The Journal of Clinical Endocrinology & Metabolism, 2024, 109, 2520–2529

https://doi.org/10.1210/clinem/dgae081
Advance access publication 19 March 2024
Clinical Research Article

Long-Term Glucocorticoid Exposure and Incident

Cardiovascular Diseases—The Lifelines Cohort
Eline S. van der Valk,1,2 Mostafa Mohseni,1,2 Anand M. Iyer,1,2 Maartje J. B. van den Hurk,1,2
Robin Lengton,1,2 Susanne Kuckuck,1,2 Vincent L. Wester,3 Pieter J. M. Leenen,4 Willem A. Dik,4,5
Jenny A. Visser,1,2 Maryam Kavousi,6 Mina Mirzaian,7 Sjoerd A. A. van den Berg,1,7

Downloaded from https://academic.oup.com/jcem/article/109/10/2520/7631418 by guest on 07 November 2024

and Elisabeth F. C. van Rossum1,2
1
Obesity Centre CGG, Erasmus MC, University Medical Center Rotterdam, 3000 CA Rotterdam, The Netherlands
2
Department of Internal Medicine, Division of Endocrinology, Erasmus MC, University Medical Center Rotterdam, 3000 CA Rotterdam,
The Netherlands
3
Department of Internal Medicine, Division of Geriatric Medicine, Erasmus MC, University Medical Center Rotterdam, 3000 CA Rotterdam,
The Netherlands
4
Department of Immunology, Erasmus MC, University Medical Center Rotterdam, 3000 CA Rotterdam, The Netherlands
5
Laboratory of Medical Immunology, Erasmus MC, University Medical Center Rotterdam, 3000 CA Rotterdam, The Netherlands
6
Department of Epidemiology, Erasmus MC, University Medical Center Rotterdam, 3000 CA Rotterdam, The Netherlands
7
Department of Clinical Chemistry, Erasmus MC, University Medical Center Rotterdam, 3000 CA Rotterdam, The Netherlands
Correspondence: Elisabeth F. C. van Rossum, MD, PhD, Department of Internal Medicine and Obesity Center CGG (Centrum Gezond Gewicht), Room Rg-5.
P.O. Box 2400, Erasmus MC, University Medical Center Rotterdam, 3000 CA Rotterdam, The Netherlands. Email: [email protected]

Abstract
Context: Long-term glucocorticoid levels in scalp hair (HairGCs), including cortisol and the inactive form cortisone, represent the cumulative
systemic exposure to glucocorticoids over months. HairGCs have repeatedly shown associations with cardiometabolic and immune
parameters, but longitudinal data are lacking.
Design: We investigated 6341 hair samples of participants from the Lifelines cohort study for cortisol and cortisone levels and associated these
to incident cardiovascular diseases (CVD) during 5 to 7 years of follow-up. We computed the odds ratio (OR) of HairGC levels for incident CVD via
logistic regression, adjusting for classical cardiovascular risk factors, and performed a sensitivity analysis in subcohorts of participants < 60 years
and ≥ 60 years of age. We also associated HairGC levels to immune parameters (total leukocytes and subtypes).
Results: Hair cortisone levels (available in n = 4701) were independently associated with incident CVD (P < .001), particularly in younger
individuals (multivariate-adjusted OR 4.21, 95% CI 1.91-9.07 per point increase in 10-log cortisone concentration [pg/mg], P < .001). All
immune parameters except eosinophils were associated with hair cortisone (all multivariate-adjusted P < .05).
Conclusion: In this large, prospective cohort study, we found that long-term cortisone levels, measured in scalp hair, represent a relevant and
significant predictor for future CVD in younger individuals. These results highlight glucocorticoid action as possible treatment target for CVD
prevention, where hair glucocorticoid measurements could help identify individuals that may benefit from such treatments.
Key Words: hair glucocorticoids, cardiovascular diseases
Abbreviations: BMI, body mass index; CVD, cardiovascular disease; HairGC, glucocorticoid levels in scalp hair; HPA, hypothalamic-pituitary-adrenal;
IL-, interleukin; IQR, interquartile range; OR, odds ratio; WC, waist circumference.

The hypothalamic-pituitary-adrenal (HPA)-axis is crucial for that is present in nearly all cells of the human body (1). Cortisol
homeostasis and survival, as its end products, glucocorticoid can be converted to the inactive hormone cortisone, and vice ver­
hormones, have pleiotropic effects throughout the body and sa, by 11-beta-hydroxysteroid dehydrogenase, which is one of
are important regulators of metabolism and immune function. the mechanisms that regulates the net glucocorticoid effect at tis­
Upon physiological or psychological stress, the HPA-axis is acti­ sue level (2).
vated starting with release of corticotropin-releasing hormone Pathological disturbances in the amounts of circulating
(CRH) and adrenocorticotropic hormone (ACTH) from the glucocorticoids cause a variety of serious clinical symptoms.
hypothalamus and anterior pituitary, respectively. This subse­ In case of severe glucocorticoid excess, known as the
quently leads to increasing levels of circulating glucocorticoid Cushing syndrome, patients develop features resembling the
hormones. metabolic syndrome, including abdominal obesity, insulin re­
In humans, the main active glucocorticoid hormone is cortisol. sistance, and hypertension (3). It has been hypothesized that
Cortisol binds to the glucocorticoid receptor, a nuclear receptor also in the general population increased glucocorticoid action

Received: 18 August 2023. Editorial Decision: 6 February 2024. Corrected and Typeset: 19 March 2024
© The Author(s) 2024. Published by Oxford University Press on behalf of the Endocrine Society.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which
permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited.
The Journal of Clinical Endocrinology & Metabolism, 2024, Vol. 109, No. 10 2521

contributes to metabolic derangements, perhaps to a lesser Manenschijn et al (11)). With alpha set at 0.05 and 1-β set at
extent (4). However, historically it has been difficult to estab­ 0.90, an estimated 1056 participants were required for each
lish a clear relation between the levels of circulating gluco­ quartile. Accounting for an expected response rate of 85%
corticoid and anthropometric measurements in the general (known from the Lifelines second visit) and for anticipated tech­
population, and conflicting results were found when gluco­ nical difficulties in the HairGC measurements, we estimated
corticoid measurements as in blood, saliva, or urine were re­ that we required 6000 to 6500 hair samples for the current
lated to, for example, body mass index (BMI) and waist study.
circumference (WC) (5). However, all these measurements To enable future associations with genetic data, we selected
concern short-term or point measurements and may not the 6341 participants from whom hair samples from 2014 as
represent long-term exposure to glucocorticoids. An interest­ well as genome-wide association study (GWAS) data were avail­
ing perspective may come from hair glucocorticoid (HairGC) able. The follow-up visit was scheduled to take place between
analyses, including the measurements of both cortisol and its October 2019 and April 2020, but eventually this was extended,
inactive form cortisone in proximal scalp hair. This is a reli­ as site visits were postponed due to COVID-19 measures. For

