Correspondence on
Breastmilk, PCBs, Psychodevelopment

THE LANCET 20apr02

mindfully.org note:
Be sure to read the authors reply at the end.

Effect of polychlorinated biphenyls on psychodevelopment

Sir

Jens Walkowiak and colleagues (Nov 10, p 1602)1 show significantly raised serum polychlorinated biphenyl (PCB) concentrations (sum of the three higher-chlorinated PCB congeners 138, 153, and 180) in children breastfed long term (>4 months, mean PCB concentration 1·77 ng/mL) compared with those breastfed short term (>2 weeks to 4 months, 0·68 ng/mL) and those bottle-fed (0·36 ng/mL) at age 42 months. They conclude that this postnatal PCB exposure was a causative factor in various kinds of mental and motor developmental delay at age 30 months and 42 months.

I doubt this causative relation. As Joseph and Sandra Jacobson state in their Nov 10 Commentary,2 as have other workers,3 neurotoxic effects of PCBs have been seen previously only in relation to smaller quantities transmitted across the placenta. The greater sensitivity of the developing fetal brain seems to render the fetus much more vulnerable to small quantities of different neurotoxicants.

There has been a significant decline in prenatal accumulation of PCBs in Germany in the past 15-20 years.4 Thereby, cord-blood reference values obtained in 1998 correspond to those presented by Walkowiak and colleagues (mean 0·39 ng/mL). In only 1994, cord-blood reference values obtained from the German general population were higher (0·90 ng/mL) than those presented for children aged 42 months, who were breastfed short term. 15 years ago, cord-blood reference values (1·37 ng/mL) were nearly as high as the values now presented for infants breastfed long term, who Walkowiak and colleagues deem to have neurodevelopmental impairments because of high PCB concentrations. Did all children born before the prohibition of PCBs in Germany in 1989 therefore have some kind of neurodevelopmental delay? Surely not.

I believe that the small amounts of PCBs today transmitted to fetuses across the placenta or to infants via breastmilk produce no neurotoxic effects, at least not by themselves. The developing brain is exposed to various xenobiotics, and it is additional effects that may lead to adverse reactions. For example, children of smoking mothers may have increased prenatal uptake of PCBs.5 The cocarcinogenic properties of PCBs and tobacco may lead to long-lasting adverse biological effects that have still to be elucidated.

Furthermore, mean PCB concentration at age 42 months was lower than that in cord blood in bottle-fed infants (0·36 vs 0·39 ng/mL) in Walkowiak and colleagues' study. This finding needs explanation because it contradicts previous findings on the kinetics of PCBs in human beings. Because of their chemical persistence and lipophilic properties, PCBs accumulate in the food chain. The main source of uptake for men is through diet, continuously from birth to adulthood. Although a higher PCB body burden in breastfed than in bottle-fed children seems reasonable, there is no explanation why children aged 42 months should have lower PCB concentrations than at birth.

Lastly, exposure to environmental tobacco smoke significantly increases the prenatal5 and postnatal uptake of PCBs in human beings (unpublished data). Therefore, this factor should be excluded for the children in Walkowiak and colleagues' study.

* Gerd-Michael Lackmann, Zentrum für Kinderheilkunde, Heinrich-Heine University, 40225 Düsseldorf, Germany (e-mail:gmlackmann@t-online.de)

1 Walkowiak J, Wiener J-A, Fastabend A, et al. Environmental exposure to polychlorinated biphenyls and quality of the home environment: effects on psychodevelopment in early childhood . Lancet 2001; 358: 1602-07. [Text]

2 Jacobson JL, Jacobson SW. Postnatal exposure to PCBs and childhood development . Lancet 2001; 358: 1568-69. [Text]

3 Landrigan PJ. Pesticides and polychlorinated biphenyls: an analysis of the evidence that they impair children's neurobehavioral development. Mol Genet Metabol 2001; 73: 11-17. [PubMed]

4 Lackmann GM. Polychlorinated biphenyls and hexachlorobenzene in full-term neonates: reference values updated. Biol Neonate 2002; 81: 82-85. [PubMed]

5 Lackmann GM, Angerer J, Töllner U. Parental smoking and neonatal serum levels of polychlorinated biphenyls and hexachlorobenzene. Pediatr Res 2000; 47: 598-601. [PubMed]

Sir,

Jens Walkowiak and colleagues1 have tested two general hypotheses: do environmental levels of exposure to PCBs adversely affect psychomotor and mental development during early childhood; and, if such effects are found, can they be counteracted by a favourable home environment?

