ABSTRACT
This study aimed to develop a prediction model to identify patients with candidemia who were at high risk of failing fluconazole treatment. Adult patients in the United States with candidemia who received fluconazole during hospitalization were selected from the Cerner Health Facts Hospital Database (04/2004 to 03/2013). Fluconazole failure was defined as switching/adding another antifungal, positive Candida culture ≥10 days after fluconazole initiation, or death during hospitalization. Patients were randomized into modeling and validation samples. Using the modeling sample, a regression analysis of least absolute shrinkage and selection operator was used to select risk predictors of fluconazole failure (demographics, Candida species, initiation of fluconazole before positive culture and after admission, and comorbidities, procedures, and treatments during the 6 months before admission and fluconazole initiation). The prediction model was evaluated using the validation sample. We found that of 987 identified patients (average age of 61 years, 51% male, 72% Caucasian), 49% failed and 51% did not fail fluconazole treatment. Of those who failed, 70% switched or added another antifungal, 21% had a second positive Candida test, and 42% died during hospitalization. Nine risk factors were included in the prediction model: days to start fluconazole after admission, Candida glabrata or Candida krusei infection, hematological malignancy, venous thromboembolism (VTE), enteral nutrition, use of nonoperative intubation/irrigation, and other antifungal use. All but VTE were associated with a higher risk of failure. The model's c-statistic was 0.65, with a Hosmer-Lemeshow test P value of 0.23. In summary, this prediction model identified patients with a high risk of fluconazole failure, illustrating the potential value and feasibility of personalizing candidemia treatment.
KEYWORDS: Candida, failure, fluconazole, candidemia, risk score, invasive candidiasis
INTRODUCTION
Tgenus Candida is the most common cause of hospital-acquired bloodstream infections in the United States and a major cause of morbidity and mortality worldwide (1–3). In addition to high morbidity and mortality, candidemia is also associated with a high economic burden of nearly $300 million annually in the United States (4). Compared to a matched cohort of hospitalized patients without candidemia, hospitalized adult patients with candidemia incurred higher hospital charges, by $39,331, and had hospital stays that were 10.1 days longer (5).
Echinocandins (e.g., caspofungin, micafungin, and anidulafungin), fluconazole, and liposomal amphotericin B (AmB) are currently the recommended first-line therapies for the treatment of Candida infections (6, 7). Historically, fluconazole was the recommended first-line therapy in the Infectious Diseases Society of America's (IDSA) treatment guideline for Candida infections, and thus it was the most commonly used treatment (8). In recent years, however, fluconazole-resistant species, such as Candida krusei and Candida glabrata, have been isolated with increasing frequency (9, 10). This is cause for concern (11–13), and recent reports have found that approximately 7% of all Candida bloodstream isolates tested were fluconazole resistant (14). Furthermore, the frequency of resistance among these strains appears to be increasing. An analysis of 313 isolates of Candida glabrata over a 10-year period demonstrated an increase in resistance to fluconazole from 18% to 30% (15, 16). In response to these findings, along with a patient-level clinical trial review showing a mortality benefit when using echinocandins (17), the 2016 update to the IDSA guidelines for the treatment of candidiasis now recommends echinocandins as first-line therapy (6). Fluconazole is considered an acceptable alternative first-line therapy in patients who are not critically ill and who are not likely to have a fluconazole-resistant Candida strain, and the IDSA recommends transition from first-line echinocandin or AmB to fluconazole, usually within a week, for patients who have stabilized and have non-fluconazole-resistant strains (6). The latest IDSA guideline may potentially lead to a gradual shift from first-line fluconazole to first-line echinocandin therapy; however, presently fluconazole is still widely prescribed (4).
An important question in the clinical management of candidemia is how to determine the choice of an effective first-line therapy for an individual patient, i.e., how to determine whether a patient should receive fluconzole versus an echinocandin or other therapy. While microbiological testing (species identification and antifungal susceptibility testing) of admitted patients' Candida cultures can most accurately inform treatment decisions (empirical therapy), waiting for the results of such tests can delay potentially life-saving treatment (18). Some studies have reported a correlation between higher patient mortality and the delay of appropriate antifungal treatment, with earlier initiation associated with lower mortality (19, 20). The odds of mortality increased by 50% with an antifungal therapy initiation delay of just 1 day from the onset of symptoms (19). However, a retrospective analysis assessing mortality rates and the time to initiation of appropriate therapy for candidemia found no correlation with mortality (21). Nevertheless, nosocomial invasive fungal disease has one of the highest rates of inappropriate therapy, and this is associated with increased mortality; thus, early and accurate identification of the proper therapy is of paramount importance (22, 23).
The objective of the current study was to develop a risk score model, based on patients' clinical profiles, to predict the risk of failing fluconazole therapy among patients with candidemia. Such a model will identify patients who are not suitable for fluconazole therapy, allow early initiation of alternative therapies, and may lead to improved clinical outcomes and potential reductions in health care costs among these high-risk patients.
RESULTS
Baseline characteristics.
