Dr. Strand has received consultant fees, speaking fees, and/or honoraria (less than $10,000 each) from AbbVie, Inc., Amgen, Janssen, Sanofi-Aventis, and UCB, and has served on the advisory boards for AbbVie, Inc., Amgen Corporation, AstraZeneca, BiogenIdec, Biotest, Celgene, Genentech/Roche, Incyte, Janssen, Lilly, Novartis Pharmaceuticals, NovoNordisk, Pfizer, Sanofi-Aventis, and UCB.
Sumati A. Rao,
AbbVie, Inc., North Chicago, Illinois
AbbVie, Inc., 1 North Waukegan Road, D-GMH1, AP31-1 NE, North Chicago, IL 60064. E-mail: email@example.com
Dr. Ruderman has received consultant fees, speaking fees, and/or honoraria (less than $10,000 each) from Amgen, Janssen, Celgene, CVS/Caremark, and Genentech/Roche, and (more than $10,000) from AbbVie, Inc., and has served on the advisory boards for AbbVie, Inc., Amgen, BMS, Human Genome Sciences, and Pfizer.
New classification criteria for axial spondyloarthritis (SpA) have been validated by the Assessment of SpondyloArthritis international Society (ASAS) working group. We applied these criteria to estimate prevalence of SpA in randomly selected, retrospectively reviewed medical records from representative US rheumatology practices.
Rheumatologists from 101 US practices identified at-risk patients, ages 18–44 years, with chronic back pain. Medical records were reviewed against ASAS criteria. The proportion of patients meeting ASAS criteria was compared to an estimate of the total number of at-risk patients treated at participating sites and, following weighting, was extrapolated to 5,520 US rheumatology practices. US Census data were used to estimate national prevalence.
In a sample of 816 randomly selected records, 514 (63%) at-risk patients (95% confidence interval [95% CI] 59.6–66.3%) met ASAS criteria. By applying this proportion to 1,217,097 Americans estimated at risk, 766,652 were projected to meet ASAS criteria. This projection corresponds to a national prevalence of 0.70% (95% CI 0.38–1.1%) or 701 per 100,000 individuals. The prevalence estimates of ankylosing spondylitis and nonradiographic axial SpA are 0.35% (95% CI 0.18–0.554%) and 0.35% (95% CI 0.18–0.554%), respectively. Rheumatologists diagnosed axial SpA in 491 (60%) of those at risk, corresponding to 0.67% (95% CI 0.36–1.01%) prevalence overall. However, of 514 patients meeting ASAS criteria, 124 (24%) were undiagnosed by rheumatologists.
This is the first systematic epidemiology study of axial SpA using ASAS criteria. Better recognition of axial symptoms is needed, as rheumatologists' expert clinical diagnoses are not always in agreement with ASAS criteria.
Spondyloarthritis (SpA) comprises rheumatic diseases characterized by pain and inflammation of the spine and sacroiliac joints. As a group, SpA is one of the most common rheumatic diseases, with prevalence estimated at 0.3–2.5% in Western Europe ([1, 2]). The definition of SpA is evolving; however, the Assessment of SpondyloArthritis international Society (ASAS) working group recently established classification criteria to distinguish 2 broad categories of SpA ([3-5]). The first category, peripheral SpA, includes diseases affecting predominantly peripheral joints, such as reactive arthritis, psoriatic arthritis, and undifferentiated SpA. The second category, axial SpA, includes those diseases with predominantly axial involvement, including ankylosing spondylitis (AS) and nonradiographic axial SpA, in which the prominent symptom is back pain.
Axial SpA is associated with a large burden of disease, including substantial pain, stiffness, and impaired physical function ([6, 7]). Despite this burden, a substantial delay between disease onset and diagnosis and treatment often occurs, particularly in those with axial SpA lacking radiographic manifestations ([6, 8]). Because this delay appears related in part to under-recognition of SpA symptoms (), availability of criteria specific to axial SpA may result in earlier diagnosis and treatment and better understanding of its epidemiology ([2, 10]).
Relatively little is known regarding the epidemiology of axial SpA, particularly in the US (). One recent US study () estimated prevalence of axial SpA to be approximately 1% of the adult population, based on an arthritis questionnaire developed from the Amor Classification Criteria for SpA and European Spondyloarthropathy Study Group (ESSG) criteria included in the 2009 to 2010 US National Health and Nutrition Examination Survey. Availability of improved ASAS classification criteria provides a new means to facilitate understanding of axial SpA epidemiology. This study sought to examine the prevalence of axial SpA based on these criteria, using clinical data from patients in representative US rheumatology practices. This is the first study to estimate prevalence of axial SpA based on ASAS criteria, and also includes an assessment of the degree of consistency between rheumatologists' expert clinical diagnoses of axial SpA and these criteria.
