It has been proposed that the intrauterine hormonal milieu is an important determinant of the risk for breast cancer in adult life. 1 Two main arguments for such an association are that estrogens are important as promoters of breast cancer and that estrogen levels increase up to ten times during pregnancy. 1 This hypothesis has been tested in both animal and epidemiologic studies. Offspring of rats fed a high-fat diet during pregnancy are more susceptible to developing breast tumors. 2 Several analytic epidemiologic studies that used birth weight as a proxy variable for the intrauterine hormonal milieu supported the hypothesis. 3–6
It has been assumed, however, rather than demonstrated, that birth weight serves as a good substitute for the intrauterine hormonal milieu. Although some studies have found that pregnancy estrogen levels increase with birth weight, data examining the strength of the association are still lacking at present. 7–11
Of the three major pregnancy estrogens, estrone, estriol, and estradiol, estriol constitutes about 90% of the estrogens produced during pregnancy, 12 and, as opposed to estrone and estradiol, estriol reaches higher levels in the fetus than it does in the mother. 13 Estriol is therefore a good marker of fetal estrogen exposure. Several maternal factors, such as age, parity, and smoking habits are reported to influence both birth weight and pregnancy estrogen levels. 14–16 A possible association between birth weight and pregnancy estriol levels may therefore be mediated through such maternal factors.
To assess the relations among pregnancy estriol levels, maternal characteristics, infant birth weight, and placental weight, we analyzed data from a Swedish-Norwegian prospective study on risk factors for being born small for gestational age (SGA).
Subjects and Methods
The subjects in our study consisted of multiparous women in the three Scandinavian cities of Bergen, Trondheim, and Uppsala who had a singleton pregnancy and had registered for antenatal care before the 20th week of gestation from January 1986 through March 1988. All were fluent in Swedish or Norwegian. Of 5,722 eligible women, we selected a 10% random sample (N = 561) for a more detailed follow-up, which included repeated clinical examinations; interviews; and blood samples taken at 17, 25, 33, and 37 weeks of gestation. We divided the other 90% into two groups; one group was defined as being at high risk for having SGA children, and the other group included the remaining women. The high-risk group (N = 1384) received the same scrutiny as the 10% random sample. We defined maternal age as completed years at delivery. We collected information about smoking at the 17th week of gestation and stratified the women into smokers and nonsmokers depending on whether they reported daily smoking or not at this interview. We collected information about birth characteristics from birth records that were routinely completed by physicians and midwives at delivery. We calculated ponderal index as 100 × birth weight (gm) per birth length cubed (cm3). We used the first day of the last menstrual cycle to estimate gestational age, with a sonogram of the fetal biparietal diameter performed at the first visit available to validate this date. If no date was remembered, or if it was discrepant by more than 14 days, we used the sonogram data to calculate gestational age. We classified birth weight for gestational age as SGA, appropriate for gestational age, and large for gestational age, with SGA infants defined as those with birth weights more than 2 standard deviations below the mean birth weight for gestational age, according to a Swedish reference curve. 17 Infants appropriate and large for gestational age were defined as those within and those above 2 standard deviations, respectively. Further details of the study have been published elsewhere. 18,19
We randomly selected 127 from the 10% random sample, and a subgroup of women having infants with birth weight less than the 15th percentile, for hormonal analyses of the blood samples drawn during pregnancy. This procedure yielded 234 women, 136 of whom were originally in the 10% random sample and 98 of whom were originally in the group defined as at high risk for an SGA birth. The mean gestational age at delivery for the two groups was 39.8 and 39.7 weeks, respectively; of all 234 women, 10 were assessed once, 35 twice, 84 three times, and 105 four times. One observation was excluded from the subsequent analysis owing to erratic values. We restricted our study to the 188 mothers whose estriol levels were assessed at least three times during pregnancy. This left us with a range of birth weights sufficient for the subsequent analysis.
