Journal of Human Hypertension (2002) 16, 439–444  2002 Nature Publishing Group All rights reserved 0950-9240/02 $25.00 www.nature.com/jhh ORIGINAL ARTICLE Long-term persistence with antihypertensive drugs in new patients

E Degli Esposti1, A Sturani2, M Di Martino3, P Falasca1, MV Novi1, G Baio4, S Buda3 and M Volpe5 1Health Directorate, Ravenna Local Health Unit, Ravenna, Italy; 2Hypertension Unit, Department of Nephrology, S.M. delle Croci Hospital, Ravenna, Italy; 3CliCon S.r.l., Health, Economics and Outcomes Research, Ravenna, Italy; 4University of Siena, R.M. Goodwin Faculty of Economics, EPRIS Research Group, Siena, Italy; 5Chair of Cardiology, Department of Experimental Medicine and Pathology, University ‘La Sapienza’, Rome, Italy

The objective of this study was to investigate stay-on- 7.6% were restarted on a treatment in the third year. Per- therapy patterns over 3 years among patients pre- sistence with treatment was influenced by: age of scribed different classes of antihypertensive drugs for patient (persistence rate increasing proportionately with the first time. A retrospective analysis of information advancing years), type of drug first prescribed recorded in the drugs database of the Local Health Unit (persistence rate higher with angiotensin II antagonists, of Ravenna (Italy) was carried out on 7312 subjects progressively lower with ACE-inhibitors, beta-blockers, receiving a first prescription for , beta-blockers, calcium channel blockers and diuretics), gender of calcium channel blockers, angiotensin-converting patient (persistence was better in males), age of general enzyme (ACE) inhibitors or angiotensin II antagonists practitioner (GP) (the younger the GP, the better the per- between 1 January and 31 December 1997. Patients sistence rate) and gender of GP (better stay-on-therapy were followed up for 3 years. All prescriptions of antihy- rate with male GP prescribing). In the case of patients pertensive drugs filled during the follow-up periods treated continuously, mean daily dose increased pro- were considered. The patients continuing or discontinu- gressively over the 3 years. With adequate markers, ing the initial treatment, the duration of treatment, and helpful data can be collected from prescription claims the doses taken were all calculated, as well as main fac- databases for the purpose of monitoring the persistence tors influencing the persistence rate. The drugs pre- of patients in continuing their , and the qual- scribed were predominantly ACE-inhibitors, followed by ity of antihypertensive treatment in a general practice calcium channel blockers, diuretics, beta-blockers and setting. angiotensin II antagonists. A total of 57.9% of patients Journal of Human Hypertension (2002) 16, 439–444. DOI: continued their initial treatment during the 3-year follow- 10.1038/sj/jhh/1001418 up period, 34.5% discontinued the treatment, whilst

Keywords: hypertension; antihypertensive drugs; administrative databases

Introduction year and only two-thirds of those who remain under care take enough medication to control their blood Persistence with treatment is a crucial element in pressure.5,6 Many factors can influence the extent of determining the success of any long-term therapy. adherence to antihypertensive therapy. Persistence The full benefit of antihypertensive treatment 1–3 rates are inversely correlated with the number of observed in randomised clinical trials can be drugs prescribed, complexity of dosage regimen, and obtained only when hypertensive patients keep tak- with the cost of drug; whilst are directly correlated ing their medication as a matter of regular habit. with the tolerability of the treatment, and with a Frustratingly, low compliance remains a major strong and trusting physician–patient relationship.7 therapeutic problem in subjects treated for the first The aim of the present study was to investigate 4 time with antihypertensive drugs. Many patients stay-on-therapy patterns over 3 years, among with hypertension drop out of care inside the first patients who were receiving different classes of anti- hypertensive drugs for the first time.

