ABG interpretation: a 6-step method with worked examples (kPa)

What are the normal ABG values in the UK (kPa)?

Ranges vary slightly between labs and analysers. Always use your own lab's ranges. These are typical UK values:

Converting: to change kPa to mmHg, multiply by 7.5. To change mmHg to kPa, multiply by 0.133.

Step 1: who is the patient, and what oxygen are they on?

Write down the clinical context (for example, COPD, diabetes, vomiting, sepsis) and how much oxygen the patient is breathing. A PaO2 of 11 kPa is normal on room air (FiO2 0.21) but very poor on 60% oxygen.

Step 2: is there respiratory failure? Type 1 vs type 2

• Hypoxaemia: PaO2 below the normal range. Respiratory failure is usually defined as PaO2 below 8 kPa.
  • Type 1: low PaO2 with a normal or low PaCO2. This is a gas-exchange problem, as in pneumonia, pulmonary embolism or pulmonary oedema.
  • Type 2: low PaO2 with a high PaCO2 (above 6.0 kPa). This is a ventilation problem, as in COPD, opioid toxicity or neuromuscular weakness.
• Is the PaO2 appropriate for the FiO2?
  • Rule of thumb: on low to moderate oxygen (about 21–40%), the PaO2 in kPa should be roughly the FiO2 % minus 10. On 28% oxygen, expect a PaO2 of about 18 kPa if the lungs are healthy. The rule breaks down at higher oxygen concentrations; use the A–a gradient instead.
  • The formal measure is the alveolar–arterial (A–a) gradient:
    • PAO2 (kPa) ≈ 95 × FiO2 − PaCO2 ÷ 0.8
    • A–a gradient = PAO2 − PaO2. Normally under about 2 kPa in young adults, rising with age (a rough guide in mmHg is age ÷ 4 + 4) to about 3–4 kPa in older adults.
    • A raised gradient points to a gas-exchange problem. A normal gradient with hypoxaemia suggests hypoventilation.

Oxygen targets (British Thoracic Society):

• Aim for 94–98% in most acutely ill patients.
• In those at risk of hypercapnic respiratory failure (for example, COPD), aim for 88–92% while waiting for the blood gas.
  • If the PaCO2 is normal, and there is no history of needing NIV or invasive ventilation, aim for 94–98%, and repeat the gas in 30–60 minutes to check the PaCO2 hasn't risen.
  • If the PaCO2 is raised, keep 88–92% and repeat the gas in 30–60 minutes.

A new BTS guideline on oxygen targets is in development, so check for an update.

Step 3: is the pH acidaemic or alkalaemic?

• Below 7.35: acidaemia.
• Above 7.45: alkalaemia.
• A normal pH doesn't exclude a disorder. Fully compensated or mixed disorders can give a normal pH.

Step 4: what is the primary disorder?

The base excess is printed on every UK blood gas. Unlike the measured bicarbonate, it isn't shifted by acute changes in PaCO2, so it is a cleaner marker of the metabolic component.

The primary change is the one that explains the pH. If both the PaCO2 and the HCO3⁻ move in the direction that pushes the pH the same way (for example, high PaCO2 and low HCO3⁻), you have a mixed disorder.

Step 5: how do you check compensation (in kPa)?

These bedside rules were derived in mmHg. The kPa versions below are straight unit conversions (1 kPa ≈ 7.5 mmHg), and the baseline PaCO2 is about 5.3 kPa (40 mmHg).

If the measured value falls outside the expected range, there is a second disorder. For example, in metabolic acidosis a PaCO2 higher than Winter's formula predicts means a coexisting respiratory acidosis, such as an exhausted patient.

Step 6: how do you calculate the anion gap and delta ratio?

The anion gap:

• Formula: anion gap = Na⁺ − (Cl⁻ + HCO3⁻).
• Some labs add K⁺, and that raises the normal range by about 4.
• Correct for a low albumin: add about 0.25 × (40 − albumin in g/L). Low albumin hides a raised gap.

