Objectives We aimed to identify soluble biomarkers that differentiate psoriatic arthritis (PsA) from osteoarthritis (OA).
Methods Markers of cartilage metabolism (cartilage oligomeric matrix protein [COMP], hyaluronan), metabolic syndrome (adiponectin, adipsin, resistin, hepatocyte growth factor [HGF], insulin, leptin) and inflammation (C-reactive protein [CRP], interleukin-1β [IL-1β], IL-6, IL-8, tumour necrosis factor alpha [TNFα], monocyte chemoattractant protein-1 [MCP-1], nerve growth factor [NGF]) were compared in serum samples from 201 patients with OA, 77 patients with PsA and 76 controls. Levels across the groups were compared using the Kruskal-Wallis test. Pairwise comparisons were made with Wilcoxon rank-sum test. Multivariate logistic regression analyses were performed to identify markers that differentiate PsA from OA. Receiver operating characteristic (ROC) curves were constructed based on multivariate models. The final model was further validated in an independent set of 73 PsA and 75 OA samples using predicted probabilities calculated with coefficients of age, sex and biomarkers.
Results Levels of the following markers were significantly different across the three groups (p<0.001)—COMP, hyaluronan, resistin, HGF, insulin, leptin, CRP, IL-6, IL-8, TNFα, MCP-1, NGF. In multivariate analysis, COMP (OR 1.24, 95% CI 1.06 to 1.46), resistin (OR 1.26, 95% CI 1.07 to 1.48), MCP-1 (OR 1.10, 95% CI 0.07 to 1.48) and NGF (OR<0.001, 95% CI <0.001 to 0.25) were found to be independently associated with PsA versus OA. The area under the ROC curve (AUROC) for this model was 0.99 compared with model with only age and sex (AUROC 0.87, p<0.001). Similar results were obtained using the validation sample.
Conclusion A panel of four biomarkers may distinguish PsA from OA. These results need further validation in prospective studies.
- psoriatic arthritis
Statistics from Altmetric.com
What is already known about this subject?
Patients with osteoarthritis (OA) may present similar to patients with psoriatic arthritis (PsA) with distal interphalangeal joint involvement and back pain.
What does this study add?
The study found that markers of cartilage metabolism, metabolic syndrome and inflammation are differentially expressed in patients with PsA, OA, and healthy controls. A panel of four biomarkers, cartilage orligometric matrix protein, resistin, monocyte chemoartractant protein-1 and nerve growth factors provdie addotopma; discriminatory value in distinguishing psA from OA form that provided by age sex.
How might this impact on clinical practice or future developments?
These biomarkers taken together with other features may provde a valuable tool for diagnosis and management of patients with OA and PsA.
Psoriatic arthritis (PsA) is an immune-mediated inflammatory musculoskeletal disease associated with joint, ligament and tendon pain, stiffness and swelling that leads to damage to the peripheral, axial and entheseal structures, resulting in reduced quality of life and life expectancy.1 2 The point of entry into the healthcare system for individuals affected by PsA is often through family physicians or dermatologists. Physicians with limited experience in this area may find it difficult to diagnose PsA, which can be confused with osteoarthritis (OA) due to the involvement of the large joints of the lower extremities and proximal (PIP) and distal interphalangeal (DIP) joints of the hands as well as often normal blood levels of acute phase reactants in both conditions.3 The identification of soluble biomarkers that distinguish PsA from OA can aid in the development of tools for physicians to better identify patients with PsA.
We hypothesised that soluble markers of cartilage metabolism, metabolic syndrome and inflammation may differentiate patients with PsA from those with OA. We aimed to detect biomarkers associated with PsA by measuring serum levels of 15 markers of cartilage metabolism (cartilage oligomeric matrix protein [COMP], hyaluronan), metabolic syndrome (adiponectin, adipsin, resistin, hepatocyte growth factor [HGF], insulin, leptin) and inflammation/immune response (C-reactive protein [CRP], interleukin-1β [IL-1β], IL-6, IL-8, tumour necrosis factor alpha [TNFα], monocyte chemoattractant protein [MCP-1], nerve growth factor [NGF]) in patients with PsA, OA and healthy controls (HC).
