Long term survival after lung transplantation is limited by the development of chronic allograft dysfunction which manifests as bronchiolitis obliterans (BO). BO consists of heterogeneously distributed areas of obliterated respiratory and terminal bronchioles that lead ultimately to a decline in forced expiratory volume in 1 second (FEV₁), graft failure and recipient death.^1–4 It is thought that earlier diagnosis and more timely treatment of BO could improve long term survival;^1–3,5,6 however, the heterogeneous distribution of BO within the transplanted lungs renders invasive diagnosis by transbronchial biopsy unreliable, with reported sensitivities as low as 17–28%.^3,7,8 In an attempt to identify BO earlier, a functional surrogate for this structural abnormality—bronchiolitis obliterans syndrome (BOS)—has been defined as a progressive decline in FEV₁.⁹ Unfortunately, identifying patients using BOS criteria still may not identify subjects early enough in the development of airflow limitation due to the distribution of the BO process. Investigators have recently turned their attention to computed tomographic (CT) scoring systems because it is thought that direct evaluation of anatomical markers may allow earlier detection of BO than indirect measurements such as FEV₁.

There are, however, limitations to CT scoring systems, predominately due to sensitivity and specificity for disease progression as well as the high inter-observer and intra-observer variability of the score itself. For example, it has been suggested that air trapping is the most sensitive and specific CT abnormality for the early detection of BOS^10–14 while exhibiting the highest inter-observer agreement.^14–16 However, more recent work has not always confirmed these findings¹⁷ and there is no consensus on how to score air trapping. For example, Bankier et al evaluate gas trapping as 0–20%, 20–40%, 40–60%, 60–80% or 80–100% of the lobe involved,^14,16 while Siegel et al score the lobes as 0%, 1–25%, 26–50%, 51–75%, and 76–100%.¹³ Furthermore, scores for mosaic pattern of attenuation, airway wall thickening, and bronchiectasis are individually less sensitive and specific than scores for air trapping,^10–14,17 and the intra- and inter-observer agreement individual scores for bronchiectasis, airway wall thickening, and mosaic pattern have not been evaluated in lung transplant recipients. Therefore, given the variety of CT abnormalities seen in BO/BOS, it may be that a composite CT score (CT_BO score) will be more sensitive and specific than a CT air trapping score (CT_AT score) alone for the early detection of BO.

The aims of the present study were (1) to determine the intra- and inter-observer agreement of a CT_BO score and a CT_AT score, (2) to determine the cross sectional relationship between the CT_BO score and CT_AT score with FEV₁, and (3) to relate the CT_BO score at baseline to changes in FEV₁ and changes in the CT_BO score over the course of 1 year. Our hypotheses were that (1) the CT_BO and CT_AT scores would show significant associations with FEV₁ in lung transplant recipients and (2) the CT_BO score at baseline would predict changes in FEV₁ and changes in the CT_BO score over the course of 1 year.

METHODS

Subjects

The baseline CT scan (first scan following transplantation) and the first annual surveillance CT (follow up) scan of 38 consecutive subjects who received a single or double lung transplant at our centre since 2000 were studied. CT scans were excluded when performed for the diagnosis of acute clinical events, when there were incidental CT findings of acute disease (such as pneumonia), or when the recipient had coincident clinical or bronchoscopic evidence of an acute event (acute rejection, infection). We did not routinely perform bronchial provocation testing on our lung transplant recipients. None of the subjects included in the study had clinical manifestations of asthma after transplantation and none had significant bronchodilator responses according to ATS criteria on spirometry. The study was approved by the clinical ethics review board of the University of British Columbia.

CT scans

All CT scans were performed on a GE Lightspeed Ultra scanner (General Electric Medical System, Milwaukee, WI, USA). Inspiratory images were acquired at suspended inspiration from lung apex to base using 1.25 mm slice thickness at 10 mm intervals. Expiratory 1.25 mm thick images were acquired at end of exhalation at the level of the aortic arch, the carina and 2 cm above the hemidiaphram. Images were acquired using 150 mA, 120 kVp, 1 second scan time, reconstructed using a high spatial frequency algorithm (“Bone”) and an appropriate field of view. For the first part of the study each baseline CT scan was assigned a random identification number, blinded for patient characteristics and reviewed using a medical imaging workstation (Leonardo Workstation, Siemens AG Medical Solutions, Erlangen, Germany). For the second part of the study each pair of baseline plus follow up scans was assigned another random identification number, blinded for patient characteristics and reviewed as per baseline.

