The Association of Third-Generation Cephalosporin Use and Invasive Candidiasis in Extremely Low Birth-Weight Infants

The Cohort

Data were collected prospectively from ELBW neonates born weighing 401 to 1000 g from September 1, 1998, to December 31, 2001 at Neonatal Research Network sites. Trained research personnel at participating centers collected sociodemographic, pregnancy, and delivery data soon after birth on all live-born ELBW infants. Maternal and infant data were collected using common definitions developed by the investigators and described in the study Manual of Operations and in previous publications.¹¹ Infant data were collected prospectively from birth until 120 postnatal days, discharge, or death. The registry includes data on late-onset sepsis, infecting organisms, and antibiotic therapy. During the study period, infection surveillance was expanded to include detailed data on maternal intrapartum antibiotic use, results of all blood and cerebrospinal fluid cultures, postnatal age at infection, antibiotic use, presence of indwelling catheters, and other risk factors for infection. Blood cultures were processed by the clinical microbiology laboratories at each center using either the Bactec (Becton Dickinson) or BacT/Alert (Organon Teknika) systems. Decisions to obtain cultures through catheters or peripherally were at the discretion of the bedside clinician. The usual practice among participating centers is to inoculate blood culture bottles with ≥0.5 mL of blood.

Neonates eligible for inclusion in this study of late-onset infection were ≤1000-g birth weight and survived beyond postnatal day 3. The institutional review boards at each center approved participation in the registry.

Definitions

Invasive candidiasis was defined as a positive culture from blood or cerebrospinal fluid. It is acknowledged that this definition underestimates the total burden of invasive candidiasis, because cultures from other normally sterile body fluids are not included. We defined BSAs as both third-generation cephalosporins and carbapenems, consistent with previous work.² Additional analyses were done with cephalosporins and carbapenems considered separately. We defined empirical BSA use as antibiotics prescribed with no positive culture at the time of antibiotic initiation.

Analysis

We used multivariable logistic regression to assess BSA contribution to individual infant risk by using each infant as a unit of measure. We then assessed demographic and treatment variables, including initial antibiotic course lasting >5 days, and any BSA exposure initiated after the third postnatal day. A complete listing of all of the variables tested in the model is included in Table 2 and its legend (noncontributing variables).

We evaluated the centers' frequencies of antibiotic courses started on postnatal day 1 that lasted for >5 days, days treated with BSAs per infant, and days treated with BSAs when there was no positive culture during the treatment period per infant and correlated these variables with the incidence of candidiasis at the centers. We also assessed correlations between candidiasis and third-generation cephalosporins alone and carbapenems alone. Center demographic characteristics assessed for correlations with candidiasis included average gestational age and birth weight of day-3 survivors. We also assessed the correlation between the mortality of the first 3 postnatal days and candidiasis and correlations between mortality among infants surviving beyond postnatal day 3 and the antibiotic variables at the centers.

We evaluated potential confounding variables that could possibly contribute to the risk of candidiasis. We assessed correlations between days with central lines (percutaneous, umbilical venous or arterial, Broviac, or other) per ELBW infant with the center incidence of candidiasis. Because incidence of candidiasis is higher among ELBW infants with birth weight <750 g, we assessed correlations between center incidence of candidiasis and center proportion of ELBW infants alive on postnatal day 3 whose birthweight was 401 to 750 g. Because candidiasis risk may be increased with gastrointestinal pathology, we used the incidence of necrotizing enterocolitis (NEC) as a marker of gastrointestinal pathology, and measured the correlation between the incidence of NEC at the centers and the incidence of candidiasis.

Individual centers may have had a rationale based on the incidence of Gram-negative infection or mortality to preferentially use BSAs. We assessed correlations between antibiotic variables and the incidence of Gram-negative infection. In addition, because of high mortality for infants with Gram-negative infection and the possibility that not all Gram-negative infections result in a positive culture, we assessed whether center BSAs use was correlated with center mortality.

We calculated Pearson's product moment correlation for variables with normally distributed incidence and Spearman's rank correlation coefficients for those with nonnormally distributed incidence. All of the data were analyzed using SAS version 8.02 (SAS Institute, Inc, Cary, NC) at RTI International.

