Year in review 2011: Critical Care - cardiology

We review key research papers in cardiology and intensive care published during 2011 in Critical Care and quote related studies published in other journals whenever appropriate. Papers are grouped into the following categories: cardiovascular therapies, mechanical therapies, biomarkers, prognostic markers, hemodynamic monitoring, cardiovascular diseases, microcirculation, hypertension in critically ill patients, and miscellaneous.


Introduction
We review key research papers in intensive care cardiology published during 2011 in Critical Care. Related studies published in other journals are also discussed, whenever appropriate.

Cardiovascular therapies
Inotropic agents are used to increase oxygen delivery in the perioperative setting but their impact on mortality is not well defi ned. In a retrospective study including 1,326 cardiac surgery patients, those exposed to inotropes had, as expected, a higher unadjusted mortality rate than those patients not exposed [1]. After adjustment, inotrope exposure was still associated with increased hospital mortality and renal dysfunction. A propensity score-matched analysis yielded similar results. A limitation of this trial is that this type of analysis only takes into account measured variables, but other factors may play a role. Post opera tive inotrope exposure may thus be associated with worse outcomes, but this should be tested in interventional trials.
Should nonadrenergic inotropic agents be preferred? A meta-analysis evaluated the eff ects of levosimendan, versus control, in patients after percutaneous or surgical cardiac revascularization [2]. Th e meta-analysis included 729 patients from 17 studies. Levosimendan increased the cardiac index. Compared with controls, levosimendan treatment was associated with a mortality reduction after coronary revascularization (odds ratio = 0.40, 95% confi dence interval (CI) = 0.21 to 0.76) and a reduction in the length of ICU stay. An important limitation of this analysis is that most of the studies included were small sized, and that there was an important heterogeneity in dosing and time of administration of levosimendan as well as in drugs used in the control arms. Th is potentially benefi cial eff ect of levosimendan should be evaluated in a large-scale randomized trial.
Another meta-analysis evaluated the renal eff ects of carperitide, an atrial natriuretic peptide, and nesiritide, a B-type natriuretic peptide (BNP) [3]. Th e systematic review included 15 studies (11 with carperitide and four with nesiritide), of which nine were included in the metaanalysis. Th ere was no diff erence in mortality rates. Both drugs increased urine output, creatinine clearance and glomerular fi ltra tion rate, and reduced the use of diuretics, renal replace ment therapy and length of ICU and hospital stay. Unfortu nately most trials were small sized and these results should be confi rmed in a larger randomized trial.
Anticoagulation during use of mechanical devices is a continuous challenge. Ranucci and colleagues retro spectively compared a conventional heparin-based anticoagula tion protocol with a bivalirudin-based protocol in 21 patients submitted to extracorporeal membrane oxygenation (ECMO) after cardiac operations [4]. Patients treated with bivalirudin achieved signifi cantly better coagu lation variables. Th romboembolic complications did not diff er between groups but bivalirudin patients required less blood product transfusions. Accordingly, bivalirudin appears to be an attractive alternative to conventional heparin, but this agent should be tested in a randomized trial.

Mechanical therapies
Two papers report single-center experience of using mechanical interventions. One study looked at the potential benefi t of routine coronary angiography after out-of-hospital cardiac arrest (OHCA). Cronier and colleagues evaluated a series of 111 consecutive hemodynamically stable patients resuscitated from OHCA due to ventricular fi brillation and treated with mild therapeutic hypothermia [5]. Emergency coronary angiography was performed in most patients, regardless of the electro cardiogram pattern. Most patients (73%) had coronary heart disease, but this incidence was lower in patients <45 years old than in other groups (41% vs. 81%; P = 0.01). Percutaneous coronary intervention was associated with survival. Th ese results were extended later this year in a larger database of 1,040 OHCA patients in Germany that reported similar benefi ts whatever the initial rhythm [6]. Th e potential benefi ts of a combination of angiography and mild therapeutic hypo thermia were discussed in an accompanying editorial [7]. Th e other paper described the use of an intra-aortic balloon pump, ECMO and continuous-fl ow left ventricular assist devices in six patients with peripartum cardio myopathy [8]. Two patients showed partial recovery and could be weaned off the intra-aortic balloon pump while four patients were implanted with a left ventricular assist device, including the ECMO patient. Th ree left ventri cu lar assist device patients were successfully transplanted. Mechanical support should be considered in patients with peripartum cardiomyopathy not responding to medical therapy. Th e choice of the support and optimal timing of implantation is still under debate.

