Journal of Hospital Infection xxx (2015) 1e13

Available online at www.sciencedirect.com

Journal of Hospital Infection

journal homepage: www.elsevierhealth.com/journals/jhin

Systematic review and meta-analysis of the risk of

microbial contamination of parenteral doses prepared
under aseptic techniques in clinical and
pharmaceutical environments: an update
P.D. Austin a, b, *, K.S. Hand b, c, M. Elia a, d, e
a
Faculty of Medicine, University of Southampton, Southampton, UK
b
Southampton Pharmacy Research Centre, University Hospital Southampton NHS Foundation Trust, Southampton, UK
c
Faculty of Health Sciences, University of Southampton, Southampton, UK
d
Institute of Human Nutrition, University of Southampton, Southampton, UK
e
National Institute for Health Research, Southampton Biomedical Research Centre, University Hospital Southampton NHS
Foundation Trust, Southampton, UK

A R T I C L E I N F O S U M M A R Y

Article history: Background: Administration of parenteral doses with microbial contamination can lead to
Received 9 February 2015 infective morbidity or death.
Accepted 2 April 2015 Aim: To test whether aseptic preparation of parenteral doses or additives to sterile doses
Available online xxx undertaken in dedicated pharmaceutical rather than clinical environments reduces the
risk of microbial dose contamination.
Keywords: Methods: Data identified from a systematic review were examined using random effects
Aseptic meta-analyses, and t-tests were used to compare dose contamination frequencies.
Batch Findings: In all, 16,552 doses from 34 studies (33 records) were identified. For all the data
Contamination combined there was a significantly higher frequency of contamination of doses prepared in
Environment clinical than in pharmaceutical environments {3.7% [95% confidence interval (CI): 2.2, 6.2;
Individual lots N ¼ 10,272 doses] vs 0.5% (95% CI: 0.1, 1.6; N ¼ 6280 doses); P ¼ 0.007}. Contamination of
doses was significantly higher when prepared as individual lots than as part of a batch in
pharmaceutical environments [2.1% (95% CI: 0.7, 5.8; N ¼ 168 doses) vs 0.2% (95% CI: 0.1, 0.9;
N ¼ 6112 doses); P ¼ 0.002]. There was a significantly higher frequency of dose contamination
if additions were made to sterile parenteral doses in clinical environments [risk ratio: 2.121
(95% CI: 1.093, 4.114); P ¼ 0.026]. The overall quality of the studies was judged to be low.
Conclusion: Reported rates of parenteral dose contamination were orders of magnitude
higher than accepted reference standards, which may increase infection risk. The limited
evidence on contamination rates supports dose preparation in pharmaceutical rather than
clinical environments, and does not support batch preparation in clinical environments.
ª 2015 The Healthcare Infection Society. Published by Elsevier Ltd. All rights reserved.

* Corresponding author. Address: Pharmacy Department, South-

ampton General Hospital, Tremona Road, Southampton SO16 6YD, UK.
Tel.: þ44 (0) 23 80 777 222 and ask for bleep 1361.
E-mail address: [email protected] (P.D. Austin).

http://dx.doi.org/10.1016/j.jhin.2015.04.007
0195-6701/ª 2015 The Healthcare Infection Society. Published by Elsevier Ltd. All rights reserved.

Please cite this article in press as: Austin PD, et al., Systematic review and meta-analysis of the risk of microbial contamination of parenteral
doses prepared under aseptic techniques in clinical and pharmaceutical environments: an update, Journal of Hospital Infection (2015), http://
dx.doi.org/10.1016/j.jhin.2015.04.007
2 P.D. Austin et al. / Journal of Hospital Infection xxx (2015) 1e13

Introduction further surgery due to endophthalmitis as a consequence of

contaminated intravitreal injections in the USA; an outbreak of
Administration of a parenteral dose with microbial contam- bloodstream infections requiring withdrawal of relevant stock
ination may result in infective morbidity and death. Recent due to contaminated intravenous analgesia in Taiwan; and
examples include: postoperative sepsis after inadequate deaths in newborns as a consequence of contaminated paren-
aseptic handling of intravenous anaesthetic; loss of vision or teral nutrition in France and the UK.1e4 This means it is

Table I
The terms and number of results for the literature search undertaken on February 10th, 2014 to identify parenteral doses prepared under
aseptic techniques in clinical and pharmaceutical environments
Database Search row Search terms Results (N)
Medline (OvidSP) 1 (syringe or syringes).mp. or syring*.tw. 21,041
2 (bag or bags).mp. or bag*.tw. 18,909
3 (infusion or infusions).mp. or infus*.tw. 258,728
4 (vial or vials).mp. or vial*.tw. 6066
5 (microbial or microbiological).mp. or micro*.tw. 2,017,452
6 (bacterium or bacteria).mp. or bact*.tw. 598,873
7 (fungus or fungi).mp. or fung*.tw. 134,157
8 (contaminated or contamination).mp. or contam*.tw. 166,465
9 prepared.mp. or prep*.tw. or manufactured.mp. or manuf*.tw. or 1,110,191
compounded.mp. or compound*.tw.
10 (1 and 5) or (1 and 6) or (1 and 7) or (1 and 8) or (2 and 5) or (2 and 6) or 19,123
(2 and 7) or (2 and 8) or (3 and 5 and 9b) or (3 and 6 and 9c) or (3 and 7) or
(3 and 8) or (4 and 5) or (4 and 6) or (4 and 7) or (4 and 8)
11 limit 10 to English language 17,662
Embase Classic 1 (syringe or syringes).mp. or syring*.tw. 32,435
and Embase (OvidSP)
2 (bag or bags).mp. or bag*.tw. 30,560
3 (infusion or infusions).mp. or infus*.tw. 352,153
4 (vial or vials).mp. or vial*.tw. 10,079
5 (microbial or microbiological).mp. or micro*.tw. 2,221,231
6 (bacterium or bacteria).mp. or bact*.tw. 952,057
7 (fungus or fungi).mp. or fung*.tw. 252,530
8 (contaminated or contamination).mp. or contam*.tw. 253,438
9 prepared.mp. or prep*.tw. or manufactured.mp. or manuf*.tw. or 1,697,262
compounded.mp. or compound*.tw.
10 (1 and 5) or (1 and 6) or (1 and 7) or (1 and 8) or (2 and 5) or (2 and 6) or 23,099
(2 and 7) or (2 and 8) or (3 and 5 and 9b) or (3 and 6 and 9c) or (3 and 7) or
(3 and 8) or (4 and 5) or (4 and 6) or (4 and 7) or (4 and 8)
11 limit 10 to English language 20,824
The Cochrane Library #1 “syringe” or “syringes” or syring* 1364
(Wiley Online Library)a
#2 “bag” or “bags” or bag* 4467
#3 “infusion” or “infusions” or infus* 36,233
#4 “vial” or “vials” or vial* 1325
#5 “microbial” or “microbiological” or micro* 66,334
#6 “bacterium” or “bacteria” or bact* 25856
#7 “fungus” or “fungi” or fung* 2759
#8 “contaminated” or “contamination” or contam* 3453
#9 “prepared” or prep* or “manufactured” or manuf* or “compounded” or 64,682
compound*
#10 (#1 and #5) or (#1 and #6) or (#1 and #7) or (#1 and #8) or (#2 and #5) or 3760
(#2 and #6) or (#2 and #7) or (#2 and #8) or (#3 and #5 and #9b) or (#3 and
#6c) or (#3 and #7) or (#3 and #8) or (#4 and #5) or (#4 and #6) or (#4 and
#7) or (#4 and #8)
a
All document search.
b
The combination of search terms 3 and 5 yielded 57,265 results in Medline, 33,340 results in Embase, and 7464 results in the Cochrane
Library, and 4848, 671, and 876 results respectively when search term 9 was also included in the combination.
c
The combination of search terms 3 and 6 returned 6363 results in Medline and 9036 results in Embase, and 689 and 784 results respectively
when search term 9 was included in the combination. The third search term was not required for the combination of search terms 3 and 6 in the
Cochrane Library since 1565 results were returned.

