Abstract
Pulmonary magnetic resonance imaging using hyperpolarised 129Xe gas (XeMRI) can quantify ventilation inhomogeneity by measuring the percentage of unventilated lung volume (ventilation defect per cent (VDP)). While previous studies have demonstrated its sensitivity for detecting early cystic fibrosis (CF) lung disease, the utility of XeMRI to monitor response to therapy in CF is unknown. The aim of this study was to assess the ability of XeMRI to capture treatment response in paediatric CF patients undergoing inpatient antibiotic treatment for a pulmonary exacerbation.
15 CF patients aged 8–18 years underwent XeMRI, spirometry, plethysmography and multiple-breath nitrogen washout at the beginning and end of inpatient treatment of a pulmonary exacerbation. VDP was calculated from XeMRI images obtained during a static breath hold using semi-automated k-means clustering and linear binning approaches.
XeMRI was well tolerated. VDP, lung clearance index and the forced expiratory volume in 1 s all improved with treatment; however, response was not uniform in individual patients. Of all outcome measures, VDP showed the largest relative improvement (−42.1%, 95% CI −52.1–−31.9%, p<0.0001).
These data support further investigation of XeMRI as a tool to capture treatment response in CF lung disease.
Abstract
Hyperpolarised 129Xe MRI is emerging as an important imaging biomarker in cystic fibrosis. Ventilation distribution, measured using this technique, improves after treatment of a pulmonary exacerbation in children with cystic fibrosis. http://ow.ly/E6Gc30nFMlb
Introduction
Overall survival and lung function have improved dramatically over the past decades in patients with cystic fibrosis (CF) [1]. Traditional pulmonary function tests (PFTs) such as spirometry are now often normal in the paediatric age range [2]. As a consequence, the use of spirometric indices as endpoints in clinical trials or for monitoring individual response to therapy, especially in the paediatric population, can be problematic [3]. The lung clearance index (LCI), an outcome measure generated from the multiple-breath washout (MBW) test, reflects ventilation inhomogeneity [4, 5] and has been demonstrated to be more sensitive than spirometry for detecting and monitoring early lung disease in paediatric CF patients [6–8]. Despite its higher sensitivity, MBW testing provides only a whole-lung assessment of ventilation inhomogeneity, and imaging-based measures of lung function may be better at providing insight into the regional nature of disease and detecting subtle, regional pathology.
Hyperpolarised noble gas functional pulmonary magnetic resonance imaging (MRI) has shown considerable advancement over the past decade for the quantification of regional ventilation defects [9], alveolar–capillary gas diffusion [10] and pulmonary microarchitecture [11]. While 3He was previously the most commonly used inhaled contrast agent in this field, more centres are now using hyperpolarised 129Xe for functional pulmonary MRI (XeMRI) owing to the lower cost and higher availability of 129Xe as well as advances in hyperpolarisation physics that have allowed for greater polarisation efficiency of this gas [12]. The most validated hyperpolarised gas MRI technique is static ventilation imaging, which allows for the quantification of ventilation “defects” as a percentage of the total thoracic volume during a breath hold, or ventilation defect per cent (VDP). VDP can be determined using various analytical techniques [9, 13, 14] and has been demonstrated to be more sensitive than spirometry at identifying lung disease in paediatric CF patients [13, 15–17]. In addition, VDP appears to be more sensitive than spirometry and MBW at detecting longitudinal progression of CF lung disease [18]. However, high sensitivity of a test does not necessarily imply responsiveness of that test to treatment. Previous studies using 3He MRI (HeMRI) outcomes in interventional settings in CF have demonstrated treatment responsiveness of this technique [19–21]. To date, no studies have investigated the ability of XeMRI to assess the efficacy of a therapeutic intervention in the CF population. If this imaging modality is to be used to direct clinical care or as an outcome measure in clinical trials, its ability to detect response to treatment must be further characterised.
In this study, we used a common clinical scenario in CF, in which improvement with treatment is expected, to assess whether XeMRI can detect a treatment response. We performed XeMRI and physiological (spirometry, plethysmography and MBW) testing before and after treatment in children with CF admitted to hospital for treatment of a pulmonary exacerbation. We used this known treatment paradigm to determine 1) whether XeMRI is safe and feasible in a sicker population of paediatric CF patients, 2) if XeMRI can detect changes in regional ventilation distribution following this treatment, and 3) how changes in VDP correlate with changes in spirometric and MBW outcome measures.
