Andrew P. Walden, Intensive Care Unit, Royal Berkshire Hospital, London Road, Reading RG1 5AN, Berkshire, UK. Email: firstname.lastname@example.org
Despite 50–60% of intensive care patients demonstrating evidence of pleural effusions, there has been little emphasis placed on the role of effusions in the aetiology of weaning failure. Critical illness and mechanical ventilation lead to multiple perturbations of the normal physiological processes regulating pleural fluid homeostasis, and consequently, failure of normal pleural function occurs. Effusions can lead to deleterious effects on respiratory mechanics and gas exchange, and when extensive, may lead to haemodynamic compromise. The widespread availability of bedside ultrasound has not only facilitated earlier detection of pleural effusions but also safer fluid sampling and drainage. In the majority of patients, pleural drainage leads to improvements in lung function, with data from spontaneously breathing individuals demonstrating a consistent symptomatic improvement, while a meta-analysis in critically ill patients shows an improvement in oxygenation. The effects on respiratory mechanics are less clear, possibly reflecting heterogeneity of underlying pathology. Limited data on clinical outcome from pleural fluid drainage exist; however, it appears to be a safe procedure with a low risk of major complications. The current level of evidence would support a clinical trial to determine whether the systematic detection and drainage of pleural effusions improve clinical outcomes.
Discontinuation of mechanical ventilation is a key moment in the recovery from critical illness, accelerating the process of rehabilitation and reducing the morbidity attributable to ventilator-induced lung injury and nosocomial infection.1–3 The process of weaning may be hampered by a combination of pathologies affecting the respiratory, cardiovascular, abdominal and neuromuscular systems in addition to psychological dysfunction.4 Components of the respiratory system include the airways, lung parenchyma, pleural cavity and chest wall. One poorly studied element of respiratory dysfunction that may delay weaning is the pleural space. Observational studies have demonstrated that 50% to 60% of patients who are mechanically ventilated develop pleural effusions of varying but poorly defined causes.5 Animal and human studies suggest that pleural effusions can have significant deleterious effects on respiratory mechanics and gas exchange.6 Reversal of these abnormalities through thoracocentesis has not been studied in a randomized controlled trial, but evidence exists from both self-ventilating and mechanically ventilated patients of improvements in gas exchange and measures of lung function.5,7–11 This review seeks to explore the aetiology and physiological effects of pleural effusions in this setting and the potential benefits of pleural fluid drainage in the discontinuation of mechanical ventilation.
NORMAL MECHANISMS OF PLEURAL FLUID PRODUCTION AND TURNOVER
The major function of the pleura and the pleural space is to facilitate inflation and deflation of the lungs by reducing friction with the chest wall.12 The pleural space is bordered by the parietal pleura, lining the chest wall, and the visceral pleura, which covers the surface of the lung. The distance between the two surfaces varies between 10 and 24µm and the pleura itself is lined by mesothelial cells projecting microvilli into the pleural space (Fig. 1). Normal pleural fluid content is of the order of 0.1–0.2 mL/kg but represents an enormous potential space with pleural collections of up to 4 L possible (∼50 mL/kg). Normal pleural fluid is of alkaline pH (pH 7.6) with a low protein content (<1.5g/dL).12
The traditional model of pleural fluid production and turnover is based on the work of Starling and Tubby on serous cavities and extended to the pleural space by Neergard.13 In this model of microvascular water and fluid filtration, there is a balance between hydrostatic and colloid osmotic forces governed by the equation:
Where Jv represents the water flux, Kf is the filtration coefficient, PH and π are the hydrostatic and colloid osmotic pressures respectively of the different cavities and σ is the solute reflection coefficient of the membrane. This paradigm proposes that pleural fluid is produced as a filtrate by the vessels around the parietal pleura and then reabsorbed through the visceral pleura (Fig. 2); however, mathematical modelling has shown this to be too simplistic. The differential absorption of fluid and protein due to the semipermeable nature of the visceral pleura would lead to accumulation of protein over time.12
Indeed, experimental evidence suggests that there is little reabsorption of pleural fluid through the thicker visceral pleura, leading to a functional separation of the pulmonary interstitium and pleural space.12 The major processes of pleural fluid homeostasis appear to be microcirculatory filtration in the parietal layer and lymphatic drainage of fluid from the pleural space by parietal lymphatics. Interestingly, in animal models of lung injury, when extravascular lung water exceeds a certain threshold, fluid will filter the other way across the visceral pleura into the pleural cavity.14 An important feature of pleural lymphatics under normal physiological conditions is that they are pulsatile due to a combination of intrinsic smooth muscle rhythmogenic activity and extrinsic tissue pressure oscillations occurring as a result of respiratory movements.15 This creates subatmospheric pressures up to −10 cm H2O with the lymphatics acting like a vacuum capable of increasing pleural fluid drainage up to 20-fold above baseline. In addition, a hydraulic gradient exists within the lung as a result of a higher concentration of parietal lymphatics on the diaphragmatic and mediastinal surfaces of the lung pleura, resulting in flow of fluid within the pleural space.16 Surprisingly, the final destination of the lymphatic drainage is uncertain. It is believed there may be communication between the pleural and peritoneal lymphatics, but little is known of the microscopic lymphatic drainage route.
