Drug Safety 7 (5): 364-373, 1992 0114-5916/92/0009-0364/$05.00/0 © Adis International Limited. All rights reserved. ORS1 99 Adverse Effects of Diuretics Edward D. Freis Veterans Administration and Georgetown University Medical School, Washington, D.C., USA Contents Summary 1. Do Thiazide Diuretics Cause Myocardial Infarction or Sudden Death? 2. Thiazide Diuretics and Electrolyte Balance 2.1 Potassium 2.2 Magnesium . Effect of Thiazide Diuretics on Monitored Cardiac Arrhythmic Activity . Diuretics in the Presence of Acute Myocardial Infarction . Diuretics and Cholesterol . Diuretics and Blood Glucose Control 6.1 Hyperinsulinism and Insulin Resistance . Hyperuricaemia and Gout . Quality of Life - Sexual Function 9. Conclusions and Clinical Implications Ra & Ww oO Summary Analysis of the available evidence indicates that diuretics do not increase coronary heart dis- ease morbidity and mortality. The multiclinic trials supporting the cardiotoxicity hypothesis are few in number and flawed in design. The majority of the trials, including the well designed trials, indicate no excess of coronary heart disease (CHD) events in diuretic-treated patients compared with those given other drugs or placebo. Recent studies indicate no increase in cardiac arrhythmias after diuretic treatment. Also, al- though depletion of intracellular potassium and magnesium occurs in patients with congestive heart failure even without diuretics, intracellular concentration of these ions is not significantly reduced by diuretics in patients with uncomplicated hypertension. Modest elevations of serum cholesterol may occur during the first 6 to 12 months of treatment with thiazide diuretics. However, after this time these elevations fall to or below the pretreatment level. The fall may be greater in patients receiving other drugs but the differences are small and their clinical significance is questionable. The incidences of hyperglycaemia and diabetes were only minimally increased in long term clinical trials while the importance of hyperinsulinism and insulin resistance in causing CHD remains unproven in patients. Thiazides remain, therefore, a safe and effective treatment for patients with hypertension. Adverse Effects of Diuretics 365 Few drugs in common use today have been ac- cused of more serious adverse effects than the thia- zides and related diuretics. The allegations include increased risk of myocardial infarction and sudden death due to either associated hypokalaemia, hy- pomagnesaemia, hypercholesterolaemia or hyper- insulinism. Additional reactions include hypergly- caemia with aggravation of diabetes as well as the nonketotic, hyperglycaemic syndrome. Hyperuri- caemia is common but seldom leads to gout. In addition, diuretics have been associated with ad- verse effects on quality of life including impotence, fatigue and weakness. This review assesses the va- lidity of these alleged reactions. 1. De Thiazide Diuretics Cause Myocardial Infarction or Sudden Death? The multiclinic trials have shown a striking re- duction in stroke but with few exceptions they have shown little benefit in preventing the complica- tions of coronary heart disease (CHD) [table I]. It has been proposed that the difference might be due to adverse effects of the thiazides which were used as 1 treatment arm in the major clinical trials. It is widely believed that the above hypothesis has gained support from the Multiple Risk Factor Prevention Trial (Multiple Risk Factor Interven- tion Trial Research Group 1985b). The patients were randomly assigned to 2 groups: the special intervention (SI) group treated in hypertension clinics in large teaching hospitals, and the usual care (UC) group who were referred to their usual community care facilities. There was no essential difference in CHD mortality in the 2 groups (table I). However, subgroup analysis of the patients who had minor baseline resting electrocardiogram (ECG) abnormalities revealed insignificantly greater CHD mortality in the SI than in the UC patients. Could this be due to thiazide treatment which was given more frequently to the SI patients? The hypothesis seems unlikely for the following reasons: (a) another retrospective analysis from the same study (Multiple Risk Factor Intervention Trial Research Group 1985a) in patients who had ab- normal baseline exercise ECGs indicated a 57% higher CHD mortality in the UC patients than in the SI group (p = 0.002). This was directly contra- dictory to the previous subgroup analysis; (b) the results of nonrandomised subgroup analysis are known to be unreliable, as was shown by these 2 opposing results; (c) the trial was not a clear com- parison of diuretics versus placebo or