An observational study of intensivists' expectations and effects of fluid boluses in critically ill patients.

BACKGROUND: Fluid bolus therapy (FBT) is common in ICUs but whether it achieves the effects expected by intensivists remains uncertain. We aimed to describe intensivists' expectations and compare them to the actual physiological effects.
METHODS: We evaluated 77 patients in two ICUs (Sweden and Australia). We included patients prescribed a FBT ≥250 ml over ≤30 minutes. The intensivist completed a questionnaire on triggers for and expected responses to FBT. We compared expected with actual values at FBT completion and after one hour.
RESULTS: Median bolus size (IQR) was 300 ml (250-500) given over a median (IQR) of 21 minutes (15-30 mins). Boluses were 57% Ringer´s Acetate and 43% albumin (40-50g/L). Hypotension was the most common trigger (47%), followed by oliguria (21%). During FBT, 55% of patients received noradrenaline and 38% propofol. Intensivists expected a median MAP increase of 2.6 mmHg (IQR: -3.1 to +6.8) at end of bolus and of 1.3 mmHg (-3.5 to + 4.1) after one hour. Intensivist´s' expectations were judged to be accurate if they were within 5% above or below measured values. At FBT completion, 33% of MAP expectations were overestimations and 42% were underestimations. One hour later, 19% were overestimations and 43% were underestimations. Only 8% of expectations of measured urine output (UO) were accurate and 44% were overestimations. Correction for sedation or vasopressors did not modify these findings.
CONCLUSIONS: The physiological expectations of intensivists after FBT carried a high risk of both over and underestimation. Since the physiological effect FBT was often small and did not meet clinical expectations, a reassessment of its rationale, effect, duration, and role appears justified.

Introduction

Fluid bolus therapy (FBT) is common in critically ill patients [1, 2]. Dynamic indices of preload (e.g., stroke volume variation, pulse pressure variation, vena cava variability) appear to have acceptable predictive values for the immediate post bolus response but, due to practical limitations, they are often not available. Instead, intensivists use a broad range of measures, such as hypotension, tachycardia, oliguria or lactate levels, to inform the decision to administer FBT [3-5]. The FENICE trial described indications for FBT in critically ill patients, with 59% of FBT initiated because of hypotension, followed by oliguria and lactate clearance. The only commonly used hemodynamic measure was central venous pressure (CVP) (25%), which has been shown to be low value in guiding fluid therapy [6]. There is also no clear consensus on which patients will respond to FBT or on the optimal rate of volume and infusion rate of a fluid bolus (FB) [3, 7–10]. Finally, how the decision-making is performed and what response ICU-practitioners expect from a FB has not been investigated [3, 8–10]. As a positive fluid balance is associated with increased mortality in intensive care unit (ICU) patients [7, 11], it appears desirable to study current practice patterns for FBT and what factor influence them so that unnecessary FBT could be avoided. The relationship between intensivists’ expectations and actual quantitative hemodynamic effects may help guide this process but has never been described.

As a new way of describing the practice and rationale for FBT, we aimed to compare the quantitative expectations of treating intensivists with the actual effects of FBT in critically ill patients. We aimed to do this in a pragmatic fashion in which treatment modalities and goals for FBT were left at the intensivist discretion.

Material and methods

Study design

This was a prospective, observational multi-center cohort study conducted in a Swedish tertiary center (Södersjukhuset) and an Australian university teaching hospital (Austin Hospital). Patients were included between May and September 2017 in Australia and in Sweden between October 2017 to October 2018 and March to April 2019 due to availability of technical equipment to extract data only during these timeframes limiting possible inclusions. The planned study population was 100 patients, but inclusion stopped at 77 since the equipment was not available to use for further study inclusions.

Ethical approval was obtained from the Ethical Review Board of Stockholm (Tomtebodavägen 18A Solna, Sweden), EPN 2017/1133-31 on 4/8/2017 and the Austin Health Human Research Ethics Committee (145 Studley Road Heidelberg Victoria, Australia), LNR/17/Austin/94 on 17/5/2017. The study was registered on Clinicaltrials.gov, NCT03178578. The need for informed consent was waived with permission from the Ethical Review Board. Consent for publication was not applicable.

Inclusion criteria: Admission to the ICU and age 18 years or older. The clinician has prescribed a FB of ≥250 ml in ≤30 minutes. Exclusion criteria: Patients in whom death was considered imminent (within 24 hours) and/or if the treating intensivist declined to participate.

Treatment

When the treating intensivist chose to give a FB (defined as 250 mL or more of fluids over 30 minutes or less, according to intensivist’s choice), the patient was included. The decision to initiate FBT, the volume, rate of infusion and type of fluid was left entirely to the discretion of the treating physician. No specific monitoring or test of responsiveness was mandated but was left to the intensivist’s discretion, as this was part of the study outcome. The intensivist then completed a questionnaire (see S1 Appendix) and selected a primary and a secondary FBT trigger from a pre-specified list: hypotension, tachycardia, oliguria, low CVP, high lactate levels, low mixed venous oxygen saturation (SvO2) or central venous oxygen saturation (ScVO2) and low cardiac index (CI).