Causes of a high anion gap acidosis. The GOLDMARK mnemonic covers most of them:

• Glycols (ethylene and propylene)
• Oxoproline (from regular paracetamol)
• L-lactate
• D-lactate
• Methanol
• Aspirin
• Renal failure
• Ketoacidosis (diabetic, alcoholic or starvation)

Causes of a normal anion gap (hyperchloraemic) acidosis:

• diarrhoea and other gut bicarbonate losses;
• renal tubular acidosis;
• acetazolamide;
• ureteric diversion into the colon;
• large volumes of 0.9% sodium chloride.

The delta ratio: (anion gap − 12) ÷ (24 − HCO3⁻).

Four worked examples

Example 1: a 19-year-old with type 1 diabetes, vomiting, on room air

pH 7.20 · PCO2 2.5 kPa · HCO3⁻ 7 · Na⁺ 135 · K⁺ 4.6 · Cl⁻ 100 · glucose 28 mmol/L

In DKA a venous gas is enough for pH, bicarbonate and potassium (JBDS guidance); an arterial sample isn't needed unless there is a breathing or oxygenation problem.

1. Context: on room air; diabetes and vomiting.
2. Oxygenation: not assessed on a venous gas (saturations on room air are normal).
3. pH: acidaemia.
4. Primary disorder: HCO3⁻ is low, so this is a metabolic acidosis.
5. Compensation: Winter's formula gives 0.2 × 7 + 1.07 = 2.47 kPa (range 2.2–2.7). The measured PaCO2 of 2.5 is in range, so the respiratory compensation is appropriate.
6. Anion gap and delta ratio:
   • Anion gap = 135 − (100 + 7) = 28, which is high.
   • Delta ratio = (28 − 12) ÷ (24 − 7) = 16 ÷ 17 ≈ 0.94, consistent with a high anion gap acidosis (typical of DKA).

Conclusion: a high anion gap metabolic acidosis with appropriate compensation, consistent with diabetic ketoacidosis. Confirm with blood ketones and follow your trust's DKA protocol (JBDS). The potassium guides how much potassium goes into the fluids.

Example 2: a 72-year-old with COPD, more breathless, on 28% oxygen

pH 7.29 · PaCO2 9.2 kPa · PaO2 7.1 kPa · HCO3⁻ 32

1. Context: COPD exacerbation on 28% oxygen.
2. Oxygenation: a PaO2 of 7.1 kPa is below 8 and the PaCO2 is above 6.0, so this is type 2 respiratory failure. On 28%, a healthy PaO2 would be about 18 kPa. A PaO2 of 7.1 kPa usually means saturations below 88%, so the patient needs more oxygen, not less.
3. pH: acidaemia.
4. Primary disorder: PaCO2 is high, so this is a respiratory acidosis.
5. Compensation: 9.2 kPa ≈ 69 mmHg, a rise of about 29 mmHg.
   • An acute rise would take the HCO3⁻ to about 27.
   • A chronic rise would take it to about 36.
   • The measured 32 lies between the two, so this is acute-on-chronic respiratory acidosis.

Conclusion: acute-on-chronic hypercapnic (type 2) respiratory failure.

• Increase the oxygen (for example, the next Venturi step) to reach 88–92%. Don't leave the patient below target.
• Give optimal medical treatment: bronchodilator nebulisers driven by air, with oxygen given separately if needed, plus steroids and antibiotics if indicated. Repeat the gas within about an hour and get a senior review.
• UK guidance (BTS/ICS) recommends non-invasive ventilation when the pH stays below 7.35 and the PaCO2 above 6.5 kPa despite optimal medical therapy.

Example 3: a 24-year-old with sudden breathlessness and tingling fingers, on room air

pH 7.55 · PaCO2 3.3 kPa · PaO2 13.5 kPa · HCO3⁻ 21

1. Context: on room air; sudden breathlessness with tingling fingers.
2. Oxygenation: the PaO2 looks normal, but check the gradient. The A–a gradient is (95 × 0.21 − 3.3 ÷ 0.8) − 13.5 = 15.8 − 13.5 ≈ 2.3 kPa (about 17 mmHg). For a 24-year-old the expected gradient is about 24 ÷ 4 + 4 = 10 mmHg (1.3 kPa), so this is mildly raised.
3. pH: alkalaemia.
4. Primary disorder: PaCO2 is low, so this is a respiratory alkalosis.
5. Compensation: 3.3 kPa is about 25 mmHg, a fall of about 15 mmHg from 40. In acute respiratory alkalosis HCO3⁻ falls about 2 mmol/L per 10 mmHg, so the expected fall is about 3, giving 24 − 3 = 21. The measured 21 matches, so this is acute respiratory alkalosis.

