211
Risk Factors for Candidemia in Patients with Acute Lymphocytic Leukemia Herve M. Riehet,* Antoine Andremont, Cyrille Tanerede, Jose Luis Pieo, and William R. Jarvis
From the Hospital Infections Program. Center for Infectious Diseases, Centers for Disease Control, Atlanta. Georgia; and the Service de Microbiologie Medicate and Departement d'Hematologie, Institut Gustave Roussy, Villejuif, France
Candidiasis and aspergillosis are the most common fungal infections in neutropenic patients; in fact Candida and Aspergillus species account for almost 90 % of all fungal infections reported in this population. Previous studies indicate that 10%-40% of patients who die with acute leukemia have evidence of disseminated fungal disease at autopsy; fungi are the cause of death in an estimated 21% of these patients [1-8]. Because acute lymphocytic leukemia (ALL) was the most common underlying disease in patients with candidemia at the Institut Gustave Roussy in 1987, the risk factors for candidemia in these patients were analyzed.
Methods Background. The Institut Gustave Roussy is a 420-bed, private, nonprofit, tertiary care referral hospital for patients with cancer. From 1982 to 1985 the incidence of candidemia among patients at the hospital increased from 0.3% to 2.9 %; a total of 30 patients developed candidemia during this interval, and rv50% of these patients died. Risk factors for candidemia were analyzed and control measures were instituted [9]. Despite these efforts, 24 patients developed candidemia in 1987, with a mortality of 53 %.
Received 27 March 1990; revised 29 May 1990. Use of trade names is for purposes of identification only and does not imply endorsement by the U.S. Public Health Service or the U.S. Department of Health and Human Services. * Current address: Laboratoire de Bacteriologic, Hotel Dieu, B.P. 1005, 44035 Nantes Cedex, France. Reprints and correspondence: Dr. William R. Jarvis, Hospital Infections Program, Mail StopClO, Centers for Disease Control, Atlanta, Georgia 30333. Reviews of Infectious Diseases 1991;13:211-5 This article is in the public domain.
Definitions. We defined a case-patient as any patient who had at least one blood culture positive for Candida species between 1 January 1986 and 31 December 1987. Patients were considered to have lung infection if they had clinical signs of lower respiratory infection associated with a bronchial washing culture positive for Candida species and to have central venous catheter-related infection if the semiquantitative culture of the catheter grew >15 colonies of Candida species. Death was considered attributable to candidal infection if the patient died of septic shock associated with candidemia or if autopsy showed histopathologic evidence of invasive candidiasis. Case ascertainment. To determine the background rate of candidemia and to identify case-patients, we reviewed the records of the microbiology laboratory for Candida-positive blood cultures from 1 January 1983 to 31 December 1987. In addition, we reviewed all autopsy and histopathologic reports for invasive candidiasis. Selection ofcontrolpatients. Sixteen controls (two for each case-patient) were randomly selected (by means of a table of random numbers) from patients with ALL who were hospitalized at the Institut Gustave Roussy at least once in 1986. The year was chosen so that the follow-up period would be at least 2 years. Retrospective survey. To compare the evolution of leukemia in case-patients and controls, we conducted a retrospective survey comparing the medical history of these patients 6 months, 1 year, and 2 years after the initial diagnosis. Patients who died during a given period were excluded from subsequent analysis. Case-control study. In this study we analyzed risk factors for candidemia during the hospitalization in which the candidemia occurred. We compared hospitalizations of casepatients with those of control patients whose hospitalizations
Downloaded from http://cid.oxfordjournals.org/ at Cambridge University on September 8, 2015
Between 1983 and 1987 the overall incidence of candidemia at the Institut Gustave Roussy, a tertiary care referral hospital for patients with cancer, increased from 0.1% (7 of 6,801) to 0.32% (24 of 7,515) (P = .009). Because acute lymphocytic leukemia (ALL) was the most common underlying disease in patients with candidemia, risk factors for candidemia were analyzed in this subset of patients. A case-eontrol study comparing the eight ALL patients who had candidemia with 18ALLcontrol patients revealedthat previousbacteremia, prolonged neutropenia, prolonged fever, prolonged administration of antimicrobial agents, treatment with multiple antimicrobial agents, and a relatively high concentration of Candida organisms in stool were significant risk factors for candidemia. In a logistic regression analysis, however, only receipt of vancomycin and/or imipenem was identified as an independent risk factor for candidemia. Further analysis showed that administration of vancomycin promoted proliferation of Candida organisms in the gastrointestinaltract and that this proliferationwasassociatedwith an increasedrisk of candidemia.
Richet et al.
