OUP user menu

11β-Hydroxysteroid dehydrogenase type 2 activity is associated with left ventricular mass in essential hypertension

Nicola Glorioso, Fabiana Filigheddu, Paolo Pinna Parpaglia, Aldo Soro, Chiara Troffa, Giuseppe Argiolas, Paolo Mulatero
DOI: http://dx.doi.org/10.1093/eurheartj/ehi070 498-504 First published online: 15 December 2004


Aims Left ventricular mass (LVM) is under the control of aldosterone and angiotensin II in experimental hypertension, but the effect of aldosterone on LVM is controversial in essential hypertension (EH). Some EH patients show a mild impairment of 11β-hydroxysteroid dehydrogenase type 2 (11β-HSD2) activity without clinical features of the syndrome of apparent mineralocorticoid excess, where the incomplete cortisol-to-cortisone conversion leads to glucocorticoid-mediated mineralocorticoid effects. The mineralocorticoid receptor and 11β-HSD2 are co-expressed in human heart. We investigated whether LVM may be regulated by glucocorticoids in EH patients.

Methods and results The ratio between 24 h urinary tetrahydro derivatives of cortisol and cortisone (THFs/THE), plasma renin activity, 24 h urinary aldosterone, blood pressure, and LVM indexed for height2.7 (LVMh2.7) were analysed in 493 never-treated hypertensives and 98 normotensives. THFs/THE was associated with LVMh2.7 in hypertensives and normotensives (r=0.32, P<0.001, and r=0.17, P=0.04, respectively) and persisted after adjusting for confounders (multiple regression analysis). Body mass index, sex, recumbent plasma renin activity, and THFs/THE accounted for 26.1% of LVMh2.7 variation. Urinary aldosterone was not correlated with LVMh2.7.

Conclusion We suggest that glucocorticoids may take part in the regulation of LVM in EH patients as a function of 11β-HSD2 activity, and contribute to the target organ damage associated with essential hypertension.

  • Cardiac mass
  • 11β-Hydroxysteroid dehydrogenase
  • Mineralocorticoids
  • Glucocorticoids
  • Blood pressure


Mineralocorticoids and glucocorticoids exhibit the same affinity for the mineralocorticoid receptor (MR) in vitro;1 in vivo, only mineralocorticoids bind to the MR in aldosterone-selective tissues due to the presence of 11β-hydroxysteroid dehydrogenase type 2 (11β-HSD2).2 11β-HSD2 protects the MR, inactivating cortisol to cortisone, whilst the ubiquitous 11β-HSD type 1 isozyme regulates glucocorticoid activity.3 11β-HSD2 activity is severely impaired in the autosomal recessive syndrome of apparent mineralocorticoid excess (SAME), where cortisol binds to the MR and acts as a mineralocorticoid.46 Thus, a defective 11β-HSD2 activity may result in glucocorticoid-dependent, MR-mediated effects. A mild defect of 11β-HSD2 activity is present in ∼20% of patients with essential hypertension (EH) in the absence of clinical features of SAME7,8 and is also associated with salt sensitivity.9,10

Left ventricular mass (LVM) is increased in 20–25% of EH patients, but is only in part related to blood pressure (BP) values, and predicts morbid events.1113 In hypertensive heart disease, both cardiac myocyte growth and fibrosis develop in the overloaded left ventricle.14 While cardiac myocyte growth seems to be related to ventricular loading, in experimental hypertension the fibrous process is likely to be under the control of the renin–angiotensin–aldosterone system and, in particular, of aldosterone and angiotensin II.15 In EH, the evidence supporting a role for aldosterone in LVM regulation is less strong,16 while in primary aldosteronism the literature is conflicting.17,18

It can be hypothesized that ligands other than aldosterone, such as cortisol, activate the cardiac MR in essential hypertensives with reduced 11β-HSD2 activity, thus exerting mineralocorticoid effects on cardiac mass. The co-existence of both the MR and the 11β-HSD2 in human heart strongly supports this hypothesis.19,20

The aim of the present study was to evaluate the hypothesis that glucocorticoids may take part in the regulation of the cardiac mass in relation to 11β-HSD2 activity in EH.



