Ultrasound-measured brachial artery reactive hyperemia in critically ill patients: an observational study

Abstract

Background

Ultrasound-measured brachial artery reactive hyperemia (RH) is independently predictive of hospital mortality in critically ill patients with sepsis. Its association with mortality is uncertain in critically ill patients in general.

Methods

This was a combined case-control and prospective cohort study. Ultrasound was used to measure brachial artery reactive hyperemia in 150 critically ill patients at a single academic medical center and in 44 control subjects without acute illness. Measurements were compared in cases versus controls, septic vs. non-septic critically ill patients, and hospital survivors vs. non-survivors. Follow-up measurements were obtained 3–5 days later in a sub-sample of patients.

Results

RH was calculated as the percent change in pre- vs. post-ischemic brachial artery velocity-time integral measured by Doppler ultrasound. RH was impaired in critically ill compared to control subjects (194 [179–210] vs. 369 [314–433]%, p < 0.001; results expressed as mean [95% confidence interval]) but similar in septic compared to non-septic patients (196 [177–217] vs. 199 [170–233], p = 0.88). RH was significantly lower in hospital non-survivors compared to survivors (144 [120–173] vs. 204 [187–222], p = 0.003). Multivariable analysis showed that the difference between survivors and non-survivors was not confounded by age or comorbidities (odds ratio for hospital death = 0.26 per log unit rise in RH, 95% confidence interval = 0.08–0.83, p = 0.02). The magnitude of RH improved over 3–5 days in hospital survivors (n = 63, 204 [180–232] vs. 239 [208–275], p = 0.02), but did not change in non-survivors (n = 11, 133 [107–165] vs. 128 [75–220].

Conclusions

Reactive hyperemia of the brachial artery is impaired in undifferentiated critically ill patients, lower in non-survivors compared to survivors, and independently associated with hospital mortality. Brachial artery reactive hyperemia improves significantly over time in survivors but not in non-survivors.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13054-025-05646-7.

Key messages

• Brachial artery reactive hyperemia is impaired in critically ill patients.

• The association between reactive hyperemia and mortality is independent of age and comorbidities.

• The relationship between RH and mortality is not mediated through an intermediate effect of vasopressor infusions.

• Brachial artery reactive hyperemia improves over time in hospital survivors but does not change in hospital non-survivors.

• Brachial artery reactive hyperemia is a practical method for measuring microvascular dysfunction in critical illness.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13054-025-05646-7.

Background

Measurements of limb reactive hyperemia after stagnant ischemia provide non-invasive assessment of microvascular function [1, 2]. Reactive hyperemia predicts cardiovascular outcomes in healthy individuals [3] and other patient populations [4–7]. We previously found that ultrasound measurements of brachial artery reactive hyperemia (hereafter referred to as RH) predict sepsis mortality [8]. It is unknown whether RH also predicts mortality in non-septic critically ill patients, and whether follow-up measurements offer additional prognostic insight.

The purpose of this study was to determine whether RH differs between patients with and without a clinical diagnosis of sepsis, to determine whether RH is associated with mortality in a general critically ill population including non-septic patients, and to assess the usefulness of follow-up imaging. We hypothesized that RH would be similar in septic and non-septic critically ill patients, associated with mortality in this more general population, and that RH would improve over time in survivors but would be persistently diminished in non-survivors.

Methods

Study design and participants

The current study is a product of a larger, ongoing, long-term single-center prospective cohort and case-control study. The study design and the clinical characteristics and outcomes of most of the study subjects (138 ICU patients and 43 control subjects) have been reported previously [9]. This paper reports brachial artery ultrasound measurements, performed beginning in 2013 and ending with the onset of the coronavirus pandemic in February 2020.

Consecutive patients admitted to the medical or surgical intensive care units (ICUs) at the University of Rochester Medical Center were eligible for enrollment if they had at least one acute organ dysfunction (Table E1, these patients, as expected for most newly admitted ICU patients, also had two or more systemic inflammatory response syndrome (SIRS) criteria [10]). A healthy control group was also recruited. Exclusion criteria are listed in Table E2. Enrolled patients were followed for the duration of the hospital stay.

