Critical Care Nephrology · Two-Week Intensive · Lecture 7 of 9
Acid-Base Disorders & DKA
From physiology to the bedside — and the dialysis connection
40 minutesNephrology FellowsWeek 2
Based on Koyner, Handbook of Critical Care Nephrology (2021) · NTUH Yunlin Branch
Learning objectives
By the end of this session you will be able to…
Read any blood gas with a fixed six-step algorithm and name every disorder present.
Apply Winter's formula, the respiratory rules, the albumin-corrected anion gap and the delta ratio without a calculator.
Work up a high-anion-gap acidosis with lactate, ketones and the osmolal gap — and say when bicarbonate is defensible.
Run a DKA or HHS protocol with correct potassium-first sequencing and gap-based endpoints.
Recognise toxic alcohol and salicylate poisoning and choose antidote, alkalinisation or extracorporeal removal.
Koyner Ch 25, 27, 28, 37 · Ronco Ch 64–71
01
Reading the gas
Every equation on the next ten slides exists to answer one question: how many disorders does this patient have?
Koyner Ch 25 · Ronco Ch 64–65
Acid-base is not a quiz topic in the ICU — it is prognosis
~25%of critically ill patients have a metabolic alkalosis, primary or as compensation
56%mortality with lactic acidosis vs 39% with unmeasured-anion and 29% with hyperchloraemic acidosis
1.21odds ratio for hospital death per 5 mmol/L of HCO₃⁻ above 30 mmol/L
0.4%case fatality in adult DKA (2014) — but ~17% in HHS
Key point
The mechanism of an acidosis carries more prognostic weight than its severity. Two patients with pH 7.20 — one from saline, one from lactate — are not the same patient.
Gunnerson KJ et al. Crit Care. 2006;10:R22 · Libório AB et al. Intensive Care Med. 2015;41:479 · Benoit SR et al. MMWR. 2018;67:362 · Fadini GP et al. Diabetes Res Clin Pract. 2011;94:172 · Ronco Ch 67, 69 · Koyner Ch 37
Henderson–Hasselbalch, and the version you can do in your head
pH = 6.1 + log₁₀ ( [HCO₃⁻] ÷ (0.03 × PaCO₂) ) [H⁺] nmol/L ≈ 24 × PaCO₂ ÷ [HCO₃⁻] — the bedside Henderson equation
[H⁺] nmol/L ≈ 10^(9 − pH) → 7.60 = 25 · 7.50 = 32 · 7.40 = 40 · 7.30 = 50 · 7.20 = 63 · 7.10 = 80 · 7.00 = 100
Normal 24 × 40 ÷ 24 = 40 nmol/L = pH 7.40. Between pH 7.28 and 7.45 the shortcut "[H⁺] rises 1 nmol for every 0.01 fall in pH below 7.40" is accurate enough for the bedside.
Use it two ways: verify that a reported gas is internally consistent, and predict the pH you will get if you change one variable — "change one, solve for the other"
The lower the starting HCO₃⁻ or PaCO₂, the more violently pH moves for the same absolute change. At HCO₃⁻ 8, losing 2 more mmol/L is a catastrophe; at HCO₃⁻ 24 it is noise
A blood-gas HCO₃⁻ is calculated from the measured pH and PaCO₂ — that does not make it invalid. If you believe the pH and the PaCO₂ you have no choice but to believe the HCO₃⁻
Venous vs arterial: HCO₃⁻ ~1 mmol/L higher, PaCO₂ ~4 mmHg higher, pH ~0.03 lower. Never interpret an arterial gas with a venous bicarbonate
Koyner Ch 25 (Rodby) · Treger R et al. Clin J Am Soc Nephrol. 2010;5:390 · Ronco Ch 65
The six-step algorithm — same order, every gas, every time
1
Is there acidaemia or alkalaemia? pH <7.35 or >7.45. A normal pH with an abnormal HCO₃⁻ or PaCO₂ means two primary disorders cancelling each other — never "no disorder".
2
Which variable moved with the pH? That one is primary. HCO₃⁻ down with acidaemia → metabolic acidosis; PaCO₂ up with acidaemia → respiratory acidosis.
3
Is compensation appropriate? Apply the rule for that disorder. Over- or under-shoot outside the stated band is a second disorder — compensation is a prediction, not an excuse.
4
Anion gap, corrected for albumin — on every gas, even when the pH is normal and even when the primary problem is respiratory.
5
Delta gap / delta ratio — does the rise in the gap account for the fall in bicarbonate? If not, a third disorder is hiding.
6
Osmolal gap when the high gap is unexplained by lactate, ketones or uraemia. Send it early — the gap closes as the alcohol is metabolised.
Koyner Ch 25 · Berend K, de Vries APJ, Gans ROB. N Engl J Med. 2014;371:1434 · Ronco Ch 65
Expected compensation — the six coefficients, memorised
Metabolic acidosis expected PaCO₂ = 1.5 × [HCO₃⁻] + 8 (± 2) — Winter's formula
Metabolic alkalosis expected PaCO₂ = 0.7 × [HCO₃⁻] + 20 (± 5)
Acute respiratory acidosis [HCO₃⁻] = 24 + 1 × (PaCO₂ − 40) ÷ 10
Chronic respiratory acidosis [HCO₃⁻] = 24 + 4 × (PaCO₂ − 40) ÷ 10
Acute respiratory alkalosis [HCO₃⁻] = 24 − 2 × (40 − PaCO₂) ÷ 10 floor ≈ 18
Chronic respiratory alkalosis [HCO₃⁻] = 24 − 5 × (40 − PaCO₂) ÷ 10 floor ≈ 14
Some texts use 3.5 rather than 4 for chronic respiratory acidosis; the difference is 1.5 mmol/L at PaCO₂ 70 and never changes a decision. The equivalent bedside form of the alkalosis rule — PaCO₂ rises ~0.7 mmHg per 1 mmol/L of HCO₃⁻ above 24 — agrees within a few mmHg.
Key point
Compensation is fast for metabolic disorders (minutes, limited only by CSF equilibration) and slow for respiratory ones — hours to excrete bicarbonate in respiratory alkalosis, but days to generate new bicarbonate in respiratory acidosis. Timing tells you whether "acute" or "chronic" is the right rule.
