Non-genetic risk factors associated with glenohumeral osteoarthritis-related shoulder arthroplasty Authors: Abigail G Carey-Ewend, Xiaoyi Xu, Vy Pham, Jay D Keener, Elizabeth L Yanik Published: 2026 ABSTRACT AbstractObjectiveGlenohumeral osteoarthritis (OA) is a leading cause of disabling shoulder pain, but little is known about glenohumeral OA risk factors. In the UK Biobank, we identified factors associated with OA-related shoulder arthroplasty (OA SA), as a measure of end-stage glenohumeral OA.DesignThe UK Biobank collects extensive baseline survey data and follow-up through linked hospital records. OA SA was defined as presence of a SA procedure code with an OA diagnosis. Associations between each potential risk factor and OA SA were estimated with multivariable Cox regression. Identified risk factors were also evaluated with OA-related THA/TKA to compare the magnitude of associations across joints.ResultsAmong 494,203 people, there were 652 OA SA cases. Each decade increase in age was associated with more than twice the OA SA risk (Hazard ratio [HR] = 2.38, 95%CI = 2.04–2.76). Higher body mass index (BMI) was strongly associated with higher risk (HR for obese vs. healthy BMI = 3.08, 95%CI = 2.47–3.84). BMI associations were weaker for THA and stronger for TKA. OA SA risk increased consistently with increasing manual work frequency. Vigorous physical activity was only associated with elevated risk at the highest levels. The highest levels of manual work and physical activity were more strongly associated with OA SA than THA/TKA.ConclusionsSimilar factors may drive OA in the shoulder as do the knee and hip, but the magnitude of associations differed by joint. Importantly, some of these factors could be amenable to intervention, such as better job design to reduce upper extremity burdens and weight loss programs to reduce obesity.Keywords: Glenohumeral osteoarthritis, Shoulder arthroplasty, Risk factors, Epidemiology, Obesity, Physical activity, Occupation Introduction To date, a majority of the research investigating risk factors for osteoarthritis (OA) has focused on the knees, hips, and hands [1,2]. OA of the glenohumeral joint also afflicts a significant portion of the population, with some cadaver and imaging studies reporting its presence in 32.8% of patients over 60 years old [3]. In the US, shoulder arthroplasty (SA) rates are growing even more rapidly than hip or knee arthroplasty, which may partly reflect an increasing burden of glenohumeral OA [4]. However, relatively little is known regarding the risk factors that influence development and progression of glenohumeral OA. Given the high prevalence of glenohumeral OA and the substantial disability caused by the resulting upper extremity pain and movement limitation, an understanding of the contributors to glenohumeral OA could have major public health impacts [1]. Numerous risk factors are known to contribute to hip and knee OA risk, including age, sex, obesity, manual labor, and physical activity levels, among others [2,5]. A better understanding of risk factors can aid in identifying high-risk populations. Furthermore, the identification of modifiable risk factors, such as obesity, could guide interventions to reduce the burden of glenohumeral OA. Our study aims to identify risk factors associated with glenohumeral OA-related SA, as a measure of end-stage glenohumeral OA, using the UK Biobank, a 500,000-person population-based cohort in the United Kingdom. We hypothesized that many risk factors for hip and knee OA would also influence glenohumeral OA risk, but that the strength of the association would differ by joint. Methods Study population The UK Biobank cohort enrolled half a million participants between 2006 and 2010 when they were ages 40–69 years [6]. After being consented, surveys collected data on patient socio-demographics, health history, health behaviors, occupation, and health status at a baseline assessment visit [7]. Anthropometric measures were taken, including height and weight for calculation of body mass index (BMI) [6]. In addition, UK Biobank’s ongoing linkage to the UK National Health Service (NHS) allows for updated information regarding hospital admissions, procedures, and diagnoses, and linkage to national death registries provides information on deaths [7]. Sex was based on the sex recorded by the central NHS registry, but in some cases was updated by the participant. For the present study, we excluded participants who self-reported rheumatoid arthritis, psoriatic arthropathy, or systemic lupus erythematosus during their baseline assessment visit, as these conditions may lead to shoulder arthroplasty (SA) unrelated to OA. Participants who had hospital records of diagnoses of rheumatoid arthritis, psoriatic arthritis, or systemic lupus erythematosus or underwent SA procedures before joining the UK Biobank were excluded. Participants who withdrew from the cohort prior to our analyses were also excluded as required by the UK Biobank. Outcome definition The UK NHS database codes patient procedures using the operations, procedures, and interventions (OCPS-4) classification system. Diagnoses are coded using the International Classification of Diseases, 10th revision (ICD-10). We defined OA-related SA as the presence of an OCPS-4 code for a primary SA procedure consistent with OA treatment combined with an OA diagnosis at or before the time of the procedure (hemiarthroplasty, total SA (Table 1)). Reverse SA cases were not included as they could be