Acceptable Risks in Pediatric Research: Views of the US Public Free

The survey instrument was drafted on the basis of a comprehensive review of the literature. We then solicited input from content experts and revised the instrument. The draft survey underwent 2 rounds of review by the National Institutes of Health Department of Bioethics Empirical Research Laboratory. The revised survey was piloted in 4 rounds of testing by using Amazon Mechanical Turk, with a total of 555 respondents. Revisions were made after each round.

The final survey addressed 4 domains: (1) respondent characteristics, (2) risks and benefits of phase I pediatric oncology research, (3) risks and societal benefits of net-risk pediatric research, and (4) attitudes toward research in general (adapted from Rubright et al, 2011).⁹  This article reports on the results from the first, third, and fourth domains. The results from the second domain are reported elsewhere.¹⁰ 

To assess which risks are considered acceptable, the survey described 4 scenarios involving procedures that pose increasing levels of risk: blood draw, bone marrow biopsy, kidney biopsy, and a single dose of an experimental drug (for the survey design, see Supplemental Fig 2). To minimize the duration of the survey, respondents were randomly assigned to 1 of the 4 scenarios. To assess whether the social value of the research influenced respondents’ views regarding the acceptability of the risks, each of the 4 scenarios described the respective procedure being used in 3 studies with increasing levels of potential social value: (1) development of treatments with reduced side effects, (2) development of treatments that extend life by several months to several years, and (3) development of curative treatments.

In each of the 4 scenarios, participants were asked to imagine they have a child aged 10 years with advanced cancer who has a high chance of dying in the next few months. The child’s doctors are giving her “the best treatment available.” In addition, she is eligible for a study that will not benefit her but is designed to collect data that might help to improve care for “children with the type of cancer your child has” (see Supplemental Information for complete wording). Each scenario included a test question to assess whether respondents understood that children would not benefit medically from participating in the described studies.

Participants were then asked whether they thought it was appropriate to enroll children in the 3 studies and whether they would enroll their own child. We also asked participants whether they thought that, in general, it can be acceptable to expose children to research risks for the benefit of other children. Those who answered yes were asked explicitly whether studies with higher social value can justify exposing children to greater research risks.

The risks of bone marrow and kidney biopsy were based on the rates of adverse events described in the literature.^11–15  Bain¹²  reported an adverse event rate of 0.08% for bone marrow biopsy. On the basis of these data, we described the risks as “1 in 1,000 children who undergo a bone marrow biopsy experience bleeding, and some of these children require a blood transfusion.” For kidney biopsy, Corapi¹³  reported a 3.5% rate of hematuria, a 0.9% rate of transfusion, and a 0.6% rate of angiographic intervention. Varnell¹⁵  reported an 11% to 18% rate of hematoma, 0.9% rate of transfusion, and a 0.7% rate of postbiopsy interventions to address complications. We thus described the risks as “1 in 10 patients who undergo kidney biopsies experience minor bleeding, ∼1 in 100 will require a blood transfusion, and 1 in 100 will require surgery to correct problems caused by it.” The risks of an experimental drug were described as a 1 in 100 chance of death and a high chance of significant nausea, pain, or hospitalization.^16–26  This description permitted us to compare respondents’ views of these risks in net-risk research with their views of the same risks in a potentially beneficial trial described in the second domain.¹⁰ 

We recruited participants from Ipsos KnowledgePanel, the largest online research panel representative of the US population. KnowledgePanel is used frequently by academic, government, and commercial researchers and consists of 55 000 members. Ipsos randomly recruits panel members through address-based sampling methods. Members are provided with Internet and hardware if needed, thus allowing recruitment of harder-to-reach individuals. Study samples are constructed to be representative by weighting responses to the geodemographic benchmarks from the most recent US Census Bureau Current Population Survey. After the data are collected, Ipsos addresses any differential nonresponses that affect representativeness by adjusting the weights. The present analysis is based on these adjusted weights.

The survey was conducted from June 23, 2020, to July 14, 2020. Selected members were notified by e-mail of the survey, and 3 reminders were sent to nonrespondents. To be eligible, participants had to be aged ≥18 years and have English-language proficiency.

The study was deemed exempt from US research regulations by the National Institutes of Health intramural Institutional Review Board (IRB). No personally identifiable information was collected. Potential participants were informed that participation was voluntary. Those who agreed to participate were informed they could skip any question and they could stop at any time. Each Mechanical Turk participant during our pilot testing was compensated $2. KnowledgePanel participants during the main phase were compensated on the basis of their membership agreement.

