Pulse pressure versus continuous flow monitoring for fluid responsiveness prediction during passive leg raising: a prospective diagnostic accuracy study

Abstract

Background

Accurate prediction of fluid responsiveness is essential in managing circulatory shock. While passive leg raising (PLR) is recommended in international guidelines, its diagnostic validity depends critically on the monitoring modality used to detect hemodynamic responses. In resource-limited settings, manual blood pressure measurement remains the predominant available technology, yet its adequacy for PLR interpretation has not been rigorously evaluated in an adequately powered study with an independent reference standard.

Objective

To compare the diagnostic accuracy of manual pulse pressure change (ΔPP%) versus continuous flow monitoring (USCOM-1 A, ΔSV%) for predicting fluid responsiveness during PLR in intensive care unit (ICU) patients with undifferentiated shock, using echocardiographic cardiac output assessment as an independent reference standard.

Methods

Prospective single-center diagnostic accuracy study enrolling consecutive ICU patients with undifferentiated shock at Sina Educational and Medical Center, Tehran, Iran (2020–2021). A standardized PLR protocol was performed with simultaneous blinded measurements by three independent operators: manual blood pressure via calibrated sphygmomanometer, continuous flow monitoring via USCOM-1 A, and echocardiographic velocity-time integral (VTI) measurement as the reference standard. Fluid responsiveness was defined as ≥ 15% cardiac output increase following 500 mL crystalloid challenge. Primary outcome was area under the receiver operating characteristic curve (AUC). The study followed Standards for Reporting Diagnostic Accuracy Studies (STARD) 2015 guidelines (Ethics approval: IR.SBMU.PHARMACY.REC.1399.316).

Results

Of 124 patients analyzed (mean age 58.1 ± 19.8 years, 65.3% female), 58 (46.8%) were fluid responders. Continuous flow monitoring demonstrated acceptable diagnostic accuracy (AUC 0.712, 95% confidence interval [CI] 0.622–0.802; sensitivity 81.0%; specificity 61.8%; negative predictive value [NPV] 79.2%). Manual pulse pressure demonstrated significantly inferior discrimination (AUC 0.601, 95%CI 0.502–0.700; DeLong test P = 0.029) with critically low sensitivity (32.8%) and clinically inadequate NPV (57.8%). Bootstrap-corrected performance estimates confirmed the robustness of findings (USCOM AUC 0.708; Manual pulse pressure [PP] AUC 0.596). Mean pulse pressure remained essentially unchanged during PLR (+ 0.4 ± 4.6 mmHg, P = 0.287) despite significant proportional increases in both systolic (+ 3.7%) and diastolic (+ 5.6%) pressures, and despite echocardiographically confirmed stroke volume augmentation (left ventricular outflow tract [LVOT] VTI + 15.9%). The correlation between ΔPP% and ΔSV% was weak (r = 0.39, R²=0.15). Baseline method comparison between USCOM-derived and echocardiography-derived cardiac output showed low paired bias, although the percentage error exceeded the conventional interchangeability threshold, supporting interpretation of USCOM as a directional flow-monitoring index rather than a fully interchangeable substitute for echocardiography.

Conclusions

In this single-center study, manual pulse pressure demonstrated insufficient sensitivity (32.8%) and inadequate NPV (57.8%) for reliable fluid responsiveness prediction during PLR. Our findings suggest it should be interpreted with considerable caution, particularly when the test result is negative. Continuous flow monitoring provides superior and clinically acceptable diagnostic accuracy when available, although USCOM should be interpreted as a directional flow-monitoring index rather than a fully interchangeable substitute for echocardiographic cardiac output measurement. The proportional pressure increase phenomenon provides a mechanistic explanation for this diagnostic limitation. These findings require multicenter validation before definitive practice recommendations can be issued.

Background

The clinical problem: shock and fluid management

Circulatory shock represents acute circulatory failure resulting in inadequate cellular oxygen delivery and life-threatening tissue hypoperfusion. The Global Burden of Disease Study estimated 48.9 million cases of sepsis annually with 11 million sepsis-related deaths in 2017, representing 19.7% of global mortality [1]. When incorporating all shock etiologies, the annual incidence in intensive care units approaches 30–40% of all admissions with mortality ranging from 20% to 60% [2, 3].

The cornerstone of initial shock management across all etiologies remains hemodynamic optimization through judicious fluid resuscitation based on the Frank-Starling principle: increasing preload through fluid administration augments stroke volume and cardiac output in patients positioned on the ascending limb of the Frank-Starling curve [4, 5]. However, this therapeutic approach faces a persistent clinical paradox. Inadequate fluid resuscitation leads to tissue hypoxia and organ dysfunction, while excessive fluid administration causes equally severe harm through acute respiratory distress syndrome, prolonged mechanical ventilation, acute kidney injury, and increased mortality risk [6, 7]. The critical question becomes: which patients, at which timepoint, will derive benefit from additional fluid administration?

