Affiliations: [a] Department of Neurology, Friedrich-Baur-Institute, Ludwig-Maximilians-University of Munich, Germany
| [b] Institute for Medical Information Processing, Biometry and Epidemiology, Ludwig-Maximilians-University of Munich, Germany
Correspondence:
[*]
Correspondence to: Haris Babačić, Friedrich-Baur-Institute, Department of Neurology, Interdisciplinary Centre for Neuromuscular diseases Ludwig-Maximilians-University of Munich, Ziemssenstrasse 1a, 80336 Munich, Germany. Tel.: +49 89 44005 7470; Fax: +49 89 44005 7402; E-mail: [email protected].
Abstract: Background/Aim: Pulmonary function tests are used for screening respiratory insufficiency in patients with myotonic dystrophy (DM). We analysed the agreement between two different approaches in assessment of abnormal findings of forced vital capacity (FVC), forced expiratory volume in the first second (FEV1), maximal inspiratory pressure (MIP) and maximal expiratory pressure (MEP), in DM patients. Methods: We used Cohen’s κ- and Bangdiwala’s B- statistic to compare the agreement between different cut-off values recommended by experts (ENMC) and the cut-off values based on the reference range (RR). We further analysed their sensitivity (Sn) and specificity (Sp) in detecting symptoms associated with respiratory insufficiency. Results: The observed agreement was: 1) for FVC: κ= –0.002, B = 0.406; 2) for FEV1: κ= 0.944, B = 0.946; 3) for MIP: κ= 0.625, B = 0.674; and 4) for MEP: κ= 0.241, B = 0.373. Overall, RR cut-off values showed higher sensitivity, whereas the ENMC values showed higher specificity in detecting symptoms of respiratory involvement. Conclusions: The two approaches showed perfect agreement in assessment of FEV1, substantial agreement for MIP, and weak agreement for FVC and MEP. RR is an established method of assessment for spirometry and should be favoured because it takes variability within the population into account. Further development and validation of regression equations for RR calculations of predicted maximal respiratory pressures, with corresponding lower limits of normal, is required.The B statistic is more robust in assessing agreement between two diagnostic methods, resolving the issue of the κ paradox.
