The authors respond:
We appreciate the comments and suggestions by Weinberg1 regarding our analysis of arsenic in drinking water and intrauterine growth.2 We agree that the social and other contextual “site-to-site” differences can be important factors, in addition to individual-level variables. Weinberg cites as an example the potentially differential effect of an important covariate such as prenatal care and suggests including interaction terms in the analysis. Although we neglected to mention it in the article, we did examine interaction terms for exposure and covariates in the data analysis, with particular attention to those factors that contributed substantially to the birthweight variability. Because none of these product terms were significant or suggestive of important differences, we did not include them in the final model. In particular, the interaction term of city by Kessner index indicated virtually no heterogeneity. We also ran the final models (without interaction terms) separately for each city, and the coefficients for the levels of Kessner index showed a similar trend and similar magnitude.
Weinberg1 raises concerns about social differences between the 2 cities. In fact, the differences in income and education presented in Table 1 of our article2 are not substantial. It may not have been clear from the article that the 2 study groups were quite homogeneous and were similar within the spectrum of socioeconomic levels in Chile. Although some variation in income and education categories were observed (Table 1), all the women in the study were of relatively midlow to low income, with 89% in Antofagasta and 95% in Valparaiso living in households earning less than US$700 monthly. Second, the fact they were all recruited from prenatal clinics of the Public Health Service in Chile is also an indicator of socioeconomic status and of a rather homogeneous population, because women from higher income brackets or with work-related health insurance are likely to get prenatal care outside this system. Finally, we performed the regression analysis, including an interaction term for income by city, and found no evidence of interaction.
We agree with Weinberg1 that the lack of variation in individual exposure is a limitation in this study. That the problem is not unique to our study is suggested by the recent commentary on the difficulties of assessing exposures that occur from drinking water.3 However, the contrast in exposure between the cities is very clear given their differing arsenic concentrations, as well as the fact that all women drank water from the same public supply in each city. It is also reassuring that a recent publication4 reported very similar findings in a study of arsenic exposure and birthweight in Taiwan, lending additional support for a possible causal association.
1.Weinberg C. Arsenic and drinking water [Letter]. Epidemiology
2.Hopenhayn C, Ferreccio C, Browning S, et al. Arsenic exposure from drinking water and birthweight. Epidemiology
3.Steenland K, Moe C. Epidemiology and drinking water, are we running dry? [Commentary]. Epidemiology
4.Yang CY, Chang CC, Shyue T, et al. Arsenic in drinking water and adverse pregnancy outcome in an arseniasis-endemic area in northeastern Taiwan. Environ Res