Updates in Therapeutics 2015: The Pharmacotherapy Preparatory Review & Recertification Course

Transcription

1 Updates in Therapeutics 2015: The Pharmacotherapy Preparatory Review & Recertification Course Study Designs: Fundamentals and Interpretation Kevin M. Sowinski, Pharm.D., FCCP Purdue University, College of Pharmacy Indiana University, School of Medicine West Lafayette and Indianapolis, IN Page 1-469

2 Outline Introduction What and why you need to know Validity, Bias and Confounding Clinical Study Designs Observational Interventional Clinical Trials Analysis and Interpretation Summary Measures of Effect Miscellaneous

3 Issues in Study Design Study Purpose: Descriptive vs. Analytical Time Orientation Prospective Retrospective Investigator Orientation Interventional vs. Quasiexperimental Experimental Setting RCTs Observational Trials Page 1-472

4 Relative Strength of Evidence: Hierarchy of Study Designs Page 1-472

5 Validity in Study Design Internal validity Validity within the confines of the study methods Does the study design adequately and appropriately test/measure what it purports? Does the study adequately and appropriately address bias, confounding, and measurement of end points? External validity Validity related to generalizing the study results outside of the study setting Can the results be applied to other groups, patients, or systems? Addresses issues of generalizability and representativeness Page 1-473

6 Bias in Study Design Systematic, non-random variation in study methodology and conductance introducing error in interpretation Selection bias: Arise from selection of subjects Sampling bias Observation or information bias Recall bias Interviewer bias Misclassification bias Page 1-473

7 Confounding in Study Design Variable that impacts the independent/dependent variable altering the ability to determine the true effect on outcome May hide or exaggerate a true association All relevant information should be collected and evaluated Controlling for confounding during the design of a study Randomization Restriction Matching Controlling for confounding during the analysis of a study Stratification Multivariate analysis Page 1-473

8 Causality Temporality Strength Biological gradient: Dose-response Consistency Specificity Plausibility Coherence Analogy Experiment Page 1-474

9 Types of Clinical Study Design Case Reports/Case Series Document and describe experiences, novel treatments and unusual events Hypotheses generation Example: QT interval prolongation associated with FQ antibiotics Case report: One patient Case series: > 1 patient with a similar experience or multiple case reports combined Sufficient detail to recognize same/ similar cases Is IRB approval required? Page 1-474

10 Types of Clinical Study Design Observations Study Designs Case-control Cohort Cross-sectional Design does not involve investigator intervention, only observation Observational study designs investigate association, not causation Pages

11 Observations Study Designs Case-control Determine the association between exposures (risks) and disease (condition) Classic example: Aspirin use and Reye s Syndrome Often referred to as retrospective studies Useful to study exposures in rare diseases or ones that take long periods to develop Critical assumptions to minimize bias: Cases are selected to be representative of those who have the disease Controls are representative of the population without the disease and are as identical as possible to the cases (- Dz) Information is collected from cases and controls in the same way Pages

12 Observations Study Designs Case-control Advantages Inexpensive and can be conducted quickly Allows investigation of multiple possible exposures /associations Disadvantages Confounding must be controlled for Observation and recall bias: looking back to remember Selection bias: Case selection and control matching is difficult Measure of Association: Odds Ratio Pages

13 Observations Study Designs Cohort Study Determine the association between exposures/ factors and disease/condition development. Estimation of the risk of outcome (RR between the exposure groups) and study outcome of interest in those with and without an exposure Risk of an event or development of a condition relative to exposure Classic example: Framingham study Describes the incidence or natural history of a disease/condition and measures it in time sequence Pages

14 Observations Study Designs Cohort Study: Retrospective Begins and ends in the present but involves a major backward look to collect information about events that occurred in the past Advantages: Less expensive and time-consuming; no loss to follow-up, ability to investigate issues not amenable to a clinical trial or ethical or safety issues Disadvantages: Only as good as the data available, little control of confounding variables through nonstatistical approaches, recall bias Pages

15 Observations Study Designs Cohort Study: Prospective Begin in the present and progress forward, collecting data from subjects whose outcome lies in the future. Advantages: Easier to control for confounding factors, easier to plan for data collection Disadvantages: More expensive and timeintensive, loss of follow-up, difficult to study rare diseases/conditions at a reasonable cost Pages

16 Observations Study Designs Cross Sectional or Prevalence Study Identify the prevalence or characteristics of a condition in a group of individuals Snapshot in time Advantages: Easy design, data collected at one time Questionnaire, interview, or other available information Disadvantages: Does not allow a study of factors in individual subjects over time, difficult-to-study rare conditions Pages 1-477

