Renal Failure and Leukocytosis Are Predictors of a Complicated Course of Clostridium difficile Infection if Measured on Day of Diagnosis.

2012·0.12 MB·English

by  Bauer, Martijn P.

SUPPLEMENT ARTICLE
Renal Failure and Leukocytosis Are Predictors
Clostridium difficile
of a Complicated Course of
Infection if Measured on Day of Diagnosis
MartijnP.Bauer,1,aMarjoleinP.M.Hensgens,1,aMarkA.Miller,2DaleN.Gerding,3,4MarkH.Wilcox,5,6AdamP.Dale,5
WarrenN.Fawley,5EdJ.Kuijper,1andSherwoodL.Gorbach7,8
1LeidenUniversityMedicalCenter,TheNetherlands;2DivisionofInfectiousDiseases,JewishGeneralHospital,McGillUniversity,Montreal,Quebec,
Canada;3HinesVeteransAffairsHospital,4LoyolaUniversityChicagoSchoolofMedicine,Maywood,Illinois;5DepartmentofMicrobiology,The
GeneralInfirmary,OldMedicalSchool,6LeedsInstituteofMolecularMedicine,UniversityofLeeds,UnitedKingdom;7OptimerPharmaceuticals,Inc.,
SanDiego,California,and8TuftsUniversitySchoolofMedicine,Boston,Massachusetts
NonsevereClostridiumdifficileinfection(CDI)andsevereCDI,whichcarriesahigherriskthannonsevereCDI
fortreatmentfailureandCDIrecurrence,aredifficulttodistinguishatthetimeofdiagnosis.Toinvestigatethe
prognostic value of 3 markers of severe CDI suggested by recent guidelines (fever, leukocytosis, and renal
failure),weusedthedatabaseof2randomizedcontrolledtrials,whichcontainedinformationfor1105patients
with CDI. Leukocytosis (risk ratio [RR], 2.29; 95% confidence interval [CI], 1.63–3.21) and renal failure (RR,
2.52; 95% CI, 1.82–3.50) were associated with treatment failure. Fever, although associated with treatment
failure (RR, 2.45; 95% CI, 1.07–5.61), was rare. Renal failure was the only significant predictor of recurrence
(RR,1.45;95%CI,1.05–2.02).Differenttimingofmeasurementsofleukocytecountandserumcreatininelevel
aroundtheCDIdiagnosisledtoadifferentseverityclassificationinmanycases.Inconclusion,bothleukocyto-
sisandrenalfailureareusefulpredictors,althoughtimingofmeasurementisimportant.
Clostridium difficile infection (CDI) has become an complicationsorrecurrencescanguidethechoiceand
increasing problem in many hospitals in the Western duration of therapy. In 2009, a prediction rule for
world during the past decade. C. difficile causes diar- recurrences, incorporating age, comorbid conditions,
rheaandcolitis,withatendencytorecurafterinitially andthenecessitytocontinueincitingantibiotictherapy,
successful antimicrobial therapy. Furthermore, gut waspublished[1].Thisrulewasderivedfromandwas
inflammationmaybesoseverethatantimicrobialthera- validated in 2 cohorts of 44 and 64 patients, respect-
py is not effective; in such cases, complications such ively. The relatively small sample sizes challenge the
ashypotension, perforation,and toxicmegacolon may credibility of this rule. Several risk factors for compli-
develop. Several risk factors for CDI have been identi- cations of CDI and prediction rules based on these
fied,ofwhichtheuseofantibioticsisthemostimpor- factors have been described, but unfortunately, none
tant.Predictingwhichpatientsareatriskfordeveloping ofthesepredictionruleshavebeenvalidated[2–6].
Thechoiceofanappropriateendpointforapredic-
tion rule for complicated and/or recurrent CDI has
aM.P.B.andM.P.M.H.contributedequallytothiswork.
beenproblematic.Theclinicaljudgmentofwhetherto
Correspondence: E. J. Kuijper, Department of Medical Microbiology, Leiden
University Medical Center, PO Box 9600, 2300 RC, Leiden, The Netherlands attributeendpointssuchasCDI-relatedmortalityand
([email protected]).
intensivecareunitadmissionmaybehighlysubjective,
ClinicalInfectiousDiseases 2012;55(S2):S149–53
especially for elderly patients, who are often admitted
©TheAuthor2012.PublishedbyOxfordUniversityPressonbehalfoftheInfectious
DiseasesSocietyofAmerica.Allrightsreserved.ForPermissions,pleaseemail: with severeillness and usuallyhavesignificantcomor-
[e-mail]edistributedunder
bid conditions. End points concerning the resolution
thetermsoftheCreativeCommonsAttributionNon-CommercialLicense(http://
creativecommons.org/licenses/by-nc/3.0), which permits unrestricted non- andrecurrenceofdiarrheaneedaprecisedefinitionof
commercialuse,distribution,andreproductioninanymedium,providedtheoriginal diarrheaandquantitativemeasurementofstoolvolume
workisproperlycited.
