Abstract
The assessment of intravascular volume status remains a challenge for clinicians. Peripheral i.v. analysis (PIVA) is a method for analysing the peripheral venous waveform that has been used to monitor volume status. We present a proof-of-concept study for evaluating the efficacy of PIVA in detecting changes in fluid volume.
We enrolled 37 hospitalized patients undergoing haemodialysis (HD) as a controlled model for intravascular volume loss. Respiratory rate (F0) and pulse rate (F1) frequencies were measured. PIVA signal was obtained by fast Fourier analysis of the venous waveform followed by weighing the magnitude of the amplitude of the pulse rate frequency. PIVA was compared with peripheral venous pressure and standard monitoring of vital signs.
Regression analysis showed a linear correlation between volume loss and change in the PIVA signal (R2=0.77). Receiver operator curves demonstrated that the PIVA signal showed an area under the curve of 0.89 for detection of 20 ml kg−1 change in volume. There was no correlation between volume loss and peripheral venous pressure, blood pressure or pulse rate. PIVA-derived pulse rate and respiratory rate were consistent with similar numbers derived from the bio-impedance and electrical signals from the electrocardiogram.
PIVA is a minimally invasive, novel modality for detecting changes in fluid volume status, respiratory rate and pulse rate in spontaneously breathing patients with peripheral i.v. cannulas.
Editor’s key points
Peripheral i.v. analysis (PIVA) is a new method of monitoring body fluid volume by analysis of the peripheral venous waveform.
In this preliminary study, there was good correlation between fluid volumes removed during haemodialysis and volumes determined by PIVA but poor correlation with peripheral venous pressure, pulse or respiratory rate.
Fluid removal during haemodialysis does not represent blood loss and further data are required to whether PIVA correlates with existing measures of blood volume status or cardiac output.
Unlike some other modalities, PIVA appears useful in patients breathing spontaneously.
More recently, peripheral i.v. analysis (PIVA), a method for analysing the peripheral venous waveform, instead of venous pressure, has been effectively utilized in monitoring overall fluid volume status.10,11 PIVA has been shown to detect as little as 6% blood volume loss in human model of haemorrhage model.11 Further, PIVA has been shown to reflect blood volume changes more sensitively than standard vital sign monitoring during haemorrhage, resuscitation and iatrogenic volume overload in a porcine species.10
In this study, using haemodialysis (HD) as a model for controlled, quantifiable fluid loss, we compared PIVA, PVP and standard vital signs during HD. We also aimed to demonstrate proof-of-concept for PIVA as volume monitoring in spontaneously breathing individuals.
Methods
The study was performed in accordance with the Vanderbilt Medical Center Institutional Review Board. Thirty-seven adult inpatients receiving intermittent HD (Fresenius 2008 Series Dialysis Machines, Waltham, MA, USA) three times a week were enrolled and informed consent obtained. The nephrologist as per clinical assessment determined the dialysate content, ultrafiltration rate and goal for volume removal.
Patients with decompensated moderate–severe congestive heart failure, patients presenting with hypotension [mean arterial pressure (MAP) <55mmHg] and patients receiving mechanical ventilation were excluded from the study. All patients had a peripheral i.v. cannula (Smiths Medical, Mundelein, IL, USA) ranging from 22G to 18G placed in the upper extremity before HD, per routine. All cannulae were in the arm opposite an arteriovenous fistula or central venous dialysis catheter. IV cannulae were dedicated to monitoring via a pressure transducer (Edwards Lifesciences, Irvine, CA, USA) connected to a portable data acquisition device. Each cannula was flushed hourly and the site inspected to ensure proper position and patency.
Peripheral venous waveform data were obtained and recorded continuously 15 min before, throughout and up to 15 min after HD. Times were noted when the dialysis machine was paused or turned off. The PIVA signal was only analysed during periods when the HD pump was paused (between 5 and 15 min) or turned off. Standard vital signs including pulse rate, non-invasive cuff arterial blood pressure, respiratory rate and oxygen saturation were measured using an IntelliVue bedside monitoring system (Philips North America Corp, Andover, MA, USA). Blood pressure measurements were obtained every 15 min per standard HD protocol.
