Factor analysis of acute kidney injury in patients administered liposomal amphotericin B in a real-world clinical setting in Japan

Liposomal amphotericin B (L-AMB) is a broad-spectrum antifungal drug that is used to treat fungal infections. However, clinical evidence of its use in patients with renal failure is limited. Here, we aimed to identify factors associated with acute kidney injury (AKI) in patients administered L-AMB. We retrospectively utilized a combination of Diagnosis Procedure Combination data and laboratory data obtained from hospitals throughout Japan between April 2008 and January 2018. In total, 507 patients administered L-AMB were identified. After L-AMB treatment initiation, AKI, which was defined as a ≥ 1.5-fold increase within 7 days or ≥ 0.3 mg/dL increase within 2 days in serum creatinine according to the KDIGO criteria, was recognized in 37% of the total patients (189/507). The stages of AKI were stage 1 in 20%, stage 2 in 11%, and stage 3 in 7%. Five factors were associated with AKI of all stages: prior treatment with angiotensin-converting enzyme inhibitors/angiotensin-receptor blockers or carbapenem; concomitant administration of catecholamines or immunosuppressants; and ≥ 3.52 mg/kg/day of L-AMB dosing. Serum potassium < 3.5 mEq/L before L-AMB therapy was associated with severe AKI of stage 2 and 3. Altogether, these factors should be carefully considered to reduce the occurrence of AKI in patients administered L-AMB.


Scientific RepoRtS
| (2020) 10:15033 | https://doi.org/10.1038/s41598-020-72135-y www.nature.com/scientificreports/ Terminology Criteria for Adverse Events (CTCAE) standard, at the start of L-AMB treatment and concomitant drug treatment are the factors associated with the occurrence of decreased renal function 10 . However, large-scale clinical evidence of AKI-associated factors in patients administered L-AMB in Japan is limited. Therefore, based on a combination of Diagnosis Procedure Combination (DPC) data and laboratory data obtained from hospitals throughout Japan, we defined AKI as a ≥ 1.5-fold increase within 7 days or ≥ 0.3 mg/dL increase within 2 days in serum creatinine levels and sought to examine the factors associated with AKI in patients treated with L-AMB.

Results
Characteristics of patients administered L-AMB. By applying the inclusion and exclusion criteria and based on the definitions described in "Methods", we identified 507 patients administered L-AMB treatment throughout Japan (Fig. 1). Only patients treated in public or private hospitals containing ≥ 200 beds were included in the analysis; thus, patients treated in university hospitals were excluded. The patient characteristics of the study population are shown in Table 1. A total of 332 cases (65%) were male patients, and 175 cases (35%) were female patients. Mean age was 66 years-old, mean body mass index (BMI) was 21.5 ± 3.7 kg/m 2 , and mean baseline estimated glomerular filtration rate (eGFR) was 96.1 ± 46.2 mL/min. More than half of the examined patients (63%, 320/507) were administered L-AMB in the hematology department. Fungal infections were observed in 36% (184/507) of patients, with aspergillosis as the most common mycosis. The median period of L-AMB administration was 11 days and median daily dose was 2.5 mg/kg/day ( Table 2).

Occurrence of acute kidney injury (AKI) in patients administered L-AMB.
To evaluate the occurrence of AKI in patients after L-AMB therapy initiation, we defined the occurrence of AKI as a ≥ 1.5-fold increase within 7 days or ≥ 0.3 mg/dL increase within 2 days in serum creatinine levels between the day after L-AMB treatment initiation and 7 days following treatment termination (Fig. S1). As shown in Table 3, the overall occurrence of AKI was 37% (189/507). AKI patients were assigned stages 1-3 according to the KDIGO AKI criteria (stage 1: ≥ 1.5 to < twofold or ≥ 0.3 mg/dL creatinine increase, stage 2: ≥ 2 to < threefold creatinine increase, stage 3: ≥ threefold creatinine increase or ≥ 4.0 mg/dL of creatinine or the initiation of dialysis) between the date of the minimum creatinine reference measurement for AKI and 7 days following L-AMB treatment termination (the maximum period was 89 days). The stages of AKI were stage 1 in 20%, stage 2 in 11%, and stage 3 in 7%. Among patients with AKI stage 3, 2 cases required dialysis. AKI occurred at approximately 9 days after the start of L-AMB administration in all patients. For 2 patients requiring dialysis, treatment began 20 days after L-AMB therapy initiation.
