Features of Axial Spondyloarthritis in Two Multicenter Cohorts of Patients with Psoriasis, Uveitis, and Colitis Presenting with Undiagnosed Back Pain

Abstract

Objective

We aimed to assess the following: (1) the frequency of axial spondyloarthritis (axSpA) according to extra‐articular presentation and HLA‐B27 status, (2) clinical and imaging features that distinguish axSpA from non‐axSpA, and (3) the impact of magnetic resonance imaging (MRI) on diagnosis and classification of axSpA.

Methods

The Screening for Axial Spondyloarthritis in Psoriasis, Iritis, and Colitis (SASPIC) study enrolled patients in two multicenter cohorts. Consecutive patients with undiagnosed chronic back pain attending dermatology, ophthalmology, and gastroenterology clinics with psoriasis (PsO), acute anterior uveitis (AAU), or inflammatory bowel disease (IBD) were referred to a local rheumatologist with special expertise in axSpA for a structured diagnostic evaluation. The primary outcome was the proportion of patients diagnosed with axSpA by the final global evaluation.

Results

Frequency of axSpA was 46.7%, 61.6%, and 46.8% in patients in SASPIC‐1 (n = 212) and 23.5%, 57.9%, and 23.3% in patients in SASPIC‐2 (n = 151) with PsO, AAU, or IBD, respectively. Among those who were B27 positive, axSpA was diagnosed in 70%, 74.5%, and 66.7% of patients in SASPIC‐1 and in 71.4%, 87.8%, and 55.6% of patients in SASPIC‐2 with PsO, AAU, or IBD, respectively. All musculoskeletal clinical features were nondiscriminatory. MRI was indicative of axSpA in 60% to 80% of patients and MRI in all patients (SASPIC‐2) versus on‐demand (SASPIC‐1) led to 25% fewer diagnoses of axSpA in patients who were HLA‐B27 negative with PsO or IBD. Performance of the Assessment of SpondyloArthritis International Society classification criteria was greater with routine MRI (SASPIC‐2), though sensitivity was lower than previously reported.

Conclusion

Optimal management of patients presenting with PsO, AAU, IBD, and undiagnosed chronic back pain should include referral to a rheumatologist. Conducting MRI in all patients enhances diagnostic accuracy.

INTRODUCTION

Diagnosis of axial spondyloarthritis (axSpA) is delayed for almost a decade, leading to a substantial burden of illness and a risk of irreversible spinal damage. ¹ Patients with back pain who concomitantly have psoriasis (PsO), acute anterior uveitis (AAU), or inflammatory bowel disease (IBD) may initially present to dermatology, ophthalmology, or gastroenterology clinics, respectively, when there may be an opportunity to diagnose the condition much sooner. Earlier diagnosis and treatment might not only alleviate spinal symptoms earlier in the disease course but also help to make more informed decisions aimed at therapeutics that are effective for both axSpA and the comorbid condition. Recent referral recommendations advocate referral to a rheumatologist of patients with chronic back pain, symptom onset at less than 45 years of age, and extra‐articular features related to axSpA. ² However, nonspecific back disorders are common, and access to a rheumatologist is limited in many parts of the world. Moreover, features indicating inflammatory back pain (IBP) are relatively nonspecific, and HLA‐B27 appears to be less strongly associated with axSpA in patients with PsO and IBD. ³ , ⁴ , ⁵ Previous reports evaluating the prevalence of axSpA, distinguishing clinical features, and associations with HLA‐B27 have been largely retrospective case studies of selected populations. ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹² , ¹³ , ¹⁴ There are limited prospective studies in which the Assessment of Spondyloarthritis International Society (ASAS) classification criteria ¹⁵ have been used to identify patients with axSpA, but magnetic resonance imaging (MRI) evaluation of the sacroiliac joints (SIJs) was only conducted when considered clinically indicated as opposed to being done routinely for all patients. ¹⁶ , ¹⁷ However, the performance of the ASAS criteria has not been assessed in these disease subgroups and may be adversely impacted by the lack of association with HLA‐B27. Moreover, MRI findings in the SIJs that are indicative of axSpA do not correlate well with clinical features of IBP and may be present even in asymptomatic patients with AAU, PsO, or IBD. ¹⁸ , ¹⁹

We conducted two multicenter studies (Screening for Axial Spondyloarthritis in Psoriasis, Iritis, and Colitis [SASPIC] cohorts 1 and 2) in which consecutive patients with undiagnosed chronic back pain attending dermatology, ophthalmology, and gastroenterology clinics with PsO, AAU, and IBD, respectively, were referred to a local rheumatologist with special expertise in axSpA for a structured clinical evaluation and imaging. Our aims were threefold: (1) to establish the frequency of axSpA according to extra‐articular presentation and HLA‐B27 status; (2) to identify the clinical, laboratory, and imaging features that distinguish patients with axSpA from those with nonspecific back pain; and (3) to assess the impact of MRI evaluation on diagnosis and classification of axSpA. In SASPIC cohort 1 (SASPIC‐1), MRI evaluation was conducted only if deemed necessary by the local rheumatologist and the findings used to inform the diagnosis, as routinely performed in a real‐world setting. In SASPIC cohort 2 (SASPIC‐2), all patients had MRI of the SIJs, and scans were evaluated first by local radiologists and then by central readers. The reports were provided to local rheumatologists to aid in establishing the final diagnosis.

PATIENTS AND METHODS

Study design and patient eligibility

The SASPIC study enrolled patients in two cohorts, first SASPIC‐1 and then SASPIC‐2. The two multicenter studies were conducted in Canada at 10 sites in SASPIC‐1 and at 8 sites in SASPIC‐2, with 7 sites contributing patients to both cohorts. Consecutive adult patients ≤45 years of age diagnosed with PsO, AAU, or IBD attending dermatology, ophthalmology, or gastroenterology practices, respectively, were recruited if they presented with chronic undiagnosed back and/or buttock pain of three months’ duration or longer and signed consent forms. Exclusion criteria were absence of chronic back pain, rheumatologist‐established diagnosis of axSpA, history of recent spinal trauma, contraindication to or inability to have MRI evaluation, and treatment with any biologic or targeted synthetic disease‐modifying antirheumatic drug (DMARD) within 12 weeks before screening.

