Epidemiological explorations on Clostridium difficile Infection
Goorhuis, A.
Citation
Goorhuis, A. (2011, October 12). Epidemiological explorations on Clostridium difficile Infection. Retrieved from https://hdl.handle.net/1887/17925
Version: Corrected Publisher’s Version
License: Licence agreement concerning inclusion of doctoral thesis in the
Institutional Repository of the University of Leiden
Downloaded from: https://hdl.handle.net/1887/17925
Note: To cite this publication please use the final published version (if applicable).
Chapter 8
General Discussion
154
Outset of the research
The research that has resulted in this thesis started in the second half of 2006. At this time, the Mirst large CDI outbreaks with the hyper virulent type 027/NAP1/BI (type 027) strain had just been described in Canada and the The United States 1,2. In addition, a then very recent outbreak with this strain in the Stoke Mandeville hospital in The United Kingdom had demonstrated that outbreaks seemed to evolve to a global emergence of this type 3. In Canada, an increased incidence was noted since 1991, especially among elderly patients (Migure 1).
Current Concepts
n engl j med 359;18 www.nejm.org october 30, 2008 1933
isolates from five facilities, and 82% of stool sam- ples from the Quebec outbreak were positive for the same strain.
4,7This epidemic strain was ini- tially identified in the 1980s by restriction endo- nuclease analysis and named BI, but is currently referred to as North American Pulsed Field type 1 (NAP1) and PCR ribotype 027 (i.e., BI/NAP1/027, or NAP-1/027).
7Three bacterial factors have been implicated in outbreaks of C. difficile infection caused by the virulent NAP-1/027 strain: increased production of toxins A and B, fluoroquinolone resistance, and production of binary toxin. Toxins A and B are the major virulence determinants of C. difficile;
indeed, toxin-negative strains are nonpathogenic.
Toxins A and B are transcribed from a pathoge- nicity locus that comprises five genes: two toxin genes, tcdA (toxin A) and tcdB (toxin B), and three regulatory genes, one of which (tcdC) encodes a putative negative regulator of toxin transcription (Fig. 2A and 2B).
8,9TcdC protein appears to in- hibit toxin transcription during the early, expo- nential-growth phase of the bacterial life cycle.
NAP-1/027 isolates that were obtained from pa- tients during recent outbreaks of C. difficile infec- tion carry deletion mutations in the tcdC inhibitory gene that have been associated with an increase by more than a factor of 10 in the production of toxins that mediate colonic tissue injury and inflammation in C. difficile infection (Fig. 2C).
7,9,10These toxins bind to the surface of intestinal epithelial cells, where they are internalized and catalyze the glucosylation of cytoplasmic rho pro- teins, leading to cell death (Fig. 2D).
11All NAP-1/027 isolates from the 1980s and 1990s, like those from recent outbreaks, carry tcdC mutations.
4,7In contrast, high-level resistance to gatifloxacin and moxifloxacin is evident in recent isolates but not in historic NAP-1 strains. Resis- tant strains may have a competitive advantage in a hospital environment where fluoroquinolone use is widespread.
12This theory is supported by the finding in the Quebec outbreak that the odds ratio for fluoroquinolone use in patients with C. difficile infection, as compared with con- trol subjects, was 3.9 (95% confidence interval [CI], 2.3 to 6.6), which was virtually the same as the odds ratio (3.8) for the use of cephalosporin (95% CI, 2.2 to 6.6), a longtime leading antibi- otic class predisposing to C. difficile infection.
4This observation suggests that limiting fluoroquino - lone use may help to contain outbreaks caused by NAP-1/027, as was reported earlier for the re-
striction of clindamycin in an outbreak caused by a clindamycin-resistant strain.
13Another potential virulence determinant of NAP-1/027 strains is the production of a third toxin, binary toxin, that is unrelated to the patho- genicity locus that encodes toxins A and B.
14Previously, about 6% of C. difficile clinical isolates produced binary toxin, homologous to the iota toxin of C. perfringens and comprised of a 48-kD enzymatic component and a 99-kD binding com- ponent. Binary toxin has enterotoxic activity in vitro, but its role, if any, in the pathogenesis of C. difficile infection is not clear.
14-16C. difficile strains that produce binary toxin in the absence of toxins A and B do not appear to be pathogenic. Nonethe- less, the finding that NAP-1/027 epidemic strains produce binary toxin has raised renewed specu- lation that this toxin may act synergistically with toxins A and B in causing severe colitis.
4,5,7,14-16E xpanding Epidemiology
C. difficile infection predominantly affects elderly and frail hospital and nursing home patients (Fig. 1).
2,3However, a recent advisory from the Centers for Disease Control and Prevention warns of a risk of the infection in populations not previ- ously considered at risk.
17These include young
22p3 880
Annual Incidence per 100,000
860
200
140 160 180
100 120
80 60 40 20 0
1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003
<17 yr 18–64 yr
≥65 yr Total
AUTHOR:
FIGURE:
JOB:
4-C H/T
RETAKE
SIZE ICM
CASE
EMail Line
H/TCombo
Revised
AUTHOR, PLEASE NOTE:
Figure has been redrawn and type has been reset.
Please check carefully.
REG F
Enon
1st 2nd 3rd Kelly
1 of 4
10-30-08 ARTIST: ts
35918 ISSUE:
Age
Figure 1. Annual Incidence (per 100,000 Population) of C. difficile Infection in Sherbrooke, Quebec, 1991–2003.
