Received 18 May 2006/Returned for modification 16 August 2006/Accepted 23 August 2006

Transcription

1 JOURNAL OF CLINICAL MICROBIOLOGY, Nov. 2006, p Vol. 44, No /06/$ doi: /jcm Copyright 2006, American Society for Microbiology. All Rights Reserved. Measurement of Serum D-Arabinitol/Creatinine Ratios for Initial Diagnosis and for Predicting Outcome in an Unselected, Population-Based Sample of Patients with Candida Fungemia Siew Fah Yeo, 1,3 Sharon Huie, 3 Andre N. Sofair, 3 Sheldon Campbell, 2,4 Amanda Durante, 3 and Brian Wong 1,3 * Infectious Diseases Section 1 and the Laboratory Service, 2 VA Connecticut Healthcare System, West Haven, Connecticut, and the Departments of Internal Medicine 3 and Laboratory Medicine, 4 Yale University School of Medicine, New Haven, Connecticut Received 18 May 2006/Returned for modification 16 August 2006/Accepted 23 August 2006 D-Arabinitol (DA) is a useful diagnostic marker for candidiasis in patients with neutropenia and other high-risk groups, but its use in unselected patients with a broad range of underlying diseases and conditions has not been studied. We used an automated enzymatic fluorometric assay to measure serum DA/creatinine ratios (DA/cr s) in 30 healthy adults, 100 hospitalized controls without Candida fungemia, and 83 patients from a study of all Candida fungemias in Connecticut between October 1998 and September Sixty-three of 83 (76%) fungemic patients and 11 of 100 (11%) nonfungemic controls had serum DA/cr s >3.9 M/mg/dl (mean 3 standard deviations for 30 healthy adults). High serum DA/cr s were less frequent in patients with cancer or fungemia caused by the DA nonproducer Candida glabrata than in patients with cancer or fungemia caused by a DA producer, C. albicans, C. tropicalis, orc. parapsilosis. The serum DA/cr was first >3.9 M/mg/dl before, on the same day as, or after the first positive blood culture was drawn for 30 (36%), 22 (27%), and 11 (13%) fungemia patients, respectively. Mortality did not differ significantly among the patients with high or normal initial or peak serum DA/cr s, but mortality was higher if any serum DA/cr value was >3.9 M/mg/dl 3 or more days after the onset of fungemia (18/27 versus 4/24 patients, respectively; P < 0.001). We conclude that serum DA/cr s are useful both for the initial diagnosis of Candida fungemia and for prognostic purposes for unselected patients with a broad range of underlying diseases and conditions. The incidence of invasive candidiasis has increased dramatically in recent years, but the timely and accurate detection of this infection is difficult. Conventional culture-based clinical methods may take several days to become positive and are not very sensitive for the detection of invasive disease (1, 13, 14). Alternative approaches, such as immunologic detection of macromolecular fungal antigens or PCR assays for fungal nucleic acid sequences, have been described; but these methods are not yet sufficiently sensitive and specific to be widely adopted into clinical practice (14, 20). It has been known for many years that several medically important Candida species produce large amounts of the five-carbon polyol D-arabinitol (DA) in culture (2, 8), and several groups have shown that serum DA concentrations and serum D-arabinitol/creatinine ratios (DA/cr s) are higher in animals and humans with invasive Candida infections than in uninfected or colonized controls (4, 6, 8, 11, 12, 17 19). However, DA measurements are not widely used for the diagnosis of invasive candidiasis. One reason for this is that the methods used to measure DA levels in serum in early studies were cumbersome and required the use of instruments and reagents that are seldom available in clinical laboratories. * Corresponding author. Present address: Division of Infectious Diseases, Oregon Health and Science University, 3181 SW Sam Jackson Park Road (NRC-3), Portland, OR Phone: (503) Fax: (503) wongbri@ohsu.edu. Published ahead of print on 6 September Switchenko et al. (12) developed an automated enzymatic method for quantifying DA in human serum, and Walsh et al. (16) used this method in a large prospective study of high-risk, neutropenic cancer patients and bone marrow transplant recipients. That study established that most patients with serious Candida infections had high