212
Results Descriptive epidemiology. Between 1983 and 1987 the overall incidence of candidemia increased from 0.1% (seven of 6,801) to 0.32 % (24 of 7,515) (P = .009). Of the 24 patients who developed candidemia in 1987, eight (33 %) had
ALL, three (13%) had lymphoma, one (4%) had acute myelogenous leukemia, one (4%) had lymphosarcoma, and 11 (46 %) had a solid tumor. Candida albicans, Candida tropicalis, and Candida pseudotropicalis were responsible for 16 (67%), six (25%), and two (8%) of the candidemia cases, respectively. Because of the broad variety of underlying diseases and cancer treatment modalities (e.g., of patients with solid tumors, six underwent surgery, two had bone marrow transplantations, and three received conventional chemotherapy), we sought a more homogeneous population in which to analyze risk factors for candidemia. We chose case-patients diagnosed as having ALL because this was the largest group and because the treatment of ALL is standardized both nationally and internationally. Description of cases. Case-patients were 3-22 years of age (median, 15 years); six (75 %) were male. Seven patients (88 %) had had at least one relapse of leukemia. The mean interval between the diagnosis of ALL and the occurrence of candidemia was 16 months (range, 0-31 months; table 1). C. albicans was responsible for four cases of candidemia, C. tropicalis for two, and C. pseudotropicalis for two. Candida species were present in the gastrointestinal tract of six casepatients (75%) at the time of candidemia; the infection spread to the lung in three patients (38 %) and to the spleen in one (13 %). For three patients (38 %), death was attributable to candidemia. Retrospective survey. We compared the evolution of leukemia in case-patients and controls during the first 2 years after the diagnosis of ALL. During the first 6 months, case-patients did not relapse significantly more frequently than controls (25 % vs. 13%; P = .6) and mortality among the two groups did not differ (13% vs. 0; P = .3); one case-patient, whose prognosis was poor, developed fatal candidemia after his first induction chemotherapy. During the first year after diagnosis, no significant differences between case-patients and controls were observed in the number of courses of chemotherapy (mean ± SD, 1.6 ±
Table 1. Characteristics of ALL patients with candidemia at the Institut Gustave Roussy in 1987.
Case no. 1 2 3 4 5
6 7 8
Age (y)/sex
17/F 31M 131M 8/F 211M 221M 81M 191M
Interval from ALL Dx to No. of ALL candidemia (mo) relapses 1 1 1 1 I 2 2 0
31 9 18 29 15
8 20 0
Candida species CA CA CP CA CT CT CA CP
No. of Duration of positive candidemia (d) blood cultures 1 8 5 9 3 12
6 4
Primary site
Secondary involvement
Outcome
GI Cath Resp GI GI GI GI GI
Lung None None None Spleen None Lung Lung
Death* Cure Cure Death Cure Death Cure Death
1 3 4 4
6 7 5 3
NOTE. Abbreviations: Dx = diagnosis; CA = C. albicans; CP = C. pseudotropicalis; CT = C. tropicalis: GI respiratory tract. '" Mycelial hyphae resembling Aspergillus species were seen on smears of lung specimens obtained post-mortem.
=
gastrointestinal tract; Cath
=
catheter; and Resp
=
Downloaded from http://cid.oxfordjournals.org/ at Cambridge University on September 8, 2015
lasted at least 20 days (the median duration of case-patient hospitalization before the occurrence of candidemia). The medical records of controls and case-patients were reviewed for the same details. The association among the administration of vancomycin, the colonization of the gastrointestinal tract with Candida species, and the occurrence of candidemia was analyzed. To correlate the number of Candida organisms per gram of feces with the risk of candidemia, we included all patients with ALL who had at least three quantitative stool cultures performed during the first 3 weeks of hospitalization. First, we assessed the risk of candidemia associated with an increase of >4 logs in the number of Candida organisms in stool. Then we looked at the association between the receipt of vancomycin and a subsequent increase in the number of Candida organisms in stool. Microbiologic studies. Stool specimens from leukemic patients were systematically cultured once or twice a week during hospitalization: a 0.2-g volume of feces was diluted at 1:100, a O.1-mL volume of dilution was inoculated onto a Sabouraud agar plate with gentamicin (Bio-Merieux; Charbonnieres les Bains, France), and colonies were enumerated after 48 hours of incubation at 35°C. Statistical methods. We used the Student's t-test and Fisher exact test (two-tailed) to look for significance of association and the X2 test for linear trend to assess secular trends. To identify the independent importance of potential risk factors significantly associated with candidemia in the univariate analysis, we conducted a stepwise logistic regression analysis, using BMDP statistical software (Department of Biomathematics, University of California, Los Angeles).
RID 1991;13 (March-April)
RID 1991;13 (March-April)
Candidemia in Acute Lymphocytic Leukemia
1 vs. 1.4 ± 0.6; P = .7), the frequency of complete remission (57% vs. 69%; P = .7), the frequency of relapse (43% vs. 31 %; P = .7), or mortality (14 % vs. 6 %; P = .5). Only one control patient was lost to follow-up after I year of evolution of his ALL. When case-patients (n = 6) and control patients (n = 16) were compared during the second year of evolution of their leukemia, several differences were observed. Case-patients were less likely to achieve complete remission of leukemia (0 vs. 71%; odds ratio [OR], not calculable [NC]; 95% confidence interval [CI], NC to 0.06; P = .01); they experi-
213
enced relapse of leukemia significantly more often (100% vs. 43%; OR, NC; 95% CI, 1.02 to NC; P = .02) and were significantly more likely to have a bone marrow relapse (100% vs. 21%; OR, NC; 95% CI, 2.4 to NC; P = .002). Overall mortality was higher among case-patients than among controls (100% vs. 31%; P = .002; OR, undefined; 95% CI, 2.4 to undefined). Stratified analysis showed that the difference in mortality between case-patients and controls was related to deaths from candidemia (three of eight vs. 0 of 16; P = .03). Mortality due to causes other than candidemia was not significantly different in case-patients and controls (63 %
Table 2. Comparison of case-patients and controls hospitalized and assessed with regard to risk factors for candidemia: Institut Gustave Roussy, 1987.