Spirometric tests

Spirometric tests were performed according to ATS guidelines.¹⁸ For the purposes of this study, FEV₁ was expressed as a percentage of the average of the two best FEV₁ values obtained after lung transplantation.⁹ BOS was defined according to the International Society of Heart and Lung Transplantion guidelines with BOS stage 0 as FEV₁ >80% post-transplant baseline value and stages 1, 2 and 3 equivalent to FEV₁ 66–80%, 51–65%, and <50% post-transplant baseline, respectively.⁹

CT scoring

To establish the CT scoring system, two observers (PAJ, JDD) independently scored the baseline CT scans in a random and blinded fashion. To test for intra-observer variation, one observer re-scored all baseline CT scans after 1 month.

Our CT scoring system is presented in table 1 and illustrative examples are shown in fig 1. The CT scans were viewed using display settings of window, −500 Hounsfield Units (HU) and level, 1500 HU. Inspiratory scans were evaluated for severity and extent of central and peripheral bronchiectasis; extent of central and peripheral mucus plugging; severity and extent of central and peripheral airway wall thickening; extent of consolidation; and extent of mosaic pattern. Expiratory scans were evaluated for the extent of air trapping. Each of the five lobes (including the lingula as a sixth “lobe”) was evaluated separately using the inspiratory CT image while six lung zones (upper, middle, lower left and right) were scored using the expiratory images. In single lung transplant recipients, only the lobes of the transplanted lung were scored.

Abnormalities were defined according to recommendations of the nomenclature committee of the Fleischner Society. To score peripheral bronchiectasis and airway wall thickening, “peripheral” was defined as less than 2 cm from the costal and diaphragmatic pleura. Visible airways abutting the mediastinal pleura were scored as bronchiectasis. Peripheral mucus plugging was evaluated using the radiological appearance of “centrilobular nodules” or “tree-in-bud” pattern rather than using a peripheral location. Central mucus plugging was scored if mucus was seen in identifiable bronchi. Mild airway wall thickening was defined as an airway wall thickness greater than 2 mm in the hilum, 1 mm in the central, and 0.5 mm in the peripheral lung. Mild bronchiectasis was defined as a bronchial lumen diameter greater than the diameter of the adjacent pulmonary artery or as a lack of tapering between bronchial generations.

Scores for bronchiectasis, mucus plugging, airway wall thickening, air trapping, and a composite CT_BO score were calculated in a similar manner to Brody et al.¹⁹ In brief, for each lobe a bronchiectasis score, mucus plugging score, and airway wall thickening score were calculated by combining the abnormalities and severity of the abnormalities in the central and peripheral lung. Next, the lobe scores for bronchiectasis, mucus plugging, airway wall thickening, consolidation, mosaic pattern, and air trapping were summed to produce a total maximum of 108, 54, 36, 18, 18 and 18, respectively. The composite score was calculated by adding the component scores together for a total maximum of 252. The maximum total scores and maximum component scores were expressed on a scale of 0–100 for statistical analysis.

For the second part of the study, after 3 months the baseline CT scans were combined with the follow up CT scans, randomised, and scored using the above scoring system to assess the predictive value of the scoring system for disease progression. The readers did not have information as to which were baseline scans and which were follow up scans. The scores from this reading were also used to evaluate the intra-observer agreement.

Statistical analysis

Intra-observer and inter-observer agreement of scores for CT components, CT_AT score, and CT_BO score were calculated using intraclass correlation coefficients. An intraclass correlation coefficient of more than 0.8 represents good agreement.

Linear regression was used to model the association between CT score and FEV₁ measured at baseline. The regression analysis was conducted for the CT_BO score and the CT_AT score and for each observer separately. The regression coefficient was a measure of association, showing the mean decrease in CT score for every additional percentage change in FEV₁. The analysis was repeated for follow up measurements and was conducted for observers 1 and 2 separately. The linear regression was also used to model the association between the baseline CT_BO score and FEV₁ at follow up and the baseline CT_BO score and CT_BO score at follow up. Finally, the linear regression was used to model the association between the baseline CT_AT score and FEV₁ at follow up and the baseline CT_AT score and CT_AT score at follow up. The analysis was conducted for observers 1 and 2 separately.

A p value of <0.05 was considered significant and all data are presented as mean (SD, range) unless indicated otherwise.

