Induction of Insulin Resistance by Beta- Blockade but Not ACE-Inhibition: Long- Term Treatment with Atenolol Or Trandolapril

Home , Insulin resistance

Journal of Human Hypertension (2000) 14, 175–180  2000 Macmillan Publishers Ltd All rights reserved 0950-9240/00 $15.00 www.nature.com/jhh ORIGINAL ARTICLE Induction of insulin resistance by beta- blockade but not ACE-inhibition: long- term treatment with atenolol or trandolapril

R Reneland, E Alvarez, P-E Andersson, A Haenni, L Byberg and H Lithell Uppsala University, Uppsala, Sweden

The effects on glucose metabolism by the beta-blocker and high-density lipoprotein cholesterol (؊13% vs atenolol and the angiotensin-converting enzyme (ACE)- ؉0.7%) also differed signiﬁcantly between atenolol and inhibitor trandolapril were investigated in a randomised trandolapril. Results after 8 weeks were similar. Glucose double-blind parallel group study of patients with pri- tolerance was not affected by either drug. Atenolol mary hypertension. Twenty-six patients were treated reduced diastolic blood pressure (DBP) better than tran- with 50–100 mg atenolol and 27 patients with 2–4 mg dolapril (؊15.3 mm Hg vs ؊6.6 mm Hg for supine DBP The difference in effect on .(0.012 ؍ trandolapril o.d. Intravenous glucose tolerance tests, after 48 weeks, P euglycaemic hyperinsulinaemic clamps and serum lipid insulin sensitivity between the drugs corresponded to measurements were performed after 8 and 48 weeks of 25% of the baseline values of insulin sensitivity, and active treatment. After 48 weeks insulin sensitivity was persisted over 48 weeks of treatment. The choice of anti- reduced by 23% by atenolol while it remained hypertensive treatment could inﬂuence the risk of dia- unchanged during trandolapril treatment (؉0.5%, betes associated with treated hypertension. Journal of .for difference between treatments, Human Hypertension (2000) 14, 175–180 0.0010 ؍ P (ANCOVA). The effect on triglycerides (؉22% vs ؊8.5%

Keywords: angiotensin-converting enzyme inhibitors; beta-blockers; human; insulin; randomised clinical trial

Introduction duration of 6 months or less. Therefore, we wanted to study the metabolic effects of the ACE inhibitor It has been shown that antihypertensive treatment 1,2 trandolapril and the beta-blocker atenolol in a ran- increases the risk for diabetes mellitus. One poss- domised double-blind parallel group trial with end- ible mediator of that effect may be the impaired point assessment after 8 and у48 weeks to see insulin sensitivity associated with the use of beta- whether the effects differed between the long and blockers and diuretics, the dominating drug classes short term. in these epidemiological studies.3–5 Furthermore, recently a randomised intervention study showed that treatment based on the angiotensin-converting Patients and methods enzyme (ACE)-inhibitor, captopril, was associated Patients and protocol with a lower incidence of diabetes than conventional antihypertensive treatment.6 Patients with Patients were recruited through advertisements in primary hypertension are insulin resistant.7,8 Elev- the local newspapers. At the initial screening visit ated plasma insulin levels, an indirect marker of inclusion and exclusion criteria (Table 1) were insulin resistance, are associated with increased risk assessed, informed consent was obtained, and any of coronary heart disease.9–12 Therefore, metabolic ongoing antihypertensive medication was stopped. factors may contribute to the excess risk for coronary After that, blood pressure (BP) was monitored at reg- heart disease associated with even well-controlled ular visits at the clinic. When the diastolic BP (DBP) treated hypertension.13,14 However, there has been reached 95 mm Hg single-blind placebo adminis- some uncertainty as to whether the metabolic effects tration was initiated. As soon as a patient had met of antihypertensive drugs persist during continuing the BP criterion for inclusion in the active phase, treatment since most metabolic studies have had a randomisation was performed and baseline investigations were carried out. Intravenous glucose tolerance tests and the euglycaemic, hyperinsulinaemic Correspondence: Richard Reneland, Department of Public Health clamps were performed on separate days, so that the and Caring Sciences/Geriatrics Unit, Uppsala University, P O Box 609, S-751 25 Uppsala, Sweden investigations were performed during a period of 4– Received 10 March 1999; revised and accepted 24 September 8 days. All metabolic investigations were performed 1999 in the morning after an overnight fast and before Metabolic effects of trandolapril and atenolol R Reneland et al 176 Table 1 Inclusion and selected exclusion criteria Insulin sensitivity measurements

