Defining the Outcomes EEL Relationship between objectives and endpoints Choosing the primary endpoint(s) Types of endpoints (outcomes) Hard vs soft endpoints Single vs multiple primary endpoints Composite endpoints Surrogate endpointsAscertainment of Endpoints SSS Every clinical trial has a primary question or questions that the trial is being conducted to evaluate. These primary questions, as well as any secondary and tertiary questions, are clearly defined and stated in advance. These objectives correspond to a study endpoint(s) for each participant. Endpoints are the post-randomization measurements required to address the intervention trial objectives.Ascertainment of Endpoints SSS8858588585885858955>5>——0—000s$q@sSomM?$*h—_—_— Have to be clearly defined and reproducible. Specified in advance, before ever see the data from the trial. Wide range: binary, continuous, ordinal, counts, time to event or length of survival. Actual level, change from baseline. In general, collect more data rather than less. Example: better to collect blood pressure, not binary hypertension.Hard vs Soft Endpoints ————SESEoxXoRE Hard endpoints: Well defined and measured without observer judgment. Example: all cause mortality, myocardial infarction confirmed by WHO criteria. Soft endpoints: Requires subjective assessments by investigator or patient. Example: level of pain, feeling better. Soft endpoints more prone to bias in ascertainment between treatment groups, especially if the interventions are not blinded; will need to build in objective criteria for assessment.Single vs Multiple Primary Endpoints SSE Single vs multiple primary endpoints: depends on the scientific question being evaluated. Example: RCT of succimer treatment of lead-exposed children aged 12-33 months. Primary endpoint: mean 1Q score on a standard test 36 months after randomization. Example: RCT of vitamin D supplementation. Primary outcomes: total cardiovascular disease, total cancer, side effects. With multiple outcomes, will affect sample size needed to answer each of the questions.Composite Endpoints Composite endpoints: single endpoint made up of multiple possible events (Example: total cardiovascular events: first ewent of myocardial infarction, stroke, CV death) Advantage: increases number of events seen in trial, thus increasing power for a given sample size. Disadvantage: don’ t know which component is primarily driving any relationship seen.Composite Endpoints SS SSS Disadvantage: if one component of the composite endpoint is disproportionately represented but not related to the intervention, will mask the true effect you are trying to see. (Example: composite outcome is all congenital malformations - but no drug causes every type of congenital malformation, may masking a true relationship with one particular rare type of malformation,Surrogate Endpoints ———— Surrogate endpoints are laboratory measurements or clinical sign or intermediate markers used as a substitute for a clinically meaningful endpoint. Advantage: Can shorten length and cost of RCT; further our understanding of pathophysiology (intermediate markers).Surrogate Endpoints ry Disadvantage: Utility depends on strength of association between surrogate and primary clinical endpoint. Changes of intervention on surrogate marker must accurately predict changes in clinically meaningful endpoint. Measure changes in an intermediate factor in the main pathway that determines clinical outcome, Examples: CD4 count or viral load, AIDS; blood pressure, stroke; but not effect of bypass grafting on graft patency rather than subsequent cardiovascular events; or hone mineral density as a surrogate for fracture risk.Minimizing Bias in Ascertainment of Outcomes Blind in as many ways as possible. Blind (participant) and double-blind (investigator) techniques (also termed masked, double-masked). Triple blind: blinding in assessment and adjudication of outcome. Blinding prevents bias due to use of co-interventions and biased ascertainment of outcomes. Adds complexity and cost, but importantly adds to the validity of the trial results.Minimizing Bias in Ascertainment of Outcomes Objective criteria (validated outcome measure) (especially important if blinding and/or placebo can't be done). Need high retention, low loss to follow-up (LFTU): bias if the response rate differs between the intervention groups, and among those who did and did not have the outcome,Validation of Endpoints ESE Endpoints should be defined ina validated, reproducible and scientifically accepted way. Exam If participant self-reports a trial clinical endpoint, request permission to obtain relevant medical records; reviewed by an Endpoints Committee of physicians blinded to randomized treatment assignment, using prespecified objective criteria (i.