Box 1. Significance & Innovations
This is the first study to assess prevalence of axial spondyloarthritis (SpA) based on new Assessment of SpondyloArthritis international Society (ASAS) classification criteria for nonradiographic axial SpA.
We estimated that axial SpA affects nearly 0.7% of the US population ages 18–44 years in a retrospective chart review of at-risk patients in US rheumatology practices.
A majority of patients in this age group who are referred to rheumatologists for inflammatory back pain are likely to meet the ASAS criteria for axial SpA.
Half of all SpA manifestations in this younger age group are not radiographically confirmed.
Approximately 24% of patients meeting ASAS criteria were not diagnosed by rheumatologists, indicating a need for increased awareness of these guidelines.
PATIENTS AND METHODS
Overall study design.
This was a retrospective cohort study with the objective of estimating the prevalence of axial SpA in at-risk patients treated by a representative sample of US rheumatologists between January 1985 and August 2011. We applied these prevalence estimates to US rheumatology practices in general and used US Census data to obtain a national prevalence estimate for Americans ages 18–44 years.
A retrospective medical chart review was performed on a random sample of at-risk patients in which clinical data documented in the medical records were screened against ASAS criteria to obtain a local prevalence estimate of the number of patients meeting criteria at each site. Patients were also classified according to whether or not a rheumatologist's expert clinical diagnosis of axial SpA had been made, irrespective of medical record documentation in support of ASAS criteria. Following poststratification weighting on region, a national at-risk pool was calculated by applying local site at-risk estimates to all US rheumatology practices. US estimates of the number of SpA cases were calculated by applying the proportion of afflicted individuals at participating sites to the national at-risk pool treated in rheumatology practices. This figure was then applied to US Census data to obtain a national prevalence estimate.
Participating sites and data collection.
A total of 101 rheumatologists were recruited from a pool of 5,520 US sites identified in a national database () using a systematic random recruitment methodology. Census region (Northeast, Midwest, West, and South) was used to stratify physician recruitment to ensure geographic representation. Rheumatologists were recruited via telephone and underwent screening prior to study acceptance and provision of study materials. To be eligible for acceptance, rheumatologists must have 1) been currently treating patients with axial SpA, 2) had infrastructure available to identify patients at risk and abstract data from records, 3) not acted as a consultant or clinical investigator for a sponsor involved in the manufacture of medications for treatment of axial SpA/AS, and 4) agreed to all study rules, including involvement in resolution of data queries and data validation processes. Referral sites conducting axial SpA clinical trials were excluded to reduce potential artificial inflation of prevalence estimates.
Data obtained from sites included total patient caseload, number of patients currently treated (prior 12 months) meeting inclusion criteria for the study, i.e., at risk for axial SpA, defined as chronic back pain of at least 3 months' duration in the age range of 18–44 years, and proportion diagnosed with axial SpA. The data collection procedure adhered to recommended best practices for retrospective chart review studies (). An initial pilot study was performed to test the feasibility of the case report form (CRF) and to make required edits. Instruction materials outlining patient selection methods were included with the CRF. Once complete, data validations were performed on a random selection of 30% of CRFs, checking key variables against source documents. Additionally, machine validations and logic checks were performed to ensure no contradictory data were present and to identify outliers and abnormal data that required further review or followup.
Rheumatologists selected up to 10 currently treated at-risk patients with sufficient medical record data available. A random selection method utilizing a combination of date of last office visit and patient birth month as selection variables was employed. Site personnel abstracted patient demographics (age, sex, race/ethnicity, weight, height, and marital status) and documented the presence of SpA clinical features specific to the ASAS criteria: inflammatory back pain, arthritis, uveitis, psoriasis, enthesitis, dactylitis, Crohn's disease/ulcerative colitis/inflammatory bowel disease, alternating buttock pain, limited lumbar spine motion and chest expansion, and other clinical parameters, including HLA–B27 positivity, family history, C-reactive protein, erythrocyte sedimentation rate, magnetic resonance imaging (MRI), radiographic sacroiliitis assessment, and response to treatment with nonsteroidal antiinflammatory drugs. Data were submitted to a central analytics organization, either on paper or via the internet, and were subjected to a validation process to ensure accuracy of data collected.
