Peripheral bone erosion and atlantoaxial instability as factors associated with pannus formation in rheumatoid arthritis: a cross-sectional risk stratification study.

OBJECTIVES: Clinical and radiologic predictors of atlantoaxial pannus formation in rheumatoid arthritis (RA) remain unidentified. We aimed to develop a risk stratification model using routinely collected data. METHODS: Our cohort comprised 242 RA patients evaluated with cervical magnetic resonance imaging (MRI) at Hanyang University Medical Center (2005 to 2015). Pannus was assessed on midsagittal T1-weighted sequences. We examined 16 candidate variables covering demographics, radiographic measurements, serology, and medication; chained-equation multiple imputation handled missing values. A parsimonious logistic regression model (bone erosion, anterior atlanto-dens interval (ADI), and age; events per variable 20.3) served as the primary analysis. Fivefold cross-validated area under the receiver operating characteristic curve (AUC) measured discrimination; 200 bootstrap resamples quantified overfitting, and calibration was assessed by the Hosmer-Lemeshow test. RESULTS: We identified pannus in 61 of 242 patients (25.2%). Bone erosion dominated, odds ratio 6.47 (95% confidence interval 3.18 to 13.16, P < 0.001). Cross-validated AUC for the parsimonious model reached 0.768 (optimism-corrected 0.769), and calibration was satisfactory (Hosmer-Lemeshow P = 0.575). A two-factor stratification by bone erosion and ADI separated subgroups whose pannus rates spanned 6.0% (95% CI 2.6% to 13.2%) to 45.7% (95% CI 34.6% to 57.3%; Cochran-Armitage trend P < 0.001). CONCLUSIONS: Bone erosion on hand and foot radiographs outperformed all 15 competing variables in each of the three analytic frameworks applied. Pairing the erosion finding with ADI values obtained from lateral cervical films produces a clinically transparent risk classification; this two-variable scheme may guide decisions on cervical MRI referral. Key Points • Peripheral bone erosion on hand and foot radiographs was the factor most strongly associated with atlantoaxial pannus in 242 rheumatoid arthritis patients (OR 6.47, P < 0.001). • A parsimonious logistic regression model (bone erosion, ADI, age) achieved a cross-validated AUC of 0.768 with good calibration. • A two-factor stratification separated subgroups with pannus rates of 6.0% (95% CI 2.6% to 13.2%) to 45.7% (95% CI 34.6% to 57.3%) (Cochran-Armitage P < 0.001), a clinically actionable risk gradient from routine plain radiographs.