Spatial cluster mapping and environmental modeling in pediatric inflammatory bowel disease

Geographical (geospatial) clusters have been observed in inflammatory bowel disease (IBD) incidence and linked to environmental determinants of disease, but pediatric spatial patterns in North America are unknown. We hypothesized that we would identify geospatial clusters in the pediatric IBD (PIBD) population of British Columbia (BC), Canada and associate incidence with ethnicity and environmental exposures.

To identify PIBD clusters and model how spatial patterns are associated with population ethnicity and environmental exposures.

One thousand one hundred eighty-three patients were included from a BC Children's Hospital clinical registry who met the criteria of diagnosis with IBD ≤ age 16.9 from 2001-2016 with a valid postal code on file. A spatial cluster detection routine was used to identify areas with similar incidence. An ecological analysis employed Poisson rate models of IBD, Crohn's disease (CD), and ulcerative colitis (UC) cases as functions of areal population ethnicity, rurality, average family size and income, average population exposure to green space, air pollution, and vitamin-D weighted ultraviolet light from the Canadian Environmental Health Research Consortium, and pesticide applications.

Hot spots (high incidence) were identified in Metro Vancouver (IBD, CD, UC), southern Okanagan regions (IBD, CD), and Vancouver Island (CD). Cold spots (low incidence) were identified in Southeastern BC (IBD, CD, UC), Northern BC (IBD, CD), and on BC's coast (UC). No high incidence hot spots were detected in the densest urban areas. Modeling results were represented as incidence rate ratios (IRR) with 95%CI. Novel risk factors for PIBD included fine particulate matter (PM_2.5) pollution (IRR = 1.294, CI = 1.113-1.507, P < 0.001) and agricultural application of petroleum oil to orchards and grapes (IRR = 1.135, CI = 1.007-1.270, P = 0.033). South Asian population (IRR = 1.020, CI = 1.011-1.028, P < 0.001) was a risk factor and Indigenous population (IRR = 0.956, CI = 0.941-0.971, P < 0.001), family size (IRR = 0.467, CI = 0.268-0.816, P = 0.007), and summer ultraviolet (IBD = 0.9993, CI = 0.9990-0.9996, P < 0.001) were protective factors as previously established. Novel risk factors for CD, as for PIBD, included: PM_2.5 air pollution (IRR = 1.230, CI = 1 .056-1.435, P = 0.008) and agricultural petroleum oil (IRR = 1.159, CI = 1.002-1.326, P = 0.038). Indigenous population (IRR = 0.923, CI = 0.895-0.951, P < 0.001), as previously established, was a protective factor. For UC, rural population (UC IRR = 0.990, CI = 0.983-0.996, P = 0.004) was a protective factor and South Asian population (IRR = 1.054, CI = 1.030-1.079, P < 0.001) a risk factor as previously established.

PIBD spatial clusters were identified and associated with known and novel environmental determinants. The identification of agricultural pesticides and PM_2.5 air pollution needs further study to validate these observations.