Abstract
Objective To characterize the relationship between the frequency of idiopathic osteoarthritis (OA) and characteristics including demographics, comorbidities, military service history, and physical health in a veteran population.
Methods We performed a cohort study in the Million Veteran Program (MVP) using International Classification of Diseases, 9th and 10th revision codes to define the frequency of site-specific OA across 3 joints or unspecified OA in veterans with respect to demographics (eg, age, sex, race and ethnicity), military service data, detailed electronic health records, military branch, and war era.
Results We validated previous reports of sex- and age-dependent differences in OA frequency, and we identified that unspecified OA was associated with a higher frequency of 16 Deyo-Charlson comorbidities. These associations generally persisted within each isolated joint site–specific OA. Depending on military branch, prior military engagement was differentially associated with the frequency of OA. Prior United States Army and Navy service were associated with higher and lower risk, respectively, of OA across all joint sites; however, multivariable-adjusted models adjusting for a range of covariates, including age, sex, and ancestry, reversed the apparent protective effect of prior Navy service.
Conclusion These findings highlight the breadth of factors associated with OA in the MVP veteran population and suggest that physical status may be a modifiable risk factor for OA. This work may help in the design of strategies to optimize appropriate detection, intervention, treatment, and even rehabilitation for OA in veterans and the general population.
Osteoarthritis (OA) is the most prevalent form of arthritis and one of the leading causes of disability around the world.1 An estimated 14% of the US adult population has OA,2 but this frequency is reportedly up to double in US military veterans.3 OA is associated with the presence of a range of comorbidities,4 reduced quality of life, and heightened mortality.5 The severity and multisystem impact of OA is evident across tissues and includes its characteristic articular cartridge breakdown, synovitis, and inflammatory remodeling of the underlying bone6 and overlying skeletal muscle.7 Further, the cyclical effects of pain and limited mobility on sedentarism and disability contribute to a progressively declining functional state,2 often persisting even after individuals with OA in operable joint sites (eg, knee, hip) undergo elective surgery. Risk factors for the development of OA, including older age, sex, trauma, obesity, genetics, activity patterns, and anatomical differences,1 remain incompletely understood. In veterans, differential physical demands and rates of combat-related trauma across military branches could be associated with risk of OA in later life. Supporting this, joint-related mobility impairment and disability contribute to high rates of discharge and associated compensation costs for the military.8 Although existing studies have focused on posttraumatic OA (PTOA) in the military population, few have explored OA that is not attributable to an injury, which could be valuable in clarifying how other factors influence the risk of OA among a diverse population of military veterans.
Current treatment options for OA are limited, with conservative medical treatments providing temporary benefits mostly focused on pain and symptom management, or lifestyle interventions that are difficult to sustain.9 Concerningly, OA-related pain and disability is a central contributor to depression in older veterans.10 Most individuals with OA eventually elect to undergo joint replacement surgery as the disease progresses11 in order to mitigate chronic pain and disability. Knee arthroplasty was recently reported to be the fifth most common surgical procedure according to the US Department of Veterans Affairs (VA).12 However, a large proportion of veterans continue to experience pain and poor physical function, even after surgery and years into rehabilitation.13 Further, decision quality associated with elective joint replacement surgery in veterans is lower than that in the general population,14 suggesting a need to improve patient care at multiple stages of OA.
As the knee is not the only joint commonly affected by OA, there remains a need to understand the contributors and comorbidities associated with OA at other sites (eg, hip, spine). The Million Veteran Program (MVP) cohort was established to enable biomedical research related to physical and mental wellness in US military veterans, with the goal of enrolling 1 million participants. Leveraging this robust dataset, our research team recently identified a handful of novel genetic loci associated with OA across genetic ancestry groups, replicated previously identified loci, and discovered new loci, some of which were uniquely associated to OA in only a given ancestry group.15 Additionally, the MVP database provides an opportunity to investigate the role of lifestyle factors unique to the military experience, such as service era, branch, and active duty status. To gain a clearer understanding of OA in military veterans, we investigated multiple characteristics derived from military service histories and electronic health records in relation to idiopathic, site-specific, and unspecified OA in a population of military veterans.
