Functional hemodynamic tests: a systematic review and a metanalysis on the reliability of the end-expiratory occlusion test and of the mini-fluid challenge in predicting fluid responsiveness

We included articles published in the English language, in indexed scientific journals, from 1966 to June 2018. Reviews, case reports, and studies published in abstract form were not included. Only studies performed in adults were eligible for inclusion.

Only studies that compared the reliability of the FHT to a FC, as the gold standard for assessing fluid responsiveness, were included. The definition of a FHT was a standardized hemodynamic maneuver affecting cardiac function and/or heart-lung interactions and used to assess fluid responsiveness. The definition of a FC was a fixed quantity of fluid administered in a defined time to change a hemodynamic variable by a predetermined threshold. We included only the following hemodynamic variables as potential indicators of a positive FC: cardiac output (CO); SV; their indexed values (CI and SVI) or SV surrogates, i.e., aortic velocity-time integrals; and aortic blood flow, as assessed by either transthoracic or trans-esophageal echocardiography.

We excluded those studies in which FHTs were performed in patients with an open chest or with atrial fibrillation. We did not impose exclusion criteria regarding the modality or the absence of mechanical ventilation.

We independently searched the MEDLINE, EMBASE, and Cochrane Database of Systematic Reviews using the following search criteria: (fluid AND responsiveness) OR passive AND leg AND raising) OR end-expiratory AND occlusion AND test) OR pulse AND pressure AND variation) OR stroke AND volume AND variation) OR (dynamic AND indices OR indexes)) OR mini-fluid challenge) OR functional AND hemodynamic AND monitoring) OR (fluid AND challenge). Filters: Humans; English; Adult: 19+ years.

The references for all included papers, review articles, commentaries, and editorials on this topic were also reviewed to identify other studies of interest that were missed during the primary search. Two of the authors (FT and GM) independently performed the evaluation of titles and abstracts. The articles were then subdivided into three subgroups: “included” and “excluded” (if the two examiners agreed with the selection) or “uncertain” (in case of disagreement). In case of “uncertain” classification, a further examination was performed by an expert (AM) and a conclusive decision was made.

We used a standardized data form to extract the data from all included studies, recording (1) the characteristics of the investigated population, (2) the methods used to perform the FHT test and to assess its hemodynamic effect, (3) the modalities of FC administration and the definition of fluid responsiveness, and (4) the area under the receiver operating characteristic (ROC) curve (AUC) and all the statistical data obtained by the ROC curve analysis (i.e., sensitivity, specificity, Youden index, positive and negative predictive values, positive and negative likelihood ratios). For those studies in which more than one method of hemodynamic monitoring was used to estimate flow parameters, we reported only the data obtained by the technique considered to be the most reliable, according to the following scale: pulmonary artery catheter or calibrated technique > cardiac echocardiography performed by experts (both transthoracic or trans-esophageal) > uncalibrated technique or esophageal Doppler probes > bioimpedance or bioreactance.

The QUADAS-2 scale was used to assess the risk of bias of the included studies [32]. Two expert authors (AM and MC) independently examined the studies using predefined criteria, which are reported in Additional file 1: Table S1.

For each criterion, the risk of bias was judged as high (3 points), unclear (2 points), or low (1 point). If the answers to all signaling questions for a domain were “yes,” then the risk of bias was judged as “low.” If any signaling question was answered “no,” the potential risk of bias was defined as indicated in Additional file 1: Table S1. The sum of these points was used to calculate the global risk of bias.

Studies were included in the highest risk of bias group if the sum of the points obtained by the risk of bias and applicability judgment assessment was higher than the median value for all the studies.

Statistical analysis was conducted on the summary statistics described in the selected articles (e.g., means, medians, proportions), and therefore, the statistical unit of observation for all the selected variables was the single study and not the individual patients.

The descriptive statistics of individual studies used different statistical indicators for central tendency and variability, whereas absolute and relative frequencies were adopted for qualitative variables. Quantitative variables were summarized with means (standard deviation, SD) or medians (25th–75th interquartile range, IQR) according to their distribution.

