Impact of Common and Understudied Technical Errors on Diffusing Capacity of the Lung for Carbon Monoxide

Introduction

The diffusing capacity of the lung for carbon monoxide (D_LCO) provides an important measurement of the transfer of gas from inhaled air to the bloodstream. Accurate D_LCO measurements are essential for diagnosing and managing lung diseases. ¹ Although American Thoracic Society/European Respiratory Society (ATS/ERS) standards ² define acceptable maneuvers, the impact of common errors beyond slow or submaximal inhalation, variable breath-hold times, changes in intrathoracic pressure (exceeding ± 40 to 50 cm H₂O ³ ), and improper alveolar sampling remains under characterized. We prospectively simulated seven errors to quantify their directional effects on D_LCO, alveolar volume (V_A), and carbon monoxide transfer coefficient (K_CO).

Methods

This study employed a within-subject experimental design and enrolled three experienced pulmonary function technologists, each acting as their own control. The UCSF Institutional Review Board approved the study (approval number 23-38743) and the subjects provided informed consent. Each error’s three trials with two associated control efforts were performed in separate sessions for each error type. Single-breath D_LCO followed ATS/ERS standards, including the four-minute minimum between efforts. The equipment to measure D_LCO was manufactured by nSpire. The test gas composition was 0.3% CO, 0.3% CH₄, 21% O₂, and balance nitrogen. We simulated seven common errors observed clinically (Fig. 1): (1) inspiratory leak; (2) early inhale; (3) leak during breath-hold to <85% of vital capacity (VC), the threshold specified in the ATS/ERS technical standard as secondary criteria for acceptability and as grade B (VC was from forced spirometry efforts performed in the same session); (4) leak during breath-hold but maintenance of ≥85% of VC; (5) leak around the mouthpiece with a closed valve; (6) leak during exhalation; and (7) nonzeroed sensors. Primary outcomes were D_LCO, V_A, and K_CO relative to the session’s control values. Given the small sample size, heterogeneous baselines between subjects, and expected variability within a subject and error type due to the inherent difficulty in replicating the error exactly, we report within-subject qualitative changes rather than inferential statistics.

Fig. 1.

Representative single-breath D_LCO tracings for all errors. Sampled concentrations of methane (CH₄%, green) and carbon monoxide (CO%, blue) rise during inspiration, plateau over a ∼10-second breath-hold (starting at the vertical red line marking breath-hold start [BHTS]), then wash out on expiration. The red trace shows spirometry throughout the maneuver, with residual volume arbitrarily set at 1 L per manufacturer settings. The upper horizontal blue line labeled VC denotes the predicted threshold for 100% vital capacity obtained from forced spirometry testing, and the lower blue horizontal line represents 85% of VC (ATS/ERS Grade B). The thin green tracing at the 1.0 L Y axis indicates mouth pressure (two horizontal blue lines above and below the 1.0 L line indicate ±60 cm H₂O). The left vertical axis is the volume of inspired air in liters; the right vertical axis is the gas concentration based on a percentage of the test gas concentration. The bold red lines on the right indicate the alveolar sample window.

Results

Errors clustered into three patterns (Fig. 2).

Fig. 2.

D_LCO, V_A, and K_CO across all errors. Scatter plots display D_LCO, V_A, and K_CO values for three subjects (S1–S3) across all tested error types (black: control efforts; red: experimental efforts). Shifts relative to control represent measurement artifacts introduced by the maneuver rather than changes in underlying lung function.

1. (1)

   Increased D_LCO due to dilution/volume artifacts: Inspiratory leak and early inhale generally increased D_LCO, driven primarily by increases in V_A; K_CO increased slightly with an inspiratory leak and did not change consistently with an early inhale.

2. (2)

   Increased D_LCO due to loss of lung volume artifacts: Breath-hold leaks through the circuit to <85% of VC and leaks around the mouthpiece against a closed valve consistently increased D_LCO with stable V_A and increased K_CO, consistent with the known volume dependence of K_CO. ⁴

3. (3)

   Breath-hold leaks with lung volume maintained at ≥85% of VC produced smaller, inconsistent D_LCO shifts, typically with slight decreases in V_A. Leaks during exhalation and nonzeroed sensors were variable across subjects and produced directionally inconsistent changes in all three metrics.

In brief, inspiratory leak and early inhale → ↑D_LCO via ↑V_A; breath-hold leak <85% VC and closed-valve leak → ↑D_LCO via ↑K_CO; other errors inconsistent.

Discussion

Our simulations show that commonly observed deviations in pulmonary function testing can systematically inflate D_LCO through two mechanistic archetypes. First, dilution/volume artifacts through inspiratory leaks and early inhalation primarily increase V_A with little change in K_CO, yielding higher D_LCO (D_LCO = K_CO × V_A). Second, loss of lung volume artifacts through breath-hold leaks to <85% VC and leaks against a closed valve leave V_A similar or slightly lower but raise K_CO, again inflating D_LCO. The magnitude of D_LCO inflation exceeds reported inter-session variability. ⁵ Errors that preserved lung volume at ≥85% VC, leaks during exhalation, and nonzeroed sensors produced small and directionally inconsistent shifts that have the potential to be device- and subject-dependent.

During an inspiratory leak or early inhale, ambient air mixes with the test gas before the alveolar sample is drawn. CO and CH₄ concentrations measured in the alveolar sample decrease as external air dilutes these gases, falsely indicating gas mixing in a larger lung volume. In addition, an inspiratory leak underestimates the inspired volume (VI) as measured by the spirometer. The net impact of a low measured VI with CH₄ dilution was an increase in the V_A. CO continues to diffuse during the breath-hold while CH₄ remains inert, producing a larger fall in alveolar CO than CH₄, an increased CH₄:CO ratio in the alveolar sample, and consequently a higher K_CO. In practice, this yields an increased D_LCO with unchanged V_A and unchanged or increased K_CO.

Exhalation to <85% VC or loss of lung volume via leak with a closed valve (where the amount of volume lost is unknown) reduces alveolar volume. K_CO is known to increase at lower lung volumes ³ due to proportionally greater recruitment of capillary blood volume and alveolar-capillary surface area relative to the loss of ventilated volume. The nSpire equipment in this study uses the highest inspiratory volume attained during the maneuver to calculate V_A regardless of a subsequent decline in volume. Hence, V_A is similar, but K_CO increases, so D_LCO increases. Exhalation to <85% VC accentuates this effect; when subjects remained ≥85% VC, changes were smaller and less consistent, aligning with the volume-dependence of K_CO.

Because this study involved only three healthy technologists on a single manufacturer and model of equipment, with characteristics that may differ from modern systems, our observations should be considered hypothesis-generating. The nSpire equipment has a valve that is closed manually by operator trigger. In contrast, a system that automatically closes the valve will not lead to leaks with a decline in lung volume visible on the spirometry tracing, but there will still be a risk of a leak against a closed valve. These findings serve as useful educational tools for identifying quality-control issues rather than as generalizable conclusions for larger patient groups. Effects in patients with lung disease, different dead-space fractions, or altered hemoglobin may differ. Future work should quantify effect sizes across devices and patient phenotypes, and develop automated QC flags.

In conclusion, common technical errors can inflate D_LCO through two predictable mechanisms: V_A inflation (dilution) and K_CO elevation at low volume. An elevated D_LCO may generate concern ⁶ and a normal D_LCO is generally reassuring, yet these data show that it can be artifactually high due to subtle errors. Teaching these subtleties and embedding them into quality control practices can prevent reporting or misinterpreting erroneous data.