Abstract
Background
Pediatric acute myeloid leukemia (AML) chemotherapy increases the risk of life-threatening complications including septic shock (SS). We hypothesized an area-based measure of social determinants of health, the Social Disorganization Index (SDI), would be associated with SS and SS-associated death (SS-death).
Methods
Children treated for de novo AML on two Children’s Oncology Group (COG) trials at institutions contributing to the Pediatric Health Information System (PHIS) database were included. SDI was calculated using residential ZIP code data from the US Census Bureau. SS was identified using PHIS resource utilization codes. SS-death was defined as death within 2 weeks of an antecedent SS event. Patients were followed from 7 days after start of chemotherapy until the first of: end of front-line therapy, death, relapse, or removal from study. Multivariable-adjusted Cox regressions estimated hazard ratios (HR) comparing time to first SS by SDI groups.
Results
The assembled cohort included 700 patients with 207 (29.6%) sustaining at least one SS event. There were 233 (33%) in SDI-5 group (highest disorganization). Adjusted time to incident SS did not statistically significantly differ by SDI [reference = SDI-1; SDI-2 HR 0.84 (95% Cl 0.51, 1.41); SDI-3 HR 0.70 (95% Cl 0.42, 1.16); SDI-4 HR 0.97 (95% CI 0.61, 1.53); and SDI-5 HR 0.72 (95% CI 0.45, 1.14)]. Nine (4.4%) patients with SS experienced SS-death; 7 (78%) were in SDI-4 or 5.
Conclusions
In a large, nationally representative cohort of trial-enrolled pediatric patients with AML, there was no significant association between SDI and time to septic shock.
Keywords: Pediatric AML, Septic Shock, Social Disorganization Index, SDOH, COG
Precis:
Time to first septic shock event was not associated with the social disorganization index (SDI), an area-based measure of social determinant of health, in children with acute myeloid leukemia treated on two COG trials. Septic shock-associated mortality is rare in children with AML and differences by SDI were not able to be detected.
INTRODUCTION
Severe infections occur frequently in children with acute myeloid leukemia (AML) receiving intensive chemotherapy. Population-based cancer studies consistently identify infection as the leading cause of treatment-related morbidity and mortality.1 A life-threatening complication of severe infections is septic shock (SS), defined as sepsis with cardiovascular organ dysfunction.2 Interventions to monitor patients more closely, intensify supportive care, or reduce treatment intensity can improve outcomes.3,4 Previous studies detected increased risk for infection-related mortality in Black children with AML,5 and increased odds of severe sepsis in Hispanic children with AML6 compared to white children. Additionally, mortality was higher among children from minoritized backgrounds who have cancer5,7–9 and among hospitalized children without cancer.10,11 However, race is an inconsistently defined and applied social construct and is often used in research as a proxy for other social determinants of heath (SDOH).12 Children living in disorganized environments can sustain early toxic stress with subsequent altered immune responses and increased risk for adverse health outcomes.13,14 We hypothesize this prior exposure of chronically altered immune response from a disorganized environment may impact infectious complications, like SS, in children with leukemia regardless of where they receive treatment. Further investigation is warranted to better understand the relationship between SDOH and SS outcomes in children with AML.
Individual-level SDOH have not been routinely collected in administrative databases or clinical trials. However, neighborhood-level data have been used as proxies for individual SDOH and have been associated with individual health outcomes.15,16 There are multiple composite neighborhood indexes17,18 comprised of different component domains of SDOH and validated at varying levels of census-based geographic granularity. The Social Disorganization Index (SDI)19 measures neighborhood disorganization, which is driven in part by structural racism resulting in policies and laws driving differential access to resources and opportunities.20 This index is validated at the ZIP code level and has been used to proxy neighborhood SDOH in investigating teen alcohol use19, syphilis infection21, and neurodevelopment in infants22 outcomes. To our knowledge, neighborhood disorganization has not been evaluated in association with SS incidence or outcomes in children with AML. The SDI may allow us to risk stratify patients and guide the development of interventions.
The primary objective of this study was to determine whether neighborhood disorganization, measured by the SDI, was associated with time to SS for children with AML treated on two Children’s Oncology Group (COG) trials. Our secondary objective was to determine whether neighborhood disorganization was associated with septic shock-associated mortality (SS-death). We hypothesized that higher neighborhood disorganization would be associated with higher hazards of SS and SS-death.
METHODS
Data sources
The Pediatric Health Information System (PHIS) is an administrative database that contains inpatient, emergency department, and observation unit information from over 50 not-for-profit, tertiary care pediatric hospitals.23 Data include demographics, dates of service, discharge disposition and vital status, diagnosis codes, and daily inpatient billing data for medications, laboratory tests, imaging procedures, clinical services/procedures, and supplies. Data are deidentified at the time of submission. Quality of submitted data is ensured through a joint effort between the Children’s Hospital Association, Truven Health Analytics, and participating hospitals.
COG is a pediatric cooperative oncology group funded by the National Cancer Institute with over 200 participating centers in the United States, Canada, Europe, New Zealand, and Australia. We have merged COG AAML0531 and AAML1031 trial data with PHIS administrative data.24,25 AAML0531 randomized 1,028 eligible patients to standard chemotherapy ± gemtuzumab for treatment of de novo AML from August 2006 to June 2010.26 AAML1031 trial randomized 1,231 eligible patients to standard chemotherapy ± bortezomib from February 2011 and January 2016.27 Both studies captured extensive diagnostic, leukemia biology, and outcomes data.
Study population
This study used a previously assembled cohort of pediatric patients with de novo AML aged 0–19 years, who enrolled on COG AAML0531 or AAML1031 trial and received treatment at a PHIS participating institution24,25,28 from 2006–2015. Patients with Down syndrome, those with an enrollment date before their treating hospital started contributing inpatient data to PHIS, those who transferred care after treatment initiation, those with incomplete billing data or less than 7 days of therapy, or those with an invalid or non-residential ZIP code (e.g., PO box and/or areas not defined by the Census) were excluded. Patients who did not survive 7 days after start of chemotherapy were excluded.