The overall incidence of C. difficile infection was relatively stable during the period from 1991 through 2002, although there was a gradual increase in the rate of infection among elderly patients (≥65 years). In 2003, the popu- lation incidence increased by a factor of 4, as compared with 2002. This in- crease was especially evident in the elderly. Data are from Pépin et al.3
The New England Journal of Medicine
Downloaded from nejm.org at LEIDS UNIVERSITY MEDISCH CENTRUM on January 12, 2011. For personal use only. No other uses without permission.
Figure 1. Annual Incidence (per 100,000 Population) of C. dif-icile Infection in Sherbrooke, Quebec (Canada), 1991–2003 4.
In The Netherlands, the Mirst outbreak occurred in 2005, in a hospital in Harderwijk, as was described by Kuijper and by Debast 5,6. At that time, an important question was if the type 027 strain truly was a hyper virulent strain, or that the severity of disease that was associated with this strain was mainly caused by increased age and co-‐morbidities of hospitalized
patients in the modern era.
155
The problem with the earliest outbreak reports from Canada, The United States and the United Kingdom, was the fact that the morbidity and mortality that was observed and that was associated with the type 027 strain was measured against historical controls, instead of
against patients who were infected with another strain type of C. dif-icile at the same time.
It was speculated that the use of Mluoroquinolones played a part in the global emergence of the type 027 strain, because it was found to be resistant to this class of antibiotics. It was however not known whether other strain types had a similar resistance to this antibiotic class.
In fact, because all the characteristics and risk factors that were reported for the type 027 strain had been observational, without the availability of an adequate control group (i.e.
simultaneously infected patients with other strain types than the type 027 strain), there was a need for studies that included such a control group, in order to identify true risk factors, associated with this new disease entity. At the start of this research, several studies had concluded that exposure to Mluoroquinolones was a major risk factor for the development of CDI due to ribotype 027 strains 7-‐9. However, these studies only included matched control patients who did not have CDI.
Another question was if there were differences between epidemic (outbreak) settings and settings with a low prevalence of CDI and whether there would be separate risk factors for CDI in these settings
Furthermore, there was a need for robust typing methods, capable of discriminating within outbreak strains, in order to investigate the spread and magnitude of these outbreaks.
Hyper virulent Clostridium dif-icile in The Netherlands.
As was described in chapter 2, the Mirst type 027 strains that were found to be circulating in The Netherlands date from 2002. It was not until 2005 that an increase in the number of infections by this type was noticed. In fact, in that year, the type 027 strain heralded its
presence with a bang, by causing the Mirst recognized outbreak with this strain in a hospital in
156
Harderwijk. In the study period thereafter, an increasing number of affected hospitals was reported, with an increasing number of outbreaks. The same held for other European countries. By 2008, outbreaks had been observed in 16 different countries (Migure) 10.
CDI during a 10-year period. Analysis of historic BI/
NAP1/027 isolates has shown them to be virtually iden- tical to the recent outbreak strains, with the exception that the older isolates were not resistant to the newer fluoroquinolone antibiotics, gatifloxacin, and moxifloxa- cin (current epidemic strains are resistant to these anti- biotics).
Epidemiology in the Hospital, Community, and Animals
Most CDI that has been attributed to the BI/
NAP1/027 strain has been found in hospitals and is commonly associated with outbreaks or epidemics. This certainly was the case in Pittsburgh, Quebec, Stoke Mandeville, and numerous other North American hospi- tals.
3,5– 8In the hospital setting, CDI occurs most fre- quently in elderly patients; it has been shown that risk and mortality of CDI from the BI/NAP1/027 strain in- creases with patient age (Table 1).
2,7The BI/NAP1/027 strain was not observed to cause disease in the commu-
Figure 1. Extent of BI/NAP1/027 distribution. (A) States in the United States that have had !1 hospital that has reported CDI caused by the BI/NAP1/027 epidemic strain as of October 2008 (red).116(B) Percent- age of C difficile isolates in Canadian provinces (no data are available for territories) with the BI/NAP1/027 strain in the 2005 Canadian Nosoco- mial Infection Surveillance Program (CNISP) survey.97 (C) Hospitals in Europe reporting outbreaks (stars) and sporadic cases (circles) of CDI caused by the BI/NAP1/027 strain.89
Figure 2. Frequency and mortality of CDI. (A) National estimates of US short-stay hospital discharges with C difficile infection as any listed diagnosis or the primary hospitalization diagnosis, based on the na- tional inpatient sample117and unpublished data, 2008). (B) C difficile–
related mortality based on listings on US death certificates from 1999 to 2004, with age-adjusted mortality rates per million people for the same years (shown in box).118
1914 O’CONNOR ET AL GASTROENTEROLOGY Vol. 136, No. 6
Figure 2. Distribution of Clostridium dif-icile type 027 by country in Europe* until June 2008 10.
During the years 2005 and 2006, an increased awareness among physicians and infection preventionists across the Netherlands led to the implementation of rapid testing algorithms and infection control protocols 5. To date, possibly due to the successful
implementation of these measures, the rate of CDI caused by type 027 has decreased, as is illustrated in Migure 3 11. By contrast, the ‘common’ types 001 and 014 remain prominently present in Dutch hospitals. Type 078 is currently the third most common PCR ribotype in the Netherlands and other European countries, whereas its occurrence before 2005 was very rare, as was discussed in chapter 5.
www.eurosurveillance.org 1
R a p i d c o m m u n i c a ti o n s
D E C R E A S E O F H Y P E R V I R U L E N T C L O S T R I D I U M D I F F I C I L E P C R
R I B O T Y P E 0 2 7 I N T H E N E T H E R L A N D S
M P Hensgens1, A Goorhuis1, D W Notermans2, B H van Benthem2, E J Kuijper (e.j.kuijper@lumc.nl)1
1. National Reference Laboratory for Clostridium difficile, Leiden University Medical Center, Leiden, the Netherlands 2. Rijksinstituut voor Volksgezondheid en Milieu (National Institute for Public Health and the Environment; RIVM), Centrum
Infectieziektebestrijding (Centre for Infectious Disease Control; Cib), Bilthoven, the Netherlands
This article was published on 12 November 2009.
