FormalPara Key Points

The literature on ceftriaxone use in neonates suffers from poor generalizability, small sample sizes, and inconsistent/passive outcome assessment. The poor quality of the literature makes it difficult to interpret the risk from ceftriaxone in neonatal patients.

Ceftriaxone may be fairly safe in low-risk healthy infants.

Comparable alternative agents are available, which may mean the risks of using ceftriaxone in neonates may outweigh the benefits.

1 Introduction

Ceftriaxone is a commonly used parentally administered antibiotic for acute bacterial infections such as pneumonia, urinary tract infections, sepsis, and meningitis [1, 2]. This is because it has broad-spectrum activity against both Gram-positive and Gram-negative bacteria. Like other third-generation cephalosporins, it is more effective than the preceding generations of cephalosporins against Gram-negative bacteria [2]. In addition, it has excellent tissue distribution and reaches clinically effective concentrations in the cerebrospinal fluid [2]. However, despite its long serum half-life, simple dosing schedule, favorable safety profile, and clinical usefulness in many types of infections, the use of ceftriaxone in neonates has been limited by concerns about toxicity related to pulmonary ceftriaxone–calcium precipitates associated with cardiopulmonary events, hyperbilirubinemia, biliary sludge, and pseudolithiasis [2, 3].

In September 2007, the US FDA issued an alert to healthcare professionals that patients, regardless of age, should not receive ceftriaxone and solutions containing calcium within 48 h of each other [4]. The alert was based on FDA post-marketing reports of five neonatal and three infant deaths, including those born prematurely, related to the interaction between ceftriaxone and calcium-containing products and autopsy evidence of crystalline materials in the pulmonary and or renal vasculatures in three of the cases. The alert was later modified by the FDA in April 2009 to state that ceftriaxone and calcium-containing solutions may be sequentially administered in patients aged >28 days if the infusion lines are thoroughly flushed between infusions with a compatible fluid. This advice was based on the results of in vitro studies using neonatal and adult plasma conducted by Roche Pharmaceuticals [5].

Ceftriaxone–calcium precipitates are not the only adverse event that may be of concern when using the medication in neonates. The long half-life of ceftriaxone results partially from its binding to serum albumin. Ceftriaxone competes with bilirubin for albumin binding and also displaces it from its binding sites [6–9]. This can lead to an increase in free bilirubin and erythrocyte-bound bilirubin in neonates at risk, even within therapeutic ranges [10]. Because of the potential of ceftriaxone to induce bilirubin encephalopathy, it is recommended that ceftriaxone be avoided in neonates, including premature infants, at high risk for kernicterus [9, 10]. Reversible biliary sludge or pseudolithiasis has been reported as a potential complication of ceftriaxone treatment since 1986 [11]. The sludge occurs when solutes in the bile precipitate as a mixture of particulate matter such as calcium bilirubinate granules or calcium–ceftriaxone complexes [12, 13]. The sludge is often asymptomatic but may also present as biliary colic, cholecystitis, cholangitis, or pancreatitis [14]. In this study, we reviewed the literature for the risk of hyperbilirubinemia and pseudolithiasis associated with use of ceftriaxone among neonates, including preterm infants, as well as for the risk of cardiopulmonary events related to the formation of ceftriaxone–calcium precipitates associated with concomitant infusion of ceftriaxone and calcium.

2 Methods

2.1 Eligibility Criteria

An article was eligible for inclusion in this review if (1) it was an in vivo study of ceftriaxone that included human neonates (age ≤30 days); (2) it was a full peer-reviewed published report, and not an editorial, conference abstract, review, or letter; (3) the full text was available to the authors; (4) it was written in English; and (5) it included outcomes on relevant adverse events in neonatal patients. With regard to the final criteria, studies that included patients who fell within a wide age range of neonates and older children (e.g., age 2 days to 16 years) were excluded if they did not clearly state the age of patients who experienced adverse events. No limitations on article publication date were applied.

