How Should Researchers Think About Very Unlikely but Very Serious Harms?

What Should You Do With a Harm That Is Extremely Unlikely but Potentially Devastating?

Researchers often become uncomfortable when a risk has two seemingly contradictory characteristics: it is very unlikely, but if it occurs, the consequences could be severe.

A one-in-a-million possibility of a minor inconvenience is usually not the same ethical problem as a one-in-a-million possibility of death, permanent disability, severe psychological harm, or serious legal consequences.

The difficulty is that researchers cannot simply multiply probability by severity and call the result an ethical judgment. Rare events are also difficult to estimate, especially when evidence is limited. The right response is not to ignore a very unlikely serious harm, nor to treat every remote possibility as decisive, but to assess its magnitude, probability, uncertainty, preventability, and consequences systematically.

Low Probability Does Not Automatically Make a Serious Harm Ethically Negligible

Why Rare Serious Harms Are Difficult to Assess

Probability and Magnitude Are Separate Dimensions

The Belmont Report warns that expressions such as “small risk” or “high risk” can be ambiguous because they may refer to both the chance of experiencing harm and the severity of the harm itself. It therefore calls for explicit assessment of both probability and magnitude.

These dimensions interact, but they should not be collapsed into a vague label such as “low risk.” Saying that an event is unlikely tells you something about frequency. It does not tell you how much harm would occur if your estimate proves wrong.

Rare Does Not Mean Impossible

In research, very low probabilities are often estimates rather than directly observed frequencies. A study may involve too few participants to detect rare adverse events. Relevant evidence may come from earlier studies, clinical experience, surveillance systems, pharmacovigilance, laboratory evidence, or biological plausibility.

Belmont specifically recommends determining whether investigators' estimates of probability are reasonable based on known facts and other available studies.

When evidence is weak, researchers should not manufacture confidence by assigning an arbitrarily tiny number. Sometimes the appropriate conclusion is that the probability is uncertain.

Small Samples Can Hide Rare Events

Suppose an adverse event has a true probability of 1 in 10,000. A study enrolling 200 participants might observe zero events. That observation does not prove that the event cannot occur.

The example is statistical rather than a specific regulatory threshold, but the ethical point is important: absence of an observed event is not the same as evidence of zero risk.

Researchers should therefore interpret rare-event evidence in light of study size, duration, population, exposure, prior evidence, and other sources of information.

Severity Can Make a Remote Possibility Worth Serious Attention

A remote possibility of serious harm may warrant particular safeguards precisely because the consequence is difficult to reverse.

Imagine two hypothetical risks, each estimated at 0.01%. One causes a mild temporary rash. The other causes permanent neurological injury. The probability is identical in this illustration. The ethical assessment is not.

This is why risk-benefit assessment must examine the characteristics of the harm, not merely its estimated frequency.

Uncertainty Is Itself Relevant to Decision-Making

A probability estimate can be uncertain for different reasons. Perhaps a related intervention has been studied extensively but not in this population. Perhaps there are signals from laboratory evidence but little human data. Perhaps the sample is too small to establish a reliable event rate.

Researchers should distinguish an event that is confidently estimated to be rare from an event whose probability is simply poorly known.

The practical consequence is not that uncertain risks must automatically be treated as certain. Rather, uncertainty should be made visible when deciding whether safeguards are adequate and whether the research remains justified.

Look for Evidence That the Event Is Related to the Research

Not every bad event occurring during research is caused by the research.

OHRP's guidance on unanticipated problems uses a specific framework for determining when an incident represents a previously unrecognized increase in research risk. It considers whether the event was unexpected, whether it was related or possibly related to research participation, and whether it suggests that the research places participants or others at greater risk than was previously known or recognized.

This distinction is useful even before an event occurs. Researchers should examine the evidence connecting the risk to the research procedure rather than treating every adverse outcome in a participant as a research effect.

Ask Whether the Serious Harm Can Be Prevented or Detected Early

Probability is not the only design variable. Researchers may be able to reduce the consequences of a rare serious event through screening, monitoring, emergency procedures, stopping rules, dose adjustment, staff training, or other safeguards.

