Outbreak of Bacteremia Due to Burkholderia contaminans Linked to Intravenous Fentanyl From an Institutional Compounding Pharmacy

Importance  Many health care facilities compound medications on site to fulfill local demands when customized formulations are needed, national supply is critically low, or costs for manufactured pharmaceuticals are excessive. Small, institutional compounding facilities may perform the same high-risk procedures as large distributors of compounded medications.

Objectives  To investigate an outbreak related to contamination of compounded sterile preparations and to determine processes to prevent future outbreaks.

Design, Setting, and Participants  We performed an outbreak investigation of inpatients at Duke University Hospital from August 31 through September 6, 2012. The investigation included a case-control study, compounding facility inspection and environmental sampling, observation of a mock compounding demonstration, and microbiologic and molecular testing of sequestered medication.

Exposures  Intravenous fentanyl prepared by an institutional compounding pharmacy.

Main Outcomes and Measures  Microbiologic and molecular evidence of contamination of a compounded sterile preparation and failure of routine sterility testing.

Results  Blood cultures of 7 patients during a 7-day period at Duke University Hospital yielded pan-susceptible Burkholderia cepacia complex bacteria. The risk factor common to all patients was receipt of continuous fentanyl infusion prepared by our institutional compounding pharmacy (odds ratio, 11.22; 95% CI, 2.09-∞; P = .01). The outbreak was terminated after sequestration of compounded fentanyl. An intensive evaluation of the compounding facility, its practice, and its procedures was completed. Investigators evaluated the clean room, collected targeted microbiologic samples within the compounding pharmacy environment, and observed a mock demonstration of compounding practice. The B cepacia complex was found in the anteroom sink drain and pH probe calibration fluid from the compounding clean room. Multiple microbiologic analyses of sequestered fentanyl initially failed. Ultimately, a batched, vacuum-assisted filtration method produced B cepacia complex from a single lot. Molecular analyses using repetitive element polymerase chain reaction and pulsed-field gel electrophoresis confirmed a clonal Burkholderia contaminans strain from patients, fentanyl, and environmental samples.

Conclusions and Relevance  An outbreak of B contaminans bacteremia was linked to contamination of locally compounded intravenous fentanyl. Health care facilities that house institutional compounding facilities must be vigilant in efforts to prevent, recognize, and terminate medication-related outbreaks.

The inherent risk involved in compounding sterile pharmaceuticals is well established and has resulted in multiple previously reported outbreaks.^1-3 The recent multistate outbreak of fungal meningitis due to contaminated corticosteroids from a nationally distributing compounding pharmacy resulted in 64 deaths, 750 illnesses, and increased scrutiny from the public and regulators.^1,4 However, smaller institutional compounding pharmacies perform these same high-risk compounding procedures and may produce substantial volumes of medications. Medical institutions are increasingly challenged in their ability to obtain adequate supplies of safely prepared, essential medications. Consequently, an increasing number of health care facilities have established compounding pharmacies to fulfill local demand when national supply is critically low, customized formulations are needed, or costs for manufactured pharmaceuticals are excessive.

The US Pharmacopeia (USP) categorizes pharmacy compounding processes according to risk of patient harm due to contamination events.⁵ Nonsterile-to-sterile preparations are considered highest risk because they involve nonsterile raw ingredients and are subsequently used in sterile body spaces (eg, intravenous and epidural). These preparations must undergo a sterilization procedural step and sterility testing before patient use. A sterility testing method that guarantees compounded sterile preparations are free of contamination and safe for use in patients does not exist. The USP provides guidelines for sterility testing in chapter 71 but also directly states, “These Pharmacopeial procedures are not by themselves designed to ensure that a batch of product is sterile or has been sterilized.”^{6(p 85)} Contamination may be sporadic, organism burden low, and fastidious organisms difficult to propagate in standard microbiologic culture. In fact, obtaining definitive microbiologic proof of a contaminated product can be a distinct challenge, even when strongly implicated by epidemiologic data in the setting of an outbreak. We present a single-center, multiunit outbreak investigation of 7 cases of bacteremia due to Burkholderia contaminans attributed to contamination of locally compounded intravenous fentanyl.

Duke University Hospital is a 924-bed, tertiary care hospital located in Durham, North Carolina. The Duke Compounding Facility is an on-site, high-volume compounding pharmacy that generates approximately 150 000 doses of medications annually. Approximately 50% of preparations are sterile; of these, 85% are nonsterile-to-sterile preparations and considered highest risk.^6,7 The Duke Clinical Microbiology Laboratory notified the Program for Infection Prevention and Healthcare Epidemiology on September 6, 2012, that 4 patients had bacteremia due to unusual Burkholderia cepacia complex (BCC) bacteria during a 6-day period. The Duke University institutional review board determined that this investigation was exempt from institutional review board approval and that patient consent was not required.

