Healthy Post-Menarchal Adolescent Girls Demonstrate Multi-Level Reproductive Axis Immaturity

Abstract

Context

Menstrual irregularity after menarche has been attributed to immature estrogen positive feedback activity (E+FB) but data are conflicting.

Objective

To determine the hypothalamic–pituitary–ovarian contributions to menstrual irregularity in adolescents.

Methods

Twenty-three healthy girls [aged 12.8 to 17.6 years; 0.4 to 3.5 years postmenarche; body mass index (BMI) percentile, 41.0 to 99.3] underwent serial hormone measurements and pelvic ultrasounds during two consecutive menstrual cycles. Hormones and follicle growth were compared with 65 adult historic controls with ovulatory cycles (OVs).

Results

Girls had anovulatory cycles (ANOVs; 30%), OVs with a short luteal phase (short OVs; 22%), or OVs with normal luteal phase (normal OVs; 48%) without differences in cycle length, chronologic or gynecologic age, or BMI. Adolescents showed a spectrum of E+FB [midcycle LH adjusted for preovulatory estradiol (E2)]; only normal OV girls were comparable to adults. All OV girls had lower E2, progesterone, and gonadotropins during the luteal phase and luteal-follicular transition compared with adults. Normal OV girls also had lower follicular phase LH and FSH levels, a longer follicular phase, a slower dominant follicle growth rate, and smaller estimated preovulatory follicle size than adults. Follicular phase E2 and inhibin B levels were lower in normal OV girls than in adults even after adjusting for differences in FSH and follicle size.

Conclusions

Early postmenarchal girls with normal OVs demonstrate mature E+FB but continue to have lower gonadotropin levels, diminished ovarian responsiveness, and decreased corpus luteum sex steroid synthesis compared with adults, indicating that reproductive axis maturity requires coordinated development of all components of the hypothalamic–pituitary–ovarian axis.

Investigation of hormone dynamics and follicle growth in early post-menarchal girls reveals intact estrogen positive feedback but decreased LH and FSH levels and ovarian responsiveness.

During the first few years after menarche, the hypothalamic–pituitary–ovarian axis is not fully mature and previous studies suggest that approximately half of menstrual cycles are anovulatory (1). The reported prevalence of ovulatory cycles (OVs) in adolescents at more advanced gynecologic ages (years since menarche) is inconsistent across studies (2–13), possibly due to differences in the methods used to identify OV, sample size, as well as subject ethnicity, nutritional status, and age at menarche (14). Importantly, few studies (7, 14, 15) have investigated hormone dynamics in contemporary adolescent girls who have matured in an era of increasing rates of obesity and earlier puberty, two factors that may influence the tempo of hypothalamic–pituitary–ovarian axis maturity.

Estrogen positive feedback activity was traditionally considered to be the major limiting factor in establishing regular OVs after menarche (13, 16–18). However, this concept is at odds with some more recent reports of spontaneous or estrogen-induced LH surges in premenarchal girls that are of comparable magnitude to surges in adult women (7, 14, 19–23). In addition, small increases in pregnanediol (Pd) glucuronide, the primary urinary metabolite of progesterone (P4), have been observed in many early adolescent menstrual cycles, suggesting ovulation with residual immaturity in corpus luteum (CL) formation and/or function (9, 24). We therefore hypothesized that other components of hypothalamic–pituitary–ovarian axis regulation contribute to reproductive axis immaturity. For example, the dynamic secretory profile of FSH, particularly during the luteal-follicular transition (LFT), when levels rise fourfold in adult women (25), is essential for initiating folliculogenesis and controlling the number of follicles reaching maturity. Moreover, for an LH surge to induce ovulation, it must be precisely coordinated with follicle growth and estradiol (E2) production, and the granulosa cells of the preovulatory follicle must express LH receptors to luteinize (26).

The current study was designed to investigate the reproductive hormone profiles that underlie the diverse menstrual cycle patterns of early postmenarchal girls and to contrast these profiles with those of normal adult women.

Materials and Methods

Subjects

Healthy girls, aged 12.8 to 17.6 years, 0.4 to 3.5 years postmenarche, were studied. A maximum gynecologic age of 3.5 years was chosen based on the longitudinal study by Vihko and Apter (27) of 200 girls demonstrating that 41% of girls had achieved OVs by the end of the fourth gynecologic year. We reasoned that we would capture a sufficient number of girls with OVs and anovulatory cycles (ANOVs) by including girls within or just after the third gynecologic year.

Subjects were recruited from the community by advertising in newspapers, on community bulletin boards, on the Internet, or through direct mail (postcard) marketing. Subjects were not on any medications known to interfere with reproductive hormones and had never been treated for a menstrual condition. Two subjects were on centrally acting medications (methylphenidate, a catecholamine reuptake inhibitor, and sertraline, a selective serotonin reuptake inhibitor). All subjects had normal thyroid, prolactin, and androgen levels (28); no inflammatory acne; and were nonhirsute (Ferriman-Gallwey score <8). They did not exercise excessively [defined as running >20 miles per week or its equivalent as determined using the Compendium of Physical Activities (29, 30)] and had no history of an eating disorder, stress fracture, or osteopenia/osteoporosis. All subjects were nonsmokers. Relevant medical history was obtained in person from the subject and a parent.