Koyner Ch 25, Table 25.2 · Ronco Ch 65, 69 · Berend K et al. N Engl J Med. 2014;371:1434
Over- or under-shoot means a second disorder — three gases
Two disorders: metabolic acidosis plus respiratory acidosis. The patient is tiring — this is an intubation decision
Common pitfall
Compensation never returns the pH to normal, and it never overshoots. A "compensated" gas with a normal pH but abnormal HCO₃⁻ and PaCO₂ is two primary disorders, not one well-compensated one.
Koyner Ch 25 · Ronco Ch 65, 70
The anion gap — and why it lies in the ICU
AG = Na⁺ − (Cl⁻ + HCO₃⁻) normal 8 ± 4 mEq/L (12 ± 4 if K⁺ is included)
AGcorrected = AG + 2.5 × (4.4 − albumin g/dL)Albumin and phosphate are the weak acids that normally fill the gap. An ICU albumin of 2.0 g/dL lowers the expected gap by ~6 mEq/L — enough to hide a clinically important lactic acidosis behind a "normal" AG of 12.
What lowers the gap without any acid-base disease
Hypoalbuminaemia — the dominant ICU cause, and the one you must correct for
Unmeasured cations: severe hypercalcaemia, hypermagnesaemia, lithium
Cationic paraproteins (IgG myeloma)
Bromide and iodide interference, and chloride over-measurement in high salicylate levels
Discipline points
Use one convention (with or without K⁺) and stay in it — critically ill patients run K⁺ <3 or >6 often enough to matter
The AG is a poor screening test for lactate below 4 mmol/L — measure lactate, never estimate it
Do not over-read a gap of 14–16; the reference range genuinely varies by analyser and laboratory
In DKA the gap, not the glucose, is the disease marker — track it
Kraut JA, Madias NE. Clin J Am Soc Nephrol. 2007;2:162 · Figge J et al. Crit Care Med. 1998;26:1807 · Koyner Ch 25 · Ronco Ch 64
The delta gap: the step that finds the third disorder
ΔAG = AGcorrected − 10 ΔHCO₃⁻ = 24 − [HCO₃⁻]
Delta ratio = ΔAG ÷ ΔHCO₃⁻Interpretation bands: <0.8 · 0.8–1.6 · >1.6. Koyner's equivalent rule of thumb: ΔAG > 1.5 × ΔHCO₃⁻ means a hidden metabolic alkalosis — the 1.5 allows for the intracellular buffering that blunts the bicarbonate fall in lactic acidosis.
Ratio
Meaning
Classic clinical pairing
< 0.8
Bicarbonate fell more than the gap rose → a second, normal-gap acidosis is also present
Lactic acidosis plus diarrhoea; DKA resuscitated with several litres of 0.9% saline
Bicarbonate fell less than the gap rose → a metabolic alkalosis (or chronic respiratory acidosis) is co-existing
DKA with vomiting or NG suction; lactic acidosis in a diuretic-treated patient
Key point
If ΔHCO₃⁻ is zero or negative — a high anion gap with a normal or raised bicarbonate — the patient has a metabolic acidosis and a metabolic alkalosis, and the acidosis is completely invisible on the bicarbonate. The delta gap is the only step that finds it.
Koyner Ch 25, Table 25.3 · Ronco Ch 64 · Berend K et al. N Engl J Med. 2014;371:1434
The osmolal gap: a time-limited window on the toxic alcohols
Calculated osmolality = 2 × Na⁺ + glucose/18 + BUN/2.8 (+ ethanol/4.6 if present)
Osmolal gap = measured (freezing-point) osmolality − calculated SI: 2 × Na⁺ + glucose + urea (all mmol/L) + 1.25 × ethanol. A "normal" gap spans roughly −9 to +19, so a gap of 12 proves nothing. A gap >30–40 is the number that should make you reach for fomepizole.
The gap reflects the parent alcohol; the anion gap reflects its acid metabolite. Early = big osmolal gap, small anion gap. Late = the reverse. A patient can present in the crossover with neither impressively raised
Measure osmolality by freezing-point depression — vapour-pressure osmometers do not detect volatile alcohols and will read falsely normal
Other unmeasured osmoles: mannitol, glycerol, sorbitol, isopropanol (ketosis without acidosis), and propylene glycol — the vehicle in IV lorazepam, diazepam, phenobarbital and phenytoin infusions
Lactic acidosis and ketoacidosis raise the gap only trivially — do not use a modest gap to invoke a toxin
The trap
A normal osmolal gap never excludes toxic alcohol ingestion. If the history or the unexplained gap fits, treat and send levels — do not let a reassuring gap close the case.
Koyner Ch 27 · Ronco Ch 64, 71 · Aabakken L et al. Hum Exp Toxicol. 1994 (osmolal gap range)
Stewart: three independent variables, and where it genuinely helps
SIDapparent = (Na⁺ + K⁺ + Ca²⁺ + Mg²⁺) − (Cl⁻ + lactate⁻) ≈ 40 mEq/L normally
ATOT = the weak acids — albumin and phosphate
SIG = SIDapparent − SIDeffective ≈ 0 in health, 5 ± 5 in critical illness
Stewart's claim: pH is determined by exactly three independent variables — SID, ATOT and PaCO₂. HCO₃⁻ and H⁺ are dependent outputs, not levers.
Where it explains what Henderson–Hasselbalch cannot
Saline acidosis — 0.9% NaCl has a SID of 0, so infusing it drags plasma SID down from 40. Nothing was "added"; chloride simply crowded out bicarbonate (Lecture 03)
Hypoalbuminaemic alkalosis — losing ATOT is alkalinising, and routinely masks a coexisting acidosis
Unmeasured anions of sepsis — a raised SIG with normal lactate; its magnitude tracks mortality
What it does not do
It does not change a single treatment decision that a corrected anion gap plus a lactate would have missed
SIDeffective needs ionised Ca²⁺, Mg²⁺, phosphate and albumin — rarely all available at 3 a.m.
No trial has shown Stewart-guided management improves outcome
The practical synthesis
Keep Henderson–Hasselbalch as the working language. Borrow three Stewart habits: correct the gap for albumin, look at the chloride before calling an acidosis unexplained, and read the fluid you are prescribing as an acid-base intervention.