PIVA and PVP data were sampled at a rate of 500 Hz and downloaded and analysed via LabChart 7 (ADInstruments, Colorado Springs, CO, USA). Fast Fourier transformation of the peripheral venous signal was measured at baseline and throughout HD. The magnitude of the amplitude of the frequencies F0, corresponding to respiration rate and F1, corresponding to pulse rate, was calculated. Amplitude magnitudes were averaged over 16 s to create 8 K windows with window overlapping of 50% with Hann (cosine-bell).
Statistical analysis was performed with JMP Pro 11 (Cary, NC, USA) and GraphPad Prism (La Jolla, CA, USA). To create receiver operator curves (ROCs), data were inputted into a nominal logistic regression model.
Results
Clinical demographics
Thirty-seven patients were enrolled. Three patients were excluded for peripheral i.v. infiltration during the HD study. IV infiltration was identified by loss of venous waveform signal. In all three subjects, there was visual evidence of i.v. infiltration when flushed with <5 ml of normal saline during the hourly i.v. assessment. Five additional patients were excluded from the PIVA signal linear regression analysis because of intradialytic hypotension, defined by a MAP <55 mm Hg.
All subjects were patients admitted to the hospital undergoing HD (Table 1). The mean age was 62 yr (sd=12 yr), with 24 (65%) men and 13 (35%) females. Average BMI was 30 (sd=7). Significant medical conditions included: 20 patients with a diagnosis of diabetes (54%), 35 patients with a diagnosis of hypertension (95%), 10 patients with a diagnosis of left heart failure (27%) and six patients with a diagnosis of right heart failure (16%).
Patient characteristics for all 37 patients, presented as mean (standard deviation) or n (%). ACE/ARB, angiotensin-converting-enzyme inhibitor/angiotensin II receptor blockade; CCB, calcium channel blockade; EF, ejection fraction (obtain from transthoracic echocardiographic examination)
| Characteristic | n=37 |
|---|---|
| Enrolled | 37 |
| IV infiltration | 3 |
| Hypotension | 5 |
| Analysed | 29 |
| Age (yr) | 62 (12) |
| Sex (%) | |
| Male | 24 (65) |
| Female | 13 (35) |
| Height (cm) | 175 (10) |
| Weight (kg) | 93 (24) |
| BMI (kg m−2) | 30 (7) |
| ASA classification | |
| I | 0 |
| II | 0 |
| III | 21 |
| IV | 16 |
| Medical conditions (%) | |
| Diabetes | 20 (54) |
| Hypertension | 35 (95) |
| ACE/ARB | 14 (38) |
| Beta-blockade | 22 (59) |
| CCB | 12 (32) |
| Diuretic | 9 (24) |
| Left heart failure | 10 (27) |
| EF ≥50% | 4 (36) |
| EF 30–49% | 5 (45) |
| EF ≤29% | 1 (9) |
| Right heart failure | 6 (16) |
| Characteristic | n=37 |
|---|---|
| Enrolled | 37 |
| IV infiltration | 3 |
| Hypotension | 5 |
| Analysed | 29 |
| Age (yr) | 62 (12) |
| Sex (%) | |
| Male | 24 (65) |
| Female | 13 (35) |
| Height (cm) | 175 (10) |
| Weight (kg) | 93 (24) |
| BMI (kg m−2) | 30 (7) |
| ASA classification | |
| I | 0 |
| II | 0 |
| III | 21 |
| IV | 16 |
| Medical conditions (%) | |
| Diabetes | 20 (54) |
| Hypertension | 35 (95) |
| ACE/ARB | 14 (38) |
| Beta-blockade | 22 (59) |
| CCB | 12 (32) |
| Diuretic | 9 (24) |
| Left heart failure | 10 (27) |
| EF ≥50% | 4 (36) |
| EF 30–49% | 5 (45) |
| EF ≤29% | 1 (9) |
| Right heart failure | 6 (16) |
Patient characteristics for all 37 patients, presented as mean (standard deviation) or n (%). ACE/ARB, angiotensin-converting-enzyme inhibitor/angiotensin II receptor blockade; CCB, calcium channel blockade; EF, ejection fraction (obtain from transthoracic echocardiographic examination)
| Characteristic | n=37 |
|---|---|
| Enrolled | 37 |
| IV infiltration | 3 |
| Hypotension | 5 |
| Analysed | 29 |
| Age (yr) | 62 (12) |
| Sex (%) | |
| Male | 24 (65) |
| Female | 13 (35) |
| Height (cm) | 175 (10) |
| Weight (kg) | 93 (24) |
| BMI (kg m−2) | 30 (7) |