RESULTS

Study population

The mean age at transplantation of the 38 lung transplant recipients included in the study was 43.7 (SD 13.1, range 12.7–64.6) years. The interval between transplantation and baseline CT scans was 44 (SD 33, range 2–120) months and the interval between baseline and follow up CT scans was 11.2 (SD 4.7, range 2.3–17.4) months. At the time of the baseline CT scan, 22, 10, 4 and 2 patients were in BOS stages 0, 1, 2 and 3, respectively, by spirometric criteria. One patient did not have a follow up CT scan and another was excluded from follow up analysis because of biopsy proven acute rejection. Other subject characteristics are given in table 2.

Inter- and intra-observer agreement

The inter-observer and intra-observer agreement for the CT_AT score, CT_BO score, and other component scores are shown in table 3. An intraclass correlation coefficient above 0.80 was considered to represent good agreement. The inter-observer agreement was above 0.80 for the CT_AT score and the bronchiectasis CT score. However the CT_BO score and consolidation, mucus plugging, airway wall thickening, and mosaic pattern component scores were below 0.80. The intra-observer agreement after 1 month was good for the CT_AT score and the CT_BO score but was below 0.80 for mucus plugging, airway wall thickening, and mosaic pattern component scores (table 3).

Relationship between baseline FEV₁ and CT_BO score or CT_AT score

There was a significant association between FEV₁ and both the CT_BO score and the CT_AT score measured at baseline (fig 2), with a higher (more damage) CT score corresponding to a lower (worse) FEV₁ value. The baseline CT_BO score increased by 0.20 (p = 0.0001, observer 1) and 0.26 (p<0.0001, observer 2) and the baseline CT_AT score increased by 0.55 (p = 0.02, observer 1) and 0.55 (p = 0.02, observer 2) for each percentage decrease in baseline FEV₁ (fig 2). The follow up CT_BO score increased by 0.25 (p<0.0001, observer 1) and 0.27 (p<0.0001, observer 2) and the follow up CT_AT score increased by 0.36 (p = 0.002, observer 1) and 0.63 (p<0.0001, observer 2) for each percentage decrease in follow up FEV₁.

We divided the subjects into a group without BOS (FEV₁ >80% baseline) and a group with BOS (FEV₁ <80% baseline) and arbitrarily set a CT score of >5 as abnormal. In patients without BOS, 12 of 22 had a CT score >5 (55% of patients with normal FEV₁ had an abnormal CT score). Air trapping alone was present in 19 of the 22 subjects without BOS (86%). In subjects with BOS, 13 of 16 had an abnormal CT score >5 (81%) and air trapping alone was present in 16 (100%).

Relationship between baseline CT_BO score and changes in CT_BO score and FEV₁

There were significant associations between the baseline CT_BO score and both FEV₁ and the CT_BO score measured after 1 year. The mean FEV₁ at follow up decreased by 1.55% (observer 1) or 1.37% (observer 2) of baseline for every additional unit in the CT_BO score at baseline (p<0.0001). The mean CT_BO score at follow up increased by 1.25 units (observer 1) or 1.12 units (observer 2) for every additional unit in the CT_BO score at baseline (p<0.0001).

Relationship between baseline CT_AT score and changes in CT_AT score and FEV₁

There were significant associations between the baseline CT_AT score and both FEV₁ and the CT_AT score measured after 1 year. The mean FEV₁ at follow up decreased by 0.27% (observer 1) or 0.24% (observer 2) of baseline for every additional unit in the CT_BO score at baseline (p = 0.0003 and p = 0.0004, respectively). The mean CT_BO score at follow up increased by 0.74 units (observer 1) or 0.68 units (observer 2) for every additional unit in the CT_BO score at baseline (both p<0.0001).

DISCUSSION

The aims of the present study were to determine the intra- and inter-observer agreement of the CT_BO score and CT_AT score, to determine the cross sectional association between the CT_BO score and CT_AT score with FEV₁, and to relate the CT_BO score at baseline to changes in FEV₁ and changes in the CT_BO score over the course of 1 year. Our hypotheses were that (1) the CT_BO score and CT_AT score would show significant associations with FEV₁ and (2) the CT_BO score at baseline would predict changes in FEV₁ and the CT_BO score over the course of 1 year. This study was not designed to determine if the CT_BO score is more useful than the CT_AT score.

