Impact of Time-to-Antibiotic Delivery in Pediatric Patients With Cancer Presenting With Febrile Neutropenia

Abstract

PURPOSE

Febrile neutropenia (FN) in pediatric patients with cancer can cause severe infections, and prompt antibiotics are warranted. Extrapolated from other populations, a time-to-antibiotic (TTA) metric of <60 minutes after medical center presentation was established, with compliance data factoring into pediatric oncology program national rankings.

METHODS

All FN episodes occurring at Vanderbilt Children's Hospital (2007-February 2022) and a sample of episodes from Colorado Children's Hospital (2012-2019) were abstracted, capturing TTA and clinical outcomes including major complications (intensive care unit [ICU] admission, vasopressors, intubation, or infection-related mortality). Odds ratios (ORs) were adjusted for age, treatment center, absolute neutrophil count, hypotension presence, stem-cell transplant status, and central line type.

RESULTS

A total of 2,349 episodes were identified from Vanderbilt (1,920) and Colorado (429). Only 0.6% (n = 14) episodes required immediate ICU management, with a median TTA of 28 minutes (IQR, 20-37). For the remaining patients, the median TTA was 56 minutes (IQR, 37-90), and 54.3% received antibiotics in <60 minutes. There were no significant associations between TTA (<60 or ≥60 minutes) and major complications (adjusted OR, 0.99 [95% CI, 0.62 to 1.59]; P = .98), and a TTA ≥60 minutes was not associated with any type of complication. Similarly, TTA, when evaluated as a continuous variable, was not associated with a major (OR, 0.99 [95% CI, 0.94 to 1.04]; P = .69) nor any other complication in adjusted analysis.

CONCLUSION

There is no clear evidence that a reduced TTA improves clinical outcomes in pediatric oncology FN and thus it should not be used as a primary quality measure.

INTRODUCTION

Febrile neutropenia (FN) during pediatric cancer therapy can be a life-threatening emergency^1-3 and is also a leading cause of hospital admission.^2,4 Patients are instructed to seek immediate medical evaluation with fever onset, and standard practice mandates empiric intravenous (IV) antibiotics in those who are severely neutropenic (absolute neutrophil count [ANC] <500 cells/µL¹). Time to antibiotic (TTA) refers to the time from patient arrival at a medical center until the time when empiric parenteral antibiotics are administered.^5-7 TTA in adult cancer patient guidelines recommend this to be within 60 minutes of presentation.^8-10 Although pediatric FN guidelines do not set an optimal TTA, many pediatric hospitals across the United States track TTA as a primary quality measure, extrapolating the adult oncology-derived benchmark.^6,7,11,12 Additionally, US News and World Report collects data on TTA as a best practice measure in pediatric oncology, using the 60-minute metric data when generating hospital rankings.¹³ To meet this metric, a plethora of quality improvement projects have been implemented, using considerable medical resources aimed at reducing TTA.^6,11,14-17

CONTEXT

• Key Objective

• A time to antibiotic (TTA) <60 minutes after medical center presentation is widely targeted by medical centers and directly affects program national rankings, but its relation with clinical outcomes is not well established.

• Knowledge Generated

• Examining the TTA of 2,349 episodes from two large medical centers revealed that the 60-minute time metric is not associated with developing a major adverse clinical outcome (defined as intensive care unit admission, vasopressors, intubation, or infection-related mortality) or other clinical outcomes.

• Relevance

• There is no clear evidence that a reduced TTA or any specific TTA improves clinical outcomes in pediatric oncology febrile neutropenia and thus it should not be used as a primary quality measure. Ill appearing patients may potentially benefit from immediate antibiotic administration but non-severely ill appearing patients can be safely assessed with lab evaluation and use antimicrobial stewardship in treatment decisions.

TTA has its roots in the treatment of adult patients with bacterial meningitis and sepsis, where TTA appeared to reduce mortality and adverse outcomes.^18,19 Those guidelines have subsequently been applied to the pediatric oncology population and are predicated on the underlying assumption that a faster TTA positively affects clinical outcomes. This premise, however, is not supported by strong evidence. The evidence dictating the optimal TTA in adult and pediatric patients with FN is largely based on a few studies with small sample sizes and inconsistent findings associating a shorter TTA with decreased mortality or other adverse outcomes, frequently finding no associations.^{5,7,11,17,20-32} Additionally, a significant confounding factor in the assessment of TTA is triage bias, wherein patients who are severely ill appearing at presentation to a medical center will consistently receive immediate medical attention, leading to a faster TTA.⁵

Although the intent of TTA metric of <60 minutes in pediatric patients with FN is to improve care, the inverse may paradoxically result when clinicians, who are under some pressure to meet this high-priority institutional standard, place orders for broad-spectrum antibiotics. Examples of unintended outcomes when driven to expedite TTA include inappropriate antibiotic selection for severely neutropenic patients, unnecessary administration of painful intramuscular antibiotic injections, and that non-neutropenic patients may receive empiric broad-spectrum IV antibiotics that are not indicated as ANC results may not be available within the allotted 60-minute window. Prompt availability of antibiotics for patients with FN remains crucial as it is undeniable that ill-appearing or unstable patients warrant rapid antibiotic delivery; however, clinically stable patients may benefit from full review of available clinical data before the choice of antibiotics. Universal administration of broad-spectrum antibiotics to all febrile pediatric oncology patients can select for multidrug–resistant gram-negative bacteria over time, contributing to the growing problem of antimicrobial resistance and additionally can lead to increased Clostridioides difficile infections.³³

Given these challenges created by a TTA metric of <60 minutes in pediatric oncology patients, we evaluated if a faster TTA, either treated as a continuous variable or at a 60-minute cut point is associated with favorable clinical outcomes in pediatric patients with FN.

