Journal of Perinatology (2013) 33, 529–532 & 2013 Nature America, Inc. All rights reserved 0743-8346/13 www.nature.com/jp

ORIGINAL ARTICLE Results of a two-center, before and after study of piperacillin–tazobactam versus ampicillin and gentamicin as empiric therapy for suspected sepsis at birth in neonates p1500 g E Chong1, J Reynolds1, J Shaw2, L Forur1, P Delmore2, H Uner3, BT Bloom4 and P Gordon1

OBJECTIVE: We changed from ampicillin and gentamicin (AG) to piperacillin–tazobactam (PT) for routine treatment of suspected early-onset sepsis. The rationale for this change included ototoxic and renal toxic effects of gentamicin, resistance to gentamicin in late-onset infections and emergence of ampicillin resistant Escherichia coli. A before and after study was designed before the start of PT administration to monitor whether PT was associated with altered outcomes within the 501 to 1500 g birth weight (Very Low Birth Weight) population. METHOD: Both unmatched and matched comparisons of AG (2007 to 2009) and PT (2010 to 2011) exposed infants are reported. Cohorts were evaluated for initial effectiveness for congenital infections, subsequent morbidities and mortality. RESULT: Data from 714 patients were collected (499 AG and 215 PT in the unmatched and 301 AG and 183 PT in the matched cohorts). No significant differences in demographics or initial Apgar scores were noted in the unmatched or matched comparisons. There were significant differences in many of the outcomes of interest in both the matched and unmatched comparisons including less necrotizing enterocolitis (NEC) and less diaper rash with PT versus AG. The only adverse finding with PT was a small, but statistically significant elevation in alkaline phosphatase. CONCLUSION: Use of PT as the initial empiric antibiotic for very low birth weight infants was not associated with adverse microbiological outcomes. There was no increase in major morbidities. Although outcomes were superior in p1500 g infants treated with PT when compared with AG, the study design does not allow us to conclude that others will see a reduction in NEC or diaper rash if they implement this alternative.

Journal of Perinatology (2013) 33, 529–532; doi:10.1038/jp.2012.169; published online 17 January 2013 Keywords: necrotizing enterocolitis; ampicillin; gentamicin; piperacillin–tazobactam; diaper rash

INTRODUCTION inducible hearing loss.11 Boles, et al.12 proposed that neonates about The most commonly used antibiotic combination for therapy of to receive gentamicin should receive genetic screening for these presumed early-onset neonatal sepsis is ampicillin and gentamicin polymorphisms. This would be costly and difficult to accomplish. (AG), which is based upon the rationale that the majority of Second, gentamicin perturbs renal electrolyte excretion and 13–15 bacterial pathogens which infect newborns are derived from prolonged exposure causes neonatal hypocalcemia. Finally, vaginal flora. The most frequent offending gram negative bacteria aminoglycosides relax the ductus arteriosus in a mouse model of 16 include Escherichia coli, Haemophilus influenza and Citrobacter sp. ductal closure (but this finding has not been formally investigated The most prevalent gram positive organism is Group B Strepto- in humans). coccus, although this organism has been in decline since the The only available Cochrane Analysis to address the question of advent of intrapartum prophylaxis with ampicillin.1–3 antibiotic superiority for early sepsis concluded that no antibiotic There are many caveats to AG use that are of concern. For regime could be recommended over another based on currently 17 example, the dose of ampicillin commonly recommended is available data. Some antibiotic choices might not be as good as insufficient to treat Listeria meningitis, the worst complication of AG. A large retrospective study found higher mortality with 18 Listeria sepsis.4,5 Antenatal exposure to ampicillin increases the risk ampicillin and cefotaxime versus AG. In choosing an alternative of necrotizing enterocolitis (NEC)8 and ampicillin resistant E. coli is for our institutions, the following considerations were given: 18 prevalent.1–3 Prolonged postnatal exposure to empiric antibiotics (1) cephalosporins were excluded due to the Clark study, 16 in neonates has been associated with an increased risk of NEC.6–8 (2) aminoglycosides were excluded due to the Reese study, (3) Ampicillin also prolongs bleeding times through a direct effect on ampicillin was excluded due to E. coli resistance and the association 1–3,6–8 platelets.9,10 with NEC, (4) the antibiotic(s) would preferably not require An alternative to gentamicin is of interest for several reasons. As blood draws for levels, (5) the antibiotic(s) would have reasonable much as two percent of the population may have variable 12S coverage of the common pathogens associated with early newborn ribosomal polymorphisms known to be associated with gentamicin sepsis, (6) The antibiotic(s) cost would not be prohibitive, (7) the

