The moral limits of medical research with children
Westra, A.E.
Citation
Westra, A. E. (2011, June 30). The moral limits of medical research with children. Retrieved from https://hdl.handle.net/1887/17752
Version: Corrected Publisher’s Version
License: Licence agreement concerning inclusion of doctoral thesis in the Institutional Repository of the University of Leiden
Downloaded from: https://hdl.handle.net/1887/17752
Note: To cite this publication please use the final published version (if applicable).
ChApter 4
how best to define
the concept of minimal risk
a. e. westra1, J. m. wit1, r. n. sukhai1, i. d. de beaufort2
1. Department of Pediatrics, Leiden University Medical Center;
2. Department of Medical Ethics and Philosophy of Medicine, Erasmus Medical Center Rotterdam
Submitted
INTRODUCTION
Some studies necessary for improving pediatric care cannot directly benefit the children involved. For example, pharmacokinetic studies and studies into pathophysiologic mechanisms often do not directly benefit their subjects. From an ethical point of view, involving children in such studies is complex: considering the assumption that children cannot give informed consent, it is not easy to justify research risks that are faced solely for research purposes. Yet a total ban on such studies would seriously harm future sick children. As a compromise, the drafters of the US Federal Regulations and those of many other rules and regulations have chosen to allow review boards to approve research that does not directly benefit the children involved only when the risks are minimal or, under certain conditions, constitute a minor increase over minimal.76
The concept of ‘minimal risk’ was not drawn from common language. Thus, how should pediatric researchers, when designing studies, and review boards, when reviewing them, understand and use this concept? How should one judge whether for instance a lumbar puncture, an anorectal manometry, or a bronchial provocation test can be considered minimal risk? An adequate definition of the concept seems vital, not only to support those involved in designing and reviewing pediatric studies, but also to appropriately protect children from research risks, and to let research involving acceptable risk levels be approved.
The US approach was to define minimal risk in relation to other situations with risks that one may regard as minimal. The definition reads as follows: ‘the probability and magnitude of harm or discomfort anticipated in the research are not greater in and of themselves than those ordinarily encountered in daily life or during the performance of routine physical or psychological examinations or tests.’77 Thus, to determine whether certain research risks are minimal, researchers and review boards are supposed to compare them with the risks of daily life and of routine examinations.
However, it soon appeared that not all persons concerned apply the provided definition in the same way. Janofsky and Starfield surveyed pediatric department chairmen and clinical research center directors and discovered a great variation in their assessment of risk.78 For example, when asked to assess the risks presented by a skin biopsy in a newborn (using the definition provided by the US Federal Regulations), 55% of the respondents classified the procedure as minimal risk, whereas 45% thought that it involved a minor increase over minimal risk, or even more. More recently, Shah and colleagues revealed that the application of ‘minimal risk’ by review board chairs is also variable.79 For example, 48% of the review board chairs classified MRI without sedation as involving minimal risk, whereas 44% thought that it involved a minor increase over minimal risk or more (8% answered ‘I don’t know’).
The respondents might have agreed more often if they had been reviewing real protocols, and the results do not provide information about consistency
within review boards. Nonetheless, the fact that the concept of minimal risk as currently defined in the US Federal Regulations is not consistently applied raises serious ethical concerns, considering the important threshold function of this concept. An interesting debate in the pediatric and bioethical literature revealed that the inconsistency is not surprising, because the definition poses some serious normative and practical problems. Nevertheless, so far, the US Federal Regulations have remained unchanged. Moreover, the EU working group that has provided guidance for the application of the European Clinical Trials Directive with regard to research in children has recently adopted the same definition.34
In this paper, we first provide an overview of the problems posed by the US definition of minimal risk, and of the solutions proposed by other authors.
This will reveal that even after implementing the most promising solutions, two issues would remain. In the second part of the paper, we will argue how we think these two remaining issues could best be solved. We will propose to distinguish between the risks of discomfort and the risks of harm, and to define for each of these two concepts what may be regarded as ‘minimal’.
THE PROBLEMS POSED By THE US DEFINITION
The US definition of minimal risk poses three main problems. The first and second could, as we will show, satisfactorily be solved. For the third and most fundamental problem, however, no fully satisfactory solution has yet been proposed.
