Energy-containing beverages: reproductive hormones and ovarian function in the BioCycle Study

Abstract

Background: Energy-containing beverages are widely consumed among premenopausal women, but their association with reproductive hormones is not well understood.

Objective: The objective was to assess the association of energy-containing beverages, added sugars, and total fructose intake with reproductive hormones among ovulatory cycles and sporadic anovulation in healthy premenopausal women.

Design: Women (n = 259) in the BioCycle Study were followed for up to 2 menstrual cycles; they provided fasting blood specimens during up to 8 visits/cycle and four 24-h dietary recalls/cycle.

Results: Women who consumed ≥1 cup (1 cup = 237 mL) sweetened soda/d had 16.3% higher estradiol concentrations compared with women who consumed less sweetened soda (86.5 pg/mL compared with 74.4 pg/mL, P = 0.01) after adjustment for age, BMI, race, dietary factors, and physical activity. Similarly elevated estradiol concentrations were found for ≥1 cup cola/d and noncola soda intake. Neither artificially sweetened soda nor fruit juice intake ≥1 cup/d was significantly associated with reproductive hormones. Added sugar above the average US woman's intake (≥73.2 g/d) or above the 66th percentile in total fructose intake (≥41.5 g/d) was associated with significantly elevated estradiol but not consistently across all models. No associations were found between beverages, added sugars, or total fructose intake and anovulation after multivariate adjustment.

Conclusions: Even at moderate consumption amounts, sweetened soda is associated with elevated follicular estradiol concentrations among premenopausal women but does not appear to affect ovulatory function. Further research into the mechanism driving the association between energy-containing beverages and reproductive hormones, and its potential implications for women's health, is warranted.

INTRODUCTION

Sweetened soda intake, the largest contributor of fructose and added sugar in the US diet, has markedly increased over the past few decades, from approximately two 8-oz (8 oz = 237 mL) servings/wk in 1942 to approximately two 8-oz servings/d in 2000 (1). Women of childbearing age in the United States derive, on average, >23% of their daily energy from beverage sources (2). Due to overwhelming experimental and epidemiologic evidence indicating that excessive consumption of sweetened beverages and added sugars are associated with weight gain, metabolic syndrome, and cardiovascular risk factors (3–5), the American Heart Association (AHA)⁴ recently expanded its recommendation to limit intake of added sugars, defined as sugars and syrups added to foods during processing or preparation and includes sugars and syrups added at the table (6). According to the AHA, a woman who is moderately physically active, aged 19–30 y, should limit her consumption of added sugars to between 8 and 12 teaspoons/d (∼32–48 g/d), equivalent to approximately a 12-oz (12 oz = 355 mL) can of nondiet soda, which is significantly below the current usual intake of 18.3 teaspoons/d (73.2 g/d) for a US woman (6).

Both human and animal studies have shown that diets high in sugars, particularly in liquid form with fructose implicated as the major risk factor, result in dyslipidemia and insulin resistance (7–12), which are well-known risk factors for hormone and ovulatory disorders (13). Research assessing the effect of energy-containing beverages on premenopausal reproductive hormone levels (14–16) and ovulatory function (13), however, is sparse. Two studies have found no association between soda and premenopausal reproductive hormones (15, 16), but inferences are limited by small sample sizes and/or inadequate methods to evaluate the phase of the menstrual cycle when the hormones were measured. Women participating in the Nurses’ Health Study were shown to have greater risk of ovulatory disorder with caffeinated soda intake (13), but no significant results with total caffeine or with caffeinated coffee or tea were found. These findings suggest that relevant components may be specific to soda. The relation between these components and reproductive hormones is of interest for assessing not only their effects on female reproductive function (13) but also for better understanding hormone-related chronic diseases (17, 18). Research has indicated that women with high follicular estradiol concentrations have increased risk of breast cancer (17).

The objective of this study was to determine whether energy-containing beverages, added sugars, and total fructose are related to serum concentrations of reproductive hormones and ovulatory function in a cohort of 259 healthy, premenopausal women by using a standardized method to time serum sample collections according to the phase of the menstrual cycle.

SUBJECTS AND METHODS

Study population

The BioCycle Study, conducted in 2005–2007, followed 259 women from western New York State for up to 2 complete menstrual cycles. The study population, materials, and methods have been previously described in detail (19). In summary, healthy women aged 18–44 y had to be regularly menstruating (self-reported cycle length between 21 and 35 d for each menstrual cycle in the past 6 mo) to participate. Women who reported conditions known to affect menstrual cycle function such as polycystic ovary syndrome, uterine fibroids, or current or recent use of hormonal contraception (ie, 3 mo before study entry) were excluded. Women with previously known ovulatory disorders were excluded, but sporadic anovulation (n = 35 women, 42 cycles) was observed in the study population (20). The University at Buffalo Health Sciences institutional review board approved the study and served as the institutional review board designated by the NIH for this study under a reliance agreement. Written informed consent was obtained from all participants.

