Non-invasive haemodynamic monitoring to individualise fluid resuscitation in adult diabetic ketoacidosis: A randomized controlled pilot trial

Abstract

BACKGROUND

Diabetic ketoacidosis (DKA) requires rapid fluid resuscitation; however, current guideline-based protocols depend on static assessments of volume status. We need a new fluid resuscitation protocol to correct DKA and alleviate the DKA-related symptoms such as fever, mania, inflammation and depression caused by the imbalance of water and electrolytes.

AIM

To evaluate whether non-invasive haemodynamic monitoring with the ultrasonic cardiac output monitor (USCOM) integrated with passive leg raising can optimise fluid resuscitation in adults with DKA without compromising metabolic recovery.

METHODS

We undertook an embedded, comprehensive review of studies on rehydration for DKA and a proof-of-concept in an emergency department in Jiangsu, China. In the first stage, adults with DKA were allocated (1:1) to USCOM-guided resuscitation or sham USCOM monitoring. In the intervention group, additional fluid boluses (250 mL crystalloids over 15 minutes) were given only when stroke volume increased by ≥ 10% during passive leg raising. The primary outcome was net fluid balance at 24 hours, 48 hours, and 72 hours. Secondary outcomes included urine output, length of hospital stay, glycaemic control, urinary ketone clearance, and adverse events. In the second stage, we systematically retrieved clinical studies on fluid resuscitation in DKA from the PubMed database and performed an in-depth analysis of the available evidence. We conducted a systematic review of the evidence to increase externality validation and generalization.

RESULTS

Fifty participants with a mean age of 48.0 ± 16.2 years (42% males) were enrolled. Compared with the control group, the USCOM-guided group exhibited a significantly lower net fluid balance at 48 hours (-1246.8 mL; P = 0.007) and 72 hours (-1759.5 mL; P = 0.001), associated with greater urine output in the first 24 hours (+952.6 mL; P = 0.002). Length of hospital stay was reduced by 1.92 days (P = 0.034). Non-significant differences were observed between groups in blood glucose trajectories, urinary ketone clearance, or adverse event rates. The systematic review identified 20 clinical studies investigating fluid resuscitation in DKA, involving 8402 patients. These included 1 randomized controlled trial, 3 observational studies, 1 nested cohort study, and 15 cohort studies, providing important insight into DKA rehydration strategies.

CONCLUSION

Dynamic, non-invasive haemodynamic monitoring is a feasible approach to optimising fluid resuscitation for DKA and supports the need for larger multicentre trials.

Key Words: Diabetic ketoacidosis; Fluid resuscitation; Non-invasive haemodynamic monitoring; Doppler cardiac output monitoring; Inflammation; Fever

Core Tip: Diabetic ketoacidosis requires urgent fluid resuscitation, yet standard protocols only adopt static volume assessment. This Chinese emergency department mixed-methods study develops a non-invasive doppler strategy combining ultrasonic cardiac output monitoring and passive leg raising for personalised adult diabetic ketoacidosis rehydration. Compared with guideline routine care, it safely lowers 72-hour net fluid balance and hospital stays without harming metabolic or renal recovery. Feasible at bedside, it reduces iatrogenic fluid overload and lessens resource pressure in under-resourced emergency settings without invasive monitors.

INTRODUCTION

Across health care systems, diabetic ketoacidosis (DKA) places a substantial burden on emergency departments, intensive care units, and general hospital wards. Hyperglycaemic crises account for a significant proportion of diabetes-related emergency admissions and unplanned hospital bed-days, often affecting younger adults and socially disadvantaged populations who require high-acuity monitoring and complex metabolic support[1-3]. In low- and middle-income countries, where emergency and critical care resources are limited, and referral pathways may be constrained, recurrent episodes of DKA can overwhelm emergency departments, delay treatment for other acute diseases, and generate catastrophic out-of-pocket expenditures for households[4,5]. In this context, even modest reductions in hospital length of stay or in the need for high-intensity monitoring could yield meaningful improvements in health system capacity and financial protection.

Rapid and effective fluid resuscitation is a cornerstone of DKA management, with the goals of restoring circulating volume, improving tissue perfusion, and supporting renal function[2,6]. In clinical practice, clinicians must balance the risk of cerebral or pulmonary oedema from overly aggressive fluid administration against the risk of persistent hypoperfusion and delayed metabolic recovery resulting from inadequate resuscitation[7]. Observational studies suggest that inappropriate fluid management contributes to iatrogenic complications in up to 30% of patients experiencing hyperglycaemic crises[4,5], underscoring the need for more precise and individualised haemodynamic guidance.

Most current DKA management protocols rely on static physiological indicators, such as blood pressure, heart rate, and urine output. However, these measures do not reliably reflect intravascular volume status in patients with hyperosmolar states[2]. Although invasive monitoring techniques, including pulmonary artery catheterisation, can provide detailed haemodynamic information, they are costly, technically demanding, and carry procedural risks, limiting their routine use in emergency care, particularly in resource-limited settings[8]. Non-invasive strategies such as passive leg raising (PLR) and pulse pressure variation offer dynamic assessment of fluid responsiveness; however, their accuracy may be reduced in spontaneously breathing patients and in those with arrhythmias, both of which are common among adults presenting with DKA[9,10].

The ultrasonic cardiac output monitor (USCOM) is a portable doppler-based device that measures aortic blood flow and provides real-time estimates of cardiac output (CO) and stroke volume. It requires minimal operator training and has demonstrated good inter-observer reliability[11]. Its non-invasive design, short learning curve, and relatively low cost make USCOM particularly well-suited for emergency departments and healthcare facilities with limited resources[12]. USCOM-guided assessment of fluid responsiveness using PLR has demonstrated clinical value in the management of sepsis and perioperative care settings[13-16]. However, its role in DKA resuscitation has not been systematically evaluated, and no randomized trial has assessed a standardised protocol in this population. Building on concerns raised by paediatric DKA trials regarding fluid infusion rates and neurological complications[7], we designed a single-centre pilot to test whether USCOM-guided fluid therapy, using a Δ stroke volume (SV) ≥ 10% threshold during PLR, can reduce fluid overload while preserving metabolic stability in adults with DKA. To our knowledge, this is the first pilot trial to evaluate a dynamic, non-invasive haemodynamic monitoring strategy for fluid resuscitation in DKA.

MATERIALS AND METHODS

Trial setting

This was a two-arm, 1-to-1, randomized parallel-controlled, single-blind, and pilot clinical trial.

This pilot trial was conducted in Jiangsu Province, China, with participants recruited from the Emergency Department of Yangzhou University Affiliated Hospital. The study complied with the Declaration of Helsinki and received approval from the hospital’s Institutional Review Board, approval No. 2022-YKL3-06-004. Written informed consent was obtained from all participants before enrollment. The study protocol was registered with the Chinese Clinical Trial Registry (No. ChiCTR2500103388) and was conducted in accordance with the Consolidated Standards of Reporting Trials guidelines.

Eligibility criteria

Adults presenting to the emergency department with DKA were screened for eligibility. Participants who met all inclusion criteria and did not meet any exclusion criteria were enrolled in this pilot trial. Individuals identified during comprehensive clinical evaluation as meeting exclusion criteria were advised to withdraw, and the rationale was fully explained.

