Impact of D-index and L-index on pulmonary infection in induction chemotherapy for acute lymphoblastic leukemia and lymphoblastic lymphoma

Abstract

Objectives: The D-index and the L-index, calculated as the area over the neutrophil and lymphocyte curves, respectively, reflect both the intensity and duration of cytopenia. We, retrospectively, investigated the impact of these indexes on pulmonary infection (PI) in induction chemotherapy for acute lymphoblastic leukemia (ALL) and lymphoblastic lymphoma (LBL).

Methods: We included 92 patients (ALL 83, LBL 9) from two institutions. We calculated the D-index and cumulative D-index until the development of PI (c-D-index), which enables real-time risk assessment for infection. We also calculated the L-index (35), defined as the area over the lymphocyte curve during lymphopenia (<700/µl) until day 35 and the cumulative-L-index until the development of PI (c-L-index).

Results: Eight patients developed PI on day 20 (median). Two patients were strongly suspected to have bacterial pneumonia, and the others were suspected to have pulmonary fungal infection. The D-index and the L-index (35) in patients with PI were higher than those in patients without PI (7230 ± 4734 vs. 4519 ± 3416, P = 0.041 and 15 458 ± 5243 vs. 8920 ± 5901, P = 0.018), while the c-D-index and the c-L-index were not significantly different. Although the c-L-index did not have predictive value for PI, c-D-index, when treated as a dichotomous variable with a cutoff value of 5589 as determined by a receiver operating characteristic curve analysis, showed a significant difference between two groups (P = 0.045). This association became clearer when we focused on suspected pulmonary fungal infection.

Discussion and conclusion: In induction chemotherapy for ALL/LBL, c-D-index with a cutoff value of 5589 might have predictive value for the development of PI.

Keywords:

• D-index
• L-index
• Pulmonary infection
• Acute lymphoblastic leukemia
• Lymphoblastic lymphoma

Introduction

The D-index and L-index, calculated as the area over the neutrophil and lymphocyte curves, respectively, reflect both the intensity and the duration of cytopenia. The D-index is based on a graph that shows the serial absolute neutrophil counts during neutropenia (<500/µl) and is calculated as the area surrounded by the neutrophil curve and the horizontal line at 500/µl.¹ The cumulative D-index (c-D-index), defined as the cumulative D-index from the start of neutropenia until the development of infection, can be used for the real-time assessment of the risk for infection. The c-D-index has been shown to have a high negative predictive value for invasive mold infection in acute myeloid leukemia (AML) patients undergoing induction chemotherapy,¹ and for pulmonary infection in hematopoietic stem cell transplantation (HSCT) recipients.² The L-index is defined as the area surrounded by the lymphocyte curve and the horizontal line at an absolute lymphocyte count (ALC) of 700/µl during lymphopenia (ALC<700/µl).³ In allogeneic HSCT recipients, the L-index from the start of conditioning to day 30 (L-index (30)) showed a significant association with cytomegalovirus reactivation.³

Infectious complications are major problem in acute leukemia patients undergoing intensive chemotherapy,^4–6 especially in AML patients undergoing induction chemotherapy due to long-lasting profound neutropenia.⁷ On the other hand, the use of corticosteroid may play an additional role in the development of infectious complications in acute lymphoblastic leukemia (ALL).⁸ However, it is still unclear how these risk factors influence pulmonary infections in ALL patients.

In this study, we evaluated the impact of the D-index and L-index on pulmonary infection in ALL and lymphoblastic lymphoma (LBL) patients undergoing induction chemotherapy. The L-index was considered to be a surrogate marker of the effect of corticosteroid use. The D-index is considered to have the strongest impact on the development of invasive mold infection among pulmonary infections. However, we often encounter a situation that we should consider antifungal therapy for nonspecific pulmonary infection in order to avoid treatment delay, even though their computed tomography (CT) findings do not fulfill the EORTC/MSG criteria. Therefore, we set our primary goal in this study as pulmonary infection of any kinds, being closer to our daily practice.

