Coadministration of Liposomal Amphotericin B and Contrast Medium Does Not Increase Risk of Kidney Injury

ABSTRACT

Intravenous radiographic contrast medium and amphotericin B are commonly required in the care of patients with fungal infections. Both interventions have proposed nephrotoxicity through similar mechanisms. We systematically examined patients who received coadministration of liposomal amphotericin B (AmBisome; GE Healthcare) and intravenous contrast medium within a 24-h period and compared the results for those patients with the results for patients who underwent non-contrast medium studies. We found 114 cases and 85 controls during our study period. Overall, no increased risk of renal injury was seen with coadministration of these 2 agents. Adjustment for age, baseline kidney function, and other clinical factors through propensity score adjustment did not change this result. Our observations suggest that, when clinically indicated, coadministration of contrast medium and liposomal amphotericin B does not present excess risk compared with that from the administration of liposomal amphotericin B alone.

INTRODUCTION

Patients with impaired immunity, particularly impaired cellular immunity, are at high risk for fungal infection. A diagnosis of fungal infection often requires radiographic studies with intravenous (i.v.) contrast medium (here called contrast) and empirical or directed treatment with antifungals, including amphotericin B formulations. Unfortunately, the use of both i.v. contrast and amphotericin B is limited by renal impairment. Since these diagnostic and therapeutic elements may be used concurrently during a fungal illness, concerns about nephrotoxicity and renal impairment are paramount.

Two main mechanisms for the nephrotoxicity associated with amphotericin B have been proposed. The therapeutic mechanism of action involves the increased permeability of fungal cell membrane elements, leading to cell death. Amphotericin B exerts its fungicidal effect by binding to the ergosterol found in fungal cell membranes but not in animal cell membranes. In humans, however, binding of the amphotericin B to cholesterol in mammalian cell membranes leads to renal tubular acidosis and altered renal ion gradients (1). Animal studies have shown another mechanism involving direct vasoconstriction of the afferent glomerular arteriole, with a resultant reduction in the glomerular filtration rate (GFR) (2). Several mechanisms have been proposed to explain this effect, although systematic studies have generally been retrospective and inconclusive. Existing studies to explain this effect on GFR are largely confounded by the other nephrotoxins commonly administered to critically ill patients who have systemic fungal infections (1).

Salt loading has emerged as a preferred strategy for avoiding amphotericin B-related nephrotoxicity (3). This approach is thought to cause the reduced sensitivity of the tubuloglomerular feedback mechanism through volume expansion, thus preserving the GFR. However, other signs of tubular dysfunction (e.g., hypokalemia, hypomagnesemia, renal tubular acidosis) persist. The volume expansion may also decrease atrial natriuretic peptide release, thus acting on the vasoconstrictive mechanism of amphotericin B (2). Lipid formulations of amphotericin B (such as AmBisome [GE Healthcare], used at our institution and in the present study) are less nephrotoxic and are predominately used, yet the same mechanisms involved in volume expansion may cause renal injury. Because of the important reduction in nephrotoxicity, salt-loading strategies are not universally used with liposomal formulations.

Radiographic studies with i.v. contrast are an important diagnostic tool when a systemic fungal infection is suspected (4). However, contrast is also associated with acute kidney injury (AKI). The exact mechanism of this is unclear but is also proposed to be mediated by renal vasoconstriction (5, 6) and hypoxia (7), as well as alterations in rheologic factors and hemodynamics (8, 9). Overall, the mechanisms seem to be similar to those for acute tubular necrosis, although the renal injury is typically less severe than that associated with necrosis induced by other toxins (10). This kind of renal injury is generally reversible; however, it often is a part of multiple renal insults, and certainly in patients with known renal disease (e.g., chronic kidney disease [CKD], diabetic nephropathy), the risk appears to be much greater (11, 12).

