Invasive fungal infection, most commonly due to Candida species, is increasingly common in preterm babies in neonatal intensive care.1^–6 The estimated incidence in very low birthweight (VLBW) babies is between 2% and 4%, but it may be as high as 10% in extremely low birthweight (ELBW) babies. Fungal sepsis has much higher mortality, 21–32%, than bacteraemia, and it is also associated with markedly higher rates of adverse neurodevelopmental outcomes.6^–10

Fluconazole prophylaxis reduces the incidence and mortality from invasive fungal infection,11^–19 but widespread use of antifungals might increase the emergence of resistance.3 10 20^–23 Studies to date are too short to fully assess the potential for fluconazole resistance and there remain reservations about treating all babies to protect a few.12 Fluconazole seems to be well tolerated in prophylactic doses,11^–19 but there are associations with rise in liver transaminases, cholestasis, toxic epidermal necrolysis and Steven–Johnson syndrome, and interactions with other medications.2 13 24^–26 Selective use of antifungal prophylaxis in the subset of VLBW babies at very high risk of fungaemia may be preferable to minimise the risk of adverse effects.2 12 Although comparatively little data are available on selective antifungal prophylaxis, two recent studies reported reductions in invasive fungal infection with a fluconazole prophylaxis policy targeting VLBW babies with additional risk factors.16 17 Recognised additional risk factors for acquiring fungal sepsis in preterm infants include third generation cephalosporin use, fungal colonisation, prolonged broad spectrum antibiotic use, parenteral nutrition with lipids, endotracheal intubation, male gender, central venous catheter use and number of days in situ, previous bacterial blood stream infections, postnatal steroids, gastrointestinal pathology, H₂-receptor antagonists, shock and coagulopathy.2 25 27^–30

In October 2003, the neonatal intensive care unit (NICU) at the Royal Maternity Hospital, Belfast, developed a guideline for antifungal prophylaxis in VLBW babies with additional risk factors for fungal sepsis. We conducted this retrospective study to determine if the use of selective fluconazole prophylaxis was effective in reducing invasive fungal infection in high-risk babies. We also wanted to establish if there has been an increase in fluconazole resistance.

METHODS

This retrospective study was conducted at a regional maternity unit. We included all babies <1500 g birth weight admitted to the Royal Maternity Neonatal Unit in Belfast, 1 year before and after 1 October 2003 when the guideline was implemented. Babies were excluded if they were admitted for short periods only (<48 h).

The antifungal prophylaxis guideline states that a VLBW baby should be considered for antifungal prophylaxis if they have any of the following additional risk factors:

• under treatment with a third generation cephalosporin;

• under treatment for more than 10 consecutive days with a systemic broad spectrum antibiotic;

• fungal colonisation from surface sites and a central venous catheter in situ.

Eligible babies were given antifungal prophylaxis at the discretion of the attending neonatologist. The fluconazole prophylaxis was given at any time during the admission and there was no specific guidance regarding prophylaxis early in the infant’s course. The prophylaxis regimen was fluconazole 6 mg/kg for 3 weeks. Dose interval was once every 72 h for the first 2 weeks of life. In babies aged between 2 weeks and 3 weeks, the dose interval was every 48 h. Babies more than 3 weeks of age were given daily fluconazole. Fluconazole was administered orally when enteral feeds were established. The guideline did not stipulate monitoring for adverse effects. This was left to the discretion of the attending neonatologist.

Data were collected on birth weight, estimated gestational age, identification of risk factors for fungal sepsis (in accordance with the NICU guideline), the number receiving prophylaxis, adherence to the dosing regimen and the incidence of fungal sepsis. The presence of other risk factors, not included in the guideline but identified in previous studies,2 25 27^–30 were also collected on eligible babies. The source, species and sensitivity patterns of Candida cultured was documented. Data were manually collected from the available medical records and the Belfast Link Laboratories microbiology database.

