Positive association of serum prolactin concentrations with all-cause and cardiovascular mortality

Abstract

Aims

Increased serum prolactin (PRL) concentrations have been associated with adverse cardiovascular risk profiles, but the relation between PRL and mortality risk is unknown.

Methods and results

We evaluated 3929 individuals (1946 men and 1983 women) aged 20–81 (mean 50.3 years) from the population-based Study of Health in Pomerania (SHIP). Associations of continuous [per standard deviation (SD) increase] and categorized (sex-specific tertiles) serum PRL concentrations with all-cause and cause-specific mortality were analysed separately for men and women by age- and multivariable-adjusted Cox regression models. During a median follow-up period of 10.1 years (38 231 person-years), 419 deaths (10.7%), 132 cardiovascular deaths (3.4%), and 152 cancer deaths (3.9%) were observed. After multivariable adjustment, we observed a positive association of PRL with all-cause mortality in men and women [hazard ratio (HR) per SD increase: 1.17, 95% confidence interval (CI): 1.07–1.29 and HR: 1.22, 95% CI: 1.03–1.46, respectively]. Similarly, individuals with PRL concentrations in the highest tertile (when compared with lowest PRL tertile) experienced the highest mortality risk (men: HR, 1.75; 95% CI, 1.32–2.32; women: HR, 1.66; 95% CI, 1.08–2.56), with a significant trend across PRL tertiles (P­ for trend <0.05). Cause-specific mortality analyses yielded similar associations for cardiovascular death in both sexes, but for cancer death only in men.

Conclusion

This is the first study to report an independent positive association of PRL concentrations with all-cause and cardiovascular mortality. Further studies are required to confirm our findings and to elucidate the potential role of PRL as a useful biomarker of cardiovascular risk and mortality assessment.

Prolactin, Mortality, Cardiovascular disease, Population-based cohort

Introduction

Prolactin (PRL) is a multifunctional pituitary hormone, which has metabolic actions that are not confined to the lactating mammary gland. With receptors expressed in nearly all organs, PRL is involved in numerous physiological and pathophysiological processes spanning the reproductive, metabolic, osmoregulatory, and immunoregulatory systems.^1–3 In patients with prolactinomas, pathologically increased serum PRL concentrations are associated with an adverse cardiovascular risk profile, typically characterized by insulin resistance, low-grade inflammation, and impaired endothelial function.^4,5 In women, higher serum PRL concentrations are positively associated with systemic blood pressure, aortic stiffness,⁶ and incident hypertension.⁷ We previously reported from our population-based sample positive associations between serum PRL concentrations, inflammatory biomarkers,⁸ and anthropometric measurements,⁹ respectively. But whether these previously shown associations between serum PRL and adverse cardiovascular risk profiles directly translate into an increased mortality risk is unknown. Therefore, we investigated associations of serum PRL concentrations with all-cause and cause-specific mortality over a 10-year follow-up period of a cohort of 3929 men and women from the population-based Study of Health in Pomerania (SHIP).

Methods

Study population

Study of Health in Pomerania is a population-based cohort study in West Pomerania, a region in northeastern Germany. Details regarding the SHIP study design, recruitment, and procedures have been published previously.^10,11 In brief, from the total population of West Pomerania comprising 213 057 inhabitants in 1996, a two-stage stratified cluster sample of adults aged 20–79 was drawn. The net sample (without migrated or deceased persons) comprised 6265 eligible subjects. After written informed consent was obtained, 4308 (2192 women) participants were examined (response proportion 68.8%) between 1997 and 2001. The study conformed to the principles of the Declaration of Helsinki and was approved by the Ethics Committee of the University of Greifswald. Of the 4308 baseline participants, we excluded participants due to the following reasons: PRL not measured (n = 166), PRL concentrations <0.5 µg/L (n = 3), PRL concentrations >30 µg/L (n = 67),¹² pregnancy (n = 3), and missing covariate data (n = 140). This resulted in a final study sample of 3929 individuals (1946 men and 1983 women).

