European Heart Journal (1992) 13 {Supplement G), 72-80
Diuretics and cardiovascular risk factors M. MOSER
Yale University School of Medicine, New Haven, Connecticut, U.S.A. KEY WORDS: Hypertension, diuretics, cardiovascular risk factors, clinical trials. Thiazide diuretics are efficacious, either as monotherapy or in combination with other antihypertensive drugs. They reduce blood pressure in a high percentage of hypertensive patients with minimal subjective side effects. There is increasing evidence that the use of diuretics, singly or in combination, will reduce morbidity and mortality associated with essential hypertension in both young and elderly subjects. Although diuretics may induce some changes in the plasma lipid profile, serum uric acid concentration and glucose metabolism, there is little evidence that these changes are of clinical significance. The increase in serum cholesterol concentration has rarely persisted in any trial beyond the first year of treatment. The incidence of diabetes mellitus in diuretic treated subjects is only about ]%, even when large doses are used. Gout may be precipitated in susceptible subjects, but is uncommon. For these reasons, diuretics should remain a preferredfirst-stepdrug of choice in the management of hypertension. Introduction Debate continues regarding the specific role of diuretic therapy in the management of hypertension despite everincreasing evidence that the use of these agents reduces morbidity and mortality in hypertensive patients1'"31. High doses of thiazide diuretics result in an increase in serum uric acid levels, a decrease in serum potassium, a shortterm increase in cholesterol and low density lipoprotein (LDL) levels, and probably a long-term increase in serum triglyceride levels'41. In addition, these medications increase insulin resistance in patients with hypertension who may have preexisting glucose tolerance and insulin utilization abnormalities15'61. But are these effects of clinical importance? Do they increase cardiovascular risk or minimize or even negate the benefits of blood pressure lowering? A recent meta analysis171 and several new studies in the elderly18"10' have cast doubt on the argument that coronary heart disease events (CHD) have not been reduced when diuretics are used as initial therapy in hypertensive individuals. In fact, the recent trials present strong evidence of benefit when these medications are used. But there are other questions to be addressed. Are all diuretics equally effective in lowering blood pressure and similar in their effects on electrolytes, glucose metabolism and lipids? Would lower dosages of the commonly used diuretics produce fewer metabolic effects while maintaining their antihypertensive capabilities, and perhaps reduce CHD events still further? Has the use of diuretics decreased CHD events? Approximately 75 000 patients have been studied in various clinical trials, all of which have used diuretics as initial therapy. Some compared diuretics to betablockers'9""1, some were placebo controlled18""131, while others compared a vigorously treated to a less rigorously Correspondence: Marvin Moser, MD, FACP, Davis Avenue Medical Center, 33 Davis Avenue, While Plains, New York 10605, U.S.A. 019S-668X/92/OG0072 + 09 $08.00/0
treated group1'41 or a control group1'51. Either a betablocker, alpha methyl-dopa, reserpine, or hydralazine were used as second or third-step medications in the trials. Most of these studies were not specifically designed to address the question of whether or not CHD mortality was reduced with antihypertensive therapy. Several trials included only patients without evidence of target organ involvement or other complications of vascular disease1"t121. It could have been predicted therefore that the number of cardiovascular events would be small and that the benefits of treatment would be difficult to prove given the short duration of the studies (3-5 years) and the numbers of patients evaluated. For example, in the Australian Study1'21, the placebo group of patients experienced approximately one-third the number of deaths compared to an age- and sex-matched group of the Australian population, making it difficult to demonstrate an even lower mortality in treated subjects. It is, therefore, not surprising that many of the individual hypertension treatment trials failed to show a statistically significant decrease in CHD events in treated compared to control groups. Speculations were advanced suggesting that it was the medication used rather than the trial design that resulted in these findings. Yet, in trials that included elderly subjects18"101 or subjects with evidence of pretreatment target organ involvement1'41 where the numbers of complications were sufficient to demonstrate differences in outcome, a decrease in CHD events has been reported in treated groups. Observational estimates that a decrease of 5—6 mmHg in diastolic blood pressure over a long period of time would decrease stroke deaths by approximately 35-40% were realized in the clinical trials; a 42% decrease was noted in the 3-5 year trials with a decrease in diastolic pressure of this magnitude17161. On the other hand, epidemiological or observational data suggested that a decrease in diastolic pressure of 5-6 mmHg should decrease CHD events by approximately 20-25%. The 14 randomized clinical trials analysed by Collins et al.m prior to the Systolic Hypertension in the Elderly (SHEP/81, © 1992 The European Society of Cardiology
Diuretics and cardiovascular risk factors
400
- (a)
k
«, 200 o "5 o 1-00 --
400 - ( b )
100 -
- „a
CD
| a
050
0-50 -
0-25 -
025 -
Baseline DBP category
•
2 00 -
t
1 1 76
2 1
84
3 1 91
4 1
98
5 1
105
73
Baseline DBP category
1 1
76
D
2
3
1 84
1 91
a
4 1
5
98
105
1
Approximate mean usual DBP (mmHg) Figure 1 Risks of stroke and coronary heart disease (CHD), estimated from combined studies. MacMahon el al. Lancet 1990; 335: 765. (a) Stroke and usual diastolic blood pressure (DBP) (seven prospective observational studies — 83 events), (b) Coronary heart disease and usual diastolic blood pressure (nine prospective observational studies — 4856 events).
