Abstract
Because psoriatic arthritis (PsA) is an inflammatory disease of joints, serum soluble biomarkers specific for chronic joint and bone inflammation may predict future disease severity and response to therapy, thereby informing stratified medicine approaches. The objectives of our systematic review were to determine whether serum soluble bone and cartilage turnover biomarkers are (1) associated with PsA or psoriatic spondyloarthropathy; and (2) associated with disease activity, disease severity, or clinical phenotype. Ten studies met eligibility criteria. Matrix metalloproteinase (MMP)-3, Dickkopf (DKK)-1, macrophage colony-stimulating factor (M-CSF), crosslinked telopeptide of collagen-1, and tumor necrosis factor-related apoptosis-inducing ligand were associated with PsA, with equivocal results for osteoprotegerin (OPG) and bone alkaline phosphatase (ALP). MMP-3, DKK-1, M-CSF, CPII:C2C (ratio of cartilage degradation vs byproduct formation), and possibly OPG were associated with PsA independently of psoriasis. C1-2C (a neoepitope released when type 2 cartilage is degraded by collagenases) was associated with both tender and swollen joint counts, and bone morphogenetic protein-4 with patient global assessment of disease, pain score, and the Bath Ankylosing Spondylitis Disease Activity Index. Bone ALP was associated with disease activity. M-CSF and receptor activator of nuclear factor-κB ligand were associated with several plain radiographic features. No studies have investigated biomarker associations specifically with axial PsA.
Psoriatic arthritis (PsA) is a chronic inflammatory musculoskeletal disorder with characteristic patterns of peripheral and axial joint inflammation and extraarticular manifestations that can include skin psoriasis, psoriatic nail disease, enthesitis, dactylitis, or uveitis. As such, candidate serum soluble biomarkers specific for chronic joint and bone inflammation may predict future disease severity and response to therapy, thereby informing stratified medicine approaches. However, identifying and monitoring biomarkers in PsA is difficult because of the heterogeneity of PsA disease. In PsA, bone loss can occur in the form of bone erosion, osteolysis, and bone mineral density (BMD) loss1. Bone formation can occur in the form of osteoproliferation, ankylosis, and syndesmophytes.
Several bone and cartilage turnover biomarkers might be of interest in PsA. Some directly cause bone resorption through their enzymatic or cytokine properties, e.g., matrix metalloproteinase (MMP)-3 enzymatically degrades the extracellular matrix of bone and cartilage2. Osteoprotegerin (OPG) is a glycoprotein secreted by osteoblasts and stromal cells, acting as a decoy receptor to receptor activator of nuclear factor-κB ligand (RANKL), thereby inhibiting osteoclastogenesis, resulting in reduced bone resorption. Others are byproducts of bone resorption, thereby acting as markers of the process, e.g., crosslinked telopeptide of collagen (CTX)-1 is the product of excess metalloproteinase degradation of type 1 collagen. There are several byproducts of cartilage turnover: C2C and C1-2C are neoepitopes that are released when type 2 cartilage is degraded by collagenases; CPII is released during procollagen 2 synthesis; and CPII:C2C is the ratio of cartilage degradation versus byproduct formation3.
Although there have been several editorial review articles, to the best of our knowledge, there have been no systematic reviews published on the clinical and prognostic value of serum soluble bone turnover biomarkers in PsA. The objectives of this systematic review were to determine whether serum soluble bone and cartilage turnover biomarkers are (1) associated with PsA or psoriatic spondyloarthropathy (PsSpA); and (2) associated with disease activity, disease severity, or clinical phenotype in PsA cases versus healthy controls, and PsA versus cutaneous psoriasis without arthritis (PsC).
MATERIALS AND METHODS
Methods of analysis and eligibility criteria were specified in advance and documented in an a priori protocol. Our study aligns with “The PRISMA Statement for Reporting Systematic Reviews and Meta-Analyses of Studies That Evaluate Health Care Interventions”4.
Inclusion criteria
We included cohort, case-control, cross-sectional studies and randomized, controlled trials published in the form of a journal paper, journal abstract, or conference abstract that compared the serum concentration of bone and cartilage turnover biomarkers in PsA cases to that in healthy controls, with or without an additional PsC comparator group.
Study participants with PsA must have fulfilled classification criteria for PsA (Classification for Psoriatic Arthritis5, or Moll and Wright6) or PsSpA7.
