Predose Samples and Dose Attribution

Why Predose Status and Dose Attribution Need Separate Decisions

A concentration sample collected just before a scheduled dose is often called a predose sample. That label describes its position relative to the dose it was collected before. It does not, by itself, identify the dose that most recently preceded the sample. In a repeated-dose regimen, a predose sample may follow several earlier doses; the most recent actual administration is the relevant preceding dose for a sample-to-dose attribution.

These distinctions matter when preparing pharmacometric data. A scheduled dose may have been missed, held, delayed, or only partly administered. A sample can therefore be before a planned dose without being before an actual dose at that scheduled time. Similarly, recorded times can be tied or imprecise. A reproducible dataset should distinguish what the records establish from what requires review, rather than infer a dose from the nominal regimen alone.

Definition: A sample is predose relative to a target dose when its sample event precedes that dose event under the established event-ordering convention and any applicable protocol definition of a predose collection. The preceding dose for a sample is the latest confirmed actual administration that precedes the sample in the ordered event history. These are different relationships: a sample may be predose relative to an upcoming dose while following an earlier administration.

Build the Attribution From the Recorded History

Start with the subject’s dose and sample records and the regimen history needed to interpret them. The earlier material on reconstructing dose histories distinguishes actual administrations from regimen plans and from dose events that did not occur. Keep that distinction here: a planned dose is not an actual administration unless the records confirm that it was given. A missed or held dose contributes no administration event. For a partial dose, retain the confirmed administration and its actual amount; do not substitute the planned amount.

Represent each confirmed administration by its recorded time and its position in the event ordering. Represent a sample in the same time basis and ordering. For a bolus, the dose event is ordinarily represented at its administration time. For an infusion, retain the start and end information: a sample during an ongoing infusion is not simply a predose or a sample after a completed infusion. It should be identified as occurring during administration when the records establish that relationship, and reviewed or represented appropriately for the intended analysis.

Definition: Let $s$ denote a sample event and let $D$ be the finite set of confirmed actual administration events in the relevant history. Let $\prec$ be the event ordering formed from recorded time and the established convention for ordering simultaneous events. If one or more administrations precede the sample, its preceding administration is the latest member of that set: $$ d^* = \max_{\prec}\{d_j \in D : d_j \prec s\}. $$ The ordering convention is part of the definition: if recorded times tie, the established simultaneous-event order determines which event is later. If no administration precedes the sample, there is no preceding actual dose in the available history. If the necessary ordering cannot be established, attribution is unresolved rather than uniquely defined.

When timestamps are sufficiently precise, the elapsed time from the attributed dose to the sample can be calculated as $s-d^*$, using a consistent time unit. This is elapsed time, not a substitute for checking whether the history is complete or whether a dose was actually administered. If dates, time zones, or time precision are incompatible, do not report an apparently exact interval.

Confirm the events.
Identify the sample record and the confirmed administration records. Separate actual administrations from planned, missed, or held doses; preserve actual partial amounts.

Put events on a comparable timeline.
Check date, clock time, time zone, and time precision. Apply the established ordering convention for events with tied recorded times.

Determine sample relationships.
Check whether the sample precedes a target dose and separately find the latest actual administration preceding the sample. Check for an ongoing infusion.

Record the result and its certainty.
Store the attributed administration, time since that administration when calculable, predose status relative to the target dose, and a review flag when the evidence is insufficient or conflicting.

Why the Ordered Selection Is Unique

The maximum in the definition is taken under the full event ordering, not under recorded time alone. This matters when two administrations have the same timestamp. A time-only rule could return two maximizers, even though the established event order distinguishes them.

The justification is direct. The set of administrations preceding a particular sample is finite. If the event-ordering convention places every relevant pair in a definite order, that finite nonempty set has exactly one latest event. Therefore the selected preceding administration is unique. If simultaneous events are not ordered, or records do not establish which event came first, the required ordering is not available; a unique selection cannot be justified from those records. In that case the correct data-preparation outcome is a review flag, not an arbitrary tie-break.

Worked Example: A Predose Sample After the Previous Day’s Dose

Suppose a hypothetical subject has a confirmed 100 mg oral administration at 08:00 on Day 1 and another confirmed 100 mg administration at 08:00 on Day 2. A sample is recorded at 07:55 on Day 2, before the Day 2 administration. Assume the date and time are precise and there are no other administrations in the history.

The sample precedes the Day 2 dose, so it is predose relative to that target dose. The only confirmed administration preceding the sample is the Day 1 dose. Thus the sample follows the Day 1 dose; it is not attributed to the Day 2 dose that has not yet occurred.

The elapsed time is $23$ hours and $55$ minutes. In hours: $$ (24 \times 1) - \frac{5}{60} = 24 - 0.08333 = 23.9167\ \text{h}. $$ The sample is therefore predose relative to the Day 2 administration and was collected 23.92 h after the preceding confirmed administration. These statements describe event timing; they do not, on their own, establish a concentration interpretation or prove that the regimen was followed outside the recorded events.

Handling Tied Times, Changed Regimens, and Uncertainty

A tied recorded time is not automatically an ambiguous attribution. If the established event-ordering convention orders the simultaneous events, use that ordering consistently. For example, if two confirmed administration events both have a recorded time of 08:00, the convention places administration A before administration B, and a sample occurs at 09:00, then B is the latest preceding administration under that ordering. However, two administrations recorded at the same time may also indicate a duplicate record, a data-entry problem, or an unusual dosing history. The deterministic attribution does not remove the need to investigate a clinically unexpected history.

