Secondary research Flashcards
Narrative review
-carried out by experts in the field of study but are guided by their own opinion. These are broad reviews and provide a qualitative summary of the studies
Systematic reviews
- focus on a narrow question
- comprehensive and specific data collection
- uniform criteria for study selection
- quantitive synthesis of data- not always possible
Meta-analysis
- statistical combination of individual study data into a quantitive summary
- extended systematic review
Literature search
- must use multiple databases
- cross check the reference list of each individual study retrieved
- hand search for materials unidentfied online
- approach experts to comment on any missing studies
- identify grey literature- e.g conference abstracts, presentations and posters
Inclusion criteria for studies in a systematic review should consider
- types of study designs to include
- types of subjects to include
- types of publications
- language restrictions
- types of interventions
- time frame for included studies
Basic steps for meta-analysis
- literature search
- establishing criteria for including and excluding studies
- recording of data from the individual studies
- statistical analysis of the data
Combining individual trial data in meta-analysis
- methods used for meta-analysis use a weighted average of results
- weighting refers to the significance attached to each study based on multiple factors
- may used fixed or random effects model to combine outcomes from different studies
Fixed effects model
- assumes that all the studies share the same common treatment effect (homogeneous)
- assumes only random error within studies could explain observed differences
- ignores between-study variations
- can only be applied if heterogeneity can be safely excluded by testing for it
Random effects model
- assumes all studies do not share the same common treatment effect
- assumes each study shows a different effect which are normally distributed around the true mean
- gives greater weight to small studies
- susceptible to publication bias and results in wide less precise confidence intervals
Q statistic
-significant heterogeneity
Which model to use?
- if there is an absence of significant heterogeneity (Q statistic), both fixed and random effects model have similar confidence issues
- if very heterogenous then random effects model will yield wider confidence intervals
Fixed effects statistics
- Mantel-Haenszel and Peto ratios are used in fixed effect analysis
- Mantel-Haenszel is useful even when wide differences exist between individual studies in the ratios of the size of two groups
- can be used in cohort/case control designs too
- Peto- mostly restricted to reviewing RCTs as it can produce biased results in unequal groups
Clinical heterogeneity
-differences in the study that results in uneven outcomes and is describable but not measurable
Methodological heterogeneity
- refers to heterogeneity resulting from the differential use of study methodology
- these may lead to different conclusions in different studies, even though the clinical characteristics are the same
- methodological heterogeneity is describable but does not need quantification
Statistical heterogeneity
- variation in trial outcomes
- homogeneous sample refers to a set of individual studies that have comparable outcomes without much variation
- heterogeneity refers to the presence of significant variation among the individual studies in a sample
Test for statistical heterogeneity
- can be graphical or statistically
- graphical = forest plot or L’Abbe plot or Galbraith plot
- statistically is chi squared, chochran’s Q,
Forest plot
-blobbogram
L’Abbe plot
- modified scatter plot wherein CER is plotted against EER from individual trials included in the meta-analysis
- treatment better than control will be above the line and control better than treatment will be below the line of equality (diagonal)
Galbraith plot
- an alternative to a forest plot
- on the horizontal axis one plots 1/standard error of the study effect estimate
- on vertical axis one plots the study effect estimate divided by the standard error (log odds ration/SE=standard normal deviate)
- distance from the middle line indicates SD
- major limitation of this approach is that the statistical tests lack the power to detect heterogeneity in most meta-analyses
Forest plot
- horizontal axis is the odds ratio or relative risk
- vertical line at 1.0 is the line of null effect
- numbers below 1 favour treatment
- numbers above 1 favour control
- a box and line represents the study
- the bigger the box the bigger the study
- the line represents the 95% confidence interval
Chi squared
- tests of heterogeneity can be tested with a Q test or I2
- chi squared tests provide a test of significance for heterogeneity but do not measure it
Cochran’s Q
-calculates the weighted sum of the squared differences between individual study effects and pooled effects across studies
I2 statistic
- describes the percentage of variation across studies that are due to heterogeneity rather than chance
- high P suggests that the heterogeneity is insignificant and that one can perform meta-analysis
Funnel plot
- detects publication bias
- shows relation between the effect size and precision of the individual studies in a meta-analysis
- large studies get published, small ones dont which can produce an asymmetrical funnel
Failsafe N
-calculates the number of zero effect studies that would be required to nullify the mean effect seen in a meta-analysis
Cumulative meta-analysis
- this can be used to assess the potential impact of publication bias in tilting or nullifying the effect
- sort the studies from largest to smallest, plot forest plot as you go along
The trim and fill procedure
- developed by Duvall and Tweedie
- helps to assess whether the effect would change if the bias were to be removed
- sensitivity analysis where missing studies are imputed and added to the analysis and then the effect size is recomputed
Location bias
- not being located due to citation habits, the database used, the key terms used or multiple replications of data
- location of information not the studies!
Inclusion bias
-refers to reviewers tendency to include studies they agree with
Sensitivity analysis
- e.g failsafe N
- publication bias can be examined
Blobbogram
- Forest plot
- the combined or pooled effect size is given by various stats including OR, RR, standardised mean differences, Cohens d
- confidence interval of individual line width estimates and combined estimates(lozenge width) should be noted
- rectangle size is noted as it shows assigned weight
- heterogeneity results should be noted -absence of heterogeneity is indicated by vertical linerarity of rectangles or non signdicant chi square test for heterogeneity
Effect sizes for continuous variables
-there are two effect size measures for continuous variables
1. simple difference between the mean values (DM)- mean value of group X minus the mean value of group Y
2. standardised difference between the mean values (SMD)- carried out by dividing the DM by the pooled standard deviation of the two groups in the basic formula
SMD=(mean value of groupX/mean value of group Y
Meta-analyses standard format
-must follow QUORUM statement
Standard format for RCTs
-CONSORT statement
Meta-regression analyses
-refers to a technique of regression wherein regression model is applied to meta-analysis to analyse which characteristics of the studies actually contributed to overall effect size