Four hundred and ninety-three never-treated EH patients and 98 normotensive subjects were studied at the Hypertension and Cardiovascular Prevention Center, University of Sassari, Italy, after written informed consent and protocol approval by the local Ethics Committee. Patients and subjects were from Sardinia, were unrelated, white, and of both genders.

Inclusion criteria for the hypertensive cohort were: (i) sitting systolic/diastolic BP>140/90 mmHg (average of three measurements: see below) at first visit as well as over the 8-week observational period; (ii) no previous antihypertensive therapy: the patients had either a recent finding of high BP or were hypertensives left untreated but monitored by GPs over the years; and (iii) hypertension onset before the age of 60 years to exclude late-onset hypertension.

Exclusion criteria for the hypertensive cohort were: (i) symptomatic hypertension requiring immediate treatment, according to international guidelines;21 (ii) clinical, laboratory, or instrumental evidence of major co-morbidities such as diabetes, cardiac abnormalities, kidney, or liver disease; (iii) secondary hypertension; and (iv) inadequate cardiac sonographic readings.

Inclusion criteria for the normotensive cohort were: (i) sitting BP ≤130/85 mmHg (average of three measurements: see below) at first visit as well as over the 8-week observational period; (ii) age range and sex distribution as for the hypertensive cohort; and (iii) absence of family history of hypertension as well as of cardiovascular diseases.

Exclusion criteria for the normotensive cohort were: (i) evidence of chronic systemic diseases; (ii) chronic use of any drug interacting with the cardiovascular system.

Age range and sex distribution of the normotensive cohort was similar to that of the hypertensive cohort. Normotensives were recruited from and live in the same geographical areas as hypertensive patients. They were either unrelated members of our patients' families (i.e. husbands or wives) or normotensive subjects approaching our facilities in the frame of the prevention programmes our centre normally carries out.

After the first visit, an 8-week observational period was started for two reasons. First, to carry out a complete clinical, laboratory, and instrumental diagnostic work-up to rule out the most frequent forms of secondary hypertension before including the patients in the study cohort. Secondly, to maximize the chance of diagnosing a true hypertensive state. As reported in the literature, the probability of overestimating the BP values is up to 40% at the first visit.22 During this period, 24-h urinary Na+ (checked by reproducibility of creatinine clearance), seated BP (average of three measurements at 2 min intervals, by the same nurse, using a validated automatic electronic device), and body mass index (BMI) were measured every 2 weeks under patients' usual diets. At the end of the observational period, LVM was evaluated by echocardiography (see below), venous blood and 24-h urine were collected for creatinine, plasma renin activity (PRA, after 1 h recumbence and 1 h standing),23 and aldosterone were measured. The ratio between the urinary tetrahydroderivatives of cortisol and cortisone (THFs/THE) as an index of 11β-HSD2 activity was also calculated. At the same time, a subgroup of 167 hypertensives and 33 normotensives, randomly selected from the starting cohorts, had their 24-h ambulatory BP monitored (Spacelabs 90207). Office BPs, 24-h urinary Na+, heart rate, and BMI at the end of the observational period were taken as reference values. Patients presenting with severe hypertension at the first visit were controlled the day after: when BP was confirmed to be ≥180/110 mmHg (n=147, 29.8%) the evaluation of the patient was completed within 1 week to start the pharmacological treatment as soon as possible according to international guidelines.21

Personnel handling patients and subjects were unaware of laboratory data.


All exams were performed with a phased-array 2.5–3.5 MHz transducer (Sonolayer SSH-160A Toshiba, Japan) by the same sonographer. Left ventricular (LV) dimensions were measured by M-mode according to the American Society of Echocardiography.24 The average of at least four consecutive measurements at the end-diastolic phase (R-wave peak of ECG trace) was considered to determine interventricular septal thickness (IVST), left ventricular minor axis (LVD), and posterior wall thickness (PWT). LVM was calculated by the cube formula25 and indexed for height2.7 (LVMh2.7),26 and body surface area (LVMBSA).27 LV hypertrophy (LVH) was defined as LVMh2.7>51 g/m2 in both sexes. Relative wall thickness (RWT) was expressed by the ratio IVST+PWT/LVD, with 0.45 as the cut-off between concentric and eccentric patterns.27,28

