Krystallenia I. Alexandraki; Gregory A. Kaltsas; Apostolos-Ilias Vouliotis; Theodoros G. Papaioannou; Lauren Trisk; Athanasios Zilos; Márta Korbonits; G. Michael Besser; Aris Anastasakis; Ashley B. Grossman
Posted: 05/16/2011; Clin Endocrinol. 2011;74(5):558-564. © 2011 Blackwell Publishing
Abstract and Introduction
Objective Hypercortisolaemia is associated with an increased risk of cardiovascular disease (CVD), either through a direct action on the myocardium or by increased traditional cardiovascular risk factors. The aim of this study was to investigate whether the alterations in the ECG in Cushing's disease (CD) are predictable from risk factor analysis alone.
Design In 79 patients with a diagnosis of CD, retrospectively recruited, ECG features [corrected for heart rate QT (QTc), QTc dispersion (QTcd), left ventricular hypertrophy (ECG-LVH), right ventricular hypertrophy (ECG-RVH)], systolic (SBP) and diastolic (DBP) blood pressure were assessed. Biochemical, hormonal (cortisol at 09·00 h or cortisol day curve, CDC) and carbohydrate abnormalities (CHA), history of hypertension and cardiovascular disease were recorded. For comparison reasons, a group of 42 healthy subjects matched for gender, age and body mass index previously subjected to ECG assessment were selected.
Results In patients with CD, we noted the following prevalence: metabolic syndrome 39%, hypertension 81%, CVD 21·5%, hypercholesterolaemia 37%, hypertriglyceridaemia 29%, CHA 41%, but a history of cardiac dysrhythmia was only noted in a single patient. No difference in QTc or QTcd was shown between patients with normal or low potassium levels. QTcd >50 ms was associated with both increased ECG-LVH and ECG-RVH. When compared to the control group, patients had longer QTcd (P < 0·001), more prevalent LVH (P < 0·001) and RVH (P = 0·001), and higher SBP and DBP (P < 0·001), but similar QTc. Both CD and ECG evidence of LVH predicted prolonged QTcd, but the association of CD with a prolonged QTcd was independent of other risk factors, including hypertension.
Conclusions Prolonged QTcd in association with ECG evidence of LVH appears to be the specific feature of CD. This may be relevant in the choice of medical therapy for CD and for consideration of treatment of the comorbidities that are associated with hypercortisolaemia.
Untreated hypercortisolaemia, Cushing's syndrome (CS), is associated with the presence of a number of cardiovascular (CV) risk factors, increased CV morbidity, and a mortality rate that is fourfold higher than that expected in the normal population.[1–3] In terms of CV morbidity, this may result from the increased prevalence of traditional metabolic risk factors[2,3] or reflect a direct glucocorticoid effect on the myocardium,[4–7] particularly as glucocorticoid receptors have been demonstrated in the human heart.[4,6]
Analysis of the QT interval measured from electrocardiographs (ECGs) and ECG features of left (ECG-LVH) and right ventricular hypertrophy (ECG-RVH) represent established and readily assessable indicators of CV pathology. Despite the fact that the clinical value of QT abnormalities as indicators of repolarization abnormalities has been questioned, their role as predictors of cardiovascular disease (CVD),[9–16] along with cardiac death and overall mortality,[17–20] appear to be generally accepted.
We speculated that the increased CV morbidity in Cushing's disease (CD) is associated with changes in the QT interval and have questioned whether any such changes were specific to CD or may be secondary to the known risk factors. This is particularly important as recent data have reported increased QTc in patients with CD treated with the recently introduced somatostatin analogue pasireotide (Novartis, Basel, Switzerland: data on file), as well as other established and frequently used medical treatments for CD (http://www.azcert.org/medical-pros/drug-lists/bycategory.cfm; http://www.sads.org.uk/drugs_to_avoid.htm).
