D-Index as a Marker of Bloodstream Infections in Patients with Allogeneic Hematopoietic Stem Cell Transplantation

Abstract

Neutropenia is a risk factor for development of infections, however, the direct effect of neutropenia on development of bloodstream infection (BSI) is not known. D-index, which is area between the neutrophil time curve and a neutrophil count of 0.5 × 10⁹/L, incorporates the combined effect of severity and duration of neutropenia. We aimed to evaluate whether D-index can be used as a marker for BSI in patients with allogeneic stem cell transplantation.

We conducted a retrospective cohort study of patients undergoing allogeneic stem cell transplantation between January 1, 2005, and September 30, 2015. The primary outcome measure was the development of BSI within 30 days of transplantation.

A total of 714 patients were included in the study of whom 101 developed BSI. Patients with BSI had a significantly higher median D-index value compared with patients who did not have BSI (4,990 versus 3,570, P<.001). As a marker, the performance of the D-index was similar to that of the duration of profound neutropenia (P=0.18) and significantly better than the total duration of neutropenia (P=0.001).

The D-index performed better than the total duration of neutropenia as a marker for BSI in patients with allogeneic stem cell transplantation. There was no difference between D-index and, a more easily calculable indicator, duration of profound neutropenia.

INTRODUCTION

Infection remains a major cause of morbidity and mortality in patients with hematologic malignancy.¹ Myelosuppression from underlying malignancy, chemotherapy, or hematopoietic stem cell transplantation (HSCT) increases risk of infection for patients with cancer compared with the general population.² Two critical risk factors for the development of infection in this group include neutropenia and disruption of the mucocutaneous barrier either from insertion of a central line or chemotherapy-associated mucositis.^2,3 A disruption of the mucocutaneous barrier can facilitate translocation of bacteria from the skin or gut to the bloodstream, and is a well-recognized risk factor for bloodstream infection (BSI).^3,4 However, the impact of neutropenia alone on the development of BSI is less clear.

Extensive research and experience with managing neutropenic fever has reduced the associated mortality rate to 10% ¹. However, more than 80% of patients with hematologic malignancy still develop neutropenic fever during the course of their treatment ⁵. BSI is a significant contributor to neutropenic fever and is documented in 10% to 25% of neutropenic fever episodes ⁵. In an international multicenter study, BSI was the cause of neutropenic fever in 29% of patients who had hematologic malignancy or underwent HSCT ⁶. Neutropenic fever complicated with BSI is associated with a higher mortality risk than neutropenic fever alone ⁷. Thus, predicting the risk of BSI in this patient population can help identify resources for prevention and encourage close monitoring and timely initiation of therapy.

The severity and duration of neutropenia has been used clinically as a marker to predict the onset of invasive infections in patients with neutropenia. To evaluate the effects of the dynamics of neutropenia on infection, Portugal et al. ⁸ designed a measure, termed the D-index that considers the combined effects of the severity and duration of neutropenia. The D-index represents the area between the neutrophil count-time curve and a neutrophil count of 0.5 × 10⁹/L. These investigators showed that the D-index successfully predicted invasive mold infection (IMI) in patients with acute myeloid leukemia (AML). Subsequently, the D-index has been evaluated as a marker for pulmonary infections, mucositis, and BSI in different cohorts of patients with hematologic malignancies ^9–14. In this study, we evaluated the utility of the D-index as a marker for development of BSI in patients who had undergone allogeneic HSCT.

METHODS

The study was approved by the Mayo Clinic Institutional Review Board (protocols 19–002832 and 15–006880).

Study Design and Outcome Measures

We conducted a retrospective cohort study of patients who had undergone allogeneic HSCT at a single center from January 1, 2005, through September 30, 2015. We used the Mayo Clinic HSCT database and the Advanced Cohort Explorer, a clinical data search tool maintained by Mayo Clinic, to identify patient records with appropriate codes from International Classification of Diseases, ninth and tenth revisions. Patients with a previous allogeneic transplant were excluded from the study.

