Serum Levels of OPG, RANKL, and RANKL/OPG Ratio in Patients with Ankylosing Spondylitis: A Systematic Review and Meta-analysis.

ABSTRACT Objectives: To investigate the role of osteoprotegerin (OPG), receptor activator of nuclear factor-kB ligand (RANKL), and RANKL/OPG ratio in the pathogenesis of ankylosing spondylitis (AS). Methods: Studies that compared serum levels of OPG, RANKL, and RANKL/OPG ratio between AS patients and healthy controls were gathered. Pooled standardized mean differences (SMDs) with 95% confidence intervals (CIs) were calculated by the random-effects model. Results: Twenty studies containing 1592 AS patients and 1064 healthy controls were included in this meta-analysis. Serum levels of OPG, RANKL, and RANKL/OPG ratio in AS patients were significantly higher than that in normal controls (OPG: SMD = 0.401, 95%CI = 0.026–0.777, p = 0.036; RANKL: SMD = 1.116, 95%CI = 0.510–1.723, p < 0.001; RANKL/OPG ratio: SMD = 0.691, 95%CI = 0.084–1.299, p = 0.026, respectively). Subgroup analysis suggested that Asian AS patients and patients with elevated ESR (ESR >20 mm/h) had higher serum OPG levels compared to normal controls. Asian patients, CRP >10 mg/L, ESR >20 mm/h, duration of disease ≤8 years, and BASDAI score >4 points subgroups showed increased RANKL levels compared to controls. Conclusions: Serum levels of OPG, RANKL, and RANKL/OPG ratio may be used as potential susceptible biomarkers for AS, but they could be influenced by race, inflammatory factors, and disease activity of AS patients.


Introduction
Ankylosing spondylitis (AS) is the most common form of spondyloarthritis that usually affects young adults, and the prevalence of AS is 0.1%-1.4% in general populations (Braun and Sieper, 2007;Tang et al., 2018). AS is an immune-mediated inflammatory autoimmune arthritis characterized by damaged axial skeleton, sacroiliac joints, and spine attachment points, which can lead to loss of joint function and disability (Mahmoudi et al., 2016).
OPG and RANKL (also called TNF-related activation-induced cytokine, TRANCE) are members of tumor necrosis factor-α superfamily (Simonet et al., 1997). RANKL is mainly secreted by osteoblasts and activated T cells. RANKL induces the maturation and activation of osteoclasts by binding to receptor activator of nuclear factor-κB (RANK) on osteoclast precursors and causes bone resorption (Gamal et al., 2018). Similarly, OPG is predominantly produced by osteoblasts, and it could inhibit the activation of osteoclast that competitively binds RANKL and blocks the interaction between RANKL and its receptors (Lacey et al., 1998). RANKL/RANK/OPG system is considered as the key signaling system in regulating osteoclast activity. Imbalance of RANKL/RANK/OPG system may be directly involved in bone remodeling and bone loss of many diseases, such as osteoporosis, osteopetrosis, rheumatoid arthritis, chronic arthritis (Liu et al., 2018;Mou et al., 2015), and the pathogenesis of osteoporosis in AS (Kim et al., 2006).
Current findings concerning serum levels of soluble OPG, RANKL (sRANKL), and RANKL/OPG ratio in AS patients versus healthy controls were inconsistent. Thus, the purpose of the present meta-analysis is to comprehensively evaluate an association between biomarkers of bone metabolism (serum OPG and RANKL levels and RANKL/ OPG ratio) and AS susceptibility or clinical outcomes (inflammatory factors and disease activity).

