To the Editor,

Certain allogeneic hematopoietic stem cell transplantation (allo-HSCT) recipients develop candidemia even while undergoing antifungal therapy; this is known as breakthrough candidemia (BC) [1]. Despite advances in antifungal agents, BC remains a fatal complication in allo-HSCT recipients with a mortality rate of 40–50% [2]. The purpose of this study was to describe the clinical characteristics and identify the risk factors of BC in allo-HSCT using a matched-pair analysis due to the limited data available. This study was performed in accordance with the Declaration of Helsinki and approved by the ethics committee of our hospital. Statistical analyses were performed with EZR [3].

From August 2006 to August 2017, 765 patients underwent 866 allo-HSCT procedures at our institution and their data were retrospectively reviewed. BC was defined as the isolation of any Candida species from at least one blood culture when a patient had received antifungal agents for at least 3 days prior to detection of the first positive blood culture [4]. Detailed transplantation procedures were described elsewhere [5]. The standard prophylaxis for fungal infection during the allo-HSCT consisted of fluconazole or itraconazole. The antifungal agents used could be changed at a physician’s discretion.

Blood culture samples were processed using the BacT/ALERT 3D automated system (bioMérieux, Marcy l'Étoile, France) with an incubation period of 7 days. All Candida isolates were recovered on Sabouraud Dextrose Agar (Nikken Biomedical Laboratory Inc., Kyoto, Japan) at 35 °C and were identified to the species level through the use of CHROMAgar Candida medium (Becton Dickinson Company, Ltd., Tokyo, Japan), API 20 C AUX (bioMérieux), and the MALDI Microflex LT mass spectrometer (Bruker Daltonik GmbH, Bremen, Germany). Antifungal susceptibility testing was performed via the broth microdilution method using an ASTY kit (Kyokuto Pharmaceutical Industrial Co., Ltd., Japan). Serum (1,3)-beta-D-glucan (BDG) was measured by the Fungitec G test MK II (Nissui Pharmaceutical Co., Ltd., Tokyo, Japan), using a cutoff value of 20 pg/ml [6].

The nutritional risk index (NRI), which was originally derived from the serum albumin concentration and the ratio of present to usual body weight (UBW), is known to represent the global nutritional status; its clinical significance has been validated in patients with various diseases [7]. Serum albumin was evaluated using the most recent data prior to the initiation of conditioning (median 1 day, range 0–21 days before the first day of conditioning). Body weight (BW) was evaluated on the morning of the first day of conditioning. However, owing to the difficulty in identifying UBW of allo-HSCT recipients, we used ideal body weight (IBW) instead of UBW. NRI was calculated using the following equation: 14.89 × albumin (g/dl) + 41.7 × (BW/IBW). IBW was calculated from height and body mass index (BMI) of 22 (kg/m2). When the patient’s BW exceeded his/her IBW, the ratio of BW to IBW was set to 1.

Among the 765 patients who received allo-HSCT, BC was identified in 20 patients (2.6%); the cumulative all-cause mortality rate was 30.0% (95% CI: 14.7–54.9) and 60.0% (95% CI: 40.0–80.7) at 30 and 100 days, respectively, from the onset of BC. Clinical characteristics of the patients are shown in Table 1. Candida glabrata (n = 6), C. parapsilosis (n = 5), C. guilliermondii (n = 4), C. albicans (n = 4), and other Candida species (n = 1) were isolated. The median period from allo-HSCT to the onset of BC was 28 days. Just before the onset of BC, 10 patients showed a lower BDG value than the cutoff (20 pg/ml), while the median value was 20.6 pg/ml. The median duration from the measurement of BDG to the onset of BC was 3.5 days (range 0–30).

Table 1 Causative Candida species and clinical characteristics of patients with breakthrough candidemia
Full size table

Antifungal agents administered at the onset of BC included itraconazole (n = 8), micafungin (n = 7), voriconazole (n = 3), liposomal amphotericin B (n = 1), and voriconazole with liposomal amphotericin B (n = 1); the median administration period was 46 days (range 7–718). Regarding susceptibilities of Candida isolates to antifungal agents used at the onset of BC, three out of seven evaluable isolates were categorized as resistant, according to the Clinical and Laboratory Standards Institute (CLSI) M60 interpretive breakpoints [8]. After the onset of BC, antifungal agents were changed in all patients, apart from two who died within 4 days of BC detection. Combination therapy was administered in seven cases. Patients who survived longer than 30 days after the onset of BC (n = 14) received antifungal agents for a period of at least 14 days, with a median of 23 days (Supplementary Table 1).

