- Email: [email protected]
Medication adherence and health care costs with the introduction of latanoprost therapy for glaucoma in a medicare managed care population
The American Journal of Geriatric Pharmacotherapy
M.J. Bhosle et al.
Medication Adherence and Health Care Costs with the Introduction of Latanoprost Therapy for Glaucoma in a Medicare Managed Care Population Monali J. Bhosle, MSI; Gregory Reardon, PhD2; Fabian T. Camacho, MS3; Roger T. Anderson, PhD3; and Rajesh Balkrishnan, PhD I tDepartment of Pharmacy Practice and Administration, The Ohio State University, Columbus, Ohio; 21nformagenics Inc., Worthington, Ohio; and 3Department of Public Health Sciences, Wake Forest University School of Medicine, Winston-Salem, North Carolina
ABSTRACT Background: Latanoprost, a prostaglandin inhibitor, is increasingly being used in the therapeutic management of glaucoma. However, there is scant literature examining the cost and outcome ramifications of latanoprost. Objectives: This study examined the medication use behavior (medication-related persistence and adherence rates) and costs associated with the introduction of latanoprost therapy in a treatment-naive older population (aged _>65 years) enrolled in Medicare. Methods: The study employed a retrospective observational cohort design and used administrative claims data fi'om a Medicare health maintenance organization (HMO), which off~red complete coverage to enrollees, including prescription benefits. The case group consisted of patients with glaucoma who began latanoprost therapy. The control group consisted of enrollees with glaucoma who started any therapy other than latanoprost. Both groups were followed up fbr 1 year befbre and after initiation of therapy. Bivariate and multivariate techniques incorporating health care utilization in the year before the start of new therapy were used to determine the study outcomes. Results: The case group comprised 101 patients (mean age, 77.60 years), while the control group included 168 patients (mean age, 77.59 years). There were no significant differences across the 2 groups with respect to age, sex, general health scores on the 12-item Short-Form Health Survey, severity of comorbidity, or the proportion of respondents with perception of worsened health. Introduction of latanoprost therapy was associated with higher medication persistence (hazard ratio, 0.90; 95% CI, 0.68-0.98) and adherence rates (mean [SD], 0.51 [0.26] vs 0.40 [0.25]; P < 0.001) compared with patients starting other glaucoma medication. Furthermore, there were no additional increases in total health care costs in the entire population associated with the introduction of latanoprost therapy, after adjusting fbr group and time effects, as well as other confbunders (mean [SD], $4718.24 [$8982.92] vs $4046.55 [$6505.39]). Conclusions: Latanoprost therapy offered improved medication use behavior in these older adults enrolled in a Medicare HMO. There were no significant additional increases in overall health care costs as a result of introduction of latanoprost therapy, after adjusting for group and time effects, as well as other baseline confbunders in this study cohort. (Am J Geriatr Pharmacother. 2007;5:100-111) Copyright © 2007 Excerpta Medica, Inc. Key words: glaucoma, open-angle glaucoma, medication adherence, elderly. Accepted for publication February 13, 2007. Printed in the USA. Reproduction in whole or part is not permitted.
I00
June 2007
Volume 5 • Number 2
doi: I 0.1016/j.amjophann,2007.05.004 1543-5946/$32.00
Copyright © 2007 Excerpta Medica, Inc.
M.J. Bhosle et al.