Inclusion Exclusion criteria Whole body sensitivity to insulin was measured by criteria the euglycaemic hyperinsulinaemic clamp pro- cedure according to DeFronzo et al,15 with minor Age 18–75 years Diabetes mellitus modifications. Insulin (Actrapid Human, Novo, DBP 95–115 mm Hg Lipid-lowering drugs, hormones, etc Copenhagen, Denmark) was infused at a rate of Stable weight Secondary hypertension 2 (Ͻ2.5 kgs change Recent cardiovascular event 56 mU/(min/m body surface area). The amount of during run-in Cardiac arrhythmias glucose infused to maintain the target glucose level period) Unstable angina/congestive heart failure during the period of assumed steady-state was Informed consent Mitral/aortic valve stenosis; defined as glucose uptake (M; mg/kg/min). Adjust- cardiomyopathy ment for the steady-state insulin concentration Renal or hepatic dysfunction Hereditary angioneurotic edema defined the insulin sensitivity index (M/I; Haematologic or electrolyte disturbances mg/min/kg/(mU/l). The intra-individual coefficient Treatment with enzyme inducing agents of variation for the insulin sensitivity index in our Treatment with antihypertensive drugs laboratory was 14%. Pregnancy/lactation/child-bearing potential Hypersensitivity to any of the study drugs Intravenous glucose tolerance test (IVGTT) Each patient’s response to an intravenous glucose load was assessed by a 90-min intravenous glucose drug intake. The clinical characteristics of the two tolerance test. The glucose load was an intravenous treatment groups are reported in Table 2. After base- injection within 1.5 min of 300 mg glucose per kg of line investigations, active treatment was initiated body weight in a 50% glucose solution. The disap- with 2 mg trandolapril or 50 mg atenolol with reg- pearance rate of glucose was expressed as a k value = × ular BP follow-up visits every 4 weeks (weeks 0–16) calculated from the formula k ln2 100/t1/2 where or 8 weeks (weeks 16–48). At each of these visits t1/2 is the time required for the glucose concentration to be halved. Plasma glucose was measured by the adverse events and concomitant medication were  monitored by a trained nurse. After 4 weeks, the glucose dehydrogenase method (Gluc-DH , Merck, dose was doubled if the DBP was Ͼ90 mm Hg. Eight Darmstadt, Germany). Insulin was assayed in EDTA weeks into the active phase, investigations were plasma in duplicate using an enzymatic-immuno-  repeated and patients were dropped out if their DBP logical assay (Enzymmun , Boehringer Mannhein, was Ͼ95 mm Hg on two consecutive measurements. Germany) performed in a ES300 automatic analyser A third assessment of end-points was performed (Boehringer Mannheim). The mean fasting plasma after 48 weeks of active treatment. glucose and insulin values were calculated from three blood samples drawn 5 min apart prior to the injection of glucose. Peak insulin response was Office blood pressure defined as the mean of insulin values measured in Supine BP and heart rate were measured in tripli- the samples drawn at 4, 6, and 8 min and the insulin cate by trained personnel. Cuff sizes of 12 × 45 or increment is reported as the difference between the 15 × 45 cm were used depending on the arm circum- peak and mean fasting value. Area under the ference. The three measurements on each occasion incremental curve (AUC) for glucose and insulin were averaged and the mean value of the DBP was was calculated by a trapezoid method. the basis for treatment decisions. Normalisation was р defined as a DBP 90 mm Hg at the time-point at Lipid and lipoprotein measurement which efficacy was assessed (8 or 48 weeks). In addition, responders were defined as patients hav- Cholesterol and triglyceride concentrations in serum ing normalised or reduced their DBP у10 mm Hg were assayed by enzymatic techniques (Instru- during the observation period. mentation Laboratories, Lexington, MA, USA) in a Monarch 2000 centrifugal analyzer. High-density lipoproteins (HDL) were separated by precipitation Table 2 Baseline characteristics of the study participants accord- with magnesium cloride/phosphotungstate. Low- ing to treatment (means s.d.) density lipoprotein (LDL) cholesterol was calculated using Friedewald’s formula: LDL = serum choles- Atenolol Trandolapril terol-HDL-(0.45 serum triglycerides). (n = 26) (n = 27)

