Twelve participants were recruited from the prevalent chronic HD population—London Health Sciences Center Renal Program, London, Ontario, Canada. Patients were included if they were > 18 years of age, receiving chronic HD therapy three times per week for > 3 months and had minimal to no urine output (< 250mL/24 hours).

Dialysis treatments were delivered in a single center (St. Joseph’s Hospital, London, Ontario, Canada) by a single operator (JDP). HD was administered using the Fresenius 5008 system, with high-flux polysulfone dialyzer and according to the participant’s routine HD prescription. Treatments were delivered midweek (Wednesday or Thursday) during the short interdialytic period. The majority of the treatments lasted for 4 hours (10/12); however, two patients received slightly shorter treatments. Dialysate parameters were programmed according to patient’s individual prescriptions; the dialysate sodium range was between 137 and 140 mmol/L; the dialysate potassium was either 1.5 or 3 mmol/L; all patients had a dialysate calcium of 1.25 mmol/L; and bicarbonate ranged from 35 to 40 mmol/L. Anticoagulation was achieved using low-molecular-weight heparin (Dalteparin), with doses between 2,500 and 5,000 units. Dialysate flow was 500mL/minute, and the temperature was set to 36.5°C for all treatments. For each session, net ultrafiltration was set on an individual basis according to the patients’ achievable ideal dry weight. Blood pump speed varied between 330 and 400mL/minute. Four participants were dialyzed via arteriovenous fistula, one via loop graft, and the remaining seven via a central venous catheter. Intradialytic weight gain ranged from 1.1–2.8 kg (see Table 1 for details). Dialysis monitoring parameters were obtained in accordance with program policy with intradialytic measures documented every 30 minutes and more often as required. Intradialytic hypotension (IDH) was defined as a reduction in systolic blood pressure (SBP) of ≥ 20 mmHg and/or ≤ 100 mmHg in association with typical symptoms of hypotension, such as nausea, lightheadedness, or cramping, requiring intervention by a care provider.

The forehead area of interest was cleansed using 70% isopropyl alcohol. The optical oximetry sensor was placed on the patient’s forehead midline, approximately 2.5 cm above the level of the nose. Once securely attached, the patient was asked to sit comfortably in their typical dialysis position (e.g., feet elevated, chair reclined) for approximately 5 minutes to establish a resting state, at which point a baseline PPG measurement was captured. Baseline measures are taken (a) to ensure proper sensor placement and adequate PPG signal and (b) as the basis for individualized hemodynamic comparison of variations throughout HD. Continuous hemodynamic data were then captured throughout the entire HD treatment. Any HD events/interventions were annotated on the CVI monitoring device for each participant, including the initiation and completion of HD, timing of echocardiogram and CT, and symptoms experienced by the participant (e.g., dizziness, lightheadedness, cramping, nausea, headache, pain/discomfort, decrease in blood pressure, or any interventions delivered by the care provider such as ultrafiltration changes, position change or fluid resuscitation requirements). CVI-derived variables were extracted from the CVI-generated output files on a case-by-case basis using a customized semi-automated pipeline designed using R Studio (R Core Team (2022). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL: http://www.R-project.org/ and RStudio Team (2022). RStudio: Integrated Development for R. RStudio, PBC, Boston, MA URL http://www.rstudio.com/) using the tidyverse package (27). All acquired data were analyzed post hoc (FRS, MHJ, DW JDP).

Dynamic CT images of the heart were acquired at baseline and peak HD treatment timepoints. Participants were aligned on a CT bed in supine position and scans were performed between 75%–75% R–R interval, prospectively ECG-gated while the participants were free breathing. For the quantification of myocardial blood flow, iodinated contrast agent (Isovue 370) was delivered intravenously during the image acquisition and the delivery of contrast was traced with 32 scans every 1–2 heartbeats. The scanner setting for all dynamic CT images are as listed: display field of view = 45.0 cm; tube voltage = 100–120 kV; tube current = 100 mA; detector coverage = 160 mm; gantry period = 0.28 s; slice thickness = 2.5 mm.

Following the imaging visit, the dynamic images were processed (LH) utilizing the proprietary ASiR algorithm (Resolution CT console, GE Healthcare) to alter the slice thickness to 2.5 mm with the aim of increasing the signal-to-noise ratio. The reconstructed images were correlated for residual cardiac and respiratory motion using a 3D non-rigid registration algorithm on a proprietary workstation (GE proprietary software, advantage workstation, GE Healthcare). Myocardial blood flow maps of the dynamic images were generated with the application of the Johnson–Wilson–Lee model of tracer kinetics for each HD timepoint (28). Seven slices of the left ventricular myocardium were selected and delineated for the absolute measurement of global myocardial blood flow. For each timepoint, the seven slices were averaged, and the mean global myocardial blood flow measurements were recorded for analysis.