Downloaded from https://academic.oup.com/jcem/article/109/10/2520/7631418 by guest on 07 November 2024

able and easily applicable technique that enables retrospect­ the current research, we used data of participants who had their
ive quantification of the cumulative systemic exposure to visit before December 2021. During the COVID-19 pandemic,
glucocorticoids over weeks to months (6-8). There is now participants were invited to fill in an additional online question­
consistent evidence that long-term HairGC measurements naire (COVID questionnaire). This questionnaire specifically
are correlated to measures of obesity such as BMI and WC concerned questions about health and behavior related to
(9, 10), as well as cardiovascular diseases (CVD) (11-13), COVID-19, but also involved questions related to chronic dis­
but only from cross-sectional studies, as longitudinal data eases and in particular CVD.
are lacking. Also, early research focused solely on cortisol, For the Lifelines cohort study, written informed consent was
but now evidence is mounting that cortisone may show equal provided by participants and study approval was obtained
or even stronger associations with metabolic parameters such from the Medical Ethics Committee of the University Medical
as BMI and WC (9, 14). Center Groningen, Groningen, The Netherlands (2007/152).
For obesity, we have demonstrated in longitudinal analyses The Lifelines database was created using the open source
that higher HairGC levels are associated with future weight MOLGENIS software 9.2.3, built on 2022-04-01 (22).
gain (15). There, we also found first evidence that immune pa­
rameters, that is, leukocyte counts and interleukin-6 (IL-6) Hair Glucocorticoid Measurements
levels, are linked with HairGC levels. This supports evidence We performed glucocorticoid measurements in hair as previously
from experimental models indicating close links between the described, using liquid chromatography–tandem mass spectrom­
stress response and immune system (16-18). It is, however, etry (LC-MS/MS) adapted with minor modifications (6, 23). To
less clear whether chronically elevated systemic glucocorticoid ensure qualitatively valid measurements, we excluded hair sam­
levels, reflected in HairGC, also translate to an increased risk ples weighing less than 7.5 mg, where there had been a problem
in cardiovascular endpoints (19). In the current research, our in sample preparation, sample matrix, and/or chromatography,
aim was to investigate whether long-term glucocorticoid levels or where mass spectrometry yielded no reliable peak, an aberrant
are associated with incident CVD during follow-up. Our sec­ ion-ratio, or co-elution. For both hair cortisol and cortisone, the
ondary aim was to investigate the relation between long-term lower limit of quantification in our assay is currently considered
glucocorticoid levels and cellular immune parameters. 2.5 pg/mg. The minimum signal-to-noise ratio was set at 10.

Methods Cardiometabolic Measurements

Participants All anthropometric measurements were performed by techni­
We analyzed hair samples in a subset of the Lifelines cohort cians trained using standard operating protocols. Height (m)
study. Lifelines, established in 2007, is a multidisciplinary pro­ and weight (kg), measured without shoes using standard scales,
spective population-based cohort study examining in a unique were used to calculate BMI (kg/m2). WC was measured while
3-generation design the health and health-related behaviors of standing upright, using a SECA 201 measurement tape that
167 729 persons living in the North of the Netherlands. It em­ was placed between the lowest rib and the iliac crest around
ploys a broad range of investigative procedures in assessing the the bare stomach.
biomedical, sociodemographic, behavioral, physical, and psy­ At the third assessment (follow-up visit), participants were
chological factors that contribute to the health and disease of asked whether they had developed any new CVD since they
the general population, with a special focus on multi-morbidity filled in the last questionnaire (yes/no). We considered patients
and complex genetics (20, 21). who answered “yes” to this question as cases with incident
In 2014, participants were invited for a second assessment, CVD. Also, participants who reported the presence of cardio­
which took place between 2014 and 2019 and was considered vascular diseases at any questionnaire that was performed
baseline for the current study. At this visit, hair samples were ob­ after the scalp hair collection, and who did not report this at
tained from all participants who gave permission and had at the earlier visits, were considered as cases with incident
least 3 cm of hair, following standard operation procedures. CVD. A detailed overview of the composition of the outcome
The sample was cut as closely to the scalp as possible, at the is provided in Supplementary data 1 (24).
posterior vertex, and subsequently stored in a dark and dry
place until it was analyzed. Prior to the HairGC measurements, Routine Laboratory Measurements
we performed a power calculation under the assumptions According to routine procedure in the Lifelines cohort, blood
that the expected incidence of CVD would be 3% in the highest was collected and analyzed in the clinical chemistry labora­
cortisol quartile, and 1% in the lowest quartile (based on tory of the University Medical Center Groningen (20).
2522 The Journal of Clinical Endocrinology & Metabolism, 2024, Vol. 109, No. 10

Table 1. Baseline characteristics of the study sample, stratified by sex

Level Female Male P Missing (%)

n 4758 1571
a
Hair cortisone levels, pg/mg, median [IQR] 5.19 [4.02, 7.09] 6.72 [5.15, 9.59] <.001 25.9
Hair cortisol levels (%)a < 2.5 pg/mg 1619 (76.4) 461 (70.9) .005 56.2
> 2.5 pg/mg 499 (23.6) 189 (29.1)
Age, years, mean (SD) 53.21 (10.06) 54.29 (10.88) <.001 0.2
BMI (kg/m2), mean (SD) 26.31 (4.62) 26.46 (3.64) .239 0.3
Waist circumference (cm), mean (SD) 87.94 (12.13) 96.10 (10.80) <.001 0.3
Corticosteroid use in past 3 months (%) no 3505 (80.6) 1197 (85.4) <.001 9.3
yes 846 (19.4) 204 (14.6)

Downloaded from https://academic.oup.com/jcem/article/109/10/2520/7631418 by guest on 07 November 2024

