NIOSHTIC-2 Publications Search

Re: DDT and neurodevelopment: results inconclusive to effect policy change.

Eskenazi-B; Marks-AR; Bradman-A; Johnson-C; Jewell-NP
Pediatrics 2006 Sep; 118(1):P3R2291
We thank Dr. Goga for her thoughtful feedback. First, we agree that no single epidemiologic investigation should be considered conclusive and that our study results clearly need to be replicated in other populations, but we also believe that all data from well-conducted studies, conclusive or not, should be considered for policy decisions. We try to address each of her specific concerns below: 1. Sample: There were 281 children with DDT/E measurements and Bayley assessments at all three ages. When multivariate analyses were limited to these children and compared to those for the full sample available at each time, the patterns of associations were similar and, in most cases, the coefficients and p-values were stronger, despite the smaller sample sizes. Also, we conducted longitudinal analyses, as stated in the text [1], and the results were similar. 2. Loss to follow-up/non-inclusion in analysis: While 601 women were enrolled in the study [1], 20 miscarried, 5 neonates/fetuses died, and 43 dropped (mostly due to relocation) before delivery, leaving 531 followed past delivery and 476 followed to 6 months. We performed Bayley assessments at least once on 459 children, or on 86% of infants alive and in the study following delivery. Blood levels of organochlorines were measured for 426 women (rather than 526 as stated by Dr. Goga), yielding an overlap with Bayley assessments of 364 children; those not analyzed were due to lack of a stored serum sample of adequate volume. As suggested, to investigate potential bias introduced by non-inclusion, we made the following comparisons: a. We compared the baseline demographic characteristics of those in the original cohort and the sample included in the paper (i.e., those who had both DDT/E levels and Bayley scores) and found that women not included were more likely to smoke and were less likely to be married or living as married. Controlling for these factors did not alter our results. b. We compared exposure levels in those who had a Bayley assessment performed at all time points and those who did not and found no significant differences in DDT or DDE levels. c. We compared the Bayley scores of those who had DDT/E levels and those who did not and found that Bayley scores did not differ at 24-months where we observed the strongest associations. However, children without DDT/E measurements had lower 12-month MDI. We further note that selection bias arising from non-inclusion solely on the basis of high or low Bayley scores would tend to bias the association with DDT/E exposure towards the null. Selection bias on the basis of exposure is irrelevant since all regression analyses already condition on the observed values of the explanatory variables. 3. Measuring exposure: It is not possible to know the proportion of the DDT/E levels observed in this population due to agricultural exposure versus anti-malaria efforts. Women in our study who were born in coastal Mexico had significantly higher levels of DDT and DDE than those who did not, and levels increased with the number of years spent outside of the U.S. [2]. In northern and central Mexico, agricultural DDT use declined in mid-1970s in response to U.S. import standards for OC residues [3], and in 1990 the Cicoplafest commission seriously restricted DDT use, which was then banned nationwide in 1997 for purposes other than public health use inside dwellings. DDT use for malaria control continued in coastal areas until 2000. To date, very little information has been published on exposure resulting from IRS with DDT. Background levels in individuals living in African nations with historical or current DDT use appear to be at least as high [4,5] as those observed in our population (please see our response to Kahiira). Thus, although we do not know the precise source of the DDT in our population, if populations living in areas with IRS spraying have maternal serum levels comparable to or higher than those in our study, then the associations we observed would still be relevant. We do not currently have postnatal exposure information, but we attempted to account for postnatal exposure by using as a proxy the duration of breastfeeding, which is the primary source of postnatal exposure in children in the U.S. We also aimed to estimate postnatal exposure more precisely by examining the interaction between maternal levels of DDT/E and the duration of breastfeeding. 4. Measuring and interpreting outcome: The Bayley Scales of Infant Development II is a well-known assessment of infant development. We agree that any assessment of infants can have limited predictive validity, and we do not know the predictive validity of this test in this population. These issues necessitate both follow-up of these children into school age and similar studies in other populations. In the meantime, this study indicates a relationship within our population between maternal serum levels of DDT and neurodevelopment by age two years. Unless there was a systematic bias introduced in the assessment, i.e., all those with high exposure were assessed in a different manner, these associations merit consideration. 