Presence of peripheral arthritis and male sex predicting continuation of anti–tumor necrosis factor therapy in ankylosing spondylitis: An observational prospective cohort study from the South Swedish arthritis treatment group register




To examine clinical characteristics as possible predictors of long-term treatment continuation with adalimumab, etanercept, and infliximab in ankylosing spondylitis (AS) patients who had never taken biologics treated in clinical practice.


Patients in southern Sweden with active AS starting biologic therapy for the first time between October 1999 and December 2008 (n = 243, 75% men) were included in a structured clinical followup over 2 years. Patients with clinical spondylitis had not responded to at least 2 nonsteroidal antiinflammatory drugs, whereas patients who also had peripheral arthritis (n = 121) had additionally failed at least 1 conventional disease-modifying antirheumatic drug (DMARD) treatment course. The mean ± SD age at inclusion was 43 ± 12 years, with a mean ± SD disease duration prior to treatment of 16 ± 12 years.


The 2-year drug continuation rate was 74%. Male sex (hazard ratio [HR] of premature discontinuation 0.36 [95% confidence interval (95% CI) 0.19–0.68]) and the presence of peripheral arthritis (HR 0.49 [95% CI 0.27–0.88]) were found to be significant predictors of better drug survival. Furthermore, a trend was seen for more favorable drug continuation on treatment with etanercept as compared with infliximab (HR 0.50 [95% CI 0.25–1.04], P = 0.062), whereas no differences were found comparing the 3 anti–tumor necrosis factor agents in other ways. Higher baseline C-reactive protein level (HR 0.99 [95% CI 0.97–1.00], P = 0.12) and concomitant treatment with nonbiologic DMARDs (HR 0.61 [95% CI 0.34–1.10], P = 0.10) also showed trends to entail better drug adherence.


AS patients in this study have an excellent 2-year drug survival rate of 74%. Significant predictors for treatment continuation in this study were male sex and the presence of peripheral arthritis.


Ankylosing spondylitis (AS) is a chronic inflammatory disease predominantly affecting the sacroiliac joints and spine. A large proportion of patients also experience peripheral arthritis, usually an asymmetric oligoarthritis mostly involving the lower extremities, and a minority of patients experience extraarticular events such as anterior acute uveitis and aortic regurgitation together with conduction abnormalities. The prevalence of AS is higher in men, with a sex ratio of women to men of approximately 2–3 to 1 (1).

Prior to the introduction of tumor necrosis factor (TNF) inhibitors, the treatment arsenal in AS was limited to physiotherapy and nonsteroidal antiinflammatory drugs (NSAIDs). Apart from a restricted potential of sulfasalazine to alleviate peripheral arthritis, available nonbiologic disease-modifying antirheumatic drugs (DMARDs) have insufficient effects in AS (2). In contrast, blockers of TNF activity have been shown in randomized placebo-controlled trials to effectively improve several aspects of the disease, including disease activity, spinal mobility, peripheral arthritis, and enthesitis (3–9), whereas the inhibition of radiographic progression is less well established.

Considering the high cost and rare, but potentially significant, toxicity of anti-TNF agents, patients should be carefully selected for treatment. In this context, identification of patient characteristics at baseline with an ability to predict treatment outcome measured by response rates or drug continuation is clinically relevant. Several previous reports have studied predictors of treatment response, finding shorter disease duration, lower age, HLA–B27 positivity, lower Bath Ankylosing Spondylitis Functional Index (BASFI), and various indicators of active inflammation, i.e., higher C-reactive protein (CRP) level/erythrocyte sedimentation rate (ESR), higher Bath Ankylosing Spondylitis Disease Activity Index (BASDAI), and presence of spinal inflammation on magnetic resonance imaging, at baseline to predict a more favorable outcome (10–16). Data on clinical characteristics predicting drug continuation, however, remain sparse. Observational studies have shown anti-TNF continuation rates, regardless of withdrawal reason, to be higher in AS than in rheumatoid arthritis (RA) (17–19). However, 20–30% of patients with AS quit therapy within 2 years of treatment initiation (17–20). Also, an earlier study addressing AS limited to axial symptoms found higher baseline CRP levels to predict a lower risk of stopping anti-TNF therapy due to inefficacy (21).

