Impact of fluid overload and infection on respiratory adverse event development during induction therapy for childhood acute myeloid leukemia

Abstract

Background:

Treatment-related morbidity and mortality occur frequently in childhood acute myeloid leukemia (AML) induction. Yet the contributions of respiratory adverse events (AEs) within this population are poorly understood. Furthermore, the roles of fluid overload (FO) and infection in AML pulmonary complications have been inadequately examined.

Objectives:

To describe the incidence, categories, and grades of respiratory AEs and to assess the associations of FO and infection on respiratory AE development in childhood AML induction.

Methods:

We retrospectively examined the induction courses of a cohort of de novo pediatric AML patients for any NCI CTCAE grade 2–5 respiratory AE, FO, and systemic/pulmonary infection occurrence. Demographic, disease, and treatment-related data were abstracted. Descriptive, univariate, survival, and multivariable analyses were conducted.

Results:

Among 105 eligible subjects from 2009–2016, 49.5% (n = 52) experienced 63 discrete respiratory AEs. FO occurred in 28.6% of subjects (n = 30), with half occurring within 24 hours of hospitalization. Positive FO status < 10 days (aHR 5.5, 95% CI 2.3–12.8), ≥ 10 days (aHR 13, 95% CI 4.1–41.8), and positive infection status ≥ 10 days into treatment (aHR 14.9, 5.4–41.6) were each independently associated with AE development.

Conclusions:

We describe a higher incidence of respiratory AEs during childhood AML induction than previously illustrated. FO occurs frequently and early in this course. Late infections and FO at any timeframe were strongly associated with AE development. Interventions focused on prevention and management of FO and infectious respiratory complications could be instrumental in reducing preventable treatment-related morbidity and mortality.

INTRODUCTION

Supportive care enhancements, including mandatory hospitalization during neutropenic periods, empiric broad spectrum antibiotics with fever, preemptive fungal coverage with prolonged fever, and restrained corticosteroid use, account for a share of the current progress in childhood AML outcomes^1–3. Nevertheless, treatment-related complications still explain a substantial proportion of morbidity and mortality within this population^4–10, particularly during induction chemotherapy, with adverse event (AE) incidence frequently underestimated on clinical trials^11–13. A substantial proportion of mortality in childhood AML occurs during the initial induction period with frequency of induction death estimated at 4–11%^4–10,14. The incidence of respiratory AEs in childhood AML and their contribution to treatment-related morbidity and mortality is heretofore poorly understood.

In adult AML, hyperleukocytosis, FAB M4 and M5 morphology, and poor performance status at diagnosis have been identified as risk factors for respiratory AE development during the induction phase¹⁵. Leukemia-specific lung involvement (pulmonary leukostasis, pulmonary leukemic infiltration, acute lysis pneumopathy^15–22) is often associated with respiratory AE development early in this induction course. In childhood AML, hyperleukocytosis has been associated with severe pulmonary leukostasis¹⁶, hypoxia¹⁷, pulmonary hemorrhage¹⁷, and acute lysis pneumopathy¹⁸.

Iatrogenic fluid overload (FO), underlying cardiogenic pathologies²³, and chemotherapy toxicity^24,25 have been associated with pulmonary edema as a cause of morbidity during AML induction. The association between FO and respiratory AE incidence in childhood acute leukemia has been scarcely explored to date.

Pulmonary infections account for over a third of all infections observed during childhood AML induction²⁶. Up to 4% of children experience life-threatening or fatal pulmonary or upper respiratory infections during AML induction, a large proportion of which can be attributed to invasive fungal infection (IFI)²⁷. Even in the absence of a pulmonary infectious source, sepsis frequently predisposes to respiratory AE development through interstitial inflammation and a cytokine-induced increase in alveolar capillary membrane permeability^28,29.

We hypothesized that the incidence and severity of respiratory AEs in childhood AML induction are under recognized and that both FO and systemic or pulmonary infection are potentially modifiable independent risk factors for respiratory AE development in this population. As such, our aims were: 1) To describe the incidence, categories, and grades of respiratory AEs during the course of induction therapy in childhood AML and 2) To assess the associations of FO and systemic and pulmonary infection on respiratory AE development during induction AML therapy.

