Association between osteoarthritis and dyslipidaemia: a systematic literature review and meta-analysis

Abstract

Objectives

We aimed to investigate the prevalence of dyslipidemia in patients with osteoarthritis (OA) and whether OA and dyslipidemia are associated.

Methods

We performed a systematic literature review and a meta-analysis, including cross-sectional, cohort and case–control studies, to assess the number of patients with OA and/or dyslipidemia. We calculated the mean (±SD) prevalence of dyslipidemia in patients with and without OA and the risk of dyslipidemia (OR, 95% CI) among patients with OA.

Results

From 605 articles screened, 48 were included in the analysis (describing 29 cross-sectional, 10 cohort and 9 case–control studies). The mean prevalence of dyslipidemia was 30.2%±0.6% among 14 843 patients with OA and 8.0%±0.1% among 196 168 without OA. The risk of dyslipidemia was greater with than without OA overall (OR 1.98,95% CI 1.43 to 2.75, p<0.0001) and with knee OA (OR 2.27, 1.33 to 3.89, p=0.003) and hand OA (OR 2.12, 1.46 to 3.07), p<0.0001).

Conclusion

The risk of dyslipidemia was twofold greater with than without OA, so lipid disturbances could be a risk factor for OA. Such a result supports the individualisation of the metabolic syndrome-associated OA phenotype.

Key messages.

What is already known about this subject?

• Metabolic disturbances such as obesity or diabetes mellitus are associated with osteoarthritis (OA), but data about the link between OA and lipid disturbances remain conflicting.

What does this study add?

• This is the first systematic review and meta-analysis demonstrating an association between OA and dyslipidemia. This result reinforces the concept of the metabolic syndrome-associated OA phenotype.

How might this impact on clinical practice?

• This study emphasises the need to screen and manage cardiovascular comorbidities, especially lipid disturbances in patients with OA in clinical daily practice.

Introduction

Osteoarthritis (OA) is the most common joint disease and a major cause of pain and disability. It is currently considered a disease with multiple distinguishable phenotypes: post-traumatic, ageing-related, genetic and metabolic syndrome (MetS)-associated OA.¹ Metabolic OA, the most commonly studied phenotype, is defined by the association between OA and MetS, associating obesity, hyperglycaemia with insulin resistance, dyslipidemia and hypertension.² Metabolic OA mainly affects middle-aged people (45–65 years) and leads to knee, hand and generalised OA. The association between OA and MetS has been reported in several epidemiological studies.^{3 4} The pathophysiological link between both diseases could be chronic low-grade systemic inflammation occurring in both conditions.⁵

The association of OA with each MetS component has been investigated.⁶ Obesity and overweight are independently linked to hand OA, with a twofold increased risk.⁷ This association suggests the release of inflammatory mediators by adipose tissue adipokines. We recently reported an association between OA and diabetes mellitus, with a 1.46-fold increased risk of OA with diabetes mellitus and a 1.41-fold increased risk of diabetes mellitus with OA.⁸ The link between both pathologies could be explained by the action of pro-inflammatory cytokines and oxidative stress occurring in both diseases.^9–12

The link between OA and the other components of MetS remains debated. Experimental studies have suggested that lipid disturbances could be involved in OA pathophysiology,¹³ but epidemiological studies revealed heterogeneous results.

With a systematic literature review and meta-analysis, we aimed to investigate the prevalence of dyslipidemia in patients with OA and assess whether OA and dyslipidemia are associated.

Methods

The systematic review was registered on PROSPERO (CRD: 42016037290).

Literature search

We performed a systematic search of articles in MEDLINE via PubMed, EMBASE and the Cochrane library. The keywords used for the PubMed search were (((‘Dyslipidemias’[Mesh] OR ‘Hypertriglyceridemia’[Mesh]) OR ‘Hypercholesterolemia’[Mesh]) OR ‘HDL’[All Fields] OR ‘LDL’[All Fields] OR ‘Triglycerides’[All Fields] OR ‘Hyperlipidemias’[Mesh]) OR ‘Cholesterol’[Mesh] OR ‘Metabolic Syndrome X’[Mesh] AND ‘Osteoarthritis’[Mesh] AND (‘humans’[MeSH Terms] AND (English[lang] OR French[lang])). No time limit was set for publication date, and articles published up to 1 January 2016 were searched. We also searched the abstracts from international meetings of the American College of Rheumatology (ACR), European League Against Rheumatism, Société Française de Rhumatologie, European Society of Cardiology, Endocrine Society’s Annual Meeting and European Congress of Endocrinology.

