Point-of-Care Ultrasound-Guided Conservative Fluid Therapy Improved the Outcome of Critically Ill Patients: A Before-and-After Study

Objectives To study the impact of uid balance on the outcome of critically ill patients. Methods Critically ill patients managed with point-of-care ultrasound were compared with those managed without. Distended internal jugular veins and inferior vena cava with reduced collapsibility were taken as signs of hypervolemia. Results Compared with critically ill patients admitted before application of point-of-care ultrasound assessment (from March, 2019 through October, 2019; 291 cases), cases admitted after (from November, 2019 through June, 2020; 285 cases) had signicantly lower in-ICU mortality (34.7% vs 26.7%, p=0.038; Fisher’s exact test), together with a dramatic change from overall positive uid balance to negative one (for cumulative uid balance during ICU stay, 2820±1381ml vs -10±39ml; p=0.001). Multiple logistic regression showed that cumulative uid balance during ICU stay, Acute Physiology and Chronic Health Evaluation (APACHE) II score and Sequential Organ Failure Assessment (SOFA) score on admission were independent risk factors for in-ICU mortality (p<0.001, p<0.001 and p=0.043 respectively). After controlling for disease severity, Cox hazard ratio of cases with a negative cumulative uid balance during ICU stay was 0.683 (95% condence interval 0.475-0.981; p=0039).


Introduction
Fluid therapy was of paramount importance in the management of critically ill patients. Fluids in excessive amounts could have deleterious effects on several organ functions, including acute kidney injury and worsening respiratory function [1]. The danger of volume overload had been noted in some critically ill patients, namely sepsis/septic shock patients [2][3][4]. However, the focus of intensivists was usually limited to resuscitation phase immediately after hospitalization of critically ill patients; intensivists paid little attention to uid balance after initial hemodynamic stabilization. Fluid removal or de-resuscitation after hemodynamic stabilization was seldom put into practice [4,5]. And too often, uid balance of critically ill patients other than sepsis/septic shock cases was not the concern of intensivists. Up to date, the danger of volume overload was not well understood and appreciated.
In previous studies, internal jugular vein was found to be a good surrogate of central venous pressure (CVP) [6][7][8][9][10][11]. Distention of internal jugular veins had the same implications as distention of super cial neck veins, and the only difference was that ultrasound could probe deeper. Although visual inspection of jugular veins as an estimate of CVP had been an integral part of physical examination, its major limitation was that jugular veins were not always observable, especially in obese patients [12]. Although the disagreement persisted as to the interpretation of size and collapsibility of inferior vena cava [13][14][15][16], distension and reduced collapsibility of inferior vena cava pointed to normo-/ hyper-volemia and were proved to be very bene cial in guiding uid therapy [17]. The present study was carried out to compare the outcome of critically ill patients before and after the implementation of point-of-care ultrasound examination of both internal jugular veins and inferior vena cava.

Results
From March, 2019 through June, 2020, 576 cases were enrolled. Cases enrolled before application of point-of-care ultrasound were not signi cantly different from those enrolled after application of ultrasound, in respect to age, principal diagnoses, and APACHE II and SOFA scores on admission (Table  1).
On admission, more than half cases had a serum BNP>400 pg/ml or in the gray zone (between 100 and 400 pg/ml). 40% cases had a peak serum BNP>400 pg/ml. Both serum BNP on admission and peak serum BNP of non-survivors were signi cantly higher than that of survivors (Fig 1).
Cumulative uid balance during ICU stay for cases enrolled before application of ultrasound was signi cantly more than that for cases enrolled after application of ultrasound; there was a dramatic change from overall positive uid balance to negative one (Fig 2). After application of point-of-care ultrasound, intravenously-administered furosemide was markedly increased; signi cantly more furosemide was administered intravenously during the rst three days to attain a negative uid balance during ICU stay (Fig 3). In-ICU mortality of cases enrolled before application of ultrasound was signi cantly higher than that of cases enrolled after application of ultrasound (34.7% vs 26.7%, p=0.038; Fisher's exact test); cases with positive cumulative uid balance during ICU stay had a markedly higher in-ICU mortality than those with negative one (Fig 4).
After controlling for APACHE II and SOFA scores on admission, Cox hazard ratio of cases with a negative cumulative uid balance during ICU stay was 0.683 (95% con dence interval 0.475-0.981; p=0039).
Multiple logistic regression showed that cumulative uid balance during ICU stay, APACHE II score and SOFA score on admission, not day 2 and day 3 cumulative uid balance, were independent risk factors for in-ICU mortality. APACHE II score and SOFA score on admission was not collinear; condition index was <10 (8.67; linear regression). Studentized residuals was <2 (Fig 5). APACHE II and SOFA scores on admission, cumulative uid balance during ICU stay, day 2 and day 3 cumulative uid balances all could predict in-ICU mortality (Fig 6).