• >
the- STOP"1 and the recent MRC1'01 studies showed a based on careful follow-up observations. It is also possstatistically significant decrease ofl 4% in CHD events. In ible, of course, that CHD will not be impacted sufficiently the latest MRC trial in the elderly1101 CHD events were by blood pressure lowering alone; other risk factors may reduced by 44%, strokes by 31%, and overall cardio- have to be reduced. But at present, there are few data to vascular deaths by 29% with low dose diuretic therapy substantiate a negative impact of diuretic therapy on compared to placebo. In the STOP study (using different cardiovascular risk. beta blockers or a diuretic), stroke risk was reduced by 47%, fatal myocardial infarction or sudden death by 50% and overall mortality by 43%. If the SHEP study had Effect of thiazide diuretics on specific cardiovascular risk been included in the recent meta-analysis, the resultant factors decrease in CHD events would have been approximately (1) Reducing blood pressure reduces cardiovascular 16%, if data from the most recent MRC110' study had been risk. The use of diuretics in hypertensive persons will included, this percentage would have been even higher reduce blood pressure to as great or greater degree as any and closer to epidemiological estimates. But to many of the other available antihypertensive agents, including observers, these data continued to suggest a shortfall in beta-blockers, alpha beta-blockers, alpha blockers, ACE benefit and elicited explanations that the drugs used inhibitors, and calcium entry blockers. A loop diuretic (mainly diuretics) may have induced metabolic changes like furosemide, while highly potent as a diuretic, is less that minimized benefit. Logic suggests that the so-called effective as an antihypertensive agent over the longshortfall was more probably related to the short duration term1"1. There are few comparative long-term data of the trials. Epidemiological data had assumed a long available with bumetanide or torasamide, a long-acting term (perhaps over decades) decrease of 5—6 mmHg in loop diuretic. diastolic blood pressure (DBP) to achieve a decrease of (2) The use of thiazide diuretics prevents progression of 20-25% in CHD, not a decrease in pressure for only 3-5 mild to more severe hypertension, thereby reducing the years. potential for cardiovascular complications1201. If it is the short duration of the trials that accounted for (3) Despite some equivocal evidence, lowering of blood the less than expected benefits for CHD, one might ques- pressure with diuretics will result in the prevention of or tion why stroke deaths decreased to the predicted level reversal of left ventricular hypertrophy814-2'1221. Thus, over the 3-5 year period of time of the clinical trials. The a major risk factor for CHD and serious arrhythmias slope of increase in stroke events with increases in blood appears to be positively affected (the effect of diuretics pressures is a steep one (Fig. 1). A minimal decrease in on this important cardiovascular risk factor is, however, pressure can therefore be expected to have a major effect less consistent than with ACE inhibitors or several of on stroke increase. The slope is less dramatic with CHD the beta-blockers, centrally acting agents and calcium events; it may take a longer period of time to decrease antagonists1231). CHD deaths with the same reduction in blood pressure. (4) Opinions differ on the clinical significance of the Some credence is given to this assumption by the 8- and effects of diuretics on lipids and glucose metabolism. The 10-year data from the HDFP"71 and MRFIT"81 studies use of thiazide diuretics over a period of less than one year respectively. CHD events were reduced further in the will result in an increase of about 5-7% in serum choleslonger-term follow-up. These new data, however, are not terol levels with a minimal, if any, change in high density
74
M. Moser
Table I Effect of diuretic-based therapy on serum cholesterol in long-term clinical trials* Total cholesterol level (mg dl~') Trial
Berglund and Anderson MRC trial Men Active treatment Placebo Women Active treatment Placebo MAPHY study HDFP SC group Oslo Active treatment Control MRFIT SI group UC group HAPPHY trial EWPHE study Active treatment Placebo MRC in elderly* Active treatment Placebo
Duration (years)
6
Difference Baseline
iMter treatment
267
255
-12
245 244
245 239
0 -5
261 260 244
260 256 243
-1 -4 -1
232
223
-9
272 278
273 280
+1 +2
254 254 242
236 240 242
-18 -14 0
256 259
238 239
-18 -20
228 228
232 232
+4 +4
3+
6 5 4 6 4 3 5
All of the trials used diuretics in doses equivalent to 50 mg or more/day of hydrochlorothiazide. 'Estimated from available data.