The BCC isolates were identified by fatty acid analysis (Sherlock Microbial Identification System; MIDI Inc). Identity of BCC was supported by matrix-assisted laser-desorption ionization–time of flight mass spectrometry (BioMerieux) with an identity score of 99%. A case was defined as a patient with a blood culture positive for BCC that matched the unique susceptibility profile of the clustered infections. We reviewed microbiology records from September 15, 2011, through September 15, 2012, to identify cases and determine the baseline incidence rate of bloodstream infection due to BCC. Both an epidemiologic curve and a line listing of common exposures and clinical features for each case patient were constructed. Investigators conducted interviews with nurses and managers on affected units to identify additional exposures or aberrancies in patient care.

A case-control study was performed to determine risk factors for BCC bacteremia during the outbreak period. Case patients were matched 3:1 with control patients on the first day of a positive blood culture collection (day 0). Controls were selected from unit rosters run on day 0 and matched to cases by 2 factors: unit of residence on day 0 and similar number of days of residence (±3 days) on that unit. Event date and time for controls were defined as the first positive blood culture collection date and time for their matched case. Exposure data were abstracted from patient medical records and recorded on standardized case record forms.

Potential causal exposures occurring before event time were analyzed for association with case status using 2-sided χ² or Fisher exact tests for categorical variables or the t test or Wilcoxon rank sum test for continuous variables, as appropriate. Exposures with significant association (P < .15) were further evaluated by exact logistic regression conditional on matched factors to calculate odds ratios, 95% CIs, and 2-sided P values. All analyses were performed using SAS statistical software, version 9.3.1 (SAS Institute Inc).

A formal review of the compounding process and distribution of fentanyl ensued. Investigators conducted surveys of the compounding facility, observed a mock demonstration of the fentanyl compounding process, and performed targeted environmental sampling of surfaces, solutions, and sink areas. All environmental cultures were processed in the Duke Clinical Microbiology Laboratory. Fluids were inoculated into blood culture bottles (BacTec Plus Aerobic/F; BD); surface swabs were streaked onto sheep blood agar (SBA) and B cepacia select agar and then placed into trypticase soy broth. Cultures were incubated at 35°C.

A sample of 23 randomly selected fentanyl syringes (10%) (10, 9, and 4 per each sequestered lot) were immediately sent to a US Food and Drug Administration–registered external reference laboratory for confirmatory sterility testing. The Duke Clinical Microbiology Laboratory also performed additional culturing of sequestered fentanyl samples in the 4 weeks after the outbreak.

The first attempt to isolate the outbreak pathogen included direct inoculation of 10 mL of fentanyl into a blood culture bottle and 1 mL onto SBA from a single syringe (3 syringes per lot). In the second attempt, 52 mL of fentanyl was pooled from 2 syringes into conical tubes and centrifuged (8, 6, and 2 per lot). The volume was decanted down to 3 mL, and 2.5 mL was inoculated into a blood culture bottle and 0.5 mL onto SBA. For the third attempt, the contents from 3 syringes from the same lot were batched together and vacuum filtered similar to methods used to isolate Legionella species from water⁸ and outlined by the USP.⁶ The shrink-wrap was removed from the exterior of 3 randomly selected syringes using a sterile technique. The contents of 3 syringes (78 mL) were batched together and vacuum filtered through a 0.22-µm filter. The filter was sterilely removed, placed into a 50-mL centrifuge tube containing 5 mL of Mueller Hinton II broth, and vortexed for 1 minute to free bacteria from the filter. A Mueller Hinton plate and B cepacia select agar plate were each inoculated with 150 μL of the broth, and the filter was placed directly onto a separate Mueller Hinton plate. As each syringe was emptied, the syringes and syringe caps were placed into a sterile 1-L flask containing Mueller Hinton II broth. The procedure was repeated for all 3 lots separately. All cultures were incubated at 35°C.

Seven cases were identified during 7 days from August 31 through September 6, 2012. Clustered BCC isolates demonstrated susceptibility to all antibiotics tested, including aminoglycosides. This unusual susceptibility pattern distinguished the outbreak pathogen from typical clinical isolates and suggested an environmental origin. Only 1 patient had bacteremia due to BCC in the preceding calendar year; this isolate’s drug susceptibilities differed significantly from the clustered isolates.