Biomarkers
Several studies evaluated in various settings the usefulness of biomarkers to understand pathophysiological processes, to guide therapy or to refi ne diagnosis.

Biomarkers as an inside to pathophysiology
Post-cardiac surgery vasoplegia frequently occurs, especially after prolonged cardiopulmonary bypass. A perioperative course of copeptin (a vasopressin precursor) and vasopressin plasma concentrations were studied in 64 consecutive patients submitted to elective cardiac surgery [9]. Patients with vasoplegia had lower vasopressin concentrations 8 hours after surgery compared with controls. Interestingly, these patients had a significantly higher copeptin plasma concentration, hyponatremia and a lower ejection fraction before cardiopulmonary bypass compared with controls. Th ey also experienced more complex surgery. A preoperative copeptin concentration >9.43 pmol/l predicted the occurrence of postoperative vasoplegia. Th ese results suggest that preoperative activa tion of the vasopressin system before surgery may induce vasopressin defi ciency after surgery, through depletion of endogenous hypothalamopituitary stores.
Microvascular permeability in burn injury leads to central hypovolemia and tissue edema. In 38 consecutive patients with severe burn injury, BNP and proteinuria were used to assess evolution of microvascular permeability [10]. BNP increased from admission to day 3 and then reached a plateau. Patients with higher BNP had lower proteinuria and received less fl uids than the other patients. Th is observation suggests that high BNP/low proteinuria can be used as a marker of microvascular permeability. Even though low permeability is associated with a better outcome, it is not clear whether these biomarkers may help to guide resuscitation.

Biomarkers as a guide for therapy
BNP and N-terminal pro-B-type natriuretic peptide (NT-proBNP) are often used to identify patients with increased hydrostatic pressure. To evaluate whether changes in BNP and NT-proBNP may help to better characterize the response to therapy, these were measured at presentation, 24 hours, 48 hours and dis charge in 171 consecutive patients presenting to the emergency department with acute decompensated heart failure [11]. BNP and NT-proBNP levels were higher in nonsurvivors than in survivors at all time points (all P <0.001). Treatment reduced BNP and NT-proBNP levels in survivors by more than 50% (P <0.001), while treatment in nonsurvivors did not lower BNP and NT-proBNP levels. Evolution of these markers can be used to early assess treatment effi cacy.
C-type natriuretic peptide and N-terminal pro-C-type natriuretic peptide (NT-proCNP) may refl ect vascular activation in sepsis. Th e ability of NT-proCNP (measured at ICU admission, day 3 and day 7) to identify sepsis was evaluated in 273 critically ill patients [12]. Patients with sepsis had higher NT-proCNP levels than nonsepsis patients. NT-proCNP was strongly associated with infl am matory parameters and severity of disease. Elevated NT-proCNP levels at admission and day 3 predicted death, while NT-proCNP decline after the initial measure ment was associated with reduced mortality compared with stable or increasing levels. Although interesting, this bio marker was not superior to conventional prognostic indicators.
Cell-free DNA is detected in blood in many diseases, including sepsis and pancreatitis, but also in healthy individuals. Cell-free DNA can originate from necrotic cells or apoptotic processes, but is also released by living cells. Levels of plasma cell-free DNA and their prognostic value were evaluated in 580 patients needing mechanical ventilation in 25 Finnish ICUs [13]. Blood samples were taken at study admission (day 0) and on day 2. Plasma cell-free DNA concentrations were similar at day 0 and day 2. Concentrations at admission were signifi cantly higher in 90-day nonsurvivors than survivors, and remained an independent predictor of 90-day mortality in a multivariate logistic regression analysis. However, its clinical benefi t as a prognostic marker seems to be limited given the relatively low area under the curve (AUC) (0.624, 95% CI = 0.572 to 0.676). Th is AUC is indeed lower than in patients after OHCA (0.796, 95% CI = 0.701 to 0.890) [14], which was discussed in the accompanying editorial [15].