DOI:10.1093/cid/cis340 and frequency, which may be difficult to obtain.
(cid:129) (cid:129)
PredictorsofComplicatedCDI CID 2012:55 (Suppl2) S149
Furthermore, the parameters included in a prediction rule For patients with rectal collection devices, volume was
should be objective, routinely measured in clinical practice, convertedtonumberofUBMsbydividingthevolumeby60mL
andavailableatthemomenttheruleisapplied(ie,whenCDI androundinguptothenearestwholenumber.Attheend-of-
isdiagnosed). therapy visit, an investigator assessed the success of therapy.
ArecentguidelinebytheSocietyforHealthcareEpidemiol- Clinical failure was defined as the persistence of diarrhea, the
ogy of America and the Infectious Diseases Society of need for additional therapy for CDI, or both on the basis of
America recommends that age, peak leukocyte count, and theopinionoftheinvestigator[10].RecurrenceofCDI(deter-
peak serum creatinine level be taken into account as potential mined by use of the same criteria as for enrollment [ie, >3
indicators of a complicated course of CDI when treatment is UBMsper24hoursandapositivestooltoxintestresult])was
started [7]. The European Society for Clinical Microbiology assessed during the mean follow-up duration (±SD) of 28±2
and Infectious Diseases has issued a guidance document for days after completion of therapy. At enrollment, temperature,
thetreatmentofCDIthatalsolistsqualitativeandquantitative leukocytecount,andserumcreatininelevelwerecollected.
symptoms,signs,laboratoryparameters,andradiologicalfind- To assess whether the timing of laboratory measurements
ings that may reflect more severe disease with associated couldinfluencetheirprognosticvalue,weusedthedatabaseof
higherriskforcomplicationsandrecurrences[8].Threequan- a prospective cohort study performed at Leeds Teaching Hos-
titative parameters for diagnosing severe colitis were included: pitalin2007.Inthisdatabase,104consecutiveadultinpatients
body temperature >38.5°C, leukocyte count >15×109/L, and with CDI (defined on the basis of the presence of unformed
serum creatinine level >50% above baseline; however, these stoolandapositiveC.difficiletoxintestresult)wereincluded.
cutoffshavenotbeenconfirmedprospectively. On days −3 to +3 relative to day 0 (the day the diarrheal
Inthepresent study,wesoughttoinvestigatethevalue of3 sample was collected), leukocyte count and serum creatinine
quantitativeseveritycriteriainpredictingthefailureofantimi- level were recorded. Data from a minimum of 2 leukocyte
crobial therapy and the recurrence of CDI after initially suc- counts and creatinine levels on different days were required
cessful treatment. Furthermore, we aimed to investigate forpatientstobeincludedintheanalyses.
whether leukocyte count and serum creatinine level fluctuate In both analyses, we defined fever as a core body tempera-
earlyinthecourseofaCDIepisodeandthereforewhetherthe ture >38.5°C and leukocytosis as a leukocyte count >15×109
timing of their measurements can influence whether severity leukocytes/L. Because the pre-CDI serum creatinine level was
criteriaaremet.Forouranalyses,weusedthedatabasefrom2 not known foreach patient, we substituted the 50% creatinine
largerandomizedclinicaltrialsthatusedastrictobjectivedefi- level increase with a fixed value of the creatinine level
nition of diarrhea and the database of a prospective single- >133μmol/L (>1.5mg/dL). This served as a proxy for renal
center cohort study that recorded sequential leukocyte counts failure.
andserumcreatininelevelsaroundthedateofCDIdiagnosis.