other drugs, as only 56% of the SI group received diuretics while 33% of the UC group received similar treatment, over half of whom were given rather large doses; (d) the results were skewed in the UC group with minor resting ECG abnormalities in that they ex- hibited an abnormally low CHD mortality, lower than the UC group without such baseline abnor- malities; (e) there were no correlations between CHD mortality and hypokalaemia or with size of the diuretic dose. Also, the more reliable total ran- domised trial showed no difference in CHD mor- tality between the SI and UC groups. The above evidence, therefore, provides no convincing sup- port for the existence of cardiotoxicity due to thia- zides. The Metoprolol Atherosclerosis Prevention in Hypertension Study (MAPHY trial) found more CHD mortality with thiazides than with metopro- lol (Wilkstrand et al. 1988). However, the MAPHY trial was a subgroup of the Heart Attack Primary Prevention in Hypertension trial (HAPPHY trial). The parent HAPPHY study reported no signifi- cant difference between the diuretic- and £- blocker-treated groups in CHD mortality (Wil- helmsen et al. 1987). The other subgroup that made up the HAPPHY trial (receiving atenolol), there- fore, must have had fewer CHD deaths with thia- zides than with @-blockers. The Oslo trial (Holme et al. 1984) indicated a higher CHD death rate with thiazides, but this study was too small to draw any valid conclusions (6 deaths per 1000 patient years in thiazide recipients versus 2 in the placebo group). These observations indicate that the evidence for diuretic-induced CHD mortality derived from only a minority of the clinical trials is weak and contains many contradictions. On the other hand, there were 9 other therapeutic trials which found quite different results (table I). The largest and one 366 Drug Safety 7 (5) 1992 Table |. Coronary heart disease (CHD) mortality and morbidity with thiazides versus placebo or other drugs (from Freis 1989, published with permission from the American Medical Association) Trial No. of Diuretic and dose All CHD events Fatal CHD events patients (mg) — thiazide other thiazide other Trials associating thiazides with increased CHD risk MRFIT (Multiple Risk Factor . 2 4788 Chiorthalidone; 29,2 17.7 Intervention Trial Research Group hydrochlorothiazide 50 1985b) Oslo trial (Holme et al. 1984) 747 Hydrochlorothiazide 50 20° 13 6° 2 MAPHY (Wilkstrand et al. 1988) 3 234 Hydrochlorothiazide 50-100; 434 36 bendroflumethiazide 5-10 Trials indicating thiazides do not increase CHD risk EWPHE (Amery et al. 1985) 840 Hydrochlorothiazide 25-50; tac 23 triamterene 50-100 MRC (Medical Research Working 17 354 Bendroflumethiazide 10 §.2¢ 6.5 2.5¢ 2.3 Party on Mild Hypertension 1985) Veterans Administration (Veterans 380 Hydrochlorothiazide 100 118 13 6° 11 Administration Cooperative Study Group on Antihypertensive Agents 1970) Australian trial (Report by the 3 427 Chlorothiazide 500-1000 70° 88 2e 8 Management Committee 1980) Public Heaith Service (Smith 1977) 785 Hydrochlorothiazide 50 ge 7 2e 2 HDFP (Williams et al. 1986) 10 940f Chlorthalidone 25-100 1314 1489 HAPPHY (Wilhelmsen et al. 1987) 6 569 Bendroflumethiazide § 9.5¢ 10.6 41 4.4 hydrochlorothiazide 50 IPPPSH9 (The IPPPSH 6 372 Collaborative Group 1985) MPPCD? (Miettiner et al. 1985) 1 203 Hydrochlorothiazide 50 a Subgroup with baseline ECG abnormality among 12 888 total patients. b Per 1000 patients. c Per 1000 person-years of observation. d_ Total deaths due to myocardial infarction plus other ischaemic heart disease. @ Number of events. f Total patients. See text for analysis of subgroups with and without resting ECG abnormalities. g Randomised to receive placebo or oxprenolol, but diuretics were soon added in 67% of the oxprenolol group and 82% of the placebo group. Numerical data were not given but no significant association found between diuretic usage and cardiovascular events. h_ Coronary events tended to be accumulated in subgroups treated with £-blocking agents or clofibrate but there were few in those receiving probucol or diuretics. Abbreviations: MRFIT = Multiple Risk Factor Intervention Trial; ECG = electrocardiogram; MAPHY = Metoprolol Atherosclerosis Prevention in Hypertension study (a subgroup of the Heart Attack Primary Prevention in Hypertension [HAPPHY] trial); EWPHE = European Working Party on High Blood Pressure in the Elderly; MRC = Medical Research Council of Great Britain; HDFP = Hypertension Detection and Follow-up Program; IPPPSH = International Prospective Primary Prevention Study in Hypertension: MPPCD = Multifactorial Primary Prevention Trial of Cardiovascular Disease in Middie-Aged Men. of the best controlled