Prevalence of axial spondylarthritis in the United States: Estimates from a cross-sectional survey
The US national prevalence of spondylarthritis (SpA) was estimated for 2 published sets of classification criteria: the Amor criteria and the European Spondylarthropathy Study Group (ESSG) criteria. These 2 SpA criteria sets have been the most widely utilized in previous population-based studies of SpA.
The US SpA prevalence estimates were based on a representative sample of 5,013 US adults ages 20–69 years who were examined in the US National Health and Nutrition Examination Survey (NHANES) 2009–2010.
The overall age-adjusted prevalence of definite and probable SpA by the Amor criteria was 0.9% (95% confidence interval [95% CI] 0.7–1.1%), corresponding to an estimated 1.7 million persons (95% CI 1.4–2.1 million persons). The age-adjusted prevalence of SpA by the ESSG criteria was 1.4% (95% CI 1.0–1.9%), corresponding to an estimated 2.7 million persons (95% CI 1.9–3.7 million persons). There were no statistically significant sex differences in SpA prevalence. The SpA prevalence among non-Hispanic white persons was 1.0% (95% CI 0.7–1.5%) by the Amor criteria and 1.5% (95% CI 1.0–2.3%) by the ESSG criteria. SpA prevalence could not be reliably estimated in other race/ethnicity subgroups due to sample size imitations.
The SpA prevalence estimates are in the range of SpA prevalence estimates reported elsewhere in population-based surveys and it is likely that SpA may affect up to 1% of US adults, a prevalence similar to that reported for rheumatoid arthritis. The current US SpA prevalence estimates may be lower than the true value because the NHANES 2009–2010 data collection did not capture a complete set of the elements specified in the 2 SpA criteria sets.
Spondylarthritis (SpA) is a common form of inflammatory arthritis among adults (1). The classically described form of SpA included only the syndrome of ankylosing spondylitis (AS), but currently SpA is seen as a family of syndromes that includes not only AS, but also spondylitis associated with inflammatory bowel disease and psoriasis, and reactive SpA due to enteric or urogenital infections. An undifferentiated form of SpA is also currently described in the literature. The elements of the SpA case definition are a history of chronic axial pain/stiffness or inflammatory back pain, spinal mobility measurement, imaging studies of the spine and pelvis, and serologic testing, especially for the HLA–B27 biomarker and C-reactive protein. The major classification criteria for SpA in epidemiologic studies have been the Amor criteria (2) and the European Spondylarthropathy Study Group (ESSG) criteria (3).
The recent development of new drug treatments for SpA with increased efficacy for the disease has refocused specific attention on the clinical aspects of SpA and its clinical and epidemiologic case definition. Recently, new classification criteria have been proposed for SpA (4). There are a number of studies that provide population-based prevalence estimates for SpA in the published literature (5); however, no national SpA prevalence estimate is currently available in the US. The primary goal of this study was to provide contemporary US national prevalence estimates for SpA using the Amor and ESSG classification criteria. These current population-based SpA prevalence estimates are based on the data collected by the US National Health and Nutrition Examination Survey (NHANES) for its 2009–2010 survey cycle.
Significance & Innovations
The current study findings for the first time provide US national prevalence estimates for spondylarthritis (SpA).
Using published case definition criteria, the US SpA prevalence estimates were determined using a representative sample of 5,013 adults 20–69 years of age examined in the 2009–2010 National Health and Nutrition Examination Survey.
Results show that SpA may affect up to 1% of US adults, a prevalence similar to that reported for rheumatoid arthritis. No significant sex differences were noted.
MATERIALS AND METHODS
The NHANES is a cross-sectional survey monitoring the health and nutritional status of the civilian, noninstitutionalized US population (6). Data were collected via household interviews and direct, standardized physical examinations and biologic specimen collection in specially equipped mobile examination centers. The NHANES samples were selected through a complex multistage probability design. The survey design involved the selection of primary sampling units (PSUs; counties), census tract segments within PSUs, households within census tract segments, and persons within households, with known probability of selection at each stage. Each annual sample is nationally representative; however, NHANES data are publicly released for 2-year survey periods to protect confidentiality and increase statistical reliability. The NHANES 2009–2010 oversampled major US demographic subgroups, including Hispanic persons and non-Hispanic black persons as well as persons with low income. The NHANES 2009–2010 researchers operated with approval from the National Center for Health Statistics Ethics Review Board.
In the NHANES 2009–2010, 6,684 persons in the study target age range of 20–69 years were screened for participation in the survey and 5,106 were interviewed. There were 3 eligible interviews where the arthritis questionnaire (ARQ) was not performed (item nonresponse), leaving a total of 5,103 complete arthritis interview records available for analysis for an overall survey response rate of 76.4%. Demographic data were collected in the NHANES household interview and included the respondent's age and self-designated race/ethnicity. The participant's sex was as observed by the NHANES interviewers. The demographic distribution of the study sample by age, sex, and race/ethnicity is provided in Table 1.
