Comparison of real-world adherence, healthcare resource utilization and costs for newly initiated valsartan/amlodipine single-pill combination versus angiotensin receptor blocker/calcium channel blocker free-combination therapy

Abstract

Objective:

To compare adherence, healthcare costs and utilization of valsartan/amlodipine single-pill combination (SPC) and angiotensin-receptor blocker/calcium-channel blocker multiple-pill free-combination (ARB + CCB FC) therapy using real-world data.

Methods:

A retrospective study (January 1, 2007 to April 30, 2009) was conducted using US commercial healthcare insurance claims. Patients were assigned to two cohorts: ‘valsartan/amlodipine SPC cohort’ and ‘ARB + CCB FC therapy cohort’. The primary endpoints were adherence and persistence. The secondary endpoints were 1-year healthcare costs and utilization.

Results:

Out of 12,628 eligible patients 3259 (26%) were included in the valsartan/amlodipine SPC cohort and 9369 (%74) in the ARB + CCB FC cohort. Risk-adjusted adherence rates were higher for valsartan/amlodipine SPC patients [OR: 1.38, 95% CI: (1.24, 1.53)]. The Cox proportional hazard model showed that valsartan/amlodipine SPC cohort patients were less likely to discontinue medication (HR: 0.87, p < 0.001). Comparison between the groups also yielded that total healthcare costs of valsartan/amlodipine SPC patients were 16–20% lower than ARB + CCB FC therapy patients (p < 0.0001).

Limitations:

Since claims data are collected for payment purposes rather than research purposes, the study is bound by limitations for the retrospective analysis. For example, the presence of a claim for a filled prescription does not indicate that the medication was consumed or taken as prescribed. Data on health behaviors and patient lifestyle were not available. Over-the-counter medications and clinical disease severity were not available in the dataset. Incorrect coding is also a possibility. However, we have used previously validated datasets where these effects are minimal. Heterogeneity of the sample may create bias in our estimates, however, we have used advanced statistical methods to control for observed and unobserved bias.

Conclusion:

The real-world use of valsartan/amlodipine SPC was associated with better adherence and persistence relative to ARB + CCB FC therapy among patients with hypertension. Moreover, patients taking single-pill combination therapy had lower healthcare costs and utilization.

Key words::

• Adherence
• Amlodipine
• Angiotensin receptor blocker
• Calcium channel blocker
• Hypertension
• Real-world
• Utilization
• Valsartan

Introduction

An estimated one-third of adults in the United States have hypertension (HTN)1. High blood pressure has significant consequences as HTN increases the risk of stroke, cerebrovascular and cardiovascular disease, end-stage renal disease, and death.

Monotherapy using only one class of these medications is not always adequate to reach target blood pressure and approximately 60% of treated individuals remain hypertensive. Combination therapy can facilitate a more rapid lowering of blood pressure to attain target blood pressure as well as minimize potential adverse side-effects2. The Seventh Report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure (JNC 7) guidelines recommend that a patient with systolic blood pressure ≥20 mmHg or diastolic blood pressure 10 mmHg above goal begin treatment with two or more antihypertensive agents, either as separate drugs or in fixed-dose combinations.

However, studies have demonstrated that patient adherence decreases with regimen complexity3. JNC 7 recommends that the use of fixed-dose combinations may be more convenient and simplify the treatment regimen, and may cost less than the individual components prescribed separately4. In contrast, compared to angiotensin receptor blocker/calcium channel blocker free-combination (ARB + CCB FC) therapy, compliance is improved in patients using fixed-dose or single-pill combinations potentially translating into better clinical outcomes5.

Relative adherence to antihypertensive therapy and its effect on healthcare costs and utilization for patients who use valsartan/amlodipine single-pill combination (SPC) therapy is not well-understood from a population perspective. In this study, we compared the adherence rates, healthcare utilization and cost associated with valsartan/amlodipine SPC and ARB + CCB FC therapy in a managed-care population. Propensity score matching (PSM) and standard multivariate regression analysis were used to control for heterogeneity of these two populations. We also employed more advanced statistical procedures, such as instrumental variable (IV) analysis, to determine the sensitivity of our estimates.