Cortisol-to-cortisone conversion rate

Urinary tetrahydrocortisol (THF), allo-tetrahydrocortisol (alloTHF), and tetrahydrocortisone (THE) were measured as described,8 and the THFs/THE was used as an index of 11β-HSD2 activity.29

Statistical analysis

Only age, 24-h urinary aldosterone, and LVD were normally distributed (Shapiro–Wilk). The other variables, with the exception of hypertension duration, diastolic BP, heart rate, urinary Na+ and K+, LVMBSA, and PRA, were normalized by reciprocal, reciprocal square, reciprocal square root, logarithmic, or square root transformation. Continuous variables were tested by the two-sided Student's t-test or Wilcoxon rank-sum test and Pearson or Spearman correlation on the basis of their distribution. To test whether the association between LVMh2.7 and THFs/THE persisted after adjustment for confounders, multiple regression analysis was run including age, sex, BMI, systolic and diastolic BP, and recumbent PRA as independent variables.

We then ran subsequent models including THFs/THE as the first independent variable and added the remaining variables one at a time, based on a clinical and scientific background.12,30,31 Starting from the variables which had given rise to the greatest P-values in the multiple regression analysis, only those giving a statistically significant increase in R2 were kept in the model. In the multiple regression analysis LVMh2.7 was preferred to LVMBSA as the dependent variable because it correlates better with fat-free mass;32 moreover, the impossibility of normalizing LVMBSA made this variable inappropriate for this test. The assumptions of homoscedasticity and normality of LVMh2.7 and residuals were not violated. When THFs/THE was considered as a discrete trait, a cut-off of 1.17 (50th percentile) was chosen.

With the current sample sizes the power was >0.90 for both the hypertensive and normotensive cohorts and >0.80 for the ambulatory blood pressure monitoring (ABPM) substudies, according to Dupont and Plummer.33

Tests of significance were two-sided, and P<0.05 was assumed as significant. Intercooled Stata 7.0 and PS-Power and sample size calculations 2.1.31 for Windows were used.


Table 1 shows the characteristics of the cohorts. The statistical differences between hypertensives and normotensives with regard to recumbent PRA, LV measurements, and THFs/THE are consistent with the literature.8,30,34

View this table:
Table 1

Characteristics of the hypertensive and normotensive cohorts

n 49398
Age, yearsa47±1047±9
Sex, M/F294/19958/40
Hypertension duration, yearsb2 (1 to 6)
BMI, kg/m2c26.8 (26.3–27.0)27.0 (26.5–27.3)
SBP, mmHgc156 (155–157)120 (119–122)**
DBP, mmHgb104 (100 to 110)77 (73 to 83)**
Heart rate, b.p.m.b76 (68 to 83)74 (70 to 80)
UNa, mmol/24 hb146 (115 to 177)141 (117 to 180)
UK, mmol/24 hb48 (40 to 60)49 (42 to 60)
rPRA, AI ng/mL/hb1.07 (0.8 to 1.4)1.4 (1.08 to 1.7)*
uPRA, AI ng/mL/hb2.1 (1.7 to 2.5)2.2 (1.7 to 2.6)
UAldosterone, µg/24 ha13.1±7.813.5±7.3
THFs/THEc1.20 (1.16–1.23)0.98 (0.93–1.02)**
THF, µg/24 hc1041 (967–1120)1071 (978–1113)
a-THF, µg/24 hc1108 (1022–1120)824 (754–864)*
THE, µg/24 hc1998 (1863–2165)2292 (2176–2419)*
LVMBSA, g/m2b97 (84 to 109)84 (72 to 99)**
LVMh2.7, g/m2.7c44.8 (44.1–45.5)38.3 (37.5–39.2)**
IVST, mmc10.3 (10.1–10.5)9.7 (9.4–10.0)**
PWT, mmc9.0 (8.9–9.2)8.7 (8.6–8.8)*
LVD, mma49±4.147±4.0**

UNa, UK, UAldosterone, 24-h urinary sodium, potassium, and aldosterone; uPRA, rPRA, upright and recumbent plasma renin activity; THFs/THE, 24-h urinary THF+alloTHF/THE; THF, tetrahydrocortisol; a-THF, allo-tetrahydrocortisol; THE, tetrahydrocortisone.


bMedian and interquartile range (IQR).

cMean and 95% confidence interval (CI).