Social Disorganization Index (SDI)
The primary exposure was the SDI, an area-based composite measure of SDOH domains. We calculated the SDI with data from the 2013 Census Bureau’s American Community Survey (ACS) and validated for 5-digit ZIP Code Tabulation Areas (ZCTA). ACS data were first reported in 2011, however this 2011 version had fewer questions and less comprehensive geographic coverage.29 Given that ACS data provide 5-year estimates,29 2013 ACS 5-year estimates were used to calculate the SDI as this year would encompass both clinical trial periods (2006–2010 and 2011–2016) and allow for comparison. The SDI is calculated as the sum of the following proportions: overall unemployment, households receiving public assistance, low-income persons (<100% and at 100–149% poverty level), high school dropouts, female-headed households, renter-occupied houses, and moved households within the last year.19 For this analysis, patients’ 5-digit home ZIP code recorded on first admission, using COG data, was used to link to the ZCTA-level SDI. To define their exposures to neighborhood disorganization at the ZIP code level, patients were grouped into one of 5 SDI categories (SDI-1 being least and SDI-5 being most disorganized) based on their US nationally normed SDI.
Outcomes
The primary outcome of interest was time from 7 days after start of first course of chemotherapy to first SS event during front-line chemotherapy. We did not assess SS during the first 6 days of therapy due to potential overlap with severity of presentation at diagnosis that may be misclassified. We used our previously developed and validated method to detect SS events in PHIS.30 Our algorithm for SS detection incorporates PHIS resource utilization (RU) codes (Figure 1) with comparable sensitivity and positive predictive value to published papers using diagnostic (ICD-9) codes.31–33 Unlike approaches based on diagnosis codes alone,33 our augmented approach using diagnosis and RU codes allows for determination of the date of onset of SS. PHIS data only provide information on whether a medication was ordered and billed; not whether it was administered to the patient. By design, our cohort used a permissive definition of septic shock, including fluid-responsive shock, as fluid-sensitive shock is clinically relevant and likely indicative of significant infectious or inflammatory events. Thus, our algorithm is designed to capture both fluid-sensitive and fluid-refractory SS.
Figure 1: Septic shock identification.
Vasopressor RU patterns were used to identify SS events in PHIS cohorts. Note: single-agent epinephrine with asparaginase within ± 3 days was considered anaphylaxis, not septic shock.
The secondary outcome of interest was time to SS-death, defined as death within 2 weeks of an antecedent SS event and without evidence of relapse.34 This analysis was restricted to patients with a first SS event.
Covariates
Patient characteristics using COG data included age,35,36 sex,37 insurance type (private, public, or other),38,39 race-ethnicity [Hispanic, non-Hispanic (NH)-Black, NH-white, and other],40 COG study and risk group by COG study.
Statistical analyses
Differences in demographic characteristics by SDI group were assessed using Fisher’s exact test. To evaluate potential confounding, each covariate’s association with exposure and outcome was evaluated and those demonstrated associations with both exposure (P<0.2 or absolute difference in proportion >10%) and outcome (P<0.2 or HR >1.5 or HR <0.67) were retained in the final adjusted model.
Time to SS event (primary outcome) and time to SS-death (secondary outcome) by SDI group were evaluated using the Kaplan-Meier method and compared by SDI group using log-rank tests. Patients were censored at the first of: relapse, death, completion of therapy or removal from the study due to induction failure or patient or clinician preference; all these dates were obtained from COG data. Cox proportional hazard regression was used to estimate unadjusted and adjusted hazard ratios (HRs) with 95% confidence intervals (CIs) for both outcomes. The proportional hazards assumption was assessed by visual inspection of the log-log plot and by the score test using the Schoenfeld residuals.
Stratified analysis by trial was performed as a post-hoc sensitivity analysis to accommodate potential heterogeneity between trials. Additionally, sensitivity analyses were performed restricting follow-up to end of course 1, as previous studies have demonstrated disparities in outcomes during induction.40 Finally, we performed sensitivity analysis collapsing the SDI categories into 3 groups: low (SDI 1 and 2), medium (SDI 3) and high (SDI 4 and 5). All analyses were performed using STATA (StataCorp. 2019. Stata Statistical Software: Release 16. College Station, TX: StataCorp LLC). The study was considered exempt by the Children’s Hospital of Philadelphia Institutional Review Board.
RESULTS
Patient characteristics
The initial eligible cohort included 771 patients diagnosed with de novo AML between October 18, 2006 and December 31, 2015. Excluded patients included patients with Down syndrome (n=2), patients who enrolled on study at hospitals not participating in PHIS and then transferred care to PHIS hospitals (n=33), and patients with incomplete billing data (n=20). Nine patients were excluded due to no ZIP code reported, invalid ZIP code, or absent ZCTA. Only patients who enrolled in either trial and started therapy (2 enrolled and never started therapy) and had at least 7 days of treatment (5 had less than 7 days) were included in the final analysis. The final study population for analysis included 700 patients. Table 1 summarizes the distribution of patient characteristics by SDI group. Higher proportions of Hispanic (37.8%) and NH-Black (17.2%) patients were in SDI-5 compared to the other SDI group (p<0.001). A higher proportion of publicly insured patients (48.1%) were in SDI-5 compared to the other SDI groups (p<0.001). Distributions of age, sex, trial participation, and AML risk groups were more comparable across SDI groups.
Table 1.