The following bone and cartilage turnover biomarkers were included [as defined in Medical Subject Headings (MeSH), EMTree, or key terms]: OPG, MMP-3, sclerostin, Dickkopf (DKK)-1, bone alkaline phosphatase (ALP), osteocalcin (OC), macrophage colony-stimulating factor (M-CSF), RANKL, collagen type II, extracellular matrix proteins, glycoproteins, procollagen, amino-terminal propeptide of procollagen type III (PIIINP), CTX-1, tumor necrosis factor-related apoptosis-inducing ligand (TRAIL), bone morphogenetic protein (BMP), and cartilage oligomeric matrix protein (COMP).
Outcome variables were (1) peripheral and/or spinal involvement: clinical symptoms, radiographic disease; (2) disease severity as measured by axial and/or peripheral radiographic disease; and (3) disease activity: tender joint counts, swollen joint counts, Bath Ankylosing Spondylitis Disease Activity Index (BASDAI), Bath Ankylosing Spondylitis Functional Index (BASFI), Bath Ankylosing Spondylitis Metrology Index, enthesitis, C-reactive protein (CRP), composite scores, and other outcome measures.
Exclusion criteria
We excluded studies without a healthy control (HC) group. We excluded studies in which participants were being treated with biological agents, because tumor necrosis factor inhibitors8,9,10, but not conventional disease-modifying antirheumatic drugs (DMARD)10,11, have been reported to directly influence serum bone and cartilage turnover biomarkers12, thereby confounding results when comparing HC with PsA cases.
Searches
The search date was February 1, 2014. The following databases were searched using key indexing terms: Medline (1950–present), Embase (1974–present), and the Cochrane Controlled Trials Register (1993–present). The following MeSH, EMTree, or key term stems were used: arthritis, psoriatic, psoriatic spondyloarthritis, biological markers, and bone turnover markers. No language restrictions were applied to publications.
The reference lists of all papers fulfilling inclusion criteria and all review articles were scrutinized for any references not identified in the original database search, but still meeting the inclusion criteria. Two key authors (VC, OF) were contacted to determine whether any important unpublished or unindexed papers (e.g., conference proceedings) should be screened.
Study selection
Two reviewers (DJ, RS) independently assessed abstracts for inclusion in the review. Where there was disparity in opinion, the full paper was obtained and consensus for inclusion or exclusion was reached (DJ, RS). An assessment was made at this point for potential publication bias or selective reporting within studies. Two reviewers (DJ, AN) independently extracted data from the papers onto a standardized data-extraction Excel spreadsheet that was initially pilot-tested. The papers were critically appraised using the Critical Appraisal Skills Programme toolkit13 for cohort and cross-sectional studies, including sources of bias, both at study and outcome level. DJ and AN reached consensus on the data for use in the subsequent analyses.
Synthesis of results
The primary summary measure was OR for serum biomarker levels in PsA versus healthy controls or PsA versus PsC, including p values for the analyses. The secondary summary measures were OR or Spearman rho correlation coefficient for clinical outcomes in PsA versus healthy controls or PsA versus PsC, including 95% CI and p values for the analyses.
We initially intended to combine the results of different studies mathematically as a metaanalysis, including tests for heterogeneity. However, because of the differing methods used in the included studies (cohorts, laboratory techniques using ELISA or immunoassays with different reference ranges, statistical analyses) and a lack of homogeneity in the reporting of results, it was not possible to combine the results of different studies statistically. Therefore, we have reported the results of the studies quantitatively, without metaanalysis.
RESULTS
Search results
There were 155 unique studies identified; 10 of these met the eligibility criteria and were included in the systematic review. Two papers14,15 that met inclusion criteria were unobtainable from several libraries (including the British Library) or on contacting the first authors, and insufficient detail of results were provided in the abstract to allow inclusion in the systematic review. Four papers were excluded because they did not have a healthy control group for comparison with the PsA group16,17,18,19. Eight papers were excluded because the PsA cases were using biological agents, and either did not have a healthy control comparator group or did not provide prebiological initiation biomarker data8,9,10,12,19,20,21,22. The remaining 131 articles were excluded because they did not fulfil several eligibility criteria. No further articles were identified on scrutinizing the reference list of included articles or by the recommendation of the 2 key authors (VC, OF).
