Validation of self-reported perception of proximity to industrial facilities: MCC-Spain study

Castelló, Adela
Pérez-Gómez, Beatriz
Lora, David
Lope, Virginia
Castaño-Vinyals, Gemma
Vitelli-Storelli, Fecundo
Dierssen Sotos, Trinidad
Amiano, Pilar
Guevara, Marcela
Moreno, Victor
Lozano-Lorca, Macarena
García Tardón, Adonina
Alguacil, Juan
Hernández-García, Marta
Chirlaque, María Dolores
Ardanaz, Eva
Ibarluzea, Jesús
Gómez Acebo, Inés
Molino, Antonio J.
O'Callaghan-Gordo, Cristina
Aragonés, Nuria
Kogevinas, Manolis
Pollán, Marina
García-Pérez, Javier
Self-reported data about environmental exposures can lead to measurement error. Objectives To validate the self-reported perception of proximity to industrial facilities. Methods MCC-Spain is a population-based multicase-control study of cancer in Spain that recruited incident cases of breast, colorectal, prostate, and stomach cancer. The participant’s current residence and the location of the industries were geocoded, and the linear distance between them was calculated (gold standard). The epidemiological questionnaire included a question to determine whether the participants perceived the presence of any industry at ≤1 km from their residences. Sensitivity and specificity of individuals' perception of proximity to industries were estimated as measures of classification accuracy, and the area under the curve (AUC) and adjusted odds ratios (aORs) of misclassification were calculated as measures of discrimination. Analyses were performed for all cases and controls, and by tumor location, educational level, sex, industrial sector, and length of residence. Finally, aORs of cancer associated with real and self-reported distances were calculated to explore differences in the estimation of risk between these measures. Results Sensitivity of the questionnaire was limited (0.48) whereas specificity was excellent (0.89). AUC was sufficient (0.68). Participants with breast (aOR(95%CI) = 2.03 (1.67;2.46)), colorectal (aOR(95%CI) = 1.41 (1.20;1.64)) and stomach (aOR(95%CI) = 1.59 (1.20;2.10)) cancer showed higher risk of misclassification than controls. This risk was higher for lower educational levels (aOR<primary vs. university (95%CI) = 1.78 (1.44;2.20)), among younger participants (aOR22-54 years vs. 73-85 years (95%CI) = 1.32 (1.09;1.60)), and for some industrial sectors: pharmaceutical (aOR(95%CI) = 29.02 (19.52;43.14)), galvanization (aOR(95%CI) = 14.14 (6.78;29.47)), and ceramic (aOR(95%CI) = 12.73 (7.22;22.44)). Participants living ≤1 year in the study area showed a lower risk of misclassification ((aOR≤1 vs. >15 years (95%CI) = 0.56 (0.36;0.85)). The use of self-reported proximity vs. real distance to industrial facilities biased the effect on cancer risk towards the nullity. Conclusions Self-reported distance to industrial facilities can be a useful tool for hypothesis generation, but hypothesis-testing studies should use real distance to report valid conclusions. The sensitivity of the question might be improved with a more specific formulation ​
This document is licensed under a Creative Commons:Attribution - Non commercial - No Derivate Works (by-nc-nd) Creative Commons by-nc-nd4.0