serum DA/cr s and that changes in serum DA/cr s over time correlated with the responses to antifungal therapy. One problem with the automated enzymatic DA assay is that it requires that an enzyme be purified from the pathogenic fungus Candida tropicalis. Also, most past studies have examined the diagnostic value of DA measurements primarily with high-risk patients who were treated in one or a few specialized institutions (5, 6, 8 12, 16 18), so it is not yet known if serum DA measurements are also useful for patients with a broad range of underlying conditions. In an earlier study, we modified the automated enzymatic DA assay by (i) substituting recombinant C. albicans DA dehydrogenase, which can be produced in large amounts in Escherichia coli, for a DA dehydrogenase that was purified from C. tropicalis and (ii) quantifying NADH production fluorimetrically, which is more sensitive than the original colorimetric method (21). In the present study, we used this modified method to assess the value of serum DA measurements as an initial diagnostic method and also to assess the prognoses for 83 patients who had a wide range of underlying conditions and in whom Candida fungemia was diagnosed and treated in teaching hospitals and community hospitals throughout the state of Connecticut. The specific objectives of this study were 3894

2 VOL. 44, 2006 D-ARABINITOL LEVELS IN CANDIDA FUNGEMIA 3895 to determine (i) the sensitivity and the specificity of elevated serum DA levels in all patients with Candida fungemia and in relevant subgroups, (ii) when during the course of Candida fungemia the serum DA levels first became elevated, and (iii) if serial serum DA levels correlated with mortality within 30 days of the onset of fungemia. MATERIALS AND METHODS Study design. The present study was conducted in conjunction with a population-based study of the epidemiology of Candida fungemia in Connecticut that was sponsored by the Centers for Disease Control and Prevention (CDC) (7). From October 1998 through September 2000, all 33 acute-care hospitals in Connecticut were asked to report all cases of Candida fungemia to a central data center and to save all Candida strains isolated from blood cultures. From October 1998 through September 1999, we also asked each participating hospital s clinical laboratories to retrieve and save unused portions of each fungemic patient s serum that had been submitted for routine testing both before Candida fungemia was documented and thereafter. Thirty of 33 hospitals agreed to collect and store serum specimens for this study. To be included in this study, a patient was required to have (i) at least one blood culture positive for Candida; (ii) serum specimens available from at least two different days, the first of which was from the same day that the first blood sample positive by culture was drawn, 1 day earlier, or 2 days earlier; and (iii) medical records that were available for review. Single serum specimens were collected from 100 hospitalized control patients from whom blood samples for cultures were obtained for suspected infection but whose blood cultures did not yield Candida. We reviewed each fungemic patient s medical records for information on age, gender, underlying conditions, treatment with steroids or other immunosuppressive drugs, use of antifungal drugs either for prophylaxis or for therapy, and survival for 30 days after the onset of fungemia. A Charlson comorbidity score was calculated for each patient (3). The nonfungemic control patients medical records were reviewed for information on age, gender, underlying diseases, the results of laboratory tests other than blood cultures, and the final outcome of hospitalization. Single serum specimens were also collected from 30 healthy hospital employees. D-Arabinitol measurements. Serum specimens were stored at 4 C at the participating hospitals until they were retrieved by study personnel. They were then transported to the study laboratory at the VA Connecticut Healthcare System, where they were stored at 80 C. The specimens were thawed, diluted 1:1 (vol/vol) with 10 mm sodium citrate (ph 4.0), boiled for 10 min, and cooled on ice; and the supernatants obtained by centrifugation at 10,000 g for 10 min were analyzed for D-arabinitol by a rapid, automated enzymatic fluorometric