The sample size was based on the primary outcome question in the second domain.¹⁰  This calculation yielded an accrual target of 1600 participants. For the present findings, we surveyed the same respondents. The study was administered by using the survey platform Qualtrics, and data were analyzed by using SAS (version 9.4) (SAS Institute, Inc, Cary, NC). Descriptive statistics were generated for the sample’s sociodemographic characteristics. Weighted summary statistics, such as proportions, were calculated by using SAS procedure SURVEYFREQ for respondents’ views on the acceptability of the described risk/social value combinations.

Associations between respondents’ sociodemographic variables and their views were tested by using χ² tests. A P value of <.05 was considered statistically significant. Respondents’ sex and race and ethnicity are based on self-report. Given that clinical research has an unfortunate history of exploiting individuals from minority groups, we assessed whether individuals from minority groups have different views regarding net-risk pediatric research.

To assess whether respondents thought greater social value could justify increased risks, we determined how many respondents indicated that the specified procedure was not appropriate in a study with the lowest level of potential social value but was appropriate in a study with either or both of the higher levels. For example, in the kidney biopsy scenario, we assessed how many respondents indicated that it is not appropriate to expose children to a kidney biopsy in a study with the potential to identify treatments with fewer side effects, but it is appropriate in a study with the potential to identify treatments that can extend life and/or in a study to find a cure. Associations between risk level and the proportion of respondents with these response patterns were tested by using χ² tests.

Of 2508 individuals who were sent the link to the survey, 1658 participated (response rate = 66.1%). Overall, 844 (50.9%) participants identified as female and 814 (49.1%) as male; 1334 (80.5%) identified as White, 165 (10%) as Black, 74 (4.5%) as Asian American, and 85 (5.1%) as other (Table 1).

The proportion of respondents who indicated that the respective research procedure was “probably” or “definitely appropriate” in children increased as the study’s potential social value increased (Table 2). For all 12 risk/social value combinations, especially those involving higher risks, the proportion of respondents who indicated that they would probably or definitely enroll their child (Table 3) was lower than the proportion indicating that the study was probably or definitely appropriate.

The proportion of respondents finding the study appropriate in children decreased as the risks increased across the 4 scenarios (Fig 1). Among respondents who answered the test question correctly (indicating understanding that the study would not benefit their child), 87.4% considered it probably or definitely appropriate for children to undergo the least risky procedure (ie, blood draw) in a study with the potential to identify treatments with fewer side effects, whereas 33.5% considered it appropriate for children to undergo the riskiest procedure (ie, single dose of an experimental drug) in a study with the same social value.

As the risks increased across the scenarios, increases in the potential social value of the study had a larger effect (P < .001) on the proportion of respondents finding the research procedure appropriate. For example, just 8.4% of respondents indicated that it is not appropriate to expose children to a blood draw in a study with the lowest level of social value but indicated that it can be appropriate in a study with the higher (ie, extended life) or highest level of potential social value (ie, cure). The proportion of respondents who changed their views in the same way was 10.9% for a bone marrow biopsy, 16.5% for a kidney biopsy, and 30.0% for an experimental drug. These findings suggest that many respondents regard net-risk pediatric research that poses low risks as generally acceptable but regard net-risk pediatric research that poses somewhat higher risks as acceptable only when it has significant social value.

Respondents were also asked whether, in general, they think it can be appropriate to expose children to research risks to gather information that might benefit others. Overall, 84.5% of respondents answered yes, whereas 15.5% thought this is never appropriate. Although there were several statistically significant associations between respondents’ views and their demographic characteristics (Table 4), support for net-risk pediatric research was high in all demographic groups, and the statistically significant differences do not seem ethically or clinically significant. For example, 79.5% of Hispanic respondents endorsed net-risk pediatric research compared with 85.5% of non-Hispanic respondents. Although this difference was statistically significant, both groups expressed strong support for this research. Given this strong support across all demographic groups, we did not conduct multivariable analyses to assess whether specific variables were more likely to be associated with support than others.

Of the respondents who answered yes, 39.1% indicated that risks to participants should be low no matter how high the potential social value of the study, whereas 60.9% indicated that it can be acceptable to expose children to higher risks in studies with the potential to offer greater benefits to children in the future.

Some commentators argue that clinical research is ethical only when the participants provide informed consent. The Nuremberg Code, for example, states that informed consent is “essential” to ethical research. Although this approach protects children from excessive research risks, it also precludes studies needed to improve pediatric clinical care. More recent guidelines thus permit research with children provided it has an appropriate risk/benefit profile and a legally authorized representative consents on behalf of the child.