The Partners Human Research Committee approved the study and signed informed assent and consent were obtained from each subject and her parent, respectively.

Protocol

Method comparison and validation

Recognizing the impracticality of obtaining daily blood specimens from adolescents, we first performed a pilot study to determine the feasibility of using saliva, urine, and dried blood spots (DBS) for hormone assays. Seven subjects participated in five study visits (one visit/week) to contemporaneously measure LH and FSH in serum and DBS and E2 and P4 (or its primary urinary metabolite, Pd) in serum, DBS, urine, and saliva (n = 35 samples). Twenty-one additional samples were obtained from 12 subjects who completed blood draws and collected DBS and/or urine on the same day during intensive cycle monitoring (see below).

Menstrual cycle monitoring

Seven subjects from the pilot study and 16 additional subjects participated. Hormone levels were measured during two consecutive menstrual cycles from approximately cycle day 10 in cycle 1 to cycle day 20 in cycle 2. Subjects underwent one to three blood draws per week (for LH, FSH, E2, P4) in clinic according to their availability and were asked to collect DBS (for LH, FSH) and dried urine strips (for E2, Pd) at home on the remaining days. Levels from DBS and urine samples, with the exception of urine Pd, were then converted to serum-equivalents (see Data Analysis). Note that saliva samples were not used in the full protocol because of poor assay precision at the low salivary hormone levels found in adolescents (see “Results”). Inhibin B (INHB) was measured in a subset of serum samples (those with adequate sample volume) in the early follicular phase (EFP) (two to six samples per subject; 91 in total). Samples were obtained every 1 to 2 days during cycle 1 and week 1 of cycle 2. Nineteen of the 23 girls had additional hormone (17.5 ± 5.6 days) and sonographic data (2.1 ± 0.9 scans) that allowed us to determine ovulatory status in cycle 2. Data were available for 87.2% ± 3.0% of days in cycle 1 and 77.7% ± 4.3% of days in cycle 2.

One investigator (J.M.A.) performed all trans-abdominal pelvic ultrasounds (two to five/subject) using a 5-2 MHz curved transducer (Phillips HDII XE system, Andover, MA) and recorded follicle counts, maximum follicle diameter, the presence of a CL and free fluid in the posterior cul-de-sac, and ovarian volume. Follicles were scanned in multiple planes to determine maximum diameter. Ovarian volume was calculated using the formula for the volume of an ellipsoid (L × H × W × 0.523).

Subjects took daily iron supplements and hemoglobin was monitored throughout the study at four time points.

Assays

Serum

Serum was analyzed for LH, FSH, E2, P4, and total testosterone (TT) using a chemiluminescent microparticle immunoassay (CMIA) (Architect; Abbott Diagnostics, Abbott Park, IL) (31). The E2 (32) and TT (33) assays have been calibrated against liquid chromatography/tandem mass spectrometry (LC/MS). To confirm the accuracy of low P4 values by CMIA in this study, 23 samples with P4 of 0.4 to 10.3 ng/mL were reanalyzed using LC/MS (Esoterix, Inc., Endocrine Sciences, LabCorp, Calabasas Hills, CA). All P4 values <0.6 ng/mL (n = 3) by CMIA were undetectable by LC/MS. After excluding these undetectable values, however, there was a tight correlation between the two methods (slope = 0.9, intercept = 0.06, r² = 0.9). INHB was measured at the University of Virginia Center for Research in Reproduction Ligand Assay and Analysis Core using an enzyme-linked immunoassay (Beckman Coulter, Inc., Brea, CA).

DBS

LH and FSH were analyzed using a solid-phase, two-site chemiluminescent immunometric assay (Siemens, Berlin, Germany). E2 and P4 were analyzed using an enzyme immunoassay (DRG International, Inc., Springfield, NJ) (34, 35).

Dried urine strips

E2 and Pd levels were determined by gas chromatography-MS/MS (Agilent 7000B) after extraction and enzymatic hydrolysis of sulfate and glucuronide conjugates (36).

Saliva

E2 and P4 were measured using an enzyme immunoassay (DRG International, Inc.) with C-18 column extraction and concentration (34, 37).

Gonadotropins are expressed in international units per liter, as equivalents of the Pituitary Second International Reference Standard 80/552. See Supplementary Table 1 and Supplementary Methods for additional details pertaining to assay performance and sample collection and processing.

