Why a Protocol Deviation Can Change the Interpretation
A food-effect or multiple-dose result answers a conditional question: what exposure was observed under specified meal, dosing, and sampling conditions? If those conditions differ from the ones intended, the measured pharmacokinetic endpoints may no longer support that interpretation. A high \(C_{\max}\), for example, might reflect a formulation difference, a different meal, or a different interval between the meal and dose. The concentration alone cannot identify which explanation is correct.
The central task is therefore to reconstruct what actually happened before interpreting an endpoint. Check the meal, dose, and sample records against the protocol and against each other. A deviation does not automatically invalidate every observation, but it can weaken a comparison, change the meaning of an endpoint, or make the source of a difference uncertain. The earlier material on meal deviations and the worked interpretation of fed and fasted results provide context for this diagnostic approach.
The examples below use hypothetical data. Where a numerical pharmacokinetic model is used, it is stated explicitly; the model is an aid to understanding a possible distortion, not a universal prediction for every drug. In general, the direction and size of a deviation’s effect depend on the drug, formulation, meal, timing, and endpoint.
Unmatched Meals: A Comparison of More Than Formulations
A fed-state comparison is interpretable as a formulation comparison only when the meal conditions are sufficiently comparable across the treatments being contrasted. Relevant features include meal composition and amount, as well as the interval between starting or finishing the meal and taking the dose. If one treatment is given with a high-fat meal and another with a lower-fat meal, the observed contrast combines any formulation difference with the effect of those different meal conditions. The comparison does not isolate either component.
This is not fixed by simply labelling both observations “fed.” That label does not establish that the participants received comparable meals or followed the same administration schedule. Review meal records, not just the planned meal category. For interpretation, distinguish a valid matched-condition contrast from a descriptive comparison of observations collected under different conditions.
Worked Example: An Unmatched Fed Comparison
Suppose, hypothetically, the reference formulation has an AUC of 130 exposure units with a specified high-fat meal. Under the same meal and schedule, the test formulation would have an AUC of 95% of the reference value. Instead, the test is administered with a different meal, under which its hypothetical AUC is 123.5 exposure units. A separate reference observation under the different meal condition is not available.
The matched high-fat comparison would be \(0.95 \times 130 = 123.5\) exposure units. But the observed value of 123.5 for the test under a different meal does not establish that its exposure is 95% of the reference under the high-fat meal: it is the result of the stipulated formulation and meal combination. Without a reference observation under the test meal condition, the formulation and meal contributions cannot be separated from these data.
The arithmetic illustrates the issue: an apparent contrast against the high-fat reference would be \(123.5/130=0.950\), but that numerical ratio compares different meal conditions. It is not, on its own, an interpretable matched-meal formulation contrast. The appropriate conclusion is that the conditions are unmatched and the intended comparison is uncertain, not that the test formulation is equivalent or different.
Incorrect Dose-to-Meal Timing
The prescribed relation between meal and dose is part of the condition being studied. A dose taken before, during, or after a meal may not represent the same condition as a dose taken at the specified interval. Changing that interval can alter absorption rate and therefore \(T_{\max}\) and \(C_{\max}\); depending on the drug and circumstances, it may also affect AUC. There is no general rule that a timing error must raise or must lower every endpoint.
A useful diagnostic is to reconstruct elapsed time from recorded timestamps rather than relying on a general note such as “dose with breakfast.” Check whether the timestamp refers to meal start, meal completion, or another event, and compare it with the actual dose time. Then interpret each endpoint in light of the deviation. A changed interval makes the observed condition different from the planned one; it does not by itself quantify what the endpoint would have been under correct timing.
Worked Example: A Dose Taken Later Than Scheduled
In a hypothetical fed study, the protocol specifies that the dose be taken 30 minutes after the start of the meal. A participant starts eating at 08:00, finishes at 08:25, and takes the dose at 09:00. The elapsed time from meal start to dose is \(09{:}00-08{:}00=60\) minutes, which is 30 minutes later than specified. The elapsed time from meal completion is 35 minutes.