Materials and Methods
Patients and Metabolic Variables
Seventy-nine patients [61 women, age: 40·76 ± 1·38 years, range: 18–72] from the CD patient database of the Department of Endocrinology, St Bartholomew's Hospital, diagnosed after the completion of a full diagnostic work-up, as previously described, and with an adequate ECG performed on their first admission to the hospital, were retrospectively studied. The study protocol was approved by the Institutional Committee on Human Research as a case record review study [audit numbers 08–76/09–143].
In the patient group, biochemical [total cholesterol (TC), triglycerides, glucose] and hormonal [09·00 h serum cortisol and cortisol day curve (CDC, involving measuring serum cortisol levels at 5 fixed times through a day)] measurements were recorded. Hypertension was defined a blood pressure measurement of ≥130 mmHg systolic (SBP) and/or ≥85 mmHg diastolic (DBP) or history of antihypertensive treatment. Abnormalities of carbohydrate metabolism (CHA) were defined as diabetes mellitus (DM), impaired glucose tolerance (IGT) or impaired fasting glucose (IFG). Hypercholesterolaemia was defined as TC levels >5·2 mm or a history of treatment for hypercholesterolaemia. Hypertriglyceridaemia was defined as triglycerides levels ≥1·7 mm or a history of treatment for hypertriglyceridaemia. To characterize the population studied metabolically and since HDL-cholesterol was not assessed, the presence of the metabolic syndrome (MS) was estimated from any three of the following four parameters: (1) hypertriglyceridaemia, (2) hypertension, (3) glucose ≥5·6 mm and (4) presence of abdominal obesity as registered in the case notes (because waist circumference was not routinely measured). Body mass index (BMI) was calculated [BMI = weight (kg)/height (m)2]. The ethnic group and drug history was also recorded. In particular, the medications used at the time of ECG assessment were registered and compared to a contemporary published list of drugs generally accepted to be associated or weakly associated with prolonged QT interval (http://www.azcert.org/medical-pros/drug-lists/bycategory.cfm; http://www.sads.org.uk/drugs_to_avoid.htm). The period between the time of occurrence of the first feature(s) of CS and diagnosis was defined as 'CD duration'.
The patient population was divided into subgroups according to the presence or absence of a history of overt CVD (angina, myocardial infarction, arrhythmias, stroke, thromboembolic disease, cardiac failure) and/or hypertension, according to low or normal range levels of potassium and according to their gender.
For comparison reasons, a gender, age and BMI-matched group of 42 healthy subjects (29 women; age: 41·76 ± 0·84 years) was selected from the database of the 1st Department of Cardiology, Hippokration Hospital, Athens. None of these subjects had ever received chronic treatment with glucocorticoids or drugs known to affect glucose, lipid metabolism or blood pressure. All these subjects fasted and abstained from smoking, caffeine and ethanol intake for at least 12 h before evaluation. All subjects in this database were Caucasian.
Analysis of the Electrocardiograms
In all subjects, a standard 12-lead ECG was performed under similar conditions (supine position and normal respiration) at a paper speed of 25 mm/s. Because of the documented problems in reproducibility using automated screening of the QT, all measurements were performed manually by two investigators and repeated twice; average values were computed for each of the ECG features.[8,25] The degree of concordance between the two investigators was 95%; the degree of concordance for each individual investigator was 97% and 98%.
ECGs were obtained at the first admission of the patient group and before any type of treatment for CS. ECGs were only accepted if all 12 leads could be analysed. All patients were admitted for investigation of CS, and all of them were haemodynamically stable without symptoms of any acute CV event. Patients with known CVD as already reported were included in the study, because CVD is included in the previous medical history of patients with untreated CS; however, they were analysed as a separate subgroup.