Failure of static parameters and the PLR solution

Traditional static hemodynamic parameters have proven inadequate for guiding fluid decisions. Central venous pressure, long taught as the primary preload indicator, demonstrates a pooled AUC of only 0.56 in meta-analyses, essentially equivalent to chance [8]. Pulmonary artery occlusion pressure shows equally poor predictive utility, while routine pulmonary artery catheter use has not demonstrated mortality benefit and carries procedural risk [9]. The fundamental limitation of static parameters is their failure to account for individual patient position on the Frank-Starling curve.

Recognizing these limitations, the critical care community shifted toward dynamic assessment of fluid responsiveness. Passive leg raising, systematically validated and refined by Monnet and Teboul, emerged as an elegant solution through gravity-mediated transfer of blood from the lower extremities and splanchnic compartment to the central circulation, creating rapid, reversible autotransfusion without actual fluid administration [10, 11]. This approach offers reversibility with no risk of iatrogenic fluid loading, rapid onset with peak effect typically occurring within 60–90 s, repeatability, and applicability in both ventilated and spontaneously breathing patients [10, 11].

The evidence base supporting PLR is substantial. Meta-analyses have demonstrated high diagnostic performance of PLR for predicting fluid responsiveness, particularly when PLR-induced changes are assessed using real-time cardiac output or stroke volume monitoring [12–14]. However, a critical caveat exists: the highest diagnostic performance of PLR has been demonstrated when the maneuver is interpreted using real-time changes in cardiac output or stroke volume, whereas PLR-induced changes in arterial pulse pressure show lower diagnostic performance [13, 14].

The evidence-practice gap

The diagnostic accuracy of PLR is strongly dependent on the measurement tool used to detect the hemodynamic response [13, 14]. This creates a clinically important dilemma between evidence-supported practice and global healthcare reality. Prior studies and meta-analyses demonstrate excellent PLR performance when interpreted using real-time changes in cardiac output or stroke volume, including pulse contour analysis, esophageal Doppler, bioreactance, or echocardiographic assessment [12–14]. In contrast, many hospitals worldwide, particularly in low- and middle-income countries, have limited access to advanced continuous cardiac output monitoring and often rely on intermittent blood pressure measurement, basic monitoring, and clinical reassessment [15–17]. Recent expert consensus statements for sepsis care in resource-limited settings similarly emphasize pragmatic clinical assessment and simple tests to guide resuscitation when advanced hemodynamic monitoring is unavailable [17]. This creates a critical evidence-practice gap: the PLR evidence base was built largely with monitoring technologies that may be unavailable in many ICUs globally.

Physiological basis for pulse pressure limitations

The theoretical basis for using pulse pressure (PP = systolic blood pressure − diastolic blood pressure) to estimate stroke volume changes derives from the simplified Windkessel relationship, in which pulse pressure is related to stroke volume and arterial compliance [18]. This relationship becomes unreliable in critically ill patients because arterial compliance and vascular tone are dynamic rather than fixed and may be profoundly modified by aging, inflammation, vasopressors, sympathetic activation, and shock physiology [18, 19]. Additionally, PLR may trigger autonomic and vascular responses that affect systolic and diastolic pressures differently. If both pressure components rise proportionally, net pulse pressure may remain unchanged despite increased stroke volume. Manual blood pressure measurement also requires time to complete, introducing timing uncertainty during PLR’s short-lived peak hemodynamic effect [10, 11].

Study rationale and objectives

Despite recommendations favoring dynamic assessment of fluid responsiveness, including PLR when appropriately monitored [3, 20], adequately powered and methodologically rigorous evidence comparing manual pulse pressure versus validated flow-based methods for PLR interpretation remains limited. Existing studies evaluating noninvasive blood pressure during PLR have been limited by differences in measurement technique, study population, sample size, or lack of independent reference standards [21, 22].

We designed this prospective diagnostic accuracy study to address whether manual pulse pressure change (ΔPP%) can substitute for continuous flow monitoring (USCOM-1 A, ΔSV%) in predicting fluid responsiveness during PLR in ICU patients with undifferentiated shock, using independent echocardiographic assessment as the reference standard.

We hypothesized that manual pulse pressure would demonstrate significantly lower sensitivity than flow monitoring due to the discontinuous nature of cuff measurement, complex pressure-flow relationships in critically ill patients, and proportional increases in both systolic and diastolic pressure during PLR potentially obscuring net pulse pressure change. Our study addresses prior limitations through an independent echocardiographic reference standard, blinded independent measurement, adequate statistical power, consecutive enrollment, and comprehensive diagnostic accuracy analysis following STARD 2015 guidelines [23].

