Mild hyperprolactinemia in women with polycystic ovary syndrome. Insights from a large cross-sectional study

Highlights

• •Hyperprolactinemia is an uncommon feature in polycystic ovary syndrome, despite this condition is argued to induce mild prolactin elevations.
• •Venipuncture stress and macroprolactin are the primary causes of hyperprolactinemia in polycystic ovary syndrome.
• •Polycystic ovary syndrome can still be diagnosed even with mild elevation in circulating prolactin when suggestive symptoms are present.

Abstract

Background

Hyperprolactinemia is an exclusion criterion for polycystic ovary syndrome (PCOS), albeit PCOS itself is argued to induce mild hyperprolactinemia. We aimed to study the prevalence and causes of hyperprolactinemia in patients with PCOS.

Methods

We conducted a cross-sectional study including 336 premenopausal patients with PCOS and 90 nonhyperandrogenic controls referred to our clinics (referral population). We also studied an unselected population of premenopausal individuals who attended our center for voluntary blood donation (14 patients with PCOS and 207 non-hyperandrogenic controls). The main outcome measure was the percentage of individuals with hyperprolactinemia.

Results

As a whole, 39 out of 647 participants showed increased basal prolactin concentrations (6.0%, 95%CI: 4.4; 8.1) regardless of having PCOS or being a control, in both referral and unselected populations. In the referral population, 18 out of 31 individuals with hyperprolactinemia (58.0%, 95%CI: 40.8; 73.6) showed normal prolactin concentrations after appropriate resting, suggesting venipuncture stress-related hyperprolactinemia, and another nine participants (29.0%, 95%IC: 16.1; 46.6) did so after pre-analytical polyethylene-glycol precipitation of serum, indicating macroprolactinemia. There were differences in these figures between patients with PCOS and controls. In the unselected population, three out of eight participants with hyperprolactinemia (37.5%, 95%IC: 13.7; 69.4) had macroprolactinemia, and stress-related hyperprolactinemia accounted for another 62.5% (95%IC: 30.6; 86.3) of cases.

Conclusions

Hyperprolactinemia is equally likely among patients with PCOS and non-hyperandrogenic individuals. The most common causes of mild hyperprolactinemia in this population are venipuncture stress and macroprolactinemia that must not preclude a diagnosis of PCOS if suggested by signs and symptoms.

Introduction

The diagnosis of polycystic ovary syndrome (PCOS) requires exclusion of hyperprolactinemia [1,2]. The rational underlying this requirement is that hyperprolactinemia may induce hypogonatropic hypogonadism in premenstrual women, mimicking the menstrual disturbances of PCOS. Such an effect is mediated by several mechanisms, including inhibition of gonadotropin-releasing hormone via down regulation of the kisspeptin-KISS1 gene signaling system, positive estrogen feedback on luteinizing hormone secretion, and blocking the effects of gonadotropins at the gonadal level [3].

On the contrary, women with PCOS – particularly when weight excess is absent – often show higher circulating luteinizing-hormone concentrations and increased pulse frequency when compared with non-hyperandrogenic women [4], even though a low dopaminergic tone has been argued as an explanation for inappropriate prolactin (PRL) secretion in some women with PCOS [5]. Despite this opposed pathophysiological mechanisms, many physicians still consider PCOS itself to be a cause of mild hyperprolactinemia [6,7] as supported by some clinical guidelines [8] and even clinical trials [9]. Hence, the pathophysiological link between two conditions sometimes presenting with similar symptoms, i.e.: ovulatory dysfunction, is unclear beyond the fortuitous coincidence of two prevalent entities such as PCOS and hyperprolactinemia [10].

Hyperprolactinemia is also considered a rare cause of hirsutism [11,12]. PRL through its own receptor [13] might stimulate adrenal androgen secretion interfering with progesterone synthesis, thereby favoring the delta 5 pathway [14], as well as down-regulating aromatase activity in ovarian granulosa cells [15]. However, PRL also inhibits 5α-reductase activity in vitro [16].