METHODS
Demographics. This study used an observational cohort design. Participants were curated from the MVP database and all data were captured between 1990 and 2019. Basic demographic information was included on the MVP survey, which has been described previously.16 All participants had previously consented to sharing their deidentified data for research. The work described in this manuscript received ethical and study protocol approval from the Veterans Affairs Central Institutional Review Board as well as the University of Alabama at Birmingham, in accordance with the principles outlined in the Declaration of Helsinki. See the Supplementary Material (available with the online version of this article) for information on adherence to Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) cohort study guidelines.
Cases of OA were identified from electronic health record data using International Classification of Diseases (ICD), 9th and 10th revision codes. Inclusion and exclusion criteria were adapted from Zengini et al.17 Specific ICD codes used for the identification of participants without OA, those with unspecified OA, and those with OA across each of 3 joint sites of interest were derived from the Genetics of Osteoarthritis (GO) consortium18 and are presented in previous work by our research team,15 as well as in Supplementary Table S1 (available with the online version of this article). Briefly, site-specific OA was defined as cases in which an individual had ≥ 1 diagnostic code for an unspecified OA category, along with ≥ 1 diagnostic code for a specific subtype based on the location of the affected joint (eg, hip, knee, and/or spine), registered ≥ 30 days apart.19 We used the definition requiring the presence of 2 OA codes to increase the specificity of defining the presence of OA.19 Individuals with 2 codes for unspecified OA separated by 30 days were collapsed into the unspecified OA category. Non-OA participants were defined as individuals who did not have any of the ICD codes for OA. Veterans presenting with ≥ 1 subtype of OA were also included in the dataset, as were veterans with prior OA-related joint replacements of the hip or knee. Participants without OA were identified as those in the dataset that did not have any of the preset OA inclusion criteria or any ambiguous codes potentially related to the presence of OA or various non-OA arthropathies. Trauma-based arthropathies were identified by searching for any instance of the string “trauma” in ICD codes among those with OA (see Supplementary Table S1 for complete list of PTOA ICD codes). Individuals with PTOA were then filtered from the dataset in order to focus on idiopathic OA. As shown in Supplementary Figure S1, the individuals with PTOA accounted for ~6% of the dataset.
Military service. Variables related to military service were collected, including whether the individual performed any prior service (coded as active, reserves, not applicable, or missing [ie, not provided by an individual who otherwise completed the survey]), which branch(es) they served in (eg, Army, Marine Corps, National Guard), and whether no prior service data were available (importantly, MVP participants may also include veterans’ spouses without any active service history). Calendar years of military service were divided into “eras” based on the primary US military focus of the associated time period (eg, World War II [WWII], Korean War, Persian Gulf War, modern War on Afghanistan/Iraq). The military service era data points were either framed as yes/no (ie, no missing) or provided automatically by the VA system, leading to a complete dataset, whereas the branch data have a higher percent missing. In order to enable positive responses to multiple options in both of these categories (ie, served in ≥ 1 branch and/or era), as well as to enable an analysis with a higher sensitivity, each of these responses was coded as binary (yes/no) and tested as a separate subcategory. MVP survey data describing military branch and service type were not available for ~35% of participants; however, missing data for these variables was not an exclusion criterion, as the remaining numbers were still sufficient to power analyses of interest.
Comorbidities and health status. The presence of 17 comorbid conditions across body systems were assessed. Conditions included unspecified diseases of perivascular, cerebrovascular, renal, and pulmonary origin, as well as cancer, HIV or AIDS, diabetes mellitus, and others, as shown in Table 1. In order to quantify the collective burden of comorbidities, the commonly used Deyo-Charlson Comorbidities Index (DCCI) was applied,20 and veterans were grouped into categories based on DCCI scores of 0, 1, 2, or ≥ 3. Veterans were also asked to rate their overall fitness status. The question, “how would you rate your current physical fitness status?” had the following answer options: (1) very good, (2) fairly good, (3) satisfactory, (4) fairly poor, (5) very poor, and (6) missing. The physical demand of their present job was evaluated by the question, “how physically strenuous is your work/job (paid and unpaid)?” with the following options: (1) very light (mainly sitting), (2) light (mainly walking), (3) medium (lifting, carrying light loads), (4) heavy manual work (climbing, carrying heavy loads), and (5) missing.