METHODS

Study design and population

Patients ≤ 21 years of age with newly diagnosed de novo AML who underwent conventional induction chemotherapy at Children’s Healthcare of Atlanta (CHOA) and identified through the cancer registry between March 2009 and December 2016 were eligible for this study. In addition to classical AML cases (AML with recurrent genetic abnormalities, AML not otherwise specified), Down syndrome, and therapy-related myeloid neoplasm cases were also included. Patients with acute promyelocytic leukemia (APL), relapsed AML, preceding myelodysplastic syndrome, receipt of any aspect of induction chemotherapy treatment at another institution, or receipt of an unconventional therapeutic approach were excluded. Subjects were followed from initial hospital day until completion of day 42 of induction chemotherapy, initiation of the subsequent chemotherapy cycle if prior to day 42, or death from any cause. Study data were collected from electronic health records (EHR) and managed using REDCap® electronic data capture tools hosted at CHOA. The study was approved by the Emory University IRB.

Study variables

The primary outcome of interest was the occurrence of any grade 2–5 respiratory AE at any point during the follow up period as defined by the NCI CTCAE v4.0. Multiple discrete respiratory AEs could be attributed to a single subject. Occurrence of a subsequent AE required complete resolution of the prior AE in terms of symptomatology, oxygen/ventilation requirements, and abnormal imaging findings. Subjects were removed from the risk set from the day of respiratory AE onset until its resolution. Respiratory AEs were ascertained by examining each subject’s induction course for development of: 1) an oxygen requirement, 2) a positive pressure requirement (high-flow nasal cannula, CPAP, BiPAP, mechanical ventilation), 3) β−2 agonist requirement for bronchospasm, 4) apnea, 5) dyspnea without oxygenation or ventilatory needs, or 6) imaging occurrence of pleural effusion without oxygen or ventilatory needs. Each AE was then characterized in terms of start date, end date, required interventions (e.g. maximum respiratory support, thoracentesis, diuresis), and chest radiology reports. Category of AE was based on the most specific and severe CTCAE grade. For example, if a subject experienced grade 4 hypoxia, grade 4 dyspnea, and grade 4 pulmonary edema, the selected category and grade would be grade 4 pulmonary edema; if a subject experienced grade 4 hypoxia and grade 4 pulmonary edema and progressed to develop grade 4 acute respiratory distress syndrome (ARDS), the selected category and grade would be grade 4 ARDS. Time-to-event was measured relative to the first day of hospitalization. Characterization and grading of AEs was performed by a single reviewer (LM).

Baseline laboratory values were collected on the date of hospital admission. FO and infection state were captured as time dependent variables. Subjects entered FO state either on the date of first furosemide administration if they received ≥ 3 doses of furosemide within a 72 hour period, or on the date of attaining ≥ 10% weight gain from the lowest previously documented weight. Subjects exited FO state if ≥ 48 hours passed between furosemide doses or weight returned below the 10% threshold. Subjects entered positive infection state on the date of first fever in the context of a positive blood culture, positive respiratory viral panel, chest radiographic findings for pneumonia, chest computed tomography findings for IFI, or clinical sepsis or, if afebrile, at the time of microbiologic or imaging indication of infection. Microorganisms from positive blood cultures and respiratory viral panels were determined through microbiology reports. Standard guidelines were used to define IFI³⁰. Clinical sepsis was determined on the date of culture-negative, clinically-suspected sepsis documentation within a patient progress note or on the problem list. The infection status window closed when blood culture or chest imaging normalized and/or associated fever resolved for ≥ 24 hours.

