Zdolsek, Joachim MD, PhD; Li, Yuhong MD, PhD; Hahn, Robert G. MD, PhD
To maintain hydration status in balance is a key challenge in the treatment of diseases that involve excessive fluid loss. Normal hydration is also important in patients scheduled for anesthesia and surgery. Serious dehydration, particularly if dominated by volume losses from the extracellular fluid (ECF) space, can result in increased hemodynamic effects of anesthetic drugs, poor tissue perfusion, and oliguria. The consequences of moderate dehydration are less well known, although the administration of approximately 2 L more or less of crystalloid fluid during abdominal surgery affects the incidence of postoperative complications.1
The judgment of whether dehydration is present is usually based on physical inspection of skin turgor. An increase in the blood hemoglobin (Hgb) or serum creatinine concentration is consistent with dehydration, but a recent control measurement might not be available. Clinical methods, such as skin turgor, may disclose dehydration amounting to 5% to 10% of the body weight, but it is unlikely that they can reliably detect smaller variations in fluid balance in preoperative surgical patients.
The primary hypothesis of the present exploratory study was that volume kinetic analysis of a small amount of crystalloid fluid could be used to detect fluid losses amounting to 2% of body weight. The kinetic analysis used serial analyses of the blood Hgb concentration to quantify the distribution and elimination of the infused fluid.2–8
A secondary hypothesis was that dehydration could be indicated by noninvasive monitoring of the pleth variability index (PVI) using pulse oximetry upon a change of body position.9–12 For this purpose, conscious volunteers were studied both when being normo-hydrated and when dehydration had been induced by IV furosemide.
Forty experiments were performed on 10 male volunteers with a mean age of 22 years (range 19–37), a body weight of 80 kg (range 75–100), and a body mass index of 24.9 kg/m2 (range 21.8–28.4). None of the subjects had any disease or took daily medication. The Ethics Committee of Linköping University approved the protocol (Ref. M114-09, ClinicalTrials.gov Identifier NCT01062776). Each volunteer gave his consent for participation after being informed about the study both orally and in writing.
Table 1 shows a flow diagram for the study. On the day of the experiment, the volunteers had fasted since midnight and then drank 800 mL (approximately 10 mL/kg) of water at 6:00 AM. The purpose of the fluid intake was to correct any preexisting dehydration while allowing the body time to void and to arrive at its desired degree of hydration 2 hours later. The experiments started at 8:00 AM in a single room at the Department of Intensive Care at Linköping University Hospital. To achieve maximum comfort, an OPN Thermal Ceiling radiant warmer (Aragona Medical, River Vale, NJ) had been placed approximately 1 m above the bed.
A cannula was placed in the cubital vein of one arm to infuse fluid and another cannula was inserted in the other arm for sampling blood. Plasma volume expansion was then induced by administering acetated Ringer's solution (Baxter Medical AB, Kista, Sweden) IV over 15 minutes using an infusion pump (Volumat MC Agilia; Fresenius Kabi, Bad Homburg, Germany).
All 10 volunteers underwent 4 experiments in a random order determined by the sealed envelope method, and separated by at least 2 days. The experiments were designed as follows:
1. Infusion of 5 mL/kg acetated Ringer's solution in the euhydrated state
2. Infusion of 10 mL/kg acetated Ringer's solution in the euhydrated state
3. Infusion of 5 mL/kg acetated Ringer's solution in the dehydrated state
4. Infusion of 10 mL/kg acetated Ringer's solution in the dehydrated state
The Ringer's solution had the following ionic content: sodium 130, chloride 110, acetate 30, potassium 4, calcium 2, and magnesium 1 mmol/L.
The fluid bag was kept on the radiant warmer above the bed to maintain a temperature of 27°C to 30°C when entering the volunteer.
Volume depletion was induced by repeated small IV doses of 5 mg of furosemide (Furix 10 mg/mL; Nycomed, Stockholm, Sweden) provided every 10 to 30 minutes until the urinary excretion amounted to between 1.5 and 2.0 L. Sixty minutes elapsed between the last dose of furosemide (total dose 25 mg, range 15–35) and the infusion of Ringer's solution.
The volunteers were weighed to the nearest 0.01 kg on an electronic scale (Vetek AB, Väddo, Sweden) just before, and at the end of, the dehydration period.