Correspondence: E Degli Esposti, MD, c/o CliCon S.r.l., Via San Materials and methods Vitale, 5, 48100 Ravenna, Italy. E-mail: edegliespostiȰiol.it Received 15 November 2001; revised and accepted 27 February This is a record linkage study, in which a retrospec- 2002 tive analysis was conducted using the database of Utilisation of antihypertensive drugs E Degli Esposti et al 440 dispensed drugs held by an Italian Local Health Unit received at least three prescriptions for drugs acting (LHU of Ravenna, Italy). LHUs are government on the cardiovascular system (ATC code: C01), and organisations which provide healthcare services to as asthmatic or having chronic obstructive pulmon- the entire population living in a given area and work ary disease (COPD) if they had received at least three toward the achievement of public health objectives prescriptions for respiratory drugs (ATC code: R03). pursued by the National Health Service (NHS). All The presence of cardiovascular diseases (acute myo- drug prescriptions written for the residents of cardial infarction, which is ICD 9 code 410, coronary Ravenna (approximately 350 000 inhabitants), tot- disease, ICD 9 code 411–414, heart failure, ICD ally or partially reimbursed by the NHS, are docu- 9 code 428–429) was determined by analysing hos- mented in the database. An automatically assigned pital admissions in the 5 years preceding inclusion national health number identifies each subject. in the study. Since 1 January 1996, the data file has been gather- ing all prescriptions dispensed, each bearing the NHS number, birth date and gender of the single Analysis of persistence patient. Also recorded, moreover, are the date of pre- We identified three patterns of persistence attribu- scription, the general practitioners (GPs) identifi- table to patients: cation code, the Anatomical-Therapeutic-Chemical (ATC) code of the drug/s, number of packs dis- (1) continuers—those who received at least two pre- pensed, number of tablets in each pack, milligrams scriptions for antihypertensive drugs during of active principle in each tablet, and cost per pack. each year of the 3-year follow-up period; The result is a database providing a long-term record (2) discontinuers—those who received at least two of all drugs prescribed for each patient. All infor- prescriptions for antihypertensive drugs during mation on antihypertensive drugs dispensed is col- the first year (lost in second year) or during the lected on a routine basis to enable computation of first and the second year (lost in third year) of the reimbursement payable to the pharmacists, who the follow-up period; supply the drugs to patients. (3) retreaters—those who received at least two pre- All patients enrolled in the present study were at scriptions for antihypertensive drugs during the least 20 years of age, and had been prescribed anti- first and the third year of the follow-up period, hypertensives for the first time in the period having suspended their therapy during the between 1 January 1997 and 31 December 1997, second year. namely diuretics (ATC code C03), beta-blockers (ATC code C07), calcium channel blockers (ATC For continuers, we also calculated the mean daily code C08), angiotensin-converting enzyme (ACE)- dose (MDD) (tablets/day) taken by each patient as inhibitors (ATC code C09A), or angiotensin II antag- the number of tablets of antihypertensive drugs pur- onist, ie losartan (ATC code C09C), which was the chased each year, divided by 365. only angiotensin II antagonist available in Italy dur- ing that period. Patients were excluded if they had Statistics received a prescription for any during the previous 12 months. Results are expressed as mean values (± standard Subjects were excluded if on the date of their first deviation). We performed all statistical analyses prescription they had been dispensed two or more using SPSS-Windows version 10.0 (SPSS Inc, classes of antihypertensive agents, or a combination Chicago, USA), considering P values less than 0.05 product including two different classes of antihyper- as significant. tensive drugs. Also excluded were those who had In the case of continuers, the hypothesis that received a single prescription or only two prescrip- mean daily dose is unaffected by 1 year’s stay-on- tions for the antihypertensive drugs considered in therapy was verified, in the case of continuers, using the present study, indicating a course of treatment a nonparametric Friedman test for dependent lasting less than 7 days. samples.8 Persistence was assessed using survival The observation period for each patient was 3 analysis methods.9 Having tested the proportional years starting from the date of the first prescription. hazards assumption, a Cox regression model was Patients who died or moved away during the follow- used: the number of days elapsing from the first to up period were excluded. Prescriptions for other the last prescription was taken as an estimate of the antihypertensive drugs filled during the follow-up duration of treatment, and the follow-up period was period, used in combination with or as substitutes limited to 1005 days. The following independent for the active principles studied in this paper, were variables were considered: age of patient, gender of also taken into account. patient, comorbidity if any, previous hospitalis- On the basis of prescriptions received in the year ations if any, class of drug prescribed at enrolment, preceding the period of inclusion in the study, administrative district, age and gender of GP, num- patients were defined as diabetic if they had ber of patients on the GPs list. The variables to be received at least three prescriptions for antidiabetic included in the model were selected thereafter drugs (ATC code: A10), as cardiopathic if they had adopting a stepwise method,10 which in the pro-

