Computed cardiopulmonography and the idealized lung clearance index, iLCI2.5, in early-stage cystic fibrosis

Abstract

This study explored the use of computed cardiopulmonography (CCP) to assess lung function in early-stage cystic fibrosis (CF). CCP has two components. The first is a particularly accurate technique for measuring gas exchange. The second is a computational cardiopulmonary model where patient-specific parameters can be estimated from the measurements of gas exchange. Twenty-five participants (14 healthy controls, 11 early-stage CF) were studied with CCP. They were also studied with a standard clinical protocol to measure the lung clearance index (LCI_2.5). Ventilation inhomogeneity, as quantified through CCP parameter σlnCl, was significantly greater (P < 0.005) in CF than in controls, and anatomical deadspace relative to predicted functional residual capacity (DS/FRCpred) was significantly more variable (P < 0.002). Participant-specific parameters were used with the CCP model to calculate idealized values for LCI_2.5 (iLCI_2.5) where extrapulmonary influences on the LCI_2.5, such as breathing pattern, had all been standardized. Both LCI_2.5 and iLCI_2.5 distinguished clearly between CF and control participants. LCI_2.5 values were mostly higher than iLCI_2.5 values in a manner dependent on the participant's respiratory rate (r = 0.46, P < 0.05). The within-participant reproducibility for iLCI_2.5 appeared better than for LCI_2.5, but this did not reach statistical significance (F ratio = 2.2, P = 0.056). Both a sensitivity analysis on iLCI_2.5 and a regression analysis on LCI_2.5 revealed that these depended primarily on an interactive term between CCP parameters of the form σlnCL*(DS/FRC). In conclusion, the LCI_2.5 (or iLCI_2.5) probably reflects an amalgam of different underlying lung changes in early-stage CF that would require a multiparameter approach, such as potentially CCP, to resolve.NEW & NOTEWORTHY Computed cardiopulmonography is a new technique comprising a highly accurate sensor for measuring respiratory gas exchange coupled with a cardiopulmonary model that is used to identify a set of patient-specific characteristics of the lung. Here, we show that this technique can improve on a standard clinical approach for lung function testing in cystic fibrosis. Most particularly, an approach incorporating multiple model parameters can potentially separate different aspects of pathological change in this disease.

Keywords: laser absorption spectroscopy; log-normal lung; lung function testing; multibreath washout; nitrogen washout.

Graphical abstract

Figure 1.

Schematic illustrating model and parameter estimation technique. There are three submodels, the first is a model for the blood gas dissociation curves, the second is a model of the circulation and body gas stores, and the third is a model of the lung that includes inhomogeneity (three of the total 125 compartments are illustrated). The size of the body gas stores and some of the initial parameter predictions are determined by the participant’s physical characteristics. The model is driven through metabolic consumption of O₂ and production of CO₂. This, along with body size, determines cardiac output. The respiratory flows recorded by the molecular flow sensor drive the ventilation of the lungs, and the inspiratory composition of the gas is set to match that recorded with the molecular flow sensor. Simulated expiratory compositions are then calculated during the execution of the model. These give rise to simulated molar flows for each gas species during expiration and can be compared with those measured by the molecular flow sensor. A nonlinear optimization routine then adjusts the parameters between successive runs of the model so as to minimize the error between the simulated and experimentally recorded molar flows for the different gases during expiration. Cd_i, vascular conductance for the i^th lung unit; CL_i, compliance for the i^th lung unit; DS_i deadspace for the i^th lung unit; DSMFS apparatus deadspace for the molecular flow sensor; VA_i, alveolar volume for the i^th lung unit.

Figure 2.

Example of the fit of the model to the data. A: measured and model-simulated cumulative tidal flows for the different gas species at the mouth. The switch from breathing air to breathing pure O₂ occurs at ∼11 min. The model responses are essentially overlaid on the data. B–D: 1-min periods on an expanded scale from the record in A, illustrating the results during the air-breathing phase, and early and late in the N₂ washout phase. E: cumulative residuals (measured minus simulated) illustrating the residual error in the model fit for the gas exchange data. Data are from participant number 17, who was a 22-yr-old female with cystic fibrosis. The sum of squared errors for the fit was 0.078 (L/s)². STPD, standard temperature and pressure, dry.

Figure 3.

Comparison for selected parameters between the healthy control (HC) and cystic fibrosis patient (CF) groups. A: plot of the measured forced expiratory volume in one second (FEV₁) % predicted for the HC (n = 14) and CF (n = 9) groups; n represents number of participants. B–E: plots showing parameters determined by computed cardiopulmonography for the HC (n = 14) and CF (n = 11) groups. Functional residual capacity (FRC) has been plotted as a percentage of the predicted value. Data are shown for each individual, as well as boxes illustrating the interquartile range, together with a horizontal line indicating the mean value. The whiskers illustrate the spread of data outside of the interquartile range and extend to the lowest or highest data point that is within three times the interquartile range of the lower or upper border of the box. Data from the older (>30 yr) participants with CF are indicated by circles surrounding the symbols. Data from participants with CF with enlarged deadspace (C) are indicated by the use of a dot within the symbols. The statistical significance of differences between the groups was determined using a Welch t test. DS, end-inspiratory deadspace; CF, cystic fibrosis; F, female; FRCpred, predicted FRC; σlnCl, standard deviation for the natural logarithm of the standardized lung compliance; HC, healthy control; M, male; σVD, standard deviation for the standardized deadspace.