Material and methods

Patients and treatment procedure

Patients with ALL and LBL who underwent induction chemotherapy at Saitama Medical Center between January 1997 and June 2013, and at NTT Medical Center between January 2001 and September 2011, were consecutively enrolled in this retrospective study. This study was approved by the Institutional Review Board of Saitama Medical Center, Jichi Medical University. Ninety-two patients (83 ALL and 9 LBL patients) were included in this study.

Patients underwent intensive combination chemotherapies as induction remission. The protocols of prospective studies by the Japan Adult Leukemia Study Group (JALSG) were mainly used: JALSG ALL97 in 9 patients,⁹ JALSG ALL202 in 27 (ALL202U 7, ALL202O 20),¹⁰ JALSG PhALL202 in 3,¹¹ and JALSG PhALL208 in 9 (unpublished). Other regimens included hyper CVAD/MA regimens in 13 patients,¹² and other original institutional regimens in 31. The key drugs were similar in JALSG protocols; cyclophosphamide, anthracyclines, vinca alkaloids, and corticosteroid. l-asparaginase and imatinib mesylate were added in philadelphia chromosome-negative and -positive cases, respectively.

Granulocyte colony-stimulating factor was administered in most of the chemotherapies. Prophylaxis against bacterial and fungal infections was performed in most of the patients. The former consisted of fluoroquinolone, mostly levofloxacin. The latter consisted of fluconazole (n = 54), itraconazole (n = 27), and other antifungal agents (n = 4). Seven patients did not receive antifungal prophylaxis. Prophylaxis against pneumocystis pneumonia consisted of trimethoprim/sulfamethoxazole (n = 78) and inhalation of pentamidine (n = 3).

D-index, L-index, and pulmonary infection

The D-index was calculated based on a graph that plotted the absolute neutrophil counts over the course of the episode of neutropenia.¹ The D-index was calculated as the difference between the observed area under the curve (AUC), which was calculated by the trapezoidal method, and the expected neutrophil area (500/µl × days with neutropenia). The c-D-index was calculated as the cumulative D-index from the start of neutropenia until the development of infection in patients with pulmonary infection, whereas the c-D-index was equal to the D-index in patients without infection. We also evaluated the durations of neutropenia (<500/µl) and profound neutropenia (<100/µl) in each chemotherapy cycle. We also calculated the L-index (35), which was defined as the area over the lymphocyte curve during lymphopenia (<700/µl) until day 35. In the same manner as in a previous study, lymphopenia was defined as ALC less than 700 /µl.³ The L-index was calculated until day 35 because lymphocyte recovery was not observed around 1 month after the start of chemotherapy in more than half of the cases and the data after day 35 were not available in most cases. The c-L-index was calculated as the cumulative L-index from the start of lymphopenia until the development of pulmonary infection, whereas the c-L-index was equal to the L-index (35) in patients without infection. Pulmonary infection was defined as new pulmonary infiltrate observed by chest X-ray or chest CT regardless of microbiological evidence.

Statistical considerations

We divided the patients into those with and without pulmonary infection. We assessed the impact of the D-index and L-index on pulmonary infection along with other epidemiological and clinical factors in these two groups. Differences between groups were examined using Fisher's exact test for dichotomous variables and the Mann–Whitney U test or Student's t-test for continuous variables. A P-value of <0.05 was considered to be significant. To assess the ability of the D-index and L-index to predict infection, we performed a receiver operating characteristic (ROC) curve analysis and calculated the positive and negative predictive values in this patient population.

All statistical analyses were performed with EZR version 1.26 (Saitama Medical Center, Jichi Medical University, Accessed 1 October 2014, at http://www.jichi.ac.jp/saitama-sct/SaitamaHP.files/statmedEN.html), which is a graphical user interface for R (The R Foundation for Statistical Computing, version 3.1.1).¹³ More precisely, it is a modified version of R commander (version 2.1-2) that was designed to add statistical functions that are frequently used in biostatistics.

