Secondary Logo

Journal Logo


Measurement of Extravascular Lung Water in Hemodialysis Patients Using Blood Ultrasound Velocity and Optical Density Dilution

Garland, Jocelyn S.; Kianfar, Cynthia; Nesrallah, Gihad; Heidenheim, Paul; Lindsay, Robert M.

Author Information
  • Free


In hemodialysis patients, volume homeostasis is an important clinical problem. The aim is to have patients at an ideal “dry weight” postdialysis, but current methods for accurately measuring dry weight are disappointing. Dry weight is usually clinically determined in most centers and is often defined as the lowest weight a patient can tolerate without developing symptoms or hypotension. It is usually measured by simple trial and error and may be imprecise as reflected either by intradialytic hypotensive episodes on the one hand and volume overload, often with pulmonary edema and poor blood pressure (BP) control, on the other. 1,2 Numerous techniques have been attempted to estimate the optimal weight of hemodialysis patients, including biochemical markers (atrial natriuretic peptide, 3 cyclic guanidine monophosphate 4), vena cava diameter, 5 electrical bioimpedance, 6 etc, but each of these techniques has limitations. 7 We therefore decided to use a novel approach in attempting to measure extravascular lung water (EVLW), postulating that EVLW changes may mirror the clinical volume spectrum seen in these patients.

The ability to estimate a measurement of EVLW is not new, and numerous methods have been described in the literature. Techniques using dye or radiolabeled substances have not found clinical acceptance, however, because of cumbersome methods, inaccuracies of measurements, and the necessary requirement of invasive monitoring devices. Krivitski and coworkers 8 have developed a technique whereby EVLW may be calculated from an analysis of concurrent transients in blood ultrasonic velocity and, separately, electrical impedance during the passage of an injected bolus of hypertonic (5%) saline. The hypertonic saline bolus, having passed through the lung, has an ultrasound velocity transient that is biphasic if measured beyond the left ventricle (e.g., carotid artery in an experimental animal). This transient is compatible with water moving into the hypertonic bolus from the lung parenchyma, thereby leaving the lung parenchyma hypertonic. Subsequently, as the bolus leaves the lung vasculature, water passes from the blood into the tissue to return the lung tonicity to its baseline, giving a moment when net movement is zero, an instant of osmotic equilibrium. The concurrent measurements of electrical impedance track the sodium chloride transient. A theoretic basis for the calculation of EVLW is derived from the water transferred to the blood, the amount of sodium chloride moved from blood to the lung, and the increase in blood osmolality measured at the moment of equilibrium. It also requires a knowledge of the cardiac output (CO). Details of this theoretic basis are given elsewhere. 8

Krivitski and coworkers 8 have tested this method in separate studies in animals and achieved agreement between obtained EVLW values and those obtained gravimetrically following animal sacrifice.

The purpose of this study was (1) to determine if EVLW measurement by this method is feasible in hemodialysis patients and (2) to determine the range of EVLW values in this patient population.


Because a CO measurement is necessary to calculate EVLW, CO and EVLW were sequentially measured in 18 hemodialysis patients, 15 with forearm arteriovenous fistulae and 3 with Gore-Tex grafts, all known to be well functioning and with no access recirculation at a dialysis circuit blood flow (Qb) = 300 ml/min. These patients (10 men, 8 women; age 59 ± 8 years) were chosen on the basis of known hemodynamic stability during hemodialysis treatments without significant cardiac disease and judged not to be in pulmonary edema on clinical grounds at the time of study. Baseline BP as well as predialysis weight, ideal body weight, postdialysis weight, and BP were recorded. Values of EVLW and CO were obtained for each patient at the beginning (within the first 20 min) and at the end (within the last 45 min) of a single dialysis session. The EVLW values obtained were recorded as an absolute and also normalized to the patient’s weight at the time of measurement (ml/Kg). To do this, the predialysis weight was taken and adjusted for ultrafiltration and fluid balance to the measurement time.

