Commentary: Safety of participants in non-therapeutic research must be ensured

When retrospectively evaluating research what matters is not the harm that actually resulted from the research, but the risk to which researchers exposed participants when all the knowledge available at the time is taken into account. At least five questions are relevant to this discussion. Was there known to have been a risk to participants before the study began, and what was the magnitude of that risk as evaluated by the evidence available at the time?

There was evidence that a reduction in the pressure of inspired oxygen might be causally related to sudden infant death 1 at the time Parkins and colleagues began their study. Is it reasonable to impose a risk of death on healthy infants to gain more knowledge about physiological responses to hypoxia? It could be argued that monitoring procedures removed this risk. Even if the study design were perfect, the chance of human and mechanical error2 could not be entirely removed.

This study is an example of non-voluntary, non-therapeutic research. It is generally accepted that the risk posed to participants by such research must be minimal. 3 4 The Royal College of Physicians suggests that participants in this type of research should be exposed to no more risk than that taken by a passenger flying on an aeroplane.3 Indeed, the justification presented by the researchers for exposing normal infants to hypoxia is that “[m]any infants travelling on aeroplanes or to holidays at high altitude are exposed to similar or even more markedly reduced partial pressures of inspired oxygen. Yet this exposure is considered safe.”

There are several problems with this argument. In the first place, researchers may have access to information which is not available to the public. Flying in an aeroplane may be more dangerous for some people—for example, those with emphysematous bullae. If an airline or responsible authority was unaware of the risks to travellers with emphysema they might allow them to travel on aeroplanes without restrictions. However, this would not provide justification for an interventional study which exposed these travellers to lower air pressures.

In the second place, even when information on risk is available some people behave recklessly; it would be opportunistic for researchers to take advantage of such behaviour. A prospective interventional study of behaviour during actual drink driving would be unethical even if resuscitation were available and there were no shortage of willing participants.

There is a related problem that occurs when judgments about the reasonableness of risk are based on assumptions drawn from behaviour. People judge that some risks are worth taking, but it is up to them to make that evaluation. Though driving a car or flying in an aeroplane does entail risk, it is wrong to assume that a person would take on this risk to participate in research. This is illustrated by the public's reaction to the scandal surrounding bovine spongiform encephalopathy. People may choose not to engage in an activity with a very small risk of death if they perceive that the benefits are outweighed by the risks. Were the parents in this study explicitly told that participation entailed a small risk to their infant's life? Participants must be scrupulously informed of such risks.

Standards of practice cannot be used to define the appropriate level of safety that should be provided to participants in research. We should look to the inherent risk. There are some concerns raised by this study by Parkins et al. Firstly, why was a saturation of ⩽80% for ⩾1 minute chosen as the criterion for ending exposure to hypoxia, and what evidence is there that it is safe to expose infants to hypoxia? Hypoxia was clearly clinically significant in some infants who were described as becoming “severely hypoxaemic.” Indeed, one required supplemental oxygen for 1 hour.

Secondly, the methods section states: “Infants and monitors were observed continuously by an experienced paediatrician. According to our protocol, exposure to hypoxia would end if oxygen saturation fell to ⩽80% for ⩾1 minute.” The results section states: “Oxygen saturation had been ⩽80% for ⩾1 minute in three infants.” It is not clear from the protocol whether there was a definite upper limit to the time an infant might spend at an oxygen saturation below 80%. How long had oxygen saturation been ⩽80% in these infants?