Among 6,051 patients who had at least one positive blood culture indicating a Candida infection during the index hospitalization, a total of 987 patients met all inclusion criteria and were included in the final study sample (Fig. 1). Of those included, 488 (49%) patients experienced fluconazole failure. Among the patients who experienced fluconazole failure, 70% had switched or added another antifungal therapy, 21% had a subsequent positive blood test for Candida, and 42% died in the hospital. Patient demographics were similar between patients with and without fluconazole failure (Table 1). Both groups had a mean age of 61 years and were ∼51% male and 70% Caucasian. The percentage of patients infected with Candida albicans was 55.4%, with Candida glabrata was 22.1%, with Candida krusei was 0.8%, with Candida parapsilosis was 11.1%, and with other Candida species was 11.3%. Patients with fluconazole failure, compared to those without failure, had a higher rate of Candida glabrata infection (27.5% versus 16.8%; P < 0.001) and a longer delay in initiating fluconazole from admission (22.9 versus 12.8 days; P < 0.001) (Table 1). Compared to patients without fluconazole failure, a higher proportion of patients who experienced fluconazole failure had hematological malignancies (3.5% versus 1.4%; P < 0.05), received nonoperative intubation or irrigation (21.1% versus 12.6%; P < 0.001), received mechanical ventilation (34.4% versus 24.2%; P < 0.001), or were administered other antifungal therapies (22.1% versus 16.2%; P < 0.05) (Table 2) during the baseline period (i.e., the 6-month period before the index date). Other baseline patient characteristics, conditions, and procedures were not significantly different between the two groups.
FIG 1.
Sample selection for patients with candidemia during hospitalization. Footnote 1: the day of initiation of fluconazole treatment was defined as the index date, and the hospitalization was defined as the index hospitalization. Footnote 2: for patients with several encounters meeting all sample selection criteria listed in the figure, a random encounter was selected for analysis to capture both the early and late use of fluconazole.
TABLE 1.
Patient demographics and index hospitalization characteristics
| Demographic or characteristic | Value for patient group |
P valueb | ||
|---|---|---|---|---|
| Total, fluconazole use (n = 987) | With fluconazole failurea (n = 488) | Without fluconazole failure (n = 499) | ||
| Fluconazole failure [n (%)] | 488 (49.4) | 488 (100.0) | ||
| Switched or added other antifungal | 343 (34.8) | 343 (70.3) | ||
| Second positive blood test | 103 (10.4) | 103 (21.1) | ||
| Death in hospital | 205 (20.8) | 205 (42.0) | ||
| Patient demographics | ||||
| Age at index date (mean ± SD) | 61.10 ± 17.12 | 60.95 ± 16.55 | 61.24 ± 17.68 | 0.700 |
| Male [n (%)] | 507 (51.4) | 252 (51.6) | 255 (51.1) | 0.866 |
| Race/ethnicity [n (%)] | ||||
| African-American | 210 (21.3) | 107 (21.9) | 103 (20.6) | 0.622 |
| Caucasian | 708 (71.7) | 341 (69.9) | 367 (73.5) | 0.200 |
| Other | 69 (7.0) | 40 (8.2) | 30 (6.0) | 0.255 |
| Insurance type [n (%)] | ||||
| Commercial | 55 (5.6) | 21 (4.3) | 34 (6.8) | 0.086 |
| Medicare | 315 (31.9) | 159 (32.6) | 156 (31.3) | 0.657 |
| Medicaid | 104 (10.5) | 49 (10.0) | 55 (11.0) | 0.616 |
| Otherc | 112 (11.3) | 63 (12.9) | 49 (9.8) | 0.126 |
| Unknown | 401 (40.6) | 196 (40.2) | 205 (41.1) | 0.769 |
| Index hospitalization characteristics | ||||
| Index year [n (%)] | ||||
| 2005–2007 | 295 (29.9) | 160 (32.8) | 135 (27.1) | 0.049* |
| 2008–2010 | 461 (46.7) | 245 (50.2) | 216 (43.3) | 0.029* |
| 2011–2013 | 231 (23.4) | 83 (17.0) | 148 (29.7) | <0.001* |
| Candida species isolated [n (%)] | ||||
| C. albicans | 547 (55.4) | 256 (52.5) | 291 (58.3) | 0.064 |
| C. glabrata | 218 (22.1) | 134 (27.5) | 84 (16.8) | <0.001* |
| C. krusei | 8 (0.8) | 6 (1.2) | 2 (0.4) | 0.173 |
| C. parapsilosis | 110 (11.1) | 48 (9.8) | 62 (12.4) | 0.196 |
| Other | 112 (11.3) | 47 (9.6) | 65 (13.0) | 0.093 |
| Time (days) from admission until start of fluconazole treatment (mean ± SD) | 17.76 ± 48.25 | 22.88 ± 66.52 | 12.76 ± 15.20 | <0.001* |
| Fluconazole use prior to positive blood culture [n (%)] | 847 (85.8) | 429 (87.9) | 418 (83.8) | 0.062 |
Fluconazole failure was defined as switching to or adding another antifungal therapy following fluconazole initiation, a positive blood culture for Candida infection during the index hospitalization and at least 10 days following fluconazole initiation, or death during index hospital visit. These three criteria are not mutually exclusive.
P values were estimated using the Wilcoxon signed-rank test for continuous variables and chi-square test for binary variables. *, P < 0.05.
“Other” insurance includes other government, other nongovernment, undesignated HMO or PPO, self-pay, or worker's compensation.
TABLE 2.