lipoprotein (HDL) but an increase in LDL and triglyceride levels'241. Many of the reported studies lacked a control group or included only small numbers of subjects. Elsewhere we have reviewed cholesterol changes in the longterm clinical trials, somewhat different findings emerge'25'. Although none of these trials (in the 1970s and 1980s) was specifically designed to study lipids and lipid fractions, data were collected in large numbers of subjects over a 2-5 year period. Table 1 summarizes the data on cholesterol changes in diuretic-based trials of longer than one year's duration. None showed a definite increase in cholesterol levels when thiazide diuretics were used. In many trials, other agents were used concomitantly; since beta-blockers were used in several of them, it would have been expected that if an adverse effect had occurred, it would have been exaggerated rather than minimized. In one 58-week trial'261, there was an increase of 5% in total cholesterol levels (and 10% in LDLs). However, as the author of this trial states, the increase occurred in subjects with initial low levels; in those with baseline cholesterols of 240-250 mg. dl" 1 no change was noted. This suggests regression to the mean, as noted by other investigators'14-271 rather than a specific medication effect. In the French Study of 20 months' duration'271, high baseline cholesterol levels decreased; low baseline levels increased on a diuretic with a potassium sparing component. Overall there was no change during the study period. The hypertension detection and follow-up programme (HDFP) data are also of interest. In this study, baseline elevated cholesterol levels decreased on thiazide diuretic
therapy while a slight increase in serum cholesterols was noted in subjects with low baseline levels'281 (Fig. 2). This study suggests that thiazide therapy need not be avoided in patients with hyperlipidaemia, as suggested by some physicians'291. In two studies (MRC and MRFIT)"1-301, cholesterol levels remained the same or decreased on thiazide therapy but not to the extent that they had in the placebo or less vigorously treated groups. It was suggested that the degree of lipid lowering (which is noted frequently in any clinical trial) may have been blunted by thiazide therapy. This may have been true in the MRC study where, for example, a difference of 3mg.dl~' between placebo and diuretic therapy in women and 5 mg. dl~' in men was noted. The MRFIT data are confusing and difficult to interpret'31'. This was not only a study of hypertension therapy but one in which individuals could enter the trial with elevated cholesterol levels and/or a smoking history. The patients who experienced the greatest decrease in cholesterol levels were those who lost the most weight and who entered the study with the highest cholesterol levels (Table 2). The non-smokers in MRFIT, who were not on treatment for hypertension at 6 years, were either not hypertensive or only mildly hypertensive at baseline. It can be assumed that they were admitted to the trial primarily because of elevated serum cholesterol concentrations. It is in this group of non-smokers and 'non-treated' subjects that the greatest difference was noted in the decrease in cholesterol levels ( — 23 in non-treated compared to — l l m g . d l " 1 in treated subjects) and in weight loss
Diuretics and cardiovascular risk factors
Baseline
75
2 3 Years of therapy
Figure 2 Effect of diuretic-based therapy on cholesterol levels in hypertension detection and follow-up programme participants'211. D = cholesterol at baseline < 2 2 0 m g . d l - ' (n = 270); • = cholesterol at baseline 221-249mg. d l 1 (n = 207); • = cholesterol at baseline 250-279 mg. dl-' (n = 113); A =cholesterol at baseline > 280 mg . dl -' (n = 74).
Table 2
Effect of diuretic-based therapy on serum cholesterol in MRFIT special intervention subjects Total cholesterol levels (mg. dl"1)
Hypertensive at baseline
Smokers On treatment for hypertension at 6 years Not on treatment for hypertension at 6 years Non-smokers On treatment for hypertension at 6 years Not on treatment for hypertension at 6 years
Change in body weight (kg)
No
Baseline
6 years
Difference
549
225
216
-9
-11
266
234
219
-15
-20
818
246
235
-11
-1-9
293
254
231
-23
-3-9
Adapted from1321.
( — 8-6 in non-treated compared to —4-2 lbs in treated subjects). Baseline cholesterol levels were also higher in the non-treated subjects. A logical explanation for at least some of the difference in outcome might be that the patients themselves, as well as the investigators, expended more effort in trying to correct their weight and their one major risk factor (elevated cholesterol level). HDL levels in the MRFIT study were decreased by less than 1 -0 mg. dl"' in the thiazide treated group compared to the non-treated group — again this result is confounded by the complicated nature of the study and the observation that changes in HDL correlated closely with weight gain or loss — those who gained the most weight experienced a decrease in HDL; those who lost weight showed an increase. None of the observed changes in lipids between the treated and the non-treated groups achieved statistical significance. Some studies have suggested that the thiazide-induced effect on cholesterol levels can be eliminated by a low-fat/ low-cholesterol diet1"1, while others have reported a
decrease in serum cholesterol levels following withdrawal of thiazide diuretics'341; the usually observed weight gain following diuretic withdrawal suggests that haemodilution may explain, at least part, of the chemical changes. Extracellular fluid volume remains contracted to a slight degree during long-term diuretic therapy141. In addition to the MRFIT1301 and MRC1"1 studies, which appear to indicate a thiazide-induced blunting of the beneficial effects of diet, a subset of patients from the HAPPHY study have been reported as demonstrating a long-term (42 months) adverse effect of thiazides on cholesterol levels'351. However, only a very small number of patients had been followed for this length of time. Cholesterol levels increased from 231 to 279 mg. dl" 1 in diuretic treated subjects but with standard deviations of ± 43 and ± 76 respectively. These types of data should not be used to prove an hypothesis or indeed in any scientific argument to make a point. In studies reported in 19791361 we noted that, while mean serum cholesterols did not change significantly in diuretic
76 M. Moser
Table 3
Mean changes in 311 patients treated with indapamide 2-5 mg . day~'
Laboratory value (mg.dl"')
Baseline
Week 12
Mean change
Week 40
Mean change from baseline
Blood, urea, nitrogen Glucose Uric acid Triglycerides Cholesterol Creatinine Calcium
14-7 93-7 5-99 155 6 227-1 102 9-45
161 100-5 6-98 1568 2321 1-07 9-64
1-4 6-8 0-99 1-2 50 005 019
15-5 98-3 6-90 166-5 225-4 106 9-63
0-8 4-6 0-91 10 9 -1-7 0-04 018
*No clinically significant differences between these results and those with HCTZ; exact data not available. Adapted fromf".