Hormone assessment

Women provided fasting blood specimens during up to 8 visits/cycle for 1 (n = 9) or 2 (n = 250) menstrual cycles, with visit timing assisted by the use of fertility monitors to correspond to menstruation, midfollicular, late-follicular, luteinizing hormone (LH)/follicle-stimulating hormone (FSH) surge, ovulation, early luteal, midluteal, and late luteal phases (21). Specifically, study visits were initially scheduled to occur during key phases of the menstrual cycle by using an algorithm accounting for each woman's self-reported cycle length. Fertility monitors (Clearblue Easy Fertility Monitor; Inverness Medical) measured estrone-3-glucuronide and LH in urine, starting on calendar day 6 after menses and continuing for 10–20 d. The monitor algorithm determines “peak fertility” on the basis of predetermined cutoffs during the first cycle, but in subsequent cycles adjusts the cutoff criteria according to the woman's specific hormone concentrations. Monitor indications of low, high, and peak fertility were used to time midcycle visits. Total estradiol, progesterone, LH, FSH, and sex hormone binding globulin (SHBG) were measured via Immulite 2000 Solid Phase competitive chemiluminescent enzymatic immunoassay (CLEIA; Siemens). The lower limits of detection for estradiol, progesterone, and SHBG were 20 pg/mL, 0.2 ng/mL, and 2 ng/mL, respectively. The albumin assay was tested on the Beckman LX20 auto analyzer (Beckman Coulter) by using bromcresol purple methodology. Calculation of free estradiol (ie, bioavailable estradiol) was performed via the equation proposed by Södergård et al (22) by using total estradiol, SHBG, and albumin concentrations. All hormonal analyses were conducted by Kaleida Health Laboratories in Buffalo, NY. Across the study period, the CVs for these tests were as follows: <10% for estradiol, SHBG, and insulin; <5% for LH, FSH, and albumin; and <14% for progesterone. Insulin resistance was calculated by using the homeostasis model assessment (23). We defined anovulation as any cycle with a peak progesterone concentration ≤5 ng/mL and no observed serum LH peak on the mid- or late luteal phase visits (n = 42 of 509 cycles; 8.3%). Study protocol compliance was high, with 94% of the participants completing 7 or 8 visits/cycle.

Dietary assessment

Participants completed a 24-h dietary recall (24HDR) at the clinic after fasting blood specimen collection during the 4 visits corresponding to menstruation, the midfollicular phase, ovulation, and the midluteal phase. Food and beverage intake [including sweetened and artificially sweetened (diet) sodas and citrus and noncitrus fruit juices] was collected, and nutrient data were analyzed by using the Nutrition Data System for Research (NDSR; version 2005; Nutrition Coordinating Center, University of Minnesota). Trained dietary interviewers, using both open-ended questions and visual amount-estimation tools (taking into account ice in beverages and amount unconsumed), captured food and beverage quantities consumed. The NDSR takes into account beverage type, brand (when necessary), and serving size in calculating nutrient information [eg, 1 cup (1 cup = 237 mL) of cola contains 26.7 g added (and total) sugar, 1 cup of 100% apple juice contains 24.2 g total sugar, and 1 cup of 100% orange juice contains 20.2 g of total sugar]. We assessed citrus and noncitrus juice separately because grapefruit juice has been shown to increase endogenous estrogen concentrations (24). In addition to added sugars, the NDSR provides information on daily free fructose and sucrose intake. Because half of the disaccharide sucrose is fructose (which is split from glucose in the small intestine), we calculated total fructose equal to the intake of free fructose plus half the intake of sucrose (25). Further abstraction was done from the raw 24HDR data to discriminate between cola and noncola soda because this information is not included in standard NDSR output and caramel coloring agents in colas are thought to affect health outcomes (26–28). The consumption of sweetened fruit and energy drinks among BioCycle Study participants was too limited to adequately assess. Compliance was high, with 87% completing 4 dietary recalls/cycle and 99% completing 3 dietary recalls.

Covariate assessment

At study enrollment, adiposity measurements, including waist circumference, waist-to-hip ratio, and BMI, were obtained by trained study staff by using standardized protocols, whereas age, race, and reproductive history were obtained by using validated questionnaires (19). For prospectively measured covariates, participants were provided a diary in which they were instructed to record daily vigorous exercise (min) and cigarettes smoked (number). The majority of participants (89%) completed ≥75% of their daily diaries.

Statistical analyses

Descriptive statistics were compared between previously used cutoffs (16) (≤1 cup) of sweetened soda intake and tertiles of daily total fructose intake, averaged over the 2 cycles under study. We assessed differences by using ANOVA per the Satterthwaite method for unequal variances (29) and Fisher's exact chi-square tests where appropriate (30). The Tukey method was used to test significant differences between each pair of groups for both the total fructose ANOVA and chi-square tests to account for multiple comparisons (30, 31). Reproductive hormone concentrations were log-transformed for statistical analyses. Linear mixed models were used to evaluate the association between soda (sweetened or artificially sweetened, cola and noncola), juice, added sugars, and total fructose intake (16) and follicular and luteal phase–specific log serum concentrations of free and total estradiol, luteal progesterone, and periovulatory LH and FSH. Generalized linear mixed models were used to estimate the odds of anovulation on the basis of beverage consumption, added sugars, and total fructose consumption. We assessed the association between intakes of ≥1 cup/d compared with intakes of <1 cup/d for sweetened or artificially sweetened beverages (16) and above the midpoint of AHA-recommended added-sugar intake amounts for moderately physically active US women, aged 19–30 y (recommended range: 38–42 g/d; cutoff for analysis: ≥40.0 g/d compared with <40.0 g/d) (32), as well as intakes above the usual added-sugar intake for US women (≥73.2 g/d compared with <73.2 g/d) (6), with hormone concentrations and anovulation. We additionally compared cutoffs at the 33rd percentile (≥28.4 g/d compared with <28.4 g/d), the 50th percentile (median; ≥33.5 g/d compared with <33.5 g/d), and the 66th percentile (≥41.5 g/d compared with <41.5 g/d) for total fructose. Random intercepts were included in the models to account for between-women variation in baseline hormone concentrations and within-woman correlations across cycles. Models evaluating reproductive hormones were restricted to ovulatory cycles (20) because the hormonal patterns for anovulatory cycles are distinctly different from ovulatory cycles.

