To assess the differences in disease activity as measured by the Bath Ankylosing Spondylitis Disease Activity Index (BASDAI) between patients with and without peripheral arthritis/enthesitis. To investigate whether scores on the BASDAI change by omitting the 2 questions on peripheral disease.
Disease activity was assessed on a 10-cm visual analog scale and by BASDAI. Alternative BASDAIs were constructed by omitting the peripheral joints question and/or the enthesitis question. Correlations between the alternative BASDAIs and other measures of disease activity were calculated. Generalized estimating equations (GEE) were used to assess whether having peripheral arthritis influenced BASDAI and alternative BASDAI scores, and to assess whether peripheral arthritis influenced the score of the individual questions of the BASDAI.
At baseline, the BASDAI was calculated in 214 patients. In patients with peripheral arthritis (n = 56), the mean (SD) BASDAI score was 4.4 (2.3) as compared with 3.1 (1.9) (P < 0.0001) in the patients without peripheral arthritis (n = 158). The relationship between arthritis and the BASDAI score appeared to be truly longitudinal (GEE regression coefficient β = 0.64; 95% confidence interval 0.28–1.00). Peripheral arthritis was significantly longitudinally associated with all separate item scores of the BASDAI. Omitting the peripheral joints and/or enthesitis question from the BASDAI questionnaire only partially explained the difference in BASDAI score between the 2 groups.
Disease activity measured by the BASDAI is higher in patients with concomitant peripheral disease compared with patients with disease restricted to the axial skeleton. The increased BASDAI score in patients with peripheral arthritis is partially explained by increased overall disease activity as well as by a disproportionate contribution of the peripheral joints question to the overall score.
Ankylosing spondylitis (AS) is a chronic rheumatic condition characterized by a wide spectrum of symptoms (1). These symptoms can be limited to the axial skeleton (spinal disease) or can be more extensive, whereby AS can present with peripheral arthritis, enthesitis, dactylitis, and uveitis (extraspinal disease). For measuring disease activity in AS, no gold standard currently exists. Objective measures such as C-reactive protein (CRP) and Westergren erythrocyte sedimentation rate (ESR) poorly correlate with clinical disease activity (2). The Assessment in Ankylosing Spondylitis (ASAS) working group selected in 1998 a core set of single instruments to assess disease activity and disease severity in AS (3). Combined instruments are not included in the core set. The progress reports stated that the value of combined instruments, such as the Bath Ankylosing Spondylitis Metrology Index (4, 5) and Bath Ankylosing Spondylitis Disease Activity Index (BASDAI) (6), should be investigated because the combined instruments might not have the same validity in all types of disease, e.g., in spinal or extraspinal disease.
The BASDAI (6) is a widely used subjective measure to assess disease activity that proved to be valid, reproducible, and responsive to change (7, 8). The BASDAI consists of 6 questions, of which one deals with complaints in peripheral joints (question 3) and one with enthesitis (question 4). The peripheral joints question is worded, “How would you describe the overall level of pain/swelling in joints other than neck, back or hips you have had.” The enthesitis question is worded, “How would you describe the overall level of discomfort you have had from any areas tender to touch or pressure.” Both questions relate to the past week.
We hypothesized that patients with peripheral arthritis would show a higher BASDAI score compared with patients with AS restricted to the axial skeleton, and that this higher score might be attributed to a higher score on the BASDAI questions about peripheral joints and entheses. The first aim of the present study was to investigate whether a difference exists in degree of disease activity as measured by the BASDAI between patients with and without peripheral arthritis or enthesitis; in other words, to assess whether arthritis is a determinant of disease activity. The second goal was to investigate whether this difference would change by omitting the questions on arthritis and/or enthesitis. Finally, the correlation was assessed between the BASDAI, including and excluding the questions on arthritis and enthesitis, and other measurements of disease activity in patients both with and without peripheral arthritis or enthesitis.
PATIENTS AND METHODS
In our study, we used data from the Outcome in Ankylosing Spondylitis International Study (OASIS) cohort, an international, longitudinal, observational study on outcome of AS with followup visits according to a fixed protocol. Data from this cohort have been reported previously (2, 9). Consecutive outpatients with an established diagnosis of AS according to the modified New York criteria (10) were included in 1996. We used data from visits at baseline (T0), 6 months (T6), 12 months (T12), 18 months (T18), and 24 months (T24). All patients completed a number of questionnaires and underwent a clinical examination.
Measures of disease activity.