treated subjects, there were some patients whose levels decreased or increased by as much as 20-40 mg. dl" 1 . Serum lipid levels should be reviewed within 3—6 months after thiazide therapy is started to detect the few patients who will experience a marked increase in cholesterol or LDL levels. This should not add to cost of care since serum analyses need only be obtained 1-2 times a year while on therapy. It is questionable whether the minimal long-term effects of thiazide diuretics on serum lipids could explain the alleged shortfall in CHD events in the clinical trials or that these have a definite adverse effect on cardiovascular risk. The authors of the recent meta-analysis'71 state that their review suggests a 1% increase in cholesterol levels in diuretic-treated subjects. Even if we accept the Lipid Research Council analysis that a 1 % increase in cholesterol levels increases CHD by 2%(3?1 it would not explain an alleged shortfall of 6-8%. As noted, the shortfall (which may not even exist to any great extent if the new studies are included in the analysis) is more easily explained by the short duration of and the type of subjects in the clinical trials. Do dosages of diuretics or the type of diuretic used alter metabolic effects? The clinical trials used diuretics in dosages considerably higher (the equivalent of 50-100 mg of hydrochlorothiazide (HCTZ) or chlorthalidone (C)) than presently recommended (12-5-50mg.day" 1 ). Data suggest that lower doses will have less effect on potassium and uric acid levels'381 but not on lipid parameters'3639-401. Lower dosages (equivalent to 25 mg. day"' of HCTZ) may lower blood pressure to a degree almost equivalent to higher dosages. In Kochar's study1391 systolic blood pressure rose by 5 mmHg over a 6-month period when the dosage of HCTZ was reduced from 50 to 25 mg. day"'; glucose and lipid levels were not changed significantly but serum potassium levels rose and serum uric acid decreased. Some data suggest that indapamide, an indolin compound with vasodilating properties, may not adversely affect serum lipid levels'41'121. Many of these studies lacked controls or were open label. In a study of 260 patients, which compared indapamide to HCTZ, data indicated
that indapamide had no adverse effects on lipids at . 40 weeks (note an increase of 5mg.dl"' at 12 weeks) (Table 3)'361. Although comparative information is not published (and apparently is not available) on the HCTZ subjects, the authors clearly state that 'there was no clinically significant difference between the treatment groups for the laboratory assessments.' In another report'431 it was noted that indapamide in doses of 2-5mg. day" 1 caused an increase of about 4% in serum cholesterol levels, equivalent to that of HCTZ in a randomized study. An 8% increase was noted when 50 mg. day"' was used. In another study of a small number of subjects'441 lipid fractions did not change with indapamide but blood pressure was not lowered. In most of the comparative studies, a small dose of indapamide has been compared to relatively large doses of a thiazide (50 mg or more). While it is possible that indapamide causes fewer adverse effects on lipids than the thiazides, available data are not convincing. It is of interest to compare a study of a new formulation of metolazone, a long-acting diuretic with indapamide'431. Short-term (6 weeks) results indicate an increase in cholesterol of 12 and 23mg.dl"', LDL of 3 and 15mg.dl"', and triglycerides of 36 and 37 mg . dl~' with metolazone and indapamide respectively, indicating that the shortterm effects of indapamide on lipids may be more, not less, pronounced than with metolazone. Shorter-acting loop diuretics, such as furosemide or bumetanide, or longer-acting loop diuretics, such as torasamide, may have variable but somewhat less shortterm effects on serum lipid levels when compared to HCTZ. These agents have not been studied extensively. Torasemide in non-diuretic doses of 2-5 mg appears to have an antihypertensive effect (statistically significant when compared to placebo)1401. Cholesterol, LDL and HDL levels are insignificantly changed when compared to placebo. Increases in dosages to 5mg.day"' will lower blood pressure to a greater extent in some patients, without definite changes in lipid fractions. In comparative studies using 2-5 or 50mg in a single daily dose, blood pressure lowering appears to be equivalent (in most studies) to that of 50-100 mg of HCTZ147-481. Decreases in serum potassium and increases in uric acid appear to be less with 2-5 and 5-0mg.day" 1 of torasemide when
•
Diuretics and cardiovascular risk factors M. MOSER