indicative of a large rotator cuff tear or rotator cuff arthropathy, instead of OA, particularly during the earlier calendar years of 2006–2016 when 59% of the follow-up and 61% of the cases occurred in this study [8,9]. SA cases were not considered OA-related if ICD-10 codes denoted avascular necrosis, or humeral head fractures. Table 1OPCS classification of surgical operations and procedures (4th revision) codes used to identify shoulder arthroplasty procedures consistent with treatment for glenohumeral osteoarthritis.Table 1OPCS codesProcedure descriptionW491, W493, W498, W499, W501, W503, W508, W509, W511, W513 W518, W519Prosthetic replacement of head of humerusW494, W504, W515Resurfacing hemiarthroplasty of head of humerusW961, W963, W964, W968, W969, W971, W973, W974, W978, W979, W981, W983, W985, W988, W989Total prosthetic replacement of shoulder jointO061, O063, O068, O069, O071, O073, O078, O079, O081, O083, O088, O089Hybrid prosthetic replacement of shoulder jointW431, W433, W434, W438, W439 with shoulder codeaTotal prosthetic replacement of joint using cementW441, W443, W444, W448, W449 with shoulder codeaTotal prosthetic replacement of joint not using cementW451, W453, W455, W458, W459 with shoulder codeaTotal prosthetic replacement of joint not elsewhere classifiedW521, W523, W528, W529 with shoulder codeaPrimary prosthetic replacement of articulation of bone using cement not elsewhere classifiedW531, W533, W538, W539, W541, W543, W548, W549 with shoulder codeaProsthetic replacement of articulation of boneW581, W582, W588, W589 with shoulder codeaOther reconstruction of jointO060, O070, O080 with O062, O072, or O082Conversion to hybrid prosthetic replacement of shoulder jointO062, O072, W962, W972, W982 with W490, W500, W510, W580, W582, or W970Conversion to total prosthetic replacement of shoulder jointaShoulder codes: Z691 = Head of humerus, Z813 = Glenohumeral joint, Z814 = Shoulder joint, Z891 = Shoulder not elsewhere classified. OPCS classification of surgical operations and procedures (4th revision) codes used to identify shoulder arthroplasty procedures consistent with treatment for glenohumeral osteoarthritis. Shoulder codes: Z691 = Head of humerus, Z813 = Glenohumeral joint, Z814 = Shoulder joint, Z891 = Shoulder not elsewhere classified. Risk factors of interest Risk factors of interest were chosen a priori, including: age, race/ethnicity, level of education, BMI, sex, smoking status, physical activity (measured in metabolic equivalent tasks minutes per week based on International Physical Activity Questionnaire), frequency of occupational manual/physical labor, income, and area deprivation based on residence as measured by the Townsend Deprivation Index [10,11]. These variables were selected based on risk factors associated with OA at other joints, and socioeconomic factors that may impact one’s likelihood to receive an OA diagnosis and subsequent surgery. Because prior studies have shown that the relationship between physical activity and OA risk may be non-linear, we chose to incorporate physical activity level as a categorical variable, instead of continuous [12]. All risk factors with missing data were categorical (i.e. no missing data for age or Townsend Deprivation Index), and for these variables missing data indicators were included in regression models to keep a consistent analytic population across analyses. Statistical analysis Follow-up for OA-related SA started at the participant’s first UK Biobank assessment visit and ended at the first occurrence of: OA-related SA, first diagnosis of rheumatoid arthritis/psoriatic arthritis/systemic lupus erythematosus, death, loss-to-follow-up, or December 31, 2022 (the last date with available hospital data at the time of analysis). Baseline characteristics were described for participants with and without OA-related SA during follow-up. Cox proportional hazards regression was used to calculate hazard ratios and corresponding 95% confidence intervals as estimates of associations between risk factors and OA-related SA risk. We evaluated the proportional hazards assumption for each risk factor by testing interaction terms with log (time) in Cox models, with P-values<0.10 considered evidence of non-proportional hazards. When evidence of non-proportional hazards was observed, stratified log-log survival plots across time were evaluated and interval-specific hazard ratios were considered in models. Initially, age-adjusted Cox models were run for each candidate risk factor individually as age is the most well-established risk factor for glenohumeral OA and may confound many of the associations with other risk factors. Second, a multivariable Cox model was conducted that included age and all risk factors with evidence of an association with OA-related SA based on age-adjusted models (P < 0.10). Statistical significance was defined as a p value < 0.05. Several sensitivity analyses were conducted to evaluate the robustness of our results. First, prevalent chronic shoulder pain could lead to changes in occupational demands, physical activity, and BMI, potentially inducing associations with OA-related SA that represent reverse causation. As such, sensitivity analyses were conducted that excluded participants who reported three or more months of neck/shoulder pain at the UK Biobank baseline assessment visit. Second, we ran the multivariable Cox model after multiple imputation of missing data. In the primary analysis, missing data indicators were used as this allowed us to retain