Predictive performance of PiCCO system and blood gas parameters for early prognosis of patients with sepsis

Purpose: Invasive hemodynamics monitor directed uids resuscitation remains misgiving issues. This study aims to explore the predictive value of the PiCCO system (pulse indicator continuous cardiac output device) and blood gas parameters on the early prognosis of patients with sepsis. Methods: 42 patients with sepsis were included from January 2013 to January 2015. All patients were stratied into survivor group (n=29) and nonsurvivor group (n=13) based on seven day-mortality. The PiCCO and blood gas parameters at enrollment and 24 hours were compared between two groups. The predictive performance of these parameters was distinguished with Area Under the Receiver Operating Characteristic Curve (AUC). Results: At 24 hours after enrollment, the HR (97.27±22.07 vs. 120.20±20.56), extravascular lung water index (EVLWI) (7.32±2.96 vs. 15.9±11.2), and lactic acid (Lac) (1.62±0.92 vs. 6.33±5.83) level were signicantly lower in survivor group (P < 0.05), whereas the cardiac index (CI) (3.67±0.85 vs. 2.98±0.73) and PaO 2 /FiO 2 (242.8±89.68 vs.136.07±78.01) increased signicantly. Meanwhile PaO 2 /FiO 2 was negatively correlated with EVLWI (r= -0.673, P < 0.01). The AUC of the combination of Lac with PaO 2 /FiO 2 , HR, EVLWI, pulmonary vascular permeability index (PVPI) and Lac at 24 hours were 0.853, 0.739, 0.776, 0.764, and 0.794. Conclusions: The PiCCO and blood gas parameters exhibit superior predictive capability for early prognosis in patients with sepsis, and the combination Lac with PaO 2 /FiO 2 was noninferior under the circumstance of unavailability with PiCCO.


Introduction
Although numerous therapeutic approaches have been employed, sepsis remains one of the major causes of death in the intensive care unit (ICU) [1]. As recommended by the Surviving Sepsis Campaign guidelines, early uid resuscitation occupies the core of the therapeutic bundle [2]. Nevertheless, the contrary pieces of the evidence challenge the reliability of the uid treatment. Fluid overload can increase the risk of organ dysfunction, including acute lung injury and acute renal injury, thus lead to high mortality [3]. Hence, it is crucial to ensure uid balance based on the stable circulation system. The PiCCO (Pulse index continuous cardiac output) system as the representative of invasive techniques permits the monitoring of hemodynamic status including vascular tone, uid load and cardiac function, following responding comprehensively physiological condition, such as uid responsiveness, oxygenation and pulmonary oedema. Although the application of PiCCO emerged promptly in recent times, some disputes remain yet clari ed. Whether the static parameters derived from PiCCO can present full disclosure of the volume evolution, the threshold of some settings can be the best guidance to clinical practice, and this technique can authentically improve or predict the clinical outcomes [4]. Furthermore, the risk of invasive manipulation and high expense under some circumstance drive the implementation of PiCCO into irresolute. The blood gas analysis, as a routine laboratory test characterized by easy accessibility, minimally invasiveness, and low price, still keep strong vitality to assess the oxygenation status and acid-base balance for critically illness patients [5].
We conducted this study to re ect the forecast value of PiCCO system on early clinical outcome in patients with sepsis and further compare the link between PiCCO with blood gas parameters.

Study design
This retrospective study was performed in a 29-bed mixed adult ICU at a tertiary teaching hospital from January 2015 to January 2018. The hospital ethics committee endorsed this study.

Study Participants
This study encompassed 51 adult patients (age ≥ 18 years) who met the clinical criteria of sepsis in Sepsis-2 consensus de nitions [6]. Subjects were discarded meeting one of the following conditions: admission owing to trauma, pregnancy, or poison, under treatment with immunosuppressant agents, chemotherapy or radiotherapy within 30 days before inclusion, and refusal to cardiopulmonary resuscitation.

Intervention
After all patients diagnosed with sepsis, the clinicians instantly initiated the treatment of Surviving Sepsis Campaign Bundle, including obtaining microbial samples from infectious sites, administration of broadspectrum antibiotics, uid resuscitation, commence a vasopressor titration et cetera. Prior to uid infusion, all cases were implemented with PiCCO device, and the PiCCO survey was executed at least every 8 hours, or depended on the clinician decision when the drastic vital sign change. Blood gas analysis was obtained along with every PiCCO inquiry.

Data collection
According to seven day survival or not after enrollment, all patients were classi ed into survivor group (n=29) and nonsurvivor group (n=13). The following clinical characteristics were assembled: sex, age, infection position, underlying disease, such as hypertension, chronic cardiovascular disease, chronic obstructive pulmonary disease, chronic renal failure, and diabetes mellitus. Meanwhile the Acute physiology and chronic health practical guidance system II (APACHE II) and the Sequential Organ Failure Assessment (SOFA) score were calculated for patient within rst 24 hours.

