University of Groningen Clinical and spinal radiographic outcome in axial spondyloarthritis Maas, Fiona

(1)

University of Groningen

Clinical and spinal radiographic outcome in axial spondyloarthritis

Maas, Fiona

IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below.

Document Version

Publisher's PDF, also known as Version of record

Publication date: 2017

Link to publication in University of Groningen/UMCG research database

Citation for published version (APA):

Maas, F. (2017). Clinical and spinal radiographic outcome in axial spondyloarthritis: Results from the GLAS cohort. Rijksuniversiteit Groningen.

Copyright

Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).

Take-down policy

If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.

Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.

(2)

Chapter 3

Reduction in spinal radiographic progression

in ankylosing spondylitis patients receiving

prolonged treatment with TNF-α inhibitors

Fiona Maas Suzanne Arends Elisabeth Brouwer Ivette Essers Eveline van der Veer Monique Efde Peter van Ooijen Rinze Wolf Nic Veeger Hendrika Bootsma Freke Wink Anneke Spoorenberg

(3)

ABSTRACT

Objective: To evaluate the course of spinal radiographic progression up to 8 years of follow-up in a large cohort of ankylosing spondylitis (AS) patients treated with TNF-α inhibitors. Methods: Consecutive patients from the Groningen Leeuwarden AS (GLAS) cohort starting TNF-α inhibitors between 2004-2012 were included. Baseline and biannual radiographs were randomized with radiographs of TNF-α naïve AS patients and scored in chronological time order according to mSASSS. The course of radiographic progression (linear or non-linear) was investigated using generalized estimating equations. Primary analysis was performed in patients with complete data over 4, 6, and 8 years of follow-up. Sensitivity analysis was performed after single linear imputation of missing radiographic data and after adjusting for patient characteristics with possible influence on radiographic progression.

Results: At baseline, median mSASSS of 210 included AS patients was 2.8 (IQR: 0.0-12.0), mean mSASSS 10.0 ± 15.5. During the first 4 years, radiographic progression followed a linear course (estimated mean progression rate was 1.7 for 0-2 and 2-4 years). A deflection from a linear course was found in patients with complete and imputed data over 6 and 8 years. Estimated mean 2-year progression rate reduced from 2.3 to 0.8 in patients with complete 8-year data. The same pattern was found after adjustment for baseline mSASSS, presence of syndesmophytes, gender, HLA-B27 status, age, symptom duration, smoking duration, BMI, disease activity, and NSAID use.

Conclusions: This observational cohort study in AS patients receiving long-term TNF-α inhibitors showed a reduction in spinal radiographic progression after more than 4 years of follow-up.

(4)

3 INTRODUCTION

Ankylosing spondylitis (AS) is a chronic inflammatory rheumatic disease that primarily affects the axial skeleton. Excessive bone formation such as syndesmophytes and ankylosis of the spine is an important disease outcome of AS [1]. It is associated with impaired spinal mobility and decreased ability to perform daily activities [2]. Therefore, reducing spinal radiographic progression is an important goal in the treatment of AS.

Since the introduction of tumor necrosis factor-alpha (TNF-α) inhibitors, multiple studies have been performed to investigate whether these drugs influence spinal radiographic progression. Previous open-label extension studies with 2 years of follow-up have found similar radiographic progression in AS patients treated with TNF-α inhibitors compared to patients from a historical cohort not treated with TNF-α inhibitors [3-6]. In a small study in 33 AS patients treated with infliximab for 4 years, a lower progression rate was observed compared to published data from a historical cohort [7]. Although this was an indirect comparison in which radiographs were not scored by the same readers, the authors concluded that infliximab may decelerate radiographic progression [7]. In our previous analysis in 176 AS patients receiving TNF-α inhibitors, in which radiographs were scored with unknown time sequence, radiographic progression followed a linear course over a mean follow-up duration of 3.8 ± 1.8 years [8].

Two studies with radiographic data up to 8 years of follow-up reported about a possible relationship between TNF-α inhibitors and less spinal radiographic progression in AS over time [9,10]. A small retrospective study with 8 years of follow-up reported less progression during 4 to 8 years of follow-up in 22 AS patients treated with TNF-α inhibitors in direct comparison to 34 patients from a historical cohort [9]. Analyses of a large prospective study in 334 AS patients with 1.5 to 9 years of follow-up indicated that the use of TNF-α inhibitors was significantly associated with less spinal radiographic progression. This was especially seen in patients with more than 3.9 years of follow-up [10].

Although the results of these long-term follow-up studies seem promising, they are under scientific debate because of methodological reasons. It is known that radiographic progression in AS is overall slow and highly variable between patients [11,12]. Therefore, differences in patient numbers at the different time points during follow-up can affect the outcome measure of interest, in this case radiographic progression. Furthermore,

(5)

studies using direct comparisons with historical cohorts are of limited value since patient characteristics and treatment regimens may differ over time [13]. Differences in patient numbers and characteristics within studies should be taken into account during the analyses and interpretation of the results. Ideally, long-term randomized placebo-controlled trials should be performed, but this is not ethical due to a proven major effect of TNF-α inhibitors on clinical outcome, especially on disease activity. Prospective observational studies with standardized longitudinal assessments are, therefore, the second best option to obtain more robust data on the long-term course of spinal radiographic progression during treatment with TNF-α inhibitors.

Data from a prospective historical cohort of AS patients mainly treated with non-steroidal anti-inflammatory drugs (NSAIDs) and with radiographic assessment every 2 years, showed a linear course of spinal radiographic progression over 12 years of follow-up [11]. To evaluate a possible deceleration of spinal radiographic progression during treatment with TNF-α inhibitors, comparable analysis should be performed in AS patients receiving prolonged treatment with TNF-α inhibitors.

The aim of the present study was to evaluate the course of spinal radiographic progression up to 8 years of follow-up in a large prospective observational cohort of AS patients treated with TNF-α inhibitors in daily clinical practice.

METHODS

This study included consecutive outpatients from the Groningen Leeuwarden AS (GLAS) cohort who started treatment with TNF-α inhibitors because of active disease between November 2004 and April 2012. Patients had lateral radiographs of the cervical and lumbar spine available at baseline and after at least one follow-up visit at 2, 4, 6 or 8 years.

The GLAS cohort is an ongoing prospective longitudinal observational cohort study in the North of the Netherlands with a standardized assessment and management protocol

[8,14,15]. All patients were 18 years or older, fulfilled the modified New York criteria for AS, and the ASAS criteria to start TNF-α inhibitors (Bath AS Disease Activity Index (BASDAI) ≥4 and/or active disease based on expert opinion) [16].

(6)

3

Type, dose, and frequency of TNF-α inhibitor use were recorded during follow-up. Duration

of TNF-α inhibitor exposure was expressed as the percentage of follow-up time. Patients who stopped or switched between different TNF-α inhibitors during follow-up were included in the analysis.

Patients were clinically evaluated at baseline, after 3 months, and then every 6 months. At baseline, gender, age, symptom duration, time since diagnosis, HLA-B27 status, years of smoking, and history of extra-articular manifestations were recorded. At all visits including baseline, disease activity was assessed with BASDAI, ASDAS_CRP, physician’s and patient’s global assessment (GDA), C-reactive protein (CRP), and erythrocyte sedimentation rate (ESR). Furthermore, body weight and height were measured to calculate body mass index (BMI), concomitant use of NSAIDs or disease-modifying antirheumatic drugs (DMARDs) were recorded, and the ASAS-NSAID index was calculated [17].

