Ankylosing spondylitis (AS) is a complex chronic disease where spinal inflammation typically causes lasting back pain with progressive loss of spinal mobility through sacroiliac and paraspinal calcification [1]. A significant number of patients also develop extra spinal complications affecting the eyes, peripheral joints, or intestines [2, 3]. The disease presents in the second or third decade of life and with no cure available. AS remains a lifelong condition with an unpredictable disease course that will negatively affect quality of life for many years in a relatively young group of patients [4, 5]. The clinical efficacy of various cytokine blocking therapies suggests that spinal inflammation in AS is at least in part driven by cytokines, but how these cytokines contribute to typical inflammatory back pain and spinal rigidity is less well understood [6,7,8]. High levels of TNF mRNA have been detected in SI joint biopsies and GWAS studies have suggested involvement of genetic pathways in AS especially for IL-23, IL17A, and TNF [9,10,11], but translational studies of the molecular pathways involved have been equivocal [12,13,14]. We investigated whether serum levels of four AS-associated cytokines could serve as useful markers of active disease, spinal function, and patient-reported health in the management of AS patients.


In a cross-sectional study design we included AS patients who fulfilled New York criteria [15]. Clinical data registered included demographics, spinal mobility measurements with patient-reported health measures assessed by Bath Functional Index (BASFI), Dougados Functional Index (DFI), Likert scale for pain and Modified Health Assessment Questionnaire (M-HAQ), as well as routine laboratory results (hemoglobin, ESR, CRP, albumen, and creatinine levels).

Disease activity was categorized according to the validated ASDAS-CRP classification (< 1.3 = inactive disease/1.3–2.1 = low disease activity/> 2.1 = high to very high disease activity) [16, 17], while serum concentrations for four AS-related cytokines (IL-6, TNF, IL-23, and IL-17A) were subsequently determined in stored aliquots (– 20 °C) by a quantitative sandwich immunoassay (Human DuoSet, R&D Systems, Minneapolis, MN, USA) [18]. Cytokine results for duplicate runs in each lot were averaged with all sera collected within a 3-day time frame from the collection of clinical and QOL data. All data and bio samples were collected during an outpatient visit prior to any treatment with biologics.

Statistics descriptors are given as numbers (%) or median (interquartile range). Comparisons were done by Chi-square (with Yates correction) and Kruskal–Wallis test, and correlations used Spearman rank coefficients. SPSS v 23 software was used with alpha levels set at 5%.

All procedures performed in studies involving human participants were in accordance with REK-NORD 2012/1589 and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. Informed consent was obtained from all individual participants included in the study.


Patient characteristics demonstrated the expected predominance of male patients and high frequency of HLA- B27 and demonstrate that 20 years into the disease course only 14% of patients were classified as having inactive disease with 31% still showing a high level of disease activity (Table 1). There was no difference in the distribution of age, gender, and disease duration across all three disease activity categories. Peripheral arthritis was more frequent in patients with high disease activity, but no significant difference was seen for other extra spinal manifestations. Overall, 41% of patients were receiving disability benefits, but the frequency was low in patients with inactive disease.

Table 1 Clinical characteristics of AS patients (n = 164) overall and by disease activity category. Numbers indicate counts (%) or median values (interquartile range)

Spinal function and patient-reported health measures were all significantly worse in patients with high disease activity (Table 2). A Spearman rank correlation analysis between absolute scores for ASDAS-CRP and these disease manifestations demonstrated that the association for ASDAS CRP was strongest with all spinal mobility measures except for lateral flexion (Table 3). Patient-reported health measures demonstrate excellent agreement between DFI, M-HAQ, and BASFI, but had lower agreement with ASDAS-CRP than mobility scores. In a similar analysis, ASDAS-CRP correlated significantly with levels of ESR and WBC and inversely with Hb and albumin levels with significant differences observed between ASDAS categories (Table 4) (Suppl Table 1). The individual cytokine levels, however, were not significantly different across disease activity categories (Fig. 1) and there was no correlation between absolute level of ASDS-CRP and any of the cytokine levels, even though cytokine levels were strongly interrelated (Table 4). M-HAQ, BASFI, and DFI scores did not correlate with cytokine levels (suppl Table 2).

