research protocol vs. research practice: case studies richard t. hull, ph.d. professor emeritus...
TRANSCRIPT
Research Protocol vs. Research Practice: Case
StudiesRichard T. Hull, Ph.D.
Professor Emeritus
Department of [email protected]
(716) 759-6692
(C) 2001. All rights reserved
A talk given at the Center for Tomorrow, April 26, 2001
at the
Protection of Human Subjects Workshop
Introduction
Institutional Review Boards have as their chief, if not sole, role the protection of human subjects.
Traditionally, this function has rested on overseeing the design of research, recruitment of subjects, informed consent, investigation of incidents, and issues of compensation.
This presentation illustrates some of the more interesting ways in which research that is well designed and well vetted by IRBs can go awry.
The aim is to draw some morals that may help sharpen IRBs’ review of research protocols and monitoring of research.
Equipoise: A Key Concept
A physician who recommends to a patient that the latter enroll in a research study should be in a state of
equipoise: the belief that the result of being randomized to either arm of the study poses no
greater risk to the patient than undergoing standard treatment. Otherwise, to recommend entry in a study when the physician believes another treatment is in the patient’s best interest would violate a primary
duty to the patient: to recommend the best available treatment.
Equipoise applies to subjects
Similarly, a subject who is enrolled in a study should believe that the state of medical knowledge is such that the known risks and benefits of being assigned to the control or to the experimental arm of the study are balanced, so that no personal risk is greater given one assignment than the other.
A First Study of AZT AZT (zidovudine, retrovir, AZido-
deoxyThimidine)was the first drug to be developed to treat HIV+ individuals.
The following study sought to evaluate AZT as a potential drug to decrease the concentration of virus particles in the blood of HIV+ individuals.
It was a 2-armed, double-blind, multi-center study that used HIV+ patients as subjects.
The following slides report events that have not been recorded in published
studies. The author’s representation of them is based on journalistic reporting and discussions in a variety of IRB and
educational contexts.
Early AZT Study Design
HIV+ Subjectsto be
randomly distributed
into 2 groups
Group #1 toreceive weeklysupply of AZT
in opaquecapsules
Group #2 to receive weekly
supply of placebo
in identical opaque
capsules
Researchers to compare
levels ofviral
particlesin blood
of thesubjects in
both groups
HIV+ particlesin AZT group
are lower
HIV+ particles
in placebo group are
higher
AND
AZT isconfirmedas a usefultreatmentfor HIV-1
relatedAIDS
IF
THEN
AZT Study Results in 2 Centers
HIV+ Subjectswere randomly
distributedinto 2 groups
Group #1 received weekly
supply of AZTin opaquecapsules
Group #2 received weekly
supply of placebo in identical opaque
capsules
Researchers comparedlevels of
viralparticlesin blood
of thesubjects in
both groups
Quantity ofHIV+ particlesin AZT group
was lower
Quantity ofHIV+ particles
in placebo group was higher,
INDICATINGAZT was effective
in reducing theviral load
associated withdevelopment of
AIDS
AND
AZT Study Results in 3rd Center
HIV+ Subjects
were randomly
distributed into
two groups
#1 received weekly supply
of AZTin opaquecapsules
#2 received weekly supply
of placeboin identical
opaquecapsules
Researchers comparedlevels of
viralparticlesin blood
of thesubjects in
both groups
Concentration of
HIV-1 particlesin AZT group
and in placebo
group were equal to one
another, indicating noeffect of AZTover placebo
Why was there differences in the data from the third center? The striking differences in results raised
the question, Why did the individuals in the trial at the third center show no difference between experimental and control groups?
Before you go on to the next slide, speculate on why such a difference might have arisen.
Upon Investigation
HIV+ Subjectsrandomly
assigned into two groups Group #2 gets
weeklysupply of placebo
in identical opaque
capsules
Concentrations of
HIV -1 particlesin AZT group
and in placebogroup were equal to one
another, indicating noeffect of AZTover placebo
Researchers comparedlevels of
viralparticlesin blood
of thesubjects in
both groups
Group #1 gets weekly supply
of AZT in opaque
capsules
Subjects meetafter
pickingup weeklydosage,
mixcapsules in
abowl, and
eachdraw out a
number = toweekly supply
All subjects
nowtakingsub-
clinicaldose of
AZT; groups
no longer differ
Lauritsen, 1987
Subjects’ Rationale
Believed AZT was only hope of survival. Each had 50/50 chance of receiving
AZT under the experimental design. By redistributing AZT and placebo, each
had 100% chance of receiving AZT, thus maximizing chance of survival.
