Pharmacoeconomics of Antihypertensive Drug Treatment: An Analysis of How Long Patients Remain on Various Antihypertensive Therapies

Abstract

The objective of the research was to perform a clinical practice‐based analysis of how long patients remain on various antihypertensive drugs. An administrative database listing of patient baseline characteristics, drug prescriptions, and hospital admissions was used. All new users of antihypertensive drugs, ≥20 years of age, receiving a first prescription for diuretics, β blockers, calcium channel blockers, angiotensin‐converting enzyme inhibitors, or angiotensin II receptor antagonists between January 1, 2000, and December 31, 2000, were included and observed for 365 days. Persistence was defined as a duration of therapy >273 days. A total of 14,062 patients were included in the study, 39.7% of whom remained on treatment (persistent patients). Persistent patients were more likely to be older, taking other drugs for concurrent disorders, hospitalized for cardiovascular diseases, and initially prescribed angiotensin II receptor antagonists. Persistent patients accounted for 80.6% of the overall cost for antihypertensive drugs. Factors associated with drug cost were age, pattern of persistence, number of prescribed classes, and specific medication at enrollment. Measuring persistence with treatment is needed to evaluate the appropriateness and the cost‐effectiveness of drug use.

Randomized and controlled trials have demonstrated the efficacy of antihypertensive drug therapy in reducing the risk for cardiovascular morbidity and mortality. ¹ , ² , ³ , ⁴ , ⁵ The cost‐effectiveness of antihypertensive treatment has also been calculated as net value between costs allocated to treatment and reduced costs of cardiovascular morbidity. ⁶ , ⁷ , ⁸ , ⁹ Both methodologies balance efficacy and cost‐effectiveness of adequate pharmacologic preventive care. Notwithstanding, in clinical practice one finds a considerable percentage of hypertensive patients whose history of drug administration does not correspond to an effective treatment regimen. ¹⁰ , ¹¹ , ¹² , ¹³ In particular, failure to continue on a particular drug regimen was discovered as the most representative type of drug noncompliance because some studies indicate that about half the subjects exposed to antihypertensive drugs discontinue treatment within a few months. ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ Substandard compliance with treatment can be considered inappropriate drug use that produces therapeutic and economic consequences. The main effect is that blood pressure is not effectively controlled and the risk of further cardiovascular events is not reduced. ¹⁸ , ¹⁹ , ²⁰ In addition, drug costs allocated to noncompliant patients suggests a waste of resources because therapy cannot be correlated with an effective outcome. ²¹ , ²² , ²³

The aim of this study was to perform a persistence‐based pharmacoeconomic evaluation of patients receiving antihypertensive drug treatment. Specific objectives were represented by measuring persistence with antihypertensive drug treatment and profiling who did and did not remain on therapy based on patient‐related factors and the initially prescribed class of drug. In addition, the overall cost of antihypertensive drugs among treated subjects in relation to the pattern of persistence and the annual average cost of drug treatment in relation to individual and therapeutic characteristics were calculated.

Data Source

The study subjects were recruited from individuals registered with the Local Health Unit (LHU) of Ravenna (an area located in northeastern Italy with 359,000 inhabitants) and eligible over the study period. The local ethics committee of the LHU of Ravenna approved this study. The LHU is a body delegated by Italy's National Health System (NHS) to provide health care to a specific geographic area. The LHU, being a point of delivery for the central health system, has an information network that routinely measures expenditures generated by the dispensing of drugs to registered patients. In particular, this administrative/accounting‐type archive is used for recording the funds that pharmacies are entitled to receive from the LHU as refunds for drugs that are reimbursable by the NHS and dispensed free of charge to the public. Since January 1, 1996, this archive has been structured so the prescriptions recorded are attributed in each case to the patient/recipient.

