Research Design: RCT and Cohort Studies John Q. Wong, MD, MSc 22 Jun 2010
PH Course Requirement Two research proposals ◦RCT ◦Cohort No need to implement
Objective At the end of the session, the students will be able to: ◦Design a randomized controlled trial (RCT) ◦Design a cohort study
Session outline RCT ◦Example ◦Exercise Cohort Study ◦Example ◦Exercise
Design elements Research hypothesis Study design Outcome variable Exposure variable Methods of analysis Possible sources of bias and confounding
Design elements Research hypothesis ◦Clear statement Study design ◦Appropriate for the hypothesis ◦Sampling ◦Randomization Outcome variable ◦Relevant to clinical practice ◦Diagnosis test used ◦Accuracy of diagnosis Exposure variable ◦Measurement of exposure ◦Accuracy Methods of analysis ◦Sample size computation ◦Statistical test Possible sources of bias and confounding ◦Potential biases Magnitude and direction ◦Potential confounding
RCT Example
Hypothesis To assess whether supplementation with calcium cholecalciferol (vitamin D3) reduces the risk of fracture in women with one or more risk factors for fracture of the hip
Study design Randomized controlled trial Stratified randomization by practice
Outcome variable Primary outcome measure ◦All clinical fractures Secondary outcome measure ◦Adherence to treatment ◦Falls ◦Quality of life
Exposure variable Intervention ◦Daily oral supplementation using 1000 mg calcium with 800 IU cholecalciferol and information leaflet on dietary calcium intake and prevention of falls (experimental group) ◦Leaflet only (control group) Pill count
Methods of analysis: sample size computation Incidence of outcome (fractures) among unexposed (control group) = 10% Difference in outcome between groups detectable = 34% Power = 80% Level of confidence = 95%, two-tailed Dropout rate = 20%
Methods of analysis: statistical tests Intention-to-treat analysis Comparison of groups at baseline ◦Chi-square ◦T-test Survival analysis: time to first fracture
Potential sources of bias and confounding Potential bias ◦Follow-up bias ◦Selection bias ◦Information bias Potential confounders ◦Age, weight, previous fracture, smoker, health status, maternal hip fracture, fall in previous 12 months, SF-12 scores, estimated dietary calcium intake
RCT Exercise Hypothesis ◦To evaluate whether treating personal clothing/sheets worn by nomads is protective against malaria infection
Cohort Example
Hypothesis To identify determinants of vaccine- associated paralytic poliomyelitis (VAPP)
Study design Retrospective cohort study All cases of acute flaccid paralysis reported in Poliomyelitis Eradication Surveillance System (PESS)
Outcome variable Occurrence of VAPP Clinical classification criteria
Exposure variable Having received OPV between 4 and 40 days before the onset of AFP (exposed) Had not received the vaccine or who received it outside of this period (non- exposed)
Methods of analysis No sampling done Incidence of VAPP RR of VAPP occurrence in relation to ◦First OPV dose ◦Total number of vaccine doses given ◦Exposure to OPV Logistic regression
Potential sources of bias and confounding Bias ◦Selection bias due to incompleteness of the PESS ◦Misclassification bias in diagnosis of VAPP Confounders ◦Gender, fever, GI and upper respiratory signs, months of onset, extremity paralyzed, number of OPV doses received before the paralysis, type of vaccine poliovirus isolated, age in years
Cohort Exercise To determine the association between social networks and all-cause and cause- specific mortality among middle-aged Japanese