The aim of the present study was to assess the potentials of various clinical characteristics to predict drug survival in AS patients who had never taken biologics treated with adalimumab, etanercept, or infliximab (in this study referred to as anti-TNF therapy) in clinical practice. Furthermore, possible associations of the identified predictors with treatment withdrawal due specifically to adverse events or lack of response were investigated.


Study design.

Patients with active AS starting anti-TNF therapy for the first time between October 1999 and December 2008 (n = 243) were monitored as part of the South Swedish Arthritis Treatment Group (SSATG) Register, a large, prospective, observational study cohort involving 11 rheumatology units (22). Patients eligible for inclusion had a clinical diagnosis of AS according to treating physicians specialized in rheumatology. The diagnosis was validated by using an independently selected cohort of AS patients with a valid diagnosis at one of the treatment centers in southern Sweden, as described in a previous publication (23), and subsequently linked to the SSATG Register. The Regional Ethics Board of the University of Lund approved this linkage (514/2007). The validation revealed that 90% (26 of 29 patients) of the anti-TNF–treated patients fulfilled the modified New York criteria for AS (24). Anti-TNF therapy was initiated due to high disease activity, but no predefined level of disease activity was required. Instead, decisions to initiate therapy with a certain agent were left solely to the treating physicians, with the support of national and international guidelines when they began to emerge (2, 25). The disease activity level currently recommended by Swedish guidelines for consideration of anti-TNF therapy is a BASDAI score ≥4 for more than 4 weeks (25). Patients with AS limited to axial symptoms had not responded to at least 2 NSAID regimens prior to anti-TNF therapy initiation, whereas those also experiencing peripheral arthritis had additionally failed treatment with at least 1 conventional DMARD, mainly sulfasalazine. None of the patients had any previous experience of biologic DMARDs. Glucocorticoids could be administered systemically or by local injections prior to or during the study period and concomitant use of NSAIDs or nonbiologic DMARDs was allowed. Patients (n = 8) with psoriatic skin disease were excluded. Peripheral arthritis was defined as arthritis in the extremities distal to the hips and shoulders.

The quality control character of the SSATG Register makes it part of the legislative documentation demanded in Sweden. Therefore, no formal ethical approval was required for this study.

Etanercept was administered twice weekly with a 25 mg subcutaneous dosage initially; later, often 50 mg once weekly was used. Infliximab was infused at 3 mg/kg at 0, 2, and 6 weeks, and then every 8 weeks. Depending on efficacy, the dosage of infliximab could then be increased in steps of 100 mg to a maximum of 500 mg administered at 4–8-week intervals. The average dosage after 6 months was ∼4.5 mg/kg every 8 weeks. Adalimumab was administered as a 40 mg subcutaneous dosage every other week.

At the start of anti-TNF therapy, baseline characteristics were reported by the treating physicians using a standardized protocol. This included information on demographics, disease duration defined as the onset of clinical symptoms, symptoms, data on past and present antirheumatic therapy, results of laboratory examinations, and clinical variables allowing the calculation of disease activity according to the BASDAI (26). Functional status was assessed by the BASFI (27) and by the validated Swedish version of the Health Assessment Questionnaire (HAQ) (28). Furthermore, the results of self-scored visual analog scales for pain (VAS pain) and general health (VAS global) were also reported, along with evaluators' global assessments of disease activity on a 5-grade Likert scale (global evaluation). Incomplete baseline or followup data did not exclude subjects from entry into the study cohort.

