Applied statistics and interpreting results Flashcards
Types of data used/found in clinical trials
Binary- two types of responses (yes/no)
Continuous- blood pressure, cholesterol
Time-to-event- death, pregnancy
What is categorical data
Nominal (categories unordered)
Ordinal (categories ordered)
Describe nominal data
Gives averages and means
Discrete- finite values possible
Continuous- all values are possible (decimals)
Parametric vs non-parametric testing
Parametric= assume normal distribution, give significant result more often
Non-parametric= compare rank order, not influenced by outliers
What is a null hypothesis
No difference exists in treatment A and B
Nothing happens
What is the alternate hypothesis
Difference exists between treatment A and B
Describe type 1 error alpha
False positive
Falsely rejecting Ho and detecting a significant difference when there is no difference
Describe type 2 error beta
False negative
Falsely accept Ho and not notice a difference when there is a difference
What is power (1-beta)
Probability of correctly rejecting Ho and detecting difference when difference exists.
Higher the power, the better.
How to measure the effect of binary date
Risk difference (absolute risk difference)
Relative risk
Odds ratio
Number needed to treat
Number needed to harm
What is absolute risk
Absolute risk difference - difference between risk of event in intervention and control group
Absolute risk reduction- treatment is effective and reduces unwanted event
Absolute risk increase- treatment does not work, increased risk of event
What is NNT
Number needed to treat
Number of patients needed to be treated to produce one additional successful outcome
Compare efficacy of treatments
NNT= 1/ARR
Best for comparing treatment with placebo
What is NNH
Number needed to harm
Number of patients needed to be treated to produce one ADR
What is relative risk
AKA risk ration, risk of event occuring
1.0 is middle number
Lower= death
Higher= cure
What is odds ratio
Probability of event occuring over risk of event not occuring
What is selection bias
Choosing people to under/overrepresent a study
Not true representation of population
What is attrition bias
Selective dropout of participants who differ from others.
E.g. those experiencing ADR
What is P value
Probability by chance
Stat sig= P<0.05
How often chance gives favourable results
What is the confidence interval
Estimate of the precision of results
95% CI = range of values within which we are 95% confident the true population estimate lies
Narrower CI indicates precise and reliable results
What is a significant result in relation to CI
Sig if CI does not cross null value
> 1 for RR
0 for ARR
Statistical significance vs clinical significance
Stat= size of effect and 95% CI in relation to Ho (no diff between treatment and control)
Clin= size of effect and 95% CI in relation to a minimum effect considered to be important
What is intention to treat in RCT
Final analysis is more robust if all patients are included (even those who didn’t complete)