Two Limits, Two Different Questions
Bioequivalence criteria use limits on an exposure ratio to support a regulatory decision. Clinical pharmacology asks a different question: what change in exposure, for a particular medicine and patient context, could produce a meaningful change in therapeutic benefit or harm? These questions are related, but a regulatory acceptance limit is not automatically a measured boundary at which clinical harm begins.
The distinction matters because the limits may be chosen using different evidence and for different purposes. A prespecified regulatory interval provides a consistent decision standard for evaluating a bioequivalence study. A clinical harm boundary, if it can be estimated, concerns consequences for a specified outcome, population, regimen, and exposure measure. A ratio alone does not establish those consequences.
A clinical harm boundary is an exposure difference, or range of differences, at which a clinically meaningful change in a specified benefit or harm outcome is expected or considered unacceptable for a defined context. Its location depends on clinical evidence and the outcome being considered; it may be uncertain, asymmetric, or not expressible as one universal ratio.
The conventional 80–125% limits are an example of a regulatory acceptance interval on the ratio scale. They are not a claim that every exposure ratio between 0.80 and 1.25 is clinically harmless for every medicine, patient, and endpoint, nor that harm necessarily occurs immediately outside those limits. The earlier material on how regulatory margins are chosen and prespecified explains their role as decision standards. Here the focus is on why that role should not be mistaken for a direct clinical measurement.
What the Ratio Limits Say—and Do Not Say
Suppose \(R\) denotes the test-to-reference ratio for a specified exposure measure. An acceptance interval such as \([0.80,1.25]\) defines the ratio range used by the applicable regulatory rule. It says where the rule places its limits. It does not, by itself, describe the dose-response curve, the probability of an adverse event, or the range of exposure differences that would be clinically unimportant.
The distinction is especially important because a bioequivalence analysis commonly concerns an average comparison on the log scale. A conclusion about that population-level comparison is not a direct statement about every participant's individual response. Clinical outcomes may depend on baseline risk, adherence, concomitant treatment, pharmacodynamic sensitivity, and other factors that are not encoded in a single exposure ratio.
Why 80–125% Is Symmetric on the Log Scale
The conventional ratio limits are reciprocal: \(1/1.25=0.80\). Taking natural logarithms converts multiplicative differences into additive differences. This reveals the mathematical symmetry of the interval, but that symmetry does not demonstrate a clinically symmetric response.
The calculation is direct: \(1/1.25=0.80\), so \(\ln(0.80)=\ln(1/1.25)=-\ln(1.25)\). The two ratio limits therefore lie the same distance from zero on the log scale. This is a property of the chosen multiplicative interval. It does not show that a 20% reduction and a 25% increase have equal clinical consequences. Clinical response can be nonlinear, and the consequences of moving below and above a reference exposure can differ.
Worked Example: Translating a Regulatory Interval
Suppose a hypothetical product has a prespecified regulatory interval of 0.80 to 1.25 for a particular exposure ratio. Translate the interval to the log scale and state what the translation establishes.
Step 1: Take the logarithm of each limit.
Step 2: Check the reciprocal relationship. Since \(1/1.25=0.80\), the lower log limit is the negative of the upper log limit. Thus the interval is symmetric around zero for the log ratio.
Interpretation. This calculation establishes the mathematical form of the regulatory interval. It does not locate a clinical harm boundary, because no clinical outcome, dose-response relationship, or patient-specific risk information has been supplied. Calling the limits “clinical thresholds” would go beyond the information in the example.
Clinical Boundaries Require Clinical Evidence
A clinical boundary needs a defined question. For example: for which population, treatment regimen, exposure measure, and outcome is a change considered important? Evidence might include exposure-response analyses, pharmacodynamic studies, dose-ranging data, safety information, or clinical outcome data. The evidence must be relevant to the product and use being considered; a boundary inferred for one endpoint or population should not be transferred automatically to another.
Even with relevant evidence, a boundary may not be a sharp biological dividing line. A change in exposure may alter risk gradually rather than create a sudden transition from “safe” to “harmful.” Estimates may also be imprecise. In some settings, the evidence may support a range of concern or different limits for increases and decreases, rather than one definitive pair of ratios.
A clinical boundary and a regulatory acceptance interval can therefore have different widths and locations. The regulatory interval is an operational standard for a defined decision. The clinical boundary is a conclusion—or sometimes an uncertain hypothesis—about consequences. Neither should be substituted for the other without a justified regulatory and clinical argument.
Worked Example: A Clinical Concern Range Wider Than a Regulatory Interval
Assume, purely for illustration, that a clinical evidence review for a hypothetical medicine identifies an exposure ratio range from 0.70 to 1.40 within which no material change in a specified outcome is expected for a defined population and regimen. Separately, suppose the applicable prespecified regulatory interval is 0.80 to 1.25. Compare the limits and interpret a ratio of 1.30.
