Effectiveness of Nursing and Non‐Pharmacological Interventions in Reducing Intradialytic Hypotension Among Haemodialysis Patients: A Systematic Review and Meta‐Analysis Authors: Mi‐Kyoung Cho, Yoon Hee Cho Published: 2026 ABSTRACT ABSTRACTAimThis study aimed to systematically evaluate the effectiveness of non‐pharmacological interventions in reducing intradialytic hypotension among adult patients undergoing haemodialysis and to examine their implications for nursing practice.DesignSystematic review and meta‐analysis.Data SourcesPubMed, EMBASE, CINAHL, SCOPUS, DBpia and RISS were searched for studies published between January 2000 and July 2025.Review MethodsRandomized controlled trials and quasi‐experimental studies evaluating non‐pharmacological interventions were included. Methodological quality was assessed using Joanna Briggs Institute critical appraisal tools. Random‐effects meta‐analysis and subgroup analyses were conducted and certainty of evidence was evaluated using GRADE.ResultsSixteen studies were included in the meta‐analysis. Non‐pharmacological interventions were associated with modest improvements in systolic blood pressure during haemodialysis, although substantial heterogeneity was observed. Dialysis technology‐based interventions showed a significant improvement in systolic blood pressure, whereas nurse‐led interventions showed a numerically similar effect but did not reach statistical significance. The overall certainty of evidence was very low.ConclusionNon‐pharmacological interventions may support haemodynamic stability during haemodialysis; however, improvement in systolic blood pressure represents a surrogate haemodynamic outcome and does not necessarily indicate a reduced incidence of intradialytic hypotension. Further well‐designed studies using clinically meaningful and patient‐centred outcomes are needed.Implications for the Profession and/or Patient CareNurses can contribute to intradialytic hypotension prevention through assessment, monitoring, early recognition of haemodynamic instability, timely bedside interventions and patient education.ImpactThe findings highlight the potential complementary roles of nurse‐led care and dialysis technology‐based approaches in the individualized management of adults undergoing haemodialysis.Reporting MethodPRISMA 2020.No Patient or Public ContributionPatients or members of the public were not involved because this study was a systematic review and meta‐analysis of previously published studies.Trial and Protocol RegistrationPROSPERO (CRD420251150743) Introduction Intradialytic hypotension (IDH) is one of the most common complications among haemodialysis patients, occurring in approximately 15%–50% of all dialysis sessions (Shoji et al. 2004; Flythe et al. 2015). Beyond causing transient discomfort, IDH can result in serious clinical consequences, including the accumulation of metabolic toxins due to inadequate dialysis, a decline in residual renal function, myocardial ischemia and neurological injury caused by cerebral hypoperfusion (Shoji et al. 2004; McIntyre 2010; Burton et al. 2009). Recurrent episodes of hypotension further increase patients' fatigue and anxiety, often necessitating early termination of dialysis sessions or adjustment of ultrafiltration volume, thereby reducing overall treatment efficiency (Shoji et al. 2004; Flythe et al. 2015; Lindsay et al. 2006; Kanbay et al. 2020). Over time, these complications contribute to cardiovascular morbidity, higher hospitalization rates and increased mortality (Shoji et al. 2004). Consequently, IDH has been recognized as a major determinant of both prognosis and health‐related quality of life in haemodialysis patients (Lindsay et al. 2006; Kanbay et al. 2020; Wang, Yao, et al. 2023). Although systolic blood pressure (SBP) is the most commonly reported hemodynamic outcome in studies evaluating interventions for IDH, it represents a surrogate marker of hemodynamic stability rather than a direct measure of clinically important outcomes, such as the incidence of IDH episodes, treatment interruption or patient quality of life (Kanbay et al. 2020; Wang, Yao, et al. 2023). Nevertheless, SBP remains the most consistently reported hemodynamic outcome in intervention studies of IDH and therefore provides a practical and comparable measure for quantitative synthesis in meta‐analysis (Kanbay et al. 2020). The pathophysiology of IDH is multifactorial (Habas et al. 2025). Excessive fluid removal during haemodialysis leads to a reduction in plasma volume, while inadequate cardiovascular compensatory responses further contribute to the development of hypotension. In addition, various physiological and environmental factors—including impaired autonomic function, underlying cardiovascular disease, postprandial state, vasodilation and high dialysate sodium concentration—interact to increase susceptibility to IDH (Habas et al. 2025). The condition occurs more frequently among older adults and patients with diabetic nephropathy, heart failure or autonomic neuropathy, owing to decreased vascular compliance and blunted sympathetic responses (Hryciw et al. 2021). Therefore, IDH cannot be explained by a single mechanism, and prevention of its occurrence should include both pharmacological and non‐pharmacological strategies. Over the years, various strategies have been implemented to prevent IDH in clinical settings. Pharmacological approaches such as vasopressors (e.g., midodrine) and plasma expanders (e.g., albumin, mannitol) have been used (Kanbay et al. 2020; Hryciw et al. 2021), but their potential side effects and cost have hindered their widespread use (Kanbay et al. 2020; Costa‐Pinto et al. 2023). In contrast, non‐pharmacological interventions have been studied as safer and more sustainable approaches from both nursing and technical perspectives. Several systematic reviews and meta‐analyses have examined specific non‐pharmacological strategies for IDH, such as cold dialysate (Subhash et al. 2025), sodium profiling (Dunne 2017) or biofeedback‐assisted dialysis (Mambelli et al. 2024). However, these reviews have typically focused on single technical or device‐based approaches and have not comprehensively synthesized the full spectrum of non‐pharmacological interventions. Furthermore, existing evidence syntheses have not explicitly distinguished interventions initiated, delivered or managed by nurses during routine haemodialysis care. Despite these diverse approaches, comprehensive reviews of the effectiveness of non‐pharmacological interventions for preventing IDH remain inadequate (Kanbay et al. 2020; Hamrahian et al. 2023). Most existing studies have been conducted at single centres with small sample sizes, limiting the generalizability of their findings. Moreover, heterogeneity in intervention methods, intensity, timing and outcome measures has led to inconsistent results across studies (Mc Causland and Charytan 2024). Many of these studies were also quasi‐experimental in design, lacking methodological rigour such as randomization and blinding, which further limits the strength of the evidence (Hamrahian et al. 2023; Peyronel et al. 2020). From a nursing perspective, IDH is not merely a physiological phenomenon. The occurrence of hypotension during haemodialysis is closely related to nurses' clinical judgement, intervention skills and monitoring ability, and its prevention constitutes a core component of nursing practice (Ozen and Cepken 2020; McIntyre et al. 2021). Therefore, it is essential to comprehensively