Quantitative interpretation

Given that there seems to be no consensus at least among experts about the meaning of minimal risk, some have proposed quantitative strategies for assessing permissible risks. Nicholson argues:

The evidence presented of the inaccuracies in investigators’ perceptions of risk suggests that it is important to work towards an arithmetical categorization of risk and away from such categories as negligible, minimal, minor increase over minimal, and so forth, which are impossible to define and rely heavily on perceptions of risk, particularly those of the professionals involved. (Nicholson, 1986, p.118)

That is, one might reduce variability and ambiguity in risk determination with the help of a quantitative interpretation of the permissible degree of risk.

How would this work? Nicholson proposes that the worthiness of a given trial be assessed by appealing to a complex arithmetical formula that proceeds as follows: one multiplies the length of time for which each harm potentially caused by a research procedure would last by its valuation on the Rosser scale (Rosser & Kind, 1978), which assigns a numerical value to different states of illness. The product is then multiplied by the probability that the harm will occur. This preliminary result represents the disutility score for each harm that an individual might face in the trial. The disutility scores are then multiplied by the number of subjects in a trial and then added together which results in a final disutility score for the whole project (Nicholson, 1986, p.108-110).

Once the final disutility score of a trial is calculated, it is evaluated by reference to a second set of numerical values that prescribe permissible levels of risk for non-

therapeutic procedures. For Nicholson, minimal risks are those that pose less than 1 per million risk of death, less than 10 per million risk of major complication, and less than 1 per thousand risk of minor complication (Nicholson, 1986, p.119). Nicholson claims that these values represent permissible degrees of risk because they correspond to the degree of risk that is generally considered negligible in medicine (Nicholson, 1986).

Nicholson’s system for risk evaluation is not purely arithmetic or purely quantitative. It relies on the judgment that negligible risks in medical practice are a morally appropriate comparator for risks in research as well as the idea that the risks considered negligible in medicine can be represented quantitatively. While Nicholson recognizes that values will inevitably enter into judgments about risk (Nicholson, 1986, p.81), he endorses the use of quantitative calculations whenever possible to reduce the ambiguity and variability generated by intuitive risk assessments. Thus, for Nicholson, the appropriate ethical assessment of clinical research—and non-therapeutic research procedures—depends on a quantitative calculation that is then assessed by comparison to the degree of risk that is considered acceptable in medical practice.

Nicholson’s quantitative strategy is unsuccessful. It relies on at least two

problematic claims. First, it is not clear that the degree of risk considered negligible in medicine can be reflected quantitatively. Nicholson arrives at his calculations by

examining statistics collected about the risks of certain medical procedures and doctors’ as well as the general public’s reactions to these risks. For example, he appeals to the research of the Committee on Safety of Medicines to find statistical information about the risks of taking medications. The Committee on Safety of Medicines examined data about the number of deaths reported as being caused by a particular drug during a given period and the number of prescriptions for the drug in the same period; they found that 22 drugs produced a risk of death per prescription no greater than 1 per million, 6 drugs produced a risk greater than 10 per million, and 1 drug produced a risk of 160 per million

(Nicholson, 1986, p.89).

Nicholson claims that these statistics serve as some evidence that both doctors and the general public deem a risk of death of one per million as negligible in medical practice (Nicholson, 1986, p.89). But this conclusion is unwarranted. It is not clear whether doctors prescribe drugs that cause a risk of 1 per million of death because they deem this degree of risk negligible or whether they are simply unaware of the risks of certain prescriptions. The results of the Committee on Safety of Medicines may well contribute to changes in prescribing practices. Further, this example cites evidence based on an examination of 29 medications, which is a very small subset of the medications prescribed by physicians on a regular basis. Even if evidence about these 29 medications

revealed that physicians deem negligible a 1 per million risk of death per prescription, it is not clear whether, and if so why, these statistics can be generalized to broader

prescribing practices. Consequently, it is hard to see how the Committee on Safety of Medicines’ report translates into statistics about the degree of risk generally considered negligible in medical practice.

Moreover, while a doctor’s decision to prescribe a medication may well take into account the statistical likelihood of a medication resulting in death, her prescribing habits are likely influenced by different or multiple factors. A physician’s judgment also

depends on particularities of the situation, including a given patient’s risk factors, allergies, and circumstantial or environmental factors that make a given medication a better choice for her. The physician’s subsequent judgment about what medication to prescribe is a qualitative judgment based on a consideration of the various benefits and harms of various medical options. It is problematic to appeal to a quantitative calculation that would exclude these considerations in medical practice. Doing so would result in poorer medical care. If the degree of risk considered permissible in clinical research is based on the degree of risk considered negligible in practice, then it should be calculated using more than statistics about risks of death and injury. Making risk assessments based on these factors alone runs the risk of neglecting other important features of a trial. Thus, it is difficult, at best, to accurately reflect the risks considered negligible in medicine quantitatively and even if one could, doing so is undesirable.

Second, Nicholson’s proposal depends on the problematic idea that the degree of risk considered negligible in medical practice is a morally relevant comparator for

permissible risks in research. The goal of clinical practice is to treat patients; accordingly, the interventions prescribed and administered in practice aim to offer a patient the

prospect of direct benefit. It follows that some risks may be deemed negligible in clinical medicine not because there is something inherently acceptable about this degree of risk but because the risk is far outweighed by the potential benefit of a treatment. But this is not necessarily the case in clinical research. Some research procedures are administered exclusively in the interest of answering the scientific question and do not aim to offer a subject the prospect of direct benefit. Given that risks associated with procedures administered in clinical practice may be considered negligible because of the benefits

these procedures offer and that some research procedures do not offer the prospect of benefit, the degree of risk considered negligible in medicine is not an appropriate comparator for the degree of risk that is considered ethically permissible in research.

More generally, quantitative interpretations for minimal risk must either arbitrarily choose a number to represent morally permissible degrees of risk or they must appeal to another kind of justification about what serves as a useful comparator for morally permissible risks in research. The former option is unjustifiable and in the latter, it is the second comparative judgment that serves as the moral foundation for the interpretation. That is, the success of the proposal ultimately depends on the strength of the moral comparator chosen for permissible risks in research, which is a qualitative judgment. Thus, quantitative strategies are ill-equipped to serve as interpretations for minimal risk. Insofar as Nicholson’s proposal fails to offer a convincing reason why the risks deemed negligible in medicine (were it possible to reflect these quantitatively) serve as a useful comparator for morally permissible risks in research and there are good reasons to think that it should not be the appropriate comparator, his proposal fails to meet the

requirement that an interpretation limit risks to a defensible degree. Consequently, it is not a successful interpretation of minimal risk.