Literature search strategy
PubMed, Web of Science, and three Chinese databases-Wanfang, Chinese National Knowledge Infrastructure (CNKI), and VIP Database were searched and updated until February 2018 to identify relevant studies that investigated the relationships between RANKL/RANK/OPG system and AS. Searching keywords included "osteoprotegerin" or "OPG" or "receptor activator of nuclear factor-kB ligand" or "RANKL" or "TNFSF11" or "TRANCE" or "OPGL" or "ODF" and "Ankylosing spondylitis" or "Spondylitis, Ankylosing [MeSH Terms]" or "Ankylosing Spondylarthritis [Entry Terms]" or "Rheumatoid Spondylitis [Entry Terms]" or "AS." To include all available studies, we contacted each corresponding author of articles that offered insufficient data for meta-analysis. All references of searched studies were also reviewed to identify additional studies. Finally, only published studies with full text were included.

Inclusion and exclusion criteria
Studies that met the following criteria were included: (1) PICOS criteria should be met in the studies. P: Ankylosing spondylitis; I: Enzyme-linked immunosorbent assay; C: Healthy control; O: Serum levels of OPG, RANKL; S: Case-control or Cross-sectional study or Clinical cohort; (2) offered sufficient data to calculate the mean and standard deviation (SD) of serum levels of RANKL and OPG and/or RANKL/OPG ratio for each group; (3) AS diagnoses fulfilled the 1984 Modified New York criteria (mNYC) or 1987 American College of Rheumatology modified AS diagnostic criteria; (4) written in English or Chinese; (5) peer-reviewed studies with full text.
Studies that met the following criteria were excluded: (1) insufficient data; (2) comment, review, and abstracts; (3) animals or in vitro study; (4) quality score assessed by Newcastle-Ottawa Scale (NOS) score less than 5.

Data extraction
According to the selection criteria, two researchers (Mengya Chen and Xingxing Hu) independently extracted data. Any discrepancy in data extraction was discussed by a third researcher (Meng Wu). The following data were extracted if available: first author, publication year, sample size, geographic location, study design, measurement, and mean ± SD for serum OPG, RANKL concentrations and RANKL/OPG ratio in AS patients and healthy controls, and disease duration, C-reactive protein (CRP) and erythrocyte sedimentation rate (ESR) levels and the Bath AS Disease Activity Index (BASDAI) in AS patients.

Statistical analysis
Serum levels of OPG and RANKL were presented as pg/mL. Standardized mean difference (SMD) and 95% confidence intervals (CIs) of serum OPG and RANKL levels were calculated for each study. In some studies, only median and range were obtained, so we transformed them into mean ± SD. Detailed description is provided in the previous article (Hozo et al., 2005). In short, m = Median, a = the smallest value (minimum), b = the largest value (maximum), and n = the size of the sample.
Considering the possibility of heterogeneity among studies, the effect of heterogeneity was evaluated using Cochran's Q-test and I 2 statistic (Lee and Bae, 2017). The fixed effect model was adopted if the heterogeneity was not statistically significant (I 2 < 50% or p> 0.1), otherwise the random effects model was used. Potential publication bias was investigated using Begg's test or Egger's linear regression test by visual examination of the funnel plot, and an asymmetric Funnel plot or p < 0.05 in Egger's test suggests possible publication bias (Egger et al., 1997). Sensitivity analysis was performed to assess the stability of results after excluded an article. Newcastle-Ottawa Scale (NOS) was used to evaluate the quality of the included studies by two observers (Mengya Chen and Xingxing Hu) independently based on three main items: the selection of the study groups (0-4 points), the comparability of the two groups (0-2 points), and the determination of the exposure (0-3 points), with a perfect score of 9. A meta-regression was conducted to further explore the source of heterogeneity. Statistical analyses were performed using STATA 14.0 (StataCorp, College Station, TX, USA). A two-sided p value ≤ 0.05 was considered statistically significant.