Please cite this article in press as: Austin PD, et al., Systematic review and meta-analysis of the risk of microbial contamination of parenteral
doses prepared under aseptic techniques in clinical and pharmaceutical environments: an update, Journal of Hospital Infection (2015), http://
dx.doi.org/10.1016/j.jhin.2015.04.007
 P.D. Austin et al. / Journal of Hospital Infection xxx (2015) 1e13 3
important to implement safe procedures in routine practice to search term, truncated search terms, and combination of three
prevent inadvertent microbial dose contamination. For search terms only if more than 5000 results were returned for
example, the risk of contamination is expected to be lower any two search term combination. Three databases were used
when procedures are undertaken in an environment with a low for all available years: Medline from 1946 onwards using
density of microbes than one with a high density. Therefore, it is OvidSP; Embase from 1947 onwards using OvidSP; and the
often recommended to move aseptic preparation of parenteral Cochrane Library. Attempts were made to identify further
doses away from a clinical environment (with a higher density of papers by hand searching.
microbes) into a specially designed pharmaceutical environ- The literature search included studies that involved microbial
ment [with a lower density of microbes (and particulates)] in contamination with bacteria and/or fungi. The studies involved
line with recognized standards operating in countries such as preparation of doses for parenteral administration to patients
the USA or the UK.5e7 For example, in the immediate area used prepared under aseptic techniques, including simulation
to prepare parenteral medicines there could be more than 90 studies. Studies were excluded if they were not reported in the
times the number of colony-forming units falling on to a 90 mm English language, if they only involved animals, or if they re-
diameter trypticase soy agar plate in a 4 h period in a clinical ported the rate of contamination of infusate stock (an infusate in
environment than is allowed by the standards applied to phar- a single container used to prepare multiple doses) rather than
maceutical environments in some countries.7,8 The use of a actual or simulated prepared doses (for example, contamination
pharmaceutical environment is particularly important in the of multi-dose vials after repeated use). Studies were also
preparation of batch doses, which carry the risk of contami- excluded if they involved the use of blood or a blood component,
nating multiple lots of individual doses, and where there is likely if there was freezing/thawing of prepared doses, or if there was
to be a period of storage before administration to patients. reuse of equipment during dose preparation (except when used
However, since pharmaceutical environments meeting recog- in the preparation of a single batch). For an environment to
nized standards for aseptic dose preparation are costly and qualify as a pharmaceutical environment the recognized stan-
require operational expertise that may not always be readily dard of the cabinet in which the doses were prepared and the
available, they are not always used. Therefore, there is a need room in which that cabinet was situated had to be specified in the
to balance the advantages and disadvantages of the procedures record (journal article). When a single record reported more
and the environments in which they are undertaken in order to than one outcome, for example when using different prepara-
obtain the desirable effects in routine clinical practice. tion environments, each outcome was included as a separate
In 2009 we published a systematic review with meta-analysis study. Consistent data within the same record were combined
to summarize published frequencies of contamination of only if whole groups of data could be combined.
parenteral doses prepared in clinical and pharmaceutical en- The search terms (including variations and truncated terms)
vironments under aseptic techniques.9 This provided some ev- and number of results are shown in Table I. In brief, each of
idence favouring dose preparation in the pharmaceutical four search terms was combined with each of four further
environment, but the conclusions were weakened by the small search terms, unless a combination returned more than 5000
number of studies which were generally of low quality. It is results, in which case a third search term was added in an
possible that some earlier studies may have been missed attempt to capture the most relevant results. It can be seen
because the initial review used only one database search engine from Table I that a third search term was required on five oc-
(PubMed from 1947 onwards). Since our review, a considerable casions. Additional papers were sought through cross-
amount of new information has become available which needs referencing and discussion with experts in the field.
to be incorporated into the analyses, and clinical concern about The literature search identified 42,246 records (17,662 from
methods to reduce morbidity and cost due to infections has Medline, 20,824 from Embase and 3760 from the Cochrane Li-
been increasing. For example, an international initiative has brary) and 28,020 after duplicates had been removed. The title
sought to rationalize and harmonize standards for aseptic and abstract (if necessary and accessible) of each of the 28,020
preparation of parenteral doses throughout Europe.10,11 It is identified records was evaluated and excluded if it did not meet
clear that the existing evidence base needs to be reviewed, the above inclusion criteria. This left 137 records, which were
updated, and clarified by providing a more precise definition of individually subjected to a full text review to confirm relevance
the pharmaceutical environment.9 Therefore, the aim of this and compliance with the above criteria to yield a final total of 34
study was to clarify and extend the evidence base to address the studies from 33 records.8,12e43 Each of the final 19 studies from
following three hypotheses: one, the risk of infective contam- 17 records12,14e16,18e22,24,33,36,43e47 identified in our 2009
ination is different for aseptic preparation in a clinical and search9 were identified in the present search but five of those
pharmaceutical environment; two, the risk is also different for studies from four records were excluded due to inadequate and/
aseptic preparation of individual and batch doses within the or inadequately described pharmaceutical environments. The
same type of environment; and three, the risk is different for methodological stages of the search are shown in Figure 1. As
additives rather than no additives to sterile doses prior to previously,9 the included studies were divided into groups ac-
administration. We also sought to consider future research cording to whether doses were prepared in a clinical or phar-
needs in light of the current evidence base. maceutical environment (hypothesis 1), whether doses had been
prepared as individual lots or as part of a batch (hypothesis 2),
and whether doses had been sampled without or before admin-
Methods istration or during or after administration to a patient (due to a
risk of contamination from manipulations after preparation and
The literature search was undertaken on 10 February 2014 potential differences in time between preparation and sampling
with a wider protocol than that undertaken by the previous which may have affected recovery of damaged microbial cells).
review.9 The present literature search used an additional Doses were considered to be either contaminated or not

Please cite this article in press as: Austin PD, et al., Systematic review and meta-analysis of the risk of microbial contamination of parenteral
doses prepared under aseptic techniques in clinical and pharmaceutical environments: an update, Journal of Hospital Infection (2015), http://
dx.doi.org/10.1016/j.jhin.2015.04.007
4 P.D. Austin et al. / Journal of Hospital Infection xxx (2015) 1e13

Identification

42,246 records identified through 0 additional records identified

database searching through other sources

28,020 records after duplicates removed

Screening

28,020 records screened 27,883 records excluded

Eligibility
137 full-text records 104 full-text records
assessed for eligibility excluded primarily
because of:
Inadequate description of
33 records (34 studies) environment,
included in qualitative methodology of results
synthesis (N = 18)
Included Lack of relevance
(N = 15)
Lack of relevant data
34 studies included in (N = 15)
quantitative synthesis (a Reuse of equipment or
series of meta-analyses). infusate
These included event rate (N = 16)
summaries of single arm Unacceptable or
studies and a meta-analysis inconsistent environment
of six studies with a control (N = 14)
group (two arms) Case report or
commentary
(N = 14)
Inconsistent methodology
(N = 9)
Deliberate contamination
of infusate or component
(N = 3)

Figure 1. The methodological stages of the literature search used to identify studies that report the rate of contamination of doses
prepared under aseptic techniques in clinical and pharmaceutical environments.

contaminated without any attempt to identify the density of any obtained by logarithmic (logit) transformation. When there was
micro-organisms present; the types of micro-organisms, where zero contamination in a group, a value of 0.5 contaminated
these were reported, are briefly summarized. doses was used to overcome the mathematical difficulties
Two of the authors (P.D.A. and K.S.H.) independently assessed associated with logarithmic transformation (the log10 of zero is
the quality of the included studies using the GRADE system with minus infinity). Data amalgamation and the meta-analyses
subsequent discussion to resolve any disagreement.48,49 The were undertaken using a random effects model and the soft-
recommendations of the UK National Health Service Centre for ware Comprehensive Meta Analysis version 2 (Biostat, Engle-
Reviews and Dissemination (CRD) and Cochrane as well as the wood, NJ, USA). One-group meta-analyses were used for
PRISMA guidelines for reporting systematic reviews were hypotheses 1 and 2 and a two-group meta-analysis was used for
considered at all stages during this review.50e52 hypothesis 3 due to the nature of the available studies. The
random effects model was chosen because of the clinical het-
Statistical analysis erogeneity of the studies, but the I2 statistic is also presented.
Comparisons between group means were undertaken using
The point estimate, standard error and 95% confidence in- unpaired t-tests, with a two-tailed P < 0.05 considered sta-
terval for the contamination rate of each separate group was tistically significant.