Table 1. Study sample for the National Health and Nutrition Examination Survey 2009–2010 US arthritis survey
|Total study sample||5,103 (100.0)|
|Age groups|| |
| 20–35 years||1,649 (32.3)|
| 36–49 years||1,539 (30.2)|
| 50–69 years||1,915 (37.5)|
| Male||2,472 (48.4)|
| Female||2,631 (51.6)|
|Race/ethnicity subgroups|| |
| Mexican American||1,024 (20.1)|
| Other Hispanic||576 (11.3)|
| Non-Hispanic white||2,244 (44.0)|
| Non-Hispanic black||963 (18.9)|
| All others||296 (5.8)|
SpA criteria data collection.
The NHANES 2009–2010 arthritis questionnaire and examination data collection were developed to provide population-based prevalence estimates for the 2 principal published SpA classification criteria sets that were current at the time the instrument was being developed. These are the Amor (2) and the ESSG criteria (3). An additional set of SpA classification criteria, the Assessment of SpondyloArthritis international Society (ASAS) criteria (4), was published after the study had been fielded and was not included in this study. The NHANES 2009–2010 arthritis-related questionnaire and examination data collection protocol are described separately (7) and copies of the household interview arthritis questionnaire and pain diagrams are available at the NHANES web site (8, 9).
Table 2 shows a detailed listing of the specific data elements that compose the Amor and ESSG SpA classification criteria. Table 2 also indicates which data elements were collected in the NHANES 2009–2010. The NHANES 2009–2010 had limitations in terms of the content that could be included in the survey, and a number of items pertinent to the Amor and ESSG SpA classification criteria could not be fielded. The NHANES data collection fully supported the ESSG inflammatory spinal pain case definition and the Amor back pain/stiffness case definition (we defined a history of morning spinal stiffness for the Amor criteria as stiffness in the upper, middle, or lower back lasting >30 minutes); however, radiologic or spinal imaging studies could not be fielded in the NHANES 2009–2010. Also, there was no data collection on peripheral arthritis, reactive arthritis, and family history of SpA-related disorders, balanitis, or dactylitis. HLA–B27 data were collected only for the 2009 survey year, so they could not be used in the present analysis. Overall, there was data collected for 6 of the 11 Amor criteria or 11 of the 18 possible points in the Amor classification system (3 items were fully supported by the NHANES data collection and 3 were partially supported). An Amor score of 5 points denoted probable SpA and a score of 6 points denoted definite SpA. Amor SpA prevalence estimates in the data tables were based on combining both definite and probable cases because of the small sample size for the definite Amor SpA cases.
Table 2. Amor and ESSG case definitions for SpA*
|Amor SpA criteria elements†|| || |
| Nocturnal spinal pain or morning stiffness||1||Yes|
| Buttock/alternating buttock pain||1 or 2‡||Yes|
| Heel pain/enthesiopathy||2||Yes|
| Acute diarrhea at or prior to SpA onset||1||No|
| Radiologic sacroiliitis||3||No|
| HLA–B27 positive or family/genetic background||2||No|
| Psoriasis, balanitis, inflammatory bowel disease||2||Yes§|
| Positive response to NSAIDs||2||Yes|
| Asymmetric oligoarthritis||2||No|
|ESSG SpA criteria elements|| || |
| Inflammatory spinal pain|| ||Yes|
| Alternating buttock pain|| ||Yes|
| Enthesopathy|| ||Yes|
| Urethritis, cervicitis, diarrhea|| ||No|
| Radiologic sacroiliitis|| ||No|
| Positive family history|| ||No|
| Psoriasis|| ||Yes|
| Inflammatory bowel disease|| ||Yes|
The questionnaire variables used in the Amor and ESSG case definitions included a self-report of physician-diagnosed psoriasis, a history of physician-diagnosed iritis plus confirmation of a characteristic symptom pattern and treatment with prescription eye medication, and a history of physician-diagnosed inflammatory bowel disease with colonoscopy. A history of heel pain was classified using a heel pain diagram with areas marked for the pain specific to plantar fasciitis and/or Achilles tendonitis (ARQ4) (9). A history of localized heel or posterior calcaneal area pain in 1 of these 2 specific areas lasting ≥2 weeks was classified as plantar fasciitis or Achilles tendonitis.
The ESSG classification system defines SpA as a history of inflammatory spinal pain by ESSG criteria plus the presence of 1 additional ESSG SpA criteria element. The NHANES 2009–2010 data collection fully supported data collection required to define ESSG inflammatory spinal pain and captured 4 of the 7 additional ESSG SpA criteria elements. The SpA prevalence estimates were therefore based on subsets of some of the most commonly occurring items that compose the Amor and ESSG SpA criteria.
