Patients

Patients were recruited for the cross-sectional component of the PSOAS cohort from Cedars-Sinai Medical Center, Los Angeles, California; Stanford University, Palo Alto, California; the National Institutes of Health, Bethesda, Maryland; the University of Texas Medical School at Houston, Houston, Texas; and the University of California, San Francisco, California. Inclusion criteria were: meeting the modified New York criteria for AS¹⁰; having AS for ⩾20 years, dated from the time of onset of persistent inflammatory low back pain or restricted spinal motion; and being willing to participate in a genetic study of AS. Patients with a history of inflammatory bowel disease or psoriasis were not excluded. A total of 302 men and 100 women with AS met these inclusion criteria and were studied.

Clinical features

All participants completed extensive questionnaires about their medical and personal histories, and at each site one of the study rheumatologists performed a clinical evaluation on each participant. Clinical data included age, sex, ethnicity (white vs other), educational level, marital status, smoking status (current, former or non-smoker), pack years of smoking and comorbid medical conditions. Participants also reported their level of recreational physical activity in their teens and twenties, relative to their same-sex peers. An occupational physical activity score for each patient was computed, based on the level of physical activity (little, moderate, much) in each job they had worked at and the number of years spent in each job.¹¹ Additional data included duration of AS, age at onset of AS symptoms, family history of AS in a first-degree relative, history of total hip arthroplasty, history of iritis and human leucocyte antigen (HLA)-B27 status. Past and current medication history included the use of methotrexate, sulphasalazine, intra-articular steroids, prednisone, infliximab and etanercept. A history of peripheral arthritis was not collected because such information would not be verifiable or valid.

Functional status

To measure functional disability, participants were asked to complete both the Bath Ankylosing Spondylitis Functional Index (BASFI)¹² and the Health Assessment Questionnaire for the Spondyloarthropathies (HAQ-S).¹³ Both the BASFI and the HAQ-S have been demonstrated to have good reliability and construct validity.^12–,¹⁷ The BASFI is a 10-item scale on which respondents rate the degree of difficulty they have in performing certain tasks, using visual analogue scales from 0 (easy) to 100 (impossible). The mean of the 10 responses is the BASFI score. The HAQ-S is a 25-item scale on which respondents rate the degree of difficulty they have performing tasks in 10 functional areas, with responses ranging from 0 (no difficulty) to 3 (unable to do). The average of the highest score in each of the 10 function categories is the HAQ-S score.

Radiographs

Radiographs of the pelvis (anterior–posterior), lumbar spine (anterior–posterior and lateral) and cervical spine (lateral) were obtained from each participant and were scored using the Bath Ankylosing Spondylitis Radiographic Index Spine (BASRI-spine) score¹⁸ by a single musculoskeletal radiologist (TJL). The range of scores for the BASRI-spine is 2–12. Inter-reader and intra-reader reliability testing for BASRI scoring was performed on a subset of radiographs, with adjudication between all investigators for inclusion and exclusion of subjects. The BASRI was the best characterised and most standardised instrument available at the time this study started; the statistical adjustments below were made for its limitations.

Statistical analysis

The unadjusted male–female comparisons were done using t tests for continuous measures that were normally distributed Wilcoxon’s rank sum tests for continuous measures that were not normally distributed, and χ² tests for categorical measures. For the analysis of radiographic data, the BASRI-spine score was considered an ordered categorical variable. As some scores had few patients and the BASRI-spine scores are known to be subject to a ceiling effect,¹⁹ the BASRI-spine scores were grouped into six ordered categories: 2–4, 4.5–6.5, 7–8.5, 9–10, 11–11.5 and 12. Odds ratios (ORs) for association of gender with the BASRI-spine category were calculated by ordinal logistic regression. These ORs represent the likelihood that men were more likely than women to be in a higher BASRI-spine category (1–6) than in the same or in a lower category. Multivariate modelling of the association of gender with BASRI-spine category included either age or duration of AS in separate models, because age and duration were strongly correlated in this cohort. Additional variables in the age-adjusted models were weighted occupational physical activity score, family history and pack years of smoking, while additional variables in the duration-adjusted models were age of onset, weighted occupational physical activity score, family history and pack years of smoking. The associations of BASRI-spine scores with functional measures were performed using Spearman’s correlation coefficients.