Figure 1 details the flow of study selection in the systematic review, and Table 1 summarizes the characteristics of the 10 studies meeting the eligibility criteria of the systematic review.
Study selection in the systematic review. MeSH: Medical Subject Headings; DJ and RS: independent reviewers.
Characteristics of the 10 studies meeting eligibility criteria of the systematic review.
Comparison of biomarkers levels in PsA cases versus healthy controls
The results of comparisons between biomarkers levels in PsA cases versus healthy controls are shown in Table 2. The serum concentration of several biomarkers (MMP-3, DKK-1, M-CSF, CTX-1, and TRAIL) was significantly higher in PsA versus healthy controls, whereas the serum concentration was not significantly different in PsA versus healthy controls for RANKL, BMP, OC, PIIINP, COMP, C1-2C, and CPII:C2C. The results for OPG and ALP were equivocal.
Comparison of biomarkers levels in PsA cases vs healthy controls.
Comparison of biomarker levels in PsA cases versus PsC cases
The results of comparisons between biomarkers levels in PsA versus PsC are given in Table 3. The serum concentration of MMP-3, DKK-1, M-CSF, and CPII:C2C was significantly higher in PsA versus PsC. The results for OPG were equivocal.
Comparison of biomarkers levels in PsA cases vs PsC.
Association of biomarker levels with demographic variables
Franck and Ittel28 demonstrated 2 biomarkers to be higher in male versus female patients with PsA: ALP (mean serum concentration 137 U/l in males vs 91 U/l in females, p < 0.05) and OC (mean serum concentration 3.62 ng/ml in males vs 2.28 ng/ml in females, p < 0.05). However, Hofbauer, et al25 did not corroborate the findings for OC, demonstrating OC levels to be no different in male and female patients with PsA (23.7 ng/ml in males vs 23.1 ng/ml in females, p = 0.82). However, the populations of the 2 studies differed, with Hofbauer, et al excluding patients treated with DMARD or corticosteroids and taking fasting blood samples, whereas Franck and Ittel took unfasted samples and included patients treated with DMARD and corticosteroids. Similarly, the association between OPG levels and sex were conflicting, with higher levels in females versus males in the study by Hofbauer, et al (6.7 pmol/l in females vs 2.09 pmol/l in males, p = 0.001)25, but no difference by sex in the smaller study by Dalbeth, et al (mean serum concentrations or p values not stated)11. No correlation has been reported between sex and DKK-111, M-CSF11, RANKL11, PIIINP25,29, or cross-laps25.
No association has been reported for PIIINP and age of patient with PsA at the time of sampling (p = 0.925)11,29, DKK-1, RANKL, M-CSF, or OPG (p values or Spearman rho correlations not stated) and body weight in kg11.
Association of biomarker levels with clinical variables
Three studies investigated the association between PsA disease duration and serum biomarkers26,28,29. However, none of the studies defined whether duration was analyzed as a continuous or categorical variable. Disease duration was positively associated with serum CTX-1 concentrations (r = 0.670, p = 0.009)26, but not with OPG26, ALP26, PIIINP29, or OC26,28.
Chandran, et al demonstrated a positive correlation between C1-2C and both tender joint counts and swollen joint counts10. However, p values, Spearman rho correlation coefficients, and tender or swollen joint counts per unit increase in C1-2C were not stated. Grcevic, et al reported a positive association between BMP-4 and both patient global assessment of disease (r = 0.54, p = 0.02) and pain score on a visual analog scale (r = 0.49, p = 0.04)23. No such associations were found between the same variables and either BMP-2 or BMP-623.
Association of biomarkers levels with laboratory variables
An association was demonstrated between CRP levels and both CTX-126 and TRAIL25 in patients with PsA, but not with either MMP-327 or OPG25. Erythrocyte sedimentation rate (ESR) was positively associated with both CTX-126 and OPG25. TRAIL was not associated with ESR levels25. Two studies consistently showed ALP to be positively associated with OC28,30. Serum creatinine levels were not associated with DKK-1, RANKL, M-CSF, or OPG levels in the single study that tested for this correlation11.