1 pH 7.28 → acidaemia. Henderson check: 24 × 24 ÷ 11 = 52 nmol/L → pH 7.28 — the gas is internally consistent 2 HCO₃⁻ 11 is low and moves with the pH → primary metabolic acidosis 3 Winter's: 1.5 × 11 + 8 = 24.5 ± 2 → 22.5–26.5. Measured PaCO₂ 24 → respiratory compensation is appropriate; no respiratory disorder 4 AG = 138 − (92 + 11) = 35. Albumin 4.2 → correction +0.5 → corrected AG ≈ 35. Clearly raised 5 ΔAG = 35 − 10 = 25 · ΔHCO₃⁻ = 24 − 11 = 13 · ratio = 25 ÷ 13 = 1.9 → >1.6 → a metabolic alkalosis is also present 6 Osmolal gap not needed — β-OHB 6.1 mmol/L accounts for the unmeasured anion
Final diagnosis — three findings, one gas
High-anion-gap ketoacidosis + metabolic alkalosis from four days of vomiting, with appropriate respiratory compensation. Implied pre-illness bicarbonate ≈ 24 + (25 − 13) = 36 mmol/L. The pH of 7.28 badly understates how much acid this patient is carrying — treat the gap, and expect the bicarbonate to fall further as the alkalosis is corrected with chloride.
Koyner Ch 25, 37 · Ronco Ch 64
02
Metabolic acidosis
The gap tells you which list to open; the list tells you which treatment actually works.
B3 inborn errors of metabolism, occasionally unmasked in adults
Lactate is a stress signal, not an oxygen probe
Adrenaline-driven aerobic glycolysis through β₂ receptors raises lactate with entirely adequate oxygen delivery — which is why adrenaline infusions, salbutamol and pure catecholamine stress all raise it. Persistent hyperlactataemia after adequate resuscitation is more often continuing cytokine and β₂ drive than continuing dysoxia. Conversely, hypoperfusion can be present with a normal lactate. Trend it as a prognostic marker; do not treat it as a perfusion readout.
Ronco Ch 67 · Jones AE et al. JAMA. 2010;303:739 · Jansen TC et al. Am J Respir Crit Care Med. 2010;182:752 · Koyner Ch 25
The lactic acidoses you will actually be consulted about
Cause
Mechanism
Clue
Action
Metformin (MALA)
Inhibits mitochondrial complex I; >90% renally cleared, so accumulates in AKI
Lactate often >10–15 with pH <7.1 and a modest illness
Stop drug; haemodialysis for shock, obtundation, pH <7.0 or lactate >20 (EXTRIP)
Thiamine deficiency
Pyruvate dehydrogenase cannot run without thiamine
Stop the infusion immediately; supportive care, often KRT
Linezolid
Inhibits mitochondrial protein synthesis
Usually >1–2 weeks of therapy; cytopenias alongside
Stop drug; acidosis is reversible
β₂-agonists / adrenaline
Aerobic glycolysis, not dysoxia
Severe asthma on continuous salbutamol; adrenaline-supported shock
Recognise it — do not escalate resuscitation to chase this lactate
Malignancy
Tumour glycolysis ± hepatic infiltration
Bulky lymphoma or leukaemia; lactate falls only with tumour response
Treat the cancer (Lecture 09 covers tumour lysis)
Ronco Ch 67 · Koyner Ch 25, 28 · Mirrakhimov AE et al. Crit Care Res Pract. 2015;2015:260385 · EXTRIP metformin recommendations, 2015
Bicarbonate for severe acidaemia: what BICAR-ICU actually showed
BICAR-ICU · Lancet 2018
389 ICU patients with pH ≤7.20, HCO₃⁻ ≤20, PaCO₂ ≤45 and SOFA ≥4 or lactate ≥2, randomised to 4.2% sodium bicarbonate targeting pH >7.30 vs no alkali. Primary composite (death by day 28 or ≥1 organ failure at day 7): 66% vs 71%, p = 0.24 — no overall benefit. Day-28 mortality 45% vs 55%, p = 0.09.
BICAR-ICU · pre-specified AKIN stage 2–3 stratum
In the 182 patients with AKIN 2–3, the primary outcome fell to 70% vs 82% (p = 0.046) and day-28 mortality to 46% vs 63% (p = 0.028). Across the whole trial, bicarbonate reduced kidney replacement therapy by day 28: 35% vs 52% (p < 0.001). Costs: metabolic alkalosis, hypocalcaemia, hypernatraemia.
Pitfall — the bicarbonate reflex
Bicarbonate is not a vasopressor and it does not treat a cause. It generates CO₂ (worse if minute ventilation is fixed), can worsen intracellular and CSF pH, shifts the oxyhaemoglobin curve left, drops ionised calcium, and each 50 mmol ampoule carries 2000 mOsm/L of sodium load. In DKA specifically the evidence trends towards harm — prolonged ketosis and greater potassium requirement.
Jaber S et al. Lancet. 2018;392:31 · Chua HR, Schneider A, Bellomo R. Ann Intensive Care. 2011;1:23 · Koyner Ch 25
If you do give base: the arithmetic, and the isovolaemic route
Textbook deficit = (24 − [HCO₃⁻]) × 0.5–0.8 × weight (kg) — do not replace it
Bedside rule: in an 80 kg adult, one 50 mmol ampoule raises [HCO₃⁻] by ≈ 1 mmol/L
Isotonic drip: 3 ampoules (150 mmol) in 1 L of D5W ≈ 130 mmol/L NaHCO₃ — ~770 mL delivers 100 mmol
Two ampoules is usually enough to get a patient out of the acidaemia danger zone. Aim to buy time, not to normalise a number.
