Soy formulas do not appear to estrogenize human infants, but such an effect might be overlooked because the term newborn is already responding to the mother’s circulating estrogen. The responses of the fetal breast and genitalia form part of the Ballard score for estimating gestational age (Ballard et al. 1979, 1991). If soy formula were to prolong physiologic estrogenization, it would not be noticed unless it was striking or specifically sought. The U.S. National Toxicology Program Center for the Evaluation of Risks to Human Reproduction Expert Panel recently reviewed the existing human data on the developmental toxicity of soy formula. The panel found the data insufficient to conclude whether soy formula would produce reproductive toxicity. They also noted that the existing human data were “limited by small samples sizes, conflicting results, and historical studies that used soy infant formula of different composition than the formula being sold currently” (Rozman et al. 2006).

In designing a study to test the hypothesis that soy formula prolonged physiologic estrogenization, we found the natural history of estrogen effects ex utero not sufficiently well described. We thus undertook a pilot study of breast and genital anatomy, milk secretion, and vaginal cytology over the first 6 months of life, which we report here. The purpose of this study was explicitly to examine methods and to assess which of the characteristics examined might exhibit declines from younger to older children, a pattern consistent with response to the withdrawal of maternal estrogen.

Children were recruited equally among the three feeding regimens and seven age groups (Table 1). Children recruited into the breast-feeding category at older ages (3–6 months) were allowed supplementation with cow’s milk formula; however data on the proportion of children who were “pure” breast-fed versus those who received supplements of cow’s milk were not collected.

Assessment of breast bud diameter using beads appeared valid and repeatable. For the validity of breast bud assessment, one examiner estimated the diameter of the hidden bead correctly in 40 of 45 trials (89%) and was off by at most one bead (2 mm); the other estimated correctly in 35 of 45 trials (78%) and was also off by at most one bead. Accuracy increased with experience; the rates of correct estimation in the second session were 93% and 87%, respectively. We also assessed repeatability through triplicate measurements of each breast. Although necessarily contemporaneous, the triplicate measurements on a single breast never varied (75 breasts with buds present).

Average breast bud diameter did not differ between left and right breasts. Maximum difference in an individual was 4 mm in girls and 15 mm in boys. All 12 subjects < 2 days old had palpable buds in both breasts. Breast bud diameter was also maximal within 1 week after birth and was smaller in older children (Figure 1); boys and girls had the same average diameter (Table 2). The proportion of children with palpable buds was smaller in older children (trend p < 0.0001 and 0.0003 for boys and girls, respectively); at 1 month or after, only 5 of 20 girls (25%) and 1 of 20 boys (5%) had a palpable breast bud.

Adipose tissue was palpable on at least one breast of every child in every age interval. The proportion palpable on both breasts was smaller in older children, however (trend p < 0.0001 and 0.0005 in boys and girls, respectively). Few children had detectable milk secretion at the nipple: 9 of 44 boys (20%) and 5 of 44 girls (11%). All children with detectable breast secretion were older than 2 days but younger than 42 days.

Female genitalia evidenced some change in appearance across age intervals but showed no evidence of postpartum regression. The proportion of girls with detectible vaginal discharge was constant across age intervals; 38 of 44 girls (86%) exhibited vaginal discharge overall, but no blood was observed in any vaginal discharge.

We collected evaluable vaginal wall cell samples from all 44 visits. The maturation index of vaginal wall cells was maximal in newborns, lowest in the 1-month olds, and somewhat higher in the older girls (Figure 2). The curvature was statistically significant (p < 0.0001). In addition, feeding regimen appeared to influence the trajectory of maturation index (p = 0.07). Differences among feeding regimens appeared minimal among girls younger than about 30 days; among girls older than that, however, girls fed soy formula tended to have higher maturation index.

Not all tissues that show developmental response to estrogen reverted when maternal estrogen was withdrawn at birth. Although breast tissue was maximal in newborns and smaller in older children, external genitalia showed few differences by age. Vaginal bleeding, from withdrawal of maternal estrogen’s trophic effect on the infant’s endometrium, was not observed. Infants synthesize some estrogen, and it may mitigate the effects of withdrawal of maternal estrogen. Vaginal wall cells are responsive to maternal estrogen, appear to lose that effect by 1 month, and then may respond both to estrogen synthesized by the girl and to soy formula. We looked for milk secretion in infants because we thought exogenous estrogen might inhibit prolactin’s promotion of milk synthesis (Guyda and Friesen 1973), but milk secretion was rare. We propose that a longitudinal set of physical examinations of girls aimed at detecting the effects of exogenous estrogen should be performed weekly to fortnightly for the first 1–2 months. Vaginal cytology specimens need to be collected for > 6 months. Milk secretion or vaginal bleeding should be asked about because they have a plausible endocrine basis, but will be infrequent.

