What Is the Evidence to Support the Association Between Metabolic Syndrome and Osteoarthritis? A Systematic Review

Abstract

Objective.

There is conflicting evidence on the association between metabolic syndrome (MetS) with the risk of osteoarthritis (OA). We aimed to systematically summarize the empirical evidence and discuss challenges in research methodologies in addressing this question.

Methods.

We performed a systematic literature review based on PubMed, Embase, Web of Science, and the Cochrane Database of Systematic Reviews on published epidemiologic studies that examined the association between MetS and the risk of OA. We included cross-sectional studies, case–control studies, and cohort studies with appropriate covariate adjustments. We extracted information on prevalence, incidence, crude and adjusted effect estimates, and the 95% confidence intervals from the articles, or this information was provided by the authors. We listed the main methodologic issues existing in current literature and provided recommendations for future research on this topic.

Results.

We identified 7 eligible studies on knee OA, 3 on hip OA, and 3 on hand OA. In studies that adjusted for body mass index or weight, MetS was not significantly associated with the risk of knee OA. No significant associations were reported for hip OA. For hand OA, the data were sparse and insufficient to reach a conclusion. Studies were mostly cross-sectional, exposure included only 1 time measurement, few studies had incident outcomes, and covariate adjustment was often insufficient.

Conclusion.

Our review was unable to reach a definitive conclusion due to insufficient data, although the data suggest that knee and hip OA are not associated with MetS. Future longitudinal studies with incident OA cases, repeated measurement of MetS, and appropriate covariate adjustment are needed.

INTRODUCTION

Worldwide prevalence of osteoarthritis (OA) is increasing due to aging and the obesity epidemic. Understanding the etiology of OA and subsequent systemic consequences of OA is important, because it confers a significant public health burden. Metabolic syndrome (MetS), a cluster of several cardiometabolic risk factors and a common accompaniment of obesity, is defined as central obesity, dyslipidemia, impaired fasting glucose, and hypertension. Substantial evidence suggests that MetS is associated with an increased risk of cardiovascular disease, type 2 diabetes mellitus, and cancer. Emerging evidence also links MetS with the risk of OA. Obese individuals are at high risk of developing OA, not only in the knee but also in non–weight-bearing joints such as the hand. The association between obesity and hand OA has suggested that the loading conferred by obesity on weight-bearing joints is not the sole explanation for the high risk of OA among obese individuals. Evidence has emerged that chronic inflammation, insulin resistance, and production of abnormal adipocytokines from adipose tissues (such as tumor necrosis factor, interleukin [IL]-1, IL-6, leptin, and adiponectin) may play a role in the etiology of OA. However, the potential mechanisms underlying this association are unclear.

Early reports on the link between MetS and OA appeared in 1990, and the number of articles focusing on this issue has increased dramatically over the past 15 years (see Supplementary Figure 1, available on the Arthritis Care & Research web site at http://onlinelibrary.wiley.com/doi/10.1002/acr.23698/abstract). Review articles published on this topic have focused mostly on the potential biologic mechanisms underlying this link. There has been little critical examination of the evidence, including quality of study design and statistical analyses. Whether a link between MetS and OA really exists is still under debate (1). There are at least 2 major study design concerns that are highly relevant to studies of MetS and OA. First, since the loading conferred by obesity is likely to be an independent cause of OA (especially of knee and hip OA), studies examining metabolic factors with OA need to include an adjustment for weight or body mass index (BMI). Second, most published studies are cross-sectional and therefore provide only limited evidence for causality. The strongest evidence for causal relations comes from high-quality longitudinal studies with incident OA as outcomes (2). Thus, the goal of our review was to provide a systematic summary of evidence, discuss challenges in epidemiologic study design and issues regarding the study of an association between MetS and OA, discuss the difficulty in analyses and evaluate strengths of the existing evidence, and suggest future directions.

MATERIALS AND METHODS

Data sources and searches.

We conducted a systematic literature review on epidemiologic studies using PubMed, Web of Science, the Cochrane Database of Systematic Reviews, and Embase. In addition to the articles found, we searched references of all identified articles. The search strategies for the PubMed search are shown in Supplementary Appendices 1–3, available on the Arthritis Care & Research web site at http://onlinelibrary.wiley.com/doi/10.1002/acr.23698/abstract. All searches were conducted for published literature up to May 16, 2018. We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement and the guidelines for performing a meta-analysis and systematic review of observational studies in epidemiology.