When base is defensible
pH ≤7.10–7.20 with haemodynamic compromise, while the cause is being fixed
Severe acidaemia with AKIN 2–3 — the one BICAR-ICU stratum that benefited
Hyperkalaemia with acidaemia; tricyclic and other sodium-channel-blocker toxicity; salicylate (see later)
Normal-gap acidosis with ongoing bicarbonate loss — here you are genuinely replacing what is being lost
Why CKRT is often the better alkali route
Severe lactic acidosis can need ≥100 mmol NaHCO₃ per hour — close to 1 L/h of isotonic drip. That volume is not survivable for long
CVVH with 35 mmol/L bicarbonate replacement fluid in a patient whose HCO₃⁻ is 5: each litre exchanged nets 35 − 5 = 30 mmol of base
At 4–5 L/h that is 120–150 mmol/h delivered isovolaemically and isosmotically
Modality choice, dose and citrate issues are Lectures 08 and 09
Koyner Ch 25 (Rodby) · Kraut JA, Kurtz I. Am J Kidney Dis. 2001;38:703 · Jaber S et al. Lancet. 2018;392:31
Ketoacidosis beyond diabetes
Alcoholic ketoacidosis
Malnourished drinker, binge then abstinence, vomiting and abdominal pain. Glucose low-normal; β-OHB dominant so the urine nitroprusside test under-reads. Often a triple disorder: ketoacidosis + vomiting alkalosis + respiratory alkalosis.
Treat: saline with dextrose, thiamine before glucose, magnesium, phosphate. Insulin is not needed — endogenous insulin resumes with the glucose load.
Starvation ketosis
Mild, gap rarely above 15–18, bicarbonate rarely below 18. Common in prolonged fasting, hyperemesis, post-operative patients and on glucose-free CKRT solutions (30–60 g of glucose lost per day).
Treat: feed. It is not a reason for an insulin infusion, and a rising gap on CKRT should prompt a look at the fluid you chose.
Euglycaemic DKA — the ICU trap
Ketoacidosis with glucose <200–250 mg/dL. SGLT2 inhibitors are the modern cause (≈4.9 vs 2.3 DKA events per 1000 person-years vs DPP4 inhibitors); also pregnancy, fasting, vomiting, pancreatitis, prior insulin dosing.
Treat: insulin and dextrose together from the start. Gaps are often >30 and the infusion runs far longer than in classic DKA. Hold the SGLT2 inhibitor before major surgery.
Koyner Ch 25, 37 · Fralick M, Schneeweiss S, Patorno E. N Engl J Med. 2017;376:2300 · Coutrot M et al. Intensive Care Med. 2018;44:1185 · Ronco Ch 66
Normal-anion-gap acidosis: two urine tests decide it
Urine anion gap (UAG) = UNa + UK − UCl — a proxy for urinary NH₄⁺
Urine osmolal gap = measured Uosm − [ 2 × (UNa + UK) + urea/2.8 + glucose/18 ] · urinary NH₄⁺ ≈ half the gap
A negative UAG (−20 to −50) means the kidney is excreting NH₄⁺ with chloride — an appropriate response, so the loss is extrarenal. A positive UAG means it is not — a renal acidification defect.
Negative UAG — extrarenal loss
Diarrhoea, high-output ileostomy, pancreatic or biliary fistula
Large-volume 0.9% saline — the most common cause on any ICU round (Lecture 03)
Proximal (type 2) RTA also gives a negative UAG once bicarbonaturia stops
Positive UAG — renal defect
Distal (type 1) and hyperkalaemic (type 4) RTA
Advanced CKD and AKI with reduced ammoniagenesis
When the UAG lies: any non-chloride anion excreted with NH₄⁺ — hippurate (toluene sniffing), β-OHB, 5-oxoproline — makes the UAG falsely positive. Use the urine osmolal gap instead; >400 mOsm/kg confirms adequate ammoniagenesis
Ronco Ch 55, 68 · Koyner Ch 25 · Brunner R et al. Crit Care. 2015;19:148
Renal tubular acidosis in the critically ill
Type
Lesion
Serum K⁺
Urine pH on acid load
Serum HCO₃⁻
ICU causes
Treatment
1 · distal
α-intercalated cell cannot maintain the H⁺ gradient — gradient, polarity, backleak or ammoniagenesis defect
Low
>5.5 (inappropriately alkaline)
Can fall <10
Amphotericin B (backleak), Sjögren, SLE, lithium, toluene, obstruction
Alkali as potassium citrate 1–3 mEq/kg/day; corrects hypocitraturia and nephrocalcinosis risk
2 · proximal
Failure to reclaim filtered HCO₃⁻ (NBC-1, NHE3, carbonic anhydrase)
Low, and falls further on alkali
<5.5 once serum HCO₃⁻ is below the reduced threshold
Plateaus around 15–18
Fanconi from ifosfamide, cisplatin, tenofovir, aminoglycosides, valproate, myeloma; acetazolamide, topiramate
Large alkali doses (10–15 mEq/kg/day) plus generous K⁺; treat the cause first
4 · hyperkalaemic
Voltage defect from aldosterone deficiency or resistance; hyperkalaemia itself suppresses ammoniagenesis
Stop the offending drug, low-K⁺ diet, potassium binder, loop diuretic; fludrocortisone only in true deficiency
Ronco Ch 68 (Corey & Eckstein) · Koyner Ch 25 · Soleimani M, Rastegar A. Am J Kidney Dis. 2016 core curriculum
03
Metabolic alkalosis
The commonest acid-base disorder nobody is consulted about — until the patient cannot be weaned.
Koyner Ch 25 · Ronco Ch 69
Classify it in one step: urine chloride
Chloride-responsive · urine Cl⁻ <20 mmol/L
Gastric losses — vomiting, high-output NG suction (loss of low-SID fluid raises plasma SID)
Diuretics, after the drug has worn off — the maintenance phase of chloride depletion
Contraction / chloride-depletion alkalosis — properly called chloride depletion, since chloride replacement corrects it despite persistently high aldosterone
Post-hypercapnic alkalosis — days of renal bicarbonate generation, then the ventilator normalises PaCO₂ in an hour and leaves the bicarbonate behind
Normotensive: Bartter and Gitelman syndromes, severe magnesium or potassium depletion, cisplatin tubulopathy
Alkali load: citrate from massive transfusion or regional citrate anticoagulation, sodium bicarbonate, milk–alkali
Active loop or thiazide diuretic raises urine Cl⁻ >20 even in a chloride-responsive alkalosis — the test is uninterpretable until the drug is off
Ronco Ch 69 · Luke RG, Galla JH. J Am Soc Nephrol. 2012;23:204 · Koyner Ch 25
Treating alkalosis: chloride first, everything else after
Chloride deficit (mEq) = 0.2 × weight (kg) × (Cl⁻desired − Cl⁻measured)
HCl (mEq) = 0.5 × weight (kg) × ([HCO₃⁻]actual − [HCO₃⁻]desired)Both "spaces" are approximations and both equations carry appreciable error — treat them as a starting dose, then re-measure.