The questionnaire also asked the intensivist to specify what changes in these variables they expected at completion of the bolus and one hour later. Only the expectations and actual results of the first bolus (at any time in ICU-course) after inclusion were measured and any further boluses given to the same patient were not considered. Patients received all care at the discretion of the treating intensivist, including all measurements and monitoring.

Monitoring

Data was extracted from the medical information systems Clinisoft® (GE Healthcare, Barringgton, Illinois, USA, TakeCare® (CompuGroup Medical, Koblenz, Germany) or PowerChart® (Cerner Corporation, Kansas City, Missouri, USA) regarding mean arterial pressure (MAP), heart rate (HR), CI, CVP, urine output per hour (UO) and laboratory values such as lactate, SvO2 or ScVO2 and creatinine for comparison. Monitoring equipment was IntelliVue MP70® and IntelliVue MX800® (Philips Healthcare, Eindhoven, Netherlands). Monitoring and laboratory values were ordered at the intensivist’s discretion and per departmental protocol, meaning that there was no mandated measurement of any advanced hemodynamic values, and these would be used as clinically indicated. Departmental standards for blood pressure measurement included intraarterial measure via arterial line (Merit Medical, South Jordan, UT, USA and ITL Biomedical, Mulgrave, Australia) and for cardiac output measures include calibrated thermodilution PiCCO® (Pulsion Medical Systems, Feldkirchen, Germany) and pulmonary artery catheter (Edwards Lifesciences, Irvine, CA, USA). Measurements were collected for 5 minutes before the onset of the fluid bolus as a baseline and until 60 min after completion of the fluid bolus. Any infusions of vasoactive medications, sedatives and diuretics given were also recorded and accounted for as confounders for hemodynamic changes. Any further fluids administered during the study period were registered and accounted for as confounders. Fluids were recorded as colloids, crystalloids, or maintenance fluids.

Outcomes

Primary outcome

Accuracy of the intensivist’s expectations on the physiological effect of FBT at completion of the FB.

Secondary outcomes

To describe the accuracy of the intensivist’s expectations of the physiological effects of FBT one hour after the completion of the FB. The trigger for the FBT and changes in CO, HR, MAP, CVP, and lactate, UO, ScvO2 or SvO2.

Statistical analysis

We studied a convenience sample, planned to consist of 100 patients but resulting in 77 patients. Normality was tested using the Shapiro-Wilk test. Categorical data was presented as fractions and percentages and continuous data was presented as means with SD or median with IQR, and 95% CI depending on underlying distribution. Linear regression models were performed to determine relationship between the levels of change in sedation and vasoactive medications and on each of the variables used by the clinicians, and a correction was then made for the relevant levels of sedation and vasoactive medications. Imputation for missing data was not performed as the datasets were largely complete.

Intensivist´s’ expectations were judged to be accurate if they were within 5% above or below measured values, as the authors judged this to be a clinically relevant interval. For purposes of defining clinical effectiveness and whether expectations were under- or overestimations, MAP, CO, CVP, ScvO2 and UO were expected to increase after FBT, and HR and lactate to decrease.

Bland-Altman plots were used to describe the relationship between measured and expected values. We performed unplanned subgroup analyses on the patients where the indication for a FB was hypotension, to explore reasons for the unexpectedly low expectations in this group. An alpha value of 0.05 was considered statistically significant.

Statistical analysis was performed using IBM SPSS Statistics® version 25 for Windows (IBM Co., Armonk, NY, USA.

Results

We studied 77 patients, 18 in Australia and 59 in Sweden (see for study flow chart). Of these, 46 were male, with a median age of 68 years and a median SOFA score of 8 (). 24 patients were monitored with invasive hemodynamic monitoring (other than invasive blood pressure monitoring), and 44 patients were receiving mechanical ventilation. Bolus fluid was either Ringers Acetate (57%) and albumin 40-50g/L (43%), with a median (IQR) volume of 300ml (250 to 500) ml and a median (IQR) duration of administration time of 22 minutes (15 to 30 mins). 42% of participating intensivists were ICU consultants, 15% were fellows and 43% were registrars.

Study flow chart.

Consort flow chart of patients screened.

Expectations of MAP were accurate in 25% of cases at FBT completion and in 37% of cases after one hour. At FBT completion, 33% of MAP expectations were overestimations and 42% were underestimations. A scatter plot of the relationship between expected and measured MAP at fluid bolus completion by main reason for bolus is illustrated in . One hour later, 19% were overestimation and 43% were underestimations. For HR, expectations were accurate in 52% of cases at the end of the bolus and 39% of cases after one hour. In 38% of cases, expectations overestimated the effect at FBT completion, and in 39% after one hour. In the case of UO, expectations were accurate in only 8% of cases after one hour, 44% were overestimation and 48% were underestimations (). Scatter plots of the relationship between expected and measured UO by main reason for bolus at fluid bolus completion is illustrated in .

Scatter plot of measured and expected MAP after fluid bolus.

Scatter plot comparing measured and expected MAP after fluid bolus. Line represents perfect fit. MAP = Mean arterial pressure.

Scatter plot of measured and expected UO one hour after fluid bolus.

