- Top of page
- Patients and Methods
Merging data from existing electronic patient records, and electronic hospital discharge and cause of death registers, is a fast and relatively inexpensive method for comparing different treatments with regard to clinical outcome. This study compared the effects of antihypertensive treatment with candesartan or losartan on cardiovascular disease (CVD) using Swedish registers. Patients without previous CVD who were prescribed candesartan (n=7329) or losartan (n=6771) for hypertension during 1999–2007 at 72 Swedish primary care centers were followed for up to 9 years. Both medications were given according to current recommendations, and there was no difference observed in achieved blood pressures. The authors have previously shown that candesartan lowered the risk of all CVD (primary composite end point) more so than losartan (adjusted hazard ratio, 0.86; 95% confidence interval, 0.77–0.96). Candesartan also had a significantly better effect with regards to reducing the development of heart failure, cardiac arrhythmias, and peripheral arterial disease. In the present analysis, the authors found that candesartan, compared with losartan, reduced the risk of all CVD, irrespective of sex, age, previous antihypertensive treatment, baseline blood pressure, and presence of diabetes. These clinical findings may reflect differences between candesartan and losartan in their binding characteristics to the angiotensin type 1 receptor. J Clin Hypertens (Greenwich). 2011;13:189–197. © 2010 Wiley Periodicals, Inc.
The extent of risk reduction in cardiovascular disease (CVD), which is obtained during pharmacologic treatment of hypertension, is directly related to the degree of blood pressure (BP) lowering.1–3 The risk for CVD would therefore be expected to be similar for similar BPs irrespective of the antihypertensive drug used.
Prospective randomized clinical trials have, however, reported differences in the incidence of CVD at similar BP reductions during antihypertensive treatment with drugs from different pharmacologic classes.4–6 There are at least 3 possible explanations for these findings. First, the BPs reported were office BPs, which may not necessarily represent potential 24-hour BP differences between treatments. Second, some drugs may have vascular protective effects, which give clinical benefits in addition to those expected from the achieved BP reductions. Third, a drug may have unwanted effects, which offset the positive effects of lowered BP.
There are few randomized trials that have compared the effects on CVD of drugs within the same pharmacologic class. This is probably because it is generally thought that it is unlikely that similar drugs have important differences with regard to clinical outcomes. The cost and time needed for such trials have also discouraged their initiation.
An alternative to a prospective randomized clinical trial is an observational register-based study. The possibility of merging data from existing electronic patient records, and electronic hospital discharge and cause of death registers, is a fast and relatively inexpensive method for comparing different treatments regarding clinical outcome. A further advantage is that such a study reflects the use of different drugs in real life, ie, without the artificial selection of investigators, patients, and rigorous treatment schedules that are characteristic of randomized clinical trials.
We have recently compared the effects of antihypertensive treatment with two angiotensin II type 1 (AT1) receptor blockers, candesartan and losartan, on CVD using existing electronic patient registers.7 Patients without previous CVD who were prescribed candesartan (n=7329) or losartan (n=6771) for hypertension in 1999–2007 at 72 Swedish primary care centers were followed for up to 9 years (median follow up, 2.0 years). Both drugs were given according to current recommendations, and there was no differences observed in the achieved BPs. We found, however, that candesartan lowered the risk of all CVD (primary composite end point) more so than losartan (adjusted hazard ratio [HR], 0.86; 95% confidence interval [CI], 0.77–0.96). Candesartan also had a significantly better effect with regard to the prevention of heart failure, cardiac arrhythmias, and peripheral arterial disease.
The present study is an extension of our previously reported investigation.7 The aim was to assess whether the risk reduction in all CVD, which was observed in the entire patient population during candesartan compared with losartan treatment, also existed in subgroups of patients, defined by baseline characteristics such as sex, age, previous antihypertensive treatment, baseline BP, and presence of diabetes.
Patients and Methods
- Top of page
- Patients and Methods
The study protocol was reviewed and approved by the Regional Research Ethics Committee in Uppsala, Sweden, and registered with ClinicalTrials.gov, number NCT00620178. The study methods have previously been described in detail.7 A summary is given below.
