Real-world diffusing capacity simulation data: how well do they fit current European Respiratory Society/American Thoracic Society acceptability criteria?

To the Editor:

Diffusing capacity of the lungs for carbon monoxide (D_LCO) is an important pulmonary function test for the diagnosis and management of obstructive, restrictive and pulmonary vascular disease. The 2017 European Respiratory Society (ERS)/American Thoracic Society (ATS) standards for single-breath carbon monoxide uptake in the lungs recommend that a weekly D_LCO simulation test be performed with a calibrated 3-L syringe [1]. This type of simulation provides quality control values for both D_LCO and alveolar volume (V_A). According to ERS/ATS standards, an acceptable simulated D_LCO is <0.5 mL·min⁻¹·mmHg⁻¹ and an acceptable simulated V_A is 3±0.3 L, under ambient temperature, pressure, dry (ATPD) conditions [1]. The ERS/ATS D_LCO standards document states that the simulated V_A should be reported under BTPS (body temperature, pressure, saturated) conditions; however, in this type of simulation, volumes should be reported under ATPD conditions.

We previously reported a case from a single system where the simulated V_A was 8–11 standard deviations above the measured mean due to a leak in a gas sampling collection bag, yet all of these values were within the ERS/ATS limits of acceptability [1, 2]. We suggested that D_LCO and V_A simulation limits based on actual performance rather than fixed arbitrary values may provide better quality control of D_LCO devices. The purpose of this study was to analyse D_LCO and V_A simulation data from multiple devices and laboratories, and compare the performance of these devices against the ERS/ATS recommended simulation limits.

We analysed D_LCO and V_A simulation data collected from three different types of D_LCO systems: Medisoft BodyBox (three devices) and SpiroAir (one device) (Sorinnes, Belgium) (ComPAS software; Morgan Scientific, Haverhill, MA, USA), and Platinum Elite (nine devices) (MGC Diagnostics Corporation, St Paul, MN, USA). The Medisoft devices are classical systems that use plastic bags for the collection of discrete gas samples (referred to as “expiratory bag” in this letter), whereas the MGC devices were equipped with either a rapid gas analyser (RGA) (six devices) or gas chromatograph (GC) (three devices). Data were collected from four clinical laboratories (St Louis University, St Louis, MO, USA; University of Vermont, Burlington, VT, USA; Elliot Health System, Manchester, NH, USA; and St Joseph Hospital, Nashua, NH, USA). Data were collected between 2017 and 2021.

D_LCO and V_A simulations were performed with a 3-L calibration syringe with a current accuracy certification according to each laboratory's protocol. The Medisoft/Morgan Scientific devices perform the D_LCO simulation with a full syringe of test gas in patient testing mode while the MGC devices perform the D_LCO simulation with 1 L air mixed with 2 L test gas in a simulation test mode. Consecutive measurements performed on different days were retrospectively collected from each device. D_LCO or V_A simulation values identified as statistical outliers that would likely prompt corrective action were excluded. Data from each type of device were compared to the 2017 ERS/ATS limits [1] and then to each other. Following individual comparisons, the data from all devices were pooled and compared to the 2017 ERS/ATS limits [1]. The use of 3-sd limits (common in laboratory medicine) [3] were considered in comparison to the fixed limits as recommended by the ERS/ATS standards.

Commercially available software was used to perform statistical analysis (Prism, version 4.0; GraphPad Software, San Diego, CA, USA). Grubb's test was applied to identify statistical outliers. Mean and standard deviation were calculated for each type of device, differences between device types was assessed with the Kruskal–Wallis (one-way ANOVA) test and Dunn's multiple comparison post-test if necessary. A p-value <0.05 was considered significant.

A total of four D_LCO measurements (two expiratory bags and two GCs) and five V_A measurements (one expiratory bag, two RGAs and two GCs) were removed from analysis after being identified as outliers. Following outlier removal, 1157 D_LCO (expiratory bag, n=286; RGA, n=707; GC, n=164) and 1158 V_A (expiratory bag, n=287; RGA, n=706; GC, n=165) simulation tests were analysed. The mean D_LCO and V_A values from different devices and when pooled collectively were within the ERS/ATS simulation limits (figure 1). The left panels of the figure shows the pooled D_LCO and V_A simulation data in comparison to the ERS/ATS limits. There were some differences between devices. If a 3-sd range is used, the high end of the D_LCO range marginally exceeds the ERS/ATS limit (<0.5 mL·min⁻¹·mmHg⁻¹) for the expiratory bag (0.58 mL·min⁻¹·mmHg⁻¹), RGA (0.81 mL·min⁻¹·mmHg⁻¹) and for the pooled data (0.75 mL·min⁻¹·mmHg⁻¹).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1

