Methods

ASGNI collected data prospectively in a multi-centre surveillance study of babies in neonatal units in Australia and New Zealand.18 Data collected routinely include birth weight and gestational age, postnatal age at the time of onset of sepsis, blood and/or cerebrospinal fluid (CSF) culture results, underlying diagnoses and outcome of infection (survival or death). Invasive neonatal fungemia was defined as a baby with: (1) clinical signs of sepsis, for example, apnoea, bradycardia, fever, hypothermia, tachycardia, tachypnea; and (2) a positive culture of a fungus from blood and/or CSF.18 Fungal meningitis is defined as a clinical presentation consistent with meningitis and either a positive culture of fungus from CSF or a positive fungal blood culture in association with a raised CSF white blood count (which we defined as >109 white cells).18

When possible, data were collected on the breakdown of admissions to neonatal units by birth weight categories, to calculate incidence of infection.

Deaths were reported by clinicians as: death definitely due to the infection; infection probably contributed to death; or death unrelated to infection. In this paper, we report mortality as babies whose death was definitely or probably due to the invasive fungal infection. In practice, however, all babies in the study who died were reported as “death definitely due to fungal infection” (see Results section).

We used the computer software package SPSS.V.15.0 for statistical analysis. Categorical variables are expressed as proportions (percentages) and numerical data are expressed as means (standard deviations), medians and ranges. We used Pearson’s χ² test or Fisher’s exact test (two tailed) to compare categorical variables and Student’s t test to compare means.

We obtained prospective approval for the study from the Royal Alexandra Hospital for Children, Sydney Institutional Ethics Committee (approval number 93060). Informed consent was not deemed necessary by the Ethics Committee because the data are recorded and analysed without identifying data.

Results

The number of VLBW babies (<1500 g) was not available for every unit reporting to ASGNI for every year, so we used only data from those hospitals which supplied birth weight stratifications. The total number of participating hospitals was 15 over the 14 years and the median number of hospitals which provided data each year was 7.5 (range 4–11). Approximately 2250 babies with birth weight <1500 g are born each year in Australia19 and approximately 500 in New Zealand,20 so our study population represents about 38% of all the babies born in Australia and New Zealand during this time period. The hospitals included at least one main tertiary neonatal centre in each of five States in Australia and two main tertiary centres in New Zealand.

We were able to obtain data on both antifungal prophylaxis and birth weight for 10 of the 15 units, representing 85.2% (12 607/14 788) of all babies <1500 g and 84.4% of all babies <1000 g (5018/5498). This included 102 (86.4%) of 118 babies <1500 g and 91 (85.8%) of 106 babies <1000 g with fungal infection (see table 1). The only antifungal used for prophylaxis was oral nystatin. Hospital 1 changed from no prophylaxis to oral nystatin for babies <27 week gestation or <750 g birth weight, hospital 2 changed from selective nystatin prophylaxis for babies with bronchopulmonary dysplasia to no prophylaxis and hospital 3 stopped using oral nystatin prophylaxis briefly, but then re-started it (see table 1). Hospitals 1, 2 and 3 are included twice in table 1, for the periods when they did or did not use prophylaxis. Two hospitals always used selective oral nystatin (hospital 4 for babies on broad spectrum antibiotics and hospital 5 for those receiving parenteral nutrition or with a central catheter). Five hospitals (hospitals 6–10) used no prophylaxis. There was no statistical difference in birth weight (p = 0.3), gestational age (p = 0.1) or age at diagnosis (p = 0.3) between units which did or did not use prophylaxis.

We identified 14 778 babies of birth weight <1500 g from 1 January 1993 through 31 December 2006. During this period there were 118 episodes of invasive fungal infection (IFI), giving an incidence for babies <1500 g of 0.80% (95% CI 0.66 to 0.94%). The incidence of IFI was 0.54% (95% CI 0.4 to 0.7%) for babies <1500 g in units using nystatin prophylaxis and 1.23% (95% CI 0.8 to 1.6%) for babies <1500 g in units using no antifungal prophylaxis (p<0.001), see table 1.