Scatter plot comparing measured and expected UO one hour after fluid bolus. Line represents perfect fit. UO = Urine output.

Intensivists expected a median MAP increase of 2.6 mmHg (IQR: -3.1 to +6.8 mmHg) at end of the FB and of 1.3 mmHg (-3.5 to + 4.1 mmHg) after one hour. Expectations for CI were a median (IQR) increase of 0.06 L/min/m2 (-0.05 to +0. L/min/m2) at bolus completion and a median increase of 0.00 L/min/m2 (-0.12 to +0.20 L/min/m2) after one hour. Expectations for HR were a median (IQR) decrease of -4.2 bpm (-11.0 to 0.0 bpm) at bolus completion and a median decrease of -3.7 bpm (-9.6 to +0.2 bpm) after one hour. Expectations are further described in . Expectations for the primary reason for FBT are described in S1 Table in S1 Appendix. Boxplots of the relationship between expected and measured MAP as well as UO by main indication for bolus are displayed in Figs .

Boxplot of difference between measured and expected MAP by indication for bolus.

Boxplot showing difference between expected and measured MAP by main reason for administration of fluid bolus. Positive values indicate a higher value for expected MAP compared to measured, and the inverse for negative values. MAP = Mean arterial pressure. CVP = Central venous pressure. ScvO2 = Central venous oxygen saturation. ScvO2 = Mixed venous saturation.

Boxplot of difference between measured and expected UO by indication for bolus.

Boxplot showing difference between expected and measured MAP by main reason for administration of fluid bolus. Positive values indicate a higher value for expected UO compared to measured, and the inverse for negative values. UO = Urine output. CVP = Central venous pressure. ScvO2 = Central venous oxygen saturation. ScvO2 = Mixed venous saturation.

Among patients where the indication for FBT was hypotension, 58% were on a noradrenaline (NA) infusion, at bolus completion, however, the MAP expectation for patients without NA was 1.5 (-3.0 to +6.2) vs. 3.3 (-3.9 to +10.0) with NA. In detail, for 5/36 of these patients, the intensivist had very small expectations of MAP increment (expected change between -2 and -2mmHg). The baseline MAP for these five patients was 69.0 (60.3 to 74.5) mmHg. Two of these had infusions of NA, and levels of NA was unchanged after FBT.

The most common primary trigger for FBT was hypotension (47%), followed by low urine output (21%). The most common secondary triggers were low urine output and tachycardia (both 17%) ().

For the primary reason for bolus administration, the estimation was accurate in 22% of cases at FBT completion and 47% were overestimations. After one hour, the effect estimation for the primary reason for bolus administration was accurate in 29% of cases and 31% were overestimations. For the secondary reason, accuracy was 20% at FBT completion, with 43% being overestimations. After one hour the estimation for the secondary reason was accurate in 22% of cases, with 31% being overestimations.

Scatter plots for MAP one hour after FB and HR at both time points are presented in S1–S3 Figs in S1 Appendix. All show limited correlation. Bland-Altman plots of bias and limits of agreement, between measured and expected MAP and HR after FB and after one hour as well as UO after one hour are displayed in S4–S8 Figs in S1 Appendix.

The median (IQR) MAP change after a fluid bolus was 4.4 (-2.3 to +9.7) mmHg and, after one hour, 2.5 (0.7 to 6.6) mmHg. The median (IQR) HR response after a fluid bolus was -1.0 (-5.4 to +1.5) bpm and after one hour -0.2 (-5.2 to +2.9) bpm. The median (IQR) UO response to fluids after one hour was -15.0 (-60.0 to +15.0) ml (see ).

Median (IQR) volume of other fluids was 70 (16 to 113) ml. In addition, 3% of patients were on CRRT and 5% of patients received diuretics within one hour after the fluid bolus. Two patients received boluses of 40 mg and 10 mg Furosemide respectively, one patient had an infusion of 17mg/h continuously during the period and one patient had an infusion of 20 mg/h started. Median (IQR) change in NA was 0.00 (0.00 to 0.00) after the bolus and 0.00 (-0.01 to +0.01) one hour after the bolus, which could otherwise have been a confounder for evaluating response regarding MAP. Six patients were treated with an inotrope during the study period (Four had infusions of milrinone, one dobutamine and one adrenaline), without any dose changes.

The median difference between expectations and outcome is presented in S2 Table in S1 Appendix. All values after correction for levels of NA and propofol are presented in S3 Table in S1 Appendix as they all were either not statistically significant or did not differ from the uncorrected values. ScvO2 or SvO2 was only measured in 3 patients and the sample size was therefore too small to make adequate analyses.

Discussion

Key findings

In this study, the most common triggers for FBT were hypotension, oliguria, and tachycardia. However, in only 25% of cases were expectations of MAP met, underestimation of effect occurred for 42% of boluses, and overestimation for 33%. The clinical expectations of intensivists in relation to urinary output were not met in >90% of cases. These results did not change significantly after correction for administered sedation and vasopressors. Finally, the predictive accuracy of the intensivists’ expectations both directly after FBT and at the follow-up one hour later was low. Regardless of accuracy, stated expectations were generally surprisingly small, yet FBT therapy was still initiated.