Study selection.

Both authors (SL and DTF) independently evaluated each study’s eligibility and study quality. Discrepancies were resolved by consensus. To be included, studies had to: 1) report data from an original, peer-reviewed study; 2) be of cross-sectional, case–control, or prospective cohort design using a noninstitutionalized adult population (age >18 years); 3) be a study of humans with and without OA; 4) characterize participants as to whether they had MetS or not; 5) define OA as clinical OA, knee or hip replacement due to OA, or symptomatic OA or radiographic OA, with the latter 2 including imaging evidence of OA; 6) report an association between the 2 conditions; and 7) have adjustment for confounding factors, including adjustment for BMI or weight. For studies published in languages other than English, we reviewed the English abstract and if the full article was needed, we asked a native speaker to translate the article into English. For multiple articles published from the same study, we reviewed all but presented details only from the most recent qualified article.

Data extraction and quality assessment.

Information from each selected study was extracted by both authors independently. We evaluated each study based on study design, study population, exposure and outcome definitions, confounding control, bias assessment, and statistical methods, as well as the study-defined effect estimates.

Data synthesis and analysis.

The odds ratio or hazard ratio was reported in eligible studies. Due to the limited numbers of longitudinal studies with adequate quality and the heterogeneity of these studies, there was not sufficient data for a meta-analysis. To be consistent with most eligible studies, we presented results on MetS as a binary variable (yes or no).

RESULTS

We found 506 studies on the topic of knee/hip/hand OA and MetS from PubMed, 702 studies from Web of Science, 0 from the Cochrane Database of Systematic Reviews, and 555 studies from Embase up to May 16, 2018. More than 1 article was published based on data from the Research on Osteoarthritis/Osteoporosis Against Disability (ROAD) cohort (3). We used the one with the most detailed analyses on numbers of MetS components that was a prospective cohort study with multivariable adjustment (3).

MetS with knee OA.

For knee OA, 7 studies met our inclusion criteria, of which 2 were cross-sectional (4,5), 1 was a case–control study (6), and 4 were cohort studies (7–10) (Table 1). Most of the reported effect estimates suggested a null association of MetS with knee OA after adjustment for BMI or weight (5,7,8,10). Despite meeting all of our selection criteria, the quality of the following studies and results needs to be interpreted with caution. The cross-sectional study performed by Shin using the fifth Korean National Health and Nutrition Examination Survey demonstrated a 1.49-fold increased risk for knee OA, and results became nonsignificant after further adjustment for body weight or BMI (4). This study did not account for sampling weights in the statistical analysis, even though the study, a nationwide survey, had a stratified, multistage probability sampling design. Contrary to other literature, the small Fasa Osteoarthritis Study showed that the odds ratio between MetS and OA paradoxically increased from 6.8 to 10.9 after adjusting for age, sex, and BMI (6).

Table 1.

Study characteristics of knee osteoarthritis studies on the association between metabolic syndrome and knee osteoarthritis (OA)^*

