This study was a multicenter study conducted across four tertiary hospitals. This involved five in-person clinic visits (baseline, 3, 6, 9, and 12 months) and daily home-based body composition measurements over 1 year. Eligible participants were adult patients (aged 18 years or older) who underwent breast cancer surgery an axillary lymph node dissection or sentinel lymph node biopsy and were therefore considered at risk of developing lymphedema to monitor longitudinal fluid changes. Prophylactic mastectomy and reconstruction were permitted. Participants were excluded if they had a bilateral cancer diagnosis, were pregnant, had an infection (ie, cellulitis), or had an implanted electrical medical device (ie, cardiac pacemaker). The participants were in the self-care phase of treatment. Enrollment took place between February 2023 and July 2024 at four institutions: the REHAB Hospital of the Pacific (Honolulu, Hawai’i), Shirley Ryan AbilityLab (Chicago, Illinois), One Oncology Tennessee (Nashville, Tennessee), and the University of Minnesota (Minneapolis, Minnesota). Recruitment was planned to be evenly distributed across the four regions; however, the number of enrolled participants varied by site during the recruitment period (see Figure 1).

This prospective, patient-driven, real-world observational study was reviewed and approved by the institutional review boards of all participating sites. Ethical approvals were obtained from the University of Hawai‘i Institutional Review Board (REHAB Hospital of the Pacific; IRB No. IRB2022-00055), the Northwestern University Institutional Review Board (Shirley Ryan AbilityLab; IRB No. STU00216604), the BRANY Institutional Review Board (One Oncology Tennessee; IRB No. 24-08-131-652), and the University of Minnesota Institutional Review Board (IRB No. STUDY00021973). The study was conducted in accordance with the principles of the Declaration of Helsinki and all applicable human research regulations. Written informed consent was obtained from all participants prior to their participation in the study.

Clinical assessments were conducted as part of routine post–breast cancer care at each participating site. At enrollment, participants underwent standard clinical evaluations to characterize their baseline fluid status, including determination of the ISL stage of lymphedema. In-clinic BIA measurements were performed. All clinical data collected during these visits reflected information routinely obtained during standard patient care and were not collected solely for research purposes. Participants enrolled in the study received standardized instruction on the use of the home-based BIA device, and the initial home-based BIA measurement was performed in the clinic at the time of training. After returning home, participants were instructed to perform self-administered BIA measurements twice daily: once in the morning within 1 hour after waking as part of a consistent daily routine and once in the evening within 1 hour before bedtime to capture diurnal variations within a consistent daily routine. Adherence to this long-term self-monitoring schedule was encouraged to reflect real-world engagement without clinical supervision (see Figure 1).

Consistent with routine LE surveillance practices, participants attended in-clinic follow-up visits every 3 months over a 12-month period (at months 3, 6, 9, and 12). During these visits, lymphedema status was reassessed, and participants received guidance on self-management based on clinical findings, while continuing home-based BIA measurements throughout the monitoring period. Participants were referred for further treatment as clinically indicated based on in-clinic evaluation findings, including range-of-motion deficits, lymphedema symptoms, tissue or integumentary issues, and reduced function or mobility.

Conventional criteria were used to define the onset and stage of BCRL across participating institutions. In routine clinical practice, BCRL onset in the at-risk arm was assessed using circumferential tape measurements and volume-based criteria comparing the ipsilateral and contralateral upper extremities. Although site-specific clinical assessments were conducted according to local practice at each institution, lymphedema staging in this study was standardized using the widely accepted ISL classification system, including subclinical lymphedema (stage 0) and stages 1‐3. Stage 0 was defined as a latent or subclinical condition characterized by impaired lymphatic transport without clinically apparent swelling, whereas stages 1 through 3 were defined according to ISL criteria based on the presence and progression of clinically observable edema and tissue changes. In this study, particular emphasis was placed on monitoring participants at stages 0 and 1, as these stages represent the subtle physiological fluid fluctuations that home-based BIA is designed to track over time. Throughout the study period in the clinical setting, all participants were managed while remaining asymptomatic or classified as stage 0 or 1, and no cases of stage 2 or 3 lymphedema were observed.