Fluid Recruitment Maneuver
To examine whether pulse oximetry could be used to detect dehydration, 3 pulse oximeter variables were recorded every 5 minutes before and after a change in body position from semistanding (0–10 minutes) to the recumbent position (11–25 minutes), a maneuver that relieves hypovolemia by recruiting interstitial fluid from the legs to increase the plasma volume by approximately 6% within 30 minutes.13
The test was performed as the first part of each experiment and, in experiments 2 and 4, after dehydration when the volunteer awaited an infusion of Ringer's solution (Table 1). “Semistanding” meant that the volunteer was standing but leaning against the bed, a position chosen to safeguard against fainting in the presence of dehydration.
During the fluid infusion experiments, venous blood (4 mL) was withdrawn to measure the Hgb concentration and the hematocrit (Hct) on a CELL-DYN Sapphire (Abbott Diagnostics, Abbott Park, IL). Sampling was performed every 5 minutes during the first 50 minutes, and then at 60, 70, 90, 105, and 120 minutes. The baseline sample was drawn in duplicate, and the mean of the 2 concentrations was used in subsequent calculations. The differences between the 2 baseline samples for all 40 experiments were used to calculate the coefficient of variation (precision) for the Hgb measurement, which was 0.8%.14 Seventeen samples were taken. A small discard volume of blood was drawn before each blood collection to preclude an admixture of rinsing solution, and 2 mL of 0.9% saline was then injected to prevent clotting.
The subjects voided when they entered the study room and this urine was discarded. They voided freely in the lying position throughout the experiments. The total volume was recorded when the volunteers emptied their bladders completely at the end of the dehydration period and at the end of the infusion experiment.
A Radical-7 ™ pulse oximeter (Masimo Corp., Irvine, CA) was applied to the index or middle finger of one hand and it was left in the same place throughout the experiment (software version SET V188.8.131.52, Handheld R.184.108.40.206 and D-station R220.127.116.11). A repeated-use probe with a programmed obsolescence of 60 hours was used. This apparatus measured the perfusion index (PI), venous Hgb concentration, and the PVI. The display was set to average the data every 6 seconds.
PI represents the pulse amplitude. PVI is the breath-induced variation in PI and is calculated as (PImax − PImin)/PImax over several respiratory cycles.9 Upon inclusion in the study, the volunteers were tested to meet a minimum requirement of PI >2 (maximum value 14) as peripheral vasoconstriction distorts the noninvasive Hgb analysis.
Pulse oximetry data are given for the fluid recruitment maneuver only. The relationship between the invasive and noninvasive Hgb analyses obtained during the infusion experiments have been reported previously and were found to be inadequate for use in volume kinetic analysis.15
Volume kinetic analyses were performed according to 1-volume and 2-volume models (Fig. 1). In the first one, the infusion of fluid at the rate Ro expanded the functional body fluid volume Vc to vc. Removal of the fluid occurred by means of baseline fluid losses (Clo) that were preset to 0.4 mL/min, and by a dilution-dependent elimination clearance mechanism (Cl).2 The differential equation for the model is:
Equation (Uncited)Image Tools
The 2-volume model added a peripheral compartment Vp that expanded to vp by exchanging fluid with vc at a rate governed by a distribution clearance, Cld. The differential equations for this model are:
Equation (Uncited)Image Tools
Equation (Uncited)Image Tools
The Hgb-derived fractional plasma dilution was used as the input Hgb(t) at time t:
Equation (Uncited)Image Tools
where Hgb and Hct are the blood Hgb concentration and the Hct at baseline, respectively. A correction for the loss of sampled Hgb was applied.7 Previous work has defined these models2 and given analytical3 and matrix solutions4 to the equations used.
Curve-fitting was performed by least-squares regression, based on the analytical solutions to the differential equations, and programmed into MATLAB 4.2 (MathWorks Inc., Natick, MA). No weights were used because the analyses were based on blood Hgb changes of between 0 and 7%–8% in which the analytical precision of the Hgb assay was practically identical.
Three approaches were used for the presentation: 1 all curves analyzed by the 1-volume model based on all 16 data points, 2 all curves analyzed by the optimal model, being either the 1- or 2-volume model as determined by the F test,2–4 and 3 the 1-volume model based on 4 data points only.