Actual Effect Under original design, each subject had a
50/50 chance of receiving (what was believed to be) a clinically effective dose of AZT.
Under subjects’ “redesign”, no subject would receive what was believed to be a clinically effective dose of AZT.
If subjects at other centers had done likewise, AZT might well have been declared a failure.
Moral
An IRB must protect subjects from the effects of their own
misunderstanding; subjects must have equipoise with respect to each
arm of the experimental design.
Surfactant Study in Preemies Children born prematurely often suffer
from respiratory distress syndrome, due to the immature development of the lungs.
Their lungs lack normal concentrations of surfactant, a natural coating of mature lung tissue that facilitates the transport of O2 and CO2 into and out of hemoglobin in red blood cells.
This study was designed to see if surfactant, administered artificially to the lungs of premature neonates, would improve survival and health upon discharge from the Neonatal Intensive Care Unit (NICU).
It was a multi-center, double-blind study with two arms: preemies who received surfactant, and preemies who received placebo saline.
The following slides report events that have not been recorded in published
articles. The author’s representation of them is based on personal
communications and discussions in a variety of IRB and educational contexts.
Premature Infant Surfactant Study Design
Through parental consent, infants
of varying degrees
of prematurity were
to be recruited into a
trial of lungsurfactant derivedfrom fetal lambs.
Infants randomized
into two groups
Group #1to receive surfactantin NICU
Group #2to receive
saline placebo in
NICU
Groups to be
comparedas to
survivalin NICU
Group #1has a highersurvival rate
in NICU
Group #2 has a lowersurvival rate
in NICU
Surfactantconfirmed
as abeneficial
aid topreemies
AND
IF
THEN
Premature Infant Surfactant Study Results in 2 Centers
Through parental consent, infants
of varying degreesof prematurity were
recruited into a trial of lung
surfactant derivedfrom fetal lambs.
Infants randomized into two groups
Group #1received surfactant
in NICU
Group #2receivedsaline
placebo inNICU
Researcherscompared
groupsas to
survivalin NICU
Group #1had a highersurvival rate
in NICU
Group #2 had a lower
survival rate
in NICU
Looks likesurfactantconfirmed
as abeneficial
aid topreemies
Kwong, et al., 1985; Enhorning et al, 1985
Premature Infant Surfactant Study Results in 3rd Center
Through parental consent, infants
of varying degreesof prematurity were
recruited into a trial of lung
surfactant derivedfrom fetal lambs.
Infants randomized into two groups
Group #1received surfactant
in NICU
Group #2received
saline placebo in
NICU
Researcherscompared
groupsas to
survivalin NICU
Group #1had no highersurvival rate
in NICUthan did
Group #2;the same
percentagessurvived and
died withsaline placebo
as withsurfactant
Surfactantnot
confirmedas a
beneficialaid to
preemies
Shapiro et al., 1985
Why did the results in the third center differ from the others?
Noncompliance of subjects, the reason in the first study, couldn’t be the reason here, since neonates were passive participants.
Before going on to the next slide, hypothesize possible explanations for the difference in results from the third center involved in the trial.
Upon Investigation
Through parental consent, infants
of varying degrees
of prematurity were
recruited into a trial of lung
surfactant derivedfrom fetal lambs.
Infants randomized
into two groups
The most premature infants in Group #1
and Group #2
receivedsurfactant;the least
premature infants in Group #1
and Group #2 received
saline placebo
Group #1has no higher
survival ratein NICUthan did
Group #2;the same
percentagessurvived and
died withsaline
placeboas with
surfactant
Surfactantnot
confirmedas a
beneficialaid to
preemies
Groupscompare
das to
survivalin NICU
Resident inNICU foundhe could identify
surfactant and salineplacebo by
shaking vialsand seeing
which frothsup. Decided
to give surfactant tosickest kids
in both groups.
The Resident’s Rationale
His primary obligation was to his patients. Surfactant offered the sickest kids a better
chance of survival than did the saline. With the design, half the sickest kids would
be deprived of their best chance of survival by receiving saline placebo.
So, surfactant should be given to the sickest.
The Actual Result
Resident-as-therapist’s aims for patients at hand were realized.
Resident-as-researcher’s aims for future patients were frustrated.
If all researchers in the protocol had acted as the resident did, surfactant would have been declared ineffective.