Similar to previous studies,^14,16 our data show good inter-observer and intra-observer agreements for the CT_AT score. However, also similar to a previous study,¹³ the intraclass correlation coefficient for the composite CT_BO score in our study was borderline. This was related to the relatively low level of inter-observer agreement in scoring mucus plugging, airway wall thickening, and mosaic pattern. A number of factors may be responsible for the disagreements in scoring airway wall thickening and mosaic pattern. Firstly, most of our patients had mild (BOS-0 and BOS-1) airflow obstruction. As the CT scans showed only subtle abnormalities, this makes scoring more difficult than in cystic fibrosis where the abnormalities are more pronounced, although in cystic fibrosis studies the scores for airway wall thickening and mosaic pattern were also not very reproducible.^15,20 Secondly, although both observers had substantial expertise with interpretation and scoring of chest CT scans, they had limited experience of reading CT scans from lung transplant recipients. However, this situation may accurately reflect the typical clinical setting where chest CT scans are often read by radiologists with limited experience in lung transplant CT interpretation. Inter-observer agreement may be better for observers in large transplant centres who are more experienced in evaluating CT scans of lung transplant recipients. Alternatively, it may be best to combine these subjective scoring systems with a computerised analysis of lung parenchyma^21,22 and airways^23,24 which could combine the clinical impression with objective quantitative values.

The most important findings of this study are that both the CT_BO and CT_AT scores are significantly associated with FEV₁, and both scores predicted the clinical course of a patient over the year following the CT scan. A 1% higher CT_BO score at baseline predicts a 1.55% faster worsening in FEV₁ and a 1.25% faster worsening in the CT_BO score over the coming year (observer 1). Similarly, a 1% higher CT_AT score at baseline predicts a 0.27% faster worsening in FEV₁ and a 0.74% faster worsening in the CT_BO score over the coming year (observer 1). This finding suggests that both the composite CT_BO score and the CT_AT score could potentially identify BO earlier than FEV₁. We cannot determine from our study whether the CT_BO score is more useful than the CT_AT score, and it would therefore be prudent for future longitudinal studies evaluating the usefulness of subjective CT interpretation in BO to include a composite CT score as well as an air trapping CT score alone.

Nevertheless, these results support the concept that CT scanning is a valuable tool in the evaluation and follow up of lung transplant recipients. Lung function is currently the “gold standard” for detecting lung allograft dysfunction, but it is an indirect measurement and can only give a global assessment of the pulmonary condition. The major advantage of CT scanning is that it is a direct measure of lung structure and allows for the identification of structural abnormalities associated with chronic allograft dysfunction, including bronchiectasis, airway wall thickening, mucus in small and large airways, and air trapping due to small airway abnormalities. Furthermore, CT imaging allows for the detailed analysis of the regional distribution of pathological processes such as BO. This is particularly pertinent in single lung transplant recipients in whom physiological measures such as FEV₁ are confounded by the contribution from the native lung. Even in double lung transplant recipients, lung function tests may be insensitive in those with a heterogeneous distribution of damage, especially when the abnormalities are located in the most peripheral airways. For these reasons, we suggest that CT scanning could identify BO earlier than FEV₁, at a time when changes in immunosuppression may result in improved clinical outcomes. The use of clearly defined CT parameters (particularly a composite CT score that quantifies numerous lung components), possibly in combination with quantitative CT measures, may therefore have an important role as a standardised outcome for research trials involving BO.

In this study we did not analyse our data using the BOS 0-p stage. Because FEV₁ declines later in the disease process, this new category (BOS 0-p) was added based on forced expiratory flows between 25% and 75% of forced vital capacity (FEF_25–75).³ However, its prognostic usefulness has been debated,^17,25 with variable positive and negative predictive values reported. In view of this uncertainty, our statistical analysis was performed using FEV₁ as a continuous variable rather than being based on BOS stages, and hence this did not affect our analysis.

Potential limitations of the study include the relatively small study population, the short follow up, and the variation in timing of the baseline CT scan. A larger number of patients followed over a longer time frame would be useful to help characterise the potential role of CT scans for the early detection of chronic lung allograft dysfunction. Such a study could also examine the optimal interval between CT scans to detect BO earlier than FEV₁. Finally, our analysis may be limited by the fact that only three expiratory CT images were obtained and it may be advantageous in future studies to increase this number.

In conclusion, we systematically evaluated inter-observer and intra-observer agreement for qualitative CT scoring of a variety of abnormalities including a composite CT_BO score in a patient population with predominantly mild abnormalities. Both the composite CT_BO score and the CT_AT score had good or fairly good inter-observer and intra-observer agreement. Our findings indicate a potential role for a composite CT score, as well as an air trapping score alone, in lung transplant recipients.