A long-term prospective cohort study of Australian adults has identified several biological, clinical, and socioeconomic determinants associated with a lower risk of undergoing surgical joint replacement in weight-bearing joints. Analyzing data from the Geelong Osteoporosis Study, researchers tracked 2,882 eligible men and women aged 20 to 96 years over a median follow-up of 16.7 years, spanning an interquartile range from 9.7 to 23.2 years. During this period, 223 participants (7.7 percent) underwent incident hip or knee arthroplasty. Using time-dependent multivariable Cox proportional hazards regression with chronological age as the primary time scale and sex stratification, the investigators found that lower body mass index, lower circulating levels of procollagen type 1 N-terminal propeptide, lower dual-energy X-ray absorptiometry spine bone mineral density, an absence of recent falls, and no history of cancer were each associated with a reduced risk of joint replacement.

Hip and knee replacements represent definitive interventions for advanced joint disease, primarily undertaken to relieve intractable pain and restore functional mobility when conservative options fail. End-stage osteoarthritis is the prevailing clinical indication, characterized by progressive loss of articular cartilage, synovial inflammation, joint stiffness, swelling, and physical limitation. Osteoarthritis affects roughly 528 million individuals worldwide, predominantly those older than 55 years. In Australia, it is the most prevalent arthritic disorder, affecting approximately 2.1 million people. In 2022, osteoarthritis accounted for 19 percent of the total Australian disease burden and 1.2 percent of domestic deaths, with knees and hips representing the principal sites of involvement for about 9 percent of adults. Management imposes an immense economic burden: between 2019 and 2020, Australia spent approximately 3.9 billion Australian dollars managing osteoarthritis, including 2.3 billion Australian dollars specifically on knee and hip arthroplasties. With no curative medical therapies available, identifying modifiable determinants remains critical to reduce surgical demand.

Most prior evidence derives from clinical samples of patients receiving specialized care for severe joint disease, introducing selection bias and limiting generalizability. To provide broader population-based estimates, the investigators utilized prospective data from the Geelong Osteoporosis Study in southeastern Australia. Potential participants were randomly selected from the Commonwealth electoral roll within the Barwon Statistical Division, requiring a minimum of six months of local residency. Initial recruitment enrolled 1,494 women aged 20 to 94 years between 1993 and 1997 with a 77 percent participation rate, and 1,539 men aged 20 to 96 years between 2001 and 2006 with a 67 percent participation rate. The Barwon Health Human Research and Ethics Committee approved the study, and participants provided written informed consent. Exclusions comprised prior joint replacements (44 women, 59 men), migration (one woman), missing joint records (one woman), declining medical record access (39 men), insufficient data (one man), and shoulder replacements (two women, four men), establishing an analytic cohort of 1,446 women and 1,436 men.

Women were reassessed at two, four, six, eight, ten, and fifteen years post-baseline, and men at five and fifteen years; analytic models utilized female data from baseline, ten-year, and fifteen-year visits, and male data from baseline, five-year, and fifteen-year visits. Standing height was measured using a Harpenden stadiometer, weight on electronic scales, body mass index was calculated as kilograms per square meter, and waist circumference was recorded using an anthropometric tape. Bone mineral density of the spine and hip and total body lean mass were evaluated using dual-energy X-ray absorptiometry scanners, transitioning from DPX-L to Prodigy Pro devices with phantom-based cross-calibration. Appendicular lean mass was adjusted for height squared. Fasting venous blood samples yielded concentrations of procollagen type 1 N-terminal propeptide, C-terminal telopeptide of type I collagen, high-sensitivity C-reactive protein, glucose, and serum lipids. To facilitate clinical interpretation of protective associations, body mass index, appendicular lean mass index, and procollagen type 1 N-terminal propeptide concentrations were reverse-coded per one-unit decrease.

Surgical endpoints were identified via linkage with the Barwon Joint Registry and Medicare records, confirmed by full-body dual-energy X-ray absorptiometry scans when available. The outcome was defined as the first primary hip or knee replacement per participant, including bilateral procedures and one partial knee replacement, while revision procedures were excluded, without segregating indications such as osteoarthritis versus fracture. Deaths were ascertained via linkage with the Australian Institute of Health and Welfare National Deaths Index. Incident procedures occurred in 135 women (79 hips, 56 knees) and 88 men (52 hips, 36 knees). Time-dependent Cox proportional hazards regression models updated covariates across waves, using age as the primary time scale and sex stratification. Missing covariate values were imputed using predictive mean matching via the mice package in R. Predictor selection was validated using forward stepwise Cox regression and penalized LASSO regression, which demonstrated consistent variable selection and concordance indices of 0.64, while the multivariable model achieved a concordance index of 0.66.

In the multivariable model, each one-unit reduction in body mass index was associated with a 4 percent lower joint replacement risk (hazard ratio 0.96, 95 percent confidence interval 0.94 to 0.99). Substituting appendicular lean mass index for body mass index due to collinearity showed a similar positive association with surgical hazard. When stratified by joint site, lower body mass index strongly protected against knee replacement (hazard ratio 0.94, 95 percent confidence interval 0.91 to 0.96, p < 0.001), but showed no association with hip replacement (hazard ratio 1.01, 95 percent confidence interval 0.99 to 1.04, p = 0.345). Age-stratified analyses revealed further nuances: among participants aged 40 to 59 years, lower body mass index was associated with reduced surgical hazard (hazard ratio 0.97, confidence interval 0.93 to 1.01), whereas among those aged 60 years and older, lower body mass index was linked to an increased hazard ratio of 1.04 (confidence interval 1.00 to 1.08), suggesting that low body weight in older adults may reflect frailty, sarcopenia, or unintentional weight loss.