Table 2 Spinal function and health status overall and in relation to disease activity status at visit
Table 3 Spearman rank coefficients for the correlation between absolute ASDAS-CRP scores and a range of clinical disease features in AS
Table 4 Spearman rank coefficients for the correlation between absolute ASDAS-CRP scores and a range of biomarkers for inflammation
Fig. 1
figure 1

Cytokine levels (mean with 95% CI) by disease activity category (ASDAS-CRP). All p values > 0.2 by Kruskal–Wallis test


This cross-sectional study demonstrates that many AS patients not on biologics continue to have significant disease activity far into their disease course. While disease activity impacted both spinal function and patient-reported health and correlated strongly with acute phase reactants, none of these findings was paralleled by the levels of any of four AS-related cytokines.

In this typical Caucasian cohort of AS patients not (yet) receiving biological therapy, inactive disease was seen in only a low proportion of patients with a prolonged disease course. This confirms that AS does not burn out for at least a decade following an often-delayed diagnosis and also highlights the limitations of synthetic DMARD therapy, which were prescribed for approximately 25% (data not shown) of our AS patients [5, 19,20,21]. The high number of patients receiving disability payments illustrates how continued and/or recurrent disease activity in AS impacts patients’ lives. We did find that patients with inactive disease less frequently received disability payments and even though the relationship between disease severity and non-employment is not straightforward, it strengthens the argument for early therapeutic intervention to attain inactive disease in AS and reduce the associated cost to society [22, 23].

The development of BASDAI and ASDAS criteria for active disease in AS have allowed for better definition of disease flares and facilitated investigation of the mechanisms of flares and their impact on patients [24,25,26]. In this cross-sectional study, we found disease activity contributes to reductions in spinal mobility test results (Tables 2 and 3), suggesting that changes in spinal mobility in AS are not necessarily a sign of structural changes but can be impacted by inflammation as well [27,28,29,30]. This finding has clinical relevance in that it supports the view that patients with limited function may still improve with effective anti-inflammatory drug therapy.

Patient-reported health measures and outcomes have become a key aspect of clinical studies and trials [31, 32]. In this cross-sectional study, it was clear that while there was a high level of agreement between the three patient health questionnaires, they were in much lower agreement with scientific definitions of disease activity, spinal mobility measures, acute phase reactants, and had virtually no agreement with cytokine levels. Thus, while objective disease measures may reflect aspects of reduced quality of life, they only partially underwrite the reduction in quality of life reported by patients [33, 34]. This discrepancy indicates that further refinement of treatment strategies will require incorporation of other outcome targets than scores or surrogates for disease activity [35,36,37].

Standard laboratory markers of the acute phase response in AS correlated significantly with ASDAS-CRP scores as well as with each other, but the correlation coefficients were modest. This illustrates the considerable unexplained variation in the levels of inflammatory markers in AS patients, which is unrelated to disease activity and may reflect other factors such as lifestyle, coping skills, and smoking [38, 39]. Similarly, cytokines levels were strongly interrelated in this AS cohort, reflecting their interaction in mitigating disease in AS patients [40,41,42], but there was a complete disconnect between cytokine levels and clinical disease activity, laboratory findings, and patient-reported health measures. Other, but not all, studies have reported a similar lack of clinical association for cytokines [1, 42, 43], but given the efficacy of anti-cytokine therapy in AS [6, 44, 45], it is hard to reconcile this discrepancy between pathophysiological and clinical findings. We have earlier also shown that TNF levels do not predict response to TNFi treatment in AS [18], but given the increased levels of TNF mRNA in sacroiliac joints [9], it can be argued that there is no or only limited overflow of site-specific inflammation markers in AS similar to the lack of specific biomarkers found in rheumatoid arthritis [46]. Similarly, as IL-6 targeting has not been successful in AS, the observed lack of correlation between IL-6 and ASDAS-CRP levels in this study confirms that the effect of IL-6 on CRP in AS is not straightforward and depends on genetic factors [38, 47]. Taken together, these data indicate that new diagnostic efforts are needed to allow for in situ measurements or imaging of these inflammatory markers in AS. Methodological differences regarding cytokine assays certainly play a role as well, while it remains possible that what is considered axial SpA/AS actually comprises a diverse group of conditions [48].

The limitations of this study lie in the exclusive Caucasian makeup of the cohort cross-sectional design and the lack of longitudinal data, which may better reflect the effects of chronic stimulation by cytokines [41, 42]. The strength of this study lies in the comprehensive and uniform data collection including a range of patient-reported health measures.


In conclusion, disease activity by ASDAS-CRP score remains a significant problem in a large number of conservatively treated AS patients. Reduction of spinal mobility and QoL in these patients is at least partly due to inflammation and may be amenable to effective treatment. Serum cytokine levels are not helpful to guide personalized management in AS.