examine IDH prevention strategies from a nursing perspective and to establish a foundation for evidence‐based practice (Peyronel et al. 2020). Against this background, the present study aims to systematically analyse the effectiveness of interventions for preventing hypotension in haemodialysis patients by synthesizing findings from randomized and quasi‐experimental studies. Furthermore, it seeks to examine the effects according to intervention type and study design, thereby providing clinically applicable evidence for future practice. In particular, this study is significant in that it comprehensively compares and analyzes various non‐pharmacological interventions, offering a scientific foundation for integrated nursing and dialysis care aimed at improving patient safety and treatment efficiency. In this review, nonpharmacological interventions refer to all strategies that prevent or alleviate IDH without the use of medications, including dialysis regimen modifications, device‐based approaches and behavioural or physical interventions. For the purposes of this review, nurse‐led interventions were defined as nonpharmacological interventions that nurses independently initiate, directly deliver or continuously manage during haemodialysis sessions. These include patient assessment and monitoring, positioning, application of compression devices, patient education and symptom‐triggered responses based on clinical judgement. These interventions highlight the role of nursing practice within the broader category of nonpharmacological strategies. Methods Eligibility Criteria The population was restricted to adults (≥ 18 years) receiving maintenance haemodialysis for at least 3 months. Eligible interventions comprised non‐pharmacological strategies intended to reduce the occurrence or severity of IDH. The Comparison groups included participants receiving standard care, alternative interventions not classified as non‐pharmacological or sham procedures. The primary outcome was SBP, while secondary outcomes included intradialytic reduction in SBP, mean arterial pressure (MAP), diastolic blood pressure (DBP), intradialytic weight gain (IDWG) and symptom scores related to IDH. Given the heterogeneity in the definitions of IDH across the included studies, this review adopted author‐defined criteria for IDH, typically involving symptoms and/or SBP thresholds (e.g., ≥ 20 mmHg drops or nadir SBP < 90 mmHg), without applying a uniform operational definition. For studies with more than one intervention arm (study IDs: 3, 10, 12, 13 and 16), each intervention group was analysed independently to compute individual effect sizes. Both randomized controlled trials (RCTs) and quasi‐experimental studies were eligible because non‐pharmacological interventions for IDH have been evaluated using both designs in clinical haemodialysis settings. Quantitative synthesis was limited to studies providing sufficient information to derive an effect size, including sample sizes, means, standard deviations or other relevant statistics. Ineligible studies were those involving: (1) participants younger than 18 years; (2) patients with acute renal failure or those receiving peritoneal dialysis; (3) interventions involving pharmacological treatment (e.g., midodrine or sertraline) or studies without blood pressure outcomes; (4) an effect estimate could not be derived from the reported data; (5) full text was unavailable, the report was an unpublished manuscript or conference abstract or the publication language was neither English nor Korean; or (6) only the incidence or frequency of IDH was reported, without pre‐ and post‐intervention blood pressure measurements. Information Sources and Search Strategy Six electronic databases—PubMed, EMBASE, CINAHL, SCOPUS, DBpia and RISS—were searched for relevant studies published from January 1, 2000, through July 31, 2025. The search was restricted to publications in English or Korean. The database search was carried out from September 19 to October 2, 2025. The search strategy was designed to identify experimental studies evaluating the effects of non‐pharmacological interventions on blood pressure among patients receiving haemodialysis. For PubMed, MeSH terms and free‐text keywords were combined with the Boolean operators AND and OR. Search concepts were organized around the PICO‐SD components: (1) Population: “renal dialysis”[MeSH Terms] OR “haemodialysis”[Other Term] OR “dialysis”[Other Term]; Intervention (I): “Nursing Care”[MeSH Terms] OR “Patient Care Planning”[MeSH Terms] OR “non‐pharmacological”[Other Term] OR “non‐drug”[Other Term] OR “cool dialysate”[Other Term] OR “dialysate temperature”[Other Term] OR “ultrafiltration profiling”[Other Term] OR “UF profiling”[Other Term] OR “sodium modelling”[Other Term] OR “positioning”[Other Term] OR “biofeedback”[Other Term] OR “exercise”[Other Term] OR “stretching”[Other Term] OR “aromatherapy”[Other Term]; Outcome (O): “hypotension”[MeSH Terms] OR “intradialytic hypotension”[Other Term] OR “dialysis hypotension”[Other Term] OR “IDH”[Other Term]; Study Design (SD): “randomized controlled trial”[MeSH Terms] OR “Non‐Randomized Controlled Trials as Topic”[MeSH Terms] OR “clinical trial”[MeSH Terms] OR “Controlled Clinical Trial”[MeSH Terms]. The search was limited to human studies involving adults and older adults, published in English or Korean between January 1, 2000 and July 31, 2025. Search terms were tailored for each database to ensure compatibility with its indexing schema and retrieval functions. Database‐specific search details are presented in Table S1. Both authors contributed to the development and implementation of the searches. The review was prospectively registered in PROSPERO (CRD420251150743), where the protocol record is publicly available: https://www.crd.york.ac.uk/PROSPERO/view/CRD420251150743. Selection Process Study selection followed PRISMA 2020 (Page et al. 2021). Titles and abstracts of records retrieved from the databases were independently reviewed by two reviewers for potential eligibility. Differences in screening decisions were settled through discussion, with a third reviewer involved when agreement could not be achieved. Screening and selection were managed manually in Microsoft Excel 365 (Microsoft Corporation, Redmond, WA, USA). The software was used to manage records, remove duplicates and organize citations for screening. Records retained after title and abstract screening were subsequently reviewed in full text to determine their eligibility. Reference lists of pertinent systematic reviews were also examined for additional studies not retrieved through the database searches. Study selection was performed without automated or machine‐learning–based screening tools. Figure 1 summarizes the study selection pathway in accordance with the PRISMA 2020 flow diagram. PRISMA flow diagram. Data Collection Process and Data Items Extracted information was organized in a structured Microsoft Excel spreadsheet (Microsoft Corporation, Redmond, WA, USA). One reviewer entered the data into the extraction form, and a second reviewer independently checked the entries for accuracy. Data collection covered four categories: study characteristics (publication year, country, centre, study design, IRB approval and funding), sample characteristics (total N and group allocation), intervention characteristics (type, duration, number of sessions, outcome measurement timing and control group intervention) and outcome variables. For meta‐analysis, post‐intervention