Age (years) 56 (9) 53 (10) Plasminogen activator inhibtor-1 (PAI-1) activity Sex (f/m) 7/19 7/20 BMI (kg/m2) 28 (4) 27 (3) PAI-1 activity was analysed in 18 atenolol treated and 20 trandolapril treated patients with a two-step ء(17) 158 ء(Systolic BP (mm Hg) 169 (14 Diastolic BP (mm Hg) 104 (4.9) 100 (4.5) indirect enzymatic assay (Spectrolyse/pL PAI kits, Insulin sensitivity index 4.8 (3) 5.3 (3) (mg/kg/min/100 mU) Biopool AB, Umea˚, Sweden). The activity is given Fasting plasma glucose 5.7 (0.6) 5.5 (0.4) in U/mL, where one unit is the amount of PAI-1 that (mmol/l) inhibits one international unit of single chain tissue- type plasminogen activator.

Journal of Human Hypertension Metabolic effects of trandolapril and atenolol R Reneland et al 177 Clinical laboratory Baseline homogeneity All other blood and urinary analyses were perfor- Baseline characteristics are reported in Table 2. The med at the Department of Clinical Chemistry, Uni- two groups were similar with the exception of BP versity Hospital, Uppsala, using routine laboratory which was higher in the atenolol treated group. methods. Insulin sensitivity Anthropometric measurements Adjusted means for each treatment group and P- Height was measured to the nearest whole centi- values with 95% confidence intervals for treatment metre and body weight was measured to the nearest differences are reported in Table 3. The within- 100 g. The body mass index (BMI) was calculated as group changes in selected variables are shown in the ratio of the weight in kg to the square of the Figure 1. Atenolol reduced insulin sensitivity by height in meters. 18% after 8 weeks and 23% after 48 weeks whereas trandolapril was neutral in this respect (−1.9% and +0.5% after 8 and 48 weeks, respectively). The dif- Statistical analysis ferences in treatment effect were highly significant. The predetermined primary target variables were the differences in effect on glucose uptake and insulin Glucose tolerance sensitivity index between the two treatment There was no persistent effect on glucose tolerance modalities. Data processing was performed using the by either treatment. At 8 weeks the trandolapril statistical program package SAS version 6.04 for group had significantly higher k-values at the IVGTT personal computers. All tests used have been two- (Table 3), a result that seemed to be dependent on a sided. When necessary, variables were logarithmi- transient increase (+12%) in the k-value by trandola- cally tranformed to allow hypothesis testing using pril. Both treatments had similar effects on serum parametric tests. If this was not feasible, Wilcoxon’s potassium concentrations. signed rank test or Mann–Whitney’s U-test was used. Testing for differences between treatment groups and between treatments and placebo was Lipid metabolism performed according to the principle of intention-to- treat using a model for analysis of covariance using The lipid changes followed the changes in insulin factors for treatment and study period with adjust- sensitivity. There was a significantly more favour- ment for baseline values. Testing for baseline differ- able effect on triglycerides and HDL-cholesterol by ences in nominal variables between groups was trandolapril than atenolol at 8 weeks as well as at done using the ␹2 test or Fisher’s exact test, as appro- 48 weeks (Table 3). There was, however, a tendency priate. Changes over time within each treatment towards an improvement in triglyceride levels with group were calculated from the observed means. time in the atenolol group since 48-week values did not differ significantly from baseline (1.66 vs 1.36, Statistical significance was accepted for P-values of = Ͻ0.05 or a 95% confidence interval that excluded P 0.09). the value for no difference. If no mention of statistical significance is indicated in tables, figures or text, Plasminogen activator inhibitor-1 (PAI-1) levels it means not significant by default. In the atenolol group, PAI-1 levels increased by 27% after 48 weeks. During trandolapril treatment PAI- Results 1 levels decreased by 6%. The difference between treatments was not significant (P = 0.067). Patient flow and drop-outs