Study selection

We selected articles published in English or French that described observational studies of adults (>18 years of age) with cohort, case–control and cross-sectional designs. Studies were included if they specified the number of patients with OA and dyslipidemia and/or the prevalence or incidence of OA in patients with dyslipidemia and/or dyslipidemia in patients with OA, and/or the mean values of parameters of dyslipidemia in patients with and without OA and/or the existence or not of an association between OA and dyslipidemia. We excluded non-observational studies (therapeutic trials, reviews, letters and case reports). Articles that did not mention the number of patients with OA or dyslipidemia and those that did not evaluate the link between the two diseases were excluded. The selection of articles was based on titles and abstracts, then full texts.

Data synthesis

We extracted the following data: publication data (title of the article, first author, journal and publication date), study design (type of study, year(s) of inclusion, study quality score), population (total number of patients included, mean age and sex of patients), methodology of articles (the definition used for OA and dyslipidemia, OA location) and data needed for statistical analysis (number of patients with OA and/or dyslipidemic patients; mean total cholesterol (TC), low-density lipoprotein (LDL), high-density lipoprotein (HDL) and triglyceride (TG) levels (mg/dL or mmol/L); and number of patients receiving statins, number with MetS and number with obesity or mean body mass index (BMI) in kg/cm²). The quality of the study was estimated by using the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) scale, the score expressed in percentage of positive answers in relation to the number of items selected.¹⁴

Statistical analysis

First, we performed a descriptive analysis of the prevalence of dyslipidemia in patients with and without OA and used the number of patients with dyslipidemia and total number with and without OA. To estimate this prevalence from cohort longitudinal prospective studies, we used baseline data. Prevalence was expressed as mean±SD Second, we calculated the mean TC, LDL, HDL and TG levels in patients with and without OA. Third, for studies examining an association between OA and dyslipidemia, we calculated the risk of dyslipidemia with OA by estimating the overall OR with 95% CIs. The data were extracted from studies examining the number of dyslipidemic patients with and without OA. We used Revman V.5.3 for the meta-analysis with a fixed-effects model. Heterogeneity was assessed by the I² index; with I²>50% (high heterogeneity), we used a random-effects model, and with I² <50% (low heterogeneity), we used a fixed-effects model. With strong heterogeneity, we used a randomised-effects analysis. To investigate potential publication bias, we have performed the funnel plot. The association was considered positive with OR >1, and the result was considered statistically significant with p≤0.05. We performed sensitivity and subgroup analyses.

Results

Characteristics of studies included

The selection of articles is reported in the flow chart (figure 1). We identified 605 publications; 48 articles (including 13 abstracts) from 43 studies were included (2 articles from the SEKOIA study, 4 from the FRAMINGHAM study and 2 from the National Health and Nutrition Examination Survey III). One abstract¹⁵ was obtained from the EMBASE database and not from screening congress abstracts. The 48 articles described 29 cross-sectional, 10 cohort and 9 case–control studies. Among them, 29 articles involved the OA population and 19 the general population (table 1). We did not find any studies based on a cohort of patients with dyslipidemia, which explains why the prevalence or relative risk of OA in patients with dyslipidemia was not calculated. Table 2 shows the definitions of OA and dyslipidemia in selected studies.

Figure 1.

Flow chart of articles in the study

Table 1.