Echocardiography was performed by a trained investigator (LH) prior to commencing HD and again at peak-HD stress (i.e., approximately 20 minutes prior to the end of HD) using commercially available equipment (1.5–3.6 MHz M4S probe, Vivid-q, GE Medical Systems, Soningen, Germany). Standard apical two- and four-chamber views of the left ventricle were recorded for offline analysis using semi-automated software (EchoPac, GE Healthcare)—2D speckle tracking software. Images were anonymized and analyzed in a random order by a trained investigator (JDP), and a random sample of these images was analyzed in random order by a second appropriately trained investigator (LH) to determine estimates of interobserver reliability. Three cardiac cycles were analyzed for each timepoint and the segmental strain values were derived from 12 left ventricular segments. Myocardial segments with a ≥ 20% reduction in longitudinal strain (between pre- and peak-HD stress) were determined to have developed a HD-induced regional wall motion abnormality (RWMA). The presence of two or more RWMAs was defined as myocardial stunning in accordance with previously published methods (17, 22). Poor-quality images were removed from the analysis, and any segment that was not visible or in which the software was not able to accurately track speckles was not counted as an RWMA.

Statistical analysis was performed using JASP [Netherlands (version 0.14.1)]. Descriptive statistics are expressed as mean standard deviation, median or percent. All data were tested for normality using the Shapiro–Wilk test. Comparisons of related outcomes at two different timepoints were performed using the paired t-test for parametric data and the Wilcoxon signed-rank test for non-parametric data. Bivariate correlation was assessed using Pearson’s correlation coefficient for parametric data and Spearman’s coefficient for non-parametric data. An alpha error of less than 5% (p < 0.05) was statistically significant. Graphs were created using Prism GraphPad (version 9.4.0).

Baseline clinical characteristics are shown in Table 1 . All participants received conventional chronic HD treatment three times per week. All participants were anuric. The mean age of our population was 67.2 ± 13.4 years, the mean dialysis vintage of the population was 54.8 ± 40.7 months, and nine participants were male. The causes of renal failure included hypertension (58%) and diabetes (33%). Other causes included IgA nephropathy, hepatorenal syndrome, and toxicity. Other comorbidities included coronary artery disease (50%) and congestive heart failure (25%). All patients were taking either monotherapy or combination antihypertensive/cardiac medication ( Table 1 ). The mean systolic BP prior to the initiation of HD was 143.3 ± 22.3 mmHg, the diastolic BP was 64 ± 13 mmHg, and the intradialytic nadir systolic BP was 101.3 ± 15.9 mmHg. The average interdialytic weight gain was 1.8 ± 0.66 kg, the mean ultrafiltration (UF) rate was 735.6 ± 174.8 mL/hour, and the mean UF mL/kg/hour was 8.5 ± 2.9. One participant had intradialytic symptoms due to vasal vagal response to choking/coughing (preexisting issue). This was not related to intradialytic hypotension; however, a 200 mL bolus of fluid was delivered due to brief unresponsiveness. Seven treatments required reductions to UF rate related to non-symptomatic reductions in BP. There were no treatments that met our definition of IDH. The average intradialytic reduction in systolic BP was 41.7 ± 14.5 mmHg in the absence of symptoms, with a mean relative blood volume (RBV) reduction of 41.7% ± 14.5%.

The PPG waveform produced an embedded raw pulse strength (PS) measurement at baseline. From this baseline measurement, the mean intradialytic PS change was −10.9% ± 33.8%, whereas the lowest PS reduction from baseline was found to be −57.5% ± 22.2%. On average, it took 119.7 ± 53.5 minutes for participants in the population to reach their lowest PS threshold. Participants spent on average 61.6% ± 34.4% of HD treatment with a PS reduction of 10% from baseline. Ten participants had a further reduction in PS of 20% from baseline, spending 53.5% ± 32% of HD treatment at that threshold. The same participants had a further PS reduction of 30% and 40% for 40.6% ± 29.6% and 28.3% ± 24.6% of HD, respectively. A further PS reduction of 50% was observed in nine participants, who spent 15.2% ± 18.2% of HD at this threshold ( Table 2 ).