Potential confounders were determined a priori by using directed acyclic graphs. Age (continuous), BMI (continuous), race (white, black, Asian, and other), total energy intake (continuous), physical activity (continuous), and the previously described Mediterranean Diet Score (33) (continuous) were included in our final models. For all sodas, we additionally adjusted for total caffeine intake. Additional covariates including cigarette use, gravidity, and insulin resistance and adjustment for other relevant beverages were considered but did not appreciably alter the estimates (34), nor did adjustment for waist circumference or waist-to-hip ratio compared with BMI. Because estradiol, progesterone, LH, and FSH concentrations change over the cycle in response to complex feedback mechanisms with other hormones, traditional regression adjustment for other hormone concentrations is inadequate. Therefore, we additionally conducted marginal structural model analyses that adjusted for other reproductive hormones through stabilized inverse probability of exposure weights (35, 36) to appropriately account for the time-varying confounding factors affected by prior exposure (eg, follicular FSH affects ovulatory LH, which in turn affects luteal progesterone concentrations). Weighted linear mixed-effects models with random intercepts were used to estimate the parameters of the marginal structural model.

To assess how total fructose, added sugar, and energy-containing beverage intake affect hormonal patterns, we used nonlinear mixed models with harmonic terms. Although the linear mixed models allow for estimation of mean percentage differences, these harmonic models can additionally evaluate differences in amplitude (ie, difference between nadir and peak concentrations) and timing of phase shifts while taking into account between- and/or within-subject variation (37).

Sensitivity analyses were conducted to assess the effects of continuous and alternative cutoffs for beverage intake, added sugar, and total fructose on reproductive hormones and anovulation. For reproductive hormones, we additionally assessed the effects when including anovulatory cycles. All analyses were performed in SAS version 9.2 (SAS Institute) except for the nonlinear mixed models with harmonic terms, which were performed in R version 2.13.1 (R Foundation for Statistical Computing).

RESULTS

Beverage and fructose consumption

Over 2 cycles, 69% of the women consumed soda (52% exclusively sweetened, 27% exclusively artificially sweetened, and 21% both sweetened and artificially sweetened) and 81% consumed fruit juice (59% exclusively citrus, 11% exclusively noncitrus fruit juice, and 30% both citrus and other fruit juice). Mean (±SD) intakes of added sugars and total fructose across both cycles were 57.2 ± 26.9 g/d and 35.4 ± 13.7 g/d, respectively. Sweetened soda intake was positively associated with black race, fructose and added sugar intake, and both follicular estradiol and follicular free estradiol concentrations and inversely associated with daily exercise, alcohol, percentage of calories from protein, fiber intake, and Mediterranean Diet Score (Table 1). Fructose intake over the study period was positively associated with total energy intake, percentage of calories from carbohydrates, and added sugars and inversely associated with percentage of calories from protein (Table 1).

TABLE 1.

Characteristics of women participating in the BioCycle Study by average fructose and sweetened soda intake across 2 menstrual cycles¹

		Sweetened soda intake2			Total fructose intake3
	Total	<1 cup/d	≥1 cup/d	P	<28.4 g/d	28.4–41.5 g/d	>41.5 g/d	P
No. of participants [n (%)]	259	244 (94.2)	15 (5.8)		86 (33.2)	88 (34.0)	85 (32.8)
Age (y)	27.3 ± 8.24	27.3 ± 8.3	26.5 ± 7.1	0.71	26.8 ± 8.1	27.0 ± 8.0	28.1 ± 8.6	0.54
Race [n (%)]				0.01				0.02
White	154 (59.5)	147 (60.3)	7 (46.7)		43 (50.0)	60 (68.2)	51 (60.0)
Black	51 (19.7)	43 (17.6)	8 (53.3)a		16 (18.6)	13 (14.8)	22 (25.9)
Asian	37 (14.3)	37 (15.2)	0 (0.0)a		21 (24.4)	10 (11.4)	6 (7.1)
Other	17 (6.6)	17 (7.0)	0 (0.0)		6 (7.0)	5 (5.7)	6 (7.1)
BMI (kg/m2)	24.1 ± 3.9	24.0 ± 3.8	25.3 ± 4.1	0.20	23.8 ± 3.9	24.3 ± 3.8	24.1 ± 3.9	0.74
Waist-to-hip ratio	0.75 ± 0.06	0.75 ± 0.06	0.78 ± 0.06	0.08	0.74 ± 0.05	0.75 ± 0.06	0.75 ± 0.05	0.31
Waist circumference (cm)	74.7 ± 8.7	74.5 ± 8.7	79.1 ± 8.7	0.06	73.4 ± 8.8	74.8 ± 8.4	76.0 ± 8.9	0.13
Nulligravid [n (%)]	177 (69.4)	165 (68.8)	12 (80.0)	0.56	60 (72.3)	63 (72.4)	54 (63.5)	0.35
Cigarette use [n (%)]	41 (15.8)	40 (16.4)	1 (6.7)	0.48	15 (17.4)	17 (19.3)	9 (10.6)	0.25
Vigorous exercise (min/d)	14.7 ± 21.9	15.4 ± 22.4	4.7 ± 5.0	<0.001	16.9 ± 24.8	15.6 ± 18.7	11.7 ± 21.9	0.26
Diet
Total energy (kcal)	1613.3 ± 367.3	1602.6 ± 366.2	1786.9 ± 353.1	0.06	1378.9 ± 320.5a,b	1649.4 ± 322.0a,c	1813.1 ± 324.0b,c	<0.001
Alcohol (g)	2.8 ± 5.5	2.9 ± 5.6	1.1 ± 1.7	0.004	3.1 ± 7.0	3.1 ± 5.4	2.1 ± 3.3	0.37
Carbohydrates (% of energy)	50.8 ± 7.3	50.7 ± 7.3	52.0 ± 7.0	0.52	47.5 ± 7.8a,b	51.3 ± 7.0a,c	53.6 ± 5.5b,c	<0.001
Protein (% of energy)	15.8 ± 3.1	15.9 ± 3.1	14.0 ± 2.1	0.02	17.5 ± 3.4a,b	15.6 ± 2.5a,c	14.3 ± 2.5b,c	<0.001
Fat (% of energy)	33.8 ± 5.6	33.7 ± 5.6	34.6 ± 6.1	0.56	34.9 ± 6.2	33.6 ± 5.4	32.9 ± 4.9	0.06
Fiber (g/d)	13.6 ± 5.6	13.9 ± 5.6	9.4 ± 3.4	0.001	11.8 ± 4.3a,b	15.0 ± 6.0a	14.1 ± 5.8b	<0.001
Fructose (g/d)	35.4 ± 13.7	34.2 ± 12.9	55.3 ± 11.8	<0.001	20.8 ± 5.1a,b	34.5 ± 4.1a,c	51.2 ± 7.9b,c	<0.001
Added sugars (g/d)	57.2 ± 26.9	55.0 ± 25.6	93.5 ± 22.0	<0.001	32.8 ± 12.9a,b	55.9 ± 14.7a,c	83.3 ± 23.2b,c	<0.001
Mediterranean Diet Score	3.3 ± 0.7	3.3 ± 0.7	3.1 ± 0.5	0.04	3.2 ± 0.8	3.5 ± 0.8	3.3 ± 0.6	0.06
Biomarkers
Total E2 (pg/mL)
Follicular5	67.0 (40.0–142.0)6	65.0 (39.0–135.0)	83.0 (46.0–180.0)	0.01	65.0 (40.0–136.0)	61.0 (38.0–132.0)	73.0 (43.0–151.5)	0.24
Luteal7	107.0 (66.0–160.5)	106.0 (63.0–159.0)	123.0 (84.0–171.0)	0.14	104.0 (66.0–157.0)	116.0 (70.0–167.5)	107.0 (63.0–163.0)	0.78
Free E2 (pg/mL)
Follicular5	1.05 (0.64–2.27)	1.03 (0.63–2.19)	1.29 (0.71–2.93)	0.01	1.03 (0.65–2.28)	0.98 (0.60–2.12)	1.17 (0.67–2.39)	0.32
Luteal7	1.63 (1.03–2.38)	1.60 (0.99–2.35)	1.94 (1.34–2.62)	0.08	1.59 (1.04–2.34)	1.68 (1.03–2.40)	1.66 (0.99–2.48)	0.98
Progesterone8 (ng/mL)	7.0 (2.0–11.3)	7.1 (1.9–11.3)	7.3 (2.6–11.1)	0.53	6.9 (1.9–11.6)	7.2 (2.1–11.2)	7.1 (2.2–11.0)	0.61
LH9 (ng/mL)	8.7 (5.5–17.4)	8.7 (5.4–17.0)	8.7 (5.7–22.3)	0.34	8.6 (5.5–17.4)	8.8 (5.6–15.7)	8.8 (5.4–18.7)	0.95
FSH9 (mIU/mL)	6.5 (4.6–9.6)	6.5 (4.7–9.6)	6.3 (4.3–10.1)	0.33	6.4 (4.6–9.3)	6.3 (4.6–9.3)	6.6 (4.7–10.2)	0.62
Insulin resistance (mmol/L)	1.5 (1.1–2.1)	1.5 (1.1–2.1)	2.0 (1.6–2.1)	0.17	1.6 (1.1–2.2)a	1.3 (1.1–1.9)a,b	1.6 (1.2–2.3)b	0.03