Thirdly, part of the reason for performing this study was because the researchers became aware of two infants who had died after travelling on an intercontinental flight. Why then did the follow up of infants exposed to hypoxia last only about 10 hours, given that one infant died 40 hours after travelling by aeroplane? Should any non-human or epidemiological research, systematic overview, or computer modelling have been done before the study to better estimate the risk to participants or to eliminate the need to use human participants?Piglets have been used as models for the physiological response of infants to hypoxia.1 Could the risk have been reduced in any other way? Researchers could have asked parents of infants who were scheduled to travel on aircraft if their infants could participate. This increases the infants' risk by increasing their total exposure to hypoxia, but these infants and their parents would have the most interest in the results of the study. The results might have been relevant: parents of those infants who tolerated hypoxia poorly might have decided not to expose their infant to the risk of air travel. Were the potential benefits of this study worth the risks? Was the study design adequate to increase understanding of responses to hypoxia in infants in aircraft and at high altitude?The authors assert that there is nothing to suggest that a reduction in the fraction of inspired oxygen in reduced barometric pressure (as occurs in an aeroplane) does not have the same effect as a reduction in the fraction of inspired oxygen in constant atmospheric pressure (as in their experiment). Yet they admit that further study during an airline flight or at high altitude (or presumably in a hypobaric chamber) will be necessary. This raises a question about the design of their study: why wasn't the study done under the conditions described above instead of exposing some infants to risk in what must be described as a preliminary study? Were the infants' parents made aware of all the relevant evidence, in particular evidence of the extent of the risk to the infants, and could the parents decide freely to participate or not based on the evidence of risk?

Concerns were expressed by the editorial committee before the paper was accepted for publication that because some parents already had a therapeutic relationship with the authors they might feel conscious or unconscious pressure to participate in the study. This is a difficult issue to evaluate because potential participants who are in a therapeutic relationship with the investigators may have the most to gain from a study and may have the strongest desire to participate for reasons of rational self interest or altruism. However, the Declaration of Helsinki requires that “informed consent should be obtained by a physician who is not engaged in the investigation and who is completely independent of this official relationship.”5

I have raised concerns in this commentary over whether the risk to the infant was fully disclosed to parents. Doctors should now have serious concerns about infants being exposed to even mild hypoxia. The study by Parkins et al addresses an important issue and will no doubt add to the information available on the effects of hypoxia on infants. A balance must always be struck between discouraging relevant research which might eliminate continuing harm and making that research as safe and ethical as possible.

Commentary: Ethical approval of study was warranted

When the research ethics committee first reviewed the project proposed by Parkins and colleagues our immediate reaction was to reject the proposal because of fears about the possible danger to infants involved in the study. After our initial discussion, however, we recognised that the study might provide important information, not only on the sudden infant death syndrome but also on the safety of air travel for infants. It was also clear that the study could not be done on participants other than infants. We decided to invite Professor Southall to attend a committee meeting to respond to our concerns about doing non-therapeutic research on infants. The potential for risk had been made clear in the original submission; we hoped that Professor Southall would provide further information on the degree of risk that it was anticipated that infants would be exposed to.

Professor Southall attended the meeting on the 26 August 1992. After the meeting, committee members were convinced of the importance of the study, and reassured about the degree of monitoring and supervision that would occur during the infants' exposure to hypoxia. We were assured that exposure would end immediately if a baby became ill or experienced an unacceptably long period of apnoea or hypoxia and that appropriate treatment would be given if required. It was also established that parents would be informed of the nature and potential risks of the study in easily understood terms and that no coercion would be used to persuade parents to allow their infants to participate in the study.

The initial protocol indicated that only families in which an infant had died of the sudden infant death syndrome or in which an infant had had an apparent life threatening event would be asked to participate. We were later requested to permit the inclusion of a control group of healthy infants who had no known risk factors. This caused further heart searching debate, but we accepted that these healthy infants would be at less risk than those from families in which an infant had previously died of the sudden infant death syndrome or had had an apparent life threatening event; the control group was also exposed to less danger than a young child would be on a transatlantic flight. The committee was satisfied that all parents would be approached in a sympathetic manner and that requests for participation would include contacting the family's general practitioner.

Committee members were fully aware of the strict guidelines on the involvement of children in non-therapeutic research. We were also concerned about the potential for harm. However, after a final discussion, and after scrutinising the modified parent information and consent forms, we were convinced that the study should be allowed to proceed. We also feel that we would make the same decision today.

Authors' reply

We considered that many healthy infants are exposed to airway hypoxia without apparent difficulties while travelling on airline flights or during holidays at high altitude when we assessed the risks that infants between the ages of 1 and 6 months would be exposed to in our study. It is not thought of as reckless to take infants on aeroplanes or on holidays at high altitude; no guidelines state that healthy infants should not be exposed to these activities.