For the QT interval measurements, the Tangent method was used: the end of the T-wave was defined as the intersection of a tangent to the steepest slope of the last limb of the T-wave and the baseline. The Tangent method was also useful for patients with co-existing hypokalaemia and U waves in resting ECGs (five patients had hypokalemia and prominent U-wave in the ECG). If a U wave appeared immediately after the T wave, the QT interval was measured at the nadir between T and U waves, and thus the 'second hump' (U wave) was excluded; QT intervals were measured in all 12 leads and corrected for heart rate (QTc) with Bazett's formula: QT interval/square root of the RR interval. However, in the standard 12-lead ECG, only two of the six extremity leads are actually recorded, because the other four leads are derived mathematically from these two leads. Thus, if we find the shortest QT interval in one of the extremity leads, the other five extremity leads must have the same QT interval. Hence, we took the shortest QT value recorded from the extremity leads and the mean value of the other five QT intervals. Furthermore, we excluded QT interval measurements from lead V1 because very often the T wave has an isoelectric segment in that lead. Hence, two QT-interval values were recorded from the extremity leads, and another five were recorded from the precordial leads (V2–V6). Thus, we obtained a maximal value for QTc in a lead and a minimal QTc value in another lead: QTc dispersion (QTcd) was defined as the maximum QTc interval (QTc max) minus the minimum QTc interval (QTc min) in seven leads (i.e., the five precordial, the shortest extremity lead and the median of the other five extremity leads).[10,18,27–29] The prevalence of patients with QTcd >80 ms and >50 ms was calculated, as these thresholds have been associated with increased mortality.[16,19,30]
To increase sensitivity (because of the female preponderance in CD), we preferred to use the Sokolow-Lyon product criterion for women and the Cornell criterion for men to diagnose ECG-LVH in both groups. ECG-RVH was defined by the use of any two combinations of the following criteria: (1) R/S ratio in lead V5 or V6 less than or equal to 1; (2) S V5 or V6 greater than or equal to 7 mm; (3) right-axis deviation of more than +90 degrees or (4) P pulmonale.
We planned to exclude from the study any ECGs with intraventricular conduction defects and delays (IVCDs) or bundle-branch block (BBB), as both IVCDs and BBB affect the QRS patterning, and may impact the accuracy of ECG criteria for LVH. However, none was found in our patient cohort.
Values are presented as mean value ± standard error (±SE). Statistical significance was accepted at a P-value<0·05. The distribution of normality of continuous variables was assessed graphically by histograms and statistically by the nonparametric Kolmogorov–Smirnov test. Comparisons between control and patient groups were made by the unpaired t-test or the Mann–Whitney U tests for normally or non-normally distributed variables, respectively. Correlations between categorical variables were estimated by the chi-square test or by Fisher's exact test, as appropriate. In the patient population, independent variables (ECG indices, SBP, DBP, age, gender, BMI, duration of the disease, lipids, fasting glucose, morning cortisol, CDC) associated with the presence of ECG-LVH and ECG-RVH were evaluated by univariate and multiple logistic regression analysis because those variables were dichotomous, and QTc and QTcd by univariate and multiple linear regression analysis because those variables were continuous. In the total population, independent variables (ECGs indices, SBP, DBP, age, gender and BMI) associated with the presence of LVH and RVH were evaluated by univariate and multiple logistic regression analysis, and QTc and QTcd by univariate and multiple regression analysis. Analysis of variance (anova) with Bonferroni correction and the Kruskal–Wallis test with Conover-Inman correction were used for multiple group comparison for normally and non-normally distributed variables. Analysis was performed using spss (version 16·01; SPSS, Inc., Chicago, IL, USA) for Windows XP (Microsoft Corp.).
The metabolic and CV features of the patients with CD are shown in Table 1. The median value of 'CD duration' was 3 years (range <1–39 years). No correlation was found between CD duration and any of the parameters studied.
On analysis of the previous medical history of the patients, we noted that 16 patients had a history of CVD: three with deep vein thrombosis, two with stroke, four with angina pectoris, three with myocardial infarction and four with cardiac failure, while one more patient had a single episode of atrial fibrillation.
Patients with CD had a longer QTcd (P < 0·001), more prevalent ECG indices of LVH (P < 0·001) and RVH (P = 0·001), higher SBP (P < 0·001) and DBP (P < 0·001) values, but a shorter QTcmin (P = 0·004), compared to the control group (Table 2).