BIA measurements were obtained using an in-clinic device (InBody 770 [IB770]; InBody Co, Ltd) and a home-based device (BWA ON; InBody Co, Ltd). IB770 is a body composition analyzer widely used in body composition research and provides standard BIA-derived parameters, including fat-free mass, soft lean mass, body fat mass, and ECW/TBW ratios, with the ability to measure segmental ECW/TBW values. IB770 measurements were used as an in-clinic reference to evaluate the consistency of this metric across devices. Both systems are multifrequency bioimpedance analyzers designed to support region-specific assessment relevant to lymphedema surveillance. IB770 was used for standardized in-clinic assessments as part of routine clinical care, whereas BWA ON was used for self-administered home measurements, with results displayed via a dedicated mobile app on the participant’s personal smartphone. Measurement data generated by BWA ON were assigned a unique participant-specific identifier and encrypted prior to transmission. Encrypted data were securely transferred through a cloud-based system to the participant’s affiliated institution, where access was restricted to authorized health care professionals and researchers for clinical monitoring and research purposes. No directly identifiable personal information was stored or displayed within the app beyond what was required for secure data linkage.

Evaluation criteria were defined a priori to distinguish device-level agreement from longitudinal home-based measurement characteristics. The primary outcome was the agreement between the home-based BIA device (BWA ON) and the in-clinic reference device (IB770) to assess whether the home-based system provided measurements suitable for clinical monitoring. This comparison was based on paired measurements obtained during the initial in-clinic visit, at which both devices were used under standardized conditions. The secondary outcome focused on within-day variation in home-based BIA measurements obtained using BWA ON during daily self-monitoring. Specifically, differences in segmental ECW/TBW ratios between morning (before-noon) and afternoon (after-noon) measurements were evaluated to identify inherent diurnal variations that could influence the interpretation of long-term trends. In addition, analyses incorporated limb dominance as a clinically relevant factor, reflecting conventional bioimpedance interpretation practices in patients with lymphedema, in which dominant and nondominant limbs may exhibit systematic physiological differences. Accordingly, secondary analyses examined whether diurnal changes in segmental ECW/TBW differed depending on whether the affected (at-risk) limb corresponded to the dominant or nondominant upper extremity. The feasibility of the monitoring system was evaluated based on participant adherence to the twice-daily measurement protocol.

The analytic dataset consisted of repeated measurements of clinical lymphedema status (ISL stage), in-clinic BIA data (IB770), and longitudinal home-based BIA data (BWA ON). To evaluate agreement between the in-clinic and home-based BIA devices, the Lin concordance correlation coefficient (CCC) was used as the primary measure of agreement, supplemented by Pearson correlation coefficients and intraclass correlation coefficients (ICCs; two-way mixed-effects model with absolute agreement). Bland-Altman analysis was additionally performed to assess mean differences and 95% limits of agreement.

To illustrate the heterogeneity of home-based BIA change patterns within the ISL stage framework, grouped bar charts were used to visualize coexisting directional changes across stage transition categories. For diurnal variation analyses, paired t tests were conducted for only days where both morning (before-noon) and afternoon (after-noon) measurements were available to compare mean ECW/TBW differences at the individual level.

To account for the hierarchical and repeated-measures structure of the full longitudinal dataset, linear mixed-effects models (LMMs) were applied. In these models, measurement timing (before-noon vs after-noon) and limb status (affected vs unaffected) were treated as fixed effects, while participant IDs were included as random effects to control for within-individual correlation. Data distributions were examined and visualized using box plots and grouped bar charts to facilitate comparison across measurement conditions. Statistical significance was defined as a two-tailed P value of <.05. All statistical analyses were conducted using Python (version 3.13).

A total of 37 female patients were initially enrolled in the study. However, two individuals who self-identified as ambidextrous were excluded from the analysis, as limb dominance was one of the key analytical variables in this study. For outcome analyses, participants were categorized based on ISL staging; however, stage 0 was further stratified to reflect longitudinal clinical status during the monitoring period. Specifically, participants who remained consistently within stage 0 throughout follow-up were distinguished from those who met the criteria for stage 0 but experienced transient fluctuations in clinical status during the study period. This stratification was applied to account for the known heterogeneity of subclinical lymphedema and to better capture differences in daily ECW distribution patterns observed through home-based bioimpedance monitoring.

At the time of program enrollment, all cases were unilateral with respect to the affected arm (right or left). In participants who had previously undergone bilateral mastectomy, those with a clearly defined affected and unaffected arm based on clinical management were included and monitored accordingly. For the final analysis, participants were categorized according to whether the affected arm was the dominant or nondominant limb, as limb dominance was a key variable of interest in this study. Baseline participant characteristics were therefore summarized by affected-side dominance (dominant vs nondominant). No statistically significant differences were observed across the four resulting groups with respect to age, body weight, or BMI.

Adherence to the patient-led home-based BIA protocol is summarized in Table 2. Over the 12-month monitoring period, 35 participants across four geographically diverse sites (Honolulu, Hawai’i; Chicago, Illinois; Nashville, Tennessee; and Minneapolis, Minnesota) proactively contributed a total of 16,425 measurements using the BWA ON device. This high volume of data yielded an overall adherence rate of 64.3% relative to the recommended twice-daily measurement schedule.