DKA was diagnosed when the Chinese Guidelines for Diagnosis and Treatment of Hyperglycaemic Crises were fulfilled.

The inclusion criteria: (1) Age ≥ 18 years; (2) Current diagnosis of DKA; (3) Laboratory confirmation with blood glucose > 13.9 mmol/L, serum ketones ≥ 3 mmol/L or positive urine ketones, arterial pH < 7.3 and bicarbonate < 18 mmol/L; and (4) Systolic blood pressure ≥ 90 mmHg without vasopressor support.

The exclusion criteria included conditions that contraindicated non-invasive ultrasound assessment, including the following: (1) Cardiopulmonary comorbidities such as class III-IV heart failure or severe chronic obstructive pulmonary disorder; (2) Structural valvular disease (aortic or pulmonary stenosis or regurgitation); (3) Thoracic deformities; and (4) A history of thoracic surgery.

All fluid resuscitation was administered by attending emergency physicians in designated treatment areas within the emergency department.

Intervention and comparator

All participants received standardised fluid resuscitation in accordance with the 2022 Chinese Consensus Guidelines for DKA and Hyperosmolar Hyperglycaemic State, with adjustments guided by crucial signs and laboratory results (https://diab.cma.org.cn/cn/zhinangongshi.aspx).

The intervention consisted of USCOM-guided, PLR-based assessment of fluid responsiveness before each planned fluid bolus. Haemodynamic measurements were obtained from the left parasternal window using the aortic velocity-time integral with a USCOM 1A device (USCOM Ltd, Australia). SV was calculated as velocity-time integral × aortic valve area (cm²), and CO as SV × heart rate.

The fluid responsiveness threshold was defined as a PLR-induced change in SV (ΔSV) ≥ 10%; fluid non-responsiveness was defined as ΔSV < 10%. A positive PLR response (ΔSV ≥ 10%) triggered administration of an additional 250 mL bolus of crystalloid over 15 min, whereas no further fluid loading was given when ΔSV < 10%. This ΔSV ≥ 10% criterion was adopted from Marik et al[17] to enhance the reliability of haemodynamic assessment. This threshold is also supported by the meta-analytic evidence from Monnet et al[14], who reported a pooled sensitivity of 0.85, specificity of 0.91, and an area under the receiver operating characteristic curve of 0.95 ± 0.01 for PLR-induced changes in CO using a threshold of approximately 10%.

In the intervention group, clinicians used real-time feedback from PLR manoeuvre and USCOM-derived haemodynamic values to determine whether the fluid responsiveness threshold was met before each resuscitation step. If ΔSV ≥ 10%, the patient was considered fluid-responsive, and a 250 mL crystalloid bolus was administered; if ΔSV < 10%, no additional fluid was given, and the patient remained under close observation.

In the control group, individuals received guideline-based fluid resuscitation without assessment of ΔSV. A sham device, visually similar to the active USCOM 1A and sharing the same user interface and workflow, displayed randomly generated haemodynamic values to preserve participant blinding. Clinicians in the control arm did not use these sham values to guide fluid administration and continued rehydration until the planned volume was achieved, based on standard clinical assessment.

Outcomes

The prespecified primary result was fluid resuscitation volume, summarised as cumulative fluid intake, net fluid balance, and urine output. Net fluid balance was defined as total fluid administered minus urine output. Volumes were recorded at 24 hours, 48 hours, and 72 hours following enrollment.

Secondary results included the incidence of complications, time to urinary ketone clearance, blood glucose concentrations, USCOM-derived haemodynamic parameters, and length of hospital stay. Complications were defined as any clinical adverse event occurring between the initiation of fluid resuscitation and hospital discharge, regardless of whether the event was considered related to the intervention. Only the type of complication was recorded; the number of events per patient was not documented.

USCOM parameters (SV and CO) were measured in the intervention group at baseline (before the first fluid bolus) and at each subsequent resuscitation step, with values obtained directly from the device display following accurate probe positioning. Blood glucose and urinary ketone levels were assessed by the hospital laboratory using venous blood samples. Blood glucose concentrations were reported at baseline, 24 hours, and 72 hours following admission. Urinary ketone clearance was defined as the time to resolution of DKA, indicated by completion of the final fluid resuscitation episode.

Continuous variables are presented as mean standard deviation (SD) if normally distributed, or median (interquartile range) otherwise. The timing of measurements for each outcome is summarised in Table 1.

Table 1 Schedule of enrollment, interventions, and outcome assessments.

Outcome	Admission	First fluid resuscitation	Day 1	Day 2	Day 3	Last fluid resuscitation	Day 4	…	Day 13	Day 14	Discharge
Fluid replacement	-	-	1, 2	1, 2	1, 2	-	1	1	1	1	-
Days of hospitalisation	-	-	-	-	-	-	-	-	-	-	1, 2
Complications	-	1	1	1	1	1	1	1	1	1	1, 2
Time to urine ketone conversion	-	-	1	1	1	-	1	1	1	1	1, 2
Blood glucose	1, 2	-	1, 2	1	1, 2	-	1	1	1	1	1
USCOM technical parameters	-	1, 2	1	1	1	1, 2	-	-	-	-	-

¹Indicates that the outcome was measured at the corresponding time point.

²Indicates that data from that time point are reported in the results section.

Timing of measurements for prespecified trial outcomes from admission through day 14 and hospital discharge, including fluid replacement, length of hospitalisation, complications, time to urinary ketone clearance, blood glucose and ultrasonic cardiac output monitor-derived haemodynamic parameters. USCOM: Ultrasonic cardiac output monitor.

Harms

The ultrasound device used in this trial is non-invasive and poses no direct risk to participants.

Sample size

The sample size calculation assumed a 30% reduction in fluid intake at 48 hours in the intervention group compared with the control group, corresponding to an effect size of d = 0.8, with α = 0.05, β = 0.20 (power 80%), an SD (σ) of 1834 mL, and an expected mean difference (Δ) of 1247 mL. Allowing for 20% attrition, 25 individuals per group were required. The required sample size was derived using the standard formula for comparing two independent means.

Randomisation and blinding

Randomisation was performed by a practising emergency physician, who generated random numbers between 0 and 1 using a computer program. Individuals with a random number between 0 and 0.5 were assigned to the intervention group, and those with a number between 0.5 and 1 were assigned to the control group. Allocation findings were communicated by telephone to the treating physicians. Recruitment was ongoing until each group had enrolled 25 eligible participants who met the inclusion criteria and did not meet any exclusion criteria, ensuring a 1:1 allocation.

Participants were blinded to their treatment allocation. Binding was achieved through two measures. First, participants in the control group were assessed using a sham USCOM device identical in appearance to the active device, but that did not provide functional haemodynamic measurements. Second, participants were not informed about the intended effects of USCOM-guided fluid management. The attending emergency physician managing fluid therapy was aware of group allocation and whether the fluid responsiveness threshold had been met during each assessment. However, the data analyst remained blinded to group assignments throughout the statistical analysis.

Statistical analysis

USCOM measurements were performed by two trained emergency physicians, with each value calculated as the mean of three consecutive readings to minimise inter-operator variability. Laboratory parameters, such as pH, bicarbonate, and ketones, were measured by a blinded technician using standardised assays.

