Energy Availability Is Associated With Luteinizing Hormone Pulse Frequency and Induction of Luteal Phase Defects

Abstract

Objective

Determine the interrelations between reductions in energy availability (EA), luteinizing hormone (LH) pulse frequency, and the induction of menstrual disturbances in previously sedentary, ovulatory women.

Methods

Secondary analysis of a randomized controlled trial consisting of a 3-month controlled diet and supervised exercise program. EA was calculated daily by measured energy intake (kcal) and exercise energy expenditure (kcal) normalized to fat-free mass (kg) and averaged during baseline and each of 3 intervention menstrual cycles. Blood samples were obtained every 10 minutes for 24 hours in the early follicular phase before the intervention and after 3 months of diet and exercise (n = 14). LH pulse dynamics were assessed by Cluster. Linear mixed models determined whether EA predicts LH pulse frequency and LH pulse frequency predicts luteal phase defects (LPDs).

Results

Subjects were 20 ± 1 years old, 165.1 ± 1.4 cm tall, and weighed 58.9 ± 1.5 kg. LH pulse frequency decreased from 0.82 ± 0.06 pulses/h to 0.63 ± 0.09 pulses/h (P = 0.048) as a result of the intervention which produced modest (-3.2 ± 0.6 kg) weight loss. EA, averaged across a menstrual cycle, predicted LH pulse frequency (P = 0.003) such that a single-unit decrease in EA was associated with a 0.017 pulses/h decrease in LH pulse frequency. LH pulse frequency in cycles with LPDs was 49% of that observed in cycles with no menstrual disturbances and for every 0.1-unit decrease in LH pulse frequency, the odds of having an LPD were 22× greater than having an optimal ovulatory cycle (P = 0.01).

Conclusions

Modest reductions in EA over a prolonged period are associated with decreased LH pulse frequency and the induction of menstrual disturbances.

Reproductive function in women is dependent on an operational hypothalamic-pituitary-gonadal axis, wherein gonadotropin-releasing hormone released from the hypothalamus in turn stimulates the pulsatile release of luteinizing hormone (LH) from the pituitary to induce ovulation. In situations of low energy availability (EA), in which there is inadequate energy intake to support all physiological functions, fuel is repartitioned away from metabolically costly processes including growth and reproduction in favor of those most necessary for immediate organism survival (i.e., cellular maintenance, thermoregulation, and locomotion) (1). As a result of this fuel repartitioning, the reproductive axis can be suppressed, leading to the development of menstrual disturbances. In fact, there is evidence for a causal role of low EA in the induction of menstrual disturbances (2, 3) that can range in severity from subclinical luteal phase defects to clinical disturbances including oligomenorrhea and amenorrhea. Menstrual disturbances present a clinical concern in that the associated hypoestrogenic state contributes to long-term health risks such as impaired bone health (4).

The mechanistic link between low EA and menstrual disturbances is proposed to act through a slowing of LH pulse frequency, which is a proxy indicator of decreased gonadotropin-releasing hormone pulsatility (5). Initial cross-sectional investigations observed conflicting results and report low (6), high (7), and similar (8–10) basal LH concentrations in oligo-amenorrheic athletes compared with healthy women. However, when examining the pulsatile secretion of LH, exercising women with menstrual disturbances displayed slower LH pulse frequency compared with cyclic athletes and with sedentary women with regular menstrual cycles (11–13). Further, short-term interventions demonstrated that the pulsatile release of LH in women is disrupted in response to 5 days of EA restriction below 30 kcal/kg lean body mass (LBM)/d (14, 15). However, it remains unclear how reduced EA over a longer duration relates to LH pulsatility and how a slowing of LH pulsatility translates to alterations in menstrual cycle quality.