1
Give chloride with the right cation — 0.9% NaCl if volume-depleted, KCl if potassium-depleted (which is nearly always). Replace magnesium or the potassium will not stay.
2
Stop generating it — hold or reduce the loop/thiazide diuretic, add a proton-pump inhibitor to cut gastric acid loss in NG suction, reconsider the citrate load.
3
Acetazolamide 250–500 mg IV or PO every 12 h when the patient is volume-overloaded and cannot take chloride — it makes the kidney excrete sodium relative to chloride, lowering plasma SID. Expect hypokalaemia and monitor it. DIABOLO (JAMA 2016) found it did not significantly shorten ventilation in COPD, so use it for the biochemistry, not for the outcome.
4
Add a potassium-sparing agent — spironolactone or amiloride — when mineralocorticoid excess is the mechanism.
5
Rescue for pH ≥7.60 with arrhythmia: dilute HCl through a central line (vein damage is real and the product is hard to obtain), or KRT — haemodialysis against a low-bicarbonate bath, or CVVHD/CVVH with low-bicarbonate fluid. Watch K⁺ and ionised calcium closely; alkalaemia already lowers ionised calcium.
Ronco Ch 69 · Koyner Ch 25 · Faisy C et al. JAMA. 2016;315:480 (DIABOLO)
04
Respiratory acid-base
The ventilator is a dialysis machine for CO₂ — and the fastest way to change any pH on the unit.
Koyner Ch 25 · Ronco Ch 70
Causes on both sides of the PaCO₂
Respiratory acidosis — the pump fails
Drive: sedatives and opioids, brainstem stroke or mass, obesity–hypoventilation, hypothyroidism, and metabolic alkalosis itself suppressing respiration
Pump: Guillain–Barré, myasthenia, high cord injury, ICU-acquired weakness, residual neuromuscular blockade, kyphoscoliosis — and hypophosphataemia or hypokalaemia weakening the diaphragm (Lecture 06)
Lung and circuit: COPD, severe asthma, ARDS, pneumonia, pneumothorax, auto-PEEP, increased dead space, exhausted soda lime
Hypoventilation may be relative: a fixed set minute ventilation that cannot rise when fever raises CO₂ production
Respiratory alkalosis — drive is too high
Hypoxaemia: pneumonia, pulmonary embolism, oedema, altitude. Always exclude it first
Direct centre stimulation: sepsis (often the earliest sign, before hypotension), pain, fever, CNS lesions, salicylate, progesterone in pregnancy
Iatrogenic: over-ventilation on controlled modes, excessive extracorporeal CO₂ removal
Consequences: cerebral vasoconstriction, falling ionised calcium with tetany, left-shifted oxyhaemoglobin curve, hypokalaemia and hypophosphataemia from intracellular shift
Permissive hypercapnia — and the nephrology reading of a low bicarbonate
Lung-protective ventilation at 6 mL/kg predicted body weight often means PaCO₂ 50–80 with pH 7.20–7.30. That is accepted, not treated — hypercapnia may itself be protective, and giving bicarbonate to a patient with fixed minute ventilation simply generates more CO₂. Avoid it with raised intracranial pressure; be cautious in severe pulmonary hypertension and RV failure; extracorporeal CO₂ removal is the escape route. Mirror image: a chronic respiratory alkalosis lowers HCO₃⁻ to 14–18 by renal excretion — appropriate compensation, not an acidosis, and not a reason for alkali. If HCO₃⁻ sits below the chronic prediction, send a salicylate level.
Common pitfall
Normalising PaCO₂ quickly in a chronically hypercapnic patient leaves the renally generated bicarbonate stranded → post-hypercapnic metabolic alkalosis, with suppressed respiratory drive and a failed wean. Come down slowly, and give chloride.
Ketogenesisβ-oxidation to acetyl-CoA, then β-hydroxybutyrate and acetoacetate — strong acids
→
04
Osmotic diuresisLoss of water, Na⁺, K⁺, phosphate and magnesium — the deficits you must replace
In HHS the residual insulin is enough to suppress ketogenesis but not hyperglycaemia — so the tempo is days, not hours, and the dominant lesion is osmolality rather than acid.
Diagnostic thresholds
DKA: ketonaemia/ketonuria + pH <7.30 + HCO₃⁻ ≤18 + hyperglycaemia (variable — may be <250)
Urine ketones: sensitivity 99%, negative predictive value 100%. An anion gap >16 has 92% sensitivity
Measure β-hydroxybutyrate: it dominates early and the nitroprusside strip does not detect it — a "worsening" ketone strip during treatment can mean improvement
Venous gases are as accurate as arterial here — stop arterial sampling for this indication
Stupor with osmolality <320 → look for another diagnosis
Koyner Ch 37 (Topf et al.) · Kitabchi AE et al. Diabetes Care. 2009;32:1335 · Schwab TM et al. Ann Emerg Med. 1999;34:342
The protocol: volume, potassium, insulin — in that order
01
VolumeDeficit up to 5 L in DKA. Moderate disease: 20 mL/kg bolus, then 500 mL/h for 4 h, then 250 mL/h. Severe: wide open until perfused. Isotonic crystalloid; balanced solutions have shown no advantage in RCTs, and lactated Ringer's took longer to bring glucose below 250 (410 vs 300 min).
→
02
PotassiumTotal body deficit 3–5 mmol/kg despite a high presenting serum K⁺. Start 20–30 mEq KCl per litre once K⁺ falls below 5.2 and urine output is confirmed. If K⁺ <3.3, hold insulin until it is corrected.
→
03
Insulin0.1 U/kg bolus then 0.1 U/kg/h — or no bolus with 0.14 U/kg/h; a retrospective comparison found no difference in hypoglycaemia, rate of glucose fall or length of stay. Target a glucose fall of 50–75 mg/dL/h.
→
04
DextroseAdd 5–10% dextrose when glucose reaches 200–250 mg/dL and cut insulin to 0.02–0.05 U/kg/h. Hold glucose at 150–200 and keep the insulin running — you are treating ketosis, not sugar.
