A cross-sectional study of newly-delivered male infants and their mothers was conducted in 2002–2003 in Tapachula, a city in the state of Chiapas, México. Women were recruited during the postpartum period at both of the city’s hospitals, which also serve the surrounding area. About fifty percent of births in Tapachula occur in these hospitals (14
). Women at the Social Security Hospital were more frequently urban, and had, on average, higher socioeconomic status and more education than women at the General Hospital (data not shown). If the eligibility criteria were met, the mothers were invited to participate and sign an informed consent form. The study protocol was approved by Institutional Review Boards at the Instituto Nacional de Salud Pública (National Institute of Public Health) in México and the National Institute of Environmental Health Sciences in the United States.
The eligibility criteria were chosen to exclude subjects for whom complicating medical conditions might have affected anogenital distance in male offspring or our ability to measure it. These criteria were determined a priori
, before any data or determinants of anogenital distance were available. Exclusion criteria for the mother were age greater than 35 years; pre-eclampsia or pregnancy-related diabetes or hypertension; seizure disorders requiring daily medication; history of repeated urinary tract infections; psychiatric, kidney, or cardiac disease; and non-speaker of Spanish. Infants were excluded if female, gestational age at delivery as estimated by the Capurro scale (15
) or the medical record (based on last menstrual period) was < 36 weeks, birthweight was < 2,500 g, pregnancy was not singleton, Apgar score at 5 minutes was ≤ 6, or child was admitted to the neonatal intensive care unit. Of the subjects who were invited to participate, 95 percent did so, resulting in 872 mother-infant pairs. Of these, the first 91 were enrolled when a preliminary anthropometric measurement protocol was in place. We excluded these from the present analysis because their measurements were not comparable, leaving 781 observations.
A questionnaire was administered to the mothers about socio-demographic characteristics, reproductive history, maternal health status, and various exposures. Maternal serum DDE and DDT was quantitated after solid phase extraction (C18
column purification), using gas chromatography and mass spectrometry (16
). For DDE the limit of detection was 0.2 μg/L; the recovery was 97 percent; and the between-assay coefficient of variation at 10 μg/L was seven percent. For DDT the limit of detection was 0.2 μg/L; the recovery was 97 percent; and the between-assay coefficient of variation at 2.5 μg/L was six percent. Total serum lipid was estimated based on levels of serum cholesterol and triglycerides (18
), which were measured using standard enzymatic methods. Measurements of weight and height were performed on the mothers and newborns. In addition, we measured infant anogenital distance and penis size.
The technique for measurement of anogenital distance and penis size has been described in detail elsewhere (11
) and is summarized briefly here. Three measures of anogenital distance were taken: anterior base of penis to anus (AGD1), posterior base of penis to anus (AGD2), and posterior of scrotum to anus (ASD) (see ). In addition, we measured penile width (PW) and stretched penile length (PL). The anogenital and PW measurements were performed using Swiss Precision Cali Max Vernier calipers, from Bel-Art Products, Pequannock, NJ, USA. The calipers were read in increments of 1 mm. The penile measurements were done when the newborn’s penis was flaccid. Each measurement was taken on two occasions; the first set of readings was recorded in the questionnaire, and after these were completed, the second set was taken and noted on a sheet that was attached to each subject’s file. On each of these two occasions, PL was usually measured in duplicate, yielding up to four recorded values. Over 80 percent of children were examined before they were 6 hours old. With one exception, all examinations were conducted before 34 hours of age; the remaining examination was conducted 7 days after birth. The anthropometrists received special training before measuring anogenital and penile dimensions, length, and weight, and they received periodic retraining during the study (equipment for length and weight measures have been reported previously (11
)). Twenty-two anthropometrists participated in the study. The reliabilities of the measures (the fraction of the variability that is true variability rather than measurement variability) were: AGD1, 0.91; AGD2, 0.88; ASD, 0.85; PW, 0.77; and PL, 0.76 (11
). The small variation in measures among replicates, and due to observers, is described in detail elsewhere (11
A) Anterior of penis to center of anus distance (AGD1), B) Posterior of penis to center of anus distance (AGD2), C) Posterior of scrotum to center of anus distance (ASD)
All 781 infants had measurements of each of the three anogenital distances; in all but two (AGD1 and AGD2) or three (ASD) cases, duplicate measurements were taken. All infants also had PW available (in duplicate in all but three cases). For 541 infants, all four PL measurements were available. Of the remainder, 43 infants had two measurements on the first and one on the second occasion, 188 had two measurements on the first and none on the second, three had one on each occasion, five had only one on the first occasion, and one had no measurements.
The goal of the analysis was to test the hypothesis that maternal DDE concentration is associated with anogenital distances and penile measurements in newborn male infants. Significance tests were based on F-tests in linear regression models, testing the hypotheses that the coefficient(s) were zero. All tests were two-sided.
Means of the replicates were used for all analyses. However, subjects were excluded from analyses of a particular anthropometric measurement if replicates differed by 30 percent or more; there were two such cases for ASD, one for PW, and three for PL. In the primary analysis, DDE was expressed on a per gram serum lipid basis, and subjects were placed in categories defined by < 3 μg/g, 3-<6 μg/g, 6-<9 μg/g, and ≥ 9 μg/g. These categories were chosen without examination of outcomes; the top category was chosen to contain a reasonable fraction of the children, with the remaining categories being of equal width. Finer categories (width of 1 ug/g rather than 3 ug/g, with the top categories collapsed to contain at least 20 children), were examined subsequently (see below). Analyses were also conducted with DDE concentration or logarithm of DDE concentration included as a linear term in the models.
When examining the relationship of anogenital and penile dimensions to DDE, we looked at both crude and adjusted relationships. Adjustment factors were those items that had previously been shown to be related to the anthropometric measurements in these data (11
); potential predictors examined included infant birth weight and length, gestational age, maternal height and prepregnancy body mass index, maternal age, parity, maternal education, maternal marital status, household income per capita, urban or rural residence, and hospital. Those that were significant at p<0.20 in models including all non-DDE predictors were included in the present models (19
). For anogenital distances, these factors were birth weight (included as a linear term), gestational age (categorized as 36–37, 38, 39, 40, 41+ weeks), urban vs rural residence, and hospital. For penile measurements, they were birth weight (linear), maternal age (linear), maternal height (linear), and parity (categorized as 1, 2–3, 4+). In both cases, we also adjusted for anthropometrist as a random effect, to account for observer-to-observer variability. Linear regression models were fitted using the MIXED procedure in SAS, version 9.00 (SAS Institute, Inc., Cary, North Carolina). All categorical variables were modeled with indicator variables.
Levels of DDT and DDE were highly correlated (Spearman r = 0.84), so we avoided fitting models that included both terms. However, we also conducted analyses for DDT similar to those described above, and in addition examined models that included DDE and the ratio of DDT to DDE.