Racial-Ethnic Differences in Osteoarthritis Pain and Disability: A Meta-Analysis

Abstract

Osteoarthritis (OA), a leading cause of disability and pain, affects 32.5 million Americans, producing tremendous economic burden. While some findings suggest that racial/ethnic minorities experience increased OA pain severity, other studies have shown conflicting results. This meta-analysis examined differences in clinical pain severity between African Americans and Non-Hispanic Whites with osteoarthritis. Articles were initially identified October 1–5, 2016, and updated May 30, 2018 using PubMed®, Web of Science™, PsycINFO, and the Cochrane Library. Eligibility included English-language peer-reviewed articles comparing clinical pain severity in adult Black/African American and Non-Hispanic White/Caucasian patients with osteoarthritis. Non-duplicate article abstracts (N=1,194) were screened by four reviewers, 224 articles underwent full text review, and 61 articles reported effect sizes of pain severity stratified by race. Forest plots of standard mean difference showed higher pain severity in African-Americans for studies using the Western Ontario and McMasters Universities Osteoarthritis Index (WOMAC) (0.57; 95%CI: 0.54, 0.61) and non-WOMAC studies (0.35; 95%CI: 0.23, 0.47). African-Americans also showed higher self-reported disability (0.38; 95%CI: 0.22, 0.54) and poorer performance testing (−0.58; 95%CI: −0.72,−0.44). Clinical pain severity and disability in OA is higher amongst African Americans and future studies should explore the reasons for these differences to improve pain management.

Perspective

This meta-analysis shows that differences exist in clinical pain severity, functional limitations and poor performance between African Americans and Non-Hispanic Whites with osteoarthritis. This research may lead to better understanding of racial/ethnic differences in OA-related pain.

Introduction

Osteoarthritis (OA) affects 32.5 million individuals and is among the leading causes of pain and disability in the United States, with annual direct (i.e., medical expenditure) and indirect (i.e., loss earnings) costs estimated at $16,000 per patient and an annual societal economic impact of $486 billion.¹³⁴ OA is a significant cause of musculoskeletal pain, stiffness, swelling, and restricted range of motion ³⁷, and adversely impacts mobility ¹⁵, activities of daily living ¹²⁷, quality of life ², and psychological functioning.⁵⁴ Although any joint can develop OA, the knee, hand, and hip joints are typically affected, with the knee being the most commonly affected joint.³⁶ In addition, older women aged ≥55 have a higher prevalence and severity of OA.¹³²

There is considerable evidence that pain severity and disability vary considerably across people and often correlate poorly with the severity of joint changes observed radiographically.^16,90 The discordance between radiographic findings and OA-related pain and disability suggests that factors other than joint pathology influence OA-related pain and disability.⁷⁰ While the mechanisms underlying OA-related pain and disability remain poorly understood, OA progression can be influenced by person-level risk factors such as age, sex, obesity, genetics, race/ethnicity, and socioeconomic status, in addition to joint-level risk factors such as injury, occupational, malalignment, and abnormal loading of the joints.^{6,77,92,109,111}

One risk factor that appears to contribute importantly to OA-related pain and disability is racial and ethnic background.^{3,5,52,79,108} For instance, a greater proportion of racial/ethnic minorities, specifically African Americans (AAs), have both radiographic and symptomatic knee OA, compared to their non-Hispanic White (NHW) counterparts.^39,79,108 Moreover, several studies have reported greater pain severity ^{5,34,52,56,113} and greater pain-related disability ^{22,49,113,129} among AA, compared to NHW individuals with OA. Furthermore, older AA women are disproportionately affected by functional limitations and disability from OA in comparison to other NHW women with OA ^7,22,46 Longitudinal data from the 1998–2004 Health and Retirement Study, revealed that AA women with arthritis but no disability at baseline experienced disability at almost twice the rate of their NHWs counterparts over the ensuing 6-year period.¹²⁶ However, other studies have found conflicting results, with some findings showing NHWs having higher incidence of hand or hip OA ^38,103 and pain severity ^122,139 compared to AAs, while other studies found equivalent pain severity among AA and NHW individuals with OA.^11,20 In light of the conflicting evidence in the literature, it is important to better characterize racial/ethnic differences in OA-related pain.

The objectives of the current meta-analytic review were to systematically examine differences in clinical pain severity between AAs and NHWs with OA and to quantify the magnitude of these effects. A secondary aim was to examine race group differences in OA-related disability in these same studies.

Methods

Information Sources

Articles were initially identified October 1–5, 2016 using MEDLINE® (PubMed®), Web of Science™ Core Collection, PsycINFO (EBSCOhost Web), and the Cochrane Library. The search process was again repeated November 9, 2017 and May 30, 2018 to identify additional articles published since the initial literature search. Backward and forward references of key articles were investigated in Web of Science for additional studies. Major concepts of chronic pain, racial disparities, and osteoarthritis were searched in combinations of MeSH terms, PsycINFO descriptors, and truncated or phrase searched keywords in the title and abstract when permitted. The detailed search strategies employed for each data source are included in the Appendix.

Articles from data sources were selected for screening if they involved osteoarthritis, chronic pain, older adults and examined racial or ethnic differences. Candidate articles were randomly divided amongst four reviewers and screened for eligibility.

