Occurrence of benzodiazepines and nonbenzodiazepines usage among newly-onset of peritoneal dialysis versus hemodialysis patients Authors: Wen-Teng Lee, Mingchih Chen, Yu-Wei Fang, Wei-Shan Chang, Kai-Yuan Hsiao, Ben-Chang Shia, Hung-Hsiang Liou, Ming-Hsien Tsai Published: 2025 ABSTRACT BackgroundThe use of central nervous system depressants, including benzodiazepines (BZDs) and nonbenzodiazepine hypnotics (Z-drugs), is increasingly observed in chronic dialysis patients. Hemodialysis (HD) and peritoneal dialysis (PD) patients differ in physiological stress and neuropsychiatric profiles, yet both groups are exposed to medications such as BZDs and Z-drugs, whose adverse effects including sedation, cognitive impairment, and increased mortality, may further compound the psychological vulnerability and survival risks inherent to these dialysis populations. However, comparative patterns of their use across dialysis modalities remain insufficiently understood.MethodsIn this retrospective study, we enrolled patients newly diagnosed with end-stage renal disease who met the criteria for catastrophic illness in the National Health Insurance Research Database (NHIRD) between January 1, 2007, and December 31, 2017. PD and HD patients were propensity score–matched in a 1:3 ratio. Kaplan–Meier methods were used to estimate the new-onset BZDs and Z-drugs usage rate and to construct survival curves. Cox proportional hazards regression models were applied to calculate hazard ratios (HRs) and 95% confidence intervals (CIs) for new-onset BZDs and Z-drugs use between PD and HD groups. Follow-up ended no later than December 31, 2020.ResultsThe study compares BZDs and Z-drugs usage between PD and HD patients across three different time intervals: within 90 days, within 270 days, and after 270 days. Initially, PD patients had a significantly lower incidence of BZDs usage [aHR 0.858 (95% CI 0.740–0.993)]. However, this advantage diminished over time and reversed after 270 days, while PD patients showing higher rates of BZDs usage [aHR 1.446 (95% CI 1.286–1.627)]. In contrast, Z-drugs usage rate showed no statistically significant difference between the two groups across all intervals.ConclusionsWhen compared to patients on HD, those on PD exhibit a gradual rise in BZDs usage over time, which raises concerns about concurrent psychological symptoms. Further research is required to explore the neuropsychological mechanisms underlying psychological problems in patients on long-term PD.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12882-025-04656-5. Introduction Patients undergoing chronic dialysis often face a complex interplay of physiological and neuropsychiatric stressors that contribute to the use of central nervous system (CNS) depressants, particularly benzodiazepines (BZDs) and nonbenzodiazepine hypnotics (Z-drugs) [1–3]. Notably, the use of BZDs and Z-drugs has been linked to adverse effects such as sedation, confusion, and impaired cognition, which may further heighten the vulnerability of dialysis patients. Previous studies have highlighted safety concerns even with short-term use of these agents in end-stage renal disease (ESRD) populations [4, 5]. Hemodialysis (HD) patients frequently experience abrupt fluid and electrolyte shifts, intradialytic hypotension [6–9], and metabolic fluctuations, which could induce alternations in patient awareness, consequently affecting their physiological stability and leading to a deterioration in sleepiness [9–11]. These factors, along with repeated hospital interactions and reduced functional independence, may prompt the prescription of CNS depressants in this populations [12–15]. In contrast, PD patients typically experience fewer acute hemodynamic shifts [16] but face distinct challenges, disrupted circadian rhythms, cognitive decline, and depressive symptoms [17]. Modality-specific factors, such as nocturnal alarms from automated peritoneal