Candidal meningitis may complicate systemic candidiasis in the premature neonate. We conducted a 10-year retrospective review of 106 cases of systemic candidiasis in neonates to define the incidence, clinical features, laboratory findings, treatment, and outcome of candidal meningitis. Twenty-three of the 106 neonates had candidal meningitis (0.4% of admissions to the neonatal intensive care unit). The median gestational age was 26.2 weeks, the median birth weight was 820 g, and the median age at the onset of illness was 8 days. Clinical disease was severe and commonly was manifested by respiratory decompensation. Findings of cerebrospinal fluid (CSF) analyses varied: pleocytosis was inconsistent, hypoglycorrhachia was common, gram staining was uniformly negative, and Candida was isolated from 17 neonates (74%). Each infant was treated with amphotericin B (median cumulative dose, 30 mg/kg); 5 also received flucytosine therapy. In conclusion, initial clinical features of candidal meningitis are indistinguishable from those of other causes of systemic infection in premature neonates, and normal CSF parameters do not exclude meningitis. Timely initiation of amphotericin B monotherapy was associated with an excellent outcome.
Although overall survival rates among very low birth weight (VLBW) premature infants have improved during the past decade, nosocomial infections continue to have an adverse effect on the outcome for these developmentally immuno-com-pro-mised hosts . Of these infections, disseminated or systemic candidiasis is a major cause of morbidity and mortality. A shift in the pathogenic species of Candida causing systemic neonatal disease recently has been reported, with more infections caused by non-albicans species [2, 3]. CNS involvement is a complication of disseminated candidiasis that may be associated with enhanced morbidity, and the number of cases of candidal meningitis in VLBW infants appears to be increasing [3, 4]. However, little is known about risk factors, clinical features, laboratory findings, treatment, and outcome of candidal meningitis in these infants.
We conducted a retrospective review of premature infants hospitalized from 1989 to 1999 in a neonatal intensive care unit (Texas Children's Hospital, Houston). The study population included those infants who had either Candida species isolated from CSF or abnormal CSF parameters and Candida isolated from a normally sterile site.
Patients and Methods
Neonates with systemic candidal infection were identified by review of records of the Microbiology Laboratory at Texas Children's Hospital for the period January 1989 through June 1999. The medical records of all infants who had Candida species isolated from a normally sterile site were reviewed. Cases of candidal meningitis were subsequently identified.
Definitions. Infants were considered to have either definite or probable candidal meningitis on the basis of clinical and laboratory criteria. A case was defined as definite if it met 1 of the following 4 sets of criteria: (1) pleocytosis (WBC count, >45 cells/mm3 ), Candida species isolated from CSF, and no intraventricular hemorrhage (IVH); (2) pleocytosis, isolation of Candida from CSF, IVH, and no candidal dermatitis; (3) pleocytosis, a sterile CSF culture, isolation of Candida from a normally sterile site, and no IVH; or (4) isolation of Candida from CSF with or without IVH and no pleocytosis, candidemia, or candidal dermatitis. Candidal dermatitis was defined as the presence of classical features (black or white crust and plaques) over the lumbar area with or without confirmation by a skin biopsy specimen that demonstrated fungal invasion beyond the stratum corneum epidermidis .
A case was defined as probable if there was either Candida isolated from CSF, no pleocytosis, and candidal dermatitis, or if there was Candida isolated from CSF, pleocytosis, candidal dermatitis, and IVH. The criteria that categorized cases as definite or probable are summarized in table 1.
Systemic (or disseminated) candidal infection was defined as isolation of Candida from blood, CSF, peritoneal or joint fluid, or skin biopsy specimen.
Exclusion criteria. Infants who had IVH, pleocytosis, and a sterile CSF culture were excluded from the study. Infants whose diagnosis of candidal meningitis was established after death were not included. However, these patients are briefly described and included in the incidence calculation.
