Sources of error in blood gas analysis

Essentially, this chapter is a short list of reasons as to why your ABG (or venous biochemistry)  measurement may be wrong, i.e. not reflective of what is happening in the patient. In 99% of cases, it is a problem with the collection storage and transport of the sample, because these are factors which are subject to human input and thus human error. The self-calibrating blood gas analyser is a dutiful and dependable servant; some sort of failure in its internal workings will only rarely contribute to the error (and usually it will be because some idiot human has improperly calibrated it). Lastly, a tiny fraction of errors are begot by the physicochemical limitations of the measurement method (eg. when an absurdly high bromine concentration interferes with chloride measurement and returns a spuriously raised chloride concentration value). This whole thing enjoys a rich full discussion in the section dedicated to arterial blood gas analysis.

But is it relevant for the exam, asks the tired exam candidate, ready to quit at 2am? Unfortunately, with the examiners deciding not to repeat or repurpose previous questions, the  established pattern of ever-increasing esotericism in the written papers suggests that yes, this kind of weirdness will appear, and in fact we are already seeing signs of it in historical SAQs:

  • Question 18 from the second paper of 2022, where the college asked what one might expect from ABG samples taken following a substantial delay, or those taken from extremely hypothermic patients.
  • Question 3.4 from the first paper of 2021, "leukocyte larseny"
  • Question 20.2 from the second paper of 2017, where the lactate electrode was fooled by the ethylene glycol in the sample
  • Question 3.2 from the second paper of 2015, where there was a delay in the processing of a venous sample

Sources of error due to inappropriate sample collection and handling

Source of error in blood gas collection and handling

  • Sample contaminated with bubbles
    • Published information on the errors of ABG sampling arising from the presence of air bubbles comes from Biswas et al (1982). After two minutes, the presence of air bubbles or froth in the syringe had resulted in a significant increase of PaO2 and a decrease in PaCO2.
  • Sample contaminated with venous blood
  • Sample clotted
  • Sample contains too much heparin (liquid heparin dilutes the sample, and causes pH changes)
  • Haemolysis en route to the ABG analyser
  • Inappropriate type of syringe used (gas-tight syringes are needed, rather than evacuated tubes)
  • The sample contains an excessive number of leukocytes, and they have consumed all the oxygen (eg. leukaemia blasts)
  • The sample took too long to transport, and blood cell metabolism has changed the gas concentration (see the section on delayed processing below)
  • The sample was transported to the lab by a pressurised pneumatic system - this has the tendency to amplify errors caused by gas bubbles (Woolley and Hickling, 2003).
  • Sample was chilled with ice (and then analysed at body temperature values)
    • Low temperatures also make the ABG syringe polymer more gas-permeable, allowing gases to exchange with the atmosphere.

Sources of error in venous biochemistry sample collection and handling

  • Clotted sample
  • Haemolysed by inappropriately small needle
  • Haemolysed by syringe vacuum
  • Inappropriate collection tube

The consequences of a delay in sample processing

Delayed venous sample processing

Question 3.2 from the second paper of 2015 presents the candidates with a venous blood sample which demonstrates characteristic features of a delay in processing. A good reference is a 2008 paper by Tanner et al, examining the delayed processing of samples collected in rural and remote areas. The studies have discovered that over 4-24 hours of storage various changes take place. These changes (and the reasons behind them) were as follows:

Escape of cellular contents due to haemolysis

  • Potassium increases. After 24 hours, Tanner et al found the original K+ level of 3.8 increased to 8.0.
  • Phosphate increases. In the same study, the PO4- went from an average value of 1.36 to 4.36.
  • Total protein increases
  • LDH increases

Compartment shift

  • Sodium decreases

Metabolism by live cells

  • Acidosis develops (bicarbonate drops)
  • LDH increases (also because of anaerobic metabolism)
  • Glucose decreases (consumed by RBCs)
  • Lactate increases (produced by RBCs)

Consequences of delay in arterial blood gas processing

An ideal reference for this exists in the Journal of Critical Care (Woolley and Hickling, 2003). It is unfortunately not available as free full text, but the author was able to get access to the article anyway, and summarise its contents.

  • Some of the changes listed above apply:
    • pH decreases
    • HCO3- decreases
    • SBE decreases
    • Lactate increases
    • Glucose decreases
  • PaO2 decreases and PaCO2 increases as a consequence of RBC metabolism (in the experiments by Biswas et al (1982), the PaO2 fell by up to 40% in samples which were stored at room temperature for twenty minutes. This ameliorated by storing the sample at 4°C, but all sorts of other problems develop as a consequence of this).