| ASA classification | |
| I | 0 |
| II | 0 |
| III | 21 |
| IV | 16 |
| Medical conditions (%) | |
| Diabetes | 20 (54) |
| Hypertension | 35 (95) |
| ACE/ARB | 14 (38) |
| Beta-blockade | 22 (59) |
| CCB | 12 (32) |
| Diuretic | 9 (24) |
| Left heart failure | 10 (27) |
| EF ≥50% | 4 (36) |
| EF 30–49% | 5 (45) |
| EF ≤29% | 1 (9) |
| Right heart failure | 6 (16) |
| Characteristic | n=37 |
|---|---|
| Enrolled | 37 |
| IV infiltration | 3 |
| Hypotension | 5 |
| Analysed | 29 |
| Age (yr) | 62 (12) |
| Sex (%) | |
| Male | 24 (65) |
| Female | 13 (35) |
| Height (cm) | 175 (10) |
| Weight (kg) | 93 (24) |
| BMI (kg m−2) | 30 (7) |
| ASA classification | |
| I | 0 |
| II | 0 |
| III | 21 |
| IV | 16 |
| Medical conditions (%) | |
| Diabetes | 20 (54) |
| Hypertension | 35 (95) |
| ACE/ARB | 14 (38) |
| Beta-blockade | 22 (59) |
| CCB | 12 (32) |
| Diuretic | 9 (24) |
| Left heart failure | 10 (27) |
| EF ≥50% | 4 (36) |
| EF 30–49% | 5 (45) |
| EF ≤29% | 1 (9) |
| Right heart failure | 6 (16) |
Peripheral venous waveform signal analysis
The peripheral venous waveform was obtained when the HD pump was paused or discontinued, as the dialysis pump resulted in dampening of the peripheral venous waveform signal (Fig. 1A and C). Fast Fourier transform analysis revealed peaks at frequencies corresponding to respiratory rate (F0) and pulse rate (F1) (Fig. 1B and D). The F1 amplitude magnitude was noted to significantly decrease as volume was removed during HD (Fig. 1B and D). There was excellent correlation between the F0 frequency and the monitored respiratory rate (Fig. 2). Similarly, the F1 frequency correlated with pulse rate obtained by a 5-lead electrocardiogram (Fig. 2).
Peripheral i.v. analysis (PIVA) before and after dialysis. Representative peripheral venous waveform obtained from a peripheral i.v. cannula before (A) and after (C) haemodialysis. Fast Fourier transform (FFT) of the corresponding peripheral venous waveform before (B) and after (C) haemodialysis. The frequencies (Hz) F0 and F1 correspond to the respiratory rate and heart rate, respectively. The amplitude of the F1 peak decreases from before (B) to after (D) dialysis, reflecting volume removal.
Peripheral i.v. analysis (PIVA) correlation with respiration rate and heart rate. The correlation between the low frequency (A, n=37) F0 signal with the monitored respiratory rate (thoracic impedence measurements) and the higher frequency (B, n=37) F1 signal with the monitored heart rate (electrocardiogram) as compared with the signal obtained with the IntelliVue bedside monitoring system (Philips North America Corp., Andover, MA, USA).
Peripheral venous waveform signal and volume
Comparisons were made between the percentage change in PIVA signal and PVP with the amount of volume removed via ultrafiltration. Regression analysis revealed a linear correlation between the amount of volume removed and the percentage change in PIVA signal (R2=0.77) (Fig. 3A). However, there was no significant correlation between the PVP, MAP or pulse rate with volume removed during ultrafiltration (Fig. 3B–D, respectively). ROCs demonstrated that the PIVA signal showed an area under the curve (AUC) for detection of 20 ml kg–1 of 0.890, whereas the AUC for PVP was 0.482 (Fig. 4).
Correlation between dialysate volume removed ml kg-1 and change in PIVA signal. This regression analysis represents the linear correlation between the amount of dialysate removed and the percentage change in PIVA signal (R2=0.77) (A; n=31). There was no significant correlation between the amount of intravascular volume removed and change in mPVP (R2=0.03) (B; n=34), MAP (R2=0.013) (C; n=34) and heart rate (R2=0.0003) (D; n=35). PIVA, peripheral i.v. analysis; MAP, mean arterial pressure; mPVP, mean peripheral venous pressure.