For the 1-volume model, the output was the optimal estimates of Vc and Cl. The elimination half-life of infused fluid (T1/2) was calculated as 0.693 Vc/Cl. For the 2-volume model, the output parameters were Vc, Vp, Cl, and Cld. The elimination T1/2 was 0.693 Vc/Cl.
The renal clearance (Clr) of the infused fluid was determined as the urine volume excreted during the 2-hour infusion experiment divided by the area under the plasma dilution-time curve.2,4 The Clr was calculated for only 38 experiments because 2 volunteers were unable to void at the end of the 10 mL/kg infusions during euhydration.
Hence, there were 4 clearances to consider. Clo and Clr were model-independent while there was one elimination Cl in the 1-volume model and another in the 2-volume model. The former Cl is higher because the 1-volume model is expected to be statistically justified when elimination is efficient (resulting in high urinary excretion). The elimination Cl might also be higher in the 1-volume model because some fluid has distributed to a peripheral compartment (by the mechanism Cld), although not in large enough amounts to change the slope of the dilution-time curve markedly. If the urine collection is complete, Clr is close to the Cl of the optimal model16 except during surgery when extravascular retention of fluid might occur so that Clr < Cl.7
The result of the very first experiment in the series deviated markedly from the others, showing a very low fluid clearance despite euhydration. The volunteer had been up all night before this experiment session and had been fasting for a period; both factors were unknown to us at the time. Therefore, this experiment was repeated several months later when the volunteer followed the instructions for normal sleep and food intake.
The results are given as the mean (SD), except when there is a skewed distribution whereby data are reported as the median (25th–75th percentiles).
Two-way analysis of variance (ANOVA) was used to examine whether the kinetic parameters differed depending on the infused fluid volume or dehydration. Logarithm-transformed (Cl, Clr, and T1/2) or square root–transformed data (urine and ratio urine/infused fluid) were applied if the distribution of the data was skewed. After transformation, the distribution was close to normal.
Changes were reported as the median and the 95% confidence interval (CI) for the difference between paired data. Differences between paired and unpaired samples were examined by the Wilcoxon matched-pair test and the Mann-Whitney U test, respectively.
Logistic regression was used to select the probability range where sensitivity and specificity of various kinetic parameters to indicate dehydration were as similar as possible. The middle of this range was selected as the cutoff point, which was then used to calculate the 95% CI for the sensitivity and specificity in question.
The study was powered (90%) based on retrospective data17 in which the blood Hgb was 96.9% (SD 2.9%) of baseline 30 minutes after the end of the 30-minute infusion of Ringer's solution in euhydrated volunteers. The corresponding Hgb level was 91.4% (3.1%) when the same volunteers underwent the infusion experiment again after an overnight fast. The current study strove to detect a difference of 5% in Hgb levels, depending on the hydration status at 60 minutes, with a power of 90% and P < 0.05.
The F test determined whether the 1-volume or 2-volume kinetic model was statistically justified.2–4 P < 0.05 was considered statistically significant.
Furosemide promoted urinary excretion amounting to 1.71 (0.26) L (range 1.38–2.43 L), which represented 2.1% (0.4%) of the body weight (range 1.5%–3.0%).
The body weight decreased by 2.0 (0.4) kg (range 1.5–3.0 kg) from the dehydration procedure. The difference between measured weight loss and measured urine excretion is attributable to evaporation and metabolism (Clo).
The Hgb concentration increased by 7.9% (SD 2.1%, range 2.1%–11.5%) during the dehydration procedure. Assuming that the blood volume amounts to 7.5% of the body weight, this Hgb change corresponds to a reduction of the blood volume by 0.50 (0.13) L. Hence, 30% (9%) of the voided volume was derived from the blood. There were no differences between the 5 mL/kg and 10 mL/kg experiments with respect to these data.
Dehydration reduced the elimination Cl by more than half (median ratio 0.36; 95% CI 0.25–0.40) and more than doubled T1/2 (mean 3.3, 95% CI 2.5–6.2).
T1/2 was 23 minutes in euhydration and 76 minutes in dehydration (Wilcoxon matched-pair test, P < 0.001; Table 2).