Discussion
In the present study, two senior physicians of the same ICU, with similar background, were sent to preside over ICU daily works of Jinniu Hospital a liated to Sichuan Provincial Hospital. And the patients enrolled before application of point-of-care ultrasound were not signi cantly different from those enrolled after, in respect to principal diagnoses, disease severity on admission and age. These patients were comparable, except for the application of point-of-care ultrasound and physicians' uid preference.
The serum BNP values of these patients suggested that more than half cases probably experienced congestive heart failure during ICU stay. Volume removal was of paramount importance in these cases.
The present study showed that cumulative uid balance during ICU stay was an independent risk factor of in-ICU mortality. After application of point-of care ultrasound, a dramatic change from positive uid balance to negative one occurred. And the change in uid balance was associated with a signi cant reduction in in-ICU mortality. After controlling for APACHE II score and SOFA score on admission, negative cumulative uid balance during ICU was protective.
It was reasonable to direct uid therapy with point-of-care ultrasound examination of internal jugular veins and inferior vena cava. Point-of-care ultrasound examination of internal jugular veins should not be viewed as a brand-new concept; distention of internal jugular veins had the same implications as distention of super cial neck veins. Point-of-care ultrasound examination of internal jugular veins was better than visual inspection of super cial neck veins, in that it could probe deeper. One major limitation of visual inspection was that jugular veins were not always observable, especially in obese patients [12].
Distension and reduced collapsibility of inferior vena cava pointed to normo-/ hyper-volemia and were proved to be very bene cial in guiding uid therapy [17].
Why cumulative uid balance during ICU stay notably affected the outcome of critically ill patients? There were two reasons worth mentioning. First, some critically ill patients were hypervolemic from the very beginning of their hospitalization. Second, catabolism and disuse.
Previous studies showed that many critically ill patients were essentially hypervolemic. For example, in critically ill anemic patients, hypervolemia was common [18]. "Plasma volume excess", in addition to "hemoglobin de ciency", often characterized anemia when plasma volume was measured [19]. Plasma volume could increase by 70% in chronic severe anemia [20]. In heart failure with preserved ejection fraction, expanded total blood volume together with true anemia was common [21,22]. In volume overload decompensated chronic heart failure, plasma volume expansion dilution-related "anemia" predominated [23]. Dilution in consequence of sodium and water retention was an important reason for development of anemia in chronic heart failure [24]. In patients with chronic kidney disease, anemia was often complicated with uid retention and overhydration, resulting in increased risk of cardiovascular morbidity and mortality and chronic kidney disease progression [25]. Expansion of plasma volume and hemodilution were important features of anemia due to hemopoietic diseases, such as leukemia [26]. In the present study, many patients had elevated serum BNP on admission.
Most critically ill patients experienced a catabolic course which was not easily reversed by medication or nutritional support. Critically ill patients developed muscle atrophy rapidly [27], as a result of both anabolic resistance (reduced stimulation of muscle protein synthesis to a given dose of protein/amino acids) and enhanced skeletal muscle breakdown[28], or as a result of iatrogenic protein undernutrition [29], and preferentially affecting the lower limbs [30]. Both catabolism and disuse in bedridden cases resulted in breakdown of proteins and reduction of dry body weight, leading to production of "endogenous water" and release of water from degraded proteins and fats. Fluid derived from catabolism and disuse should be removed to avoid volume overload and tissue edema.
In critically ill postoperative patients and patients with severe nutritional depletion, a dramatic change in body composition was observed. Reduction of body weight was accompanied by a marked increase in extracellular water (by 4.0-5.1 liters in 31-34 days) [31]. Rapid and massive total protein loss was accompanied by retention of large amounts of uids in critically ill patients with major traumatic injury or severe sepsis. Trauma patients had retained > 6L and sepsis patients > 12L of uids, mainly in the extracellular compartment, by the time they were hemodynamically stable. And in elderly patients (> 60 years) the period of extracellular water expansion was markedly prolonged compared to younger patients (< 40 years) [32].
Fluids in excessive amounts could have deleterious effects on several organ functions, including acute kidney injury and worsening respiratory function [1]. The danger of volume overload was already noted in some critically ill patients, namely sepsis/septic shock patients [2][3][4]. The present study was in line with these studies, suggesting that conservative uid strategy was superior to liberal one, and that uid removal or de-resuscitation after hemodynamic stabilization should be put into practice.
The negative impact of positive cumulative uid balance could not be overestimated. Many hypervolemic patients died when given diuretics. If uid management could be improved, these cases would add to the difference in survival between cases enrolled before application of point-of-care ultrasound and those enrolled after. The diuresis of these deceased patients was not well titrated, in respect of timing and dosing of diuretic administration. Admittedly, the uid balance was di cult to maintain in these critically ill patients. Studies conducted by nephrologists showed that shift of uid between intravascular and extravascular compartments could be surprisingly rapid; plasma re ll rate reached up to 20.1 ± 4.0 ml/min during hemodialysis [33][34][35]. When re-absorption far exceeded the rate of volume removal, a "well-controlled" pace of negative uid balance would result in hemodynamic instability.
One main problem was that physicians had di culty adapting to such a paradigm shift. For some physicians, uid challenge and positive uid balance were still the only solution when facing shock and lactic acidosis. The improvement in outcome would be greater if all physicians had abandoned uid preference.
Another limitation of the present study was that 28-day or 90-day mortality was not recorded. However, the contribution of post-ICU treatment to 28-day or 90-day mortality was di cult to estimate, and physicians' and surgeons' uid preferences would have major impact on these outcome measures.
In conclusion, volume overload might become increasingly evident in the course of illness. If not curbed, volume overload would minimize the chance of survival of critically ill patients. Overgenerous uid administration should be avoided, and uid balance should be meticulously managed during the whole ICU stay. More studies are needed to address the importance of uid management during the whole ICU stay.

