Methods

Experimental Approach to the Problem

To assess the effects of menstrual cycle phase, symptoms, motivation to train, and readiness to perform on kinematic outputs, a prospective longitudinal repeated-measures study design was implemented. For each participant, the study duration spanned four menstrual cycles, with participant's individual timeline determined by their menstrual cycle length. This included two baseline cycles, followed by two training cycles, which involved 22 resistance training sessions completed across approximately 8 weeks (Fig. 1). During the baseline cycles, two complete menstrual cycles were tracked using calendar-based counting and urinary ovulation tests. During the training cycles, menstrual cycles were tracked in accordance with best-practice recommendations [24], which included recording the onset of menstrual bleeding, using urinary ovulation tests, and assessing retrospective serum concentrations of estradiol and progesterone. Three-repetition maximum (3RM) strength and load–velocity profile (LVP) assessments of the bench press and trap bar deadlift were completed at baseline and after approximately 4 (mid-training testing session) and 8 weeks of training (post-training testing session) to support the quantification of kinematic outputs during resistance training sessions. Kinematic outputs were monitored during each training session, with the fastest repetition from each set used to assess changes in training performance across menstrual cycle phases. ‘Expected’ performance (measured in m·s−1 and based upon the LVP of each participant) was compared to the observed velocity of the barbell from each set, while menstrual cycle symptoms, motivation to train, and readiness to perform were measured using a self-reported questionnaire upon arrival to each training session. This study was registered with the trial number ACTRN12626000365369.

Fig. 1: An example of the study design for a hypothetical participant with an idealized 28-day menstrual cycle. This schematic illustrates the sequencing of two baseline cycles followed by two training cycles. During the baseline cycles, participants’ menstrual cycles were monitored using calendar-based counting and urinary ovulation testing. During the two training cycles, participants completed 22 resistance training sessions, with 3RM and LVP assessments conducted at baseline, mid training (~ week 4), and post training (~ week 8). Blood samples were collected during the early follicular, ovulatory, and mid-luteal phases. Menstrual cycles were then retrospectively classified into six phases, and training sessions were assigned to the corresponding phase in which they were performed. 3RM three-repetition maximum, LVP load–velocity profile, s resistance training session. The droplet symbol indicates first day of menstrual bleeding, * indicates ovulation testing, and the test tube symbol indicates blood sample

Participants

Twenty-eight healthy, resistance trained females aged 18–40 years were recruited through advertisements distributed across a university and the broader community (e.g., social media and local gyms), and through word of mouth. To qualify for inclusion, participants were required to (1) be experiencing natural and regular (i.e., length of 21–35 days) menstrual cycles; (2) have engaged in at least two resistance training sessions per week for at least 6 months prior to the study’s commencement; and (3) have not used any type of hormonal contraceptive within the 3 months prior to the study’s commencement. Participants were excluded if they (1) were pregnant or had conditions affecting ovarian function, including known menstrual dysfunction, known endocrine disorders, or chronic diseases, or (2) had a diagnosed injury in the previous 6 months that would affect exercise performance. A total of 60 participants enrolled and were screened for eligibility in this study, with 28 included in the final analysis (refer to Supplementary Material 1; see the electronic supplementary material). Of the 28 participants included in the final analysis, 24 were classified as recreationally active (tier 1) and four as trained (tier 2) [25]. Participants were not compensated for participation in this study.

This study was conducted at Australian Catholic University (Brisbane, Australia) between February 2023 and June 2025. Prior to the study beginning, participants received detailed information of the experimental procedures and provided written informed consent. Participants were asked to abstain from any resistance exercise not included in the prescribed training program during the intervention period. All testing and training sessions were completed in the same laboratory, with the same equipment, researcher present, and testing procedures. Additionally, prior to all testing occasions, participants refrained from caffeine and for the 24 h before testing, participants were asked to consume their habitual diet, and no alcohol was to be consumed. Participants were required to complete > 90% of the total number of workouts to be included in the analysis. Ethics approval was granted by the Australian Catholic University’s Human Research Ethics Committee (Ethics number: 2022-2639H).

