Relationship between hypothalamic–pituitary–adrenal axis activity and menstrual irregularities in women with polycystic ovary syndrome

Article information

Korean J Fertil Steril. 2026;.cerm.2025.08347

Publication date (electronic) : 2026 January 22

doi : https://doi.org/10.5653/cerm.2025.08347

Ali Zeynettin^,1, Orhan Balikci¹, İsmail Demir²

¹Department of Internal Medicine, İzmir City Hospital, İzmir, Türkiye

²Department of Internal Medicine, İzmir City Hospital, University of Health Sciences İzmir Faculty of Medicine, İzmir, Türkiye

Corresponding author: Ali Zeynettin Department of Internal Medicine, İzmir City Hospital, İzmir, Türkiye Tel: +90-5079215185 E-mail: alizeynettin@gmail.com

Received 2025 June 29; Revised 2025 August 29; Accepted 2025 September 4.

Abstract

Introduction

Polycystic ovary syndrome (PCOS) is the most common endocrine disorder among women of reproductive age. PCOS is characterized by anovulation, hyperandrogenism, and polycystic ovarian morphology. Its prevalence ranges from 6% to 10%, depending on the diagnostic criteria used, and it represents a significant reproductive health concern [1,2].

PCOS has a multifactorial etiopathogenesis that includes genetic predisposition, insulin resistance, hyperandrogenism, chronic inflammation, and hypothalamic–pituitary dysfunction [3,4]. One of its most prominent clinical manifestations is menstrual irregularity. Oligomenorrhea, amenorrhea, and anovulatory cycles are frequently observed, adversely impacting both fertility and quality of life [1].

These irregularities are not due solely to hormonal disturbances; psychosocial factors such as stress, depression, and anxiety are also implicated [2]. Increased perceived stress in individuals with PCOS may influence both metabolic and endocrine functions. The hypothalamic–pituitary–adrenal (HPA) axis plays a central role in the physiological response to stress through cortisol secretion; under chronic stress, this regulation may be impaired—potentially elevating cortisol levels, suppressing the release of gonadotropin-releasing hormone, disrupting the luteinizing hormone (LH)/follicle-stimulating hormone (FSH) balance, and contributing to ovulatory dysfunction [3,4]. Beyond simple between-group differences, however, it remains unclear whether perceived stress is independently associated with menstrual irregularity once key confounders (including age, adiposity, medication use, and metabolic/inflammatory status) are accounted for.

Cortisol also influences glucose metabolism, insulin resistance, and inflammatory responses [5]. Therefore, the impact of stress on PCOS pathophysiology should be considered not only psychologically but also biochemically, raising the possibility that stress-related menstrual disturbance in PCOS may be mediated primarily through metabolic and inflammatory pathways rather than representing a direct effect of perceived stress.

Recent studies have reported correlations among 10-item Perceived Stress Scale (PSS-10) scores, homeostatic model assessment of insulin resistance (HOMA-IR), and the levels of serum cortisol, adrenocorticotropic hormone (ACTH), and high-sensitivity C-reactive protein (hs-CRP) [6,7]. Nevertheless, many prior reports are limited by small sample sizes and incomplete adjustment for confounders; few have simultaneously evaluated subjective (PSS score-based) and biomarker-based stress measures using multivariable modeling with appropriate control for multiple testing.

Accordingly, our primary objective was to test whether PSS score is independently associated with menstrual irregularity in PCOS after adjustment for clinical and biochemical covariates. Secondary objectives were to (1) quantify the relationships among PSS score, HPA axis markers (cortisol, ACTH), and metabolic/inflammatory markers (HOMA-IR, hs-CRP), and (2) compare the discriminatory performance of candidate predictors using receiver operating characteristic (ROC) analysis. We pre-specified a comprehensive multivariable approach to mitigate confounding and multiple-testing artifacts. Based on this rationale, we hypothesized that although PSS score would correlate with menstrual irregularity, it would not remain an independent predictor after adjustment, whereas metabolic/inflammatory markers would.

