Individual participant data meta-analysis of metabolomics on sustained knee pain in primary osteoarthritis patients

Abstract

Objectives

Knee pain is the major driver for OA patients to seek healthcare, but after pursuing both conservative and surgical pain interventions, ∼20% of patients continue to report long-term pain following total knee arthroplasty (TKA). This study aimed to identify a metabolomic signature for sustained knee pain after TKA to elucidate possible underlying mechanisms.

Methods

Two independent cohorts from St John’s, NL, Canada (n = 430), and Toronto, ON, Canada (n = 495) were included in the study. Sustained knee pain was assessed using the WOMAC pain subscale (five questions) at least 1 year after TKA for primary OA. Those reporting any pain on all five questions were considered to have sustained knee pain. Metabolomic profiling was performed on fasted pre-operative plasma samples using the Biocrates Absolute IDQ p180 kit. Associations between metabolites and pair-wise metabolite ratios with sustained knee pain in each individual cohort were assessed using logistic regression with adjustment for age, sex and BMI. Random-effects meta-analysis using inverse variance as weights was performed on summary statistics from both cohorts.

Results

One metabolite, phosphatidylcholine (PC) diacyl (aa) C28:1 (odds ratio = 0.66, P = 0.00026), and three metabolite ratios, PC aa C32:0 to PC aa C28:1, PC aa C28:1 to PC aa C32:0, and tetradecadienylcarnitine (C14:2) to sphingomyelin C20:2 (odds ratios = 1.59, 0.60 and 1.59, respectively; all P < 2 × 10⁻⁵), were significantly associated with sustained knee pain.

Conclusions

Though further investigations are needed, our results provide potential predictive biomarkers and drug targets that could serve as a marker for poor response and be modified pre-operatively to improve knee pain and surgical response to TKA.

Rheumatology key messages.

• A significant proportion of knee OA patients report treatment-resistant sustained pain that continues after total knee arthroplasty.

• A meta-analysis of metabolomics highlighted phosphatidylcholine metabolism as a key driver of sustained pain.

• The findings have potential utility in prediction and treatment of sustained pain.

Introduction

OA impacts over 30% of the worldwide population above 60 years old [1, 2]. Joint pain acts as a major driver for patients to seek healthcare [3]. The aetiology of joint pain in OA is complex; multiple factors contribute to pain perception including joint pathology, biomechanics, psychological state, comorbidities and sensitization [4–6]. This complexity is further highlighted by discordance between pain sensation and radiographic OA severity, which are weakly correlated [7] despite prevalence of plain radiograph as an assessment tool for OA, indicating more factors than just structural changes drive pain.

Conservative treatments for OA, focusing on pain management, are limited. First-line treatments include exercise, weight management and physiotherapy to improve joint function and supporting structures, indirectly alleviating pain. Pharmacological therapies such as oral and/or topical non-steroidal anti-inflammatory drugs, acetaminophen and IA CS injections directly target pain [8]. When conservative strategies are ineffective, surgery is considered. Total joint arthroplasty (TJA) is the most effective treatment for end-stage OA; burden of primary and revision TJA are steadily rising in many countries due to increasing OA prevalence and population longevity. Rates for total knee arthroplasty (TKA) specifically are high; yearly, over 75 000 TKAs are performed in Canada [9] and over 700 000 in the USA [10]. These numbers are expected to rise, with a projected increase of 673% by 2030 [11].

Despite its effectiveness, up to 44% of patients still report pain 3–4 years post-TKA [5]. Current literature on chronic post-TKA pain suggests biological, surgical and psychosocial risk factors [3, 5, 12]; however, the cause of the treatment-resistant sustained pain remains elusive. Given the volume of affected patients and associated costs, understanding the mechanisms of this pain warrants urgent attention and provides opportunities to develop strategies to identify patients at risk of sustained pain, tailor pre- and post-surgical treatments to improve outcomes and overall quality of life, and reduce the burden on the healthcare system. To elucidate possible underlying mechanisms, we conducted an individual participant data meta-analysis of metabolomics on sustained knee pain in primary OA patients using two independent cohorts.