All clinical data were analysed to assess differences in outcomes between the intervention and control groups. Statistical analyses were two-sided with a significance level of 0.05 and were conducted using the Statistical Package for the Social Sciences software (version 27.0; IBM Corp.). Normally distributed continuous variables are reported as mean ± SD and compared using independent-samples t-tests. Non-normally distributed variables are reported as median and interquartile range and compared using the Mann-Whitney U test. Categorical variables are expressed as n (%) and compared using the χ² or Fisher’s exact test, as appropriate.

Predefined subgroup analyses were conducted by age, with 60 years as the stratification cut-off.

Embedded systematic review

This pilot trial was embedded within a broader mixed-methods project that also included a systematic review of the international literature on fluid resuscitation in DKA. The purpose of this embedded evidence synthesis was to contextualise the trial results, identify gaps in current practice and research, and explore the generalisability of haemodynamic-guided strategies across settings.

The systematic review search strategy employed in this study drew inspiration from Bo and Zhao[18] approach to establishing the role of inflammation in platelets. This means that searching the PubMed database is applicable when constructing a specific theory[18]. We searched the PubMed database using the terms “DKA” and “rehydration”, consistent with the methodology[19]. We identified studies that directly examined relationships between DKA and rehydration practices through systematic searching, following the method described. The detailed methodology for conducting a systematic review, such as literature search strategies, screening procedures, the formulation of research questions, and approaches to addressing them, typically demonstrates universality across various research fields and topics. The specific methodology we employed was largely based on a research protocol published by Oxford University Press entitled ‘The role of inflammation in depression: A scoping review protocol on mechanisms, evidence, and therapeutic potential’. This research protocol primarily describes how to use a scoping review to explore the role of inflammation in depression[20]. Although this disease model does not pertain to DKA, the research methodology is transferable. Consequently, we have not provided an extensive description of the systematic review methodology in this DKA study.

Eligible study designs included randomized controlled trials, cohort studies, and retrospective observational studies. This mixed embedding study was not registered. In the past, this systematic review approach of embedding case reports was widely favoured by researchers. In this pilot study, we plan to continue using this widely popular method and include the results of the systematic review in the results section. This embedded system review method can often be used as an additional means of external validation when only a single case report is available or when the number of recruited participants is small. In this viewpoint, we have identified a number of case report articles to verify whether an embedded systematic review of similar studies helps to enhance the credibility of the article. For example, a study examining complex case such as acute-on-chronic liver failure, artificial liver support and psychotherapy; this particular case report presents a highly complex instance of acute-on-chronic liver failure accompanied by persistent fever and manic symptoms. As psychotherapy and artificial liver support were not mentioned in multicentre clinical trials, this could raise public doubts regarding the credibility of this single case report. By incorporating a summary of a systematic review of similar conditions within the discussion section, the authors not only reinforce the role of psychotherapy and artificial liver support in the management of acute-on-chronic liver failure but also broaden the readers’ perspective[21]. This strategy is equally applicable to our DKA fluid resuscitation study; as this study is sufficiently novel, it is likely to give rise to similar concerns. This embedded systematic review of similar studies is an appropriate approach.

Two reviewers independently screened titles and abstracts, assessed full texts for eligibility, and extracted data on sample size, patient age group, clinical setting, fluid regimens (rate, volume, and composition), outcomes, and complications. Disagreements between reviewers were resolved through discussion and consensus or, when necessary, through adjudication by a third reviewer. The primary outcomes of interest were neurological complications, acute kidney injury, mortality, and markers of fluid overload such as pulmonary oedema and prolonged hospital length of stay. The overall review process followed procedures similar to those outlined in the scoping review methodology previously published by Bo[20].

Because of anticipated heterogeneity in populations, interventions, and outcome definitions, we prespecified a structured narrative synthesis rather than a formal quantitative meta-analysis. Findings were summarised by clinical theme, age group, and outcome domain, and similarities and differences in direction and magnitude of effects were described qualitatively (Supplementary Tables 1 and 2). Harvest/effect-direction plot was constructed by the study design and outcome domain. This method of integrating evidence from highly heterogeneous studies was borrowed from a previous study by Zhuang and Bo[22] on care issues for diabetic patients during the coronavirus disease 2019 period.

RESULTS

Participant flow

From 6 March 2022 to 1 August 2025, 69 adults presenting to the emergency department with suspected DKA were screened for eligibility. Among these, 50 met all inclusion criteria. None were excluded and were randomized in a 1:1 ratio to the USCOM haemodynamic monitoring group (intervention, n = 25) or the sham USCOM group (control, n = 25). Participant flow through the trial, including reasons for exclusion before randomisation, is illustrated in Figure 1. No participants withdrew during the 14-day treatment and follow-up period, and all randomized participants were included in the final analysis.

Figure 1 Pilot trial profile and participant flow. Consolidated Standards of Reporting Trials-style flow diagram indicating the screening of adults presenting with suspected diabetic ketoacidosis, eligibility assessment, randomisation to ultrasonic cardiac output monitor-guided or control fluid resuscitation, allocation, follow-up and inclusion in the primary and secondary analyses. Boxes indicate reasons for exclusion before randomisation. No participants withdrew after randomisation during the 14-day treatment and follow-up period.

Baseline data

Baseline demographic, clinical, and laboratory characteristics were comparable between the two groups (Table 2). Mean age was slightly higher in the USCOM group compared to that in the control group (51.8 ± 13.5 years vs 44.1 ± 18.1 years), but this difference did not reach statistical significance. Body mass index, blood pressure, heart rate, respiratory rate, and oxygen saturation were balanced between groups, as were sex distribution and crucial biochemical markers at presentation. These results support the validity of between-group comparisons for treatment outcomes.

Table 2 Baseline demographic and clinical characteristics of the study population, mean ± SD.

Parameter	USCOM group (n = 25)	Control group (n = 25)	Statistical effect size	P value
Age (years)1	51.8 ± 13.5	44.1 ± 18.1	1.70	0.093
BMI (kg/m2)1	24.1 ± 2.5	24.8 ± 4.3	-0.74	0.464
SBP (mmHg)1	127.1 ± 16.1	132.5 ± 23.1	-0.95	0.346
DBP (mmHg)1	77.8 ± 10.6	76.0 ± 15.5	0.47	0.641
MAP (mmHg)1	94.2 ± 11.5	94.8 ± 16.8	-0.15	0.881
Heart rate (bpm)1	102.2 ± 19.4	104.1 ± 18.1	-0.35	0.731
Respiratory rate (breaths/minute)1	24.7 ± 6.1	22.5 ± 5.1	1.35	0.180
SpO2 (%)1	96.6 ± 2.1	96.2 ± 1.4	0.80	0.431
Sex (male/female)2	11/14	10/15	0.04	0.396

¹Independent-samples t-tests.

²Pearson’s χ² or Fisher’s exact tests.

Baseline demographic, vital sign, and clinical characteristics of participants randomized to ultrasonic cardiac output monitor-guided vs control management. Data are presented as mean ± SD unless otherwise stated. P values compare groups using appropriate parametric or non-parametric tests (for instance, independent-samples t-tests for continuous variables and χ² or Fisher’s exact tests for categorical variables), as indicated by superscripts in the table. BMI: Body mass index; SBP: Systolic blood pressure; DBP: Diastolic blood pressure; MAP: Mean arterial pressure.