assay. The key reagent in this assay is a recombinant C. albicans DA dehydrogenase that is produced in E. coli and purified by dye ligand affinity chromatography. Formal validation studies have established that this assay can quantify DA in human serum with high degrees of accuracy and precision (21). To correct for the effects of renal function on serum DA levels, creatinine was measured by the modified Jaffee alkaline picrate reaction with an Hitachi 717 chemistry instrument, and the DA/cr s were calculated and expressed as the DA concentration ( M)/creatinine concentration (mg per dl) (6, 16, 17, 21). Microbiology methods. All Candida isolates were transferred to Sabouraud s dextrose agar and stored. Subcultures were sent to the CDC laboratories, where the species were determined and the organisms were tested for their susceptibilities to several antifungal drugs (7). Statistical methods. The Mann-Whitney test was used for continuous variables, and the Fisher exact test was used for discontinuous variables. All P values were two tailed, and a P value 0.05 was considered significant. Human investigation committee approvals. This study was approved by the Human Investigations Committees of the VA Connecticut Healthcare System, Yale University School of Medicine, the Connecticut Hospital Association, and all participating hospitals. RESULTS Characteristics of study subjects. The mean standard deviation (SD) age of the 30 healthy hospital employees was years (range, 21 to 63 years). The mean SD age of the 100 hospitalized, nonfungemic controls was years (range, 8 months to 87 years). These controls included 44 patients with gram-positive bacteremia, 16 with gram-negative bacteremia, 2 with fungal infections other than candidiasis, 19 with nonbacteremic bacterial infections (8 patients with urinary tract infections; 4 patients with pneumonias; 2 patients with intravascular catheter infections; and 1 patient each with peritonitis, cellulitis, groin infection, wound infection, and cholangitis), and 23 with no known infection. The total exceeds 100 because some patients had more than one infection. Eighty-three patients with Candida fungemia qualified for this study, and 345 serum specimens were available from these patients (mean, 4.16 serum specimens per patient; median, 4 serum specimens per patient; range, 2 to 8 specimens per patient). The mean SD age of the fungemic patients was years (range, 1 to 87 years). The fungemic patients included 39 (47%) males, 3 (3.6%) with human immunodeficiency virus infection or AIDS, 15 (18%) with neutropenia, 28 (34%) with cancer, and 25 (30%) who had undergone abdominal surgery during the current hospitalization. Prior to the onset of fungemia, 80 patients (96%) had received antibacterial drugs, 34 (41%) had received corticosteroids or other immunosuppressive drugs, 22 (26%) had received antifungal prophylaxis, 70 (84%) had at least one indwelling intravenous catheter, and 60 (72%) had at least one central venous catheter. Forty-eight fungemic patients were infected with C. albicans, 10 with Candida tropicalis, 10 with Candida glabrata, 7 with Candida parapsilosis, 3 with Candida lusitaniae, 3 with Candida krusei, and 2 with undetermined Candida species. Seventy patients (84%) received systemic antifungal therapy after fungemia was recognized. Forty fungemic patients (48%) died within 30 days of the onset of fungemia, including 33 who received systemic antifungal therapy and 7 who did not. Serum DA/cr s for initial diagnosis of Candida fungemia. The mean SD DA/cr s for 30 healthy hospital employees was M/mg/dl, and we defined the upper limit of normal as 3 SDs above the mean for this group (3.9 M/mg/dl). The serum DA/cr s for the nonfungemic controls with gram-negative bacteremia, gram-positive bacteremia, other infections, or no known infection were not significantly different; so the serum DA/cr s for these groups were combined. The mean serum DA/cr and the proportion of patients with serum DA/ cr s 3.9 M/mg/dl were higher for the hospitalized nonfungemic controls than for the healthy adults, but these differences were not quite significant (Table 1). The 11 nonfungemic control patients with serum DA/cr s 3.9 M/mg/dl included 6 patients with gram-positive or gram-negative bacteremia, 2 with Candida species