Reviewing the literature in 1992 we found that non-indigenous infants born at altitude were at an increased risk of sudden death and mountain sickness.1 We had also undertaken2 and were aware of studies3 linking airway hypoxia to apparent life threatening events. We also knew of two infants who had died of the sudden infant death syndrome shortly after an airline flight. We thought that by studying healthy infants in an environment of controlled hypoxia we might be able to elucidate issues relevant to the sudden infant death syndrome, apparent life threatening events, and the effects of respiratory infection. We did not believe that this information could be obtained through animal experiments (such as those mentioned in the commentary by Savulescu; these were published 3 years after our study began).

Research on children with cystic fibrosis has shown that hypoxia at sea level can accurately predict oxygen saturation during air travel.4 Other studies have examined oxygen saturation at high altitude but mainly in indigenous populations which have a genetic adaptation to living in hypoxic conditions.5 We considered performing our study in a hypobaric chamber but felt that this would cause difficulties in monitoring the infants, and might increase the risks to the infants because of difficulties in access.

Asking parents of infants who were scheduled to fly on aircraft to participate in the study might have created alarm or anxiety in parents before any results were known. Access to information about infants who are scheduled to fly is protected and difficult or impossible to obtain.

The facts about the study and its risks were presented clearly to the families. Parents were initially contacted by letter from a doctor who was not involved in their clinical care (KJP). They were invited to contact us for further information using a prepaid envelope. A more detailed discussion with a member of the research team then occurred and the parents were given written information. If they agreed to participate, consent was obtained. All parents were aware that there was a potential risk of their infant's blood oxygen saturation falling during exposure to 15% oxygen. They knew that their baby would be closely monitored by an experienced paediatrician and that if blood oxygen saturation fell below a threshold value the exposure would be ended. Before consenting to participate and when appropriate, families were informed that a proportion of infants studied earlier in the project had had episodes of desaturation when exposed to 15% oxygen. The families of those infants who had episodes of desaturation during our study were advised against taking the infant on an aeroplane or to high altitude until the infant was older; this is a potential benefit of being included in the study. All families knew that we had concerns about the safety of infants during airline travel; they knew that these concerns included a small risk of sudden death. Parents knew that they could withdraw their child from the study at any time without needing to justify their decision.

The degree of airway hypoxia that is safe for infants to be exposed to is unknown. We considered known baseline oxygen saturation levels at altitude6 and normal ranges for episodes of desaturation in healthy infants7 to guide us somewhat empirically in choosing a threshold value of oxygen saturation of ⩽80% for ⩾1 minute. Airway hypoxia was discontinued as soon as possible in each infant who showed this degree of desaturation; it should be remembered that this required the tent to be opened and the gas mixture to be removed from around the baby. No infant remained at ⩽80% in 15% oxygen for longer than 126 seconds.

Of the four infants in whom exposure to hypoxia was discontinued early, one infant had a sibling who had died of the sudden infant death syndrome and was already being monitored at home. Oxygen saturation levels in all four infants remained within the normal range during subsequent monitoring. We believed that monitoring the infants for a longer period in hospital would not have been ethically appropriate because they might be exposed to additional risks (for example, the risk of acquiring an infection in hospital). The two infants who had died following an aircraft flight were not monitored so we are unaware of the duration and degree of hypoxaemia to which they might have been exposed.

Although Savulescu's commentary raises the spectre of human or mechanical error, we took every precaution to ensure that the infants were safe. These included the use of a special medical gas mixture of 15% oxygen and 85% nitrogen instead of air diluted with nitrogen, continuous monitoring of the partial pressure of inspired carbon dioxide to identify rebreathing, and continuous monitoring of the partial pressure of inspired oxygen to ensure adequate ventilation of the tent with the gas mixture. The study was done in a room near the intensive care unit. There was also continuous surveillance by an experienced paediatrician of the readings from the pulse oximeter, transcutaneous monitoring of the partial pressure of carbon dioxide, monitoring of respiratory movement, and electrocardiography.

Although Milner reports in his editorial that British Airways identified no deaths on the undisclosed number of flights involving infants, this is low quality information. It is not accurate, as shown by the personal communication cited in our paper. Infant stimulation and the attention paid to an infant during an airline flight may delay potentially serious consequences of the flight until after the plane's arrival. British Airways would not have access to information on infants after arrival and did not seem to know about either of the two cases of the sudden infant death syndrome that were described in our report.