In all fifty of our patients, the device yielded REMS-BMD values at the femoral neck before total hip arthroplasty, as well as values of REMS-BMD measured in the metal material of the endoprosthesis. The postoperative “femoral neck” REMS-BMD values were significantly lower compared with the preoperative values. The Bland–Altman’s analysis indicated that the REMS-BMD values measured in the “femoral neck” after surgery were consistently and statistically significantly lower than the REMS-BMD values measured before surgery. The coefficient of determination in multiple linear regression analysis of the REMS-BMD measurements from the femoral neck was higher before total hip arthroplasty compared to REMS-BMD measurements taken after the surgery. For interpretation of the results, it is essential to note that the REMS-BMD values of the patient are derived from models stored in the proprietary device’s database, where the Osteoporosis Score is categorized by 5-year age intervals, sex, and BMI. The algorithm contains a spectral validation step. In this step, the spectra not similar enough to one of the reference models should be rejected [12]. Arguably, this should lead to rejection of metallic implant scans. Our findings suggest that the unique quality of the material being measured was not rejected and may hinder the accurate assessment of the Osteoporosis Score. If the metallic materials were classified as outside the models, the REMS-BMD would likely be calculated based solely on the patient’s age, sex, and BMI. Consequently, the 2.1% differences in REMS-BMD values before and after hip replacement, as determined by Bland–Altman’s analysis, could be explained by the dominant contributions of sex, age, and BMI to the determination of REMS-BMD.

Across previously published studies, demographic and body composition indices typically explained between 20 and 40% of BMD variance [15, 16], while some studies explained up to 60–80% based on the specific variables included in the models, measurement sites, and population characteristics, making direct comparisons between studies challenging [17,18,19]. In this study, pre-surgery age, sex, and BMI explained about 89–90% of the variability in REMS-BMD of the femoral neck (Adj Rsqr = 0.898). This substantial impact of age, sex, and BMI persisted in postoperative measurements, even with metal implants occupying most of the former bone volume (Adj Rsqr = 0.888). There was strong agreement between calculated and measured femoral neck REMS-BMD values (Adj Rsqr = 0.902), and Bland–Altman analysis showed no systematic bias, indicating that a simple model incorporating age, sex, and BMI can accurately predict femoral neck REMS-BMD and T-scores in both pre- and postoperative settings, with excellent agreement to measured values and no evidence of systematic bias.

We were unable to compare the REMS-BMD and DXA-BMD values in our patients with osteoarthritis affecting the proximal femur. In key echographic studies, correlations between DXA-BMD (measuring bone mineral) and REMS-BMD values (predicted based on Osteoporosis Score, sex, age, and BMI) have shown both sensitivity and specificity above 90% for both lumbar spine and femoral neck [12, 13, 20, 21]. However, the European multicenter study involving 4307 Caucasian women aged 30 to 90 years showed a larger variation in BMD measurements obtained through the REMS method compared to DXA [22]. In some studies, the REMS-BMD method classified more women as “osteoporotic” compared to the DXA-BMD. In patients with type 2 diabetes mellitus, DXA-BMD values were significantly higher than those in the control group, while REMS-BMD values were lower. Consequently, the proportion of women classified by REMS as “osteoporotic” was 47.0%, compared to 28.0% as measured by DXA [23]. Additionally, a study involving postmenopausal women with radiological osteoarthritis of the lumbar spine found that the REMS-BMD T-score was significantly lower than the DXA-BMD T-score for both the lumbar spine and all femoral subregions [24, 25]. Furthermore, in patients with disuse-related osteoporosis, the femoral neck REMS-BMD was significantly lower than the DXA-BMD value [26]. The discrepancies observed between the results of REMS-BMD and DXA-BMD measurements in some studies could be attributed to REMS’s capability to overcome common artifacts, or to the quality of bone other than in osteoporosis [27]. Additional research is needed to evaluate the reliability of REMS in assessing REMS-BMD in patients whose DXA-BMD values deviate from the expected correlation between REMS-BMD and DXA-BMD.

A strength of this study is the high predictive accuracy of the regression models, supported by strong statistical performance and agreement analyses, using readily available clinical variables (age, sex, BMI). However, we conducted a single-center observational study involving a well-defined and homogeneous group of patients with hip osteoarthritis. This focused approach allowed for a targeted evaluation of REMS in a specific surgical population. The real-world performance of REMS in a more heterogeneous population cannot be inferred from our cohort, and external validation in broader cohorts is required to confirm generalizability. Additionally, as the device does not provide the Osteoporosis Score data, we were unable to determine how the material of the endoprosthesis influences the spectral changes in the radiofrequency ultrasound signal or the subsequent determination of the Osteoporosis Score. We also lack a corresponding set of DXA values in our patients to illustrate the artificial increases in BMD post-hip replacement by DXA, in contrast to evaluable REMS-derived BMD.

The significant differences observed in the REMS-BMD results when the operator inputted either their actual or fictitious age and BMI (see Fig. 5) are attributed to the algorithm used for calculating REMS-BMD. When an incorrect age or BMI is input, the algorithm automatically aligns the measured spectral deformation with inappropriate reference models.