Control of Breathing and Dyspnea

Breathing is generated automatically by brainstem circuits, then adjusted by chemical signals, voluntary commands and reflexes from working muscles. For a breathless patient, ask which input is driving ventilation and whether the output matches the need.

Rhythm and inputs

Brainstem · chemoreceptors · behavior

Sagittal brainstem with pontine and medullary respiratory regions and descending motor output to the diaphragm and intercostal muscles. — Regional anatomy, using historical pneumotaxic and apneustic terminology. This overview does not resolve individual rhythm-generating cell groups. OpenStax College, Respiratory Centers of the Brain, CC BY 3.0. Unmodified.

Where the rhythm comes from

The preBötzinger complex in the medulla is central to inspiratory rhythm generation. Parafacial circuits contribute to active expiration. These interacting networks are more specific than the classical regions in the illustration.

Three families of input

Chemical: central and peripheral chemoreceptors respond to CO₂ and pH. The carotid bodies provide the principal acute hypoxic response.
Behavioral: speech, voluntary breathing and emotion modify automatic breathing. Cat recordings show interaction within the automatic system, not a simple independent bypass.
Exercise: central command and muscle afferents increase ventilation with metabolic demand.

Dyspnea adds perception. Respiratory motor command and sensory feedback contribute to breathing discomfort. Gas exchange and symptom intensity can diverge.

Both extremes exist

Interpret in clinical context

01

Too little drive

Hypoventilation

Congenital central hypoventilation syndrome combines alveolar hypoventilation with autonomic dysregulation. Most patients have a pathogenic PHOX2B variant, but rare non-PHOX2B forms are recognized. Specialist evaluation integrates phenotype and genetic testing.

02

Too much drive

Hyperventilation

Central neurogenic hyperventilation was described in 1959 in patients with brainstem disease. In practice it is reached only after more common causes of hyperventilation have been excluded (chapter 5).

Why start here: a respiratory rate, a saturation and a chest film describe the output of this system. The chapters that follow show how to read the gas to infer the input, because the cause of breathlessness is often outside the lungs.

At a fixed PCO₂, the ventilatory response to falling oxygen becomes much steeper around an alveolar PO₂ of 50–60 mmHg. This is a descriptive range, not an on–off threshold or a rule for excluding cardiopulmonary disease.

Read the response

Oxygen and CO₂ interact

Ventilation against alveolar PO₂ (schematic)

Marker follows the resting-PCO₂ curve. Timing is illustrative.

Schematic of the hyperbolic relationship in healthy adults breathing at fixed PCO₂ (isocapnic hypoxia). Ventilation changes little until alveolar PO₂ falls to about 50–60 mmHg, then rises steeply. At higher PCO₂ the same fall in PO₂ produces a much larger response. The curve shapes follow the published shape parameters. The ventilation axis is illustrative.

01

Use it at the bedside

Read the shape, then the limits

Normal PaO₂ makes severe hypoxemia an unlikely dominant driver. It does not establish zero carotid-body activity, normal oxygen delivery or absence of lung disease.
Hypocapnia blunts the hypoxic response. Do not extrapolate measurements at resting PCO₂ ±5 mmHg to extreme hypocapnia or diagnose a central lesion from respiratory rate alone.
Hypoxia and hypercapnia interact. How steeply ventilation rises as PO₂ falls depends on the PCO₂ at which it is measured.

Limits: these are laboratory curves from ten healthy volunteers using a non-steady-state technique. They describe normal physiology, not any one patient's response.

What the laboratory curves showed

Ten healthy volunteers

Shape

Hyperbolic fit. Ventilation rose steeply below about 50–60 mmHg. The fitted asymptote at 30–40 mmHg is a mathematical model feature, not infinite ventilation in a patient.

Speed

Ventilation adjusted to a change in alveolar PO₂ within 18–23 seconds.

Dependence on CO₂

The shape parameter A, where higher means a steeper response, was 180 at resting PACO₂, 453 at 5 mmHg above resting and 30 at 5 mmHg below. The same hypoxia drives ventilation much more at higher PCO₂ and much less at lower PCO₂.