17 Incidence and Prevalence Incidence Measure of the probability of developing a disease Incidence rate: Number of new cases of disease per population in a specified time Measured in persons/year Prevalence Measure of the number of individuals who have a condition/disease at any given time. Point prevalence (date) vs. Period prevalence (year, month) Pages 1-478

18 Interpreting Relative Risks/Odds Ratios Estimate the magnitude of association between exposure and disease. Key point:this is not cause and effect but association RR (a.k.a risk ratio): Cohort studies OR: Case control studies (estimate of the RR) The RR and OR are interpreted on the basis of their difference from 1. If the 95% CI includes 1, no statistical difference is indicated Pages 1-478

19 Interpreting Index of Risk Direction of risk Magnitude of Risk RR OR Interpretation < 1 < 1 Negative association, RR: Risk lower in exposed group, OR: Odds of exposure is lower in diseased group =1 =1 No association, RR: Risk is the same, OR: Odds of exposure is the same > 1 > 1 Positive association,, RR: Risk is greater in exposed group OR: Odds of exposure is greater on diseased group RR OR Interpretation % reduction in the risk/odds No difference in risk/odds % increase in the risk/odds fold (or 200%) increase in the risk/odds Pages 1-478

20 PPA Study Interpretation Cases (+ stroke) n=383 Controls ( stroke) n=750 Adjusted OR (95% CI) Appetite suppressant: Women ( ) Appetite suppressant: Men 0 0 Appetite suppressant: Either ( ) PPA: Women ( ) PPA: Men ( ) PPA: Either ( ) Interpret the point estimate and 95% CI in all cases? What does the point estimate mean? What does the CI mean? Which ones are statistically significant? Pages N Engl J Med 2000;343:

21 Observations Study Designs Causation REMEMBER: association not causality Considerations when evaluating causality Statistical significance observed? Strength of the association? Dose-response? Temporal relationship? Have the results been consistently shown? Biologic plausibility? Experimental (animal, in vitro, etc.) evidence? Page 1-480

22 Observational Study Design Summary of Characteristics Study Design Exposure or Outcome? Measure of Association Major Advantages Major Disadvantages Case Report/ Case Series Outcome Generate new information about natural history of Dz ID new disease/condition Usually can t measure rates of association Case-control Outcome OR Study relatively rare diseases Not practical for studying Low cost and short duration rare exposures Inability to study multiple outcomes in one study Cohort Exposure RR Study relatively rare exposures Study temporal associations Estimate direct risk estimates Not practical for studying rare diseases Increased cost and longer duration (prospective) Crosssectional N/A Prevalence Low cost and short duration Temporal associations can t be established Adapted from Pharmacotherapy 2010;30:

23 Interventional Study Design Randomized, Controlled Trials Make intervention and evaluate cause and effect Design allows assessment of causality Minimizes bias through randomization and/or stratification Parallel vs. crossover design Crossover provides practical and statistical efficiency. Crossover is not appropriate for certain types of treatment questions. Effect of treatment on a disease that worsens during the study period Pages

24 Interventional Study Design Randomized, Controlled Trials Examples: Clinical trial: Comparison of two drugs, two behavioral modifications... Educational intervention: Online course versus lecture class format Health care intervention: RPh vs. non-rphbased health care team Pages

25 Interventional Study Design Randomized, Controlled Trials Are the results of the study valid? What were the results? Can I apply the results of this study to my patient population? Will they help me care for my patients? Other issues related to RCT Subgroup analyses Primary, Composite and Surrogate Endpoints Superiority, Equivalence, Non-Inferiority Pages

26 Randomized, Controlled Trials Subgroup Analysis Important part of controlled clinical trials Often overused and over-interpreted Many potential pitfalls in identifying and interpreting: Failure to account for multiple comparisons or adjust p-values Problems with sample size, power, classification, and lack of assessment of interaction Page 1-483

27 Randomized, Controlled Trials Primary and Composite End Points Primary end point: crucial design decision What does the following statement mean? ramipril reduces the rate of death, MI, stroke, revascularization, cardiac arrest, HF, complications related to DM, and new cases of DM in high-risk patients. Treating 1000 patients with ramipril for 4 years prevents about 150 events in around 70 patients. Was there a reduction in all the end points or just some? Are all the outcomes just as likely to occur? Why would this trial have been interested in all of these outcomes? Page 1-483