Description of the 48 articles studies selected for analysis

Osteoarthritis population			General population
Type of study	Author	Year	Author	Year
Cross-selectional	Stürmer et al25	1998	Davis et al26	1988
	Racaza et al65	2012	Han et al27	2013
	Erb et al66	2004	Dahaghin et al41	2007
	Eymard et al28	2015	Haugen et al42	2015
	Shea et al29	2015	Inoue et al30	2011
	Salamon et al50	2015	Cemeroglu et al22	2014
	Abourazzak et al20	2015	Meek et al51	2014
	Juge et al17*	2015	Al-Arfaj31	2003
	Rollefstad et al23*	2014	Suri et al48	2010
	Saunders et al53*	2013	Puenpatom et al4	2009
	Nuñez et al32*	2012	Hart et al19	1995
	Shukurova et al67*	2014	Maddah et al24	2015
	Salaru et al33*	2013	Engström et al34	2009
	Kemta Lekpa et al35*	2014	Yoshimura et al3	2012
	Niu et al36*	2015	Nielen et al54	2012
	Haugen et al43*	2013	Marshall et al44	2015
	Courties et al45*	2014	Hussain et al37	2014
Cohort	Gandhi et al49	2014	Sowers et al21	2009
	Laires et al38*	2015	Massengale et al46	2012
	Thelier–Deloison et al15*	2012
Case–control	Soran et al16	2008
	Cheras et al18	1997
	Mishra et al39	2012
	Oliviero et al52	2012
	Addimanda et al47	2012
	Philbin et al55	1996
	Irshad et al56	2014
	Zayed et al40	2013
	Cheng et al57*	2013

*Data from a congress.

Table 2.

Characteristics of the 48 included articles: definitions of osteoarthritis (OA) and dyslipidemia, outcomes and Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) study quality

Author	OA definition	Dyslipidemia definition	Outcome	STROBE study quality (%)
Stürmer et al25	Arthroplasty or KL≥2	TC≥240 mg/dL and/or statin therapy	MV in OA+ Association of OA and dyslipidemia	53
Racaza et al65	ACR or Cq and Rx	–	NPD in OA+	42
Erb et al66	Cq and Rx	–	MV in OA+	50
Eymard et al28	ACR Cq and Rx KL scale	History of dyslipidemia	NPD in OA+	82
Shea et al29	Cq and Rx	–	NPS in OA+ MV in OA+	78
Salamon et al50	ACR	–	NPD in OA+ MV in OA+	72
Abourazzak et al20	KL≥2	HDL<50 mg/dL TG≥150 mg/dL	NPD in OA+	66
Juge et al17*	Rx	–	NPD in OA+	NA
Rollefstad et al23*	History of OA	–	MV in OA+ and OA– Association of OA and dyslipidemia	NA
Saunders et al53*	KL scale	TC>4 mmol/L	NPD in OA+ Association of MV and KL scale	NA
Nuñez et al32*	–	Hypercholesterolemia (ND)	NPD in OA+	NA
Shukurova et al67*	-	Hypercholesterolemia (ND)	NPD in OA+	NA
Salaru et al33*	ACR	–	NPD in OA+	NA
Kemta Lekpa et al35*	ACR	–	NPD in OA+	NA
Niu et al36*	Arthroplasty or KL≥2	HDL<40 mg/dL in M;<50 mg/dL in W TG>150 mg/dL	Association of OA and dyslipidemia	NA
Haugen et al43*	KL≥2	Low HDL and HTG (ND)	NPD in OA+ Association of OA and dyslipidemia	NA
Courties et al45*	KL≥2	–	NPD in OA+	NA
Gandhi et al49	Cq and Rx	HDL<35 mg/dL in M,<40 mg/dL in W; TG≥150 mg/dL	NPD in OA+	52
Laires et al38*	–	–	NPD in OA+	NA
Thelier–Deloison et al15*	History of OA	–	NPD in OA+ and OA– Association of OA and dyslipidemia	NA
Soran et al16	Cq and Rx	–	MV in OA+ and OA– Association of OA and dyslipidemia	65