INTRODUCTION
Glaucoma, sometimes referred to as a "silent blinder," is the second leading cause of blindness overall and the leading cause among blacks and Hispanics in the United States. 1-3 Glaucoma is usually asymptomatic until late in its course. As many as 50% of glaucoma patients in developed countries are unaware that they have the disease. 4 The incidence of glaucoma is predicted to be between 1.1% and 3% of the US population, with the incidence increasing to --6% in patients aged _>65 years, s The major risk factors for glaucoma are age _>65 years, family history of glaucoma, certain systemic diseases (eg, hypel"cension, diabetes mellitus, hyperthyroidism), myopia, and black descent. 6 The most commonly occurring fbrm of the disease in western populations is primary open-angle glaucoma, an increase in imraocular pressure (IOP) with resulting damage to the optic nerve, deterioration of visual fields, and eventual loss of vision. Open-angle glaucoma affects >2 million Americans, a finding that is expected to rise to >3 million by 2020 considering the rapid aging of the US population. 7 Treatment fbr glaucoma is usually focused on IOP reduction with medications, as this is believed to prevent or arrest deterioration of the optic nerve and visual fields by reducing aqueous humor production or increasing aqueous outflow. Laser and surgical procedures are mostly reserved for patients who do not respond to drug therapy. Some of the factors that could impact achieving optimal IOP in a patient may include effectiveness of a medication, disease progression, and patient adherence, s,6 In recent years, prostaglandins have emerged as the mainstay of treatment fbr open-angle glaucoma, and latanoprost is fi'equcntly used as primary therapy. .8,9 Latanoprost belongs to the category of prostaglandin analogues and reduces IOP by increasing the uveoscleral flow without affecting the aqueous flow. Latanoprost was introduced in the US market in late 1996 as an antiglaucoma medication and has been widely used as an adjunctive and often primary therapy, possibly due to its QD administration schedule and effective IOP-lowering e f f e c t s . 9 The drug reduces IOP by 30% in patients with open-angle glaucoma and is well tolerated in most patients. 9 In a randomized study of 35 patients with open-angle glaucoma, Polo et al 1° compared the effect on IOP at the end of 2 weeks and 3 months of latanoprost (n = 18) versus timolol plus dorzolamide (n = 17). In this patient cohort, latanoprost (0.005%), administered QD, showed sig*Trademark: Xalatan ® (Pfizer Inc., New York, New York).
The Am.erican Journal of Geriatric Pharmacotherapy
nificantly better long-term reduction (at the end of 3 months of therapy) in IOP than combined therapy with timolol plus dorzolamide (timolol 0.5%, BID; dorzolamide 2%, BID). Similar to latanoprost, other prostaglandin analogues (eg, bimatoprost, travoprost [both approved in 2001]) effectively manage glaucoma by lowering IOP levels, s All these prostaglandins offer a convenient QD dosage schedule. Common adverse effects of these medications include increased iris pigmentation s and hyperemia, predominantly in predisposed eyes. Previous clinical trials indicated hyperemia rates ranging f?om 5% to 15% for latanoprost, 11 15% to 45% for bimatoprost, 12 and 35% to 50% for travoprost. ~3 Latanoprost was the only prostaglandin analogue approved by the US Food and Drug Administration (FDA) as a first-line agent for open-angle glaucoma and ocular hypertension when the present study was conducted ~4 and was generally regarded as a first-line agent by glaucoma spedalists soon after its introduction in 1996. is Bimatoprost was approved by the FDA in 2006 as a first-line therapy fbr elevated IOP-assodatcd openangle glaucoma or ocular hypertension. 16 Medication use behavior, as determined by medicationrelated persistence and adherence among glaucoma patients, is an important issue. Persistence to medication refers to the duration of continuous medication use as prescribed, 17 and adherence refers to the extent to which a person's behavior (eg, following prescribed medication regimen and a diet, and/or executing lifestyle changes) corresponds with advice/recommendations fi'om health care professionals, is Both adherence and persistence to pharmacotherapy arc important in treating chronic conditions, as patients' nonadherence to the treatment regimen may result in poor outcomes, including treatment failure, resistance to the therapy, medication overdose, disease progression, otherwise preventable hospitalizations, and unnecessary medical expenditures. 19 Overall, patients' fhilure to follow the prescribed medication regimen for their chronic condition may cost as much as $300 billion to the US health c a r e s y s t e m s . 19 The lack of overt symptoms early in the stage of disease may tend to decrease the medication-related persistence/ adherence among glaucoma patients. 4 The estimated medication nonadhcrence rate ranged between 23% and 59% fbr glaucoma patients, 2°-22 and the rate was found to be lower (23%) in Medicaid-enrolled patients older than 65 years. 22 In one retrospective study of glaucoma patients enrolled in a large national health care provider obtained through a third-party anonymous source (PharMetrics Inc., Watertown,
101
The American Journal of Geriatric Pharmacotherapy
M.J. Bhosle et al.
