Start With the Question Being Compared
Fed and fasted results are interpretable only after identifying the contrast they represent. Comparing the same formulation given under fed and fasted conditions addresses a food effect. Comparing test and reference formulations under the same fed condition addresses fed-state formulation performance. These are different questions, even if both studies collect plasma concentration data after dosing.
For a food-effect comparison, the useful basic unit is the participant’s pair of results: one exposure summary under fasting conditions and one under fed conditions. For a fed-state formulation comparison, it is the participant’s pair of results after receiving test and reference formulations under the same meal condition. A difference between two condition-level averages does not, by itself, answer both questions.
The earlier material on distinguishing food-effect and fed-state bioequivalence questions establishes this central distinction. Here, the emphasis is on reading paired results alongside the study conduct that produced them. The numerical summaries below are descriptive; they do not substitute for a formal bioequivalence analysis or establish regulatory acceptability.
How to Summarize a Paired Comparison
Suppose each of \(n\) participants has an exposure summary under each of two conditions. Let \(Y_{i,F}\) be participant \(i\)’s summary when fasted and \(Y_{i,D}\) the summary when fed. The values must refer to the same endpoint, units, dose, and analysis rules. A participant-level difference is \(d_i=Y_{i,D}-Y_{i,F}\). A positive \(d_i\) means that participant’s observed value was higher in the fed condition.
A descriptive ratio of the two condition means, \(\bar{Y}_{D}/\bar{Y}_{F}\), can express the relative scale of the summaries, provided the fasted mean is positive. It is not the same quantity as the mean of participants’ individual ratios. State which quantity is reported and keep the units clear: a ratio has no units, whereas a difference in AUC or \(C_{\max}\) retains the endpoint’s units.
The paired structure matters because each participant supplies both measurements. With complete pairs, the mean of the paired differences equals the difference between the two condition means:
This identity is a direct consequence of distributing the sum across subtraction. It does not mean the pairing is irrelevant: pairing is needed to show how condition-specific values correspond within participants, to describe variation in individual changes, and to support an appropriate paired analysis if one is later required. If the two condition means come from different sets of participants, this identity does not apply to a paired mean difference.
The interpretation relies on comparable measurements and credible study conduct. In a crossover, the same participant should receive the conditions under the planned dose and sampling procedures; treatment order should be randomized or otherwise controlled, and the washout should be adequate for the drug and design. Predose restrictions, meal composition and timing, dose timing relative to the meal, sample timing, and endpoint calculation must follow the protocol. These conditions support the comparison; they do not prove that every potential source of bias has been removed.
Worked Example: Describing a Food Effect on AUC
In a hypothetical crossover study, four participants receive the same formulation once fasted and once after the standardized meal. Their AUC values are in mg·h/L. Assume the meal and dose timing were protocol-compliant, the same AUC definition and sampling schedule were used in both periods, all four pairs are complete, and washout was adequate.
| Participant | Fasted AUC | Fed AUC | Fed − fasted |
|---|---|---|---|
| 1 | 8 | 12 | 4 |
| 2 | 10 | 15 | 5 |
| 3 | 9 | 11 | 2 |
| 4 | 11 | 14 | 3 |
The fasted mean is \((8+10+9+11)/4=38/4=9.50\) mg·h/L. The fed mean is \((12+15+11+14)/4=52/4=13.0\) mg·h/L. The paired differences average \((4+5+2+3)/4=14/4=3.50\) mg·h/L, which agrees with \(13.0-9.50=3.50\) mg·h/L.
The descriptive ratio of condition means is \(13.0/9.50=1.368\). In this hypothetical sample, mean AUC was 3.50 mg·h/L higher in the fed condition, and the fed mean was about 1.368 times the fasted mean. This describes the observed comparison for this formulation and study setup. It does not show that food caused the difference beyond uncertainty, establish a formal food-effect conclusion, or demonstrate bioequivalence.
Keep Formulation Performance Separate
A food-effect result for one formulation cannot be used as a substitute for a formulation comparison. To describe fed-state formulation performance, compare the formulations under the same fed conditions. The meal and the dose timing relative to it should be common to both treatments, as should the endpoint and sampling rules. A paired crossover can make this comparison within participants, but its interpretation still depends on adequate conduct and the planned design.
Worked Example: Comparing Formulations Under Fed Conditions
In a separate hypothetical crossover, four participants receive test and reference formulations at the same dose, each under the same standardized fed condition. Assume adequate washout, valid predose samples, matched sampling procedures, and no meal or dose-timing deviations. The AUC values are in mg·h/L.
| Participant | Test AUC | Reference AUC | Test − reference |
|---|---|---|---|
| 1 | 9 | 10 | −1 |
| 2 | 10 | 10 | 0 |
| 3 | 8 | 9 | −1 |
| 4 | 11 | 12 | −1 |
The test mean is \((9+10+8+11)/4=38/4=9.50\) mg·h/L. The reference mean is \((10+10+9+12)/4=41/4=10.25\) mg·h/L. The mean paired difference is \((-1+0-1-1)/4=-3/4=-0.750\) mg·h/L, equal to \(9.50-10.25\). The descriptive test-to-reference ratio of means is \(9.50/10.25=0.9268\).
The observed test mean is lower than the reference mean in this sample. This is a fed-state formulation comparison, not an estimate of the food effect: both formulations were administered fed. The descriptive ratio alone does not establish bioequivalence; a formal conclusion requires the appropriate analysis and decision criteria, which are not calculated here.