Sensitivity and specificity of PIVA vs PVP for volume of dialysate removed. Receiver operator curve (ROC) of PIVA signal (AUC=0.890) and PVP (AUC=0.482) in patients with a 20 ml kg–1 volume loss during dialysis. PIVA, peripheral i.v. analysis; PVP, peripheral venous pressure; AUC, area under the curve.
Eight patients were excluded from analysis. Five patients developed intradialytic hypotension (systolic blood pressure <90 mm Hg, MAP <50 mm Hg). Of note, in these five patients, there were changes in the PIVA signal before hypotension, including low F1 amplitude magnitudes. Three patients were excluded for dislodgement or infiltration of the i.v. cannula during the study. Of note, there was a complete loss of the PIVA signal in these patients during the study (data not shown).
Discussion
Using a model of patients undergoing HD, volume status was assessed with PIVA. We found that: (a) PIVA changes linearly with volume loss in spontaneously breathing patients undergoing dialysis; (b) PVP, a surrogate for CVP, failed to correlate with volume loss; and (c) PIVA accurately and continuously measures respiratory rate and pulse rate through a standard peripheral i.v. cannula. PIVA, using venous waveform characteristics, may offer a minimally invasive approach for continuous volume assessment in spontaneously breathing patients via a standard peripheral i.v. cannula.
There is no gold standard for measuring fluid status in patients undergoing dialysis.1 Current clinical, laboratory and diagnostic modalities used to assess volume are limited.1 Clinical assessment such as arterial pressure, pulse rate, body weight, oedema, jugular venous pressure and lung auscultation used to aid in the assessment of volume status have limitations.12–14 Imaging techniques such as inferior vena cava diameter with ultrasound requires technical expertise and is impractical for continuous volume assessment.15,16 Dynamic volume responsiveness monitors, such as pulse pressure variation, rely on intrathoracic changes induced by positive pressure ventilation and thus have limited utility in spontaneously breathing individuals and in patients requiring lung protective mechanical ventilation.2,16,17
The wave intensity magnitude is lowest in veins; hence, venous waveform analysis was not rigorously examined until appropriate sensing and amplifying technologies became available.18 Venous waves are generated by the cardiac cycle and propagation of the wave (and harmonics) is modulated by venous compliance (C = ΔV/ΔP, where V=volume and P=pressure).11 The venous system is a low-pressure, highly compliant system that can accommodate large changes in volume with only minimal changes in pressure.19 Compensatory vasoconstriction of the venous system to volume depletion diverts blood from the periphery to the central vasculature in order to maintain cardiac output and ultimately end-organ perfusion.19 This may explain the insensitivity of arterial blood pressure, pulse rate and PVP to volume depletion. However, the amplitude of the venous waveform and harmonics obtained through fast Fourier transformation are highly sensitive to changes in venous compliance because of changes in volume.11,16 Mild hypovolaemia leads to modulation of the cardiac frequency (F1) of the peripheral venous waveform.16 The proposed mechanism involves a decrease in venous blood volume so that the venous waveform is no longer transmitted back from the right atrium.16
There were several interesting secondary findings throughout this study. Five of the 37 patients had episodes of intradialytic hypotension, defined by a systolic blood pressure <90 mm Hg or MAP <55 mm Hg. A unique feature in all five patients was a notable decrease in the PIVA signal amplitude before the decrease in non-invasive blood pressure that was obtained every 15 min. Intradialytic hypotension occurs in approximately 6–7% of HD sessions.20 During HD, the ultrafiltration volume sometimes exceeds the entire plasma volume. However, the total plasma volume may only decrease by 20% because of mobilization of interstitial fluid into the intravascular compartment during ultrafiltration. Patients with limited interstitial fluid may not be able to adequately refill the intravascular compartment, resulting in hypotension.21 As it is difficult to measure the interstitial fluid, closed-loop systems with real-time ultrafiltration rate adjustments have been used based on measured blood volume. This technique has been shown to be effective for reducing intradialytic hypotension.22,23 More data are needed to understand whether PIVA can predict hypotension in patients undergoing HD or be useful as a closed-loop system for adjusting the rate and volume of ultrafiltration.