Massachusetts), 23 glaucoma pharmacotherapy requiring >2 administrations per day was one of the important predictors of nonadherence. In this population of 1784 patients who had received at least 2 latanoprost prescription refills while on no other glaucoma medication and in whom a second glaucoma medication was then added to their latanoprost regimen, mean (SD) refill intervals fbr latanoprost were 40.6 (21.8) days before the addition of a second drug and 4Z4 (24.4) days after the addition of a second drug. Compared with other therapies, the use of latanoprost (with its QD administration schedule and lower systemic effects) may be associated with improvements in medication use behavior. 24 In a study by Nordstrom et al, 2s patients continuously enrolled for at least 1 year before the initial glaucoma diagnosis date (UnitedHeakhCare) were identified from the Ingenix Research Database between July 1, 1995, and December 31, 2001. In this retrospective claims data analysis, the study cohort consisted of patients who were newly diagnosed with open-angle (diagnosed glaucoma, n = 3623) or strspect (n = 1677) glaucoma and who received at least one dispensing of a topical ocular hypotensivc medication. Prostaglandin analogues had higher medication-related persistence compared with J]-blockcrs (hazard ratios fbr discontinuation, 0.40 [95% CI, 0.35-0.44] fbr diagnosed glaucoma patients and 0.44 [95% CI, 0.37-0.52] fbr patients with suspect glaucoma). However, none of these studies reported medication use behavior associated with the introduction of latanoprost therapy in patients aged >65 years. Improved medication-related persistence and adherence are associated with improved therapy outcomes and lower utilization of health care services, which in turn reduce treatment-related and overall health-related costs in patients with chronic diseases. 2°-23 However, previous research has not tested whether the frequency of dosing and tolerability of therapy, as effects isolated from product or therapy class selection, are related to adherence or persistence. The primary objective of the current study was to compare the medication use behavior (persistence and adherence) and costs impact in Medicare-aged glaucoma patients who were newly treated with latanoprost versus other glaucoma medications. Because of its acceptance as a first-line agent and as the mostprescribed alternative to traditional standard therapies, 8-1° including the [3-blocker class, and newer agents such as dorzolamide/timolol, we chose to compare the use of latanoprost with other ocular hypotensive agents.
102
MATERIALS AND METHODS Study Population The target population induded older adults (aged _>65 years) with primary open-angle glaucoma enrolled in a Medicare health maintenance organization (HMO) in the southeastern United States. This managed care plan was operational in late 1996 and was the sole provider of medical care to its enrollees ("lock-in" risk-benefit plan). On enrollment in the plan, the enrollee was mailed a questionnaire that induded questions about demographics, clinical conditions, health service utilization in the year before joining the plan, lifestyle, depression, and quality of life. Since its inception in 1996, the response rate for this questionnaire has been >80%. 26 A comprehensive range of preventive, diagnostic, and treatment services is provided by plan providers that includes all necessary office visits. Prescription drugs are covered according to a 2- or 3-tiered open formulary. For prescription medications prescribed by a plan physician and obtained at a plan pharmacy, patients pay a $6 generic or $12 name brand copayment per prescription unit or refill. The 12-item Short-Form Health Survey (SF-12) was used to assess general health status. The SF-12 is a self administered questionnaire that evaluates symptoms, flmctioning, and quality of lifE. A higher score indicates better quality of life in patients, and 0 indicates worst possible quality of life (eg, near-death or worse-than-death experience, especially in elderly patients with chronic conditions and comorbidities). Enrollment in the plan was tracked fi'om 1996 until December 2002. During this period, there were a total of 14,654 patients aged _>65 years. The mean (SD) age of this cohort was 74.69 (6.81) years. Eligible patients were required to have a previous medical claim coded for open-angle glaucoma during the study period (International Classification of Diseases, Ninth Revision [ICD-9] code 365.1), and at least one prescription fill for a glaucoma medication during the study period.
Study Design The study had a retrospective observational cohort design. The study compared medication persistence, medication adherence, and health care costs among 2 patient cohorts starting a new glaucoma medication. The study covered the period from January 2000 to December 2002 (36 months). The earliest and specific glaucoma medication dispensed in calendar year 2001 was identified as the index agent. The time frame extended from 1 year before (preindex) to i year after
M.J. Bhosle et al.