Patient population. During the 10.5-year study period, there were 6171 neonates admitted to the neonatal intensive care unit (1980 infants had birth weights <1500 g). Of the 106 infants who had systemic candidiasis, 32 (30%) had Candida isolated from CSF, pleocytosis with Candida isolated from normally sterile site other than CSF, or postmortem evidence of CNS infection caused by Candida. Twenty-nine (91%) of these 32 infants had birth weights of <1500 g. Six infants were excluded (see exclusion criteria). Three infants were found at postmortem examination to have CNS involvement without prior documentation of candidemia; each had disseminated candidiasis. Of these 3 infants, 1 received a single dose of amphotericin B just before death because of a history of Candida albicans urinary tract infection, and the other 2 infants did not receive antifungal therapy. Thus, the incidence of candidal meningitis in the population of our neonatal intensive care unit was 0.4%, and among infants weighing <1500 g at birth, it was 1.1%.
The medical records of the 23 infants with an antemortem diagnosis of definite or probable candidal meningitis were reviewed. Data recorded included demographic information, clinical findings at the onset of systemic candidiasis, occurrence of predisposing factors for invasive candidiasis (hyperglycemia [blood sugar level, >150 mg/dL], insulin use, antibiotic use, total parenteral nutrition and intralipid use, methylxanthine or steroid use, and candidal colonization), duration of blood or CSF cultures yielding Candida species, and results of diagnostic studies to define disseminated infection. In addition, Candida species and clinical outcome were recorded.
Twenty-three infants met our inclusion criteria: 17 had definite and 6 had probable meningitis caused by Candida species. Each group of infants had similar clinical presentations, underlying medical conditions (except for the more frequent occurrence of candidal dermatitis in probable cases [P = .02; Fisher's exact test]), and therapy. Therefore, both groups were combined for further analysis.
Clinical features and predisposing factors. The clinical characteristics of the 23 neonates with candidal meningitis are summarized in table 2. Each neonate was born before 37 weeks of gestation. Gestations ranged from 23 to 34.9 weeks (median, 26.2 weeks), and birth weights ranged from 600 to 2135 g (median, 820 g). Age at the onset of illness ranged from 5 to 167 days (median, 8 days). Thirteen infants were male (57%). Of these premature neonates, 12 were white, 6 were African-American, and 5 were Hispanic.
Each neonate had umbilical arterial and venous catheters inserted shortly after birth, and the mean duration of their use was 7.7 days (range, 1–15 days) for arterial catheters and 7.2 days (range, 4–9 days) for venous catheters. Seven infants (30%) had a percutaneous central venous catheter present at the time of diagnosis; the catheter was removed from 5 of these infants when the diagnosis of invasive candidiasis was made. In the other 2 infants, the catheter was retained because blood cultures were sterile. Total parenteral nutrition and intralipid use each was documented in 21 cases (91%). Eleven infants (48%) had documented hyperglycemia, 7 (64%) of whom required insulin therapy. Each of the 23 infants received broad-spectrum antibiotic therapy before the onset of candidemia (mean duration, 8 days [range, 3–17 days]). Five infants (22%) had received steroid therapy primarily for weaning from ventilator support. Seventeen neonates (74%) were receiving or had received caffeine citrate therapy. Three infants (13%) had documented colonization of the respiratory tract with C. albicans before the onset of systemic candidiasis caused by this species.
Three infants (13%) had congenital abnormalities; each had definite candidal meningitis. One infant had a bicuspid aortic valve, 1 had a ventricular septal defect and hypospadias, and 1 had congenital ichthyosis. Ten infants (43%) had IVH of varying severity documented by cranial ultrasonography; in 7 of these infants, IVH was detected after the onset of candidal meningitis.
Clinical disease was severe in most infants and most commonly was heralded by respiratory decompensation. A substantial proportion (61%) of infants had an increase in their oxygen requirement or in the number of episodes of apnea and bradycardia. Four infants required reintubation and mechanical ventilation. Metabolic acidosis (pH ≤ 7.2) was recorded in 7 cases (30%), and a decrease in urinary output was recorded in 4 cases (19%). Hypotension necessitating the initiation of inotropic support was noted in 5 cases (22%).
One neonate (4%) had concomitant suppurative thrombophlebitis, and another 3 (13%) had candidal osteomyelitis. Necrotizing enterocolitis developed in 3 infants (13%); in 2 of these cases, culture of peritoneal fluid obtained at surgery yielded Candida. An additional neonate had an isolated bowel perforation, but peritoneal fluid cultures did not yield Candida. Each of the 6 infants with a diagnosis of probable meningitis also had clinical evidence of candidal dermatitis, and this condition was biopsy proven in 3 cases .