Serum samples were obtained from the biobank of the University of Toronto PsA Program (PsA) and University Health Network Arthritis Program (OA). All patients with PsA had psoriasis confirmed by a dermatologist and they satisfied the Classification for Psoriatic Arthritis criteria.4 Patients with OA did not have inflammatory arthritis and HCs were healthy volunteers. All HCs completed screening questionnaires to rule out the presence of autoimmune conditions and family history of psoriasis and PsA. The discovery set included 77 PsA, 201 OA and 76 HC. An independent set of 73 PsA and 75 OA patient samples was included for validation set. No patient with PsA was receiving treatment with a biological agent at the time of serum collection, but may have received these agents after the sample was drawn. Samples were obtained at the time of knee or hip joint replacement surgery (OA) or at the time of clinical assessment (PsA, HC), and stored at −80°C until laboratory assays were conducted.
Serum ELISA and multiplex assays
Markers of cartilage metabolism (COMP and hyaluronan) were measured using Quantikine immunoassays according to the manufacturer’s instructions (R&D Systems, Minneapolis, MN). Markers of metabolic syndrome (adiponectin, adipsin, resistin, HGF, insulin and leptin) and inflammation (CRP, IL-1β, IL-6, IL-8, TNFα, MCP-1 and NGF) were measured using magnetic bead Luminex panels (EMD Millipore, Billerica, MA) according to the manufacturer’s instructions. Data were acquired using the Luminex 200 system (Luminex, Austin, TX) and analysed with the Bio-Plex Manager software (Bio-Rad Laboratories, Hercules, CA). All samples were measured in duplicate and marker levels were quantified in relation to a fivefold serially diluted standard provided with the kits, using a five-parameter logistic regression curve.
Marker levels in serum were compared across the three groups in the discovery set using the Kruskal-Wallis test. Additional pairwise comparisons were made with Wilcoxon rank-sum test to identify the source of differences. In order to identify markers that could differentiate patients with PsA from patients with OA, we performed multivariate logistic regression analysis adjusting for age and sex. At the first stage, the regression models included age, sex and 1 of the 15 markers as covariates; markers having p≤0.1 were entered into a multivariate model in the second stage and backward elimination was carried out to obtain a reduced model. Further, internal cross-validation was performed using the PsA and OA samples. The samples were divided into three subsamples within each set (PsA: 25, 26, 26; OA: 67, 67, 67) and each pair of subsamples was pooled and used as the training set with the remaining subsamples used as the testing set. The same procedure of variable selection was performed on each training set: a logistic regression with age and sex was fitted for each biomarker; those with p≤0.1 were further considered in a multivariate model adjusting for age and sex; the resulting multivariate model was then simplified using stepwise backward elimination. In the validation set, predicted probabilities were estimated using the linear predictors from the model developed in the discovery set. Logistic regression analyses were also conducted to generate models to compare the interaction between age and biomarkers as well as determine the effects of age and sex alone in both the discovery and validation sets. Discriminative ability was assessed by way of receiver operating characteristic (ROC) curves based on the multivariate models. We compared the area under the curve (AUC) of the model based simply on age and sex with that based on age, sex and marker levels using the DeLong’s test for two correlated ROC curves.
A summary of the demographics and disease characteristics of the study subjects can be found in table 1. The discovery set included 77 patients with PsA, 201 OA and 76 HC. A significant difference in age between the three groups was found (p<0.0001). The validation set consisted of 73 patients with PsA and 75 patients with OA, where patients with OA were significantly older than patients with PsA (p<0.0001). All patients with PsA had active PsA and a total of nine patients (two in discovery and seven in validation set) also had concomitant OA.
Comparison of serum marker levels between PsA, OA and HC
When we compared the levels of the 15 markers between the three groups (PsA, OA and HC), 12 markers were differentially expressed. The marker levels and results of the comparison across and between the subject groups are summarised in table 2. Nine markers (CRP, resistin, HGF, IL-6, IL-8, insulin, MCP-1, TNFα and COMP) were elevated in patients with PsA compared with patients with OA. CRP, HGF, IL-6, leptin, NGF, TNFα and COMP were also increased in PsA compared with HC. Four markers (IL-6, leptin, NGF and hyaluronan) were present in higher levels in patients with OA as compared with HC and two of these markers (NGF and hyaluronan) were also high in patients with OA compared with patients with PsA. In contrast, resistin, HGF, IL-8, insulin and MCP-1 were reduced in OA serum compared with HC.