For example, an adverse reaction may be rare but have a much better outcome if recognized promptly. A study may therefore need a monitoring system capable of identifying the event early rather than merely estimating its probability.

The 2024 Declaration of Helsinki requires researchers to implement measures to minimize risks and burdens and to continuously monitor, assess, and document them. It also requires researchers to be confident that risks and burdens can be satisfactorily managed.

Reversibility Matters

A rare reversible harm and a rare irreversible harm present different ethical concerns even when their estimated probabilities are similar.

Researchers should consider how quickly harm occurs, whether treatment exists, whether damage can be reversed, whether recovery is expected, and what the consequences would be if intervention failed.

These characteristics help explain why focusing only on percentage probability can give a misleading impression of overall risk.

A Very Rare Severe Harm May Still Be Compatible With Ethical Research

The presence of a serious rare risk does not automatically make a study unethical. Medical and other research can involve outcomes that are possible but difficult to reduce below a certain level.

The key questions are whether the risk is genuine, whether the research is necessary and scientifically valuable, whether exposure has been minimized, whether safeguards are proportionate, and whether the residual risk is permissible for the participant population.

The Belmont Report explicitly states that risk can perhaps never be entirely eliminated, while emphasizing careful consideration of alternative procedures and especially strong justification when significant risk of serious impairment exists.

Do Not Use “Less Than One Percent” as a Universal Safety Rule

A numerical threshold can sound reassuring because it appears precise. But there is no general ethical rule saying that a risk becomes acceptable whenever its probability falls below one percent, one in a thousand, or any other universal figure.

The acceptable level depends on the magnitude of harm, participant population, potential direct benefit, scientific and social value, alternatives, safeguards, and applicable ethical or regulatory limits.

Precision is useful only when it represents actual knowledge.

Rare Severe Harms Matter Especially When There Is No Direct Participant Benefit

Suppose a research-only procedure offers no prospect of benefit to the participant and carries a very small possibility of serious injury. The scientific value of the knowledge becomes particularly important, but so does the level of risk the applicable framework permits when participants receive no direct benefit.

This is one reason the question of research without personal benefit requires more than a simple statement that the event is rare.

Do Not Hide Serious Rare Events in Technical Language

Researchers sometimes describe rare severe outcomes in a way that technically mentions them but makes their significance difficult to understand. Consent information and risk assessments should distinguish likelihood from seriousness in language participants and reviewers can understand.

The 2024 Declaration of Helsinki requires prospective participants to be informed in plain language about anticipated benefits and potential risks and burdens and to be given sufficient information to make a voluntary decision.

Separate Pre-Study Assessment From an Unexpected Event During the Study

Before recruitment, researchers are estimating foreseeable risks from existing evidence. During the study, new information may reveal that an event is more frequent, more severe, or more plausibly related to the research than previously recognized.

Those are related but distinct tasks. OHRP's guidance explains that an unexpected event that is related or possibly related to research and indicates greater risk than previously recognized may constitute an unanticipated problem requiring consideration of protocol changes or other corrective action.

The later guides in this collection address what to do when new risks or information actually emerge. Here, the focus is the initial reasoning: do not dismiss a serious possibility simply because its estimated probability is small.

How to Treat a Rare but Serious Adverse Event in a Study Design

The point is not to convert the event into a single score. The serious consequence remains visible even though the probability is low and uncertain.

Common Mistakes When Assessing Rare Serious Harms

Do Not Let a Low Percentage Hide a High-Consequence Event

When a possible harm is both rare and serious, write the assessment so that a reviewer can see why you believe the probability is low, how uncertain that estimate is, and what you have done to control the consequences.

A useful risk statement often becomes clearer when written as two sentences rather than one label: the event is expected to be rare, and the consequence would be serious if it occurred. That is much more informative than simply calling the study “low risk.”

Before Calling a Serious Harm “Very Unlikely,” Check These

Frequently Asked Questions About Rare but Serious Research Harms

Do Not Let “Very Unlikely” Become a Synonym for “Unimportant”

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.