trials was the Medical Re- 1985). Propranolol, thiazide or placebo were ran- search Council (MRC) trial of Great Britain (Medi- domly assigned. There was no difference in CHD cal Research Working Party on Mild Hypertension events in the thiazide recipients compared with the Adverse Effects of Diuretics 367 propranolol or placebo group. However, stroke was reduced by 69% in the thiazide group and by 27% in the propranolol-treated patients compared with the placebo group. Although a subgroup of non- smokers showed fewer CHD deaths in the pro- pranolol as compared with the thiazide patients, the HAPPHY trial found no’ such difference be- tween @-blockers and thiazide among nonsmokers. Another well controlled trial, the Australian trial (Report of the Management Committee 1980), noted a trend toward a lower incidence of nonfatal CHD events in the thiazide recipients as compared to those receiving placebo. The results of these 9 trials (table I) indicated no significant difference in CHD morbidity and mortality between thiazide re- cipients and those receiving placebo or other drugs. Therefore, the majority of the trials including those which were best controlled outweigh the contrary evidence supplied by the few flawed trials. An ad- ditional recent study, the large SHEP trial in el- derly patients, with isolated systolic hypertension (SHEP Cooperative Research Group 1991) re- ported a smaller number of CHD events in di- uretic-treated patients compared with placebo. The questions of hypokalaemia, hypomagnesaemia, elevations of cholesterol with thiazides and insulin resistance are discussed later but it seems evident that if thiazides increase CHD risk from any cause then a higher incidence of CHD events with thia- zides should have been manifested in the majority of the clinical trials. 2. Thiazide Diuretics and Electrolyte Balance 2.1 Potassium The intracellular concentration of potassium is many times greater than the extracellular content. This gradient is maintained by a Nat,K+-ATPase metabolic pump which actively extrudes sodium from the cells but retains potassium. Therefore, ex- tracellular potassium concentrations bear little re- lation to the intracellular content. Sodium and po- tassium loss in the urine following continuous thiazide administration increases during the first 2 or 3 days but then comes back into balance with intake, thus preventing excessive body losses dur- ing continuous treatment (Papademetriou 1984). A review of many studies indicated that after long term administration of thiazide diuretics the re- duction of intracellular potassium approximates only 5%, which is physiologically unimportant (Kassirer & Harrington 1977). The Nernst equation expresses the electrical po- tential across the cell membrane. The transmem- brane voltage is related to the ratio of various ions inside to outside the cell. A reduction in the ratio of potassium outside to that inside the cell (such as occurs with thiazide diuretics) increases the ne- gativity of the resting membrane potential which stabilises or reduces its electrical excitability (Guy- ton 1981). Thus, predominantly extracellular hy- pokalaemia should reduce rather than increase cardiac arrhythmic activity. 2.2 Magnesium Thiazide-induced magnesium deficiency had been widely considered to be a cause of cardiac arrhythmias. However, interest in this hypothesis has declined in recent years because of the follow- ing considerations. The influence of diuretics on intracellular mag- nesium and potassium is controversial. Some in- vestigators have found only negligible reductions (Araoye et al. 1978; Bergstrom et al. 1973) while others have reported a deficit of both ions (Dyck- ner & Webster 1979; Lim & Jacob 1972). As in- dicated in the previous section, biologically insig- nificant changes were found in intracellular potassium with thiazide treatment of patients with uncomplicated hypertension (Kassirer & Harring- ton 1977) and intracellular magnesium concentra- tions paralleled potassium concentrations (Lim & Jacob 1972). Therefore, as is the case with potas- sium, only small reductions in intracellular mag- nesium would be expected with thiazide adminis- tration. Also, thiazides should not have much influence on magnesium excretion because these diuretics act in the early distal tubules while mag- 368 Drug Safety 7 (5) 1992 nesium is absorbed predominantly in the loop of Henle (Ryan 1986). The controversial findings of different investi- gators on reduction in potassium and magnesium is probably the result of patient selection. The stud- ies which found deficiencies of these ions were conducted in patients who mostly had congestive heart failure or myocardial