Figure 4.

Calculation of the idealized lung clearance index from the simulated tracer gas washout protocol. A: alveolar volume against time, throughout the simulated protocol, showing the regular breathing pattern. B: cumulative volume uptake of the theoretical tracer gas against time. This oscillates to and from zero for the first 2.5 min as it is breathed in and out, after which it is progressively washed out by air (which contains no tracer gas). C: plot during the washout phase of the tracer gas fraction at the mouth (normalized relative to the starting fraction) against the cumulative expired volume.

Figure 5.

Repeatability of measurements made with the Ecomedics Exhalyzer D (ExD) device and with computed cardiopulmonography (CCP). A and B: repeatability of FRC measurement made with the ExD (HC, n = 13; CF, n = 6) and with CCP (HC, n = 11; CF, n = 8), respectively; n represents number of participants. C and D: repeatability of the lung clearance index (LCI_2.5) made with the ExD and the idealized-LCI_2.5 (iLCI_2.5) made using CCP, respectively. Patient numbers and statistical comparisons, as for A and B. Shaded areas indicate the 95% confidence intervals for the mean and for the values for ±1.96 SD. CCP, computed cardiopulmonography; CF, cystic fibrosis; FRC, functional residual capacity; HC, healthy control.

Figure 6.

Effect of varying the simulated breathing pattern on the values calculated for the idealized lung clearance index, iLCI_2.5. Participant numbering is ordered by value of iLCI_2.5 at a respiratory rate of 8 breaths per minute. Vertical broken line separates the HC participants to the left, and participants with CF to the right. The tidal volume of 10.5 mL/kg of ideal body weight is the midpoint of the range (8–13 mL/kg) considered acceptable for the standard clinical measurement of LCI_2.5. Also shown are the experimentally determined values for LCI_2.5 from the standard washout procedure where the breathing frequency differed between participants. There was a positive correlation (r = 0.46, P = 0.031) between the value for the LCI_2.5 relative to the iLCI_2.5 (at 8 breaths/min) and the participant’s breathing frequency during the LCI_2.5 measurement. CF, cystic fibrosis; HC, healthy control; RR, respiratory rate; VT, tidal volume.

Figure 7.

Comparison of FRC and LCI_2.5 measured using ExD with FRC and iLCI_2.5 determined using CCP. A: comparison of FRC values between the two techniques (HC, n = 13; CF, n = 9). n represents number of participants. B: comparison of iLCI_2.5 with LCI_2.5 (HC, n = 13; CF, n = 9). C: box and whisker plots as for Fig. 2 comparing FRC values for HC and CF groups from ExD (HC, n = 13; CF, n = 9) and from CCP (HC, n = 14; CF, n = 11). D: box and whisker plots comparing values for ExD LCI_2.5 with those for CCP iLCI_2.5 for HC and CF groups. Participant numbers as for C. CCP, computed cardiopulmonography; CF, cystic fibrosis; ExD, Ecomedics Exhalyzer D; FRC, functional residual capacity; iLCI_2.5, idealized lung clearance index; HC, healthy control; LCI_2.5, lung clearance index.

Figure 8.

Breakdown of the contributions of different aspects of lung inhomogeneity to overall values for iLCI_2.5. A: example of the simulated washout profiles, showing normalized tracer gas fraction against the expired volume measured in lung turnovers. The different records are for 1) a homogenous lung with no deadspace (DS); 2) a lung with DS but without either deadspace inhomogeneity (σVD) or ventilation inhomogeneity (σlnCL); 3) a lung with DS and σVD, but without σlnCL; 4) a lung with σlnCL but with no DS; and 5) a lung with all inhomogeneities present. Horizontal broken line indicates 1/40^th of starting fraction for tracer gas. B and C: stacked bar charts illustrating the contribution of each aspect of lung inhomogeneity to the total iLCI_2.5 for each participant in absolute terms in B, and in percentage terms in C. Participants have been numbered in ascending order of iLCI_2.5. The older participants with CF are numbers 19, 21, 22, and 25. The vertical line indicates the division between the HC group (participants 1–14) to the left, and the CF group (participants 15–25) to the right. CF, cystic fibrosis; HC, healthy control.

Figure 9.

Relationship between ExD LCI_2.5 and selected model variables from CCP. A: LCI_2.5 vs. FRC % predicted. B: LCI_2.5 vs. DS/FRCpred. C: LCI_2.5 vs. σlnCL. D: LCI_2.5 vs. σlnCL*(DS/FRC). Participants: HC, n = 13; CF, n = 9. Broken lines illustrate simple linear regressions, significant for B–D, but not A. Multiple linear regression was significant for σlnCL*(DS/FRC) (F = 147, P < 0.001) and DS/FRCpred (F = 7.5, P < 0.02), indicating interaction between different aspects of lung function is the most important overall determinant of LCI_2.5. CCP, computed cardiopulmonography; CF, cystic fibrosis; DS, deadspace; ExD, Ecomedics Exhalyzer D; FRC, functional residual capacity; HC, healthy control; LCI_2.5, lung clearance index.