Clustering of pediatric onset inflammatory bowel disease in Finland: a nationwide register-based study.

BACKGROUND: The incidence of pediatric inflammatory bowel disease (PIBD) has increased dramatically during the past decades. This implies involvement of environmental factors in etiology but lends no clues about specific agents. We evaluated clustering in time and place of residence at PIBD onset using a case-control setting with comprehensive nationwide register data. METHODS: We included all PIBD cases diagnosed at ages < 18 years during 1992-2017 (3748 cases; median age of 14.6; 2316 (58%) with ulcerative colitis (UC), 1432 with Crohn's, and 18,740 age- and sex-matched controls) and constructed complete residential histories (including coordinates) from the national database until the date of the diagnosis of the case assigned as index date for the controls. Using the coordinates of the addresses of the subjects and the diagnosis/index dates, we evaluated clustering in time and place using the Knox test. Four temporal (2, 4, 6, 12 months) and four distance (0.25, 0.5, 1, 5 km) thresholds were used, and results were calculated separately for Crohn´s disease and UC. Similar analyses were conducted using the addresses at birth and the addresses five years before the diagnosis or index date. Based on the threshold values displaying the most clustering in the Knox test, logistic regression models were built to identify whether sex, age at diagnosis or the year of diagnosis affected the probability of belonging to a cluster. To analyze clustering in time and place throughout the residential histories, we used Jacquez's Q with an open-access python program pyjacqQ. RESULTS: The mean number of residencies until the index date was 2.91 for cases and 3.05 for controls (p = 0.0003). Knox test indicated residential clustering for UC with thresholds of 500 m between locations and time-period of four months (p = 0.004). In the regression analysis, sex, age at diagnosis or year of UC diagnosis did not show differences between the clustered and other cases. Jacquez Q analyses showed higher than expected frequency of clustered cases throughout residential histories (p < 10- 8). CONCLUSION: Our findings suggest that the incidence of PIBD, especially of UC, exhibits clustering in locations of residencies over time. For the clustered cases, environmental triggers warrant future studies.