Arthritis & Rheumatism, Volume 63,
November 2011 Abstract Supplement

Abstracts of the American College of
Rheumatology/Association of Rheumatology Health Professionals
Annual Scientific Meeting
Chicago, Illinois November 4-9, 2011.

The BILAG-2004 Systems Tallya Novel Way of Analysing the BILAG-2004 Scores.

Yee1,  Chee-Seng, Gordon1,  Caroline, Isenberg2,  David A., Griffiths3,  Bridget, Teh4,  Lee- Suan, Bruce5,  Ian N., Ahmad6,  Yasmeen

University of Birmingham, Birmingham, United Kingdom
University of Southampton, Southampton, United Kingdom
St. Thomas' Hospital, London, United Kingdom
Lupus Research Unit, The Rayne Institute, Kings College London School of Medicine, London, United Kingdom
MRC Biostatistics Unit, Cambridge, United Kingdom
University College London, London WC1E 6JF, United Kingdom
Freeman Hospital, Newcastle Upon Tyne, United Kingdom
Royal Blackburn Hospital, Blackburn, United Kingdom
A, Manchester, United Kingdom
The Department of Rheumatology, Betsi Cadwaladr University Health Board (West), Llandudno, LL30 1LB, UK, Wales, United Kingdom
University College London, London, United Kingdom
Sheffield Center Rheumatic Dis, Sheffield South Yorkshire, United Kingdom
Royal National Hospital, Bath, United Kingdom


The BILAG-2004 index has 9 separate categorical system scores which pose a challenge to longitudinal analysis of disease activity in clinical studies. We propose a novel method to record the number of systems in which activity increases, decreases or remains the same between two time points and express this as a tally (BILAG-2004 systems tally, BST). The aim of this exploratory analysis was to assess the appropriateness and performance of this novel method.


This was an analysis of a multi-centre longitudinal study of SLE patients where data were collected on the BILAG-2004 index and therapy at every visit. External responsiveness with multinomial logistic regression was used to examine the appropriateness of BST with change in therapy as the reference standard. It was anticipated that the tally of worsening or persistently active disease would be significantly associated with treatment increase and the tally of improving activity would be significantly associated with treatment reduction. Results were reported in coefficient and 95%CI. Receiver operating characteristics (ROC) curves analysis was used to assess the performance of this approach.


There were 1414 observations from 347 patients. BST was a data driven simplification of the BILAG-2004 system scores and its derivation was suggested from external responsiveness analysis of the original 9 system scores. It had 6 components (Table) and showed the expected significant association with change in therapy. This was further simplified into 3 components (simplified BST, sBST), by combining persistent significant activity and deteriorations (major and minor) into one category (number of systems with active/worsening disease), and combining the improvements (major and minor) into another category (number of systems with improving disease). Similarly, sBST showed the expected significant association with change in therapy: active/worsening disease component with treatment increase (coefficient 2.08, 95%CI: 1.72–2.43) and improving disease component with treatment reduction (coefficient 0.43, 95%CI: 0.21–0.65). ROC curves analyses demonstrated that both versions had similar good performance characteristics in predicting increase in therapy with area under curve (AUC) for BST 0.83 and sBST 0.81, whereas the AUC for the original BILAG-2004 index systems score was 0.75.

Table. External responsiveness of the BILAG-2004 systems tally with multinomial logistic regression

BILAG-2004 systems tallyNumber of ObservationsIncrease in Therapy Coefficient (95% CI)Decrease in Therapy Coefficient (95% CI)
Number of systems with major deterioration (change of Grade B/C/D/E to A or Grade D/E to B)1702.82 (2.34, 3.30)-0.22 (-0.79, 0.35)
Number of systems with minor deterioration (change of Grade C to B)1031.88 (1.22, 2.55) 
Number of systems with persistent significant activity (no change from Grade A or B)791.64 (1.21, 2.06)-0.02 (-0.63, 0.59)
Numbers of systems with minor improvement (change of Grade A to B or B to C)188-0.28 (-0.73, 0.17)-0.38 (-0.79, 0.03)
Number of systems with major improvement (change of Grade A to C/D or Grade B to D)1730.18 (-0.26, 0.63)0.33 (0.04, 0.62)
Number of systems with persistent minimal or no activity (change of Grade C/D/E to C/D/E)141400


The novel BST and sBST provide an alternative method to analyse BILAG-2004 disease activity longitudinally. Both are clinically relevant and meaningful which may provide a more efficient way of analysing clinical trials results. With better efficiency, fewer patients will be required in trials without jeopardising the power of the study and this will reduce the cost of running such studies.

To cite this abstract, please use the following information:
Yee, Chee-Seng, Gordon, Caroline, Isenberg, David A., Griffiths, Bridget, Teh, Lee- Suan, Bruce, Ian N., et al; The BILAG-2004 Systems Tallya Novel Way of Analysing the BILAG-2004 Scores. [abstract]. Arthritis Rheum 2011;63 Suppl 10 :605

Abstract Supplement

Meeting Menu