Experimental Study Design: Randomized Controlled Trial Sakda Arj-ong Vallipakorn, MD. , PhD. Pediatrics, Pediatric Cardiology, Pediatric Emergency Medicine Emergency Medicine and Family Medicine Pawin Numthavaj, MD Section for Clinical Epidemiology and Biostatistics Faculty of Medicine Ramathibodi Hospital Mahidol University History of Clinical Trials

• 562 BC: • King Nebuchadnezzar ordered his men to eat only meat and drink but some young men objected and eat vegetables • Vegetarians better nourished than meat-eaters 1747: and Trial

• Surgeon on a ship • Scurvy among sailors

On the 20th of May 1747, I selected twelve patients in the scurvy, on board the Salisbury at sea. Their cases were as similar as I could have them. They all in general had putrid gums, the spots and lassitude, with weakness of the knees. They lay together in one place, being a proper apartment for the sick in the fore-hold; and had one diet common to all, viz. water gruel sweetened with sugar in the morning; fresh mutton-broth often times for dinner; at other times light puddings, boiled biscuit with sugar, etc., and for supper, barley and raisins, rice and currants, sago and wine or the like. Two were ordered each a quart of cyder a day. Two others took twenty-five drops of elixir vitriol three times a day … Two others took two spoonfuls of three times a day … Two of the worst patients were put on a course of sea-water … Two others had each two oranges and one given them every day … The two remaining patients, took … an electary recommended by a hospital surgeon … The consequence was, that the most sudden and visible good effects were perceived from the use of oranges and ; one of those who had taken them, being at the end of six days fit for duty … The other was the best recovered of any in his condition; and … was appointed to attend the rest of the sick. Next to the oranges, I thought the cyder had the best effects … (Dr James Lind's “Treatise on Scurvy” published in in 1753) Definition of RCT

• Experiment in which individuals are randomly allocated to receive or not receive an •experimental diagnostic, •preventive, •therapeutic, •or palliative procedure and then followed to determine the effect of the intervention Phases of

Phase Phase Phase I Phase II III IV • First in testing • 30-100 patients • 100+ patients • Postmarketing new • Preliminary • Efficacy and • Observational/ intervention in safety, safety, safety controlled trial human efficacy • Usually • Effectiveness, • 10-30 healthy controlled side effects people • Identify dose, drug metabolism Types of Trial Design

1. Parallel 2. Cross-over 3. Group allocation (Clustered) 4. Factorial design 1. Parallel Design

• Treatment & control group • Each person is randomly assigned to one treatment group • Randomization: reduce treatment selection bias Parallel Design

Patient

Randomization

Treatment A Treatment B

Outcome Outcome Measurement Measurement 2. Crossover Design

• Randomization of order in which treatments are received • AB or BA

• Patients serves as their own control • Fewer patients needed • Use in chronic disease with constant intensity of underlying disease • Look for short-term treatment effects

Crossover Design Patient

Randomization

Treatment A Treatment B

Outcome Outcome Measurement Measurement Washout

Treatment B Treatment A

Outcome Outcome Measurement Measurement Disadvantages of Crossover Design

• Treatment cannot have permanent effect • Potential “carry-over” effect • Need for “washout” period • Dropouts are more significant • Analysis may be more difficult 3. Group Allocation Design

• “Cluster” design • Unit of randomization = “Group of patients” (Community, hospital, school)

• Use when individual randomization is not feasible ex. contamination Group Allocation Design

Intervention A Intervention B Intervention B Intervention A 4. Factorial Design

• Test two interventions simultaneously • Economical way • Determine treatment interaction • Assumption: treatments are independent (no interaction) Factorial Design

Treatment B + Treatment B -

A, Treatment A + A & B A, Not B with or without B

Not A, Treatment A - B, Not A Not A & Not B with or without B

B, Not B, With or without A With or without A Hypothesis of trial

• Superiority Trial • Equivalence Trial • Non-inferiority Trial Superiority Trial

• Null Hypothesis (H0) • No difference between treatments • A = B

• Alternate Hypothesis (HA) • There is a difference between treatments • A ≠ B Equivalence Design

