Public Citizen: The SUPPORT Study was Even Worse than We Thought
Michael Carome and Sidney Wolfe, 05/21/2013

Public Citizen: The SUPPORT Study was Even Worse than We Thought

(Research) Permanent link

In his April 18 Bioethics Forum article, John Lantos criticized the findings of the Department of Health and Human Services Office for Human Research Protections that the conduct of the Surfactant, Positive Pressure, and Oxygenation Randomized Trial (SUPPORT) violated HHS regulations regarding informed consent. He also attacked Public Citizen’s position that the conduct of the study was unethical because parents of babies enrolled in the study were not informed of the purpose, nature, and risks of the research.

Lantos claimed that OHRP and Public Citizen were “deeply misinformed” about the design of the SUPPORT study. We acknowledge that as of April 10 when we first wrote to HHS Secretary Kathleen Sebelius about this study, Public Citizen only had access to limited excerpts from the study protocol and one institutional review board-approved consent form that were presented in OHRP’s March 7, 2013, letter to the University of Alabama at Birmingham (UAB).    

We subsequently obtained the complete SUPPORT study protocol and consent forms that were approved by 22 IRBs and used at approximately two dozen neonatal intensive care units . Our analysis of these documents not only affirms the appropriateness of OHRP’s determination that the UAB consent form failed to mention the reasonably foreseeable serious risks of the oxygen experiment, but it demonstrates that the consent form deficiencies were far more significant than those discussed in OHRP’s letter. Indeed, the important, material factual omissions from the consent forms, combined with mischaracterizations of the study, misled the parents of babies enrolled in the study about the true purpose, nature, and risks of the research. 

Study Overview

The study involved two simultaneous complex experiments. In one experiment, the babies were randomly divided into two groups, each receiving a different treatment to assist breathing. Babies in one group were treated with a face mask, called a continuous positive airway pressure (CPAP) mask. Babies in the other group were intubated and given the drug surfactant, which helps lungs stay open, and placed on mechanical ventilation.

For the other experiment, babies in both the CPAP and mechanical-ventilation groups were further randomly divided into a low-oxygen group and a high-oxygen group (the oxygen experiment). For the low-oxygen group, the investigators tried to maintain oxygen saturation levels in a low target range (85% to 89%), and for the high-oxygen group in a high target range (91% to 95%), rather than adjust each baby’s oxygen levels within the broader range of 85% to 95% — identified as the “normal” range by the investigators — to meet the baby’s individual needs, as would have been the case if the baby had not been in the study.

Consent Form Deficiencies

The serious deficiencies in the SUPPORT study consent forms become readily apparent when information in the protocol, as well as statements made by the investigators, is compared with information in the consent forms.

The oxygen experiment’s purpose was to test the hypothesis that “relative to infants managed with a higher Sp02 range . . . the use of a lower Po02 range (85% to 89%) will result in an increased survival without the occurrence of threshold ROP [retinopathy of prematurity, which can  cause blindness] and/or the need for surgical intervention.” Consistent with this hypothesis, the primary endpoint was a composite of death and ROP. Knowing that death was a key endpoint was critical for understanding the true purpose of the experiment.

In contrast to the protocol, while most of the consent forms noted that the purpose of the oxygen experiment was to assess the effects of the two oxygen targets on the incidence of ROP, none disclosed that another key purpose was to determine whether babies were more likely to die in the low- or high-oxygen group. The incomplete descriptions of the purpose of the oxygen experiment in the consent forms were therefore misleading.

The protocol summarized the oxygen experiment’s design as follows: “A prospective comparison of a lower SpO2 range (85% to 89%) with a higher more conventional SpO2 range (91% to 95%) until the infant is no longer requiring ventilatory support or oxygen.”

Thus, the investigators themselves considered the high oxygen saturation target range to be “more conventional” treatment for premature babies receiving routine standard of care, which implicitly meant that the low oxygen saturation target range was more unconventional.

This clear protocol characterization of the relative difference between the low- and high-oxygen targets contrasts sharply with the statements in most of the consent forms implying that both ranges were equally conventional, such as the following from the UAB consent form: “[Both target] saturations are considered normal ranges for premature infants.”

Condensed, incomplete descriptions of oxygen treatment, such as the one above, were misleading because they lacked an appropriately detailed explanation of the usual standard of care in critically ill premature infants and the segment of the broader 85% to 95% “normal” target oxygen saturation range that was most commonly targeted as part of routine NICU care at the medical centers conducting the study. Of note, one consent form suggested that an oxygen saturation range of 88% to 92% was most commonly targeted in many NICUs, whereas another stated that the saturation was usually targeted at 88% to 94%. Disclosures of the oxygen saturation ranges most commonly targeted for premature infants at a particular institution should have been included in all consent forms to allow parents to have better understood how the experimental interventions would have altered the care of their babies.

