Measures: Randomized Trials
HESC 401 Epidemiology
Objectives � 1)Describe and list the purposes of randomized trials
� 2) Explain the importance of selection of subjects in randomized trials and how non-randomized selection can introduce bias.
� 3) What is a comparison group and describe the different types of comparison groups.
� 4) Explain randomization and the different types
� 5) What is stratified randomization?
� 6) Describe “blinding” and the importance of blinding.
� 7) Describe crossover, factorial design and noncompliance
Randomized Trials � There are many types of study designs that can be
used to assess the relationship between an exposure (e.g. smoking) and outcome (lung cancer).
� Starting with chapter six, we will begin to look at various types of study designs.
� Generally study designs/epidemiology is divided into Observational studies and Experimental studies.
� Randomized trials fall into Experimental studies.
Randomized Trials � The randomized trial is considered the ideal design
for evaluating both the effectiveness and the side effects of new forms of intervention.
� When comparing randomized trials with other study designs, it is generally considered the “Gold Standard” in terms of study designs.
� Its design has major applicability to studies outside the clinical setting, such as community-based trials. This is why the term randomized trial is used and not randomized clinical trial.
Purposes of Randomized Trials � 1) They can be used in evaluating new drugs and other
treatments of disease, including tests of new health and medical care technology.
� For example, before Lipitor was approved to treat high cholesterol, several RCT’s were conducted to test the safety and efficacy of Lipitor.
Purposes of Randomized Trials � 2) They can be used to assess new programs for screening
and early detection, or new ways of organizing and delivering health services.
� For example, before screening is implemented in the general population, a randomized trial is conducted to whether those getting the screening compare d with those not getting the screening have a reduced risk of the disease that is being screened for.
� In the 1990’s there was a big controversy on whether women should receive mammograms at the age of 40. Randomized trials suggested it didn’t make a difference in terms of risk for breast cancer to be screened earlier, still many women supported screening at an early age.
Figure 7-1: This figure shows the general outline of a randomized trial. You have a study population, in which persons are randomly assigned into two groups. They either receive the current treatment (which maybe nothing) or the new treatment. The outcomes of both groups are either improve or do not improve.
Ethical Issues for RCTs � One ethical issue remains when conducting
randomized trials is whether to use “no treatment” at all for the control/comparison group or to use “current treatment”.
� Even though a “control/no treatment” group would “ideal” when comparing to the treatment group, it is unethical to with-hold or not give treatment to those that are ill and therefore “current treatment” is generally used as the control/comparison group.
James Lind � An interesting and important historical story related to
epidemiology and randomized trial is related to scurvy. � Scurvy was a disease that killed many British sailors.
James Lind, a British doctor in 1747 was interested in studying this.
� By chance, he found out that limes/lemons (high in Vitamin C) protected against Scurvy. So he designed a randomized trial and gave some crew members lemons/limes and others nothing. Those that received the limes faired so much better and didn’t get Scury.
� Therefore, all British sailors after this discovery were served lemon juice as part of their diet and till this day, they are called “limeys”.
Selection of Subjects � According to Gordis, “the criteria for determining
who will or will not be included in the study must be spelled out with great precision, and in writing.”
� There should be no element of subjective decision- making on the part of the investigator in deciding who is included or not included in the study.
� Any study must be replicable by others, just as is the case with laboratory experiments.
Allocation of Subjects to
Treatment Groups � Studies without Comparison
� In this type of study, no comparison is made with an untreated group or with a group that is receiving some other treatment. This is called a case series or case study.
� However, comparison is important because we want to be able to conclude that there is causal relationship between the treatment (the drug or screening) and eventual outcome (reduced diabetes or reduced cancer).
� Without a comparison group, it is really difficult to deduce whether the relationship that was observed was real or just due to the characteristics of the persons being studied.
Allocation of Subjects to Treatment
Groups (cont’d) � Studies with Comparison
� Historical controls- comparison group from the past � Example: There is a therapy today that we believe will be
quite effective, and we would like to test it in a group of patients. We need a comparison group. For comparison, we go back to the records of patients with the same disease who were treated before the new therapy became available.
� Problems with historical controls � 1) The quality of the data regarding treatment for historical
controls may not be as good or collected with the same accuracy as for the “current” treatment group.