Although irreversible renal injury is uncommon with either agent, coadministration of i.v. contrast and amphotericin B formulations may confer a greater risk than administration of either agent alone. The transient decrease in renal clearance caused by either agent may increase the toxicity level of the other agent, as observed with coadministration of angiotensin-converting enzyme inhibitors (ACEs) and nonsteroidal anti-inflammatory drugs (13). In a preliminary analysis performed as part of a quality improvement project, we identified 38 patients who were coadministered an i.v. contrast and an amphotericin B formulation within 24 h. Among these patients, 19 met the Acute Kidney Injury Network (AKIN) stage I renal injury criterion (creatinine increase, ≥0.3 mg/dl), and 5 did not have recovery, reaching a new baseline renal function (14). Although the absolute rate of injury was low, the proportion of injuries was remarkable. It was more severe than that from ACE-angiotensin receptor blocker associations, which are part of our routine screening for radiologic studies, previously reported (15). Given these preliminary data, we conducted a larger study to evaluate whether coadministration of contrast with liposomal amphotericin B (AmBisome) is a preventable risk factor for iatrogenic renal injury.

RESULTS

In total, 425 computed tomography (CT) scans occurred within 24 h of liposomal amphotericin B coadministration during the study period. The study was restricted to index cases only, so 199 unique patients were available for analysis, with 85 patients (43%) being involved in noncontrast studies and 114 patients (57%) being involved in contrast studies (Table 1). Patients with noncontrast studies had a median age significantly greater than that for patients who received contrast (median age, 58 years [noncontrast patients] versus 54 years [contrast patients]; P = 0.03). The chronic illness burden, measured with the Charlson comorbidity index, was similar across the groups (P = 0.36). The rates of human immunodeficiency virus infection were similar across the groups (1 in the contrast group and 4 in the noncontrast group; P = 0.04). Likewise, in the contrast and the noncontrast groups, the rates of solid organ transplant (33.3% versus 41.4%; P = 0.24) and bone marrow transplant (40.2% versus 41.9%; P = 0.86) were similar. The median baseline estimated GFR was slightly greater for patients who received contrast (92 patients) than patients who did not receive contrast (83 patients) (P = 0.006), and the prevalence of previously diagnosed chronic kidney disease was greater for patients who did not receive contrast, although the latter difference did not achieve statistical significance (19% for noncontrast patients versus 11% for contrast patients; P = 0.10). Propensity score matching yielded a cohort of 65 patients in the contrast group and 65 in the noncontrast group. Following matching, all clinical variables were similar between the contrast and noncontrast groups.

TABLE 1.

Demographic characteristics of cohort before and after propensity score adjustment^a

Characteristic	Unadjusted			1:1 Matched
	Value(s) for:		Unadjusted P valueb	Value(s) for:		Matched P value
	Contrast group (n = 114)	Noncontrast group (n = 85)		Contrast group (n = 65)	Noncontrast group (n = 65)
Median (IQR) age (yr)	54 (38–62)	58 (47–69)	0.03	57 (40–63)	55 (43–66)	0.53
No. (%) female patients	66 (58)	54 (64)	0.42	38 (58)	36 (55)	0.73
Renal function
Median (IQR) baseline Scr level (mg/dl)	0.8 (0.7–0.9)	0.9 (0.7–1.2)	0.003
Median (IQR) baseline eGFR (ml/min/1.73 m2)	92 (75–123)	83 (62–105)	0.006	87 (74–117)	87 (70–114)	0.40
No. (%) of patients with the following preexisting comorbidity:
Nonmetastatic malignancy	79 (69)	55 (65)	0.49	45 (69)	43 (66)	0.72
Metastatic solid tumor	7 (6)	1 (1)	0.08	1 (2)	1 (2)	0.99
DM with vascular complication	3 (2)	6 (7)	0.17	2 (3)	3 (5)	0.57
DM without complication	11 (10)	15 (18)	0.10	7 (11)	7 (11)	0.99
PVD	9 (8)	10 (12)	0.36	5 (8)	5 (8)	0.99
CKD	12 (11)	16 (19)	0.10	9 (14)	6 (9)	0.37
CVD	8 (7)	9 (11)	0.37	3 (5)	4 (6)	0.66
CHF	21 (18)	17 (20)	0.78	12 (18)	11 (17)	0.81
COPD	27 (24)	25 (29)	0.36	17 (26)	16 (25)	0.84
Mild liver disease	14 (12)	6 (7)	0.23	8 (12)	6 (9)	0.53
Moderate liver disease	9 (8)	7 (11)	0.51	7 (11)	4 (6)	0.37
Myocardial infarction	4 (3.5)	4 (4)	0.67	2 (3)	1 (2)	0.57
Ulcer	7 (6)	1 (1)	0.08	1 (2)	1 (2)	0.99
Rheumatic disease	5 (4)	5 (5)	0.63	4 (6)	4 (6)	0.99
Median (IQR) Charlson comorbidity index	4 (2–6)	4 (3–7)	0.36	4 (3–6)	3 (2–6)	0.51
Median (IQR) no. of days of amphotericin B exposure before CT	2 (0–8)	1 (0–7)	0.17	3 (0–8)	1 (0–8)	0.69