The patients were continuously enrolled during the study period and were monitored during 2 years from treatment initiation, with scheduled followup visits at weeks 2, 6, 12, 24, 52, and 104. Withdrawals from anti-TNF therapy were prospectively reported by the treating physicians and classified as being due to adverse events, lack of response/inefficacy, including both primary and secondary inefficacy, or miscellaneous, with the latter comprising reasons such as pregnancies, patient decisions, poor compliance, withdrawal due to disease remission, and other unspecified causes. No formal criteria were required to terminate treatment, and the decision was based on the judgment of the treating physician. Patients stopping therapy prior to surgery were considered to remain on treatment if the duration of the pause did not exceed 6 months. To minimize followup loss, data for all of the patients included were checked every 6 months. In case of incomplete clinical data for a period of more than 250 days, a request to report the missing data or provide the time and reason for potential drug discontinuation was sent to the treating physician.

Statistical analysis.

Differences in baseline characteristics were assessed using chi-square analysis for ordinal variables and Mann-Whitney U test or Kruskal-Wallis test for continuous variables.

Drug survival data were estimated according to Kaplan-Meier techniques and further analyzed with log rank statistics for comparing different subgroups. The potentials of baseline characteristics to predict premature termination of anti-TNF therapy were assessed using a backward deletion Cox proportional hazards model. In order to avoid bias due to changes in therapy indications over time, the analysis was stratified according to calendar year of treatment initiation. Colinearity was examined using Spearman's correlation test and covariate correlations <0.2 were required. Covariates entered in the model were chosen based on correlation, previous reports (10–17, 19, 21), and clinical relevance, and comprised age, sex, disease duration, clinical phenotype (isolated axial spondylitis versus presence of peripheral arthritis), concomitant nonbiologic DMARD use (yes versus no), type of anti-TNF therapy (infliximab versus etanercept, infliximab versus adalimumab, and etanercept versus adalimumab), CRP level, and BASDAI, BASFI, HAQ, VAS global, and global evaluation scores.

Furthermore, Cox proportional hazards models were then applied to elucidate whether identified predictors were associated with treatment withdrawal due to inefficacy or adverse events. Results of predictor analyses are shown as the hazard ratio (HR; 95% confidence interval [95% CI]) unless stated otherwise. Two-tailed P values less than 0.05 were considered statistically significant.


During the study period, 243 patients with active AS who had never taken biologics started treatment with a first anti-TNF inhibitor, with 113 receiving infliximab, 91 receiving etanercept, and 39 receiving adalimumab. Men constituted 75% of the study sample and the mean ± SD age at inclusion was 43 ± 12 years, with a mean ± SD previous disease duration of 16 ± 12 years. Fifty percent of patients (n = 122) experienced limited axial disease, whereas the other half also showed symptoms of peripheral arthritis. Baseline characteristics at the start of anti-TNF therapy, split according to this difference in clinical phenotype, are summarized in Table 1. As expected, previous and concomitant DMARD use was significantly more frequent in patients with peripheral arthritis (P < 0.01 and P = 0.02, respectively). Apart from this, a higher mean ESR in subjects displaying peripheral symptoms (P = 0.04) constituted the only significant difference at baseline. Regarding treatment, no differences in age, sex, ESR, CRP level, HAQ, global evaluation, VAS pain, VAS global, BASDAI, or BASFI were present at inclusion for patients receiving infliximab, etanercept, or adalimumab. Disease duration showed slightly higher values for patients treated with infliximab (mean ± SD 18 ± 12 years) than for etanercept (mean ± SD 14 ± 11 years) and adalimumab (mean ± SD 14 ± 13 years; P = 0.04).

Table 1. Patient characteristics at the start of anti-TNF therapy*
 Spondylitis onlyPeripheral arthritis 
n = 122Missing data, no.n = 121Missing data, no.P
  • *

    Values are the mean ± SD unless otherwise indicated. Anti-TNF = anti–tumor necrosis factor; VAS = visual analog scale; ESR = erythrocyte sedimentation rate; CRP = C-reactive protein; BASDAI = Bath Ankylosing Spondylitis Disease Activity Index; BASFI = Bath Ankylosing Spondylitis Functional Index; HAQ = Health Assessment Questionnaire; DMARDs = disease-modifying antirheumatic drugs; MTX = methotrexate.

