Diabetes mellitus and chronic kidney disease (CKD) represent a challenging clinical intersection with significant morbidity and mortality implications. Approximately 40% of patients with end-stage renal disease (ESRD) have diabetes as a primary aetiology, creating complex glycaemic management challenges that intensify when patients progress to maintenance haemodialysis (MHD).1,2

Traditional glycaemic monitoring methodologies demonstrate significant limitations in the haemodialysis (HD) population. Haemoglobin A1c (HbA1c), the standard glycaemic control marker, exhibits reduced reliability in ESRD due to altered red blood cell lifespan, low haemoglobin and uraemic toxins.3,4 Evidence on utility of continuous glucose monitoring (CGM) in diabetic management in dialysis and post-renal transplant patients is limited.5–7

Furthermore, self-monitoring of blood glucose (SMBG) provides only intermittent measurements, potentially missing critical glycaemic excursions that occur during haemodialysis sessions. These limitations underscore the necessity for more comprehensive monitoring strategies that can elucidate the unique glycaemic patterns in this population. CGM is a great innovation to manage glucose levels in such patients. Literature is sparse on the usage of CGM in MHD patients.8,9

This prospective observational study aims to comprehensively characterize glycaemic variability in patients with type 2 diabetes mellitus undergoing MHD using CGM. The study was performed in SRM Medical College Hospital and Research Centre and was approved by Institutional Ethical Committee (SRMIEC-ST0923-1472). The sample size was calculated using Cochran's formula: n=Z2×p×(1−p)/d2, where Z is the standard normal variate (1.96 at 5% type 1 error), p is the expected proportion in the population, and d is the absolute error. Based on previous studies, we considered the prevalence of significant glycaemic variability in haemodialysis patients with type 2 diabetes mellitus. Mirani et al. (2010) reported significant glycaemic variability in 80% of their study population in 10 insulin-treated patients on haemodialysis.10 Similarly, Jung et al. (2010) observed pronounced glycaemic fluctuations in 75% of their cohort of 20 haemodialysis patients with type 2 diabetes.11 Using a conservative estimate of p=0.75, with an absolute precision of 15% and 95% confidence level, the minimum required sample size was calculated as 24.5, rounded up to 25 participants. Stable MHD patients aged between 18 and 70 years with more than 3 months dialysis vintage, with adequate kt/V and target haemoglobin above 10.5g/dl were included in the study. Type 1 diabetes mellitus patients, candidates on peritoneal dialysis, acute on CKD cases requiring HD, patients with erratic glucose measurements, patients in intensive care units, patients in sepsis and pregnant females were excluded.

The study included 25 patients, of whom 68% (n=17) were males and 32% (n=8) were females. The mean age of the study participants was 62.2±10 years. The mean duration of diabetes was 230.4±79.3 months. The average dialysis vintage of the study population was 18.1±14.2 months. The baseline characteristics of the study population are described in Table 1.

For statistical analysis, the study population was divided into insulin-treated (n=13) and OHA (linagliptin)-treated (n=12) groups. In both the groups, glycaemic excursions were observed, with dialysis days having high mean glucose values than non-HD days, and the observation was more prominent in the insulin-treated group (Table 2). The mean glucose levels were lower in the nocturnal period in both the groups (Table 3). Both the treatment groups had lower mean glucose levels in late intradialytic period (Table 4).

On further analysis, it was noticed that the overall glycaemic variability, glycaemic variability in both dialysis and non-dialysis days were lower in linagliptin-treated group (Table 5). Hypoglycaemia was more prominent in the insulin-treated group when compared to the linagliptin-treated group (Table 6). On dialysis days, the glycaemic variability parameters were substantially higher in the insulin-treated group compared to linagliptin-treated group. Such a variability was not observed during non-HD days. Table 7 describes the comprehensive temporal glycaemic distribution across both the treatment modalities, in terms of percentage and absolute duration metrics.

Fig. 1 illustrates the temporal patterns of mean glucose concentrations during the first 4h of HD sessions compared with time-matched periods on non-dialysis days, stratified by antidiabetic treatment modality. A consistent pattern of significantly lower glucose values was observed on haemodialysis days compared to non-dialysis days across all patient groups (p<0.001). The nadir in glucose concentrations occurred approximately 90min into haemodialysis (81.3mg/dL for all patients), representing a 53.9% reduction from time-matched non-dialysis values (186.3mg/dL).

Fig. 1.

Comparison of intradialytic and time-matched non-dialysis glycaemic profiles stratified by antidiabetic treatment modality. This figure illustrates the temporal patterns of mean glucose concentrations during the first 4h of HD sessions compared with time-matched periods on non-dialysis days, stratified by antidiabetic treatment modality. A consistent pattern of significantly lower glucose values was observed on haemodialysis days compared to non-dialysis days across all patient groups (p<0.001). The nadir in glucose concentrations occurred approximately 90min into haemodialysis (81.3mg/dL for all patients), representing a 53.9% reduction from time-matched non-dialysis values (186.3mg/dL).

