Systemic sclerosis (SSc), also known as scleroderma, is a connective tissue disease characterized by vascular disease, immunologic abnormalities and fibrosis. The hallmark clinical feature consists of thickening of the skin. Internal organs such as the lungs, gastrointestinal tract and kidneys are also frequently affected. Pulmonary fibrosis is now a significant cause of morbidity and is the leading cause of mortality in patients with SSc [1].

High resolution computed tomography (HRCT) has become the gold standard for diagnosis of SSc related interstitial lung disease (SSc-ILD) especially for early stage disease. It is a much more sensitive diagnostic test than traditional chest radiograph, especially for detection of early or mild disease [2,3]. The most common radiologic pattern of lung disease in SSc is diffuse parenchymal lung disease characterized by prominent ground-glass opacities and fine interstitial reticular markings with lower lung predominance. The pattern is similar to that found in idiopathic non-specific interstitial pneumonia (NSIP) [4-6]. As the disease progresses, ground-glass opacifications get replaced with coarser interstitial reticulations, traction bronchiectasis and bronchiolectasis. Honeycombing also develops with time [7]. This pattern closely resembles idiopathic pulmonary fibrosis (IPF). Figure 1 illustrates the typical HRCT abnormalities that can be found in SSc- ILD.

Figure 1: Typical HRCT abnormalities in SSc-ILD. A. Pure ground glass opacities can be seen peripherally, predominantly in the right middle and left lower lobes; B. Subpleural thickened reticular markings are associated with traction bronchiectasis and bronchiolectasis; C. Ground glass opacities are mixed with fine reticular septal thickening diffusely; D. Extensive macrocystic honeycombing has replaced most of the left lower lobe and is associated with significant volume loss.

Different systems for evaluating SSc-ILD on HRCT have been developed over the past 20 years. Several scoring methods have been used to characterize and quantify the disease, correlate with common clinical parameters, prognosticate patients, assess disease progression and evaluate response to treatment. This article reviews the different published scoring systems and their applications.

Comparative scoring

Most of the scoring systems published for SSc-ILD have been developed from methods previously used to assess IPF. One of the first reading methods used to assess SSc-ILD was published by Wells et al. [8] (Table 1). They used a comparative grading system evaluating the extent of one HRCT abnormality (parenchymal disease) relative to another (reticular disease). The goal was to determine how different HRCT abnormalities correlate with lung histology, specifically inflammatory changes and fibrosis, on biopsy. This scoring method was based on a similar protocol published by Müller et al. [9] correlating findings on HRCT of IPF with histopathology.

Table 1: Comparative scoring method, Wells et al. [8].

An even simpler scoring method was published by Morelli et al. [15]. They divided the lungs into three zones: apices to main carina, main carina to inferior pulmonary venous confluence, and pulmonary veins to diaphragms, representing upper, middle and lower zones respectively. A score of 0 was given if no abnormality was found in a zone, a score of 1 for any abnormality found in SSc-ILD except for honeycombing (ground-glass, reticular markings, bronchiectasis) and a score of 2 if honeycombing was present. The global score was obtained by adding up these zonal scores. Comparative scoring methods assume that a higher score or higher grade represent greater severity of disease.

Semi-quantitative scoring

Semi-quantitative scoring methods have been developed to provide more precise assessment of quantity and type of ILD abnormalities. One scoring system that has been used in several studies was developed and published by Warrick et al. [16] (Table 2).

Table 2: Semi-quantitative scoring method: Warrick et al. [16].

This scoring system combines severity and extent of disease. Different abnormalities corresponding to increasingly severe disease are given increasingly high scores. Extent is determined based on the total number of bronchopulmonary segments involved for each abnormality. The greater the number of segments involved, the higher the extent score. These scores are combined to obtain a global score.

Application of this scoring system, however, requires more advanced knowledge of pulmonary anatomy and proficiency in identifying bronchopulmonary segments. This may not be generally useful for clinicians.

Kazerooni et al. [25] published a scoring method to assess HRCT and to correlate with pathology in IPF. Variations of this method have since been used by several groups in patients with SSc [7,26-31]. Ooi et al. [29] evaluated correlation between HRCT findings and clinical markers of disease activity. Their scoring system, which is similar to that of Kazerooni et al. [25] is detailed in Table 3.

Table 3: Semi-quantitative scoring : Ooi et al. [29].