  • Comparing the 2 groups.

  • From the onset of clinical symptoms.

Age, years42 ± 12 44 ± 12 0.29
Men, no. (%)91 (75) 90 (74) 0.97
Disease duration, years15 ± 12 16 ± 11 0.31
VAS pain, mm64 ± 201161 ± 23140.42
VAS global, mm63 ± 221160 ± 21140.45
Global evaluation, mm2.0 ± 0.792.1 ± 0.7100.50
ESR, mm/hour24 ± 201031 ± 25100.04
CRP level, mg/liter19 ± 201126 ± 30100.19
BASDAI score5.4 ± 2.0425.4 ± 1.9660.99
BASFI score4.2 ± 2.1434.4 ± 2.1680.75
HAQ score0.72 ± 0.54110.79 ± 0.56170.40
Previous number of DMARDs1.2 ± 0.9 1.7 ± 1.1 < 0.01
Present antirheumatic therapy, no. (%)     
 Anti-TNF monotherapy55 (45) 36 (30) 0.02
 Concurrent sulfasalazine use20 (16) 15 (12) 0.48
 Concurrent MTX use41 (34) 61 (50) 0.01
Type of anti-TNF therapy, no. (%)     
 Infliximab use55 (45) 58 (48) 0.75
 Etanercept use48 (39) 43 (36) 0.63
 Adalimumab use19 (16) 20 (17) 0.98

At inclusion, female patients displayed significantly lower baseline mean ± SD CRP levels (16.5 ± 20.8 versus 24.4 ± 26.5 mg/liter; P = 0.02) and reported higher mean ± SD VAS pain scores (68.8 ± 15.9 versus 60.0 ± 22.6 mm; P = 0.02) than male patients.

In Figure 1, associations of clinical phenotype with treatment discontinuation due to any reason, lack of response, and adverse events analyzed by the Kaplan-Meier technique are displayed. Overall, 94% of patients with peripheral arthritis compared with 90% of patients with axial disease only remained on the initial anti-TNF therapy after 6 months (P = 0.05). Corresponding drug survival rates at 12 and 24 months were 86% compared with 74% and 78% compared with 68% for peripheral arthritis and isolated spondylitis, respectively (P = 0.05). Due to the active monitoring/active followup mechanisms described above, no patients were lost to followup without information on date of and reason for termination of therapy. At 24 months, as seen in Figure 1, Kaplan-Meier estimated treatment withdrawals were almost equally distributed between adverse events (11%) and inefficacy (13%), whereas only a few percentages were reported as being due to miscellaneous reasons (data not shown). In the latter group, only 1 patient was reported to stop treatment following clinical remission of disease. In less than 5% of cases, treatment cessation was reported as being caused by both adverse events and lack of response. These were all reclassified as being due to adverse events.

Figure 1.

Anti–tumor necrosis factor discontinuation due to all reasons, treatment failure, and adverse events, split according to isolated axial disease and combined axial and peripheral disease.

In order to study whether patients with the most favorable response are actually the ones who remained on therapy, a subgroup analysis of VAS global changes was studied. Relative changes in VAS global scores have previously been shown to correlate well with treatment response, as measured by the Assessment of SpondyloArthritis International Society criteria (29, 30). Therefore, patients were divided into 2 groups based on changes in VAS global scores during the first year of treatment. The greatest relative change from baseline in VAS global scores at 3, 6, or 12 months of followup was used for each patient, splitting the study population into those with a maximal VAS global improvement above or below the median. Drug continuation rates at 12 and 24 months of therapy were then calculated for the 2 groups. A median VAS global improvement of 48.5% compared with baseline was seen during the first year of treatment (n = 173). After 12 months, 93% of the patients with favorable VAS global improvement remained on treatment, as opposed to 79% of those with a less favorable VAS global response (P < 0.01). Corresponding figures at 24 months of therapy were 85% and 68%, respectively (P < 0.01).