Baseline risk factors
| Baseline conditiona | No. (%) among patient group |
P valuec | ||
|---|---|---|---|---|
| Total with fluconazole use (n = 987) | With fluconazole failureb (n = 488) | Without fluconazole failure (n = 499) | ||
| CCI (mean ± SD) | 1.58 ± 2.33 | 1.66 ± 2.42 | 1.51 ± 2.24 | 0.598 |
| All diagnosesd | ||||
| Hematological malignancy | 24 (2.4) | 17 (3.5) | 7 (1.4) | 0.034* |
| Solid malignancy | 109 (11.0) | 56 (11.5) | 53 (10.6) | 0.669 |
| Neutropenia | 12 (1.2) | 9 (1.8) | 3 (0.6) | 0.075 |
| Renal failure (including hemodialysis) | 254 (25.7) | 120 (24.6) | 134 (26.9) | 0.416 |
| Hepatitis | 55 (5.6) | 33 (6.8) | 22 (4.4) | 0.107 |
| Gastroenteritis | 69 (7.0) | 36 (7.4) | 33 (6.6) | 0.638 |
| Abscess | 79 (8.0) | 34 (7.0) | 45 (9.0) | 0.235 |
| Venous thromboembolism | 52 (5.3) | 22 (4.5) | 30 (6.0) | 0.290 |
| Diabetes | 152 (15.4) | 75 (15.4) | 77 (15.4) | 0.978 |
| Type I | 15 (1.5) | 8 (1.6) | 7 (1.4) | 0.761 |
| Type II | 145 (14.7) | 69 (14.1) | 76 (15.2) | 0.628 |
| Severe sepsis/septicemia | 157 (15.9) | 73 (15.0) | 84 (16.8) | 0.421 |
| Graft-vs-host disease | 1 (0.1) | 1 (0.2) | 0 (0.0) | 0.494 |
| Liver disease (moderate to severe) | 27 (2.7) | 18 (3.7) | 9 (1.8) | 0.070 |
| Disorder of fluid, electrolyte, or acid-base balance | 251 (25.4) | 115 (23.6) | 136 (27.3) | 0.183 |
| Disease of lung | 138 (14.0) | 67 (13.7) | 71 (14.2) | 0.821 |
| Anemia | 250 (25.3) | 115 (23.6) | 135 (27.1) | 0.208 |
| Protein-calorie malnutrition | 131 (13.3) | 60 (12.3) | 71 (14.2) | 0.371 |
| Essential hypertension | 189 (19.1) | 91 (18.6) | 98 (19.6) | 0.692 |
| HIV | 4 (0.4) | 3 (0.6) | 1 (0.2) | 0.369 |
| Invasive fungal infection | 435 (44.1) | 216 (44.3) | 219 (43.9) | 0.906 |
| Candidemiae | 409 (41.4) | 204 (41.8) | 205 (41.1) | 0.818 |
| Other fungal infectionsf | 58 (5.9) | 26 (5.3) | 32 (6.4) | 0.469 |
| Procedures | ||||
| Parenteral nutrition | 177 (17.9) | 81 (16.6) | 96 (19.2) | 0.280 |
| Enteral nutrition | 74 (7.5) | 44 (9.0) | 30 (6.0) | 0.073 |
| Organ transplant | 11 (1.1) | 6 (1.2) | 5 (1.0) | 0.734 |
| Stem cell transplant | 1 (0.1) | 1 (0.2) | 0 (0.0) | 0.494 |
| Use, replacement, or removal of central venous catheters | 391 (39.6) | 192 (39.3) | 199 (39.9) | 0.863 |
| Use of drain | 7 (0.7) | 2 (0.4) | 5 (1.0) | 0.452 |
| Use of prosthetic device | 17 (1.7) | 7 (1.4) | 10 (2.0) | 0.492 |
| Operation on larynx and trachea | 62 (6.3) | 33 (6.8) | 29 (5.8) | 0.538 |
| Use of nonoperative intubation or irrigation (excluding mechanical ventilation) | 166 (16.8) | 103 (21.1) | 63 (12.6) | <0.001* |
| Ileostomy | 21 (2.1) | 13 (2.7) | 8 (1.6) | 0.248 |
| Mechanical ventilation | 289 (29.3) | 168 (34.4) | 121 (24.2) | <0.001* |
| Treatments | ||||
| Corticosteroids | 58 (5.9) | 30 (6.1) | 28 (5.6) | 0.720 |
| Other antifungalsg | 189 (19.1) | 108 (22.1) | 81 (16.2) | 0.019* |
Baseline factors were defined as those occurring between 6 months prior to the index hospitalization admission and the index date. CCI, Charlson comorbidity index.
Fluconazole failure was defined as (i) switching to or adding another antifungal therapy following fluconazole initiation, (ii) a positive blood culture for Candida infection during the index hospitalization and at least 10 days following fluconazole initiation, or (iii) death during index hospital visit. These three criteria are not mutually exclusive.
P values were estimated using the Wilcoxon signed-rank test for continuous variables and Chi-square tests for binary variables. *, P <0.05.
Includes primary and secondary diagnoses.
ICD-9-CM codes 112.4 (candidiasis of lung), 112.5 (disseminated candidiasis), 112.8x (e.g., candidal endocarditis, candidal meningitis), and 112.9 (candidiasis of unspecified site). These infections were not necessarily the index candidemia infection.
ICD-9-CM codes 111.2 (tinea blanca) and 114x-118 (e.g., coccidioidal meningitis, blastomycosis, opportunistic mycoses).
Other antifungals included amphotericin, flucytosine, itraconazole, voriconazole, posaconazole, caspofungin, anidulafungin, and micafungin.
Development of a model estimating risk factors for fluconazole failure.
A total of 46 candidate risk factors were evaluated for inclusion in the model to predict fluconazole failure. Nine factors—infection with one of two Candida species (Candida glabrata and Candida krusei), time to initiate fluconazole, hematological malignancy, venous thromboembolism, enteral nutrition, use of nonoperative intubation or irrigation, use of mechanical ventilation, and use of other antifungals—were selected via the LASSO method (a regression analysis of the least absolute shrinkage and selection operator) to be included in the final prediction model for fluconazole failure (Table 3). Among these evaluated risk factors, the following were associated with a higher risk of fluconazole failure: longer time to start fluconazole after hospital admission, Candida glabrata infection, Candida krusei infection, hematological malignancy, enteral nutrition, use of nonoperative intubation or irrigation, use of mechanical ventilation, and use of other antifungals (Table 3). Conversely, venous thromboembolism was associated with a lower risk of fluconazole failure.