� 2) Many things other than treatment change over time such as living conditions, quality of nutrition and this in itself could affect the results and if we don’t have information on these changes, then it’s difficult to truly assess the relationship between treatment and outcome using historical controls.
Simultaneous Nonrandomized Controls
� Due to the problems posed by historical controls and the difficulties of dealing with changes over calendar time, an alternative approach is to use simultaneous controls that are not selected in a randomized manner.
� Selecting controls in a nonrandomized fashion:
� One way is to assign patients by the day of the month on which the patient is admitted to the hospital.
� One of the goals of randomization is to eliminate the possibility that the investigator will know what the assignment of the next patient will be, because this knowledge introduces the possibility 0f selection bias.
Randomization � It is the best approach in the design of a trial.
� In simple terms, it means tossing a coin to decide the assignment of a patient to a study group.
� The critical element is the unpredictability of the next assignment.
� By randomizing, you are making sure that the two groups (the treatment and control/comparison) group are “equal” in terms of age, ethnicity, sex, etc.
How to randomize? � There are different methods for randomizing subjects. When conducting research today, many
researchers use computers to randomize subjects.
� Another option is to use the table below:
� Based on Gordis: How do we use this table? Let us say that we are conducting a study in which there will be two groups: therapy A and therapy B. In this example, we will consider every odd number an assignment to A and every even number an assignment to B. We close our eyes and put a finger anywhere on the table, and write down the column and row number that was our starting point. We also write down the direction we will move in the table from that starting point (horizontally to the right, horizontally to the left, up, or down). Let us assume that we point to the “5” at the intersection of column 07 and row 07, and move horizontally to the right. The first patient, then, is designated by an odd number, 5, and will receive therapy A. The second patient is also designated by an odd number, 3, and will receive therapy A. The third is designated by an even number, 8, and will receive therapy B, and so on. Note that the next patient assignment is not predictable; it is not a strict
alternation, which would be predictable.
� 00-04 05-09 10-14 15-19
00 56348 01458 36236 07253
01 09372 27651 30103 37004
02 44782 54023 61355 71692
03 04383 90952 57204 57810
04 98190 89997 98839 76129
05 16263 35632 88105 59090
06 62032 90741 13468 02647
07 48457 78538 22759 12188
08 36782 06157 73084 48094
09 63302 55103 19703 74741
The importance of randomization � One of the most important aspects of randomization
is that it not only “equalizes” the two groups by age, gender and ethnicity, but also it “equalizes” the two groups on factors that we may not know about the subjects such as their immune function, genetic make-up and general lifestyle.
Stratified Randomization � First, stratify (stratum = layer) our study population
by each variable (e.g. age and sex) that we consider important, and then randomize participants to treatment groups within each stratum.
� The next slide shows Figure 7-4.
� The first layer is stratified by sex: males and females.
� From those groups, the second layer is stratified by age.
� Finally, from the age groups, randomly select two groups: new treatment and current treatment.
Figure 7-4 Stratified Randomization
Data Collection on Subjects � Treatment (Assigned & Received)
� We first need to know which treatment group the patient was assigned to, and then which therapy the patient actually received.
� Example: It is important to know:
� If a patient was assigned to receive treatment A, but did not comply
� If a patient agreed to be randomized, but later changed their mind and refused to comply
� If a patient who was not assigned to receive treatment A, but took treatment A on their own, without realizing it.
Data Collection on Subjects (cont’d)
� Outcome � According to Gordis, “the need for comparable
measurements in all study groups is particularly true for measurements of outcome.” These include improvement and any apparent side effects.
� In other words, it is critical that measurement of outcome (whether it be cholesterol levels, type II diabetes or cancer risk) is consistent and based on strict guidelines/criteria for both the treatment and the control group. If the criteria for measuring outcome are not the same across both groups then it may look like the treatment group did better (the outcome ) when they really didn’t.
Masking (Blinding) � When a subject is masked (blinded), they do not
know which group they will be assigned to.
� One way to accomplish this is by using a placebo, which is an inert substance that looks, tastes, and smells like the active agent.
� However, this does not always guarantee that the patients are blinded.
� Some patients may try to determine whether they are taking the placebo or the active drug.