Three patients had dislodgement of their peripheral i.v. access during their HD session. In these three subjects, there was a complete loss of the venous waveform signal. While the timing of the signal loss to the i.v. malpositioning event was not known, it is possible that PIVA may be used to identify i.v. cannula integrity.
In this study, we used HD as a model for volume loss. PIVA has the potential to monitor volume status in other clinical conditions such as perioperative goal-directed fluid therapy, congestive heart failure optimization, acute kidney injury and trauma. While none of the subjects in our study received anaesthesia and were spontaneously breathing, the utility of PIVA in the presence of anaesthesia has been previously evaluated in subjects undergoing cardiac surgery with cardio-pulmonary bypass. It was found that the PIVA signal amplitude does decrease slightly with the induction of general anaesthesia. However, compared with the effects of anaesthesia, haemorrhage caused a much greater reduction in the F1 PIVA amplitude.24
The clinical model of volume removed during dialysis represents a heterogeneous group of patients with a broad range of co-morbidities (Table 1). The diverse patient demographics in this study with differing co-morbidities suggest that PIVA is applicable to a wide range of hospitalized patients. However, it is also not known what the effect of acute treatment with vasoactive drugs (such as vasodilators) that may impact venous caliber or viscoelasticity would have on the PIVA signal. These questions require future study.
In this study, we compared pulse rate and respiratory rate obtained with PIVA vs standard electrocardiogram and pulse oximeter technology. There was an exceptional correlation between the modalities. PIVA may provide an option for continuous pulse rate and respiratory rate evaluation in patients outside of the intensive care unit who may be intermittently monitored.
There were limitations to this study. This was a prospective observational study in a limited cohort of patients undergoing HD to determine if PIVA signal correlated with changes in volume status, rather than absolute values of volume status in these patients. While the changes in the PIVA signal correlated well with changes in volume because of dialysis, the ability of PIVA to correlate with static measurements of volume status such as pulmonary artery pressures are not known. All patients presenting for HD had unknown and likely variable baseline volume status. Regardless of this limitation, there remained a strong correlation between volume removal and change in PIVA signal.
The requirement of a peripheral i.v. limits the use of PIVA to patients with peripheral i.v. cannula. PIVA signal could only be analysed when the HD machine was inactive likely a result of noise and signal dampening created by the HD pump. Central venous pressures were not obtained because of the absence of central venous catheterization or use of these catheters for HD. However, peripheral venous pressure measurements served as the surrogate for CVP. Pre- and post-weights were not obtained because of immobility and inaccurate bed measurements in this inpatient population. However, absolute volume removed during ultrafiltration was measured.
Conclusion
Peripheral venous analysis (PIVA) represents a novel, minimally invasive modality for volume assessment in patients with peripheral venous cannulas. PIVA detected as little as 20 ml kg–1 of ultrafiltration. An important point is that this does not represent blood loss. This represented the volume of pre-dialysis patients that, typically, present in a volume overloaded state. We have demonstrated that during fluid removal, there is no significant change in PVP, a surrogate for CVP. Importantly, volume changes were detected in spontaneously breathing individuals, possibly offering a significant advantage over dynamic monitoring modalities. In addition to volume status, PIVA accurately assesses pulse rate and respiratory rate, potentially serving as a novel continuous monitor in hospitalized patients via a standard peripheral i.v. cannula.
Future studies are warranted to correlate PIVA to cardiac output and to determine whether PIVA predicts fluid responsiveness. This was an observational study, thus PIVA was not used to guide therapeutic decisions.
Authors’ contributions
Study design, data analysis, product development and manuscript preparation (all drafts): K.M.H.
Study design, patient recruitment and manuscript preparation (all drafts): B.D.A.
Study design, data analysis, product development and manuscript preparation (all drafts): F.B. and C.M.B.
Product development and manuscript preparation (final draft): R.B. and I.B.
Study design, data analysis, product development, patient recruitment and manuscript preparation (all drafts): S.E.
Declaration of interests
I.B. is an employee (Medical Affairs department) and shareholder of Baxter Healthcare Corporation. K.H., R.B., F.B., C.B., and S.E. all have financial interests in the form of grants and contracts from Baxter Healthcare Corporation.
Funding
This work was supported though funding from Baxter International Inc.
Comments