Journal of Human Hypertension Utilisation of antihypertensive drugs E Degli Esposti et al 441 cesses of removal uses the Wald statistic (P- Table 1 Characteristics of the patients = = value(in) 0.05, P-value(out) 0.10). Mean Male Co- Previous Mean age age morbidity events of GP Results (years) (%) (%) (%) (years)

Applying the inclusion and exclusion criteria noted Diuretics 65.4 ± 16.8 32.5 15.0 13.6 46.5 ± 7.4 above (Figure 1), 7312 patients were identified, Beta-blockers 54.4 ± 13.9 38.5 4.2 9.9 47.1 ± 7.8 numbering 3154 males (equivalent to 43.1%) and Calcium 62.3 ± 15.2 50.1 8.6 17.2 46.6 ± 7.6 4158 females (equivalent to 56.9%), aged between ACE-inhibitors 61.1 ± 14.8 47.3 7.8 12.8 46.4 ± 7.4 20 and 100 years. The class of antihypertensive drug Angiotensin II 60.5 ± 13.9 42.4 7.1 13.6 47.2 ± 7.4 most frequently selected as first choice was that of antagonist ACE-inhibitors (prescribed for 2418 patients, equiv- alent to 33.1% of the population considered in the study), followed by calcium channel blockers (prescribed for 1882 patients, equivalent to 25.7% in staying on therapy was influenced significantly of the population), diuretics (prescribed for 1648 by five variables: the age of the patient, with older patients, equivalent to 22.6% of the population), patients tending to continue longer with treatment beta-blockers (prescribed for 1166 patients, equival- than younger patients; the drug on which the patient ent to 15.9% of the population) and angiotensin II was started, with those who were prescribed an antagonists (prescribed for 198 patients, equivalent angiotensin II antagonist tending more to persist to 2.7% of the population). The characteristics of the with antihypertensive therapy than those taking an five cohorts of patients, identified according to the ACE-inhibitor, a beta-blocker, a calcium channel drug on which they were started, are indicated in blocker or a ; the gender of the patient, with Table 1. As regards the extent to which individuals male patients tending to show more persistence than persisted with their treatment, 4231 patients female patients; the age of the GP, inasmuch as (equivalent to 57.9%) continued, 2525 patients patients registered with a young GP appear to stay (equivalent to 34.5%) discontinued (1999 after the on therapy longer than patients registered with an first year and 526 after the second), and 556 patients older GP; and the gender of the GP, inasmuch as (equivalent to 7.6%) were restarted on a therapy patients registered with a male GP appear to stay on during the third year of the follow-up period, having therapy longer than patients registered with a female suspended treatment during the second year. The GP. The first of these two variables (age of patient stepwise process (Table 2) showed that persistence and class of drug initially prescribed) are those hav-

Figure 1 Flow chart of excluded patients.

Journal of Human Hypertension Utilisation of antihypertensive drugs E Degli Esposti et al 442 Table 2 Cox regression model: predictors of persistence Table 3 Mean daily dose (MDD) purchased per year in continuers and number of antihypertensive classes ␤a s.e. Wald Hazard ratio statistic (CI 95%)b Year 1 Year 2 Year 3

