Pulmonary Consequences of Acute Kidney Injury

As the link between AKI and pulmonary complications has emerged over the past several decades, so has the notion that AKI is a complex multisystem disorder resulting in a series of both traditional complications and nontraditional complications (Table 1). This classification is useful when considering the potential mediators of and remedies for systemic complications of AKI, including respiratory complications of AKI (Table 2). The traditional complications of AKI include volume overload; electrolyte abnormalities such as hyperkalemia, metabolic acidosis, hyperphosphatemia, and hypocalcemia; and the clinical syndrome of uremia. These traditional complications of AKI all generally are correctable by RRT. However, AKI increasingly is recognized to result in multiple systemic derangements and a variety of nontraditional complications have been reported in human beings, animal models, or both, including increased susceptibility to sepsis,^,  immune dysfunction, and injury to or dysfunction of a variety of organs including the heart,^,  liver,^{,  ,}  intestine,^,  and brain. It has been proposed that these multisystem nontraditional complications of AKI are responsible for its high mortality, which persists despite RRT.^, 

Although our ability to study the effects of AKI in human beings generally is limited to observational data, the specific effects of AKI, rather than a lack of renal function per se, can be gleaned from comparisons between acutely ill patients with AKI and those with end-stage renal disease (ESRD). ESRD patients are subjected to the same traditional consequences of renal failure as AKI patients including electrolyte abnormalities, fluid overload, and uremia. In contrast, only AKI patients are subject to the nontraditional systemic effects of AKI. As such, differences in outcomes between AKI and ESRD patients may be mediated by these nontraditional effects of kidney injury.

Multiple studies have established that AKI in hospitalized patients carries a higher risk of mortality than ESRD.^{,  ,  ,}  For example, one study compared critically ill patients without renal failure, with ESRD, with AKI not requiring RRT, and with AKI requiring RRT, showing mortality rates of 5%, 11%, 23%, and 58%, respectively, despite similar illness severity (Acute Physiology And Chronic Health Evaluation [APACHE] III) scores in the ESRD and AKI patients. Another study showed that the increased mortality of AKI relative to ESRD occurred despite a significantly older age and rate of comorbid diabetes and hypertension in the ESRD cohort. Furthermore, three of these studies found that mechanical ventilation is needed in a higher percentage of AKI patients relative to ESRD patients, and in all three studies mechanical ventilation was found to be a risk factor for mortality.^{,  ,}  In one study, mechanical ventilation was the sole factor on multivariate analysis driving the increased mortality in AKI patients relative to ESRD patients. AKI was considered to contribute to respiratory failure via a combination of fluid overload, as shown by increased weight gain between ICU admission and continuous renal replacement therapy (CRRT) initiation, and increased inflammation, as evidenced by depressed serum albumin. Significantly lower levels of serum albumin in AKI patients were observed, despite the fact that hypoalbuminemia is common and is associated with increased mortality and increased inflammation in ESRD patients. Thus, the poor outcomes of AKI versus ESRD patients with respiratory failure illustrate how nontraditional complications of AKI can combine with the traditional complications of AKI to produce harm.

When considering pulmonary complications of AKI, the most important traditional complication of AKI is volume overload. Volume overload from any cause, including severe renal failure with preserved cardiac function, has the potential to cause hydrostatic (often referred to as cardiogenic) pulmonary edema. Volume overload resulting from AKI was linked long ago to respiratory failure from pulmonary edema.^,  Studies in the 1950s showed that some patients with pulmonary edema and renal failure improve with fluid removal alone.^,  However, it has long been suspected that volume overload does not explain all cases of pulmonary edema in patients with AKI. For example, two historical case series collectively reported nine patients with pulmonary edema in the setting of severe renal dysfunction but in the absence of increased pulmonary capillary wedge pressure, suggesting that increased pulmonary capillary permeability was responsible for the edema.^,  Human autopsy studies, including a large study comparing more than 100 patients with AKI with more than 400 non-AKI controls, have shown that pulmonary edema in patients with AKI is associated with leukocyte infiltration,^{,  ,}  proteinaceous edema fluid,^{,  ,} hyaline membrane formation,^,  and diffuse alveolar damage, leading some investigators in that era (the 1950s–1980s) to consider “uremic pneumonitis”^,  a cause of ARDS.