The same care applies to regimen changes and interruptions. Regimen history helps identify which administrations were expected, but expectations do not establish that a dose occurred. Use records of actual administration to select the preceding dose. If an expected dose was held, the sample may follow an earlier administration by a longer interval than the nominal schedule suggests. If a dose was delayed, use its confirmed time rather than its planned time. If only the planned time is available, record that actual timing is not established and flag the attribution as uncertain where that uncertainty affects the conclusion.

Worked Example: Selecting the Latest Dose Under a Tied-Time Order

Consider a hypothetical history with two confirmed administration events at 08:00 on the same date. The established event ordering places event A before event B when recorded times tie. A sample is recorded at 09:00. Assume the sample time is precise, both administrations are valid records, and the simultaneous-event convention applies to these events.

Both administrations precede the sample. Their recorded clock times are equal, so a rule that selects only the largest time would leave both as candidates. Under the full event ordering, however, $A \prec B \prec s$. The latest preceding administration is B. A time-since-dose calculation based only on clock times gives: $$ 09{:}00 - 08{:}00 = 1.00\ \text{h}. $$ This elapsed time is the same from either tied timestamp, but the dose-event attribution is B because the ordering convention resolves the event tie. The duplicated or simultaneous administration records should still be checked against the source history; if their validity or order is not established, flag the attribution for review rather than silently selecting B.

Worked Example: A Missed Dose and an Uncertain Timestamp

In a hypothetical regimen, doses are planned for 08:00 each day. A confirmed administration is recorded at 08:00 on Day 1. The Day 2 dose is documented as held, and there is no confirmed administration on Day 2. A sample is collected on Day 3, but its time is recorded only as “morning”; the next confirmed administration is recorded at 08:00 on Day 3.

The held Day 2 dose is not an administration and cannot be selected as the sample’s preceding dose. If the Day 3 sample is known to have occurred before the 08:00 administration, its preceding confirmed dose is the Day 1 administration. If it occurred after 08:00, the Day 3 administration is the latest preceding dose. “Morning” does not determine which case applies, so the records do not support a unique attribution.

A correct preparation records the Day 2 dose as held, retains the two confirmed administrations, and flags the sample-to-dose attribution as unresolved because sample timing is insufficient. It must not assume that the sample was predose merely because it was collected in the morning or because a dose was planned for that day. No exact elapsed time can be calculated from “morning”; reporting a numerical interval would imply precision absent from the records.

Diagnostics, Alternatives, and Limits

Useful checks focus on whether the event history supports the assigned relationship. Review samples assigned to a dose that occurs after the sample, samples with no preceding administration despite a known prior treatment history, and samples whose calculated elapsed time is negative. Also inspect repeated same-time administrations, unexplained gaps, inconsistent date or time-zone conventions, and samples falling within an infusion interval. Such checks identify records needing investigation; they do not establish which source record is correct.

A transparent alternative to automatic attribution is manual review of every sample against source records. This may be appropriate for a small dataset or when the event ordering is incomplete, but it is labor-intensive and can produce inconsistent decisions unless the review rules are documented. Automated selection is reproducible when actual administrations and a complete event ordering are available. It is not a remedy for missing history, uncertain timestamps, or mistaken event coding. In those cases, flagging and resolving the underlying data are preferable to adding an undocumented rule.

Predose classification may also depend on a protocol-defined collection window or target-dose definition. Apply that definition when it is available; do not invent a universal time window. If the records establish only that a sample occurred before a dose, describe that temporal relationship without claiming that it meets a protocol-specific predose criterion. Likewise, attribution identifies a temporal relationship in the recorded history; it does not demonstrate adherence, determine the biological contribution of each prior dose, or establish causality.

Do not equate scheduled and administered. A regimen plan cannot replace a confirmed administration record.
Do not use time alone to break a tied event time. Apply the established event ordering, or flag the relationship if no valid order is available.
Do not treat every pre-dose sample as having no prior dose. In a repeated regimen, it commonly follows an earlier administration.
Do not report more timing precision than the records support. Date-only or imprecise times may prevent both a unique attribution and an exact elapsed-time calculation.
Do not force ordinary predose/postdose labels onto an ongoing infusion. Preserve the during-infusion relationship for appropriate handling.

Key takeaway: Determine predose status relative to a target dose, and preceding-dose attribution from confirmed actual administrations, as separate decisions. Select the latest prior administration under the complete event ordering; when the history or ordering is insufficient, flag the case for review.

Check Your Understanding

Answer each question using the event history and ordering information given. Solutions follow the questions.

A confirmed dose occurs at 10:00, and a sample occurs at 09:50 the next day before a confirmed 10:00 dose. Which dose precedes the sample, and what is the elapsed time?
Two confirmed administrations share a recorded time. The established event ordering places dose C before dose D, and a sample occurs later. Which administration is latest before the sample?
A dose is planned for 08:00 but documented as held. Can it be selected as the preceding actual administration for a sample at 09:00?
A sample time is recorded only as “afternoon,” while a confirmed dose on the same date is recorded at 13:00. What should be concluded about the sample’s relationship to that dose?

Solutions. (1) The preceding dose is the confirmed 10:00 administration on the prior day. The elapsed time is $23$ h $50$ min, or $23 + 50/60 = 23.83$ h. The sample is predose relative to the next day’s 10:00 dose. (2) Dose D is latest under the established ordering, provided the records are valid and the order applies; investigate the unusual tied-time history if warranted. (3) No. A held dose is not an administration; select the latest confirmed actual administration before the sample, or report that none is established. (4) The record does not establish whether the sample preceded or followed the 13:00 dose. Do not assign predose status or a unique preceding dose based on “afternoon” alone; flag the timing relationship for review.