Efficacy analyses

Both groups achieved resolution of DKA within three days of initiating fluid resuscitation. Total fluid intake over the first 72 hours was similar in the USCOM and control groups, indicating that the protocol did not simply restrict overall fluid delivery (Table 3; Figure 2). However, the timing and effective utilisation of fluids differed. At 72 hours, ongoing fluid infusion volumes were lower in the USCOM group than in the control group.

Figure 2 Effects of ultrasonic cardiac output monitor-guided fluid resuscitation on net fluid balance, urine output, hospital stay, and complications. A: Net cumulative fluid replenishment over 72 hours in ultrasonic cardiac output monitor (USCOM)-guided and control groups, with error bands representing standard deviations; B: Urine output trajectories over the first 72 hours, with shaded areas indicating standard deviations; C: Kaplan-Meier curves illustrating time to hospital discharge, demonstrating earlier discharge trends in the USCOM-guided group; D: Incidence of key complications, indicating no signal of harm in major safety endpoints. Panels depict cumulative net fluid balance and urine output at 24 hours, 48 hours, and 72 hours, the total 72 hours net fluid balance, hospital length of stay, and the incidence of prespecified complications (hypoglycaemia, hypokalaemia, pulmonary oedema, acute kidney injury, and arrhythmia) in USCOM-guided and control groups. Bars indicate mean values, with error bars representing standard deviations. P values are from two-sided independent-samples t-tests for continuous outcomes and χ² or Fisher’s exact tests for categorical outcomes, consistent with the analyses reported in Tables 3, 4 and 6.

Table 3 Fluid intake, net fluid balance, and urine output in ultrasonic cardiac output monitor-guided vs control groups, mean ± SD.

Time point	USCOM group (n = 25)	Control group (n = 25)	t-value	Cohen’s d	Mean difference	95%CI	P value
Fluid replenishment (mL)
24 hours	6514.04 ± 1925.03	6230.80 ± 1433.69	0.590	0.167	283.24	-681.97 to 1248.45	0.558
48 hours	3052.68 ± 933.51	3010.96 ± 1386.33	0.125	0.035	41.72	-630.37 to 713.81	0.901
72 hours	1751.04 ± 544.65	2328.88 ± 733.72	-3.162	-0.894	-577.84	-945.29 to -210.39	0.003a
Total 72 hours	11317.76 ± 2436.41	11570.64 ± 2726.75	0.35	0.10	-252.88	-1217.82 to 1723.58	0.730
Net fluid replenishment (mL)
24 hours	2681.68 ± 1833.34	3351.00 ± 1232.93	-1.515	-0.428	-669.32	-1557.76 to 219.12	0.136
48 hours	2934.12 ± 1529.80	4180.96 ± 1833.98	-2.610	-0.738	-1246.84	-2207.22 to -286.46	0.012a
72 hours	2916.36 ± 1542.48	4675.84 ± 2021.83	-3.459	-0.978	-1759.48	-2782.10 to -736.86	0.001a
Urine output (mL)
24 hours	3832.36 ± 1015.11	2879.80 ± 1148.91	3.107	0.879	952.56	336.05-1569.07	0.003a
48 hours	2800.24 ± 1411.51	2181.00 ± 1112.64	1.723	0.487	619.24	-103.51 to 1341.99	0.091
72 hours	1768.80 ± 839.23	1834.00 ± 978.60	-0.253	-0.072	-65.20	-583.61 to 453.21	0.801

^aP < 0.05.

Overall, compared with the control group, the ultrasonic cardiac output monitor (USCOM) group demonstrated significantly lower net fluid replenishment at 48 hours and 72 hours, significantly lower fluid replenishment at 72 hours, and significantly higher urine output at 24 hours. Values are presented as mean ± SD. Independent-samples t-tests were used to compare USCOM and control groups at each time point. The mean difference was calculated as the USCOM group minus the control group; therefore, negative values indicate lower values in the USCOM group. The 95% confidence interval represents the confidence interval around the mean difference. Effect sizes were estimated using Cohen’s d, calculated from the pooled SD. P values are two-sided and were not adjusted for multiple comparisons. USCOM: Ultrasonic cardiac output monitor; CI: Confidence interval.

Net fluid balance diverged progressively between groups. At 24 hours, net fluid accumulation was numerically lower with USCOM guidance; by 48 hours, this difference was statistically significant. At 72 hours, the USCOM group had a significantly lower net fluid balance than controls, corresponding to a mean reduction of 1759.48 mL (Table 3; Figure 2). Therefore, USCOM-guided management reduced cumulative net fluid accumulation by nearly 1.8 L over 72 hours without restricting total fluid input in the early phase of resuscitation.

Despite the lower net fluid balance, early diuresis was enhanced in the USCOM group. At 24 hours, urine output was higher in the USCOM group compared to that in the control groups, with a trend towards higher output at 48 hours. By 72 hours, urine output was similar in the two groups (Table 3; Figure 2). These findings suggest that USCOM-guided resuscitation improved effective renal perfusion and promoted earlier mobilisation of excess fluid while avoiding persistent fluid overload.

Metabolic control and DKA resolution were not compromised by the intervention. Blood glucose trajectories over 24 hours, 48 hours, and 72 hours were similar in the USCOM and control groups, with non-significant between-group differences at any time point (Table 4). Moreover, time to urinary ketone clearance was comparable (6.04 ± 2.29 days vs 6.28 ± 2.59 days; mean difference -0.24 days; 95% confidence interval (CI): -1.15 to 1.63; P = 0.730). Notably, these findings indicate that minimising net fluid accumulation with haemodynamic guidance did not delay DKA resolution or impair glycaemic control.

Table 4 Blood glucose trajectories, urinary ketone clearance, and length of hospital stay, mean ± SD.

Time point	USCOM group (n = 25)	Control group (n = 25)	t-value	Cohen’s d	Mean difference	95%CI	P value
Blood glucose (mmol/L)
24 hours	16.07 ± 3.62	15.99 ± 4.51	0.07	0.02	0.08	-2.25 to 2.41	0.944
48 hours	15.81 ± 3.54	14.84 ± 4.15	0.89	0.25	0.97	-1.23 to 3.17	0.377
72 hours	13.85 ± 4.09	13.32 ± 3.23	0.51	0.14	0.53	-1.56 to 2.62	0.612
Time for urinary ketone bodies to turn negative (day)
Discharge	6.04 ± 2.29	6.28 ± 2.59	0.35	0.10	-0.24	-1.15 to 1.63	0.730
Length of hospitalisation (day)
Discharge	7.8 ± 2.77	9.72 ± 3.38	2.18	0.62	-1.92	0.15-3.69	0.034a

^aP < 0.05.

Blood glucose concentrations at 24 hours, 48 hours, and 72 hours after admission, time to urinary ketone clearance, and total length of hospitalisation in ultrasonic cardiac output monitor-guided and control groups. Data are presented as mean ± SD, with mean differences and 95% confidence interval. P values are from two-sided independent-samples t-tests. “Discharge” denotes the time point at which the participant left the hospital. Time to urinary ketone clearance is defined as the cumulative number of hospital days from admission until the last recorded urinary ketone test is negative. USCOM: Ultrasonic cardiac output monitor; CI: Confidence interval.

