We designed an observational, prospective, proof-of-concept study. Anthropometric, clinical, laboratory and instrumental measurements, together with MBG evaluation, were performed at baseline before starting a mid-week HD session. MBG was then measured at established time-points during this session, in relationship with the absolute and hourly weight loss achieved. Patients then entered a prospective phase in which IDH episodes were systematically recorded during the following fourteen HD sessions (30-day follow-up).

All patients undergoing chronic HD treatment at the Dialysis Unit of the University Hospital of Catanzaro, Italy were screened for eligibility. Inclusion criteria were a regular HD treatment with a rhythm of 3.30 to 4 h sessions/thrice a week, a stable dry-weight and an unchanged therapeutic scheme for at least 3 months before enrolment. Exclusion criteria were dialysis vintage 6 months, recent history of hospitalization for cardiovascular or infectious diseases, symptomatic pre-dialysis hypotension and severe cognitive or physical impairment.

The dialyzer employed was a Flexya dialysis monitor (Bellco®, Mirandola, Italy) with a standard bicarbonate buffer, standard sodium concentration set at 140 meq/L and dialysate temperature maintained at 36.5 ${}^{\circ }$C. All patients received a bolus (25–30 IU/Kg) of unfractioned heparin at the beginning of the dialysis procedure, which was followed by a maintenance hourly dose (500–2000 U). The study was conducted in agreement with the Declaration of Helsinki and was approved by the local Ethical Committee (Comitato Etico “Sezione Area centro” Regione Calabria-Catanzaro, Italy). A fully informed consent was obtained from all participants.

Common biochemical data were measured following the standard methods used in the routine clinical laboratory. Adequacy of dialysis was assessed using Kt/V, calculated as the natural logarithm of the ratio between initial and final urea concentration.

Human MBG was measured in the blood using an enzyme linked immunosorbent assay (ELISA) commercially available kit (BlueGene Biotech, Shanghai, China), according to the manufacturer’s instructions. The enzymatic reactions were quantified in an automatic microplate photometer and centralized in the same laboratory (Clinical pathology lab, “Mater-Domini” University Hospital-Catanzaro, Catanzaro, Italy). MBG measurements were made immediately before starting the treatment, after the first hour, after the second hour and at immediately before patient’s disconnection from the circuit (dialysis end). All tests were made blind and in duplicate and levels were expressed as nmol/L. MBG was also measured in a small group of 10 healthy matched subjects.

Blood pressure (BP) was monitored using an automated sphygmomanometer integrated in the dialysis machine. Pre-dialysis blood pressure was measured in the non-access arm at the beginning of HD, after 10-minutes rest with the patient seated in the dialysis chair before placement of a dialysis needle. Further BP measurements were automatically performed every 10 minutes or upon patient’s request in case of symptoms suggestive of hypotension.

Volume status was clinically appraised by lung auscultation and evaluation of peripheral oedema [15]. Lung auscultation was performed anteriorly and posteriorly in each hemithorax in the seated position. To evaluate crackles, the following scale was used: (1) no crackles; (2) uncertainness about the presence of fine crackles; (3) definite fine crackles at lung bases; (4) moderate crackles; and (5) bilateral, diffuse crackles. For clinical oedema, the following scale was used: (1) no clinical edema; (2) slight pitting (2-mm depth) with no visible distortion; (3) somewhat deeper pit (4 mm) with no readily detectable distortion; (4) noticeably deep pit (6 mm) with the dependent extremity full and swollen; and (5) very deep pit (8 mm) with the dependent extremity grossly distorted.

In addition to clinical evaluation, volume status was also estimated instrumentally by bioimpedance analyses, by echocardiography and by ultrasound evaluation of lung US-B lines and inferior vena cava diameter.

After the baseline assessment, patients were prospectively followed up to 30 days and any IDH episode occurred during the following fourteen dialysis sessions was recorded.

IDH episodes were defined as a systolic blood pressure drop greater than 20 mmHg or a decrease in mean arterial pressure of 10 mm Hg during dialysis with symptoms such as dizziness, headache, confusion, nausea or sweating requiring nurse intervention [3].

Statistical analyses were performed using the SPSS package (version 24.0; IBM corporation, Chicago, IL, USA), the MedCalc Statistical Software (version 14.8.1; MedCalc Software bvba, Ostend, Belgium) and the GraphPad prism software (version 8.4.2, GraphPad Software, San Diego, CA, USA).

Data were presented as mean $\pm$ standart deviation SD for normally distributed values (at Shapiro-Wilk test), median [interquartile range (IQR)] for variables with skewed distribution or frequency percentage for dichotomous values. Differences between groups were determined by the unpaired T-test for normally distributed values, the Mann-Whitney U test for non-parametric values and the chi-square followed by a Fisher’s exact test for frequency distributions. Temporal trends in MBG levels during dialysis were evaluated by analysis of variance (ANOVA) for repeated measures. The Pearson (R) correlation coefficient was employed to test correlations between MBG levels and, respectively, total and hourly weight loss during HD and the absolute number of IDH episodes during follow-up. Before testing correlations, skewed variables were log-transformed to approximate normal distribution. A Receiver Operating Characteristics (ROC) analysis was employed to calculate the area under the curve (AUC) for different MBG measurements, considering the occurrence of IDH as status variable. AUCs were compared by a non-parametric approach. The best cut-off values were computed by the Youden index. Kaplan-Meier curves considering the time to the first IDH episode were generated for patients with MBG above or below the optimal, ROC-derived thresholds and compared by a Log-Rank test. Univariate Cox-regression analyses were performed to test time-dependent associations between variables which were different at baseline between subgroups of patients and the occurrence of the first IDH episode. All results were considered significant if the p value was $\le$0.05.