Eligibility Criteria

Inclusion:

Eligibility included all English-language peer-reviewed articles from clinical trials and observational studies on clinical pain severity and race. Patient populations were adults ≥ 18 years with osteoarthritis that included both Black/African-Americans and Non-Hispanic White/Caucasians patients. The use of the WOMAC (Western Ontario McMaster Osteoarthritis Index) or other standard measure such as the Visual Analog Scale (VAS), Numeric Rating Scale (NRS) or Arthritis Impact Measurement Scales 2 (AIMS2) was required for clinical pain severity.^17,99,138 Articles that reported bivariate effect sizes of clinical pain severity stratified by race were included in final qualitative synthesis. Quantitative analysis used forest plots and consisted of unique studies reporting standard mean, standard deviation and sample size.

Exclusion:

Study populations with only one racial/ethnic group were not included, nor were studies comparing other racial groups (e.g. Whites vs. Asians). Only publications reporting primary data were included; thus review articles, meta-analyses, and grey literature were excluded from the review.

Data Extraction & Measures

The primary outcome of this study was standard mean difference (SMD) of clinical pain severity between Black/African-Americans and White/Caucasians. The SMD and 95% confidence interval was calculated using Cochrane Collaboration Review Manager (RevMan v 5.3).¹¹⁸ The unadjusted mean pain score, standard deviation, and sample size were extracted for each article included and used to generate forest plots. Articles that had missing data were kept for qualitative synthesis, along with articles that had odds ratio, proportions or other type of effect size instead of mean pain score. The type of assessment used to evaluate pain severity was categorized as Western Ontario McMaster Osteoarthritis Index (WOMAC), Numeric Rating Scale (NRS), Visual Analog Scale (VAS), or Other.^17,99,138

The following study characteristics were extracted during article review: name of first author, journal, title and year of publication, study design, type of osteoarthritis, and source of data or study population. To investigate additional characteristics, each article was assessed for whether pain was treated as a primary outcome or a secondary outcome, and race was treated as a primary predictor variable or a control variable. We also assessed if the study reported on disability and socioeconomic status like income, education or employment.

Data Synthesis

Studies using WOMAC to assess pain were assumed to have a fixed distribution of effect sizes. When calculating non-WOMAC studies, a random distribution was assumed since the pooled effect size reflected multiple types of pain scales. For most pain measures, higher scores indicate worse outcome or most severe pain. To permit consistent interpretation, self-reported pain scales for which higher scores reflect better outcomes were recalculated by subtracting the mean score from the maximum possible value for that scale (e.g. SF-36v2). Self-reported disability scales were treated in a similar manner and recalculated as needed. However, for performance testing measures (e.g. Short Physical Performance Battery) lower scores are indicative of worse outcome or level of functioning, and the original values were retained. Any article using a single composite scale for both pain and disability was only attributed to the pain analyses. If an article reported using a pain and/or disability subscale, then it was reported in the pain and disability forest plots, respectively.

To assess evidence of bias, RevMan 5.3 was used to generate funnel plots of the standard error by SMD. Funnel plots for pain and disability measures are in the Appendix.

Results

Study Selection

A total of 1,194 non-duplicate articles were screened and 224 articles underwent full text review (Figure 1). Of these, 61 articles reported effect sizes of pain severity stratified by race, 42 articles reported mean pain scores, and 28 unique studies were identified for quantitative analysis. No unique, eligible articles were discovered from reference lists.

Figure 1:

PRISMA Diagram

Study Characteristics

The most common clinical pain measure (N=31) was the WOMAC (Western Ontario and McMaster Universities Osteoarthritis Index); both the pain subscale and full scale were used.¹⁷ Several articles (N=40) used other pain scales to assess pain including VAS/NRS (N=12), Arthritis Impact Measurement Scales (AIMS2; N=4), Lequesne Index of Severity for Osteoarthritis (N=2), Philadelphia Geriatric Center (PGC) Pain Scale (N=2), and Medical Outcomes Study 36-item Short Form (SF-36v2) Bodily Pain subscale (N=2).^91,96,99,138 In some articles (N=9), a single pain question was utilized. Examples include: “On most days do you have pain, aching, or stiffness in your [study joint]?”; “Have you had any joint pain in your [study joint] in the last month lasting most of the month?”; “Have you had any pain in your [study joint] on most days for at least 6 weeks?”; “Do you have any chronic, recurrent, or long-lasting pain, more than aches and pains that are fleeting and minor?”. For a complete list of clinical pain measures and study characteristics, refer to Table 1.

Table 1 :