Potential risk factor Previous consolidation Relapse Bone marrow relapse Granulocytopenia Outcome of chemotherapy Success Failure Central catheterization Parenteral nutrition Intralipid Fever Antibiotic therapy before candidemia Vancomycin Imipenem Trimethoprim-sulfarnethoxazole Antifungal therapy before candidemia Mycostatin Amphotericin B (oral) Amphotericin B (iv) Other nosocomial infection Bacteremia Urinary tract infection Mucitis Diarrhea Colonization with Candida species Outcome Complete remission Failure of chemotherapy Death Continuous variables No. of relapses Duration of neutropenia (d) < 100 leukocytes 3 logs in Candida organisms in stool, whereas none of six ALL patients who did not receive vancomycin had such an increase (OR, undefined: 95 % CI, 1.37 to undefined; P = .01). None of the eight ALL patients who had an increase of Candida organisms in stool after receiving vancomycin had a stool culture positive for Candida species before receiving vancomycin. Within 10 days after initiation of vancomycin therapy, the number of Candida organisms in stool rose from o to 107/g of feces in two patients, from 0 to lQ6/g in three patients, from 0 to 5 x 104 in one patient, and from 0 to 3 x 1Q4 in one patient. Because the stratified analysis showed that receipt of prophylactic oral amphotericin B protected against candidemia, we assessed the effect of this treatment on the number of Candida organisms in stool. None of the four ALL patients who received vancomycin and oral prophylactic amphotericin B at the same time had an increase of Candida in stool, whereas all of eight ALL patients who received only vancomycin had an increase (OR, undefined; 95 % CI, undefined to 0.34; P = .002).
Discussion Candida species have been shown to be important nosocomial pathogens associated with a significant increase in length of hospital stay and 38% attributable mortality [6]. Multivariate analysis of risk factors for nosocomial candidemia among inpatients at a teaching hospital showed that the number of antibiotics received before infection, the isolation of Candida species from sites other than blood, prior hemodialysis, and prior use of Hickman catheters were the best predictors of candidemia [7]. In a study of hospitalized adult patients without leukemia, Bross et al. showed that the presence of a centralline or bladder catheter, the administration oftwo or more antibiotics, azotemia, transfer from another hospital, diarrhea, and candiduria were significant risk factors for candidemia [8]. Several groups of investigators have tried to analyze risk factors for candidemia in patients with cancer. However, in studies of systemic candidiasis in cancer patients [9], of fungemia at a cancer hospital [10], and-more recently-of can-
Downloaded from http://cid.oxfordjournals.org/ at Cambridge University on September 8, 2015
vs. 31%;P = .2). When times of first relapse incase-patients and controls were compared, numbers of patients in relapse were similar after 6 months (two of eight vs. two of 16; P = .6) and after 1 year (one of eight vs. four of 16; P = .6) but were significantly different in the second year of evolution of ALL (four of eight vs. one of 16; P = .03). Case-control study. This study analyzed risk factors for candidemia during the hospitalization in which it occurred. Although numerous risk factors were associated with candidemia (including previous consolidation chemotherapy, relapse, site of relapse, previous bacteremia, prolonged neutropenia, failure of chemotherapy, prolonged fever, administration of multiple antimicrobial agents before onset of candidemia, prolonged antimicrobial therapy, receipt of vancomycin and/or imipenem and/or trimethoprim-sulfamethoxazole, receipt of intravenous amphotericin B before collection of the first Candida-positive blood sample, and a relatively large number of Candida organisms in stool), among the categorical variables, only the receipt of imipenem, vancomycin, and amphotericin B reached statistical significance (lower 95 % CI, >1; table 2). No association with candidemia was found for the presence of indwelling lines, the administration of parenteral nutrition, or administration of any of the five other antibiotic and the 13 antineoplastic agents analyzed. Stratified analysis showed that receipt of vancomycin was not a significant risk factor for candidemia in ALL patients who received prophylactic oral amphotericin B (four of four vs. five of 10; P = .2) but that it was a significant risk factor for candidemia in ALL patients who did not receive such prophylaxis (four of four vs. one of eight; OR, undefined; 95 % CI, 1.5 to undefined; P = .01). In an assessment of the independent risk factors for candidemia, two logistic regression analyses were performed. In the first analysis, previous consolidation, relapse, site of relapse (bone marrow vs. other sites), duration of neutropenia «100Jeukocytes/mm 3) , duration of fever, and number of Candida organisms in stool were entered in the model; only relapse (OR, 2,170; P = .016) and the duration ofneutropenia (OR, undefined; P = .013) appeared as independent risk factors and best predicted candidemia. In the second model, we added the antimicrobial agents (vancomycin, imipenem, and amphotericin B) significantly associated with candidemia in the univariate analysis; receipt of vancomycin (OR, 275; P = < .001) and receipt of imipenem (OR, 25; P = .046) were the only independent risk factors for infection. Vancomycin as a riskfactor for candidemia. Because experimental studies in animals have shown that administration of vancomycin to mice promotes colonization of the gastrointestinal tract with C. albicans and subsequent dissemination of the fungi to other organs, we analyzed the association among vancomycin administration, gastrointestinal tract colonization with Candida species, and occurrence of candidemia. This study included all patients with ALL who had at least three quantitative stool cultures performed during the first 3 weeks
RID 1991;13 (March-April)
RID 1991;13 (March-April)
Candidemia in Acute Lymphocytic Leukemia
are unavailable; nevertheless, in our study, the number of Candida organisms was significantly higher in patients receiving this antibiotic. If further studies confirm our findings, candidemia in ALL patients may be prevented by limiting the use of vancomycin and imipenem and/or including oral amphotericin B in the regimen when these drugs are used in neutropenic patients.