Because the present study employed an intensive longitudinal repeated‑measures design, to provide statistical transparency regarding the adequacy of the sample size, a design‑based sensitivity analysis was conducted to estimate the smallest effect size that the study was capable of detecting. Given the final sample size (N = 28), repeated measures, and a moderate‑to‑high within‑participant correlation in velocity expression across sessions (ρ ≈ 0.5–0.7), which reflects the high consistency typically observed in velocity‑based resistance training [26–29], the study was adequately powered to detect within‑participant effects corresponding to standardized mean differences of approximately d = 0.38–0.45. To aid practical interpretation, these detectable standardized effects were converted into absolute velocity units using the residual standard deviation (SD) of mean concentric velocity derived from the mixed‑effects models, which represents typical unexplained within‑participant, session‑to‑session variability. Based on this residual variability, the smallest detectable effect corresponded to an absolute difference of approximately ~ 0.02 m·s⁻1.

Procedures

Menstrual Cycle Monitoring

Two baseline menstrual cycles of monitoring were completed by each participant prior to initiating the resistance training program. During the baseline cycles, participants recorded the onset of menstruation and used mid-stream ovulation tests (Fertility2Family, Australia) to determine the urinary peak of luteinizing hormone. For each participant, ovulation testing commenced on a predetermined day based on their typical cycle length, and they were instructed to complete one test per day until a positive test was identified. All ovulation testing was performed following the manufacturer’s guidelines, and all test results were photographed by the participant and sent to the research investigator for visual confirmation. Participants who were unable to demonstrate a positive ovulation test across two baseline cycles were excluded from further participation. The information obtained from the baseline cycles was used to confirm that participants experienced a regular ovulatory cycle prior to initiating the training intervention. When the regularity of the menstrual cycle was confirmed over two baseline cycles, the resistance training program began on the first day of menstrual bleeding in the third cycle.

During the third and fourth menstrual cycles (i.e., training cycles), participants’ menstrual cycles were tracked using a three-step approach [30, 31]. This included a combination of using calendar-based counting, urinary ovulation testing, and retrospective serum analysis of 17β-estradiol and progesterone concentrations. Using calendar-based counting and ovulation testing methods in combination allowed follicular (begins at the onset of menses) and luteal (post-ovulation) phases to be identified. In addition, venous blood samples were collected at three timepoints throughout each training cycle. Following recommendations from Janse de Jonge et al. [31], the three timepoints used for venous blood sampling were as follows: day 1 to day 4 of the menstrual cycle (early follicular phase), within 24–48 h of a positive ovulation test (ovulatory phase), and 7–9 days after a positive ovulation test (mid-luteal phase). Participants’ menstrual status was retrospectively classified as either eumenorrheic or naturally menstruating. For naturally cycling participants to be classified as eumenorrheic they had to meet the criteria outlined in Supplementary Material 2 (see the electronic supplementary material).

Menstrual Cycle Phase Classification

The menstrual cycle was broken down into six predefined phases. The classification of these phases was chosen based on current published guidelines [24], and selected to coincide with the key fluctuations in ovarian hormones that occur across the menstrual cycle. These phases were: phase 1 (early follicular), phase 2 (mid-follicular), phase 3 (late follicular/ovulatory), phase 4 (early luteal), phase 5 (mid-luteal), and phase 6 (late luteal). Phases 1, 3, and 5 were retrospectively verified using serum 17β-estradiol and progesterone concentrations and correspond to formally recognized phases described by Elliott-Sale et al. [24]. Phase 1 was equivalent to phase 1, and phase 3 was equivalent to phases 2–3 in previous research by Elliott-Sale et al. [24]. Phase 5 was aligned with phase 4 in previous research by Elliott-Sale et al. [24], with a slightly broader window applied in the present study. Additionally, phases 2, 4, and 6 were estimated and used to represent transitional hormonal changes between the verified phases. Including six phases allowed training data to be evaluated across the full menstrual cycle, providing a more continuous and ecologically valid representation of real-world training. Table 1 summarizes each phase, including definitions and calculation criteria. Training sessions were then assigned to the corresponding phase for each participant, allowing performance to be compared across the full cycle.