TABLE 3.
Risk prediction modela
| Selected risk factor | Estimate | SE | Odds ratiob |
|---|---|---|---|
| Time (days) to start fluconazole after admission | 0.02 | 0.01 | 1.02 |
| Candida species | |||
| C. glabrata | 0.68 | 0.20 | 1.98 |
| C. krusei | 2.10 | 1.12 | 8.20 |
| Baseline period conditions | |||
| Diagnoses | |||
| Hematological malignancies | 1.18 | 0.60 | 3.25 |
| Venous thromboembolism | −0.67 | 0.41 | 0.51 |
| Procedures | |||
| Enteral nutrition | 0.44 | 0.32 | 1.56 |
| Nonoperative intubation or irrigation (excluding mechanical ventilation) | 0.43 | 0.29 | 1.54 |
| Mechanical ventilation | 0.12 | 0.24 | 1.13 |
| Treatments | |||
| Other antifungals | 0.34 | 0.21 | 1.40 |
The model diagnostic statistics for the validation sample resulted in a c-statistic estimate of 0.65 and a Hosmer-Lemeshow P value of 0.23.
The odds ratio is the exponent value of the estimate.
Evaluation of the risk-prediction model.
The c-statistic of the risk prediction model was 0.65, suggesting good discrimination between patients who failed fluconazole therapy and those that did not. The P value from the Hosmer-Lemeshow test in the validation sample was 0.23, indicating an adequate fit of the model to the data. The sensitivities, specificities, positive predictive values (PPVs), and negative predictive values (NPVs) for the model, based on different fluconazole failure probability cutoffs ranging from 0.1 to 0.9, are shown in Table 4. The probability cutoff for maximizing both sensitivity (56%) and specificity (68%) was 0.5.
TABLE 4.
Evaluation of the prediction model
| Cutoff of predicted failure risk scorea | No. of observed failures | No. of predicted failures | No. with correctly predicted failure | % true positives (sensitivityb) | % true negatives (specificityc) | Positive predictive valued (%) | Negative predictive valuee (%) |
|---|---|---|---|---|---|---|---|
| 0.1 | 163 | 330 | 163 | 100 | 0 | 49 | |
| 0.2 | 163 | 326 | 162 | 99 | 2 | 50 | 75 |
| 0.3 | 163 | 296 | 158 | 97 | 17 | 53 | 85 |
| 0.4 | 163 | 224 | 127 | 78 | 42 | 57 | 66 |
| 0.5 | 163 | 145 | 91 | 56 | 68 | 63 | 61 |
| 0.6 | 163 | 80 | 53 | 33 | 84 | 66 | 56 |
| 0.7 | 163 | 34 | 25 | 15 | 95 | 74 | 53 |
| 0.8 | 163 | 13 | 10 | 6 | 98 | 77 | 52 |
| 0.9 | 163 | 5 | 5 | 3 | 100 | 100 | 51 |
The cutoff of the predictive failure rate used to define fluconazole failure. For example, a cutoff of 0.4 indicates that all patients with a risk score higher or equal to 0.4 were defined as experiencing fluconazole failure.
Sensitivity is the number of correctly predicted failures divided by the number of observed failures.
Specificity is the number of correctly predicted nonfailures divided by the number of observed nonfailures.
Positive predictive value is the number of correctly predicted failures divided by number of predicted failures.
Negative predictive value is the number of correctly predicted nonfailures divided by the number of predicted nonfailures.
DISCUSSION
The increasing prevalence of fluconazole-resistant Candida species, as well as the lack of comprehensive identification of certain patient risk factors for fluconazole failure, jeopardize the effective, timely treatment of and may increase mortality for patients with candidemia. This study constructed a prediction model to evaluate the risk of fluconazole failure among patients with candidemia and subsequently identified the following risk factors: infection with Candida glabrata or Candida krusei species, time to initiate fluconazole, hematological malignancy, venous thromboembolism, enteral nutrition, use of nonoperative intubation or irrigation, use of mechanical ventilation, and use of other antifungals before initiating fluconazole. With the exception of venous thromboembolism, which actually was a protective risk factor, all of these factors were associated with a higher risk of fluconazole failure and represent a constellation of risk factors that can aid in screening candidates for fluconazole treatment.
The findings of this study are supported by several prior studies noting certain risk factors for fluconazole failure or mortality. The Candida species C. glabrata and C. krusei are known to be resistant to fluconazole (6, 24), and this supports the suggestion of using their positive presence as predictors of fluconazole failure. Almirante et al. reported that intubation after Candida infection and hematological malignancy, two risk factors noted here for fluconazole failure, were associated with a higher risk of mortality (25). In addition, Kabbani reported that the presence of hematologic malignancy is also associated with a higher risk of fluconazole failure (24). However, these studies primarily identified single or a few risk factors, while the current study incorporated multiple risk factors simultaneously into a more comprehensive tool for evaluating the risk of failing fluconazole treatment. The finding of venous thromboembolism as a protective factor was unexpected. A potential explanation could be that heparin is commonly used for patients with thromboembolism (26). Interestingly, heparin has been reported to bind to C. albicans via the virulence factor Int1 and to alter candidal surface structures (27, 28), resulting in significant attenuation of Candida biofilm formation and thus its infective persistence (29). This might help explain the protective effects of thromboembolism observed in the model.
