Meta-Analysis of Invasive versus Non-Invasive Techniques to Predict Fluid Responsiveness by Passive Leg Raising in the Critically Ill

Objective: To analyze the accuracy and specificity of recent studies to compare the ability of predicting fluid responsiveness with Passive Leg Raising (PLR) by using invasive or non-invasive techniques during passive leg raising. Data Sources: MEDLINE, EMBASE and the Cochrane Database of Systematic Reviews were systematically searched. Study Selection: Clinical trials that reported the sensitivity, specificity and area under the receiver operating characteristic curve (AUC) between the responder and non-responder induced by passive leg raising and Volume Expansion (VE) in critical ill patients were selected. 246 studies were screened, 14 studies were included for data extraction, which met our inclusion criteria. Data Extraction: Data were abstracted on study characteristics, patient population, type and amount of VE, time of VE, definition of responders, position, techniques used for measuring hemodynamic change, number and percentage of responders, the correlation coefficient, sensitivity, specificity, best threshold and area under the ROC curve (AUC). Meta-analytic techniques were used to summarize the data. Data Synthesis: A total of 524 critical ill patients from 14 studies were analyzed. Data are reported as point estimate (95% confidence intervals). The pooled sensitivity and specificity of invasive techniques were 80% (73% 85%) and 89% (84% 93%) respectively with the area under the sROC of 0.94. While, the pooled sensitivity and specificity of non-invasive techniques were 88% (84% 92%) and 91% (86% 94%) respectively with the area under the sROC of 0.95. The pooled DOR of invasive techniques was 32.2 (13.6 76.8), which was much lower than that of non-invasive techniques with the value of 64.3 (33.9 121.7). Conclusions: The hemodynamic indexes changes induced by PLR could reliably predict fluid responsiveness. Non-invasive hemodynamic techniques with their accuracy and safety can benefit the daily work in ICUs. Because the numHow to cite this paper: Si, X., Cao, D.Y., Wu, J.F., Chen, J., Liu, Z.M., Chen, M.Y., Bin, O.Y. and Guan, X.D. (2016) Meta-Analysis of Invasive versus Non-Invasive Techniques to Predict Fluid Responsiveness by Passive Leg Raising in the Critically Ill. International Journal of Clinical Medicine, 7, 736-747 http://dx.doi.org/10.4236/ijcm.2016.711080 Received: October 11, 2016 Accepted: November 19, 2016 Published: November 22, 2016 Copyright © 2016 by authors and Scientific Research Publishing Inc. This work is licensed under the Creative Commons Attribution International License (CC BY 4.0). http://creativecommons.org/licenses/by/4.0/ Open Access

X. Si et al.

Introduction
Fluid therapy is an essential part in Intensive Care Unit (ICU) to survive patients with hypovolemia.In fact, that's not easy.Studies have shown that about 50% of critically ill patients do not exhibit the desired effect [1].How to assess intravascular volume accurately has been a critical problem.
Passive Leg Raising (PLR) is a reversible maneuver that mimics rapid Volume Expansion (VE) by shifting venous blood from the lower limbs toward the intrathoracic compartment [2].Thus, PLR increases the cardiac preload.PLR has been validated to predict fluid responsiveness, but it requires the determination of CO or its surrogates with a fast-response device, because the hemodynamic changes may be transient [3] [4].
There are a lot of "fast-response devices" and all of them can be divided into 2 categories: invasive and non-invasive.Invasive hemodynamic techniques such as transpulmonarythermodilution (PiCCO), Vigileo, arterial BP transducer, pulmonary artery catheter are widely used in intensive units.Over the past few years, new techniques assessed for rapid and non-invasive prediction of fluid responsiveness have been introduced in clinical practice.Transthoracic echocardiography (TTE), transesophageal echocardiography (TEE), transthoracic Doppler ultrasonography (USCOM), Bioreactance technology-based system (NICOM), Continuous Non-invasive Arterial Pressure (CNAP) have been developed to predict fluid responsiveness.
Evidence shows that various studies have confirmed the ability of predicting fluid responsiveness by these techniques, but the predictive value of the hemodynamic response after PLR as a dynamic index of fluid responsiveness between invasive and noninvasive techniques has not been compared yet.The aim of this systematic review is to answer the question: can non-invasive techniques be better than invasive ones to be used as a tool for predicting volume responsiveness in critically ill during PLR maneuver and VE? Data reporting conformed to the Standards for Reporting of Diagnostic Accuracy (STARD) [5].

Study Selection
Only full-text articles in indexed journals were included.Reviews, chapter, case reports, reference network and studies published in abstract form were excluded.No language restriction was imposed.We included only studies with patients admitted in intensive care unit (ICU).Children and pregnant women would be excluded.Articles were collected by one reviewer and crosschecked by another reviewer and references of included papers were examined to identify other studies of interest.

Inclusion Criteria
We included full-text studies with the following criteria: 1. PLR was performed and followed with VE; 2. the number of patients and boluses had been counted; 3. the reference standard of predicting fluid responsiveness had been described; 4. the number of responsive patients and non-responsive patients had been counted; 5. sensitivity, specificity and the threshold of the index in identifying those patients who subsequently responded to VE (responders) had been calculated.

Data Extraction and Quality Assessment
Data were extracted using a structured data collection sheet including the following items: authors, year of publication, study setting, population, age of patients, number of patients included, ventilation mode, cardiac rhythm (sinus vs. arrhythmias), type and amount of VE, time of VE, definition of responders, position, assessments used for measuring hemodynamic change, number of VE administered, number and percentage of responders, sensitivity, specificity, best threshold and area under the ROC curve (AUC).We use QUADAS-2 (quality assessment of diagnostic accuracy-2) [6] to assess the quality of included studies on diagnostic accuracy in systematic reviews.The checklist was structured with 4 parts: patient selection, index test, reference standard and flow and timing.