Demographics and health status of veterans with and without unspecified OA in the Million Veteran Program.
Statistical tests. All analyses were performed on the VA Informatics and Computing Infrastructure platform using SQL Server Management Studio (Microsoft) and RStudio (R version 4.3.1; R Foundation for Statistical Computing). All numeric data are presented as mean (SD) and categorical variables as total n (%) relative to the entire group.
Univariate modeling. For numerical values (eg, age and DCCI), a t test was performed to compare values between veterans with and without OA. For more nuanced examination, age was subdivided into 10-year categories. All categorical and binary (yes/no) outcomes were compared between veterans with and without OA using a chi-square goodness of fit analysis of proportions. Each categorical characteristic was also compared individually between veterans with and without site-specific OA (ie, for the purposes of this investigation, joint sites were not directly compared to one another). This was done in order to establish whether a given joint might be differentially associated with a demographic characteristic in comparison to a control condition. Significant differences were declared at P < 0.05 and statistical trends (P < 0.10) were acknowledged.
Multivariable-adjusted modeling. To investigate the independent influence of each variable of interest on the likelihood of unspecified and site-specific OA, multivariable-adjusted models were constructed. Variables included in the multivariable-adjusted models were age group, sex, ancestry, BMI (calculated as weight in kilograms divided by height in meters squared), DCCI score, military branch, and military service type. Additional models were constructed to examine interactions with ancestry and sex with each variable of interest as well as sex-stratified and ancestry-stratified models.
RESULTS
Unspecified OA: descriptives and unadjusted estimates. Table 1 shows the demographic information for veterans with and without unspecified OA. Each demographic variable was associated with unspecified OA, including female sex; African American, American Indian or Alaska Native ancestry; and non-Hispanic/Latine ethnicity. The rates of previous joint replacement were higher in veterans with unspecified OA than in those without OA for both hip and knee arthroplasties. Only 116 and 74 individuals without OA had undergone hip (0.06%) and knee (0.04%) replacement surgery, respectively.
Mean DCCI was higher (2.4 [SD 2.5]) in veterans with unspecified OA than in those without (1.1 [SD 1.8]; Table 1). Frequency of OA differed across categorical DCCI levels, reflecting this same trend. Specifically, 37.70% of veterans with unspecified OA had a DCCI category of ≥ 3 in contrast with only 14.56% of veterans without OA. The presence of 16/17 Deyo-Charlson comorbidities was positively associated with unspecified OA. The only comorbidity that had a higher frequency among veterans without OA than in veterans with unspecified OA was HIV/AIDS, although this constitutes a small portion of the sample.
The frequency of individuals with unspecified OA also differed from those without unspecified OA in terms of physical status and physical demand at work (Table 2). In general, those with OA reported being in categories of lower fitness and were more likely to report prior active duty or involvement in any military service. Veterans with unspecified OA were more likely to have served in the Army and National Guard and less likely to have served in the Air Force, Coast Guard, Navy, and multiple branches. A nonsignificant trend toward this pattern was seen in those who had served in Public Health Service. Veterans with unspecified OA were more likely to have served between May 1975 to July 1990 or during the Vietnam War era but significantly less likely to have served in every other military era except for pre-WWII (November 1941 and earlier), for which no statistical difference was detected (Table 2). Missing military branch/service type data are presented in the Table 2 footnotes.
Physical demand, status, and military service in veterans with and without unspecified OA.