Pitfall — insulin before potassium
Insulin drives potassium intracellularly within minutes. Starting it at K⁺ 3.1 converts a survivable metabolic emergency into a cardiac arrest. Equally, in a profoundly volume-depleted patient — usually HHS — insulin given before fluid pulls water back into cells and can precipitate cardiovascular collapse. Fluid, then potassium, then insulin.
Koyner Ch 37 · Kitabchi AE et al. Diabetes Care. 2009;32:1335 · Van Zyl DG et al. QJM. 2011;105:337 · Williams V et al. Crit Care. 2020;24:1 (SPinK)
Ending the infusion: the gap decides, not the glucose
Resolution criteria — glucose <200 and two of three
pH >7.30
Serum HCO₃⁻ >15 mmol/L
Anion gap <12
Hyperglycaemia usually corrects within 6 h; ketosis typically persists ~12 h. The insulin must outlast the sugar
Transition to subcutaneous insulin
Give the first basal subcutaneous dose 1–2 hours before stopping the infusion — IV regular insulin has a half-life of minutes and an unbridged gap re-opens the ketosis
In HHS continue IV insulin until osmolality and mental status normalise, and overlap by 2 h
Insulin-naive patients: roughly 0.5–0.6 U/kg/day total, split basal/prandial; restart the home regimen if it was adequate
Do not transition overnight or while the patient is still not eating
Why the bicarbonate lags
After the gap closes, most patients are left with a hyperchloraemic normal-gap acidosis — ketoanions were excreted in the urine (losing potential bicarbonate) and several litres of 0.9% saline finished the job. It is self-limited, needs no treatment, and must not be read as failed treatment. This is exactly the delta-ratio <0.8 pattern from Section 01.
Koyner Ch 37 · Kitabchi AE et al. Diabetes Care. 2009;32:1335 · Adrogué HJ et al. JAMA. 1989;262:2108
Complications to anticipate before they happen
Hypokalaemia
The commonest preventable death in DKA. Deficit 3–5 mmol/kg; check K⁺ every 2 h for the first 6 h. Hold insulin below 3.3.
Hypophosphataemia
High at presentation, then insulin drives it in. Deficit ~1 mmol/kg. Randomised data show no clinical benefit from routine replacement and real harm (hypocalcaemia, hypomagnesaemia) — replace only below 1 mmol/L or with respiratory or cardiac weakness (Lecture 06).
Hypoglycaemia
From failing to add dextrose at 200–250 while the infusion runs. Hourly glucose, standing dextrose order.
Cerebral oedema
Overwhelmingly paediatric (<1% of episodes but a large share of childhood DKA deaths). Avoid rapid osmolality swings; suspect it with headache, bradycardia and falling consciousness after initial improvement.
Hyperchloraemic acidosis
From the saline, not from the disease. Expected, self-limited, no treatment.
Thrombosis & AKI
Both DKA and HHS are prothrombotic — give prophylactic (not therapeutic) anticoagulation. AKI occurs in ~50% of ICU DKA admissions; half recovers within 24 h with fluid, ~3% need acute dialysis, and survivors lose GFR faster afterwards (Lecture 01).
HHS and the two situations that break the protocol
Feature
DKA
HHS
Tempo
Hours to 1–2 days
Days to weeks of osmotic diuresis
Glucose
Variable; may be <250 (euglycaemic)
Usually >600, often >650
pH / HCO₃⁻ / ketones
<7.30 / ≤18 / strongly positive
>7.30 / >18 / minimal
Osmolality
Usually <320
>320, typically >350; correlates linearly with conscious level
Fluid deficit
Up to ~5 L
9–12 L — replace over 24–48 h, not in the first hour
Insulin
Start once K⁺ is secured
Wait until volume is at least partly restored; bolus rarely needed
Mortality
<1% overall, >5% in the elderly
~17% — largely the precipitant and the age
Sodium: use Katz, but only for orientation
Adjusted Na⁺ = measured Na⁺ + 0.016 × (glucose − 100). If it is high or high-normal, switch to 0.45% NaCl. Never use the adjusted sodium to compute the anion gap — it tells you where the sodium will land after correction, nothing more (Lecture 05).
Dialysis-dependent ESKD is a different disease
No osmotic diuresis → no volume depletion, no large K⁺ or phosphate deficit, and presenting glucose is higher (~836 vs 659 mg/dL). These patients are often fluid overloaded from intracellular-to-extracellular water shift. Treat with insulin largely alone (0.05–0.07 U/kg/h), give 250–500 mL boluses only if genuinely hypovolaemic, and reconsider every standing potassium order.
Koyner Ch 37 · Katz MA. N Engl J Med. 1973;289:843 · Schaapveld-Davis CM et al. Clin Diabetes. 2017;35:202 · Fadini GP et al. Diabetes Res Clin Pract. 2011;94:172
06
Toxic alcohols & salicylate
Two poisonings where the nephrologist's phone call changes the outcome, and where waiting for the level is the error.
Koyner Ch 27–28 · Ronco Ch 71, 98–99
Methanol vs ethylene glycol: same mechanism, different organ
Target organ: the kidney — calcium oxalate deposition, AKI, and needle- or envelope-shaped crystals on urine microscopy
Hypocalcaemia with a prolonged QT as calcium is consumed; urine may fluoresce under Wood's lamp if the product contained fluorescein (unreliable)
Give thiamine 100 mg and pyridoxine 50–100 mg to shunt glyoxylate away from oxalate
Read the two gaps as a clock
Osmolal gap high with a normal anion gap = early, parent alcohol still unmetabolised — the best moment to block ADH and the worst moment to be reassured. Anion gap high with a closing osmolal gap = late, the metabolite is already made. Ethanol co-ingestion delays everything by competing for ADH, and must be added back into the calculated osmolality.
Fomepizole, dialysis, and the other dialysable toxins
1
Block ADH now — do not wait for a level. Fomepizole 15 mg/kg IV load, then 10 mg/kg every 12 h for doses 2–4, then 15 mg/kg every 12 h (it induces its own metabolism via CYP2E1). Indicated for suspected ingestion, methanol >20 mg/dL, or a confirmed ingestion with an osmolal gap >10 — or strong suspicion plus two of: pH <7.3, HCO₃⁻ <20, osmolal gap >10.