Statistical analysis
Data were represented with either frequency and percentage for categorical variables or mean and standard deviation (SD) and interquartile range (IQR) for continuous variables. Two groups were compared with the t test or the the non-parametric Mann-Whitney U test, as appropriate. Dependency relationships among the hemodynamic and blood gas parameters were identi ed by person correlation analysis. All parameters exhibited signi cant difference in comparison were further evaluated for the predictive value via receiver-operating characteristic (ROC). All data processing and statistical analyses were performed with SPSS 13.0.
Compared to the single parameter, the combination of Lac and PaO 2 /FiO 2 at 24 hours was superior for early prognosis. The AUC was 0.853, with a sensitivity of 83.3%, a speci city of 84.6%, and the Youden' s Index of 0.679, as seen in Figure 3.

Discussion
The main fact of this study is that certain hemodynamic or blood gas parameters measured at 24 h after sepsis onset were associated with worse early clinical outcomes.
Although the absolute mortality has decreased from 35.0% to 18.4% in patients with severe sepsis in the areas of Australia and New Zealand during the last decades [7], one recent systematic review revealed the ICU mortality of sepsis in Europe and North America remains 37.3%, where the morbidity was estimated at 10.4% [8]. It is similar to these studies that the 28-day mortality was 42.9% in our study. The factor that average age over 65 in our patient population could contribute to the relatively high mortality, according to the results from some surveys launched in China [1,9].
To date, the lack of recognition in the underlying pathophysiology mechanism falls us into the dilemma of treatments for sepsis, of which to maintain the stable hemodynamics occupy the critical core in the early period. In order to achieve the goal of volume status ful lled with oxygen delivery-consumption balance, active uid resuscitation is always the basis of treatment bundle; meanwhile, the varied methods to monitor the uid infusion could overcome some side-effects perhaps to follow, such as pulmonary edema or cardiac overload.
Due to those signi cant drawbacks, including air embolism, permanent occlusion, pseudoaneurysm and so on, occurred in fewer than 1% of the clinical procedures, PiCCO has been the invasive hemodynamic monitor system generally accepted [10]. The uid challenge strategics directed via PiCCO present fully compatible with the philosophy of titration therapy. It is a remarkable fact that numerous actively advocated studies virtually focus on one or more speci c parameters yield from the PiCCO techniques rather than itself [11]. Theoretically, the ability to integrate a series of hemodynamic data provide the most valuable advantage of PiCCO; however, the misinterpretation of these data causes the wrong direction of treatment. Zhang et al. asserted that the uid management schedule based on PiCCO parameters couldn't improve the clinical outcome in patients with sepsis and/or acute respiratory distress syndrome [4]. Conversely, Sánchez-Sánchez argued that the improper diagnostic criteria for the intrathoracic blood volume index (ITBVI) led the negative conclusion in Zhang's study [12], which alert the clinical practitioners should be cautious about PiCCO data analysis. But as shown in our research and the other similar studies, the EVLWI and PVPI among the PiCCO parameters exhibited the outstanding predictive power for the prognosis in patients with sepsis [13].
Despite many apparent advantages of the invasive hemodynamic monitor, it doesn't mean that inexpensive examination could be neglected. We found that EVLWI exerted signi cant correlation with oxygenation index, which calculated based on the data from blood gas. EVLWI re ects the situation that the liquid content extravasate outside of the pulmonary vasculature into pulmonary alveolar and interstitium, following by ventilation-perfusion imbalance. Some pathophysiological factors, including lung resection, pulmonary vessels obstruction, and higher positive end-expiratory pressure, may potentially generate the trust-less data of EVLWI [14]. PaO2/FiO2 and Lac widely veri ed in the multiple intensive care scenarios [15][16][17], likewise present the affordable alternatives to invasive methods in our results, especially combined both themselves.

Limitations And Strengths
Several limitations should be clari ed when the application of our instructional results, containing retrospective study design, small sample size, and patients population con ned to sepsis. Even if considering these de ciencies, our study offers another non-inferior choice for prediction of the patients with sepsis under the circumstance without costly PiCCO system.

Conclusions
The hamodynamic parameters including HR, EVLWI, and PVPI exhibit superior predictive performance for early prognosis in patients with sepsis, and the combination Lac with PaO 2 /FiO 2 possess identical diagnostic value. Scatter plot of correlation between PaO2/FiO2 and EVLWI at 24 hours after enrollment Figure 3 ROC cure of the predictive value of combination of Lac and PaO2/FiO2 at 24 hours after enrollment for early prognosis