Absolute or relative interpretation?
The first problem associated with the US definition is that this definition can be interpreted in more than one way. It is clear that the concept of ‘minimal risk’ is meant to function as a threshold; ‘as low as possible’ (after risk minimization) is not enough. However, it is unclear whether this threshold depends on the daily risks of an average child (the so-called ‘absolute interpretation’) or on the daily risks of the specific type of children to be enrolled in the study (the so-called ‘relative interpretation’).80 In other words: is the threshold meant to be fixed or variable?
Although this is a rather fundamental question, which any definition of minimal risk should address, the drafters of the US Regulations have chosen not to take sides. To solve this problem, several federally-charged committees have argued for the absolute interpretation.36;81;82 This choice seems right: review boards should be prevented from approving high risk studies as minimal risk just because the children who will be enrolled in the study experience high risk in their daily lives.
The definition confuses the descriptive and the normative
The second problem associated with the US definition is that the choice for
‘comparator activities’ such as daily life and routine examinations also involves
a normative choice.83 Use of this definition implies that what a child encounters during daily life and during routine examinations is also considered acceptable in the context of research without direct benefit (provided that all other requirements for ethical research, such as scientific validity, are fulfilled).8 Why should this be the case? Daily risks that have been identified and quantified appeared to be fairly high: playing football, for instance, involves a 1/250 risk of injury, and when adolescents drive on rural highways under wet conditions, they face a 4/million risk of mortality.84 In addition, children commonly have experiences at school that threaten their self-image, such as getting bullied by peers.85 We accept most of these daily risks only because they are compensated by potential benefits and/
or because eliminating them from our lives is impossible.86 The risks of routine physical of psychological examinations on the other hand, are very low.86;87 Just as one may wonder why risks as high as daily risks should be considered acceptable in the context of pediatric research without direct benefit, one may also ask why risks as low as the risks of routine examinations should be part of the definition.
To solve this problem, several authors have tried to identify comparator activities that are normatively ‘correct’. That is, they have tried to identify activities that are sufficiently similar to participating in research without direct benefit and whose risks are deemed acceptable. For example, Nelson and Ross, following Freedman and Ackerman, have proposed that review boards should compare research risks with the risks of those situations in daily life in which scrupulous parents would intentionally expose their child to risks.14;88-91 Examples of such situations are drying the dishes, and a day out with the family. Wendler has proposed to compare research risks with the risks faced during charitable activities: just as research without direct benefit, charitable activities involve facing risks for the benefit of others.86;87 Together with Rid and other colleagues, Wendler has also suggested the ‘reasonable’ interpretation of the daily risk standard. This interpretation concerns the risks of those activities of daily life that are appropriate for ordinary individuals, even in contexts that do not offer the potential for personal benefit.92 This interpretation seems both normatively correct and feasible (as it is general enough to provide a considerable number of examples), which makes it the most promising solution to the second problem of the US definition.
How to compare research risks with those of comparator activities?
However, after solving both the first problem and the second problem associated with the US definition, a third problem would remain: properly comparing research risks with the risks of comparator activities is very difficult. How should one compare the likelihoods and the magnitudes of all potential harms resulting from the research with the likelihoods and the magnitudes of all potential harms resulting from a (reasonable) daily life? Although this problem is a practical rather than a normative problem, it is no less important to solve, considering the fact
that this could significantly improve the quality of ethical review of pediatric research studies.
Recently, Rid and colleagues have attempted to solve this third problem by offering a systematic method for making the required comparisons.92 The four steps of this method are: 1) identifying all potential harms posed by the research;
2) categorizing the magnitude of each potential harm into one of seven harm levels on a ‘harm magnitude scale’; 3) quantifying or estimating the likelihood of each potential harm, and 4) comparing the likelihood of each potential harm with the likelihood of harms with the same magnitude occurring as a result of the comparator activity. The harm magnitude scale was developed specifically for this purpose in cooperation with several clinicians, bioethicists and other experts, using discomfort, duration, required treatment, and effect (on ability to perform activities, and on ability to pursue life goals) as important parameters. Its seven categories include negligible harm (e.g., mild nausea, or transient pain), small harm (e.g., headache for several hours, or common cold), moderate harm (e.g., bone fracture, or major hemorrhage), significant harm (e.g., ligament tear of knee with permanent instability, or intensive care for several weeks), major harm (e.g., psychotic episode, or loss of finger), severe harm (e.g., major depression, or paraplegia), and catastrophic harm (e.g., death).92
By emphasizing the importance of assessing risks systematically and by developing the harm scale, Rid and colleagues have made a very valuable contribution to the debate on minimal risk. However, we believe that their comparison method is unable to completely solve the problem that properly comparing research risks with the risks of comparator activities is very difficult. If this very systematic and detailed method were implemented, two issues would remain.