Statistical methods

Descriptive statistics were calculated to characterize the cohort. Subjects were entered into the respiratory AE group based on the occurrence of ≥ 1 respiratory AE at any time during the follow-up period. Respiratory AE cases were classified into early AE and late AE cases based on occurrence < 10 days or ≥ 10 days from hospitalization. Comparisons were made using two-sample T tests, χ² tests, Fischer’s exact tests, and Mann-Whitney U tests where appropriate. The log-rank test was used to estimate respiratory AE-free survival, fluid overload-free survival, and infection-free survival, censoring for death from other causes. A multivariable Cox proportional hazard model for recurrent events was used to estimate the instantaneous hazard ratios between the independent variables and respiratory AE development, incorporating significant independent variables on bivariate analysis (p < 0.05) and clinically relevant variables. A Heaviside function³¹ within the model allowed for estimation of hazard ratios in reference to AEs occurring both < 10 days and ≥ 10 days. Subsequent AEs within each subject were not considered independent of each other in multivariable analysis. For all tests described, a p value < 0.05 was considered statistically significant. All computations were performed using SAS System v9.4 (2012, SAS Institute, Cary, NC, USA).

RESULTS

Patient characteristics and clinical presentation

Between March 25, 2009 and December 31, 2016, 129 pediatric patients with de novo pediatric AML were identified through the institutional cancer registry at CHOA and 105 met study inclusion criteria (Fig. 1). The median age at diagnosis was 7 years (range, 0–20 years), the female-to-male ratio was 3:2, and 52% were white. Cytogenetic and molecular abnormalities occurred at similar frequencies when compared to the pediatric AML population overall. Eleven patients (10.5%) had acute megakaryoblastic leukemia (AMKL) with trisomy 21 and 8 (7.6%) had therapy-related myeloid neoplasm (Table 1). The median time from hospital admission until chemotherapy initiation was 2 days (range 0–7 days).

Figure 1.

CONSORT flow diagram demonstrating the selection of eligible subjects.

Table 1.

Baseline patient characteristics for patients with de novo AML (n = 105).

Characteristic	N	%
Sex
Female	63	60
Male	44	40
Age at diagnosis, (median, range), y	7	(0–20)
Race/ethnicity
White, non-Hispanic	54	51.9
Black, non-Hispanic	36	34.6
Hispanic	7	6.7
Asian or PI	5	4.8
Native American	2	1.9
BMI category
≥ 95th percentile	15	14.4
5–94.99th percentile	84	80.8
< 5th percentile	5	4.8
FAB morphology
M0/M1	3	2.9
M2	10	9.5
M4	11	10.5
M5	32	30.5
M6	1	1.0
M7	22	21.0
tAML	8	7.6
Not classified	18	17.1
Cytogenetic classification
Normal	39	37.5
Trisomy 21	8	7.7
t(8;21)	12	11.5
inv(16)	8	7.7
Monosomy 7	10	9.6
Monosomy 5/del(5q)	1	1.0
Complex cytogenetics	5	4.8
KMT2A rearrangement	21	20.2
WBC, (median, range), × 109/L	22.1	1.0–444.0
Peripheral blasts, (median, range), %	41	0–94
Hgb, (median, range), g/dL	8.1	2.5–13.7
Platelets, (median, range), × 109/L	50	3–314
Creatinine, (median, range), mg/dL	0.5	0.2–21.2
Uric acid, (median, range), mg/dL	3.9	1.8–31.3
PT, (median, range), sec	15.5	12.9–28.5
D-dimer, (median, range), ng/dL	1086	110–10000

Five deaths occurred during the induction course (4.8% induction death at days 2, 5, 19, 25, and 73 from the initiation of chemotherapy). Two deaths were attributed to ARDS and one to hypoxia not otherwise specified, each in the context of IFI with pulmonary involvement. A fourth death occurred in the context of grade 5 apnea secondary to left frontal intraparenchymal hemorrhage and uncal herniation in a coagulopathic patient. In total, 52 subjects (49.5%) experienced 63 discrete grade 2–5 respiratory AEs, with 90.5% attributed to events categorized as grade 3–5 (Table 2). Fig. 2A displays the estimation of respiratory AE-free survival for the first event, with 36.2% (n = 38) experiencing a respiratory AE by hospital day 10. Among the grade ≥ 3 AEs, 42.9% were hypoxia (n = 27), 15.9% were pulmonary edema (n = 10), 9.5% were pleural effusion (n = 6), and 1.6% were ARDS (n = 1). Nine subjects (8.6%) experienced multiple respiratory AEs, 2 of whom experienced ≥ 3 AEs. Among the 2^nd and 3^rd events (n = 11), 1 was grade 5 (grade 3 hypoxia on day 2 and died of grade 5 ARDS on day 28 in the context of IFI and FO), 9 were grade 3 or 4, and 1 was grade 2. Among all respiratory AEs, 33.3% required high-flow nasal cannula (n = 3), BiPAP/CPAP (n = 3), or mechanical ventilation (n = 15) as the highest level of respiratory support. Of the 63 respiratory AEs, the median time to development was 4 days from presentation (range 1–43) with a large peak noted at day 1 and a smaller peak on day 17.