A common cause of a hyperprolactinemia misdiagnosis in women with hyperandrogenic symptoms is the presence of macroprolactinemia [17,18]. In the same line, another likely etiology for mild hyperprolactinemia among premenopausal women is venipuncture stress [19,20]. Their presence may lead to the exclusion of PCOS with all the negative consequences that such a misdiagnosis could carry to the health of these patients. We hereby aimed to address the prevalence of hyperprolactinemia, macroprolactinemia, and venipuncture stress-hyperprolactinemia in women with PCOS compared with non-hyperandrogenic women presenting with regular menses, both at the clinical setting and in unselected women from the general population.

Materials and methods

We conducted an observational cross-sectional study using archived samples and clinical data derived from two distinct populations: i) a group of consecutive premenopausal women reporting to our clinic because of signs and symptoms of androgen excess and appropriate controls (referral population) [21]; and ii) a group of female volunteers reporting spontaneously for blood donation, composed of premenopausal women in whom PCOS was screened in order to assess the prevalence of the syndrome (unselected population) [22].

Subjects

Referral population

Three hundred and thirty-six consecutive premenopausal women reported from 2006 to 2022 to our Reproductive Endocrinology clinic because of symptoms of functional androgen excess or hyperandrogenemia. According to 2023 International PCOS Network criteria [2], these individuals were diagnosed with PCOS using state-of-the-art methodology, including liquid chromatography mass spectrometry/mass spectrometry (LC-MS/MS) assays for diagnosing biochemical hyperandrogenism, sonography for the assessment of polycystic ovarian morphology (PCOM), or anti-müllerian hormone (AMH) immunoassays when the former was not available, as described earlier [21]. We used the 95th percentile of AMH in our local control group of non-hyperandrogenic women as cut-off value [55.9 pM (7.8 ng/mL)] [21]. Other androgen excess or related disorders were systematically excluded including non-classic congenital adrenal hyperplasia, androgen-secreting neoplasms, androgenic/anabolic drug abuse, Cushing syndrome, thyroid dysfunction and, as will be seen, true hyperprolactinemia. To this purpose, blood extraction for the measurement of serum PRL was conducted between 08:00 and 9:00 in a quiet room (basal PRL). An intravenous cannula was placed in the ante-cubital fossa and the baseline sample was withdrawn. Patients remained seated for 15 min, and then a second sample was collected (resting PRL) from the indwelling cannula. When we faced basal and resting PRL concentrations above 25 µg/L and ≤ 25 µg/L, respectively, a diagnosis of stress-induced hyperprolactinemia was established. In the subset of women in whom resting PRL remained elevated, serum was also assayed after precipitation with polyethylene glycol (PEG) to check for the presence of macroprolactinemia [23].

Hyperprolactinemia was defined by basal PRL values above 25 µg/L. For LC–MS/MS and AMH assays, biochemical hyperandrogenism and upper limit of normality, respectively, were defined by values > 95th percentile of a sample of control premenopausal women composed of female volunteers recruited from the hospital’s staff, and overweight or obese women seeking advice for weight loss at our department. These controls had regular menses, showed no signs of hyperandrogenism, and were phenotyped using the same methods used to diagnose PCOS in patients. Patients and controls had no history of oophorectomy or hysterectomy, nor had received treatment with hormonal contraceptives, antiandrogens, or insulin sensitizers for at least 6 months before sampling. This medication history also accounted for recent use of drugs inducing hyperprolactinemia.

All patients and controls provided informed consents allowing us to include their data in a database for research purposes, including this study. That research database was approved by the local Ethics Committee from Hospital Universitario Ramón y Cajal (Date of approval: 7-March-2002; Reference number: 12/02). The informed consent was revised and approved again by the same local Institutional Review Board on 2022, March 10.