¹All dietary variables, including fructose, added sugars, and sweetened soda intake, were assessed 4 times during each cycle (corresponding to menstruation, midfollicular phase, ovulation, and midluteal phase clinic visits) via 24-h dietary recall. Fructose was equal to the intake of free fructose plus half of the intake of sucrose. Mean values or frequencies sharing a common superscript letter are significantly different, P < 0.05 (Tukey's test). E2, estradiol; FSH, follicle-stimulating hormone; LH, luteinizing hormone.

²Calculated by using Student's t test for continuous variables (Satterthwaite method when variance was unequal) and exact chi-square tests for categorical variables. Reproductive hormones were log-transformed for normality for statistical analyses. 1 cup = 237 mL.

³Calculated by using ANOVA for continuous data and exact chi-square tests for categorical data.

⁴Mean ± SD (all such values).

⁵Follicular includes menstruation, midfollicular, late follicular, LH/FSH surge, and ovulation phase visits.

⁶Median; IQR in parentheses (all such values).

⁷Luteal includes early luteal, midluteal, and late luteal phase visits.

⁸Progesterone assessment includes early luteal, midluteal, and late luteal phase visits.

⁹FSH and LH include the late follicular, LH/FSH surge, and ovulation phase visits.

Reproductive hormones

Women who consumed ≥1 cup sweetened soda/d had 16.3% higher concentrations of follicular total estradiol (86.5 pg/mL compared with 74.4 pg/mL, P = 0.01) and 12.5% higher concentrations of follicular free estradiol (1.35 pg/mL compared with 1.20 pg/mL, P = 0.03) compared with women who consumed <1 cup/d after multivariate adjustment, including total caffeine intake (Table 2). Results were similar for cola and noncola intake, with consumption of ≥1 cup/d associated with elevated follicular free and total estradiol concentrations. Neither intake of artificially sweetened soda nor intake of citrus/noncitrus fruit juice (≥1 cup/d) was significantly associated with reproductive hormone concentrations, with the exception of citrus juice, which showed a higher concentration of luteal progesterone for ≥1 cup/d compared with a lower intake (Table 2). No significant associations were found between added sugar (either above the AHA-recommended limit or usual intake for US women compared with a lower intake) and reproductive hormones (Table 3). Women above the 66th percentile in fructose intake (41.5 g/d) had 9.3% higher follicular free estradiol concentrations (1.29 pg/mL compared with 1.18 pg/mL, P = 0.04) compared with those consuming <41.5 g/d (Table 3), but no significant differences were found at the 33rd or 50th percentiles (data not shown). Sensitivity analyses showed that women who averaged 2–3 cups sweetened soda/d had a 43.0% higher follicular total estradiol (106.4 pg/mL compared with 74.4 pg/mL, P = 0.02) and a 39.2% higher follicular free estradiol (1.67 pg/mL compared with 1.20 pg/mL, P = 0.03) compared with women who averaged <1 cup/d. No significant associations were found between 1-cup increments in artificially sweetened soda or juice intake, or with continuous added sugar or total fructose intake, and reproductive hormones. Similar results were found when additionally adjusted for relevant phase-specific hormones with the use of marginal structural models, with the exception of a significantly higher LH for ≥1 cup/d of sweetened soda intake compared with a lower intake and a significantly higher LH and FSH for added sugar above the usual intake for US women or above the 66th percentile in fructose intake compared with a lower intake.