Summary of Included Articles

Author	Journal	Year	OA Site	Dataset/ Population	Race	Sample Size	Parameter	Disab	SES	Pain Scale	Forrest Plot?	Results
Albert1	J Cross Cult Gerontol	2008	Hip, Knee	Allegheny County, PA	Primary predictor	B:284, W:267	Mean Pain Score	No	No	VAS	Yes	B>W
Allen7	Osteoarthritis Cartilage	2009	Hip ,knee	JoCo OA Project	Primary predictor	B:442, W:926	Mean Pain Score, Mean Difference(β)	Yes	Yes	WOMAC	Yes	B>W*
Allen8	Osteoarthritis Cartilage	2010	Hip, Knee	SeMOA, Durham VAMC	Primary predictor	B:221, W:270	Mean Pain Score, Mean Difference(β)	Yes	Yes	AIMS2	Yes	B>W*
Allen5	J Rheumatol	2012	Knee	JoCo OA Project	Primary predictor	B:891, W:1857	Mean Difference(β)	No	Yes	WOMAC	No	B>W*
Ang12	Med Care	2002	Hip, Knee	Cleveland VAMC	Primary predictor	B:262, W:334	Mean Pain Score	Yes	Yes	WOMAC, Lequesne	Yes†	B=W
Ang10	J Rheumatol	2003	Hip, Knee	Cleveland VAMC	Primary predictor	B:262, W:334	Mean Pain Score, Mean Difference(β)	Yes	Yes	WOMAC, Lequesne	Yes	B=W
Ang13	J Rheumatol	2009	Hip, Knee	Indianapolis VAMC, Wishard Hospital	Primary predictor	B:260, W:425	Proportion by pain status	Yes	Yes	WOMAC	No	B=W
Blanco20	J Am Geriatr Soc	2012	Unspec	EAS	Primary predictor	B:56, W:157	Median Pain Score	Yes	Yes	TPI	No	B=W
Burns22	J Natl Med Assoc	2007	Knee	FAST	Primary predictor	B:131, W:387	Mean Pain Score	Yes	No	Knee Pain Scale	Yes	B=W
Cano26	J Pain	2006	Unspec	Detroit Metropolitian, MI	Primary predictor	B:58, W:69	Mean Pain Score	Yes	Yes	MPI	Yes	B>W*
Colbert31	Artdritis Care Res (Hoboken)	2013	Knee	OAI	Primary predictor	B:595, W:3100	Mean Pain Score	Yes	Yes	WOMAC	Yes	B>W*
Collins32	Osteoartdritis Cartilage	2014	Knee	OAI	Control	B:413, W:1290	Prevalence by pain status (mild, mod, severe)	Yes	Yes	WOMAC	No	B>W
Conroy33	Artdritis Care Res (Hoboken)	2012	Knee	Healtd ABC Study	Control	B:381, W:477	Proportion w/ radiographic OA pain	No	No	WOMAC	No	B<W
Cruz-Almeida35	Artdritis Rheumatol	2014	Knee	UPLOAD Study	Primary predictor	B:147, W:120	Mean Pain Score, Mean Difference(β)	Yes	Yes	WOMAC,V AS, GCPS	Yes	B>W*
Dillon38	Am J Phys Med Rehabil	2007	Hand	NHANES III	Control	B:486, W:1379	Proportion w/ Symp OA	Yes	Yes	ACR/ NHANES criteria	No	B<W
Dominick42	J Natl Med Assoc	2004	Unspec	Durham VAMC	Primary predictor	B:61, W:141	Mean Pain Score	Yes	Yes	WOMAC	Yes†	B>W*
Eberly45	PLos One	2018	Knee	UNM Ortdopaedics Clinic	Primary predictor	B:10, W:191	Mean Pain Score	No	No	NRS	Yes	B>W
Flowers48	Arthritis Res Tder	2018	Knee, Hip Spine	JoCo OA Project	Primary predictor	B:735, W:1410	Proportion w/ Symptoms	No	No	Single pain question	No	B>W
Gandhi51	J Rheumatol	2008	Hip, Knee	Toronto Western Hospital	Primary predictor	B:20, W:1488	Mean Pain Score	Yes	No	WOMAC	Yes	B>W
Glover52	Arthritis Rheum	2012	Knee	UPLOAD Study	Primary predictor	B:45, W:49	Mean Pain Score	Yes	No	WOMAC	Yes†	B>W*
Glover53	Clin J Pain	2015	Knee	UPLOAD Study	Control	B:141, W:115	Mean Pain Score	Yes	No	WOMAC	Yes†	B>W*
Golightly56	Aging Clin Exp Res	2005	Hip, Knee, Back, Foot, Ankle	Durham VAMC	Primary predictor	B:61, W:141	Mean Pain Score, Mean Difference(β)	Yes	Yes	WOMAC	Yes	B>W*
Golightly55	Int J Behav Med	2015	Hip, Knee, Hand	JoCo OA Project, Durham VAMC	Control	B:59, W:94	Mean Pain Score	No	No	VAS	Yes	B>W*
Goodin58	Health Psychol	2013	Knee	UPLOAD Study	Primary predictor	B:67, W:63	Mean Pain Score	No	Yes	WOMAC	Yes†	B>W*
Goodin57	Psychosom Med	2014	Knee	UPLOAD Study	Primary predictor	B:122, W:103	Mean Pain Score	No	Yes	VAS	Yes	B>W*
Groeneveld60	Arthritis Rheum	2008	Hip, Knee	Philadelphia VAMC, Pittsburg	Primary predictor	B:450, W:459	Mean Pain Score	Yes	Yes	WOMAC, VAS	Yes	B>W*
Han63	Am J Phys Med Rehabil	2018	Knee	OAI	Control	B:237, W:749	Proportion by pain severity	Yes	Yes	KOOS Pain	No	B>W*
Hausmann66	Race Soc Probl	2013	Hip, Knee	Philadelphia VAMC, Pittsburgh VAMC	Primary predictor	B:127, W:303	Proportion by pain severity quartile	No	Yes	WOMAC	No	B>W
Hausmann65	Arthritis Care Res(Hoboken)	2017	Knee	VA Musculoskeletal Disorders Cohort	Primary predictor	B: 50,172, W:473,170	Proportion by pain severity	No	No	NRS	No	B>W*
Herbert68	Ann of Behav Med	2014	Knee	UPLOAD Study	Control	B:81, W:87	Interclass Correlation	Yes	Yes	WOMAC	No	B>W*