The primary outcome measure was the development of BSI within 30 days after transplantation. Secondary outcome measures were the development of Clostridioides difficile infection (CDI) at any time after transplantation, development of mucositis within 30 days after transplantation and death within 100 days after transplantation.

Definitions and CDI Detection

Neutropenia was defined as an absolute neutrophil count (ANC) less than 0.5 × 10⁹/L and profound neutropenia as ANC less than 0.1 × 10⁹/L ⁵. The neutrophil engraftment was defined as ANC more than 0.5 × 10⁹/L for three consecutive days. Duration of neutropenia was calculated from ANC less than 0.5 × 10⁹/L till engraftment. Duration of profound neutropenia was calculated from ANC less than 0.1 × 10⁹/L till ANC more than 0.1 × 10⁹/L. Oral mucositis grade (range, 0–3) was determined by using the Bearman et al. ¹⁵ criteria. BSI was defined as the growth of a known bacterial or fungal pathogen (Centers for Disease Control and Prevention−National Healthcare Safety Network list) ¹⁶ from 1 or more blood culture samples. In case of growth of a potential contaminant from skin, the determination of clinical significance (confirmed BSI) was adjudicated by 3 independent reviewers (P.V., J.C.O., A.J.T.).

Before November 2011, Premier Toxin A/B enzyme immunosorbent assay (Meridian Bioscience) was used to detect C. difficile in the stool. In 2011, our institution switched to a laboratory developed real-time polymerase chain reaction (PCR) with fluorescent resonance energy transfer hybridization probes targeting the gene regulating toxin production (tcdC) ¹⁷. This test detects C. difficile toxins A and B but does not differentiate between them. It also cannot differentiate between toxin-hyperproducing C. difficile strains (eg, B1, NAP1) and other strains. Since October 2015, we have also used the FilmArray gastrointestinal panel (BioFire Diagnostics), in addition to PCR, to diagnose CDI cases.

D-Index Calculation

The D-index (Ae – Ao) was calculated by the trapezoidal method as the difference between the expected (Ae) area under the curve (AUC), if the patient did not have neutropenia, and observed (Ao) AUC from the neutrophil-time curve.⁸ The D-index₀ was calculated from day 0 (day of transplantation) until engraftment and was assigned a value of 0 if the patient did not have neutropenia during this period (Figure 1). The relevant calculations were adopted from Abad et al. ¹⁸. We also calculated D-index (as the term D-index was used in previous studies) from the start of neutropenia until engraftment, for patients who were already neutropenic on day 0, to include the burden of neutropenia before transplantation (Supplementary Figure 1).

Figure 1:

D-index is the area over the neutrophil curve and expresses the neutrophil deficit. Adapted from Kimura et al. [11]. This figure is available under the Creative Commons CC-BY-NC-ND license. Copyright © 2010 American Society for Blood and Marrow Transplantation. Published by Elsevier Inc. Abbreviations: Ae, expected area under the curve; Ao, observed area under the curve.

Statistical Methods

The D-index₀, the duration of neutropenia (ANC <500/mcL), and the duration of profound neutropenia (ANC <100/mcL) were calculated from transplantation (day 0) to engraftment. Comparisons of the groups with and without BSI were evaluated with the nonparametric Wilcoxon rank sum test for continuous variables and the Pearson χ² test or Fisher’s exact test for categorical variables, as indicated. We used receiver operating characteristic (ROC) curves to show the ability of each measure of neutropenia to separate subjects with and without BSI. The difference between D-index₀ and duration of neutropenia, in terms of discriminative ability, was determined by using the nonparametric approach of DeLong, DeLong, and Clarke-Pearson for comparing areas under the ROC curve ¹⁹. Additionally, we identified an optimal threshold for the D-index₀ by selecting the value that maximized the sum of sensitivity and specificity; this value was chosen from the range of possible thresholds along the ROC curve. Data analysis was conducted by using SAS statistical software (version 9.4, SAS Institute Inc.). For all analyses, P values <.05 were considered statistically significant.