For the selected studies, we planned to perform (1) a metanalysis in order to determine the pooled AUC and the pooled sensitivity and specificity of the FHT as a predictor of fluid responsiveness and (2) a metanalysis in order to determine the pooled correlation between the changes in the flow hemodynamic parameters after FHT and the changes after FC administration. The FC was the exposure variable, and clinical and hemodynamic characteristics were considered as the outcome variables. Fixed effect models were used. In-between study heterogeneity was assessed through the I² indicator. Bias assessment graphs were plotted, and Egger’s regression analysis was used to evaluate the publication bias. Student’s t test or Mann-Whitney test for parametric or non-parametric distributions were respectively used to assess a difference in mean values between responders and non-responders.

The pooled AUC of the EEOT from two studies conducted in the OR [33, 34] and six [23, 43, 46, 48‐50] in the ICU was 0.96 (95%CI 0.92–1.00). The pooled sensitivity of the test was 0.86 (95%CI 0.74–0.94), with I² of 75% (95%CI 43–85%), and the pooled specificity was 0.91 (95%CI 0.85–0.95), with I² of 35% (95%CI 0–69%). The median threshold identified was a 5% (4–8%) increase in the considered variable. The funnel plot of the included studies testing the EEOT shows a significant likelihood of publication bias (see Additional file 1: Figures S1 and S2).

The pooled AUC of the mini-FC obtained from two studies conducted in the OR [35, 36] and five [29, 40, 41, 44, 45] in the ICU was 0.91 (95%CI 0.85–0.97). The pooled sensitivity of the test was 0.82 (95%CI 0.76–0.88), with I² of 26.9% (95%CI 0–69%), and pooled specificity was 0.83 (95%CI 0.77–0.89), with I² of 34% (95%CI 0–71%). The median threshold identified was a 5% (3.0–7.0%) increase in the considered variable.

The funnel plot for the included studies testing the mini-FC shows a small likelihood of publication bias (see Additional file 1: Figures S3 and S4). Moreover, it was possible to calculate a pooled correlation of r = 0.68 (95%CI 0.41–0.84) between the changes in the cardiac flow parameters after mini-FC application and after FC administration from data obtained from 6 studies [29, 36, 40, 41, 44, 45].

Of the tests studied, the EEOT showed the highest sensitivity and specificity [pooled AUC of 0.96 (95%CI 0.92–1.00); pooled sensitivity and specificity of 0.86 (95%CI 0.74–0.94) and 0.91 (95%CI 0.85–0.95), respectively, with a best threshold of 5% (4.0–8.0%) of increase in SV or its surrogates; see Fig. 3 and Table 3]. In the two studies reporting an AUC higher than 0.90, the percentage of patients included in the gray zone was 17–20% [34, 46] (see Table 3).

This FHT was first proposed by Monnet et al. [23] and predicts fluid responsiveness by assessing changes in CO, or its surrogates, following a few second interruption to mechanical ventilation. In preload-dependent patients, this maneuver increases venous return and right ventricular and then subsequently left ventricular stroke volume. The potential drawbacks of this FHT include that it may be limited by patient positioning, the baseline tidal volume ventilation adopted, and the hemodynamic effects of residual spontaneous breathing efforts. Only one study used the EEOT to assess fluid responsiveness in prone ICU patients with moderate ARDS, reporting an AUC of 0.65 (0.46–0.84) [43]. Prone positioning affects the venous return by compressing the inferior cava vein and changing the intra-abdominal pressure [51‐53], which may reduce the changes in CO and SV seen in response to the ventilatory challenge and limit the reliability of the EEOT.

The change in intrathoracic pressure may be insufficient to adequately increase right ventricular preload when a lung-protective ventilation strategy is used. Also, if the neural trigger for ventilation is preserved, a 15- to 30-s expiratory hold would result in a progressive increase in inspiratory pressure [54], affecting the venous return and the reliability of the FHT. Unfortunately, data regarding these issues is limited and contradictory.