Those two calculations show why the protocol’s reference point matters: “35 minutes after finishing” is not the same statement as “30 minutes after starting.” If the planned rule was measured from meal start, this administration deviated by 30 minutes. The pharmacokinetic endpoints can still be calculated from the samples, but they describe the observed timing condition. Without additional evidence, one cannot calculate the participant’s counterfactual \(C_{\max}\) or AUC under the planned timing, nor attribute any difference from another treatment solely to formulation.
Sampling Before Steady State
For a multiple-dose interpretation that presumes steady state, sampling must occur after concentrations have approached their repeated-dose pattern under the specified regimen. “Several doses were given” is not enough to demonstrate this. The expected approach depends on the drug’s disposition and dosing interval; the planned criterion, available concentration history, and adherence should all be reviewed. The earlier material on establishing steady state and sampling across a dosing interval gives the relevant context for those checks.
A short derivation shows why early sampling can give a systematically different profile. Assume linear, time-invariant one-compartment disposition, instantaneous doses of size \(D\), elimination rate constant \(k\), and dosing interval \(\tau\). Let \(q=e^{-k\tau}\). At a fixed time after the \(n\)th dose, the contributions from preceding doses form a geometric sum:
Here \(V\) is the model’s apparent distribution volume. Dividing the finite-dose expression by its steady-state limit gives \(C_n/C_{\mathrm{ss}}=1-q^n\). Because \(0<q<1\), this fraction is less than one for finite \(n\). Under these assumptions, early repeated-dose concentrations have not yet reached the steady-state pattern. Oral absorption, variable dosing, or nonlinear pharmacokinetics can require a different description, so this calculation is illustrative rather than a universal correction.
Worked Example: Two Doses Are Not Necessarily Steady State
Assume the hypothetical model above, a 24-hour dosing interval, and an elimination half-life of 12 hours. Then \(q=e^{-k\tau}=2^{-24/12}=0.25\). After two doses, the concentration at the same post-dose time as the steady-state comparison is
Thus, under this model, the two-dose concentration is 93.75% of its steady-state value at that comparison time. The difference is 6.25% of the steady-state value. The calculation does not establish that two doses are adequate for a real study: it shows that a specific dosing history and half-life can leave a measurable gap. For study interpretation, verify the prespecified steady-state rationale and actual sampling schedule rather than declaring steady state from dose count alone.
Missed Doses and the Meaning of the Observed Profile
A missed dose changes the dosing history that generated subsequent concentrations. It can reduce exposure over an interval and alter trough concentrations or peak-to-trough patterns. The impact depends on when the dose was missed, when sampling occurred, the drug’s disposition, and whether later doses were taken as scheduled. A missed dose near sampling may matter more to a particular endpoint than one much earlier; there is no single adjustment that applies to all regimens.
The following simple calculation illustrates the mechanism. Retain the stated one-compartment, linear, instantaneous-dose assumptions. At steady state, just before a scheduled dose, the trough concentration is \(C_{\min,\mathrm{ss}}=(D/V)q/(1-q)\). If that scheduled dose is omitted, the concentration at the next scheduled dose time is the prior trough multiplied by \(q\), because no new dose was added at the missed time. This model isolates the effect of one omission; it is not a general estimator for oral drug exposure.
Worked Example: One Omitted Dose Before a Trough Sample
Let \(D/V=10\) mg/L and \(q=0.25\), using the illustrative model above. With all doses taken, the steady-state trough is
If the dose at the scheduled time is omitted, then at the next scheduled dose time the model gives \(qC_{\min,\mathrm{ss}}=0.25(3.333)=0.8333\) mg/L, before any dose at that time. The trough is therefore lower than the expected steady-state trough in this scenario. A measured low trough cannot be attributed to formulation or food without checking dose records and timing. In an actual study, absorption and other departures from this simplified model may change the magnitude and time course of the effect.
How to Diagnose and Respond to Deviations
A disciplined review starts with records and the intended comparison, not with an attempt to explain an unexpected result after the fact. Check meal composition and completion times, dose timestamps and dose records, predose restrictions, sampling times, and the sequence of events for each participant and period. Check whether reported concentrations and endpoints are consistent with that history. A suspicious concentration pattern can prompt a record review, but by itself it does not prove that a dose was missed or a meal was different.