The first issue is that comparing research risks with the risks of comparator activities is difficult when dealing with risks of discomfort. In addition to the risks of harm, and in addition to the discomforts these harms may cause, almost all research procedures one may think of involve a fairly high likelihood of momentary negative physical (e.g., pain, or itchiness) and/or psychological (e.g., fear, boredom, or embarrassment) experiences. We believe that such risks of discomfort must be assessed carefully, because, although unrelated to ‘real’ harm, they concern negative experiences for the children involved. The comparison method described by Rid et al. reveals that when using a comparator-based definition, this is difficult. Rid and colleagues do not use the term discomfort but their category of ‘negligible harm’, considering the provided examples, seems to refer to what usually are called discomforts. The comparator harm in this category (stemming from their reasonable interpretation of the daily life standard), then suggests an acceptable likelihood of 100%. What does this mean? Should a 100% chance of transient pain always be regarded as minimal risk, even if the pain is very severe, and/or repetitive? Also the category of ‘small harm’ includes some discomforts as examples, such as anxiety. The comparator harm in this category then suggests an acceptable likelihood of approximately one out of four. Does this mean that
studies should be classified as involving more than minimal risk if more than a quarter of the subjects may experience some anxiety, even if very little?
The second issue concerns the level of preciseness at the other end of the spectrum: the risks of more serious harm. When assessing such risks, which usually involve very low likelihoods, data on the risks of comparator activities may be very helpful: they can provide an indication of the likelihoods of the harms that may occur during the comparator activity. The level of preciseness that can be offered, however, depends on the available number of comparator harms: to construct exact cut-off likelihoods for several harm categories (by magnitude), one needs numerous comparator harms for each category. In that situation, one could, for example, calculate the average likelihood for that type of harm, or one could choose to use the highest likelihood. Unfortunately though, no matter which comparator activity is used, the available number of comparator harms is very low at this end of the spectrum. Rid and colleagues’ use of the available data on the reasonable risks of daily life results in the following combinations of magnitude and likelihood counting as ‘minimal’: moderate harm: 7/10,000, significant harm: 8/100,000, major harm: 8/100,000,000, severe harm: 3/10,000,000, and catastrophic harm: 2/1,000,000. This list seems rather arbitrary. Note, for example, that the acceptable likelihood for severe harm (e.g., paraplegia) is higher than that for major harm (e.g., loss of finger). Perhaps, until more data become available, aiming for such a level of preciseness is not realistic. This is not necessarily problematic, as the exact likelihoods of the harms that may occur during research procedures often are not known either, but it maybe helpful to be aware of this limitation.
In sum, the systematic comparison method proposed by Rid and colleagues cannot completely solve the the third problem associated with the US definition (i.e., that comparing research risks with the risks of comparator activities is very difficult), because of two remaining issues: 1) comparing research risks with the risks of comparator activities is difficult when dealing with risks of discomfort; and 2) regarding the risks of more serious harm, aiming for a high level of preciseness is not realistic.
ANOTHER TyPE OF DEFINITION COULD SOLVE THE REMAINING ISSUES
So far, we have argued that adequately defining the concept of ‘minimal risk’ is vital, and that the definition provide by the US Federal Regulations poses some serious problems, one of which seems difficult to solve completely. In the second part of this paper, we will describe how we think the two remaining issues best can be solved. We will propose to distinguish between the risks of discomfort and the risks of harm, and to define for each of these two concepts what can be regarded as ‘minimal’ by ‘incorporating’ the comparisons at stake.
Figure 1. The risks of discomfort and the risks of harm posed by a research procedure.