The extent of each abnormality is estimated for each lobe. A semi-quantitative score is assigned (0-4) based on the approximate percentage of disease for each one. This scoring method makes a distinction between pure ground-glass opacities and ground-glass mixed with reticular disease. Differentiating between pure groundglass opacity and mixed reticular disease, and between fibrosis and honeycombing, may permit more precise assessment of the relationship between particular abnormalities and clinical parameters.

Pandey et al. [30] used the same scoring method to assess the relationship between ILD and peak systolic pulmonary arterial pressure (sPAP). They scored any ground-glass disease, reticular abnormalities and honeycombing in each of 5 lobes (scoring lingula with left upper lobe) using the same 0 to 4 semi-quantitative score. They also weighted each score for relative lobar volume using correction factors. Weighting the scores based on relative volume of each lobe may allow for more accurate estimation of global disease. A score of 3 for fibrosis in the lingula may not represent the same total amount of disease as a score of 3 in the left lower lobe, for example.

Goldin et al. [27] published the scoring method that was used to assess and follow-up ILD in the Scleroderma Lung Study population. This was also based on Kazerooni’s method. Instead of scoring lobes, however, they scored 3 anatomical zones (upper, middle and lower zones) in each lung (Table 4).

Table 4: Semi-quantitative scoring; Scleroderma Lung Study [27].

The sum of these grades in all 6 zones make up the global score for each abnormality. In this case, pure ground-glass disease is scored as opposed to ground-glass mixed with reticular disease. Disease progression was assessed subsequently by comparing CT scans in a blinded fashion. Disease in each zone was qualitatively compared from one scan to another and each abnormality was scored as better, same or worse. Semi-quantitative scores were not used to evaluate progression in this case [27].

Quantitative scoring

Wells et al. [32] proposed a quantitative scoring method, using estimation of extent of disease as a percentage of the anatomical region assessed, which consisted of 5 separate levels. This method had been previously developed in patients with IPF [33]. Details are outlined in Table 5.

Table 5: Quantitative and semi-quantitative scoring: Wells et al. [32].

Figure 2 is an example of a typical HRCT scan of a patient with SSc. This scan can be used as an example to illustrate how these scoring methods can be applied and how they differ.

Figure 2: HRCT scan of a patient with SSc-ILD. This figure represents four cuts of an HRCT for a patient with SSc: A. Level of the aortic arch; B. Main carina; C. Pulmonary venous confluence; D. Origin of the diaphragm. The disease is more severe in the lower lung zones, with more pronounced reticular fibrosis and peripheral honeycombing.

Whole pulmonary lobes, bronchopulmonary segments or anatomical zones are scored in the comparative scoring system and the semi-quantitative methods described above. All the HRCT slices would be required to adequately assess disease extent in each lobe or zone, however for convenience, four slices will be used here and scored independently. Using Wells’ comparative score, this HRCT would get graded as 4, 5, 3 and 4 for images A, B, C and D respectively, due to the predominance of reticular pattern over ground glass opacities. In contrast, using Warrick’s score, this scan would receive a severity score of 10 out of 15 because of the presence of ground glass opacities (score of 1), irregular pleural margins (score of 2), septal lines (score of 3) and honeycombing (score of 4). It would be impossible to score extent of disease given that bronchopulmonary segments canot be identified with only four cuts. When combining the images and applying the method used by the Scleroderma Lung Study group, a grade of 1 (1- 25% involvement) is obtained for ground glass opacity, 2 (26-50%) for fibrosis and 1 for honeycombing (although technically, right and left lungs should be scored separately). Finally, when applying Well’s quantitative method, global disease extent is estimated at 25%, with a grade of coarseness of reticulation of 2 because of the microcystic honeycombing.

The construct validity of a scoring system can be defined as its ability to predict an accepted measure of an underlying construct such as disease severity. The most common outcomes used to assess the validity of the different scoring methods included pulmonary function tests, findings on bronchoalveolar lavage, histopathology and survival. Details of correlation between HRCT scores for each method and clinical parameters can be found in the supplementary table (appendix 1).

