Clinical trials are experiments designed to evaluate new interventions to prevent or treat disease in humans. The interventions evaluated can be drugs, devices (e.g., hearing aid), surgeries, behavioral interventions (e.g., smoking cessation program), community health programs (e.g. cancer screening programs) or health delivery systems (e.g., special care units for hospital admissions). We consider clinical trials experiments because the investigators rather than the patients or their doctors select the treatment the patients receive. Results from randomized clinical trials are usually considered the highest level of evidence for determining whether a treatment is effective because trials incorporates features to ensure that evaluation of the benefits and risks of treatments are objective and unbiased. The FDA requires that drugs or biologics (e.g., vaccines) are shown to be effective in clinical trials before they can be sold in the US. The course will explain the basic principles for design of randomized clinical trials and how they should be reported. In the first part of the course, students will be introduced to terminology used in clinical trials and the several common designs used for clinical trials, such as parallel and cross-over designs. We will also explain some of the mechanics of clinical trials, like randomization and blinding of treatment. In the second half of the course, we will explain how clinical trials are analyzed and interpreted. Finally, we will review the essential ethical consideration involved in conducting experiments on people.
Lecture 3: Masking
Loading...
来自 Johns Hopkins University 的课程
Design and Interpretation of Clinical Trials
1141 个评分
从本节课中
Randomization and Masking
This week we discuss two key design features of randomized clinical trials used to protect against bias, randomization and masking.
与讲师见面
Janet Holbrook, PhD, MPHAssociate Professor, Epidemiology
Bloomberg School of Public Health
Lea T. Drye, PhDAssistant Scientist, Epidemiology
Bloomberg School of Public Health
Welcome.
This session will cover the use of masking, also
known as blinding, as a feature of a clinical trial.
The purpose of masking is to reduce information bias.
In a nutshell, it helps to ensure that we treat participants assigned to the
experimental and control treatments the same throughout
the trial and objectively evaluate their outcomes.
I'm going to go to the outline of the lecture
and the first segment is going to be on
the definition of masking, the rationale for using masking
in clinical trials and some of the advantages and disadvantages.
First masking is synonymous with the term blinding which you may be
more familiar with that is used in clinical trials to blind treatment.
But, for various reasons, I think
masking is really a better term.
What masking means is that the
treatment assignment is not known after randomization.
And so it's important to recognize that this isn't
about not knowing the treatment assignment before someone comes in.
But it's after randomization.
So, if someone's assigned a treatment, but it's
mass, so they don't know which treatment it is.
And that someone who doesn't know can be the patient.
It can be the patient and the clinical personnel.
Or include people who are evaluating outcomes in patients.
It can include data processors, analysts, and
even the independent review committee and sponsors.
And so we talk about single, double, triple and quadruple masking,
and I'll go over specific definitions of those in a few slides.
But I do want to emphasize again, this is not the same as
the concealment of treatment assignment prior to randomization.
That's what kind of design maneuver we used to prevent selection bias.
So you can still have that early concealment in an unmasked trial, that
once the person's assigned to the treatment,
everybody knows what treatment that person's on.
So that's a different thing than masking, which is once they're in the trial.
So it's sort of an extension
of that protection.
And I'll try to elaborate on that in the next few slides.
So what's the rational for keeping people unaware of
the treatment assignment during the trial is to reduce
any bias that might be related to prior knowledge
or beliefs about the treatment effects on the performance trial.
Not who gets into the trial and who gets assigned to which group.
But once they're in the trial that everyone
gets followed in the same manner.
So it promotes objectivity in the data reporting.
So when people are doing ad-hoc reports about whether
they've had an adverse event or those types of things
that come to you sort of unscheduled will be unbiased
or if people don't actually know what treatment they're on.
So, if they don't know that they're
on the aspirin treatment maybe they're not as
likely to go ahead and report Gastric distress.
Also it promotes objectivity in the data collection and follow-up.
So that the study schedule is maintained in all treatment groups To the same degree
and not somewhat influenced by knowing what treatment assignment people are on.
You might think oh, they're on the new treatment so
you better get them in more often and make sure they
don't miss visits.
So if we masked the treatments people shouldn't be influenced by that.
Also on the other treatments people may be recieving and the
test they recieve during the trial And even their own individual behaviors.
If we mask the treatments, we're more likely to have those
kind of external forces outside of the trial be the same.
Because you can imagine someone who's about to enroll in a clinical
trial may be more energized to change other
factors in their life that would affect their treatment.
Like if you were going to enroll in a trial of
hypertension, well you might decide to make some dietary modifications, because
now, you know, you're more knowledgeable about hypertension and the
staff has gotten you educated on the things you can do.
And if you were randomized to the Placebo or the control group,
let's say if it was unmasked so it's
just a control group, you're going to be on observation.
You might still be energized to change your lifestyle some.
It also promotes objectivity in outcome assessment and in data interpretation.
And by data interpretation, I'm meaning data interpretation
as we go through the trial and monitor it.
Certainly at the end of the trial we're
going to want to know what the treatment assignments are.
