The 'Normal' Anion Gap in a Vomiting Patient with an Albumin of 2.1
A 62-year-old woman is admitted at 16:20 from a nursing home with three days of vomiting and poor oral intake. Her medications are furosemide 40 mg twice daily and lisinopril 20 mg daily. A stat basic metabolic panel and an arterial blood gas are drawn at the same time. The analyzer reports sodium 132 mmol/L, potassium 3.1 mmol/L, chloride 88 mmol/L, bicarbonate 34 mmol/L, glucose 96 mg/dL, lactate 1.4 mmol/L, and albumin 2.1 g/dL (reference 3.5-5.0). The blood gas gives pH 7.47, pCO2 48 mmHg, pO2 72 mmHg on room air. Middleware auto-calculates the anion gap as 10 mmol/L and flags it as normal. The resident pages the laboratory at 17:05 and asks only whether the potassium needs replacing. The night tech is alone with 40 other specimens queued, and the nursing home chart lists no BUN or creatinine from the past six weeks.
1. Is the reported anion gap of 10 mmol/L a valid conclusion? Interpret the full acid-base and electrolyte picture and state what you would report.
Read the worked answer
Recompute from the reported numbers. AG = Na - (Cl + HCO3) = 132 - (88 + 34) = 10 mmol/L. Taken at face value that looks normal, and that is exactly the trap.
Correct for albumin. Albumin carries most of the unmeasured anion charge, so the gap falls about 2.5 mmol/L for every 1 g/dL the albumin is below 4.0. Corrected AG = 10 + 2.5 x (4.0 - 2.1) = 10 + 4.75, or about 14.8 mmol/L. That is above the usual 8-12 interval, so an unmeasured anion is present.
Check the compensation. Expected pCO2 for a metabolic alkalosis is roughly 0.7 x HCO3 + 21 = 0.7 x 34 + 21 = 44.8 mmHg, with a +/-2 margin. The measured 48 mmHg sits slightly above that range, meaning there is a mild additional respiratory acidosis (opioids, sedation, COPD, splinting, or hypoventilation) rather than pure compensation.
Synthesize. This is a metabolic alkalosis from vomiting plus a loop diuretic (hypochloremia of 88 fits) with a superimposed high-anion-gap metabolic acidosis. Lactate is normal, so the likely contributors are ketosis from starvation, uremia (no renal function is on file), or a drug/toxin such as salicylate.
Report the albumin-corrected gap with an interpretive comment rather than endorsing the middleware's 'normal' flag.
2. Write out the substance of the 17:05 return call to the resident. What is said first, and what do you ask for?
Read the worked answer
Structure the call so the safety-critical information cannot be missed:
- Identify and verify. Give your name and department, the patient's name, medical record number, date of birth, and the collection time. Ask the resident to read the identifiers back.
- Lead with the corrected gap, not the potassium. 'Her anion gap is reported as 10, but her albumin is 2.1, so the albumin-corrected gap is about 15. She has a metabolic alkalosis from vomiting and furosemide, plus a high-gap acidosis on top of it.'
- Give the numbers so the resident can act independently: sodium 132, chloride 88, bicarbonate 34, pH 7.47, pCO2 48 - slightly under-compensating, so watch for hypoventilation.
- State the gap in the workup. Ask for BUN and creatinine (nothing on file for six weeks), a beta-hydroxybutyrate or ketone check, and a salicylate level if there is any question of ingestion. Suggest a repeat panel in 4-6 hours once fluids are running.
- Answer the original question. The potassium of 3.1 is real and is driven by the alkalosis and the diuretic, but replacing it without addressing the gap and the volume depletion risks under-treating the acidosis.
- Close the loop. Ask for a read-back of the corrected gap and the requested add-on tests, record the name and title of the person receiving the call, the time, and the read-back in the LIS comment or the call log. A critical or unexpected result that is called without a documented read-back is not defensible on review.
3. The resident later asks why the laboratory does not simply have the middleware replace the anion gap with an albumin-corrected value. Recommend a policy and justify it.
Read the worked answer
Recommendation: display both, suppress neither. The middleware should show the measured gap and, immediately beside it, an albumin-corrected gap whenever albumin is available, with an explicit interpretive comment. The uncorrected value should not be deleted.
Why not auto-replace:
- Reference intervals change with the formula. A corrected gap is only useful if the clinician knows which reference interval applies. Reporting a corrected value under a label whose reference range was derived for uncorrected values invites misreading.
- Albumin is not always available or current. If the albumin was drawn 12 hours earlier, or is from a different encounter, the correction silently imports an assumption.
- Delta checks and trending break. Historical comparison depends on a consistent calculation.
- The correction factor is an estimate, approximately 2.5 mmol/L per 1 g/dL, and varies with the assay and with extremes of albumin.
Practical build: flag only when the measured gap is within reference limits but the corrected gap exceeds them - that is the clinically dangerous discordance. Add a standard comment: 'Anion gap corrected for albumin 2.1 g/dL is approximately 15 mmol/L; interpret with the clinical picture.' Validate the rule with the medical director and a sample of 30-50 stripped and spiked specimens before go-live, document the validation, and include the change in competency training. Version-control the rule so that every report can be traced to the logic in force on the day it was issued.