TABLE 2.

Serum concentrations of reproductive hormones according to energy-containing beverage intake (n = 467 ovulatory cycles)¹

	Sweetened soda		Artificially sweetened soda		Cola soda		Noncola soda		Citrus fruit juice		Noncitrus fruit juice
Hormone	≥1 cup/d	<1 cup/d	≥1 cup/d	<1 cup/d	≥1 cup/d	<1 cup/d	≥1 cup/d	<1 cup/d	≥1 cup/d	<1 cup/d	≥1 cup/d	<1 cup/d
E2 (pg/mL)
Follicular2	86.5 (77.5, 96.5)*	74.4 (70.1, 79.0)*	81.5 (73.0, 91.8)	75.2 (70.8, 79.8)	86.5 (78.3, 95.6)*	74.4 (70.1, 79.0)*	86.3 (78.2, 95.2)*	74.4 (70.2, 78.9)*	75.8 (71.5, 80.4)	78.5 (68.9, 89.4)	76.1 (71.8, 80.6)	75.6 (65.1, 88.0)
Luteal3	113.3 (100.5, 127.7)	105.6 (98.5, 114.4)	113.3 (99.5, 129.0)	106.7 (99.5, 114.4)	112.2 (99.5, 125.2)	106.7 (98.5, 113.3)	111.6 (99.7, 124.9)	106.3 (98.9, 114.1)	107.8 (100.4, 115.7)	102.0 (89.2, 116.6)	107.8 (100.5, 115.6)	97.1 (82.3, 114.5)
Free E2 (pg/mL)
Follicular2	1.35 (1.21, 1.51)*	1.20 (1.14, 1.27)*	1.26 (1.13, 1.42)	1.21 (1.15, 1.28)	1.35 (1.22, 1.49)*	1.20 (1.13, 1.27)*	1.34 (1.22, 1.48)*	1.20 (1.13, 1.27)*	1.22 (1.15, 1.29)	1.21 (1.14, 1.28)	1.22 (1.15, 1.29)	1.20 (1.04, 1.40)
Luteal3	1.79 (1.60, 1.99)	1.65 (1.55, 1.77)	1.70 (1.51, 1.92)	1.67 (1.57, 1.79)	1.72 (1.55, 1.92)	1.67 (1.55, 1.77)	1.72 (1.55, 1.91)	1.66 (1.56, 1.78)	1.68 (1.57, 1.79)	1.68 (1.57, 1.79)	1.68 (1.58, 1.79)	1.56 (1.33, 1.83)
Progesterone4 (ng/mL)	5.0 (4.0, 6.2)	4.7 (4.2, 5.4)	4.6 (3.7, 5.9)	4.8 (4.2, 5.4)	4.5 (3.7, 5.6)	4.8 (4.2, 5.4)	4.5 (3.7, 5.6)	4.8 (4.2, 5.4)	4.9 (4.3, 5.5)*	3.8 (3.0, 4.9)*	4.8 (4.3, 5.4)	4.0 (3.0, 5.5)
FSH5 (mIU/mL)	6.9 (6.2, 7.6)	7.0 (6.7, 7.4)	7.2 (6.5, 8.1)	7.0 (6.6, 7.3)	9.9 (8.7, 11.2)	7.0 (6.7, 7.4)	6.8 (6.2, 7.5)	7.0 (6.7, 7.4)	10.2 (9.5, 10.9)	9.9 (8.2, 11.9)	10.3 (9.6, 11.0)	8.4 (6.8, 10.4)
LH5 (ng/mL)	10.8 (9.4, 12.6)	10.1 (9.4, 10.8)	9.9 (8.4, 11.6)	10.2 (9.5, 10.9)	6.8 (6.2, 7.5)	10.3 (9.6, 11.0)	9.9 (8.7, 11.2)	10.2 (9.5, 11.0)	7.0 (6.6, 7.4)	7.1 (6.3, 8.1)	7.0 (6.7, 7.4)	6.4 (5.6, 7.4)

¹All values are geometric means; 95% CIs in parentheses. Adjusted for age, BMI, race, total energy intake, Mediterranean Diet Score, and physical activity by using linear mixed-effects models on the log scale of hormones. Sodas were additionally adjusted for total caffeine intake. Beverage intake was assessed at 4 times during each cycle (corresponding to menstruation, midfollicular phase, ovulation, and midluteal phase clinic visits) via 24-h dietary recall. Anovulation is any cycle with peak progesterone concentration ≤5 ng/mL and no observed serum LH peak on the mid- or late luteal phase visits (n = 42 cycles). 1 cup = 237 mL. *P < 0.05. E2, estradiol; FSH, follicle-stimulating hormone; LH, luteinizing hormone.

²Follicular includes menstruation, midfollicular, late follicular, LH/FSH surge, and ovulation phase visits.

³Luteal includes early luteal, midluteal, and late luteal phase visits.

⁴Progesterone assessment includes early luteal, midluteal, and late luteal phase visits.

⁵FSH and LH include the late follicular, LH/FSH surge, and ovulation phase visits.

TABLE 3.