We previously conducted a 3-month, prospective, randomized controlled trial (RCT) to examine the role of energy deficiency resulting from diet and exercise on menstrual cyclicity and reproductive function in premenopausal, untrained women. The RCT study design allowed us to explore the underlying mechanisms that accompany the very beginning stages of reproductive dysfunction caused by low EA. We have documented that this intervention produced modest weight loss, induced menstrual disturbances (16), and influenced LH pulsatility (17), but we do not yet know how changes in EA across the 3-month intervention are related to suppressed LH pulse dynamics, and, in turn, how changes in LH pulsatility relate to the induction of menstrual disturbances in the context of exercise and caloric restriction. As such, the purpose of this investigation was 1) to assess how EA, averaged from daily measures over a menstrual cycle, relates to LH pulse frequency measured within the same or a subsequent menstrual cycle and 2) to determine if LH pulse frequency is associated with the induction of actual menstrual disturbances. We hypothesized that EA would be positively associated with LH pulse frequency and that reductions in LH pulse frequency would be related to the induction of menstrual disturbances.

Methods

Study design

This investigation was part of a larger, prospective, RCT consisting of a 3-month controlled diet and supervised exercise program designed to assess the role of energy deficiency (-0%, -15%, -30%, -60%) on menstrual function (16). Premenopausal, sedentary women were recruited and inclusion criteria were 1) nonsmoking, 2) no history of serious medical conditions, 3) no evidence of disordered eating or current eating disorder, 4) age 18 to 30 years, 5) weighing 45 to 75 kg with body fat 15% to 35% and body mass index (BMI) 18 to 25 kg/m², 6) no medication use that would alter metabolic hormone levels, 7) no hormonal contraceptive use for previous 6 months, 8) no significant weight loss (±2.3 kg) in previous year, 9) <1 h/wk of purposeful aerobic exercise for past 6 months, and 10) documentation of at least 2 ovulatory menstrual cycles during screening. All subjects signed an informed consent form approved by the Pennsylvania State University.

Subjects completed a 3-month diet and exercise intervention following a screening and 4-week baseline monitoring period. The screening process lasted 2 to 3 menstrual cycles and included assessments of medical history, menstrual history, physical activity, eating attitudes, and psychological status. An endocrine screening panel was performed to rule out other endocrinopathies that may influence menstrual function. Following the baseline monitoring period, subjects were randomly assigned to either a control group that did not exercise, a group that exercised but was fed supplemental calories to maintain their weight, or 1 of 4 groups of varying degrees of energy deficit designed to produce modest weight loss. Subjects included in this analysis were those who were randomized to an exercising group and completed the 24-hour blood sampling at the start of and following the intervention.

Menstrual status

Menstrual calendars were used throughout the study to record menstrual cycle length and symptoms. Daily urine samples were collected during the baseline and 3 intervention cycles according to previously published methods (18). Analysis of estrogen and progesterone metabolites, estrone-1-glucuronide (E1G) and pregnanediol glucuronide (PdG), and midcycle LH were used to determine ovulatory status and the presence of luteal phase defects (LPD) (16, 19, 20). Luteal phase defects were either short luteal phases (≤8 days in length, defined as the interval between the urinary LH surge and start of the next menses (20)) or inadequate luteal phases (PdG peak concentration <5.0 µg/ml) (16). Anovulatory cycles were defined as cycles that did not have an adequate preovulatory E1G peak, did not have a midcycle LH surge, and no luteal rise in PdG above 2.49 µg/mL, and oligomenorrheic cycles were those that lasted ≥36 days in length (16). Cycles in which no menstrual disturbances were observed are referred to as being optimal ovulatory. Depending on cycle length and hormonal characteristics, it is possible that a single cycle may have multiple disturbances, for example a cycle could be classified as being both oligomenorrheic and having a LPD.

Energy balance and energy availability

Baseline energy needs were assessed during the early follicular phase of the baseline cycle as the sum of resting metabolic rate (RMR) and daily physical energy expenditure assessed by accelerometer (21). A one week “calibration” period was used to adjust food intake if a participant’s body weight varied. During the intervention, those assigned to the energy deficit group were provided a diet with a macronutrient composition of 55% carbohydrate, 30% fat, and 15% protein and lower in calories than what would be required to maintain baseline body weight, whereas the caloric intake for those in the exercising control group was intended to maintain body weight. Participants also began supervised exercise training 5 days/week at an intensity of 70% to 80% maximal heart rate determined from maximal exercise tests. Total exercise energy expenditure was determined by using the OwnCal feature on the Polar S610 heart rate monitor (Polar Electro Oy, Kempele, Finland) (22).