References 1. Meyers JD. Infection in bone marrow transplant recipients. Am J Med 1986;81(Suppl lA):27-8 2. Myerowitz RL, Pazin GJ, Allen CM. Disseminated candidiasis: changes in incidence, underlying diseases and pathology. Am J Clin Pathol 1977;68:29-38 3. Meunier-Carpentier F, Kiehn TE, Armstrong D. Fungemia in the immunocompromised host. Changing patterns, antigenemia, high mortality. Am J Med 1981;71:363-70 4. Bodey GP. Candidiasis in cancer patients. Am J Med 1984;77:13-9 5. Degregorio MW, Lee WMF, Linker CA, Jacobs RA, Ries CA. Fungal infections in patients with acute leukemia. Am J Med 1982;73:543-8 6. Wey SB, Mori M, Pfaller MA, Woolson RF, Wenzel RP. Hospitalacquired candidemia. The attributable mortality and excess length of stay. Arch Intern Med 1988;148:2642-5 7. Wey SB, Mori M, Pfaller MA, Woolson RF, Wenzel RP. Risk factors for hospital-acquired candidemia. A matched case-control study. Arch Intern Med 1989;149:2349-53 8. Bross J, Talbot GH, Maislin G, Hurwitz S, Strom BL. Risk factors for nosocomial candidemia: a case-control study in adults without leukemia. Am J Med 1989;87:614-20 9. Maksymiuk AW, Thongprasert S, Hopfer R, Luna M, Fainstein V, Bodey GP. Systemic candidiasis in cancer patients. Am J Med 1984;77(Suppl 4D):20-7 10. Horn R, Wong B, Kiehn TE, Armstrong D. Fungemia in a cancer hospital: changing frequency, earlier onset, and results of therapy. Rev Infect Dis 1985;7:646-55 II. Komshian SV, Uwaydah AK, Sobel JD, Crane LR. Fungemia caused by Candida species and Torulopsis glabrata in the hospitalized patient: frequency, characteristics, and evaluation of factors influencing outcome. Rev Infect Dis 1989;1l:379-90 12. Karabinis A, Hill C, Leclercq B, Tancrede C, Baume D, Andremont A. Risk factors for candidemia in cancer patients: a case-control study. J Clin Microbiol 1988;26:429-32 13. Kennedy MJ, Volz PA. Effect of various antibiotics on gastrointestinal colonization and dissemination by Candida albicans. Sabouraudia 1985;23:265-73 14. Kennedy MJ, Volz PA. Ecology of Candida albicans gut colonization: inhibition of candida adhesion, colonization, and dissemination from the gastrointestinal tract by bacterial antagonism. Infect Immun 1985;49:654-63
Downloaded from http://cid.oxfordjournals.org/ at Cambridge University on September 8, 2015
didemia in hospitalized patients [11], no case-control design has been used to compare patients with candidemia or fungemia and those not infected; thus the identification of risk factors was precluded. Risk factors for candidemia in a general population of patients with cancer were analyzed by Karabinis and associates [12], who performed a case-control study at the Institut Gustave Roussy, comparing candidemic patients with matched control patients. Several risk factors for candidemia were identified after multivariate analysis: peripheral blood cultures positive for Candida species, central catheterization, and neutropenia. However, the incidence of candidemia at this institution increased significantly from 1983 to 1987; the relative risk of developing candidemia among patients hospitalized in units treating hematologic malignancies (i.e., the hematology unit and the pediatric bone marrow transplant unit) was 8.25 times higher in 1987 than in 1983. For this reason and also because ALL was the most common underlying disease in cancer patients with candidemia in 1987, we chose to analyze the risk factors for candidemia in patients with ALL. Focusing on candidemia in patients with ALL helped us to find a possible explanation for the increased risk and to identify preventable risk factors, such as relapse, prolonged neutropenia, and antibiotic administration. The more aggressive chemotherapies now used for patients with relapse could induce neutropenia of longer duration and thus fever and antibiotic administration of longer duration. The results of our logistic regression analysis showed that the receipt of vancomycin and/or imipenem was an independent risk factor for candidemia; we also found that the administration of vancomycin promoted the proliferation of Candida organisms in the gastrointestinal tract and that this proliferation was associated with an increased risk of candidemia. All of these findings are consistent with experimental reports. Studies in animals have shown that the anaerobic population in the cecum of mice receiving vancomycin, penicillin, or clindamycin decreases; in contrast, the enteric bacillary population increases. In addition to having an effect on the bacterial flora, these three antibiotics were associated with C. albicans proliferation in the gut and subsequent dissemination of fungi from the gastrointestinal tract to other organs. The antibiotics that reduced anaerobic population levels- but not numbers of enteric bacilli or aerobes - predisposed animals to mucosal colonization with C. albicans [13, 14]. Such data about the use of imipenem
215
Risk Factors for Candidemia in Patients with Acute Lymphocytic Leukemia Herve M. Riehet,* Antoine Andremont, Cyrille Tanerede, Jose Luis Pieo, and William R. Jarvis
From the Hospital Infections Program. Center for Infectious Diseases, Centers for Disease Control, Atlanta. Georgia; and the Service de Microbiologie Medicate and Departement d'Hematologie, Institut Gustave Roussy, Villejuif, France
Candidiasis and aspergillosis are the most common fungal infections in neutropenic patients; in fact Candida and Aspergillus species account for almost 90 % of all fungal infections reported in this population. Previous studies indicate that 10%-40% of patients who die with acute leukemia have evidence of disseminated fungal disease at autopsy; fungi are the cause of death in an estimated 21% of these patients [1-8]. Because acute lymphocytic leukemia (ALL) was the most common underlying disease in patients with candidemia at the Institut Gustave Roussy in 1987, the risk factors for candidemia in these patients were analyzed.