RESULTS

Patient Characteristics

We identified 4,339 patients who underwent HSCT during the study period; of these, 789 had allogeneic HSCT. A final cohort of 714 patients in whom the D-index₀ was available from Abad et al. ¹⁸ were included in the study (Figure 2). Baseline patient characteristics and comorbid conditions are summarized in Table 1. Forty-two percent of patients were female, 96% were white, and the median age at the time of transplantation was 53 years (range, 18–75 years). AML was the most common reason for HSCT (n=263 [37%]), followed by myelodysplastic syndrome or myeloproliferative disorder (n=156 [22%]), and chronic myeloid leukemia (n=32 [4.5%]). Seventy-nine patients (11%) had a prior autologous HSCT. Patients in our cohort had a low prevalence of comorbid conditions; the most common conditions were diabetes mellitus (n=48 [7%]), chronic renal insufficiency (n=28 [4%]), and chronic obstructive pulmonary disease (n=16 [2%]).

Figure 2:

Screening of the Hematopoietic Stem Cell Transplantation (HSCT) Cohort. Abbreviations: HSCT, hematopoietic stem cell transplantation; IFI, invasive fungal infection.

Table 1:

Baseline Characteristics of the Study Cohort (N=714).

Characteristic	No BSI (n=613)a	BSI (n=101)a	P Value
DEMOGRAPHICS
Sex, male	352 (57.42)	61 (60.40)	.58
Age, median (IQR), y	53 (42–59)	53 (43 –59)	.58
Race/ethnicity			.28
White	589 (96.08)	96 (95.05)
Black	6 (0.98)	3 (2.97)
Asian	7 (1.14)	2 (1.98)
Middle Eastern	4 (0.65)	0 (0)
Other	7 (1.14)	0 (0)
Comorbid conditions
COPD	14 (2.28)	2 (1.98)	.85
DM	38 (6.20)	10 (9.90)	.17
CKD	21 (3.43)	7 (6.93)	.09
CMV serostatus			.43
D+/R−	69 (11.26)	10 (9.90)
D+/R+	175 (28.55)	32 (31.68)
D−/R+	182 (29.69)	29 (28.71)
D−/R−	154 (25.12)	20 (19.80)
Unknown/R−	10 (1.63)	4 (3.96)
Unknown/R+	23 (3.75)	6 (5.94)
CHARACTERISTIC OF TRANSPLANTATION
Primary reason for transplantationb			.33
Myeloid	383 (62.48)	68 (67.33)
Lymphoid	107 (17.46)	16 (15.84)
Other	123 (20.07)	17 (16.83)
Type of graft			.11
Cord blood	28 (4.57)	8 (7.92)
Bone marrow	55 (8.97)	14 (13.86)
Peripheral blood	530 (86.46)	79 (78.22)
T-cell−depleting agent (e.g., alemtuzumab or antithymocyte globulin)	28 (4.57)	7 (6.93)	.31
Transplant donor			.01
Related	311 (50.73)	38 (37.62)
Unrelated	302 (49.27)	63 (62.38)
HLA typing			.67
Matched (10 of 10)	418 (68.19)	71 (70.30)
Mismatch or haploidentical	195 (31.81)	30 (29.70)
TRANSPLANT PROPHYLAXIS
GVHD prophylaxis			.004
Tac + MTX	209 (34.09)	52 (51.49)
CSA + MTX	275 (44.86)	33 (32.67)
CSA + MMF	66 (10.77)	11 (10.89)
CSA alone	12 (1.96)	3 (2.97)
Other	51 (8.32)	2 (1.98)
Antibacterial gram-negative prophylaxis			.69
Cefdinir	20 (3.26)	3 (2.97)
Ciprofloxacin	13 (2.12)	0 (0)
Levofloxacin	562 (91.68)	95 (94.06)
Moxifloxacin	2 (0.33)	0 (0)
Other	14 (2.28)	3 (2.97)
None	2 (0.33)	0 (0)
High-dose corticosteroids (>1 mg/kg per d for 14 d or >2 mg/kg per d for 7 d) within 100 d	80 (13.05)	18 (17.82)	.20
Duration of corticosteroids, median (IQR), d	19 (14–26)	18 (14–33)	.56
LABORATORY DATA
Neutropenia	603 (98.37)	100 (99.01)	.63
Profound neutropenia	579 (94.45)	99 (98.02)	.13
Duration of neutropenia, median (IQR), d
ANC <0.5 x 109/L	13 (11–17)	16 (11.5–19.5)	.002
ANC <0.1 x 109/L	10 (8–13)	13 (9.25–17)	<.001
D-index0, median (IQR), d•neutrophil/mcL	3,540 (2,685–4,670)	4,800 (3,253–6,273.75)	<.001
D-index	3,570 (2,685–4,850)	4,990 (3,258–6,798)	<.001
OUTCOME
Mucositis	286 (46.66)	53 (52.48)	.28
Grade I	48 (16.78)	5 (9.43)
Grade II	189 (66.08)	29 (54.72)
Grade III	48 (16.78)	19 (35.85)
Not described	1 (0.35)	0 (0)
Acute GVHD	313 (51.06)	48 (47.52)	.51
Grade I	13	93
Grade II	17	101
Grade III	12	75
Grade IV	6	44
Chronic GVHD	227 (37.03)	27 (26.73)	0.45
Death within 100 d	48 (7.83)	20 (19.80)	<.001