CO measurements were performed using the Transonic Hemodialysis Monitor, HDO1 (Transonic Systems Inc, Ithaca, NY). This method uses the indicator (ultrasound velocity) dilution (normal saline) principle, is easily performed on hemodialysis patients during dialysis, gives reproducible results (coefficient of variation < 10%), and its use in our hemodialysis population has been described elsewhere. 9 The EVLW measurements were made using a Transonic Hemodialysis Monitor adapted by the addition of an Optical Density Monitor (Transonic Systems Inc, Ithaca, NY) with its probe attached to the patient’s arterial blood line. The optical probe emits light from a diode at 540 nm wave length and, after passing through the blood, is received by a photocell. The light on its passage through blood is partly absorbed and partly scattered. The amount of absorption and scattering determine the decreased amount of light arriving at the photocell; this is referred to as the optical density of blood. The injection of isotonic saline decreases the optical density of blood. The reverse occurs after hypertonic saline injection, which alters the osmotic behavior of erythrocytes, and this causes increased scattering of light. The Optical Density Monitor will record changes identical to those measured by electrical impedance as was originally employed by Krivitski 8 in his animal experiments. The use of the electrical impedance probe is invasive, being attached to a blood vessel wall and therefore not suitable for patient studies.

The velocity of sound through blood (1,560–1,585 m/sec) is dependent upon its density and is sensitive to the addition of isotonic saline, which will make it decrease. This is detected by ultrasound probes placed on the patient’s venous and arterial blood lines that record dilution curves after calibration and measure injections of isotonic saline. Calibration injections are needed to test the sound velocity properties of blood. They are made into the venous line and recorded by the probe on the venous line. Measurement injections of 0.9% saline permit the calculation of CO. Injections are made into the venous line and recorded by the probe on the arterial line. In contrast, the sound velocity of blood is scarcely changed following the injection of 5% saline (ultrasound velocity of 1,573 m/sec) and, thus, what is detected by the ultrasound probes on the venous line will show only a small (if any) dilution curve. However, a change in optical density has occurred after both the isotonic bolus (decrease) and by the hypertonic bolus (increase), and these changes are recorded by the optical probe placed on the patient’s arterial blood line. These separate probes on the arterial line can therefore track the concurrent change in optical density and ultrasound velocity following the injection of saline of two different concentrations. The hypertonic saline passing through the lung will create a flux of water from the lung tissue to the blood vessels. The amount of water withdrawn from the lung depends upon the amount of water in lung tissue and upon the amount the osmotic pressure has increased (after hypertonic saline) in the blood vessels at the time the bolus passes through the lung capillaries. The approach to calculating the amount of water in lung tissue (the EVLW) is based on the calculations of the water flux and the increment of the osmotic pressure. This is complicated by the fact that the reversed flux of sodium from the capillary blood to the lung tissue has to be accounted for as well as the water flux. In addition, it must be appreciated that this dilution technique can only calculate lung water from perfused areas. Therefore, the EVLW calculated is “functional,” not anatomical, lung water.

The experimental protocol first required calibration measurements. This entailed temporarily moving the optical probe from the arterial line to the venous line (only one probe was available). Then 10 ml of normal saline at 37° C was injected into the patient’s venous blood line, the ultrasound and optical density curves being recorded by the probes on the venous side. Next, 10 ml of 5% saline was injected into the patient’s venous blood line. The calibration curves were instantly obtained by ultrasound and optical density probes. The areas under these calibration curves are required to calculate a calibration coefficient (Ks0) for each EVLW measurement (see equation below). The optical density probe was rapidly switched to the arterial line and, after 45 to 60 seconds, 30 ml of normal saline (37° C) was injected into the venous line, and the respective ultrasound and optical dilution curves were obtained from the corresponding probes on the arterial line. This protocol was repeated once each time per patient at both beginning and end of dialysis. In each patient, the measurements were completed during the first 15 minutes and last 45 minutes of the hemodialysis session. Ultrasound velocity and optical density transients were recorded and the results applied to the Krivitski equation, 8 through which a measurement of EVLW was calculated for each data set. The average value of each pair of measurements was used in the data analysis.