Please cite this article in press as: Austin PD, et al., Systematic review and meta-analysis of the risk of microbial contamination of parenteral
doses prepared under aseptic techniques in clinical and pharmaceutical environments: an update, Journal of Hospital Infection (2015), http://
dx.doi.org/10.1016/j.jhin.2015.04.007
 Table II
dx.doi.org/10.1016/j.jhin.2015.04.007
doses prepared under aseptic techniques in clinical and pharmaceutical environments: an update, Journal of Hospital Infection (2015), http://
Please cite this article in press as: Austin PD, et al., Systematic review and meta-analysis of the risk of microbial contamination of parenteral

Summary of studies that reported the frequency of microbial contamination of parenteral doses prepared under aseptic techniques in clinical and pharmaceutical environments
Study Country Dose Individual or Preparation Administration Additives/ Control group
batch preparation environmenta to patientsb repackaging group (no additives)
(or treated as
Total Contaminated Total Contaminated
one or the other)
doses (N) doses (N) doses (N) doses (N)
Austin et al.12 England Growth medium Batch Pharmaceutical No 1002 0 e e
Austin et al.8 England Growth medium Batch Clinical No 778c 19c e e
Aydin et al.13 Turkey Propofol with and Batch Clinical No 1920d 1d e e
without lidocaine
Bach et al.14 Germany Anaesthetic agents Individual Clinical Yes 1228* 47 e e

P.D. Austin et al. / Journal of Hospital Infection xxx (2015) 1e13

Breheny et al.15 Australia Parenteral nutrition Individual Clinical Yes 150* 0 96 0
Burke et al.16 USA 5% w/v glucose Batch Clinical No 95 27 e e
Choy et al.17 USA Various, including Individual Pharmaceutical No 150e 3e e e
parenteral nutrition
D’Arcy et al.18 Ireland Various Individual Clinical Yes 61* 34 40 5
Dominik et al.19 Germany Contrast media Batch Clinical No 1000 9 e e
Driver et al.20 USA Various for obstetric Batch Clinical No 756 0 e e
theatre use
Ernerot et al.21,y Sweden Various Batch Clinical No 50 0 e e
Ernerot et al.21,y Sweden Various Individual Clinical Yes 131* 3 40 2
Farrington et al.22 England Midazolam or propofol Individual Clinical Yes 100* 7 e e
Fleer et al.23 The Netherlands Parenteral nutrition Batch Clinical Nof 428 81 e e
Hernandez-Ramos Mexico Various Individual Clinical Yes 1011* 60 e e
et al.24
Jackson and Gallo25 USA Insulin Batch Clinical No 159g 0
Jacobson et al.26 USA Filgrastim (G-CSF) Batch Pharmaceutical No 60h 0h e e
Khalili et al.27 Iran Crystalloid fluids Individual Clinicali No 92j 1j e e
Kundsin et al.28 USA Not stated Individual Clinical Yesk 432* 5 247 1
Letcher et al.29 USA ‘Medications’ Individual Clinical Yes 224m,* 13m 142m 5m
Lorenz et al.30 Austria Propofol Individual Clinical Nof 80n 5n e e
Macias et al.31 Mexico Variousp Individual Clinical Yes 101* 8 e e
Madeo et al.32 England Epidural analgesia Individual Clinical Yes 46q,* 3q e e
Magee et al.33 England Anaesthetics, Batch Clinical No 195 8 e e
0.9% w/v sodium
chloride and
growth medium
Micard et al.34 France Meglumine gadoterate Batch Pharmaceutical No 20 0 e e
Poretz et al.35 USA 0.9% w/v sodium Individual Clinicali Yesj 110* 10 50 2
chloride in
5% w/v glucose in
Ringer’s lactate
Soong36 Australia Propofol Individual Clinical Yes 5* 0 e e
Spiliotis et al.37 Greece Parenteral nutrition Individual Clinical No 80r 0r e e
(continued on next page)

5
 6
Table II (continued )
dx.doi.org/10.1016/j.jhin.2015.04.007
doses prepared under aseptic techniques in clinical and pharmaceutical environments: an update, Journal of Hospital Infection (2015), http://
Please cite this article in press as: Austin PD, et al., Systematic review and meta-analysis of the risk of microbial contamination of parenteral

Study Country Dose Individual or Preparation Administration Additives/ Control group

batch preparation environmenta to patientsb repackaging group (no additives)
(or treated as
Total Contaminated Total Contaminated
one or the other)
doses (N) doses (N) doses (N) doses (N)
Stjernstrom et al.38 Sweden ‘Saline’ and Batch Clinical No 100 0 e e
growth medium
Thomas et al.39 USA Growth medium Batch Pharmaceutical No 2030s 7s e e
Urbano et al.40 Japan Growth medium Individual Pharmaceutical No 18 0 e e
van Doorne et al.41 The Netherlands Growth medium Batch Pharmaceuticalt No 3000 1 e e
van Grafhorst et al.42 The Netherlands Growth medium Batch Clinicali No 650 151 e e

P.D. Austin et al. / Journal of Hospital Infection xxx (2015) 1e13

Yorioka et al.43 Japan Electrolytes and Individual Clinical Yes 290* 0 e e
dobutamine
* Asterisks indicate doses that were sampled during or after administration, and the absence of an asterisk indicates doses that were sampled without or prior to administration.
y
Different aspects examined within the same record.
a
A clinical environment includes hospital wards or operating theatres; to be classified as a pharmaceutical environment, the record must state compliance with a recognized standard
for both the preparation cabinet and immediate room surrounding that cabinet environment.
b
‘Yes’ if the doses were sampled during or after administration and ‘no’ if the doses were sampled without or prior to administration.
c
In this study, 19 of 276 doses prepared by nurses were contaminated; zero of 502 doses prepared by a pharmacy operator were contaminated.
d
The data from part 2 of this study have been excluded since they involved unacceptable methodology (a delay in drawing up the dose).
e
Of the 150 prepared doses, 52 were parenteral nutrition and all of the three contaminated doses were parenteral nutrition.
f
These doses were administered to patients after they had been sampled.
g
This record reports that one additional prepared dose was misplaced and not tested.
h
Includes only those doses prepared in a standardized pharmaceutical environment.
i
Only the data from a clinical environment are included since the nature of the pharmaceutical environment used is unacceptable/unclear.
j
Excludes data from vial residues.
k
Simulated patient administration.
m
The data reported from the containers rather than from the associated giving sets.
n
The data from sample 2 of group I have been excluded since they represented the same doses, and data from group II have been excluded due to unacceptable conditions.
p
Study design excluded patients receiving electrolytes, antibiotics or cancer chemotherapy.
q
Data reported from cases without reuse of administration sets.
r
Data from sampling immediately after dose preparation, not those same doses sampled after infusion (when three contaminated samples were identified).
s
Data from standardized pharmaceutical conditions since the environment used for the negative control doses is unclear.
t
Excludes the doses prepared in uncontrolled pharmaceutical environments.
 P.D. Austin et al. / Journal of Hospital Infection xxx (2015) 1e13 7