Study implications

To our knowledge this is the first study to compare intensivists’ quantitative expectations with the actual response to a fluid bolus (FB). Our findings imply that intensivists’ expectations of FBT effects are often inaccurate for MAP and markedly so for UO and that, for these two variables, both overestimation and underestimation of effect are common. Intensivists’ expectations generally correlate poorly or very poorly with the actual effect. This emphasizes the difficulty in predicting response to fluids, and that many FB interventions will be given without achieving the intended goal. Finally, our study suggests that, as in other studies, the most common triggers for FBT in the ICU remain hypotension, oliguria, and tachycardia.

Expectations were surprisingly small, and could even be negative for several variables, including MAP, the most common indication for a FB. There might be several explanations for this. If the patient is unstable, merely avoiding a further decrease might seem like a reasonable expectation. For a hypotensive patient on vasopressors, expectation of a MAP-response to a FB might manifest as a reduced dose of NA instead of a higher MAP, as this will be titrated down. However, vasopressor doses were not in fact decreased, and the expectations for MAP among patients receiving NA were higher rather than lower.

Observational bias might also be a factor, as knowledge of being monitored might cause intensivists to lower their stated expectations, for fear that they might be found markedly inaccurate [12]. Another explanation might be that FBT is a norm in the ICU environment and intensivists feel there is peer pressure to use this therapy even in cases where their clinical judgment is that the response will be small. Another reason that intensivists feel comfortable with small expectations might be that, at the bedside, a FB might seem like a harmless intervention, one that might be administered without much consequence. If this is the case it might explain some of the FBT expectations. However, such thinking would be problematic in light of the connection between positive fluid balance and long-term morbidity and mortality [7, 11]. More research is needed to unravel such aspects of the psychology of decision-making leading to FBT prescription.

Relationship with previous studies

Fluid responsiveness has been extensively studied and is a well-accepted concept. However, the number of patients that are fluid responders in the ICU remains low. Also, measurements of CO and SV are often not available in patients who are considered in need of FBT. In this study, we have therefore focused on the goals that intensivists themselves have chosen as target, an assessment that better describe if clinically relevant goals are met, as it takes intensivists’ evaluation and judgement into account.

To our knowledge, no previous studies have investigated the accuracy of expected treatment response to FBT. Unfortunately, while the proportion of fluid responders has been described as 50% in the ICU, the accuracy of predicting this response after FBT was only in the range of 8–52% depending on the parameter chosen. This is disappointing but aligned with other areas of clinical practice. For example, palliative care doctors predicted patient survival inaccurately in 34% of cases and overestimate survival time in 51% of cases [13].

As in the FENICE trial, MAP was the commonest trigger for FBT [9] in our cohort, and the use of more advanced parameters for guiding FBT remained low. This shows the continuing problem of using inaccurate parameters to predict and measure response to fluid administration. However, in the FENICE study, CVP was commonly used [9], while in our population, CVP was used in only 3% of cases. This likely reflects changes in practice due to evidence against using CVP as a marker for fluid administration [3, 6, 14]. Also, in our study, compared to both FENICE and international surveys, bolus volume was both slightly smaller and administered over longer period [8, 9].

Study strengths and limitations

The strengths of our study include the pragmatic and novel study design. Another strength of our study is that we corrected statistically for changes in vasopressors and sedatives to exclude them as potential confounders. Also, our study recruited from two ICUs in two different countries, both a university centre and a tertiary centre. This reduces the risk that local treatment algorithms and traditions influenced the results and adds external validity. Moreover, we evaluated all effects for up to one hour after FB, as previous trials have shown a decrease in the effect of fluid in this timeframe [15, 16]. Our study also included evaluation of UO, which is one of the most commonly used triggers for guiding FBT and an important clinical marker for renal function.

A limitation of our study is that the sample is a convenience sample and relatively small. Also, we had originally planned for a sample of 100 patients, but due to issues with the equipment used to extract the monitoring data we unfortunately had to end the study at 77 patients. However, the findings are clear and unlikely to be materially altered by a larger sample.

Compared to some previous trials defining a fluid bolus and the standard definition of fluid responsiveness, the mean bolus size in this group was smaller, and given rather slowly (300 ml given over 21 minutes). This might affect the absolute size of the fluid response and would perhaps mean that fewer patients would have a rise in SV or CO > 10%. However, the intensivists chose both the fluid volume and rate of infusion and had this in mind when considering their expectations. Thus, their predictive accuracy is adjusted to that bolus size and reflects clinical practice. We described accuracy both as the percentage of expectations that were within the interval of +/-5%, as well as by mean bias in standard Bland-Altman plots with wider limits of agreement. These results were slightly divergent, probably due to a larger effect of outliers on the mean bias in the Bland-Altman plots. However, both methods show that it is difficult to estimate what the effect of a FB on MAP, HR and UO will be.

Conclusion

Hypotension was the most common indication for giving fluids, yet MAP expectations were met in only 25% of cases. The other common reasons for fluid boluses were poor urine output and tachycardia, with expectation met even less frequently. The clinical expectations of intensivists were not met to a substantial extent and showed poor accuracy at the end of the bolus and after one hour. Since the clinical effect is often small and does not meet clinical expectations, a reassessment of the rationale, effect, duration, and role of FBT in ICU patients appears justified.