Methods

This retrospective cross-sectional study was conducted between 2023 and 2024 at the Internal Medicine Outpatient Clinic of SBÜ İzmir Bozyaka Training and Research Hospital. A total of 296 women of reproductive age with PCOS, diagnosed according to the Rotterdam criteria, were included. Medication use at the time of evaluation (metformin, antiandrogens) was recorded a priori for adjustment in multivariable analyses. Women receiving systemic glucocorticoids were excluded. Additional exclusion criteria were pregnancy, lactation, current use of hormonal contraception, Cushing syndrome, thyroid dysfunction, pituitary adenoma, adrenal tumor, diabetes mellitus, chronic inflammatory disorders, systemic corticosteroid therapy, and any other endocrinological or metabolic disease.

Data were retrieved from the hospital’s electronic medical record system. Demographic characteristics and anthropometrics were recorded. Biochemical and hormonal profiles included fasting glucose, insulin (for HOMA-IR), lipid parameters, hs-CRP, cortisol, ACTH, LH, FSH, and total testosterone. Cortisol and ACTH levels were measured from fasting samples obtained on cycle day 2–3 whenever feasible. All assays were performed in the institutional laboratory using standardized protocols. Psychological stress was assessed using the validated Turkish version of the PSS-10 (Eskin et al. [8]). Scores were retrieved from the digital record and used as continuous variables.

Menstrual status was classified as follows: regular menstruation (cycle interval of 21 to 35 days), oligomenorrhea (cycles >35 days), or amenorrhea (absence of menstruation for ≥3 consecutive months). The primary outcome was menstrual irregularity, analyzed as a binary endpoint (amenorrhea vs. others) for the main model; a secondary, ordinal analysis across the three categories was pre-specified. Analyses were conducted using IBM SPSS Statistics ver. 28.0 (IBM Corp.). Normality was assessed with the Shapiro–Wilk test. Group comparisons employed analysis of variance or the Kruskal–Wallis test for continuous variables and the chi-square or Fisher exact test for categorical variables; appropriate post hoc tests were applied. Pairwise associations were summarized with the Spearman rho and visualized as a correlation heatmap; cells with false discovery rate (FDR)–adjusted p-values less than 0.05 were annotated.

For the primary question, multivariable logistic regression (amenorrhea vs. other categories) was fitted a priori, adjusting for age, body mass index (BMI) or waist circumference, medication use (metformin), LH/FSH, total testosterone, HOMA-IR, and hs-CRP. As a sensitivity analysis, a proportional-odds (ordinal) model across the three menstrual categories was also fitted with the same covariates. Multicollinearity was assessed using variance inflation factors (VIF), and model parsimony followed events-per-variable considerations. Discrimination was quantified by ROC area under the curve (AUC) with 95% confidence intervals (CIs); optimal cut-offs were derived using the Youden index. Exact two-sided p-values are reported; multiplicity was controlled with the Benjamini–Hochberg FDR.

Ethical approval was obtained from the SBÜ İzmir Bozyaka Training and Research Hospital Ethics Committee (File No. 29.04.2025/28; Decision No. 2025/34; Date: 30.04.2025), and the study was conducted in accordance with the Declaration of Helsinki. Given the retrospective design and use of de-identified data, the requirement for informed consent was waived.

Results

A total of 296 women with PCOS were included. Participants were divided into three groups according to menstrual status: regular menstruation (n=95), oligomenorrhea (n=99), and amenorrhea (n=102). The mean participant age was 31.7±8.6 years, with no significant difference between groups (p=0.921). Although BMI and waist circumference were higher in the amenorrhea group, these differences were not statistically significant (BMI, p=0.425; waist circumference, p=0.372) (Table 1).

Table 1.

Comparison of clinical and biochemical parameters according to menstrual status groups in women with PCOS

Among endocrine parameters, LH, FSH, and the LH/FSH ratio were analyzed across the three groups. The LH/FSH ratio was numerically higher in the amenorrhea group compared with regular menstruation; however, the overall between-group difference did not reach statistical significance (p=0.672). In addition, correlations between the LH/FSH ratio and HPA axis hormones were not significant (LH/FSH vs. cortisol: r=0.01, p=0.931; LH/FSH vs. ACTH: r=0.05, p=0.369), indicating no clear coupling at the gonadotropin level in this dataset (Table 1, Figure 1).

Figure 1.

Spearman correlation heatmap among 10-item Perceived Stress Scale (PSS-10) score, cortisol, adrenocorticotropic hormone (ACTH), homeostatic model assessment of insulin resistance (HOMA-IR), high-sensitivity C-reactive protein (hs-CRP), total testosterone, and luteinizing hormone (LH)/follicle-stimulating hormone (FSH) ratio. Cells with false discovery rate (FDR)-adjusted p<0.05 are marked with an asterisk.

