Assessment of fluid responsiveness in spontaneously breathing patients: a systematic review of literature

Patients who increase stoke volume or cardiac index more than 10 or 15% after a fluid challenge are usually considered fluid responders. Assessment of fluid responsiveness prior to volume expansion is critical to avoid fluid overload, which has been associated with poor outcomes. Maneuvers to assess fluid responsiveness are well established in mechanically ventilated patients; however, few studies evaluated maneuvers to predict fluid responsiveness in spontaneously breathing patients. Our objective was to perform a systematic review of literature addressing the available methods to assess fluid responsiveness in spontaneously breathing patients. Studies were identified through electronic literature search of PubMed from 01/08/2009 to 01/08/2016 by two independent authors. No restrictions on language were adopted. Quality of included studies was evaluated with Quality Assessment of Diagnostic Accuracy Studies tool. Our search strategy identified 537 studies, and 9 studies were added through manual search. Of those, 15 studies (12 intensive care unit patients; 1 emergency department patients; 1 intensive care unit and emergency department patients; 1 operating room) were included in this analysis. In total, 649 spontaneously breathing patients were assessed for fluid responsiveness. Of those, 340 (52%) were deemed fluid responsive. Pulse pressure variation during the Valsalva maneuver (∆PPV) of 52% (AUC ± SD: 0.98 ± 0.03) and passive leg raising-induced change in stroke volume (∆SV-PLR) > 13% (AUC ± SD: 0.96 ± 0.03) showed the highest accuracy to predict fluid responsiveness in spontaneously breathing patients. Our systematic review indicates that regardless of the limitations of each maneuver, fluid responsiveness can be assessed in spontaneously breathing patients. Further well-designed studies, with adequate simple size and power, are necessary to confirm the real accuracy of the different methods used to assess fluid responsiveness in this population of patients. Electronic supplementary material The online version of this article (10.1186/s13613-018-0365-y) contains supplementary material, which is available to authorized users.


Background
Intravascular volume expansion is a common intervention in critically ill patients [1]. Patients who will benefit from intravascular volume expansion, i.e., will boost stroke volume (SV) after a volume expansion, have both ventricles in the ascending portion of the Frank-Starling curve, characterizing a preload dependency [1,2]. Nevertheless, nearly 50% of critically ill patients will not benefit from an intravascular volume expansion [2,3]. Conversely, an accurate assessment of fluid responsiveness prior to volume expansion is critical to avoid fluid overload, which has been associated with increased morbidity and mortality in critically ill patients [4][5][6].
The concept of predicting fluid responsiveness was initially reported in deeply sedated patients under volumecontrolled mechanical ventilation with tidal volume Open Access *Correspondence: chavesrcf@hotmail.com 1 Intensive Care Unit, Hospital Israelita Albert Einstein, Av. Albert Einstein, 627/701, 5th Floor, São Paulo, SP 05651-901, Brazil Full list of author information is available at the end of the article (VT) of at least 8 ml/Kg and positive end-expiratory pressure (PEEP) lower than 10 cm H 2 O [7]. Nonetheless, since many patients in the intensive care unit (ICU) are not under such conditions, for many years the presence of spontaneous breathing or inspiratory efforts, with or without an endotracheal tube, was considered a major limitation to assess fluid responsiveness in critically ill patients [8].
Knowledge on the interaction between heart, lung and abdominal compartment is critical to understanding the concept of fluid responsiveness [9,10]. In spontaneous breathing patients without mechanical ventilation, intrathoracic pressure decreases, while venous return and stroke volume increases during inspiration [10]. On the other hand, at expiration, intrathoracic pressure increases, while venous return and stroke volume decreases [10]. Thus, quantifying stroke volume variation, between respiratory cycles could be used to assess fluid responsiveness [1].
Thus, our primary objective was to perform a systematic review addressing the available methods for fluid responsiveness assessment in spontaneously breathing patients. A secondary objective was to summarize the performance of available methods to assess fluid responsiveness in spontaneously breathing patients.

Methods
This systematic review was reported following the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines [30].

Eligibility criteria
Articles were selected for inclusion if they evaluated fluid responsiveness in spontaneous breathing adult patients. Articles were assessed for eligibility if one of the following standard definitions of fluid responsiveness and fluid challenge was adopted: increase in stroke volume (SV) ≥ 10% and/or cardiac output (CO) ≥ 10% and/ or cardiac index (CI) [31] ≥ 10% and/or aortic velocity-time integral (VTI) ≥ 10% after a fluid challenge [2,32]. Fluid challenge was considered adequate if at least 250 ml over 30 min of intravenous (I.V.) fluid was infused [2,33]. Spontaneously breathing was defined as patients without any ventilatory support, patients on noninvasive mechanical ventilation or patients on invasive mechanical ventilation in a spontaneous mode. Patients in the following clinical scenarios were included: ICU, emergency department (ED) and operating room.