Results

Patients

The characteristics of the patients are summarized in Table 1. Among the 92 patients, 8 developed pulmonary infection during or within 1 week after recovery from neutropenia (>500/µl). The median onset was day 20 after the start of chemotherapy (range, 2–27). Among the eight patients with pulmonary infection, three and one were classified as probable and possible invasive pulmonary mold infections, respectively, according to the European Organization for Research and Treatment of Cancer/Invasive Fungal Infectious Cooperative Group and the National Institute of Allergy and Infectious Disease Mycoses Study Group (EORTC/MSG) revised criteria.¹⁴ Although the other two patients showed nonspecific pulmonary infiltrate on chest CT, pulmonary fungal infection was clinically suspected and antifungal treatment with voriconazole was started. The remaining two patients who developed nonspecific pulmonary infiltrate were successfully treated with doripenem. In the four patients with nonspecific pulmonary infiltrate, serum markers such as Aspergillus galactomannan and 1,3-β-d-glucan were negative, and bronchial lavage was unavailable There was no difference in the clinical or epidemiological characteristics between patients with and without pulmonary infection. Two of three patients with probable invasive pulmonary mold infections were receiving itraconazole for prophylaxis and the other was not taking any antifungal agent. The patient with possible invasive pulmonary mold infections is on itraconazole prophylaxis. The two patients with clinically suspected pulmonary fungal infection were receiving itraconazole and fluconazole, respectively, and the remaining two patients with suspected bacterial infections were on fluconazole prophylaxis.

Table 1 Clinical and epidemiological characteristics of the study patients

	Total cases (n = 92)	Patients with PI (n = 8)	Patients without PI (n = 84)	P–value*^Citation1
Age, median (range)	49.5 (15–90)	48 (20–71)	49.5 (15–90)	1.00
Sex
Male	39	3	36	1.00
Female	53	5	48
Type of disease
Acute lymphoblastic leukemia	83	8	75	1.00
Lymphoblastic leukemia	9	0	9
Cytogenetics
Poor risk	30	4	26	0.427
Ph-positive	26	3	23	0.427
WBC count at presentation (/µl), median (range)	11 465 (500–612 200)	18 995 (500–163 500)	11 465 (600–612 200)	0.879
LDH at presentation (IU/l), median (range)	654 (89–9390)	1129 (193–6805)	648 (89–9390)	0.489
Fever at presentation	51	6	45	0.291
Neutrophil count at presentation (/µl), median (range)	1741 (0–21 938)	1689 (0–15 660)	1749 (0–21 938)	0.493
ALC at presentation (/µl), median (range)	1597 (0–214 488)	1660 (255–5457)	1592 (0–214 488)	0.897
Antifungal prophylaxis
None or FLCZ	61	4	57	0.435
Antimold agents	31	4	27
Total dose of corticosteroid (in PSL equivalent, mg)	1800 (100–2900)	2007 (100–2683)	1795 (300–2900)	0.708
Duration of G-CSF use (days)	7 (0–44)	8 (2–27)	7 (0–44)	0.506

*¹Patients with PI were compared to those without PI.

PI, pulmonary infection; Ph, Philadelphia chromosome; WBC, white blood cell; ALC, absolute lymphocyte count; FLCZ, fluconazole; PSL, prednisolone; G-CSF, granulocyte colony-stimulating factor.

D-index and L-index

The results regarding neutropenia indexes and L-index are summarized in Table 2. The mean D-index in this cohort was 4754, which was lower than those in induction for AML and approximately equal to those in consolidation chemotherapy for AML reported in the previous studies.^1,15 Two patients with a low c-D-index (0 and 727) were strongly suspected to have bacterial pneumonia because they were successfully treated with antibiotics. Although the D-index and the L-index (35) in patients with PI were significantly higher than those in patients without PI (P = 0.041 and P = 0.018, respectively), there was no difference in the c-D-index or c-L-index between the two groups. The c-L-index in patients with PI was rather lower than the L-index (35) in patients without PI.

