Abstract
Objective To investigate the association between cigarette smoking and hip joint involvement in ankylosing spondylitis (AS).
Methods This case-control study compared patients with AS with and without hip involvement, as defined by the Bath Ankylosing Spondylitis Radiology Hip Index. Logistic regression analysis, subgroup analysis, and sensitivity analysis were conducted to estimate the association between smoking and hip involvement in AS.
Results This study included 103 patients with hip involvement (cases) and 89 patients without hip involvement (controls). In univariate analysis, patients who had juvenile-onset AS (JAS), were younger, were male, had peripheral arthritis history, or had cigarette exposure were prone to hip involvement. After adjusting for confounding factors, JAS (adjusted odds ratio [aOR] 2.52, 95% CI 1.26-5.06), male sex (aOR 2.89, 95% CI 1.14-7.33), and cigarette smoking (aOR 7.23, 95% CI 2.27-23.05) were regarded to be independently associated with hip involvement in AS. Moreover, patients who smoked with exposure of less than 10 pack-years were 2.2 times more likely to have hip involvement than those without (aOR 2.21, 95% CI 1.09-4.47). This association was reproduced in subgroup analyses of males and propensity score–matched subjects, and it withstood sensitivity analysis.
Conclusion Smoking is a novel independent risk factor for hip involvement in AS; even exposure of less than 10 pack-years could contribute to increased prevalence of hip involvement in AS, which underlines the significance of smoking cessation in patients with AS, especially for JAS.
Ankylosing spondylitis (AS) is the prototype of axial spondyloarthritis (axSpA), which constitutes a group of chronic inflammatory diseases that primarily affect the spine, sacroiliac joints, and peripheral joints.1 Hip involvement, a prevalent manifestation of AS, is associated with spinal damage, functional impairment, increased disease burden, and poor prognosis.2,3 However, the precise mechanisms involved in the development of hip joint damage in AS are still not fully understood. In addition, only a few risk factors have been reported, such as juvenile-onset AS (JAS), male sex, history of peripheral arthritis or enthesitis, and raised serum C-reactive protein (CRP) levels,2-4 which hampers the early detection of patients and early access to proper treatments, such as tumor necrosis factor inhibitors.5,6
Cigarette smoking has been reported to be closely connected to AS. It has been shown to be associated with higher disease activity, worse physical function, poorer quality of life, and, more importantly, severe spinal damage.7,8 In addition, spinal damage in AS has been shown to be associated with hip involvement. A Chinese study reported that severe spine structure damage was a predictor of radiological hip involvement,9 and data from a large international study also pointed out that patients with severe hip involvement were more likely to have spinal damage.2 However, the association between cigarette smoking and hip involvement in AS has not been fully investigated. The identification of such a modifiable factor could not only promote our understanding of the pathogenesis in AS, but also facilitate early detection and treatment of high-risk patients. Herein, we rendered the hypothesis that smoking is associated with hip involvement in AS, and we conducted a case-control study to provide evidence for it.
METHODS
Patients and data sources. Patients enrolled in this study were from the Chinese Ankylosing Spondylitis Prospective Imaging Cohort (CASPIC), a nationwide, longitudinal cohort launched in 2016. The study protocol has been previously demonstrated.10 Patients were enrolled at the outpatient rheumatology clinics at the First Medical Center of the Chinese People’s Liberation Army (PLA) General Hospital, a tertiary referral center in Beijing, China. Before they participated in this study, patients were adequately informed that the data they provided would be saved and used in scientific research only.
Adult patients aged 18 years or older were enrolled from May 2017 to September 2020 if they fulfilled the 1984 modified New York criteria for AS.11 Exclusion criteria were as follows: (1) refusing or unable to complete the survey, (2) providing incomplete or poor-quality data, and (3) having AS in combination with other diseases.
Data collection. The following demographic data and disease features were acquired from questionnaires and medical records: age at onset, disease duration, peripheral arthritis history, Bath Ankylosing Spondylitis Functional Index (BASFI),12 Bath Ankylosing Spondylitis Disease Activity Index (BASDAI),13 Ankylosing Spondylitis Disease Activity Score with erythrocyte sedimentation rate (ASDAS-ESR),14 and HLA-B27 status. Further, the following radiographs were obtained at enrollment to confirm the diagnosis and assess the severity of structural damage: radiographs of the anteroposterior (AP) pelvis, lumbar spine (AP and lateral), and cervical spine (lateral). The severities of hip and spine impairments were assessed by the Bath Ankylosing Spondylitis Radiology Hip Index (BASRI-h)15 and the modified Stoke Ankylosing Spondylitis Spinal Score (mSASSS),16 respectively. After a training session and a consensus meeting to master the scoring methods, all radiographs were read separately and blindly by 2 rheumatic disease physicians (ZH and YL).