Type	Longitudinal	Longitudinal	Longitudinal	Longitudinal
Author, year (ref.)	Hellevik et al, 2018(7)	Niu et al, 2017 (8)	Monira Hussain et al, 2014 (9)	Engstrom et al, 2009 (10)
Study	Nord-Trondelag Health Study 2 linked to the Norwegian Arthroplasty Register	Framingham Heart Study Offspring cohort	Melbourne Collaborative cohort study	Malmo Diet and Cancer Study
Design	Prospective cohort	Prospective cohort	–	Prospective cohort
Country	Norway	US	Australia	Sweden
Baseline	1995–1997	1992–1995	2003–2007	1991–1994
No.	62,661	991 with no OA at baseline	20,430: 660 TKR, 19,208 no joint replacement	5,171
Mean age	49 years	54.2 years	68 years	57.5 years
Follow-up	15.4 years	10 years	6.8 years	12.4years
Exclusion	956 excluded due to previous joint replacement of hip or knee (n = 796), missing date of operation (n = 158), or emigration during baseline period (n = 2)	Excluded knees with prevalent OA at baseline	Missing anthropometric measurements	Excluded participants with history of OA
Exposure	Joint Interim Statement definition of MetS: presence of ≥3 of the following: waist circumference ≥102 cm for men and ≥88 cm for women, systolic blood pressure ≥130 mm Hg or diastolic blood pressure ≥85 mm Hg, or treatment of previously diagnosed hypertension, triglycerides ≥1.7 mmol/liter, HDL cholesterol men <1.1 mmol/liter, women <1.3 mmol/liter, and glucose ≥5.6 mmol/liter or self-reported type 2 diabetes mellitus	Assessment of MetS in 1990–1993, NCEP-ATP III criteria for MetS; ≥3 of the following: abdominal obesity (waist circumference, men ≥102 cm, women ≥88 cm), high triglyceride levels (≥150 mg/dl), low HDL cholesterol (men <40 mg/dl, women <50 mg/dl), high blood pressure (systolic ≥130 mm Hg or diastolic ≥85 mm Hg, or treatment for high blood pressure), high fasting glucose (≥110 mg/dl or diagnosis of diabetes mellitus); MetS was also defined using the modified ATP III criteria (high fasting glucose based on the cut point of ≥100 mg/dl)	International Federation of Diabetes definition of MetS: central obesity (waist circumference men ≥94 cm, women ≥80 cm) and any 2 of the following: raised serum triglyceride ≥1.7 mmol/liter, reduced serum HDL cholesterol (men <1.03 mmol/liter, women <1.29 mmol/liter, or specific treatment for lipid abnormalities), raised blood pressure (systolic ≥130 mm Hg or diastolic ≥85 mm Hg, or treatment of previously diagnosed hypertension), and impaired fasting glucose ≥5.6 mmol/liter or previously diagnosed type 2 diabetes mellitus	NCEP-ATP III criteria for MetS: presence of any 3 components: high waist circumference (men ≥102 cm, women ≥88 cm), low HDL(men <1.03 mmol/liter, women <1.29 mmol/liter, or lipid-lowering medication), hypertension (≥130/85 mm Hg or treatment for hypertension), hyperglycemia (fasting plasma glucose ≥5.6 mmol/liter), hypertriglyceridemia (≥1.7 mmol/liter or treatment)
Outcome	Incident of total knee replacement: 1,111 total	Incident radiographic OA (men: n = 35; women: n = 31)for K/L grade ≥2; incident symptomatic OA (men: n = 18, women: n = 32) present when a knee developed new combination of radiographic OA and knee pain	660 total knee replacements	89 knee OA, defined as a first knee arthroplasty or high tibia osteotomy, in combination with a diagnosis of OA
Covariates	Sex, smoking, physical activity, education, BMI	Age, sex, education, current smoking, physical activity, alcohol consumption, BMI (kg/m2)	Age, sex, country of birth, level of education, physical activity, BMI	Age, sex, BMI, smoking, CRP level, physical activity
Statistical analysis	Cox regression model stratified by age group (<50, 50–69.9, ≥70 years)	Sex-specific analysis; generalized estimating equation	Cox proportional hazards models	Cox proportional hazards model
