Purchase this article with an account.
C. Wilson, K. D. Cocker, M. J. Moseley, A. G. Smith, A. L. Ells, G. E. Quinn, A. R. Fielder, Plus Disease Collaborative Group; A New Approach to Classifying Plus Disease. Invest. Ophthalmol. Vis. Sci. 2007;48(13):3103.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Plus disease is the new driver for retinopathy of prematurity (ROP) treatment yet diagnosis is based on a single image published in 1984. Plus is a constellation including congestion and tortuosity of vessels at the posterior pole. The aim of this study was to describe retinal vessel morphology associated with development and also with ROP.
This study had two arms. (1) Grading plus disease with current guidelines of plus, preplus and normal. (2) Quadrantic grading of vessel morphology-venous width (VW), arteriolar tortuosity (AT), venous tortuosity (VT), arterial width (AW) on a scale of 0-10. One hundred Retcam (Massie Labs) images from babies ranging from 31 to 42 weeks postmenstrual age (PMA) were displayed in random order on a web interface linked to an online database. Images were graded using a panel of 10 readers: pediatric ophthalmologists ('experts'), trainee ophthalmologists, a medical student and a neonatal nurse. Readers' scores were compared with recorded clinical ROP stage.
An overall average reading of all readers (OAR) for VW, AT, VT & AW for all quadrants of one image correlated with clinical staging yielding a Spearman's rho of 0.62 (n=39) (p<0.05). VW, AT, VT & AW alone also gave rho values >0.5 (n=39, p<0.05) when correlated with clinical staging. Inter-reader error between experts for plus disease grading had a Fleiss kappa (FK) statistic of 0.18, and a Kendall’s coefficient of concordance (KCC) of 0.62. Inter-reader agreement between experts for vascular changes ranged from 0.01 to 0.13 (FK) and 0.35 to 0.66 (KCC) for VW, AT, VT & AW in each quadrant. Inter-reader agreement between non-experts for plus disease grading was 0.09 (FK) and 0.48 (KCC). The OAR for images of babies without ROP was significantly correlated (Spearman's rho = 0.49, p=0.02) with gestational age but not with PMA at image acquisition (Spearman's rho = 0.18, p=0.40).
Absolute agreement between readers is poor, yet they tended to agree on the intrinsic order of the grading scale. Similar results for expert and non expert readers indicate promise for screening applications where expertise is limited. PMA appears to be associated with OAR, suggesting minimal width and tortuosity change in development of babies without ROP. As OAR correlated with clinical ROP staging this may prove a successful new grading system for plus disease. An improved grading system should enhance plus disease diagnosis and improve treatment of ROP.
This PDF is available to Subscribers Only