Cases were defined as meeting ASAS criteria based upon the schematic presented in Figure 1. The case definition includes presence of HLA–B27, ASAS imaging criteria (MRI and/or radiographs), and other SpA features. Cases were subdivided as nonradiographic axial SpA or AS based on presence or absence of sacroiliitis of grade ≥2 on plain radiographs. The base-case model assumed that the absence of radiographic results in a chart was equivalent to a negative result. An alternative approach excluded patients without radiographic findings from the analysis. Classification was subject to documentation of all required data in the medical record. We also captured rheumatologists' expert clinical diagnoses of axial SpA where present.
We obtained site prevalence estimates of patients in the chart review sample. Following poststratification weighting based on geographic region, we multiplied the average number of at-risk patients per study site by the total number of regional rheumatology practices (Table 1). The national at-risk population was estimated from the sum of regional at-risk populations. This total at-risk population was multiplied by the proportion of patients within the study sample with either a rheumatologist's expert clinical diagnosis of axial SpA or who met ASAS criteria for axial SpA upon chart review to obtain the total number of patients with a diagnosis of axial SpA or who met ASAS criteria in US rheumatology practices. Finally, US Census data indicated a total population of 109,377,973 in the target age range. This number was used as the denominator to determine a national prevalence rate (Table 2) (). Importantly, this denominator takes into account the entire US population in the at-risk age group, rather than only those with chronic back pain evaluated by a rheumatologist. We anticipated our data to be sensitive both to estimates of the population at risk as reported by sites and site prevalence determined on chart review. This uncertainty was modeled around our prevalence point estimate using 95% confidence intervals (95% CIs) in a two-way fashion. Agreement between ASAS-consistent SpA and rheumatologists' expert clinical diagnoses was examined via a kappa statistic.
Table 1. Geographic distribution of at-risk patients within US rheumatology practices
aEstimate provided by rheumatologists.
Geographic representation of rheumatology practices
Table 1 shows the disposition of the physician sample by census region. The sample reflected adequate geographic representation such that only small weighting factors were required to correct for variance from actual distribution of US rheumatology practices. A total of 816 charts were reviewed. Table 3 shows the demographics for individuals included in the study sample. Two-thirds of patients within the at-risk sample were male. Patients were also predominantly white, with a mean age of 36 years. These demographics are not necessarily representative of national statistics, but they reflect the patient population referred to rheumatologists for treatment of chronic low back pain.
Table 3. Patient demographics (study sample)*
ASAS sample (n = 816)
Census (n = 109,377,973)
NHANES (9) (n = 5,103)
ASAS = Assessment of SpondyloArthritis international Society; NHANES = National Health and Nutrition Examination Survey.
Current age, years
Site prevalence in at-risk patients.
Of the 816 patients, 514 (63.0%) met ASAS criteria for axial SpA (Table 2). Of these, 124 (24.1%) did not have a documented rheumatologist's diagnosis of axial SpA. Conversely, 101 (33.4%) patients without medical record documentation sufficient to classify them as meeting ASAS criteria had a documented axial SpA diagnosis. Of the 124 patients who were classified with axial SpA according to the ASAS classification criteria but not by rheumatologist diagnosis, 94 (76%) had radiographic results available in their charts and 30 (24%) did not. Of these 94 patients, 52 (55%) had positive findings: conventional radiographic finding of sacroiliitis sufficient to fulfill the modified New York criteria for AS, and 42 (45%) had negative radiographs. Of the 52 patients with positive radiographs, 45 had a physician diagnosis of AS in their charts. Of the 42 patients with negative radiographs, 32 had a physician diagnosis of AS in their charts. Among 30 patients with no reports of radiographic results in the charts, 24 were diagnosed with AS by the physician.
Concordance between ASAS criteria and rheumatologists' expert clinical diagnoses yielded a kappa of 0.41, or low to moderate agreement. Use of a broad definition of SpA that included patients meeting ASAS criteria as well as those with a rheumatologists' expert clinical diagnosis yielded a site prevalence estimate of 75.5% of at-risk individuals. Prevalence of nonradiographic axial SpA by ASAS documentation in reviewed charts was 31.5% (n = 257).
National prevalence estimates.