Statistical Analysis
We used RevMan 5.2 (Cochrane Collaboration, Oxford, UK) to make the QUADAS-2 scale to assess quality of studies on diagnostic accuracy to be included in systematic reviews.To calculate pooled values of sensitivity, specificity, diagnostic odds ratio (DOR) and area under summary receiver operating characteristic (sROC) curve we used Me- We used the Cochran Q statistic [7] to evaluate heterogeneity between studies.When the value of p less than or equal to 0.10 and I 2 more than 50%, it could be regarded as heterogenerity significantly and a random effect model was used to perform metaanalysis.For sensitivity and specificity, the Spearman correlation coefficient between those two parameters was calculated to evaluate a threshold effect determining heterogeneity [8].
For each study, sensitivity, specificity, positive likelihood ratio (+LR), negative likelihood ratio (−LR), and DOR were calculated after constructing a 2 × 2 contingency table.Pooled estimates with 95% confidence intervals (CIs) were calculated using a random-effects model.A summary receiver operating characteristic (sROC) curve was drawn according to the regression model proposed by Moses et al. [9] and it was performed to assess the interaction between sensitivity and specificity.The area under the sROC curve (AUC) was obtained to assess the diagnostic performance of hemodynamic techniques.Potential presence of publication bias was tested using the Egger [10] and Begg test [11].

Process of Study Selection
The initial search yielded 246 articles after the first query in the three databases.Among them, 86 were excluded for not directly concerning item of interest.In the 160 full-articles, 103 were excluded because they were reviews, chapters or abstracts.16 were excluded because they didn't perform PLR and another 14 were excluded because they didn't use VE. 13 were excluded because they didn't satisfy our inclusion criteria.Therefore, 14 studies [12]- [25] were included for final analysis.

Characteristics of Included Studies
The clinical characteristics of the 14 included studies were summarized in Table 1 and main results were reported in Table 2 and Table 3.The results of QUADAS-2 were showed in Figure 1.All the included 14 studies were prospective studies with enrollment of patients with sign of inadequate tissue perfusion.We found good compliance with appropriate population selection, index test adequately described, appropriate reference standard, and adequate flow and timing.Population selection bias was minimized, as the inclusion criteria of the patients were close.However, no study described the blinding of the assessors to the outcome measurement of the results. 2 out of the 14 included studies didn't report the lasting time of PLR.

Diagnostic Accuracy of Invasive Techniques
We first divided the 14 studies into 2 groups: invasive group  [22] reported by the same technique only one was included in the meta-analysis in order to avoid duplication of sample size.
After excluded the threshold effect with spearman correlation coefficient = 0.233 and p = 0.546 (>0.05), we used Moses-Shapiro-Littenberg method to draw the symmetrical summary ROC curve (SROC) with AUC of 0.94.
The threshold for predicting fluid responsiveness varied between 5 and 15%.

Publication Bias
The result of Egger test and Begg test showed that the potential publication bias was significant (P > 0.05), which indicated a potential for publication bias.

Strengths and Limitations
The strengths of our meta-analysis lie in the methods adhering to recent guidelines for diagnostic reviews [6] [29] [30], as well as in the advanced statistical methods used [31], which analyze all reported thresholds, sensitivity, specificity and their correlated results simultaneously.Also no other review has compared the ability to predict fluid responsiveness between these 2 groups of hemodynamic techniques.The results of our review could guide the using of the techniques to assess patients' volume statue in our clinical practice.
Limitations still exist in our meta-analysis.First, the pooling of diagnostic accuracy data inevitably contributed to sources of bias [7], which could be revealed in the significant amount of statistical heterogeneity across studies.Second, the number of patients included in the present trials was small (14 studies, 524 patients).A better review needs larger sample of studies.Third, the criteria of the included studies are based on clinical manifestation and the confounding factors such as cardiac function, respiratory function, severity of disease have not been analyzed.

Conclusion
The hemodynamic indexes induced by PLR can well discriminate between fluid responders and non-responders regardless of arrhythmia and ventilation mode.Non-invasive hemodynamic techniques with their accuracy and safety can benefit the daily work in ICUs.
Two authors independently performed a search in MEDLINE (using PubMed as the search engine, from 1947), EMBASE (from 1974) and the Cochrane Database of Systematic Reviews for prospective studies in January 2014 with the following key words: X. Si et al. "Passive leg raising" AND (fluid therapy OR fluid responsiveness OR fluid expansion OR fluid load* OR volume therapy OR volume responsiveness OR volume expansion).

taDiSC 1 . 4 (
Unit of Clinical Biostatisticsteam of the Ramon y Cajal Hospital, Madrid, Spain).P-values of less than 0.05 were considered statistically significant.Publication bias was performed by STATA statistical software 12.0 (StataCorp, College Station, TX).

Figure 2 .
Figure 2. Forest plots of pooled sensitivity between invasive and non-invasive techniques.

Figure 4 .
Figure 4. Summary receiver operating characteristics curve for the ability of non-invasive techniques discriminate between responders and non-responders.

Table 1 .
Main characteristics of the included studies.

Table 2 .
Pooled results for predictive capacity of invasive hemodynamic techniques.

Table 3 .
Pooled results for predictive capacity of non-invasive hemodynamic techniques.