Methods

Data sources and study population

HIPAA-compliant pharmacy and medical administrative claims data from a proprietary US health plan database, associated with i3 Global, were used for this study. i3 Global has access to a proprietary research database containing claims and enrollment data dating back to 1993, with the opportunity to link patient and physician survey data to pharmacy and medical claims, medical record data, socioeconomic measures, and clinical laboratory results. For 2005, data relating to approximately 14 million individuals with both medical and pharmacy benefit coverage are available. An additional 8 million enrollees with medical benefits only are also available. Underlying information is geographically diverse across the United States, and is updated frequently. Ingenix research activities utilize de-identified data from the research database except in limited instances where applicable law allows the use of patient identifiable data.

Patients were selected from the database for the index period beginning July 2006 through April 2009. The patient identification period was between January 1, 2007 and April 30, 2008, allowing 6 months pre-period and 12 months follow-up time. To be eligible for the sample, patients were required to have evidence of hypertension- a claim with HTN diagnosis (International Classification of Diseases, 9th Revision, Clinical Modification [ICD-9-CM]: 401.xx–404.xx), have continuous enrollment in a health plan for at least 6 months prior to the index admission date and 12 months following the index date or until death (whichever came first), be age 18 years or older as of the year of the index date, and not have a diagnosis code of pregnancy at any time during the index period. The index date was defined as the date of the first fill for valsartan/amlodipine SPC or ARB + CCB FC therapy during the patient identification period.

Two mutually exclusive cohorts were defined. Patients in the valsartan/amlodipine SPC cohort had at least two filled prescriptions for valsartan/amlodipine SPC therapy with total days' supply of at least 90 days. Patients who had evidence of valsartan/amlodipine or ARB + CCB use in the pre-index period and evidence of ARB + CCB FC therapy use in the follow-up period were excluded from the valsartan/amlodipine SPC cohort. Similarly, the ARB + CCB FC cohort included patients with at least two filled prescriptions for ARB and CCB (including the index prescription) during the follow-up period with days of supply at least 90 days. Note that only the days the patient was on both ARB and CCB were counted. Therefore the patient will be on these medications at the same time. Patients with evidence of valsartan/amlodipine or ARB + CCB use in the pre-index period and evidence of valsartan/amlodipine use in the follow-up period were excluded from the ARB + CCB FC therapy cohort.

Baseline variables

Patient demographics (including age, gender, geographic location, and plan type) were captured from enrollment data in the claims database. Several variables were created to determine baseline clinical characteristics. Using the Deyo6 adaptation of the Charlson Comorbidity Index7, baseline comorbidity was measured for each patient. This adaptation uses a scale of 0–28, representing the sum of individual weighted scores for each comorbid condition. Moreover, a count of total unique medications was calculated to assess the overall pill burden.

Individual comorbidities at the baseline were identified using ICD-9-CM codes: coronary heart disease (CHD) (ICD-9-CM: 410–414), myocardial infarction (MI) (ICD-9-CM: 410), peripheral vascular disease (PVD) (ICD-9-CM: 443), stroke (ICD-9-CM: 434.91), ischemic heart disease (IHD)(ICD-9-CM: 410-414), congestive heart failure (CHF) (ICD-9-CM: 428), dyslipidemia (ICD-9-CM: 272), diabetes (ICD-9-CM: 250), hypertension (ICD-9-CM: 401.x), obesity (ICD-9-CM: 278), chronic kidney disease (ICD-9-CM: 585), and smoking (ICD-9-CM: 305.1).

Use of concomitant medication was determined from pharmacy claims files. In particular, we identified ACE-inhibitors, beta-blockers, anti-diabetics, diuretics (excluding HCTZ) and other antihypertensives.