The differences between controls and patients subgroups with (A) CVD, with and without hypertension, (B) hypertension only and (C) without CVD or hypertension are shown in Table 2. QTcd was prolonged in all the subgroups (P < 0·001), and no relevant difference was seen in QTc, QTcd, ECG-LVH or ECG-RVH between these groups besides the fact that QTcd was more prolonged in subgroup B compared to subgroup C. It is of note that patients of subgroup C still had a prolonged QTcd compared to the controls, despite the fact that they were younger (P = 0·02; Table 2). In this specific group, no patient had DM, although one had IFG and another IGT. In addition, there was a statistically significant difference between the groups regarding sex (women: A: 56%, B: 78%, C: 100%, P = 0·01).
In 44 (56%) patients, the QTcd value was more than >50 ms and in 5 (6·3%) more than >80 ms (maximal value 87 ms). Interestingly, in six out of seven patients on treatment with drugs associated with prolonged QT, QTcd value was >50 ms (85·7%) but only in one >80 ms (14·3%). When the subgroups of patients with QTcd ≥50 ms or <50 ms were compared, there was a higher prevalence of ECG-LVH (P = 0·004) and ECG-RVH (P = 0·01) in the first subgroup. By contrast, only 2/42 (4·8%) subjects from the control group had a QTcd >50 s, and all were <60 ms. Hence, the sensitivity of QTcd >50 ms to identify a patient with CD was 50·6%, the specificity 95·2%, the positive predictive value (PPV) 95·6% and the negative predictive value 53·3%. However, there was no difference in ECG-LVH or ECG-RVH prevalence between the subgroups of QTcd ≥80 ms and <80 ms.
With regard to the drug history, 28 patients were on no medication, while the remaining patients were on treatment for oedema and/or hypertension (diuretics, potassium-sparing diuretics, aldosterone antagonists, β-adrenoceptor blockers, calcium antagonists, angiotensin-converting enzyme inhibitors), potassium replacement, nitrates, anticoagulants, aspirin, statins and fibrates, sulphonylureas, metformin or insulin, analgesics, psychiatric drugs or drugs for digestive problems. However, only seven patients were on treatment with drugs generally accepted or weakly associated with prolonged QT, namely sotalol, hydrochlorthiazide, sertraline, clomipramine, imipramine, chlorpromazide or fluoxetine. All these patients were women, and all were included in the hypertension-only subgroup. The only differences that were observed when this group was excluded from analysis were that the QTcd in the hypertensive group was reduced and the difference between patients with CVD alone and HTN alone was no longer significant (Table 2). As no other difference was observed in the results, we have included those seven patients in the final analysis.
Seventy-one of the 79 patients had records of plasma potassium levels; the mean level was 3·8 ± 0·68 mm (range: 2·6–5 mm). Fifty-seven (80%) patients had normal potassium levels (median value: 3·9 mm; range: 3·5–5 mm), while 14 had low potassium levels (median value: 3·2 mm; range: 2·6–3·4 mm). No difference was found in the QTcd between these groups (normal potassium: 52·34 ± 2·15 ms vs low potassium: 51·66 ± 5·57 ms, P = 0·90) or in QTc (normal potassium: 392·11 ± 3·11 ms vs low potassium: 388·21 ± 8·10 ms, P = 0·60). In the low-potassium group, six patients (42·9%) had QTcd >50 ms and two patients (14·3%) had QTcd >80 ms, while in the normal-potassium group 35 (61·4%) had QTcd >50 ms and three patients (5·3%) had QTcd >80 ms; these difference were not statistically significant (P = 0·24 and 0·25, respectively). Furthermore, no correlation was found between potassium levels and QTcd or QTc in the whole group of patients or in each subgroup of potassium levels.
Regarding the comparison between female and male patients with CD, no significant difference was found in any of the parameters studied. Men had QTcd 51·38 ± 3·09 ms vs women 51·82 ± 2·29 ms (P = 0·92) and QTc 391·11 ± 3·39 ms vs 391·8 ± 3·13 ms, respectively (P = 0·92). Interestingly, the accepted difference between women and men in QT interval was seen only in the control group in QTcd with men having 15·46 ± 1·87 ms vs women 23·1 ± 2·63 ms (P = 0·02) but not in QTc with men having 385·77 ± 6·28 ms vs 398·79 ± 4·44 ms, respectively (P = 0·11).