Association of biomarkers levels with composite indices
Franck and Ittel reported an association between disease activity and both ALP (mean serum concentration in patients with “no disease activity” 69 U/l vs 148 U/l in patients with “high” disease activity, p < 0.005) and OC (mean serum concentration in patients with “no disease activity” 2.2 ng/ml vs 3.92 ng/ml in patients with “high” disease activity, p < 0.05), although the numbers of patients in these groups were very small (4 vs 98, respectively) and no definition of “no” versus “high” disease activity was stated28. Disease Activity Score at 28 joints using CRP (DAS28-CRP) was not associated with DKK-1, M-CSF, RANKL, OPG11, BMP-2, BMP-4, or BMP-623. BMP-4 was associated with BASDAI in the 1 study that tested this correlation (r = 0.46, p = 0.04)23. No association was demonstrated between BASDAI and either BMP-2 or BMP-623. COMP positively correlated with the Psoriasis Area and Severity Index (PASI) in 1 study (analyses not stated in paper)10. No association was demonstrated between DAS28-CRP and DKK-1, M-CSF, RANKL, OPG, BMP-2, BMP-4, or BMP-611. No association was demonstrated between BASFI and BMP-2, BMP-4, or BMP-623.
Association of biomarkers levels with radiographic variables
Four studies10,11,23,25 investigated the association between biomarker levels and radiographic variables, with 2 studies providing the majority of the data10,11 (Table 4). Joint space narrowing was associated with RANKL (p < 0.05) and M-CSF (p < 0.01), but not with DKK-1 or OPG in the 1 study testing these associations11. Similarly, osteolysis (defined as pencil-in-cup deformity) was associated with both RANKL (p < 0.05) and M-CSF (p < 0.05), but not with DKK-1 or OPG in the 1 study testing these associations11. Osteoproliferation was not associated with RANKL, M-CSF, DKK-1, or OPG in the 1 study testing these associations11. Two studies tested for association between peripheral radiographic erosions and serum biomarkers. M-CSF was positively associated with peripheral erosions (p < 0.001)11, but the remaining biomarkers were not (MMP-3, DKK-1, RANKL, OPG, COMP, C2C, C1-2C, CPII)10,11. The modified van der Heijde score for PsA (VDH) is a composite score encompassing joint space narrowing and erosions in peripheral radiographs31. Dalbeth, et al demonstrated an association between the VDH and both M-CSF (p < 0.01) and RANKL (p < 0.05), but not with DKK-1 or OPG11.
Association of biomarkers levels with radiographic variables.
Three studies10,11,23 tested for association between radiographic sacroiliitis and several biomarkers; no associations were found with MMP-3, RANKL, OPG, BMP-2, BMP-4, BMP-6, COMP, C2C, C1-2C, and CPII, and results were equivocal for DKK-1 and M-CSF10,11,23.
No association was found between BMD at the hip, lumbar spine or femur, and several biomarkers, including DKK-1, RANKL, M-CSF, OPG, or TRAIL (Table 4)25.
DISCUSSION
A summary of biomarkers that are associated with PsA and its clinical variables is shown in Figure 2. The following biomarkers were associated with PsA: MMP-3, DKK-1, M-CSF, CTX-1, and TRAIL, and the results were equivocal for OPG and ALP. MMP-3, DKK-1, M-CSF, CPII:C2C, and possibly OPG were associated with PsA independently of PsC. ALP was associated with male sex in PsA. CTX-1 was associated with disease duration, C1-2C with both tender and swollen joint counts, and BMP-4 with both patient global assessment of disease and pain score. CRP was associated with both CTX-1 and TRAIL, ESR with both CTX-1 and OPG, and ALP with OC. Disease activity was associated with ALP and possibly OC, BASDAI was associated with BMP-4, and skin score (PASI) correlated with COMP. The following biomarkers were associated with radiographic features: M-CSF with the PsA-modified VDH composite score, joint space narrowing, peripheral radiographic erosions, and osteolysis; RANKL with the VDH composite score, joint space narrowing, and osteolysis.
Summary of serum soluble bone and cartilage turnover biomarkers showing association with psoriatic arthritis. CTX-1: crosslinked telopeptide of collagen-1; BMP-4: bone morphogenetic protein-4; C1-2C: a neoepitope released when type 2 cartilage is degraded by collagenases; ALP: bone alkaline phosphatase; TRAIL: tumor necrosis factor-related apoptosis-inducing ligand; OPG: osteoprotegerin; OC: osteocalcin; MMP-3: matrix metalloproteinase-3; DKK-1: Dickkopf-1; M-CSF: macrophage colony-stimulating factor; COMP: cartilage oligomeric matrix protein; RANKL: receptor activator of nuclear factor-κB ligand; PsA: psoriatic arthritis; CPII:C2C: ratio of cartilage degradation vs byproduct formation.