28 Randomized, Controlled Trials Composite End Points Positives for using composite end points? Problems? Difficulties in interpreting composite end points Misattribution of statistically beneficial effects of composite measure to each of its component end points Dilution of effects, negative results for relatively common component of composite end point hide real differences in other end points. Undue influence exerted on composite end point by softer component end points Averaging of overall effect Should all end points weigh the same, or death weigh more? Results for each individual end point should be reported with the results for the composite Page 1-483

29 Randomized, Controlled Trials Surrogate End Points Parameters thought to be associated with clinical outcomes BP reduction and stroke prevention LDL-C reduction and CV death reduction Statins vs. hormone replacement therapy PVC suppression and mortality reduction Surrogate outcomes predict clinical outcomes Short-duration studies with surrogate end points may be too small to detect uncommon AEs Page 1-483

30 Randomized, Controlled Trials Superiority vs. Equivalence vs. Non-inferiority Superiority: Detect a difference between Txs Typical design in a clinical trial. Equivalence: Confirm the absence of meaningful difference(s) between Txs What difference is important? Non-inferiority: Investigate whether a Tx is not clinically worse (no less effective, or inferior) May be the most effective, or have a similar effect. Useful when placebo is not possible due to ethical reasons Page 1-484

31 Randomized, Controlled Trials Non-inferiority Design: ONTARGET Telmisartan, ramipril, or combination in patients with a high risk of VDz Is telmisartan non-inferior in the incidence of CV deaths? Non-inferior difference defined as < 13% Essentials of non-inferiority design Control group (ramipril) must be effective Study similar to previous study with control (HOPE) and with equal doses, clinical conditions, and design Adequate power is essential, and usually, larger sample sizes are required. Page 1-484

32 Randomized Trial of ERT-P for Secondary Prevention of CHD in Postmenopausal Women Randomized trial of ERT-P for secondary prevention of CAD in postmenopausal women Objective: Does ERT-P therapy alter the risk of CHD in postmenopausal women with established CHD? Randomized, blinded, placebo controlled CEE mg/day plus MPA 2.5 mg/day (ERT-P) and placebo n=2763 with CAD < 80; mean age = 66.7 years Follow-up averaged 4.1 years; 82% of HRT still taking at the end of 1 year; 75% 3 years Pages JAMA 1998;280:605 13

33 Randomized Trial of ERT-P for Secondary Prevention of CHD in Postmenopausal Women End points Primary: Nonfatal MI, CHD death Secondary: Many, including all-cause mortality. Are these composite outcomes appropriate? Surrogate end point: LDL-C lowered Statistical analysis: Baseline characteristics: t-test and Chi-square Power analysis and sample size calculation Kaplan-Meier with Cox proportional hazards model, intention to treat Pages JAMA 1998;280:605 13

34 Baseline Characteristics Demographics Age, mean±sd, yrs White, % Education, mean±sd, yrs ERT-P (n=1380) 67± ±3 Placebo (n=1383) 67± ±3 p-value Statistical analysis: Baseline characteristics: t-test and Chi-square Pages JAMA 1998;280:605 13

35 Surrogate Endpoints Change in Lipid Profiles after 1 year p<0.001 Statistics: No documented test for above comparison mean LDL-C decreased.p<0.001 for the difference between groups) Pages JAMA 1998;280:605 13

36 What is appropriate test?? Mean LDL (and others) decrease? A. Wilcoxon signed rank test B. Chi-square test C. Two-way ANOVA D. Two-sample t-test

37 Randomized, Controlled Trials Composite Endpoint ERT-P Placebo HR (95% CI) Primary CHD events ( ) CHD death ( ) Any thromboembolic event ( ) Gall bladder disease ( ) Statistics: Kaplan-Meier with Cox proportional hazards model, intention to treat Significant time trend: More CHD events in the treatment group than in placebo in year 1 and fewer in years 4 and 5 Which are statistically different? Yes No Conclusions Pages JAMA 1998;280:605 13

38 Common Approaches to Analyzing Clinical Trials: Intention to treat Compares outcomes based on randomization How they were intended to be treated Treatment effects under usual conditions. Analogous to routine clinical practice in which a patient receives a prescription but may not adhere to the drug Conservative estimate (may underestimate) of differences in treatment Most common approach to assessing clinical trial results Pages

39 Common Approaches to Analyzing Clinical Trials: Per-Protocol Analysis Those who do not complete/adhere to treatment are not included in the final analysis Provides additional information about treatment effectiveness and more generous estimates of differences Subject to several issues such as lower sample size and definitions of adherence Results are more difficult to interpret Page 1-486

