Inter-observer reliability for radiographic assessment of early osteoarthritis features: the CHECK (cohort hip and cohort knee) study

Abstract

To calculate inter-observer reliability between four different trained readers and an experienced reader on early radiographic osteoarthritis (OA) features in our early OA cohort hip and cohort knee (CHECK) cohort. Four readers were trained by a radiologist and experienced reader to score radiographic OA features. After this training they scored the CHECK cohort. Of the 1002 participants, 38 were scored by all readers. Five different angle radiographs (three for the knee, two for the hip) at three different time points were scored and compared. Inter-observer reliability was evaluated between each of the four trained readers and the experienced reader. Separate radiographic OA features and of overall Kellgren & Lawrence (K&L) scores. In addition, reliability of progression of radiographic was determined in K&L scores and joint space narrowing (JSN). For hip and knee there was substantial inter-observer reliability on overall K&L scores. In the knee, JSN was scored with fair to moderate reliability, osteophytes with moderate to nearly perfect reliability, and other features with fair to substantial reliability. In the hip, reliability ranged from substantial to nearly perfect. Moderate inter-observer reliability was found for progression of OA in both knee and hip, with slightly better reliability for progression based on K&L scores than on separate features. Good inter-observer reliability can be achieved between trained readers and an experienced reader. Although JSN in the knee is scored with lower inter-observer reliability than osteophytes, this does not seem to influence overall K&L scoring. In the hip all features showed good reliability.