IMPAIRED PLEURAL FLUID TURNOVER AND PLEURAL FAILURE IN MECHANICALLY VENTILATED PATIENTS
As parietal pleural filtration and parietal lymphatic drainage form the bulk of pleural fluid cycling, in a steady state, these two processes must be balanced. Even in situations of pathological fluid filtration across the visceral pleura, or high extravascular lung water, the pleural lymphatics have the capacity to increase fluid and protein removal by as much as 20-fold the baseline rate.15 Therefore, an imbalance of production and drainage is required to lead to pleural fluid accumulation. Mechanical ventilation leads to multiple perturbations of these processes, and consequently, failure of normal pleural function (Fig. 3). Critical illness is often associated with systemic inflammation leading to capillary leak. This results in a reduction in the reflection coefficient, creating greater filtration of fluid and solute. Increased filtration will be augmented further by decreases in plasma oncotic pressure due to extravasation of oncotically active molecules and administration of crystalloid solutions. While the visceral pleura contributes little to pleural fluid flow in health, this may not be the case with some pathologies. Aberle et al. investigated the radiological evidence of pleural effusions in acute respiratory distress syndrome, finding it present in 38% of patients.17 The same group examined the effects of pleural fluid accumulation in a sheep model of oleic acid-induced lung injury (sheep visceral pleura is similar in thickness and function to humans). These investigators found that pleural fluid formation occurred 5 h after lung injury, led to moderate size pleural effusions, and from analysis of the protein content, came from the pulmonary interstitium across the visceral pleura.14 Indeed the rate of pleural fluid production was increased 25 times compared with baseline.
Other factors will affect removal of pleural fluid. The pleural lymphatics are the main mechanism for removal of fluid from the pleura,12 the natural rhythmogenicity being augmented by negative pressure ventilation. In critically ill patients on positive pressure ventilation, this process is impeded. Prolonged recumbency alters bulk flow across the pleura and consequently impairs the usual lymphatic drainage. Patients may have high intra-abdominal pressures, particularly following abdominal surgery, or the presence of ileus, leading to gut wall oedema and impaired lymphatic drainage. Not all these factors may be present at the same time, and this may account for the size range of effusions observed. However, recognizing that there is a physiological substrate that predisposes to the formation of pleural effusions in critically ill patients is an important concept. Indeed, even in the presence of an established cause for a pleural effusion, the presence of these factors may result in a size larger than might be expected in a spontaneously breathing, ambulatory patient.
INCIDENCE OF PLEURAL EFFUSION IN MECHANICALLY VENTILATED PATIENTS
There are inherent problems with diagnosing pleural effusions in recumbent, mechanically ventilated patients. Clinical examination is difficult to perform and it is not possible to perform erect postero-anterior chest radiographs in this group of patients. Clinical examination and chest X-ray (CXR) changes alone are an insensitive method for detecting their presence. In a prospective study comparing clinical examination with CXR and ultrasound against the gold standard of computerized tomogram in patients with acute respiratory distress syndrome, CXR added nothing to clinical examination in the diagnosis of pleural effusion.18 The sensitivity and diagnostic accuracy of clinical examination were 42% and 61% respectively but only 39% and 47% for CXR. This compares with ultrasound where the values were 92% and 93%. These are important facts to take into consideration when examining any data. Three case series have attempted to determine the incidence of pleural effusions in critically ill medical patients.19,20 Fartoukh et al. used only clinical examination and CXR changes to determine the presence of pleural effusions.20 They examined 1351 medical intensive care unit (ICU) admissions prospectively identifying effusions in only 113 (8.4%). However, in a study using ultrasound in addition to examination and CXR findings of 100 consecutive admissions staying for longer than 24 h on the ICU, 62 had evidence of pleural effusions.19 Recently, the routine use of ultrasound in a medical ICU in the United States demonstrated an incidence of 49% in all admissions.