Two-way ANOVA showed that dehydration (P < 0.001), but not the infused volume (being 5 or 10 mL/kg), had a statistically significant influence on Cl and T1/2. Contrast estimates yielded that dehydration decreased Cl by 1.49 (95% CI 0.90–2.08) mL/min/kg and increased T1/2 by 67 (38–96) minutes. The nonsignificant effect of the larger fluid volume resulted in a change of Cl by 0.15 (−0.44 to 0.75) mL/min/kg and of T1/2 by 10 (−19 to 40) minutes.
The best capacity to distinguish dehydrated from euhydrated volunteers was obtained when parameters were derived from the 5 mL/kg experiments (Fig. 2, A and B). With optimal cutoff points, there was complete separation between the data on Cl and T1/2 depending on the state of hydration (Table 3). The sensitivity and specificity were at least 10% lower with the infusion of 10 mL/kg.
Optimal Kinetic Model
The 2-volume kinetic model was statistically justified (by the F test) in 22 of the 40 experiments (1 in group 1, 5 in group 2, and 8 each in groups 3 and 4). The 1-volume model was appropriate in the other 18 experiments (Table 2).
Dehydration (P < 0.001), but not the volume of infused fluid (P = 0.45–0.82), had a statistically significant influence on Cl and T1/2 when the data from the optimal model in each experiment were pooled (2-way ANOVA; Figs. 3 and 4).
Using a cutoff point of 40 minutes, the T1/2 of the pooled optimal model fully discriminated between euhydration and dehydration when 5 mL/kg had been infused (Fig. 2C). For the infusion of 10 mL/kg, the sensitivity and specificity were only 83% and 80%, respectively.
The Clr and the excreted/infused fluid ratio were both lower in the presence of dehydration (P < 0.0001 and P < 0.03) and with the larger fluid volume infusion (P < 0.003 and P < 0.01; 2-way ANOVA; Table 2).
Clr readily distinguished between euhydration and dehydration (Table 3), while the excreted urine volume alone showed sensitivity and specificity of only 60% to 70%.
Comparisons of the plasma dilution in euhydration and dehydration suggest that the fractional plasma dilution values are separated by the largest amount approximately 60 minutes after the completion of the 15-minute fluid infusion (Fig. 4, C and D). The Hgb concentration was then 98.2% (mean) of baseline in euhydration and 95.8% in dehydration (Wilcoxon matched-pair test, P < 0.001).
The 1-volume kinetic analysis was then repeated but based on only the 4 points in time when the within-series and between-series differences in Hgb concentration were greatest (0, 15, 60, and 70 minutes). Here, V and Cl attained 9% and 18% lower values than in the 16-point analysis, but the capacity of the kinetic model to identify dehydration of approximately 2% was largely retained (Table 3).
The difference between the blood Hgb level at 60 to 70 minutes and the baseline value could identify dehydration as well as the 4-point kinetic analysis (Table 3).
A comparison between the pulse oximetry variables obtained at baseline of the fluid recruitment maneuver performed before and after dehydration (in experiments 3 and 4) showed that the dehydration procedure decreased the PI while the Hgb increased (both P < 0.01); the PVI was unchanged.
The noninvasive PI, Hgb, and PVI showed the same pattern upon change of body position regardless of hydration status. The only consistent change afterward was a progressive decrease of the Hgb concentration that eventually reached 5% to 6% (repeated-measures ANOVA, P < 0.001; Fig. 5).
The results showed that kinetic analysis of a small bolus infusion of Ringer's solution could be used to disclose that a healthy volunteer was volume depleted by an average of 2.1% of the body weight. Cl and T1/2 discriminated well between the fluid infusions that were performed in the euhydrated and dehydrated states. The smaller infusion (5 mL/kg) yielded a more reliable discrimination than the larger one (10 mL/kg).
This novel approach is based on the belief that the volume status can be determined by studying how the body handles a fluid challenge.17 Euhydration results in an effort to regain the preinfusion hydration level. In dehydration, the body retains infused fluid, which decreases Cl and prolongs T1/2 for acetated Ringer's solution.
Previous work has suggested this to be the case. The T1/2 of crystalloid fluid was only half as long in semifasting volunteers18 and laparoscopic patients19 when volume depletion was prevented by a previous infusion of the same fluid (15 vs 28 minutes and 19 vs 38 minutes, respectively). In fasting volunteers subjected to hemorrhage, the T1/2 of acetated Ringer's solution increased from 22 minutes (no hemorrhage) to 45 minutes (500 mL) and 62 minutes (900 mL).5 However, the administered fluid volumes in those studies were 2 to 5 times larger than in the present study.