Healthcare cost and utilization during the pre-index period were computed as the combined health plan and patient-paid amounts. Costs were calculated as total costs (medical + pharmacy), medical costs, and pharmacy costs. Costs were adjusted using the annual medical care component of the Consumer Price Index to reflect inflation between 2007, 2008 and 2009. Costs were separated between plan-paid and patient-paid. In terms of healthcare utilization, we calculated the percentage of patients who had inpatient, ER, physician office, and other outpatient visits, and use of at least one prescription.

Outcomes variables

Medication use patterns including proportion of days covered (adherence to the prescribed regimen) and discontinuation (persistency) of therapy were measured using claims for the index medication inclusive of fills on the index date8.

Adherence with index therapy was measured using proportion of days covered (PDC) and calculated as the proportion of days a patient had the index drug within the 12-month follow-up period. Persistence with therapy was measured as days from the index date to therapy discontinuation date, which was defined as a gap in therapy of 30 or more days.

All-cause healthcare cost and resource utilization were calculated using methods similar to those used in the pre-index period.

Statistical analysis

Baseline variables were compared between the two groups using chi-squared testing for dichotomous and polychotomous variables. Risk adjustment was done using multivariate analysis and propensity score matching (PSM). Adherence to medication using proportion of days covered was analyzed using logistic regression. Cox proportional hazards regression was used for persistence of therapy to calculate the risk of discontinuation.

For the propensity score match, we applied a number of possible matching techniques. Using guidelines provided by Baser9, one-to-one matching created the best comparable sample. Therefore, risk-adjusted healthcare resource utilization and cost were compared using one-to-one PSM.

Sensitivity analysis

Although effectively controlling for observable bias, neither PSM nor regression adjustment addresses bias caused by imbalances in unmeasured factors. Since retrospective observational studies do not contain a clinical severity measure, choosing between two treatments may be influenced by disease severity, which is not controlled in the model. Although we controlled for baseline comorbidities and baseline healthcare cost and utilization as a proxy for severity, none of these measures can replace the clinical severity measure. Therefore, by using instrumental variable (IV) regressions as a sensitivity analysis, we showed how sensitive our results were for unobservable measures. Previously validated, we have used doctors’ prescribing pattern as an instrument10,11.

Results

After applying the inclusion and exclusion criteria, 12,628 patients were eligible for the analysis (Table 1). 26% of this group was in the valsartan/amlodipine SPC cohort (n = 3259), and 74% was in the ARB + CCB FC therapy cohort (n = 9369).

Table 1.  Patient selection.

Criteria for patient selection	Patient count	Patient count
	V/A SPC	ARB + CCB MPC
# of patients eligible in the health plan for at least 1 day during the intake period (1/1/2007 – 4/30/2008). 6-month pre-period (7/1/2006 – 12/31/2006) and 12-month follow-up (4/30/2008 – 4/30/2009)	8274	57450
# of patients at least 18 years or older as of index date	8274	57406
# of patients with diagnosis of HTN during baseline period (401.xx–404.xx)	7423	43851
# of patients with ≥1 or more fills for drug(s) of interest – Exforge (valsartan/amlodipine) SPC or ARB + CCB FC therapy, during follow-up period with total days supply ≥ 90 days	5513	31331
Patients with continuous enrollment with pharmacy and medical benefits in baseline and follow-up periods (at least 6 months pre-index and 12 months post-index)	3844	22097
Patients without diagnosis of pregnancy from pre-period through follow-up period	3837	22076
Patients without valsartan/amlodipine use from pre-period through follow-up period	N/A	21716
Patients without ARB + CCB use from the pre-period through the follow-up period	3308	N/A
Patients with at least 2 fills of drugs	3259	21597
Newly initiated patients	3259	9369
Propensity score-matched group	3071	3071

There were significant differences in the observed baseline characteristics between the two groups. Patients in the valsartan/amlodipine SPC cohort were younger (mean age of 54 vs. 61 years), mostly male (59.5 vs. 52.3%), and living in the southern United States (68.82 vs. 57.60%; p < 0.001, all comparisons) (Table 2).