Results	HR 0.89 (95% Cl 0.63–1.26) for age <50 years; HR 1.16 (95% Cl 0.96–1.41) for age 50–69.9 years; HR 1.27 (95% Cl 0.85–1.90) for age ≥70 years	Radiographic OA: ATP III criteria: men RR 1.4 (95% Cl 0.8–2.5), women RR 0.8 (95% Cl 0.5–1.4); modified ATP II criteria: men RR 1.2 (95% Cl 0.7–2.0), women RR 0.8 (95% Cl 0.5–1.3); symptomatic OA: ATP III criteria: men RR 0.8 (95% Cl 0.4–1.6), women RR 1.2 (95% Cl 0.7–2.1); modified ATP II criteria: men RR 1.3 (95% Cl 0.7–2.5), women RR 1.1 (95% Cl 0.6–1,9)†	HR 1.92 (95% Cl 1.59–2.32) before adjusting for BMI; HR 1.24(95% Cl 1.02–1.52) after adjusting for BMI	MetS (yes vs. no) for all: RR 2.1 (95% Cl 1.3–3.3) before adjusting for BMI, RR 1.1 (95% Cl 0.7–1.8) after adjusting for BMI; MetS (yes vs. no) men RR 1.4(95% Cl 0.6–1.3) before adjusting for BMI, RR 0.6 (95% Cl 0.2–1.5) after adjusting for BMI; MetS (yes vs. no) women RR 2.5 (95% Cl 1.5–4.4) before adjusting for BMI, RR 1.4 (95% Cl 0.8–2.6) after adjusting for BMI
Author, year (ref.)	Askarl et al, 2017 (6)	Vlsseret al, 2015(5)	Shin, 2014(4)
Study	Fasa Osteoarthritis Study	Netherlands Epidemiology of Obesity study	Fifth Korean National Health and Nutrition Examination Survey (2010)
Design	Case-control, matched for sex	Cross-sectional analyses of baseline measurements	Cross-sectional analyses
Country	Iran	The Netherlands	Korea
Baseline	2013	2008	2010
No.	131 OA patients, 262 matched controls	5,002 participants with a self-reported BMI ≥27 kg/m2 In the greater area of Leiden; 1,671 inhabitants from Lelderdorp were invited Irrespective of BMI	2,363
Mean age, years	OA group 52.9 years, controls 55.5 years	45–65 years	63.4years
Follow-up	NA	NA	NA
Exclusion	Controls who had radiographic complication caused by OA in the knee and hip were not included
Exposure	NCEP-ATP III definition of MetS: presence of ≥3 of the following: systolic blood pressure ≥130 mm Hg and/or diastolic blood pressure ≥85 mm Hg; triglyceride ≥150 mg/dl; high density lipoprotein cholesterol (men <40 mg/dl, women <50 mg/dl); fasting blood sugar ˃100; waist circumference women ˃88 cm, men ˃102 cm	NCEP-ATP III definition of MetS: presence of ≥3 of the following: elevated waist circumference (men ≥102 cm, women ≥88 cm), elevated triglycerides (≥1.7 mmol/liter or treatment for elevated triglycerides), reduced HDL cholesterol (men <1.03 mmol/liter, women <1.3 mmol/liter, or treatment for reduced HDL cholesterol), elevated blood pressure (systolic ≥130 mm Hg, diastolic ≥85 mm Hg or antihypertensive medication), elevated fasting glucose (≥5.6 mmol/liter or glucose lowering medication)	NCEP-ATP III definition of MetS: presence of ≥3 of the following: waist circumference men ≥90 cm, women ≥85 cm; triglyceride level ≥150 mg/dl or medication use; HDL cholesterol men <40 mg/dl, women <50 mg/dl or medication use; blood pressure of 130/85 mm Hg or greater or medication use; fasting glucose ≥100 mg/dl or medication use
Outcome	131 OA patients, 262 controls matched for sex; diagnosis of OA based on K/L score ˃1; controls were from referrals to orthopedic clinic of the hospital without symptoms of pain and/or stiffness In knee or hip	Clinical OA defined according to ACR clinical criteria; presence of a knee prosthesis was considered knee OA	Presence of radiographic knee OA defined as a K/L grade ≥2
Covarlates	Sex, age, BMI	Age, sex, height, smoking, education, ethnicity	Age, sex, Income, smoking, alcohol consumption, physical activity, BMI
Statistical analysis	Multiple logistic regression model	Logistic regression for cross-sectional analyses	Logistic regression
Results	MetS (yes vs. no): RR 6.8 (95% Cl 4.1–11.4) before adjusting for age, sex, and BMI; RR 10.9 (95% Cl 5.5–21.8) after adjusting for age, sex, and BMI	RR 1.56 (95% Cl 1.24–1.97) before adjusting for weight; RR 1.08 (95% Cl 0.85–1.39) after adjusting for weight	RR 1.49 (95% Cl 1.23–1.79) before adjusting for BMI; RR 0.92 (95% C 0.74–1.13) after adjusting for BMI; RR 1.49 (95% Cl 1.23–1.79) before adjusting for weight; RR 1.04 (95% Cl 0.84–1.27) after adjusting for weight