• Objective: Show that intervention response falls sufficiently close to control group response • “Detectable difference”

• Null Hypothesis (H0) • There is a difference between treatments • A ≠ B • Alternate Hypothesis (HA) • There is no difference between treatments • A=B Non-inferiority Design

• Objective: determine if the treatment is at least as good as standard treatment

• Null Hypothesis (H0) • Treatment A is worse than treatment B • A < B • Alternate Hypothesis (HA) • Treatment A is equal to or superior to treatment B • A ≥ B • Hypothesis of non-inferiority is one-sided • Sample size is smaller than equivalency trial The role of Δ in superiority, equivalence and non-inferiority trials. Schumi and Wittes Trials 2011 12:106 doi:10.1186/1745-6215-12-106 Non-inferiority design confidence interval

Schumi and Wittes Trials 2011 12:106 doi:10.1186/1745-6215-12-106 Outcomes in RCT

• Primary Outcome • Secondary Outcome

• Measurement of Outcome • Dichotomous • Time-to-event • Rates • Composite Measures Primary Outcome

• Reflects primary objective, primary hypothesis • “Design” variable • Sample size is based on this outcome

• Should be relevant and likely to be influenced by treatment • Measurement should be accurate and reliable • Evaluation in all participants Secondary Outcome

• Other important potential effects • Safety • In-between mechanism of effect Different types of outcome measurement

• Dichotomous outcome • Yes or No • Disease or No disease • Cutoff value for continuous measurement

• Time-to-event outcome • Dimension of time in addition to dichotomous outcome • Ex: time to death, time to hospital discharge • Allow for censoring (some patient loss follow up) • Rates • Dichotomous, but allow repeated outcome

• Composite measurement • Two or more events related to disease process

• Continuous variable • Value or changes from baseline • Typically more powerful than discrete outcome • Therefore, smaller n needed Efficacy or Effectiveness

• Efficacy: in ideal condition/best circumstance • Effectiveness: in real situation

• Example • Asthma trial • Efficacy measurement: FEV1 • Effectiveness measurement: hospitalization rate • Vaccine trial • Efficacy measurement: Confirmed cases of dengue infection with virological studies • Effectiveness measurement: Clinical cases Propose of Randomization

• Protection from “Selection bias” • Set two groups to be “comparable “in both known characteristics that can be measured and unknown and unmeasurable characteristics • By following methods: Random

• Process in which there is a probability associated with every legitimate outcome • Coin toss • Dice throw Rationale for Randomization

• Avoid selection bias • Comparable treatment group with known or unknown confounders • Assures statistical tests will have valid significance level • Defined time-point for trial entry Randomization Scheme

1. Simple Randomization 2. Restricted Randomization • Block Randomization • Stratification 3. Adaptive Randomization 1. Simple Randomization

• Complete randomization • Every participant has same chance to be in the treatment group • Each new participant is independent of previous participants assignment • Ideal statistical randomness Methods for Generating Randomization Number

Recommended • Random number table • Computer generated

Methods that are not recommended: • Dice • Coin-toss • May become non-random and do not leave an audit trail

35 Random Number Table

36 • Advantages: • Each assignment is completely unpredictable • Long run: tends to produce similar number of patients in both groups • Risks of simple randomization • Imbalance of number of participants in each group • Imbalance of number of participants with and without confounding factors • Less number of participants – higher risk of imbalance 2. Restricted Randomization

• Blocking • Block: list of treatment assignments that achieve the allocation ratio • A:B 1:1 • Block of 4 should have 2A:2B • Possibility: AABB ABAB ABBA BABA BAAB BBAA • Block of 6 should have 3A:3B • Possibility: AAABBB … • Randomization: Random block possibility Treatment A = 7 Treatment B = 5 Treatment A = 6 Treatment B = 6 • Advantages • Ensure balance of treatments over time • Protect against time-related changes e.g. learning curve of surgeon • If trial is stopped early, we still have balanced groups • Analysis are more powerful • Disadvantages • Prediction of next assignment • Especially unblinded trial • Stratification • Ensure balance in treatment assignments within subgroups • Site of study • Gender • Age group