Unselected population comprised of female volunteer blood donors

Phenotyping procedures for this population are detailed elsewhere [22]. In brief, investigators attended the blood bank facilities of our center on every afternoon of the work week, inviting all premenopausal women reporting spontaneously for blood donation to participate in the study. Almost all women agreed, with no differences in race, ethnicity or socioeconomic status between women who agreed and the very few who refused participation [22]. We recruited only women who were found suitable for blood donation. We scored hirsutism using the modified Ferriman-Gallwey method also used for the referral population [24]. The presence or absence of acne and androgenic alopecia was recorded, and weight, height, waist, and hip circumferences were measured. A history form was completed, including menstrual dating and irregularity, history of hirsutism and acne, reproductive and gynecological history, and use of medication including oral contraceptive pills, among other data. According to their medication history, in women who were receiving hormonal contraceptives, their menstrual history before treatment and the reason for treatment were recorded, even though the subjects included in the present study had no history of oophorectomy or hysterectomy, nor had received treatment with hormonal contraceptives, antiandrogens, or insulin sensitizers for at least 6 months before sampling. They were not recently used drugs inducing hyperprolactinemia, as well.

None of the subjects were younger than 18-year-old, which is the minimum legal age for blood donation in Spain. In these women, a single blood sample was extracted. Serum and plasma aliquots were stored at − 80 °C until assayed. Androgens were assessed using the same methods above described in the referral population. Since these women were not evaluated in the follicular phase of their menstrual cycle when local normative values were available, circulating AMH levels were not measured in this study subgroup. Serum extractions for non-fasting basal PRL measurements were performed between 12:00 and 14:00, which were the opening hours of our blood bank facilities. In those women with hyperprolactinemia, serum was precipitated with PEG to rule out the presence of macroprolactinemia. Those women with persisting hyperprolactinemia after PEG precipitation were re-evaluated by obtaining two serum samples separated by 15 min, in order to obtain a valid resting PRL measurement.

The study was approved by the local Ethics Committee from Hospital Universitario Ramón y Cajal (Date of approval: 7-November-2005; Minute number: 162), and written informed consent was obtained from all the subjects.

Assays

The technical specifications assays used in measuring sex steroids, lipid profiles, and insulin sensitivity index are detailed elsewhere [21]. Serum AMH concentrations were measured using an automated immunochemiluminescent assay on a Cobas e601 ® analyser (Elecsys ®, Roche Diagnostics, Germany). The assay limits of detection and lower limit of quantification were 0.07 and 0.21 pM, respectively. The intra-assay and inter-assay CVs were < 4 %. The limit above the measuring range was 164 pM.

Serum PRL was assayed using an automated immunochemiluminescence assay (Immulite, Diagnostic Products Corporation, Los Angeles, CA), with 6.2 % and 8.5 % intra-assay and interassay coefficients of variation. Precipitation with PEG was conducted as follows [23]: 200 μL of serum was mixed with 200 μL of a 25 % (wt/vol) solution of 6,000 kDa PEG (Merck-Schuchardt, Art. 807491, Hohenbrunn, Germany) in phosphate buffered saline at pH 7.4. Samples were mixed for 1 min in a vortex mixer and incubated at room temperature for 10 min and immediately centrifuged for 30 min at 1,800 × g. The supernatant was separated and assayed for PRL as described. The result was multiplied by 2 to compensate for the 1:1 dilution during the first step. An untreated aliquot of each serum was run within the same assay. A decrease ≥ 40 % in the serum PRL level after precipitation with PEG was considered indicative of macroprolactinemia.

Power calculations

We used the application provided by the Program of Research in Inflammatory and Cardiovascular Disorders from the Institut Municipal d’Investigació Mèdica, Barcelona, Spain (https://www.datarus.eu/aplicaciones/granmo/) for power calculations. Foreseen a potentially higher prevalence of hyperprolactinemia in women with PCOS than in non-hyperandrogenic control women, after setting α at 0.05 and power at 0.80, the inclusion of 350 women with PCOS and 297 non-hyperandrogenic control women would be able to detect as statistically significant a proportion difference ≥ 4.5 % in the first group with respect to the second, by using the Poisson approximation. This calculation assumes a local prevalence of hyperprolactinemia among premenopausal women of 4.1 %, as previously reported [19].