Yale University School of Medicine, New Haven, Connecticut, U.S.A. KEY WORDS: Hypertension, diuretics, cardiovascular risk factors, clinical trials. Thiazide diuretics are efficacious, either as monotherapy or in combination with other antihypertensive drugs. They reduce blood pressure in a high percentage of hypertensive patients with minimal subjective side effects. There is increasing evidence that the use of diuretics, singly or in combination, will reduce morbidity and mortality associated with essential hypertension in both young and elderly subjects. Although diuretics may induce some changes in the plasma lipid profile, serum uric acid concentration and glucose metabolism, there is little evidence that these changes are of clinical significance. The increase in serum cholesterol concentration has rarely persisted in any trial beyond the first year of treatment. The incidence of diabetes mellitus in diuretic treated subjects is only about ]%, even when large doses are used. Gout may be precipitated in susceptible subjects, but is uncommon. For these reasons, diuretics should remain a preferredfirst-stepdrug of choice in the management of hypertension. Introduction Debate continues regarding the specific role of diuretic therapy in the management of hypertension despite everincreasing evidence that the use of these agents reduces morbidity and mortality in hypertensive patients1'"31. High doses of thiazide diuretics result in an increase in serum uric acid levels, a decrease in serum potassium, a shortterm increase in cholesterol and low density lipoprotein (LDL) levels, and probably a long-term increase in serum triglyceride levels'41. In addition, these medications increase insulin resistance in patients with hypertension who may have preexisting glucose tolerance and insulin utilization abnormalities15'61. But are these effects of clinical importance? Do they increase cardiovascular risk or minimize or even negate the benefits of blood pressure lowering? A recent meta analysis171 and several new studies in the elderly18"10' have cast doubt on the argument that coronary heart disease events (CHD) have not been reduced when diuretics are used as initial therapy in hypertensive individuals. In fact, the recent trials present strong evidence of benefit when these medications are used. But there are other questions to be addressed. Are all diuretics equally effective in lowering blood pressure and similar in their effects on electrolytes, glucose metabolism and lipids? Would lower dosages of the commonly used diuretics produce fewer metabolic effects while maintaining their antihypertensive capabilities, and perhaps reduce CHD events still further? Has the use of diuretics decreased CHD events? Approximately 75 000 patients have been studied in various clinical trials, all of which have used diuretics as initial therapy. Some compared diuretics to betablockers'9""1, some were placebo controlled18""131, while others compared a vigorously treated to a less rigorously Correspondence: Marvin Moser, MD, FACP, Davis Avenue Medical Center, 33 Davis Avenue, While Plains, New York 10605, U.S.A. 019S-668X/92/OG0072 + 09 $08.00/0
treated group1'41 or a control group1'51. Either a betablocker, alpha methyl-dopa, reserpine, or hydralazine were used as second or third-step medications in the trials. Most of these studies were not specifically designed to address the question of whether or not CHD mortality was reduced with antihypertensive therapy. Several trials included only patients without evidence of target organ involvement or other complications of vascular disease1"t121. It could have been predicted therefore that the number of cardiovascular events would be small and that the benefits of treatment would be difficult to prove given the short duration of the studies (3-5 years) and the numbers of patients evaluated. For example, in the Australian Study1'21, the placebo group of patients experienced approximately one-third the number of deaths compared to an age- and sex-matched group of the Australian population, making it difficult to demonstrate an even lower mortality in treated subjects. It is, therefore, not surprising that many of the individual hypertension treatment trials failed to show a statistically significant decrease in CHD events in treated compared to control groups. Speculations were advanced suggesting that it was the medication used rather than the trial design that resulted in these findings. Yet, in trials that included elderly subjects18"101 or subjects with evidence of pretreatment target organ involvement1'41 where the numbers of complications were sufficient to demonstrate differences in outcome, a decrease in CHD events has been reported in treated groups. Observational estimates that a decrease of 5—6 mmHg in diastolic blood pressure over a long period of time would decrease stroke deaths by approximately 35-40% were realized in the clinical trials; a 42% decrease was noted in the 3-5 year trials with a decrease in diastolic pressure of this magnitude17161. On the other hand, epidemiological or observational data suggested that a decrease in diastolic pressure of 5-6 mmHg should decrease CHD events by approximately 20-25%. The 14 randomized clinical trials analysed by Collins et al.m prior to the Systolic Hypertension in the Elderly (SHEP/81, © 1992 The European Society of Cardiology
Diuretics and cardiovascular risk factors
400
- (a)
k
«, 200 o "5 o 1-00 --
400 - ( b )
100 -
- „a
CD
| a
050
0-50 -
0-25 -
025 -
Baseline DBP category
•
2 00 -
t
1 1 76
2 1
84
3 1 91
4 1
98
5 1
105
73
Baseline DBP category
1 1
76
D
2
3
1 84
1 91
a
4 1
5
98
105
1
Approximate mean usual DBP (mmHg) Figure 1 Risks of stroke and coronary heart disease (CHD), estimated from combined studies. MacMahon el al. Lancet 1990; 335: 765. (a) Stroke and usual diastolic blood pressure (DBP) (seven prospective observational studies — 83 events), (b) Coronary heart disease and usual diastolic blood pressure (nine prospective observational studies — 4856 events).