Case patients were located on 5 different intensive care and bone marrow transplant units (Table 1). All patients exhibited fever before incident blood culture; 3 of 7 (43%) had hypotension. All patients had more than 1 set of positive blood cultures; 4 of 7 had documented bacteremia on more than 1 calendar day. The median duration of hospitalization before the onset of bacteremia was 6 days (range, 4 hours to 24 days).

All 7 patients received compounded fentanyl via patient-controlled analgesia (PCA) or continuous infusion in the 24 hours preceding onset of bacteremia for the purposes of pain control or sedation while ventilated. During the outbreak period, compounded intravenous fentanyl was used for PCA or continuous infusion. In contrast, intravenous push fentanyl administered on an as-needed basis was obtained from manufactured vials. No other exposures common to all case patients were identified. Receipt of continuous-infusion intravenous fentanyl on day 0 was highly associated with the development of BCC bloodstream infection (odds ratio, 11.22; 95% CI, 2.09-∞; P = .01) (Table 2). The specific syringe number of fentanyl administered to patients was handwritten on the PCA flow sheet for 4 of 7 patients. All 4 of these patients received fentanyl from lot A within the 2 days before the incident blood culture collection.

Remaining fentanyl syringes were immediately removed from circulation on September 7, 2012, and the hospital secured an alternate manufactured supply. A total of 228 of 591 syringes (38.6%) from 3 lots were recovered and held for additional testing. No further cases of BCC were identified after fentanyl sequestration. All patients were treated with appropriate antibiotic therapy and demonstrated resolution of bacteremia. Two of 7 patients died before discharge, although their deaths were due to underlying conditions and not BCC bacteremia. The number of patients who received syringes from the implicated lot and the number of syringes per lot administered to each patient were unknown.

The usual compounding procedure for preparation of intravenous fentanyl from nonsterile raw materials was conducted on site in a clean room.⁶ Raw, nonsterile fentanyl powder was diluted into sterile water. The pH was adjusted using a pH probe and sodium hydroxide. To achieve a sterile end product, the 50-μg/mL fentanyl solution was filtered through a 0.22-µm filter into previously sterilized 26-mL syringes. Filled syringes were shrink-wrapped to ensure they remained tamper free. Lots of approximately 200 were produced in batches of 25. Every 26th syringe underwent sterility testing, which was performed on site by pharmacy staff using direct inoculation of trypticase soy broth. Exact volumes for each sterility test varied and were unknown. Sterility was determined by the clarity of the broth after a 14-day period of incubation; clear solution was deemed sterile while cloudy solution was deemed contaminated. Each lot was quarantined until results of sterility tests were known. No failed sterility tests were recognized before or during the outbreak period.

Syringes were transported from the compounding facility to a secured pharmacy storage area and then to dispensary machines on hospital units. Verification of syringe count at each transfer point was well documented; however, the final unit location of specific syringes or lots was unknown.

Investigators collected 35 environmental cultures from inside the clean room, anteroom, and glassware preparation room: 15 from solutions and 20 from surface swabs. Cultures taken from fluid used for pH probe calibration and the surface of the sink drain in the anteroom yielded BCC with identical susceptibility patterns as patients’ isolates (Table 3).

A mock demonstration of the compounding procedure was completed after resolution of the outbreak. Investigators noted possible mechanisms for contamination from surfaces, equipment, and gloves in the compounding room. These mechanisms included but were not limited to the following: gloves were not changed between compounding steps (instead gloves were disinfected using alcohol and then used to manipulate sterile items), the pH probe calibration fluid was not routinely changed, and the pH probe was submerged into the entire volume of the preparation instead of testing small aliquots at a time. Staff members expressed a belief that the final filtration step would be adequate to resolve possible contamination events that occurred before this step.

Sterility tests at the reference laboratory revealed no growth. Cultures in the Duke Clinical Microbiology Laboratory yielded no growth on the first 2 attempts. The BCC with pan-susceptibility was isolated from 1 of the 3 suspected lots using the batched filtration technique. Broth culture from the syringes and caps from this lot also yielded BCC.

Eleven outbreak-related BCC isolates from the 7 case patients, 2 environmental cultures, and 2 fentanyl cultures were analyzed at the Cystic Fibrosis Foundation’s Burkholderia cepacia Research Laboratory and Repository at the University of Michigan. Species-specific polymerase chain reaction and DNA sequencing of the recA gene^9,10 identified the isolates as B contaminans. Genotyping analyses that used repetitive element polymerase chain reaction and pulsed-field gel electrophoresis demonstrated that all 11 isolates were clonal (eFigure in the Supplement).

