Medical Statistics Exam Technique and Coaching, Part 2 Richard Kay Statistical Consultant RK Statistics Ltd 22/09/2019

Short Answer Questions You have been asked to work with a statistician to determine the sample size for a phase III parallel group superiority trial in heart failure patients of a new drug “X” versus placebo. The primary endpoint will be all cause mortality after one year of treatment Apart from a type I error (statistical significance level), list 3 other critical direct assumptions/pieces of information you will need to tell your statistician so a sample size can be calculated for this study design (3 marks) 22/09/2019

Q1, 2013 (continued) Complete the table below to show what change in the 3 assumptions you have listed in the question part a) would result in a smaller sample size (1½) If the primary endpoint was changed to a composite of all cause mortality plus cardiovascular morbidity after one year, what would be the expected impact on sample size? (½) Assumption from part a) Change in assumption to achieve a smaller sample size Type I error (e.g. p<0.01) Increase (e.g. p<0.05) 22/09/2019

Q1, 2013 (continued) In another phase III trial, a biomarker (raised protein level ) was evaluated for sensitivity and specificity in predicting hospitalisations for heart failure. The results showed this biomarker to have 70% sensitivity but 95% specificity to hospitalisations for heart failure Briefly explain what is meant by a “70% sensitivity and 95% specificity for heart failure hospitalisation” (2) How could this biomarker be used in future trials? (3) 22/09/2019

Q6, 2014 Define Absolute Risk Reduction (ARR) (1) Define Relative Risk (RR), also known as a risk ratio (1) What is a hazard ratio and when is it used in preference to a RR? (2) In the context of a comparison between two treatments explain the term ‘null value’ What would be the null value be for an RR? What would be the null value be for an ARR? (3) 22/09/2019

Q6, 2014 (continued) If the point estimate for a RR of drug X versus placebo for all-cause mortality was 0.5 (in favour of drug X), briefly explain, giving an example for each, what the 95% CI values for the RR could be when the p value for the result was: Highly statistically significant Not statistically significant (3) For the above scenario comparing drug X with placebo, give an example of a point estimate and the 95% CI if the RR result was now statistically significant in favour of placebo (1) 22/09/2019

Accept Null Hypothesis Reject Null Hypothesis Q9, 2015 Regarding Statistical Error In the context of testing drug A on mortality in a placebo controlled study, give the “null hypothesis” (1) In the table below, there are cells labelled W, X, Y, Z. If you could replace these letters using just one of the three terms “Correct decision”, “Type I error” or “Type II error” indicate what term should be given to each letter. (2) Which letter could be considered the “Regulator’s Risk” and why? (1½) Which letter could be considered the “Sponsor’s Risk” and why? (1½) Accept Null Hypothesis Reject Null Hypothesis Null Hypothesis is true W X Null Hypothesis is false Y Z 22/09/2019

Regarding Missing Data Q9, 2015 (continued) Regarding Missing Data List 2 reasons why missing data is of concern in a clinical trial (1) List 3 ways to handle missing data in an Intention to Treat analysis for continuous data (1½) List 1 way to handle missing data in an Intention to Treat analysis for binary data (e.g. dead or alive) (½) To assess the impact of missing data, what would you ask your statistician to do? (1) 22/09/2019

Briefly describe why we perform (3) Q8, 2016 Briefly describe why we perform (3) A superiority study A non-inferiority study An equivalence study What do you understand by the non-inferiority margin? (1) A non-inferiority study was performed with a new drug (X) versus a licensed reference drug (Y). The non-inferiority margin for relative risk (RR) on all-cause mortality (the primary endpoint) was 1.30. 22/09/2019

Test drug (X) is non-inferior to reference (Y) Q8, 2016 (continued) Draw a diagram to show the position of the non-inferiority margin and illustrate the position of the 95% confidence interval for the RR result for the following scenarios: (6) Test drug (X) is non-inferior to reference (Y) Test drug (X) meets the criteria for non-inferiority but the reference drug (Y) is better Test drug (X) does not meet the criteria for non-inferiority and the reference drug is better Test drug (X) is superior to reference (Y) The result is inconclusive (neither X nor Y is superior and the criteria for non-inferiority are not met) *you need only draw one diagram, but please clearly label each scenario 22/09/2019

b) What would be the null hypothesis for this trial? (1) Q5, 2018 In a phase 3 clinical trial, 1000 patients were treated for one year with a new investigational drug X, whilst another 1000 patients received placebo for a year. The event of interest for the primary endpoint was a composite of all-cause mortality and non-fatal myocardial infarction (MI). After one year, 50 patients in the Drug X group were dead or had a non-fatal MI versus 100 patients in the placebo group. a) Briefly describe 2 key safeguards which should make a placebo control arm ethically acceptable in this trial (2) b) What would be the null hypothesis for this trial? (1) 22/09/2019

Q5, 2018 (cont.) c) What do you understand by a type 1 and a type 2 error and give a typical value you would expect the statistician to assume for the type 2 error in this trial? (3) d) Show how you would calculate the relative risk reduction (RRR) for drug X on death and non-fatal MI in this trial using the information given. Note: you do not need to calculate the actual value, full marks will be awarded by showing the method (2) e) The absolute risk reduction (ARR) in this trial was 5%. Calculate the Number Needed to Treat (NNT) and explain what the NNT means in the context of this trial (2) 22/09/2019

Q8, 2017 a) What do you understand by an event driven (tine to event) study? (2) b) With respect to the assumptions made for the event rate, list 2 important trial logistics that must be in place for a time to an event trial (2) c) If during the trial the event rate is lower than expected, briefly describe 4 different options available to youto ensure the trial can be completed (4) d) For 2 of the options you have listed in c), briefly describe what you must do to implement that option 22/09/2019