Introduction

Joint pain is the key clinical feature of osteoarthritis (OA) [1–3] and plays a major role in the decision for a total joint arthroplasty, but the causes of OA pain are not well established [4] and correlation between pain and structural joint damage is weak [5]. A possible link between joint pain and the adipokine leptin, which is considered as a key mediator in the well established association between obesity and osteoarthritis (OA) [6–13], has been indirectly suggested in studies on weight loss. Weight reduction consistently resulted in decreased joint pain in patients with OA [14, 15] and simultaneously decreased blood leptin levels [16–18]. Furthermore, leptin is particularly elevated in women and in obese patients [10, 19], and in these two patient groups, higher OA pain levels have been reported [20–26]. Studies from outside the musculoskeletal field have lent further support to a possible association between pain and leptin concentrations. Leptin levels in peritoneal fluid correlated positively with pain severity in endometriosis patients [27]. Furthermore, recent studies in rodents [28–30] suggested that spinal leptin might be involved in the pathogenesis of neuropathic pain.

So far two clinical studies have evaluated the association between adipokines and OA pain. Gandhi et al. [31] did not find a clear association between intra-articular leptin concentrations and pre-operative pain, but they reported that the synovial fluid (SF) adiponectin/leptin ratio predicted pain. Massengale et al. [32] showed that leptin serum concentration was associated with the intensity of pain in patients with chronic hand OA.

The objective of this study was to investigate if leptin may influence joint pain in OA patients. We hypothesized that SF leptin concentrations correlate with pain severity, and mediate the association between joint pain and (1) female gender and (2) obesity.

Materials and methods

Study design and patient population

We conducted a cross-sectional study including patients with end-stage OA of the hip and knee who presented for a total joint arthroplasty at a large orthopaedic centre between January and December 2010. Only patients with primary hip or knee OA were eligible. The study was approved by the local ethics committee and informed consent was obtained from all patients. Overall, synovial fluid (SF) was collected from 250 patients. Of those 219 had completed the pain scores and they were included in the final analysis.

Exposures of interest

Exposure of interest was the intra-articular (SF) concentration of leptin obtained at the time of joint replacement surgery (hip or knee). It was measured by a sandwich ELISA technique using the DuoSet ELISA Development Systems (R&D Systems, Abingdon, UK) according to the manufacturer’s protocols. The minimum detectable cytokine concentration for these assays was estimated to be 31 pg/ml for leptin.

Sex and body mass index (BMI) were the clinical factors of interest. BMI was analysed as continuous and dichotomised variable (BMI < 30 kg/m2 = non-obese vs. ≥ 30 kg/m2 = obese).

Outcome of interest

One outcome of interest was the pre-operative pain level measured with the Western Ontario McMaster Universities Osteoarthritis Index (WOMAC) pain score [33]. We employed the 5-point Likert scale version of the WOMAC 3.0. We analysed the score both as continuous (0–100, 100 = no pain) and as dichotomized variable (> 25 vs. ≤ 25 = quartile with greatest pain). We also used the visual analog scale (VAS) for pain assessment. The scale was evaluated as continuous (0–10, 0 = no pain) and as a dichotomized variable (< 7 vs. ≥ 8 = quartile with greatest pain). Both patient-assessed instruments are validated and widely used. In addition, the physician assessed pain item of the Harris hip score (HHS) (0–44, 44 = no pain) [34] was evaluated in patients with hip OA, and the physician-assessed pain item of the American Knee Society score (AKSS) (0–50, 50 = no pain) [35] was evaluated in patients with knee OA.

Co-variates and potential confounders

The following variables were assessed: (a) age; (b) American Society of Anesthesiologists (ASA) score investigated as binary variable (ASA < 3 vs. ≥ 3); (c) OA site (hip vs. knee); (d) presence of diabetes; (e) Medical Outcomes Study Short Form-12 (SF-12), a patient-administered generic health-related quality of life measure comprising a mental (mcs) and physical component score (pcs) [36]; the summary measures range from 0 to 100 (100 = best); (f) presence of a contra-lateral arthritic or operated joint; (g) Charnley disability category C for the hip and AKSS disability category C for the knee; the category C is defined as presence of multiple joint disease or other disabilities leading to difficulties in ambulation [35, 37]; (h) radiological severity of OA (Kellgren-Lawrence grades 0–4 and presence of osteophytes, subchondral sclerosis and minimal joint space width [minJSW]) [38]; (i) serum (blood) leptin concentration in ng/ml; and (j) synovial fluid adiponectin concentration in μg/ml.