Data analysis

Comparison of serum with saliva, urine, and DBS measurements

To identify the best noninvasive methods of hormone measurement, we fit log-transformed hormone levels derived from DBS and serum (FSH, LH) or urine and serum (E2) using Passing-Bablok nonparametric regression models (Analyze-it 3.90.7, Leeds, UK) (Supplementary Figs. 1 and 2). The strongest correlations were between DBS and serum LH and FSH (both r = 0.9, P < 0.001) and between urine and serum E2 (r = 0.8, P < 0.001). The fitted models then allowed conversion of DBS and urine levels to serum-equivalents for use in the full protocol [serum FSH = 0.89*(DBS FSH^0.88); serum LH = 0.43*(DBS LH^1.16); serum E2 = 41.85*(urine E2^1.08)]. P4 in DBS and saliva and Pd in urine did not correlate well with serum P4, particularly at low levels (0 to 2 ng/mL). In the full protocol, P4 was expressed only as serum P4 [12.3 ± 1.0 (mean ± SE) samples per subject]. Urine Pd was also used (without interconversion) in the full protocol (18.6 ± 2.2 samples per subject) because the assay appeared to perform well in the range indicative of ovulation according to serum P4 (i.e., P4 > 1.65 ng/mL as described below).

Menstrual cycle classification

To establish biochemical criteria for OVs in adolescents, we examined serum P4 and urine Pd measurements from 37 cycles with (i) sonographic evidence of ovulation based on collapse of the dominant follicle, detection of a CL, and/or increased free fluid in the posterior cul-de-sac or (ii) sonographic evidence of anovulation, including absent follicle growth, regression of the dominant follicle without subsequent follicle growth, or identification of a large (≥30 mm) unruptured follicle within 2 days of menses (38–40). We then performed receiver-operator characteristic curve analysis and identified cut-offs with the highest combined sensitivity and specificity: P4 = 1.65 ng/mL (sensitivity 85%, specificity 96%) and Pd = 390.5 µg/g Cr (sensitivity 73%, specificity 96%) (Supplementary Fig. 3). Areas under the curve were 0.96 for P4 (95% CI, 0.89 to 1.0; P < 0.0001) and 0.93 for Pd (95% CI, 0.85 to 1.0; P < 0.0001).

Cycles were determined to be ovulatory if they met the following criteria: (i) a midcycle (MC) LH peak; (ii) a preovulatory E2 peak; and (iii) a serum P4 >1.65 ng/mL or urine Pd >390.5 µg/g Cr. A peak was defined as the first LH value ≥mean + 2 SD of the six preceding values, as previously described (41). Sonographic signs of ovulation also provided supporting evidence. The day of ovulation was defined as the cycle day meeting at least three of the following criteria (42): (i) day of the MC LH peak; (ii) day of the MC FSH peak; (iii) day of/day after the preovulatory E2 peak (defined as highest E2 value); (iv) day on which serum P4 doubled or exceeded 0.8 ng/mL (41), (v) 3 days before a sustained, threefold increase in urine Pd above mean follicular phase (FP) levels (modified Kassam algorithm) (43); and (vi) day of/day before follicle collapse on ultrasound. Urine Pd levels were lagged by 1 day to account for hormone metabolism and excretion (44). The length of the FP was calculated as the number of days from menses onset up to and including the estimated day of ovulation. The length of the luteal phase (LP) was calculated as the number of days after ovulation until the day before next menses.

In ANOVs, an LH peak was also identified by an LH value at least two SDs above the mean of the six preceding values (41). The length of the FP was equal to total cycle length. Luteinization of a dominant follicle was defined as a serum P ≥ 1 ng/mL (based on the average P4 within 36 hours of LH surge initiation in adults) (41) or a threefold rise in urine Pd above the mean FP level (45).

Comparisons among adolescent subgroups and with historic adult controls

Gonadotropin and sex steroids in adolescents were compared with corresponding data from 65 women, ages 18 to 34 years, with regular OVs who underwent daily blood sampling for one to two cycles (94 cycles total) (46). INHB measurements were available in 11 of these women (47). Follicle growth rates and E2 levels relative to follicle size were compared between adolescents and 45 ovulatory women who underwent two to six serial transvaginal ultrasounds and a subset (n = 21) who had concurrent daily blood samples in the week before ovulation (48).

The same assay platforms for serum gonadotropins and sex steroids in adolescents were used previously in adult controls (see Supplementary Methods). INHB in adults was measured using a different enzyme-linked immunoassay (Serotec, Oxford, UK) and levels were converted to values from the current assay (49).

Baseline characteristics were compared among adolescent groups using one-way analysis of variance. FP hormones were normalized to a 14-day FP (42) and all hormones were natural-log-transformed before analysis. For LH, FSH, and E2, cycle day numbering was centered on the MC LH peak of cycle 1 (day 0) and compared across all groups during the late FP (LFP, −5 to −2), MC (−1 to +1), LP (+2 to menses) of cycle 1 and in days 1 to 7 of cycle 2 (EFP). P4 levels were compared at MC and during the LP.

ANOVs were either centered to MC (day 0) using the LH peak (n = 3; LH 13.9, 16.9, and 24.2 IU/L, respectively), the day with the highest E2 level (n = 3; E2 84, 91, or 232 pg/mL, respectively), or the first day with evidence of luteinization of the dominant follicle (n = 1; defined as serum P4 ≥1 ng/mL or threefold rise in urine Pd) to allow for comparisons with OVs.