CAUSES OF PLEURAL EFFUSIONS IN CRITICALLY ILL PATIENTS
Light has defined an approach to the diagnosis of pleural effusion and developed criteria for the differentiation of pleural effusions into transudates and exudates.21 These criteria have been extensively validated in spontaneously breathing patients, although this is not the case for mechanically ventilated patients. Light's criteria represent the best available evidence for critically ill individuals; however, changes in capillary permeability and plasma oncotic pressure may affect their accuracy in mechanically ventilated patients. While this may make the distinction between transudates and exudates difficult, the same principles should apply as for non-ventilated patients to ensure a systematic approach and limit diagnostic inaccuracy. Evidence of pleural space infection should always prompt adequate drainage in addition to appropriate antibiotics. Two studies have directly addressed the aetiology of effusions in critically ill patients using similar criteria.19,20 The diagnostic classification was based to a large degree on clinical grounds and the range of diagnoses was wide. While it is important to attempt a classification of the causes, categories such as atelectasis, hypoalbuminaemia, fluid overload, cardiac failure and postoperative effusions are all diagnoses of exclusion and the diagnostic certainty must be low from the criteria used (Table 1). Hypoalbuminaemia and fluid overload are common findings in critically unwell patients and diagnosis of cardiac failure on the basis of clinical examination and CXR has a very poor yield in mechanically ventilated patients.18 No documentation of left ventricular function on echocardiography was stated.19,20 The diagnosis of atelectasis raises issues of causality—did the pleural effusion lead to collapse of the underlying lung or did the potential space caused by lung collapse lead to pleural fluid accumulation? This question was addressed by Doelken et al. who measured pleural pressure and elastance during thoracocentesis. These investigators found that the pressure remained low in all patients studied. This would favour effusion causing collapse, as the underlying lung must be expanding freely for the pressure to remain low.22 This concept is also supported by the rapid and large volume of fluid that is quickly drained at thoracocentesis.5
Table 1. Categorization of causes of pleural effusion in intensive care patients. Summary of causes and how diagnoses were reached in the two studies specifically identifying the cause of pleural effusions in intensive care patients
The pressure in the pleural space is usually subatmospheric, but as fluid accumulates, it becomes positive. As the volume within the pleural space increases, other structures inside the thorax must accommodate this. The chest wall, lungs, diaphragm and heart are all distensible, and both the volume of pleural fluid present and the compliance of the different components determine the effects on these structures. As the lung is separated from the chest wall, a vertical pressure gradient of 1 cm H2O per cm will develop such that the effects will be more marked in dependant areas.6
Effects on respiratory mechanics and gas exchange
No data exist on how pleural effusions affect respiratory mechanics in mechanically ventilated patients; however, there are animal data and studies in ambulatory, non-ventilated patients with pleural effusion from which some useful information can be derived.
In humans, restrictive ventilatory defects with reductions in the vital capacity, functional residual capacity and total lung capacity have been observed in case–control series,23,24 resulting in hypoxaemia and a reduction in carbon monoxide transfer factor. This may represent the uncoupling of the chest wall from the lung in the presence of effusion, attenuating the effects of normal negative pressure ventilation. This effect may differ in positive pressure ventilation.
In a mechanically ventilated dog model, there was a relationship between the amount of fluid instilled into the pleural space and a reduction in the functional residual capacity; however, this was not in proportion to the volume of fluid instilled.25 Due to higher compliance of the chest wall and diaphragm for a given quantity of fluid instilled, the effects were to reduce the functional residual capacity and total lung capacity by one third and increase chest wall volume by two thirds. A study on anaesthetized, ventilated pigs showed that hypoxaemia occurs with relatively small volumes of saline instilled into the pleural cavity and in a dose-dependent manner.26 This was a consequence of increased intrapulmonary shunting with very little effect on cardiac output.
In an elegant study in mechanically ventilated pigs, Graf et al. used measures of respiratory mechanics in combination with computerized tomogram measurements of lung volume to study the effects of pleural effusions on ventilation and lung recruitment.27 There was a dose-dependent decrease in the lung volume as determined by computerized tomogram, with pleural effusion partially reversed with the application of positive end expiratory pressure. In addition with low levels of positive end expiratory pressure, the proportion of tidal ventilation was high with its associated risks of ventilator-induced lung injury. The clinical implications of this are clear with application of positive end expiratory pressure offsetting to some degree the effects on lung collapse and reducing tidal ventilation.
