Article Text

THU0143 Performance of Prediction Matrices for Rapid Radiologic Progression in Daily Practice
  1. D. De Cock1,
  2. S. Meyfroidt1,
  3. G. Vanderschueren2,
  4. L. Lateur2,
  5. J. Joly3,
  6. R. Westhovens1,3,
  7. P. Verschueren1,3
  1. 1Skeletal Biology and Engineering Research Center, Department of Development and Regeneration, KU Leuven
  2. 2Radiology
  3. 3Rheumatology, University Hospitals Leuven, Leuven, Belgium


Background Rapid radiologic progression (RRP) is an important radiologic hallmark for structural damage in rheumatoid arthritis (RA) patients. For this purpose, various prediction matrices to detect risk on RRP were developed. However, these matrices were developed and, in general, verified in clinical trials and not in daily clinical practice, a setting for which they were conceived and in which they should be used.

Objectives To compare 4 different prediction matrices (ASPIRE CRP matrix, ASPIRE ESR matrix, SWEFOT matrix and the BEST matrix) to predict RRP in patients of a Belgian cohort in daily practice at year one, year two and over these two years.

Methods 74 consecutive DMARD naïve early RA patients were included between 2001 and 2007 with X-rays of both hands and feet at baseline, year one and year two. Patients received initial combination therapy with steroids (ICTS) or DMARD monotherapy (IMT). The RX images were scored according to the modified Sharp/van der Heijde (SvH) method by three readers separately: two experienced radiologists (LL and GVDS) and one clinical researcher (DDC). A total Sharp score (TSS) progression of equal or higher than 5 per year was considered RRP.

The Aspire matrices use SJC28, RF status (U/mL) and CRP status (mg/dL) or ESR status (mm/h). The Best matrix consists of baseline erosion scores, RF and ACPA presence combined and CRP status (mg/L). The Swefot matrix consists of gender, RF presence, ACPA presence and ESR status (mm/h). The Aspire and Best matrices have a different risk probability for patients on mono- or combination therapies.

Patients were systematically placed in each of the matrices yielding a RRP probability. The performance of the matrices was tested by Receiver Operating Characteristic (ROC) curve analysis in which the Area Under the Curve (AUC) reflected the discriminating power. A matrix with an AUC of less than 0.5 was considered as having no predicting value, an AUC between 0.5 and 0.7 was considered to have moderate predicting value and an AUC higher than 0.7 was regarded as having good predicting value.

Results In the first year, the BEST and the ASPIRE ESR/CRP matrices yielded an AUC between 0.5 and 0.7, which gives them moderate predicting value. The SWEFOT matrix lacked any predicting value. In the second year, the ASPIRE CRP and SWEFOT matrix displayed an AUC between 0.5 and 0.7, which gives them moderate predicting value. The BEST and ASPIRE ESR yielded an AUC equal or lower than 0.5 which means they have no predictive value at all. Over two years, the ASPIRE CRP matrix displayed an AUC higher than 0,7 which corresponds to good predicting value. The BEST and ASPIRE ESR matrix yielded an AUC between 0.5 and 0.7, which gives them moderate predicting value. The SWEFOT matrix lacked any predicting value.

Conclusions The ASPIRE CRP matrix generally showed the highest discriminating power to predict RRP in a cohort of early RA patients. The overall performance of all matrices was however disappointing to fully rely on in daily practice. More reliable prediction models should be developed to detect RRP.

Disclosure of Interest None Declared

Statistics from

Request permissions

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.