Computed tomography-based assessment of radiographic progression in spine and sacroiliac joints after pregnancy in women with ankylosing spondylitis

We aimed to investigate whether pregnancy affects radiographic progression in patients with ankylosing spondylitis (AS) based on computed tomography (CT) evaluations. This retrospective study included women with AS aged 19–49 years who underwent at least two CT evaluations of the whole spine or sacroiliac joints (SIJs) at intervals of 2–4 years. The delivery group was restricted to women who had the rst CT ~ 2 years before delivery and the second CT ~ 2 years after delivery. The CT Syndesmophyte Score (CTSS) (0-522) and SIJ scores (0–40) were used to radiographic progression. A total of 21 women in the delivery group and 38 women in the controls were included. The median (Q1-Q3) CTSS at baseline in the delivery group and controls was 19 (16–23) and 20 (13.25–27.75), and the median progression was 1 (0–3) and 0 (0–1) during the median 2.9-year follow-up, respectively. The median SIJ score at baseline in the delivery group and controls was 13 (8–22) and 11 (6–22), and the median progression was 1.5 (0–3) and 1 (0–2), respectively. The CTSS and SIJ scores signicantly increased in both groups; however, no difference in absolute score changes per time-point was observed. Pregnancy do not affect the radiographic progression in AS.


Introduction
Ankylosing spondylitis (AS) is a chronic in ammatory arthritis affecting the sacroiliac joints (SIJs) and the spine, commonly occurring in childbearing age 1 . Although AS was previously believed to predominantly affect men, recent studies have shown that the incidence of AS in women is as high as one in every three women 2 . Pregnancy is a major issue faced by female AS patients of childbearing age.
Mechanical stress is one of the causes of AS 3 . During pregnancy, the mechanical load on the spine and pelvis increases because of the gravid uterus. During pregnancy and after childbirth, almost half of pregnant women experience pain in the lumbar and/or sacroiliac area and/or in the symphysis pubis 4 .
More women with AS discontinue medications during pregnancy than women with systemic lupus erythematosus and rheumatoid arthritis 5 , and tumor necrosis factor inhibitor (TNFI) discontinuation in early pregnancy has been identi ed as a risk factor for disease ares during pregnancy 6 . Previous studies on the effects of pregnancy in patients with AS have focused on disease activity during pregnancy, pregnancy outcomes (e.g., rates of cesarean section [CS], preterm birth, and preeclampsia), and fetal outcomes (e.g., low birth weight and congenital abnormalities) 6-8 . However, to our knowledge, the possible in uence of pregnancy on radiographic progression in patients with AS has not yet been documented.
Recently, the Computed Tomography (CT) Syndesmophyte Score (CTSS) was developed to analyze spinal damage in patients with AS 9 . Although the modi ed Stoke Ankylosing Spondylitis Spine Score (mSASSS) includes the evaluation of lateral views of the cervical and lumbar spine segments, the CTSS has the advantage of evaluating bone proliferation in the entire spine, including the thoracic spine, at four quadrants (anterior, posterior, left, and right) with excellent reliability 10, 11 . With respect to imaging evaluation of the SIJs, CT has higher sensitivity and reliability than conventional radiography in detecting sacroiliitis 12,13 .
In this study, we aimed to investigate whether pregnancy and delivery affect the radiographic progression of the spine and SIJs in women with AS based on CT evaluations. To compare radiographic progression after delivery, we classi ed the patients into two groups: delivery group and control group. The delivery group was restricted to women who had the rst CT evaluation at a maximum of 2 years before delivery and the second CT evaluation at a maximum of 2 years after delivery. Controls were de ned as women who had two CT evaluations with 2-4-year intervals.

Methods
Demographic characteristics (age and smoking), clinical data (disease duration and extra-articular manifestations), method of delivery (vaginal delivery [VD] or CS), laboratory data (human leukocyte antigen-B27 [HLA-B27], C-reactive protein [CRP], and time-averaged CRP), and treatments were collected from the complete medical records. Time-averaged CRP was de ned as the mean CRP level during the CT intervals 15 . The normal range of CRP was de ned as 0-0.5 mg/dL. CRP elevations related to infections or medical interventions, based on a retrospective chart review, were excluded from the analysis. The treatment strategies used were classi ed as continuous, on-demand, or discontinued treatment.

Ethics approval
This study was conducted in accordance with the Declaration of Helsinki and approved by the institutional review boards (IRBs) of Gangdong Kyung Hee University Hospital (2019-08-012) and Konkuk University Medical Center (KUH1011006). Written informed consent was waived by IRB of Gangdong Kyung Hee University Hospital and Konkuk University Medical Center owing to the retrospective nature of the study and because the analysis used anonymous clinical data.