The presence of infection was determined by consensus of at least two experienced clinicians during detailed retrospective chart review including all clinical, imaging, and microbiological data. Enrolled patients with infection had sepsis according to Sepsis-2 criteria (infection + two or more SIRS criteria + acute organ dysfunction) [10]. Shock of any type was defined as the requirement for vasopressor agents or shock index > 1.0 (shock index = heart rate divided by systolic blood pressure). Infected patients meeting these shock criteria were classified as septic shock. Alternative diagnoses were specified during this retrospective chart review for the patients who were deemed non-septic and are listed in Table E6.

Informed consent was obtained from subjects or surrogate decision-makers. This study is approved by the University of Rochester Research Subjects Review Board (“The Severe Systemic Inflammatory Syndrome Cohort [SSIC] Study”, approval number 00204, initial approval date 03/17/2010). All research procedures were in accordance with the ethical standards of the University of Rochester Research Subjects Review Board and with the amended Helsinki Declaration of 1975. Compliance with STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines for reporting observational studies is shown in Table E3.

Measurements

Reactive hyperemia was measured by clinical research staff. The initial staff member was trained by a registered sonographer by performing RH on healthy control subjects under direct supervision until, in the opinion of the sonographer, this staff member was proficient. This initial staff member then trained other research staff members.

Initial RH was measured as soon as possible after consent was obtained, within 72 h of meeting study inclusion criteria. Follow-up RH was measured 3–5 days later. RH was measured using the previously reported technique [8]. In brief, measurements occurred in hospital rooms (cases) or an outpatient facility (controls). The supine subject’s arm was extended and the brachial artery was imaged just proximal to the antecubital fossa. The pulse-wave Doppler gate was positioned at a 60-degree angle within the center of the arterial lumen. Images were acquired using a General Electric Vivid 7 ultrasound machine with a M12L linear-array transducer (GE Medical Systems, Milwaukee, WI, USA) or a Mindray M9 diagnostic ultrasound system with a L12-4s linear array transducer (Shenzhen Mindray bio-medical electronics Co., Ltd., Shenzhen, China).

A deflated sphygmomanometer cuff was placed around the forearm 1–2 cm distal to the antecubital fossa. The brachial artery above the antecubital fossa was imaged and resting pulse-wave spectral Doppler recordings were obtained prior to cuff inflation. The cuff was rapidly inflated to 200 mmHg for 5 min, then rapidly and completely deflated. Post-ischemic pulse-wave spectral Doppler 2-D recordings were immediately acquired consecutively for the first 15 s after cuff release. The baseline and post-occlusion images were digitally stored for later analysis.

Doppler tracings were analyzed with ImageJ 1.52d software (National Institutes of Health, Bethesda, MD, USA http://imagej.nih.gov/ij) to measure velocity-time integral (VTI) over a single cardiac cycle (units = centimeters/cardiac cycle). Investigators visually defined the borders of each waveform allowing the software to calculate area under the curve (AUC). The average of the three maximal VTIs before cuff inflation defined baseline velocity, and the average of the three maximal VTIs post-deflation defined hyperemic velocity. Brachial artery reactive hyperemia was defined as the percent change in hyperemic velocity relative to baseline velocity, thereby indexing the measurement to potential differences in the baseline brachial artery blood flow or cardiac cycle time between comparison groups. Subjects with missing initial post-cuff deflation pulse-waves or inadequate Doppler signals were excluded.

Statistical analysis

The primary independent variable was RH measured within 72 h of meeting inclusion criteria. The distribution of RH was right-skewed so natural log transformation was applied to approximate normality. The primary outcome variables were presence of critical illness (critically ill patients vs. healthy controls), sepsis (patients with sepsis vs. non-septic critically ill patients), and hospital mortality (hospital survivors vs. non-survivors). Other secondary dependent variables included lactate, vasopressor days from day 0–28, ICU-free days from day 0–28, and ventilator-free days from day 0–28. Transformed variables are expressed as mean ± 95% confidence intervals (CI) after back-transformation to the original scale. Other continuous variables are expressed as mean ± standard deviation (SD) or median (inter-quartile range [IQR]), as appropriate. Categorical variables are expressed as number (percentage).

