Introduction

Monitoring techniques in the intensive care unit (ICU) are primarily used to identify disease patterns and titrate therapies. Understanding of pathophysiological processes may help limit disease progression and promote recovery, so early detection of physiological alterations to guide therapeutic interventions should improve outcomes; however, doubts have been raised about the utility of some monitoring systems, e.g., the pulmonary artery catheter (PAC) [1–3], in the ICU.

In evidence-based medicine, the randomized controlled trial (RCT) represents the ideal study design to validate an intervention; however, demonstration of utility of a monitoring system may be difficult and few monitoring systems have been evaluated by RCT. We performed a systematic review of the literature for RCTs that have evaluated the impact of monitoring systems on outcomes in critically ill patients and in perioperative patients undergoing major procedures.

Methods

We conducted a comprehensive search for all publications of prospective RCTs that focused on the impact of monitoring systems in adult critically ill patients and adult perioperative patients undergoing major procedures. A highly sensitive search strategy was carried out in the Cochrane Central Register of Controlled Trials (CENTRAL) and in MedLine using the approach shown in the electronic supplement. Eligibility assessment and data abstraction were performed independently in an unblinded standardized manner by two reviewers (G.A.O and R.L.C). Interrater reliability between the abstractors for RCT selection was evaluated using the κ statistic. Discrepancies about classification and outcome evaluation were resolved by author consensus.

A study was selected when it compared a monitored group of patients with a non-monitored control group or when the health-care team was unaware of monitoring measurements and so was unable to use them to guide therapy. We defined outcome in general, including morbidity and mortality, optimization of therapeutic strategies, complications, costs, and quality of life. We excluded animal and laboratory studies, non-randomized trials, reviews, letters, meta-analyses, guidelines or commentaries, trials comparing monitoring methods without a non-monitored control group, trials evaluating the impact of medicaments or therapies on measurements obtained by a monitoring system, and pediatric studies.

Selected studies were separated into those evaluating hemodynamic (or tissue perfusion), respiratory, neurological, or metabolic variables and, according to the results obtained for each of their outcome objectives, were classified as positive, neutral, or detrimental. Additionally, we classified studies evaluating mortality as primary or secondary outcome studies.

Results

Study selection

A total of 4175 articles were identified and 67 were included (Fig. 1). Interobserver agreement for selection and final classification of the studies was high: κ statistic 0.90. Seven studies [4–10], which assessed two monitoring devices in the same protocol, were enrolled in two different categories.

Fig. 1
figure 1

Study selection process

Full size image

Hemodynamic and perfusion monitoring (Fig. 2)

Arterial pressure

No RCTs evaluating the impact of arterial pressure monitoring on outcomes when used in the ICU or operating room were identified (Table 1).

Fig. 2
figure 2

Summary of outcomes in randomized controlled trials (RCTs) on hemodynamic and perfusion monitoring

Full size image
Table 1 Randomized controlled studies of hemodynamic and perfusion monitoring techniques. ARDS, acute respiratory distress syndrome; CI, cardiac index; CVP, central venous pressure; EVLW, extravascular lung water; ICU, intensive care unit; IQ, interquartile; LOS, length of stay; NA, not applicable; PAC, pulmonary artery catheter; PAOP, pulmonary artery occlusion pressure; SD, standard deviation; SOFA, sequential organ failure assessment; SVR, systemic vascular resistance
Full size table

Central venous pressure

Two RCTs evaluated the impact of CVP monitoring on outcome in the ICU [8, 11]. One RCT compared conventional intraoperative fluid management with transesophageal Doppler or CVP monitoring to optimize intraoperative fluid therapy in patients with hip fracture [8]. Although in the conventional group a catheter was placed and CVP recorded by the investigator, the clinician was unaware of these measurements and so unable to use them to guide therapy. Patients in the CVP and transesophageal Doppler groups had a significantly higher intraoperative fluid balance; however, there were no differences in mortality or morbidity (p = 0.24), although fluid administration guided by CVP monitoring shortened the time before patients were fit for discharge. In the other RCT, fluid challenge targeted at keeping the CVP > 5 mmHg during renal transplant surgery resulted in a greater frequency of onset of graft function within the first three postoperative days than in a control group without CVP monitoring [11].