^*Studies were presented by the order of study design and publication year. ref. = reference; TKR = total knee replacement; MetS = metabolic syndrome; HDL = high-density lipoprotein; NCEP-ATP III = National Cholesterol Education Program Adult Treatment Panel III revised guideline; K/L = Kellgren/Lawrence; BMI = body mass index; CRP = C-reactive protein; HR = hazard ratio; 95% CI = 95% confidence interval; RR = risk ratio; NA = not applicable; ACR = American College of Rheumatology.

^†For both radiographic and symptomatic OA, results from Niu et al did not report RR before adjustment for BMI. We were therefore unable to present RRs with and without BMI adjustment.

MetS with hip OA.

For hip OA, 3 studies were eligible, all of which were cohort studies (7,9,10) (Table 2). All studies had large sample sizes and a long duration of follow-up. Results before and after BMI adjustment were consistently null. Overall, there was no association between MetS with hip OA.

Table 2.

Study characteristics of hip osteoarthritis (OA) studies on the association between metabolic syndrome and hip OA^*

Type	Longitudinal	Longitudinal	Longitudinal
Author, year (ref.)	Hellevlk et al, 2018(7)	Monlra Hussain et al, 2014 (9)	Engstrom et al, 2009 (10)
Study	Nord-TrØndelag Health Study 2 linked to the Norwegian Arthroplasty Register	Melbourne Collaborative cohort study	Malmo Diet and Cancer Study
Design	Prospective cohort	Prospective cohort	Prospective cohort
Country	Norway	Australia	Sweden
Baseline	1995–1997	2003–2007	1991–1994
No.	62,661	562 primary hip replacement, 19,208 with no joint replacement	5,171
Mean age	49 years	68 years	57.5 years
Follow-up	15.4 years	6.8 years	12 years
Exposure	Joint Interim Statement definition of MetS: presence of ≥3 of the following: waist circumference men ≥102 cm, women ≥88 cm; systolic blood pressure ≥130 mm Hg or diastolic blood pressure ≥85 mm Hg, or treatment of previously diagnosed hypertension; triglycerides ≥1.7 mmol/liter; HDL cholesterol men <1.1 mmol/liter, women <1.3 mmol/liter; glucose ≥5.6 mmol/liter or self-reported type 2 diabetes mellitus	Internatlonal Federation of Diabetes definition of MetS: central obesity (waist circumference men ≥94 cm, women ≥ 80 cm), and any 2 of the following: raised serum triglyceride ≥1.7 mmol/liter, reduced serum HDL cholesterol (men <1.03 mmol/liter, women <1.29 mmol/liter, or specific treatment for lipid abnormalities), raised blood pressure (systolic ≥130 mm Hg or diastolic ≥85 mm Hg, or treatment of previously diagnosed hypertension), and Impaired fasting plasma glucose ≥5.6 mmol/liter or previously diagnosed type 2 diabetes mellitus	NCEP-ATP III definition of MetS: presence of ≥3 components: waist circumference (men ≥102 cm, women ≥88 cm), low HDL (men <1.03 mmol/liter, women <1.29 mmol/liter, or lipid-lowering medication), hypertension (≥130/85 mm Hg or treatment for hypertension), hyperglycemia (fasting plasma glucose ≥5.6 mmol/liter) and hypertriglyceridemia (≥1.7 mmol/liter or treatment)
Outcome	Incldent total hip replacement	Total hip replacement (n = 562)	Arthroplasty due to severe hip OA (n = 120)
Covarlates	Sex, smoking, physical activity, education, BMI	Age, sex, country of birth, level of education, physical activity, BMI	Age, sex, smoking, physical activity, CRP, BMI
Statistical analysis	Cox regression model stratified by age group (<50,50–69.9, ≥70 years)	Cox proportional hazards model	Cox proportional hazards model
Results	HR 0.58 (95% Cl 0.40–0.83) for age <50 years; HR 0.93 (95% Cl 0.79–1.10) for age 50–69.9 years; HR 0.83 (95% Cl 0.65–1.14) for age ≥70 years	MetS not significantly associated with hip replacement before and after adjusting for BMI; HR 1.19(95% Cl 0.95–1.49) before adjusting for BMI; HR 1.00 (95% Cl 0.78 1.27) after adjusting for BMI	For both men and women, HR 1.00 (95% Cl 0.60–1.50) before adjusting for BMI, HR 0.7 (95% Cl 0.4–1.2) after adjusting for BMI; for men only, HR 0.9 (95% Cl 0.4–1.8) before adjusting for BMI, HR 0.7 (95% Cl 0.3–1.6) after adjusting for BMI; for women only, HR 1.0 (95% Cl 0.6–1.7) before adjusting for BMI, HR 0.7 (95% Cl 0.4–1.3) after adjusting for BMI

^*Studies were presented by the order of study design and publication year. See Table 1 for abbreviations.

MetS with hand OA.

