Research published in JAMA Paediatrics demonstrates that prenatal exposure to to phthalates was associated with language delay in children.
A Roundup accompanied this Before the Headlines.
Title, Date of Publication & Journal
‘Association of prenatal phthalate exposure with language development in early childhood’ by Carl-Gustaf Bornehag et al. will be published in JAMA Pediatrics at 15:00 UK time on Monday 29 October 2018.
Study’s main claims – and are they supported by the data
The main claim of this study is that higher prenatal exposure to some (but not other) phthalates is associated with delayed language development in young children. Eight phthalates were looked at, and two of them were found to be associated with delayed language development. Also, even though there is some evidence of an association between certain phthalate exposure and delayed language development, as this is an observational study we cannot be certain that any of these observed associations were causal. It is possible that various other factors could account for the observed associations. The authors tried to account for some of these factors but there may be others that they did not measure that could account for the association, such as diet, socioeconomic factors, smoking, etc. Also, the study does not directly measure exposure of the foetus – they measure the concentration in the mother’s urine as an approximation.
So, this study does not show that phthalates cause delayed language development in children, and nor does it claim to.
The authors analysed data from two separate cohorts. For one of the cohorts the authors found statistically significant associations between higher levels of two different types of phthalate (out of the eight compounds investigated) and delayed language development. Doublings of urinary concentration of phthalates referred to as MBP and MBzP were associated with 29% and 26% higher odds of delayed language development, respectively. (Odds ratios approximate relative risks when the outcome is uncommon. Only 6% of children have delayed language development so this is probably a reasonable assumption.) However, the authors do not find evidence of a dose response relationship (increases in odds of delayed language development as phthalate exposure increases) when they divided phthalate exposure into quartiles, which might have been expected if the relationship had been simple or causal.
The associations were investigated in two cohorts. The main claim is supported by one of the cohorts looked at, but not by the other. The second cohort looked at was small and did not provide any statistically significant results, so does not lend any support to the associations identified in the larger cohort (though this may have been because it was too small).
Language delay was defined as a vocabulary of 50 words or fewer at an age of 30 months. Phthalate exposure was estimated from phthalate concentrations in a single urine sample from each mother that was taken in the first trimester of pregnancy.
The authors present results from analysis of two cohorts.
Language development was assessed by a validated method in both cohorts.
Phthalate in the mothers’ urine was assessed by standardised laboratory methods.
Language development was only assessed at one point in child development.
The study assumes that first trimester foetal exposure to phthalates can be assessed from just one measure of phthalate concentration in maternal urine. The authors do not discuss whether this is a sound assumption. There is no direct measurement of foetal exposure to phthalates – only the concentration measured in the mothers’ urine.
The authors do not provide any data in the main paper on how much urinary phthalate levels differed between the women looked at. Without this information it is difficult to interpret these results. The supplementary material contains some additional data on phthalate distribution but it may be incomplete.
The authors also present the two associations for which there were statistically significant associations (MBzP and MBP) by quartile of phthalate distribution. They do not find evidence of a dose response relationship. However, they do find a statistically significant higher odds of delay in language development in those in the top quartile of phthalate exposure compared with the bottom quartile. Although the odds of delay in language development in the top quartile is about two times that in the lowest quartile, there is a large degree of uncertainty in the estimate.
There were issues with missing data in both studies. These reduced the ability of the studies to detect statistically significant associations. Missing data could potentially affect the observed associations. The authors describe the missing data and its possible effects. The proportion of missing data in one study meant that it was too small to detect any statistically significant results.
The authors’ abbreviations and classification of phthalates is confusing. For example, they talk about “DBP” in the text but “DBP” is not mentioned in Table 3 or Figure 1. Instead they use “MBP” etc. The relationships between these different phthalates are not explained in the main text, but there is some information in Supplementary Table 1.
Odds = the ratio of something happening to it not happening.
Quartile = each of four equal groups into which a population is divided according to the ordered distribution of phthalate levels.
Any specific expertise relevant to studied paper (beyond statistical)?
I have experience of observational studies, but I do not have any experience of the study area.