Herbert67	Clin J Pain	2017	Knee	UPLOAD Study	Primary predictor	B:56, W:35	Mean Pain Score, Mean Difference(β)	No	Yes	VAS	Yes†	B>W*
Ibrahim72	Artdritis Care Res (Hoboken)	2001	Hip, Knee	Cleveland VAMC	Primary predictor	B:260, W:332	Mean Pain Score	No	Yes	WOMAC	Yes	B=W
Ibrahim74	Med Care	2002	Hip, Knee	Cleveland VAMC	Primary predictor	B:262, W:334	Mean Pain Score	Yes	Yes	WOMAC, Lequesne	Yes†	B=W
Ibrahim73	Arthritis Rheum	2002	Hip, Knee	Cleveland VAMC	Primary predictor	B:262, W:334	Mean Pain Score	Yes	Yes	WOMAC, Lequesne	Yes†	B=W
Ibrahim71	J Gerontol A Biol Sci Med Sci	2003	Hip, Knee	Cleveland VAMC	Primary predictor	B:135, W:165	Mean Pain Score	Yes	Yes	WOMAC, VAS	Yes†	B=W
Johannes76	J Pain	2010	Unspec	US Households	Control	B:2073, W:21,992	Proportion w/ chronic pain	No	Yes	Single pain question	No	B<W*
Jones78	J Cross Cult Gerontol	2008	Unspec	Philadelphia VAMC, Pittsburgh VAMC	Primary predictor	B:459, W:480	Mean Pain Score	Yes	Yes	WOMAC	Yes†	B>W*
Jordan80	J Rheumatol	2009	Hip	JoCoOA Project	Control	B:999, W:2069	Proportion w/ Symp OA	No	No	ACR/ NHANES criteria	No	B>W*
Kwoh86	Arthritis Res Ther	2015	Knee	Pittsburgh VAMC	Primary predictor	B:285, W:514	Mean Pain Score	Yes	Yes	WOMAC	Yes	B>W*
Lachance87	Osteoartdritis Cartilage	2001	Knee	MBHS, SWAN- Michigan Center	Primary predictor	B:323, W:506	Proportion w/ pain and OA	No	No	Arthritis Questionn aire	No	B>W*
Lavernia89	J Artdroplasty	2004	Hip, Knee	Orthopedic Institute at Mercy Hospital	Primary predictor	B:88, W:157	Mean Pain Score	Yes	Yes	WOMAC, VAS, SF-36	Yes	B>W
Lavernia88	Clin Ortdop Relat Res	2010	Hip, Knee	Orthopedic Institute at Mercy Hospital	Primary predictor	B:49, W:282	Mean Pain Score	Yes	No	WOMAC, SF-36	Yes	B>W*
MacFarlane95	J Clin Rheumatol	2018	Knee	VITAL Trial	Primary predictor	B:285, W:785	Mean Pain Score	Yes	Yes	WOMAC	Yes	B>W*
McIlvane97	Aging Ment Healtd	2007	Knee, Back, Hand	Tampa Metropolitan, FL	Primary predictor	B:77, W:98	Mean Pain Score	Yes	Yes	AIMS2	Yes†	B>W
Mcllvane98	J Gerontol Psych Soc Sci	2008	Unspec	Tampa Metropolitan, FL	Primary predictor	B:77, W:98	Mean Pain Score	Yes	Yes	AIMS2	Yes	B>W
Mingo101	J Nat Med Assoc	2008	Unspec	Tampa Metropolitan, FL	Primary predictor	B:77, W:98	Mean Pain Score	Yes	No	AIMS2	Yes†	B>W
Moss102	Osteoarthrisis Cartlidge	2016	Hip	JoCo OA Project	Control	B:500, W:2278	Incidence Rate Symp OA	No	Yes	Single pain question	No	B<W*
Murphy104	Osteoarthrisis Cartlidge	2010	Hip	JoCo OA Project	Control	B:496, W:2260	Incidence Rate Symp OA	No	Yes	Single pain question	No	B>W
Murphy106	Arthritis Care Res	2016	Knee	JoCo OA Project	Primary predictor	B:319, W:1199	Lifetime Risk Symp OA	No	Yes	Single pain question	No	B<W
Neuburger110	J Public Health (Oxf)	2012	Hip, Knee	PROMs; HES Database	Control	B:995, W:107176	Mean Pain Score	Yes	Yes	OHS, OKS	Yes	B>W*
Parmelee114	J Aging Health	2012	Knee	Philadelphia Metropolitan, PA; Philadelphia VAMC	Primary predictor	B:94, W:269	Mean Pain Score	Yes	No	PGC	Yes	B>W
Parmelee112	Sleep Health	2017	Knee	Western/Central Alabama; Long Island, NY	Primary predictor	B:96, W:128	Mean Pain Score	No	Yes	5-point scale	Yes	B>W*
Petrov115	J Pain	2015	Knee	UPLOAD Study	Control	B:88, W:52	Mean Pain Score	Yes	Yes	WOMAC	Yes†	B>W
Siedlecki123	Pain Manag Nurs	2009	Unspec	Nortdeast Ohio	Primary predictor	B:36, W:24	Mean Pain Score	No	Yes	VAS, MPQ-SF	Yes	B>W; B<W
Smitd125	Arthritis Care Res (Hoboken)	2016	Knee	Western Alabama; Long Island, NY	Control	B:39, W:81	Mean Pain Score	Yes	Yes	PGC	Yes	B>W*
Song128	Arthritis Care Res (Hoboken)	2013	Knee	OAI	Primary predictor	B:286, W:1603	Mean Pain Score	No	Yes	WOMAC	Yes	B>W*
Sowers130	Am J Epidemiol	2006	Knee	MBHS, SWAN- Michigan Center	Control	B:211, W:669	Proportion w/ pain	Yes	No	Single pain question	No	B>W*
Vina135	Osteoarthrisis Cartlidge	2018	Knee	OAI	Primary predictor	B:778, W:3498	Mean Pain Score	Yes	Yes	WOMAC, NRS	Yes	B>W*
Weng137	Arthritis Rheum	2007	Knee	Los Angeles VAMC	Primary predictor	B:54, W:48	Mean Pain Score	Yes	No	WOMAC	No	B>W*
Wright140	J Musculoskelet Pain	2015	Neck, Should er	JoCo OA Project	Control	B:523; W:1149	Proportion w/ symptoms	Yes	No	Single pain question	No	B<W*
Yang141	BMC Complement Alt Med	2012	Knee	OAI	Primary predictor	B:508, W:2075	Mean Pain Score	Yes	Yes	KOOS Pain	Yes	B>W*