Conclusion: acute respiratory alkalosis with a mildly raised A–a gradient. That is a gas-exchange problem until proven otherwise, so pulmonary embolism must be actively considered with a proper clinical risk assessment. Don't label this as anxiety. (Even a normal gradient doesn't exclude a PE.)

Example 4: a 45-year-old with a week of vomiting, on room air

pH 7.52 · PaCO2 6.4 kPa · PaO2 11.2 kPa · HCO3⁻ 38 · Na⁺ 137 · K⁺ 3.0 · Cl⁻ 88

1. Context: on room air; prolonged vomiting.
2. Oxygenation: PaO2 normal.
3. pH: alkalaemia.
4. Primary disorder: HCO3⁻ is high, so this is a metabolic alkalosis.
5. Compensation: the expected PaCO2 is 0.7 × 38 + 20 = 46.6 mmHg ≈ 6.2 kPa (range about 5.5–6.9). The measured 6.4 is in range, so the compensation is appropriate. A raised PaCO2 here is compensation, not lung disease.

Conclusion: a hypochloraemic, hypokalaemic metabolic alkalosis from vomiting. Replace fluid, chloride and potassium.

Test yourself: three quick ABGs

Work through each one with the six steps before you read the answer. These cases were written by our team.

Case 1. A 72-year-old man with COPD is breathless, on 28% oxygen through a Venturi mask. pH 7.31 · PaCO2 8.2 kPa · PaO2 8.9 kPa · HCO3⁻ 31 mmol/L · base excess +6

Answer

• Primary disorder: acidaemia with a high PaCO2, so a respiratory acidosis.
• Compensation: the PaCO2 is 2.9 kPa above 5.3. An acute rise would lift the HCO3⁻ to about 26; a chronic one to about 33. A measured 31 lies between, which suggests acute-on-chronic respiratory acidosis. The positive base excess fits long-standing renal compensation.
• Oxygenation: hypercapnic (type 2) respiratory failure. On 28% oxygen, healthy lungs would give a PaO2 of about 18 kPa, so 8.9 shows a large gas-exchange problem too.
• Next steps:
  • Keep saturations at 88–92%.
  • Give optimal medical treatment, and repeat the gas.
  • British Thoracic Society guidance is to start non-invasive ventilation if the pH stays below 7.35 and the PaCO2 above 6.5 kPa despite optimal medical treatment.

Case 2. A 24-year-old woman has tingling fingers and feels anxious. She is breathing room air. pH 7.55 · PaCO2 3.4 kPa · PaO2 13.8 kPa · HCO3⁻ 22 mmol/L · base excess −1

Answer

• Primary disorder: alkalaemia with a low PaCO2, so a respiratory alkalosis.
• Compensation: an acute fall of 1.9 kPa predicts an HCO3⁻ about 3 lower, around 21. The measured 22 fits an acute process.
• A–a gradient: PAO2 ≈ 95 × 0.21 − 3.4 ÷ 0.8 ≈ 15.7 kPa, so the gradient is about 1.9 kPa. That's normal for her age.
• Interpretation: this fits hyperventilation. But a normal A–a gradient doesn't fully exclude a pulmonary embolism, so only put it down to anxiety after you've considered the alternatives.

Case 3. A 45-year-old man has had diarrhoea for five days. He is breathing room air. pH 7.29 · PaCO2 3.9 kPa · PaO2 12.9 kPa · HCO3⁻ 14 mmol/L · Na⁺ 138 · Cl⁻ 114 · K⁺ 3.1 · base excess −11

Answer

• Primary disorder: acidaemia with a low HCO3⁻ and a negative base excess, so a metabolic acidosis.
• Compensation: Winter's formula predicts a PaCO2 of 0.2 × 14 + 1.07 ≈ 3.9 kPa (range about 3.6–4.1). The measured 3.9 shows appropriate respiratory compensation, so there is no second respiratory disorder.
• Anion gap: 138 − (114 + 14) = 10 mmol/L, which is normal. So this is a normal anion gap (hyperchloraemic) metabolic acidosis, which fits bicarbonate loss from the gut.
• Also note: the low potassium, also from gut losses. Replace it, and keep in mind that as the acidosis corrects, potassium moves back into cells, so the level can fall further.

