Synovial Fluid Biomarkers in Knee Osteoarthritis: A Systematic Review and Quantitative Evaluation Using BIPEDs Criteria

Abstract

Objective

The aim of this systematic review was to analyze the evidence about the efficacy of the several synovial fluid (SF) biomarkers proposed for knee osteoarthritis (OA), categorizing them by both molecular characteristics and clinical use according to the BIPEDs criteria, to provide a comprehensive and structured overview of the current literature.

Design

A systematic review was performed in May 2020 on PubMed, Cochrane Library, and Embase databases about SF biomarkers in patients with knee OA. The search was limited to articles in the last 20 years on human studies, involving patients with knee OA, reporting SF biomarkers. The evidence for each selected SF biomarker was quantified according to the 6 categories of BIPEDs classification.

Results

A total of 159 articles were included in the qualitative data synthesis and 201 different SF biomarkers were identified. Among these, several were investigated multiple times in different articles, for a total of 373 analyses. The studies included 13,557 patients with knee OA. The most promising SF biomarkers were C4S, IL-6, IL-8, Leptin, MMP-1/3, TIMP-1, TNF-α, and VEGF. The “burden of disease” and “diagnostic” categories were the most represented with 132 and 106 different biomarkers, respectively.

Conclusions

The systematic review identified numerous SF biomarkers. However, despite the high number of studies on the plethora of identified molecules, the evidence about the efficacy of each biomarker is supported by limited and often conflicting findings. Further research efforts are needed to improve the understanding of SF biomarkers for a better management of patients with knee OA.

Introduction

Osteoarthritis (OA) is a degenerative joint disease characterized by progressive deterioration and loss of articular cartilage with concomitant structural and functional changes in the entire joint, including synovium, meniscus (in the knee), periarticular ligaments, and subchondral bone. ¹ Clinical features of OA are mostly determined by signs and symptoms of inflammation including pain, effusions and secondarily resulting in stiffness and loss of mobility, often associated with significant functional impairment. ² The diagnosis of OA is currently based on clinical symptoms and radiographic criteria (e.g., joint space narrowing, osteophytes, subchondral sclerosis). ³ However, these parameters are difficult to detect at early stages of the disease and most often are recognized only in the advanced stages; thus, the diagnosis of OA is often delayed, when joint tissue destruction is irreversible and conservative treatments are less effective. ⁴ Moreover, monitoring the progression, as well as therapeutic management of OA, are also complex. Patients are followed by radiographs and magnetic resonance to evaluate the imaging progression of joint degeneration, and there is no examination that can predict the effects and/or the benefits of the current treatments used for OA, such as exercise, nonsteroidal anti-inflammatory drugs (NSAIDs), or injective therapies with corticosteroids, hyaluronic acid, or platelet-rich plasma.^5,6 Considering these limitations, alternative parameters, that is, biomarkers, that may be more sensitive in early stages of OA have been pursued as diagnostic tools and to evaluate disease progression. ⁷

A biomarker is defined as “a characteristic that is objectively measured and evaluated as an indicator of normal biological processes, pathogenic processes, or pharmacologic responses to a therapeutic intervention,” ⁸ and it can be represented by radiographic, histologic, physiologic, or molecular characteristics. ⁹ In particular, molecular biomarkers are widely studied in knee OA, as their values could reflect dynamic and quantitative changes in joint remodeling and therefore disease progression. The BIPEDs classification has been introduced to identify key parameters for the evaluation and qualification of the utility of these biomarkers, aimed at providing a common language and structure to communicate knowledge and advances related to biomarkers for both clinical and research applications. ¹⁰ Accordingly, biomarkers in knee OA can be extracted from serum, urine, or synovial fluid (SF) samples, and significant research efforts have been spent to identify the most suitable molecules, though often controversial findings on their efficacy have been reported among the different studies in the literature. ¹¹ In this light, while serum and urine parameters are prone to systemic influences that hinder the identification of local changes, measuring biomarkers in SF of the knee could provide more information about the condition of the diseased joint, being a direct representation of the changes in the intraarticular environment. Moreover, SF biomarkers are in higher concentrations than in blood or urine, thus able to prove higher sensitivity and specificity. ¹² For this reason, evidence on several SF biomarkers is rapidly increasing, with many recent publications aimed either at confirming the potential of previous preliminary findings or identifying new molecules as suitable knee OA biomarkers.

The aim of this systematic review was to analyze the evidence about the efficacy of the several SF biomarkers proposed for knee OA, categorizing them according to both molecular characteristics and clinical use by applying the BIPEDs criteria, in order to provide a comprehensive and structured overview of the current evidence about SF biomarkers in knee OA.

Methods

Search Strategy and Screening Criteria

A systematic review was performed on the literature about SF biomarkers in patients with knee OA. The search was conducted on PubMed, Cochrane Library, and Embase databases on May 18, 2020, by using the following parameters: “synovial fluid AND biomarker AND knee AND osteoarthritis.” Literature search was limited to the last 20 years (2000-2020). The PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analysis) guidelines were used; ¹³ a flowchart of the study selection for qualitative data synthesis is reported in Figure 1 . The screening process was conducted separately by 2 independent observers (AB and GM). In the first step, the articles were screened by title and abstract. The following inclusion criteria for relevant articles were used during the initial screening of titles and abstracts: human studies, involving patients with knee OA, reporting SF biomarkers, written in English language. Exclusion criteria were articles written in other languages, preclinical studies, biomarkers from other samples (e.g., serum, urine), studies reporting other diseases and other joints without evaluating knee OA patients, and reviews. In the second step, the full texts of the selected papers were screened according to the previously mentioned inclusion/exclusion criteria. Moreover, the reference lists from the selected papers and from the systematic reviews found during the screening were also checked, and all selected studies were included in the qualitative data synthesis.

Figure 1.

Flowchart of the study selection.

Classification of Biomarkers

Afterwards, selected SF biomarkers were categorized according to Lotz et al., ⁷ based on their nature and molecular origin, into 4 classes: (1) biomarkers associated with the metabolism of collagen in cartilage (type II collagen) or subchondral bone (type I collagen); (2) biomarkers associated with the metabolism of aggrecan in cartilage; (3) biomarkers related to a range of non-collagenous proteins that have a role in other metabolic pathways in the joint, including glycoproteins, proteoglycans, metalloproteinases, and advanced glycation end-products, as well as hyaluronan, which is a constituent of both cartilage and synovium; and (4) biomarkers associated with other processes, such as inflammation or fibrosis.

