1. The randomized clinical trial: an unbeatable standard in clinical research? Vianda S. Stel¹, Kitty J. Jager¹, Carmine Zoccali², Christoph Wanner³, Friedo W. Dekker 1,4 1 ERA–EDTA Registry, Department of Medical Informatics, Academic Medical Center, University of Amsterdam, Amsterdam, The Netherlands 2 CNR–IBIM Clinical Epidemiology and Pathophysiology of Renal Diseases and Hypertension, Renal and Transplantation Unit, Ospedali Riuniti, 89125 Reggio Cal., Italy 3 University of Würzburg, Division of Nephrology, University Clinic, Würzburg, Germany 4 Department of Clinical Epidemiology, Leiden University Medical Centre, Leiden, The Netherlands Kidney International: series on epidemiology

2. Background Clinical epidemiological studies often investigate whether an exposure, e.g. a medical intervention, gene, or environmental factor, is related to the development or progression of a disease To address these research questions, clinical epidemiology makes use of different study designs Purpose: To discuss whether the randomized clinical trial (RCT) is indeed the gold standard among epidemiological studies

3. Types of Epidemiological Studies Case report and case series PASTFUTURE PRESENT START STUDY Assessment of exposure in one or more patients Assessment of outcome in one or more patients Assessment of exposure Assessment of outcome

4. Types of Epidemiological Studies Cross sectional study PASTFUTURE PRESENT START STUDY Selection of study population and assessment of exposure and outcome at same time

5. Types of Epidemiological Studies Case-control study PASTFUTURE PRESENT START STUDY Selection of cases and controls based on outcome Assessment of exposure in cases and controls

6. Types of Epidemiological Studies PASTFUTURE PRESENT START STUDY Prospective cohort study Prospective cohort study Composition of cohort and assessment of exposure Assessment of outcome in exposed and unexposed group Retrospective cohort study Assessment of outcome in exposed and unexposed group Composition of cohort and assessment of exposure

7. Types of Epidemiological Studies Randomized clinical trial PASTFUTURE PRESENT START STUDY Composition of cohort and allocation of exposure by randomization Assessment of outcome in experimental and control group RCT should preferably be double blind

8. Clinical example We will discuss the advantages and disadvantages of different study designs by answering the following research question based on the study of Wanner et al¹: Is the use of a statin associated to less cardiac mortality in patients with type 2 diabetes mellitus who receives hemodialysis? Exposure = use of statin Outcome = cardiac mortality ¹Wanner C, Krane V, Marz W et al. Atorvastatin in patients with type 2 diabetes mellitus undergoing hemodialysis. N Engl J Med 2005; 353: 238-248.

9. Case series and case report Answering the clinical example Therapeutic effects of simvastatin on hyperlipidemia in CAPD patients Di Paolo B, Del Rosso G, Catucci G, Vocino V, Terenzio MG, Bonomini M, Suriani K, Albertazzi A. A causal link between hypercholesterolemia due to elevated plasma concentrations of LDL and VLDL remnants of CAPD patients has been established. The effects of 24 weeks of treatment with Simvastatin, a new HMG coenzyme A-reductase inhibitor (at 20 and 40 mg/day) on serum lipid, lipoprotein, and apolipoprotein A-I and B concentrations, as well as safety parameters and subjective side effects, were evaluated in eight patients (mean duration CAPD 24.80 +/- 7.50 months, age 54.50 +/- 13.70 years). Conclusion: It appears to be a promising drug for the effective control of hyperlipemia in a large proportion of hypercholesterolemic patients, reducing their cardiovascular morbidity while on CAPD. ASAIO Trans. 1990 Jul-Sep;36(3):M578-80

10. Case report and case series Answering the clinical example Case report and case series would not provide a definitive answer to the question whether a statin reduces cardiac mortality in type 2 diabetes patients on hemodialysis An important reason for this is that the study is performed in only a few patients, without a control group. The observations may fit the hypothesis and may lead to important progress in medical scientific knowledge The study of Di et al. for instance was important to generate ideas for further studies

11. Cross sectional study Answering the clinical example Cardioprotective medication use in hemodialysis patients Miller LM, Hopman WM, Garland JS, Yeates KE, Pilkey RM …This was a cross-sectional study of a cohort of 185 prevalent hemodialysis patients… …There were 185 patients enrolled. Sixty-six (35.7%) patients had diabetes and 89 (48.1%) patients had established coronary artery disease (CAD). Those with CAD were more likely to be prescribed an ACEI or an angiotensin II receptor blocker (P=0.026), a beta-blocker (P<0.001), ASA (P<0.001) or a statin (P=0.001) than those without CAD. There were no differences in the use of these medications between diabetic and nondiabetic patients… Conclusion: Many hemodialysis patients are not prescribed cardioprotective medications. Given the high cardiovascular mortality in this high-risk population, more attention to reducing cardiovascular risk is warranted. Can J Cardiol 2006: 22 (9): 775-760