Methods

Study participants

Patients were derived from two cohorts: the Newfoundland Osteoarthritis Study (NFOAS), recruited from patients undergoing TKA between 2011 and 2017 at the Health Sciences Centre and at St Clare’s Mercy Hospital in St John’s, Newfoundland and Labrador (NL), Canada, and the Longitudinal Evaluation in the Arthritis Program, Osteoarthritis Study (LEAP OA), recruited from patients undergoing TKA between 2014 and 2017 at Toronto Western Hospital in Toronto, Ontario (ON), Canada. Primary knee OA diagnosis for the NFOAS cohort was based on the ACR diagnostic criteria and was confirmed by the attending orthopedic surgeon and pathology reports of the removed articular cartilage following surgery. Primary knee OA diagnosis for the LEAP OA cohort was also based on ACR clinical criteria for knee OA classification [13]. Ethics approval for the study was received from the Health Research Ethics Authority of NL (11.311) and University Health Network (REB 16-5759) and informed written consent was received from all study participants.

Sustained pain

To classify sustained pain, we used the WOMAC Likert 3.0 pain subscale, administered to patients at least 1 year post-surgery. This subscale consists of five questions, each rated on a scale from 0–4, with 0 being no pain and 4 being severe pain, assessing self-reported pain when patients are walking on a flat surface, going up and down stairs, in bed at night, sitting or lying, and standing upright [14]. Patients with sustained pain were primarily considered to be those who reported at least one point in each of the five questions. Three other pain patterns were also considered to maximize robustness of results and minimize potential misclassification: patients who reported at least one point in one question, patients who reported at least one point in the ‘pain when sitting and/or lying’ question and patients who reported at least one point in the ‘pain at night while in bed’ question. Patients without pain for all four definitions were those who reported no pain in any question of the WOMAC pain subscale.

Metabolomic data collection

Blood samples were collected pre-operatively after 8h of fasting. Plasma was separated from whole blood following a standard protocol and stored at –80°C freezers until analysis [15]. Metabolomic profiling was performed using the Biocrates AbsoluteIDQ p180 kit (BIOCRATES Life Sciences AG, Innsbruck, Austria) which quantifies 186 metabolites including acylcarnitines, amino acids, biogenic amines, lysophosphatidylcholines (lysoPC), phosphatidylcholines (PC), sphingomyelins (SM) and more (supplementary Table S1, available at Rheumatology online). Details of the metabolic profiling method used in this kit [16] and the in-house reproducibility assay [17] have been reported previously. Briefly, reproducibility was assessed using mean coefficient of variation (CV) for all metabolites in 23 samples. Average CV for all metabolites was 0.07 (0.05); 90% of metabolites had a CV >0.10. Metabolomic profiling was done using an API4000 Qtrap^® tandem mass spectrometry instrument (Applied Biosystems/MDS Analytical Technologies, Foster City, CA, USA) equipped with Agilent 1100 HPLC system at The Metabolomics Innovation Centre, AB, Canada (https://www.metabolomicscentre.ca). The analytical and quality control (QC) process for the profiling was completed using the MetIDQ software package as part of the AbsoluteIDQ p180 kit. Metabolite concentrations were reported as μMol.

Statistical methods

Prior to analysis, metabolomic data underwent a strict QC procedure. Metabolites were removed from analysis if >10% of samples had values below the limit of detection. For metabolites with <10% of samples having values below the limit of detection, missing values were imputed using the metabolite mean for the cohort. Metabolites that passed QC in both cohorts (n = 137 metabolites) were included in final analysis. Raw metabolite data for both cohorts was standardized to the mean before testing for association with sustained pain using a generalized estimating equation for the NFOAS cohort, as some participants had both knees included, and logistic regression for the LEAP OA cohort as there was only one knee included per individual participant. Models in both cohorts were adjusted for age, sex and BMI. Pairwise ratios (n = 18 632) were generated from raw metabolite data and standardized to the mean before testing for association with sustained pain using a generalized estimating equation for the NFOAS cohort and logistic regression for the LEAP OA cohort, with adjustments for age, sex and BMI. For metabolites and metabolite ratios, summary statistics from each cohort were subjected to random-effects meta-analysis modelling using inverse variance as weights. For individual metabolites, a Bonferroni correction for multiple testing was applied and significance was defined as α = 0.00037. For metabolite ratios, significance was defined using the proposed metabolome-wide significance level (α = 2 × 10⁻⁵) [18]. The analysis was performed using R Version 4.0.3 with built-in functions [19] and package geeM [20].

