Arterial blood gas interpretation

• An arterial blood gas (ABG) is a sample of arterial blood that measures pH, oxygen (PaO2), and carbon dioxide (PaCO2), and calculates bicarbonate and base excess; it is the gold standard for oxygenation, ventilation and acid-base status.
• It is most often taken from the radial artery after a brief check of collateral hand circulation, with firm pressure afterwards; errors in collection and handling are common, and pre-analytical rules matter as much as the interpretation.
• Read every gas the same way: pH, then the parameter that matches the pH, then compensation, then oxygenation, then the A-a gradient.
• A low oxygen level with a normal alveolar-arterial (A-a) gradient points to hypoventilation or a low inspired oxygen tension; a widened gradient points to mismatch of ventilation and perfusion, shunt or, rarely, diffusion impairment.
• Pulse oximetry can usually exclude hypoxaemia when the saturation is 92% or higher, but it is unreliable at low values and cannot exclude hypercapnia, so an ABG is still required. ^[3]

By the end of this page you should be able to:

1. State the indications and contraindications for arterial puncture and perform or assist a safe radial sample.
2. List the normal adult ranges for arterial pH, PaO2, PaCO2, bicarbonate and base excess, and record the inspired oxygen fraction and temperature with the sample.
3. Verify the internal consistency of an ABG report using the Henderson or Henderson-Hasselbalch relationship.
4. Read an ABG in a fixed order and classify the primary disturbance as respiratory or metabolic, acute or chronic, and uncomplicated or mixed.
5. Apply the alveolar gas equation to calculate the A-a gradient and use it to place the cause of hypoxaemia.
6. Explain why compensation never fully corrects the pH and how mixed disorders are recognised.
7. Cross-check ABG numbers against the renal acid-base topics and avoid the common pre-analytical errors.
8. Perform the aftercare, complications surveillance and low-resource adaptations described below.

An arterial blood gas is not only a number generator. It answers three separate clinical questions: is the patient hypoxaemic (oxygenation, PaO2), are they ventilating adequately (PaCO2), and is their acid-base balance disturbed (pH and bicarbonate). Arterial blood, coming directly from the pulmonary circulation, reflects respiratory conditions far better than venous blood can; a venous sample is useful for screening pH and bicarbonate but cannot reliably exclude hypercapnia or hypoxaemia, because venous PaCO2 runs 4.4 to 8.6 mm Hg higher than arterial and venous PO2 is roughly 35 to 50 mm Hg lower. ^[2][3]

This page treats the procedure of obtaining the sample and the bedside method of reading it. The metabolic causes of acidosis and alkalosis, the anion gap, the osmolar gap and renal tubular acidosis are treated in the companion renal topic; the two use the same compensation formulas, and the rules here are consistent with them. ^[5]

The simpler and safer non-invasive tests come first. ABG sampling is indicated when the result will change assessment or treatment, most commonly for the following reasons. ^[3]

• Critically ill patients with cardiorespiratory or metabolic dysfunction.
• Patients with oxygen saturation below 92% in whom hypoxaemia may be present.
• A deteriorating saturation that requires an increasing fraction of inspired oxygen.
• Patients at risk of hypercapnia (for example chronic obstructive pulmonary disease, neuromuscular disease, opioid overdose, chest wall disease).
• Breathless patients in whom a reliable oximetry signal cannot be obtained.
• Monitoring of the response to oxygen therapy, exercise testing, or a documented disease process.
• Suspected or confirmed respiratory failure and acute respiratory distress syndrome.
• Diabetic ketoacidosis and other severe metabolic illnesses, where the gas is part of overall assessment even when non-invasive monitoring looks acceptable.