All primary care centers in Sweden use electronic patient records, which are regularly updated with information on patients’ disease history, symptoms, findings of physical examinations, laboratory data, diagnoses, and medication. International codes are used for diagnoses (International Classification of Diseases [ICD]) and medications (Anatomical Therapeutic Chemical classification system [ATC]). Patients are identified using their social security number, which is unique for each Swedish individual and provides their date of birth and sex. The social security number is also used for the identification of patients in many other registers, such as the electronic Swedish Hospital Discharge Register and Cause of Death Register. Specific information from electronic patient records and these national registers provided the basis for the study.
Men and women with hypertension, aged 18 years and older, who were prescribed candesartan (ATC: C09CA06, C09DA06) or losartan (ATC: C09CA01, C09DA01) from January 1, 1999, to December 31, 2007, at 72 primary care centers were eligible for the study. Exclusion criteria were prior CVD, treatment with drugs indicating CVD (vitamin K antagonists, digitalis glycosides, or nitrates), ongoing malignancy, or prescription of a renin-angiotensin system (RAS)–inhibiting drug, other than the study medications, within 1 week after index prescription. CVD was defined as heart failure (ICD-10: I50; ICD-9: 428), cardiac arrhythmias (ICD-10: I46-48; ICD-9: 427), peripheral artery disease (ICD-10: I70-71, I74; ICD-9: 440-441, 444), chronic ischemic heart disease (ICD-10: I20.9, I25.1; ICD-9: 413-414), myocardial infarction (ICD-10: I21-23; ICD-9: 410-411, 429), stroke (ICD-10: I61, I63-64, G45; ICD-9: 431-435), and unstable angina (ICD-10: I20.0; ICD-9: 411). Coronary revascularization procedures were also included in the definition of CVD. They were identified using the Nordic Medico-Statistical Committee (NOMESCO) Classification of Medical Procedures (NCMP) and Surgical Procedures (NCSP).
Information on smoking, an important risk factor for CVD, was unavailable in a substantial proportion of the electronic patient records and could not be used in the study. This limitation was partly overcome by recording and analyzing data on chronic lower respiratory disease (ICD-10: J40-45; ICD-9: 490-496), which is closely related to smoking. The prevalence of smoking in Sweden is among the lowest in the developed world (current smokers: men, 17%; women, 23%).8
The 72 participating centers were selected from primary care centers in Sweden, which used the most common information technology system for electronic patient records,9 aiming for a representative sample of Swedish primary care patients. All patient records were scanned using a special software system10 in order to identify the study population, ie, all patients who fulfilled the inclusion criteria and were free from exclusion criteria. The first prescription of the study medication within the study period was defined as the index prescription (treatment onset). Data used in the study were extracted and entered into a study database. No discrepancies were found when the database was manually compared with data from 100 patient records at 25 of the primary care centers. This analysis studied 5 pairs of subgroups of patients. These subgroups were defined by the following baseline characteristics: sex (men/women), age (65 and older/younger than 65 years), previous antihypertensive treatment (yes/no), baseline BP (high/low), and diabetes (yes/no). High BP was defined as systolic ≥160 mm Hg and/or diastolic ≥100 mm Hg, and low BP as systolic <160 mm Hg and diastolic <100 mm Hg. Diabetes was defined by the relevant ICD codes (ICD-10: E10–E14; ICD-9: 250) or the use of antidiabetic medication (ATC: A10).
BP measurements in Sweden are performed according to standardized methods, using the manual Korotkoff method or automatic measurements. The baseline value for BP was calculated as the mean of the last 3 measurements during the period from 15 months before index prescription until 14 days after. BP at 6 months was calculated as the mean of measurements 2 weeks to 6 months after the index prescription. From 12 months onward, the 6-month BPs were calculated as the mean of measurements from 6 months before to 6 months after the specific time point.
The primary end point of all CVD was a composite that consisted of CVD morbidity, CVD mortality, and coronary revascularization procedures.
The study database is owned and managed by the Department of Public Health and Caring Sciences, Uppsala University, Uppsala, Sweden. It contains data from the 72 primary care centers, the Swedish Hospital Discharge Register, and the Cause of Death Register, all merged using the patients social security number. The data were processed and analyzed by an independent statistical contract company.
