Synovial fluid adipokines are associated with clinical severity in knee osteoarthritis: a cross-sectional study in female patients with joint effusion

We conducted a cross-sectional study with systematic inclusion of female patients with symptomatic primary KOA according to American College of Rheumatology criteria [20] who visited our hospital for a monographic OA consultation. The women were aged 50–85 years and had evidence of significant joint effusion based on physical examination and confirmed by ultrasound (≥4 mm on midline suprapatellar line). Symptomatic OA was defined as pain intensity rated as ≥4 on a 10-cm visual analogue scale despite the use of prescribed analgesic drugs for at least 3 months. Only patients reporting persisting knee effusion or with documented effusion in several consultations were entered into the study. Patients with secondary OA were excluded, such as those with a history of trauma, meniscal injury, inflammatory rheumatic or septic conditions, previous knee surgery, any condition that could interfere with pain perception, systemic glucocorticoid intake in the last 6 months, or intra-articular glucocorticoid or hyaluronic acid injection in the last 3 or 6 months, respectively. The recruitment period was October 2013 to June 2015. We included only female patients to homogenize the sample, as there are differences between men and women related to pain perception, anthropometric measures, and fat content and distribution that might influence the adipokine profile [21‐23]. This study was approved by the local ethics committee at the Parc Tauli Sabadell University Hospital. All patients included were verbally informed about the study and signed informed consent forms.

Information on the following variables was collected: age, physical exercise (never, occasional [less than 150 minutes per week], or regular), tobacco exposure, and KOA symptom duration. Each participant’s medical history, specifically regarding the presence of cardiovascular risk factors, was recorded. A diagnosis of hypertension, dyslipidemia, or diabetes was established if already diagnosed or if the participant was receiving treatment for any of these conditions. Anthropometric measurements included weight (kg), height (cm), body mass index (BMI) (kg/m²), waist circumference (WC) (cm), hip circumference (cm), waist-to-hip ratio (WHR), and percentage of body fat measured using a bioimpedance analyzer (BC-418 MA; Tanita, Arlington Heights, IL, USA) according to a standard protocol. Obesity was defined as a BMI ≥30 kg/m². In accordance with the modified criteria of the National Cholesterol Education Program Adult Treatment Panel III, MetS was defined as having three or more of the following conditions: fasting plasma glucose ≥100 mg/dl or treatment with glucose-lowering drugs, arterial blood pressure ≥130/85 mmHg or antihypertensive medication, fasting plasma triglycerides ≥150 mg/dl (1.7 mmol/L) or drug treatment for hypertriglyceridemia, high-density lipoprotein (HDL) cholesterol <50 mg/dl (1.28 mmol/L) or drug therapy to raise HDL cholesterol concentration, and WC ≥88 cm [24]. Radiographic severity was evaluated by anteroposterior knee x-ray examination with the patient in standing position performed in the last 18 months and graded according to the Kellgren-Lawrence (KL) scale (grades 1–4). Time between radiologic evaluation and visit was considered for adjustment in the statistical analysis. Two rheumatologists (JC, CO) evaluated x-rays independently. The Lequesne algofunctional index, a simple and validated questionnaire for pain and disability in KOA with scores ranging from 0 (best) to 24 (worst), was used to assess clinical severity.

Joint aspiration was performed during the visit and at the same time of the day for proper evaluation of synovial adipokines. Synovial fluid was analyzed to ensure noninflammatory fluid (joint cell count <2500 cells) and absence of microcrystals. Synovial samples were stored at −80 °C. Seven adipokines and three inflammatory markers were measured by enzyme-linked immunosorbent assay (ELISA) in accordance with the manufacturers’ recommendations for synovial fluid dilutions: leptin (Biocompare, South San Francisco, CA, USA), adiponectin (eBioscience, San Diego, CA, USA), resistin (RayBiotech, Norcross, GA, USA), osteopontin (eBioscience), visfatin (Phoenix Pharmaceuticals, Burlingame, CA, USA), omentin (CUSABIO, Wuhan, China), chemerin (Elabscience, Bethesda, MD, USA), hs-CRP (DRG Diagnostics, Marburg, Germany), and IL-6 and TNF-α (Milliplex HCYTOMAG-60 K-03; Merck Millipore, Billerica, MA, USA) (see Additional file 1 for detailed description). For technical reasons related to ELISA technology (configuration of the plates used), none of these markers could be assessed at the same time for all patients. In order to control for technical variability, the rounds of measurement were considered as an adjustment factor in the statistical analyses.

Clinical data and laboratory parameters and their association with Lequesne index score were summarized using nonparametric methods. Medians, interquartile ranges, and Spearman’s correlations (r) were used for continuous measures, while frequencies and Mann-Whitney or Kruskal-Wallis tests were applied to categorical variables. For estimation of adjusted effects, linear models were fitted with suitable transformation of explanatory variables when necessary in order to fit the model assumptions. The covariates included in the multivariate analyses were age at recruitment, KL grade, time from visit to date of radiologic assessment, and KOA symptom duration. Using this model as a starting point, a stepwise algorithm was used sequentially with the rest of the potential confounders in two stages: first, only anthropometric and metabolic parameters were evaluated; second, inflammatory markers were considered in the resulting model, then each adipokine was added one at a time to the resulting model to assess its association with KOA severity (Table 3, “Adjusted effects” column). Finally, all adipokines were added to the model to assess their association with the Lequesne index after controlling for the anthropometric, metabolic, and inflammatory markers found to be informative in the previous steps, as well as for the rest of the adipokines (“Multivariate model” column in Table 3). For interpretation purposes, the partial correlation coefficient (PCC) was used as a measure of association for continuous variables. Associations for adipokines and inflammatory markers were adjusted by measurement round in all cases. Owing to high collinearity observed between adiponectin and omentin (PCC 0.792), only one of them at a time was included in a linear model. When selection of confounders was needed, a stepwise algorithm was carried out using Akaike’s information criterion for model selection. Associations were assessed in the linear models using the corresponding F and Wald tests. Tests were performed at the 5 % significance level. All statistical analyses were conducted using R software (see Additional file 2 for detailed description).