2
Ethanol is the fallback where fomepizole is unavailable: 600 mg/kg load then 66–154 mg/kg/h, targeting a blood ethanol of ~100 mg/dL, checked every 4–6 h. It needs an infusion, sedates the patient, and causes hypoglycaemia — fomepizole is preferred wherever it exists.
3
Haemodialysis (EXTRIP, methanol): coma, seizures, new visual deficits; arterial pH ≤7.15; persistent acidosis despite antidote; anion gap >24; methanol >70 mg/dL on fomepizole, >60 on ethanol, or >50 with no ADH blocker. Same principles for ethylene glycol; isopropanol needs supportive care only.
4
Intermittent HD, not CKRT. These are small, water-soluble, minimally protein-bound molecules — IHD clears them several-fold faster and is explicitly preferred. Use CKRT only if haemodynamics forbid IHD, and expect a much longer run. Continue the ADH blocker throughout and redose fomepizole every 4 h during dialysis — it is dialysed out. Stop when methanol <20 mg/dL with clinical improvement.
5
Fomepizole alone may be enough. With early blockade, preserved kidney function and no significant acidosis or visual signs, many ethylene glycol ingestions never need dialysis. The counter-argument for methanol: its half-life under ADH blockade is 40–50 hours, so early dialysis shortens both fomepizole exposure and hospital stay.
6
Others worth knowing now, covered fully in Lecture 09: lithium (HD if level >5.0 mmol/L, confusion or seizures; expect rebound and re-check at 12 h), valproate, metformin (shock, obtundation, pH <7.0 or lactate >20), barbiturates, and salicylate below.
Koyner Ch 27–28 · Roberts DM et al. Crit Care Med. 2015;43:461 (EXTRIP methanol) · Decker BS et al. EXTRIP lithium, 2015 · Barceloux DG et al. J Toxicol Clin Toxicol. 2002;40:415
Salicylate: read the gas, and think twice before intubating
Typical gas: pH 7.44 · PaCO₂ 18 · HCO₃⁻ 12 · Na⁺ 140 · Cl⁻ 102
Henderson: 24 × 18 ÷ 12 = 36 nmol/L → pH 7.44 ✓ · AG = 140 − (102 + 12) = 26
If respiratory alkalosis were primary: chronic rule predicts HCO₃⁻ = 24 − 5 × 2.2 = 13; acute predicts 19.6 — measured 12 is below both
If metabolic acidosis were primary: Winter's predicts PaCO₂ = 1.5 × 12 + 8 = 26 ± 2 — measured 18 is far below
Both routes give the same answer: a primary respiratory alkalosis and a primary high-anion-gap metabolic acidosis. An alkalaemic pH with an anion gap of 26 is salicylate until proved otherwise.
Mechanism and presentation
Direct medullary respiratory stimulation → tachypnoea and hyperpnoea; mitochondrial uncoupling → lactate, ketones and hyperpyrexia
Tinnitus, nausea, vomiting, agitation, then coma; non-cardiogenic pulmonary oedema in severe cases
Salicylate falsely raises chloride on some analysers, spuriously lowering the calculated gap
Chronic toxicity in elderly patients is missed for days — levels are lower and the picture looks like sepsis or delirium
Why intubation can kill
The patient's minute ventilation — often 30–40 L/min — is what is keeping the pH alkalaemic and salicylate ionised and out of the brain. Sedating and ventilating at a conventional rate collapses that compensation, the pH falls, non-ionised salicylate floods the CNS, and patients arrest peri-intubation. If the airway must be secured: pre-treat with bicarbonate, use a rapid-onset agent, keep apnoea time minimal, and set a very high minute ventilation to match or exceed the patient's own. Better still — dialyse rather than intubate.
Alkalinise: 1–2 mEq/kg bicarbonate bolus then an isotonic bicarbonate infusion, target urine pH 7.5–8.0 and serum pH 7.45–7.55. Ion-trapping cuts tubular reabsorption and keeps salicylate out of the CNS. It fails without aggressive potassium repletion — hypokalaemia forces the kidney to reabsorb K⁺ in exchange for H⁺ and the urine will not alkalinise
Give dextrose even with a normal blood glucose — CSF glucose can be low while serum is normal
Haemodialysis (EXTRIP): altered mental status; hypoxaemia needing oxygen; level >100 mg/dL (>90 with kidney impairment). If supportive care is failing, also consider at >90 mg/dL, >80 with kidney impairment, or systemic pH ≤7.20. Run at least 6 h or until symptoms improve and the level is <19 mg/dL, and continue bicarbonate between sessions
Juurlink DN et al. Ann Emerg Med. 2015;66:165 (EXTRIP salicylate) · Koyner Ch 28 · Ronco Ch 71
07
Cases, poll & wrap-up
Three consults that all begin with the same sentence — "the bicarbonate is abnormal" — and end in three different places.
Koyner Ch 25, 27–28, 37 · Ronco Ch 64–69
Case 1 · 44 M, found confused in a garage, no history available
Run all six steps. What is the diagnosis, and what do you order in the next ten minutes?
Think 60 seconds · full working on the next slide
Case 1 — worked gas 2
1 pH 7.16 → acidaemia. Henderson: 24 × 20 ÷ 7 = 69 nmol/L → pH 7.16 ✓ 2 HCO₃⁻ 7 moves with the pH → primary metabolic acidosis 3 Winter's: 1.5 × 7 + 8 = 18.5 ± 2 → 16.5–20.5. Measured 20 → appropriate; no respiratory disorder 4 AG = 142 − (106 + 7) = 29. Albumin 2.0 → + 2.5 × (4.4 − 2.0) = +6 → corrected AG = 35 5 ΔAG = 25 · ΔHCO₃⁻ = 17 · ratio = 1.5 → a single high-gap acidosis, no hidden alkalosis or normal-gap acidosis 6 Calculated osm = 2(142) + 108/18 + 22/2.8 = 284 + 6 + 8 = 298 · measured 341 → osmolal gap = 43
Lactate 3.2 explains ~3 of 25 mEq/L; ketones negative; creatinine and urea do not account for the rest
Diagnosis
Toxic alcohol ingestion — methanol or ethylene glycol — caught in the crossover phase with both gaps still open. Check urine microscopy for oxalate crystals and examine the visual fields and fundi.