(postindcx) the start date (index date) of latanoprost index therapy (comprising the case group) and the same time intervals for the non-latanoprost index medications (collectively comprising the control group). To assure that the patient was new m the index agent, patients in both the case cohort and control cohorts were required to not have a previotts fill for the specific index agent for a period of 1 year before the index date. Patients aged <65 years old and those who had a prescription claim fbr any open-angle glaucoma prescription in the prcindex period were excluded fi'om the study. M e a s u r e m e n t and O u t c o m e s
Prescription refill patterns were used to derive measures of adherence and persistence under the assumption that a prescription filled was a prescription taken. Pharmacy records have been shown to be valid measures of cumulative exposure and gaps in medication supply. 27 We performed random checks on pharmacy claims to compare days of prescription supply against the quantity of medication dispensed to determine if there was a reasonable fit with the figure entered, which we found true in all cases examined. We further confirmed days of supply for each medication based on bottle size using the methodology followed by Nordstrom ctal. 2s The days' supply calculations were based on the fbllowing assumptions: for most nonprostaglandins, up to 5 mL were assumed to provide 60 days of therapy; up to 10 mL were assumed to provide 90 days of therapy; and >10 mL were assumed to provide 120 days of therapy. The drugs other than prostaglandins are used QD (eg, timolol gel, levobunolol 0.5%) or BID (eg, timolol solution, levobunolol 0.25%) depending on the product formulation and the ophthalmologist's/manufacturer's recommended dosing. The prostaglandins are indicated for QD use. Because the nonprostaglandin products prescribed for study patients were generally BID fbrmulations, we adjusted the expected days' formulation for prostaglandins to last longer, though not twice as long, as the nonprostaglandins. Accordingly, for prostaglandins, we assumed that: up to 2.5 mL provided 60 days of therapy, up to 5 mL provided 90 days of therapy, and > 5 mL provided 120 days of therapy. Working Definition of Persistence Rate Persistence to the new therapy was calculated with survival analysis techniques. We obtained the time to discontinuation of initial therapy by finding the number of days from the starting date to the end of the last dispensing (ie, last fill date plus days' supply)
The Am.erican Journal of Geriatric Pharmacotherapy
of the index agent befbre a gap in prescription supply. In the analysis, discontinuation of therapy (discontinuation date) was defined as no further index drug refill 60 days (if dispensed 1 bottle) or 120 days after the last prescription fill (if dispensed >1 bottle). Patients who reached their study end date without discontinuing the initially prescribed therapy were censored as of the end date. The Kaplan-Meier plots showed the time to discontinuation of therapy for the 2 study cohorts (case vs control) as well as fbr latanoprost and the individual medication classes formed by subcategorizing the control group as [8-blockers, combination dorzolamidc/ timolol, and other medication class (Table I). Working Definition of Adherence Rate An index of glaucoma medication utilization for the postindex year was computed for each patient based on the Med-Total approach by Stcincr ct al. 2s The Medication Possession Ratio (MPR) was calculated as the total nmnber of days of glauconla prescription supply dispensed in this period divided by the number of days between the index date and the dispensing date of the last prescription in this period. The number of days a person was in the hospital was subtracted from the denominator because any drug taken during this time may have been dispensed by the hospital; inpatient pharmacy claims were not available for observation. As suggested by Stciner et al, MPRs >1 or <0 were truncated to i and 0, respectively. Health Care Costs The study examined costs associated with total health care utilization that included all medical and pharmacy claims, as opposed m examining glaucomarelated utilization and costs. This was done to avoid problems inherent in attributing all billings for health care service utilization to a specific condition due to miscoding of diagnosis. 29 Patients were fbllowcd up for complete health care service utilization (ie, hospitalizations, emergency dcpamncnt visit, outpatient physician visits, utilization of glaucoma medication) and costs 1 year before and 1 year after the index date. Actual costs paid by the plan m providers for medical, hospital, and pharmacy services were used m estimate total costs for each of these years. Statistical Analysis Descriptive (frequencies and percentages) analyses of baseline characteristics and univariate (paired t tests and Z2) analyses were conducted comparing demographics and study outcome variables across groups.
103
The Am.erican Journal of Geriatric Pharmacotherapy
M . f Bhosle et al.
Table I. Other glaucoma medications used in the study (control group comprised patients who started any therapy other than latanoprost*; case group consisted of patients who started with latanoprost therapy [2.5 mL, a 60 days' supply]).