Results

Descriptive statistics

In total, 430 knee joints belonging to 363 patients from the NFOAS cohort and 495 knee joints belonging to 495 patients from the LEAP OA cohort were included in the final analysis. The average age of NFOAS patients was 65.2 (7.5) years and the average BMI was 34.9 (6.9) kg/m²; 57.4% of patients were female. The average time to follow-up was 4.0 (1.3) years. In total, 9.8% of patients experienced sustained pain, 24.9% of patients reported pain in at least one question, 15.0% of patients reported pain while sitting or lying, and 15.7% of patients experienced pain while in bed. In total, 67 patients had two knee joints included in final analysis. The sustained pain status of 63/67 patients (94%) was concordant between both knee joints; the rate of sustained pain was lower in patients with both knees replaced than those with one knee replaced although it was not statistically significant (supplementary Table S2, available at Rheumatology online).

The average age of the LEAP OA cohort was 65.5 (8.4) years and the average BMI was 30.8 (6.0); 57.2% of patients were female. The average time to follow-up was 1.0 (0.1) years. In total, 50.0% of patients experienced sustained pain, 76.2% of patients reported pain in at least one question, 63.0% of patients reported pain while sitting or lying and 64.0% of patients experienced pain while in bed.

Clinical assessment

Younger age was significantly associated with sustained pain and all alternate pain definitions in the NFOAS cohort (P < 0.05; Table 1 and supplementary Table S3, available at Rheumatology online); there was no statistically significant difference in sex or BMI (P > 0.05).

Table 1.

Comparison of demographic factors between sustained knee pain case and control in the NFOAS and LEAP OA cohorts

	NFOAS			LEAP OA
	Controls	Cases	P	Controls	Cases	P
N (%)	323 (90.2)	35 (9.8)	NA	118 (50)	118 (50)	NA
Age (years)	65.7 (7.7)	61.8 (5.6)	0.004	66.1 (7.8)	64.1 (8.7)	0.06
Sex (% female)	57.5	60	0.92	56.8	58.4	0.89
BMI (kg/m2)	34.7 (7.1)	35.3 (5.1)	0.61	29.6 (5.2)	31.8 (6.6)	0.004

Values are either mean (s.d.) or percentage. P-values were obtained by Student’s t-test (age, BMI) or Chi-squared test (sex). NFOAS: Newfoundland Osteoarthritis Study; LEAP OA: Longitudinal Evaluation in the Arthritis Program, Osteoarthritis Study; NA: not applicable.

Higher BMI was significantly associated with sustained pain and all alternate pain definitions in the LEAP OA cohort (P < 0.05; Table 1 and supplementary Table S3, available at Rheumatology online); there was no statistically significant difference in age or sex (P > 0.05). Age and sex distributions between the NFOAS and LEAP OA cohorts were relatively consistent; there was no significant difference in either variable between the two cohorts while BMI was significantly higher in the NFOAS cohort than in the LEAP OA cohort (P < 2.2 × 10⁻¹⁶; Table 2).

Table 2.