METHODS

Cohort

After obtaining Institutional Review Board (IRB) approval, all pediatric patients diagnosed with any childhood cancer treated at the Monroe Carell Jr Children's Hospital at Vanderbilt (VCH) from January 2007 to February 2022 were identified, and FN episodes that occurred were abstracted from the electronic health record (EHR). At Children's Hospital Colorado (CHCO), this study used the Colorado Sepsis Treatment and Recognition Registry (CSTRR), a database which contains retrospectively collected EHR data from pediatric emergency department (ED) encounters with ED clinician concern for possible sepsis and a subset of pediatric oncology patients who underwent evaluation for fever at the CHCO ED from May 2012 to February 2019.^34,35 Although this limited data set only included encounters in which a serum lactate was assessed within 2 hours of ED arrival, this was standard institutional practice to obtain in all EDs at CHCO during the study period regardless of clinical appearance. This study was approved by Colorado Multiple IRB (21-2600). Inclusion criteria were applied, and deidentified clinical data were compiled and securely transferred to the main study site for analysis.

Fevers were defined in accordance with the Infectious Diseases Society of America as any temperature reported or recorded ≥38.0°C for >1 hour or ≥38.3°C for any duration or any temperature ≥38.0°C that was acted on by the providing physician.³⁶ Neutropenia was defined as an ANC of <500 cells/µL. Excluded from our study were FN episodes that occurred <7 days from a prior FN episode, patients who already received any IV antibiotics <24 hours before FN fever onset, patients without a central venous catheter (CVC), and patients for whom no accurate TTA could be discerned from the EHR. In addition, two episodes where the TTA was over 24 hours were excluded (TTAs of 1,461 and 3,525 minutes) as it was felt that it would not be comparable with other episodes; in these episodes, neither resulted in a major complication nor had a bacterial infection.

FN episodes at VCH were abstracted from four different health care settings: the VCH ED, clinic, inpatient, and presentation to an outside hospital. At CHCO, only ED encounters were included.

Outcomes and Main Exposure

The primary outcome assessed was developing a major complication defined as infection-related mortality and/or the need for intensive care unit (ICU) care, vasopressor support, or endotracheal intubation within the following timeframes: ≤6 hours, ≤24 hours, or ≤7 days from febrile presentation. The secondary outcomes included hospital length of stay, all-cause mortality, need for bolus IV fluid, amount of bolus fluids received within 24 hours of presentation, and supplemental oxygen administration within 7 days (only captured in VCH cohort). TTA was defined as the time from arrival to the medical center to the time of IV antibiotic administration for patients presenting with fever to the ED or the outpatient setting. For patients who developed a fever during admission or after medical center arrival, TTA was defined as time of fever onset to the time of IV antibiotic administration. The TTA variable was evaluated both nominally (<60 minutes and ≥60 minutes) and continuously. To evaluate if a later TTA cutoff would show additional associations, the analysis was repeated using the nominal cutoff of <90 minutes versus ≥90 minutes.

Statistical Analysis

A detailed statistical analysis plan and power calculations are presented in the Data Supplement (Table S1, online only). In brief, patients' characteristics were summarized by and compared between TTA groups using the Wilcoxon rank sum test for continuous variables and the Chi square/Fisher exact test when comparing nominal variables. To evaluate the associations between TTA and study end points, both univariate and multivariable adjusted (applies to endpoints with effected sample size >10) logistic regressions and linear regressions were used as appropriate. Odds ratios (ORs), mean differences and their associated 95% CIs, and P values estimated from these models were reported. In the multivariable models, a priori selected covariates on the basis of clinical knowledge including patients' age, study site, ANC, hypotension status, post–stem-cell transplant status, and CVC type were considered. In addition, to control for potential triage bias or differences in clinical practice between study sites, patients clinically unstable at presentation (needing immediate ICU level care) were analyzed separately and excluded from the main analysis, and three specific preplanned subgroup analyses defined by study site and ultimate blood stream infection (BSI) status were carried out. Finally, to further control the potential heterogeneity of the episodes analyzed, we conducted ad hoc subgroup analyses among those (1) with initial hypotension but not requiring immediate ICU care, (2) inpatient, (3) outpatient, (4) presented before 2015, (5) presented in 2015 or later, and (6) the first episode from each patient. For each outcome, the episodes with missing outcome values were excluded from the corresponding analysis, and the numbers of events were reported. No missing values in covariates were observed. For the primary outcome, any major complication within 7 days or infection-related mortality, a post hoc power analysis carried out using the Fisher exact test to estimate the minimal differences in event proportions and ORs that the study could detect indicated that our study has 80% power to detect an OR of 1.66 (or higher) or 0.51 (or lower). Bias analyses were also conducted for the primary outcomes using the E-values assessment implemented in EValue package in R software.³⁷

All reported P values were raw P values, and no multiple testing adjustments were applied. Two-sided P values <.05 were considered statistically significant. All analyses were conducted using R software version 4.1.³⁸

RESULTS

Patient Population

A total of 2,349 episodes were identified. In 14 episodes (0.6%), the patient was characterized as immediately decompensated and required ICU management on arrival, leaving a remaining cohort of 2,335 episodes, with 1,912 from VCH and 423 from CHCO. Baseline characteristics for the 14 episodes requiring immediate ICU management were like the overall cohort other than being older (median 14.3 years v 8.0 years, P < .01) and more commonly having ALL (64.3% v 37.0%, P = .04; Table 1). Baseline characteristics were also similar between VCH and CHCO cohorts except for differences in malignancy type, CVC type, presence of hypotension, and presenting ANC (Table 1). Baseline characteristics for those not immediately decompensated were similar between those who received antibiotics in <60 minutes versus ≥60 minutes, except for some differences in cancer diagnosis and CVC type (Table 2). TTA of <60 minutes was more common in episodes occurring after 2015 (P < .01) and in the CHCO cohort (P < .01).