A multiunit outbreak of B contaminans bacteremia was caused by contaminated intravenous fentanyl prepared by an institutional compounding pharmacy. Rapid recognition of the uncommon pathogen by microbiology staff, prompt investigation by the infection prevention team, and a quick decision to sequester circulating compounded fentanyl led to termination of the outbreak within 24 hours of recognition. The outbreak pathogen was isolated from the compounding facility anteroom sink and pH probe calibration fluid sampled from the clean room. In addition, B contaminans was isolated from the fentanyl solution itself using a batched filtration method. Genetic identity was confirmed among patient, environmental, and medication isolates.

Burkholderia contaminans are gram-negative, aerobic, nonsporulating bacteria that have previously been isolated from water, soil, animals, industrial products, and contaminated pharmaceutical products.¹⁰Burkholderia contaminans was identified as a novel species in 2009 with use of multilocus sequence typing.^10,11 It has rarely been described as a human pathogen in cystic fibrosis,^12,13 cholangitis,¹⁴ and health care–associated outbreak-related infections of the lung, wound, and bloodstream.¹⁵ Although more than 50 different hospital outbreaks due to BCC species have been reported in the medical literature, only 1 due to B contaminans, specifically, has been reported.¹⁵ Previous sources of BCC outbreaks have included nonsterile pharmaceutical products, nasal spray,¹⁶ inhaled medications,¹⁷ chlorhexidine,¹⁸ diluted alcohol,¹⁹ alcohol-free mouthwash,²⁰ intravenous heparin,²¹ and bottle stoppers.²² The single other B contaminans outbreak was linked to prefabricated moist washcloths in a German university hospital system.¹⁵ The current B contaminans outbreak is unique because it involved intravenous medications prepared in an institutional compounding pharmacy with microbiologic and molecular proof of an environmental source. The uniqueness of this organism triggered the investigation. Outbreaks relating to compounding practice may go unnoticed if they are caused by more commonly encountered human pathogens that also survive in compounding facility environments (eg, Stenotrophomonas and Pseudomonas aeruginosa).

Compounding of sterile preparations holds inherent risks to patient safety. The highest-risk preparations involve nonsterile raw materials, many procedural steps, human hands, and multiple opportunities for errors and contamination.⁷ Compounded intravenous medications have been previously implicated in multiple highly publicized multistate outbreaks: Serratia marcescens in magnesium sulfate,²Sphingomonas paucimobilis in fentanyl,³ and, most recently, Exserohilum rostratum in corticosteroid preparations.¹ These outbreaks have brought attention to the collective need for more robust guidelines and regulation for compounding practice in large distributors. To our knowledge, there has been no formal evaluation delineating the scope of compounding practice limited to institutional pharmacies in the United States.²³ Moreover, the incidence of outbreaks related to small-scale compounding practice is unknown. In a voluntary survey of compounding pharmacies completed in April 2013, a total of 29% self-reported a patient incident that involved compounding errors during a 5-year period.²⁴ Ten percent reported recalls conducted for in-house preparations in 1 year.²⁴ Outbreaks of this type are underreported for several possible reasons: poor detection, limited effort or ability to fully investigate, medicolegal implications, or fear of damage to institutional reputation.

The Food and Drug Administration has suspended or closed several US compounding pharmacies because of safety concerns after the fungal meningitis outbreak.^1,25,26 These actions have required medical centers that relied on these facilities to seek alternate sources of supply, which may paradoxically include establishment or expansion of a local compounding pharmacy. In a January 2013 survey of 236 compounding pharmacies by the Office of the Inspector General, 56% of respondents reported that they were considering altering their sourcing practices for compounded medications.⁷ Fifty-four percent were planning on increasing hospital capacity to prepare compounded medications on site. This outbreak, however, is a reminder that institutional compounding pharmacies also must address the risk of contamination and potential harm to vulnerable, hospitalized populations.

This investigation identified multiple opportunities to improve future practice in the Duke Compounding Facility. These opportunities included consultation with an external expert in compounding pharmacy safety, ending contracts with consultants who had proven ineffective in execution of USP 797, revision of sterility testing protocols, new sterility testing equipment, reassignment of sterility testing and interpretation to dedicated microbiology staff, redesign and reconstruction of the clean room to ensure optimized air and workflow, staff retraining regarding concepts of sterility, and an intensive plan for routine peer review and audit.