A more disturbing component of the oxygen experiment was the procedure under which the medical team caring for each study baby was according to the protocol, intentionally given inaccurate information about the baby’s oxygen saturation levels by using pulse oximeters miscalibrated across the wide range of oxygen saturations between 85% and 95%. Medical teams apparently were instructed to try to maintain the oxygen saturation for study babies between displayed readings of 88% and 92%. However, for any displayed oxygen saturation level between 88% and 92%, the actual oxygen saturation was 85% to 89% in the low oxygen group and 91% to 95% in the high-oxygen group. The 5% to 6% difference in the actual saturation levels between groups represented clinically important differences in the babies’ actual blood oxygen content, which certainly could have adversely impacted clinical decisions made by the medical teams caring for these babies.

Half of the 22 IRB-approved consent forms did not disclose that the medical teams caring for babies in this experiment would be intentionally given inaccurate information about the babies’ oxygen levels. Moreover, none of the consent forms described how this experimental procedure could have impacted important clinical decisions related to the babies’ care.

The combined experimental procedures of randomly assigning infants to interventions using narrow low or high oxygen saturation targets plus the use of miscalibrated pulse oximeters represented a clear departure from the standard of care that these critically ill infants would have received had they not been enrolled in the study. Yet, the majority of the consent forms included statements assuring subjects’ parents that all research procedures used in the study were “standard of care.”

Study risks

Surprisingly, the “Risks and Benefits” section of the study protocol listed no risks of the oxygen experiment. Nevertheless, other parts of the protocol, as well as recent statements made by the study investigators, reveal that the oxygen experiment posed significant risks to the premature infants who were enrolled.

For example, the “Background” section of the protocol indicated the following: “[O]xygen toxicity can result in increased risk for [chronic lung disease], retinopathy of prematurity (ROP) and other disorders. Alternatively, oxygen restriction may impair neurodevelopment.”

Some study investigators also recently stated: “Death was included in the primary outcome because it competes with retinopathy . . .

Thus, the investigators were aware when designing the study that manipulating oxygen therapy to target narrow low and high oxygen saturation ranges in premature infants could have had different effects on these acknowledged competing risks. Targeting the high-oxygen range could have increased the risk of ROP, whereas targeting the low-oxygen range could have increased the risk of brain injury and death. Furthermore, the study’s statistical analysis plan revealed that the death rate was an important variable to be measured for the oxygen experiment.

A small minority of the IRBs that reviewed the study actually appear to have recognized some of the risks posed by the oxygen experiment because two consent forms identified ROP as a risk of the high-oxygen group. However, all of the consent forms failed to disclose that death was a risk of the research, even though an increased risk of death was reasonably foreseeable based on the information presented in the protocol.

Finally, perhaps the most troubling aspect of the consent forms was the universal failure to disclose the additional reasonably foreseeable substantial risks of harm to the infants resulting from providing the medical teams with inaccurate information about the babies’ blood oxygen saturation levels. The blood oxygen saturation level is a critically important clinical parameter monitored in such patients. This experimental procedure had the potential to adversely impact important clinical decisions, such as whether to intubate a baby and start mechanical ventilation or whether to extubate an intubated baby and discontinue mechanical ventilation. Inappropriate decisions regarding intubation and extubation certainly could have exposed babies to foreseeable risks of harm. (For specific examples of how babies could have been harmed see: Carome M, Wolfe S, Macklin R. Report prepared for Secretary of Health and Human Services Kathleen Sebelius:

The following protocol statement indicates that the investigators recognized the potential for inappropriate intubations and extubations in study infants: “CPAP infants who require intubation three times, for any criteria, will have all subsequent treatment including subsequent extubations and any further re-intubations performed using unit Standard of Care. This addition is to prevent such infants from being exposed to further protocol driven intubations and extubations.”

Even more disturbing, the majority of the consent forms included extraordinarily misleading statements like the following: “Because all of the treatments proposed in this study are standard of care, there is no expected increase in risk for your infant.”

Our conclusions

Lantos states that the “most important ethical issues for the future of clinical research . . . are the issues of honesty, transparency, and safety.” Public Citizen agrees. Disturbingly, the consent forms used in the SUPPORT study failed to present an honest and transparent disclosure of the purpose, nature, and risks of the oxygen experiment, thereby leading parents to believe the experiment was safer than it was.

It is neither OHRP’s determination nor Public Citizen’s critique regarding the study that poses a threat to biomedical research and the advancement of medical knowledge and innovation. Rather, the real threat to such scientific endeavors is unethical research, which undermines the public’s trust in the motives and conduct of researchers. Conformance with the fundamental ethical principles for conducting human subjects research must never be sacrificed in the quest to advance medical knowledge.

Lantos’s accusation that “criticism [of the study] . . . discredits bioethics” is completely without merit. The findings of consent-form deficiencies by OHRP and Public Citizen were well-founded and carefully explained. Indeed, as more information about the study has become available, the apparent degree of seriousness of the consent failures has only escalated. Attacking messengers who raise serious, well-reasoned concerns about the ethics of a study does not discredit bioethics. Instead, it discredits the attackers.

For a more detailed discussion of the problems with the SUPPORT study, see

Michael Carome, M.D., and Sidney Wolfe, M.D. are, respectively, Deputy Director and Director of Public Citizen’s Health Research Group.

Posted by Susan Gilbert at 05/21/2013 02:30:26 PM