Seventy-one patients were Caucasian. To control for racial differences, a group of 71 Caucasian patients with CD was compared to the control population (only Caucasian), and similar results were observed as for the total group (data not shown).
In the patient population, univariate analysis showed that the presence of ECG-LVH (b = 9·78, P = 0·03) and ECG-RVH (b = 12·48, P = 0·002) were predictors of QTcd. However, multivariate analysis revealed that only ECG-LVH (b = 11·20, P = 0·02) remained a predictor of QTcd. In univariate analysis, only CDC (OR = 1·004, CI = 1·0–1·007, P = 0·04) was a predictor of ECG-LVH, while only triglycerides levels (OR = 3·14, CI = 1·23–7·99, P = 0·02) were a predictor of ECG-RVH.
In the total population of controls and patients, univariate analysis showed that the presence of CD (b = 30·99, P < 0·001), SBP (b = 0·37, P < 0·001), DBP (b = 0·57, P < 0·001), ECG-LVH (b = 2431, P < 0·001) and ECG-RVH (b = 21·44, P < 0·001) were predictors of QTcd. However, in multivariate analysis only the presence of CD (b = 24·05, P < 0·001) and ECG-LVH (b = 10·51, P = 0·02) remained predictors of QTcd (Table 3). In the multivariate analysis, only DBP remained a predictor of ECG-LVH and ECG-RVH (Table 3).
In this study, analysis of the baseline ECG characteristics of patients with CD at diagnosis demonstrated that QTcd but not QTc was strongly associated with CD along with the well established association of LVH with QTcd. In addition, hypercortisolaemia was found to be associated with the presence of LVH, implying an interplay of CS pathophysiology with ECG characteristics before the introduction of any therapy. As CD is associated with an increased risk of CVD and mortality, ECG features might represent an easily assessable CV-risk marker present early in the natural history of CD.
We found a high prevalence of hypertension and CVD in CD, in accordance with previous studies[2,26] in addition to a deleterious metabolic profile.[2,23] The presence of these factors may well contribute to the increased morbidity and mortality seen in untreated CD and emphasise the need for early and effective intervention to treat the disease. We did not find any clear relationship between these risk factors and the degree of hypercortisolaemia, but we were dealing with a population of patients with homogeneously increased levels of cortisol. Interestingly, the duration of the disease in this study did not predict any ECG abnormality implying that the presence of these risk factors is not a function of the length of excessive glucocorticoid exposure.[3,34] Nevertheless, it is important to note that many of these risk factors may remain present even years after the cure of CD, suggesting that continued vigilance is required in monitoring their presence and in initiating treatment.[6,35]
In terms of ECG features, a direct comparison of the patient group and controls revealed prolonged QTcd and ECG features of LVH in patients with CD. However, the significant difference in QTcd remained present even in patients without overt CVD or hypertension, suggesting that the conventional risk factors alone could not explain, on their own, all of the cardiac morbidity in CS. It is also of note that QTc was not as sensitive as QTcd in revealing any difference among the subgroups studied. Regression analysis confirmed the well-known association between QTcd and LV mass.[36,37] On the other hand, it has been previously suggested that the relationship between prolonged QTc and future cardiac mortality may be attributed to ventricular electrical instability and abnormal repolarization, while increased QTcd reflects electrical inhomogeneity as a result of myocardial ischaemia, ventricular hypertrophy or dilatation, autonomic neuropathy, peripheral vascular disease or hypertension.[9–14,17,18,39] However, in our study, only a single female patient had a history of a cardiac dysrhythmia. Hence, our findings support the studies that dispute the validity of QT interval as an indicator of repolarization abnormalities, but are in keeping with more recent evidence suggesting its role as predictor of CV risk not always related to electrical disturbances.