^†Met eligibility, but removed from final analysis due to duplicate data

^*Indicates statistical significance

Dataset Abbreviations:Einstein Aging Study(EAS), Fitness Arthritis in Seniors Trial(FAST), Health, Aging and Body Composition(Health ABC), Study, Johnston County Osteoarthritis Project(JoCo OA), Michigan Bone Health Study(MBHS),National Health Examination and Nutrition Survey(NHANES),National Health Service Hospital Episode Statistics(HES), database, National Health Service Patient Reported Outcome Measures Programme(PROMs), Osteoarthritis Initiative(OAI), Self-Management of OsteoArthritis in Veterans Study(SeMOA), Study of Women’s Health Across the Nation(SWAN-Michigan Center), University of New Mexico Health Sciences Center Orthopaedic Clinic(UMN Orthopaedics Clnics), Understanding Pain and Limitations in Osteoarthritic Disease(UPLOAD), Study, VITamin D and OmegA-3 Trial(VITAL)

Pain Scale Abbreviations:American College of Rheumatology(ACR), Arthritis Impact Measurement Scales(AIMS2), Graded Chronic Pain Scale(GCPS), Knee Injury and Osteoarthritis Outcome Score(KOOS), Lequesne Index of Severity for Osteoarthritis(Lequesne), McGill Pain Questionnaire Short Form(MPQ-SF), Medical Outcomes Study 36-item Short Form(SF-36v2), Numeric Rating Scale(NRS), Oxford Hip Score(OHS), Oxford Knee Score(OKS), Philadelphia Geriatric Center Pain Scale(PGC), Sickness Impact Profile(SIP) , Total Pain Index(TPI), Visual Analog Scale(VAS), Western Ontario Macmaster University Osteoarthritis Index(WOMAC)

Among our sample, 45 articles treated race as a primary predictor, while 16 articles treated race as a control variable. Predominantly, race was a dichotomous variable of African-American/Black versus White/Caucasian; however, a few studies included additional stratification by Hispanic ethnicity (N=15) or race as a categorical variable that included Asian/Pacific Islander or Other (N=6).^{38,45,53,57,58,63,65,67,68,76,89,110,112,115,125,135}

A large subset of articles (N=40) assessed disability or functional limitations through self-report, performance testing or a combination of modalities. Most articles measured function simultaneous with pain using the same composite scales. Function-specific assessments included WOMAC, AIMS2, Short Physical Performance Battery (SPPB), Sickness Impact Profile (SIP), Medical Outcomes Study 12-item Short Form (SF-12) Physical Component, Knee Injury and Osteoarthritis Outcome Score Physical Function (KOOS-PF), Disability of Arm, Shoulder and Hand (DASH) Questionnaire, Stanford Health Assessment Questionnaire Disability Index (HAQ) and SF-36v2 Physical Function.^{17,18,50,61,62,69,96,99,119,136} Performance testing included timed walking, gait, stair climb, standing, and grip strength.

SES was reported in 43 articles using a combination of different measures. Education (N=41), income (N=27) and employment status (N=16) were most commonly reported as dichotomous or categorical measures. Health insurance status was reported by 4 articles; additional characteristics included geographic region and internet access.^{8,13,45,76,86,89,95} One study calculated the Index of Multiple Deprivation (IMD), a scale based on zip code that incorporates a multiple SES factors previously mentioned along with environment, crime and barriers to housing and other social services.¹¹⁰

Synthesis of Results

The qualitative synthesis included 61 articles reporting analyses of pain stratified by race. Of these, 46 showed higher pain severity in Blacks/African Americans, but only 32 were statistically significant. Contrastingly, 7 articles reported higher pain in Whites/Caucasians, with 3 having a statistically significant difference. One article reported higher pain scores in African-Americans using the VAS and in Caucasians using the MPQ-SF.¹²³ The remaining 9 articles concluded that there was no race group difference in pain.

