PDF
Citation
Print
INTRODUCTION
Management of patients in shock poses great challenges for physicians of all relevant specialties, including emergency physicians, cardiologists, and intensivists, as it requires rapid decision-making that is not always obvious. One of the main difficulties is differential diagnosis among the types of shock, i.e., hypovolemic, cardiogenic, obstructive, or distributive, as it largely depends on difficult measurements of cardiac output (CO) and pulmonary capillary wedge pressure. The boundaries between them are not always clear, and symptoms often overlap [1,2].
The Swan-Ganz catheter (SGC) or a similarly designed thermodilution pulmonary artery catheter (PAC) is the gold standard for CO measurement. However, this technique has several limitations and complications, mostly due to its invasive nature and general lack of availability in healthcare settings outside the intensive care unit (ICU) [3]. Several methods of non-invasive or minimally invasive CO measurement have been developed; however, none are interchangeable with PAC and thermodilution [4]. Because of the lack of a “one size fits all” solution, every device and method should be used, and results should be interpreted according to the patient’s clinical condition [5,6].
Several different methods are available for estimating CO by pulse contour analysis, based on either models of human circulation or other wave-analysis algorithms [7-9]. Among them, the ARCSolver algorithm [10] combined with a built-in automated, brachial cuff–based oscillometric device (Mobil-O-Graph, IEM GmbH) allows simultaneous non-invasive estimation of brachial blood pressure (BP) and central hemodynamic parameters (e.g., aortic BP, aortic stiffness, pressure wave reflections), including CO. The Mobil-O-Graph (MG) apparatus is commercially available, has received U.S. Food and Drug Administration and European (Conformité Européenne) approval, and has been validated by the British Hypertension Society and European Society of Hypertension for brachial BP measurement [11,12]. Its algorithms have been positively evaluated for feasibility, accuracy, and reproducibility of brachial BP, aortic BP, aortic stiffness, and pressure-wave reflection in both static and ambulatory modes of operation [13-16]. Moreover, it has been shown that, although the MG apparatus underestimates CO, its measurements exhibit satisfying reproducibility and an adequate level of agreement with the gold-standard method (SGC-thermodilution) [17].
We evaluated the potential clinical benefit of non-invasive bedside CO monitoring, as assessed by the MG apparatus, in critically ill patients with shock. This is important in settings where an SGC is not available and the diagnostic resources are limited, such as remote health care units. The aim of the present study was to test the hypothesis that decisions based on CO as assessed by this device are clinically useful and non-harmful. We compared decisions among critically ill patients under shock for basic therapeutic interventions based on standard ICU monitoring (SIM) alone or in combination with CO data acquired from either the SGC and thermodilution technique or an MG apparatus.
MATERIALS AND METHODS
The study protocol complies with the Declaration of Helsinki, and it was approved by the Institutional Review Board of Laikon Hospital (No. 790; Athens; 27 Jun, 2016) and the National Data Protection Authority (Protocol No. ΓΝ/ΕΞ/4507-1/03-08-2016, Athens; 3 Aug, 2016). Informed consent was obtained from the legally authorized representatives of the patients, upon enrollment to the study.
Study Population
Twenty-nine patients consecutively hospitalized with shock in the ICU of the Laiko General Hospital of Athens were initially enrolled. All patients required CO monitoring using a PAC (thermodilution) by clinical indication for diagnostic and therapeutic purposes as part of hemodynamic monitoring due to shock of undetermined etiology. SGCs were placed and the type of shock was determined based on international guidelines [18,19]. Nine patients were excluded due to inadequate collection of data (Figure 1).
Study Protocol
The study began in March 2016 and ended in June 2018. Eligibility criteria required that the recruited patients be hospitalized with shock in the ICU of Laiko Hospital and be hemodynamically monitored with an SGC by clinical indication, according to the judgment of the attending intensivist. Soon after restoring a relatively stable hemodynamic status, simultaneous CO estimation was performed in each patient by SGC and a non-invasive, automated, oscillometric, cuff sphygmomanometer (MG). Data collection was consistent, and the primary researcher attempted to obtain informed consent from the legal representatives of the patients. Data from the 20 patients who met the final eligibility criteria were retrospectively provided to three experienced intensivists (physicians A, B, and C) who were asked to make decisions for each patient based solely on these data. Six possible therapeutic interventions were reported considering the need to modify (start/stop or increase/decrease) blood transfusions; use of inotropes, vasoconstricting agents, and diuretics; and administration of oxygen or fluids. Physician A made guideline-based decisions by evaluating data acquired from SIM (BP, body temperature, electrolyte levels, central venous pressure, oxygen saturation, respiratory rate etc.) and an SGC (CO, total peripheral resistance, reflection coefficient, cardiac index etc.). Physician B made therapeutic decisions based on data acquired from SIM and MG measurements, while physician C (SIM) made treatment decisions by evaluating the commonly available ICU data, as presented in Table 1, without any knowledge of CO.