Identification of factors associated with the occurrence of AKI in patients receiving L-AMB. To identify the factors associated with AKI in patients administered L-AMB, we evaluated 62 patient characteristics, including sex, age, disease history, patient condition before and during L-AMB treatment, prior and concomitant use of drugs, and L-AMB dosing and period. By performing univariate regression analysis with the 62 patient characteristics, we identified 27 candidate factors associated with AKI (p < 0.2) ( Table 4). These included < 65 years old, disease history (hypertension, heart failure), baseline patient conditions (catecholamine therapy, ≥ 60 mL/min eGFR, and < 3.5 mEq/L potassium), patient conditions during L-AMB administration Scientific RepoRtS | (2020) 10:15033 | https://doi.org/10.1038/s41598-020-72135-y www.nature.com/scientificreports/ (catecholamine treatment and < 3.5 serum potassium), prior treatment with immunosuppressants, steroids, angiotensin-converting enzyme (ACE) inhibitors/angiotensin-receptor blockers (ARBs), diuretics, several antibiotics, and cytotoxic antineoplastic agents, concomitant treatment with immunosuppressants, steroids, ACE inhibitors/ARBs, diuretics, several antibiotics, and ≥ 1,000 mL/day of fluid replacement, and the daily dose of L-AMB. By using stepwise regression, 7 of the 27 variables identified by univariate logistic regression were selected. In addition to the 7 variables, 2 clinically important variables (eGFR ≥ 60 mL/min and serum potassium < 3.5 mEq/L prior to L-AMB therapy), which were identified by univariate logistic regression (p < 0.2), were simultaneously evaluated in the final multivariate regression model. As shown in Table 5, five independent factors that were significantly associated with AKI in patients treated with L-AMB (p < 0.05) were identified through multivariate regression analysis. These factors included: catecholamine treatment during L-AMB administration ( (Tables S3, S4).

Discussion
L-AMB is a broad-spectrum antifungal drug that is used for the treatment of invasive fungal infections. However, clinical evidence of its use in patients with renal failure is limited. In this study, we aimed to identify factors associated with the occurrence of AKI in patients administered L-AMB. Based on a combination of claims data and laboratory data, we identified the following five AKI-associated factors: prior treatment with ACE inhibitors/ ARBs or carbapenem, concomitant administration of catecholamines or immunosuppressants, and ≥ 3.52 mg/ kg/day of L-AMB dosing. Moreover, serum potassium < 3.5 mEq/L before L-AMB therapy was associated with severe AKI of stage 2 and 3. Therefore, patients with those factors might require frequently monitoring for AKI. As our dataset consisted of many patients (507 patients) administered L-AMB, and was obtained from hospitals throughout Japan, the factors identified in this study could serve as reliable indicators.
Several studies have reported an association between factors identified in this study and the occurrence of AKI. First, immunosuppressant calcineurin inhibitors are known to increase the risk of renal injury 11 . In the present study, we found that calcineurin inhibitors, such as tacrolimus (58%; 26/45 patients) and cyclosporin (29%;  Second, consistent with our finding that the use of catecholamines during L-AMB administration was associated with AKI, patients with shock treated with catecholamines often causes renal failure 12 . Third, ACE inhibitors/ ARBs may increase serum creatinine owing to the decrease in intraglomerular pressure, resulting in an increase in the occurrence of renal dysfunction, especially when the deterioration of general conditions, such as infection, occurs 13 . Furthermore, carbapenems, such as meropenem and imipenem/cilastatin, have been reported to cause an increase in serum creatinine and serum urea 14 . As imipenem is highly nephrotoxic, it is used in combination with the nephrotoxicity-reducing drug cilastatin. However, in this study, as the occurrence of AKI was high in patients treated with meropenem (44%, 91/206 patients) and imipenem/cilastatin (54%, 47/87 patients) prior to L-AMB therapy initiation compared to all patients (37%), those drugs may be used for critically ill patients. We showed that ≥ 3.52 mg/kg/day of L-AMB doses is significantly associated with the development of AKI. L-AMB may induce AKI through tubular injury and renal vasoconstriction 15,16 . Tubular injury may be induced by intramembranous pore formation or vacuolation of the epithelial cells in the distal convoluted tubule 17 , while renal vascular resistance may be increased by activating the tubuloglomerular feedback mechanism 18 . Previous studies with a small population did not identify a correlation between L-AMB dosage and the occurrence of   www.nature.com/scientificreports/ AKI 9,10 . Since the present study is based on data retrieved from a large patient cohort, our results might be more reliable than those obtained with a small cohort. Altogether, L-AMB daily dose and the cumulative dose should be carefully considered.