dialysis systems, may further interfere with rest and psychological adaptation, contributing to cumulative neuropsychiatric burden over time. Despite growing use of BZDs and Z-drugs in dialysis populations [2, 18], limited evidence exists regarding their differential usage patterns across dialysis modalities and over time. In practice, BZDs and Z-drugs are prescribed based on physician judgment considering age, comorbidities, and symptom burden. Prescribing patterns also vary across countries and guidelines; for example, BZDs are not strongly recommended in elderly patients, and Japan has reported much higher prescribing rates among dialysis patients compared to other regions [19]. Understanding how these agents are prescribed in PD versus HD patients may reflect evolving clinical and psychological needs. Our study aims to compare the incidence and time-dependent trends of BZDs and Z-drugs usage between PD and HD patients using Taiwan’s National Health Insurance Research Database (NHIRD), offering real-world evidence to support more informed and modality-specific prescribing practices. Methods Data source The NHIRD serves as a robust asset for appraising chronic illnesses and medical interventions or occurrences from Taiwan. This repository encompasses patient attributes such as age and sex, as well as data on outpatient visits, hospital admissions, diagnoses, and medical procedures. In adherence to the principles of the Declaration of Helsinki, this study was conducted rigorously and was approved by the Ethics Committee of Shin-Kong Wu Ho-Su Memorial Hospital (IRB No. 20221102R). Study design and study population Our study method’s flowchart is shown in Fig. 1. We enrolled patients with newly ESRD who met the criteria for NHIRD catastrophic illness certification between January 1, 2007, and December 31, 2017. The index date was defined as the first date of maintenance dialysis recorded in NHIRD after catastrophic illness certification for ESRD. To ensure that enrolled patients were indeed receiving maintenance dialysis, we required at least 24 HD sessions within 3 months for the HD group, and for the PD group, at least one PD record in each of three consecutive months within the first 3 months after the index date, reflecting ongoing treatment rather than isolated sessions. Our exclusion criteria were as follows: (1) Age < 20 years, (2) Mortality within the 6 months after the index date, (3) Conversion from PD to HD for any reason, (4) Kidney transplantation performed before the index date, (5) Neuropsychiatric associated comorbidities including cerebrovascular accident, cancer, dementia, depression, Parkinson’s disease, schizophrenia, and epilepsy within 1 year before the index date, (6) Prescription of BZDs or Z-drugs within 6 months before the index date, and (7) Missing information. After applying these criteria, the remaining participants from the PD and HD groups are entered into the propensity score matching (1:3 PSM) with sex, age, and Charlson comorbidity index (CCI) score. In terms of follow-up duration, participants were observed for a maximum of 1,080 days after the index date. The tracking endpoint was set no later than December 31, 2020. Fig. 1Study flowchart illustrating patient selection from the National Health Insurance Research Database (NHIRD) Study flowchart illustrating patient selection from the National Health Insurance Research Database (NHIRD) Covariates Baseline covariates for both PD and HD patients were extracted from claims records, including 30 comorbidities, the Charlson comorbidity index (CCI), and confounding drugs, all of which are presented in Table 1. The coding for comorbidities, CCI, drugs, and outcomes was performed per the ninth and tenth revisions of the International Classification of Diseases (ICD-9 and ICD-10, respectively). The comorbidities, CCI, and outcome coding details are presented in Appendix 1. The drug coding utilized