Abstract
Introduction
Infection is the most important cause of treatment-related mortality in leukemia. Recent adult data have associated gut microbiome alterations with systemic infection. As environmental and dietary factors influence the microbiome, we sought to explore neighborhood factors associated with blood stream infection (BSI) and specifically mucosal barrier injury (MBI) BSI in patients receiving treatment for leukemia.
Methods
We assembled two retrospective cohorts, ALL and AML patients, treated at the Children's Hospital of Philadelphia between 2006-2015. Type of leukemia and therapy start and end dates were manually abstracted from the electronic medical record (EMR). Then, we used an automated REDCapTM-Epic data query tool to interrogate the EMR and capture patient demographics, residential addresses and all blood cultures. Data were extracted from date of presentation until end of therapy, stem cell transplant, relapse or death. The primary outcome was incident BSI. A secondary outcome was incident mucosal barrier injury (MBI) BSI per the CDC definition. To define neighborhood factors, addresses at treatment start were matched to geo-coordinates that linked each patient to a census-derived block group. Block groups were then connected with a priori chosen socioeconomic variables (poverty and neighborhood disorganization index), expenditure data (spending on fresh produce, fast food and tobacco products) and environmental characteristics (air quality indices and lead exposure) hypothesized to impact the microbiome. Neighborhood poverty was defined as the proportion of people in the block group with income below the federal poverty level.
Risk of first BSI and first MBI BSI were estimated from a generalized linear model with a binary distribution. Spline regression was used to confirm a linear association between neighborhood poverty and BSI risk. Several additional exploratory analyses were performed to evaluate the impact of neighborhood factors on BSI at specific timepoints including BSI within 30 days of therapy start and outpatient BSIs (defined by a positive blood culture collected while outpatient or on the day of admission).
Results
There were 378 patients with ALL and 74 with AML. The median proportion of people below the poverty limit in patient neighborhoods was 5% (IQR: 2-12%) in ALL and 7% (IQR: 2-18%) in AML. Demographics by disease and neighborhood poverty are shown in table 1.
One fifth of ALL patients and 68.9% of AML patients had at least one BSI. Overall, 51.6% of incident BSI events were classified as MBI BSI (ALL 36%, AML 74.5%). The median time to first BSI was 91 days in ALL and 65 days in AML. In ALL patients, the onset of BSI was in the outpatient setting 54.7% of the time, whereas only 15.7% of BSIs were outpatient in AML.
Among patients with AML, the proportion of neighborhood poverty was inversely associated with incident BSI and MBI BSI: the relative risk per percent increase in population below the poverty level was .27 (95% CI = .06 - 1.16) and .12 (95% CI = .01 - 1.08), respectively. These estimates were robust to adjustment. No other neighborhood factors were associated with BSI or MBI BSI. For ALL patients the association between neighborhood poverty and BSI was similar to that found in the AML cohort but with less precision (RR = .48, 95% CI: .07 - 3.18). The point estimates using specific BSI timepoints were less suggestive of an association in either direction.
Conclusions
In a cohort of children with AML, we found that the proportion of neighborhood poverty was inversely associated with BSI and MBI BSI. While not statistically significant, the magnitude of the association was striking. A similar trend was noted in ALL patients. Other neighborhood factors - including diet and tobacco expenditures, pollution and lead exposure - showed no association.
Although poverty is generally associated with poor health outcomes, mechanisms can be proposed to explain an association between poverty and BSI in either direction. Poverty may lead to gut dysbiosis and subsequent BSI by limiting access to healthy foods; alternatively, poverty-associated environmental exposures and less antibiotic use could increase microbiome diversity and protect against BSI. These data suggest the latter. Larger, geographically diverse studies are needed to further explore this association, as well as translational studies evaluating the human microbiota in the context of neighborhood factors.
Fisher:Merck: Research Funding; Pfizer: Research Funding.