infarction complicated by heart failure (Dyckner & Webster 1979; Lim & Jacob 1972). It has been known for many years that patients with congestive heart failure have significant def- icits of intracellular potassium (and, therefore, magnesium) even in the absence of diuretics. Iseri et al. (1952) and Cort and Mathews (1954) found marked reductions in potassium in muscle tissue taken by biopsy in patients with congestive heart failure and severe myocardial infarction in the era before thiazides were available. Treatment with magnesium- and potassium- sparing diuretics may possibly have a place in the treatment of congestive heart failure, although evi- dence from well controlled studies is lacking. It is also possible that the use of thiazides in combi- nation with potassium-sparing diuretic may be helpful in elderly diabetics whose diet may be po- tassium poor. However, in patients with uncom- plicated hypertension there does not appear to be any indication for routine replacement therapy. 3. Effect of Thiazide Diuretics on Monitored Cardiac Arrhythmic Activity Early studies (Holland et al. 1981; Hollifield & Slaton 1981) indicating increased cardiac arrhyth- mias following thiazide diuretics have been refuted by subsequent better-designed trials. Papademe- triou et al. (1983) selected 16 hypertensive patients receiving hydrochlorothiazide 50mg twice daily whose serum potassium levels averaged 2.8 mmol/ L. The ECG was monitored for 24h following which hypokalaemia was normalised by appropriate treatment for 4 weeks. There were no significant ECG differences between the hypokalaemic and the normokalaemic periods, including couplets and runs of ventricular tachycardia. The lack of effect of thiazides on cardiac arrhythmias was confirmed by other investigators (Lief et al. 1984; Madias et al. 1984). Papademetriou et al. (1985) also found that thiazides caused no increase in arrhythmias in thiazide-treated patients with hypertension and left ventricular hypertrophy. 4. Diuretics in the Presence of Acute Myocardial Infarction The stress of an acute myocardial infarction is often associated with increased levels of plasma catecholamines (Struthers et al. 1983). Because catecholamines reduce the levels of plasma potas- sium, severe myocardial infarction is often accom- panied by hypokalaemia, even in the absence of thiazide diuretics. The question has been raised that thiazides may increase the risk of fatal arrhythmias by aggravating the hypokalaemia. This hypothesis is not supported, however, by the multiclinic trials. The majority of these trials have found that the incidence of fatal myocardial infarction is no greater in thiazide recipients than in patients receiving other regimens (table I). Mor- tality is, however, related directly to the level of circulating catecholamines (Karlsberg et al. 1981). These results suggest that the arrhythmias are not due to the hypokalaemia per se, but rather to the direct arrhythmogenic effects of catecholamines on the heart in the setting of myocardial cell injury, tissue ischaemia and other factors related to the myocardial insult. Nordrehaug et al. (1985) found that catecholamines were raised in the initial pe- riod following an infarction. Plasma potassium levels were inversely related to ventricular tachy- cardia during this early period. However, patients receiving diuretics at this time did not show a sig- nificant increase in ventricular arrhythmias. These investigators also showed that 8 hours following the infarction, when catecholamine levels had fallen, the degree of hypokalaemia also was not re- lated to the incidence of cardiac arrhythmias, again suggesting that other factors were the cause of the arrhythmias. Another reason for doubting the importance of hypokalaemia in worsening the course of acute Adverse Effects of Diuretics 369 myocardial infarction is the negative results ob- tained with potassium replacement therapy (Fletcher et al. 1968; Rogers et al. 1979). ‘Polar- izing solutions’ have been used in the past which contained potassium, glucose and insulin. Intra- venous infusion of these solutions in sufficient amounts to raise the plasma potassium level to normal failed to prevent ventricular arrhythmias or reduce mortality in patients with acute myo- cardial infarction. 