The GLAS cohort was approved by the local ethics committees of the Medical Center Leeuwarden (MCL) and the University Medical Center Groningen (UMCG). All patients provided written informed consent according to the Declaration of Helsinki.

Radiological assessment

Two trained readers (FM and IE) scored the radiographs in chronological time order using the modified Stoke AS Spine Score (mSASSS) [18,19]. Readers were not familiar with the patients and all patient characteristics were removed from the radiographs. To avoid potential reader bias due to the knowledge of the applied therapy, radiographs of included patients were randomized and scored with radiographs up to 8 years of AS patients from a historical cohort, not treated with TNF inhibitors.

In case of radiographs with ≤3 missing vertebral corners, the missing vertebral corner was substituted by the score of the vertebral corner at the previous or next observation added or subtracted by the mean progression score of the corresponding spinal segment (score 0 as minimum, score 3 as maximum) as proposed by Ramiro et al. [4]. If the vertebral corner at the previous or next observation was also missing, the mean score of the spinal segment was used to substitute the missing corner. Radiographs with >3 missing vertebral corners were excluded from the analysis. Patients with complete ankylosis (total mSASSS of 72) at baseline were excluded from the analysis because these patients could not show progression in mSASSS during follow-up.

(7)

The average mSASSS total score of the two readers was used in the analysis. Inter-observer reliability for mSASSS status scores was very good with intraclass correlation coefficient (ICC; two-way mixed effects model, single measures, absolute agreement) of 0.974 or higher. Inter-observer reliability for mSASSS progression scores was also good with ICC’s of 0.712 or higher. Bland-Altman plots revealed no systematic error. When there was discrepancy in mSASSS total scores beyond the 95% limits of agreement, a third independent experienced reader (AS) scored the radiographs. The vertebral corner score of the primary reader closest to the third reader was used to calculate the final mSASSS total score.

The smallest detectable change (SDC: 1.96*SD_{∆(progression score)/(√2*√k),}in which k represents the number of readings) [20], was 2.3, 2.7, 3.2, and 4.3 for 0-2, 0-4, 0-6, and 0-8 time intervals, respectively.

Statistical analysis

Descriptive statistics and generalized estimating equations (GEE) were used to evaluate spinal radiographic progression. Spinal radiographic progression for the different time intervals (0-2, 0-4, 0-6 or 0-8 years) was expressed as mean ± SD and median (interquartile range (IQR)). The SDC, see above, was used to detect the number of patients (%) who showed radiographic progression above the measurement error. In addition, the number of patients who developed new non-bridging or bridging syndesmophytes according to both readers was reported.

The GLAS cohort is still ongoing and therefore, not all patients had already reached 4, 6, and 8 years of follow-up (Figure 1). Since characteristics of patients who are included more recently might differ from those included at the start of the cohort, the course of radiographic progression was analyzed separately in three groups; patients with 4, 6, and 8 years of follow-up. GEE with linear and different non-linear (quadratic, cubic, square root, logarithmic, and exponential) time functions were conducted to investigate whether spinal radiographic progression follows a linear or non-linear course [8,11]. The exchangeable correlation matrix was used to take into account the high within patient correlations. First, the linear and non-linear relationships between radiographic damage and time were investigated using Wald statistics. A time function with p-value ≤0.05 was considered as statistically significant. Subsequently, the mean estimated mSASSS status and progression scores of the models were calculated based on the estimated intercept and regression coefficients. Data were presented for the significant model with the best fit for the data based on the corrected

(8)

3

quasi-likelihood information criterion (QICC). The model with the smallest QICC represents

the model with the smallest probability for lost information and therefore the best fit for the data [21].

Complete case analysis was used to avoid potential bias due to differences in patient numbers and characteristics at the different time points during follow-up, as recommended by Machado et al. [15]. Sensitivity analyses were performed after single linear imputation of missing radiographic data in patients with missing data at one or more intermediate follow-up visits. Missing data were substituted by the mean mSASSS of the previous and next observation of the patient assuming a linear course during the imputed time interval. In order to investigate the robustness of the results, all models were adjusted for patient characteristics with possible influence on radiographic progression (i.e. baseline mSASSS, presence of syndesmophytes, gender, HLA-B27 status, age, symptom duration, smoking duration, BMI, disease activity, and NSAID use over time) [22-24]. Statistical analysis was performed with IBM SPSS Statistics 22 (SPSS, Chicago, IL, USA).

RESULTS

In total, 210 of 266 AS patients who started TNF-α inhibitor treatment because of active disease had baseline radiographic data (Figure 1). Baseline characteristics were shown in Table 1. Patients who were excluded because of missing radiographs (n=38) or >3 missing vertebral corners (n=10) at baseline had similar baseline characteristics as the 210 included patients, except for age (mean age 46 vs. 42 years, p<0.05). Eight patients were excluded because of complete ankylosis. As expected, these 8 patients were older (mean age 55 vs. 42, p<0.01) and had longer symptom duration than the included patients (median 38 vs. 14 years, p<0.01).

Of 210 included patients, 160, 98, and 45 had reached 4, 6, and 8 years of follow-up, respectively. In total, 27 (13%) patients were lost to follow-up (Figure 1). Patients with 6 and 8 years of follow-up used NSAIDs more frequently and had higher CRP levels at baseline than patients with shorter follow-up time. Patients with 8 years of follow-up also had higher ASDAS and more spinal radiographic damage at baseline (Supplementary Table S1).

(9)

Figure 1. Flowchart of included AS patients: patients with 4, 6, and 8 years of follow-up including (complete) radiographic data.

(10)

3

Patients were exposed to TNF-α inhibitors during >97% of their follow-up time; median

exposure duration was 3.9, 5.9, and 7.8 years during 4, 6, and 8 years of follow-up, respectively. All disease activity measures improved significantly after starting TNF-α inhibitor treatment and remained stable at the group level during follow-up (p<0.001, data not shown). As a result, the use of NSAIDs decreased rapidly over time; mean ASAS-NSAID index decreased from 67 to 43, 30, 23, 11 at 2, 4, 6, and 8 years, respectively (median ASAS-NSAID index decreased from 60 to 0 over time).

Of 160, 98, and 45 patients who had reached 4, 6, and 8 years of follow-up, 110, 53, and 19 had complete radiographic data at all 2-year time points, respectively (Figure 1). Patient characteristics were comparable between patients with and without complete data, except for lower age in patients with versus without complete 8-years data (Supplementary Table S2).

Spinal radiographic progression

At baseline, the median mSASSS of all included patients was 2.8 (IQR: 0.0-12.0) and the mean mSASSS was 10.0 ± 15.5 (Table 2). One or more syndesmophytes were present in 108 (54%) patients, of which 60 patients had at least one bridging syndesmophyte. During the first 2 years of follow-up, the mean spinal radiographic progression was 1.6 ± 2.8 mSASSS units. Forty (25%) patients showed progression above the SDC and 42 (26%) developed one or more new non-bridging or bridging syndesmophytes during 2 years of follow-up. Patients with 8 years of follow-up had a mean progression rate of 7.0 ± 6.3 mSASSS units. Of these patients, 24 (59%) patients showed progression above the SDC and 27 (66%) developed one or more new non-bridging or bridging syndesmophytes (Table 2).