Figure 1 (upper portion) gives an example of the ultrasound dilution curve obtained following a measurement injection of 30 ml of 0.9% saline; the CO is calculated from this. The lower portion shows dilution of the optical density by this isotonic injection. Figure 2 shows an example of the concurrent ultrasound and optical dilution curves obtained following a measurement injection of 10 ml of 5% saline. The first (upper) curve, which is biphasic, shows the water flux from lung to blood (upward curve) and then the reverse movement of water back into the lung tissue (downward curve). The second (lower) shows the optical density change in blood following the hypertonic saline injection. What is measured by the arterial probe is the net result of that injected less that which moved into the lung tissue. Figure 3 superimposes the dilution curves recorded by the ultrasound arterial sensor. Curve 1 is that obtained following the injection of isotonic saline (e.g.,Figure 1 upper part), but the curve has been normalized to have the same area under the curve as does curve 2, which is that obtained following the injection of hypertonic saline (e.g.,Figure 2 upper part). Curve 1 has also been inverted to allow the area between these curves to be measured and the time taken for the two curves to cross to be identified. Figure 4 superimposes the dilution curves recorded by the optical arterial sensor. Curve 1 is that obtained from isotonic saline (e.g.,Figure 1 lower part) but again normalized and inverted as above. Curve 2 is that obtained from hypertonic saline (e.g.,Figure 2 lower part). In both Figures 3 and 4 the time when the curves cross is the point in time when the net movement of sodium and water is zero, the instant of osmotic equilibrium. Knowledge of this point (tp), together with the areas between the curves, allows the calculation of lung water volume from the equation: MATHwhere

  1. CO is cardiac output;
  2. hp and point tp, are calculated from dilution curves in either Figure 3 or 4 (same for both);
  3. 95 Sp and 95 Op are the areas between hypertonic curve and normalized isotonic curve for ultrasound and optical sensors, respectively (Figures 3 and 4);
  4. S0 9 is the area under the curve following isotonic saline by ultrasound sensor (Figure 1, upper part); and
  5. Calibration coefficient Kso =Ko/ Ks where Ko and Ks are the ratios of areas under dilution curves produced by hypertonic and isotonic calibration injections for optical (K0) and ultrasound velocity (Ks) sensors, respectively.
Figure 1
Figure 1:
An example of the dilution curves obtained following the injection of 30 ml of 0.9% (isotonic) saline obtained by ultrasound (upper) and optical density (lower) probes.
Figure 2
Figure 2:
An example of the dilution curves obtained following of 10 ml of 5% (hypertonic) saline obtained by ultrasound (upper) and optical density (lower) probes.
Figure 3
Figure 3:
The superimposition of dilution curves recorded by the arterial ultrasound probe. Curve 1 is the isotonic curve normalized to the area under the hypertonic curve (curve 2) and inverted. This allows the area between the curves (ΔSp) to be calculated and the time point (tp) where the curve lines intersect to be defined. This, together with similar information from Figure 4, allows the calculation of extravascular lung water (EVLW).
Figure 4
Figure 4:
The superimposition of dilution curves recorded by the arterial optical density probe. Curve 1 is the isotonic curve normalized to the area under the hypertonic curve (curve 2) and inverted. This allows the area between the curves (ΔOp) to be calculated and the time point (tp) where the curve lines intersect to be defined. This, together with similar information from Figure 3, allows the calculation of extravascular lung water (EVLW).

Statistical Analyses

Student’s t-test for paired data was used to compare early versus late dialysis systolic and diastolic BP, body weights, CO, and EVLW values. Values are expressed as the mean ± standard deviation and statistical significance was defined as p < 0.05. Linear regression analysis was used to determine relationships between variables in both early and late dialysis periods.

All patients gave informed consent to participate in this study.


In all 18 patients, the experimental protocol provided early and late dialysis values for EVLW and CO. The mean and standard deviations are shown in Table 1, which also gives corresponding values for systolic and diastolic BP and for patient weights.

Table 1
Table 1:
Early and Late Dialysis Values for EVLW, Cardiac Output, Blood Pressure, and Weight for Each Patient

The absolute values for EVLW show a significant decline from early (260 ± 49 ml) to late (230 ± 48 ml) dialysis (p = 0.011). The absolute decrements across dialysis in EVLW did not show a significant relationship to the corresponding weight changes (r = 0.0128) (Figure 5). The early dialysis absolute EVLW values did show a reasonable correlation with corresponding values for CO (r = 0.51, p = 0.029) (Figure 6); this relationship was lost by late dialysis. When EVLW values were normalized (to Kg body weight), while the mean late dialysis value was lower (3.02 ± 1.04 ml/Kg) than the mean early dialysis value (3.29 ± 1.0 ml/Kg), this change was not significantly different (p = 0.073). A strong correlation existed between the early and late normalized EVLW values (r = 0.83, p < 0.001). The CO values also fell significantly (p = 0.033) over the course of dialysis as did systolic (p = 0.039), but not diastolic, BP.

