Prognostic factors among critically ill patients with community-acquired acute bacterial meningitis and acute kidney injury

Objective To investigate prognostic factors among critically ill patients with community-acquired bacterial meningitis and acute kidney injury. Methods A retrospective study including patients admitted to a tertiary infectious disease hospital in Fortaleza, Brazil diagnosed with community-acquired bacterial meningitis complicated with acute kidney injury. Factors associated with death, mechanical ventilation and use of vasopressors were investigated. Results Forty-one patients were included, with a mean age of 41.6 ± 15.5 years; 56% were males. Mean time between intensive care unit admission and acute kidney injury diagnosis was 5.8 ± 10.6 days. Overall mortality was 53.7%. According to KDIGO criteria, 10 patients were classified as stage 1 (24.4%), 18 as stage 2 (43.9%) and 13 as stage 3 (31.7%). KDIGO 3 significantly increased mortality (OR = 6.67; 95%CI = 1.23 - 36.23; p = 0.028). Thrombocytopenia was not associated with higher mortality, but it was a risk factor for KDIGO 3 (OR = 5.67; 95%CI = 1.25 - 25.61; p = 0.024) and for mechanical ventilation (OR = 6.25; 95%CI = 1.33 - 29.37; p = 0.02). Patients who needed mechanical ventilation by 48 hours from acute kidney injury diagnosis had higher urea (44.6 versus 74mg/dL, p = 0.039) and sodium (138.6 versus 144.1mEq/L; p = 0.036). Conclusion Mortality among critically ill patients with community-acquired bacterial meningitis and acute kidney injury is high. Acute kidney injury severity was associated with even higher mortality. Thrombocytopenia was associated with severer acute kidney injury. Higher urea was an earlier predictor of severer acute kidney injury than was creatinine.

Conclusion: Mortality among critically ill patients with communityacquired bacterial meningitis and acute kidney injury is high. Acute kidney injury severity was associated with even higher mortality. Thrombocytopenia was associated with severer acute kidney injury. Higher urea was an earlier predictor of severer acute kidney injury than was creatinine. another etiologic agent, but its role in CABM is usually restricted to extremes of age and immunocompromised patients. (6) The classic triad of altered mental status, neck stiffness and fever is being slowly replaced by a tetrad that also includes headache. Although only a few patients present with all these classic findings, at least 2 of them will be present in 95% of cases. (3) Definite diagnosis relies on cerebrospinal fluid (CSF) analysis, but no diagnostic procedure should prevent early treatment of suspected cases.
Overall mortality in CABM ranges from 8.5 to 25%. (7)(8)(9)(10) Patients may require intensive care, mostly due to severe impairment of mental status, septic shock and organ failure. (10) In this group of patients, mortality rises to 40 -56%. (11,12) Among septic patients, mortality can be as high as 77.4%. (13) Acute kidney injury (AKI) has a significant prevalence (6 -23%) in intensive care units (ICU) (14,15) and is also an important determinant of outcome in this group of patients. (16,17) Nevertheless, there are very few studies assessing AKI in critically ill patients with meningitis. The aim of this study was to investigate predictors of poorer outcome in critically ill patients with CABM and AKI.

METHODS
This is a retrospective study including patients admitted to the ICU of a tertiary care infectious disease hospital in Fortaleza (CE), Brazil with a confirmed diagnosis of CABM complicated with AKI in the period between January 2003 and December 2015. This study was reviewed and approved by the Ethics Committee of Hospital São José de Doenças Infecciosas (CAAE: 12565613.4.0000.5054).
Community-acquired bacterial meningitis was diagnosed by CSF analysis. Patients who presented with previous kidney disease or used nephrotoxic drugs that were not related to the current hospitalization were excluded from the study. Acute kidney injury was diagnosed and classified according to Kidney Disease: Improving Global Outcomes (KDIGO) criteria. (18) To determine the impact of CABM infection on the kidney, renal function was assessed using serum creatinine and urea as well as urine output quantification. Further data included patient demographics, drugs in use, and inhospital survival. Hemoglobin, hematocrit, white blood cell count, platelet count, sodium, potassium, aspartate aminotransferase (AST), alanine aminotransferase (ALT) and arterial blood gas analysis were among the laboratory data collected in the study.
Clinical and laboratory data were compared between groups, such as survivors versus non-survivors, severe versus non-severe AKI (KDIGO 3 versus non-KDIGO 3), vasopressors use versus non-vasopressors use, mechanical ventilation (MV) versus non-MV. Parameters on admission and at AKI diagnosis were also correlated with poor outcome in the following 48 hours.
Statistical analysis was performed using Statistical Package for Social Science (SPSS), version 20.0 for Windows (IBM, USA). The Kolmogorov-Smirnov test was applied to sort out the continuous variables that follow a normal distribution from the variables that do not follow a Gaussian curve. Student's t or Mann-Whitney tests were applied accordingly. Likewise, chi-squared or Fisher's exact tests were performed to make comparisons between categorical variables of different groups of patients. Variables that presented a relationship in the chi-squared or Fisher's exact tests (p < 0.05) underwent multivariate analysis by binary logistic regression. Continuous variables were recoded into categorical variables so that chi-squared or Fisher's exact tests could be performed; if statistical significance was found, binary logistic regression was carried out. Significance was set on 5% (p < 0.05). For the assessment of mortality outcome, the variables entered in the logistic regression analysis were "need for vasopressors" and "KDIGO 3"; for respiratory outcome, the variables "thrombocytopenia" and "sodium levels above 140mEq/L" were analyzed; for renal outcome, "thrombocytopenia" was entered into regression analysis; and for circulatory outcome (need for vasopressors), the variable "hematocrit above 35%" was analyzed.