Event Lower Upper

Study name rate limit limit Event rate and 95% CI

Austin et al. 2013 0.024 0.016 0.038

Aydin et al. 2002 0.001 0.000 0.004
Burke et al. 1986 0.284 0.203 0.383
Dominik et al. 1995 0.009 0.005 0.017
Driver et al. 1998 0.001 0.000 0.010
Ernerot et al. 1973 0.010 0.001 0.138
Fleer et al. 1983 0.189 0.155 0.229
Jackson et al. 1990 0.003 0.000 0.048
Magee et al. 1995 0.041 0.021 0.080
Stjernstrom et al. 1978 0.005 0.000 0.074
van Grafhorst et al. 2002 0.232 0.201 0.266
Batch doses in clinical environments 0.027 0.011 0.062
Bach et al. 1997* 0.038 0.029 0.051
Breheny et al. 1990* 0.003 0.000 0.051
D'Arcy et al. 1973* 0.557 0.432 0.676
Ernerot et al. 1973* 0.023 0.007 0.069
Farrington et al. 1994* 0.070 0.034 0.140
Hernandez-Ramos et al. 2000* 0.059 0.046 0.076
Khalili et al. 2013 0.011 0.002 0.073
Kundsin et al. 1973* 0.012 0.005 0.027
Letcher et al. 1972* 0.058 0.034 0.097
Lorenz et al. 2002 0.063 0.026 0.142
Macias et al. 2012* 0.079 0.040 0.150
Madeo et al. 1999* 0.065 0.021 0.184
Poretz et al. 1974* 0.091 0.050 0.161
Soong 1999* 0.083 0.005 0.622
Spiliotis et al. 1989 0.006 0.000 0.091
Yorioka et al. 2006* 0.002 0.000 0.027
Individual doses in clinical environments 0.047 0.025 0.084
Austin et al. 2006 0.000 0.000 0.008
Jacobson et al. 1996 0.008 0.001 0.118
Micard et al. 2001 0.024 0.001 0.287
Thomas et al. 2005 0.003 0.002 0.007
van Doorne et al. 1994 0.000 0.000 0.002
Batch doses in pharmaceutical environments 0.002 0.001 0.009
Choy et al. 1982 0.020 0.006 0.060
Urbano et al. 2013 0.026 0.002 0.310
Individual doses in pharmaceutical environments 0.021 0.007 0.058
0.00 0.50 1.00

Figure 2. Forest plot and summary statistics of the frequency of the contamination rates of parenteral doses prepared aseptically in
clinical and pharmaceutical environments. Asterisks indicate doses that were sampled during or after administration, and the absence of
an asterisk indicates doses that were sampled without or prior to administration. CI, confidence interval.

Results low quality, primarily due to small sample size34,36,40 and

limited procedural detail and high contamination rate.18
Quality of studies
Overview of the rate of contamination of doses
For the purpose of this review, both raters graded all of the prepared in clinical and pharmaceutical environments
included studies as low to very low quality, with three dis-
agreements within these categories. After discussion, the ma- A grand total of 16,552 doses eligible for inclusion were
jority of the studies were graded as low quality primarily because identified from 34 studies taken from 33 records, which are
they were non-randomized, and four studies were graded as very summarized in Table II.8,12e43 The single record12 identified

Please cite this article in press as: Austin PD, et al., Systematic review and meta-analysis of the risk of microbial contamination of parenteral
doses prepared under aseptic techniques in clinical and pharmaceutical environments: an update, Journal of Hospital Infection (2015), http://
dx.doi.org/10.1016/j.jhin.2015.04.007
8 P.D. Austin et al. / Journal of Hospital Infection xxx (2015) 1e13
through other sources in our previous review9 was identified by seven studies (seven records12,17,26,34,39e41) had been prepared
the present literature search. Excluding control groups, this in pharmaceutical environments. When all the data were
represents an increase of 133% in the number of doses (16,552 combined, there was a significantly higher frequency of
vs 7101), 79% in the number of studies (34 vs 19), and 94% in the contamination of doses prepared in clinical than in pharma-
number of records (33 vs 17) from the 2009 review.9 If the five ceutical environments [2.5% (95% CI: 1.2, 5.5; n ¼ 6383 doses)
studies not meeting the inclusion criteria of the present review (I2 ¼ 95.69%; P < 0.001) vs 0.5% (95% CI: 0.1, 1.6; n ¼ 6280
are withdrawn from the first review, there is an increase of doses) (I2 ¼ 69.18%; P ¼ 0.003); P ¼ 0.044]. The between-study
173% in the number of doses (16,552 vs 6074), 143% in the contamination was more variable in the clinical than in the
number of studies (34 vs 14), and 154% in the number of records pharmaceutical environment (range: 0.1e28.4% vs 0.0e2.6%
(33 vs 13). respectively).
Of the total 33 records, only seven involved head-to-head Doses sampled during or after administration
comparisons. One record21 compared batch doses and individ- It was not possible to compare doses prepared in clinical and
ual lots in a clinical environment and six records15,18,21,28,29,35 pharmaceutical environments that had been sampled during or
compared additives and no additives to sterile doses in a clin- after administration due to lack of data in the pharmaceutical
ical environment. environment.
Figure 2 shows the forest plot obtained when all the study Individual doses
data were combined in a meta-analysis grouped according to All identified doses. The analysis involved 4309 doses from
environment (pharmaceutical or clinical) and type of dose 18 studies (18 records14,15,17,18,21,22,24,27e32,35e37,40,43). Of
preparation (individual or batch). The majority (94%) of the these, 4141 doses from 16 studies (16 records14,15,18,21,
22,24,27e32,35e37,43
doses that had been prepared as individual lots in clinical en- ) had been prepared in clinical environments
vironments (N ¼ 4141) had been sampled during or after and 168 doses from two studies (two records17,40) had been
administration, and all of the other doses had been sampled prepared in pharmaceutical environments. When all the data
without or prior to administration. When only the 4141 doses were combined, there was a non-significantly higher frequency
prepared as individual lots in clinical environments were of contamination of doses prepared in clinical than in phar-
included there was a non-significantly higher frequency of maceutical environments [4.7% (95% CI: 2.5, 8.4; N ¼ 4141
contamination of doses sampled without or prior to adminis- doses) (I2 ¼ 91.64%; P < 0.001) vs 2.1% (95% CI: 0.7, 5.8; N ¼ 168
tration than during or after administration [5.3% (95% CI: 2.7, doses) (I2 ¼ 00.00%; P ¼ 0.856); P ¼ 0.190]. The between-study
10.0; N ¼ 3889 doses) (I2 ¼ 93.07%; P < 0.001) vs 2.3% (95% CI: contamination was more variable in the clinical than in the
0.5, 10.1; N ¼ 252 doses) (I2 ¼ 56.45%; P ¼ 0.101); P ¼ 0.314]. pharmaceutical environment (range: 0.2e55.7% vs 2.0e2.6%
Nineteen of the 22 studies with contamination reported the respectively).
type of microbe.8,13,14,16e19,21e24,27e33,35,39,41,42 In pharma- Doses sampled without administration or prior to adminis-
ceutical environments this was limited to coagulase-negative tration. There were 420 doses from five studies (five re-
staphylococci (including Staphylococcus epidermidis), Bacillus cords17,27,30,37,40) that had been sampled without administration
spp., and Propionibacterium spp.17,39 The same microbes were or prior to administration, of which 252 doses from three studies
identified in clinical environments, where more pathogenic (three records27,30,37) had been prepared in clinical environ-
microbes were also found, including Staphylococcus aureus, ments and 168 doses from two studies (two records17,40) had
Serratia marcescens, Klebsiella spp., Enterobacter spp., and been prepared in pharmaceutical environments. When all the
fungi (including Candida spp.).13,14,16,19,21e24,27,29e31,33,35,42 data were combined there was a non-significantly higher fre-
quency of contamination of doses prepared in clinical than in
Hypothesis 1. Dose preparation in a clinical compared to a
pharmaceutical environments [2.3% (95% CI: 0.5, 10.1; N ¼ 252
pharmaceutical environment
doses) (I2 ¼ 56.45%; P ¼ 0.101) vs 2.1% (95% CI: 0.7, 5.8; N ¼ 168
Individual and batch doses combined
doses) (I2 ¼ 00.00%; P ¼ 0.856); P ¼ 0.923]. The between-study
All identified doses. The analysis involved 16,552 doses from
contamination was more variable in the clinical than in the
34 studies (33 records8,12e43). Of these, 10,272 doses from 27
pharmaceutical environment (range: 0.6e6.3% vs 2.0e2.6%
studies (26 records8,13e16,18e25,27e33,35e38,42,43) had been pre-
respectively).
pared in clinical environments and 6280 doses from seven
Doses sampled during or after administration. It was not
studies (seven records12,17,26,34,39e41) had been prepared in
possible to compare doses prepared in clinical and pharma-
pharmaceutical environments. When all the data were com-
ceutical environments that had been sampled during or after
bined there was a significantly higher frequency of contami-
administration due to lack of data in the pharmaceutical
nation of doses prepared in clinical than in pharmaceutical
environment.
environments [3.7% (95% CI: 2.2, 6.2; N ¼ 10,272 doses)
Batch doses
(I2 ¼ 95.35%; P < 0.001) vs 0.5% (95% CI: 0.1, 1.6; N ¼ 6280
All identified doses. The analysis involved 12,243 doses from
doses) (I2 ¼ 69.18%; P ¼ 0.003); P ¼ 0.007]. The between-study
16 studies (16 records8,12,13,16,19e21,23,25,26,33,34,38,39,41,42). Of
contamination was more variable in the clinical than in the
these, 6131 doses from 11 studies (11 records8,13,16,19e21,23,25,
pharmaceutical environment (range: 0.1e55.7% vs 0.0e2.6% 33,38,42
) had been prepared in clinical environments and 6112
respectively).
doses from five studies (five records12,26,34,39,41) had been
Doses sampled without or prior to administration. There
prepared in pharmaceutical environments. When all the data
were 12,663 doses from 21 studies (21 records8,12,13,16,17,
19e21,23,25e27,30,33,34,37e42 were combined, there was a significantly higher frequency of
) that had been sampled without
contamination of doses prepared in clinical than in pharma-
administration or prior to administration, of which 6383 doses
ceutical environments [point estimate: 2.7% (95% CI: 1.1, 6.2;
from 14 studies (14 records8,13,16,19e21,23,25,27,30,33,37,38,42) had
N ¼ 6131 doses) (I2 ¼ 96.48%; P < 0.001) vs 0.2% (95% CI: 0.1,
been prepared in clinical environments and 6280 doses from
0.9; N ¼ 6112 doses) (I2 ¼ 56.49%; P ¼ 0.056); P < 0.001]. The

