Insert your own work and answers into this WORD file. If you use SAS to answer a question, then please cut and paste your SAS program statements and the relevant SAS output into this WORD file.
The problems for this homework are typical of those encountered by statisticians working with clinical researchers. So if you'll humor me, I have written this assignment as though it is a day in the life of a project statistician. Here is your agenda:
1. Meet with an investigator planning a new study:
Dr. Stockler is planning a multi-center trial to compare an extended release (XR) formulation of a drug to the approved immediate release (IR) version in patients with advanced Parkinson's disease. The primary outcome will be change from baseline in a Parkinson disease rating scale after 24 weeks of treatment. The rating scale has a maximum of 108 points, with a higher score indicating more severe symptoms. He shows you results from earlier clinical studies: at 24 weeks, mean changes (SD) of -10.8 (8.32) and -11.1 (7.56) points in the XR and IR studies, respectively. (You note these differences mean the patients have more severe symptoms at 24 weeks than at baseline.) These studies both had dropout rates of approximately 20 %. Dr. Stockler says his study will be an equivalence trial because he hopes the results will lead to patients being able to take the XR medication once a day instead of the IR twice a day. He asks how many patients are needed to have 90% power to show that the difference in the mean change from baseline at 24 weeks, XR group minus IR group is not significantly less than -2.5 points. If the difference in the mean change between the groups significantly favors the XR group, he would want to use the results to promote the XR formulation.
- Explain to Dr. Stockler the difference between superiority, non-inferiority and equivalence trials and identify which he is proposing, given the information he has provided.
- Discuss the implications of the expected dropout rate. How will the statistical analysis account for any missing data at Week 24?
- What is your recommendation for the sample sizes? (Show your work: program statements, relevant output, calculations and state assumptions.)
2. Investigator question
Dr. Shaul shows you a journal submission she is reviewing that concludes two drugs are equivalent based on a mean difference remaining within the upper and lower limits of a confidence interval. She wants you to help her draft a response to the authors requiring that they correct their calculations and the resulting inferences because they claimed 95% confidence the drugs are equivalent and clearly used 1.64 as the Z value in calculating the upper and lower limits of the interval. Shouldn't this be 1.96, she asks? What is your response to Dr. Shaul?
3. Prepare a treatment allocation schedule for manufacturing to package study drug
A study site is ready for study drug supplies. 50 patients are to be randomized equally to two treatment groups, with the number randomized to each treatment equal after every 10 patients. Produce a treatment allocation schedule for the pharmacist that meets these requirements.
4. Second investigator planning a study
Dr. Liu asks for the appropriate sample sizes for a study in men with prostate cancer who are experiencing PSA increase after undergoing radical therapy with curative intent. The men do not have signs or symptoms of metastases and questions remain about the clinical importance of PSA doubling, so treatment may have no benefit or even be more harmful than surveillance for these men. Thus, the double blind study will randomize men to treatment with study drug or placebo. During the 24 month treatment period, PSA levels are regularly assessed. The proposed primary outcome is time to PSA doubling. Without treatment, it is expected that 50% of the men will experience PSA doubling by the end of the 24 month period; it is hoped that in the treated group only 25% will have PSA doubling. The investigator would like to have 90% power to detect a difference of this magnitude in the hazard ratio using a test at the 0.05 level of significance. He expects 30% of men in the placebo group and 20% in the treatment group will withdraw from the study by the end of the 24 months, based on results in a study with another drug vs placebo.
What sample size do you advise for each treatment group?
You are also asked whether a difference of this magnitude in the dropout rates will affect the validity of the results and if so, what can be done. What is your response?
Finally, the investigator asks would there be any advantage to stratifying on initial PSA values as well as by clinic?
5. Answer VP for Marketing
You are in meeting to review the final results from a completed study which compared two therapies over 5 years. There were 3 planned looks at the data: 2 interim and this final analysis. The protocol stated that O'Brien-Fleming stopping rules were to be used maintaining an overall significance level of 0.05. The study was triple-masked until the final analysis was completed. An independent statistician who performed the interim analyses for the Data Safety and Monitoring Committee has reported the p-values for the primary response variable as 0.07 and 0.065 at 18 months and 3 years. At the meeting, the final result is presented: p=0.048 for the primary outcome variable, favoring your company's investigational drug. With the study complete and a p-value <0.05, the VP for Marketing is eager to write a press release and claim superiority of the new treatment. However, the independent statistician will not agree to support a conclusion of superiority for the company's product based on these results. The VP for Marketing complains the independent statistician is completely unreasonable and looks to you for support. "Something must be done!" the VP mutters as he considers his stock options on his smartphone. How do you respond?
6. Analyze data from a multi-center trial and draft a conclusions statement based on the results.
Data from a trial involving patients at Topeka, St. Louis, Chicago, Minneapolis, Indianapolis and Kansas City clinics require statistical analysis. The study is a comparison of 2 active treatments, with response indicated as success or failure as shown in the SAS code on the next page. The data analysis outlined in the protocol for this variable is to provide a 95% confidence interval for the odds of success with the treatment B as compared to treatment A, adjusted for clinic. Perform this analysis and provide the pooled CI for the odds ratio for successful treatment, if appropriate. What is your conclusion, based on these results?
SAS code for the data is given
proc format;
value centfmt 1='Topeka'
2='St. Louis'
3='Minneapolis'
4='Chicago'
5='Indianapolis'
6='Kansas City';
value trtfmt 0='Treatment A'
1='Treatment B';
value respfmt 0='Failure'
1='Success';
run;
data here;
input center trt response count;
format center centfmt. trt trtfmt. response respfmt.;
cards;
1 0 0 22
1 0 1 32
1 1 0 26
1 1 1 26
2 0 0 16
2 0 1 26
2 1 0 24
2 1 1 19
3 0 0 20
3 0 1 5
3 1 0 18
3 1 1 10
4 0 0 12
4 0 1 23
4 1 0 26
4 1 1 10
5 0 0 13
5 0 1 25
5 1 0 24
5 1 1 24
6 0 0 23
6 0 1 15
6 1 0 13
6 1 1 20
;
run;
7. It's been a day. Collect your paycheck and head home J