The blood samples were kept frozen at –70°C and were subjected to analysis only after the study was completed. Estriol in serum was assayed with a reagent kit from Amersham International UK [subsequently Kodak Clinical Diagnostics Ltd, Amersham UK, Oestriol (total) II radioimmunoassay kit]). The assay measures both free and glucuronic acid conjugates of estriol, and the assay precisions, expressed as within- and between-assay coefficients of variation, were 2.6% and 4.3%, respectively. All of the serum samples from each patient were analyzed in the same series and assayed twice. The recorded results are the mean values between each duplicate set.
We estimated the individual total estriol load during pregnancy by the area under the curve (AUC) for the estriol levels over time. The general form these curves followed was an exponential function: MATH 1 We estimated the parameters of the exponential function by ordinary linear regression on the log-estriol values for all 188 subjects with three or more estriol determinations. Because the infants were born up to 6 weeks after the last estriol assessment, we corrected the AUC by assuming a constant estriol level between week 37 and parturition (PT):MATH 2 We validated this modeling procedure by assuming a piecewise linear function for the estriol levels of the mothers for whom we had all 4 estriol assessments. The AUCs calculated were essentially the same for both models. Use of the exponential model thus enabled us to include the women with only three estriol assessments in our analyses (Figure 1).
Because the calculated total estriol load [corrected value for AUC (cAUC)] distribution was skewed, we based all statistical analyses on logarithmically transformed cAUC values. The cAUC was analyzed by one-way analysis of variance with the explanatory variables categorized. Additionally, we performed a multiple linear regression analysis to consider all of the explanatory variables and interactions among them simultaneously.
Mean estriol values consistently increased with gestational age (Table 1). Maternal age affected estriol levels only marginally. Among smokers, compared with nonsmokers, there was a 20% reduction in mean estriol levels at the first estriol assessment (week 17), and this difference increased to 25% and 30% later. The difference between smokers and nonsmokers did not increase with increasing number of cigarettes smoked (data not shown). Mean estriol levels increased with augmenting birth weight category. In the lowest birth weight category (<2,500 gm), mean estriol level in week 37 was 392 nmol/liter, compared with 775 nmol/liter in the highest birth weight category (>4,500 gm). There was also a clear dose-response relation between birth weight for gestational age and estriol levels during pregnancy. Ponderal index category was also, but to a lesser extent, related to mean estriol levels. A similar relation was seen between placental weight and mean estriol levels. The estriol levels were assessed in the 17th, 25th, 33rd, and 37th weeks of gestation, and there was no increase in mean estriol levels at these points in time with increasing gestational age at delivery (data not shown). The variability of the measurements was substantial, however (Table 1).
We then calculated the geometric mean cAUC for the infants when grouped according to maternal age, maternal smoking, birth weight, birth weight for gestational age, ponderal index, and placental weight, respectively (Table 2). The category limits for birth weight, ponderal index, and placental weight were determined to give equally sized groups among all women participating in the cohort study (N = 1945), regardless of whether their blood samples were being analyzed. For children weighing more than 4.5 kg at birth, the mean cAUC was double that of the point estimate for children weighing less than 2.5 kg. There was a 28% reduction in mean cAUC among the smokers as compared with the nonsmokers. Increasing placental weight category was associated with increasing cAUC, and the same pattern was observed to a lesser extent for ponderal index. No trend could be detected with augmenting maternal age category.
We entered the variables placental weight, birth weight, ponderal index, maternal age, and maternal smoking into a linear regression model. Smoking status and birth weight were independently associated with cAUC. The relative impact on cAUC was 0.8 for smoking (smokers vs nonsmokers; 95% confidence interval = 0.72–0.90) and 1.2 per kilogram for birth weight (95% confidence interval = 1.1–1.3). When birth weight was in the model, placental weight and ponderal index did not add any explanatory power to the equation.
