The protocol for the INVEST in Bone Health trial has been described elsewhere (Miller et al., 2021). In brief, INVEST in Bone Health was a 12-month randomized controlled trial conducted from September 2019 through April 2024. Participants were randomly assigned in a 1:1:1 fashion to engage in dietary weight loss alone (WL), weight loss plus resistance training (WL+RT), or weight loss plus weighted vest use (WL+VEST). All study procedures were approved by the Wake Forest University School of Medicine institutional review board (IRB# 00058279), and all participants provided written informed consent.

Participants were community-dwelling older adults (aged 60–85 years) who were weight stable (no WL >5% within the previous 6 months) and living with obesity [body mass index (BMI): 30–40 kg/m²)] or who were both overweight (BMI: 27 - <30 kg/m²) and had at least one obesity-related comorbidity and willing to engage in study procedures. Interested individuals were not eligible if they engaged in regular resistance and/or aerobic exercise for >60 min on> 5 days per week over the prior 6 months, had severe cardiometabolic disease, musculoskeletal impairments that precluded weighted vest use or exercise training, cognitive impairment, or used a prescription osteoporosis or WL medication within the previous year. In-depth characteristics of the sample are available in a prior publication (Beavers et al., 2025).

Participants (N = 150 recruited between July 2020 and February 2023) were randomly assigned using random permuted blocks of size 3, 6, 9, and 12 and stratified by sex to one of three groups: WL, WL+VEST, or WL+RT. The WL intervention involved the provision of a nutritionally complete partial meal replacement program by Medifast, Inc. (the “4 & 2 & 1 Plan®”) with the aim of inducing a 10% loss of initial body mass within 12 months. Participants in all conditions engaged in group nutrition classes (weekly for the first 6 months, biweekly for the latter 6 months) that were designed to provide social support, education on nutrition concepts, and practice in dietary behavior self-regulation strategies. Those in the WL+VEST group were asked to wear a weighted vest for a target duration of 8 h per day during the most active part of the day. The vest was initially unloaded and then adjusted weekly based on the total amount of weight lost using 1/8th-pound blocks with a maximum load of 10% of initial body mass. On average, 78% ± 30% of the weight that was lost was replaced in the vest (Beavers et al., 2025). Participants were asked to keep a daily log of wear time, vest load, and any complications associated with wear. As reported previously, average self-reported daily wear time across the 12 months was 7.1 ± 1.5 h per day (Beavers et al., 2025). Of relevance to this analysis, participants received no specific instructions on modifying their daily sitting or standing behaviors. Finally, those in the WL+RT condition attended a supervised and center-based progressive RT program on three non-consecutive days each week. Participants aimed to complete three sets of 10–12 repetitions of eight upper- and lower-body exercises at 70%–75% of their one repetition max.

Participants self-reported key demographic characteristics, including age and sex, at baseline. BMD at the hip was a focal outcome of the INVEST in Bone Health trial. The primary outcome of the trial was 12-month change in quantitative computed tomography (CT)-acquired total hip trabecular volumetric bone mineral density (vBMD); accordingly, these measures were collected at baseline and months 6 and 12. Total hip integral (hereafter referred to simply as “total hip”) and cortical vBMD were also collected at each time point. In addition to CT, participants completed dual energy x-ray absorptiometry (DXA) assessments at baseline, 6 months, and 12 months. From this, a measure of total hip areal bone mineral density (aBMD) was extracted, as was aBMD of the femoral neck and lumbar spine. To compute BMI at each time point, height was measured without shoes to the nearest 0.1 cm via stadiometer, and body mass was recorded using a calibrated and certified balance beam scale to the nearest 0.1 kg. Finally, to measure posture, participants wore the ActivPAL 4 accelerometer (PAL Technologies, Glasgow, Scotland) on the midline of their non-dominant thigh for 7 days at each time point (baseline, 6 months, 12 months). The ActivPAL 4 provides an excellent classification of sedentary postures, standing, and stepping behaviors (O'Brien et al., 2022). In this study we summed daily time spent standing and stepping (i.e., time spent upright) to provide insight on total time spent experiencing the load from the vest. ActivPAL data were downloaded via PALBatch version 9 following a 24-h wear protocol and behaviors were classified via the CREA version 1.3 algorithm. Data were averaged across valid days (i.e., days with at least 20 h of wear time as is standard for the ActivPAL 24-h wear protocol) at a given time point, and at least 3 valid days were required for inclusion in analyses.