isolated in culture from the sputum and stool, 1 with a wound infection, and 2 with no known infection. Four of the 11 nonfungemic controls with high serum DA/cr s had multiorgan failure and rapidly deteriorating renal function. The 83 patients with Candida fungemia had higher peak serum DA/cr s and were more likely than the nonfungemic controls to have peak serum DA/cr s 3.9 M/mg/dl (Table 1). Because the analysis of multiple specimens per fungemic patient and only one specimen per control was a potential source of bias, we repeated the analysis using only the initial DA/cr for each fungemic patient (defined as the DA/cr on the first day of fungemia or the day closest to that day, depending on availability of serum specimens). The initial DA/cr and the propor-

3 3896 YEO ET AL. J. CLIN. MICROBIOL. Group No. of patients TABLE 1. Serum D-arabinitol/creatinine ratios Mean SD ( M/mg/dl) Serum DA/cr P b No. of patients/ total no. (%) DA/cr 3.9 M/mg/dl a Healthy adults /30 (0) Nonfungemic controls /100 (11) All fungemic patients 83 Highest /83 (76) Initial /83 (48) C. albicans 48 Highest /48 (83) Initial /48 (46) C. tropicalis 10 Highest /10 (80) Initial /10 (70) C. glabrata 10 Highest /10 (40) Initial /10 (30) 0.12 C. parapsilosis 7 Highest /7 (71) Initial /7 (57) Other Candida species d 8 Highest /8 (75) Initial /8 (50) a Mean 3 SDs for healthy subjects. b P versus nonfungemic controls by the Mann-Whitney test. c P versus nonfungemic controls by the Fisher exact test. d Includes C. krusei (n 3), C. lusitaniae (n 3), and undetermined Candida species (n 2). P c tion of patients with high initial DA/cr s were also higher for the fungemic patients than for the nonfungemic controls (Table 1). C. albicans, C. tropicalis, and C. parapsilosis produce DA in culture; but C. glabrata does not (2). When the fungemic patients were analyzed according to Candida species, the groups infected with C. albicans, C. tropicalis, and C. parapsilosis had higher initial and peak DA/cr s than the nonfungemic controls; and most of the patients in all of these groups had serum DA/cr s 3.9 M/mg/dl. In contrast, the 10 patients with C. glabrata fungemia had initial and peak serum DA/cr s lower than those observed for patients infected with any other individual species, and most of these patients had serum DA/cr s 3.9 M/mg/dl (Table 1). Compared to the 65 patients whose fungemias were caused by a DA-producing species (C. albicans, C. tropicalis, or C. parapsilosis), the 10 C. glabrata-infected patients had lower peak serum DA/cr s ( versus M/mg/dl [means SD]; P 0.003), and they were also less likely to have a peak serum DA/cr 3.9 M/mg/dl (4/10 [40%] versus 53/65 [82%]; P 0.009). Lastly, when the values for the 10 patients with C. glabrata fungemia were excluded, the peak and initial DA/cr values among the 73 remaining patients were 3.9 M/mg/dl for 59 (81%) and 37 (51%) patients, respectively. Since the fungemic patients in this study had a broad range of underlying diseases and conditions, we correlated the presence or absence of high serum DA/cr s with a series of host and microbial characteristics. Patients with cancer were less likely than patients without cancer to have high serum DA/cr s; but there were no significant correlations between a high serum DA/cr and persistent fungemia, a Charlson comorbidity score 3, prior abdominal surgery, administration of antifungal prophylaxis, infection with a Candida strain that was resistant to fluconazole, treatment with corticosteroids or other immunosuppressive drugs, neutropenia, the presence of an indwelling intravenous catheter, or treatment in community rather than teaching hospitals (Table 2). Timing of elevated DA/cr s. To assess the timeliness of high serum DA/cr s, we compared the dates on which each fungemic patient s first positive blood culture was drawn and on which the serum DA/cr was first 3.9 M/mg/dl. The serum DA/cr was 3.9 M/mg/dl 1 or more days before the first positive blood culture was drawn in 30 patients (36%), on the same day that the first positive blood culture was drawn in 22 patients (27%), and 1 or more days after the first positive blood culture was drawn in 11 patients (13%) (Fig. 1). Twenty patients (24%) never had high serum DA/cr s. DA/cr and outcome. Because serum DA/cr s are known to rise in proportion to the fungal load in animals with experimental C. albicans infections (15, 17), we also correlated the serum DA/cr with outcomes, both at the onset of Candida fungemia and thereafter. Between 3 days before and 2 days after the onset of fungemia, the serum DA/cr s for patients who survived for 30 days after the onset of fungemia did not