Data collection

Synovial fluid (1 ml) was taken during the intervention by the operating surgeon by direct aspiration through the joint capsule, after skin incision, in order to minimize blood contamination. Blood samples (5 ml) were obtained during the pre-operative examination, at the same time as the pre-operative blood draw. A medical secretary was responsible for SF and blood sample transport. Patients’ samples were made anonymous prior to transport to the laboratory. All blood and SF samples were immediately centrifuged, aliquoted and frozen at −80 °C, until they were measured.

Since 1996 all patients undergoing total hip arthroplasty (THA) and since 1998 all those undergoing total knee arthroplasty (TKA) in the Geneva University Hospitals have been routinely enrolled in a prospective hospital-based cohort. The patients included in this study are part of the two cohorts. Information about baseline characteristics including medical and orthopaedic co-morbidities is routinely documented on specifically designed data collection forms. Pre-operative score and questionnaire assessment is routinely performed for all patients undergoing joint arthroplasty surgery. Harris hip score in case of hip OA and AKSS score in knee OA were assessed by the orthopaedic surgeon in charge of the patient the day before surgery. The WOMAC and the SF-12 were sent to all patients seven to ten days prior to surgery. The medical secretary was involved in sending and collecting scores and questionnaires and in data entry. The assessment of the radiographic severity of OA and identification of specific OA features was performed on pre-operative radiographs by two orthopaedic surgeons (GP, DS) blinded to the patient’s laboratory and clinical results.

Statistical analysis

The distribution of leptin concentrations (synovial fluid and serum) was right-skewed. To allow for comparison with previous literature on leptin and OA, we reported both mean (standard deviation, ± SD) and median (interquartile range, IQR) concentrations.

The association between SF leptin and preoperative pain levels was evaluated (1) as continuous variable using the Spearman’s correlation coefficient and (2) as categorical variable using quartiles. In particular, we compared pain levels in the 4th quartile (> 19.6 ng/ml) to those in the combined 1st to 3rd quartiles (≤ 19.6 ng/ml). First, we evaluated the percentage of patients with severe pain (defined as WOMAC ≤ 25 and VAS ≥ 8) in the various leptin categories and computed relative risks (RR) of severe pain by leptin category. Second, we calculated mean pain score differences using linear regression analysis. Because differences in pain levels could be related to other factors (confounding factors), we adjusted the analysis for age, diabetes, ASA score, contra-lateral arthritic joint, OA site and disability category C. In a second model we additionally adjusted for the SF-12 mental component score. This was done to separately show the influence of mediating psychological factors (mental health) on pain expression. We first analysed the data separately in patients with hip and knee OA, but decided the priority was to combine the results in the absence of effect modification by OA joint. Baseline characteristics are presented according to leptin quartiles.

In the second study objective, we explored whether leptin could mediate the association between pain and (1) female gender and (2) obesity in patients with OA. We first assessed crude pain differences by gender and by BMI categories and then we examined if adjusting for SF leptin concentration in a multivariate regression model would mitigate these univariate associations.

Results

Overall, 219 patients were evaluated, 123 with hip (56.2 %) and 96 with knee OA. Mean age was 72 (±9) years, and 59 % were women (n = 130). Mean BMI was 28.0 (±4.9) kg/m2, 27.3 kg/m2 in the hip group and 29.1 kg/m2 in the knee group. The median SF leptin concentration was 8.7 (IQR 3.0, 19.6) ng/ml and the mean concentration 15.8 (±21.8) ng/ml. In addition, mean SF leptin concentrations were 22.9 (±25.6) ng/ml in women compared to 5.4 (±5.9) ng/ml in men, and they were 12.4 (± 15.6) ng/ml in patients with hip OA compared to 20.1 (± 27.2) ng/ml in those with knee OA.