Studies using comparative scoring methods found correlation between higher HRCT severity grades and low DLCO, low FVC and low TLC [2,11,14]. When compared to biopsy specimen, mixed groundglass and reticular thickening (HRCT grade 3) corresponded to more inflammatory changes on histopathology. In contrast, a predominance of reticular disease (HRCT grade 4) corresponded to predominant fibrosis histopathologically [8]. CT discriminated correctly between inflammation and fibrosis in 80% of biopsy specimens. This scoring method suggested a relationship between particular CT abnormalities and histopathological disease. Wells et al. [36] also showed a correlation between extent of disease (in both IPF and SSc) as scored by CT, and neutrophil levels within each lobe.

Using semi-quantitative scoring systems, global score on HRCT has been shown consistently to be significantly and inversely correlated with TLC, DLCO and FVC% predicted [17,19-20,22,29]. A global score of 7 using the score described by Warrick et al. [16] was predictive of PFT abnormalities with a positive predictive value of 0.82, a sensitivity of 0.6 and specificity of 0.83 [17]. Global scores have also been correlated with elevated pulmonary arterial pressures, poor exercise tolerance and gastroesophageal reflux [23,29-30]. When assessing each abnormality separately, fibrosis scores were again correlated with low FVC, DLCO and TLC. One study reported correlation between ground-glass disease and pulmonary function test abnormalities [37]. In contrast, in the Scleroderma Lung Study, pure ground-glass disease did not correlate well with pulmonary function tests, but was associated with evidence of alveolitis on bronchoalveolar lavage, albeit weakly [27]. These results suggest that pure ground-glass on CT scan may be reversible and may represent inflammation, but that mixed ground-glass corresponds to microscopic fibrosis, as it is equally correlated with PFT abnormalities as reticular disease.

Extent of disease assessed using the quantitative scoring method by Wells et al. [32] also correlated inversely with DLCO, FVC and TLC. This scoring method was applied to a cohort of 215 subjects with SSc to help predict prognosis [35]. Global extent of disease greater than 20% was associated with significantly increased mortality (hazard ratio (HR) of death 2.48) and worse progression-free survival. When combined with FVC of less than 70% predicted, HR was 3.46, and this was highly statistically significant.

In terms of assessing disease progression and response to treatment, the Scleroderma Lung Study group found that patients with more severe fibrosis at baseline (or higher fibrosis scores) had a greater mean of decline in FVC compared to groups with no or moderate fibrosis [38]. The pulmonary fibrosis score progressed in a greater proportion of patients on placebo compared to those receiving cyclophosphamide, but there was no difference in ground-glass or honeycombing [39]. Severity of fibrosis, but not ground glass, at baseline correlated with better response to treatment with cyclophosphamide [40]. The results of this study, again, put into doubt the interpretation that ground glass changes represent those pathologic abnormalities which perhaps lead to fibrosis and which may be reversible. Launay et al. [7] using a scoring system similar to the one used in the Scleroderma Lung Study showed that over time, patients had higher mean extent scores and higher proportion of fibrosis, which was associated with lower DLCO and oxygen saturation on follow-up, but not with FVC decline. Kim et al. [41] used the quantitative scoring method described by Wells and found that change in HRCT disease was weakly associated with decrease in DLCO but not with FVC. In contrast, the scoring system used by Warrick et al. [16] when applied to a small cohort (n = 13) of patients receiving cyclophosphamide, did not show progression of HRCT scores despite worsening PFTs [21]. This may have been due to the small sample studied.

There is significant variability in the different scoring systems in the literature, although many variations of similar methods have been described. Several systems have been found to have moderate to good inter-rater reliability (Table 6) [6,20,27,35,39]. Intra-rater test-retest reliability has also been shown in one scoring system to be good [20]. Assessing change in disease severity over time however has been shown to be much less reliable.

Table 6: Inter-observer agreement.

Several quite different scoring systems for HRCT abnormalities in SSc have been used. We have classified these scoring methods as comparative, semi-quantitative or quantitative. Comparative methods assess one abnormality (ground-glass) relative to another (fibrosis) to determine disease severity. Semi-quantitative methods are characterized by estimating disease extent and assigning a grade, with higher grades corresponding to a higher percentage of disease, whereas the actual percentage of lung involvement is used in quantitative methods.