Comparison of the sustained pain prevalence and demographic factors between NFOAS and LEAP OA Cohorts

	NFOAS	LEAP OA	P
Sustained pain cases (%)	9.8	50.0	<2.2 × 10–16
Pain in one question cases (%)	24.9	76.2	<2.2 × 10–16
Pain while in bed cases (%)	15.7	64.0	<2.2 × 10–16
Pain while sitting or lying cases (%)	15.0	63.0	<2.2 × 10–16
Age (years)	65.2 (7.5)	65.5 (8.4)	0.5
Sex (% female)	57.4	57.2	0.99
BMI (kg/m2)	34.9 (6.9)	30.8 (6.0)	<2.2 × 10–16

Values are either mean (s.d.) or percentage. P-values were obtained by Student’s t-test (age, BMI) or Chi-squared test (sex, sustained pain). NFOAS: Newfoundland Osteoarthritis Study; LEAP OA: Longitudinal Evaluation in the Arthritis Program, Osteoarthritis Study.

Metabolomic association analysis

We tested 137 metabolites and 18 632 metabolite ratios for association with sustained pain in each cohort and subsequently performed a meta-analysis. The volcano plots in Fig. 1 present the meta-analysis results of individual metabolites and metabolite ratios using the primary sustained pain definition; volcano plots for the other pain patterns can be found in supplementary Figs S1–S3, available at Rheumatology online. With the pre-defined significance level, we found that one metabolite and three metabolite ratios were significantly associated with sustained pain (Fig. 2): PC diacyl (aa) C28:1, PC aa C32:0 to PC aa C28:1, PC aa C28:1 to PC aa C32:0, and acylcarnitine C14:2 to SM C20:2 ratios. In addition, acylcarnitine C16:2 to SM C20:2 ratio was associated with reporting pain while in bed (Fig. 3). Similar effect sizes and directions were found for the significant metabolite and metabolite ratios between the two cohorts with very little heterogeneity, as measured by I² (Figs 2 and 3).

Figure 1.

Meta-analysis of individual metabolites and metabolite ratios for sustained pain. Volcano plots of the meta-analysis results on individual metabolites (a) and metabolite ratios (b) for sustained pain. *Blue dashed line indicates the pre-defined significance level and the coloured dots indicate the significant metabolite/metabolite ratios. P-values were obtained by generalized estimating equation adjusted for age, sex and BMI for the NFOAS cohort and by logistic regression adjusted for age, sex and BMI for the LEAP OA cohort, from which summary statistics were meta-analyzed with a random-effects model using inverse variance as weights. NFOAS: Newfoundland Osteoarthritis Study; LEAP OA: Longitudinal Evaluation in the Arthritis Program, Osteoarthritis Study

Figure 2.

Forest plot for the significant metabolite and metabolite ratios associated with sustained pain. ORs and CIs were obtained by generalized estimating equation adjusted for age, sex and BMI for the NFOAS cohort and by logistic regression adjusted for age, sex and BMI for the LEAP OA cohort. Meta-analysis ORs and CIs were calculated from cohort summary statistics with random-effects meta-analysis modeling using inverse variance as weights. NFOAS: Newfoundland Osteoarthritis Study; LEAP OA: Longitudinal Evaluation in the Arthritis Program, Osteoarthritis Study; PC: phosphatidylcholine; SM: sphingomyelin; OR: odds ratio

Figure 3.

Forest plot for the significant metabolite ratios associated with other pain patterns. ORs and CIs were obtained by generalized estimating equation adjusted for age, sex and BMI for the NFOAS cohort and by logistic regression adjusted for age, sex and BMI for the LEAP OA cohort. Meta-analysis ORs and CIs were calculated from cohort summary statistics with random-effects meta-analysis modeling using inverse variance as weights. NFOAS: Newfoundland Osteoarthritis Study; LEAP OA: Longitudinal Evaluation in the Arthritis Program, Osteoarthritis Study; SM: sphingomyelin; OR: odds ratio