Methods
We conducted this study in Intensive Care Unit (ICU) of Jinniu Hospital a liated to Sichuan Provincial Hospital. This ICU was a 24-bed setting for medical and surgical patients. This hospital had no anesthesia recovery room and emergency intensive care unit. The Ethical Committee of Jinniu Hospital a liated to Sichuan Provincial Hospital approved the study protocol. The informed consents were obtained from the patients or their close relatives. The study was conducted in accordance with the Declaration of Helsinki.
From March, 2019 on, senior physicians from Emergency Intensive Care Unit (EICU) of Sichuan Provincial Hospital were sent to preside over daily works of this ICU, rst by Dr Deng for 8 months, then by Dr Chen (the corresponding author) for another 8 months. During the rst 8 months, point-of-care ultrasound was seldom used. Point-of-care ultrasound was widely used during the late 8 months. Distended internal jugular veins and inferior vena cava were viewed as signs of hypervolemia, together with raised CVP or brain natriuretic peptide (BNP) or radiological/clinical ndings of pulmonary edema.
Cross-sectional area of internal jugular vein and diameter of inferior vena cava were measured with Mindray M9 (Shenzhen Mindray Bio-medical Electronics Co., Ltd., Shenzhen, Guangdong Province, China). Internal jugular vein was measured at cricoid level. Inferior vena cava was measured just below the xyphoid process, and within 2 cm from right atrium. They were both measured at end-expiration. To measure internal jugular vein, the probe was gently placed on the skin; internal jugular vein was readily deformed by any pressure applied to it. Cross-sectional areas of internal jugular veins were measured because many of them were in irregular shapes, and only a small number of them were rounded. Patients were routinely assessed with point-of-care ultrasound within 10 hrs after admission. Collapsibility index of inferior vena cava was assessed with M-mode ultrasound. Whenever distended internal jugular vein with a collapsibility index (cross-sectional area)<50% and/or inferior vena cava diameter>15 mm with a collapsibility index(diameter)<50% were detected, negative uid balance became a goal. Different from previous studies [13][14][15][16][17], distended inferior vena cava with reduced collapsibility index was taken as a sign of hypervolemia, not a sign of lack of uid responsiveness. Loop diuretics were frequently used to achieve a negative uid balance, together with uid restriction, and renal replacement therapy when indicated. Patients were re-assessed daily until becoming hemodynamically stable.
Inclusion criteria: a successive cohort of patients admitted to ICU were included.
Exclusion criteria: patients died or discharged against medical advice within 10 hrs after admission or referral to ICU; hemodynamically stable surgical patients, either emergent or elective, referred to EICU for anesthesia recovery only, experiencing an uneventful recovery, and with an ICU stay shorter than 3 days. These surgical patients were referred to ICU simply because there was no anesthesia recovery room. All patients with an ICU stay <`10 hrs had a positive uid balance.
Outcome measures: The primary endpoint was in-ICU mortality. The survivors were referred to medical or surgical ward. Non-survivors included patients brought home on hospice, to be with the family before death. It was stipulated by local authority that patients died in hospitals had to be cremated. Some patients were brought home before death because they wanted to be buried, not cremated. They died on the way home or immediately back home. Non-survivors also included patients referred to Sichuan Provincial Hospital and died within 72 hrs after referral. The secondary endpoints were length of ICU stay (LOS), length of ventilator support (invasive and/or non-invasive).