Eligible patients were assessed by local rheumatologists with special expertise in spondyloarthritis (SpA). This consisted of a history, physical examination, C‐reactive protein (CRP), HLA‐B27, and an anteroposterior pelvic radiograph done within the last six months. MRI of the SIJs (oblique coronal STIR and T1‐weighted spine echo [T1W SE]) and spine (sagittal STIR and T1W SE of thoracolumbar and cervicothoracic halves) was ordered as deemed appropriate by the rheumatologist in SASPIC‐1, whereas all patients in SASPIC‐2 had MRI evaluation of the SIJs within four weeks of the clinical evaluation of the patient. After clinical assessment, the rheumatologist determined whether the patient's back and/or buttock pain was consistent with IBP (yes/no and numerical rating scale [NRS; 0–10] with anchors: 0 being no probability of IBP and 10 being definite IBP). The rheumatologist also completed an electronic case report form that included all the clinical axSpA variables in the 2009 ASAS classification criteria. ¹⁵ The diagnostic conclusion and level of confidence in the diagnosis was assessed at each stage of the diagnostic process by the following question: this patient's clinical presentation indicates a diagnosis of axSpA (yes/no and the degree of confidence on a −10 to +10 NRS, with −10 being definitely not axSpA, 0 being I cannot decide if the patient has axSpA or not, and +10 being definitely has axSpA). In SASPIC‐1, this question was addressed after the following evaluations: stage 1 after completion of the history and physical examination; stage 2 after review of the CRP, HLA‐B27, and pelvic radiograph; and Stage 3 after review of the MRI and local radiologist report in those patients who had this evaluation (Supplementary Figure 1). In SASPIC‐2, the diagnostic global question was addressed after the same three sequential steps as in SASPIC‐1 but also included a fourth stage in which the rheumatologist completed the question after receiving central reader evaluations of the pelvic radiograph and MRI scan. All patients provided informed consent. The study was approved by an institutional review board or ethics committee at each study site and conducted according to the Declaration of Helsinki and the International Council for Harmonisation guidelines.

Central reader imaging evaluation

Images were independently evaluated by two experienced readers (rheumatologist [WPM] and a musculoskeletal radiologist [RGL]) for the presence or absence of radiographic sacroiliitis according to the modified New York (mNY) criteria. ²⁰ Once radiographs had been assessed, readers could evaluate the presence or absence of active inflammatory and structural changes on MRI indicative of axSpA. Discrepancies in these assessments were independently evaluated by an experienced adjudicator (rheumatologist [UW]). Readers also provided a standard narrative report to local rheumatologists for patients recruited to SASPIC‐2. Patients and/or public were not involved in any steps of the design, conduct, analysis, and results dissemination of this study.

Outcomes

The primary outcome of the study was the proportion of patients diagnosed with axSpA among the three groups of patients after final global evaluation. Secondary outcomes included proportions of patients with a diagnosis of axSpA stratified by the presence/absence of HLA‐B27, clinical and imaging variables that distinguished axSpA from non‐axSpA among the three groups, impact of on‐demand versus systematic MRI evaluation on diagnosis and classification of axSpA, and the proportion of patients fulfilling the ASAS 2009 classification criteria for axSpA.

Statistical analysis

The proportion of patients diagnosed and classified with axSpA was calculated out of the total number of patients evaluated at each of the rheumatology sites. Descriptive analysis of demographic variables, clinical features of axSpA, and imaging and laboratory variables were tabulated and summarized by means, SD, medians, interquartile ranges, and by numbers and percentages for categorical variables. Statistically significant differences between patients diagnosed with and without axSpA were determined by using Mann–Whitney U test for continuous variables and χ² test for categorical variables. Significance tests were conducted at significance level α = 0.05. All statistical analyses were conducted in SAS Studio version 9.4 (SAS Institute).

RESULTS

A total of 212 patients (10 sites) were recruited to SASPIC‐1 between February 22, 2013, and March 5, 2018, of whom 45 (21.2%) had PsO, 73 (34.4%) had AAU, and 94 (44.3%) had IBD. For SASPIC‐2, 151 patients (eight sites, seven sites participated in SASPIC‐2) were recruited between May 28, 2018, and September 27, 2022, of whom 51 (33.8%) had PsO, 57 (37.7%) had AAU, and 43 (28.5%) had IBD. The percentage per site diagnosed with axSpA varied from 15.8% to 100% in SASPIC‐1 and from 20% to 71.4% in SASPIC‐2 (Supplementary Table 1).

Demographic and clinical characteristics

The proportions of patients diagnosed with axSpA were 46.7%, 61.6%, and 46.8% for those with PsO, AAU, and IBD, respectively, in SASPIC‐1, whereas the respective proportions were 23.5%, 57.9%, and 23.3% for SASPIC‐2 (Tables 1 and 2; Supplementary Tables 2 and 3). A higher proportion of male patients was evident among all groups in patients diagnosed with axSpA, and this was particularly notable in SASPIC‐2. In both cohorts, having PsO and IBD preceded back symptoms (76.2% and 68.2%, respectively, of patients with axSpA in SASPIC‐1, and 91.7% and 70%, respectively, of patients with axSpA in SASPIC‐2), whereas the opposite was evident for having AAU (35.6% and 18.2% of patients with axSpA in SASPIC‐1 and ‐2, respectively). Mean duration of back symptoms before a diagnosis of axSpA was similar in both cohorts and varied from six to nine years among the subgroups. This delay in diagnosis was shortest in patients with IBD in both cohorts and was longest in those with PsO in SASPIC‐1 and in those with AAU in SASPIC‐2. The majority of patients diagnosed with axSpA in all groups and in both cohorts reported back pain severity >5 on a 0 to 10 scale. The mean percentage of patients receiving nonsteroidal anti‐inflammatory drugs (NSAIDs) and reporting back pain severity >5 was 76.6% and ranged from 66.7% to 87.5% of the three subgroups in SASPIC‐1 and ‐2.

Table 1.

Demographic and clinical disease characteristics of patients presenting with undiagnosed back pain in an inception cohort of patients with PsO, AAU, and IBD recruited to SASPIC‐1 according to final local rheumatologist diagnosis of axSpA versus non‐axSpA^*