Characteristics of children with acute myeloid leukemia according to Social Disorganization Index (SDI).
| Lowest <<< <<>> >>> Highest disorganization disorganization | |||||||
|---|---|---|---|---|---|---|---|
| Total | SDI 1 | SDI 2 | SDI 3 | SDI 4 | SDI 5 | p-valuea | |
| N=700 (100) | n=103 (14.7) | n=103 (14.7) | n=114 (16.3) | n=147 (21.0) | n=233 (33.3) | ||
| Age in yrs, n (%) | 0.62 | ||||||
| <1 | 64 (9.1) | 4 (3.9) | 11 (10.7) | 11 (9.6) | 14 (9.5) | 24 (10.3) | |
| 1–2 | 132 (18.9) | 16 (15.5) | 14 (13.6) | 23 (20.2) | 29 (19.7) | 50 (21.5) | |
| 3–9 | 169 (24.1) | 33 (32.0) | 22 (21.4) | 22 (19.3) | 36 (24.5) | 56 (24.0) | |
| 10–15 | 173 (24.7) | 24 (23.3) | 27 (26.2) | 30 (26.3) | 36 (24.5) | 56 (24.0) | |
| 15+ | 162 (23.1) | 26 (25.2) | 29 (28.2) | 28 (24.6) | 32 (21.8) | 47 (20.2) | |
| Sex, n (%) | 0.44 | ||||||
| Female | 336 (48) | 50 (48.5) | 45 (43.7) | 63 (55.3) | 66 (44.9) | 112(48.1) | |
| Male | 364 (52) | 53 (51.5) | 58 (56.3) | 51 (44.7) | 81 (55.1) | 121(51.9) | |
| Race-ethnicity, n (%) | <0.001 | ||||||
| Hispanic | 136 (19.4) | 8 (7.8) | 8 (7.8) | 12 (10.5) | 20 (13.6) | 88 (37.8) | |
| NHb Black | 66 (9.4) | 2 (1.9) | 1 (1.0) | 7 (6.1) | 16 (10.9) | 40 (17.2) | |
| NH White | 421 (60.1) | 81 (78.6) | 80 (77.7) | 81 (71.1) | 94 (63.9) | 85 (36.5) | |
| Other | 77 (11) | 12 (11.7) | 14 (13.6) | 14 (12.3) | 17 (11.6) | 20 (8.6) | |
| Insurance*, n (%) | <0.001 | ||||||
| Private | 386 (55.1) | 78 (75.7) | 74 (71.8) | 64 (56.1) | 82 (55.8) | 88 (37.8) | |
| Public | 224 (32.0) | 14 (13.6) | 19 (18.4) | 34 (29.8) | 45 (30.6) | 112(48.1) | |
| Other | 90 (12.9) | 11 (10.7) | 10 (9.7) | 16 (14.0) | 20 (13.6) | 33 (14.2) | |
| COG trial, n (%) | 0.41 | ||||||
| AAML 0531 | 340 (48.6) | 59 (57.3) | 48 (46.6) | 52 (45.6) | 68 (46.3) | 113(48.5) | |
| AAML 1031 | 360 (51.4) | 44 (42.7) | 55 (53.4) | 62 (54.4) | 79 (53.7) | 120(51.5) | |
| Risk Group, n (%) | 0.39 | ||||||
| 0531 Low | 75 (10.8) | 14 (13.7) | 11 (10.8) | 7 (6.1) | 17 (11.6) | 26 (11.2) | |
| 0531 Intermediate | 206 (29.6) | 35 (34.3) | 30 (29.4) | 38 (33.3) | 36 (24.7) | 67 (28.8) | |
| 0531 High | 59 (8.5) | 10 (9.8) | 7 (6.9) | 7 (6.1) | 15 (10.3) | 20 (8.6) | |
| 1031 Low | 285 (40.9) | 40 (39.2) | 41 (40.2) | 53 (46.5) | 60 (41.1) | 91 (39.1) | |
| 1031 High | 72 (10.3) | 3 (2.9) | 13 (12.7) | 9 (7.9) | 18 (12.3) | 29 (12.4) | |
| Missing | 3 (0.4) | 1 (1) | 1 (1) | 0 (0) | 1 (0.7) | 0 (0) | |
| Median follow up times in days (IQR**) | |||||||
| 167 (133,214) | 173 (141,221) | 168 (122,202) | 170 (139,221) | 170 (115,217) | 165 (132,210) | ||
Categorical variables compared by Fisher’s exact test
NH= Non-Hispanic
Public insurance includes Medicaid, Medicare, and no means of payment (no insurance); Private insurance includes Medicare and private insurance (together), military sponsored, private insurance, and veterans sponsored insurance; Other insurance includes categories other, self-pay, and unknown.
IQR= interquartile range
Follow up time and distribution of events
The median on-therapy follow-up for the cohort was 167 days [interquartile range (IQR) 133, 214], and was similar by SDI groups (Table 1).
There were 207 (29.6%) patients with a SS event before completing follow-up or experiencing a censoring event. There were 33 (4.7%) patients censored for relapse [SDI-1 n=2 (1.9%), SDI-2 n=6 (5.8%), SDI-3 n=4 (3.5%), SDI-4 n=8 (5.4%), and SDI-5 n=13 (5.8%) ] and two patients (one each in SDI-3 and SDI-4) died without an antecedent SS event. Competing risk analyses for relapse and non-relapse mortality were not performed given the small numbers of competing events in each SDI group and similar distribution across SDI groups.
Association of SDI with first septic shock event
Figure 2 shows Kaplan-Meier estimates for time to first SS according to neighborhood disorganization. SDI was not significantly associated with time to first SS event (log rank, p=0.54).
Figure 2:
Kaplan-Meier time to first septic shock event estimates for children with acute myeloid leukemia according to Social Disorganization Index (SDI). X-axis indicated time since start of chemotherapy + 7 days, in days. Y-axis indicates probability of being septic shock free. Line pattern indicates SDI group. Solid line: SDI 1 (least disorganized); small dash : SDI 2; long dash : SDI 3; short dash dot: SDI 4; long dash dot dot: SDI 5 (most disorganized).