Moreover, all SF biomarkers were classified in the 6 categories of BIPEDs classification and each of these was summarized in separate paragraphs and tables. BIPEDs classification was initially proposed by Bauer et al. in 2006 ¹⁰ (later modified by Kraus et al. ¹⁴ ) to provide specific biomarker definitions with the goal of improving the ability to develop and analyze OA biomarkers and to communicate these advances within a common framework. In detail, 6 categories of biomarkers have been identified:

• Burden of disease biomarkers: assess the severity or extent of disease, typically at a single point in time, among individuals with OA. Studies of these markers require comparison with one or more gold standard methods of determining disease severity, such as clinical or radiographic criteria (i.e., Kellgren-Lawrence classification).

• Investigative biomarkers: classified investigative if there is insufficient information to allow inclusion into one of other categories.

• Prognostic biomarkers: a prognostic biomarker predicts OA onset in patients without the disease or OA progression in those with existing disease. Their evaluation requires longitudinal studies showing an association of the marker at baseline with the risk of development or progression of OA.

• Efficacy of intervention biomarkers: those that can be applied in randomized controlled trials to assess the changes associated with pharmacological treatments. These markers can be indicative or predictive of treatment efficacy.

• Diagnostic biomarkers: a diagnostic biomarker can distinguish between individuals with and without OA with good sensitivity and specificity. Studies of these markers must include individuals with and without OA, and those with conditions sometimes confused with OA, such as rheumatoid arthritis.

• Safety biomarkers: those able to reflect tissue and/or organ toxicity following a specific treatment, to monitor for local and systemic adverse effects both early and advanced, to set therapeutic dosages that do not impact on physiology and to understand what “safe” ranges are for joint tissue biomarkers.

According to this classification, a single biomarker may meet multiple criteria for different uses; thus, biomarkers have been classified into one or more categories.

Results

In the initial search, 515 studies were identified. Subsequently, all abstracts were screened and selected according to the inclusion and exclusion criteria ( Fig. 1 ): 329 abstracts were excluded and 5 articles were identified through the reference lists, which gave a total of 191 full-text articles assessed for eligibility. Afterwards, further 32 articles were excluded, either because they were reviews, studies not written in English, studies that described biomarkers not obtained from SF samples, or studies that included other diseases without reporting about OA. Thus, 159 articles^12,15-172 were included in the qualitative data synthesis; among these, 138 within cross-sectional and 21 within longitudinal studies.

The studies included 13,557 patients. A total of 789 patients with knee injury (mainly meniscal or anterior cruciate lesions), 755 controls (donors or volunteers), and 431 with knee rheumatoid arthritis were also reported as comparative group. In the included studies, 201 different SF biomarkers were identified. Among these, several were investigated multiple times in different articles, for a total of 373 analyses of SF biomarkers reported in the literature in the last 20 years. In the following paragraphs, the identified biomarkers are grouped according to their molecular characteristics (Lotz et al. ⁷ ), and the results documented for each biomarker are summarized according to the BIPEDs classification.

Molecular Characteristics

SF biomarkers are categorized into 4 classes based on their nature and molecular origin: the most represented biomarkers were those related to “other processes” as inflammation or fibrosis. In particular, inflammation biomarkers were the most studied, with cytokines and chemokines above all: IL-6 was the most studied biomarker (17 articles), followed by IL-8 (12 articles), TNF-α (9 articles), and other less reported ones. Other SF biomarkers included in this category were growth factors (GFs), adipokines, metabolites, and other proteins related to bone or cartilage metabolism. Other articles focused on SF biomarkers related to collagen metabolism, with the most studied being CTX-II (5 articles). In the category of SF biomarkers related to aggrecan metabolism, the most considered were chondroitin 4-sulfate (7 articles), chondroitin 6-sulfate (6 articles), and keratan sulfate (4 articles). Finally, the most common SF biomarkers related to other non-collagenous proteins included COMP and MMP-3 (10 articles for each), TIMP-1 and MMP-1 (9 articles for each), MMP-13 (6 articles), and hyaluronan (5 articles). More details are reported in Table 1 .

Table 1.

Selected SF Biomarkers Classified Based on Their Molecular Characteristics.

Biomarkers related to collagen metabolism

CIM; CIIM; Collagenase 2; CTXI**; CTXII****; Deoxypyridinoline*; PICP; PIICP*; Pyridinoline*

Biomarkers related to aggrecan metabolism

ADAMTS-4; ADAMTS-5; Aggrecan; Aggrecan Epitope 846*; ARGS-aggrecan**; Chondroitin 4-sulfate******; Chondroitin 6-sulfate*****; Dermatan sulfate; Keratin sulfate***

Biomarkers related to other non-collagenous proteins

COMP*********; Fibulin-3; Hyaluronan****; MWHA; Osteocalcin*; Osteogenic Protein 1***; Osteopontin*; sRAGE; YKL-40 Matrix metalloproteinases: MMP-1********; MMP-2*; MMP-3*********; MMP-7; MMP-8; MMP-9; MMP-12; MMP-13***** Tissue inhibitors of metalloproteinases: TIMP-1********; TIMP-2*; TIMP-3; TIMP-4 Extracellular matrix: Glypican-3; Syndecan4; sBGN; sDCN; ucMGP; Tenascin C*; TSG-6**

Biomarkers related to other processes

Cytokines: Activin A*; Chemerin**; HMGB-1; IFNγ; IL-1*; IL-1β***; IL-2; IL-5; IL-6****************; IL-7; IL-8***********; IL-10**; IL-11; IL-12p70; IL-13; IL-15; IL-17; IL-18***; IL-34; Inhibin A; MIF; TNF-α******** Chemokines: CCL2****; CCL3*; CCL4; CCL5***; CCL11; CCL13; CCL20; CCL22; CX3CL1*; CXCL10**; CXCL12*; CXCL13 Complement system: C3 fragment; C3f Des-Arg; C3bBbP; C4d; sTCC Growth factors: Adrenomedullin; BDNF; CTGF; FGF-2*; FGF-21; HGF; Hif-1α*; G-CSF; LIF; LTBP2; M-CSF; PDGF-AA; sCD40L; TGF-β1**; VEGF****** Adipokines: Adiponectin; FABP4; Leptin****; Omentin-1*; RBP-4; Resistin***; Visfatin Bone or cartilage metabolism: BMP-2; CD-RAP*; DKK1***; Exosomal lncRNA PCGEM1; Exosomal lncRNA HOTAIR; Exosomal lncRNA GAS5; FRZB*; Ghrelin; Glypican-3; Gremlin-1**; Irisin; Lubricin; NTx; OPG; OPG/OCIF; Sclerostin*; Thymosin β4 Metabolites: 3-Hydroxybutyrate; Citrate; Creatine; Ethanol; Ethanolamine; Fructose; Hexanolycarnitine; Malate; Methionine; N-Phenylacetylglycine; O-Acetylcarnitine MicroRNA: MicroRNA-140-3p; MicroRNA-140-5p; miR-23a-3p*; miR-24-3p; miR-27a-3p; miR-27b-3p; miR-29c-3p; miR-30c-5p; miR-34a-5p; miR-100-5p; miR-186-5p; miR-200c-3p; miR-1826 Oxidative or Nitrosative stress: 3-Nitrotyrosine; 8-Isoprostane F2α; GPx; GSH; Iron; MPO; Nitrotyrosine; SOD; TBARs; sVAP-1; Vitamin E OTHER: α-MSH; α(II)CB11B; ADA; Albumin; anti-CCP; ATP; Autotaxin; Bradykinin*; Calprotectin; Camp; CD14*; CD163; CGRP*; Chlorinated peptides; Endoglin; Haptoglobin; Hs-CRP; HSPA1A; iCAM-1; IgA-RF; IgG-RF; IgM-RF; ITI-H4; LBP; LPS; Neuropeptide Y*; Oncostatin M; P Selectin; Pentosidine; Periostin*; PGE2*; Phospholipids; RANKL; S100A12; Salic Acid; Sphingolipids; sTNFR-1; sTNFR-2; Substance P; TWEAK; UCMA; vCAM-1; White blood cells

