From a total of 382 visits by 165 children, we collected 381 urine samples, 359 saliva samples, and 88 blood samples. For example, among 165 children contributing urine samples, 66 were in once, 28 twice, 25 three times, and 46 four times (). We collected saliva from 152 children at 359 visits (but we have data on 361 sample aliquots as two visits (samples) were represented by two aliquots each). We have fewer saliva samples than urine samples because of the difficulty collecting a sufficient volume of saliva from the youngest, particularly breast-fed children. Expecting blood draw to be unacceptable to many parents, we designed for 62 samples but collected 88 blood samples from 72 children. Because the blood-draw procedure turned out to be acceptable, we substituted a blood sample for a saliva sample for some infants who failed to produce enough saliva.

| **Table 3**Number of infants with single or multiple visits by biological matrix. |

Most samples could be analyzed for all three analytes; however, only 77 of 88 blood samples were adequate for analysis, and equol urinary concentrations could not be estimated for four urine samples because of a chromatographic interference. For these urine samples, the ratio between quantitation and confirmation ions for equol was outside the normal ranges. Therefore, these equol results, which did not fulfill the QA/QC criteria, could not be reported. Nevertheless, many isoflavone concentrations were censored at the method’s limit of detection (). Concentrations of equol, which requires metabolism of daidzein by the intestinal microflora, were below the limit of detection in 100% of blood and saliva samples for all three feeding groups. We observed detectable concentrations of equol in only 35 (9%) of 377 urine samples: two from breast-fed children, 27 from cow-milk-formula-fed children, and six from soy-formula-fed children. These 35 samples represented 26 children who ranged in age from two to 377 days. Because of the high proportion of non-detectable values, we do not consider equol further.

| **Table 4**Proportion of samples from each matrix and feeding method where concentrations of daidzein, genistein, or equol were below the limit of detection. |

Concentrations of daidzein were undetectable in blood or saliva for at least 93% of samples from children fed breast milk or cow-milk formula but was undetectable in fewer than 17% of samples from children fed soy formula (). Similarly, genistein concentrations were undetectable in blood or saliva for at least 90% of samples from children fed breast milk or cow-milk formula but undetectable for fewer than 9% of samples from children fed soy formula. The proportion on non-detectable values was somewhat lower in urine than the other matrices. Among breast-fed children, concentrations of daidzein and genistein were non-detectable in 70% and 51% of samples, respectively; among cow-milk-formula-fed children the corresponding proportions were 22% and 9%, and among soy-formula fed children, daidzein and genistein concentrations exceeded the limit of detection in every sample. Because of non-detectable values, we compare the feeding groups using only urinary concentration data, and we compare urine, blood, and saliva using only the soy-formula-fed group.

In general, based on median values (), urinary concentrations of genistein and daidzein were about 500 times higher in the soy-formula-fed infants than in the cow-milk-formula-fed infants. Based on 75^{th} percentiles (medians were below the limit of detection; ), urinary concentrations of genistein and daidzein were about six to ten times higher in infants fed cow-milk formula than in infants fed breast milk. Genistein concentrations in urine were relatively constant through time for all three feeding methods (). We observed a slight decrease in urinary genistein concentrations with age among the soy-formula-fed infants but a slight increase with age among infants in the other two feeding groups (). Genistein concentrations were highly correlated with daidzein concentrations in all three matrices (); and concentration trajectories of daidzein exhibited the same weak temporal patterns as genistein (data not shown). These patterns are consistent with the expectation that older infants in each feeding group are beginning to have more varied diets and no longer consume exclusively soy or exclusively non-soy diets.

| **Table 5**Quartiles of daidzein and genistein concentrations among samples from each matrix and feeding method. |

We compared isoflavone concentrations between blood, saliva, and urine using only the soy-formula-fed infants to minimize potential biases from censored data. For both daidzein and genistein, urine had the highest median concentrations, followed by blood, and saliva had the lowest median concentrations (). For genistein, the urine:blood:saliva median concentration ratios were about 900:40:1; for daidzein, the corresponding ratios were about 660:80:1. All three matrices showed a slightly decreasing trend in genistein concentration with age (, ); the same pattern holds for daidzein (data not shown).

We examined the correlation of isoflavone concentrations between matrices and again restricted attention to infants fed soy formula ( and ). Our design which limited the number of blood samples available provides the most information about the saliva-urine correlations and less about correlations involving blood. Generally speaking, urine concentrations were more strongly correlated with blood or saliva concentrations for genistein than for daidzein. This difference may be because the half-life of blood genistein (8.4h) was longer than that of daidzein (5.8h).(

Watanabe *et al.*, 1998) The blood-saliva correlations were, however, comparable for both analytes (). Nevertheless, the correlations between matrices for either analyte were strikingly lower that the correlation between the two analytes in the single matrix (compare and ).

| **Table 6**Correlations (Kendall’s tau-b) of genistein or daidzein levels between matrices for samples from soy-fed infants (number of paired samples). |

We also looked for relationships between isoflavone concentrations and the levels of sex hormone binding globulin and certain hormones (estrone, estradiol, testosterone) and gonadotropins (LH, FSH) in children fed soy formula. We found no significant correlations, either separately by sex or combined. The relationship between genistein and testosterone () was typical of those between the isoflavones and hormones.