The core set of variables selected by the ASAS working group (3) were all assessed.
The BASDAI consists of 6 questions: the first on fatigue; the second on pain of the spine and hips; the third on pain or swelling of the peripheral joints; the fourth on enthesitis; and the last 2 on severity and duration of morning stiffness, respectively (6). The questions are answered on a 10-cm visual analog scale (VAS). The mean of the 2 scores regarding morning stiffness counts as 1 variable. The final score is defined by calculating the mean of the 5 items. Scores range from 0 (best) to 10 (worst). In case of missing values for questions 1–4, the BASDAI was not calculated and reported as missing. In case of 1 missing value of question 5 or 6, the remaining question counted as the mean of questions 5 and 6. We calculated alternative BASDAIs in 3 different ways. First, by omitting question 3, the peripheral joints question (BASDAI-without-joints); second by omitting question 4, the enthesitis question (BASDAI-without-entheses); and third by omitting both the peripheral joints and the enthesitis question (BASDAI-without-joints-entheses). The mean of the remaining items was calculated to obtain the value of these alternative BASDAIs.
Disease activity was also assessed by a single-item 10-cm VAS concerning the degree of disease activity during last week (patient) and on the day of physical examination (physician).
Function was measured by the Bath Ankylosing Spondylitis Functional Index (11). During clinical examination, the physician performed a painful (53 joints) and swollen (44 joints) joint count. Peripheral arthritis was defined as the presence of at least 1 swollen joint. Peripheral arthritis data on followup visits at T6, T18, and T24 were available only for patients in The Netherlands (n = 137). Pain was recorded in 2 ways: pain of the spine due to AS last week, on average, using a 10-cm VAS; and night pain last night using a 4-point Likert scale. Patients' global wellbeing was measured on a 10-cm VAS. ESR was assessed using the Westergren method (normal range male 0–7 mm/hour, female 0–12 mm/hour) and CRP by the turbidimetric method (normal range 2–9 mg/liter). The lowest detection limit for CRP was 2 mg/liter and patients with undetectable levels were assigned 0.
Each study visit was analyzed independently. Independent t-tests were used to compare patients with (A+ group) and without (A− group) peripheral arthritis with regard to scores on the BASDAI and alternative BASDAIs, and on the separate questions of the BASDAI.
Correlation between the original and alternative BASDAIs and disease activity by patients was calculated using Pearson's correlation coefficient. Correlation between the original and alternative BASDAIs and disease activity by physician, ESR, and CRP were calculated using Spearman's correlation coefficient because of skewness in data distribution.
In general, longitudinal data sets are characterized by observations with high variability between patients and rather low variability within patients (i.e., the BASDAI score at T0 is highly correlated with the BASDAI score at T12). Because of the high within-patient correlation, longitudinal relationships cannot be analyzed with ordinary regression methods. Generalized estimating equations (GEE) is a regression technique for studying intervariable relationships in longitudinal studies; this technique takes into account time, as well as time-independent and time-dependent covariates (12). The advantage of using GEE over other methods is that GEE uses all available longitudinal data, allows unequal numbers of repeated measurements and unequal time intervals, and does not require multivariate normality of the outcome variable. GEE does require an a priori “working” correlation structure to adjust for the within-subject correlation operating in repeated-measurement designs. A correlation structure must be chosen on the basis of the actual data set. Here we chose an exchangeable correlation structure.
Autoregressive GEE was used to assess the effect of having peripheral arthritis on the total score of the BASDAI and alternative BASDAIs, as well as to assess the effect of having peripheral arthritis on the individual question scores of the BASDAI. In autoregressive analysis, each value of the dependent variable at Tx is adjusted for the value of the dependent variable at Tx−1. The effect of having peripheral arthritis on the different scores is given as a regression coefficient (β). Betas are standardized by dividing the beta by the standard error (β/SE), which allows comparison of different scales and a different standard error.
At baseline, 217 patients were included. Table 1 presents characteristics of the patients. The median score, interquartile range, and range of the individual questions of the BASDAI at T0 are presented in Figure 1. All individual questions of the BASDAI scored significantly lower in the A− group as compared with the A+ group. This effect was observed at all time points. Particularly, the individual BASDAI question scores on peripheral joints, entheses, and stiffness were sensitive to having peripheral arthritis.
Table 1. Baseline characteristics and scores on ASAS core set measures*
Study population, n = 217
ASAS = Assessment in Ankylosing Spondylitis working group; IBD = inflammatory bowel disease; BASFI = Bath Ankylosing Spondylitis Functional Index; VAS = visual analog scale; IQR = interquartile range; ESR = erythrocyte sedimentation rate; CRP = c-reactive protein.