The actual difference in reality is not that great.  It is possible to demonstrate this directly, in the following scenario. The two gases are arterial, taken at 35% FiO2 from an ICU patient; the sample was analysed once, then left next to the ABG machine for a half-hour or so, and then accidentally analysed again by someone who thought it was fresh.

Same blood gas analysed twice

The sample on the left was first, the one on the right was the thirty minute delay. The ambient temperature was probably close to 20 degrees. As one can see, the differences are trivial. During that half hour, this anaemic patient's few remaining RBCs managed to metabolise only 0.1mmol/L of glucose, and produced a barely measurable increase in lactate and CO2.

ABG interpretation in the severely hypothermic patient

This is not really a source of error per se, but a need to adjust one's one mind to take into account the changes occurring in hypothermia, which then influence the expectations one has from the gas. Consider:

  • The lower the temperature, the higher the gas solubility.
  • The higher the solubility, the lower the partial pressure.
  • Thus:
    • PaO2 drops by 5mmHg for every degree below 37°C.
    • PaCOdrops by 2mmHg for every degree below 37°C.
  • pH increases by 0.015 for every degree below 37°C because temperature affects the ionisation constant of water

Additionally, one needs to decide whether one uses the alpha-stat or the pH-stat models of ABG interpretation, which, in a (very crude) summary, are:

  • Alpha-stat: you warm or correct all samples to 37°C 
  • pH-stat: you correct all samples to the patient's actual body temperature
  • Of these, alpha-stat is preferred by cardiac anaesthetists for mild hypothermia, and pH-stat for deep hypothermic circulatory arrest so that the extra CO2 helps cerebral blood flow (according to Baraka, 2004)

Blasts in the sample: "leukocyte larceny"

This has come up in the exam only once, in Question 3.4 from the first paper of 2021. The college had asked for an explanation as to why the patient, whose SpO2 was 98%, had an arterial blood gas showing a PaO2 of 45 mmHg and an SaO2 of 67%. Conveniently, just before asking this question, they also revealed that his blood count was 52% blasts. 

This spurious hypoxemia is usually referred to as "leukocyte larceny", on the count of the leukocytes being responsible for misappropriating all the oxygen in the sample during its transport. It is described in some detail by Sacchetti et al (1990) and Fox et al (1979), of whom the latter appears to be responsible for coining the expression. According to the data collected by Fox et al, the rate of metabolism can be very rapid. For blood samples with the highest WCC counts, the investigators observed a PaO2 drop of up to 72 mmHg. 

Instrument errors:

  • Inappropriate electrolyte temperature
  • Poorly calibrated ABG analyser
  • Inadequate quality control and maintenance of the electrodes (they have a measurable lifespan; for instance the Clark electrode needs to be changed every 3 years as its Teflon membrane becomes coated with proteinaceous filth).
  • Poorly compensated Clark electrode nonlinearity.

Limitations of measurement technique

pH measurement

Heparin of any sort: Given its extreme anionic charge and near-total dissociation at physiologic pH, it would be surprising if heparin failed to interfere in some way with the strong ion difference in a blood gas sample. Indeed, it would appear that even a "thin film" of it skews the results in the direction of a metabolic acidosis. The quoted study is from 1985, a grim and primitive age when intensive care staff would have to manually heparinise the ABG syringe prior to collecting the sample, sending it to a distant analyser with a porter.

Sodium measurement

​Lithium heparin: Cations other than sodium could potentially exist in the sample in such concentrations as to fool the supposedly sodium-selective ceramic electrode membrane. Of these the most likely culprit is lithium from lithium heparin. These days the arterial blood gas syringe typically contains a very small amount of dry heparin, about 20-25 units per ml.

Chloride measurement

Halogen ions: the measured Cl- increases in presence of halogen ions, such as bromine and iodine. This is a selectivity failure of the supposedly selective chloride-sensitive electrode membrane, which will accept any halogen if their concentration is high enough. 

Salicylates: in massive salicylate overdose there is enough salicylic acid around to compete with chloride for the ionophore in the chloride-selective electrode membrane. In other words, salicylate will infiltrate your chloride-selective electrode, and produce a spurious hyperchloraemia. This effect obviously scales with salicylate levels; for a sense of scale, salicylate levels of 600 mg/L  produce a 15% increase in the ABG chloride measurement.

