None of the proposed criteria for a positive spinal magnetic resonance imaging (MRI) scan are clinically useful in distinguishing patients with nonradiographic axial spondyloarthritis (nr-axSpA) from those with nonspecific back pain, new research indicates.
The diagnostic utility of all cutoff values for corner inflammatory (CIL) or several corner fatty lesions (CFL) as proposed in the literature at ≥2/≥3 CIL or ≥6 CFL were not clinically useful for differentiating nr-axSpA from nonspecific back pain.
Only a relatively high threshold of ≥6 CFL was found to have moderate to substantial diagnostic utility, but this threshold was limited by decreased sensitivity in the setting of clinically suspected nr-axSpA.
"These findings question definitions of a positive MRI in spondyloarthritis based on spine MRI alone without taking into account concomitant lesions on sacroiliac joint (SIJ) MRI," Ulrich Weber, MD, of the in Edmonton, and colleagues observed.
Two independent cohorts consisting of 130 consecutive patients 50 years of age or less were recruited for the study, which was published online in . Patients had been newly referred to two university outpatient clinics between 2008 and 2011.
Cohort A, from the in Switzerland, was made up of 42 back pain patients for evaluation of clinically suspected SpA and were matched with 20 healthy controls.
Cohort B was made up of 88 patients who presented to a university ophthalmology department at the University of Alberta in Canada. Those who had a history of past or present back pain lasting at least 3 months were again referred to rheumatology for assessment of SpA.
In both cohorts, a local rheumatology expert classified patients as having nr-axSpA, ankylosing spondylitis (AS), or nonspecific back pain.
Because of the heterogeneity of the two cohorts, investigators analyzed the cohorts separately.
None of the spinal thresholds of ≥2/≥3 CIL or ≥6 CFL showed clinically relevant diagnostic utility in distinguishing nr-axSpA patients from those with nonspecific back pain in either cohort (positive likelihood ratio 1.38-2.36).
Further, limited diagnostic utility was observed in nr-axSpA patients using a threshold of ≥4 CIL (sensitivity 0.34/0.19, specificity 0.91/0.93, and positive likelihood ratio of 3.83/2.72 for cohorts A and B, respectively), the authors add.
A ≥3 CIL and a ≥6 CFL did have some diagnostic utility in AS patients (positive likelihood ratio 2.44/4.03 and 2.49/2.53 for cohort A and B, respectively) but it was small, researchers note.
A threshold of ≥6 CIL was moderately to substantially useful in diagnosing nr-axSpA and AS patients in both cohorts (positive likelihood ratio 13.26/6.74 and 29.56/14.67 for cohort A and B, respectively), but this was achieved at the cost of a drop in sensitivity relative to ≥2/≥3 CIL alone (range 1.47-1.74/1.38-2.36).
Moreover, using a combination of ≥4 CIL plus ≥2/≥3/≥4 CFL yielded a moderate to substantial diagnostic utility in nr-axSpA patients in cohort A (positive likelihood ratio 7.74-13.26), but its utility in cohort B was small (2.59-3.19).
Similarly, the combination of ≥2/≥3 CIL plus ≥2/≥3/≥4 CFL resulted in a minimal and clinically irrelevant increase in diagnostic utility in nr-axSpA patients (positive likelihood ratio 2.58-2.95/1.57-2.34 for cohorts A and B, respectively) compared with ≥2/≥3 CIL alone.
"The present study analyzed two SpA inception cohorts separately because of significant differences in symptom duration and disease activity which reflects the broad clinical spectrum of nr-axSpA patients presenting in daily routine," the authors write.
"And the impact of disease characteristics on diagnostic utility of MRI features challenges uniform definitions for a positive spinal MRI in disorders with a heterogeneous clinical spectrum as in SpA."
Disclosures
The authors declared that they had no competing interests in relation to the article.
Primary Source
Arthritis and Rheumatology
Weber Ulrich, et al "Diagnostic utility of candidate definitions for a positive MRI of the spine in patients with axial spondyloarthritis" Arth Rheum 2014; DOI 10.1002/art.39001.