Which of the 5 criteria of G.R.A.D.E is considered the greatest threat to the validity of a systematic review and meta-analysis?
Publication bias / missed RCTs
What are the 5 criteria in G.R.A.D.E?
Indirectedness
Imprecision
Inconsistency
Publication bias
Risk of bias
What does Grade 2b recommendation mean?
Grade 1 = strong
Grade 2 = weak
A = high certainty (consistent across RCTs)
B = moderate certainty (RCTs but some limitations)
C = low certainty (serious flaws)
What are self controlled case series?
A study where individuals act as their own control
Comparisons are made within individuals
Only individuals who experience the event are included
Often used to identify temporal association between exposure and event eg vaccination side effects eg GBS
When does concealment occur?
At randomisation
Recruiter blinded
When does blinding occur?
After randomisation
Need to specify who was blinded
How do you calculate relative risk reduction?
1 - relative risk (expressed as %)
When are Hazard ratios generated?
From time-to-event/ survival analysis ie relative risk averaged over time
What studies can produce odds ratios?
Case control
Cross-sectional studies
Logistic regression
What is the Cox Proportional Hazards Model?
A statistical technique used in survival analysis to evaluate the effect of various factors on time until event
Produces hazard ratios which compare the risk of an event occuring based on different predictor variables
What is the Log Rank test?
Used in survival analysis to compare the distribution of time to event in two or more independent samples
It answers the question - is there a significant different between two Kaplan-Meier curves
How do you calculate relative risk?
Experimental event rate / control event rate
How do you calculate relative risk reduction?
1 - RR
How do you calculate absolute risk difference?
Control event rate - experimental event rate
How do you calculate NNT?
100 / risk difference (in %)
ie
100 / control event rate % - experimental event rate %
How do you calculate positive likelihood ratio?
True positive rate / false positive rate
= sens / (1 - sp)
How do you calculate negative likelihood ratio?
(1-sp) / sn
How do you calculate power in a study?
1-beta error
1-0.20 = 0.80 (80%)
Minimal clinical important difference (delta)
Why would you use composite outcome measures?
If low event rate to improve power of study
What statistical test do you employ for numerical univariate values (eg blood pressure, FEV1)?
Unpaired T-test
What statistical test do you employ for categorical univariate analysis (eg dead vs alive)?
Chi squared test
What are examples of parametric tests?
T-test
Chi squared
What are non-parametric tests?
Distribution free tests
Mann-Whitney U
Wilcoxon’s Rank and Kendall’s S
Signed Rank Sum test
What statistical test do you use for multivariate analysis?
Logistic regression or linear regression