SF leptin concentrations were divided into quartiles with the highest quartile corresponding to levels > 19.6 ng/ml. The proportion of women increased strongly with rising leptin levels, up to 94 % in the highest quartile (Table 1). Similar results were seen for obesity, with the greatest proportion (59 %) of obese patients in the highest quartile. In addition, patients in the highest leptin category had less often diabetes, and more often knee OA, a contralateral arthritic joint, as well as an ASA score of 3–4. The SF-12 mcs and pcs scores were the lowest in this group. The radiographic severity of hip and knee OA did not differ substantially according to SF leptin quartiles (Table 1).

Table 1 Pre-operative patient characteristics and radiological features according to leptin concentration (in quartiles) in synovial fluid
Full size table

Synovial fluid leptin and pain

There was a weak correlation between SF leptin concentrations (continuous) and WOMAC pain (Spearman’s correlation coefficient r = −0.182) as well as between leptin and VAS pain (r = 0.115). Patients with high SF leptin concentrations (highest quartile: >19.6 ng/ml) reported higher pain levels on both the WOMAC and VAS pain scales. In the highest leptin quartile, 45 % of patients had a WOMAC pain score ≤ 25 compared to 20 % in the other quartiles (RR 2.3, 95 % CI 1.5; 3.5), and 42 % had a VAS pain score ≥ 8 compared to 21 % in the other quartiles (RR 2.0, 95 % CI 1.3–3.0) (Figs. 1 and 2). This association between SF leptin levels and pain (shown for WOMAC as percentage of patients ≤ 25; highest quartile vs. others) was apparent both in women (46 vs. 21 %) and men (33 vs. 19 %), obese (45 vs. 23 %) and non-obese patients (46 vs. 19 %), and in hip (57 vs. 24 %) and knee OA patients (37 vs. 14 %) (Figs. 3, 4 and 5).

Fig. 1
figure 1

Proportion of patients with WOMAC pain scores ≤ 25 (highest pain levels) according to intra-articular leptin concentration (in quartiles)

Full size image
Fig. 2
figure 2

Proportion of patients with VAS pain scores ≥ 8 (highest pain levels) according to intra-articular leptin concentration (in quartiles)

Full size image
Fig. 3
figure 3

Proportion of WOMAC pain scores ≤ 25 in women and men according to intra-articular leptin concentration (highest quartile [leptin > 19.6 ng/ml] vs. all other quartiles [≤19.6 ng/ml])

Full size image
Fig. 4
figure 4

Proportion of WOMAC pain scores ≤ 25 in obese and non-obese patients according to intra-articular leptin concentration (highest quartile [leptin > 19.6 ng/ml] vs. all other quartiles [≤19.6 ng/ml])

Full size image
Fig. 5
figure 5

Proportion of WOMAC pain scores ≤ 25 in patients with hip and knee osteoarthritis according to intra-articular leptin concentration (highest quartile [leptin > 19.6 ng/ml] vs. all other quartiles [≤19.6 ng/ml])

Full size image

The unadjusted mean score difference between patients in the highest leptin quartile compared to those in the other quartiles was −9.6 (95 % CI −15.1; −4.0; standardized effect size 0.54) for the WOMAC and 0.8 (95 % CI 0.2; 1.3; standardized effect size 0.44) for the VAS (Table 2). The association remained virtually unchanged after adjusting for potential confounders such as age, ASA score, contra-lateral arthritic joint, OA site, disability category C and diabetes (WOMAC adjusted mean difference −9.1 [95 % CI −14.8 to −3.5] and VAS adjusted mean difference 0.7 [95 % CI 0.2–1.3]). Additional adjustment for the SF-12 mental component score (WOMAC adjusted mean difference −6.8 (95 % CI −12.2 to −1.3) and VAS adjusted mean difference 0.6 [95 % CI 0–1.2]) slightly lowered the estimates, but the association remained significant. A greater degree of pain was seen in patients with high leptin levels as assessed by both the patient-assessed and the joint-specific physician-assessed scores HHS and AKSS (Table 2).

Table 2 Pre-operative pain levels according to leptin concentrations in synovial fluid
Full size table

Significant associations between increased joint pain as measured with WOMAC and (1) female gender and (2) obesity observed in univariate analyses were substantially reduced after adjusting for SF leptin concentrations. The mean difference in WOMAC for women vs. men decreased from 4.7 to 1.6 and from 7.0 to 4.1 for obese vs. non-obese after correcting for leptin concentrations, suggesting that these associations might partially be mediated by leptin (Table 3).