Unspecified OA: adjusted estimates. In general, significant associations from the univariate analyses were robust in multivariable-adjusted models that adjusted for age group, sex, ancestry, BMI, DCCI score, military branch, and military service type (Table 3). Independent of all other factors, frequency of OA was higher in older age groups relative to the youngest group studied. However, risk was modified such that those in the 40-49 years, 50-59 years, and 60-69 years age groups had the highest risks of OA (OR 4.4, OR 6.1, and OR 4.8, respectively), whereas those aged ≥ 70 years experienced a relatively attenuated higher OA risk (OR 3.6-3.8). These associations persisted when the data were stratified by ancestry and sex, but no interactions were found. BMI showed a mild but significant relationship with the risk of OA (OR 1.1) that persisted for both ancestry- and sex-adjusted and stratified models. Likewise, the heightened OA risk associated with higher comorbidity burden, as indicated by DCCI, was preserved across both sexes as well as both ancestries. None of the unadjusted associations with military branch type were changed when stratified for race or sex (Supplementary Table S2, available with the online version of this article).
ORs for OA at multiple joint sites in veterans from the Million Veteran Program as determined by multivariable-adjusted models.
Site-specific OA: descriptives and unadjusted estimates. In the present MVP dataset, OA was most frequent at the knee joint, accounting for approximately 22.6% of MVP veterans (Table 4). As Figure 1 shows, there were often multiple occurrences of OA at joint sites within an individual, particularly in the hand, finger, and thumb. Taken together, OA at some hand site (hand, finger, or thumb) accounted for 3% of veterans in our study. Figure 1 also shows the large number of individuals who had medical records with unspecified OA and no additional data regarding site-specific subtype. Frequency of OA across all demographic categories (age, sex, and race and ethnicity) varied within each site relative to the control group, generally in the direction of the trends seen for unspecified OA. The proportion of veterans who had undergone prior joint replacement was significantly different within each site vs control (Table 4). Although joint sites were not compared to one another, the overwhelming majority of prior joint replacement for operable joints occurred within their respective sites (19.77% for knee and 29.18% for hip).
Demographics of 344,242 veterans from the MVP by site-specific OA.
UpSetR plot showing the frequencies of site-specific OA in the Million Veteran Program. Overlap across sites is represented by the dot plot at the bottom, where intersecting occurrence of OA at ≥ 2 sites is connected by a line. Hand, finger, and thumb OA cooccurred very often and, as such, were collapsed. Knee OA was the most common individual subtype, followed by spine and then hip. Electronic medical records showing generalized OA without a specific site designation are considered “unspecified” and constituted the second-largest group. Combinations of cooccurring site OA are shown by other bars on the right. Only the top 16 intersections are shown for space accommodations. OA: osteoarthritis.
As with unspecified OA, mean DCCI was significantly higher for each site-specific OA (2.5-3–fold higher proportion than veterans without OA), and the distribution across DCCI severity subcategories also differed (Supplementary Table S3, available with the online version of this article). Of the 17 comorbidities assessed, most reflected the patterns seen in unspecified OA. Veterans with OA at all sites other than hip (ie, knee and spine) were less likely to have HIV/AIDS than veterans without OA (Supplementary Table S3). The frequency of metastatic solid tumors and OA was higher among veterans with knee and hip OA but not spine OA. For each site-specific type of OA, the proportion of individuals differed across physical demand at work categories (Supplementary Table S3), such that load-bearing OA sites (eg, knee, spine, and hip) had a higher proportion of individuals in the “very light” work demand category than those without. Veterans with OA at any site generally reported lower physical fitness than those without OA.
For knee and hip OA, generally fewer veterans had been in reserves and more had an active service history than controls. In comparison to veterans without OA, rates of prior Army and Marine Corps service were higher across all site OA subtypes (Table 5). Rates of prior Navy service, as well as service across the category indicating multiple military branches, were lower for all OA subtypes. Knee OA frequency only was lower in veterans who had served in the Coast Guard, and there was a lower proportion of veterans with prior Air Force service for all OA sites. Missing military branch/service type data are available in the footnotes of Table 5.
Prior military service among veterans with site-specific OA from the MVP.