Methods

Study design and setting

This prospective diagnostic accuracy study was conducted from 2020 to 2021 in the medical-surgical ICU of Sina Educational and Medical Center, Tehran, Iran, a tertiary academic hospital affiliated with Tehran University of Medical Sciences. This study was extracted from the master’s thesis in nursing of Mansoureh Ashoub and conducted with financial support from the Research Vice-Chancellor of Shahid Beheshti University of Medical Sciences, Tehran, Iran. The study followed STARD 2015 guidelines [23] and received prospective ethics approval from the Ethics Committee of Shahid Beheshti University of Medical Sciences (Ethics Code: IR.SBMU.PHARMACY.REC.1399.316). The study was conducted in full accordance with the Declaration of Helsinki (2013 revision).

Written informed consent was obtained from all patients or their legally authorized representatives before enrollment. In cases where patients were unable to provide informed consent due to altered mental status, sedation, or mechanical ventilation, written informed consent was obtained from legally authorized representatives as approved by the ethics committee.

Sample size calculation

Sample size was calculated to detect a clinically meaningful difference between two correlated AUC values using the method of Obuchowski and McClish [24]. Based on prior literature, we anticipated an AUC of 0.75 for flow monitoring and 0.60 for pulse pressure [12–14], with an estimated correlation coefficient of 0.40 between tests, two-sided α of 0.05, power of 90%, and expected prevalence of fluid responsiveness of 50% [2, 3]. This calculation yielded a required sample size of 118 patients. We targeted enrollment of 126 patients to account for anticipated 5–7% technical failures.

Participants

Inclusion criteria comprised adults aged ≥ 18 years admitted to the ICU with undifferentiated shock, defined as ≥ 2 of the following: MAP < 65 mmHg or vasopressor requirement to maintain MAP ≥ 65 mmHg, serum lactate ≥ 2.0 mmol/L, oliguria defined as urine output < 0.5 mL/kg/hour for ≥ 2 consecutive hours, skin mottling score ≥ 2 on the validated 0–5 scale [25], or acute alteration in mental status. Shock etiology had to be clinically uncertain within the first 2 h of ICU admission to ensure truly undifferentiated shock cases were enrolled.

Exclusion criteria included cardiac arrhythmias precluding reliable Doppler assessment, significant valvular heart disease, contraindications to PLR or supine positioning, technical exclusions due to inadequate USCOM or echocardiographic windows, pregnancy, do-not-resuscitate status, or refusal of informed consent.

Study protocol and blinding strategy

The study employed blinded, independent assessment by three operators performing simultaneous measurements while remaining blinded to each other’s results throughout the study. Operator 1 performed manual blood pressure measurements, blinded to USCOM and echocardiography results. Operator 2 performed continuous flow monitoring, blinded to manual blood pressure and echocardiographic findings. Operator 3 performed echocardiographic measurements and was independent from other operators and blinded to all index test results and responder classification.

Study timeline and fluid challenge protocol

The standardized protocol followed this sequence: initial resuscitation according to institutional ICU protocols, consistent with Surviving Sepsis Campaign recommendations [20], including crystalloid-based resuscitation, typically targeting approximately 30 mL/kg when clinically appropriate, followed by baseline measurements by all three operators with the patient semi-recumbent at 45 degrees after a 3-minute stabilization period, standardized PLR maneuver with simultaneous measurements during the maneuver, fluid challenge initiated as clinically feasible after PLR completion (median 12 min, IQR 8–18 min; range 5–30 min), and post-fluid measurements by all three operators simultaneously following fluid challenge completion.

All patients received 500 mL crystalloid (normal saline or lactated Ringer’s solution). Crystalloids distribute across both intravascular and interstitial compartments, while colloids remain predominantly intravascular due to higher oncotic pressure. This difference influences osmotic and hydrostatic pressure balance between intravascular and interstitial fluid compartments, potentially affecting hemodynamic response magnitude. This crystalloid choice reflects standard clinical practice and ensures applicability to commonly used resuscitation fluids.

The fluid challenge was administered to all enrolled patients regardless of PLR result as per the study protocol approved by the ethics committee, to avoid incorporation bias and ensure an independent reference standard. During the interval between PLR and fluid challenge, the following restrictions were enforced: no change in vasopressor infusion rates, no additional fluid boluses, no ventilator changes, and no invasive procedures. Patients who experienced significant clinical deterioration or improvement during this interval requiring protocol deviation (n = 2) were excluded from the final analysis.

Index test protocols

Manual pulse pressure measurement was performed using calibrated aneroid sphygmomanometers and appropriately sized cuffs. Baseline measurements were obtained after stabilization in the semi-recumbent position. During PLR, cuff inflation was initiated at 60 s post-maneuver with measurement completion targeted at 85–90 s. Change in pulse pressure percentage (ΔPP%) was calculated as [(PP_PLR − PP_baseline) / PP_baseline] × 100.

Continuous flow monitoring was performed using USCOM-1 A^® with a 2.2 MHz continuous-wave Doppler probe via suprasternal approach targeting the ascending aorta. Continuous recording was maintained from 30 s before through 120 s after PLR initiation. Baseline values represented the mean of final 30 s pre-PLR, and peak PLR response was the mean during the 85–95 s window. Change in stroke volume percentage (ΔSV%) was calculated as [(SV_PLR − SV_baseline) / SV_baseline] × 100.