The initial RCT design relied on calculated energy balance to induce varying degrees of energy deficiency. To estimate energy balance during the intervention, measurements of intake, RMR, non-exercise physical activity, and exercise training energy expenditure were repeated and used to calculate daily values. The calories for RMR, non-exercise physical activity, and exercise were summed to represent 24-hour energy expenditure. RMR was assessed during the follicular phase of baseline, intervention cycle 2, and the poststudy period. Non-exercise physical activity was calculated every other week from subjects wearing the RT3 for 24 h/d for 7 days (except for training bouts, swimming, and bathing). Exercise was measured during each training bout as described above. Energy balance was calculated daily as (intake [kcal] – [RMR + nonexercise PA + Ex (kcal)]) using the most recent data available and updated continually throughout the intervention as new measurements were obtained.

For the current investigation, EA was utilized as the index of energetic status in order to be related to previously conducted short-term studies (4-5 days) on LH pulsatility (14, 15). Energy availability for a given day was calculated each day as (energy intake [kcal] – (exercise energy expenditure [kcal] – RMR [kcal/min] × duration of exercise bout [minutes])/fat-free mass [FFM] [kg]). On nonexercise days, EA was calculated based on an assumed exercise energy expenditure of 0 (21). Determinations of daily EA incorporated new measures of RMR and/or FFM as soon as those measurements occurred. Daily EA was then averaged across each menstrual cycle.

Body composition

Body weight was measured to the nearest 0.01 kg using a digital scale (Seca, Hamburg, Germany) twice per week wearing standard shorts and t-shirt. Underwater weighing was used to determine body composition during the follicular phase (days 1-7) in the baseline and poststudy periods ,as described previously (23).

Twenty-four hour serial blood sampling

Before and following the 3-month intervention, subjects reported to the General Clinical Research Center at 7:30 am during the early follicular phase of their menstrual cycle following an overnight fast and having not exercised for the previous 24 hours. An intravenous catheter was inserted into a forearm vein and blood samples were obtained every 10 minutes for 24 hours. Subjects remained in a supine position with their upper body and head slightly elevated. All postural changes were recorded and meals were provided at 9:00 am (breakfast), 12:00 pm (lunch), 6:00 pm (dinner), and 9:00 pm (snack) that were consumed within 30 minutes. Total calories over the day represented 85% of each subject’s weight maintenance intake to account for negligible physical activity during the procedure. Each subject consumed a 500-calorie dinner and the rest of the daily calories were distributed such that 43% were at breakfast, 49% at lunch, and 4% at snack. The macronutrient composition of the food over the entire day averaged 55% carbohydrate, 30% fat, and 15% protein (17). Food intake during the preintervention and postintervention 24-hour repeated sampling procedures was identical in meal and total caloric content and in macronutrient composition. Preintervention serial samples were obtained during the baseline (n = 8) and at the beginning of intervention 1 (n = 6) cycles and the postintervention serial samples were obtained after the end of the intervention (n = 14). All 24-hour sampling was performed in the early follicular phase (5.4 ± 0.2; range 2-8 days).

Assays: LH

LH was assessed in serum samples from the 24-hour serial blood draw at 10-minute intervals from 8:00 am to 8:00 am according to the Siemens (Deerfield, IL) kit for Immulite. Assay sensitivity is 0.1 mIU/mL. The intra-assay and inter-assay coefficients of variation are 5.7% and 12.3%. All samples from a given subject were assayed in the same run.

Data analysis

Twenty-four hour LH concentrations were analyzed for pulse frequency using the pulse detection algorithm Cluster (CLUSTER 8) (24). A 2 × 1 pulse configuration using a t statistic value of 2.0 for upstroke and downstroke was used and missing data were linearly interpolated between 2 adjacent values (17). Data were screened before analysis to test statistical assumptions and identify outliers. Paired t tests were used to determine changes from pre- to postintervention. Generalized linear mixed models were used to determine if EA is predictive of LH pulse frequency measured within the same or a subsequent menstrual cycle, and whether LH pulse frequency is predictive of the presence of LPDs in the same or prior cycle. Whenever possible, EA and menstrual cycle status of the same cycle in which LH was measured were used for analysis. Only in situations in which energy or menstrual status data were not available (post) were the previous cycle data (EX3) used (n = 14 of 28 observations). All analyses were performed using SPSS (v. 24.0; SPSS, Armonk, NY). All data are expressed as means ± standard error of the mean and P ≤ 0.05 was considered statistically significant.