Moral
An IRB must protect subjects from the effects of researchers’ confusion of roles, and
insure that researchers have equipoise with respect to each arm of the experimental
design and do not abandon equipoise and move from the research role to the
therapeutic role with subjects.
Breast Cancer
Strikes 100,000 women in the US annually, and several million women world-wide.
Standard treatment was the Halstead radical mastectomy, in which the entire breast, lymph nodes under the arm, and pectoral muscles underneath the breast are removed.
Standard treatment left women disfigured and with swelling of the upper arm and decreased movement.
A controversy arose among surgeons in the early 1970s over whether less radical surgical treatment might have equally good results expressed in terms of survival of the disease.
Bernard Fisher designed a multi-center protocol to answer this question.
Bernard Fisher’s Radical Mastectomy vs. Simple
Mastectomy Trial
Marquis, 1989
RADICALMASTECTOMY
PatientPool
MD askspt to
participate
MD doesn’t
ask
Pt givesinformedconsent
Pt refuses
RANDOMIZATION
SIMPLE MASTECTOMY
RADICALMASTECTOMY
Results of Fisher’s First Study 1765 patients were enrolled in the study. After 10 years, there was no better
disease-free survival with radical mastectomy than with simple mastectomy.
As a result, many women have been subsequently spared unnecessarily deforming surgery.
Marquis, 1989
Participating surgeons had equipoise; they believed, prior to the study, that there was no good reason to prefer one treatment over another.
Surgeons could thus respect their patient’s freely-given, informed consent to participate under randomization without violating any of their moral duties to patients.
Thus, the study was conducted ethically and with truly beneficial results.
Marquis, 1989
Fisher’s Study of Simple Mastectomy vs. Lumpectomy
Another question emerged in the 1970s, namely, whether equally good results could result in early cases of breast cancer from removal of only the tumor and a small surrounding margin of breast, with or without radiation therapy.
Fisher designed a second study to answer this question.
Marquis, 1989
Fisher’s Second Study
Pt w/ tumors
<4 cm
MD askspt to
participate
MD doesn’t
ask
Pt givesinformedconsent
Pt refuses
RANDOMIZATION
Lumpectomyw/ radiation
Other Treatment
Lumpectomyw/o radiation
Other treatment
Simple Mastectomy
OtherTreatment
after Marquis, 1989
Results of Second Study
Enrollment in the study was so slow that the study could not be completed as designed, because:
Women who were faced with the consequence of either radical mastectomy or lumpectomy being decided by a randomization device would not consent.
Modification of the Study
Statistician Marvin Zelen thought the reasons for poor enrollment might involve uncertainty of subjects about what treatment would result from randomization.
He suggested randomization prior to recruitment.
Fisher’s Third Study: a Prerandomized Clinical Trial
RANDOMIZATION
MD chooses
to participate
MD doesn’tchoose toparticipate
Lumpectomywith
radiation
Lumpectomy w/o
radiation
Simple mastectomy
Informedconsent
requested
Informedconsent
requested
Informedconsent
requested
YES
YES
YES
NO
NO
NO
Lumpectomyw/ radiation
Other Treatment
Lumpectomyw/o radiation
Other treatment
Simple Mastectomy
OtherTreatment
Pt w/ <4 cm.lump
Marquis, 1989
Results of Third Study A sufficient number of subjects enrolled. Study established that when both
lumpectomy groups were combined, disease-free survival after mastectomy was no better than disease-free survival after lumpectomy.
Study also established that disease-free survival after lumpectomy with radiation was better than mastectomy.
Why Did the Third StudySucceed in Enrolling a Sufficiently
Large Sample? The two studies are identical in terms of
known risks and benefits, so women should enroll in the third study no more frequently than in the second.
Thus, the third study ought to have had no greater enrollment unless it presented to potential participants a different characterization of the risks and benefits.
Marquis, 1989
Before going on to the next slide, speculate on the possible explanations of why the third study succeeded in enrolling sufficient numbers of participants, while the second study did not succeed.
Speculations as to the answer
No woman was confronted with choosing to participate in a trial in which she would be randomized between lumpectomy and mastectomy groups.
No physician had to approach a patient with what may seem to be a gambling device.
Physicians could choose to recommend participation or non-participation based on a judgment of the patient’s best interests.
Patients would feel more comfortable in consenting to be in the study because they would know what treatment they would receive before consenting to participate.
Marquis, 1989
Prerandomization mimics the purely therapeutic context; the physician can “assist” the patient’s decision by emphasizing the advantages of the treatment to which the patient has been assigned.