Treatment-specific analysis revealed differential patterns, with insulin-treated patients (n=13) demonstrating both higher absolute glucose values and greater dialysis-associated decrements compared to linagliptin-treated patients (n=12). Notably, the insulin group exhibited more pronounced post-nadir glycaemic recovery, suggesting differential counter-regulatory responses. These findings underscore the substantial haemodialysis-induced perturbations in glucose homeostasis and highlight treatment-modality differences in glycaemic dynamics during dialysis.

CGM systems represent a paradigm shift in diabetes management in the last decade. These systems measure interstitial glucose concentrations at frequent intervals (typically every 5–15min), generating comprehensive glycaemic profiles across various physiological states and clinical interventions.12,13 The ambulatory glucose profile (AGP) analysis derived from CGM data facilitates the evaluation of critical metrics including mean glucose concentrations, time in range (TIR), glycaemic variability and hypoglycaemic episodes, and such parameters that have demonstrated prognostic significance beyond traditional glycaemic markers.14,15,16

Glucose, a 180Da molecule readily crosses the dialyser membrane depending upon the concentration gradient. Emerging research demonstrates that HD sessions induce substantial alterations in glucose metabolism through multiple mechanisms: glucose removal into dialysate, improved insulin sensitivity, reduced insulin clearance, and decreased counter-regulatory hormone production.17,18 The intricate pathophysiological interplay between these conditions manifests in altered glucose homeostasis, insulin metabolism, and medication pharmacokinetics, necessitating specialized glycaemic monitoring approaches.19,20 These physiological perturbations potentially manifest as heightened glycaemic variability and increased hypoglycaemic risk, particularly during and immediately following dialysis sessions. Multiple studies have documented distinct glycaemic patterns between dialysis and non-dialysis days, with potentially higher hypoglycaemic risk during nocturnal periods20–22 (Table 8).

The pharmacological management of diabetes in MHD patients presents unique challenges, with both insulin therapy and OHAs demonstrating specific advantages and limitations. While insulin remains the predominant therapeutic modality in this population, its complex pharmacokinetics in renal failure and dialysis may contribute to enhanced glycaemic variability.20 Conversely, select OHAs with minimal renal clearance, such as linagliptin and teneligliptin present potential alternatives, though their comparative effects on glycaemic variability patterns remain incompletely characterized in current literature.21

Our findings demonstrate that both insulin and linagliptin-treated patients exhibit significantly higher mean glucose concentrations and increased glycaemic variability during HD days compared to non-dialysis days. Absolute glycaemic variability as measured by standard deviation was significantly higher on dialysis days in both groups, with a more pronounced effect in insulin-treated patients (p=0.037). This observation is consistent with previous research by Bomholt et al. (2023) who documented higher mean glucose levels and standard deviations on dialysis days in their cohort of 25 diabetic haemodialysis patients.19 Similarly, Mirani et al. (2010) reported increased glycaemic variability metrics during dialysis days in their study of 10 insulin-treated patients.10 However, our study extends these findings by demonstrating a treatment-modality effect, with insulin-treated patients exhibiting a significantly greater dialysis-associated hyperglycaemic response (mean difference: 39.8mg/dL versus 28.7mg/dL, p=0.043) and higher absolute glycaemic variability compared to linagliptin-treated patients.

The physiological mechanisms underlying haemodialysis-associated hyperglycaemia and increased glycaemic variability are multifactorial. Abe et al. (2007) proposed that dialysis-induced inflammatory cytokine release and counter-regulatory hormone secretion contribute to transient insulin resistance.2 Additionally, Kazempour-Ardebili et al. (2009) documented significantly reduced insulin clearance during HD in their study on 17 diabetic patients, potentially leading to hyperinsulinemia and subsequent post-dialysis hypoglycaemia.14 Our observation of treatment-modality differences in dialysis-associated glycaemic perturbations suggests that exogenous insulin administration may amplify these physiological responses compared to oral agents with different pharmacokinetic profiles.

The nocturnal hypoglycaemic burden observed in our study warrants particular attention. Both treatment groups demonstrated significantly lower mean glucose concentrations and higher hypoglycaemic risk during nocturnal periods, with insulin-treated patients exhibiting a significantly higher nocturnal hypoglycaemic burden compared to linagliptin-treated patients (12.4% versus 9.8% of monitoring time, p=0.047). These findings highlight the complex dynamics of hypoglycaemic risk in diabetic-haemodialysis patients and underscore the importance of treatment-specific risk assessment and preventive strategies, particularly during high-risk periods such as nocturnal hours and early intradialytic phases. This finding aligns with the results of Jung et al. (2010), who documented frequent nocturnal hypoglycaemic events in their cohort of 20 haemodialysis patients with type 2 diabetes, particularly in insulin-treated individuals.11 The heightened nocturnal hypoglycaemic risk may reflect the combined effects of reduced caloric intake during sleep, prolonged post-dialysis insulin sensitivity, and the pharmacokinetic properties of intermediate or long-acting insulin formulations that peak during night-time hours.