Cheras et al18	Cq and Rx	–	MV in OA+ and OA– Association of OA and dyslipidemia	75
Mishra et al39	KL scale ACR	–	MV in OA+ and OA– Association of OA and dyslipidemia	58
Oliviero et al52	ACR	–	MV in OA+ and OA– Association OA and dyslipidemia	67
Addimanda et al47	Cq KL scale	LDL≥130 mg/dL and/or CT≥240 mg/dL and/or statin therapy	NPD in OA+ and OA– Association of OA and dyslipidemia	75
Philbin et al55	Questionnaire Radiological Danielson scale	LDL≥160 mg/dL and/or HDL≤35 mg/dL	NPD in OA+ and OA– NPS in OA+ and OA– Association of OA and dyslipidemia MV in OA+ and OA–	73
Irshad et al56	KL scale	TC≥200 mg/dL and/or TG≥150 mg/dL	NPD in OA+ and OA– Association of OA and dyslipidemia MV in OA+ and OA–	47
Zayed et al40	ACR	–	MV in OA+ and OA– Association of OA and dyslipidemia	56
Cheng et al57*	–	–	Association of OA and dyslipidemia	NA
Davis et al26	Rx	–	MV in OA+ and OA– Association of OA and dyslipidemia	67
Han et al27	History of OA by physician	HDL<40 mg/dL in M,<50 mg/dL in W; TG≥150 mg/dL	MV in OA+ and OA– Association of OA and dyslipidemia	84
Dahaghin et al41	KL≥2, ACR, Cq	–	MV in OA+ and OA– Association of OA and dyslipidemia	69
Haugen et al42	KL≥2	–	NPS in OA+ MV in OA+	84
Inoue et al30	KL≥2	HDL<40 mg/dL in M,<50 mg/dL in W; TG≥150 mg/dL	NPD in OA+ and OA– Association of OA and dyslipidemia MV in OA+ and OA–	69
Cemeroglu et al22	≥3 articulations with KL≥2	TC>200 mg/dL LDL>100 mg/dL HDL<40 mg/dL TG>150 mg/dL	NPD in OA+ and OA– MV in OA+ and OA– NPS in OA+ and OA– Association of OA and dyslipidemia	59
Meek et al51	Codes	–	MV in OA+ NPS in OA+	78
Al-Arfaj31	KL≥2	TC≥220 mg/dL	NPD in OA+ and OA– Association of OA and dyslipidemia	50
Suri et al48	Pathria and Weishaupt scale	TC≥240 mg/dL	NPD in OA+ and OA– Association of OA and dyslipidemia	72
Puenpatom et al4	Codes Rx History of OA by physician	Codes or HDL<40 mg/dL in M,<50 mg/dL in W; or TG≥150 mg/dL	NPD in OA+ and OA– Association of OA and dyslipidemia	69
Hart et al19	KL≥2	–	Association of OA and dyslipidemia	78
Maddah et al24	KL≥2	TC≥5 mmol/L and TG≥2 mmol/L and HDL≤1 mmol/L in M, ≤1.1 mmol/L in W	NPD in OA+ and OA– Association of OA and dyslipidemia MV in OA+ and OA–	72
Engström et al34	Codes: arthroplasty for hip or knee OA	HDL<1.03 mmol/L in M,<1.29 mmol/L in W; TG≥1.7 mmol/L or statin therapy	NPD in OA+ and OA– Association of OA and dyslipidemia	79
Yoshimura et al3	KL≥2	HDL≤40 mg/dL	MV in OA+ and OA– Association of OA and dyslipidemia	91
Nielen et al54	Codes	Codes: hypercholesterolemia	NPD in OA+ and OA– NPS in OA+ and OA–	81
Marshall et al44	KL scale	Codes	NPD in OA+ NPS in OA+	74
Hussain et al37	Joint replacement	HDL<1.03 mmol/L in M,<1.29 mmol/L in W; HTG≥1.7 mmol/L	NPD in OA+ and OA– Association of OA and dyslipidemia	85
Sowers et al21	KL≥2	HDL≤45 mg/dL or LDL>160 mg/dL or TG>200 mg/dL	MV in OA+ and OA– Association of OA and dyslipidemia	70
Massengale et al46	–	TC≥240 mg/dL	NPD in OA+ and OA–	78