^*Asterisks indicate the number of further articles analyzing the specific biomarker. α-MSH = melanocyte stimulating hormone; α(II)CB11B = cluster of differentiation molecule 11B; ADA = adenosine deaminase; ADAMTS-4 = aggrecanase-1; ADAMTS-5 = aggrecanase-2; anti-CCP = anti–cyclic citrullinated peptide; BDNF = brain-derived neurotrophic factor; BMP = bone morphogenetic protein; CCL = chemokine (C-C motif) ligand; CD = cluster of differentiation; CD-RAP = cartilage derived retinoic acid sensitive protein; CGRP = calcitonin gene-related peptide; CIM = collagen type I-specific neoepitope; CIIM = collagen type II-specific neoepitope; COMP = cartilage oligomeric matrix protein; CTGF = connective tissue growth factor; CTX = C-terminal telopeptide; CXCL = chemokine (C-X-C motif) ligand; DKK = dickkopf-1; FABP4 = fatty acid-binding protein 4; FGF-2 = basic fibroblast growth factor; FGF-21 = fibroblast growth factor-21; FRZB = frizzled-related protein; G-CSF = granulocyte-colony stimulating factor; GPx = glutathione peroxidase; GSH = glutathione; HGF = hepatocyte growth factor; Hif-1α = hypoxia-inducible factor-1α; HMGB-1 = high-mobility group box 1; Hs-CRP = high sensitivity C-reactive protein; HSPA1A = heat shock protein family A (Hsp70) member 1A; iCAM-1 = intercellular adhesion molecule 1; IFNγ = interferon-γ; IgA-RF = IgA rheumatoid factor; IgG-RF = IgG rheumatoid factor; IgM-RF = IgM rheumatoid factor; IL = interleukin; ITI-H4 = inter-alpha-trypsin inhibitor heavy chain 4; LBP = LPS binding protein; LIF = leukemia-inhibitory factor; LPS = lipopolysaccharide; LTBP2 = latent-transforming growth factor beta-binding protein 2; M-CSF = macrophage-colony stimulating factor; MIF = macrophage migration inhibitory factor; MMP = matrix metalloproteinases; MPO = myeloperoxidase; MWHA = high-molecular-weight HA; NTx = N-terminal telopeptide; OPG = osteoprotegerin; OPG/OCIF = osteoprotegerin/osteoclastogenesis inhibitory factor; PDGF-AA = platelet-derived growth factor; PGE2 = prostaglandin E2; PICP = procollagen type II carboxy-terminal propeptide; PIICP = procollagen type II carboxy-terminal propeptide; RANKL = receptor activator of nuclear factor kappa-Β ligand; RBP-4 = retinol binding protein 4; S100A12 = S100 calcium-binding protein A12; sBGN = soluble biglycan; sDCN = soluble decorin; sCD40L = soluble CD40 ligand; SOD = superoxide dismutase; sRAGE = soluble receptor for advanced glycation end products; sTCC = soluble terminal complement complex; sTNFR = soluble tumor necrosis factor receptor; sTCC = soluble terminal complement complex; sVAP-1 = soluble vascular adhesion protein-1; TBARs = thiobarbituric acid reactive substances; TGF-β1 = transforming growth factor-β; TIMP = tissue inhibitors of metalloproteinases; TNF-α = tumor necrosis factor-α; TSG-6 = TNF-stimulated gene 6 protein; TWEAK = TNF-like weak inducer of apoptosis; UCMA = upper zone of growth plate and cartilage matrix associated; ucMGP = uncarboxylated matrix Gla-protein; vCAM-1 = vascular cell adhesion molecule-1; VEGF = vascular endothelial growth factor; YKL-40 = chitinase-3-like protein 1.

BIPEDs Classification

Selected biomarkers were classified according to BIPEDs classification in 6 categories, as reported in the following paragraphs and tables, summarizing both their clinical and imaging results.

Burden of Disease Biomarkers

The “burden of disease” category is the most represented in the literature of SF biomarkers: 132 different biomarkers reported in 96 studies were identified, for a total of 190 analyses on burden of disease SF biomarkers. Kellgren-Lawrence classification was the most commonly used reference parameter for the radiographic severity, applied in 82 studies; other less commonly used scales were Ahlbäck classification (2 articles) and Tomihisa Koshino’s scoring system (1 article). Moreover, in some articles also arthroscopic severity was evaluated with the Outerbridge classification (2 articles) or with the International Cartilage Repair Society classification (1 article). Last, the knee OA severity was also respectively evaluated with scintigraphy imaging, ultrasonographic findings, macroscopy and histology (1 article each).

For clinical severity, the most used reference parameter was the WOMAC (Western Ontario and McMaster Universities Osteoarthritis Index) scale (26 articles); other less commonly used parameters were VAS (visual analogue scale) in 7 articles, followed by Numeric Rating Scale (NRS), NPQ (Neurophysiology of Pain Questionnaire) scale, KSS (Karolinska Sleepiness Scale) scale, and Watanabe’s pain score (1 article each).