Culture proven Candida sepsis was defined as presence of a blood culture positive for Candida species or a clinically septicaemic baby with a urine culture or central line tip culture positive for Candida species. A diagnosis of probable Candida sepsis was made if the infant was clinically septicaemic with no positive cultures but there was additional evidence of fungi such as fungal balls on ultrasound scan. Clinical septicaemia was diagnosed by the attending neonatologist.

Antibiotic susceptibility testing for Candida

Antibiotic susceptibility of fungal isolates was tested using Sensititre Yeastone (TREK diagnostic systems, West Sussex, UK). The sensitivity range for Candida species for fluconazole were: sensitive, minimum inhibitory concentration (MIC) ⩽8 μg/ml; intermediate, MIC 16–32 μg/ml; resistant ⩾64 μg/ml.

Statistical analysis

The main outcome of interest was a positive diagnosis of invasive Candida infection. We analysed the data using the Fisher exact test. Group characteristics were compared by using either the Student t test or the Fisher exact test where appropriate on Windows SPSS (version 15.0).

Research ethics

Ethical approval was not required as this was a retrospective chart and computer records study, in keeping with our trust’s research governance policy. The data collected did not affect patient care, was anonymous, and was securely collected and stored.

RESULTS

A total of 121 babies admitted to the NICU before the implementation of the guideline were eligible for inclusion in the study, and 110 charts were available for data collection. Five infants were excluded as they were admitted for less than 48 h before transfer for surgery. A total of 107 babies admitted to the NICU after the implementation of the guideline were eligible and 102 charts were available for data collection. None of these babies were excluded. Although a few records were missing, these did not include infants with Candida sepsis identified in the microbiology database. These babies had similar characteristics to the babies whose charts were available. None of the babies admitted before the implementation of the guideline received fluconazole prophylaxis. Among those admitted after the implementation of the guideline, 14 of the 31 eligible babies received prophylaxis. Eight of these 14 babies received the correct dose, interval and duration of treatment. In the remaining six there were minor errors in dosage (3/14) and duration of treatment (5/14).

Incidence and mortality from invasive Candida infection

Following the introduction of the guideline, there was a significant reduction in culture proven invasive Candida infection in babies eligible for fluconazole prophylaxis (6% before and 0% after introduction of the guideline, p = 0.03). One baby admitted after the implementation of the guideline developed probable Candida sepsis. The diagnosis was made based on clinical septicaemia, ultrasound images suggestive of fungal balls in the renal tract and the steady improvement of the clinical course after administration of liposomal amphotericin B. No fungi were cultured from the urine or blood but fluconazole prophylaxis had been given. The reduction in invasive Candida infection (culture proven and probable) was not statistically significant. Table 1 shows the total number of babies who died and the mortality from Candida sepsis and other causes in the periods before and after the implementation of the guideline. Other causes of death included necrotising enterocolitis, bronchopulmonary dysplasia and intraventricular haemorrhage. A statistical analysis was undertaken to determine the effect of deaths from Candida sepsis on the mortality in the two groups (p = 0.2). The source and species of positive invasive Candida cultures is shown in table 2. All six babies with invasive Candida infection were ELBW and would have been eligible for prophylaxis if it have been available at that time.

We found significant differences between the infants who received fluconazole prophylaxis and those who were eligible but did not receive prophylaxis: infants receiving prophylaxis had a lower gestational age, two or more risk factors and more had received a third generation cephalosporin (see p values in bold in table 3). There was also a tendency for ELBW babies to receive prophylaxis (table 3).

DISCUSSION

Here we have presented data on antifungal prophylaxis given to a subset of VLBW babies with recognised additional risk factors for fungal sepsis. The results for the entire group of babies eligible for fluconazole prophylaxis are shown, rather than the babies who were selected for treatment, as there were no suitable controls for the selected group. We found a significant reduction in culture proven Candida sepsis in a group of babies at particularly high risk, following the introduction of antifungal prophylaxis to a NICU. The retrospective nature of this study, relatively small numbers of eligible babies and low incidence of invasive Candida infection make this study underpowered to draw firm conclusions about the impact of selective fluconazole prophylaxis. Other limitations were the lack of controls and routine monitoring for adverse effects. However, the reduction in total cases of Candida sepsis (culture proven and probable) from 5% to 1% may be important, given the high morbidity and mortality associated with this condition.