Oxygen itself is a drug with a defined target range: 88 to 92% saturation in chronic respiratory disease with potential hypercapnia, and 92 to 96% in other clinical situations. ^[3] When oxygen is given, a repeat gas is needed if hypercapnia is a risk. ^[3]

Contraindications are relative and must be weighed against the clinical situation. The absolute bars to sampling are an inadequate blood supply to the hand after a modified Allen (collateral circulation) test, and puncture through or distal to an infected area. Avoid sampling through inflamed or cellulitic skin, from a limb with severe peripheral vascular disease, Raynaud syndrome, an arteriovenous fistula, or an arterial graft, and avoid a limb with compromised drainage such as a fracture. Take extra precautions with abnormal coagulation, medium to high dose anticoagulant therapy, or clotting disorders, which increase the risk of haematoma. A negative modified Allen test means do not use that radial artery. Sampling from the brachial artery and, especially, the femoral artery, requires training and supervision; femoral puncture should not be performed outside hospital. ^[3][6]

The radial artery is the preferred and safest default. It is superficial, palpable, has good collateral flow from the ulnar artery through the palmar arch, and is surrounded by few structures that can be injured. The brachial artery, larger and deeper, is a second choice in skilled hands. The femoral artery is large and always accessible but gives the highest complication rate, is hard to compress for long enough, and should be reserved for emergencies under direct supervision. Other sites (ulnar, posterior tibial, dorsalis pedis, axillary) are only used by skilled operators. ^[3][6]

Before radial puncture, the modified Allen test checks that the ulnar artery can supply the hand if the radial artery were to thrombose. Occlude both arteries with the hand open and relaxed; release pressure over the ulnar artery; prompt return of colour to the palm and fingers means collateral flow is adequate. A hand that stays pale is a contraindication to that radial puncture. ^[3][6]

• A pre-heparinised arterial blood gas syringe matched to the required volume, with a small needle (needle gauge selection is a balance between success rate and the complication rate, which increases with larger needles).
• Skin preparation as per local policy, gloves, apron and face protection as required.
• Clean dry gauze for immediate pressure after the puncture.
• A capped container with ice or slurry is useful when transport will be delayed, in the light of the storage rules below.
• A method of labelling the syringe with patient, time, oxygen concentration, temperature if required, and mode of ventilation. ^[1][5]

Do not assemble and leave a needle uncapped; activate safety mechanisms immediately after withdrawal.

Explain the procedure, its purpose and the expected discomfort, and obtain informed consent. A calm, well-lit private area, a stable surface and a comfortable position help, because the pain of an arterial puncture can make the patient hyperventilate or hold their breath, which itself acutely lowers PaCO2 and creates a misleading result. ^[2][5]

1. Verify patient identity, indications, current oxygen or ventilator settings, and the required tests.
2. Position the patient supine or comfortably, with the arm on a stable surface; avoid a clenched fist, breath holding or crying in children, since these alter ventilation and distort the sample. ^[5]
3. Perform and document a modified Allen test if the radial site is chosen.
4. Prepare the tray and clean hands; use personal protective equipment as required by local policy.
5. Allow the patient's respiratory state to stabilise before sampling, particularly if they have just been moved or distressed.

The following sequence is a teaching outline. Follow the current local standard operating procedure and the World Health Organization guidance on drawing blood that your department uses. ^[1]

1. Clean the planned site and allow it to dry fully; do not touch the site again.
2. Prefill a heparinised syringe to its designed volume if necessary; if liquid heparin is used, hold the syringe vertical with the needle upward and expel all excess heparin and air before sampling. ^[1]
3. Palpate the artery with the non-dominant index finger; inform the patient.
4. Insert the needle at a shallow angle a short distance distal to the palpating finger, without piercing the far wall, and advance until blood flashes back.
5. Keep the needle steady and allow the arterial pressure to fill the syringe to the appropriate volume; do not pull the plunger. ^[5]
6. Withdraw the needle smoothly; place clean dry gauze immediately and apply firm pressure until bleeding has stopped. Use longer compression if the patient is anticoagulated, has a bleeding disorder, or is hypertensive. ^[6]
7. Activate the needle safety device, then mixing is needed without trapping air bubbles: cap the syringe, expel any visible air by holding the syringe vertical with the needle upward, and gently roll or invert the syringe (typically 10 inversions) to mix the anticoagulant. ^[1][5]
8. Label the syringe, record the current oxygen concentration and (for ventilated patients) the mode and settings on the request, and transport immediately. Temperature and pacemaker or other relevant data should accompany the sample per local policy.
9. Inspect the puncture site for bleeding, thank the patient, and document.