Citation style for this article: Hensgens MP, Goorhuis A, Notermans DW, van Benthem BH, Kuijper EJ. Decrease of hypervirulent Clostridium difficile PCR ribotype 027 in the Netherlands. Euro Surveill. 2009;14(45):pii=19402. Available online: http://www.eurosurveillance.org/ViewArticle.aspx?ArticleId=19402
!FTERTHElRSTOUTBREAKSOF#LOSTRIDIUMDIFlCILE0#2RIBOTYPE
.ORTH !MERICAN PULSED lELD TYPE RESTRICTION ENDONUCLEASE
ANALYSIS GROUP ") IN THE .ETHERLANDS IN A NATIONAL
SURVEILLANCEPROGRAMMEFOR#DIFlCILEINFECTION#$) WASSTARTED
&URTHERMORE NATIONAL GUIDELINES WERE DEVELOPED TO RAPIDLY
RECOGNISETYPEINFECTIONSANDPREVENTFURTHERSPREAD4HE
MEANINCIDENCEOF#$)MEASUREDINHOSPITALSREMAINEDSTABLE
THROUGHOUTTHEYEARSANINCIDENCEOFPERADMISSIONS
WASSEENINAND"ETWEEN!PRILAND*UNE
ATOTALOFSAMPLESWEREAVAILABLEFOR0#2RIBOTYPING!
DECREASEWASSEENINTHENUMBERANDINCIDENCEOFTYPEAFTER
THESECONDHALFOF)NTHElRSTHALFOFTHEPERCENTAGE
OFTYPEISOLATESAMONGALL#$)DECREASEDTOWHEREAS
TYPEINCREASEDTO4YPEWASPRESENTINOF
THEISOLATESAND#DIFlCILETYPESLIGHTLYINCREASEDTO
7ECONCLUDETHATCURRENTLYTHEREISASIGNIlCANTDECREASEINTYPE
ASSOCIATED#$)INTHE.ETHERLANDS
3INCE THE NEW HYPERVIRULENT STRAIN OF #LOSTRIDIUM DIFlCILE 0#2RIBOTYPE.ORTH!MERICANPULSED lELDTYPE.!0 RESTRICTIONENDONUCLEASEANALYSIS2%! GROUP")WASFOUNDINTHE
5NITED3TATESAND#ANADAINALARGENUMBEROFCOUNTRIES
WORLDWIDEREPORTED#DIFlCILEINFECTIONS#$) DUETOTHISTYPE;=
3EVERALREPORTSINDICATEDTHAT#$)DUETOTYPEISASSOCIATED
WITHAHIGHERMORBIDITYANDMORTALITYANDALSOHASTHETENDENCYTO
RELAPSEMOREFREQUENTLY; =!NOVERVIEWPUBLISHEDIN*ULY
REVEALEDTHATTYPEWASDETECTEDIN%UROPEANCOUNTRIES
ANDWASASSOCIATEDWITHOUTBREAKSIN"ELGIUM&INLAND&RANCE 'ERMANY)RELAND,UXEMBOURGTHE.ETHERLANDS3WITZERLANDAND
THE5NITED+INGDOM;=!SOF*ULYOUTBREAKSHAVEALSOBEEN
REPORTEDIN!USTRIA;=AND$ENMARK;=
3OONAFTERTHElRSTOUTBREAKSINTHE.ETHERLANDSINA
NATIONALSURVEILLANCEPROGRAMMEFOR#DIFlCILEWASINITIATEDBY
THE,EIDEN5NIVERSITY-EDICAL#ENTRE,5-# ANDTHE#ENTREFOR
)NFECTIOUS $ISEASE #ONTROL OF THE .ATIONAL )NSTITUTE FOR 0UBLIC
(EALTHANDTHE%NVIRONMENT!LLMEDICALMICROBIOLOGISTSINTHE
.ETHERLANDSWEREREQUESTEDTOSEND#DIFlCILEISOLATESTOTHE
$UTCHNATIONALREFERENCELABORATORYATTHE,5-#FORRAPID0#2
RIBOTYPINGANDCHARACTERISATIONINCASEOFANOUTBREAKMORETHAN
TWO#$)CASESWITHINONEWEEKINONEDEPARTMENT ORWHENAPATIENT
SUFFEREDFROMSEVERE#$))NADDITIONAPROSPECTIVETHREEYEAR LONG
SURVEILLANCESTUDYOFTHEINCIDENCEOF#$)ANDTHEDISTRIBUTIONOF
!"#$#DIFlCILE0#2RIBOTYPESWASSTARTEDIN$UTCHHOSPITALS
IN*UNE
)NTHEPERIODBETWEEN!PRILAND*UNEATOTALOF
SAMPLESWERESUBMITTEDTOTHEREFERENCELABORATORYOFWHICH
N WEREAVAILABLEFOR0#2RIBOTYPING/FTHOSE
SAMPLESHADBEENSUBMITTEDBYMEDICALMICROBIOLOGISTS
BECAUSEOFEITHERSEVEREDISEASEORA#$)OUTBREAKWHEREASTHE
REMAININGWEREPARTOFTHENATIONALSURVEILLANCESTUDY3INCE
NO DIFFERENCE IN THE DISTRIBUTION OF VARIOUS 0#2 RIBOTYPES WAS
FOUNDBETWEENTHETWOSURVEILLANCESYSTEMSWEREPRESENTTHE
DATACOMBINED4HEREASONFORTHISEQUALDISTRIBUTIONISTHATMOST
HOSPITALSTHATENCOUNTEREDANOUTBREAKORACASEOFSEVERE#$) CONTINUEDTOSUBMITSAMPLESONAREGULARBASISTHEREAFTER
4HE&IGUREDEPICTSTHEDISTRIBUTIONOFTHElVEMOSTCOMMON0#2
RIBOTYPESINTHE.ETHERLANDSBETWEEN!PRILAND*UNE
!LTHOUGHTHETOTALNUMBEROFSUBMITTEDSAMPLESINCREASEDFROM
INTHESECONDQUARTEROFTOASTEADYNUMBERBETWEEN
ANDAFTERTHElRSTQUARTEROFADECREASEINTHENUMBEROF
TYPEISOLATESHASBEENOBSERVEDSINCETHESECONDHALFOF
)NTHEHOSPITALSPARTICIPATINGINTHECONTINUOUSSURVEILLANCEA
0 50 100 150 200 250 300
2nd 3rd 4th 1st 2nd 3rd 4th 1st 2nd 3rd 4th 1st 2nd 3rd 4th 1st 2nd
2005 2006 2007 2008 2009
Quarter of year
Number of isolates
Other types (n=1,236) Type 002 (n=125) Type 014 (n=358)
Type 001 (n=351) Type 078 (n=301) Type 027 (n=417)
F i g u r e
C. difficile PCR ribotypes in the Netherlands, April 2005 – June 2009 (n=2,788)
Figure 3.