The left panels shows pooled diffusing capacity of the lung for carbon monoxide (D_LCO) and alveolar volume (V_A) simulation data from multiple devices and laboratories compared to European Respiratory Society (ERS)/American Thoracic Society (ATS) acceptability ranges. The right panels show box-and-whisker plots of D_LCO and V_A simulation data from different devices assessed with the Kruskal–Wallis (one-way ANOVA) test and Dunn's multiple comparison post-test. The red lines in the right panels represent the measured 3 standard deviation range. The solid lines in the lower panels represent the V_A simulation target and the dashed lines in all panels represent the ERS/ATS acceptability ranges. ATPD: gas conditions at atmospheric temperature, pressure, dry; Exp bag: expiratory bag; RGA: rapid gas analyser; GC: gas chromatograph.

The measured 3-sd V_A range for the RGA, GC and pooled data matched the ERS/ATS-recommended variance of ±0.3 L, and were 10.4%, 10.3% and 10.3% of the mean, respectively. However, the 3-sd V_A range for the RGA, GC and pooled data were offset below the ERS/ATS fixed range (3±0.3 L) because the mean V_A for these devices were <3 L: 2.89, 2.92 and 2.91 L, respectively.

The measured 3-sd range for the expiratory bag devices (±0.15 L, 5% of the mean 2.98 L) was tighter than the ERS/ATS range. Using the ERS/ATS V_A range for these devices would create acceptability boundaries of −5.6 and +6.4 sd from the measured mean. Using a ±3-sd range may be more appropriate than the fixed ERS/ATS V_A range for devices with less variance.

Comparison of D_LCO and V_A simulation data between device types revealed statistically significant differences between all device types (figure 1, right panels). While it is unclear if these differences are clinically important, they may be important with regards to establishing acceptable ranges for V_A simulation.

Our data support the limits of acceptability for D_LCO simulation (<0.5 mL·min⁻¹·mmHg⁻¹) as recommended by the ERS/ATS D_LCO standards. However, a potential limitation of this recommendation is that there is no consideration of limits on negative D_LCO values which may indicate gas analyser malfunction. Indeed, all four D_LCO outliers that were removed from this study were negative values (expiratory bag: −0.32, −0.32; GC: −0.34, −0.30) but satisfy the ERS/ATS single-sided limit of <0.5 mL·min⁻¹·mmHg⁻¹. While a V_A range of ±0.3 L perfectly fit the 3-sd range for the RGA, GC and pooled data, the mean V_A were <3 L, resulting in an offset of the 3-sd range from the ERS/ATS range. However, the 3-sd V_A range from the expiratory bag devices were significantly tighter than the ERS/ATS recommended ranges.

It is possible that a methodological difference might be responsible for the varying results between devices. For example, the expiratory bag devices performed the D_LCO simulation with a full syringe of test gas in patient testing mode while the RGA and GC devices used in this study performed the D_LCO simulation with 1 L air mixed with 2 L test gas in a simulation test mode. Factors that might cause the simulated V_A to be <3 L include the volume of dead space correction applied to either the syringe or the D_LCO system, as well as the mechanism supplying the inspired gas (e.g. reservoir bag or demand valve). Manufacturers must perform these simulations on their equipment and provide guidance for their customers to achieve ERS/ATS-compliant simulation data.

Performing a D_LCO simulation test weekly, or whenever a problem is suspected, provides a quick means for troubleshooting a complex system. Any simulation data outside of the acceptability limits should prompt users to take the system out of service until the problem can be resolved. This requires that the acceptability limits accurately represent the performance of the D_LCO system under normal operating conditions. Subtle D_LCO system malfunctions may escape detection by calibration procedures so failure to perform weekly D_LCO simulation and biological control testing risks reporting inaccurate patient data. Because D_LCO thresholds are used in many clinical scenarios, inaccurate data may have serious consequences for patients [4, 5].

Our data, collected from 13 devices and four clinical laboratories under real-world conditions, indicate that the ERS/ATS acceptability limits for D_LCO simulation are appropriate. However, a limit on negative values should be added. For many of the devices we examined, the absolute V_A ranges fit the ERS/ATS recommendation but were offset below 3±0.3 L because the mean V_A was offset below 3 L. Unfortunately, our study was limited to a few devices; additional research on other D_LCO systems and the cause of the mean V_A offset is needed. Manufacturers must ensure that their systems produce simulation data that fit the ERS/ATS targets. A V_A range ±3 sd may be more appropriate for devices with a measured range tighter than ±0.3 L. Based on the data presented, we suggest that future ERS/ATS D_LCO technical standards recommend a D_LCO target that accounts for negative values and a V_A target of 3±0.3 L or ±3 sd, whichever is smaller.