Of the 118 babies with invasive fungal infection, 106 (89.8%) had birth weight <1000 g, see fig 1. The overall incidence of fungal infection in babies with birth weight <1000 g was 1.78% (106/5948) (95% CI 1.44 to 2.12%). The incidence of IFI was 1.23% (95% CI 0.92 to 1.54%) for babies <1000 g in units using prophylaxis and 2.67% (95% CI 1.87 to 3.37%) for babies <1000 g in units using no prophylaxis (p<0.001).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Invasive fungal infection by birth weight.

For the 3 hospitals that changed antifungal policy during the study period (hospitals 1, 2 and 3 in table 1), 32 of 4659 babies <1500 g who received nystatin prophylaxis or 0.69% (95% CI 0.45 to 0.93%) developed fungal infections compared with 16 of 1417 babies <1500 g or 1.13% (95% CI 0.58 to 1.68%) who did not receive nystatin prophylaxis (χ² = 2.2, p>0.05). For babies <1000 g, the rate of fungal infections was 1.23% (95% CI 0.78% to 1.68%) if they received nystatin prophylaxis and 3.25% (95% CI 1.68 to 4.82%) if they did not receive antifungal prophylaxis (χ² = 9.2, p<0.005).

There was significant variation between individual hospitals in the incidence of fungal infection in babies <1500 g (range 0–2.37%, p<0.005). The median gestational age at birth of infants with invasive fungal infection was 25 weeks (range 22–31 weeks), see fig 2. The median birth weight was 720 g (range 344–1345 g). The median age of first positive culture was 15 days (range 1–90 days), see fig 3. Fifty-four per cent of infections occurred within 14 days of birth, 28% occurring between days 6 and 10.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Number of invasive fungal infections by gestational age.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Age at culture diagnosis.

Regarding concurrent infections, three babies (2.5%) had renal fungal infections, eight babies (6.8%) had necrotising enterocolitis, three (2.5%) had peritonitis and soft tissue abscesses and one baby had fungal pneumonia.

We found a non-significant trend to a lower incidence of fungal infections over time, which may reflect increasing use of oral nystatin prophylaxis over the study period.

All fungal infections were due to Candida species. Seventy-four infections (62.7%) were due to C. albicans and 39 (33.1%) due to C. parapsilosis. Three babies had C. glabrata infection and one each had C. guillermondi and C. famata. There was no significant difference in gestational age, birth weight or age at culture diagnosis between babies with C. albicans and C. parapsilosis infections.

The mortality attributed to fungal sepsis was 16.5% (37 of 224 babies). All 37 babies were reported as “death definitely due to fungal infection”. The mortality for babies <1500 g was 16.1% (19 of 118) and for babies <1000 g was 17.0% (18 of 106), p>0.05. The mortality was higher at 38.5% (5 of 13) for babies whose first culture was positive before seven days of age, than the mortality of 15.5% (13 of 84) for babies with positive cultures from seven to 28 days and 4.8% (1 of 21) for babies with first positive culture after 28 days (p<0.05).

In all, 82 of 118 infected babies (69.5%) had lumbar punctures (LP). Sixteen babies grew a Candida species from the CSF and a further four babies had positive blood cultures with an abnormally high CSF white blood cell count, giving a total of 20 babies with meningitis (16.9% of the 118 infected babies and 24.4% of the 82 who had an LP). Four (27%) of the 15 babies with meningitis had a normal CSF white cell count. Two babies with fungal meningitis had a negative blood culture. Nine of 20 babies (45%) with meningitis died from their infection.

Discussion

Very low birth weight infants are at greatly increased risk of invasive fungal infection compared to more mature infants. The incidence of fungal infection, however, appears to be higher in the USA and Italy than in the UK, Australia and New Zealand. The reported incidence was 10% of babies <1500 g in the USA and Italy9 10 and 5.1–26% of US babies <1000 g8 11 but only 2.1% of babies <1500 g and 1% of babies <1000 g in the UK.12 The British Paediatric Surveillance Unit study depends on passive reporting and may have under-estimated the true rate of fungal infection, whereas the US and Italian studies used active surveillance. On the other hand, our study used active surveillance from neonatologists and we report a low incidence of fungal infection, even when antifungal prophylaxis is not used. However, we also found that selective or universal prophylaxis using oral nystatin was associated with a rate of invasive fungal infection less than half than the rate without prophylaxis for babies <1500 g and babies <1000 g.