Most commentators agree that research which offers children the potential for medical benefit is ethically appropriate. More controversial is net-risk pediatric research, in which children are exposed to risks to collect data that might benefit others. Some commentators argue that net-risk pediatric research is always unethical. Others argue it can be acceptable but only when the risks are minimal. Still others permit a minor increase over minimal risk. This study, which assessed the views of a representative sample of the US public, contributes a number of findings toward resolving this important debate.

First, 84.5% of respondents indicated that, in general, it can be appropriate to expose children to research risks to gather information that might benefit others. This strong support for net-risk pediatric research was consistent across all demographic groups. It suggests, contrary to the views of critics, that net-risk pediatric research can be ethically acceptable.

Second, ∼91% of respondents indicated that it can be acceptable to have children undergo a research blood draw to collect data that might benefit future patients. This finding provides strong support for net-risk pediatric research that poses minimal risk. This finding is consistent with previous surveys conducted in less representative populations.^27–30  For example, Wendler and Jenkins²⁷  found that 79% of a convenience sample of parents were willing to have their child participate in net-risk research that poses a chance of headache.

Third, permitting net-risk pediatric research that poses a minor increase over minimal risk is controversial. Moreover, what constitutes a minor increase is not defined, and there is significant variability across IRBs regarding which risks are considered approvable in this category.³¹  We found strong support for net-risk pediatric research that poses somewhat higher-than-minimal risks. For example, ∼69% of respondents indicated that it can be acceptable to have children undergo a bone marrow biopsy to collect data that might benefit future patients. This finding provides strong support for net-risk pediatric research that poses a minor increase over minimal risk, provided it has important social value. It also suggests that one way to clarify this level of risk, and potentially reduce variability across IRBs, would be to cite the risks of specific procedures, such as bone marrow biopsy, as benchmarks for which risks are approvable in this category.

The present level of support for more-than-minimal-risk pediatric research is somewhat higher than the support found in less representative populations. Wendler and Jenkins²⁷  found that 26% of respondents were willing to have their child participate in net-risk research involving a broken leg, and Freibott et al²⁹  found that 31% of parents with a child in intensive care approved of a research brain MRI with sedation.

Fourth, we found surprisingly strong support for net-risk pediatric research that poses even greater risks, including a 1 in 100 chance of death. For example, ∼50% of respondents stated that this level of risk could be acceptable in pediatric studies that have the potential to identify treatments that extend life for future patients. In addition, 1 in 4 participants indicated a willingness to enroll their own child in such studies. These responses concern children with a life expectancy of only a few months. Hence, they do not support exposing children with an average life expectancy to the same risks. Instead, these findings likely reflect a willingness to evaluate risks in light of the specific participants’ circumstances: a risk of death in a child with a few months to live is arguably different from a risk of death in a child with average life expectancy.

At the same time, these findings suggest many members of the US public do not regard a minor increase over minimal risk as the absolute upper limit on acceptable net risks in pediatric research. In this regard, US regulations permit pediatric research that is not approvable in the categories of minimal risk, minor increase over minimal risk, or prospect of direct benefit to be submitted to a federal panel for review.^32–34  The regulations do not impose an explicit risk limit on approval in this fourth category. Instead, they mandate that these studies must undergo extensive additional review and be consistent with “sound ethical principles.”³⁴  Our finding that public support for higher risks increases as the social value of the research increases suggests that pediatric studies that pose risks somewhat greater than a minor increase over minimal might be approvable in this category but only when they have the potential to collect data that have high social value and that cannot be obtained in a less risky way.

The present findings are subject to 4 important limitations. First, respondents were asked to evaluate net-risk research with children who have a few months to live. Support for net-risk pediatric research may be lower in children who have longer to live. Second, our findings are based on hypothetical scenarios. Respondents’ views may differ when faced with actual studies. Third, we assessed respondents’ views regarding net-risk research with the potential to benefit children who have the same type of cancer as the participants. Respondents may have different views regarding research designed to benefit children with other conditions. Fourth, respondents had to have English-language proficiency. Future research will be needed to assess whether the findings reflect the views of other members of the US public.

Exposing children to research risks to collect data that might benefit others is critical to improving pediatric medical care but remains controversial. Some argue that net-risk pediatric research is always unethical; others consider it acceptable but only when the risks are minimal. In the current study, we found that a representative sample of the US public disagrees. They overwhelmingly support low-risk pediatric research and also support somewhat greater risks, including risks that exceed a minor increase over minimal, provided the research has high social value. Stakeholders assessing pediatric clinical trials, and those applying the US regulations, should be aware of these findings, including the importance of evaluating the acceptability of net risks in pediatric research in light of the study’s social value.

IRB

institutional review board