Four markers differentiate patients with PsA from patients with OA
The results of multivariate logistic regression analysis comparing PsA and OA serum levels in the discovery set are found in table 3. COMP, resistin, MCP-1 and NGF were identified as markers that are significantly different between patients with PsA and OA. The ROC curve constructed using the model with age, sex and the four biomarkers (COMP, resistin, MCP-1, NGF) had an AUC of 0.9984 (figure 1A). The model containing age and sex alone had an AUC of 0.8727. There was a significant difference in the AUC of the two ROC curves (p<0.001). MCP-1 was consistently identified as a PsA marker during cross-validation. The first fold identified HGF, MCP-1 and COMP and the AUC was 0.9896, the second identified resistin, MCP-1, NGF and COMP (AUC 1.000) and the third internal validation identified MCP-1 and NGF (AUC 0.9719), respectively.
Biomarker validation in an independent set
To further validate these findings, the four biomarkers identified in the multivariate model of the discovery set (COMP, resistin, MCP-1 and NGF) were measured in an independent set of patients with OA (n=75) and PsA (n=73) (validation set, table 4). HGF was also measured because it satisfied the criteria for backward elimination in part of the cross-validation of the discovery set. No differences in COMP were observed. Resistin, HGF and MCP-1 were elevated in patients with PsA compared with patients with OA while NGF was reduced, similar to the discovery set. The performance of the multivariate model developed in the discovery set was assessed by calculating predicted probabilities with coefficients of age, sex, COMP, resistin, MCP-1 and NGF. The resulting ROC curve had an AUC of 0.9896 (figure 1B). The model containing age and sex alone had an AUC of 0.8296. There was a significant difference in the AUC of the two ROC curves (p<0.001). Thus, we confirmed that the biomarkers provide additional discriminatory value for discriminating PsA from OA over demographic features (age and sex).
Because of the significant difference in age between the patients with PsA and OA in this study, we computed the logistic regression models to compare the interaction between each of the four biomarkers (COMP, resistin, MCP-1 and NGF) and age to predict disease status. There were no significant interactions between age and the four biomarkers observed in the discovery and validation sets (online supplementary table 1).
OA is the most commonly occurring form of arthritis. Due to the similarities in the distribution of joints involved as PsA and the lack of biomarkers, it may be difficult to distinguish between the two arthritides leading to delayed diagnosis and inappropriate treatment. Both OA and PsA can affect similar joint regions such as the DIP and PIP joints in the hands, cervical and lumbar spine and large joints of the lower limbs.3 Inflammation of the small joints of the hands and feet was in fact difficult to discriminate between PsA and OA using high-resolution MRI.5 New bone formation as osteophytes in OA and syndesmophytes in PsA and joint erosions are typically found in both conditions and are usually distinguished by imaging. However, a variant of OA called diffuse idiopathic skeletal hyperostosis involves the formation of osteophytes that are similar to syndesmophytes in some patients with PsA.6
PsA and OA also share common risk factors and pathological triggers. Trauma and injury are well known to trigger OA and the relationship between trauma and PsA, known as the Koebner’s effect, is well documented.7 Obesity is a primary risk factor for disease onset in OA and is associated with increased risk and severity of arthritis in PsA.8 9 OA and PsA typically commence around age 50, particularly for women affected by OA coinciding with the drop in oestrogen levels.10 11 Imaging studies have revealed an age-related thickening of the normal enthesis and ligaments after age 40.3 12 The enthesis is one of the earliest distinguishable sites of hand OA and enthesitis has been hypothesised to be the initiation site of pathogenesis in PsA.13 14 These overlapping aetiological processes in PsA and OA may contribute to a difficulty in distinguishing these disorders in the clinic