*P<0.05, **P<0.001.

LVM analysis

Hypertensive cohort

Concentric LVH was present in 32 hypertensives (6.5%) and eccentric LVH in 97 (19.7%); 31 hypertensives (6.3%) showed concentric remodelling. Normal heart geometry was shown in 333 hypertensives (67.6%). LVMh2.7 correlated with THFs/THE (Table 2), age (r=0.14, P<0.001), BMI (r=−0.27, P<0.001), systolic BP (r=−0.09, P=0.04), average 24-h ambulatory mean BP (rho=0.156, P=0.033), and recumbent PRA (rho=−0.25, P<0.001). LVMBSA was associated with THFs/THE (Table 2), sex (rho=−0.18, P<0.001), and recumbent PRA (rho=−0.22, P<0.001). Duration of hypertension, diastolic BP, and 24-h urinary aldosterone did not correlate with LVM. Twenty-four hour urinary Na+ correlated with LVD (rho=0.15, P=0.027), but not with LVMh2.7 (rho=0.04, P=0.31).

View this table:
Table 2

Pearson/Spearman correlation betweenTHFs/THE and clinical/sonographic characteristics

Hypertensives (n=493)Normotensives (n=98)
Hypertension duration0.090.04
Heart rate−0.050.28−0.020.55

Normotensive cohort

Eight out of 98 normotensives (8.2%) showed LVH.30 The low numbers did not allow analysis of the cardiac patterns.

Analysis of THFs/THE

THFs/THE ranged from 0.27 to 2.64 in the hypertensive, and 0.33 to 1.61 in the normotensive, cohort. Assuming the mean value +2SD of the THFs/THE in the normotensives cohort [(0.98)+(2×0.28)=1.54] as the upper normal limit, we found that n=119 hypertensive patients (24.1%) showed a THFs/THE higher than normal. These THFs/THE values are very close to those found in our previous paper examining different cohorts.8 If the upper limit is set as 2.00 on the basis of the cut-off for SAME,6 then n=33 (6.7%) hypertensive patients showed higher values, in the absence of clinical/laboratory features of SAME.

Consistent with our previous data,8 alloTHF and THE were significantly different in hypertensives as opposed to normotensives (Table 1) and thus cause the changes of THFs/THE.

Sex, BMI, and duration of hypertension correlated with THFs/THE. The latter also showed an inverse correlation with recumbent PRA (Table 2). Figure 1 shows the partial regression leverage plot of recumbent PRA against THFs/THE after both have been adjusted for confounders (BMI, sex, age) in the hypertensive cohort.

Figure 1 Partial regression leverage plot of recumbent PRA (rPRA) against THFs/THE (logarithmic transformation) after both have been adjusted for confounders (BMI, sex, age). Data refer to the hypertensive cohort only (n=493).

Analysis of the association between THFs/THE and LVMh2.7

THFs/THE as a continuous variable

A positive association of THFs/THE with LVMh2.7 was present in both hypertensive and normotensive cohorts (Table 2), as well as when the cohorts were combined (n=591, r=0.38, P<0.001).

In hypertensives, multiple regression analysis, including at the same time THFs/THE, systolic and diastolic BP, age, sex, BMI, and recumbent PRA, showed significant independent association of LVMh2.7 with THFs/THE [β=0.40 (SE=0.04), P<0.001], BMI [β=−3.84 (SE=0.64), P<0.001], sex [β=0.04 (SE=0.01), P=0.033], whereas only a trend was evident with regard to recumbent PRA [β=−0.03 (SE=0.02), P=0.07]. No association of LVMh2.7 with age, SBP, or DBP was present (P>0.05). The model containing THF/THE, sex, BMI, and recumbent PRA accounted for 26.1% of the variation in LVMh2.7, while the same model without THFs/THE explained only 10.7%. Figure 2 depicts the relationship between LVMh2.7 and THFs/THE after both have been adjusted for confounders. The association of THFs/THE with LVM was present also in our hypertensive cohort, in patients with recumbent PRA values both below and above 1.07 ng/mL/h (the median of recumbent PRA in our hypertensive cohort): multiple regression analysis in the two sets of patients gave a β coefficient of 0.45 and 0.33, respectively, P<0.001 for both.