TABLE 1.

Patients' Demographics and Clinical/Laboratory Characteristics by Study Site and Clinical Condition at Presentation (overall cohort, N = 2,349)

Characteristic	Primary Study Cohort: Clinically Stable at Presentation				Presented Immediately Clinically Unstable (n = 14)	P b
	Vanderbilt (n = 1,912)	Colorado (n = 423)	Combined (N = 2,335)	P a
Patient age, years, median (IQR)	8.0 (4.0-13.0)	7.0 (3.5-12.4)	8.0 (4.0-13.0)	.10	14.3 (10.5-16.0)	<.01
Sex, No. (%)				.81		.79
Male	1,023 (53.5)	229 (54.1)	1,252 (53.6)		8 (57.1)
Female	889 (46.5)	194 (45.9)	1,083 (46.4)		6 (42.9)
Cancer diagnosis,c No. (%)
ALL	681 (35.6)	184 (43.5)	865 (37.0)	<.01	9 (64.3)	.04
AML	128 (6.7)	1 (0.2)	129 (5.5)	<.01	2 (14.3)	.15
Lymphoma/other leukemia	225 (11.8)	34 (8.0)	259 (11.1)	.03	0 (0.0)	.19
Solid tumor	736 (38.5)	152 (35.9)	888 (38.0)	.33	2 (14.3)	.07
CNS tumor	142 (7.4)	52 (12.3)	194 (8.3)	<.01	1 (7.1)	.87
Post–stem-cell transplant, No. (%)	85 (4.4)	16 (3.8)	101 (4.3)	.54	2 (14.3)	.07
Days post-transplant,d median (IQR)	71.0 (8.0-452.0)	32.0 (5.8-145.0)	65.0 (7.0-350.0)	.10	138.5 (127.2-149.8)	.62
Central line type, No. (%)				<.01		.76
Port	1,312 (68.6)	377 (89.1)	1,689 (72.3)		11 (78.6)
Peripherally inserted central catheter	62 (3.2)	6 (1.4)	68 (2.9)		0 (0.0)
External tunneled central catheter	538 (28.2)	40 (9.5)	578 (24.8)		3 (21.4)
Absolute neutrophil count, cells/µL, median (IQR)	20.0 (0.0-110.0)	30.0 (10.0-120.0)	20.0 (0.0-110.0)	<.01	5.0 (0.0-70.0)	.30
Hypotension,e No. (%)	105 (5.5)	10 (2.4)	115 (4.9)	<.01	14 (100.0)	<.01
Time to antibiotics, minutes, median (IQR)	60.0 (39.0-101.0)	48.0 (35.0-59.0)	56.0 (37.0-90.0)	<.01	28.0 (20.0-37.0)	<.01
Time to antibiotics <60 minutes, No. (%)	949 (49.6)	318 (75.2)	1,267 (54.3)	<.01	11 (84.6)	.03
Year episode occurred, No. (%)				<.01		.99
2007-2011	627 (32.8)	0 (0.0)	627 (26.8)		4 (28.6)
2012-2015	517 (27.0)	149 (35.2)	666 (28.6)		4 (28.6)
2016-2022	768 (40.2)	274 (64.8)	1,042 (44.6)		6 (42.9)
Location at presentation, No. (%)
Emergency department	1,023 (53.5)	423 (100.0)	1,446 (61.9)	<.01	12 (85.7)	.17
VCH clinic	481 (25.2)	0 (0.0)	481 (20.6)		1 (7.1)
Outside hospital	82 (4.3)	0 (0.0)	82 (3.5)		1 (7.1)
VCH inpatient	326 (17.1)	0 (0.0)	326 (14.0)		0 (0.0)
Bacterial bloodstream infection isolated, No. (%)	274 (14.3)	51 (12.1)	325 (13.9)	.22	8 (57.1)	<.01
Primary outcome: Composite complication,f No. (%)	53 (2.8)	43 (10.2)	96 (4.1)	<.01	14 (100.0)	<.01

Abbreviations: ICU, intensive care unit; VCH, Monroe Carell Jr. Children's Hospital at Vanderbilt.

^a Comparing between study sites. Continuous variables with P value from the Wilcoxon Rank Sum test. Categorical variables with P value from the Pearson χ² test.

^b Comparing the primary study cohort (N = 2,335) with those who presented immediately clinically unstable (n = 14). Continuous variables with P value from the Wilcoxon Rank Sum test. Categorical variables with P value from the Pearson χ² test.

^c A patient may have multiple cancer diagnoses. Each diagnosis was tested individually.

^d Among those who had stem-cell transplant only.

^e Before or within 1 hour of presentation to medical center.

^f Composite complications were defined as having any of the following complications: (1) ICU admission, (2) need for vasopressors to support blood pressure, (3) need for intubation, and (4) having infection-related mortality.

TABLE 2.