Hospital length of stay was significantly shorter among patients managed with USCOM guidance fluid resuscitation. Mean length of stay was 7.80 ± 2.77 days in the USCOM group compared with 9.72 ± 3.38 days in the control group, obtaining a mean reduction of 1.92 days (95%CI: 0.15-3.69; P = 0.034; Table 4; Figure 2). Extrapolated to a cohort of 100 patients, this corresponds to approximately 192 hospital bed-days saved, representing an estimated 3.1% increase in effective bed capacity under fixed resources.

Within the USCOM group, haemodynamic measurements confirmed that the protocol obtained its intended physiological effect. SV increased from 29.8 ± 4.5 mL at baseline to 68.6 ± 9.7 mL following resuscitation (mean difference 38.8 mL; 95%CI: 33.2-44.4; P < 0.001), and CO rose from 3.0 ± 0.7 L/minute to 5.4 ± 1.0 L/minute (mean difference 2.4 L/minute; 95%CI: 1.9-2.9; P < 0.001; Table 5). These alterations support the role of non-invasive doppler monitoring in identifying truly fluid-responsive patients and optimising cardiac performance during DKA resuscitation.

Table 5 Haemodynamic effects of ultrasonic cardiac output monitor-guided fluid resuscitation in the intervention group, mean ± SD.

Parameters	Baseline	Post-resuscitation	t-value	Cohen’s d	Mean difference	95%CI	P value
SV (mL)	29.8 ± 4.5	68.6 ± 9.7	14.82	2.96	38.8	33.2-44.4	< 0.001
CO (L/minute)	3.0 ± 0.7	5.4 ± 1.0	12.00	2.40	2.4	1.9-2.9	< 0.001

Stroke volume and cardiac output were measured by ultrasonic cardiac output monitor at baseline (before the first rehydration therapy) and after completion of protocolised fluid resuscitation in the ultrasonic cardiac output monitor group. Data are paired means ± SD, with within-patient mean differences and 95% confidence interval. P values are from two-sided paired t-tests, chosen to minimise inter-individual variability when comparing the same patient before and after treatment. Baseline indicates measurements before the first rehydration therapy; post-resuscitation indicates measurements obtained before the last rehydration treatment. SV: Stroke volume; CO: Cardiac output; CI: Confidence interval.

Complications

All participants received standardised diabetes management, including insulin treatment and electrolyte replacement, according to guideline-based protocols. The overall safety profile was similar between groups (Table 6; Figure 2). Rates of hypoglycaemia were 24% in the USCOM group and 20% in controls, and hypokalaemia occurred in 24% and 20%, respectively (both P = 0.733). No cases were observed in the USCOM group compared with one case in the control group; acute kidney injury was observed in 8% of USCOM-treated patients and 12% of controls, with non-statistically significant between-group differences. Arrhythmias occurred in 20% of patients in the USCOM group and in 12% of patients in the control group (P = 0.440).

Table 6 Incidence of complications in ultrasonic cardiac output monitor-guided vs control groups, n (%).

Complication	USCOM group (n = 25)	Control group (n = 25)	χ2	Cohen’s d	Absolute risk difference	95%CI	P value
Hypoglycemia1	6 (24)	5 (20)	0.12	0.086	+0.04	-0.19 to 0.27	0.733
Hypokalemia1	6 (24)	5 (20)	0.12	0.086	+0.04	-0.19 to 0.27	0.733
Pulmonary oedema2	0 (0)	1 (4)	-	0.252	-0.04	-0.12 to 0.04	0.313
Acute kidney injury2	2 (8)	3 (12)	-	0.134	-0.04	-0.21 to 0.09	0.640
Arrhythmia2	5 (20)	3 (12)	-	0.274	+0.08	-0.12 to 0.28	0.440

¹Pearson’s χ² tests.

²Fisher’s exact tests as indicated.

Incidence of prespecified complications - including hypoglycaemia, hypokalaemia, pulmonary oedema, acute kidney injury and arrhythmia - from the initiation of fluid resuscitation until hospital discharge. Values are the number of patients (%) in each group, together with absolute risk differences and corresponding 95% confidence interval. USCOM: Ultrasonic cardiac output monitor; CI: Confidence interval.

This information indicates that USCOM-guided fluid resuscitation reduced net fluid accumulation and hospital stay without increasing the overall incidence of prespecified complications.

Subgroup analyses

Prespecified subgroup analyses stratified participants by age (< 60 years vs ≥ 60 years) to explore whether the advantages of USCOM guidance differed between younger and older adults (Tables 7 and 8).

Table 7 Fluid intake, net fluid balance, and urine output in participants younger than 60 years, mean ± SD.

Time point	USCOM group (n = 17)	Control group (n = 21)	t-value	Cohen’s d	Mean difference	95%CI	P value
Fluid replenishment (mL)
24 hours	6637.35 ± 2220.25	6232.43 ± 1402.32	0.69	0.22	404.92	-794.0 to 1603.8	0.497
48 hours	3035.59 ± 921.23	3141.10 ± 1403.68	-0.27	0.09	-105.51	-907.6 to 696.6	0.787
72 hours	1727.76 ± 456.78	2300 ± 699.01	-2.91	0.95	-572.24	-973.1 to -171.3	0.006a
Total 72 hours	11400.71 ± 2677.67	11669.24 ± 2840.30	-0.30	0.10	-268.53	-2079.5 to 1542.5	0.773
Net fluid replenishment (mL)
24 hours	2759.12 ± 2007.14	3200.76 ± 1081.40	-0.82	0.27	-441.64	-1555.0 to 671.7	0.424
48 hours	2748.94 ± 1400.17	4036.62 ± 1816.50	-2.40	0.78	-1287.68	-2376.0 to -199.4	0.021a
72 hours	2796.71 ± 1579.76	4442.81 ± 2007.73	-2.76	0.89	-1646.10	-2855.7 to -436.5	0.009
Urine output (mL)
24 hours	3878.24 ± 1083.97	3031.67 ± 1144.61	2.32	0.76	846.57	108.7-1584.4	0.026a
48 hours	3045.76 ± 1584.61	2305.24 ± 1134.93	1.68	0.54	740.52	-153.2 to 1634.2	0.102
72 hours	1680 ± 635.42	1899.05 ± 1016.13	-0.81	0.25	-219.05	-768.5 to 330.4	0.423

^aP < 0.05.

Subgroup analysis of cumulative fluid intake, net fluid balance, and urine output at 24 hours, 48 hours, and 72 hours and over the first 72 hours among participants aged < 60 years in ultrasonic cardiac output monitor-guided and control groups. Data presentation, definitions of time intervals, and statistical methods are as described for Table 3. USCOM: Ultrasonic cardiac output monitor; CI: Confidence interval.

Table 8 Fluid intake, net fluid balance, and urine output in participants aged 60 years or older, mean ± SD.