Reference standard protocol

Echocardiographic stroke volume assessment using LVOT-VTI was selected as the independent reference standard based on its validation against thermodilution and widespread recognition as a reliable non-invasive reference standard for ICU cardiac output monitoring [26, 27]. LVOT diameter was measured in the parasternal long-axis view and averaged over 3 cardiac cycles. LVOT-VTI was measured in the apical 5-chamber view using pulsed-wave Doppler and averaged over 3 cycles. Stroke volume was calculated as π × (LVOT diameter/2)² × LVOT-VTI, and cardiac output as stroke volume × heart rate. Fluid responsiveness was defined as a ≥ 15% increase in cardiac output after a 500 mL crystalloid challenge. This threshold exceeds reported Doppler LVOT-VTI measurement imprecision and least significant change for transthoracic echocardiography in critically ill patients, supporting its use as a clinically meaningful response threshold consistent with prior fluid responsiveness studies [12, 13, 28]. Inter-rater reliability was ICC 0.91 (95%CI 0.83–0.96), assessed in 15 randomly selected cases by independent duplicate measurement.

Passive leg raising technique

PLR was performed using the standardized technique described by Monnet and Teboul [10, 11]. Initial position was semi-recumbent with trunk at 45 degrees and legs horizontal. A coordinated two-person team simultaneously lowered the trunk from 45° to 0° and raised legs from 0° to 45°, completed in < 5 s using an electric ICU bed, and maintained for 2 min to capture peak hemodynamic effects.

Data collection and variables

Baseline variables included age, sex, BMI, APACHE II score [29], SOFA score [30], serum lactate, creatinine, hemoglobin, arterial blood gas analysis, and comorbidities. Shock etiology was adjudicated retrospectively at 24–48 h by two independent intensivists using all available clinical, microbiological, imaging, and laboratory data, classified per Sepsis-3 criteria [31], with disagreements resolved by a third senior intensivist.

Statistical analysis

Statistical analysis was performed using IBM SPSS version 28.0, MedCalc version 20.1, and R version 4.2.1. Descriptive statistics used mean ± SD, median [IQR], and frequency (%) as appropriate. Primary analysis employed ROC curves for ΔPP% and ΔSV% against the echocardiographic reference standard with AUC calculated using the DeLong method [32]. Bootstrap validation was performed with 1,000 iterations. Bland–Altman analysis assessed baseline method comparison between USCOM-derived and echocardiography-derived cardiac output. Percentage error was interpreted in accordance with established cardiac output method-comparison principles [33].

Sensitivity analyses were performed restricted to patients with fluid challenge initiated ≤ 15 min after PLR and patients with no vasopressor dose changes during the interval. Pre-specified subgroup analyses included mechanical ventilation status, age, shock etiology, and baseline cardiac output. Missing data were minimal (< 2%), occurred randomly, and complete case analysis was deemed appropriate.

Quality assurance

Monthly calibration of sphygmomanometers was performed throughout the study. Operators completed standardized training and proficiency procedures. Inter-rater reliability was formally assessed through independent duplicate measurements in randomly selected cases. Protocol adherence was monitored through structured checklists, and double data entry with discrepancy resolution ensured data integrity.

Results

Patient enrollment and flow

During the study period, 178 consecutive patients with undifferentiated shock were screened, of whom 52 were excluded. The most common exclusions were cardiac arrhythmias (n = 18) and inadequate acoustic windows (n = 11). The remaining 126 patients were enrolled and underwent the complete protocol. Two patients were subsequently excluded from final analysis, yielding 124 patients for primary analysis (Fig. 1). Among the analyzed cohort, 58 patients (46.8%) met the pre-specified definition for fluid responsiveness.

Fig. 1.

Study flow diagram. Of 178 consecutive patients with undifferentiated shock screened during the study period, 52 were excluded. Of 126 enrolled patients, 2 were subsequently excluded from final analysis, yielding 124 patients for primary analysis. Among these, 58 (46.8%) were classified as fluid responders and 66 (53.2%) as non-responders based on the echocardiographic reference standard

Baseline patient characteristics

The study population represented a critically ill cohort with undifferentiated shock (Table 1). Mean age was 58.1 ± 19.8 years with 65.3% female. Severity scores confirmed significant critical illness with APACHE II 18.4 ± 6.8 and SOFA 8.2 ± 3.4. Serum lactate averaged 3.1 ± 1.8 mmol/L. Shock etiology demonstrated septic shock as most common (50.8%), followed by hypovolemic (22.6%), cardiogenic (14.5%), and mixed/indeterminate etiologies (12.1%).

Table 1.