Table 2.

Respiratory adverse events by NCI CTCAE v4.0 category and grade occurring over the course of the induction period.

	Respiratory AEs (n = 63)
Category, N (%)	Grade 2 n = 6	Grade 3–4 n = 53	Grade 5 n = 4	Any Grade ≥ 2 n = 63
Hypoxia	0 (0)	25 (47.2)	1 (25)	26 (41.3)
Pulmonary edema	1 (16.7)	9 (17.0)	0 (0)	9 (14.3)
Pleural effusion	1 (16.7)	6 (11.3)	0 (0)	6 (9.5)
Dyspnea	3 (50)	4 (7.5)	0 (0)	4 (6.3)
Apnea	0 (0)	2 (3.8)	1 (25)	3 (4.8)
ARDS	0 (0)	1 (1.9)	2 (50)	3 (4.8)
Pulmonary hemorrhage	0 (0)	1 (1.9)	0 (0)	1 (1.6)
Bronchospasm or laryngospasm	1 (16.7)	4 (7.5)	0 (0)	4 (6.3)

Figure 2.

Kaplan-Meier survival curves according to outcome, including survival free from (A) respiratory AE development and (B) fluid overload (grey) and infection (red).

In univariate analysis, development of a respiratory AE was more likely in children with age ≥ 10 years at diagnosis (p = 0.0005), male gender (p = 0.014), increased WBC at diagnosis (p = 0.002), elevated uric acid (p < 0.0001), and elevated PT (p < 0.0001). Patients with FO (p < 0.0001), infection (p = 0.002), or TLS (p = 0.004) were also more likely to develop respiratory AEs (Table 3). The difference between the respiratory AE and no respiratory AE groups in regards to FAB morphology (p = 0.002) was driven by a lower rate of respiratory AE occurrence among subjects with Down syndrome. Early respiratory AE occurrence was more likely in those with elevated uric acid (p = 0.026), PT (p = 0.0009), and D-dimer (0.034) at diagnosis. Infection occurred in 36.8% of early AEs and 60% of late AEs (p = 0.071) (Supplemental Table 1).

Table 3.

Baseline characteristics and co-morbid conditions by experience of a respiratory adverse event (AEs).

	Respiratory AE, n = 52		No respiratory AE, n = 53
Characteristic	N	%	N	%	P
Sex					0.014*
Female	25	48.1	38	71.7
Male	27	51.9	15	28.3
Age category					0.0005*
< 10 years	20	38.5	38	71.7
≥ 10 years	32	61.5	15	28.3
Race/ethnicity					0.0275*
Black, non-Hispanic	23	45.1	13	24.5
All other races and ethnicities	28	54.9	40	75.5
BMI category					0.290
≥ 95th percentile	10	19.2	5	9.6
5–94.99th percentile	39	75.0	45	86.5
< 5th percentile	3	5.8	2	3.9
FAB morphology					0.002*
M0, M1, M2	8	15.4	5	9.4
M4, M4eo, M5	26	50	17	32.1
M7	5	9.6	17	32.1
tAML	3	5.8	5	9.4
Not classified	10	19.2	8	15.1
WBC category					0.002*
< 100 × 109/L	34	65.4	48	90.6
≥ 100 × 109/L	18	34.6	5	9.4
Hgb, (mean, SEM), g/dL	7.7	0.35	8.2	0.34	0.295
Platelet category					0.074
< 30 × 109/L	19	36.5	11	20.8
≥ 30 × 109/L	33	63.5	42	79.3
Creatinine, (mean, SEM) mg/dL)	0.99	0.40	0.44	0.03	0.172
Uric acid, (median, IQR), mg/dL	4.55	3.7–7.1	3.45	2.75–4.50	< 0.0001*
PT, (median, IQR),sec	16.75	15.5–17.7	14.9	14.3–15.7	< 0.0001*
Comorbidities
Fluid overload	29	55.8	1	1.9	< 0.0001*
Infection	28	53.9	13	24.5	0.002*
Tumor lysis syndrome	14	26.9	3	5.7	0.004*