Biomarkers for diagnosis
High-sensitivity troponin has been reported to facilitate the early diagnosis of acute myocardial infarction [16]. Th e diagnostic performances of high-sensitivity troponin and conventional troponin for the diagnosis of acute myocardial infarction were compared in 317 consecutive patients presenting to the emergency department with chest pain suggestive of acute myocardial infarction [17]. Th e pretest probability of acute myocardial infarction was assessed without knowledge of troponin results. In the low to moderate pretest probability group, highsensitivity troponin levels ≥0.014 μg/l identifi ed acute myocardial infarction with a higher sensitivity than cardiac troponin I. As specifi city of high-sensitivity troponin was lower, the AUCs were similar for highsensitivity troponin and cardiac troponin I. Interestingly, the positive predictive value was lower for highsensitivity troponin than for cardiac troponin I (47% vs. 78%, P <0.05) but the negative predictive value was not diff erent (99% vs. 95%) for the entire population, with similar trends in diff erent pretest probability subgroups. Th ese results cast some doubts on the diagnostic accuracy of high-sensitivity troponin compared with cardiac troponin I.

Prognostic factors in cardiovascular diseases
Pulmonary embolism remains associated with signifi cant mortality. Th e prognostic value of markers of right ventricular dysfunction in pulmonary embolism was evaluated in a meta-analysis that retrieved eight echocardiographic marker-based studies (n = 1,249), three computed tomography marker-based studies (n = 503) and seven natriuretic peptide-based (BNP) studies (n = 582) [18]. Th e presence of right ventricular dys function determined by echocardiography and biological markers but not by computed tomography was associated with short-term mortality in patients with pulmonary embolism without hemodynamic compro mise on admission. However, the current prognostic value in clinical practice remains very limited due to insuffi cient pooled positive and negative likelihood ratios.
Acute heart failure is another acute life-threatening disease leading to ICU admission. A large database of 4,153 patients hospitalized with acute heart failure was used to evaluate factors associated with outcome [19]. De novo heart failure was seen in 58.3% of the patients. Mortality varied from 62.7% in cardiogenic shock, to 16.7% in right heart failure, to 7.1% in pulmonary edema, to 6.1% in high-output heart failure, to <2.5% in hypertensive or diastolic heart failure. Invasive ventilation and age >70 years were the most important predictive factors for mortality, with or without cardiogenic shock.

Cardiac output measurements
Analysis of the arterial pressure trace is a common way to non-invasively determine cardiac output (CO). Cali bration is often required. Th e number of injections required at each calibration point was evaluated in 91 hemo dynamically stable patients monitored by a PiCCO 2 device (Pulsion, Munich, Germany) [20]. At least three injections should be performed, which allows precision to drop below 10%: 8% (6 to 12%) for CO, 8% (6 to 14%) for global end-diastolic volume and 8% (7 to 14%) for extravascular lung water. Th ese results are important, both for routine and investigational use.
Th e impact of vasopressor use on accuracy of pulse contour measurements of CO was evaluated in 330 data pairs of pulse contour and transpulmonary thermodilution CO measurements from 73 noncardiac surgery critically ill patients [21]. Pulse contour-derived CO was recorded immediately before calibration. Agreement between the two methods was analyzed according to norepinephrine dosage and a time interval between calibrations of up to 24 hours. Pulse contour CO had a mean bias of 0.16 l/minute with a percentage error of 38%. Bias was not aff ected by the norepinephrine dosage or the time elapsed between calibrations.
Some of the usually calibrated devices also have an uncalibrated version. Th ese devices are mostly dedicated to evaluating changes in CO. Th e noncalibrated pulse power CO (LiDCO; LiDCO, London, UK) was compared with trans pulmonary thermodilution (PiCCO; Pulsion) in 42 patients [22]. Several points were obtained before and after cardiopulmonary bypass. Before bypass, the percen tage error was very high (86%). Changes in CO were not adequately tracked. Th is study is unfortunately limited by the acquisition of multiple (and variable) numbers of observations per patient.
Th ere is a continuous search to develop new devices that non-invasively measure CO. An innovative technique based on measurement of spatial patterns of voltage changes distributed over the thoracic skin has been presented [23]. When applying a weak electric current over the thorax, emptying and fi lling of the ventricles during the cardiac cycle induce cyclic voltage changes on the thoracic skin, generating specifi c twodimensional spatial patterns. Analysis of these patterns may enable measurement of changes in ventricular volume. In a pig model, CO was measured with the new device and an ultrasonic fl ow probe positioned around the ascending aorta. Correlation between the two methods was excellent (r = 0.978). Bias was minimal (0.114 l/minute) and limits of agreement were satisfactory (0.55 l/minute). Th is device therefore appears promising but requires further validation.