The details of the BASRI-h are demonstrated below. Regarding the mSASSS, the anterior corners of the lumbar and cervical vertebra were scored using the following criteria: 1 point for squaring, erosion, or sclerosis; 2 points for syndesmophytes; and 3 points for bridging syndesmophytes. The total score ranged from 0 to 72.
Ethics and consent. The study protocol was approved by the Ethical Committee of the Chinese PLA General Hospital (approval no. S2016-049-02), and informed consent for participation was obtained from all patients at enrollment. The study was performed in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines and complied with the Declaration of Helsinki.
Exposure. Regarding smoking status, patients were designated as nonsmokers or ever-smokers. Nonsmokers were subjects who reported no smoking history. Patients who were actively smoking at enrollment or had previously smoked were identified as ever-smokers. Specific information on smoking, including smoking duration and average number of cigarettes per day, was also collected. Ever-smokers were categorized by smoking duration (< 15 vs ≥ 15 years) and smoking intensity (< 15 vs ≥ 15 cigarettes per day).17 Cumulative cigarette exposure in pack-years was calculated by multiplying the number of years a patient smoked by the packs smoked per day. Ever-smokers were classified by cumulative cigarette exposure into 2 groups: < 10 or ≥ 10 pack-years.
Cases and controls. BASRI-h scores were represented by 5 categories: 0 = no change (normal); 1 = possible focal joint space narrowing (suspicious); 2 = definite narrowing, leaving a circumferential joint space > 2 mm (minimal); 3 = narrowing but circumferential joint space < 2 mm or bone-on-bone apposition < 2 cm (moderate); and 4 = bone deformity or bone-on-bone apposition > 2 cm or total hip replacement (severe). Patients with BASRI-h values of 2 or higher were defined as positive for hip involvement (ie, cases),15 whereas patients with BASRI-h values of less than 2 were defined as negative for hip involvement (ie, controls).
In practice, some patients would be classified as positive for hip involvement by 1 observer but negative by the other. Patients with disputed results were adjudicated by an expert reader to make the final decision. However, the definition of hip stage in radiographs leaves much room for subjective interpretation: the interobserver 
ranged from 0.59 to 0.60 for experienced observers, even after the training session and consensus meeting to minimize discrepancy.18 Additionally, it was possible for patients with disputed results to have hip involvement, whereas their joint structural damage may not have been severe enough to meet the BASRI-h criteria. For these reasons, in our study, patients with consistent results from 2 observers were admitted into the main analysis, whereas patients with disputed results were only included in the sensitivity analysis to evaluate the reliability of results from the main analysis.
Statistical analysis. Quantitative variables were reported as mean (SD) or median (IQR) for normally or nonnormally distributed data, respectively. Categorical variables were described as frequencies and percentages. An independent-samples t test or the Mann-Whitney U test was used to compare continuous variables between cases and controls, and the chi-square test was used for categorical variables.
In the multivariable analysis, binary logistic regression (ie, backward-conditional) was conducted to compare the proportion of smoking exposure in cases to that in controls and to estimate the odd ratios (ORs) and 95% CIs. Then, multivariable analysis was reperformed in male patients and in matched cases and controls to exclude the confounding effects of sex and other heterogenetic factors. Matched cases and controls were selected by propensity score matching at a ratio of 1:1, by age, sex, disease duration, and JAS.
Sensitivity analysis was performed as 2 exaggerated situations were set: patients with disputed results were assumed to all have hip involvement or not; they were then merged with identified cases or controls. For all analyses, a P value of less than 0.05 was considered statistically significant. Statistical analysis was performed with SPSS Statistics (version 22; IBM Corp.).
RESULTS
We enrolled 338 patients with AS in our study; 27 patients were excluded owing to poor-quality radiographs, 51 patients were excluded for lacking previous medical records, and 32 patients were excluded because they were combined with other diseases. Finally, 228 patients were admitted into the analysis. A flowchart of the analytical approach is presented in Figure 1.
Flowchart of analytical approach.
The median age of the 228 patients was 31.0 (IQR 26.0-38.0) years, and disease duration was 10.0 (IQR 6.0-14.0) years. Out of 228 patients, 190 (83.3%) were male and 211 (92.5%) were HLA-B27 positive. In total, 103 patients (45.2%) were identified as having hip involvement by both observers, 36 patients (15.8%) were rated as positive for hip involvement by only 1 observer, and 89 patients (39%) were rated as negative for hip involvement by both. Demographics and other disease-related characteristics of the 103 cases and 89 controls are detailed in Table 1.
Characteristics of cases and controls in main analysis and propensity score–matched analysis.