Effects on haemodynamics
Experimental animal models demonstrate that high-volume pleural effusions can have significant effects on haemodynamics. In pigs, escalating quantities of saline introduced into the pleural space increases intrapleural pressure, leading to lung collapse and a concomitant rise in pulmonary vascular resistance.28 This ultimately leads to right ventricular failure and cardiovascular collapse. While this degree of compromise is probably an uncommon cause of cardiovascular failure in critically unwell patients, in respiratory failure with acute cor pulmonale,29 it may contribute to the overall deleterious effects on pulmonary vascular resistance, blood pressure and cardiac output. In dogs, direct compressive effects have been shown with experimental effusions when intrapleural pressure exceeds intrapericardial pressure, right ventricular diastolic collapse can occur, causing tamponade.30 Similar effects of right ventricular diastolic collapse have been observed as incidental findings in patient case series.31–33 Haemodynamic collapse rapidly improves following drainage in cardiac surgical patients,34 hepatic hydrothorax and malignant effusion.35 It may be that many patient parameters alter the haemodynamic response to a pleural effusion. The volume status, lung and chest wall compliance and pre-existing cardiac function could all influence the outcome in addition to the effusion size and rate of accumulation.
PHYSIOLOGICAL IMPROVEMENTS FOLLOWING THORACOCENTESIS
Spontaneously breathing patients
Several studies have examined the effects of thoracocentesis in spontaneously breathing, ambulatory individuals where they have acted as their own controls.7–10 These studies have selected patients with evidence of pure pleural disease to remove any confounding effects from underlying neuromuscular or parenchymal lung pathology. While this does not reflect the heterogeneity of lung disease in critically ill patients, it provides useful insight into the improvements in mechanics and gas exchange that might be expected. Brown et al. showed small but significant improvements in functional residual capacity, total lung capacity and the partial pressure of oxygen7 following thoracocentesis, in contrast to Karetzky et al.8 who found no significant changes in partial pressure of oxygen or partial pressure of carbon dioxide either immediately or 4 h after drainage. In the latter study, the observed alterations showed a wide spread, suggesting that there may be responders and non-responders. In a separate study of 33 patients, improvements in partial pressure of oxygen and a reduction in intrapulmonary shunting were observed for 24 h following drainage, but there was no effect on dead space ventilation.9 Inversion of the hemidiaphragm is a marker of effusion size and might be expected to have more profound effects on respiratory mechanics. When examining this cohort of patients, large volumes of fluid were drained (1610 ± 510 mL) with marked improvements in forced expiratory volume in 1 s, forced vital capacity and partial pressure of oxygen.10 The conclusion is that while not all individuals may improve following pleural drainage, a majority of patients appear to show better ventilation perfusion matching as collapsed lung under a pleural effusion reinflates. Another clear finding from these studies is the invariable symptomatic improvement following thoracocentesis of even small volumes of fluid. In the context of unblinded, uncontrolled studies, it is difficult to ascertain how much of this benefit is attributable to fluid drainage or a ‘placebo’ effect. However, reducing the excursion of the chest wall may lead to less stimulation of stretch receptors and also allow the respiratory muscles to work on a more efficient portion of the length–tension curve, reducing work of breathing.36 The improved coupling of the chest wall and diaphragm to the lung following drainage may also lead to improvements in ventilation and the work of breathing.
Mechanically ventilated patients
In spite of the high incidence of pleural effusions in ICU and the stated adverse effects on ventilation perfusion and respiratory mechanics, there is limited evidence on the role of thoracocentesis in mechanically ventilated patients.
A recent systematic review identified five studies describing the effect of thoracocentesis on oxygenation.11 Pooled data showed a mean improvement in the partial pressure of oxygen to inspired oxygen concentration ratio of 18% (95% confidence interval: 5–33%), corresponding to an increase of 4.1 kPa (95% confidence interval: 0.8–7.3 kPa) among a total of 118 patients. More recently, Walden et al.5 performed a total of 15 thoracocentesis in 10 mechanically ventilated patients with pleural effusions estimated to be greater than 800 mL in volume using ultrasound. Following thoracocentesis, there was a 40% increase in the partial pressure of oxygen from 10.9 ± 1.4 kPa to 15.3. ± 4.1 kPa (P < 0.05) with a 34% increase in the partial pressure of oxygen to inspired oxygen concentration ratio from 22.5 ± 7.5 kPa to 31.7 ± 9.7 kPa (P < 0.05). The effects on partial pressure of oxygen to inspired oxygen concentration ratios were maintained for a period of 48 h (Table 2).