Notably, while adherence varied across sites—ranging from 48.9% (Chicago, Illinois) to 78.4% (Minneapolis, Minnesota)—all sites generated sufficiently dense longitudinal data to support the feasibility of extended, patient-driven monitoring in a real-world setting. Furthermore, as described in Table 2, the diurnal distribution of measurements was remarkably consistent across all regions. Before-noon measurements accounted for 56% of the total dataset, while after-noon measurements comprised 44%. This balanced distribution indicates that participants maintained a stable measurement routine regardless of their clinical site, ensuring sustained engagement with both daily measurement windows throughout the monitoring period. This consistency confirms that the home-based system is practical for capturing both morning and afternoon fluid status without clinical supervision.

To evaluate the agreement between the in-clinic BIA system (IB770) and the home-based device (BWA ON), paired measurements obtained at the enrollment visit were analyzed. Correlation analyses demonstrated strong positive associations between the two devices across all parameters, with correlation coefficients exceeding 0.99 for weight and percent body fat, and 0.86 for segmental ECW/TBW ratios. Of the 35 participants, two were excluded from the agreement analysis due to significant data entry errors in height (unit confusion between feet and inches), which resulted in physiologically implausible outliers. Therefore, the final agreement analysis was conducted with 33 participants.

Bland-Altman analyses showed minimal mean bias (0.001) for ECW/TBW measurements in both the affected and unaffected arms, with narrow 95% limits of agreement (Figure 2). ICC and CCC values further supported substantial agreement between the systems. Collectively, these findings indicate that home-based BIA measurements obtained using the BWA ON device are highly consistent with professional in-clinic standards. This high level of technical agreement supports the use of the home-based system as a reliable tool for high-frequency longitudinal monitoring, providing data quality comparable to standardized clinical assessments.

To evaluate longitudinal trends in ECW distribution over the 12-month monitoring period, participants’ clinical lymphedema status was defined using ISL stage assessments obtained during five scheduled in-clinic visits for lymphedema management. These visits, occurring at approximately 3-month intervals, served as the reference standard for evaluating changes in edema status over time. Based on ISL staging across consecutive visits, temporal changes were categorized into four clinical transition patterns: worsening (stages 0-1), improvement (stages 1-0), stable low-risk status (persistent stage 0), and stable persistent status (persistent stage 1). These categories were designed to capture the full spectrum of LE progression and regression, as well as periods of clinical stability.

Corresponding to each clinical visit interval, home-based bioimpedance data were analyzed to determine whether segmental ECW trends measured by the BWA ON system exhibited concordant directional changes. The ECW/TBW ratio was used as the primary bioimpedance-derived indicator, with values exceeding 0.390 considered indicative of abnormal extracellular fluid accumulation. For each intervisit interval, changes in the ECW/TBW ratio were classified according to directional trends relative to the previous interval: a “decrease” reflected a shift toward edema improvement, an “increase” indicated a shift toward worsening, and minimal change was defined as “stable” status maintenance.

These bioimpedance-based transition patterns were examined in relation to the corresponding ISL stage transitions to assess whether home-based ECW/TBW trends reflected clinically observed changes. As illustrated in Figure 3, the results demonstrated that ISL stage classification alone does not fully capture the underlying heterogeneity of BWA ON change patterns. Across all stage transition categories, BWA ON measurements did not follow a uniform directional pattern. Notably, heterogeneous distributions were observed even within clinically stable groups (0→0 and 1→1), revealing substantial variability in bioimpedance behavior that remained undetected within the same clinical stage framework. These distinct distributions across clinical risk contexts suggest that patient-led monitoring provides a more granular level of physiological information than intermittent clinical staging.

This analysis investigated whether ECW/TBW ratios obtained in a home-based environment demonstrated systematic variation according to measurement time within a day. Furthermore, we assessed how these ratios were influenced by the interaction between measurement timing, limb dominance (dominant vs nondominant), and BCRL status (stage 0 vs stage 1). These associations were evaluated using LMMs to account for the hierarchical nature of repeated measurements within participants.

Initially, we examined whether the ECW/TBW ratio of the affected arm varied according to measurement time (before-noon vs after-noon) and whether it differed depending on whether the affected arm corresponded to the dominant or nondominant limb. The results of the stepwise LMM are presented in Table 4. The estimated mean ECW/TBW ratio of the affected arm at the before-noon time point was 0.380 (intercept coefficient, P<.001). Compared to before-noon measurements, the after-noon ECW/TBW ratio was significantly decreased by 0.001 (P<.001). The between-group variance was minimal (group variance=0.000; P<.001), indicating that variability in the ECW/TBW ratio was primarily associated with measurement timing rather than interindividual differences.