Methods
Patient population
Patients with CF between the ages of 8 and 18 years admitted to hospital for intravenous antibiotic treatment of a physician-diagnosed pulmonary exacerbation were eligible. Exclusion criteria were the inability to perform PFTs or MRI, pregnancy, supplemental oxygen use or a forced expiratory volume in 1 s (FEV1) <40% of predicted at the time of screening. Local ethics board and Health Canada approval was obtained (SickKids REB #1000049033). The study was registered at ClinicalTrials.gov (NCT02606487). The responsible physician determined the treatment regimen. Healthy controls were recruited as part of a separate study (SickKids REB #1000048243, ClinicalTrials.gov NCT02740868).
Study protocol
On the day of testing, inhaled medications were held until testing was complete. Akron pulmonary exacerbation severity (APES) scores were assessed at the time of recruitment [22]. A Likert scale of wellness (1–10) was collected before and after treatment. PFTs, MBW and MRI were performed on the same day within 48 h of the initiation and completion of antibiotic treatment. Spirometry and plethysmography were performed according to American Thoracic Society/European Respiratory Society (ATS/ERS) standards [23, 24]. Nitrogen MBW testing was performed on the Exhalyzer D (Ecomedics, Dürnten, Switzerland) in supine and seated postures, according to standard quality control criteria [5, 25].
Hyperpolarised 129Xe gas was generated as described (Model 9800; Polarean, Durham, NC, USA) and was dosed at 10% of the participant's total lung capacity (TLC), diluted in ultra high purity nitrogen to a total volume of 1 L for each dose as previously described [26]. XeMRI and conventional proton images were acquired at 3T (Siemens Prisma, Erlangen, Germany) in the coronal plane using a flexible vest radiofrequency coil (for XeMRI images; Clinical MR Solutions, Brookfield, WI, USA) or a flexible torso array and spine coil (for proton images; Siemens, Erlangen, Germany). The detailed MRI protocol is described in the supplementary material.
Image analysis
Seven slices from each image set (the centre slice plus three flanking slices) were chosen for VDP analysis. XeMRI and proton MRI images were aligned using a semi-automated segmentation approach [26]. VDP was calculated using two approaches: the k-means clustering approach described by Kirby et al. [9] and the mean-anchored, linear binning approach described by Collier et al. [27]. The VDP threshold in the linear binning technique was estimated from the mean signal intensity distribution of the 10 healthy controls (supplementary material).
Statistical analysis
Statistical analyses were performed using STATA v14 (Stata Corp., College Station, TX, USA). Within-subject absolute changes in outcome measures were post-treatment values minus the pre-treatment values. Within-subject relative changes were the absolute changes divided by the pre-treatment values. Within-subject absolute and relative outcome measure changes are reported with 95% confidence intervals. Correlation of outcome measures was assessed with simple linear regression and the coefficient of determination (R2) is reported.
Results
Patient population
Between December 2016 and August 2017, 33 children with CF aged between 8 and 18 years admitted to the Hospital for Sick Children for treatment of a pulmonary exacerbation were screened; 20 were enrolled and 15 completed the study (figure 1). Of the enrolled subjects, one could not perform MBW testing, one was unable to correctly complete the breath-hold manoeuvre for XeMRI, and three had their diagnoses of pulmonary exacerbation revised by their treating physicians and were discharged from hospital prior to completing a full course of antibiotic therapy. Data from the 15 participants who completed the study were included in the final analysis and their characteristics are shown in table 1. The median age was 14 years, and baseline lung function was relatively preserved, with a median best FEV1 % pred in the 6 months preceding the pulmonary exacerbation of 85.0% (interquartile range (IQR) 61.0–93.0%).
Study flow diagram demonstrating screening, enrolment and study completion. FEV1: forced expiratory volume in 1 s; MBW: multiple-breath washout; XeMRI: magnetic resonance imaging using hyperpolarised 129Xe gas.
Baseline and admission characteristics of the participants (n=15)
At the time of admission, the median APES was 11.0 (IQR 8.0–14.0). Mean±sd FEV1 % pred at the time of admission was 62.9±13.9%, which represented a median relative drop of 18.0% (IQR −8.0–−31.0%) from baseline (best FEV1 in the past 6 months). Antibiotic treatment duration was 13 or 14 days for 14 out of 15 participants and 22 days for one participant.
