Abstract
Objective
Patients with osteoarthritis (OA) affecting multiple joints experience greater pain than those with single-joint disease, yet most research examines isolated joints, leaving a gap in multi-joint disease. This study aimed to describe radiographic hip (rHOA) and knee OA (rKOA) within UK Biobank (UKB), exploring interrelationships across joints and associations with joint pain, body size, race and deprivation.
Design
This cross-sectional study applied machine learning to hip and knee dual-energy X-ray absorptiometry scans. Radiographic OA (rOA) was defined as custom grades ≥2. Joint pain was assessed through self-reported questionnaires. Logistic regression models examined bilateral and cross-joint associations, as well as associations with joint pain. Adjustments were made for age, sex, race, height, weight and deprivation.
Results
Among 59,475 individuals (mean age 65 years; range 45–85; SD 7.7; 52.8% female), rHOA prevalence was 4,098 (6.9%, right-side) and 4,841 (8.1%, left-side). The corresponding estimates for rKOA were 3,750 (6.3%) and 4,220 (7.1%). Overall, increasing grades of rOA and number of joints affected were more strongly associated with concurrent joint pain (four-joints: OR 4.3 [95%CI 2.6–7.2]). Regarding joint-interrelationships, bilateral associations were stronger at the knee (26.1 [24.1–28.2]) than the hip (5.6 [5.2–6.1]). Cross-joint associations (hip-knee) were weaker. BMI was more strongly associated with rKOA (1.57 [1.52–1.61]) than rHOA (1.05 [1.02–1.09]). Greater height was positively associated with rHOA but appeared protective for rKOA.
Conclusions
Radiographic hip and knee OA exhibit distinct patterns of interrelationship, associations with symptoms and risk factors, suggesting heterogeneity in disease processes and the need for joint-specific treatment.
Keywords: Osteoarthritis, Epidemiology, Multi-joint, Pain, Deprivation
Introduction
Osteoarthritis (OA) is a painful, chronic condition that limits mobility and activities of daily living, contributing substantially to disability worldwide(1). It affects over 565 million people globally, with prevalence projected to rise by 60–100% by 2050, due to an ageing population and rising rates of obesity, the latter of which is estimated to account for approximately 20% of OA cases(2). More broadly, both incident and prevalent OA have a diverse risk factor profile including genetics, joint injury, deprivation and race. For example, prevalence rates are higher in more deprived areas, potentially reflecting other health determinants including obesity(3) and Black-Americans experience a greater burden of both radiographic and symptomatic knee OA(4). However, these studies are based on data from several decades ago(4), meaning up-to-date estimates from contemporary population cohorts are justified.
Historically, lower-limb OA has largely been considered an asymmetric monoarthritic disease; yet OA can affect more than one joint, termed multi-joint OA. Those with multi-joint OA have been shown to experience greater pain and functional impairment than individuals with single-joint disease(5). Complementary research has challenged traditional assumptions that disease onset and progression are driven solely by degeneration of the articular joint surface, proposing instead that OA should be considered a systemic disease, which fits with the concept of multi-joint OA(6). Studying the interrelatedness of OA across joint types within the same individual, where systemic risk factors are shared, may provide important insights into this theory.
However, new insights from genetic studies have suggested that while OA may have systemic components, it is not a single homogenous condition(7). The largest genome-wide association study shows that genetic loci confer risk differentially across joint sites, indicating that OA at different anatomical locations may have distinct underlying biological mechanisms(8). For example, GDF5, a key regulator of joint development and chondrogenesis, is preferentially associated with knee rather than hip OA(8–10). If it were the case that OA at different joints were driven by separate pathways, then systemic risk factors such as obesity might show varying associations across these sites.
UK Biobank (UKB) offers a unique opportunity to investigate the prevalence and predictors of multi-joint OA at scale. In particular, high-resolution dual-energy X-ray absorptiometry (DXA) scans of the hip and knee, which are akin to radiographs, are available(11). Previous studies have demonstrated that DXA-derived measures of rOA are clinically relevant but until now the knee and hip joints have not been studied together(12, 13). In addition, the phenotypic breadth of UKB allows further investigation of important risk factors, like deprivation, BMI and race. Therefore, the aim of this cross-sectional study was to i) undertake a descriptive analysis of radiographically defined prevalent hip and knee OA, ii) assess their associations with joint pain, iii) examine the interrelationship of OA across the hip and knee joints, iv) explore associations with BMI, height and weight, and v) identify any differences according to deprivation and race.
Materials and Methods
Population
UK Biobank is a longitudinal cohort study that aimed to recruit a population of ~500,000 individuals aged 40–69 between 2006–2010, from across the UK. Participants underwent extensive genetic and physical phenotyping and completed questionnaires on socio-demographic, lifestyle and environmental factors(14). Initiated in 2014, the extended imaging study sought to gather medical imaging data, including hip and knee DXA scans, from 100,000 UKB participants. All participants with bilateral hip and knee DXA scans at the time of data extraction (January 2024), were included in this cross-sectional study. UKB obtained ethical approval from the National Information Governance Board for Health and Social Care and Northwest Multi-centre Research Ethics Committee (11/NW/0382). All participants provided informed consent for the collection and use of their data and for follow-up via linkage to electronic healthcare records.
DXA-derived radiographic osteoarthritis
Image Analysis
High-resolution bilateral hip and knee DXA scans were obtained using Lunar iDXA scanners (GE Healthcare, Madison, WI, USA), with participants lying supine. DXA-imaging of other joints such as the hands, shoulders and ankles, were not available. Validated machine-learning algorithms, previously trained on large manually annotated datasets, were applied to these scans to delineate joint shape and to characterise the features of rOA (Figure 1, Supplementary Methods)(12, 13). Measurements included osteophyte area and joint space width, while subchondral sclerosis and cysts were not included because they are infrequently and less reliably visualised on DXA. All images underwent manual review to verify the automated measurements and osteophyte segmentation (Supplementary Methods).
Figure 1 -.