Table 2.  Demographic characteristics.

	V/A SPC		ARB + CCB MPC		p-value
	n/mean	%/SD	n/mean	%/SD
# of patients, n (%)	3259 (26%)		9369 (74%)
Age, years
Mean	54.37	10.72	60.77	11.87	0.0000
Median	54		60
Gender, n (%)
Male	1939	59.50%	4895	52.25%	0.0000
Female	1320	40.50%	4474	47.75%	0.0000
Geographic region, n (%)
Northeast	293	8.99%	879	9.38%	0.5070
Midwest	431	13.22%	1839	19.63%	0.0000
South	2243	68.82%	5397	57.60%	0.0000
West	292	8.96%	1254	13.38%	0.0000
Plan type, n (%)
PPO	439	13.47%	2393	25.54%	0.0000
HMO/POS	2192	67.26%	5014	53.52%	0.0000
EPO	492	15.10%	982	10.48%	0.0000
IND	136	4.17%	972	10.37%	0.0000
Other	0	0.00%	8	0.09%	0.0952
Duration of follow-up, months
Mean	507.82	78.48	594.07	143.78	0.0000
Median	512		581

As reported in Table 3, in terms of clinical characteristics, at the baseline, the valsartan/amlodipine SPC cohort had a lower mean Charlson Comorbidity Score (0.60 vs. 1.37) and a lower overall pill burden (13.43 vs. 19.45). They had lower rates of congestive heart failure (1.78 vs. 7.06%), diagnosed diabetes (21.54 vs. 36.11%), dyslipidemia (55.97 vs. 60.23%), and chronic kidney disease (2.30 vs. 11.24%). Fewer patients in the valsartan/amlodipine SPC cohort used beta-blockers (26.6 vs. 37.35%), anti-diabetics (15.99 vs. 29.12%) and diuretics (6.51 vs. 20.08%, p < 0.01, all comparisons) but had higher rates of other anti-hypertensives (45.97 vs. 29.66%, p < 0.01, all comparisons).

Table 3.  Baseline clinical characteristics.

	V/A SPC		ARB + CCB MPC		p-value
	n/mean	%/SD	n/mean	%/SD
Number of patients, n (%)	3259		9369
Baseline comorbid conditions, n (%)
CHD	363	11.14%	1912	20.41%	0.0000
MI	19	0.58%	155	1.65%	0.0000
PVD	56	1.72%	374	3.99%	0.0000
Stroke	27	0.83%	192	2.05%	0.0000
CHF	58	1.78%	661	7.06%	0.0000
Dyslipidemia	1824	55.97%	5643	60.23%	0.0000
Diabetes	702	21.54%	3383	36.11%	0.0000
Hypertension	3069	94.17%	8816	94.10%	0.8797
Obesity	229	7.03%	721	7.70%	0.2124
Chronic kidney disease	75	2.30%	1053	11.24%	0.0000
Smoke	107	3.28%	343	3.66%	0.3163
Baseline concomitant medications, n (%)
ACE-Inhibitors	459	14.08%	1917	20.46%	0.0000
Beta-blockers	867	26.60%	3487	37.22%	0.0000
Anti-diabetics	521	15.99%	2489	26.57%	0.0000
Diuretics (excluding HCTZ)	212	6.51%	1585	16.92%	0.0000
Other anti-hypertensives	1498	45.97%	3168	33.81%	0.0000
overall pill burden, mean (SD)	13.43	12.13	19.45	16.46	0.0000
Deyo–Charlson Score, mean (SD)	0.60	1.07	1.37	1.75	0.0000
Baseline costs and resource use
Plan paid medical + pharmacy plan costs, mean (SD)	$3330	$8981	$6764	$20536	0.0000
Plan paid pharmacy costs mean (SD)	$765	$1322	$1232	$2121	0.0000
Patient paid medical + pharmacy costs, mean (SD)	$681	$887	$990	$1453	0.0000
Patient paid pharmacy costs mean (SD)	$306	$290	$412	$376	0.0000
Total medical + pharmacy costs  (plan paid + patient paid), mean (SD)	$3969	$9396	$7724	$21092	0.0000
Total pharmacy costs mean (SD)	$1032	$1463	$1604	$2280	0.0000
Patient paid medical + pharmacy co-pay, mean (SD)	$391	$337	$517	$446	0.0000
Patient paid pharmacy co-pay mean (SD)	$297	$281	$399	$366	0.0000
Patients w/ ≥1 ER visit, n (%)	169	5.19%	614	6.55%	0.0053
Patients w/ ≥1 inpatient stay, n (%)	174	5.34%	1499	16.00%	0.0000
Patients w/ ≥1 physician office visit, n (%)	3192	97.94%	9132	97.47%	0.1285
Patients w/ ≥1 other outpatient visit, n (%)	1281	39.31%	5200	55.50%	0.0000
Patients w/ ≥1 prescription fill, n (%)	3085	94.66%	8978	95.83%	0.0056