A positive correlation with the knee OA severity was found for 101 analyses of SF biomarkers, a negative correlation for 32, and no correlation for 57 analyses of SF biomarkers. Among the most studied biomarkers of this category, IL-6 was analyzed in 8 studies, reporting controversial results: a negative correlation with radiographic or clinical severity of knee OA were reported in 4 cases, but a positive correlation was reported in 2 cases, with no correlation in the remaining 2 cases. IL-8 was analyzed in 6 studies, showing a positive correlation in only one case for both radiographic and clinical severity, while no correlation in the other 5 cases. MMP-1 levels in SF were analyzed in 6 studies, showing a negative correlation with radiographic OA severity in 2 cases, no correlation in 2 cases, while a positive correlation was reported in 2 cases. Similarly, MMP-3 and MMP-13 were investigated in 5 articles each, showing correlation with OA severity in only 2 cases, while no correlation was reported in the other cases. Conversely, consistent results have been found for some biomarkers: SF levels of VEGF were investigated in 7 studies, showing a positive correlation with OA severity in 6 cases and no correlation in the last one. CCL2, Leptin, and TIMP-1 were analyzed in 4 studies each, showing a positive correlation with OA severity in 3 cases and no correlation in the last one. Results for each biomarker are summarized in Table 2 , while more detailed findings are reported in the supplementary material.

Table 2.

Burden of Disease SF Biomarkers.

Biomarker	References	Imaging	Clinical	Biomarker	References	Imaging	Clinical
α-MSH	25	↓	↓	KS	42	\	\
ADAMTS-5	171	−	\	LBP	55	\	↑
Adiponectin	20	\	\	Leptin	39, 77, 94, 130	↑ \ \ ↑	− \ ↑ −
Adrenomedullin	51	↑	−	LPS	55	↑	↑
ATP	119	−	↑	Lubricin	53	\	\
Autotaxin	72	↑	↑	M-CSF	171	−	\
BDNF	90	\	−	MiRNA-140-3p	38	↓	−
BMP-2	74	↑	↑	MiRNA-140-5p	38	↓	−
Bradykinin	171	−	\	MIF	54	↑	↑
C3 fragment	36	↓	−	miR-186-5p	56	↑	−
C3f Des-Arg	36	↓	−	miR-23a-3p	56	↑	−
Calprotectin	157	−	↑	miR-24-3p	56	↑	−
CCL2	12, 21, 81, 109	\ ↑ \ \	↑ − ↑ −	miR-27a-3p	56	↑	−
CCL3	109	\	−	miR-27b-3p	56	↑	−
CCL5	68	\	−	miR-29c-3p	567	↑	−
CCL13	73	↑	−	miR-34a-5p	56	↑	−
CCL20	18	↑	↑	MMP-1	32, 68, 100, 106, 126, 150	↓ \ ↑ ↓ \ \	− − − − − −
CD-RAP	148	↓	−	MMP-2	68, 100	\ ↑	− −
CGRP	70, 171	↑ −	↑ \	MMP-3	21, 100, 126, 150, 171	\ ↑ \ \ −	↑ − − − \
Chemerin	109, 164	↑ ↑	− −	MMP-7	100	\	−
Cl-peptides	138	↓	−	MMP-8	100	↑	−
C4S	137	↓	−	MMP-9	100	↑	−
C6S	137	↓	−	MMP-12	100	\	−
CIM	46	−	↑	MMP-13	40, 77, 100, 126, 171	\ \ ↑ \ −	− − \ − ↑
CIIM	165	↑	−	MPO	138	↓	−
Collagenase 2	40	\	\	Neopterin	96, 171	↑ −	↑ \
COMP	26, 77, 128, 137	↑ \ \ ↓	− \ − −	Neuropeptide Y	80	↑	↑
CTGF	103	↑	−	NTx	150	\	−
CTXI	44	\	\	Omentin-1	98, 105	− ↓	− ↓
CTXII	44	↑	\	OPG	135	↑	−
CX3CL1	71, 93	↑ ↑	↑ −	OPG/OCIF	152	↑	−
CXCL10	110	↓	−	OP-1	42, 132	↑ ↑	− −
CXCL12	63, 101	↑ ↑	− −	Osteopontin	116, 129	↑ ↑	− −
Deoxypyridinoline	150	↑	−	P selectin	122	↑	−
Dermatansulfate	137	↓	−	Periostin	69	↑	−
DKK1	43, 115	↑ ↓	− −	PICP	150	\	−
Endoglin	127	↑	−	Pyridinoline	150	\	−
Exosomal lncRNA GAS5	28	\	\	RANKL	113	↑	−
Exosomal lncRNA HOTAIR	28	\	\	Resistin	62, 77	↑ \	↑ \
Exosomal lncRNA PCGEM1	28	↑	↑	S100A12	97	↑	↑
FGF-21	75	↑	−	Salic Acid	83	↑	−
Fibulin-3	35	↑	−	Sclerostin	43, 86	\ ↓	− −
Ghrelin	45	↓	↓	sICAM-1	21	↑	↑
Glypican-3	24	↓	−	sRAGE	120	↓	−
GREM 1	48	↑	−	sTNFR-1	85	−	↓
HA	42, 126	\ \	− −	sTNFR-2	85	−	↓
Haptoglobin	60	↑	−	Substance P	171	−	\
HGF	126	\	−	sVAP1	22	↑	−
HIF-1α	47, 78	↑ ↑	− −	sVCAM-1	21	↑	↑
HMGB-1	65	↑	↑	Syndecan 4	160	↑	−
IFNγ	109	\	−	Tenascin-C	143	↑	−
IL-1	40	\	−	TGF-β1	126	\	−
IL-1β	44, 109, 171	− \ ↓	↓ \ ↓	Thymosin β4	95	↑	−
IL-2	109	\	−	TIMP-1	21, 100, 126, 150	↑ ↑ \ ↑	↑ − − −
IL-5	109	\	−	TIMP-2	100	\	−
IL-6	40, 44, 46, 68, 109, 153, 166, 171	\ \ − ↑ \ ↓ − ↓	− \ ↑ − \ − ↓ ↓	TIMP-3	100	↑	−
IL-7	106	\	−	TIMP-4	100	\	-
IL-8	40, 44, 68, 109, 153, 166	\ \ ↑ \ \ −	− \ − − − ↑	TNF-α	40, 44, 171	\ − \	− ↑ ↓
IL-10	40	\	−	TWEAK	32	↓	−
IL-12p70	109	\	−	UCMA	19	↑	↑
IL-15	133	↓	−	ucMGP	57	↓	−
IL-17	66	\	↑	VEGF	21, 30, 40, 77, 82, 106, 126	↑ \ \ ↑ ↑ ↑ \	↑ ↑ − \ − − −
IL-18	39	↑	\	Visfatin	114	↑	−
IL-34	27	↑	↑	YKL-40	64	−	↑
Irisin	59	↓	−

“↑” = positive correlation; “↓” = negative correlation; “\” = no correlation; “−” = not evaluated.