Statistics

Continuous variables were shown as means ± SD and 95% confidence intervals (95%CI, lower limit; upper limit). We tested the distribution of continuous variables for normality using the Kolmogorov–Smirnov test. Logarithmic transformation was applied to ensure normality, as needed. Discrete variables were presented according to their absolute, relative frequency, and 95%CI using the Wilson method without continuity correction. Continuous variables were compared by univariate general linear models introducing as fixed factors the type of population (referral vs. unselected) and the diagnosis of the women (PCOS vs. controls). The odds ratios (OR)/Exp(β) (95%CI) for hyperprolactinemia according to diagnosis (PCOS vs. control) or population (unselected vs. referral) and their interaction were calculated using logistic binary regression. To address possible associations between basal clinical or biochemical and PRL levels, we implemented linear general regression models introducing as fixed independent variables diagnosis and type of study population, and introducing one by one other clinical and biochemical variables such as age, body mass index (BMI), waist circumference, total and calculated free testosterone, androstenedione, and dehydroepiandrosterone-sulfate (DHEA-S). In the referral population, we also checked for potential associations of serum PRL with lipid parameters, insulin sensitivity index, and serum estradiol as measured by LC-MS/MS.

We used SPSS Statistics 22.0 (SPSS Ibérica, Madrid, Spain) for the analyses. P < 0.05 was considered statistically significant.

Results

The baseline characteristics of women with a diagnosis of PCOS and non-hyperandrogenic control women are displayed in Table 1. When considering the referral and unselected populations as a whole, women with PCOS (n = 350) had higher circulating androgens compared with control women (n = 297). Overall, women with PCOS also showed higher BMI and waist circumference, and were more likely obese, compared with non-hyperandrogenic women. However, this finding resulted mostly from the marked differences observed in the unselected population, as indicated by the statistically significant interaction found in the general lineal model.

Table 1.

Anthropometric, sex steroids, lipid and insulin sensitivity profiles of non-hyperandrogenic control women and women with polycystic ovary syndrome from our clinic (referral population) and voluntary donors (unselected population) cohorts.

	Referral population						Unselected population						PCOS		Population		Interaction
	Control women			Women with PCOS			Control women			Women with PCOS
	(n = 90)			(n = 336)			(n = 207)			(n = 14)			F/U/Exp(β)	P	F/Exp(β)	P	F/Exp(β)	P
Age, years	29	±	6	27	±	6	23	±	7	28	±	7	10.3	0.001	8.2	0.004	0.8	0.369
BMI, kg/m2	27.2	±	7.4	27.8	±	7.3	24.2	±	3.7	27.6	±	5.3	4.5	0.035	2.6	0.105	2.0	0.155
Obesity, n (%)	28 (31)			111 (33)			17 (8)			4 (29)			1.1	0.730	0.2	<0.001	4.1	0.042
WC, cm	81	±	16	82	±	17	78	±	9	85	±	10	3.3	0.071	0.0	0.887	2.7	0.099
Total T, nM	0.8	±	0.3	1.2	±	0.6	0.7	±	0.3	1.1	±	0.6	34.8	<0.001	8.4	0.004	0.2	0.625
SHBG, nM	54	±	25	44	±	27	50	±	39	32	±	23	9.0	0.003	3.2	0.074	0.8	0.377
Calculated free T, pM	14	±	5	24	±	13	13	±	7	26	±	12	52.6	<0.001	0.1	0.776	0.5	0.105
Androstenedione, nM	4.5	±	1.5	6.6	±	2.7	3.5	±	1.4	5.1	±	1.7	43.4	<0.001	19.1	<0.001	0.1	0.781
DHEA-S, µM	5.0	±	2.4	6.5	±	3.3	3.4	±	2.0	4.3	±	3.0	7.8	0.005	21.4	<0.001	0.6	0.460
Estradiol, pM	208	±	184	147	±	107	−			−			11,502	<0.001	−	−	−	−
AMH, pM	22.4	±	15.4	48.6	±	35.1	−			−			6,516	<0.001	−	−	−	−
ISI-Matsuda	7.3	±	3.6	6.3	±	4.5	−			−			11,456	0.003	−	−	−	−
Total cholesterol, mM	4.6	±	0.9	4.6	±	0.9	−			−			14,610	0.623	−	−	−	−
HDL-cholesterol, mM	1.5	±	0.4	1.4	±	0.3	−			−			11,895	0.025	−	−	−	−
LDL-cholesterol, mM	2.7	±	0.7	2.8	±	0.8	−			−			13,939	<0.001	−	−	−	−
Triglycerides, mM	0.9	±	0.5	1.0	±	0.6	−			−			13,211	0.066	−	−	−	−
PCOS phenotype
Classic	−			238 (71)			−			13 (93)			−	−	−	−	−	−
Ovulatory	−			23 (7)			−			0 (0)			−	−	−	−	−	−
Normoandrogenic	−			75 (22)			−			1 (7)			−	−	−	−	−	−
Basal PRL, ug/L	13	±	7	15	±	9	10	±	7	11	±	7	1.3	0.259	7.7	0.006	0.2	0.644