All 12 participants exhibited treatment-induced myocardial ischemic injury, defined as myocardial stunning. The number of left ventricular segments that underwent a 20% reduction in longitudinal strain ranged between three and 12, mean 6.1 ± 2.9 ( Table 2 ). The lowest PS reduction during HD was associated with the development of RWMAs [p = 0.03, r = 0.63 ( Figure 2 )]. In addition, there were trending patterns in terms of time spent at PS reduction thresholds and the number of RWMAs, with statistical significance being reached with a PS reduction of 50% from baseline [(p = 0.048, r = 0.58), Tables 2 , 3 ]. When stratifying participants by the mean number of RWMAs (severity), although no statistical significance was reached between groups, those who developed a higher number of RWMAs had a lower PS reduction [3–6 RWMAs, PS −47.4 ± 25.9; 7–12 RWMAs, PS −67.9 ± 12.5, p = 0.09 ( Table 2 ; Figure 2B )] and spent more time at each PS threshold ( Table 2 ). In addition, there was a greater reduction in global myocardial perfusion in those with an increased number of RWMAs. Furthermore, increasing numbers of RWMAs were associated with longer HD vintage [7–12 RWMAs = 77.6 ± 17.7 months, whereas 3–6 RWMAs = 48 ± 39 months ( Table 1 )].

The mean global perfusion before HD was 93.9 ± 28.5 mL/minute/100 g, whereas the mean global perfusion at peak-HD stress was reduced to 74.2 ± 13.3 mL/minute/100 g [(p = 0.002), Tables 2 , 3 ]. The average change in global perfusion between the two timepoints was a reduction of −17.1% ± 16.9% ( Table 2 ). This trend was also observed in 10 participants, in whom we observed a reduction in global cardiac perfusion at peak-HD stress; however, in two participants, we observed a very slight increase in global perfusion (3.6mL/minute/100 g, 4.2mL/minute/100 g). Direct changes in intradialytic global myocardial perfusion were associated with reductions in PS [p = 0.048, r = 0.58 ( Table 3 ; Figure 3A )]. In addition, patients who developed more RWMAs also had a lower reduction in global perfusion [3–6 RWMAs, global perfusion −12 ± 15.2; 7–12 RWMAs, global perfusion −21.5 ± 17.7, p = 0.48 ( Table 2 ; Figure 3B )].

UF rates were set in a linear profile using the patients’ achievable prescribed ideal body weight. The mean UF rate for the group studied was 735.6 ± 174.8 mL/hour, equating to 8.5 ± 2.9mL/kg of body weight per hour (mL/kg/hour), which was used for the following correlations seen in Table 3 . Rate of fluid removal was associated with percutaneous perfusion measures of lowest PS reduction [p = 0.003, r = −0.78 ( Figure 4A )], time spent at 10% reduction threshold (p = 0.033, r = 0.63), and time spent at 20% reduction threshold [(p = 0.034, r = 0.61), Table 3 ]. UF was also associated with the number of HD-induced RWMAS [p = 0.026, r = 0.64 ( Figure 4B )], and also with direct changes in myocardial perfusion [p = 0.001, r = −0.83 ( Figure 4C )]. Notably, increased myocardial perfusion was related to a very minimal ultrafiltration requirement in one participant.

There were no relationships found among population demographics and lowest PS reduction nor global perfusion—including age (p = 0.174 and p = 0.313, respectively) and HD vintage (p = 1.0 and p = 0.543, respectively). There were also no associations between traditional macrovascular measures of HD-associated stability and reductions in PS—including pre-HD systolic BP (p = 0.560), intradialytic systolic BP reduction (p = 0.564), nadir BP (p = 0.974), or minimum RBV (p = 0.534). In addition, we found no relationship among pre-HD systolic BP, intradialytic systolic BP reduction, nadir BP, and changes in global perfusion (p = 0.144, p = 0.451, and p = 0.151, respectively). There was, however, a slight association between changes in global perfusion and relative blood volume [RBV (p = 0.049, r = 0.58)]. Details can be found in Table 3 .

Intradialytic imaging is extremely difficult to incorporate into research and, although our study cohort was small, the results (indirect vs. direct perfusion) are impactful. PPG technology does have some limitations. Since the skin is thermally regulated, it is unknown if ambient room temperature or comfort measures (warm blankets) had any impact on the PPG waveforms. In addition, changes in patient positioning plays a role in PPG outputs; for example, the reclined position increases pre load, whereas the standing position decreases pre load.

In conclusion, our study shows that the skin (an accessible microcirculatory vascular bed) is a surrogate for direct measures of organ perfusion. PPG technology is a well-accepted option that can be used for enhanced intradialytic monitoring at the microcirculatory level, providing a window of opportunity for the preemptive adapting of therapy and individualizing of treatment. This proactive approach may result in safer HD delivery, with improved clinical outcomes where current methods fail.