These chloride-electrode-fooling shenanigans can manifest in the form of a negative anion gap. We are fortunate that, even as they scrape the bottom of the physiology barrel to find new questions, CICM examiners have never resorted to this weirdness for any past paper SAQs. Negative anion gaps has certainly appeared (for example Question 29.1 from the first paper of 2021, and Question 3.1 from the second paper of 2013) but they have been paltry (around -6) and each time associated with lithium toxicity, unconsciousness and seizures. Bromine intoxication could be equally likely, as a reader had pointed out (thanks Stuart), but in these scenarios the chloride level ends up being absurdly high. How high? Ng et al (1992) reported a case of brominated dextromethorphan overdose which presented with a reported chloride level of 261.

Haematocrit measurement

Haemolysis: because the ABG machine shreds the red blood cells with 30Hz ultrasound in order to get to their haemoglobin, it will not notice if there is already free haemoglobin in the bloodstream (eg. due to haemolysis).  

Carboxyhaemoglobin measurement

 Foetal haemoglobin and carboxyhaemoglobin have a very similar absorptivity spectrum. Thus, foetal haemoglobin interferes with the spectrophotometric measurement of carboxyhaemoglobin and falsely increases the measured FCOHb. Conversely, the presence of carbon monoxide poisoning in a pregnant woman might mask a foeto-maternal haemorrhage.

Lactate measurement

The amperometric measurement of lactate using the lactate-sensitive electrode relies on the use of lactate oxidase. This enzyme catalyses the reaction which converts lactate into pyruvate, producing hydrogen peroxide which is reduced at the measurement cathode. Glycolic acid, the metabolic byproduct of ethylene glycol metabolism, also acts as the substrate for this enzyme. Therefore, in ethylene glycol poisoning the lactate measurement by the blood gas analyser will be spuriously elevated. The formal lactate measurment by use of a lactate dehydrogenase enzyme assay will still yield a correct result. The difference between the "formal" and the ABG lactate is described as the "lactate gap", and is a well known phenomenon of ethylene glycol toxicity. This source of ABG error is made famous by its use in a classic ABG question (Question 1.12) from the 2003 edition of Data Interpretation in Intensive Care Medicine by Bala Venkatesh et al. It had also made an appearance in Question 20.2 from the second paper of 2017. Interestingly, the Reference Manual for the local ABG analyser lists a large number of molecules which can potentialy cause interference with lactate measurement- notably ascorbic acid, bilirubin, citrate, EDTA, ethanol, heparin, glucose, paracetamol, salicylate and urea.

Methaemoglobin measurement

Hyperlipidaemia: which in general tends to produce unusual reading in the spectrophotometer, giving rise to abnormally high total haemoglobin readings and negative carboxyhaemoglobin numbers. These extra fatty acids tend to scatter the light, interfering with the absorption spectrum and giving the impression that a lot more light has been absorbed. This can be seen in central venous gases taken accidentally from the TPN lumen. The elevation is typically modest - the author's own recent experience with a pancreatitis patient yielded a serum triglyceride level of 39.9, and a paltry FMetHb of 2.4%. One would obviously need a truly monstrous hyperlipidaemia.

Isosulfant blue: otherwise known as "patent blue" - a dye used in surgery to identify lymphatic vessels, eg. for sentinel node biopsy- causes the same sort of problem: it is blue enough to interfere with spectrophotometry.

Influence of albumin on acid-base interpretation

Question 9.3 from the first paper of 2021 asked about the ways in which albumin influences acid-base analysis. The specific question was about hypoalbuminaemia:  "how does this finding alter your interpretation of acid-base data?" The basic premise is that plasma proteins, among which albumin is the unquestioned king, are largely non-volatile acids, and loss or gain of albumin leads to alkalaemia and acidaemia, respectively. For a more thorough overview of the biochemistry behind this, as well as for experimental data which describes the magnitude of this effect, one is directed to the excellent paper by Rossing et al (1986).

In summary:

Anion gap measurement: Albumin probably contributes about 75% to the total anion gap value, which means hypoalbuminaemia should result in a smaller anion gap. Thus, when analysing an anion gap result, one needs to down-adjust one's expectation of what a normal anion gap should be for the hypoalbuminaemic patient. For this, there are multiple possible equations, but the easiest to remember is the 4:1 rule: for every 4g/L decrease in serum albumin, the normal expected anion gap decreases by 1. Thus, at an albumin of 32, the expected anion gap is 10 instead of 12. 

Delta ratio calculation requires the calculation of the anion gap, and is therefore influenced by hypoalbumnaemia indirectly.