Feature	Patients with PsO			Patients with AAU			Patients with IBD
	axSpA (n = 21)	Non‐axSpA (n = 24)	P value	axSpA (n = 45)	Non‐axSpA (n = 28)	P value	axSpA (n = 44)	Non‐axSpA (n = 50)	P value
Male, n (%)	15 (71.4)	9 (37.5)	0.04	28 (62.2)	13 (46.4)	0.23	23 (52.3)	20 (40)	0.30
Age, mean (±SD), y	36.3 (±6.2)	36.7 (±5.8)	0.84	33.4 (±5.3)	35.0 (±6.5)	0.24	34.0 (±7.6)	35.9 (±7.0)	0.22
Age at onset of back symptoms, mean (±SD), y	27.2 (±9.2)	30.0 (±8.6)	0.29	25.0 (±7.4)	25.8 (±7.7)	0.67	27.4 (±8.6)	28.9 (±7.3)	0.36
Symptom duration, mean (±SD), y	9.2 (±7.9)	6.7 (±7.5)	0.28	8.4 (±6.1)	9.3 (±7.6)	0.59	6.6 (±5.6)	6.9 (±6.3)	0.80
Age at onset of extra‐articular symptoms (PsO, AAU, and IBD), mean (±SD), y	19.7 (±7.4)	20.3 (±7.9)	0.80	28.2 (±6.0)	32.0 (±7.3)	0.021	25.7 (±8.6)	23.4 (±8.8)	0.20
Family history of SpA, n (%) a
AS	1 (4.8)	2 (8.3)	1.00	6 (13.3)	2 (7.1)	0.70	0	1 (2)	1.00
PsO	10 (47.6)	15 (62.5)	0.38	5 (11.1)	8 (28.6)	0.07	9 (20.4)	4 (8)	0.13
IBD	3 (14.3)	2 (8.3)	0.65	5 (11.1)	3 (10.7)	1.00	15 (34.1)	10 (20)	0.16
AAU	0	0	na	9 (20.0)	5 (17.9)	1.00	2 (4.5)	2 (4)	1.00
Back pain characteristics
Severity of current back pain, (0–10 NRS), mean (±SD) b	6.6 (±2.3)	5.2 (±2.4)	0.04	5.1 (±2.9)	4.9 (±2.4)	0.78	6.0 (±2.6)	5.3 (±4.0)	0.11
Current back pain >5 (0–10 NRS), n (%)	15 (71.4)	11 (45.8)	0.13	24 (53.3)	14 (50)	0.81	26 (59.1)	24 (48)	0.31
Inflammatory back pain, n (%)	18 (85.7)	17 (70.8)	0.30	40 (88.9)	17 (60.7)	0.008	40 (90.9)	25 (50)	<0.001
Respond to NSAIDs within 48 hours, n (%)
Yes	11 (64.7)	15 (75)	0.72	26 (81.3)	15 (78.9)	1.00	17 (58.6)	21 (58.3)	1.00
Mostly complete	0	1 (5)	1.00	3 (9.4)	3 (15.8)	0.66	1 (3.4)	0	0.45
Very good (>50% relief)	4 (23.5)	6 (30)	0.73	11 (34.4)	4 (21.1)	0.36	8 (27.6)	5 (13.9)	0.22
Some benefit (20%–50% relief)	7 (41.2)	6 (30)	0.51	8 (25)	7 (36.8)	0.53	5 (17.2)	12 (33.3)	0.17
Minimal (<20% relief)	0	2 (10)	0.49	4 (12.5)	1 (5.3)	0.64	3 (10.3)	4 (11.1)	1.00
No response	6 (35.3)	5 (25)	0.72	6 (18.8)	4 (21.1)	1.00	12 (41.4)	15 (41.7)	1.00
Treatment, n (%)
Currently receiving NSAIDs	14 (66.7)	11 (45.8)	0.23	23 (51.1)	5 (17.8)	0.006	9 (20.5)	9 (18)	0.80
Currently receiving csDMARDs c	2 (9.5)	2 (8.3)	1.00	0	0 (0)	na	2 (4.5)	5 (10)	0.44
Previously received bioDMARDs d	1 (4.8)	3 (12.5)	0.61	0	0 (0)	na	4 (9.1)	9 (18)	0.25
Peripheral and extra‐articular symptoms, n (%)
Heel pain	8 (38.1)	8 (33.3)	0.77	6 (13.3)	5 (17.9)	0.74	7 (15.9)	6 (12)	0.77
History of enthesitis	5 (23.8)	7 (29.2)	0.75	7 (15.6)	5 (17.9)	1.00	3 (6.8)	14 (28)	0.01
History of peripheral arthritis	9 (42.9)	10 (41.7)	1.00	9 (20)	2 (7.1)	0.19	9 (20.5)	8 (16)	0.60
History of extra‐articular symptoms
AAU	2 (9.5)	0	0.21	45 (100)	28 (100)	na	1 (2.3)	0	0.47
Crohn disease colitis	0	0	na	1 (2.2)	0	1.00	25 (56.8)	28 (56)	1.00
Ulcerative colitis	0	0	na	1 (2.2)	0	1.00	19 (43.2)	22 (41)	1.00
PsO	21 (100)	24 (100)	na	0	0	na	3 (6.8)	1 (2)	0.34
Dactylitis	1 (4.8)	0	0.47	1 (2.2)	0	1.00	0	2 (4)	0.50
Physical examination
Enthesitis, n (%)	6 (28.6)	7 (29.2)	0.23	9 (20)	6 (21.4)	0.26	12 (27.3)	20 (40)	0.28
Number of SPARCC enthesitis points, mean (±SD)	1.2 (±1.7)	0.6 (±1.2)	0.17	0.3 (±1.0)	0.8 (±1.6)	0.22	0.5 (±1.3)	1.0 (±1.6)	0.05
Dactylitis, n (%)	1 (4.8)	2 (8.3)	1.00	2 (4.4)	0	0.52	1 (2.3)	1 (2)	1.00
Hip involvement, n (%) e	2 (9.5)	2 (8.3)	1.00	3 (6.7)	1 (3.6)	1.00	7 (15.9)	6 (12)	0.77
PsO, n (%)	21 (100)	24 (100)	na	0	1 (3.6)	0.38	6 (13.6)	1 (2)	0.05
Peripheral arthritis, n (%)	7 (33.3)	6 (25)	1.00	2 (4.4)	0	0.52	3 (6.8)	3 (6)	1.00
Number of swollen joints, mean (±SD)	1.0 (±2.1)	0.9 (±2.0)	0.83	0.6 (±0.9)	0.0 (±0.0)	na	0.05 (±0.2)	0.1 (±0.3)	0.55
BASMI, mean (±SD)	2.7 (±0.8)	2.1 (±0.7)	0.001	2.4 (±1.0)	2.0 (±0.7)	0.05	2.5 (±1.3)	2.2 (±0.8)	0.23
EDASMI, mean (±SD)	2.3 (±1.4)	2.0 (±0.9)	0.34	2.1 (±1.4)	1.8 (±1.0)	0.24	2.0 (±1.5)	1.9 (±0.9)	0.78
Laboratory
B27 positive, n (%) f	7 (33.3)	3 (13.0)	0.16	38 (84.4)	13 (46.4)	0.001	16 (37.2)	8 (16.3)	0.03
CRP abnormal (≥6 mg/L), n (%)	7 (33.3)	7 (29.2)	1.00	16 (35.6)	9 (32.1)	0.81	19 (43.2)	12 (24)	0.05
CRP, mean (±SD)	5.8 (±6.4)	6.7 (±9.6)	0.71	7.1 (±10.7)	3.8 (±3.0)	0.05	11.9 (±20.0)	4.7 (±7.8)	0.03

^*AAU, acute anterior uveitis; AS, ankylosing spondylitis; axSpA, axial spondyloarthritis; BASMI, Bath Ankylosing Spondylitis Metrology Index; bioDMARD, biologic disease‐modifying antirheumatic drug; CRP, C‐reactive protein; csDMARD, conventional synthetic disease‐modifying antirheumatic drug; EDASMI, Edmonton Ankylosing Spondylitis Metrology Index; IBD, inflammatory bowel disease; na, not applicable; NRS, numerical rating scale; NSAID, nonsteroidal anti‐inflammatory drug; PsO, psoriasis; SASPIC‐1, Screening for Axial Spondyloarthritis in Psoriasis, Iritis, and Colitis cohort 1; SpA, spondyloarthritis; SPARCC, Spondyloarthritis Research Consortium of Canada.

^a AS, PsO, AAU, and IBD are included.

^b Severity of current back pain is measured as the amount of back and/or hip pain during the last week on a 0 to 10 (most severe) NRS.

^c csDMARDs include methotrexate and salazopyrin.

^d bioDMARDs include etanercept, golimumab, adalimumab, infliximab, and certolizumab.

^e Hip involvement is defined as groin pain and/or restricted range of internal rotation.

^f One and two patients were missing HLA‐B27 data in the cohorts with PsO and IBD, respectively.

Table 2.