^*p < 0.05

FO status occurred in 28.6% of subjects (n = 30) with a median duration of 8 days (range 1–34 days). Among these subjects, 50% developed FO within the first 24 hours of hospitalization (Fig. 2B). FO was more likely in patients aged ≥ 10 years (p = 0.001), those with infection at any time (p = 0.005), with WBC ≥ 100 × 10⁹/L (p = 0.0001), with low Hgb (0.0008) and elevated uric acid (p = 0.0001), PT (p < 0.0001), and D-dimer (p < 0.0001) at diagnosis (Supplemental Table 2). Infection occurred in 39% of subjects (n = 41), with 53 discrete infections observed among those individuals. Of those who experienced infection, 39% (n = 16) did so within 4 days of initial presentation while an additional 48.8% (n = 20) developed their first infection between days 15–30 (Fig. 2B). Bacteremia and culture-negative sepsis accounted for 42.9% of infections (n = 24) and radiographic pneumonia, disseminated fungal infection, and PCR-positive viral respiratory infections accounted for 57.1% of infections (n = 32). Organisms responsible for bacteremia cases included Viridans streptococci (n = 7), Escherichia coli (n = 3), Methicillin-resistant Staphylococcus aureus (n = 1), Staphylococcus epidermidis (n = 1), Enterobacter cloacae (n = 1), Rothia mucilaginosa (n =1), Bacillus cereus (n = 1), and Actinomyces odontolyticus (n = 1). Fungemia cases included Candida parapsilosis (n = 1) and Candida tropicalis (n = 1). Among all respiratory AEs, 16 (25.4%) occurred while in the FO state and without concurrent infection. Among these, 25% were grade 3 or 4 hypoxia not otherwise specified (n = 4), 25% were grade 3 or 4 pulmonary edema (n = 4), and 37.5% were grade 2–4 pleural effusion (n = 6).

In a multivariable Cox model for recurrent events adjusted for age, sex, race/ethnicity, WBC and platelet count at diagnosis, and infection status, FO status was significantly associated with increased hazard for respiratory AE development before day 10 (HR 5.5, 95% CI 2.3–12.8) and from day 10 onward (HR 13.0, 95% CI 4.1–41.8). Infection status was significantly associated with increased hazard for respiratory AE development from day 10 onward (HR 14.9, 95% CI 5.4–41.6) (Table 4).

Table 4.

Hazard^% for recurrent events of Grade 2–5 respiratory adverse event.

	n = 105
Characteristic	aHR	95% CI	P
Sex			0.001*
Female	0.47	0.26–0.84
Male	1	-
Race/ethnicity			0.087
Black, non-Hispanic	1	-
All other races and ethnicities	0.69	0.40–1.20
Age at diagnosis			0.314
< 10 yr	0.76	0.41–1.39
≥ 10 yr	1	-
WBC at diagnosis			< 0.001*
< 100 × 109/L	0.39	0.21–0.71
≥ 100 × 109/L	1	-
Platelet count at diagnosis			0.160
< 30 × 109/L	1	-
≥ 30 × 109/L	1.40	0.78–2.51
Fluid overload			< 0.0001*^ < 0.0001*^
No	1	-
< 10 days	5.46	2.33–12.76
≥ 10 days	13.0	4.05–41.81
Infection			0.292& < 0.0001*&
No	1	-
< 10 days	1.45	0.66–3.20
≥ 10 days	14.9	5.37–41.58

^*p < 0.05

^%Multivariable cox regression model for recurrent events incorporating a Heaviside function for early versus late respiratory AEs.