Figure 5
Figure 5:
The scatter plot of change in extravascular lung water (EVLW) (ml)versus the change in weight over hemodialysis. No relationship is shown.
Figure 6
Figure 6:
The correlation between extravascular lung water (EVLW) (ml) and cardiac output (L) during early dialysis before ultrafiltration. A statistically significant correlation is shown.


This study has successfully used a new technique to obtain an estimate of EVLW in hemodialysis patients using blood ultrasound velocity and optical density dilution. This method has never before been tried in humans, and we are encouraged to report that EVLW measurements using this technique are feasible; moreover, they are not difficult to accomplish by noninvasive means. In 18 hemodialysis patients judged clinically not to be in pulmonary edema, the mean EVLW was found to be 260 (± 49) ml or 3.29 (± 1.0) ml/Kg early in dialysis, before fluid removal. The method used in this experiment is virtually identical to that described by Krivitski and coworkers. 8 In their work in animals, a normalized ratio of EVLW to body weight of 3.1 ml/kg was obtained in nonpulmonary edema rabbits, and reasonable agreement was obtained between repeated measurements (mean difference 7.6% ± 11%). The measurements obtained by Krivitski were found to agree with postmortem gravimetric measurements at animal sacrifice. In the case of our measurements, there is not a suitable gold standard available to which these results can be compared, as no other technique in measuring EVLW is similar; however, our results certainly concur with those of Krivitski’s animal data using this same method, as well as other investigators who have used different techniques of measuring EVLW. For example, Sibbald et al., 11 in measuring EVLW by thermal dilution, which is known to overestimate EVLW measurements, 10 determined EVLW to be 5.6 ml/kg in ventilated nonpulmonary edema intensive care unit patients.

These 18 patients all had fluid removed during the hemodialyses during which the measurements were made. The amount of fluid removed was variable (Table 1) and depended upon the clinical need at the time. No attempt was made to remove a standard volume. It is interesting that the absolute mean value for EVLW (but not the normalized value) did decrease significantly with fluid removal, indicating that EVLW is sensitive to treatment. This change in EVLW, however, did not correlate with the amount of fluid removed; this is not surprising as the maintenance of an EVLW volume is likely a complex physiologic process and fluid removal during dialysis ultrafiltration will likely influence cardiopulmonary hemodynamics in a similar complex fashion. For example, we have shown, using sequential bioimpedance in addition to hemodynamic measurements, that during dialysis ultrafiltration fluid is predominantly removed from the extracellular fluid of the periphery, especially the legs. This is affected by increasing the peripheral resistance and the vascular refilling rate from the periphery and is done to preserve the central blood volume, i.e., that of the cardiopulmonary circulation plus the great vessels. 12 It is, therefore, probable that low values for EVLW do not occur; rather, they increase with volume overload and the development of pulmonary edema. For these pilot studies, relatively healthy and stable dialysis patients were chosen because saline injections were used, and the robustness of the equipment for measurements during a possible BP change was not known. It would be likely, therefore, that these patients would maintain a “normal” EVLW over the course of an uncomplicated dialysis procedure; the study results support this.

The fall in CO over the course of a dialysis (Table 1) with ultrafiltration has also been studied by us. 9 When fluid is removed at a rate that causes a reduction in blood volume (as measured by an on-line hematocrit monitor [Crit-Line, In-Line Diagnostics, Ogden, Utah]) of approximately 10%, the CO will fall to levels recorded here; when the plasma refilling rate equals the dialysis ultrafiltration rate to maintain blood volume, the CO is maintained. BP change then depends upon the peripheral vascular resistance response. CO values in dialysis patients are higher than normal because of the arteriovenous access. 9

The advantages of measuring EVLW and CO by this new method are that it is noninvasive and easy to use. It does not require the use of a central line and can be done very quickly at the patient’s bedside. In this preliminary work in humans, we did not set out to test the reproducibility of this method per se. However, CO measurements are required to calculate EVLW, and the measurements of CO at our institution have been found to be reproducible to within ± 10% using the Transonic Hemodialysis Monitor HDO1. 9 Therefore, we are confident in using these CO values as variables in the equation to calculate EVLW. If we take the early and late dialysis EVLW values and use those as a quasi “test–retest” data set, then they show a strong linear correlation (r = 0.83;p < 0.001) and a mean difference of only 8.2% (± 5.1%). Thus, we feel this measurement is reproducible.

