Epi 6: 1. Explain the purposes of randomization, placebo groups, and double blind procedures in a clinical trial.

Randomization: Ideally, randomization will achieve treatment group comparability on other (non-treatment) factors related to the study outcome. 1. Primary pursoe is to prevent bias in allocating subjects to treatment groups (avoid predictability). 2. Secondary purpose is to achieve comparability btwn the groups. Groups can differ on important prognostic characertersitics (possible confounding factors) stratified randomization where patients are randomized within strata of the factor (such as age gender) Placebo groups: that patients given only inert substances (e.g., sugar pills) will often show subsequent clinical improvement when compared with similar patients not so treated. This phenomenon, termed the placebo effect, must be taken into account in clinical trial design if effects in the intervention group are to be ascribed to the intervention itself and not to the generic effect of being treated. The usual method to accomplish this is to use an inert treatment that is indistinguishable from the primary intervention in the control group. Thus, the only difference between groups is the specific intervention under study. Doulbe blind procedures: in double blinding procoedures botht the patient and the investigator are prevented from knowing which treatment group the subject is assigned. It prevents observation bias and miaximizes the comparibiility of treatment groups with regard toevaluation of study outcomes -most important when reported subjective outcomes and side effects -placebo aids in study blinding Historical control trial: All pts assigned to same treatment group and compared with a historical ctrl group (previous pt who received standard treatment. Advantages: 1. requires fwere new pts than RCT 2. can be carried out when randomization not ethically feasible Disadvantages: inability to ensure comparability of treatment grps due to 1. lack of uniform inclusion citeria in ctrl grp 2. changes over time in type of pts available 3. changes over time in ancillary pt care 4. inferior data quality in ctrl group 5. differences in evaluatin response to treatment Tned to exaggerate the value of a new reatement, thus validity of positive result may b equestionable. Once getting positive result, ethically more difficult to do a RCT

2. Describe a historical control trial and its primary advantages and disadvantages relative to a randomized clinical trial (RCT)

3. Describe factorial and crossover designs, explain why they are used, and when they shouldnt be used

4. Explain the purpose of a comparability analysis in RCT

than in the first place. Randomized clinical trials (RCT): ots assugbed ti treatnebt giryos ysugb rabdinuzatuib scgene Factorial design: patients are randomly a ssigned to all the possible combinations of 2 mor more treatments. Advantage: One can estimate the effects of each treatment separetly. 2 or more treatments may be compared in single trial using same number of patients. Disadvantage: assumes no interaction between the treatments Crossover design: each pt receives both treatments and serve as his/her control/ Pts are randomly assigned to one of the 2 treatement orders Advangtage: within-pt comparison allow f or more preicision and smaller sample sizes Disadvantage: not sutibale for diseases that are unstabl, surgical procedures or long term drug therapy -cant be used if there is signifnicant carryover effecsffrom one treatment period to athe other Comparability analysis: B/c randomization doesnt insure that treatment groups are blaanced wrt to eery coufnfounding factor, it is essential to exmaie the treatment groups for comparability ofn baseline characteristics. If subjnest not compatible on certain factors, these data become improrant at the analysis stage whent reament groups are comated on the study outcome. The prognositcfactors that differ between groups are used a covariates in the statistical model -use data and stratification or regression methods to adjust for group differences If subjects are not comparable on certain factors, must account for this using stratification or regression methods at time of analysis Treatment group: Compliacne may differ btwn treatment groups due to differences in side effects. Actual treatment may not be same as treatment to which subject was randomly assigned. Patients should NOT be regrouped and analyzed according to treatment actually received, because conffoudning may result. Comllaijnt subjects ad non compliant subjects may be different on prognosticfactors. Intent to treat analysis compares the treatments groups w/o rgard to the degree of the eventual compliance. The ada is analyzed according ot the treatment assigned. It is indeded to reduce risk of comfounding . Scondar y analsys compliance based analysis lmit analysis to treatment compliant subjects A. Scientific standards If a trial is of sufficiently poor quality that it cannot make a meaningful contribution to medical knowledge, then it should be declared unethical. (Pocock, 1983)

5. Describe how data would be analyzed for patients who vary from the assigned treatment in a RCT and explain why this is necessary (intent to-treat

6. Discuss the medical-ethical considerations involved in design, conduct, analysis and reporting of clinical trials

A clinical trial: 1. Should only involve treatments for which there is scientific uncertainty about the relative efficacy of treatments. 2. Should represent an important question to answer. B. Informed consent Written informed consent using an official document approved by the facilitys Institutional Review Board (IRB). This consent includes clearly stated (lay language) descriptions of: 1. The trial design 2. Possible risks and benefits associated with the treatment 3. Any procedures conducted as part of the trial 4. Patient and investigator responsibilities during the trail C. Data Safety & Monitoring Board (DSMB) -A committee responsible for monitoring the progress of the study from scientific and ethical viewpoints. -Composed of medical experts, biostatisticians, medical ethicists. -Evaluates interim analyses ongoing recruitment and study progress, violations of study protocol, adverse experiences, possible interim analyses of study outcomes -Should the study continue, or should it be stopped?