Age, years, mean (SD)
Duration of complaints, years, mean (SD)
Time since diagnosis, years, mean (SD)
HLA B27, present/absent/no data
History of IBD, present/absent/no data
History of uveitis, present/absent/no data
History of psoriasis, present/absent/no data
BASFI, 0–10, mean (SD)
VAS pain of the spine, 0–10 cm, mean (SD)
Night pain, 0–4 Likert, median (IQR)
Chest expansion, cm, mean (SD)
10-cm Schober, cm, mean (SD)
Occiput-to-wall distance, cm, median (IQR)
VAS patient global, 0–10 cm, mean (SD)
Peripheral arthritis, present/absent†
ESR, mm/hour, median (IQR)
CRP, mg/liter, median (IQR)
VAS physician on disease activity, 0–10 cm, median (IQR)
VAS patient on disease activity, 0–10 cm, mean (SD)
Duration of morning stiffness, minutes, mean (SD)
Table 2 shows the scores on the BASDAI, the alternative BASDAIs, and the patients' VAS concerning the degree of disease activity; it also presents the median with interquartile ranges on ESR, CRP, and physicians' single-item VAS concerning the degree of disease activity at baseline. All scores were significantly higher in the A+ group compared with the A− group. By omitting the questions on joints and entheses, the scores in the A− group increased slightly (BASDAI-without-joints or BASDAI-without-entheses) to moderately (BASDAI-without-joints-entheses) and only slightly in the A+ group if both questions were omitted. Consequently, the difference in scores between the A+ group and the A− group decreased somewhat, particularly if both questions were omitted.
Table 2. Scores on the BASDAI and alternative BASDAIs, ESR, CRP, patients' and physicians' disease activity in patients with (A+ group) and without (A− group) peripheral arthritis at baseline*
A+ group n = 56
A− group n = 158
Difference between A+ and A− group
Data reported as mean (SD) unless otherwise noted. All scores (except ESR and CRP) range from 0 to 10. BASDAI = Bath Ankylosing Spondylitis Disease Activity Index; ESR = erythrocyte sedimentation rate; CRP = C-reactive protein; IQR = interquartile range; VAS = visual analog scale.
ESR, mm/hour, median (IQR)
CRP, mg/liter, median (IQR)
Patient VAS on disease activity, cm
Physician VAS on disease activity, cm, median (IQR)
Table 3 presents the correlation between the various BASDAIs and ESR, CRP, and the patients and the physicians' single-item VAS concerning the degree of disease activity at T0. The correlations between ESR and CRP and the various BASDAIs were low in both groups. The correlations between the patients' disease activity and all BASDAIs were consistently higher than those between physician disease activity and the BASDAIs. By leaving out the peripheral joint question, the enthetitis question, or both the peripheral joint and the enthesitis questions, there was no gain in correlation with the patient disease activity in the group without peripheral disease; however, there was substantial loss in the group with peripheral disease.
Table 3. Correlations between patients' and physicians' disease activity, ESR, CRP, BASDAI, and alternatives at time 0*
VAS physician on disease activity
VAS patient on disease activity
BASFI = Bath Ankylosing Spondylitis Functional Index; for additional abbreviations, see Table 2. P < 0.01.
Group without peripheral arthritis
Group with peripheral arthritis
Longitudinal data analysis by GEE allowed us to investigate time trends in scoring by individual patients (Table 4). From this table it can be seen that the average patient without arthritis at T0 will score 0.64 of 10 (6.4% of the scale) points higher at T6 months if he or she has arthritis at that time point. It can also been seen that excluding the joint and/or enthesitis question(s) from the original BASDAI questionnaire did not significantly change this longitudinal relationship between peripheral arthritis and scoring: the regression coefficient was 0.66 if the BASDAI was calculated without the joint question, 0.66 without the enthesitis question, and 0.68 without both questions. This analysis suggests that peripheral arthritis in patients with AS reflects higher disease activity in general, as measured by the BASDAI questionnaire. The relationship between BASFI and true peripheral arthritis was not statistically significant, suggesting that function is less influenced by having arthritis.