CT scoring system
The CTSS, ranging from 0 to 552, was used to compare the radiographic progression of the whole spine 9 . The anterior and posterior quadrants from the bottom half of C2 to the top half of S1 were scored in the coronal and sagittal planes (23 vertebral units [VUs] in total), scoring eight quadrants per VU. The score re ects the height of a syndesmophyte relative to the intervertebral disc space (IDS). For every quadrant, the score was 0 if no syndesmophyte was present, 1 if a syndesmophyte was present but did not reach 50% of the IDS, 2 if the syndesmophyte reached or crossed 50% of the IDS, and 3 if the syndesmophyte bridged the IDS.
The SIJs were divided into the left and right segments and the iliac and sacral segments, for a total of four segments. The scoring of the radiographs considered erosions, joint space alterations, and sclerosis. Erosions were graded from 0 to 3 (grade 0, no erosions; grade 1, small, isolated erosions [1][2] or a questionable single erosion; grade 2, de nite erosions [3-5; <3 mm] or a larger single erosion [>3 mm]; and grade 3, multiple [>5] or con uent erosions). Each of the four segments was independently assessed for the slice with the maximum number of erosions in both the transverse and coronal views. By adding the number of erosions in all four segments in both transverse and coronal views, the total erosion score (TES) was calculated (0-24). The joint space alterations were graded from 0 to 4 (grade 0, no joint space changes; grade 1, questionable widening or narrowing; grade 2, pseudo-widening; grade 3, partial ankyloses; and grade 4, extensive/total ankyloses). Ankylosis was de ned as >1 cm of marrow contiguous between the sacral and iliac sides. Extensive ankylosis was de ned as affecting at least half of the joints. Sclerosis in the coronal view was graded from 0 to 2 (grade 0, no sclerosis; grade 1, questionable or little sclerosis [≥5 mm]; and grade 2, evident sclerosis [≥10 mm]) 13 . The total SIJ score ranged from 0 to 40.
The CT scans were independently scored by two readers (K.A. Lee and S.Y. Lee). The readers were blinded to patient characteristics, clinical data, and ndings of other imaging modalities.

CT scanning technique
The CT scans and radiographs were read using a Digital Imaging and Communications in Medicine viewer. CT evaluations were performed with contiguous slices of 2-mm thickness. All spinal CT scans included a complete vertebral column. Because this study utilized preexisting scans, radiation exposure was not controlled and images of the SIJs were not reformatted into a semicoronal view.

Statistical analysis
Statistical analyses were performed using SPSS for Windows (version 22.0; SPSS Inc., Chicago, IL, USA).
Continuous variables are expressed as medians (Q1-Q3), and categorical variables are presented as frequencies and proportions. The results were compared using Pearson's chi-square test, Fisher's exact test, and the Mann-Whitney U test or Kruskal-Wallis test, as appropriate, for signi cant differences in patient characteristics between the delivery group and the control group. Interreader reliability was assessed using the intraclass correlation coe cient (ICC). The CT scores between the two time points were compared using the paired Wilcoxon test. The average score was used for the analysis. Statistical signi cance was set at p<0.05.

Results
Baseline characteristics A total of 45 women with AS had both spine and pelvic CT evaluations at baseline and follow-up (minimum 2 years, maximum 4 years): delivery group (n=17) and controls (n=28). Fifteen women with AS had only pelvic CT at baseline and follow-up: delivery group (n=5) and controls (n=10). Table 1 summarizes the baseline characteristics of the study population. The median (Q1-Q3) age of the AS patients in the delivery and control groups was 32 (30-34.25) and 30 (26-34.25) years, with a median (Q1-Q3) disease duration of 3 (0.87-6.0) and 1 (0.17-3.0) years, respectively. The median (Q1-Q3) interval time between the rst and second CT in the delivery and control groups was 2.96 (2.33-3.47) and 2.95 (2.43-3.33) years, respectively.
No apparent differences were observed between the two groups in terms of HLA-B27 positivity, extraarticular manifestations, CRP levels at baseline CT, medications, and smoking status. The median CRP levels at follow-up CT and the time-average CRP levels were higher in the delivery group than in controls (0.11 vs. 0.05 mg/dL [p=0.034] and 0.36 vs. 0.17 mg/dL [p=0.006], respectively); however, the levels were within the normal range in both groups. In the delivery group, 19 women (90.5%) discontinued AS medications for a median (Q1-Q3) of 12 (9.5-24) months because of pregnancy and/or breastfeeding (n=17) and inactive disease (n=2). In the control group, seven women discontinued treatment during the follow-up period because of inactive disease. The median (Q1-Q3) duration of treatment discontinuation was 14 (10.5-18) months.