Table 3 Association between WOMAC pain and (1) sex and (2) body mass index
Full size table

Correlation with serum leptin

Synovial fluid leptin levels strongly correlated with serum leptin levels (correlation coefficient r = 0.875). Overall, the median serum leptin concentration was 15.8 (IQR 5.7, 36.9) ng/ml and the mean concentration 27.9 (±34.3) ng/ml. In addition, mean serum leptin concentrations were 38.7 (±39.5) ng/ml in women compared to 11.6 (±13.2) ng/ml in men.

Discussion

We found that OA patients with high SF leptin concentrations reported substantially more OA pain. Furthermore, the increased levels of pain observed in women and in obese patients seemed to be associated with high intra-articular leptin concentrations. However, due to the cross-sectional design a causal relation between leptin and pain cannot be ascertained.

Pain is certainly the most important symptom of OA, but also the least well understood [39]. Recent models imply that OA pain consists of a nociceptive stimulation of the joint and of peripheral sensitization, which can be accompanied by spinal and central sensitization [2, 40–43]. Leptin may affect the OA pain response in at least two possible ways. First, leptin appears to be a mediator of the immuno-inflammatory response in different experimental models. The pro-inflammatory cytokine-like functions of leptin in OA [10, 44–47] could explain its role in peripheral sensitization. Second, animal models suggest an involvement of leptin in the pathogenesis of pain at the spinal level and a possible role in the development of neuropathic pain [28–30]. Our clinical results support these experimental findings in patients with lower limb OA.

Our findings are in accordance with Massengale et al. [32] who, in a cohort of patients with chronic hand OA, reported an association between serum leptin levels with pain intensity but not with radiographic severity. Ghandi et al. [31] have investigated the relation between leptin and pain. They did not find a clear association between intra-articular leptin concentrations and pain in a linear regression analysis, but reported that the synovial fluid (SF) adiponectin/leptin ratio predicted pain. We found that the association between intra-articular leptin concentration and pain intensity was not linear. Pain intensity was similar within the lower three quartiles, but substantially higher in the highest quartile of leptin concentration.

Clinical implications of these findings might be: (1) exercise and weight loss programs should be particularly targeted to women, who constitute the vast majority of the high leptin level group; and (2) prospective studies are warranted to test the clinical usefulness of serum leptin as a biomarker (as a proxy for intra-articular leptin level) to identify subgroups of patients, who would particularly benefit from exercise and weight loss programs, to detect the presence of neuropathic pain reported to be common in knee OA patients [48, 49], as well as to predict persistence of pain after total joint arthroplasty.

The main limitation of this analysis is its cross-sectional design, and therefore the study is not suited to establish a causal link between leptin and increased pain. Leptin is a pleiotropic cytokine with central and peripheral effects involved in the regulation of energy intake and expenditure, of bone metabolism and inflammatory responses as well as of stress response [50]. Despite the above mentioned evidence from experimental studies supporting a possible causal effect of leptin on pain, other explanations for our findings including inverse causation and residual confounding need to be explored in future studies.

Second, we evaluated pain with two widely used, patient-assessed validated instruments. Because physician evaluation of pain has been found to be less influenced by the presence of co-morbidities [51], we additionally included one hip- and one knee-specific physician-assessed pain sub-score. However, information about the quality of pain (pain on movement, pain on rest, morning stiffness) was not routinely recorded. Third, depression and comorbidities, in particular the presence of OA at multiple sites, have been identified as important factors influencing the perception and severity of pain [20, 52–54].

A depression-specific instrument was not used in our study, but we collected information on mental health status using the SF-12 mental component score. Moreover, specific information about OA in other joints except for the contra-lateral joint was not systematically available. However, we assessed (and adjusted for) the presence of multiple joint disease and other medical disabilities leading to difficulties in ambulation.

Conclusion

Our results suggest that high leptin concentrations may affect the level of joint pain in OA of the lower limbs. Furthermore, elevated leptin levels may potentially explain the well established association between increased pain and female sex or obesity. However, the causal relation between leptin and joint pain needs to be confirmed in a longitudinal study.