Oxygen content

In four subjects ventilation was a linear function of arterial oxygen content, which suggests but does not prove that chemoreceptor tissue PO₂, partly set by oxygen delivery, is an important factor.

In a breathless patient with a high respiratory rate, read pH and bicarbonate first to distinguish primary hyperventilation from metabolic compensation. Then consider the direction of the PaCO₂ and the size of the alveolar–arterial (A–a) oxygen gradient.

Sort the pattern

Acid–base first

Sort breathlessness by PCO₂ and the A–a gradient

	A–a gradient normal for age	A–a gradient widened
PCO₂ low	Consider increased drive or metabolic compensationConsider metabolic acidosis, CNS disease, pain, sepsis, salicylates, pregnancy or hepatic disease. Anxiety is a diagnosis of exclusion.	Consider gas-exchange impairment plus increased driveLung or pulmonary vascular disease: embolism, interstitial disease, pneumonia, edema, shunt.
PCO₂ high	Consider alveolar hypoventilationNeuromuscular weakness, drug or brainstem depression, obesity or chest-wall restriction.	Consider combined gas-exchange and ventilatory failureCOPD, severe asthma or advanced parenchymal disease. Fatigue is only one possible contributor.

A physiologic sorting aid, not a validated decision rule. Low PaCO₂ may be compensation for metabolic acidosis. Normal blood gases or a normal A–a gradient cannot exclude pulmonary embolism. Respiratory alkalosis has both pulmonary and extrapulmonary causes.

Calculate and interpret

Room air · sea level

01

Compute the gradient

Record FiO₂ and barometric pressure

Approximate alveolar PO₂: PAO₂ = FiO₂ × (barometric pressure − 47) − PaCO₂ ÷ 0.8. Assuming a respiratory exchange ratio of 0.8, on room air at sea level this is about 150 − 1.25 × PaCO₂. The simplified equation is less accurate at high FiO₂.
A–a gradient = PAO₂ − PaO₂.
Compare with the age-expected value. A widely used rule of thumb for the upper limit on room air is (age ÷ 4) + 4 mmHg.

Worked example (room air, sea level, age 80). PCO₂ 15 and PO₂ 100 mmHg. PAO₂ = 0.21 × (760 − 47) − 15 ÷ 0.8 = 149.7 − 18.8 = 131 mmHg. A–a gradient = 131 − 100 = 31 mmHg. Age-expected ≈ 80 ÷ 4 + 4 = 24 mmHg. The gradient is mildly above the rule of thumb. The PaO₂ of 100 looks reassuring partly because hyperventilation has raised the alveolar PO₂ to 131.

Reference values vary. Hardie reported mean PaO₂ of 77.0 mmHg in men and 73.5 mmHg in women among 146 healthy adults over 70. These population means do not validate an upper-limit formula for an individual.

Where the gradient misleads

Assumptions matter

Pitfall	What to do
Supplemental oxygen	Document the oxygen device and FiO₂. Nasal-cannula FiO₂ is uncertain, and the room-air age rule does not apply reliably on oxygen. Do not withdraw needed oxygen to calculate a gradient.
Altitude or weather	Use the actual barometric pressure rather than 760 mmHg.
Choice of formula	Several age equations exist. Treat the expected value as an estimate, not a diagnostic exclusion threshold.
Masked widening	A normal PaO₂ can coexist with a widened gradient when a low PaCO₂ has raised the alveolar PO₂. Compute the gradient rather than judging PaO₂ alone.

In some patients with chronic lung disease, giving oxygen raises PaCO₂. The long-taught explanation, that oxygen removes the hypoxic drive and the patient stops breathing, is not supported as the main mechanism, and the response to hypoxemia is still titrated oxygen.

At the bedside

Titrate · repeat gases · reassess

Titrate oxygen to 88–92% in COPD or other risk of hypercapnic respiratory failure pending blood gases. Adjust to the patient’s gas results and prior history. Critical illness may initially require high-concentration oxygen while urgent assessment and ventilatory support are arranged.
Repeat blood gases within 30–60 minutes after starting or increasing oxygen in patients at risk of hypercapnia, even if the initial PaCO₂ is normal. Reassess sooner if deteriorating. Rising PaCO₂ with acidosis requires urgent assessment for ventilatory support. Step oxygen down to the target rather than abruptly stopping it.
This chapter concerns CO₂ retention. The hypothetical example in chapter 5 is hypocapnic and asks a different physiologic question.