For hand OA, the only longitudinal study showed a null association (11). However, this study was of small sample size and studied only whites. We did not have sufficient data to reach a definitive conclusion in this review (5,11,12) (Table 3). The 2016 study by Tomi et al focused on patients with HIV and may not be generalizable to the general population (12). Although not included in our review, the Netherlands Epidemiology of Obesity study provided important evidence on adiposity, particularly visceral fat, associated with a 1.3-fold elevated risk for hand OA in men (13).

Table 3.

Study characteristics of hand osteoarthritis studies on the association between metabolic syndrome and hand osteoarthritis (OA)^*

Type	Longitudinal	Cross-sectional	Cross-sectional
Author, year (ref.)	Strand et al, 2018(11)	Tomi et al, 2016(12)	Visser et al, 2015(5)
Study	Framingham OA Study	Metabolic Syndrome and Fibrosls- Osteoarthritls study (METAFIB-OA)	Netherlands Epidemiology of Obesity (NEO) study
Design	Longitudinal analyses	Cross-sectional single-center study on patients with HIV Infection	Cross-sectional analyses of baseline measurements
Country	US	France	The Netherlands
Baseline	1992–1995	2011	2008
No.	586	458 with HIV infection, ages 45–64years, from January 2011 to December 2012 from outpatient clinics in France	5,002 participants with self-reported BMI ≥27 kg/m2 In the greater area of Leiden; 1,671 Inhabitants from Lelderdorp were invited irrespective oftheir BMI
Mean age	50–75 years	45–64 years	45–65 years
Follow-up	2 years	NA	NA
Exclusion	Excluded hand OA at baseline	–	–
Exposure	American Heart Association/National Heart, Lung, Blood Institute clinical/laboratory MetS definition: presence of ≥3 of the following: central obesity (waist circumference men ≥102 cm, women ≥88 cm), hypertension (systolic blood pressure ≥130 mm Hg, diastolic blood pressure ≥85 mm Hg, and/or antihypertensive treatment), diabetes mellltus (fasting blood glucose ≥100 mg/dl and/or anti-diabetes treatment), elevated triglycerides (≥150 mg/dl), and low HDL (men <40 mg/dl, women <50 mg/dl, and/or cholesterol-lowering treatment); we created an alternative definition based on clinical and laboratory measurements only (I.e., excluding treatment-related information), since Individuals with MetS who are adequately treated theoretically may have a lower risk of OA	Internatlonal Diabetes Federation criteria: central obesity (waist circumference men ≥94 cm, women ≥80 cm [In Europe] or with ethnicity-specific values for other groups); plus any 2 of the following: triglycerides ≥1.7 mmol/liter, or specific treatment for this lipid abnormality; HDL cholesterol men <1 mmol/liter, women <1.3 mmol/liter, or specific treatment for this lipid abnormality; increased blood pressure with systolic ≥130 or diastolic ≥85 mm Hg, or treatment of previously diagnosed hypertension; increased fasting plasma glucose ≥100 mg/dl (5.6 mmol/liter), or previously diagnosed type 2 diabetes mellitus	ATP III criteria for MetS: presence of ≥3 of the following: elevated waist circumference (men ≥102 cm, women ≥88 cm), elevated triglycerides (≥1.7 mmol/liter ortreatment for elevated triglycerides), reduced HDL cholesterol (men <1.03 mmol/liter men, women <1.3 mmol/liter, or treatment for reduced HDL cholesterol), elevated blood pressure (systolic ≥130 mm Hg, diastolic ≥85 mm Hg or antlhyperten- sive medication), elevated fasting glucose (≥5.6 mmol/liter or glucose lowering medication)
Outcome	Incldent radiographic hand OA (n = 56); ≥2 distal or proximal Interphalangeal joints with K/L grade ≥2	Any patient with ≥1 finger joint scored K/L grade ≥2 was considered to have radio- graphic hand OA	Clinical OA defined according to ACR clinical criteria; bony and soft swellings as well as deformities of distal Interphalangeal, proximal interphalangeal, metacarpophalangeal, carpometacarpal, and wrist joints were assessed
Covarlates	Age, sex, BMI	Age, sex, previous hand trauma, HIV duration, CD4 level, HIV viral load, duration of exposure to protease inhibitors	Age, sex, height, smoking, education, ethnicity; analyses on surrogates for systemic process were adjusted for weight
Statistical analysis	Logistic regression	Logistic regression	Weighted analyses; logistic regression for cross sectional analyses
Results	MetS definition Including treatment RR 0.98 (95% Cl 0.66–1.46) before adjusting for BMI; RR 0.91 (95% Cl 0.58–1.44) after adjusting for BMI; MetS clinical/laboratory definition: RR 1.00 (95% Cl 0.67–1.50) before adjusting for BMI; RR0.92 (95% Cl 0.58–1.49) after adjusting for BMI	RR 2.18 (95% Cl 1.26–3.96) for hand OA	RR 1.62 (95% Cl 1.23–2.13) before adjusting for weight; RR 1.46 (95% Cl 1.06–2.02) after adjusting for weight

^*Studies were presented by the order of study design and publication year. See Table 1 for abbreviations.