ACR, American College of Rheumatology; Cq, clinical; HDL, high-density lipoprotein; HTG, hypertriglyceridemia; KL, Kellgren and Lawrence; LDL, low-density lipoprotein; M, men; MV, mean values of lipid profile; NA, if the data were issued only from congress; ND, not defined; NPD, number of patients with dyslipidemia; NPS, number of patients with statin therapy; OA+, patients with osteoarthritis; OA–, patients without osteoarthritis; Rx, radiography; TC, total cholesterol; TG, triglycerides; W, women.

*Data from a congress.

The median STROBE quality score was 69.1% (range 42%–91%). Nine articles had a STROBE quality score <60% (table 3).

Table 3.

Characteristics of the population on the 48 included articles: number, age, gender, overweight proportion

Author	Sample size (N = number of total patients; n= number of patients with OA)	Mean age (years) in OA+ and OA– patients	Gender in OA+ and OA– patients (% of F)	Overweight proportion (%) or BMI (kg/m2) in OA+ and OA–
Stürmer et al25	n=809	–	OA+: F: 62.3%	–
Racaza et al65	n=859	OA+: 62.9	OA+: F: 74.5%	–
Erb et al66	N=250 n=64	OA+: 57.3±10.1	OA+: F: 62.5%	OA+: 30.9±7.6 kg/m2
Eymard et al28	n=559	OA+: 62.8	OA+: F: 70.1%
Shea et al29	n=791	OA+: 74.25±4.5	OA+: F: 62.3%	OA+: 27.28 kg/m2
Salamon et al50	N=927 n=344	–	OA+: F: 83.4%	OA+: 29.5 kg/m2
Abourazzak et al20	n=130	OA+: 56.7±8.1	OA+: F: 100%	OA+: 32.54±2.9 kg/m2
Juge et al17 *	n=147	OA+ : 75.8±10	OA+: F: 68.7%	OA+: 27.2 kg/m2
Rollefstad et al23 *	N=626 n=469	OA+: 64.1±8.6 OA–: 63.3±9.3	OA+: F: 73.1% OA–: F: 58%	–
Saunders et al53 *	n=57	–	–	–
Nuñez et al32 *	n=260	OA+: 69.8±8	OA+: F: 79.2%	–
Shukurova et al67 *	n=1243	OA+: 56.1±7.9	–	OA+: 61.6% of OP
Salaru et al33 *	n=61	OA+: 64.9±2.7	OA+: F: 77%	OA+: 60.6% of OP
Kemta Lekpa et al35 *	n=148	OA+: 57±10.6	OA+: F: 75%	OA+: 53% of OP 30.8±5.6 kg/m2
Niu et al36 *	n=1091	OA+: 62	OA+: F: 55.5%	–
Haugen et al43 *	n=748	OA+: 58.1	–	OA+65.7% of OP
Courties et al45 *	n=869	OA+: 54±7	OA+: F : 72%	–
Gandhi et al49	n=1502	OA+: 55.3±15.5	OA+: F: 48.8%	OA+: 27.3 kg/m2
Laires et al38 *	n=197	OA+: 67±8.6	OA+: F: 79.2%	–
Thelier–Deloison et al15 *	n=112 n=26	–	–	OA+: 100% of OP
Soran et al16	N=66 n=36	OA+: 40.9±2.5 OA–: 39±4.7	OA+: F: 72.2% OA–: F : 66.7%	OA+: 29.9±3.3 kg/m2 OA–: 27.6±3.8 kg/m2