Please cite this article in press as: Austin PD, et al., Systematic review and meta-analysis of the risk of microbial contamination of parenteral
doses prepared under aseptic techniques in clinical and pharmaceutical environments: an update, Journal of Hospital Infection (2015), http://
dx.doi.org/10.1016/j.jhin.2015.04.007
 P.D. Austin et al. / Journal of Hospital Infection xxx (2015) 1e13 9
between-study contamination was more variable in the clinical P < 0.001) vs 2.7% (95% CI: 1.1%, 6.2%; N ¼ 6131 doses)
than in the pharmaceutical environment (range: 0.1e28.4% vs (I2 ¼ 96.48%; P < 0.001); P ¼ 0.299]. The between-study
0.0e2.4% respectively). contamination was more variable for doses prepared as indi-
Doses sampled without or prior to administration. Since all vidual lots than as part of a batch (range: 0.2e55.7% vs
of the identified doses had been sampled without or prior to 0.1e28.4% respectively).
administration, a comparison of doses prepared in clinical and Doses sampled without or prior to administration. There
pharmaceutical environments that had been sampled without were 6383 doses from 14 studies (14 records8,13,16,19e21,23,25,
27,30,33,37,38,42
or prior to administration yields the same results as all of the ) that had been sampled without or prior to
data combined (above). administration, of which 252 doses from three studies (three
Doses sampled during or after administration. It was not records27,30,37) had been prepared as individual lots and 6131
possible to compare doses prepared in clinical and pharma- doses from 11 studies (11 records8,13,16,19e21,23,25,33,38,42) had
ceutical environments that had been sampled during or after been prepared as part of a batch. When all the data were
administration due to lack of data in either the clinical or combined, there was a non-significantly higher frequency of
pharmaceutical environment. contamination of doses prepared as part of a batch than as
individual lots [point estimate: 2.7% (95% CI: 1.1, 6.2; N ¼ 6131
Hypothesis 2. Dose preparation as individual lots or as part of
doses) (I2 ¼ 96.48%; P < 0.001) vs 2.3% (95% CI: 0.5, 10.1;
a batch
N ¼ 252 doses) (I2 ¼ 56.48%; P ¼ 0.101); P ¼ 0.856]. The
Clinical and pharmaceutical environments combined
between-study contamination was more variable for doses
All identified doses. The analysis involved 16,552 doses from
prepared as part of a batch than as individual lots (range:
34 studies (33 records8,12e43). Of these, 4309 doses from 18
0.1e28.4% vs 0.6e6.3% respectively).
studies (18 records14,15,17,18,21,22,24,27e32,35e37,40,43) had been
Doses sampled during or after administration. It was not
prepared as individual lots and 12,243 doses from 16 studies (16
possible to compare doses prepared as individual lots and as
records8,12,13,16,19e21,23,25,26,33,34,38,39,41,42) had been prepared
part of a batch that had been sampled during or after admin-
as part of a batch. When all the data were combined there was
istration due to lack of data for doses prepared as part of a
a significantly higher frequency of contamination of doses
batch.
prepared as individual lots than as part of a batch [4.4% (95% CI:
Pharmaceutical environments
2.5%, 7.6%; N ¼ 4309 doses) (I2 ¼ 90.77%; P < 0.001) vs 1.3%
All identified doses. The analysis involved 6280 doses from
(95% CI: 0.5%, 3.0%; N ¼ 12,243 doses) (I2 ¼ 96.68; P < 0.001);
seven studies (seven records12,17,26,34,39e41). Of these, 168
P ¼ 0.022]. The between-study contamination was more vari-
doses from two studies (two records17,40) had been prepared as
able for doses prepared as individual lots than as part of a batch
individual lots and 6112 doses from five studies (five re-
(range: 0.2e55.7% vs 0.0e28.4% respectively).
cords12,26,34,39,41) had been prepared as part of a batch. When
Doses sampled without administration or prior to admin-
all the data were combined, there was a significantly higher
istration. There were 12,663 doses from 21 studies (21
frequency of contamination of doses prepared as individual lots
records8,12,13,16,17,19e21,23,25e27,30,33,34,37e42) that had been
than as part of a batch [2.1% (95% CI: 0.7, 5.8; N ¼ 168 doses)
sampled without administration or prior to administration, of
(I2 ¼ 00.00%; P ¼ 0.856) vs 0.2% (95% CI: 0.1, 0.9; N ¼ 6112
which 420 doses from five studies (five records17,27,30,37,40) had
doses) (I2 ¼ 56.49%; P ¼ 0.056); P ¼ 0.002]. The between-study
been prepared as individual lots and 12,243 doses from 16
contamination was more variable for doses prepared as part of
studies (16 records8,12,13,16,19e21,23,25,26,33,34,38,39,41,42) had
a batch than as individual lots (range: 0.0e2.4% vs 2.0e2.6%
been prepared as part of a batch. When all the data were
respectively).
combined, there was a non-significantly higher frequency of
Doses sampled without or prior to administration. Since all
contamination of doses prepared as individual lots than as part
of the identified doses had been sampled without or prior to
of a batch [point estimate: 2.7% (95% CI: 1.2, 6.0; N ¼ 420
administration, a comparison of doses prepared as individual
doses) (I2 ¼ 28.53%; P ¼ 0.231) vs 1.3% (95% CI: 0.5, 3.0;
lots and as part of a batch that had been sampled without or
N ¼ 12,243 doses) (I2 ¼ 96.68%; P < 0.001); P ¼ 0.231]. The
prior to administration yields the same results as all of the data
between-study contamination was more variable for doses
combined (above).
prepared as part of a batch than as individual lots (range:
Doses sampled during or after administration. It was not
0.0e28.4% vs 0.6e6.3%, respectively).
possible to compare doses prepared as individual lots and as
Doses sampled during or after administration. It was not
part of a batch that had been sampled during or after admin-
possible to compare doses prepared as individual lots and as
istration, since no relevant doses that had been prepared as
part of a batch that had been sampled during or after admin-
either individual lots or as part of a batch had been identified.
istration due to lack of data for doses prepared as part of a
batch. Hypothesis 3. Undertaking additions to terminally sterilized
Clinical environments doses
All identified doses. The analysis involved 10,272 doses from The maximum expected contamination rate of doses
27 studies (26 records8,13e16,18e25,27e33,35e38,42,43). Of these, terminally sterilized according to appropriate and validated
4141 doses from 16 studies (16 records14,15,18,21,22,24,27e32, procedures is one per million.53
35e37,43
) had been prepared as individual lots and 6131 doses Clinical and pharmaceutical environments combined
from 11 studies (11 records8,13,16,19e21,23,25,33,38,42) had been It was not possible to combine doses from clinical and
prepared as part of a batch. When all the data were combined, pharmaceutical environments, since no studies that reported
there was a non-significantly higher frequency of contamina- the contamination rate of sterile doses with and without ad-
tion of doses prepared as individual lots than as part of a batch ditives undertaken in pharmaceutical environments had been
[4.7% (95% CI: 2.5%, 8.4%; N ¼ 4141 doses) (I2 ¼ 91.64%; identified.

