Statistics Flashcards
What is the difference between effectiveness and efficacy?
Effectiveness - how well a treatment works in the practice of medicine
Efficacy - measure of how well a treatment works in clinical trials or laboratory studies.
How do you calculate standard error of mean?
How is SEM used in calculation of 95% CI?
SEM = SD/square root (n)
n = sample size
Therefore SEM becomes smaller as sample size increases.
95% confidence interval:
Lower limit = mean - (1.96xSEM)
Upper limit = mean + (1.96xSEM)
Describe the level of evidence of following:
1a 1b 2a 2b 3 4
Ia - evidence from meta-analysis of randomised controlled trials
Ib - evidence from at least one randomised controlled trial
IIa - evidence from at least one well designed controlled trial which is not randomised
IIb - evidence from at least one well designed experimental trial
III - evidence from case, correlation and comparative studies
IV - evidence from a panel of experts
Describe the following grading of recommendations:
A
B
C
Grade A - based on evidence from at least one randomised controlled trial (i.e. Ia or Ib)
Grade B - based on evidence from non-randomised controlled trials (i.e. IIa, IIb or III)
Grade C - based on evidence from a panel of experts (i.e. IV)
Calculation of:
LR+
LR-
LR+ = Sn/1-Sp LR- = 1-Sn/Sp
Difference between accuracy and precision
Accuracy is the degree to which on average a test measurement represents the true value.
Precision is the degree to which further measurements or calculations show the same or similar results.
Test can be precise in terms of repeatability and reproducibility, but may not be accurate (ie, far from the true value)
Positive and negative likelihood ratio
LR+ = Sn/1-sp
LR- = 1-Sn/Sp
Phases of clinical trials
Preclinical
Phase 1-4
Preclinical - in vitro or in vivo (animal study) to see whether a potential compound has a scientific merit for further development as a therapeutic drug.
Phase 1:
Toxicity screening for maximum tolerated dose in humans N 3-10
Phase 2:
Test for positive effect in humans for fixed dose, reduced dose from phase 1 N 10-20
Phase 3:
Larger trial 1000-3000, usually multicentre trial, aimed at collecting more data on efficacy and safety. Longer follow up. Aimed at getting drug approved if the data shows good safety and tolerability and efficacy.
Phase 4: post marketing surveillance to check for long term safety.
Define following bias:
Selection bias
Selection bias - bias that occurs because of the way in which people are selected to go into various exposure groups in a study.
eg: recruitment on hospitalized patients only may result different outcome to treatment when compared with patients in the community at large.
Distribution of data according to standard deviation 1/2/3
1 SD = 66%
2 SD = 95%
3 SD = 99.7%
What does p value of <0.05 actually mean?
It means that if null hypothesis was actually true, then you would observe the difference in 5% due to random chance. P value does NOT mean the error rate of correctly rejecting the null hypothesis.
P value CANNOT 100% help you to reject the null hypothesis - it provides confidence to reject the null hypothesis.
This statistical significance could mean either:
- There really is a true difference, or
- There is not a true difference, but the result has been obtained simply by chance.
In order to correctly reject the null hypothesis, you actually need to interpret the data to see whether it is clinically significant, or do repeat studies to confirm the findings.
4 criteria for viability of a screening test
- Disease must have a pre-symptomatic phase detectable by screening
- Intervention at the pre-symptomatic phase should change the natural history of the disease to reduce morbidity and mortality
- Screening needs to be inexpensive and easy to administer and acceptable to patient
- Screening has to have high sensitivity and specificity
Lead time bias vs length time bias
Lead time bias - screening allows earlier diagnosis which means it automatically increases ‘survival time’ ie time from diagnosis to death, even if natural history is not changed.
Length time bias - if a disease has a long pre-clinical phase, this may be indicative of slow disease progression. Longer lead time makes it more likely to be detected by screening and therefore will have a better prognosis than those who are not detected (ie, more aggressive cancers with shorter lead time). Therefore factually longer survival caused by slower disease progression gives the impression that the screening was beneficial.
Describe t test and ANOVA
t-test compares 2 means.
t-test can be paired or unpaired:
Paired - compare two dependent measurements from the same sample, eg: before and after measurement of blood pressure, or measuring L and R feet
unpaired - compares two independent groups or samples. eg: birthweight of male and female children.
Analysis of variance (ANOVA) compares more than 2 means
Describe multivariate analysis
Statistical method in which effect of multiple variables can be taken into account.
eg: effect of smoking and age in lung cancer.
There are 2 types of multivariate analysis:
- Multiple regression - used when the outcome is a continuous rather than dichotomous variable (eg: height, blood pressure, blood glucose)
- Logistic regression - used when the outcome is dichotomous (alive or dead, complication or no complication etc)