If the question is whether food affects each formulation differently, the design must include both formulations in both food conditions. A fed comparison alone cannot reveal how the same formulation behaves fasted, and a food-effect comparison for only one formulation cannot establish that the effect is the same for another. The earlier material on formulation-by-food interactions addresses that distinct question.
Study Conduct Is Part of the Result
Exposure summaries do not explain themselves. A numerical difference can be interpreted as a food-related contrast only when the fed condition represents the planned meal and dosing procedure. Earlier material on fasting restrictions, meal composition, dose timing, and meal deviations supplies the operational detail; when interpreting results, check the actual conduct rather than relying solely on the protocol description.
Worked Example: A Meal-Timing Deviation
Suppose a hypothetical crossover specifies dosing 30 minutes after the start of a standardized meal in the fed period. Three participants follow that timing and have paired AUC values shown below. A fourth participant takes the dose 90 minutes after the meal starts. The fasted periods were conducted as planned. Values are in mg·h/L.
| Participant | Fasted AUC | Fed AUC | Fed-period conduct |
|---|---|---|---|
| 1 | 10 | 14 | Protocol timing |
| 2 | 9 | 13 | Protocol timing |
| 3 | 11 | 15 | Protocol timing |
| 4 | 10 | 18 | Dose at 90 minutes |
For the three participants with protocol-timed fed dosing, the fasted mean is \((10+9+11)/3=10.0\) mg·h/L and the fed mean is \((14+13+15)/3=14.0\) mg·h/L. Their mean paired difference is \((4+4+4)/3=4.00\) mg·h/L. Including participant 4 gives a fasted mean of \((10+9+11+10)/4=10.0\) mg·h/L, a fed mean of \((14+13+15+18)/4=15.0\) mg·h/L, and a mean paired difference of \((4+4+4+8)/4=5.00\) mg·h/L.
The arithmetic shows that the observed summary changes when the deviation is included. It does not establish that delayed dosing caused participant 4’s higher AUC; individual variability and other factors remain possible explanations. The defensible interpretation is that the all-participant descriptive result includes a fed period that did not follow the planned timing. Whether to exclude or otherwise handle that period must follow prespecified study rules and a reasoned review, not a desire to obtain a preferred estimate.
Useful checks include confirming actual meal start and finish times, dose times, meal completion, predose concentrations, sample times, missing samples, and endpoint calculations. Check that each participant’s fed and fasted values are genuinely paired and that the same units and endpoint definition were applied. Examine the participant-level differences rather than only the two means: a mean shift driven by one unusual pair may tell a different descriptive story from a consistent shift across participants. Such checks diagnose interpretive concerns; they do not by themselves replace a statistical analysis.
Alternatives, Limits, and a Common Pitfall
| Approach | Question it addresses | Key caution |
|---|---|---|
| Paired fed–fasted summaries for one formulation | How observed exposure differs by food condition | Requires interpretable pairs and comparable conduct |
| Paired test–reference summaries under fed conditions | How observed exposure differs by formulation when fed | Does not estimate a food effect |
| Separate fed and fasted participant groups | A condition contrast between groups | Does not use within-participant pairing and may differ in participant composition |
| Formal model-based analysis | Estimation and inference under a specified design and model | Requires appropriate prespecification, assumptions, and decision criteria |
A common mistake is to see a higher fed mean and conclude that the test formulation performs better. That reasoning confuses condition with formulation: if the same formulation was used in both periods, the contrast concerns food conditions. The correction is to name the treatments being compared before interpreting the numbers, then state the result conditionally and descriptively.
These summaries are limited. A small or selectively presented set of participants may not represent the study population; missing or invalid periods can disrupt pairing; and meal or dosing deviations can undermine the intended contrast. Arithmetic means and their ratios are useful for illustrating observed patterns, but they are not a substitute for the later formal log-scale and crossover analyses. Do not infer statistical significance, bioequivalence, or a causal mechanism from a descriptive table alone.
Check Your Understanding
Answer each question using the comparison and the study conduct described.
- A participant receives the same formulation fasted and fed. What question does the paired comparison address?
- In a fed crossover, four test AUC values are 8, 9, 10, and 11 mg·h/L; reference values are 9, 9, 11, and 11 mg·h/L. Calculate both means, the mean paired difference (test minus reference), and the ratio of means.
- Why is a fed mean from a test formulation not, by itself, evidence that the test formulation is bioequivalent to a reference formulation?
- One fed period used a meal different from the protocol meal. What should be checked before interpreting the fed–fasted contrast?
Solutions. 1. It addresses the observed food effect for that formulation, because the same formulation is compared across fed and fasted conditions. 2. The test mean is \((8+9+10+11)/4=38/4=9.50\) mg·h/L; the reference mean is \((9+9+11+11)/4=40/4=10.0\) mg·h/L. The paired differences are \(-1,0,-1,0\), so their mean is \(-2/4=-0.500\) mg·h/L, equal to \(9.50-10.0\). The ratio of means is \(9.50/10.0=0.9500\). These are descriptive results, not a formal bioequivalence conclusion. 3. Bioequivalence requires a comparison with the reference under an appropriate design, analysis, and decision criteria; a test-only mean supplies no reference comparison. 4. Verify the meal’s composition and timing, the actual dose time relative to the meal, other relevant conduct and sampling details, and the prespecified handling of the deviation. The result should be interpreted with the deviation disclosed; it should not be silently treated as protocol-compliant.