Baseline healthcare utilization and costs for patients in the valsartan/amlodipine SPC cohort were significantly lower. In particular, they had fewer hospitalizations, outpatient visits, and ER visits. Therefore, baseline healthcare costs for patients in the SPC group were significantly lower (p < 0.01, all comparisons).

Table 4 presents summary values of unadjusted outcome values overall, for patient groups who have higher and lower adherence. Without adjustment, regardless of what adherence category they belong to, healthcare costs and utilizations were significantly lower for valsartan/amlodipine SPC users relative to ARB + CCB FC users.

Table 4.  Patient post-index cost and utilization (unadjusted).

	V/A SPC		ARB + CCB MPC		p-value
	n/mean	%/SD	n/mean	%/SD
All patients n	3259		9369
Post-index costs and resource use
Plan-paid medical + pharmacy plan costs, mean (SD)	7452.48	18890.05	12712.83	33065.89	0.0000
Plan-paid pharmacy costs, mean (SD)	1918.15	2648.82	3349.52	4677.43	0.0000
Patient-paid medical + pharmacy costs, mean (SD)	1668.39	1568.97	2170.95	1903.81	0.0000
Patient-paid pharmacy costs, mean (SD)	924.51	647.94	1181.32	795.32	0.0000
Total medical + pharmacy costs (plan-paid + patient-paid), mean (SD)	9088.34	19616.54	14866.34	33892.40	0.0000
Total pharmacy costs (plan-paid + patient-paid), mean (SD)	2820.95	3007.72	4491.18	4992.10	0.0000
Patient-paid medical + pharmacy copay, mean (SD)	1078.82	753.36	1391.83	955.07	0.0000
Patient-paid pharmacy copay, mean (SD)	885.95	628.06	1142.33	780.56	0.0000
Patients w/ ≥1ER visit, n (%)	243	7.46%	909	9.70%	0.0001
Patients w/ ≥1 inpatient stay, n (%)	276	8.47%	1578	16.84%	0.0000
Patients w/ ≥1 physician office visit, n (%)	3189	97.85%	9285	99.10%	0.0000
Patients w/ ≥1 other outpatient visit, n (%)	1736	53.27%	6504	69.42%	0.0000
Patients w/ ≥1 prescription fill, n (%)	3259	100.00%	9369	100.00%
Patients with PDC ≥80% n(%)	1526		3831
Post-Index costs and resource use
Plan-paid medical + pharmacy plan costs, mean (SD)	7664.12	20543.81	10221.40	21747.17	0.0001
Plan-paid pharmacy costs, mean (SD)	2361.18	2965.49	3554.78	4738.52	0.0000
Patient-paid medical + pharmacy costs, mean (SD)	1821.61	1663.25	2118.18	1598.97	0.0000
Patient-paid pharmacy costs, mean (SD)	1088.56	695.75	1276.97	828.84	0.0000
Total medical + pharmacy costs (plan-paid + patient-paid), mean (SD)	9445.56	21146.99	12331.02	22495.78	0.0000
Total pharmacy costs (plan-paid + patient-paid), mean (SD)	3421.17	3321.60	4790.96	5028.68	0.0000
Patient-paid medical + pharmacy copay, mean (SD)	1226.64	781.02	1451.47	940.72	0.0000
Patient-paid pharmacy copay, mean (SD)	1043.03	671.25	1233.63	814.22	0.0000
Patients w/ ≥1ER visit, n (%)	120	7.86%	327	8.54%	0.4222
