The objective of this systematic review was to gather and discuss the main studies available in literature, on lead levels in human milk. One difficulty faced in the selection of articles was the heterogeneity of methods used in the different studies. This difficulty, however, is common in systematic reviews. As new inclusion criteria are added, a gain in specificity is achieved but also a loss in sensitivity, compromising the objective of showing the scenario of publications on a specific subject. Because of the substantial heterogeneity and the limitations in methodology in the original studies, we considered a quantitative pooling inappropriate. Hence, the qualitative systematic review was conducted on the evidence available. Although multiple search strategies were used to identify relevant studies, some publications may have been missed.
Among the different studies, a diversity of epidemiologic design and analytic methodologies to measure lead in breast milk could be observed, making data comparison difficult and not allowing for a precise evaluation of estimates. Hence, determining whether these gaps are a result of analytic limitations or if the data truly reflect different types of exposure is difficult. Several studies are limited by the small number of samples, non-adjustment for confounding variables, and short follow-up periods. Other factors include differences in the age groups studied, parity, lactation stage, as well as bias in selecting the participants. Not every author followed international definitions and parameters as inclusion criteria, which also brought difficulty in comparing these studies. All these factors were fundamental for establishing the criteria for the selection of studies for this review.
The articles analyzed in the present review were characterized, in their majority, as cross-sectional studies. Note that only the study by Ettinger et al /13
/ had a longitudinal design.
Detection levels are comparable, and most studies have reported these limits, which can be seen in . Nevertheless, the ways in which values lower than the detection limit are reported and used in statistical analysis were not presented in all papers. These numbers must be reported, either as non-detectable
or by using half their values in statistic calculations. Another possibility would be to divide the limit of detection value by the square root of 2 or, in a more liberal way, use the detection limit value itself in these calculations /31
/. However, these options must be carefully and clearly described and the studies have not presented the option used, with the exception of Koyashiki et al /32
/, which used half the detection limit for statistics calculations.
Considering that the concentrations of lead in milk did not necessarily present a normal distribution, the best measures of central tendencies to be used would be the median and the geometric mean /33
/, which are less influenced by high values than the arithmetic mean. Nevertheless, a great portion of the studies analyzed in the present review presented the arithmetic mean as their central measure.
In the articles conducting analyses of lead in milk, the highly sensitive techniques like ICP-MS and GFAAS could guarantee results that would present good accuracy and precision.
The sample preparation steps are usually the most common source of analytic error and usually increase the overall imprecision of the method /34
/. We must also consider that because of the low levels of metal in milk, one problem was the contamination of samples during collection or analytical determination as milk is a substance that can be easily contaminated. Much of the older data and values > 3 parts per billion of lead in breast milk reported during the last 15 years have been questioned /4,5,18
/. Using a comparisons of ratios expressed as percentage of lead concentration in milk to that in blood constitutes a method to validate results. Data showing a milk/blood ratio of > 15% should be treated with caution, however, because of possible contamination during the collection of the samples or during laboratorial processing and analysis /4–5
When determining lead in a biological matrix, fat content in milk is also a complicating factor because it changes during the course of lactation. A methodological challenge would be to establish the most appropriate time for the milk to be used as a good exposure biomarker. According to Needham and Wang /31
/, for monitoring purposes, mature milk must be collected only when its fat content is set, which happens 2 weeks after birth. Therefore, mature milk would be more appropriate when compared to colostrum, which might explain the greater variability in lead concentrations found in the studies of colostrum presented in .
Lead in breast milk from mothers with current ongoing exposure to lead or mothers exposed by the redistribution of cumulative maternal bone lead stores could pose a potential health risk to the infant. Ettinger et al. /14
/ found, however, that even among a population of women with relatively high cumulative lifetime exposures to lead as assessed by bone lead levels, breast milk levels were low. Nevertheless, breast milk was found to exert a strong influence on infant blood lead levels over and above the influence of maternal blood lead. Thus, the question arises of what exactly is the risk to child health? Maternal blood lead levels < 10 μg L−1
are considered no cause for concern as the amount of lead in breast milk is only ≤5% of that in her blood. Within the range of lead concentrations reported in the studies reviewed here, the effects are expected to be small /14,35
/. Thus, breast-feeding should not be discouraged.