Patients' Demographics and Clinical/Laboratory Characteristics by TTA Groups (primary cohort, N = 2,335)

Characteristic	TTA ≥60 Minutes (n = 1,068)	TTA <60 Minutes (n = 1,267)	Total (N = 2,335)	P
Patient age, years	7.7 (4.0-13.0)	8.0 (4.0-13.0)	8.0 (4.0-13.0)	.64
Sex				.17
Male	556 (52.1)	696 (54.9)	1,252 (53.6)
Female	512 (47.9)	571 (45.1)	1,083 (46.4)
Cancer diagnosisa
ALL	400 (37.5)	465 (36.7)	865 (37.0)	.71
AML	76 (7.1)	53 (4.2)	129 (5.5)	<.01
Lymphoma/other leukemia	123 (11.5)	136 (10.7)	259 (11.1)	.55
Solid tumor	365 (34.1)	523 (41.3)	888 (38.0)	<.01
CNS tumor	104 (9.7)	90 (7.1)	194 (8.3)	.02
Post–stem-cell transplant	50 (4.7)	51 (4.0)	101 (4.3)	.44
Days post-transplantb	34.0 (6.2-184.8)	99.0 (8.0-449.5)	65.0 (7.0-350.0)	.17
Central line type				<.01
Port	726 (68.0)	963 (76.0)	1,689 (72.3)
Peripherally inserted central catheter	32 (3.0)	36 (2.8)	68 (2.9)
External tunneled central catheter	310 (29.0)	268 (21.2)	578 (24.8)
Absolute neutrophil count, cells/µL	20.0 (0.0, 120.0)	20.0 (0.0, 100.0)	20.0 (0.0, 110.0)	.72
Hypotensionc	57 (5.3)	58 (4.6)	115 (4.9)	.40
Time to antibiotics, minutes	96.0 (74.0-135.0)	39.0 (29.0-48.0)	56.0 (37.0-90.0)	<.01
Year episode occurred				<.01
2007-2011	351 (32.8)	276 (21.8)	627 (26.8)
2012-2015	419 (39.2)	247 (19.5)	666 (28.6)
2016-2022	298 (28.0)	744 (58.7)	1,042 (44.6)
Bacterial bloodstream infection isolated	152 (14.2)	173 (13.7)	325 (13.9)	.69
Study site				<.01
Vanderbilt	963 (90.2)	949 (74.9)	1,912 (81.9)
Colorado	105 (9.8)	318 (25.1)	423 (18.1)
Location at presentation				<.01
Emergency department	604 (56.6)	842 (66.5)	1,446 (61.9)
VCH clinic	193 (18.1)	288 (22.7)	481 (20.6)
Outside hospital	60 (5.6)	22 (1.7)	82 (3.5)
VCH inpatient	211 (19.8)	115 (9.1)	326 (14.0)

NOTE. Continuous variables were presented as median (IQR) with P value from Wilcoxon Rank Sum test. Categorical variables were presented as frequency (percent) with P value from the Pearson χ² test.

Abbreviations: TTA, time to antibiotic; VCH, Monroe Carell Jr. Children's Hospital at Vanderbilt.

^a A patient may have multiple cancer diagnoses. Each diagnosis was tested individually.

^b Among those who had stem-cell transplant only.

^c Before or within 1 hour of presentation to medical center.

The median TTA for the cohort was 56 minutes (IQR, 37-90), and 54.3% (n = 1,267) received antibiotics in <60 minutes. The TTA in the group of patients requiring immediate ICU care was faster with a median of 28 minutes (IQR, 20-37), with 84.6% receiving antibiotics in <60 minutes and the longest delay being 67 minutes. Mortality for this group was high (57.1%, n = 8), and 75% of patients who died received antibiotics in <60 minutes. The two patients who died with TTA ≥60 minutes included one patient who was determined on immediate arrival to already be too decompensated to benefit from antibiotics before their death and another patient whose TTA was exactly 60 minutes.

Association Between TTA <60 Minutes Versus ≥60 Minutes and Complications

Receiving antibiotics in <60 minutes more commonly led to developing a major complication within 6 hours, 24 hours, or 7 days after presentation (4.7%, 3.9%, 2.4%) versus ≥60 minutes (3.5%, 2.5%, 1.6%); however, this was not statistically significant (Table 3). In univariate analysis, factors significantly associated with developing a major complication included older age, lower ANC, presenting with hypotension, and the CHCO site (Table 4). In multivariable-adjusted analysis, there was no association with a TTA ≥60 minutes and developing a major complication within 6 hours, 24 hours, or 7 days (OR 0.99 [95% CI, 0.62 to 1.59], OR, 0.93 [95% CI, 0.54 to 1.6], OR, 0.96 [95% CI, 0.49 to 1.89]; P values .98, .78, .91). There were similarly no associations between a TTA ≥60 minutes and any individual major complication at any time point (ICU admission, pressor support, need for intubation, or infection-related mortality) nor was a TTA ≥60 minutes associated with all-cause mortality, length of hospital stay, or need for oxygen within 7 days (VCH cohort only). In a site-specific subanalysis, no additional associations were found other than in the VCH cohort, a TTA <60 minutes was associated with vasopressor support in <24 hours in adjusted analysis (P = .04; Data Supplement, Table S2). Additionally, in the nonadjusted CHCO analysis, a TTA ≥60 minutes was associated with the need for intubation within 24 hours or 7 days, P = .02 (Data Supplement, Table S3). A subset analysis of only those who ultimately developed a BSI (n = 325) revealed no significant association between a TTA ≥60 minutes and any complication (Data Supplement, Table S4). Similar findings in both baseline characteristics and their association with complications were observed from all ad hoc subgroup analyses (Data Supplement, Tables S5-S16). In addition, using 90 minutes as the nominal cutoff did not yield any new associations (Data Supplement, Table S17).

TABLE 3.