Patients w/ ≥1 inpatient stay, n (%)	113	7.40%	487	12.71%	0.0000
Patients w/ ≥1 physician office visit, n (%)	1495	97.97%	3798	99.14%	0.0004
Patients w/ ≥1 other outpatient visit, n (%)	821	53.80%	2565	66.95%	0.0000
Patients w/ ≥1 prescription fill, n (%)	1526	100.00%	3831	100.00%
Patients with PDC<80% n (%)	1733		5538
Post-Index costs and resource use
Plan-paid medical + pharmacy plan costs, mean (SD)	7266.45	17310.38	14437.78	38933.09	0.0000
Plan-paid pharmacy costs, mean (SD)	1529.06	2266.06	3207.48	4629.79	0.0000
Patient-paid medical + pharmacy costs, mean (SD)	1533.38	1468.22	2207.46	2088.25	0.0000
Patient-paid pharmacy costs, mean (SD)	779.73	564.48	1115.11	764.35	0.0000
Total medical + pharmacy costs (plan-paid + patient-paid), mean (SD)	8773.79	18162.85	16620.18	39823.31	0.0000
Total pharmacy costs (plan-paid + patient-paid), mean (SD)	2292.42	2589.50	4283.81	4956.49	0.0000
Patient-paid medical + pharmacy copay, mean (SD)	949.51	703.49	1350.65	962.78	0.0000
Patient-paid pharmacy copay mean (SD)	748.66	552.46	1079.22	750.02	0.0000
Patients w/ ≥1ER visit, n (%)	123	7.10%	582	10.51%	0.0000
Patients w/ ≥1 inpatient stay, n (%)	163	9.41%	1091	19.70%	0.0000
Patients w/ ≥1 physician office visit, n (%)	1694	97.75%	5487	99.08%	0.0000
Patients w/ ≥1 other outpatient visit, n (%)	915	52.80%	3939	71.13%	0.0000
Patients w/ ≥1 prescription fill, n (%)	1733	100.00%	5538	100.00%

Both risk adjustments for propensity score matched regression and multivariate regression were done to control for factors such as age, gender, geographic region, baseline pill burden, baseline comorbidities, Deyo–Charlson comorbidity index and co-payments.

A total of 3071 patients in each treatment group were matched. The area under receiver-operating characteristic curve was 0.82, suggesting that the logistic model had adequate power to discriminate. The Hosmer–Lemeshow test statistic was insignificant (p = 0.358), confirming the overall suitability of the model. For the IV approach, controlling for other covariates, correlation between physician/prescribing patterns and the choice between single pill versus multiple pill was significant, confirming the strength of the instrument (p < 0.001)12.

The risk-adjusted data are shown in Table 5. Adherence rates for the index therapy were higher for patients who were treated with valsartan/amlodipine SPC relative to ARB + CCB FC therapy. Odds ratios for likelihood of adherence to index therapy ranged from 1.38 to 1.45 depending on the model that we estimated, but all yielded significant relationships in the same direction.

Table 5.  Association between single-pill vs. multiple-pill combination status to adherence (or persistence) to medication.