In the present study, serum potassium levels < 3.5 mEq/L (hypokalemia) prior to L-AMB treatment were associated with stages 2 and 3 AKI. Importantly, chronic, persistent hypokalemia is associated with AKI through vacuolar degeneration of proximal renal tubule cells and distal renal tubule cells 19 . This degeneration may be caused by impaired angiogenesis 20 , tubular cytoplasmic vacuolization 21 , and/or interstitial scarring caused by renal cytogenesis 22 . Therefore, intervention for hypokalemia prior to L-AMB administration is essential for reducing the occurrence of AKI.
Baseline eGFR was found to be an important indicator of renal function. Importantly, baseline eGFR levels did not correlate with the duration and daily dose of L-AMB administration (p = 0.414 and p = 0.387, respectively, the Jonckheere-Terpstra trend test). These results suggest that L-AMB could be administered without dosage adjustment.
During L-AMB treatment, patients frequently showed abnormal serum levels of blood components, such as hematopoietic cells or electrolytes. Therefore, we opted to focus on patients with serum levels of blood components within the range of the CTCAE standard on the day of and within 7 days before L-AMB treatment initiation, and examined those levels between the day after L-AMB treatment initiation and 7 days following treatment termination. By investigating blood cell parameters (WBC and PLT) and the levels of serum biochemicals (total protein, BUN, uric acid, AST, ALT, LDH, γ-GTP, ALP, total bilirubin, glucose, Ca, Na, K, and Cl), we observed abnormal decreases below or increases over the standard values for these parameters in 41-78% of patients administered L-AMB. Severe abnormalities, as defined by a CTCAE grade ≥ 3 except for death, were observed in 2-33% of patients administered L-AMB. Notably, lower baseline eGFR was correlated with a higher occurrence of abnormal serum levels of chloride (p = 0.045, the Cochran-Armitage trend test).
In this study, we identified 2 patients who underwent dialysis after the completion of L-AMB treatment. Those patients died immediately after the completion of dialysis without renal recovery. One patient received dialysis for 2 days started from 6 days after L-AMB therapy termination and died on the day after completion of dialysis, while another patient received dialysis for 3 days initiated from 5 days after L-AMB therapy termination and died 4 days after completion of dialysis. In these patients, serum creatinine levels on the day before death did not recover to the levels observed before AKI (data not shown).
This study had several limitations owing to its retrospective nature. First, the generalizability of the findings presented herein requires further discussion. This is because the database used to analyze the current data did not contain data from university hospitals where infectious disease experts may work and patients with severe infectious disease and comorbidities may visit for treatment, and facilities that had less than 200 beds. Moreover, hospital transfers of patients could not be tracked. Thus, our study findings may not be fully representative of patients treated throughout Japan. Second, we could not evaluate AKI by assessing decreased urine volume and/ or AKI biomarkers 23 , as these parameters could not be obtained from the database used in this study. Therefore, further large-scale prospective studies, which include the value of serum creatinine as well as urine volume and AKI biomarkers, are required to verify the results. Finally, for the sensitivity analysis performed to identify the factors related to stages 2 and 3 AKI, although more variables (two additional variable) than the permissible number (8) were identified, all factors were included in the multivariate regression model to adjust all important confounding factors.