Abbreviations: ANC, absolute neutrophil count; CKD, chronic kidney disease; CMV, cytomegalovirus; COPD, chronic obstructive pulmonary disease; CSA, cyclosporine; D, donor; DM, diabetes mellitus; GVHD, graft-versus-host disease; HLA, human leukocyte antigen; IQR, interquartile range; MMF, mycophenolate; MTX, methotrexate; R, recipient; Tac, tacrolimus; +, positive; −, negative.

^aData are shown as No. (%) unless otherwise indicated.

^bMyeloid includes acute myeloid leukemia, myelodysplastic syndromes, chronic myelogenous leukemia. Lymphoid includes acute lymphoblastic leukemia, chronic lymphocytic leukemia, Hodgkin lymphoma, non-Hodgkin lymphoma and other lymphomas.

Indices of Neutropenia

Nearly all patients (n=703 [98%]) had neutropenia, and many (n=678 [95%]) developed profound neutropenia. The median duration of neutropenia was 13 days (range, 0–90 days), and the median duration of profound neutropenia was 10 days (range, 0–81 days). For those experiencing profound neutropenia, the duration was <10 days for 353 patients (52%), 10 to 20 days for 285 (42%), and >20 days for 41 (6%). A majority of patients (n=661 [94%]) experienced neutrophil engraftment. The median D-index₀ was 3594 (IQR range, 2690–4878.25). Distribution of D-index₀ and D-index is shown in Supplementary Figures 2 and 3.

Development of BSI (Primary Outcome)

A majority of patients (n=529 [74%]) had no microbial growth from central line or peripheral blood cultures within the first 30 days after transplantation. The remaining 185 patients had at least 1 positive blood culture (Supplementary Table 1), and of these, 101 (54%) were clinically adjudicated as a true BSI. For 87 patients, the same organism was identified from venipuncture culture and central line culture. For 8 cases in which only the central line culture was positive and 6 cases in which only the peripheral culture was positive, the cultures were clinically adjudicated as true BSIs. The median time from HSCT to detection of BSI was 9 days (range, 1–25 days). Consistent with current literature, coagulase-negative Staphylococcus was the most commonly identified pathogen (n=30), followed by Enterococcus sp. (n=18) and Escherichia coli (n=10) (Figure 3).^5,20

Figure 3.

Pathogens Identified from Bloodstream Infections.