• >
the- STOP"1 and the recent MRC1'01 studies showed a based on careful follow-up observations. It is also possstatistically significant decrease ofl 4% in CHD events. In ible, of course, that CHD will not be impacted sufficiently the latest MRC trial in the elderly1101 CHD events were by blood pressure lowering alone; other risk factors may reduced by 44%, strokes by 31%, and overall cardio- have to be reduced. But at present, there are few data to vascular deaths by 29% with low dose diuretic therapy substantiate a negative impact of diuretic therapy on compared to placebo. In the STOP study (using different cardiovascular risk. beta blockers or a diuretic), stroke risk was reduced by 47%, fatal myocardial infarction or sudden death by 50% and overall mortality by 43%. If the SHEP study had Effect of thiazide diuretics on specific cardiovascular risk been included in the recent meta-analysis, the resultant factors decrease in CHD events would have been approximately (1) Reducing blood pressure reduces cardiovascular 16%, if data from the most recent MRC110' study had been risk. The use of diuretics in hypertensive persons will included, this percentage would have been even higher reduce blood pressure to as great or greater degree as any and closer to epidemiological estimates. But to many of the other available antihypertensive agents, including observers, these data continued to suggest a shortfall in beta-blockers, alpha beta-blockers, alpha blockers, ACE benefit and elicited explanations that the drugs used inhibitors, and calcium entry blockers. A loop diuretic (mainly diuretics) may have induced metabolic changes like furosemide, while highly potent as a diuretic, is less that minimized benefit. Logic suggests that the so-called effective as an antihypertensive agent over the longshortfall was more probably related to the short duration term1"1. There are few comparative long-term data of the trials. Epidemiological data had assumed a long available with bumetanide or torasamide, a long-acting term (perhaps over decades) decrease of 5—6 mmHg in loop diuretic. diastolic blood pressure (DBP) to achieve a decrease of (2) The use of thiazide diuretics prevents progression of 20-25% in CHD, not a decrease in pressure for only 3-5 mild to more severe hypertension, thereby reducing the years. potential for cardiovascular complications1201. If it is the short duration of the trials that accounted for (3) Despite some equivocal evidence, lowering of blood the less than expected benefits for CHD, one might ques- pressure with diuretics will result in the prevention of or tion why stroke deaths decreased to the predicted level reversal of left ventricular hypertrophy814-2'1221. Thus, over the 3-5 year period of time of the clinical trials. The a major risk factor for CHD and serious arrhythmias slope of increase in stroke events with increases in blood appears to be positively affected (the effect of diuretics pressures is a steep one (Fig. 1). A minimal decrease in on this important cardiovascular risk factor is, however, pressure can therefore be expected to have a major effect less consistent than with ACE inhibitors or several of on stroke increase. The slope is less dramatic with CHD the beta-blockers, centrally acting agents and calcium events; it may take a longer period of time to decrease antagonists1231). CHD deaths with the same reduction in blood pressure. (4) Opinions differ on the clinical significance of the Some credence is given to this assumption by the 8- and effects of diuretics on lipids and glucose metabolism. The 10-year data from the HDFP"71 and MRFIT"81 studies use of thiazide diuretics over a period of less than one year respectively. CHD events were reduced further in the will result in an increase of about 5-7% in serum choleslonger-term follow-up. These new data, however, are not terol levels with a minimal, if any, change in high density
74
M. Moser
Table I Effect of diuretic-based therapy on serum cholesterol in long-term clinical trials* Total cholesterol level (mg dl~') Trial
Berglund and Anderson MRC trial Men Active treatment Placebo Women Active treatment Placebo MAPHY study HDFP SC group Oslo Active treatment Control MRFIT SI group UC group HAPPHY trial EWPHE study Active treatment Placebo MRC in elderly* Active treatment Placebo
Duration (years)
6
Difference Baseline
iMter treatment
267
255
-12
245 244
245 239
0 -5
261 260 244
260 256 243
-1 -4 -1
232
223
-9
272 278
273 280
+1 +2
254 254 242
236 240 242
-18 -14 0
256 259
238 239
-18 -20
228 228
232 232
+4 +4
3+
6 5 4 6 4 3 5
All of the trials used diuretics in doses equivalent to 50 mg or more/day of hydrochlorothiazide. 'Estimated from available data.