Baseline descriptive statistics [i.e., mean (standard deviation) for continuous variables, n (%) for count variables] were computed for the sample overall and for each group. To investigate whether time spent upright moderated the impact of the intervention on BMD, we fit a series of linear mixed models, each including a participant-specific random intercept. For each model, change in BMD [computed as change in aBMD (mg/m²) or vBMD (mg/m³) from baseline to 6 or 12 months] was included as the dependent variable and predictors included the baseline BMD value, visit (6 or 12 months), group, and average daily upright duration. Additionally, models included the interaction between upright duration and group to give insight into whether the relationship between group assignment and change in BMD differed by the amount of time participants spent upright at a given time point. Additional covariates included BMI (to investigate the relationship between upright duration and BMD independent of body mass), age, and sex. Interactions were retained in the model at p < 0.10 as determined by marginal omnibus test. Significant interactions were examined via plots of fixed predicted values and post hoc tests contrasting estimated marginal trends between groups. Additionally, to aid in interpretation, coefficients involving upright time were multiplied by 30 to represent a 30-min difference in average daily upright duration. Statistical significance was established at p < 0.05 (note that as an exploratory analysis, no adjustment for multiple comparisons was employed). To investigate the influence of age and sex on each model, we conducted a final set of sensitivity analyses without these covariates. Their removal did not influence our results nor interpretation. All analyses were completed in R version 4.5.1 (R Core Team, 2021).

Participant descriptive statistics are displayed in Table 1. A total of 150 participants completed the parent trial (see the primary outcomes manuscript (Beavers et al., 2025) for a description of the complete sample). Of these, 134 (WL n = 43, WL+RT n = 45, WL+VEST n = 46; 67 ± 4.8 years of age) had sufficient data for inclusion in DXA and/or CT models. Of these participants, 131 had sufficient DXA data for inclusion (WL n = 41; WL+RT n = 45; WL+VEST n = 45) and 132 participants had sufficient CT data for inclusion in analysis (WL n = 43, WL+RT n = 44, WL+VEST n = 45; see Supplementary Tables S1 and S2 for participant characteristics). Those without sufficient data for inclusion in DXA models (n = 3) were 70.66 ± 2.96 years of age; two were female and one was male; one identified as Black American, one as White, and one as more than one race; one had a college education and two achieved a post-graduate education; and had a BMI of 35.67 ± 2.15 kg/m (Villareal et al., 2005). None had sufficient baseline accelerometer data to derive baseline upright time (notably, all included participants had sufficient accelerometer data at later timepoints to be included in analyses). Those without sufficient data for inclusion in CT models (n = 2) were 68.74 ± 5.98 years of age; both were female and White; one had a college education and the other a post-graduate education; had a BMI of 33.05 ± 2.33 kg/m² and spent 251.61 ± 15.68 min upright at baseline.

Fixed effects from mixed models for DXA variables can be found in Table 2. The model for total hip aBMD revealed a significant interaction between group and time spent upright (p = 0.023), and post hoc contrasts revealed this was driven by a significant difference in the relationship between average daily upright time and change in total hip aBMD between WL and WL+VEST (p = 0.009) and between WL and WL+RT (p = 0.043) but not between WL+RT and WL+VEST (p = 0.450). Predicted associations between upright time and change in total hip aBMD are displayed in Figure 1. Investigation of estimated marginal trends revealed a significant negative association between average daily upright time and aBMD within WL (B = −0.08, p = 0.001) suggesting that a 30-min higher average duration of upright time was associated with a 2.4 mg/cm² decrease in total hip aBMD. By contrast, there was no significant association between upright time and hip aBMD in the WL+VEST group (B = 0.013, p = 0.603) or WL+RT (B = −0.013, p = 0.584) condition. Put differently, a 30-min higher average duration of upright time was associated with a 0.39 mg/cm² increase in total hip aBMD for WL+VEST, and a 0.39 mg/cm² reduction in total hip aBMD for WL+RT.