The present study used LASSO regression analysis to identify potential risk factors. In contrast with traditional model selection approaches, such as stepwise selection or use predefined covariates, the LASSO method balances parsimony and generalizability with the predictive power of the model (30). Of the identified risk factors for fluconazole failure, some were more commonly observed in the study population (use of mechanical ventilation, use of nonoperative intubation or irrigation, infection with Candida glabrata, and use of other antifungals) than other factors (infection with Candida krusei, hematological malignancy, use of venous thromboembolism, and enteral nutrition). For example, mechanical ventilation was commonly observed among 29% of patients with fluconazole use and was significantly higher in patients with fluconazole failure (34%) than in those without failure (24%). Additionally, 22% of all patients in this study had Candida glabrata infection, with 28% and 17% in the fluconazole failure and nonfailure groups, respectively. Around 20% of the total study sample had used other antifungals during the baseline period, and a significantly higher proportion of patients with fluconazole failure used other antifungals than did those without failure (22% versus 16%, respectively). In addition, we observed that although some patients were identified as having Candida species likely to be resistant to fluconazole, fluconazole was still used among these patients (Table 1), further underscoring the need for appropriate therapy.
Our study findings complement the most recent updates to treatment guidelines for Candida infection management, which suggest that fluconazole be only administered to stable patients unlikely to have resistant infections (6), and our study provides a clinically accessible tool to assist in identification of such individual patients. The approach outlined in this study, with the development of risk prediction scores based on identified risk factors, has successfully been utilized to identify patients who are at high risk of developing candidemia. These methods include a score for peritoneal Candida infection among patients in intensive care units (31), a Candida colonization index for patients undergoing surgery (32), a five-pronged clinical score (33), and the bedside scoring system, known as the Candida score (34). These systems were considered to be useful in the selection process for antifungal prophylaxis among at-risk patients (35); the addition of a scoring system to aid in the identification of patients at risk of fluconazole failure, such as the one proposed here, may further improve patient outcomes.
The high clinical and economic burdens associated with candidemia underscore an unmet need for individualized treatment regimens in order to improve quality of care. An informed understanding of which patients with candidiasis might be most at risk for treatment failure may improve patient outcomes. This may be especially true for patients receiving fluconazole prior to microbiological testing, as well as patients transitioning from other first-line therapies; despite the latest IDSA guidelines recommending it for only a subset of patients (6), fluconazole is still widely prescribed (4). The findings of this study suggest that personalized treatment of candidemia based on patient and disease characteristics is feasible, and our findings may have implications for the treatment of this life-threatening infection.
This study was subject to common limitations of retrospective observational studies based on hospital databases, such as data omissions or coding errors. Additionally, the risk factors of fluconazole failure included in this study were subject to data availability and limitations in the Cerner database. Potential risk factors that were not available in the database could not be included in this study (e.g., previous surgery, history of intravenous drug abuse). Future studies using alternative data sources with rich clinical information (e.g., fever and blood cell counts) are warranted to validate the current model. The current definition of fluconazole failure has its limitations, but it was nevertheless used given its clinical relevance and the consideration that infection data were not always systematically collected in clinical practice to confirm a treatment failure, thus use of only infection data and mortality would underestimate the failure rate. As an external validation, the fluconazole failure rate observed in our study was consistent with rates reported in prior literature (36). The current study only included patients who initiated intravenous fluconazole (although they could switch to oral fluconazole) and did not include patients who initiated oral fluconazole, in order to help select patients who used fluconazole for treatment as opposed to prophylaxis. Future research might be conducted to evaluate the predictors of oral fluconazole failure.
Conclusions.
This study identified important clinical factors that can be used in combination to predict the risk of fluconazole failure among individual hospitalized patients. The prediction model identified patients with a high risk of fluconazole failure, illustrating the potential value and feasibility of personalizing candidemia treatment to improve patient outcomes.
MATERIALS AND METHODS
Data source.
This study used the Cerner Health Facts Hospital Database, a commercially available electronic deidentified health record database used in over 150 hospitals throughout the United States. The database contains 110 million records spanning from 04/01/2004 to 03/31/2013 and includes patient demographic information as well as medical (admission, discharge, diagnoses, procedures, etc.), hospital pharmacy, and laboratory data (including microbiology and pathogen information). Data collected were compliant with the Health Insurance Portability and Accountability Act and the Declaration of Helsinki (1964, amended 2008); no ethical review was required.
Sample selection and construction.
Adult patients aged 18 years and older who had at least one positive blood culture indicating Candida infection during a hospitalization were selected. Patients were further required to have initiated intravenous fluconazole treatment during their hospital stay and no more than 5 days before the positive blood culture reading for Candida infection. Included patients were also required to have at least one nonmissing underlying diagnosis during their hospitalization. For patients who met all sample selection criteria and had multiple hospitalization records, one hospitalization per patient was randomly selected. This was to allow capture of both the early and late use of fluconazole during the course of the Candida infection and to avoid confounding in the results due to correlation if all observations from the same patient were included.
Key definitions.