lipoprotein (HDL) but an increase in LDL and triglyceride levels'241. Many of the reported studies lacked a control group or included only small numbers of subjects. Elsewhere we have reviewed cholesterol changes in the longterm clinical trials, somewhat different findings emerge'25'. Although none of these trials (in the 1970s and 1980s) was specifically designed to study lipids and lipid fractions, data were collected in large numbers of subjects over a 2-5 year period. Table 1 summarizes the data on cholesterol changes in diuretic-based trials of longer than one year's duration. None showed a definite increase in cholesterol levels when thiazide diuretics were used. In many trials, other agents were used concomitantly; since beta-blockers were used in several of them, it would have been expected that if an adverse effect had occurred, it would have been exaggerated rather than minimized. In one 58-week trial'261, there was an increase of 5% in total cholesterol levels (and 10% in LDLs). However, as the author of this trial states, the increase occurred in subjects with initial low levels; in those with baseline cholesterols of 240-250 mg. dl" 1 no change was noted. This suggests regression to the mean, as noted by other investigators'14-271 rather than a specific medication effect. In the French Study of 20 months' duration'271, high baseline cholesterol levels decreased; low baseline levels increased on a diuretic with a potassium sparing component. Overall there was no change during the study period. The hypertension detection and follow-up programme (HDFP) data are also of interest. In this study, baseline elevated cholesterol levels decreased on thiazide diuretic
therapy while a slight increase in serum cholesterols was noted in subjects with low baseline levels'281 (Fig. 2). This study suggests that thiazide therapy need not be avoided in patients with hyperlipidaemia, as suggested by some physicians'291. In two studies (MRC and MRFIT)"1-301, cholesterol levels remained the same or decreased on thiazide therapy but not to the extent that they had in the placebo or less vigorously treated groups. It was suggested that the degree of lipid lowering (which is noted frequently in any clinical trial) may have been blunted by thiazide therapy. This may have been true in the MRC study where, for example, a difference of 3mg.dl~' between placebo and diuretic therapy in women and 5 mg. dl~' in men was noted. The MRFIT data are confusing and difficult to interpret'31'. This was not only a study of hypertension therapy but one in which individuals could enter the trial with elevated cholesterol levels and/or a smoking history. The patients who experienced the greatest decrease in cholesterol levels were those who lost the most weight and who entered the study with the highest cholesterol levels (Table 2). The non-smokers in MRFIT, who were not on treatment for hypertension at 6 years, were either not hypertensive or only mildly hypertensive at baseline. It can be assumed that they were admitted to the trial primarily because of elevated serum cholesterol concentrations. It is in this group of non-smokers and 'non-treated' subjects that the greatest difference was noted in the decrease in cholesterol levels ( — 23 in non-treated compared to — l l m g . d l " 1 in treated subjects) and in weight loss
Diuretics and cardiovascular risk factors
Baseline
75
2 3 Years of therapy
Figure 2 Effect of diuretic-based therapy on cholesterol levels in hypertension detection and follow-up programme participants'211. D = cholesterol at baseline < 2 2 0 m g . d l - ' (n = 270); • = cholesterol at baseline 221-249mg. d l 1 (n = 207); • = cholesterol at baseline 250-279 mg. dl-' (n = 113); A =cholesterol at baseline > 280 mg . dl -' (n = 74).
Table 2
Effect of diuretic-based therapy on serum cholesterol in MRFIT special intervention subjects Total cholesterol levels (mg. dl"1)
Hypertensive at baseline
Smokers On treatment for hypertension at 6 years Not on treatment for hypertension at 6 years Non-smokers On treatment for hypertension at 6 years Not on treatment for hypertension at 6 years
Change in body weight (kg)
No
Baseline
6 years
Difference
549
225
216
-9
-11
266
234
219
-15
-20
818
246
235
-11
-1-9
293
254
231
-23
-3-9
Adapted from1321.
( — 8-6 in non-treated compared to —4-2 lbs in treated subjects). Baseline cholesterol levels were also higher in the non-treated subjects. A logical explanation for at least some of the difference in outcome might be that the patients themselves, as well as the investigators, expended more effort in trying to correct their weight and their one major risk factor (elevated cholesterol level). HDL levels in the MRFIT study were decreased by less than 1 -0 mg. dl"' in the thiazide treated group compared to the non-treated group — again this result is confounded by the complicated nature of the study and the observation that changes in HDL correlated closely with weight gain or loss — those who gained the most weight experienced a decrease in HDL; those who lost weight showed an increase. None of the observed changes in lipids between the treated and the non-treated groups achieved statistical significance. Some studies have suggested that the thiazide-induced effect on cholesterol levels can be eliminated by a low-fat/ low-cholesterol diet1"1, while others have reported a
decrease in serum cholesterol levels following withdrawal of thiazide diuretics'341; the usually observed weight gain following diuretic withdrawal suggests that haemodilution may explain, at least part, of the chemical changes. Extracellular fluid volume remains contracted to a slight degree during long-term diuretic therapy141. In addition to the MRFIT1301 and MRC1"1 studies, which appear to indicate a thiazide-induced blunting of the beneficial effects of diet, a subset of patients from the HAPPHY study have been reported as demonstrating a long-term (42 months) adverse effect of thiazides on cholesterol levels'351. However, only a very small number of patients had been followed for this length of time. Cholesterol levels increased from 231 to 279 mg. dl" 1 in diuretic treated subjects but with standard deviations of ± 43 and ± 76 respectively. These types of data should not be used to prove an hypothesis or indeed in any scientific argument to make a point. In studies reported in 19791361 we noted that, while mean serum cholesterols did not change significantly in diuretic
76 M. Moser
Table 3
Mean changes in 311 patients treated with indapamide 2-5 mg . day~'
Laboratory value (mg.dl"')
Baseline
Week 12
Mean change
Week 40
Mean change from baseline
Blood, urea, nitrogen Glucose Uric acid Triglycerides Cholesterol Creatinine Calcium
14-7 93-7 5-99 155 6 227-1 102 9-45
161 100-5 6-98 1568 2321 1-07 9-64
1-4 6-8 0-99 1-2 50 005 019
15-5 98-3 6-90 166-5 225-4 106 9-63
0-8 4-6 0-91 10 9 -1-7 0-04 018
*No clinically significant differences between these results and those with HCTZ; exact data not available. Adapted fromf".