conclusions
Based on a combination of claims data and laboratory data obtained throughout Japan, we identified five AKI-related factors: prior treatment with ACE inhibitors/ARBs or carbapenem, concomitant administration of catecholamines or immunosuppressants, and ≥ 3.52 mg/kg/day of L-AMB dosing. Moreover, serum potassium < 3.5 mEq/L before L-AMB therapy was associated with severe AKI in patients with stages 2 and 3 AKI. Therefore, those factors should be carefully considered to attenuate the development of AKI in patients administered L-AMB. Study population. Patients administered L-AMB during hospitalization were selected according to a set of criteria. The inclusion criteria were: L-AMB treatment for the first time, age ≥ 18 years-old on the first day of the month of L-AMB treatment initiation, at least two measurements of serum creatinine performed between 6 days before L-AMB treatment initiation and 7 days after L-AMB treatment termination, at least one measurement of serum creatinine performed between the day after L-AMB treatment initiation and 7 days after L-AMB treatment termination, and at least one measurement of serum creatinine performed between 180 and 7 days before L-AMB treatment initiation including the period before hospitalization. The exclusion criteria were: unavailability of height or weight data on the day of or within 1 year of L-AMB treatment initiation, which were required to calculate eGFR, a L-AMB dose > 6 mg/kg/day, dialysis on the day of or within 3 days prior to L-AMB administration initiation, and ≥ 1.5-fold increase within 7 days or ≥ 0.3 mg/dL increase within 2 days in serum creatinine on the day of or within 7 days before L-AMB treatment initiation to exclude patients who developed AKI before the initiation of L-AMB therapy.
Assessments. L-AMB treatment duration was defined as the time from treatment initiation to discontinuation. L-AMB discontinuation was defined as an administration interval ≥ 8 days. According to the KDIGO AKI criteria 25 , we defined AKI as a ≥ 1.5-fold increase within 7 days or ≥ 0.3 mg/dL increase within 2 days in serum creatinine between the day after L-AMB treatment initiation and 7 days following L-AMB administration termination (Fig. S1). As the time for a creatinine measurement in a day could not be obtained from database, ≥ 1.5-fold or ≥ 0.3 mg/dL increases in serum creatinine were evaluated for a total of 8 or 3 days, respectively. Urine volume was not employed as part of the definition because we could not obtain this information from the database. AKI patients were assigned three stages based on the KDIGO AKI criteria between the date of the minimum creatinine reference measurement for AKI and 7 days following L-AMB treatment termination (the maximum period was 89 days). The criteria were: stage 1, ≥ 1.5-to < 2-fold increase or ≥ 0.3 mg/dL increase in serum creatinine; stage 2, ≥ 2-to < 3-fold increase in serum creatinine; stage 3, ≥ threefold increase in serum creatinine, ≥ 4.0 mg/dL of serum creatinine or the initiation of dialysis 25 . Adjusted eGFR was calculated by using the following formula for Japanese which correlates with other formulas such as CKD-EPI and MDRD 26 : where, Body surface area (m 2 ) = 0.007184 × [weight (kg)] 0.425 × [height (cm)] 0.725 , Cr = the concentration of serum creatinine (mg/dL). For calculation of adjusted eGFR at baseline, baseline serum creatinine was defined as the minimum level of serum creatinine measured between 180 and 7 day before the initiation of L-AMB therapy, including the period before hospitalization 27 . eGFR levels were divided into six groups: ≥ 90 mL/min; ≥ 60 to < 90 mL/min; ≥ 45 to < 60 mL/min; ≥ 30 to < 45 mL/min; ≥ 15 to < 30 mL/min; and < 15 mL/min. To examine the abnormality of blood components during L-AMB treatment, we opted to focus on patients with serum levels of blood components within the range of the CTCAE standard on the day of and within 7 days before L-AMB treatment initiation, and evaluated the serum levels of blood cell parameters (WBC and PLT) and the levels of serum biochemicals (total protein, BUN, uric acid, AST, ALT, LDH, γ-GTP, ALP, total bilirubin, glucose, Ca, Na, K, and Cl) between the day after L-AMB treatment initiation and 7 days following treatment termination. Abnormal serum levels were defined as a decrease and increase in the above and below standard values, respectively, while severe abnormality was defined as a Grade ≥ 3 CTCAE, except for those of deceased patients. The missing values, including baseline serum creatinine, were not used and complemented for analysis.