Identifying studies
An electronic literature search was carried out by two authors through a computerized blinded search on PubMed. The following search strategy was applied: ((("hemodynamics"[MeSH Terms] OR "hemodynamics"[All Fields]) AND ("respiration"[MeSH Terms] OR "respiration"[All Fields] OR "cell respiration"[MeSH Terms] OR ("cell"[All Fields] AND "respiration"[All Fields]) OR "cell respiration"[All Fields]) AND ("cardiac output"[MeSH Terms] OR ("cardiac"[All Fields] AND "output"[All Fields]) OR "cardiac output"[All Fields]))). Literature search was limited to a period of time (01/08/2009 to 01/08/2016) and to "human. " No restrictions on language were adopted. Additionally, we hand-searched the reference lists of the included studies to identify other relevant studies.

Study selection
Prospective studies that reported sensitivity, specificity, cutoff value of each maneuver to assess fluid responsiveness, number of patients included and frequency of fluid responsiveness and non-fluid responsiveness patients were included in this systematic review. Review articles, editorials, studies assessing fluid responsiveness during mechanical ventilation and studies that did not report outcomes of interest were excluded.

Data extraction
Two authors independently screened all retrieved citations by reviewing their titles and abstracts (RCFC and FTM). Then, the reviewers independently evaluated the full-text manuscripts for eligibility using a standardized form. Reviewers independently extracted the relevant data from the full-text manuscripts and assessed the risk of bias using a standardized form. Any disagreement between the authors was resolved by a third author (ASN).

Quality assessment
The quality of each study was evaluated by the Quality Assessment of Diagnostic Accuracy Studies tool (QUA-DAS) [34]. Details of the quality assessment are reported in Additional file 1.

Primary objective
The primary objective was to report the available methods to assess fluid responsiveness in spontaneously breathing patients.

Secondary objectives
Secondary objectives were to assess diagnostic performance and build a receiver operating characteristics curve (ROC curve) of methods available to assess fluid responsiveness in spontaneously breathing patients.
Pulse pressure variation was calculated as the difference in pulse pressure maximal (PPmax) and pulse pressure minimal (PPmin) over the respiratory cycle divided by the mean between PPmax and PPmin [∆PP = (PPmax − PPmin)/(PPmax + PPmin)/2] [16,19,20]. Passive leg raising consists in moving the patient from the 45° semirecumbent position to a horizontal position with the lower limbs lifted 30°-45° relative to the trunk [1,18]. PLR was determined as the difference between baseline and the highest value induced during the PLR or after the PLR [21,23,27]. Inferior vena cava collapsibility index represents the difference in the vena cava maximum diameter (IVCmax) and vena cava minimum diameter (IVCmin) divided by the vena cava maximum diameter over the respiratory cycle [cIVC = (IVCmax − IVCmin)/(IVCmax)] [26,27]. Valsalva maneuver consists of sustaining a forced expiration effort against a closed mouth [16]. Forced inspiratory breaths consist of three respiratory cycles of deep inspiration immediately followed by slow passive expiration [20]. Deep inspiration maneuver consists of slow continuous inspiration strain (5-8 s) followed by slow passive exhalation [19].

Statistical analysis
The number of patients included, study design, setting, inclusion and exclusion criteria, time and type of fluid infused, the best cutoff value of each maneuver and definition of fluid responders were extracted from published studies. The accuracy of each diagnostic test was assessed with sensitivity (Sens), specificity (Spec), positive predictive value (PPV), negative predictive value (NPV), positive likelihood ratio (LR +), negative likelihood ratio (LR −), AUC along with its standard deviation (SD) or 95% confidence interval (95% CI). Whenever not reported, accuracy, PPV, NPV, LR + and LR − were calculated using the Review Manager (RevMan) [computer program]-version 5.3-Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014.
A receiver operator characteristics curve (ROC curve) was constructed using the sensitivity and specificity of each maneuver extracted from included study using Meta-DiSc version 1.4 (Universidad Complutense, Madrid, Spain) [35]. Methods for fluid responsiveness assessment were classified according to their accuracy [area under the receiver operating characteristics curve (AUC)]. AUC from 0.90 to 1.00 was considered excellent, from 0.80 to 0.89 adequate, from 0.70 to 0.79 fair, from 0.60 to 0.69 poor and from 0.50 to 0.59 failure [36].

Search results
The initial search strategy identified 537 studies (Fig. 1). After screening the reference lists of the included studies, 9 potentially relevant articles were included and 546 potentially relevant articles were selected. Fifteen prospective studies (649 patients in total) were included in this systematic review after the exclusion of 531 studies  (307 studies had no data on outcome of interest, 111 studies did not regard spontaneously breathing patients, 75 studies did not access fluid responsiveness, and 38 were review articles or editorials) (Fig. 1).

Methods for fluid responsiveness assessment
Thirty-four maneuvers for predicting fluid responsiveness in spontaneously breathing patients were reported (Tables 1 and 2). Studies that adopted pulse pressure variation to assess fluid responsiveness are summarized in Table 1. Studies that adopted echocardiography maneuvers, pulse contour analysis or noninvasive cardiac output monitor (NICOM ® ) are summarized in Table 2.