2.2 Search Strategy

PubMed and EMBASE were searched for relevant articles in March 2016. The full search strategy, including search terms and filters for each database can be viewed in Appendix 1 in the Electronic Supplementary Material (ESM). The final search terms were reviewed and approved by one of the authors (LL), who is a specialist in pediatric infectious diseases.

2.3 Study Selection Process

Search results were imported into a reference manager database, and duplicate references were removed. References were then imported into a web-based system called Presentation and Evaluation of Evidence-Based Research (PEER) developed by the American Academy of Orthopaedic Surgeons (AAOS). Two authors (RS, PD) reviewed the abstracts and included any article for full-text review that appeared relevant to the topic. A large number of abstracts were returned, so we divided them equally between the two reviewers. The full-text articles were then reviewed by one author and excluded if they did not meet the inclusion criteria.

2.4 Data Extraction

The following data elements were extracted from each of the full-text articles for the neonatal patients: treatment type, route of administration, number of patients, average age or age range of patients, gestational age, length of follow-up, number of adverse events, and type of adverse events. These data were entered directly into a Microsoft® Excel spreadsheet.

2.5 Risk of Bias Evaluation

The risk of bias for each of the included studies was assessed using the Quality of Prognosis Studies (QUIPS) tool [15]. Although originally developed to evaluate prognostic risk factor studies, the instrument comprises five quality domains that are directly applicable to pharmacoepidemiological research. These include study participation (external validity/generalizability), attrition, exposure measurement bias, outcome measurement bias, confounding, and bias in statistical analysis (including selective outcome reporting).

3 Results

3.1 Search Results

The initial search resulted in 478 abstracts that were included for review (Fig. 1). After abstracts were reviewed, 66 full-text articles were retrieved to determine whether they met the inclusion criteria. Nine of these 66 full-text articles met the inclusion criteria. Of the 57 articles we excluded, 32 were because they did not include neonates or specify the age of the population, seven were because they were in vitro studies, eight were because they did not report adverse events, and six were because the full-text article was unavailable.

Fig. 1
figure 1

Study selection process

Full size image

3.2 Risk of Bias

Table 1 presents the risk of bias in each QUIPS domain for the nine included studies, along with reasons for the ratings. While the risk of bias due to attrition was generally low, a lack of comparison groups, poor outcome reporting, and small sample sizes were common and contributed to the overall low quality of the evidence base. However, the most consistent methodological flaws in the neonatal ceftriaxone literature were low external validity and high risk of information bias due to inconsistent outcome measurement.

Table 1 Risk of bias assessment using the quality of prognosis studies (QUIPS) instrument
Full size table

The risk of external validity bias was either moderate or high in eight of nine studies. Four of those specifically chose patients who met pre-specified criteria that qualified them for outpatient treatment [9, 16–18]. One study had low external validity because they evaluated spontaneous reports, which may not be easily generalizable to all neonatal patients receiving ceftriaxone [19]. One study stratified patients by whether they were aged ≤3 days [20]. The older group in this study included 12 patients who were originally in the younger group and received a second treatment episode when they were older. The risk of bias was rated as high for this group because it included patients who survived the first treatment episode. One published case report was also unlikely to be generalizable to other neonatal populations [21]. Two studies did not adequately describe the inclusion criteria and were rated as having an unclear risk of bias [22, 23].

Every study had at least some risk of outcome misclassification. The most common reason for bias was that diagnostic testing was not standardized for all patients [9, 16–18, 22, 23]. The one study that investigated ceftriaxone compared with other antibiotics was at very high risk of differential misclassification bias because patients who received ceftriaxone were required to have bilirubin levels tested before hospital discharge, whereas the comparison groups were not [18]. One article relied on cases spontaneously reported to the FDA and was rated at high risk of outcome misclassification bias because autopsy information was only available for four of the five neonates [19]. Another published case report was rated at unclear risk of bias because it lacked a clearly defined outcome definition [21]. Finally, one study was considered at high risk because some events were presented as prevalent outcomes [20].