Measures

Data about age, sex, smoking habits, educational status, medical history, menopausal status, parity (number of children), and hormone replacement therapy (HRT) [using the Anatomical Therapeutic Chemical (ATC) codes G03C, G03D, and G03F] were collected by computer-assisted personal interviews. Type 2 diabetes was defined based on self-reported physician's diagnosis or use of antidiabetic medication (ATC code A10) in the last 7 days, or glycated haemoglobin (HbA1c) concentrations >6.5%. Prevalent cardiovascular disease (CVD) was defined by a summative score comprising information about (i) peripheral artery disease defined according to the four stages proposed by Rose et al.,¹³ (ii) heart failure based on recommendations of the New York Heart Association,¹⁴ (iii) angina pectoris defined according to Rose Angina Questionnaire,¹³ and a recall of physician's diagnoses of (iv) stroke and (v) myocardial infarction. Waist circumference (WC) was measured to the nearest 0.1 cm using an inelastic tape midway between the lower rib margin and the iliac crest in the horizontal plane, with the subject standing comfortably with weight distributed evenly on both feet. After a resting period of at least 5 min, systolic and diastolic blood pressure was measured three times on the right arm of seated subjects by use of an oscillometric digital blood pressure monitor (HEM-705CP, Omron Corporation, Tokyo, Japan). The interval between the readings was 3min. The mean of the second and third measurements was calculated, and hypertension was defined as systolic or diastolic blood pressure of >140 or >90 mmHg, respectively, or the use of antihypertensive medication (ATC codes C02, C03, C04, C07, C08, and C09).¹⁵ Estimated glomerular filtration rate (eGFR) was calculated using the modification of diet in renal disease equation¹⁶: eGFR (mL/min/1.73 m²)= 186 × serum creatinine (mg/dL)^−1.154 × age (years)^−0.203 × (0.742 if female).

Analysis of serum prolactin

Serum PRL was measured from frozen sera of participants at the baseline examination. Non-fasting blood samples were drawn from the cubital vein in the supine position between 7.00 a.m. and 6.00 p.m. Samples were stored at −80°C until analysis using a chemiluminescent immunometric assay on an Immulite 2500 analyzer (Ref. L5KPR, DPC Biermann GmbH, Bad Nauheim, Germany). An aliquot of two alternating levels of a third party commercial control material (Bio-Rad Lyphochek Immunoassay Plus Control, lot 40151 and lot 40152; Bio-Rad, Munich, Germany) was included in each series in single determination. During the course of the study, the inter-assay coefficient of variation was 5.6% with a systematic deviation of −4.3% at the 6.3 µg/L level and 4.3% with a systematic deviation of −6.4% at the 14.5 µg/L level.

Serum low-density lipoprotein (LDL) cholesterol was measured by applying a precipitation procedure using dextran sulphate (Immuno, Heidelberg, Germany) on an Epos 5060 (Eppendorf, Hamburg, Germany). Total cholesterol concentrations were measured photometrically (Hitachi 704, Roche Diagnostics, Mannheim, Germany). HbA1c concentrations were determined by high-performance liquid chromatography (Bio-Rad Diamat). Non-fasting serum glucose concentrations were determined enzymatically using reagents from Roche Diagnostics (Hitachi 717, Roche Diagnostics). High-sensitive C-reactive protein (hsCRP) was determined immunologically on a Behring Nephelometer II with commercially available reagents from Dade Behring (Dade Behring, Eschborn, Germany). Plasma fibrinogen concentrations were assayed according to Clauss using an Electra 1600 analyzer (Instrumentation Laboratory, Barcelona, Spain). White blood cell count (WBC) was performed within 60 min after blood sampling with a Coulter Max M analyzer (Coulter Electronics, Miami, USA).⁸ All assays were performed according to the manufacturers' recommendations by skilled technical personal. The laboratory takes part in official quarterly German external proficiency testing programs.

Mortality follow-up

Information on vital status was collected from population registries at annual intervals from time of enrolment into the study through 15 December 2009. Subjects were censored at either death or loss to follow-up, and the number of months between baseline examination and censoring was used as follow-up length. Death certificates were requested from the local health authority at the place of death. Causes of deaths were coded by a certified nosologist according to the International Classification of Diseases, 10th revision and classified according to CVD (I10–I79) and cancer (C00–C97).¹⁷ Additionally, two internists (H.W. and M.D.) independently validated the underlying cause of death in each case and performed a joint reading in cases of disagreement. A third internist (H.V.) finally decided in cases of still existing disagreement.