Strong ion gap (SIG) is similar to the anion gap, as it represents the presence of unmeasured anions. It is the difference between the apparent strong ion difference (SIDa) and the effective strong ion difference (SIDe). Albumin plugs into the SIDe equation,  which is is the combination of bicarbonate and Atot (the combination of all weak nonvolatile acids, of which albumin and phosphate are the most important). Thus: 

SIDe = HCO3-+albumin+PO4-

SIDa = ([Na+] + [K+] + [Mg2+] + [Ca2+]) – [Cl-]  


SIDe is normally -40 mEq/L;  a low albumin will decrease the SIDe, and therefore increase the strong ion gap.  However, the ionisation of albumin is different, depending on the pH, and this also needs to be adjusted:

[alb-] = [alb g/l] × (0.123 × pH – 0.631)

Base excess is influenced by albumin concentration, as albumin is an acidic protein. Hypoalbuminaemia therefore has an alkalinising influence, and tends to increase the base excess. In general, every 10 g/L fall in albumin will increase the base excess by 2.5 mEq/L, i.e. the normal base excess for a patient with an albumin of 20g/L should be 5.0 and not 0.  The specific equation used by Story et al (2004) was:

albumin effect (meq litre–1)=0.25×(42–[albumin]g litre–1)

and the main reason it is mentioned here is because one of the et al was Rinaldo Bellomo. 

Base excess gap, another mathematical tool which looks for unmeasured anions or cations,  will therefore also be affected by albumin levels. This gap is calculated using the following equation:

BEG = BEactual - BEcalculated - lactate

where BEcalcuated is a base excess value adjusted by a sum of contributions from sodium chloride free water protein and albumin. Ergo, albumin plays a minor but important role in this gap.


Of particular interest is this pamphlet released by Radiometer, about avoiding preanalytical errors in blood gas testing.

Baird, Geoffrey. "Preanalytical considerations in blood gas analysis." Biochemia medica 23.1 (2013): 19-27.

Hutchison, A. S., et al. "Too much heparin: possible source of error in blood gas analysis." British medical journal (Clinical research ed.) 287.6399 (1983): 1131.

Ordog, Gary J., Jonathan Wasserberger, and S. Balasubramaniam. "Effect of heparin on arterial blood gases." Annals of emergency medicine 14.3 (1985): 233-238.

Ng, Y-Y., et al. "Spurious hyperchloremia and decreased anion gap in a patient with dextromethorphan bromide." American journal of nephrology 12.4 (1992): 268-270.

Biswas, C. K., et al. "Blood gas analysis: effect of air bubbles in syringe and delay in estimation." Br Med J (Clin Res Ed) 284.6320 (1982): 923-927.

Woolley, Andrew, and Keith Hickling. "Errors in measuring blood gases in the intensive care unit: effect of delay in estimation." Journal of critical care 18.1 (2003): 31-37.

Hankinson, S. E., et al. "Effect of transport conditions on the stability of biochemical markers in blood." Clinical Chemistry 35.12 (1989): 2313-2316.

Tanner, Melissa, et al. "Stability of common biochemical analytes in serum gel tubes subjected to various storage temperatures and times pre-centrifugation." Annals of Clinical Biochemistry 45.4 (2008): 375-379.

Marwick, J., R. O. C. Elledge, and A. Burtenshaw. "Ethylene glycol poisoning and the lactate gap." Anaesthesia 67.3 (2012): 299-299.

Jacob, Jeena, and Eric J. Lavonas. "Falsely normal anion gap in severe salicylate poisoning caused by laboratory interference." Annals of emergency medicine 58.3 (2011): 280-281.

Ng, Yee-Yung, et al. "Spurious hyperchloremia and decreased anion gap in a patient with dextromethorphan bromide." American journal of nephrology 12.4 (1992): 268-270.

Sacchetti, Alfred, et al. "Leukocyte larceny: spurious hypoxemia confirmed with pulse oximetry." The Journal of emergency medicine 8.5 (1990): 567-569.

Fox, Michael J., et al. "Leukocyte larceny: a cause of spurious hypoxemia." The American journal of medicine 67.5 (1979): 742-746.

Story, David A., Hiroshi Morimatsu, and Rinaldo Bellomo. "Strong ions, weak acids and base excess: a simplified Fencl–Stewart approach to clinical acid–base disorders." British journal of anaesthesia 92.1 (2004): 54-60.

Rossing, T. H., N. Maffeo, and V. Fencl. "Acid-base effects of altering plasma protein concentration in human blood in vitro." Journal of applied physiology 61.6 (1986): 2260-2265.

Berend, Kenrick. "Diagnostic use of base excess in acid–base disorders." New England Journal of Medicine 378.15 (2018): 1419-1428.