Past or current smoking (%) no 3670 (84.7) 1153 (82.7) .087 9.7
yes 664 (15.3) 241 (17.3)
Alcohol use, categorized (%) ≤ 1 drink per day 2509 (58.3) 651 (46.7) <.001 10.1
1-2 drinks per day 659 (15.3) 405 (29.1)
2 drinks per day 137 (3.2) 202 (14.5)
nondrinker 999 (23.2) 136 (9.8)
HbA1c (mmol/mol), mean (SD) 37.25 (4.88) 37.60 (5.33) .017 1.2
Fasting blood glucose (mmol/L), mean (SD) 5.07 (0.93) 5.36 (0.96) <.001 1.7
HDL-cholesterol (mmol/L), mean (SD) 1.71 (0.42) 1.38 (0.34) <.001 1.1
LDL-cholesterol (mmol/L), mean (SD) 3.33 (0.91) 3.45 (0.90) <.001 1.1
Systolic blood pressure (mmHg), mean (SD) 128.18 (17.12) 133.44 (15.67) <.001 0.3
Diastolic blood pressure (mmHg), mean (SD) 72.88 (8.93) 78.06 (9.17) <.001 0.3
Past history of cardiovascular diseases (%) not mentioned 4577 (96.2) 1450 (92.3) <.001 0.0
present 181 (3.8) 121 (7.7)
Past history of diabetes mellitus (%) not mentioned 4594 (96.6) 1501 (95.5) .078 0.0
present 164 (3.4) 70 (4.5)
Leukocytes, 10E9/L 6.09 (1.70) 6.16 (2.15) .210 1.1
Monocytes, 10E9/L 0.48 (0.14) 0.55 (0.16) <.001 1.9
Lymphocytes, 10E9/L 2.03 (0.62) 1.98 (1.30) .027 1.9
Neutophilic granulocytes, 10E9/L 3.29 (1.21) 3.30 (1.11) .676 1.9
Eosinophilic granulocytes, 10E9/L 0.19 (0.13) 0.21 (0.14) <.001 1.9

Parametric data are presented as mean(SD), nonparametric data are presented as median (IQR; interquartile range).
Abbreviations: BMI, body mass index; HDL, high-density lipoprotein; LDL, low-density lipoprotein.
a
Flowchart of hair cortisol and cortisone results is given in Fig. 1A and 1B.

Other Relevant Covariates analyses (> 2.5 pg/mg or < 2.5 pg/mg). Descriptive statistics
By using questionnaires, we obtained self-reported data concern­ of the baseline measurements are reported as median and
ing education level (classified as low/medium/high according to interquartile range (IQR) for nonparametric data and mean
the standardized cutoffs applicable to the Dutch education sys­ ± SD for parametric data. We investigated via linear regres­
tem, Standaard onderwijsindeling); alcohol use based on ques­ sion whether anthropometric data at baseline were associated
tions regarding the number of drinking days per week and the with hair cortisone and cortisol levels.
number of drinks per occasion (subsequently classified as First, we used univariate logistic regression to estimate the
“Non-drinkers,” “0-1 drinks per day,” “1-2 drinks per day,” influence of HairGC levels and classical cardiovascular risk
or “More than 2 drinks per day”); and past or current smoking factors on the outcome incident CVD (present/absent).
(yes/no). Participants were also asked whether they used any Next, we used stepwise logistic regression to estimate the in­
corticosteroid-containing drugs in the last 3 months before tak­ fluence of HairGC levels on incident CVD. First, age was
ing the hair sample (yes/no) and whether they had been diag­ added to the model, including the interaction term
nosed with diabetes on any of the previous visits (yes/no). age*HairGC level (Model A). To investigate sex-differences,
we performed exploratory analysis in males and females sep­
arately, but refrained from further sex-specific analyses due to
Statistical Analysis a lack of statistical power, particularly for males. Also, insuf­
Analyses were performed using R Studio (Version 4.2.2). ficient data were available on menopausal status of females.
HairGC results below
√ the lower limit of quantification were The final analyses were corrected for age, including the
replaced by 2.5/ 2 (25), and because of non-normality they interaction term age*HairGC, sex, corticosteroid use, WC,
were 10-log transformed for all analyses. As the majority of current smoking, systolic blood pressure, high-density lipo­
cortisol data were below the lower limit of quantification, protein (HDL)- and total cholesterol, having a past history
we used cortisol as a dichotomous variable in longitudinal of diabetes mellitus, and having a past history of CVD
The Journal of Clinical Endocrinology & Metabolism, 2024, Vol. 109, No. 10 2523

A Hair corsol analysis, Lifelines n=6341

Reason exclusion n
Less than 7.5mg hair: 889
Other reasons*: 1819
Ion-rao: 220

Further analysis, n=3413

Higher than 2.5pg/mg, n=1330

Reason exclusion n
Signal to noise <10, co-eluon or no peak: 637

Downloaded from https://academic.oup.com/jcem/article/109/10/2520/7631418 by guest on 07 November 2024

Corsol <2.5pg/mg Quantave data:
n=2083 n=693

B Hair corsone analysis, Lifelines n=6341

Reason exclusion n
Less than 7.5mg hair: 889
Other reasons*: 654
Ion-rao: 61

Further analysis, n=4747

Higher than 2.5pg/mg, n=4576

Reason exclusion n
Signal to noise <10, co-eluon or no peak: 46
Corsol <2.5pg/mg Quantave data:
n=171 n=4530

Figure 1. Flowchart of hair cortisol (A) and hair cortisone analysis (B). *Indicates problem with sample preparation, sample matrix, and/or
chromatography.

Figure 2. Heatmap showing the predicted probability of cardiovascular diseases during follow-up, according to the interaction between age at baseline
and hair cortisone levels (pg/mg, 10-log transformed) at baseline. Red color indicates high probability of new cardiovascular diseases (CVD), purple
indicates lower probability.
2524 The Journal of Clinical Endocrinology & Metabolism, 2024, Vol. 109, No. 10

Table 2. Results of the multivariate logistic regression for the and cortisone levels (adjusting for age, sex, BMI, and recent
outcome self-reported incident cardiovascular diseases during the 5 corticosteroid use), and whether they were independent pre­
to 7 years of follow-up, for (a) hair cortisone (as continuous
predictor, including the interaction term hair cortisone*age) and (b)
dictors for incident CVD by adding them to Model B.
hair cortisol (dichotomized as above/below the lower limit of
quantification [2.5pg/mg])
Results
Characteristic ORa 95% CIa P value Hair Glucocorticoids and Anthropometrics at
Hair cortisone levels (total n = 2878, number of events = 214)
Baseline
Hair cortisone level (pg/mg), 83.3 2.36, 2855 .015
Baseline characteristics of the full cohort from whom hair
10-log samples were obtained (n = 6341), are shown in Table 1. In
Age, per 10 years 2.64 1.55, 4.51 <.001
brief, the population was predominantly female (75%), the
mean age was 53 years, and the mean BMI was 26 kg/m2.
Hair cortisone level (pg/mg), 0.52 0.28, 0.96 .038
10-log * Age, per 10 years Hair cortisol levels were below the lower limit of quantifica­