Cases were younger (median age 30.0 yrs vs 34.0 yrs, P = 0.04) and had a lower age at disease onset (median age 19.0 yrs vs 23.0 yrs, P = 0.01) than controls. There was higher percentage of males (92.2% vs 73%, P = 0.01) among cases than controls, as well as a higher rate of JAS (43.7% vs 20.2%, P = 0.01) and peripheral arthritis history (35% vs 18%, P = 0.01). However, cases and controls did not differ in terms of disease duration, HLA-B27–positive rate, the BASFI, the BASDAI, the ASDAS-ESR, and the mSASSS. Regarding smoking status, there was a greater proportion of ever-smokers among cases than controls (46.6% vs 29.2%, P = 0.01). However, ever-smokers among cases had shorter smoking durations (10.4% vs 30.8% with ≥ 15 yrs, P = 0.03) and less cumulative smoking exposure (10.4% vs 26.9% with ≥ 10 pack-yrs, P = 0.07) than those among controls, although they shared a similar extent of smoking intensity (39.6% vs 34.6% with ≥ 15 cigarettes/day, P = 0.68; Table 1).
Risk factors of having hip involvement in patients with AS. Univariate analysis revealed that JAS (OR 3.06, 95% CI 1.60-5.85), male sex (OR 4.39, 95% CI 1.86-10.36), and ever-smoking status (OR 2.12, 95% CI 1.16-3.85) were associated with elevated hip involvement risk in AS. The association between JAS and hip involvement remained statistically significant in all models; the same was true for the association between sex and hip involvement. Ever-smoking status was associated with 7.2-fold and 3.7-fold higher odds of having hip involvement than nonsmoking status in models 2 and 4, respectively, where the models were adjusted for JAS, sex, and smoking duration or pack-years (model 2: adjusted OR [aOR] 7.23, 95% CI 2.27-23.05; model 4: aOR 3.74, 95% CI 1.51-9.27). However, this association became insignificant in other models (Table 2).
Risk factors of having hip involvement in patients with AS.
Association between hip involvement and cumulative smoking exposure. In univariable analysis, smokers with exposure of less than 10 pack-years had significantly greater odds of having hip involvement compared with nonsmokers (OR 2.59, 95% CI 1.35-4.97). However, this association lost statistical significance among smokers with exposure of 10 or more pack-years (OR 0.82, 95% CI 0.25-2.73). After adjustment for JAS, sex, and disease duration, the association between smoking and higher risk of hip involvement remained among smokers with exposure of less than 10 pack-years (aOR 2.21, 95% CI 1.09-4.47), whereas it was still insignificant among smokers with exposure of 10 or more pack-years (aOR 0.70, 95% CI 0.20-2.45).
Subgroup analysis in males. A similar analysis was conducted among male patients. Results restricted to males were similar to the main results: after adjustment for JAS, smoking duration or JAS, pack-years in model 2 or 4, ever-smokers had 6.5-fold and 3.5-fold higher odds, respectively, of having hip involvement than nonsmokers (model 2: aOR 6.51, 95% CI 2.01-21.06; model 4: aOR 3.49, 95% CI 1.40-8.75; Table 3).
Risk factors of having hip involvement in male patients with AS.
Propensity score–matched analysis. In total, 55 cases and 55 controls were matched and their disease manifestations were comparable, except for a higher percentage of ever-smokers and peripheral arthritis history among the cases (Table 1). Ever-smokers among the cases had shorter smoking durations (8% vs 35% with ≥ 15 yrs smoking; P = 0.03) and less cumulative smoking exposure (8% vs 30% with ≥ 10 pack-yrs smoking; P = 0.06) than those among the controls. However, ever-smokers among the cases and controls shared a similar extent of smoking intensity (44% vs 35% with ≥ 15 cigarettes/day; P = 0.55). The association between ever-smoking and hip involvement was insignificant in the unadjusted model (OR 1.39, 95% CI 0.68-2.83) and in the model adjusted for smoking intensity (aOR 0.68, 95% CI 0.13-3.38). However, it became statistically significant when the model was adjusted for smoking duration (aOR 7.89, 95% CI 1.75-35.63) and pack-years (aOR 3.02, 95% CI 1.03-8.82).
Sensitivity analysis. Sensitivity analysis was performed to assess the influence of the all-or-none imputation of patients with suspicious hip involvement on the main result. In our first sensitivity analysis, a complete case scenario was set, in which those 36 patients were all assumed to have hip involvement (139 cases and 89 controls). When the models were adjusted for JAS, sex, and smoking duration or pack-years, ever-smokers had 5.9-fold or 3.0-fold higher odds of having hip involvement than nonsmokers (model 2: aOR 5.88, 95% CI 1.90-18.20; model 4: aOR 2.99, 95% CI 1.24-7.19). Secondly, those patients were assumed to all be free from hip involvement (103 cases and 125 controls). These significant associations were reproduced in the same multivariate models (model 2: aOR 6.02, 95% CI 2.26-16.05; model 4: aOR 3.80, 95% CI 1.69-8.51; Table 4).