Analyses
METHODS The intention-to-treat population that received at least 1 dose
ofstudymedicationwasusedfortheanalysis.Distributionsof
Databases the continuous variables of temperature, leukocyte count, and
The database from 2 randomized controlled phase III trials creatinine level were compared for patients with and patients
comparing vancomycin with fidaxomicin for the treatment of without clinical treatment failure and recurrence. Nonnor-
CDI wasused toassessthepredictivevalue offever,leukocyte mally distributed variables were compared with the Mann-
count, and serum creatinine level [9, 10]. Patients were Whitney U test. Proportions were compared with the χ2 test.
recruited in the United States, Canada, and Europe (Clinical- Riskratios(RRs)and 95%confidenceintervals(CIs)werecal-
Trials.gov registry number NCT00314951, April 2006–July culated for the associations of fever, leukocytosis, and renal
2008, United States and Canada; and ClinicalTrials.gov failure with the outcome parameters. Kaplan-Meier survival
registry number NCT00468728, April 2007–November 2009, curves were constructed to investigate the association of fever,
United States, Belgium, Canada, France, Germany, Italy, leukocytosis, and renal failure with the time to resolution of
Spain, Sweden, and United Kingdom). Patients with CDI, diarrhea(expressedinhoursfromthefirstdoseoffidaxomicin
defined on the basis of diarrhea (>3 unformed bowel move- or vancomycin). The log-rank test was used to test the differ-
ments [UBMs] per day) and a positive stool toxin test for ence between the survival curves. Cox regression was used to
C. difficile, were randomly assigned to receive 125mg of van- calculatehazardratios(HRs)with95%CIs.Receiveroperating
comycin 4 times daily or 200mg of fidaxomicin twice daily characteristic curves were constructed to assess the validity of
for 10 days. The numbers and times of UBMs were recorded the cutoffs used to define categorical variables. Variability of
during treatment and for 2 days after an end-of-therapy visit. leukocyte counts and serum creatinine levels were compared
(cid:129) (cid:129)
S150 CID 2012:55 (Suppl2) Baueretal
within patients and expressed in absolute differences. All ana- Table 1. Determinants of Clinical Treatment Failure Among
lyses were carried out in SPSS for Windows software, version PatientsWithClostridiumdifficileInfection
17.0(SPSS,Chicago,IL).
Variable MedianValue IQR Pa
RESULTS Continuous,outcome
Temperature(°C)
There were 1105 patients with CDI in the clinical trial data- Failure 36.8 36.4–37.2 .180
base. Patients treated with vancomycin (n=566) or fidaxomi- Cure 36.7 36.4–37.1
Leukocytecount(×109leukocytes/L)
cin (n=539) had similar median values for temperature,
Failure 10.5 6.8–17.4 .002
leukocyte count, and serum creatinine level and were evenly
Cure 8.9 6.5–12.1
distributed across the groups with respect to dichotomized
Creatininelevel(μmol/L)b
continuous variables (data not shown). Fever was rare; only
Failure 80 62–150 .005
1.2% of patients (13 of 1102) had atemperature >38.5°C. The Cure 71 62–97
median treatment duration was 11 days for each group.
Categorical,cutoff Failurec RRd 95%CI
Overall, 143 patients (13%) experienced clinical treatment
Fever,temperature
failure at the end of treatment. Of the 962 patients who were
>38.5°C 4/13 2.45 1.07–5.61
cured after treatment, 194 (20%) experienced recurrence a
≤38.5°C 137/1089
mean(±SD)of28±2daysaftertreatment.
Leukocytosis,leukocytelevel
The median leukocyte count and creatinine level were sig- >15×109leukocytes/L 38/153 2.29 1.63–3.21
nificantly higher in patients with clinical treatment failure; ≤15×109leukocytes/L 90/829
temperature distributions in patients with and those without Renalfailure,creatininelevel
treatment failure were almost identical. In addition, dichoto- ≥133μmol/Lb 41/160 2.52 1.82–3.50
mous categories of fever, leukocytosis, and renal failure all <133μmol/Lb 91/896
showed significantcorrelation withtreatment failure (Table1).
Abbreviations:CI,confidenceinterval;IQR,interquartilerange;RR,riskratio.
Themediancreatininelevelwassignificantlyhigherinpatients aComparisonbetweenpatientswithclinicaltreatmentfailureandthosewith
with recurrence, and this parameter was the only significant clinicalcure.
bCreatinine conversion: 1μmol/L is equal to 0.0113mg/dL. Therefore,
predictorofrecurrence(Table2).Differentcutoffsforthecon-
133μmol/Lisequalto1.50mg/dL.
tinuous variables of temperature, leukocyte count, and creati- cDataareno.ofpatientswithfailure/overallno.
ninelevel,assessedbyreceiveroperatingcharacteristics,didnot dFortheassociationwithfailure.
leadtohigherrelativerisksandthereforebetterperformancein
thepredictionofclinicaltreatmentfailureorrecurrentCDI.