treated subjects, there were some patients whose levels decreased or increased by as much as 20-40 mg. dl" 1 . Serum lipid levels should be reviewed within 3—6 months after thiazide therapy is started to detect the few patients who will experience a marked increase in cholesterol or LDL levels. This should not add to cost of care since serum analyses need only be obtained 1-2 times a year while on therapy. It is questionable whether the minimal long-term effects of thiazide diuretics on serum lipids could explain the alleged shortfall in CHD events in the clinical trials or that these have a definite adverse effect on cardiovascular risk. The authors of the recent meta-analysis'71 state that their review suggests a 1% increase in cholesterol levels in diuretic-treated subjects. Even if we accept the Lipid Research Council analysis that a 1 % increase in cholesterol levels increases CHD by 2%(3?1 it would not explain an alleged shortfall of 6-8%. As noted, the shortfall (which may not even exist to any great extent if the new studies are included in the analysis) is more easily explained by the short duration of and the type of subjects in the clinical trials. Do dosages of diuretics or the type of diuretic used alter metabolic effects? The clinical trials used diuretics in dosages considerably higher (the equivalent of 50-100 mg of hydrochlorothiazide (HCTZ) or chlorthalidone (C)) than presently recommended (12-5-50mg.day" 1 ). Data suggest that lower doses will have less effect on potassium and uric acid levels'381 but not on lipid parameters'3639-401. Lower dosages (equivalent to 25 mg. day"' of HCTZ) may lower blood pressure to a degree almost equivalent to higher dosages. In Kochar's study1391 systolic blood pressure rose by 5 mmHg over a 6-month period when the dosage of HCTZ was reduced from 50 to 25 mg. day"'; glucose and lipid levels were not changed significantly but serum potassium levels rose and serum uric acid decreased. Some data suggest that indapamide, an indolin compound with vasodilating properties, may not adversely affect serum lipid levels'41'121. Many of these studies lacked controls or were open label. In a study of 260 patients, which compared indapamide to HCTZ, data indicated
that indapamide had no adverse effects on lipids at . 40 weeks (note an increase of 5mg.dl"' at 12 weeks) (Table 3)'361. Although comparative information is not published (and apparently is not available) on the HCTZ subjects, the authors clearly state that 'there was no clinically significant difference between the treatment groups for the laboratory assessments.' In another report'431 it was noted that indapamide in doses of 2-5mg. day" 1 caused an increase of about 4% in serum cholesterol levels, equivalent to that of HCTZ in a randomized study. An 8% increase was noted when 50 mg. day"' was used. In another study of a small number of subjects'441 lipid fractions did not change with indapamide but blood pressure was not lowered. In most of the comparative studies, a small dose of indapamide has been compared to relatively large doses of a thiazide (50 mg or more). While it is possible that indapamide causes fewer adverse effects on lipids than the thiazides, available data are not convincing. It is of interest to compare a study of a new formulation of metolazone, a long-acting diuretic with indapamide'431. Short-term (6 weeks) results indicate an increase in cholesterol of 12 and 23mg.dl"', LDL of 3 and 15mg.dl"', and triglycerides of 36 and 37 mg . dl~' with metolazone and indapamide respectively, indicating that the shortterm effects of indapamide on lipids may be more, not less, pronounced than with metolazone. Shorter-acting loop diuretics, such as furosemide or bumetanide, or longer-acting loop diuretics, such as torasamide, may have variable but somewhat less shortterm effects on serum lipid levels when compared to HCTZ. These agents have not been studied extensively. Torasemide in non-diuretic doses of 2-5 mg appears to have an antihypertensive effect (statistically significant when compared to placebo)1401. Cholesterol, LDL and HDL levels are insignificantly changed when compared to placebo. Increases in dosages to 5mg.day"' will lower blood pressure to a greater extent in some patients, without definite changes in lipid fractions. In comparative studies using 2-5 or 50mg in a single daily dose, blood pressure lowering appears to be equivalent (in most studies) to that of 50-100 mg of HCTZ147-481. Decreases in serum potassium and increases in uric acid appear to be less with 2-5 and 5-0mg.day" 1 of torasemide when
•