Patient age** −0.024 0.001 408.4 0.976 (0.974 to MDD (tablets/day) 0.8 ± 0.56 0.95 ± 0.67 1 ± 0.68 0.978) Mean no. of classes 1.68 1.52 1.54 Class of drug – 138.9 – prescribed (angiotensin II antagonist)** ACE inhibitors* 0.456 0.14 10.6 1.577 (1.198 to 2.076) Beta-blockers** 0.626 0.143 19.2 1.869 (1.414 to Table 4 Mean daily dose (MDD) and number of antihypertensive 2.472) classes purchased per year according to the starting drug in con- Calcium channel 0.729 0.141 26.9 2.073 (1.574 to tinuers blockers** 2.731) Diuretics** 0.965 0.141 47.1 2.624 (1.992 to Year 1 Year 2 Year 3 3.457) Patient gender −0.112 0.037 9.4 0.894 (0.832 to (female)* 0.961) Diuretics GP age* 0.006 0.002 9.9 1.006 (1.002 to Average no. of classes 1.75 1.52 1.55 prescribed 1.011) ± ± ± GP gender (female)* −0.093 0.043 4.6 0.911 (0.836 to Mean MDD (tabs/day) 0.69 0.5 0.9 0.6 1.02 0.69 0.992) Beta blockers Average no. of classes 1.6 1.46 1.5 Variables are listed by order of entry in the model; in brackets is prescribed indicated the reference category of categorical variables (hazard Mean MDD (tabs/day) 0.73 ± 0.47 0.92 ± 0.91 0.98 ± 0.76 = ratio 1). Calcium channel a b Related to 1 year age increment. Adjusted for other variable in blockers Ͻ Ͻ ␹2 = Ͻ table. *P 0.05, **P 0.001. Global 550.55, P 0.001. Average no. of classes 1.68 1.54 1.57 prescribed Mean MDD (tabs/day) 0.97 ± 0.73 1.06 ± 0.75 1.12 ± 0.74 ing the greatest influence on persistence. When per- sistence is analysed on the basis of the drug pre- ACE-inhibitors Average no. of classes 1.64 1.52 1.54 scribed, remaining constant on the effect of other prescribed covariates (Figure 2), it is shown that patients who Mean MDD (tabs/day) 0.76 ± 0.47 0.92 ± 0.54 1 ± 0.58 were started on angiotensin II antagonists display a Angiotensin II significantly greater persistence with antihyperten- antagonists sive therapy than those who were started on the Average no. of classes 2 1.74 1.58 other four classes of drug. prescribed ± ± ± Among the 4231 patients classified as continuers, Mean MDD (tabs/day) 0.8 0.39 1 0.61 1.12 0.74 mean daily dose increased significantly over the 3 years of treatment (Friedman statistic = 625.867, P Ͻ 0.001), whereas the number of classes of drug prescribed fell significantly from the first to the second year, remaining steady thereafter from the second to the third year (Table 3). This pattern was similar across the five groups of patients identified on the basis of the drug prescribed at enrolment (Table 4). The analysis conducted on the basis of MDD (Table 5) showed that the cumulative percent- age of patients taking a MDD р1 tablet per day dropped substantially from 75.5% to 60.2% during 3 years of follow-up; correspondingly the cumulat- ive percentage of patients taking a MDD Ͼ1 tablet per day increased from 24.5% to 39.8%.

Table 5 Mean daily dose (MDD) distribution (percentage values)

Tables per day Year 1 Year 2 Year 3

Ͻ0.5 31.9 21.9 18.1 0.5 р tab Ͻ 1 43.6 44.4 42.1 р Ͻ Figure 2 Cox regression model, persistence with treatment by 1 tab 2 20.5 27.1 31.4 у class of drug (adjusted for age and gender of patient, and age and 2 4.0 6.6 8.4 gender of GP).