Marquis, 1989
Ethical Problems Physicians recommending to their patients
participation in a randomized clinical trial with two or more arms should have equipoise - should not, in their best medical judgment, believe one treatment to be better for the patient than another.
This trial, however, permitted physicians to pretend that they were recommending the treatment to which their patient had been prerandomized as in the patient’s interest over other possible treatments.
Patients’ have a tendency to “therapeutic misconception” (tendency to read consent forms, sign them, and still believe the treatment they are receiving was chosen for them because their doctor thought it best was not countered in this study).
So, this trial involved patients being randomized to a therapy they would view as much less desirable (because of the differing, permanent consequences for their bodies) in the absence of their physicians’ belief that the less desirable treatment offered a therapeutic advantage.
Marquis, 1989
Hence, genuine informed consent was not achieved, explaining why prerandomi-zation succeeded in this case: doctors violated their duties to their patients by not ensuring their under-standing that the treatment they were to receive was chosen for them by a random selection device, not by their physicians.
A Paradoxical Result
This study had results that were of benefit to hundreds of thousands of women who now can choose less disfiguring surgery knowing they are not reducing their chances for survival.
Based on the failure of the comparable previous study, these results could not have been gained without violating the participants’ rights.
Possible Moral #1 An IRB must insure that the experimental
design does not surreptitiously violate a patient’s right to a fully-informed, freely-given informed consent through design
variations such as prerandomization, and must resist the temptation to subordinate duties to the subject to the beneficial aims
of the research.
Possible Moral #2 An IRB should be prepared, in cases where highly beneficial results of a study
cannot be gained in any other way, to subordinate the research’s duties to the
subject to the goals of the study by permitting such devices as
prerandomization to be employed when straightforward ethical design fails to
obtain sufficient participation.
What is your choice?
Moral #2 is in violation of the federally mandated principles of research, which require that the subject’s autonomy always be respected even when achieving a result of great public benefit would result from sacrificing the autonomy of an individual subject.
In an informal survey at this conference, 40% of IRB members asked which moral to draw from Fisher’s third study, chose the second moral.
References
G. Enhorning, A. Shennan, F. Possmayer, M. Dunn, C. P. Chen, & J. Milligan. 1985. “Prevention of neonatal respiratory distress syndrome by tracheal instillation of surfactant: a randomized clinical trial. ” Pediatrics 76 (2): 145-153.
B. Fisher, Madeline Bauer, Richard Margolese, et al. 1985a. “Five-year results of a randomized clinical trial comparing total mastectomy and segmental mastectomy with or without radiation in the treatment of breast cancer.” New England Journal of Medicine 312 (11), 674-681.
B. Fisher, Carol Redmond, Edwin R. Fisher, et al. 1985b. “Ten-year results of a randomized trial comparing radical mastectomy and total mastectomy with or without radiation.” New England Journal of Medicine 312 (11), 674-681.
M. S. Kwong, E. A. Egan, R. H. Notter & D. L. Shapiro. 1985. “Double-blind clincial trial of calf lung surfactant extract forthe prevention of hyaline membrane disease in extremely premature infants. ” Pediatrics 76 (4): 5895-592.
J. Lauritsen. 1987. “AZT on trial.” New York Native, 19 October, paragraph 31. [http://www.virusmyth.com/aids/data/jltrial.htm]
D. Marquis. 1989. “An Ethical Problem Concerning Recent Therapeutic Research on Breast Cancer.” Hypatia 4 (2), 140-155.
D. Marquis. 1986. “An argument that all prerandomized clinical trials are unethical.” Journal of Medicine and Philosophy 11 (4), 367-384.
D. Marquis. 1983. “Leaving therapy to chance: An impasse in the ethics of randomized clinical trials.” Hastings Center Report 13 (4), 40-47.
D. L. Shapiro, R. H. Notter, F. C. Morin 3rd, K. S. Deluga, L. M. Golub, R. A. Sinkin, K. I. Weiss, & C. Cox. “Double-blind, randomized trial of a calf lung surfactant extract administered at birth to very premature infants for prevention of respiratory distress syndrome. ”Pediatrics 76 (4): 593-599.
M. Zelen. 1977. “Statistical options in clinical trials.” Seminars in Oncology 4 (2), 441-446. M. Zelen. 1979. “A new design for randomized trials.” New England Journal of Medicine 300 (22), 1242-
1245.