Table 2 Neutropenia indexes and L-index

	Total cases (n = 92)	Patients with PI (n = 8)	Patients without PI (n = 84)	P-value*
Day of neutropenia (<500/µl)	13.7 ± 9.0	17.8 ± 9.44	13.3 ± 8.91	0.187
Day of profound neutropenia (<100/µl)	6.8 ± 6.0	10.6 ± 9.6	6.5 ± 5.5	0.061
D-index	4754 ± 3600	7230 ± 4734	4519 ± 3416	0.041
Cumulative days of neutropenia (<500/µl)	13.2 ± 8.7	12.1 ± 6.6	13.3 ± 6.9	0.707
Cumulative days of profound neutropenia (<100/µl)	6.55 ± 5.42	7.4 ± 4.8	6.5 ± 5.5	0.657
c-D-index	4568 ± 3388	5091 ± 3237	4519 ± 3416	0.65
D-index
≥5589		6	22	0.009
<5589		2	62
c-D-index
≥5589		5	22	0.045
<5589		3	62
L-index (35)	9345 ± 6050	15458 ± 5243	8920 ± 5901	0.018
c-L-index	8723 ± 5828	6945 ± 5115	8920 ± 5901	0.366

*Patients with PI were compared to those without PI.

PI, pulmonary infection.

In the ROC curve analysis, the AUC were 0.705, 0.65, and 0.648 for the D-index, days of neutropenia and days of profound neutropenia, respectively (Fig. 1A). AUC values were 0.6, 0.511, and 0.582 for the c-D-index, cumulative duration of neutropenia, and cumulative duration of profound neutropenia, respectively (Fig. 1B). The ROC curve was closest to the left corner of the plot when the thresholds for the D-index, days of neutropenia, and days of profound neutropenia were 5589, 15, and 6, respectively. With the use of these cutoff values, the sensitivity and specificity for predicting pulmonary infection were 75 and 73.8%, 75 and 61.9%, and 87.5 and 47.6%, respectively (Table 3). The positive and negative predictive values were 21.2 and 96.9%, 15.6 and 96.3%, and 13.6 and 97.6%, respectively (Table 3). Similarly, the ROC curve was closest to the left corner of the plot when the thresholds for the c-D-index, cumulative duration of neutropenia, and cumulative duration of profound neutropenia were 5589, 15, and 6, respectively. With the use of these cutoff values, the sensitivity and specificity for predicting pulmonary infection were 62.5 and 73.8%, 62.5 and 61.9%, and 75 and 47.6%, respectively. The positive and negative predictive values were 18.5 and 95.4%, 13.5 and 94.6%, and 12.0 and 95.2%, respectively. We did not perform an ROC curve analysis for the L-index (35) and c-L-index because the c-L-index did not have predictive value for the development of pulmonary infection.

Figure 1 Receiver operating characteristic curve analyses. Receiver operating characteristic curves comparing the D-index with the days of neutropenia (<500/µl , N500) and profound neutropenia (<100/µl, N100) (A), and comparing the cumulative D-index (c-D-index) with the cumulative durations of neutropenia (<500/µl, N500) and profound neutropenia (<100/µl, N100) (B) as predictors of pulmonary infection.

Table 3 Predictive values of each parameter for pulmonary infection

	CO value	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)
D-index	5589	75.0	73.8	21.4	96.9
Days of neutropenia (<500/µl)	15	75.0	61.9	15.8	96.3
Days of profound neutropenia (<100/µl)	6	87.5	47.6	13.7	97.6
c-D-index	5589	62.5	73.8	18.5	95.4
Cumulative duration of neutropenia	15	62.5	61.9	13.5	94.6
Cumulative duration of profound neutropenia	6	75.0	47.6	12.0	95.2

CO, cutoff; PPV, positive predictive value; NPV, negative predictive value.