5. Diuretics and Cholesterol Numerous investigators have demonstrated an increase in serum cholesterol during the first few months of treatment with thiazide diuretics (Ames & Hill 1976; Schoenfeld & Goldberger 1964). The elevations were small, averaging less than 0.28 mmol/L. Long term studies, usually longer than | year, have shown either no change in cholesterol or a slight decrease compared with baseline (table IJ). Other long term studies have shown a greater drop in serum cholesterol with other drugs than with thiazides (Lasser et al. 1984; Williams et al. 1986). Again, the differences were small, averaging 0.10 to 0.18 mmol/L with the other drugs as compared to thiazides and were probably clinically unimpor- tant. It should also be noted that the various clinical trials used considerably larger doses of diuretics than are used today. Cholesterol levels may be fur- ther reduced with smaller doses during long term treatment. 6. Diuretics and Blood Glucose Control It has been recognised for many years that thia- zide diuretics may raise blood glucose but the in- cidence, severity and duration of the rise have been unclear. The best evidence should be found in long term clinical trials. After 5 years of diuretic treat- ment the MPPCD (Helsinki) trial found a mean rise 1h after a glucose load of only 0.3 mmol/L from baseline as compared to 0.7 mmol/L in the control group (Miettiner et al. 1985). The HAPPHY trial involving approximately 6000 patients reported an incidence of diabetes over a 45-month period of 6.1 per 1000 patient-years in diuretic recipients compared with 6.9 per 1000 patient-years in those on 6-blockers (Wilhelmsen et al. 1987). In the MRC trial there were no significant changes in casual blood glucose levels, although there was a slight increase at 1 year followed by a fall after 2 years (Medical Research Working Party on Mild Hypertension 1977). On the other hand, the HDFP trial reported an incidence of 1.6% of diabetes or hyperglycaemia in chlorthalidone re- cipients compared with 0.1% in those receiving re- serpine and none in those receiving methyldopa (Williams et al. 1986). A 10-year controlled trial indicated that low doses of thiazides are not diabetogenic (Berglund et al. 1986). Thus, aside from HDFP and consid- ering the natural increase of diabetes in middle- aged to elderly patients, the reported increased in- cidence of diabetes with thiazides is low and in some studies nonexistent (Berglund et al. 1986). The hyperosmolar nonketotic syndrome is char- acterised by an acute, severe elevation of blood glucose occurring without ketosis. It is not limited to diabetic patients. It usually occurs in patients receiving diuretics (Curtis et al. 1972). These patients often develop severe dehydration and oc- casionally coma. The syndrome is manifested by polyuria and markedly elevated levels of blood glu- cose. Rapid rehydration is indicated. It is often possible to transfer the patient to a chemically dif- ferent diuretic such as chlorthalidone in place of a thiazide or vice-versa without recurrence of the hyperglycaemic state. 6.1 Hyperinsulinism and Insulin Resistance The association between hypertension and dia- betes represents a major health problem affecting 2.5 million Americans (Working Party on Hyper- tension in Diabetes 1987). The presence of hyper- tension in diabetes greatly increases the risk of both CHD and nephropathy. Nephropathy but not CHD has been favourably influenced by reduction of blood pressure (Christlieb 1982; Mogensen 1989). 370 Drug Safety 7 (5) 1992 Table il. Changes in serum cholesterot level with short term versus long term treatment of hypertension with diuretics (from Freis 1989, published with permission from the American Medical Association) Trial No. of Diuretic and dose Duration of Mean change in patients (mg) treatment cholesterol (mmol/L) Trials indicating a rise in cholesterol Ames and Hill (1976) 74 Chlorthalidone 25-100 1-3 months +0.28 VA-NHLBI (Goldman et al. 1980) 302 Chiorthalidone 50-100 1 year +0.26 Grimm et al. (1981) 57 Chiorthalidone 100 or 1.5-3 months +0.18 hydrochlorothiazide 100 Studies indicating no change or a fall in cholesterol EWPHE (Amery et al. 1985) 190 Hydrochlorothiazide 25-50 2 years -0.52 Framingham study (Kannel et al. 288 Thiazides 2 years ~0.16 1977) MRC (Medical Research Working 17 354 Bendroflumethiazide 10 3 years +0.054 Party on Mild Hypertension 1985) MRFIT (Multiple Risk Factor 10215 Chlorthalidone or 6 years ~0.23 Intervention Trial Research Group hydrochlorothiazide 50-100 1985b) MPPCD (Miettiner et al. 1985) 1 203 Hydrochiorothiazide 50 5 years 0 Oslo triat (Holme et al. 1984) 300 Hydrochlorothiazide 50 3 years 0 HAPPHY (Wilhelmsen et al. 1987) 6 669 Bendroflumethiazide 5 or 1 year 9 hydrochlorothiazide 50 Trials indicating an early rise followed by a fall below baseline VA propranolol- 147 Hydrochlorothiazide 50-200 2 months +0.16 hydrochtorothiazide (Veterans 1 year —-0.08 Administration Cooperative Study Group on Antihypertensive Agents 1982a,b) Alcazar et al. (1982) 236 Hydrochlorothiazide 50-100 1-3 months >oe 1-2 years