Abbreviations, AMH, anti-müllerian hormone, BMI, body mass index, DHEA-S, dehydroepiandrosterone-sulphate, HDL, high density lipoprotein, ISI, insulin sensitivity index, LDL, low density lipoprotein, PRL, prolactin, SHBG, sex hormone-binding globulin, T, testosterone, WC, waist circumference. Local reference ranges and normality cut-offs for adult premenopausal women in follicular phase: androstenedione (≤ 7.4 nM); anti-müllerian hormone (≤ 55.9 pM); calculated free testosterone (≤ 23.2 pM); dehydroepiandrosterone-sulfate (≤ 9.1).

µM), estradiol (77 to 554 pM); total testosterone (≤ 1.6 nM); HDL-cholesterol (≥ 1.3 mM); LDL-cholesterol (≤ 3.3 mM) insulin sensitivity index (≥ 3.5); prolactin (≤ 25 µg/L); sex hormone-binding globulin: (11 to 180 nM); total cholesterol (≤.5.2 mM); triglycerides (≤ 1.7 mM).

Data are shown as means ± SD or counts (%). Continuous variables were compared by one‐way ANOVA, except for ISI-Matsuda, lipid parameters, estradiol, and basal prolactin comparisons which were performed by U Mann-Whitney test. Logarithmic transformations were applied as needed. Dichotomous variables were compared by applying a binary logistic regression model. Free testosterone was calculated by Vermeulen’s formula using a default albumin level of 43 g/L [36].

There were no statistical differences in mean basal PRL concentrations between women with PCOS and controls from referral (n = 426) and unselected populations (n = 221) (Table 1), although mean PRL concentrations were mildly lower in blood donors than in women recruited at the clinical level. Thirty-nine out of 647 study subjects (6.0%, 95%CI: 4.4; 8.1) had hyperprolactinemia. We found no differences in the frequency of hyperprolactinemia between women with PCOS and non-hyperandrogenic control women (OR: 2.20, 95%CI: 0.25; 19.23; P = 0.477; Fig. 1, left panel). Also, no differences were observed between referral (31 out of 426, 7.3%, 95%CI: 5.2; 10.0) and unselected populations (eight out of 221, 3.8% 95%CI: 1.9; 7.2) (OR: 1.68, 95%CI: 0.52; 5.44; P = 0.387] (Fig. 1, right panel). There was no significant interaction between diagnosis and type of study population on the presence of hyperprolactinemia either [Exp(β): 0.65, 95%CI: 0.06; 7.03; P = 0.722].

Fig. 1.

Distribution of hyperprolactinemic women considering all study populations as a whole and as a function of study subgroups. The figures above the bars display frequencies (%) and their 95% confidence intervals. The numbers inside the bars show counts. The figures under the x-axis indicate the individuals included in each study subgroup.