Baseline patient characteristics (N = 124)

Characteristic	Overall (N = 124)	Responders (n = 58)	Non-Responders (n = 66)	P-value
Age, years	58.1 ± 19.8	54.8 ± 18.2	60.9 ± 20.8	0.088
Age ≥ 65 years, n (%)	47 (37.9)	18 (31.0)	29 (43.9)	0.139
Female sex, n (%)	81 (65.3)	39 (67.2)	42 (63.6)	0.676
BMI, kg/m²	24.3 ± 4.7	23.8 ± 4.2	24.7 ± 5.1	0.301
APACHE II	18.4 ± 6.8	17.9 ± 6.2	18.8 ± 7.3	0.479
SOFA score	8.2 ± 3.4	7.8 ± 3.1	8.5 ± 3.6	0.254
Lactate, mmol/L	3.1 ± 1.8	3.4 ± 1.9	2.9 ± 1.7	0.122
Septic shock, n (%)	63 (50.8)	24 (41.4)	39 (59.1)
Hypovolemic shock, n (%)	28 (22.6)	20 (34.5)	8 (12.1)
Cardiogenic shock, n (%)	18 (14.5)	4 (6.9)	14 (21.2)
Mixed/Indeterminate, n (%)	15 (12.1)	10 (17.2)	5 (7.6)	0.018†
Mechanical ventilation, n (%)	52 (41.9)	27 (46.6)	25 (37.9)	0.323
Vasopressor use, n (%)	88 (71.0)	38 (65.5)	50 (75.8)	0.205
NE dose, µg/kg/min‡	0.15 [0.08–0.29]	0.12 [0.06–0.23]	0.18 [0.10–0.34]	0.031§
Heart rate, bpm	98.4 ± 18.5	101.8 ± 19.2	95.5 ± 17.4	0.062
MAP, mmHg	97.0 ± 8.7	95.9 ± 8.2	97.9 ± 9.1	0.201
Pulse Pressure, mmHg	53.3 ± 11.2	51.8 ± 10.2	54.6 ± 11.9	0.165
USCOM Cardiac Output, L/min	3.9 ± 1.6	3.4 ± 1.2	4.2 ± 1.7	< 0.001§
USCOM Stroke Volume, mL	39.4 ± 18.2	33.8 ± 14.1	44.2 ± 20.0	0.002§
LVOT VTI, cm	14.5 ± 3.4	13.2 ± 3.1	15.6 ± 3.4	< 0.001§
Echo CO, L/min	3.5 ± 1.7	2.9 ± 1.3	4.0 ± 1.8	< 0.001§

†Chi-square test

‡Among patients receiving norepinephrine

§Statistically significant

Critically, fluid responders demonstrated significantly lower baseline cardiac output by both USCOM (3.4 ± 1.2 versus 4.2 ± 1.7 L/min, P < 0.001) and echocardiography (2.9 ± 1.3 versus 4.0 ± 1.8 L/min, P < 0.001), while baseline pulse pressure showed no significant difference between groups (51.8 ± 10.2 versus 54.6 ± 11.9 mmHg, P = 0.165). This baseline pressure-flow dissociation provides early evidence of the fundamental decoupling between pressure and flow parameters in this population.

Hemodynamic response to passive leg raising

PLR induced significant hemodynamic changes across all measured parameters, but with markedly different patterns between pressure-based and flow-based measurements (Table 2). All patients experienced statistically significant increases in both systolic and diastolic blood pressure components during PLR. However, mean pulse pressure remained essentially unchanged, increasing by only 0.4 ± 4.6 mmHg (P = 0.287). This occurred because diastolic pressure increased proportionally more than systolic pressure (+ 5.6% versus + 3.7%).

Table 2.

Hemodynamic changes during passive leg raising

Parameter	Baseline	Peak PLR (90s)	Absolute Change	% Change	P-value
SBP, mmHg	132.5 ± 20.4	137.3 ± 20.6	+ 4.8 ± 8.1	+ 3.7%	< 0.001*
DBP, mmHg	79.2 ± 15.5	83.6 ± 14.8	+ 4.4 ± 7.0	+ 5.6%	< 0.001*
Pulse Pressure, mmHg	53.3 ± 11.2	53.7 ± 11.5	+ 0.4 ± 4.6	+ 0.75%	0.287
MAP, mmHg	97.0 ± 8.7	101.5 ± 9.1	+ 4.5 ± 6.2	+ 4.6%	< 0.001*
Heart rate, bpm	98.4 ± 18.5	96.6 ± 18.0	−1.8 ± 5.1	−1.8%	< 0.001*
USCOM Stroke Volume, mL	39.4 ± 18.2	41.8 ± 17.5	+ 2.4 ± 6.1	+ 6.1%	< 0.001*
USCOM Cardiac Output, L/min	3.9 ± 1.6	4.0 ± 1.7	+ 0.2 ± 0.8	+ 4.2%	< 0.05*
LVOT VTI, cm	14.5 ± 3.4	16.8 ± 3.6	+ 2.3 ± 1.8	+ 15.9%	< 0.001*
Echo CO, L/min	3.5 ± 1.7	4.0 ± 1.8	+ 0.5 ± 0.9	+ 13.8%	< 0.001*

*Paired t-test

In contrast, flow-based measurements demonstrated hemodynamic augmentation during PLR. USCOM revealed a 6.1% increase in stroke volume (P < 0.001), with a modest derived cardiac output increase after accounting for the concomitant decrease in heart rate (+ 4.2%, P < 0.05). Echocardiographic measurements showed a 15.9% increase in LVOT-VTI (P < 0.001), corresponding to an estimated cardiac output increase of 13.8% (P < 0.001). The simultaneous observation of unchanged pulse pressure alongside significant echocardiographically confirmed stroke volume augmentation directly demonstrates pressure-flow dissociation.