^aAbbreviations: CHF, chronic heart failure; CKD, chronic kidney disease; COPD, chronic obstructive pulmonary disease; CT, computed tomography; CVD, coronary vascular disease; DM, diabetes mellitus; eGFR, estimated glomerular filtration rate; IQR, interquartile range; PVD, peripheral vascular disease; Scr, serum creatinine.

^bBold indicates statistical significance.

Dosing of liposomal amphotericin B ranged from 3 to 7.5 mg/kg of body weight intravenously every 24 h on the basis of the indication (median, 4.8 mg/kg; interquartile range [IQR], 4.7 to 5.2] mg/kg).

The outcomes before and after propensity score adjustment are shown in Table 2. Before propensity score adjustment, patients in the noncontrast group had significantly higher post-CT peak serum creatinine (Scr) values than patients in the contrast group (median, 1.2 versus 1.1; P = 0.004) and lower nadir estimated GFR values than patients in the contrast group (median, 53 versus 65; P = 0.03). These differences were no longer present after propensity score matching. The rates of AKI stages I and II were similar between the contrast and noncontrast groups before and after propensity score adjustment (P value range, 0.69 to 0.88 for stage I and 0.57 to 0.89 for stage II). A nonsignificant tendency toward more stage III AKI was found in contrast group patients and continued after propensity score adjustment. Dialysis and end-stage renal disease outcomes were rare.

TABLE 2.

Unadjusted and propensity score-adjusted outcomes^a

Characteristic	Value(s) for:		OR (95% CI)b	P valuec
	Contrast group	Noncontrast group
No. of patients
Unadjusted	114	85	NA	NA
Stratified	114	85	NA	NA
1:1 matched	65	65	NA	NA
Median (IQR) peak Scr level (mg/dl)
Unadjusted	1.1 (0.8–1.4)	1.2 (0.9–1.7)	NA	0.04
1:1 matched	1.1 (0.8–1.4)	1.1 (0.8–1.6)	NA	0.55
Median (IQR) nadir eGFR (ml/min/1.73 m2)
Unadjusted	65 (48–91)	53 (40–80)	NA	0.03
1:1 Matched	62 (49–92)	58 (45–86)	NA	0.38
Median (IQR) change in Scr level (ng/dl)
Unadjusted	0.2 (0.1–0.4)	0.2 (0.1–0.5)	NA	0.78
1:1 matched	0.2 (0.1–0.4)	0.2 (0.1–0.4)	NA	0.72
No. (%) of patients with AKIN stage I renal injury
Unadjusted	19 (17)	16 (19)	0.86 (0.41–1.80)	0.69
Stratified	NA	NA	0.93 (0.41–2.15)	0.88
1:1 matched	12 (18)	13 (20)	0.87 (0.32–2.42)	0.80
No. (%) of patients with AKIN stage II renal injury
Unadjusted	7 (6)	7 (8)	0.73 (0.25–2.16)	0.57
Stratified	NA	NA	0.75 (0.22–2.59)	0.89
1:1 matched	4 (6)	5 (8)	0.80 (0.21–2.98)	0.74
No. (%) of patients with AKIN stage III renal injury
Unadjusted	6 (5)	1 (1)	4.67 (0.55–39.5)	0.12
Stratified	NA	NA	10.5 (0.64–174)	0.20
1:1 matched	3 (5)	0	NA	NA
No. (%) of patients undergoing dialysis
Unadjusted	1 (1)	0 (0)	NA	NA
Stratified	NA	NA	NA	NA
1:1 matched	0 (0)	0 (0)	NA	NA
No. (%) of patients with ESRD
Unadjusted	1 (0.9)	0 (0)	NA	NA
Stratified	NA	NA	NA	NA
1:1 matched	0 (0)	0 (0)	NA	NA