Patients with higher PC aa C28:1 concentration were significantly less likely to have sustained pain (P = 2.60 × 10⁻⁴), associated with a 34% reduction in the risk of sustained pain per s.d. increase in the metabolite concentration (Fig. 2). Two of the three metabolite ratios associated with sustained knee pain involved the same metabolites—PC aa C28:1 and PC aa C32:0. Patients with a higher ratio of PC aa C32:0 to PC aa C28:1 were significantly more likely to have sustained pain, with a 59% increase in risk per s.d. increase in the metabolite ratio (P = 1.21 × 10⁻⁵); its reciprocal ratio was associated with a 40% decrease in risk per s.d. (P = 1.36 × 10⁻⁵) (Fig. 2). In addition, patients with higher ratio of C14:2 to SM C20:2 were more likely to have sustained pain, with a 59% increase in risk per s.d. increase in the metabolite ratio (P = 1.77 × 10⁻⁵) (Fig. 2). This ratio was also associated with reporting pain while in bed and pain while sitting or lying, with a 61% increase and 48% increase in risk per s.d. increase, respectively (P = 3.63 × 10⁻⁶ and P = 1.99 × 10⁻⁵) (Fig. 3). Further, we found that patients with a higher ratio of acylcarnitine C16:2 to SM C20:2 were significantly more likely to report pain while in bed, with a 55% increase in risk per s.d. increase in the metabolite ratio (P = 1.97 × 10⁻⁵) (Fig. 3).

The identified metabolite/ratios were not associated or were weakly associated with age and BMI but the difference between males and females was significant (supplementary Table S4, available at Rheumatology online).

Discussion

In this study, we examined the prevalence of sustained knee pain using a novel definition in two large cohorts. We used meta-analysis to assess the association of sustained pain with metabolites and metabolite ratios as proxies for enzymatic reactions to offer insight into systemic metabolite relationships and metabolism inside the joint as surrogates for SF metabolite ratios [21]. Our results have potential uses as predictive pre-operative biomarkers of sustained pain and in understanding molecular mechanisms underlying sustained pain in primary knee OA.

We considered four definitions for sustained pain. Our primary and most conservative definition, which considered patients to have sustained pain if they reported any pain in each question on the WOMAC pain subscale, had lower misclassification error and revealed a greater number of metabolomic differences between pain groups. No significant associations were detected with our least strict definition, pain reported in a minimum of one question. The two other pain definitions considered pain at rest, which is related to neuropathic pain after TJA [22]. We observed similar effect sizes and directions for the significant metabolite and ratios in all four definitions. Pain during activity was assessed using the two activity-related questions (pain while walking on a flat surface; going up and down stairs). No significant associations were found (data not shown).

Reported rates of sustained knee pain vary. A previous meta-analysis showed an average of 20% of patients reporting long-term sustained pain post-TKA [23]. These studies classify pain using various patient-reported outcome measures (WOMAC, Visual Analogue Scale, and Oxford Knee Score, etc.). For comparable knee OA studies using the WOMAC, sustained pain definitions ranged from minimum one or two points in any one question [5, 24] to strict ‘non-responder to TJA’ definitions such as a change score less than seven points out of 20 [25]. Average follow-up times ranged from 6 to 41 months [5, 26]. Our definition acts as an intermediate with more stringent cut-off for sustained pain than one or two points but less conservative than the non-responder definitions which can exclude patients with sustained pain post-TJA that is less severe than their pre-surgical pain.

Rates of sustained knee pain varied between our two cohorts. LEAP OA cohort rates were significantly higher in all categories than the NFOAS cohort; 9.8% of NFOAS patients and 50.0% of LEAP OA patients experienced sustained pain under our primary definition. Rates in the NFOAS cohort were consistent with literature while the LEAP OA cohort saw higher rates than typically reported. Factors influencing the differences between the two cohorts and the literature include patient expectations [27], ethnicity, lifestyle and surgeons at/population served by recruiting hospitals; the NFOAS cohort was recruited at tertiary care centres while the LEAP OA cohort was recruited at a quaternary care centre. While different follow-up time in the two cohorts could be a factor, when considering only NFOAS patients with similar follow-up time to the LEAP OA cohort, rates were consistent with the full NFOAS cohort. Obesity could be another factor, but the NFOAS cohort had a significantly higher BMI than the LEAP OA cohort, with similar age and sex distribution. The rate might also be biased by another potentially affected yet to be replaced knee joint. However, a subset of the NFOAS cohort with both knees replaced showed that the concordance rate of the sustained knee pain status was 94% and the rate of the sustained pain in patients with both knees replaced was actually lower than those with one knee replaced. Though reasons for the rate difference between the two cohorts remain elusive, it could be considered a strength of the current study as effect size and direction of the significant metabolite and ratios found in our meta-analysis were nearly identical in the two individual cohorts, indicating a robust association validated in two different cohorts, strengthening the connection between the metabolite/ratios and sustained pain.