New applications
A new application for transpulmonary thermodilution was proposed. Severely burned children often develop circulatory failure and increased permeability. Th is prospective cohort study evaluated the feasibility of transpulmonary thermodilution (PiCCO) for hemo dynamic measurements in 69 severely burned children [24]. CO signifi cantly increased from admission and was highest 3 weeks post burn. Th e intrathoracic blood volume index and the extravascular lung water index begin to increase at days 8 to 9 post burn, and the extravascular lung water index was signifi cantly higher in patients who did not survive burn injury. Th ese data indicate that the hyper dynamic state occurs 1 week post burn injury.

Echocardiography
Several articles used echocardiography to assess hemodynamic function in critically ill patients. Assessment of left ventricular function is an area of intense interest. Several echocardiographic indices for the assessment of left ventricular systolic function were compared in 50 patients with shock [25]. Transthoracic echocardiographic evaluation was obtained daily for 7 days. All measured indices were easy to obtain. Th e eyeball ejection fraction was the most reliable and easiest index to obtain for the evaluation of systolic function.
Transthoracic echocardiography can be used for the diagnosis of left ventricular thrombosis in the postoperative care unit [26]. Five cases were observed in 160 patients. Th rombi were observed in three of 35 patients with ischemic cardiomyopathy and in two of 21 patients with dilated cardiomyopathy. Of note, these thrombi were absent in preoperative echocardiograms. Th ese thrombi were probably facilitated by local low-fl ow condi tions and the postoperative procoagulant state. Clini cians should be aware of this potential postoperative complication.

Central venous oxygen saturation and lactate
Even though low central venous oxygen saturation values are associated with a poor outcome in patients with septic shock, the prognostic value of high central venous oxygen saturation values is not well determined. Th e relation between maximal central venous oxygen saturation levels within the fi rst 72 hours after the onset of shock and ICU survival was evaluated in a singlecenter retrospective study including 152 patients in a 2-year period [27]. Th e maximal level of central venous oxygen saturation was higher in nonsurvivors compared with the survivors (85% (78 to 89%) vs. 79% (72 to 87%), respectively; P = 0.009). As the accompanying editorial commented [28], multiple causes may contribute to an increased central venous oxygen saturation, including microcirculatory alterations [29,30], decreased oxygen handling by the cells due to mitochondrial dysfunction [31], and excessive inotropic/fl uid therapy.
Evolution of lactate levels over the fi rst 24 hours may refl ect adequacy of resuscitation. Th e prognostic value of time-weighted lactate levels was evaluated in 5,041 consecutive critically ill patients from four Australian university hospitals [32]. Both the time-weighted average lactate and the change in lactate over the fi rst 24 hours were independently predictive of hospital mortality. Th e combination of both computations signifi cantly outperformed (P <0.0001) static indices of lactate concentration, such as admission lactate, maximum lactate and minimum lactate. Th e same group recently reported in the same database that time-weighted relative hyperlactatemia, defi ned as time-weighted lactate between 1 and 2 mmol/l, is also associated with a signifi cant increase in risk of death [33]. Altogether these results show that dynamic indices of hyperlactatemia in the fi rst 24 hours following ICU admission are associated with outcome and are superior to static measurements of blood lactate levels.