Visual summary of the methodology used in this study
A: Machine learning methods were applied to bilateral hip and knee DXA scans from 59,475 UK Biobank participants. Point-placement was used to derive minimum joint space width (mJSW). Reproduced by kind permission of UK Biobank ©. Osteophytes were identified through shaded colours. Semi-quantitative grades for osteophytes and joint space narrowing were derived from osteophyte area and mJSW and combined to form radiographic osteoarthritis grades.
B: Interrelatedness of OA across joints and associations with joint pain.
C: Associations with BMI, height, weight, deprivation and racial background.
Generation of radiographic osteoarthritis
Previously validated grading systems for hip and knee OA in the UKB were used to derive rOA(12, 13). Osteophytes were semi-quantitatively graded 0–3 for each anatomical joint based on their area (mm2). The hip osteophytes were assigned at the lateral acetabulum, the superolateral femoral head and the inferomedial femoral head(15). Knee osteophytes were graded at the medial and lateral femur and the medial and lateral tibia. Minimum joint-space width measurements were obtained at the superior hip and medial knee compartments. These were used to grade joint-space narrowing from 0–3 as previously published(13, 16). An overall radiographic OA (rOA) score was created for each joint by summing grade scores for osteophytes and joint-space narrowing. Grades of radiographic hip OA (rHOA) were assigned using thresholds: grade 0 (sum = 0), grade 1 (sum = 1), grade 2 (sum = 2–3), grade 3 (sum = 4–6), grade 4 (sum = 7–12). At the knee, the osteophyte score was halved, due to four potential sites for osteophytes, before combining with joint-space narrowing grade to form the rOA score. The following thresholds were used for radiographic knee OA (rKOA): grade 0 (sum = 0), grade 1 (sum = 1), grade 2 (sum: >1–≤3), grade 3 (sum: >3–≤4.5) and grade 4 (sum: >4.5). Unless otherwise stated, rOA refers to grade ≥2 at the specified joint(12, 13).
Demographic and clinical variables
Age, height and weight measures were obtained at the same assessment as the DXA (Supplementary Methods). Race was determined using a self-reported questionnaire, with the responses assigned to six categories (Supplementary Table 1). When used as a covariate, race was incorporated into the analyses as a binary variable (White vs. non-White), due to the low numbers of non-White participants. However, associations between race and rOA were explored using the individual categories, with White as the reference group. The Townsend Deprivation Index (TDI) was used as a proxy for social determinants of health. Higher TDI scores, a postcode-based indicator of socioeconomic deprivation, signify greater deprivation(17). TDI was incorporated as a categorical variable based on pre-determined quintiles of deprivation(18) (Supplementary Table 2), because UKB is known to be less deprived than the general population. Self-reported chronic hip and knee pain were obtained through a questionnaire administered during the imaging visit (Supplementary Methods). Hospital diagnosed OA was derived from ICD-10 codes obtained from inpatient admissions, as previously published(12, 13). Joint pain and hospital diagnosed OA were not side-specific variables. As 207 (23%) of hospital diagnoses exist before DXA, analyses were conducted cross-sectionally.
Statistical analysis
Descriptive statistics were used to summarise the demographic characteristics, presented as means, standard deviations (SD) and ranges for continuous variables and as frequencies for categorical and binary variables. Pearson’s correlation coefficient (r) was used to compare minimum joint-space width at the hips and knees. Logistic regression was used to assess the associations between rOA grades 1–4 and self-reported joint pain, separately for each joint. Similarly, logistic regression was applied to examine the interrelationships of rOA (grades ≥2), between bilateral joints (e.g., with OA in the left hip or knee modelled as the outcome and OA in the corresponding right joint as the exposure) and ipsilateral joints. Logistic regression was used to assess associations of rOA at each joint (present vs absent) with TDI quintiles and race (White vs non-White). Further details on model adjustment and sensitivity analyses are available in Supplementary Methods. Results are reported as odds ratios (ORs) with 95% CIs. All statistical analyses were performed using Stata version 18 (StataCorp, College Station, TX, USA).
Results
Population characteristics
In total, 59,475 individuals (mean age: 65.1 years, SD: 7.7, range: 45–85 years), consisting of 31,414 females (52.8%), were included in this study (Supplementary Figure 1). The mean height was 169.7 cm (range: 124–204) and the mean body weight was 75.2 kg (34–200). Mean BMI was 26.0 kg/m2 (12.8–70.9) with 16.0% having a BMI ≥ 30 kg/m2. The frequency of rHOA grade ≥2 was 4,098 (6.9%) for the right and 4,841 (8.1%) for the left hip. The corresponding frequency of rKOA grade ≥2 was 3,750 (6.3%) and 4,220 (7.1%). Of those with features of rHOA, 1,285 (16.8%) had bilateral involvement. Likewise, 2,112 (36.1%) had bilateral rKOA. Generally, the prevalence of rOA increased with age (Supplementary Table 3).
Among those with rOA affecting at least one joint, single-joint involvement was present in 8,745 (69.6%) individuals, two-joint involvement in 3,369 (26.8%) individuals, three-joint involvement in 394 (3.1%) individuals, and involvement in all four joints was found in 61 (0.5%) individuals. Of the 3,369 participants with two-joint rOA involvement, only 488 (14.5%) had one hip and one knee joint affected (i.e., non-cognate). Radiographic knee involvement was approximately twice as prevalent in females than in males, whereas radiographic hip involvement was twice as prevalent in males. Hospital diagnosed HOA was noted in 899 (1.5%) of individuals, while hospital diagnosed KOA was seen in 2,413 (4.1%) of individuals. Hip pain was self-reported in 4,924 individuals (8.3%), while knee pain was more common in 8,795 individuals (14.8%) (Table 1).
Table 1 -.