Pulmonary artery catheter

Sixteen studies assessed a possible influence of the PAC on ICU outcomes: seven in patients undergoing cardiac or major peripheral vascular surgery [6, 9–16]; three in high-risk surgical patients [4, 10, 17]; three in mixed-ICU populations [1, 18, 19]; two in patients with acute lung injury (ALI)/acute respiratory distress syndrome (ARDS) [3, 20]; and one in patients with congestive cardiac failure [2]. Of 15 studies evaluating mortality as a primary or secondary outcome when a PAC was placed [1, 2–4, 6, 9, 10, 12–15, 17–20], nine used predefined hemodynamic targets [2–4, 9, 10, 14–17]; of these, only two [4, 17] demonstrated a positive effect on mortality. Mortality outcomes were not improved in studies that did not include a hemodynamic goal protocol.

In high-risk surgical patients, Shoemaker et al. [4] demonstrated lower mortality when a PAC-protocol group was compared with a PAC control group (4 vs. 33%, p < 0.01), as well as with a CVP+PAC-control group (4 vs. 28%, p < 0.02). In this study, the PAC was inserted early (preoperatively) and the PAC protocol established before the development of organ dysfunction. There were no differences in mortality between a CVP group and the PAC control group. Complications, duration of mechanical ventilation, as well as ICU and hospital stays, were less in the PAC-protocol group when compared with the CVP+PAC control group. Again in high-risk surgical patients, Sandham et al. [10] reported no significant differences in hospital mortality among groups managed with or without a PAC (7.8 vs. 7.7%) despite having a PAC-directed protocol. Pearson et al. [6] showed no significant differences in morbidity or mortality, but increased costs, in patients undergoing cardiac surgery when compared with management with a CVP catheter, but no predefined hemodynamic targets were established.

Studies on PAC monitoring in patients undergoing abdominal aortic surgery [12, 13] did not show improved complication rates, duration of intensive care, postoperative hospital stay, or mortality, but no preoperative optimization was conducted and predefined hemodynamic goals were not established. In a study by Joyce et al. [13], all the serious postoperative cardiac complications occurred in patients with impaired left ventricular function, independent of PAC use.

Preoperative hemodynamic optimization with a PAC has been evaluated in four RCTs in patients undergoing major vascular surgery [9, 14–16] and one in patients with proximal hip-fracture repair [17]. Two of these RCTs demonstrated a positive impact on ICU outcomes [14, 17]. Berlauk et al. [14] showed fewer intra- and postoperative adverse events, less early graft thrombosis, and lower mortality rates than in the control group. Likewise, Shultz et al. [17] showed a lower mortality rate in patients undergoing surgical repair of proximal hip fracture. In contrast, three RCTs [9, 15, 16] showed no improvement in morbidity, intraoperative events, ICU stay, hospital stay, or mortality when preoperative optimization was guided by a PAC.

In ALI/ARDS, Richard et al. [20] found no differences in morbidity or mortality in patients randomized to PAC or no PAC. The ARDS Network recently reported similar mortality and organ function measures in patients with ALI managed by PAC-guided compared with CVP-guided therapy, but higher complication rates [3]. In a mixed ICU population, Rhodes et al. [19] also found similar mortality rates, organ dysfunction, and ICU or hospital stays, in patients managed with or without a PAC. Likewise, Harvey et al. [1] found no significant differences in hospital mortality or length of ICU and hospital stays in patients managed with or without a PAC in a mixed ICU population. In these studies, no predefined hemodynamic goals were proposed in the PAC groups.

In patients with severe symptomatic and recurrent heart failure [2], use of a PAC did not significantly affect days alive and out-of-hospital in the first 6 months, mortality, number of days hospitalization, or quality of life.