Routine internal sterility testing, as well as sterility testing at an external reference laboratory, failed to detect the contaminant for this outbreak. We believe that inadequate volume of sampling and the insensitive direct inoculation method were the main causes of failed microbiologic detection. Routine internal sterility testing before the outbreak, as well as confirmatory testing at the external reference laboratory, did not fully follow USP guidelines due to inadequate volumes and proportion of media. Investigators had difficulty isolating the organism despite concerted efforts. Therefore, this contaminant may not have been identified even with the methods outlined by the USP. Possible theories explaining the failure of initial culturing methods include the presence of a growth inhibitor, sporadic distribution of the contaminant and inadequate sampling, or inadequate volumes.

This outbreak further supports the long-held principle that epidemiologic data can be highly reliable in the face of negative microbiologic test results. Initial negative culture results cast doubt among some investigators that fentanyl was the common source despite the convincing epidemiologic data. Furthermore, negative results from the external reference laboratory led to the initial conclusion that additional efforts at microbiologic testing would ultimately fail. We had difficulty finding external guidance for proper sampling of sequestered medication or meaningful assessment of confirmatory test validity. This experience highlights the need for reference laboratories and the USP to provide further recommendations in determining optimal methods for confirmatory sterility testing when contamination is suspected. A batched filtration approach may prove important for future outbreaks and should be considered early in the process given initial unsuccessful attempts using direct inoculation.

This investigation was limited in some ways. First, investigators were unable to fully track individual fentanyl lot or syringe numbers from pharmacy to patient and could not determine an accurate attack rate. This lack of information complicated the outbreak investigation and impaired identification of at-risk patients. Second, a robust analysis of host risk factors was not possible because of the small number of cases. Third, some cases may have gone undetected due to ascertainment problems related to blood culturing practices, concomitant antibiotic use, or poor recognition of nonspecific infectious symptoms that occur frequently in complex, hospitalized patient populations. Finally, the specific mechanism of contamination could not be definitively determined. A sporadic contamination event may have occurred as a result of contact with nonsterile gloves or the pH probe. There was no evidence to suggest diversion of fentanyl; however, the affected lot was not tested for dilution. Each syringe was shrink-wrapped, and no evidence of tampering or missing volume was detected. We cannot explain why the final filtration step failed to remove the contaminant. Human error or equipment malfunction during the filtration procedure may have gone undetected for the particular lot that caused the outbreak.

Despite these limitations, this experience can inform others who must address similar outbreaks and should serve as a warning for institutions considering a move to local compounding as part of routine practice. Medical institutions that produce high-risk compounded sterile preparations must be vigilant in efforts to prevent, detect, and terminate medication-related outbreaks. Such investigations should include a comprehensive epidemiologic analysis, the use of careful and determined microbiologic testing, and identification and remedy of practices that are not safe.

Accepted for Publication: October 23, 2013.

Corresponding Author: Rebekah W. Moehring, MD, MPH, Campus Box 102359, Duke University Medical Center, Durham, NC 27710 (Rebekah.moehring@duke.edu).

Author Contributions: Drs Moehring and Sexton had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Moehring, Schell, Thomann, Hazen, Chen, Sexton.

Acquisition of data: Moehring, Lewis, Isaacs, Schell, Thomann, Althaus, Hazen, LiPuma, Chen, Sexton.

Analysis and interpretation of data: Moehring, Lewis, Thomann, Hazen, Dicks, LiPuma, Chen, Sexton.

Drafting of the manuscript: Moehring, Lewis, Isaacs, Hazen, Sexton.

Critical revision of the manuscript for important intellectual content: Moehring, Lewis, Schell, Thomann, Althaus, Hazen, Dicks, LiPuma, Chen, Sexton.

Statistical analysis: Moehring.

Administrative, technical, or material support: Moehring, Isaacs, Schell, Thomann, Hazen, LiPuma, Chen, Sexton.

Study supervision: Moehring, Hazen, Chen, Sexton.

Published Online: February 3, 2014. doi:10.1001/jamainternmed.2013.13768.

Conflict of Interest Disclosures: None reported.

Previous Presentation: This investigation was previously presented at IDWeek; October 4, 2013; San Francisco, California.

Additional Contributions: The authors sincerely thank the Duke University Hospital infection preventionists, nursing staff, and unit leaders. Special thanks to Duke Clinical Microbiology Laboratory staff for their recognition of the outbreak, dedicated efforts, and extra hours during the investigation. Thank you to the many pharmacy staff who participated in and contributed to this investigation. Thank you to our hospital leadership, pharmacy leadership, and risk management team. We thank Alexander Kallen, MD, MPH (Centers for Disease Control and Prevention, Atlanta, Georgia), and Zack Moore, MD, MPH (North Carolina Department of Health, Raleigh), for their advice during the investigation. These individuals did not receive compensation for their help.