1Department of Pediatrics, Tulane University School of Medicine Pediatrics: Neonatology, New Orleans, LA, USA; 2Department of Pediatrics, Wesley Medical Center, Wichita KS, USA; 3Department of Pediatrics, Kansas University of School of Medicine-Wichita, Wichita KS, USA and 4Wichita Medical Research and Education Foundation, Wichita, KS, USA. Correspondence: Dr P Gordon, Department of Pediatrics, Tulane University School of Medicine Pediatrics: Neonatology 1430 Tulane Avenue New Orleans, LA 70112 USA. E-mail: [email protected] Received 16 October 2012; revised 3 November 2012; accepted 7 November 2012; published online 17 January 2013 Study of piperacillin–tazobactam versus ampicillin/gentamicin E Chong et al 530 antibiotic(s) would not be commonly reserved or restricted by Statistics infectious disease specialists for highly resistant organisms and (8) Statistical methods included Mann À Whitney U, t-test and analysis of the antibiotic(s) would ideally already have a history of use (and thus, variance for nominal and continuous variables. Chi square was used for some reassurance of safety) in the neonatal population. categorical variables. Descriptive statistics are presented by group for Piperacillin–tazobactam (PT) is a synthetic penicillin derivative continuous variables. Significance was set at a P-valuep0.0012 using the combined with a beta-lactamase inhibitor. It is known to be Bonferroni correction (1/43 variables  0.05 P-value for base line signifi- especially effective for the lung and intestinal infections (which cance) to adjust for multiple comparisons. are common sources of entry for fetal infection due to amniotic aspiration or swallowing), and is frequently used in combination RESULTS with gentamicin in neonates to treat pneumonia. It covers the Population pathogen families seen in early neonatal sepsis.19,20 Levels are not A total of 714 patients were identified by our search, 499 exposed required for dosing, and its spectrum of coverage includes to AG and 214 exposed to PT. Of these, a maximum match of PT anaerobes, potentially making it favorable to enteric flora patients to two AG patients could be found for 118 PT patients, balance. Downsides of PT include variable penetration to the and 236 AG patients. 65 patients were added to each group with cerebrospinal fluid and potential to affect the emergence of drug the subsequent 1:1 match. resistance. We felt the available data supported a change to PT monotherapy as initial broad spectrum coverage for presumed early sepsis in the newborn with the caveat that infants with Demographics positive cultures would have immediate conversion to narrow Unmatched and matched patient cohorts were compared for spectrum therapy. This report describes a Phase IV, post differences in gestation, birth weight, mode of delivery or gender implementation, matched and unmatched comparison of the AG (Table 1). No differences were found in either cohort. Race was historical cohort with the PT implementation cohort. stable across the study period at both institutions