The confounding domain was only relevant to two studies because the others did not include an unexposed or comparison group. Wagner et al. [18] investigated ceftriaxone compared with other antibiotics and was at high risk of confounding by indication because a patient could only be given ceftriaxone if his/her bilirubin level was <8 mg/dl. The correlational examination by Martin et al. [9] was a bivariate analysis that was too small to adjust for potential confounders and thus can only present crude associations that do not account for differences between the treated groups.

3.3 Biliary Adverse Events

Of the studies and reports included from the published literature, six reported on potential adverse events associated with biliary function as a result of ceftriaxone therapy.

In 1985, James et al. [22] reported on the use of intravenous or intramuscular ceftriaxone 50 mg/kg/day in place of ampicillin plus gentamicin in 104 premature neonates with suspected sepsis. The average gestational age of the patients was 32 ± 7.4 weeks. Treatment was stopped after 48 h if infant conditions improved but was continued for up to 5 days if suspicion of infection persisted. The safety of ceftriaxone was assessed at the end of treatment. Ten of the 104 neonates who received ceftriaxone died. One of the deaths was due to infection; none of the remaining nine deaths were attributed to ceftriaxone. Nine of the 104 neonates exhibited clinical deterioration, and treatment was changed from ceftriaxone to gentamicin plus penicillin or ampicillin. How the authors defined clinical deterioration is unclear; nor do they indicate whether the deterioration was associated with ceftriaxone-related adverse events.

In 1992, Bradley et al. [17] performed a cohort study of 46 neonates with a clinical diagnoses of sepsis, pneumonia, or urinary tract infection. The data on patient ages were not complete; therefore, the average patient age was not reported. Only low-risk patients suitable for outpatient therapy were enrolled (see Table 1 for low-risk criteria). Additionally, patients had to survive and respond to antibiotics other than ceftriaxone in an inpatient setting for 48 h to be included in the study. This limited the external validity of the study finding to patients considered to be at low risk. For treatment, intramuscular or intravenous ceftriaxone 20–50 mg/kg/day was administered to patients. Higher doses were recommended at study onset, and the lower dosage levels were later suggested as pharmacologic data became available. The duration of treatment depended on the type and severity of infection. This study did not include regular kidney and liver function laboratory tests or investigations of bilirubin levels during follow-up, leaving it at high risk for outcome misclassification bias. The authors did state that no treated patient developed incident or worsening jaundice as a result of therapy.

In 1993, Martin et al. [9] performed an in vivo study assessing bilirubin–albumin binding interaction with ceftriaxone in plasma samples from 14 non-jaundiced neonates aged <11 days and requiring antibacterial therapy. Gestational ages ranged from 33 to 42 weeks (average 37 ± 3). The included patients received intravenous ceftriaxone 50 mg/kg/day infused over 30 min. Baseline bilirubin, ceftriaxone, reserve albumin concentration, plasma bilirubin toxicity levels, and displacement factor were tested prior to ceftriaxone administration, during drug infusion, and 15 min following infusion; six patients received follow-up 60 min after infusion. The authors found the plasma bilirubin toxicity index increased significantly immediately following infusion of ceftriaxone (0.64 to 0.96; p < 0.0001) but returned to baseline levels thereafter. Reserve albumin levels decreased from 91.9 to 38.6 mmol/l immediately after infusion and to 58.3 mmol/l at 30 min after infusion. There was a statistically significant correlation between higher ceftriaxone plasma concentration levels and lower reserve albumin concentration (p = 0.0095). Based on these results, there appears to be evidence for albumin-binding competition between the ceftriaxone antibiotic and bilirubin.