This figure summarizes the ways in which a research procedure can negatively affect the research subjects. Although the figure should not be regarded as a flow chart, it may be helpful when reviewing a medical research study. Risks of discomfort (which are also known by the term ‘burden’) differ from risks of harm because they concern subjective and
momentary experiences, because the likelihood usually is relatively high, and because the magnitude may vary considerably among the subjects. Thus, when assessing the risks of discomfort, one should consider what exactly is imposed on the research subjects, how likely it is that this will cause any type of discomfort in the subjects, and in how many of the subjects this discomfort will be considerable. Risks of harm, on the other hand, can best be assessed the other way around. Thus, when assessing the risks of harm, one should consider the magnitudes of all potential harms, and then estimate how likely these harms are.
Medical research study
Procedure 1 Procedure 2
Risks of discomfort
(burden)
Procedure 3
Likelihood Magnitude
Risks harm of
Magnitude Likelihood
This figure summarizes the ways in which a research procedure can negatively affect the research subjects. Although the figure should not be regarded as a flow chart, it may be helpful when reviewing a medical research study. Risks of discomfort (which are also known by the term ‘burden’) differ from risks of harm because they concern subjective and momentary experiences, because the likelihood usually is relatively high, and because the magnitude may vary considerably among the subjects. Thus, when assessing the risks of discomfort, one should consider what exactly is imposed on the research subjects, how likely it is that this will cause any type of discomfort in the subjects, and in how many of the subjects this discomfort will be considerable. Risks of harm, on the other hand, can best be assessed the other way around. Thus, when assessing the risks of harm, one should consider the magnitudes of all potential harms, and then estimate how likely these harms are.
Figure 1. The risks of discomfort and the risks of harm posed by a research procedure.
Minimal risk of discomfort
Above, we have argued that comparing research risks with the risks of comparator activities is difficult when dealing with risks of discomfort. We believe that this difficulty is due to the fact that discomforts are not just ‘low-magnitude harms’, and thus should not be treated as such (See Figure 1). Risks of discomfort differ
from risks of harm because they concern subjective and momentary experiences, and because the likelihood usually is relatively high. Several codes and regulations therefore already explicitly distinguish expected discomforts from risks of harm and speak of ‘minimal risk and burden’ (the term ‘burden’ referring to the expected discomforts).25;32;33 We do not suggest using the term ‘burden’, because a risk of discomfort involves a likelihood component, which (at least technically) makes it a risk. However, we agree that the risks of discomfort and the risks of harm can best be defined and assessed separately.
Importantly, the magnitude of the experienced discomfort may vary considerably among the subjects. Empirical studies assessing discomfort levels related to various procedures have shown that whereas a certain part of the subjects experiences little discomfort during that procedure, another part experiences considerable discomfort.93;94 This variety is not linked to the type of discomfort: it is not the case that, for example, pain is always less intense than anxiety. Thus, rather than regarding some types of discomfort as minimal risk even if occurring in all cases, and other types of discomfort only in case the likelihood is less than one out of four, we propose to define minimal risk of discomfort as either
‘little discomfort’ (even if occurring in all cases), or ‘considerable discomfort’ if the likelihood is less than one out of four.
Accordingly, we propose that ‘minimal risk of discomfort’ applies if: ‘Empirical data, expert opinions and/or the procedural characteristics (e.g., invasiveness;
disturbance of normal routines) suggest that at most a quarter of the persons concerned will experience considerable discomfort’ (See Table 1). The term
‘considerable’ encompasses both the magnitude and the duration of the discomfort, and must be considered in light of the fact that the subjects have to undergo the procedure solely for research purposes. It is important to realize in this respect that ‘minimal risk’ is not the upper level of acceptable risk, but some sort of baseline level: most codes and regulations allow for exceptions to this requirement.
Table 1. Risks of discomfort and risks of harm that may be regarded as ‘minimal’.
Minimal risk of (physical and/
or psychological) discomfort Empirical data, expert opinions and/or the procedural characteristics (e.g., invasiveness; disturbance of normal routines) suggest that at most a quarter of the persons concerned will experience considerable discomfort.