When patients were stratified by the presence or absence of BSI, we did not identify any significant differences in the demographic characteristics between the two groups (Table 1). The use of antibacterial gram-negative prophylaxis and duration of corticosteroid use also was similar for both groups. Only 10 patients (2%) in the non-BSI group and 1 (1%) in the BSI group did not have neutropenia during the study period. The simple occurrence of neutropenia (ANC <0.5 × 10⁹/L) or profound neutropenia (ANC <0.1 × 10⁹/L) was not associated with the development of BSI.

We noted some significant differences between the two groups. More patients in the BSI group received HSCT from an unrelated donor compared with the non-BSI group (P=.01). The duration of neutropenia was significantly different between groups; the median time for the non-BSI group was 13 days (interquartile range [IQR], 11–17 days), whereas for the BSI group, it was 16 days (IQR, 12–20 days) (P=.002). Similarly, the median duration of profound neutropenia was significantly different between groups (non-BSI, 10 [IQR, 8–13] days; BSI, 13 [IQR, 9–17] days; P<.001). Patients with BSI had a significantly higher median D-index₀ value (4,800 [IQR, 3253–6273]) compared with patients who did not have BSI (3,540 [IQR, 2,647–4,610]) (P<.001). Subgroup analysis was performed for patients with grade III mucositis which also demonstrated higher median D-index₀ in patients with BSI (5210 [IQR,4750–7175]) when compared to patients without BSI (4285.25[IQR,2982.5–5640]) which reached statistical significance (P=0.014).

As shown in the ROC curves (Figure 4), as a marker, the performance of the D-index₀ was similar to that of the duration of profound neutropenia and slightly better than the duration of neutropenia. The area under the ROC curve was 0.66 for the D-index₀, 0.67 for the duration of profound neutropenia, and 0.60 for the duration of neutropenia (P<0.001 for any difference). ROC analysis also showed that a D-index₀ cutoff of 3,578 provided a sensitivity of 69%, a specificity of 50%, and a negative predictive value of 91%.

Figure 4.

Receiver operating characteristic curves comparing the D-index₀ (red) and D-index (green) with the duration of neutropenia (blue) and profound neutropenia (orange) until the development of a bloodstream infection. Abbreviations: AUC, Area under the curve. *P values represent the comparison between the two indices.

Secondary Outcomes

There were 141 patients who developed CDI, 339 had mucositis and 68 patients died within 100 days of transplantation. The median D-index₀ was significantly higher for patients who died within 100 days after transplantation compared with those who survived (4,542 vs 3,574; P=.004). The duration of profound neutropenia was also significantly longer for patients who died within 100 days, but the duration of neutropenia was not associated with increased mortality rates. The D-index₀, duration of neutropenia, and duration of prolonged neutropenia were higher in patients with mucositis compared to those without mucositis. The D-index₀, duration of neutropenia, and duration of prolonged neutropenia were not significantly different in the subset of cases with CDI (Table 2).

Table 2.

Primary and Secondary Outcomes with Respective Measured Variables.

Outcome	D-Index0		D-Index		Duration of Neutropenia, d		Duration of Profound Neutropenia, d
	Median (IQR)	P Value	Median (IQR)	P Value	Median (IQR)	P Value	Median (IQR)	P Value
BSI		<.001		<.001		.002		<.001
Present	4,800 (3,253–6273)		4,990 (3,258–6,798)		16 (12–20)		13 (9–17)
Absent	3,540 (2,685–4,670)		3,570 (2,685–4,850)		13 (11–17)		10 (8–13)
Mucositis		.01		.006		.01		<.001
Present	3,734 (2,920–5,085)		3,800 (2,925–5,260)		14 (11–17)		11 (9–14)
Absent	3,450 (2,496–4,615)		3,515 (2,595–4,885)		13 (10–17)		10 (8–13)
CDI		.46		.50		.90		.90
Present	3,675 (2,858–5,127)		3,755 (2,863–5,153)		14 (11–17)		10 (8–13)
Absent	3,590 (2,685–4,863)		3,610 (2,688–5,175)		13 (11–17)		10 (8–14)
Death		.004		<.001		.18		<.001
Deceased	4,542 (2,783–6,604)		5,032 (3,079 – 9,003)		15 (11–20)		13 (9–18)
Alive	3,574 (2,689–4,726)		3,588 (2,690–4,943)		13 (11–17)		10 (8–13)

Abbreviations: BSI, bloodstream infection; CDI, Clostridioides difficile infection.