Downloaded from https://academic.oup.com/jcem/article/109/10/2520/7631418 by guest on 07 November 2024

tion (2.5 pg/mg) in the majority of participants in whom valid
Sex, male 1.07 0.75, 1.51 .7
measurements could be performed (n = 2083; 33%). In 693
Waist circumference 1.00 0.99, 1.02 .8
participants (11%), hair cortisol levels were > 2.5 pg/mg. A
Any corticosteroid use in last 1.06 0.72, 1.53 .7 flowchart of hair cortisol results is provided in Fig. 1A. For
3 months, yes
cortisone, 171 (2.7%) participants had cortisone levels below
Current smoking, yes 0.96 0.60, 1.48 .9 the lower limit of quantification. Baseline hair cortisone levels
Systolic blood pressure, per 1.19 1.09, 1.30 <.001 were quantifiable in a total of 4530 participants (71.4%, flow­
10 mmHg increase chart provided in Fig. 1B).
Past history of diabetes 2.13 1.15, 3.79 .012 In individuals with quantifiable hair cortisol levels (n = 693),
mellitus there was a positive association between hair cortisol levels and
HDL-cholesterol, mmol/L 0.80 0.53, 1.19 .3 BMI (r = 0.099, beta = 1.872, P = .0097), and between hair cor­
Total cholesterol, mmol/L 0.92 0.79, 1.08 .3 tisol levels and waist circumference (r = 0.109, beta = 5.92,
Past history of cardiovascular 3.00 1.75, 4.99 <.001 P = .004). For cortisone, we found significant baseline associa­
diseases tions between hair cortisone and BMI (r = 0.075, beta = 1.51,
Hair cortisol levels (total n = 1739, number of events = 140) P < .00001) and between hair cortisone and WC (r = 0.137,
Hair cortisol levels 8.34 .79, 84.5 .075 beta = 7.67, P < .00001).
> 2.5 pg/mg
Age, per 10 years 1.80 1.42, 2.28 <.001 Hair Glucocorticoids and Incident Cardiovascular
Hair cortisol levels 0.69 .46, 1.04 .078 Events During Follow-Up
> 2.5 pg/mg * Age,
per 10 years
A total of 4218 participants had follow-up data, from either
the follow-up visit and/or the COVID questionnaire or 2B
Sex, male 1.08 .69, 1.65 .7
questionnaire (Supplemental data 1 & 2 (24)). A total of
Waist circumference 1.00 .98, 1.01 .7 348 individuals reported one or multiple new cardiovascular
Any corticosteroid use in last 0.79 .48, 1.27 .4 disease manifestations during follow-up (incident CVD).
3 months, yes This was specified by 53 participants as a heart attack
Current smoking, yes 1.00 .55, 1.72 >.9 (15%), 30 participants specified it as “narrowing of the arter­
Systolic blood pressure, 1.01 1.00, 1.03 .010 ies in the legs” (9%), 30 participants as “stroke” (9%), and 47
mmHg reported “heart failure” (14%). The remaining cases were un­
High-density lipoprotein 0.61 .35, 1.02 .063 specified, mainly because these answers were obtained from
cholesterol, mmol/L questionnaires where a specification was not requested.
Total cholesterol, mmol/L 0.98 .80, 1.18 .8 In univariate logistic regression analysis (results provided in
Past history of diabetes 2.07 .94, 4.27 .058 Supplemental data 3 (24)), baseline hair cortisone levels were sig­
mellitus nificantly associated with incident CVD during follow-up
Past history of cardiovascular 3.27 1.64, 6.19 <.001 (OR = 3.08 [95% CI, 1.79-5.20] P < .001) on a continuous scale
diseases as well as in quartiles. Here, hair cortisol levels did not predict in­
cident CVD (P = .73), neither when dichotomized (P = .30).
Results of a multivariate logistic regression model are provided, where incident
cardiovascular diseases in 5-7 years follow-up was used as outcome. Included
covariates were age, sex, waist circumference (per cm increase), self-reported
corticosteroid use in the past 3 months (yes/no), smoking, systolic blood pressure Age-Specific Analysis
(odds ratio presented per 10 mmHg increase) and self-reported past history of In a multivariate logistic regression model (Model A), we investi­
diabetes mellitus (yes/no). P values were significant at P < .05 (presented in bold).
a
Odds ratios (OR) and 95% confidence intervals (95% CI) are presented. gated the association between baseline cortisone levels on the out­
come incident CVD, correcting for age and the interaction hair
cortisone*age. We found that hair cortisone, age, and the inter­
(Model B). Next, we performed separate sensitivity analysis in action between age and cortisone were all significantly associated
participants aged < 60 years and aged ≥ 60 years, also correct­ with the outcome (P = .01, P < .001, and P = .03, respectively),
ing for covariates as described in Model B. indicating that having higher hair cortisone levels increases the
Lastly, we investigated whether cellular immune parame­ probability of incident CVD in an age-dependent manner. A heat­
ters (ie, total leukocytes, neutrophils, lymphocytes, mono­ map showing the interaction between age and hair cortisone level
cytes, and eosinophils) were associated with hair cortisol on the probability of incident CVD is provided in Fig. 2.
The Journal of Clinical Endocrinology & Metabolism, 2024, Vol. 109, No. 10 2525

Downloaded from https://academic.oup.com/jcem/article/109/10/2520/7631418 by guest on 07 November 2024

Figure 3. A, Graphical representation of the predicted probability of developing new cardiovascular diseases (CVD) during follow-up, according to
baseline hair cortisone levels (pg/mg, 10-log transformed), corrected for age (per 10 years increase), sex, waist circumference, corticosteroid use in the
last 3 months, smoking, systolic blood pressure (per 10 mmHg increase), and past medical history of diabetes mellitus, in a logistic regression model
with incident CVD as the outcome variable. Red line indicates predicted probability, with red bands indicating 95% CI. B, Graphical representation of the
predicted probability of developing new cardiovascular diseases (CVD) during follow-up, according to baseline hair cortisone levels (pg/mg, 10-log
transformed), split by age category (< 60 years or ≥ 60 years), including the interaction term age category*hair cortisone. Analyses are corrected for sex,
waist circumference, corticosteroid use in the last 3 months, smoking, systolic blood pressure (per 10 mmHg increase), high-density lipoprotein (HDL)
cholesterol, total cholesterol, past medical history of diabetes mellitus, and past medical history of CVD, in a logistic regression model with incident CVD
as the outcome variable. Red line indicates predicted probability, with red bands indicating 95% CI for the younger individuals. Blue line indicates
predicted probability for the elder individuals, with blue bands indicating 95% CI.