Literature review

For the literature review, a MEDLINE search from 1966 to 2005 was performed, including the terms “ankylosing spondylitis”, “men”, “women” and “gender”. The studies published before 1966 that are included in the review were identified in the bibliographies of the selected articles. To be included, the study had to report results comparing clinical and/or radiographic features between men and women. Comparisons of specific clinical and radiographic features were examined separately, and the results were summarised.

Clinical features and outcomes

In total, 302 men and 100 women with AS were studied in this cohort. Women had a significantly earlier age at disease onset, but the disease duration at the time of study entry was nearly equal between the genders (table 1⇓). Women more frequently reported family histories of AS in first-degree relatives than men. Men were more likely to be current or former smokers. There were no statistically significant differences in educational level, weighted occupational physical activity, marriage status, ethnicities or HLA-B27 positivity (table 1⇓). The frequency of total hip arthroplasty between the two groups was not statistically different (9.6% of men vs 13% of women, p = 0.34). The frequency of iritis was not statistically different (42% of men vs 44% of women, p = 0.74). In terms of drug usage, women were significantly more likely to have been treated with intra-articular steroids, sulphasalazine in the past, prednisone in the past and currently, and infliximab in the past. More women had current or past use of methotrexate, but the values did not reach statistical significance (table 2⇓).

Radiographic results

Radiographic severity was worse among men. The unadjusted median BASRI-spine score for men was 10 vs 6.5 for women (p<0.001 by Wilcoxon’s rank sum test.) Adjusting for age, men were significantly more likely than women to be in a higher BASRI-spine category than in the same or a lower category, with an OR of 3.86 (95% CI 2.53 to 5.90, p<0.001). After adjusting for duration of disease instead of age, the OR was still significantly high at 4.10 (95% CI 2.69 to 6.26, p<0.001). In multivariate adjusted analyses, men were still more likely to be in a higher BASRI-spine category than women, with an OR of 3.81 (95% CI 2.45 to 5.93, p<0.001) after adjusting for age and with an OR of 3.83 (95% CI 2.46 to 5.96, p<0.001) after adjusting for duration of disease (table 3⇓).

Functional outcomes

Functional limitations as measured by the BASFI were not statistically different between men and women in unadjusted analysis (table 4⇓). Results comparing HAQ-S scores were comparable and also not statistically different (data not shown). Using analyses adjusted for age, duration of AS or the above multivariate model, the results were still not statistically different for both the BASFI and HAQ-S scores (data not shown). However, including the BASRI-spine scores in the multivariate model revealed that women had statistically significant worse functional scores as measured by both BASFI (table 4⇓) and HAQ-S (data not shown), thus indicating that, at the same level of radiographic damage, women had worse functional scores. The results of an additional analysis, which was performed to evaluate the associations of BASRI-spine scores with functional measures in men and women, using Spearman’s correlation coefficients, revealed that radiographic damage has a similar correlation with functional impairment in men and women (table 5⇓). Thus, although radiographic damage correlates with functional impairment in men and women, women report more functional limitations at the same level of radiographic damage.

Literature search results

In all, 14 studies comparing clinical features and 13 studies comparing radiographic features of men and women with AS were identified.^{1,2,14,18,20–},³²

Clinically, the most consistent trends that emerged were that women had more neck and peripheral joint pain (table 6⇓). Several studies examined root joints (hips and shoulders) separately, whereas others considered them all together. The remarks on overall course tended to describe women as having either the same or milder course, although these conclusions were not based on objective measurements.

In terms of radiographic differences (table 7⇓), the features that stood out from the literature review are that women tend to have less thoracic and lumbar spinal radiographic severity, and in the more recent studies, lower BASRI scores. The studies that looked specifically at findings on cervical spine and peripheral joints (including hip and shoulder), however, did not report evidence of women having more radiographic damage in these areas. Thus, the literature review data of clinical differences indicating increased pain in the cervical spine and peripheral joints did not necessarily correlate with radiographic changes.

The studies done before 1976 used no criteria for patient inclusion,^20–,²³ and of the remaining studies, only the study by Kidd et al³⁰ used the modified New York criteria. Furthermore, the earlier of these studies did not use statistical methods to quantify their observations, and their conclusions were only qualitative.^20–,²⁴ In later studies, statistical comparisons were made, but multivariate adjustments were not in use until the late 1980s.^{2,14,18,30–},³² Frequently, the studied cohorts had wide ranges of disease durations and age ranges among the patients. Finally, most of these studies were carried out using retrospective chart review, except for two case–control studies^29,30 and one questionnaire study.²