the full study population while examining the prevalence and patterns of missingness, and as missing data was rare (<3% missingness) for most variables. However, for both income and physical activity missingness was more common and the use of missing data indicators could introduce bias if data is not missing completely at random. Third, as the UK Biobank is an aging population, we conducted a sensitivity analysis using a Fine-Gray sub-distribution model to account for the competing risk of death. To evaluate whether risk factor associations differed based on joint, we conducted analogous analyses to evaluate risk factor associations with OA-related total hip arthroplasty (THA) and OA-related total knee arthroplasty (TKA). THA and TKA codes were identified with the codes used by the UK National Joint Registry. We limited to those cases with an OA diagnosis by the time of the procedure and excluded cases occurring concurrently with a fracture diagnosis. A multivariable Cox regression model was run for OA-related THA and for OA-related TKA, both with the same covariates included as in the final OA-related SA model. All analyses were conducted using SAS v. 9.4 (SAS Institute, Cary, North Carolina, USA). Data visualizations were performed in R using the ggplot2 v3.4.4 package [13]. Ethical approval The NHS Research Ethics Committee approved the UK Biobank. The Washington University Institutional Review Board determined the current study to be exempt from IRB review because it involved use of a limited dataset and execution of a Data Use Agreement. Results Population characteristics There were 502,520 enrolled and consented UK Biobank participants; 164 of whom were excluded because they had withdrawn from the UK Biobank by the time of this study (Fig. 1). A total of 7936 participants were excluded due to a previous history of lupus, rheumatoid arthritis, or psoriatic arthritis. Finally, 217 participants with a history of SA prior to UKB enrollment were excluded. After exclusions, 494,203 participants were included in the analytic sample with a mean follow-up period of 13.7 years (ranging from 1 to 17 years). There were 652 subjects that underwent a primary OA-related SA. In NHS records after UK Biobank enrollment and prior to their SA date, 23 people had a diagnosis of shoulder dislocation or instability, 15 people had a prior diagnosis of humeral head/proximal humerus fracture, and 3 had a prior diagnosis of rotator cuff injury. The median time to shoulder arthroplasty was 7 years (interquartile range: 5–10 years) (Cumulative incidence curve shown in Supplemental Fig. 1). UK Biobank study population selection. Flow chart showing the number of UK Biobank participants excluded when each was exclusion criterion applied and the total number of participants remaining in the final analytic study population. Compared to the 493,511 participants who didn’t undergo OA-related SA, cases were older (mean age 61.6 vs. 56.5 years), more likely to be female (69.6% vs. 54.2%), and more likely to be of white race (96.8% vs. 94.1%) (Table 2). Cases were also less likely to have a college/university education (24.1% vs. 32.2%), reported less income, and were more likely to work physically-demanding jobs (11.6% of cases reported jobs that always involve heavy physical work) (Table 2). Cases had higher BMI (41.7% vs. 24.2% categorized as obese), and fewer cases had never smoked (48.6% vs. 54.6%). Physical activity did not differ significantly between OA-related SA cases and non-cases; over a third of participants in both groups reported no vigorous activity in a week. Table 2Characteristics at the baseline assessment visit for 494,203 people in the UK Biobank with and without osteoarthritis (OA)-related shoulder arthroplasty.Table 2CharacteristicOA-related shoulder arthroplastyP-valueYesNoTotal652493,551Age, median years (IQR)63 (59, 66)58 (50, 63)<0.0001Sex, N (%) Male198 (30.37%)226,254 (45.84%)<0.0001 Female454 (69.63%)267,297 (54.16%)Race/Ethnicitya, N (%) White631 (96.78%)464,283 (94.07%)0.0192 Black<57922 (1.61%) Asian6 (0.92%)11,242 (2.28%) Mixed<52892 (0.59%) Other5 (0.77%)4488 (0.91%)Highest level of educationa, N (%) O levels/GCSES or CSEs94 (14.42%)81,829 (16.58%)<0.0001 A levels20 (3.07%)26,797 (5.43%) NVQ, HND, HNCs75 (11.5%)61,918 (12.55%) Other professional qualifications109 (16.72%)71,087 (14.4%) College/University157 (24.08%)159,086 (32.23%) None of the list173 (26.53%)82,887 (16.79%)Townsend deprivation index, median (IQR)−2.2 (−3.5, 0.4)−2.1 (−3.6, 0.5)0.907Body mass index (BMI)a, N (%) Underweight/Healthy (BMI <25 kg/m2)120 (18.4%)162,550 (32.93%)<0.0001 Overweight (25≤ BMI <30 kg/m2)250 (38.34%)208,666 (42.28%) Obese (30 kg/m2≤ BMI)272 (41.72%)119,326 (24.18%)Smokinga, N (%) Never317 (48.62%)269,263 (54.56%)<0.0001 Former271 (41.56%)169,523 (34.35%) Current53 (8.13%)51,901 (10.52%)MET vigorous activity (mins/week)a, N (%) 0 MET minutes206 (42.47%)161,105 (40.71%)0.0846 1–240 MET minutes61 (12.58%)54,340 (13.73%) 241–960 MET minutes101 (20.82%)98,214 (24.82%) 961–10080 MET minutes117 (24.12%)82,062 (20.74%)Summed minutes activity categorya, N (%) 0–55 min118 (24.33%)102,635 (25.94%)0.1806 56–100 min117 (24.12%)102,070 (25.79%) 101–180 min125 (25.77%)105,245 (26.6%) 181+ minutes125 (25.77%)85,771 (21.67%)Job involves heavy manual/physical worka, N (%) Never/rarely120 (55.56%)184,201 (64.98%)0.0062 Sometimes53 (24.54%)61,029 (21.53%) Usually18 (8.33%)19,318 (6.82%) Always25 (11.57%)18,910 (6.67%)Annual income in poundsa, N (%) Less than 