Patients infected with CDI caused by type 027 were signiMicantly older and used more Mluoroquinolones, compared to patients infected with non-‐027 ribotypes. At the same time, on a national level, the pre-‐epidemic use of Mluoroquinolones increased by almost 50% (the mean number of deMined daily doses per 10.000 patient-‐days in four hospitals increased from 4357 to 6471). The study described in chapter 2 was the Mirst published report that conMirmed that use of Mluoroquinolones was a risk factor for type 027-‐CDI, also in direct comparison to patients who had CDI caused by other types.
The investigation of the dual outbreak with types 017 and 027 in Amersfoort, described in chapter 6, again conMirmed that use of Mluoroquinolones is a strong risk factor speciMically for type 027, but not for other types. Several explanations were discussed in chapter 2, such as a higher level of Mluoroquinolone resistance among type 027 strains 12, probably associated with a single transition mutation in gyrA 13, that was not found among 2 pre-‐epidemic Dutch type 027 strains, dating from 2002. However, apart from a selection advantage of the type 027 strain due to its resistance to Mluoroquinolones, Mluoroquinolones may in itself contribute to
158
CDI by inducing increased spore an cytotoxin production 14. A recent report by Kallen et al.
demonstrated a signiMicant decline in C. dif-icile cases when a complete restriction of Mluoroquinolones was initiated in a hospital, after a failure of other interventions to reduce CDI cases 15. This was also the case during the type 027 outbreak in Harderwijk in The
Netherlands, where the ban on use of Mluoroquinolones was a major component in controlling the outbreak, as illustrated in Migure 4 6.
Results
Description of the outbreak
The background incidence of CDI in St Jansdal Hospital was 3.8 patients per 10 000 admissions in 2004. In 2005, a more than ten-fold increase in the incidence of CDI was observed (Fig. 1). In this study, we included the first 45 patients diag- nosed with CDI in 2005. In total, 50 patients with CDI were diagnosed during the outbreak. Faeces were cultured, and C. difficile isolates were identified as toxinotype III and PCR ribotype 027. In addition, the strain had the binary toxin genes and contained an 18-bp deletion in the toxin regulator gene tcdC. The isolates were resistant to erythromycin (MIC >256 mg/L) and ciprofloxacin (MIC >32 mg/L), and susceptible to clindamycin (MIC 2 ml/L) and metronidazole (MIC 0.19 mg/L).
A multidisciplinary hospital outbreak management team (OMT) was formed to coordinate measures to control the epidemic. Special folders informed medical personnel in the hospital. In addition, all clinicians were informed personally.
The medical microbiologist and infection control practitioner organized special meetings on the involved wards with the nursing staff. The cleaning team received special instructions for intensified cleaning procedures from the infection control practitioner. All measures were described in a CDI hospital guideline by the OMT.
Measures taken by the OMT to control the epidemic (from 1 May 2005 onwards) included isolation of all patients with diarrhoea (until two tests, 24 h apart, gave negative results for C. difficile toxin), hand washing with water and soap, use of chlorine-containing disinfectant (0.1% sodium hypochlorite), and cohorting of all C. difficile-infected patients on a separate ward. In addition, from 7 July 2005 until 14 September 2005, a complete ban on all fluoroquinolones was established, and the use of cephalosporins and clindamycin was limited.
The course of the epidemic, including the time-scheme of all infection control measures taken and the use of antibiot- ics in the hospital, are depicted in Fig. 1. The outbreak came to an end in September 2005. After the re-introduction of fluoroquinolones, however, a temporary increase in CDI was noticed.
