Why Meal Deviations Matter
A fed study asks a specific question under a specified condition: how does a formulation perform when administered with the planned meal and dosing schedule? The earlier material on standardizing meal composition and timing explains how that condition is defined. This tutorial addresses what follows when a participant does not experience it as planned. An incomplete meal, a dose taken at the wrong time, vomiting, or additional food may change the conditions around drug administration. The resulting concentration data are still observations, but their connection to the intended fed condition may be uncertain.
A deviation does not automatically make data unusable. Its importance depends on what happened, when it happened, what the protocol prespecified, and whether the event could alter drug absorption or the interpretation of the comparison. A defensible assessment distinguishes recording a deviation from deciding how the data may be used. The latter should follow prespecified rules and be documented, not chosen after seeing whether a result is favorable.
A Framework for Assessing a Deviation
The relevant standard is the protocol for the particular study, not a universal threshold for meal completion or timing. The protocol should describe the intended meal and administration schedule, any permitted windows, relevant food restrictions, how deviations are recorded, and how their consequences are handled. A percentage of a meal consumed or a number of minutes outside a schedule cannot be declared acceptable in general: its meaning depends on the protocol and on the drug and formulation.
For each event, establish the facts before judging its impact. Useful records include the meal start and finish times, dose time, amount of the meal consumed, what food or drink was taken, the time and nature of vomiting, and the participant's account of events. Record uncertainty as uncertainty rather than turning an estimate into a precise measurement. Then ask whether the event could plausibly alter the condition relevant to the study and whether the same type of departure occurred across formulations or periods.
Record the event and its timing using source information where possible. Separate observed facts from estimates or missing details.
Identify the exact meal, timing, or restriction that was not met, and determine whether the protocol allows a window or specifies a response.
Ask whether the event could change absorption or make it unclear whether the observation represents the intended fed condition. Consider the event's timing and extent, not just its label.
Retain, qualify, or exclude data only as justified by protocol and analysis rules. Preserve the event record and explain any impact on interpretation.
This framework separates two questions that are often confused: whether a participant departed from the protocol, and whether that departure invalidates a particular observation for a particular purpose. A deviation can be reportable but have little plausible effect on the fed condition. Conversely, an event whose exact size is unknown may materially undermine attribution of observed exposure to the planned meal and dose.
Why Unequal Deviations Complicate a Comparison
The interpretive problem can be seen with a simple contrast. Let \(Y_T^*\) and \(Y_R^*\) denote the outcomes that would be observed for the test and reference formulations under the intended fed condition. Let \(Y_T\) and \(Y_R\) denote the outcomes actually observed. The observed difference can be written exactly as
The first term is the difference under the intended condition. The bracketed term represents the difference between the deviations' effects on the two observed outcomes. This is an accounting identity, not an assumption that meal effects are biologically additive or known. If both periods were truly conducted as planned, the bracketed term is zero. If meal-related departures differ between periods, the observed contrast need not represent the intended fed-condition contrast. Even when the same type of event occurs in both periods, equal effects cannot be assumed without justification.
This is why a deviation's direction cannot usually be predicted from the event label alone. Food may alter absorption in ways that depend on drug properties and formulation, while vomiting may make the amount of dose retained uncertain. The event may increase, decrease, delay, or have little discernible impact on a measured outcome. An analyst should not “correct” a result by guessing the direction or size of an effect.
Worked Example: An Incomplete Meal
Suppose a hypothetical protocol specifies an 800-kcal meal, with the dose administered after the meal is completed. It requires staff to record the amount consumed but gives no numeric tolerance for an incomplete meal. One participant consumes an estimated 600 kcal before dosing. What can be concluded?
The consumed fraction is \(600/800=0.75\), or 75% of the specified meal. This calculation describes the departure; it does not establish that 75% is acceptable or that the observation must be excluded. The protocol has not defined a numerical completion threshold.
The record should state the estimated amount consumed, whether the estimate is reliable, and that dosing occurred before the specified meal was fully completed. The reviewer should check whether the protocol or a prespecified deviation-handling rule addresses this situation. If it does, that rule governs. If it does not, the event should be reviewed and its consequences qualified rather than silently reclassified as a compliant fed administration.
If the same participant's comparator period had a fully completed meal, the fed conditions were not clearly matched. The observed comparison therefore cannot be assumed to isolate formulation performance under the planned meal condition. The data may remain useful for other purposes, but the intended fed-state interpretation is weakened.
Timing, Vomiting, and Unscheduled Food
Incorrect timing can arise even when the participant eats the correct meal. The relevant comparison is between the actual meal and dose timestamps and the schedule, including any permitted window, stated in the protocol. A dose given before a required meal, after an unplanned delay, or before meal completion may represent a different condition from dosing at the planned point. The elapsed time should be calculated from recorded times, while recognizing that a timestamp difference alone does not tell us the biological size of the resulting effect.
Vomiting raises a distinct concern: it may make the amount of the administered dose that was retained uncertain. Its relevance depends, among other factors, on when it occurred relative to dosing, whether the event was witnessed or described reliably, and the protocol's instructions. Do not infer that a dose was wholly lost, partly absorbed, or unaffected from the mere fact that vomiting occurred. If the event's timing or details are unknown, say so and limit the interpretation accordingly.
Unscheduled food or caloric drinks may also alter the planned conditions. Record what was consumed and when, including whether the item was a prohibited snack, a meal, or a beverage with calories. A small intake far from dosing and a substantial meal near dosing should not be treated as equivalent merely because both are “unscheduled food.” The study protocol and the event's timing guide the assessment; there is no general amount below which additional food can be presumed irrelevant.
