Respiratory Failure & Oxygen Therapy

In plain language, respiratory failure means the blood "doesn’t have enough oxygen or has too much carbon dioxide." Two words need separating first. Hypoxemia is low oxygen in arterial blood — a low PaO₂ on the blood gas. Hypoxia is oxygen insufficient at the tissue level to maintain homeostasis. Hypoxemia is one specific route to hypoxia, and the ABG is how you see it: the arterial blood gas is the gold standard for diagnosing respiratory failure, reading pH, PaO₂, PaCO₂, and HCO₃ from one sample.

The fork on that ABG defines the two types. Type 1 (hypoxemic) failure is a PaO₂ below 60 mmHg with a normal or decreased PaCO₂ — the lungs cannot load oxygen, but they are still blowing off CO₂. Type 2 (hypercapnic) failure is a PaCO₂ above 45 mmHg with a pH below 7.35 — the bellows are failing, so CO₂ accumulates and the blood turns acidic. Learn the threshold as 45, not 50: every current source defines hypercapnia and Type 2 failure at 45 mmHg, the upper limit of the normal PaCO₂ range. Type 2 is at its core a ventilation problem ("won’t breathe" or "can’t breathe"); Type 1 is a gas-exchange problem inside the lung.

Time matters as much as type. Acute failure "happens quickly and without much warning" and "qualifies as a medical emergency requiring intensive care hospitalization"; chronic failure is usually managed at home unless severe. The kidney tells you which is which: renal bicarbonate reabsorption in the proximal tubule "is a slow process," so in acute respiratory acidosis HCO₃ rises only about 1 mEq/L for every 10 mmHg rise in PaCO₂ over minutes, whereas in chronic retention it rises about 4 mEq/L per 10 mmHg over days, pulling the pH back to the low-normal range. Acute-on-chronic failure is the COPD patient whose HCO₃ is already high and whose pH has dropped below 7.35 — a new insult on top of compensated retention. The stakes are real: acute respiratory failure strikes about 1,275 per 100,000 US adults a year, AECOPD is the third most common cause of these admissions, in-hospital mortality is about 12% overall, and once intubated it climbs to 38% for AECOPD and 48% for pneumonia.

Hypoxemia has exactly five mechanisms, and each one predicts how the patient will respond to oxygen. (1) Hypoventilation — too little fresh gas reaches the alveoli (airway obstruction, CNS depression, obesity hypoventilation, neuromuscular disease); PaCO₂ rises with a normal A–a gradient, and it is "readily corrected with supplemental oxygen." (2) V/Q mismatch — ventilation and perfusion are unevenly matched: low-V/Q units in chronic bronchitis or pulmonary edema, high-V/Q (dead-space-like) units in pulmonary embolism or emphysema; the A–a gradient is raised and it corrects with 100% O₂. It is the most common cause of Type 1 failure (ARDS, COPD, heart failure, PE). (3) Right-to-left shunt — blood passes the lung without meeting ventilated alveoli, anatomically (ASD, VSD, PDA, AV malformation) or physiologically (pneumonia, atelectasis, ARDS); it is "notably resistant to correction with supplemental oxygen." True shunt does not improve with oxygen, and giving FiO₂ 1.0 is one way clinicians tell shunt from mismatch. (4) Diffusion impairment — interstitial edema, inflammation, or fibrosis thickens the membrane; "moderate-to-substantial supplemental oxygen may be necessary." (5) Low inspired oxygen — altitude; normal gradient.

The A–a gradient is the arithmetic behind that sorting: alveolar PAO₂ (from FiO₂, barometric pressure, and PaCO₂) minus the measured PaO₂, with a normal value that grows with age (roughly age ÷ 4, plus 4). A normal gradient with hypoxemia points to hypoventilation or low FiO₂ — the lung itself is fine, the air just isn’t arriving. A raised gradient points inside the lung: V/Q mismatch, shunt, or a diffusion defect. One more asymmetry explains why diffusion disease drops oxygen long before it traps CO₂: carbon dioxide diffuses more readily across the alveolar-capillary membrane than oxygen, so a thickened membrane costs the patient oxygen first. Whatever the mechanism, untreated hypoxia ends the same way — organ dysfunction, brain damage, and cardiac arrest — which is why the early cues on this page are an emergency-in-waiting, not a note for rounds.

Hypercapnia is mostly arithmetic: PaCO₂ is proportional to CO₂ production divided by alveolar ventilation, and alveolar ventilation is minute ventilation minus dead-space ventilation. So CO₂ climbs when the rate falls, when tidal volume falls, or when dead space rises — and "unless there is a significant loss of pulmonary ventilation, metabolic processes will not induce hypercapnia." Most hypercapnia is therefore the pump failing, not the body overproducing CO₂. Two counter-intuitive corollaries: rapid shallow breathing can raise CO₂ by ventilating mostly dead space (a rising dead-space-to-tidal-volume ratio), and fever, sepsis, thyrotoxicosis, or overfeeding raise CO₂ production but rarely cause hypercapnia alone.