Total testosterone and dehydroepiandrosterone sulfate (DHEAS) levels tended to increase with the severity of menstrual irregularity, with the highest values observed in the amenorrhea group; nevertheless, between-group differences were not statistically significant (total testosterone, p=0.893; DHEAS, p=0.223) (Supplementary Figure 1). Accordingly, markers of hyperandrogenism showed a graded but statistically non-significant pattern across menstrual categories in this cohort (Table 1).

Cortisol levels did not differ significantly between groups (p=0.912); nevertheless, a strong positive correlation was observed between PSS score and cortisol (r=0.807, p<0.001), consistent with HPA axis activation in relation to stress perception. Similarly, ACTH levels displayed no significant group difference (p=0.433) but correlated positively with PSS (r=0.722, p<0.001), supporting pituitary-level involvement associated with perceived stress (Figure 1).

PSS-10 scores also did not differ significantly across menstrual status groups (Supplementary Figure 2)

HOMA-IR increased in parallel with menstrual irregularity (with mean values rising from the regular to the amenorrhea group); however, the overall difference across groups was not statistically significant (p=0.308). hs-CRP levels showed a similar non-significant trend (e.g., 3.07 mg/L in regular menstruation vs. 3.20 mg/L in amenorrhea; p=0.388), compatible with low-grade inflammation without clear categorical separation. PSS scores increased with the severity of menstrual irregularity but did not differ significantly between groups (p=0.543); the amenorrhea group had a mean PSS score of 20.12, compared with 19.39 for the participants with regular menstruation. In correlation analyses, PSS demonstrated significant positive associations with HOMA-IR (r=0.619, p<0.001) and hs-CRP (r=0.609, p<0.001), whereas the association with triglycerides was not significant (r=0.00, p=0.938) (Table 1, Figure 1).

Markers of renal/metabolic burden displayed modest differences: the urea/creatinine ratio was borderline higher across groups (overall p=0.050), whereas uric acid level showed no evidence of a between-group difference (p=0.971). These findings suggest that any renal/metabolic load related to menstrual status in this sample was subtle (Table 1).

Multivariable logistic regression analysis identified HOMA-IR as an independent predictor of amenorrhea after adjustment for age, BMI, waist circumference, total testosterone, and PSS (odds ratio [OR], 1.86; 95% CI, 1.08 to 3.19; p=0.025). In contrast, total testosterone (OR, 1.00; 95% CI, 0.98 to 1.01; p=0.629) and waist circumference (OR, 1.03; 95% CI, 0.98 to 1.08; p=0.305) did not remain independent predictors after adjustment; age and BMI were also non-significant. Notably, PSS was not independently associated with amenorrhea (OR, 0.97; 95% CI, 0.91 to 1.02; p=0.240) once metabolic and anthropometric covariates were considered (Table 2, Figure 2).

Table 2.

Multivariable logistic regression predicting amenorrhea and single-marker receiver operating characteristic

Figure 2.

Adjusted odds ratios (ORs) for amenorrhea from the multivariable model (including age, body mass index [BMI], waist circumference, total testosterone, homeostatic model assessment of insulin resistance [HOMA-IR], and 10-item Perceived Stress Scale [PSS-10] score). Points show adjusted ORs, horizontal bars indicate 95% confidence intervals, and the dashed line denotes OR=1.

According to ROC analysis, HOMA-IR as a single marker showed modest discrimination for amenorrhea versus others (AUC, 0.554). The Youden-optimal cut-off was 4.02, above which the likelihood of amenorrhea increased (sensitivity, 0.14; specificity, 0.98). By comparison, ROC analysis of the PSS score revealed poor discriminatory power (AUC, 0.57), indicating limited predictive value when used alone; PSS may contribute more meaningfully in multivariable or biomarker combination models rather than as a single threshold (Figure 3).

Figure 3.

Receiver operating characteristic curve for homeostatic model assessment of insulin resistance (HOMA-IR) (amenorrhea vs. others). Area under the curve (AUC), 0.554; Youden-optimal threshold, 4.02 (sensitivity, 0.14; specificity, 0.98). The diagonal line represents random classification (AUC, 0.50).