• Disadvantages: too many strata may lead to imbalances in overall treatment allocation M 1:3 F 3:1

= Treatment 3. Adaptive Randomization

• Probability of assignment to treatment group is based on current balance and/or composition of the group Minimization • First patient is random • Next patient assignment is based on the treatment that yields the smallest imbalance

Play the winner • Change allocation ratio to favor the better treatment based on the primary outcome • Preferentially assigns patients to better treatment • Need to be able to assess outcome of treatment quickly Allocation Concealment

• Technique of ensuring that implementation of the random allocation sequence occurs without knowledge of which patient will receive which treatment • knowledge of the next assignment could influence whether a patient is included or excluded based on perceived prognosis.

46 47 Rationale for unequal randomization

• Why 2:1, 3:1, 4:1 ?

• Acquire as much experience with new treatment as possible • Incentive fore recruitment – patient has more chance to get study drug • Cost issue Everything is not as simple as it seems

• In many times, patients decided something different that proposed protocol • Protocol violation Example

50 Intention-To-Treat

• ITT analysis includes every subject who is randomized according to randomized treatment assignment. It ignores noncompliance, protocol deviations, withdrawal, and anything that happens after randomization • ITT analysis is usually described as “once randomized, always analyzed”. • Analyzed data based purely on randomization • Ignore • Ineligibility • Complete nonadherence • Treatment termination • Treatment switches • Partial adherence

Perspect Clin Res. 2011 Jul-Sep; 2(3): 109–112 52 Subgroup analysis in RCT

• Check for consistency of treatment effect across characteristics • If significance tests are performed to assess effect between multiple subgroups, Type I error rate is inflated • Probability that p-value is <0.05 in one or more subgroups is greater than 5% by chance alone • More subgroups, more significance effect can be found Guidelines for subgroup analysis

• Pre-specification of subgroup analysis • Calculation of sample size for subgroup analysis • Reporting number of subgroup analyses that were performed • Adjustment for multiple comparison • Reporting of confidence interval instead of just p Value Blinding method

• Blinding (Masking) • Treatment assignment is not known after randomization • Reduce bias related to prior knowledge or beliefs about treatment effects on the performance of the trial and on the reporting of outcome • Not the same as concealment • Single blind: participant • Double blind: participant + clinical investigator • Triple blind: participant + clinical investigator + others (evaluators, data analyst, …)

• Best to show who is blinded to treatment Advantages/Disadvantages of Blinding

Advantages Disadvantages • Protect against performance / • May be impossible reporting bias • May not reflect real clinical care practices (concomitant) • Increased complexity, cost Selection of Control Therapy

• Placebo • More efficient in determining efficacy • Use when standard of care is not universal

• Active control • Risk of harm in absence of therapy • Use when effective therapy exists • Clinically relevant question is “whether the new therapy is better than standard”, not “is better than nothing” Regulation of Trials

• Transparency in conducting and reporting of Trials • Clinical trial register – register as you are going to conduct the trial • Reporting of trial – use standardized reporting of trial: CONSORT Clinical Trial Register – ClinicalTrials.gov Reporting of RCT: CONSORT

http://www.consort-statement.org/consort-2010

Ethical Issue

• Institutional Review Boards (IRB) or Ethics Committee (EC) • Protection of Human Subjects • Institution has the responsibility “for protecting the rights and welfare of human subjects of research conducted at or sponsored by the institution” • IRB approval of research is necessary before first patient enrollment IRB (EC) Submission

• Research question and rationale • Protocol • Risks • Consent / assent • Recruitment procedures and material

• Ramathibodi: Intranet - http://intra.rama.mahidol.ac.th/ethic/index.php

Advantages/Disadvantages of RCT over Observational Study Design

Advantages Disadvantages • Internal validity: • External validity: • Avoid selection bias • “Selected” population • Avoid confounding • Rigid protocols • Controls similar to exposure case • Can’t evaluate potentially harmful exposure • Reduce information bias • Often too small to detect side effects • Blinding • Usually too short to see long term effect • Detect small to moderate effect • Expensive ceb-rama.org

Questions? Thank You