A total of 175 patients were recruited into the drug- Blood pressure free run-in period, and 53 were randomised and included in the intention-to-treat analysis. The over- Treatment effects on blood pressure are reported in whelming majority of drop-outs during the pre-ran- Table 3. Blood pressure was well controlled in both domisation phase were patients who did not reach groups with a tendency towards a larger blood press- inclusion BP when their habitual antihypertensive ure reduction in the atenolol group. At 48 weeks the medication was discontinued. Twenty-six patients response rates were 67% and 88% for trandolapril in the trandolapril group and 23 patients in the aten- and atenolol, respectively (P = 0.10) and the normal- olol group experienced adverse events. The most isation rates were 63% in the trandolapril group and common were headache (ﬁve out of 26 patients in 85% in the atenolol group (P = 0.12). the atenolol group, four out of 27 in the trandolapril group), cough (one out of 26 for atenolol, seven out Body weight of 27 for trandolapril), clinically relevant hyperlipidaemia (nine out of 26 for atenolol, two out of 27 Body weight was unchanged in the atenolol group for trandolapril). There were two discontinuations after 8 weeks but had increased by 2.0 kg after 48 due to adverse events, one patient with depression weeks. Trandolapril had no effect on body weight and impotence in the atenolol group and one patient (−0.5 kg at 48 weeks, P = 0.19). Atenolol and trandol- with intolerable cough in the trandolapril group. april had signiﬁcantly different effects on body

Journal of Human Hypertension Metabolic effects of trandolapril and atenolol R Reneland et al 178 Table 3 Comparison between treatment effects. Adjusted means

Variable Period (weeks) Atenolol Trandolapril 95% CI P-value

Weight (kg) 0 84.4 84.4 8 84.9 84.2 0.13 48 86.5 83.9 0.0002

Insulin sensitivity (mg/min/kg/100 mU) 0 4.8 5.3 8 4.1 5.0 0.26 to 1.53 0.008 48 3.8 5.1 0.55 to 2.0 0.001

Fasting insulin (mU/l) 0 15.5 13.0 8 15.6 14.8 −3.7 to 2.1 0.58 48 16.7 14.9 −5.0 to 1.39 0.28

Fasting plasma glucose (mmol/l) 0 5.69 5.48 8 5.54 5.51 −0.26 to 0.18 0.73 48 5.65 5.69 −0.20 to 0.28 0.72

K-value at IVGTT 0 1.14 1.25 8 1.09 1.36 0.08 to 0.46 0.009 48 1.22 1.31 −0.06 to 0.26 0.23

Insulin AUC (IVGTT) 0 4349 3093 8 4038 3686 −1052 to 347 0.33 48 4515 4074 −1094 to 212 0.19

Glucose AUC (IVGTT) 0 900 899 8 950 863 −143 to −30 0.004 48 942 860 −134 to −29 0.004

Serum potassium (mmol/l) 0 3.88 3.95 8 4.09 4.13 −0.09 to 0.18 0.51 48 4.07 4.13 −0.06 to 0.19 0.34

Serum triglycerides (mmol/l) 0 1.36 1.54 8 1.92 1.44 −0.80 to −0.14 0.008 48 1.73 1.35 −0.75 to −0.03 0.04

Serum LDL-chol (mmol/l) 0 3.42 3.96 8 3.66 3.81 −0.11 to 0.42 0.25 48 3.61 3.73 −0.10 to 0.33 0.30