All group comparisons were performed using linear mixed models to account for repeated measures. Prespecified pairwise comparisons of adolescent groups and adults were performed using the two-sample t test. To capture the preovulatory E2 rise that starts ~60 hours before the MC surge (50, 51), E2 from day −3 to 0 was also compared by linear mixed models. Mean LH at MC, adjusted for the mean preovulatory E2 level, was used as an index of E2 positive feedback and was compared between groups using analysis of covariance.

Dominant follicle growth rates and maximum preovulatory follicle size during cycle 1 (day −7 to 0) were derived from a linear mixed model (52). The day on which a dominant follicle (≥10 mm) first appeared was predicted from the model and corresponding CIs estimated by the Fieller method (53). Two indices of ovarian maturity were compared across groups using linear mixed models: (i) cycle 2 EFP INHB adjusted for FSH; and (ii) FP E2 adjusted for same-day FSH and follicle diameter (cycles 1 and 2).

All analyses were conducted in JMP Pro 13.2 and SAS 9.4 (Cary, NC). Results are reported as geometric mean ± SE or the geometric mean difference (95% CI) for log-transformed data.

Results

Subject characteristics

Twenty-three healthy adolescents participated (Table 1 and Supplementary Table 2). Subjects had Tanner V breast development and a gynecologic age of 0.4 to 3.5 years (1.7 ± 0.2). They were 14.4 ± 0.2 years old and racially diverse (15 white, 5 black, 3 Asian; 25% Hispanic). Four girls (17%) were overweight and nine (39%) were obese; the remaining 10 girls (44%) were of normal weight. Subjects were nonhirsute, had no more than mild acne, and had normal TT levels (determined at the screening visit), with no differences between the groups [(mean ± SE) normal OV, 19.2 ± 3.0 ng/dL; short OV 20.0 ± 3.5 ng/dL; and ANOV 22.0 ± 5.9 ng/dL; P = 0.9].

Table 1.

Baseline Characteristics of Subjects and Comparisons of Cycle 1 Subgroups

	Ovulatory, Normal Luteal (Normal OV)	Ovulatory, Short Luteal (Short OV)	Anovulatory (ANOV)	Difference Among Groups, P-Valuea
n	11	5	7
Chronologic age, y	14.8 ± 0.8	14.0 ± 0.3	14.0 ± 0.5	0.48
Age at menarche, y	12.6 ± 0.2	12.9 ± 0.3	12.5 ± 0.2	0.58
Gynecologic age, y	1.9 ± 0.3	1.9 ± 0.5	1.5 ± 0.4	0.68
BMI percentile	79.8 ± 9.3	78.9 ± 5.7	83.3 ± 8.2	0.91
Cycle length, d	39.4 ± 3.9	25.2 ± 1.1	35.9 ± 4.4	0.10
FP length, d	27.8 ± 4.0	17.4 ± 1.2	–	0.11
LP length, d	11.6 ± 0.4	7.8 ± 0.8	–	–b
Peak serum P4, ng/mLc	7.1 ± 1.3	3.4 ± 0.4	1.4 ± 1.1	–b
Peak urine Pd, µg/g Crc	1276.7 ± 153.9	588.0 ± 89.0	316.4 ± 145.5	–b

^aOne-way ANOVA.

^bUsed to define the groups of adolescents.

^cTo convert P4 to SI units (nmol/L), multiply by 3.18; for Pd (nmol/mmol creatinine), multiply by 0.35.

Menstrual cycle patterns

Using intensive biochemical and sonographic monitoring of cycle 1, we grouped cycles into three different patterns: (i) OVs with normal (10 to 14 days) LP length (designated “normal OV”); (ii) OVs with short (5 to 9 days) LP length (“short OV”) (54, 55); and (iii) anovulatory cycles (“ANOV”) (Table 1 and Supplementary Table 2). All but one OV (cycle length, 77 days) fell within the 21- to 45-day range typical of adolescent menstrual cycles (1), as did all but two ANOVs (cycle lengths, 20 and 60 days) (Supplementary Fig. 4a). There was significant overlap in cycle length among these groups, indicating that cycle length, when within this range, is a poor predictor of ovulatory status in adolescents. In this sample, chronologic age, gynecologic age, and body mass index (BMI) percentile did not differ significantly among groups (Table 1), and there was no correlation between gynecologic age and cycle length (slope = 2.9, intercept = 30.0, r² = 0.07, P = 0.2) (Supplementary Fig. 4b). The presence of one OV did not necessarily signal reproductive maturity as four (21.1%) subjects had one OV and one ANOV during the monitoring period (57.9% had two OVs and 21.1% had two ANOVs). Cycle dynamics also appeared to influence the chance of ovulation in the subsequent cycle: subjects with an OV in cycle 1 were more likely to have another OV (in cycle 2) than subjects with an ANOV in cycle 1 (84.6% vs 33.3%, P = 0.046) (Supplementary Table 2). One subject with ANOVs and two subjects with normal OVs had enlarged ovaries (12.8 to 17.3 cm³), defined as an average ovarian volume >12 cm³ in adolescents (56).