The data available on each prescription include the patient's national health number, the prescribing physician's number, the anatomical‐therapeutic‐chemical code of the drug purchased, the number of packs, the number of units per pack, the dosage, the unit cost per pack, and the prescription date. The identification of the patient given by the personal health number, cross‐checked with the registry office and hospital database, allows all the information to be integrated with the patient's sex, date of birth, and any record of previous hospitalizations for cardiovascular disease (CVD).

Study Design

This was a cohort study that included only the records of patients who received antihypertensive drugs for the first time. Those enrolled for the study were new users, ≥20 years of age, who had received a first prescription for one of the following classes of antihypertensive drugs between January 1 and December 31, 2000 (the enrollment period): diuretics, β blockers, calcium channel blockers (CCBs), angiotensin‐converting enzyme (ACE) inhibitors, or angiotensin II receptor antagonists (AIIAs). New users were defined as patients who had not been prescribed any antihypertensive drugs in the 12 months preceding the enrollment date (date of the first prescription within the enrollment period). Patients who died or moved away were not included in the summary. In addition, individuals were excluded if, on the enrollment date, they had filled prescriptions for two or more classes of antihypertensive agents or for a combination product including two different classes of antihypertensive drug (e.g., nondiuretic/diuretic combination).

A history of three or more prescriptions for drugs acting on the cardiovascular system, for antidiabetes, or antiasthmatic or chronic obstructive pulmonary disease (COPD) during the 12 months preceding the enrollment date was used to define patients as cardiac, diabetic, or having asthma/COPD, respectively.

Analysis of Persistence

The follow‐up period for each patient was 12 months, starting from the enrollment date. The duration of therapy (days elapsing between the first and the last prescription) and the mean daily dose were calculated. ¹⁴ , ²⁴ , ²⁵

Persistence with treatment was arbitrarily defined as a duration of treatment of >273 days from the enrollment date. ¹⁴ , ²⁴ , ²⁵ According to established criteria, patients were divided into three categories: continuers, patients who persisted with drug treatment, using the class prescribed at enrollment even if combined with another class; switchers, patients who persisted with drug treatment but changed to a drug other than that initially prescribed at enrollment; and discontinuers, patients who discontinued drug treatment. Patients who persisted with drug treatment by maintaining, combining, or switching the class of drug initially prescribed at enrollment were considered persistent patients. Discontinuers were considered nonpersistent patients.

Cost Analysis

Direct costs of antihypertensive drugs were taken into account and evaluated at NHS purchase prices. Costs were expressed as overall and average values. The currency reference was the Euro (ɛ) (ɛ1=$1.117). The mean price per pack of the drugs belonging to the five pharmacologic classes, weighted for the number of packs administered for each individual drug, was ɛ3.83 for diuretics, ɛ10.91 for β blockers, ɛ14.64 for CCBs, ɛ11.54 for ACE inhibitors, and ɛ29.50 the AIIAs.

A multiple linear regression analysis was performed on persistent patients taking at least 75% of the recommended daily dose. The following independent variables were considered: age, sex, CVD, diabetes, asthma/COPD, previous hospitalizations for CVD, pattern of persistence, number of different classes of drug prescribed in the study period, and class of drug prescribed at enrollment. ²⁶

Statistical Analysis

The effect of continuous variables is expressed in mean values±standard deviation. Statistical significance of proportions was calculated using Pearson's χ ² test; statistical significance of averages was calculated using analysis of variance. All p values quoted are two‐sided; p<0.05 was considered significant. ²⁷

A Cox regression model was used to estimate the impact of several risk factors and to test for confounders.

RESULTS

A total of 19,995 subjects were enrolled in the study. Of these, 5933 subjects (29.7% of enrollees) were excluded: 3673 (18.4% of enrollees) because they were taking a class of drug not included in the study at enrollment, 1443 (7.2% of enrollees) because they were combining several classes at enrollment, 480 (2.4% of enrollees) because they died, and 337 (1.7% of enrollees) because they moved away during the follow‐up period. A total of 14,062 patients were thus included in the study, 6098 men (43.4%) and 7964 women (56.6%) with an average age of 56.9±17.7 years (range 20–105 years). ACE inhibitors were the class of drug most commonly prescribed as first‐time therapy (28.0%), followed by CCBs (23.8%), diuretics (23.8%), β blockers (17.6%), and AIIAs (6.9%) (Table I). Patient demographics and characteristics by class of drug prescribed at enrollment are presented in Table I.