(11)

Table 1. Baseline characteristics of all included AS patients and of patients with complete data over 4, 6 or 8 years. All patients n=210 Complete 4yr data n=110 Complete 6yr data n=53 Complete 8yr data n=19 Male gender 145 (69) 79 (72) 38 (72) 16 (84) Age (yrs) 41.6 ± 11.5 41.2 ± 11.7 39.7 ± 11.1 39.3 ± 10.7

Symptom duration (yrs) 14 (8-24) 15 (7-23) 15 (7-21) 17 (9-23)

Time since diagnosis (yrs) 6 (1-15) 5 (1-14) 5 (1-14) 9 (3-17)

HLA-B27+ 160 (78) 88 (80) 43 (81) 17 (90)

BMI (kg/m2) 25.8 ± 4.3 26.2 ± 3.9 25.3 ± 3.8 24.9 ± 2.0

Smoking duration (yrs) 12 (0-23) 13 (0-24) 13 (0-26) 13 (0-27)

NSAID use 164 (80) 89 (81) 45 (85) 18 (95) ASAS-NSAID index 60 (25-100) 69 (17-100) 67 (38-100) 50 (25-100) DMARD use 38 (18) 21 (19) 16 (30) 5 (26) First TNF-α inhibitor Infliximab 28 (13) 17 (15) 12 (23) 5 (26) Etanercept 132 (63) 71 (65) 35 (66) 14 (74) Adalimumab 50 (24) 22 (20) 6 (11) 0 (0) BASDAI (0-10) 6.0 ± 1.7 6.0 ± 1.6 5.8 ± 1.7 6.0 ± 1.4 ASDASCRP 3.7 ± 0.8 3.7 ± 0.7 3.8 ± 0.8 4.0 ± 0.6 CRP (mg/L) 13 (4-22) 12 (4-22) 14 (7-25) 17 (12-40) ESR (mm/hr) 21 (10-34) 20 (9-34) 21 (11-34) 21 (11-37) Patient’s GDA (0-10) 7 (5-8) 7 (6-8) 7 (5-8) 7 (5-8)

mSASSS (range 0-72) mean 10.0 ± 15.5 10.7 ± 16.0 8.2 ± 12.9 10.0 ± 12.9

median 2.8 (0.0-12.0) 3.6 (0.0-15.8) 3.7 (0.0-11.4) 5.4 (1.0-17.1)

≥1 syndesmophyte 108 (54) 60 (55) 28 (53) 12 (63)

Values are presented as number of patients (%), mean ± SD, or median (IQR).

Abbreviations: AS: ankylosing spondylitis; HLA: human leukocyte antigen; BMI: body mass index; NSAID: non-steroidal anti-inflammatory drug; ASAS: Assessment of SpondyloArthritis international Society; DMARD: disease-modifying antirheumatic drug; BASDAI: Bath AS disease activity index; ASDAS: AS disease activity score; CRP: C-reactive protein; ESR: erythrocyte sedimentation rate; GDA: global disease activity; mSASSS: modified Stoke AS spine score.

(12)

3

Table 2. Baseline damage and spinal radiographic progression in all AS patients who started treatment

with TNF-α inhibitors; observed data.

n Mean ± SD Median (IQR) SDC >SDC n (%) new syndesmophytes n (%)

Baseline mSASSS 210 10.0 ± 15.5 2.8 (0.0-12.0)

mSASSS progression 0-2yr 163 1.6 ± 2.8 0.0 (0.0-2.1) 2.3 40 (25) 42 (26)

mSASSS progression 0-6 yr 80 4.2 ± 4.8 2.4 (0.0-7.3) 3.2 35 (44) 45 (55)

Abbreviations: mSASSS: modified Stoke AS spine score; SDC: smallest detectable change.

Complete case analyses

Baseline mSASSS and progression rates of patients with complete 4, 6, and 8-years data are shown in Table 3 and Figure 2. The observed data suggest a reduction in spinal radiographic progression over time.

In patients with complete mSASSS data over 4 years of follow-up, GEE analysis revealed that there was a linear relationship between spinal radiographic damage and time (Supplementary Figure S1A). Based on the estimated parameters of the linear time model, the mean progression rates were estimated to be 1.7 mSASSS units for 0-2 and 2-4 years (Table 4). Although the mean estimated status and progression scores of non-linear time models were similar to the observed status and progression scores, these non-linear models were not statistically significant (Supplementary Table S3).

In patients with complete mSASSS data over 6 years of follow-up, GEE analysis revealed that there was a non-linear relationship between spinal radiographic damage and time (Supplementary Figure S1B). The mean estimated status and progression scores of all the non-linear time models showed a reduction in spinal radiographic progression over time (Supplementary Table S3). Based on the estimated parameters of the non-linear time model with the best fit for the data, the mean progression rates were estimated to be 1.7, 1.6, and 1.1 during 0-2, 2-4, and 4-6 years, respectively (Table 4).

(13)

Table 3. Baseline damage and spinal radiographic progression in AS patients with complete data over 4, 6, or 8 years of follow-up; observed data.

Complete 4yr data n Mean ± SD Median (IQR)

Baseline mSASSS 110 10.7 ± 16.0 3.6 (0.0-15.8)

mSASSS progression 0-2yr 110 1.8 ± 3.0 0.5 (0.0-2.3)

Baseline mSASSS 53 8.2 ± 12.9 3.7 (0.0-11.4)

Baseline mSASSS 19 10.0 ± 12.9 5.4 (1.0-17.1)

Abbreviation: mSASSS: modified Stoke AS spine score.

Figure 2. Observed individual spinal radiographic progression of patients with complete data over 4

(14)

3

Table 4.

R

elationship bet

w

een spinal radiog

raphic damage and time with GEE estimat

ed mean mSASSS status scor

es and pr

og

ression scor

es of the model

with the best fit f

or the data in AS patients with complet

e and imput ed data o ver 4, 6, or 8 y ears of f ollo w-up . Follo w -up 4 y ears 6 y ears 8 y ears Complet e case analy sis (n=110) Sensitivit y analy sis (n=132) † Complet e case analy sis (n=53) Sensitivit y analy sis (n=80) † Complet e case analy sis (n=19) Sensitivit y analy sis (n=41) † Rela

tion with time

Linear Linear Non-linear Non-linear Non-linear Non-linear Estima ted mean (95% CI) Estima ted mean (95% CI) Estima ted mean (95% CI) Estima ted mean (95% CI) Estima ted mean (95% CI) Estima ted mean (95% CI)

mSASSS status sc

or es 0yr 10.7 (7.4-14.1) 11.5 (8.4-14.6) 8.2 (4.2-12.1) 8.7 (5.4-11.9) 9.9 (2.9-17.0) 13.4 (7.2-19.6) 2yr 12.4 (8.6-16.2) 13.3 (9.8-16.8) 9.9 (5.2-14.7) 10.4 (6.5-14.2) 12.2 (2.2-22.3) 16.2 (7.9-24.5) 4yr 14.1 (9.8-18.3) 15.0 (11.1-18.9) 11.5 (5.8-17.2) 11.9 (7.3-16.5) 13.7 (1.4-25.9) 18.3 (7.4-29.1) 6yr – – 12.6 (5.4-19.7) 12.9 (7.2-18.6) 14.7 (0.7-28.7) 19.7 (6.0-33.4) 8yr – – -– 15.5 (0.0-31.1) 20.5 (3.4-37.5) mSASSS prog ression sc or es 0-2yr 1.7 (1.2-2.1) 1.7 (1.3-2.1) 1.7 (1.0-2.6) 1.7 (1.1-2.3) 2.3 (-0.7-5.4) 2.8 (0.6-4.9) 2-4yr 1.7 (1.2-2.1) 1.7 (1.3-2.1) 1.6 (0.6-2.5) 1.5 (0.8-2.3) 1.4 (-0.7-3.6) 2.1 (-0.4-4.6) 4-6yr – – 1.1 (-0.4-2.5) 1.0 (-0.1-2.1) 1.0 (-0.8-2.8) 1.4 (-1.5-4.3) 6-8yr – – -– 0.8 (-0.8-2.4) 0.8 (-2.5-4.0) Abbr eviations: GEE: G eneraliz

ed estimating equations; mSASSS: modified St

ok

e AS spine scor

e; CI: confidence int

er

val

.