The source population consisted of 40 chronic HD patients. After the eligibility screening 11 patients were excluded because refused to participate, showed an unstable dry weight over the last 3 months, had too short dialysis vintage or were not on a regular HD regimen.

The final study cohort included 29 prevalent HD patients. The majority of them were male (72.4%) and 27.5% were diabetics. Mean age was 65.4 $\pm$ 12.2 years. The median dialysis vintage was 31 months (IQR 18–55). Patients displayed on average a good dialysis adequacy with a Kt/V of 1.45 $\pm$ 0.26 and no overt clinical or instrumental evidence of volume excess. Five patients (17.2%) were on treatment with angiotensin-converting enzyme (ACE)-inhibitors, ten (34.5%) were on Angiotensin-receptor blockers while eighteen (62%) received calcium-channel blockers or beta-blockers, respectively.

Echocardiography data indicated the presence of a mildly increased left ventricular (LV) mass with normal LV chamber dimensions. Systolic LV and right ventricular (RV) functions were preserved, while we observed a mild impairment in LV diastolic dysfunction, along with a mild increase in systolic pulmonary artery systolic pressure.

MBG levels in HD patients were significantly higher than those measured in healthy controls (0.75 $\pm$ 0.29 nnmol/L vs. 0.59 $\pm$ 0.19 nnmol/L; p = 0.03, Fig. 1). Tables 1,2 summarizes the main cohort data.

Fig. 1.

MBG levels in healthy controls, in the whole cohort of HD patients and in subgroups of HD patients with or without IDH episodes during follow-up. * p = 0.03 vs. controls; ** p = 0.0008 vs. hypotensive HD patients; ${}^{\mathrm{\S }}$p = 0.19 vs. controls.

During the 30-day follow-up, 19 patients (65.5%) experienced at least one IDH with a total of 73 episodes recorded altogether (mean number of IDH per patient 3.47 $\pm$ 1.60; Incidence rate 0.65 IDH events/person/month; 95% confidence interval (CI) 0.50–0.85). Median time to first IDH episode from the baseline assessment was 7.0 days (95% CI 5.0–15.4). All events were resolved by appropriate therapeutic management with no clinical sequelae. At baseline, HD patients with IDH had a significantly increased serum phosphate, serum potassium, ESR, LDL and total cholesterol (p ranging from 0.01 to 0.04) while they were less over-hydrated at bioimpedance (OH 3.95 [1.6–4.3] L vs. 2 [0.52–2.7] L; p = 0.04), had a lower number of US-B lines and a lower pulmonary crackles score (both p = 0.04) and showed a higher E/e’ (8.25 $\pm$ 2.81 vs.10.45 $\pm$ 3.43; p = 0.05) at the echocardiography assessment. This latter result paired well with the prevalence of left atrial enlargement, that was found in 53% of IDH patients compared to 30% of patients not experiencing IDH. No further relevant differences were noticed with respect to the other laboratory, clinical, instrumental and echocardiography parameters. The two subgroups were also comparable in terms of baseline anti-hypertensive therapy.

MBG levels at baseline were markedly reduced in HD patients experiencing IDH during follow-up as compared to others (0.65 $\pm$ 0.28 nmol/L vs. 0.94 $\pm$ 0.21 nmol/L; p = 0.008) while values in this subgroup were apparently not different from those measured in healthy controls (0.65 $\pm$ 0.28 nnmol/L vs. 0.59 $\pm$ 0.19 nnmol/L; p = 0.19).

In the whole cohort, significant, inverse correlations were found between the absolute number of IDH episodes per person and, respectively pre-dialysis MBG (R –0.55; p = 0.001), 2 h MBG (–0.50; p = 0.007) and HD-end MBG (–0.59; p = 0.001).

Tables 1,2 summarize the main characteristics of the two study subgroups. Fig. 1 illustrates MBG levels in HD patients categorized according to IDH occurrence during follow-up.

MBG levels at the end of the HD session were comparable to those measured at baseline (0.90 [0.35–0.96] nmol/L vs. 0.75 $\pm$ 0.29 nmol/L; p = 0.26) with an overall trend not attaining statistical significance (p = 0.07). The average weight loss at dialysis end was 2.45 $\pm$ 0.33 Kg with mean hourly weight losses ranging from 0.52 $\pm$ 0.20 to 0.65 $\pm$ 0.12 Kg. No significant correlations were found between MBG and such absolute or partial weight losses.