Summary Statistics of Fever and Neutropenia Complications by Time to Antibiotics Groups and Their Associations Among Those Who Were Not Immediately Decompensated at Presentation (primary cohort, N = 2,335)

Type of Complication	Complication Rate			Unadjusted Associationsa		Adjusted Associationsb
	TTA ≥60 (n = 1,068)	TTA <60 (n = 1,267)	P	Effectsc (95% CI)	P	Effectsc	P
Composite complicationd
Any major complication within 7 days or infection-related mortality (nYes = 96)	37 (3.5)	59 (4.7)	.15	0.73 (0.48 to 1.12)	.15	0.99 (0.62 to 1.59)	.98
Any major complication within 24 hours or infection-related mortality (nYes = 76)	27 (2.5)	49 (3.9)	.07	0.64 (0.4 to 1.04)	.071	0.93 (0.54 to 1.6)	.783
Any major complications within 6 hours or infection-related mortality (nYes = 48)	17 (1.6)	31 (2.4)	.15	0.64 (0.35 to 1.17)	.15	0.96 (0.49 to 1.89)	.905
Individual complication
Any ICU admission
Within 7 days from presentation (nYes = 92)	37 (3.5)	55 (4.3)	.28	0.79 (0.52 to 1.21)	.279	1.1 (0.68 to 1.78)	.691
Within 24 hours from presentation (nYes = 69)	23 (2.2)	46 (3.6)	.04	0.58 (0.35 to 0.97)	.038	0.83 (0.46 to 1.5)	.546
Within 6 hours from presentation (nYes = 40)	12 (1.1)	28 (2.2)	.04	0.5 (0.25 to 0.99)	.048	0.79 (0.36 to 1.72)	.547
Need for vasopressors to support blood pressure
Within 7 days from presentation (nYes = 56)	22 (2.1)	34 (2.7)	.33	0.76 (0.44 to 1.31)	.328	0.93 (0.51 to 1.7)	.814
Within 24 hours from presentation (nYes = 46)	15 (1.4)	31 (2.4)	.07	0.57 (0.3 to 1.06)	.075	0.69 (0.34 to 1.39)	.3
Within 6 hours from presentation (nYes = 5)	1 (0.1)	4 (0.3)	.25	0.3 (0.03 to 2.65)	.276	0.39 (0.03 to 4.45)	.452
Need for intubation
Within 7 days from presentation (nYes = 17)	7 (0.7)	10 (0.8)	.7	0.83 (0.31 to 2.19)	.705	1.02 (0.37 to 2.86)	.964
Within 24 hours from presentation (nYes = 10)	5 (0.5)	5 (0.4)	.79	1.19 (0.34 to 4.11)	.787	1.77 (0.45 to 6.88)	.413
Within 6 hours from presentation (nYes = 7)	3 (0.3)	4 (0.3)	.88	0.89 (0.2 to 3.98)	.878	1.48 (0.29 to 7.54)	.634
Infection-related mortality during febrile neutropenia episode (nYes = 8)	5 (0.5)	3 (0.2)	.34	1.98 (0.47 to 8.31)	.35	1.93 (0.45 to 8.35)	.38
Other complications
All-cause mortality during febrile neutropenia episode (nYes = 16)	9 (0.8)	7 (0.6)	.4	1.53 (0.57 to 4.12)	.401	1.38 (0.5 to 3.8)	.529
Length of hospital stay (days)e
Those already admitted at episode start (nTotal = 317)	13.0 (7.0-21.5)	12.0 (7.5-20.0)	.74	–0.23 (–7.22 to 6.77)	.949	0.02 (–6.99 to 7.02)	.996
Those presenting outpatient (nTotal = 2,002)	4.0 (3.0-7.0)	4.0 (2.8-7.0)	.72	–0.11 (–0.69 to 0.48)	.723	0.09 (–0.5 to 0.67)	.77
Need for oxygen of any type within 7 daysf (nYes = 93)	43 (4.5)	50 (5.3)	.41	0.84 (0.55 to 1.28)	.415	0.8 (0.52 to 1.23)	.317
Need for any bolus fluids within 24 hoursg (nYes = 843)	343 (32.3)	500 (40.4)	<.01	0.7 (0.59 to 0.84)	<.001	0.91 (0.75 to 1.1)	.321
Amount of bolus fluids needed within 24 hours, mL/kg (nTotal = 2,298)h	0.0 (0.0-15.2)	0.0 (0.0-20.0)	<.01	–2.4 (–3.59 to –1.22)	<.001	–0.62 (–1.69 to 0.45)	.257

NOTE. For the complications measured with continuous values, median (IQR) by TTA status, and P value from Wilcoxon Rank Sum test were reported. For the complications measured with yes/no status, frequency (rate) of yes were presented.

Abbreviations: ICU, intensive care unit; OR, odds ratio; TTA, time to antibiotic.

^a Results from univariate logistic regression or linear regression as appropriate (depended on the complications).

^b Results from multivariable logistic regression or linear regression as appropriate (depended on the complications). Patient's age, study site, absolute neutrophil count, hypotension status, post–stem-cell transplant status, and central line type were adjusted.

^c Presented were ORs and 95% CIs of developing a complication for comparing those with TTA ≥60 to those who had TTA <60 minutes. For complications measured with continuous values, reported were mean differences and 95% CI for comparing those with TTA ≥60 to those who had TTA <60 minutes.

^d Composite complications were defined as having any of the following complications: (1) ICU admission, (2) need for vasopressors to support blood pressure, (3) need for intubation, or (4) having infection-related mortality.

^e Sixteen patients did not have hospital stay information recorded.