Description of C. difficile-associated disease cases
From 1 April 2005 until the end of August 2005, a total of 45 patients met the case definition of CDI. Clinical charac- teristics of the CDI cases are given in Table 1. Thirty-five patients developed diarrhoea during their stay in the hospital (mean duration of hospital stay prior to development of symptoms was 13 days). Of the ten patients admitted with diarrhoea, nine patients had healthcare-associated CDI, as they had been hospitalized in the same hospital within the preceding 3 months. The only patient who had not been
Start infection control measures
Ban on all
Reintroduction of fluoroquinolones
2nd ban on
8 9
10 150
fluoroquinolones fluoroquinolones
5 6 7
75
2 3 4
0 1
Jan Feb March April May June July Aug Sept Oct Nov Dec Jan* Feb Months
0
2005 2006
DDD/100 bed-days per month
Cefuroxim IV Ciprofloxacin PO + IV Incidence CDI
CDI incidence per 10 000 admissions per month
FIG. 1.Course of the epidemic and dynamics of antibiotic use in St Jansdal Hospital. DDD, defined daily dose; PO, oral administration; IV, intra- venous administration.
4 Clinical Microbiology and Infection CMI
ª2009 The Authors
Journal Compilationª2009 European Society of Clinical Microbiology and Infectious Diseases, CMI
Figure 4. Course of the epidemic and dynamics of antibiotic use in St Jansdal Hospital. DDD, deMined daily dose; PO, oral administration; IV, intravenous administration.
Use of Mluoroquinolones cannot be the only explanation for the rapid spread of the type 027 throughout hospitals. As was described in chapter 6, almost 70% of the patients with CDI caused by type 027 had not used Mluoroquinolones. Other studies point to increased spore production of the type 027 strain which, because it increases environmental contamination, probably facilitates the spread of the strain throughout hospitals. The role of the environment in C. dif-icile transmission has been noted in investigations of CDI in both Western Europe and
the United States 16. A very recent study by Merrigan et al., showed that the strains in the hyper virulent (type 027) clade of C. dif-icile not only sporulated earlier and with greater efMiciency than other strains, but also produced robust amounts of toxin 17. Another recent study by Dumford et al.. demonstrated how broadly C. dif-icile spores are disseminated. During an outbreak of the type 027 strain in their hospital, C. dif-icile spores were detected on
computer keyboards, nursing stations and 31% of physician work areas 18.
Mortality associated with the type 027 strain is high. In chapter 6, it was observed that all cause mortality in an outbreak setting with type 027 was 26% after 30 days, while
mortality due to non-‐outbreak strains was 3% after 30 days. Although this large difference will undoubtedly be largely associated with the severity of other co-‐morbidities, the difference was still signiMicant after adjustment for age, ward and Charlson co-‐morbidity score, which measures chronic disease severity at baseline. The attributable mortality rate of 7% among patients with CDI caused by type 027 is comparable to the rates that have been reported in the literature (table 1) 1. In one of the Mirst studies that described a large outbreak with type 027 in Canada, it was also clear that the attributable mortality rate increased with age, along with the incidence (table 1) 1 .
nity until recently. In a typing study of isolates from hospital-onset and health care facility–associated CDI, 54% of these isolates and 50% of community-associated C difficile isolates were NAP1 (the predominant type).
9In a Foodborne Diseases Active Surveillance Network surveil- lance study, 22% of 60 community-associated–C difficile isolates from 5 states were NAP1, suggesting that this epidemic strain can be isolated from patients in the community as well as in the hospital.
10Nonetheless, in the community, CDI caused by any strain is relatively rare compared with the rate in hospitals.
Potential community sources of CDI include soil; salt, fresh, and tap water; pet animals and food animals; and meats and vegetables.
11The BI/NAP1/027 strain was identified in meat in Canada and in Arizona.
12,13How- ever, the most common C difficile strains found in animals and meat are toxinotype V, PCR ribotype 078, and REA type BK.
12–14These strains share with type BI/NAP1/027 the presence of binary toxin and tcdC gene deletions. To date there has been no conclusive evidence that consum- ing food contaminated with C difficile has led to clinical CDI in human beings. The means by which the current epidemic BI/NAP1/027 strain has become so widely dis- tributed in multiple countries so quickly has not been determined, but a common vehicle such as food remains an enticing avenue for further research.
Patient Risk Factors
The greatest risk factors for CDI caused by the BI/NAP1/027 strain are advanced patient age, hospital- ization, and exposure to specific antimicrobials, espe- cially fluoroquinolones and cephalosporins.
2,6,15,16The specific fluoroquinolones that have been identified as risk factors include levofloxacin, moxifloxacin, gatifloxacin, and ciprofloxacin, presumably as a result of the fluoro- quinolone-resistance present in the epidemic strain.
Cephalosporin antibiotics, to which virtually all C difficile strains are resistant, have also been implicated as a risk in hospitals in which the epidemic strain is present, includ- ing use for surgical prophylaxis.
2,17Exposures to stomach acid–reducing agents such as histamine type 2 blockers and proton pump inhibitors have been identified incon- sistently as risk factors for CDI in hospitals in which the
epidemic strain is present.
2,16,18Other than the new flu- oroquinolone resistance, the organism-specific factors that have enabled the BI/NAP1/027 strains to be so successful recently in their dissemination and ability to cause severe CDI remain largely speculative but are the subject of intensive research efforts.
Organism Virulence Factors Regulation of Toxin A and Toxin B Production
All virulent C difficile strains carry a 19.6-kb patho- genicity locus (PaLoc), which contains 5 genes (Figure 3):
tcdR, tcdB, tcdE, tcdA, and tcdC The tcdA and tcdB genes encode toxins A and B, respectively; these toxins are monoglucosyltransferases that modify rho proteins in host cells leading to collapse of the actin cytoskeleton and cell death.
19The tcdE gene encodes a holin-like pro- tein that is proposed to facilitate release of toxins A and B from the bacterial cell, because these toxins do not possess signal peptides.
