Epidemiology of Osteoarthritis: Literature Update 2022–2023

Abstract

Purpose of Review:

This review highlights recently published studies on osteoarthritis (OA) epidemiology, including topics related to understudied populations and joints, imaging, and advancements in artificial intelligence (AI) methods.

Recent Findings:

Contemporary research has improved our understanding of the burden of OA in typically understudied regions, including ethnic and racial minorities in high-income countries, the Middle East and North Africa (MENA) and Latin America. Efforts have also been made to explore the burden and risk factors in OA in previously understudied joints, such as the hand, foot, and ankle. Advancements in OA imaging techniques have occurred alongside the developments of AI methods aiming to predict disease phenotypes, progression, and outcomes.

Summary:

Continuing efforts to expand our knowledge around OA in understudied populations will allow for the creation of targeted and specific interventions and inform policy changes aimed at reducing disease burden in these groups. The burden and disability associated with OA is notable in understudied joints, warranting further research efforts that may lead to effective therapeutic options. AI methods show promising results of predicting OA phenotypes and progression, which also may encourage the creation of targeted disease modifying OA drugs (DMOADs).

Introduction

As the global age-adjusted prevalence of OA continues to increase (1,2), recent work has attempted to more thoroughly quantify the disease’s impact in previously understudied populations. In this review, we draw attention to studies that investigated OA in the Middle East and North Africa (MENA), Latin America, and among racial and ethnic minority groups in the United States. We also describe literature that focused on OA in typically understudied joints, including the hand, foot, and ankle. Imaging continues to be a vital instrument in quantifying OA progression and determining the success of disease-modifying OA drug (DMOAD) clinical trials. We highlight a few articles describing recent advances in imaging techniques, including the use of imaging in advanced artificial intelligence (AI) methods. While still relatively new, AI methodologies have the potential to redefine both the clinical care of OA and future development of DMOAD.

Article Selection

Given that this paper serves to summarize updates in OA literature, we did not use traditional systematic review methods to determine article inclusion and exclusion criteria. We restricted searching in online databases to articles published since 2022, however, many review articles published in this time frame included individual manuscripts published within the past five years. We describe several such studies in this update to provide a summary of what is known on a specific topic, particularly on OA areas that have received relatively less attention in the research space.

Understudied Populations

Recent estimates suggest that 7.6% of the world’s population (595 million individuals) had OA in 2020 (1). Two recent studies were the first to investigate the prevalence of knee and hip OA specifically in the Middle East and North Africa (MENA) (3,4). The prevalence of both knee and hip OA were found to have increased dramatically, 2.9-fold and 3.1-fold, respectively, over the past three decades in this region (3,4). Another study used a rapid evidence assessment to examine the burden of OA in Latin America (5). The authors concluded that those with knee OA in Mexico reported higher levels of functional disability and worse Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) scores compared to those in Argentina and Brazil, likely in part due to differences in healthcare systems and infrastructure between countries (5). Analyses into the prevalence and disability associated with OA globally are the key first steps to understanding the depth and impact of the disease. Attention will also need to be focused on understanding how public health interventions aimed at preventing OA and mitigating associated disability can be successfully implemented in lower- and middle-income countries. Interventions will need to be tailored specifically to such regions as they face different social determinants of health compared to their higher-income counterparts.

Investigation into the prevalence of OA among racial and ethnic minorities has thus far been relatively lacking. A recent review investigated racial/ethnic, socioeconomic, and geographic disparities in knee and hip OA (6). Some research based in the United States suggests that pain and disability may be higher among African American compared with White individuals, with a meta-analysis reporting a standard mean difference of 0.57 (95% CI, 0.54–0.61) in WOMAC scores (6,7). Additionally, OA-related pain may be more severe among Asian Americans compared to White Americans (6,8). Differences in pain and function between racial and ethnic groups may be partially explained through disparities in depressive symptoms, income and related socioeconomic factors, and involvement in physically demanding occupations (6). A rapid systematic review investigating differences in pharmacologic management of OA among racial and ethnic minorities found that non-selective non-steroidal anti-inflammatory drugs (NSAIDs) were more commonly used among African American compared to White participants with OA (9). Importantly, the use of non-selective NSAIDs increases risk of some toxicities, including cardiovascular and gastroduodenal toxicity. While more expensive than non-selective NSAIDs, COX-2 selective NSAIDs carry somewhat less risk and therefore may be a better treatment option for OA patients with certain comorbidities (9).