40 Common Approaches to Analyzing Clinical Trials: As-Treated Analysis Analyzed by the actual intervention received Subjects who were in the treatment group but did not take treatment would be analyzed as if they were in the placebo group. This analysis essentially ignores the randomization process for those who did not adhere to the study design Page 1-486

41 Systematic reviews Summary that uses explicit methods to perform a comprehensive literature search, critically appraise it, and synthesize the literature Differs from a standard literature review Key is a well-documented and described systematic review. Some systematic reviews will attempt to statistically combine results from many studies Differs from other reviews, which combine evaluation with opinions Page 1-486

42 Meta-analysis Systematic review that uses statistical techniques to summarize the results of the evaluated studies These techniques may improve on: Calculation of effect size Increase statistical power Interpretation of disparate results Reduce bias Answers to questions that may not be addressable with additional study Pages

43 Issues related to meta-analysis Reliant on criteria for inclusion of previous studies and statistical methods to ensure validity. Details of included studies are essential. Page 1-487

44 Elements of trial methods Research question Identification of available studies Criteria for trial inclusion/exclusion Data collection and presentation of findings Calculation of summary estimate Page 1-487

45 Forest plot Page Pharmacotherapy 2010;30:119-26

46 Elements of trial methods Assessment of heterogeneity Statistical heterogeneity χ 2 and Cochrane Q are common tests for heterogeneity. Assessment of publication bias Sensitivity analysis Page 1-487

47 Summary Measures of Effect Absolute vs. Relative Differences Absolute differences or absolute changes Relative differences or relative changes Absolute differences are more important than relative differences Authors highlight the differences observed in their trials with relative differences because they are larger Why? Larger numbers are more convincing Most drug advertisements (both directly to patients and to health care professionals) quote relative differences Page 1-487

48 Heart Outcomes Prevention Evaluation (HOPE) Study Randomized, double blind, placebo controlled study 9297 high-risk patients received ramipril or placebo daily; average follow-up of 5 years Primary outcome: Composite of MI, stroke, or death from cardiovascular causes Page N Engl J Med 2000;342:145 53

49 Absolute vs. Relative Differences Risk reduction of Primary End Point Proportion of Patients Placebo Ramipril % RRR 4 % ARR Time (Days) Page Data extracted and adapted from NEJM 2000;342:145-53

50 Summary Measures of Effect Number Needed to Treat Another means to characterize changes or differences in absolute risk Definition: The reciprocal of the ARR NNT = 1/(ARR) Rounded to the next highest whole num Applied to clinical outcomes with dichotomous data Cautions: Assumes the baseline risk is the same for all (or that it is unrelated to RR) Extrapolation beyond studied time duration NNTs should only be provided for significant effects Number needed to harm Page 1-488

51 NNT Application HOPE study Results NNT = 1/( ) = 1/0.038 = 26.3, rounded up to 27 Interpretation NNT for each endpoint Outcome Ramipril (%) Placebo (%) Relative Risk RRR ARR NNT Combined Death from CV causes Myocardial infarction Stroke Page N Engl J Med 2000;342:145 53

52 Reporting Guidelines for Clinical Studies The Consolidated Standards of Reporting Trials (CONSORT) Updated in 2010 Improve, standardize and increase the transparency of the reporting of clinical trials The CONSORT statement and checklist The Flow Diagram Extensions to the Consort Statement Pages

53 Reporting Guidelines for Clinical Studies Strengthening the reporting of observational studies in epidemiology (STROBE) statement Initially published in 2007 Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Initially published in 1996, renamed in 2009 Enhancing the Quality and Transparency of Health Research (EQUATOR) network Pages

54 Pharmacoeconomics Description and analysis of the costs of drug therapy to health care systems and society Identifies, measures and compares the costs/consequences of pharmaceutical products and services Outcomes research: identify, measure and evaluate the end results of health care services Types of pharmacoeconomic analyses Page 1-490

55 Pharmacoeconomic Studies Type of Analysis Outcome Objective Cost-minimization Equal Lowest cost alternative Cost-benefit Cost-effectiveness Monetary Worth the cost? Clinical Units (BP reduction, etc.) Greatest net benefit alternative Most cost-effective alternative Cost-utility Utility, QALY Greatest benefit alternative Page 1-490

56 Sensitivity/Specificity/Predictive Values Sensitivity: Proportion of true positives that are correctly identified by a test Specificity: Proportion of true negatives that are correctly identified by a test Positive Predictive Value: Proportion of patients with a positive test who are given a correct diagnosis Negative Predictive Value: Proportion of patients with a negative test who are given a correct diagnosis Page 1-490

57 Summary