Table 2 presents unadjusted and adjusted (adjusting for confounders: race-ethnicity and insurance status) HRs and 95% Cls for first SS event by SDI group. There was no evidence that the proportional hazards assumption was violated either by formal statistical tests (p>0.5) or visual inspection of Schoenfeld residuals.
Table 2.
Hazard ratio and 95% CI for septic shock event in children with AML according to SDI group.
| SDI group | Unadjusted HR (95% CI) | Adjusted* HR (95% CI) |
|---|---|---|
| SDI 1 | Reference | Reference |
| SDI 2 | 0.85 (0.51–1.42) | 0.84 (0.51–1.41) |
| SDI 3 | 0.77 (0.46–1.28) | 0.70 (0.42–1.16) |
| SDI 4 | 1.12 (0.71–1.75) | 0.97 (0.61–1.53) |
| SDI 5 | 0.91 (0.59–1.39) | 0.72 (0.45–1.14) |
Abbreviations: SDI, social disorganization index; AML, acute myeloid leukemia; CI, confidence interval; HR, hazard ratio.
Adjusted for patient’s race-ethnicity and insurance status
Septic shock-associated mortality in relation to patient characteristics and SDI
Table 3 summarizes the distribution of patient characteristics for the subset of patients (n=207) in the cohort that had at least one SS event by SDI group. The distribution of patients by SDI group was: 15% SDI-1, 13.5% SDI-2, 14.5% SDI-3, 24.2.0% SDI-4 and 32.9% SDI-5. Similar to the full cohort, most patients (32.9%) were in SDI-5.
Table 3.
Characteristics of children with acute myeloid leukemia who had at least one septic shock event according to group of Social Disorganization Index (SDI).
| Lowest <<< <<>> >>> Highest disorganization disorganization | |||||||
|---|---|---|---|---|---|---|---|
| Total | SDI 1 | SDI 2 | SDI 3 | SDI 4 | SDI 5 | p-valuea | |
| N=207 (100) | n=31 (15.0) | n=28 (13.5) | n=30 (14.5) | n=50 (24.2) | n=68 (32.9) | ||
| Age in yrs, n (%) | 0.49 | ||||||
| <1 | 10 (4.8) | 1 (3) | 3 (11) | 1 (3) | 3 (6) | 2 (3) | |
| 1–2 | 27 (13) | 2 (6) | 0 (0) | 6 (20) | 8 (16) | 11 (16) | |
| 3–9 | 43 (21.8) | 8 (26) | 6 (21) | 5 (17) | 13 (26) | 11 (16) | |
| 10–15 | 54 (26.1) | 6 (19) | 9 (32) | 7 (23) | 10 (20) | 22 (32) | |
| 15+ | 73 (35.3) | 14 (45) | 10 (36) | 11 (37) | 16 (32) | 22 (32) | |
| Sex, n (%) | 0.49 | ||||||
| Female | 99 (47.8) | 14 (45) | 13 (46) | 19 (63) | 22 (44) | 31 (46) | |
| Male | 108 (52.2) | 17 (55) | 15 (54) | 11 (37) | 28 (56) | 37 (54) | |
| Race-ethnicity, n (%) | <0.001 | ||||||
| Hispanic | 46 (22.2) | 6 (19) | 2 (7) | 4 (13) | 6 (12) | 28 (41) | |
| NHb Black | 27 (13.0) | 1 (3) | 1 (4) | 3 (10) | 7 (14) | 15 (22) | |
| NH White | 111 (53.6) | 20 (65) | 21 (75) | 20 (67) | 31 (62) | 19 (28) | |
| Other | 23 (11.1) | 4 (13) | 4 (14) | 3 (10) | 6 (12) | 6 (9) | |
| Insurance*, n (%) | 0.002 | ||||||
| Private | 105 (50.7) | 20 (65) | 19 (68) | 19 (63) | 27 (54) | 20 (29) | |
| Public | 64 (30.9) | 4 (13) | 7 (25) | 6 (20) | 14 (28) | 33 (49) | |
| Other | 38 (18.4) | 7 (23) | 2 (7) | 5 (17) | 9 (18) | 15 (22) | |
| COG trial, n (%) | 0.037 | ||||||
| AAML 0531 | 100 (48.3) | 19 (61) | 10 (36) | 13 (43) | 18 (36) | 40 (59) | |
| AAML 1031 | 107 (51.7) | 12 (39) | 18 (64) | 17 (57) | 32 (64) | 28 (41) | |
| Risk Group, n (%) | 0.44 | ||||||
| 0531 Low | 22 (10.7) | 8 (26) | 2 (7) | 0 (0) | 4 (8) | 8 (12) | |
| 0531 Intermediate | 63 (30.7) | 10 (32) | 7 (26) | 11 (37) | 11 (22) | 24 (35) | |
| 0531 High | 15 (7.3) | 1 (3) | 1 (4) | 2 (7) | 3 (6) | 8 (12) | |
| 1031 Low | 83 (40.5) | 9 (29) | 12 (44) | 17 (57) | 23 (47) | 22 (32) | |
| 1031 High | 22 (10.7) | 3 (10) | 5 (19) | 0 (0) | 8 (16) | 6 (9) | |
| Missing | 2 (1) | 0 (0) | 1 (4) | 0 (0) | 1 (2) | 0 (0) | |
Categorical variables compared by Fisher’s exact test
NH= Non-Hispanic
Public insurance includes Medicaid, Medicare, and no means of payment (no insurance); Private insurance includes Medicare and private insurance (together), military sponsored, private insurance, and veterans sponsored insurance; Other insurance includes categories other, self-pay, and unknown.