This was a post-hoc analysis regarding the potential clinical value of non-invasively derived CO to guide decision-making process, as this study was initially designed to compare MG-derived CO versus the invasive standard. The experienced ICU physicians made decisions based on current guidelines and their clinical experience, evaluating hemodynamic parameters as presented in the text and Table 1. Their main goal was to identify the primary etiology of shock and determine a central treatment strategy based on the above described predefined six actions (blood transfusion use of inotropes, vasoconstricting agents, diuretics; and administration of oxygen or fluids).
Invasive and Non-invasive Measurements of CO
PAC-thermodilution was performed as previously described [20-22]. Non-invasive estimation of CO was performed using the MG device. All patients were placed in the supine position and two consecutive measurements were performed at a 1-minute interval using appropriately sized brachial cuffs in a temperature-controlled room (20 °C–23 °C). The technique records sequential brachial pressure waves using an appropriate cuff and the oscillometry principle [10]. Initially, the device measures the brachial BP conventionally; the cuff, which is equipped with a high-fidelity pressure sensor (MPX5050, Freescale Inc.), then reinflates to the level of the diastolic BP for approximately 10 seconds, at which point it acquires the corresponding brachial pressure waveforms. The sensor is connected to a 12-bit analog-to-digital converter by means of an active analog-band bass filter (0–25 Hz). The digitalized pressure signals are processed and analyzed, employing a three-step algorithm (ARCSolver algorithm). Initially, the single pressure waves are assessed for plausibility by examining minima position and corresponding wavelengths. To identify minima, an iterative approach involving the analysis of higher-order time derivatives of the pressure signal is used. In the second phase, all single pressure waves are compared to one another to recognize artifacts; subsequently aortic pulse waves are generated. Stroke volume (SV) was derived from pulse contour analysis determined with oscillometry. The SV is proportional to the area of the flow curve at the systolic phase. The aortic flow was calculated from the pressure using the three-element windkessel model, taking into account aortic characteristic impedance, aortic compliance, and peripheral resistance. The SV was derived from the time lag between the aortic pressure and flow curve, and CO was calculated by multiplying the SV with the heart rate [10,23,24].
Statistical Analysis
Continuous variables are presented as mean± standard deviation (SD), and categorical variables as absolute frequencies and percentage (%). Cohen’s kappa agreement coefficient was used to assess agreement between different treatment decisions and was preferred over simple frequencies as factors in the possibility of the agreement occurring by chance [25,26]. The kappa statistic has been the most frequently used method of inter-rater agreement, despite its limitations, for the past 50 years. Many extensions have been proposed [27-29], but none have been embraced by health researchers. A non-parametric Mann-Whitney test [27] was used to compare the number of interventions per patient, as a normal distribution from a Silk-Shapiro test [28] was not found for the number of agreed interventions. P-values lower than 0.05 indicated resent statistical significance. Statistical analysis was performed in IBM SPSS (IBM Corp.).
RESULTS
Of the 29 patients initially enrolled in the study undergoing CO monitoring by PAC-thermodilution, 9 were excluded due to missing data. Descriptives of the remaining 20 patients are presented in Table 2. The hemodynamic parameters during the CO measurement are provided in Table 1. During CO measurement 75% of the patients were under vasoconstricting treatment and 5% under inotropes. The ICU survival rate of the studied population was 65%. The therapeutic decisions based on sources of data by each physician (A, B, and C) are delineated descriptively in Figure 2, and Cohen’s kappa agreement coefficients of treatment decisions between them are shown in Figures 3B and 4.