In 1995, Van Reempts et al. [20] published a clinical report on a study of the safety of ceftriaxone in 80 neonates with 92 infectious episodes. The average gestational age of the entire sample was 33.8 ± 4.4 weeks. Included subjects were in one of two treatment groups: group A included neonates aged <72 h and given ceftriaxone and ampicillin; group B included neonates aged >72 h and given ceftriaxone and vancomycin. Two patients in group A were diagnosed with incident cases of hyperbilirubinemia after the ceftriaxone and ampicillin infusion, but they did not require exchange transfusion. No new cases of hyperbilirubinemia were observed among group B neonates. Six patients combined from groups A and B had positive sonographic signs of biliary sludge, but all cases resolved spontaneously within 2 weeks after detection. Whether these were incident or prevalent cases is unclear.

In 2000, Wagner et al. [18] reported on a prospective cohort study of 95 neonates diagnosed with sepsis, presumed sepsis, pneumonia, or uncomplicated meningitis who met inclusion criteria for outpatient treatment with antibiotics, which included non-jaundiced infants. Their average gestational age was 38.4 ± 2.1 weeks, and the average age at hospital discharge was 5.2 days (range 4–12). Patients were treated with ampicillin and gentamicin (n = 53) or ceftriaxone (n = 20) or other antibiotics (n = 22). Four of the 20 (20 %) who commenced treatment with ceftriaxone before discharge had to be switched to other antibiotics because of elevated bilirubin (>8 mg/dl), but none required exchange transfusion, and no infant was readmitted to an inpatient setting for hyperbilirubinemia. Bilirubin levels were only measured for neonates who received ceftriaxone. It was not possible to compare effects on bilirubin between the groups in this study.

Takei et al. [21] reported a case in 2013 describing a 15-day-old neonate’s inpatient admission due to group B Streptococcus infection. She was born at 39 weeks’ gestation. Prior to admission, the patient carried a previous diagnosis of jaundice. She received parenteral ceftriaxone 60 mg/kg/day and ampicillin 120–200 mg/kg/day. The patient’s overall health improved due to the ceftriaxone–ampicillin therapy; however, she showed signs of progressive jaundice. Subsequent gallbladder and biliary tract ultrasound tests did not reveal any abnormalities. The patient was later diagnosed with congenital biliary atresia, a childhood disease of the liver in which one or more bile ducts are abnormally narrow, blocked, or absent [24]. This makes it unlikely the patient’s exposure to ceftriaxone was related to the progressed jaundice while admitted for care.

3.4 Cardiopulmonary Adverse Events

Of the studies and reports included from the published literature, three reported adverse cardiopulmonary events as a result of ceftriaxone therapy.

In the previously mentioned 1995 paper by Van Reempts et al. [20], three neonates in group A died from adverse events not attributed to the infection for which they were being treated. These adverse cardiopulmonary events included perinatal asphyxia and persistent pulmonary hypertension. No infants in group B died from cardiopulmonary adverse events. Therefore, death due to cardiopulmonary events occurred in 4.1 % of the total sample. Additionally, thrombocytosis was reported in 11 neonates (13.8 %) spanning both groups. The authors note this primarily occurred after cessation of ceftriaxone therapy, but there is no clear indication of incident versus prevalent cases or whether the event was due to surgery or recovery from infection.

In 2009, Bradley et al. [19] published an assessment of cardiopulmonary adverse event reports in neonates administered intravenous ceftriaxone concurrent with administration of calcium-containing solutions. The authors searched the FDA Adverse Event Reporting System (AERS) database for voluntary case reports of probable adverse drug events in this population. The search generated eight case reports from healthcare professionals submitted to the FDA by manufacturers of ceftriaxone as a result of mandatory reporting requirements. Five infants were aged <3 weeks, two were in their second month of life, and one patient was of unknown age at the time of the event. Five cases were preterm, with the earliest birth occurring at a gestational age of 30 weeks. These eight neonates and young infants from outside of the USA experienced cardiorespiratory arrest while receiving intravenous ceftriaxone and intravenous calcium. All infants received intravenous ceftriaxone, seven of the eight infants received intravenous calcium gluconate, and one received calcium as a part of hyperalimentation. The dosage of ceftriaxone varied and was not consistently reported; however, at least three infants received multiple doses. Seven of the eight cases died, and autopsy findings from four of the five infants with available information indicated the presence of crystalline material or white precipitate in the lungs. Overall, the reporting of underlying disease, dose of medication administered, and other medication usage during the time period in these case reports was limited or inconsistent. We also cannot ascertain a total population administered ceftriaxone and calcium-containing solutions because this is an event-driven reporting system; therefore we do not know the true population risk.