Our physical findings, like those of others (Alexander et al. 1992; Donovan et al. 2007), show more variability than Ballard reported (Ballard et al. 1979, 1991) in term infants. Ballard reports labia majora cover the clitoris and minora at term, which we saw in all girls older than 17 days but only in half of the younger girls. Ballard saw deep scrotal rugae at term, but most of our boys had “good” rugae, only one had “deep” rugae.

We measured testicular volume in infants 48 hr to 6 months of age using orchidometer beads that were smaller and in smaller increments (0.5 mL) than the norm. Mean testicular volume (0.7 mL) in our younger age intervals was less than others report, but was similar by 6 months of age (1.04 mL). Average testicular volume from tissue sections in 1-year-olds was 1.1 mL (range 0.3–1.9 mL) in autopsy material from 36 Danish children who died accidentally (Müller and Skakkebæk 1983). Orchidometric assessment in Taiwanese newborns showed mean volume of 1.3 ± 0.3 mL (Chin et al. 1998). Chemes (2001) and Sak et al. (1993) reported infant testicular volumes of approximately 1 mL. In a small but longitudinal study, Cassorla et al. (1981) measured testicular volumes with orchidometer beads of the same sizes as ours monthly in 10 U.S. boys from birth to 6 months of age. Volume increased from birth (1.1 mL) to 2–3 months (2.0 mL), and then declined (1.5 mL at 6 months) (Cassorla et al. 1981). The increase is plausibly related to the reported peak in testosterone at 2 months (Grumbach 2005). Assuming the peak occurs regularly, detecting it may require the smaller variability in longitudinal data, and will require monthly, or more frequent, exams.

Ballard reported that term infants of both sexes had breast buds of 5–10 mm diameter and palpable adipose breast tissue (Ballard et al. 1979, 1991). Among 21 term infants followed by McKiernan and Hull (1981), mean breast bud diameter was 9 mm at birth, 13 mm at 2 weeks, and receded to 10 mm at 1 month, with breast buds in girls larger than in boys. We did not observe an increase in breast bud size postnatally nor a sex differential, and can offer no explanation for these differences. Among 1,126 Danish infants with median age of 3 months, 79% had breast buds, girls more often than boys; buds were larger in girls (5.7 vs. 4.9 mm) and associated with endogenous estradiol levels (Schmidt et al. 2002). McKiernan and Hull (1981) observed that most infants still had palpable adipose breast tissue at 6 months, as did ours, and that most 1-week-old term infants had some milk secretion, but ours did not.

We know of no other studies that report vaginal maturation index from vaginal wall cells in infants, although it can be used both in the diagnosis and the evaluation of treatment for precocious puberty in girls (Comite et al. 1981), and should have a shorter response time than, say, breast size. The components of vaginal maturation index are usually used as a measure of estrogen effects in adult women (Hammond 1977; van der Laak et al. 1999). In a “normal child,” one would expect almost 100% parabasal cells, few intermediate, and 0% superficial cells, for a vaginal maturation index of nearly 0%, whereas a peripubertal girl may have a virtually no parabasal, 90% intermediate, and about 10% superficial for a vaginal maturation index of about 55%. Infant girls’ cells responded as expected to their in utero exposure to maternal estrogen (Farage and Maibach 2006); at 1 month, this effect was not apparent. Older girls had higher vaginal maturation index values, which may represent the effect of their own synthesis of estrogen. Older soy-fed girls had somewhat higher vaginal maturation index values than girls of the same age on other diets. This observation, though perhaps a chance occurrence, is plausible and deserving of follow-up. We recommend use of vaginal maturation index in infant girls. A possible parallel measure in boys, called the urocytogram (Preeyasomba and Kenny 1966), exploits the hormonal responsiveness of urethral cells from spun urine. We did not evaluate this end point but now would consider using it.