Ask any lab scientist which sample problem they see most, and “haemolysed” will be near the top. The authors of an RCPAQAP survey of 68 Australian laboratories describe haemolysis as the commonest ongoing pre-analytical problem. Here’s what it is, why it matters, and what labs do about it.
What is haemolysis?
Haemolysis is the breakdown of red blood cells, which releases haemoglobin and everything else inside the cell into the surrounding serum or plasma. Once enough haemoglobin is free, the serum turns pink or red. That happens from roughly 0.3–0.5 g/L of free haemoglobin, depending on the study you read. Interference starts below the level you can see, though, so “it looked fine to me” is not a quality check.
In vitro vs in vivo
In vitro haemolysis happens outside the body, during collection, transport or processing. It’s an artefact. The patient’s blood is fine, but the sample isn’t. This is by far the most common type.
In vivo haemolysis happens inside the patient, for example in haemolytic anaemias or transfusion reactions. One frequently quoted study put it at about 3% of haemolysed samples (Carraro et al, cited by Koseoglu et al). Heireman and colleagues give a figure of under 2%. Here the free haemoglobin is genuinely in the circulation, so recollecting won’t fix it, and the finding may be clinically important.
How do you tell the two apart? Clues to in vivo haemolysis include haemolysis that turns up in repeat, carefully collected samples, low haptoglobin, raised unconjugated bilirubin and a raised reticulocyte count.
Why it pushes potassium up (and what else it affects)
Red cells are full of things plasma has much less of. Roche’s ISE method sheet notes red cells contain about 25 times more potassium than plasma. Rupture enough of them and the measured potassium climbs while the patient’s true potassium hasn’t changed. This is pseudohyperkalaemia, the haemolysis effect with the highest stakes. A falsely high result can trigger unnecessary treatment, and a haemolysed “normal” can hide real hypokalaemia.
It interferes in several ways (Farrell & Carter):
- Release of cell contents: raises potassium, LD, AST and phosphate.
- Spectral interference: haemoglobin absorbs strongly at certain wavelengths, adding to (or confusing) photometric readings.
- Chemical interference: haemoglobin can take part in or inhibit reactions. Falsely low bilirubin has been reported with a diazo method, for example.
- Dilution: in gross haemolysis, released fluid can slightly lower some analytes.
How sensitive is each test? In one study on an Olympus analyser, LD and AST were affected even at haemolysis too slight to see (free haemoglobin under 0.5 g/L). Potassium and bilirubin changes became clinically meaningful above about 1 g/L. At up to 4.5 g/L, potassium rose by an average of 1.48 mmol/L. These numbers are method- and analyser-specific.
HIL indices: how analysers “see” haemolysis
Modern analysers estimate the haemolysis (H), icterus (I) and lipaemia (L) indices by measuring the sample’s absorbance at several wavelengths. The EFLM Working Group for the Preanalytical Phase strongly discourages visual assessment, and recommends automated H-index measurement on all serum and plasma samples. CLSI C56 gives guidance on using HIL indices as indicators of interference.
There’s a catch: different analysers report the H-index in different units. Understanding of how H-index units relate to haemoglobin concentration has been poor, and labs have varied widely in how they act on them. Converting instrument units to g/L of free haemoglobin is recommended to help harmonise practice.
Comment, suppress or recollect?
Labs set analyte-specific H-index thresholds from manufacturer data or local studies, and review them when the platform changes. The European recommendations above describe a tiered approach:
- No meaningful bias: release the result.
- Bias that matters analytically but not clinically: release with a comment stating the direction of the likely error, for example “Value possibly increased by haemolysis. Consider recollecting another sample.”
- Bias beyond what’s clinically acceptable: suppress that result, explain why, and request a new sample.
- Gross haemolysis (≥10 g/L free haemoglobin): suppress all clinical chemistry results and request recollection.
The same recommendations also strongly discourage “correcting” results with formulas based on the H-index, because the amount released varies too much between patients. Australia has its own AACB/RCPA-endorsed consensus statement (Badrick et al, 2016). It calls for quantitative H-index measurement, documented assay-specific thresholds and clear report comments.
Potassium shows why the middle tier exists. As the recommendations point out, failing to report a genuinely high or low potassium from a mildly haemolysed sample may be worse than releasing it with a comment.
Common causes
Most in vitro haemolysis comes from how blood is collected and handled:
- Collecting through intravenous cannulas. A CDC Laboratory Medicine Best Practices meta-analysis found straight-needle venipuncture produced much less haemolysis in emergency departments than drawing from IV starts.
- Forcing blood through a needle, for example pushing a syringe sample into a tube. The Olympus study above used exactly this to create haemolysed samples.
- Difficult or traumatic collections, small-gauge needles, and vigorous shaking rather than gentle inversion.
- Transport and processing. A meta-analysis found pneumatic tube systems cause small average rises in potassium (about 0.04 mmol/L) and LD (about 10 U/L), with speed and distance playing a part. Each lab has to validate its own system.
One lookalike to remember: refrigerating unspun whole blood can raise potassium without any haemolysis at all. One study measured a mean rise of about 1.0 mmol/L after 6 hours at 4 °C. A normal H-index doesn’t rule out pseudohyperkalaemia.
Key takeaways
- Haemolysis is mostly an in vitro artefact. In vivo haemolysis is uncommon but real.
- It raises potassium, LD, AST and phosphate, and interferes spectrally and chemically.
- Use the H-index, not your eyes. Thresholds are analyte- and platform-specific.
- Release, comment or suppress based on clinical impact. Don’t “correct” with a formula.