Orders now
Fomepizole 15 mg/kg IV immediately — do not wait for levels
Folinic acid, thiamine and pyridoxine (cover both alcohols until the level returns)
Send methanol and ethylene glycol levels; repeat gas and gap hourly
Arrange intermittent haemodialysis: an anion gap of 29 (corrected 35) is above the EXTRIP threshold of 24, and the acidosis is persisting. Redose fomepizole every 4 h on dialysis
Roberts DM et al. Crit Care Med. 2015;43:461 · Koyner Ch 27–28 · Ronco Ch 64, 71
Case 2 · 24 F, type 1 diabetes, insulin pump failure, 12 h of vomiting
Kussmaul respiration, dry mucous membranes, heart rate 128, blood pressure 96/58. The admitting team has written for an insulin infusion and 1 L of 0.9% saline.
The infusion is about to be connected. What do you change, and what do you calculate first?
Hands up — then justify the order of the three drugs
Case 2 — the first hour, and the arithmetic behind it
Henderson: 24 × 17 ÷ 6 = 68 nmol/L → pH 7.17 ✓ · Winter's: 1.5 × 6 + 8 = 17 ± 2 → measured 17, appropriate
AG = 130 − (96 + 6) = 28 · ΔAG = 18 · ΔHCO₃⁻ = 18 · ratio = 1.0 → pure high-gap ketoacidosis
Katz-adjusted Na⁺ = 130 + 0.016 × (520 − 100) = 137 — the hyponatraemia is dilutional, not a sodium problem
The corrected sequence
Hold the insulin. K⁺ 3.1 is below the 3.3 threshold
20 mL/kg isotonic crystalloid over the first hour, then 500 mL/h; confirm urine output
20–30 mEq KCl per litre, recheck K⁺ within 60 min; start insulin only once K⁺ ≥3.3
Then 0.1 U/kg/h (bolus optional); add 5–10% dextrose when glucose reaches 200–250 and cut to 0.02–0.05 U/kg/h
Stop when glucose <200 plus two of: pH >7.30, HCO₃⁻ >15, gap <12 — with subcutaneous basal insulin given 1–2 h beforehand
No bicarbonate at pH 7.17: no benefit above pH 7.1, and a signal of prolonged ketosis and greater potassium requirement
What else to anticipate
Creatinine 1.8 is mostly haemodynamic — about half of DKA-associated AKI resolves within 24 h with fluid
Expect a residual hyperchloraemic acidosis after the gap closes; do not restart insulin for it
Find the precipitant: pump failure here, but screen for infection, myocardial infarction and pancreatitis. A raised lipase is present in a third of DKA cases and is often not pancreatitis
Prophylactic anticoagulation; magnesium and phosphate checked but not reflexly replaced (Lecture 06)
Koyner Ch 37 · Kitabchi AE et al. Diabetes Care. 2009;32:1335 · Chua HR et al. Ann Intensive Care. 2011;1:23 · Orban JC et al. PLoS One. 2014;9:e110925
Case 3 · 68 M, day 5 after oesophagectomy, cannot be weaned
The urine chloride is 45 — does that make this chloride-resistant? What are you going to do?
Think 45 seconds · discuss before the takeaways
Quick poll
Post-operative patient, 6 L of 0.9% saline in 12 hours. pH 7.29 · PaCO₂ 30 · HCO₃⁻ 14 · Na⁺ 140 · Cl⁻ 114 · albumin 4.4 · lactate 1.1.
A. High-anion-gap metabolic acidosis · B. Normal-anion-gap metabolic acidosis with appropriate respiratory compensation · C. Metabolic acidosis plus a primary respiratory alkalosis · D. Primary respiratory alkalosis with renal compensation
Hands up — then name the single number that rules out each wrong answer
Key takeaways
Six steps, every gas: pH → primary disorder → compensation → albumin-corrected gap → delta ratio → osmolal gap. Steps 4 and 5 are the ones that get skipped and the ones that find the extra disorder.
Compensation is a prediction. Outside the band, or a normal pH with abnormal HCO₃⁻ and PaCO₂, means a second primary disorder — not good compensation.
Correct the gap for albumin (+2.5 per g/dL below 4.4) and measure lactate rather than inferring it. Lactate is a stress signal, not an oxygen probe.
Bicarbonate did not improve the primary outcome in BICAR-ICU, but reduced kidney replacement therapy and helped the AKIN 2–3 stratum. In DKA it trends towards harm — reserve it for pH <6.9.
DKA: fluid, then potassium, then insulin. Hold insulin below K⁺ 3.3, add dextrose at glucose 200–250, and end the infusion on the gap, with subcutaneous basal insulin overlapped by 1–2 hours.
An unexplained gap plus an osmolal gap is fomepizole now, before the level returns; an alkalaemic pH with a gap of 26 is salicylate, and intubating that patient without matching their minute ventilation can be fatal.
References & further reading
Where to go deeper
Jaber S, Paugam C, Futier E, et al. Sodium bicarbonate therapy for patients with severe metabolic acidaemia in the intensive care unit (BICAR-ICU): a multicentre, open-label, randomised controlled, phase 3 trial. Lancet. 2018;392(10141):31–40.
Kitabchi AE, Umpierrez GE, Miles JM, Fisher JN. Hyperglycemic crises in adult patients with diabetes. Diabetes Care. 2009;32(7):1335–1343.
Berend K, de Vries APJ, Gans ROB. Physiological approach to assessment of acid-base disturbances. N Engl J Med. 2014;371(15):1434–1445.
Kraut JA, Madias NE. Serum anion gap: its uses and limitations in clinical medicine. Clin J Am Soc Nephrol. 2007;2(1):162–174.
Mehta AN, Emmett JB, Emmett M. GOLD MARK: an anion gap mnemonic for the 21st century. Lancet. 2008;372(9642):892.
Roberts DM, Yates C, Megarbane B, et al. Recommendations for the role of extracorporeal treatments in the management of acute methanol poisoning: a systematic review and consensus statement (EXTRIP). Crit Care Med. 2015;43(2):461–472.
Juurlink DN, Gosselin S, Kielstein JT, et al. Extracorporeal treatment for salicylate poisoning: systematic review and recommendations from the EXTRIP workgroup. Ann Emerg Med. 2015;66(2):165–181.