HPLC sequences, injections and reinjections: control and data integrity

A sequence stops. On restart, some vials are reinjected and others analyzed for the first time. The final report may look orderly while uncertainty remains about earlier activities. Control means reconstructing all relevant work and justifying the results used, without replacing the original chronology with the latest one.

1. Link sample, preparation, vial and injection

The sample identifies the received or collected material; the preparation records how the analytical solution was made. The vial is the physical container placed in a position; the sequence plans activities; each performed injection generates an event and associated data. These objects are related but not interchangeable.

An autosampler position alone does not identify a sample: that position may be reused. Retain sample, preparation and vial identifiers, sequence and injection links, instrument, method/version, operator and relevant times. If one preparation supplies several vials or a vial is replaced, the relationship must remain explicit.

Relationships should support two directions: from result to signal and source material, and from sample to all relevant injections, including those not reported as the final result. A plausibly named file does not demonstrate completeness.

2. Plan the sequence according to the method

Define standards, controls, blanks and samples, order, planned replicates and applicable criteria using the approved method. Assess vial identities, positions, preparations, solution stability, storage conditions and usable times. This article does not prescribe a universal injection count or a standard sequence suitable for every analysis.

Before starting, verify that the selected configuration matches authorized work. Record sequence changes and relevant reasons; after starting, preserve the relationship between the original plan and actual activities. A scheduled row does not prove an injection occurred.

System suitability controls and instrument troubleshooting remain within PL-01. Here, the concern is traceability of their data, decisions and sequence links: a later successful check does not erase earlier acquisitions.

3. Manage interruptions and distinguish repeats

Record when and where work stopped, which injections were completed, partial or not started, available alerts and the known reason or reason still under investigation. Preserve generated data, including partial signals, with their status. “Instrument error” is insufficient if it does not clarify impact and available evidence.

A reinjection takes another injection from the same solution/preparation; a new preparation repeats preparative operations on material from the original sample; resampling concerns another sample or other batch units under the relevant plan. The hypothesis tested, introduced variability and required records differ. Reprocessing an already acquired signal is a separate activity.

Before restarting, establish purpose, scientific justification, required authorization and method conditions: solution stability, system status, controls and data to assess. A technical software restart is not analytical authorization. If an OOS occurred, reinjection or retesting must form part of the relevant investigation; do not repeat until a passing result appears.

Labels such as “trial”, “test” or “invalidated” do not automatically exclude data from the GMP record. Assess actual purpose, material used and relationship to the activity. Invalidation requires a reasoned conclusion and does not remove the original record. Do not use product samples for exploratory testing intended to predict the outcome to accept.

4. Reconcile the activity ledger

The following teaching ledger uses invented identifiers. It shows only the sample portion: method-required standards, blanks and controls must remain documented in the complete record. Link every row to actual records; do not use it to create missing data retrospectively.

Compare the plan, actual acquisitions, processed results and report. Look for orphan identifiers, unexplained repeats, injections omitted from summaries and results without an associated signal. Total row count is insufficient: two datasets can have equal counts yet refer to different samples.

Use CDS event review to examine changes and interruptions. For different results derived from the same signal, apply reprocessing controls, without confusing them with a new injection.

5. Simulated case: restart without losing history

In this case, P1 has already been acquired, P2 has a partial signal and P3 was not injected. The laboratory does not rename S2 as though it were the only sequence. It assesses the interruption cause and solution stability before deciding whether the proposed restart is justified.

If evidence and the method support reinjecting P2, D03 retains its link to D02; for P3, D04 is the first acquisition. If stability is unproven or impact remains uncertain, the plan changes through the relevant authorized decision. This case does not automatically authorize reuse of vials or earlier results.

The conclusion identifies results used, data excluded from final evaluation with reasons, anomalies and linked investigations. Keep all relevant records, the review outcome and responsibilities. A readable report must allow retrieval of the complete record, not replace it.

6. Sources and example limitations

Sources checked on 1–2 October 2026. Context: GMP laboratory for human medicinal products. No company method is available; the ledger is an original example, not an approved reinjection protocol.

• European Commission — EU GMP Chapter 6, revision 28 March 2014, in operation from 1 October 2014, §§6.7–10 and 6.15–17: documentation, method and test records.
• FDA — Data Integrity and Compliance With Drug CGMP, final December 2018, Q12–14: generated data, trials and retention of processing records.
• FDA — Investigating OOS Test Results, revision 1, final May 2022, §§III.B, IV.B and V: hypotheses, retesting, resampling and conclusions within CDER chemistry-testing scope. FDA guidance is nonbinding and must be read alongside applicable rules.