18,000175 (26.84%)94,584 (19.16%)<0.0001 18,000 to 30,999168 (25.77%)106,144 (21.51%) 31,000 to 51,999101 (15.49%)109,297 (22.15%) 52,000 to 100,00051 (7.82%)85,417 (17.31%) Greater than 100,00017 (2.61%)22,724 (4.6%)P-values are based on chi-squared tests for categorical variables and Wilcoxon rank-sum tests for continuous variables.A level = Advanced level, AS level = Advanced Subsidiary level, CSE = Certificate of Secondary Education, GCSE = General Certificate of Secondary Education, HNC = Higher National Certificate, HND = Higher National Diploma, NVQ = National Vocational Qualification, O level = Ordinary level, MET = metabolic equivalent task, IQR = Interquartile Range.aThere are 2729 participants with missing values for race, 9971 participants with missing values for education, 3019 participants with missing values for BMI, and 2875 participants with missing values for smoking. There are 97,997 participants with missing values in “MET vigorous activity (mins/week)” and “Summed minutes activity category”. There are 210,529 participants who were retired or unemployed at the baseline assessment center visit and so did not have a response for “Job involves heavy manual or physical work”. There are 75,525 participants with missing values for income. Characteristics at the baseline assessment visit for 494,203 people in the UK Biobank with and without osteoarthritis (OA)-related shoulder arthroplasty. P-values are based on chi-squared tests for categorical variables and Wilcoxon rank-sum tests for continuous variables. A level = Advanced level, AS level = Advanced Subsidiary level, CSE = Certificate of Secondary Education, GCSE = General Certificate of Secondary Education, HNC = Higher National Certificate, HND = Higher National Diploma, NVQ = National Vocational Qualification, O level = Ordinary level, MET = metabolic equivalent task, IQR = Interquartile Range. There are 2729 participants with missing values for race, 9971 participants with missing values for education, 3019 participants with missing values for BMI, and 2875 participants with missing values for smoking. There are 97,997 participants with missing values in “MET vigorous activity (mins/week)” and “Summed minutes activity category”. There are 210,529 participants who were retired or unemployed at the baseline assessment center visit and so did not have a response for “Job involves heavy manual or physical work”. There are 75,525 participants with missing values for income. Risk factor associations In age-adjusted Cox regression analyses, increasing age was associated with 2.7 times the risk of OA-related SA per decade (HR 2.71, 95% CI 2.40–3.06), and men had half the risk of OA-related SA compared to women (HR 0.50, 95% CI 0.43–0.59; Table 3). Lower levels of education were significantly associated with OA-related SA risk compared to those with a college or university education. Participants making higher incomes had lower OA-related SA risk compared to those in the lowest income category. Compared to participants with underweight or healthy BMIs, those with an overweight BMI had 46% higher risk of OA-related SA (HR 1.46, 95% CI 1.18–1.82), and those with an obese BMI had nearly three times the risk (HR 2.89, 95% CI 2.33–3.58, Table 3). In looking at differences in occupational manual labor, those who reported higher levels of heavy physical labor in their job were found to be at increased risk for OA-related SA (HR 2.15, 95% CI 1.40–3.31, for those reporting jobs that ‘always’ compared to ‘never or rarely’ involved heavy physical labor). In the evaluation of proportional hazards, only age and vigorous physical activity showed evidence of non-proportional hazards across time. Stratified log-log survival plots indicated that older age consistently increased risk throughout follow-up, but that differences in risk were largest at earlier follow-up, when the population was younger (Supplemental Fig. 2A and 2C). By contrast, stratified log-log survival plots indicated that associations with vigorous physical activity were strongest 2–5 years after the baseline assessment visit (Supplemental Fig. 2B and 2D). In multivariable analyses with time-interval specific hazard ratios, age was associated with OA-related SA risk across time, but the association was stronger during the first six years of follow-up (Supplemental Table 1). By contrast, the highest levels of vigorous physical activity were only significantly associated with OA-related SA risk after the first 5 years of follow-up (Supplemental Table 1). Table 3Associations between potential risk factors and osteoarthritis-related shoulder arthroplasty in 494,203 UK Biobank participants.Table 3Age-adjusted Cox regressionMultivariate Cox regressionHR (95% CI)P-valueHR (95% CI)P-valueAge at enrollmentper decade2.71 (2.40, 3.06)<0.00012.38 (2.04, 2.76)<0.0001SexFemaleReferenceReferenceMale0.50 (0.43, 0.59)<0.00010.46 (0.38, 0.54)<0.0001Race/EthnicityWhiteReferenceReferenceBlack0.28 (0.07, 1.13)0.07390.21 (0.05, 0.84)0.0271Asian0.53 (0.24, 1.18)0.1180.56 (0.25, 1.25)0.1566Mixed1.17 (0.38, 3.65)0.78321.05 (0.34, 3.26)0.9348Other1.20 (0.50, 2.89)0.68551.03 (0.43, 2.50)0.9438Highest level of educationCollege/UniversityReferenceReferenceO levels/GCSES or CSEs1.08 (0.83, 1.39)0.57310.82 (0.63, 1.07)0.1457A levels0.75 (0.47, 1.19)0.21480.65 (0.40, 1.03)0.0658NVQ, HND, HNCs1.26 (0.96, 1.66)0.10191.07 (0.81, 1.42)0.6395Other professional qualifications1.31 (1.03, 1.68)0.02971.08 (0.84, 1.38)0.5604None of the list1.39 (1.11, 1.73)0.00360.98 (0.77, 1.25)0.8854Townsend deprivation index1.01 (0.99, 1.04)0.2593––Body mass index (BMI)Underweight or healthyReferenceReferenceOverweight1.46 (1.18, 1.82)0.00061.63 (1.31, 2.03)<0.0001Obese2.89 (2.33, 3.58)<0.00013.08 (2.47, 3.84)<0.0001SmokingNeverReferenceReferenceFormer1.16 (0.99, 1.36)0.07641.18 (1.00, 1.39)0.0485Current1.02 (0.76, 1.37)0.88471.10 (0.82, 1.47)0.5400MET vigorous activity (mins/week)0 MET minutesReferenceReference1-240 MET minutes0.91 (0.68, 1.21)0.50981.02 (0.76, 1.35)0.9165241-960 MET minutes0.87 (0.69, 1.11)0.25470.98 (0.77, 1.24)0.8590961-10080 MET minutes1.25 (0.99, 1.56)0.05811.45 (1.15, 1.82)0.0018Summed minutes activity category0–55 minReferenceReference56–100 min0.99 (0.77, 1.28)0.9633101–180 min1.00 (0.78, 1.28)0.9773181+ minutes1.16 (0.90, 1.49)0.2537Job involves heavy manual or physical workNever/rarelyReferenceReferenceSometimes1.32 (0.95, 1.82)0.09711.25 (0.90, 1.73)0.1845Usually1.42 (0.87, 2.33)0.16541.37 (0.83, 2.27)0.2199Always2.15 (1.40, 3.31)0.00051.99 (1.27, 3.10)0.0026Retired/Unemployed1.64 (1.31, 2.06)<0.00011.39 (1.10, 1.77)0.0065Annual income in poundsLess than 18,000ReferenceReference18,000 to 30,9990.94 (0.76, 1.16)0.56081.07 (0.86, 1.33)0.575331,000 to 51,9990.71 (0.55, 0.91)0.00720.90 (0.69, 1.17)0.438152,000 to 100,0000.56 (0.40, 0.77)0.00040.79 (0.56, 1.12)0.1907Greater than 100,0000.73 (0.44, 1.21)0.22081.12 (0.66, 1.89)0.6834Age-adjusted models included each covariate individually in an age-adjusted model. The multivariate Cox regression model included all the covariates listed with results in that column (covariates with P-value <0.10 in age-adjusted models).HR = hazard ratio, 95%CI = 95% confidence intervals, A level = Advanced level, AS level = Advanced Subsidiary level, CSE = Certificate of Secondary Education, GCSE = General Certificate of Secondary Education, HNC = Higher National Certificate, HND = Higher National Diploma, NVQ = National Vocational Qualification, O level = Ordinary level, MET = metabolic equivalent tasks. Associations between potential risk factors and osteoarthritis-related shoulder arthroplasty in 494,203 UK Biobank participants. Age-adjusted models included each covariate individually in an age-adjusted model. The multivariate Cox regression model included all the covariates listed with results in that column (covariates with P-value <0.10 in age-adjusted models). HR = hazard ratio, 95%CI = 95% confidence intervals, A level = Advanced level, AS level = Advanced Subsidiary level, CSE = Certificate of Secondary Education, GCSE = General Certificate of Secondary Education, HNC = Higher National Certificate, HND = Higher National Diploma, NVQ = National Vocational Qualification, O level = Ordinary level, MET = metabolic equivalent tasks. In the final multivariable model including all risk factors with P-value<0.10 in age-adjusted analyses, all associations remained significant, except for education and income (Table 3). Additionally, former smoking and high levels of physical activity were statistically significantly associated with OA-related SA in the multivariable model. Former smokers had a 18% higher risk compared to never smokers (HR 1.18, 95% CI 1.00–1.39), while current smokers were not at significantly elevated risk. When compared to individuals with no reported weekly vigorous activity (minutes/week), those with over 961 min/week of vigorous activity had a 45% higher risk of OA-related shoulder arthroplasty (HR 1.45, 95% CI 1.15–1.82). In sub-analyses, excluding individuals reporting ≥3 months of prevalent neck/shoulder pain at baseline, associations were similar, but slightly stronger for physical activity and heavy physical labor (Table 4). The risk of OA-related SA was 56% higher among those reporting the highest vigorous physical activity level (HR 1.56, 95% CI 1.16, 2.11) compared to those reporting no vigorous physical activity. Workers with jobs that ‘always’ involved heavy physical labor had more than twice the risk of OA-related SA (HR 2.36, 95% CI 1.34, 4.14 Table 4) compared to those whose job never or rarely involved heavy physical labor. Table 4Multivariate Cox model associations with osteoarthritis-related shoulder arthroplasty among non-chronic shoulder/neck pain participants without prior shoulder history.Table 4HR (95% CI)P-valueAge at enrollmentPer decade2.32 (1.90, 2.84)<0.0001SexFemaleReferenceMale0.37 (0.29, 0.47)<0.0001Race/EthnicityWhiteReferenceBlack0.20 (0.03, 1.46)0.1131Asian0.38 (0.09, 1.51)0.1684Mixed1.32 (0.33, 5.29)0.6997Other1.17 (0.37, 3.66)0.7897Highest level of educationCollege/UniversityReferenceO levels/GCSES or CSEs0.82 (0.58, 1.15)0.2415A levels0.59 (0.31, 1.10)0.0961NVQ, HND, HNCs1.03 (0.71, 1.51)0.8671Other professional qualifications1.00 (0.72, 1.39)0.9947None of the list0.96 (0.70, 1.32)0.8047Body mass index (BMI)Underweight or healthyReferenceOverweight1.63 (1.23, 2.17)0.0007Obese2.91 (2.18, 3.88)<0.0001SmokingNeverReferenceFormer1.34 (1.07, 1.67)0.0094Current1.40 (0.96, 2.04)0.0843MET vigorous activity (MET mins/week)0Reference1–2400.93 (0.63, 1.38)0.7217241–9601.06 (0.78, 1.45)0.6950961–100801.56 (1.16, 2.11)0.0037Job involves heavy manual or physical workNever/rarelyReferenceSometimes1.18 (0.75, 1.83)0.4747Usually1.24 (0.61, 2.51)0.5475Always2.36 (1.34, 4.14)0.0028Annual income in poundsLess than 18,000Reference18,000 to 30,9991.26 (0.94, 1.68)0.118331,000 to 51,9990.90 (0.62, 1.28)0.546252,000 to 100,0000.90 (0.57, 1.41)0.6400Greater than 100,0001.40 (0.74, 2.67)0.3041The multivariate Cox regression model