In the OR, the EEOT performed better in a study using a mean tidal volume of 6.8 ml/kg [34], when compared to another study using 8.2 ml/kg [33]. In the ICU, the median tidal volume in those studies enrolling supine patients was 6.8 ml/kg (6.1–7.3). The EEOT failed to predict fluid responsiveness in the study of Myatra et al. using a 6-ml/kg ventilation [49], whereas Jozwiak et al. reported an AUC of 0.98 (0.85–1.0) using a 6.2-ml/kg ventilation. Interestingly, these two latter studies reported a comparable mean total respiratory system compliance in the enrolled patients (28 vs. 36 ml/cmH₂O, respectively).

Monnet et al. reported an EEOT failure as high as 22.5%, due to the patient effort against an occluded airway [23]. However, none of the other studies using this FHT reported this failure rate. Four of the five studies reported no spontaneous breathing activity during assisted-controlled ventilation (see Table 1), implying the level of sedation was inhibiting neural triggering. None of these studies reported a flowchart showing the overall number of excluded patients, limiting the assessment of EEOT reliability during visible spontaneous breathing activity, which is a potential drawback for assessing fluid responsiveness.

The mini-FC showed a pooled AUC of 0.91 (95%CI 0.85–0.97). The pooled sensitivity and specificity were 0.82 (95%CI 0.76–0.88) and 0.83 (95%CI 0.77–0.89), respectively, with a best threshold of 5% (3.0–7.0%) increase in SV or its surrogates, see Fig. 4 and Table 3. These values of the pooled ROC curve imply a moderate overlap in the distribution of responders and non-responders.

In the two studies reporting an AUC higher than 0.90, the percentage of patients included in the gray zone was approximately 14–19% [35, 36] (see Table 3). Moreover, the performance of this FHT was comparable under stable conditions in the OR (using uncalibrated tools) and in more unstable ICU patients (using calibrated tools) (see Table 1).

The dose of the mini-FC was not fixed. Most of the studies used a bolus of 100 ml infused over 60 s, but Wu et al. demonstrated that a 10% of change in SV following the infusion of a 50-ml bolus in 10 s reliably predicted fluid responsiveness [40].

Some may argue that the mini-FC should not be considered an appropriate FHT, since the response to the first small aliquot of fluids is actually included into the response to the final volume administered, therefore not predicting the response to the whole FC, but only to a part of it. However, recent studies have shown different components of FC, related to the response (the extent of SV increase) and sustainability of the hemodynamic effect (the effect of SV over time) [55‐57]. The mini-FC allows a dynamic evaluation of fluid administration, preventing inappropriate administration and allowing a tailored infusion. Moreover, this FHT has been also used in a different functional manner. In fact, Mallat et al. [45] demonstrated that a reduction in PPV [AUC = 0.92 (0.81–0.98)] or SVV [AUC = 0.91 (0.80–0.97)] following a mini-FC test was a better predictor of fluid responsiveness than an increase in CO. The cut-offs identified by the ROC curve for the changes in PPV and SVV are even smaller (2.0%) than the changes in CO (5.2%), implying a high precision of measurement, whichever hemodynamic tool is used.

All the other FHTs reported in the literature affect both right ventricular preload and afterload, by briefly altering intrathoracic pressure and, as a consequence, venous return and pulmonary vascular resistance.

The RSVT is based on the delivery of consecutive pressure-controlled inspiratory breaths, using incremental peak inspiratory pressures (up to 30 cmH₂O) and plotting the minimal values of the systolic arterial pressure recorded after each breath against the related airway pressures (offline slope calculation) [28, 37]. Despite promising results obtained in both the OR and ICU [28, 37], the integration of respiratory and hemodynamic signals required to allow an online computation of the RSVT has never been achieved at the bedside.