Student’s t-test and the Wilcoxon rank-sum test were used to compare continuous variables between groups as appropriate. The χ²-square test was used to compare categorical variables. Pearson’s or Spearman’s correlation coefficients were used as appropriate to assess relationships between continuous variables. Analysis of variance was used to assess differences in RH between cases and controls while controlling for differences between the groups in age, sex, and heart rate.

Logistic regression was used to assess associations between RH and hospital mortality. We used clinical judgment and previous reports [8] to select age and Charlson comorbidity index [11] as potential confounding variables (see Figure E1 for hypothesized causal diagram). The relationship between RH and the dependent variable is reported as the odds ratio (OR) of the outcome per one unit increase in the natural log-transformed RH, with the corresponding 95% CI. RH was treated as a continuous variable in the logistic regression model to maximize statistical power.

We next evaluated whether there was a direct relationship between low RH and hospital mortality, independent of mediation whereby low RH is indirectly associated with vasopressor requirement, which is in turn associated with higher hospital mortality (Figure E1). We hypothesized that RH had a direct association with mortality, without the requirement for mediation by vasopressors. We used Stata’s mediate function for calculation of the total effect, natural direct effect, and natural indirect effect of the association between low RH and hospital mortality, with or without mediation by vasopressor administration [12]. We decomposed these effects under the conceptual framework of a presumed direct association between RH and mortality, while evaluating the possibility of an indirect mediating effect of vasopressors on the association between RH and hospital mortality [13]. For this analysis, RH was dichotomized according to the median value.

Formal power calculations were not performed for the determination of sample size. Missing values for ancillary clinical variables and laboratory measurements were not imputed. Instead, patients with data missing for a particular analysis were excluded from that analysis. A p value of < 0.05 was accepted as statistically significant. Statistical analyses were performed using Stata Statistical Software: Release 18 (StataCorp LP, College Station, TX:).

Results

From July 2013 through February 2020, 272 critically ill patients were enrolled (Fig. 1). Of these, initial RH was attempted in 200 patients and successfully performed in 150 patients. The most common reasons RH was not performed were unavailability of ultrasound devices or trained research staff, and the most common reason for unsuccessful measurements was insufficient cuff inflation times and unsuccessful image capture immediately after cuff deflation (Fig. 1). Brachial artery reactive hyperemia was initially measured 37 (29–44) hours after inclusion criteria were met.

Fig. 1.

Patient screening and enrollment flow diagram. Control subjects were screened for inflammation or vascular insufficiency of the target extremity or hemolytic disorders (these exclusion criteria were absent in all screened subjects). *These exclusions were eliminated in July 2015. ICU = intensive care unit; SIRS = systemic inflammatory response syndrome; RH = ultrasound-measured brachial artery reactive hyperemia (percent change); RBC = red blood cell; OR = operating room

Characteristics of the enrolled patients are shown in Table 1. The sources of sepsis are shown in Table E4, and categorized reasons for ICU admission in non-septic patients are shown in Table E5. The specific diagnoses of the non-septic patients are shown in Table E6. Study clinicians were unable to confidently determine the presence of sepsis in 11 patients. Hence, these unclassifiable subjects were excluded from the corresponding descriptive statistics in Table 1 and the sepsis analysis described below. Forty-four healthy control subjects were enrolled. Lactate concentration was not measured in 34 ICU patients and six control subjects.

Table 1.

Baseline characteristics of critically ill patients^a values are median (interquartile range) for continuous variables or number (percent) for categorical variables