There were 9 SS-deaths. Due to the limited number of events, Kaplan-Meier estimates and Cox regression models were not performed. Table 4 summarizes the distribution of patient characteristics who had SS-deaths by SDI group. Seven of the 9 patients (78%) belonged in SDI-4 or 5, and 7 out of 9 patients (78%) were NH white. Five out of 9 patients (56%) had SS-death in cycle 1 of therapy.
Table 4.
Characteristics of children with acute myeloid leukemia who had septic shock-associated mortality according to group of Social Disorganization Index (SDI).
| Total | SDI 1 | SDI 2 | SDI 3 | SDI 4 | SDI 5 | |
|---|---|---|---|---|---|---|
| N=9 (100) | n=0 (0) | n=1 (11) | n=1 (11) | n=3 (33) | n=4 (44) | |
| Age in yrs, n (%) | ||||||
| <10 | 6 (67) | 0 (0) | 0 (0) | 1 (100) | 2 (67) | 3 (75) |
| 10+ | 3 (33) | 0 (0) | 1 (100) | 0 | 1 (33) | 1 (25) |
| Sex, n (%) | ||||||
| Female | 6 (67) | 0 (0) | 1 (100) | 0 (0) | 1 (33) | 4 (100) |
| Male | 3 (33) | 0 (0) | 0 (0) | 1 (100) | 2 (67) | 0 (0) |
| Race-Ethnicity, n (%) | ||||||
| Hispanic | 1 (11) | 0 (0) | 0 (0) | 0 (0) | 0 (0) | 1 (25) |
| NH Black | 0 (0) | 0 (0) | 0 (0) | 0 (0) | 0 (0) | 0 (0) |
| NH White | 7 (78) | 0 (0) | 1 (100) | 1 (100) | 2 (67) | 3 (75) |
| Other | 1 (11) | 0 (0) | 0 (0) | 0 (0) | 1 (33) | 0 (0) |
| Insurance, n (%) | ||||||
| Private | 6 (67) | 0 (0) | 0 (0) | 1 (100) | 3 (100) | 2 (50) |
| Public | 1 (11) | 0 (0) | 0 (0) | 0 (0) | 0 (0) | 1 (25) |
| Other | 2 (22) | 0 (0) | 1 (100) | 0 (0) | 0 (0) | 1 (25) |
| Trial, n (%) | ||||||
| AAML0531 | 5 (56) | 0 (0) | 0 (0) | 0 (0) | 1 (33) | 4 (100) |
| AAML1031 | 4 (44) | 0 (0) | 1 (100) | 1 (100) | 2 (67) | 0 (0) |
| Course number, n (%) | ||||||
| 1 | 5 (56) | 0 (0) | 1 (100) | 0 (0) | 2 (50) | 2 (50) |
| >1 | 4 (44) | 0 (0) | 0 (0) | 0 (0) | 2 (50) | 2 (50) |
Sensitivity Analysis
Primary analyses were repeated by trial (AAML0531 and AAML1031). Supplemental Figure 1 and supplemental Figure 2 shows Kaplan-Meier estimates for time to first SS according to SDI by trial. SDI was not significantly associated with time to first SS event for AAML0531 (log rank, p=0.264) or for AAML1031 (log rank, p=0.117). In our adjusted model for AAML0531 there is a decreased hazard of SS for SDI-2–5 when compared to SDI-1 that approaches statistical significance. In contrast, the adjusted HR for SDI-2–5 compared to SDI-1 for AAML1031 were not statistically significant (Supplemental Tables 1–4). Additionally, SDI was not significantly associated with time to first SS event when we restricted follow up to end of cycle 1 nor was there a significant association when we collapsed the SDI categories into 3 groups.
DISCUSSION
In a large, nationally representative cohort of trial-enrolled pediatric patients with AML, there was no significant association between SDI and time to SS in children treated on two COG trials. Thus, the SDI was not a useful metric to risk stratify these patients. This study is the first to investigate the relationship of neighborhood disorganization proxied by the SDI, an area-based measure of SDOH, and frequency of SS in children with AML.
The general pediatric literature has mixed findings on the association of area-based measures of SDOH and infection-related outcomes, including sepsis. One study found no significant association between neighborhood social vulnerability and prolonged lengths of intensive care unit stays for children with severe sepsis and underlying medical complexity, not restricted to patients with cancer.41 Nationally, roughly 80% of pediatric patients receive AML care in the inpatient setting.42 Therefore, we assume most of our population of patients with AML was already admitted to the hospital following receipt of chemotherapy. Once hospitalized, disparities in access to medical care may be less impactful; thus, the care our study population received at large children’s hospitals might mitigate the impact of neighborhood SDOH on SS outcomes. In contrast, in a cohort of infants presenting to medical care for fever and infection evaluation, patients exposed to a high neighborhood poverty status had higher rates of bacterial infection compared to patients from low poverty neighborhoods.43
An adult study found that sepsis-related mortality was higher among patients living in low-income neighborhoods compared to patients from higher income neighborhoods and that neighborhood poverty, lack of insurance, and lower formal education status were independently associated with sepsis-related mortality.44 This study used community data for the city of Baltimore, creating a more demographically granular area of measurement. This method differs from ours, which used a less granular 5-digit ZIP-code based measure with a higher risk of misclassification, which may have biased our results towards the null.
Fifty-four percent of the patients in our cohort came from neighborhoods assigned to SDI-4 or 5 (more disorganized neighborhoods), and this finding was similar across the two trials. To our knowledge, this is the first time this has been reported in a pediatric AML population. There are some studies that applied the area deprivation index in pediatric acute lymphoblastic leukemia45 and pediatric central nervous system tumors46 populations, and their cohorts had proportionate numbers of patients by area deprivation index quartiles. Further investigation is needed to understand the distribution of social disorganization among trial-enrolled pediatric AML patients.