Sensitivit

y analysis af

ter single linear imputation of missing values in all patients completing the f

ollo

w-up per

iod

(15)

In patients with complete mSASSS data over 8 years of follow-up, GEE analysis also revealed that there was a non-linear relationship between spinal radiographic damage and time (Supplementary Figure S1C). The mean estimated status and progression scores of all the non-linear time models showed a reduction in spinal radiographic progression over time (Supplementary Table S3). Based on the estimated parameters of the non-linear time model with the best fit for the data, the mean progression rates were estimated to be 2.3, 1.4, 1.0 and 0.8 during 0-2, 2-4, 4-6, and 6-8 years, respectively (Table 4, Figure 2C).

Sensitivity analysis after single linear imputing of missing radiographic data, assuming an ongoing linear course during the imputed time interval, revealed similar patterns in the course of spinal radiographic progression; the course was linear in patients with 4 years of follow-up and non-linear with reducing progression rates in patients with 6 and 8 years of follow-up (Table 4 and Supplementary Table S4).

Importantly, all these time models remained statistically significant after adjustment for patient characteristics with possible influence on radiographic progression, including baseline mSASSS, presence of syndesmophytes, gender, HLA-B27 status, age, symptom duration, smoking duration, BMI, disease activity, and NSAID use over time (data not shown).

DISCUSSION

This prospective standardized longitudinal observational cohort study showed overall slow spinal radiographic progression during long-term treatment with TNF-α inhibitors in daily clinical practice. During the first 4 years of follow-up, spinal radiographic progression followed a linear course according to GEE modeling. Interestingly, a deflection from a linear course was found in patients with 6 years and 8 years of follow-up. Both complete case analyses and sensitivity analyses (after single linear imputation and after adjustment for patient characteristics that could affect radiographic progression) clearly showed the same pattern; a linear course with stable progression rates in patients with 4 years of follow-up and a non-linear course with reducing progression rates over time in patients with 6 and 8 years of follow-up.

These results may refer to a delayed effect of TNF-α inhibitors on radiographic progression and support the TNF brake hypothesis. Presence of already triggered repair processes can first

(16)

3

lead to continuation of bone formation. Subsequently, long-term inhibition of inflammation

by TNF-α inhibitors may result in a reduction of new bone formation over time [25,26]. Our results demonstrate that more than 4 years of follow-up seems necessary to show any potential effect of TNF-α blocking therapy on the inhibition of spinal radiographic progression in AS. The results are consistent with previous findings. In a retrospective study of 54 AS patients, a significantly lower spinal radiographic progression rate during 4 to 8 years of follow-up was found in patients treated with TNF-α inhibitor compared to patients from a historical cohort, whereas there was no difference during the first 4 years [9]. A prospective study in 334 AS patients found that TNF-α inhibitor use was significantly associated with less spinal radiographic progression, especially in patients with a follow-up duration of more than 3.9 years [10]. However, in both studies TNF-α naïve patients showed a remarkably high radiographic progression rate after 4 years. This may be inflicted by selection or completion bias caused by missing data or losses to follow-up.

As demonstrated in a historical cohort of which the majority of patients were not treated with TNF-α inhibitors over a period of 12 years, spinal radiographic progression was linear at the group level, but highly variable between individual patients [11]. This emphasizes that in this heterogeneous and slowly progressive disease, differences in patient numbers and characteristics within studies can have a high impact on the radiographic progression rates at the group level. Therefore, potential bias that influences radiographic progression should be taken into account during the analysis and interpretation of radiographic data in AS. In the present study, we analyzed the course of spinal radiographic progression in three different groups; in patients with 4, 6 and 8 years of follow-up. This was done to exclude potential selection bias caused by patient characteristics of those with longer follow-up. In addition, we performed complete case analyses and sensitivity analyses after single linear imputation of missing radiographic data. These methods are recommended to avoid potential bias due to differences in patient numbers and characteristics at the different follow-up time points [13]. Although we assumed ongoing linear progression during the imputed time interval, GEE after single linear imputation revealed similar patterns as complete case analysis; a reduction of spinal radiographic progression over time. In addition, adjustments for patient characteristics that could affect radiographic progression did not result in different patterns in the course of spinal radiographic progression. This strengthens the results found in complete case analysis.

(17)

In an earlier analysis of our cohort in which patients with 2 to 6 years of follow-up (mean 3.8 years) were included, we could not find a reduction in spinal radiographic progression

[8]. This can be explained by the different methodology. All patients were included in one analysis, irrespective of complete data, and radiographs were scored with unknown time sequence leading to negative progression rates and a higher SDC. Most important, fewer patients had reached 6 years of follow-up resulting in less power to explore the course of radiographic progression over 6 years. In the present study, all radiographs were scored in chronological time order after new training sessions in which consensus about abnormalities due to the disease process was refreshed. Scoring in chronological time order is considered as the most sensitive method to score spinal radiographic progression and avoids negative progression rates in cohort studies [27]. Potential bias caused by knowledge of the readers that all radiographs originated from patients treated with TNF-α inhibitors was ruled out by randomization and scoring the radiographs together with radiographs of patients from a historical inception cohort not treated with TNF-α inhibitors.

The prospective longitudinal observational design of our study allowed longitudinal modeling of data obtained from daily clinical practice. With GEE, different time functions were modeled and the results of the significant time model with the best fit for the data, based on the smallest QICC, were presented. The QICC is not an absolute value but it can be used to compare the goodness of fit of different models [21]. It should be noted that it was not the aim of our study to estimate the most suitable and robust regression function (e.g. exponential or logarithmic) to predict spinal radiographic outcome in AS patients treated with TNF-α inhibitors. Different time models were used to explore whether the course of spinal radiographic progression was linear or non-linear. All non-linear models in patients with 6 and 8 years data revealed the same pattern, i.e. a reduction in spinal radiographic progression. The QICC values were rather similar. Therefore, it can be concluded that a non-linear time model is a valid representation of the course of spinal radiographic progression in AS patients with 6 and 8 years of follow-up treated with TNF-α inhibitors.

Previously, an association between continuous NSAID use and a reduction of spinal radiographic progression has been found in AS. Less radiographic progression with continuous NSAID therapy was most pronounced in patients with elevated CRP levels

[24,28]. Very recently, the Effects of NSAIDs on RAdiographic Damage in Ankylosing Spondylitis (ENRADAS) trial investigated primarily the effect of continuous versus on-demand use of NSAIDS on 2-year radiographic progression [29]. This RCT could not confirm

(18)

3

the earlier findings. In our cohort, the use of NSAID decreased rapidly over time and patients

discontinued their NSAIDs due to the good clinical effect of TNF-α inhibitors on disease activity, including CRP. This resulted in very low ASAS-NSAID scores over time. Therefore, influence of NSAIDs on spinal radiographic progression could not be observed in our study. The content of this manuscript underlines that it is challenging to evaluate spinal radiographic progression in AS. Well-organized large prospective cohorts with radiographic assessment every 2 years during long-term follow-up are very valuable in the field of AS. For the evaluation of spinal radiographic progression in this slowly and very heterogeneous disease, we recommend to take the following aspects into account: 1) radiographs should be scored in chronological time order together with radiographs of patient not treated with TNF-α inhibitors in order to reduce the measurement error and to exclude reader bias, 2) the smallest detectable change should be noted, 3) a flow chart should be presented of included patients, drop-outs, missing data of the primary outcome and differences in patients characteristics over time, 4) analyses should be stratified for different follow-up periods when patient characteristics differ, 5) complete case analysis and sensitivity analysis after imputation of missing values should be performed to diminish completion bias, 6) analyses should be adjusted for patient characteristics with possible influence on radiographic progression.