Mortality and common morbidities of prematurity METHODS We compared the cohorts for incidence of NEC, severe Data collection intraventricular hemorrhage, periventricular leukomalacia, alive We performed a retrospective before and after practice change, cohort and having not failed the newborn hearing screen, length of stay study involving 2 NICUs (Wesley Medical Center, Wichita, KS and Tulane and mortality (Table 1). We found that there were significant Lakeside Hospital, New Orleans, LA, USA). Both institutions received Institutional Review Board approvals for this study. We report both improvement in incidence of NEC with PT treatment compared unmatched and matched comparisons of AG exposed (2007 to 2009) and with AG treatment in both the unmatched and matched analyses. PT exposed infants (2010 to 2011). Inclusion criteria were inborn infants with birth weight of 501 to 1500 g, who had suspected systemic infection Miscellaneous outcomes and were receiving either AG or PT. During these two periods there were no changes in the indications for initiation of antibiotic coverage. We We assessed systemic rash, diaper rash (data from Wesley patients examined the impact on the Very Low Birth Weight population for two only), hypoglycemia and hyperglycemia (Table 1). We found reasons. First, we had existing intervention and outcome data with significantly fewer diaper rashes in PT exposed infants within both standardized definitions on all of these infants within our Vermont Oxford the unmatched and matched analyses Network database and second, these infants stayed well beyond the initial empiric phase and could be monitored for intermediate term impact on colonization or infection as well as toxic effects. Patients with methicillin Renal function and electrolytes resistant staphylococcus aureus colonization were to be excluded from this We analyzed data related to renal function and electrolytes study if they were treated with vancomycin during the empiric phase of around the time of antibiotic exposure and throughout hospita- therapy (a concern that was exceedingly rare). lization (Supplementary Appendix 1). AG treatment was asso- Paper and electronic records from the Neodata system were reviewed at ciated with lower mean calcium values in both the unmatched Tulane, where as a full electronic data extraction was used at Wesley. and matched cohorts and met significance for hypocalcemia in Cohorts were evaluated for composite morbidities and mortalities, the unmatched analysis, but was just above significance in the incidence of sepsis, patent ductus arteriosus treated with indomethacin, ibuprofen or ligation, NEC, ventilator-associated pneumonia (defined as matched cohort. per Vermont Oxford Network) and types of rash (diaper rash and systemic rash – rash outside the diaper area). Other profiles evaluated included Liver function outcomes neutropenia (o2500 mm À 3), thrombocytopenia (o100 000 mm À 3), 1 We assessed a panel of traditional markers for liver function for elevated or critically low glucose (o45 or 4125 mg dl À ), abnormal À 1 differences between AG and PT treated patients (Supplementary calcium (o8or411 mg dl ), abnormal sodium (o130 or 4150 mEq l À 1), creatinine 41.5 mg dl À 1, direct bilirubin 42 mmol l À 1, Appendix 2). There were no differences in total or direct bilirubin, aspartate aminotransferase 42 Â upper limit normal, alanine aminotrans- aspartate aminotransferase or alanine aminotransferase. However, ferase 42 Â upper limit normal or alkaline phosphatase 45 Â upper limit there was a significant increase in mean alkaline phosphatase normal on Days 1, 2 and 7. associated with PT use (but not incidence of elevated alkaline phosphatase) in both the unmatched and matched analyses. Matching Comparison groups included both the unmatched before and after cohorts Cardio-pulmonary interventions and outcomes as well as matched cohorts, consisting of one to one or two to computer- We assessed a number of cardio-pulmonary outcomes, including: generated patient matches (depending on the number of patients available in each cohort) matched by gestation at birth (same week), use of positive end-expiratory pressure devices devices, intubation, and, birth weight (±50 g). A SAS (Cary, NC) matching program was used to ventilator days, ventilator-associated pneumonia, patent ductus optimize the total number of infants available for analysis. The time period arteriosus, patent ductus arteriosus ligation and hypotension. There offered a limited number of patients and the originally planned 2:1 match were no significant differences (Supplementary Appendix 3). on the criteria of gestational age (same week) and birth weight (within 50 g) were limiting, so we used a multiple sequence algorithm to match as many total patients altering the number of matches of PT infants to two Empiric antibiotic exposure patients and to one patient until the total available for analysis could not There was no increase in exposure to the initial antibiotic. The be increased. median duration of the initial course of antibiotics was 3 days for

Journal of Perinatology (2013), 529 – 532 & 2013 Nature America, Inc. Study of piperacillin–tazobactam versus ampicillin/gentamicin E Chong et al 531 Table 1. Demographics, major morbidities and miscellaneous outcomes

Unmatched Matched

Amp and Gent Pip À Tazo Amp and Gent Pip À Tazo N 499 214 Unmatched P-value 301 183 Matched P-value

Demographics Birth weight 1.058±0.292 1.045±0.300 0.5967 1.079±0.289 1.071±0.295 0.7701 Gestation 28±2.6 28±2.7 0.9751 28±2.4 28±2.5 0.8423 Male 53% 47% 0.1678 54% 48% 0.1953 Vaginal delivery 28% 27% 0.9218 28% 28% 0.9441

Major Morbidities NEC 12.0% 1.4% o0.0001 11.0% 1.1% o0.0001 Severe IVH 6.1% 5.7% 0.8536 3.9% 6.7% 0.1722 PVL 1.4% 0% 0.0806 0.7% 0% 0.2686 Alive and hearing 66% 74% 0.095 66% 76% 0.0588 Length of stay 57±39 57±31 0.9153 55±37 77±31 0.5098 Mortality 10.2% 6.0% 0.0744 10.0% 4.9% 0.0478