Never recap a used needle with two hands; use the one-hand scoop or a safety device. ^[6]

• Apply firm pressure until the site is dry, then an occlusive dressing.
• Inspect the site, the colour and warmth of the fingers, and sensation after radial puncture; a persistent haematoma, swelling, pain, pallor or loss of pulses requires escalation.
• Dispose of sharps, clean, and document the sample and any difficulty encountered.
• Repeat or correct sampling errors rather than interpreting a doubtful gas. ^[2]

Minor events are common and major ones rare. Anticipate and recognise the following, all of which are documented complications of arterial sampling. ^[6]

• Pain, which may itself trigger hyperventilation and an acutely low PaCO2. ^[2]
• Haematoma or local haemorrhage, especially with anticoagulation or coagulopathy; inadequate seal because of high arterial pressure. ^[6]
• Arteriospasm, which may occur even after successful needle entry.
• Vasovagal syncope; reduce risk by explaining the procedure and keeping the patient supine. ^[5]
• Nerve injury from deep or redirected needles.
• Arterial occlusion, thrombosis or distal ischaemia, particularly with a negative collateral test.
• Air embolism, thromboembolism and, rarely, pseudoaneurysm formation.
• Infection or nosocomial bacteraemia from inadequate skin preparation.
• Erroneous results from non-arterial sampling, air bubbles, incorrect anticoagulation, haemolysis or storage errors, which must be treated as complications of the process rather than of the patient. ^[1][5]

Most errors in blood gas results are made before the sample reaches the analyser. Three rules are consistent across sources. ^[1][5][6]

• Air bubbles. Any bubble must be expelled immediately after withdrawal, before cooling or mixing, because gas exchange across the bubble reduces PaCO2 and increases PaO2. A bubble occupying as little as 1% of the syringe volume can seriously affect the PaO2 reading. ^[5]
• Timing and temperature. Analyse room-temperature samples within 15 minutes, and samples stored on ice within 1 hour; when pO2 and pCO2 are critical, analysis within 5 minutes is preferred. Some sources advise not routinely cooling the sample and using glass syringes with ice slurry only if analysis will be delayed beyond 30 minutes, whereas others advise ice if immediate analysis is impossible. Follow your laboratory's policy. ^[1][5][6]
• Anticoagulant. Use the anticoagulant the syringe was designed for. Liquid heparin dilutes the sample and lowers electrolytes, glucose and haematocrit; only 0.05 mL heparin is needed per mL blood, and over heparinising or underfilling the dead space both introduce error. Dry balanced heparins minimise ionic binding. Mix thoroughly by about 10 inversions to avoid clotting. ^[1][5]

Unchilled, delayed samples suffer ongoing cellular metabolism: PaCO2 rises 3 to 10 mm Hg per hour, pH falls, and PaO2 falls. ^[2] Mixed or non-arterial sampling, haemolysis or inadequate labelling void the result. ^[6]

Venous blood gas analysis is less invasive and provides rapid pH, bicarbonate, lactate and electrolytes, but it cannot accurately estimate PaO2 or PaCO2; a venous PCO2 below 40 mm Hg makes hypercapnia unlikely but does not exclude it. Capillary (earlobe or fingertip) gas is an alternative when arterial puncture fails: it gives reliable pH and PaCO2 but variably underestimates PaO2. In a setting without a blood gas machine, venous PCO2, clinical assessment and repeated clinical review carry most of the load. ^[2][3]

These ranges come from the cited sources; individual laboratories and age groups vary slightly, so compare with the local reference interval. ^[1][2]

On room air at sea level, the arterial PO2 is normally about 95 mm Hg after accounting for the small physiological right-to-left shunt, and the equivalent alveolar value is about 100 mm Hg. ^[4] Arterial PO2 declines with age even in health. ^[2]