Investigative Biomarkers

This systematic review found 41 different SF biomarkers without sufficient information to allow inclusion into one of the other categories (among these, some biomarkers have SF levels correlated with serum levels, others changed in SF following exercise or arthroscopic surgery or high tibial osteotomy, and others had levels correlating to other biomarkers). For this reason, these biomarkers were classified in the “investigative” category. Among these, CD-RAP, COMP, MMP-1, MMP-3, MMP-13, RBP-4, and TIMP-1 showed higher levels in SF samples compared to serum. Conversely, Pentosidine showed lower levels in SF samples compared to serum. Moreover, the SF levels of Adiponectin, Bradykinin, IL-6, IL-18, Oncostatin M, PGE2, Resistin, TNF-α, and TSSG-6 showed a significant correlation with the SF levels of other biomarkers for cartilage degradation or inflammation. More detailed findings for investigative SF biomarkers are reported in the supplementary material.

Prognostic Biomarkers

Only 8 studies described “prognostic” SF biomarkers for knee OA, identifying 9 biomarkers, out of the 14 investigated ones, correlating with OA progression.

CD14 and CD163 were positively associated with osteophyte progression at 3 years of follow-up, supporting the central role of inflammation as a determinant of progression risk of OA. Similarly, baseline PIICP level showed a direct positive correlation with radiographic joint space narrowing over 4 years. TSG-6 at baseline predicted OA progression over a period of 3 years. IL-6, IL-8, and TNFα were investigated in subjects with previous meniscectomy; higher or over time increasing levels of IL-6 and TNF-α, but not IL-8, were associated with an increased risk of radiographic OA progression after 18 years. Following meniscectomy, ARGS concentrations were inversely associated with both future loss of joint space and pain worsening after 18 years with low levels associated with increased risk of progression and higher levels with decreased risk. With regard to biomarkers associated to anterior cruciate ligament (ACL) rupture, acute and chronic SF concentrations of aggrecan, COMP, MMP-3, and TIMP-1, collected at the time of ACL reconstruction, failed to predict knee OA 16 years after ACL injury. Finally, CTXI and CTXII concentrations were associated with a rapid clinical progression in OA patients with obesity and depression with respect to patients without comorbidities. More detailed findings of the studies on each of these biomarkers are reported in Table 3 .

Table 3.

Prognostic SF Biomarkers.

Biomarker	References	Patients	Follow-up	Results
Aggrecan	50	88 patients post ACL injury	16 Years	Neither acute nor chronic concentrations of SF aggrecan were associated with the development of radiographic knee OA at the 16-year follow-up.
ARGS aggrecan	104	141 patients post meniscectomy	18 Years	Having decreasing levels of SF ARGS over time was associated with an increased risk of loss of joint space and pain worsening but showed no association with other patient-reported outcomes or osteophyte progression.
CD14	76	159 OA	3 Years	SF CD14 levels were positively associated with osteophyte progression at 3 years.
CD163	76	159 OA	3 Years	SF CD163 levels were positively associated with osteophyte progression at 3 years.
COMP	50	88 patients post ACL injury	16 Years	Neither acute nor chronic concentrations of SF COMP were associated with the development of radiographic knee OA at the 16-year follow-up.
CTXI	169	600 OA	2 Years	CTXI was associated with a rapid clinical progression in OA patients with obesity and depression respect to patients without comorbidities.
CTXII	169	600 OA	2 Years	CTXII was associated with a rapid clinical progression in OA patients with obesity and depression respect to patients without comorbidities.
IL-6	61	132 patients post meniscectomy	18 Years	In subjects with previous meniscectomy, higher or over time increasing SF levels of IL-6 seems to be associated with increased risk for progression of radiographic OA.
IL-8	61	132 patients post meniscectomy	18 Years	No correlation of SF IL-8 levels and increased risk for progression of radiographic OA.
MMP-3	50	88 patients post AC -injury	16 Years	Neither acute nor chronic concentrations of SF MMP-3 were associated with the development of radiographic knee OA at the 16-year follow-up.
PIICP	147	110 OA	4 Years	Quantification of SF PIICP was able to predict subsequent radiographic progression in early knee OA.
TIMP-1	50	88 patients post ACL injury	16 Years	Neither acute nor chronic concentrations of SF TIMP-1 were associated with the development of radiographic knee OA at the 16-year follow-up.
TNF-α	61	132 patients post meniscectomy	18 Years	In subjects with previous meniscectomy, higher or over time increasing SF levels of TNF-α seems to be associated with increased risk for progression of radiographic OA.
TSG-6	91	132 OA	3 Years	TSG-6 is a promising biomarker for OA progression, for the correlation with the joint space narrowing.
	163	813 OA	17 Months	TSG-6 activity can potentially be used to track longitudinal changes in disease status and inform treatment decisions.

OA = osteoarthritis; SF = synovial fluid; ACL = anterior cruciate ligament.

Efficacy of Intervention Biomarkers

Only 7 studies described “efficacy of intervention” SF biomarkers for knee OA, identifying a total of 20 different markers with composite findings.

Biomarkers related to aggrecan metabolism were the most represented, such as C6S, C4S, aggrecan, and keratan sulfate. C4S and Tenascin-C levels in SF before HA injection showed a significant inverse correlation with the improvement of VAS pain after treatment in knee OA patients. Moreover, positive correlations were noted between the concentrations of C6S and aggrecan before HA injection and the improvement of the Japanese Orthopaedic Association (JOA) score 1 month after injection. SF levels of white blood cells before steroid injection showed a significant correlation with the reduction in KOOS (Knee Injury and Osteoarthritis Outcome Score) pain score after treatment, identifying persons more likely to benefit from an intra-articular steroid injection.

Looking at changes in SF biomarkers after treatment, concentrations of α(II)CB11B, C4S, C6S, IL-6, MMP-1, and MMP-9 showed a significant decrease after HA injections. Conversely, SF concentrations of COMP, HA, and osteocalcin increased significantly after HA injections. SF CTXI were significantly reduced in knee OA patients treated with oral colchicine compared to placebo. More detailed findings of the studies on each of these biomarkers are reported in Table 4 .

Table 4.

Efficacy of Intervention SF Biomarkers.

