A diabetic man with a normal glucose and a swollen leg

Working diagnosis: diabetic ketoacidosis precipitated by left leg cellulitis, with acute kidney injury, in a patient taking an SGLT2 inhibitor.

• pH 7.16 with bicarbonate 8 mmol/L and ketones 5.8 mmol/L satisfies the biochemical definition of DKA (glucose above 11 mmol/L, pH below 7.3, bicarbonate below 18 mmol/L, ketones above 3 mmol/L).
• Anion gap = 131 − (96 + 8) = 27 mmol/L, a high-anion-gap metabolic acidosis. The pCO2 of 22 mmHg is the expected respiratory compensation, so there is no second respiratory disorder. Delta ratio = (27 − 12)/(24 − 8) ≈ 0.9, consistent with pure ketoacidosis without a large additional metabolic alkalosis or lactic acidosis.
• Corrected sodium = 131 + 0.3 × (12.9 − 5.5) ≈ 133 mmol/L. This is not the dilutional picture of hyperosmolar hyperglycaemic state, and the effective osmolality is far too low for HHS. HHS is excluded by the acidosis, the high ketones and the relatively preserved sodium.
• Glucose is only 12.9 mmol/L because empagliflozin is producing glycosuria that masks hyperglycaemia: euglycaemic DKA. A near-normal glucose must never be used to exclude DKA in anyone taking an SGLT2 inhibitor, and this is the single most testable trap in this scenario.
• Potassium 5.7 mmol/L with peaked T waves does not mean potassium excess. Insulin deficiency, hyperosmolality and AKI shift potassium out of cells while vomiting and urinary losses deplete total body stores; expect a sharp fall once insulin is running.

Precipitant: the dusky, blistered shin with fever and hypotension is the likely source and also raises the question of necrotising fasciitis. That needs a bedside surgical review now, not an MRI that delays theatre.

First hour

• Two large-bore cannulae. 0.9% sodium chloride 1 L over 60 minutes; a balanced crystalloid is an acceptable alternative but the 4 mmol/L potassium in Ringer's lactate is a minor concern at K 5.7. Reassess after each litre - this man has AKI and is at risk of pulmonary oedema.
• Blood and urine cultures, lactate, repeat gas, continuous ECG, urinary catheter for hourly output, daily weights.
• Sepsis bundle within 1 hour: broad-spectrum antibiotics after cultures (for example piperacillin-tazobactam, or ceftriaxone plus flucloxacillin, adjusted to local resistance data) and urgent surgical assessment of the leg. Fluid resuscitation serves both DKA and sepsis; do not under-resuscitate an adult for fear of cerebral oedema.
• Hyperkalaemia: peaked T waves with a normal QRS do not require calcium. Give 10 mL of 10% calcium gluconate only if the QRS widens. Do not give bicarbonate for the acidosis.

Insulin and glucose

• Fixed-rate infusion of soluble insulin 0.1 units/kg/h (about 6 units/h). No bolus in a hypovolaemic patient with AKI.
• Because glucose is already 12.9 mmol/L, run 5% or 10% dextrose alongside the saline from the outset and target glucose 8–12 mmol/L. Otherwise hypoglycaemia will force you to stop insulin before ketogenesis is switched off. The endpoint is ketone clearance and anion gap closure, not a glucose number.
• Hold empagliflozin and metformin.

Monitoring

• Potassium and glucose hourly for 4 hours, then 2-hourly; venous gas, bicarbonate and ketones 2-hourly.
• Add 20–40 mmol/L potassium chloride to fluids once K falls below 5.5 with adequate urine output; if K falls below 3.3, hold insulin and replace potassium first.

Resolution criteria: ketones below 0.6 mmol/L, pH above 7.3, bicarbonate above 18 mmol/L and anion gap 12 mmol/L or less, with the patient eating. A normal glucose is not resolution.

Transition

• Give the subcutaneous basal insulin (glargine, his usual 22 units or an adjusted dose) 1–2 hours before stopping the infusion. Stopping the infusion without basal cover is the classic cause of rebound ketosis.
• Continue the infusion until the biochemical criteria are met even if glucose is normal, adjusting dextrose rather than stopping insulin.
• Metformin can restart once renal function returns towards baseline and he is eating; empagliflozin only after full recovery, with explicit sick-day rules (stop during vomiting, sepsis or reduced intake).

If glucose had been 28 mmol/L the principles are identical, but the practical emphasis shifts: no dextrose initially, add it only when glucose falls below about 14 mmol/L; expect a larger fluid deficit and a more marked fall in potassium; and the euglycaemic trap no longer applies, so the diagnosis is easier. The common error in both versions is stopping the insulin infusion when the glucose normalises while the anion gap is still open.

Also in this patient: watch for over-resuscitation and pulmonary oedema given the AKI, hypokalaemia during treatment, and progression of the leg - repeat surgical review if pain or skin changes worsen despite antibiotics.