What the physiology studies found

Mechanisms and their limits

Aubier 1980: COPD in acute respiratory failure, air then 100% oxygen

Minute ventilation fell early (by 18 ± 2% of control, through both tidal volume and rate), then recovered to 93 ± 6% of the room-air value by 15 minutes. PaCO₂ nevertheless rose by 23 ± 5 mmHg, with no significant correlation between the change in ventilation and the change in PaCO₂. The authors attributed the rise mainly to increased VA/Q inhomogeneity.

Hanson 1996: computer model compared with a prior case series

Oxygen alters hypoxic pulmonary vasoconstriction and modulates the Haldane effect, changing physiologic dead space. The model found these changes sufficient to account for the hypercarbia.

Robinson 2000: 22 patients in an acute exacerbation

Twelve retained CO₂ (PaCO₂ rose 8.3 ± 5.6 mmHg) and ten did not. Ventilation fell in retainers (9.0 to 7.2 L/min, P = 0.007) but not in non-retainers, while V/Q dispersion widened in both groups. The authors concluded that an overall reduction in ventilation characterizes oxygen-induced hypercapnia. Alveolar dead space rose in retainers only.

Abdo and Heunks 2012: review

The review argues against loss of hypoxic drive as the principal explanation and recommends titrated oxygen. Reduced ventilation can still contribute.

The honest synthesis is multifactorial. V/Q redistribution with added dead space, the Haldane effect and, in some patients, a modest fall in ventilation all contribute, and the studies disagree about their relative weight. Aubier found ventilation back near baseline while PaCO₂ still climbed. Robinson found a ventilation fall that separated retainers from non-retainers. Loss of hypoxic drive alone does not adequately explain these findings.

What the outcome data show

Austin trial · Cochrane review

Austin and colleagues used cluster randomization of paramedics to give either 8–10 L/min oxygen by non-rebreather mask or oxygen titrated to a saturation of 88–92% in 405 patients with a presumed COPD exacerbation (214 with confirmed COPD). Mortality was 9% on the high-flow regimen and 4% on titrated oxygen (relative risk 0.42, 95% CI 0.20–0.89), and 9% versus 2% in confirmed COPD (relative risk 0.22, 0.05–0.91). Lower acidosis and hypercapnia were found in the small per-protocol blood-gas subgroup. These secondary outcomes did not differ significantly in the intention-to-treat analysis.

Grade the evidence honestly. The 2020 Cochrane review found this to be the only eligible trial and judged the evidence low-certainty, with unclear to high risk of bias.

A low PaCO₂ shows that effective alveolar ventilation is high relative to CO₂ production. It does not by itself identify why the patient is breathless.

Read it in three steps

Physiology before diagnosis

01

A hypothetical blood gas

Room air · sea level

pH7.53

PaCO₂25 mmHg

PaO₂100 mmHg

Calculated HCO₃⁻≈20 mmol/L

Illustrative values, not a patient record. Consider a 60-year-old with persistent tachypnea and clear lungs. A room-air arterial sample at sea level shows pH 7.53, PaCO₂ 25 mmHg and PaO₂ 100 mmHg. Calculated bicarbonate is about 20 mmol/L.

Acid–base

Alkalemia with hypocapnia: a primary respiratory alkalosis is present. Determine whether metabolic disease also contributes. Blood-gas bicarbonate is calculated from pH and PaCO₂, whereas a chemistry panel measures total CO₂.

Ventilation

Relative alveolar ventilation: at constant CO₂ production, changing PaCO₂ from 40 to 25 mmHg corresponds to a 40/25 = 1.6-fold increase. Respiratory rate alone cannot quantify alveolar ventilation because tidal volume and dead space also matter.

