Gas analysis,
from photon to plateau.

A molecule absorbs infrared radiation only if its vibration changes its dipole moment: only if the distribution of charge across the molecule is asymmetrical, or becomes asymmetrical as the bonds stretch and bend. In practice that means gases whose molecules contain two or more different atoms. Carbon dioxide, nitrous oxide, water vapour and the halogenated volatile agents all qualify. Oxygen, nitrogen and helium do not: they are elementary or symmetrical, have no dipole moment, and are effectively transparent to infrared.

Each absorbing gas does so at its own characteristic wavelengths. Choosing a wavelength where the gas of interest absorbs strongly, and the other gases in the mixture do not, is what lets one gas be measured in the presence of the others.

Infrared light of the chosen wavelength is passed through a chamber of known path length containing the sample. The transmitted intensity, I, falls exponentially from the incident intensity, I₀, with both the concentration of the absorbing gas, c, and the distance the light travels through it, the path length l, which the design fixes. Written as absorbance, A, the same law is a straight line: absorbance is proportional to concentration and path length.

ε is the extinction coefficient: how strongly one mole of this particular gas absorbs at this particular wavelength, in L·mol⁻¹·cm⁻¹. It is a fixed property of the molecule, not of the sample or the instrument, and it is the term that makes the method gas-specific. Carbon dioxide has a large ε near 4.3 µm and a negligible one elsewhere, which is exactly why a narrow optical filter can pick it out of a mixture; a gas with ε = 0 at the chosen wavelength, such as oxygen, cannot be measured this way at all.

The bench has five parts, in the order the light meets them:

1. Infrared source: a heated element emitting a broad infrared spectrum.
2. Optical filter, or a rotating filter wheel, selecting the narrow band near 4.3 µm. Selectivity here decides how much nitrous oxide interference gets through. A rotating wheel also chops the beam into an alternating signal, which is easier to amplify and drifts less than a steady one.
3. Sample chamber of fixed path length, with sapphire windows. Ordinary glass cannot be used because glass itself absorbs infrared.
4. Reference path or reference cell, containing carbon dioxide-free gas, so the analyser measures a ratio rather than an absolute intensity and is insensitive to drift in the source and detector (a double-beam instrument).
5. Photodetector and processor, converting transmitted intensity to a displayed concentration or partial pressure.

Zeroing and calibration

The analyser is zeroed against a carbon dioxide-free gas, usually room air, which sets the value of I₀ and compensates for gradual soiling of the windows and ageing of the source. It is calibrated against a gas mixture of known carbon dioxide concentration, which sets the scale. Zeroing corrects the offset; calibration corrects the gain. Both are needed.

Response time

The total delay between gas leaving the airway and the number changing on the screen is the transit time along the sampling line plus the rise time of the analyser itself, conventionally the time to move from 10% to 90% of a step change. A small analysing chamber, of about 1 mL, keeps the rise time under about 400 ms. The transit time only delays the signal, but if the total approaches the duration of expiration, the plateau is never reached and end-tidal carbon dioxide is under-read. This is why neonatal and high-respiratory-rate monitoring is where sidestream sampling most often fails.

The slight upslope of phase III is real and worth explaining rather than drawing flat. Alveoli with longer time constants and lower ventilation–perfusion ratios empty later in expiration and carry more carbon dioxide, so the concentration creeps upward across the plateau.

The two angles

The α angle, between phases II and III, is normally about 100–110°. It increases with obstruction to expiratory flow and with ventilation–perfusion mismatch, together with a steeper phase III. The β angle, between phase III and the inspiratory downstroke, is normally about 90°, and it increases with rebreathing.

Normal values

The arterial to end-tidal difference exists because alveolar dead space, alveoli that are ventilated but not perfused, contributes carbon dioxide-free gas to the expirate and dilutes the alveolar sample. In healthy people with even ventilation the gradient is under 5 mmHg, small and reasonably stable, which is what makes end-tidal carbon dioxide a useful surrogate for arterial carbon dioxide.

It becomes unreliable exactly when the gradient changes. It widens with hypovolaemia, reduced cardiac output, pulmonary embolism, high airway pressures or excessive positive end-expiratory pressure, chronic lung disease, increasing age and the upright posture. In those situations a falling end-tidal value does not mean a falling arterial value; it may mean the opposite. And with severe heterogeneity of ventilation, end-tidal PCO₂ can even exceed arterial.

That is also the basis for judging the adequacy of ventilation from the trace. Hypoventilation raises the end-tidal value and hyperventilation lowers it, but only while the gradient is stable, so a falling trace in a shocked or embolised patient is a circulatory sign, not a ventilatory one.

Rebreathing from equipment deserves a special mention. If all the carbon dioxide produced is retained and distributed through the body stores, arterial PCO₂ can rise no faster than about 0.4 to 0.8 kPa a minute, which limits how quickly rebreathing becomes dangerous but also means the rise is easy to miss without a capnograph.

A sudden change of the trace to a straight line gives prompt warning of a failure of ventilation or of the circulation, such as an air embolus, while a progressive rise of end-tidal carbon dioxide points to inadequate ventilation or the onset of malignant hyperthermia.