Biomarker	References	Type of treatment	Results
α(II)CB11B	151	20 OA treated with HA injection	SF α(II)CB11B concentration showed a significant decrease after HA injections.
Aggrecan	139	38 OA treated with HA injection	Positive correlations were noted between the concentrations of aggrecan before HA injection and the improvement of the JOA score after injection (after 1 month).
C4S	79	55 OA treated with HA injection	C4S and the SF volume were significantly decreased after intraarticular (IA) HA treatment. These improvements of SF properties indicated that IA HA had potential of affecting for the joint cartilage and synovium with knee OA.
	125	51 OA treated with HA or steroid	No significant differences between 2 groups. C4S Concentrations in SF decreased slightly after injection therapy in both groups. There was no correlation between changes in the biomarkers and clinical results.
	136	28 OA treated with HA injection	Inverse correlations were observed between the levels of C4S before HA injection and improvement of VAS after the 5 injections.
	139	38 OA treated with HA injection	C4S decreased after injection, although these decreases were not significant.
C6S	79	55 OA treated with HA injection	C6S and the SF volume were significantly decreased after IA HA treatment. These improvements of SF properties indicated that IA HA had potential of affecting for the joint cartilage and synovium membrane with knee OA.
	125	51 OA treated with HA or steroid	No significant differences between 2 groups. C6S Concentrations in SF decreased slightly after injection therapy in both groups. There was no correlation between changes in the biomarkers and clinical results.
	136	28 OA treated with HA injection	No significant correlation was seen between levels of C6S before injections and improvement of VAS after the 5 injections.
	139	38 OA treated with HA injection	Positive correlations were noted between the concentrations of C6S before HA injection and the improvement of the JOA score after injection (after 1 month).
CD14	33	109 OA treated with Colchicine or placebo	No significant differences between 2 groups.
COMP	152	20 OA treated with HA injection	COMP levels were significantly higher after HA injections.
CTXI	33	109 OA treated with Colchicine or placebo	SF CTXI were significantly reduced in the Colchicine group, but not the placebo-treated group after 16 weeks.
CTXII	33	109 OA treated with Colchicine or placebo	No significant differences between 2 groups.
HA	79	55 OA treated with HA injection	No correlation.
	125	51 OA treated with HA or steroid	HA concentrations in SF increased significantly after therapy in the Na-HA group but remained unchanged in the CS group. There was no correlation between changes in the biomarkers and clinical results.
IL-6	33	109 OA treated with Colchicine or placebo	No significant differences between 2 groups.
	79	55 OA treated with HA injection	IL-6 and the SF volume were significantly decreased after IA HA treatment. These improvements of SF properties indicated that IA HA had potential of affecting for the joint cartilage and synovium membrane with knee OA.
IL-8	33	109 OA treated with Colchicine or placebo	No significant differences between 2 groups.
IL-18	33	109 OA treated with Colchicine or placebo	No significant differences between 2 groups.
Keratansulfate	136	28 OA treated with HA injection	No significant correlation was seen between levels of KS before injections and improvement of VAS after the 5 injections.
	79	55 OA treated with HA injection	No correlation.
MMP-1	151	20 OA treated with HA injection	SF MMP-1 concentration showed a significant decrease after HA injections.
MMP-9	125	51 OA treated with HA or steroid	MMP-9 concentrations in SF decreased after injection therapy in the Na-HA group but were unchanged in the CS group. There was no correlation between changes in the biomarkers and clinical results.
Osteocalcin	151	20 OA treated with HA injection	Osteocalcin levels were significantly higher after HA injections.
Tenascin-C	136	28 OA treated with HA injection	Inverse correlations were observed between the levels of TN-C before HA injection and improvement of VAS after the 5 injections.
TNF-α	33	109 OA treated with Colchicine or placebo	No significant differences between 2 groups.
TIMP-1	125	51 OA treated with HA or steroid	TIMP-1 concentrations did not change significantly after therapy in either group. There was no correlation between changes in the biomarker and clinical results.
White blood cells	51	55 OA treated with steroid injection	With each category increase in SF WBC there was a greater mean reduction in KOOS pain score after steroid injection. Total SF WBC count may predict response to anti-inflammatory treatment.

OA = osteoarthritis; HA = hyaluronan; SF = synovial fluid; JOA = Japanese Orthopaedic Association; VAS = Visual Analogue Scale; WBC = white blood cells; KOOS = Knee Injury and Osteoarthritis Outcome Score.

Diagnostic Biomarkers

“Diagnostic” SF biomarkers were studied in 45 articles focusing on 106 different biomarkers, for a total of 139 analyses on this category. Seventy-eight biomarkers were studied for the diagnosis between OA patients and healthy controls; among these, 42 presented significantly higher levels in the SF of OA patients compared with healthy controls, 21 presented significantly lower levels in OA compared to healthy controls, while for 15 biomarkers no differences were found with the control. Fifty-five biomarkers were investigated for the differential diagnosis between OA and immune-mediated inflammatory arthritis (rheumatoid arthritis, psoriatic arthritis, pseudogout), among these, 5 presented significantly higher levels in OA patients versus inflammatory arthritis, 35 presented significantly lower levels in OA patients versus inflammatory arthritis, while 15 showed no statistically significant differences versus control groups. Finally, 49 biomarkers were studied to differentiate the SF of OA patients from the SF of knee injured patients; among these, 17 presented significantly higher concentrations in SF of OA patients, 13 presented lower concentration in SF of OA patients, while no differences were found for 19 biomarkers. IL-6 was the most studied biomarker also for this category, with 7 studies evaluating its diagnostic usefulness: this cytokine showed significantly higher levels in OA patients when compared with healthy controls or knee injury patients (4 studies). Moreover, SF levels of IL-6 were significantly lower in OA patients when compared with rheumatic patients (3 studies). Similarly, IL-8 showed higher levels in SF of patients with knee OA when compared with healthy controls or knee injury patients (3 studies), and lower levels when compared with rheumatic patients (3 studies). Furthermore, MMP-1, MMP-3, and CCL5 were significantly increased in the SF of OA patients compared with healthy controls and knee injury patients (3 studies for each one). Conversely, some SF biomarkers, including cAMP, CD-RAP, COMP, ITI-H4, and MMP-7, were significantly higher in OA patients compared with rheumatic patients (one study for each one). Results for each biomarker are summarized in Table 5 , while more detailed findings are reported in the supplementary material.

Table 5.

Diagnostic SF Biomarkers.