Intradialytic glycaemic variability revealed interesting patterns, with both treatment groups demonstrating progressive glycaemic stabilization from early to late intradialytic periods. This finding is consistent with observations by Hissa et al. (2021), who documented decreasing glycaemic variability during the course of HD sessions in their study of 14 diabetic patients.8 However, our study further identified a treatment-modality effect on this stabilization pattern, with linagliptin-treated patients exhibiting a more pronounced reduction in glycaemic variability during late intradialytic periods compared to insulin-treated patients (SD reduction: 20.9mg/dL versus 5.4mg/dL, p=0.028). This differential response may reflect the distinct pharmacokinetic alterations of oral agents versus insulin during HD, as well as potential differences in counter-regulatory responses to dialysis-induced glucose shifts.

The mechanisms underlying higher glycaemic variability with insulin therapy in MHD patients are likely multifactorial. Hayashi et al. (2021) documented that haemodialysis alters insulin receptor sensitivity patterns and insulin clearance rates, potentially leading to unpredictable insulin action.4 In contrast, select oral agents with stable pharmacokinetic profiles and glucose-dependent mechanisms of action, such as dipeptidyl peptidase-4 (DPP-4) inhibitors may confer more predictable glycaemic responses in this population.23,24

The absence of significant correlations between traditional clinical parameters (age, diabetes duration, dialysis vintage, BMI) and glycaemic variability metrics in our study deserves consideration. Vos et al. (2012) similarly reported poor predictive value of conventional clinical parameters for glycaemic variability in their study including 15 diabetic CKD patients.1

The consistent pattern of higher glycaemic variability metrics in insulin-treated patients across multiple contexts suggests a fundamental treatment-associated difference in glycaemic stability that may have important implications for individualized diabetes management in HD patients. Our findings have significant clinical implications. First, the documented differences in treatment modalities concerning glycaemic variability offer an evidence-based foundation for the selection of individualized antidiabetic regimens. For patients experiencing significant glycaemic fluctuations while on insulin therapy, transitioning to suitable oral agents with minimal renal clearance, such as linagliptin may enhance glycaemic stability. Secondly, the identification of high-risk periods for hypoglycaemia, specifically during nocturnal hours and early intradialytic phases in insulin-treated patients, necessitates specific preventive strategies. These strategies may include bedtime snacks, reduced overnight insulin doses, and glucose monitoring with appropriate insulin adjustments. Thirdly, the observed progressive glycaemic stabilization during HD sessions suggests that administering insulin post-dialysis might mitigate the risk of hypoglycaemia without compromising overall glycaemic control. Finally, our findings underscore the value of CGM as a clinical tool for identifying patient-specific glycaemic patterns and treatment-associated risk profiles that remain undetected with conventional monitoring approaches.

This study demonstrated significant differences in glycaemic variability based on antidiabetic treatment modality in HD population. Patients receiving insulin therapy exhibited significantly higher glycaemic variability compared to those on linagliptin. Both treatment groups demonstrated increased glycaemic variability during HD days compared to non-dialysis days, with insulin-treated patients showing a more pronounced dialysis-associated glycaemic perturbation. Notably, insulin therapy was associated with significantly higher nocturnal hypoglycaemia periods. The differential patterns of glycaemic variability between treatment modalities underscore the need for tailored antidiabetic regimens for HD patients, potentially favouring OHA therapies in select patients to minimize glycaemic fluctuations and nocturnal hypoglycaemic risk.

AA and VJ conceived the idea and designed the study. AA, RR and RG collected the data and did statistical analysis. AA, VJ and MGG did literature search. AA, RR, RG and MGG prepared the manuscript. VJ reviewed the paper and approved the final version of it.

This study was conducted in accordance with the Declaration of Helsinki and was approved by the Institutional Ethics Committee of SRM Medical College Hospital and Research Centre (Approval Number: SRMIEC-ST0923-1472). Written informed consent was obtained from all participants prior to their inclusion in the study. All participants were fully informed about the study's aims, procedures, potential risks, and their right to withdraw at any time without any consequences.

Patient consent was taken from all of the patients included in the study.

Not applicable.

Not applicable.

There is no funding (institutional, private and/or corporate financial support) for the work reported in their manuscript.

The authors declare no competing interests.

All data related to this study and statistical analysis are available with the first author and corresponding author. Data can be provided based on request.