Raising intrathoracic pressure by increasing peak inspiratory pressure using either a Valsalva maneuver [42]or the end-expiratory occlusion pressure [47] or by performing a LRM are all FHTs that induce a sudden change in right ventricular preload and afterload. LRMs have been successfully applied in the OR, showing a comparable AUC in neurosurgery [38] and general abdominal surgery [39]. However, Biais et al. found that the best threshold to define fluid responsiveness was a 30% reduction in SV, but De Broca et al. showed only a 16% reduction was required [39], suggesting caution in the interpretation on this FHT.

Finally, more recently, Myatra et al. successfully enhanced the reliability of baseline indexes of fluid responsiveness by increasing the tidal volume from 6 to 8 ml/kg for 1 min (the tidal volume challenge) [49]. This simple and quick FHT could be used in patients undergoing protective ventilation but should be tested in larger ICU populations both with and without spontaneous breathing activity.

The EEOT and the mini-FC could be appropriately used in different clinical scenarios, especially when the PLR is unsuitable or in adjunct to that. In Fig. 5, we propose a step-by-step clinical algorithm in patients who would benefit from FC administration in the OR and the ICU.

The comparability of the included studies is limited by the heterogeneity of FC administration used as the reference point (see Table 1). Aya et al. have previously demonstrated that a FC should be at least 4 ml/kg [55]. For this reason, some patients enrolled in those studies adopting a smaller dose of FC (3.7 ml/kg [34]; 3.3 ml/kg [35, 38]) may be underchallenged, which may have affected the observed rate of fluid responsiveness and, in turn, the ROC curve construction.

Another potential source of bias is related to the different hemodynamic tools used to assess both fluid responsiveness and FHT reliability. In fact, when considering the median cutoff value identifying responders from non-responders (about 5% for both the EEOT and the mini-FC), the accuracy of measurement of the changes in CO, or its surrogates, is of pivotal importance. For example, the negative results of Guinot et al. [33], conducted in the OR, have been questioned as the esophageal Doppler does not measure the change in aortic diameter and could therefore underestimate the change in SV during either the EEOT or the FC [59].

Additionally, the reliability of different calibrated and uncalibrated tools in tracking the dynamic trends of CO may not be consistent and may be below the boundaries of the accuracy and precision of the Critchley-Critchley criteria [60, 61]. For instance, the reproducibility of the measurements obtained by the different hemodynamic tools has never been reported in the included studies. This implies that small changes in CO, or its surrogates, after a FHT may be inaccurately detected in the OR, where the hemodynamic monitoring is usually performed with uncalibrated tools, whereas the use of calibrated techniques by means of thermodilution could reduce the risk of imprecise measurements in ICU.

All the included studies had a small-sized single-center design and enrolled a median number of patients rather small [39 (IQR 25–50)], and about 43% of the included studies were classified in the subgroup with the highest risk of bias, mainly because of the drawbacks related to the patient selection, according to the QUADAS-2 score (see Table 2). This limitation along with the use of different cutoff values, thresholds, and measurement techniques to assess fluid responsiveness potentially produced heterogeneity in the response to the FC administration. As confirmed, the proportion of responders ranged between 30 and 71% across the included studies. The bedside application is also limited in those potentially misclassified patients (roughly 20% in the reported studies) included in the gray zone of the ROC curve, where the predictive power of the FHT is rather low. Another source of heterogeneity may be related to the different sample sizes of the included studies, as confirmed by the large interquartile ranges of the I². Finally, we did not include non-full-text studies, studies not in English, and unpublished studies, and this systematic review was not prospectively registered in PROSPERO, an international database of systematic reviews in health and social care, increasing the overall risk of reporting bias.

For all these key aforementioned reasons, despite the increasing number of studies in this field, the clinical applicability and utility of the FTHs should be assessed by a large multicentric trial. Although pooling a few data from studies carried out in different settings could bias the interpretation of the findings, the identification of the current evidence, associated to a careful assessment of the confounding factors, could help in designing future studies.