In the referral population, 18 out of 31 women with hyperprolactinemia normalized PRL concentrations after resting (58.0%, 95%CI: 40.8; 73.6), and another nine women normalized PRL concentrations after PEG precipitation (29.0%, 95%CI: 16.1; 46.6). There were no statistically significant differences in these figures between women with PCOS and non-hyperandrogenic control women (Fig. 2). Only four women, all of them with a PCOS diagnosis, had “true” hyperprolactinemia. Two of them presenting at the time of the study with PRL values after resting and after PEG precipitation of 32 and 39 µg/L, respectively, showed normal basal and resting PRL at later assessments during follow-up. Another patient who presented with basal and resting PRL concentrations of 88 and 89 µg/L showed a pituitary microadenoma on nuclear magnetic resonance imaging. This woman was treated with cabergoline for 32 months; during treatment pituitary imaging no longer revealed the microadenoma and PRL values after PEG precipitation decreased to 28 µg/L even after withdrawing cabergoline. The remaining woman had PRL levels of 49 µg/L after resting and PEG precipitation and was diagnosed of idiopathic hyperprolactinemia after normal pituitary imaging. Cabergoline treatment was given for 6 months, and PRL concentrations normalized even after dopamine agonist withdrawal.

Fig. 2.

Distribution of women with stress-induced hyperprolactinemia and macroprolactinemia among hyperprolactinemic women considering all study population as a whole and as a function of study subgroups. The figures above the bars display frequencies (%) and their 95% confidence intervals. The numbers inside the bars show counts. The figures under the x-axis indicate the individuals included in each study subgroup.

In the unselected population, three out of eight women with hyperprolactinemia showed normal PRL concentrations after PEG precipitation (37.5%, 95%CI: 13.7; 69.4) (Fig. 2). PRL after PEG precipitation ranged from 27 to 43 µg/L in the remaining five women, one within the PCOS group, and four from the non-hyperandrogenic control group. These women were re-evaluated obtaining two new serum samples for measuring basal and resting PRL. All of them showed normal serum PRL concentrations in the re-evaluation. Therefore, stress-related hyperprolactinemia accounted for 62.5% (95%CI: 30.6; 86.3) cases of hyperprolactinemia in this study population (Fig. 2).

After adjusting by PCOS diagnosis and type of study population in linear regression models, basal PRL levels only showed significant associations with age (standardized β: −0.10; P = 0.013), androstenedione (standardized β: 0.09; P = 0.048), and DHEA-S (standardized β: 0.29; P < 0.001). When these variables were introduced together within a single model, only DHEA-S maintained a weak statistically significant association with PRL (Fig. 3). When PRL levels after resting and/or PEG precipitation were introduced as the dependent variable (“true” serum prolactin), these findings were virtually the same (Fig. 3). Accordingly, mean serum PRL was higher in women with abnormally elevated DHEA-S levels (n = 90) compared with those with normal concentrations (n = 557) when all the women in the study were considered as a whole (14.4 ± 10.3 vs. 10.9 ± 5.4 µg/L; P < 0.001). After excluding those two women who required transitory treatment with cabergoline, the model also retained the association between DHEA-S and serum PRL. When these associations were tested only in those women derived from the referral population, there was no significant impact of being a patient or control on “true” PRL levels, but we observed a significant association between them and age (standardized β: −0.12; P = 0.018) and DHEA-S (standardized β: 0.23; P < 0.001). Furthermore, these coefficients virtually remained identical after adjusting by circulating estradiol, which did not show any significant association with PRL concentrations (standardized β: 0.03; P = 0.484). Finally, we did not find any associations between circulating PRL and lipid profiles or insulin sensitivity indices in those women from our referral population (data not shown).

Fig. 3.

