Background Several models have been proposed to refer patients (pts) with possible axial spondyloarthritis (axSpA) from primary care to the rheumatologist.
Objectives To compare and evaluate the performance of all published referral models for axSpA in the Leiden SPondyloarthritis Caught Early (SPACE) cohort.
Methods Ten referral models were found in literature and were tested in the Leiden SPACE cohort, which includes pts with back pain (≥3 months, ≤2 years, onset <45 years; n=192) referred to the rheumatology outpatient clinic of the LUMC. Imaging was omitted from all models if used, as it is unfeasible for screening in primary care. The Brandt model refers pts if HLA-B27+ or inflammatory back pain (IBP)+ (based on 3 criteria1) present; in Brandt I, pts are IBP+ if one IBP criterion present, in Brandt II if 2 etc. The Braun model registers presence of psoriasis (PSOR), buttock pain (BP) and improvement of back pain by exercise; referral if ≥2 features present. If ≤1 feature present HLA-B27 is tested; positive pts are referred.2 In the Braun alt. model, alternating BP is used instead of BP. MASTER refers pts if ≥2/4 features are present: IBP, HLA-B27, family history (FH) for ankylosing spondylitis, good response to NSAIDs.3 RADAR refers pts if ≥2/5 features are positive: IBP, HLA-B27, FH for SpA, good response to NSAIDs, extra-articular manifestations (EAM). RADAR 2/3 refers pts if ≥2/3 of the following are positive: IBP, good response to NSAIDs, EAM.4 CaFaSpA refers pts with ≥1 or ≥2 points (pt) (good response to NSAIDs, FH for SpA and IBP (ASAS); all 1pt).5 The performance of the models was evaluated (sensitivity, specificity, positive predictive value (PPV), positive likelihood radio (LR+)) using classification by ASAS axSpA criteria as external standard. For erroneously referred pts (i.e. not fulfilling the ASAS criteria, “false positive”; FP), post-test probability (PTP) for axSpA was calculated based on LR product for presence of SpA features (LR product ≥78 (equaling PTP ≥80%) as cut-off for probable axSpA). Erroneously not referred (“false negative”; FN) pts met ASAS axSpA criteria.
Results In total, 74/192 pts fulfilled the ASAS axSpA criteria; 48 with imaging+. Most models performed well with regard to sensitivity/specificity. The Braun alt. model has the most balanced sensitivity/specificity and highest LR+. All models including HLA-B27 yield ASAS imaging+ FN pts, 14-23% with radiographic sacroiliitis. The PPV of the models is generally low, indicating that many pts are referred erroneously. On the other hand, of these “FP pts”, 6-16% have a PTP ≥80% for axSpA and are therefore correctly referred.
Conclusions Most referral models performed well in SPACE. However, this cohort includes pts already referred from primary care, probably causing overestimation of performance. All models yield pts, who have advanced disease as indicated by fulfilling ASAS imaging+ but are not referred. On the other hand, large numbers of pts referred unnecessarily might lead to a burden for health care systems. Further studies should be conducted in primary care to evaluate these models in their target population.
Brandt ARD 2007;66:1479-84.
Braun Rheum 2013;52:1418-24.
Poddubnyy J Rheumatol 2011;38:2452-60.
Sieper ARD 2013;72:1621-7.
Weel AC&R 2013 Sep 19.
Disclosure of Interest None declared