Autosampler qualification: the tests to run on a new autosampler

Autosampler qualification is the documented proof that an injector does what your methods need. For a new HPLC autosampler, check the delivery and installation (IQ), then run an operational qualification (OQ) covering injection precision (peak area RSD of six injections), injection volume accuracy and linearity, carryover and, if the tray is cooled, sample temperature accuracy. One public set of typical limits, from the European OMCL network's guideline, is RSD ≤1.0%, carryover ≤0.2% and tray temperature within ±3 °C.

Key points

• USP <1058> in its 2017 version is still the official chapter as far as we could verify in October 2026. A lifecycle revision, retitled Analytical Instrument and System Qualification, is in progress.
• The core autosampler OQ tests are area precision, volume accuracy and linearity, carryover and tray temperature. One injection sequence covers precision and carryover.
• Typical OQ limits are far looser than manufacturer specifications (1.0% vs 0.25% RSD, 0.2% vs 0.002% carryover in our example). Set your own limits from what your methods need, for example the USP <621> repeatability limit of 0.73% for five injections at B = 2.0.
• Weighing is the most direct volume test, but only with an evaporation control and the water density at the measured temperature.
• Requalify at a documented interval and after moves, major repairs and failure trends. System suitability does not replace the OQ.

What USP <1058> requires, and where its revision stands in 2026

USP general chapter <1058> Analytical Instrument Qualification is the reference most regulated labs use. The chapter in force became official on 1 August 2017 (USP-NF). It is an informational chapter: it gives a risk-based framework, not fixed test limits.

Two ideas from it shape every autosampler protocol: the 4Qs model (DQ, IQ, OQ, PQ), and risk-based grouping into Group A apparatus, Group B instruments and Group C computerised systems. An HPLC system with a chromatography data system is the usual Group C example (PQMS), so the autosampler gets the full treatment.

The revision is still moving. USP published a restructured draft in Pharmacopeial Forum 51(2) in 2025, retitled Analytical Instrument and System Qualification, with comments due by 31 May 2025 (ECA). It keeps the 4Qs and the groups but places them inside a three-stage life cycle ending in ongoing performance verification (McDowall, Spectroscopy). The draft drew nearly 300 comments; after a workshop on 23-25 June 2026, USP planned to publish an enhanced version in PF 52(6) or PF 53(1) (USP workshop agenda). Check USP-NF Online before you write a new protocol.

Europe's Official Medicines Control Laboratories (the OMCL network coordinated by the EDQM) follow their own public guideline, which other labs may use on a voluntary basis. It uses four levels (selection, installation and release, periodic and motivated checks, in-use checks) and avoids the DQ/IQ/OQ/PQ terms because labs used them inconsistently (OMCL core document). Level III corresponds roughly to an OQ, and the liquid chromatography annex (R7, in force since 15 November 2023) lists typical acceptance limits with example test procedures (OMCL Annex 1).

DQ, IQ, OQ and PQ applied to an autosampler

The question for each stage is quoted from Burgess and Jones (1998).

Which tests belong in an autosampler OQ

The table sets typical OQ limits from the OMCL annex next to one manufacturer's specification, the Thermo Scientific Vanquish Split Sampler C/CT, to show the gap.

OMCL marks linearity and accuracy as needed only when a method injects different volumes in the same analysis. A 2008 LCGC series on HPLC performance qualification calls a carryover limit of ≤0.1% common and expects under 0.05% "from a well designed autosampler in good repair" (Dowling et al., LCGC).

The gap is deliberate. A generic limit has to fit every LC system in a network, so passing at 1.0% RSD shows that the injector is not broken, not that it is good enough for a method with a 0.73% limit. OMCL says other justified approaches are acceptable, so a lab that runs assays can justify a tighter internal limit.

How to run the injection precision test

OMCL Method 2 is a good default because it also gives you carryover. Use 50 µg/mL caffeine in water as reference solution (b) and 0.5 mg/mL as reference solution (a), on an RP-18 column with water/acetonitrile 85:15 at 1.0 mL/min, 40 °C and 273 nm. Inject 20 µL on HPLC or 10 µL on UHPLC with a standard flow cell. The sequence is:

1. Blank (mobile phase)
2. Six injections of reference solution (b)
3. One injection of reference solution (a)
4. Blank 2
5. One injection of reference solution (b)

Fill the vial well above the needle depth, use a fresh septum and let the baseline settle before the first injection. If the first injection of the day is always low, that is a finding to fix (often air in the metering syringe), not a reason to discard it. See the autosampler troubleshooting guide for the usual causes.

Worked example: area RSD from six injections

mean = sum(A_i) / n
SD   = sqrt( sum((A_i - mean)^2) / (n - 1) )
RSD% = 100 × SD / mean

The mean is 152 418. The squared deviations sum to 726 083, so SD = √(726 083 / 5) = 381 and RSD = 100 × 381 / 152 418 = 0.25%. That passes the 1.0% limit with room to spare.

Now replace injection 6 with one short injection of 150 480 (about 1.2% below the other five, as from a small air bubble). The RSD rises to 0.56%. That still passes the OQ, but it uses two thirds of a 0.85% system suitability budget, which is why a single bad injection deserves a look even when the test passes.