Methods Background. The Institut Gustave Roussy is a 420-bed, private, nonprofit, tertiary care referral hospital for patients with cancer. From 1982 to 1985 the incidence of candidemia among patients at the hospital increased from 0.3% to 2.9 %; a total of 30 patients developed candidemia during this interval, and rv50% of these patients died. Risk factors for candidemia were analyzed and control measures were instituted [9]. Despite these efforts, 24 patients developed candidemia in 1987, with a mortality of 53 %.
Received 27 March 1990; revised 29 May 1990. Use of trade names is for purposes of identification only and does not imply endorsement by the U.S. Public Health Service or the U.S. Department of Health and Human Services. * Current address: Laboratoire de Bacteriologic, Hotel Dieu, B.P. 1005, 44035 Nantes Cedex, France. Reprints and correspondence: Dr. William R. Jarvis, Hospital Infections Program, Mail StopClO, Centers for Disease Control, Atlanta, Georgia 30333. Reviews of Infectious Diseases 1991;13:211-5 This article is in the public domain.
Definitions. We defined a case-patient as any patient who had at least one blood culture positive for Candida species between 1 January 1986 and 31 December 1987. Patients were considered to have lung infection if they had clinical signs of lower respiratory infection associated with a bronchial washing culture positive for Candida species and to have central venous catheter-related infection if the semiquantitative culture of the catheter grew >15 colonies of Candida species. Death was considered attributable to candidal infection if the patient died of septic shock associated with candidemia or if autopsy showed histopathologic evidence of invasive candidiasis. Case ascertainment. To determine the background rate of candidemia and to identify case-patients, we reviewed the records of the microbiology laboratory for Candida-positive blood cultures from 1 January 1983 to 31 December 1987. In addition, we reviewed all autopsy and histopathologic reports for invasive candidiasis. Selection ofcontrolpatients. Sixteen controls (two for each case-patient) were randomly selected (by means of a table of random numbers) from patients with ALL who were hospitalized at the Institut Gustave Roussy at least once in 1986. The year was chosen so that the follow-up period would be at least 2 years. Retrospective survey. To compare the evolution of leukemia in case-patients and controls, we conducted a retrospective survey comparing the medical history of these patients 6 months, 1 year, and 2 years after the initial diagnosis. Patients who died during a given period were excluded from subsequent analysis. Case-control study. In this study we analyzed risk factors for candidemia during the hospitalization in which the candidemia occurred. We compared hospitalizations of casepatients with those of control patients whose hospitalizations
Downloaded from http://cid.oxfordjournals.org/ at Cambridge University on September 8, 2015
Between 1983 and 1987 the overall incidence of candidemia at the Institut Gustave Roussy, a tertiary care referral hospital for patients with cancer, increased from 0.1% (7 of 6,801) to 0.32% (24 of 7,515) (P = .009). Because acute lymphocytic leukemia (ALL) was the most common underlying disease in patients with candidemia, risk factors for candidemia were analyzed in this subset of patients. A case-eontrol study comparing the eight ALL patients who had candidemia with 18ALLcontrol patients revealedthat previousbacteremia, prolonged neutropenia, prolonged fever, prolonged administration of antimicrobial agents, treatment with multiple antimicrobial agents, and a relatively high concentration of Candida organisms in stool were significant risk factors for candidemia. In a logistic regression analysis, however, only receipt of vancomycin and/or imipenem was identified as an independent risk factor for candidemia. Further analysis showed that administration of vancomycin promoted proliferation of Candida organisms in the gastrointestinaltract and that this proliferationwasassociatedwith an increasedrisk of candidemia.
Richet et al.