means, SDs and group‐specific Ns were collected for SBP (primary) and for secondary outcomes including intradialytic reduction in SBP, MAP, DBP, IDWG and symptom scores. SBP was designated as the primary outcome because it provided sufficient continuous data across studies for quantitative evaluation of the hemodynamic response to non‐pharmacological interventions using standardized mean differences. Study Risk of Bias Assessment Risk of bias was appraised with the Joanna Briggs Institute (JBI) Critical Appraisal Checklists specific to randomized controlled trials (RCTs) (Barker et al. 2023) and quasi‐experimental studies (Quasi‐E) (Tanriver‐Ayder et al. 2021). Prior to formal appraisal, two reviewers reviewed the guidance notes and reached consensus on the interpretation of each item. To ensure consistency in scoring, one RCT and one Quasi‐E study were independently evaluated as pilot cases. The two reviewers subsequently completed separate risk‐of‐bias appraisals. The JBI checklists contain 13 items for RCTs and nine for Quasi‐E studies, each evaluated as “Yes,” “No,” “Unclear,” or “Not Applicable.” For scoring purposes, a “Yes” response was assigned one point, whereas all other responses received zero points. Item scores were then summed to derive an overall quality score for each study. Effect Measures Hedges' g was selected as the effect measure, with corresponding 95% confidence intervals (CIs) reported for each estimate (Barker et al. 2024; Tanriver‐Ayder et al. 2021). Hedges' g was selected due to its bias correction, which was particularly advantageous when synthesizing results from trials with relatively small sample sizes (Murad et al. 2019). For each comparison, the effect estimate was derived from post‐intervention data. Synthesis Methods Pooled estimates were obtained under a random‐effects framework. Between‐study variance (τ 2) was estimated using restricted maximum likelihood (REML) (Murad et al. 2019). Primary analyses were performed using IBM SPSS Statistics for Windows, Version 30.0 (IBM Corp., Armonk, NY, USA). Hedges' g was used as the standardized mean difference (SMD), and each estimate was accompanied by a 95% confidence interval (CI) (Lin and Aloe 2021). Hedges' g was chosen to reduce small‐sample bias (Lin and Aloe 2021; Taylor and Alanazi 2023). For multi‐arm studies in which intervention arms shared a common control group, each arm was analysed as a separate comparison. To account for the statistical dependency introduced by shared control participants, the variance (vi) for comparisons originating from the same study was adjusted in accordance with recommendations from the Cochrane Handbook. This adjustment was implemented using the metafor package (version 4.4–0, Maastricht University, Netherlands) in R, which was used exclusively for variance correction in multi‐arm studies and for calculating 95% confidence intervals for heterogeneity indices. Statistical significance was set at p < 0.05 (Flechner and Tseng 2011). Heterogeneity was characterized using I 2 and Cochran's Q (Ruppar 2020), with the magnitude of I 2 interpreted according to the reference values proposed by Higgins et al. (Ruppar 2020). The 95% confidence intervals for I 2 were estimated using the meta package in R (version 4.5.1). Subgroup analyses and meta‐regression were conducted to identify potential sources of the observed heterogeneity. Subgroup analyses were conducted according to prespecified study characteristics, including publication year, region, intervention type, study design and methodological quality. Study design was included to examine the contribution of differences between RCTs and Quasi‐E studies to heterogeneity. Additionally, to explore whether intervention delivery characteristics influenced treatment effects, interventions were further classified into nurse‐led interventions and dialysis technology‐based interventions. Nurse‐led interventions were defined as interventions that could be initiated and delivered by dialysis nurses without modification of dialysis prescriptions or dialysis machine settings. Dialysis technology‐based interventions were defined as interventions requiring modification of dialysis machine settings, dialysis prescriptions or physician‐prescribed dialysis treatment. The same REML random‐effects approach was retained for subgroup estimates. Sensitivity analysis comprised three approaches: (1) sequential omission of individual studies (leave‐one‐out analysis); (2) reanalysis after removal of outliers detected with the Galbraith plot; and (3) comparison of the primary REML estimate with estimates from the DerSimonian–Laird random‐effects and fixed‐effects models. Reporting Bias Assessment Assessment of publication bias was restricted to the primary outcome (SBP), for which more than 10 studies were available. Funnel plot asymmetry was examined graphically and with Egger's regression test. Duval and Tweedie's trim‐and‐fill procedure was additionally used to estimate the impact of potentially missing studies on the pooled effect (Duval and Tweedie 2000). Certainty of Evidence Evidence certainty was determined for each outcome according to the GRADE framework (Guyatt et al. 2008). Ratings were informed by five domains—risk of bias, inconsistency, indirectness, imprecision and publication bias—and assigned to one of four levels: high, moderate, low or very low. Two reviewers completed the GRADE evaluation independently, with differences in judgement settled by consensus. Results Study Selection Screening was therefore undertaken for 582 records, of which 547 did not proceed beyond title and abstract review. The remaining 35 records underwent full‐text review, and 19 were removed because the reported data did not permit calculation of an effect size. The quantitative synthesis consequently comprised 16 articles. Five articles (study IDs 3, 10, 12, 13 and 16) contributed more than one eligible intervention arm; these arms were analysed separately, resulting in 26 effect‐size comparisons. Figure 1 summarizes the selection process in the PRISMA 2020 flow diagram. Study Characteristics The meta‐analysis comprised 16 studies published from 2000 to 2025; nine appeared after 2016 and ten originated from Asian countries. Twelve studies recruited participants from a single centre, while 15 reported Institutional Review Board (IRB) approval. Twelve studies disclosed receiving external funding. Study designs comprised four quasi‐experimental crossover studies, six crossover RCTs and six parallel‐group RCTs. Eight studies included sample sizes of 50 or more participants. The types of interventions varied: three studies applied dialysate cooling, six utilized sodium profiling and seven investigated other non‐pharmacological strategies such as adjusting dialysis frequency, UF profiling, acupoint or exercise‐based interventions. The methodological quality scores ranged from 6 to 11 (Table 1). TABLE 1Descriptive summary of the included studies.Study IDAuthorYearCountryCentresIRBFundResearch designParticipantsCharacteristics of participantsEG intervention methodsSessionsCG intervention methodOutcomeQuality score1Arasnezhad et al 2024 IranSingle‐centreYesYesRCT (cross‐over) N = 16 (E:16, C:16)Maintenance HD ≥ 6 months with > 20% IDH episodes in the past monthA/D‐UF profiling + Na profiling16 sessionsConventional HD (standard sodium profile)BP, MAP, IDH episode, sodium change, Wt change92McCausland et al. 2022 USASingle‐centreYesYesRCT N = 139 (E:69, C:70)Hospitalized, maintenance HD ≥ 90 daysLower DNa 