Maternal Fish Consumption, Mercury Levels, and Risk of Preterm Delivery

Fei Xue; Claudia Holzman; Mohammad Hossein Rahbar; Kay Trosko; Lawrence Fischer


Environ Health Perspect. 2007;115(1):42-47. 

In This Article


At the time of enrollment, 41% of women were < 25 years of age, 42% had ≤ 12 years of education, and 43% were insured by Medicaid, a health-related public assistance program ( Table 1 ). The percentage of women from racial/ethnic backgrounds other than non-Hispanic white and African American was only 9%. The "other" ethnic groups were included with non-Hispanic whites in the final models. Alternative analyses excluding these other groups showed similar results to that of the more inclusive final models.

In approximately the first 6 months of pregnancy, 11% of women in this sample did not eat any fish, 25% ate two fish meals or fewer, and only 50% ate more than nine fish meals. The mean level of total fish consumption was 19.6 meals/6 months, considerably higher than the median (50th percentile = 9.0 meals/6 months), suggesting a right skewing of the distribution ( Table 2 ). Canned fish was the most frequently consumed fish category, with 25% of women eating ≥ 12 meals/6 months, followed by bought fish, with 25% eating ≥ 6 meals/6 months. Only 9.2% of women reported consumption of sport-caught fish during the first 6 months of pregnancy.

Mercury levels in maternal hair ranged from 0.01 to 2.50 µg/g, with a mean of 0.29 µg/g and a median of 0.23 µg/g. Approximately 20% of women had levels > 0.38 µg/g ( Table 3 ). Mercury levels were divided into quintiles and total fish consumption and consumption of each fish category were divided into four levels: 0 meals/6 months (reference group), 1-5 meals/6 months, 6-23 meals/6 months, and ≥ 24 meals/6 months. Women with higher levels of total fish consumption were more likely to have mercury levels in the upper quintiles ( Table 3 ). The mean and median hair mercury levels for the 109 women who did not consume fish during this period in pregnancy were 0.15 µg/g and 0.13 µg/g, respectively. Interestingly, 10% of women who reported not eating fish during pregnancy had hair mercury levels in the 4th and 5th quintiles.

In multicovariate analyses, higher mercury levels were significantly associated with older maternal age (≥ 25 years), white and "other" ethnicity, not being insured by Medicaid, and residing in communities 3, 4, and 5 (Figure 2), even after adjusting for total fish consumption. Maternal mercury levels were not significantly related to gestational week at enrollment or smoking before or during pregnancy. We reevaluated the association between total fish consumption and mercury levels in a model that included maternal covariates related to mercury levels in hair ( Table 4 ). The adjusted mean mercury continued to be significantly higher in women who consumed fish compared with that in women who did not consume fish in the first 6 months of pregnancy, and mean mercury levels increased as levels of fish consumption increased. In another model that included consumption of each fish category along with the other maternal covariates related to mercury levels in this sample, consumption of canned fish, bought fish, and sport-caught fish were each positively associated with mercury levels in maternal hair. Adjustment for gestational week at enrollment did not appreciably alter these associations.

Figure 2.

Mercury levels (95% CI) in hair collected at mid-pregnancy according to maternal characteristics, adjusted for all the other covariates in the figure and total fish consumption. Mercury levels were transformed to a natural log and then transformed back as exp [mean of ln(mercury)]. *p < 0.05.

In the final analyses, we assessed maternal mercury levels at mid-pregnancy in relation to gestational week at delivery. Women who delivered very preterm (< 35 weeks) were more likely to have had hair mercury levels at or above the 90th percentile (0.55-2.5 µg/g) than were women who delivered at term (≥ 37 weeks), even after adjusting for maternal characteristics and total fish consumption [odds ratio (OR) = 3.0; 95% confidence interval (CI), 1.3-6.7] ( Table 5 ). These results remained relatively unchanged in models that adjusted for gestational week at enrollment and fish categories. The association between maternal mercury levels and very preterm delivery was not evident at lower threshold levels of mercury (i.e., quintile cut-points), and mercury levels were not associated with delivery of a moderately preterm infant (35-36 weeks).


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.