Lactate and sepsis: what does NICE say now?

NICE replaced its 2016 sepsis guideline (NG51) in November 2025:

• NG253 covers people aged 16 and over who are not pregnant or recently pregnant;
• NG254 covers children;
• NG255 covers pregnancy.

In NG253's adult algorithm, for someone at moderate risk, a venous lactate over 2 mmol/L (or evidence of acute kidney injury) counts as hypoperfusion and moves them to high-risk management. Pregnant and recently pregnant people (NG255) and children (NG254) have their own criteria. Use the guideline for your patient group and your local sepsis pathway.

ABG or VBG: which do you need?

Common sampling errors

• Air bubbles in the syringe: the PaO2 drifts towards room-air values and the PaCO2 falls. Expel air straight away.
• Delay before analysis: cells keep metabolising, so the PaO2 falls, the PaCO2 rises and the lactate rises. Run the sample promptly.
• Too much liquid heparin: dilutes the sample and lowers the PaCO2 and bicarbonate. Use pre-heparinised syringes.
• Venous sample sent as arterial: a low PaO2 with normal saturations on the monitor should make you suspect this.

Carbon monoxide: why the gas machine matters

Pulse oximeters can't tell carboxyhaemoglobin from oxyhaemoglobin, so saturations can look normal in carbon monoxide poisoning. The blood gas analyser's co-oximetry measures carboxyhaemoglobin (COHb) directly. Check it in anyone with smoke exposure or unexplained headache and confusion, and give high-flow oxygen.

How can UKMLARevisions help?

• Test yourself: free UKMLA AKT questions, with every option explained.
• Find your gaps: questions are mapped to the MLA content map, so you can see which conditions and presentations you've covered.
• Make it stick: turn the normal ranges and compensation rules into a flashcard set, and let Smart Revision bring back the questions you get wrong.
• Practise under exam conditions: full AKT mocks, Paper 1 and Paper 2, 100 questions and 2 hours each.

Sources (checked 18–19 October 2026)

• British Thoracic Society / Intensive Care Society: Davidson AC et al. Guideline for the ventilatory management of acute hypercapnic respiratory failure in adults. Thorax 2016;71(Suppl 2):ii1–ii35.
• Reference ranges (NHS labs, plus LITFL for base excess):
  • Liverpool University Hospitals blood gases
  • North Bristol NHS Trust (Severn Pathology) blood gases
  • LITFL: base excess vs standard base excess
• Oxygen targets:
  • O'Driscoll BR et al. BTS guideline for oxygen use in adults in healthcare and emergency settings. Thorax 2017;72(Suppl 1):i1–i90.
  • BTS target oxygen saturation guideline (in development)
  • NICE QS10 statement 6
• NIV in COPD: Davidson AC et al. BTS/ICS guideline for the ventilatory management of acute hypercapnic respiratory failure in adults. Thorax 2016;71(Suppl 2):ii1–ii35.
• Compensation rules:
  • Brandis K. Bedside rules for assessment of compensation.
  • Albert MS, Dell RB, Winters RW. Ann Intern Med 1967;66(2):312–322.
• A–a gradient: RCEMLearning: A–a gradient
• Anion gap:
  • University Hospitals Sussex: anion gap
  • Higgins C. Clinical aspects of the anion gap
  • FICM. Albumin correction for the anion gap
  • Figge J et al. Crit Care Med 1998;26(11):1807–10.
• Delta ratio: LITFL: delta ratio
• GOLDMARK: Mehta AN, Emmett JB, Emmett M. Lancet 2008;372(9642):892.
• Normal anion gap causes: NHS Scotland IMD metabolic acidosis protocol
• DKA: Joint British Diabetes Societies for Inpatient Care. The management of diabetic ketoacidosis in adults (current edition).
• Sepsis:
  • NICE NG253
  • NG253 algorithm (NEWS2)
  • NG254 update information