	Control subjects, n = 44	Critically ill patients, n = 150	p	Presence of sepsisb, c		p	Hospital mortality		p
				Yes, n = 95	No, n = 44		Died, n = 21	Survived, n = 129
Age	58 (52–65)	64 (52–71)	0.04	65 (50–73)	61 (56–68)	0.005	71 (68–76)	61 (50–69)	< 0.001
female sex	24 (54)	59 (39)	0.07	37 (39)	18 (41)	0.83	9 (43)	50 (39)	0.72
Race			0.08			0.23			0.48
Caucasian	39 (91)	121 (81)		81 (85)	33 (75)		17 (81)	104 (81)
African- American	1 (2)	22 (15)		11 (12)	7 (16)		4 (19)	18 (14)
Unknown	3 (7)	7 (5)		3 (3)	4 (9)		0	7 (5)
Ethnicity			0.34			0.15			0.96
Hispanic	0 (0)	6 (4)		3 (3)	3 (7)		1 (5)	5 (4)
Non- Hispanic	43 (98)	138 (92)		86 (91)	41 (93)		19 (90)	119 (92)
Unknown	1 (2)	6 (4)		6 (6)	0		1 (5)	5 (4)
MAP at time of RH (mm Hg)	---	80 (74–89)		77 (71–88)	86 (76–92)	0.01	71 (66–81)	82 (74–90)	< 0.001
Vasopressor use at time of RH	---	32 (21)		28 (29)	3 (7)	0.003	11 (52)	21 (16)	< 0.001
Heart rate at time of RH (bpm)	67 (62–78)	88 (77–104)	< 0.001	88 (76–104)	90 (78–104)	0.89	84 (72–92)	88 (78–104)	0.30
Shock present at time of RHd	---	45 (30)		36 (38)	7 (16)	0.01	11 (52)	34 (26)	0.02
Charlson comorbidity index	---	3 (1–5)		3 (2–5)	3 (1–4)	0.37	5 (3–7)	3 (1–4)	< 0.001
APACHE II score	---	20 (16–25)		22 (17–29)	17 (12–22)	< 0.001	28 (23–30)	19 (15–24)	< 0.001
SOFA score	---	5 (3–8)		6 (4–9)	4 (2–6)	< 0.001	11 (8–14)	5 (2–8)	< 0.001
ICU length of staye	---	4 (2–7)		4 (2–8)	3 (2–6)	0.02	7 (6–12)	3 (2–7)	< 0.001
Hospital length of stay	---	9 (6–18)		10 (7–22)	8 (5–13)	0.02	9 (7–22)	9 (6–16)	0.81

^avalues are median (interquartile range) or number (percent); ^bpresence or absence of sepsis could not be confidently determined in 11 patients; ^csources of infection for septic patients are shown in Supplemental Table E4 and the ICU admission categories for non-septic patients are shown in Supplemental Table E5; ^dshock was defined as Shock of any type was defined as the requirement for vasopressor agents or shock index greater than 1.0 (shock index = heart rate divided by systolic blood pressure); ^eICU length of stay missing in the two subjects never admitted to ICU. RH = ultrasound-measured brachial artery reactive hyperemia (percent change)

RH in critically ill patients vs healthy controls Healthy subjects were younger, more likely to be female, and had lower heart rates compared to critically ill subjects (Table 1). RH was lower in critically ill vs. healthy subjects (Table 2; Fig. 2). The difference in RH between critically ill patients and healthy controls remained significant after adjustment for differences in age, sex, and heart rate (Table 2).

Table 2.

Comparison of RH measurements and results: critically ill patients vs. control subjects

Variable		Critically ill (n = 150)		Healthy (n = 44)		p
Baseline brachial artery VTI		15.4 [12.3–19.8]		19.2 [14.8–27.4]		0.002
Hyperemic brachial artery VTI		49.4 [35.8–60.7]		99.9 [81.6–117.3]		< 0.001
RH
Unadjusted		194 [179–210]		369 [314–433]		< 0.001
Adjusteda		200 [184–217]		335 [284–395]		< 0.001

All values are mean [95% confidence interval] after transformation back to their original scale

^a Adjusted for age, sex, and heart rate

VTI = velocity-time integral: RH = ultrasound-measured brachial artery reactive hyperemia (percent change)

Fig. 2.

RH comparisons in healthy control vs. critically ill subjects. Natural log-transformed RH is plotted on the y-axes. Box plots show the median (horizontal line) and 25th and 75th percentiles (lower and upper limits of the box). The dots represent outliers beyond the whiskers designating the 10th and 90th percentiles. Comparisons were made with the Student’s t-test. RH = ultrasound-measured brachial artery reactive hyperemia (percent change)

Relationships between RH and sepsis RH was similar in septic and non-septic patients (196 [177–217], n = 95 vs. 199 [170–233], n = 44, p = 0.88).