The index hospitalization was defined as a hospitalization encounter that met all sample selection criteria. The index date was defined as the date of initiation of fluconazole treatment as the first-line treatment during the index hospitalization. The baseline period was defined as the period 6 months prior to the index hospital admission through the index date (fluconazole initiation). Patients with fluconazole failure were defined as patients who met any of the following criteria: (i) switched to or added another antifungal therapy following fluconazole initiation; (ii) had a subsequent positive blood culture for Candida infection during the index hospitalization and at least 10 days following fluconazole initiation; or (iii) died during the period from index date to end of index hospitalization. For criterion ii, the IDSA treatment guidelines do not specify a time for the definition of fluconazole failure (8). The National Committee on Clinical Laboratory Standards (now the Clinical Laboratory Standards Institute) proposed in 1998 to define fluconazole failure as lack of response to systematic fluconazole therapy within 14 days (37). Given that the initiation of fluconazole was allowed within 5 days of a positive blood culture in the current study, a time frame of 10 days was used to evaluate the recovery from candidemia based on a blood culture result.
Study measures.
Patient-level measures were collected from the database and included demographics (age, sex, and race), baseline period risk factors (Charlson comorbidity index [CCI] [38], diagnoses, procedures, and treatments received) within 6 months prior to the index date, and characteristics of the index hospitalization (index year, Candida species, time to initiate fluconazole from admission, and prior fluconazole use). All study measures were compared between patients with and without fluconazole failure by using Wilcoxon signed-rank tests for continuous variables and chi-square tests for categorical variables.
Development of a prediction model for risk of fluconazole failure.
An optimal set of risk factors for predicting fluconazole failure were selected from a list of candidate risk factors by using the LASSO approach (30), which permitted simultaneous selection of a set of risk factors instead of predefining risk factors or stepwise selection. Candidate risk factors included patient demographics, such as age, sex, and race, index year, index hospitalization characteristics, such as Candida species during the index hospitalization, time to start fluconazole from admission, prior fluconazole use, and conditions (diagnoses, procedures, and treatments) during the baseline period. Baseline conditions with frequencies of more than 1% in the entire sample were considered candidate risk factors.
Once the optimal list of risk factors was selected, a prediction model was then developed using a cross-validation approach. Total samples were partitioned into modeling and validation samples stratified by fluconazole failure status at a ratio of 2 to 1. A logistic regression model was developed in the modeling sample, and model performance was evaluated based on the c-statistic and the Hosmer-Lemeshow test (39) in the validation sample. Sensitivity, specificity, PPV, and NPV were estimated for various cutoffs of the predictive failure rate that defined fluconazole failure, using the validation sample (40).
Statistical analyses.
Statistical analyses were performed within the program SAS version 9.4 (SAS Institute Inc., Cary, NC) and R version 3.1.2 (open source; The R Foundation for Statistical Computing). A two-tailed P value of 0.05 was used to determine significance.
ACKNOWLEDGMENTS
This study was funded by Astellas Pharma.
Medical writing assistance was provided by Shelley Batts, an employee of Analysis Group, Inc., who was ultimately paid by the sponsor.
REFERENCES
- 1.Concia E, Azzini AM, Conti M. 2009. Epidemiology, incidence and risk factors for invasive candidiasis in high-risk patients. Drugs 69(Suppl 1):S5–S14. doi: 10.2165/11315500-000000000-00000. [DOI] [PubMed] [Google Scholar]
- 2.Horn DL, Fishman JA, Steinbach WJ, Anaissie EJ, Marr KA, Olyaei AJ, Pfaller MA, Weiss MA, Webster KM, Neofytos D. 2007. Presentation of the PATH Alliance registry for prospective data collection and analysis of the epidemiology, therapy, and outcomes of invasive fungal infections. Diagn Microbiol Infect Dis 59:407–414. doi: 10.1016/j.diagmicrobio.2007.06.008. [DOI] [PubMed] [Google Scholar]
- 3.Magill SS, Edwards JR, Bamberg W, Beldavs ZG, Dumyati G, Kainer MA, Lynfield R, Maloney M, McAllister-Hollod L, Nadle J, Ray SM, Thompson DL, Wilson LE, Fridkin SK, Emerging Infections Program Healthcare-Associated Infections and Antimicrobial Use Prevalence Survey Team. 2014. Multistate point-prevalence survey of health care-associated infections. N Engl J Med 370:1198–1208. doi: 10.1056/NEJMoa1306801. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Neoh CF, Slavin M, Chen SC, Stewart K, Kong DC. 2014. Echinocandins in the treatment of candidaemia and invasive candidiasis: clinical and economic perspectives. Int J Antimicrob Agents 43:207–214. doi: 10.1016/j.ijantimicag.2013.08.010. [DOI] [PubMed] [Google Scholar]