Figure 2 provides the forest plot of adjusted ORs from the multivariable model, and Table 1 summarizes the clinical and biochemical parameters by menstrual group. Collectively, these results indicate that while PSS score correlates strongly with HPA axis markers and metabolic/inflammatory markers, its standalone predictive value for menstrual irregularity is limited, and in adjusted models HOMA-IR emerges as the most consistent independent predictor of amenorrhea in this cohort.

Discussion

This study examined the associations between menstrual irregularity and HPA axis activity, metabolic parameters, and PSS scores in women with PCOS. Menstrual irregularity was examined from multiple perspectives. Across menstrual categories, total testosterone, HOMA-IR, and hs-CRP displayed graded increases consistent with the pathophysiology of PCOS; however, in this cohort, overall between-group differences did not consistently reach statistical significance, indicating a pattern compatible with metabolic/inflammatory activation but lacking clear categorical separation [9]. Especially in the amenorrhea group, the overall biomarker profile suggests that menstrual dysregulation is not solely ovarian in origin but is also closely related to the endocrine, inflammatory, and metabolic systems [10].

One of the most striking findings was the robust alignment of perceived stress with biological measures: the PSS-10 score was strongly and positively correlated with HOMA-IR and hs-CRP, cortisol, and ACTH levels [11]. This constellation supports the view that perceived stress, beyond a subjective construct, reflects a physiologically active stress response [12]. Cohen et al. [13] showed that stress activates the HPA axis and affects metabolic balance, and Chandran et al. [14] emphasized that chronic stress can increase insulin resistance. Our results extend these observations by simultaneously evaluating subjective (PSS score-based) and biomarker-based stress measures and by testing independence in multivariable models. The crude association between PSS and menstrual irregularity attenuated and lost statistical significance after adjustment for key covariates, suggesting that neuroendocrine and metabolic pathways are the proximal correlates of cycle disturbance rather than perceived stress per se [15].

With respect to classification performance, single-marker discrimination was modest. Although prior research has reported relatively high AUCs for metabolic indicators in selected phenotypes [16-18], our ROC analysis of HOMA-IR for amenorrhea versus others yielded limited discrimination (AUC, 0.55). Several factors may account for this difference, including phenotype definition (such as amenorrhea vs. regular menses or three-category coding), cohort characteristics, timing of sampling within the cycle, and differences in pre-analytical/laboratory procedures. Nonetheless, the directional signal remains biologically consistent: insulin resistance relates to more severe menstrual disturbance, but reliance on HOMA-IR alone is insufficient for robust risk stratification, underscoring the need for multi-marker approaches.

Collectively, our findings support an integrated pathway—perceived stress aligns with HPA axis activation (cortisol/ACTH) and with insulin resistance/inflammation (HOMA-IR, hs-CRP), whereas menstrual irregularity is more closely linked to the metabolic/inflammatory arm [19]. In multivariable regression, HOMA-IR retained an independent association with amenorrhea after adjustment for age, adiposity, medication use, gonadotropins, and androgens, whereas PSS did not remain independently associated. These results are compatible with a mediation framework in which stress-related biology influences menstrual function primarily through metabolic and inflammatory pathways [3-20]. The observed rise in hs-CRP across categories, even without a statistically significant group difference, is consistent with the low-grade inflammation frequently described in PCOS [21].

The multivariable results also clarify the role of other covariates. Total testosterone and waist circumference did not retain statistical significance after adjustment, and age and BMI were likewise non-significant. This pattern suggests that while hyperandrogenism and central adiposity are important features of PCOS, their apparent crude relationships with amenorrhea may be explained by shared variance with insulin resistance in this cohort. Clinically, these data argue against using a single anthropometric or androgen measure in isolation for menstrual risk stratification, instead favoring parsimonious models that emphasize metabolic/inflammatory readouts in conjunction with clinical features.

PSS, despite low standalone discriminatory power (AUC, 0.57) and the lack of an adjusted association with menstrual category, remained correlated with HOMA-IR and hs-CRP in continuous analyses. This pattern underscores a practical message: subjective stress tracks with objective metabolic/inflammatory strain even if it does not function as an independent categorical predictor of menstrual status. The positive correlations between PSS and cortisol/ACTH further support HPA axis engagement in stress-related physiology [12]. Under chronic stress, hyperactivation of the HPA axis may inhibit ovulatory cycles via corticotropin-releasing hormone-/ACTH-mediated mechanisms, consonant with the framework proposed by Kyrou and Tsigos [5] and with subsequent observations linking chronic stress to ovulatory dysfunction [20].

