Background High-resolution CT (HRCT) is well-established as a sensitive method of detecting and characterizing of interstitial lung disease in systemic sclerosis (ILD-SSc). Several visual scales have been developed for scoring the severity and extent of ILD-SSc. All of these score systems reflects varying complexity of lung abnormalities. No agreement has yet been reached on how ILD should be scores, especially on serial CT.
Objectives 1) to determine the inter-rater reliability for measures of three radiologic indexes separately; 2) to evaluate the agreement of radiological indexes for assessment of pneumofibrosis progression using Cohen’s kappa changes in lung damage on serial CT
Methods We compared three radiological indexes invented by Kazerooni EA et al. (1), Wells AU et al. (2) and Warrick JH et al. (3). Interstitial and alveolar scores of Kazerooni were additionally summed into overall index. The agreement of radiologic indexes was evaluated using Cohen’s kappa (4). Patients In 2006-2008 we prospectively followed 110 pts with ILD-SS. Nobody had rapid progression of ILD. HRCT was performed twice over one year by one expert (L.V.N), who graded ILD on pair CT-scans as unchanged in 77 cases (70%), worsened in 20 (18%) and improved in 13 (12%). The expert recruited CT scans with definite improvement (6 pts) and progression of ILD (6 pts). The final cohort consisted of 12 pts, mean age 42±13.2, 11 females, diffuse form – 7, limited -5. Overall 24 scans were respectively evaluated by means of three scoring systems by the expert and three investigators - experienced thoracic radiologists, who were blinded to any patient information and chronology of HRCT investigations. The radiological indexes were collected in protocols for statistical analysis.
Results Intraclass correlation coefficients (ICC) for measures of three radiologic indexes are shown in table. Wells’ index has the best inter- rater reliability, estimated as ICC. All indexes applied with one-year interval didn’t mirror changes on CT scans. From a statistic’s point of view, Well’s index was better than others, so we used it for evaluating of Cohen’s kappa. There were no significant changes in medians of Well’s index detected by three radiologists at initial and follow up assessments. Agreement in the detection of ILD progression between pairs of observers was not satisfactory. The strength of the agreement was poor, slight or fair (kappa -0, 5; 0, 12; 0, 25).
Conclusions Our experiments displayed that applied score systems failed to detect evolution of ILD on CT scans in pts with mild ILD-SS. In may be that one year follow up period is relatively short to reveal small radiological changes on HRCT scans in pts with slow progression.
Kazerooni EA et al. Am J Roentgenol 1997, 169:977-83.
Warrick JH et al. Rheumatol, 1991, 18:1520-1528.
Wells AU et al. Am J Roentgenol 1993, 161:1159-1165.
Smidt N., et al. BMC Medical Research Methodology, 2006, 6: 12-20.
Disclosure of Interest None Declared
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.