Worked Example: Dose Taken Outside the Planned Window
In a hypothetical study, the protocol requires dosing 30 minutes after the recorded start of the meal, with a permitted window of 25 to 35 minutes. A participant begins eating at 08:00 and takes the dose at 08:45. Assess the timing deviation.
The actual interval is \(08{:}45-08{:}00=45\) minutes. The permitted interval is 25 to 35 minutes, so the dose was 10 minutes beyond the upper limit: \(45-35=10\) minutes. This is outside the stated window, even though it is only 15 minutes later than the protocol's target of 30 minutes.
The record should include both the target and actual interval and note that the dose was outside the allowed window. It should also establish whether the participant had completed the meal by 08:45 and whether any other food was consumed. The arithmetic establishes a protocol deviation; it does not quantify how much exposure may have changed. Handling follows the prespecified rule, and the observation should not be described as fully compliant with the planned timing.
Worked Example: Vomiting and Unscheduled Food in One Period
In a hypothetical crossover study, a participant vomits 20 minutes after dosing in one period. The participant cannot recall whether vomiting began 10 or 30 minutes after the dose. In the other period, the participant eats a snack 40 minutes after dosing despite a protocol restriction. How should these events be interpreted?
For the vomiting event, the reported range spans materially different intervals after dosing. Because the exact time and retained dose are unknown, the event cannot be converted into a reliable estimate of absorbed amount. Record the uncertainty and apply the protocol's vomiting rule; do not assume complete dose loss or full absorption.
For the snack, record its type, estimated amount if known, and time. The snack occurred 40 minutes after dosing, but that fact alone does not establish that it had no effect. It is a separate protocol departure in the other period. Because the two periods have different events, the comparison may be less informative about the planned fed condition. A reviewer should assess each event under the same prespecified criteria, document the limits on interpretation, and avoid attributing any observed difference solely to formulation.
Diagnostics, Alternatives, and Limits
A useful review checks for patterns as well as individual events. Compare meal completion, timing, vomiting, and additional food across participants and periods, particularly by formulation. Confirm that timestamps are internally consistent and that source records support derived intervals. Look for missing meal records, retrospective estimates, repeated departures, or events clustered in one treatment period. A deviation listing can reveal these issues, but a listing alone does not establish their effect on exposure.
There are three broad approaches to handling affected observations, and none is universally correct:
| Approach | When it may be appropriate | Main limitation |
|---|---|---|
| Retain with documentation | The deviation is minor under the protocol, or the prespecified rules call for retaining the observation. | Retention does not make the administration compliant; interpretation may still need qualification. |
| Qualify the interpretation | The data are recorded and potentially informative, but the intended meal condition is uncertain. | Qualification describes uncertainty; it does not remove or quantify its effect. |
| Exclude under a prespecified rule | The protocol or analysis plan defines the event as grounds for exclusion from a stated analysis. | Exclusion reduces available information and can bias a comparison if applied selectively or after results are known. |
A common alternative is to analyze only participants or periods judged compliant. This can be useful when the rule was defined in advance and applied uniformly, but it is not automatically superior to retaining all observations. If exclusions depend on treatment period, participant characteristics, or observed results, the remaining comparison may no longer represent the planned study population. Conversely, including a period with an important, poorly characterized meal departure without qualification can overstate how directly the result answers the intended question. Report the rule, the events, and the consequences; do not choose an approach simply because it produces a preferred result.
The principal limitation is that documentation cannot reconstruct an unobserved compliant outcome. A precise account of a deviation improves transparency, but it does not reveal exactly what exposure would have occurred under the planned meal. The aim is therefore not to manufacture certainty: it is to preserve the distinction between the protocol condition, what actually happened, and what the data can reasonably support.
Common Mistake
Wrong reasoning: “The participant ate most of the meal, so the period is effectively fed and can be treated as compliant.” This substitutes an informal judgment for the protocol and assumes that the uneaten portion is inconsequential.
Correction: Report the amount consumed and compare it with the protocol's requirements and prespecified rules. If no threshold was specified, do not invent one after the event. Determine whether the departure could affect the intended interpretation, apply the same review principles across periods, and qualify the conclusion when the planned condition cannot be established.
Check Your Understanding
Answer each question, then compare your reasoning with the solutions below.
- A protocol specifies a meal of 700 kcal but gives no completion tolerance. A participant consumes 560 kcal. What fraction was consumed, and does that calculation alone establish compliance?
- A protocol permits dosing 20 to 30 minutes after meal start. The meal starts at 12:10 and the dose is given at 12:44. Is the dose within the window?
- Why should vomiting after dosing not automatically be treated as complete loss of the dose?
- One formulation period has an incomplete meal and the other has a fully completed meal. What interpretive concern follows?
- Why can excluding every noncompliant period after reviewing the results create a problem?
Solutions. (1) \(560/700=0.80\), so 80% was consumed. This describes the fraction but does not establish compliance; the protocol has no stated tolerance, so the event requires documentation and assessment. (2) The interval is \(12{:}44-12{:}10=34\) minutes, which is 4 minutes beyond the upper limit of 30 minutes; it is outside the window. (3) The amount retained depends on timing and other details that may be unknown, so complete loss cannot be inferred from vomiting alone. (4) The periods do not clearly represent the same planned fed condition, so the observed contrast may not isolate the intended formulation comparison. (5) Post hoc exclusions can be selective and may alter the comparison or discard information in a way that depends on observed events or results; rules should be prespecified and applied consistently.