Table 2. Summary of studies examining the effects of thoracocentesis on oxygenation. P-value not reported in original paper calculated by Goligher et al.11
PEEP (cm H2O)
Volume drained (mL, mean ± SD)
A-a gradient; alveolar arterial gradient; FiO2, inspired oxygen concentration; PaO2, partial pressure of oxygen; PEEP, positive end expiratory pressure; SD, standard deviation.
Four studies have examined the effects of thoracocentesis on respiratory mechanics with variable consequences on peak inspiratory pressures22,37,38 but with consistent improvements in dynamic compliance5,38,39 (Table 3). One study also showed improvements in dead space ventilation 24 h after thoracocentesis, suggesting that improvements in ventilation perfusion ratio matching continue and are not confined to improved shunting.5
Table 3. Summary of studies examining effects of thoracocentesis on respiratory mechanics. P-value not reported in original paper calculated by Goligher et al.11
VD, dead space volume; VT, tidal volume; VD/VT, dead space fraction.
In summary, thoracocentesis appears to improve respiratory mechanics but not in proportion to the amount of fluid aspirated. It is an effective procedure in terms of symptom relief, and in larger series, appears to result in improved oxygenation. In ventilated patients, there appears to be an improvement in oxygenation that may be sustained over a period of time as a consequence of altered ventilation perfusion relationships.
EFFECT OF THORACOCENTESIS ON CLINICAL OUTCOMES IN MECHANICALLY VENTILATED PATIENTS
The effects of thoracocentesis on clinical outcome in mechanically ventilated patients have not been well studied. In a systematic review examining the utility and safety of thoracocentesis in mechanically ventilated patients, Goligher et al.11 found 19 observational and no controlled trials.20,22,37–53 None of the studies reported duration of mechanical ventilation or mortality. One study of 44 patients with pleural effusions compared length of stay in ICU between patients with greater or less than 500 mL of pleural fluid drained and found no difference.50 Goligher et al. concluded that given the absence of controlled trials, the effect of thoracocentesis on duration of mechanical ventilation and length of stay in ICU and hospital could not be determined.11
SAFETY OF ULTRASOUND-GUIDED THORACOCENTESIS
The safety of thoracocentesis in mechanically ventilated patients has been extensively studied. Tension pneumothorax and haemothorax are the most serious complications; however, under ultrasound guidance, the reported complication rate is reassuringly low. Three studies have been performed by ultrasound-trained intensivists specifically examining complications. The largest included 205 patients and 255 procedures over a 3-year period;52 the rate of pneumothoraces was 4.8%. The requirement for tube placement was only 0.78%. Mayo et al.39 quoted a pneumothorax rate of 3/232 procedures (1.3%) with no haemothoraces and Lichtenstein et al. had a zero complication over a series of 45 procedures.45
Several other studies have recorded similarly low complication rates. In a series quantifying effusion size, there was also a zero complication rate in 55 patients albeit this study was performed by a level III ultrasonographer and thoracocentesis failed in six patients.51 A similar complication rate was found in a study of parapneumonic effusions in mechanically ventilated patients, with 2/94 (2%) of patients developing a haemothorax.48 In a surgical ICU with a series of 135 patients having 338 procedures, the complication rate was again similar, with four patients (1.2%) developing pneumothoraces and four developing haemothoraces.54
In a systematic review of studies of thoracocentesis in ventilated patients, Goligher et al.11 found the pooled risk of pneumothorax was 3.4% (95% confidence interval: 1.7–6.5)—20 events in 14 studies including 965 patients. The pooled risk of haemothorax was 1.6% (95% confidence interval: 0.8–3.3%)—four events in 10 studies with 721 patients. It should be noted that some of the older studies did not use ultrasound guidance, while others used ultrasound to mark puncture sites or real-time ultrasound guidance. Goligher et al. found that the use of ultrasound was not associated with a reduction in risk of pneumothorax (odds ratio 0.32; 95% confidence interval: 0.08–1.19).
This data should reassure intensive care physicians that with individuals appropriately trained in the use of ultrasound, the risk of procedural harm is low.
Mechanical ventilation in critically ill individuals provides the physiological substrate for the development of pleural failure and accumulation of pleural fluid. In addition, many of the conditions that lead to ICU admission may be associated with pleural effusions. The incidence in unselected patients is of the order of 50–60%, and drainage leads to variable improvements in respiratory mechanics, reflecting the heterogeneity of underlying disease in this patient cohort. Significant effects on oxygenation are apparent when results of studies are pooled. The lack of outcome data is striking and there is sufficient evidence to support a trial examining the effects of drainage on clinically relevant variables such as duration of mechanical ventilation.
Thanks to Rebecca Clemens for her helpful insights.