Individual Risk Factors

Risk factors for invasive candidiasis in individual infants were assessed. Results from the logistic regression are depicted in Table 2. Treatment with BSAs was associated with an increased risk of invasive candidiasis. Risk-adjusted center odds of candidiasis varied widely, with 6 of 11 centers having higher odds of candidiasis than the reference center. The reference center had the lowest incidence of candidiasis and the largest subject population. As expected, lower gestational age increased candidiasis odds. Initial antibiotic course >5 days was not significant at the individual infant level. Maternal hypertension was protective, whereas rupture of membranes >24 hours was associated with increased candidiasis odds. In a separate multivariable analysis with a conditional logistic regression model that included an individual center, using blood culture as the unit of measure, exposure to BSAs in the 7 days before culture was significantly associated with candidiasis (hazard ratio: 1.675; P = .0012).

Correlations: Center Demographics and Candidiasis

Demographics varied substantially among centers, but centers average gestational age and birth weight were weakly correlated with incidence of candidiasis. Center mortality before or after postnatal day 3 did not correlate with incidence of candidiasis (Table 3).

Correlations: Center Antibiotics and Candidiasis

Initial antibiotic courses starting on postnatal day 1 and lasting >5 days ranged from 26.9% to 69.3% (Table 3). The use of initial antibiotic courses >5 days, overall use, and use with negative cultures were weakly correlated with incidence of candidiasis after postnatal day 3. Center BSAs use after postnatal day 3 varied widely and was significantly correlated with the center incidence of candidiasis. With cephalosporins alone, the correlation between average cephalosporin days and candidiasis was 0.67 (P = .017). The correlation coefficient between candidiasis and the average days of cephalosporin use with negative cultures was 0.68 (P = .015). Carbapenem use was rare in this cohort and ranged from 0 days to an average of 1.15 days per infant. Correlation coefficients were not significantly different without carbapenem data. Carbapenem use alone was not significantly correlated with incidence of candidiasis.

Correlations: Central Lines and Other Possible Confounders

Central line days per infant in each center (range, median days: 13–27.5 days) were also not correlated with candidiasis (correlation coefficient [CC]: 0.21; P = .51), although the center with the highest median line days was also the center with highest incidence of candidiasis (Table 3). The proportion of postnatal day 3 ELBW survivors whose birthweight fell between 401 and 750 g (range: 32.14%–56.29%) at a center was not correlated with center incidence of candidiasis (CC: 0.34; P = .28). Incidence of NEC (range: 6.21%–16.06%) was not correlated with candidiasis (CC: −0.15; P = .64)

Correlations: BSA Use With Gram-Negative Infections

To assess whether centers had a rationale for the prevalent use of BSAs based on a high incidence of Gram-negative infections, we assessed correlations between BSAs use and the incidence of Gram-negative infections (Table 3). The center incidences of Gram-negative bacterial infections (range: 5.9%–16.8%) were correlated with neither the incidence of candidiasis at the center (CC: −0.07; P = .84), nor the use of BSAs by the center (CC: −0.07; P = .83).

Correlations: BSA Use and Mortality

Correlations between center mortality among infants surviving beyond postnatal day 3 (range: 9%–26.8%) and average BSAs days per infant and average BSAs days with negative cultures per infant were weakly positive (CCs: 0.505 and 0.511, respectively) but not significant (both P = .09; Table 3).

Limitations

There are several limitations to this study. We have limited information on the impact of carbapenems and other broad-spectrum antibiotics, such as piperacillin-tazobactam, because these agents were used infrequently in this cohort. We did not analyze other center practices that could possibly be risk factors for candidiasis, including hand washing, unit staffing practices, or colonization rates among staff or infants. Postnatal steroid use was not included in the analysis. Timing of steroid use relative to timing of infection was not collected. This limitation is tempered by the fact that previous multivariable regression analyses have not shown an association between steroid use and Candida species colonization or invasive candidiasis.^2,5 Data were unavailable on other possible confounders, including days of hyperalimentation, timing of use of H2 blockers before and after candidiasis, and use of intralipids. No center reported an outbreak of candidiasis or reported using fluconazole prophylaxis during the study period.