Models did not reveal significant interactions between group and upright time for aBMD at the femoral neck or spine (p = 0.158 and p = 0.109 respectively). For femoral neck aBMD, there was a significant main effect for visit [B = −5.52; p = 0.027; 95% CI (−10.39, −0.65)], such that aBMD decreased by 5.52 mg/cm² between months 6 and 12 after adjusting for baseline. The main effect for upright time approached statistical significance [B = −0.04; p = 0.065; 95% CI (−0.08, 0.00)] such that a 30-min higher average duration of upright time was associated with a 1.2 mg/cm² reduction in femoral neck aBMD.

The fixed effects from mixed models for CT variables can be found in Table 3. The model for integral vBMD did not reveal a significant group by upright time interaction (p = 0.601), and therefore the interaction was not retained in the final model. There was a significant main effect for time [B = −2.88, p < 0.001, 95% CI (−3.80, −1.97)] such that integral vBMD decreased by 2.88 mg/cm³ between months 6 and 12 after controlling for baseline. Regarding trabecular vBMD, the mixed model revealed an interaction between group assignment and time spent upright was retained in the final model (p = 0.094). Exploratory post hoc contrasts indicated this was driven by significantly different relationships between upright time and change in vBMD between the WL vs. WL+VEST conditions (p = 0.043); no other contrasts were significant (all p ≥ 0.080). This interaction is illustrated in Figure 2. Investigation of estimated marginal trends revealed a non-significant negative relationship between trabecular vBMD and average daily upright time among those in the WL condition (B = −0.005, p = 0.077), such that a 30-min higher duration of upright time was associated with a 0.15 mg/cm³ decrease in vBMD. Among those in the WL+VEST condition, a 30-min higher duration of upright time was associated with a non-significant 0.12 mg/cm³ increase in vBMD (B = 0.004, p = 0.267). There was also a significant main effect for visit [B = −1.47; p < 0.001; 95% CI (−1.87, −1.07)] such that trabecular vBMD decreased by 1.47 mg/cm³ between months 6 and 12 after controlling for baseline. Note that fitting the model without the interaction did not meaningfully change this interpretation (i.e., the main effect for visit was retained, no other relationships were significant). Finally, there was no group by upright time interaction for cortical vBMD of the hip (p = 0.480), nor were there any significant main effects of interest in the model without the interaction.

There are several notable strengths to this secondary analysis. For instance, INVEST in Bone Health was a novel randomized controlled trial that allowed for comparisons of dietary weight loss alone or paired with either weighted vest user or resistance training. Additionally, upright time was quantified objectively using the ActivPAL accelerometer, which provides high-quality classification of posture and avoids bias associated with self-report (Prince et al., 2020; Schaller et al., 2016). There are also several notable limitations. As a secondary analysis, we did not conduct an a priori power analysis and therefore it may be that there was insufficient power to detect significant interactions between time spent upright and group assignment on CT-derived measures of BMD. Additionally, the sample was comprised of primarily female, White, and college-educated participants. As these factors are associated with different daily time-use patterns (e.g., males, individuals who do not identify as White, and those with lower levels of education engage in higher levels of occupational physical activity) (Gudnadottir et al., 2019), it would be beneficial to investigate the impact of weighted vest use on individuals belonging to more varied sociodemographic groups and living in more diverse environments. Finally, future work would benefit from the development of objective markers of vest use (e.g., sensor-derived vest wear) to facilitate examination of the amount and duration of daily vest use while in an upright posture required to sustain BMD during weight loss.