Autosampler carryover: causes, how to measure it and how to cut it

Autosampler carryover is analyte from one injection that turns up in the next. You measure it by injecting a blank after your highest standard and dividing the blank's peak area by the standard's. Bioanalytical methods under ICH M10 must keep that blank at or below 20% of the LLOQ response, which for a 1000-fold calibration range works out at about 0.02% carryover. Most carryover comes from the needle, the needle seat or seal and the injection valve rotor seal, and you cut it with a wash solvent that actually dissolves the analyte, enough rinse cycles, a sensible injection order and timely seal replacement.

Key points

• Carryover (%) = (first blank area after the high standard, minus any area in a blank run before it) ÷ high standard area × 100.
• The limit that matters scales with your calibration range: 20% of the LLOQ response is 0.2% carryover for a 100-fold range but 0.002% for a 10,000-fold range.
• Three blanks in a row tell you the mechanism: a fast drop means residual liquid, a slow drop means adsorption, a flat line means contamination rather than carryover.
• Fix the chemistry first (wash solvent strength and pH), then the hardware (needle seal, needle, rotor seal), then the sequence (order and blanks).

What is autosampler carryover?

ICH M10 defines carry-over as "an alteration of a measured concentration due to residual analyte from a preceding sample that remains in the analytical instrument". You see it as a peak at the analyte's retention time in a blank injected after a concentrated sample.

John Dolan's LCGC column on autosampler carryover separates two kinds, and the difference decides the fix:

• Dilution carryover. A tiny residue of the previous sample stays in the needle, loop or valve and is washed out by the next injection. It falls fast: if the first blank shows 1%, the second shows about 0.01%.
• Adsorptive carryover. The analyte sticks to a surface (needle, rotor seal, loop, column) and bleeds off slowly: 1%, then 0.9%, then 0.8%. Extra blanks barely help.

A third pattern is often mistaken for carryover: contamination, a small constant peak that does not shrink with repeated injections and also appears in blanks run before any standard. Dolan's illustration of why carryover matters: in a tablet assay with a ±2% specification, a 101.5% tablet injected after a 100% standard with 1% carryover reads 102.5% and fails.

How to calculate carryover

Carryover (%)          = (A_blank - A_preblank) / A_high × 100
Blank vs LLOQ (%)      = (A_blank - A_preblank) / A_LLOQ × 100
IS in blank (%)        = A_IS,blank / mean A_IS × 100

Largest carryover that meets a limit L (fraction of LLOQ response),
assuming a linear response:
Carryover_max ≈ L × LLOQ / ULOQ

A_high is the area of the highest standard (the ULOQ, or the highest concentration you expect in real samples), A_blank the area in the first blank after it and A_preblank any area at the same retention time in a blank injected before the standard.

To measure it: equilibrate, inject a blank in the sample solvent or matrix, then the highest standard, then three blanks, all with the same volume, method and integration parameters. Repeat the set a few times, because blank areas near the noise are imprecise.

Worked example: an LC-MS assay checked against ICH M10

Section 3.2.6 of ICH M10 says carry-over in the blank samples following the highest calibration standard "should not be greater than 20% of the analyte response at the LLOQ and 5% of the response for the IS". Suppose a method runs from 1 to 1000 ng/mL:

• Carryover relative to the ULOQ: 980 ÷ 2,480,000 × 100 = 0.040%.
• Blank relative to the LLOQ: 980 ÷ 2,500 × 100 = 39%. This fails the 20% criterion.
• Internal standard: 3,100 ÷ 410,000 × 100 = 0.76%. This passes the 5% criterion.
• Blank 2 is about 32 times smaller than blank 1, so this is dilution carryover.

A carryover of 0.04% sounds small, yet it fails, because this range allows no more than 0.20 × 1/1000 = 0.02%. The wash has to remove at least half of the residue left now. Because the second blank is already clean, a blank after each high sample would also pass, and M10 accepts that as a validated mitigation, but it costs instrument time on every run.

For scale, Agilent's user information for its Multisampler states that a carryover of 0.001% "should be achievable" even with a triple quadrupole detector. By the M10 arithmetic that covers about a 20,000-fold range, with the wash set up well.

What limit to use outside bioanalysis

For assay, impurity, cleaning or reaction-monitoring methods, set the limit with the same arithmetic: the carried-over amount should be a small fraction of the smallest result you report. If the highest standard is 50 µg/mL, you report down to 0.05 µg/mL and you accept up to 10% of that from carryover, the limit is 0.10 × 0.05 ÷ 50 = 0.01%. As a rule of thumb, Dolan writes that most workers would find carryover greater than 0.05-0.1% unacceptable.

Where does carryover come from?