CONCLUSIONS

Our prospective observational cohort study of AS patients with longstanding disease treated with TNF-α inhibitors showed overall slow and linear spinal radiographic progression during the first 4 years of follow-up. A deflection from a linear course in radiographic progression was observed in patients with 6 and 8 years of follow-up. These findings suggests that long-term inhibition of inflammation with TNF-α inhibitors diminishes new bone formation over time in AS patients with longstanding disease.

(19)

KEY MESSAGES

••

It is important to take into account potential bias that affects radiographic progression in the analysis and interpretation of radiographic data in AS.

••

Spinal radiographic progression followed a linear course during the first 4 years of follow-up in AS patients with longstanding disease treated with TNF-α inhibitors.

••

A deflection from a linear course was observed in AS patients with 6 and 8 years of follow-up which may refer to a delayed effect of TNF-α inhibitors.

(20)

3

1. van der Heijde D, Calin A, Dougados M, Khan

MA, van der Linden S, Bellamy N. Selection of instruments in the core set for DC-ART, SMARD, physical therapy, and clinical record keeping in ankylosing spondylitis. Progress report of the ASAS Working Group. Assessments in Ankylosing Spondylitis. J Rheumatol 1999;26:951-4.

2. Machado P, Landewé R, Braun J, Hermann KG, Baraliakos X, Baker D, et al. A stratified model for health outcomes in ankylosing spondylitis. Ann Rheum Dis 2011;70:1758-64.

3. Baraliakos X, Listing J, Rudwaleit M, Brandt J, Sieper J, Braun J. Radiographic progression in patients with ankylosing spondylitis after 2 years of treatment with the tumour necrosis factor alpha antibody infliximab. Ann Rheum Dis 2005;64:1462–6.

4. van der Heijde D, Landewe R, Baraliakos X, Houben H, van Tubergen A, Williamson P, et al. Radiographic findings following two years of infliximab therapy in patients with ankylosing spondylitis. Arthritis Rheum 2008;58: 3063–70. 5. van der Heijde D, Landewe R, Einstein S, Ory P,

Vosse D, Ni L, et al. Radiographic progression of ankylosing spondylitis after up to two years of treatment with etanercept. Arthritis Rheum 2008;58:1324–31.

6. van der Heijde D, Salonen D, Weissman BN, Landewé R, Maksymowych WP, Kupper H, et al. Assessment of radiographic progression in the spines of patients with ankylosing spondylitis treated with adalimumab for up to 2 years. Arthritis Res Ther 2009;11:R127.

7. Baraliakos X, Listing J, Brandt J, Haibel H, Rudwaleit M, Sieper J, et al. Radiographic progression in patients with ankylosing spondylitis after 4 yrs of treatment with the anti-TNF-alpha antibody infliximab. Rheumatology (Oxford) 2007;46:1450–3.

8. Maas F, Spoorenberg A, Brouwer E, Bos R, Efde M, Chaudhry RN, et al. Spinal radiographic progression in patients with ankylosing spondylitis treated with TNF-α blocking therapy: a prospective longitudinal observational cohort study. PLoS One 2015;10:e0122693.

9. Baraliakos X, Haibel H, Listing J, Sieper J, Braun J. Continuous long-term anti-TNF therapy does not lead to an increase in the rate of new bone formation over 8 years in patients with ankylosing spondylitis. Ann Rheum Dis 2013;27:1-6.

10. Haroon N, Inman RD, Learch TJ, Weisman MH, Lee M, Rahbar MH, et al. The impact of tumor necrosis factor α inhibitors on radiographic progression in ankylosing spondylitis. Arthritis Rheum 2013;65:2645-54.

11. Ramiro S, Stolwijk C, van Tubergen A, van der Heijde D, Dougados M, van den Bosch F, et al. Evolution of radiographic damage in ankylosing spondylitis: a 12 year prospective follow-up of the OASIS study. Ann Rheum Dis 2015;74:52-9. 12. Baraliakos X, Listing J, von der Recke A, Braun J.

The natural course of radiographic progression in ankylosing spondylitis--evidence for major individual variations in a large proportion of patients. J Rheumatol 2009;36:997-1002. 13. Machado P. Anti-tumor necrosis factor and new

bone formation in ankylosingn spondylitis: the controversy continues. Arthritis Rheum 2013;65:2537-40.

14. Arends S, Brouwer E, van der Veer E, Groen H, Leijsma MK, Houtman PM, et al. Baseline predictors of response and discontinuation of tumor necrosis factor-alpha blocking therapy in ankylosing spondylitis: a prospective longitudinal observational cohort study. Arthritis Res Ther 2011;13:R94.

(21)

15. Arends S, Spoorenberg A, Houtman PM, Leijsma MK, Bos R, Kallenberg CG, et al. The effect of three years of TNFalpha blocking therapy on markers of bone turnover and their predictive value for treatment discontinuation in patients with ankylosing spondylitis: a prospective longitudinal observational cohort study. Arthritis Res Ther 2012;14:R98.

16. Braun J, Davis J, Dougados M, Sieper J, van der Linden S, van der Heijde D; ASAS Working Group. First update of the international ASAS consensus statement for the use of anti-TNF agents in patients with ankylosing spondylitis. Ann Rheum Dis 2006;65:316-20.

17. Dougados M, Simon P, Braun J, Burgos-Vargas R, Maksymowych WP, Sieper J, et al. ASAS recommendations for collecting, analysing and reporting NSAID intake in clinical trials/epidemiological studies in axial spondyloarthritis. Ann Rheum Dis 2011;70:249-51.

18. Spoorenberg A, de Vlam K, van der Linden S, Dougados M, Mielants H, van de Tempel H, et al.Radiological scoring methods in ankylosing spondylitis. Reliability and change over 1 and 2 years. J Rheumatol 2004;31:125-32.

19. Wanders AJ, Landewe RB, Spoorenberg A, Dougados M, van der Linden S, Mielants H, et al. What is the most appropriate radiologic scoring method for ankylosing spondylitis? A comparison of the available methods based on the Outcome Measures in Rheumatology Clinical Trials filter. Arthritis Rheum 2004;50:2622-32.

20. Bruynesteyn K, Boers M, Kostense P, van der Linden S, van der Heijde D. Deciding on progression of joint damage in paired films of individual patients: smallest detectable difference or change. Ann Rheum Dis 2005;64:179-82.

21. Cui J. QIC program and model selection in GEE analyses. The Stata Journal 2007;7:209-20. 22. Arends S, Spoorenberg A, Brouwer E, van

der Veer E. Clinical studies on bone-related outcome and the effect of TNF-α blocking therapy in ankylosing spondylitis. Curr Opin Rheumatol 2014;26:259-68.