Biomarker	References	Controls	Knee Injury	RA	Biomarker	References	Controls	Knee Injury	RA
3-Hydroxybutyrate	67	↓	−	−	IL-11	111	↑	−	\
3-Nitrotyrosine	29	−	\	−	IL-13	167	↑	−	−
8-Isoprostane F2α	29	−	↑	−	IL-18	134	−	−	↓
Activin A	118	−	−	↓	Inhibin A	118	−	−	↓
ADA	117	−	−	↓	Iron	124	−	\
ADAMTS-4	12	↑	↑	−	ITI-H4	161	−	−	↑
Aggrecan ARGS	131	↑	−	−	Leptin	118, 124, 161	− − ↑	− ↑ −	↓ − −
anti-CCP	140	−	−	↓	LIF	112	↓	−	\
C3bBbP	49	↑	\	↓	Malate	67	↓	−	−
C4d	49	↑	\	↓	Methionine	67	↓	−	−
cAMP	134	−	−	↑	MicroRNA-140-3p	38	↓	−	↓
CCL2	68	\	−	−	MicroRNA-140-5p	38	↓	−	↓
CCL3	167	↑	−	−	miR-100-5p	17	↓	−	−
CCL4	167	↑	−	−	miR-1826	17	↓	−	−
CCL5	68, 112, 167, 172	↑ − ↑ −	− \ − ↑	− − − −	miR-200c-3p	17	↓	−	−
CCL11	167	↓	−	−	MMP-1	68, 100, 172	↑ ↑ −	− − ↑	− \ −
CCL22	167	↑	−	−	MMP-2	68, 100	\ ↑	− −	− \
CD-RAP	148	−	−	↑	MMP-3	100, 155, 172	↑ − −	− − ↑	\ \ −
Chemerin	89	−	−	\	MMP-7	100	↑	−	↑
Cl-peptides	138	↑	−	−	MMP-8	100	↑	−	\
CIIM	165	↑	−	−	MMP-9	100	↑	−	\
Citrate	67	↑	−	−	MMP-12	100	\	−	\
Collagenase 2	40	−	\	−	MMP-13	40, 100	− ↑	\ −	− \
COMP	26, 111, 168	↑ \ −	− − −	− \ ↑	MPO	138	↑	−	−
Creatine	67	\	−	−	Neuropeptide Y	80	↑	−	−
CTXII	144	↑	−	−	Nitrotyrosine	99	↑	↓	↓
CXCL10	167, 172	↑ −	− ↑	− −	O-Acetylcarnitine	66	↓	−	−
Deoxypyridinoline	146	−	−	\	Osteocalcin	111	\	−	\
DKK1	31, 43	\ −	− ↑	− −	OP-1	111, 172	↑ −	− ↑	↓ −
Ethanol	67	↓	−	−	Osteopontin	129	↑	−	−
Ethanolamine	67	↓	−	−	PDGF-AA	167	↑	−	−
Exosomal lncRNA GAS5	28	\	−	−	Periostin	162	−	−	↓
Exosomal lncRNA HOTAIR	28	\	−	−	PGE2	134	−	−	↓
Exosomal lncRNA PCGEM1	28	↑	−	−	Phenylacetylglycine	67	↓	−	−
FABP4	34	↑	−	−	Pyridinoline	146	−	−	↓
FGF-2	167	\	−	−	RANKL	113	−	−	\
Fructose	67	↑	−	−	Sbgn	58	−	↑	↓
FRZB	31	↓	−	−	sCD40L	167	↑	−	−
G-CSF	167	↓	−	−	Sclerostin	43	−	\	−
GPx	124	−	\	−	sDCN	58	−	\	\
GREM 1	31	\	−	−	SOD	124	−	\	−
GSH	124	−	\	−	sTCC	49	↑	\	↓
Haptoglobin	60	−	↑	−	TBARs	124	−	\	−
Hexanolycarnitine	67	↓	−	−	TGF-β1	167	\	−	−
Hif-1α	47	↑	−	−	TIMP-1	100, 172	↓ −	− \	↓ −
hs-CRP	118	−	↓	−	TIMP-2	100	\	−	↓
HSPA1A	172	−	↑	−	TIMP-3	100	↓	−	↓
IgA-RF	141	−	↓	−	TIMP-4	100	↓	−	↓
IgM-RF	156	−	↓	−	TNF-α	40, 89, 172	− − −	\ − \	− ↓ −
IL-1	40, 111	− ↓	\	− ↓	VEGF	40, 77	− −	↑ −	− ↓
IL-6	40, 68, 89, 111, 153, 167, 172	− ↑− \ − ↑ −	↑ − − − − − ↑	− − ↓ ↓ ↓ − −	Visfatin	114	↑	−	−
IL-8	40, 68, 89, 111, 112, 153, 172	− ↑ − \ − − −	\ − − − ↑ − ↑	− − ↓ ↓ − ↓−	Vitamin E	124	−	↓	−
IL-10	40	−	↓	−

“↑” = biomarker levels were significant higher in OA patients; “↓” = biomarker levels were significant lower in OA patients; “\” = no significant differences; “−” = not evaluated; RA, rheumatoid arthritis.

Discussion

The main finding of this systematic review is that in the literature SF biomarkers related to knee OA are numerous, with 201 identified molecules, and “burden of disease” and “diagnostic” categories being the most studied, compared to “prognostic” and “therapeutic” ones. However, while many biomarkers were found to correlate with disease features, each result is supported by limited and sometimes controversial findings.

The research on SF biomarkers can foster the scientific progress in the field of knee OA. In fact, SF presents potential key features because it is in direct contact with cartilage and synovial layer of the involved joint. In a joint with OA, the disease alters the composition of SF, with abnormal matrix turnover resulting in the release into the joint fluid of many molecules and fragments of matrix components derived from tissues such as the articular cartilage, bone, and synovium. ¹⁷³ Some of these fragments can be assayed before they are even detectable in other samples such as serum and urine. ¹⁷⁴ Therefore, SF can reflect more closely the alterations caused by OA in a singular joint, without undergoing systemic alterations due to other physiological or pathological processes originating from other tissues. ¹⁷⁵ Moreover, many biomarkers produced by collagen II synthesis or degradation or by synovium turnover are found in the SF in higher concentrations than serum or urine, such as CD-RAP, COMP, MMP-1, MMP-3, MMP-13, and TIMP-1,^141,142 offering the possibility of having a higher sensitivity and specificity of these markers in SF samples. Accordingly, the detection of biomarkers in SF seems more favorable than in other body fluids. On the other hand, the evaluation of SF also presents some limitations; while it bathes the intrinsic structures of diarthrodial joints offering the unique opportunity to study the entire joint, it is more difficult to obtain in the clinical practice, and its evaluation sometimes limited by the small quantity. ⁴ For this reason, researchers are looking for new tools that can improve biomarkers detection and measurement in SF, such as a recent permanent magnet introduced by Garraud et al. ¹⁷⁶ With this technique, biomarkers may be collected from SF by injecting magnetic particles that are functionalized to bind with biomarker molecules within the fluid and then collecting these particle-biomarker conjugates from the fluid via a permanent magnet. Thus, this approach could allow for a more reliable collection of biomarkers from highly viscous biological fluids or complex anatomical locations without the need to remove the fluid from the patient.