Shimadzu's carryover guide says to replace the high-pressure valve's needle seal first because "it is the culprit in many instances", and warns that PEEK fittings can slip above about 4000 psi. A Thermo Fisher poster on carryover and contamination adds poor drainage of the injector valve waste line, a worn syringe plunger and needles that go too deep into the sample. Agilent advises that a wash vial should have no septum, while Dolan notes that a silicone or PTFE-faced silicone septum wipes the needle like a squeegee and foil or PTFE-film septa wipe less well (see HPLC vials, caps and septa). For where each part sits, see how an HPLC autosampler works and the 6-port injection valve explained.

How to find which part is causing carryover

Work from the cheapest test to the most expensive one.

1. Run the diagnostic sequence: pre-blank, high standard, three blanks, then one run with no injection at all (Shimadzu calls it a null injection).
2. Read the pattern using the table below.
3. Vary the blank volume. Shimadzu's reading: if the carryover peak grows with a larger blank volume, the blank is contaminated; if it stays the same, the residue is most likely on the outside of the needle; if it falls, look inside the needle and loop.
4. Take the column out of the question. Replace it with a zero-dead-volume union and repeat the standard and blanks. If carryover remains, it is in the hardware. With the column back, run the gradient again without injecting: a peak in this second gradient means the column bed or inlet frit is holding the analyte.
5. Swap parts in order of cost and likelihood: needle seal or seat, then needle and loop, then rotor seal. If many unrelated compounds carry over, suspect a physical fault rather than chemistry.

How to choose a needle wash solvent

The wash only works if it dissolves the analyte better than the sample solvent does. Principles from Dolan and Shimadzu:

• Strength. For reversed-phase methods use a wash at least as strong as the mobile phase at elution, often 80-100% methanol or acetonitrile. Shimadzu calls isopropanol an excellent wash for many applications, fatty acids among them.
• pH. Choose a pH that keeps the analyte soluble. A volatile acid or base (formic acid, acetic acid, ammonium formate, ammonium acetate, ammonium hydroxide) at 0.1-1% by volume is usually enough; it does not need to be a true buffer.
• No non-volatile salts. Shimadzu says never to use phosphate buffers in the wash, and Agilent warns that salt left on the needle and seat leads to blockages.
• Freshness. Change the wash solvent regularly (Dolan suggests weekly), and purge the wash lines so they hold no air.

Multi-rinse systems often pair a strong wash with a weaker one close to the starting mobile phase, so the next sample does not go in with a slug of strong solvent that distorts early peaks (see how much you can inject). Check that every wash is miscible with the mobile phase and the sample solvent.

Wash modes, volumes and rinse cycles

External wash: dip or flowing rinse

Dolan's column reproduces Shimadzu data for chlorhexidine:

How long the needle stayed in the dip did not matter; fresh, flowing solvent did. Agilent's flush port, for example, holds 0.68 mL and is refilled by a 6 mL/min pump in 7 s, and Agilent suggests 2-3 s of washing for routine work and 10-20 s for complete washing.

Internal rinse: several small rinses beat one large one

The inside of the needle and sample line is cleaned by pulling wash solvent through it. If each rinse perfectly mixes a residual volume V_r with a rinse volume V_w, the fraction of residue left after n rinses is:

f = (V_r / (V_r + V_w))^n

With 2 µL of residue and 120 µL of rinse solvent in total:

• one 120 µL rinse: f = 2 ÷ 122 = 1.6% of the residue left
• three 40 µL rinses: f = (2 ÷ 42)³ = 0.011% of the residue left

Real needles are not perfectly mixed and adsorbed analyte does not follow a dilution law, so measured gains are far smaller than this idealised model, but the direction holds: for the same solvent volume, more cycles clean better than one big rinse. For difficult compounds, rinse both before and after drawing the sample if your autosampler offers it.

Why the injector design changes the wash strategy

In flow-through-needle designs the needle sits in the high-pressure path, and Agilent notes that the flow path, including the inside of the needle, is continuously flushed in normal operation. In fixed-loop designs with a low-pressure needle, such as the Elutrolab AS-48, the mobile phase backflushes the loop into the column but does not flush the needle and sample line, so their cleanliness depends on the external dip and the internal rinse cycles (0-3 on the AS-48). Dolan's advice to leave the valve in the inject position for the whole run, for maximum flushing of the loop, applies here too.

For the rotor seal, Agilent suggests switching the injection valve at high organic after the last peak, then switching back once starting conditions have stabilised, so the bypass groove holds starting mobile phase; this matters most below 0.5 mL/min.

Injection order and blanks

• Order samples from low to high concentration where the study design allows, as Dolan suggests.
• Put a blank after any sample you expect to be high, before the next low one. ICH M10 lists this as a mitigation and says that if carry-over is unavoidable, study samples should not be randomised.
• Run a blank after the highest calibrator in every run. M10 asks that carry-over during study sample analysis be assessed and reported, and this blank gives you the number.
• Re-inject the sample that follows an over-curve sample, after a blank, rather than trusting its first result.

Materials that reduce adsorption

If you change a wetted material to solve carryover, re-run the diagnostic sequence and file the result with your autosampler qualification records, so the next person knows why the part is different.