138 mmol/L≤ 6 sessionsHigher DNa 142 mmol/LBP, IDH episode113Ebrahimi et al. 2017 IranSingle‐centreYesYesRCT (cross‐over) N = 80 (Ea:80, Eb:80, Ec:80, C:80)Maintenance HD ≥ 6 monthsEa: 35°C + Na 138 mmol/L, Eb: 37°C+ Na ramping, Ec: 35°C + Na ramping3 sessionsConventional HD (37°C, Na 138 mmol/L)BP, IDH episode, clinical symptoms114Garg et al. 2022 Canada84 centresYesYesCluster‐RCT N = 15,413 (E:8000, C:7413)Maintenance HD ≥ 90 days at participating centresPersonalized cooler dialysateNot reportedConventional HD (standard dialysate 36.5°C)BP, Cardiovascular death, hospitalization115Gil et al. 2014 Korea9 centresYesYesQuasi‐E (cross‐over) N = 60 (E:60, C:60)Maintenance HD > 3 months with ≥ 25% IDH episodes in the past monthUF profiling (blood‐volume feedback)‐hemocontrol biofeedback system24 sessionsConventional HDBP, MAP, IDH episode, interventions, pre/post HD Wt, IDWG, URR, electrolytes76Guo et al. 2025 ChinaSingle‐centreYesNot reportedRCT N = 99 (E:50, C:49)Maintenance HD ≥ 3 months with ≥ 3 IDH episodes in the past 4 weeksIntermittent haemodiafiltration36 sessionsConventional HDBP, MAP, IDH episode, Haemoglobin, Albumin, side effects107Kim et al. 2014 KoreaSingle‐centreYesNot reportedQuasi‐E (cross‐over) N = 24 (E:24, C:24)Maintenance HD ≥ 2 years, urine output ≤ 100 mL/dayLower DNa 135 mmol/LAbout 90 sessionsConventional HD (DNa 140 me/L)BP, weight gain, IDH episode, muscle cramps, headache68Kim and Kim 2018 KoreaSingle‐centreYesNoRCT N = 59 (E:30, C:29)Maintenance HD, ≥ 2 IDH in last 3 HDElastic compression stockings1 sessionUsual care (HD without compression stockings)BP, PR, Hypotensive symptom score89Kim et al. 2019 KoreaSingle‐centreYesYesQuasi‐E (cross‐over) N = 11 (Ea:11, Eb:11, C:11)HD ≥ 3 months, BMI ≤ 35, stable BPEa: Aerobic exercise, Eb: Resistance exercise1 sessionUsual care (HD without exercise)BP during exercise, post 20 h ambulatory BP810Kotanko et al. 2015 USA, Canada21 centresYesYesRCT N = 245 (Ea:125, Eb:45, Ca:120, Cb:42)Maintenance HDEa: frequent daily HD, Eb: frequent nocturnal HDAbout 312 sessionsConventional HD (3 times/week)BP, antihypertensive drug use1011Maggiore et al. 2002 9 countries in Europe27 centresYesYesRCT (cross‐over) N = 95 (E:95, C:95)Maintenance HD ≥ 6 months with ≥ 25% symptomatic IDH in the past monthIsothermal dialysis12 sessionsThermoneutral dialysisBP, HR, IDH episode, interventions, URR, electrolytes1012Meira et al. 2010 Brazilsingle‐centreYesYes RCT (cross‐over) N = 22 (Ea:22, Eb:22, C:22)Maintenance HD ≥ 3 months with ≥ 2 episodes of IDH or cramps in the preceding monthEa: step sodium profile, Eb: linear sodium profile12 sessionsFixed Na 139 mmol/LBP, Wt change, IDWG813Ribeiro et al. 2021 BrazilSingle‐centreYesYes RCT (cross‐over) N = 8 (Ea:8, Eb:8, C:8) Maintenance HD ≥ 3 months (short‐daily HD, 5 sessions/week)Ea: Low‐intensity isometric Handgrip Exercise, Eb: Moderate‐intensity isometric Handgrip Exercise1 sessionUsual care (HD without exercise)BP, HR, HRV, myocardial oxygen consumption index714Tsai et al. 2018 TaiwanSingle‐centreYesYesRCT N = 27 (E:14, C:13)HD ≥ 3 months with ≥ 15% IDH in past 2 monthsHerbal Acupoint Therapy (HAT) patch12 sessionsSham‐HATBP, fatigue, symptomatic IDH episode, early termination rate1015Zhou et al. 2006 ChinaSingle‐centreYesNot reported RCT (cross‐over) N = 8 (Ea:8, Eb:8, Ec:8, C:8) Maintenance HD ≥ 6 months with ≥ 22% IDH in past 3 monthsEa: Na profile, Eb: UF profile Ec: Na + UF profile10 sessionsFixed Na 138 mmol/L + constant UFIDH episode, RBV change, BP, HR, stroke volume, cardiac output716Park et al. 2000 KoreaSingle‐centreNot reportedNot reportedQuasi‐E (cross‐over) N = 14 (Ea:14, Eb:14, C:14) HD ≥ 3 months with prior IDH experienceEa: Hypernatremic HD (Na 147 me/L), Eb:Sequential hypernatremic HD (Na 147 → 144 → 140 me/L)6 sessionsConventional HD (DNa 137 mmol/L)BP, PR, IDH episode, UFR, IDWG, electrolytes8 Note: Superscripts a, b, and c were divided by intervention.Abbreviations: A/D‐UF, Ascending/Descending Ultrafiltration; BP, blood pressure; C and CG, control group; DNa, dialysate sodium; E and EG, experimental group; HD, haemodialysis; HR, heart rate; HRV, heart rate variability; IDH, intradialytic hypotension; IRB, institutional review board; MAP, mean arterial pressure; PR, pulse rate; Quasi‐E, quasi‐experimental; RBV, relative blood volume; RCT, randomized controlled trial; UF, Ultrafiltration; UFR, ultrafiltration rate. Descriptive summary of the included studies. RCT (cross‐over) RCT (cross‐over) N = 8 (Ea:8, Eb:8, C:8) RCT (cross‐over) N = 8 (Ea:8, Eb:8, Ec:8, C:8) N = 14 (Ea:14, Eb:14, C:14) Note: Superscripts a, b, and c were divided by intervention. Abbreviations: A/D‐UF, Ascending/Descending Ultrafiltration; BP, blood pressure; C and CG, control group; DNa, dialysate sodium; E and EG, experimental group; HD, haemodialysis; HR, heart rate; HRV, heart rate variability; IDH, intradialytic hypotension; IRB, institutional review board; MAP, mean arterial pressure; PR, pulse rate; Quasi‐E, quasi‐experimental; RBV, relative blood volume; RCT, randomized controlled trial; UF, Ultrafiltration; UFR, ultrafiltration rate. Risk of Bias in Studies Of the 16 studies, 12 were RCTs, with a mean JBI quality score of 9.33 on the 13‐item appraisal tool. The highest score observed was 11 and the lowest was 7. The most frequently satisfied items were Item 11 (intention‐to‐treat analysis) and Item 12 (appropriate statistical analysis), both of which were reported in 11 of the 12 RCTs. In contrast, Item 5, concerning blinding of intervention deliverers, was unclear across all RCTs. Only four RCTs met the criterion for allocation concealment (Item 2) and three for participant blinding (Item 4). Four studies were quasi‐experimental designs, with a mean quality score of 7.25 out of 9 (range: 6–8). Most of the items were generally well addressed, except for Item 8 (reliable outcome measurement), which was not reported in any of the quasi‐experimental studies (Table 2). TABLE 2Quality assessment of the included studies.Study IDJoanna Briggs Institute of critical appraisal tools checklist for randomized controlled trialsTotal scoreAuthor (year)Study design123456789101112131Arasnezhad et al. (2024) RCT (cross‐over) 101001011111192McCausland et al. (2022)RCT1101011111111113Ebrahimi et al. (2017) RCT (cross‐over) 1011011111111114Garg et al. (2022)Cluster RCT1110011111111116Guo et al. (2025)RCT1110011101111108Kim and Kim (2018)RCT1010010101111810Kotanko et al. (2015)RCT00100000101111011Maggiore et al. (2002) RCT (cross‐over) 11100111011111012Meira et al. (2010) RCT (cross‐over) 1010010101111813Ribeiro et al. (2021) RCT (cross‐over) 1010001011110714Tsai et al. (2018)RCT10110111011111015Zhou et al. (2006) RCT (cross‐over) 10100101011107Total1141130107106111212109.33Study IDJoanna Briggs Institute of critical appraisal tools checklist for quasi‐experimental studiesTotal scoreAuthor (year)Study design1234567895Gil et al. (2014)Crossover Quasi‐E11110110177Kim et al. (2014)Crossover Quasi‐E10101110169Kim et al. (2019)Crossover Quasi‐E111111101816Park et al. (2000)Crossover Quasi‐E1111111018Total4343344047.25Abbreviations: RCT, randomized controlled trial; Quasi‐E, quasi‐experimental. Quality assessment of the included studies. RCT (cross‐over) RCT (cross‐over) RCT (cross‐over) RCT (cross‐over) RCT (cross‐over) RCT (cross‐over) Abbreviations: RCT, randomized controlled trial; Quasi‐E, quasi‐experimental. Results of Syntheses Effect of Non‐Pharmacological Interventions on SBP The primary analysis of SBP comprised 11 studies, providing 17 comparisons