Statistical analysis

The baseline characteristics of the study population are presented for men and women as percentages or median (25th and 75th percentiles). The χ²-test (categorical data) or the Mann–Whitney U-test (continuous data) was performed for intergroup comparisons by sex. We used restricted cubic splines to detect a possible non-linear dependency of the log hazard function on PRL concentrations.¹⁸ Three knots were pre-specified located at the 10th, 50th, and 90th percentiles, as recommended by Stone and Koo.¹⁹ Spline functions were estimated separately for men and women using multivariable Cox proportional hazard regression models,²⁰ including age, smoking habits, educational level, WC, LDL cholesterol, total cholesterol, eGFR, type 2 diabetes, hypertension, and CVD. Based on these results, we performed subsequent Cox proportional hazard regression models including PRL concentrations categorized into sex-specific tertiles using the lowest PRL tertiles as the reference group. The P for trend test was conducted by including PRL tertiles as an ordinal score to the regression models. To increase statistical power, we additionally modelled continuous PRL concentrations, estimating the effect on mortality risk per 1SD increase. Effect estimates were presented as hazard ratio (HR) and 95% confidence interval (95% CI). The proportional hazard assumption was confirmed for all variables using Schoenfeld's tests as well as by visual inspection of graphed Schoenfeld residuals and log–log plots.²¹ Interaction terms between age and each covariate were investigated using multivariable models and retained in the models when P-values were <0.10.

To analyse the improvement in mortality risk prediction after the inclusion of a continuous PRL variable into the multivariable model, we calculated C-statistics (95% CI) using STATA's stcox post-estimation command “estat concordance” and tested the differences using the “roccomp” command. Several sensitivity analyses were performed. To evaluate the potential influence of medication-induced hyperprolactinaemia on our revealed estimates, we additionally excluded 411 individuals treated with antipsychotics (ATC code N05A), antidepressants (N06A), verapamil (C08DA01), methyldopa (C02AB), reserpine (C02LA01), metoclopramide (A03FA01 or N02CX59), domperidone (A03FA03), or cimetidine (A02BA01).²² To analyse the impact of acute illness or underlying comorbidity on the observed associations, we also excluded deaths within the first year of follow-up (n = 22) or individuals with baseline CVD (n = 722), respectively. Based on the covariates considered, we calculated inverse probability weights and included them into the multivariable models to evaluate possible non-response bias due to missing data.²³ To assess the potential confounding effect of the diurnal cycle in PRL secretion, we additionally adjusted multivariable models for the time of blood sampling. Finally, we estimated multivariable models in women, additionally adjusting for parity, menopausal status, and HRT. We further explored potential mediating pathways of inflammation, blood pressure, insulin resistance, and obesity in the observed association between PRL and mortality by additionally including hsCRP, plasma fibrinogen, WBC, systolic blood pressure, HbA1c, glucose, and WC into age- and sex-adjusted Cox regression models.²⁴ Two-sided probability values <0.05 were considered statistically significant. All statistical analyses were performed using Stata 11.0 (Stata Corporation, College Station, TX, USA).

Results

During 38 231 person-years (median: 10.1 years; 25th percentile: 9.4; 75th percentile: 10.7) of follow-up, 419 individuals (10.7%, 136 women) died, reflecting an overall crude mortality rate of 10.7 deaths per 1000 person-years. Of these, 132 individuals (3.4%, 45 women) died from cardiovascular causes and 152 individuals (3.9%, 41 women) from cancer. Baseline characteristics of the study sample are presented stratified by sex (Table 1). Multivariable Cox regression models revealed a positive association between PRL and mortality showing that each SD increment in PRL was associated with a higher all-cause (men: HR, 1.17; 95% CI, 1.07–1.29; women: HR, 1.22; 95% CI, 1.12–2.06) and cardiovascular mortality risk (men: HR, 1.28; 95% CI, 1.07–1.53; women: HR, 1.52; 95% CI, 1.12–2.06), but not cancer mortality risk (men: HR, 1.18; 95% CI, 0.99–1.39; women: HR, 0.89; 95% CI, 0.59–1.35) (Table 2). Similarly, spline analyses also showed a positive association with a more steep increase in the predicted log hazard as a function of PRL among men (Figure 1). Unadjusted Kaplan–Meier survival analyses reflected this more pronounced linear association yielding significantly shorter survival times among men with PRL concentrations in the highest PRL tertiles, but not among women with (Figure 2).