4 VOL. 44, 2006 D-ARABINITOL LEVELS IN CANDIDA FUNGEMIA 3897 TABLE 2. Factors or conditions associated with elevated serum DA/cr levels Factor or condition No. of patients with DA/ cr s 3.9 M/mg/dl a / total no. (%) P b Present Absent Persistent fungemia 25/30 (83) 38/53 (72) 0.29 Cancer 17/29 (58) 46/55 (84) 0.03 Charlson comorbidity score 3 34/42 (81) 29/41 (71) 0.31 Abdominal surgery 21/25 (84) 42/58 (72) 0.40 Antifungal prophylaxis 19/22 (83) 44/60 (73) 0.25 Fluconazole MIC 32 g/ml 16/17 (94) 52/74 (70) Steroids or immunosuppressive 27/34 (79) 36/49 (73) 0.61 drugs Neutropenia 11/15 (73) 47/62 (76) 0.99 Indwelling intravenous catheter 48/61 (79) 15/22 (68) 0.38 Treated in community hospital 49/65 (75) 14/18 (78) 0.99 a Mean 3 SDs for healthy adults. b P calculated by the Fisher exact test. differ significantly from the serum DA/cr s for patients who did not survive. However, the patients who did not survive had significantly higher serum DA/cr s by 3 days after the onset of fungemia than the patients who survived (Fig. 2). Since these results suggested that the serum DA/cr may be useful for prognoses, we also compared the 30-day mortality rates for the patients who had normal or high serum DA/cr s at the onset of fungemia and at several intervals thereafter. Mortality was not significantly different for the fungemic patients who had high or normal initial or peak serum DA/cr s. In contrast, among the patients from whom at least one serum specimen was available 3 or more days after the onset of fungemia, the 30-day mortality was much higher if the DA/cr was 3.9 M/mg/dl in any of these late specimens than if all of these late serum DA/cr s were normal (Table 3). DISCUSSION FIG. 2. Serum DA/cr and outcome. The median DA/cr declined into the normal range in the 43 patients who survived for 30 days after the onset of fungemia but not in the 40 patients who died. Day 0 was the day the first positive blood culture was drawn. The table shows how many serum samples were analyzed per group on each day and the P values for the difference between the two groups, as determined by the Mann-Whitney test. The overall goal of this project was to assess the usefulness of serum DA/cr measurements for the initial diagnosis of Candida fungemia and for predicting the outcomes for patients with a broad range of underlying diseases or conditions. Several groups have assessed DA as a diagnostic marker for patients with serious Candida infections, but most of these prior studies were conducted with special high-risk populations and/or at one or a few specialized referral centers (5, 6, 8, 10, 11, 16). Moreover, most previous studies included too few patients whose serum DA levels were measured more than once to correlate changes in serum DA levels with outcomes. The present study differed from most earlier studies in two ways. First, our patients were drawn from a population-based survey of patients in whom Candida fungemia was diagnosed in more than 30 hospitals throughout the state of Connecticut, so they better represented the broad range of patients who develop Candida fungemia. Second, in order to correlate initial and subsequent serum DA/cr s with outcomes, we required that serum specimens be available from at least two different days, the first of which was between 0 and 2 days before the first positive blood culture was drawn. Our analysis of the sensitivity of serum DA/cr analysis as an initial diagnostic method showed that 63 of 83 (76%) patients with Candida fungemia and 25 of 30 (83%) patients with persistent fungemia had at least one serum DA/cr 3.9 M/mg/dl (mean 3 SDs of the value for healthy hospital workers). These results were very similar to those reported by Walsh et al. (16), who found that 31 of 42 (74%) neutropenic cancer patients with Candida fungemia and 25 of 30 (83%) patients with persistent Candida fungemia had one or more serum DA/cr s 4.0 M/mg/dl (mean 3 SDs of the value for TABLE 3. Serum DA/cr values