Determinants of serum prolactin levels. The upper panel figures show the main determinants of serum basal prolactin concentrations and serum prolactin concentrations after ruling out stress-induced hyperprolactinemia and macroprolactinemia in those women with hyperprolactinemia, respectively. These associations were analyzed by linear regression models adjusting by study population, and PCOS or control status. The low panel figures show the correlation between circulating dehydroepiandrosterone-sulfate and serum prolactin. Solid and dashed black lines represent β of simple linear regression and its 95% confidence interval.

Discussion

The first objective in this study sought to report the prevalence of hyperprolactinemia among women with a clinical picture that led finally to a diagnosis of PCOS. This ranging from 5.4 to 11.0% without significant differences compared to non-hyperandrogenic women with regular menses either from our referral population or from a group of unselected voluntary blood donors. Stress-hyperprolactinemia was the most common cause of elevated basal PRL levels accounting in the whole group of study subjects independently either of being women with PCOS or controls and their study population, followed by the presence of macroprolactinemia. In short, the actual prevalence of “real” hyperprolactinemia in the whole study population was only 0.6% corresponding to four cases among referral women with PCOS.

These findings confirm our previous reports communicating stress-induced hyperprolactinemia and macroprolactinemia as the most common etiologies in premenopausal non-pregnant women with elevated PRL levels both in women from the general population [19] and in women presenting with hyperandrogenic symptoms [17]. In conceptual agreement, Hayashida et al. [25] reported a prevalence of stress-induced hyperprolactinemia and macroprolactinemia of 52.9% and 47.1% among 34 out of 277 Brazilian women with PCOS presenting with increased PRL concentrations at the clinical setting.

A large Dutch study [26] recently reported 15 cases of hyperprolactinemia from 1,429 women with PCOS, for an overall prevalence of 1.1%. There were no statistically significant differences when these patients were compared with non-hyperandrogenic women with regular menses. Five of those 15 women were diagnosed with PRL-secreting pituitary adenomas, yet the authors did not report any specific cause of increase PRL levels in the remaining 10 women [26]. Another similar study found that hyperprolactinemia prevalence was 11.6% (76 cases) among 657 Korean women with PCOS defined by Rotterdam criteria [27]. Thirteen of them (17%) were diagnosed with a pituitary adenoma, suggesting 53.9 µg/L as an optimal cut-off for detecting prolactinomas derived from receiver operating characteristic (ROC) analysis. However, the diagnosis of the remaining 53 women was not clarified, even though 40 of them presented with normal pituitary imaging [27]. In agreement with this report, and our own data, a similar prevalence (14.4%) was reported in patients from United Kingdom [28]. In this communication, 12 patients were receiving medications known to affect PRL levels and two were pregnant. Although apparently not systematically evaluated, a case of stress-related hyperprolactinemia was reported [28]. Screening for macroprolactin was performed in 19 patients with hyperprolactinemia and normal menstrual cycles, being positive in only one case. A pituitary adenoma was observed in 25 of these women. In this series, a PRL cut-off value of 47.5 µg/L could correctly identify all patients with prolactinomas, albeit with a low specificity. Eighteen women with mild hyperprolactinemia did not have an identifiable cause for their elevated PRL levels. Of note, these latter women did not have any significant difference with patients with PCOS and normal PRL levels in terms of clinical, hormonal, or metabolic variables [28]. In a French cohort [18], 50 out of 179 (28%) women fulfilling Rotterdam PCOS criteria presented with hyperprolactinemia. When a second PRL measurement was extracted on a separate day, only 21 of them remained above the normal range (11.7%), suggesting that 58% of cases could have been associated to stress. Five had a microprolactinoma. Based on the ROC curve, a PRL threshold of 60 µg/L showed 100% sensitivity for the diagnosis of microprolactinoma [18]. Another five women had a diagnosis of macroprolactinemia, four cases were drug-induced, and six mild PRL elevations were finally categorized as idiopathic. In conceptual agreement with our findings, these reports support the presence of hyperprolactinemia in approximately one out of 10 women fulfilling PCOS criteria, although those idiopathic cases or “PCOS-induced” are anecdotic, since most of them are likely associated to venipuncture stress or the presence of macroprolactinemia. Therefore, a causal patho- physiologic link between PCOS and hyperprolactinemia is seriously questioned in light of these data.