Table I.

Demographics and Characteristics of Patients by Class of Drug Prescribed at Enrollment

				ACE
	Diuretics	β Blockers	CCBs	Inhibitors	AIIas
Patients (n [%])	3344 (23.8)	2471 (17.6)	3341 (23.8)	3938 (28.0)	968 (6.9)
Mean age* (yrs [SD])	61.00 (19.05)	52.12 (15.20)	56.08 (18.21)	56.78 (17.17)	57.70 (15.87)
Gender* (% male)	34.7	40.4	47.4	48.4	46.9
Patients with heart disease† (%)	1.5	0.8	0.9	1.3	0.8
Diabetics* (%)	1.6	1.0	1.3	3.8	2.5
Asthmatics* (%)	3.6	1.4	1.6	1.9	2.5
Previous CV hospital admissions* (%)	8.3	6.8	10.6	8.4	8.6
Patients with two or more comorbidities* (%)	3.3	1.0	2.3	2.3	2.4
CCBs=calcium channel blockers; ACE=angiotensin‐converting enzyme; AIIAs=angiotensin II receptor antagonists; CV=cardiovascular; *p<0.01; † p<0.05

Analysis of Persistence

Within the population observed, 60.3% of patients discontinued treatment, 30.9% continued treatment, and 8.8% switched treatment (Table II). Among the discontinuers, 83.3% of patients interrupted the treatment after a single prescription. Among the continuers, 81.1% of patients maintained the enrollment therapy throughout the follow‐up period (no combination). Patient demographics and characteristics by pattern of persistence are presented in Table II. The percentage of continuers, switchers, and discontinuers varied significantly among the five study classes (p<0.001) (Table III).

Table II.

Patient Demographics and Persistence Patterns

	Continuers	Switchers	Discontinuers
Patients (n [%])	4340 (30.9)	1235 (8.8)	8487 (60.3)
Mean age* (yrs [SD])	63.80 (14.53)	63.24 (14.41)	52.39 (18.19)
Gender† (% male)	43.2	42.3	43.6
Patients with heart	1.9	1.7	0.6
disease* (%)
Diabetics* (%)	3.1	2.3	1.6
Asthmatics† (%)	2.3	2.3	2.1
Previous CV hospital	13.2	12.0	5.7
admissions* (%)
Patients with two or more	4.2	2.7	1.3
comorbidities* (%)
CV=cardiovascular; *p=0.001; † p=nonsignificant

Table III.

Persistence Patterns for Class of Drug Prescribed at Enrollment

	Continuers (n [%])	Switchers (n [%])	Discontinuers (n [%])
Diuretics	866 (25.9)	243 (7.3)	2235 (66.8)
β Blockers	913 (36.9)	160 (6.5)	1398 (56.6)
CCBs	891 (26.7)	286 (8.6)	2164 (64.7)
ACE inhibitors	1267 (32.2)	418 (10.6)	2253 (57.2)
AIIAs	403 (41.7)	128 (13.2)	437 (45.1)
Total	4340 (30.9)	1235 (8.8)	8487 (60.3)
CCBs=calcium channel blockers; ACE=angiotensin‐converting enzyme; AIIAs=angiotensin II receptor antagonists