When we treated the D-index and the c-D-index as dichotomous variables with a cutoff value of 5589 as determined by ROC analysis, the difference between the two groups became statistically significant (P = 0.009, 0.045, respectively) (Table 2). Furthermore, when we excluded the two patients who were suspected to have bacterial pneumonia, to identify the utility of these indexes for pulmonary fungal infection, the D-index in patients with PI was significantly higher than that in patients without PI (mean 8632 vs. 4519, P = 0.007) (Table 4). The c-D-index tended to be higher in patients with PI with borderline significance (mean 6667 versus 4519, P = 0.132). When we treated the c-D-index as a dichotomous variable with a cutoff value of 5589, the difference between the two groups became statistically significant (P = 0.0086) (Table 4).

Table 4 Analysis of the neutropenia indexes after excluding two patients with bacterial pneumonia

	Patients with PI (n  =  6)	Patients without PI (n  =  84)	P-value
Day of neutropenia (<500/µl)	20.7 ± 9.14	13.3 ± 8.91	0.056
Day of profound neutropenia (<100/µl)	13.0 ± 9.91	6.5 ± 5.5	0.01
D-index	8632 ± 4673	4519 ± 3416	0.007
Cumulative days of neutropenia (<500/µl)	15.5 ± 2.51	13.3 ± 8.9	0.558
Cumulative days of profound neutropenia (<100/µl)	9.33 ± 3.78	6.5 ± 5.5	0.215
c-D-index	6667 ± 1643	4519 ± 3416	0.132
D-index
≥5589	6	22	0.0006
<5589	0	62
c-D-index
≥5589	5	22	0.0086
<5589	1	62

PI, pulmonary infection.

Discussion

In this study, we retrospectively analyzed the impact of neutropenia and lymphopenia on the development of pulmonary infection during induction chemotherapy for ALL and LBL.

Our results showed that both the D-index and the L-index (35) in patients with pulmonary infection were higher than those in patients without PI. This is a larger cohort study than previous studies that showed the relationship between the degree of neutropenia and infection.^1,2,15 However, the implications of these results for each index seem to be different.

There was no significant difference in the c-D-index between the two groups, which is different from the results in the previous studies.^1,2 One of the reasons for this discrepancy might be the difference in the severity of neutropenia. The mean D-index in this cohort was 4754, which was lower than those in induction chemotherapy for AML and in HSCT recipients. The mean D-index of 4754 was approximately equal to those in patients with AML who received consolidation chemotherapy with high-dose cytarabine, where the impact of the c-D-index was less significant.¹⁵ Another possible explanation is the development of bacterial pneumonia. The c-D-index was lower than 5589, which was the cutoff value determined by ROC curve analysis, in three of the eight patients with pulmonary infection. Two of these three patients with an extremely low c-D-index (0 and 727) were strongly suspected to have bacterial pneumonia, since they were successfully treated with antibiotics.

While the c-D-index has some limitations, as described above, the c-D-index, when treated as a dichotomous variable with a cutoff value of 5589 as determined by ROC curve analysis, showed a statistically significant difference between patients with and without pulmonary infection. The negative predictive value of the c-D-index for pulmonary infection was 95.4%. The negative predictive value was highest in the study of patients who were receiving remission induction therapy for AML.¹ A possible explanation is that the end point in that study was limited to proven or probable mold infection, which leaded to the high negative predictive value of the D-index. Another explanation is that neutropenia is the sole risk factor in the chemotherapy for AML, whereas the use of steroid or immunosuppressants may affect the incidence of pulmonary infection in patients undergoing chemotherapy for ALL or undergoing allogeneic stem cell transplantation. However, we think that the c-D-index is useful in terms of its high negative predictive value in our study. We believe that it is unnecessary to use empirical antifungal therapy for patients with febrile neutropenia whose c-D-index are lower than 5589 and who do not have clinical evidence of invasive fungal infection. On the other hand, the c-D-index might be beneficial as a threshold to start antifungal therapy when it exceeds 5589 in the absence of a strong evidence for invasive fungal infection by imaging studies or serological tests, so that the risk of overlooking the invasive fungal infection or delay of antifungal treatment can be reduced in diagnostic-driven antifungal approach. With regard to the difference between c-D-index and simple duration of neutropenia, ROC curve analysis showed that the D-index and the c-D-index are slightly more useful than simple duration of neutropenia. This is also reported in the first JCO paper.¹ We believed that both depth and duration of neutropenia calculated as the D-index are important to evaluate the intensity of neutropenia.