Across all joint sites, veterans with OA were less likely to have served in a military era after 1990 (Table 5). Veterans of the Vietnam War and the following era tended to have higher frequency of OA at all sites. Other eras presented more variable patterns in site-specific OA (Table 5). Briefly, veterans with spine OA were more likely to have served in the Persian Gulf War (25.58%) and WWII (1.26%). Veterans with hip OA were more likely to have served in the Korean War (8.91%) and the following era, whereas other types of site-specific OA tended to be less frequent for those eras. Hand OA was less frequent in those that served during the Korean War. Knee, spine, and hand OA frequencies were lower in veterans who served in the era immediately after WWII. For those with service histories in November 1941 or earlier, OA frequency was higher for knee and hip OA.
Site-specific OA: adjusted estimates. The association between comorbidity burden and risk of OA at all joint sites was upheld in multivariable-adjusted models (Table 3). Prior Army service, independent of all other factors, was associated with ORs of 1.2 (knee OA), 1.1 (hip OA), and 1.3 (spine OA, hand OA). Prior Marine Corps service was also independently associated with ORs indicating elevated OA risk across all sites (OR 1.3 for all sites). However, the multivariable-adjusted model reversed the apparent protective effect of prior Navy service for spine and hand OA (both OR 1.1) and eliminated its significance altogether for hip and knee OA. No significance was found for knee OA frequency in Coast Guard veterans after adjustments in the multivariable models, although hip (OR 1.3) and hand (OR 1.5) OA were more common in Coast Guard veterans. Additionally, negative associations with prior Air Force service did not persist in the multivariable-adjusted analysis. In fact, hand OA reached significance in the opposite direction (OR 1.1). See Supplementary Tables S4-S7 (available with the online version of this article) for detailed results of multivariable-adjusted models.
DISCUSSION
OA is a growing public health concern21 exacerbated by a rapidly aging population. Many patients with OA have completed military service and may have faced unique environmental exposures, physical challenges, and other triggers.10 In our present research, we studied the frequency of OA across joint sites within different demographic and other groups in veterans in the MVP. Our findings support the association of sex and ancestry with OA in the veteran population; specifically, female and Black or African American and American Indian or Alaska Native veterans in MVP were more likely to have OA, in keeping with prior reports.22 Further, OA is generally accompanied by higher comorbidity burden across multiple physiological systems.
The MVP cohort provides the unique and valuable opportunity to report on the relationship between military service and OA frequency in later life.15 In this cohort of veterans, we provide evidence that OA may differentially affect certain military branches and eras, with prior service showing a protective effect in some cases and a deleterious effect in others. Veterans with unspecified OA as well as knee and spine OA were more likely to have served in the Army, Marine Corps, or National Guard compared to other branches. The higher relative frequency of OA in the Army, Marine, and National Guard may be due to various exposures that make these veterans more prone to developing OA.23,24 The higher rates of knee and spine OA in these groups may be a product of the more intense physical strain on these joints during service and possibly higher rates of back or limb injury. A limitation of our study is that, despite PTOA being an exclusion criterion, we may not have captured all veterans who developed OA after joint trauma. Active-duty service members in the Army and Marines have a higher rate of injuries and physical trauma compared to those in the Navy and Air Force,25 which may partially contribute to the development of OA.
Multivariable-adjusted model results showed that many service types that appeared partially protective against OA in unadjusted models were not independently significant. This highlights the need to establish a more nuanced understanding of contributors to the relationships across demographics such as age, sex, ancestry, and joint health after service. However, caution should be taken when comparing frequencies across military branches or extrapolating to the general population, as our study did not perform this direct comparison.