Interview sample weights were used to account for differential probabilities of selection and the complex NHANES sample design, and to obtain SpA prevalence estimates and their SEs that were representative of the noninstitutionalized US population ages 20–69 years. The sample weights accounted for the unequal selection probabilities of subgroups and adjusted for nonresponse and noncoverage. Data set assembly and statistical analysis were performed using SAS (release 9.2, SAS Institute) and SUDAAN (release 10.0, Research Triangle Institute). Ninety-five percent confidence intervals (95% CIs) were constructed using the logit transformation because the SpA prevalence estimates were low (<10%). Age adjustment was carried out by the direct method using the year 2010 Census Bureau projections for the US civilian, noninstitutionalized population. SEs were estimated by Taylor series linearization. The equality of the prevalence of SpA by selected demographic variables was tested (univariately) at the α = 0.05 level using a Student's t statistic with the appropriate df. The NHANES analytical guidelines were used to set the criteria for minimum acceptable numbers of cases and df in subgroups, adequate group sample sizes (based on design effect and the specified proportion), and relative SEs (RSEs) to assess statistical stability of computed estimates. Estimates with an RSE ≥30% and sample sizes or df less than recommended are designated in Table 3 as potentially unreliable and should be interpreted with caution (10).
Table 3. Prevalence of SpA in US adults 20–69 years of age: the NHANES 2009–2010*
|Amor SpA|| || || || || |
| Overall (definite and probable)†||5,103||43||0.9||0.1||0.7–1.1|
| Age group|| || || || || |
| 20–49 years||3,188||17||0.5‡||0.2||0.2–1.2|
| 50–69 years||1,915||26||1.5||0.3||1.0–2.3|
| Sex†|| || || || || |
| Race/ethnicity group†|| || || || || |
| Mexican American||1,024||6||0.6‡||0.3||0.2–1.7|
| Non-Hispanic white||2,244||24||1.0||0.2||0.7–1.5|
| Non-Hispanic black||963||9||0.9‡||0.3||0.4–1.8|
|ESSG SpA|| || || || || |
| Age group|| || || || || |
| 20–49 years||3,188||49||1.5||0.2||1.1–2.0|
| 50–69 years||1,915||21||1.3||0.4||0.7–2.5|
| Sex†|| || || || || |
| Race/ethnicity group†|| || || || || |
| Mexican American||1,024||15||1.5‡||0.5||0.7–3.0|
| Non-Hispanic white||2,244||38||1.5||0.3||1.0–2.3|
| Non-Hispanic black||963||9||0.9‡||0.3||0.4–1.8|
Table 3 shows the 2009–2010 prevalence estimates for SpA in the US overall and according to the major US demographic subgroups in the NHANES sample design. By the Amor criteria, the weighted, age-adjusted US national prevalence estimate for combined definite and probable SpA for US adults 20–69 years of age was 0.9% (95% CI 0.7–1.1%). This corresponded to an estimated 1.7 million persons (95% CI 1.4–2.1 million persons) in the 20–69 years age group. The crude prevalence of definite SpA by the Amor criteria was very low (n = 5 cases, crude % = 0.1), and a statistically reliable prevalence estimate could not be made (data not shown). The available data set sample sizes in some instances did not permit reliable Amor SpA estimates to be produced for sex and race/ethnicity. For the Amor SpA criteria, reliable prevalence estimates could be made for the age subgroup 50–69 years, for women, and for non-Hispanic white persons. These estimates were 1.5% (95% CI 1.0–2.3%), 1.3% (95% CI 0.9–1.8%), and 1.0% (95% CI 0.7–1.5%), respectively.
For the ESSG criteria, the overall US national age-adjusted SpA prevalence estimate was 1.4% (95% CI 1.0–1.9%) in the 20–69 years age group. This corresponds to an estimated 2.7 million persons (95% CI 1.9–3.7 million persons). The ESSG SpA prevalence estimate for men was 1.1% (95% CI 0.6–2.0%) and for women was 1.7% (95% CI 1.2–2.5%). There was not a significant prevalence difference either by age groups (t = 0.51, P = 0.62) or between men and women (t = 1.39, P = 0.18). A detailed analysis of SpA prevalence by race/ethnicity could not be reliably performed for the ESSG SpA criteria due to sample size limitations. ESSG SpA prevalence could be reliably estimated only for the non-Hispanic white person subgroup (1.5% [95% CI 1.0–2.3%]).