A Cochrane meta-analysis concluded that fluconazole prophylaxis for all VLBW or ELBW babies considerably reduces the incidence of fungal sepsis.11 12 More recent studies have included additional risk factors. Bertini et al observed a marked reduction in invasive candidiasis (7.6% to 0%) when fluconazole prophylaxis was given to VLBW babies with central venous catheters in situ (38% of all VLBW babies in the study period).17 Uko et al reported similar results when a shorter course of lower-dose fluconazole prophylaxis (3 mg/kg) was given to VLBW babies receiving systemic antibiotics for more than 3 days (51% of all VLBW babies in the study period).18 All of the studies quoted report similar reductions in the incidence of Candida sepsis; however, none adopted a risk-based policy as selective as this guideline (14% of all VLBW babies in the study period).

To achieve greater selection of babies at highest risk of fungal sepsis our guideline only included what are considered the most important risk factors: cephalosporin treatment, fungal colonisation, prolonged antibiotic use and the presence of a central venous catheter.11 25 29 30 H₂-receptor antagonists and postnatal dexamethasone are rarely used in our unit and were not included in the guideline. The guideline was not designed to be prescriptive and a large proportion of eligible babies in our NICU were considered not to require antifungal prophylaxis by the attending neonatologist. For example, two VLBW infants were started on a cephalosporin for suspected necrotising enterocolitis, but this was discontinued after 48 h once the clinical picture made this improbable. Eleven babies were eligible because of prolonged antibiotic use as their only risk factor. None of these babies received prophylaxis. It is difficult to comment on the decisions of individuals in a retrospective study. However, in practice, fluconazole prophylaxis was administered to infants with lower birth weights, with two or more risk factors and, in particular, those receiving a third generation cephalosporin.

It is possible that other factors may have influenced the decrease in invasive fungal infection. Although there was no change in infection control policy in the NICU, an audit of blood stream infections in this unit during the same period showed a modest 26% reduction in laboratory-confirmed blood stream infections but no reduction in those not confirmed by the laboratory. In addition, prevention of horizontal transmission of Candida by hand infection control measures has had only limited success in decreasing Candida colonisation and sepsis in neonates.31 32

Fluconazole prophylaxis gives rise to concerns that species which are inherently resistant to fluconazole will be selected out, or that there will be a gradual increase in resistance among previously sensitive strains. A strain of C parapsilosis, less susceptible to fluconazole, has been identified as causing blood stream infections in a NICU where fluconazole prophylaxis has been practised for 10 years.33 Azole resistance in Candida albicans has also been reported, and Candida resistance has been found in preterm animals.34 35 However, similar to most previous studies,11^–17 we found no evidence of the emergence of fluconazole-resistant strains, and all isolated Candida remained fully susceptible to fluconazole. As the study period was short it will be necessary to continue surveillance with these issues in mind.

In conclusion, the incidence of Candida sepsis in the NICU noticeably decreased in the 12 months following the introduction of an antifungal prophylaxis guideline, using fluconazole, targeted at a subset of VLBW babies at higher risk of invasive fungal infection. This study highlights the practicalities of implementing a fluconazole prophylaxis policy that attempts to appropriately identify those infants at greatest risk, while not exposing infants at lesser risk to a potentially harmful medication. Our results suggest that this risk-based approach correctly identifies those infants most likely to benefit from fluconazole prophylaxis. Achieving the correct balance has implications for cost effectiveness, emergence of resistant species and reducing exposure to potentially harmful medications. We are unable to prove any causal relationship due to the lack of power and retrospective nature of this study. However, a multicentre randomised controlled trial would elucidate the importance of particular risk factors in the development of fungal sepsis and the role of antifungal prophylaxis.