In total, we found that one metabolite and three metabolite ratios were significantly associated with sustained knee pain. It appears PC aa C28:1 is the key driver for the association as it was not only associated with sustained knee pain individually but also involved in two of the three significant metabolite ratios. PCs have been linked to pain and to OA in previous studies through the conversion of PCs to lysoPCs, releasing long-chain fatty acids such as arachidonic acid which act as precursors for pro-inflammatory, anti-inflammatory and pain mediators [15, 28–31]. Dysregulation of these lipid mediators, resulting in a persistent state of inflammation, has been proposed to be involved in the pathologies of OA and other chronic diseases, inflammatory pain syndromes and neuropathic pain [32–34]. Thus, chronic inflammatory pain, possibly leading to neuropathic pain and central sensitization, could be a contributor to sustained pain through PC metabolism.

The reciprocal PC aa C28:1 to PC aa C32:0 ratios provided further insights into the potential contributions of PC metabolism to sustained pain. Alterations in PC metabolism have previously been shown to be associated with OA [35]. An increased ratio of the saturated PC relative to the monounsaturated PC could result from increased conversion of monounsaturated fatty acid side chains to saturated or polyunsaturated side chains, alterations in production of downstream signalling mediators, or another mechanism which would preferentially remove unsaturated PCs from circulation over saturated PCs, such as lipid peroxidation [36]. Desaturation and lipid peroxidation are both associated with increased oxidative stress, which promotes neuro-inflammation, pain sensitization and chronic pain [37]. Pain sensitization has been highlighted as a possible mechanism underlying chronic pain following TJA in OA, especially in pain at rest [22]; clinical depression and pain at multiple sites are other suggested indicators of a pain sensitization phenotype in chronic pain patients [5]. Significantly higher rates of clinical depression and severe multisite musculoskeletal pain have been seen in non-responders to TJA in the NFOAS cohort [38]; when these factors were added to the current linear regression models for NFOAS patients with available data, P-values of significant metabolites did not substantially change, and clinical depression was also associated with sustained pain (P < 0.05; data not shown). The involvement of PC aa C28:1 as a significant metabolite and the presence of two significant reciprocal ratios involving this PC indicates a robust association which is not commonly seen in similar metabolomic studies [39], highlighting PCs as a metabolite group of interest for future investigations.

The link between SMs and acylcarnitines is less clear; while both compounds are involved in lipid and energy metabolism, no direct connection between them has been linked to pain. Increased acylcarnitines are often observed when β-oxidation is increased, such as in fasting states [40]. Accumulations of acylcarnitines are associated with oxidative stress and inefficient β-oxidation [41] which can lead to inflammatory and neuropathic pain and promote insulin resistance, type II diabetes mellitus and other conditions associated with chronic inflammation [42]. Reduced medium- and long-chain acylcarnitines have previously been associated with OA, OA severity and comorbidity with diabetes [43, 44]. Though seemingly elevated in this study, it is possible acylcarnitines seen in sustained pain patients are low when compared with non-OA controls or other subtypes of OA patients while still elevated when compared with SM C20:2. SM is a cell membrane constituent which has previously been associated with pain conditions and has a proposed role in inflammatory signalling as part of membrane lipid rafts [45, 46]. Sphingomyelinases (SMase), activated in part by pro-inflammatory cytokines, break SM into phosphocholine and ceramide, which is heavily involved in inflammation and apoptosis [47]. Decreased SM could indicate increased ceramide due to SMase activity, resulting in inflammation and subsequent pain. Altered SM metabolism and products of ceramide metabolism are suggested to influence development of neuropathic pain and sensitization [48] and could contribute to sustained pain. Interestingly, incomplete β-oxidation is linked to an accumulation of ceramides and an increase in proinflammatory molecules [41], offering a possible indirect link between the acylcarnitines and SM ratio and sustained pain.