Phase number | Phase name | Description | Calculation
Phase 1 | Early follicular phase | Low concentrations of estrogen and progesterone | Menstruation onset to day 5
Phase 2 | Mid-follicular phase | Rising estrogen and low progesterone | Days between the early follicular phase and the late follicular/ovulatory phase
Phase 3 | Late follicular/ovulatory phase | High/peaking followed by medium/falling estrogen and low progesterone | Day − 2 before a positive ovulation test to day of and day + 1 following a positive test
Phase 4 | Early luteal phase | Rising estrogen and progesterone | Days between the late follicular/ovulatory phase and the mid-luteal phase
Phase 5 | Mid-luteal phase | High estrogen and progesterone | Days 5 through to 9 following a positive ovulation test
Phase 6 | Late luteal phase | Falling estrogen and progesterone | Days between the mid-luteal phase and the onset of the next menstruation

Familiarization

Approximately 1 week before testing, participants completed a familiarization session. During this session, they were guided through the procedures used in the study’s testing sessions. To ensure proficiency and standardization of exercise technique, participants performed two sets of five repetitions with a self-selected load (~ 50% of 1RM) in the bench press and trap bar deadlift. They were also familiarized with performing the concentric (lifting) phase of each repetition with maximal intended velocity (i.e., as fast as possible), followed by a controlled eccentric (lowering) phase of approximately 2 s.

Assessment of Load–Velocity Profiles for the Quantification of Resistance Training Performance

To assess changes in resistance training performance, maximal dynamic strength (3RM) and LVPs were established in the bench press and trap bar deadlift exercises at baseline and after 4 and 8 weeks of training. LVPs, which were developed according to previously established protocols [32], are an established method of monitoring changes in training performance.

Prior to establishing each participant’s LVP and 3RM, participants performed a standardized warm up consisting of 5 min on a cycle ergometer (Wattbike Pro, Nottingham, England) against a self-selected light resistance. This was followed by dynamic stretches and ten repetitions with an empty barbell (Australian Barbell Company, Mordialloc, VIC, Australia) of the relevant exercise. Following this, participants performed the LVP and 3RM protocol. Using information obtained from each participant’s recent training diaries, three repetitions at 20, 40, and 60%, two repetitions at 80%, and one repetition at 90% of the participant’s estimated 1RM were completed. Next, participants had five attempts to establish their 3RM. For the mid and post testing sessions, values from training and the previous testing occasion were used to inform load progressions. All repetitions completed during each testing session were monitored using a linear position transducer (GymAware Power Tool; Kinetic Performance Technologies, Canberra, Australian Capital Territory, Australia). The GymAware Power Tool has demonstrated excellent levels of validity and reliability across a range of different loads and exercises [33]. It is commonly used as a gold-standard assessment tool for the quantification of mean velocity when resistance training, with mean error compared to three-dimensional motion capture reported to be 0.01–0.03 m·s⁻1 [34]. The LVP was developed using three to five loads, with the lightest being the closest available load to their estimated 40% of 1RM and the heaviest being their fastest repetition from their 3RM attempt (i.e., ~ 93% [35]). The fastest repetition at each load was recorded for the LVP regression equation. Additionally, it should be noted that for the trap bar deadlift, only repetitions that had an external load added were used in the LVP as this standardized the range of motion used. Collectively, this approach ensured that the loads that were trained were being accurately represented (i.e., the full spectrum of loads and velocities that were used during training were captured during testing). Participants received velocity feedback after each repetition [36, 37], and were instructed to use a controlled eccentric tempo of ~ 2 s, before performing the concentric phase of each repetition with maximal intent. The 3RM determined for each exercise was then used to estimate the 1RM using guidelines from the National Strength and Conditioning Association [35].