Reliability
The interreader reliability is shown in Supplementary Table S1. The ICC for the baseline status score was 0.979 for the whole spine and 0.848-0.963 for the spine segments. The ICC for the change scores was 0.926 for the whole spine and 0.848, 0.940, and 0.647 for the cervical, thoracic, and lumbar spine segments, respectively. Very good reproducibility was observed in the total score (0.997-0.999), TES (0.994-1.000), joint space narrowing (JSN) score (0.996-0.998), and sclerosis score (0.878-0.992) of the SIJs.
However, no differences in the absolute changes in the total, erosion, JSN, and sclerosis scores of the SIJs per time point were observed between the delivery group and controls (p=0.471) ( Figure 2C,D). We analyzed the interval changes of SIJs according to the delivery method (CS [n=9] vs. VD [n=3]) in the delivery group (data on the delivery method were missing for nine patients). The changes in the total SIJ score, TES, and JSN and sclerosis scores were comparable between the CS and VD groups (data not shown).

Discussion
The present study assessed, using CT, the radiographic progression of the spine and SIJs in women with AS with and without delivery. We found that the interval changes in the CT scores of the spine and SIJs were not signi cantly different between AS patients with and without delivery. The interaction between biomechanical stress and the innate immune response has been proposed as a cause of AS. According to animal models, mechanical strain can contribute to entheseal in ammation and new bone formation 3 .
During pregnancy, the combination of mechanical and hormonal changes causes excessive load on the spine and pelvis and alterations of the pelvic ligaments, leading to pelvic and low back pain 16 . Ursin et al. reported that 70% of pregnant women with axial spondyloarthritis in the third trimester did not take any medication. Discontinuation of TNFIs has been reported to be associated with disease ares in parous women with AS 8 . Therefore, we hypothesized that biomechanical stress and discontinuation of treatment could affect the radiographic progression of the SIJs and spine in patients with AS during pregnancy. However, our study results showed no association between childbirth and structural damage in women with AS.
A major advantage of using CT for the assessment of syndesmophytes is that CT visualizes the whole spine in the sagittal and coronal axes with many slices. CT imaging studies have shown that more extensive syndesmophytes and more progression occur in the thoracic spine than in the cervical or lumbar spine 17 . Indeed, the CTSS is a more detailed score for syndesmophyte growth, indicating a height reaching > 50% or < 50% of the IDS, than the mSASSS, which indicates only the presence or absence of syndesmophytes 9 . Radiographic assessment of spinal progression in AS has been reported to have limited reliability 11 . However, we found that the CT scores had excellent interobserver reliability, consistent with a previous CTSS study 9 . With respect to SIJs, conventional radiography underestimated sacroiliitis compared with CT in the SIMACT (SacroIliac Magnetic resonance CT) study 13 . Although CT provides complete visualization of structural damage, concerns about radiation exposure have traditionally limited its use in AS patients. However, the development of low-dose CT has led to a reduction in radiation dose and noise. The effective dose in low-dose CT of the whole spine was estimated to be approximately 4 mSv, which is still higher than that of radiography (4 mSv: total dose from cervical, lumbar, and pelvic radiography) 9 . Diekhoff et al. showed that the mean radiation exposure with low-dose CT of the SIJs was similar to that with conventional radiography (0.51 vs. 0.52 mSv), with a higher detection rate for sacroiliitis 13 . Indeed, an oblique scanning method for the SIJs could avoid radiation exposure to the ovaries and require fewer slices to cover the whole SIJs compared with axial scanning 18 . Improvement in software together with the ability to reduce the radiation dose could make CT a feasible method for the evaluation of structural damage in the spine and SIJs in the near future.
Spinal progression was found to be associated with the presence of baseline syndesmophytes, male sex, old age, smoking, high disease activity at baseline, and time-averaged CRP levels 19,20 . An observational 2-year follow-up study on patients with AS using data from the SIAS (Sensitive Imaging in Ankylosing Spondylitis) study showed a mean (standard deviation [SD]) CTSS change score of 16 (21). In our study, the median (Q1-Q3) CTSS change in the delivery group and controls was 1 (0-3) and 0 (0-1), respectively, during a similar interval. In the SIAS study, 84% of the enrolled patients were men with a mean (SD) age of 50 (9.8) years, and 37% of the patients had elevated CRP levels at baseline and higher CTSS (whole spine mean [SD] score: 163 [126]). However, in our study, all patients with AS were women of a young age with a much lower CTSS at baseline (median [Q1-Q3] in the delivery group and controls: 19 [16][17][18][19][20][21][22][23] and 20 [13.25-27.75], respectively). Indeed, the median time-averaged CRP levels were within the normal range in both groups. The discrepancy in baseline characteristics could explain the different results of CTSS changes.