A blood gas report can be internally inconsistent because of pre-analytical or transcription error. Check the Henderson relationship, hydrogen ion equals 23.9 (or 24) times (PaCO2 divided by bicarbonate) and also 10 to the power of (9 minus pH), equivalently pH equals 6.1 plus log of (bicarbonate divided by 0.0301 times PaCO2). ^[1][2]

For a quick estimate, subtract the two digits after the decimal point of the pH from 80 to approximate the hydrogen ion in nmol/L (a pH of 7.23 gives about 57). Equivalently, start from a hydrogen ion of 40 nmol/L at pH 7.40 and multiply by 1.25 for each 0.1 unit fall in pH, or divide by 1.25 for each 0.1 unit rise. If the pH calculated from the reported bicarbonate and PaCO2 differs from the reported pH by more than about 0.05, suspect an error and repeat the test. ^[1][2]

Obtain the clinical history in parallel. A gas is interpreted against the patient: chronic respiratory acidosis and acute ketoacidosis look very different to the same numbers in a previously well patient. ^[5]

Apply the steps in the same order every time, and do not jump ahead to the anion gap before the respiratory component is understood. ^[1]

The "CLEAR" bedside order (Check oxygenation, Look at pH, Evaluate the primary disorder and compensation, Assess anion gap, Rule out additional disorders) is an equivalent route through the same information. ^[1]

Oxygenation is read alongside, not instead of, ventilation. First estimate the oxygen level expected for the patient's age and the inspired oxygen: predicted room-air PaO2 is about 100 minus 0.3 times the age in years, or roughly five times the percentage of inspired oxygen. For this, classify hypoxaemia as mild at a PaO2 of 60 to 79 mm Hg, moderate at 40 to 59 mm Hg, and severe below 40 mm Hg. ^[1][5] A saturation of at least 92% practically excludes hypoxaemia (defined as SaO2 under 90% or PaO2 under 60 mm Hg), though pulse oximetry can under-read by up to about 8% and miss a low saturation in up to 28% of cases, and its accuracy worsens in severe hypoxaemia, shock and poor perfusion. ^[3]

The A-a gradient quantifies gas exchange. The alveolar oxygen tension is not measured directly; it is estimated from the alveolar gas equation, PAO2 = (FiO2 times (barometric pressure minus water vapour pressure of 47 mm Hg)) minus (PaCO2 divided by the respiratory exchange ratio, usually 0.8). On room air at sea level this is roughly 100 mm Hg, and the blood leaving the alveoli is slightly lower at about 95 mm Hg because of normal physiological shunt. ^[4] The gradient is PAO2 minus PaO2. ^[4]

Interpretation follows three categories:

• Hypoxaemia with a normal PAO2 points to alveolar hypoventilation (raised PaCO2) or a low inspired oxygen (high altitude, low atmospheric pressure). ^[4]
• Hypoxaemia with a normal A-a gradient points to hypoventilation or a low inspired oxygen tension; the alveoli and the blood are both low together. ^[4][5]
• Hypoxaemia with a widened A-a gradient points to ventilation-perfusion mismatch, right-to-left shunt, or (rarely in intensive care) diffusion impairment. Shunt is distinguished by its poor response to supplemental oxygen. ^[4]

A normal value is under 15 mm Hg in young adults and under (age in years divided by 4) plus 4 in older adults. It must always be compared with the age-adjusted expectation, not with a single cutoff, and the expectation rises at higher inspired oxygen concentrations. An A-a gradient below 20 mm Hg in a patient with hypoxaemic respiratory failure indicates an extrapulmonary cause (central nervous system depression, neuromuscular failure, chest wall or pleural disease). ^[4][5]

Identify the single primary process, then test compensation. PaCO2 is the respiratory acid and bicarbonate the metabolic base; they move in opposite directions in a normal gas. ^[5]

