot
Describe the different features of a forest plot
- Horizontal line at bottom is the scale measuring tx effect
- Vertical line in the middle si where tx and control have same effect and there is no different between the two
- For each study there is an ID, data is then divided into the experimental and control group, a weight is given
- There is then a label above the graph to tell you what ratio has been used and the data is shown using a line and also numerically
- Each study has a blob which is placed where the data measure the effect
- The horizontal line running through the blob is called a confidence interval and is a measure of how we think this result may vary with chance
- The wider the line is, the less confident and if it overlaps the vertical no effect line, then insufficient evidence to support a difference in the two interventions
What are the 3 types of heterogeneity
- clinical
- methodological
- statistical
What is clinical heterogenity
the way study has been conducted
variation in participants, interventions, outcomes, study design
e.g age, baseline disease severity, dosage etc
What is methodological heterogeneity
Variation in methods used
* Variations in study design and risk of bias
* e.g use of allocation concealment
What is statistical heterogeneity
Excessive variation in results of studies
Variation in the intervention effects being evaluated in the different studies and is a consequence of clinical or methodological diversity
Manifests itself in the observed intervention effects being more different from each other than one would expect due to chance alone
How do we identify heterogeneity
o if studies are estimating the same thing, we would expect confidence intervals to overlap to a large extent
o statistical heterogeneity may appear in a forst plot as poor overlap of confidence intervals
o Chi squared test
o P<0.1 demonstrates statically significant heterogeneity and then it may be inappropriate to pool data
o I2 statistic is the percentage variation due to heterogeneity rather than chance and <50% is acceptable
What is subgroup analysis
- Studies divided into sub groups
- Should be decided in advance, not after seeing data
- Done where it is suspected in advance that certain features may alter the effect of an intervention
What is the main aim of subgroup analysis
- The main aim of a subgroup analysis is to identify either consistency of or large differences in the magnitude of treatment effect among different categories of patients.
What are examples of subgroups that could be utilised in subgroup analysis
o gender
o age groups
o specific disease subtypes
What is sensitivity analysis
Repeating metaanalysis
.g if some studies eligibility is dubious due to a lack of full details: run the metaanalysis the first time using the dubious studies and the second time using the non-dubious studies
* Need for sensitivity analysis usually decided in review process but can be predecided in the protocol
What info can the sensitivity analysis give us
- When sensitivity analyses shows that the overall result and conclusions are not affected by the ‘dubiosness’ then the results can be regarded wit a higher degree of sensitivity
- On the other hand, it may show that the missing information from these dubious studies has a large impact on results and so the search for the missing info can be undergone e.g contacting trial authors
What is the fixed effect
o Assumes studies are so similar that they are effectively different parts of one large study
o Assumes that the true answer for each study is the same
o If I2 <50% (heterogeneous) then can use this model of meta-analysis
What is the random effect
o Assumes studies are slightly different
o Assumes that the true answer for each study will be slightly different
o More conservative/wider CI
o If I2 >50% then can use this model of meta-analysis
What is GRADE
- Used to estimate certainty of evidence
o high
o moderate
o low
o very low
What are the 5 factors that can lower the quality grade
o high or unclear risk of bias – due to design issues or poor conduct of studies
o inconsistency between studies – heterogeneity
o indirectness – pico
o imprecision – numbers and CI
o publication bias – likely that negative/null results not published