Table 4. Longitudinal relationship between having peripheral arthritis and score on the BASDAI and its alternatives, and the BASFI in patients with ankylosing spondylitis: first order autoregressive analysis using generalized estimating equations*
Scores for BASDAI and its derivates as well as BASFI range from 0 to 10. In all analyses, the values of the dependent variable at the previous time point is included as the autoregressive variable; the coefficient for the autoregressive variable is not presented in the table. BASDAI = Bath Ankylosing Spondylitis Disease Activity Index; BASFI = Bath Ankylosing Spondylitis Functional Index; 95% CI = 95% confidence interval.
Dependent variable in the generalized estimating equations.
Table 5 shows the longitudinal relationship between true peripheral arthritis and the 6 item scores from the BASDAI separately. The β in this table gives the amount of change of the calculated variable when assessment of arthritis changes from “no arthritis” to “arthritis” and thus the model can be interpreted as follows: A patient without arthritis will have an average increase on the fatigue question of the BASDAI of 0.75 if he has arthritis and an average increase of 0.71 on the spine and hip pain question. As can be seen, presence of arthritis increased the score on all individual questions. The effect of having arthritis was the lowest on the question on spinal pain and the largest on the peripheral joints question.
Table 5. Longitudinal relationship between having peripheral arthritis and score on the individual questions of BASDAI in patients with ankylosing spondylitis: first order autoregressive analysis using generalized estimating equations*
Scores on individual BASDAI questions range from 0 to 10. In all analyses, the values of the dependent variable at the previous time point is included as the autoregressive variable; the coefficient for the autoregressive variable is not presented in the table. BASDAI = Bath Ankylosing Spondylitis Disease Activity Index; 95% CI = 95% confidence interval.
Dependent variable in the generalized estimating equations.
The BASDAI includes the components fatigue, spinal and hip pain, peripheral joint pain and swelling, enthesitis, and morning stiffness (both severity and duration). The designers of the BASDAI considered these as important and common symptoms in AS (6). In a 6-week, double-blind, placebo-controlled nonsteroidal antiinflammatory drug study, the BASDAI proved to be a valid and appropriate composite index to define disease activity in AS (7). In this study however, only patients with axial disease were accepted and those with active peripheral involvement were excluded.
Because of the individual weight of each question of the BASDAI, we hypothesized that patients with AS limited to the axial skeleton would have lower scores on the BASDAI compared with patients with concomitant peripheral disease. We further hypothesized that this would be due to a lower scoring on both the peripheral joint question and the enthesitis question. In our study, patients with peripheral arthritis indeed reported higher disease activity compared with patients without peripheral arthritis as measured by the BASDAI. Although this difference is most obvious in the peripheral joints question and the enthesitis question, it is also present in questions on fatigue, duration and intensity of morning stiffness, and spinal pain. Cautious observation of the data shows that the higher BASDAI in patients with peripheral arthritis might be explained by 2 separate phenomena. 1) Patients with peripheral arthritis have higher overall disease activity, which is reflected by higher scores in the nonarthritis items. 2) Having peripheral arthritis—and a higher score on the peripheral arthritis question—contributes disproportionately to the BASDAI score, which is reflected by the highest standardized regression coefficient for the peripheral joints question.
As a consequence of these separate effects, the difference in BASDAI between patients with and those without arthritis only partially disappears by omitting the joints and enthesitis questions from the BASDAI questionnaire.
Similarly, different aspects of peripheral joint disease in AS have been reported in the past. In 1958, Wilkinson and Bywaters concluded that the main factor influencing prognosis in AS was the presence or absence of extraspinal joint involvement (13). Carette et al observed in 1983 in their study in 150 war veterans with AS that early peripheral joint disease predicts disease severity (14). Claudepierre et al reported in 1998 that disease activity in spondylarthropathy was linked to peripheral joint disease, disease duration, and dietary habits (15). However, Claudepierre et al evaluated a French version of the BASDAI in which a high score on spinal pain did not correlate with pain in peripheral joints (16). In our study, we found a significant correlation coefficient between question 2 (spinal and hip pain) and question 3 (joint pain) of 0.36 at T0 and ranging from 0.61 to 0.67 at T6, T12, T18, and T24. Differences in study population could be an explanation of the difference in correlation found between the mentioned questions of the BASDAI.
Our most important conclusion is that disease activity as measured with the BASDAI is higher in patients with peripheral joint disease compared with patients with disease limited to the axial skeleton, and thus that peripheral arthritis is an expression of higher disease activity. Modification of the BASDAI by omitting the question(s) on peripheral joint pain and/or enthesitis results in neither disappearance of this difference nor in a better correlation between BASDAI and other measures of disease activity.