included all covariates in the table.HR = hazard ratio, 95%CI = 95% confidence intervals, A level = Advanced level, AS level = Advanced Subsidiary level, CSE = Certificate of Secondary Education, GCSE = General Certificate of Secondary Education, HNC = Higher National Certificate, HND = Higher National Diploma, NVQ = National Vocational Qualification, O level = Ordinary level, MET = metabolic equivalent tasks. Multivariate Cox model associations with osteoarthritis-related shoulder arthroplasty among non-chronic shoulder/neck pain participants without prior shoulder history. The multivariate Cox regression model included all covariates in the table. HR = hazard ratio, 95%CI = 95% confidence intervals, A level = Advanced level, AS level = Advanced Subsidiary level, CSE = Certificate of Secondary Education, GCSE = General Certificate of Secondary Education, HNC = Higher National Certificate, HND = Higher National Diploma, NVQ = National Vocational Qualification, O level = Ordinary level, MET = metabolic equivalent tasks. Sensitivity analyses using multiple imputation for missing values and Fine-Gray sub-distribution models to account for the competing risk of death resulted in similar association estimates as the primary analysis and the same significant risk factors were identified with the exception of former smoking (P = 0.06 with multiple imputation and P = 0.07 with sub-distribution model) (Supplemental Tables 2 and 3). Comparison to associations with OA-related TKA and THA We performed multivariable Cox regression analyses that included the same factors as the OA-related SA model in Table 3, but with OA-related TKA (n = 19,691 cases) and THA (n = 17,258 cases) as the outcomes (Fig. 2, Table 5). Of the OA-related SA cases included in our primary analyses, 151 were also OA-related TKA cases and 103 were also OA-related THA cases. Obese BMI showed the largest HR for OA-related TKA (HR 5.02), followed by SA (HR 3.08), and then THA (HR 1.73). For occupational manual labor, association patterns were similar across all joints; but the effect estimates for jobs that ‘always’ require heavy manual labor appeared strongest for the knee (HR 1.66) and shoulder (HR 1.99, Fig. 2). Vigorous physical activity for 241–960 and > 961 MET minutes per week were significantly associated with higher TKA and THA risk. At the highest vigorous physical activity levels (>961 MET minutes per week), the association for OA-related SA appeared significantly stronger than for TKA (HR 1.16) and THA (HR 1.14, Fig. 2). Comparison of associations for body mass index (A), Manual Labor (B), and Physical Activity (C) with OA-related Hip, Knee, and Shoulder Arthroplasty Squares represent hazard ratios and lines represent 95% confidence intervals. The reference category for body mass index was underweight/healthy weight (<25 kg/m2). The reference category for heavy manual labor is job that ‘Never/Rarely’ involves heavy manual/physical work. The reference category for vigorous physical activity is 0 MET minutes/week. BMI = body mass index. Table 5Associations of risk factors with OA-related hip and knee arthroplasty in the UK Biobank.Table 5HR (95%CI)∗THATKAAge at enrollmentPer decade2.18 (2.11–2.24)2.33 (2.26–2.4)SexFemaleRefRefMale0.67 (0.65–0.70)0.81 (0.78–0.84)Race/EthnicityWhiteRefRefBlack0.36 (0.28–0.47)0.75 (0.64–0.88)Asian0.15 (0.11–0.21)1.32 (1.18–1.49)Mixed0.68 (0.50–0.92)0.69 (0.51–0.92)Other0.40 (0.29–0.55)1.01 (0.83–1.23)Highest level of educationCollege/UniversityRefRefO levels/GCSES or CSEs1.01 (0.96–1.06)1.23 (1.17–1.3)A levels0.93 (0.85–1.01)1.03 (0.95–1.12)NVQ, HND, HNCs1.03 (0.97–1.09)1.36 (1.28–1.43)Other professional qualifications1.09 (1.03–1.15)1.29 (1.23–1.36)None of the list0.95 (0.90–1.00)1.35 (1.29–1.42)Body mass indexUnderweight/HealthyRefRefOverweight1.34 (1.28–1.40)2.38 (2.26–2.50)Obese1.73 (1.65–1.81)5.02 (4.77–5.29)SmokingNeverRefRefFormer1.10 (1.06–1.14)1.07 (1.03–1.10)Current1.09 (1.02–1.16)0.82 (0.77–0.88)MET vigorous activity (mins/week)0RefRef1–2400.98 (0.93–1.03)0.94 (0.89–0.99)241–9601.06 (1.02–1.11)1.05 (1.01–1.10)961–100801.14 (1.09–1.20)1.16 (1.11–1.21)Job involves heavy manual/physical workNever/rarelyRefRefSometimes1.10 (1.03–1.17)1.34 (1.26–1.42)Usually1.21 (1.10–1.34)1.53 (1.40–1.68)Always1.24 (1.12–1.38)1.66 (1.51–1.81)Retired/Unemployed0.98 (0.94–1.03)1.09 (1.04–1.14)IncomeLess than 18,000RefRef18,000 to 30,9990.96 (0.77–1.21)1.01 (0.81–1.26)31,000 to 51,9991.00 (0.79–1.28)1.06 (0.84–1.34)52,000 to 100,0000.84 (0.62–1.14)0.61 (0.43–0.87)Greater than 100,0000.72 (0.40–1.30)0.37 (0.15–0.91)The multivariate Cox regression models include all covariates in the table.HR = hazard ratio, 95%CI = 95% confidence intervals, A level = Advanced level, AS level = Advanced Subsidiary level, CSE = Certificate of Secondary Education, GCSE = General Certificate of Secondary Education, HNC=Higher National Certificate, HND = Higher National Diploma, NVQ = National Vocational Qualification, O level = Ordinary level, MET = metabolic equivalent tasks. Associations of risk factors with OA-related hip and knee arthroplasty in the UK Biobank. The multivariate Cox regression models include all covariates in the table. HR = hazard ratio, 95%CI = 95% confidence intervals, A level = Advanced level, AS level = Advanced Subsidiary level, CSE = Certificate of Secondary Education, GCSE = General Certificate of Secondary Education, HNC=Higher National Certificate, HND = Higher National Diploma, NVQ = National Vocational Qualification, O level = Ordinary level, MET = metabolic equivalent tasks. Discussion The increasing rates of SA and the effect of glenohumeral OA on morbidity and quality of life necessitate research to identify risk factors that can point to prevention opportunities [14]. The large UK Biobank population allowed for a detailed study of risk of glenohumeral OA-related SA, as a measure of end-stage glenohumeral OA. We found that older age, female sex, overweight/obese BMI, high levels of vigorous physical activity, and jobs involving consistent heavy manual labor were all associated with higher rates of OA-related SA. Female sex was significantly associated with higher risk of OA-related SA. While this trend matches the higher overall OA rates in female populations, prior studies of sex-specific glenohumeral OA rates have found mixed results [[15], [16], [17]]. Men are more likely to be employed in jobs requiring heavy manual labor, and this may drive higher rates of glenohumeral OA in populations where these jobs are common. In the UK Biobank, heavy manual labor was associated with OA-related SA, but only a small portion of the population held such jobs. Higher BMI was strongly associated with OA-related SA. Risk increased in a dose-response manner across BMI categories with obese individuals having nearly three times the risk of OA-related SA compared to those with a healthy BMI. Although the glenohumeral joint is not considered a weight-bearing joint, increased adiposity leads to a chronic inflammatory state in the body, which can lead to injury to various joints. Similar associations have been reported in literature describing the positive association between adipose tissue and hand OA [18]. A higher body mass might also increase the biomechanical load on the shoulder for some activities, even if it is not considered a weight-bearing joint. Our findings are also supported by prior literature which showed that OA patients with BMI >30 are at higher risk of having SA than those with BMIs between 25 and 30 [5]. In comparing to associations observed for the weight-bearing knee and hip joints, we found that obesity associations with OA-related SA were stronger than those observed for OA-related THA, but weaker than those observed for OA-related TKA. This finding is interesting as BMI is often thought to affect weight bearing joints more than non-weight bearing joints. It has been shown that hip OA is strongly driven by joint morphology and factors such as BMI may be less influential if obesity and inflammation alter risk less in patients with hip dysplasia or hip impingement [19]. Further, the shoulder joint has a higher pre-disposition for instability than the hip joint, and shoulder instability is a known risk factor for glenohumeral OA. Among patients with shoulder instability, elevated BMI has been associated with heightened OA risk [20]. There remains a need for further research into the effects of obesity related stress on the glenohumeral joint. Physical activity was only associated with higher risk of OA-related SA among people self-reporting the highest levels of vigorous physical activity. This association appeared specific to OA-related SA risk 6+ years after UK Biobank enrollment, potentially because in earlier OA-related SA cases subclinical symptoms could have been influencing physical activity at enrollment. We identified the highest levels of vigorous physical activity as participants reporting >961 MET mins/week. As an example, moderate effort cycling (12–13.9 mph) for 120 min per week would be sufficient to reach this cut-off for vigorous activity. The increase seen in the highest activity group may indicate that overexposure to activity and exercise puts the joint at risk of developing OA. Previous literature has noted that athletes who are exposed to repetitive high-level motions and activities have higher prevalence of OA [21]. Similar patterns have been observed for knee and hip OA, both in our study and in the prior literature, in which risk is only notably elevated in people participating in the highest levels of activity [12]. Further studies are needed to define specific higher risk exercise and/or physical activities that represent true risk factors for glenohumeral OA. In a similar vein, those who reported having physically-demanding occupations had higher risk of OA-related SA, with risk increasing in a dose-response manner with increasing frequency of heavy manual labor. While the association between manual labor and OA is well-documented for the weight-bearing joints, there is little research that explores glenohumeral OA and the role of physical demands [16,22,23]. Our findings support a positive association between glenohumeral OA and manual labor and the mechanism may be similar to that involved in weight bearing joints that experience repeated exposure to heavy physical demands. The limitations of this study include the small proportion of non-white participants included in the UK Biobank population [24]. While there was some indication of differences in OA-related SA risk by race, these comparisons were generally underpowered. UKB also lacks relevant radiographic images and patient-reported outcomes, such as pain and function measures, hence our decision to focus on OA-related SA as the primary outcome for this study. Because our outcome was defined by a surgical procedure, results could be influenced by differences in propensity to seek out surgical treatment for OA. However, healthcare access is more uniform in the UK than the US, and we