Medication
No. of Patients Who Started with the Medication
Bottle Size (No. of Patients Who Started on the Bottle Size)
Usual Dosage
Carbonic anhydrase inhibitor Dorzolamide hydrochloride ophthalmic solution Brinzolamide ophthalmic suspension
16 I0 6
5 mL (n - 5), 10 mL (n - 5) 5 mL (n - 4), 10 mL (n - 2)
TID TID
t-Blockers Timolol maleate Timolol maleate (gel-forming solution) Levobunolol hydrochloride ophthalmic solution Metipranolol ophthalmic solution Carteolol hydrochloride
69 39 7 21 I I
5 mL (n = 39) 5 mL (n - 7) 5 mL (n = 9), I0 mL (n = 9), 15 mL (n = 3) I0 mL (n = I) I0 mL (n -- I)
BID QD BID BID BID
Other prostaglandin analogues Bimatoprost ophthalmic solution Travoprost solution
21 II I0
2.5 mL (n - 7), 5 mL (n - 4) 2.5 mL (n - 10)
QD QD
ot-Agonists Apraclonidine solution Brimonidine tartrate ophthalmic solution Propine solution (dipivefrin hydrochloride) Combination dorzolamide/timolol Dorzolamide hydrochloride/timolol maleate ophthalmic solution Unoprostone isopropyl ophthalmic solution
6 I 5 0
5 mL (n - I) 5mL(n- I),10mL(n-3),15mL(n-
I)
TID TID BID
55 55
I
5 mL (n = 12), 10 mL (n = 43)
BID
5mL(n=
BID
I)
*Days' supply assumptionsfor the control gpoup were as follows: up to 5 mL 60 days;up to I0 mL, 90 days;and >10 mL 120 days.
Survival and multiple regressions were used to determine the medication-related persistence, adherence, and cost impact of latanoprost therapy compared with other glaucoma medications. Survival techniques were used to analyze predictors of medication-related persistence. Cox proportional hazards models were constructed fbr latanoprost versus other glaucoma medication groups (model 1) and for individual medication class using [3-blockers as the reference group (model 2). Other variables included in the models were patients' demographics (eg, age, sex). All refills of the index prostaglandin were tracked during the postindex year. Persistence was identified as the duration of continuous treatment with the index medication. 25 However, within the regression model, patients who experienced the introduction of a new ocular hypotensive in the postindex year were classified as such with the dichotomous variable "combination therapy." If a patient had switched f?om the index agent to a new agent, this would have been indicated
104
in the model as a discontinuation of the index agent, rather than a "switch" per se. The fbllowing regression model fbr adherence was used for each of the 2 outcomes: outcome = f (demographic and clinical confounders, treatment group indicator, year indicator, year x group interaction), where outcomes were annual M P R scores and total annual health care costs. The year indicator and year x group interaction variables were included only in the cost regression, as adherence was measured the postindex year only. The year x group interaction represents the relative cost impact of one therapy compared with another in the cost regression. The regression controlled for demographic and clinical confounders, including age, sex, total number of prescription medications during the 2-year period, and propensity for high health care utilization in the preindex year. Because most patients did not have other comorbid conditions listed, we used their health care utilization in the preindex year as a risk stratifier.
M.J. Bhosle et al.
The dependent variable in the cost impact regression was annual health care costs reimbursed by Medicare. The distribution of health care costs was skewed (as determined by the Shapiro-Wilk test3°). To avoid potential estimation problems in the model such as heteroscedasticity, the natural logarithm of cost was used as the adjusted dependent variable for the analyses. Parameter estimates obtained from regressing log-transformed costs on covariates were interpreted by using the antilog of model estimates. 31,32 Multiple linear regressions using the postindex year data were employed to obtain the differences in treatment adherence between the 2 groups fbr both sets of comparisons. This regression adjusted for the same set of confounders described earlier. The dependent variable was adherence to the glaucoma medication regimen in the postindex year. The regressions also checked for the presence of multicollinearity (linear relation between predictor variables) and autocorrdation (correlation between sequential values [in this case, yearly values] of variables). Correlation coefficients (R 2) between the predictor variables were not greater than 0.3, indicating that multicollinearity effects were not significant. Results of the Durbin-Watson test performed on the regressions were <2 fbr all the regressions, indicating that there was no auto correlation problem within the data. All statistical analyses were conducted using the Stata 9.0 software (StataCorp LP, College Station, Texas).
The American Journal of Geriatric Pharmacotherapy
Table II. Baseline characteristics of the study population.