Comparisons of hormone dynamics and folliculogenesis

Girls with OVs in cycle 1

Girls with normal OVs and short OVs demonstrated the dynamic changes in LH, FSH, E2, and P4 across the menstrual cycle that typify mature OVs in adults, including an increase in FSH during the LFT, a biphasic increase in E2, a MC LH peak, and a peri-ovulatory and postovulatory increase in P4 or Pd (Table 2; Fig. 1). However, the level of hormonal detail afforded by the current studies revealed significant differences in the magnitude of these hormone changes between adult women and adolescents.

Table 2.

Mixed Model Estimates^a of Hormone Dynamics by Phase

	LFP			MC			LP			EFP
	Mean	95% CI	P	Mean	95% CI	P	Mean	95% CI	P	Mean	95% CI	P
LH, IU/L
Adult	6.0	5.4–6.7	Reference	18.2	16.1–20.5	Reference	3.8	3.5–4.2	Reference	4.4	4.0–4.9	Reference
Normal OV	6.8	5.1–9.2	0.43	16.7	11.3–24.6	0.67	2.5	1.9–3.3	0.005	2.8	2.1–3.8	0.006
Short OV	3.4	1.9–6.1	0.06	12.7	7.3–22.1	0.21	1.5	0.9–2.3	<0.001	2.7	1.8–4.3	0.04
ANOV	7.6	5.2–11.1	0.26	7.2	4.5–11.7	<0.001	2.2	1.4–3.3	0.009	2.3	1.6–3.4	0.002
FSH, IU/L
Adult	4.8	4.5–5.2	Reference	7.3	6.7–7.8	Reference	4.0	3.7–4.2	Reference	6.5	6.1–7.0	Reference
Normal OV	4.6	3.8–5.5	0.62	5.7	4.5–7.2	0.06	1.5	1.2–1.9	<0.001	4.1	3.4–5.1	<0.001
Short OV	3.5	2.4–5.0	0.08	5.8	4.1–8.1	0.20	1.7	1.3–2.3	<0.001	4.3	3.2–5.7	0.005
ANOV	4.1	3.2–5.2	0.21	3.4	2.5–4.5	<0.001	1.8	1.3–2.3	<0.001	2.7	2.1–3.5	<0.001
E2,b pg/mL
Adult	133.2	120.7–147.0	Reference	154.4	139.2–171.2	Reference	98.2	90.8–106.3	Reference	54.0	49.3–59.2	Reference
Normal OV	51.2	40.0–65.5	<0.001	123.5	89.5–170.3	0.20	59.2	46.3–75.5	<0.001	16.9	12.9–22.2	<0.001
Short OV	77.9	48.5–125.1	0.03	133.4	82.6–215.5	0.56	64.1	43.5–94.6	0.04	26.8	18.2–39.4	<0.001
ANOV	47.7	34.5–66.0	<0.001	108.0	73.5–158.7	0.08	87.4	61.9–123.2	0.51	35.2	25.0–49.6	0.02
P4,b ng/mL
Adult				1.0	0.9–1.3	Reference	5.1	4.4–5.9	Reference
Normal OV				0.7	0.4–1.4	0.29	3.2	1.9–5.4	0.09
Short OV				0.5	0.2–1.7	0.23	2.0	0.9–4.4	0.02
ANOV				0.4	0.2–1.	0.06	1.0	0.4–2.5	<0.001

^aMeans and 95% CIs are estimated values from the mixed model, taking into account the repeated measures across participants. Log-transformed values were transformed back to original units for presentation here.

^bTo convert E2 to SI units (pmol/L), multiply by 3.67; for P4 (nmol/L), multiply by 3.18.

Figure 1.

LH, FSH, E2, and P4 levels during two consecutive menstrual cycles in adolescent girls with either normal OV (left, n = 11) or short OV (right, n = 5) cycles. Levels represent serum or serum equivalents as determined from DBS or urine samples. Cycle days are centered to the MC LH peak of cycle 1 (day 0) and shown during the first week of cycle 2 (day 1 = menses). Adolescents: filled circles are arithmetic mean values and error bars are ±1 SE. Both groups are compared with historic adult controls (n = 65). Adults: Dashed line denotes arithmetic mean level and shaded area is ±1 SD. *P < 0.05; **P < 0.01; ***P < 0.001 for cycle phase (i.e., LFP, MC, LP, EFP). To convert E2 to SI units (pmol/L), multiply by 3.67; for P4 (nmol/L), multiply by 3.18.