In the quantitative analysis, forest plots were created from 28 studies reporting mean pain scores. Among WOMAC studies (Figure 2), the overall standard mean difference showed higher pain severity in African-Americans with a moderate effect size (0.57; 95%CI: 0.54, 0.61). Similar results were found in non-WOMAC studies, but with a slightly lower effect size (0.35; 95%CI: 0.23, 0.47) (Figure 3). The forest plots of 21 studies from our secondary analysis of disability measures are presented in Figures 4 and 5. The findings were similar to those observed for pain measures, with African-Americans showing higher, moderate effect size for self-reported disability (0.38; 95%CI: 0.22, 0.54) and poorer performance testing (−0.58; 95%CI: −0.72,−0.44).

Figure 2:

Forest Plot of WOMAC Pain Measures Higher scores=worst outcome/most severe pain

Figure 3:

Forest Plot of Non-WOMAC Pain Measures Studies that reported using more than one pain scale are listed multiple times. Higher scores=worst outcome/most severe pain. Scales used by Lavernia 2004, 2010 (SF-36v2), Neuberger 2012 (OHS, OKS), and Yang 2012 (KOOS) are reverse coded, therefore mean scores were recalculated by subtracting from the maximum possible value for that scale.

Figure 4:

Forest Plot of Self-Reported Disability Measures Studies using WOMAC: Allen 2009; Ang 2003; Cruz-Almeida 2014; Gandhi 2008; Golightly 2005; Lavernia 2004; Lavernia 2010; MacFarlane 2018; Vina 2018. Studies using AIMS2: McIlvane 2008, Parmelee 2012; Smith 2016. Studies using SF-12 or SF-36v2: Groeneveld 2008; Kwoh 2015; Yang 2012. Other scales used include SIP: Cano 2006; GCPS: Cruz-Almeida 2014; and KOOS: Yang 2012. Studies the reported using more than one pain scale are listed multiple times. Higher scores=worst outcome/less mobility; The SF-12, SF-36v2 and KOOS are reverse coded, therefore mean scores were recalculated by subtracting from the maximum possible value for that scale.

Figure 5:

Forest Plot of Performance Testing Disability Measures Performance measures include Gait velocity: Blanco 2012; Walking test: Burns 2007; and Short Physical Performance Battery (SPPB): Cruz-Almeida 2014. Lower scores= worst outcome/less mobility.

Visual inspection of funnel plots for pain and disability showed a high concentration of studies fell within the 95% CI line. Minimal horizontal scatter was present; a result of heterogeneity. Studies with larger sample sizes closer to apex had a smaller standard error; therefore, exhibited more precision. Both plots were symmetrical around the SMD; however, the absence of studies near the middle and bottom of the funnel indicate the presence of publication bias.

Discussion

The objectives of this meta-analytic review were to characterize differences in clinical pain severity and disability between AAs and NHWs with OA and to quantify the magnitude of these effects. Overall, clinical pain severity in OA was higher amongst AAs with studies using WOMAC and non-WOMAC scales. Similarly, AAs had greater self-reported disability than NHWs, greater disability in performing and doing functional tasks (e.g. timed walking test). On average, the effect sizes for the observed race differences in both pain and disability were moderate in magnitude, indicating that AAs with OA are more likely to experience significantly greater pain and disability, compared to their NHW counterparts.

Several large representative studies provide evidence that AAs are more likely to have radiographic and symptomatic OA and greater severity of knee/hip OA than NHWs.^{38,39,79,80,131} The current findings reveal that, in addition to differences in prevalence, AAs with OA also report greater pain and disability, and these racial/ethnic differences are moderate in magnitude. However, there was some variability in the findings across studies. Of the 61 studies included in the current meta-analysis, 75% observed higher pain severity among AAs, whereas 11% found that NHWs experienced higher pain, and 15% found no difference in pain between the races. Inconsistencies in race differences in OA-related pain may emerge due to a variety of methodological factors, including sample characteristics, measures of pain used, and analytic approach, like variables being controlled for in the analyses. Specifically, the populations from which samples are recruited (e.g. clinic vs. community, veteran vs. civilian), health status, comorbidities, and sociodemographic characteristics (e.g., income, educational attainment) may differentially impact outcomes.⁸¹ For example, several studies that found no race differences in pain were based on veteran samples and enrolled patients only with moderate to severe symptoms.^{10,12,13,71–74} Such sampling methods may be important for achieving other study aims, but they would likely attenuate the magnitude of observed race group differences in pain. Also, the studies using the WOMAC (Figure 2) showed larger group differences than studies using other measures. However, this could reflect the influence of the affected joint, since the WOMAC is used exclusively for lower extremity OA (hip and knee) while studies in Figure 3 reflect multiple OA sites. Regarding analytic approaches, some studies have controlled for a variety of potential confounders, including radiographic measures, and SES, while others have not. Hence, the analytic approach may contribute to the inconsistent findings. Given that OA-related pain contributes to greater functional limitation and disability ⁷⁵, understanding the potential mechanisms underlying racial/ethnic differences in OA is important to improve pain management.

Disparities in pain management among racial/ethnic minorities, specifically AAs have been documented.^9,40 Despite higher rates of OA-related pain and disability, AAs undergo hip and knee joint replacements at lower rates than NHWs.^29,44,133 Moreover, racial disparities in prescribing medications for OA have been documented.^{41,59,100,121} The unmet need of pain management in AA compared to NHW likely contributes to higher levels of pain and disability in this cohort.