The objective of this study was to provide current nationally representative US prevalence estimates for SpA using 2 widely published epidemiologic SpA classification criteria, the Amor and the ESSG criteria. The NHANES 2009–2010 data were used for this purpose. The overall age-adjusted national prevalence of SpA among US adults 20–69 years of age was 0.9% by the Amor criteria and 1.4% by the ESSG criteria. There was little evidence in the data for an SpA prevalence difference by age or by sex using the ESSG SpA case definition. For the Amor criteria, this latter assessment was precluded by the available study sample size. Sample size considerations limited the possibility of any analysis by race/ethnicity for both sets of the study criteria. Both the overall Amor and ESSG SpA prevalence estimates are in the approximate range of the previously published US National Arthritis Data Workgroup estimates for SpA prevalence (0.35–1.30% of the US population). This prevalence range was calculated by summing published prevalence estimates for the major component subtypes of SpA (1). Studies in the literature regarding other international prevalence estimates for SpA have been recently reviewed (5).
The methodologic strengths of the NHANES are its nationally representative sample of both men and women, adequate response rates, oversampling of older persons and ethnic subgroups, and standardized, quality-controlled protocol for data collection. It is therefore a useful venue for US population-based prevalence estimation for chronic medical conditions such as SpA; however, there are important limitations to the current study. Chief among these is the fact that detailed followup clinical examinations were not performed on the individuals in the survey who were classified as having SpA. Therefore, the validity of the SpA prevalence estimates reported here rests on the prior validation of the 2 sets of SpA classification criteria previously published in the literature.
A further limitation of this study is the available study sample size in relation to the absolute magnitude of the SpA prevalence estimates. A 2-year NHANES survey cycle is designed to permit statistically reliable estimation (RSE <30%) for disorders with a ≥10% prevalence in each of the major demographic subgroups in the NHANES sample design. For disorders with much lower prevalence (i.e., the approximate 1% prevalence seen here for SpA), a 2-year NHANES survey sample may permit a statistically reliable overall US prevalence estimate to be calculated as well as overall prevalence estimates for some major demographic subgroups such as sex. However, the sample sizes in a typical 2-year data set are insufficient to permit estimation of more detailed demographic trends, for example, an analysis of age trends by sex. Also, the available sample sizes often preclude detailed analytic studies.
Another significant limitation is the practical constraint that the NHANES 2009–2010 was not able to collect data for all of the data items required for each of the 2 SpA criteria (Table 2). For example, pertinent radiologic imaging studies were not performed and HLA–B27 data were collected only for a single survey year (2009) and could not be included in the present analysis. However, it should be noted that the lack of radiologic imaging studies and/or HLA–B27 data is not a study limitation with respect to ESSG SpA prevalence estimation as it might be for the Amor classification system because radiologic imaging and HLA–B27 are not included as criteria in the ESSG system. There were no data collected on peripheral arthritis or family history of AS or other SpA-related diseases. This lack of item availability appeared to impact the Amor SpA prevalence estimates more than the ESSG estimates. The current SpA prevalence estimates may therefore represent a lower bound for the true prevalence of SpA in the US. Also, because the NHANES 2009–2010 arthritis survey was fielded prior to publication of the new ASAS classification criteria for SpA (4), prevalence estimates for these newer SpA criteria could not be calculated.
Finally, the NHANES 2009–2010 arthritis questionnaire was created for a population-based survey. This is a different mode of administration than medical interviews conducted in a clinical setting. The NHANES questionnaire was administered by professional field interviewers. The survey questions themselves were designed to be clear and readily understandable to persons with a low literacy level. Also, the NHANES interviewers did not have medical backgrounds and therefore were unable to carry out any detailed probing, which is customarily done during clinical interviews. While population-based interviewing is not the same as clinical interviewing in the context of patient care, both can be employed in research studies and each has its relative strengths and weaknesses. The approach we used is standard for population-based research. The advantage of this approach is that specific questions were asked precisely as scripted and were administered to each participant in a standardized way by professional interviewers. The questions used in the arthritis questionnaire underwent formal cognitive testing to help assure that the questions were well constructed, unambiguous, and well understood by persons of all education levels.
The current study findings provide for the first time US national prevalence estimates for SpA. These estimates should be interpreted in light of the strengths and limitations of the NHANES data collection protocol. Because the NHANES data collection was incomplete, the US SpA prevalence estimates may be a lower bound for the true prevalence of SpA among US adults. Nevertheless, these US SpA prevalence estimates will be useful for future public health planning.
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 published. Dr. Weisman 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. Reveille, Witter, Weisman.
Acquisition of data. Reveille.
Analysis and interpretation of data. Reveille, Witter, Weisman.
The efforts of the Spondyloarthritis Association of America and the Spondyloarthritis Research and Treatment Network to help support and field the 2009 NHANES study are gratefully acknowledged. Also especially acknowledged is the generous voluntary participation of the US residents who have given their personal time to make the NHANES possible. We also gratefully acknowledge the editorial support and assistance from Alisa Wilson, PhD, Charles F. Dillon, MD, PhD, and Rosemarie Hirsch, MD.