Persistence with treatment was related to age (for each year that age increased the risk of discontinuing the treatment decreased by an average of 2.2%), by whether or not patients were on chronic drug therapy for heart disease or diabetes (subjects not under treatment for heart disease and diabetes showed, respectively, a 66.6% and 39.4% higher risk of discontinuing the treatment than subjects on such treatment), by a history of previous hospitalizations for CVD, by the presence of two or more comorbidities, and by the class of antihypertensive drug prescribed at enrollment (Table IV). Patients who were given an AIIA as the first‐line class showed a greater tendency to stay on antihypertensive therapy compared with those who were enrolled on ACE inhibitors (in whom the risk of discontinuing treatment was 38.6% greater), on CCBs (a 66.3% greater risk), and on diuretics (an 85.3% greater risk), respectively. Gender and the concurrent use of antiasthmatic drugs, on the other hand, were not significantly related to persistence with treatment.

Table IV.

Cox Regression Model Applied to Persistence With Treatment

Factor	p Value	Hazard Ratio (95% CI)*
Age	<0.0001	0.978† (0.976–0.979)
Cardiac disease (present)	0.0002	1.666 (1.270–2.185)
Diabetes (present)	0.0002	1.394 (1.171–1.658)
Previous CV hospital admissions  (present)	<0.0001	1.507 (1.369–1.658)
Patients with two or more	<0.0001	1.630 (1.341–1.981)
comorbidities (present)
Drug class at enrollment
AIIAs	<0.0001	——
ACE inhibitors	<0.0001	1.386 (1.250–1.538)
CCBs	<0.0001	1.663 (1.499–1.846)
Diuretics	<0.0001	1.853 (1.670–2.056)
CI=confidence interval; CV=cardiovascular; AIIAs=angiotensin II receptor antagonists; ACE=angiotensin‐converting enzyme; CCBs=calcium channel blockers *figures in parentheses indicate the reference category of categorical variables (hazard ratio=1), hazard ratio is adjusted for other variables in the table, global χ 2 =1841.451, p<0.0001; †related to 1‐year age increment

Analysis of Costs

The total cost of the study cohort was ɛ1,238,752.37, of which ɛ745,328.31 was for continuers (60.2%), ɛ253,293.08 was for switchers (20.4%), and ɛ240,130.98 was for discontinuers (19.4%).

The average annual cost of a patient exposed to treatment was ɛ88.09 (95% confidence interval [CI], 86.10–90.08) (Table V). In regard to the class initially prescribed, the average annual cost varied from ɛ33.45 (95% CI, 30.97–35.93) for diuretics, to ɛ63.40 (95% CI, 59.94–66.86) for β blockers, to ɛ104.43 (95% CI, 100.07–108.79) for the CCBs, to ɛ108.25 (95% CI, 104.43–112.09) for ACE inhibitors, and ɛ201.53 (95% CI, 191.24–211.81) for AIIAs (p<0.001) (Table V). On the basis of the persistence patterns, the average annual cost was ɛ171.73 (95% CI, 167.43–176.04) for continuers, ɛ205.10 (95% CI, 196.85–213.34) for switchers, and ɛ28.29 (95% CI, 27.62–28.97) for discontinuers (p<0.001) (Table V).

Table V.

Annual Average Cost by Class of Drug Prescribed at Enrollment and Persistence Pattern

	Average Cost for Continuers (ɛ [95% CI])	Average Cost for Switchers (ɛ [95% CI])	Average Cost for Discontinuers (ɛ [95% CI])	Average Cost for Study Cohort (ɛ [95% CI])
Diuretics	65.09	153.10	8.17	33.45
	(58.67–71.52)	(137.59–168.62)	(7.60–8.75)	(30.97–35.93)
β Blockers	109.29	158.73	22.52	63.40
	(102.46–116.12)	(139.61–177.84)	(21.24–23.79)	(59.94–66.86)
CCBs	234.63	199.62	38.24	104.43
	(224.78–244.47)	(183.45–215.78)	(36.78–39.70)	(100.07–108.79)
ACE inhibitors	196.28	237.53	34.76	108.25
	(189.69–202.86)	(222.28–252.79)	(33.53–35.99)	(104.43–112.09)
AIIAs	326.16	268.07	67.10	201.53
	(313.05–339.27)	(241.55–294.59)	(62.89–71.31)	(191.24–211.81)
Total	171.73	205.10	28.29	88.09
	(167.43–176.04)	(196.85–213.34)	(27.62–28.97)	(86.10–90.08)
CI=confidence interval; CCBs=calcium channel blockers; ACE=angiotensin‐converting enzyme; AIIAs=angiotensin II receptor antagonists

