Fifty-four studies were included (n=9,228 for toxicity analyses). There were 21 RCTs (n=3,721), 16 prospective cohort studies (n=2,153), 11 retrospective cohort studies (n=3,269), 3 case-control studies (n=96) and 3 case series (n=46). Not all enrolled patients were included in the toxicity analysis, hence the discrepancy in the total number of patients analysed.
Infusion-related reactions (48 studies).
The frequency of fever ranged from 4 to 100% for AmB, from 0 to 36% for L-AmB, and from 12 to 88% for ABCD. Most of the studies were too heterogeneous to pool. Fever was less commonly associated with fluconazole than AmB (4 homogeneous studies).
Nausea with or without vomiting was most common with itraconazole (mean 19.7%, range: 6 to 24). There was no difference in nausea between fluconazole and AmB (2 studies) or between L-AmB and conventional AmB (2 studies).
Bronchospasm or cough without dyspnoea or hypoxia was more common with itraconazole (9.4%; n=233) and ABLC (8.5%; n=307) than with AmB (7.2%; n=1073).
The meta-analysis showed no difference in rash between fluconazole and AmB (3 RCTs).
Discontinuation of the study drug due to transfusion-related reaction was twice as common with ABCD (3.9%) and ABLC (3.7%) than with conventional AmB (1.4%).
The studies used a wide variety of regimens. Using the most common criteria, L-AMB was the least nephrotoxic of the lipid-soluble formulations (15%; n=797) and AmB was the most nephrotoxic (33.2%, 95% CI: 30.7, 36).
The meta-analysis showed that L-AmB was less nephrotoxic than conventional therapy; the OR (3 RCTs) was 0.40 (95% CI: 0.30, 0.54).
Most studies did not define the severity or only described potassium replacement.
No studies reported discontinuation of the study drugs due to hypokalaemia. One patient taking ABCD died of cardiac arrhythmia.
The studies defined hepatotoxicity in a variety of ways or gave no criteria (n=2,695).
Patients with underlying liver damage were excluded from itraconazole studies, but itraconazole was stopped in 1.6% due to hepatotoxicity (5 studies, n=442). Across other treatments, discontinuations due to hepatotoxicity ranged from 0.2% with AmB to 1.1% with ABLC. The meta-analysis showed no difference in hepatotoxicity between L-AmB and AmB (based on 2 RCTs in neutropoenic patients).