of non‐pharmacological interventions for IDH prevention during haemodialysis. SBP was significantly higher following non‐pharmacological intervention than in the control condition (Hedges' g = 0.31, 95% CI: 0.06 to 0.56; p = 0.015). Heterogeneity was substantial (Q = 95.29, df = 16, p < 0.001; I 2 = 83.0%, 95% CI: 74.1 to 88.9) (Figure 2). Across studies, adverse effects directly attributable to the interventions were infrequently reported. When noted, they were generally mild (e.g., thirst, cold sensation, cramping) and transient, suggesting that the interventions were overall well tolerated. Forest plot of the effects of non‐pharmacological interventions on systolic blood pressure. ES, Effect size; SE, Standard error; CI, Confidence interval. Superscripts a, b, and c were divided by intervention. Subgroup analyses by intervention type: (A) nurse‐led interventions and (B) dialysis technology‐based interventions. Subgroup Analysis of Non‐Pharmacological Interventions on SBP According to Study Characteristics Given the substantial heterogeneity, subgroup analyses examined variation in effect estimates by publication year, region, number of study centres, IRB approval, funding status, study design, sample size, intervention type, intervention duration, number of sessions, timing of outcome measurement, comparator type and methodological quality. The pooled estimates favoured the intervention in studies published after 2016 (Hedges' g = 0.59, 95% CI: 0.34 to 0.85), studies conducted in Asia (Hedges' g = 0.53, 95% CI: 0.33 to 0.72), and single‐centre studies (Hedges' g = 0.42, 95% CI: 0.19 to 0.65). Estimates were also significant for studies with IRB approval (Hedges' g = 0.33, 95% CI: 0.04 to 0.62; p = 0.024) and those with no funding or with funding status not reported (Hedges' g = 0.44, 95% CI: 0.15 to 0.73; p = 0.003). Across study‐design categories, statistical significance was reached only for quasi‐experimental crossover studies (Hedges' g = 0.36, 95% CI: 0.10 to 0.62; p = 0.008). For nurse‐led interventions, Hedges' g was 0.30, with a 95% CI from −0.07 to 0.66 and no observed heterogeneity (I 2 = 0.0%). For dialysis technology‐based interventions, Hedges' g was 0.34, with a 95% CI from 0.01 to 0.68; heterogeneity was high (I 2 = 86.9%). In terms of intervention types, significant effects were found in studies evaluating increased dialysis frequency, intermittent haemodiafiltration (IHDF) and UF profiling, although a single study represented each. In contrast, cooling dialysate, sodium profiling and stimulation‐based interventions (e.g., exercise, acupoint stimulation) did not yield statistically significant pooled effects. For interventions lasting less than 2 months, Hedges' g was 0.39 (95% CI: 0.13 to 0.66). The estimate was 0.45 for interventions delivered in fewer than 12 sessions (95% CI: 0.23 to 0.67) and 0.28 for outcomes assessed immediately after intervention (95% CI: 0.08 to 0.48). For studies using another non‐pharmacological intervention as the comparator, Hedges' g was 0.58 (95% CI: 0.29 to 0.88) (Table 3). TABLE 3Subgroup analysis of non‐pharmacological interventions on systolic blood pressure by study characteristics.VariablesCategory K Study ID n ES95% CIZ p Lower limitUpper limitYear< 201685, 7, 10a, 10b, 12a, 12b, 16a, 16b 608−0.01−0.290.27−0.100.921≥ 201691, 3a, 3b, 3c, 6, 8, 13a, 13b, 145490.590.340.854.51< 0.001CountryAsia111, 3a, 3b, 3c, 5, 6, 7, 8, 14, 16a, 16b 7350.530.330.725.37< 0.001America & Europe610a, 10b, 12a, 12b, 13a, 13b 422−0.23−0.490.03−1.760.078CentreSingle141, 3a, 3b, 3c, 6, 7, 8, 12a, 12b, 13a, 13b, 14, 16a, 16b 7050.420.190.653.61< 0.001Multi‐centres35, 10a, 10b 452−0.09−0.670.50−0.290.772IRBNot reported216a,16b 420.21−0.310.740.790.427Yes151, 3a, 3b, 3c, 5, 6, 7, 8, 10a, 10b, 12a, 12b, 13a, 13b, 1411150.330.040.622.250.024FundNo or not reported56, 7, 8, 16a, 16b 2480.440.150.733.000.003Yes121, 3a, 3b, 3c, 5, 10a, 10b, 12a, 12b, 13a, 13b, 149690.32−0.070.721.600.110Research designQuasi‐E (crossover)45, 7, 16a, 16b 2100.360.100.622.670.008RCT (crossover)81, 3a, 3b, 3c, 12a, 12b, 13a, 13b 4300.10−0.390.580.400.691RCT56, 8, 10a, 10b, 145170.71−0.171.591.590.113Participants< 5091, 7, 12a, 12b, 13a, 13b, 14, 16a, 16b 2270.45−0.231.131.310.192≥ 5083a, 3b, 3c, 5, 6, 8, 10a, 10b 9300.33−0.010.681.890.059Intervention delivery characteristicsNurse‐led interventions48, 13a, 13b, 141100.30−0.070.661.590.111Dialysis technology‐based interventions131, 3a, 3b, 3c, 5, 6, 7, 10a, 10b, 12a, 12b, 16a, 16b10470.340.010.681.990.047Intervention typeDialysate temperature23a, 3c 2400.54−0.051.131.780.074Frequency210a, 10b 332−0.41−0.70−0.11−2.730.006IHDF16990.800.391.213.84< 0.001Sodium profiling71, 3b, 7, 12a, 12b, 16a, 16b 3360.71−0.411.831.250.211UF profiling (blood‐volume feedback)151200.480.120.842.590.010Nurse‐led interventions48, 13a, 13b, 141100.30−0.070.661.590.111Intervention duration< 2 months121, 3a, 3b, 3c, 8, 12a, 12b, 13a, 13b, 14, 16a, 16b 5610.390.130.662.910.004≥ 2 months55, 6, 7, 10a, 10b 5990.15−0.330.630.610.543Intervention session< 12 sessions93a, 3b, 3c, 8, 13a, 13b, 14, 16a, 16b 4720.450.230.673.99< 0.001≥ 12 sessions81, 5, 6, 7, 10a, 10b, 12a, 12b 6850.52−0.451.491.060.290Outcome measurement timeImmediately111, 5, 7, 8, 12a, 12b, 13a, 13b, 14, 16a, 16b 4060.280.080.482.680.007Delayed63a, 3b, 3c, 6, 10a, 10b 7510.30−0.150.761.300.194Control group interventionNo131, 3a, 5, 6, 8, 10a, 10b, 12a, 12b, 13a, 13b, 16a, 16b 9220.25−0.070.571.560.119Yes43b, 3c, 7, 143150.580.290.883.87< 0.001Quality scores< Mean81, 5, 7, 8, 12a, 12b, 13a, 13b 3370.62−0.261.501.380.167≥ Mean93a, 3b, 3c, 6, 10a, 10b, 14, 16a, 16b 8200.27−0.060.611.590.113 Note: Superscripts a, b and c were divided by intervention.Abbreviations: CI, confidence interval; ES, effect size; IHDF, intermittent haemodiafiltration; IRB, institutional review board; K, number of effect‐size comparisons; N, number of participants; Quasi‐E, quasi‐experimental study; RCT, randomized controlled trials; UF, ultrafiltration. Subgroup analysis of non‐pharmacological interventions on systolic blood pressure by study characteristics. Note: Superscripts a, b and c were divided by intervention. Abbreviations: CI, confidence interval; ES, effect size; IHDF, intermittent haemodiafiltration; IRB, institutional review board; K, number of effect‐size comparisons; N, number of participants; Quasi‐E, quasi‐experimental study; RCT, randomized controlled trials; UF, ultrafiltration. Meta‐Regression Analysis of Non‐Pharmacological Interventions on SBP Meta‐regression was subsequently performed to determine whether study characteristics could explain the observed heterogeneity. Studies published before 2016 showed a significantly smaller pooled effect size compared with those published after 2016 (Estimate = −0.61, p = 0.042). Similarly, studies conducted in America and Europe demonstrated significantly smaller pooled effect sizes than those conducted in Asia (Estimate = −0.75, p = 0.008). No other covariates were statistically significant in the meta‐regression model (Table 4). TABLE 4Meta‐regression analysis of non‐pharmacological interventions on systolic blood pressure.Covariates (Ref.)EstimateSE95% CI t p Lower limitUpper limitYear (Ref.: ≥ 2016)−0.610.28−1.20−0.02−2.220.042Country (Ref.: Asia)−0.750.24−1.28−0.23−3.080.008Centre (Ref.: Multi)0.510.38−0.301.321.350.197IRB (Ref.: Yes)−0.110.61−1.411.18−0.190.885Fund (Ref.: Yes)0.120.40−0.730.960.290.777Quasi‐E research design (Ref.: RCT (crossover))−0.190.47−1.200.83−0.390.700RCT research