Demographic and clinical disease characteristics of patients presenting with undiagnosed back pain in an inception cohort of patients with PsO, AAU, and IBD recruited to SASPIC‐2 according to final local rheumatologist diagnosis of axSpA versus non‐axSpA^*

Feature	Patients with PsO			Patients with AAU			Patients with IBD
	axSpA (n = 12)	Non‐axSpA (n = 39)	P value	axSpA (n = 33)	Non‐axSpA (n = 24)	P value	axSpA (n = 10)	Non‐axSpA (n = 33)	P value
Male, n (%)	8 (67)	18 (46)	0.32	25 (76)	6 (25)	<0.001	6 (60)	12 (36)	0.28
Age, mean (±SD), y	36.3 (±6.5)	35.4 (±7.2)	0.71	34.8 (±6.4)	35.2 (±6.9)	0.84	30.6 (±7.8)	34.0 (±9.0)	0.28
Age at onset of back symptoms, mean (±SD), y	28.7 (±8.1)	28.4 (±7.6)	0.91	25.3 (±8.2)	27.1 (±6.9)	0.40	24.2 (±25.6)	28.1 (±9.9)	0.10
Symptom duration, mean (±SD), y	7.6 (±9.9)	7.0 (±5.8)	0.85	9.5 (±7.3)	8.1 (±7.7)	0.50	6.4 (±38.9)	6.0 (±7.0)	0.86
Age at onset of extra‐articular symptoms (PsO, AAU, and IBD), mean (±SD), y	20.5 (±6.8)	20.9 (±9.3)	0.88	32.5 (±7.4)	29.5 (±8.1)	0.16	20.4 (±8.4)	23.4 (±9.6)	0.38
Family history of SpA, n (%) a
AS	1 (8.3)	1 (2.6)	0.42	6 (18.2)	1 (4.2)	0.22	0	4 (12.1)	0.56
PsO	5 (41.7)	21 (53.8)	0.52	5 (15.2)	6 (25.0)	0.50	2 (20.0)	9 (27.3)	1.00
IBD	1 (8.3)	7 (17.9)	0.66	4 (12.1)	2 (8.3)	1.00	4 (12.1)	7 (21.2)	1.00
AAU	3 (25.0)	2 (5.1)	0.08	2 (6.1)	2 (8.3)	1.00	1 (10.0)	0	0.23
Back pain characteristics
Severity of current back pain, (0–10 NRS), mean (±SD) b	6.4 (±2.2)	5.6 (±2.3)	0.30	4.9 (±2.8)	5.0 (±2.3)	0.88	5.8 (±2.8)	5.4 (±1.9)	0.57
Current back pain >5 (0–10 NRS), n (%)	8 (66.7)	22 (56.4)	0.74	16 (48.5)	12 (50.0)	1.00	6 (60)	16 (48.5)	0.72
Inflammatory back pain, n (%)	9 (75)	27 (69.2)	1.00	28 (84.8)	19 (79.2)	0.73	8 (80)	22 (66.7)	0.70
Respond to NSAIDs within 48 hours, n (%)
Yes	9 (81.8)	21 (75.0)	1.00	20 (87.0)	11 (73.3)	0.40	7 (100)	9 (64.3)	0.12
Mostly complete	1 (9.1)	0 (0)	0.28	4 (17.4)	2 (13.3)	1.00	2 (28.6)	0	0.10
Very good (>50% relief)	1 (9.1)	10 (35.7)	0.13	7 (30.4)	3 (20.0)	0.71	2 (28.6)	4 (28.6)	1.00
Some benefit (20%–50% relief)	3 (27.3)	5 (17.9)	1.00	7 (30.4)	2 (13.3)	0.27	3 (42.9)	4 (28.6)	1.00
Minimal (<20% relief)	4 (36.4)	6 (21.4)	1.00	2 (8.7)	4 (26.7)	0.19	0	1 (7.1)	1.00
No response	2 (18.2)	7 (25.5)	1.00	3 (13.0)	4 (26.7)	0.40	0 (0)	5 (35.7)	0.12
Treatment, n (%)
Currently receiving NSAIDs	5 (41.7)	20 (51.3)	0.74	18 (54.5)	9 (37.5)	0.28	4 (40)	6 (18.2)	0.21
Currently receiving csDMARDs c	0	2 (5.1)	1.00	0	0 (0)	na	0	5 (15.2)	0.32
Previously received bioDMARDs d	1 (8.3)	2 (5.1)	0.56	1 (3.0)	0 (0)	1.00	5 (50)	6 (18.2)	0.09
Peripheral and extra‐articular symptoms, n (%)
Heel pain	1 (8.3)	8 (20.5)	0.67	3 (6.1)	6 (25.0)	0.06	0 (0)	5 (15.2)	0.32
History of enthesitis	1 (8.3)	11 (28.2)	0.25	5 (15.2)	3 (12.5)	1.00	1 (10)	4 (12.1)	1.00
History of peripheral arthritis	3 (25)	14 (25.9)	0.73	2 (6.1)	3 (12.5)	0.64	0 (0)	5 (15.2)	0.32
History of extra‐articular symptoms
AAU	0	2 (5.1)	1.00	33 (100)	24 (100)	na	2 (20)	1 (3)	0.13
Crohn disease colitis	0	0	na	1 (3.0)	0	1.00	4 (40.0)	19 (57.6)	0.47
Ulcerative colitis	0	0	na	0	0	na	7 (70)	14 (42.4)	0.16
PsO	12 (100)	39 (100)	na	1 (3)	1 (4.2)	1.00	0	1 (3)	1.00
Dactylitis	0	1 (2.6)	1.00	0	0	na	1 (10)	0	0.23
Physical examination
Enthesitis, n (%)	5 (41.7)	14 (35.9)	0.74	4 (12.1)	8 (33.3)	0.10	0 (0)	8 (24.2)	0.17
Number of SPARCC enthesitis points, mean (±SD)	1.4 (±2.9)	0.9 (±1.4)	0.22	0.4 (±1.2)	1.0 (±1.5)	0.047	0 (±0)	0.7 (±1.6)	0.02
Dactylitis, n (%)	3 (25)	1 (2.6)	0.04	0	0	na	0	0	na
Hip involvement, n (%) e	3 (25)	4 (10.3)	0.33	5 (15.2)	1 (4.2)	0.38	1 (10)	4 (12.1)	1.00
PsO, n (%)	11 (91.7)	34 (87.2)	1.00	0 (0)	1 (4.2)	0.42	1 (10)	1 (3)	0.42
Peripheral arthritis, n (%)	4 (33.3)	6 (15.4)	0.22	1 (3.0)	2 (8.3)	0.57	0	0	na
Number of swollen joints, mean (±SD)	1.1 (±1.7)	0.3 (±0.8)	0.15	0.1 (±0.3)	0.1 (±0.3)	0.79	0 (±0)	0 (±0)	na
BASMI, mean (±SD)	2.4 (±1.0)	2.1 (±0.9)	0.36	2.2 (±1.0)	1.9 (±0.9)	0.35	3.0 (±1.2)	2.1 (±0.9)	0.009
EDASMI, mean (±SD)	2.8 (±1.1)	2.1 (±1.3)	0.08	1.9 (±1.2)	1.9 (±0.9)	0.85	2.7 (±1.4)	2.0 (±1.2)	0.15
Laboratory
B27 positive, n (%)	5 (41.7)	2 (5.1)	0.005	26 (78.8)	15 (62.5)	0.24	5 (50)	4 (12.1)	0.02
CRP abnormal (≥6 mg/L), n (%)	3 (25.0)	2 (5.1)	0.08	8 (24.2)	1 (4.2)	0.06	7 (70)	5 (15.2)	0.002
CRP, mean (±SD)	7.9 (±11.8)	2.2 (±2.6)	0.13	6.9 (±11.5)	2.1 (±3.3)	0.03	25.7 (±32.0)	4.1 (±8.3)	0.06