Geometric mean difference in log serum concentrations of reproductive hormones according to added sugar and fructose intake (n = 467 ovulatory cycles)¹

	Added sugar (AHA recommended limit)		Added sugar (usual intake for US women)		Fructose (66th percentile)
Hormone	≥40.0 g/d	<40.0 g/d	≥73.2 g/d	<73.2 g/d	≥41.5 g/d	<41.5 g/d
Total E2 (pg/mL)
Follicular2	77.1 (72.3, 82.2)	74.4 (69.1, 80.0)	79.3 (73.2, 86.0)	74.7 (70.2, 79.5)	80.1 (74.2, 86.4)	73.8 (69.2, 78.7)
Luteal3	107.6 (99.6, 116.2)	106.5 (97.9, 116.0)	114.1 (103.3, 126.1)	104.9 (97.5, 112.9)	107.0 (97.4, 117.4)	107.2 (99.6, 115.5)
Free E2 (pg/mL)
Follicular2	1.23 (1.15, 1.31)	1.21 (1.12, 1.30)	1.26 (1.17, 1.37)	1.20 (1.13, 1.28)	1.29 (1.19, 1.39)*	1.18 (1.11, 1.26)*
Luteal3	1.67 (1.56, 1.80)	1.67 (1.55, 1.81)	1.80 (1.63, 1.75)	1.63 (1.53, 1.75)	1.70 (1.56, 1.85)	1.66 (1.55, 1.78)
Progesterone4 (ng/mL)	4.9 (4.2, 5.6)	4.6 (4.0, 5.4)	5.0 (4.2, 6.0)	4.8 (4.2, 5.5)	4.6 (3.9, 5.5)	4.8 (4.2, 5.5)
FSH5 (mIU/mL)	7.1 (6.7, 7.5)	6.8 (6.4, 7.3)	7.0 (6.5, 7.6)	6.9 (6.5, 7.3)	7.2 (6.7, 7.7)	6.9 (6.5, 7.3)
LH5 (ng/mL)	10.1 (9.4, 11.0)	10.2 (9.3, 11.3)	10.2 (9.1, 11.3)	10.2 (9.4, 11.0)	10.1 (9.2, 11.1)	10.2 (9.4, 11.1)

¹Values are geometric mean differences; 95% CIs in parentheses. Adjusted for age, BMI, race, total energy intake, Mediterranean Diet Score, and physical activity by using linear mixed-effects models on the log scale of hormones. Anovulation is any cycle with peak progesterone concentration ≤5 ng/mL and no observed serum LH peak on the mid- or late luteal phase visits (n = 42 cycles). Nutrient intake was assessed at 4 times each cycle (corresponding to menstruation, midfollicular phase, ovulation, and midluteal phase clinic visits) via 24-h dietary recall. Fructose was equal to the intake of free fructose plus half of the intake of sucrose. *P < 0.05. AHA, American Heart Association; E2, estradiol; FSH, follicle-stimulating hormone; LH, luteinizing hormone.

²Follicular includes menstruation, midfollicular, late follicular, LH/FSH surge, and ovulation phase visits.

³Luteal includes early luteal, midluteal, and late luteal phase visits.

⁴Progesterone assessment includes early luteal, midluteal, and late luteal phase visits.

⁵FSH and LH include the late follicular, LH/FSH surge, and ovulation phase visits.

Nonlinear harmonic models mirrored results from the linear mixed models for sweetened soda intake and both total and free estradiol (Figure 1). Women who consumed, on average, ≥1 cup sweetened soda/d had 13.3% (P = 0.04) higher mean free estradiol concentrations levels compared with woman who consumed <1 cup/d after adjustment for age, BMI, race, total energy intake, Mediterranean Diet Score, and physical activity (Table 4). Although no significant associations were found for total fructose intake and reproductive hormones, women who consumed ≥73.2 g/d of added sugar had 9.4% (P = 0.01) higher mean free estradiol concentrations compared with women with intakes <73.2 g/d (Table 4;Figure 1) in addition to greater LH amplitude (P = 0.01) (Table 4).

FIGURE 1.

Adjusted mean serum concentrations of reproductive hormones across the menstrual cycle according to sweetened soda (A, B) intake [≥1 cup/d (1 cup = 237 mL) compared with a lower intake] or added sugar (C, D) (≥73.2 g/d compared with a lower intake; usual intake for American women) based on nonlinear harmonic models centered on the day of ovulation (n = 467 cycles). Harmonic models were adjusted for age, BMI, race, dietary factors, and physical activity. Figures display total and free estradiol for white women with a mean age of 27 y, a BMI (in kg/m²) of 24.1, total energy intake of 1614 kcal, a Mediterranean Diet Score of 2.9, and mean daily exercise of 14.5 min. Nonlinear harmonic models allowed a flexible approach to study cyclic patterns over time and can not only evaluate mean percentage differences in hormone concentrations across the menstrual cycle by sweetened soda or added sugar intake category but can also assess differences in peak hormone concentrations (ie, amplitude) and timing of the peak (ie, phase shift). Estimates, 95% CIs, and P values are shown in Table 4.

TABLE 4.

Associations between sweetened soda, added sugar, total fructose, and hormones by using nonlinear mixed models with harmonic terms¹