The study of SF biomarkers is paramount also for other reasons, as it may allow to shed some light about OA pathogenesis, thanks to insights into the interaction of different biomarkers in the complex cascade which finally leads to OA. Following this concept, inflammatory biomarkers were the most studied, confirming the interest on these mediators produced by articular tissues in OA, being implicated in the pathogenesis. ¹⁷⁷ In fact, IL-6 and IL-8 were the most studied, with 17 and 12 studies, respectively. Unfortunately, despite the high research interest due to the direct involvement in the disease processes, research led to often conflicting results, and the same was found for the other molecular categories. For example, MMPs are implicated in the loss of articular cartilage in OA, ¹⁷⁸ but while they have been found to be often significantly higher in SF of advanced stage of knee OA compared with early stages, they were at times unrelated to knee OA severity or even to OA presence.^68,126,150 Among other molecules correlated with the pathogenesis, adipokine leptin showed in most studies a positive correlation with both diagnosis and severity of knee OA.^{39,94,118,130} Moreover, VEGF showed in many studies a positive correlation with radiographic grading, ultrasonographic findings and functional status in knee OA, confirming the role of angiogenesis in the pathogenesis of OA.^{21,40,77,82,106}

Overall, the literature analysis underlined a plethora of molecules investigated as SF biomarkers, but each one with a limited number of studies. Accordingly, while some molecules emerged as possible suitable candidates, most of them have either weak or even controversial findings. In this regard, biomarkers in serum and urine have often shown discordant results in SF. For example, COMP is a constituent of hyaline articular cartilage released from damaged cartilage into SF and hence into the circulation, and as such it was frequently studied in serum, with a recent meta-analysis showing elevated serum COMP in patients with knee OA where it was sensitive to OA disease severity. ¹⁷⁹ However, the few studies on COMP levels in SF could not confirm such positive results, with the majority of the studies being not able to correlate OA severity and prognosis.^{50,77,111,128} Similarly, CTX-II, a marker related to collagen type II degradation, was the molecule most extensively investigated in urinary samples, 66 times in total as reported in a review by Van Spil et al., ¹¹ with a strong correlation with knee OA grade and diagnosis. However, these results are weakly confirmed in SF, where the literature analysis showed only 2 out of 3 studies supporting a similar correlation.^44,144

Controversial results are probably due to the current literature limitations, but they can be also related to the complexity of the OA processes, which are complicated and multidimensional and may affect differently the whole articular environment. In this light, evaluating one biomarker in the absence of a clear understanding of its interaction with other molecules can lead to erroneous conclusions. Each biomarker could play a small role in the summative outcome of interest; thus, the measurement of complex, composite biomarkers may enable better predictions. Different molecules or, even more, different modalities to assess OA biomarkers may offer complementary information, and aggregate markers may provide superior performance. New technologies are being develop such as interactome test, a method that allows the visualization and interpretation of complex interactions among large numbers of molecules and could therefore be used as a method of integrating multiple “omic” datasets to provide an understanding of the organizational complexity within the biological system. ¹⁸⁰ The study of groups of proteins or proteins and metabolites that are highly interconnected can be used to identify key functions within an interactome network. ¹⁸¹ Accordingly, several studies are now focused on applying metabolomics in serum and in urine samples to detect a metabolic fingerprint of OA.^182,183 However, this aspect is still not investigated in SF samples, with only preliminary evidence showing the ability of the metabolic profiles to distinguish OA patients from the controls in a small population, thus supporting the need for further efforts to develop this promising biomarker approach in large-scale metabolomic studies. ⁶⁷ Similarly, the association between biological markers and imaging markers could provide better results. For example, Dam et al. ¹⁸⁴ suggested that a combination of biochemical and MRI-based markers can improve diagnosis and prognosis of knee OA and may be useful to select high-risk patients for inclusion in disease-modifying osteoarthritis drugs (DMOAD) clinical trials. Although promising, the analysis of such heterogeneous aspects could seem challenging with the current technologies, but the emerging advances in machine learning protocols may open a completely new horizon for a multidimensional evaluation of OA patients. ¹⁸⁵

The limitations of the current systematic review reflect those of the included studies and their limited quality. In fact, while the literature analysis was able to identify a large number of studies and 201 molecules, only a few papers supported each biomarker, thus impairing the possibility to draw definitive conclusions in this field. High-quality studies for a large-scale investigation of these SF biomarkers are crucial in order to increase the knowledge about the applicability of these biomarkers and even favor the understanding of OA pathogenesis. In particular, this systematic review underlined the paucity of longitudinal studies on biomarkers. In fact, only 21 studies, out of the 159 included, evaluated the biomarker changes over time. Considering that the very high interindividual variation in biomarker levels can most accurately be determined in a longitudinal fashion, research efforts should focus on further studies of this type. Moreover, further studies are needed to better understand the biomarker patterns in the early stages of OA, including the evaluation of biomarker changes in joints affected by ACL injuries, meniscal tears, and focal chondral lesions. In these clinical settings, changes in biomarker patterns have been demonstrated, showing the possibility to identify subsets of patients exhibiting dysregulation of the inflammatory response and an increased risk of posttraumatic osteoarthritis.^186-188 Another limitation is that the literature presents scattered and incomplete information about several aspects, such as sex, type of OA, BMI, age, and their role in biomarker OA correlation. All these factors might influence the biomarker results, and future studies should investigate how these key aspects can affect the role of the plethora of molecules as synovial OA biomarkers.

Research efforts could identify new and more sensitive/specific biomarkers, or improve existing biomarker assays, with better technologies and platforms favoring a more consistent and standardized detection and measurement. In fact, most biomarkers are currently in an explorative stage, and there are still no “normal ranges” used by the various authors to foster study comparisons and the adoption of a common reference for different studies. Moreover, most of the literature retrieved did not provide conclusions on the sensitivity and specificity of the selected biomarkers. Due to the limitations and heterogeneity in data reporting, it was not possible to perform a meta-analysis of the results. On the other hand, the systematic review documented a rapidly evolving field, underlining potential and limitations of several molecules and identifying promising areas for future studies on biomarkers. To our knowledge, this is the first systematic evaluation of the literature on SF biomarkers carried out in accordance with the BIPEDs classification. BIPEDs classification aims at improving our collective ability to match a biomarker with its appropriate purpose to enable greater speed, efficiency, and precision in the development of useful diagnostic and therapeutic technologies and strategies, in the end providing a sound foundation for the development and implementation of public health policies. In this light, the BIPEDs classification in this systematic review could simplify the use of SF biomarkers for therapeutic development, clinical research, and patient care.

In conclusion, the systematic review of the literature identified numerous SF biomarkers. According to BIPEDs classification, out of the 201 SF biomarkers identified, the “burden of disease” category is the most represented with 132 different biomarkers, followed by the 106 “diagnostic” biomarkers, while “efficacy of intervention” and “prognostic” SF biomarkers were less investigated, being only 20 and 14, respectively. The most promising SF biomarkers were C4S, IL-6, IL-8, Leptin, MMP-1/3, TIMP-1, TNF-α, and VEGF. However, despite the large number of studies on the plethora of identified molecules, the evidence about the efficacy of each biomarker is supported by limited and often conflicting findings, and further research efforts are needed to improve the understanding of SF biomarkers for a better management of patients affected by knee OA.