^*AAU, acute anterior uveitis; AS, ankylosing spondylitis; axSpA, axial spondyloarthritis; BASMI, Bath Ankylosing Spondylitis Metrology Index; bioDMARD, biologic disease‐modifying antirheumatic drug; CRP, C‐reactive protein; csDMARD, conventional synthetic disease‐modifying antirheumatic drug; EDASMI, Edmonton Ankylosing Spondylitis Metrology Index; IBD, inflammatory bowel disease; na, not applicable; NRS, numerical rating scale; NSAID, nonsteroidal anti‐inflammatory drug; PsO, psoriasis; SASPIC‐2, Screening for Axial Spondyloarthritis in Psoriasis, Iritis, and Colitis cohort 2; SpA, spondyloarthritis; SPARCC, Spondyloarthritis Research Consortium of Canada.

^a AS, PsO, AAU, and IBD are included.

^b Severity of current back pain is measured as the amount of back and/or hip pain during the last week on a 0 to 10 (most severe) NRS.

^c csDMARDs include methotrexate and salazopyrin.

^d bioDMARDs include etanercept, golimumab, adalimumab, infliximab, and certolizumab.

^e Hip involvement is defined as groin pain and/or restricted range of internal rotation.

More patients with axSpA in the groups with AAU and IBD were deemed to have IBP with a probability >5 on a 0 to 10 scale, and this was particularly evident in SASPIC‐1 (Supplementary Data). This was not consistently evident in those with PsO, and the proportion with high probability of IBP was even greater in those diagnosed with non‐axSpA in the group with PsO of SASPIC‐2. Only a minority of patients reported a complete or very good response to NSAIDs, and this did not differ significantly according to diagnosis. For patients with PsO in both cohorts, a complete or very good NSAID response was even reported more frequently in those diagnosed with non‐axSpA.

Physical examination did not reveal any significant differences in frequencies of patients with axSpA on the basis of Spondyloarthritis Research Consortium of Canada entheseal tender points, peripheral arthritis, or hip involvement between those with and without axSpA in any group (Tables 1 and 2). There were numerically more patients with enthesitis on physical examination in patients with non‐axSpA than those with axSpA, with the sole exception of the subgroup with PsO in SASPIC‐2, in which more patients and a higher mean enthesitis count were evident in those diagnosed with axSpA. Peripheral arthritis was more frequent on physical examination in the group with PsO, especially in SASPIC‐2, though without significant differences between those diagnosed with and without axSpA. Dactylitis was more frequent in patients with PsO diagnosed with axSpA in SASPIC‐2 but not in SASPIC‐1. Overall, musculoskeletal physical examination findings were not helpful in differentiating patients with axSpA versus non‐axSpA.

HLA‐B27 was present in 33.3%, 84.4%, and 37.2% of patients diagnosed with axSpA who had PsO, AAU, and IBD, respectively, in SASPIC‐1, whereas the respective frequencies were 41.7%, 78.8%, and 50% in SASPIC‐2 (Tables 1 and 2). It was significantly more frequent in patients with axSpA versus non‐axSpA among those with AAU (84.4% vs 46.4%) and IBD (37.2% vs 16.3%) in SASPIC‐1 and among those with PsO (41.7% vs 5.1%) and IBD (50% vs 12.1%) in SASPIC‐2. CRP was mostly higher in patients with axSpA versus non‐axSpA, this being more evident in SASPIC‐2, although only a minority of patients with axSpA had a level greater than normal. Among patients with IBD, there were 52 (55.3%) with Crohn disease and 42 (44.7%) with ulcerative colitis in SASPIC‐1 (Supplementary Table 4) and 23 (53.5%) with Crohn disease and 20 (46.5%) with ulcerative colitis in SASPIC‐2 (Supplementary Table 5). There were few differences among these subgroups according to diagnosis of axSpA, except for the level of CRP, which was significantly higher between patients with Crohn disease diagnosed with axSpA versus non‐axSpA.

Imaging characteristics

Radiographic sacroiliitis meeting the mNY criteria was recorded by local readers in 42.9%, 57.8%, and 56.8% of patients diagnosed with axSpA who had PsO, AAU, or IBD, respectively, in SASPIC‐1 (Table 3). The corresponding percentages in SASPIC‐2 were 33.3%, 42.4%, and 60%. Values for patients with non‐axSpA were <5% in SASPIC‐1 and 7.7%, 16.7%, and 6.1% for patients with PsO, AAU, and IBD, respectively, in SASPIC‐2. Radiographic sacroiliitis in patients with axSpA was recorded less frequently by central readers, especially in SASPIC‐1 (PsO, AAU, and IBD: 15%, 33.3%, and 25.6% for SASPIC‐1 and 41.7%, 36.4%, and 50% for SASPIC‐2). Only two patients with non‐axSpA were recorded by central readers with radiographic sacroiliitis, both being in the AAU group of SASPIC‐1.

Table 3.

Imaging features of patients presenting with undiagnosed back pain in two inception cohorts of patients recruited to SASPIC‐1 and ‐2 according to final local rheumatologist diagnosis of axSpA versus non‐axSpA^*

Imaging variable	Patients with PsO			Patients with AAU			Patients with IBD
	axSpA	Non‐axSpA	P value	axSpA	Non‐axSpA	P value	axSpA	Non‐axSpA	P value
SASPIC‐1 (n = 212)
Radiography
Total	21 (46.7)	24 (53.3)	–	45 (61.6)	28 (38.4)	–	44 (46.8)	50 (53.2)	–
Radiographic sacroiliitis by local readers a	9 (42.9)	1 (4.2)	0.003	26 (57.8)	1 (3.6)	<0.001	25 (56.8)	2 (4.1)	<0.001
Radiographic sacroiliitis by central readers b	3 (15)	0 (0)	0.086	15 (33.3)	2 (7.1)	0.010	11 (25.6)	0 (0)	<0.001
Unilateral grade 2 sacroiliitis by central readers c	1 (5)	2 (8.3)	1.0	9 (20)	1 (3.6)	0.078	6 (14.3)	2 (4.2)	0.139
MRI of SIJs (n = 132)
Total	14 (43.8)	18 (56.3)	–	29 (63.0)	17 (37.0)	–	23 (42.6)	31 (57.4)	–
MRI of SIJs indicative of axSpA by central readers d	6 (42.9)	1 (5.6)	0.027	18 (62.1)	1 (5.9)	<0.001	9 (39.1)	0 (0)	<0.001
MRI positive by ASAS definition by central readers e	4 (28.6)	1 (5.6)	0.14	13 (44.8)	0 (0)	<0.001	7 (30.4)	0 (0)	0.001
MRI of the spine (n = 84)
Total	9 (47.4)	10 (52.6)	–	17 (58.6)	12 (41.4)	–	14 (38.9)	22 (61.1)	–
MRI of the spine indicative of axSpA by central readers f	2 (22.2)	1 (10)	0.58	9 (52.9)	0 (0)	0.003	4 (28.6)	0 (0)	0.017
SASPIC‐2 (n = 151)
Total	12 (23.5)	39 (76.5)	–	33 (57.9)	24 (42.1)	–	10 (23.3)	33 (76.7)	–
Radiographic sacroiliitis by local readers g	4 (33.3)	3 (7.7)	0.044	14 (42.4)	4 (16.7)	0.039	6 (60)	2 (6.1)	<0.001
Radiographic sacroiliitis by central readers c , h	5 (41.7)	0 (0)	<0.001	12 (36.4)	0 (0)	0.001	5 (50)	0 (0)	<0.001
Unilateral grade 2 sacroiliitis by central readers c	1 (8.3)	1 (2.6)	1.0	3 (9.4)	1 (4.3)	0.632	1 (10)	1 (3)	1.0
MRI of SIJs indicative of axSpA by central readers i	9 (75)	0 (0)	<0.001	21 (63.6)	0 (0)	<0.001	8 (80)	0 (0)	<0.001
MRI positive by ASAS definition by central readers j	5 (41.7)	0 (0)	<0.001	17 (51.5)	0 (0)	<0.001	5 (50)	1 (3)	0.001