Results

Descriptive and energetic characteristics at the start and end of the intervention are presented in Table 1. As expected, the intervention resulted in significant decreases in weight, BMI, and fat mass (P < 0.001). Average weight loss was 3.2 ± 0.6 kg, and due primarily to decreases in fat mass (-3.1 ± 0.5 kg) because FFM (-0.1 ± 0.5 kg) did not change across the intervention (P > 0.05). Energy availability decreased, as designed, across the intervention from 38.3 ± 2.0 kcal/kgFFM/d to 28.1 ± 2.4 kcal/kgFFM/d (P < 0.001) and averaged 30.9 ± 2.1 kcal/kgFFM/d for all subjects across the intervention.

Table 1.

Demographic and Energetic Characteristics (n = 14 subjects)

	Pre	Post	P Value
Age (y)	20.4 ± 0.1	----	----
Height (cm)	165.1 ± 1.4	----	----
Age of menarche (y)	11.9 ± 0.4	----	----
Gynecological age (y)	8.4 ± 0.8	----	----
Weight (kg)	58.9 ± 1.5	55.3 ± 1.3	<0.001
BMI (kg/m2)	21.6 ± 0.6	20.4 ± 0.6	<0.001
Body fat (%)	28.2 ± 1.3	24.2 ± 1.3	0.099
Fat mass (kg)	16.6 ± 0.9	13.5 ± 1.0	<0.001
Fat free mass (kg)	42.0 ± 1.0	41.8 ± 0.9	0.778
EA (kcal/kgFFM/d)*	38.3 ± 2.0	28.1 ± 2.4	<0.001
Energy deficit (%)	----	-34.7 ± 8.3	----
VO2max (kcal/kg/min)	36.9 ± 1.2	43.0 ± 1.6	<0.001

*Pre values represent average EA for the cycle in which LH was assessed, either baseline (n = 8) or Intervention cycle 1 (n = 6).Abbreviations: BMI, body mass index; EA, energy availability; FFM, fat-free mass; LH, luteinizing hormone; VO₂max, maximal oxygen consumption.

Bold indicates a significant difference (P < 0.05) pre vs. post.

Menstrual cycle changes

Reproductive characteristics are presented in Table 2. There was no change in overall cycle length or follicular phase length from baseline to intervention cycle 3 (P > 0.05); however, there was a significant shortening of the luteal phase as a result of the intervention (P = 0.007). Over the course of the intervention, 8 women (57%) developed any type of menstrual disturbance. Across all menstrual cycles observed, the most common disturbances were LPDs (n = 4 subjects, 12 cycles), followed by oligomenorrhea (n = 4 subjects, 5 cycles) and then anovulation (n = 3 subjects, 4 cycles). Four cycles, in 4 subjects, had a combination of disruptions (oligomenorrhea/luteal phase defect, oligomenorrhea/anovulation). One subject did have evidence of luteal phase deficiency during their baseline cycle. Statistical tests were run with and without this participant with similar results; therefore, the subject remains in all analyses.

Table 2.