To study the predictive value of baseline characteristics for continuation of anti-TNF therapy (adalimumab, etanercept, and infliximab), a Cox proportional hazards model was computed. Results of the predictor analysis are shown in Table 2. Male sex (HR for discontinuation 0.36 [95% CI 0.19–0.68]) and presence of peripheral arthritis (HR 0.49 [95% CI 0.27–0.88]) were found to be statistically significant predictors. Furthermore, a trend was seen for more favorable drug continuation for treatment with etanercept as compared with infliximab (HR 0.50 [95% CI 0.25–1.04], P = 0.06), whereas no differences were found comparing the 3 anti-TNF agents in other ways. Higher baseline CRP level (HR 0.99 [95% CI 0.97–1.00], P = 0.12) and concomitant treatment with nonbiologic DMARDs (HR 0.61 [95% CI 0.34–1.10], P = 0.10) also showed trends to entail better drug adherence, although again nonsignificantly. No associations with drug survival levels were found for age, disease duration, and BASDAI, BASFI, HAQ, VAS global, and global evaluation scores at baseline.

Table 2. Cox proportional HRs with 95% CIs and levels of significance for stopping anti–tumor necrosis factor treatment for any reasons*
CovariateHR95% CIP
  • *

    HR = hazard ratio; 95% CI = 95% confidence interval; DMARD = disease-modifying antirheumatic drug; CRP = C-reactive protein.

Peripheral vs. axial arthritis0.490.27–0.880.02
Male vs. female sex0.360.19–0.68< 0.01
Concomitant DMARD therapy0.610.34–1.100.10
CRP level0.990.97–1.000.12
Etanercept vs. infliximab0.500.25–1.040.06
Infliximab vs. adalimumab1.400.58–3.420.45

Based on the above findings, a subgroup analysis was performed. Sex, clinical phenotype, CRP levels, concomitant DMARD use, and type of anti-TNF agent were entered into further Cox proportional hazards models to reveal possible associations with treatment withdrawal due to adverse events or lack of efficacy. Results of these subanalyses, presented in Table 3, show male sex (HR 0.38 [95% CI 0.15–0.96]) and presence of peripheral arthritis (HR 0.36 [95% CI 0.15–0.89]) to entail a significantly higher chance not to stop treatment due to inefficacy (32 events and 211 censored cases). Conversely, no correlations were seen with discontinuation due to adverse events (27 events and 216 censored cases). Therefore, the protective impact on drug continuation of male sex and peripheral arthritis as seen in Table 2 is explained mainly by a lower risk of ceasing therapy due to inefficacy.

Table 3. Subgroup analysis with Cox proportional HRs, 95% CIs, and levels of significance for stopping anti–tumor necrosis factor treatment split according to withdrawals because of adverse events or treatment failure*
CovariateHR95% CIP
  • *

    HR = hazard ratio; 95% CI = 95% confidence interval; DMARD = disease-modifying antirheumatic drug; CRP = C-reactive protein.

Withdrawal due to inefficacy   
 Peripheral vs. axial arthritis0.360.15–0.890.03
 Male vs. female sex0.380.15–0.960.04
 Concomitant DMARD therapy0.500.20–1.250.14
 CRP level0.990.97–1.020.70
 Etanercept vs. infliximab0.700.22–2.220.55
 Infliximab vs. adalimumab1.620.44–5.930.47
Withdrawal due to adverse events   
 Peripheral vs. axial arthritis1.060.44–2.580.90
 Male vs. female sex0.530.19–1.430.21
 Concomitant DMARD therapy0.660.26–1.640.37
 CRP level0.990.96–1.010.27
 Etanercept vs. infliximab0.460.16–1.350.16
 Infliximab vs. adalimumab0.370.08–1.860.23


This study identifies the presence of peripheral arthritis and male sex as predictors of better anti-TNF therapy continuation in patients with active AS, mainly due to a lower risk of ceasing treatment because of insufficient therapeutic response. Two years after starting a first anti-TNF remedy, 74% of patients with AS remained on therapy.