Most of the systems assessed several abnormalities but generally looked at ground glass, reticular interstitial thickening, bronchiectasis and bronchiolectasis and honeycombing. The major differences between systems relate to: 1) the anatomical regions assessed: either entire lung zones [27], whole lobes [29-31] or a defined number of HRCT slices [32], and 2) the use of a severity score roughly related to a broad range of percentage involvement of a segment [27,29] versus a severity score directly related to the actual percentage involvement (to the nearest 5%) [32]. An advantage of scoring lobes instead of zones is the ability to correlate lobar scores with findings on histopathology or BAL. This can be a useful tool in the study of the pathogenesis of disease. Assessing disease extent using a limited number of thinsection CT slices [32] may result in decreased accuracy compared with scoring entire lobes or zones. However, Kazerooni et al. [25] showed similar accuracy when using a 3-slice method compared to a full CT in patients with IPF. This may be due to relative gradual change of disease severity over the lung fields. Whether this applies to SSc-ILD remains to be determined.

Most of the systems have demonstrated adequate reliability and validity so deciding which system to use may be based on other factors. Many have been shown to have good inter-rater reliability, although chest radiologists do better than clinicians and trainees and the agreement between observers for change over time is less than that for the score at one time. The implications of these findings are that scoring requires some expertise and that using HRCT scores for longitudinal multi-center trials may be limited by considerable error. Simpler scoring systems that have little inter-observer variability may be easier to use in a clinical setting, especially by physicians who are not radiologists. Assigning a percentage of disease is easier than identifying disease extent by counting broncho-pulmonary segments, which requires more advanced knowledge of pulmonary anatomy. Similarly, scoring lobes can be more complex for the clinician than scoring upper, middle and lower lung zones. However, simpler methods with less variability will be less sensitive to small changes over time and will have less discriminant validity.

Semi-quantitative scoring systems, using grades that correspond to ranges in percentage, automatically causes approximation of the regional extent of disease. Therefore there is inherent imprecision to these methods. Six percent involvement of fibrosis would be graded the same way as 24% involvement of a region, but these are clearly different and may have different clinical implications. Semi-quantitative systems, compared to quantitative scores using actual percentage of disease may also be less sensitive in detecting small changes. This may impair the ability of these methods to assess disease progression over time. Variations that may be clinically or physiologically significant, but still remain within the same scoring range, will be unaccounted for. This is even more problematic for a global score that groups different abnormalities. Progression of one abnormality, for example fibrosis or honeycombing, but with simultaneous regression of another, for example ground-glass opacity, may be reflected as a lower global score but in fact may represent worsening overall lung physiology. If we knew that ILD in SSc clearly progressed through a series of stereotypic changes, then comparative systems may provide a different method of assessing progressive disease. For example, a progression from ground glass to honeycombing to subpleural cysts may equate with disease progression and increased severity [8].

Scoring systems quantifying different abnormalities separately allow for measuring contribution of each abnormality, which cannot be done by scoring global percentage of disease only. This is important in view of recent evidence that ground glass opacities do not necessarily correspond to reversible disease [13]. Although pure ground glass may represent inflammatory changes, it may also represent very fine reticular fibrosis beyond the resolution of current HRCT technique. Determining the correlation of each abnormality with clinical parameters may provide more insight into the relationship between specific CT findings and underlying disease pathophysiology and disease progression. The relationship of each abnormality with prognosis, histology, or clinical parameters can then be assessed empirically in any study, without any a priori assumption about the meaning of the abnormality.

A good scoring system should be reliable, valid, sensitive to change and have prognostic implications. Scoring systems for research purposes should adhere closely to these principals and thus may require a high level of complexity. In a clinical setting, the issue of practicality, ease of use and applicability by most physicians who are following patients suffering from SSc is important.

Of the scoring systems discussed, the method described by the investigators of the Scleroderma Lung Study most closely approximates the requirements for a research scoring system. One of the limitations of this method might be a lack of sensitivity given the semi-quantitative nature of the method, especially when assessing for small changes over time.

In the clinical setting, the quantitative method described by Wells and colleagues may be more useful given that it is easier to apply and correlates with prognosis. Further studies however are required to evaluate how sensitive it is to detect change over time, and therefore how it can be used in assessing response to therapy.

Judging by the number of published CT scoring methods, there is little consensus among research groups regarding the optimal way to assess ILD in SSc. A method that would be uniformly adopted for research would allow for pooling of studies and could dramatically improve the strength of the evidence available. A scoring method, perhaps not the same as the best research method, that can be easily applied in clinical settings, could be integrated as part of a routine evaluation of patients with SSc.