• If the pH is acidaemic (under 7.35), an elevated PaCO2 (over 45 mm Hg) points to respiratory acidosis, while a low bicarbonate (under 22 mmol/L) points to metabolic acidosis.
• If the pH is alkalaemic (over 7.45), a low PaCO2 (under 35 mm Hg) points to respiratory alkalosis, while a raised bicarbonate (over 26 mmol/L) points to metabolic alkalosis.
• A normal pH with abnormal PaCO2 and bicarbonate means a mixed disorder with two opposing processes. ^[1]

Acute terms are separated from chronic using the rate of compensation. Breath holding or pain during sampling can acutely lower PaCO2 and create a spurious respiratory alkalosis, so the clinical picture matters. ^[2][5]

Compensation is the "other system" returning the pH toward normal, and it never fully succeeds: the maximum response usually restores only 50 to 75% of the pH, and compensation never overshoots. Acute respiratory compensation occurs within 6 to 24 hours and chronic compensation within 1 to 4 days, so an abnormal bicarbonate with acute respiratory change implies a mixed picture rather than compensation. ^[5]

For respiratory disorders, the ratio of the change in hydrogen ion to the change in PaCO2 (delta H+ divided by delta PaCO2) separates acute (above 0.8) from chronic (below 0.3) and acute-on-chronic (0.3 to 0.8). ^[5]

In metabolic acidosis, the minimum (complete) compensation formula (Winter's) is also written PaCO2 = 1.5 × [HCO3−] + 8, with a ±2 mm Hg range. ^[1] The rate of compensation differs by source: some quote acute respiratory acidosis bicarbonate increments of 1 mmol/L and chronic of 3.5 mmol/L per 10 mm Hg; others quote 1 to 2 and 3 to 4. ^[1][5] Both conventions are acceptable as long as they are applied consistently. ^[1][5]

Mixed disorders are present when compensation is inadequate, when PaCO2 and bicarbonate change in opposite directions, or when the pH is extremely abnormal (below 7.20 or above 7.55), which implies two acidosis or two alkalosis insults moving the hydrogen ion in the same direction. ^[1][2][5]

For every 1 mmol/L rise in the anion gap, bicarbonate should fall by 1 mmol/L. Add the change in anion gap to the measured bicarbonate (the corrected bicarbonate); a result above 24 to 26 mmol/L reveals a hidden metabolic alkalosis behind the acidosis, and the delta gap (delta anion gap minus delta bicarbonate) is normal between minus 6 and plus 6, above +6 suggesting an additional metabolic alkalosis and below −6 an additional non-anion-gap acidosis. ^[1] The renal acid-base topic covers the anion gap adjustments for albumin, the osmolar gap, the urine anion gap, the RTA types and the definitions of chloride-responsive and chloride-resistant metabolic alkalosis; use the same normal anion gap of 8 to 12 mmol/L and the same compensation formulas. ^[1][5]

If the anion gap is 20 mmol/L or higher, a metabolic acidosis is present regardless of pH or bicarbonate. ^[1]

Example 1 - chronic respiratory acidosis in an exacerbation. A 68-year-old man with chronic obstructive pulmonary disease presents with confusion. He is breathing air. The arterial blood gas shows pH 7.30, PaCO2 70 mm Hg, bicarbonate 33 mmol/L, PaO2 52 mmHg. The pH is acidaemic and the PaCO2 is high, so the primary disturbance is respiratory acidosis. Bicarbonate of 33 is high for a normal person, so renal compensation has occurred; using the chronic ratio of about 3.5 mmol/L bicarbonate per 10 mm Hg rise in PaCO2, this fits chronic disease with acute worsening rather than an acute event alone. ^[5]