23. Ramiro S, van der Heijde D, van Tubergen A, Stolwijk C, Dougados M, van den Bosch F, Landewé R. Higher disease activity leads to more structural damage in the spine in ankylosing spondylitis: 12-year longitudinal data from the OASIS cohort. Ann Rheum Dis 2014;73:1455-61.

24. Kroon Landewé R, Dougados M, van der Heijde D. Continuous NSAID use reverts the effects of inflammation on radiographic progression in patients with ankylosing spondylitis. Ann Rheum Dis 2012;71:1623-9.

25. Sieper J, Appel H, Braun J, Rudwaleit M. Critical appraisal of assessment of structural damage in ankylosing spondylitis: implications for treatment outcomes. Arthritis Rheum 2008;58:649-56.

26. Maksymowych WP, Chiowchanwisawakit P, Clare T, Pedersen SJ, Østergaard M, Lambert RG. Inflammatory lesions of the spine on magnetic resonance imaging predict the development of new syndesmophytes in ankylosing spondylitis: evidence of a relationship between inflammation and new bone formation. Arthritis Rheum 2009;60:93-102

27. van Tuyl LH, van der Heijde D, Knol DL, Boers M. Chronological reading of radiographs in rheumatoid arthritis increases efficiency and does not lead to bias. Ann Rheum Dis 2014;73:391-5.

28. Poddubnyy D, Rudwaleit M, Haibel H, Listing J, Märker-Hermann E, Zeidler H, et al. Effect of non-steroidal anti-inflammatory drugs on radiographic spinal progression in patients with axial spondyloarthritis: results from the GErman SPondyloarthritis Inception Cohort (GESPIC). Ann Rheum Dis 2012;71:1616-22.

29. Sieper J, Listing J, Poddubnyy D, Song IH, Hermann KG, Callhoff J, et al. Effect of continuous versus on-demand treatment of ankylosing spondylitis with diclofenac over 2 years on radiographic progression of the spine: results from a randomised multicentre trial (ENRADAS). Ann Rheum Dis 2016;75:1438-43.

(22)

3 SUPPLEMENT

AR

Y FILES

Supplemen tar y T able S1. Baseline charac ter istics

of AS patients who did or did not (y

et) r eached 4, 6, and 8 y ears of f ollo w-up . 4 y ears 6 y ears 8 y ears Ye s n=160 No n=50 Ye s n=98 No n=112 Ye s n=45 No n=165 M ale gender 117 (73) 28 (56)* 72 (74) 73 (65) 36 (80) 109 (66) Age (yrs) 41.2 ± 10.7 42.0 ± 13.9 41.2 ± 10.1 41.9 ± 12.7 41.2 ± 8.9 41.7 ± 12.2 Sympt om dur ation (yrs) 14 (7-23) 15 (9-30) 15 (7-23) 14 (8-25) 18 (9-24) 13 (7-23) Time sinc e diag nosis (yrs) 5 (1-14) 7 (1-20) 7 (1-15) 5 (1-15) 10 (3-18) 5 (1-13) HLA -B27+ 125 (79) 35 (78) 77 (78) 83 (78) 36 (80) 124 (78) BMI (kg/m 2) 26.3 ± 3.8 24.6 ± 5.3* 25.9 ± 3.8 25.8 ± 4.6 25.4 ± 3.5 25.9 ± 4.4 Smok ing dur ation (yrs) 12 (0-23) 5 (0-21) 14 (0-25) 10 (0-20) 15 (0-24) 10 (0-22) NSAID use 130 (82) 34 (72) 85 (88) 79 (73)* 43 (96) 121 (76)* ASAS -NSAID inde x 60 (25-100) 58 (16-100) 54 (25-100) 67 (0-100) 50 (42-100) 67 (4-100) DM ARD use 31 (19) 7 (14) 23 (24) 15 (13) 11 (24) 27 (16) BASD AI (0-10) 6.1 ± 1.6 6.0 ± 1.7 5.9 ± 1.7 6.1 ± 1.6 6.1 ± 1.6 6.0 ± 1.7 ASD ASCRP 3.8 ± 0.8 3.6 ± 1.0 3.8 ± 0.8 3.6 ± 0.9 4.0 ± 0.7 3.6 ± 0.8* CRP (mg/L) 13 (5-22) 11 (3-19) 15 (7-23) 11 (3-21)* 16 (9-31) 11 (3-21)* ESR (mm/hr) 20 (9-34) 18 (8-34) 21 (12-35) 18 (8-34) 24 (11-36) 20 (10-34) Pa tien t’s GD A (0-10) 7 (5-8) 7 (5-8) 7 (5-8) 7 (6-8) 7 (5-8) 7 (6-8) mSASSS (r ange 0-72) mean 10.7 ± 15.8 7.5 ± 14.4 10.8 ± 15.2 9.3 ± 15.8 13.6 ± 16.2 9.0 ± 15.2* median 3.5 (0.0-14.9) 2.1 (0.0-6.5) 4.0 (0.0-16.8) 2.2 (0.0-11.3) 6.0 (1.0-20.8) 2.4 (0.0-10.8)* ≥1 syndesmoph yt e 88 (56) 20 (46) 56 (57) 52 (51) 30 (67) 78 (50)* Values ar e pr esent

ed as number of patients (%), mean ± SD

, or median (IQR). *p≤0.05 compar ed t o patients who r eached the f ollo w-up time . Abbr eviations: AS: ank

ylosing spondylitis; FU: f

ollo

w-up; HLA: human leuk

oc

yt

e antigen; BMI: body mass index; NSAID: non-st

er oidal anti-inflammat or y drug; ASAS: A ssessment of SpondyloAr thr itis int er national S ociet y; DM ARD: disease -modifying antir

heumatic drug; BASD

AI: Bath AS Disease A

ctivit y I ndex; ASD AS: AS Disease A ctivit y S cor e; CRP : C-r eac tiv e pr ot ein; ESR: er ythr oc yt e sedimentation rat e; GD A: global disease ac tivit y; mSASSS: modified St ok e AS Spine S cor e.

(23)