Blood Transfusion
SIM+MG-based decisions (physician B) agreed with decisions based only on SIM (physician C) in 18 patients (90%, k=0.35, P<0.05), and SIM+SGC-based decisions (physician A) agreed with those based only on SIM in 17 patients (85%, k=0.48, P<0.05) (Figure 2). The kappa agreement between physicians A and B was 0.46 (P=0.015) (Figure 3). When examining only decisions suggesting that “action is needed” based on the extra data (SGC or MG), the kappa agreement between the modification of decisions by SGC data (from physician C to physician A) and MG data (from physician C to physician B) was 0.62 (P<0.001) (Figure 4).
Inotropes
Combined SIM+MG-based decisions (physician B) agreed with SIM-only decisions (physician C) in 19 patients (95%, k=0.83, P<0.001), and SIM+SGC-based decisions (physician A) agreed with SIM-only decisions in 15 patients (75%, k=−0.14, P=0.53) (Figure 2). The kappa agreement between physicians A and B was 0.23 (P=0.264) (Figure 3). When examining only those decisions suggesting that “action is needed” based on the extra data (SGC or MG), the kappa agreement between the modification of decisions by SGC data (from physician C to physician A) and MG data (from physician C to physician B) was 0.29 (P=0.005) (Figure 4).
Fluids
SIM+MG-based decisions (physician B) agreed with SIM-only decisions (physician C) in 15 patients (75%, k=0.7, P<0.01), and SIM+SGC-based decisions (physician A) agreed with those based only on SIM in 16 patients (80%, k=0.61, P<0.01) (Figure 2). The kappa agreement between physicians A and B was 0.68 (P=0.002) (Figure 3). When examining only those decisions suggesting that “action is needed” based on the extra data (SGC or MG), the kappa agreement between the modification of decisions by SGC data (from physician C to physician A) and MG data (from physician C to physician B) was 0.48 (P=0.014) (Figure 4).
Diuretics
SIM+MG-based decisions (physician B) agreed with SIM-only decisions (physician C) in 14 patients (70%, k=0.21, P=0.329), and SIM+SGC-based decisions (physician A) agreed with SIM-only decisions in 15 patients (75%, k=0.39) (Figure 2). The kappa agreement between physicians A and B was 0.66 (P<0.01) (Figure 3). When examining only those decisions suggesting that “action is needed” based on the extra data (SGC or MG), the kappa agreement between the modification of decisions by SGC data (from physician C to physician A) and MG data (from physician C to physician B) was 0.65 (P<0.001) (Figure 4).
Vasoconstriction
SIM+MG-based decisions (physician B) agreed with SIM-only decisions (physician C) in 14 patients (70%, k=−0.18, P=0.3), and SIM+SGC-based decisions (physician A) agreed with SIM-only decisions in 8 patients (40%, k=−0.1) (Figure 2). The kappa agreement between physicians A and B was 0.32 (P=0.026) (Figure 3). When examining only those decisions suggesting that “action is needed” based on the extra data (SGC or MG), the kappa agreement between the modification of decisions by SGC data (from physician C to physician A) and MG data (from physician C to physician B) was 0.44 (P=0.002) (Figure 4).
Only two patients received no therapeutic intervention, in one of which all three physicians decided unanimously to not administrate anything. The mean (±SD) difference between the SGC- and the MG-measured CO was 1.3 L (±1.39). The number of clinical interventions per patient between SGC-based and MG-based CO (Figure 3A) did not differ significantly (P=0.521). Additionally, the numbers of interventions per patient for which physician C (decisions based only on SIM data) agreed with physician A (SIM+SGC) and physician B (decisions based on SIM+MG data) agreed with physician A (SIM+SGC) were not significantly different, independent of the level of CO difference (greater or less than 1.3 L) (P=0.851).
DISCUSSION
In this pilot study, we enrolled 20 patients in shock under treatment after restoring relatively stable hemodynamic conditions and demonstrated for the first time that decisions based on SIM- and MG-acquired data were closer overall to those based on SIM- and SGC-acquired data than were the decisions based on SIM only. Furthermore, it was shown that there is a benefit in the supplementary use of MG for early diagnosis and resuscitation of critically ill patients in shock of unknown etiology, particularly regarding supplementation of fluids, diuretics, and vasoconstrictive agents compared with relying solely on clinical data.