^f Only available among Vanderbilt patients, n_Total = 1,912.

^g Thirty-seven patients did not have information on if they need bolus fluids within 24 hours, n_Total = 2,298.

^h Thirty-seven patients did not have information on amount of bolus fluids within 24 hours.

TABLE 4.

Patients' Demographics and Clinical/Laboratory Characteristics by Primary Outcome Status (primary cohort, N = 2,335)

Characteristic	Did Not Develop a Complication (n = 2,239)	Developed a Complication (n = 96)	P a
Time to antibiotics, minutes, median (IQR)	56.0 (37.0-91.0)	52.0 (36.0-74.0)	.08
Time to antibiotics <60 minutes, No. (%)			.15
No	1,031 (96.5)	37 (3.5)
Yes	1,208 (95.3)	59 (4.7)
Patient age, years, median (IQR)	7.4 (4.0-13.0)	12.3 (7.7-16.0)	<.01
Sex, No. (%)			.07
Male	1,192 (95.2)	60 (4.8)
Female	1,047 (96.7)	36 (3.3)
Cancer diagnosis
ALL, No. (%)			.76
No	1,411 (96.0)	59 (4.0)
Yes	828 (95.7)	37 (4.3)
AML, No. (%)			.09
No	2,119 (96.1)	87 (3.9)
Yes	120 (92.9)	9 (7.0)
Lymphoma/other leukemia, No. (%)			.44
No	1,993 (96.0)	83 (4.0)
Yes	246 (95.0)	13 (5.0)
Solid tumor, No. (%)			.16
No	1,381 (95.4)	66 (4.6)
Yes	858 (96.6)	30 (3.4)
CNS tumor, No. (%)			.71
No	2,052 (95.8)	89 (4.2)
Yes	187 (96.4)	7 (3.6)
Post–stem-cell transplant, No. (%)			.34
No	2,144 (96.0)	90 (4.0)
Yes	95 (94.1)	6 (5.9)
Days post-transplant,b median (IQR)	65.0 (7.0-394.0)	79.0 (11.5-214.0)	.52
Central line type, No. (%)			.14
Port	1,625 (96.2)	64 (3.8)
Peripherally inserted central catheter	67 (98.5)	1 (1.5)
External tunneled central catheter	547 (94.6)	31 (5.4)
Absolute neutrophil count, median (IQR)	20.0 (0.0-110.0)	10.0 (0.0-30.0)	<.01
Hypotension,c No. (%)			<.01
No	2,155 (97.1)	65 (2.9)
Yes	84 (73.0)	31 (27.0)
Year episode occurred, No. (%)			.06
2007-2011	611 (97.4)	16 (2.6)
2012-2015	633 (95.1)	33 (4.9)
2016-2022	995 (95.5)	47 (4.5)
Bacterial bloodstream infection isolated, No. (%)			<.01
No	1,967 (97.9)	43 (2.1)
Yes	272 (83.7)	53 (16.3)
Study site, No. (%)			<.01
Vanderbilt	1,860 (97.2)	53 (2.8)
Colorado	380 (89.8)	43 (10.2)
Location at presentation, No. (%)			.14
Emergency department	1,382 (95.6)	64 (4.4)
VCH clinic	470 (97.7)	11 (2.3)
Outside hospital	78 (95.1)	4 (4.9)
VCH inpatient	309 (94.8)	17 (5.2)

Abbreviation: VCH, Monroe Carell Jr. Children's Hospital at Vanderbilt.

^a Continuous variables were presented with P value from the Wilcoxon Rank Sum test. Categorical variables were presented with P value from the Pearson χ² test.

^b Among those who had steam-cell transplant only.

^c Before or within 1 hour of presentation to medical center.

Association Between TTA as a Continuous Variable and Complications

The median TTA was faster in those who developed a major complication within 6 hours, 24 hours, and 7 days (51-52 minutes v 56 minutes), and in univariate analysis, a faster TTA was significantly associated with a major complication at all time points (OR, 0.95 [95% CI, 0.9 to 0.99], OR, 0.94 [95% CI, 0.89 to 0.99], OR, 0.91 [95% CI, 0.84 to 0.98]; P values .03, .02, .02). However, these findings did not persist after multivariable adjustment (OR, 0.99 [95% CI, 0.94 to 1.04], OR, 0.99 [95% CI, 0.93 to 1.05], OR, 0.96 [95% CI, 0.88 to 1.05]; P values .69, .80, .40; Table 5). TTA was not associated with individual specific major complications in adjusted analyses. Lower TTA was initially associated with the need for ICU admission within 6 or 24 hours and need for vasopressors within 24 hours or 7 days, but the association did not persist after adjustment. Faster TTA was associated with the need for bolus intravenous fluids within 24 hours, and TTA was inversely correlated with the amount of bolus fluids needed within 24 hours, but neither finding persisted in adjusted analysis. In a site-specific analysis, no additional new associations were found other than in the CHCO cohort, where there was an association between delayed TTA and the need for intubation within 24 hours and 7 days (Data Supplement, Table S18 and S19). In the CHCO cohort only in adjusted analysis, there was also a small association between prolonged TTA (each 10 minutes) and the increased amount of IVF given at 24 hours (mean difference, 0.77 [95% CI, 0 to 1.53], P = .049). A subset analysis of only those who ultimately developed a BSI (n = 325) revealed no significant associations between TTA and any complication (Data Supplement, Table S20). Similar findings were observed from all ad hoc subgroup analyses (Data Supplement, Tables S21-S26).

TABLE 5.

