Anyone who has spent some time in a cardiac intensive care unit understands the physiologic appeal of the intra-aortic balloon pump (IABP). Anecdotally its use improves multiple clinical endpoints that for years have been considered surrogates for patient important outcomes. And yet, despite these physiologic advantages when examined in a rigorous fashion, IABPs have failed to demonstrate any advantages in patient survival (1). This is not the first time that an elegant physiological principle has failed to lead to its expected clinical conclusions. It is with this in mind that we turn our eye to fluid responsiveness, the current darling of resuscitative strategies in patients presenting in various states of hemodynamic compromise.
Like the IABP its physiologic appeal makes fluid responsiveness an attractive tool to use in the Emergency Department. But techniques to identify the sweet spot on the Frank-Starling curve such as passive leg raise (PLR) are arduous, time consuming, and do not lend themselves to chaotic nature of the Emergency Department. As such some have proposed respiratory variation of the inferior vena cava (IVC) on bedside ultrasound (US) as an adequate surrogate to these more complex strategies to guide fluid administration. The following is a discussion of these tools and the data supporting their use. In a recent piece on iSepsis, Dr. Paul Marik provides an excellent review of the literature and the evidence against the use of IVC US. This post will serve as an alternative perspective.
In a recent article published by the the Journal of Critical Care, Corl et al (2) examined both the use of IVC variation and passive leg raise (PLR) to predict fluid responsiveness (defined as > 10% increase in cardiac index as measured by NICOM). The authors enrolled a convenience sample of spontaneous breathing patients being admitted to the ICU with signs of circulatory failure (SBP<90 mmHg, or MAP< 65 mmHg for >30 min, urine output <0.5 ml/kg/h, heart rate > 120 bpm for >30 min, and/or serum pH < 7.3 or lactic acid > 2 meq/l).
The authors enrolled 124 patients over a 2-year period. Overall, 49.2% of the cohort were considered fluid responsive. In this cohort a change in IVC diameter of > 25% optimally predicted responsiveness to a 500 cc fluid bolus with a sensitivity, specificity, LR+, LR- and AUROC of 87%, 81%, 4.56, 0.16, and 0.84, respectively. This was compared to the PLR which significantly underperformed when compared to previous studies with an AUROC of 0.68.
Before discussing the clinical applicability of these results there are a number of methodological issues to discuss.
First the 25% cutoff was not determined in a prospective fashion. What this means is after the authors obtained the results of the study they went back and identified the threshold that demonstrated the best diagnostic accuracy (in this case 24.6%). It is unlikely that this specific threshold would demonstrate the same diagnostic prowess when tested in an external cohort. Ideally a follow-up study prospectively validating this threshold would be required before using this cutoff clinically. Interesting to note, the threshold used to determine fluid responsiveness using the PLR (increase in CO of 10%), was determined prospectively and may be one of the factors that contributed to its underperformance.
Second, the authors utilized an increase in cardiac output of 10% as determined by a NICOM monitor as their gold standard for fluid responsiveness. And while this is not an uncommon metric used to define fluid responsiveness, it relies on the validity of this non-invasive device which has shown moderate correlation when compared to the gold standard of a pulmonary artery catheter (PAC), a tool that in RCTs has never demonstrated its use improves patient oriented outcomes (3).
Third, while 49% of this cohort was identified as fluid responsive, is in line with most ICU cohorts, it is important to recognize that these are not the typical patients encountered in the Emergency Department with circulatory compromise. The median amount of fluid received in this cohort prior to enrollment was 4000 mL. This is vastly different from Emergency Department patients who typically have yet to undergo any resuscitative efforts. As such, the rate of fluid responsiveness in an Emergency Department cohort prior to fluid resuscitation is far higher (approximately 80%) (4). How this change in prevalence would affect the accuracy of IVC variation or PLR is unclear.
These methodological inadequacies are not isolated to the Corl et al paper and are pervasive throughout the literature examining fluid responsiveness. Take for example PLR, which has consistently demonstrated the strongest data supporting its use (5). In a recent meta-analysis by Cherpananth et al which comprised 23 studies and over 1,000 patients examining the use of PLR to predict fluid responsiveness, the authors reported a pooled sensitivity, specificity, and AUROC of 86%, 92%, and 0.95, respectively. And while these numbers seem diagnostically robust, a closer look at the primary trials suggest PLR’s true diagnostic capabilities are likely to be less ideal than reported.
The Lakhal et al trial published in Intensive Care Medicine in 2010 serves as a nice example. These authors enrolled 102 adult mechanical ventilated patients admitted to the ICU. The authors reported an AUROC of 0.89 using a gold standard of 10% increase in CO measured by PAC. In this case the diagnostic threshold was a 7% change in cardiac output (CO) with PRL using a PICCO monitor. But this 7% cutoff was determined retrospectively as it yielded the ideal diagnostic characteristics and again likely would perform suboptimally when validated in a prospective fashion. This technique of retrospectively determining the ideal diagnostic characteristics is ubiquitous throughout the trials included in the Cherpananth meta-analysis. In fact, the diagnostic threshold utilized to determine fluid responsiveness varied widely from study to study, ranging from 5-15%, each trial group retrospectively selected the threshold that provided the optimal performance. Given this, it is not unreasonable to conclude that there is not an ideal cutoff, and when used clinically PLR will perform far worse than the results described in this meta-analysis.