Parameter
Case Group ( n - 100)
Contr-olGroup ( n - 168)
Age, y Mean (SD) Range
77.60 (6.54) 65-98
77.59 (6.72) 65-96
Male sex, %
30
35
59.50 (21.31) 25-I00
62.36 (21.98) 0-I 00
3.02 (2.58) 0-I 2
3.09 (2.36) 0-I 2
Proportion of respondents perceiving worse health Mean (SD) Range
0.037 (0.19) 0-1
0.075 (0.26) 0-1
Total no. of prescription medications* Mean (SD) Range
36.98 (28.41) 0-140
31.75 (24.53) 0-149
0.80
0.75
SF-12 general health scores Mean (SD) Range Charlson's comorbidity severity index score Mean (SD) Range
Proportion on monotherapy
SF-12 = 12-item Short-Form Health Sur~ey. *P < 0.05;meantotal number of prescriptionstaken by patients in caseand control groups reflecLsa mix day supplies.
RESULTS
The study sample represented -0.02% of the elderly population (n = 14,546) in the health care plan. Table II compares the adjusted descriptive characteristics of the 2 study populations. The case group comprised 101 patients, while the control group included 168 patients. One outlier patient in the case group who had total health care costs of $338,000 in the postindex year was dropped f~om this model since such high costs usually represent end-of-life care. There were no significant differences across the 2 groups with respect to age, sex, general health scores on the SF-12, severity of comorbidity, or the proportion of respondents with perception of worsened heakh. A total number of mean prescription medications during the 2-year period were higher for patients in the case group compared with patients in the control group (36.98 vs 31.75; P < 0.05). Bivariate tests indicated no significant differences in the health care costs in both the years between the 2 comparison groups (cases vs controis) (Table III).
The regression analysis for persistence and adherence rates included data from the postindex year only. Starting with latanoprost therapy was associated with a slightly better persistence rate compared with starting with other medications (hazard ratio of discontinuation, 0.90; 95% CI, 0.68-0.98) (Figure 1, Table IV [model 1]). Furthermore, patients' discontinuation rate of initially prescribed medication was lower for latanoprost compared with ]3-blockers (hazard ratio of discontinuation, 0.92; 95% CI, 0.69-0.96) (Figure 2, Table IV [model 2]). Persistence rate to combination dorzohmide/tinlolol was higher compared with the IB-blockers, as indicated by a hazard ratio of discontinuation for combination dorzolamide/timolol compared with 13-blockers of 0.89 (95% CI, 0.84-0.98). It is noteworthy that the Kaplan-Meier curves are flat for the first 60 days because patients could not discontinue their therapy during that time. The covariate "total nmnber of prescription medications" had a significantly lower risk of discontinuation in models 1
105
The Am.erican Journal of Geriatric Pharmacotherapy
M.J. Bhosle et al.
and 2; each additional prescription used by the patient was associated with a 1% and 4% decrease in probability of discontinuation, respectively. Table V displays the results of the comparison of adherence rates and health care costs between the case group and the comrol group using multiple regression analysis. The mean adherence rate for the case group was significantly higher than that of the control group after adjusting fbr other variables ([3 = 0.057; P < 0.01). Introduction of latanoprost therapy was not associated with a significant increase in costs over other glaucoma medications. Sensitivity
Table III. Study outcomes (unadjusted). Case Group (n - I00)
Parameter
Analysis
The appropriateness of the specification of the dependent variable for the cost impact analysis was examined. This was done by observing both logged (natural logarithm of annual health care costs as a dependent variable) and unlogged (annual health care costs as a dependent variable) spedfications of the annual health care costs in the regression model used for comparing total health care costs between the case and control groups. Both variations of the spedfication produced an identical set of significant predictor variables. The direction of effect of the significant variables remained
Control Group (n - 168)
Total health care costs in preindex year, $ Mean (SD) Range
3749.97 (6892.24) 3097.51 (4983.15) 0-44,739.38 0-37,190.37
Total health care costs in postindex year; $ Mean (SD) Range
4718.24 (8982.92) 4046.55 (6505.39) 144.51-73,779.71 0-45,974.35
Persistence rate to new medication therapy (postindex year) Mean (SD) Range
0.75 (0.32) 0-1
0.68 (0.36) 0-1
Adherence rate to new medication therapy (postindex year-)* Mean (SD) Range
0.51 (0.26) 0-1
0.40 (0.25) 0-1
*P < 0.001.