Tracer methods for measuring body fluid volumes, such as deuterium and bromide, have a precision of 1% to 2% if performed carefully,20 but they involve expensive analytical procedures and may require very long mixing times. Moreover, they do not provide information about the hydration level the body strives to maintain. Clinical methods, such as assessment of skin turgor and measurement of serum osmolality, are not sensitive enough to detect minor deficits in body fluid volume.
The dehydration test presented here overcomes most of these problems, which makes our approach useful for research. Less-frequent blood sampling than the 16 points used in this exploratory study would be desirable to reach clinical efficacy. Interestingly, useful results seem to have been obtained from only 4 samples, and 2 samples indicating a shift of the Hgb baseline may suffice to detect dehydration (Table 3). However, with fewer samples, the results will be more sensitive to occasional errors in sampling technique and analysis of the relatively minor changes in Hgb concentration studied here. However, the relatively long period of time (60 minutes) between the first and the second blood sample cannot be significantly shortened (Fig. 4, C and D).
The choice of dehydration method and kinetic model deserves a comment. Between 50% and 70% of patients require blood volume optimization before surgery, regardless of whether they fast overnight or ingest fluid preoperatively,21–23 and the blood volume is greatly affected by changes in the ECF volume.24 This explains our use of a better method of inducing dehydration than fasting overnight. Euhydration at baseline was ensured by letting each volunteer drink 800 mL of tap water and giving him 2 hours to excrete any excess fluid; controlled dehydration was induced by furosemide. Because the plasma constitutes 1/3 of the expandable ECF volume,2 the increase in Hgb concentration during the dehydration procedure confirms that furosemide primarily reduced the ECF volume. Hence, the type of dehydration studied here better reflects fluid losses by vomiting and diarrhea (volume depletion) than by evaporation alone.
The 1-volume model offers simplicity and stability, and therefore, clinical applicability. The kinetics may also be handled by adding a second compartment to the model. The biexponential form of the plasma dilution-time curve is typical of fluid retention and can be expected in hypovolemia,5 anesthesia,8 and in anesthesia and surgery.7 Several of the present experiments with 10 mL/kg showed the biexponential form, as did most curves collected during dehydration.
The present work shows the ability of both the 1-volume and the statistically justified (optimal) model to detect dehydration. The 2-volume model, which overall proved to be optimal in more than half of the experiments, yielded 15% longer elimination T1/2 for acetated Ringer's solution than the 1-volume model did. Because the 2-volume model was most common in dehydration, the difference in T1/2 between euhydration and dehydration increased when also considering the 2-volume model. Although the statistical comparison between the pooled Cl and T1/2 can be questioned on theoretical grounds, the data disclose the marked differences in kinetic parameter values that will be found when studying the hydration status of a volunteer by using the optimal model. They also illustrate that the T1/2 of Ringer's solution exceeding 40 minutes serves well as a cutoff point between euhydration and dehydration, regardless of the choice of volume kinetic model. Even the model-independent Clr solved the task well whereas the urinary excretion alone was hardly useful. Table 3 shows future investigators the sensitivity and specificity of kinetic parameters that are capable of detecting dehydration.
Infusing a large fluid volume causes greater depression of the Hgb level but does not seem to improve the usefulness of the dehydration test.17 The reason might be efficient elimination of small volumes of fluid in euhydrated volunteers. The urinary excretion even slightly exceeded the infused volume after 2 hours of the experiment with 5 mL/kg (data not shown), which resulted in a high elimination Cl. Here, there was a marked predominance of the 1-volume model over the 2-volume model (relationship 9/1), which makes peripheral accumulation of fluid less likely. Dehydration-induced renal conservation of fluid apparently contrasted very well and consistently against high-elimination Cl.
The secondary hypothesis was that pulse oximetry could serve as a tool to reveal dehydration. This was not possible with the approaches used here. Reduction of body water changed the baseline for PI and Hgb, but the values differed too much between individuals to be useful as a test of dehydration (Fig. 5).