Circulating biochemical markers and radiological skeletal metrics provided critical insights into bone metabolism and structural disease progression. Lower levels of procollagen type 1 N-terminal propeptide, a serum bone turnover marker, conferred a 31 percent reduction in joint replacement risk (hazard ratio 0.69, 95 percent confidence interval 0.50 to 0.96). In joint-specific models, this protective association was more pronounced for hip replacement (hazard ratio 0.55, confidence interval 0.29 to 1.03) than knee replacement (hazard ratio 0.76, confidence interval 0.31 to 1.87), reflecting active subchondral bone remodeling. Adjusting for clinical osteoarthritis attenuated the propeptide association to borderline significance (hazard ratio 0.55, confidence interval 0.30 to 1.02). Additionally, lower lumbar spine bone mineral density predicted reduced surgical risk (hazard ratio 0.84, confidence interval 0.77 to 0.92). Higher spine bone mineral density often reflects osteophytes and subchondral sclerosis that artificially elevate dual-energy X-ray absorptiometry readings, indicating that lower readings signaled joints unburdened by advanced structural disease rather than true skeletal protection.

Physical fragility markers also influenced surgical probability. Participants reporting no fall history experienced a reduced risk of joint replacement (hazard ratio 0.74, 95 percent confidence interval 0.55 to 0.99), an association significant for hip arthroplasty but not knee arthroplasty, matching prior fracture predicting hip procedures. The absence of a cancer history was initially associated with a lower surgical risk (hazard ratio 0.66, confidence interval 0.48 to 0.90) in time-dependent Cox regression; however, this association lost statistical significance in a Fine-Gray competing risk model (subdistribution hazard ratio 0.92, confidence interval 0.61 to 1.39, p = 0.70) because higher competing mortality among cancer patients curtailed elective arthroplasty. Lifestyle factors, including smoking, alcohol consumption, and physical activity, showed no independent associations. Although lower dietary calcium intake was initially associated with reduced surgical hazard (hazard ratio 0.74, confidence interval 0.52 to 1.04), this association attenuated after adjusting for calcium supplement use (hazard ratio 0.75, confidence interval 0.53 to 1.06, p = 0.108), pointing to confounding by indication and reverse causality as symptomatic individuals were advised to increase calcium intake.

Socioeconomic status, evaluated through the Index of Relative Socioeconomic Disadvantage and grouped into three analytical tiers, exhibited a counterintuitive positive relationship with surgical risk. Compared with low socioeconomic status, medium and high socioeconomic status were associated with increased joint replacement risk (hazard ratio 1.91, confidence interval 1.04 to 3.50; hazard ratio 1.88, confidence interval 1.13 to 3.15). A significant interaction between sex and high socioeconomic status (hazard ratio 0.48, confidence interval 0.25 to 0.91) indicated that this elevated hazard was substantially more pronounced among men. Absolute incidence rates ranged from 2.76 per 1,000 person-years in low-socioeconomic men to 6.02 in medium-socioeconomic women. Because 60 to 70 percent of Australian joint replacements occur in private hospitals, this pattern reflects private insurance coverage, financial capacity, and surgical affordability rather than greater biological disease severity, highlighting potential unmet surgical needs among disadvantaged populations.

Key strengths of the investigation include its prospective population-based design, long-term follow-up, and dynamic time-dependent modeling of physiological exposures. Limitations involve the modest overall number of surgical events, inability to classify replacements by specific indication or laterality, self-reported lifestyle and comorbidity variables, and geographic concentration in southeastern Australia. Ultimately, joint replacement represents an administrative healthcare-utilization milestone reflecting surgical access and clinical decision-making alongside biological disease progression.

Subject of Research: Longitudinal cohort analysis of biological, lifestyle, and socioeconomic determinants of weight-bearing joint replacement risk

Article Title: Key determinants for lowering the risk of joint replacement in weight-bearing joints: a population-based cohort study

Article References: Suleiman, A., Lotfaliany, M., Williams, L. J., Stuart, A. L., Page, R., Gill, S., Budehal, A. P., & Pasco, J. A. (2026). Key determinants for lowering the risk of joint replacement in weight-bearing joints: a population-based cohort study. Arthroplasty, 8(1), Article 45. https://doi.org/10.1186/s42836-026-00403-9

Image Credits: AI Generated

DOI: 10.1186/s42836-026-00403-9

Keywords: Joint Replacement, Osteoarthritis, Geelong Osteoporosis Study, Body Mass Index, Procollagen Type 1 N-terminal Propeptide, Bone Mineral Density, Socioeconomic Status, Knee Replacement, Hip Replacement, Longitudinal Cohort, Bone Turnover, Barwon Joint Registry, Physical Activity, Dietary Calcium, Hazard Ratio, Subchondral Remodelling, Competing Risk Analysis, Orthopaedic Surgery