Sex-Specific Analysis of quantification, having high hair cortisol (> 2.5 pg/mg)
In a separate univariate analysis within female participants only was in trend associated with future CVD (P = .072 for
(n = 2441, 165 CVD cases) hair cortisone levels were still signifi­ hair cortisol, P < .001 for age, and P = .075 for the inter­
cantly associated with CVD (P < .001, OR = 3.05 [95% CI, action term cortisol*age; Table 2B). In the subgroup of par­
01.56-5.79]). In males, the association lost significance ticipants with known cortisol data, cortisone was still
(n = 704, 78 CVD cases, P = .2, OR = 1.81 [95% CI, significant in univariate analysis but not in multivariate
0.65-4.91]). Due to lack of power, we refrained from further analyses according to Model A and B (n = 1847, P = .02
sex-specific analyses, but we used “sex” as a covariate in all for univariate analysis; n = 1847 and P = .11 for Model A;
multivariate models. and n = 1695 and P = .2 in Model B). In the subgroup of
participants without past CVD, hair cortisone levels were
Fully Adjusted Models still significantly associated with incident CVD (P < .019
for hair cortisone, P = .057 for hair cortisone*age when
In the fully adjusted model (using Model B, including classical
used in Model B).
cardiovascular risk factors), hair cortisone was still associated
with incident CVD (Table 2A, graphical representation of pre­
dicted probability for incident CVD by hair cortisone levels Sensitivity Analysis in Age Categories
shown in Fig. 3A). Within the subgroup of individuals < 60 years, hair cortisone
Cortisol was not significantly associated with CVD in was strongly and significantly associated with incident
Model B when used on a continuous scale (P = .193 and CVD after adjustment for relevant covariates via Model B
P = .209 for the interaction term cortisol*age). However, (OR = 4.21 [95% CI, 1.91-9.07], P = .005, Table 3). In elder­
when dichotomized in those below and above the lower limit ly individuals (≥ 60 years), hair cortisone levels were not
2526 The Journal of Clinical Endocrinology & Metabolism, 2024, Vol. 109, No. 10

Table 3. Results of the sensitivity analyses in subgroups based on (r = 0.025, P = .19). After adjustment for age, sex, BMI, and
age (categorized as < 60 years and ≥ 60 years) corticosteroid use, only lymphocytes were significantly associ­
ated with hair cortisol (P = .04, Table 5).
Predicting incident cardiovascular diseases by baseline hair cortisone,
split by age category
Immune Parameters at Follow-Up
Characteristic OR 95% CI P value When total leukocytes, neutrophils, and monocytes were
< 60 years (total n = 2229, number of events = 125) added to Model B with cortisone as predictor, only monocytes
were a significant independent risk factor for incident CVD
Hair cortisone level (pg/mg), 10-log 4.21 1.91, 9.07 <.001
(OR = 3.37 [95% CI, 1.19-9.36], P = .021), leukocytes and
Sex, male 1.14 .71, 1.79 .6
neutrophils were not significant (Supplemental data 4A &
Waist circumference 1.01 .99, 1.02 .5 4B (24)). In the same model with dichotomized cortisol,
Any corticosteroid use in last 1.13 .68, 1.81 .6 none of the immune parameters were significant predictors
3 months, yes of incident CVD.

Downloaded from https://academic.oup.com/jcem/article/109/10/2520/7631418 by guest on 07 November 2024

Current smoking, yes 0.92 .52, 1.52 .7
Systolic blood pressure, per 1.24 1.10, 1.40 <.001
10 mmHg increase Discussion
High-density lipoprotein 1.12 .66, 1.88 .7 We prospectively evaluated the association between baseline
cholesterol, mmol/L hair glucocorticoid levels and incident cardiovascular diseases
Total cholesterol, mmol/L 0.94 .76, 1.15 .5 during 5 to 7 years of follow-up in a large population-based
Past history of diabetes mellitus 3.31 1.37, 7.24 .004 cohort study. Hair cortisone levels showed strong associations
Past history of cardiovascular 3.03 1.32, 6.25 .005 with incident CVD, even after adjustment for known predic­
diseases tors of CVD. This effect was driven in particular by the young­
≥ 60 years (total n = 649, number of events = 89) er individuals, as hair cortisone was highly significantly and
Hair cortisone level (pg/mg), 10-log 0.67 .23, 1.81 .4 relevantly associated with incident CVD in the subcohort of
Sex, male 1.07 .61, 1.85 .8
individuals < 60 years of age. The magnitude of the associ­
ation was comparable to other known risk factors for CVD,
Waist circumference 1.00 .97, 1.02 .7
such as a history of diabetes mellitus and higher systolic blood
Any corticosteroid use in last 0.93 .51, 1.63 .8 pressure. Hair cortisol levels were not significantly, but in
3 months, yes
trend, associated with incident CVD, perhaps due to lack of
Current smoking, yes 0.88 .34, 1.99 .8 statistical power. Hair cortisone and hair cortisol were both
Systolic blood pressure, per 1.20 1.05, 1.37 .006 associated with multiple cellular immune parameters in cross-
10 mmHg increase sectional analyses; in longitudinal analysis, only monocyte
High-density lipoprotein 0.56 .29, 1.04 .075 count was an independent predictor of incident CVD.
cholesterol, mmol/L Although it has been established that higher hair gluco­
Total cholesterol, mmol/L 0.96 .76, 1.22 .7 corticoid levels, reflecting chronic systemic exposure, are asso­
Past history of diabetes mellitus 1.58 .64, 3.62 .3 ciated on a cross-sectional level with cardiometabolic
Past history of cardiovascular 3.33 1.58, 6.80 .001 parameters, such as BMI, WC, metabolic syndrome, and
diseases CVD (9, 11, 14), the relevance of these findings was unsure
as longitudinal studies were lacking (4, 19). Our study is
Results of a multivariate logistic regression are provided, where hair cortisone now the first to fill this gap, providing evidence on the relevant
levels (pg/mg, 10 log transformed), sex, waist circumference (per cm increase),
self-reported corticosteroid use in the past 3 months (yes/no), smoking, systolic clinical outcome of incident CVD. This endorses the concept
blood pressure (odds ratio presented per 10 mmHg increase) and self-reported that higher exposure to endogenous glucocorticoid levels,
history of diabetes mellitus (yes/no) were used as predictors. Odds ratios (OR) and
95% CI are presented. P values were significant at P < .05 (presented in bold).
also in individuals without clear clinical endogenous hyper­
cortisolism, may not only represent a risk factor for weight
gain (15), but also for clinically relevant endpoints, such as in­
cident CVD. For morning plasma cortisol, it has already been
significantly associated with CVD after adjustment for rele­
prospectively demonstrated that this was associated with
vant covariates (Table 3, visual representation in Fig. 3B).
higher risk for CVD (26). However, those odds ratios were
much lower than those observed in the current study, indicat­
Immune Parameters at Baseline ing that perhaps a measure of cumulative exposure, which is
Baseline hair cortisone levels were significantly associated believed to be assessed by hair glucocorticoid analyses, is of
with baseline total leukocytes, neutrophils, and monocytes higher interest than morning (time point) measurements of
(r = 0.065, r = 0.076, and r = 0.11 respectively; P = .0001, cortisol. Importantly, it should be noted that despite the asso­
P = 3E-7, and P = 5.3E-14, respectively), and in trend to lym­ ciations that were found, definite evidence for causation is still
phocytes (P = .078) and eosinophils (P = .067). After adjust­ lacking.
ment for age, sex, BMI, and corticosteroid use, hair Interestingly, the association seems to apply particularly to
cortisone was significantly associated with leukocytes, neutro­ incident CVD at a younger age (≤ 60 years at baseline). For
phils, monocytes, and lymphocytes (all P < .001, Table 4). traditional cardiovascular risk factors, relative risk contribu­
Similarly, hair cortisol levels were associated with baseline tions are known to decrease with age (27, 28), suggesting a
leukocytes, neutrophils, lymphocytes, and eosinophils multifactorial etiology, whereas age per se is strongly associ­
(r = 0.045, r = 0.042, r = 0.046, and r = 0.038, respectively; ated with increased CVD risk. A high a priori risk in older in­
P = .018, P = .03, P = .017, and P = .047), but not monocytes dividuals may thus attenuate the relative contribution of other
Table 4. Cross-sectional linear regression models investigating the associations between hair cortisone levels and immune parameters, adjusting for age, sex, recent corticosteroid use, and body
mass index (BMI, kg/m2)