Table 1

Baseline characteristics of the study sample stratified by sex

Variable	Men (n = 1946)	Women (n = 1983)	P-value
Age (years)	51.9 (37.4; 65.2)	49.2 (35.9; 62.3)	<0.001
Serum prolactin (µg/L)	4.9 (3.6; 6.8)	6.4 (4.5; 9.3)	<0.001
Waist circumference (cm)	95.2 (87.5; 102.9)	81.5 (72.8; 92.0)	<0.001
Current smoker (%)	33.6	26.9	<0.001
Educational status (%, years)
<10	42.6	37.5	<0.001
=10	40.1	46.9
>10	17.4	15.6
Systolic blood pressure (mmHg)	140.5 (129.0; 153.0)	126.5 (114.5; 142.5)	<0.001
Diastolic blood pressure (mmHg)	85.0 (78.0; 93.5)	80.0 (73.5; 87.5)	<0.001
Hypertension (%)	61.4	37.9	<0.001
Total cholesterol (mmol/L)	5.69 (4.94; 6.43)	5.65 (4.88; 6.49)	0.616
LDL cholesterol (mmol/L)	3.54 (2.83; 4.26)	3.42 (2.68; 4.21)	0.001
eGFR (mL/min/1.73 m²)	83.1 (73.8; 92.9)	74.9 (66.7; 84.2)	<0.001
Type 2 diabetes (%)	12.1	8.9	0.001
Cardiovascular disease (%)	18.4	18.4	0.974
Hormone replacement therapy (%)	NA	20.7	<0.001

Data are given as percentages or median (25th and 75th percentiles). LDL, low-density lipoprotein cholesterol; eGFR, estimated glomerular filtration rate.

P-values were calculated from χ²-test for categorical data, and Whitney–Mann U-test for continuous data.

Table 2

Cox regression models for the association of serum prolactin with all-cause and cause-specific mortality in men and women

	Per SD increase	Prolactin tertiles			P for trend
		First tertile	Second tertile	Third tertile
Men (n = 1946)
All-cause mortality (# deaths)	283	95	81	107
Age-adjusted HR (95% CI)	1.16 (1.05; 1.28)*	(Ref.)	0.90 (0.67; 1.22)	1.57 (1.19; 2.07)*	0.002
Multivariable-adjusted HR (95% CI)a	1.17 (1.07; 1.29)*	(Ref.)	0.97 (0.72; 1.31)	1.75 (1.32; 2.32)*	<0.001
Cardiovascular disease mortality (# deaths)	87	26	29	32
Age-adjusted HR (95% CI)	1.23 (1.02; 1.47)*	(Ref.)	1.05 (0.62; 1.78)	1.83 (1.09; 3.07)*	0.012
Multivariable-adjusted HR (95% CI)a	1.28 (1.07; 1.53)*	(Ref.)	1.08 (0.63; 1.85)	2.16 (1.27; 3.67)*	0.003
Cancer mortality (# deaths)	111	38	30	43
Age-adjusted HR (95% CI)	1.12 (0.94; 1.32)	(Ref.)	0.82 (0.51; 1.32)	1.64 (1.06; 2.53)*	0.032
Multivariable-adjusted HR (95% CI)a	1.18 (0.99; 1.39)	(Ref.)	0.97 (0.60; 1.57)	1.99 (1.27; 3.10)*	0.004
Women (N = 1983)
All-cause mortality (# deaths)	136	52	46	38
Age-adjusted HR (95% CI)	1.16 (0.97; 1.38)	(Ref.)	1.17 (0.78; 1.73)	1.47 (0.96; 2.23)	0.078
Multivariable-adjusted HR (95% CI)a	1.22 (1.03; 1.46)*	(Ref.)	1.30 (0.87; 1.96)	1.66 (1.08; 2.56)*	0.020
Cardiovascular disease mortality (# deaths)	45	18	10	17
Age-adjusted HR (95% CI)	1.27 (0.95; 1.71)	(Ref.)	0.72 (0.33; 1.55)	1.97 (1.01; 3.84)*	0.174
Multivariable-adjusted HR (95% CI)a	1.52 (1.12; 2.06)*	(Ref.)	0.79 (0.35; 1.76)	2.84 (1.38; 5.89)*	0.045
Cancer mortality (# deaths)	41	15	18	8
Age-adjusted HR (95% CI)	0.87 (0.59; 1.30)	(Ref.)	1.57 (0.79; 3.11)	1.03 (0.43; 2.45)	0.738
Multivariable-adjusted HR (95% CI)a	0.89 (0.59; 1.35)	(Ref.)	1.60 (0.79; 3.26)	1.05 (0.44; 2.54)	0.722