XeMRI tolerability and feasibility
The gas dosing volumes used in this study, as a fraction of TLC, are shown in supplementary table S1. All test procedures were well tolerated. Transient, self-resolving oxygen desaturation (arterial oxygen saturation measured by pulse oximetry (SpO2) of <88% lasting up to 10 s) was observed in 15 out of 30 XeMRI scans (50%). The median oxygen saturation nadir for all scans was 88% (IQR 81–91%). No desaturations lasted longer than 10 s and no scans were aborted owing to adverse events. Analysable images were acquired in all but one scanned participant. Representative ventilation defect maps generated using the k-means and linear binning techniques and the corresponding signal histograms are shown in figure 2.
Examples of images from magnetic resonance imaging (MRI) using hyperpolarised 129Xe gas (XeMRI), ventilation defect per cent (VDP) mask overlays and a XeMRI signal intensity histogram from the same participant. a) Unprocessed XeMRI signal. b) Colourised proton MRI (purple) and XeMRI (green) acquisitions after registration (i.e. alignment). c, d) Ventilation maps calculated using the k-means (c) and linear binning (d) techniques demonstrating ventilated (green) and unventilated (red) lung. e) XeMRI signal intensity histogram plotting the distribution of normalised voxel intensity, with low signal intensity being shown on the left; the mean healthy intensity distribution is overlaid (dashed line) and the calculated VDP threshold is demonstrated by the vertical line.
Imaging and physiological responses to antibiotic treatment
Imaging, spirometric, plethysmographic, MBW and symptom score outcomes all improved significantly following treatment (table 2). VDP (calculated using the linear binning technique) showed the greatest mean relative improvement of all outcome measures (−42.1%, 95% CI −57.3–−26.8%, p<0.0001), dropping from a mean of 6.8% to 3.8%. In most participants, the distribution of XeMRI ventilation signal intensity shifted towards the healthy range (the derivation of the healthy distribution can found in supplementary figure S1) but did not completely return to a healthy distribution in any case (supplementary figure S2). Seated LCI showed a significant mean improvement following treatment (−9.1%, 95% CI −15.5–−2.6%, p=0.009), but there was no overall change observed in supine LCI (−2.7%, 95% CI −9.6–4.2%, p=0.4). The distribution of individual relative changes in imaging and physiological outcomes is shown in figure 3. Absolute FEV1 and LCI values correlated with VDP (supplementary table S2). However, there was no correlation between the magnitude of change of the imaging and physiological outcomes.
Physiological and imaging outcome measures before and after treatment
Relative improvement of lung clearance index (LCI), forced expiratory volume in 1 s % predicted (FEV1 % pred) and ventilation defect per cent (VDP) (k-means and linear binning techniques) following treatment. Individual relative changes (100% × [pre-treatment value–post-treatment value]/pre-treatment value) are shown by hollow circles. Improvement is always depicted as a positive change regardless of the nature of the outcome measure. The dashed horizontal line indicates “no change”. The mean relative change (solid circle) and 95% confidence intervals are shown.
Unprocessed XeMRI data, signal intensity histograms and physiological data are shown from three cases that exemplify the complementarity of the data obtained using the techniques (figure 4). In patient 20, there was an improvement in FEV1, LCI and VDP (figure 4a). Qualitative analysis of the XeMRI signal histogram generated by the linear binning technique demonstrated resolution of a low-intensity signal plateau seen at the left half of the histogram (figure 4b). In patient 9, treatment resulted in an increase in FEV1, while the LCI and VDP worsened (figure 4c). Qualitative assessment of the XeMRI signal histogram showed a distinct distribution pattern with a very high frequency of lower intensity signal (suggesting widespread regions of poorly ventilated lung) that did not change substantially with treatment (figure 4d). Finally, in patient 13, FEV1 did not improve following treatment, but both VDP and LCI decreased (figure 4e), and the post-treatment XeMRI signal intensity histogram shifted towards the normal range with reductions in focal defects on the XeMRI images (figure 4f).
Pre- and post-treatment ventilation images from magnetic resonance imaging (MRI) using hyperpolarised 129Xe gas (XeMRI) (three slices shown) (a, c, e) and signal intensity histograms (b, d, f). a, b) Participant 20; c, d) participant 9; e, f) participant 13. b, d, f) Pre-treatment distribution is shown in dark grey and post-treatment distribution is shown in white; the mean healthy distribution is overlaid with a dashed curve and the calculated ventilation defect per cent (VDP) threshold is identified by a dashed vertical line. FEV1: forced expiratory volume in 1 s; LCI: lung clearance index.