DISCUSSION

In our current review, most evidence pointed to a null association of MetS with knee and hip OA. For hand OA, the data were limited and conflicting and were not sufficient to allow us to reach a definitive conclusion. Our systematic review showed that the strongest evidence came from a few longitudinal studies (7–10). More rigorous longitudinal evidence is needed.

Overall, we found methodologic deficiencies in most studies examining MetS and OA. Few studies used longitudinal data with sufficient sample sizes to assess associations between MetS and OA (7–10), and even fewer studies excluded prevalent OA cases, or prevalent joint replacement at baseline (7,8,10). The adequacy of control for confounding varied considerably across studies. The definition of MetS and its relevant exposure window were not clear, with existing studies having only 1 time measurement of MetS in adulthood. Since obesity, sedentary lifestyle, and MetS may increase as a consequence of knee or hip OA, cross-sectional studies examining MetS and OA in these joints are limited in their ability to make causal inferences. Evidence was mainly from the US, Europe, and Asia, and future studies from black, Hispanic, and other minority populations are needed.

Of cross-sectional and longitudinal studies of MetS and knee or hip OA, several showed associations unadjusted for BMI or weight (4,5,9,10). In unadjusted analyses from all of these studies, there was a significantly increased risk of OA, and in all, this association diminished greatly and became nonsignificant in all but 1 when analyses were adjusted for BMI (4,5,9,10). In the Fasa Osteoarthritis Study, the odds ratio paradoxically increased from 6.8 to 10.9 after adjustment for age, sex, and BMI (6). Despite consistent findings from the literature that age is a strong risk factor for OA, this study showed a paradoxically reduced risk of OA with advanced age (6).

Inferring causality from observational studies is challenging and is based on multiple assumptions. Hill criteria include strength, consistency, specificity, temporality, biologic gradient, plausibility, coherence, experimental evidence, and analogy (2). Among all components, temporality is the most important consideration. Evidence from cross-sectional studies contributes less when compared with longitudinal studies, with exposure preceding outcome and confounder control (14). The evidence is further strengthened if there is a longitudinal study with control for baseline confounders (14). When there is a potential feedback between the exposure and outcome (for example, hip and knee OA leading to obesity) over time, cross-sectional studies are subject to reverse causation and cannot be used for drawing inferences. When reverse causation is highly unlikely, cross-sectional studies provide some evidence. If study investigators have a clear rationale on the temporal ordering of the exposure and outcome, and the confounding variables are likely to temporally precede the exposure and outcome, then cross-sectional evidence can be helpful (14).

Knee and hip OA can lead to changes in lifestyle, such as reduced physical activity level and weight gain, that may increase the subsequent risk for MetS. Whether there is a bidirectional relationship between MetS and OA is currently unclear. For studies on MetS as a risk factor for OA, differentiating incident from prevalent knee OA is therefore important.

Prevalent hand OA cases may be less subject to reverse causation, because the hand is a non–weight-bearing joint, and hand OA might not be likely to lead to dramatic lifestyle changes when compared with knee or hip OA. However, an absence of evidence does not prove a null association, and more data on the longitudinal association of MetS with hand OA are needed. Currently, only 1 study investigated this association longitudinally (11) (Table 2).

The current evidence focused on incident OA as defined by a Kellgren/Lawrence grade or total joint replacement, which is already late in the pathogenesis of joint pathology. For studies using joint replacement as outcomes, excluding prevalent joint replacement at baseline is important. Hellevik et al (7) excluded prevalent joint replacement at baseline and reported effect estimates for incident joint replacement. Future studies examining earlier stages of disease (possibly with magnetic resonance imaging) and joint pain are needed.