Latanoprost . . . . OGM 1.00 ..........................
4-J
.............
0.75 -
.., |--%
L
'
OO
!
0.50 -
O O O-
,. . . . , "I
o O_
i L--'I
0.25 . . . . . . ,.,. . . . . . . . . . . . . . .
~ ...........................
,o
0-I
I
I
I
I
0
100
200
300
400
Days After Glaucoma Treatment Initiation Figure I. Proportion of patients who remained on the initially prescribed glaucoma medication therapy. Note: KaplanMeier curve adjusted for age, sex, total number of prescription medications, and combination therapy shows time to discontinuation after starting the new therapy for the 2 study cohorts (latanoprost group vs other glaucoma medication [OGM] group).
106
M.J. Bhosle et al.
The American Journal of Geriatric Pharmacotherapy
Table IV. Adjusted Cox proportional models for time to discontinuation of the initially prescribed glaucoma medication therapy.* Values are given as hazard ratio (95% CI). Model I (Latanoprost vs OGMs)
Model 2 (Based on Medication Class)
0.90 (0.68-0.98)t Reference category -
0.92 (0.69-0.96)# Reference category
1.01 (0.68-1.51) 1.00 (0.99-1.00) I. 16 (0.87-1.54)
0.89 (0.84-0.98)t 1.95 (I,37-1.77)# 1.00 (0.67-1.49) 1.00 (0.99-1.00) I. 13 (0.85-1.51)
0.99 (0.97-0,99)*
0.96 (0,88-0.99)#
0.86 (0.64-1.15)
0.88 (0.66-1.18)
Variable Latanoprost OGMs 13-Blockers Combination clorzolam (dorzolam ide/timolol) Other medication class# Age Age squared§ Male sex Total no, of prescription medications Combination therapy (>1 glaucoma medication in postindex group)
OGMs - other glaucoma medications. *Per~sistencewas calculated using only postindex year dat~ tP < 0.05. Other medication classFor model 2 included carbonic anhydr~se inhibitors,other pros~aglandin analogues, 0~-agonis~s,and unoprostone isopr~)pylophthalmic solution. §Age squared = age. age..
. . . . j3-Blockers Latanoprost Combination dorzolamide/timolol .... Other
1.00 -
f~
. . . . . . .
0.75 -
. . . .
I
e°m
O| O c-
0.50 -
. . . . . .
! I
O
O O-
La _ _
~
L-------- ..._1
! . . . .
o O_
i-~_~--~___~__ ~_,,~.-~~
7
I ' !
0.25 -
"-I I__ . . . . . . . . . .
I
- I
_ _
-,
I. . . . . . . .
L
L. L° L. . . . . L.
OI
I
0
100
I 200
I 300
I 400
Days after Glaucoma Treatment Initiation Figure 2. Proportion of patients who remained on the initially prescribed glaucoma medication therapy (by medication class). Note: Kaplan-Meier curve adjusted for age, sex, total number of prescription medications, and combination therapy shows time to discontinuation after starting the new therapy (latanoprost, J3-blockers, combination dorzolamide/timolol, other medications).
107
The American Journal of Geriatric Pharmacotherapy
M.J. Bhosle et al.
Table V. Comparison of adherence rates and total health care costs between case and control groups.*
Dependent Variables (13 [SE-])
Predictor Variables
Adherence Ratet
Age Age squared Male sex
0.05 (0.04)
-0.14 (0.18)
-0.00033 (0.0003)
0.0009 (0.001)
-0.055 (0.04)
0.32 (0. II)
High health care costs (>$4000) in preindex year Latanoprost
Natural Log of Total Health Care Costs
NA
2.21 (0.13)¢
0.057 (0.03)t
0.14 (0.17)
Year
NA
-0.62 (0.15)¢
Latanoprost group-year interaction
NA
0.085 (0.23)
Total no. of prescription medications
--0.003 (0.007)~
Combination therapy (>1 glaucoma medication in the postindex group)
Constant
0.009 (0.002)~
0.21 (0.03)
0.064 (0.16)
-1.90 (1.63)
12.14 (1.68)
0.052
0.25
Adjusted R2
NA = not applicable. *Reference groups (omitted categories) include:control, female sex,costs _~$4000/yearin preindex yea~,patients with no combination therapy,and postindex ye~ t Adherence was calculated usingonly postindex year data. ~:P < 0.01. §P < 0.05.