Fluid responsiveness
Optimization of fl uid therapy is a topic of intense research. Cardiac fi lling volumes and pressures are often used to guide fl uid administration. Th e relative value of cardiac fi lling volume and pressures for predicting fl uid responsiveness was evaluated in 32 patients after cardiovascular surgery [34]. Regardless of the ejection fraction, baseline central venous pressure was lower in responding events. Th e pulmonary artery occlusion pressure was more useful than the global end-diastolic volume index for predicting fl uid responsiveness when the ejection fraction was low, but when the ejection fraction is near normal the volume index is more useful than the occlusion pressure. Could fi lling pressures be determined on chest X-ray imaging? Th e vascular pedicle width is determined on chest X-ray imaging and is a noninvasive measurement of intravascular volume status. Th e vascular pedicle width and central venous pressure or pulmonary artery occlusion pressure were measured in 152 patients with acute lung injury [35]. Th is retrospective cohort is a substudy of the Fluid and Catheter Treatment Trial [36]. Th e vascular pedicle width correlated with the pulmonary artery occlusion pressure and central venous pressure. A vascular pedicle width ≥72 mm identifi ed a pulmonary artery occlusion pressure ≥18 mmHg with an AUC of 0.69 (P = 0.001) but, admittedly, with modest sensitivity and specifi city (61% for both).
Dynamic indices better predict fl uid responsiveness than static indices. A closed-loop fl uid-management algor ithm using a patient simulator including a pulse pressure variation output was tested in three phases [37]. In diff erent hemorrhage scenarios, the algorithm better maintained hemodynamics compared with no management. In the second phase, compared with 20 practicing anesthesiologists for the management of a simulated hemorrhage scenario, the algorithm intervened earlier, gave more fl uids and better preserved CO. Th e algorithm was relatively insensitive to noise and artifacts.
Dynamic indices may be limited by several factors. Respiratory variation in plethysmography, as a consequence of stroke volume variations, can non-invasively predict volume responsiveness [38]. Vasoconstriction altered this signal in 67 patients [39]. Plethysmography variability was related to pulse pressure variation in patients not receiving norepinephrine, while this relationship was lost in patients treated with norepinephrine. In the latter, plethysmography variations were unable to predict the response to fl uids. Th e tidal volume may also aff ect pulse pressure variations [40]. Lakhal and colleagues observed in 65 patients with acute respiratory distress syndrome that variations in pulse pressure fail to predict fl uid responsiveness [41]. Th e importance of the driving pressure was illustrated well [42]: pulse pressure variation adequately predicted fl uid responsiveness in the subgroup of patients experiencing a respiratory variation in pulmonary artery occluded pressure >4 mmHg, refl ecting large changes in pleural pressure. When these indices are ineff ective, one may consider using the passive leg raising test. Interestingly, this test can be applied even in the most severe patients such as patients treated with venovenous ECMO [43]. In 17 patients, an increase in stroke volume >10% during passive leg raising predicted the response to fl uids.
Th e increase in arterial pressure during fl uid administration depends not only on the increase in CO but also on arterial tone. However, predicting whether a given patient would present increased arterial pressure before performing fl uid administration remains challeng ing. Monge García and colleagues present an elegant method to evaluate vascular tone [44]. Th ese authors computed dynamic arterial elastance as the slope of pulse pressure variations divided by stroke volume variations over the same respiratory cycle. In 25 preload responsive patients, dynamic arterial elastance >0.89 predicted an increase in arterial pressure in response to volume loading in preload-dependent patients with a sensitivity of 93.75% (95% CI = 69.8 to 99.8%) and a specifi city of 100% (95% CI = 66.4 to 100%).