20The regulation of toxin production in C difficile is not completely characterized, but current data show that toxin expression is controlled by the positive regulator, TcdR, its antagonist, TcdC, and the global regulator, CodY (Figure 3).
21–24During logarithmic-phase growth, tcdR, tcdB, tcdE, and tcdA are expressed (at a low level) from a single transcript that originates from a putative promoter upstream of tcdR, at a rate proportional to cell density, whereas the expression of tcdC is maximal (Figure 3A).
25In stationary phase, the expression of tcdC de- creases, whereas tcdR, tcdA, and tcdB expression increases.
During this time, tcdR, tcdA, and tcdB are transcribed from individual promoters in a TcdR-dependent manner (Fig- ure 3B).
22,23,25,26TcdR is a 22-kDa protein that is part of group 5 of the sigma 70 factor family, which has similarities to extracy- toplasmic function–like sigma factors.
27,28This protein binds to the RNA polymerase holoenzyme
22to facilitate expression of tcdR, tcdA, tcdB, and probably tcdE. TcdC, a 26-kDa protein, is likely to be an antisigma factor that acts as a negative regulator of toxin production.
24,29The tcdC structural gene is located downstream of the other 4 PaLoc genes and is encoded on the opposite strand (Fig- ure 3).
25TcdC is a membrane-associated protein that most likely interacts with and sequesters TcdR to inhibit transcription of the toxin genes.
24,30Previous analysis of tcdC polymorphisms among C dif- ficile clinical isolates found that most strains that carried nonsense mutations or deletions in tcdC also had variant tcdA and tcdB genes.
31The investigators observed that these strains were persistent in the hospital population during a 2-year period, were responsible for outbreaks of CDI, and conferred higher levels of in vitro cytotoxicity than some other clinical isolates. Thus, it was proposed that the variant tcdC genes in these toxin-variant strains may have been responsible for their higher levels of cy-
Table 1. CDI Rates and Mortality Increase With Patient AgeAge
CDI rate per 1000 admissions
Attributable 30-day mortality rate, %
!40 y 3.5 2.6
41–50 y 11.2 1.2
51–60 y 20.0 3.2
61–70 y 24.4 5.1
71–80 y 38.3 6.2
81–90 y 54.5 10.2
"90 y 74.4 14.0
Adapted from Loo et al.2
May 2009 CDI CAUSED BY EPIDEMIC BI/NAP1/027 STRAIN 1915
Table 1. CDI rates and mortality increase with age 1
160
To date, the signiMicance of this new epidemic strain is not completely deMined. The type 027 strain also seems to be associated with infection in individuals not previously considered at risk, including young and previously healthy persons who were not exposed to a healthcare environment or antimicrobial therapy 19,20. Furthermore, a study in Canada showed an
increase in mortality among patients infected with type 027-‐CDI, particularly patients
between the ages of 60 and 90 years 21. In addition, a recent case control study by Sundram et al. reported that early mortality from CDI was four times higher with ribotype 027 strain than with the next most common strain (ribotype 106) in their hospital. Ribotype 027 strains constituted 45% of the C. dif-icile isolates in their hospital 22. Contrary to these observations are two studies that have shown that, in a non-‐epidemic setting, the type 027 strain is not associated with more severe disease 23,24. To elucidate these issues, there is a need for studies that investigate the incidence and the exact cause of death at longer term follow-‐up, among patients infected with these strains.
Epidemic CDI
In chapter 6, general risk factors for CDI were investigated, in comparison to patients who had no CDI and in comparison to patients who had diarrhea that was not caused by CDI.
Risk factors for CDI included increased co-‐morbidity, hematological malignancy, nasogastric intubation and use of antibiotics, especially high exposure to cephalosporins and clindamycin.
These factors have previously been recognized, although studies lacked appropriate control groups of non-‐CDI diarrheal patients 25-‐28. Risk factors for diarrhea in general were prior abdominal surgery, co-‐existing diseases of the digestive system and low exposure to 1st generation cephalosporins (i.e prophylactic use). An interesting observation in the study described in chapter 6 was the fact that use of cephalosporins was a risk factor for CDI, regardless which PCR-‐ribotype was involved. By contrast, use of clindamycin was a very strong risk factor for CDI caused by type 017, but not for other types and use of
Mluoroquinolones was a very strong risk factor only for CDI caused by type 027. This Minding underscores the importance of culturing strains of Clostridium dif-icile in the case of an increased incidence of CDI, in order to perform antibiotic susceptibility testing (and possibly PCR-‐ribotyping, to recognize outbreaks). This information will help in choosing a sensible antibiotic restriction policy, if needed (for example: a restriction in the use of clindamycin in the case of an outbreak with type 017, versus a restriction in the use of Mluoroquinolones in the case of an outbreak with type 027).
A worrying development is the occurrence of CDI due to clindamcyin-‐resistant type 027 strains that has been observed in outbreaks in Ireland, as described in chapter 3. Resistance to this antimicrobial agent increases the risk for CDI in patients. Another problem is that
clindamycin is an important alternative when cephalosporins can not be prescribed to patients, for example because of a ban on these antibiotics during an outbreak of CDI. If
physicians would then prescribe clindamycin, without further knowledge about the resistance proMile of the Clostridium dif-icile strain that is involved, an outbreak could further be
facilitated. Its use would then be an important factor contributing to the persistence and spread of the outbreak. These clindamcyin-‐resistant type 027 strains probably reMlect the emergence of a new clone, because MLVA clearly differentiated between clindamycin-‐
susceptible and -‐resistant isolates.