Descriptive statistics of study population
| Combined | Male | Female | |
|---|---|---|---|
| Mean [SD, Range] | Mean [SD, Range] | Mean [SD, Range] | |
| Age | 65.1 [7.7, 45–85] | 65.8 [7.8, 46–85] | 64.5 [7.6, 45–85] |
| Height (cm) | 169.7 [9.4, 124–204] | 176.8 [6.6, 140–204] | 163.4 [6.4124–198] |
| Weight (kg) | 75.2 [15.0, 34–200] | 83.0 [13.2, 44–171] | 68.3 [13.1, 34–200] |
| BMI (kg/m 2 ) | 26.0 [4.3, 12.8–70.9] | 26.5 [3.9, 15.0–54.7] | 25.6 [4.7, 12.8–70.9] |
| Frequency [%] | Frequency [%] | Frequency [%] | |
| BMI ≥ 30 kg/m 2 | 9,485 [16.0] | 4,571 [16.3] | 4,914 [15.6] |
| Race | |||
| White | 54,836 [92.2] | 25,945 [92.5] | 28,891 [92.0] |
| Asian | 651 [1.1] | 393 [1.4] | 258 [0.8] |
| Black | 395 [0.7] | 162 [0.6] | 233 [0.7] |
| Mixed | 297 [0.5] | 94 [0.3] | 203 [0.7] |
| Chinese | 154 [0.3] | 59 [0.2] | 95 [0.3] |
| Unknown | 3,142 [5.3] | 1,408 [5.0] | 1,734 [5.5] |
| TDI quintiles | |||
| 1 (least deprived) | 24,999 [42.0] | 12,061 [43.0] | 12938 [41.2] |
| 2 | 14,862 [25.0] | 6,948 [24.8] | 7914 [25.2] |
| 3 | 10,103 [17.0] | 4,671 [16.7] | 5432 [17.3] |
| 4 | 6,585 [11.1] | 3,012 [10.7] | 3573 [11.4] |
| 5 (most deprived) | 2,926 [4.9] | 1,369 [4.9] | 1557 [5.0] |
| Right Hip rOA | 4,098 [6.9] | 2,853 [9.2] | 1,515 [4.8] |
| Left Hip rOA | 4,841 [8.1] | 3,305 [11.8] | 1,536 [4.9] |
| Right Knee rOA | 3,750 [6.3] | 1,215 [4.3] | 2,535 [8.0] |
| Left Knee rOA | 4,220 [7.1] | 1,369 [4.9] | 2,851 [9.1] |
| Hospital diagnosed Hip OA | 899 [1.5] | 376 [1.3] | 523 [1.7] |
| Hospital diagnosed Knee OA | 2,413 [4.1] | 1,261 [4.5] | 1,152 [3.7] |
| Hip Pain | 4,924 [8.3] | 1,795 [6.4] | 3,129 [10.0] |
| Knee Pain | 8,795 [14.8] | 4,098 [14.6] | 4,697 [15.0] |
| Joints Affected | |||
| 0 | 46,906 [78.9] | 21,613 [77.0] | 25293 [80.5] |
| 1 | 8,745 [14.7] | 4,656 [16.6] | 4089 [13.0] |
| 2 | 3,369 [5.7] | 1,585 [5.7] | 1784 [5.7] |
| 3 | 394 [0.67] | 182 [0.7] | 212 [0.7] |
| 4 | 61 [0.1] | 25 [0.1] | 36 [0.1] |
| One Hip and One Knee | 488 [0.8] | 242 [0.4] | 246 [0.4] |
| Bilateral rHOA | 1,285 [2.2] | 938 [3.3] | 347 [1.1] |
| Bilateral rKOA | 2,112 [3.6] | 637 [2.3] | 1,475 [4.7] |
Abbreviations: BMI – body mass index, TDI – Townsend Deprivation Index, OA – osteoarthritis, rHOA – radiographic hip osteoarthritis, rKOA – radiographic knee osteoarthritis, SD – standard deviation
Hospital diagnosed OA was derived from ICD-10 codes obtained from inpatient admissions
Radiographic OA and associations with joint pain
Both the hip and knee joints demonstrated strong positive associations between grade of rOA and self-reported joint pain (Figure 2). Individuals with the most radiographically advanced changes (grade 4) had an 11- to 16-times increase in the odds of reporting chronic hip pain and 9- to 15-times increase in the odds of reporting chronic knee pain, as compared with those with no evidence of rOA in the assessed joint. For those with minimal rKOA (grade 1), there was a 2-times increase in the odds of knee pain, while little evidence was seen for an association between grade 1 rHOA and hip pain (Supplementary Table 4). Across grades 1–3, the knee generally showed a stronger relationship with pain than the hip. Interestingly, rHOA showed moderate associations with knee pain, while rKOA showed no associations with hip pain (Supplementary Table 4).
Figure 2 -.

Logistic regression results for the associations between grades of radiographic osteoarthritis and self-reported joint pain
Adjusted for age, sex, race, height, weight and deprivation
rHOA – radiographic hip osteoarthritis, rKOA – radiographic knee osteoarthritis
In fully adjusted models, a greater number of joints affected by OA was associated with increased odds of experiencing hip and/or knee pain. Each additional OA joint exhibited higher odds of pain (Figure 3). Individuals with all four joints involved had four times the odds of reporting pain compared with those without rOA (one joint: OR of 1.57 [95%CI 1.49–1.66], two joints: 2.58 [2.40–2.78], three joints: 3.22 [2.63–3.95], four joints: 4.29 [2.57–7.16]). Similar associations were seen when looking at knee and hip pain separately (Supplementary Figure 2). For those with knee pain, 75.5% did not have rKOA, likewise for those with hip pain, 77.4% did not have rHOA (Supplementary Table 5).
Figure 3 -.