Establishing the precise mechanisms of pulmonary edema in human beings with AKI is challenging because of the complexities of assessing volume status, the frequent presence of multiple insults, and uncertainty regarding the onset of AKI. In this regard, animal models of AKI have been particularly informative to characterize the nature of lung injury after AKI and elucidate potential mechanisms. Notably, over the past 2 to 3 decades a large body of data (discussed later) has accumulated showing that AKI can induce inflammatory lung injury and nonhydrostatic pulmonary edema. Similar to the historical human case series and autopsy data, the majority of animal models of AKI-induced acute lung injury (ALI) are characterized by higher wet/dry lung weight ratios,^{,  ,}  along with inflammatory changes such as neutrophilic infiltration,^{,  ,  ,  ,}  increased circulating levels of inflammatory mediators,^{,  ,} increased bronchoalveolar lavage (BAL) fluid protein content,^{,  ,,}  and increased pulmonary vascular permeability as measured by Evans blue dye accumulation in the lung.^{,  ,  ,}

As a result, a modern approach to respiratory dysfunction in the setting of AKI has been proposed in which pulmonary edema is considered to exist on a spectrum between pure hydrostatic pulmonary edema, which can be corrected through volume removal by diuresis or ultrafiltration, and nonhydrostatic edema, for which volume removal is ineffective (Figs. 1 and 2). In any given patient, a combination of hydrostatic and nonhydrostatic pulmonary edema may co-exist; in such a patient, volume removal will improve oxygenation and lung function to a degree, beyond which further volume removal may be detrimental by potentially leading to decreased renal perfusion, hypotension, and prolongation of AKI.

Although conceptually straightforward, the impact of volume resuscitation or volume removal in a given patient with AKI and respiratory failure remains a clinically problematic issue, far more complicated than the truism that dry lungs are happy lungs. There is a paucity of data informing clinicians on when to give fluid or when to stop giving fluid to patients with AKI.^,  Although a full discussion of fluid management in the setting of AKI and respiratory failure is beyond the scope of this review, a few salient points can be made.

First, appropriate volume resuscitation in the setting of AKI may be beneficial to both the kidneys and the lungs. For example, a study comparing differing amounts of fluid administration in the first 7 days after ischemic AKI in mice showed that mice that received adequate fluid resuscitation recovered from ischemic AKI sooner and this recovery was accompanied by more rapid resolution of lung inflammation; in contrast, under-resuscitated mice with ischemic AKI had a prolongation of both AKI and inflammatory lung injury. However, despite these animal data and general consensus that, in many patients, adequate volume resuscitation is essential for the prevention of or for shortening the course of AKI, data supporting this notion are limited.^, 

In contrast, there is a large and increasing number of studies showing that volume overload in AKI is associated with poor outcomes, including an increased risk of death; although the observational nature of these studies subject the data to the risk of residual confounding, in all these studies fluid overload remained an independent predictor of mortality after adjustment for other covariates including severity of illness.^{,  ,,  ,  ,}  More importantly, prevention of volume overload has been associated with a decreased duration of mechanical ventilation in settings ranging from adults with ARDS treated with diuretics to infants started pre-emptively on peritoneal dialysis (PD) after congenital heart surgery.^,  The former study, the Fluid and Catheter Treatment Trial (FACTT), showed, via randomization of 1,000 patients to one of two different fluid strategies, that maintaining an approximately net even fluid balance over 1 week (compared with an approximately net positive balance of 1 L/d in the control group) led to improved oxygenation and decreased severity of lung injury and shortened the duration of mechanical ventilation and ICU stay by more than 2 days each. There were no differences in overall mortality or in the rates of shock, AKI, or other nonpulmonary organ failure, but there was a nonsignificant trend toward decreased need for RRT in the conservative management group compared with the control.