212
Results Descriptive epidemiology. Between 1983 and 1987 the overall incidence of candidemia increased from 0.1% (seven of 6,801) to 0.32 % (24 of 7,515) (P = .009). Of the 24 patients who developed candidemia in 1987, eight (33 %) had
ALL, three (13%) had lymphoma, one (4%) had acute myelogenous leukemia, one (4%) had lymphosarcoma, and 11 (46 %) had a solid tumor. Candida albicans, Candida tropicalis, and Candida pseudotropicalis were responsible for 16 (67%), six (25%), and two (8%) of the candidemia cases, respectively. Because of the broad variety of underlying diseases and cancer treatment modalities (e.g., of patients with solid tumors, six underwent surgery, two had bone marrow transplantations, and three received conventional chemotherapy), we sought a more homogeneous population in which to analyze risk factors for candidemia. We chose case-patients diagnosed as having ALL because this was the largest group and because the treatment of ALL is standardized both nationally and internationally. Description of cases. Case-patients were 3-22 years of age (median, 15 years); six (75 %) were male. Seven patients (88 %) had had at least one relapse of leukemia. The mean interval between the diagnosis of ALL and the occurrence of candidemia was 16 months (range, 0-31 months; table 1). C. albicans was responsible for four cases of candidemia, C. tropicalis for two, and C. pseudotropicalis for two. Candida species were present in the gastrointestinal tract of six casepatients (75%) at the time of candidemia; the infection spread to the lung in three patients (38 %) and to the spleen in one (13 %). For three patients (38 %), death was attributable to candidemia. Retrospective survey. We compared the evolution of leukemia in case-patients and controls during the first 2 years after the diagnosis of ALL. During the first 6 months, case-patients did not relapse significantly more frequently than controls (25 % vs. 13%; P = .6) and mortality among the two groups did not differ (13% vs. 0; P = .3); one case-patient, whose prognosis was poor, developed fatal candidemia after his first induction chemotherapy. During the first year after diagnosis, no significant differences between case-patients and controls were observed in the number of courses of chemotherapy (mean ± SD, 1.6 ±
Table 1. Characteristics of ALL patients with candidemia at the Institut Gustave Roussy in 1987.
Case no. 1 2 3 4 5
6 7 8
Age (y)/sex
17/F 31M 131M 8/F 211M 221M 81M 191M
Interval from ALL Dx to No. of ALL candidemia (mo) relapses 1 1 1 1 I 2 2 0
31 9 18 29 15
8 20 0
Candida species CA CA CP CA CT CT CA CP
No. of Duration of positive candidemia (d) blood cultures 1 8 5 9 3 12
6 4
Primary site
Secondary involvement
Outcome
GI Cath Resp GI GI GI GI GI
Lung None None None Spleen None Lung Lung
Death* Cure Cure Death Cure Death Cure Death
1 3 4 4
6 7 5 3
NOTE. Abbreviations: Dx = diagnosis; CA = C. albicans; CP = C. pseudotropicalis; CT = C. tropicalis: GI respiratory tract. '" Mycelial hyphae resembling Aspergillus species were seen on smears of lung specimens obtained post-mortem.
=
gastrointestinal tract; Cath
=
catheter; and Resp
=
Downloaded from http://cid.oxfordjournals.org/ at Cambridge University on September 8, 2015
lasted at least 20 days (the median duration of case-patient hospitalization before the occurrence of candidemia). The medical records of controls and case-patients were reviewed for the same details. The association among the administration of vancomycin, the colonization of the gastrointestinal tract with Candida species, and the occurrence of candidemia was analyzed. To correlate the number of Candida organisms per gram of feces with the risk of candidemia, we included all patients with ALL who had at least three quantitative stool cultures performed during the first 3 weeks of hospitalization. First, we assessed the risk of candidemia associated with an increase of >4 logs in the number of Candida organisms in stool. Then we looked at the association between the receipt of vancomycin and a subsequent increase in the number of Candida organisms in stool. Microbiologic studies. Stool specimens from leukemic patients were systematically cultured once or twice a week during hospitalization: a 0.2-g volume of feces was diluted at 1:100, a O.1-mL volume of dilution was inoculated onto a Sabouraud agar plate with gentamicin (Bio-Merieux; Charbonnieres les Bains, France), and colonies were enumerated after 48 hours of incubation at 35°C. Statistical methods. We used the Student's t-test and Fisher exact test (two-tailed) to look for significance of association and the X2 test for linear trend to assess secular trends. To identify the independent importance of potential risk factors significantly associated with candidemia in the univariate analysis, we conducted a stepwise logistic regression analysis, using BMDP statistical software (Department of Biomathematics, University of California, Los Angeles).
RID 1991;13 (March-April)
RID 1991;13 (March-April)
Candidemia in Acute Lymphocytic Leukemia
1 vs. 1.4 ± 0.6; P = .7), the frequency of complete remission (57% vs. 69%; P = .7), the frequency of relapse (43% vs. 31 %; P = .7), or mortality (14 % vs. 6 %; P = .5). Only one control patient was lost to follow-up after I year of evolution of his ALL. When case-patients (n = 6) and control patients (n = 16) were compared during the second year of evolution of their leukemia, several differences were observed. Case-patients were less likely to achieve complete remission of leukemia (0 vs. 71%; odds ratio [OR], not calculable [NC]; 95% confidence interval [CI], NC to 0.06; P = .01); they experi-
213
enced relapse of leukemia significantly more often (100% vs. 43%; OR, NC; 95% CI, 1.02 to NC; P = .02) and were significantly more likely to have a bone marrow relapse (100% vs. 21%; OR, NC; 95% CI, 2.4 to NC; P = .002). Overall mortality was higher among case-patients than among controls (100% vs. 31%; P = .002; OR, undefined; 95% CI, 2.4 to undefined). Stratified analysis showed that the difference in mortality between case-patients and controls was related to deaths from candidemia (three of eight vs. 0 of 16; P = .03). Mortality due to causes other than candidemia was not significantly different in case-patients and controls (63 %
Table 2. Comparison of case-patients and controls hospitalized and assessed with regard to risk factors for candidemia: Institut Gustave Roussy, 1987.