All testing sessions were performed with a minimum of 48 h of rest and were conducted at the university’s research laboratory under the direct supervision of the same investigator, at the same time of day for each participant (± 2 h), and under controlled environmental conditions. Furthermore, testing sessions were scheduled to occur at the same phase of the menstrual cycle for all participants (late luteal phase), determined as 12–14 days following a positive ovulation test. This allowed an average of 2 days (± 2) prior to the onset of menstrual bleeding for the subsequent cycle, with the initiation of each training block commencing with the onset of menstrual flow for each training cycle.

Resistance Training Protocol and Recording of Kinematic Outputs

Participants completed two training cycles that involved approximately 8 weeks of one-on-one supervised resistance training. The training program followed a block periodization model, which consisted of two 4-week mesocycles of increasing relative intensity, with an unloading week during each week of testing [38]. To align the training program with each participant’s individual menstrual cycles, the program was scheduled so that the first training session of each block was initiated by the onset of menstrual bleeding, and testing occurred during the final week of the menstrual cycle. Furthermore, sessions were scheduled according to follicular and luteal phases, with the luteal phase starting the day after ovulation. By using this scheduling approach, the number of training sessions completed in each phase could be standardized, regardless of differences in menstrual cycle length or phase duration between participants and between cycles for the same participant.

All participants were programmed a total of 22 resistance training sessions across the intervention period. At least 1 day of rest was scheduled between sessions. However, due to interindividual variations in cycle and phase length, it was not always feasible to strictly adhere to this rest period between sessions. All participants completed the same training program, whereby each workout consisted of the same exercise selection, sequence, repetitions, sets, intensities, and total recovery time. The characteristics of the 8-week resistance training program can be found in Supplementary Material 3 (see the electronic supplementary material).

Each workout consisted of six exercises, starting with primary exercises (i.e., exercises that underwent 3RM strength testing), followed by secondary compound exercises, and then auxiliary exercises. Training intensities for primary exercises were prescribed using loads that correspond to a percentage of estimated 1RM. Additionally, for all primary exercises, all repetitions completed across the program were monitored using a linear position transducer, and participants received verbal encouragement and velocity feedback after each repetition [36]. For secondary compound exercises, training was prescribed using the repetitions in reserve method, whereas auxiliary exercises were prescribed using repetition maximum zones. For all exercises, participants were instructed to control the eccentric phase and to execute the concentric portion using maximal intent.

Assessment of Training Performance

Changes in training performance were assessed by comparing observed barbell velocities achieved during each training session with expected velocities. Observed velocity was determined as the fastest repetition achieved for each set in the primary exercises performed. Expected velocities for each exercise were derived from participants' LVPs. However, an important consideration was to control for changes in strength that may be a natural result of the training program. Therefore, to improve the accuracy of the velocity estimations at each load throughout training, an expected velocity value was calculated daily. The expected velocity value considered changes in the LVP at baseline versus mid testing, and at mid testing versus post testing. This was achieved by first calculating the estimated velocity for the %1RM used during each training session across the block determined from the baseline LVP. The same process was then repeated using the mid-testing LVP. The difference between these two estimated velocities for each load represented the total change in velocity across the block. Assuming this change occurred linearly across training sessions, this total change was divided by 11 (i.e., the number of training sessions between the baseline and mid-testing profiles) to determine the average change in velocity per session. The expected velocity for each session was then calculated using the following formula:where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\nu}{s}$$\end{document}νs is the expected velocity for session s_s, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\nu}{\mathrm{b}\mathrm{a}\mathrm{s}\mathrm{e}}$$\end{document}νbase is the estimated velocity at the prescribed %1RM from the baseline LVP, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\nu}{\mathrm{m}\mathrm{i}\mathrm{d}}$$\end{document}νmid is the corresponding value from the mid-testing LVP. An example of the calculation of expected velocities and their comparison with observed velocities for a representative participant is provided in Supplementary Material 4 (see the electronic supplementary material).