7. Discuss advantage of using life method tables for describing clinical trial outcomes instead of event proportions or event rates based on person-time.

Statisitcial approaches to describing event data: proportion of events: Data: event occur? (yes/no) #events/#subjects H0: Proportion of events equal in two treatment/exposure groups Disadvantage: Ignores information about time until event occurrence; ignores varying lengths of follow-up. Mean survival time: Data: time to events (survival time) H0: Mean survival time is equal in two treatment/exposure groups Disadvantage: Results biased by variable/incomplete follow-up (censoring), since not all survival times are known exactly Person-time methods: Data: event rate (#events/person-time of observation) H0: Event rate is equal in two treatment/exposure groups Advantage: Corrects for variable/incomplete follow-up (censoring). Disadvantage: Information about the pattern of events (I.e., event rates in different parts of the follow-up period) is not reflected by the single summary statistic (the event rate over

the entire follow-up period) Life-table methods: Data: Survival time (for patients with events); time until censoring (for patients without events) H0: The survival curve is equal in two treatment/exposure groups Advantage: Corrects for variable/incomplete follow-up (censoring); makes inference about the entire pattern of survival (over follow-up) 8. Explain censoring as its defined in life table analysis and recognize circumstances under which a subject is censored in a clinical trial Censored patients are those who do not have the event during the study follow up, all pateitns are either censored or have the event of interest Causes of censoring: 1) lost to follow up and don't know if dea/alaive 2? Died from a competing cause of death ceonsored at date of their death 3) study follow up ends before pt experienced the event - for censored subjects , only minimal survival time is known so life table method uses data on whole cohort (both censored and noncensored) - The life table method attempts to equalize time effects through stratification, that is, subdividing the interval of observation into smaller periods to permit the estimation of the probability of the outcome of interest, frequently mortality, during each period among those alive at the start of each period Cumulative survival probability: Median survival time: time at which half of cohort has survived. Time at which cumulative survival probability = 0.50. Draw horizontal line across curve at S = 0.50 on y-axis, drop vertical line from intersection of hori. line w/ plotted curve, & read survival time on x-axis at intersection of vertical line. Median survival time in example = 12.17 years Start all subjects at time t = 0 Divide follow-up period into time intervals At each interval, calculate Cumulative Survival Probability Use all subjects data Cumulative Incidence = 1 - Cumulative Survival Probability Survival Curve: Plot Cumulative Survival Probabilities over time Tick marks representing when subjects were censored A confidence interval around survival probability at each time period. This confidence interval is function of number of subjects at risk at each time period. Since this number is decreasing over time, confidence intervals become wider (less precise) over time

9. Interpret and use graph based on life table procedures (estimate median survival times, obtain estimates of cumulative survival probability at specified follow up times, compare the reliability of short-term vs long=germ survival rates)

10. Interpret the meaning of tick mars, number of patients at risk, and confidence intervals when hey are displayed on a life table graph

number of subjects at risk at beginning of each time period. Because we are losing subjects (to death and censoring) over follow-up period, number of subjects at risk will decrease over time. 11. Recognize when appropriate to use long rank test, Wilcoxon test, and Cox proportional hazards regression for survival time data. Testing equality of survival curves 1. Log rank test: chi-square test; comparing proportions of deaths in two treatment groups at multiple, equallyweighted time points (I.e., at each time interval of the lifetable). Particularly appropriate when the relative risk for an event in the two groups is constant over the follow-up period. 2. Wilcoxon test: like a log rank test, but earlier time periods in the follow-up are weighted more heavily than later time periods (because there are more persons under observation at earlier times) Log-rank Test When the relative risk is constant over time Wilcoxon Test Earlier time periods are more heavily weighted because as time goes by, the # of subjects decreases Adjustment for covariates 1. Stratification on possible confounders stratified log-rank or stratified Wilcoxon test tests for differences in survival curves (e.g. stratify on gender), adjusted for the stratification factor. 2. Proportional hazards model (Cox regression model): a type of regression model for survival outcomes. The effects of multiple factors (treatment group, confounders) can be modeled simultaneously. Assumption: The relative risk does not change over the follow-up time