Menstrual Cycle Characteristics of All Subjects Across the Intervention

	Baseline			Cycle 1			Cycle 2			Cycle 3
Subject	Follicular Phase Length	Luteal Phase Length	Cycle Length	Follicular Phase Length	Luteal Phase Length	Cycle Length	Follicular Phase Length	Luteal Phase Length	Cycle Length	Follicular Phase Length	Luteal Phase Length	Cycle Length
1	16	13	29	18	14	32	17	12	29	18	13	31
2	14	13	27	18	11	29	16	15	31	11	14	25
3	17	13	30	17	11	28	19	12	31	23	10	33
4	17	12	29	15	11	26	30	6	36	21	7	28
5	23	12	35	21	15	36	----	----	59	----	----	----
6	21	6	27	27	3	30	29	8	37	32	2	34
7	15	11	26	14	11	25	21	7	28	15	13	28
8	13	15	28	----	----	30	----	----	28	14	13	27
9	18	11	29	24	4	28	20	6	26	17	7	24
10	13	15	28	16	13	29	13	15	28	16	11	27
11	14	13	27	15	15	30	16	11	27	24	10	34
12	21	10	31	16	10	26	22	3	25	19	3	22
13	14	13	27	15	12	27	14	18	32	15	12	27
14	14	13	27	14	10	24	23	11	34	----	----	41
Mean ± SEM	16.4 ± 0.9	12.1 ± 0.6*	28.6 ± 0.6	17.7 ± 1.1	10.8 ± 1.1	28.6 ± 0.8	20.0 ± 1.6	10.3 ± 1.3	32.2 ± 2.3	18.8 ± 1.6	9.6 ± 1.2*	29.3 ± 1.4

Each line represents an individual subject; length of phase and cycle are expressed in days.

Abbreviation: SEM, standard error of the mean.

*Significant difference (P < 0.05) baseline vs. cycle 3.

LH

LH data are presented in Table 3. As a result of the intervention, and demonstrated in Fig. 1, LH pulse frequency was reduced from preintervention levels (P = 0.048). Further, EA averaged over an entire menstrual cycle was determined to be a significant positive predictor of LH pulse frequency as assessed in the same or subsequent menstrual cycle (n = 28 cycles; P = 0.003) and is represented in Fig. 2. Specifically, a single-unit decrease in EA was associated with a 0.017 pulses/h decrease in LH pulse frequency.

Table 3.

LH Characteristics of Each Cycle (n = 14)

	Pre	Post
LH pulse frequency (pulses/h)	0.82 ± 0.06	0.63 ± 0.09
LH pulse amplitude (mIU/mL)	9.1 ± 0.6	9.8 ± 1.0
LH mean (mIU/mL)	5.0 ± 0.4	4.3 ± 0.5
Day of sampling	5.6 ± 0.5	5.3 ± 0.3

For LH statistics, 8 subjects’ samples were collected during the baseline period and 6 subjects’ samples were collected at the start of Intervention cycle 1 before the beginning of the exercise intervention; data from baseline and Intervention cycle 1 were combined and are presented as baseline.

Abbreviation: LH, luteinizing hormone.

Bold indicates a significant difference (P < 0.05) pre vs. post.

Figure 1.

Individual 24-hour LH pulse profile during pre- and postintervention. Subject was in an energy deficit group in which energy availability declined from 33.4 kcal/kgFFM/d to 26.3 kcal/kgFFM/d and LH pulse frequency decreased by 0.36 pulses/h. *Pulses identified by Cluster analysis (24). Abbreviations: FFM, fat-free mass; LH, luteinizing hormone.

Figure 2.

Scatterplot demonstrating the association between EA and LH pulse frequency. Based on mixed model analysis, n = 28 cycles with each subject included twice. EA was determined to be a positive predictor of LH pulse frequency. Abbreviations: EA, energy availability; FFM, fat-free mass; LH, luteinizing hormone.

LH pulse frequency did not predict the presence of any type of menstrual disturbance within the same or previous cycle (n = 28 cycles; P = 0.10), but did relate to the most commonly observed menstrual disturbance, LPDs (n = 25 total cycles; 6 LPDs; P = 0.01) such that for every 0.1-unit decrease in LH pulse frequency, the odds of having an LPD were 22 times greater than having an optimal ovulatory cycle. Average LH pulse frequency in optimal ovulatory cycles (n = 19 cycles) was 0.85 ± 0.06 (range, 0.15-1.18) and, in cycles with an LPD (n = 6 cycles), average LH pulse frequency was 0.42 ± 0.12 (range, 0.00-0.79) as shown in Fig. 3. LH pulse frequency in cycles with an LPD was 49% of the LH pulse frequency in cycles that did not have any type of menstrual disturbance.

Figure 3.

Difference in LH pulse frequency between cycles classified as optimal ovulatory and as LPDs. Based on mixed model analysis, LH pulse frequency was determined to significantly predict the presence or absence of an LPD. Dots represent n = 25 cycles, with 11 participants having 1 cycles included (pre, post), which could either both be classified as optimal ovulatory or LPD, or 1 as optimal ovulatory and 1 as LPD. Abbreviations: LH, luteinizing hormone; LPD, luteal phase defect.