In normal OV adolescents (Figs. 1a–1d; Table 2; Supplementary Figs. 5a and 6a), E2 was lower compared with adults during the LFP (days −5 to −2; P < 0.001) and in the immediate preovulatory period [days−3 to 0; percent difference in means (95% CI): −35.3% (−51.5 to −13.7), P = 0.003; modeled estimate of E2 means (95% CI): normal OV, 115.0 pg/mL (87.3 to 151.4); adult, 117.7 (162.9 to 194.0)], but the MC LH peak remained robust (P = 0.2). Indeed, the magnitude of the LH peak controlled for preovulatory E2 was not different from adult cycles [percent difference in means −12.4% (−36.6 to 21.0), P = 0.4], consistent with mature estrogen positive feedback activity in normal OV girls. Despite a normal LH peak and LP length, however, normal OV girls showed a tendency toward lower P4 levels (P = 0.09) and did not attain adult levels of E2 (P < 0.001) during the LP. Normal OV adolescents showed a precipitous drop in FSH and LH after the MC LH peak. FSH and LH remained lower in normal OV than in adults during the LP (FSH, P < 0.001; LH, P = 0.005) and subsequent EFP of cycle 2 (FSH, P < 0.001; LH, P = 0.006) (i.e., the LFT). E2 levels were also lower (P < 0.001) in adolescents during the EFP of cycle 2 compared with adults.

In short OV adolescents (Figs. 1e–1h; Table 2; Supplementary Figs. 5b and 6b), E2 was not different from adults during the immediate preovulatory period [percent difference in E2 means (95% CI): −20.7% (−50.7 to 27.7), P = 0.3; modeled estimate of mean in short OV, 141.0 pg/mL (88.3 to 225.1)] but was lower than in adults when the entire LFP was included (P = 0.03). The MC LH peak was not statistically lower (P = 0.2). However, the LH peak adjusted for preovulatory E2 was less in short OV girls than adults [−37.1% (−60.2 to −0.7); P = 0.047], raising the possibility of immature positive feedback activity in this group. FSH, LH, and E2 fell rapidly after MC in short OV girls and remained lower than in adults during the LP (P < 0.05 for all) and EFP of cycle 2 (P < 0.05 for all). P4 in short OV girls was lower (P = 0.02) than in adults during a LP of 5 to 9 days’ duration.

In normal OV adolescents and adults, the dominant follicle first appeared 6 to 7 days before MC in cycle 1 [estimated median (interquartile range): 7.0 days (5.5 to 7.5) vs 7.0 days (6.0 to 7.0) in adolescents vs adults; P = 0.8], indicating that the longer FP observed in the adolescents was due to a longer time required for small antral follicle growth and/or selection of the dominant follicle rather than for dominant follicle growth [FP: 27 days (20 to 28) vs 16 days (14 to 16.5) in adolescents vs adults; P < 0.001]. The dominant follicle grew more slowly in normal OV adolescents than in adults (1.5 ± 0.2 vs 2.2 ± 0.1 mm/d; P < 0.001; Fig. 2) such that the predicted size of the preovulatory follicle was smaller in adolescents than adults (mean preovulatory follicle diameter, 20.3 ± 1.1 vs 25.2 ± 0.4 mm; P < 0.001). Follicular growth in short OV girls (1.4 ± 1.2 mm/d; preovulatory diameter, 20.6 ± 4.8 mm) largely mirrored that of normal OV girls (test for difference in regression lines for folliculogenesis, P = 0.9).

Figure 2.

Dominant follicle diameter plotted against cycle day for adults and normal OV adolescents. Lines and 95% confidence bands were estimated from a linear mixed model. Adults, dashed line; normal OV adolescents, solid line. Filled circles are individual data points for normal OV adolescents. The slope of a line is the estimated follicle growth rate; the intercept at cycle day 0 is the estimated follicle size at ovulation. The dominant follicle grew more slowly in adolescent girls with normal OV cycles than in adults resulting in a smaller estimated follicle size at ovulation in cycle 1.

Both normal and short OV adolescents had signs of ovarian immaturity that were unrelated to diminished FSH stimulation. Mid- to late-follicular E2 was lower in normal and short OV girls compared with adults even after adjusting for both lower FSH and smaller follicle size [percent difference in means: normal OV vs adult, −39.4% (−60.1 to −8.0), P = 0.02; short OV vs adult, −46.6% (−71.0 to −1.9), P = 0.04]. Cycle 2 EFP INHB was lower than in adults before (P < 0.01 for normal and short OV; Fig. 3) and after adjusting for lower FSH levels [percent difference in means: normal OV vs adult, −58.0% (−75.0 to −29.6), P = 0.002; short OV vs adult, −42.6% (−64.5, −7.3), P = 0.02].

Figure 3.

INHB levels plotted against cycle day for adults and adolescents during the EFP of cycle 2. Adults: arithmetic mean levels, (○); dashed lines delineate ±1 SE. Adolescents: individual data points, normal OV (●), short OV (△), or ANOV (□). Adult and adolescent subjects each contributed data to only a subset of cycle days. INHB levels were lower in all adolescent groups compared with adults during the first 7 d of cycle 2. INHB values are equivalent in SI units (ng/L).