Logistic regression results for the associations between number of joints affected by radiographic osteoarthritis and self-reported hip and/or knee pain
Adjusted for age, sex, race, height, weight and deprivation
Joint site interrelationships
When comparing the odds of rOA between contralateral and ipsilateral joints, both the hip and knee demonstrated robust bilateral associations (Figure 4, Supplementary Table 6), consistent with the descriptive distribution of joint involvement (Supplementary Table 7). In fully-adjusted models, rHOA in one hip was associated with more than five times higher odds of rHOA in the other hip. Associations at the knees were much larger, with more than a 26-fold increase in the odds of right or left rKOA if the other joint was affected. Cross-joint associations, although considerably weaker, were still present: for individuals with right rHOA, the adjusted odds ratios were 1.25 (95%CI 1.10–1.41) for right rKOA and 1.36 (1.22–1.53) for left rKOA. Similar cross-joint (non-cognate) associations were observed for rHOA when rKOA was present. In sex stratified analyses, males showed stronger bilateral associations at the knees, while females showed stronger bilateral associations at the hip (Supplementary Table 8).
Figure 4 –

Logistic regression results for the odds of radiographic osteoarthritis based on presence at each joint
Adjusted for age, sex, race, height, weight and deprivation
rHOA – radiographic hip osteoarthritis, rKOA – radiographic knee osteoarthritis
Left and right minimum joint-space widths at the medial knee compartments showed a strong positive correlation with each other (r=0.73, 95%CI 0.73–0.74), while at the hips the positive correlation was slightly weaker (r=0.66, 0.66–0.67). In comparison, the different anatomical joints demonstrated weaker correlations (r~0.27–0.30) (Supplementary Figure 3).
Associations with BMI, height and weight
BMI was strongly associated with rKOA (grades ≥2), with fully adjusted odds ratios per one standard deviation increase of 1.57 (95%CI 1.52–1.61) and 1.58 (1.53–1.62) for the right and left knees, respectively. In contrast, BMI was weakly associated with rHOA, with corresponding odds ratios of 1.05 (1.02–1.09) and 1.07 (1.03–1.10), for the right and left hips, respectively. Associations with obesity (BMI ≥30 kg/m2) showed a similar pattern, with the magnitude of association more than twice as large for the knees (right: 2.45 [2.27–2.65], left: 2.44 [2.27–2.62]), compared with the hips (right: 1.14 [1.05–1.24], left: 1.15 [1.06–1.24]). Likewise, rKOA showed stronger associations with weight. Greater height was associated with higher risk of rHOA (right: 1.33 [1.26–1.39], left: 1.27 [1.21–1.33]), but was associated with lower risk of rKOA (right: 0.82 [0.77–0.86], left: 0.83 [0.79–0.87] (Table 2).
Table 2 -.
Logistic regression results for the associations between radiographic osteoarthritis and BMI, obesity, height and weight
| Right rHOA ≥ 2 | Left rHOA ≥ 2 | Right rKOA ≥ 2 | Left rKOA ≥ 2 | |||||
|---|---|---|---|---|---|---|---|---|
| Unadjusted | Adjusted | Unadjusted | Adjusted | Unadjusted | Adjusted | Unadjusted | Adjusted | |
| OR [95% CI] | OR [95% CI] | OR [95% CI] | OR [95% CI] | OR [95% CI] | OR [95% CI] | OR [95% CI] | OR [95% CI] | |
| 1.06 [1.03 – 1.10] | 1.05 [1.02 – 1.09] | 1.10 [1.07 – 1.13] | 1.07 [1.03 – 1.10] | 1.47 1.43 – 1.52] | 1.57 [1.52 – 1.61] | 1.49 [1.45 – 1.53] | 1.58 [1.53 – 1.62] | |
| (Kg/m 2 ) | ||||||||
| 1.09 [1.00 – 1.18] | 1.14 [1.05 – 1.24] | 1.12 [1.04 – 1.21] | 1.15 [1.06 – 1.24] | 2.20 [2.04 – 2.37] | 2.45 [2.27 – 2.65] | 2.20 [2.05 – 2.36] | 2.44 [2.27 – 2.62] | |
| (BMI ≥30) | ||||||||
| 1.26 [1.22 – 1.30] | 1.09 [1.05 – 1.13] | 1.35 [1.32 – 1.39] | 1.10 [1.06 – 1.15] | 1.21 [1.18 – 1.25] | 1.76 [1.69 – 1.82] | 1.23 [1.20 – 1.27] | 1.77 [1.71 – 1.84] | |
| (Kg) | ||||||||
| 1.42 [1.38 – 1.47] | 1.33 [1.26 – 1.39] | 1.55 [1.50 – 1.60] | 1.27 [1.21 – 1.33] | 0.75 [0.73 – 0.78] | 0.82 [0.77 – 0.86] | 0.76 [0.74 – 0.79] | 0.83 [0.79 – 0.87] | |
| (cm) | ||||||||
For BMI and obesity adjustments were made for age, sex, race and deprivation. Height and weight were additionally adjusted for each other.
For BMI, height and weight the ORs are per standard deviation (SD) increase.
BMI – body mass index, SD – standard deviation, OR – odds ratio, CI – confidence interval, P – p-value, rHOA – radiographic hip osteoarthritis, rKOA – radiographic knee osteoarthritis
Associations with deprivation and race
Prevalence of rOA was similar across different quintiles of deprivation and racial groups (Supplementary Table 9). Overall, deprivation (TDI quintiles) showed no consistent association with rOA at the hips or knees. For right rHOA, quintile 2 showed a slightly lower odds compared to quintile 1, the least deprived quintile, with an OR of 0.92 (0.85–1.00). Quintiles 3–5 showed no association with right rHOA (Supplementary Table 10). For left rHOA, the results were similarly null, as were those for both right and left rKOA. Equally, there was little evidence of consistent associations between rHOA and the different racial backgrounds. However, there was some evidence that Asian race is a risk factor for rKOA (right: 1.65 [1.20–2.26], left: 1.55 [1.14–2.10]) (Supplementary Table 11).
Discussion
This cross-sectional study examined rOA of the hips and knees, to characterise joint interrelationships and describe associations with sex, pain, body size and deprivation. Our findings highlight differences between rHOA and rKOA. Overall, at each anatomical joint, increasing radiographic severity was associated with increasingly higher odds of self-reported joint pain. Likewise, a greater number of joints affected also saw increasing associations with pain. Notably, almost a third of patients with rOA experienced multi-joint involvement. Pronounced bilateral associations were observed, particularly at the knee but also at the hip, while cross-joint associations between the hip and knee, despite being weaker, were still present. Established risk factors for OA demonstrated clear joint-specific patterns, with BMI and weight being more strongly associated with rKOA than rHOA. Contrastingly, height was more strongly associated with rHOA but appeared protective in rKOA.