Shock is a condition with high mortality and is challenging to manage, not only in the ICU, but also in the emergency room, the hospital ward, the primary care units and, generally, in any setting without access to ICU equipment. Its prognosis largely depends on timely reversal [29], and primary resuscitative efforts to stabilize the patient’s breathing and circulation and formulation of a working hypothesis must take place before determination of a definitive cause of shock [30]. Nevertheless, ongoing differential diagnosis can be complicated, and the frequent coexistence of more than one overlapping cause can hinder classification of shock into one of four major types [31], as seen in eight patients of this study (Table 2). Knowledge of CO and continuous monitoring can answer many initial diagnostic questions, including the etiology of hypotension [32].
Cohen’s kappa agreement between the SGC- and MG-derived decisions of physicians A and B, respectively, indicated a weak level of agreement (0.2<k<0.4) on supplementation or discontinuation of inotropes and vasoconstrictive agents, moderate agreement (0.4<k<0.6) on the need for blood transfusion, a substantial level of agreement (0.6<k<0.8) on the administration of fluids and diuretics, and a perfect agreement on supplementation of oxygen (k=1) in these 20 patients. Interestingly, apart from the need for blood transfusion, Cohen’s kappa agreement coefficients between the SGC- and MG-derived decisions were slightly stronger than the ones between SGC-derived and SIM-based decisions in all therapeutic interventions (Figure 3). The difference was greater, particularly in decisions regarding inotropes and vasoactive agents, where the kappa agreement between physicians A (SGC+SIM) and C (SIM only) was negative. This observation is consistent with clinical experience and published literature, as there is no easy way to predict the patients’ need or response to these medications with SIM.
Agreement percentages between all decision-making methods in these 20 patients were relatively high (>75% in most interventions, aside from the use of vasoconstrictive agents). However, as seen in Figure 2, high agreement percentages tended to be associated with patients for whom clinicians agreed not to take action regarding a specific intervention. We, therefore, compared the variation in MG- and SGC-acquired data to the decisions based on SIM only. Kappa agreements between SGC- and MG-derived decisions were stronger—although not to a statistically significant level—in the interventions regarding blood transfusion, inotropes, and vasoconstrictive agents and weaker for fluid-related interventions (Figure 4) compared with those presented in Figure 3B.
Limitations
This pilot study had a limited sample size that did not allow thorough statistical analysis. Moreover, because this study was initially designed to use and focused on the agreement of CO between the non-invasive and invasive methods, without using hard outcome end points, we were unable to evaluate the final clinical impact of these clinical decisions. Another major limitation of the present post-hoc analysis is that the three certified and experienced ICU physicians provided their clinical decisions not in a real-time setting, but based only on available data (Table 1). We acknowledge that the present results may have been affected by subjective criteria, namely the personal judgment of the intensivists, with respect to application of clinical guidelines, although such judgments are always present in clinical decisions. Finally, it must be noted that the present results were derived from patients in shock under treatment, after restoring relatively stable hemodynamic conditions and cannot be extrapolated to all conditions. However, these hemodynamic conditions represent a large population with similar hemodynamic status and are common in clinical practice.
Conclusions
Among 20 critically ill patients in the setting of an ICU, CO as assessed by the MG apparatus led to clinical decisions that had a fair agreement with those produced by the gold standard of CO measurement (SGC). Due to the post-hoc analysis of this pilot study, no generalized conclusions can be safely extrapolated. However, if this observation is confirmed by larger, randomized studies on a variety of critically ill patients with different underlying comorbidities, MG could be introduced into daily clinical practice as a low-cost, first-line, bedside, non-invasive, user-friendly technique for simultaneous continuous BP and CO monitoring. Such an approach could prove useful to physicians who serve in remote and rural primary care units. In such settings, physicians often do not have access to or expertise with an SGC or other modalities for CO assessment (e.g., cardiac ultrasound) or, even worse, have access to less than the herein evaluated SIM data. Clinical trials should be performed in such settings with mortality endpoints to test the clinical value of CO monitoring with MG apparatus.
KEY MESSAGES
▪ It has been previously shown that cardiac output (CO) assessment with a Mobil-O-Graph (MG) apparatus is reproducible and comparable to measurements from a Swan-Ganz catheter (SGC) with thermodilution.
▪ Therapeutic decisions based on standard intensive care unit (ICU) monitoring (SIM) and MG-derived CO data showed higher levels of agreement with those based on SIM and SGC CO measurements compared with those based on SIM alone.
▪ If these findings are confirmed by larger studies, MG could be an effective first-line bedside method for simultaneous continuous monitoring of blood pressure and CO in critically ill patients outside the ICU.
XML Download