The probability of resolution of diarrhea within 10 days of whereas only 56 of 403 (13.9%) experienced a recurrence after
treatment was slightly lower in patients with renal failure, successful fidaxomicin treatment (P<.001). In patients with
compared with patients without renal failure (HR, 0.83; 95% renalfailureatbaseline,fidaxomicintherapywasassociatedwith
CI, .68–1.02; Figure 1). Neither fever (HR, 1.08; 95% CI, a 60% reduction in the frequency of recurrences (8 of 54
.61–1.91) nor leukocytosis (HR, 1.02; 95% CI, .84–1.24) was [14.8%]) relative to vancomycin (24 of 65 [36.9%]; P=.007).
associated with a lower probability of resolution of diarrhea. Likewise,inpatientscategorizedashavingleukocytosisorsevere
Although creatinine level distributions were similar between CDI, the incidence of recurrence was more than double for
patients treated with fidaxomicin and those treated with van- patients cured with vancomycin, compared with those treated
comycin, we repeated the analysis of renal failure as a predic- successfullywithfidaxomicin(P<.01foreachcomparison).
tor of resolution of diarrhea stratified according to treatment Becauseleukocytosisandrenalfailureatthetimeofdiagno-
group and found similar results (vancomycin: HR, 0.80 [95% sis were shown to be the strongest predictors, we investigated
CI, .61–1.05]; fidaxomicin: HR, 0.88 [95% CI, .66–1.19]). the stability of these parameters during a 6-day interval
Because recurrences occurred less often in patients treated around diagnosis. In the population from the database of
withfidaxomicin,theCIiswidestinthatgroup. Leeds Teaching Hospital, the highest mean leukocyte count
Clinicaltreatmentfailureratesweresimilarinthefidaxomicin (13.4×109leukocytes/L)wasfoundonthedayofCDIdiagno-
and vancomycin treatment groups regardless of clinical status, sis. Within the interval from 3 days before to 3 days after the
using the 3 severity factors. Recurrence was significantly more diagnosis of CDI, the mean difference between the highest
frequentfollowingvancomycintreatment,comparedwithfidax- and lowest leukocyte counts was 6.4×109 leukocytes/L.
omicin treatment. In patients without renal failure, 93 of 402 Twenty of 86 patients (23.3%) had a minimum to maximum
patients(23.1%)curedbyvancomycintherapyhadarecurrence, leukocyte count range >10×109 leukocytes/L, and 33 (38.4%)
(cid:129) (cid:129)
PredictorsofComplicatedCDI CID 2012:55 (Suppl2) S151
Table 2. Determinants of Clostridium difficile Infection
Recurrence
Variable MedianValue IQR Pa
Continuous,outcome
Temperature(°C)
Norecurrence 36.7 36.4–37.1 .827
Recurrence 36.7 36.4–37.0
Leukocytecount(×109leukocytes/L)
Norecurrence 8.8 6.5–12.1 .276
Recurrence 9.1 6.6–12.8
Creatininelevel(μmol/L)b
Norecurrence 71 62–97 .008
Recurrence 80 62–115
Categorical,cutoff Recurrencec RRd 95%CI
Fever,temperature
>38.5°C 1/9 0.55 .09–3.51
≤38.5°C 192/952
Leukocytosis,leukocytelevel
>15×109leukocytes/L 22/115 1.00 .67–1.50 Figure 1. Kaplan-Meier analysis of time to resolution of diarrhea for
≤15×109leukocytes/L 141/739 patients with andwithout renal failure. Thehazard ratiowas 0.83 (95%
Renalfailure,creatininelevel confidenceinterval,.68–1.02).
≥133μmol/Lb 32/119 1.45 1.05–2.02
<133μmol/Lb 149/805
infrequentinourstudytobeausefulpredictor,butitsassoci-
Abbreviations:CI,confidenceinterval;IQR,interquartilerange;RR,riskratio.
aComparison between patients with recurrence and those without atedrelativeriskwassignificant.
recurrence. In previous studies, leukocytosis and renal failure were also
bCreatinine conversion: 1μmol/L is equal to 0.0113mg/dL. Therefore, associatedwithcomplicationsandrecurrenceofCDI[3,11–13].