\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\nu}_{s}={\nu}_{\mathrm{b}\mathrm{a}\mathrm{s}\mathrm{e}}+\left(\frac{{\nu}_{\mathrm{b}\mathrm{a}\mathrm{s}\mathrm{e}}-{\nu}_{\mathrm{m}\mathrm{i}\mathrm{d}}}{11}\right)\times \left(s-1\right),$$\end{document}νs=νbase+νbase-νmid11×s-1,

Menstrual Cycle Symptoms

Upon arrival at each training session, menstrual cycle symptoms were assessed using a self-report questionnaire. Specifically, the Menstrual Distress Questionnaire (MDQ) was used to capture the presence and severity of menstrual cycle symptoms [39]. The MDQ consists of 46 items across eight subscales, with responses rated on a 6-point scale from 0 (no experience of the symptom) to 5 (acute or partially disabling experience). These subscales measure specific domains of menstrual distress including the following: pain, concentration, behavioral change, autonomic reactions, water retention, negative affect, arousal, and control [39].

Motivation to Train and Readiness to Perform

Motivation was measured on a 100-mm visual analog scale (VAS) anchored by the words ‘none at all’ (0) to ‘maximal’ (100) [40, 41]. Participants were required to answer the question ‘How would you rate your motivation to train in this current moment?’ Motivation was defined as the feeling of determination to complete the training session at that moment [42]. Readiness to perform was also measured on a 100-mm VAS anchored from ‘none at all’ (0) to ‘maximal’ (100). Participants were required to answer the question ‘How ready do you feel in this current moment to train?’ Participants were familiarized with the scales and were provided with specific examples to contextualize motivation to train and readiness to perform.

Blood Sampling and Analysis

Venous blood samples were collected from an antecubital vein following an 8-h fast. Blood samples were collected directly into two 5-mL serum separator tubes. Following inversion, the whole blood was allowed to clot for 30 min and then centrifuged (Megafuge 8R, ThermoFisher Scientific, ANZ) for 10 min at 1100_g_. The remaining serum was split into aliquots and stored at − 80 °C until batch analysis. To minimize the risk of sample degradation, each aliquot per timepoint was thawed only once at the time of analysis. 17β-estradiol and progesterone were measured using enzyme-linked immunosorbent assay (ELISA) kits (17β-estradiol ELISA Kit [ab108667; sensitivity: 8.68 pg/mL] and human progesterone ELISA kit [ab108670; sensitivity: 0.05 ng/mL]). Standardized protocols were followed for the analysis of each hormone in accordance with manufacturer recommendations. Intra-assay coefficients of variation (CVs) were calculated for the 17β-estradiol and progesterone assays using the formula %CV = (SD/mean) × 100, based on the optical density (background-corrected) values for standards, duplicates, and control samples provided in the ELISA kits. The calculated average intra-assay %CV for estrogen (5.95%) and progesterone (3.73%) were both within the typical ranges reported by the manufacturer (Abcam values: < 9% for estrogen and < 4% for progesterone).

Statistical Analysis

To examine whether resistance training performance differed from expected values across menstrual cycle phases, separate linear mixed-effects models were constructed for the bench press and deadlift exercises. The dependent variable was the difference between observed and expected mean concentric velocity (‘Difference’), where positive values indicated performance better than expected and negative values indicated performance worse than expected.

For each exercise, menstrual cycle phase (six levels) was included as a fixed effect, and participant identity was included as a random intercept to account for repeated observations within individuals:where:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathrm{D}\mathrm{i}\mathrm{f}\mathrm{f}\mathrm{e}\mathrm{r}\mathrm{e}\mathrm{n}\mathrm{c}\mathrm{e}}{ij}$$\end{document}Differenceij is the observed minus expected mean concentric velocity for participant i at session j,\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta}{0}$$\end{document}β0 is the fixed intercept (reference menstrual cycle phase),\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta}{k}$$\end{document}βk are the fixed effects for menstrual cycle phase (with one phase treated as the reference category),\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathrm{M}\mathrm{C}.\mathrm{P}\mathrm{h}\mathrm{a}\mathrm{s}\mathrm{e}}{k,ij}$$\end{document}MC.Phasek,ij are indicator variables for the remaining menstrual cycle phases,\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${u}{0i} \sim N(0,{\sigma}{u}^{2})$$\end{document}u0i∼N(0,σu2) is the participant-specific random intercept accounting for repeated measurements,\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\varepsilon}_{ij} \sim N(0, {\sigma }^{2})$$\end{document}εij∼N(0,σ2) is the residual error term.