Increasing rOA severity was associated with progressively higher odds of chronic joint pain, replicating smaller studies in UKB(12, 13). These mirror associations seen in other cohorts, where increasing Kellgren and Lawrence (K/L) grade was associated with even greater odds of consistent knee pain. In the Multicentre Osteoarthritis Study, K/L grade 4 was associated with 151 (43–526) times higher odds of frequent knee pain compared to grade 0(19). The corresponding odds ratio in the Framingham cohort was 73 (16–331)(19). Consistent with these associations with joint pain, these rOA measures have also been reported to be predictive of joint replacement(12, 13).
In general, rKOA showed stronger associations with knee pain than rHOA with hip pain. Minimal rKOA (grade 1) was associated with increased odds of knee pain, no such association was seen for grade 1 rHOA. Previous research has shown that patients with early radiographic knee changes can experience pain and functional impairment similar to those with end-stage structural changes, while hip pain may only be strongly associated with more severely degraded joints (K/L ≥3)(20, 21). A cross-sectional study found 26% of rKOA joints (KL ≥2) but only 7% of rHOA joints were painful(22). These differences could reflect the unique biomechanics of each joint. For example, the knee bears greater load and may be more sensitive to subtle changes caused by early disease(23). Another reason might be that hip pain has more frequent non-OA causes (e.g., bursitis or tendinopathy) than knee pain, particularly in women, explaining its higher prevalence in females, despite them having less rHOA than men(24). Nevertheless, our results and others show robust associations with hip pain in those with rHOA ≥2, reinforcing the clinical importance of radiographic changes at the hip(25). Interestingly, rHOA showed modest associations with knee pain, while rKOA did not demonstrate a reciprocal association with hip pain. This aligns with existing understanding that pain originating from the hip is often referred to regions including the knee(26).
The slightly lower occurrence of rKOA compared to rHOA, contrasts with previous studies, including the Johnston County OA Project and the Rotterdam Study, which generally report a higher prevalence of radiographic and symptomatic knee than hip OA(27–29). Similarly, recent UK-based clinical estimates suggest that approximately 5.4 million people are affected by knee OA, compared with 3.2 million with hip OA(30). Given that the prevalence of pain and hospital diagnosed OA at the knee in UKB was double that of hip, this discrepancy likely reflects differences in how early radiographic change presents with symptoms between the joints. Furthermore, it is difficult to make grading systems that are perfectly aligned. For example, our grading approach may have favoured the detection of hip disease, where more weight was given to the presence of osteophytes. The observed sex-differences in the prevalence of rHOA and rKOA are consistent with previous estimates, with rKOA more common among females and rHOA more common among males(24). These differences may be due to variations in joint shape, hormonal factors and pain sensitivity. Obesity is unlikely to account for these differences in our cohort, as obesity prevalence was similar between the sexes (M: 16.3%, F:15.6%).
Robust bilateral associations were observed at both knee and hip joints but were larger at the former. Supporting this finding, a previous study found that 80% of participants with unilateral rKOA progressed to bilateral disease over 12 years(31), while in another study only 40% of patients with unilateral rHOA developed bilateral rHOA over 9 years(32). In contrast, cross-joint associations between the hip and knee were considerably weaker, consistent with an earlier finding that clinically-diagnosed hip OA is only modestly associated with incident knee OA (Hazard ratio (HR) of 1.36) and vice versa for knee OA and incident hip OA (HR 1.66)(33). These findings support studies suggesting the genetic architecture of OA is joint-specific, with distinct loci contributing towards disease risk at different anatomical sites(34). For example, the largest genome wide association study to date, found 151 independent associations for hip and 146 for knee OA(35).
Consistent with knee and hip OA being distinct conditions, we observed different associations between common anthropometric risk factors and rOA at the hip and knee. Research has previously established that high BMI is more strongly associated with clinically-diagnosed knee than hip OA(36), which likely reflects anatomical and pathophysiological differences with the knees being more vulnerable to increased mechanical load and obesity-induced chronic inflammation(23, 37). Previously it has also been demonstrated how the knee and hips vary in terms of susceptibility to occupational risk factors such as shift work and heavy manual work(38). These different risk factor profiles suggest that joint-specific treatment strategies should be considered. Indeed, a recent randomised controlled trial has shown that treatment with the GLP-1 agonist semaglutide, a weight-loss injectable, can significantly reduce pain in knee OA patients(39). These findings have been corroborated in pre-clinical studies, suggesting both weight-dependent and weight-independent mechanisms of action(40). Whether similar benefits extend to the hip remains uncertain, and our data suggests it cannot be assumed that these drugs will work equally well for hip OA. Randomised controlled trials for hip OA are certainly warranted.
The association between hip OA and height suggests shared mechanisms with growth. Genetic studies have identified loci associated with hip OA, height and cartilage thickness, supporting a role for growth-related pathways in the development of hip disease(41, 42). Height-associated loci have also been related to skeletal development and bone morphology(43), which are considered risk factors for HOA. Taller stature may therefore reflect developmental processes that both influence joint morphology and increase susceptibility to hip OA. Evidence regarding height and its association with radiographic and clinical knee OA remains conflicting, with studies identifying taller stature to be associated with both increased and decreased risk(44, 45). In this study, increased height appeared protective of knee OA.