Oxygen transfer

Oxygen transfer: PAO₂ ≈ 149.7 − 25/0.8 = 118.5 mmHg. A–a ≈ 18.5 mmHg, near the age rule of 60/4 + 4 = 19 mmHg. This supports relatively preserved resting oxygen transfer in this setting, but does not exclude PE or an exertional abnormality.

02

Choose the next question

Use the clinical context

Systemic or metabolic drive: fever, sepsis, metabolic acidosis, pregnancy, liver disease, pain or a drug/toxin exposure such as salicylate.
Pulmonary or cardiac disease: use pretest probability and disease-specific pathways. A reassuring resting gas does not end the evaluation.
Neurologic or behavioral drive: pursue CNS assessment when neurologic findings or otherwise unexplained persistent hyperventilation support it. Central neurogenic hyperventilation is a rare diagnosis of exclusion. Anxiety should not be assigned solely because chest tests are normal.

Resolve the remaining questions

Time course · metabolic disease · filling pressures

What a low bicarbonate does not tell you

Renal adaptation to sustained hypocapnia lowers bicarbonate. A coexisting metabolic acidosis can also lower it. Review the clinical time course, electrolytes, anion gap, lactate, renal function and relevant exposures before attributing the bicarbonate entirely to compensation or using it to date the disorder.

The HFpEF question: grade 1 diastolic dysfunction neither establishes nor excludes HFpEF. Symptoms and signs require objective evidence of elevated filling pressures, at rest or with provocation, rather than an echo grade alone. See Heart Failure and CPET.

When breathlessness persists after the obvious chest testing, combine symptoms, examination, pretest probability and blood-gas physiology to choose the next test.

A bedside sequence

Stabilize · sample · interpret

Look and count.
Rate, work of breathing, speech, mental state and saturation. Stabilize hypoxemia, shock or exhaustion first.
Get the gas.
Use an arterial sample when oxygen-transfer or ventilatory assessment is needed. Record oxygen delivery and do not delay stabilization.
Compute the A–a gradient
with known FiO₂. Use the age estimate on room air (chapter 3).
Interpret.
Start with pH and compensation, then PCO₂ and the A–a gradient.
Target the next test.
Use the pattern to refine the differential, never as an exclusion rule.

Choose the next test

Pattern plus pretest probability

If the gas shows	Consider	Related guides
Low PCO₂, widened gradient	Lung or pulmonary vascular disease: embolism, pulmonary hypertension, interstitial disease, edema, shunt.	Pulmonary Embolism & VTE, Pulmonary Hypertension, CPET
Low PCO₂, normal gradient	Metabolic compensation or increased drive: sepsis, salicylates, pregnancy, hepatic disease, pain, CNS disease or anxiety after exclusion of other causes. PE remains possible.	Electrolytes & Acid–Base Emergencies
High PCO₂, normal gradient	Pump failure: neuromuscular weakness, drug or brainstem depression, obesity or chest-wall restriction.	Ventilator Management
High PCO₂, widened gradient	Combined gas-exchange and ventilatory failure. Assess the cause, titrate oxygen and consider ventilatory support (chapter 4).	Ventilator Management

Keep the symptom in context

Cardiac function · perception

Ask the cardiac question explicitly

A normal resting echocardiogram and a normal chest do not exclude a cardiac contribution. Heart failure with preserved ejection fraction is defined by elevated filling pressures at rest or with provocation, and exertional testing can show what rest does not. Relevant guides: Heart Failure, Echocardiography, Right Heart Cath & Hemodynamics and Cardiopulmonary Exercise Testing.

Dyspnea is a sensation, not a gas

The 2012 American Thoracic Society statement distinguishes air hunger, breathing effort and chest tightness, which can arise through different mechanisms. Dyspnea engages cortico-limbic networks as well as sensory and motor pathways. A normal gas does not make the symptom unreal. Ask about air hunger, effort and chest tightness, what provokes them, and how they limit activity.

Limits of this guide. It is an educational synthesis of physiology and bedside reasoning, not a treatment protocol. The sorting table is a teaching aid, not a validated decision rule. The worked example is hypothetical. A normal A–a gradient cannot exclude pulmonary embolism.