The pump can fail at any link in the chain. Central drive — sedatives ("alcohol, benzodiazepines, opiates"), encephalitis, stroke, tumor, spinal cord injury. Neural and neuromuscular transmission — Guillain-Barré, myasthenia gravis, ALS, botulism, organophosphate poisoning, polio, tetanus, transverse myelitis. Chest wall and pleura — flail chest, kyphoscoliosis, hyperinflation, large effusions, obesity. Muscle — diaphragmatic paralysis, muscular dystrophy, and the exhausted diaphragm of a long asthma attack or COPD exacerbation. Obesity-hypoventilation syndrome deserves its own line: BMI above 30 with awake daytime PaCO₂ above 45 mmHg once other causes are excluded, about 90% with coexisting sleep apnea, driven by the mechanical load on the diaphragm plus a blunted CO₂ drive.

Drive itself is a CO₂ system. Arterial PCO₂ is the chief determinant of the respiratory drive — central chemoreceptors in the medulla read the CSF pH that arterial CO₂ produces, while the peripheral carotid and aortic bodies sense mainly low oxygen and supply only about 15% of the total drive. Opioids attack the center: μ-receptors in the medulla "reduce the medullary response to hypercarbia and decrease the respiratory response to hypoxia," and opioids "primarily target the preBötzinger complex," the rhythm generator — the result is a diminished stimulus to breathe and, eventually, apnea. Benzodiazepines act on GABA receptors and depress "all neural functions, including the respiratory pacemaking system." Tolerance is asymmetric: opioid-tolerant patients "do not develop complete tolerance to the loss of the hypoxic stimulus," which is why long-term users still die of overdose. And the nursing trap follows directly from the physiology: as CO₂ rises the patient does not gasp — CO₂ is a CNS depressant, so the patient gets sleepier, breathes less, and retains more, while supplemental oxygen keeps the SpO₂ reassuringly normal. "In these clients, SpO₂ may be normal, but this should not provide false reassurance."

Hemoglobin loads oxygen along a sigmoid curve. The flat upper portion means saturated hemoglobin at high PO₂ — extra oxygen adds almost nothing; the steep middle means a small fall in PO₂ releases a lot of oxygen. Acidosis, rising CO₂, fever, and 2,3-DPG shift the curve right (the Bohr effect, less affinity, more unloading); alkalosis, low CO₂, hypothermia, and carbon monoxide — which hemoglobin binds 200 to 300 times more avidly than oxygen — shift it left. The nursing number falls out of the curve: adequate tissue oxygenation is "generally achieved" at a PaO₂ of about 60 mmHg, which corresponds to a saturation of about 90% (the mid-to-high 80s on some devices — treat the pairing as approximate). Below that knee the patient is on the steep part, and saturation falls fast. Cyanosis is a late sign for the same reason: skin turns blue only when deoxygenated hemoglobin exceeds about 5 g/dL, roughly an arterial saturation under 80–85%. Anemia masks it — by the time an anemic patient looks blue, "oxygen saturation is very low" — and polycythemia exaggerates it.

"Oxygen is classified as a drug requiring prescription and skilled administration." It has a dose (FiO₂), a delivery system, a target, and toxicity. Low-flow devices mix oxygen with room air, so the FiO₂ "varies with patient minute ventilation"; high-flow systems exceed the patient’s peak inspiratory flow and deliver a consistent concentration. The targets are the whole game. For acutely ill adults not at risk of CO₂ retention, nursing sources carrying the BTS Emergency Oxygen Guideline set SpO₂ 94–98% — and every source shares the upper limit of 98%, then "wean to the lowest FiO₂ that holds the target." For patients at risk of hypercapnia — COPD, obesity hypoventilation, neuromuscular and chest-wall disease — the target is 88–92% (BTS and GOLD, as carried by StatPearls and the Brill & Wedzicha review). The evidence is not subtle: in a prehospital randomized trial, titrating COPD patients to 88–92% instead of high-flow oxygen produced "a 78% reduction in mortality for those patients with confirmed COPD," with less acidosis and hypercapnia.

Why does uncontrolled oxygen raise a retainer’s CO₂? Not the explanation you may have memorized. The accepted mechanisms: (1) supplemental oxygen releases hypoxic pulmonary vasoconstriction, so blood flows into poorly ventilated alveoli and V/Q mismatch worsens — more dead space, less CO₂ clearance; (2) the Haldane effect — oxygenated hemoglobin carries less CO₂, shifting the CO₂ dissociation curve right and accounting for up to 25% of the PaCO₂ rise; (3) a small, transient dip in ventilatory drive (about a 20% fall in minute ventilation) that recovers while PaCO₂ keeps climbing. The classic "hypoxic drive" story — that chemoreceptors go CO₂-blind and the patient breathes only from hypoxia, so oxygen switches breathing off — is a rejected theory, called "persistent and unfounded" in the oxygen-therapy literature. The practical rule is identical under every explanation: titrate to 88–92%, confirm with an ABG, and never withhold oxygen from a hypoxemic patient because of it. Excess oxygen has its own price list: absorption atelectasis (high FiO₂ washes nitrogen out of alveoli, which then collapse into shunt), airway drying without humidification, free-radical oxygen toxicity (pleuritic chest pain, substernal heaviness, cough, dyspnea from tracheobronchitis), hyperoxemia on ventilators associated with higher mortality, and — in a normoxic heart attack — larger infarcts. Use the lowest concentration that relieves tissue hypoxia and avoid prolonged FiO₂ above 0.6.