Discussion
In this study, we assessed the ability of pulmonary XeMRI to detect changes in ventilation distribution following inpatient treatment of a CF pulmonary exacerbation. The study procedure was well tolerated in all cases. We demonstrate responsiveness of this outcome measure, reflected by a significant decrease in poorly ventilated lung (quantified by VDP) following treatment. The change in VDP signal was substantial, with a relative improvement of ∼40%, suggesting that this technique can generate outcomes that are robustly responsive to treatment regardless of the image analysis technique that is used. Importantly, the change in VDP did not correlate with changes in MBW or spirometry outcomes, suggesting that these modalities provide complementary information.
To our knowledge, this is the largest study to investigate hyperpolarised gas functional pulmonary MRI as a tool to monitor treatment response in CF, and the first using XeMRI. One other study has assessed the ability of hyperpolarised HeMRI to detect a treatment response to a pharmacological intervention. In a double-blind study of eight CF patients with gating mutations, Altes et al. [19] demonstrated that HeMRI can identify a decrease in ventilation defects after treatment with ivacaftor, a cystic fibrosis transmembrane conductance regulator (CFTR) modulator. Previous HeMRI studies failed to demonstrate a net treatment response to airway clearance therapies using global ventilation defect measures, though the location of the defects changed with treatment [20, 21]. Advanced regional XeMRI image analysis techniques [28] may be able to provide a more detailed physiological quantification of treatment responsiveness; however, data from the ivacaftor study and the current trial suggest that global hyperpolarised gas MRI measures like VDP are responsive to changes in pulmonary ventilation distribution following treatment in CF. More evidence in the interventional setting is needed to define how responsive this technique is to treatment, compared to other available testing modalities.
Another potential clinical role for hyperpolarised gas MRI may be in longitudinal monitoring of CF lung disease. The rate of FEV1 decline in CF is low, even in adulthood [29], and spirometry is typically stably normal in younger children [30]. Techniques such as MBW are more sensitive than spirometry for detecting the progression of lung disease over time in young children [30], but recent data have suggested that HeMRI (and VDP in particular) may be even more sensitive than MBW for detecting the longitudinal progression of disease [18]. Especially in the era of CFTR modulators, in which we expect the annual rate of lung function decline to be further diminished (or potentially halted) [31], highly sensitive tools to monitor lung function progression over time will become increasingly important to clinical care.
In this study, we confirmed correlations between VDP and both LCI and FEV1 [15, 17]. However, the correlation between VDP and LCI was weaker in the current study than previously observed. This could be because the participants in this study had more severe lung disease, with lower FEV1 and higher LCI than our group's previous assessment of stable CF participants [15]. Additionally these children were also experiencing an acute pulmonary exacerbation. These differences could result in higher variability of measures of ventilation inhomogeneity [32, 33]. Contrastingly, a recent study by Smith et al. [17] interrogating regional physiology in children with CF using HeMRI and MBW techniques showed good correlation between VDP and LCI, even in those with worse lung function. However, in this same paper, the authors also demonstrated that as imaged lung volume approached TLC, measured VDP decreased. Measurements in our study were obtained at a lung volume corresponding to ∼80% of TLC (supplementary table S1), which is lower than the 60% of TLC that was targeted by Smith et al. [17]. It is therefore possible that imaging at a higher lung volume introduced a “ceiling effect” and that VDP was relatively underestimated, especially in smaller patients, thereby decreasing imaging–physiology correlation. This same phenomenon may also have blunted the observed magnitude of treatment response. Gas contrast dosing in future studies should be based on subject lung volumes to minimise this potential confounding effect.
Interestingly, while the absolute physiological and imaging outcome measures were correlated, the magnitude of changes in VDP did not correlate with changes in LCI or FEV1, suggesting that these measures are providing complementary information about treatment response. Examples of individual cases with conflicting results for spirometry, MBW and XeMRI results (figure 4) demonstrate the complementary nature of these tests. Future studies should assess the relationship and combined utility of XeMRI-derived and physiological outcome measures in predicting long-term outcomes and/or disease progression to better define which measurement(s) should be used to best guide treatment decisions.