MBG variations were then analyzed in HD patients who experienced IDH during follow-up and who did not. The average weight loss achieved at the end of this HD session was not different between the two subgroups (2.32 $\pm$ 0.50 Kg vs. 2.66 $\pm$ 0.86 Kg; p = 0.23), as well as the hourly weight losses (p ranging from 0.16 to 0.55).

MBG levels remained roughly unchanged in HD patients with no documented IDH episodes during follow-up (baseline MBG vs. HD-end: 0.94 $\pm$ 0.21 nmol/L vs. 0.96 [0.94–0.98] nmol/L; p = 0.66; p for trend = 0.26). Conversely, a statistically significant linear trend in this biomarker was observed in HD patients with IDH (p = 0.02). In these individuals, there was an apparent increase in MBG levels after the first hour of dialysis (0.72 [0.56–1.00] nmol/L vs. 0.65 $\pm$ 0.28 nmol/L; p = 0.04), while such levels progressively decreased throughout the rest of the session, being even more reduced at HD end as compared to baseline (0.45 [0.24–0.95] nmol/L vs. 0.65 $\pm$ 0.28 nmol/L; p = 0.04). Predictably, MBG trends in IDH and non-IDH HD patients were remarkably different (p = 0.004). Fig. 2 shows changes in MBG during the HD session in subgroups of patients.

Fig. 2.

Changes in MBG levels throughout a single HD session in patients with or without IDH episodes during follow-up.p for trend in non-hypotensive HD = 0.26; p for trend in hypotensive HD = 0.02; p for comparison between trends = 0.004.

The diagnostic capacity of MBG to identify HD patients experiencing IDH episodes was assessed for each time-point measurement by separate ROC analyses (Fig. 3 and Table 3).

Fig. 3.

ROC curves of MBG levels measured prior to dialysis start (A), 1 h after dialysis start (B), 2 h after dialysis start (C) and at the end of dialysis session (D) to detect patients experiencing IDH episodes. White dots indicate the best cut-off value (Youden Index).

Baseline MBG levels displayed a remarkable diagnostic capacity to discriminate such individuals with an AUC of 0.808 [95% CI 0.634 to 0.982; p = 0.0005] and a best cut-off of $\le$0.82 nmol/L (Sensitivity 73.7% [95% CI 48.8–90.9], Specificity 90.0% [95% CI 55.5–99.7]). MBG levels at 2 h and at HD end held a similar diagnostic performance with an AUC of 0.850 [95% CI 0.699 to 1.000; p 0.0001] and 0.856 [95% CI 0.718 to 0.994; p 0.0001] and best cut-off values of $\le$0.92 nmol/L (Sensitivity 88.2% [95% CI 63.6–98.5], Specificity 70.0% [95% CI 34.8–93.3]) and $\le$0.55 nmol/L (Sensitivity 64.7% [95% CI 38.3–85.8], Specificity 100% [95% CI 69.2–100]), respectively. These AUCs were not statistically different each other (p ranging from 0.07 to 0.92). On the contrary, MBG measurement at 1 h of dialysis were apparently not able to identify IDH patients (AUC 0.642 [95% CI 0.426 to 0.858]; p = 0.19).

Kaplan-Meier survival curves of IDH-free patients were then generated for individuals categorized according to the optimal, ROC-derived MBG thresholds at each time-point measurement (Fig. 4). 1 h MBG levels were excluded from this analysis as apparently not discriminant.

Fig. 4.

Kaplan-Meier survival curves of IDH-free patients according to the optimal ROC-derived cut-off for MBG measured prior to dialysis start (A), 2 h after HD start (B) and at the end of the HD session (C).

Lower baseline MBG resulted strongly associated with a higher risk of IDH throughout the follow-up period (Log-rank 11.162; p = 0.0008) with a crude HR of 4.37 [95% CI 1.73 to 10.99]. Similarly, reduced MBG levels at 2 h (Log-rank 9.312; p = 0.02) and, more markedly, at HD end (Log-rank 23.215; p 0.0001), predicted a higher risk of IDH with crude HRs of 6.51 [95% CI 2.51 to 16.89] and 6.68 [95% CI 2.10 to 21.25], respectively.

Parameters that were different at baseline between subgroups of HD patients with or without IDH episodes during follow-up (Tables 1,2) were tested in a Cox-regression analysis to evaluate time-dependent associations with the outcome of interest. As summarized in Table 4, in this analysis serum phosphate (HR 1.665; 95% CI 1.077 to 2.571; p = 0.02), LDL cholesterol (HR 1.016; 95% CI 1.001 to 1.031; p = 0.03), total cholesterol (HR 1.014; 95% CI 1.001 to 1.027; p = 0.03), E/e’ (HR 1.137; 95% CI 1.000 to 1.416; p = 0.05) and, above all, pre-dialysis MBG (HR 0.155; 95% CI 0.040 to 0.598; p 0.0001), 2 h MBG (HR 0.099; 95% CI 0.021 to 0.454; p = 0.002) and HD-end MBG (HR 0.068; 95% CI 0.014 to 0.322; p = 0.0004) were significantly associated to IDH episodes, while serum potassium, ESR, OH, US-B lines and the pulmonary crackles score failed to reach statistical significance.