Laboratory and diagnostic imaging findings. Results of the initial CSF analyses for the 23 neonates are summarized in table 3. Nine infants (39%) with definite meningitis had pleocytosis (see Patients and Methods), and in 6 of these infants, mononuclear forms were the predominant WBC type. Twelve infants (52%), 7 of whom had definite and 5 of whom had probable meningitis, had a normal CSF WBC count (range for infants with definite meningitis, 0–40 cells/mm3 [mean, 5 cells/mm3]; range for infants with probable meningitis, 5–36 cells/mm3 [mean, 14 cells/mm3]) at the initial lumbar puncture. Eight infants with definite and 2 with probable meningitis had CSF RBC counts of >3000 cells/mm3. CSF specimens from 2 infants were evaluated only by culture because of gross contamination with blood. For 21 infants, the mean CSF glucose concentrations were 95 mg/dL, and protein concentrations were 214 mg/dL. Five of 21 infants had CSF glucose levels of <45 mg/dL, and each had definite meningitis. Gram staining of CSF was negative for yeast in all instances.
For the 17 infants who had Candida isolated from CSF, additional CSF cultures were performed to document sterility. The second CSF specimens were collected a mean of 4 days after the initial culture specimen, and CSF specimens from 13 (76%) of the 17 infants were sterile. Of the 4 infants whose CSF cultures remained positive, 2 had third CSF specimens that were sterile (cultures were performed 5 and 13 days later). For 1 infant, lumbar puncture was not repeated, and CSF from the other infant was sterile at the fifth lumbar puncture, which was done 1 week after the first procedure. CSF specimens were collected again from 5 infants whose CSF cultures were initially sterile; in each case, CSF remained sterile.
At least 1 blood specimen for culture was collected from each infant at the onset of illness. Eight infants had sterile blood cultures; 4 had probable candidal meningitis. Blood cultures remained positive for Candida for a mean of 5.2 days (range, 1–9 days). C. albicans was isolated from blood or CSF from 22 (96%) of the 23 infants. Candida glabrata was isolated from blood and CSF from the remaining infant (patient 10; table 2). The mean peripheral WBC count at the onset of illness was 20,400 cells/mm3 (range, 4400–71,400 cells/mm3), with a count of >15,000 cells/mm3 in most infants (61%). Thrombocytopenia (platelet count, <150,000 cells/mm3) was present in 7 infants (30%) at the onset of symptoms.
Echocardiography, to exclude intracardiac vegetations or vascular thrombi, was performed for 13 patients (57%). Two studies were abnormal; 1 infant had an intracardiac vegetation on the atrial septum, and 1 had a 5-mm thrombus at the junction between the superior vena cava and the innominate vein. Twenty infants (87%) underwent indirect ophthalmoscopic evaluation. Findings were abnormal for only 1 patient; this infant required vitrectomy for endophthalmitis (patient 21; table 2). Abdominal or renal ultrasonography was performed for 22 infants (96%); 3 (14%) had abnormalities detected. One infant had multiple hypoechogenic lesions in the liver, 1 had increased renal medullary echogenicity, and 1 had a single echogenic abnormality in the liver.
CT of the head was performed only for 5 infants (22%). Studies were done during antifungal therapy for 4 of these infants. One infant (patient 12; table 2) had communicating hydrocephalus with multiple microabscesses; a follow-up study performed 2 weeks later revealed multiple small parenchymal calcifications with no progression of hydrocephalus. Another infant (patient 8; table 2) had hydrocephalus with radiolucencies in both cerebral hemispheres and in the posterior fossa that was detected 2 months after completion of antifungal therapy. This infant had grade 4 IVH and died of multiorgan failure 1 week after this study. Three infants had studies with normal results.