of AS patients with complet

e or missing radiog

raphic data dur

ing 4, 6, and 8 y ears of f ollo w-up . 4 y ears FU (n=160+4 lost t o FU) 6 y ears FU (n=98+9 lost t o FU) 8 y ears FU (n=45+2 lost t o FU) Complet e r adiog raphic da ta Complet e r adiog raphic da ta Complet e r adiog raphic da ta Ye s n=110 No n=54 Ye s n=53 No n=54 Ye s n=19 No n=28 M ale gender 79 (72) 40 (74) 38 (72) 41 (76) 16 (84) 22 (79) Age (yrs) 41.2 ± 11.7 41.4 ± 9.1 39.7 ± 11.1 42.7 ± 8.4 39.3 ± 10.7 43.5 ± 7.7* Sympt om dur ation (yrs) 15 (7-23) 13 (8-24) 15 (7-21) 15 (8-24) 17 (9-23) 19 (10-27) Time sinc e diag nosis (yrs) 5 (1-14) 5 (1-14) 5 (1-14) 8 (2-17) 9 (3-17) 12 (3-18) HLA -B27+ 88 (80) 41 (77) 43 (81) 42 (79) 17 (90) 20 (71) BMI (kg/m 2) 26.2 ± 3.9 26.7 ± 3.6 25.3 ± 3.8 26.3 ± 3.7 24.9 ± 2.0 25.7 ± 4.2 Smok ing dur ation (yrs) 13 (0-24) 12 (0-22) 13 (0-26) 12 (0-24) 13 (0-27) 16 (2-24) NSAID use 89 (81) 43 (84) 45 (85) 108 (81) 18 (95) 135 (81) ASAS -NSAID inde x 69 (17-100) 50 (25-100) 67 (38-100) 50 (25-100) 50 (25-100) 67 (50-100) DM ARD use 21 (19) 11 (20) 16 (30) 8 (15) 5 (26) 6 (21) BASD AI (0-10) 6.0 ± 1.6 6.2 ± 1.7 5.8 ± 1.7 6.1 ± 1.7 6.0 ± 1.4 6.1 ± 1.7 ASD ASCRP 3.7 ± 0.7 3.9 ± 0.8 3.8 ± 0.8 3.8 ± 0.8 4.0 ± 0.6 3.9 ± 0.7 CRP (mg/L) 12 (4-22) 14 (7-23) 14 (7-25) 15 (6-22) 17 (12-40) 16 (7-22) ESR (mm/hr) 20 (9-34) 23 (13-36) 21 (11-34) 21 (12-36) 21 (11-37) 24 (9-36) Pa tien t’s GD A (0-10) 7 (6-8) 7 (5-8) 7 (5-8) 7 (5-8) 7 (5-8) 8 (4-7) mSASSS (r ange 0-72) mean 10.7 ± 16.0 10.2 ± 15.0 8.2 ± 12.9 13.8 ± 16.9 10.0 ± 12.9 18.1 ± 10.4 median 3.6 (0.0-15.8) 3.0 (0.0-13.5) 3.7 (0.0-11.4) 6.0 (0.9-27.4) 5.4 (1.0-17.1) 10.4 (1.0-35.8) ≥1 syndesmoph yt e 60 (55) 30 (59) 28 (53) 35 (65) 12 (63) 20 (71) Values ar e pr esent

, or median (IQR).

*p≤0.05 compar

ed t

o patients with complet

e radiog

raphic data.

Abbr

eviations:

AS: ank

ollo

oc

yt

heumatic drug; BASD

ctivit y I ndex; ASD AS: AS Disease A ctivit y S cor e; CRP : C-r eac tiv e pr ot ein; ESR: er ythr oc yt e sedimentation rat e; mSASSS: GD A: global disease ac tivit y; modified St ok e AS Spine S cor e.

(24)

3

of AS patients with complet

e or missing radiog

raphic data dur

ing 4, 6, and 8 y ears of f ollo w-up . 4 y ears FU (n=160+4 lost t o FU) 6 y ears FU (n=98+9 lost t o FU) 8 y ears FU (n=45+2 lost t o FU) Complet e r adiog raphic da ta Complet e r adiog raphic da ta Complet e r adiog raphic da ta Ye s n=110 No n=54 Ye s n=53 No n=54 Ye s n=19 No n=28 M ale gender 79 (72) 40 (74) 38 (72) 41 (76) 16 (84) 22 (79) Age (yrs) 41.2 ± 11.7 41.4 ± 9.1 39.7 ± 11.1 42.7 ± 8.4 39.3 ± 10.7 43.5 ± 7.7* Sympt om dur ation (yrs) 15 (7-23) 13 (8-24) 15 (7-21) 15 (8-24) 17 (9-23) 19 (10-27) Time sinc e diag nosis (yrs) 5 (1-14) 5 (1-14) 5 (1-14) 8 (2-17) 9 (3-17) 12 (3-18) HLA -B27+ 88 (80) 41 (77) 43 (81) 42 (79) 17 (90) 20 (71) BMI (kg/m 2) 26.2 ± 3.9 26.7 ± 3.6 25.3 ± 3.8 26.3 ± 3.7 24.9 ± 2.0 25.7 ± 4.2 Smok ing dur ation (yrs) 13 (0-24) 12 (0-22) 13 (0-26) 12 (0-24) 13 (0-27) 16 (2-24) NSAID use 89 (81) 43 (84) 45 (85) 108 (81) 18 (95) 135 (81) ASAS -NSAID inde x 69 (17-100) 50 (25-100) 67 (38-100) 50 (25-100) 50 (25-100) 67 (50-100) DM ARD use 21 (19) 11 (20) 16 (30) 8 (15) 5 (26) 6 (21) BASD AI (0-10) 6.0 ± 1.6 6.2 ± 1.7 5.8 ± 1.7 6.1 ± 1.7 6.0 ± 1.4 6.1 ± 1.7 ASD ASCRP 3.7 ± 0.7 3.9 ± 0.8 3.8 ± 0.8 3.8 ± 0.8 4.0 ± 0.6 3.9 ± 0.7 CRP (mg/L) 12 (4-22) 14 (7-23) 14 (7-25) 15 (6-22) 17 (12-40) 16 (7-22) ESR (mm/hr) 20 (9-34) 23 (13-36) 21 (11-34) 21 (12-36) 21 (11-37) 24 (9-36) Pa tien t’s GD A (0-10) 7 (6-8) 7 (5-8) 7 (5-8) 7 (5-8) 7 (5-8) 8 (4-7) mSASSS (r ange 0-72) mean 10.7 ± 16.0 10.2 ± 15.0 8.2 ± 12.9 13.8 ± 16.9 10.0 ± 12.9 18.1 ± 10.4 median 3.6 (0.0-15.8) 3.0 (0.0-13.5) 3.7 (0.0-11.4) 6.0 (0.9-27.4) 5.4 (1.0-17.1) 10.4 (1.0-35.8) ≥1 syndesmoph yt e 60 (55) 30 (59) 28 (53) 35 (65) 12 (63) 20 (71) Values ar e pr esent

, or median (IQR).

*p≤0.05 compar

ed t

o patients with complet

e radiog

raphic data.

Abbr

eviations:

AS: ank

ollo

oc

yt

heumatic drug; BASD

ctivit y I ndex; ASD AS: AS Disease A ctivit y S cor e; CRP : C-r eac tiv e pr ot ein; ESR: er ythr oc yt e sedimentation rat e; mSASSS: GD A: global disease ac tivit y; modified St ok e AS Spine S cor e. Supplemen tar y T able S3. Obser

ved and estimat

ed mean mSASSS o

ver time of diff

er

ent time models in patients with complet

e radiog raphic data o ver 4, 6, and 8 y ears of f ollo w-up . Obser ved da ta

GEE time models

Linear Q uadr atic Cubic Squar e r oot Logarithmic Exponen tial Sta tus sc or es Pr og ression sc or es Sta tus sc or es Pr og ression sc or es Sta tus sc or es Pr og ression sc or es Sta tus sc or es Pr og ression sc or es Sta tus sc or es Pr og ression sc or es Sta tus sc or es Pr og ression sc or es Sta tus sc or es Pr og ression sc or es Complet e 4 yr da ta (n=110) 0yr 10.7 10.7 10.7 – 10.7 10.7 10.7 2yr 12.5 1.8 12.4 1.7 12.5 1.8 – – 12.5 1.7 12.5 1.8 12.5 1.8 4yr 14.0 1.5 14.1 1.7 14.0 1.5 – – 14.0 1.5 14.0 1.5 14.0 1.5 p-value <0.001 0.276 – 0.276 0.276 0.276 QIC C 94898 94899 – 94899 94899 94899 Complet e 6 yr da ta (n=53) 0yr 8.2 8.3 8.2 8.2 8.2 8.2 8.2 2yr 9.8 1.6 9.8 1.5 10.0 1.8 9.8 1.6 10.0 1.8 10.0 1.8 9.9 1.7 4yr 11.6 1.8 11.3 1.5 11.4 1.4 11.6 1.8 11.4 1.4 11.4 1.4 11.5 1.6 6yr 12.6 1.0 12.8 1.5 12.6 1.2 12.5 0.9 12.6 1.2 12.6 1.2 12.6 1.1 p-value <0.001 0.063 0.075 0.104 0.121 0.028 QIC C 45373 45370 45369 45371 45371 45368 Complet e 8 yr da ta (n=19) 0yr 10.0 10.4 10.0 9.9 9.9 9.9 10.1 2yr 12.2 2.2 11.8 1.4 12.1 2.1 12.2 2.3 12.2 2.3 12.2 2.3 11.9 1.8 4yr 13.7 1.5 13.2 1.4 13.7 1.6 13.7 1.5 13.7 1.5 13.7 1.4 13.6 1.7 6yr 14.6 0.9 14.6 1.4 14.8 1.1 14.7 1.0 14.7 1.0 14.7 1.0 15.0 1.4 8yr 15.5 0.9 16.0 1.4 15.4 0.6 15.5 0.8 15.5 0.8 15.5 0.8 15.5 0.5 p-value <0.001 0.067 0.159 0.052 0.048 0.104 Q ICCC 21215 21202 21203 21201 21201 21205

