Large-scale use of therapeutic protocols for early intervention has been hampered by logistic and conceptual problems. Installing invasive hemodynamic monitoring is labor-intensive per se. Use of hemodynamic management protocols assumes that appropriate goals and interventions are known. Traditional invasive hemodynamic monitoring using a pulmonary artery catheter has also been questioned [14-16].

Recently, less invasive techniques have been introduced for monitoring cardiac output, stroke volume or their surrogates [17,18]. An arterial pressure waveform-based method allows continuous estimation of cardiac output without calibration [19-21] and facilitates rapid installation of minimally invasive cardiac output (MICO) measurement in patients with an arterial line. The MICO should reflect trends in cardiac output, even if the absolute values may not be accurate.

All emergency admissions to the ICU were screened. Patients with aortic regurgitation, with an intra-aortic balloon pump or with a pulmonary artery catheter on admission to ICU were excluded. All other patients at least 18 years of age, hemodynamically unstable at or within six hours of ICU admission, and with an arterial line, were eligible for inclusion. Since no additional interventions or deviations from usual care were included, the Ethics Committees of the hospitals (Finnish centers) and the Canton of Bern (Swiss center) approved deferred written informed consent (from the patient or the family; two centers) or deferred written informed consent and prior to study inclusion a statement of no objection to the study from an independent physician (one center).

Hemodynamic instability was defined as the presence of any of the following five criteria: 1) clinically relevant hypotension (systolic arterial pressure < 90 mmHg, acute symptomatic decrease of blood pressure, suspicion of cerebral hypoperfusion, or acute reduction in urinary output related to blood pressure decrease; 2) clinical signs of hypovolemia with or without hypotension (peripheral vasoconstriction, decreased venous filling); 3) oliguria (excluding patients with established oliguric renal failure), with diuresis < 0.5 ml/kg estimated body weight/hour; 4) elevated blood lactate (> 50% above upper normal limit) and clinical suspicion of hypoperfusion; 5) acute alteration of mental status related to hemodynamic alterations.

At baseline, clinical signs of hypovolemia and clinically relevant hypotension were the most frequent criteria of hemodynamic instability, with no significant differences between the groups (Table ​(Table2).2). The number of patients with at least three criteria of hemodynamic instability was higher in the MICO group (P = .05; Table ​Table22).

There were no significant differences between the study groups in volumes or types of fluids received for hemodynamic support in the first six hrs (Table ​(Table3).3). The MICO group received slightly more colloids than the control group after 6 hrs (P = .01) and for the whole 24 hrs (P = .01). There were no differences in the frequency or type of inotropic or vasopressor drugs given, but vasodilators were used more frequently in the control group (P = .02; Table ​Table3).3). A pulmonary artery catheter was inserted in 19% of the patients in the MICO group and in 26% in the control group (P = .11).

At six hours, 45 patients (22%) in the MICO group and 52 patients (28%) in the control group were hemodynamically stable (P = .24). In each group, 19 patients (10%) became stable before 6 hrs and remained stable thereafter (see Additional file 1/Table ​/Table2);2); 34 patients in each group were stable at each time point (6 hrs and 24 hrs). The Kaplan-Meier curves for reaching hemodynamic stability (the time point after which none of the instability criteria were present or reappeared; see Additional file 1/Figure ​/Figure1)1) for 6 hrs showed no difference between the groups (P = .250). The presence of individual criteria of hemodynamic instability at baseline and mechanical ventilation at baseline were compared between stable and unstable patients at 6 hrs, and variables with differences with a P-value less than .10 were included in logistic regression analysis. Only an acute symptomatic decrease in blood pressure at baseline (P = .050) and increased blood lactate at baseline (P = .002) were significant predictors of instability at six hours (for additional details, see Additional file 1/Results).

In the MICO group, stroke volume increased marginally [baseline: 65 mL (20 mL), 6 hrs: 65 mL (20 mL), 24 hrs: 69 mL (20 mL); P = .002] and stroke volume variation decreased slightly over time (baseline: 16% (12%); 6 hrs: 15% (10%); 24 hrs: 14% (10%): P = .05), whereas cardiac output did not change significantly (baseline: 5.9 (2.0) L/minute, 6 hrs: 5.9 (1.8) L/minute; 24 hrs: 6.1 (1.8) L/minute: P = .23].

Neither the ICU mortality nor the hospital mortality was significantly different between the groups (ICU mortality MICO 18%, control 12%, P = .122, relative risk = 1.497

95% confidence interval 0.925 to 2.42; hospital mortality MICO 26%, control 21%, P = .34, risk = 1.209, 95% confidence interval .843 to 1.735). There were more deaths from randomization until 24 hrs after end of study monitoring in the MICO group (MICO n = 16 vs. control n = 4, P = .011). All but one of these deaths (in the MICO group) occurred after withdrawal of treatment due to poor prognosis of the underlying condition (for details, see Additional file 1/Table ​/Table33).

The 24-hr Kaplan-Meier curves (Figure ​(Figure2)2) indicated that patients in the control group were more likely to reach hemodynamic stability (P = .033). After excluding the early deaths, the difference was not significant (P = .11). When stratified according to number of instability criteria at baseline (≥ 3 vs. < 3 criteria), the difference between MICO and control groups was not significant (P = .79 for ≥ 3 criteria; P = .12 for < 3 criteria).

Differences between survivors and non-survivors were tested for study group, presence of each of the individual criteria of hemodynamic instability at baseline and 6 hrs, mechanical ventilation at baseline and 6 hrs, stability at 6 hrs and 24 hrs, and time to achieve hemodynamic stability. Those with a P-value less than .10 were included in logistic regression analysis. Only oliguria (P = .000) and increased lactate (P = .01) at baseline were significant predictors of hospital mortality (for additional details, see Additional file 1/Results and Additional file 1/Tables 4-8).