So what are the levels of masking?
Well there can be many but I'm going to just go over these three.
Normally when we're talking about a single
mask trial, we're talking about a trial where
the participant is masked to the treatment
but everyone else knows what the treatment is.
In a double mask trial, it's usually the
participant and the clinical investigators are masked to treatment.
So, the people administering the treatment, and
perhaps the same ones who are evaluating the
participant's outcomes are also masked.
In a triple masked trial, that can mean that the participant
and the clinical investigator and others are masked to the trial.
And those others might be independent people who evaluate outcomes.
Let's say if there are some sort of x-rays or MRI's
that need to be evaluated, that those people are also masked.
Or the data analysts are masked.
So, they just know
that there's treatment group a and there's treatment
group b, but they don't know which is which.
Or the data monitoring committee, or the sponsers.
So, what are the advantages and disadvantages of masking?
Well, I think I've just gone over the advantages.
So, that we're protecting against performance and reporting bias.
But, what are the disadvantages?
Why wouldn't we do this if it's a
good tool to help us have more objective results?
Well, the disadvantages
include, that it may be logistically or ethically impossible to do the masking.
And we'll talk about that in a few
slides when, when we talk about decisions to mask.
It also may not reflect the clinical care practices as well.
Or the real world clinical conditions where people know what treatments
they're on, and their healthcare providers know what treatments they're on.
That may not be as real world scenario.
And it certainly
increases the logistical complexity of mounting
a trial and getting the treatments together.
And with that increased complexity, you can
imagine that also goes along with increased cost.
So, I just want to go over how the level of
masking is related to the bias protection we get from masking.
So if we don't have any masking, we do have the potential
for bias in all of those areas that are listed as rows.
Data reporting, collection follow up, and so on.
Now, if we have a single masked trial
where the participants mask, well, we have some protection
against data reporting bias that patients are not, maybe, more likely to
report specific adverse events, if they don't know what treatment they're on.
But, again, maybe not any protection for data collection and follow up.
We may be able to protect somewhat
against differential behaviors in the treatment groups.
But all the other concomitant things that
go along with healthcare would not be masked.
And you can see
once we jump up to double masking, we get
a lot more protection from masking, because we are actually
not only masking the participant, but we're masking the people
who are actually conducting the trial and evaluating the outcomes.
So there's a big gain from going from single masking to double masking.
And triple masking is again when we have other
people who are involved in the data analysis and
interpretation masked to the trial as we go on.
And that's a little bit controversial whether
it's really good to have everybody masked
in a clinical trial where your giving
people drugs and you're following their health outcomes.
You know, somebody's gotta know what's going
on to protect the participants against undue risks.
So, how do we decide whether we're going to mask treatments in a clinical trial?
I've certainly been involved in a number of trials that are unmasked trials, and I
still think that they were good trials
that added to the scientific evidence about treatment.
But the first decision point is whether it's ethical to mask a trial.
Is masking the trial going to expose participants to
risks by giving them some sort of placebo treatment,
a sham surgery, or an IV infusion?
In some cases, those may be appropriate and okay but you have to think about that.
And in most cases sham surgery is probably not considered appropriate.
Also even if there isn't a risk associated
with giving the person a sham or placebo treatment.
The risk to the participant could be increased because
the investigator and caregivers don't know the treatment.
And they may need to know the treatment to really
appropriately ensure that the patient is protected against undue risk.
So maybe because of the monitoring for the adverse events.
So in that case it would be difficult to mask.
And you'll also have to have a viable
unmasking plan if you're going to put people
in an experimental situation, and not only
you're randomizing them the treatment, you're not
telling the tre, treatment to the participants
and the care providers and the clinical personnel.
You've gotta have a viable plan to unmask the treatment
that is kind of commiserate or with the risk involved.
And I'll talk about those unmasking plans.
So the next thing you have to think about, is it possible to mask the treatment?
Is the treatment has such distinctive
characteristics that it's just impossible to mask.
One treatment that's difficult to mask is zinc
supplementation because there's a very characteristic flavor with zinc.
And then behavioral interventions, like weight loss
programs and education programs, you just cannot mask.
It just wouldn't work.
And also maybe some community interventions.
So if you're doing interventions based on at
the community level, it may also be impossible to mask.
And then in terms of whats possible can you make treatments that seem identical?
So this is kind of related to
the does the treatment have distinctive characteristics?
Can you make a placebo treatment that is, looks exactly like the active treatment?
Or in the case when you may be comparing
two active treatments can you create a placebo for each
treatment so that you have a dummy treatment.
So patients who are randomized to either one active treatment
versus another will actually take two doses of medication maybe.
They'll take the active treatment and then the dummy for
the other treatment because you couldn't them to look exactly alike.
Also influencing whether you decide to mask is the
type of outcome you have chosen for you trial.
What are the primary outcomes?
If they're very objective outcomes it's probably less important to mask.
Certainly the most objective outcome is whether a participant alive or dead.