Please cite this article in press as: Austin PD, et al., Systematic review and meta-analysis of the risk of microbial contamination of parenteral
doses prepared under aseptic techniques in clinical and pharmaceutical environments: an update, Journal of Hospital Infection (2015), http://
dx.doi.org/10.1016/j.jhin.2015.04.007
10 P.D. Austin et al. / Journal of Hospital Infection xxx (2015) 1e13
Clinical environments There was a consistently lower frequency of contamina-
The analysis involved 1723 doses from six studies (six rec- tion of doses prepared in pharmaceutical environments
ords15,18,21,28,29,35). Of these, additions had been made to 1108 compared to clinical environments. However, this finding was
doses and no additions had been made to 615 doses. All of the not found to be statistically significant for doses prepared as
doses had been prepared as individual lots and sampled during individual lots, despite up to more than a two-fold difference
or after administration. Figure 3 shows the forest plot when all in the overall frequency of dose contamination (4.7% vs 2.1%
of the study data were combined in a meta-analysis. There was for all individual doses combined, and 2.3% vs 2.1% for only
a significantly higher frequency of contamination of doses with those individual doses sampled without administration or
additives than without additives [risk ratio: 2.121 (95% CI: prior to administration). This lack of statistical significance
1.093, 4.114); P ¼ 0.026], with a low statistical heterogeneity could at least in part be explained by the limited data
(I2 ¼ 22.50%; P ¼ 0.265). identified for doses prepared as individual lots in pharma-
Pharmaceutical environments ceutical environments (N ¼ 168) (a potential type 2 error due
It was not possible to compare sterile doses with and to inadequate statistical power), which could have been
without additives in pharmaceutical environments, since no compounded by necessary mathematical corrections during
relevant studies had been identified. the analyses (see limitations below). A consistently narrower
range of between-study frequencies of dose contamination
was found for doses prepared in pharmaceutical than in
Discussion clinical environments.
The lower frequency and variability of contamination of
This update includes more than double the number of doses doses prepared in pharmaceutical than in clinical environments
than the 2009 review,9 which we have attempted to summarize is intuitive since pharmaceutical facilities are constructed and
to help inform judgements when establishing policy and clin- operated to restrict the number of environmental microbes,
ical practice that ultimately aim to reduce patient infection incorporate specialized equipment operated by staff wearing
rates. Overall, the contamination frequency was lower when special clothing to minimize shedding of micro-organisms (and
doses had been prepared in pharmaceutical than in clinical particles) and who have more consistent and extensive training
environments, but reported rates were often unacceptably in the validation in the use of aseptic techniques.8 When re-
high in both settings. For example, the mean reported study ported, the types of micro-organisms found after preparation
frequency of microbial contamination of doses prepared under in pharmaceutical environments were generally of low patho-
aseptic techniques in pharmaceutical environments could
genicity. The same bacteria were reported in doses prepared in
be >100 times higher than that expected from following the clinical environments, but a wider range of micro-organisms
procedures recommended in Europe (>2.0% compared with was found in this setting, including various Gram-negative
0.02%),11 and >2750 times higher in clinical environments than
bacteria and fungi that have greater pathogenic potential.
that expected in a pharmaceutical environment (>55.0% Not only are limited or no environmental control procedures
compared to 0.02%). The greater number of studies identified followed in clinical environments, but the closer proximity of
in this update meant that a previously non-significant but drug preparation to patients serves as an additional source of
intuitive finding of the previous review9 achieved statistical micro-organisms that are potentially antibiotic resistant and/
significance in the present review (Hypothesis 3). or more pathogenic. Indeed, perceived benefits of pharma-
Hypothesis 1. Dose preparation in a clinical compared to a ceutical rather than clinical environments for aseptic prepa-
pharmaceutical environment ration of parenteral doses have been noted in national

Risk Lower Upper

ratio limit limit P-Value Risk ratio and 95% CI

Breheny et al. 1990* 0.640 0.013 31.985 0.823

D'Arcy et al. 1973* 4.459 1.906 10.431 0.001

Ernerot et al. 1973* 0.458 0.079 2.646 0.383

Kundsin et al. 1973* 2.859 0.336 24.331 0.336

Letcher et al. 1972* 1.648 0.600 4.524 0.332

Poretz et al. 1974* 2.273 0.517 9.993 0.277

2.121 1.093 4.114 0.026

0.01 0.1 1 10 100

Favours additives Favours no additives

Figure 3. Forest plot of random effects meta-analysis comparing contamination rates of sterile parenteral doses with and without ad-
ditives in clinical environments. Asterisks indicate doses that were sampled during or after administration. CI, confidence interval.

Please cite this article in press as: Austin PD, et al., Systematic review and meta-analysis of the risk of microbial contamination of parenteral
doses prepared under aseptic techniques in clinical and pharmaceutical environments: an update, Journal of Hospital Infection (2015), http://
dx.doi.org/10.1016/j.jhin.2015.04.007
 P.D. Austin et al. / Journal of Hospital Infection xxx (2015) 1e13 11
documents, such as in the UK, and particularly for high-risk Limitations
products such as parenteral nutrition.54e56
In addition to potential clinical benefits it is also necessary The evidence base was limited and generally based on poor
to consider the economic consequences of where doses are quality studies, weakening the conclusions of this paper. One
prepared. None of the reviewed studies undertook a cost- of the main limitations is that the studies did not primarily set
effectiveness analysis but the start-up costs for building a out to examine the hypotheses raised in this review and so did
new facility to create an appropriate pharmaceutical envir- not use the most appropriate study designs to address them.
onment would be high (e.g. several million national currency Furthermore, although there are substantially more studies in
units in Europe or the USA). There are also substantial ongoing the current review than in the 2009 review,9 there is still the
costs (including operator training, maintenance, monitoring for possibility that a type 2 error may have arisen when testing
environmental contaminants, and the need for an appropri- specific hypotheses. For example, there were only 168
ately qualified manager). In addition, logistic issues created by individual doses prepared in pharmaceutical environments
a centralized facility, such as the need to reallocate staff identified. The risk of type 2 error may have also been
resource from wards to the pharmacy department, and the increased by the need to add 0.5 contaminated doses in a
need to safely and efficiently deliver drugs to points of use, group when in reality there were no contaminated doses. For
would have to be addressed. example, the effect on the rate of contamination of doses
Hypothesis 2. Dose preparation as individual lots or as part of
prepared as individual lots in a pharmaceutical environment in
a batch one study was reported as 0.0% (zero contaminated doses
For all the doses in both clinical and pharmaceutical envir- from a total of 18 doses) but was included in the analyses as
onments combined, contamination was found to be higher in 2.4% (0.5 contaminated doses from a total of 18 doses).40 The
doses prepared as individual lots rather than as part of a batch. relevance of this mathematical complication is reduced as the
This difference was found to be significant in pharmaceutical sample size increases. Another potential limitation is that the
environments but not in clinical environments. It is intuitive studies spanned a period of >40 years (1972 to 2013), most of
that individual doses would pose a higher risk than batch doses which were more than 10 years old [79% (27 from 34 studies)],
in pharmaceutical environments since the risks of batch which raises the possibility that the overall results do not
preparation are offset by fewer environmental contaminants, exactly reflect current practice with currently used products.
less variable techniques, and the availability of specialized Finally, the general lack of head-to-head trials (seven from a
equipment. It is also intuitive that potential benefits of batch total of 33 records) has meant that in some cases less robust
preparation would be lost in an uncontrolled environment with analyses had to be used. In standard meta-analyses involving
greater contaminants where more variable techniques are head-to-head trials, the differences between two groups of
employed and no specialized equipment for batch production is individual studies are established and amalgamated (two
available. These findings support recommendations to limit the group meta-analysis). By contrast, in the present work for
expiry of parenteral doses prepared under aseptic techniques hypotheses 1 and 2 the average results from studies involving
in clinical environments, for example to 24 h in the UK, which each group were amalgamated separately (one group meta-
effectively preclude batch preparation, and which do not apply analysis) and then compared with each other. This increases
to pharmaceutical environments (although different additional the risk of bias since the products tested and conditions in the
requirements do apply).56 two comparator groups are less well matched. For example,
38% (3889 from 10,272) of the doses prepared in clinical
Hypothesis 3. The effect of undertaking additions to termi- environments had been sampled during or after administration
nally sterilized doses compared to none (from 6280) of the doses prepared in
It is reasonable to suggest that aseptic manipulations to a pharmaceutical environments, and 40% (4141 from 10,272) of
sterile dose can only increase the risk of microbial contami- doses prepared in clinical environments had been prepared as
nation, but there is limited evidence for such an effect. Unlike individual lots compared to 3% (168 from 6280) in
the 2009 review,9 which reported no significant effect of ad- pharmaceutical environments.
ditions to sterilized doses, this updated review found a
significantly higher contamination rate of sterile doses sub- Recommendations
jected to additions compared to those that were not [a risk
ratio of 1.459 (P ¼ 0.682) and 2.121 (P ¼ 0.026) respectively]. It is logical that the safest environment should be used to
This difference can be explained by use of a meta-analysis prepare parenteral doses under aseptic technique but several
based on only three studies15,18,21 with high statistical high-profile incidents (including deaths) in recent years make
heterogeneity (I2 ¼ 66.45%, P ¼ 0.055) in the 2009 review,9 and the continued lack of high-quality data in this field surprising.
a meta-analysis based on six studies15,18,21,28,29,35 with lower The limited and low-quality evidence base supports the use of
heterogeneity (I2 ¼ 22.50%, P ¼ 0.265) in the present review. pharmaceutical rather than clinical environments for aseptic
This finding is consistent with the intuitive idea that aseptic parenteral dose preparation and does not support batch
manipulations should be minimized in uncontrolled environ- preparation in clinical environments, but further data are
ments such as hospital wards whenever possible. Neverthe- required. There is a need for high-quality head-to-head trials
less, adequate protocols and training are still required when it with large sample sizes to strengthen the available evidence
is necessary to prepare doses in clinical environments under base. Such studies would better inform decisions and policies in
aseptic technique. The updated conclusion that additions to clinical practice. In particular, at the present time there are
sterile doses in clinical environments increase the contami- limited published data for doses prepared as individual lots in
nation rate is in line with the findings for the previous two pharmaceutical environments (N ¼ 168), and for doses pre-
hypotheses. pared as individual lots in clinical environments without