Example 2 - high anion gap metabolic acidosis with a second, less obvious problem. A 55-year-old man is lethargic after antifreeze ingestion. Arterial blood gas: pH 7.12, PaCO2 22 mm Hg, PaO2 85 mm Hg, bicarbonate 7 mmol/L. Electrolytes: sodium 140 mmol/L, chloride 100 mmol/L; serum osmolality 360 mOsm/kg; albumin 40 g/L. The acidaemia is explained by the low bicarbonate, so this is a metabolic acidosis. Winter's formula gives an expected PaCO2 of (1.5 × 7) + 8 = 18.5 ± 2 mm Hg, and the measured 22 is close to that, so respiratory compensation is appropriate. The anion gap is 140 − (100 + 7) = 33 mmol/L, raised, and the osmolar gap is 69 mOsm/kg. The delta gap is within normal limits, so no hidden metabolic alkalosis is unmasked here; the picture is a high anion gap metabolic acidosis with normal compensation. ^[1]

Example 3 - mixed disorder with normal pH. An ill patient has pH 7.40, PaCO2 41 mm Hg, bicarbonate 22 mmol/L, sodium 137 mmol/L and chloride 90 mmol/L. A normal pH with a near-normal PaCO2 and bicarbonate does not exclude disease: the anion gap is 137 − (90 + 22) = 25 mmol/L, raised, and the delta gap is 15 with a corrected bicarbonate of 37 mmol/L. There is therefore a hidden high anion gap metabolic acidosis with a concurrent metabolic alkalosis, a combination that can be missed if the report is not read in this order. ^[1]

• Where an analyser is unavailable, venous PCO2 below 40 mm Hg makes hypercapnia unlikely, and a rising or falling trend in clinical signs with saturation and venous gases often suffices for treatment decisions. ^[3]
• Respect the handling times even when transport is limited (analyse at room temperature within 15 minutes; iced within 1 hour; 5 minutes when critical). ^[1][5][6]
• A-a gradient calculations need the inspired oxygen fraction and barometric pressure; at altitude, PAO2 falls with atmospheric pressure. ^[4]
• Explain and report every sample fully, so the reader does not have to reconstruct the oxygen conditions later. ^[1]

• An ABG directly measures pH, PaO2 and PaCO2 and calculates bicarbonate and base excess; arterial blood, not venous blood, is needed for reliable oxygenation and ventilation assessment, while venous gas rapidly screens pH and bicarbonate. ^[2][3]
• The radial artery is preferred, with a documented modified Allen test; contraindications include a negative test, infection over the site and severe vascular disease. ^[3][6]
• Most ABG errors are pre-analytical: bubbles, inadequate mixing, wrong anticoagulant, temperature and delays; handle the sample to the rules above. ^[1][5][6]
• Normal adult ranges at sea level: pH 7.35 - 7.45, PaO2 75 - 100 mm Hg, PaCO2 35 - 45 mm Hg, bicarbonate 22 - 26 mmol/L, base excess −2 to +2, saturation 95 to 100%. ^[1][2]
• Read in order: oxygenation, pH, primary respiratory or metabolic cause, compensation, A-a gradient, then anion and delta gaps. ^[1]
• A normal A-a gradient is under 15 mm Hg in young adults and under (age divided by 4) + 4 in older adults; below 20 mm Hg in hypoxaemic respiratory failure points to an extrapulmonary cause. ^[4][5]
• A normal saturation of at least 92% effectively excludes hypoxaemia (PaO2 under 60 mm Hg), but cannot exclude hypercapnia; measure the gas when the patient is deteriorating, when hypoxaemia is present, or when hypercapnia is suspected. ^[3]
• Compensation never fully restores the pH and never overshoots; the ratio of delta hydrogen ion to delta PaCO2 separates acute from chronic respiratory disorder. ^[5]
• Acute compensation takes 6 to 24 hours and chronic 1 to 4 days; metabolic and respiratory compensation formulas must always be compared with the actual values. ^[5]
• Cross-check ABG interpretation with the renal acid-base topics; the same anion gap, urine anion gap and delta gap rules apply. ^[1][5]

Normal values in health and disease; ventilation and gas exchange; hypoxaemia; hypercapnia; oxygen therapy; acute respiratory failure; pneumonia and ARDS; COPD exacerbation; asthma; metabolic acidosis and alkalosis; renal tubular acidosis; diabetic ketoacidosis; respiratory mechanics.