Less intuitively, similar benefits of fluid restriction recently were found in patients with sepsis rather than ARDS. One trial of 150 patients with septic shock showed a benefit from implementation, after an initial resuscitation period, of a restrictive fluid strategy, with a trend toward decreased mortality and a statistically significant decrease in AKI rates. In addition, a meta-analysis of 11 trials of adults and children with ARDS, sepsis, and/or systemic inflammatory response syndrome (including FACTT) had similar results, with a conservative or “de-resuscitation” strategy of fluid management resulting in reduced ICU length of stay and more ventilator-free days without a significant effect on mortality.

Four recent, large, multicenter trials of protocolized resuscitation of patients, including three^{,  ,}  of early septic shock treated with 6-hour protocols similar to the original early goal-directed therapy (GDT) of Rivers et al, and one using a cardiac output–guided hemodynamic algorithm after major surgery, all failed to show improvement in either renal outcomes or mortality relative to usual care, emphasizing the need to carefully individualize fluid therapy to every patient rather than applying a blanket approach to all patients with a given diagnosis. However, of the trials that have shown benefit to protocolized fluid administration in the past 20 years, a few themes can be gleaned. First, the older trials showing improvement in renal or overall outcomes of GDT in adults or children with septic shock generally have resulted in increased fluids given early (ie, the first 6 hours), with differences in amounts of fluid administered largely equal in the intervention and usual care groups by 72 hours.^,  Similarly, a review of 24 trials of postoperative GDT showed a significant reduction in rates of postoperative AKI, but the overall difference between the groups in fluid administered was ultimately quite modest, averaging less than an additional 600 mL. Furthermore, the overall benefit of postoperative GDT was driven primarily by the trials in which fluid balance was near equal and/or inotropic support was used, rather than those trials that used significantly more fluid administration. Finally, the interventions in nearly all the studies were limited to the first 24 postoperative hours or less. Coupled with the more limited sepsis data, these findings suggest that fluid resuscitation to prevent AKI is most effective if given early and in a targeted manner to prevent excessive fluid administration and should be followed, outside the initial 24 hours of critical illness, by the maintenance of a net even fluid balance.^, 

With these data in mind, it is clear that accurate assessment of volume status in patients with AKI and respiratory failure, although often elusive, remains of pre-eminent importance because nephrologists frequently are faced with the dilemma of whether to administer volume or attempt, with diuresis or ultrafiltration, to remove volume.

Unfortunately, traditional measures of volume status or fluid responsiveness in patients with hypotension and/or renal failure have been shown to have only marginal clinical utility. For example, central venous pressure (CVP), once the mainstay for guiding resuscitation in sepsis, has been shown to be a very poor determinant of fluid responsiveness, with one systematic review reporting an area under the receiver operating characteristic curve for CVP for predicting improvement in cardiac index with fluids as only 0.56 (ie, only slightly better than a coin flip).

In fact, consistent with the poor overall outcomes associated with volume overload in AKI,^{,  ,,  ,  ,}  data continue to emerge that higher CVP in AKI is associated with worse renal and overall outcomes.^,  Specifically, multiple recent reports have correlated higher CVP with worsening AKI or higher AKI-related mortality in not only heart failure or ARDS, as might be expected, but also in sepsis^,  and in ICU patients as a whole. This effect has been attributed to the harmful effects of venous congestion and reduced renal blood flow and glomerular filtration rate resulting from increased renal subcapsular and interstitial pressure.^,  In sepsis, higher CVP, even within the target range of range of 8 to 12 mm Hg originally proposed by Rivers et al, is associated independently with an increased risk of AKI.^,  Similarly, in a post hoc analysis of more than 300 patients in the FACTT trial who developed AKI in the context of ARDS, CVP was statistically higher (at approximately 13 versus 11 mm Hg) in those who died relative to survivors, and higher mean CVP was associated independently in a multivariate model with a higher risk of death. Likewise, in a study of more than 4,500 medical, cardiac, and surgical ICU patients, progressively higher CVP at admission was associated with a progressively higher risk of AKI. Together, these data suggest that increased CVP is associated with worse outcomes, and some have proposed that the primary utility of CVP is to identify patients with increased CVP at risk of harm from volume overload in whom further fluid administration may be harmful or fluid removal may be beneficial. However, this approach has not been prospectively tested in clinical trials.