The included patients were eligible for analysis as long as they continued the study medication (candesartan or losartan). The observation period ended on the date when the patient died, started a new RAS-inhibiting drug, was lost to follow-up, or on December 31, 2007.
All descriptive data are given as mean or proportion. Proportional hazards regression was used to compute HRs with 95% CIs and P values for each subgroup separately. P values for interaction were calculated using the same statistical model with treatment group and factor of interest (ie, sex, age, previous antihypertensive treatment, baseline BP, and diabetes), and the interaction between treatment and factor of interest, added as independent factors into the model. The P value for the interaction is presented for each model. Kaplan–Meier curves were used to depict events during follow-up. If a patient had more than one end point, only the first was used in the survival model. Although the maximum observation was 108 months, the X-axis for the plots was truncated at 90 months because of low number of observations after this time.
All proportional hazard regressions were performed unadjusted, enabling comparison of the HR values between subgroup models.
A propensity score matching analysis11 of the eligible patients was performed as a sensitivity analysis. Patients fulfilling the inclusion and exclusion criteria where matched regarding age, sex, inclusion year, systolic BP, diastolic BP, and diabetes diagnosis, and whether previously treated with antihypertensive drugs, statins, antithrombotic drugs, thiazides, β-blockers.
- Top of page
- Patients and Methods
There were 24,943 eligible patients, of whom 11,942 (47.9%) were prescribed candesartan and 13,001 (52.1%) losartan. Of 10,843 patients with exclusion criteria, 4606 (42.5%) belonged to the candesartan group and 6237 (57.5%) to the losartan group. The remaining study population consisted of 14,100 patients; 7329 were treated with candesartan and 6771 with losartan. During the observation period, 2015 patients were discontinued from the candesartan-treated group and 2123 from the losartan group. The proportion of discontinued patients was less (P<.0001) among patients treated with candesartan (27.5%) compared with those treated with losartan (31.4%). The patient flow is summarized in Table I.
Table I. Patient Flow and Number of Patients
|Patients with inclusion criteria|
| Women or men||11,942||13,001|
| Aged ≥18 y|
| Prescribed candesartan or losartan in the time period|
| January 1, 1999–December 31, 2007|
|Patients with exclusion criteria||4606||6237|
| Previous cardiovascular disease||4144||5792|
| On-going malignancy||379||386|
| Prescribed other renin-angiotensin system inhibitor within 1 week of index prescription||83||59|
|Discontinued patients (% of study population)||2015 (27.5)||2123 (31.4)|
| Converted to another renin-angiotensin system inhibitor||788||939|
| Cessation of study drug||1071||1029|
Patient baseline characteristics were similar, or only slightly different, in patients treated with candesartan or losartan (Table II). Approximately half of the patients in each treatment group had a history of non–CVD-related hospital care and the number of days in hospital were identical in both groups (5.9 days). The number of patients hospitalized prior to inclusion because of chronic obstructive pulmonary disease (COPD) and asthma were identical (2.1% in each group). There were no differences when separately comparing COPD and asthma data (data not shown).
Table II. Baseline Characteristics
| ||Candesartan (n=7329)||Losartan (n=6771)|
|Age, y||61.7 (11.7)||62.4 (11.5)|
|Women, No. (%)||4109 (56.1)||3723 (55.0)|
|Diabetes, No. (%)||1112 (15.2)||1215 (17.9)|
|Chronic lower respiratory disease, No. (%)||156 (2.1)||142 (2.1)|
|Systolic blood pressure, mm Hg||159.9 (19.4)||159.0 (19.9)|
|Diastolic blood pressure, mm Hg||90.4 (10.3)||89.3 (10.3)|
|Body mass index, kg/m2||30.2 (5.4)||30.2 (5.3)|
|Serum creatinine, μmol/L||83.9 (19.2)||84.1 (20.9)|
|Serum cholesterol, mmol/L||5.7 (1.1)||5.7 (1.0)|
|B-glucose, mmol/L||6.2 (2.3)||6.3 (2.4)|
|Thiazides, No. (%)||1087 (14.8)||848 (12.5)|
|β-Blockers, No. (%)||1883 (25.7)||1605 (23.7)|
|Calcium antagonists, No. (%)||1104 (15.1)||968 (14.3)|
|Angiotensin-converting enzyme inhibitors, No. (%)||1459 (19.9)||1361 (20.1)|
|Angiotensin receptor blockers, No. (%)||120 (1.6)||101 (1.5)|
|History of hospital care, No. (%)||3560 (48.6)||3286 (48.5)|
At median time of follow-up (2 years), the mean BP was similar in patients treated with candesartan or losartan, both in the entire study population and in the different subgroups (Table III).
