Drug-induced new onset diabetes in hypertension: does it matter?

Hypertension is often accompanied by abdominal obesity, metabolic abnormalities and insulin resistance, increasing the risk of development of impaired glucose metabolism and type-2 diabetes [1,2]. This is well known from several observational studies in different populations [3,4]. During the last 15 years, it has also been debated whether or not the hyperglycaemia and new-onset diabetes (NOD) associated with the use of some antihypertensive drugs are of an equal importance or not for the long-term risk of cardiovascular disease as the risk of non-drug hyperglycaemia, which is so often found as a hallmark of many patients with essential hypertension. In a series of elegant studies, the late Hans Lithell (Uppsala, Sweden) investigated the effects on insulin sensitivity by various antihypertensive drugs using hyperinsulinaemic, euglycaemic clamp methodology [5,6]. He and his co-workers documented that non-selective β-receptor blockers and high-dose thiazide diuretics tended to decrease insulin sensitivity whereas agents that block the renin-angiotensin system (RAS), as well as vasodilators such as α-blockers, tended to improve it [6]. Calcium channel blockers were mostly neutral. The conclusion from these experiments was that the benefits of blood pressure-lowering could eventually be offset by derangement of glucose and lipid metabolism caused by a worsening of insulin sensitivity. The ensuing hyperglycaemia and dyslipidaemia were suspected of negatively influencing the progress of arterial atherosclerosis and macrovascular disease manifestations, such as coronary heart disease and ischaemic stroke. However, all of these clamp studies [6] were of relatively short duration and not powered to investigate effects on cardiovascular end-points. Notwithstanding, they were carefully conducted and have proven that antihypertensive drugs are able to influence different physiological and metabolic processes, for better or worse. Does drug-induced hyperglycaemia and NOD really matter with respect to clinical prognosis during long-term follow-up? This view has been advocated by data obtained from observational studies [7,8], but challenged by others [9,10]. Verdecchia et al. [8] estimated that one cardiovascular event associated with new-onset diabetes might be prevented for every 385–449 patients treated with ‘new’, rather than ‘old,’ antihypertensive drugs for approximately 4 years [8]. The counter-argument is that, both in the ALLHAT study [9] and in the SHEP follow-up study after 15 years [10], no detrimental effects of hyperglycaemia secondary to thiazide treatment (chlorthalidone) were recorded for macrovascular events. In a recent post-hoc analysis, the ALLHAT investigators concluded that fasting glucose levels increase in older adults with hypertension regardless of treatment type in their study. For those taking chlorthalidone versus other medications, the risk of developing fasting glucose levels higher than 6.9 mmol/l was modestly greater, but there was no conclusive or consistent evidence that this diuretic-associated increase in diabetes risk also increases the risk of clinical events after 2 years of follow-up [9]. On the other hand, the full ALLHAT study was of rather short duration (4.9 years) and the SHEP follow-up study did not have full control of changes in medication, lifestyle and medical conditions in the post-trial period. Furthermore, no data have ever been provided on the risk of developing microvascular complications, such as retinopathy and peripheral angiopathy, in these patients. This is important because it is well established that hyperglycaemia is strongly associated with microvascular complications in general, as shown in patients with diabetes [11]. Newer studies such as LIFE [12], VALUE [13] and ASCOT [14] have contributed data showing that agents that block the renin–angiotensin system, also in combination with a dihydropyridine calcium antagonist (ASCOT), are associated with less NOD than treatment with conventional antihypertensive drugs. This is supported by data from a recent network meta-analyses by Elliot and Meyer [15], where it was documented that, by using thiazide diuretics as reference, angiotensin-converting enzyme (ACE)-inhibitors, angiotensin-2 receptor blockers, calcium antagonists and placebo were all significantly less likely to develop on-treatment NOD, but there was no difference in relation to β-receptor blockers [15]. Taking all this evidence together, as reported in a recent ESH Newsletter on the prevention of diabetes by antihypertensive drugs [16], it was recommended to use antihypertensive drugs with a neutral or beneficial effect on metabolism in middle-aged patients at risk for the development of type-2 diabetes (e.g. with the metabolic syndrome). New evidence from a long-term Swedish study In this issue of the journal, Almgren et al. [17] provide new and interesting data from a long-term follow-up of the Gothenburg Hypertension Clinic data. The authors have followed patients with hypertension from the mid-1970s to recent years. The risk of cardiovascular end-points was significantly increased in subjects with NOD detected during the follow-up period. This report is an extension of a previous longitudinal study carried out by the same group [18]. In that study, the authors found a non-significant excess risk of coronary artery disease associated with NOD in the cohort of 686 hypertensive subjects followed for 15 years. Now, with a longer follow-up (25–28 years) and considerably more outcome events accumulated, NOD turned out to be a significant independent predictor of myocardial infarction [hazard ratio (HR) = 1.66; 95% confidence interval (CI) = 1.10–2.50] stroke (HR = 1.67; 95% CI = 1.10–2.56), and total mortality (HR = 1.42; 95% CI = 1.07–1.90). Body mass index, triglyceride levels, cholesterol and treatment with β-receptor blockers at 3 years of follow-up were independent predictors of NOD. Most of the patients had been treated with conventional antihypertensive drugs, such as thiazide diuretics or β-receptor blockers, in mono- or combination therapy, as first-line choice. Therefore, the increased risk associated with NOD in this observational study is also a reflection of the adverse metabolic profile by the drugs commonly used during the study period. The lesson to be learned is that on-treatment NOD is associated with a serious prognosis for cardiovascular risk and mortality. Because agents that block the RAS system were not commonly used during at least the first part of the follow-up, the question remains whether the detrimental effects on glucose metabolism could have been less pronounced if such drugs were used in combination with thiazide diuretics. However, the combination of a RAS blocker with a β-receptor blocker is less recommended, and the combination of a thiazide diuretic with a β-receptor blocker is even contra-indicated in patients at increased risk of diabetes according to recommendations from the NICE organization (UK) in collaboration with the British Hypertension Society, as presented in a statement from 2006 [19]. The role of thiazide diuretics Should we now abandon conventional drugs such as thiazide diuretics against this background of potential worsening of impaired glucose metabolism? Personally, I do not think so, based on evidence. Diuretics are often very effective antihypertensive agents, also in established diabetes, in which the total body sodium load is increased and the extracellular fluid volume expanded. However, diuretics that increase urinary potassium and magnesium losses can worsen hyperglycaemia because insulin secretion is impaired by potassium depletion, and insulin sensitivity in peripheral tissues may also be decreased [5,20]. The use of high-dose thiazide diuretics is reported to increase the risk of non-diabetic hypertensive patients developing diabetes by up to three-fold [21]. This does not seem to occur to the same degree with the use of low dosages [22]. Potassium depletion is particularly severe with high-dose chlorthalidone, less with furosemide and bendrofluazide, and apparently negligible with indapamide. Thiazides may also aggravate dyslipidaemia (high triglycerides, low high-density lipoprotein-cholesterol) [23], although low dosages probably carry a low risk. These drugs have also been associated with a risk of gout and impotence and are therefore generally avoided in middle-aged men at risk for hyperuricaemia or erectile dysfunction. Diuretics have been shown to successfully prevent cardiovascular disease in elderly subjects with established type-2 diabetes and systolic hypertension [24]. However, one observational study suggested that the use of diuretics increased cardiovascular mortality in hypertensive type-2 diabetics who were still hyperglycaemic in spite of treatment [25]. Overall, these drugs are effective and safe when used appropriately at low dosage in many patients, including diabetic patients, and sometimes in combination with potassium supplements or potassium-sparing drugs such as amiloride. If ineffective, diuretics should be combined with another first-line drug (e.g. an ACE inhibitor or an angiotensin-2 receptor