There were 9 SS-deaths events in the cohort, and 78% (n=7) of them were in SDI-4 or SDI-5 (highest disorganization). Seven out of 111 NH-white patients and zero out of 27 NH-Black patients had a SS-death, which contrasts with an older study that showed NH-Black patients had higher mortality from infection.5 Our SS definition was based on resource utilization codes,30 while the previous study used clinical trial adverse event reporting, which can be inaccurate.47 A previous study showed higher acuity of presentation in NH-Black children with AML compared to NH-white children, which was associated with increased mortality during Induction40. Since our study excluded the first 7 days of treatment to differentiate acuity of initial presentation from chemotherapy-associated toxicities, our study intentionally excludes sepsis present at the time of AML diagnosis. Septic shock-associated mortality is rare in children with AML receiving intensive chemotherapy, and differences by SDI were not able to be detected in this cohort.
Our stratified analysis showed evidence of heterogeneity by trial. The trials had similar clinical demographics including similar number of patients (340 for AAML0531 and 360 for AAML1031), a similar distribution of patients by SDI, and a similar proportion of patients had our primary outcome. Methodologically we used 2013 ACS 5-year estimates to calculate SDI for both trials to allow for comparison. We suspect that the heterogeneity between trials may be due to the small numbers of events across the 5 SDI groups.
The majority of studies investigating SDOH as they relate to cancer outcomes use area-based measures.48 However, individual measures of SDOH may be more useful in developing targeted patient-centered interventions. Although there was no association between SDI and hazard of SS, individual-level measures of SDOH likely play a role in pediatric cancer outcomes. Acknowledging the importance of these key risk factors, COG and the Dana-Farber Cancer Institute Acute Lymphoblastic Leukemia Consortium have incorporated individual assessments of SDOH in recent50 and upcoming phase 3 trials (clinicaltrials.gov ID: NCT03020030 and NCT03914625).
This study has several important limitations to note. First, it used an area-based measure to calculate neighborhood disorganization. Populations comprising ZIP codes and ZCTAs tends to be more heterogenous than other census geographies, such as block groups. ZIP codes and ZCTAs are large in area, and assignment of an SDI group to an individual patient or patient’s family using 5-digit ZIP code data is more vulnerable to misclassification than using SDI groups derived using data from a more granular measurement such as block groups. Additionally, only patients with a valid U.S. 5-digit zip code were included in this analysis, thus our results cannot be applied to populations outside the United States or in certain rural areas unmapped to ZIP codes. To compare results by trial, we calculated the SDI using 2013 ACS 5-year estimates, which ensures the same methodology in calculating our exposure but risks misclassification if there were changes in SDI during the early and late years of AAM0531 and AAML1031 respectively. In this dataset, we were not able to distinguish which patients were discharged early from those who remained inpatient through count nadir and recovery and thus were not able to evaluate if early discharge was associated with time to first SS. Our SS outcome is defined by an intentionally inclusive algorithm that includes fluid-responsive SS and thus may result in higher number of SS events. Finally, our study only captures patients enrolled on a clinical trial and treated at PHIS participating institutions, which are limited to free-standing academic children’s hospitals. Patients treated off trial and/or treated in a community hospital were not captured and thus our results may not include those with the greatest barriers to accessing care.
This study found no statistically significant association between an SDI derived from 5-digit ZIP code areas and hazard of SS in children being treated for AML. Additional studies are needed to further explore individual-level and other neighborhood-level measures of SDOH and SS in children undergoing cancer treatment. It may be necessary to leverage more granular measures of SDOH to identify differential risk for poor outcomes among pediatric cancer populations. The potential impact of SDOH on patient outcomes may be mitigated by care provided in the inpatient setting. Therefore, possible association between SDOH and clinical outcomes may be more apparent in cohorts of children with a cancer that is managed in the outpatient setting such as acute lymphoblastic leukemia.
Supplementary Material
Funding Support:
Jenny Ruiz was supported by an Alex’s Lemonade Stand Foundation and Northwestern Mutual Young Investigator Grant (21-23920), the Cancer Clinical Epidemiology Training Grant from the National Cancer Institute (T32-CA009679), and the Abramson Cancer Center’s Paul Calabresi Career Development Award for Clinical Oncology (K12 CA076931). Kelly Getz received support for research effort in part through an NIH Career Development Award (K01HL143153). Lena Winestone received for research effort in part through an American Society of Hematology Scholar Award
Disclosures:
BTF receives research funding from Allovir, Merck and Pfizer and serves on a Data Safety Monitoring Board for Astellas. None of these efforts are related to this publication.
Footnotes
Conflict of Interest
The rest of the co-authors have no conflict of interests to disclose.
Ethical Approval: Ethical approval was not required as study did not meet criteria for human subject research.
Data Availability Statement:
Data is available on request from authors.