Understudied Joints

The knee and hip have historically been the most studied joints in OA research. Explorations into hand OA have been significantly limited despite evidence suggesting a relatively high prevalence (10). A review investigating the prevalence, incidence, and risk factors for hand OA found that prevalence may range from 21% to 92% depending on the age and geographic location of the study cohort (11). Data from the Johnston Country Osteoarthritis Project (JoCo OA), a community-based study of adults 45+ years old, showed higher incidence of radiographic and symptomatic hand OA among White compared to African American participants [radiographic hand OA in ≥1 joint: adjusted odds ratio (OR) (95% CI) White vs. African American=2.26 (1.49–3.43). Symptomatic hand OA in ≥2 joints: adjusted OR (95% CI) White vs. African American=4.29 (2.08–8.86)] (10). Another study using data from the Osteoarthritis Initiative (OAI) and propensity score-matching methods found that African American participants experienced less severe hand OA compared to non-African American individuals (12). They also concluded that African American participants were less likely to experience hand OA in all joints compared to non-African American participants [OR (95% CI): 0.79 (0.66–0.94)] (12). A strength of this study was the use of propensity score-matching variables (including age, sex, comorbidity status, and presence of radiographic knee OA) to match African American participants to non-African Americans in a 1:3 ratio to reduce potential confounding (12).

Globally, a recent study seeking to explore the epidemiology of hand OA in the MENA found that the age-standardized prevalence of hand OA increased 2.7 times between 1990 to 2019, while globally rates have generally decreased (13). Such findings could be due to an increase in OA-related risk factors in MENA countries, including obesity and diet changes (13). This study was limited by the heterogeneity in the definition and method of diagnosing hand OA across MENA, demonstrating the need to standardize such criteria (13).

Foot and ankle OA have also been relatively understudied. In the JoCo OA, development of radiographic ankle OA was associated with male sex, increased BMI, current smoking status, and the presence of symptoms in other joints (14). Roughly a quarter of ankles with a Kellgren-Lawrence (KL) grade of 0 at baseline progressed to a KL grade of at least 1 over three years of follow-up, and previous ankle injury was associated with progression of radiographic ankle OA compared to the absence of injury history [OR=3.4 (95% CI 1.13–10.3)] (14). Evidence from the Clinical Assessment Study of the Foot (CASF) suggests that there are at least three distinct phenotypes to describe foot OA, including 1) no/minimal foot OA, 2) isolated first metatarsophalangeal (MTP) joint OA, and 3) polyarticular foot OA, however additional studies using larger cohorts are needed to complement these findings (15,16).

To address the gap in research efforts directed toward foot and ankle OA, an International Foot and Ankle Osteoarthritis Consortium (IFOAC) of clinicians and researchers was formed in 2019 (17). In 2021, the IFOAC formulated a nineteen item research agenda, focusing on foot and ankle OA diagnosis, epidemiology, burden, outcome assessment, and treatment (17). Specifically, the IFOAC noted the critical need to establish formal clinical definitions and diagnostic criteria for foot and ankle OA (17). Additionally, they described the need for longitudinal studies in diverse populations to better understand the progression of foot and ankle OA over time and to quantify their burden among diverse patient groups (17).

Historically, clinical trials investigating treatments for foot and ankle rheumatic diseases have not used a standard set of outcome measures, making it challenging, if not impossible, to compare effectiveness between trials. The International Outcome Measures in Rheumatology (OMERACT) Initiative was created in 1992 to research and promote the use of standardized outcome measures to be used clinically and in clinical trials investigating treatments for rheumatic diseases (18). In 2018, the OMERACT Foot and Ankle Working Group began investigating a standardized core outcome set (COS) that would be accepted and used specifically in the context of foot and ankle rheumatic diseases (19). The group recently published a systematic review focused on qualitative studies to better understand which outcome domains were important to patients with a foot or ankle rheumatic disease and thus might be included in the standard COS (19). The authors reported that patients are concerned with the change in appearance of their feet/ankles, activity limitations, social isolation, occupational limitations, and financial and emotional distress, among others. Clinical trials investigating treatments for foot and ankle rheumatic diseases could benefit from adapting current outcomes and endpoints to include these noted concerns of patients, including the often underappreciated psychological burden. The OMERACT Foot and Ankle Working Group recently published a protocol describing their process of identifying the outcome domains that will be included in the Core set of Outcome Measures for Foot and Ankle disorders in RheumaTic and musculoskeletal diseases (COMFORT) (18). Next steps in this process will be to identify outcome measures that appropriately capture the agreed upon domains.