design (Ref.: RCT (crossover))−0.390.42−1.300.52−0.920.371Participants (Ref.: ≥ 50)−0.040.37−0.830.75−0.120.910Intervention type (Ref.: Sodium profiling)−0.070.48−1.131.00−0.130.896Intervention type (Ref.: Dialysate temperature)−0.250.56−1.490.98−0.450.660Intervention type (Ref.: UF profiling (blood‐volume feedback))−0.200.70−1.741.34−0.280.783Intervention type (Ref.: IHDF)−0.520.72−2.101.06−0.730.482Intervention type (Ref.: Frequency)0.660.56−0.581.901.170.268Intervention duration (Ref.: ≥ 2 months)0.260.37−0.541.050.690.499Intervention session (Ref.: ≥ 12 sessions)0.200.35−0.550.950.560.582Outcome measurement time (Ref.: Delayed)0.030.37−0.770.820.070.948Control group intervention (Ref.: Other intervention)−0.290.39−1.130.54−0.750.464Quality score (Ref.: ≥ Mean)0.110.38−0.690.910.300.772Abbreviations: CI, confidence interval; IHDF, intermittent haemodiafiltration; IRB, institutional review board; Quasi‐E, quasi‐experimental study; RCT, randomized controlled trials; Ref., reference; SE, standard error; UF, ultrafiltration. Meta‐regression analysis of non‐pharmacological interventions on systolic blood pressure. Abbreviations: CI, confidence interval; IHDF, intermittent haemodiafiltration; IRB, institutional review board; Quasi‐E, quasi‐experimental study; RCT, randomized controlled trials; Ref., reference; SE, standard error; UF, ultrafiltration. Sensitivity Analyses of Non‐Pharmacological Interventions on SBP To evaluate the robustness of the results, multiple sensitivity analyses were performed, including the leave‐one‐out method, exclusion of outliers identified through the Galbraith plot and the use of alternative meta‐analytic models (Table 5). Sequential omission of individual studies produced no material change in the pooled estimate (Table 5). Five comparisons (study IDs 1, 3b, 3c, 6 and 10a) fell outside the 95% confidence limits on the Galbraith plot (Figure 3). Reanalysis without these comparisons yielded a slightly smaller effect, which remained statistically significant and retained the same direction. Estimates obtained under the REML random‐effects, DerSimonian–Laird random‐effects and fixed‐effects models were similar, with statistical significance maintained across all three models. TABLE 5Sensitivity test of non‐pharmacological interventions on systolic blood pressure.Leave‐one‐out sensitivity analysisStudy ID K ES95% CI Z p Lower limitUpper limit1160.250.020.472.170.0303a160.320.050.602.290.0223b160.280.020.552.090.0363c160.260.010.512.050.0405160.300.030.572.170.0306160.270.020.532.080.0377160.320.050.592.310.0218160.310.040.582.240.02510a160.380.160.593.43< 0.00110b160.350.090.612.630.00812a160.340.070.602.470.01312b160.350.090.612.600.00913a160.320.060.592.400.01713b160.310.050.602.290.02214160.320.060.592.380.01716a160.320.050.582.320.02016b160.320.060.592.360.018Sensitivity analysis after excluding outliers identified from the Galbraith plotStudy ID K ES95% CI Z p Lower limitUpper limitOutliers (1, 3b, 3c, 6, 10a)51.26−0.482.991.420.156Inliers (3a, 5, 7, 8, 10b, 12a, 12b, 13a, 13b, 14, 16a, 16b)120.180.010.342.040.041Sensitivity analysis by meta‐analytic modelModel (estimate) K ES95% CI Z p Lower limitUpper limitRandom effect model (REML)170.310.060.562.430.015Random effect model (DerSimonian–Laird)170.330.040.612.210.027Fixed effect model (REML)170.250.140.354.42< 0.001 Note: Superscripts a, b and c were divided by intervention.Abbreviations: CI, confidence interval; ES, effect size; K, number of effect‐size comparisons; REML, restricted maximum likelihood. Sensitivity test of non‐pharmacological interventions on systolic blood pressure. Note: Superscripts a, b and c were divided by intervention. Abbreviations: CI, confidence interval; ES, effect size; K, number of effect‐size comparisons; REML, restricted maximum likelihood. Galbrith plot of non‐pharmacological intervention on systolic blood pressure among haemodialysis patients. Precision = 1/standard error; 0.05 limit line = 95% confidence limit. Secondary Outcomes of Non‐Pharmacological Interventions Beyond the primary outcome of SBP, this meta‐analysis also assessed five additional clinical parameters (Table 6). Although intradialytic reduction in SBP pertains to the same construct as the primary outcome, it was reported as a change score rather than a post‐intervention value. It thus could not be pooled with the SBP data used in the primary analysis. Therefore, it was analysed separately. The pooled effect size for intradialytic SBP reduction was Hedges' g = 0.13 (95% CI: 0.29 to 0.89), indicating a statistically significant improvement, despite substantial heterogeneity (Q = 300.39, p < 0.001; I 2 = 98.7%, 95% CI: 98.0% to 99.1%). A significant increase was also observed in DBP (Hedges' g = 0.39, 95% CI: 0.12 to 0.65), with a high degree of heterogeneity (I 2 = 80.2%, 95% CI: 68.2% to 87.7%). In contrast, no significant effects were found for MAP, IDWG and symptom scores. TABLE 6The effect of non‐pharmacological interventions on other variables.Variables K Study ID N Overall ES95% CI Z (p) Q (p) I 2 (95% CI)Lower limitUpper limitIntradialytic reduction in SBP52, 4, 9a, 9b, 1115,7750.13−1.211.460.19 (0.852)300.39 (< 0.001)98.7 (98.0%, 99.1%)DBP151, 3a, 3b, 3c, 5, 6, 7, 8, 10a, 10b, 12a, 12b, 14, 16a, 16b 11570.390.120.652.87 (0.004)70.78 (< 0.001)80.2 (68.2%, 87.7%)MAP61, 5, 6, 15a, 15b, 15c 2710.90−0.282.081.50 (0.134)30.21 (< 0.001)83.4 (65.3%, 92.1%)IDWG91, 5, 7, 10a, 10b, 12a, 12b, 16a, 16b 628−0.13−0.500.25−0.65 (0.513)49.71 (< 0.001)83.9 (71.1%, 91.1%)Symptom32, 5, 8318−0.25−0.690.19−1.12 (0.265)6.59 (0.037)69.7 (0.0%, 91.1%) Note: Superscripts a, b and c were divided by intervention.Abbreviations: CI, confidence interval; DBP, diastolic blood pressure; ES, effect size; IDWG, intradialytic weight gain; K, number of effect‐size comparisons; MAP, mean arterial pressure; N, number of participants; SBP, systolic blood pressure. The effect of non‐pharmacological interventions on other variables. Note: Superscripts a, b and c were divided by intervention. Abbreviations: CI, confidence interval; DBP, diastolic blood pressure; ES, effect size; IDWG, intradialytic weight gain; K, number of effect‐size comparisons; MAP, mean arterial pressure; N, number of participants; SBP, systolic blood pressure. Reporting Biases For SBP, the funnel plot showed no marked asymmetry (Figure 4). Egger's regression yielded an intercept of −0.43 (t = −1.40, p = 0.183), providing no statistical indication of funnel plot asymmetry (Table 7). The trim‐and‐fill analysis estimated no missing studies; consequently, the adjusted Hedges' g remained 0.31 (95% CI: 0.06 to 0.56), identical to the original estimate. Thus, the analyses provided no indication of publication bias for the pooled SBP estimate. Publication bias was not examined for the secondary outcomes because fewer than 10 studies contributed to these analyses. For SBP, the concordant findings from the funnel plot, Egger's test and trim‐and‐fill analysis provided no indication that the pooled estimate was affected by publication bias. Funnel plot of non‐pharmacological interventions on systolic blood pressure among haemodialysis patients. Precision = 1/standard error; 0.05 limit line = 95% confidence limit. TABLE 7Publication bias test of non‐pharmacological interventions on systolic blood pressure.Egger's regression testCoeffSE95% CI t p Lower limitUpper limitIntercept−0.430.3−1.070.22−1.400.183Slope2.730.990.634.832.770.014Trim and fill method K Hg95% CI Z p Lower limitUpper limitOriginal170.310.060.562.430.015Trim