Primary outcome: diagnostic accuracy

USCOM demonstrated acceptable diagnostic accuracy (AUC 0.712, 95%CI 0.622–0.802, P < 0.001), while manual pulse pressure demonstrated significantly inferior discrimination (AUC 0.601, 95%CI 0.502–0.700, P = 0.062 versus null). The DeLong comparison confirmed statistically significant superiority of USCOM over manual pulse pressure (P = 0.029) (Table 3; Fig. 2). Bootstrap validation confirmed the robustness of these findings, with bias-corrected AUC estimates of 0.708 for USCOM and 0.596 for manual pulse pressure.

Table 3.

Primary diagnostic accuracy comparison

Index Test	AUC	Bootstrap-Corrected AUC	95% CI	P-value	Interpretation
USCOM ΔSV%	0.712	0.708	0.618–0.798	< 0.001	Acceptable
Manual ΔPP%	0.601	0.596	0.496–0.695	0.062	Poor
DeLong test	Difference = 0.111	Difference = 0.112	—	0.029	USCOM superior

Fig. 2.

Receiver operating characteristic curves for predicting fluid responsiveness during passive leg raising. Receiver operating characteristic curves comparing continuous flow monitoring using USCOM-derived stroke volume change (ΔSV%) with manual pulse pressure change (ΔPP%). Continuous flow monitoring demonstrated significantly superior discrimination (AUC 0.712, 95%CI 0.622–0.802) compared with manual pulse pressure change (AUC 0.601, 95%CI 0.502–0.700; DeLong test P = 0.029). The diagonal dashed line represents no discrimination

Clinical performance characteristics and subgroup analyses

At the optimal threshold determined by Youden Index maximization, USCOM demonstrated sensitivity of 81.0%, specificity of 61.8%, and NPV of 79.2%. Manual pulse pressure demonstrated sensitivity of only 32.8% and NPV of 57.8% (Table 4). Exploratory subgroup analyses showed consistently poor manual pulse pressure performance, but subgroup analyses were underpowered and should be considered hypothesis-generating only.

Table 4.

Performance metrics and subgroup analyses

Parameter	USCOM ΔSV%	Manual ΔPP%	P-value
AUC (95% CI)	0.712 (0.622–0.802)	0.601 (0.502–0.700)	0.029*
Sensitivity, %	81.0	32.8	< 0.001
Specificity, %	61.8	78.8	0.024
PPV, %	64.4	57.6	—
NPV, %	79.2	57.8	—
+LR	2.12	1.55	—
−LR	0.31	0.85	—
Mechanically ventilated AUC	0.758	0.628	0.089
Spontaneously breathing AUC	0.682	0.583	0.267
Age < 65 years AUC	0.724	0.631	0.074
Age ≥ 65 years AUC	0.694	0.556	0.183
Septic shock AUC	0.702	0.608	0.244
Hypovolemic shock AUC	0.743	0.621	0.312
Cardiogenic shock AUC	0.696	0.542	0.389

USCOM–echocardiography agreement

Baseline comparison demonstrated low paired bias in the original Bland–Altman analysis (0.08 L/min, 95% limits of agreement − 1.12 to + 1.28 L/min). The percentage error was 34.2%, slightly exceeding the conventional 30% threshold commonly used to support interchangeability in cardiac output method-comparison studies [33]. Therefore, USCOM and echocardiography should not be considered fully interchangeable for absolute cardiac output values. However, no proportional bias was detected (P = 0.182), and ICC was 0.88 (95%CI 0.83–0.92), supporting the use of USCOM as a directional flow-monitoring index test in this study (Fig. 3).

Fig. 3.

Bland–Altman plot of baseline cardiac output. Bland–Altman plot showing baseline method comparison between USCOM-derived and echocardiography-derived cardiac output. The original analysis demonstrated low paired bias (0.08 L/min) with 95% limits of agreement from − 1.12 to + 1.28 L/min. The percentage error was 34.2%, exceeding the conventional 30% interchangeability threshold; therefore, USCOM and echocardiography should not be considered fully interchangeable for absolute cardiac output values

Sensitivity analyses

The superiority of USCOM over manual pulse pressure was consistent across timing-based sensitivity analyses (Table 5) and across varying thresholds for fluid responsiveness (Table 6).

Table 5.