Somewhat surprisingly, we found that VDP correlated slightly better with seated MBW outcomes (supplementary table S3) than those collected in the supine posture (the posture in which the MRI was performed). This is contrary to recent data that have shown that structural CT scores correlate better with MBW taken in the supine posture [34], though Smith et al. [17] also observed slightly better correlation of VDP with seated LCI in their recent paper. The aetiology of this observation is not entirely clear but may again be related to the lung volume at which the XeMRI images were obtained. While the supine posture reduces resting lung volume, performing imaging above the functional residual capacity likely shifts the imaged lung volumes closer to those seen in the sitting position.
One final observation of this study was the impact of disease stability on the postural dependence of MBW outcomes. There is existing evidence that the posture in which MBW testing is performed influences the outcome measures that are generated, with an increase in ventilation heterogeneity seen in the supine posture [35, 36]. We confirmed this finding in the current study, showing that the LCI was significantly higher and the functional residual capacity significantly lower in a supine posture than in a seated posture. This postural dependence of MBW outcomes was diminished (or in the case of LCI, completely abolished) in the pre-treatment participants, but was recovered following therapy (supplementary table S2). These findings are similar to those of Smith et al. [35], who found that children with CF had less predictable posture-driven changes in MBW outcomes than those with no lung disease. We hypothesise that mucous plugs, which are a major radiographic feature of CF patients experiencing an exacerbation [37, 38], may shift unpredictably with postural changes, leading to less predictable changes in regional ventilation patterns in sicker patients as they change posture.
Strengths of this study include its prospective design, and the same-day measurement of multiple outcome measures including a standardised approach to the collection and analysis of XeMRI data. The primary limitation of this study is that the intervention studied is not necessarily a scenario in which XeMRI is likely to be used clinically. Similar to what has been done for other outcome measures, pulmonary exacerbations were chosen as a first step of evaluation because of a predictable improvement following treatment in most outcome measures on a group level [39]. Now that XeMRI has been shown to be responsive to treatment in this setting, future studies can assess its responsiveness to other interventions in which smaller signals in traditional PFTs are typically seen [7, 40]. Additionally, structural analysis was not performed in this study. Significant advancements in traditional proton MRI techniques have also been able to demonstrate responses in structural and perfusion scores following therapeutic interventions in CF, though this responsiveness is variable [37, 41, 42]. The combination of structural and functional MRI techniques will provide important future insights into CF disease mechanisms and treatment responsiveness [43] and should be considered in future studies investigating treatment responsiveness of MRI techniques. Finally, in contrast to current structural MRI techniques [41], the XeMRI performed in this study was limited to participants who were capable of performing an adequate breath-hold manoeuvre. While one study has demonstrated some success in free-breathing infants with HeMRI [44], this remains a significant drawback of this technique.
In conclusion, this study has shown that that XeMRI is responsive to treatment of pulmonary exacerbations in paediatric CF participants, with robust decreases observed in VDP. This demonstrates that XeMRI shows promise as an emerging imaging biomarker for monitoring treatment response in CF. In addition, XeMRI has a potential role as a research tool to better understand the complex and spatially heterogeneous pathophysiology of CF lung disease.
Supplementary material
Supplementary Material
Please note: supplementary material is not edited by the Editorial Office, and is uploaded as it has been supplied by the author.
Supplementary material ERJ-02188-2018.Supplement
Footnotes
This article has supplementary material available from erj.ersjournals.com
Author contributions: All authors have contributed significantly to the work, have participated in drafting or critically revising the manuscript, have seen and approved the final version of the manuscript, and take responsibility for the content of the manuscript.
Conflict of interest: J.H. Rayment has nothing to disclose.
Conflict of interest: M.J. Couch reports partial salary support from Siemens Healthcare through the MITACS Elevate postdoctoral fellowship programme, during the conduct of the study.
Conflict of interest: N. McDonald has nothing to disclose.
Conflict of interest: N. Kanhere has nothing to disclose.
Conflict of interest: D. Manson has nothing to disclose.
Conflict of interest: G. Santyr has nothing to disclose.
Conflict of interest: F. Ratjen has nothing to disclose.
Support statement: This study was funded by The Irwin Fund and a Canadian Institutes for Health Research Project grant #376120. J.H. Rayment was funded by a Cystic Fibrosis Canada clinical fellowship and M.J. Couch was supported by a MITACS Elevate award. Funding information for this article has been deposited with the Crossref Funder Registry.
- Received November 16, 2018.
- Accepted February 2, 2019.
- Copyright ©ERS 2019
References