Figure 2 Partial regression leverage plot of LVMh2.7 against THFs/THE after both have been adjusted for confounders (BMI, sex, recumbent PRA, age). Both THFs/THE and LVMh2.7 are mathematically transformed (square root and logarithmic, respectively). Data refer to the hypertensive cohort only (n=493).

Multiple regression analysis including either IVST, PWT, or LVD as dependent variables and THFs/THE, recumbent PRA, sex, age, BMI and BPs as explanatory variables did not show any significant association of THFs/THE with the LV components (P>0.05 in all cases).

No interactive effect of 24-h urinary Na+ with THFs/THE or with aldosterone on LVMh2.7 was shown (P=0.26 and 0.29, respectively).

THFs/THE as a discrete variable

LVMh2.7 was lower in patients with THFs/THE below the median [43.0 (95% CI 42.0–44.0), n=243, vs. 46.5 g/m2.7 (95% CI 45.5–47.6), n=250; P<0.001].


We report herein a positive association of 11β-HSD2 activity, as assessed by urinary THFs/THE, with LVM in never-treated essential hypertensives and normotensives. An inverse correlation of THFs/THE with recumbent PRA was also found suggesting that the effect of 11β-HSD2 activity is concurrently present in the heart and in the kidney.

The association between LVMh2.7 and THFs/THE persisted after adjustment for confounders and helped explain a relevant portion of LVM variability. In fact, sex, BMI, recumbent PRA, and THFs/THE explained 26.1% of the variability of LVMh2.7 vs. 10.7% explained by the same model without THFs/THE. Moreover, THFs/THE correlated not only with LVMh2.7 but also with LVMBSA and this correlation was present, although weaker, in a cohort of normotensives as well as in the combined cohorts, suggesting that glucocorticoid-mediated activation of the MR could play a role in the regulation of this cardiovascular phenotype independent of BP status.

However, it is not surprising that we failed to find any significant association of THFs/THE with IVST, PWT, and LVD analysed one at a time, as LVM calculation is a complex expression of the former.

For the present study, we measured the urinary metabolites of cortisone (THE) and cortisol, originating from its dehydration through the 5β- and 5α-reductase enzymes (THF and alloTHF, respectively). Consistent with our previous study,8 we have confirmed the concurrent reduction of 11β-HSD2 and 5β-reductase activity, leading to a reduced inactivation of cortisol to cortisone and to a greater synthesis of alloTHF: 24.1% of our essential hypertensives showed a THFs/THE ratio higher than the mean +2SD of THFs/THE measured in the normotensive cohort.

We measured THFs/THE as an index of 11β-HSD2 activity. Traditionally, THFs/THE has been used as a proper indicator of 11β-HSD2 activity in vivo, particularly in the diagnosis of SAME.5,7 Although a possible advantage of urinary free cortisol/cortisone (UFF/UFE) has been reported, THFs/THE displays lower intra-individual variability and better discrimination between salt-sensitive and salt-resistant subjects.29,35,36

While 11β-HSD2 is absent or present only in very low levels in several animal species,20 in man the MR and 11β-HSD2 are co-expressed in the heart suggesting that both the kidney and the heart have developed a specific machinery for glucocorticoid inactivation.19,20 To date, several components of the renin–angiotensin–aldosterone system have been taken into account in LVM regulation: while in experimental hypertension, cardiac fibrosis is generated mainly by aldosterone and angiotensin II, the association of aldosterone with LVM is still controversial both in primary aldosteronism17,18 and in EH.16,37 In our study, 24-h urinary aldosterone was in the normal range, and was not associated with LVMh2.7, nor with THFs/THE and PRA; however, its variations could have been missed in the 24-h urine collection.