RESULTS
A total of 41 patients were included, with a mean age of 41.6 ± 15.5 years, and 56% were males. Mean time between ICU admission and AKI diagnosis was 5.8 ± 10.6 days. Overall mortality was 53.7%. There was no statistically significant difference between survivors and non-survivors regarding age (42.72 ± 19.09 versus 41.76 ± 11.84 years, p = 0.855). The most common comorbidity was diabetes mellitus (22%), followed by arterial hypertension (14.6%) and liver disease (7.3%). One patient had history of cancer (2.4%).
Non-survivors required vasopressors during ICU stay more often than survivors (p = 0.001) and, in the multivariate analysis, need for vasopressors was a risk factor for dying (p = 0.004; OR = 12.5, 95%CI = 2.23 -70. 19). However, patients who used vasopressor drugs in the first 48 hours after AKI diagnosis did not present increased mortality (p = 0.220).
Patients who developed early AKI (in the first 24 hours after ICU admission) did not have higher mortality (p = 1.00) but required MV in the first 2 days after AKI diagnosis more often than those who had delayed AKI (p = 0.029). Likewise, patients who needed MV by 48 hours from AKI diagnosis had greater urea (44.62 ± 27.30 versus 74.05 ± 33.69mg/dL, p = 0.039) and sodium levels (138.62 ± 3.50 versus 144.11 ± 6.54mEq/L, p = 0.036) and presented a tendency to higher AST levels (50 ± 44.62 versus 330.82 ± 754.61UI/L, p = 0.052). Urea levels above 85mg/dL at AKI diagnosis tended to predict the need for MV in 48 hours (p = 0.061). In the multivariate analysis, sodium levels above 140mEq/L at the time of AKI diagnosis showed a tendency to increase the risk of requiring MV in 48 hours (p = 0.061; OR = 6, 95%CI = 0.92 -39. 18). There was no statistically significant age difference between patients who required intubation and those who did not (41.67 ± 14.54 versus 41.7 ± 19.19 years, p = 0.995). Tables 1 -4 display mean laboratory values on ICU  admission and AKI diagnosis and correlate them with  worse outcomes. Patients who developed KDIGO 3 AKI  had higher sodium levels on ICU admission, as shown in  table 2. At AKI diagnosis, the only parameter apart from creatinine and urea that was significantly different in the KDIGO 3 group was platelet count, as shown in table 4. However, patients who developed thrombocytopenia (platelet count < 150.000/mm 3 ) during the ICU stay did not have increased mortality (p = 1.00).
At the time of AKI diagnosis, thrombocytopenia was associated with a higher incidence of KDIGO 3 (p = 0.018) and need for dialysis (p = 0.044). In the multivariate analysis, it was a risk factor for KDIGO 3 (OR = 5.67, 95%CI = 1.25 -25.61; p = 0.024) and tended to increase the risk of requiring dialysis (OR = 8.77, 95%CI = 0.94 -81.67; p = 0.057). Thrombocytopenia at any point of the ICU stay was a risk factor for invasive ventilatory support (OR = 6.25, 95%CI = 1.33 -29.37; p = 0.02) in the multivariate analysis.
Hematocrit above 35% at AKI diagnosis correlated with a higher need for vasopressors (p = 0.041) and an increased risk for needing vasopressors in the multivariate analysis (p = 0.036; OR = 4.75, 95%CI = 1.11 -20.39). There was no statistically significant age difference between patients who required such drugs and those who did not (41.9 ± 13.1 versus 41 ± 20.8 years, p = 0.860).