and mortality Group No. of patients who died within 30 days/ total no. (%) P a FIG. 1. Cumulative proportion of fungemic patients with a high serum DA/cr. The serum DA/cr was first 3.9 M/mg/dl before or on the same day the first positive blood culture was drawn (day 0) in most of the patients with Candida fungemia. The serum DA/cr was always 3.9 M/mg/dl in 20 (26%) fungemic patients. Initial DA/cr 3.9 M/mg/dl b 21/40 (53) 19/43 (44) 0.59 Any DA/cr 3.9 M/mg/dl 33/63 (52) 7/20 (35) 0.21 DA/cr 3.9 M/mg/dl 3 or more days after onset of fungemia 18/27 (67) 4/24 (17) a P calculated by the Fisher exact test. b Mean 3 SDs for healthy adults. Yes No

5 3898 YEO ET AL. J. CLIN. MICROBIOL. healthy blood donors). When we analyzed the serum DA/cr s for a series of subgroups, the only significant correlations we found were that high serum DA/cr s were more frequent (i) in patients infected with the DA-producing species C. albicans, C. tropicalis, and C. parapsilosis than in those infected with the DA-nonproducing species C. glabrata (Table 1) and (ii) in patients who did not have cancer than in those with cancer (Table 2). The overall similarity of our results to those observed previously for neutropenic cancer patients and other high-risk groups (6, 8, 16) and the fact that none of the subgroups that we examined was less likely than cancer patients to have high serum DA/cr s imply that serum DA/cr measurements are as sensitive for the initial diagnosis of Candida fungemia in unselected patients with a broad range of underlying diseases and conditions as they are for the initial diagnosis of Candida fungemia in the special populations in which they have previously been studied. Since the usefulness of any diagnostic test depends on its specificity as well as its sensitivity, we also measured the serum DA/cr s for 100 hospitalized control patients in whom blood cultures that were performed for suspected infection did not yield Candida. The nonfungemic controls had higher mean serum DA/cr s and were more likely than the 30 healthy adults to have a DA/cr 3.9 M/mg/dl, but these differences were not quite significant (Table 1). Previous investigators have also reported that hospitalized controls with proven or suspected infections had higher serum DA concentrations or DA/cr s than healthy subjects (6, 8, 16). For example, 28 of the 206 (14%) nonfungemic controls studied by Walsh et al. (16) had serum DA/cr s 4.0 M/mg/dl. It is not known why hospitalized patients without evidence of serious Candida infections have higher serum DA/cr s than healthy subjects. Another group has reported that the high serum DA/cr results among nonfungemic controls patients were associated with rapidly deteriorating renal function, particularly when it was associated with preterminal, multiorgan failure (16); and 4 of the 11 nonfungemic controls in the present study who had high serum DA/cr s had multiorgan failure with deteriorating renal functions. Another possibility is that the antibiotics given to many hospitalized patients may deplete the gastrointestinal tract of the bacteria that ordinarily catabolize the DA derived from the diet or from colonizing Candida organisms. This is supported by our earlier observations that (i) rats receiving dietary DA had higher serum and urinary DA levels than rats given no dietary DA and (ii) antibiotics increased the serum and urinary DA levels in rats maintained on diets that contained DA, but not in controls receiving no dietary DA (19). Lastly, it is known that approximately one-half of patients with serious invasive Candida infections have repeatedly negative blood cultures (1); thus, it is possible that some of the nonfungemic controls in this and also earlier studies had high serum DA/cr s because they had significant Candida infections that were not detected by standard diagnostic methods. When we compared the dates on which our patients serum DA/cr s were first 3.9 M/mg/dl with the dates on which the first positive blood culture were drawn, we found that the serum DA/cr was first 3.9 M/mg/dl 1 or more days before the first positive blood culture was drawn in 36% of the patients, on the same day the first positive blood sample positive by culture was drawn in an additional 27% of the patients, and 1 or more days after the first positive blood