Step 1: Compare the intervals. The hypothetical clinical range extends below the regulatory lower limit because \(0.70<0.80\), and above the regulatory upper limit because \(1.40>1.25\). The ranges are not interchangeable.
Step 2: Locate the illustrative ratio. Since \(1.25<1.30<1.40\), the ratio 1.30 is outside the stated regulatory interval but inside the hypothetical clinical range.
Interpretation. This comparison does not establish a bioequivalence conclusion: an actual regulatory assessment uses the specified statistical criterion and study results, not a point ratio considered alone. Nor does it prove that a ratio of 1.30 is harmless in practice. The clinical range in this hypothetical example applies only to the stated outcome, population, and regimen, and is assumed to come from relevant evidence. It illustrates that a regulatory limit need not coincide with a clinical harm boundary.
Worked Example: A Clinical Concern Boundary Narrower Than a Regulatory Interval
Now suppose a different hypothetical clinical assessment identifies concern if the exposure ratio exceeds 1.10 for a specified outcome in a susceptible population. For illustration, assume the regulatory interval for the product remains 0.80 to 1.25. Compare a ratio of 1.16 with these two limits.
Step 1: Compare with the regulatory upper limit. Since \(1.16<1.25\), the ratio is inside the regulatory interval.
Step 2: Compare with the clinical concern boundary. Since \(1.16>1.10\), the ratio is beyond the stated clinical concern boundary for the specified context.
Interpretation. The two comparisons answer different questions. Being inside a regulatory interval does not disprove a clinical concern identified for a particular susceptible population or outcome. Conversely, this hypothetical comparison alone does not determine a regulatory decision; that decision depends on the applicable criteria and study evidence. The clinical boundary also cannot be generalized to other populations or outcomes without evidence that it applies.
Assumptions and Limits of Interpretation
Comparing any two limits is meaningful only when the quantities being compared refer to the same exposure measure and ratio definition. An AUC ratio, a peak-concentration ratio, and a pharmacodynamic response are not interchangeable quantities. A clinical boundary for one endpoint cannot be compared numerically with a regulatory interval for another without a defensible model linking them.
The comparison also requires that the clinical evidence apply to the relevant product, regimen, population, and outcome. If the clinical boundary is inferred from a model, the model's assumptions and uncertainty matter. For example, assuming a proportional relation between exposure and effect may be inappropriate if the actual relationship is nonlinear, has a plateau, or differs between patients.
Finally, regulatory acceptance is a rule-based conclusion, not a direct clinical-outcome experiment. Meeting a prespecified criterion supports the regulatory conclusion defined by that criterion. It does not prove that no patient can experience a meaningful difference, and it does not estimate the exact exposure at which harm begins. Clinical evaluation may require evidence and reasoning beyond the bioequivalence comparison itself.
Common Mistake and a Practical Check
Wrong reasoning: “The acceptance limits are 80–125%, so the drug is clinically harmless throughout that interval and harmful outside it.” This treats an operational statistical standard as if it were a directly observed clinical threshold.
Correction: The interval defines limits used in the prespecified regulatory decision rule. Clinical consequences require evidence for a specified outcome and context. A ratio within the interval is not, by itself, proof of clinical harmlessness for every individual or endpoint; a ratio outside it is not, by itself, proof that harm has occurred.
When interpreting a margin, ask three separate questions: What exposure measure and statistical parameter does the regulatory rule concern? What evidence supports the chosen limits as a regulatory standard? What clinical evidence, if any, identifies consequences at particular exposure differences? Keeping those questions separate prevents a decision criterion from being given a clinical meaning it was not designed to provide.
Check Your Understanding
Answer each question, then compare your reasoning with the solutions.
- What is the main difference between a regulatory acceptance interval and a clinical harm boundary?
- Why does the reciprocal relationship between 0.80 and 1.25 not prove that clinical effects are symmetric?
- A hypothetical clinical assessment suggests concern above a ratio of 1.15, while a regulatory interval extends to 1.25. Does being inside the regulatory interval disprove that clinical concern? Explain.
- Why must the exposure measure and clinical outcome be specified before comparing two limits?
- What does a ratio inside a regulatory interval establish by itself?
Solutions. (1) The regulatory interval is part of a prespecified regulatory decision rule; a clinical harm boundary concerns expected or unacceptable consequences for a defined outcome and context. (2) The reciprocal limits are symmetric after taking logarithms, but that is a mathematical property and says nothing by itself about the shape or symmetry of a clinical exposure-response relationship. (3) No. The regulatory limit and the hypothetical clinical concern boundary address different questions; the clinical evidence may indicate concern before the regulatory upper limit, while the formal regulatory decision still requires its specified statistical assessment. (4) Different exposure measures and outcomes are different quantities, and a numerical boundary for one cannot be transferred to another without a justified link. (5) It establishes only that the ratio is located within the interval. A regulatory conclusion requires the full prespecified decision criterion, and clinical consequences require relevant clinical evidence.