Rheumatologists reported currently treating mean ± SD 220 ± 240 at-risk patients. Application of region-weighted site volume estimates to all 5,520 US rheumatology practices yielded an estimated pool of 1,217,097 Americans at risk (Table 1). Prevalence estimates based on ASAS criteria obtained via chart review (63%) were applied to this rheumatologist-treated at-risk population, yielding an estimate of 766,771 at-risk patients meeting ASAS criteria for axial SpA. US Census data indicated a total population of 109,377,973 in the same age range. Applying the national ASAS estimate to US Census data in this age range yielded a prevalence estimate of 0.701% (95% CI 0.38–1.06%) or 701 per 100,000 individuals (Table 2). Approximately half (258 of 514) of the axial SpA cases in the study sample met the criteria for nonradiographic axial SpA, corresponding to a prevalence of 0.35% (95% CI 0.18–0.554%) (Table 2). When 191 patients with no documented radiographic examinations were excluded from the analysis, 173 met criteria for nonradiographic axial SpA, corresponding to a prevalence of 0.31%. A total of 256 cases met the criteria for AS, corresponding to a prevalence of 0.35% (95% CI 0.18–0.554%). Rheumatologists diagnosed axial SpA in 491 of 816 (60%) patients at risk, corresponding to a 0.67% (95% CI 0.36–1.01%) prevalence overall.
Despite recent progress in understanding the etiology, classification, and clinical manifestations of axial SpA, limited literature exists regarding its epidemiology. This is the first US study to estimate prevalence based on patient-level clinical data. It is also the first systematic study to estimate axial SpA prevalence using the newly developed ASAS criteria for axial SpA. It includes 2 independent assessments, 1 according to rheumatologists' expert clinical diagnoses and 1 based on classification from patient chart data. The prevalence estimates based on these 2 methods are approximately 0.67% and 0.70%, respectively.
Diagnostic and classification criteria of SpA are presented in Figure 2. AS is the prototypical form of axial SpA and was the first for which standardized diagnostic criteria were developed. Modified New York criteria () emphasized radiographic assessment and established definite sacroiliitis of grade ≥2 on plain radiographs as a prerequisite for the diagnosis of AS. However, it was quickly recognized that many clinical features could be present in the absence of clear radiographic documentation of damage. This observation, combined with confirmation of a genetic association with HLA–B27, resulted in the development of 2 additional criteria. The Amor Classification Criteria for SpA () and ESSG criteria () both accounted for genetic factors and clinical features in addition to presence of sacroiliitis on plain radiographs for assessment of SpA, but differed considerably in emphasis of specific clinical features. These criteria allow for a broader definition of SpA relative to the modified New York criteria, but emphasize peripheral SpA rather than specifically axial symptomatology ().
ASAS criteria for axial SpA assess the presence of inflammation, HLA–B27, and clinical features that are specific to axial SpA and take advantage of the emergence of MRI as a sensitive method for detection of inflammation in the sacroiliac joints not well demonstrated by plain radiographs () (Figure 2). Using these criteria, patients are diagnosed with axial SpA according to 2 classification arms, one based on evidence of inflammation by imaging (either radiographic or MRI) and clinical features specific to SpA, and the other requiring presence of HLA–B27 and clinical features specific to SpA. Although it is possible that patients with mechanical or other noninflammatory (or, in some cases, inflammatory ) causes of back pain may be misclassified, these criteria have been demonstrated to be sensitive (82.9%) and specific (84.4%) for classification of axial SpA ().
The prevalence estimate of 0.70% according to ASAS criteria is lower than those based on Amor and ESSG criteria (0.9% and 1.4%, respectively) obtained by Reveille et al (). Although their study included a relatively large, nationally representative sample, it relied on less specific classification criteria, and several items pertinent to those criteria were not collected. For example, there were no data on suspected cases regarding physical examinations, radiographic findings, HLA–B27, peripheral arthritis, reactive arthritis, family history of SpA-related disorders, and dactylitis. As noted, the Amor and ESSG criteria emphasize peripheral features of spondyloarthropathy and may identify patients with limited or no axial involvement. In contrast, our study was based on improved criteria with a focus on axial disease, and it included clinical assessment by practicing rheumatologists.
Importantly, nonradiographic axial SpA accounted for up to half of all axial SpA cases meeting ASAS criteria in our study. The corresponding national prevalence estimate of 0.31% to 0.35% is approximately equivalent to the prevalence of AS. This observation is striking given that the diagnosis of axial SpA is less likely to be made in the absence of radiographic findings ([6, 8]). Our data support this view; of the 124 patients in our sample who met the ASAS criteria but were not diagnosed by rheumatologists, 72 (58%) had negative or no radiographic results, while 52 (42%) had positive radiographic findings (Table 2). While the progression of this disease is not well understood, current estimates suggest that only 8–12% of patients with nonradiographic axial SpA will progress to AS over a 2-year period ([19, 20]). The consequences of underdiagnosis or diagnostic delay must be considered in the context that nonradiographic axial SpA and AS are associated with a similar burden of disease ([3, 7]).