These hypotheses are indisputably non-directional, since both allow for results to occur in either direction. Consequently, the statistical tests must be interpreted as two-tailed probabilities. After several data analyses, Walkowiak and colleagues conclude that PCBs have a significantly negative impact on children's psychomotor and mental development from age 30 months onward, and the quality of the home environment has a significant positive effect on psychomotor and mental development, also from age 30 months onwards. Unfortunately, these results cannot be ascribed scientific validity because of some serious design flaws.

First, the researchers state that they used one-tailed probabilities throughout, because the hypotheses were directional, and effects at p<0·05 were significant. Adjusting for this flaw means doubling each of the reported probability levels. This required adjustment2 changes the results strikingly. Of nine comparisons involving PCBs, six were significant with one-tailed tests, but only two remained significant with the two-tailed tests. Furthermore, the research scientist must then adjust for the nine post-hoc PCB comparisons that were done to eliminate the mistake of misinterpreting chance for significant probability. After this adjustment, no PCB comparison comes close to significance.

By contrast, Walkowiak and colleagues also made nine comparisons involving home environment; again, six were significant before but only five remained significant after two-tailed tests were done; three were still significant after adjustment for the nine multiple comparisons. But even if most PCB results, like the home results, had remained significant, that still would not be sufficient.

The major difficulty is that the home environment needs to be treated as a confounding variable, as it has been in other PCB3 and lead-level4 investigations, rather than as a separate independent variable. Joseph and Sandra Jacobson5 point to the importance of adjusting for potential confounders, especially the home environment, parental education, and maternal intelligence quotient.

Furthermore, Walkowiak and colleagues note that the home environment is an influential confounder in neurotoxicity studies, which contradicts their curious decision to omit this factor as a confounder in their study. The essential scientific question must be, what is the effect, if any, of PCB concentrations on psychomotor and mental development, after the effects of home environment have been eliminated? I predict that even the unadjusted one-tailed significant PCB effects would become invalid, because, as the researchers correctly conclude, the overall positive impact of the home environment is stronger than the negative effect of neonatal PCB exposure.

*Domenic V Cicchetti, Alan S Kaufman, Child Study Center, Yale University School of Medicine, 230 South Frontage Road, New Haven, CT 06520, USA (e-mail: dom.cicchetti@yale.edu)

1 Walkowiak J, Wiener J-A, Fastabend A, et al. Environmental exposure to polychlorinated biphenyls and quality of the home environment: effects on psychodevelopment in early childhood. Lancet 2001; 358: 1602-07. [Text]

2 Toothaker LE. Multiple comparisons for researchers. Newbury Park: Sage Publications, 1991.

3 Pantandin S, Lanting CI, Mulder PGH, Boersma ER, Sauer PJJ, Weisglas-Kuperus N. Effects of environmental exposure to polychlorinated biphenyls and dioxins on cognitive abilities in Dutch children at 42 months of age. J Pediatr 1999; 134: 33-41. [PubMed]

4 Kaufman AS. Do low levels of lead produce IQ loss in children? A careful examination of the literature. Arch Clin Neuropsychol 2001; 16: 303-41. [PubMed]

5 Jacobson JL, Jacobson SW. Postnatal exposure to PCBs and childhood development. Lancet 2001; 358: 1568-69. [Text]

Authors' reply

Sir

Gerd-Michael Lackmann summarises our main findings incorrectly. We found negative associations between PCBs in milk taken at age 2 weeks, and Bayley or Kaufman test scores taken at ages 7, 18, 30, and 42 months. These early PCB values in milk reflect maternal body burden during pregnancy and, thus, prenatal exposure. We report also a pure postnatal impact of PCB for the Kaufman test at 42 months after adjustment for prenatal PCB exposure.