and fill correction170.310.060.562.430.015 Note: Egger's regression test for zero intercept.Abbreviations: CI, confidence interval; Coeff, coefficient; K, number of effect‐size comparisons; SE, standard error. Publication bias test of non‐pharmacological interventions on systolic blood pressure. Note: Egger's regression test for zero intercept. Abbreviations: CI, confidence interval; Coeff, coefficient; K, number of effect‐size comparisons; SE, standard error. Certainty of Evidence GRADE rated the certainty of evidence as very low for all outcomes. The reasons for this rating varied by outcome and involved risk of bias, inconsistency, indirectness and imprecision (Table 8). Detailed GRADE assessments and reasons for downgrading were presented in Table S2. TABLE 8Summary of findings and certainty of evidence (GRADE).OutcomeStudies (comparisons)Effect estimate (Hedges' g, 95% CI)Risk of biasInconsistencyIndirectnessImprecisionPublication biasCertainty of evidenceSBP (primary outcome)11 (17)0.31 (0.06 to 0.56)SeriousVery seriousSeriousSeriousNo evidence of publication bias⊕◯◯◯ Very lowIntradialytic reduction in SBP4 (5)0.13 (−1.21 to 1.46)SeriousVery seriousSeriousVery seriousNo evidence of publication bias⊕◯◯◯ Very lowDBP10 (15)0.39 (0.12 to 0.65)SeriousSeriousSeriousSeriousNo evidence of publication bias⊕◯◯◯ Very lowMAP4 (6)0.90 (−0.28 to 2.08)SeriousSeriousSeriousVery seriousNo evidence of publication bias⊕◯◯◯ Very lowIntradialytic weight gain6 (9)−0.13 (−0.50 to 0.25)SeriousSeriousSeriousSeriousNo evidence of publication bias⊕◯◯◯ Very lowIDH‐related symptom scores3 (3)−0.25 (−0.69 to 0.19)SeriousSeriousNot seriousVery seriousNo evidence of publication bias⊕◯◯◯ Very low Note: Studies are presented as the number of included studies (number of effect‐size comparisons).Abbreviations: CI, confidence interval; DBP, diastolic blood pressure; GRADE, Grading of Recommendations Assessment, Development and Evaluation; IDH, intradialytic hypotension; IDWG, intradialytic weight gain; MAP, mean arterial pressure; SBP, systolic blood pressure. Summary of findings and certainty of evidence (GRADE). Note: Studies are presented as the number of included studies (number of effect‐size comparisons). Abbreviations: CI, confidence interval; DBP, diastolic blood pressure; GRADE, Grading of Recommendations Assessment, Development and Evaluation; IDH, intradialytic hypotension; IDWG, intradialytic weight gain; MAP, mean arterial pressure; SBP, systolic blood pressure. Discussion This meta‐analysis systematically synthesized evidence from randomized and quasi‐experimental studies to evaluate the effectiveness of non‐pharmacological interventions in preventing IDH among haemodialysis patients. The interventions examined in this review included cool dialysate, sodium and UF profiling, biofeedback dialysis and nurse‐led approaches. The findings indicated that non‐pharmacological interventions were associated with a statistically significant improvement in SBP during haemodialysis (Hedges' g = 0.31, 95% CI 0.06–0.56, p = 0.015). The use of SBP as the primary outcome was based on its consistent reporting across the included studies, which enabled quantitative comparison of intervention effects. However, SBP represents a surrogate hemodynamic outcome and its improvement does not necessarily indicate a reduced incidence of IDH. Although substantial heterogeneity was observed (I 2 = 83.0%), the overall pooled effect favoured the intervention groups. Nevertheless, this heterogeneity reduces confidence in the pooled effect estimates and suggests that the overall findings should be interpreted with appropriate caution. According to previous studies, cooling the dialysate induces peripheral vasoconstriction and helps maintain blood pressure during haemodialysis. Subhash et al. (Subhash et al. 2025) also demonstrated in their meta‐analysis that cool dialysate reduced the incidence of IDH and improved SBP stability, which aligns with the present findings. However, some studies also reported adverse reactions, such as shivering and discomfort among elderly patients (Subhash et al. 2025). These adverse effects may be attributed to variations in target dialysate temperature, thermoregulatory capacity and intervention duration. These findings suggest that individualized dialysate temperature adjustment based on patient tolerance is warranted. UF and sodium profiling were also found to have beneficial effects on blood pressure control, consistent with the results of Ertuglu et al. (Ertuglu et al. 2021) and Kanbay et al. (Kanbay et al. 2020). These strategies alleviate rapid fluid shifts and osmotic disequilibrium, contributing to circulatory stability. Furthermore, biofeedback dialysis—which continuously monitors changes in blood volume—was associated with improved cardiac output and reduced IDH incidence, corroborating the findings of Mambelli et al. (Mambelli et al. 2024). Subgroup analyses revealed that IHDF and UF profiling were particularly effective, whereas interventions lasting more than 12 sessions did not show significant improvements. These findings imply that the duration and patient adaptation to interventions may influence treatment efficacy. In addition, subgroup analysis according to intervention delivery characteristics distinguished dialysis technology‐based interventions from nurse‐led interventions. Dialysis technology‐based interventions demonstrated a statistically significant improvement in SBP, whereas nurse‐led interventions showed a comparable effect size but did not reach statistical significance. Because only four studies with a relatively small sample size evaluated nurse‐led interventions, this finding may reflect limited precision resulting from the small number of studies and participants; therefore, the effectiveness of nurse‐led interventions remains uncertain. Nevertheless, nurse‐led interventions may complement dialysis technology‐based approaches through individualized assessment, continuous monitoring and timely bedside interventions during haemodialysis. Differences in effect sizes were also observed across study characteristics. Recent studies and those conducted in Asian populations demonstrated relatively greater intervention effects. The enhanced effects in recent studies are likely attributable to advances in dialysis technology and improvements in clinical monitoring systems (Mitra et al. 2025). The adoption of automated sodium control systems, hemodynamic biofeedback devices and real‐time monitoring technologies may have increased the precision and adaptability of these interventions (Mitra et al. 2025). In contrast, the greater effects observed in Asian studies may have resulted from a combination of demographic and physiological factors. First, demographic or physiological characteristics may differ between Asian and non‐Asian populations, potentially affecting relative hemodynamic responses to the same absolute UF rate (Assimon et al. 2016). Second, vascular reactivity and autonomic function may vary across populations due to genetic and environmental influences, which could contribute to differences in responsiveness to interventions (Yim et al. 2012). Future research should therefore explore how population‐specific and physiological factors influence the outcomes of non‐pharmacological strategies. The findings of this study reaffirm that the pathophysiology of IDH is multifactorial (Kanbay et al. 2020; Habas et al. 2025). Excessive ultrafiltration rates that exceed plasma refill capacity