^*Values are n (%) unless indicated otherwise. AAU, acute anterior uveitis; ASAS, Assessment of Spondyloarthritis International Society; axSpA, axial spondyloarthritis; CT, computed tomography; IBD, inflammatory bowel disease; MRI, magnetic resonance imaging; PsO, psoriasis; SASPIC‐1, Screening for Axial Spondyloarthritis in Psoriasis, Iritis, and Colitis cohort 1; SASPIC‐2, Screening for Axial Spondyloarthritis in Psoriasis, Iritis, and Colitis cohort 2; SIJ, sacroiliac joint.

^a Included are four patients who had CT, of whom one had radiographic sacroiliitis.

^b A total of 17 patients were adjudicated for radiographic sacroiliitis by a third central reader.

^c Excluded are four and one patients, respectively, for whom neither CT nor x‐ray were provided for central review for SASPIC‐1 and SASPIC‐2, respectively.

^d Nine patients were adjudicated for MRI indicative of axSpA by a third central reader.

^e A total of 22 patients were adjudicated for ASAS positive MRI by a third central reader.

^f Five patients were adjudicated for MRI of the spine indicative of axSpA by a third central reader.

^g Included are three patients who had CT, of whom two had radiographic sacroiliitis.

^h Eight patients were adjudicated for radiographic sacroiliitis by a third central reader.

ⁱ Eight patients were adjudicated for MRI indicative of axSpA by a third central reader.

^j A total of 12 patients were adjudicated for ASAS positive MRI by a third central reader.

MRI of the SIJs was conducted in 132 patients (62.3%) in SASPIC‐1 and in all patients of SASPIC‐2 (Table 3). MRI evaluation in SASPIC‐1 was conducted mostly in patients who were negative for radiographic sacroiliitis (76.7%). Central readers considered the MRI indicative of axSpA in the SIJs of 42.9%, 62.1%, and 39.1% patients diagnosed with axSpA and PsO, AAU, IBD, respectively, in SASPIC‐1. The corresponding percentages in patients in SASPIC‐2 diagnosed with axSpA were 75%, 63.6%, and 80%.

MRI of the spine was conducted in 84 patients (39.6%) in SASPIC‐1, all of whom also had MRI of the SIJs (Table 3). Spinal MRI was considered indicative of axSpA in 22.1%, 52.9%, and 28.6% of patients diagnosed with axSpA who had PsO, AAU, or IBD, respectively, and in one patient (10%) diagnosed with non‐axSpA who had PsO. There were no patients with MRI indicative of axSpA confined to the spine, 11 patients for whom only the SIJs were positive, and 16 for whom both the SIJs and spine were positive. There were few differences in radiographic or MRI findings when patients with axSpA with Crohn disease and ulcerative colitis were compared (Supplementary Table 6).

Diagnosis of axSpA at sequential stages of evaluation

For patients presenting with undiagnosed back pain who were HLA‐B27 positive, axSpA was diagnosed in 70%, 74.5%, and 66.7% after final diagnostic evaluation in patients with PsO, AAU, and IBD, respectively, in SASPIC‐1, and in 71.4%, 87.8%, and 55.6% after final diagnostic evaluation in patients with PsO, AAU, and IBD, respectively, in SASPIC‐2 (Figure 1). The final diagnoses of axSpA in patients who were HLA‐B27 negative were made in 41.2%, 31.8%, and 39.7% of patients with PsO, AAU, and IBD, respectively, in SASPIC‐1, and in 15.9%, 43.8%, and 14.7% of patients with PsO, AAU, and IBD in SASPIC‐2. Diagnostic ascertainment of axSpA decreased with successive stages of evaluation irrespective of referral group or B27 positivity (Figure 1). This was particularly evident in SASPIC‐2 after diagnostic evaluation with MRI by local readers and in patients who were HLA‐B27 negative. The frequency of patients who were positive for axSpA always decreased after local reader MRI evaluation in SASPIC‐1 and ‐2, and a further relatively minor decrease was evident in some subgroups of SASPIC‐2 after central reader evaluation, particularly patients with PsO or IBD who were HLA‐B27 negative. The local rheumatologist level of confidence in ruling in and ruling out the diagnosis increased primarily after MRI evaluation (Supplementary Figure 2).

Figure 1.

Diagnosis of axSpA by local rheumatologists according to referral source and HLA‐B27 positivity at sequential stages of evaluation of patients presenting with undiagnosed back pain in two inception cohorts of patients recruited to the Screening for Axial Spondyloarthritis in Psoriasis, Iritis, and Colitis co. (A) SASPIC 1 (all patients). (B) SASPIC 1 patients with MRI scan. (C) SASPIC 2. axSpA, axial spondyloarthritis; IBD, inflammatory bowel disease; MRI, magnetic resonance imaging; SASPIC 1, Screening for Axial Spondyloarthritis in Psoriasis, Iritis, and Colitis cohort 1; SASPIC 2, Screening for Axial Spondyloarthritis in Psoriasis, Iritis, and Colitis cohort 2. Color figure can be viewed in the online issue, which is available at http://onlinelibrary.wiley.com/doi/10.1002/art.42967/abstract.

Performance of the 2009 ASAS classification criteria in SASPIC‐1 and ‐2

Sensitivity and specificity of the ASAS classification criteria using the final rheumatologist evaluation as gold standard were 66.4% and 79.4% in SASPIC‐1, and performance was somewhat better in analyses of patients with high diagnostic confidence and those with symptom duration of five years or longer (Table 4). Performance was comparable in male and female patients for the overall criteria and for the clinical arm. For the imaging arm, sensitivity in male patients (50%) was almost double that in female patients (27.3%) at the expense of slightly lower specificity (92% in male patients vs 100% in female patients). When only the subset of patients in SASPIC‐1 who had MRI were analyzed, the performance of the criteria was similar to the entire SASPIC‐1.

Table 4.