Potential risk factor Previous consolidation Relapse Bone marrow relapse Granulocytopenia Outcome of chemotherapy Success Failure Central catheterization Parenteral nutrition Intralipid Fever Antibiotic therapy before candidemia Vancomycin Imipenem Trimethoprim-sulfarnethoxazole Antifungal therapy before candidemia Mycostatin Amphotericin B (oral) Amphotericin B (iv) Other nosocomial infection Bacteremia Urinary tract infection Mucitis Diarrhea Colonization with Candida species Outcome Complete remission Failure of chemotherapy Death Continuous variables No. of relapses Duration of neutropenia (d) < 100 leukocytes 3 logs in Candida organisms in stool, whereas none of six ALL patients who did not receive vancomycin had such an increase (OR, undefined: 95 % CI, 1.37 to undefined; P = .01). None of the eight ALL patients who had an increase of Candida organisms in stool after receiving vancomycin had a stool culture positive for Candida species before receiving vancomycin. Within 10 days after initiation of vancomycin therapy, the number of Candida organisms in stool rose from o to 107/g of feces in two patients, from 0 to lQ6/g in three patients, from 0 to 5 x 104 in one patient, and from 0 to 3 x 1Q4 in one patient. Because the stratified analysis showed that receipt of prophylactic oral amphotericin B protected against candidemia, we assessed the effect of this treatment on the number of Candida organisms in stool. None of the four ALL patients who received vancomycin and oral prophylactic amphotericin B at the same time had an increase of Candida in stool, whereas all of eight ALL patients who received only vancomycin had an increase (OR, undefined; 95 % CI, undefined to 0.34; P = .002).
Discussion Candida species have been shown to be important nosocomial pathogens associated with a significant increase in length of hospital stay and 38% attributable mortality [6]. Multivariate analysis of risk factors for nosocomial candidemia among inpatients at a teaching hospital showed that the number of antibiotics received before infection, the isolation of Candida species from sites other than blood, prior hemodialysis, and prior use of Hickman catheters were the best predictors of candidemia [7]. In a study of hospitalized adult patients without leukemia, Bross et al. showed that the presence of a centralline or bladder catheter, the administration oftwo or more antibiotics, azotemia, transfer from another hospital, diarrhea, and candiduria were significant risk factors for candidemia [8]. Several groups of investigators have tried to analyze risk factors for candidemia in patients with cancer. However, in studies of systemic candidiasis in cancer patients [9], of fungemia at a cancer hospital [10], and-more recently-of can-
Downloaded from http://cid.oxfordjournals.org/ at Cambridge University on September 8, 2015
vs. 31%;P = .2). When times of first relapse incase-patients and controls were compared, numbers of patients in relapse were similar after 6 months (two of eight vs. two of 16; P = .6) and after 1 year (one of eight vs. four of 16; P = .6) but were significantly different in the second year of evolution of ALL (four of eight vs. one of 16; P = .03). Case-control study. This study analyzed risk factors for candidemia during the hospitalization in which it occurred. Although numerous risk factors were associated with candidemia (including previous consolidation chemotherapy, relapse, site of relapse, previous bacteremia, prolonged neutropenia, failure of chemotherapy, prolonged fever, administration of multiple antimicrobial agents before onset of candidemia, prolonged antimicrobial therapy, receipt of vancomycin and/or imipenem and/or trimethoprim-sulfamethoxazole, receipt of intravenous amphotericin B before collection of the first Candida-positive blood sample, and a relatively large number of Candida organisms in stool), among the categorical variables, only the receipt of imipenem, vancomycin, and amphotericin B reached statistical significance (lower 95 % CI, >1; table 2). No association with candidemia was found for the presence of indwelling lines, the administration of parenteral nutrition, or administration of any of the five other antibiotic and the 13 antineoplastic agents analyzed. Stratified analysis showed that receipt of vancomycin was not a significant risk factor for candidemia in ALL patients who received prophylactic oral amphotericin B (four of four vs. five of 10; P = .2) but that it was a significant risk factor for candidemia in ALL patients who did not receive such prophylaxis (four of four vs. one of eight; OR, undefined; 95 % CI, 1.5 to undefined; P = .01). In an assessment of the independent risk factors for candidemia, two logistic regression analyses were performed. In the first analysis, previous consolidation, relapse, site of relapse (bone marrow vs. other sites), duration of neutropenia «100Jeukocytes/mm 3) , duration of fever, and number of Candida organisms in stool were entered in the model; only relapse (OR, 2,170; P = .016) and the duration ofneutropenia (OR, undefined; P = .013) appeared as independent risk factors and best predicted candidemia. In the second model, we added the antimicrobial agents (vancomycin, imipenem, and amphotericin B) significantly associated with candidemia in the univariate analysis; receipt of vancomycin (OR, 275; P = < .001) and receipt of imipenem (OR, 25; P = .046) were the only independent risk factors for infection. Vancomycin as a riskfactor for candidemia. Because experimental studies in animals have shown that administration of vancomycin to mice promotes colonization of the gastrointestinal tract with C. albicans and subsequent dissemination of the fungi to other organs, we analyzed the association among vancomycin administration, gastrointestinal tract colonization with Candida species, and occurrence of candidemia. This study included all patients with ALL who had at least three quantitative stool cultures performed during the first 3 weeks
RID 1991;13 (March-April)
RID 1991;13 (March-April)
Candidemia in Acute Lymphocytic Leukemia
are unavailable; nevertheless, in our study, the number of Candida organisms was significantly higher in patients receiving this antibiotic. If further studies confirm our findings, candidemia in ALL patients may be prevented by limiting the use of vancomycin and imipenem and/or including oral amphotericin B in the regimen when these drugs are used in neutropenic patients.