^aAbbreviations: AKIN, Acute Kidney Injury Network; CI, confidence interval; eGFR, estimated glomerular filtration rate; ESRD, end-stage renal disease; IQR, interquartile range; NA, not applicable; OR, odds ratio; Scr, serum creatinine.

^bOdds for contrast group versus noncontrast group.

^cBold indicates statistical significance.

To investigate the potential role of survival bias as a confounder, we constructed a survival plot comparing the status at the last follow-up or the survival status of the patients after their CT studies (Fig. 1). The median time of survival of patients who received i.v. contrast was 11.9 months, whereas it was 3.8 months for patients in the noncontrast group (P = 0.12).

FIG 1.

Plot of time of survival since the radiographic study. The median time of survival of patients who received intravenous contrast medium was 11.9 months, whereas it was 3.8 months for patients who did not receive it (P = 0.12). Dotted line, patients with contrast computed tomography; solid line, patients with noncontrast computed tomography.

DISCUSSION

Our findings suggest no significant increase in risk when i.v. radiographic contrast and liposomal amphotericin B (AmBisome) are coadministered within 24 h. Our initial findings (14) suggestive of a high risk did not bear out compared with the findings for an appropriate control group. In both study arms, nearly one-third of the patients had AKI, which is reflective of the high baseline morbidity of this group.

The nephrotoxicity of liposomal amphotericin B and contrast has come under increased scrutiny recently, with the suggestion that either agent may not be truly nephrotoxic and that transient laboratory abnormalities do not necessarily indicate renal injury.

Contrast-induced nephropathy was originally recognized in the 1960s when it was reported in large single-arm studies (16). In the 1980s and 1990s, several studies with matched control groups found no difference in the rates of renal injury in patient groups receiving i.v. contrast and noncontrast control groups (17, 18). Subsequent studies have not yet definitively answered questions about this issue, but a 2014 study found no association between contrast and renal injury when adjusting for baseline renal function (19).

Recent years have also seen a possible shift in the thinking about the toxicity of liposomal amphotericin B. Much of the initial interest in its nephrotoxicity was driven by the historical use of amphotericin B deoxycholate. Deoxycholate is directly toxic to the tubule and is absent in the liposomal formulation (20). Hence, for liposomal amphotericin B, the amount of attributable nephrotoxicity is not known.

Our study has several limitations. First, it may have involved selection bias, with providers selectively avoiding both contrast and amphotericin B products in patients believed to be at risk for renal injury. We attempted to adjust for this bias by performing propensity score adjustment; however, unmeasured confounders that affected our results may have been present. Second, our study is limited by its sample size because we did not achieve our target power. However, given that this study captured the data for all patients who had coadministration of contrast and liposomal amphotericin B (AmBisome) at a large institution over 13 years, we believe that this limitation could not be remedied without a larger multi-institutional study. Third, the cumulative effect of all other potential nephrotoxins is not quantifiable but may have confounded the results. Many of these patients were transferred from other facilities or received a scan directly after admission, and we did not have the ability to account for these nephrotoxins in our data. Fourth, our definition for a nephrotoxic window of 24 h for coadministration was a limitation. Although we believed that this was a reasonable restriction, we may have underestimated the window in which coadministration would be relevant.

Overall, our findings suggest that coadministration of liposomal amphotericin B and contrast does not pose an excess risk of AKI when patients are carefully selected to receive contrast. The patient population needing these interventions is at a high baseline risk for renal injury, but we did not find an observable effect of combining these 2 agents. When clinically indicated, these data show that clinicians can consider radiographic contrast and amphotericin B for selected patients.

