Statistics Flashcards
First trials in humans, single sub-therapeutic doses given
Phase O drug design
- primarily to gain pharmacokinetics and pharmacodynamics data
Healthy volunteers trial medication to determine safe dosing ranges and identify some adverse effects
Phase I drug design
100-300 participants, confirming dosing requirements and efficacy
Phase II drug design
Dosing requirements = IIa
Efficacy = IIb
100-1000s of participants, confirm safety and efficacy of drug
Phase III drug design
Confirm effectiveness vs placebo/active treatments (“gold standard”)
Involve regulators (e.g. FDA) to obtain approval
Continued pharmaco-vigilance, trial of new uses/populations, involvement of paediatric patients
Phase IV drug design
Larger populations but less controlled, longer follow up
Assess interactions with other medications
Greatest strength of case control studies?
Multiple risk factors can be assessed
Greatest threat to validity in longitudinal cohort study?
Loss to follow up
Study looking at exposures in patients with known disease?
Case control
Evaluation of participants with known exposure of interest?
Cohort study
Major strength of RCTs?
Minimises bias and confounding
Which research design can most conclusively demonstrate causality?
RCT
Test which evaluates if there is a significant difference between expected frequencies and observed frequencies in one or more sets of categorical data
Chi-square test
The degree of which two variables are related
= correlation
Test which compares the means of two different samples of data which have a normal distribution
Student’s t-test
Test determines whether there is statistically significant differences between the means of three or more independent variables
= analysis of variance
Type I error
False positive
Type II error
False negative
i.e. failure to reject an incorrect null hypothesis
Feature of a meta-analysis indicating minimal or no bias in the results?
Statistical heterogeneity
In what circumstances does an odds ratio approximately equal a relative risk ratio for a disease in the population?
Low prevalence of disease
- OR is a ratio of odds, whereas relative risk is a ratio of the two probabilities
- OR would exaggerate the risk if the disease is more common than 10%
What is risk reduction?
Difference in event rates between the control and experimental groups expressed as a proportion of the event in the untreated group
How to adjust for multiple comparisons in a study?
Bonferroni correction
Best way to reduce random error in a study?
Increase sample size
Definition of the p value
Probability of obtaining the test result observed under the assumption that the null hypothesis is true
i.e. the probability that an observed difference could have occurred by chance
Standard error
A measure of the accuracy of the sample estimate (i.e. the larger the sample size, the more likely this accurately reflects the population, so the smaller the sample error)
Terms used to describe standard deviations from the mean
Sigma score
Standard score
Z score
Measurements used for variability of data?
Range, interquartile range, standard deviation
Logistic vs linear regression
Both used to find a relationship when there are multiple variables
Logistic regression = binary outcomes (e.g. presence of disease)
Linear regression = continuous outcomes (e.g. oxygen requirement)
Which measure of association is typically used for survival analysis?
Hazard ratio
Best measure to describe the frequency of occurrence of disease in the investigation of an epidemic?
Prevalence
Positive predictive value
Calculation of true positive rate in total positive test outcomes
(True positive/All positive results)
Negative predictive value
Calculation of true negative test results within total negative test outcomes
(True negative/All negative results)
Sensitivity
Ability of a test to detect disease
= True positive test results / Positive for condition
If highly sensitive test is negative = rule out
Specificity
Test which correctly identifies people without disease
= True negative test result / Patients without disease
Positive likelihood ratio
= Sensitivity / (100% - specificity)
Negative likelihood ratio
= (100% - sensitivity) / specificity
Absolute risk reduction calculation
(events in controls/total controls) - (events in treatment/total treatment)
Number needed to treat
NNT = 100 / ARR (%)
Number needed to harm
NNG = 100 / absolute risk increase (%)
Odds ratio
(ad) / (cb)
