Metabolic effects of antihypertensive treatment

Thiazide diuretics and β-adrenergic blockers, alone or in combination, have long been suspected of having an unfavourable metabolic impact: deterioration of glucose tolerance and insulin sensitivity, worsening of the serum lipid profile and potassium depletion are the chief imputations [1]. By contrast, in many clinical trials [2–6], these drugs improved the cardiovascular outcome of hypertensive patients, although it has been noted that the cardiac protection they offered was less than what could be expected on the basis of the known epidemiology [7]. Both thiazide diuretics and β-adrenergic blockers are efficacious, safe and cheap. More recently developed antihypertensive agents, such as calcium-channel blockers, angiotensin-converting enzyme inhibitors (ACE-I) and angiotensin II receptor blockers (ARB), generally have a better metabolic profile than thiazide diuretics or β-adrenergic blockers: essentially neutral or even beneficial depending on the evidence one favours [8–15]. These newer drugs, which are proving their non-inferiority to the older ones in outcomes trials [2–6,13], are efficacious and safe, but expensive. The question then arises: should the new kids on the block replace the old cronies as first choice in the pharmacological treatment of hypertension? A clear-cut answer, free of the opposing pressures of budget and marketing, depends on several subordinate questions: (i) Is the untoward metabolic profile of thiazide diuretics and β-adrenergic blockers conclusively proven? (ii) What is the size of these undesirable effects? (iii) What is their clinical relevance in which category of patients? (iv) How compelling are the cost considerations? (v) Can a general recommendation be made? In this issue of the journal, Lindholm et al. [16] provide useful additional data to approach the issue. In a head-to-head comparison, an ARB (candesartan) combined (in 71% of the cases) with a calcium antagonist (felodipine) was better than a low-dose thiazide diuretic (hydrochlorothiazide) combined (in 84% of the cases) with a β-adrenergic blocker (atenolol) on serum glucose, insulin, lipid, potassium and uric acid levels in a cohort of hypertensive patients followed up for 1 year. In particular, for comparable antihypertensive efficacy (−21/−13 versus −23/−13 mmHg, for the ‘new’ and ‘old’ regimen, respectively), the diuretic + β-blocker pair raised plasma glucose by 0.12–0.34 mmol/l (95% confidence intervals of the between-treatment difference), fasting plasma insulin by 0.6–2.9 mU/l, 2-h post-glucose plasma insulin by 14–47 mU/l, serum low-density lipoprotein (LDL)-cholesterol by 0.06–0.31 mmol/l, serum triglycerides by 0.22–0.50 mmol/l, serum apolipoprotein B by 2.8–14.0 mg/100 ml and serum uric acid by approximately 20%; at the same time, serum high-density lipoprotein (HDL)-cholesterol was lowered by 0.02–0.10 mmol/l and serum potassium by approximately 10%. Moreover, incident diabetes was more frequent with the ‘old’ than the ‘new’ combination (4.0 versus 0.5% per year). The strengths of this study are: (i) the patients were, in vast majority of cases, newly detected and drug-naïve; (ii) the clinical phenotype of the patients was of the most common variety: middle-aged men and women, overweight, probably rather sedentary, moderately dyslipidaemic but non-diabetic (by selection) and at low absolute cardiovascular risk (by inference); (iii) titration of antihypertensive therapy was structured in the protocol and conformed to standard clinical practice; (iv) the follow-up was long enough for metabolic endpoints; and (v) the results are statistically significant and internally coherent. Limitations are the relatively small size and ethnic homogeneity of the patient cohort and the lack of longer-term observation with surrogate endpoints (e.g. changes in carotid artery intima–medial thickness). On balance, this study does provide an answer to the first subordinate question: thiazide diuretics, even at low doses, and β-blockers, also in the cardioselective version, clearly cause a worsening of the metabolic profile in the context of conventional clinical usage. Having proven this concept, the next question is are these metabolic changes clinically significant? It could be argued that the rise in LDL cholesterol and triglycerides, and the attendant fall in HDL cholesterol, are quite small; in a hypertensive population, statins and/or fibrates are likely to be used liberally to correct the dyslipidaemia, particularly if cardiovascular risk is high [5,17]. Furthermore, the association of thiazides with potassium-sparing agents, or the use of gradually acting diuretics with a substantially neutral effect on potassium balance, can prevent or limit potassium depletion and allopurinol can effectively control the hyperuricaemia. This argument would be supported by calling upon those studies [15,18–22] showing that, in high-risk patients, cardiovascular (and renal) damage is driven by advanced hypertension and prognosis is essentially dependent on effective, durable blood pressure control. Finally, subsets of hypertensive patients, such as those with evidence or history of an acute coronary syndrome or those with a pronounced volume component to their hypertension, benefit from the use of diuretics + β-blockers above and beyond their effect on blood pressure. The counter-argument is based upon two main lines of reasoning. First, deterioration