There are several strengths and limitations of this study. The use of two cohorts and large sample size strengthen our findings, especially with near identical effect size and direction of the significant findings in both cohorts and the meta-analysis. We considered multiple definitions of sustained pain and identified a novel definition that reduced misclassification and showed increased metabolic differences between groups. While this novel definition was intended to identify and assess patients with pain after surgery, it does not take into account pre-surgical pain. Thus, our sustained pain group may be comprised of individuals whose pain has improved significantly alongside individuals who have improved minimally or not at all. We used a commercially available targeted metabolomics kit which quantified metabolites rather than assessing relative abundance; while the kit broadly targeted a large number of metabolites, coverage was limited when considering the scope of human metabolism and possibly missed other important metabolic markers for sustained pain in primary knee OA. In addition, method resolution was low and provided limited details on structures of PCs and SMs beyond number of carbons and double bonds; a higher resolution method is needed to determine individual side chain lengths and double bond locations. Our analysis was adjusted for age, sex and BMI, thus the results were independent of these factors. However, the levels of the identified metabolite and ratios were different between males and females, suggesting that different reference ranges are needed for males and females when using the identified metabolite/ratios to predict sustained pain risk. Lastly, the generalizability of the findings to other ethnic groups needs to be validated as the study participants included in the current study were predominately of European descent.

In conclusion, we identified one novel metabolite and three novel metabolite ratios to be associated with sustained pain in primary knee OA. These findings suggest potential roles for inflammation, oxidative stress, pain sensitization and altered lipid metabolism. Though further validation is required, they have potential utility in prediction and treatment of sustained pain and to provide additional insights into mechanisms underlying sustained pain in knee OA.

Supplementary data

Supplementary data are available at Rheumatology online.

Data availability statement

Research data are available by request subject to ethical approval.

Funding

The work was supported by the Canadian Institutes of Health Research (CIHR) (grant numbers 132178, 143058 and 153298), The Research and Development Corporation of Newfoundland and Labrador (grant number 5404.1423.102) and The Memorial University of Newfoundland Medical Research Fund. This work is also, in part, supported by Tony and Shari Fell Platinum Chair in Arthritis Research (M.K.), Canada Research Chairs (CRC) Program (Tier 1 CRC for M.K.) and the Schroeder Arthritis Institute via the University Health Network (UHN) Foundation, Toronto, Canada.

Disclosure statement: A.V.P. holds a role on the Arthritis Care & Research Editorial Board. Y.R.R. holds royalties or licenses for and has received consulting fees for Medtronic, has patents planned, issued, or pending for microRNA for OA, and holds stock options in Arthur Health Corporation. N.N.M. holds stock or stock options in Arthritis Innovation Corporation and Arthur Health Corporation. P.R. has received consulting fees and payment or honoraria for lectures, presentations, speakers bureaus, manuscript writing or educational events for AbbVie, Amgen, Celgene, Pfizer, Janssen, Novartis and Eli Lilly, and holds leadership or fiduciary roles in the Canadian Academy of Health Sciences and Canadian Council of Academies. M.K. holds grants from the Canadian Institutes of Health Research (CIHR) and Arthritis Society, has received payment or honoraria for lectures, presentations, speakers bureaus, manuscript writing or educational events for Shenzhen University, China, CARC, McGill University and Harbin International Osteoarthritis Summit, has received support for attending meetings and/or travel from the Osteoarthritis Research Society International (OARSI), and holds leadership or fiduciary roles in the OARSI Board of Directors and the Canadian Connective Tissue Society Board. All other authors (C.A.C., J.S.R., M.L., R.G., E.W.R., A.F., G.Z.) have no conflicts of interest to declare.