The majority of those with multi-joint OA in this study were bilateral knee or hip OA, which suggests they are distinct conditions. A minority of individuals with OA (7.5%) have non-cognate multi-joint disease (e.g., co-existent hip and knee OA), where the likelihood of reporting hip and/or knee pain increased with each additional joint affected. These individuals represent a more generalized pattern of disease(46). Unfortunately, equivalent imaging is not available in UKB to derive radiographic hand or shoulder OA, preventing the assessment of associations with lower limb rOA. Nevertheless, the involvement of multiple weight-bearing joints has previously been associated with greater pain severity, poorer functional outcomes and increased risk of falls(5, 47, 48). Together, these findings highlight the clinical importance of multi-joint OA as a concept and suggest those patients with higher disease burden merit the consideration of cumulative joint involvement, rather than isolated single-joint pathology.
Although higher incidence of hospital-diagnosed OA has previously been reported among more deprived populations(3), we observed little association between deprivation quintiles and rOA at either joint site. Similarly, there was no strong evidence that race was a risk factor for rHOA. However, Asian race (i.e., Indian, Pakistani and Bangladeshi ancestry) was a risk factor for rKOA, which is supported by recent large-population study that also saw greater prevalence of hospital-diagnosed KOA among South Asian individuals(49). Nonetheless, these analyses were limited by the low numbers of participants with rOA among racial minorities and a greater skew towards less deprived individuals. Notably, inequalities have previously been identified in the provision of joint replacement, with rates lower among less affluent areas, despite greater clinical need(50). Disparities in access to care exist and this study underscores the need to ensure equitable access across all quintiles of deprivation.
A key strength of this study is its large sample size, providing substantial statistical power to these analyses. The use of image-derived radiographic measures has enabled detailed assessment of disease severity and for direct comparison of the hip and knee. Limitations include the cross-sectional nature of this study, meaning unmeasured confounding (e.g., joint injury and occupation) and potential reverse causation prevent causal inference. Only knee and hip joints are imaged in UKB with no hand or shoulder imaging, and symptom data were not side-specific reducing the scope of our multi-joint rOA definitions. Spinal imaging is available in UKB but our automated methods are not optimised for these data and therefore they were beyond the scope of this study. Another limitation is the lack of diversity among the UKB population, limiting generalisability to other populations. Furthermore, only a subset of UKB participants were imaged potentially causing a selection bias. The radiographic OA grading systems were developed separately for the joints and are not necessarily equivalent, warranting caution when comparing prevalence estimates. Nevertheless, they were based on established definitions of rOA. In addition, this study uses DXA scans acquired from supine participants, which are therefore not weight-bearing, potentially limiting the assessment of joint space narrowing. Finally, DXA does not allow for the identification of subchondral sclerosis and cysts. However, the replicated strong clinical associations support the validity of these rOA measures.
In conclusion, this study demonstrates that rOA captures structurally and clinically meaningful disease, while also identifying important heterogeneity between hip and knee OA. Increasing radiographic severity and multi-joint involvement were increasingly associated with pain. Strong bilateral but weaker cross-joint associations, alongside unique relationships with BMI, height, and weight, support the existence of different phenotypes. These findings indicate the importance of tailored approaches in the monitoring and management of OA. Further research is required to explore joint site interrelationships over time and to inform interventions aimed at preventing or managing multi-joint OA. Additionally, research is warranted to elucidate the joint-specific pathomechanisms.
Supplementary Material
Acknowledgements and affiliations
This research has been conducted using the UK Biobank Resource under application number 17295. The authors would like to acknowledge the huge contribution of Dr Monika Frysz and Dr Rhona Beynon in establishing this unprecedented imaging resource in UK Biobank. Both individuals declined the offer of authorship as they have left academia.
Funding sources
BGF is supported by a Wellcome Trust Early Career Award (316390/Z/24/Z) and an Academy of Medical Sciences Starter Grant (SGL030\1057). RE, FS and MJ were supported by a Wellcome Trust collaborative award (209233/Z/17/Z). CL is funded by a Sir Henry Dale Fellowship jointly funded by the Wellcome Trust and the Royal Society (223267/Z/21/Z). NCH is supported by UK Medical Research Council [MC_PC_21003; MC_PC_21001], the National Institute for Health and Care Research [as an NIHR Senior Investigator (NIHR305844), and through the NIHR Southampton Biomedical Research Centre (NIHR203319)]. AEN and LA are supported by NIH P30AR072580.
For the purposes of open access, the authors have applied a CC BY public copyright licence to any Author Accepted Manuscript version arising from this submission.
Footnotes
Declaration of competing interests
Given their role as an Associate Editor of OAC, Amanda Nelson had no involvement in the peer-review of this article and has no access to information regarding its peer-review. Full responsibility for the editorial process for this article was delegated to another journal editor.
Declaration of Generative AI and AI-assisted technologies in the writing process
During the preparation of this work the AH and BGF used ChatGPT-5 in order to refine code to develop the figures and proofread text. After using this tool, AH and BGF reviewed and edited the content as needed and take full responsibility for the content of the publication.
Data statement
All UK Biobank data (project number 17295) will be available via their data showcase shortly after publication.