Cheras et al18	N=96 n=44	OA+: 69±9 OA–: 68±7	OA+: F 40.9% OA–: F 40.4%	OA+: 25.8 kg/m2 OA–: 24.8 kg/m2
Mishra et al39	N=100 n=28	OA+: 49.1±1.4 OA–: 49.6±1.3	OA+: M: F: 71.4% OA–: M: F: 69.4%	OA+: 23.4±0.6 kg/m2 OA–: 22.9±0.6 kg/m2
Oliviero et al52	N=77 n=16	OA+: 54.7±11.5 OA–: –	OA+: F: 68.7% OA–: –	–
Addimanda et al47	N=753 n=446	OA+: 68±8 OA–: 63.9±9	OA+: F: 92.8% OA– : F : 97.4%	OA+: 25.1±3.8 kg/m2 OA–: 24.9±3.9 kg/m2
Philbin et al55	N=69 n=46	OA+: 65.8±9.3 OA–: 67.9±6.7	OA+: F: 56.5% OA–: F: 65.2%	OA+: 31.2±5.9 kg/m2 OA–: 24.6±3.2 kg/m2
Irshad et al56	N=100 n=50	–	–	–
Zayed et al40	N=80 n=40	OA+: 43.5±3.7 OA–: 44.4±3.9	OA+: F: 87.5% OA–: F: 87.5%	OA+: 37.3±5.9 kg/m2 OA–: 23.5±1.3 kg/m2
Cheng et al57 *	N=56 607 n=23 530	–	–	–
Davis et al26	N=3885 n=301	–	–	–
Han et al27	N=10 839 n=270	OA+: 64.5±10.1 OA–: 53.2±11	OA+: F: 84.8% OA–: F: 50%	–
Dahaghin et al41	n=3585	–	–	OA+26.3±3.5 kg/m2
Haugen et al42	N=1348 n=726	–	–	–
Inoue et al30	N=795 n=251	OA+: 66.3 OA–: 55.5	OA+: F: 79.3% OA–: F : 54.7%	OA+: 23.8 kg/m2 OA–: 22.8 kg/m2
Cemeroglu et al22	N=61 n=39	–	OA+: F: 100% OA–: F: 100%	–
Meek et al51	N=858 n=206	OA+: 59.2±11	OA+: F: 79.1%	–
Al-Arfaj31	N=246 n=122	–	–	–
Suri et al48	N=441 n=310	OA+: 57.8±10.6 OA–: 46.7±9.7	OA+: F: 49% OA–: F: 39%	–
Puenpatom et al4	N=7714 n=975	OA+: 69.6 OA–: 41.3	OA+: F: 61.3% OA–: F: 51.3%	OA+: 66.9% of OP OA–: 34.8% of OP
Hart et al19	N=979 n=118	–	OA+: F: 100%
Maddah et al24	N=625 n=244	OA+: 61.2 OA–: 48.0	OA+: F: 89.8% OA–: F: 73.8%
Engström et al34	N=5194 n=209	OA+: 59.9 OA–: 57.6	OA+: F: 66.5% OA–: F: 58.4%	OA+: 27.9 kg/m2 OA–: 25.37 kg/m2
Yoshimura et al3	N=1690 n=71	OA+: 67.3±8.2 OA–: 58.2±11.8	OA+: F: 74.6% OA–: F: 58.6%	OA+: 23.6±2.9 kg/m2 OA–: 22.4±3.2 kg/m2
Nielen et al54	N=175 956 n=4040	OA+: 69.8 OA–: 51	OA+: F: 68.7% OA–: F: 50.4%	–
Marshall et al44	N=1076 n=341	OA+: 69.0	OA+: F: 80.4%	–
Hussain et al37	N=20 430 n=1222	OA+: 68.3±7.7 OA–: 64.8±8.6	OA+ : F: 66.2 OA–: F: 59.5%	OA+: 76.8% of OP, 28.6±5.0 kg/m2 OA–: 62.6% of OP, 26.8±4.5 kg/m2
Sowers et al21	N=664 n=53	OA+: 50±5 OA–: 47±8	OA+: F: 100% OA–: F: 100%	OA+: 35.6±11.1 kg/m2 OA–: 27.3±8.4 kg/m2
Massengale et al46	N=2477 n=466	–	OA+: F: 58.2% OA–: F: 46.6%	–