References 1. Meyers JD. Infection in bone marrow transplant recipients. Am J Med 1986;81(Suppl lA):27-8 2. Myerowitz RL, Pazin GJ, Allen CM. Disseminated candidiasis: changes in incidence, underlying diseases and pathology. Am J Clin Pathol 1977;68:29-38 3. Meunier-Carpentier F, Kiehn TE, Armstrong D. Fungemia in the immunocompromised host. Changing patterns, antigenemia, high mortality. Am J Med 1981;71:363-70 4. Bodey GP. Candidiasis in cancer patients. Am J Med 1984;77:13-9 5. Degregorio MW, Lee WMF, Linker CA, Jacobs RA, Ries CA. Fungal infections in patients with acute leukemia. Am J Med 1982;73:543-8 6. Wey SB, Mori M, Pfaller MA, Woolson RF, Wenzel RP. Hospitalacquired candidemia. The attributable mortality and excess length of stay. Arch Intern Med 1988;148:2642-5 7. Wey SB, Mori M, Pfaller MA, Woolson RF, Wenzel RP. Risk factors for hospital-acquired candidemia. A matched case-control study. Arch Intern Med 1989;149:2349-53 8. Bross J, Talbot GH, Maislin G, Hurwitz S, Strom BL. Risk factors for nosocomial candidemia: a case-control study in adults without leukemia. Am J Med 1989;87:614-20 9. Maksymiuk AW, Thongprasert S, Hopfer R, Luna M, Fainstein V, Bodey GP. Systemic candidiasis in cancer patients. Am J Med 1984;77(Suppl 4D):20-7 10. Horn R, Wong B, Kiehn TE, Armstrong D. Fungemia in a cancer hospital: changing frequency, earlier onset, and results of therapy. Rev Infect Dis 1985;7:646-55 II. Komshian SV, Uwaydah AK, Sobel JD, Crane LR. Fungemia caused by Candida species and Torulopsis glabrata in the hospitalized patient: frequency, characteristics, and evaluation of factors influencing outcome. Rev Infect Dis 1989;1l:379-90 12. Karabinis A, Hill C, Leclercq B, Tancrede C, Baume D, Andremont A. Risk factors for candidemia in cancer patients: a case-control study. J Clin Microbiol 1988;26:429-32 13. Kennedy MJ, Volz PA. Effect of various antibiotics on gastrointestinal colonization and dissemination by Candida albicans. Sabouraudia 1985;23:265-73 14. Kennedy MJ, Volz PA. Ecology of Candida albicans gut colonization: inhibition of candida adhesion, colonization, and dissemination from the gastrointestinal tract by bacterial antagonism. Infect Immun 1985;49:654-63
Downloaded from http://cid.oxfordjournals.org/ at Cambridge University on September 8, 2015
didemia in hospitalized patients [11], no case-control design has been used to compare patients with candidemia or fungemia and those not infected; thus the identification of risk factors was precluded. Risk factors for candidemia in a general population of patients with cancer were analyzed by Karabinis and associates [12], who performed a case-control study at the Institut Gustave Roussy, comparing candidemic patients with matched control patients. Several risk factors for candidemia were identified after multivariate analysis: peripheral blood cultures positive for Candida species, central catheterization, and neutropenia. However, the incidence of candidemia at this institution increased significantly from 1983 to 1987; the relative risk of developing candidemia among patients hospitalized in units treating hematologic malignancies (i.e., the hematology unit and the pediatric bone marrow transplant unit) was 8.25 times higher in 1987 than in 1983. For this reason and also because ALL was the most common underlying disease in cancer patients with candidemia in 1987, we chose to analyze the risk factors for candidemia in patients with ALL. Focusing on candidemia in patients with ALL helped us to find a possible explanation for the increased risk and to identify preventable risk factors, such as relapse, prolonged neutropenia, and antibiotic administration. The more aggressive chemotherapies now used for patients with relapse could induce neutropenia of longer duration and thus fever and antibiotic administration of longer duration. The results of our logistic regression analysis showed that the receipt of vancomycin and/or imipenem was an independent risk factor for candidemia; we also found that the administration of vancomycin promoted the proliferation of Candida organisms in the gastrointestinal tract and that this proliferation was associated with an increased risk of candidemia. All of these findings are consistent with experimental reports. Studies in animals have shown that the anaerobic population in the cecum of mice receiving vancomycin, penicillin, or clindamycin decreases; in contrast, the enteric bacillary population increases. In addition to having an effect on the bacterial flora, these three antibiotics were associated with C. albicans proliferation in the gut and subsequent dissemination of fungi from the gastrointestinal tract to other organs. The antibiotics that reduced anaerobic population levels- but not numbers of enteric bacilli or aerobes - predisposed animals to mucosal colonization with C. albicans [13, 14]. Such data about the use of imipenem
215