Even more basic assessments of fluid balance such as daily weight and charted fluid balance have significant limitations, with recent studies showing poor day-to-day correlation between each other in ICU patients in general or in patients specifically on CRRT, even after correcting for insensible fluid losses. Considering that insensible losses are not accounted for routinely, reliance on charted fluid balance has the potential to induce significant hypovolemia even if RRT is used to maintain seemingly even fluid balance over a prolonged period.

A variety of dynamic measures of fluid responsiveness, such as pulse pressure variation, passive leg raise, or ultrasonographic inferior vena cava collapsibility, have been developed and reported to be superior to static measures of fluid status in caring for ICU patients,^,  although whether they are useful in guiding fluid therapy specifically in AKI is not as clear.

Furthermore, these dynamic measures of fluid responsiveness have been primarily validated for deciding whether to give fluid, not whether to remove fluid, with only a few small studies specifically addressing fluid removal in AKI.^{,  ,}  The first such study involved 39 hemodynamically stable ICU patients with renal failure, including 85% with AKI, in which the passive leg raise was shown to accurately predict subsequent intradialytic hypotension with attempted ultrafiltration with intermittent RRT. The second was a pilot study of 24 patients with decompensated CHF and diuretic resistance, in which inferior vena cava collapsibility similarly was shown to predict hypotension during attempted slow continuous ultrafiltration. More recently, a feasibility trial was performed in 32 ICU patients on the use of transpulmonary thermodilution to estimate cardiac index and related measures and found it moderately useful (with areas under the receiver operating characteristic curves of 0.65-0.75) to predict the ability to achieve ultrafiltration goal during sustained low-efficiency dialysis without a significant increase in vasopressor requirement. None of these studies reported outcomes beyond short-term hypotension or tolerance of fluid removal. A variety of other technologies have been proposed to guide volume removal during dialysis of patients with AKI, but either have been found to be unhelpful, as with continuous blood volume monitoring, are promising but as-of-yet unproven, as with the use of biomarkers to assess for recurrent or additional renal injury from excess volume removal, or, as in the case of bioimpedance or lung ultrasonography, have not yet been studied in AKI apart from prognostication or as adjunctive measures of impaired oxygenation.^,  A study of the use of bioimpedance with lung ultrasound to guide fluid removal with CRRT in AKI currently is underway (ClinicalTrials.gov identifier: NCT02384525). As it currently stands, no one method—dynamic or static, traditional or novel—for diagnosing fluid overload or any one strategy for determining when to start or stop removing fluid, whether by diuretics or ultrafiltration, has been proven superior to any other. Thus, further study is needed to help determine which measures of fluid status are most useful to guide fluid removal and to correlate them with not only short-term risk of hypotension or tolerance of fluid removal, but more meaningful long-term outcomes such as renal recovery or overall survival.

The nontraditional effects of AKI on the lung appear to involve a complex inflammatory cascade, of which four components have been best studied, including increases in levels of circulating cytokines, most prominently interleukin 6 (IL6) and IL8, pulmonary endothelial cell apoptosis, renal necroinflammation, and T-lymphocyte recruitment and activity.^,  These inflammatory processes have been shown most clearly in animal models to be important in bringing about the neutrophil recruitment, endothelial cell injury, and increased capillary permeability characteristic of inflammatory pulmonary edema (Fig. 3). Each is discussed separately in detail later.

AKI in human beings is associated with increased serum levels of IL6,^{,,  ,}  IL8,^{,  ,,}  and tumor necrosis factor (TNF). Increased serum levels of IL6 and IL8 after cardiac surgery in children can be detected as soon as 2 hours after surgery, well before AKI is appreciated by an increase in serum creatinine concentration; furthermore, these early increases in serum IL6 and IL8 are associated with both subsequent AKI development and a prolonged need for mechanical ventilation. IL6 likewise has been shown to predict AKI in patients with severe sepsis. In adult ICU patients, AKI, in comparison with ESRD or non–renal failure controls, is associated with increased levels of IL6 and IL8, which in turn predict a higher risk of mortality.