Data source
The process of study selection is summarized in Figure 1. A total of 226 relevant papers were identified according to the searching strategy, and 108 duplicate records were initially excluded. After screening the titles and abstracts, 98 articles were excluded. Finally, 20 studies (Bai et al., 2017;Chen et al., 2010;Dhir et al., 2013;Franck et al., 2004;Grisar et al., 2002;Hou et al., 2018;Jadon et al., 2017;Klingberg et al., 2014;Kong et al., 2010;Korkosz et al., 2013;Kwon et al., 2012;Li et al., 2013;Mou et al., 2015;Serdaroğlu Beyazal et al., 2016;Sveaas et al., 2015;Taylan et al., 2012;Wen et al., 2016;Xu et al., 2014;Zhang et al., 2010Zhang et al., , 2015 assessing serum OPG (n = 19), RANKL (n = 11) levels, and RANKL/OPG ratio (n = 7) in AS patients and healthy controls were included in the meta-analysis. All included studies were published between 2002 and 2018. Among the 20 studies, there were 1592 AS patients, mostly diagnosed by the modified New York classification criteria (mNYC) in 1984, and 1064 healthy controls. Serum levels of OPG and RANKL were measured by enzyme-linked immunosorbent assay (ELISA) in all studies. NOS scores of included studies mainly ranged from 5 to 8, and one article was excluded because of a NOS score <5. Detailed characteristics of the included studies are shown in Table 1.

Heterogeneity test results
Heterogeneity was significant in this meta-analysis (all p < 0.001, Table 2); therefore, the random-effect models were performed.

Meta-regression analysis
Meta-regression for RANKL/OPG ratio was not conducted due to the limited number of studies. The source of heterogeneity was explored by adding country (China and others), publication year, NOS score, and sample size (subjects ≤100 and >100) in meta-regression. The results of meta-regression did not reveal the source of heterogeneity (Table 3). Metaregression of race, disease, CRP, ESR, and BASDAI had performed to further explore the source of heterogeneity between different subgroups. However, no any heterogeneity was found between different subgroups (Table S1).
Publication bias and sensitivity analysis A significant publication bias was found for serum RANKL levels (Egger's test: p= 0.002, Begg's test: p= 0.013) but not for serumOPG levels and RANKL/OPG ratio using Egger's and Begg's tests. Sensitivity analysis showed that no significant changes were found by removing one single study in sequence, indicating that our results were statistically robust ( Figure 3).