Various degrees of blinding � Ways to minimize these problems:
� Single-blinded study � Blinding (masking) subjects to their own group assignment (either
intervention or control/comparison) � Minimizes bias introduced by subjects � No effect on bias on part of investigators
� Double-blinded study � Neither the subjects nor the investigators are aware of subjects’ group
assignments � Overcome both sources of bias � Use a sealed code for assignments of subjects that is only broken at
conclusion of study � STANDARD for Randomized Clinical Trialss
� Triple-blinded study � Subjects, investigators, and those analyzing data are unaware of
subjects’ group assignments � Minimizes data manipulation done to support hypothesis
Crossover � Two types: planned and unplanned.
� Basically what crossover means is to “switch” the treatment and control group after a certain period of time.
� The planned crossover example shows a new treatment being compared with a current treatment.
� Subjects are randomized into these two groups.
� The subjects are observed over a certain period of time on one therapy, and after any changes are measured, they are switched to the other therapy.
The subjects are being randomized into new and current treatment
The subjects are observed over a certain period of time on one therapy, and after any changes are measured, they are switched to the other therapy.
Both groups are then observed again for a certain period of time.
Crossover (cont’d) � The planned crossover allows researchers to observe
changes in Group 1 patients while they were on the new treatment compared to changes while they were on the current treatment, and vice versa with Group 2.
� This allows each person to serve as his/her own control.
Cautions To Be Taken in Planned
� 1) Carryover: If a subject is changed from therapy A to therapy B and observed under each one, the observations under therapy B will be valid only if there is no carryover from therapy A. � The washout period (e.g. the period where the drug or
treatment is not present in the blood of the subjects) must be long enough to be sure that none of therapy A, or its effects, remains
Cautions To Be Taken in Planned
Crossover (cont’d) � 2) The order in which the therapies are given may
elicit psychological responses.
� We want to be sure that any differences observed are indeed due to the agents being evaluated, and not to any effect of order.
� 3) Planned crossover is not possible if the new therapy is surgical or if it cures the disease.
Unplanned Crossover � Figure 7-6 on the next slide shows the design of a
randomized trial of coronary bypass surgery, comparing it with medical care for coronary heart disease.
� In unplanned crossover, randomization is carried out after informed consent has been obtained.
Figure 7-6 Unplanned crossover in a randomized study of cardiac bypass surgery
Unplanned Crossover (cont’d) � However, some subjects assigned by the
randomization to bypass surgery may start to have second thoughts and decide to not have the surgery, shown in Figure 7-7 A on the next slide.
� These subjects then become crossovers into the medical care group. The condition of some of these subjects may begin to deteriorate and urgent bypass surgery may be required (Figure 7-7 B).
Downloaded from: StudentConsult (on 2 March 2009 11:42 PM)
© 2005 Elsevier
Figure 7-7 A-D
Unplanned Crossover (cont’d) � In Figure 7-7 C, the patients on the left are now
treated surgically and the subjects on the right are treated medically.
� If we want to carry out an intention to treat analysis, we would compare the groups in Figure 7-7 D. We would compare the patients according to their original assignments following randomization.
� This is a problem ,unless we analyze according to the treatment that the patients actually received.
� Basically, a factorial design is just a “combination” of testing more than one type of treatment/drug.
� Figure 7-8 on the next slide show the factorial design for studying the effects of two treatments.
� We are assuming that two drugs are to be tested, the anticipated outcomes for the drugs are different, and their modes of action are independent. So one can use the same study population for testing both drugs.
Figure 7-8. Based on the figure above, you can see that two drugs are being tested simultaneously. This is the advantage of a factorial design, because if there are two drugs that might be effective, it is advantageous to study them together instead of spending lots of money, time and effort to measure them separately.
Noncompliance � Patients may, at first, agree to be randomized.
However, they may not comply with the assigned treatment.
� Noncompliance can be either overt or covert.
� Overt: Some people will simply voice their refusal to comply or may stop participating in the study. These non-compliers are also called dropouts from the study.
� Covert: Others will just stop taking the agent assigned without admitting it to the investigator /staff of the study.
Noncompliance � Noncompliance can be a real issue for many studies,
therefore many trials work vigorously to make sure that either the participants are being compliant or to measure the compliance.
� Some ways to measure compliance is to count the number of pills the participants have or to measure their blood work and to have regular follow-up interaction/calls with the intervention group.