The initial Vanguard Study of the National Children's Study was conducted during 2009-2010 in 7 locations in the United States. A goal was to evaluate the feasibility and yield of a household-based sampling design to recruit pregnant women. A multistage area probability sampling design was used to identify study locations (generally, counties) that were subsequently divided into smaller geographical units, termed segments. Between 7 and 18 segments were selected in each location, and dwelling units within segments were listed. A household-based recruitment process was implemented, which included enumeration of households to identify age-eligible women, pregnancy screening to identify pregnant women eligible for immediate enrollment and nonpregnant women for telephone follow-up, and administration of informed consent to eligible women. After a recruitment period of 17-20 months, 67,181 (89%) households were enumerated, which identified 34,172 (88%) age-eligible women to whom the pregnancy screener was administered. Among those who completed the screener, 2,285 women became eligible for enrollment, of whom 1,399 (61%) enrolled. Although response rates were fairly high at initial contact and among pregnant women, the overall yield was lower than anticipated. In particular, telephone follow-up of nonpregnant women was not a practicable strategy for prospective recruitment of newly pregnant women.