References
- 1.Hunter DJ, Bierma-Zeinstra S. Osteoarthritis. The Lancet. 2019;393(10182):1745–59. [DOI] [PubMed] [Google Scholar]
- 2.Global, regional, and national burden of osteoarthritis, 1990–2020 and projections to 2050: a systematic analysis for the Global Burden of Disease Study 2021. Lancet Rheumatol. 2023;5(9):e508–e22. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Reyes C, Garcia-Gil M, Elorza JM, Mendez-Boo L, Hermosilla E, Javaid MK, et al. Socio-economic status and the risk of developing hand, hip or knee osteoarthritis: a region-wide ecological study. Osteoarthritis and Cartilage. 2015;23(8):1323–9. [DOI] [PubMed] [Google Scholar]
- 4.Abdullah Y, Olubowale OO, Hackshaw KV. Racial disparities in osteoarthritis: Prevalence, presentation, and management in the United States. Journal of the National Medical Association. 2025;117(1):55–60. [DOI] [PubMed] [Google Scholar]
- 5.Nelson AE. Multiple joint osteoarthritis (MJOA): What’s in a name? Osteoarthritis Cartilage. 2024;32(3):234–40. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Collins KH, Haugen IK, Neogi T, Guilak F. Osteoarthritis as a systemic disease. Nature Reviews Rheumatology. 2026;22(2):105–17. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Boer CG, Hatzikotoulas K, Southam L, Stefánsdóttir L, Zhang Y, Coutinho de Almeida R, et al. Deciphering osteoarthritis genetics across 826,690 individuals from 9 populations. Cell. 2021;184(18):4784–818.e17. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Hatzikotoulas K, Southam L, Stefansdottir L, Boer CG, McDonald M-L, Pett JP, et al. Translational genomics of osteoarthritis in 1,962,069 individuals. Nature. 2025;641(8065):1217–24. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Coveney CR, Maridas D, Chen H, Muthuirulan P, Liu Z, Jagoda E, et al. Complex Regulatory Interactions at GDF5 Shape Joint Morphology and Osteoarthritis Disease Risk. Arthritis Rheumatol. 2025;77(11):1488–502. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Faber BG, Frysz M, Zheng J, Lin H, Flynn KA, Ebsim R, et al. The genetic architecture of hip shape and its role in the development of hip osteoarthritis and fracture. Hum Mol Genet. 2025;34(3):207–17. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Littlejohns TJ, Holliday J, Gibson LM, Garratt S, Oesingmann N, Alfaro-Almagro F, et al. The UK Biobank imaging enhancement of 100,000 participants: rationale, data collection, management and future directions. Nature Communications. 2020;11(1):2624. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Beynon RA, Saunders FR, Ebsim R, Faber BG, Jung M, Gregory JS, et al. A novel classifier of radiographic knee osteoarthritis for use on knee DXA images is predictive of joint replacement in UK Biobank. Rheumatology Advances in Practice. 2025;9(1). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Faber BG, Ebsim R, Saunders FR, Frysz M, Lindner C, Gregory JS, et al. A novel semi-automated classifier of hip osteoarthritis on DXA images shows expected relationships with clinical outcomes in UK Biobank. Rheumatology (Oxford). 2022;61(9):3586–95. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Bycroft C, Freeman C, Petkova D, Band G, Elliott LT, Sharp K, et al. The UK Biobank resource with deep phenotyping and genomic data. Nature. 2018;562(7726):203–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Faber BG, Ebsim R, Saunders FR, Frysz M, Lindner C, Gregory JS, et al. Osteophyte size and location on hip DXA scans are associated with hip pain: findings from a cross sectional study in UK Biobank. Bone. 2021:116146. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Beynon RA, Saunders FR, Ebsim R, Faber BG, Jung M, Gregory JS, et al. A novel classifier of radiographic knee osteoarthritis for use on knee DXA images is predictive of joint replacement in UK Biobank. Rheumatol Adv Pract. 2025;9(1):rkaf009. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Townsend P, Phillimore P, & Beattie A. Health and Deprivation: Inequality and the North. London: Routledge; 1988. 236 p. [Google Scholar]
- 18.Hippisley-Cox J. Cut offs for Townsend quintiles in England and Wales and QResearch 2011 [23/11/25]. Available from: https://www.qresearch.org/media/1056/cut-offs-for-townsend-quintiles-in-england-and-wales-and-qresearch.pdf.
- 19.Neogi T, Felson D, Niu J, Nevitt M, Lewis CE, Aliabadi P, et al. Association between radiographic features of knee osteoarthritis and pain: results from two cohort studies. Bmj. 2009;339:b2844. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Jones LD, Bottomley N, Harris K, Jackson W, Price AJ, Beard DJ. The clinical symptom profile of early radiographic knee arthritis: a pain and function comparison with advanced disease. Knee Surgery, Sports Traumatology, Arthroscopy. 2016;24(1):1795. [DOI] [PubMed] [Google Scholar]
- 21.Iidaka T, Muraki S, Akune T, Oka H, Kodama R, Tanaka S, et al. Prevalence of radiographic hip osteoarthritis and its association with hip pain in Japanese men and women: the ROAD study. Osteoarthritis Cartilage. 2016;24(1):117–23. [DOI] [PubMed] [Google Scholar]
- 22.Pereira D, Severo M, Santos RA, Barros H, Branco J, Lucas R, et al. Knee and hip radiographic osteoarthritis features: differences on pain, function and quality of life. Clinical Rheumatology. 2016;35(6):1555–64. [DOI] [PubMed] [Google Scholar]
- 23.Hall M, van der Esch M, Hinman RS, Peat G, de Zwart A, Quicke JG, et al. How does hip osteoarthritis differ from knee osteoarthritis? Osteoarthritis and Cartilage. 2022;30(1):32–41. [DOI] [PubMed] [Google Scholar]
- 24.Faber BG, Macrae F, Jung M, Zucker BE, Beynon RA, Tobias JH. Sex differences in the radiographic and symptomatic prevalence of knee and hip osteoarthritis. Front Endocrinol (Lausanne). 2024;15:1445468. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Rondas GA, Macri EM, Oei EH, Bierma-Zeinstra SM, Rijkels-Otters HB, Runhaar J. Association between hip pain and radiographic hip osteoarthritis in primary care: the CHECK cohort. Br J Gen Pract. 2022;72(723):e722–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Dibra FF, Prieto HA, Gray CF, Parvataneni HK. Don’t forget the hip! Hip arthritis masquerading as knee pain. Arthroplast Today. 2018;4(1):118–24. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Odding E, Valkenburg HA, Algra D, Vandenouweland FA, Grobbee DE, Hofman A. Associations of radiological osteoarthritis of the hip and knee with locomotor disability in the Rotterdam Study. Ann Rheum Dis. 1998;57(4):203–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Nelson AE, Hu D, Arbeeva L, Alvarez C, Cleveland RJ, Schwartz TA, et al. Point prevalence of hip symptoms, radiographic, and symptomatic OA at five time points: The Johnston County Osteoarthritis Project, 1991–2018. Osteoarthritis and Cartilage Open. 2022;4(2):100251. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Nelson AE, Hu D, Arbeeva L, Alvarez C, Cleveland RJ, Schwartz TA, et al. The Prevalence of Knee Symptoms, Radiographic, and Symptomatic Osteoarthritis at Four Time Points: The Johnston County Osteoarthritis Project, 1999–2018. ACR Open Rheumatol. 2021;3(8):558–65. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Arthritis UK. The State of Musculoskeletal Health 2025. 2025. p. 41.