culture was drawn in an additional 12% of patients (Fig. 1). Since only limited numbers of samples were available from the patients in this study, it is likely that even better results could be obtained with more frequent sampling. Nevertheless, our results were quite similar to those of Walsh et al. (16), who found that the serum DA/cr was first 4.0 M/mg/dl before, on the same day, and after the first positive blood culture was drawn in 35%, 15%, and 50% of neutropenic cancer patients with Candida fungemia, respectively. The close similarity between our overall results and those reported by others indicates that serum DA/cr measurements are at least as sensitive, specific, and timely for the initial diagnosis of Candida fungemia in unselected patients with a broad range of underlying diseases and conditions as they are in high-risk patients, such as those with neutropenia and cancer. Another goal of this study was to determine if the serum DA/cr could be used for prognostic purposes. Since the clinical data available for this study were obtained in retrospect by reviewing the records from many different community and teaching hospitals, the only clinical outcome variable that was available in all cases was whether or not the patient survived. Mortality within 30 days was slightly higher in patients who had high initial or peak serum DA/cr s than in patients who had normal initial or peak serum DA/cr s, but these differences were not significant. In contrast, the 30-day mortality was much higher in patients who had high serum DA/cr s 3 or more days after the onset of fungemia than in patients whose serum DA/cr s were all normal 3 or more days after the onset of fungemia (67% versus 17%; P 0.001). An earlier study showed that the serum DA/cr correlated with the therapeutic responses to antifungal treatment in neutropenic cancer patients and that trends in the serum DA/cr over time correlated with mortality (16). However, that earlier study did not show that the serum DA/cr on any day had clear prognostic implications. Our results are also consistent with the strong correlations previously observed between the serum DA/cr and the fungal load in untreated and treated animals with experimental candidiasis (15, 17). Because we had only retrospective clinical data, it was not possible to determine if patients with better or poorer prognoses could have been differentiated as well as or better by the use of alternative clinical or other laboratory criteria as they were from their serum DA/cr s. Nevertheless, our findings suggest that serial DA/cr measurements may be useful for identifying patients with a poor prognosis who may benefit from alternative antifungal drugs, higher drug doses, and/or combination therapy. Serial DA/cr measurements may also be useful as a surrogate marker for therapeutic responses in comparative treatment trials. In summary, this population-based study of unselected patients with Candida fungemia showed that serum DA/cr measurements are as sensitive, specific, and timely for the initial detection of Candida fungemia in patients with a broad range of underlying diseases and conditions as has previously been reported in cancer patients and other special populations. In addition, the serum DA/cr s determined 3 or more days after the onset of fungemia correlated well with the 30-day mortality. We conclude from these results that serum DA measurements are more broadly applicable than has previously been

6 VOL. 44, 2006 D-ARABINITOL LEVELS IN CANDIDA FUNGEMIA 3899 appreciated, both for the initial diagnosis of serious Candida infections and for prognostic purposes. ACKNOWLEDGMENTS This work was supported by grants from the U.S. Department of Veterans Affairs (to B.W.), the National Institutes of Health (grant R01 AI to B.W.), the Centers for Disease Control and Prevention (to A.S.), and Pfizer Pharmaceuticals (to B.W.). We thank the laboratory personnel in all of the participating Connecticut hospitals for their assistance with collecting, saving, and storing the Candida isolates and serum specimens. R. Hajjeh and colleagues at the Centers for Disease Control and Prevention assisted with the handling and analysis of the extensive clinical database that was assembled for this study, and they also determined the species of all of the Candida isolates and tested these isolates for antifungal drug susceptibility. Vivian Wu provided valuable technical assistance. REFERENCES 1. Berenguer, J., M. Buck, F. Witebsky, F. Stock, P. A. Pizzo, and T. J. Walsh Lysis-centrifugation blood cultures in the detection of tissue-proven invasive candidiasis: disseminated versus single organ infection. Diagn. Microbiol. Infect. Dis. 17: Bernard, E. M., L. J. Christianssen, S. F. Tsang, T. E. Kiehn, and D. Armstrong Rate of arabinitol production by pathogenic yeast species. J. Clin. Microbiol. 