To calculate EVLW, as stated above, the CO must be known. We did find, in the beginning dialysis period, that a moderate correlation existed between EVLW and CO. This relationship, although expected, was not seen in the late dialysis period. Five patients (36%) received fluid (normal saline) boluses towards the end of their dialysis treatment because of symptomatic hypotension before the EVLW measurements were carried out for the end dialysis period (Table 1). It is possible that these five patients had a subsequent rise in EVLW that could have affected our results. Certainly, our sample size is small, and greater than one third of patients received IV fluids (isotonic saline only). However, it is also possible that the relationship between predialysis EVLW and CO was biased because of possible mathematical coupling because a CO measurement is required to calculate EVLW. Further exploration of this issue is needed in future studies.

The principle finding of this study is that in 18 hemodialysis patients without signs or symptoms of pulmonary edema, the mean normalized ratio of lung water to body weight was 3.29 ml/Kg in early dialysis and 3.02 ml/Kg toward the end of dialysis. Further testing of this method is certainly required to ascertain the range of normalized lung water to body weight ratios in different clinical scenarios (for example in patients who are in pulmonary edema). Once the values for these ratios are obtained, the clinical value of this measurement may be determined. The ability to noninvasively and easily measure EVLW in hemodialysis patients by this technique may be a clinically useful adjunct in the overall assessment of dry weight and hemodynamic status. This may be especially so in cases when the BP is low and there is suspicion of pulmonary edema, and the decision to raise or lower the target weight is difficult.


The authors wish to thank Drs. N. Krivitski and V. Kislukhin (Transonic Systems Inc, Ithaca, NY) for their technical and intellectual support.


1. Charra B, Laurent G, Chazot C, et al: Clinical assessment of dry weight. Nephrol Dial Transplant 11: (Suppl.) 16–19, 1996.
2. Fishbane S, Natke E, Maesaka JK: Role of volume overload in dialysis refractory hypertension [Abstract]. Am J Kidney Dis 28: 257–261, 1996.
3. Anderson JV, Raine AE, Proudler A, Bloom SR: Effect of hemodialysis on plasma concentrations of atrial natriuretic peptide in adult patients with chronic renal failure. J Endocrinol 110: 193–196, 1986.
4. Lauster F, Heim JM, Drummer C, Fulle HJ, Gerzer R, Schiffl H: Plasma cGMP level as a marker of the hydration state in renal replacement therapy. Kidney Int 41: (Suppl.) 57–59, 1993.
5. Mandelbaum A, Ritz E: Vena cava diameter measurement for estimation of dry weight in haemodialysis patients. Nephrol Dial Transplant 11: (Suppl.) 24–27, 1996.
6. Ho LT, Kushner RF, Schoeller DA, Gudivaka R, Spiegel DM: Bioimpedance analysis of total body water in hemodialysis patients. Kidney Int 46: 1438–1442, 1994.
7. Jaeger J, Mehta R: Assessment of dry weight in hemodialysis. J Am Soc Nephrol 10: 397–403, 1999.
8. Krivitski N, Kislukhin V, Dobson A, Gleed R, Rawson R, Robertshaw D: Volume of extravascular lung fluid determined by blood ultrasound velocity and electri-cal impedance dilution. ASAIO J 44: M535–M540, 1998.
9. Pandeya S, Lindsay RM: The relationship between cardiac output and access flows during hemodialysis. ASAIO J 45: 135–138, 1999.
10. Wallin C, Rösblah P, Leksell L: Quantitative estimation of errors in the indicator dilution measurement of extravascular lung water. Intensive Care Med 23: 469–475.
11. Sibbald WJ, Warshawski FJ, Short AK, Harris J, Lefcoe MS, Holliday RL: Clinical studies of measuring extravascular lung water by the thermal dye technique in critically ill patients. Chest 83: 725–731, 1983.
12. Shulman T, Heidenheim AP, Kianfar C, Shulman S, Lindsay RM: Preserving central blood volume: Changes in body fluid compartments during hemodialysis. ASAIO J 47: 615–618, 2001.
Copyright © 2002 by the American Society for Artificial Internal Organs