Discussion

The results of this investigation highlight that reducing EA by 10 units from 38 kcal/kgFFM/d to 28 kcal/kgFFM/d via an intervention similar to that which may be used by women trying to lose weight, is sufficient to significantly reduce LH pulse frequency and increase the likelihood of developing LPDs. This investigation also extends the initial findings of Loucks et al. (14, 15) who reported that reductions in EA over 4 to 5 days were sufficient to reduce LH pulse frequency, by examining longer term changes in EA over 3 to 4 menstrual cycles. Our findings confirm that reductions in LH pulse frequency are related to an increased likelihood of developing a menstrual disturbance (LPD). Specifically, this 3-month RCT of controlled diet and exercise reduced EA, resulting in decreased LH pulse frequency and the induction of clinically detectable oligomenorrhea and subclinical LPDs and anovulation.

Decreasing EA from 38 kcal/kgFFM/d to 28 kcal/kgFFM/d, a 26% decline, over the course of a 3-month controlled diet and exercise program significantly reduced LH pulse frequency by 23% with a nonsignificant 8% increase in LH pulse amplitude and nonsignificant 14% reduction in mean LH. We observed that the relation between LH pulse frequency and EA was linear across a physiological range of EA. Comparatively, Loucks et al. (14, 15) demonstrated that reducing EA from 45 kcal/kgLBM/d to below 30 kcal/kgLBM/d for 5 days reduced 24-hour LH pulse frequency by 10% to 32% and increased LH pulse amplitude by 21% to 36%. Specifically, 5-day reductions in EA by 33%, 56%, and 78% suppressed LH pulse frequency by 0, 18%, and 32%, respectively (14). The aforementioned findings suggest that milder declines in EA (26% vs. 33%-78%; 38 to 28 kcal/kgFFM/d vs. 45 to 30, 20, 10 kcal/kgLBM/d), maintained over a longer period (for at least 3 months vs. 4-5 days), may cause similar or greater reductions in LH pulse frequency, but perhaps smaller increases in LH amplitude compared with short-term severe energy restriction. Our findings also suggest, contrary to Loucks et al. (14), that declines in LH pulse frequency occur along a range of EA in a dose-response fashion that does not involve being below a particular threshold of EA. The absence of a clear threshold below which LH pulse frequency abruptly declines is visually and statistically supported by our findings such that no clear threshold is evident in Fig. 1 and that a significant linear model was fit between EA and LH pulse frequency. As such, prospective measurements of EA over variable periods in exercising subjects appear to be necessary to improve our understanding of the downstream effects of changes in EA on reproductive function.

Because only LH pulsatility was assessed as a surrogate marker of reproductive function in the studies by Loucks et al. (14, 15), it remained unclear how short-term reductions in EA for 4 to 5 days because of diet and exercise, and the resultant slowing of LH pulse frequency, would translate into the actual induction of menstrual disturbances. In the current investigation, LH pulse frequency significantly predicted the presence of an LPD compared with an optimal ovulatory menstrual cycle in which no menstrual disturbances were present. Additionally, cycles in which LPDs were observed had an LH pulse frequency 51% lower than those classified as optimal ovulatory, a difference in LH pulse frequency that is greater than what was observed by Loucks et al. (14) when EA was decreased from 45 to 10 kcal/kgLBM/d. Further, based on the current investigation, a 10-unit decline in EA maintained over 3 months, which is similar in magnitude to what may be implemented by women to cause moderate weight loss compared with short-term, severe laboratory-implemented restriction that may be difficult to sustain for longer than a few days, is sufficient to reduce LH pulse frequency by 0.17 pulses/h and subsequently translates into 37× greater odds of having an LPD than an optimal ovulatory menstrual cycle. It is hypothesized that larger declines in EA would likely result in more severe and or frequent menstrual disturbances because the severity of menstrual disturbance is often consistent with the magnitude of energy deficiency such that the most severe menstrual disturbances (i.e., amenorrhea) also have the most severe signs of energy deficiency (25), and the frequency of menstrual disturbances increases in accordance with a greater magnitude of energy deficiency (16). The association between LH pulse frequency and menstrual cycle status further highlights the importance of monitoring EA over an extended period (as opposed to a 1-point-in-time measure) because declines in EA and LH pulse frequency are related to the induction of menstrual disturbances.