STROBE statement—checklist of items that should be included in reports of observational studies.

(DOCX)

Click here for additional data file.

(DOCX)

Click here for additional data file.

(DOCX)

Click here for additional data file.

30 Dec 2021

31 Jan 2022

Answers to reviewers' comments for authors regarding the manuscript “An observational study of intensivists’ expectations and effects of fluid boluses in critically ill patients”

Dear Editor,

We thank you and the referees for the comments regarding our paper with the above reference number and we are very grateful that you give us the possibility to revise and improve the manuscript. We have closely looked at the reviewer’s comments and will provide answers point-by-point.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

We have conformed to the style requirements to the best of our ability, and would appreciate further instructions if we are still not fulfilling them.

2. We note that you have indicated that data from this study are available upon request. PLOS only allows data to be available upon request if there are legal or ethical restrictions on sharing data publicly. For more information on unacceptable data access restrictions, please see http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions.

The datasets from this study are not completely anonymized and cannot be shared openly, per the ethical approval obtained at the Stockholm and Melbourne ethical boards. We understand this is not ideal but the after consultation with the legal staff of both the hospital and the university we are not allowed to share a “minimal” dataset either as this would still be considered personal data since it would contain data such as age, gender, date of admission to the hospital and ICU as well as type of surgery, which would be sensitive information.

Requests for access to the data should be made to the Research Data Office at Karolinska Institutet via rdo@ki.se and if permitted by law and ethical approval, decided on a case by case basis, the data can shared.

In your revised cover letter, please address the following prompts:

a) If there are ethical or legal restrictions on sharing a de-identified data set, please explain them in detail (e.g., data contain potentially sensitive information, data are owned by a third-party organization, etc.) and who has imposed them (e.g., an ethics committee). Please also provide contact information for a data access committee, ethics committee, or other institutional body to which data requests may be sent.

Please see above.

b) If there are no restrictions, please upload the minimal anonymized data set necessary to replicate your study findings as either Supporting Information files or to a stable, public repository and provide us with the relevant URLs, DOIs, or accession numbers. For a list of acceptable repositories, please see http://journals.plos.org/plosone/s/data-availability#loc-recommended-repositories.

We will update your Data Availability statement on your behalf to reflect the information you provide.

Please see above, restrictions apply

3. Your ethics statement should only appear in the Methods section of your manuscript. If your ethics statement is written in any section besides the Methods, please move it to the Methods section and delete it from any other section. Please ensure that your ethics statement is included in your manuscript, as the ethics statement entered into the online submission form will not be published alongside your manuscript.

The ethics statement has been moved from Declarations to Methods.

4. Please include captions for your Supporting Information files at the end of your manuscript, and update any in-text citations to match accordingly. Please see our Supporting Information guidelines for more information: http://journals.plos.org/plosone/s/supporting-information.

Captions have been added for the Supporting Information.

5. Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript. If you need to cite a retracted article, indicate the article’s retracted status in the References list and also include a citation and full reference for the retraction notice.

We have reviewed our list of references and not found any instances of papers that have been retracted and no replacements have been made.

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: The authors conducted a prospective, observational multi-center cohort study to investigate, whether the quantitative expectations of treating intensivists coincide with the actual observed effects of FBT in critically ill patients. The authors included 77 patients.

They concluded that physiological expectations of intensivists after FBT carried a high risk of both over and underestimation, so agreement seem to be rather weak.

The data and presentation are suitable, to support the conclusion. The analysis of the data is by appropriate sound scientific descriptive methods.

I do have some minor comment, which the authors may consider.

The Bland Altman approach is suitable here, but leaves the reader with the two agreement axis accuracy and precision. Lin's concordance coefficient would summarize the two axis. There is also a version for repeated measures similar to Bland Altman. May be the authors would like to consider this measure, with 95% CI.

While the Bland-Altman model is not perfect it is useful for this type of data, and while presenting both accuracy and precision is somewhat more complex also gives the reader a fuller understanding of the findings. We collaborated with a statistician regarding the statistical models and were advised against using a correlation model as the degree of correlation would not only reflect the accuracy but also the heterogeneity and spread of our sample population. Also, to our knowledge the Bland-Altman model is frequently used for presenting accuracy and precision and might be familiar to interpret for the reader, whereas Lin´s concordance coefficient might be less so.

L29: 77 patients are enrolled, but 100 are planned in the protocol. It would be informative to discuss this difference. (See patient flow chart later)

We agree and have added this to the Methods section (page 6 “We studied a convenience sample, planned to consist of 100 patients but resulting in 77 patients”) and Discussion in the Study strengths and limitations section (page 17, “Also, we had originally planned for a sample of 100 patients, but due to issues with the equipment used to extract the monitoring data we unfortunately had to end the study at 77 patients”).

L39: The authors define the cutoff by 5% above or below the measured values. The problem with this definition is, that there will be almost always observations outside the 5% margin defined by the empirical distribution. Although Bland Altman's approach is data driven as well the cutoff by SD seem to be more robust. Could you please clarify the definition?