Imaging and Artificial Intelligence

Imaging allows for the identification and quantification of structural changes in joints with OA (20). Radiographs remain the standard imaging modality to assess OA progression, either through the use of the KL grading system or through loss in the semi-quantitative joint space width (JSW) (21). JSW loss is the recommended study endpoint for Phase III clinical trials investigating the efficacy of potential disease modifying OA drugs (DMOAD), despite recognized limitations including lack of direct assessment of cartilage damage, challenges around reproducibility, and poor sensitivity to detect change over time (21). Magnetic resonance imaging (MRI), on the other hand, is not typically recommended in clinical settings, but its ability to visualize the entire joint of interest and its high sensitivity to longitudinal changes makes it an appealing option for OA research (21,22).

The availability of large image repositories has spurred an increase in research using artificial intelligence (AI) to study OA diagnosis and outcome prediction (23). A number of reviews (24–28) have described the current state of AI and OA research, so we will only detail a few recent studies here (29–33). A new study using data from the OAI investigated the relationship between MRI-based three-dimensional texture of the infrapatellar fat pad (IPFP) and the future development of knee OA among knees with KL grade ≤1 across 48-months of follow-up (29). The authors found that predictive models using voxel-based quantitative texture features of IPFP were more predictive of knee OA development compared to traditional clinical characteristics and MRI Osteoarthritis Knee Score (MOAKS) imaging markers [Area under the curve (AUC) ≥ 0.75 for IPFP vs. AUC ≤ 0.69 for the traditional clinical models for all timepoints; AUC=0.75 for IPFP vs. AUC=0.50–0.57 with MOAKS at baseline] (29).

Given the high variability in OA disease etiology, symptoms, and progression, recent efforts have focused on classifying the disease into distinct phenotypes based on a combination of observable factors (i.e. risk factors and structural progression) and unobservable characteristics (i.e. genetics and biomarkers) (30,31). Future therapeutics may be developed with specific phenotypes in mind, a process that could lead to further success in DMOAD development. Using OAI data, a recent study used two machine learning techniques, deep embedding clustering (DEC) and multiple factor analysis with clustering (MFAC), in combination with clinical variables, to identify different knee OA phenotypes (32). DEC methods resulted in 5 phenotype groups, while MFAC identified 3 phenotypes. The disease clusters were similar between the two methods: both resulted in a phenotype characterized by higher body mass index (BMI), higher comorbidity burden, and lower levels of physical activity, as well as a grouping characterized by younger and more physically active participants (32). Results from this study were limited by lack of an external validation dataset. The study’s clustering methods were also limited by the OAI dataset’s sample size and narrow diversity of patient characteristics (32). Findings indicate that such clustering methods may be useful in determining participant inclusion criteria for DMOAD clinical trials.

Dunn et al. tested the use of peripheral blood epigenetic biomarker data in elastic net models to predict knee OA progression using data from the Osteoarthritis Biomarkers Consortium (OABC) (33). Of note, this study used two entirely independent cohorts, including one from the Johnston County OA Project and one from a previous OAI methylation dataset for external validation of their tested prediction model. Models using OABC data had robust performance in predicting radiographic progression [AUC: 0.94 ± 0.004] and pain progression [AUC 0.97 ± 0.004] (33). Models used in this study were able to accurately predict future radiographic and/or pain progression between two and five years from single blood samples collected at baseline. The authors also note that prediction remained high when models were reduced to a smaller number of genomic regions, suggesting that more expensive and larger genome sequencing methods may not be necessary to retain accuracy (33). This study’s promising results indicate the viability of more cost-effective methods for accurate OA progression and symptom prediction models, which could be useful tools for determining participant inclusion criteria for DMOAD clinical trials.

Conclusions

The burden and disability associated with OA is clear. New evidence suggests that disparities in terms of OA-related burden, symptoms, and treatments among racial and ethnic minority groups exist, but have yet to be thoroughly quantified. Further work is also needed to understand the specific modifiable risk factors and barriers to care that affect these communities. OA research has also historically been lacking in certain joints, including the hand, foot, and ankle. Ongoing efforts to create standardized diagnostic criteria, clinical definitions, and patient reported outcomes for OA affecting these joints are vital to accurately enumerate their burden and compare findings between potential treatments. Finally, advances in imaging techniques and AI methodologies have been significant in recent years. As AI techniques continue to advance, they will likely play a major role in determining future directions for OA research and therapy development.

Key Points.

• Recent studies have investigated the burden of OA in lower- and middle-income countries, however, further work is needed to understand how existing treatment options integrate into the current healthcare infrastructure in these regions.

• Standardized clinical definitions and diagnostic criteria for many understudied joints, including hand, foot, and ankle OA, have been lacking, and there are notable ongoing efforts to fill these needs.

• Artificial intelligence (AI) techniques are proving valuable in predicting OA progression and outcomes, indicating their potential use in defining participant inclusion criteria in disease-modifying OA drug (DMOAD) clinical trials.