Biomarkers can serve as objective measurements that are indicators of normal biological or pathogenic processes. In this study, the aim was to identify soluble biomarkers that would aid in the distinction between PsA and OA. A panel of four markers (COMP, resistin, MCP-1 and NGF) was found to discriminate patients with PsA from patients with OA. Previous studies have also supported these findings. Ross and colleagues have observed high expression of MCP-1 in the synovium of patients with PsA supported by a positive correlation between T cell numbers in synovial fluid (SF) and MCP-1 levels.15 MCP-1 was also elevated in PsA compared with OA serum,16 which may reflect the chronic inflammatory burden in PsA. NGF has been found in the SF of patients with OA, spondyloarthritis and rheumatoid arthritis.17 Inhibition of NGF has been shown to provide extensive pain relief in OA and two anti-NGF antibodies, tanezumab and fasinumab, have shown tremendous promise in clinical trials for the treatment of pain in OA.18 19 NGF plays a crucial role in the chronic pain of OA20–22 and this was reflected in the consistently high levels of NGF in OA compared with PsA serum in this study. COMP has previously been found to be elevated in PsA compared with control serum and may be a prognostic marker in the preclinical diagnosis of PsA.23–25 In OA, previous studies have pointed to the use of serum COMP as a marker for early cartilage lesions in the knee as it is negatively correlated with OA disease duration.26 27 A recent meta-analysis determined that high levels of COMP increase the probability of developing knee OA.28 Resistin was previously shown to be elevated in PsA compared with control serum and released from OA cartilage, where it correlated with SF expression of IL-6, matrix metalloproteinase-1 (MMP-1) and MMP-3.29 30 Obesity is a common predisposing factor for PsA and OA, although the role of resistin is not well understood. We observed consistently high levels of resistin in patients with PsA compared with patients with OA but not controls, consistent with previous reports.31–33
The large sets of patients with PsA and OA available at the same institution provided a unique opportunity for this study which cannot be easily obtained. The validation of the results in two independent groups of patients and consistently high sensitivity and specificity of the biomarker panel is promising. However, this study is not without limitations. The patients with OA included in this study were those undergoing joint replacement surgeries and may not be representative of the general population of patients with OA. Thus, it is necessary to replicate these results using patients with OA seen in family practice or rheumatology clinics. Also, the effect of other variables such as measurements of cardiovascular health and the use of medications was not accounted for in these findings.
With the development of biological therapies, the inclusion of soluble biomarkers in the diagnosis and treatment of PsA and OA would facilitate a personalised medicine approach to patient care. Anti-TNF treatments have been highly efficacious in PsA but do not relieve symptoms of OA.34–36 The failure of these anti-TNF agents in some patients with PsA may in part be due to a misdiagnosis of OA. Eventually, the availability of a biomarker panel could facilitate more accurate diagnosis to target appropriate therapeutic strategies. However, the clinical feasibility of this biomarker panel needs to be determined. Future studies are needed to evaluate this panel in a prospective cohort of patients.
In conclusion, PsA and OA sometimes affect the same anatomical regions which make them difficult to distinguish in a subgroup of patients. We identified a panel of four biomarkers (COMP, resistin, MCP-1 and NGF) that when combined provide additional discriminatory value to that provided by age and sex alone. These markers were consistently validated internally within the discovery set as well through an independent set of patients. After further verification and validation, these markers could provide a valuable tool for the improved diagnosis and management of both OA and PsA thus resulting in improved patient outcomes.
Handling editor Josef S Smolen
Presented at The abstract for this manuscript (Abstract 1657) has been presented at the American College of Rheumatology (ACR)/AHRP Annual Meeting Poster Session B held on 14 November 2016 (https://acrabstracts.org/abstract/soluble-biomarkers-may-differentiate-psoriatic-arthritis-from-osteoarthritis/?msg=fail&shared=email).
Contributors All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. All authors were likewise involved in the study conception and design, acquisition of data as well as analysis and interpretation of data. DDG had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Funding The University of Toronto Psoriatic Arthritis Program is supported by a grant from the Krembil Foundation.
Competing interests None declared.
Patient consent for publication Obtained.
Ethics approval This study was approved by the University Health Network Research Ethics Board according to the principles of the Declaration of Helsinki.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.