We searched MEDLINE (1946–January 2012), EMBASE (1950–January 2012), the Cochrane database for systematic reviews, Cumulative Index to Nursing and Allied Health Literature (CINAHL1982–January 2012) and the Cochrane library (Cochrane Central Register of Controlled Trials, Cochrane database of systematic reviews, and Database of abstracts of reviews of effects) for any clinical study about fluconazole use that involved at least one paediatric patient (≤17 years). Any study with involvement of a paediatric age group participant taking at least a single dose of fluconazole was eligible. Only studies with a report of safety after exposure to fluconazole in the paediatric patients were included. There was no restriction on the language of publication of the articles as translations to extract relevant data were done; where translations were not possible, abstracts containing relevant data were used. Also included in this review were any clinical study, whether comparative or non-comparative, randomised controlled trials (RCTs) or case reports and also letters to the editors that documented exposure of a paediatric patient to fluconazole and reported on safety. Included articles and extracted data were validated by two reviewers. Search terms comprised free text words and subject headings. These included terms relating to azole or imidazole or fluconazole, adverse effects or adverse drug reactions or side effects, pharmacokinetics and drug interactions.

Data extracted from each study included the year of publication, type of study, number of paediatric patients exposed, age of paediatric patients exposed, doses of fluconazole used, route of administration and safety data. The safety data extracted were occurrence of any adverse event (AE), any drug interactions, any withdrawal due to AEs and any drug-related death.

To minimise the risk of bias, we assessed the quality of included RCTs using the CONSORT checklist for reporting of harm [¹²]. All RCTs with scores of ≥6 out of nine criteria were considered to provide good quality safety reporting. Cohort studies were scored using the STROBE checklist [¹³], where a score of >70 % is considered to be good. Case series were evaluated using the health technology assessment checklist [¹⁴], and all studies fulfilling the good or satisfactory criteria were included.

The relevant data were extracted onto the data extraction form. Participants in the study were grouped into paediatric age groups of preterm neonates (<36 weeks gestation, 0–27 days), full-term neonates (0–27 days, >37 weeks gestation), infants and toddlers (28 days–23 months), children (2–11 years) and adolescents (12–17 years). All reported AEs were pooled together from the various studies. The duration of treatment was grouped into <21, 21–42 and >42 days; the treatment dose was grouped into <3, 3–6 and >6 mg/kg; the route of administration was recorded as intravenous (IV), oral or IV and oral.

Meta-analysis was performed using the Stata/IC v.11 statistical package (StataCorp LP, College Station, TX). Only those studies rated as good were included in the meta-analysis. Studies with zero frequency were included in the meta-analysis by entering 0.5 to zero cells so that all of the information could be used.

The relative risk (RR) was calculated for these binary outcomes (RR >1 indicates a positive effect of fluconazole). We calculated the pooled relative risks with fixed effect models using the Mantel and Haenszel method. The heterogeneity of the model was examined by calculating the DerSimonian and Laird’s Q statistic [¹⁵] and the I2-statistic [¹⁶]. Both were compared with a chi-square distribution with degrees of freedom (df) equal to the number of trials minus one. We used the Q statistic for testing the presence of heterogeneity and the I2-statistic for estimating the degree of heterogeneity. When heterogeneity was observed, we used the with random effect models as suggested by DerSimonian and Laird [¹⁵].

The forest plots have been created for presenting the pooled effects of fluconazole. The effects of indication, age groups, dose range, route of administration and duration of treatment on risk of AEs in the fluconazole groups against the active comparator were assessed using random effect models.

Poisson regression analysis was used to test the effect of indication, age groups, dose groups, route of administration and duration of treatment on incidence of AEs and hepatotoxicity in the fluconazole group. The incidence–rate ratios (IRR) are reported from the Poisson regression analysis. All results are reported with 95 % confidence intervals (CIs), and all P values are two-tailed.

Fluconazole was either administered as prophylaxis or therapeutically. The median prophylactic dose was 3 mg/kg/day [interquartile range (IQR) 3–6 mg/kg/day] over a median period of 42 days (IQR 1.57–42 days). The median administered therapeutic dose was 6 mg/kg/day (IQR 5–6 mg/kg/day) over a median duration of 42 days (IQR 14–67 days). Therapeutic indications were invasive candidiasis, taenia capitis, fungal meningitis, urinary tract infection and other mycotic infections. The duration of treatment ranged between 1 day [⁴⁰–⁴²] and 9 years [⁶⁸]. The most common routes of administration were oral (30 %), IV (23 %) or both (28 %) (Table 1).

A total of 4,209 patients from 90 studies were exposed to fluconazole, with 794 AEs recorded in 35 studies. Hepatotoxicity was the most common AE across all age groups. About one-third of the reviewed articles exclusively involved preterm and term neonates, accounting for 2,434 fluconazole exposed neonates. A total of 307 AEs (38.6 %) were recorded in neonates, of which 295 (96.1 %) were hepatotoxic effects. Gastrointestinal events were the second most common AE documented. One hundred cases of respiratory symptoms were recorded in one study, none of which was found to be drug-related. Other adverse events identified were renal dysfunction, haematological abnormalities and rash (Table 2). The relative risk of all AEs in the fluconazole group was not statistically different from those treated with placebo (RR 1.30, 95 % CI 0.84–2.03, P = 0.238). Compared to all other antifungal drugs there was again no significant increase in the risk (RR 1.05, 95 % CI 0.62–1.80, P = 0.85) (Fig. 2a). The overall relative risk of adverse events in the fluconazole group was not significantly different within the treatment group (RR 0.82, 95 % CI 0.49–1.36, P = 0.437) or the prophylaxis group (RR 1.68, 95 % CI 0.55–5.11, P = 0.364) compared to other antifungal drugs. There were 378 recorded cases of hepatotoxicity, accounting for just under half of all the AEs across all age groups. The majority of cases (295) occurred in neonates. The relative risk of hepatotoxicity with fluconazole was 1.36 (95 % CI 0.87–2.14) and 1.43 (95 % CI 0.67–3.03) when compared with placebo and other antifungals, respectively (Fig. 2a and 2b); these relationships were not statistically significant (P = 0.175 and 0.352, respectively). However, when compared against nystatin, the only comparator to have sufficient numbers of patients for analysis, there was a significant increase in risk of hepatotoxicity with fluconazole (RR 1.92, 95 % CI 1.13–3.26, P = 0.016) (Fig. 2c).