Relationship between RH and hospital mortality RH was lower in hospital non-survivors compared to survivors (144 [120–173] vs. 204 [187–222], p = 0.003, Fig. 3). Brachial artery reactive hyperemia was independently associated with hospital mortality in multivariable logistic regression controlling for age and Charlson comorbidity score (OR = 0.26, 95% C.I. = 0.08–0.83, p = 0.02, Table 3). The predicted and actual hospital mortality by quartile of RH is shown in Table E7. The actual and modeled relationship between the full range of RH and hospital mortality is shown in Figure E2.

Fig. 3.

RH comparisons in hospital survivors vs. non-survivors. Natural log-transformed RH is plotted on the y-axes. Box plots show the median (horizontal line) and 25th and 75th percentiles (lower and upper limits of the box). The dots represent outliers beyond the whiskers designating the 10th and 90th percentiles. Comparisons were made with the Student’s t-test. RH = ultrasound-measured brachial artery reactive hyperemia (percent change)

Table 3.

Multivariable logistic regression model for hospital mortality

Explanatory variable		Odds ratio, hospital mortality		95% confidence interval		P value
RH a		0.26 b		0.08–0.83		0.02
Age		1.07 c		1.02–1.11		0.005
Charlson Comorbidity Index		1.32 d		1.08–1.61		0.006

^a natural log-transformed RH; ^binterpretation of odds ratio: there is a 74% lower risk of hospital mortality for every one natural log increase in RH; ^c per one year increase in age; ^d per one unit increase in Charlson Comorbidity Index

RH = ultrasound-measured brachial artery reactive hyperemia (percent change)

The association between RH and hospital mortality did not depend on the mediating effect of vasopressor use. In a scenario where all subjects have RH below the median value, expected hospital mortality is 17% higher than if none of the subjects have RH below the median (95% CI = 6–28%, p = 0.002). Most of this effect is attributable to the direct association between RH on mortality (14%, 95% CI = 4–24%, p = 0.008), absent any vasopressor use. Only the remaining 3% (95% CI = − 0.02–0.09) of the total effect is attributable to a mediating association between vasopressors on mortality (Table E8).

Relationship between RH and other dependent variables. There were weak but significant correlations of RH with lactate (Spearman’s rho = − 0.23, p = 0.01), number of days of vasopressor infusion (Spearman’s rho = − 0.20, p = 0.02), ICU-free days (Spearman’s rho = 0.29, p < 0.001), and ventilator-free days (Spearman’s rho = 0.28, p < 0.001).

Follow-up imaging. Follow-up RH was measured in 74 patients. The remaining 76 patients did not have follow-up imaging, most commonly because POCUS-trained study personnel were not available (Figure E3). Follow-up measurements were made 95 (72–120) hours after the initial measurement. Brachial artery reactive hyperemia improved significantly in survivors (n = 63, 204 [180–232] vs. 239 [208–275], p = 0.02), but not in non-survivors (n = 11, 133 [107–165] vs. 128 [75–220], p = 0.89, Fig. 4).

Fig. 4.

RH in survivors and non-survivors at Measurement 1 (M1) and Measurement 2 (M2). This figure excludes patients without paired M1 and M2 measurements. Box plots show the median (horizontal line), 25th and 75th percentiles (lower and upper limits of the box). The dots represent outliers beyond the whiskers that designate the 10th and 90th percentiles. * p = 0.008 for comparison of M1 between these survivors and non-survivors. ** p = 0.002 for comparison of M2 in these survivors vs. non-survivors. Paired comparisons between M1 and M2 made with paired t-test. Comparisons between survivors and non-survivors made with unpaired t-test. RH = ultrasound-measured brachial artery reactive hyperemia (percent change)

Discussion

Our study provides several important findings. First, our study indicates that RH is impaired in critically ill septic and non-septic patients. Second, we find that RH is significantly impaired in non-survivors compared to survivors. Third, we find that the association between lower RH and hospital mortality remains significant when controlling for the confounding effects of age and comorbidity. Fourth, we found that the association between RH and mortality was not mediated by concomitant vasopressor use. Fifth, remeasurement of RH 3–5 days after initial testing showed improvement in survivors but no change in non-survivors.