133μmol/Lisequalto1.50mg/dL.
cDataareno.ofpatientswithrecurrence/overallno. Therefore, both parameters could be suitable for evaluation in
dFortheassociationwithrecurrence. apredictionmodel.However,becauseofthevariablenatureof
these values around the time of CDI diagnosis, a strict defi-
nition is needed before incorporating these parameters in a
prediction rule. Early or late diagnosis could influence leuko-
patients had a minimum to maximum leukocyte count range
that included the cutoff of 15×109 leukocytes/L; therefore, a cyte count and serum creatinine level. Fever appeared not to
beausefulpredictoroffailureofCDItreatment.Thiswasalso
differenceintimingofasinglebloodsamplearounddiagnosis
could have led to a different severity classification. The mean shownbyasmallstudyin2007[14].
serum creatinine concentration was 147μmol/L on the day of Bothfeverandleukocytosisarethoughttoreflectmoresevere
inflammationofthe bowelwall.However, feverwastoo rare in
diagnosis. The mean minimum to maximum range in serum
creatinine values was 38.7μmol/L. Nineteen of 93 patients ourpatientpopulationtobeofuseasapredictor.Renalfailure
may reflect loss of effective circulating volume due either to
(20.4%)hadaminimumtomaximumrangeincreatininelevels
that included the cutoff of 133μmol/L, which could have led dehydrationbecauseofdiarrheaortoshockinthecontextofa
toadifferentclassificationinthecaseofdifferenttiming. systemic inflammatory response. Unfortunately, the predictive
value of these parameters may decrease because of underlying
illnesses and comorbid conditions. Renal failurewas present in
DISCUSSION 14%ofclinicalpatientsandwastheonlysignificantpredictorof
recurrence,anditwastheonlyparameterassociated,albeitnon-
Leukocytosis and renal failure were significant predictors of significantly,withalongertimetoresolutionofdiarrhea.Thus,
failure of CDI treatment. Only renal failure showed a trend creatininelevelmaybeagoodpredictor,alsobecauseofitsrela-
toward longer duration of diarrhea during treatment and was tivelygreaterstabilityaroundthetimeofCDIdiagnosisincom-
correlated significantly with recurrence after successful treat- parisontoleukocytosis.
ment. Both leukocyte count and serum creatinine level were Strengthsofthisstudyarethelargenumberofpatientswith
highly variable around diagnosis. Fever was found to be too CDI in the database with a well-described definition of
(cid:129) (cid:129)
S152 CID 2012:55 (Suppl2) Baueretal
diarrhea andaconsistentmeasureofUBMs.Onelimitationis 2. Fujitani S, George WL, Murthy AR. Comparison of clinical severity
that other potential predictors of severe CDI, such as age, score indices for Clostridium difficile infection. Infect Control Hosp
Epidemiol2011;32:220–8.
serum albumin level, or use of concomitant antibiotics, were
3. HenrichTJ,KrakowerD,BittonA,YokoeDS.Clinicalriskfactorsfor
notincludedinthisanalysis.Therefore,wewerenotabletode- severeClostridiumdifficile-associateddisease.EmergInfectDis2009;
velop acomplete risk score. Another limitation is the absence 15:415–22.
4. HubertB,LooVG,BourgaultAM,etal.Aportraitofthegeographic
of a baseline creatinine level for each patient, precluding us
disseminationoftheClostridiumdifficileNorthAmericanpulsed-field
fromdistinguishingbetweenchronicandacuterenalfailure. type1strainandtheepidemiologyofC.difficile-associateddiseasein
The results of our study suggest that both leukocytosis and Québec.ClinInfectDis2007;44:238–44.