Please cite this article in press as: Austin PD, et al., Systematic review and meta-analysis of the risk of microbial contamination of parenteral
doses prepared under aseptic techniques in clinical and pharmaceutical environments: an update, Journal of Hospital Infection (2015), http://
dx.doi.org/10.1016/j.jhin.2015.04.007
12 P.D. Austin et al. / Journal of Hospital Infection xxx (2015) 1e13
administration to patients (N ¼ 252). In addition, the intro- and distributors (The Orange Guide). London: Pharmaceutical
duction of a reporting system for contamination rates achieved Press; 2014.
during routine clinical practice and/or routine simulation 8. Austin P, Elia M. Improved aseptic technique can reduce variable
studies used to verify competence of operator aseptic tech- contamination rates of ward-prepared parenteral doses. J Hosp
Infect 2013;83:160e163.
nique that takes into account the sampling procedures
9. Austin PD, Elia M. A systematic review and meta-analysis of the
employed would be of benefit. Future work in this area should risk of microbial contamination of aseptically prepared doses in
also consider the risk of contamination and infection with different environments. J Pharm Pharmaceut Sci 2009;12:
different types of microbes, the clinical risks associated with 233e242.
contamination of different types of preparation (e.g. intui- 10. Pharmaceutical Inspection Convention Co-operation Scheme (PIC/
tively an intraocular preparation sounds higher risk than a S). Guide to good manufacturing practice for medicinal products.
preparation intended for bolus intravenous administration), 2014; document reference PI 009e11.
and the economic implications, including cost-effectiveness, 11. Pharmaceutical Inspection Convention Co-operation Scheme
of drug preparation in clinical and pharmaceutical (PIC/S). Recommendation on the validation of aseptic processes.
environments. 2011; document reference PI 007-6, Interpretation of data (sec-
tion 6).
12. Austin P, Dixson S. Hub fluid does not increase microbiological
Acknowledgements contamination of prepared and stored syringes. Pharmaceut J
2006;276(7383):47e49.
13. Aydin N, Aydin N, Gultekin B, Ozgun S, Gurel A. Bacterial
This study was made possible by financial support from the
contamination of propofol: the effects of temperature and lido-
Wessex Clinical Academic Careers Steering Group and the fa-
caine. Eur J Anaesthesiol 2002;19:455e458.
cilities of the National Institute for Health Research South- 14. Bach A, Motsch J, Schmidt H, et al. In-use contamination of pro-
ampton Biomedical Research Centre. The authors are also pofol. A clinical study. Eur J Anaesthesiol 1997;14:178e183.
grateful to the Health Service Library at Southampton General 15. Breheny F, McGechie D, Toohey M. Contamination of parenteral
Hospital, and in particular to P. Sands (Academic Liaison nutrition solutions not a hazard with additions made at ward level.
Librarian) for an update on the use of the literature search Anaesth Intensive Care 1990;18:547e549.
software and S. Vickerstaff (Document Delivery Library Assis- 16. Burke KG, Larson E, Maciorowski L, Adler DC. Evaluation of the
tant) for her assistance with obtaining the full text of a number sterility of thermodilution room-temperature injectate prepara-
of the records. tions. Crit Care Med 1986;14:503e504.
17. Choy FN, Lamy PP, Burkhart VD, Tenney JH. Sterility-testing
program for antibiotics and other intravenous admixtures. Am J
Conflict of interest statement
Hosp Pharm 1982;39:452e456.
None declared. 18. D’Arcy PF, Woodside W. Drug additives: a potential source of
bacterial contamination of infusion fluids. Lancet
Funding sources 1973;2(7820):96.
The time taken to undertake this study by one of the authors 19. Dominik RH, Segebade IE, Taenzer V. Risk of microbial contami-
(P.D.A.) was financially supported by the Wessex Clinical nation of iodinated contrast media on multiple use of large-
Academic Careers Steering Group, and the facilities of the volume bottles. Eur J Radiol 1995;19:198e205.
National Institute for Health Research Southampton 20. Driver Jr RP, Snyder IS, North FP, Fife TJ. Sterility of anesthetic
Biomedical Research Centre were used to undertake this and resuscitative drug syringes used in the obstetric operating
study. room. Anesth Analg 1998;86:994e997.
21. Ernerot L, Thoren S, Sandell E. Studies on microbial contamination
of infusion fluids arising from drug additions and administration.
References Acta Pharm Suec 1973;10:141e146.
22. Farrington M, McGinnes J, Matthews I, Park GR. Do infusions of
1. Muller AE, Huisman I, Roos PJ, et al. Outbreak of severe sepsis due midazolam and propofol pose an infection risk to critically ill pa-
to contaminated propofol: lessons to learn. J Hosp Infect tients? Br J Anaesth 1994;72:415e417.
2010;76:225e230. 23. Fleer A, Senders RC, Visser MR, et al. Septicemia due to
2. Centers for Disease Control and Prevention. Notes from the field: coagulase-negative staphylococci in a neonatal intensive care
multistate outbreak of postprocedural fungal endophthalmitis unit: clinical and bacteriological features and contaminated
associated with a single compounding pharmacy e United States, parenteral fluids as a source of sepsis. Pediatr Infect Dis
MarcheApril 2012. Morb Mortal Wkly Rep 2012;61:310e311. 1983;2:426e431.
3. Chiang PC, Wu TL, Kuo AJ, et al. Outbreak of Serratia marcescens 24. Hernandez-Ramos I, Gaitan-Meza J, Garcia-Gaitan E, Leon-
postsurgical bloodstream infection due to contaminated intra- Ramirez AR, Justiniani-Cedeno N, Avila-Figueroa C. Extrinsic
venous pain control fluids. Int J Infect Dis 2013;17:e718ee722. contamination of intravenous infusates administered to hospital-
4. Boseley S. Questions remain over source of ITH Pharma baby feed ized children in Mexico. Pediatr Infect Dis J 2000;19:888e890.
contamination. The Guardian. Available at: http://www. 25. Jackson EA, Gallo BM. Sterility of insulin in prefilled disposable
theguardian.com/society/2014/jun/05/questions-remain-source- syringes. Am J Hosp Pharm 1990;47:2508e2510.
ith-pharma-baby-feed-contamination; June 5th, 2014 [last 26. Jacobson PA, West NJ, Spadoni V, Maksym CJ, Pierson C. Sterility
accessed January 2015]. of filgrastim (G-CSF) in syringes. Ann Pharmacother 1996;
5. Phamaceutical Compounding e Sterile Preparations (Chapter 797). 30:1238e1242.
In: The United States Pharmacopeia: The United States Pharma- 27. Khalili H, Sheikhbabayi M, Jamalifar H, Dalili D, Samadi N. Bac-
copeial Convention 2008. terial contamination of single- and multiple-dose vials after mul-
6. Beaney AM. Quality assurance of aseptic preparation services. 4th tiple use and intravenous admixtures in three different hospitals
edn. London/Chicago: Pharmaceutical Press; 2006. in Iran. Iranian J Pharmaceut Res 2013;12:205e209.
7. Great Britain Medicines and Healthcare Products Regulatory 28. Kundsin RB, Walter CW, Scott JA. In-use testing of sterility of intra-
Agency. Rules and guidance for pharmaceutical manufacturers venous solutions in plastic containers. Surgery 1973;73:778e781.