^{^}Fluid overload < 10 days vs fluid overload at ≥ 10 days, p = 0.1366

^&Infection < 10 days vs. infection ≥ 10 days, p < 0.0001

DISCUSSION

This analysis provides the most comprehensive examination to date on the incidence of respiratory AEs within the pediatric AML population during induction chemotherapy. Among our cohort, 49.5% of children experienced at least one grade 2–5 respiratory AE during AML induction, indicating that respiratory AEs are more frequently observed in the first phase of de novo pediatric AML treatment than had been previously recognized^16,18,27. Of the 63 respiratory AEs categorized, the majority were severe. The extensive analysis of the COG AAML03P1 and AAML0531 clinical trials for de novo pediatric AML indicated 6% of pediatric patients developed grade 3–5 hypoxia during induction¹⁷. As a direct comparison, 24.8% of the subjects in our cohort (n = 26) developed grade 3–5 hypoxia during this same time course, a discordance likely related to under recognition of respiratory AEs in prior multi-institutional cohort analyses. Recent evaluation of the COG AAML0531 clinical trial data demonstrated reported AEs to be much lower than those ascertained through medical chart abstraction¹⁸ and AEs have been shown to be underreported on additional recent analyses^12,13,32.

Respiratory AEs occurred in a bimodal fashion with a large peak on day 1 from hospital admission and a smaller peak at day 20. This pattern suggests potentially discrete etiologies of earlier compared with later events. In the AML-BFM 87, 93, and 98 clinical trials, mortality up until day 14 of induction chemotherapy occurred largely in conjunction with leukostasis and bleeding while deaths occurring after day 14 were predominantly associated with infection⁴. Our multivariable analysis accounting for the bimodal distribution of AEs notes FO carried strong independent associations with respiratory AE occurrence in both early (HR 5.5) and later (HR 13.0) AEs. Alternatively, infection only demonstrated an independently significant association with respiratory AE development ≥ 10 days from hospital admission (HR 14.9).

We detected FO in nearly one third of our subjects (n = 30), with fluid overloaded subjects averaging 10.3 days within the FO state. Among all subjects, 14.3% (n = 15) developed FO within 24 hours of hospitalization, often prior to initiation of systemic chemotherapy. Most notably, patients with FO were more likely, regardless of the time frame and controlling for infection status, to develop a respiratory AE during the induction phase. In other pediatric conditions such as sepsis^{33, 34}, acute renal failure³⁵, sickle cell disease^36–38, and asthma,³⁹ FO status is known to be associated with respiratory AE development, deterioration, and poor outcomes⁴⁰. FO is an iatrogenic and modifiable risk factor, yet the construct of FO is not included in the NCI CTCAE versions 4.0 or 5.0. Furthermore, FO is not consistently described in the pediatric literature, with definitions typically relating to daily fluid balance^41–43, percent weight change from baseline⁴¹, and physical exam findings. At initial presentation, pediatric acute leukemia patients have historically received hyperhydration in an effort to prevent hyperuricemia-induced renal injury and leukostasis complications⁴⁴, a practice that may place patients at risk for iatrogenic FO, pulmonary edema, and preventable respiratory AEs.

Several diagnostic and management strategies for prevention, early diagnosis, and treatment of FO have been previously studied and may be relevant to our population. Lung ultrasound is an inexpensive, low risk, highly accessible method shown to be effective in quantifying extravascular lung volume in patients at risk for pulmonary edema^45,46. Conservative, net neutral, fluid balance strategies have been successful in a number of pediatric patient populations, including stem cell transplant recipients³⁵, ARDS patients^34,43, and post-resuscitation sepsis patients^34,43. In a meta-analysis of 11 randomized trials incorporating conservative versus standard/liberal fluid strategies in the post-resuscitative care of adults and children with ARDS, sepsis, or systemic inflammatory response syndrome, a conservative strategy was associated with increased ventilator-free days and decreased ICU stay³⁴. The Fluid and Catheter Treatment Trial (FACTT), a multicenter randomized controlled trial that compared use of liberal versus conservative fluid management in patients with ARDS, showed a conservative fluid balance significantly improved survival and ventilator-free days⁴⁷.