- 5.Zaoutis TE, Argon J, Chu J, Berlin JA, Walsh TJ, Feudtner C. 2005. The epidemiology and attributable outcomes of candidemia in adults and children hospitalized in the United States: a propensity analysis. Clin Infect Dis 41:1232–1239. doi: 10.1086/496922. [DOI] [PubMed] [Google Scholar]
- 6.Pappas PG, Kauffman CA, Andes DR, Clancy CJ, Marr KA, Ostrosky-Zeichner L, Reboli AC, Schuster MG, Vazquez JA, Walsh TJ, Zaoutis TE, Sobel JD. 2016. Clinical practice guideline for the management of candidiasis: 2016 update by the Infectious Diseases Society of America. Clin Infect Dis 62:e1–e50. doi: 10.1093/cid/civ933. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Hope WW, Castagnola E, Groll AH, Roilides E, Akova M, Arendrup MC, Arikan-Akdagli S, Bassetti M, Bille J, Cornely OA, Cuenca-Estrella M, Donnelly JP, Garbino J, Herbrecht R, Jensen HE, Kullberg BJ, Lass-Florl C, Lortholary O, Meersseman W, Petrikkos G, Richardson MD, Verweij PE, Viscoli C, Ullmann AJ, ESCMID Fungal Infection Study Group. 2012. ESCMID* guideline for the diagnosis and management of Candida diseases 2012: prevention and management of invasive infections in neonates and children caused by Candida spp. Clin Microbiol Infect 18(Suppl 7):S38–S52. doi: 10.1111/1469-0691.12040. [DOI] [PubMed] [Google Scholar]
- 8.Pappas PG, Kauffman CA, Andes D, Benjamin DK Jr, Calandra TF, Edwards JE Jr, Filler SG, Fisher JF, Kullberg BJ, Ostrosky-Zeichner L, Reboli AC, Rex JH, Walsh TJ, Sobel JD, Infectious Diseases Society of America. 2009. Clinical practice guidelines for the management of candidiasis: 2009 update by the Infectious Diseases Society of America. Clin Infect Dis 48:503–535. doi: 10.1086/596757. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Bassetti M, Taramasso L, Nicco E, Molinari MP, Mussap M, Viscoli C. 2011. Epidemiology, species distribution, antifungal susceptibility and outcome of nosocomial candidemia in a tertiary care hospital in Italy. PLoS One 6:e24198. doi: 10.1371/journal.pone.0024198. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Bassetti M, Righi E, Costa A, Fasce R, Molinari MP, Rosso R, Pallavicini FB, Viscoli C. 2006. Epidemiological trends in nosocomial candidemia in intensive care. BMC Infect Dis 6:21. doi: 10.1186/1471-2334-6-21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Quindos G. 2014. Epidemiology of candidaemia and invasive candidiasis: a changing face. Rev Iberoam Micol 31:42–48. doi: 10.1016/j.riam.2013.10.001. [DOI] [PubMed] [Google Scholar]
- 12.Goncalves SS, Souza AC, Chowdhary A, Meis JF, Colombo AL. 26 January 2016. Epidemiology and molecular mechanisms of antifungal resistance in Candida and Aspergillus. Mycoses doi: 10.1111/myc.12469. [DOI] [PubMed] [Google Scholar]
- 13.Rex JH, Rinaldi MG, Pfaller MA. 1995. Resistance of Candida species to fluconazole. Antimicrob Agents Chemother 39:1–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Cleveland AA, Farley MM, Harrison LH, Stein B, Hollick R, Lockhart SR, Magill SS, Derado G, Park BJ, Chiller TM. 2012. Changes in incidence and antifungal drug resistance in candidemia: results from population-based laboratory surveillance in Atlanta and Baltimore, 2008-2011. Clin Infect Dis 55:1352–1361. doi: 10.1093/cid/cis697. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Alexander BD, Johnson MD, Pfeiffer CD, Jimenez-Ortigosa C, Catania J, Booker R, Castanheira M, Messer SA, Perlin DS, Pfaller MA. 2013. Increasing echinocandin resistance in Candida glabrata: clinical failure correlates with presence of FKS mutations and elevated minimum inhibitory concentrations. Clin Infect Dis 56:1724–1732. doi: 10.1093/cid/cit136. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Pfaller MA, Andes DR, Diekema DJ, Horn DL, Reboli AC, Rotstein C, Franks B, Azie NE. 2014. Epidemiology and outcomes of invasive candidiasis due to non-albicans species of Candida in 2,496 patients: data from the Prospective Antifungal Therapy (PATH) registry 2004-2008. PLoS One 9:e101510. doi: 10.1371/journal.pone.0101510. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Andes DR, Safdar N, Baddley JW, Playford G, Reboli AC, Rex JH, Sobel JD, Pappas PG, Kullberg BJ, Mycoses Study Group. 2012. Impact of treatment strategy on outcomes in patients with candidemia and other forms of invasive candidiasis: a patient-level quantitative review of randomized trials. Clin Infect Dis 54:1110–1122. doi: 10.1093/cid/cis021. [DOI] [PubMed] [Google Scholar]
- 18.Parkins MD, Sabuda DM, Elsayed S, Laupland KB. 2007. Adequacy of empirical antifungal therapy and effect on outcome among patients with invasive Candida species infections. J Antimicrob Chemother 60:613–618. doi: 10.1093/jac/dkm212. [DOI] [PubMed] [Google Scholar]
- 19.Garey KW, Rege M, Pai MP, Mingo DE, Suda KJ, Turpin RS, Bearden DT. 2006. Time to initiation of fluconazole therapy impacts mortality in patients with candidemia: a multi-institutional study. Clin Infect Dis 43:25–31. doi: 10.1086/504810. [DOI] [PubMed] [Google Scholar]
- 20.Klevay MJ, Ernst EJ, Hollanbaugh JL, Miller JG, Pfaller MA, Diekema DJ. 2008. Therapy and outcome of Candida glabrata versus Candida albicans bloodstream infection. Diagn Microbiol Infect Dis 60:273–277. doi: 10.1016/j.diagmicrobio.2007.10.001. [DOI] [PubMed] [Google Scholar]