The difference in reported incidence between Australia/New Zealand and USA/Italy may in part be due to the definition of invasive fungal infection used. In our study, we only included positive blood and/or CSF cultures, because of the difficulty distinguishing positive urine cultures caused by infection from positive cultures due to contamination or colonisation of the skin or genitalia. A 12-centre US study that included only babies with blood or CSF cultures positive for Candida reported that 7% of babies <1500 g developed systemic candidiasis.21 An eight-centre randomised controlled trial of fluconazole prophylaxis in Italy found the incidence of invasive fungal infection was 13.2% in babies <1500 g in the placebo group, but these included urine infections.22 If babies with urine infections were excluded, the incidence of invasive fungal infection was 9.4%. The largest US study included only Candida bloodstream infections.11 We have already discussed the limitations of the British Paediatric Surveillance Unit study,12 but it is interesting that they report the incidence of invasive fungal infection as 1.0% in babies <1500 g and 2.1% in babies <1000 g, very similar to our reported incidence. The UK study included positive cultures from urine, bone, joint, peritoneal and pleura as well as culture positive long line tips. About a third of all UK neonatal units use systemic, oral or topical antifungal prophylaxis.13

We did not specifically request that clinicians report babies found to have unexpected fungal infection at autopsy. However, we are told by our clinicians that they no longer see babies dying from undiagnosed fungal infection, and the comparison studies did not include such babies either, so we do not believe our results under-estimate the true rate of fungal infection significantly.

We did find significant variation in incidence between hospitals, although the highest rates of 2.2% for babies <1500 g and 5.4% <1000 g were still lower than rates reported from the USA and Italy. Although this intra-hospital variation in incidence is one potential weakness of our study, we did not find any significant variation in caseload and we did not find that any one hospital contributed excessively to the numbers of cases, which would potentially have biased our conclusions.

The disparity between the relatively high incidence in the USA and Italy on the one hand, and the low incidence in the UK, Australia and New Zealand on the other hand, could be due to other factors known to affect the incidence of fungal infection.23 Neither our study, nor any of the others reported here, provided data on factors such as use of broad spectrum antibiotics, duration of antibiotics, duration of parenteral nutrition, duration of endotracheal intubation or duration of central venous catheterisation. It is possible that fungal infection is much less common in the UK, Australia and New Zealand because broad spectrum antibiotics are used less often than in the USA and Italy and for shorter periods, that central venous catheters are removed earlier and babies are extubated earlier, but we did not collect data to confirm or refute this hypothesis. Future studies could collect data to attempt to answer whether or not these factors are relevant.

We found an association between oral nystatin antifungal prophylaxis and reduced incidence of fungal infection. Our study was not designed to test the efficacy of antifungal prophylaxis and our data could be subject to bias. For example, hospitals might introduce antifungal prophylaxis and other measures, such as improved infection control, at a time of high incidence of fungal infections. A fall in incidence could be due to chance or to improved infection control. On the other hand, our findings are consistent with the only randomised controlled trial on the use of oral nystatin prophylaxis, which showed that prophylactic nystatin significantly reduced systemic fungal infections.24 A recent sequential study also found a fall in systemic fungal infection in babies <33 weeks gestation associated with the introduction of prophylactic oral nystatin.25 In addition, for the three hospitals that changed antifungal prophylaxis over the study period, we found a lower incidence of fungal infection in babies receiving oral nystatin, which was highly statistically significant for babies <1000 g.

There is strong evidence, summarised in a Cochrane systematic review,26 that prophylactic fluconazole reduces the incidence of neonatal systemic fungal infection in babies <1500 g. Some authors have reported that prophylactic fluconazole can be associated with hepatotoxicity,14 15 and that its use may potentially select for fluconazole-resistant fungi and non-albicans Candida species,16 although the hepatotoxicity is usually reversible27 and resistance has not been reported as a clinical problem to date.17 We were not able to test isolates for sensitivity to fluconazole. Oral nystatin is cheap, less toxic than fluconazole, and was shown to be highly effective at preventing Candida infection in one randomised controlled trial.24 Our study provides support that prophylaxis using the non-absorbable antifungal nystatin may reduce the incidence of systemic fungal infections in very low birth weight infants. Prescribers might feel that oral nystatin is an attractive alternative to fluconazole. We recommend that future studies of prophylactic fluconazole to prevent neonatal fungal infection should use oral nystatin, not placebo, as the comparator.23