Hemodynamic monitoring in intensive care has been controversial over the past decades. Use of the pulmonary artery catheter has markedly decreased due to studies suggesting that its routine use does not improve clinically relevant outcomes [14-16,26-31], and the costs associated with its use are substantial. Nevertheless, early hemodynamic instability is indisputably a risk factor for poor outcome [32-35]. Accordingly, alternative techniques for hemodynamic monitoring have been introduced and adopted for widespread clinical use despite the lack of proof of efficacy and despite the costs involved [17-21,36].

We chose to use the proportion of patients stable at six hours, an outcome reflecting the care process, as the primary outcome variable, because early treatment of hemodynamic instability has been perceived as crucial for improving outcome. If this cannot be achieved, any improvements in other outcomes are unlikely.

Our study can be criticized due to the lack of a treatment protocol combined with the use of a monitoring device - a feature that is common to most previous studies on hemodynamic monitoring, especially those on the pulmonary artery catheter [15,16]. Several single-center and small-scale studies have shown improved clinically relevant outcomes when hemodynamic monitoring has been combined with the use of protocols to guide treatment [1-10]. Protocolized care is problematic for various reasons, such as difficulties in defining the "right" intervention, acceptance in routine clinical use, compliance, and efficacy outside clinical trials. Furthermore, new monitoring devices are normally introduced into clinical practice without the use of proven protocols to achieve efficacy.

Despite the lack of protocol use in our study, almost half of the fluids were given during the first six hours, suggesting more intensive treatment during the early ICU stay. We further evaluated (data not shown) patients receiving little fluids for volume expansion during the first six hours (< 5 ml/kg); most of these had either a cardiac or respiratory (APACHE III diagnoses cardiac arrest, cardiac failure, acute myocardial infarction, bacterial or aspiration pneumonia) cause for ICU admission, which is likely to explain the restrictive volume supply, or were admitted postoperatively having already received treatment.

We expected that rapid availability of stroke volume and cardiac output at the bedside would help clinicians select interventions resulting in enhanced stabilization. There are several possible reasons why this failed. The reliability and accuracy of arterial pulse wave-based stroke volume and cardiac output technologies to track changes can be questioned [20,21,37]. Hence, the clinicians possibly did not trust the information. Indirectly, more frequent use of the pulmonary artery catheter to gain more hemodynamic data in the control group speaks against this. All participating centers had 24-hour-a-day, 7-day-a-week in-unit coverage by intensive care physicians, and all centers used invasive hemodynamic monitoring routinely. This may have facilitated the care process in both groups, and any potential improvements would be difficult to achieve. We consider this explanation unlikely, despite the substantially lower than expected hospital mortality (MICO 26% vs. 44%, control 21% vs. 43%). Despite the high physician coverage, the proportion of patients who became hemodynamically stable within 6 hrs and remained stable was low (only 10% in both groups), and less than half of the patients reached stability within 24 hrs. This was clearly less than expected in our power calculations, and would have reduced the chance of finding any difference. The lower than expected statistical power is probably not relevant, because not even a numeric trend in favor of MICO was seen for any relevant outcome. More importantly, the low rate of reaching stability suggests that the care process could have been improved. We consider that our results provide another example of the difficulties in changing clinical practice. In the future, combining such technologies with care protocols should be tested.

Patients in the control group were more likely to achieve hemodynamic stability at 24 hrs than patients in the MICO group. Despite lack of differences in achieving stability at six hours and in-hospital mortality, the possibility that stroke volume and cardiac output data delayed interventions should be considered. The similarity of the treatments, including the marginally higher use of colloids after six hours in the MICO group, speaks against this. The more frequent use of vasodilators in the control group is also unlikely to explain the difference at 24 hrs due to the small number of patients with vasodilators. Also, there were no differences between the groups in the frequency of each of the hemodynamic instability criteria in patients still in the ICU at 24 hrs (see Additional file 1/Tables 3-7). Subtle differences in the case mix are more likely to explain this result, especially the substantially higher number of patients in the MICO group who died early after discontinuation of treatment due to poor prognosis of the underlying disease. Most importantly, these early deaths in the MICO group markedly increased the proportion of patients unstable at 24 hrs. In addition, patients in the MICO group tended to have more instability criteria at baseline (P = .06), and more MICO patients had at least three instability criteria, suggesting more severe illness at baseline.

• MICO did not facilitate hemodynamic stabilization during the first 24 hrs of intensive care.

• Additional monitoring with MICO per se had no effect on the care process (additional monitoring, fluid and vasoactive therapy).

The Department of Intensive Care Medicine at the Bern University Hospital has, or has had in the past, research contracts with Abbott Nutrition International, B. Braun Medical AG, CSEM SA, Edwards Lifesciences Services GmbH, Kenta Biotech Ltd, Maquet Critical Care AB, Omnicare Clinical Research AG, and Orion Corporation; and research and development/consulting contracts with Edwards Lifesciences SA, Maquet Critical Care AB, and Nestlé. The money is/was paid into a departmental fund; no author receives/received any personal financial gain.

The Department of Anesthesiology and Intensive Care at Kuopio University Hospital has research contracts with Orion Corporation, Eli Lilly and Arstellas. The money is paid to the institution.

The Department of Anesthesiology and Intensive Care at Tampere University Hospital received an unrestricted grant from Eli Lilly for an international scientific symposium in 2009. The money was paid to the institution.

See related commentary by Cecconi and Bennett, http://ccforum.com/content/15/4/173