APACHE (Acute Physiology and Chronic Health Evaluation) II and SOFA (Sequential Organ Failure
Assessment) scores were calculated by the corresponding author (YC) alone, using data of the rst 24 hrs after admission to EICU. For patients died within 24 hrs after admission, data at least 30 min prior to death were taken into consideration.
Gastrointestinal loss and third space loss which was drawn with a needle and syringe or percutaneously drained were counted as output. Insensible losses were not counted.
Fluid balance=intravenous uid+oral intake-urine output-gastrointestinal loss-third space drainage.

Statistical analysis
These data were analyzed with IBM SPSS Statistics 20. APACHE II and SOFA scores of different groups were compared with Student's t-test. In-ICU mortalities of different groups were compared with chi-square test and Fisher's exact test. Fluid balances of cases enrolled before application of point-of-care ultrasound were compared with that of cases enrolled after application of ultrasound using Student's ttest. Amounts of intravenously-administered furosemide before application of point-of-care ultrasound were compared with that after application of ultrasound using Mann-Whitney U test. Serum brain natriuretic peptide (BNP) of survivors was compared with that of non-survivors using Mann-Whitney U test. Logistic regression analysis was conducted to nd out the risk factors of in-ICU mortality among APACHE II score, SOFA score, day 1 uid balance, day 2 cumulative uid balance, day 3 cumulative uid balance and cumulative uid balance during ICU stay. Logistic regression analysis did not report odds ratio and con dence interval when uid balances were in ml. Then logistic regression analysis was done after uid balances were transformed into -Ls (liters). Before logistic regression analysis, collinearity diagnosis was done for APACHE II score and SOFA score on admission with linear regression analysis. Survival curve (Cox regression model) and ROC (receiver operating characteristic) curve were plotted with IBM SPSS Statistics 20. Length of ventilator support and LOS of cases before application of ultrasound were compared with that of cases after application of ultrasound using Mann-Whitney U test.
Abbreviations ICU, intensive care unit; APACHE, Acute Physiology and Chronic Health Evaluation; SOFA, Sequential Organ Failure Assessment; CVP, central venous pressure; LOS, length of ICU stay; BNP, brain natriuretic peptide; ROC, receiver operating characteristic; AUC, area under curve.

Declarations
Ethnics approval and consent to participate The ethical committee at Jinniu Hospital approved the study protocol. The informed consents were obtained from the patients or their close relatives.

Consent for publication
Not applicable.

Competing interests
The authors declare that they have no competing interests.

Availability of data and material
The datasets during and/or analyzed during the current study are available from the corresponding author on request.
Authors' contributions HH and DL took part in data collection and analysis. ZM conducted ultrasound examination. CJ took part in data collection. CY took part in data collection and analysis, and wrote the manuscript. All authors read and approved the nal manuscript.

Funding
Dr Y Chen was supported by a grant from Sichuan Provincial Health Administration (416001004078008).   Amounts of furosemide intravenously administered. *Compared with Mann-Whitney U test. In-ICU mortalities. *Compared with chi-square test (p=0.036) and Fisher's exact test (p=0.038). For cases with positive vs. negative cumulative uid balance during ICU stay, p<0.001 by both chi-square test and Fisher's exact test.

Figure 5
Multiple logistic regression analysis showing independent risk factors of in-ICU mortality. *OR for per liter of uid. Studentized residuals≦0.971.

Figure 6
Receiver operating characteristic (ROC) curves showing the relationship between sensitivity and 1speci city in determining APACHE II and SOFA scores, cumulative uid balance during ICU stay, and uid balances of the rst three days to predict in-ICU mortality.