In animal models, IL6 has been shown to be particularly important among the factors involved in lung injury. In mice subjected to ischemic AKI or bilateral nephrectomy, IL6 has been found to mediate lung injury characterized by neutrophil infiltration, increased neutrophil chemokine production, and capillary leak.^{,  ,} These effects have been attributed specifically to circulating IL6 rather than pulmonary IL6. The primary source of the circulating IL6 in models of AKI-induced ALI had been presumed to be the injured kidney, and indeed renal IL6 messenger RNA and protein expression have been shown to be induced by both ischemic and nephrotoxic AKI.^{,  ,} However, both decreased renal clearance and increased extrarenal production by the spleen and liver and by macrophages have been shown also to contribute to increased IL6 levels after bilateral nephrectomy. Renal clearance of IL6, similar to other small proteins, appears to be mediated by metabolism by proximal tubular cells and is inhibited by tubular injury but not prerenal azotemia. A study using functional genomic analyses of the inflammatory transcriptome, which included more than 100 inflammatory genes, identified the IL6 signaling pathway as a mediator of ALI after ischemic AKI.

IL8 is a cytokine and neutrophil chemokine that also has been shown in animal studies to have a key role in the development of AKI-induced lung injury. Both lung and serum IL8 levels are increased in AKI, and, in models of AKI-induced lung injury, a reduction in IL8 results in decreased lung neutrophil infiltration and activity and decreased lung permeability.^{,  ,,  ,} The most definitive demonstration of the role of IL8 in AKI-induced lung injury comes from an additional experiment performed using the previously referenced mouse model that showed the role of circulating IL6. The investigators were able to show that IL6 acts via up-regulation of the chemokine C-X-C motif ligand 1 (CXCL1), the mouse functional analog of human IL8, on lung endothelial cells to promote neutrophil accumulation. In mice that were treated with either anti-CXCL1 antibodies or were deficient in CXCL1′s receptor (C-X-C motif chemokine receptor 2), AKI or bilateral nephrectomy resulted in significantly less neutrophil infiltration than in the respective untreated or wild-type controls.

A third cytokine shown to mediate lung injury after bilateral nephrectomy or renal ischemia-reperfusion is the high-mobility group B 1 protein, which was shown to produce its effects both via and independently of its receptor, Toll-like receptor 4.

Finally, in addition to IL6 and IL8, other inflammatory mediators increased in the lung or alveolar fluid after ischemic AKI include adhesion molecules such as intercellular adhesion molecule 1,^{,  ,} other chemokines such as macrophage inflammatory protein 2^,  and cytokine-induced neutrophil chemoattractant 2, chemokine receptors such as C-X-C motif chemokine receptor 2, and the signaling complex nuclear factor-κB.^, 

As an extension of these human and animal data showing the role of IL6 and IL8 in mediating AKI-induced lung injury, four studies have been published within the past few years investigating the effects of RRT on cytokine levels and related outcomes.^{,  ,,}  The most comprehensive of these was an animal study that evaluated the effect of RRT on both inflammatory mediators and on lung inflammation after ischemic AKI or bilateral nephrectomy. By using a novel model of PD in mice with AKI, the investigators were able to show that high-dose PD resulted in decreased serum IL6 levels, lung CXCL1 levels, lung neutrophil count and activity, and lung interstitial macrophage number without a change in the wet/dry lung weight ratio. By infusing and tracking recombinant IL6, this experiment directly showed that IL6 is cleared effectively by PD, consistent with prior human data, and that peritoneal IL6 is in equilibrium with serum IL6.

Two recent human studies have shown that RRT is associated with decreased plasma IL6 levels and with improved clinical outcomes, possibly owing to IL6 removal. In one of the previously referenced case-matched cohort studies of prophylactic PD use in infants at high risk of fluid overload after cardiopulmonary bypass for congenital heart disease, the use of PD was shown to be associated with lower serum levels of both IL6 and IL8 within 24 hours, and with more negative fluid balance, less inotrope use, and earlier sternal closure, with a trend toward a shorter duration of mechanical ventilation. Although PD use was associated with a statistically significant decrease of the serum IL6 and IL8 levels, implying clearance of the cytokines, cytokine clearance (ie, cytokine levels in the PD effluent) was not measured directly. A later similar cohort study of prophylactic PD after infant cardiac surgery did not measure cytokine levels, but did report a statistically significant shortening of ventilation time.