[9–19] There was no direct correlation between QTcd and the degree of hypercortisolaemia, but this may again reflect the rather uniform increase in hypercortisolaemia in the study population. However, the fact that prolonged QTcd could be predicted by a diagnosis of CD in the total population or by LVH that is affected by the degree of hypercortisolaemia suggests that there may be a direct impact of hypercortisolaemia on QTcd. If this speculation is valid, it may be of even greater clinical significance considering the large number of patients on glucocorticoids for medical reasons; the increasingly diagnosed cases with subclinical CS and mild hypercortisolaemia may also need to be treated even in the face of normal or treated conventional risk factors. It is of note that despite the recent debate regarding the clinical validity of QTcd, this is the feature of ECG that was found abnormal in this specific patient population and not QTc.[8,41] The real significance of the better diagnostic value of QTcd over QTc found in CD in this study cannot be explained because the debates regarding the clinical significance of these CV risk markers have not been resolved. However, it has been documented that patients with an increased QTcd, which may mirror discrepant repolarization characteristics in different areas of the heart, show increased CV mortality and morbidity and that both prolongation of the QT interval and QTcd dispersion independently affected the prognosis of CV mortality and cardiac fatal and nonfatal morbidity.
A QTcd >80 ms, which has been considered indicative of cardiac mortality,[16,30] was seen only in 6·3% of the patients. However, we found that the lower established threshold (>50 ms) was observed in 56% of patients and was related to both ECG-LVH and ECG-RVH, demonstrating high specificity and PPV. Finally, while severe hypercortisolism may be associated with hypokalaemic alkalosis, the fact that no difference was observed in QTcd or QTc between patients with normal or low potassium levels implies that prolonged QTcd is a distinct feature of CD independent of potassium levels. Furthermore, the known differences in women and men in QTcd was not seen in patients with CD as in the normal population.
Finally, the recent findings of a specific effect of the glucocorticoids acting on mineralocorticoid and/or glucocorticoid receptor have to be considered along with the results of this study on LVH and RVH prevalence. Indeed, in adults, treatment with mineralocorticoid receptor antagonists results in a reduction in cardiac hypertrophy and an increase in survival rate,[44,45] mainly because of a decrease in cardiac fibrosis. However, the research in this area is ongoing, and the role of glucocorticoids acting through the cardiac mineralocorticoid receptor remains controversial.[47,48]
This study has a number of limitations, principally because of the retrospective recruitment of patients with CD. Some of the patients were on drugs that could influence QTcd; however, since this study focused on patients with CD who may be under variable pharmacological treatment for several years before their diagnosis and thus this is considered a more representative group of patients as seen in clinical practice. Indeed, the findings persisted even if such patients were excluded. Indeed, the fact that hypercortisolaemia per se seems to be associated with a prolonged QTcd suggests that particular caution is required in such patients when drugs known to alter this parameter are administered. Another limitation was the fact that patient and control groups were from different national backgrounds. We also used a different national group for our control population for logistic reasons, but as similar changes were seen when the small number of non-Caucasian patients were excluded from the analysis, it seems highly unlikely that national identity can account for these marked changes.
In conclusion, a detailed comparison of the ECG features of a group of patients with CD and a BMI-matched control group showed ECG abnormalities along with the expected increase in cardiac risk factors. However, certain of these changes appeared to be specific to this disorder as they correlated directly or indirectly with the hypercortisolaemia independent of any known risk factors. We suggest that hypercortisolaemia per se can be cardiotoxic, and this needs to be considered when assessing the need for therapy even in cases of mild hypercortisolaemia. Prolonged QTc dispersion is classically considered to increase the risk of serious dysrhythmias such as torsades de pointes;[49–51] this may be further increased by certain drugs included in the therapeutic armamentarium of CD, while the effects of others such as somatostatin analogues and ketoconazole are under investigation (http://www.azcert.org/medical-pros/drug-lists/bycategory.cfm; http://www.sads.org.uk/drugs_to_avoid.htm); these findings should be borne in mind when considering pharmacotherapy.