Acute Kidney Injury and Urinary Biomarkers in Human Immunodeficiency Virus–Associated Cryptococcal Meningitis

Summary Acute kidney injury (AKI) occurs commonly in human immunodeficiency virus–infected patients with cryptococcal meningitis treated with 2 weeks of amphotericin B deoxycholate and is associated with mortality. Urine protein may play a role in earlier diagnosis of AKI but needs further evaluation.

in critically ill patients [15]. Neutrophil gelatinase-associated lipocalin (NGAL; also known as Lipocalin-2) is a 25-kDa protein produced in many tissues, including renal epithelial cells, and is an established sensitive early marker of acute ischemic kidney injury [16,17]. Elevated urine NGAL levels predict a composite outcome of dialysis initiation or death during hospitalization when used as a risk stratification tool in unselected emergency department patients at the time of admission to the hospital [18]. Cystatin C (CysC), a 13-kDa proteinase inhibitor, can be measured in serum and in urine and, unlike creatinine, is not affected by muscle mass or age [19]. Urine CysC is a marker of tubular dysfunction or damage [20], and elevated levels may be associated with AKI [15]. Tissue inhibitor of metalloproteinases-2 (TIMP-2) is an inducer of G1 cell-cycle arrest, predicts severe AKI, and is associated with mortality in critically ill adult patients [21]. Neutrophil gelatinase-associated lipocalin and CysC are induced during inflammation. These biomarkers have not been evaluated in the context of HIV infection and opportunistic infections. In resource-limited settings where liposomal amphotericin is not available, a urine biomarker that could predict development of AKI early and accurately could be a valuable clinical tool to tailor duration of amphotericin therapy.
We assessed these urine biomarkers together with urine protein and creatinine levels in patients with HIV-associated cryptococcal meningitis treated with amphotericin therapy and investigated association with AKI and mortality.

METHODS
Human immunodeficiency virus-infected, antiretroviral therapy (ART)-naive adults with a new diagnosis of cryptococcal meningitis were prospectively enrolled into the Cryptococcal Optimal ART Timing (COAT) trial during 2010-2012 and treated with amphotericin B, fluconazole, and ART according to the COAT trial protocol [22,23]. Baseline and follow-up serum creatinine concentrations were measured at regular intervals as per the COAT trial protocol (5-6 time points during amphotericin treatment and at 1 and 2 weeks after completion of amphotericin therapy). Estimated GFR was calculated using the Modification of Diet in Renal Disease (MDRD) study equation [24], and AKI was defined as a decrease in GFR to <60 mL/ min/1.73 m 2 at any point within 3 weeks of cryptococcal diagnosis. Urine output was not routinely monitored and could not be used to define AKI. Ten-week and 12-month mortality was ascertained.
We included 130 participants with stored urine samples, which were collected a median of 4 (interquartile range [IQR] = 4-5) days from diagnosis of cryptococcal meningitis (which is similar to time from amphotericin initiation: median = 4; IQR = 4-5 days), and measured NGAL, CysC, and TIMP-2 by enzyme-linked immunosorbent assay (R&D Systems) and protein and creatinine (Beckman Coulter DXC 800) on stored urine samples. We summarized incidence of AKI and explored risk factors for incident AKI using Cox univariate and multivariable proportional hazards models. Models were estimated with biomarkers on a continuous scale (after log 2 transformation, so that the hazard ratio is per doubling of the biomarker). Other models compared the upper quartile versus the lower 3 quartiles (an a priori comparison to assess clinical outliers). Each biomarker was first statistically tested individually and then in a multivariable model adjusted for timing of ART initiation group, age, sex, altered mental status at diagnosis, CD4 cell count, and cerebrospinal fluid quantitative cryptococcal culture at diagnosis. We examined the association of incident AKI with 12-month mortality using time-updated Cox proportional hazards models. With the time-updated models, an indicator for incident AKI was recalculated at the time of each creatinine measurement to assess the impact of developing AKI on mortality. All analyses were completed using SAS 9.3. Relevant ethics and regulatory authority approvals were obtained. Written informed consent was obtained to participate in the COAT trial, as previously described [22].

Ten-Week Outcomes in Acute Kidney Injury Patients
The outcome of abnormal renal function was assessed in survivors at 10 weeks after diagnosis of cryptococcal meningitis. Among 53 patients who developed AKI, 51% (n = 27) died before 10 weeks, and renal function recovered to eGFR >60 mL/min/1.73 m 2 in the remaining 49% (n = 26/53) by 8 weeks. All 4 patients with abnormal renal function at baseline survived and recovered renal function to eGFR >60 mL/min/1.73 m 2 by 10 weeks.