DISCUSSION
Acute kidney injury is still a poorly studied complication among patients with CABM, a disease that was associated in the present study with very high mortality at a rate higher than described in the literature. (11,12) A study performed in another city in Brazil, which included 294 CABM patients from both the ICU and general wards, highlighted a threefold increase in mortality among patients with thrombocytopenia or a positive blood culture. (19) Low platelet count also increases mortality in septic patients. (20) Thrombocytopenia was an important prognostic factor in this study , but it was not a predictor of mortality. This may be explained by the obvious differences between the samples of the two studies. Nonetheless, the fact that thrombocytopenia was a risk factor for intubation and, at AKI diagnosis, for severer AKI and dialysis corroborates that the clinician should be careful to monitor variations in platelet count.   A study by Hollestelle et al. has suggested that disseminated intravascular coagulation and consumption of von Willebrand factor and granzyme-B may be involved in the mechanism of thrombocytopenia in CABM. (21) Thrombocytopenia has been associated with severe AKI in various conditions, such as heat stroke, Hantaan virus infection, and other infectious diseases, including leptospirosis, dengue and malaria. (22)(23)(24) The mechanism of such an association is still elusive. The present study suggests that CABM follows the same pattern, since thrombocytopenia at AKI diagnosis increased almost sixfold the risk of KDIGO 3 and tended to be a risk factor for dialysis. Higher serum creatinine correlated with worse renal outcome only at AKI diagnosis, while higher serum urea predicted KDIGO 3 both on admission and AKI diagnosis. Since KDIGO diagnostic criteria for AKI diagnosis are based on creatinine rather than urea levels, it is expected that AKI is diagnosed with the rise of creatinine in the serum. However, table 2 shows that patients who eventually developed KDIGO 3 AKI already had higher urea levels on ICU admission before the rise in serum creatinine took place. This finding suggests a faster increase in urea than in creatinine levels.
Another study compared early and delayed AKI in the same ICU, but in a different period from the present study. It included 147 patients with various infectious diseases who developed AKI and drew the conclusion that delayed AKI predicted MV and showed a tendency to increase mortality, although patients who developed early AKI had higher APACHE II scores. (25) In the present study, such a tendency was not observed, and early AKI was associated with intubation within 48 hours from AKI diagnosis.
Older age was not a predictor of severe AKI, death or need for vasopressors or invasive ventilatory support. However, a study that included 65 CABM critically ill patients (who did or did not develop AKI) showed that patients who presented "adverse clinical outcomes" (e.g., neurologic sequelae or death) were older than those who did not. (11) S. pneumoniae is capable of elevating vascular endothelial growth factor (VEGF) levels in CSF. It could be involved in the pathogenesis of brain edema in some CABM patients via an increase in vascular permeability, although VEGF inhibition did not diminish cerebral edema. (26,27) However, VEGF production stimulation is not limited to the CSF, as it was first demonstrated that S. pneumoniae was capable of inducing VEGF production by human neutrophils in the peripheral blood. (28) This phenomenon may constitute a mechanism of hypotension via increased vascular permeability and third-space fluid loss. Such a proposition is consistent with the finding that higher hematocrit at AKI diagnosis elevated the risk of requiring vasopressors almost fivefold.
Some laboratory variables at AKI diagnosis seemed to predict intubation in 48 hours, such as higher urea and sodium levels. There are studies in the literature proposing an association between high urea levels and an increased need for ventilatory support. (29,30) A recent prospective study has pointed to urea levels > 49.25mg/dL as an independent risk factor for re-intubation in a surgical ICU. (27) Clark and Lettieri (30) developed a very specific clinical model to predict prolonged intubation, in which urea > 25mg/dL increased the risk of requiring MV for more than 14 days. A previous study by Milhaud et al. (12) also showed that high urea is an important prognosis factor in CABM patients.
The main limitations of this study derive from its retrospective nature and small sample size. Data regarding the etiological diagnosis of the agents causing CABG were not available. The study was conducted in only one region of Brazil, so disease patterns may be different in other regions of the globe.

CONCLUSION
Mortality among critically ill patients with communityacquired bacterial meningitis complicated with acute kidney injury is very high, and the clinician should be particularly careful to monitor prognostic factors in order to improve patient care. High hematocrit, serum urea and thrombocytopenia were associated with worse hemodynamic, ventilatory and renal outcomes, respectively. Low sodium levels at acute kidney injury diagnosis may also be an important prognostic factor in this group of patients. Higher serum urea was an earlier predictor of worse renal outcome compared to serum creatinine.