Drug continuation rates found in this study are generally in accordance with those previously reported from observational settings, most of which range between 70% and 80% continuation after 2 years (17–20). Direct comparisons are rendered difficult by the varying inclusion criteria, study group characteristics, and termination bias of different studies. However, these studies uniformly underscore the need of accounting for withdrawals to therapy when making health economic analysis of patients from registries.

In AS, peripheral joint involvement is usually manifested as an asymmetric oligoarthritis, mostly affecting the hips, knees, ankles, and shoulders. Available data regarding the prevalence of peripheral arthritis vary greatly. Depending on the inclusion criteria, the definitions of peripheral arthritis (mainly differing in respect to the hips and shoulders), the ethnicity of the study samples, and the inherent transiency of peripheral symptoms, frequencies ranging from approximately 30–60% are reported (31–35). Reviewing earlier Scandinavian studies, Gran and Skomsvoll found a 39.8% prevalence of peripheral arthritis in 100 Norwegian patients with AS (36), whereas Vinje et al report radiographic signs of arthritis in at least one joint (including the hips and shoulders) in 41% of surveyed AS patients (37). The 50% frequency of peripheral arthritis found in the present study is therefore in the higher range, but still well compatible with international findings. The ratio of men to women in our cohort was 3:1, well in accordance with available epidemiologic data (1). Overall, the present study sample therefore seems to be rather representative of the broader spectrum of patients with chronic AS.

To our knowledge, the current study is the first to show the presence of peripheral arthritis to predict more favorable anti-TNF treatment survival in AS. This impact seems to be mainly related to a lower risk of ceasing treatment due to inefficacy. Although anti-TNF therapy affects both axial and peripheral manifestations of AS (3–9), it is clinically easier to distinguish active inflammation from irreversible damage/degeneration in peripheral joints than in the spine. Therefore, it can be speculated that one reason for this finding might reside in a higher clinical accuracy in identifying patients with active inflammation most likely to respond to anti-TNF treatment in the group with peripheral symptoms. Moreover, patients experiencing peripheral as well as axial disease might be expected to be better motivated to continue treatment than those with a more limited phenotype.

In a previous study, a trend was found for the presence of peripheral arthritis to imply a somewhat lower chance to reach a BASDAI50 response after 12 weeks of treatment with infliximab or etanercept (odds ratio 0.45 [95% CI 0.20–1.02], P = 0.053) (11). However, the BASDAI is not discriminative toward peripheral arthritis because only 1 of 6 dimensions in the questionnaire directly addresses peripheral symptoms (26).

The finding that male sex entails a lower risk of premature TNF inhibitor withdrawal in AS is supported by recent results from a Czech database (19), as well as from a large observational study of patients with varied arthritic rheumatic diseases (11.2% experiencing AS) (17). In RA, several publications have indicated male sex to predict better responsiveness, mainly a greater chance to achieve remission, to treatment with conventional DMARDs as well as to biologics (38–41). This, however, remains inconclusive and may partly be explained by how remission is defined in combination with the generally lower baseline disease activity seen in male RA patients (42). The lower CRP levels and higher VAS pain scores seen in the female patients at baseline may be speculated to indicate a higher level of comorbidity with chronic pain syndromes such as fibromyalgia in this group. If so, this may affect subjectively perceived treatment efficacy and therefore willingness to adhere to therapy in a negative way. Interestingly, similar sex differences in CRP levels and VAS pain scores were also seen in RA patients in southern Sweden (42).