Physiological significance

Reactive hyperemia represents the integrated physiological responses of the microvasculature to a period of stagnant ischemia [14, 15]. The physiological mechanisms responsible for this effect include endothelium-dependent production of vasodilator prostaglandins and nitric oxide [16, 17], endothelium-independent activation of vascular smooth muscle ATP-sensitive potassium channels [18], a myogenic response [19], adenosine production [19], and rheologic abnormalities [20]. Each of these mechanisms is implicated in the pathophysiology of sepsis [21–26] and together these mechanisms generate the microvascular dysfunction driving organ dysfunction in sepsis [27] and other critical illnesses [28].

Comparison with previous studies

Reactive hyperemia is linked with inflammatory markers [29] and cardiovascular events [3–7], suggesting its potential prognostic utility in critically ill patients, epitomized by systemic inflammation. Prior studies have measured reactive hyperemia in septic patients using other methods including plethysmography [20, 30–32], transcutaneous laser Doppler measurements of erythrocyte velocity [33], reactive hyperemia peripheral arterial tonometry (RH-PAT) [34], and near-infrared spectroscopy plethysmography (NIRS) [35]. All showed significant relationships between reactive hyperemia and severity of illness or mortality. A 2019 meta-analysis of 18 studies found that reactive hyperemia (measured using various techniques) was lower in septic vs. non-septic subjects and in non-survivors compared to survivors [36].

We previously found that RH was an independent predictor of hospital mortality in septic patients [8]. Since then, two additional studies have shown associations between impaired brachial artery RH and mortality in sepsis [37, 38]. These studies and our current findings support the practical utility of brachial artery RH measurements.

Strengths and limitations

The methodological strengths of our study are the recruitment of a range of critically ill patients with and without sepsis and follow-up measurements performed in a subset of the study subjects.

Our study has several limitations. First, the observational study design precludes conclusions about causal relationships between impaired RH and adverse outcomes. It is possible that there are other confounding variables that were not considered in our analysis. Our limited sample size required a parsimonious approach to multivariable analysis. We considered age and comorbid conditions to be key potential confounding variables. We further theorized that low RH could be associated with a need for vasopressors, which in turn could be associated with mortality. Our causal mediation analysis evaluated this pathway and confirmed the direct association between RH and hospital mortality.

A second limitation of our study is that we did not routinely measure global hemodynamic variables. Therefore, we could not assess whether central filling pressures and cardiac output were linked with RH. Third, our non-septic patients all manifested signs of systemic inflammation (albeit not from a definable infectious origin). Therefore, we were unable to assess potential differences in RH between sepsis and non-inflammatory conditions. Fourth, repeat RH measurements were only possible in half of the study subjects. Finally, we were unable to assess inter-observer variability of ultrasound image acquisition. Nevertheless, from a pragmatic perspective, we found that RH independently predicts mortality, notwithstanding possible limitations in inter-observer precision.

Conclusions

In this prospective cohort study of critically ill patients, we find that RH is associated with hospital mortality. This association is not unique to septic patients, is not explained by age or chronic health conditions, and is not mediated by use of vasopressor infusions. The association between RH and adverse outcomes is reinforced by significant correlations with lactate concentrations, duration of vasopressor infusions, ICU-free days, ventilator-free days, and by temporal improvement in survivors alone. This study provides rational for use of RH as an enrollment criterion and surrogate outcome measure in pilot clinical trials aiming to improve microvascular function.

Author contributions

CS, AR, and AP had full access to the data and take responsibility for the integrity of the data. AP takes responsibility for the accuracy of the data analysis. CS, TB, JZ, OW, and AP were involved in the concept and design. All authors were involved in acquisition of data. CS, NK, TM, LI, HC, OW, MM, TB, CD, AF, JZ, and AP were involved in sepsis and ARDS designations. CS, TB, JZ, EL, AR, and AP were involved in analysis and interpretation of data. CS and AP were involved in initial draft of the article. CS, TB, JZ, and AP were involved in statistical analyses. All authors participated in critical revision and review of the manuscript.

Funding

AHA 13CRP17110114, NIH R01 HL 160723.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Informed consent was obtained from subjects or surrogate decision-makers. This study is approved by the University of Rochester Research Subjects Review Board (“The Severe Systemic Inflammatory Syndrome Cohort [SSIC] Study”, approval number 00204, initial approval date 03/17/2010).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.