Girls with ANOVs in cycle 1

ANOV adolescents demonstrated heterogeneous hormone profiles (Table 2; Supplementary Figs. 5c and 6c). Two girls had normal follicle development with a corresponding increase in E2 and a relatively robust increase in LH without ensuing follicle rupture; instead, the dominant follicle(s) continued to secrete E2 and a small amount of P4 before menses occurred. One of these subjects demonstrated persistent growth of three follicles for at least 5 days after MC, consistent with the presence of luteinized unruptured follicles (Supplementary Fig. 7). The second subject grew one dominant follicle (maximum diameter 30 mm) and had a MC LH peak of 24.2 IU/L; after MC, follicle growth ceased but E2 remained in the 130 to 150 pg/mL range and P4 rose to 3.9 ng/mL before menses occurred 12 days later. Note that without serial ultrasounds demonstrating a luteinized unruptured follicle, this cycle would have been misclassified as ovulatory. The remaining ANOV girls had either minimal follicle growth (n = 2; maximum diameter 7 and 11 mm, respectively), formed a luteinized preovulatory follicle that did not ovulate (n = 2; 27 mm and P4 1.3 ng/mL; 39 mm and P4 1.0 ng/mL), or formed a preovulatory follicle that did not luteinize or ovulate (n = 1; 22 mm, P4 0.5 ng/mL).

As a group, ANOV girls had lower E2 before MC and lower LH and FSH at MC compared with adults (Table 2). MC LH in ANOV girls remained attenuated compared with adults after adjustment for E2 [difference: −57.9% (−71.1 to −38.8), P < 0.001]. FSH and LH levels in ANOV girls fell after MC but E2 secretion was preserved relative to adults. In these ANOV girls, E2 was presumably not derived from the CL, but from one or more luteinized follicles that continued to produce E2 for several days until menses (Supplementary Fig. 7). In the EFP of cycle 2, LH, FSH, and E2 were lower in ANOV girls than in adults (Table 2).

ANOV girls as a whole did not differ from the two OV groups in dominant follicle growth rate or follicle size at MC (0.9 ± 0.3 mm/d; 20.0 ± 1.5 mm; test for difference in regression lines, P = 0.5), but EFP INHB was lower in ANOV girls than in adults both before (P < 0.01; Fig. 3) and after adjusting for lower FSH [difference: ANOV vs adult −61.7% (−76.9% to −36.7%), P < 0.001], suggesting ovarian immaturity.

Discussion

We performed detailed reproductive phenotyping of a contemporary cohort of 23 healthy early postmenarchal girls to gain insight into the normal development of the hypothalamic–pituitary–ovarian axis. Approximately three-quarters demonstrated at least one cycle in which ovulation occurred, but many of these OVs were characterized by luteal insufficiency (short LP length and/or lower P4 or Pd compared with adult women). Hormone dynamics in ANOVs were variable, as has been noted by others (14, 57), and included minimal or absent follicle growth, growth of a luteinized preovulatory follicle that was not followed by ovulation and CL formation, and/or a diminished LH peak at MC. We demonstrate that the luteinized unruptured follicle syndrome is an abnormality that falls within the adolescent luteal insufficiency spectrum. Importantly, we found that the threshold P4 value indicative of ovulation in adolescents is ~1.65 ng/mL, well below the commonly used adult standard of 4 to 5 ng/mL, suggesting that many adolescent cycles previously classified as anovulatory may have instead been ovulatory with luteal insufficiency. The episodic release of P4 by the CL (58) in the setting of slow frequency LH pulses during the LP suggests further that more frequent measurement of P4 in the LP, complemented by pelvic ultrasounds, will be necessary to accurately classify adolescent cycles in future studies, and in particular, to distinguish adolescents with short OV cycles from those with ANOV cycles with luteinized unruptured follicles. OV and ANOV cycles appear to be indistinguishable based on any routine parameters readily available to the clinician (e.g., patient’s cycle length, gynecologic age, chronologic age, or BMI).