Miscellaneous Outcomes Systemic rash 6.5% 1.1% 0.0046 6.7% 1.2% 0.0078 Diaper rash 48% 17% o0.0001 48% 18% o0.0001 Hypoglycemia 44.8% 47.4% 0.5175 44.9% 50.8% 0.2029 Hyperglycemia 37.0% 31.5% 0.146 35.2% 30.9% 0.2258

Abbreviations: Amp and Gent, ampicillin and gentamicin; IVH, intraventricular hemorrhage; PVL, periventricular leukomalacia; Pip À Tazo, piperacillin– tazobactam; NEC, necrotizing enterocolitis. Values relayed as means or percentiles. Standard deviation shown in parenthesis. Significance set at P-valueo0.0012. both groups with very little variation (Interquartile range was 0) in In our study periods, all positive blood cultures were reviewed. the unmatched and matched analysis. The organism, interpretation and antibiotic sensitivity were collected. Our first hypothesis, that PT would be sufficient for all congenital infections was correct for all cases of bacteremia. Microbiology We did not observe a change in late-onset infections or We assessed the isolates of all positive cultures throughout the sensitivities between the two treatment eras. Both institutions hospital stay and defined them as either early sepsis (day 0 to 2) or reserved PT monotherapy for early sepsis, all other use of PT late sepsis (day 3 to discharge) (Supplementary Appendix 4). (particularly later life or prolonged use) was in combination with Organisms causing infection during early sepsis included Group B gentamicin or another narrow spectrum gram negative antibiotic Streptococcus, E. coli, Citrobacteria and Pseudomonas. The E. coli as a means to discourage resistance. isolates were all resistant to ampicillin, as was Citrobacteria. One There are important limitations to our study. The first is that it is isolate of Pseudomonas was found to be resistant to AG. All of a retrospective, historical controlled observational study, and as these isolates were sensitive to PT. There were five admission such cannot be used to assert causation. Although the timing of cultures that were considered contaminants, including Bacillus our intervention correlated with a reduction in NEC and diaper species, Micrococcus species and Corynebacterium species, which rash; there may have been other variables in play, which we were were not treated beyond the limited empiric window (an average unable to capture in this detailed study (although we are not sure of 3 days). All early sepsis isolates were from the AG epoch (11) what they might be—for example, there was no known change in versus none in the PT epoch. There were also more late-onset antepartum antibiotic use or concurrent change in care across infections during the AG epoch when compared with the PT both neonatal intensive care units). epoch (Supplementary Appendix 4), but neither group reached statistical significance after the Bonferroni correction. There was CONCLUSION no change in the type of late-onset organisms or their sensitivities to either AG or PT. We also examined the median time to late- We describe a successful two center experience with PT onset sepsis to determine if PT might have altered the timing of monotherapy versus AG as empiric therapy for suspected sepsis in the unmatched cohort. It was 12 days (with an congenital sepsis in infants o1500 g. Reductions in NEC and interquartile range of 14 days) in the AG epoch versus 15.5 days diaper rash were noted in the PT treated patients. (with and interquartile range of 8 days) in the PT epoch. This was not significantly different (P-value ¼ 0.4384). CONFLICT OF INTEREST The authors declare no conflict of interest.

DISCUSSION REFERENCES In this phase IV monitoring study of a practice change, we identified no serious microbiological impact of PT use for 1 Stoll BJ, Hansen N, Fanaroff AA, Wright LL, Carlo WA, Ehrenkranz RA et al. Changes in pathogens causing early-onset sepsis in very-low-birth-weight infants. suspected neonatal sepsis. A reduction in NEC and diaper rash N Engl J Med 2002; 347: 240–247. were each associated with PT. An elevation in alkaline phospha- 2 Bizzarro MJ, Dembry LM, Baltimore RS, Gallagher PG. Changing patterns in neo- tase was also noted with PT. Further research into this novel natal Escherichia coli sepsis and ampicillin resistance in the era of intrapartum finding is required to determine if it is causally related to PT. antibiotic prophylaxis. Pediatrics 2008; 121: 689–696.