Prevalence estimates were similar between ASAS classification and rheumatologists' expert clinical diagnoses. However, the kappa statistic of 0.41 indicated low to moderate agreement in the classification of individual patients, suggesting that ASAS criteria and rheumatology experts diagnosed somewhat different patient populations. Seventy-six percent (390 of 514) of patients who fulfilled ASAS criteria were also diagnosed with axial SpA by rheumatologists, and 79% (390 of 491) of patients diagnosed by rheumatologists also met the ASAS criteria. Conversely, according to the ASAS criteria, 24% (124 of 514) of patients were undiagnosed and 21% (101 of 491) were misdiagnosed in clinical practice. This level of consistency is moderately lower than expected given the 82.9% sensitivity and 84.4% specificity reported for the ASAS criteria. Moreover, 101 of the 124 ASAS nonradiographic axial SpA patients who were not diagnosed by rheumatologists were instead diagnosed with AS, which also is included in the spectrum of axial SpA. These discrepancies underscore the importance of using a consistent set of criteria in epidemiology studies and clinical practice. The complexity of diagnosis in clinical practice was also underscored by the fact that 24 of 30 patients in our study diagnosed by a rheumatologist as having AS had no radiographic results in their chart. While it is possible that the radiographs, and the diagnosis, predated the patients' entry into the practice, it is also possible that physicians may have entered a diagnosis of AS for these patients to meet insurance guidelines for coverage of treatment with a tumor necrosis factor inhibitor.
There was wide variance in the number of “at-risk” patients in the rheumatology practices in this study, presumably due both to the fact that this number was an estimate, as well as to differences in the focus and the demographics of the practices queried. Indeed, patient demographics indicated significant bias in the current patient sample. In particular, 68% of the population in our sample was male, suggesting that men with symptoms of chronic low back pain are more frequently referred to rheumatologists than women. This may be due to a greater awareness of AS by general practitioners and a perception that AS is predominantly a disease of males (), which raises the possibility of underdiagnosis of axial SpA in women.
The current study exhibited considerable strength in its ability to identify patients with axial SpA, owing to the amount and detail of clinical information available for analysis. On the other hand, analyses depended upon a qualitative estimate of the percentage of at-risk patients in a given rheumatology practice. The prevalence estimate was calculated based on the mean estimate. Because there is substantial variance in this estimate between regions and among individual practitioners (Table 1), calculations may over- or underestimate the true prevalence of axial SpA within the target population. Diagnosis of axial SpA in our study relied on identification of elements of the ASAS criteria in patient medical records; thus, we were unable to capture disease characteristics that were not recorded, which may have led to underestimation of the prevalence of axial SpA.
Another limitation of our study is that it included only patients treated by rheumatologists and not those seen in primary care or orthopedic practices. Inclusion of these additional patients would potentially lead to a different prevalence estimate. Moreover, the focus of this study was on a specific at-risk population in the age range of 18–44 years, based on the age cutoff of 45 years in ASAS SpA criteria. Accounting for a broader age range may also result in a different prevalence estimate. Finally, the methodology did not allow an assessment of incidence, as case identification depended upon the presence of at least some symptoms of SpA required for diagnosis.
Despite these limitations, the results suggest that just under 1% of the US adult population is affected by axial SpA, in good agreement with estimates based on other methodologies and in different populations. Up to half of that population may not have radiographic evidence of axial SpA and, therefore, may be at risk for underdiagnosis. Increased awareness and use of the ASAS criteria, with emphasis on MRI and use of HLA–B27 testing, should lead to more accurate and timely diagnosis for these patients.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be submitted for publication. Dr. Rao had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Strand, Rao, Cifaldi, McGuire.
Acquisition of data. McGuire.
Analysis and interpretation of data. Strand, Rao, Shillington, Cifaldi, McGuire, Ruderman.
ROLE OF THE STUDY SPONSOR
The design, study conduct, and financial support for the study were provided by AbbVie. AbbVie participated in the interpretation of data, review, and approval of the manuscript. The study design was conceptualized in collaboration with non-AbbVie authors. Submission for publication was contingent on AbbVie approval. Medical writing and editorial assistance were provided by Eric Bertelsen, PhD, of Arbor Communications, Inc. (Ann Arbor, Michigan) and was funded by AbbVie.
Author Shillington is an employee of EPI-Q, Inc. Author McGuire is an employee of Medical Data Analytics.