Minimal risk of (physical and/
or psychological) harm Empirical data and/or expert opinions suggest that in the persons concerned, the likelihoods that the procedure(s) will cause small, moderate, and/or seriousharm are
< 1/100, < 1/10,000, and < 1/1,000,000, respectively.i
I Examples of small harm: minor psychological harm, or reversible adverse effects of medication, such as nausea and vomiting for several hours or days. Examples of moderate harm: bone fracture, or major hemorrhage. Examples of serious harm: ligament tear of knee with permanent instability, intensive care for several weeks, loss of finger, serious long-term psychological harm, paraplegia, or death.
Minimal risk of harm
Next, let us consider the definition of ‘minimal risk of harm’. At this side of the spectrum, the main issue was that regarding the risks of more serious harm, aiming for a high level of preciseness is not realistic. To solve this issue, we propose to use a modified version of the harm magnitude scale developed by Rid and colleagues, to deduct global cut-off likelihoods from their data on the reasonable risks of daily life, and to implement these likelihoods in the definition.
Of the seven harm categories described by Rid and colleagues, the category of negligible harm only involves discomfort and thus no longer plays a role in this context. Regarding the category of small harm, we suggest to also include example harms such as minor psychological harm and the possible adverse effects of medication (e.g., nausea and vomiting for several hours or days), and to use a global cut-off likelihood of < 1/100. We suggest using the category of moderate harm and its global cut-off likelihood (i.e., < 1/10,000) without modifications. Last, we propose to combine the four most serious harm categories (i.e., significant, major, severe, and catastrophic) into one category named ‘serious harm’. Taken together, the available data (8/100,000, 8/100,000,000, 3/10,000,000, and 2/1,000,000, respectively) suggest a global cut-off likelihood for this type of harm of < 1/1,000,000. Thus, we propose that ‘minimal risk of harm’ applies if:
‘Empirical data and/or expert opinions suggest that in the persons concerned, the likelihoods that the procedure(s) will cause small, moderate, and/or serious harm are < 1/100, < 1/10,000, and < 1/1,000,000, respectively’.
Points of clarification
Two features of our proposed definitions of minimal risk of discomfort and of minimal risk of harm may require further clarification. First, we speak of
‘procedure(s)’ in stead of ‘study’ because some studies involve a combination of procedures without direct benefit and procedures with direct benefit; the latter do not have to meet the requirement of minimal risk.13;63 Second, we use the phrase ‘in the persons concerned’. This is not because we believe that the acceptability of the risks at stake depends on which individuals will be enrolled in the study (which would be a so-called ‘relative interpretation’). Instead, we consider individual differences to be relevant when identifying and quantifying these actual ‘risks at stake’, because these are also not always the same for all potential subjects. For example, having to undress for teenagers generally is far more stressful than it is for prepubertal children. On the other hand, young children’s trust of authorities makes them more vulnerable to being deceived in research.85;95 In addition, during various phases of growth and development, children’s vulnerability to physical harm may vary. When aiming to accurately review proposals for pediatric research, one should take this into account.
POTENTIAL LIMITATIONS
Our proposed definitions of minimal risk of discomfort and of minimal risk of harm have several potential limitations. First, our definitions imply identifying and evaluating the likelihoods and magnitudes of all possible discomforts and harms of all procedures one by one, while searching for and using all relevant data. This may be too time consuming for often already heavily overloaded review boards.
The process may not be as time consuming as comparing all research risks with the risks of comparator activities, if one would do so according to the systematic method described by Rid et al., but will nevertheless take more time than a rather intuitive approach. Yet we believe pediatric research protocols deserve this time:
intuition alone may be biased, may lack transparency, and may fail to take into account all relevant data. Besides, the process could be completed pretty fast if review boards were to work together in getting all available data on the risks of discomforts and/or harm posed by various procedures systematically presented on a website.
The second potential limitation is that even when all relevant data are available, our definitions may not in all cases lead to unambiguous outcomes: people may disagree on whether the expected discomforts must be regarded as considerable, or on whether the potential harms must be classified as ‘small’, ‘moderate’ or
‘serious’. However, compared to the US definition, our definition offers some clear advantages in this respect: discussions within review boards, and between review boards and researchers, will be more to the point, because the risks of discomfort and the risks of harm are assessed and discussed separately.
We believe that the present paper provides a useful context for a discussion on which risks we allow children to face if solely for research purposes. We hope to have made clear that there are reasons to modify the US definition, and that the approach presented in this paper may be an appropriate candidate to this purpose.