We agree that defining this cutoff is difficult and problematic. The +-5% we arrived at was what we judged to be clinically relevant, even if this of course is arbitrary. Our consideration regarding the distribution was that the cutoff had to be kept narrow since the values for the parameters are highly correlated. Ie, the blood pressure of an individual patient after fluid bolus is highly correlated to the blood pressure before fluid bolus and as is seen in our material changes are likely to be slight. Therefore, using a larger cutoff might instead make it very likely for any reasonable estimation at all to be correct, which also would taint the results

L126: Some more details are necessary to describe the constitution of the "convenience sample" so that the effect of possible selection biases could be assessed.

A description of the convenience sample has been added to Methods, please see above. Further details on patient selection are also found in the Flow Chart Fig.1

L130: skip "in"

Fixed

L132: may be "included in the analysis

Fixed

(Comment: As this is rather a descriptive explorative analysis it suffices to describe the missing pattern. In principle multiple imputation techniques can be used to mitigate attrition bias here.)

Excluded patients are presented in Fig.1, as for missing data our set was largely complete (Two datapoints missing for MAP at bolus end and after one hour for instance), and we opted against using imputation for the few missing values that we had.

L 145f: Add a figure illustrating patient flow similar to consort flow chart.

We´re uncertain if there has been some omission of Figures, but Fig.1 should represent a Consort patient flow chart

Reviewer #2: This multi centre observational study compares the expectations of intensivists after a fluid bolus with the actual observed effects in a mixed surgical and medical intensive care population

The manuscript is clearly written, and the design is novel and interesting, and the question important.

Comments;

-The main problem is that the very low expectations of the intensivists means it is difficult to draw conclusions about them being under or over estimations. A mean MAP difference of 2.6mmhHg is arguably not clinically meaningful. The expectations of change in the other measured variables are also very low. It begs the question - did the authors ask and measure the correct question? Eg other measures of perfusion such as capillary return (Hernandez. 2018. JAMA. 2019;321(7):654-664. doi:10.1001/jama.2019.0071). Plus, the 57% that were on NA were unlikely to have had a large MAP change - as this NA titration is usually used to maintain a MAP. This would more likely be manifested as a drop in NA as the authors point out. Although this is acknowledged by the authors, it still limits the validity of their results.

We agree that the low expectations and effects of the bolus are surprising, however to our mind these are some of the main findings of the study, and not something we created by our study design. To our knowledge this has not been described in this fashion before and is in itself worth further study. Regarding the study question, we used a questionnaire with the parameters most commonly used by the clinicians at our institutions, and the indications were then chosen by the clinicans themselves. So, if they were not correct questions and measurements then they are not used in clinical practice in either of our institutions. Specifically, regarding capillary return, this measure was not used by the physicians in either setting, however it has place when no measurement of perfusion or output is used and could have been added to the questionnaire.

The use of pressors could indeed influence the change in blood pressure, and disguise any signal in the data as NA would be lowered to keep MAP in place. We collected changes in NA as a confounding variable. We have clarified this (page 13 “Median (IQR) change in NA was 0.00 (0.00 to 0.00) after the bolus and 0.00 (-0.01 to +0.01) one hour after the bolus, which would otherwise have been a confounder for evaluating response regarding MAP” and specifically for hypotensive patients at line 256 in Results levels of NA were not lower after FB than before.

-A median bolus of 300mls is rather small to expect large and clinically meaningful changes in physiology.

We agree that this may be true but this is in our opinion one of the major findings of the study. Both the size of the bolus and the expectations are chosen by the clinicians studied and not the trial investigators. We might find them small and be unsurprised by the lack of response, but these were the boluses actually used in these ICU:s and the expectations the physicians had for their response.

-What was the rate of missing data?

The missing patients are presented in Fig.1. In addition to this we only included 77 rather than 100 patients due to issues with availability of the programs to extract the data from the monitors which we have expanded on in the discussion. The rate of missing data for the individual variables was low but difficult to present as it depends on the type of variable. For instance, for urine output, there were hourly data points and a few were missing. For blood pressure, the dataset is very large as it was collected per 20-second to minute and then aggregated for the analysis, so there is very little data missing (Two datapoints for MAP at bolus end and after one hour for instance) except when there was a technical issue. Also, we chose not to consider the data as missing in the cases where clinicians did not chose to use a parameter such as for instance CO or ScvO2, as this was then by their own active choice.

Minor points

-Albumin - was the 4%? 20%? not clear

Albumin was either 40 or 50g/L, this is stated in the beginning of Results and in the relevant tables

Reviewer #3: The authors have conducted an observational study to assess the expected response from fluid bolus therapy (FBT) in patients admitted to ICU with critical medical and surgical conditions. Trigger points for FBT included commonly used hemodynamic parameters or markers of tissue perfusion such as HR, MAP, CI, lactate levels or urine output. The endpoints were increase of MAP and other commonly used hemodynamic parameters and urine output on conclusion of FBT and at one hour of completion. The authors need to be congratulated on questioning the rationale of a common scenario that is practiced universally to manage hypotension and low urine output states.