The fluid recruitment maneuver, which consisted of a change of body position from semistanding to lying down, was designed to disclose relative differences in the degree of capillary refill depending on hydration. The similar Hgb responses to this maneuver are consistent with the expectation that furosemide would dehydrate the plasma and the interstitial fluid volumes to a similar degree. An extension of the follow-up time after the change in body position could possibly have shown a dehydration-dependent difference in Hgb concentration, just as it did after the fluid infusion. However, the noninvasive Hgb technology, in contrast to invasive laboratory analysis of Hgb, does not yet offer the precision required to capture small Hgb differences in individuals.15
Interest has been focused on the PVI as an index of fluid responsiveness during mechanical ventilation,9,10 and although the PVI is a less-precise measure, it seems to work in the conscious state as well.11,12 However, we found no difference in the baseline PVI and no difference in the PVI response to the fluid recruitment maneuver that depended on the state of hydration.
The present study has a proof-on-concept design and challenges whether volume kinetics can detect dehydration and, if so, how much infusion fluid would be needed to detect dehydration. The size of the body fluid space(s) expanded by an infusion can apparently not detect dehydration, but elimination parameters can. Moreover, only a small volume of fluid needs to be infused. Future studies with other designs are needed to examine whether volume kinetics can discriminate between variable degrees of dehydration. The test also needs careful validation in patients with vascular or renal disease.
The limitations to our study include the necessity for blood sampling to distinguish safely dehydrated from euhydrated volunteers. Infusions made in the euhydrated state began at 8:30 AM, and in the presence of dehydration, they began at 11:00 AM. A bladder catheter was not used, which could have improved the quality of the urine data. The repeated experiment shows that aberrant results could be produced by an abnormal diurnal rhythm or lifestyle. Drobin's single-volume model25 was first fitted to the data but required stabilization from the urinary excretion values to work when based on only 16 samples.
The perceived role of the dehydration test is in the preoperative setting. In adults awaiting hernia surgery after an overnight fast, half of the patients had a T1/2 of Ringer's solution exceeding 40 minutes,26 which is a likely incidence. The test is unsuitable for application during general anesthesia and surgery because other mechanisms greatly prolong the T1/2 of acetated Ringer's solution under those conditions.2,7,8,27 Moreover, the fluid chosen for the test must be retained if the kidneys sense dehydration but excreted if they do not. Therefore, colloid fluids, or blood, are not viable options.
If the test shows that a patient is dehydrated, the rational treatment is to replace the missing volume by oral or IV fluid before anesthesia is induced. Alternatively, those who are dehydrated may be given more fluid perioperatively to maintain organ perfusion. In contrast, patients who are euhydrated should receive relatively less fluid to avoid tissue edema, because the body's capacity to excrete a surplus of crystalloid fluid is very poor during anesthesia and surgery.2 The dehydration test could provide the anesthesiologist with the information needed to individualize the fluid therapy. However, it is unclear at present whether such a strategy would help to reduce the complication rate or hospital time. The requirement for invasive sampling also contributes to our belief that the dehydration test will primarily be used for research.
In conclusion, the half-life of a small volume of crystalloid fluid (5 mL/kg) discriminated between euhydration and dehydration in healthy volunteers. This finding opens a possibility to disclose moderate degrees of dehydration and/or volume depletion that previously have been difficult to explore. The prevalence and pathophysiological role of such dehydration in the preoperative setting are currently unclear. Further studies are needed to investigate whether the Hgb response to a preoperative fluid challenge can be used to refine and individualize perioperative fluid therapy.
Name: Joachim Zdolsek, MD, PhD.
Contribution: This author helped conduct the study and analyze the data.
Attestation: Joachim Zdolsek has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.
Name: Yuhong Li, MD, PhD.
Contribution: This author conducted the experiments.
Attestation: Yuhong Li has seen the original study data, approved the final manuscript, and is the author responsible for archiving the study files.
Name: Robert G. Hahn, MD, PhD.
Contribution: This author helped design the study, analyze the data, write the manuscript, and arranged finance.
Attestation: Robert G. Hahn has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.
This manuscript was handled by: Tony Gin, FANZCA, FRCA, MD.
Nurses Susanne Öster and Helén Didriksson assisted during the experiments. Statistician Elisabeth Berg, Karolinska Institutet, performed many of the data analyses. The European Society of Anesthesiology supported the project.