Sensitivity analysis based on PLR-to-fluid-challenge interval

Analysis	n	USCOM AUC	Manual PP AUC	DeLong P
All patients	124	0.712	0.601	0.029
Interval ≤ 15 min	89	0.721	0.614	0.041
Interval ≤ 10 min	58	0.738	0.623	0.089
No vasopressor change	118	0.715	0.598	0.027

Table 6.

Sensitivity analysis for varying fluid responsiveness thresholds

Threshold	Prevalence	USCOM AUC	Manual PP AUC	DeLong P
ΔCO ≥ 10%	67.7%	0.695	0.614	0.187
ΔCO ≥ 12%	58.1%	0.704	0.608	0.109
ΔCO ≥ 15%	46.8%	0.712	0.601	0.029*
ΔCO ≥ 20%	31.5%	0.724	0.595	0.021*

An asterisk (*) indicates statistical significance

The proportional pressure increase phenomenon

Both systolic and diastolic blood pressure increased during PLR (+ 3.7% and + 5.6%), resulting in negligible net pulse pressure change (+ 0.75%, P = 0.287). This occurred simultaneously with increased USCOM stroke volume (+ 6.1%) and echocardiographic LVOT-VTI (+ 15.9%), providing direct empirical demonstration of pressure-flow dissociation (Fig. 4). The correlation between ΔPP% and ΔSV% was weak (r = 0.39, R²=0.15).

Fig. 4.

Proportional pressure increase phenomenon during passive leg raising. Hemodynamic changes during passive leg raising demonstrating the proportional pressure increase phenomenon. Both systolic and diastolic blood pressure increased significantly and proportionally (+ 3.7% and + 5.6%, respectively), resulting in negligible net pulse pressure change (+ 0.75%, P = 0.287). This occurred simultaneously with flow augmentation, reflected by increased USCOM-derived stroke volume (+ 6.1%) and echocardiographic LVOT-VTI (+ 15.9%), providing direct empirical demonstration of pressure-flow dissociation

Exploratory clinical outcomes

Exploratory analysis revealed that fluid-responsive patients had shorter ICU length of stay, lower cumulative fluid balance at 72 h, and a numerical trend toward lower 28-day mortality. These findings were not powered for clinical outcome inference and should be considered hypothesis-generating only.

Discussion

Principal findings

This prospective diagnostic accuracy study, one of the largest studies to date directly comparing manual blood pressure monitoring versus validated flow-based technology for PLR interpretation in undifferentiated shock, yields four principal findings requiring multicenter validation. First, manual pulse pressure demonstrated inadequate sensitivity and NPV. Second, continuous flow monitoring via USCOM provided significantly superior diagnostic accuracy. Third, the proportional pressure increase phenomenon explained the failure of pulse pressure monitoring: simultaneous rises in systolic and diastolic pressures resulted in negligible net pulse pressure change despite echocardiographically confirmed LVOT-VTI augmentation. Fourth, USCOM should be interpreted as a directional flow-monitoring index test rather than as a fully interchangeable substitute for echocardiographic cardiac output measurement.

Diagnostic performance in context

The USCOM AUC in our study (0.712), while statistically superior to manual pulse pressure, represents moderate rather than excellent diagnostic accuracy. This performance falls below AUC values reported for PLR interpreted with real-time cardiac output or stroke volume monitoring in prior meta-analyses [12–14]. Potential explanations include enrollment of an undifferentiated shock population, technical challenges of suprasternal USCOM in critically ill patients, use of an independent echocardiographic reference standard, and method-comparison limitations between USCOM and echocardiography.

Comparison with existing literature

Our findings confirm and extend prior evidence showing that PLR diagnostic performance depends critically on the monitoring modality used. Meta-analytic evidence consistently shows that PLR-induced changes in cardiac output or stroke volume outperform PLR-induced changes in arterial pulse pressure [12–14]. Lakhal et al.‘s multicenter study of 112 ICU patients with circulatory failure evaluated noninvasive oscillometric blood pressure changes during PLR, while Pickett et al. also examined noninvasive blood pressure monitoring during PLR [21, 22]. Monnet et al. further showed that arterial pressure changes were less useful than end-tidal carbon dioxide for predicting volume responsiveness during PLR [34].

Pathophysiological mechanisms

The simplified Windkessel relationship assumes relatively stable arterial compliance, an assumption often violated in shock [18, 19]. In our cohort, diastolic pressure increased proportionally more than systolic pressure, leaving pulse pressure essentially unchanged despite increased LVOT-VTI. This finding is consistent with prior physiological evidence showing that fluid-induced increases in cardiac output do not necessarily translate into parallel increases in arterial pressure because arterial load and vascular tone may change dynamically during shock resuscitation [35].