Poisson regression analysis of the effect of treatment, age group, dose, route of administration, indication and duration of treatment on the incidence of hepatotoxicity of fluconazole was performed. This showed that the incidence of hepatotoxicity with therapeutic fluconazole was significantly greater than that with prophylaxis (IRR 5.34 95 % CI 1.99–14.37, P = 0.001), while the duration of treatment had no effect. Although the incidence of hepatotoxicity in neonates on fluconazole was greater than that in children (IRR 1.33, 95 % CI 0.63–2.80), this effect was not statistically significant (P = 0.451). The incidence of hepatotoxicity appears to decrease with increasing dose (IRR 0.52, 95 % CI 0.37–0.74, P = 0.001). Patients on oral fluconazole were less likely to have hepatotoxicity than those on IV fluconazole (IRR 0.21, 95 % CI 0.92–0.47, P = 0.001).

Only three of the cohort studies reported any AE; one of which was a prophylactic study which recorded 127 cases of cholestasis in 409 fluconazole-exposed extremely low birth weight neonates. However, this study did not report the number of non-exposed neonates. Of these patients, 69 % recovered, while the others were discharged or transferred to other facilities [³⁶]. Another prophylactic cohort study recorded 60 cases of cholestasis in 140 fluconazole-exposed extremely low birth weight neonates as against 12 in 137 non-exposed neonates (P < 0.001) [³⁴].

Of all the 378 hepatotoxicity cases, resolution of symptoms was not determined in 113 (30 %) cases, while in 42 (11 %) cases, all involving neonates, there was no improvement at discharge or upon referral to another hospital (188 neonates and 35 children had completely recovered during treatment or shortly after). Therefore, 84 % of patients, for whom follow-up was complete, had resolution of symptoms. Hepatotoxicity was the most frequent reason for withdrawal of the drug. Of the 41 drug-related withdrawals,17 (42 %) were due to elevated liver enzymes [²⁶, ²⁷, ³⁰, ³⁶, ⁴², ⁶⁶, ⁶⁹].

Gastrointestinal (GI) symptoms including nausea, vomiting, abdominal pain, diarrhoea, dyspepsia, anorexia and gastritis accounted for 15.4 % of AEs (122 cases) (Table 2). There was only one recorded case of GI event in neonates. There was no statistical difference in the risk of GI events of fluconazole compared with placebo (RR 0.81, 95 %CI 0.12–5.60, P = 0.831). The risk of GI events increased, but not significantly, when fluconazole was compared with other comparator antifungal drugs (RR 1.23, 95 % CI 0.88–1.71, P = 0.235) and nystatin (RR 2.02, 95 % CI 0.66–6.23, P = 0.219). Poisson regression analysis showed that the incidence of GI AEs were lower in neonates than children (IRR 0.15, 95 % CI 0.03–0.66, p = 0.012), while dose (IRR 1.07, 95 % CI 0.85–1.39, P = 0.585) and duration of treatment (IRR 1.01, 95 % CI 0.99–1.03, P = 0.07) were unlikely to significantly affect the incidence of GI events. Although oral administration increased the incidence of GI AEs, this increase was not significant (IRR 3.22, 95 % CI 0.72–14.33, P = 0.125).

There was a decrease in mortality when fluconazole was compared with placebo, but this was not significant (RR 0.62, 95 % CI, 0.38–1.03, P = 0.067). The mortality rate between the fluconazole group and antifungal drugs was not different (RR 1.01, 95 % CI 0.72–1.41, P = 0.960) (Table 3). No cases of drug-related death were documented. There were ten reported cases of serious AEs, five of which were not treatment-related [³⁰]. The other serious AEs were five drug interactions [⁷⁰–⁷⁴]. Two interactions in children were with all-trans retinoic acid (ALTRA) and resulted in acute renal failure and pseudotumour cerebri [⁷⁰, ⁷¹]. Another case of acute renal failure in a 9-year-old child was recorded following interaction with tacrolimus [⁷³]. A 12-year-old child had syncope following co-administration with amitriptyline [⁷²]. Co-administration of fluconazole with vincristine also caused severe constipation [⁷⁴].

This research was conducted within the framework of a TINN project, a FP7 project sponsored by the European Commission.

RCT, Randomised controlled trial

^aIncluding studies involving infants up to adolescence (some of which included some neonates) and paediatric studies for which the age group was not stated

GIT, Gastrointestinal tract

^aNumber of adverse events (AEs) cut across age categories

^bObtained from a single study

^cPatients with anorexia, gastritis, dyspepsia, GI upset or a combination of any of nausea, vomiting, diarrhoea and abdominal pain

* (<0.05) statistically significant