12. Cross sectional study Answering the clinical example Cross sectional studies would not provide a definitive answer to the question whether a statin is related to cardiac disease* in type 2 diabetes patients on hemodialysis An important reason for this is that we do not know whether the exposure to statins occurred before, during, or after the onset of disease as these measurements were performed simultaneously * It is not possible to study cardiac mortality as outcome, because dead people cannot be included into cross sectional studies

13. Case-control study Answering the clinical example Hypothetical case-control study: Cases: patients with type 2 diabetes mellitus on hemodialysis who died within five years Controls: patients with type 2 diabetes mellitus on hemodialysis who survived for five years The use of statin (exposure) would be determined by looking backward in time. We may find The proportion of patients using statin was lower in the cases compared to the controls

14. Case-control study Answering the clinical example Case-control studies would not provide a definitive answer to the question whether a statin is related to cardiac disease in type 2 diabetes patients on hemodialysis The main reason for this is SELECTION BY PROGNOSIS The clinician decides who will receive (statin) therapy, and his opinion about patient prognosis may guide his prescription Even after adjustment for potential confounders, it may not be possible to make a fair comparison between cases and controls

15. Cohort study Answering the clinical example HMG-coenzyme a reductase inhibitor use is associated with mortality reduction in hemodialysis patients Mason NA, Bailie GR, Satayathum S, Bragg-Gresham JL, Akiba T, Akizawa T, Combe C, Rayner HC, Saito A, Gillespie BW, Young EW …Data were analyzed from the Dialysis Outcomes and Practice Patterns Study... …Patients prescribed statins had a 31% lower relative risk for death compared with those not prescribed statins (P < 0.0001). Statins were associated with a 23% lower cardiac mortality risk (P = 0.03) and a 44% lower noncardiac mortality risk (P < 0.0001)… …Conclusion: Statin prescription is associated with reduced mortality in HD patients, providing additional support for the value of statin therapy in this patient group. Am J Kidney Dis 2005; 45(1):119-126

16. Cohort study Answering the clinical example This cohort study suggested that statins may prevent cardiac death in hemodialysis patients We should however still be careful to conclude that statin use is associated with less cardiac mortality Like in the case-control study, the main reason for this is SELECTION BY PROGNOSIS Even after adjustment for potential confounders, it may not be possible to make a fair comparison between the exposed and unexposed group

17. Randomized clinical trial Answering the clinical example Atorvastatin in Patients with Type 2 Diabetes Mellitus Undergoing Hemodialysis Christoph Wanner, M.D., Vera Krane, M.D., Winfried März, M.D., Manfred Olschewski, M.Sc., Johannes F.E. Mann, M.D., Günther Ruf, M.D., Eberhard Ritz, M.D. We conducted a multicenter, randomized, double-blind, prospective study of 1255 subjects with type 2 diabetes mellitus receiving maintenance hemodialysis who were randomly assigned to receive 20 mg of atorvastatin per day or matching placebo. The primary end point was a composite of death from cardiac causes, nonfatal myocardial infarction, and stroke. During a median follow-up period of four years, 469 patients (37 percent) reached the primary end point, of whom 226 were assigned to atorvastatin and 243 to placebo (relative risk, 0.92; 95 percent confidence interval, 0.77 to 1.10; P=0.37). Conclusion: Atorvastatin had no significant effect on the individual components of the primary end point. N Engl J Med, 2005; 353: 238-248

18. Randomized clinical trial Answering the clinical example Using a double blind RCT, any differences in outcome can reasonably be attributed to the effect of the exposure However, in their RCT Wanner et al. found no difference in outcome between the experimental and control group, and therefore they concluded that atorvastatin did not prevent the primary end point

19. Is the RCT the unbeatable standard in clinical research? For a therapeutic research question like our clinical example, the RCT is the most appropriate study design If a therapeutic study is feasible (e.g. no ethical problems, sufficient number of patients can be included, affordable, feasible follow-up period) the RCT is almost unbeatable: the problems that may occur in the other study designs do not exist or to a lesser extent using an RCT The main advantage of an RCT is that the randomization procedure helps to prevent selection bias by the clinician Hence randomization helps, but certainly does not guarantee, to make all else equal for measured and unmeasured variables, but the differences occur by chance (and not the doctors choice) Within observational studies, however, selection by prognosis may occur, and adjustment for potential confounders in the statistical analysis may not be sufficient to make a fair comparison between the groups

20. Is the RCT the unbeatable standard in clinical research? Usually, results from observational studies are needed to come to a hypothesis that can subsequently be tested within an RCT Moreover, observational data are most often more useful than RCTs for non-therapeutic studies The next article/presentation in this series will describe the strengths of observational studies in more detail