Antifungal therapy. All infants were treated with iv amphotericin B deoxycholate. The infant with endophthalmitis (patient 21; table 2) also received intraocular amphotericin B. The initial iv dose ranged from 0.25 to 0.5 mg/kg. Within 24 h of the initiation of therapy, a daily dose of 1 mg/kg was administered. The cumulative dose of amphotericin B therapy ranged from 9 to 45 mg/kg (median, 30 mg/kg), and the duration ranged from 9 to 52 days (median, 31 days). Twenty patients (87%) received cumulative doses of 25–31 mg/kg. The infant with candidal meningitis, endophthalmitis, and arthritis received 45 mg of amphotericin B/kg. Two infants received cumulative doses of 9 mg/kg and 16 mg/kg when they developed fulminant and fatal sepsis caused by Staphylococcus aureus and Serratia marcescens. Five infants (22%) also received flucytosine treatment in dosages ranging from 75 to 100 mg/kg/d. In 3 of these 5 neonates, the CSF culture already was sterile when flucytosine was added to amphotericin B treatment. The duration of flucytosine therapy ranged from 3 to 20 days.
Nephrotoxicity manifested by a blood urea nitrogen level of ≥30 mg/dL or a serum creatinine level of ≥1 mg/dL was observed in only 7 infants (35%). Three infants had abnormal values before the initiation of antifungal therapy. Two infants had values that fell to the normal range when the daily dose of amphotericin B was decreased to 0.75 mg/kg. Renal function normalized without adjustment in the daily amphotericin B dose for the remaining 3 patients.
Outcome. Nine (35%) of 26 neonates died; 7 of these patients had definite meningitis. However, only 3 deaths (12%) could be related to candidal infection. These infants died before antifungal therapy was given, and they were found to have disseminated candidal infection at postmortem examination. Of the 6 other patients, 4 had completed antifungal therapy at the time of death and had no evidence of recurrent fungal disease. The other 2 infants developed fatal bacterial sepsis during antifungal therapy.
Premature infants, especially those weighing <1500 g at birth, are at increased risk for systemic candidiasis [1, 2, 7]. CNS involvement as an associated focus in disseminated candidal infection  was found in 25% of our neonatal intensive care unit patients with systemic candidiasis from 1989 through 1999. Candidal meningitis also has been described occasionally as a complication of neurosurgical procedures, but in neurosurgical patients, the route of infection is contiguous rather than hematogenous .
Besides preterm birth, other factors predisposing to the development of disseminated candidiasis in newborn infants include prolonged use of broad-spectrum antibiotics, use of steroids, total parenteral nutrition and intralipid administration, and invasive procedures, especially intravascular arterial or venous catheterization [10–13]. Most of our 23 infants with candidal meningitis had ≥1 of these predisposing factors.
Mucous membrane colonization with Candida also is an important antecedent event to invasive candidiasis. Up to 50% of infants who develop systemic candidiasis have been shown to be colonized with Candida [2, 14–16]. Invasive procedures that can disrupt normal mucosal integrity facilitate systemic invasion in colonized infants. Further, VLBW neonates colonized with Candida during the first week of life appear to be at a greater risk . Only 3 (13%) of our infants had colonization documented before candidal meningitis; however, in each infant, colonization occurred during the first week of life, and the site of colonization was the respiratory tract. Routine surveillance cultures for Candida and administration of oral nystatin prophylaxis for colonized neonates were proposed by Sims et al.  as a method to decrease the incidence of systemic candidiasis. However, this proposed practice would not identify all neonates at risk, would require repeated cultures of specimens from >1 site, and, most importantly, has not been evaluated for efficacy.
Most of our neonates had respiratory findings at the onset of candidal meningitis, which were manifested as an increase in the number of episodes of apnea and bradycardia and an increase in oxygen requirement. These clinical features are indistinguishable from those in infants with disseminated candidiasis without meningeal involvement  and from the features of patients with bacterial infection. Each of our patients underwent evaluation for systemic bacterial infection and was promptly treated with broad-spectrum antibiotics, typically vancomycin and gentamicin. Six infants also had presumed invasive fungal dermatitis  involving the back (this condition was skin biopsy proven in 3 infants); Candida was isolated from the CSF of each of these 6 infants. Because of the anatomic distribution of dermatitis, the CSF specimen might have been contaminated. However, the presence of invasive fungal dermatitis prompted an evaluation for disseminated fungal infection, and 2 infants were found to have candidemia. Cultures of CSF from another infant persistently yielded C. albicans, even after initiation of therapy with iv amphotericin B.