Based on the premise that limb dominance influences bioimpedance interpretation, we further stratified participants into four groups. The reference group consisted of participants whose dominant arm was affected and who consistently maintained stage 0 throughout follow-up (Same & non-BCRL). In this baseline group, the mean ECW/TBW ratio was estimated at 0.378 (P<.001), with a modest but significant afternoon decrease of 0.001 (P<.001). The results of this comprehensive interaction analysis are summarized in Table 5.

At the before-noon time point, the “different and non-BCRL” group showed a significant increase of approximately 0.003 compared with baseline (P=.047). The analysis of the interaction between time of day and group status revealed distinct diurnal patterns. Notably, the “same and BCRL” group exhibited a significantly greater reduction in the affected arm ECW/TBW ratio in the afternoon (interaction coefficient=−0.002; P<.001) than the baseline group. In contrast, the “different and non-BCRL” group reflected a significantly smaller afternoon decrease (interaction coefficient=−0.000; P<.001).

Overall, while absolute differences in mean ECW/TBW ratios across groups were relatively small, diurnal variation patterns differed significantly. The same and BCRL group exhibited the largest magnitude of afternoon decrease, suggesting that diurnal changes are closely associated with clinical BCRL status and daily loading of the dominant limb. The minimal random intercept variance indicates that variability was more strongly explained by within-individual factors—such as measurement timing—than by between-individual differences. This underscores the need for standardizing home-based measurements to a fixed morning time to ensure data interpretability and minimize inherent physiological noise in longitudinal self-monitoring strategies.

This study evaluated the technical agreement between in-clinic and home-based bioimpedance systems and characterized the longitudinal and diurnal patterns of segmental ECW/TBW ratios over 12 months. Our findings demonstrated that home-based BIA measurements using the BWA ON device are highly consistent with professional in-clinic standards (IB770), supporting the reliability of patient-led remote monitoring. However, longitudinal analysis revealed that ISL stage transitions do not always follow uniform BIA directional patterns, highlighting the inherent physiological heterogeneity within clinical stages. Notably, the linear mixed-effects model identified that measurement timing (diurnal variation) is a more significant driver of ECW/TBW variability than interindividual differences.

The strong agreement observed across all body composition parameters (Table 3), particularly the high ICC and CCC values for segmental ECW/TBW ratios, underscores the technical validity of home-based BIA. Despite the lack of clinical supervision, participants demonstrated sustained engagement, contributing over 16,000 measurements with a 64.3% adherence rate. This high volume of real-world data suggests that patient-led monitoring is not only feasible but also provides a much denser data stream than intermittent clinic visits, which are typically limited to 3-month intervals.

A key contribution of this study is the characterization of diurnal ECW/TBW fluctuations. While absolute differences between groups were small, the same and BCRL group (dominant arm affected with a history of BCRL) exhibited the most pronounced afternoon decrease (Table 5). This suggests that the dominant limb, which typically undergoes higher daily mechanical loading, may experience more dynamic fluid shifts in individuals with a history of lymphatic impairment. The near-zero group variance observed in our mixed-effects model further confirms that these fluctuations are primarily a function of measurement timing rather than inherent differences between individuals. Consequently, interpreting a single ECW/TBW value without considering the time of day may lead to clinical misinterpretation of a patient’s edema status.

Our results provide a practical evidence-based framework for standardizing home-based surveillance. Given that sustained twice-daily measurements may be logistically challenging for long-term self-care, we propose that home-based monitoring be standardized to a fixed morning time–ideally within 1 hour of waking and prior to daily physical activities. This standardization minimizes physiological “noise” caused by diurnal loading and ensures that longitudinal trends reflect true changes in lymphatic health rather than transient daily fluctuations.

This study has several limitations, including a relatively small sample size, which may limit the generalizability of the diurnal patterns across broader populations. Additionally, the 3-month interval for clinical ISL staging restricted our ability to correlate high-frequency BIA data with immediate clinical changes. Future research should use more frequent clinical assessments or digital symptom tracking to further validate the sensitivity of home-based BIA trends in capturing subclinical fluctuations.

In conclusion, home-based segmental bioimpedance analysis provides a reliable and technically valid method for longitudinal lymphedema surveillance. The significant impact of diurnal variation and limb dominance on ECW/TBW ratios highlights the necessity of standardized measurement protocols in remote health settings. By establishing consistent morning measurement routines, patient-led monitoring can provide granular, high-quality data that complement traditional in-clinic care, supporting more precise and personalized management for breast cancer survivors at risk for lymphedema.