Both the current investigation and the previous work conducted by Loucks et al. used untrained, sedentary women; however, Loucks et al. (14, 15) controlled baseline EA to be set at 45 kcal/kgLBM/d, whereas baseline EA in the current investigation varied among individuals and averaged 38 kcal/kgFFM/d. There were also slight differences in how EA was calculated between studies. In our investigation, we defined exercise energy expenditure as that above RMR rather than typical nonexercise expenditure. In light of these differences, we previously found that the difference between using RMR compared with nonexercise expenditure was not significant and found at most an average difference of 1.2 kcal/kgFFM/d between methods (21). Both investigations also used EA as the primary measure of energy status. EA, as an index, has been related to menstrual disturbances (21), but has been questioned methodologically in that it does not account for changes in nonexercise-associated thermogenesis. Energy deficit, an alternative measure of energy status that is based on energy balance, has been related to the frequency of menstrual disturbances (16) and was assessed as part of this investigation. Interestingly, this investigation found that average EA and energy deficit were highly related (r = 0.893, P < 0.001) with average EA across the entire intervention being 31 kcal/kgFFM/d that corresponded to an energy deficit of -31%. It is also possible that differences in our results compared with those of Loucks et al. (14, 15), such as the greater magnitude of change in LH pulse frequency and no significant increase in LH pulse amplitude, could be due not only to the length and degree of restricted EA, but also to differences in how EA was achieved through manipulations in diet (caloric content and macronutrient composition) and exercise (intensity and duration).

Strengths of this investigation include the initial RCT design, the assessment of 4 successive menstrual cycles including a baseline monitoring period and the 3-cycle intervention, and confirmation of menstrual status by assessing daily ovarian hormone metabolites. Menstrual cycle status can be inconsistent, and assessing a single menstrual cycle will underestimate the number of abnormal cycles (19). Additionally, this investigation further highlights the importance of screening for subclinical menstrual disturbances such as LPDs and anovulation, both of which were induced as part of this intervention, but are not readily observable without detailed hormonal analyses. The primary limitation of this investigation is that LH was sampled during the poststudy period during which energy and menstrual data were not available for the whole poststudy menstrual cycle. As such, we relied on the energetic and menstrual cycle data from the previous cycle (EX3). Also, many menstrual disturbances occurred in cycles before LH sampling and so we were unable to observe LH dynamics during the first cycle in which menstrual disturbances were induced. Additionally, although a thorough screening process was conducted, we could not rule out other subtle causes of menstrual disturbances such as unappreciated ovulatory polycystic ovary syndrome because we did not routinely assess testosterone concentration or perform ovarian ultrasounds. Last, this investigation was specifically designed to test the effect of energetic status on the induction of menstrual disturbances, but we acknowledge that there are other contributing factors to the development of menstrual disturbances, including psychosocial stress (26), gynecologic age (27), and genetics (28).

In summary, this investigation is the first long-term prospective study examining the effect of reductions in EA on LH pulse characteristics and demonstrates the importance of maintaining EA at appropriate levels, on an individual basis and over a sustained period, to maintain reproductive health. As evidenced by our 3-month RCT study design, reducing EA from 38 to 28 kcal/kgFFM/d over at least 3 months may have similar or even greater effects on reproductive function compared with short-term, acute changes in EA. Reductions in EA can consequently contribute to suppression of the reproductive axis as evidenced by a slowing of LH pulse frequency and the development of menstrual disturbances, which are known to negatively influence overall health and well-being.

Glossary

Abbreviations

BMI

body mass index

EA

energy availability

FFM

fat-free mass

LBM

lean body mass

LH

luteinizing hormone

LPD

luteal phase defect

PdG

pregnanediol glucuronide

RCT

randomized, controlled trial

RMR

resting metabolic rate.

Additional Information

Disclosure Summary: The authors have no conflicts of interest to disclose.

Data Availability

The datasets generated during and/or analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request.