To identify the risk factors of BC, matched-pair analyses with the historical control group who did not develop BC were performed. There were no significant differences in age (P = 0.88), sex (P = 0.25), or diagnosis (P = 0.12) between the patients with and those without BC. As a control group, 60 out of 745 patients without BC (non-BC group) were randomly selected at a 1:3 ratio, and matched for age, sex, and diagnosis (Supplementary Table 2). The NRI of the BC group was significantly lower than that of the non-BC group (median 88.9, range 73.7–104.0 in the BC group vs. median 96.9, range 73.0–113.2 in the non-BC group; P = 0.001). In addition, among the 20 patients with BC and 60 patients without BC, the development of BC, which was treated as a time-dependent covariate, had a negative impact on overall survival (hazard ratio [HR] 5.17, 95% confidence interval [CI] 2.75–9.73; P < 0.001; Mantel–Byar test).

For more detailed analyses, the cutoff values of continuous variables were determined as the value that maximized the sum of sensitivity and specificity on a receiver-operating characteristic curve for the onset of BC. The most discriminative cutoff values of serum albumin, BMI, and NRI were 3.6, 18.6, and 92.4, with the corresponding area under the curve values of 0.74 (95% CI: 0.61–0.86), 0.56 (95% CI: 0.40–0.71), and 0.75 (95% CI: 0.62–0.87), respectively.

Variables with a P-value of <0.05 were included in the multivariate analysis, except for BMI and albumin, which were incorporated in the NRI. Multivariate analysis using the Cox proportional hazards regression model was performed, and low NRI values (HR 9.84, 95% CI: 2.77–34.98; P < 0.001) and cord blood transplantation (CBT) (HR 4.86, 95% CI: 1.41–16.75; P = 0.012) were independent risk factors for the development of BC (Supplementary Table 3). In the low NRI group, the ratio of patients with standard risk and neutrophil recovery before conditioning, as well as those who received myeloablative conditioning and standard prophylaxis with antimicrobial agents at the initiation of conditioning, was significantly lower than that in the high NRI group (Supplementary Table 4).

In this study, low values of NRI and CBT were identified as independent risk factors for the development of BC in allo-HSCT recipients. To our knowledge, this is the first report on the association between NRI and BC in allo-HSCT, although a recent report also showed that CBT was an independent risk factor for BC in allo-HSCT recipients [9]. Faced with the challenge of identifying UBW in allo-HSCT recipients, we alternatively used IBW to calculate the NRI, a formula also applied in older patients, known as geriatric NRI [10]. Although both hypoalbuminemia and low BMI were significant factors in univariate analysis for the development of BC, low NRI had a higher HR than these two factors and was shown to be an independent factor in multivariate analysis. We speculated that low NRI would reflect conditions heavily treated for the underlying hematological disease or infection, and the malnutrition state might cause the decline of immune functions, as suggested in the previous study [11]. Although several previous studies showed that hypoalbuminemia and a low value of BMI had a negative impact on transplant-related mortality [12, 13], few studies have been available on the association between NRI and transplant outcomes. Recently, we reported the prognostic significance of the modified NRI, or geriatric NRI among allo-HSCT recipients with acute myeloid leukemia in our institution [14]. Further investigation in other cohorts is warranted on the application of NRI for allo-HSCT recipients as a measure of malnutrition.

In four out of seven evaluable cases, Candida isolates were categorized as susceptible to antifungal agents used at the onset of BC, according to the CLSI breakpoint. Also, in several previous reports, BC could develop in allo-HSCT recipients, even though the causative Candida species were susceptible to the agents in vitro [15, 16]. The CLSI breakpoints were largely based on the data from non-neutropenic patients, and clinical relevance for allo-HSCT recipients should be validated in future studies [8].

Previous research showed that delay in the initiation of antifungal therapy in patients with candidemia significantly impacted mortality [17]. However, timely diagnosis of candidemia in allo-HSCT recipients is difficult in clinical practice, because the sensitivity of blood cultures is poor, and symptoms such as fever and diarrhea are nonspecific. Furthermore, BDG was insufficient for the prediction of BC. Clarification of the risk factors for BC in allo-HSCT recipients may facilitate early intervention, including changing or adding antifungal agents empirically and may contribute to good outcomes.

Our study has the following limitations. First, this was a retrospective analysis in a single center, and there could be a selection bias. Second, we were unable to identify gene mutations in causative Candida isolates [18]. Last, we might underestimate the incidence of candidemia, because of the poor sensitivity of blood cultures [19].

In conclusion, this study revealed that BC tended to occur in patients with malnutrition prior to allo-HSCT, a state represented by NRI. Further investigation is warranted to validate the results of our study.