Likewise, the differential OA frequencies across military eras are interesting and present potentially novel insights. By necessity, this implicates the known influence of age, as older adults would have served in earlier eras than their younger counterparts. As described, age was integrated as a covariate in all multivariable-adjusted analyses. The age difference, although statistically significant, was 2 years between those with and without unspecified OA (the largest OA group) and –1 to 5 years for the site-specific OA. This may partially account for the generally lower frequencies of site-specific OA in more recent war eras vs higher rates in preceding eras such as the Vietnam War (notably, this era showed the highest proportions of overall service, likely due to implementation of the draft lottery). It is interesting to consider how different site-specific OA subtypes manifest by era, as this also implicates variables outside of the scope of typical examination, such as war strategy, weapon use, fatigue design/weight distribution, and training regimen. For example, spine OA is the most prevalent subtype (based on proportion alone) for those that served between 1990 and 2001. Continued research into the relationship between military era and chronic diseases may provide guidance toward establishing practices that protect joint health and quality of life in veterans.
Rates of previous arthroplasty specific to the affected joint were higher in those with OA than in those without for both hip and knee arthroplasties, supporting observations that OA often manifests bilaterally. This may be due to mechanical (eg, overcompensation, overuse) and/or metabolic (eg, inflammatory burden) influences, all of which warrant continued investigation.7,26 This also highlights the importance of appropriately personalized and progressive rehabilitation, as the individual’s unique joint replacement history, gait, and biomechanical patterns, among other factors, likely converge to affect the success of rehabilitation following joint replacement. In support of this, a recent report suggested that poor recovery from the initial arthroplasty procedure was a major determinant in the decision to cancel a replacement of the contralateral joint, even when medically necessary, in individuals without military background.27
We found that veterans with OA tended to have jobs that demanded lower levels of physical activity and reported lower overall fitness, but this may be a consequence of OA rather than a cause or determinant. Unfortunately, the progressive pain and disability that accompanies OA often leads to a vicious cycle wherein individuals may reduce daily activity as a strategy to mitigate pain, and the molecular consequences of reduced activity include further muscle weakness, deconditioning, and increased joint pain.7 The rehabilitation field has focused research efforts toward interventions that may influence this decline, including blood flow restriction exercise, but continued research is highly warranted.28
Our study leveraged a large sample size in a richly phenotyped dataset ripe for OA investigation. A limitation of the present work is that OA was identified using electronic health records and not radiographic assessment, which may lead to misclassification bias. We attempted to mitigate this bias by using a previously published algorithm by Zengini et al17 and requiring the presence of ≥ 2 ICD codes ≥ 30 days apart to establish the presence of OA, which has been recently shown to have high positive predictive value and specificity at the cost of sensitivity.19 Propensity matching was considered, but since this is not a pharmacoepidemiology study comparing the effect of various treatments, which is usually adjusted by including the treatment propensity, a propensity-matched design does not confer a significant advantage in the current scenario.
Although the unique military service profile of this large cohort focus has implications for ongoing and future military operations as well as veteran health, care should be taken when generalizing findings. First, participation was voluntary, meaning that not all veterans were included. Second, this study was, by design, not a population-level epidemiological investigation and should not be extrapolated to the public. A major reason for this is the overall different demographic composition of this subset of veterans, especially regarding race, ethnicity, and sex. For instance, the proportion of female individuals in the present dataset is considerably low (< 10%), especially considering that OA is generally more prevalent in female individuals. Insight into OA associations may have been obscured in the present dataset due to this low number, particularly when veterans with OA were subdivided into joint sites. As such, continued investigation is necessary to continue to establish the epidemiology of OA in a population more balanced for sex and ancestry.
Care should also be taken not to generalize findings to other types of OA, such as that affecting other joint types. For example, foot and ankle OA were not assessed in this study (as we modeled the design after the GO consortium)18 in order to provide a sense of symmetry to enable future cross-interpretation. Further, there were low numbers of foot and ankle OA, given their relatively low prevalence and association with acute trauma.29 However, this represents a valuable future direction in larger cohorts, especially as foot and ankle OA remains a relatively underrepresented area of study in OA research.30 Additionally, findings should not be generalized to PTOA, as this condition may be etiologically distinct from the idiopathic OA studied herein. However, it is possible that some of the OA cases described herein may be partially attributable to past injury during service; thus, the current results may overestimate the frequency of idiopathic OA. Ideally, investigation of PTOA would require information accounting for OA and a known precipitating injury or ligament issue. This is an area of active research for our team in conjunction with the GO consortium.31
The finding that prior military service is associated with differential rates of OA is useful; however, it is difficult to disentangle whether this is causative, or whether these veterans may have been drawn to and selected for active service partially because of their physiological resilience, or “grit.”32,33 Indeed, it is of high interest in military selection assessments to identify veterans who display advantageous physical and psychological traits such as higher grit and lower susceptibility to overload injury during basic training and subsequent deployment, both for their own health and the stability and success of future operations.34,35 Continued examination of the veteran population is likely to provide critical insight into quality of life following service and may inform existing military protocols in areas of concern.