Chapter 6 shows that when mortality is measured, without taking the involved PCR-‐
ribotype into account, mortality of outbreak strains will be under-‐estimated and mortality caused by non-‐outbreak strains will be over-‐estimated. A remarkable Minding of the study described in chapter 6 was, that when the PCR-‐ribotype that caused CDI was not taken into account, the mortality rate for these patients was high (17% after 30 days). However, when only the subgroup of the outbreak types was regarded, mortality was much higher. Among patients with CDI caused by type 027, all-‐cause mortality was 26% after 30 days and among patients with CDI caused by type 017, it was 23% after 30 days. The high disease burden of
162
nosocomial CDI at longer term follow-‐up has been mentioned earlier in an observational study 29. In that study however, no control group was included. In a very recent article by Oake et al., the independent impact of hospital acquired CDI on in-‐hospital mortality was investigated, after adjusting for the time-‐varying nature of CDI and baseline mortality risk at hospital 30. On average, patients with CDI had a 3-‐fold increase in the hazard of death. In this study, the strain-‐type that caused CDI was not taken into account. However the results of this study match those that we found for the outbreak strains, types 017 and 027, which suggests that our Mindings are probably not unique for this hospital.
In the outbreak setting described in chapter 6, C. dif-icile type 017 was associated with similar clinical presentation and outcomes as type 027. This was surprising, because type 017 lacks the toxin A gene and contains none of the proposed virulence markers typical for type 027. To date, outbreaks with type 017 have been associated with mortality rates between 4%
and 7% after 30 days, which is considerably lower than the rate that we found in our study
31-‐34. We hypothesize that yet unknown virulence markers might be involved, such as variants of TcdB, or non–toxin-‐related virulence factors 17,35-‐37. In a very recent study applying
comparative genome analysis of 14 sequences strains, SNPs that were found in 2 candidate genes with yet-‐unknown functions were associated with severe CDI 38. Interestingly, these SNPs were found to be present among type 027 strains, but also among type 017 strains that lacked toxin A.
Endemic CDI
An interesting observation in the study described in chapter 6 was that, among patients with non-‐outbreak strains, the all-‐cause mortality was much lower (3% after 30 days) and comparable to the mortality that was found among control patients. The implication of this observation is that a high mortality is mainly related to outbreak strains and/or outbreak settings, but not to endemic settings, as was observed in the study described in chapter 7. In
this endemic setting, without outbreaks or circulating strains that have earlier been
associated with increased mortality, the mortality among CDI patients after one month was 8%, which is much lower that the rates that have been described in outbreak situations
1,4,12,39,40. CDI patients described in this study were however more severely ill than non-‐CDI diarrheal patients, as illustrated by a higher leukocyte count and a higher 30-‐day mortality.
The analysis of risk factors in this endemic setting shed more light on some observations that were also made in chapter 6. Again, it was found that use of (second generation)
cephalosporins was a strong risk factor for CDI 1,41,42. On the other hand, since there were no circulating outbreak strains, use of Mluoroquinolones or clindamycin did not increase the risk for patients to contract CDI, nor did the use of proton pump inhibitors. Earlier studies that investigated the use of PPIs in association with CDI produced conMlicting conclusions 43,44. In our study, half of the non-‐CDI and control patients also used PPIs. It could be possible that use PPIs is only a weak risk factor with regard to CDI, only to be considered important in outbreak settings.
Regarding endemic versus outbreak settings, the problem exists that there are no clear deMinitions for what they imply. The endemic incidence of CDI in The Netherlands is around 18 per 10.000 admissions 11. However, the endemic incidence that has been reported in the United States was 106 per 10,000 hospital admissions, which is a factor 5 higher 41. It is very well possible that these two “endemic settings” hold different risks with regard to the
development of CDI.
Measuring dissemination of Clostridium dif-icile
MLVA had been described as a very reliable tool to type and subtype Clostridium dif-icile
45-‐48. In a study by Killgore et al., the discriminatory power was investigated of seven DNA Mingerprinting techniques, when applied to 42 C. dif-icile strains collected in four countries.
164
Only REA and MLVA had sufMicient power to distinguish strains from different outbreaks, and to discriminate between North American and European type 027 isolates 47.
In chapter 3, clonal spread of CDI caused by type 017 in a hospital in Argentina was investigated by MLVA. This clonal expansion took place over period of 5 years, during which an increase in incidence of CDI was noticed. Outbreaks with this type have been described in other parts of the world, the most recent being the outbreak described in chapter 6 31,33,50. Application of MLVA resulted in interesting information. First, MLVA was capable of discriminating 57 unique MLVA types among 71 type 017 isolates. Of these 71 isolates, 56 were from the hospital in Argentina en 15 originated from other parts of the world. It was found that 75% of the Argentinean type 017 isolates were genetically related and that they were quite distinct from type 017 isolated that originated from other parts of the world. In addition, all clonal complexes were country-‐speciMic (with the exception of one Canadian isolate).
Second, among the Argentinean isolates, clonal complexes did not show a correlation over time, but were found to be restricted to speciMic wards. This observation suggested the continuous presence of clones on these wards, providing a source for new infections at
various time intervals. This was a good example of the practical use of MLVA in understanding patterns of clonal dissemination.
The same was found in chapter 6, where with MLVA, it was possible to discern a pattern of clonal dissemination of types 027 and 017, indicating transmission of spores from the environment or asymptomatic patients. Transmission occurred despite appropriate infection control measures, with prolonged presence of clones on certain wards and throughout the hospital, up to time periods of more than a year.