Disparity in study findings
There are several potential reasons for disparity in study results. First, most of the studies have been cross-sectional rather than prospectively conducted cohort studies. A study by Young-Min, et al of early rheumatoid arthritis (RA) demonstrated that biomarkers are associated with swollen/tender joint counts and DAS only when longitudinal data were analyzed and not when cross-sectional baseline data were analyzed32. Studies have investigated differing clinical variables and used differing collection protocols (overnight fasted in 2 studies25,30) and laboratory techniques, making comparison difficult. Several studies have small samples23,24,26,29,30,33, and likely were underpowered.
We acknowledge that there may be publication bias toward studies with positive results. However, we suspect that because of several biomarkers being reported in each study, with a mixture of both positive and negative findings, that selective reporting bias may be less of a problem.
While it is common practice in studies of metabolic bone disease to test bone markers in the morning and in a fasted state, few PsA studies have undertaken testing in this manner. Clowes, et al investigated the effect of feeding versus fasting on several markers in 20 women and demonstrated little effect on bone biomarkers, except in the case of serum CTX34. Other factors influencing serum levels were circadian rhythm35, sex, oral contraceptive pill use, menstrual cycle, growth, diet, meal composition, and timing of sample after ingestion34.
Standardization of diet prior to sampling may improve measurement variability, but at the expense of feasibility.
Conflicting reports of associations between serum biomarker concentrations and demographic variables may be, in part, attributable to uncontrolled confounding because of a lack of matching or adequate adjustment for age and sex within the study design and analysis. Dalbeth, et al did not adjust for the higher proportion of women in their HC versus PsA group11. Hofbauer, et al had entirely men in its HC group, because this was a “convenience sample” derived from participants in a coronary artery study25. They also reported higher OPG serum concentration in women compared with men, because of a lack of adjustment for sex; estrogen is known to stimulate OPG production24. Ribbens, et al sex-matched their participants, and because corticosteroid use alters MMP-3 levels, analyses were made only in patients not treated with corticosteroids33.
Sharif, et al did not state the source of their HC group, which appears much younger than the PsA cohort29. Significant differences in mean age, sex, and disease duration of the patients in all 3 disease groups were noted. While Shibata, et al24 matched for age in their study, further inspection demonstrates that the HC group was much younger than both the PsA and PsC groups24.
Priorities for future research
Biomarker identification in PsA may help identify patients with PsC with subclinical arthritis and aid both prognostication and stratified medicine approaches. Biomarkers may facilitate monitoring of disease activity and treatment response, so that nonefficacious treatment is switched rather than waiting several years for radiographic progression. Our knowledge of the pathogenesis of PsA, and how it overlaps with ankylosing spondylitis and RA, may be improved through such research. Biomarkers may guide the development of new drugs, both to obtain proof of principle in an early stage of drug development and avoid reliance on slow structural damage outcomes requiring lengthy clinical trials22. Serum biomarkers may offer a more economic and readily available alternative to imaging. All such knowledge is important for the individual patient, public health, and health policy.
Despite the theoretical advantages, “novel” biochemical markers have not translated to the bedside. This may be in part attributable to a lack of longitudinal prospective studies and robust evidence of superiority over existing biomarkers, e.g., CRP. PsA is a heterogeneous disease with several subphenotypes, varied clinical course, and often comorbidities that can confound the interpretation of results36.
There is a need for longitudinal studies to identify biomarkers that correlate with or predict longterm clinical, radiographic, and functional outcomes, and treatment response. Research will be most valuable if it identifies biomarkers that fulfill the Outcome Measures in Rheumatology Clinical Trials filter: truth, discrimination, and feasibility37. It is likely that a panel of biomarkers, rather than a single biomarker, will achieve this12,38.
Acknowledgment
We thank Dr. Vinod Chandran (VC) and Professor Oliver Fitzgerald (OF) for providing their expert input to ensure no important articles were omitted during the systematic search. We thank Jason Ovens (Head of Library and Knowledge Services at the Royal United Hospital, Bath) for facilitating the systematic search and sourcing articles.
Footnotes
-
Supported by an unrestricted educational grant from Pfizer Pharmaceuticals.
- Accepted for publication September 2, 2014.