Time point	USCOM group (n = 8)	Control group (n = 4)	t-value	Cohen’s d	Mean difference	95%CI	P value
Fluid replenishment (mL)
24 hours	6252 ± 734.71	6222.25 ± 1417.23	0.05	0.03	29.75	-1274.9 to 1334.4	0.961
48 hours	3089 ± 899.41	2327.75 ± 791.66	1.43	0.87	761.25	-372.2 to 1894.7	0.184
72 hours	1800.5 ± 665.55	2503 ± 793.00	-1.63	1.00	-702.5	-1623.8 to 218.8	0.135
Total 72 hours	11141.5 ± 1807.49	11053 ± 1952.32	0.08	0.05	88.5	-2342.9 to 2520.0	0.939
Net fluid replenishment (mL)
24 hours	2517.125 ± 1217.20	4139.75 ± 1496.80	-2.04	1.24	-1622.625	-3286.0 to 39.7	0.069
48 hours	3327.625 ± 1620.86	4938.75 ± 1473.68	-1.67	1.02	-1611.125	-3686.8 to 464.5	0.127
72 hours	3170.625 ± 1318.32	5899.25 ± 1246.86	-3.44	2.10	-2728.625	-4385.3 to -1071.9	0.005a
Urine output (mL)
24 hours	3734.875 ± 761.75	2082.5 ± 517.27	3.87	2.37	1652.375	701.1-2603.7	0.002a
48 hours	2278.5 ± 491.07	1528.75 ± 395.61	2.64	1.61	749.75	116.4-1383.2	0.024a
72 hours	1957.5 ± 1096.61	1542.5 ± 474.26	0.71	0.43	415	-884.3 to 1714.3	0.495

^aP < 0.05.

Subgroup analysis of cumulative fluid intake, net fluid balance and urine output at 24 hours, 48 hours and 72 hours and over the first 72 hours among participants aged ≥ 60 years in ultrasonic cardiac output monitor-guided and control groups. Data presentation, definitions of time intervals, and statistical methods are as described for Table 3. USCOM: Ultrasonic cardiac output monitor; CI: Confidence interval.

Among individuals younger than 60 years (USCOM n = 17, control n = 21), total 72-hour fluid intake remained similar between groups, but net fluid balance at 72 hours was significantly lower in the USCOM group (2796.71 mL ± 1579.76 mL vs 4442.81 mL ± 2007.73 mL; mean difference -1646.10 mL; 95%CI: -2855.70 to -436.50; P = 0.009). Early urine output at 24 hours was higher with USCOM guidance (3878.24 ± 1083.97 mL vs 3031.67 ± 1144.61 mL; mean difference 846.57 mL; 95%CI: 108.70-1584.40; P = 0.026), consistent with improved renal perfusion.

In participants aged 60 years or older (USCOM n = 8, control n = 4), the magnitude of benefit appeared even greater, although the sample size was small. While total 72-hour fluid intake did not differ meaningfully between groups, net fluid balance at 72 hours was significantly lower in the USCOM group (3170.63 ± 1318.32 mL vs 5899.25 ± 1246.86 mL; mean difference -2728.63 mL; 95%CI: -4385.30 to -1071.90; P = 0.005). At 24 hours and 48 hours, net fluid accumulation in older patients tended to be lower and urine output higher under USCOM guidance, with multiple comparisons approaching or reaching statistical significance.

These subgroup analyses suggest that dynamic, non-invasive haemodynamic monitoring may be particularly beneficial for older adults with DKA, who are at increased risk of volume overload-related complications, while also providing meaningful reductions in net fluid balance in younger patients.

Evidence from the embedded systematic review

The search identified 25 studies on DKA rehydration. After excluding one case report, two systematic reviews, and two studies involving clinician populations, 20 primary studies were included in the evidence synthesis[23-42]. Among these, nine studies focused on children (6040 participants), ten focused on adults (2322 participants), and one included both adults and children (20 participants), obtaining a total sample of 8382 individuals. Approximately 72% of all cases came from paediatric cohorts, underscoring that most studies on DKA rehydration have focused on children, despite the substantial burden of DKA in adult populations.

To structure the synthesis, we grouped the 20 studies into three thematic domains: (1) Relationships between rehydration time and complications (n = 8); (2) Choice of rehydration modality (n = 9); and (3) Determination of rehydration volume and rate (n = 3). Across these domains, the evidence demonstrated that inappropriate fluid methods, whether excessively rapid or insufficiently tailored, were associated with neurological complications, prolonged hospitalisation, and in some cases, acute kidney injury. However, no study incorporated dynamic, non-invasive haemodynamic monitoring or used SV-guided protocols. Additionally, heterogeneity in patient populations, fluid regimens and outcome definitions limited the feasibility of formal quantitative meta-analysis for most endpoints; we therefore depend primarily on narrative synthesis, with any exploratory pooled estimates reported in Supplementary Tables 3-5.

The systematic review highlights three key gaps that our pilot trial begins to address. First, adults with DKA remain underrepresented in the literature, even though adults with comorbid heart or kidney disorders may be particularly vulnerable to fluid overload. Second, existing protocols are largely based on fixed weight-based formulas rather than real-time physiological assessment. Third, there is no randomized evidence on the use of non-invasive haemodynamic monitoring in DKA. The present USCOM-guided trial directly targets this gap by testing an individualised, doppler-based resuscitation strategy in an adult emergency cohort, thereby complementing and extending the primarily paediatric, protocol-driven evidence identified in the review. Key characteristics and lessons from the included studies are summarised in Table 9 and Supplementary Table 3.

Table 9 Evidence landscape for fluid resuscitation in diabetic ketoacidosis.

Evidence domain	Direction	Certainty	Clinical maturity	Core interpretation
Guideline-based fluid resuscitation	↑	Moderate	Established care	Fluid resuscitation remains foundational, but most protocols rely on fixed formulas and static clinical assessment
Fluid composition: Balanced crystalloids vs normal saline	↗	Low-moderate	Selective clinical use	Balanced fluids may improve acid-base or chloride profiles, but evidence remains heterogeneous
Fluid rate and volume strategies	?/↗	Low-moderate	Contextual evidence	Paediatric evidence predominates; optimal adult-specific rates and volumes remain uncertain
Rehydration timing and complications	?	Low	Contextual/risk evidence	Excessive or poorly tailored fluid strategies are associated with neurological complications, AKI, or prolonged hospitalisation, but causal inference is limited
Protocol/order-set implementation	↗	Low-moderate	Implementation evidence	Standardised protocols may improve adherence and DKA resolution but may introduce trade-offs such as hypoglycaemia
Adult DKA evidence-based	?	Low	Evidence gap	Adults remain underrepresented despite high clinical burden and comorbidity-related fluid-overload risk
Dynamic haemodynamic monitoring	↗	Low	Experimental clinical approach	No previous included study used non-invasive haemodynamic or stroke-volume-guided protocols; the present pilot trial provides proof-of-concept evidence
High-risk adults, older patients, and those with cardiac/renal comorbidity	?	Very low	Hypothesis-generating	These populations may benefit most from individualised monitoring, but direct evidence is sparse
Resource-limited emergency settings	↗	Low	Translational promise	Portable, non-invasive monitoring may be scalable when invasive monitoring is unavailable, but evidence of its implementation is needed