Total leukocytes (10 e-9/L) Neutrophil granulocytes Monocytes (10 e-9/L) Lymphocytes (10 e-9/L) Eosinophil granulocytes
(10 e-9/L) (10 e-9/L)

Predictors Estimates P Estimates P Estimates P Estimates P Estimates P

Hair cortisone levels (pg/mg), 10-log 0.48 (0.24-0.71) <.001 0.44 (0.28-0.60) <.001 0.04 (0.02-0.06) <.001 0.09 (0.00-0.17) .046 0.00 (−0.02-0.02) .781
Age, per 10 years −0.15 (−0.20-−0.10) <.001 −0.14 (−0.18-−0.11) <.001 0.01 (0.00-0.01) <.001 −0.01 (−0.03-0.01) .363 0.00 (−0.00-0.01 .483
Sex, male −0.01 (−0.14-0.11) .821 −0.05 (−0.14-0.04) .260 0.05 (0.04-0.07) <.001 −0.07 (−0.12-−0.02) .003 0.03 (0.02-0.03) <.001
Recent corticosteroid use, yes 0.27 (0.14-0.40) <.001 0.19 (0.09-0.28) <.001 0.02 (0.01-0.03) <.001 0.04 (−0.01-0.09) .127 0.03 (0.02-0.04) <.001
BMI (kg/m2) 0.07 (0.06-0.08) <.001 0.05 (0.04-0.06) <.001 0.00 (0.00-0.00) <.001 0.02 (0.01-0.02) <.001 0.00 (−0.00-0.00) .064
Observations 4243 4210 4210 4210 4210

Results of a multivariate linear regression model where hair cortisone levels (pg/mg), age, sex, recent corticosteroid use and body mass index (BMI) are included. Linear regression estimates and 95% confidence intervals are provided.
P values were significant at P < .05 (presented in bold).
The Journal of Clinical Endocrinology & Metabolism, 2024, Vol. 109, No. 10

Table 5. Cross-sectional linear regression models investigating the associations between hair cortisol levels and immune parameters, adjusting for age, sex, recent corticosteroid use and body mass
index (BMI, kg/m2)

Predictors Total Leukocytes (10 e-9/L) Neutrophil Granulocytes Monocyte (10 e-9/L) Lymphocytes (10 e-9/L) Eosinophil Granulocytes
(10 e-9/L) (10 e-9/L)

Estimates P Estimates P Estimates P Estimates P Estimates P

Hair cortisol levels (pg/mg), 10-log 0.36 (−0.04-0.76) .075 0.22 (−0.05-0.50) .112 0.01 (−0.02-0.04) .471 0.15 (0.01-0.30) .036 0.02 (−0.01-0.05) .158
Age, per 10 years −0.18 (−0.24-−0.11) <.001 −0.15 (−0.19-−0.10) <.001 0.01 (0.00-0.01) .001 −0.02 (−0.05-0.00) .058 0.00 (−0.00-0.01) .849
Sex, male 0.06 (−0.10-0.22) .499 0.02 (−0.09-0.13) .659 0.06 (0.05-0.07) <.001 −0.08 (−0.14-−0.03) .004 0.03 (0.01-0.04) <.001
Recent corticosteroid use, yes 0.29 (0.12-0.47) .001 0.21 (0.09-0.33) .001 0.02 (0.01-0.03) .005 0.03 (−0.03-0.10) .333 0.03 (0.02-0.04) <.001
2
BMI (kg/m ) 0.07 (0.06-0.09) <.001 0.05 (0.04-0.06) <.001 0.00 (0.00-0.00) <.001 0.02 (0.01-0.02) <.001 0.00 (0.00-0.00) .030
Observations 2514 2499 2499 2499 2499

Results of a multivariate linear regression model where hair cortisol levels (pg/mg), age, sex, recent corticosteroid use and body mass index (BMI) are included. Linear regression estimates and 95% confidence intervals are provided.
P values were significant at P < .05 (presented in bold).
2527

Downloaded from https://academic.oup.com/jcem/article/109/10/2520/7631418 by guest on 07 November 2024

2528 The Journal of Clinical Endocrinology & Metabolism, 2024, Vol. 109, No. 10

traditional risk factors. Of note, it is well established that found), or following psychological or physical stressors (eg,
mean daily glucocorticoid levels and hair glucocorticoid levels chronic pain, sleep deficit, endocrine disruptors) (4), and/or
generally increase with aging (10, 29). In this regard, it can be altered cortisol metabolism eg, in the liver or visceral adipose
hypothesized that higher glucocorticoid levels in younger indi­ tissue (35).
viduals may actually reflect underlying accelerated biological In Cushing disease, treating the hypercortisolism greatly re­
aging, where these individuals resemble older individuals duces cardiovascular risk (36). However, until now, in people
with regard to CVD risk. This may be mediated by an acceler­ without Cushing disease there were only modest effects of drugs
ation of the age-related increase in activation of the immune targeting glucocorticoid action (37-39). It can be speculated that
system (“inflammaging”), which is highly relevant in the de­ HairGC measurements can be used to identify individuals at
velopment of atherosclerosis (30). Indeed, in the current risk who may benefit from pharmacological or non-
work we found significant cross-sectional associations be­ pharmacological glucocorticoid-lowering interventions or pre­
tween various cellular immune parameters and glucocorticoid ventive strategies, ultimately decreasing cardiovascular risk.
levels. In the multivariate-adjusted longitudinal analyses, In summary, we demonstrate that long-term glucocorticoid