Serum HDL-chol (mmol/l) 0 1.34 1.28 8 1.11 1.28 0.07 to 0.27 0.002 48 1.14 1.31 0.07 to 0.27 0.002

Plasma PAI-1 activity (U/ml) 0 24.3 23.6 8 28.9 24.0 −13.8 to 4.1 0.29 48 30.9 22.1 −18.0 to 0.3 0.067

Systolic BP (mm Hg) 0 169 158 8 141 150 −0.0 to 17.2 0.056 48 145 154 −0.9 to 18.7 0.082

Diastolic BP (mm Hg) 0 104 100 8 87 91 0.9 to 7.6 0.12 48 87 95 1.9 to 13 0.012

95% CI, 95% conﬁdence intervals for the difference in treatment effect; IVGTT, intravenous glucose tolerance test; AUC, incremental area under the curve; LDL-chol, low-density lipoprotein cholesterol; HDL-chol, High-density lipoprotein cholesterol; PAI-1, plasminogen activator inhibitor-1.

weight after 48 weeks (P Ͻ 0.0002), but not after 8 (−35% vs −12.5%, P = 0.053). In the trandolapril weeks (P = 0.13). group the response was not different between doses (P = 0.76). Correlation analysis Discussion There was no correlation between change in serum potassium and change in either glucose tolerance or This study conﬁrms previous studies that have insulin sensitivity either in the study as a whole or reported impaired insulin sensitivity during treat- in either of the treatment groups. In the atenolol ment with atenolol,3,4 and demonstrates that the group, the low-dose group tended to decrease more response to the ACE inhibitor trandolapril is sig- in insulin sensitivity index than the high-dose group niﬁcantly more favourable in this respect.