^aAdjusted for age, smoking habits, educational level, waist circumference, low-density lipoprotein cholesterol, total cholesterol, estimated glomerular filtration rate, type 2 diabetes, hypertension, and cardiovascular disease). Parameter estimates were presented as HR and 95% confidence interval (95% CI).

*P< 0.05.

Figure 1

Predicated relative hazard function for all-cause mortality risk as a function of serum prolactin in men (brown line) and women (green line). Estimates were derived from Cox proportional hazards regression analyses with restricted cubic splines adjusted for fixed covariate levels (50 years, waist circumference: 95 cm men; 81 cm women, 5.7 mmol/L cholesterol, 3.5 mmol/L low-density lipoprotein cholesterol, estimated glomerular filtration rate: 83 mL/min/1.73 m² men; 75 mL/min/1.73 m² women, never smoker, no type 2 diabetes, no hypertension, no cardiovascular disease) and the reference point set to the median of the prolactin distribution. Spline functions were estimated separately for men and women. Broken lines indicate limits of sex-specific tertiles (T).

Figure 2

Kaplan–Meier survival curves for sex-specific tertiles of serum prolactin separately for men and women.

Subsequent multivariable Cox regression models with PRL concentrations categorized into sex-specific tertiles confirmed the positive association, showing that individuals with PRL concentrations in the highest tertile had an increased risk of all-cause (men: HR, 1.75; 95% CI, 1.32–2.32; women: HR, 1.66; 95% CI, 1.08–2.56) and cardiovascular mortality (men: HR, 2.16; 95% CI, 1.27–3.67; women: HR, 2.84; 95% CI, 1.38–5.89) (Table 2) with a significant linear trend across PRL tertiles (P < 0.05 for both outcomes). An association between higher PRL concentrations and cancer mortality was observed only in men (Table 2). None of the evaluated interaction terms were retained in the final multivariable model.

We conducted sensitivity analyses with the exclusion of individuals with potential medication-induced hyperprolactinaemia, first-year deaths, or baseline CVD, and additional adjustment for non-response bias, blood sampling time, parity, menopausal status, and HRT, respectively; but without any substantial impact on the overall estimates (Table 3). We further investigated the potential mediating effects of inflammation, blood pressure, insulin resistance, and obesity in the association of PRL with all-cause and CVD mortality (Table 3). But comparing age- and sex-adjusted Cox models with and without the intermediate variables showed that the performed mediation analyses altered the initially revealed estimates only slightly. Assessing the predictive performance, the applied multivariable model yielded a C-statistic of 0.884 (95% CI, 0.869–0.899), whereas the additional inclusion of continuous PRL variable showed a slight, but statistically not significant (P-value 0.127), improvement in the C-statistic: 0.886 (95% CI, 0.871–0.901).

Table 3

Sensitivity and mediation analyses for the association of serum prolactin concentrations with all-cause mortality

	All-cause mortality HR (95% CI)
	Men	Women
Multivariable-adjusted model (per SD increase)a	1.17 (1.07; 1.29)*	1.22 (1.03; 1.46)*
Inclusion of non-response weights	1.13 (1.03; 1.24)*	1.13 (0.94, 1.36)
Adjustment for blood sampling time	1.17 (1.07; 1.29)*	1.22 (1.02; 1.45)*
Adjustment for parity, menopausal status, HT	NA	1.26 (1.04; 1.53)*
Exclusion of baseline CVD (N = 722)	1.19 (1.04; 1.37)*	1.18 (0.93; 1.48)
Exclusion of first-year deaths (N = 22)	1.16 (1.05; 1.28)*	1.22 (1.02; 1.45)*
Exclusion of influential medications (N = 411)	1.20 (1.08; 1.34)*	1.06 (0.81; 1.39)
	Mediation analyses
	Men	Women
Age-adjusted model (per SD increase)	1.16 (1.05; 1.28)*	1.16 (0.97; 1.38)
High-sensitive C-reactive protein	1.17 (1.06; 1.28)*	1.15 (0.97; 1.38)
Plasma fibrinogen	1.15 (1.04; 1.26)*	1.16 (0.98; 1.38)
White blood cell count	1.16 (1.05; 1.27)*	1.16 (0.98; 1.38)
Systolic blood pressure	1.16 (1.05; 1.28)*	1.17 (0.98; 1.39)
Haemoglobin A1c	1.17 (1.06; 1.28)*	1.15 (0.98; 1.37)
Non-fasting serum glucose	1.16 (1.06; 1.28)*	1.15 (0.97; 1.37)
Waist circumference	1.16 (1.05; 1.28)*	1.17 (0.99; 1.39)