Patients Who Were Neutropenic on the Day of Transplantation

Of the 714 patients in the cohort, 116 (16%) were neutropenic on the day of the transplantation (day 0). The median D-index₀ differed between the patients who were and were not neutropenic on day 0 (5,320 vs 3,460). The median D-index of these 116 cases was 7,980. Similar to the D-index₀, the D-index was significantly higher in patients with BSI, mucositis and in patients who died within 100 days. The difference was not significant for patient with and without CDI. Further, BSIs were more common in patients who were neutropenic (28/116 [24%]) vs non-neutropenic (73/598 [12%]) on day 0 (P=.002).

DISCUSSION

BSI remains the leading cause of infection in patients with HSCT and is associated with a crude mortality rate of 12% to 42% ^21–24. An ideal risk assessment tool will be able to determine a subpopulation of patients at risk for BSI and provide prophylaxis or increased monitoring targeted to them. In this study, we observed that D-index₀ and D-index were significantly higher in patients with BSI compared to patients without BSI (P<.001). Moreover, D-index performed better than duration of neutropenia as a marker for occurrence of BSI (P<0.001).

Kimura et al. ¹⁴ evaluated the ability of the D-index to predict early BSI (1 week after engraftment) and pulmonary infections in patients undergoing HSCT. D-index values were higher in patients with BSI than in patients without BSI (7,102.5 vs 3,963.5), but the difference was not significant (P=.055), possibly because of the small sample size (n=58). The conclusions of their study may differ from ours because their study had a smaller number of patients with BSI (12 patients: 1 in autologous and 11 in allogeneic HSCT), lacked clinical adjudication, and studied a different duration of BSI (1 week after engraftment versus 30 days post-transplant). Their study also had a heterogeneous patient group, with 23 autologous and 35 allogeneic HSCT recipients, and a significantly different median duration of neutropenia between the groups (autologous HSCT, 6.5 days; allogeneic HSCT, 18.5 days; P<.001). We included a homogenous cohort of patients with allogeneic HSCT, with a median duration of neutropenia of 13 days. The median time to development of BSI was similar in both studies (9.5 days [range, 1–24 days] vs 9 days [range, 1–25 days]). They concluded that early BSI after HSCT may be caused by bacterial translocation from damaged mucosa or a central line and is not strongly associated with neutropenia; however, they also noted that a larger sample size may have shown a statistically significant difference.

D-index was initially proposed in 2009 by Portugal et al. ⁸ as a tool to predict IMI. They prospectively evaluated the D-index in 11 patients with AML and IMI and were able to demonstrate that it had a high negative predictive value (>97%), similar to that of galactomannan and b-D-glucan. In their study, a cumulative D-index (defined as the D-index from the start of neutropenia until the IMI manifestation) cutoff of 5,800 was able to detect IMI with fairly good sensitivity (91%) and specificity (58%). The authors recommended using the D-index to assess the pretest probability of IMI and to complement positive results of serum galactomannan and b-D-glucan tests. Since then, the D-index has been used, with mixed results, to predict invasive fungal infections (IFI) in adult or pediatric patients with AML ^10,25, in patients undergoing HSCT or reduced-intensity HSCT ^13,14,26, in patients receiving consolidation chemotherapy for AML with high-dose cytarabine ⁹, and in patients receiving induction therapy for acute lymphoblastic leukemia and lymphoma ¹¹. Of these prior studies, the largest sample size was 68 patients ¹³. Recently, in the largest study to date, our group demonstrated that D-index did not correlate with risk of early invasive fungal infection in the same cohort of 714 patients with HSCT ¹⁸. Nonetheless, a randomized controlled trial demonstrated that D-index-guided early antifungal therapy decreased the use of antifungal agents without increasing IFIs or mortality in 413 patients with malignancy when compared to empiric antifungal therapy.²⁷

Kishimoto et al. ¹² assessed the D-index in 104 chemotherapy recipients to determine its predictability of oral mucositis and odontogenic infections. Although D-index values were not associated with odontogenic infections, the group with oral mucositis had higher D-index values than those without mucositis. Similarly, in our analysis, patients with mucositis had higher D-index₀ and D-index values, and they also had a longer duration of neutropenia and profound neutropenia.