Furthermore, the impact of PRL on the phenotype of women with PCOS appears to be limited, if any, at least from a clinical point of view [29], in disagreement with other studies [7]. This apparent discrepancy may rely on inadequately controlled confounding factors in those reports, including obesity, stress-hyperprolactinemia, and macroprolactinemia among others [7]. While no influence of serum PRL on anthropometric or metabolic parameters was detected, we observed a mild, although consistent, association between DHEA-S and serum PRL levels. Women with PCOS present with adrenal hyperandrogenism without any PRL increase in a significant number of cases [30]. However, the association between adrenal hyperandrogenism and PRL might be supported by the findings of some clinical trials reporting a slight reduction in circulating DHEA-S concentrations after the administration of cabergoline to women with PCOS [31]. PRL has a direct synergistic effect with ACTH on adrenal cells to increase adrenal androgen release [32]. This stimulatory effect might be maintained even in subjects with PRL levels within the normal range [33]. Even if adrenal hyperandrogenism might influence the clinical and metabolic profile of women with PCOS [34], the role played by PRL concentrations on the cardiovascular risk of women with PCOS remains uncertain at the very best [6,35].

Among the strengths of our study, we may highlight the careful phenotyping of patients and controls using state-of-the-art methodology. In the same line, using voluntary blood donors as study population overcomes potential referral biases since: (i) blood donors are usually healthy people, not seeking medical care for any reason at the time of donation; (ii) blood donation is not rewarded in any way in Spain, and there is no bias derived from the socio-economic background; and (iii) there is no preselection of blood donors, as they report to the hospital voluntarily without any schedule. However, our study was not free of limitations: i) women with PCOS in our series mostly presented with the classic hyperandrogenic phenotype; (ii) despite a careful medication history was taken, drug-induced hyperprolactinemia might have been overlooked in very few women from our referral population since blood extractions were conducted after that first evaluation; (iii) while our aim was to estimate the prevalence of hyperprolactinemia, no pathological cases were actually identified among our unselected study population; (iv) food ingestion may have overestimated the prevalence of venipuncture stress-induced hyperprolactinemia among voluntary blood donors not presenting at fasting to blood extraction; and (v) because blood donors may not be entirely representative of the general population – in example, it is possible that subjects with a pre-existing diagnosis of hyperprolactinemia who may be on medications are less likely to volunteer for blood donation – the true prevalence of hyperprolactinemia in the general population might have been underestimated by our study.

Conclusions

Hyperprolactinemia is not more likely among women with PCOS than in non-hyperandrogenic women with regular menses. The most common causes of mildly increased PRL levels in non-pregnant premenopausal women not taking drugs are functional or analytical causes such as venipuncture stress and the presence of macroprolactinemia, and thus must not preclude a diagnosis of PCOS in women with suggestive symptoms. Moreover, both causes of hyperprolactinemia may lead to unnecessary patient concerns and expensive diagnostic and therapeutic approaches. Finally, albeit serum PRL is directly related to circulating DHEA-S, such an association has no apparent impact on the clinical or metabolic profile of women with PCOS.

CRediT authorship contribution statement

Manuel Luque-Ramírez: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Project administration, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization. Alejandra Quintero-Tobar: Writing – review & editing, Validation, Investigation, Data curation. María Ángeles Martínez-García: Writing – review & editing, Investigation, Formal analysis. Sara de Lope Quiñones: Writing – review & editing, Validation, Investigation, Data curation. María Insenser: Writing – review & editing, Investigation, Formal analysis. Lía Nattero-Chávez: Writing – review & editing, Validation, Investigation. Héctor Francisco Escobar-Morreale: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Project administration, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Luque-Ramirez & Escobar-Morreale reports financial support was provided by Carlos III Health Institute. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.