Summary Statistics of TTA by Fever and Neutropenia Complications Status and Its Association With Complications Among Those Who Were Not Immediately Decompensated at Presentation (primary cohort, N = 2,335)

Type of Complication	Summary Statistics of TTA			Unadjusted Associationsa		Adjusted Associationsb
	Did Not Develop a Complication	Developed a Complication	P	Effectsc (95% CI)	P	Effectsc	P
Composite complicationd
Any major complication within 7 days or infection-related mortality (nYes = 96)	56.0 (37.0-91.0)	52.0 (36.0-74.0)	.08	0.95 (0.9 to 0.99)	.025	0.99 (0.94 to 1.04)	.688
Any major complication within 24 hours or infection-related mortality (nYes = 76)	56.0 (37.0-91.0)	51.5 (35.8-71.0)	.06	0.94 (0.89 to 0.99)	.025	0.99 (0.93 to 1.05)	.795
Any major complication within 6 hours or infection-related mortality (nYes = 48)	56.0 (37.0-91.0)	51.0 (36.8-71.0)	.06	0.91 (0.84 to 0.98)	.019	0.96 (0.88 to 1.05)	.404
Individual complication
Any ICU admission
Within 7 days from presentation (nYes = 92)	56.0 (37.0-91.0)	52.0 (36.0-74.8)	.12	0.95 (0.91 to 1)	.038	1 (0.95 to 1.05)	.88
Within 24 hours from presentation (nYes = 69)	56.0 (37.0-91.0)	51.0 (35.0-68.0)	.04	0.93 (0.88 to 0.99)	.026	0.99 (0.93 to 1.06)	.807
Within 6 hours from presentation (nYes = 40)	56.0 (37.0-91.0)	44.5 (34.5-66.5)	.01	0.87 (0.79 to 0.97)	.009	0.93 (0.82 to 1.05)	.245
Need for vasopressors to support blood pressure
Within 7 days from presentation (nYes = 56)	56.0 (37.0-90.0)	49.0 (30.0-74.8)	.05	0.93 (0.87 to 1)	.04	0.96 (0.9 to 1.04)	.332
Within 24 hours from presentation (nYes = 46)	56.0 (37.0-90.0)	42.0 (30.2-67.5)	.02	0.9 (0.83 to 0.98)	.019	0.94 (0.85 to 1.03)	.196
Within 6 hours from presentation (nYes = 5)	56.0 (37.0-90.0)	30.0 (30.0-46.0)	.10	0.76 (0.52 to 1.13)	.176	0.74 (0.42 to 1.3)	.296
Need for intubation
Within 7 days from presentation (nYes = 17)	56.0 (37.0-90.0)	53.0 (39.0-65.0)	.64	0.93 (0.83 to 1.05)	.27	0.97 (0.85 to 1.1)	.609
Within 24 hours from presentation (nYes = 10)	56.0 (37.0-90.0)	60.0 (42.5-67.2)	.95	0.96 (0.84 to 1.1)	.582	1.02 (0.9 to 1.16)	.785
Within 6 hours from presentation (nYes = 7)	56.0 (37.0-90.0)	56.0 (45.5-66.0)	.81	0.93 (0.76 to 1.13)	.439	0.99 (0.83 to 1.2)	.955
Infection-related mortality during FN episode (nYes = 8)	56.0 (37.0-90.0)	71.0 (52.8-90.5)	.40	1 (0.89 to 1.12)	.999	1.01 (0.9 to 1.13)	.912
Other complications
All-cause mortality during FN episode (nYes = 16)	56.0 (37.0-90.0)	64.0 (49.2-80.2)	.46	1.01 (0.94 to 1.08)	.852	1.01 (0.93 to 1.09)	.884
Length of hospital stay (days)e
Those already admitted at episode start (nTotal = 317)	Spearman ρ = –0.01		.81	–0.1 (–0.46 to 0.25)	.572	–0.05 (–0.41 to 0.31)	.799
Those presenting outpatient (nTotal = 2,002)	Spearman ρ = 0.01		.82	–0.03 (–0.08 to 0.02)	.258	0 (–0.06 to 0.05)	.916
Need for oxygen of any type within 7 daysf (nYes = 93)	60.0 (39.0-101.0)	57.0 (39.5-91.5)	.53	0.99 (0.95 to 1.02)	.443	0.99 (0.96 to 1.03)	.639
Need for any bolus fluids within 24 hoursg (nYes = 843)	58.0 (39.0-99.0)	52.5 (36.0-82.0)	<.01	0.96 (0.95 to 0.98)	<.001	0.99 (0.97 to 1)	.103
Amount of bolus fluids needed within 24 hours, mL/kg (nTotal = 2,298)h	Spearman ρ = –0.09		<.01	–0.2 (–0.3 to –0.11)	<.001	–0.04 (–0.13 to 0.04)	.331

NOTE. For the complications measured with continuous values, Spearman ρ correlations with TTA and P values were reported. For the complications measured with yes/no status, TTA were presented as median (IQR) by complication status, and P value from Wilcoxon Rank Sum test were reported.

Abbreviations: FN, febrile neutropenia; ICU, intensive care unit; OR, odds ratio; TTA, time to antibiotic.

^a Results from univariate logistic regression or linear regression as appropriate (depended on the complications). For those complications with limited events (<10), no model fitting was attempted.

^b Results from multivariable logistic regression or linear regression as appropriate (depended on the complications). Patient's age, study site, absolute neutrophil count, hypotension status, post–stem-cell transplant status, and central line type were adjusted. For those complications with limited events (<10), no model fitting was attempted.