Performance of the 2009 ASAS classification criteria in patients presenting with undiagnosed back pain recruited to SASPIC‐1 and ‐2 with the final diagnosis of the rheumatologist as gold standard^*

Patient category a	Number	Overall ASAS criteria a		Imaging arm a		Clinical arm
		Sen	Spec	Sen	Spec	Sens	Spec
SASPIC‐1
All patients	212	66.4	79.4	40.9	97.1	53.6	81.4
Patients diagnosed with confidence >7 for axSpA and <−4 for non‐axSpA (−10 to +10 scale)	163	75	82.8	51.3	97.7	59.2	85.1
Patients with symptom duration of five years or longer	132	71.8	83.6	40.8	98.4	57.7	83.6
Patients with symptom duration under five years	80	56.4	73.2	41	95.1	46.2	78
Male	108	69.7	78.6	50	92.9	53	83.3
Female	104	61.4	80	27.3	100	54.5	80
Patients with MRI in SASPIC‐1
All patients	132	65.2	75.8	43.9	98.5	48.5	77.3
Patients diagnosed with confidence >7 for axSpA and <−4 for non‐axSpA (−10 to +10 scale)	99	75.6	81	58.5	98.3	53.7	82.8
Patients with symptom duration of five years or longer	78	70.7	78.4	43.9	100	51.2	78.4
Patients with symptom duration under five years	54	56	72.4	44	96.6	44	75.9
Male	66	65	76.9	55	96.2	42.5	80.8
Female	66	65.4	75	26.9	100	57.7	75
SASPIC‐2
All patients	151	85.5	82.3	69.1	99	61.8	83.3
Patients diagnosed with confidence >7 for axSpA yes and <−4 for non‐axSpA (−10 to +10 scale)	126	92.3	81.6	87.2	98.9	66.7	82.8
Patients with symptom duration of five years or longer	84	90.3	75.5	67.7	98.1	80.6	77.4
Patients with symptom duration under five years	67	79.2	90.7	70.8	100	37.5	90.7
Male	75	82.1	88.9	69.2	100	59	88.9
Female	76	93.8	78.3	68.8	98.3	68.8	80

^*ASAS, Assessment of Spondyloarthritis International Society; axSpA, axial spondyloarthritis; MRI, magnetic resonance imaging; SASPIC‐1, Screening for Axial Spondyloarthritis in Psoriasis, Iritis, and Colitis cohort 1; SASPIC‐2, Screening for Axial Spondyloarthritis in Psoriasis, Iritis, and Colitis cohort 2; Sen, Sensitivity; Spec, Specificity.

^a Central reader assessment of radiographic sacroiliitis and ASAS positive MRI were used for the analysis of the performance of the ASAS classification criteria and clinical and imaging arms.

Using the final rheumatologist diagnostic evaluation after receipt of central reader imaging data as gold standard, sensitivity and specificity of the ASAS criteria were 85.5% and 82.3%, respectively, in SASPIC‐2, and thereby higher than in SASPIC‐1. This was mainly due to higher sensitivity, which was also evident for patients diagnosed with axSpA with high confidence (Table 4). The greatest beneficial impact of the availability of central reader imaging data was observed in female patients and the subset with symptom duration of shorter than five years. The performance of the criteria varied somewhat between subgroups according to the referral source, but sensitivity was comparatively low in all three subgroups, especially in patients with IBD, and in both SASPIC‐1 and ‐2 (Table 5). Both sensitivity and specificity were greater in SASPIC‐2 for all subgroups, especially for patients with PsO, but sensitivity remained comparatively low.

Table 5.

Performance of the 2009 ASAS classification criteria in three groups of patients presenting with undiagnosed back pain recruited to SASPIC‐1 and ‐2 with the final diagnosis of the rheumatologist as gold standard^*

Patient category a	Patients with psoriasis			Patients with AAU			Patients with IBD
	Total	Sen	Spec	Total	Sen	Spec	Total	Sens	Spec
All patients of SASPIC‐1	45	42.9	83.3	73	51.1	92.9	94	36.4	90
Patients with MRI in SASPIC‐1	32	42.9	77.8	46	55.2	100	54	30.4	87.1
SASPIC‐2	51	75	94.9	57	66.7	100	43	50	90.9

^*AAU, acute anterior uveitis; ASAS, Assessment of Spondyloarthritis International Society; IBD, inflammatory bowel disease; MRI, magnetic resonance imaging; SASPIC‐1, Screening for Axial Spondyloarthritis in Psoriasis, Iritis, and Colitis cohort 1; SASPIC‐2, Screening for Axial Spondyloarthritis in Psoriasis, Iritis, and Colitis cohort 2; Sen, Sensitivity; Spec, Specificity.

^a Central reader assessment of radiographic sacroiliitis and ASAS positive MRI were used for the analysis of the performance of the 2009 ASAS classification criteria.

DISCUSSION

Data from these two cohorts include the first data incorporating systematic evaluation using MRI, have significant implications for clinical practice, and emphasize the crucial role of MRI. First, the frequency of axSpA was about 25% in those presenting with PsO or IBD and 60% in those with AAU. Second, diagnosis was delayed by up to nine years, this being comparable to the delay reported in other cohorts of patients with axSpA ¹ despite the known association with extra‐articular features and the severity of back pain. These two observations support recent referral recommendations that advocate referral to a rheumatologist of patients with chronic back pain, symptom onset at younger than 45 years of age, and extra‐articular features related to axSpA. ² , ¹⁶ Third, musculoskeletal clinical features largely failed to discriminate between axSpA and non‐axSpA, and some SpA features, such as enthesitis, were even more common in non‐axSpA. IBP in patients with PsO was more common in those diagnosed without axSpA. Fourth, good response to NSAIDs did not discriminate between axSpA and non‐axSpA, consistent with a recent report. ²¹ Fifth, family history was nondiscriminatory, a history of PsO even being higher in patients with non‐axSpA. Sixth, the frequency of HLA‐B27 was only 30% to 50% in patients with PsO or IBD diagnosed with axSpA, being closer to 50% of patients in SASPIC‐2 with axSpA for whom MRI of the SIJs was conducted in all patients. Conversely, being positive for HLA‐B27 was associated with axSpA in 56% to 88% of patients, the highest figure being evident in patients in SASPIC‐2 with AAU. Seventh, MRI was almost twice as sensitive as radiography in detecting axSpA lesions in the SIJs, being positive in 60% to 80% of patients. Spine MRI was indicative of axSpA in a substantial minority, though there were no patients with only spinal involvement. MRI was especially helpful in ruling out a diagnosis of axSpA raised on clinical grounds because the frequency of axSpA diagnoses declined after MRI in both cohorts and in all subgroups. Finally, sensitivity and specificity of the ASAS classification criteria was impacted by systematic MRI assessment, and sensitivity was comparatively low in all three subgroups, consistent with the relatively low frequency of HLA‐B27 in patients who were axSpA positive with concomitant PsO and IBD.