Table III. Mean SBP and DBP at Baseline and at Median Follow-Up (2 y)
| ||BP at Baseline SBP/DBP, mm Hg||BP at Median Follow-Up SBP/DBP, mm Hg|
|All patients||160/90 (5860)||159/89 (5370)||148/84 (2558)||148/83 (2510)|
|Previously untreated||161/90 (3000)||160/89 (2761)||150/84 (1221)||149/83 (1140)|
|Previously treated||159/91 (2860)||158/90 (2609)||146/84 (1337)||146/83 (1370)|
|Men||161/89 (3326)||160/88 (3038)||149/83 (1459)||149/82 (1388)|
|Women||159/92 (2534)||157/91 (2332)||147/85 (1099)||146/84 (1122)|
|≥65 y||165/87 (2352)||164/87 (2308)||153/81 (1047)||151/81 (1140)|
|<65 y||156/93 (3508)||155/91 (3062)||145/85 (1511)||145/85 (1370)|
|Diabetics||158/87 (988)||156/86 (1074)||148/81 (467)||148/81 (558)|
|Nondiabetics||160/91 (4872)||160/90 (4296)||148/84 (2091)||147/84 (1952)|
|High baseline BPa||173/94 (3211)||173/94 (2778)||153/84 (1307)||153/83 (1119)|
|Low baseline BPb||144/85 (2649)||144/84 (2592)||142/83 (974)||142/82 (1040)|
The observation period comprised a total of 36,339 patient-years: 18,642 patient-years of candesartan treatment and 17,697 patient-years of losartan treatment. The median time of follow-up was 2.0 years. In all patients, the incidence of a first CVD event (all CVD) was significantly lower in the candesartan group (n=575) compared with the losartan group (n=676), and the HR was 0.79 (95% CI, 0.71–0.89; P<.001) (Table IV). The incidence of all CVD was also significantly lower in patients treated with candesartan than losartan in all subgroups, except in patients 65 years and older and in those with the highest BPs at baseline (SBP ≥160 mm Hg and/or DBP ≥100 mm Hg). The 95% CI of the HR in each pair of subgroups were overlapping, and the P values for interaction/heterogeneity were all nonsignificant (>.05). Kaplan–Meier curves for cumulative incidence in the subgroups are shown in the Figure.
Table IV. Clinical Outcomes of Antihypertensive Treatment in the Entire Patient Population (All Patients) and in Subgroups
Figure Figure. Kaplan-Meier curves of the composite end point (all cardiovascular disease) for the entire patient population and for subgroups of patients. BP indicates blood pressure.
Download figure to PowerPoint
During the observation period, 3.2% of the patients treated with candesartan and 3.9% of those treated with losartan had a diagnosis of chronic lower respiratory disease. The propensity score matching analysis identified two groups of 5860 patients with equal baseline characteristics. The HR was 0.79 (95% CI, 0.672–0.921) (P<.001) for the time to first CVD event (all CVD).
- Top of page
- Patients and Methods
The analyses in this study strongly suggest that hypertensive patients who receive primary care treatment with candesartan have a lower incidence of all CVD than those treated with losartan, irrespective of sex, age, previous antihypertensive treatment, baseline BP, and whether they have diabetes. A previous analysis of the entire study population showed an adjusted risk reduction in all CVD of 14% with candesartan compared with losartan treatment.7 In the primary preventive setting, the Losartan Intervention for Endpoint Reduction in Hypertension (LIFE) study demonstrates a BP-independent risk reduction of 13% in patients with hypertension and left ventricular hypertrophy, using an end point consisting of nonfatal myocardial infarction, stroke, and cardiovascular mortality, thus demonstrating similar risk reduction levels as in the present study despite similar BP reduction.5 It should be pointed out, however, that the comparison was between losartan and atenolol.5 The clinical importance of more aggressive RAS inhibition has been demonstrated in the Effects of High-Dose Versus Low-Dose Losartan on Clinical Outcomes in Patients With Heart Failure (HEAAL) study, where losartan 50 mg was compared with losartan 150 mg when treating patients with heart failure.12 A 10% risk reduction was demonstrated for heart failure hospitalizations and deaths using the higher dose of losartan.