In a previous study, ~ 60% of women presenting with postpartum back pain without axial spondyloarthritis had positive magnetic resonance imaging (MRI) ndings for sacroiliitis according to the Assessment of SpondyloArthritis International Society de nition 21 . Eshed et al. also found that bone marrow edema (46%) and subchondral sclerosis (26%) were prevalent peripartum MRI ndings 22 .
However, we could not nd a link between these in ammatory changes in the peripartum period and the structural progression of SIJs. Although conventional radiography has limitations in evaluating the SIJs owing to their irregular outline and obliquity, CT is a sensitive method for detecting erosions, bone sclerosis, and ankylosis, which also represent the reference standards for the detection of bony alterations in AS 23 . Although radiographic progression of the SIJs was detected by CT in both AS patients with and without delivery, no signi cant differences in the interval changes of total, erosion, sclerosis, and JSN scores of the SIJs were found in our study. We expected that women with AS would have increased sclerosis in the SIJs after delivery, such as osteitis condensans ilii 24 . However, the SIJ scores for sclerosis in the delivery group did not signi cantly change. Although the score changes were small, the changes in the JSN scores in the delivery group and in the erosion and sclerosis scores in the control group were signi cant. Semiquantitative scores were used to evaluate the structural damage of the SIJs. Further studies with detailed quantitative measurements, such as the number and size of erosions or the exact thickness of sclerosis, are needed to con rm our study results.
Our study suggested that the delivery method (CS or VD) does not affect the radiographic progression of SIJs in patients with AS. A recent study using a medical claims database also showed that changes in prescription did not signi cantly differ between CS and VD 25 . The SIJs are not a part of the pelvic inlet and cannot be mechanically affected during delivery. In the present study, the rate of CS (n = 9, 75%) was much higher than that of VD (n = 3, 25%) in women with AS. However, the small sample size, especially in the VD group, limited the assessment of structural changes of the SIJs according to the delivery method.
The effect of nonsteroidal anti-in ammatory drugs (NSAIDs) on the inhibition of radiographic progression in AS is controversial. A randomized controlled trial of diclofenac failed to show that this NSAID can prevent structural progression in AS 26 . A post hoc analysis of the celecoxib trial identi ed that AS patients with elevated acute-phase reactants bene ted from treatment with celecoxib with respect to radiographic progression 27 . In our study, most AS patients in the delivery group stopped their medications after the con rmation of pregnancy. NSAIDs were the most prescribed medications, and the proportion of patients taking TNFIs was small in both groups. Indeed, the median time-averaged CRP levels were within the normal range in both groups. Owing to the low disease activity and low use of TNFIs in our study population, treatment discontinuation in pregnancy could not make a signi cant difference in structural progression between the two groups.
Although it provided insights into the association of pregnancy and delivery with radiographic progression in AS, our study had several limitations. First, the statistical power was limited by the small sample size. The small number of patients with at least two CT evaluations in 2-4-year intervals, the relatively low prevalence of AS in female patients, and the low birth rate in Korea made it di cult to enroll participants in our study 28 . Second, the retrospective design of this study is a shortcoming. Certain clinical parameters, such as the Bath Ankylosing Spondylitis Disease Activity Index and the Ankylosing Spondylitis Disease Activity Score, could not be evaluated through retrospective chart reviews. Information on the delivery method was missing for nine patients. Third, the present study utilized preexisting scans, and the radiation dose of CT was not uniform between the two centers. Images were not reformatted into an oblique coronal view oriented parallel and perpendicular to the long axis of the sacrum. A total of 24 regions of the SIJs were scored in the SIMACT study 13 . However, only four segments with coronal and sagittal views were analyzed in our study, which could have led to a low detection of radiographic progression in the SIJs.
In conclusion, present study found that pregnancy and delivery have no effects on the radiographic progression of the spine and SIJs in female patients with AS based on CT scoring methods. Our study results can be provided to women with AS who are planning to conceive, for informative and educational purposes.

Declarations Data availability
The datasets in this study are available from the corresponding author on reasonable request.
Authors' contributions KAL, HRK, and SHL were involved in study conception and design. KAL, SYL, SHK, and HSK were involved in data acquisition. KAL and SHL performed data analysis and interpretation. All authors were involved in the drafting or critical revision of the article, and all authors approved the nal version for publication. HRK and SHL had full access to all of the data in the study and takes.  Coronal-view CT scans of SIJ at baseline (a) and follow-up (b) in women with AS. New bone erosion in the right sacral side (white arrow) and sclerosis (asterisk) in the right iliac side are noted in follow-up CT.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.