How to measure injection volume accuracy by weighing

OMCL infers accuracy from the linearity run: inject five volumes from one vial, fit area against volume and divide the intercept by the slope. With 1-5 µL injections giving a slope of 30 415 area units per µL and an intercept of 2 035, the offset is 2 035 / 30 415 = 0.067 µL (r² = 0.999998), inside the 0.2 µL UHPLC limit. The catch: a constant offset is all this method detects. A metering device that delivers 2% too little at every volume changes the slope, not the intercept, and passes.

Weighing catches proportional errors as well. The method is the same principle OMCL uses for pump flow (V = m/ρ):

1. Fill two identical vials with degassed HPLC water, cap them and let them reach room temperature. Record the water temperature.
2. Weigh both on an analytical balance with 0.01 mg readability.
3. Place the test vial and the control vial in neighbouring positions. Pierce the control septum once so its evaporation path matches.
4. Run N injections from the test vial with no wash steps that draw from it.
5. Reweigh both vials at once.

m_net        = (loss of test vial) - (loss of control vial)
V_total      = m_net / rho(T)
V_injection  = V_total / N
accuracy (%) = 100 × (V_injection - V_set) / V_set

Worked example, 20 injections of 50 µL at 22.0 °C, where water has a density of 0.997773 g/mL (NIST):

Three details decide whether the number means anything:

• Evaporation. Without the control vial the result would read 0.12% high. At 20 × 1 µL the same evaporation would be a 6% error, so test small volumes with many more injections.
• Temperature. One degree of error in the density lookup shifts the result by about 0.02%, so the thermometer matters less than the evaporation control.
• What leaves the vial. The balance measures what the needle draws, not what reaches the column. Full-loop and overfill modes draw more than the set volume; see full-loop vs partial-loop injection before you choose the injection mode for this test.

The arithmetic also shows what it takes to verify a tight target. The Elutrolab AS-48 has a design target* of ±0.1 µL volume accuracy after solvent-specific calibration. At 20 × 50 µL that is ±0.2%, so the 0.12% evaporation correction above is not optional. Repeat the test with your main sample diluent if it is not water, because viscosity and volatility change how a syringe draws.

Carryover in the same sequence

OMCL compares blank 2 with the reference (b) injection that follows it. The limit is 2% of the (b) area, and because (a) is ten times more concentrated than (b), that equals 0.2% of the high standard. Worked example: (b) gives 152 000, blank 2 shows 1 100 at the caffeine retention time. That is 0.72% of (b), or 0.072% of (a): a pass against OMCL and against the ≤0.1% that LCGC calls common. For where the residue comes from and how to cut it, see autosampler carryover: causes and fixes.

How to test sample temperature accuracy

Only autosamplers with tray cooling or heating need this test. OMCL suggests a setting in the working range (for example 5 °C), waiting for equilibrium, then reading a calibrated thermometer in the tray or a probe in a vial of water, at several positions if the manufacturer describes a mapping. The typical limit is ±3 °C; the 2008 LCGC series measured air with a ±5 °C limit. The water vial tells you what samples actually see, and air recovers faster than liquid after the door opens.

How the OQ connects to USP <621> system suitability

The harmonised <621> Chromatography, official since 1 December 2022, defines system repeatability as the %RSD of at least three injections of a reference solution (USP <621>). For the assay of an active substance or excipient with a 100% target, where the monograph sets no repeatability limit, the maximum is:

%RSDmax = K × B × sqrt(n) / t(90%, n-1)
K = 0.349
B = upper content limit in the monograph minus 100%
n = number of injections (3 to 6)

Worked example: a monograph assay limit of 98.0-102.0% gives B = 2.0. With six injections, t(90%, 5) = 2.015, so %RSDmax = 0.349 × 2.0 × 2.449 / 2.015 = 0.85%. With five injections it is 0.73%. The USP-only text adds that when a monograph states an RSD limit of 2.0% or less, five replicate injections are used, and six when it is more than 2.0%. Limits stated in a monograph take precedence.

An autosampler that passes the OQ at 0.9% would therefore fail most five-injection suitability tests at B = 2.0, where the limit is 0.73%. The reverse also holds: the chapter states that no sample analysis is acceptable unless system suitability has been shown, and the LCGC authors write that "system suitability of a particular method is not a substitute for a PQ".

How often to requalify an autosampler

Neither the OMCL guideline nor the LCGC series sets one calendar interval. OMCL requires Level III checks after installation, a move or significant repair, plus periodic checks at a frequency defined in advance and documented. Many labs choose once a year, alongside preventive maintenance. The LCGC authors tie requalification to events: a move to a new bench triggers a full PQ, shipping to another site counts as a new installation, and a component failure calls for retesting what that component affects.

A reasonable mapping for the autosampler (our judgement, not a standard):

Run the same set on a second-hand unit before it goes into service (see the used autosampler buying guide), and use the same vial and septum type every time so that results stay comparable year to year (vials, caps and septa explains how septa affect evaporation).