Downloaded from https://academic.oup.com/jcem/article/109/10/2520/7631418 by guest on 07 November 2024

monocyte counts were the only cellular immune parameter levels, specifically cortisone, are associated with an increased
that was a significant risk factor for CVD, independent of risk of incident CVD in younger individuals, even after adjust­
glucocorticoid levels. This indicates tight links between ment for classical risk factors and with comparable effect
HPA-axis activation and immune activation, post aut propter sizes. Our results thus highlight the potential role of gluco­
translating to a higher risk for incident CVD. Also, we con­ corticoid action in the development of CVD and reveal hair
firmed earlier findings that monocytes are most closely linked glucocorticoid levels as promising biomarkers that may be
to the development of CVD, which is in line with their patho­ used to identify individuals with increased cardiovascular risk.
genic role (31).
Of note, the current work supports previous findings regard­
ing metabolic syndrome and waist circumference, suggesting Acknowledgments
that among HairGC, the most pronounced associations with We would like to acknowledge Cagla Koyuncu and Yolanda
cardiometabolic health are seen for the biologically inactive de Rijke for their contribution in analyzing the LC-MS data.
cortisone, instead of cortisol (9, 14). It has been hypothesized The authors wish to acknowledge the services of the
that hair cortisone levels reflect the “reservoir” of circulating Lifelines Cohort Study, the contributing research centers de­
glucocorticoids, that can be activated by the enzyme livering data to Lifelines, and all the study participants.
11-beta-hydroxysteroid dehydrogenase type 1 to biologically
active cortisol (4, 9). However, it should be noted that we
did not meet the anticipated statistical power that was required Funding
to detect a significant difference for cortisol. Thus, our cortisol Elisabeth Foundation; Netherlands Organization of Scientific
results should be interpreted with high caution, and the high Research NWO, Grant/Award Number: 91716453.
number of cortisol measurements that had to be excluded for The Lifelines initiative has been made possible by subsidy
technical reasons is a limitation of the current study. from the Dutch Ministry of Health, Welfare and Sport, the
Another important limitation of our work is that the end­ Dutch Ministry of Economic Affairs, the University Medical
point CVD has been based on self-reporting of a diagnosis, Center Groningen (UMCG), Groningen University and the
with several (structured) questionnaires that, however, used Provinces in the North of the Netherlands (Drenthe,
slightly different phrasings. It is known that self-reporting of Friesland, Groningen).
incident diseases has a high specificity, but sensitivity tends
to be lower (32). Furthermore, there is a relatively large group
of individuals from whom we did not have any follow-up Disclosures
(33%). This subgroup was not different regarding age and
There are no disclosures.
sex but had a significantly higher prevalence of all risk factors,
including hair cortisone levels. Perhaps part of the attrition is
due to increased morbidity or even (cardiovascular) mortality, Data Availability
which is, however, unknown.
Furthermore, inherent in the population of the northern Data may be obtained from a third party and are not publicly
provinces of the Netherlands, the Lifelines cohort includes available. Researchers can apply to use the Lifelines data used
relatively few ethnic minorities, hampering generalizability in this study. More information about how to request Lifelines
to other populations. Also, in line with other population stud­ data and the conditions of use can be found on their website
ies using HairGC measurements, we have an overrepresenta­ (https://www.lifelines.nl/researcher/how-to-apply).
tion of females (15, 33, 34), limiting further sex-specific
analyses due to lack of power in males.
References
Next, we did not have any data on medication use (eg, sta­
tins, antihypertensives, and glucose-lowering drugs) which 1. Lengton R, Iyer AM, van der Valk ES, et al. Variation in gluco­
may have induced a bias. corticoid sensitivity and the relation with obesity. Obes Rev.
2022;23(3):e13401.
In order to determine the future clinical consequences of our
2. Ramamoorthy S, Cidlowski JA. Corticosteroids: mechanisms of ac­
findings, we need to better understand why some individuals tion in health and disease. Rheum Dis Clin North Am. 2016; 42(1):
have increased systemic glucocorticoid exposure. Possible ex­ 15-31, vii.
planations include increased cortisol production, for example, 3. Nieman LK, Biller BM, Findling JW, et al. The diagnosis of
in response to low-grade inflammation (as suggested by the Cushing's syndrome: an endocrine society clinical practice guide­
consistent associations with immune parameters that we line. J Clin Endocrinol Metab. 2008;93(5):1526-1540.
The Journal of Clinical Endocrinology & Metabolism, 2024, Vol. 109, No. 10 2529

4. van der Valk ES, Savas M, van Rossum EFC. Stress and obesity: are 23. Odijk R, van den Berg S, Dorst K, de Rijke Y. An optimized LC-MS/
there more susceptible individuals? Curr Obes Rep. 2018;7(2): MS method for cortisol and cortisone in human hair. presented at:
193-203. Psychoneuroendocrinology 2019; Session 107(S) (2019) 1–81.
5. Rodriguez AC I, Epel ES, White ML, Standen EC, Seckl JR, Tomiyama 24. van der Valk ES, Mohseni M, Iyer AM, et al. Data from:
AJ. Hypothalamic-pituitary-adrenal axis dysregulation and cortisol Supplementary data to article “Long-term glucocorticoid exposure
activity in obesity: a systematic review. Psychoneuroendocrinology. and incident cardiovascular diseases—the Lifelines cohort”. Figshare.
Dec 2015;62:301-318. 2023. doi:10.6084/m9.figshare.24595611
6. Noppe G, Rijke YB, Dorst K, Akker ELT. LC-MS/MS-based meth­ 25. Harel O, Perkins N, Schisterman EF. The use of multiple imput­
od for long-term steroid profiling in human scalp hair. Clin ation for data subject to limits of detection. Sri Lankan J Appl
Endocrinol (Oxf). 2015;83(2):162-166. Stat. 2014;5(4):227-246.
7. Short SJ, Stalder T, Marceau K, et al. Correspondence between hair 26. Crawford AA, Soderberg S, Kirschbaum C, et al. Morning plasma
cortisol concentrations and 30-day integrated daily salivary and cortisol as a cardiovascular risk factor: findings from prospective
weekly urinary cortisol measures. Psychoneuroendocrinology. cohort and Mendelian randomization studies. Eur J Endocrinol.
2016;71:12-18. 2019;181(4):429-438.
8. Mirzaian M, van Zundert SKM, Schilleman WF, et al. 27. van Bussel EF, Hoevenaar-Blom MP, Poortvliet RKE, et al.