in our study is presented in Appendix 2, aligning with the Anatomical Therapeutic Chemical (ATC) classification system codes. Table 1The basic characteristics of the study population (after 1:3 PSM)VariablesPD (n = 3,661)HD (n = 10,983)p-value Basic characteristics Male2,033 (55.5)6,305 (57.4) 0.049 Age (Years)57.06 ± 13.6156.80 ± 13.930.323 Age group (Years)0.155 21–29105 (2.9)394 (3.6) 30–39274 (7.5)924 (8.4) 40–49769 (21.0)2,175 (19.8) 50–591,116 (30.5)3,286 (29.9) 60–69757 (20.7)2,277 (20.7) 70–79451 (12.3)1,361 (12.4) 80–89189 (5.2)566 (5.2) Underlying comorbidities Hypertension3,057 (83.5)8,974 (81.7) 0.015 Diabetes mellitus1,718 (46.9)5,193 (47.3)0.724 Hyperlipidemia1,864 (50.9)5,218 (47.5) < 0.001 Cardiovascular disease3,225 (88.1)9,482 (86.3) 0.007 Ischemic heart disease765 (20.9)2,350 (21.4)0.537 Congestive heart failure436 (11.9)1,335 (12.2)0.715 Peripheral vascular disease254 (6.9)670 (6.1)0.077 Atrial fibrillation58 (1.6)174 (1.6)1.000 Cerebral vascular disorder451 (12.3)1,326 (12.1)0.715 Asthma187 (5.1)544 (5.0)0.742 COPD222 (6.1)641 (5.8)0.641 Liver cirrhosis74 (2.0)258 (2.3)0.276 Mild liver disease375 (10.2)1,098 (10.0)0.692 Moderate or severe liver disease11 (0.3)30 (0.3)0.928 Peptic ulcer disease244 (6.7)713 (6.5)0.743 Ulcerative disease756 (20.7)2,302 (21.0)0.707 HBV14 (0.4)25 (0.2)0.165 HCV6 (0.2)15 (0.1)0.900 Constipation322 (8.8)1,130 (10.3) 0.010 Dementia27 (0.7)94 (0.9)0.562 Depression90 (2.5)251 (2.3)0.591 Hemiplegia22 (0.6)74 (0.7)0.723 GERD191 (5.2)611 (5.6)0.450 Hyperthyroidism30 (0.8)80 (0.7)0.658 Hypothyroidism40 (1.1)112 (1.0)0.778 SLE3 (0.1)17 (0.2)0.438 Gout1,088 (29.7)3,005 (27.4) 0.006 BPH25 (0.7)80 (0.7)0.865 Connective tissue disease157 (4.3)411 (3.7)0.152 Malignancy196 (5.4)505 (4.6)0.070 CCI score4.15 ± 2.214.11 ± 2.160.297 Confounding drug ACEI/ARB1,418 (38.7)4,303 (39.2)0.646 α-blockers1,253 (34.2)3,494 (31.8) 0.007 β-blockers1,697 (46.4)4,752 (43.3) 0.001 Calcium channel blockers2,664 (72.8)7,616 (69.3) < 0.001 Insulins658 (18.0)2,003 (18.2)0.738 Diuretics1,780 (48.6)6,101 (55.5) < 0.001 Lipid-lowering agents1,209 (33.0)3,323 (30.3) 0.002 Anti-coagulants1,064 (29.1)3,254 (29.6)0.530 Urate-lowering drugs1,166 (31.8)2,830 (25.8) < 0.001 NSAID484 (13.2)2,104 (19.2) < 0.001 Oral antidiabetics drugs277 (7.6)762 (6.9)0.213Abbreviation: PSM, propensity score matching; HD, hemodialysis; PD, peritoneal dialysis; COPD, chronic obstructive pulmonary disease; HBV, hepatitis B virus; HCV, hepatitis C virus; GERD, gastroesophageal reflux disease; SLE, systemic lupus erythematosus; BPH, benign prostate hypertrophy; CCI, Charlson comorbidity index. ACEI, angiotensin-converting enzyme inhibitors; ARB, angiotensin receptor blockers; NSAID, nonsteroidal anti-inflammatory drugs Table 1The basic characteristics of the study population (after 1:3 PSM)VariablesPD (n = 3,661)HD (n = 10,983)p-value Basic characteristics Male2,033 (55.5)6,305 (57.4) 0.049 Age (Years)57.06 ± 13.6156.80 ± 13.930.323 Age group (Years)0.155 21–29105 (2.9)394 (3.6) 30–39274 (7.5)924 (8.4) 40–49769 (21.0)2,175 (19.8) 50–591,116 (30.5)3,286 (29.9) 60–69757 (20.7)2,277 (20.7) 70–79451 (12.3)1,361 (12.4) 80–89189 (5.2)566 (5.2) Underlying comorbidities Hypertension3,057 (83.5)8,974 (81.7) 0.015 Diabetes mellitus1,718 (46.9)5,193 (47.3)0.724 Hyperlipidemia1,864 (50.9)5,218 (47.5) < 0.001 Cardiovascular disease3,225 (88.1)9,482 (86.3) 0.007 Ischemic heart disease765 (20.9)2,350 (21.4)0.537 Congestive heart failure436 (11.9)1,335 (12.2)0.715 Peripheral vascular disease254 (6.9)670 (6.1)0.077 Atrial fibrillation58 (1.6)174 (1.6)1.000 Cerebral vascular disorder451 (12.3)1,326 (12.1)0.715 Asthma187 (5.1)544 (5.0)0.742 COPD222 (6.1)641 (5.8)0.641 Liver cirrhosis74 (2.0)258 (2.3)0.276 Mild liver disease375 (10.2)1,098 (10.0)0.692 Moderate or severe liver disease11 (0.3)30 (0.3)0.928 Peptic ulcer disease244 (6.7)713 (6.5)0.743 Ulcerative disease756 (20.7)2,302 (21.0)0.707 HBV14 (0.4)25 (0.2)0.165 HCV6 (0.2)15 (0.1)0.900 Constipation322 (8.8)1,130 (10.3) 0.010 Dementia27 (0.7)94 (0.9)0.562 Depression90 (2.5)251 (2.3)0.591 Hemiplegia22 (0.6)74 (0.7)0.723 GERD191 (5.2)611 (5.6)0.450 Hyperthyroidism30 (0.8)80 (0.7)0.658 Hypothyroidism40 (1.1)112 (1.0)0.778 SLE3 (0.1)17 (0.2)0.438 Gout1,088 (29.7)3,005 (27.4) 0.006 BPH25 (0.7)80 (0.7)0.865 Connective tissue disease157 (4.3)411 (3.7)0.152 Malignancy196 (5.4)505 (4.6)0.070 CCI score4.15 ± 2.214.11 ± 2.160.297 Confounding drug ACEI/ARB1,418 (38.7)4,303 (39.2)0.646 α-blockers1,253 (34.2)3,494 (31.8) 0.007 β-blockers1,697 (46.4)4,752 (43.3) 0.001 Calcium channel blockers2,664 (72.8)7,616 (69.3) < 0.001 Insulins658 (18.0)2,003 (18.2)0.738 Diuretics1,780 (48.6)6,101 (55.5) < 0.001 Lipid-lowering agents1,209 (33.0)3,323 (30.3) 0.002 Anti-coagulants1,064 (29.1)3,254 (29.6)0.530 Urate-lowering drugs1,166 (31.8)2,830 (25.8) < 0.001 NSAID484 (13.2)2,104 (19.2) < 0.001 Oral antidiabetics drugs277 (7.6)762 (6.9)0.213Abbreviation: PSM, propensity score matching; HD, hemodialysis; PD, peritoneal dialysis; COPD, chronic obstructive pulmonary disease; HBV, hepatitis B virus; HCV, hepatitis C virus; GERD, gastroesophageal reflux disease; SLE, systemic lupus erythematosus; BPH, benign prostate hypertrophy; CCI, Charlson comorbidity index. ACEI, angiotensin-converting enzyme inhibitors; ARB, angiotensin receptor blockers; NSAID, nonsteroidal anti-inflammatory drugs The basic characteristics of the study population (after 1:3 PSM) Abbreviation: PSM, propensity score matching; HD, hemodialysis; PD, peritoneal dialysis; COPD, chronic obstructive pulmonary disease; HBV, hepatitis B virus; HCV, hepatitis C virus; GERD, gastroesophageal reflux disease; SLE, systemic lupus erythematosus; BPH, benign prostate hypertrophy; CCI, Charlson comorbidity index. ACEI, angiotensin-converting enzyme inhibitors; ARB, angiotensin receptor blockers; NSAID, nonsteroidal anti-inflammatory drugs Outcomes The study outcomes included the new-onset of BZDs and Z-drugs usage. We defined medication exposure as continuous use of either BZDs (e.g., midazolam, estazolam, alprazolam, diazepam, lorazepam) or Z-drugs (e.g., zolpidem, eszopiclone, zaleplon) for a minimum of 6 months from the patient’s index date, regardless of prescribing specialty. Analyses were conducted separately for BZDs and Z-drugs as class-specific outcomes, and patients could contribute to more than one analysis if initiation of different drug classes occurred at different times. Statistical analyses Categorical variables, including baseline characteristics, comorbidities, and concomitant medications, were presented as counts (percentages) and compared between groups using the chi-square test. Continuous variables were expressed as means with standard deviations (SDs) and compared using independent-sample t-tests. The new-onset of BZDs and Z-drugs usage rate were estimated using the Kaplan–Meier method and compared via the log-rank test. Cox proportional hazards regression models were used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for new-onset BZDs and Z-drugs usage between PD and HD groups. Models were adjusted for age, sex, comorbidities, and baseline medications that differed significantly after PSM. In cases where the proportional hazards assumption was violated, we performed time-stratified analyses, segmenting follow-up into intervals informed by both clinical plausibility (e.g., prescription cycle considerations) and inspection of survival curve patterns. All statistical analyses were performed using SAS software version 9.4 (SAS Institute Inc., Cary, NC, USA). A two-tailed p-value < 0.05 was considered statistically significant. Results Patient characteristics The baseline characteristics of study participants after PSM are shown in Table 1. A total of 3,661 PD patients and 10,983 HD patients were included. The mean age was 57.06 ± 13.61 years in the PD group and 56.80 ± 13.93 years in the HD group, with males comprising 55.5% and 57.4%, respectively. Although the difference in male proportion between PD and HD groups remained statistically significant (p = 0.049), the variation was minimal (< 2%) and was further adjusted for in the subsequent multivariable Cox analyses. In addition, some residual differences in comorbidities and concomitant medications were observed after matching: PD patients had higher rates of hypertension, hyperlipidemia, cardiovascular disease, and gout but lower rates of constipation, and more frequently used α-blockers, β-blockers, calcium channel blockers, lipid-lowering agents, and urate-lowering drugs, whereas HD patients more often received diuretics and nonsteroidal anti-inflammatory drugs. Comparisons of BZDs and Z-drugs use between PD and HD patients Table 2 and Fig. 2 summarize the temporal patterns of BZDs and Z-drugs usage in PD and HD patients. To account for differences in prescribing dynamics over time and proportional hazards violations, the analyses were stratified into three intervals: 0–90, 0–270, and 270–1,080 days. This stratification allowed us to capture both the early adjustment period after dialysis initiation and longer-term trends. For BZDs, distinct temporal changes were observed between the two groups. Within the first 90 days, PD patients showed a significantly lower initiation rate compared with HD patients (aHR 0.858; 95% CI, 0.740–0.993; Fig. 2B). This early advantage gradually diminished over the following months, with no significant difference detected by 270 days (aHR 0.983; 95% CI, 0.882–1.097; Fig. 2C). Beyond 270 days, however, the trend reversed: PD patients exhibited a higher incidence of BZDs use than their HD counterparts (aHR 1.446; 95% CI, 1.286–1.627; Fig. 2D). In contrast, no statistically significant differences in Z-drugs initiation were observed between PD and HD patients across all time intervals (Figs. 2E–H). Although numerical differences existed, these did not reach statistical significance, suggesting broadly comparable prescribing patterns for Z-drugs regardless of dialysis modality. Visual inspection of Kaplan–Meier survival curves corroborated these findings, showing deviations at approximately 90 days—consistent with the completion of a typical prescription cycle—and at around 270 days, where crossover between PD and HD groups became evident. These turning points supported the rationale for the chosen analytic intervals. Table 2Comparisons of BZDs and Z-drugs usage between PD and HDMedicationsPD (n = 3,661)HD (n = 10,983)PD versus HDEventsIncidence rate per 100 PYEventsIncidence rate per 100 PYHR(95% CI)aHR(95% CI) BZDs 0–90 days23125.9882831.35 0.830 (0.717–0.960) 0.858 (0.740–0.993) 0-270 days43217.091,33317.740.964 (0.865–1.075)0.983 (0.882–1.097)270–1080 days416 (3,229(a))4.58877 (9,653(a))3.17 1.447 (1.288–1.626) 1.446 (1.285–1.627) Z-drugs 0–90 days374.031555.650.714 (0.499–1.022)0.733 (0.511–1.052)0-270 days802.973093.840.773 (0.605–1.001)0.812 (0.633–1.040)270–1080 days122 (3,581(a))1.15297 (10,677(a))0.941.229 (0.995–1.517)1.262 (0.995–1.561)Abbreviations: BZD, benzodiazepines; Z-drug, non-benzodiazepines; HD, hemodialysis; PD, peritoneal dialysis; PY, person-year; HR, hazard ratio; aHR, adjusted hazard ratio; CI, confidence interval. (a) Represents the cumulative number of patients at risk for BZDs and Z-drug usage events after 270 daysNote that the adjusted hazard ratio (aHR) was calculated by accounting for underlying comorbidities and significant confounding drugs that differed statistically (p < 0.05) between the PD and HD groups from Table 1, including sex (male percentage), hypertension, hyperlipidemia, cardiovascular disease, constipation, gout, α-blockers, β-blockers, Calcium channel blockers (CCB), diuretics, lipid-lowering agents, uric acid lowering drugs and nonsteroidal anti-inflammatory drugs (NSAID) Table 2Comparisons of BZDs