We designed this unique study to reproduce the sequential process of diagnostic decision‐making in clinical practice and to document and weigh the contribution of each assessment at sequential stages in the process. The use of MRI appears to impact diagnostic decision‐making for these patients in two primary ways. The majority of patients were considered to have axSpA after clinical evaluation, especially if they were HLA‐B27 positive, and rheumatologists primarily ordered MRI evaluation in radiograph negative/equivocal patients in SASPIC‐1. ²² In patients in SASPIC‐2, a negative MRI also appeared to provide the confidence to rule out axSpA, and this is most likely the reason for the lower frequency of axSpA after MRI evaluation, especially for patients with PsO and IBD. The concept that MRI contributes to an accurate diagnosis is supported by the enhanced performance of the ASAS criteria in SASPIC‐2 for which MRI was a required evaluation, especially in patients with PsO and IBD.

Two recent reports have assessed the frequency of axSpA in patients presenting with AAU, included MRI per clinical need, and stated that the ASAS criteria were used to diagnose axSpA. In the Spanish Sentinel study, 50.2% of all patients with AAU and 71.2% of patients who were HLA‐B27 positive presented with axSpA. ¹⁷ The frequency of axSpA was lower than in the SASPIC cohorts, but only 44% had IBP in the Sentinel study given that all patients presenting with AAU were evaluated. The Irish Dublin Uveitis Evaluation Tool (DUET) study recruited discovery and validation cohorts for the development of a screening strategy for axSpA in patients presenting with AAU. ¹⁶ The frequency of axSpA was 38.5% and 37.8% in the discovery and validation cohorts, respectively. MRI was conducted according to clinical need, although it was not stated how many patients had MRI or what the MRI findings were. The frequency of axSpA was lower in the DUET study versus SASPIC, but only two‐thirds of patients in the DUET study had back pain. Neither of these nor additional studies evaluating axSpA in patients with AAU ²³ , ²⁴ systematically compared patients with back pain diagnosed with or without axSpA for discriminatory features, and none conducted systematic MRI.

A recent prospective, multicenter study conducted in Berlin applied a dermatologist‐centered screening/referral tool focusing on detecting axial involvement in patients with PsO and undiagnosed chronic back pain. ²⁵ The diagnosis of axial psoriatic arthritis (axPsA) was made in 14 of 100 patients (14%) evaluated by a rheumatologist and 9 patients (64.3%) fulfilled the ASAS classification criteria for axSpA. HLA‐B27 was present in 28.6% of those diagnosed with axPsA versus 14.8% in those without axPsA. The somewhat higher frequency of axPsA in SASPIC‐2 (23.5%) versus Berlin (14%) cohorts may reflect differences between Canadian and German rheumatologists in the perceived importance of a positive MRI for diagnosis (75% vs 100% in SASPIC vs Berlin cohorts, respectively). These reports are consistent with the prevalence of 5% to 28% for axPsA reported in patients with early‐stage disease. ²⁶ , ²⁷ , ²⁸ , ²⁹ , ³⁰ , ³¹ , ³² , ³³ , ³⁴ An important argument in favor of systematic MRI evaluation is that only 25% to 45% of patients with radiographic axPsA have been reported to have IBP, whereas axSpA‐based IBP criteria have been shown to lack specificity for axPsA. ³⁵ , ³⁶ , ³⁷ , ³⁸ , ³⁹ , ⁴⁰

There are several study limitations. First, there is no international consensus on a case definition of axPsA and little information on the sensitivity and specificity of MRI lesions in the axial skeleton. Consequently, MRI readers relied on their expertise in deciding as to the presence of a positive MRI. Second, MRI evaluation for spinal lesions was only conducted on an as‐required basis by local rheumatologists in SASPIC‐1, and only MRI of the SIJs was required in SASPIC‐2. Third, we confined systematic radiographic assessment to the SIJs, consistent with many prior studies that have relied on defining axPsA according to the mNY criteria. It is possible that radiographic findings in the spine, particularly new bone formation, may have captured additional patients with axial disease. Fourth, our conclusions are based on patients presenting with undiagnosed back pain, and we acknowledge that prospective follow‐up is likely to demonstrate higher numbers of patients diagnosed with axial inflammation. Fifth, we confined our analysis to patients presenting with back and/or buttock pain, and we acknowledge that patients who were asymptomatic may not have been detected. However, it is unclear how this might have impacted management. We also excluded patients receiving biologic DMARD for their extra‐articular disease because this could obscure clinical and imaging features of axSpA, especially inflammation. However, our findings regarding the value of MRI may be further reinforced because MRI is the most sensitive means to detect structural lesions in the SIJs, which may be the only feature pointing to axSpA in such patients. In view of the importance that rheumatologists attach to MRI, particularly central reader evaluations, there could be concern regarding circular reasoning in formulating the diagnostic conclusion that a patient has axSpA. However, it is noteworthy that rheumatologists diagnosed axSpA despite negative imaging by central readers in 20% to 36% of patients diagnosed with axSpA in SASPIC‐2, the cohort in which all patients had central MRI evaluation. Furthermore, the major change in local rheumatologist diagnostic conclusion occurred after local radiologist assessment of the MRI scan and not central reader evaluation. The impact of MRI assessment was mainly observed in patients who were HLA‐B27 negative, and this was aimed at ruling out the diagnosis. A major strength of our study is the unique study design based on a stepwise diagnostic process that enabled evaluation of the contribution of clinical and laboratory features to diagnosis independently of information provided by imaging. We observed that the performance of the ASAS classification criteria improved in SASPIC‐2, the cohort in which MRI evaluation was conducted in every patient, supporting the argument that systematic assessment of SIJs by MRI in all patients helps to correctly identify the patient with axSpA.

In conclusion, our multicenter stepwise evaluation of two Canadian cohorts of patients presenting with undiagnosed back pain and PsO, IBD, or AAU culminated in a diagnosis of axSpA in 25% of patients with PsO or IBD and 60% in patients with AAU. Diagnostic ascertainment of axSpA decreased with successive stages of evaluation irrespective of referral group or B27 status, especially after diagnostic evaluation with MRI by local readers and in patients who were HLA‐B27 negative. Our data support the benefit of recent referral recommendations that advocate referral to a rheumatologist of patients with chronic back pain and extra‐articular features related to axSpA. Our data also demonstrate that MRI evaluation of all patients contributes to accuracy of diagnosis and the performance of classification criteria because clinical features are not helpful in identification of axSpA, and frequency of HLA‐B27 is lower than in axSpA without these features. Finally, these findings have significant implications for clinical trials that recruit patients on the basis of rheumatologist opinions that do not incorporate evaluation of MRI scans.

AUTHOR CONTRIBUTIONS

All authors contributed to at least one of the following manuscript preparation roles: conceptualization AND/OR methodology, software, investigation, formal analysis, data curation, visualization, and validation AND drafting or reviewing/editing the final draft. As corresponding author, Dr Maksymowych confirms that all authors have provided the final approval of the version to be published, and takes responsibility for the affirmations regarding article submission (eg, not under consideration by another journal), the integrity of the data presented, and the statements regarding compliance with institutional review board/Helsinki Declaration requirements.

ROLE OF THE STUDY SPONSOR

AbbVie Canada and Janssen Canada had no role in the study design or in the collection, analysis, or interpretation of the data, the writing of the manuscript, or the decision to submit the manuscript for publication. Publication of this article was not contingent upon approval by Abbvie Canada or Janssen Canada.

Supporting information

Disclosure form

Appendix S1: Supporting Information.