Extravascular lung water

Eisenberg et al. [21] evaluated the impact of extravascular lung water (EVLW) monitoring to guide therapy in a mixed ICU population. In the routine management group, EVLW was measured, but results were unknown to the primary care physicians, while the protocol group used an EVLW-guided algorithm. In-hospital mortality was not significantly lower in the protocol group (54 vs. 65%). The ICU mortality was similar in all hemodynamic subsets, except for patients with an initially elevated EVLW and pulmonary artery occlusion pressure (PAOP) < 18 mmHg, where the protocol group had lower mortality rates.

Mitchell et al. [22] showed a significant decrease in ventilator days and ICU stay in critically ill patients managed with EVLW monitoring; however, differences in ICU (35 vs. 47%) and hospital (56 vs. 65%) mortality rates did not reach statistical significance.

Transesophageal echocardiography and esophageal Doppler

Six RCTs evaluated the impact of transesophageal echocardiography (TEE) or esophageal Doppler in surgical patients: two [8, 23] in patients undergoing proximal femoral fracture repair; two in patients undergoing colon surgery [24, 25]; one in patients undergoing major surgery [26]; and one in patients undergoing cardiac surgery [27]. All the studies used a protocol guided by TEE/esophageal Doppler to optimize fluid loading. A reduction in postoperative complications was reported in all the studies, with a significant reduction in in-hospital length of stay in four studies [8, 23, 26, 27]. None of the studies evaluated mortality.

Gastric tonometry

Although a number of observational studies have demonstrated that intramucosal pH (pHi) is a good indicator of morbidity [28, 29] and mortality [30], we found only five RCTs that assessed the influence of gastric tonometry on outcomes [31, 32–35]. Gutierrez et al. [31] studied 260 medical, postoperative, and trauma patients admitted to the ICU. Patients were randomized to a gastric tonometry-guided protocol or conventional treatment in which pHi values were recorded but not reported to the health-care team. In patients admitted with a low pHi, survival was similar (37 vs. 36%), whereas for those admitted with a normal pHi, it was significantly greater in the protocol than in the control group (58 vs. 42%, p < 0.01). Ivatury et al. [32] evaluated optimization of therapy in post-resuscitated trauma patients using a pHi-guided protocol: normalization of pHi in the first 24 h was associated with greater survival compared with the control group; however, in another RCT including trauma patients [35], resuscitation guided by gastric tonometry did not produce significant differences in mortality rates, ventilator days, organ dysfunction, or length of stay.

In a study by Pargger et al. [33], patients were randomized to pHi-guided therapy or control (pHi was registered but treating physicians were blinded to pHi values) after elective repair of infrarenal abdominal aneurysm. Low pHi values (< 7.32) and their persistence were predictors of major complications, but treatment aimed at correcting low pHi values did not improve postoperative outcomes. In resuscitated acutely ill adult patients admitted to the ICU, Gomersall et al. [34] showed no clinically or statistically significant differences in ICU or hospital survival, organ function, or duration of stay in an intent-to-treat analysis for patients randomized to pHi-guided therapy (pHi measurements were obtained in the control group, but not used to guide treatment).

Mixed (SvO2) and central (ScvO2) venous oxygen saturation

Only six studies have evaluated the impact of these monitoring variables on outcomes [6, 36–40].

Jastremski et al. [36] analyzed the impact and cost-effectiveness of SvO2 monitoring in medical ICU patients. There was no difference in the number of ICU days, decrease in potentially adverse hemodynamic events, or survival between the groups, although the authors still considered that SvO2 measurement could have a reasonable cost-effectiveness ratio. In a mixed population of critically ill patients, Gattinoni et al. [37] showed no differences in morbidity or mortality when comparing a normal SvO2 or supranormal cardiac index (CI) with a normal CI control group.