Journal of Human Hypertension Utilisation of antihypertensive drugs E Degli Esposti et al 443 Discussion mented in various real-world clinical practice situ- ations.18–21 The clinical, social, and economic importance of tre- It is difficult to sustain the hypothesis that drugs ating hypertension requires the implementation of can be really effective in controlling blood pressure tools for monitoring the use of antihypertensive when used in a range of doses lower than those rec- drugs so that they will give an understanding as to ommended. Although there is no positive evidence, what extent data from randomised and controlled we suppose that patients who repeat prescriptions trials are applicable in real clinical practice. It is a on a regular basis will be taking their medication, basic factor of clinical trials that, to be effective in and this is a reasonable assumption in the case of achieving blood pressure control, antihypertensive treatment for a chronic condition such as hyperten- therapy is given in adequate doses, in a continuous sion. The easiest alternative to a true evaluation is manner, and for a long period of time. With antihy- to investigate directly by asking whether or not pertensive drugs, accordingly, patient stay-on-ther- patients are actually taking the prescribed drugs. apy plays a crucial role in obtaining normotension The findings may be encouraging, but not so con- 22,23 and consequently lowering cardiovascular mor- clusive as might be expected. In our opinion, the bidity and mortality. In clinical practice, it is diffi- analysis of drug databases according to the pro- cult to have a large amount of information on blood cedures described above could provide an effective pressure control rate unless the cohort of subjects tool in evaluating patient stay-on-therapy and the considered is small, whereas utilising claims data- probable efficacy of antihypertensive treatment. bases, it becomes possible to gather data on persist- The second dimension characterising the question ence with antihypertensive therapy prescribed in a of compliance includes factors that influence per- whole population, as already shown.11,12 The issue sistence with treatment. Perseverance in matters of of stay-on-therapy can be seen as having at least two therapy depends on the complex relationship that dimensions. The first dimension is related to the exists between doctor and patient, and accordingly, proportion of patients persisting with their pre- account was taken of the information held in our scribed antihypertensive drugs, and the second database, which, albeit not exhaustive in respect of dimension is given by factors influencing its rate. all the potential compliance-determining factors in Our study showed that some 60% of patients who play, nonetheless allows an evaluation of the spe- were prescribed antihypertensives for the first time cific elements characterising two aspects of the had stayed on therapy for the duration of the follow- relationship. Firstly, one of the factors having great- est influence is the age of the patient: younger up. Moreover, the protracted follow-up period patients display a generally poorer rate of persist- enabled us to observe that around 8% of patients ence where antihypertensive treatment is con- in the study restarted on antihypertensives after a cerned, as previously reported,24 and this is more lengthy break. Other studies conducted on the basis evident in females than in males. Secondly, but no of prescription databases assess patients persisting less importantly, persistence is influenced by the with treatment at percentages ranging between 41 11,13–17 class of drug on which the patient is started. The and 86%. These results however are not effec- long-term persistence rate of patients who started tively comparable since some of the studies antihypertensive treatment on losartan, the only included new and established patients together, fol- 11,13,16 angiotensin II antagonist available in Italy at that lowed up for periods of less than 12 months, time, was significantly higher when compared with whereas others related only to new patients, fol- 14,15 those who started on ACE-inhibitors, beta-blockers, lowed up for a period of 12 months, or 54 calcium channel blockers, and diuretics respect- 17 months. The gradual increase in the average num- ively. Although this result could have been influ- ber of tablets taken over 3 years by continuers (from enced by the relative small sample size of patients 0.8 to 1 tablet/day), and the reduction in number of treated with angiotensin II antagonists, differences individuals who took less than half a tablet per day displayed are highly significant even after a 3-year (from 31.9% to 18.1% of patients treated), reflects follow-up, suggesting that the choice of initial drug in clinical practice that which in randomised con- has a significant bearing on persistence. A similar trolled trials (RCTs) is the titration procedure, pur- result has been reported previously in a 12 months sued in an attempt to achieve the target of normoten- follow-up study.12 sion. The clear difference is in the time over which The database we used in this study does not take this process takes place: generally a few weeks in into account baseline and subsequent blood press- the case of RCTs and many months (if not years) in ure levels, any appearance of side effects, or clinical clinical practice. Consequently, the presence of an outcomes. Therefore, the reasons that are able to appreciable percentage of patients who suspend explain the relationship between the persistence treatment or are undertreated, as a result of continu- with antihypertensive treatment, and the conditions ing to take antihypertensives in doses significantly of the patient, or the therapeutic decision made by lower than those usually adopted in RCTs, in some GP, are not yet clear. This represents the major limi- cases for extended periods, could help to explain the tation of our analysis. However, it can be assumed low percentage of normotensive patients docu- that drug efficacy and tolerability play a critical role