The Severinghaus electrode measures PCO₂ in blood, not in gas. Carbon dioxide diffuses across a carbon dioxide-permeable plastic membrane into a thin film of sodium bicarbonate solution held in a nylon mesh, where it forms carbonic acid and dissociates; the resulting change in hydrogen ion concentration is measured by a glass pH electrode. Because the gas must diffuse through the membrane, its response takes about 2–3 minutes, and with care it is accurate to about 130 Pa (1 mmHg). Like the oxygen electrode, it is kept at 37 °C and calibrated with humidified gas mixtures of known concentration.

A colorimetric detector uses a pH-sensitive dye, classically metacresol purple, which changes from purple to yellow in the presence of carbon dioxide. It is a disposable qualitative device for confirming tracheal placement when no capnograph is available. It gives no waveform, no number and no trend, and gastric contents, mucus or adrenaline in the tube can discolour it.

Mass spectrometry ionises the sample gas and separates the ions by mass in a magnetic field (or with four charged rods in a quadrupole instrument). It identifies and quantifies up to eight gases breath by breath, with a response under 100 ms. Nitrous oxide and carbon dioxide share a mass number of 44, so nitrous oxide is measured from its nitric oxide fragment, mass 30. The equipment is expensive, so it was shared between theatres through long sampling lines, with delay and a single point of failure; it has largely disappeared from clinical use.

Raman spectroscopy shines a laser through the sample; a small proportion of the scattered light is shifted in wavelength by an amount characteristic of each molecule. It can measure oxygen and nitrogen as well, but the equipment is large and no longer in routine theatre use. Photoacoustic spectroscopy pulses infrared radiation through the gas, causing cyclical heating, expansion and a pressure wave that a microphone detects as sound; one simple microphone serves every infrared-absorbing gas, and it appears in some multi-gas analysers.

All the agents in current use are halogenated hydrocarbons or ethers containing C–H bonds, so they all absorb near 3.3 µm. A single-wavelength analyser can therefore say how much agent is present but not which agent it is; the early analysers had to be programmed by the user for the agent in use, and the wrong agent, or a mixture, gave erroneous readings. Because the minimum alveolar concentration of desflurane, about 6%, is roughly five times that of isoflurane, about 1.15%, a misidentified agent gives a badly misleading picture of depth.

Modern analysers solve this by measuring each agent at its own set of wavelengths in the 8–12 µm range and matching the pattern against known spectra. That lets them identify the agent and quantify it, and even quantify two agents in the presence of each other.

Carbon dioxide raises the hydrogen ion concentration by its spontaneous and carbonic anhydrase-catalysed hydration, and it also reacts with free amine groups in proteins to form carbamino compounds. Because it diffuses readily across cell membranes, it generates carbonic acid and hydrogen ions inside cells as well as outside them, and the resulting pH follows the Henderson–Hasselbalch relationship.

Two rules of thumb describe the size of the change. In acute respiratory acidosis, pH falls by about 0.08 for every 10 mmHg (1.33 kPa) rise in PaCO₂, and bicarbonate rises by 1–2 mmol/L. In chronic hypercarbia, renal compensation raises bicarbonate by 3–4 mmol/L for each 10 mmHg, so pH falls by only about 0.03.

Because the net cardiovascular effect is the balance of two opposing actions, it is unpredictable, and neither “increased cardiac output” nor “myocardial depression” alone describes it.

The clinical picture in a patient with chronic hypercapnia follows from the table: a flushed skin and a full, bounding pulse with occasional extrasystoles; a raised but unreliable blood pressure; and muscle twitching and a characteristic flap of the hands. Arterial PCO₂ above 6.1 kPa (46 mmHg) should be considered abnormal in a healthy subject, although values up to 6.7 kPa (50 mmHg) can be reached by breath-holding.

There are only four causes of hypercapnia: carbon dioxide in the inspired gas (from rebreathing, since current machines carry no carbon dioxide supply); increased production, as in malignant hyperthermia or sepsis, or absorption during laparoscopy; hypoventilation, by far the commonest; and increased dead space.

Cerebral vasoconstriction with reduced cerebral blood flow is the most important effect. It reduces intracranial pressure briefly, but the reduction may be short-lived, the effect on flow is variable, and sustained vasoconstriction risks cerebral ischaemia; hyperventilation to lower intracranial pressure is therefore kept for when other measures have failed. With prolonged hypocapnia, cerebral blood flow drifts back towards baseline over a few hours as CSF pH corrects.

Hypocarbia also shifts the oxyhaemoglobin dissociation curve to the left, increasing haemoglobin affinity and impairing tissue unloading. Severe hypocapnia causes paraesthesiae and tetany from a general hyperexcitability of nerves; the fall in ionised calcium is too small to explain the tetany. In the anaesthetised patient it causes apnoea once PaCO₂ falls below the apnoeic threshold, and in animals it attenuates hypoxic pulmonary vasoconstriction, although that has not been described in humans.