	Sweetened soda (≥1 cup/d compared with <1 cup/d)			Added sugar (≥73.2 g/d compared with <73.2 g/d)			Total fructose (≥41.5 g/d compared with <41.5 g/d)
Hormone	Estimate	95% CI	P	Estimate	95% CI	P	Estimate	95% CI	P
Total E2
Mean2 (%)	13.7	0.98, 28.01	0.03	7.21	0.50, 14.36	0.03	3.77	−2.28, 10.20	0.23
Amplitude3	0.01	−0.07, 0.17	0.91	0.02	−0.03, 0.11	0.58	−0.01	−0.05, 0.06	0.68
Phase shift4 (d)	−0.35	−1.21, 0.34	0.36	0.17	−0.22, 0.45	0.36	0.07	−0.28, 0.36	0.66
Free E2
Mean2 (%)	13.3	0.38, 27.93	0.04	9.43	2.51, 16.82	0.01	5.05	−1.18, 11.66	0.11
Amplitude3	0.02	−0.07, 0.23	0.79	0.00	−0.05, 0.10	0.92	−0.03	−0.07, 0.04	0.29
Phase shift4 (d)	−0.09	−0.71, 0.42	0.75	0.19	−0.22, 0.52	0.33	−0.10	−0.52, 0.24	0.58
Progesterone
Mean2 (%)	−6.39	−17.62, 6.36	0.31	1.24	−5.71, 8.70	0.73	−2.86	−9.08, 3.77	0.39
Amplitude3	0.02	−0.08, 0.18	0.74	0.01	−0.05, 0.10	0.80	−0.04	−0.08, 0.03	0.19
Phase shift4 (d)	−0.03	−0.07, 0.00	0.13	−0.08	−0.51, 0.21	0.67	−0.02	−0.22, 0.14	0.81
LH
Mean2 (%)	3.33	−10.74, 19.62	0.66	−4.76	−11.99, 3.06	0.23	−1.79	−8.72, 5.65	0.63
Amplitude3	0.04	−0.01, 0.15	0.21	0.05	0.01, 0.12	<0.01	0.02	−0.01, 0.08	0.22
Phase shift4 (d)	−0.31	−0.96, 0.29	0.33	−0.21	−0.55, 0.11	0.21	0.18	−0.14, 0.46	0.27
FSH
Mean2 (%)	−4.07	−15.48, 8.89	0.52	−4.32	−10.65, 2.45	0.21	3.28	−3.03, 10.01	0.32
Amplitude3	0.01	−0.03, 0.11	0.82	0.01	−0.02, 0.07	0.57	0.00	−0.02, 0.05	0.83
Phase shift4 (d)	−0.32	−1.00, 0.51	0.43	−0.07	−0.46, 0.39	0.76	−0.25	−0.58, 0.16	0.22

¹Adjusted for age, BMI, total energy intake, Mediterranean Diet Score, and mean daily exercise. Soda was additionally adjusted for total caffeine intake. All models accounted for repeated measures and correlated cycles. 1 cup = 237 mL. E2, estradiol; FSH, follicle-stimulating hormone; LH, luteinizing hormone.

²Ratios were obtained from the model to compare the mean hormone concentrations across the menstrual cycle for short and long cycles with the normal cycle length group. Estimates of the percentage difference in hormone concentration and 95% CIs were obtained by subtracting 1 from each ratio and multiplying by 100.

³Amplitude difference (ie, comparing the nadir to the peak) in hormone concentrations across the menstrual cycle is interpreted as the difference in peak hormone concentrations compared with the reference category.

⁴Phase shift differences are interpreted as the difference in the number of days of a standardized 28-d cycle between the increase in hormone concentrations compared with the reference category (eg, 1.0 = hormone increase is shifted 1.0 d later).

Anovulation

No significant associations were found between intakes of energy-containing beverages (≥1 cup/d compared with <1 cup/d), added sugar (≥40.0 g/d compared with <40.0 g/d or ≥73.2 g/d compared with <73.2 g/d), or total fructose (≥41.5 g/d compared with <41.5 g/d) and ovulatory function (Table 5). Sensitivity analyses for 1-cup increments of beverage consumption yielded no significant results nor did assessing anovulation on the basis of the less conservative definition (peak progesterone of ≤5 ng/mL).

TABLE 5.

Odds of anovulation with consumption of added sugar, fructose, and fructose-rich beverages¹

	Intake amount	Multivariate-adjusted OR (95% CI)2
Beverage intake
Sweetened soda	≥1 cup/d compared with <1 cup/d	1.25 (0.52, 3.03)
Artificially sweetened soda	≥1 cup/d compared with <1 cup/d	1.07 (0.34, 3.35)
Cola soda	≥1 cup/d compared with <1 cup/d	0.65 (0.25, 1.73)
Noncola soda	≥1 cup/d compared with <1 cup/d	1.04 (0.41, 2.64)
Citrus fruit juice	≥1 cup/d compared with <1 cup/d	0.73 (0.29, 1.82)
Noncitrus fruit juice	≥1 cup/d compared with <1 cup/d	0.67 (0.23, 1.99)
Nutrient intake
Added sugar (AHA recommended limit)	≥40 g/d compared with <40 g/d	0.93 (0.53, 1.62)
Added sugar	≥73.2 g/d compared with <73.2 g/d	0.57 (0.31, 1.06)
Fructose	≥41.5 g/d compared with <41.5 g/d	0.76 (0.40, 1.42)

¹Anovulation is any cycle with peak progesterone concentration ≤5 ng/mL and no observed serum luteinizing hormone peak on the mid- or late luteal phase visits (n = 42 cycles). Intake was assessed at 4 times during each cycle (corresponding to menstruation, midfollicular phase, ovulation, and midluteal phase clinic visits) via 24-h dietary recall. Fructose was equal to the intake of free fructose plus half of the intake of sucrose. 1 cup = 237 mL. AHA, American Heart Association.

²Adjusted for age, BMI, race, total energy intake, Mediterranean Diet Score, and physical activity by using generalized linear mixed models. Sodas were additionally adjusted for total caffeine intake.

DISCUSSION

We observed that sweetened soda intake (≥1 cup/d compared with a lower intake) was associated with elevated follicular free and total estradiol across all models, but neither artificially sweetened soda nor fruit juice was significantly associated with estradiol or any of the reproductive hormones after adjustment for age, BMI, race, dietary factors, and physical activity. Added sugar above the average US woman's intake (≥73.2 g/d) or above the 66th percentile in total fructose intake (≥41.5 g/d) was associated with elevated follicular estradiol and LH but was not significant across all models. Energy-containing beverages, added sugars, and total fructose were not associated with sporadic anovulation after multivariate adjustment. Findings from this study suggest that intake of energy-containing beverages may increase follicular serum estradiol concentrations but do not interfere with ovulation among healthy premenopausal women with no known ovulatory disorders. Although speculative, given that higher concentrations of premenopausal follicular estradiol (ranging from 15 to 36 pg/mL for total estradiol and from 0.16 to 0.40 pg/mL for free estradiol) have been associated with increased breast cancer risk, further research is warranted (17).