Dependent variable	Adjusted odds ratio	95% confidence interval		p-value
Adherence to medication (PDC > 0.80)
Propensity score matched	1.38	1.24	1.53	0.0000
Regression analysis	1.42	1.30	1.55	0.0000
IV analysis (sensitivity)	1.45	1.29	1.62	0.0000
Persistence of therapy	Hazard ratio	95% confidence interval		p-value
Propensity score matched	0.87	0.83	0.92	0.0000
Regression analysis	0.86	0.82	0.9	0.0000
IV analysis (sensitivity)	0.85	0.81	0.89	0.0000

Note: (a) One-to-one propensity score matching applied. Models controlled for age, gender, geographic region, baseline pill burden, baseline comorbidities, Deyo–Charlson Comorbidity Index and copayment. (b) For adherence to medication, model logistic regression was used. Persistence of therapy is estimated using Cox model regression analysis controlled for the same factors as controlled for in the propensity score model. (c) Doctors’ prescribing patterns were used as an instrument for instrumental variable analysis.

Moreover, the Cox proportional hazard model showed that patients in the valsartan/amlodipine SPC cohort were less likely to discontinue their medication relative to patients in the ARB + CCB FC group (HR: 0.87, p < 0.001).

Table 6 shows the effect of adherence to medication on total healthcare cost. After controlling for baseline differences, a strong relationship existed between adherence and total healthcare cost. Depending on the model, patients who were compliant to their index medication had an average of 9–13% less in total healthcare cost (p < 0.001).

Table 6.  Association between adherence to medication (or single-pill combination vs. multiple-pill combination) and total healthcare costs.

Exposure variable	Estimate	Standard error	p-value
Adherence to medication (PDC > 0.80)
Propensity score matched	−0.09	0.02	0
Regression analysis	−0.13	0.02	0
IV analysis (sensitivity)	−0.13	0.03	0
Single-pill combination vs. multiple-pill combination
Propensity score matched	−0.16	0.10	0
Regression analysis	−0.20	0.02	0
IV analysis (sensitivity)	−0.18	0.02	0

Note: (a) One-to-one propensity score matching applied. Models controlled for age, gender, geographic region, baseline pill burden, baseline comorbidities, Deyo–Charlson comorbidity index and copayment. (b) Generalized linear models with log link and gamma distribution were used for the regression analysis. Regression analysis controlled for the same factors as controlled for in the propensity score model. (c) Doctors’ prescribing patterns were used as an instrument for instrumental variable analysis.

Comparison between the valsartan/amlodipine SPC and ARB + CCB FC groups yielded that total healthcare costs of patients who used the former were 16–20% lower those who used the latter (p < 0.0001).

Consistent with the adherence and healthcare cost comparison, real-world analysis of healthcare utilization in patients in both groups provided similar results. As depicted in Table 7, after risk adjustment, patients treated with valsartan/amlodipine SPC therapy had fewer ER visits (7.62 vs. 9.51%, p < 0.01), inpatient stays (8.66 vs. 10.13%, p < 0.05), physician office visits (97 vs. 99%, p < 0.001) and outpatient visits (54.12 vs. 60.31%, p < 0.001) compared to patients treated with ARB + CCB FC therapy.

Table 7.  Risk-adjusted association between single-pill vs. multiple-pill combination and healthcare resource utilization.

All patients n	V/A SPC		ARB + CCB MPC		p-value
Post-index period resource use	n/Mean	%/SD	n/Mean	%/SD
Patients with ≥1 ER visit, n (%)	234	7.62%	292	9.51%	0.0072
Patients with ≥1 inpatient stay, n (%)	266	8.66%	311	10.13%	0.0457
Patients with ≥1 physician office visit, n (%)	3007	97.92%	3035	98.83%	0.0047
Patients with ≥1 other out-patient visit, n (%)	1662	54.12%	1852	60.31%	<0.0001

Note: One-to-one propensity score matching applied. Models controlled for age, gender, geographic region, baseline pill burden, baseline comorbidities, Deyo–Charlson Comorbidity Index and co-payment.

Discussion

The prevention and management of hypertension are major public health challenges for the United States. Costs related to hypertension in the United States in 2007 were estimated to be $66.4 billion13. Therefore, it is crucial to have effective therapies with a favorable adherence and cost profile.