Angiotensin II has been reported to influence LVM through sodium retention and volume expansion.32 Both glucocorticoids and mineralocorticoids can potentiate the vasoconstrictor action of a number of pressor hormones, including angiotensin II, by up-regulation of their respective receptors.38 Moreover, glucocorticoids can potentiate angiotensin II-induced ion transport in renal tubular cells in conditions of 11β-HSD2 impairment.39 Although speculative, since we did not measure angiotensin II, all this could have contributed to the raised LVM as a function of THFs/THE observed in our study. Nevertheless, the inverse correlation between recumbent PRA and LVM does not support a major effect of angiotensin II on LVM. PRA and angiotensin II are directly related, i.e. angiotensin II is low when PRA is low. Instead, other factors, such as cortisol, could account both for the increased LVM and for the modulation, due to negative feedback, of the recumbent PRA levels. The finding that in Cushing's syndrome long-lasting exposure to increased cortisol, rather than hormone or BP levels, is a major determinant of LVM further supports our hypothesis.4042

The inverse correlation of THFs/THE with recumbent PRA in our study suggests that the ‘glucocorticoid-mediated mineralocorticoid effect’ is not confined to the heart. The absence of a correlation between THFs/THE and upright PRA does not weaken this finding: recumbent PRA reflects the salt balance more closely than upright PRA, which is influenced by sympathetic activity.

In the RALES and the EPHESUS clinical trials, circulating aldosterone was normal and Na+ status unremarkable: nevertheless, MR blockade had unquestioned benefits.43,44 Our hypothesis, which focuses on the MR however it is activated, as the main factor responsible for cardiac fibrosis, could help in the interpretation of the findings of these trials. About 20% of EH patients show a mild impairment of 11β-HSD2 activity;7,8 a similar percentage of EH patients develop LVH, a well-known risk factor for cardiac damage.1113 If a causal relationship between mild impairment of 11β-HSD2 activity, LVM, and cardiac damage were defined, aldosterone antagonists could be used rationally to reduce cardiovascular morbidity and mortality in this subgroup of EH patients.

Dietary salt intake is reported to be an independent determinant of LVM in EH45 and in the general population.46 In our study, 24-h urinary Na+ was associated with LVD, but not with LVM or with THFs/THE.

As reported by others12,47, a correlation of 24-h ambulatory mean BP with LVMh2.7 was shown in our study (rho=0.156, P=0.033); only a trend was seen for office systolic BP (r=0.08, P=0.07). The duration of hypertension was not associated with LVM in our cohort while a slight correlation with THFs/THE was seen (Table 2). Particular care was taken to assess the duration of hypertension based on previous medical documentation: the long duration of sustained untreated hypertension in our cohort (Table 1) could be explained by the patient's underestimation of the risk associated with the disease and their decision not to take any antihypertensive drugs, often in the absence of symptoms. No concerns were raised with regard to the observational period of 8 weeks prior to treatment: all the patients were accurately followed up (visits at 2-week intervals) and all the symptoms were monitored. This is consistent with the guidelines in the management of EH.21

We are aware of several limitations of our study:

  1. Myocyte hypertrophy and fibrosis are the most common features of hypertension-associated cardiac structural changes.14 Ultrasound techniques cannot distinguish between fibrous and muscular components. Thus, we cannot say which contributed most to LVM in our study, nor did we evaluate the indices of cardiac stiffness.

  2. Cardiac and renal 11β-HSD2 direct measurement was not performed.

  3. Our conclusions are drawn on the basis of correlative data, although great attention was paid to adjusting for confounders. In any case, we want to stress that the correlation between THFs/THE and LVM does not even imply a cause-and-effect relationship. On the other hand, the relatively large number of patients studied, as well as the absence of antihypertensive treatment, could represent a strength of the study.

In conclusion, we suggest that glucocorticoids may take part in LVM regulation through the MR; the activation of the latter, regulated by 11β-HSD2, could contribute to the target organ damage associated with EH.


This work was supported by MIUR (Italian Ministry of Education, University and Research, Grant FIRB #RBNE01724C). The authors are indebted to Dr Mario Palermo, Professor Paul M. Stewart, and Professor Giuseppe Bianchi for their comments and advice, and to Dr Tracy Ann Williams for editorial revision.


View Abstract