Clinical implications

Manual pulse pressure has one-directional utility: a positive response may support fluid responsiveness, but a negative response cannot reliably exclude it. In resource-limited settings where continuous flow monitoring is unavailable, clinicians should recognize that the high diagnostic accuracy reported for PLR applies primarily to PLR interpreted with real-time cardiac output or stroke volume changes, not manual pulse pressure alone. Recent expert consensus statements for sepsis care in resource-limited settings emphasize pragmatic clinical assessment and simple tests when advanced monitoring is unavailable [17].

Strengths and limitations

Strengths include sample size, independent echocardiographic reference standard, blinded independent measurements, consecutive enrollment, high protocol completion, and STARD-guided reporting. Limitations include the single-center design, demographic composition, operator dependence, and the practical interval between PLR and fluid challenge.

Non-responders received higher baseline norepinephrine doses than responders. This likely reflects greater vasodilatory shock severity and lower preload responsiveness in this group. We did not perform multivariable adjustment because the primary objective was diagnostic accuracy comparison between two index tests rather than development of an independent prediction model. With 124 patients and 58 events, extensive adjustment would also risk overfitting. Nevertheless, baseline vasopressor imbalance should be considered when interpreting generalizability.

Another important limitation concerns method comparison between USCOM and echocardiography. Although baseline Bland–Altman analysis showed low paired bias and no proportional bias, the percentage error exceeded the conventional 30% threshold used to support interchangeability [33]. Therefore, USCOM should not be interpreted as fully interchangeable with echocardiography for absolute cardiac output measurement.

Future research directions

Future research should include multicenter validation, randomized trials comparing flow-guided versus pressure-guided PLR strategies, and evaluation of simplified, low-cost flow monitoring technologies suitable for resource-limited settings.

Conclusions

In this single-center diagnostic accuracy study of 124 ICU patients with undifferentiated shock, manual pulse pressure monitoring demonstrated insufficient sensitivity and inadequate NPV for predicting fluid responsiveness during PLR. Continuous flow monitoring via USCOM-1 A provided significantly superior diagnostic accuracy compared with manual pulse pressure, although USCOM should be interpreted as a directional flow-monitoring index test rather than a fully interchangeable substitute for echocardiographic cardiac output measurement. The proportional pressure increase phenomenon provides a mechanistically coherent explanation for the poor diagnostic performance of manual pulse pressure monitoring. These findings require validation in multicenter cohorts across diverse settings before definitive changes to clinical practice guidelines can be recommended.

Abbreviations

APACHE II

Acute Physiology and Chronic Health Evaluation II

AUC

Area Under the Curve

BMI

Body Mass Index

CI

Confidence Interval

CO

Cardiac Output

DBP

Diastolic Blood Pressure

ICC

Intraclass Correlation Coefficient

ICU

Intensive Care Unit

IQR

Interquartile Range

LVOT

Left Ventricular Outflow Tract

MAP

Mean Arterial Pressure

NE

Norepinephrine

NPV

Negative Predictive Value

PLR

Passive Leg Raising

PP

Pulse Pressure

PPV

Positive Predictive Value

ROC

Receiver Operating Characteristic

SBP

Systolic Blood Pressure

SD

Standard Deviation

SOFA

Sequential Organ Failure Assessment

STARD

Standards for Reporting Diagnostic Accuracy Studies

SV

Stroke Volume

USCOM

Ultrasonic Cardiac Output Monitor

VTI

Velocity-Time Integral

ΔPP%

Percentage Change in Pulse Pressure

ΔSV%

Percentage Change in Stroke Volume

+LR

Positive Likelihood Ratio

−LR

Negative Likelihood Ratio

Authors’ contributions

Author contributionsFateme Monjazebi: Conceptualization, Methodology, Validation, Formal analysis, Writing-Original Draft, Writing – Review & Editing. Vahid Zamanzadeh: Project Administration, Methodology, Supervision, Writing – Review & Editing. Mansore Ashob: Writing-Original Draft, Data collection. Babak Jahangirifard: Data collection, USCOM, Echocardiography.

Funding

This study was conducted with financial support from the Research Vice-Chancellor of Shahid Beheshti University of Medical Sciences, Tehran, Iran. The funding source had no role in the study design; collection, analysis, or interpretation of the data; writing of the manuscript; or the decision to submit the manuscript for publication.

Data availability

De-identified individual participant data underlying reported results will be made available within 3–24 months of publication to researchers with methodologically sound proposals. Requests should be directed to the corresponding author. Requestors must sign a data access agreement. The study protocol and statistical analysis plan are also available upon reques.

Declarations

Ethics approval and consent to participate

This study was approved by the Research Council and the Research Ethics Committee of Shahid Beheshti University of Medical Sciences, Tehran, Iran (Ethics Code: IR.SBMU.PHARMACY.REC.1399.316). All procedures were conducted in full accordance with the Declaration of Helsinki (2013 revision) and relevant institutional guidelines and regulations. Written informed consent was obtained from all participants prior to enrollment. In cases where patients were unable to provide informed consent due to altered mental status, sedation, or mechanical ventilation, written informed consent was obtained from their legally authorized representatives, as approved by the ethics committee.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.