Hypertension may also exist in association with other conditions in which there are compelling indications such as the ones involving high-risk conditions that can be direct sequels of hypertension (heart failure, chronic kidney disease, recurrent stroke) or commonly associated with hypertension (diabetes, high coronary disease risk). More than two-thirds of hypertensive individuals cannot be controlled on one drug and will require two or more antihypertensive agents selected from different drug classes to reach target blood measures3. Generally, combination therapy regimens include two or more medications that are either taken separately (such as ARB + CCB FC) or in a fixed-combined formulation.

In this research, we analyzed these two groups using real-world data to see if patient adherence increases with regimen simplicity and if this translates into better economic outcomes, i.e. lower healthcare cost and utilization.

The most effective therapy will control hypertension only if patients are motivated to take the medication as directed. Confirming previous studies14 and consistent with JNC7 recommendations, patients using a product that had a more simplistic regimen were more likely to be compliant with their therapy and maintained a better persistency profile15. Healthcare costs and utilization of patients using the valsartan/amlodipine SPC were significantly lower relative to patients with the ARB + CCB FC therapy.

When comparing the treatment effect in non-experimental studies, we are always faced with overt and hidden bias11,16. Overt bias can exist because the treatment groups are different in terms of certain observable factors, such as baseline demographic and clinical characteristics. We have used both propensity score matching (PSM) and multivariate regression analysis to control for overt bias. However, there also exists hidden bias, which would be a result of failure to control for unobservable factors, such as disease severity. We used the instrumental variable (IV) approach to control for hidden bias in our study. Comparing the estimates from PSM and multivariate regression with the estimates from the IV approach showed that our estimates were robust and did not change with failure to control for unobserved bias.

Although we used advanced techniques to control for both overt and hidden bias, there are limitations to our study as is typical for any retrospective claims database17. While claims data are extremely valuable to research treatment patterns, healthcare resource utilization and cost, these data are collected for the purpose of payment administration rather than research. Presence of a diagnosis code on a medical claim is not necessarily proof of the presence of disease, as diagnoses may be incorrectly coded or included as rule-out criteria rather than actual disease. However, we applied detailed quality checks to the data set before starting the analysis in order to mitigate these problems. Data on socioeconomic status, health behaviors, and patient lifestyle were not available for the analysis. Occurrence of drug prescription fill does not guarantee actual consumption of the drug by the patient. Blood pressure and other clinical data were not available to help determine the severity or stage of hypertensive disease condition.

Conclusion

The main objective of this paper was to examine the risk-adjusted adherence, healthcare cost and utilization rates for patients who received valsartan/amlodipine SPC as compared to ARB + CCB FC therapy.

Using a retrospective analysis of health insurance claims data and controlling for confounders such as demographic and clinical factors as well as baseline healthcare use, we showed that patients who received valsartan/amlodipine SPC therapy had better adherence rates and that their risk-adjusted healthcare costs and utilization were lower than those patients who used ARB + CCB FC therapy. Our sensitivity analysis using the instrumental variable approach showed that our results were not sensitive to unobserved factors in the model.

However, results may only be generalizable to an index therapy treatment naïve patient population, as only patients who did not receive valsartan/amlodipine single-pill combination or ARB + CCB free-combination therapy during the pre-index period were included in the analysis. To be warranted, these conclusions must be supported with more extensive analyses that would show cohorts accurately representing the greater hypertensive population.

Transparency

Declaration of funding

This study was supported by Novartis Pharmaceuticals Corporations, NJ, USA.

Declaration of financial/other relationships

O.B., L.W., and L.X. are employees of STATinMED Research, which was subcontracted by i3 Global for this research. L.M.A. is an employee of Novartis Pharmaceuticals Corporation.

Acknowledgments

Elizabeth Moran, of STATinMED Research, provided editorial support.

These data were previously presented at the American Heart Association Quality of Care and Outcomes Research in Cardiovascular Disease and Stroke, 2010 Scientific Sessions, Washington, District of Columbia, USA, 19–21 May 2010.