The use of a wider composite end point in a real-life setting, as in the present study, makes the numeric comparison of HR values from randomized clinical trials difficult. The adjusted risk reduction of 14.4% may seem modest but resulted in important differences regarding the use of health care resources.13 The effect was mainly due to a marked risk reduction in heart failure (36%), cardiac arrhythmias (20%), and peripheral artery disease (39%). There was also a lower risk with candesartan treatment, although statistically nonsignificant, for all other specific CVD events (eg, chronic ischemic heart disease, myocardial infarction, stroke, hospitalization for unstable angina) and for elective coronary vascularization and cardiovascular mortality. Thus, the lower risk of CVD during antihypertensive treatment with candesartan compared with losartan was consistent, as it was apparent in all and in specific CVD events irrespective of a number of patient characteristics. Although the study was carried out in Sweden, these results should also be valid for other countries with similar patient populations and medical care.
The subgroups selected represent large groups of patients of general interest. By limiting the analysis to all CVD, the statistical power was kept high, as the number of patients with an event was large in each of the 10 subsets of patients in both treatment groups.
There was no significant difference in risk reduction with candesartan compared with losartan in any of the subgroup pairs. The 95% CI for the HR within a pair was overlapping, and the P value for the interaction/heterogeneity test was nonsignificant. The fact that the risk reduction in one subgroup (eg, patients younger than 65 ) was significant, while it was nonsignificant in the contrasting subgroup (patients 65 years and older), should not be misinterpreted as indicating a difference in the favorable effect of candesartan compared with losartan.
The considerable reduction in risk for developing heart failure (primary prevention) in patients treated with candesartan, compared with losartan, deserves additional comment. This finding is consistent with the unequivocal evidence of candesartan’s favorable effect in heart failure treatment and the less-convincing documentation for losartan.14,15 Candesartan reduces morbidity and mortality in patients with chronic heart failure and decreased left ventricular systolic function, whether or not it is administered together with an optimal dose of an ACE inhibitor.
The lower incidence of CVD in candesartan- compared with losartan-treated patients, in spite of similar recorded BP, may seem unexpected as the reduction in CVD during treatment of hypertension is generally closely related to the BP reduction.1–3 However, the results may be explained by candesartan’s tighter and longer-lasting binding to the AT1 receptor when compared with that of losartan.16,17 This translates into an effective AT1 receptor blockade and BP reduction for 24 hours after candesartan, while losartan has a shorter duration of action.18 In the present study, both drugs were prescribed once daily, and BP was measured during office hours, ie, during the daytime. It may therefore be argued that the 24-hour BP control may have been better in the candesartan compared with the losartan group, in spite of similar recorded BP values. This possibility cannot be excluded considering the extent of missing office BP measurements and the lack of BP recordings during the dosage interval (24 hours). It is also possible that effects of AT1 receptor blockade, other than the reduction of elevated BP, eg, organ protection and reduction of vascular and myocardial hypertrophy, also depend on the extent and duration of receptor blockade. This may explain why the prevention of CVD was more pronounced in patients treated with candesartan compared with those treated with losartan.
Another possible explanation for the lower incidence of CVD during treatment with candesartan is the possibility that the baseline CVD risk was lower in the candesartan compared with the losartan group. This is unlikely, however, since the HR were similar without and with adjustments for a number of known and unknown baseline risk factors in the statistical analysis.7 Furthermore, the consistent results of the propensity score matching analysis with the primary analysis support that the observed differences are accurately adjusted for and that the results are robust.