MATERIALS AND METHODS

The Mayo Clinic Institutional Review Board approved this study as a minimal-risk retrospective study. We identified patients from the adult population of patients admitted to Mayo Clinic Hospital—Rochester (Rochester, MN) between 1 December 2002 and 30 June 2015. Inclusion criteria were adult patients who received i.v. liposomal amphotericin B (AmBisome) within 24 h of CT with contrast (e.g., abdomen CT with contrast, pulmonary embolism protocol CT angiography with contrast) for the case group and without contrast (e.g., chest CT without contrast) for the control group. All patients received the same contrast product, iohexol (Omnipaque; GE Healthcare). We excluded patients younger than 18 years who had preexisting dialysis-dependent renal disease, patients who had received oral contrast without i.v. contrast, or patients who had had no research.

Current protocols in our institution do not preclude the use of contrast in renal injury, on the basis of prior research indicating that contrast is not an independent risk factor for renal injury (19). Providers are asked whether the patient has renal injury, prior contrast exposure, or known allergy, and the decision to give contrast is ultimately made by the radiology service on the basis of the indication for the study.

On the basis of our preliminary analysis, we anticipated a 25% rate of AKI from administration of i.v. contrast with liposomal amphotericin B. Assuming a 5% risk of AKI in the noncontrast group, we needed 55 cases and 55 controls. Assuming a 20% risk in the noncontrast group, we needed 315 patients in each arm to demonstrate statistical significance. We planned statistical power for the latter with an enrollment target of 630 patients (315 cases and 315 controls) for a targeted α error of 0.05 and a targeted β error of 0.80. The date range of events retrieved was based on a 1-year incidence and an estimated number of years needed to achieve this power.

Case identification was performed with Mayo Clinic's Unified Data Platform (UDP) and a pharmacy medication administration database. UDP is a database combining several aspects of the electronic health record, including laboratory findings, clinical notes, billing documents, and radiographic reports. We used UDP to identify all study types with contrast (cases) or without contrast (controls) during the study period and the dates that these studies were performed. We searched the pharmacy database to determine which of these patients also received a dose of liposomal amphotericin B in the 24 h before or after the study. We also used UDP and pharmacy records to identify potential confounders, including comorbidities, demographic characteristics, infection characteristics, creatinine values over time, concomitant nephrotoxins, length of hospitalization, and preexisting dialysis status. These data were used to prepopulate a database and to calculate the incidence of AKI using AKIN criteria, the CKD status before and after the imaging event, and the incidence of new dialysis.

Data from this database were manually verified, and charts were inspected for indications that an alternative explanation was present for any changes in renal function. Our primary endpoint was development of AKIN stage I renal injury without an alternative likely cause. Our secondary outcomes were changes in baseline kidney function sufficient to reclassify the CKD stage, a new need for dialysis (temporary or permanent), death, and length of hospitalization. For patients with more than 1 episode of concurrent liposomal amphotericin B administration and a contrast CT study, only the index event was considered.

Generation of propensity scores, stratification by decile, and 1-to-1 matching of contrast and noncontrast group patients were performed using the R package MatchIt, as previously described (21). In brief, propensity scores were estimated for each included patient from a logistic regression model of the likelihood of contrast exposure based on demographic characteristics (e.g., age, sex), baseline Scr level, preexisting comorbidities (e.g., malignancy, diabetes mellitus, CKD), Charlson comorbidity score, and the number of days of amphotericin B exposure before CT covariates. Nearest-neighbor (greedy-type) matching without replacement was performed using a caliper width of a 0.15 standard deviation of the propensity score logit.

Statistical analyses were performed using R (version 3.0.3; R Project for Statistical Computing). Continuous data were presented as the median with the IQR, and categorical data were displayed as relative frequency and percentage. Clinical characteristics and outcome rate differences between the contrast and noncontrast groups were compared before matching using the Wilcoxon rank sum test, Fisher exact test, or Pearson χ² test. The collective risk of each outcome following stratification by propensity score was determined using Cochran-Mantel-Haenszel estimates. Clinical characteristic and outcome rate differences between the contrast and noncontrast groups were compared following propensity score matching using conditional logistic regression. P values of <0.05 were considered significant.