of glucose tolerance to diabetes is not a gradual upward drift in glucose levels but an abrupt [23] and, very often, irreversible phenomenon. Prepared by long-standing insulin resistance, diabetes onset is due to a precipitous failure of β-cell function [23]. Insulin resistance augments the insulin demand exponentially [24] and both hypokalaemia and β-blockade [25] interfere with glucose-stimulated insulin release. Once diabetes, or even just impaired glucose tolerance (IGT), sets in (indeed, throughout a long pre-diabetic phase), cardiovascular risk increases three- to five-fold, especially in women [26]. Whether because diabetes progresses inexorably or because antidiabetic treatment is scarcely effective, morbidity, mortality and costs all escalate, and quality of life declines, with the appearance of diabetic micro- and macroangiopathy [15,27,28]. In the study by Lindholm et al. [16], the increase in fasting glucose levels in the patients on thiazide + β-blockers was deceptively small; however, the incidence rate of diabetes rose strikingly (although the actual numbers were necessarily small) and only post-OGTT plasma insulin and glucose levels (measured in a subgroup of randomized patients) signalled the impending small-scale epidemic. At the population level, the use of antihypertensive drugs with a documented diabetogenic potential is hardly a wise choice: treating half the hypertensive patients (approximately 20% of the population) exposes twice as many subjects as those with manifest diabetes (approximately 5% of the population). Finally, recent studies have reported that hypertensive patients treated with ACE-I or ARB have a lower incidence of diabetes than patients treated with diuretics and/or β-blockers [2,14,29,30]. Whether this protection is due to potassium sparing or some other pleiotropic effect of ACE-I remains to be determined. In any event, the extra cost of employing expensive antihypertensive drugs should be weighed against the financial burden that diabetes and its complications impose on health care systems. Second, and somewhat more general, is the issue of the metabolic syndrome (or insulin resistance syndrome, depending on one's views on its aetiology). The founding concept underlying the various definitions of the metabolic syndrome [31] is that a cluster of subclinical abnormalities carries a prognostic value larger than the sum of the values of its individual components: multiple coexisting risk factors enhance atherogenic risk in a synergistic rather than additive fashion [31]. By way of example, classical cardiovascular risk factors (e.g. the Framingham model), although present in high concentration in the diabetic population, fall short of accounting for their high event rate in every epidemiological study ever undertaken [32–34]. Even without attempting to analyse the intricacies of the pathophysiology of the metabolic syndrome, a syndrome with strongly interacting components is almost a logical necessity if it is considered that known risk factors invariably explain a relatively small fraction of cardiovascular events [35]. If the syndrome is accepted or assumed in this or that configuration, it follows that marked risk reduction can be expected from a joint attack on several, if not all, risk factors. By the same token, improving one risk (e.g. high blood pressure) at the cost of worsening another (e.g. glucose or lipids) may cancel or weaken the anticipated benefit of the intervention. How often and in what measure this actually happens in patients is difficult, if not impossible, to ascertain. The play of interactive factors is a statistical nightmare: interactions increase exponentially with the number of factors. Nevertheless, suggestive examples of simple synergism are the observations that the efficacy of pharmacological antihypertensive treatment is reduced in insulin resistant subjects [28] and potentiated by simultaneous weight loss [36] or increased physical activity. Having provided some answers to the subordinate questions, we can now return to the principal question: should newer drugs replace old ones as first-line agents in the treatment of hypertension? Recommendations are easier to give than to take. Furthermore, they generalize and, in doing so, inevitably take on a fundamentalistic overtone. Analysing and meta-analysing the myriad of clinical trials is a stimulating intellectual exercise, which occasionally can drift away from real life. We therefore only submit the minimal set of conclusions that we have drawn from the study by Lindholm et al. [16] in the context of our reading of the literature and our own biases. Severe hypertension in any patient and moderate hypertension in patients with evidence of cardiovascular or renal disease command aggressive anti-hypertensive treatment (almost invariably with drug combinations) regardless of metabolic side-effects and cost. In all other cases, including mild or recent-onset hypertension, it is a good idea to search for the metabolic syndrome [37] and, if present, to manage the whole of it, giving first choice (especially in diabetes and lesser degrees of glucose intolerance) to newer drugs to reduce the blood pressure, regardless of cost. Needless to say, newer agents would more easily replace older drugs (and the controversy fade into irrelevance) if they cost less. Obviously, however, no recommendation can (and should) replace clinical judgement for an individual patient.