This table summarises the direction, certainty, and clinical maturity of evidence across primary domains identified in the embedded narrative synthesis of studies on diabetic ketoacidosis (DKA) rehydration. Evidence was organised into domains concerning rehydration timing and complications, fluid type, fluid volume and rate, protocol implementation, adult evidence gaps, and dynamic haemodynamic monitoring. Direction symbols indicate whether the available evidence supports established or clinically relevant practice, suggests promising but preliminary findings, or remains uncertain or indirect. The clinical translation continuum distinguishes established guideline-based care, selective clinical use, contextual or implementation evidence, experimental clinical approaches, and evidence gaps requiring future trials. Overall, the evidence supports the central role of fluid resuscitation in DKA, but most existing studies rely on fixed protocols, fluid-composition comparisons, or retrospective data on implementation. Adult DKA populations are underrepresented, and no previous study in the included evidence base used non-invasive haemodynamic monitoring or stroke-volume-guided fluid resuscitation. The present ultrasonic cardiac output monitor-guided pilot trial therefore occupies the experimental clinical approach stage and provides proof-of-concept evidence for a physiology-guided strategy. Certainty categories are narrative and reflect study design, population relevance, consistency, and directness. “↑”: Established or clinically relevant evidence; “↗”: Promising but preliminary evidence; “?”: Uncertain, indirect, or heterogeneous evidence; AKI: Acute kidney injury; DKA: Diabetic ketoacidosis.

DISCUSSION

Fluid management implications

In this pilot randomized trial, USCOM-guided, PLR-based fluid resuscitation reduced cumulative 72-hour net fluid balance by approximately 1.8 L compared with guideline-based usual care, while increasing early urine output and shortening hospital length of stay by almost two days. Notably, these benefits were achieved without delaying DKA resolution, impairing glycaemic control, or increasing acute kidney injury or other recorded complications. These results suggest that combining dynamic haemodynamic monitoring into DKA protocols may improve the efficiency of fluid resuscitation, rather than merely restricting fluid administration.

The observed pattern of increased early diuresis alongside reduced net fluid accumulation suggests that USCOM-guided management may optimise renal perfusion and promote earlier mobilisation of excess fluid. For patients with DKA, who may have pre-existing microvascular or renal vulnerability, minimising unnecessary fluid overload is clinically relevant, even when short-term creatinine-based measures of kidney injury do not differ between groups. By identifying truly fluid-responsive patients using a ΔSV ≥ 10% threshold during PLR, this strategy allows clinicians to target fluid boluses to those most likely to benefit, while avoiding additional fluid loading in non-responders.

Research significance

Current DKA guidelines emphasise prompt fluid resuscitation as the initial step in management to restore circulating volume and renal function. However, they also acknowledge the limited evidence supporting dynamic haemodynamic guidance and continue to rely largely on empirical “deficit plus maintenance” formulae and static clinical signs. In adults, estimated fluid deficits of approximately 100 mL/kg are typically corrected using weight-based protocols, and in children, approximately 70 mL/kg. These formulae, however, do not account for inter-individual variation in cardiac function, vascular permeability or osmotic shifts, particularly in patients with heart failure, chronic kidney disorder or haemodynamic instability[43-46].

Our embedded systematic review of 20 primary studies involving 8382 patients demonstrated that research on DKA rehydration has predominantly focused on paediatric populations, with adults being markedly under-represented. Most existing studies compared fixed-rate protocols or fluid composition, rather than dynamic physiological guidance. Across the literature, inappropriate fluid methods, whether excessively rapid or insufficiently individualised, were associated with neurological complications, prolonged hospitalisation, and in some cases, acute kidney injury. However, none of the included studies incorporated non-invasive haemodynamic monitoring or SV-guided protocols. Against this background, the present trial provides proof-of-concept randomized evidence that a simple bedside doppler-based strategy may safely reduce net fluid accumulation and length of hospital stay in adults with DKA.

Moreover, the findings are relevant to emerging DKA phenotypes. Normoglycaemic or euglycaemic DKA associated with sodium-glucose cotransporter 2 inhibitors often occurs in patients with pre-existing cardiac dysfunction, in whom aggressive volume loading poses a heightened risk of heart failure[46]. The ability of USCOM to provide beat-to-beat information on SV and CO at the bedside provides a pragmatic way to individualise resuscitation in these high-risk populations, where traditional weight-based algorithms may be especially unsafe.

By integrating a prospective pilot trial with a structured narrative synthesis of the broader literature, this mixed-methods study links gaps in the global evidence base with directly observed clinical effects in a resource-limited emergency department. The convergence of these two components strengthens the inference that more physiologically informed, non-invasive methods may provide a realistic pathway to safer and more efficient DKA resuscitation.

Limitations and generalisability

Several important limitations related to the study population, research design, definition of the disease concept, and statistical analysis should be fully considered before it can be widely promoted globally.

Our cohort was drawn predominantly from a Han Chinese population in a tertiary emergency department with standardised protocols and experienced clinical staff. A recent meta-analysis has highlighted significant variation in adherence to DKA resuscitation guidelines across institutions, with protocol compliance ranging from 38% to 82%[47]. Practice patterns in community hospitals or primary care facilities, especially in low- and middle-income countries, may differ markedly from those in our setting. External validity therefore remains uncertain and should be explored in more diverse health systems, including lower-level hospitals and non-tertiary centres, where the burden of DKA is high, and access to intensive care resources is limited.

DKA diagnosis in the study design warrants clarification. We seek to clarify that the DKA diagnosis was not the crucial methodological determinant in this fluid resuscitation study. During patient recruitment, DKA diagnosis was assessed and confirmed by senior emergency medicine specialists. This study considered DKA primarily as a binary clinical state, namely the presence of DKA at enrollment and resolution of DKA following treatment, rather than modelling dynamic alterations in DKA severity. We acknowledge that serial assessment of essential variables reflecting DKA severity during fluid treatment, particularly at each fluid resuscitation step and daily, would have strengthened the physiological interpretation and internal validity of this proof-of-concept pilot study. In a future definitive trial, we will adopt internationally recognised diagnostic and severity classification criteria for DKA and will prospectively assess relevant crucial variables at predefined intervals throughout treatment.

The new DKA treatment endpoint and the definition of ketone clearance. We seek to clarify the rationale for our method. We previously reviewed a DKA fluid resuscitation study conducted in the United States, in which the reported treatment time was approximately three days. In our clinical context in China, however, a three-day period is insufficient to capture the full course of DKA management. According to our hospital experience, the average duration of hospital-based DKA treatment is approximately 14 days. Nevertheless, this 14-day period includes not only fluid resuscitation, but also the treatment of complications, management of precipitating conditions, and other supportive care. As the present study specifically examined the effect of titrated fluid resuscitation in DKA, the period of active fluid treatment, estimated at approximately three days in some United States-based studies, may provide a more representative treatment window for evaluating the direct effect of the fluid strategy. Additionally, some patients may continue to remain in the hospital after DKA correction because of underlying disorders or comorbid conditions; this additional time is usually included in the total length of hospital stay. Ketone clearance is also a dynamic process. Some patients with DKA may still require fluid therapy beyond day three following admission. To balance these clinical differences and uncertainties, we defined the endpoint as follows. In other words, the treatment duration in our study reflects the time required to resolve DKA-related fluid therapy and to manage DKA-related complications, rather than the entire period of hospitalisation for all underlying or concurrent conditions.