The practical response depends on the deviation and the analysis plan. A minor departure that is unlikely to affect the endpoint may be documented without changing the primary interpretation. A deviation that makes treatment conditions incomparable should be clearly identified and may warrant a prespecified sensitivity analysis or a qualified interpretation. Excluding an observation is not automatically preferable: exclusion can change the analyzed population and may introduce bias if the reason for exclusion is related to treatment or outcome. Do not silently “correct” a measured endpoint using a generic factor.
| Observed problem | What it threatens | Useful check |
|---|---|---|
| Different meal composition or amount | Attribution of an exposure contrast to formulation or food condition | Compare actual meal records with the specified composition and with the other treatment condition |
| Dose outside the specified meal interval | Interpretation of the planned fed condition; potentially peak timing and magnitude | Calculate elapsed time using the protocol’s stated meal reference point |
| Sampling before the intended steady-state criterion | Interpretation of AUC over the dosing interval, peak, or trough as steady-state endpoints | Review dose history, time allowed for accumulation, and sampling schedule |
| Missed or late dose | Interpretation of concentrations as resulting from the scheduled regimen | Reconcile dose records with sample times and inspect the affected interval |
The main alternative to interpreting every observation under the planned condition is to describe the observed conditions directly and qualify the inference. Where an analysis plan permits, a sensitivity analysis can compare conclusions with and without observations affected by a substantial deviation. These approaches answer different questions: the primary analysis follows its prespecified rules, while a sensitivity analysis examines how dependent the conclusion is on particular data-handling choices. Neither approach repairs a confounded comparison or proves what would have happened under perfect adherence.
Common Mistakes and a Practical Standard
- Wrong reasoning: Both treatments were labelled “fed,” so their meals were matched. Correction: Confirm actual meal composition, amount, and timing; the label alone is insufficient.
- Wrong reasoning: A timing deviation must increase or decrease \(C_{\max}\) in a known direction. Correction: Timing can change the absorption profile, but the direction and size depend on the drug and circumstances. Interpret the observed condition and avoid a universal prediction.
- Wrong reasoning: Sampling after a few doses means steady state. Correction: Steady state depends on the disposition and schedule; evaluate the planned criterion and actual history.
- Wrong reasoning: A low concentration proves a missed dose. Correction: A low value is a signal to investigate, not proof of its cause. Check timing, meal, sampling, and dose records.
- Wrong reasoning: Remove every deviation and the remaining data automatically give an unbiased answer. Correction: Exclusion has consequences and should follow a justified, prespecified rule where possible; report the deviation and assess sensitivity when appropriate.
Check Your Understanding
For each situation, identify the main interpretive concern and the first check you would make. Solutions follow the questions.
- One treatment was administered with the specified high-fat meal; the other was given with a lower-fat meal. Can their AUC ratio be interpreted as a matched-meal formulation comparison?
- A dose was scheduled 30 minutes after meal start. The participant took it 50 minutes after meal start. By how many minutes did the actual timing differ from the specified timing, and what does that establish?
- Under the simple repeated-dose model, \(q=0.20\). What fraction of the steady-state concentration is reached after three doses at the same comparison time?
- A trough concentration is unexpectedly low after one scheduled dose may have been missed. Name one appropriate record check and one conclusion that should not be drawn from the concentration alone.
Solutions. (1) No. The meal conditions are unmatched, so the contrast combines treatment and meal-condition differences. Check actual meal composition, amount, and timing for both treatments. (2) The difference is \(50-30=20\) minutes. This establishes that the observed dose-to-meal interval departed from the specified one; it does not calculate the endpoint under correct timing. (3) The fraction is \(1-q^3=1-(0.20)^3=1-0.008=0.992\), or 99.2%, under the stated model and assumptions. (4) Reconcile dose records and timestamps with the trough sample time. The low concentration alone does not prove a missed dose or identify its cause.