5. Measuring confounders and other variables: We considered for analyses many of the covariates mentioned by Dr. Goga, but did not consider any variables that were not considered in the literature to be related to the exposure or outcome or were possibly on the causal pathway. We considered pertinent medical conditions, including those resulting from birth trauma, and the few children affected were excluded from the study either de facto (i.e. they were not assessed or valid scores could not be calculated) or in the data analysis. We have not yet measured postnatal exposure to neurotoxicants other than lead, which was low. To account for the associations observed, postnatal exposure to other neurotoxicants would need to be associated with prenatal DDT/E levels and unrelated to prenatal levels of the same chemicals, which we did consider in analyses. The CES-D is a well-regarded tool to screen for maternal depression in epidemiologic investigations. Although available in Spanish, we agree it may not be a valid tool in an immigrant population. Nevertheless, the limitations of the instrument are consistent across this homogeneous population. We are not sure how the limitations of the CES-D would alter the conclusions of the present study on the associations of DDT and neurodevelopment. 6. Statistical analysis: We followed Dr. Goga's suggestion to examine neurodevelopment by dichotomous exposure. Thus, we compared children with prenatal DDT/E exposure in the highest quartile to children whose exposure fell in the lowest quartile, controlling for the same covariates described in the paper (see Table 1). These new results are entirely consistent with the findings presented in the paper. Specifically, as shown in Table 1, children in the highest quartile of prenatal exposure to p,p'-DDT and p,p'-DDE had significantly lower mean PDI scores at 6 and 12 months (-3.3 to -5.8 points) and borderline lower mean PDI scores for p,p'-DDT at 24 months (-2.8 points) than children in the lowest quartile. Relative to the least exposed children, children in the highest quartile of both p,p'-DDT and o,p'-DDT exposure had significantly lower mean MDI scores at 12 months (-3.4 to -4.2 points), and children most highly exposed to both isomers of DDT and to DDE, had significantly lower mean MDI scores at 24 months (-4.0 to -5.1 points). Regarding Dr. Goga's other point about the source of exposure, as stated above, the serum DDT/E measures represent body burden. It is not possible to identify the source of the exposure, nor does it call into question the observed association. Furthermore, very little information is presently available on DDT exposure resulting from IRS, and there is a need for further exposure monitoring and research. 7. Other issues of concern: As stated in the paper [1], mean Bayley PDI scores at 6, 12, and 24 months were 96.4, 106.7, and 97.8, respectively; mean MDI scores were 95.4, 100.9, and 86.0.With the single exception of the 24-month MDI, these were all close to the expected mean score of 100, which would seem to indicate that using the age-standardized scores is not entirely unwarranted. If a child performed more than two standard deviations below the norm, the child's guardian was informed, the child was referred to his or her pediatrician for further assessment, and study staff spoke to the pediatrician. Such assessments were excluded from these analyses since they were outliers. The assessments were done for research, not clinical purposes, and the parents and clinicians were informed of this. We considered them inappropriate for clinical purposes, because we translated the instrument to Spanish (yet made no other modifications to the instrument), and although the psychometricians performing the assessments were trained extensively and were clinically supervised, they did not have graduate degrees. Given the paucity of bilingual psychologists in this community, we believed it was paramount to have assessments completed by bilingual bicultural members with experience with children from their community. Thus, these assessments were purely for research purposes to enable comparison of neurodevelopment within this population. While it is important to acknowledge the shortcomings of this instrument and administration, the relevant issue here is whether the methods of administration would alter the association of DDT and neurodevelopment in this population. Assessors were blind to DDT levels and exposure-related factors, thus there is no reason to believe that they would have been systematically biased in their assessments based on children's exposures. In summary, we agree that more data are needed on IRS exposure, that the children in the present cohort should be followed to determine the persistence of the observed associations and the contribution of breastfeeding, and that other studies should be conducted to examine these associations within other populations. In the meantime, we believe that it is for policy makers to weigh the benefits and risks of DDT use using all the data available.
In-utero-exposure; Children; Demographic-characteristics; Racial-factors; Prenatal-exposure; Agriculture; Agricultural-industry; Agricultural-workers; Farmers
Publication Date
Document Type
Email Address
Funding Amount
Funding Type
Fiscal Year
Identifying No.
Issue of Publication
Source Name
Performing Organization
University of California, Berkeley
Page last reviewed: April 12, 2019
Content source: National Institute for Occupational Safety and Health Education and Information Division