No abstract

Background and Aims Monitoring of fluid, body composition and nutritional changes is important in clinical nephrology. The Body Composition Monitor (BCM; Fresenius Medical Care, Bad Homburg, Germany) measures whole-body bioimpedance and determines extracellular and intracellular resistance by using the Cole-model to estimate total body water (TBW-BCM) and its partition into extracellular and intracellular water. Both can then be used to define body composition and separate body weight into lean tissue mass, adipose tissue mass, and fluid overload. Urea kinetic modeling (UKM) allows the estimation of dialysis dose (double-pooled Kt/V), urea distribution volume (V-UKM) and dietary protein intake. We studied the bias between estimated V-UKM to anthropometric and measured TBW-BCM (Vant, TBW-BCM). Method Pre-hemodialysis (HD), electrodes for the BCM assessments were placed on the non-arteriovenous access arm and ipsilateral leg, respectively, with the patient in a supine position. Vant was calculated using the Watson equation. In addition to these assessments we entered the specified values from the most recent urea kinetic modeling (UKM) treatment into the online solute-solver calculator (http://ureakinetics.org). We chose a baseline ratio of modeled/anthropometric volume of 0.6 to 1.3 L to exclude values with data entry errors and/or UKM sampling errors. We calculated the post HD TBW-BCM by subtracting the intradialytic weight loss and adjusted these estimates by the differences in post HD weight between sessions to make both estimates comparable. We depicted the comparison between the estimated V-UKM versus the TBW-BCM in a scatter- and Bland-Altman (BA) plot (Figure). For the purpose of error investigation we studied the computed bias (V-UKM minus TBW-BCM) as a function of body mass index (BMI) and stray capacitance (td) in a BA plot. We then calculated the difference between Vant and V-UKM and illustrated the comparison in a scatter and BA plot. Results In a cross-sectional design, we studied 161 stable prevalent HD patients (61.3±14.7 years, 98 (60.9%) males, height of 167.5±10.7 cm) prior to their treatment. The regression plot showed slight agreement (R2= 0.69) and the Bland-Altman plot no systematic trends or proportional error in the main analysis (Figure 1a and b). Neither BMI or td explained bias and variance in the bias between both estimates. Vant and V-UKM plots showed agreement (R2 of 0.68) with a mean bias of -2.3±5.1 and no proportional error. Conclusion Both TBW-BCM and the V-UKM as the “bronze standard” of TBW estimation seemed to agree reasonably well. Neither body composition measurement or kinetic modeling approach showed any significant influence on the accuracy and precision of the estimate. According to BCM availability, estimated V-UKM or measured TBW-BCM could be used alternatively in practice to support clinical decision when pharmacokinetic considerations are concerned.

Background and Aims Pervasive sensing technologies allow healthcare providers to gain additional insights into patient’s status outside the clinical setting. In order to adopt widespread use of remote monitoring devices we must first study the feasibility of their use. We aim to quantify how long patients will use a wearable device before requiring an intervention to maintain the use of the device. Method In-center hemodialysis (HD) patients were enrolled from 4 clinics in New York City starting in May 2018 and followed for a period of up to 1 year. Patients ≥18 years, on HD ≥3 months, able to walk, owning a smartphone, mobile tablet or PC were enrolled. They were provided with a wrist-based monitoring device (Fitbit Charge 2). Participants were instructed on how to use the device, and sync data to their smartphone, mobile tablet, or PC. If a patient failed to sync data for 7 consecutive days, a text message or email reminder was sent from the research staff. We evaluated time to first notification using Kaplan Meier time-to-event analysis. Predictors of time to first intervention was assessed via univariate Cox Regression. Patients were censored at the end of the observation period (January 6, 2020). Socio-economic parameters such as living situation, marriage status, employment status, race, and education level were collected at the beginning of the study. Results 125 patients were enrolled into our study. 7 patients were screen-failed after enrollment. At enrollment patients were 54±12 years old with a dialysis vintage of 5.6±5.8 years. 37% lived alone, 56% were single, 59% unemployed, 64% were African-American, and 42% had an education level of some college or higher. 82% of the patients required a text message reminder to continue wearing/syncing the device. Mean and median time to first reminder were 101 days (95% CI 80 to 123) and 50 days (95% CI 35 to 70 days), respectively. The probability of being on the study without intervention is shown in Figure 1. Predictors of time to first intervention were chosen a priori and included gender, age, living situation, and education level. None of these parameters were significant predictors of time to first intervention. Conclusion We found that most patients will require at least some intervention or counseling to maintain the use of a wrist-based wearable device for remote patient monitoring. While most patients require an intervention before approximately 2 months into wear, the patients who can maintain use independently after that point are likely to do so for longer.