Variables and statistical analyses. Logistic regression analysis was conducted to identify the factors associated with AKI in patients administered L-AMB. The clinicians selected 62 variables from the patient characteristics that were considered to be related to AKI. Sex and age were obtained on the first day of the month of L-AMB therapy initiation. Diabetes, hypertension, chronic kidney disease, and heart failure were identified using the corresponding ICD-10 codes which were registered on the month of L-AMB therapy initiation. The ICD-10 codes for chronic kidney disease are summarized in Supplementary Table S6. Treatment with insulin, which was performed for patients with diabetes, was identified between the admission date and the date of L-AMB therapy initiation. Hepatic dysfunction was defined as ≥ 120 IU/L of either Aspartate transaminase (AST) or alanine transaminase (ALT) on the most recent day of L-AMB therapy initiation measured between the day of and 7 days before L-AMB initiation. Treatment with catecholamine prior to L-AMB therapy was identified on the day of or within 7 days before L-AMB therapy initiation, while concomitant catecholamine treatment during L-AMB therapy was identified between the day after L-AMB therapy initiation and the date of L-AMB therapy termination. Hypoalbuminemia (serum albumin ≤ 3 g/dL), hypokalemia (serum potassium < 3.5 mEq/L), and hyperkalemia (serum potassium > 5.0 mEq/L) before L-AMB therapy were evaluated by using the most recent value of serum albumin and potassium on the date of L-AMB therapy initiation measured between the day of and 7 days before L-AMB therapy initiation. On the other hands, hypokalemia after the initiation of L-AMB therapy was evaluated with the minimum value of serum potassium measured between the day after L-AMB therapy initiation and the date of L-AMB therapy termination. Drug treatment before L-AMB therapy was identified between the admission date and the day before L-AMB therapy initiation. Concomitant drug treatment during L-AMB therapy was identified between the date of L-AMB therapy initiation and ter- www.nature.com/scientificreports/ mination. Fluid replacement (≥ 1,000 mL/day) before L-AMB treatment initiation was identified within 7 days before L-AMB therapy initiation, while that after L-AMB therapy initiation was identified between the date of L-AMB therapy initiation and 6 days after L-AMB therapy initiation or the date of L-AMB therapy termination. We selected patients who had no missing values for variables with a P value < 0.2 in univariate logistic regression analysis conducted with the maximum number of patients. 62 variables in selected patients were then subjected to univariate logistic regression analysis. Variables with a P value < 0.2 were selected using stepwise regression according to the backward elimination selection algorithms using the Akaike information criterion (AIC). In addition to the factors selected by stepwise regression, the clinically important 4 factors: baseline eGFR < 60 mL/ min, serum potassium < 3.5 mEq/L prior to L-AMB therapy, treatment with contrast agents prior to L-AMB therapy, and concomitant treatment with non-steroidal anti-inflammatory drugs (NSAIDs) were included in the final multivariate logistic regression model when those variables had a P value < 0.2 in the univariate logistic regression analysis. Multivariate logistic regression analysis was thus conducted with the above variables. The variance inflation factor (VIF) was calculated to identify the multicollinearity of the explanatory variables. Receiver operating characteristic (ROC) curves were used to identify the cut-off values for continuous variables, including the daily and cumulative doses of L-AMB. For sensitivity analysis, patients without AKI and with stage 1 AKI or those without AKI and with stages 2 and 3 AKI were subjected to logistic regression analysis. Continuous variables are presented as average ± standard deviation. Student's t-test was employed to compare two groups for continuous variables, while the Fisher's exact test was used for two categorical variables. The correlation between baseline eGFR and treatment duration and daily dose of L-AMB was evaluated by the Jonckheere-Terpstra trend test, while that between baseline eGFR and abnormality of blood components between the day after L-AMB treatment initiation and 7 days following treatment termination was evaluated by the Cochran-Armitage trend test.