¢, WHO criteria for myocardial infarction) Only confirmed endpoints in final analyses; or stratify by level of confirmation (example: cause of death - medical records, death certificate with ICD code, family reported).Validation of Endpoints SSeS aaa Endpoints should be defined in a validated, reproducible and scientifically accepted way. Example: If participant self-reports a trial clinical endpoint, request permission to obtain relewant medical records: reviewed by an Endpoints Committee of physicians blinded to randomized treatment assignment, using prespecified objective criteria (i.e, WHO criteria for myocardial infarction). Only confirmed endpoints im final analyses; or stratify by level of confirmation (example; cause of death» medical records, death certificate with ICD code, family reported).Ascertainment of Variables ee Post-randomization endpoints Baseline characteristics To describe population Compare between treatment groups to assure control of confounding Potential effect modifiers of association; subgroup analyses Side effects Postulated biological mechanismsSummary a SSSasS.91 Defining, ascertaining, and validating outcomes key component of a RCT. But remember... Plan for recruitment always being less than expected. Plan for compliance never being 100%). Plan for retention never being 100°%. Main threats to the walidity of a RCT in the design phase = must be prepared to deal with these likely problems.The Statistical Analysis Plan es The statistical analysis plan (SAP) is an important part of every study protocol It describes: study endpoints and hypotheses, the data analysis plan, sample size considerations. In this segment, we introduce the elements of the SAP. Each element will be developed in later segments.Elements of the SAP gE —_______a_ Study Endpoints and Hypotheses Sample Size Considerations The Randomization Procedure The Plan for Data and Safety Monitoring Analysis Sample Analysis Methodsjis Study Outcomes and Hypotheses The SAP typically begins with a restatement of the study's endpoints and primary and secondary hypotheses. These determine the study endpoints, the methods of analysis, and finally, sample size considerations Earlier, we introduced the CORONARY Study, which compared the safety and efficacy of on-pump and off- pump coronary artery bypass surgery.Primary Outcome First primary aim of the CORONARY Study: Compare the incidence of the composite outcome, death, stroke, nonfatal myocardial infarction, or renal failure 30 days after CABG surgery. The primary aim includes a statement of the primary endpoint, a binary variable or time-to-event variable. ‘This leads naturally to the statistical models and methods that will be used in the data analysis.Analysis Sample a The Intent to Treat Sample Modified Intent to Treat Per-Protocol Analysis On assigned treatment Fully compliant with the study protocol As-treated AnalysisElements of the Analysis Plan SESE SSS SSS Methods for comparing baseline characteristics of treatment groups Methods for testing the primary and secondary hypotheses Planned subgroup analyses Monitoring for adverse eventsriers) Comparing the Groups at Baseline Goal is to document that the treatment groups are similar Demographic variables, disease characteristics, and prognostic variables should be considered Requires standard methods for comparing two or more groups (I tests, nonparametric tests, chi-square tests, exact tests)== Analyzing Study Outcomes LL ___EEEESE————EEE__—_————_—————— Most dependent variables encountered in clinical trials research are either 1) binary, 2) measured, or 3) time-to-event variables We will discuss statistical methods for each of these types of outcomes in week 8The CORONARY Trial The CORONARY Trial investigators used time-to event methods to analyze the primary outcome. These methods include estimation of the time-to-event distributions by the Kaplan-Meier method, the log-rank test for unadjusted comparison of the treatment groups, and the proportional hazards regression model for adjusted comparisons. We will discuss these concepts in the segments on data analysis.Data and Safety Monitoring The Data and Safety Monitoring Committee is responsible for monitoring accumulating safety and efficacy data Should the study continue? The Study may have a formal interim analysis planSample Size Calculations a Sample size calculations are a critical past of study planmung, Will the study be large enough to detect clinically meaningful differences between treatment groups with hugh probability? ‘The methods for sample size calculations depend on the type of study endpout. The thiee most important types ave 1) binary outcomes, 2) measured outcomes, and 3) time-to-event outcomes: We will devote the mext three segments to this important lopicSummary — Es Every study protocol should include or provide a inked statistical analysis plan (SAP) The SAP must be finalized before the investigators access unblinded data by treatment group Once the SAP is finalized, subsequent changes should be infrequent and their purpose fully documented