Background and Aims Preciado et al. have identified half-hourly relative blood volume (RBV) targets during hemodialysis (HD) that are associated with significantly improved patient survival. Attainment of these RBV targets would require frequent adjustments to the ultrafiltration rate (UFR) by the dialysis nurse, which is logistically not feasible. We developed a novel proportional-integral controller that takes RBV data from the commercially available CLiC® device as an input and provides UFR suggestions to guide the RBV curve into the desired targets. In this study, we investigated the degree to which the dialysis nurses accepted the UFR recommendations made by this novel feedback controller. Method We conducted a single-arm, prospective, interventional pilot study in subjects on chronic HD at three Avantus Renal Therapy Dialysis Centers in New York City. Subjects were treated with Fresenius 2008T HD machines. RBV was measured with the CLiC® device. CLiC® and HD machine data were fed into a research laptop running the UFR Feedback Controller software. The UFR recommendations (generated every 10 minutes) were evaluated by dialysis nurses who then either implemented or rejected them as they deemed clinically appropriate. Results Fifteen subjects (58.9 ± 15.3 years, 33% white, 53% black, dialysis vintage 4.1 ± 2.4 years, baseline interdialytic weight gain 2.6 ± 0.8 L, treatment time 222 ± 28 min) were studied. Fifty-six study visits (48 complete, 8 partial) from 14 subjects had analyzable data. Out of 1,038 UFR recommendations made by the Controller, 926 (89.2%) were accepted by the dialysis nurses, while 112 (10.8%) were overridden. In 48% of all 48 complete treatments, the Controller ran without a single override by the nurses. Of the other half of the treatments, about a quarter only saw one override per treatment (only half of which were due to low blood pressure or clinical symptoms). Only 1 of 59 study visits was incomplete due to a staff intervention for medical reasons that disengaged the Controller. For the 25 complete study treatments (from 11 subjects) that had at least one Controller suggestion overridden by the nurses, we analyzed the direction and magnitude of disagreement between the Controller-suggested and the implemented UFR (Fig. 1). Out of a total of 109 overridden Controller suggestions, 20 implemented UFRs were greater than the respective Controller-suggested UFRs (i.e., not indicated for medical reasons). Another 70 implemented UFRs were less than 100 mL/h (on average 49.7 mL/h) lower than the Controller-suggested UFRs (i.e., very mild disagreements). Together, these two categories make up 83% of all “disagreements” between healthcare staff and the Controller, leaving only 19 more pronounced disagreements. Of note, of the 109 UFR overrides, 65 (59.6%) were due to staff preference in the absence of low blood pressure or clinical symptoms; 41 were due to low blood pressure, 2 due to hypotension with clinical symptoms, and 1 due to clinical symptoms without low blood pressure. Conclusion The proportion of UFR recommendations generated by this UFR Feedback Controller that were accepted by the nurses was very high (≈90%). Importantly, of the relatively few cases where nurses chose to override the Controller’s recommendation, the majority (≈60%) were due to “staff preference” in the absence of low blood pressure or clinical symptoms. The high rate of “staff preference” overrides is likely owed to the fact that, for this study, the nurses exclusively attended to only one patient (the study subject) at a time for the entire HD session.

Background and Aims Preciado et al. have identified half-hourly relative blood volume (RBV) targets (at 30, 60…180 min) during hemodialysis (HD) that are associated with significantly improved patient survival. Attainment of these RBV targets would necessitate incessant adjustments to the ultrafiltration rate (UFR) by the dialysis nurse, which is logistically not feasible. We developed a novel proportional-integral controller that takes RBV data from the commercially available CLiC® device as an input and provides UFR suggestions to guide the RBV curve into the desired targets. The clinician specifies the desired UF goal and the maximum allowed upward/downward deviation from this goal, and the Controller then optimizes the RBV trajectory within the limits allowed by the clinician’s prescription. The present study is aimed to characterize the behavior of this novel feedback controller. Method We conducted a single-arm, prospective, interventional pilot study in subjects on chronic HD at three Avantus Renal Therapy Dialysis Centers in New York City. Subjects were treated with Fresenius 2008T HD machines. RBV was measured with the CLiC® device. CLiC® and HD machine data were fed into a research laptop running the UFR Feedback Controller software. The UFR recommendations (generated every 10 minutes) were evaluated by dialysis nurses who then either implemented or rejected them as they deemed clinically appropriate. The nurses were instructed to only override Controller recommendations if medically indicated, but not in an attempt to manage the subjects’ RBV trajectories themselves. Results Fifteen subjects (58.9 ± 15.3 years, 33% white, 53% black, dialysis vintage 4.1 ± 2.4 years, baseline interdialytic weight gain 2.6 ± 0.8 L, treatment time 222 ± 28 min) were studied (63 study visits, 4.2 ± 1.9 visits per subject). Of 300 analyzed RBV target timepoints, 63% had RBVs within the desired target range, 33% of the RBVs were above and 4% were below target. Stratified by timepoint, the on-target percentage increased from 37% at 30 min to 73% at 180 min into HD, while the proportion of RBVs above or below target decreased. In subjects with at least 4 complete study visits (N=8), looking at each of their first 4 complete visits, on average 71.8% of subjects were within the desired RBV target at 180 min into HD. The rate of intradialytic morbid events did not appear to be outside of the ordinary. There was no indication of adverse events related to the use of the UFR Feedback Controller. The Figure shows an example study visit where the UFR Feedback Controller modulates the UFR on an ongoing basis throughout the treatment to keep the RBV curve close to the ideal target trajectory (red line, defined by connecting the RBVs associated with the lowest all-cause mortality). Solid black line: RBV curve (left y-axis); dashed black line: UFR (right y-axis); green boxes: half-hourly RBV target ranges associated with improved survival. Conclusion The UFR Feedback Controller behaves as expected, steering the patients’ RBV curves toward the predefined target ranges where possible, while simultaneously guaranteeing that the prescribed fluid removal goal will be achieved. Preciado et al. had reported approx. one third of patients within the favorable RBV target range at 3h into HD. In contrast, while our pilot study was relatively small, with use of our novel UFR Feedback Controller, approx. 72% of subjects were within the desired RBV target range at 3h into HD. This novel UFR feedback control technology holds great promise for improving fluid management and clinical outcomes in HD patients without requiring additional staff time.