Three studies evaluated SvO2 monitoring in cardiac and vascular surgery patients [6, 38, 39]. Preoperative optimization of cardiovascular function using a target SvO2 > 65% did not reduce intra- or postoperative complications in patients undergoing elective peripheral vascular surgery [38]. Pearson et al. [6] found no significant differences in length of ICU stay, morbidity, or mortality among patients monitored by conventional PAC, SvO2 – PAC, or CVP; however, Polonen et al. [39] reported a reduction in hospital stay and postsurgical morbidity in cardiac surgery patients randomized to SvO2 and lactate monitoring in the first 8 postoperative hours. A post-hoc analysis showed improved survival at 6-months and 1-year after randomization in patients who achieved the target SvO2 and lactate values.

In an RCT evaluating early goal-directed therapy in the treatment of severe sepsis and septic shock, Rivers et al. [40] observed that patients randomized to a protocol group who received treatment guided by ScvO2, in addition to other variables, had less in-hospital mortality than a control group (30.5 vs. 46.5%; p < 0.009) and major improvements in organ function.

Oxygen delivery (DO2) and consumption (VO2)

We identified eight studies on DO2 and VO2 monitoring in ICU patients in which, for the control group, treating physicians were blinded to measurements or the measurements were not used to guide some form of therapy. Four studies in high-risk surgical and severe trauma patients [4, 41–43] showed that a deliberate perioperative increase in CI, DO2, and VO2 reduced mortality and complications; however, two studies in mixed ICU populations [44, 45] and another in high-risk surgical patients [10] showed no improvement in organ failure or mortality rates.

Other perfusion monitoring

We identified no RCT that evaluated the influence of other hemodynamic or perfusion monitoring techniques, including hepatic venous oxygen saturation, abdominal pressure monitoring, near-infrared spectroscopy, and mucosal laser Doppler flowmetry, on outcomes.

Respiratory monitoring (Fig. 3)

Pulse oximetry

We identified seven RCTs that evaluated outcomes with pulse oximetry: four in the operating or postoperative room [46–49], two in the ICU [5, 7], and one on a specialized, postsurgical hospital floor [50] (Table 2).

Fig. 3
figure 3

Summary of outcomes in RCTs on respiratory monitoring

Full size image
Table 2 Randomized controlled studies of respiratory monitoring techniques. ARDS, acute respiratory distress syndrome; ICU, intensive care unit; LOS, length of stay; PEEP, positive end-expiratory pressure
Full size table

Several studies indicated that pulse oximetry could help to detect hypoxemic events [46, 47]. Moller et al. [48] demonstrated no significant differences in late cognitive dysfunction when perioperative monitoring with pulse oximetry was employed. A large RCT evaluating pulse oximetry in 20,802 surgical patients [49] reported no differences in cardiovascular, neurological, or infectious complications, or in-hospital deaths. Niehoff et al. [5] showed the potential utility of pulse oximetry and capnography in postoperative weaning from mechanical ventilation in a subset of only 24 post-cardiac surgery patients. Finally, Ochroch et al. [50] demonstrated that continuous pulse oximetry did not prevent readmissions to the ICU when cardiac and thoracic postoperative patients were monitored in a specialized postsurgical care floor.

Capnography

We identified three studies of capnography monitoring [5, 7, 51], including two also dealing with pulse oximetry monitoring [5, 7]. In the third study in this group, Helm et al. [51] used capnography to monitor ventilation in prehospital trauma victims, and found a higher proportion of patients who were “normoventilated” at hospital admission in the monitor group than in the monitor-blind group (63 vs. 20%, p < 0.001), but final in-hospital outcomes were not reported.

Respiratory mechanics

A number of studies showed that respiratory mechanics monitoring can guide ventilator adjustments as part of protective ventilation strategies in acute respiratory failure [52–56]. Amato et al. demonstrated that a protective ventilation strategy adjusting the positive end-expiratory pressure (PEEP) to above the lower inflection point (LIP) of the static pressure-volume curve and maintaining end-expiratory plateau pressure and peak-inspiratory pressures below 20 and 40 cmH2O, respectively, improved lung function in patients with ARDS [52], increased the chance of early weaning, and reduced mortality [54]. The control groups in these two studies did not consider airway pressure vigilance.