The doubts about the validity of this pure postnatal association are understandable, because most studies stress the importance of prenatal exposure. Thus, our observations require corroboration. However, Joseph and Sandra Jacobson do state that our observations are consistent with experimental data, which is one of the requirements for inferring causality from correlations.1

Lackmann believes that the small amounts of PCBs transmitted to fetuses or to infants produce no neurotoxic effects by themselves. However, he contradicts this by admitting that neurotoxic effects of PCBs were previously shown in relation to the small quantities transmitted across the placenta to the vulnerable developing brain.

He suggests that smoking mothers or environmental tobacco smoke increase the prenatal uptake of PCBs and may have long-lasting effects. We included maternal smoking during pregnancy in our regression model.

Lackmann is surprised that the serum PCB concentrations at age 42 months in bottle-fed infants were slighly lower than those in cord-blood values. In the group as a whole, the expected postnatal increase of PCB body burden does occur. The slight decrease in the small group of bottle-fed children is probably because of dilution related to rapid early growth, and the associated increase of fat compartments.2,3

Domenic Cicchetti and Alan Kaufman suggest inappropriate testing of non-directional hypotheses, with no adjustment for multiple testing or quality of the home environment. Our hypotheses were directional. We tested for adversity of prenatal and neonatal PCB exposure, because the assumption of a beneficial impact of PCBs on motor and mental development is not biologically plausible. We also tested for a beneficial effect of the home environment because, for psychological reasons, good parental caretaking supports rather than impairs psychodevelopment. Thus, one-tailed testing is mandatory.

Six of the nine comparisons made were significant. This is very unlikely to have occurred by chance alone (p<0·0001). Division of p values by nine would be necessary to correctly interpret only one of nine comparisons. In contrast to Cicchetti and Kaufman's reasoning, we do not simply rely on p values and formal hypothesis testing to show negative developmental PCB effects; these criteria are judged least important in epidemiological reasoning.4,5 We, therefore, presented dose-response information and effect sizes additionally and took into account possible sources of bias.

Cicchetti and Kaufman may have misunderstood our statistical methods. In our two-step model building, variables in the first step were parental education, sex, maternal intelligence, as well as home environment and PCB concentration simultaneously. Therefore, in testing for PCB effects we adjusted for home environment effects and vice versa. Thus, our procedure fully conforms to what our critics rightly consider indispensable and is also in line with the Jacobsons' commentary. *G Winneke, J Walkowiak, U Kraemer, H-J Steingrüber, B Heinzow

* Medical Institute of Environmental Hygiene, and Institute of Medical Psychology, Heinrich-Heine-Universität, 40225 Düsseldorf, Germany; and State Agency of Nature and Environment of Schleswig-Holstein, Flintbeck (e-mail: gerhard.winneke@uni-duesseldorf.de)

1 Hill AB. The environment and disease: association or causation? Proc R Soc Med 1965; 58: 295-300. [PubMed]

2 Patandin S, Weisglas-Kuperus N, de Ridder MAJ, et al. Plasma polychlorinated biphenyl levels in Dutch preschool children either breast-fed or formula-fed during infancy. Am J Public Health 1997; 87: 1711-14. [PubMed]

3 Lanting Cl, Fidler V, Huisman M, Boersma ER. Determinants of polychlorinated biphenyl levels in plasma from 42-months-old children. Arch Environ Contam Toxicol 1998; 35: 135-39. [PubMed]

4 Savitz DA, Olshan AF. Multiple comparisons and related issues in the interpretation of epidemiological data. Am J Epidemiol 1995; 142: 904-08. [PubMed]

5 Thompson JA. Multiple comparisons and related issues in the interpretation of epidemiological data. Am J Epidemiol 1998; 147: 801-06. [PubMed]