Regarding the gonadotropin axis, our groupwise comparisons did not show a statistically significant difference in the LH/FSH ratio across menstrual categories, and correlations of LH/FSH with cortisol/ACTH were not significant in this dataset (Supplementary Figure 3). While previous studies have posited potential links between LH/FSH imbalance and stress physiology [22,23], our adjusted analyses emphasize that downstream metabolic pathways better account for the menstrual phenotype observed here. Nevertheless, given the well-known heterogeneity of PCOS, it remains plausible that subgroups defined by androgenic or gonadotropin profiles could exhibit different stress–cycle relationships, a hypothesis that warrants targeted exploration.

Metabolic profiling (including glucose, low-density lipoprotein, high-density lipoprotein, and triglyceride levels) remains essential for understanding the interface between stress, inflammation, and cardiometabolic risk [24,25]. In our data, triglycerides did not correlate significantly with PSS, while waist circumference did not independently predict amenorrhea once HOMA-IR and other covariates were included—aligning with the view that visceral adiposity acts primarily through insulin resistance rather than as a direct predictor. Although age did not differ significantly between groups, we retained it as a confounder based on known age-related hormonal changes [26]. As for renal/metabolic load, we noted a borderline elevation of the urea/creatinine ratio across categories without differences in uric acid, a pattern that may reflect subtle hemodynamic/metabolic burden associated with inflammation and insulin resistance [27].

From a methodological perspective, the present revision addresses the reviewer’s recommendations by (1) pre-specifying a covariate-adjusted multivariable strategy (including age, adiposity, medications, LH/FSH, testosterone, HOMA-IR, and hs-CRP); (2) assessing multicollinearity (via VIFs) to avoid overparameterization; (3) controlling multiplicity with FDR; and (4) replacing the bar plot with a correlation heatmap that displays pairwise structure and FDR-significant cells. We also report exact two-sided p-values where applicable, aligning with standard reporting practices.

This study has certain limitations. The retrospective, cross-sectional design precludes causal inference and is vulnerable to residual confounding; subjective stress measures may be affected by reporting variability and the timing of assessment; objective long-term stress biomarkers (e.g., hair cortisol) were not available; and discrimination estimates depend on phenotype definitions and may differ in external cohorts. Future research should validate parsimonious multi-marker models across centers, incorporate objective stress biomarkers, and test whether improvements in insulin resistance/inflammation mediate normalization of menstrual patterns.

In conclusion, perceived stress in PCOS aligns strongly with HPA axis activation and with metabolic/inflammatory status, but it does not independently predict menstrual irregularity after adjustment. Insulin resistance emerges as the most consistent independent correlate of amenorrhea, arguing for metabolically oriented assessment and management, while multicomponent models are likely to be required for clinically useful risk stratification [28,29].

This study indicates that menstrual dysregulation in women with PCOS relates not only to ovarian–hormonal factors but also to stress-linked metabolic and inflammatory biology involving the HPA axis. Across menstrual categories, several biomarkers (e.g., HOMA-IR, hs-CRP, androgens) displayed less favorable profiles with more severe irregularity; however, between-group differences were not consistently statistically significant in this cohort. Perceived stress (PSS-10 score) correlated strongly with cortisol, ACTH, HOMA-IR, and hs-CRP, supporting a physiological stress response. However, PSS-10 score was not independently associated with menstrual irregularity after multivariable adjustment, and single-marker discrimination was modest (e.g., HOMA-IR AUC, 0.554). These findings suggest that the influence of perceived stress on cycle regulation likely operates indirectly through metabolic and inflammatory pathways rather than as a standalone predictor.

Clinically, the results support a multidisciplinary assessment that prioritizes metabolic/inflammatory evaluation and management alongside hormonal and psychosocial care. Targeting insulin resistance and low-grade inflammation may offer the greatest leverage for improving cycle regularity, while multicomponent risk models—as opposed to any single measure—are more likely to achieve clinically useful stratification.

Supplementary material

Article information Continued