The conclusions from the study suggested that the actual response did not correlate with the expected response to MAP and urine output. For the primary reason for bolus administration (actual response), the estimation was accurate in 22% of cases at FBT completion and 47% were overestimations. After one hour, the effect estimation for the primary reason for bolus administration was accurate in 29% of cases and 31% were overestimations. For the secondary reason (predicted response), accuracy was 20% at FBT completion, with 43% being overestimations. After one hour the estimation for the secondary reason was accurate in 22% of cases, with 31% being overestimations. The effect was assessed considering the concomitant administration of vasopressor and sedation where applicable.

The authors propose that more work needs to be done in this area to define the role of FBT in ICU patients who have traditionally been treated with volume replacement with or without using CVP as marker of volume status.

The authors also suggest that a measure such as FBT is not benign as positive fluid balance in ICU settings in critical patients might correlate with adverse outcomes.

Strengths:

The study seems to be well designed with well identified goals

The concept is novel with aim to correlate the expected response obtained to FBT versus the actual effect

The statistical analysis seems to outline well the findings of the study

Limitations of the study:

Observational study in small group of patients

This is true and limits conclusions that can be drawn. Regarding the observational format though, as a main focus of our study is the clinicians expectations and goals, finding an ideal format would be difficult as this cannot be randomized or controlled

Does not identify the response status to patients’ cardiac and renal functions

We agree that cardiac and renal function are relevant to evaluate the expected response to fluid boluses. However, we chose to focus on the parameters commonly used to evaluate the response to fluid boluses. This included diuresis, which is commonly used as a marker for renal function and in this time-span of an hour to a couple of hours is to our knowledge the most commonly used marker for renal function. Regarding the cardiac function, several measurements relating to cardiac function and perfusion such as CO, ScvO2 and lactate were used. However, we did not include echocardiographic measurements of cardiac function. This could be a weakness, but these measurements would at best be bedside echocardiographies by the physician at the bedside. They would be of very varying quality and not performed in a standardized fashion and timeframe due to the observational nature of the study, making any analysis unfeasible.

Is there any difference in the response expected between colloids and crystalloids?

Considering the dataset consists of 77 patients already divided into sub-categories by reasons for fluid bolus administration, we chose not to further divide the groups to not make the groups any smaller than necessary and further weaken any possibility to draw conclusions.

Identification of difference between medical and surgical patients, as they would represent two different cohorts who might react differently to volume administration esp. in the setting of third space fluid losses (this could potentially happen in medical patients).

We agree that this is an interesting inquiry and that we would have liked to be able to compare the intensivists’ expectations in the two groups. However, our sample size is too small, especially of medical patients to allow any conclusions to be drawn from analysis.

However, despite all these reasons, the authors have presented a simple argument to define the role of FBT in critically ill patients in ICU My impression is that fluid responsiveness in critically ill patients is dependent on several factors that include patients’ cardiac/ renal status/ capillary leakiness etc. and future studies will need to direct their attention at comparable group of patients. I think the manuscript will improve in its value if some of these points are highlighted.

We wholeheartedly agree that these are all important factors for the fluid responsiveness of patients in the ICU setting, and that further studies are needed in the field. This trial however is more focused on the relationship between the physicians stated expectations and whether or not these goals are achieved, rather than if the patient in question is a fluid responder. To our knowledge this has been less extensively studied, and we feel that further study of what goes on at the bedside and what guides our decision making should be conducted as there is no consensus on what constituted optimal fluid therapy and there are clearly discrepancies between guidelines and actual practice.

8 Mar 2022

An observational study of intensivists’ expectations and effects of fluid boluses in critically ill patients

PONE-D-21-33377R1

Dear Dr. Wall,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Jaishankar Raman, MBBS, MMed, FRACS, PhD

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

16 Mar 2022

PONE-D-21-33377R1

An observational study of intensivists’ expectations and effects of fluid boluses in critically ill patients

Dear Dr. Wall:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Jaishankar Raman

Academic Editor

PLOS ONE

Table 1

Patient characteristics.

Age (years)	68 (58–78)
Male	46/77 (60%)
Height (cm)	171 (166–178)
Weight (kg)	76 (63–88)
SOFA	8 (6–10)
Comorbidities
Atrial fibrillation	9/77 (12%)
COPD	14/77 (18%)
Chronic kidney disease	6/77 (8%)
Diabetes	14/77 (18%)
Hypertension	34/77 (44%)
Ischemic heart disease	23/77 (30%)
Congestive heart failure	5/77 (6%)
Smoking	15/77 (20%)
Acute admission	57/77 (74%)
Surgical admission	57/77 (74%)
Type of surgery
Thoracic	19/77 (25%)
Abdominal	18/77 (23%)
Orthopedic	4/77 (5%)
Vascular	3/77 (4%)
Other	3/77 (4%)
Surgical admission that did not undergo surgery	30/77 (39%)
Noradrenaline infusion	42/77 (54%)
Noradrenaline dose baseline (Δg/kg/min)	0.08 (0.03–0.15)
Propofol infusion	29/77 (38%)
Propofol dose baseline (mg/kg/h)	1.33 (1.00–1.84)

Values are presented as median with (IQR) or numbers (percentages) of patients.