To our knowledge, our finding that sweetened soda consumption was positively associated with reproductive hormone concentrations is novel in humans. Although we are aware of no previous animal or human studies investigating the relation between reproductive hormones and fructose or added sugar intake, our results showing higher estradiol concentrations with sweetened soda intake mirror results from animal studies (38). The 2 previous studies that investigated soda intake and reproductive hormone concentrations in humans found null associations (15, 16); however, a comparison of these studies with ours is limited because their assessment of diet was via a single retrospective food-frequency questionnaire and their hormone measurement, captured over one menstrual cycle, included either one very early (days 1–5 after start of menses) follicular measurement (16) or 2 measurements (days 11 and 22) (15) and did not take into account menstrual cycle variability. Caramel coloring in cola sodas (which contains advanced glycation end-products) has been associated with adverse health effects in animal models (28); however, we found both cola and noncola sodas to be similarly associated with elevated estradiol, indicating that caramel coloring plays no mechanistic role in the association between sweetened soda and estradiol.

Although limited data exist on the effect of sweetened soda on reproductive hormones, other studies have shown that sweetened beverages are associated with impaired fasting glucose and metabolic syndrome (39, 40). Because sucrose and high-fructose corn syrup (HFCS) are both composed of approximately equal parts glucose and fructose, many believe that the effects on the endocrine system are equivalent (1). Others, in contrast, contend that the effects of HFCS-sweetened beverages differ from the effect of sucrose-sweetened beverages or fructose-rich beverages (eg, fruit juice) (40) due to the reactive carbonyls found in drinks containing HFCS (41). Although we could not distinguish between sodas sweetened with sucrose from those sweetened with HFCS, our results showing elevated reproductive hormone concentrations with added sugars and total fructose support the hypothesis that the effects of fructose do not differ between sucrose- or HFCS-sweetened beverages.

The question remains as to whether fructose or some other component in sweetened beverages is associated with elevated estradiol. The direct mechanism as to how fructose may affect estradiol is not known; however, research has shown that the hepatic metabolism of fructose induces hyperlipidemia, hyperinsulinemia, and uric acid production (1, 42). Although cholesterol is the precursor to the 5 major classes of steroid hormones, including estrogen (43), understanding the relation between fructose and estradiol is complicated by the fact that estrogen has been shown to protect against the development of hyperinsulinemia associated with high fructose intake (42). The inclusion of fructose intake in our multivariate models with beverages made the relation between sweetened soda, fruit juice intake, and estradiol weaker, suggesting that fructose explains part of the association. Although fructose may contribute, the trend toward elevated estradiol with artificially sweetened soda intake of ≥1 cup/d suggests that additional components may be at work, given that other studies have found that sodas (regardless of sugar content) contribute to adverse health effects (13, 44), including greater risk of ovulatory disorder infertility (13). In addition, although fruit juices are a known contributor to fructose intake, our finding of a significant association between sweetened soda and estradiol, but not fruit juice, may be due to the beneficial components of fruit juice, which include vitamins and antioxidants (44) and which are often lacking in consumers of sugar-sweetened beverages (45). Further research into the relation between different juice types [including freshness and processing (46)] and premenopausal reproductive hormones is needed.

The BioCycle Study has several strengths, including multiple measures of hormones (by using standardized methods to time cycle phase), dietary intake, and lifestyle factors over the course of 2 menstrual cycles. Although self-report of diet, including the 24HDR, is subject to measurement error (47–50), we assessed recall validity by comparing total soda intake with the averages obtained from 2 food-frequency questionnaires captured over the same time period and found a significant correlation (r = 0.71). Nevertheless, the study was limited by the relatively low consumption of added sugars and sodas. US premenopausal women consume, on average, 78 g added sugars/d, 19 oz of soda, and 3 oz of fruit juice (2), whereas the BioCycle Study participants consumed, on average, 57.2 g added sugars/d, 3 oz of soda, and 4 oz of fruit juice. Another potential limiting factor was the use of CLEIA for estradiol assessment. Although isotope dilution–gas chromatography–mass spectrometry (ID-GC-MS) techniques are considered to be the gold standard for measuring estradiol (51), the use of ID-GC-MS is most critical for populations with relatively low estradiol concentrations [ie, men, children, and postmenopausal women (54)]. BioCycle participants had estradiol concentrations within the normal premenopausal range [median follicular estradiol = 67.0 (IQR: 40.0–142.0) and median luteal estradiol = 107.0 (IQR: 66.0–160.5)]. In addition, the Immulite 2000 CLEIA has been shown to have good overall accuracy and little bias (51) compared with ID-GC-MS and thus prone to random rather than systematic error. Thus, we believe that the Immulite 2000 CLEIA was sensitive enough to detect significant differences in estradiol concentrations for sweetened soda and, in the nonlinear models, added sugar intake. Moreover, measurement error of the outcome only results in an increase of the variance. Future studies investigating the association between intakes of sweetened beverages, added sugars, and fructose and reproductive hormones should be conducted by using gold-standard techniques, for both total and free estradiol (52), to help inform the validity of our and others’ (15, 16) results.

In conclusion, ≥1 cup sweetened soda/d was associated with elevated estradiol concentrations. Added sugars and fructose, particularly in liquid form, may partially explain the effects of energy-containing beverages on reproductive hormones. Although we observed no effect on incident anovulation with moderate sweetened beverage consumption, further research investigating higher consumption along with the inclusion of women with more pronounced ovulatory disorders is warranted (13). Whereas recent research indicates that consumption of sweetened beverages and added sugars is decreasing in the Unites States (53), mean intakes among premenopausal women continue to exceed recommendations. Our findings have public health implications not only for the role that energy-containing beverages have on female fertility but also for their potential relation with a woman's future risk of chronic diseases associated with reproductive hormones (13, 17, 18).