Values in bold indicat

e the model with p<0.05 (

W ald t est) and lo w est QIC C. Abbr eviations: QIC C: cor rec ted quasi-lik elihood inf or mation cr iter ion.

(25)

Supplemen tar y T able S4. Obser

ved and estimat

ed mean mSASSS o

ver time of diff

er

ent time models in the t

otal g

roup of patients with complet

e or imput ed radiog raphic data o ver 4, 6, and 8 y ears of f ollo w-up . Obser ved da ta

GEE time models

Linear Q uadr atic Cubic Squar e r oot Logarithmic Exponen tial Sta tus sc or es Pr og ression sc or es Sta tus sc or es Pr og ression sc or es Sta tus sc or es Pr og ression sc or es Sta tus sc or es Pr og ression sc or es Sta tus sc or es Pr og ression sc or es Sta tus sc or es Pr og ression sc or es Sta tus sc or es Pr og ression sc or es Total g

roup with 4yr imput

ed da ta (n=132) 0yr 11.5 11.5 11.5 – 11.5 11.5 11.5 2yr 13.4 1.9 13.3 1.7 13.4 1.9 – – 13.4 1.9 13.4 1.9 13.4 1.9 4yr 15.0 1.6 15.0 1.7 15.0 1.6 – – 15.0 1.6 15.0 1.6 15.0 1.6 p-value <0.001 0.270 – 0.270 0.270 0.270 QIC C 120298 120295 – 120295 120295 120295 Total g

roup with 6yr imput

ed da ta (n=80) 0yr 8.7 8.8 8.7 8.7 8.7 8.7 8.7 2yr 10.2 1.5 10.2 1.4 10.4 1.7 10.3 1.6 10.4 1.7 10.4 1.7 10.4 1.7 4yr 11.9 1.7 11.7 1.4 11.8 1.4 11.9 1.6 11.8 1.4 11.8 1.4 11.9 1.5 6yr 12.9 1.0 13.1 1.4 12.9 1.1 12.9 1.0 13.0 1.2 13.0 1.2 12.9 1.0 p-value <0.001 0.024 0.177 0.043 0.051 0.009 QIC C 71889 71883 71883 71884 71884 71882 Total g

roup with 8yr imput

ed da ta (n=41) 0yr 13.5 14.1 13.4 13.4 13.4 13.4 13.6 2yr 16.0 2.5 15.9 1.8 16.2 2.8 16.2 2.8 16.3 2.9 16.4 3.0 16.0 2.4 4yr 18.5 2.5 17.6 1.8 18.3 2.1 18.3 2.1 18.2 1.9 18.2 1.8 18.2 2.2 6yr 19.5 1.0 19.4 1.8 19.7 1.4 19.7 1.4 19.5 1.3 19.5 1.3 19.9 1.7 8yr 20.5 0.9 21.1 1.8 20.5 0.8 20.4 0.7 20.5 1.0 20.6 1.1 20.4 0.5 p-value <0.001 0.001 0.868 0.001 0.001 0.001 QIC C 71201 71140 71142 71144 71146 71146

(26)

3

Supplemen tar y T able S4. Obser

ved and estimat

ed mean mSASSS o

ver time of diff

er

ent time models in the t

otal g

roup of patients with complet

e or imput ed radiog raphic data o ver 4, 6, and 8 y ears of f ollo w-up . Obser ved da ta

GEE time models

Linear Q uadr atic Cubic Squar e r oot Logarithmic Exponen tial Sta tus sc or es Pr og ression sc or es Sta tus sc or es Pr og ression sc or es Sta tus sc or es Pr og ression sc or es Sta tus sc or es Pr og ression sc or es Sta tus sc or es Pr og ression sc or es Sta tus sc or es Pr og ression sc or es Sta tus sc or es Pr og ression sc or es Total g

roup with 4yr imput

ed da ta (n=132) 0yr 11.5 11.5 11.5 – 11.5 11.5 11.5 2yr 13.4 1.9 13.3 1.7 13.4 1.9 – – 13.4 1.9 13.4 1.9 13.4 1.9 4yr 15.0 1.6 15.0 1.7 15.0 1.6 – – 15.0 1.6 15.0 1.6 15.0 1.6 p-value <0.001 0.270 – 0.270 0.270 0.270 QIC C 120298 120295 – 120295 120295 120295 Total g

roup with 6yr imput

ed da ta (n=80) 0yr 8.7 8.8 8.7 8.7 8.7 8.7 8.7 2yr 10.2 1.5 10.2 1.4 10.4 1.7 10.3 1.6 10.4 1.7 10.4 1.7 10.4 1.7 4yr 11.9 1.7 11.7 1.4 11.8 1.4 11.9 1.6 11.8 1.4 11.8 1.4 11.9 1.5 6yr 12.9 1.0 13.1 1.4 12.9 1.1 12.9 1.0 13.0 1.2 13.0 1.2 12.9 1.0 p-value <0.001 0.024 0.177 0.043 0.051 0.009 QIC C 71889 71883 71883 71884 71884 71882 Total g

roup with 8yr imput

ed da ta (n=41) 0yr 13.5 14.1 13.4 13.4 13.4 13.4 13.6 2yr 16.0 2.5 15.9 1.8 16.2 2.8 16.2 2.8 16.3 2.9 16.4 3.0 16.0 2.4 4yr 18.5 2.5 17.6 1.8 18.3 2.1 18.3 2.1 18.2 1.9 18.2 1.8 18.2 2.2 6yr 19.5 1.0 19.4 1.8 19.7 1.4 19.7 1.4 19.5 1.3 19.5 1.3 19.9 1.7 8yr 20.5 0.9 21.1 1.8 20.5 0.8 20.4 0.7 20.5 1.0 20.6 1.1 20.4 0.5 p-value <0.001 0.001 0.868 0.001 0.001 0.001 QIC C 71201 71140 71142 71144 71146 71146

Supplementary Figure S1. Estimated course of spinal radiographic progression according to GEE

modeling in patients with complete radiographic data over 4 years (A, n=110), 6 years (B, n=53), and 8 years (C, n=19) of follow-up.

Supplementary Figure S2. Estimated course of spinal radiographic progression according to GEE

modeling in patients with complete or imputed radiographic data over 4 years (A, n=132), 6 years (B, n=80), and 8 years (C, n=41).

(27)