Working with a Statistician How and Why Karen Pieper Associate Director of Clinical Trials Statistical Operations Duke Clinical Trials Institute Why? Pre-Lecture Examination: A key reason to know a statistician is: (True or False) Question 1 They will be the life of your next party. Question 2: Statisticians display the latest in fashion. Therefore interacting with one helps keep you up-to-date in what to wear. True Question 3: Statistical support can make your grant proposal or manuscript stronger. TRUE Estimating the value of an internal biostatistical consulting service –Statist Med 2000; 19:2131-2145: “our biostatistics group returns more than $6 for each dollar spent in institutional support in 1998” 8 Question 4: Statistical Support can help insure that your results are correct and valid. TRUE Some of the Key Steps in Performing Research 1. Develop Hypotheses 2. Write Grant / Protocol 3. Develop Study Materials 4. Perform the Experiment 5. Collect the data 6. Analyze the data 7. Share results 1. Develop Hypotheses Drug A will be better at decreasing myocardial infarctions than Drug B 1. Develop Hypotheses Drug A will be better at decreasing myocardial infarctions (MIs) than Drug B 1. How do you define a myocardial infarction? 2. How do you define better; at least a 20% decrease, an absolute difference of at least 5%,… ? 3. What is the time frame for developing an MI?– 30 days, 5 years…? 4. What in what sort of patients do you expect to see Drug A work better than Drug B? 5. What about patients who die? Will you count these as MIs? 2. Write Grant / Protocol 1. Sample Size Calculations- Is it ethical to expose subjects (animals, patients, healthy humans…) to an experimental treatment if you have little to no possibility of answering the question of interest. 2. Statistical methods section 3. Definitions of key factors 4. Review 3. Develop Study Materials 1. Randomization Scheme 2. Data Collection Tools 1. (True, TRUE, TRU, T, t, 1) 3. Instructions 4. Database creation What should a statistician have picked up? What should a statistician have picked up? Appears to be w.r.t. CABG 6. Analyze the data Example: A clinical trial evaluated the treatment effect of a new drug (A) versus placebo (P) in ACS patients. The primary endpoint of the trial was 30-day death or MI. Of special interest was the effectiveness of the new drug in patients who had received a PCI versus those who had not. Sample Patient 1 Randomization PCI Death or MI 30-day Assessment Sample Patient 2 Randomization Death or MI 30-day Assessment Original Table Incidence of 1 Endpoint Eptifibatide Placebo PCI < 72 hours P Odds Ratio (95% CI) (N = 606) (N = 622) 96 hours 57 (9.4) 95 (15.3) 0.002 0.576 (0.406, 0.817) 7 days 62 (10.2) 100 (16.1) 0.003 0.595 (0.424, 0.835) 30 days 70 (11.6) 104 (16.7) 0.010 0.650 (0.469, 0.901) No PCI < 72 hrs (N = 4116) (N = 4117) 96 hours 302 (7.3) 334 (8.1) 0.188 0.897 (0.763, 1.055) 7 days 415 (10.1) 452 (11.0) 0.185 0.909 (0.790, 1.047) 30 days 602 (14.6) 641 (15.6) 0.232 0.929 (0.823, 1.048) PCI = percutaneous coronary intervention; CI = confidence interval Pieper, KS. Tsiatis AA. Davidian M et.al., Circ. 109 641-646, 2004 Sample Patient 3 Randomization MI PCI 30-day Assessment New Table Time Interval Eptifibatide (N = 606) Placebo (N = 622) Absolute Reduction P-value Before PTCA Death/MI 1.7% 5.5% 3.8% < 0.001 96 hours Death/MI* 8.1% 10.9% 2.9% 0.090 7 days Death/MI* 8.9% 11.7% 2.8% 0.105 30 days Death/MI* 10.2% 12.4% 2.2% 0.235 *Composite only includes myocardial infarctions (MI) occurring after the percutaneous intevention Pieper, KS. Tsiatis AA. Davidian M et.al., Circ. 109 641-646, 2004 Logistic Regression Example: The Linearity Assumption 0.0 Logit -1.0 -2.0 -3.0 -4.0 32 48 64 80 96 112 128 Weight (kg) 144 160 176 192 Logistic Regression Example: The Linearity Assumption 0.0 Logit -1.0 -2.0 -3.0 -4.0 32 48 64 80 96 112 128 Weight (kg) 144 160 176 192 7. Share results TMQF Committee Formation and Charge Duke Chancellor for Health Affairs Victor Dzau, MD, convened an internal Duke committee to review institutional approaches to assure the quality of discovery science destined for clinical application. A panel of external experts was also convened to review the recommendations of the TMQF Committee. The report is now available at http://medschool.duke.edu/modules/som_research/in dex.php?id=22. Faculty and staff comments are welcome at tmqffeedback@duke.edu http://medschool.duke.edu/modules/som_research/index.php?id=22 How- Timing When do I consult a statistician? It’s never too early Have realistic time expectations It takes more than a couple of days to prepare an abstract or grant. Statistician may not be free to help on the day you call for information. Be Involved in all Aspects of the Research Choice of Methods Don’t be wedded to the method that you just learned about or that is usually used in the literature. Remember the Question Fishing expeditions are expensive. The more time you take, the fewer papers the statistician can work on. Remembering the hypothesis makes the paper manageable Timeliness Every time a project is put away, extra time is needed to refresh one’s memory about the project and find all relevant documents / programs. When Statisticians love working with investigators Enthusiastic about Research Appreciative Inquisitive Take time to share knowledge B&B Organization Most faculty affiliated with specific Center, Institute, Department 17 with DCRI 13 with Cancer Center 3 with VA/DGIM 1 with IGSP (about to expand) 2 with Aging Center 2 with Radiology Department Many co-located with clinical colleagues Good mix of disease specific and methodological research Expertise in medical subject area leads to statistical issues to be resolved Some areas of expertise Clinical trials design and analysis Early phase studies Adaptive designs Sequential analysis Surrogate endpoints Pharmacokinetics/pharmacodynamics Bioavailability/bioequivalence Some areas of expertise (cont) Health Services Research Cost-effectiveness analysis Decision modeling Causal inference Predictive modeling/Provider performance Genetics, genomics, proteomics, metabolomics Biomarker development CTSA Biostatistics Core Established with funding from DTMI/CTSA Partially funds several B&B faculty members and some masters statisticians to work with those needing statistical support (currently) Original Goal: Investigators paired with statisticians develop relationships Partnership breaks off from the Core with other funding Slightly revised and expanded incubator concept: Direct more of the Core funding toward education Continue to allocate faculty and staff effort to the Core “Contract” with SBRs/Depts/Centers/Institutes for percent effort and associated expenses Goal is still to develop productive relationships Contacts Diana.abbott@duke.edu Yuliya.lokhnygina@duke.edu Elizabeth.delong@duke.edu Points to Remember Research is rewarding and fun, and requires statistics Statistics is fun (contrary to preconceived biases) Clinical research is a “contact sport”—it requires contact and collaboration with many people, including statisticians. All types of research also needs statistical support for all phases of the research Make friends with a statistician. You’ll find it will be helpful in the future