None of the prior D-index studies evaluated how well it could predict CDI or death. The risk factors for the development of CDI after HSCT include graft-versus-host disease and receipt of chemotherapy or broad-spectrum antibiotics ²⁸. Neutropenia itself has not been proven to be a risk factor for CDI, which is consistent with our results showing that the D-index was not predictive of CDI. Although the D-index₀, D-index, and duration of profound neutropenia were significantly higher in patients who died within 100 days after transplantation, the duration of neutropenia was not higher. This difference further highlights the varying effects of severity and duration of neutropenia on patient outcomes.

The study is limited by its retrospective nature. The presence of a central line is a confounding factor because a central line is itself a risk factor for BSI and we did not collect data specific to the central line (e.g. duration of access, type of line). However, we believe that most (if not all) patients who underwent allogeneic HSCT had a central line. Also, in the setting of neutropenic fever, it is hard to differentiate between intra-abdominal (mucositis/neutropenic colitis) or line-related sources. We were unable to determine with certainty whether the source of BSI was the central line. The study was not designed to evaluate the effect of graft-versus-host disease prophylaxis and donor matching on the incidence of infections. Lastly, the predictive value of the D-index may be affected by the type of antimicrobial prophylaxis used.

The strengths of the study include a large sample size and homogeneous population of patients who underwent allogeneic HSCT, which increases its applicability in this particular cohort. All cases of BSI were also clinically adjudicated to ensure that they did not represent line colonization or contamination.

Manual calculation of the D-index requires complex computations; however, with the adoption of electronic medical records at most hospitals, D-index values, including cumulative D-index, could be automatically calculated and readily available for interpretation. With automation, the D-index could be measured over longer periods, capturing the effect of fluctuating neutrophil counts in patients with hematologic malignancy. Prospective monitoring on a daily basis in real-time throughout the disease course would also more accurately reflect underlying immunosuppression and could even be evaluated as a predictor for BSI. D-index data potentially could be analyzed with artificial intelligence to determine the need and type of neutropenic prophylaxis. Theoretically, automatically calculated D-index values would provide a greater benefit than the cumulative duration of neutropenia because it considers the varying severity of neutropenia, which may reflect the state of immunosuppression more accurately. However, the actual benefit will need to be determined prospectively in a study where cumulative D-index can be measured from the onset of first of multiple episodes of neutropenia till the development of BSI. In addition, a multivariate analysis including type of transplant, conditioning regimen, GVHD grade and prophylaxis, mucositis and D-index could shed more light on the direct effect of these factors on the outcomes measured.

In conclusion, the D-index performed better than duration of neutropenia, but was similar to duration of profound neutropenia as a marker for development of bloodstream infections in patients with allogeneic stem cell transplantation. The D-index, which considers the combined effect of severity and duration of neutropenia, can be a clinical tool that helps assess the risk of BSI, mucositis, and 100-day mortality. Further studies are needed to determine whether this inexpensive approach can be incorporated into the daily management of patients with neutropenia with an aim to individualize antibiotic prophylaxis.

ABBREVIATIONS

AML

acute myeloid leukemia

ANC

absolute neutrophil count

AUC

area under the curve

BSI

bloodstream infection

CDI

Clostridioides difficile infection

HSCT

hematopoietic stem cell transplantation

IFI

invasive fungal infection

IMI

invasive mold infection

PCR

polymerase chain reaction

ROC

receiver operating characteristic