Clark et al. Environmental Health( 2026) 25:12 Page 3 of 13 and socio-economic status [ 2, 50 – 53 ]. The study protocol has been approved by the Institutional Review Boards at the California Health and Human Services, University of Southern California, and Yale University.
Exposure assessment Oil and gas data were collected from Enverus’ s Drilling Info platform on March 26, 2024. Data retrieved included all California wells in the“ Wells” database( n = 253,171). We restricted our exposure dataset to directionally, horizontally, and vertically drilled oil and gas wells spud( i. e., drilled) or producing during the period of interest( 1980 – 2015) with a status of‘ active’ or‘ plugged and abandoned’. Eligible wells were those that were documented as( a) active( drilled or producing) or( b) plugged or abandoned during the exposure window of interest. For wells with a first production start date but no spud date, we assigned a spud date by subtracting the median number of days between spud and first production, from the first production date.“ Active” status includes wells that have never been designated as“ Plugged and Abandoned”; only the active time window was considered for these wells.“ Plugged and abandoned” status includes wells that have been designated as“ Plugged and Abandoned”; only the plugged and abandoned time window was considered, defined as the day after the last production date through the end of the study period. Because many plugged and abandoned wells lacked sufficient information to assign them cutoff dates, we focused on those wells clearly designated as‘ plugged and abandoned’ during our study period. This data was further cleaned to remove duplicates, resolve missing data, and fix structural errors.
We used maternal residential address at the time of birth, obtained from birth records and geocoded by the University of California Berkeley [ 54 ], to assign OGD exposure metrics individually for each case and control during the prenatal exposure window of interest, three months before conception to birth. The critical window of exposure for Ewing sarcoma is poorly understood. We chose to focus on the prenatal period both because many childhood cancers are thought to initiate prenatally [ 50, 55, 56 ] and because of the association between congenital hernia and Ewing sarcoma risk [ 57, 58 ], which also supports a potential prenatal origin. To represent OGD exposure for each subject, we calculated( 1) distance to nearest well,( 2) number of wells within a buffer zone, and( 3) an inverse distance-squared weighted( ID 2 W) well count. The ID 2 W metric was constructed using the Euclidean distance between the subject’ s home and surrounding eligible OGD wells within 1, 2, 5, and 10 km( ArcGIS 10.8.1). All distances are Euclidian using a projected coordinate system specific to California( EPSG: 3310). We selected these buffer sizes based on the hydrogeologic( supports contaminant transport distances of 2 km or less [ 30, 59, 60 ]) and epidemiologic( incorporates buffer distances up to 20 km [ 61, 62 ]) literature. Exposure metrics were averaged over the primary prenatal exposure window, three months before conception to birth, to estimate cumulative exposure. OGD exposures were operationalized as ordinal( for number of wells within a buffer zone), binary( any or no exposure), and as tertiles( low / medium / high exposure). All distance calculations were completed using the latest version of R( R version 4.4.0 as of June 2025).
Statistical analysis All statistical analyses were conducted in SAS 9.4, and all tests were two-sided with an alpha level of 0.05. We used chi-square and t-tests to identify differences in the distribution of population characteristics and OGD exposure between case and control children. We used unconditional logistic regression to estimate odds ratios( OR) and 95 % confidence intervals( CI) for the association between active, plugged and abandoned, and all OGD exposure and risk of pediatric Ewing sarcoma, adjusting for year of birth( the matching variable), sex, race, ethnicity, birth weight( continuous, per 500 g) and community-level SES.
Sensitivity analyses Because people of color are more likely to be disproportionately exposed to environmental hazards including OGD in the United States [ 63 – 70 ], we compared exposure distributions and stratified models by race and ethnicity when possible to evaluate potential exposure inequalities with respect to OGD. Rurality is an important predictor of OGD siting [ 71 ]. Additionally, we conducted a sensitivity analysis using unconditional logistic regression to estimate the association between OGD exposure and risk of Ewing sarcoma adjusting for region of birth( assigned by County) in addition to all other confounders included in the main model( year of birth [ the matching variable ], sex, race, ethnicity, birth weight [ continuous, per 500 g ] and community-level SES).
Results Population characteristics Case children were more likely to be male( 58 % vs. 51 %, p < 0.01), have a birth weight > 4000 g( 14 % vs. 11 %, p = 0.01), and be non-Hispanic White( 48 % vs. 37 %, p < 0.01; Table 1) than control children. Case children also had significantly lower social vulnerability indices than control children across all domains examined, indicating lower social vulnerability and higher SES( all p≤0.02; Table 2). When stratified by Hispanic ethnicity, this pattern remained evident among non-Hispanic cases, but not among Hispanic cases( Table 2). Most of the cases included in this study were diagnosed at 10 – 14( 33 %) and 15 – 19( 29 %) years of age.