Ranieri et al. [55] reported a decreased inflammatory response in patients undergoing a protective ventilation strategy using tidal volumes of 5–8 ml/kg and PEEP adjusted above the LIP of the static pressure-volume curve. Recently, Villar et al. [56] showed a reduction in ICU mortality (53 vs. 32%), in-hospital mortality (55 vs. 34%), and ventilator-free days at day 28 (6 vs. 11) with a strategy using low-tidal volume and PEEP adjusted according to the LIP of the static pressure-volume curve.

Other studies did not show a significant difference in mortality rates associated with limited airway pressures [57–59]. An ARDS network study [60] showed that the reduction in mortality rates in patients with ARDS was related to a limitation in tidal volumes more than plateau pressures, questioning the importance of the monitoring of airway pressures.

Neurological monitoring (Fig. 4)

Intracranial pressure

We identified no RCT evaluating the effects of intracranial pressure (ICP) monitoring on outcomes in acutely ill patients.

Fig. 4
figure 4

Summary of outcomes in RCTs on neurological monitoring

Full size image

Depth of anesthesiaand continuous electroencephalogram

No RCTs evaluating the impact of continuous depth of anesthesia and continuous electroencephalogram (EEG) monitoring devices on outcomes in the ICU were identified.

Ten studies evaluated the effects of bispectral index (BIS) monitoring during major surgical procedures, including one in patients undergoing abdominal surgery [61], two in mixed major and minor procedures [62, 63], three in cardiac surgery [64–66], two in major surgery [67, 68], one in neurosurgery [69], and one in the emergency department [70] (Table 3). All these studies showed a reduction in anesthetic recovery times and in adverse effects of the anesthetic agent.

Table 3 Randomized controlled studies of neurological monitoring techniques. ED, Emergency department; BIS, bispectral index
Full size table

Other neurological monitoring

We found no RCTs assessing the effects of jugular venous bulb saturation, transcranial Doppler, near-infrared spectroscopy, or cerebral microdialysis.

Metabolic monitoring

We identified no RCTs that assessed the impact of indirect calorimetry in the ICU.

Discussion

Therapy based only on bedside clinical observations is often subjective, sometimes inadequate, and potentially harmful [71]. A wide range of clinical and technological tools allows multiple variables to be assessed at the bedside in an invasive or non-invasive, continuous, or intermittent manner. As monitoring of important physiological variables can guide a number of therapeutic interventions, one could anticipate a strong body of evidence demonstrating that monitoring improves outcomes; however, our literature review illustrates that few RCTs document a positive impact of any monitoring system on outcome.

Multiple limitations can prevent the accurate study of the impact of monitoring on outcomes. Firstly, the need for the monitoring system may be so obvious that it has never been tested, or to test it would be considered unethical. This is true for direct information about so-called vital signs. An RCT on electrocardiogram monitoring in patients with acute myocardial infarction, or on arterial pressure monitoring in shock states, would not be acceptable. Sometimes, also, the benefit/risk profile of a monitoring system may clearly be very good, e.g., for pulse oximetry monitoring; however, it is interesting that even very large RCTs with pulse oximetry could not show significant differences in late cognitive dysfunction, morbidity, or mortality rates [48, 49]. In addition, monitoring devices may never have been subjected to an RCT because they evaluate variables considered to be a sound basis for therapeutic interventions. ICP monitoring is a good example of this. ICP monitoring has not been subjected to an RCT evaluating its impact on outcomes; however, the prognostic value of ICP monitoring has been clearly established [72, 73]. Some observational cohort studies in patients with severe head trauma have demonstrated favorable effects of aggressive ICP-guided treatment on mortality when compared with historic controls [72, 74] or with patients managed in hospitals that did not use this type of monitoring [75]. ICP monitoring has been shown to reveal episodes of intracranial hypertension in severe head trauma patients without tomographic signs of intracranial hypertension [73]. Likewise, observational studies have suggested that monitoring of jugular bulb oxygen saturation (SjO2) could enable early identification of transitory episodes of cerebral ischemia which would otherwise go undetected and untreated [76, 77] resulting in poorer neurological outcomes [78]. On the other hand, one could apply the same rationale to the PAC as the monitored cardiovascular variables have prognostic value, are less accessible by other techniques, and can be influenced by therapy.

