Trial injections: why "just checking the system" is a major GMP observation

Every analyst knows the feeling. The column has just gone on, or the instrument sat idle over the weekend, and you are not sure the first result will be good. So you put one sample in, “just to check”, before the real sequence.

In a research lab that is ordinary practice. In a GMP lab it is a major observation, and regulators have spent years explaining why.

What the regulators say

Health Canada’s risk classification guide lists two related observations under the quality control department, both rated major (Risk 2) (GUI-0023, Appendix A):

• “The practice of performing of trial sample injections was identified.”
• “The use of trial blank or standard injections was not governed by a written procedure.”

Read them together. The first concerns samples. The second accepts that blanks and standards may be injected before a run, and asks for a written procedure.

FDA’s data integrity guidance makes the same distinction. It calls the use of an actual sample “in test, prep, or equilibration runs as a means of disguising testing into compliance” a violative practice. It adds that system suitability should use a standard preparation, and that if a sample is used for suitability, it should be “a properly characterized secondary standard”, under written procedures, from a different batch than the samples being tested (FDA, 2018, Q13).

What a trial injection looks like to an inspector

An April 2026 FDA warning letter shows the pattern in detail (Warning Letter 721180, Ava Inc.). Product names are redacted in the letter, so I describe only what it states.

Investigators found HPLC and GC vial positions that analysts described as designated for trial injections. The chromatograms from them were not saved in the analytical system or in the batch record.

In one HPLC case, a trial injection met the specification, the repeat gave a better result, and the more favourable result was reported. In one GC case, a trial injection was out of specification, a second trial was also out, and a repeat came in. The passing result was reported, and the out-of-specification results were not reported or investigated.

The firm’s explanation was that the trial injections verified the condition of the system. FDA did not accept it. The letter states that “the injection of trial samples is not acceptable, in part, because all data from analysis of product samples must be retained and reviewed”, and that the firm had not shown how its system suitability would establish readiness instead.

One detail in the letter is worth noticing. The firm’s own GC procedure required a trial injection “to ensure the system is equilibrated”, and stated it “is not processed as a part of raw data.” FDA wrote that it was “particularly concerning” that the practice was in a written procedure, because the procedure directed staff away from CGMP. A method SOP written years ago can carry a data integrity problem into every run that follows it.

The remediation FDA requested was large: an independent review of all trial injections over the last five years, investigation of every out-of-specification result from them, and a GMP consultant.

Why the need behind it is real

I do not think most trial injections start as fraud. They start with a real chromatography problem: the system is not ready.

LC and LC-MS systems do need conditioning. In one published plasma LC-MS study, a single blank injection disturbed retention time, peak width and peak area, and the system needed 5 to 6 injections to recover; fewer than 3 was not enough (Martínez-Sena et al. 2019, p. 6). The effect depended on the matrix and the analyte (p. 8). That is a research study on an untargeted method, so the numbers will not transfer to your assay. The point does: the first injections after a change can behave differently, and analysts know it.

The fault is in how that need gets met: with a product sample, outside the record, with the result seen before the official run.

How to do it properly

The fix is to make conditioning and readiness visible. In practice:

1. Write the conditioning step into the method or an SOP: what is injected (mobile phase blank, standard, or a defined conditioning solution), how many injections, and when it is required. This answers GUI-0023’s second observation directly.
2. Inject it inside the sequence, so every conditioning injection is in the data system and the audit trail.
3. Use system suitability on a standard to show the system is ready, against the approved criteria. If it fails, that is a finding to investigate. I have written about reading an SST failure.
4. Never inject a product sample outside the official run. If a sample-like material is truly needed for suitability, follow FDA’s conditions: a characterized secondary standard, from a different batch, under a written procedure.
5. Review the audit trail, including injections that never reached a report. In the warning letter, the investigators found the problem in chromatograms that were not part of any official record.
6. Read your old SOPs for wording like “for information only” or “not part of raw data”. If a procedure tells analysts to make injections that are not reviewed, the procedure is the first thing to fix.

Sources

• Health Canada. Risk classification guide for drug good manufacturing practices observations (GUI-0023), Appendix A, quality control department (C.02.013–C.02.015).
• US FDA. Data integrity and compliance with drug CGMP: questions and answers, December 2018, question 13.
• US FDA. Warning Letter 721180 to Ava Inc., 14 April 2026.
• Martínez-Sena T et al. Monitoring of system conditioning after blank injections in untargeted UPLC-MS metabolomic analysis. Sci Rep 2019; 9:9822. doi:10.1038/s41598-019-46371-w.

If your methods still carry a “trial injection” step, or the team is not sure what the conditioning step should be, send me the method type on LinkedIn and I will tell you how I would write it. The validation-readiness checklist covers the other method questions.