Obesity is a risk factor that contributes to the development of OA ^19,124, as well as increases the risk of radiographic progression.^43,120 JoCoOA and Framingham found that obesity was independently associated with disability and compounded disability from knee pain.^28,47,82 In light of higher rates of obesity in AAs than NHWs with OA, studies have investigated the extent to which racial/ethnic differences in obesity contribute to disparities in OA-related pain.^{7,32,53,82,104,128} Furthermore, BMI and depression were important factors in explaining racial differences in pain and physical function.⁷

Differences in OA disease severity and phenotype may contribute to race group differences. Some evidence suggests that genetic influence on the etiology of OA may contribute to race differences in diseases outcomes.^93,142 A genome-wide association study of knee OA provided evidence for differences in the genetic architecture of knee OA between race groups.⁹³ Racial differences in the radiographic features of knee ²¹, hip ¹⁰⁷ and hand ^38,131 OA have been documented in AAs which may contribute to race differences in OA severity. Specifically, AAs have more severe tibiofemoral OA, tricompartmental OA’ and lateral joint-space narrowing, as well as a higher prevalence and severity of osteophytes and higher frequency of sclerosis than NHWs. ²¹ Similarly, compared to NHW, AAs were more likely to have a valgus thrust during walking, which could increase the risk of lateral knee OA and were more likely to have joint space narrowing, osteophytes, and cysts compared to NHWs.^27,107

Beyond differences in the OA disease process, race group differences in pain processing could contribute to greater pain severity among AAs. Indeed, considerable evidence from quantitative sensory testing (QST) in healthy adults demonstrates lower pain thresholds and tolerances, and greater experimental pain sensitivity among AAs with knee OA compared to their NHW counterparts.^{14,35,83,84,117}

Socioeconomic status (SES) represents another important explanatory variable in understanding racial disparities in OA-related pain and disability. Lower SES is associated with poorer health outcomes and increased morbidity ¹¹⁶ and is a risk factor for the development and or progression of OA.⁹⁴ It has been widely documented that AAs on average have a lower socioeconomic status compared to NHWs.^{4,23,59,64,94,116} Whether the greater OA-related pain observed in AAs is accounted for by SES has not been adequately addressed. Studies that control for SES often simultaneously adjust for other variables (e.g., BMI, age, sex); therefore, independent contribution of SES is not discernable. However, even after controlling for these characteristics, race differences in OA-related pain and disability have not been completely explained. Thus, it is well documented that low SES is an independent predictor of OA risk and confers poorer OA-related pain and disability outcomes.^{30,85,94,103,105} Low SES likely contributes to poorer pain outcomes via multiple mechanisms, including: 1) less access to optimal medical care, 2) increased levels of environmental and psychosocial stress, or 3) lower availability of beneficial resources such as exercise facilities, dietary nutrients, among others.⁹⁴ However, the extent to which SES drives racial/ethnic differences and OA pain remains unknown. Several factors may contribute to this lack of information: 1) many researchers do not adequately control for SES, 2) researchers adjust for other variables (e.g., age, BMI, sex), so it is hard to know what SES contributes, and 3) studies operationalize SES in different ways, so it is unknown which variables are important. Stratifying analyses using a standard set of SES measures could help disentangle the complex effects of race and SES.³⁵

Limitations

There are several limitations of this systematic review and meta-analysis. To improve the quality of data, sources were restricted to journal articles published in peer-reviewed journals. Including data from grey literature such as dissertations, and scientific proceedings, might have provided additional insight into the results and reduced publication bias. Additionally, race was classified using different terminology across studies (e.g., whites and other), and more consistent approaches to race classification are needed to ensure comparability of findings. Moreover, many studies did not report on race groups when reporting demographic characteristics of participants, which limited inclusion of studies in the meta-analysis. As previously mentioned, studies evaluated pain using a wide variety of pain scales, such that 31 studies used the WOMAC, 12 studies used VAS or NRS, and a handful of studies evaluated pain using a single question. Therefore, separate effect sizes were calculated for studies using the WOMAC and non-WOMAC pain scales. Lastly, although several studies have evaluated the impact of SES on OA, ^24,25 few studies have evaluated how SES impact race differences and OA-related pain outcomes.

Conclusions

Future research should investigate the mechanisms underlying these differences, including the extent to which SES affects race differences and OA-related clinical pain severity. We provide meta-analytic evidence for higher clinical pain severity among AAs compared to NHWs with knee OA, regardless of pain scale. Similarly, AAs with knee OA reported more functional limitations and showed poorer performance on functional tasks (e.g., timed walking test) compared to NHWs with knee OA. On average, the effect sizes for the observed race group differences in both pain and disability were moderate in magnitude, indicating that AAs with OA are more likely to experience significantly greater pain and disability, compared to their NHW counterparts. The first step in successfully reducing racial disparities in pain and osteoarthritis research requires adhering to consistent definitions of racial categories, as well as, making the appropriate distinction between race and ethnicity. Such research may lead to better understanding of racial/ethnic differences in OA-related pain and inform future interventions to reduce these disparities and improve pain management.