In addition to the aforementioned papers, in 1990 McCarthy et al. [16] reported on a study conducted on outpatient management of selected infants with possible sepsis. The infants ranged in age from 11 to 59 days (mean 36) and were required to have a gestational age of at least 37 weeks. Patients were administered intramuscular ceftriaxone on study days 1 and 2 and were assessed for fever, irritability, lethargy, anorexia, respiratory symptoms, vomiting, diarrhea, rash, or injection site reactions. Respiratory symptoms were reported at some point in the study period in 51 of the 86 infants in the study. However, of note, whether these respiratory symptoms can be attributed to ceftriaxone or other underlying conditions is unclear. No effort was made to describe these symptoms in great detail. Additionally, the authors do not distinguish between neonates, the population of interest for this paper, and infants aged >30 days. How the results of this paper truly add to the body of evidence for cardiopulmonary adverse events from ceftriaxone is unclear.

Martin et al. [23] administered ceftriaxone 50–100 mg/kg to 43 children, including ten neonates who ranged in age from 6 to 30 days. No bleeding complications, changes in hemostasis, neutropenia, or thrombocytopenia were found. However, the sample size was small, and these outcomes were not consistently assessed.

4 Discussion

We have reviewed the use of ceftriaxone in neonates and assessed whether serious adverse events, particularly bilirubin displacement and the risk of kernicterus, and cardiopulmonary events associated with concomitant use of calcium-containing solutions, preclude the use of ceftriaxone in this population. Pharmacokinetic and pharmacodynamic studies in neonates suggest these adverse effects are limited to ceftriaxone and are not a concern with other oxyimino-cephalosporins [25]. Ceftriaxone is a widely used extended-spectrum cephalosporin; its popularity is mainly due to its convenient (daily) dosing schedule [26]. However, the pharmacokinetics and pharmacodynamics of ceftriaxone differ in neonates, a diverse population with varied gestational age/maturity, body surface area, feeding, medications, ventilator dependence, comorbidities, and other parameters affecting drug distribution, metabolism, and side effect profiles. We found nine studies with significant methodological limitations and a relatively homogenous low-risk patient group, which limits the ability to draw inferences to such a diverse general neonatal population.

Pediatric drug trials require extensive approval processes, as they involve a vulnerable population. Thus, pediatric pharmacoepidemiological data often lags (years) behind adult data. Clinical trials conducted in children are often small, limiting the ability to detect adverse drug events. Therefore, it is often necessary to rely on case reports or observational studies in children to adequately understand whether there is a safety problem with a medication used in the pediatric population. With respect to ceftriaxone, it first came to market in 1982, but the FDA did not warn of potential serious interactions between ceftriaxone and calcium-containing solutions until 2007 [4, 5]. Importantly, safety signals may have occurred as early as 1985 for an interaction between ceftriaxone and calcium solutions that could lead to severe adverse events. For example, James et al. [22] found the cause of death in five of ten infants was reported as respiratory distress syndrome. Could this have been related to ceftriaxone? No mention was made of pulmonary autopsy findings, and there was no control group with which to compare outcomes.