The most recent pediatric TLS prevention and management guidelines recommend “ample” hydration and prophylactic rasburicase in patients at high risk for TLS development (WBC ≥ 100 × 10³ cells/μL), hydration and either allopurinol or rasburicase use in intermediate risk patients (WBC ≥ 25 × 10³ cells/μL and < 100 × 10³ cells/μL or with any WBC if LDH is ≥ 2 times the upper limit of normal), and “aggressive” hydration with rasburicase and diuresis for overt TLS and hyperuricemia⁴⁴. Of our cohort, 52.4% of subjects (n = 55) were considered intermediate or high risk for TLS based on initial WBC, yet 26.7% of subjects that experienced FO (n = 8) presented with a WBC < 25 × 10³ cells/μL, suggesting that many subjects at low risk for TLS and leukostasis received hyperhydration. These findings underscore the ambiguity of definitions of “ample” and “aggressive” hydration and that patients at high risk for TLS may not require copious fluids, particularly with the accessibility of rasburicase⁴⁸. Given the availability and efficacy of rasburicase in reduction of hyperuricemia, it may be reasonable to reassess widely instituted practices on fluid management in pediatric patients at risk for TLS, acknowledging that hyperuricemia reduction does not address the additional risks of hyperleukocytosis, leukostasis, and pulmonary leukemic infiltration.

Among our cohort, patients with positive systemic or pulmonary infection status were 14.9 times more likely, respectively and independent of known confounders, to develop a respiratory AE from day 10 AML induction onward, consistent with the risks of infection and capillary leak during prolonged severe myelosuppression^26,27,49,50. Pulmonary infections represented 36.5% of all infections during AML induction in the BFM-2004 trial²⁶. We noted a similar infection incidence, with pneumonia (n = 15), PCR-positive viral respiratory infections (n = 8), and IFI (n = 7) accounting for 28.3%, 15.1%, and 13.2% of infections in our cohort respectively. While customary supportive care now includes antibiotic and antifungal prophylaxis and treatment strategies^3,50, adjunct approaches to pulmonary hygiene (respiratory toilet, ambulation) have not been standardized.

This study’s findings should be interpreted with reference to the limitations of a retrospective single institution experience. While our data is more detailed than what is likely reported in large cooperative group settings, daily fluid balance is challenging to ascertain from EMR retrospectively and the literature lacks a clear defintion. Hence, our definition of FO did not account for the total volume, rate, or type (colloid, crystalloid) of fluids being provided. Definitions of infection in the pediatric oncology literature typically focus on microbiologically-defined infections, clinically-defined infections, and fever without a source requiring antibiotic management. For the purposes of this study, we focused on a select group of microbiologically and clinically-defined infections while omitting incidents of fever without a source and non-pulmonary infections.

However, our data supports a need for prospective management of FO and systemic or pulmonary infection as risk factors for respiratory AE development during the initial pediatric AML course. Cooperative and institutional protocols should consider providing explicit fluid management guidelines directed to recognition of FO, maintenance of net-neutral fluid balance, and standardization of grading and reporting of FO status. We suggest evaluation of potential interventions beyond antibacterial and antifungal therapies toward minimizing morbidity of FO, including scheduled incentive spirometry, positive expiratory pressure, and early frequent ambulation in all pediatric AML inpatients⁵¹. Randomization of childhood AML patients to the aforementioned supportive care interventions could be instrumental in determining their utility for future children undergoing AML treatment in an effort to reduce preventable treatment-related morbidity and mortality.

Glossary

AE

adverse event

AML

acute myeloid leukemia

AMKL

acute megakaryoblastic leukemia

APL

acute promyelocytic leukemia

ARDS

acute respiratory distress syndrome

BiPAP

bilevel positive airway pressure

BMI

body mass index

CHOA

Children’s Healthcare of Atlanta

COG

Children’s Oncology Group

CPAP

continuous positive airway pressure

EFS

event-free survival

EHR

electronic health records

FAB

French-American-British

FACTT

Fluid and Catheter Treatment Trial

FO

fluid overload

IFI

invasive fungal infection

NCI CTCAE

National Cancer Institute Common Terminology Criteria for Adverse Events

NOS

not otherwise specified

OS

overall survival

PT

prothrombin time

TLS

tumor lysis syndrome

WBC

white blood cells