and Z-drugs usage between PD and HDMedicationsPD (n = 3,661)HD (n = 10,983)PD versus HDEventsIncidence rate per 100 PYEventsIncidence rate per 100 PYHR(95% CI)aHR(95% CI) BZDs 0–90 days23125.9882831.35 0.830 (0.717–0.960) 0.858 (0.740–0.993) 0-270 days43217.091,33317.740.964 (0.865–1.075)0.983 (0.882–1.097)270–1080 days416 (3,229(a))4.58877 (9,653(a))3.17 1.447 (1.288–1.626) 1.446 (1.285–1.627) Z-drugs 0–90 days374.031555.650.714 (0.499–1.022)0.733 (0.511–1.052)0-270 days802.973093.840.773 (0.605–1.001)0.812 (0.633–1.040)270–1080 days122 (3,581(a))1.15297 (10,677(a))0.941.229 (0.995–1.517)1.262 (0.995–1.561)Abbreviations: BZD, benzodiazepines; Z-drug, non-benzodiazepines; HD, hemodialysis; PD, peritoneal dialysis; PY, person-year; HR, hazard ratio; aHR, adjusted hazard ratio; CI, confidence interval. (a) Represents the cumulative number of patients at risk for BZDs and Z-drug usage events after 270 daysNote that the adjusted hazard ratio (aHR) was calculated by accounting for underlying comorbidities and significant confounding drugs that differed statistically (p < 0.05) between the PD and HD groups from Table 1, including sex (male percentage), hypertension, hyperlipidemia, cardiovascular disease, constipation, gout, α-blockers, β-blockers, Calcium channel blockers (CCB), diuretics, lipid-lowering agents, uric acid lowering drugs and nonsteroidal anti-inflammatory drugs (NSAID) Comparisons of BZDs and Z-drugs usage between PD and HD Abbreviations: BZD, benzodiazepines; Z-drug, non-benzodiazepines; HD, hemodialysis; PD, peritoneal dialysis; PY, person-year; HR, hazard ratio; aHR, adjusted hazard ratio; CI, confidence interval. (a) Represents the cumulative number of patients at risk for BZDs and Z-drug usage events after 270 days Note that the adjusted hazard ratio (aHR) was calculated by accounting for underlying comorbidities and significant confounding drugs that differed statistically (p < 0.05) between the PD and HD groups from Table 1, including sex (male percentage), hypertension, hyperlipidemia, cardiovascular disease, constipation, gout, α-blockers, β-blockers, Calcium channel blockers (CCB), diuretics, lipid-lowering agents, uric acid lowering drugs and nonsteroidal anti-inflammatory drugs (NSAID) Fig. 2Kaplan–Meier survival curves for new-onset BZDs usage during the entire follow-up period (A), 0–90 days (B), 0–270 days (C), and 270–1,080 days (D). Corresponding survival curves for Z-drugs usage are shown in panels E–H Kaplan–Meier survival curves for new-onset BZDs usage during the entire follow-up period (A), 0–90 days (B), 0–270 days (C), and 270–1,080 days (D). Corresponding survival curves for Z-drugs usage are shown in panels E–H Discussion Our study, based on real-world database analysis, reveals that PD is associated with a reduced incidence of new-onset BZDs usage within the first 90 days after the initiation of dialysis compared to HD. Nonetheless, this advantage progressively diminishes by the 270-day mark and ultimately transforms into a reversed outcome. In contrast, the rate of Z-drugs usage does not differ significantly between PD and HD patients. These findings suggest that clinicians should be cautious with long-term BZDs prescribing in PD, as the higher incidence may reflect accumulating psychological or systemic burdens. A minor residual imbalance in sex distribution remained after PSM, likely reflecting real-world demographic patterns of dialysis modality choice in Taiwan and the structural characteristics of the NHIRD dataset, where full elimination of small baseline variation is rarely achievable. Although causality cannot be established from claims data, these findings underscore the importance of cautious long-term BZDs prescribing in PD and support the need for regular medication review and deprescribing considerations, particularly in long-term PD care. BZDs and Z-drugs are widely prescribed CNS depressants with distinct pharmacological