Journal of Human Hypertension Metabolic effects of trandolapril and atenolol R Reneland et al 179 reports of clinically relevant hyperlipidaemia in comparison to two reports in the trandolapril treated group. In contrast to the insulin sensitivity data there was a tendency towards normalisation of the triglyceride levels with sustained treatment. In a previous open long-term follow-up study we observed that the effects on insulin sensitivity as well as on lipids persisted over a period of 2–3 years.17 It is therefore possible that the apparent amelioration of the hypertriglyceridaemia with time in the present study may be the result of random fluctuations. Non-insulin-dependent diabetes as well as insulin resistance are associated with an increased cardiovascular risk.2,9–12 The fast inhibitor of fibrinolysis, Figure 1 Relative changes after 48 weeks’ treatment with atenolol PAI-1, which is elevated in these conditions, has and trandolapril for selected variables. M/I, insulin sensitivity been suggested to play a role in this association.18 index; k-value, glucose disappearance coefficient at IVGTT; TG, Although no consistent change in PAI-1 levels have serum triglycerides; HDL chol, high-density lipoprotein choles- 19 terol; s-K, serum potassium; PAI-1, plasma plasminogen activator been reported in other studies, we have previously -P Ͻ 0.05 for change from baseline. found a 17% increase in PAI-1 levels during atenoء ;inhibitor-1 lol treatment which did not quite reach statistical significance.20 The changes in PAI-1 levels in this Importantly, the effects of the two drugs remained study followed the pattern of changes in triglycer- unchanged after 11 months treatment as compared ides and insulin sensitivity. This could be an indi- to the 2-months data, indicating that the reduced cation that PAI-1 is mechanistically linked to the insulin action during atenolol is not ameliorated insulin sensitivity-metabolic complex. after prolonged treatment. The difference in effect Atenolol reduced BP better than trandolapril at by the two drugs corresponded to roughly 25% of the doses selected for this study, even if the differ- the baseline values of insulin sensitivity. A pro- ence was statistically significant only for DBP after longed difference in insulin sensitivity of this mag- 48 weeks. It is not likely that the difference in effects nitude for the duration of decades may have clinical on insulin sensitivity was due to inappropriately consequences such as an increased risk of diabetes high doses of atenolol, since, if anything, patients in mellitus. An increased risk of diabetes associated the low-dose group experienced a larger decrease in with beta-blockade and diuretic treatment has insulin sensitivity than the high-dose group. indeed been demonstrated recently.2 In the recently Are metabolic effects induced by antihypertensive published CAPPP-study ACE-inhibitor based treat- treatment of any consequence for the prognosis of ment reduced the incidence of diabetes as compared hypertensive patients? The data from a recently pub- to conventional antihypertensive treatment.6 lished randomised intervention study in hyperten- It has been speculated that the difference in meta- sives6 supports the hypothesis that the differences bolic effects between beta-blockers and ACE-inhibi- in metabolic profiles between ACE-inhibitors and tors is due to effects on potassium balance and the conventional antihypertensive treatment translates fact that potassium is not infused during hyperinsul- into a difference in diabetes incidence. On the other inaemia.16 In this study there was no relationship hand, one study has failed to show an increased risk between the change in potassium and the change in of coronary heart disease associated with diabetes insulin sensitivity or glucose tolerance during treat- diagnosed during follow-up of treated hypertensive ment. Furthermore, the two drugs induced almost patients.21 The confidence interval for the risk esti- identical increases in serum potassium, which mate was very wide in that study, however, and makes it unlikely that the different metabolic effects included a relative risk of up to 6. The same group of the two drugs were secondary to different effects reported that the on-treatment level of serum choles- on potassium homeostasis. terol, but not BP, was an independent determinant Body weight increased significantly in the ateno- of coronary heart disease in a long-term follow-up of lol-treated group but this slow increase in body treated hypertensive patients.14 Concerning patients weight developed mainly in the period between 8 that are already diabetic, the data from the UKPDS and 48 weeks, whereas the insulin resistance was antihypertensive treatment study in type 2 diabetics nearly completely developed already after 8 weeks’ indicates that BP reduction seems at least as treatment. This sequence of events indicates that the important as tight metabolic control for the difference in effect on insulin sensitivity between reduction of cardiovascular events.22,23 In the same the two drugs was not exclusively mediated by prospective study, randomised treatment with aten- changes in body weight. The impaired insulin sensi- olol was associated with higher levels of glycated tivity in the atenolol treated group was not reflected haemoglobin, a higher frequency of additional glu- in impaired glucose tolerance, as represented by cose lowering treatment and a greater weight gain fasting glucose and glucose disappearance rate dur- but this did not result in poorer outcome in terms ing the IVGTT. Lipid levels were affected in the typi- of clinical events24 and there was no recommen- cal pattern for beta-blockers by atenolol. This was dation for the use of one treatment in preference to reflected in the pattern of adverse event as nine the other. Taken together, these data suggests that in