Spontaneous breathing with ventilatory support
∆SV-PLR TE > 13% [AUC ± SD: 0.96 ± 0.03] had the highest accuracy, while ∆SV ≥ 10% [AUC (95% CI) 0.57(0.34-0.78)] had the worst accuracy to predict fluid responsiveness in mechanically ventilated patients in a spontaneous mode (3 studies totaling 77 patients) (Additional file 1: Figure S2).  Table 3 Performance of included studies that addressed pulse pressure variation to predict fluid responsiveness in spontaneously breathing patients Sens sensitivity, Spec specificity, PPV positive predictive value, NPV negative predictive value, LR + positive likelihood ratio, LR − negative likelihood ratio, AUC area under the receiver operating characteristics curve, SD standard deviation, 95% CI 95% confidence intervals, ∆PP pulse pressure variation, ∆SP systolic pressure variation, ∆PPf ∆PP during forced inspiratory effort, ∆SPf ∆SP during forced inspiratory effort, ∆PPV ∆PP during the Valsalva maneuver, ∆VSP ∆SP during the Valsalva maneuver, PPmin lowest pulse pressure, PLR passive leg raising, ∆SV-PLR PLR-induced change in stroke volume, ∆PP-PLR PLR-induced change in radial pulse pressure, ∆VF-PLR PLR-induced change in the velocity peak of femoral artery flow, ∆PPdim deep inspiration maneuver-induced change in pulse pressure, ∆VF respiratory change in velocity peak of femoral artery flow, ∆VFdim deep inspiration maneuver-induced change in velocity peak of femoral artery flow, ∆PP FB ∆PP during forced inspiratory breathing

Discussion
The main finding of this systematic review is that, regardless of intrinsic limitations of each reported maneuver, fluid responsiveness can be assessed in spontaneously breathing patients with acceptable accuracy. Approximately two-thirds (19/29) of reported maneuvers were deemed adequate or excellent to predict fluid responsiveness in spontaneous breathing patients without ventilatory support and 60% (3/5) were deemed excellent in mechanically ventilated patients in a spontaneous mode.
Moreover, approximately half of the patients included in this study were not fluid responsive. This finding reinforces the importance of assessing fluid responsiveness in critically ill patients prior to intravascular volume expansion, thus avoiding unnecessary exposure to additional fluids.
In patients with an invasive arterial line in place, dynamic parameters such as ∆PP in association with a maneuver that magnifies cyclic changes in intrathoracic pressures, i.e., deep inspiration or forced inspiratory breathing, represent important tools to assess fluid responsiveness continuously and with minimal interrater variability. [19,20] Echocardiographic maneuvers such as ∆VF, ∆SV, cIVC represent important tools to assess fluid responsiveness in patients without availability of an invasive arterial line [19,21,23,28]. Although it is operator-dependent, echocardiographic is a noninvasive technique that enables fluid responsiveness assessment with good accuracy in spontaneously breathing patients [19,21,23,28]. The main disadvantages of echocardiographic measurements are non-continuous monitoring and high inter-rater variability [18,24,27].
Reversible and noninvasive maneuvers that magnify cyclic changes in intrathoracic pressures and on transpulmonary pressure, such as Valsalva or deep inspiration maneuver, in association with ∆PP or echocardiographic measurements, improve the accuracy of the maneuvers without adverse effects, allowing clinicians at the bedside to assess preload dependency [16,19]. Nevertheless, it is important to emphasize that all reported methods to assess fluid responsiveness in spontaneously breathing patients have limitations [13,14]. The need of patients cooperation, inability to sustain deep inspiration, presence of pain, intra-abdominal hypertension, major abdominal surgery, low diaphragm strength, higher respiratory rate, low reproducibility and lack of external validation are frequently reported limitations of available methods [16].
Furthermore, transforming a continuous diagnostic index, such as ∆PP and ∆SV, into binary variables (i.e., responders or non-responders) represents an important limitation of all methods to assess fluid responsiveness [37]. The decision of whether to support or avoid volume expansion in patients with intermediate values of continuous diagnostic index could be imprecise (gray zone) [37]. These patients may benefit from a reversible maneuver, such as PLR prior volume expansion to avoid unnecessary exposure to fluids [37].
Our study has limitations. First, it is important to emphasize that the results of this systematic review should be interpreted in the context of the included studies. Furthermore, studies with small sample size, carried out in different clinical scenarios and with a heterogeneous methodology, were included in this systematic review. Finally, systematic reviews are subject to publication bias, which may exaggerate the conclusion of the study if publication is related to the strengths of the results.

Conclusion
In conclusion, our systematic review suggests that regardless of the limitations of each maneuver, fluid responsiveness could be assessed in spontaneously breathing patients. Further research with adequate sample size and power are necessary to confirm the real accuracy of the different methods available to assess fluid responsiveness in this population of critically ill patients.

Additional file
Additional file 1. A pdf file containing quality of each study was evaluated by the Quality Assessment of Diagnostic Accuracy Studies tool (QUADAS), a receiver operating characteristic curve of methods to assess fluid responsiveness in spontaneous breathing patients without any ventilatory support and in mechanically ventilated patients during a spontaneous mode.