SOFA = sequential organ failure assessment score. COPD = Chronic obstructive pulmonary disease.

Table 2

Accuracy of expectations.

Parameter	After bolus			One hour after bolus
	Correct estimates (%)	Overestimation of effect	Underestimation of effect	Correct estimates (%)	Overestimation of effect	Underestimation of effect
MAP (mmHg)	18/72 (25%)	24/72 (33%)	30/72 (42%)	27/72 (38%)	14/72 (19%)	31/72 (43%)
UO (ml)	N/A	N/A	N/A	5/61 (8%)	27/61 (44%)	29/61 (48%)
HR (bpm)	38/73 (52%)	28/73 (38%)	7/73 (10%)	29/74 (39%)	29/74 (39%)	16/74 (22%)
CI (L/min/m2)	6/23 (26%)	12/23 (52%)	5/23 (22%)	4/22 (18%)	6/22 (27%)	12/22 (54%)
Lactate (mmol/L)	2/22 (9%)	14/22 (64%)	6/22 (27%)	2/33 (6%)	18/33 (54%)	13/33 (39%)
CVP (mmHg)	5/21 (24%)	3/21 (14%)	13/21 (62%)	3/21 (14%)	8/21 (38%)	10/21 (48%)
ScvO2 (%)	2/3 (67%)	0/3 (0%)	1/3 (33%)	1/3 (33%)	2/3 (67%)	0/3 (0%)

Values are presented as numbers (percentages) of patients.

MAP = Mean arterial pressure. UO = Urine output. HR = Heart rate. CI = Cardiac index. CVP = Central venous pressure. ScvO2 = Central venous oxygen saturation. Overestimation is defined as an estimate greater than measured value for MAP, UO, CI, CVP and ScvO2, and as a measured value greater than estimated value for HR and lactate.

Table 3

Triggers for FBT and physiological expectations.

Reasons for bolus	Main reason, N (%)	Secondary reason, N (%)	Expectation, median (IQR)
			After bolus	One hour after bolus
Hypotension (mmHg)	36/77 (47%)	12/77 (16%)	2.6 (-3.1–6.8)	1.3 (-3.5–4.1)
Poor urine output (ml)	16/77 (21%)	13/77 (17%)	N/A	0.0 (-47.5–20.0)
Tachycardia (bpm)	14/77 (18%)	13/77 (17%)	-4.2 (-11.0–0.0)	-3.7 (-9.6–0.2)
Low cardiac index (L/min/m2)	5/77 (6%)	5/77 (6%)	0.06 (-0.05–0.31)	0.00 (-0.12–0.20)
High lactate (mmol/L)	4/77 (5%)	6/77 (8%)	0.0 (-0.2–0.3)	-0.2 (-0.4–0.2)
Low CVP (mmHg)	2/77 (3%)	3/77 (4%)	0.5 (-0.7–2.4)	0.0 (-1.1–1.6)
Low SvO2/ ScvO2 (%)	0/77 (0%)	1/77 (1%)	Too small sample size	Too small sample size
None given	N/A	24/77 (31%)	N/A	N/A

Values are presented as median with (IQR) or numbers (percentages) of patients.

MAP = Mean arterial pressure. UO = Urine output. HR = Heart rate. CI = Cardiac index. CVP = Central venous pressure. ScvO2 = Central venous oxygen saturation. N/A = Not applicable.

Table 4

Effects of the fluid bolus.

Parameter	Baseline	Difference from baseline after bolus, median (IQR)	Difference from baseline after 1 hour, median (IQR)
MAP (mmHg)	69.7 (64.2–78.0)	4.4 (-2.3–9.7)	2.5 (-2.2–10.0)
N baseline = 77
N after bolus = 75
N after 1 hour = 75
UO (ml/h)	60.0 (25.0–150.0)	N/A	-15.0 (-60.0–15.0)
N baseline = 73
N after 1 hour = 71
HR (bpm)	92.9 (79.0–111.0)	-1.0 (-5.4–1.5)	-0.2 (-5.2–2.9)
N baseline = 77
N after bolus = 76
N after 1 hour = 77
CI (L/min/m2)	2.68 (2.14–3.10)	0.00 (-0.13–0.09)	0.17 (-0.12–0.37)
N baseline = 24
N after bolus = 23
N after 1 hour = 22
Lactate (mmol/L)	1.3 (0.9–1.9)	0.0 (-0.3–0.1)	-0.1 (-0.3–0.2)
N baseline = 68
N after bolus = 28
N after 1 hour = 48
CVP (mmHg)	10.3 (8.0–12.9)	3.1 (1.5–5.2)	0.5 (-1.6–1.7)
N baseline = 23
N after bolus = 21
N after 1 hour = 23
ScvO2 (%)	65.0 (62.2–71.1)	Too small sample size	Too small sample size
N baseline = 4
N after bolus = 3
N after 1 hour = 3

Values are presented as median with (IQR).

MAP = Mean arterial pressure. UO = Urine output. HR = Heart rate. CI = Cardiac index. CVP = Central venous pressure. ScvO2 = Central venous oxygen saturation. N/A = Not applicable.