The formula derived from the linear regression analysis with log-cAUC as the dependent variable and maternal smoking and infant birth weight as independent variables was cAUC =e8.1+[0.18×infant birth weight (kg)] – [0.21×Smoking (No/Yes)]. We used this formula to evaluate the agreement between birth weight and smoking as predictors of antenatal estriol and observed antenatal estriol levels (Table 3). The sensitivity and specificity in predicting the highest quartile of estriol exposure were 13/47 = 0.28 and 131/141 = 0.93, respectively, whereas the sensitivity and specificity in predicting the lowest estriol exposure were 11/47 = 0.23 and 132/141 = 0.94.
There was a strong relation between birth weight category and mean estriol levels. In the 37th week of gestation, the mean estriol levels were twice as high in the mothers who eventually gave birth to infants in the highest birth weight category as compared with those of the lowest category. When gestational age was taken into account, the association between birth weight and mean estriol levels remained unchanged. The association between ponderal index category and mean estriol level was less pronounced. Because ponderal index is a ratio with one aspect of growth (that is, length) in the denominator, this is not surprising. The increased estriol levels among women ages 20–24 years that have been reported previously could not be confirmed from our data. 14 The mean estriol reduction of 20–30% among smokers throughout the whole pregnancy has not been shown previously, although a negative association between smoking and pregnancy estrogen levels has been reported in other studies. 15,20 We use smoking information collected at week 17. Because pregnant women in Scandinavia are strongly encouraged not to smoke, it is likely that some of the mothers subsequently stopped. 21 If so, we may have underestimated the reduction of estriol levels among the smokers.
Although no conclusions can be drawn from this study on an individual level, our findings show the strong impact of smoking on the intrauterine environment during pregnancy. In our study, we did not find a dose-response relation between number of cigarettes smoked and mean estriol levels. Whether this reflects a lack of precision in self-reported data on smoking or the lack of a true biological relation is not clear.
The AUC is an estimate of total estriol load during pregnancy, and our findings indicate an even stronger relation between birth weight and estriol than can be shown by individual measurements alone. With the exception of the lowest birth weight category with only four observations, every step up in birth weight category had a cAUC mean with confidence limits at the 95% level that were separated from the confidence limits of the surrounding steps. The increase in mean estriol AUC with stepwise increase in ponderal index and placental weight was not as evident. Birth weight, ponderal index, and placental weight are all related to each other, and taken one by one they are all correlated with cAUC. When put together in a regression analysis, ponderal index and placental weight added little to what was provided by birth weight alone.
The model used for determining the estriol dose over time included adding the product of the 37th week estriol value and the time between the 37th week and parturition. This probably led to an underestimation of the true estriol exposure, because estriol concentrations have been reported to continue to increase throughout pregnancy. 13 However, as we lack knowledge of the extent of the increase in estriol levels after the 37th week, we have chosen the more conservative model.
In conclusion, there is a strong relation between estriol levels and birth weight, as well as placental weight. On an individual level, birth weight may not be a good predictor of the gestational estriol exposure, but on an aggregated level, birth weight and maternal estriol are related in a dose-response-like manner. Furthermore, there is an even stronger relation between the total estriol load during pregnancy and birth weight, which gives credence to birth weight as a good proxy variable for intrauterine exposure of estriol in epidemiologic studies.
1. Trichopoulos D. Hypothesis: does breast cancer originate in utero
? [see comments]. Lancet 1990; 335:939–940.
2. Hilakivi-Clarke L, Clarke R, Onojafe I, Raygada M, Cho E, Lippman M. A maternal diet high in n-6-polyunsaturated fats alters mammary gland development, puberty onset, and breast cancer risk among female rat offspring. Proc Natl Acad Sci USA 1997; 94:9372–9377.
3. Ekbom A, Hsieh Cc, Lipworth L, Adami HQ, Trichopoulos D. Intrauterine environment and breast cancer risk in women: a population-based study. J Natl Cancer Inst 1997; 89:71–76.