Multiple testing and subgroup analyses. We seek to clarify the rationale for our stratified analytical method. The purpose of the stratified analyses was exploratory and concept-driven. Specifically, we aimed to examine which fluid-related indicators, including fluid replenishment, urine output, and net fluid balance, might be most responsive to USCOM-guided titrated fluid resuscitation. The significant finding was that this titrated strategy essentially reduced net fluid accumulation during DKA fluid treatment while maintaining clinical efficacy compared with the control group. Age-based stratification was performed to provide a preliminary conceptual assessment of whether older and younger patients differed in their response to USCOM-guided titrated fluid resuscitation. Understanding these differences may help clinicians determine whether this method is particularly valuable for patients at higher risk of fluid overload. For instance, in a patient aged 60 years or older with DKA who meets the criteria for USCOM assessment, titrated fluid resuscitation may reduce fluid burden and potentially lower the risk of fluid-related complications. For older patients with DKA who may have limited physiological reserve, this represents a promising method. In this pilot study, the subgroup analyses should therefore be interpreted as exploratory and hypothesis-generating rather than as confirmatory multiple comparisons. They were not intended to establish definitive differences in treatment outcomes between subgroups. Instead, they were undertaken to support conceptual validation of the intervention and to inform the design of future adequately powered trials. In a future large-scale clinical study, prespecified subgroup hypotheses, hierarchical testing procedures, and appropriate adjustment for multiple comparisons should be considered.

The concern is that net fluid balance is a controversial primary endpoint. We agree that this outcome primarily reflects the treatment process and does not, by itself, necessarily establish a direct connection with clinically meaningful patient results. This issue was carefully considered during the development of the study protocol. One of the primary objectives of this USCOM-guided titrated fluid resuscitation study was to evaluate fluid administration and its derived variables, including net fluid balance. At the design stage, we recognised that a more conventional disease-specific outcome for DKA might have been selected as the primary endpoint. However, based on our clinical experience, a lower cumulative fluid burden during the first three days of treatment is often associated with improved patient recovery, provided that DKA resolution and metabolic control are not compromised. The central clinical question was therefore: How can fluid administration during the first three days be reduced without impairing the therapeutic efficacy of DKA treatment, and is a titrated fluid resuscitation method feasible for this purpose? For this reason, fluid administration and net fluid balance were selected as primary process-related endpoints in this pilot study. Throughout the study, all patients received standard DKA care, including metabolic correction, electrolyte management, and monitoring for complications. We acknowledge that this primary endpoint design is relatively new. Therefore, when comparing our results with those of similar studies, it is crucial to recognise that the outcome definition used in this trial is not a traditional DKA disease-resolution endpoint, but a treatment-efficiency endpoint intended to evaluate whether USCOM-guided titration can lower unnecessary fluid exposure while preserving clinical effectiveness.

The basis for the sample size calculation. We identified an interesting issue when reviewing the original study design and theoretical sample size estimation. Based on our clinical experience with DKA fluid therapy, a lower total fluid burden is often associated with better overall clinical recovery, including shorter recovery time, fewer or milder complications, and lower medical costs[19]. Therefore, when planning the sample size calculation, we defined a 30% reduction in fluid administration within 48 h as the ideal expected effect. However, the observed findings demonstrated that total fluid administration was broadly similar between groups, with a statistically significant difference emerging only at 72 hours. During subsequent analysis, we identified that the primary effect of the titrated fluid resuscitation strategy was reflected not in a large reduction in total fluid administration, but in a reduction in net fluid balance. Moreover, this finding illustrates an essential clinical phenomenon: Total fluid administration in DKA is difficult to reduce directly and significantly through any single approach because adequate fluid replacement remains necessary for circulatory restoration and metabolic recovery. This may help explain why international guidelines recommend considering multiple clinical variables when estimating fluid replacement requirements, rather than providing a single universally precise formula for fluid calculation. The effect of USCOM-guided titrated fluid therapy may therefore lie not simply in reducing the total value of fluid administered, but in improving the efficiency and precision of fluid resuscitation by identifying patients who are truly fluid-responsive and minimising unnecessary net fluid accumulation.

This proof-of-concept pilot study incorporated an embedded systematic review to provide an external contextual validation tool. Our PubMed search did not identify any previous studies evaluating USCOM-guided or titrated fluid resuscitation in patients with DKA. Consequently, rather than directly comparing our findings with closely similar USCOM-guided DKA studies, we shifted the focus of comparison towards the historical development and research exploration of fluid resuscitation approaches in DKA. In this context, the studies identified were highly heterogeneous in terms of population, fluid strategy, intervention focus, outcome definition, and study design. A quantitative meta-analysis was therefore unfeasible or methodologically inappropriate. The most appropriate method was to classify the literature by the clinical task addressed in each fluid resuscitation study and subsequently provide a high-level synthesis of the lessons learned within each task domain. For this reason, we did not extend the search to additional databases or perform a formal risk-of-bias assessment (Supplementary Tables 1 and 2). Even if these steps had been undertaken, they would not have resolved the substantial clinical and methodological heterogeneity of the available literature, nor would they have identified directly comparable studies of USCOM-guided titrated fluid resuscitation in DKA. Our experience suggests that embedding a focused mini-systematic review within a case report or pilot clinical study can strengthen the interpretations of findings by placing them within the broader evidence landscape and providing readers with a more comprehensive framework for evaluating the results[21].

Follow-up was limited to the index hospitalisation. Although reduced fluid overload during acute resuscitation is biologically plausible as a strategy for protecting organ function, we did not measure long-term renal or neurological outcomes. Observational studies in critically ill populations have linked fluid overload to an increased risk of chronic kidney disease[39]. However, whether the approximately 1.8-litre reduction in net fluid balance observed in this study translates into sustained organ protection among DKA survivors remains unknown. Future studies should incorporate longitudinal follow-up to capture chronic kidney disorder, recurrent DKA, and functional outcomes. They should evaluate whether dynamic monitoring during the acute phase can influence longer-term clinical trajectories.

Additionally, treating clinicians could not be blinded to group allocation, potentially introducing performance bias in clinical decisions beyond the protocolised fluid boluses. We sought to minimise this risk through standardised insulin and electrolyte replacement protocols and blinded outcome assessment. Nevertheless, the possibility of residual bias cannot be excluded.

CONCLUSION

This pilot trial supports the use of non-invasive haemodynamic monitoring as a promising approach to individualising fluid resuscitation in adult DKA. The embedded systematic review served as an external contextual validation tool, confirming that the pursuit of more efficient, precise and individualised fluid replacement strategies has been a long-standing but unresolved research priority. In the present pilot study, the clinical benefits of USCOM-guided titrated fluid resuscitation were conceptually and preliminarily demonstrated, with these benefits being associated with reduced net fluid accumulation. These proof-of-concept findings provide a foundation for future large-scale clinical trials evaluating USCOM-guided titrated fluid resuscitation in adult DKA.

ACKNOWLEDGEMENTS

The authors thank Yangzhou University Affiliated Hospital for technical assistance.