Perioperative management of high-risk surgical patients
Perioperative hemodynamic optimization is recommended in high-risk surgical patients [45]. Cecconi and colleagues extend this concept somewhat [46]. Th ey randomized 40 patients submitted to elective total hip arthroplasty under regional anesthesia to receive either conventional hemodynamic therapy or goal-directed hemodynamic therapy (GDT). Patients randomized to the GDT group received a greater volume of intravenous fl uids, more red blood cell transfusions, and more dobutamine during the intraoperative period, although control group patients received greater volumes of blood replacement postoperatively. Th ere was an increased number of complications in the control group compared with GDT patients (100 vs. 80%; P = 0.05), predominantly due to a diff erence in minor complications. Th is trial is nevertheless in line with another trial in the fi eld, using the same monitoring tool [47]. In a meta-analysis including 26 randomized trials, perioperative goaldirected therapy was also found to decrease infections, including surgical site infections (pooled odds ratio = 0.58, 95% CI = 0.46 to 0.74; P <0.0001), pneumonia (pooled odds ratio = 0.71, 95% CI = 0.55 to 0.92; P = 0.009), and urinary tract infections (pooled odds ratio = 0.44, 95% CI = 0.22 to 0.84; P = 0.02) [48].
Fluids are an important part of GDT. Cuthbertson and colleagues conducted a multicenter randomized controlled trial of fl uid loading in 111 high-risk surgical patients undergoing major elective abdominal surgery (the FOCCUS study) [49]. Th e patients underwent preoperative fl uid loading with 25 ml/kg Ringer's solution in the 6 hours before surgery and the control group had no preoperative fl uid loading. Patients in the intervention group spent fewer days in hospital after surgery (mean 12.2 (standard deviation 11.5) days compared with 17.4 (20.0) days) and presented an adjusted mean diff erence of 5.5 days (median 2.2 days; 95% CI = -0.44 to 11.44; P = 0.07). Th ere was a reduction in adverse events in the intervention group (P = 0.048).
However, giving fl uids as a fi xed amount may not be optimal. Th e eff ects of a restrictive versus conventional strategy of crystalloid administration during goaldirected therapy were tested in 88 high-risk surgical patients [50]. Patients were randomized to receive 4 ml/kg/hour (restrictive) or 12 ml/kg/hour (conventional) Ringer's lactate solution as fl uid maintenance during surgery. In both groups, dobutamine and fl uid challenges were administered as necessary to optimize oxygen delivery. Th e conventional treatment group received a signifi cantly greater amount of fl uids than the restrictive group but oxygen delivery was similar in the two groups. Th e restrictive group had a 52% lower rate of major postoperative complications than the conventional group (P = 0.046). Th is further illustrates that fl uids should be given as needed but not using fi xed amounts.
Finally, how frequently is GDT applied? Current hemodynamic management practices in patients undergoing high-risk surgery in Europe and in the United States were evaluated in a survey among members of the American Society of Anesthesiologists and the European Society of Anaesthesiology [51]. CO is monitored by only 34% of American Society of Anesthesiologists and European Society of Anaesthesiology respondents (P = 0.49), while central venous pressure is monitored by 73% of American Society of Anesthesiologists respondents and by 84% of European Society of Anaesthesiology respondents (P <0.01). Th e pulmonary artery catheter is being used more frequently in the United States than in Europe in the setup of high-risk surgery (85.1% vs. 55.3%, P <0.001). Th ese results indicate a considerable gap between clinical practices in Europe and the United States. Of note, technologies required for perioperative hemodynamic optimization were used only by one-third of responders.

Microcirculation
Th is year, several trials also focused on the microcirculation. Th e microcirculatory eff ects of fl uids were investigated using near-infrared spectroscopy during a vascular occlusion test during 42 fl uid challenges in patients undergoing major abdominal surgery [52].
Before fl uids, the tissue oxygen saturation recovery slope was lower in positive fl uid challenges than in negative fl uid challenges (P = 0.02) while baseline tissue oxygen saturation did not diff er. Volume expansion increased the tissue oxygen saturation recovery slope both in positive fl uid challenges (62 ± 49%, P <0.001) and in negative fl uid challenges (26 ± 34%, P = 0.04). Hence, fl uid loading signifi cantly improved the tissue oxygen saturation recovery slope, even when associated with ineff ective changes in systemic hemodynamics.
Th e eff ects of vasopressin agonists on sublingual microcirculation were evaluated in 60 patients with septic shock [53]. Th ese patients were randomized to receive continuous infusions of either terlipressin (1 μg/kg/hour), vasopressin (0.04 U/minute) or placebo (isotonic saline). Terlipressin and vasopressin decreased norepinephrine requirements at the end of the 6-hour study period. Th ere were no diff erences in sublingual microcirculatory variables and systemic hemodynamics among the three study groups during the entire observation period.

Hypertension in critically ill patients
Euro-STAT is an observational study performed in 11 hospitals in seven European countries including 791 consecutive adult patients treated with intravenous antihypertensive therapy in the emergency department, perioperative unit or ICU [54]. Nitroglycerine was the most commonly used antihypertensive treatment (40% of patients), followed by urapidil (21%), clonidine (16%) and furosemide (8%). Hypotension occurred in 10% of patients.
What is the best therapy for acute hypertension? In a multicentric trial in 226 patients, nicardipine-treated patients more often reached target pressure within 30 minutes than labetalol-treated ones (91.7% vs. 82.5%, P = 0.039) [55]. Use of rescue medications did not diff er between nicardipine and labetalol treatment (15.5% vs. 22.4%, P = 0.183). Nicardipine may thus more rapidly control hypertension.

Miscellaneous
Is it useful to drain pleural eff usions in mechanically ventilated patients? A meta-analysis including 19 observational studies and 1,124 patients showed that eff usion drainage improved the PaO 2 :FiO 2 ratio by 18% [35]. Complication rates were low for pneumothorax (3.4%) and hemothorax (1.6%). Ultrasound guidance was not associated with a reduction in the risk of pneumothorax.