Take for example the largest of the trials by Duus et al trial published in the Critical Care Medicine in 2015, one of the few trials to prospectively determine their diagnostic threshold (7). The authors looked at 109 spontaneously breathing patients presenting to the Emergency Department in whom the treating physician intended to administer a fluid bolus. Using the NICOM as their gold standard they examined the diagnostic utility of a PLR to predict patients’ responsiveness a fluid bolus. In this case the authors predefined their threshold of PLR to predict fluid responsiveness as a 10% increase in CO, and when compared to the studies that determined optimal threshold retrospectively, the PLR diagnostic characteristics were far less impressive. The authors reported a sensitivity of 80%, specificity of 61%, and an overall accuracy of 0.74.
But of course all of these are all minor methodological observations that exist on the peripheries of the major issue. Does a strategy that employs fluid responsiveness to guide fluid therapy improve clinically important outcomes? In fact, there is no data suggesting that a fluid responsiveness guided strategy performs better than traditional methods to determine fluid administration. While it is an alluring physiologic surrogate, an increase in fluid responsiveness does not necessarily translate into improvements in survival. Approximately 90% of healthy volunteers will respond to a fluid bolus by increasing their cardiac output (8). And so one may conclude that we exist naturally in a fluid responsive state and to iatrogenically drive someone to the flat portion of their Frank-Starling curve is by definition over-resuscitation. Additionally, most studies that examine the duration of a fluid bolus’s cardiovascular augmentation demonstrate the effects are short-lived, lasting approximately 60 minutes (9). What impact can such an intervention truly have on clinical outcomes?
With this in mind let us once again examine the data comparing PLR and IVC variation as potential strategies to guide fluid administration. Many of these trials compare each tests respective performance using the AUROC. The AUROC is essentially a global assessment of a test across all possible diagnostic thresholds. But this is not how we typically utilize diagnostic tests clinically. Take for example either IVC variation or PLR to determine whether or not to administer a fluid bolus. We do not examine each test’s performance across all possible diagnostic thresholds, but rather make the decision to administer fluids based off a single dichotomous threshold. Above this threshold we administer a bolus and below it we withhold the intended therapy.
Despite its methodological inadequacies, the majority of literature suggests PLR outperforms IVC variation from the perspective of comparing AUROCs (5). But what does this mean clinically? Most of us who use IVC variation, only administer fluid to a spontaneous breathing patient if the IVC US exhibits significant respiratory variation. Trials have consistently demonstrated that when used in this manner, IVC US identifies a cohort of patients that are likely to be fluid responsive (10,11). But the contrary cannot be said with any confidence. A plethoric IVC without respiratory variation cannot rule out fluid responsiveness. This point is illustrated nicely in an article published by Airapetian et al in Critical Care in 2015. In this observational cohort, the authors examined 59 spontaneously breathing patients admitted to two ICUs in which the treating physician felt they required a fluid bolus. Authors prospectively collected IVC measurements, VTI and CO using transthoracic echo (TTE). The authors reported IVC variation predicted fluid responsiveness with an AUROC of 0.62. In contrast, PLR had a AUROC of 0.78 a far superior test from a purely statistical perspective. But using a cutoff of 42%, IVC variation demonstrated a sensitivity and specificity of 31% and 97%, respectively. What this means is IVC US will identify a cohort of patients who will almost universally respond to a fluid bolus, but it will also categorize a large portion of patients as non-responders who may actually still augment their CO with intravascular expansion (Fig. 1). While IVC variation is statistically inferior to PLR, from a clinical standpoint this failure would lead to a decrease in the total quantity of fluid administered, as it would fail to identify a subset of patients that would have otherwise been identified as “fluid responsive” by the use of PLR.
Simply put those of us that favor the use of IVC variation to guide fluid administration do so not out of ignorance, but rather because we believe it better represents a state of fluid tolerance. Utilizing strategies such as PLR, that encourage resuscitating patients to the limits of fluid responsiveness will more likely lead to over-resuscitation as patients are driven towards the flat portion of their Frank-Starling curves. In fact, the only high quality RCT comparing a fluid strategy using PLR and standard care demonstrated just that, as the patients randomized to the PLR arm received more fluid early in their resuscitation without any improvement in clinically important outcomes (4).