Downloaded from https://academic.oup.com/jcem/article/109/10/2520/7631418 by guest on 07 November 2024

Determination of cortisone and cortisol in human scalp hair using Predictive value of traditional risk factors for cardiovascular
an improved LC-MS/MS-based method. Clin Chem Lab Med. disease in older people: a systematic review. Prev Med. 2020;132:
2023;62(1):118-127. doi:10.1515/cclm-2023-0341 105986.
9. van der Valk E, Abawi O, Mohseni M, et al. Cross-sectional rela­ 28. Lind L, Sundstrom J, Arnlov J, Lampa E. Impact of aging on the
tion of long-term glucocorticoids in hair with anthropometric strength of cardiovascular risk factors: a longitudinal study over
measurements and their possible determinants: a systematic review 40 years. J Am Heart Assoc. 2018; 7(1):e007061.
and meta-analysis. Obes Rev. 2022;23(3):e13376. 29. Yiallouris A, Tsioutis C, Agapidaki E, et al. Adrenal aging and its
10. Stalder T, Steudte-Schmiedgen S, Alexander N, et al. Stress-related implications on stress responsiveness in humans. Front
and basic determinants of hair cortisol in humans: a meta-analysis. Endocrinol (Lausanne). 2019;10:54.
Psychoneuroendocrinology. 2017;77:261-274. 30. Roy P, Orecchioni M, Ley K. How the immune system shapes ath­
11. Manenschijn L, Schaap L, Van Schoor NM, et al. High long-term erosclerosis: roles of innate and adaptive immunity. Nat Rev
cortisol levels, measured in scalp hair, are associated with a history Immunol. 2022;22(4):251-265.
of cardiovascular disease. J Clin Endocrinol Metab. 2013;98(5): 31. Williams H, Mack CD, Li SCH, Fletcher JP, Medbury HJ. Nature
2078-2083. versus number: monocytes in cardiovascular disease. Int J Mol
12. Pereg D, Gow R, Mosseri M, et al. Hair cortisol and the risk for acute Sci. 2021;22(17):9119.
myocardial infarction in adult men. Article. Stress. 2011;14(1):73-81. 32. Oksanen T, Kivimaki M, Pentti J, Virtanen M, Klaukka T, Vahtera
13. Faresjo T, Stromberg S, Jones M, et al. Elevated levels of cortisol in J. Self-report as an indicator of incident disease. Ann Epidemiol.
hair precede acute myocardial infarction. Sci Rep. Dec. 31 2010;20(7):547-554.
2020;10(1):22456. 33. Jackson SE, Kirschbaum C, Steptoe A. Hair cortisol and adiposity
14. Stalder T, Kirschbaum C, Alexander N, et al. Cortisol in hair and in a population-based sample of 2,527 men and women aged 54
the metabolic syndrome. J Clin Endocrinol Metab. 2013;98(6): to 87 years. Obesity (Silver Spring). 2017;25(3):539-544.
2573-2580. 34. Feeney JC, O'Halloran AM, Kenny RA. The association between
15. van der Valk ES, van der Voorn B, Iyer AM, et al. Hair cortisol, hair cortisol, hair cortisone, and cognitive function in a population-
obesity and the immune system: results from a 3 year longitudinal based cohort of older adults: results from the Irish longitudinal
study. Psychoneuroendocrinology. 2021;134:105422. study on ageing. J Gerontol A Biol Sci Med Sci. 2020;75(2):
16. McCann SM, Karanth S, Kamat A, et al. Induction by cytokines 257-265.
of the pattern of pituitary hormone secretion in infection. 35. Papanastasiou L, Fountoulakis S, Vatalas IA. Adrenal disorders and
Neuroimmunomodulation. 1994;1(1):2-13. non-alcoholic fatty liver disease. Minerva Endocrinol. 2017;42(2):
17. McCann SM, Kimura M, Karanth S, Yu WH, Mastronardi CA, 151-163.
Rettori V. The mechanism of action of cytokines to control the re­ 36. Coulden A, Hamblin R, Wass J, Karavitaki N. Cardiovascular
lease of hypothalamic and pituitary hormones in infection. Ann N Y health and mortality in Cushing's disease. Pituitary. 2022;25(5):
Acad Sci. 2000;917:4-18. 750-753.
18. van der Meer MJ, Sweep CG, Rijnkels CE, et al. Acute stimulation 37. Rosenstock J, Banarer S, Fonseca VA, et al. The 11-beta-hydroxyste­
of the hypothalamic-pituitary-adrenal axis by IL-1 beta, TNF alpha roid dehydrogenase type 1 inhibitor INCB13739 improves hypergly­
and IL-6: a dose response study. J Endocrinol Invest. 1996;19(3): cemia in patients with type 2 diabetes inadequately controlled by
175-182. metformin monotherapy. Diabetes Care. 2010;33(7):1516-1522.
19. Iob E, Steptoe A. Cardiovascular disease and hair cortisol: a novel 38. Bianzano S, Nordaby M, Plum-Morschel L, Peil B, Heise T. Safety,
biomarker of chronic stress. Curr Cardiol Rep. 2019;21(10):116. tolerability, pharmacodynamics and pharmacokinetics following
20. Sijtsma A, Rienks J, van der Harst P, Navis G, Rosmalen JGM, once-daily doses of BI 187004, an inhibitor of 11 beta-
Dotinga A. Cohort profile update: lifelines, a three-generation co­ hydroxysteroid dehydrogenase-1, over 28 days in patients with
hort study and biobank. Int J Epidemiol. 2022;51(5):e295-e302. type 2 diabetes mellitus and overweight or obesity. Diabetes Obes
21. Scholtens S, Smidt N, Swertz MA, et al. Cohort profile: lifeLines, a Metab. 2023;25(3):832-843.
three-generation cohort study and biobank. Int J Epidemiol. 39. Yadav Y, Dunagan K, Khot R, et al. Inhibition of
2015;44(4):1172-1180. 11beta-Hydroxysteroid dehydrogenase-1 with AZD4017 in pa­
22. van der Velde KJ, Imhann F, Charbon B, et al. MOLGENIS research: tients with nonalcoholic steatohepatitis or nonalcoholic fatty liver
advanced bioinformatics data software for non-bioinformaticians. disease: a randomized, double-blind, placebo-controlled, phase II
Bioinformatics. 2019;35(6):1076-1078. study. Diabetes Obes Metab. 2022;24(5):881-890.