Sensitivity analysis of the risk factors of having hip involvement in AS.
DISCUSSION
Hip involvement is a common manifestation of AS and is associated with poor outcomes.2,3 In a large international study exploring the effects of hip involvement on AS, 20% to 30% of patients suffered from hip involvement and, more importantly, 5% to 8% of them had undergone hip replacement surgery.2 Despite the large burden of hip involvement in AS, it was poorly understood. It is noteworthy that Chinese patients with AS were reported to be especially prone to hip involvement19 and, consequently, disease burden in them is likely to be underestimated. This study has identified for the first time, to our knowledge, that smoking exposure is an independent risk factor for hip involvement in Chinese patients with AS.
Cigarette smoking is involved in the pathogenesis of AS. Smoking has been demonstrated to be intimately associated with an increase in many inflammatory markers, such as CRP. Raised CRP has been reported to not only be a strong predictor of syndesmophytes formation20 but to be associated with hip involvement in AS.4 In previous studies, smoking was demonstrated to be associated with worse physical function and poorer quality of life in AS.20,21 Additionally, we showed that smoking was associated with higher prevalence of hip joint damage in AS.
In our study, patients with hip involvement shared a similar BASFI and mSASSS to those without, which is at odds with previous studies.2,3 However, in our study, patients with hip involvement were younger than those without (median age 30.0 yrs vs 34.0 yrs, P = 0.04) and had a virtually higher percentage of JAS (43.7% vs 20.2%, P = 0.01). Based on the finding that patients with JAS are less likely to have axial symptoms and spinal damage,3,22,23 it is sensible that they had a higher percentage of hip involvement while being exempted from severe axial disease and functional impairment.
In the main analysis, cigarette exposure was independently associated with hip involvement in AS. The significant association between smoking and hip involvement remained among propensity score–matched subjects and the subgroup of males, and it withstood 2 sensitivity analyses, which vigorously enhanced the robustness of results from the main analysis. The relationship between cumulative smoking exposure and hip involvement was also demonstrated. After adjustment for confounding factors, a significant association between cigarette exposure and hip involvement was found among smokers with exposure of less than 10 pack-years (aOR 2.21, 95% CI 1.09-4.47), but not on smokers with exposure of 10 or more pack-years (aOR 0.70, 95% CI 0.20-2.45). The unelevated odds of hip involvement among the higher-exposure group might be explained by a lack of statistical power in this group, given that there were relatively low numbers of subjects in this group (only 5 cases and 7 controls).
Smoking cessation is a well-recognized principle in the treatment of SpA21 because of the reported association between smoking exposure and spinal damage, let alone that smoking is one of the greatest threats to health. However, quitting smoking is not as easy as we would expect in the real world; a multicentric cohort study reported that 40% of its patients were current smokers.24 Patient education about why and how to quit smoking remains an important and unmet clinical matter in improving patient outcomes. Experience from the field of behavioral economics could shed light on this issue.25
Our study has some limitations. First, recall bias and selection bias are inevitable, despite collection of and consultation about patients’ medical records. Although patients were enrolled from a single center, this center is a tertiary referral hospital in Beijing, and over half of the patients in the CASPIC study came from another area of China.10 Further, we adjusted our analyses for already-known confounding factors, but we cannot exclude the possibility of confounding by unknown factors. The lack of some important data in the cohort, such as alcohol drinking, psychological and socioeconomic conditions, and physical activity, poses another limitation to our study. However, we not only used different models to analyze the relationship between smoking and hip involvement, but we also selected propensity score–matched subjects to control for confounding bias. In addition, sensitivity analyses were conducted to evaluate the validity of our conclusion. In our study, the OR value observed for ever-smoking after adjustment was greater than 7, which clearly indicates that there was a sound association between cigarette exposure and hip involvement in AS. Further research might investigate whether our observations persist across racial and ethnic groups and whether the odds of hip involvement in AS increase with increasing pack-years of smoking.
In summary, this study showed that cigarette smoking is a novel independent risk factor for hip joint damage in AS. Even exposure of less than 10 pack-years could contribute to an increased prevalence of hip involvement in AS, underlining the significance of smoking cessation in AS, especially for JAS.
Footnotes
Z. Hu and Y. Li contributed equally to this paper.
The authors declare no conflicts of interest relevant to this article.
- Accepted for publication December 28, 2022.
- Copyright © 2023 by the Journal of Rheumatology
REFERENCES
DATA AVAILABILITY
All relevant data are reported in the article. Additional details could be provided by the corresponding author upon reasonable request.