profiles. BZDs act non-selectively on multiple GABA-A receptor α subunits (e.g., α1 for sedation, α2 for anxiolysis, α5 for cognitive impairment), while Z-drugs primarily target the α1 subunit, producing sedative effects with fewer anxiolytic or cognitive impacts [20, 21]. Initially developed as safer alternatives to BZDs, Z-drugs still carry risks of dependence, abuse, and withdrawal. In clinical practice, BZDs may be preferred when comorbid anxiety or depression is present [22, 23], whereas Z-drugs are often favored for their rapid onset and shorter half-life. These pharmacodynamic differences may influence prescribing decisions across dialysis modalities, particularly as PD and HD patients differ in psychological burden, treatment autonomy, and exposure to systemic stressors. In contrast, the absence of significant differences in Z-drugs use between PD and HD may reflect their common indication as short-term hypnotics, with rapid onset, short half-life, and relatively lower risk of long-term dependence. These pharmacological properties likely lead to similar prescribing patterns across modalities, explaining the comparable usage observed in our study. Although HD and PD differ in physiology and care settings, comparative data on BZDs and Z-drugs use remain limited [24, 25]. Prior studies suggest that PD patients may experience greater psychiatric burden, including depression and affective distress, which could shape long-term prescribing behaviors [4, 26–28]. Differences in dialysis schedules and self-management demands may also contribute to distinct prescribing patterns between modalities [29]. Some studies indicate that cytokines play a role in depression [30, 31] and sleep regulation [32, 33], which may help explain the higher prevalence of BZDs use among PD patients. In particular, TNF-α and IL-1β promote non-rapid-eye-movement sleep under inflammatory conditions [34], and acute peritonitis or prolonged exposure to high-glucose dialysates in PD can further elevate cytokine activity [35, 36]. Activation of the NLRP3 inflammasome and increased IL-1β levels have also been linked to sleep disturbance [37]. These mechanisms may contribute to the temporal inversion observed around 270 days, when BZDs use in PD patients surpassed that in HD patients. Chronic dialysis patients often present with aging-related comorbidities such as hypertension and metabolic syndrome, which may increase vulnerability to neuropsychiatric symptoms. In our study, PD patients showed a higher long-term incidence of BZDs use compared with HD. This trend may reflect underlying psychological or systemic conditions, together with the chronic inflammatory state in PD that could elevate psychiatric risk, underscoring the need for greater caution in prescribing. Beyond biological mechanisms, non-biological aspects such as heavier self-management demands, social isolation and long-term psychological stress in PD may also contribute to these patterns. While not directly testable in claims data, these hypotheses are supported by prior literature and warrant further prospective validation. Our study possesses some notable strengths, one of which is the use of the NHIRD, a database that distinguishes itself by providing comprehensive coverage of Taiwan’s entire population (approximately 23 million residents). Furthermore, patients with ESRD received extensive coverage under the National Health Insurance program, with dialysis being nearly classified as a catastrophic condition. This extensive coverage provided us with a representative dataset that offers valuable insights into various aspects of dialysis, including trends, comorbidities and medication screening. In our study, we focused on BZDs and Z-drugs prescriptions rather than diagnostic codes, our study better reflects real-world prescribing patterns and clinical decision-making in the dialysis populations, particularly in differentiating usage