result in intravascular volume reduction, a relationship supported by studies examining plasma refill rate and ultrafiltration dynamics during haemodialysis (Wang, Negoianu, et al. 2023; Sohn et al. 2021; Dong et al. 2024). Meanwhile, in elderly or diabetic patients, diminished compensatory vasoconstriction—likely due to impaired autonomic responsiveness and reduced vascular compliance—further compromises hemodynamic stability (Seravalle et al. 2021). These mechanisms highlight the importance of non‐pharmacological approaches as complementary strategies to pharmacologic treatment in addressing the hemodynamic causes of IDH. Clinically, maintaining haemodynamic stability during dialysis is crucial for patient comfort and for reducing the risk of recurrent ischemic stress to the myocardium and brain. Burton et al. (Burton et al. 2009) and McIntyre (McIntyre 2010) demonstrated that repetitive IDH episodes induce subclinical ischemic damage, leading to progressive cardiac dysfunction and cognitive impairment. Thus, the moderate improvements in blood pressure stability observed in this meta‐analysis may support haemodynamic stability rather than direct clinical protection (Shoji et al. 2004; McIntyre 2010; Burton et al. 2009). This systematic review and meta‐analysis adds new knowledge to the existing literature in several important ways. First, it synthesizes evidence across a broad range of non‐pharmacological interventions, explicitly incorporating interventions that are directly implementable by nurses during routine haemodialysis care. Second, by comparing effects across intervention types, study characteristics and implementation features, this review identifies conditions under which non‐pharmacological strategies are more likely to stabilize SBP. Third, the findings highlight the practical relevance of short‐term, targeted interventions that can be integrated into everyday nursing workflows. However, the clinical significance of SBP improvement should be interpreted with caution, and future studies should prioritize patient‐centred outcomes, including the frequency of IDH episodes, symptom severity and dialysis tolerance. This study has several limitations. First, although both randomized and quasi‐experimental studies were included to ensure comprehensiveness, variability in study design and the lack of a standardized definition of intradialytic hypotension across studies may have contributed to the substantial heterogeneity observed in this meta‐analysis. In addition, quality appraisal indicated that several included studies had methodological limitations, particularly inadequate randomization, allocation concealment and blinding, which may have introduced bias and influenced the pooled effect estimates. Future research should adopt standardized diagnostic criteria and outcome measures for IDH and conduct well‐designed multicentre randomized controlled trials to improve methodological rigour and comparability across studies. Second, the diversity of intervention types and patient characteristics limited subgroup comparisons. Standardized reporting of key parameters such as dialysate temperature, sodium concentration and UF targets is necessary to enhance comparability across studies. Third, most included studies were conducted in single centres with relatively small sample sizes, which limits the generalizability of the findings across different dialysis settings and patient populations. Future studies should include large multicentre randomized controlled trials to improve the external validity and generalizability of the evidence. Moreover, details of nurse‐led interventions were often insufficiently described. Future research should incorporate specific nursing parameters, including timing, frequency and decision‐making processes. Fourth, most studies focused on short‐term hemodynamic outcomes, whereas evidence on long‐term endpoints—such as cardiovascular events, residual renal function and quality of life—remains scarce. Longitudinal studies are needed to evaluate the sustained effects of non‐pharmacological interventions. Finally, although publication bias was assessed using funnel plots, Egger's regression test and the trim‐and‐fill method, the relatively small number of included studies and substantial heterogeneity may have reduced the reliability of these methods. Therefore, the findings regarding publication bias should be interpreted with caution. Overall, the certainty of evidence was rated as very low according to the GRADE approach, primarily because of methodological limitations, inconsistency and imprecision across the included studies. Implication for Nursing Practice The prevention of IDH is an important component of routine haemodialysis nursing care. Although dialysis technology‐based interventions demonstrated a statistically significant improvement in SBP in this review, their implementation generally depends on dialysis prescriptions and equipment settings. In contrast, nurse‐led interventions, including positional adjustments, compression stockings, relaxation therapy and herbal acupoint therapy, can be implemented directly at the bedside as part of routine nursing care (Marinovich et al. 2020; Bennett et al. 2015; Tsai et al. 2018). Nurses play a central role in preventing IDH through ongoing patient assessment, early recognition of hemodynamic instability, continuous monitoring during dialysis, timely bedside interventions and patient education (McIntyre et al. 2021; Kesik and Özdemir 2020). Integrating these nursing activities with dialysis technology‐based strategies may support individualized clinical decision‐making, patient management and safe haemodialysis care. Nursing education should strengthen competencies in early recognition of hemodynamic instability, patient assessment and timely intervention, while future research should further evaluate standardized nurse‐led interventions and their effects on clinically meaningful and patient‐centred outcomes. Conclusions In conclusion, this systematic review and meta‐analysis found that non‐pharmacological interventions were associated with modest improvements in SBP during haemodialysis. However, these findings should be interpreted with caution, as improvements in blood pressure represent a surrogate hemodynamic outcome and do not necessarily indicate a reduced incidence of IDH. While selected short‐term and targeted interventions may support hemodynamic stability during dialysis sessions, further well‐designed studies are required to determine whether these strategies translate into clinically meaningful reductions in IDH episodes, symptom burden and patient‐centred outcomes. From a nursing perspective, these results highlight the potential value of integrating carefully selected non‐pharmacological strategies into routine dialysis care, accompanied by standardized protocols and ongoing evaluation of clinical effectiveness. Author Contributions Mi‐Kyoung Cho: conceptualization, methodology, formal analysis, data curation, writing – review and editing, writing – original draft. Yoon Hee Cho: conceptualization, methodology, writing – original draft, writing – review and editing, validation. Funding This work was supported by Chungbuk National University NUDP program 2025. Ethics Statement The authors have nothing to report. Conflicts of Interest The authors declare no conflicts of interest. Supporting information Table S1: Complete search strategies for the electronic databases. Table S2: GRADE assessment and reasons for downgrading.