- 31.Metcalfe AJ, Andersson ML, Goodfellow R, Thorstensson CA. Is knee osteoarthritis a symmetrical disease? Analysis of a 12 year prospective cohort study. BMC Musculoskelet Disord. 2012;13:153. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Vossinakis IC, Georgiades G, Kafidas D, Hartofilakidis G. Unilateral hip osteoarthritis: can we predict the outcome of the other hip? Skeletal Radiology. 2008;37(10):911–6. [DOI] [PubMed] [Google Scholar]
- 33.Prieto-Alhambra D, Judge A, Javaid MK, Cooper C, Diez-Perez A, Arden NK. Incidence and risk factors for clinically diagnosed knee, hip and hand osteoarthritis: influences of age, gender and osteoarthritis affecting other joints. Ann Rheum Dis. 2014;73(9):1659–64. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Zhai G, Huang J. Genetics of osteoarthritis. Best Practice & Research Clinical Rheumatology. 2024;38(4):101972. [DOI] [PubMed] [Google Scholar]
- 35.Hatzikotoulas K, Southam L, Stefansdottir L, Boer CG, McDonald M-L, Pett JP, et al. Translational genomics of osteoarthritis in 1,962,069 individuals. Nature. 2025;641(8065):1217–24. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Reyes C, Leyland KM, Peat G, Cooper C, Arden NK, Prieto-Alhambra D. Association Between Overweight and Obesity and Risk of Clinically Diagnosed Knee, Hip, and Hand Osteoarthritis: A Population-Based Cohort Study. Arthritis Rheumatol. 2016;68(8):1869–75. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Shumnalieva R, Kotov G, Ermencheva P, Monov S. Pathogenic Mechanisms and Therapeutic Approaches in Obesity-Related Knee Osteoarthritis. Biomedicines. 2023;12(1). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Hashmi A, Scott S, Jung M, Meng QJ, Tobias JH, Beynon RA, et al. Associations between work characteristics and osteoarthritis: A cross-sectional study of 285,947 UK Biobank participants. Osteoarthr Cartil Open. 2025;7(1):100565. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Bliddal H, Bays H, Czernichow S, Hemmingsson JU, Hjelmesæth J, Morville TH, et al. Once-Weekly Semaglutide in Persons with Obesity and Knee Osteoarthritis. New England Journal of Medicine. 2024;391(17):1573–83. [DOI] [PubMed] [Google Scholar]
- 40.Qin H, Yu J, Yu H, Zhou C, Yuan D, Wang Z, et al. Semaglutide ameliorates osteoarthritis progression through a weight loss-independent metabolic restoration mechanism. Cell Metab. 2026. [DOI] [PubMed] [Google Scholar]
- 41.Faber BG, Frysz M, Boer CG, Evans DS, Ebsim R, Flynn KA, et al. The identification of distinct protective and susceptibility mechanisms for hip osteoarthritis: findings from a genome-wide association study meta-analysis of minimum joint space width and Mendelian randomisation cluster analyses. EBioMedicine. 2023;95:104759. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Aspden RM. Osteoarthritis: a problem of growth not decay? Rheumatology (Oxford). 2008;47(10):1452–60. [DOI] [PubMed] [Google Scholar]
- 43.Styrkarsdottir U, Stefansson OA, Gunnarsdottir K, Thorleifsson G, Lund SH, Stefansdottir L, et al. GWAS of bone size yields twelve loci that also affect height, BMD, osteoarthritis or fractures. Nature Communications. 2019;10(1):2054. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Welling M, Auvinen J, Lehenkari P, Männikkö M, Karppinen J, Eskola PJ. Association between height and osteoarthritis of the knee and hip: The Northern Finland Birth Cohort 1966 Study. Int J Rheum Dis. 2017;20(9):1095–104. [DOI] [PubMed] [Google Scholar]
- 45.Lee DH, Lee HS, Jang SH, Heu JY, Han K, Lee SW. Decreased Risk of Knee Osteoarthritis with Taller Height in an East Asian Population: A Nationwide Cohort Study. J Clin Med. 2023;13(1). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46.Finney A, Dziedzic KS, Lewis M, Healey E. Multisite peripheral joint pain: a cross-sectional study of prevalence and impact on general health, quality of life, pain intensity and consultation behaviour. BMC Musculoskelet Disord. 2017;18(1):535. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.Pavel RMS, Purza AL, Tit DM, Radu A-F, Iovanovici DC, Vasileva D, et al. Functional Burden and Quality of Life in Hip and Knee Osteoarthritis: A Cross-Sectional Study. Medicina. 2025;61(7):1155. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48.Smith RD, McHugh GA, Quicke JG, Finney A, Lewis M, Dziedzic KS, et al. The relationship between multisite peripheral joint pain and physical activity levels in older adults: A cross-sectional survey. Musculoskeletal Care. 2022;20(2):341–8. [DOI] [PubMed] [Google Scholar]
- 49.Costantini JG, Darbinian JA, Garcia EA, Chandra M, Lo JC. Racial and ethnic variation in the prevalence of knee osteoarthritis among older U.S. adults, including Asian ethnic populations. Osteoarthritis and Cartilage Open. 2026;8(2):100761. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50.Lenguerrand E, Ben-Shlomo Y, Rangan A, Beswick A, Whitehouse MR, Deere K, et al. Inequalities in provision of hip and knee replacement surgery for osteoarthritis by age, sex, and social deprivation in England between 2007–2017: A population-based cohort study of the National Joint Registry. PLoS Med. 2023;20(4):e1004210. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
All UK Biobank data (project number 17295) will be available via their data showcase shortly after publication.