So often we become consumed with the numbers, exchanging insults regarding each tests respective AUROC without truly contemplating the clinical consequences this data implies. While fluid responsiveness is a physiologically appealing resuscitative strategy, it may turn out that a more fluid restrictive approach, one that focuses on early vasopressor use and fluid tolerant strategies, lends itself to improved outcomes. Or maybe no single tool can account for the complexity that surrounds the resuscitation of the critically ill patient and what is required is a thoughtful clinician standing at the bedside. The final answer is still unclear but to continue to debate the value of bedside strategies based on their ability to identify a physiological state of unknown clinical consequence does not get us closer to answering these questions. Instead our attention should shift to focus on how these tools affect clinical care and our ability to improve patient important outcomes.
Sources Cited:
- Thiele H, Zeymer U, Neumann FJ, et al. Intraaortic balloon support for myocardial infarction with cardiogenic shock. N Engl J Med 2012;367:1287-96
- Corl KA, George NR, Romanoff J, et al. Inferior vena cava collapsibility detects fluid responsiveness among spontaneously breathing critically-ill patients. J Crit Care. 2017;41:130-137.
- Shah MR, Hasselblad V, Stevenson LW, et al. Impact of the pulmonary artery catheter in critically ill patients: meta-analysis of randomized clinical trials. JAMA. 2005;294(13):1664-70.
- Kuan WS, Ibrahim I, Leong BS, et al. Emergency Department Management of Sepsis Patients: A Randomized, Goal-Oriented, Noninvasive Sepsis Trial. Ann Emerg Med. 2016;67(3):367-378.e3.
- Bentzer P, Griesdale DE, Boyd J, Maclean K, Sirounis D, Ayas NT. Will This Hemodynamically Unstable Patient Respond to a Bolus of Intravenous Fluids?. JAMA. 2016;316(12):1298-309.
- Cherpanath TG, Hirsch A, Geerts BF, et al. Predicting Fluid Responsiveness by Passive Leg Raising: A Systematic Review and Meta-Analysis of 23 Clinical Trials. Crit Care Med. 2016;44(5):981-91.
- Duus N, Shogilev DJ, Skibsted S, et al. The reliability and validity of passive leg raise and fluid bolus to assess fluid responsiveness in spontaneously breathing emergency department patients. J Crit Care. 2015;30(1):217.e1-5.
- Miller J, Ho CX, Tang J, et al. Assessing Fluid Responsiveness in Spontaneously Breathing Patients. Acad Emerg Med. 2016;23(2):186-90.
- Nunes TS, Ladeira RT, Bafi AT, De azevedo LC, Machado FR, Freitas FG. Duration of hemodynamic effects of crystalloids in patients with circulatory shock after initial resuscitation. Ann Intensive Care. 2014;4:25.
- Airapetian N, Maizel J, Alyamani O, et al. Does inferior vena cava respiratory variability predict fluid responsiveness in spontaneously breathing patients? Crit Care 2015;13(19):400.
- Preau S, Bortolotti P, Colling D. Diagnostic accuracy of the inferior Vena Cava collapsibility to predict fluid responsiveness in spontaneously breathing patients with Sepsis and Acute Circulatory Failure. Crit Care Med 2016 Sep 30 [Epub ahead of print].
- EM Nerd-The Case of the Partial Cohort - May 24, 2020
- EM Nerd: The Case of the Sour Remedy Continues - January 20, 2020
- EM Nerd-The Case of the Adjacent Contradictions - December 23, 2019
nicely argued and physiologically rational. Reddi & Carpenter explained how we can see venous excess in the jugular pulse, US exam must have a central venous parameter that is the gold standard for venous excess.
Great post! I have long been wondering why it it seems that in every study I read on the subject it is always around 50 % of patients that are “fluid responders”? it is never 20 or 30 or 70 %. Has it got more to do with basic human physiology in combination with the intervention (i.e. 6 ml/kg of fluids) and the outcome that we have chosen (i.e. 10 % increase in CO) for the test? In the post its mentioned that 90 % of healthy volunteers meet criteria for fluid responsiveness. Most studies are done in ICU patients… Read more »
Completely agree with your argument. Similar nonsensical recommendation abound in the fluid repletion in acute pancreatitis. Guidelines recommend agressive resuscitation (some to BUN/Creatinine endpoints) in the first 24h, based on very shaky primary evidence. My own approach is to eye-ball the intravascular deficit based on (1) the patients’ looks and (2) history, bolus them with that amount (with the flow rate depending on their age, heart function, and severity of hemodynamic compromise), and start vasopressors when I feel that the patient’s still hypoperfused despite the volume resus. Thus, acute pancreatites usually receive 1-2 L depending on how long they’ve been… Read more »
Let’s not forget that when you have the echo available for the critical pt., we don’t only evaluate the VC. The information we obtain looking at the heart, lungs, abdomen, large veins, is something we don’t have with other methods like PLR. For example, To detect in a critical pt´s heart echo a diminished EF or a dilated RV associated with the interpretation of the VC is a invaluable factor for the conduct we take regarding the use of fluids.