accounted for several socioeconomic factors to mitigate this limitation. Nonetheless, specific physical demands may more heavily influence the decision to pursue shoulder arthroplasty than the decision to pursue hip/knee arthroplasty. This could contribute to differential risk factors associations by joint. The exclusion of reverse shoulder arthroplasty may slightly lower the capture rate of patients with end-stage glenohumeral OA as more complex deformities are sometimes managed with reverse arthroplasty rather than anatomic arthroplasty. During our follow-up period the majority of reverse arthroplasties were performed for diseases pertaining to the rotator cuff or fractures and exclusion of these cases creates a cleaner OA definition. However, during the calendar years covered by our study, the global trend in reverse shoulder arthroplasty utilization was rising rapidly for many indications, including for GHOA [25], [9]. Exclusion of reverse shoulder arthroplasty likely leads to under capture of OA-related SA cases during the later years of our study, particularly for older women as they compose the highest utilizers of this procedure [26]. This may explain why older age was more strongly associated with OA-related SA risk in the earlier years of our study, though associations with sex did not significantly change across time. Finally, the risk factors evaluated in this study were collected at a single time point. The role of certain factors, such as manual labor and physical activity, may be better characterized if measured over time and focused on specific activities that directly involve the shoulder. However, these single time-point measures collected in UKB have been shown to strongly correlate with long-term measures [27,28]. The strengths of this study include the large population size and the depth of collection of many risk factors. Additionally, as the UK healthcare system is a public healthcare system, linked NHS hospital data allows a comprehensive capture of SA procedures. This feature of the UK health system also means that differences in surgical rates are less likely to be driven by healthcare access disparities. Another benefit of the UK Biobank is the longevity of the study, which has over 13 follow-up years of linked hospital records on most participants. In conclusion, we found that female sex, obesity, high levels of physical activity, and physically-demanding occupations were all associated with increased risk of OA-related SA. While the large population of participants in the UK biobank gave our study power, there is room for further analysis to explore how changes in occupational physical requirements over time may affect risk of OA and subsequent SA. Our findings can motivate implementation of prevention efforts for those at higher risk of developing OA and requiring surgery, especially programs targeting weight loss and workplace interventions for those experiencing a high degree of glenohumeral repetitive motions. Author contributions AGC, JDK, and ELY were responsible for the conception and design of the study. XX and VP were responsible for statistical analysis of the data. All authors were involved in interpretation of the results. ACG and ELY were responsible for drafting the article. All authors were involved in critical revision of the article and provided final approval of the article. ELY obtained funding for the study. All authors agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy and integrity of any part of the work are appropriately investigated and resolved. Declaration of generative AI and AI-assisted technologies in the writing process Generative AI and AI-assisted technologies were NOT used in the preparation of this work. Funding This work was supported by National Institutes of Health/NIAMS grant R01AR083937 (PI: Yanik) and National Institutes of Health/NIA grant R03AG088746 (PI: Yanik). The study sponsors played no role in the study design, collection, analysis and interpretation of data, writing of the manuscript, or decision to submit the manuscript for publication. Conflict of interest JDK received royalties from Stryker and Shoulder Innovations, was a paid consultant for Stryker, has equity in Genesis, has received research funding from Stryker and the U.S. National Institutes of Health, and is on the editorial board of the Journal of Shoulder and Elbow Surgery and Seminars in Arthroplasty: JSES. ELY has received research funding from the U.S. National Institutes of Health, the Department of Defense, and the Scoliosis Research Society. AGC, XX, and VP have nothing to disclose. Acknowledgements This research has been conducted using the UK Biobank Resource (http://www.ukbiobank.ac.uk/) under Application 27034. This work uses data provided by patients and collected by the NHS as part of their care and support, and data assets made available by National Safe Haven as part of the Data and Connectivity National Core Study Programme. Handling Editor: Professor H Madry Footnotes Supplementary data to this article can be found online at https://doi.org/10.1016/j.ocarto.2026.100865. Contributor Information Abigail G. Carey-Ewend, Email: a.carey-ewend@wustl.edu. Xiaoyi Xu, Email: lexyxu@mednet.ucla.edu. Vy Pham, Email: vypham@wustl.edu. Jay D. Keener, Email: keenerj@wustl.edu. Elizabeth L. Yanik, Email: yanike@wustl.edu. Supplementary data The following is the Supplementary data to this article. Multimedia component 1 References Associated Data Supplementary Materials Multimedia component 1