^aThe multivariable model was adjusted for age, smoking habits, educational level, waist circumference, low-density lipoprotein cholesterol, total cholesterol, estimated glomerular filtration rate, type 2 diabetes, hypertension, and cardiovascular disease. Parameter estimates were presented as HRs and 95% confidence interval (95% CI). HT, hormone therapy; NA, not applicable; CVD, cardiovascular disease.

*P< 0.05.

Discussion

To our knowledge, this is the first study to show a positive association between serum PRL concentrations and increased risk of all-cause and cardiovascular mortality. This finding was independent of standard risk factors, consistent over a broad range of sensitivity analyses, and based on data from a large population-based sample.

The positive association between serum PRL concentrations and cardiovascular mortality observed in our prospective study is consistent with previous cross-sectional observations, suggesting that higher serum PRL concentrations are associated with an adverse cardiovascular risk profile. Previous small-scale cross-sectional studies showed that hyperprolactinaemic states are associated with insulin resistance,^4,5 low-grade inflammation,^4,8 impaired endothelial function,⁵ increased platelet aggregation,²⁵ increased thrombosis risk,²⁶ and dyslipidemia.²⁷ Additionally, patients with stroke, myocardial infarction, and acute coronary syndromes had significantly higher serum PRL concentrations in comparison with healthy controls.^28,29 Of note, our findings were obtained by evaluating serum PRL within the physiological range (we excluded individuals with PRL concentrations above 30 µg/L, analysing only PRL concentrations within the physiological range between 4.0 and 25.0 µg/L in women and 0.5 and 19.0 µg/L in men, respectively).¹² Potential physiological explanations for the association of higher serum PRL concentrations with subsequent mortality may relate to the broad spectrum of PRL's biological effects ranging from generation of atherogenic phenotypes,³⁰ proliferation of vascular smooth muscle cells,^31,32 and promotion of vasoconstriction³³ to greater oxidative stress that in turn promotes the fragmentation of PRL into its angiostatic and proapoptotic 16 kDa fragment. This 16 kDa fragment^34–36 adversely affects the endothelium, as well as the cardiac vasculature and cardiomyocyte function, and has also been hypothesized as a potential co-factor in the pathogenesis of peripartum cardiomyopathy.³⁷ Taken together, PRL is a hormone that can either stimulate or inhibit various stages of vessel formation or cardiac remodelling^38,39 and thereby leading to defective cardiac angiogenesis, heart failure, and subsequent mortality.^40,41

Another experimental study showed that the PRL receptor is present in the macrophages of the atherosclerotic plaque at sites of most prominent inflammation.⁴² In line with this, we observed in a previous cross-sectional investigation of our population-based sample positive associations between serum PRL concentrations and inflammatory biomarkers including WBC.⁸ Thus, markers of inflammation could possibly mediate the observed association between PRL and mortality risk. But the conducted mediation analyses, assessing mediating effects of inflammation, blood pressure, insulin resistance, and obesity, had virtually no impact on the revealed estimates, providing further evidence for an independent association between serum PRL concentrations and mortality risk. A number of sensitivity analyses further proved the observed associations, since the strength of association between serum PRL concentration and mortality remained largely unchanged after accounting for potential non-response bias, measurement bias (PRL blood sampling time), selection bias (exclusion of first-year deaths and individuals taking hyperprolactinaemia-inducing medications, respectively), and residual confounding (additional adjustment for parity, menopausal status, and HRT in women).