Journal of Human Hypertension Metabolic effects of trandolapril and atenolol R Reneland et al 180 hypertensive patients metabolic factors may or may 10 Welborn TA, Wearne K. Coronary heart disease inci- not be important, whereas in diabetic patients the dence and cardiovascular mortality in Busselton with achieved BP levels may be most important for the reference to glucose and insulin concentrations. Dia- clinical outcome of patients treated with anthypert- betes Care 1979; 2: 154–160. ensive drugs. 11 Ducimetie`re P et al. Relationship of plasma insulin levels to the incidence of myocardial infarction and In summary, trandolapril compared favourably to coronary heart disease mortality in a middle aged atenolol in its effect on insulin sensitivity and lipid population. Diabetologia 1990; 19: 205–210. profile while there was no sustained difference in 12 Despre´s J-P et al. Hyperinsulinemia as an independent effect on glucose tolerance. The difference in effect risk factor for ischemic heart disease. N Engl J Med was large enough to be in line with the difference 1996; 334: 952–957. in risk for diabetes in treated hypertensive patients 13 Lindholm L, Ejlertsson G, Schersten B. High risk of and the difference in metabolic control in type 2 dia- cerebro-cardiovascular morbidity in well treated male betes when ACE inhibitors and beta-blockers are hypertensives. A retrospective study of 40–59-year-old compared. hypertensives in a Swedish primary care district. Acta Med Scand 1984; 216: 251–259. 14 Andersson OK et al. Survival in treated hypertension: Acknowledgements follow up study after two decades. BMJ 1998; 317: 167–171. This study was supported financially by Knoll AG. 15 DeFronzo RA, Tobin JD, Andres R. Glucose clamp technique: a method for quantifying insulin secretion References and resistance. Am J Physiol 1979; 6: E214–E223. 16 Heinemann L et al. Four week administration of an 1 Skarfors ET, Selinus KI, Lithell HO. Risk factors for ACE inhibitor and a cardioselective ␤-blocker in heal- developing non-insulin dependent diabetes: a 10 year thy volunteers: no influence on insulin sensitivity. Eur follow up of men in Uppsala. BMJ 1991; 303: 755–760. J Clin Invest 1995; 25: 595–600. 2 Mykkanen L et al. Increased risk of non-insulin-depen- 17 Lind L, Pollare T, Berne C, Lithell H. Long-term meta- dent diabetes mellitus in elderly hypertensive sub- bolic effects of antihypertensive drugs. Am Heart J jects. J Hypertens 1994; 12: 1425–1432. 1994; 128: 1177–1183. 3 Pollare T et al. Metabolic effects of diltiazem and aten- 18 Juhan-Vague I, Alessi MC, Vague P. Increased plasma olol: results from a randomized, double blind study plasminogen activator inhibitor 1 levels. A possible with parallel groups. J Hypertens 1989; 7: 551–559. link between insulin resistance and atherothrombosis. 4 Pollare T, Lithell H, Selinus I, Berne C. Sensitivity to Diabetologia 1991; 34: 457–462. insulin during treatment with atenolol and metoprolol: 19 Lottermoser K et al. Antihypertensive drug treatment a randomized, double blind study of effects on carbo- and fibrinolytic function. Am J Hypertens 1998; 11: hydrate and lipoprotein metabolism in hypertensive 378–384. patients. BMJ 1989; 298: 1152–1157. 20 Haenni A, Lithell H. Urapidil treatment decreases 5 Pollare T, Lithell H, Berne C. A comparison of the plasma fibrinogen concentration in essential hyperten- effects of hydrochlorothiazide and captopril on glu- sion. Metabolism 1996; 45: 1221–1229. cose and lipid metabolism in patients with hyperten- 21 Samuelsson O et al. Diabetes mellitus and raised sion. N Engl J Med 1989; 321: 868–873. serum triglyceride concentration in treated hyperten- 6 Hansson L et al. Effect of angiotensin-converting sion-are they of prognostic importance? Observational enzyme inhibition compared with conventional ther- study. BMJ 1996; 313: 660–663. apy on cardiovascular morbidity and mortality in 22 UKPDS. Tight blood pressure control and risk of mac- hypertension: the Captopril Prevention Project rovascular and microvascular complications in type 2 (CAPPP) randomised trial. Lancet 1999; 353: 611–616. diabetes: UKPDS 38. BMJ 1998; 317: 703–713. 7 Ferrannini E et al. Insulin resistance in essential 23 UKPDS. UK prospective diabetes study 33: intensive hypertension. N Engl J Med 1987; 317: 350–357. blood glucose control with sulphonylureas or insulin 8 Pollare T, Lithell H, Berne C. Insulin resistance is a compared with conventional treatment and risk of characteristic feature of primary hypertension inde- complications in patients with type 2 diabetes. Lancet pendent of obesity. Metabolism 1990; 39: 167–174. 1998; 352: 837–853. 9 Pyo¨ra¨la¨ K. Relationship of glucose tolerance and 24 UKPDS. Efficacy of atenolol and captopril in reducing plasma insulin to the incidence of coronary heart dis- risk of macrovascular and microvascular compli- ease: results from two population studies in Finland. cations in type 2 diabetes: UKPDS 39. BMJ 1998; 317: Diabetes Care 1979; 2: 131–141. 713–720.

Journal of Human Hypertension