Hadpop Flashcards
Put the following types of study in order of quality:
Case report Case control RCT Cohort Metaanalysis
Meta analysis RCT Cohort Case control Case report
What does a forest plot represent?
Studies and their significance in a meta analysis
Square is the odds ratio/point estimate of each study Lines are confidence intervals Square size is the weighting Diamond is the pooled estimate Vertical line is the null hypothesis
What makes a result significant?
If the null hypothesis lies outside the confidence intervals
Where might bias be generated in a clinical study?
Randomisation - selection bias, healthy worker, survivor bias
Blinding ineffective
Question - recall bias
Intention to treat - dropout bias
Follow up - length time and lead time bias
Conflict of interest
What are length time and lead time bias?
Both bias associated with screening programs - make screening look more effective
Length time - less fatal things more likely to be found because they last longer
Lead time - if things are found earlier, perceived survival is increased.
Can you control for bias?
No, you can minimise its effects with blinding, good questions, good selection process but can’t control.
What are the most common confounding factors?
Age
Sex
Ethnicity
Socioeconomic status
Can you control for confounders?
Yes by using a randomised controlled trial
But try to make sure equal numbers of each characteristic are in each group
Randomisation means equally likely to be in each group so balances unknown confounders
Why are RCTs double blinded?
Reduces the placebo/nocebo effect and observer bias
Why do RCTs use controls?
To make sure we are measuring the variable we are trying to measure by making the two groups the same in every way apart from the variable.
Why are RCTs randomised?
To balance and control for unknown confounders
What is the difference between a systematic review and a meta analysis>?
Meta analysis requires similar studies
What does a funnel plot show?
X axis is effect size
y axis is study size
Expect a funnel shape in a good meta analysis
If not symmetrical it indicates publication bias
How would you interpret an odds ratio?
- The odds ratio shows you are x times as likely to die if…
- We can be 95% confident that the true value lies between x and y
- The null hypothesis is 1
- The null is not in the confidence interval so we can reject it.
- The result is significant
How would you interpret an incidence rate ratio?
- The IRR shows you are x times as likely to die if…
- We can be 95% confident that the true value lies between x and y
- The null hypothesis is 1
- The null is not in the confidence interval so we can reject it.
- The result is significant