DISCUSSION
To our knowledge, this is the first study to evaluate urine biomarkers in patients with HIV-associated cryptococcal meningitis treated with amphotericin. Acute kidney injury was a common complication and associated with high mortality. Urine protein level in the highest quartile independently predicted AKI and 12-month mortality, whereas the other biomarkers tested showed no strong associations. Urine biomarkers are being increasingly applied in clinical practice, particularly in settings where AKI is common and can potentially be mitigated by amending management strategies, such as intensive care units. Urine biomarkers have not been studied in the context of HIV-associated cryptococcosis where AKI is a common complication of amphotericin therapy and mortality is high. Urine protein can also be measured by urine dipstick, which is a cheap and readily available test in resource-limited settings. This could be a useful marker to investigate in similar settings as an adjunctive tool to aid clinical decision making. Elevated urine protein levels at initiation of therapy could prompt more intensive fluid replacement, closer monitoring of renal function, and possible decreased duration of amphotericin, but this needs to be evaluated in further studies. Acute kidney injury, as defined in this study, developed in 42% of patients in this cohort. The Kidney Disease: Improving Global Outcomes (KDIGO) published case definition of AKI consists of (1) increase in serum creatinine within 48 hours of ≥26.5 μmol/L (0.3 mg/dL) or (2) increase in serum creatinine to ≥1.5 times baseline, which is known or presumed to have occurred in the prior 7 days, or urine volume of <0.5 mL/kg/h for 6 hours [11]. We did not measure urine output in this cohort, and using serial serum creatinine values, 80% of our patients met creatinine-based KDIGO AKI case definition criteria. We used an amended case definition, which is equal to a grade 3 (severe) adverse event [25] and has clinical relevance in this critically ill cohort who required treatment with a nephrotoxic drug. Amphotericin therapy is generally interrupted or stopped when a patient's GFR declines to <60 mL/min/1.73 m 2 .
People with higher CD4 T-cell counts developed AKI more commonly in our study. The reason for this is unclear. It is possible that patients with very low CD4 T-cell counts did not survive to develop AKI, and there is thus a competing risk of death; however, CD4 count was not associated with survival in this cryptococcal cohort [22]. More women developed AKI in our  study, which is consistent with the literature, which lists women as more susceptible to AKI [26]. Tissue inhibitor of metalloproteinases-2 and urine protein levels were associated with development of AKI in univariate analysis but not after adjusting for potential confounders on a continuous scale. Urine TIMP-2 combined with insulin-like growth factor-binding protein 7 (IGFBP7) has been shown to predict development of severe AKI in acutely ill patients  within 12 hours of collection of urine sample [27]. Higher urine TIMP-2 has also been associated with mortality in acutely ill patients [21] but not in our study. Our urine samples were collected at a median of 4 days after diagnosis of cryptococcal meningitis, and AKI developed at a median of 8 days on therapy, which is a different scenario to the previous studies investigating the predictive value of TIMP-2.
Higher urine CysC was not associated with development of AKI or mortality in our study when evaluated on a continuous log 2 scale or comparing the upper quartile with lower quartiles. Cystatin C is an early marker of tubular dysfunction or damage, and it is possible that (1) the mechanism of development of AKI in our cohort was different or (2) that we measured CysC levels too long before AKI events to detect elevated levels.
We found AKI to be independently associated with mortality. Patients in this study were severely immune suppressed, and although it is possible that AKI secondary to amphotericin therapy could have played a role in mortality, there are other factors such as nosocomial sepsis, opportunistic infections, and decreased level of consciousness with decreased oral intake that likely contributed to both AKI and mortality.
A limitation of our study is that we used a nonstandard case definition of AKI. This limits comparability across studies, and importantly, patients who had normal renal function at baseline would have had to experience a much larger proportional decline in renal function before being classified as an AKI case compared with patients who entered the study with a GFR close to the threshold of 60 mL/min/1.73 m 2 . Furthermore, urine was collected after initiation of amphotericin therapy, and urine biomarker findings may reflect early subclinical amphotericin-related AKI and not purely the theoretical potential to develop AKI. The timing of best measurement of urine biomarkers is unknown, and further elucidation of the change in biomarkers over time would be of interest. Ultimately, the clinical utility of urine biomarkers would be optimal if future AKI could be predicted accurately early after diagnosis to guide dosing and duration of amphotericin therapy. There is considerable interest in shortening amphotericin regimens for treatment of cryptococcal meningitis in resource-limited settings [28,29] and a urine biomarker that could predict risk of AKI accurately and early could be valuable to guide duration of amphotericin treatment. Not all patients from the COAT trial were included because not all patients had stored urine specimens. Patients who were enrolled in the COAT trial and not included in this study had no difference in baseline CD4, HIV load, age, sex, COAT treatment arm, and cerebrospinal fluid quantitative culture (data not shown) compared with this cohort.
In conclusion, AKI occurred in 42% of HIV-infected patients treated with amphotericin B deoxycholate for cryptococcal meningitis, and AKI was associated with mortality. However, AKI was transient among survivors. Urine protein levels are easy to measure and may be useful for antecedent prediction of amphotericin-associated AKI. The utility of measuring urine protein levels for aiding decisions regarding duration of amphotericin therapy needs further evaluation.