Discussion
Accumulating evidence shows that abnormal bone remodeling and osteoporosis are frequent even in the early stage of AS (Baek et al., 2005;Calin, 1991; Davey-Ranasinghe and Deodhar, 2013). There is some evidence that OPG and RANKL are the important bone-turnover biomarkers and the relative balance of RANKL/RANK and OPG plays a pivotal role in the regulation of bone remodeling and bone loss (Kim et al., 2006). However, previous studies showed inconsistent results when comparing serum levels of  OPG and RANKL between AS patients and healthy controls. In some studies, serum levels of OPG and RANKL were elevated in AS patients ; however, serum OPG and RANKL levels have been found to be reduced in AS patients in some other studies (Franck et al., 2004;Klingberg et al., 2014). In addition, some researchers suggested that there were no definite differences between AS patients and healthy controls in serum levels of OPG (Jadon et al., 2017) and RANKL (Taylan et al., 2012). Some reasons may explain these discrepancies among studies, such as small sample size, different detection methods, and different races. Therefore, we conducted the present metaanalysis to comprehensively evaluate the relationships of serum levels of OPG, RANKL, and RANKL/OPG ratio with the development of AS. We found that serum OPG levels in AS patients were significantly higher than that in normal controls. High serum OPG concentrations might inadequately inhibit RANKmediated bone resorption explaining the presence of excessive ossification in AS patients (such as paravertebral syndesmophytes). It has also been explained that high OPG levels in AS was a compensatory manifestation of the anti-bone resorption . Serum RANKL levels were significantly higher in AS patients compared with normal controls, further validated that patients with AS was prone to osteoporosis (Klingberg et al., 2012). RANKL interacts with its receptor RANK as a trigger of downstream signaling pathways for osteoclastogenesis. These signaling pathways include three mitogen-activated protein kinases (p38 MAPK, ERK, and JNK), AKT, and auto amplified nuclear factor of activated T cells, cytoplasmic 1 (NFATc1), which lead to stimulate the activation of critical genes for osteoclastogenesis . These are the potential specific molecular mechanisms of bone loss in the early stages of AS. RANKL/OPG ratio shows a significant difference between AS patients and normal controls. An earlier study in RA demonstrated that the imbalance of RANKL/OPG ratio plays an important role in bone metabolism (Haynes et al., 2003). Radiographic damage in early stage of AS is initially characterized by erosive changes followed by a distinct anabolic skeletal response, which results in excessive bone formation (Lories et al., 2009). Accordingly, early stage of AS was prone to osteo-porosis，while AS with long disease duration was prone to paravertebral syndesmophytes.
When stratified by ethnicity, three biomarkers of bone metabolism showed no connections with AS risk in Caucasian, but existed in Asians. This may be related to ethnic differences in the predominant HLA-B27 alleles associated AS in which HLA-B27*05 and HLA-B27*02 are widely prevalent in European populations, and HLA-B27*04 is predominate in Asian populations (Bowness, 2015). The association between different HLA-B27 subtypes with AS patients is different. It is now widely believed that HLA-B27*02, HLA-B27*04, and HLA-B27*05 had positive associations with AS, but HLA-B27*06 and HLA-B27*09 had negative associations (Park et al., 2008;Lin and Gong, 2017). Patients with elevated ESR (ESR >20 mm/h) had higher serum OPG levels when compared to controls, suggesting that race and inflammatory factors may be associated with OPG. Similarly, Asanuma et al. also found a positive correlation between serum OPG levels and ESR (Asanuma et al., 2007). Serum RANKL levels were significantly higher in patients with elevated ESR (ESR >20 mm/h), high CRP (CRP >10 mg/L), short duration of disease (duration of disease ≤8 years), and high BASDAI score (BASDAI score >4 points) compared to healthy controls. Patients with higher bone erosion had short disease duration and higher inflammatory activity, because early stage of AS is acute stages of inflammation marked by mononuclear cell infiltrates and an increased number of osteoclasts (OCs) (Caparbo et al., 2018). Given these facts, we hypothesized that serum OPG and RANKL levels may be used as potential biomarkers to reflect inflammation and bone metabolism in AS.
Heterogeneity was significant in this meta-analysis, and a meta-regression was conducted to further explore the source of heterogeneity. However, the result of metaregression does not reveal the source of heterogeneity.
Egger's (p= 0.002) and Begg's test (p = 0.013) found publication bias in RANKL. Then, trim-and-fill method was used to correct the result, and two potential missing studies were added in the lower side of the funnel plot to make the plot symmetric ( Figure 4). However, the correction did not significantly reduce the serum RANKL levels in AS patients compared with controls (SMD = 0.693, 95%CI = 0.056-1.329, p = 0.033). This is the first meta-analysis to comprehensively evaluate the role of serum OPG, RANKL levels and RANKL/OPG ratio in the pathogenesis of AS. This study also had some limitations. First, we only included online published articles in this meta-analysis, some gray literature have not been taken into account and were missed. A significant publication bias has been found in this meta-analysis, but trim-and-fill method indicated that the publication bias did not significantly influence the pooled results, suggesting the results were robust. Second, we did not investigate the correlation of other factors, such as sex, smoking, body mass index (BMI), and drug use, with AS risk due to lack of availability of data. Meta-regression did not reveal any source of heterogeneity for OPG and RANKL, suggesting other uninvestigated factors may also contribute to heterogeneity. Third, the levels of OPG and RANKL affect the capacity to generate osteoclasts leading to the onset of AS, but the capacity to generate osteoclasts has not been evaluated in our study. Therefore, we cannot draw such a conclusion that the changes in the levels of OPG and RANKL result in the occurrence of AS by affecting the capacity to generate osteoclasts.

Conclusion
In conclusion, serum levels of OPG, RANKL, and RANKL/OPG ratio may be used as potential susceptible biomarkers for AS, but they could be influenced by race, inflammatory factors, and disease activity of AS patients.