4. Ekbom A, Trichopoulos D, Adami HO, Hsieh CC, Lan SJ. Evidence of prenatal influences on breast cancer risk [see comments]. Lancet 1992; 340:1015–1018.
5. Michels KB, Trichopoulos D, Robins JM, Rosner BA, Manson JE, Hunter DJ, Colditz GA, Hankinson SE, Speizer FE, Willett WC. Birthweight as a risk factor for breast cancer [see comments]. Lancet 1996; 348:1542–1546.
6. Sanderson M, Williams MA, Malone KE, Stanford JL, Emanuel I, White E, Daling JR. Perinatal factors and risk of breast cancer. Epidemiology 1996; 7:34–37.
7. Gerhard I, Fitzer C, Klinga K, Rahman N, Runnebaum B. Estrogen screening in evaluation of fetal outcome and infant’s development. J Perinat Med 1986; 14:279–291.
8. Gerhard I, Vollmar B, Runnebaum B, Klinga K, Haller U, Kubli F. Weight percentile at birth. II. Prediction by endocrinological and sonographic measurements. Eur J Obstet Gynecol Reprod Biol 1987; 26:313–328.
9. Petridou E, Katsouyanni K, Spanos E, Skalkidis Y, Panagiotopoulou K, Trichopoulos D. Pregnancy estrogens in relation to coffee and alcohol intake. Ann Epidemiol 1992; 2:241–247.
10. Hardy MJ, Humeida AK, Bahijri SM, Basalamah AH. Late third trimester unconjugated serum oestriol levels in normal and hypertensive pregnancy: relation to birth weight. Br J Obstet Gynaecol 1981; 88:976–982.
11. Chew PC, Ratnam SS, Salmon JA. Plasma oestriol in normal pregnancy in an Asian population. Br J Obstet Gynaecol 1976; 83:430–433.
12. Pauerstein CJ. Clinical obstetrics. New York: John Wiley and Sons, 1987.
13. Speroff L, Glass RH, Kase NG. Clinical Gynecologic Endocrinology and Infertility. 5th ed. Baltimore: Williams & Wilkins, 1994.
14. Panagiotopoulou K, Katsouyanni K, Petridou E, Garas Y, Tzonou A, Trichopoulos D. Maternal age, parity, and pregnancy estrogens. Cancer Causes Control 1990; 1:119–124.
15. Petridou E, Panagiotopoulou K, Katsouyanni K, Spanos E, Trichopoulos D. Tobacco smoking, pregnancy estrogens, and birth weight. Epidemiology 1990; 1:247–250.
16. Petridou E, Trichopoulos D, Revinthi K, Tong D, Papathoma E. Modulation of birthweight through gestational age and fetal growth. Child Care Health Dev 1996; 22:37–53.
17. Marsal K, Persson PH, Larsen T, Lilja H, Selbing A, Sultan B. Intrauterine growth curves based on ultrasonically estimated foetal weights. Acta Paediatr 1996; 85:843–848.
18. Markestad T, Bergsjo P, Aakvaag A, Lie RT, Jacobsen G, Hoffman HJ, Bakketeig LS. Prediction of fetal growth based on maternal serum concentrations of human chorionic gonadotropin, human placental lactogen and estriol. Acta Obstet Gynecol Scand Suppl 1997; 165:50–55.
19. Bakketeig LS, Jacobsen G, Hoffman HJ, Lindmark G, Bergsjo P, Molne K, Rodsten J. Pre-pregnancy risk factors of small-for-gestational age births among parous women in Scandinavia. Acta Obstet Gynecol Scand 1993; 72:273–279.
20. Mochizuki M, Maruo T, Masuko K, Ohtsu T. Effects of smoking on fetoplacental-maternal system during pregnancy. Am J Obstet Gynecol 1984; 149:413–420.
21. Thue E, Schei B, Jacobsen G. Psychosocial factors and heavy smoking during pregnancy among parous Scandinavian women. Scand J Prim Health Care 1995; 13:182–187.