behaviors between HD and PD patients. Methodologically, the application of PSM allowed for better control of baseline differences, while stratified time-interval analyses enabled us to examine both short-term and long-term prescribing trends. Together, these strengths provide a robust framework for evaluating how dialysis modality influences longitudinal CNS depressant use in clinical practice. However, our study also had some limitations. First, reliance on claims data limits access to clinical detail, including body mass index, disease severity, nutritional status, medication dosage, drug adherence, and residual kidney function (RKF). As PD is often associated with longer RKF preservation, direct measures of RKF (e.g., urine volume or clearance) are unavailable, and unmeasured RKF differences may partly explain modality-specific prescribing. Indication-level information and symptom severity at the time of prescribing are also not reliably captured, and diagnostic codes at the time of prescription were not used to infer clinical indication, as they are frequently billing-related rather than symptom-based in administrative datasets and may introduce greater misclassification than clarity. Although NHIRD records all reimbursed prescriptions across care settings (dialysis centers, hospitals, and community clinics/family physicians), cash-paid and over-the-counter medications are not captured; however, BZDs and most Z-drugs are prescription-only and typically reimbursed, so under-ascertainment is likely limited. Hard outcomes (e.g., mortality) were not analyzed in this study, and subgroup analyses by age, sex, or comorbidity/psychiatric burden were also not performed. Both aspects may provide additional insights and should be explored in future research. Second, among initially excluded patients, a sizable proportion had prior BZDs or Z-drugs use; whether they differ systematically from the included cohort remains uncertain, which may introduce selection bias. The smaller proportion of PD patients reflects incident ESRD identification, modality confirmation, and 1:3 matching; while cohorts are nationally representative, this distribution may affect generalizability. Residual imbalances in comorbidities and medications after PSM could still confound modality-specific prescribing patterns. Third, concurrent prescriptions or switching between BZDs and Z-drugs during follow-up could not be fully distinguished, which may have introduced some misclassification, although this is unlikely to materially alter the observed modality-specific patterns. Fourth, evidence regarding drug–drug interactions, pharmacokinetics, efficacy, and safety of psychotropic agents in ESRD remains limited, constraining causal interpretation. In addition, diagnostic codes prior to prescription were not used to define clinical indications, as they may not reliably capture actual patient conditions and could further increase misclassification risk. Finally, psychiatric comorbidities (e.g., depression, anxiety) are often under-coded in administrative data, complicating attribution of indications for BZDs or Z-drugs initiation. Mechanistic explanations discussed in the manuscript (biological and psychosocial) should therefore be viewed as hypotheses grounded in prior literature rather than conclusions supported directly by claims measures. Conclusion Our study findings suggest that patients on PD initially exhibit a lower incidence of new-onset of BZDs usage, which gradually changes to a higher incidence over an extended period. The utilization of BZDs in dialysis patients may signify the presence of concurrent psychological symptoms such as anxiety or even depression. The neuropsychological mechanisms underlying psychosocial symptoms among patients on long-term PD patients require further investigation. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1 Supplementary Material 2 Supplementary Material 2