The annual cost for antihypertensive drug treatment was related to age (for each year that age increased the cost for antihypertensive drug treatment decreased on average by ɛ0.43), to the pattern of persistence (the costs of the patients combining and of the patients switching their enrollment class were, respectively, ɛ26.35 and ɛ33.17 higher than the cost of the subjects who maintained their enrollment class), to the number of different drug classes prescribed during the follow‐up period (each class increase corresponded to an average increase of ɛ10.94), and to the class of antihypertensive drug prescribed at enrollment (Table VI). Patients who were given a diuretic as first‐line therapy showed a lower annual cost for antihypertensive drug treatment compared with those who were enrolled on β blockers (ɛ67.45 more), ACE inhibitors (ɛ168.20 more), CCBs (ɛ188.84 more), and AIIAs (ɛ278.19 more), respectively. By observing the standardized regression coefficients, it can be seen that the variables that best correlated with the cost of drug treatment were the specific medication used at enrollment and the pattern of persistence. Gender; concurrent use of cardiac, antidiabetic, or antiasthmatic drugs; and previous hospital admission for CVD were not significantly related to cost for antihypertensive drug treatment.

Table VI.

Multiple Linear Regression Analysis of Cost of Drug Treatment

Factor	Reference Category	Regression Coefficient (b)	95% CI	p Value
Age		−0.43*	(−0.81 to −0.05)	0.027
Compliance pattern	Increased cost compared
	with patients who stayed
	on initial therapy
Combining the class at		26.35	(11.37–41.34)	0.001
enrollment
Switching the class at	33.17	(18.65–47.70)	<0.001
enrollment
No. of different classes		10.94†	(2.62–19.27)	0.010
Class at enrollment	Increased cost compared
	with diuretics
β Blockers		67.45	(49.71–85.20)	<0.001
CCBs		188.84	(172.81–204.87)	<0.001
ACE inhibitors		168.20	(153.05–183.35)	<0.001
AIIAs		278.19	(258.84–297.54)	<0.001
CI=confidence interval; CCBs=calcium channel blockers; ACE=angiotensin‐converting enzyme; AIIAs=angiotensin II receptor antagonist; *related to 1‐year age increment; †related to one unit increment

DISCUSSION

An analysis of pharmaceutical use at the Ravenna LHU confirms previous findings that indicate that persistence with antihypertensive drug therapy was not adequately achieved because a high percentage of subjects discontinued treatment within 12 months of starting it. ¹⁴ , ¹⁶ , ¹⁷ , ²⁸ Nonpersistence with treatment was associated with patient‐related factors—age, concurrent chronic pharmacologic treatments, previous hospital admission for CVD—and with the initially prescribed class of antihypertensive drugs.

Analysis of the patients' demographic data and case histories showed that younger subjects with no concurrent chronic pharmacologic regimen and no previous hospital admissions for CVD were more likely to interrupt antihypertensive treatment. This may be due to the fact that subjects with concurrent chronic pharmacologic regimen and previous hospital admission for CVD are more prone to tolerate undesirable side effects of drug treatment and are more aware of the danger of inadequate treatment.