The second trial to report on the effect of RRT on cytokine levels in human beings was the highly publicized Early versus Late Initiation of Renal Replacement Therapy in Critically Ill Patients with Acute Kidney Injury (ELAIN) trial, one of two recent randomized controlled trials comparing the timing of RRT initiation in ICU patients with AKI. The trial, in a secondary exploratory analysis, measured levels of various proinflammatory cytokines including IL6 and IL8. The modality used at RRT initiation in both sets of patients was continuous venovenous hemodiafiltration, with a 1:1 ratio of dialysate and replacement fluid. Hemofiltration has been shown in a different setting to effectively clear IL6 and IL8 in human beings. In the ELAIN trial, early RRT resulted in a significant decrease in both the primary outcome of 90-day mortality, with an absolute risk reduction of more than 15%, and in the median duration of mechanical ventilation, by more than 48 hours, despite no significant difference in fluid balance between the groups. With regard to cytokines, early RRT was associated with a reduction in serum IL6 and IL8 levels by 24 hours after randomization, and, in turn, lower day 1 IL6 and IL8 levels were found to be associated with a lower risk of mortality. Again, similar to the human PD study referenced earlier, although cytokine clearance was implied by the decreased serum levels in the early RRT group, cytokine clearance (ie, effluent levels of cytokines) was not measured directly.

Although promising, the overall benefit from early RRT initiation in the ELAIN trial, whether mediated by cytokine clearance or not, must be interpreted with caution given the discrepant results of two other studies. First, the contemporaneous Artificial Kidney Initiation in Kidney Injury (AKIKI) trial, in contrast to the single-center ELAIN trial, was a larger multicenter trial of RRT timing in ICU patients with AKI, which did not show any benefit to earlier initiation of RRT (although, of note, the definitions of early versus late initiation differed significantly in the two trials). The AKIKI trial did not measure cytokine levels. Second, a third report of the effect of RRT in human beings on cytokine levels, a substudy of the Veterans Affairs/National Institutes of Health ATN trial on dose intensity of RRT for AKI in the ICU, had differing results than the earlier-described studies. Specifically, serum levels of a variety of cytokines and apoptosis biomarkers, including IL6 and TNF receptor 1 (TNFR1), were measured on day 1 and day 8 of AKI in more than 800 patients. Higher-dose RRT was associated with no overall change in the serum levels of these biomarkers in those surviving to day 8, with a reduction noted in those with high baseline levels on day 1, but an increase noted in those with low baseline levels. Higher levels of day-8 biomarkers were associated with an increased risk of death and RRT dependence, but the dose of RRT did not modulate the rates of death or renal recovery even when adjusting for baseline biomarker levels.

Of note, in contrast to the ELAIN trial, both the AKIKI and ATN trials used a mix of intermittent hemodialysis and CRRT, in roughly equal proportions.^,  In AKIKI, the CRRT patients underwent either hemofiltration or hemodialysis (with the proportion not reported); in the ATN trial most of the patients on CRRT were prescribed hemodiafiltration (with, as in ELAIN, a 1:1 ratio of dialysate and replacement fluid), although a minority (<10%) of the CRRT treatments were delivered as sustained low-efficiency dialysis.^,  Given that many cytokines are middle-weight molecules greater than 20 kDa (eg, IL6 is approximately 26 kDa), the overall clearance of cytokines via hemofiltration is likely greater than with hemodialysis because the molecular weight cut-off of modern high-flux dialyzers is typically approximately 10 kDa.^,  As such, it is possible that the different outcomes between the AKIKI and the ATN trials and the ELAIN trial could relate to the higher overall proportion of the RRT being provided as hemodialysis rather than hemofiltration, but the limited data available directly comparing the CRRT modalities suggest no difference between hemofiltration and hemodialysis on clinical outcomes in AKI.