Findings of CSF analyses for infants with candidal meningitis are variable and relate closely to the type of CNS involvement (meningeal vs. parenchymal) [10, 17]. Normal CSF parameters do not exclude CNS involvement when Candida is isolated from CSF culture, since Candida may be at a low inoculum  and the inflammatory response may be meager or delayed. The former point is underscored by the uniform failure to visualize yeast forms by gram staining of CSF samples from our patients. In contrast, if an infant has pleocytosis and Candida is not isolated by CSF culture but is isolated from another normally sterile site (i.e., blood), the infant should be presumed to have candidal meningitis. Further, it should be appreciated that although most of 21 neonates had a predominance of mononuclear cells in their CSF, other researchers have reported a predominance of polymorphonuclear cells . Hypoglycorrhachia was observed inconsistently in our patients, as has been noted by other investigators [9, 18].
Consistent with other reports [4, 15, 19], C. albicans was the most frequent Candida species isolated from our 23 infants with meningitis, accounting for 96% of cases. C. glabrata was isolated from blood and CSF from a single infant. We and other investigators [3, 8] have observed an increase in the number of cases of systemic infection caused by Candida parapsilosis in neonates, but this species rarely is associated with meningeal invasion. Faix  described 17 infants with invasive C. parapsilosis infection, but only 3 had concomitant meningitis. No meningeal involvement was found in the 4 infants with invasive infection due to C. parapsilosis who were described by Lee et al. .
CT of the head was performed for a few of our patients. Abnormal findings for those infants who underwent diagnostic imaging included ventricular dilation, microembolic phenomena, and punctate calcified lesions in the parenchyma. Intracerebral calcifications have been previously described in association with candidal meningitis in a VLBW infant, presumably as a result of calcification of parenchymal granulomas . In another report of a case of a premature infant with C. albicans meningitis, multiple intraparenchymatous abscesses were detected by cranial ultrasonography and CT . These abnormalities most probably occur late in the course of treatment, as do radiographic findings for VLBW neonates with candidal osteomyelitis .
Most experts recommend amphotericin B in combination with flucytosine for the treatment of candidal meningitis [17, 18]. This recommendation is based on the fact that flucytosine penetrates CSF better than amphotericin B does [8, 17]. However, many of our infants were critically ill and unable to safely take oral feedings or medications. Five infants did receive flucytosine treatment, but amphotericin B alone failed to sterilize CSF in only 2 of these infants at the time when treatment with this antifungal agent was initiated. Treatment with amphotericin B alone was curative in 91% of the cases where it was used, and none of 6 deaths could be related to failure of antifungal therapy. The cumulative dose of amphotericin B routinely administered to our infants was higher than previously reported [3, 7, 8] and may, in part, explain the success of amphotericin B monotherapy. The maximal daily dose of amphotericin B also was achieved more quickly than reported by other investigators [19, 24]. Fluconazole monotherapy was used for 7 VLBW neonates with candidal meningitis . However, until comparable studies document the similar or superior safety and efficacy of antifungal agents other than amphotericin B de-oxy-cholate, it remains the drug of choice for the treatment of candidal meningitis in VLBW neonates.
Adverse effects, such as nephrotoxicity and electrolyte imbalance, are reported with the use of amphotericin B. These events can be diminished by careful fluid management. In our patients with increased blood urea nitrogen or creatinine levels for whom amphotericin B dosing was altered, rapid normalization of renal function was observed. Electrolyte imbalance occurred in a small number (9%) of our patients, and causes other than amphotericin B treatment were probable in each infant.
In summary, the frequent occurrence of meningitis (25%) in our neonates with systemic candidiasis suggests that all infants with systemic candidiasis should undergo CSF analysis. If CSF culture fails to yield Candida, pleocytosis (in infants without IVH) accompanied by isolation of Candida from another normally sterile site is evidence for meningitis, and such infants should receive appropriate antifungal therapy. Prompt initiation of therapy with amphotericin B alone appears to be adequate for C. albicans meningitis in most neonates, especially those in whom sterilization of CSF occurs quickly. Further assessment of the frequency of abnormalities by CT and their relationship to neurological outcome seems warranted.