In the previous study in HSCT recipients,² most cases of pneumonia seemed to be of fungal origins.¹⁶ However, in this study, two patients were strongly suspected to have bacterial pneumonia. Therefore, we needed to perform further analysis excluding these patients to identify the utility of these indexes for pulmonary fungal infection. When we performed another analysis after excluding two patients who were suspected to have bacterial pneumonia and focused on suspected pulmonary fungal infection, this association became clearer. For pulmonary fungal infection, more prolonged severe neutropenia than in the case of bacterial pneumonia is known to be a risk factor.¹⁷ Based on these considerations, the c-D-index might be useful for predicting pulmonary fungal infection, as described previously.¹

On the other hand, in contrast to the c-D-index, the mean c-L-index in patients with pulmonary infection was rather lower than the mean L-index (35) in those without PI, while the L-index (35) was substantially higher in patients with pulmonary infection. This meant that lymphopenia became severe after the development of pulmonary infection. We thought that pulmonary infection itself, malnutrition or stress could induce prolonged lymphopenia.¹⁸ Therefore, the c-L-index before the development of pulmonary infection does not have predictive value for PI.

This study has some limitations. The first is the small number of patients evaluated, especially the number of patients with pulmonary infection. The second is the heterogeneity of the treatment protocol, especially between the two institutions and the period of treatments. The impact of the D-index is suspected to vary because of the difference in the kinetics of the neutrophil count depending on the treatment protocol. Third, although the c-D-index, when treated as a dichotomous variable, showed a significant difference, it is possible that this threshold for the c-D-index was arbitrary. However, we believe that this was a valid cutoff value, since it was similar to those in previous reports.

We think that the D-index is useful for in terms of its high negative predictive value. This index may reduce the use of unnecessary antifungal agents as used in the classic empiric antifungal therapy by discriminating very low-risk patients. We are now advancing a clinical trial of c-D-index-guided antifungal therapy with a cutoff value of 5500 in high-risk hematological patients to identify the utility of c-D-index.

In conclusion, our results showed that both the D-index and the L-index (35) were higher in patients with pulmonary infection undergoing induction chemotherapy for ALL/LBL. The c-D-index with a cutoff value of 5589 seemed to have predictive value for pulmonary infection, especially pulmonary fungal infection. However, the c-L-index did not have predictive value, since a high L-index (35) was mainly due to severe lymphopenia after the development of pulmonary infection.

Disclaimer statements

Contributors Y.I., S-I.K. K.U., and Y.K. conducted the study, collected the data, interpreted data and drafted the manuscript. Y.A., N.H., H.N., K.K., T.U., H.W., R.Y., K.K., K.S., M.A., M.S., K T-S., M. K., J.K., and S.K. collected the data and analyzed the data. H.N, R.Y., A.T., and J.N. reviewed the manuscript and gave substantial contribution to the final version. All authors reviewed the manuscript and approved the final version.

Conflicts of interest The authors declare no conflict of interest with regard to the contents of this study.

Ethics approval This study was approved by the Institutional Review Board of Saitama Medical Center, Jichi Medical University.

Acknowledgements

This work was presented in part at the 76th Annual Meeting of the Japanese Society of Hematology, Osaka, Japan, 2014 (Abstract OS-3-47).

Additional information

Funding

None.