And then, in that situation it's not as likely to be influenced
by people's opinions about the treatment.
So the more the subjective the outcome, the more
you should be thinking about trying to mask the treatment.
So if you're asking for patients evaluation of how they
feel, certainly in pain studies, it's very important to mask.
But sometimes, even when the outcome can be fairly subjective
you may not need to mask, or you can maybe institute
some bias control without masking the participants
by masking the people that evaluate the outcome.
So the people that are looking at the X-rays
or the MRIs don't know the, the treatment assignment.
So they won't be influenced.
I've done a lot of ophthalmology studies where we're taken different images of the
eye, the retina, and we haven't been able to mask a number of those studies
because of the invasiveness and the different modes of treatments.
You know, in some cases we had surgery where we put medicine in peoples eyes.
But even though we weren't able to mask the treatment
at the level the participant or the clinical investigator applying the
treatment, we were able to mask the reading center that
looked at the images and evaluated how the treatment was doing.
So it's also another thing to think about
is, what's the comparison group?
Is it a no treatment observation, or an active control?
If you're giving someone an active treatment, you
create some expectation about that this will help them.
So, it may be less important if you're comparing two active treatments
to mask than if you're comparing an active treatment to no treatment.
Because, in the no treatment
group, people know they're not getting treatment and there's some meaning
to just getting treatment that may help people have better outcomes.
Certainly you have to think about whether it's feasible.
What's the cost of masking?
What's the benefit?
Are you going to be able to do it effectively
enough to be worth the cost and is it practical?
Sometimes you're requiring participants to take
multiple doses of medication, in these kind
of double dummy designs.
And so, are participants going to adhere to it?
You know, it's hard enough to get people to take
a pill each day, much less have them take four.
And you really have to think about the practicality of it.
So, as I said in considering the ethics of masking, you
have to have a unmasking plan, if you do mask a trial.
And there's two types of unmasking
that comes up. There's the unmasking that's planned.
When the participant has completed the trial you certainly
want to let them know what treatment they were on.
Especially if they may want to continue with that
treatment if they felt that they were doing well.
And so this slide goes over that kind of unmasking.
And the strategy depends on the close out design.
So you can have a common close out design when you may enroll
patients over a period of time, but you're going to follow
everybody to a common closeout within, let's say, a month.
So, people may have different amounts of follow up time, depending on when they
were enrolled in the trial, but the
trial's going to end in November of 2010 regardless.
So, at that, you can sort of plan to
shut down all operations in terms of actively following people.
And you can have a close-out
visit and sort of inform everyone, around the same time.
But more commonly we have anniversary close out.
Where participants enroll in a trial and are followed for
a specific length of lime, let's say for six months.
So, you may be enrolling for a trial over a few
years, but than each participant is only followed for six months.
So you have people finishing the trial as you're continued to enroll.
So that's a little more difficult.
Because you don't want to be revealing the treatment assignment
to the clinic staff in a double mask trial.
But you do have an obligation to reveal it to the participant.
So things we've used is letters to participants at the exit interview
that we also have a letter that goes to their study physician in.
Of the asthma trials I've been involved in, and with
instructions on how to continue the treatment if they were
pleased with it.
Also in terms of unmasking, you should plan to find out information on what the
participant and the clinical personnel thought the
treatment assignment was before you can reveal it.
It can be informative about how effective your masking regimen was.
And it's important that you document the procedure.
And that you provide the participant and
their primary care providers with as much informsh,
em, mation, as you can about their status at the end of the trial.
It may be things like giving them the
results of tests that were done during the trial.
But then you can also have unplanned unmasking
and that's when a participant is going along in
a trial and some event happens that there's
a need to know what that participant is on.
You can commonly, you would think of it as being an adverse event.
Well, the first thing is, that we discourage that.
We really don't want to reveal treatment assignments.
And most of the time, if someone has an adverse
event, you're going to stop the drug regardless of whatever else has happened.
So, the treatment is going to be to stop the drug, and see if
the adverse event goes away, and you don't really need to unmask people for that.
But there can be extenuating circumstances,
especially if all the treatments are approved.
Well you may need to know what the patient did
poorly on to decide what the next treatment should be.
And certainly in cases of overdoses where there may be real safety concerns.
And then you might need to know exactly what the
treatment is if the two year old has gotten the
bottled opened and downed 12 pills you want to know
the difference between the sugar pills and the active substance.
And so you want to have easy access that is
appropriate to the level of risk associated with the treatments
without being too easy.
So methods we've used are if it's a lot of concern about people being able to
know immediately, you could actually have the unmasking on the study drug container.
Something that they could pull off to reveal what's inside the
bottle and that would seem to work for most overdose situations.
You'd hope the bottle would be near the two year old.
You could have people at the clinics who are semi independent and
from the trial and be responsible for it.
You can contact the parties that are doing the randomization schedule.
So you have some sort of 24 hour call line,
or a website to be able to reveal the treatment assignment.
And we have used all of those.
And the best thing has been to discourage it.
There have been very few cases where we've really had to unmask people.