Journal of Human Hypertension Utilisation of antihypertensive drugs E Degli Esposti et al 444 in the persistence with antihypertensive treatment, 10 Mignani S, Montanari A. Appunti di analisi statistica especially when considering that physicians would multivariata, societa` editrice Esculapio: Bologna, not prolong a therapy without evidence of toler- 1997. ability or efficacy. In randomised and controlled 11 Jones JK et al. Discontinuation of and changes in the clinical trials, the angiotensin II antagonist losartan treatment after start of new courses of antihypertensive drugs: a study of a United Kingdom population. BMJ showed a higher tolerability compared with that 25 1995; 311: 293–295. observed with other antihypertensive classes. The 12 Bloom BS. Continuation of initial antihypertensive tolerability profile of a drug might therefore be one medication after 1 year of therapy. Clin Ther 1998; 20: of the reasons to clarify the results of this study, but 1–11. this remains an interesting hypothesis that requires 13 Bailey JE, Lee MD, Somes GW, Graham RL. Risk fac- to be confirmed. tors for antihypertensive medication refill failure by In conclusion, we believe that a chronological patients under Medicaid managed care. Clin Ther reading of prescriptions for antihypertensive drugs 1996; 18: 1252–1262. represents a useful tool in assessing drug utilisation 14 Okano GJ et al. Patterns of antihypertensive use among from an epidemiological point of view and for the patients in the US Department of defense database initially prescribed an angiotensin-converting enzyme detection of key factors influencing persistence inhibitor or . Clin Ther 1997; with treatment. 19: 1433–1445. 15 Monane M et al. The effects of initial drug choice and comorbidity on antihypertensive therapy compliance: Acknowledgements results from a population-based study in the elderly. Funding: Ravenna Local Health Unit and Merck Am J Hypertens 1997; 10: 697–704. Sharp & Dohme Italia S.p.A. The co-operation in this 16 Christensen DB et al. Assessing compliance to anti- study of Dr Alessandro Capone is gratefully hypertensive using computer-based phar- macy records. Med Care 1997; 35: 1164–1170. acknowledged. 17 Caro JJ et al. Persistence with treatment for hyperten- sion in actual practice. Can Med Assoc J 1999; 160: References 31–37. 18 Joint National Committee on Detection, Evaluation, 1 MacMahon SW et al. The effect of drug treatment for and Treatment of Blood Pressure. The sixth report of hypertension on morbidity and mortality from cardio- the Joint National Committee on Prevention, Detec- vascular disease: a review of randomised clinical tion, and Treatment of High Blood Pressure (JNC VI). trials. Prog Cardiovasc Dis 1986; 29 (Suppl 1): 99–118. Arch Intern Med 1997; 157: 2413–2446. 2 Dahlof B et al. Morbidity and mortality in the Swedish 19 Colhoun HM, Dong W, Poulter NR. Blood pressure Trial in Old Patients with Hypertension (STOP- screening, management and control in England: results Hypertension). Lancet 1991; 338: 1281–1285. from the Health Survey for England 1994. J Hypertens 3 SHEP Cooperative Research Group: Prevention of 1998; 16: 747–752. stroke by antihypertensive drug treatment in older per- 20 Marques-Vidal P, Tuomilehto J. Hypertension aware- sons with isolated systolic hypertension: final results ness, treatment and control in the community: is the of the Systolic Hypertension in the Elderly Program ‘rules of halves’ still valid? J Hum Hypertens 1997; 11: (SHEP). J Am Med Assoc 1991; 265: 3255–3264. 213–223. 4 Miller NH, Hill M, Kottke T, Ockene IS. The multilevel 21 Cocchi R et al. Cardiovascular risks in hypertensive compliance challenge: recommendations for a call to patients: results of the Pandora project. J Nephrol action; a statement for health care professionals. Circu- 2002; 15:29–34. lation 1997; 95: 1085–1090. 22 Fletcher SW, Pappius EM, Harper SJ. Measurement of 5 Eraker FA, Kirscht JP, Becker MH. Understanding and medication compliance in a clinical setting. Arch Inter improving patients compliance. Ann Intern Med 1984; Med 1979; 139: 635–638. 100: 258–268. 23 Stephenson BJ et al. Is this patient taking the treatment 6 Mancia G et al. Blood pressue control in the hyperten- as prescribed? JAMA 1993; 269: 2779–2781. sive population. Lancet 1997; 349: 454–457. 24 O’Brian Smith E et al. Clinic attendance in the hyper- 7 Bittar N. Maintaining long-term control of blood press- tension detection and follow-up program. Hyperten- ure: the role of improved compliance. Clin Cardiol sion 1982; 4: 710–715. 1995; 18 (Suppl III): 12–16. 25 Goldberg AI, Dunlay MC, Sweet CS. Safety and toler- 8 Wayne D. Applied Nonparametric Statistics, Second ability of losartan potassium, an angiotensin II recep- Edition. PWS-KENT Publishing Company: USA, 1990. tor antagonist, compared with , 9 Marubini E, Valsecchi MG. Emmerson M. Analysing , ER, and angiotensin-converting Survival Data from Clinical Trials and Observational enzyme inhibitors for the treatment of systemic hyper- Studies. John Wiley & Sons: June, 1995. tension. Am J Cardiol 1995; 75: 793–795.

Journal of Human Hypertension