Overall, the Study indicates a weak inverted U-shaped association between maternal blood
manganese level and birth weight at the time of delivery. It potentially shows a possible
inflection point at the high exposure in the association where manganese level change from
beneficial to detrimental in relation with birth weight. On the other hand, the study found no
association between umbilical cord manganese level and birth weight at the time of delivery.
The following context will focus on the analysis concerning the association between maternal
blood manganese level and birth weight.
The study examined the association through different types of models. Upon using penalized
splines, a non-linear association was found after accounting for various potential biologic and
socioeconomic confounders of birth weight. Yet, this model found a linear association when
maternal blood manganese level up to around 3 ug/dL, after that point, a modest inverse U shape
was observed.
Multivariate regression model with quadratic polynomial function was also employed to examine
the potential U shaped dose-response relationship. The result, in consistent with the spline model,
also showed that “the birth weight increased with manganese levels up to 3.1-ug/L and a slight
reduction in weight when levels exceeded this point.” (Zota et al. 2009) After the birth weight
estimates were compared with the 3.1-ug/L point, the 5th percentiles of exposure (1.3 ug/L)
were -160g, 95%CI(-286, -33) and the 95th percentiles of exposure (4.0 ug/L) were -46g,
95%CI(-38-131).
The authors further examined the association between maternal blood manganese level and
birth weight by quintiles based on maternal blood manganese level. After adjusting for potential
confounders, the U-shape association was found, but the estimates for the 5th quintile was not
statistically significant (p-value = 0.75).
This study is important in a sense that it was the first epidemiologic study providing a
non-linear association between maternal manganese blood level and birth weight. It is also
consistent with other animal studies that showed an adverse reproductive outcome with high
manganese exposure. However, the study did have several limits especially when it comes to
the association between the manganese level exceeding the estimate point of inflection
3.1 ug/L or 5th quintile and the decreased birth weight.
First, the biomarker deployed in this study may not correctly reflect the exposure
level of the fetus since residual confounders may exist and change manganese level
that are not accounted for in this study, such as nutritional, dietary factors or other heavy
metal exposures. As mentioned by the study and given that the study focuses on the subjects
living in proximity with superfund sites, some heavy metals, such as iron, lead, zinc, may possibly
be associated or confound with the estimate in birth weight. Moreover, the manganese level was
measured at the delivery which may not accurately reflect the fetus exposure during the whole
pregnancy considering other effects such as chemicals’ half-life.
Second, from the data provided, reader can only ascertain a relatively weak association between
high manganese exposure and reduced birth weight. Under the multivariate regression model
with quadratic polynomial function, the birth weight estimate at 95th percentiles of exposure were
-46g, yet the 95% confidence interval ranged from -38g o 131, which overlapped both negative
and positive values. Furthermore, upon using quintiles to examine the association, the estimate of
5th quintile was not statistically significant. There may be multiple possible reason to explain
this weak association. One potential reason is that there were too few subjects in the 95th
percentiles and 5th quintiles. Another reason could be that the cutoff point in quintiles did not
truly reflect the biological plausibility which manipulated the estimate of the outcome.
In terms of the generalizability, this study focuses on the subjects that live near superfund site
in which various exposure to toxic substances or heavy metals are highly plausible. Yet, the study
did not account for possible exposures to other chemicals such as lead and zinc or even radioactive
and other harmful exposures. Therefore, we must be cautious when applies the result of this study
to the general public.
In conclusion, this study can hardly be the sole basis to suggest an occupational or environmental
exposure standards even though it provided a strong basis for further study. for several reasons.
First, this study can’t be generalized into the general public. Moreover, there were possibly
residual confounder existed in the study such as other chemical exposure, unknown dietary and
nutritional factors, etc. Further, the estimates in the 95th percentile and 5th quintile only
showed weak association, albeit statistically insignificant association. Therefore, this study is
not sufficient to be the basis in setting up standard until further collaborating studies can
supplement the findings.