^c Presented were ORs and 95% CIs of developing a complication for each 10 minutes increase in TTA for complications measured with yes/no status. For complications measured with continuous values, reported were mean differences and 95% CI in complication measurements for each 10-minute increase in TTA.

^d Composite complications were defined as having one of the flowing three complications: (1) ICU admission, (2) need for vasopressors to support blood pressure, and (3) need for intubation or having infection-related mortality.

^e Sixteen patients did not have hospital stay information recorded.

^f Only available among Vanderbilt patients, n_Total = 1,912.

^g Thirty-seven patients did not have information on if they need bolus fluids within 24 hours, n_Total = 2,298.

^h Among those who needed any bolus fluids within 24 hours, nine did not have actual amount recorded.

Bias Analysis

The E-values for the primary outcome of any major complications within 7 days, 24 hours, and 6 hours or infection-related mortality were 1.11, 1.34, and 1.25, respectively. Indicating that to explain away the observed OR of 0.99, 0.93, and 0.96, an unmeasured confounder that was associated with both any major complication or infection-related mortality and TTA by an OR of 1.11-1.34 or above would be required, and any weaker confounding could not alter this observed association.

DISCUSSION

In our large-scale retrospective study assessing 2,349 FN episodes in pediatric patients with cancer at two major academic children's hospitals, a faster TTA, particularly <60-minute time point, was not associated with improved clinical outcomes, including reducing mortality or ICU transfer. These results are in accordance with several recent studies assessing the impact of TTA in smaller cohorts of pediatric patients with FN.^23,26,32

A few FN studies have shown an association between a prolonged TTA and some adverse clinical outcomes but have not been consistent.³⁹ Fletcher et al showed in 1,628 FN episodes a multivariate association between a composite outcome (death, ICU admission or receiving >40 mL/kg IVF) in those receiving antibiotics between 60 and 120 minutes (32.9% of the cohort) versus ≤60 minute (20% of the cohort), P = .048. However, TTA when treated as a continuous variable was not associated with the composite outcome nor was the comparison between a TTA <60 minutes and TTA >121 minutes (47.1% of the cohort).²⁴ In a univariate comparison, Salstrom et al³⁰ (n = 220) showed an association between ICU admission and receiving antibiotics >60 minutes; however, no other significant associations were found. One adult study found significance between mortality and TTA; however, the BSI rate (37.4%) and mortality (9.4%) were much higher than expected, potentially limiting the study's generalizability.²⁹ Finally, an adult study of 3,219 patients showed decreased mortality giving antibiotics <3 hours versus 3-6 hours but not between quicker TTA periods.²¹

Empiric antibiotics are required for severely neutropenic patients with fever because of their risk for severe bacterial infection and the inability to distinguish which patients have a bacterial etiology at fever presentation.^1,2 There exists a small subset of patients with FN who have severe sepsis at presentation for whom quick action is always needed (0.6% of episodes in this cohort). For these patients, antibiotics must be given urgently as this study showed a mortality rate over 57% even when antibiotics were given rapidly.⁴⁰ For nonseverely ill-appearing patients, our data suggest it is safe to thoughtfully evaluate, consider antibiotic stewardship, and wait for ANC results before empiric treatment as opposed to providing care to meet an absolute TTA of <60 minutes. This may facilitate more appropriate antibiotic choices among neutropenic patients and can also allow for the identification of non-neutropenic patients, a population for whom there is growing evidence that may need no empiric antibiotics.⁴¹ As the TTA metric is measured from medical center arrival rather than fever onset, studying the time from fever onset could be more prognostic; however, in this study, analyzing the 326 patients from whom fever onset was precise because of already being inpatient did not yield any associations between TTA and clinical outcomes.

This study has some limitations. We only evaluated those with CVCs in place and could not evaluate those already on IV antibiotics or other rare patient-specific conditions that might affect an individual's risk of complications. We adjusted our analysis to control for major factors that can be associated with BSI infections and analyzed separately those who presented with severe sepsis to address the potential for confounding by indication but could be additional unmeasured confounders; however, bias analysis showed it would have to be significant to alter to findings of this study. Finally, these analyses were conducted in medical centers where patients are (1) counseled and usually can present quickly to attention after fever onset, (2) patients are clinically assessed quickly on arrival and is feasible to administer antibiotics expediently to those ill appearing, and (3) ANC results are usually available within 2 hours to make decisions. Additional study would be needed before generalizing outside of these prerequisites.

In conclusion, our results provide strong, practice-changing evidence that the current TTA metric of <60 minutes used by numerous pediatric hospitals does not improve quality of care in most patients. As such, TTA in its present manifestation should be reevaluated as a best practice metric, and future research toward reducing morbidity and mortality in pediatric oncology patients with FN should focus on other strategies.

AUTHOR CONTRIBUTIONS

Conception and design: George C. De Castro, Zhiguo Zhao, Chelsea C. Clark, Debra L. Friedman, Adam J. Esbenshade

Provision of study materials or patients: Debra L. Friedman, Jenna Demedis

Collection and assembly of data: George C. De Castro, Leonora R. Slatnick, Morgan Shannon, Kasey Jackson, Christine M. Smith, Daniel Whitehurst, Claire Elliott, Chelsea C. Clark, Halden F. Scott, Jenna Demedis, Adam J. Esbenshade

Data analysis and interpretation: George C. De Castro, Leonora R. Slatnick, Morgan Shannon, Zhiguo Zhao, Chelsea C. Clark, Debra L. Friedman, Jenna Demedis, Adam J. Esbenshade

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

Impact of Time-to-Antibiotic Delivery in Pediatric Patients With Cancer Presenting With Febrile Neutropenia

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/op/authors/author-center.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).