As might be expected, indicators of active inflammation at baseline have been shown to predict better anti-TNF response rates in AS (10, 11, 14–16), likely due to a greater potential for improvement during treatment. Moreover, elevated baseline CRP level was the only variable found to significantly protect against premature anti-TNF withdrawal due to inefficacy in the French study cited above (21). A similar trend for CRP levels was found in the current analysis, whereas other indicators of ongoing inflammation, i.e., BASDAI, VAS global, and global evaluation scores, were not associated with drug continuation. It could be speculated that decisions to cease treatment are more closely related to recent fluctuations in disease activity than to baseline levels, which probably explains the weaker association of baseline inflammatory parameters to drug survival than to responsiveness.

As previously shown in RA (43) and psoriatic arthritis (44), concomitant use of nonbiologic DMARDs tended to protect against premature anti-TNF treatment termination in this AS material. In contrast to the situation in RA, however, this was not clearly due to a decreased risk of side effects. At this point it should be noted that reasons for stopping treatment are scored by the treating physician, and interobserver variation in classification of stop reason therefore exists. Moreover, subgrouping of the patients reduces the statistical power and introduces problems of mass significance, making these analyses less reliable. Therefore, results regarding the cause of treatment termination must be interpreted with care, placing more emphasis on overall adherence to therapy (43). The rate of overall drug survival should be considered a surrogate marker for favorable treatment response combined with the absence or presence of adverse events; however, the rate relies solely on the clinical decision of the treating physician, and the validity of this measure could therefore always be questioned. On the other hand, our subgroup analysis clearly showed that patients with the most favorable VAS global response also had better drug survival, validating the fact that patients who remain on therapy actually have effective treatment. Furthermore, it is not likely that making rigorous criteria for when to stop or start treatment would strengthen the quality of “drug survival” in open-labeled settings because these criteria are always at risk of being manipulated to fit with the belief of the treating physician.

The trend for more favorable drug continuation with etanercept as compared with infliximab is in line with previous results from studies of patients with varied arthritic rheumatic diseases (17), RA (43), and psoriatic arthritis (44), although the opposite was noted in 92 AS patients reviewed by Brocq et al (18). This disparity of previous findings, the risk for confounding by indication, and the varying access of different TNF inhibitors during the study period suggests that this result should be interpreted with caution.

A large, recent study identified HLA–B27 positivity as a strong predictor of response to treatment with adalimumab in AS (15), and a similar, although nonsignificant, trend for treatment with infliximab or etanercept has been reported previously (11). On the other hand, a history of at least one episode of uveitis was not significantly correlated with treatment response (15). Data on HLA–B27 status and extraarticular manifestations of AS were not part of the SSATG protocol, but should be considered as possible confounders to the results found in the current study, and it would be valuable to investigate these in future studies.

The open, observational setting of this study inherently entails limitations such as possible bias regarding patient selection, assignment of treatment, and collection of clinical data (45). Decisions to start or stop therapy with a certain agent were based solely on clinical practice and experience of treating physicians, with national and international guidelines as support. Higher levels of missing data regarding certain baseline parameters (Table 1) are another result of the observational setting. Although these methodologic limitations should be borne in mind when interpreting the results, observational studies remain important in offering information from daily clinical practice. Moreover, the centralized, prospective collection and entry of SSATG data optimize uniformity of interpretation of forms and results. In conclusion, this observational study of patients with active AS treated with a first TNF inhibitor showed a good overall drug continuation, with 74% of patients remaining on treatment after 2 years. Male sex and peripheral arthritis were identified as significant predictors of treatment continuation, whereas nonsignificant trends were seen for concomitant DMARD use, higher baseline CRP levels, and treatment with etanercept as compared with infliximab. Age, disease duration, and BASDAI, BASFI, HAQ, VAS global, and global evaluation scores at baseline were not associated with continuation of anti-TNF therapy in this study.


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. Kristensen 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. Kristensen, Karlsson.

Acquisition of data. Kristensen, Karlsson, Geborek.

Analysis and interpretation of data. Kristensen, Karlsson, Englund, Petersson, Saxne, Geborek.


We are indebted to all of the colleagues and staff in the SSATG for cooperation and data supply.