Appendix

Figure A1:

Funnel Plot of WOMAC & Non-WOMAC Pain Measures

Figure A2:

Funnel Plot of Disability Measures

Table A1:

Detailed Search Strategy by Data Source

Data Source	Search String
MEDLINE® (PubMed®) Filters: English	“African Continental Ancestry Group”[MH] OR “black men”[TIAB] OR negroid[tiab] OR (negro [tiab] NOT (river OR rio)) OR blacks [TIAB] OR “black man”[TIAB] OR “black woman”[TIAB] OR “black women”[TIAB] OR “black population”[TIAB] OR “black populations”[TIAB] OR “black community”[TIAB] OR “black communities”[TIAB] OR “black patients”[TIAB] OR “black patient”[TIAB] OR “black race”[tiab] OR “black person”[tiab] OR “black persons”[tiab] OR “black people”[tiab] OR “white men”[TIAB] OR “white man”[TIAB] OR “white woman”[TIAB] OR “white women”[TIAB] OR “white population”[TIAB] OR “white populations”[TIAB] OR “white community”[TIAB] OR “white communities”[TIAB] OR “white patients”[TIAB] OR “white patient”[TIAB] OR “European Continental Ancestry Group”[Mesh] OR caucasoid [tiab] OR Caucasian* [tiab] OR Caucasians [TIAB] OR “white race”[tiab]OR “white person”[tiab] OR “white persons”[tiab] OR “white people”[tiab] OR “ethnic group” [TIAB] OR “ethnic groups” [TIAB] OR “Ethnic Groups” [MH] OR “ethnic population” [TIAB] OR “ethnic populations” [TIAB] OR Hispanic [TIAB] OR Hispanics [TIAB] OR Latino [TIAB] OR Latinos [TIAB] OR Latina [TIAB] OR Latinas [TIAB] OR “minority group” [TIAB] OR “minority groups” [MH] OR “minority groups” [TIAB] OR “minority population” [TIAB] OR “minority populations” [TIAB] OR “people of color” [TIAB] OR “ethnic minority”[tiab] OR “ethnic minorities”[tiab] OR whites [TIAB] OR “ethnic disparities” [TIAB] OR “ethnic disparity” [TIAB] OR “Minority Health” [MH] OR “minority health” [TIAB] OR “racial disparities” [TIAB] OR “racial disparity” [TIAB]OR “ethnic difference”[tiab] OR “ethnic differences”[tiab]) AND (“pain”[mesh] or pain[tiab]) AND (“Osteoarthritis”[Mesh] or osteoarth*[tiab])
Web of Science™ Core Collection Language: English Document Types: Article	#1: TS=(“minority health” OR (minority AND health) OR “minority group*” OR “ethnic group*” OR “health disparit*” OR “African American*” OR negroid OR colored OR “people of color” OR latino* OR latina* OR hispanic* OR chicano* OR chicana* OR Mexicans OR “Mexican American*” OR caucasian* OR whites OR Caucasoid) #2: TS=(pain) AND TS=(osteoarthrit*) #3: #1 and #2
PsycInfo (EBSCOhost Web) Source Type: Academic Journals Language: English	((AB osteoarth* OR TI osteoarth*) AND (( DE “Pain” OR DE “Aphagia” OR DE “Back Pain” OR DE “Chronic Pain” OR DE “Myofascial Pain” OR DE “Neuralgia” OR DE “Neuropathic Pain” OR DE “Somatoform Pain Disorder OR DE “Muscle Contraction Headache” OR DE “Trigeminal Neuralgia” ) AND ( DE “Racial and Ethnic Differences” OR DE “Health Disparities” OR DE “Latinos/Latinas” OR DE “Mexican Americans” OR DE “Blacks” OR DE “Whites” OR DE “Hispanics”) OR (AB (osteoarth* AND pain) OR TI ( osteoarth* AND pain )) AND AB (“minority health” OR (minority AND health) OR “minority group*” OR “ethnic group*” OR “health disparities” OR “African American* OR negroid OR colored OR “people of color” OR latinos OR latinas OR hispanics OR chicanos OR chicanas OR Mexicans OR “Mexican American*” OR caucasian* OR whites OR caucasoid) OR TI (“minority health” OR (minority AND health) OR “minority group*” OR “ethnic group*” OR “health disparities” OR “African American* OR negroid OR colored OR “people of color” OR latinos OR latinas OR hispanics OR chicanos OR chicanas OR Mexicans OR “Mexican American*” OR caucasian* OR whites OR caucasoid))
Cochrane Library	#1: MeSH descriptor: [Osteoarthritis] explode all trees #2: MeSH descriptor: [Pain] explode all trees #3: MeSH descriptor: [Minority Groups] explode all trees #4: MeSH descriptor: [Minority Health] explode all trees #5: MeSH descriptor: [African Continental Ancestry Group] explode all trees #6: MeSH descriptor: [Ethnic Groups] explode all trees #7:“African American” or “African Americans” or “African ancestry” or black or blacks or Caucasian or Caucasians or “ethnic group” or “ethnic groups” or “ethnic population” or “ethnic populations” or Hispanic or Hispanics or Latino or Latinos or Latina or Latinas or “minority group” or “minority groups” or “minority population” or “minority populations” or “people of color” or white or whites or “ethnic disparities” or “ethnic disparity” or “minority health” or “racial disparities” or “racial disparity”:ti,ab,kw (Word variations have been searched) #8: osteoarth*:ti,ab,kw (Word variations have been searched) #9:pain:ti,ab,kw (Word variations have been searched) #10: #1 or #8 #11: #2 or #9 #12: #3 or #4 or #5 or #6 or #7 #13: #10 and #11 and #12

Databases searched October 1–6, 2016, November 9, 2017, and May 30, 2018