antagonist), rather than given at increased dosage. Spironolactone is normally not combined with an ACE inhibitor because this increases the risk of hyperkalaemia. Furosemide is useful in patients with renal impairment (serum creatinine > 150 μmol/l) or oedema. The debated role of β-adrenergic receptor blocking agents β-receptor blockers may significantly lower blood pressure levels in diabetic patients with hypertension, even though renin release (a major target for these drugs) is commonly reduced in diabetes because of sodium and fluid retention. However, these drugs are often ineffective in patients of African descent who commonly have low-renin hypertension. Another mechanism of action is to reduce blood pressure, heart rate and cardiac output via interference with β1 and β2 receptors in the myocardium and in the vessel wall. Similar to diuretics, β-receptor blockers may aggravate both hyperglycaemia and dyslipidaemia [6,16]. These effects depend on both dosage and the degree of selectivity of the individual drug. The hyperglycaemic effect is attributed to inhibition of β2-adrenergic-mediated insulin release and decreased insulin action in peripheral tissues. The long-term risks of a non-diabetic person developing the disease may be increased by six-fold [26] and even more if given together with thiazides; thus, this combination is not recommended [19]. The metabolic side-effects of β-blockers can be reduced by using low dosages combined with other agents, particularly dihydropyridine calcium antagonists, or by intensifying efforts to decrease weight and improve physical activity. β-blockers may also interfere with the counter-regulatory effects of catecholamines released during hypoglycaemia, thereby blunting manifestations such as tachycardia and tremor and delaying recovery from hypoglycaemia in patients with diabetes [27]. β-blockers may also aggravate erectile dysfunction, and are generally contra-indicated in second- or third-degree atrio-ventricular heart block, severe peripheral vascular disease, asthma and chronic airway obstruction. However, it should be remembered that, in the UKPDS, clinical effects of atenolol were comparable to that of the ACE inhibitor, captopril [28]. Both non-selective and selective β-blockers are effective in the secondary prevention of myocardial infarction after an initial event in diabetic patients [29]. β-blockers in general are useful in patients who also have angina or tachy-dysrhythmias and, as such, adverse metabolic effects have to be weighed against these beneficial effects in selected patients. In summary, conventional anti-hypertensive drugs such as thiazide diuretics and β-receptor blockers can increase the risk of NOD, especially if used at higher dosages and in combination when given to high-risk patients with, for example, the metabolic syndrome or pre-diabetes [16]. This is associated with an increased risk of cardiovascular events in several observational studies, including the recent one from Gothenburg, Sweden [17]. Because the risk is likely to increase with a longer duration of treatment, these aspects are of considerable importance for the middle-aged patient about to start treatment for hypertension, and less important in the elderly hypertensive for whom haemodynamic control of for example systolic hypertension is of greater importance for stroke protection than metabolic side-effects. The use of lower dosages and combination therapy with agents that are neutral or beneficial for glucose metabolism might substantially decrease the risk of metabolic disturbances. This is a strong argument for the use of a combination treatment, including a low-dose thiazide diuretic and an agent that blocks the renin–angiotensin system. When the ultimate goal is to prevent arterial ageing [30,31] and consequent cardiovascular events, the implementation of lifestyle advice and a careful selection of antihypertensive drugs tailored to the need of the individual patient is mandatory. Metabolic abnormalities, which are so often associated with essential hypertension in itself, should not be worsened by antihypertensive drug therapy. Hyperglycaemia, irrespective of its origin, should be avoided or counteracted whenever possible. The choice of drugs that are less likely to induce NOD than other drugs is wise during certain circumstances, keeping in mind that type-2 diabetes cannot often be totally prevented, but merely postponed for a few years in susceptible individuals [12–14]. However, this is beneficial in the sense that the burden of hyperglycaemia can be lessened and thus also the risk for adverse outcomes.