The available ceftriaxone data in neonates are subject to great bias, mainly due to issues with study design. The current literature is primarily limited to observational studies and passive surveillance databases where under-reporting and selective reporting occur. This does not allow for incidence calculations to clearly understand the risks associated with the use of ceftriaxone in this population. Furthermore, because alternative agents are available to treat most infections for which ceftriaxone is indicated, the risk of using ceftriaxone in neonates may outweigh the benefit.

However, the adverse events reported ranged in severity, and a risk–benefit assessment may need to be stratified to determine whether the use of ceftriaxone in subgroups of neonates would be appropriate. For example, Martin et al. [9] found the moderate total serum increases in bilirubin returned to zero immediately after the drug infusion in term and near-term neonates. An interpretation of this finding may be that this does not represent a clinically significant adverse event. Moreover, alternative data suggest ceftriaxone does not bind to the same sites on albumin as bilirubin, which would decrease the likelihood of bilirubin displacement in neonates [20].

It is probable that ceftriaxone can be considered in the neonate with mild or transient hyperbilirubinemia; however, practitioners generally avoid it because of perceived risk. There are also no data on whether the interaction between calcium and ceftriaxone could be mitigated by changes in administration route [19]. For example, could combinations such as intramuscular ceftriaxone and intravenous calcium or intravenous ceftriaxone and oral calcium be acceptable?

The results of this review must be interpreted with caution because of certain limitations. The most important of these is the quality of the included studies. Conducting clinical trials in neonatal patients presents unique ethical, scientific, and regulatory challenges [27] that make designing a high-quality prospective controlled study on ceftriaxone safety in newborns extremely difficult. Therefore, the evidence base mostly consists of case series, published case reports, and spontaneous reports. In general, the infants represented a carefully selected low-risk population. Furthermore, the sample sizes of the included studies were likely too small to detect rare events. Furthermore, the methods of outcome ascertainment were mostly passive and often not consistently applied to all patients in a study.

A further limitation of this review is that studies had to be excluded when the full-text reports could not be obtained through the authors’ institutions. Resource restrictions limited our ability to obtain these additional articles, which meant that six possibly relevant studies were excluded. However, upon re-review of their abstracts, it appears that most would likely have been excluded on full-text review or would have produced the same conclusions and been subject to the same methodological limitations. These included two small case series (n = 11 and n = 21) [28, 29] that made no mention of biliary or cardiopulmonary events in the abstract; one case report of successful ceftriaxone treatment with no mention of adverse events in the abstract [30]; one article with no outcomes (adverse events or clinical) mentioned in the abstract that was likely a narrative literature review or commentary [31]; and one case series that looked at complications in children of all ages that did not appear to stratify events by age in the abstract [32] (the sixth study did not provide an abstract [33]). Given this information and the inherent ethical and methodological challenges in designing a large high-quality and generalizable trial for ceftriaxone in such a vulnerable population, it is unlikely that including these six articles would have altered the conclusions of this review. An additional limitation is that non-English language studies were not included in this review. If adverse event rates reported in non-English studies differ from those reported in articles included in this review, then our results may be biased.

It is worth noting that most of the publications included in this review are not recent. Six of the nine articles were published in or before 1995. Study reporting standards, such as the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines [34], were only developed in 2007. It is probable that the quality of outcome reporting in the reviewed studies would have been improved if they were published after reporting guidelines were developed.

The final limitation is that we did not search for unpublished literature. It is possible that event rates in unpublished studies could be higher than in published articles. The lack of controlled trials and inconsistent outcome ascertainment meant that we could not conduct a meta-analyses. Therefore, no statistical tests for selective publication could be performed, and publication bias cannot be ruled out.

5 Conclusion

Ceftriaxone remains a standard treatment for severe bacterial infections in children. It is essential that sequential or concomitant administration of intravenous ceftriaxone and calcium-containing solutions be avoided because of the risk of cardiopulmonary events. However, ceftriaxone could be considered in a subset of healthy term or near-term neonates, as the convenience of daily dosing may outweigh any clinically relevant side effects of bilirubin displacement.