We used an interrupted time series design16,17
to estimate the impact of the advisory on total fish intake as well as consumption of dark meat fish, canned tuna, shellfish, and white meat fish individually. We divided the study time into three periods. The preadvisory period ran from April 1999 through December 2000. The advisory was published in January 2001. We allowed a 3-month “wash-in” period, January–March 2001, because dietary questionnaires asked about consumption up to 3 months before the survey date. The postadvisory period ran from April 2001 through February 2002.
Study subjects were participants in Project Viva, a prospective cohort study of gestational diet, pregnancy outcomes, and offspring health. We recruited women attending their initial prenatal visit at eight urban and suburban obstetric offices of a multispecialty group practice in eastern Massachusetts. Eligibility criteria included fluency in English, gestational age less than 22 weeks at the initial prenatal clinical appointment, and singleton pregnancy. We enrolled 2609 women (63% of those eligible) between April 23, 1999, and March 1, 2002, of whom subsequently 145 women became ineligible because of multiple gestation (n = 6), transfer of obstetric care to a nonstudy site (n = 49), or they were no longer pregnant (n = 80). This analysis is limited to 2235 of the 2464 remaining subjects (91%) who completed at least one of the three dietary questionnaires during the pre- or postadvisory periods.
After obtaining informed consent, we collected demographic and health history information by interview and self-administered questionnaire. We used a combination of incentives, mail and phone follow-up, and outreach within the clinical offices to promote questionnaire completion. Institutional review boards of participating institutions approved the study. All procedures were in accordance with the ethical standards for human experimentation established by the Declaration of Helsinki.18
Participants completed semiquantitative food frequency questionnaires after each trimester of pregnancy. The semiquantitative food frequency questionnaire was modified for use in pregnancy from an extensively validated instrument.19
We previously calibrated the first-trimester questionnaire against maternal blood nutrient levels.20
For every gram of intake per day, red blood cell concentrations of long-chain omega-3 fatty acids were higher by 1.72% among 72 black women (P
= .09) and by 4.79% among 132 white women (P
< .001). In the study population, seafood comprised 85–90% of the dietary source for these omega-3 fatty acids.
The first-trimester semiquantitative food frequency questionnaire, administered at study enrollment, asked about average consumption “during this pregnancy” (ie, since the last menstrual period), and the second semi-quantitative food frequency questionnaire, administered at 26–28 weeks’ gestation, asked about consumption “during the past 3 months.” A limited third-trimester semiquantitative food frequency questionnaire focused on fish intake in the month before delivery. Questions regarding fish queried intake of “canned tuna fish (3–4 oz.),” “shrimp, lobster, scallops, clams (1 serving),” “dark meat fish, e.g. mackerel, salmon, sardines, blue-fish, swordfish (3–5 oz.),” and “other fish, e.g. cod, haddock, halibut (3–5 oz.).” Response options ranged from “never/less than 1 per month” to “1 or more servings per day.” We generated a count of monthly servings as the average within each category; for example, we coded “1–3 servings per month” as two servings per month.
Because trimester-specific analyses yielded similar results, we combined data from all three trimesters. Within each period, we determined subject characteristics (excluding the 170 subjects who completed questionnaires in both periods). We compared means using t
tests and proportions using χ2
tests. In addition, we calculated mean consumption of total fish and of each of the four fish types and determined the percentage of women who consumed more than 13 total fish servings per month (more than three servings per week). We compared consumption in the pre- and postadvisory periods using Poisson regression,21
with period as the predictor. We assumed that all missing data were missing at random.22
Next, we evaluated the effect of the advisory using an interrupted time series analysis. We performed Poisson regression with servings of fish per month as the dependent variable and period, month, and enrollment site as predictors. We accounted for repeated observations within individuals using generalized estimating equations.23
We assessed change over time (slope) in mean fish consumption within the preadvisory period and in the post- versus preadvisory period. We also determined change in level at the beginning of the postadvisory period compared with the end of the preadvisory period. Inclusion of subject race/ethnicity, age, and education did not change estimates, so we did not include these characteristics in the final models.