& 2013 Nature America, Inc. Journal of Perinatology (2013), 529 – 532 Study of piperacillin–tazobactam versus ampicillin/gentamicin E Chong et al 532 3 Puopolo KM, Eichenwald EC. No change in the incidence of ampicillin-resistant, 12 Boles RG, Friedlich P. Should patients be screened for 12S rRNA mutations before neonatal, early-onset sepsis over 18 years. Pediatrics 2010; 125: e1031–e1038. treatment with aminoglycosides? Mitochondrion 2010; 10: 391–392. 4 Marget W, Seeliger HP. Listeria monocytogenes infections–therapeutic possibi- 13 Giapros VI, Cholevas VI, Andronikou SK. Acute effects of gentamicin on urinary lities and problems. Infection 1988; 16: S175–S177. electrolyte excretion in neonates. Pediatr Nephrol 2004; 19: 322–325. 5 Hof H. An update on the medical management of listeriosis. Expert Opin Phar- 14 Chiruvolu A, Engle WD, Sendelbach D, Manning MD, Jackson GL. Serum calcium macother 2004; 5: 1727–1735. values in term and late-preterm neonates receiving gentamicin. Pediatr Nephrol 6 Cotten CM, Taylor S, Stoll B, Goldberg RN, Hansen NI, Sanchez PJ et al. Prolonged 2008; 23: 569–574. duration of initial empirical antibiotic treatment is associated with increased rates 15 Tugay S, Bircan Z, Caglayan C, Arisoy AE, Gokalp AS 2006Acute effects of gen- of necrotizing enterocolitis and death for extremely low birth weight infants. tamicin on glomerular and tubular functions in preterm neonates. Pediatr Nephrol Pediatrics 2009; 123: 58–66. 21: 1389–1392. 7 Alexander VN, Northrup V, Bizzarro MJ. Antibiotic exposure in the 16 Reese J, Veldman A, Shah L, Vucovich M, Cotton RB. Inadvertent relaxation of the newborn intensive care unit and the risk of necrotizing enterocolitis. J Pediatr ductus arteriosus by pharmacologic agents that are commonly used in the 2011; 159: 392–397. neonatal period. Semin Perinatol 2010; 34: 222–230. 8 Weintraub AS, Ferrara L, Deluca L, Moshier E, Green RS, Oakman E et al. Antenatal 17 Mtitimila EI, Cooke RW 2004Antibiotic regimens for suspected early neonatal antibiotic exposure in preterm infants with necrotizing enterocolitis. J Perinatol sepsis. Cochrane Database Syst Rev CD004495. 2012; 32: 705–709. 18 Clark RH, Bloom BT, Spitzer AR, Gerstmann DR. Empiric use of ampicillin and 9 Sheffield MJ, Lambert DK, Baer VL, Henry E, Butler A, Snow GL et al. Effect of cefotaxime, compared with ampicillin and gentamicin, for neonates at risk for ampicillin on bleeding time in very low birth-weight neonates during the first sepsis is associated with an increased risk of neonatal death. Pediatrics 2006; 117: week after birth. J Perinatol 2011; 31: 477–480. 67–74. 10 Sheffield MJ, Lambert DK, Henry E, Christensen RD. Effect of ampicillin on 19 Kuck NA, Jacobus NV, Petersen PJ, Weiss WJ, Testa RT. Comparative in vitro and the bleeding time of neonatal intensive care unit patients. J Perinatol 2009; 30: in vivo activities of piperacillin combined with the beta-lactamase inhibitors 527–530. tazobactam, clavulanic acid, and sulbactam. Antimicrob Agents Chemother 1989; 11 Ealy M, Lynch KA, Meyer NC, Smith RJ. The prevalence of mitochondrial mutations 33: 1964–1969. associated with aminoglycoside-induced sensorineural hearing loss in an NICU 20 Daley D, Mulgrave L, Munro R, Neville S, Smith H, Dimech W. An evaluation of the population. Laryngoscope 2011; 121: 1184–1186. in vitro activity of piperacillin/tazobactam. Pathology 1996; 28: 167–172.

Supplementary Information accompanies the paper on the Journal of Perinatology website (http://www.nature.com/jp)

Journal of Perinatology (2013), 529 – 532 & 2013 Nature America, Inc.