\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathrm{D}\mathrm{i}\mathrm{f}\mathrm{f}\mathrm{e}\mathrm{r}\mathrm{e}\mathrm{n}\mathrm{c}\mathrm{e}}_{ij}={\beta}_{0}+\sum_{k=1}^{5}{\beta}_{k}{\mathrm{M}\mathrm{C}.\mathrm{P}\mathrm{h}\mathrm{a}\mathrm{s}\mathrm{e}}_{k,ij}+{u}_{0i}+{\varepsilon}_{ij},$$\end{document}Differenceij=β0+∑k=15βkMC.Phasek,ij+u0i+εij,

Separate models were fitted for the bench press and deadlift to allow exercise-specific inference. Model assumptions were assessed via visual inspection of residual plots. Estimated marginal means (EMMs) for each menstrual cycle phase were obtained from the fitted models. Statistical significance was evaluated by testing whether the EMM for each phase differed from zero (i.e., expected performance), with 95% confidence intervals (CIs) reported.

Standardized effect sizes were calculated for the phase-specific deviations from expected performance using the model-based residual SD. Effect sizes were expressed as standardized mean differences (Cohen’s d) using the identity contrast, with 95% CIs derived from the fitted mixed-effects models. This approach allowed effect magnitudes to be interpreted relative to within-model variability rather than raw velocity units.

To investigate psychological predictors of training performance, additional linear mixed-effects models were fitted using an alternative dependent variable (‘Difference Pooled’), representing observed minus expected performance pooled across exercises. First, separate univariate models were fitted to examine the independent associations of perceived readiness to perform and motivation to train with performance. Subsequently, multivariate models were constructed to assess whether these associations remained after accounting for menstrual cycle phase and exercise type. The base multivariate model included menstrual cycle phase, exercise, readiness, and motivation as fixed effects, with participant included as a random intercept. To evaluate whether the effect of menstrual cycle phase differed between exercises, an interaction between menstrual cycle phase and exercise was added in a follow-up model.

To examine the association between menstrual cycle-related symptoms and resistance training performance, a series of linear mixed-effects models were fitted using the observed minus expected velocity difference as the dependent variable. Prior to modeling, correlations among symptom subscales were examined to assess shared variance. Several symptom measures were moderately to strongly correlated, indicating that inclusion of all subscales within a single model would risk unstable estimates. Accordingly, each symptom subscale was examined in a separate model.

Prior to analysis, raw subscale scores were transformed into standard T-scores (mean = 50, SD = 10) to facilitate comparison across different symptom clusters. For each symptom subscale, a unified mixed-effects model was specified that included the standardized symptom score, exercise type, and their interaction as fixed effects. Perceived readiness to perform, motivation to train, and menstrual cycle phase were included as covariates to account for their potential influence on performance. Participant identity was included as a random intercept to account for repeated observations within individuals. The interaction between symptom severity and exercise was included to determine whether the association between symptoms and performance differed between the bench press and deadlift. Fixed-effect estimates, 95% CIs, and p values were reported for all models.

Where a symptom-by-exercise interaction was observed, post hoc analyses were conducted to estimate the slope of the symptom–performance relationship separately for each exercise. Differences between these exercise-specific slopes were assessed using pairwise comparisons with Holm adjustment to control the family-wise error rate. In addition, EMMs were calculated to examine the main effect of exercise on performance, averaged across symptom severity and covariates, with pairwise comparisons similarly adjusted.

Model assumptions were evaluated through visual inspection of residuals and fitted values. Random-slope models were explored for symptom effects. However, where these models failed to converge or did not improve model fit, random-intercept-only models were retained for inference. This modeling approach allowed the independent associations of menstrual cycle symptoms with resistance training performance to be evaluated while accounting for exercise-specific effects, psychological factors, menstrual cycle phase, and repeated measurements within individuals.

Statistical significance was set at p < 0.05. All statistical analyses were conducted using R (R Foundation for Statistical Computing, Vienna, Austria, version 4.4.2).