Secondly, the extreme heterogeneity of the populations studied and the multiple therapeutic interventions that may be used can create excessive noise and limit study interpretation.

Thirdly, the way in which the monitoring system is used, and the accuracy with which monitoring data are collected and interpreted, can influence the final results. Monitoring systems cannot improve outcomes per se. The impact of a monitoring system will depend on its correct usage. There may also be significant variability in the interpretation of measured variables by physicians, perhaps particularly true for the PAC [79]. It is difficult to differentiate between poor technology and good technology used poorly. On some occasions, the frontier between a direct monitoring benefit and the benefit of a monitoring-guided therapy is difficult to determine. This is the case in some studies using CI, DO2, and VO2 as resuscitation end points in critically ill patients [80–86]. All these studies used a “supranormal” hemodynamic target in the protocol group, whereas the control groups were treated to achieve lesser hemodynamic goals; hence, these studies were not included in our review, although other studies on “supranormal” hemodynamic targets, which included a control group that received standard treatment (when the clinician was unaware of monitoring measurements or when no effort was made to correct the measures obtained through the monitoring device), were included [37, 41–45].

The benefits of monitoring can also be influenced by the timing of the interventions used to correct the altered measurements. Early goal-directed therapy guided by ScvO2 in the first 6 h after admission to an emergency department may improve mortality in severe sepsis and septic shock [40], but restoring SvO2 in critically ill patients may not show the same benefit when this is achieved later in the course of disease [37]. Likewise, increasing DO2 to supranormal levels may improve outcome in the early perioperative period [14, 17, 38, 39] but not later [37]; however, in these conditions, the question can be raised as to whether the clinical benefit may have been the result of more attentive and more aggressive therapy, with or without monitoring systems.

The safety and utility of invasive hemodynamic monitoring devices, including the PAC, have been debated for years. The PAC debate was highlighted in 1996 when the results of a prospective cohort study [87] suggested an increased 30-day mortality, greater intensity of care, and longer ICU stays in patients managed with a PAC. A recent European study, using the same methodology (based on a propensity score), could not reproduce these observations [88]. RCTs have not shown increased mortality or morbidity in patients managed with a PAC, except for one study that showed an increase in adverse in-hospital events attributable to higher rates of PAC infection [2]. The lack of demonstrated benefit with the use of PACs could be due to the lack of standardized therapeutic protocols for managing patients in the PAC groups. The same may or may not apply to other monitoring studies.

This literature review has its limitations. Firstly, monitoring is difficult to define; in particular, the distinction between monitoring and diagnostic techniques is often blurred. In addition, monitoring systems may not always reliably measure what they are supposed to. Secondly, one cannot define a homogenous patient population or setting where the monitoring system was assessed. As it is sometimes difficult to establish a distinction between the operating room and the ICU, we extended our review to patients undergoing major procedures, even though the data obtained may not apply fully to critically ill patients in the ICU; for example, data obtained by esophageal Doppler during major surgery [8, 23–26] may not be applicable to guiding fluid challenge in the ICU. Thirdly, due to the range of monitoring systems studied and the limited number and small size of available studies in several groups, we did not perform a meta-analysis, but results of meta-analyses for several individual monitoring systems, e.g., the PAC [89], have recently been published.

Conclusion

In conclusion, although it is often argued that monitoring systems should be subjected to objective evaluation, the value of conducting an RCT to demonstrate an impact of monitoring systems on outcomes may be questioned. Our literature review revealed that monitoring systems have not been well evaluated in RCTs, and that, of the few studies available, most have not yielded positive results. The literature search also revealed somewhat puzzling results, including the lack of evidence supporting the use of pulse oximetry, a widely accepted technique, but some evidence supporting the use of gastric tonometry, a technique which has not really stood the test of time. Finally, the PAC has been subjected to more evaluation than any other monitoring technique in acutely ill patients.