With a rapidly aging population structure, accompanied by a shift toward less active lifestyle choices, rates of chronic disease such as OA are expected to rise. Rehabilitation of OA remains highly variable and is too often incompletely successful, and the diagnosis of OA is based on chronic pain and loss of normal function. It remains paramount to establish better means of identifying, prehabilitating, and rehabilitating OA in veterans with a wide array of demographic characteristics and personal histories. We provide evidence that these factors, along with joint site specificity, need to be considered in greater detail for the design of a precision medicine–driven approach to managing OA. Valuable future insight from continued examination of genetics and molecular mechanisms of gene expression regulation, along with demographic lifestyle factors, may lead to clarity in diagnosing and managing OA. Future studies should consider the important interplay of these mechanisms in designing strategies for treatment, prehabilitation, and even rehabilitation of patients with OA across a diverse population.
ACKNOWLEDGMENT
The authors acknowledge the use of the US Department of Veterans Affairs (VA) Centers for Medicare and Medicaid Services dataset in this manuscript and thank the members of the VA Million Veteran Program (for the full list of members, see the Supplementary Acknowledgment, available with the online version of this article).
Footnotes
This research is based on data from the Million Veteran Program, Office of Research and Development, Veterans Health Administration, and was supported by award no. I01RX002745. This publication does not represent the views of the Department of Veterans Affairs or the US Government.
M.L.N. McDonald and J.A. Singh contributed equally as co-senior authors.
JAS has received consultant fees from ROMTech, Atheneum, ClearView Healthcare Partners, ACR, Yale, Hulio, Horizon, DINORA, Frictionless Solutions, Schipher, Crealta, Medisys, Fidia, PK MED, Two Labs, Adept Field Solutions, Clinical Care Options, Putnam Associates, Focus Forward, Navigant Consulting, Spherix, MedIQ, Jupiter Life Science, UBM LLC, Trio Health, Medscape, WebMD, Practice Point Communications, and the NIH; has received institutional research support from Zimmer Biomet Holdings; has received food and beverage payments from Intuitive Surgical and Philips Electronics North America; owns stock options in atai Life Sciences, Kintara Therapeutics, Intelligent Biosolutions, Acumen, TPT Global Tech, Vaxart, Aytu BioPharma, Adaptimmune, GeoVax Labs, Pieris, Enzolytics, Seres Therapeutics, Tonix, Abeona, and Charlotte’s Web; has previously owned stock options in Amarin, Viking, and Moderna; serves on the speaker’s bureau for Simply Speaking; was an executive member of OMERACT; serves on the FDA’s Arthritis Advisory Committee; serves as the co-chair of the Veterans Affairs Rheumatology Field Advisory Board; serves as the editor and the Director of the University of Alabama at Birmingham Cochrane Musculoskeletal Group Satellite Center on Network Meta-analysis; and has previously served as a member of the following committees: member of the ACR’s Annual Meeting Planning Committee and Quality of Care Committees; chair of the ACR Meet-the-Professor, Workshop and Study Group Subcommittee; and the co-chair of the ACR Criteria and Response Criteria subcommittee. The remaining authors declare no conflicts of interest relevant to this article.
- Accepted for publication August 5, 2024.
- Copyright © 2024 by the Journal of Rheumatology
REFERENCES
SUPPLEMENTARY DATA
Supplementary material accompanies the online version of this article.