In chapters 4 and 5, MLVA was applied among strains with identical PCR-‐ribotypes from various hospitals. Chapter 4 described the emergence of clindamycin-‐resistant type 027 strains. An interesting observation was that 10 out of 16 clindamycin-‐resistant strains were
found within a clonal complex, while these strains originated from 5 different hospitals. Van den Berg et al. demonstrated that MLVA is a very discriminatory typing technique for type 027 strains and that there was much variability among these strains 50. The fact that the
clindamycin resistant strains formed a large clonal complex, suggested that transmission of strains between different hospitals is common.
In chapter 5, the emergence of CDI caused by type 078 was described. It was speculated that regarding this strain, humans and animals share a common source. Reason for this speculation was the fact that type 078 strains from humans an animals were very similar, to the point of clonality. The high degree of genetic relatedness among type 078 isolates was surprising, because previous studies had revealed a high variation among type 017 and 027 strains 50-‐52. An important question was whether this could be extrapolated to other PCR-‐
ribotypes, for example type 078. Recently, Bakker et al answered this question by re-‐iterating all the steps regarding the use of MLVA to (sub)type type 078 strains 53. They found MLVA to be a reliable method for typing of 078 strains and corroborated by MLVA the high genetic relatedness of 102 human and 56 porcine type 078 strains from 4 European countries. This can indicate that type 078 has not been part of the spectrum of human CDI for a long enough time to develop a more-‐remote relatedness to porcine strains (or vice versa) or that it is very stable with regard to mutations. The latter is not likely, because Stabler et al. 54 observed that C. dif-icile readily undergoes genetic exchange. Using whole genome analysis, they also
hypothesized that human strains arose from those found in pigs on the basis of identiMication of a toxinogenic clade containing porcine, bovine, and human isolates.
The notion of a common source was also supported by the fact that a signiMicant number of both human and porcine strains were resistant to tetracycline, and that these strains
contained the mobile element Tn916-‐like transposon. Interestingly, this element has also been described for tetracycline-‐resistant enterococci from human and porcine origins 55.
166
Interspecies transmission or transmission through meat products have been suggested as sources of infection, but these sources have yet to be established 56-‐58.
Finally, in chapter 3, MLVA was used to discriminate between recurrence and re-‐
infection. The problem here was that when a patient was stricken by a second episode of CDI, this could have been either a new infection (re-‐infection), especially in epidemic settings with a high incidence of CDI, or the second episode could have been caused by the same strain as the one that caused the Mirst episode, because it had not been eradicated from the patients’ gut (relapse). The Minding that a major proportion of recurrences was caused by re-‐infection rather than relapse has been reported before 59,60. In chapter 3, it was observed, by use of MLVA, that a majority (56%) of recurrences was caused by a different strain (re-‐infection), despite identiMication of the identical PCR-‐ribotype in initial infection and recurrence. This observation may be of clinical relevance, because the treatment approach towards a re-‐
infection with a different strain differs from the approach for recurrent infections with the same strain. However, it remains difMicult to distinguish between a re-‐infection with a different strain and a relapse with the same endogenous strain. The Mirst could represent a relapse with a different previously unrecognized endogenous strain; the second could represent a re-‐
infection from the environment with the same circulating strain. To complicate matters even more, van den Berg et al. 61 showed that different PCR ribotypes could be found
simultaneously in stool samples of two of 23 patients with CDI. By contrast, O’Neill et al. found that all 10 cultured colonies of each of 10 patients with a Mirst episode of CDI contained the same REA type 59. In all, the Minding that the majority of recurrences were re-‐infections with a different 017 MLVA type, supports the hypothesis that the environment and possibly other patients or healthcare workers contributed to the mode of transmission signiMicantly. This is also supported by the work of Riggs et al, who found that spread through asymptomatic carriers was an important risk factor for nosocomial transmission 62.
Epidemiology revisited: type 078, humans and animals
In chapter 5, the emergence of a novel strain of C. dif-icile across The Netherlands was described. This novel strain belongs to PCR-‐ribotype 078 (type 078) and toxinotype V and was found to have similar virulence characteristics as the type 027 strain, such as the presence of tcdA, tcdB, and binary toxin genes and also a mutation (C184T) in the regulatory tcdC gene,
that results in a premature stop codon and a non-‐functional TcdC-‐protein. This mutation in toxinotype V strains had been reported elsewhere, but had never been ascribed to type 078
63,64. When the study was published, type 078 was the second-‐most frequently encountered type (13%) in The Netherlands. By June 2009, it still was the third most frequently found type
11. By contrast, in 2005, type 078 was the eleventh-‐most frequently found type in Europe 65. In a recent study by Bauer et al., 395 isolates from 73 hospitals in 26 countries were investigated by PCR-‐ribotyping. Among the 65 different PCR ribotypes that were identiMied, type 078 was the fourth most frequently encountered type (Migure) 66. From these data it can be concluded that the increase in incidence of the type 078 strain is a phenomenon that has also occurred in other parts of Europe.
The intriguing aspect about this type is the fact that it is predominant type among pigs, calves and horses 63,67-‐71. This is depicted in table 2.
Host Number of PCR-
ribotypes found Most prevalent
PCR-ribotypes References
Humans Approximately 200 014, 020, 001, 078 Bauer et al 66 Horses 10 to 12 078 (up to 35%) Keel et al 67, Arroyo et al 69
Calves 3 to 8 078 (up to 94%) Rodriguez-‐Palacios et al 71, Hammit et al 63 , Keel et al 67 Piglets 2 to 4 078 (up to 83%) Keel et al 67, Pirs et al 70
Table 2. Predominant PCR-‐ribotypes of Clostridium dif-icile per species.
168