But although our study provides first evidence that higher serum PRL concentrations within the physiological range are associated with increased risk of all-cause and cardiovascular mortality, the exact mechanisms remain to be elucidated in future studies. With only two previous prospective studies,^43,44 the epidemiological evidence about PRL and the risk of future cardiovascular events and mortality is scarce and under investigation for the greater part. A nested case–control study in the EPIC-Norfolk cohort showed no association between serum PRL concentrations and incident coronary artery disease (CAD) among healthy men and women aged 40–79 who were followed-up for an average of 7 years.⁴⁴ In apparent contrast to ours and previous patient-based investigations, results from a patient cohort of men with sexual dysfunction suggest that low-serum PRL concentrations are associated with adverse cardiovascular risk profiles⁴⁵ and increased risk of incident CVD events.⁴³ Possible explanations for the diverging results are differences in study populations (representative population-based sample of men and women from the general population vs. samples of primary care patients⁴⁴ and men with sexual dysfunction,⁴³ respectively), varying follow-up durations (mean 9.7 vs. 7.0 years,⁴⁴ and 4.4 years,⁴³ respectively), and heterogeneous ascertainment of the outcome (population-registry mortality data only vs. hospital-registry data about fatal and non-fatal CAD⁴⁴ and CVD,⁴³ respectively). In accordance with our results, a previous cross-sectional study among 76 menopausal women found a positive correlation between serum PRL concentrations and the European Society of Cardiology HeartScore, a composite index that predicts 10-year cardiovascular mortality.⁶ However, about a possible association between PRL and incident CVD in the present study could be only speculated, given the young mean age (50.3 years) and the insufficient number of incident CVD cases in our sample. But despite this evidence from observational studies, the causal interpretation of observational data has severe limitations⁴⁶ that warrants further research in independent samples from other prospective large-scale epidemiological studies and randomized controlled clinical trials to further evaluate the predictive utility of serum PRL concentrations and its potential clinical implications.

Our investigation has several strengths including a large population-based sample, its long-term follow-up, and the detailed assessment of potential confounders (including hyperprolactinaemia-inducing medications). Potential limitations include a single-serum PRL measurement-based non-fasting blood samples. Like in other large-scale population-based studies, a single blood sample was drawn whenever the participant attended (between 7.00 a.m. and 6.00 p.m.). But serum PRL concentrations have been shown to be relatively stable between 9.00 a.m. and 5.00 p.m. since PRL is secreted in a pulsatile fashion mostly during the night.⁴⁷ Therefore, we were not able to capture any effects from additional adjustment for day time of blood sampling on our revealed estimates. A single PRL measurement may have introduced some bias since serum PRL concentrations may vary over the 10-year follow-up period. But we found no indication for an interaction between age and PRL in our analyses, and measurements which are subject to within-individual variability has been shown to rather underestimate the strength of the investigated association.⁴⁸

In conclusion, we observed that higher serum PRL concentrations within the physiological range are associated with an increased risk of all-cause and cardiovascular mortality in a population-based sample of 3929 men and women who were followed for 10 years. Since the present study is the first to observe an association between PRL and mortality risk, our results should be confirmed in other independent samples. Likewise, the potential role of PRL as a useful biomarker of cardiovascular risk and mortality assessment needs to be further elucidated.

Funding

Study of Health in Pomerania is part of the Community Medicine Research net of the University of Greifswald, Germany, which is funded by the Federal Ministry of Education and Research (01ZZ9603, 01ZZ0103, and 01ZZ0403), the Ministry of Cultural Affairs as well as the Social Ministry of the Federal State of Mecklenburg, West Pomerania. The GANI_MED consortium is funded by the Federal Ministry of Education and Research and the Ministry of Cultural Affairs of the Federal State of Mecklenburg–West Pomerania (03IS2061A). This study was carried out in collaboration with the German Centre for Cardiovascular Research (GCCR) which is funded by the Federal Ministry of Education and Research and the Ministry of Cultural Affairs of the Federal State of Mecklenburg, West Pomerania, Germany.

Conflict of interest: Pfizer provided partial grant support for the determination of plasma samples and data analysis. The PRL reagent used was sponsored by DPC Biermann GmbH, Bad Nauheim, Germany.

Acknowledgements

This work is also part of the research project Greifswald Approach to Individualized Medicine (GANI_MED).