The initially prescribed class of antihypertensive drugs (first‐line therapy) was another important factor defining persistence with treatment. The reasons may derive from various factors: different tolerability profiles for the classes studied, financial incentives for prescribing specific medications, selection bias when enrolling study patients, health care provider bias in selecting medications, and other factors. ²⁹ , ³¹ However, as suggested by previous studies, ¹⁶ , ¹⁷ the most plausible is the different tolerability profile of the drugs studied. This is confirmed by the fact that the greatest persistence was obtained for the AIIAs that are known to have an as good as or better tolerability profile when compared with other drug classes. ³² , ³³ , ³⁴ As for other possible factors, in the Italian NHS no antihypertensive drug is paid for by the patient (they are all refunded by the NHS) and health care providers are not encouraged through financial incentives to prescribe a particular drug. Under these circumstances, any bias effects on the selection of medications can reasonably be considered negligible. Finally, statistical analyses were performed to assess the effect of any selection bias. Although patient‐related factors differed among the five study classes, the relationship between persistence with treatment and initial class of drug remained statistically significant when these variables were assessed.

The determination of the level of nonpersistence with antihypertensive treatment and, above all, definition of the demographic, clinical, and therapeutic characteristics of patients who discontinue therapy (risk profiling, where the risk is represented by interruption of treatment) may provide useful information for health service administrators and general practitioners. Indeed, these results should create a greater awareness of patients who are "carriers" of the characteristics most frequently associated with nonpersistence and help in the selection of first‐line drugs based on the need to maintain a patient on treatment as well as on therapeutic efficacy.

The interruption of antihypertensive drug treatment also has some important economic implications. Nonpersistent patients and the high cost for antihypertensive drugs that may be inappropriate put a heavy drain on resources. From a pharmacoeconomic point of view, the use of antihypertensive drugs that are discontinued fairly early in treatment can be interpreted as an economic indicator of inappropriateness. This can be improved by shifting patients from the nonpersistent to the persistent state. Improving persistence with treatment should be pursued in the short term for a more appropriate use of pharmacologic resources, and in the long term to reduce cardiovascular risk and the high costs associated with specialist hospital treatment.

The annual average cost for antihypertensive drug treatment was significantly affected by age, persistence pattern, and therapy at enrollment. The variations in the costs of treatment can be attributed to the purchase price of antihypertensive drugs, the prescription frequency during the observation period, and the number of other drugs prescribed in addition to the original medication. A detailed assessment of the cost of antihypertensive treatment, taking into account the factors that cause variations, is an important element for the accurate calculation of a budget for a population with significantly heterogeneous demographic and pharmacoepidemiologic characteristics that alter over time. Moreover, calculating the average cost for drug therapies is a key element for assessing the cost‐effectiveness of alternative pharmacologic agents in practice.

Some final considerations are necessary concerning the limits of this study. First, health care claim data‐bases lack data on diagnoses, so there may have been a misclassification of subjects in regard to high blood pressure. However, this limitation should only apply to patients who were prescribed these drugs just once, as it is reasonable to assume that physicians would not continue to prescribe therapy unless the patients were diagnosed as hypertensive. Second, health care claim databases lack blood pressure data. Blood pressure levels at enrollment, if available, might shed further light on the different persistence found for the drug classes studied (selection bias). It is difficult to attribute to this single factor the great variations in persistence observed among the classes studied. Finally, another factor that may have influenced the greater persistence with the newest medications (i.e., AIIAs) is the relative emphasis placed on these agents at medical meetings and in the literature. These may have influenced enthusiasm for those agents.

As opposed to previous similar real‐practice analyses, ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ , ²⁸ this study attempted to define in greater detail the patient characteristics associated with treatment persistence (risk profiling) and to assess the total and individual cost for a cohort of new users of antihypertensive drugs. Further studies are needed to evaluate other patient characteristics like blood pressure levels—useful for more accurate patient profiles—and to measure outcomes such as cardiovascular morbidity and mortality—useful for comprehensive‐effectiveness and cost‐effectiveness analyses.

Acknowledgment and disclosure:

The authors thank Dr. Vincenzo Porzio for his review of this article. This study was funded by Ravenna Local Health Unit, Ravenna, Italy, and Merck Sharp & Dohme SpA., Rome, Italy.