Although data on smoking were missing for the majority of patients and therefore not included in the analysis, it is unlikely that there were important differences in smoking habits between the treatment groups. This conclusion is supported by the observation of similar proportions of patients with chronic lower respiratory disease, which is closely related to smoking, in the candesartan and losartan treatment groups. Furthermore, there is no reason to suspect that the prescription of candesartan or losartan was dependent on the patients’ smoking habit. It is also notable that the proportion of patients with a history of hospital care for diseases other than CVD was similar in both treatment groups at baseline, indicating well-balanced treatment groups.
The method used to exclude undetected CVD in our study was critical in order to avoid confounding differences between the groups at baseline. Prior to inclusion, we did not observe any increased number of COPD and asthma hospitalizations in the losartan group, which could have indicated a higher risk of undetected heart failure patients. Patients were not only excluded because of recorded diagnoses, but also when treated with drugs indicating CVD.7 This further strengthens the method of detecting undiagnosed patients. Additionally, we have adjusted for drugs at baseline, which might be associated with the treatment of any undetected CVD, eg, differences in treatment with β-blockers, statins, antithrombotics, and diuretics.
It was decided in this study to limit the BP data for candesartan- and losartan-treated patients in the 10 subgroups and to report BP levels at baseline and at 2 years (time of median follow-up). We have previously shown, however, that the recorded BPs were similar in the entire patient populations treated with candesartan or losartan throughout the observation period and that little change in BP levels occurred after the first 2 years.7
The use of candesartan and losartan in this study was consistent with the manufacturers’ recommendations, as was the addition of thiazides.7 It is also realistic to assume that treatment was tailored for each patient in order to achieve the goal BP, which is recommended in Sweden and consistent with international guidelines (<140/90 mm Hg, or <130/85 mm Hg in patients with diabetes or end organ damage). Similar office BP values were therefore anticipated in patients treated with candesartan or losartan. However, several randomized studies using fixed doses, or forced-dose titration, have demonstrated greater BP reduction with candesartan compared with losartan.19,20
At baseline, the candesartan group had more frequent use of thiazides and β-blockers. In the losartan group, more usage of statins, antithrombotics and oral glucose–lowering drugs was observed. The treatment at baseline became similar when excluding the diabetes patients. The higher prevalence of diabetes patients in the losartan group may therefore have been the explanation for the higher use of CVD-preventive strategies (eg, statins, antithrombotics) and glucose-lowering drugs. On the other hand, physicians may be reluctant to treat diabetes patients with β-blockers due to the increased risk of hypoglycemia and thiazides due to the discussions regarding negative metabolic effects. Additionally, when adjusting for treatment differences at baseline, both using Cox linear regression model and propensity score matching, the results remain similar.7
The use of thiazides was more frequent during follow-up in the losartan compared with the candesartan treatment group.7 This should be to losartan’s advantage with respect to the prevention of CVD as indicated by observations in the LIFE study. Hydrochlorothiazide therapy, concomitant to losartan, was associated with reduced CVD morbidity and mortality21 and with improved regression of electrocardiographic left ventricular hypertrophy.22
Irrespective of whether the achieved 24-hour BP control in this study was similar in both treatment groups, it can be concluded that treatment of hypertension with candesartan reduced the risk of a first CVD event significantly more than treatment with losartan. This was the case when the two drugs were prescribed as recommended with respect to dosing and combinations with other antihypertensive agents.
Since the comparisons of CVD incidence between candesartan- and losartan-treated patients included 10 subgroups of patients, representing different baseline CVD risk, unadjusted analyses were used for this study. When the incidence of all and specific CVD events in the entire population of patients treated with candesartan or losartan were compared, survival analyses were made both with and without adjustments for possible differences in baseline CVD risk.7 Overall, results from unadjusted and adjusted analyses were similar, and the conclusions did not change. For example, the unadjusted HRs for all CVD was 0.79 (95% CI, 0.71–0.89), while it was 0.86 (95% CI, 0.77–0.96) in the primary analysis that was adjusted for baseline age, sex, diabetes, and index year. A sensitivity analysis for different discontinuation rates was omitted in the present analysis, as we previously found that the greater discontinuation rate of patients treated with losartan than candesartan in the entire patient population did not influence the results regarding all CVD.7