The ultimate step in female reproductive axis maturation has traditionally been deemed the acquisition of estrogen positive feedback: mature GnRH secretion leads to mature gonadotropin secretion and normal folliculogenesis; growth of a healthy dominant follicle then provides the exponential increase in estrogen levels required to induce an LH surge and consequent ovulation (59, 60). The current studies, however, demonstrate that even adolescent girls who have achieved this milestone (i.e., OVs, a robust LH peak at MC, and normal LP length) continue to have immature FSH and LH dynamics compared with normal adults, suggesting the GnRH secretory profile or pituitary responsiveness is not fully mature. Previous studies reported that FSH increases more slowly in ovulatory adolescents than in adults during the FP but did not compare absolute FSH levels (or LH levels) between adolescents and adults or assess gonadotropin secretion during the LFT (61). A small study of daily urinary hormone measurements in early postmenarchal girls compared with adults found no differences in FSH and LH, but groups were compared using a 3- to 6-month integrated measure without consideration of cycle phase (14). In the current study, we observed a profound decrease in FSH and LH levels compared with adults during the LFT, even in girls with normal OV cycles, which was presumably responsible for lower E2 and INHB in the EFP of the subsequent cycle in girls. The slower growth rate of the dominant follicle and smaller predicted preovulatory follicle size in adolescents compared with adults in the current study and in previous studies (61, 62) are also consistent with diminished trophic stimulation by FSH in adolescents. Thus, although the robust LH peak at MC in ovulatory adolescents demonstrates that the pituitary is capable of storing and releasing LH in response to GnRH, lower FSH (and also LH) during the LFT suggest residual hypothalamic and/or pituitary immaturity. No previous studies have rigorously investigated hormone dynamics during the LFT in adolescents. While several studies carefully documented hormone profiles in girls with frequent blood or urine sampling, they did not examine intercycle hormone dynamics (13, 14, 61, 63–67). The only reference to a potential immaturity of FSH dynamics during the LFT was made by Hayes and Johanson (63) who noted that the late luteal rise in FSH characteristic of adult cycles was only observed in girls with a gynecologic age of at least 3 years.

Although our data suggest that diminished FSH stimulation contributes to the decreased follicular growth rate in adolescents, there is also evidence for intrinsic ovarian immaturity. E2 was significantly lower during the FP in adolescents even after adjusting for their smaller follicle sizes, suggesting a developmental discordance between FSH-driven follicle growth and aromatization and/or LH-driven androgen synthesis. We had insufficient data on other potential markers of follicle maturity (P4, inhibin A) relative to follicle size to further address the possibility of ovarian immaturity. These data provide support for previous in vitro studies: Anderson et al. (68) demonstrated that secondary follicles and oocytes isolated from ovarian cortical biopsies from early postmenarchal girls grow more slowly than follicles from adult women under the same culture conditions of insulin and recombinant FSH. Inadequate follicle development is also suggested by lower luteal P4 in ovulatory girls compared with adults as the quality of the CL is thought to reflect the quality of its predecessor, the dominant follicle (69).

This study has several limitations. Given the intensive nature of these studies, particularly in pediatric subjects, we used several noninvasive, in-home sampling modalities to augment clinic blood draws and ultrasounds. Even with these additional methods, we were unable to capture hormone data and follicle size on every day. Our pilot study demonstrated strong correlations between serum and DBS for gonadotropins and between serum and urine for E2. Although P4 salivary assays have gained increasing popularity in the fertility industry, in our hands, salivary P4 could not distinguish ANOVs from OVs in adolescents. It is unclear if this was related to the assay itself or to poor collection technique (i.e., recent chewing gum, food, or liquids), but it does suggest that studies that classified adolescent menstrual cycles based solely on salivary P4 using adult cutoffs should be interpreted with caution. Similarly, we found that urine Pd was only a reliable indicator of serum P4 in the mid- to upper range (≈6 to 20 ng/mL). We compared adolescents with historic as opposed to contemporary adult controls because we had access to a large database of daily blood samples across the menstrual cycle in these very well-characterized healthy women. Although assays of serum gonadotropins and sex steroids were conducted in the same laboratory (the Reproductive Endocrine Unit Reference Laboratory at Massachusetts General Hospital), we cannot rule out the possibility that the differences observed between the two age groups may be related, in part, to different assay conditions or to differences in gonadotropin glycosylation patterns. Ultrasounds were performed trans-vaginally in adult controls and trans-abdominally in the current subjects; however, this procedural change was unlikely to introduce bias, as all scans in adolescents and in adults were performed by the same senior ultrasonographer (J.M.A.). Furthermore, antral follicle counts, which require a transvaginal approach, were neither attempted nor relevant to the current study. We studied each subject over 2 months with intensive monitoring but captured only a snapshot of the pattern of changes that occur between menarche and the establishment of mature OVs. Finally, this study was not powered to determine whether chronologic age, gynecologic age, race, or BMI influence cycle dynamics in adolescent girls although each of these questions is of interest.

In conclusion, the current studies demonstrate the heterogeneity of hormone profiles that underlie irregular menses in the early postmenarchal years and highlight immaturity of both FSH secretion and ovarian responses in adolescents. These data suggest further that maturation of all components of the hypothalamic-pituitary-ovarian axis is required to achieve one of the most critical milestones of adolescence: the mature OV of a fertile woman.

Glossary

Abbreviations:

ANOV

anovulatory cycle

BMI

body mass index

CL

corpus luteum

CMIA

chemiluminescent microparticle immunoassay

DBS

dried blood spot

E2

estradiol

EFP

early follicular phase

FP

follicular phase

INHB

inhibin B

LC/MS

liquid chromatography/tandem mass spectrometry

LFP

late follicular phase

LFT

luteal-follicular transition

LP

luteal phase

MC

midcycle

OV

ovulatory cycle

P4

progesterone

Pd

pregnanediol

TT

total testosterone