- 21.Kludze-Forson M, Eschenauer GA, Kubin CJ, Della-Latta P, Lam SW. 2010. The impact of delaying the initiation of appropriate antifungal treatment for Candida bloodstream infection. Med Mycol 48:436–439. doi: 10.3109/13693780903208256. [DOI] [PubMed] [Google Scholar]
- 22.Horn DL, Neofytos D, Anaissie EJ, Fishman JA, Steinbach WJ, Olyaei AJ, Marr KA, Pfaller MA, Chang CH, Webster KM. 2009. Epidemiology and outcomes of candidemia in 2019 patients: data from the prospective antifungal therapy alliance registry. Clin Infect Dis 48:1695–1703. doi: 10.1086/599039. [DOI] [PubMed] [Google Scholar]
- 23.Morrell M, Fraser VJ, Kollef MH. 2005. Delaying the empiric treatment of candida bloodstream infection until positive blood culture results are obtained: a potential risk factor for hospital mortality. Antimicrob Agents Chemother 49:3640–3645. doi: 10.1128/AAC.49.9.3640-3645.2005. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Kabbani S, Stein B, Hollick R, Lockhart S, Lyon GM III, Chiller T, Harrison L, Farley M. 2015. Epidemiology and risk factors for fluconazole resistance and mortality among adults with candidemia in Atlanta and Baltimore, abstr 216. IDWeek Final Program. Infectious Diseases Society of America, Arlington, VA. [Google Scholar]
- 25.Almirante B, Rodriguez D, Park BJ, Cuenca-Estrella M, Planes AM, Almela M, Mensa J, Sanchez F, Ayats J, Gimenez M, Saballs P, Fridkin SK, Morgan J, Rodriguez-Tudela JL, Warnock DW, Pahissa A, Barcelona Candidemia Project Study Group. 2005. Epidemiology and predictors of mortality in cases of Candida bloodstream infection: results from population-based surveillance, Barcelona, Spain, from 2002 to 2003. J Clin Microbiol 43:1829–1835. doi: 10.1128/JCM.43.4.1829-1835.2005. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.McRae SJ, Ginsberg JS. 2004. Initial treatment of venous thromboembolism. Circulation 110(Suppl 1):I3–I9. doi: 10.1161/01.CIR.0000140904.52752.0c. [DOI] [PubMed] [Google Scholar]
- 27.Green JV, Orsborn KI, Zhang M, Tan QK, Greis KD, Porollo A, Andes DR, Long Lu J, Hostetter MK. 2013. Heparin-binding motifs and biofilm formation by Candida albicans. J Infect Dis 208:1695–1704. doi: 10.1093/infdis/jit391. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Miceli MH, Bernardo SM, Ku TS, Walraven C, Lee SA. 2012. In vitro analyses of the effects of heparin and parabens on Candida albicans biofilms and planktonic cells. Antimicrob Agents Chemother 56:148–153. doi: 10.1128/AAC.05061-11. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Ramage G, Saville SP, Thomas DP, Lopez-Ribot JL. 2005. Candida biofilms: an update. Eukaryot Cell 4:633–638. doi: 10.1128/EC.4.4.633-638.2005. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Tibshirani R. 1996. Regression shrinkage and selection via the LASSO. J R Stat Soc B Stat Methodol 73:267–288. doi: 10.1111/j.1467-9868.2011.00771.x. [DOI] [Google Scholar]
- 31.Dupont H, Bourichon A, Paugam-Burtz C, Mantz J, Desmonts JM. 2003. Can yeast isolation in peritoneal fluid be predicted in intensive care unit patients with peritonitis? Crit Care Med 31:752–757. [DOI] [PubMed] [Google Scholar]
- 32.Pittet D, Monod M, Suter PM, Frenk E, Auckenthaler R. 1994. Candida colonization and subsequent infections in critically ill surgical patients. Ann Surg 220:751–758. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Ostrosky-Zeichner L, Sable C, Sobel J, Alexander BD, Donowitz G, Kan V, Kauffman CA, Kett D, Larsen RA, Morrison V, Nucci M, Pappas PG, Bradley ME, Major S, Zimmer L, Wallace D, Dismukes WE, Rex JH. 2007. Multicenter retrospective development and validation of a clinical prediction rule for nosocomial invasive candidiasis in the intensive care setting. Eur J Clin Microbiol Infect Dis 26:271–276. doi: 10.1007/s10096-007-0270-z. [DOI] [PubMed] [Google Scholar]
- 34.Leon C, Ruiz-Santana S, Saavedra P, Almirante B, Nolla-Salas J, Alvarez-Lerma F, Garnacho-Montero J, Leon MA, EPCAN Study Group. 2006. A bedside scoring system (“Candida score”) for early antifungal treatment in nonneutropenic critically ill patients with Candida colonization. Crit Care Med 34:730–737. [DOI] [PubMed] [Google Scholar]
- 35.Ahmed A, Azim A, Baronia AK, Marak KR, Gurjar M. 2014. Risk prediction for invasive candidiasis. Indian J Crit Care Med 18:682–688. doi: 10.4103/0972-5229.142178. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Charlier C, Hart E, Lefort A, Ribaud P, Dromer F, Denning DW, Lortholary O. 2006. Fluconazole for the management of invasive candidiasis: where do we stand after 15 years? J Antimicrob Chemother 57:384–410. doi: 10.1093/jac/dki473. [DOI] [PubMed] [Google Scholar]
- 37.Fichtenbaum CJ, Powderly WG. 1998. Refractory mucosal candidiasis in patients with human immunodeficiency virus infection. Clin Infect Dis 26:556–565. [DOI] [PubMed] [Google Scholar]
- 38.Charlson ME, Pompei P, Ales KL, MacKenzie CR. 1987. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 40:373–383. [DOI] [PubMed] [Google Scholar]
- 39.Hosmer DW, Lemeshow S, Sturdivant RX. 2013. Applied logistic regression, 3rd ed Wiley, Hoboken, NJ. [Google Scholar]
- 40.Skaik YA. 2008. Understanding and using sensitivity, specificity and predictive values. Indian J Ophthalmol 56:341. doi: 10.4103/0301-4738.41424. [DOI] [PMC free article] [PubMed] [Google Scholar]