14: Charlson, M., T. P. Szatrowski, J. Peterson, and J. Gold Validation of a combined comorbidity index. J. Clin. Epidemiol. 47: Christensson, B., G. Sigmundsdottir, and L. Larsson D-Arabinitol a marker for invasive candidiasis. Med. Mycol. 37: Christensson, B., T. Wiebe, C. Pehrson, and L. Larsson Diagnosis of invasive candidiasis in neutropenic children with cancer by determination of D-arabinitol/L-arabinitol in urine. J. Clin. Microbiol. 35: Gold, J. W., B. Wong, E. M. Bernard, T. E. Kiehn, and D. Armstrong Serum arabinitol concentrations and arabinitol/creatinine ratios in invasive candidiasis. J. Infect. Dis. 147: Hajjeh, R. A., A. N. Sofair, L. N. Harrison, M. A. Lyon, B. A. Arthington- Skaggs, S. Mirza, et al Incidence of bloodstream infections due to Candida species and in vitro susceptibilities of isolates collected from 1998 to 2000 in a population-based active surveillance program. J. Clin. Microbiol. 42: Kiehn, T. E., E. M. Bernard, J. W. M. Gold, and D. Armstrong Candidiasis: detection by gas-liquid chromatography of D-arabinitol, a fungal metabolite, in human serum. Science 206: Larsson, L., C. Pehrson, T. Wiebe, and B. Christensson Gas chromatographic determination of D-arabinitol/L-arabinitol ratios in urine: a potential method for diagnosis of disseminated candidiasis. J. Clin. Microbiol. 32: Lehtonen, L., V. J. Anttila, T. Ruutu, J. Salonen, J. Nikoskelainen, E. Eerola, and P. Ruutu Diagnosis of disseminated candidiasis by measurement of urine D-arabinitol/L-arabinitol ratio. J. Clin. Microbiol. 34: Roboz, J., R. Suzuki, and J. F. Holland Quantification of arabinitol in serum by selected ion monitoring as diagnostic technique in invasive candidiasis. J. Clin. Microbiol. 12: Switchenko, A. C., C. G. Miyada, T. C. Goodman, T. J. Walsh, B. Wong, M. J. Becker, and E. F. Ullman An automated enzymatic method for the measurement of D-arabinitol in human serum. J. Clin. Microbiol. 32: Telenti, A., J. M. Steckelberg, L. Stockman, R. S. Edison, and G. D. Robert Quantitative blood cultures in candidemia. Mayo Clin. Proc. 66: Walsh, T. J., and S. J. Chanock Diagnosis of invasive fungal infections: advances in nonculture systems. Curr. Clin. Trop. Infect. Dis. 18: Walsh, T. J., J. W. Lee, T. Sien, R. Schaufele, J. Bacher, A. C. Switchenko, T. C. Goodman, and P. A. Pizzo Serum D-arabinitol measured by automated quantitative enzymatic assay for detection and therapeutic monitoring of experimental disseminated candidiasis: correlation with tissue concentrations of Candida albicans. J. Med. Vet. Mycol. 32: Walsh, T. J., W. G. Merz, J. W. Lee, R. Schaufele, T. Sein, P. O. Whitcomb, M. Ruddel, W. Burns, J. R. Wingard, A. C. Switchenko, T. Goodman, and P. A. Pizzo Diagnosis and therapeutic monitoring of invasive candidiasis by rapid enzymatic detection of serum D-arabinitol. Am. J. Med. 99: Wong, B., E. M. Bernard, J. W. M. Gold, D. Fong, A. Silber, and D. Armstrong Increased arabinitol levels in experimental candidiasis in rats: arabinitol appearance rates, arabinitol/creatinine ratios, and severity of infection. J. Infect. Dis. 146: Wong, B., and K. L. Brauer Enantioselective measurement of fungal D-arabinitol in the sera of normal adults and patients with candidiasis. J. Clin. Microbiol. 26: Wong, B., K. L. Brauer, J. Clemens, and S. Beggs Effects of gastrointestinal candidiasis, dietary arabinitol and cortisone acetate on levels of the Candida metabolite D-arabinitol in rat serum and urine. Infect. Immun. 58: Yeo, S. F., and B. Wong Current status of nonculture diagnostic methods for invasive fungal infections. Clin. Microbiol. Rev. 15: Yeo, S. F., Y. Zhang, D. Schafer, S. Campbell, and B. Wong A rapid, automated enzymatic fluorometric assay for determination of D-arabinitol in serum. J. Clin. Microbiol. 38: