Resolve Dose-Sample Ties With Event Ordering

Why an Identical Timestamp Is Not a Complete Sequence

A dose and a pharmacokinetic sample can have the same recorded timestamp even when one occurred before the other. This can happen when records retain only minute-level or other rounded clock times, or when a study system assigns a common visit time to events that occurred in sequence. The time value alone then does not say whether the sample reflects drug exposure before the dose or after it.

That distinction matters in a longitudinal analysis. A predose sample belongs to the exposure history before the new dose; a postdose sample belongs to the history after the dose. If the records are ordered incorrectly, the analysis may associate a concentration with the wrong dose history. Sorting rows by timestamp is therefore not enough when timestamps tie.

There is no universal convention that makes every sample precede every tied dose, or every dose precede every tied sample. The appropriate ordering depends on what happened and what the source records establish. A dataset should preserve that information with an explicit sequence field or event-order code. A consistent sorting rule can then use both timestamp and event order.

Specify the Ordering Before Building the Analysis Records

For a simple tied pair, one transparent convention is to assign an order rank: 1 for an event known to occur first and 2 for the event known to occur second. Within a subject and the relevant record set, sort first by timestamp and then by rank. The rank is meaningful only because it is linked to documented event sequence; the numbers themselves do not establish what happened.

For example, if a sample is documented as predose and ties with a dose, assign the sample the earlier rank and the dose the later rank. The ordering is \((t,1)<(t,2)\), even though both events have timestamp \(t\). If the dose is documented as occurring first, the ranks are reversed. Where a dataset uses event codes rather than numeric ranks, the same principle applies: the code definitions must explicitly identify sequence and be applied consistently.

The rule should be based on the best available evidence, such as collection and administration records or a documented study procedure. A label such as “predose” can support a before-dose ordering if it refers to the actual sample collection relative to the actual administration. A visit label or a timestamp tie alone does not establish that sequence. If evidence is insufficient, retain the tie as unresolved for review rather than silently assigning an order.

Why the Ordering Changes the Dose History

A dose record changes which doses have occurred by the time a sample is collected. The following justification is a bookkeeping argument, not a claim about the size of a concentration change: immediately before a dose event, that dose is not yet part of the administered-dose history; immediately after it, the dose is part of that history. Thus, at a tied timestamp, changing the event order can change the history associated with the sample.

When the Timestamp Cannot Resolve the Tie

If there is no reliable information about which event occurred first, a deterministic processing rule cannot recover the unknown chronology. For instance, assigning every tied sample a lower rank than every tied dose creates an orderly file, but it does not demonstrate that all such samples were predose. The scientifically defensible status is unresolved until additional evidence is available.

Quality Checks, Alternatives, and Limitations

Useful checks focus on both the code definition and the event pairs to which it is applied. Review the data specification and processing logic together: a code name, numeric rank, and sorting direction must all express the same ordering. Then inspect tied dose-sample pairs, checking the source sequence where available and confirming that the final ordered records preserve it.

• Flag every dose-sample pair for the same participant with an identical recorded timestamp, rather than reviewing only examples selected by an existing code.
• Check that documented predose samples sort before the corresponding dose and documented postdose samples sort after it.
• Look for contradictory codes for equivalent documented sequences, including codes that change meaning across sources or processing steps.
• Check unresolved ties separately from resolved ties; do not let a default rank make an unknown sequence appear verified.
• After changing a code mapping, rerun the tie checks and compare the resulting event order with the source records.

A common pitfall is to assume that equal timestamps mean the sample and dose were simultaneous, so their row order is immaterial. The correction is that recorded equality may reflect limited timestamp resolution, while the biological and administrative sequence can still be meaningful. A second pitfall is to assume that sorting by timestamp is sufficient. When timestamps tie, sort order requires additional, scientifically justified sequence information.

The main alternative is to retain a shared timestamp and add a separate event-order field. This keeps the recorded time intact while representing the sequence explicitly. Replacing one event’s timestamp with a slightly earlier or later time can force a sort, but it invents time information and may falsely imply a measured interval. Use such a change only if a documented data convention specifically supports it, and preserve the original timestamp and rationale. If sequence is unknown, flagging the tie for review is preferable to either arbitrary ordering or an invented time.

Event ordering has limits. A correct rank does not recover a missing collection time, establish the actual interval between events, or prove that an event label is accurate. It also does not, by itself, determine how a particular analysis handles a dose and a sample at the same time. The ordering must be interpreted within the documented study records and the analysis method.