REFERENCES
- 1.Pole JD, Gibson P, Ethier M-C, et al. Evaluation of treatment-related mortality among paediatric cancer deaths: a population based analysis. Br J Cancer. 2017;116(4):540–545. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Goldstein B, Giroir B, Randolph A. International pediatric sepsis consensus conference: definitions for sepsis and organ dysfunction in pediatrics. Pediatr Crit care Med a J Soc Crit Care Med World Fed Pediatr Intensive Crit Care Soc. 2005;6(1):2–8. [DOI] [PubMed] [Google Scholar]
- 3.Taub JW, Berman JN, Hitzler JK, et al. Improved outcomes for myeloid leukemia of Down syndrome: a report from the Children’s Oncology Group AAML0431 trial. Blood. 2017;129(25):3304–3313. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Miller TP, Getz KD, Kavcic M, et al. A comparison of discharge strategies after chemotherapy completion in pediatric patients with acute myeloid leukemia: a report from the Children’s Oncology Group. Leuk Lymphoma. 2016;57(7):1567–1574. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Aplenc R, Alonzo TA, Gerbing RB, et al. Ethnicity and survival in childhood acute myeloid leukemia: a report from the Children’s Oncology Group. Blood. 2006;108(1):74–80. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Savage B, Thomas-Hawkins C, Cole PD, Stapleton JL, de Cordova PB. Increased Risk of Severe Sepsis in Hispanic Children Hospitalized With Acute Myeloid Leukemia. J Pediatr Oncol Nurs Off J Assoc Pediatr Oncol Nurses. 2020;37(6):349–358. [DOI] [PubMed] [Google Scholar]
- 7.Castellanos MI, Dongarwar D, Wanser R, et al. In-hospital Mortality and Racial Disparity in Children and Adolescents With Acute Myeloid Leukemia: A Population-based Study. J Pediatr Hematol Oncol. 2022;44(1):e114–e122. [DOI] [PubMed] [Google Scholar]
- 8.Winestone LE, Getz KD, Miller TP, et al. Complications preceding early deaths in Black and White children with acute myeloid leukemia. Pediatr Blood Cancer. 2017;64(12). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Winestone LE, Getz KD, Bona KO, et al. Area-Based Socioeconomic Disparities in Survival of Children with Newly Diagnosed Acute Myeloid Leukemia: A Report from the Children’s Oncology Group. Blood. 2019;134(Supplement_1):703. [Google Scholar]
- 10.Mitchell HK, Reddy A, Montoya-Williams D, Harhay M, Fowler JC, Yehya N. Hospital outcomes for children with severe sepsis in the USA by race or ethnicity and insurance status: a population-based, retrospective cohort study. Lancet Child Adolesc Heal. 2021;5(2):103–112. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Li E, Ng AP, Williamson CG, Tran Z, Federman MD, Benharash P. Assessment of Racial and Ethnic Disparities in Outcomes of Pediatric Hospitalizations for Sepsis Across the United States. JAMA Pediatr. Published online November 2022. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Flanagin A, Frey T, Christiansen SL. Updated Guidance on the Reporting of Race and Ethnicity in Medical and Science Journals. JAMA. 2021;326(7):621–627. [DOI] [PubMed] [Google Scholar]
- 13.Johnson SB, Riley AW, Granger DA, Riis J. The science of early life toxic stress for pediatric practice and advocacy. Pediatrics. 2013;131(2):319–327. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.MacGillivray DM, Kollmann TR. The role of environmental factors in modulating immune responses in early life. Front Immunol. 2014;5:434. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Adler NE, Rehkopf DH. U.S. disparities in health: descriptions, causes, and mechanisms. Annu Rev Public Health. 2008;29:235–252. [DOI] [PubMed] [Google Scholar]
- 16.Diez Roux A V, Mair C. Neighborhoods and health. Ann N Y Acad Sci. 2010;1186:125–145. [DOI] [PubMed] [Google Scholar]
- 17.Singh GK. Area deprivation and widening inequalities in US mortality, 1969–1998. Am J Public Health. 2003;93(7):1137–1143. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Andrews MR, Tamura K, Claudel SE, et al. Geospatial Analysis of Neighborhood Deprivation Index (NDI) for the United States by County. J Maps. 2020;16(1):101–112. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Byrnes HF, Miller BA, Morrison CN, Wiebe DJ, Woychik M, Wiehe SE. Association of environmental indicators with teen alcohol use and problem behavior: Teens’ observations vs. objectively-measured indicators. Health Place. 2017;43:151–157. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Jindal M, Trent M, Mistry KB. The Intersection of Race, Racism, and Child and Adolescent Health. Pediatr Rev. 2022;43(8):415–425. [DOI] [PubMed] [Google Scholar]
- 21.Bonett S, Tam V, Singapur A, Min J, Koenig HC, Wood SM. Incidence of syphilis infection and syphilis-related care utilization among adolescents and young adults living with HIV. Int J STD AIDS. 2022;33(2):136–143. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Favilla E, Faerber JA, Hampton LE, et al. Early Evaluation and the Effect of Socioeconomic Factors on Neurodevelopment in Infants with Tetralogy of Fallot. Pediatr Cardiol. 2021;42(3):643–653. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Maude SL, Fitzgerald JC, Fisher BT, et al. Outcome of pediatric acute myeloid leukemia patients receiving intensive care in the United States. Pediatr Crit care Med a J Soc Crit Care Med World Fed Pediatr Intensive Crit Care Soc. 2014;15(2):112–120. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Aplenc R, Fisher BT, Huang YS, et al. Merging of the National Cancer Institute-funded cooperative oncology group data with an administrative data source to develop a more effective platform for clinical trial analysis and comparative effectiveness research: a report from the Children’s Oncol. Pharmacoepidemiol Drug Saf. 2012;21 Suppl 2(Suppl 2):37–43. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Li Y, Hall M, Fisher BT, et al. Merging Children’s Oncology Group Data with an External Administrative Database Using Indirect Patient Identifiers: A Report from the Children’s Oncology Group. PLoS One. 2015;10(11):e0143480. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Gamis AS, Alonzo TA, Meshinchi S, et al. Gemtuzumab ozogamicin in children and adolescents with de novo acute myeloid leukemia improves event-free survival by reducing relapse risk: results from the randomized phase III Children’s Oncology Group trial AAML0531. J Clin Oncol Off J Am Soc Clin Oncol. 2014;32(27):3021–3032. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Aplenc R, Meshinchi S, Sung L, et al. Bortezomib with standard chemotherapy for children with acute myeloid leukemia does not improve treatment outcomes: a report from the Children’s Oncology Group. Haematologica. 2020;105(7):1879–1886. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Kavcic M, Fisher BT, Torp K, et al. Assembly of a cohort of children treated for acute myeloid leukemia at free-standing children’s hospitals in the United States using an administrative database. Pediatr Blood Cancer. 2013;60(3):508–511. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Census Bureau US. Understanding and Using American Community Survey Data: What All Data Users Need to Know; 2020. [Google Scholar]
- 30.Seif AE, Li Y, Cahen VC, et al. Reduced Relapse Risk in Children with Acute Myeloid Leukemia (AML) Who Experience Septic Shock (SS). Blood. 2019;134(Supplement_1):3496. [Google Scholar]
- 31.Ruth A, McCracken CE, Fortenberry JD, Hall M, Simon HK, Hebbar KB. Pediatric severe sepsis: current trends and outcomes from the Pediatric Health Information Systems database. Pediatr Crit care Med a J Soc Crit Care Med World Fed Pediatr Intensive Crit Care Soc. 2014;15(9):828–838. [DOI] [PubMed] [Google Scholar]
- 32.Weiss SL, Parker B, Bullock ME, et al. Defining pediatric sepsis by different criteria: discrepancies in populations and implications for clinical practice. Pediatr Crit care Med a J Soc Crit Care Med World Fed Pediatr Intensive Crit Care Soc. 2012;13(4):e219–26. [DOI] [PubMed] [Google Scholar]
- 33.Balamuth F, Weiss SL, Hall M, et al. Identifying Pediatric Severe Sepsis and Septic Shock: Accuracy of Diagnosis Codes. J Pediatr. 2015;167(6):1295–300.e4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Alexander S, Pole JD, Gibson P, et al. Classification of treatment-related mortality in children with cancer: a systematic assessment. Lancet Oncol. 2015;16(16):e604–10. [DOI] [PubMed] [Google Scholar]
- 35.Creutzig U, Büchner T, Sauerland MC, et al. Significance of age in acute myeloid leukemia patients younger than 30 years: a common analysis of the pediatric trials AML-BFM 93/98 and the adult trials AMLCG 92/99 and AMLSG HD93/98A. Cancer. 2008;112(3):562–571. [DOI] [PubMed] [Google Scholar]
- 36.Razzouk BI, Estey E, Pounds S, et al. Impact of age on outcome of pediatric acute myeloid leukemia: a report from 2 institutions. Cancer. 2006;106(11):2495–2502. [DOI] [PubMed] [Google Scholar]
- 37.Meshinchi S, Arceci RJ. Prognostic factors and risk-based therapy in pediatric acute myeloid leukemia. Oncologist. 2007;12(3):341–355. [DOI] [PubMed] [Google Scholar]
- 38.Ward E, Halpern M, Schrag N, et al. Association of insurance with cancer care utilization and outcomes. CA Cancer J Clin. 2008;58(1):9–31. [DOI] [PubMed] [Google Scholar]
- 39.Bradley CJ, Dahman B, Jin Y, Shickle LM, Ginder GD. Acute myeloid leukemia: how the uninsured fare. Cancer. 2011;117(20):4772–4778. [DOI] [PubMed] [Google Scholar]
- 40.Winestone LE, Getz KD, Miller TP, et al. The role of acuity of illness at presentation in early mortality in black children with acute myeloid leukemia. Am J Hematol. 2017;92(2):141–148. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Hamilton H, West AN, Ammar N, et al. Analyzing Relationships Between Economic and Neighborhood-Related Social Determinants of Health and Intensive Care Unit Length of Stay for Critically Ill Children With Medical Complexity Presenting With Severe Sepsis. Front public Heal. 2022;10:789999. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Getz KD, Miller TP, Seif AE, et al. A comparison of resource utilization following chemotherapy for acute myeloid leukemia in children discharged versus children that remain hospitalized during neutropenia. Cancer Med. 2015;4(9):1356–1364. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Yaeger JP, Moore KA, Melly SJ, Lovasi GS. Associations of Neighborhood-Level Social Determinants of Health with Bacterial Infections in Young, Febrile Infants. J Pediatr. 2018;203:336–344.e1. [DOI] [PubMed] [Google Scholar]
- 44.Galiatsatos P, Brigham EP, Pietri J, et al. The effect of community socioeconomic status on sepsis-attributable mortality. J Crit Care. 2018;46:129–133. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Schraw JM, Peckham-Gregory EC, Rabin KR, Scheurer ME, Lupo PJ, Oluyomi A. Area deprivation is associated with poorer overall survival in children with acute lymphoblastic leukemia. Pediatr Blood Cancer. 2020;67(9):e28525. [DOI] [PubMed] [Google Scholar]
- 46.Adel Fahmideh M, Schraw JM, Chintagumpala M, Lupo PJ, Oluyomi AO, Scheurer ME. Neighborhood Socioeconomic Deprivation and Mortality in Children with Central Nervous System Tumors. Cancer Epidemiol biomarkers Prev a Publ Am Assoc Cancer Res cosponsored by Am Soc Prev Oncol. 2021;30(12):2278–2285. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.Miller TP, Li Y, Kavcic M, et al. Accuracy of Adverse Event Ascertainment in Clinical Trials for Pediatric Acute Myeloid Leukemia. J Clin Oncol Off J Am Soc Clin Oncol. 2016;34(13):1537–1543. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48.H Tran Y, Coven SL, Park S, Mendonca EA. Social determinants of health and pediatric cancer survival: A systematic review. Pediatr Blood Cancer. 2022;69(5):e29546. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49.Zheng DJ, Shyr D, Ma C, Muriel AC, Wolfe J, Bona K. Feasibility of systematic poverty screening in a pediatric oncology referral center. Pediatr Blood Cancer. 2018;65(12):e27380. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50.Aziz-Bose R, Zheng DJ, Umaretiya PJ, et al. Feasibility of oncology clinical trial-embedded evaluation of social determinants of health. Pediatr Blood Cancer. 2022;69(11):e29933. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Data is available on request from authors.