*Data from a congress. BMI, body mass index; F, female; M, male; OA, osteoarthritis.

In total, 30 articles assessed the association of OA and dyslipidemia, 30 assessed the prevalence of dyslipidemia among patients with OA and 22 assessed mean lipid level values among patients with OA (table 3).

Patient characteristics

This study involved 306 044 patients. The mean age range was 39.0±4.7¹⁶ to 77.5±9.0 years.¹⁷ The mean proportion of females was 53.2% (range 40.6%¹⁸ to 100%^19–22). The localisation was the knee in 23 articles,3 15 16 19–21 24–40 hand in 9,15 22 41–47 generalised OA in 3,^{25 31 47} hip in 3,^{25 34 37} spine in 2^{48 49} and shoulder in 1.¹⁷ MetS was reported in nine articles,4 20 24 28 30 36 40 43 50 the prevalence of MetS ranged from 5%²⁴ to 97.5%.⁴⁰ The prevalence of obesity ranged from 7.8%⁵¹ to 100%^{15 40} and BMI from 22.3±2.7³⁰ to 37.3±5.9 g/cm².⁴⁰ Seven articles described the use of statin treatment (table 3).

Prevalence of dyslipidemia among patients with and without OA (table 4)

Table 4.

Main results of prevalence of dyslipidemia and mean lipid-level values in patients with osteoarthritis (OA) and non-OA patients

	Prevalence of dyslipidemia	Mean CT level (mg/dL)	Mean high-density lipoprotein level (mg/dL)	Mean low-density lipoprotein level (mg/dL)	Mean triglyceride level (mg/dL)
OA+ population	30.2%±0.7% n=14 823 n=28	245±25.1 n=6037 n=14	54.4±8.9 n=5856 n=18	126.5±20.7 n=656 n=9	137.3±80.3 n=2406 n=15
OA– population	8.0%±0.1% n=196 168 n=13	233.1±17.5 n=3763 n=3	53.1±7.5 n=412 n=7	136.9±15.9 n=451 n=2	131±27.3 n=3460 n=6

The mean prevalence of dyslipidemia was 30.2%±0.6% among 14 843 patients with OA and 8.0%±0.1% among 1 96 168 without OA. The mean prevalence with knee OA was 27.6%±1.4%,15 20 24 25 28 30–35 37 38 hand OA 37.6%±1.6%,^{22 43–47} generalised OA 30.5%±3.9%,^{25 31 47} hip OA 20%±2.1%^{25 34 37} and symptomatic OA was 21%.^{28 44}

Mean lipid-level values with and without OA (table 4)

The mean lipid-level values for patients with and without OA were for TC, 245±25.1 and 233.1±17.5 mg/dL; LDL, 126.5±20.7 and 136.9±15.9 mg/dL; HDL, 54.4±8.9 and 53.1±7.5 mg/dL; and TG, 137.3±80.3 and 131±27.3 mg/dL.

Association between dyslipidemia and OA

Overall, 30 articles indicated the presence or the absence of an association between OA and dyslipidemia; 21 (70%) showed a positive association between OA and dyslipidemia3 4 15 18 19 21 23 24 25 30 31 39 40 47 48 52–57; 12/18 articles (67%) with STROBE score >60% found a positive association.3 4 18 19 21 24 30 47 48 52 54 55 In addition, 4/7 articles19 25 31 47 that reported an OR adjusted on age and BMI found a positive association. Among the three with negative association findings after adjustment, two had a STROBE score >60%.^{34 37}

Overall risk of dyslipidemia with OA: meta-analysis

Among 204 148 patients from 13 articles,4 15 22 24 30 31 34 37 47 48 54–56 the overall OR was 1.98 (95% CI 1.43 to 2.75, p<0.0001; I²=94%), evaluated by a random-effects model (figure 2).

Figure 2.

Forest plot for dyslipidemia among patients with and without osteoarthritis (OA).

Risk of dyslipidemia with OA: sensitivity analyses

To strengthen our results, we performed four sensitivity analyses. First, we removed the studies that did not use ACR criteria or Kellgren-Lawrence grading for OA diagnosis: among 2568 patients from the six remaining articles,22 24 30 31 47 56 the risk of dyslipidemia was increased with than without OA (OR 2.64, 95% CI 2.14 to 3.26, p<0.00001, I²=0%). Second, we excluded studies with a STROBE score <60%: among 203 629 patients from the nine remaining articles,4 24 30 34 37 47 48 54 55 the risk of dyslipidemia remained increased with than without OA (OR 1.63, 1.13 to 2.36, p=0.009, I²=95%). Third, we excluded studies that specified the use of statin treatment because the definition of dyslipidemia in these studies was based on only lipid values and did not account for statin treatment. Among 41 539 patients from the 10 remaining articles,4 15 24 30 31 34 37 47 48 56 the risk of dyslipidemia remained increased with than without OA (overall OR 1.93, 1.42 to 2.61, p<0.0001, I²=87%). Fourth, we pooled the results of the articles that reported an age-adjusted and BMI-adjusted OR. Among 31 764 patients, from the four articles,31 34 37 47 there was no association between dyslipidemia and OA (OR 1.31, 95% CI 0.88 to 1.95, p<0.0001, I²=83%).

Risk of dyslipidemia with OA: subgroup analyses

We performed a subgroup analysis by OA localisation. The increased risk of dyslipidemia with OA persisted with knee OA (among 26 805 patients, OR 2.27, 1.33 to 3.89, p=0003, I²=88%)15 24 30 31 34 37 and hand OA (among 814 patients, OR 2.12, 1.46 to 3.07, p<0.0001, I²=0%)^{22 47} but not hip OA (among 24 934 patients, OR 0.86, 0.69 to 1.08, p=0.18, I²=0%).^{34 37}

Discussion

We investigated the potential association between OA and dyslipidemia with a systematic review and meta-analysis and found a 30% prevalence of dyslipidemia with OA, which seems much higher than in the non-OA population (8.0%). Furthermore, the meta-analysis revealed an increased risk of dyslipidemia, by 1.98, with than without OA and was observed with knee as well as hand OA.

The mean prevalence of dyslipidemia in hand OA was 37.6%±1.6%, much higher than the mean prevalence of 30.2%±0.6% with OA overall. Moreover, the risk of dyslipidemia was increased twofold with hand OA (OR 2.12, 95% CI 1.46 to 3.07). These results again confirm the systemic metabolic component of hand OA, as recently reported in the NEO study.⁵⁸ The pathophysiological link between hand OA and MetS might be explained by the action of the adipose-tissue source of proinflammatory cytokines and the action of visceral fat.⁵⁸

Hip OA, defined by joint replacement, was not associated with dyslipidemia possibly because of a selection bias of patients: cardiovascular comorbidities often associated with dyslipidemia might have restricted the indication for surgery due to the perioperative period. Furthermore, mechanical stress is more involved than metabolic stress in this joint.

For knee OA, the mean prevalence of dyslipidemia was 27.6%±1.4% and the association between knee OA and dyslipidemia was confirmed with increased risk of dyslipidemia (OR 2.27, 95% CI 1.33 to 3.89). The association between knee OA and MetS is sometimes conflicting. Han et al,²⁷ Inoue et al,³⁰ and Hussain et al³⁷ did not find any positive association possibly because of different OA definitions. A recent study showed that the most important risk factor of knee OA was mechanical stress (before and after adjustment for metabolic factors), which limits the identification of a systemic metabolic component in knee OA.

Our meta-analysis has some limitations. The heterogeneity between studies was high, probably because of differences in OA localisations, definition of OA and dyslipidemia, statin therapy could not have been taken into account, and types and quality of studies. Dyslipidemia referred to lipid abnormalities such as hypercholesterolemia, low HDL level, high LDL level or hypertriglyceridemia. Because of the different definitions of dyslipidemia, we chose to define dyslipidemia first by high LDL level, then low HDL level, then hypercholesterolemia and hypertriglyceridemia. To counteract this heterogeneity, we performed sensitivity analyses to check whether the association between OA and dyslipidemia persisted after removing studies with poor methodology and found that the association persisted in all sensitivity analyses. Moreover, the heterogeneity of the studies was assessed by the I² index and we adapted the method to its value. The results of the meta-analysis are not modified by removing the most heterogeneous studies (data not shown). We were not able to integrate confounding factors such as age, BMI, HTA, smoking and physical activity in the overall statistical analysis. Obesity is a major risk factor of development and progression of OA. Obesity increases the risk of OA of the weightbearing joints due to excessive mechanical stress but is also associated with dyslipidemia in MetS.⁵⁹ We identified seven articles accounting for confounding factors of dyslipidemia and OA: four showed a positive association after adjustment on age and BMI. However, when we meta-analysed the seven articles that reported an age-adjusted and BMI-adjusted OR, there was no association between dyslipidemia and OA, but raw data before adjustment on age and BMI are used. Finally, the impact of statin treatment could not be assessed because of the lack of data concerning its prescription. In fact, we have no details about statin use in dyslipidemic and non-dyslipidemic patients. However, Riddle et al did not find beneficial effect of statins on the structural progress at patients monitored for a knee osteoarthritis.⁶⁰

In this funnel plot, the distribution of common values is not heterogeneous. Likewise, we can consider that there is no major publication bias in our meta-analysis.

We demonstrated an association between dyslipidemia and OA, but the pathophysiological explanation for the causal relationship has not been clearly defined. Experimental studies suggest the existence of lipid metabolism dysfunction in OA. Mice with altered HDL metabolism showed knee OA despite abnormal weight gain.⁶¹ Gierman et al showed that dietary cholesterol intake increased spontaneous cartilage damage in mice.⁶² High LDL levels promote synovial inflammation and ectopic bone formation in mouse OA models.⁶³ Oxidised-LDL (oxLDL) could be involved in the development and progression of OA by stimulating synovial cells (macrophages, synovial fibroblasts and endothelial cells) and chondrocytes. A treatment strategy that lowers the level of oxLDL could be interesting.⁶⁴

In conclusion, this is the first systematic review and meta-analysis demonstrating an association between OA and dyslipidemia, which illustrates the role of metabolic disturbances beyond glucose metabolism in OA pathophysiology. Such a study emphasises the need to screen and manage cardiovascular comorbidities in patients with OA in clinical practice.