Please cite this article in press as: Austin PD, et al., Systematic review and meta-analysis of the risk of microbial contamination of parenteral
doses prepared under aseptic techniques in clinical and pharmaceutical environments: an update, Journal of Hospital Infection (2015), http://
dx.doi.org/10.1016/j.jhin.2015.04.007
 P.D. Austin et al. / Journal of Hospital Infection xxx (2015) 1e13 13
29. Letcher KI, Thrupp LD, Schapiro DJ, Boersma JE. In-use contami- syringes for continuous intravenous drug administration in a simu-
nation of intravenous solutions in flexible plastic containers. Am J lation model in intensive care units. Crit Care Med 2002;30:
Hosp Pharm 1972;29:673e677. 833e836.
30. Lorenz IH, Kolbitsch C, Lass-Florl C, et al. Routine handling of 43. Yorioka K, Oie S, Oomaki M, Imamura A, Kamiya A. Particulate and
propofol prevents contamination as effectively as does strict microbial contamination in in-use admixed intravenous infusions.
adherence to the manufacturer’s recommendations. Can J Biol Pharmaceut Bull 2006;29:2321e2323.
Anaesth 2002;49:347e352. 44. Lawrence J, Turner M, Gilbert P. Microbial contamination and
31. Macias A, De La Torre A, Huertas-Jimenez MA, Silva Y, Gonzalez L, growth in total parenteral nutrition solutions. J Clin Pharm
Polanco C. Endemic IV fluid contamination in hospitalized children Therapeut 1988;13:151e157.
in Mexico. A problem of serious public health consequences. Am J 45. Reiter PD. Sterility of intravenous fat emulsion in plastic syringes.
Infect Control 2012;40:e180ee181. Am J Health-Syst Pharm 2002;59:1857e1859.
32. Madeo M, Samaan AK, Allison W, Wilson JA, Martin CR. Contami- 46. Takagi J, Khalidi N, Wolk RA, Tjolsen E, de Leon R, Wesley JR.
nation of bags for continuous epidural infusion. Acute Pain Sterility of total parenteral nutrient solutions stored at room
1999;2:125e128. temperature for seven days. Am J Hosp Pharm 1989;46:973e977.
33. Magee L, Godsiff L, Matthews I, Farrington M, Park GR. Anaes- 47. Weil DC, Arnow PM. Safety of refrigerated storage of admixed
thetic drugs and bacterial contamination. Eur J Anaesthesiol parenteral fluids. J Clin Microbiol 1988;26:1787e1790.
Suppl 1995;12:41e43. 48. Guyatt GH, Oxman AD, Vist GE, et al. GRADE: an emerging
34. Micard S, Rieutord A, Prognon P, Hassan M, Brion F. Stability and consensus on rating quality of evidence and strength of recom-
sterility of meglumine gadoterate injection repackaged in plastic mendations. BMJ 2008;336(7650):924e926.
syringes. Int J Pharmaceut 2001;212:93e99. 49. Guyatt GH, Oxman AD, Kunz R, et al. What is “quality of evidence”
35. Poretz DM, Guynn Jr JB, Duma RJ, Dalton HP. Microbial contami- and why is it important to clinicians? BMJ 2008;336(7651):
nation of glass bottle (open-vented) and plastic bag (closed-non- 995e998.
vented) intravenous fluid delivery systems. Am J Hosp Pharm 50. Centre for Reviews and Dissemination. Systematic reviews: CRD’s
1974;31:726e732. guidance for undertaking reviews in health care. 3rd edn. York:
36. Soong WA. Bacterial contamination of propofol in the operating CRD; 2009.
theatre. Anaesth Intensive Care 1999;27:493e496. 51. Higgins J, Green S, editors. Cochrane handbook for systematic
37. Spiliotis J, Arvaniti A, Tsirigotis M, Angelopoulou H, Kalfarentzos F. reviews of interventions version 5.1.0 (updated March 2011).
Contamination rates of total parenteral nutrition bags prepared Cochrane Collaboration. Available at: www.cochrane-handbook.
under aseptic conditions. Intensive Ther Clin Monit org; 2011 [last accessed April 2015].
1989;10:302e303. 52. Moher D, Liberati A, Tetzlaff J, Altman DG, Group P. Preferred
38. Stjernstrom G, Gunnarsson B, Wikner H. Studies on microbiological reporting items for systematic reviews and meta-analyses: the
contamination of in-use IV-fluids. Acta Pharm Suec 1978;15:169e174. PRISMA statement. Ann Intern Med 2009;151:264e269, W64.
39. Thomas M, Sanborn MD, Couldry R. I.V. admixture contamination 53. Sharp J. Quality in the manufacture of medicines and other
rates: traditional practice site versus a class 1000 cleanroom. Am J healthcare products. London: Pharmaceutical Press; 2000.
Health-Syst Pharm 2005;62:2386e2392. p.331e360.
40. Urbano N, Modoni S, Schillaci O. Media Fill Test for validation of 54. Department of Health and Social Security Health Services Division.
autologous leukocytes separation and labelling by (99m)Tc- Chapter 8 and Chapter 10 of Health Circular HC(76)9: Report of
HmPAO. Nucl Med Biol 2013;40:104e108. the working party on the addition of drugs to intravenous infusion
41. van Doorne H, Bakker JH, Meevis RF, Marskamp A. Influence of fluids (the ‘Breckenridge Report’). London: DHSS; 1976.
background air on microbial contamination during simulated 55. Audit Commission for Local Authorities and the National Health
i.v.-admixture preparation. J Clin Pharm Therapeut 1994;19: Service in England and Wales. A spoonful of sugar (medicines
181e187. management in NHS hospitals), paragraph 46. London: Audit
42. van Grafhorst JP, Foudraine NA, Nooteboom F, Crombach WH, Commission; 2001.
Oldenhof NJ, van Doorne H. Unexpected high risk of contamination 56. National Patient Safety Agency. Patient safety alert 20 Promoting
with staphylococci species attributable to standard preparation of safer use of injectable medicines. London: NPSA; 2007.

Please cite this article in press as: Austin PD, et al., Systematic review and meta-analysis of the risk of microbial contamination of parenteral
doses prepared under aseptic techniques in clinical and pharmaceutical environments: an update, Journal of Hospital Infection (2015), http://
dx.doi.org/10.1016/j.jhin.2015.04.007