5. Miller M, Gravel D, Mulvey M, et al. Health care-associated Clos-
renal failure predict clinical treatment failure, whereas only
tridium difficile infection in Canada: patient age and infecting strain
renalfailureisapredictorofrecurrenceaftertherapy.However,
typearehighlypredictiveofsevereoutcomeandmortality.ClinInfect
these predictors are highly dependent on the timing of their Dis2010;50:194–201.
determination, hampering their use in clinical practice. We 6. PépinJ,ValiquetteL,AlaryME,etal.Clostridiumdifficile-associated
diarrheainaregionofQuebecfrom1991to2003:achangingpattern
need betterand more closely defined predictorstoconstruct a ofdiseaseseverity.CMAJ2004;171:466–72.
reliable prediction score for complicated and recurrent CDI 7. CohenSH,GerdingDN,JohnsonS,etal.SocietyforHealthcareEpi-
thatisapplicableinclinicalpractice. demiologyofAmerica;InfectiousDiseasesSocietyofAmerica.Clini-
cal practice guidelines for Clostridium difficile infection in adults:
2010 update by the Society for Healthcare Epidemiologyof America
Notes (SHEA)andtheInfectiousDiseasesSocietyofAmerica(IDSA).Infect
ControlHospEpidemiol2010;31:431–55.
Acknowledgments. YinKeanprovidedstatisticalsupport. 8. Bauer MP, Kuijper EJ, van Dissel JT. European Society of Clinical
Supplementsponsorship. Thisarticlewaspublishedaspartofasupple- Microbiology and Infectious Diseases. European Society of Clinical
mententitled“FidaxomicinandtheEvolvingApproachtotheTreatmentof MicrobiologyandInfectiousDiseases(ESCMID):treatmentguidance
ClostridiumdifficileInfection,"sponsoredbyOptimerPharmaceuticals,Inc. document for Clostridium difficile infection (CDI). Clin Microbiol
Potential conflicts of interest. S. L. G. is a part-time employee of Infect2009;15:1067–79.
Optimer Pharmaceuticals, receiving honoraria from and owning stock 9. Crook D, Peto T, Miller M, et al. Efficacyand safety of fidaxomicin
options in Cempra. M. M. is a consultant for Optimer Pharmaceuticals. (FDX) vs vancomycin (VAN) in Clostridium difficile infection (CDI)
D.N.G.holdspatentslicensedtoViroPharmaforthetreatmentandpre- in 2 randomizedcontrolledtrials (RCT)with1105patients[abstract
ventionofCDI;isaconsultantforViroPharma,Optimer,Cubist,Merck, 1417]. Presented at: Infectious Diseases Society of America 48th
Pfizer, TheraDoc, Astellas, BioRelix, and Actelion; and holds research AnnualMeeting,21–24October2010,Vancouver,Canada.
grants from GOJO, Merck, Optimer, Sanofi Pasteur, Eurofins Medinet, 10. LouieTJ,MillerMA,MullaneKM,etal.OPT-80-003ClinicalStudy
andViroPharma.M.H.W.hasreceivedhonorariaforconsultancywork, Group.FidaxomicinversusvancomycinforClostridiumdifficileinfec-
financialsupporttoattendmeetings,andresearchfundingfromAstellas, tion.NEnglJMed2011;364:422–31.
Astra-Zeneca,Bayer,bioMerieux,Cerexa,Cubist,Nabriva,Novacta,Pfizer, 11. Moshkowitz M, Ben-Baruch E, Kline Z, Shimoni Z, Niven M,
Sanofi-Pasteur, Summit, The Medicines Company, and Viropharma. All Konikoff F. Risk factors for severity and relapse of pseudomembra-
otherauthorsreportnopotentialconflicts. nouscolitisinanelderlypopulation.ColorectalDis2007;9:173–7.
AllauthorshavesubmittedtheICMJEFormforDisclosureofPotential 12. PepinJ,AlaryME,ValiquetteL,etal.Increasingriskofrelapseafter
Conflicts of Interest. Conflicts that the editors consider relevant to the treatment of Clostridium difficile colitis in Quebec, Canada. Clin
contentofthemanuscripthavebeendisclosed. InfectDis2005;40:1591–7.
13. SailhamerEA,CarsonK,ChangY,etal.FulminantClostridiumdiffi-
References cile colitis: patterns of care and predictors of mortality. Arch Surg
2009;144:433–9;discussion439–40.
1. Hu MY, Katchar K, Kyne L, et al. Prospective derivation and vali- 14. BelmaresJ,GerdingDN,ParadaJP,MiskevicsS,WeaverF,JohnsonS.
dation of a clinical prediction rule for recurrent Clostridium difficile Outcome of metronidazole therapy for Clostridium difficile disease
infection.Gastroenterology2009;136:1206–14. andcorrelationwithascoringsystem.JInfect2007;55:495–501.
(cid:129) (cid:129)
PredictorsofComplicatedCDI CID 2012:55 (Suppl2) S153