Except for the Melbourne Collaborative cohort study (9) (562 hip replacement, 660 knee replacement) and the Nord-Trøndelag Health Study (7) (1,840 hip replacement, 1,111 knee replacement), current evidence was based on a limited number of incident OA events. In general, studies may not have been large enough to give a definite answer and had low power to detect potential effect modification. Combining data from studies to leverage existing cohort resources and examine potential effect modification, such as by sex and race, may be needed.

Currently there is no consistent definition of MetS. In our review, we did not exclude studies based on the definition of MetS, and we listed all definitions of MetS that were used in the original studies. The heterogeneity in MetS definitions has contributed to published studies with different definitions and criteria for MetS. Consistency of the research findings across different MetS definitions would increase the robustness and generalizability of the results.

Current literature mainly focused on MetS as a one-time measurement, thus studying a fixed prevalent exposure. There is a need for studies examining changes in MetS with risk of OA. Questions regarding how MetS changes and how MetS in childhood or early life is related to OA risk in later life is unknown. Further, according to the developmental origins of chronic disease theory, obesity, cardiovascular disease, and type 2 diabetes mellitus develop during intrauterine exposure and fetal programming. Whether this programming is also true for OA etiology is currently unclear. Future study on the intergenerational effects of MetS on OA risk using a life course approach may be of interest.

Whether there is a potential effect modification by sex on the association between MetS and OA is currently unclear. Only 2 studies have tested for this potential effect modification. In the Framingham OA study, Niu et al (8) found no sex-specific findings. Engstrom et al (10) also found MetS and OA risk to be similar by sex. Due to the limited numbers of studies presenting sex-specific results, we cannot have a definite answer regarding effect modification by sex.

Current evidence is subject to uncontrolled and unmeasured confounding. Socioeconomic factors, dietary information, and lifestyle factors contribute to the etiology of both MetS and OA. However, existing studies have not fully considered nor were able to account for influences of these factors. Each study would have been strengthened if the authors had presented additional analyses on these as potential confounders of the association between MetS and OA and had presented information on how observed results would be changed by these potential confounders. Bias analysis methods have been developed to evaluate the robustness of evidence from observational studies (15). None of the published studies included in our search performed sensitivity or bias analyses to assess the influence of unmeasured confounding. We encourage future studies to perform sensitivity analyses and demonstrate the robustness of study findings.

The Korean National Health and Nutrition Examination Study showed a nonsignificant association between MetS and OA after further adjustment for weight or BMI. Despite a stratified, multistage probability sampling design, this study used a logistic regression model without adjusting for sampling weights, which can bias results. It would be helpful for future studies to provide results with sampling weights adjustment.

While this systematic review focused on overall MetS, many of the articles included also gave data on individual components of MetS and their relation with OA. As with MetS, there is a burgeoning literature on the potential relation of each of these with OA. Further, for each, there are unique biologic reasons to suspect a relationship. Our review is insufficient to grapple comprehensively with the evidence linking such elements of MetS as hypertension, diabetes mellitus, lipid abnormalities, and visceral adiposity with OA. Our suggestions about study design are highly relevant to the study of these issues as well.

The number of reviews on MetS and OA has grown exponentially in recent years without a critical look at the quality of the evidence. Our review focused on published studies and thus may be subject to publication bias. Further, our review focused on MetS, and future studies are needed to look into each individual component of MetS.

In conclusion, there was insufficient data from large high-quality studies on the association of MetS with OA, especially hand OA. However, the preponderance of high-quality evidence suggests that there is no association of MetS with either knee or hip OA. Future evidence from large high-quality longitudinal studies is needed. We suggest either larger longitudinal studies or a pooling of studies to permit a further examination of this association, as well as a focus on examining earlier stages of disease. If every future epidemiologic study could address the methodologic considerations mentioned above, the potential for prevention of OA may be substantial, and the results could have important public health and policy implications.

SIGNIFICANCE & INNOVATIONS.

• Articles published on the association between metabolic syndrome (MetS) and osteoarthritis (OA) have focused mostly on potential biologic mechanisms. There has been little critical examination of the evidence, including quality of study design and statistical analyses. Whether a link between MetS and OA really exists is still under debate.

• For knee and hip OA, after adjustment for body mass index or weight, most studies showed a null association.

• We found that most existing evidence on the association of MetS and OA risk was of limited quality. Either large high-quality longitudinal studies in the future or a pooling of studies will permit a further examination of the association of MetS, especially with hand OA. Clarifying the relationship could offer opportunities for prevention of OA and have important public health and policy implications.