the same. The logged specification produced more conservative estimates in the magnitude of estimation; for example, in the logged specification model, high health care costs in the prcindcx year were associated with increased annual health care costs (2.21 [0.13]; P < 0.001). The sensitivity analysis using unlogged specification indicated similar direction of effects in the association ($955Z35 [$1055.25]; P < 0.01). DISCUSSION
The results of this study highlight the importance of the introduction of latanoprost therapy in improving medication-related adherence among patients aged _>65 years with glaucoma. In this study, after adjusting for the effects of patient characteristics, patients who started with latanoprost had higher medication adherence rates compared with patients starting with
108
other glaucoma medications. Patients' time to discontinuation of initially prescribed glaucoma medication therapy was lower for the latanoprost group when compared with other glaucoma medications, thus indicating higher persistence for the latanoprost group. Further categorization of the control group indicated that latanoprost and combination dorzolamide/timolol had better persistence rates compared with ]3-blockers. In the regression analysis, a higher number of total prescription medications was associated with a lower medication adherence rate. The negative association between total number of prescriptions and patients' medication adherence has been reported in the literature. 29 Patients in the latanoprost group had a higher number of total prescription medications compared with patients in the control group; their adherence and persistence rates were higher.
M.J. Bhosle et al.
Previous retrospective cohort studies have compared the medication persistence rate in patients (aged <65 years) treated with latanoprost versus 13-blockers, brimonidine, or carbonic anhydrase inhibitors. These studies found that patients treated initially with latanoprost monotherapy remained on therapy significantly (P < 0.001 to P < 0.05) longer than those treated with comparator drugs. As reported by Schwartz et al, 33 in their retrospective cohort study of a managed care database, glaucoma suspect patients treated with timolol monotherapy were 39% (95% CI, 1.21-1.58) more likely to discontinue and 27% (95 % CI, 1.13-1.43) more likely to change therapy (P < 0.001 for both comparisons) compared with latanoprost monotherapy. A study comparing persistency rates across topical prostaglandins in >4000 patients followed up for 15 months fbund that glaucoma patients treated with latanoprost were significantly (P < 0.001) more likely to continue therapy than patients receiving either bimatoprost or travoprost. 34 Wilensky et al 3S compared persistency rates across topical prostaglandins using an algorithm to correct for inconsistent data reported in quantity supplied and days' supply field of database used (IMS Health LifeLink database wifla pharmacy claims). This study reported significantly higher (P _< 0.05) mean number of days adherent to therapy for bimatoprost-treated patients (291.2 days) compared with latanoprost-treated patients (281.0 days) but not significantly different from travoprost-treated patients (287.0 days). In the current study, introduction of latanoprost therapy was not associated with a significant increase in total health care costs compared with other glaucoma medications. These results are consistent with the previous pilot study (N = 78) conducted by Balkrishnan et a136 in glaucoma patients. This study reported no significant additional increases in total health care charges in the study population associated with the introduction of latanoprost therapy, after adjusting for group and time effects, as well as other baseline confounders. The current study, however, should be considered an initial exploration of these issues, and caution is required when interpreting these findings, due to a number of study limitations. First, the observational study design does not permit causal inference of results. In spite of confounder adjustment, the study results may be subject to some issues of treatment group selection. Also, the measure of adherence depended on pharmacy records. Even though this method has been shown to be reliable, 27 it cannot guard against
The Am.erican Journal of Geriatric Pharmacotherapy
instances of undetected adherence, such as the case when a hospitalized patient is given a generous supply of medication at time of discharge, and no record of the medication exists in the patient's pharmacy data. As our previous research with topical medications has highlighted the measurement issues associated with assessing retrospective medication use with retrospective data, we have used mukiple measures (both persistence and adherence), which have been used previously to describe medication use behavior. 37,38 Additionally, we continued the days' supply recorded in the data set by calculating days of medication supply based on bottle size of each glaucoma medication. Our algorithm fbr estimating persistence on medications other than prostaglandins assumed that most such prescriptions were dosed BID. There was a greater incidence of BID formulations of nonprostaglandins in the actual number of prescriptions dispensed. Due to the reliance of this study on a claims database, we could not observe how often ophthalmologists actually directed the patient to dose these agents. If QD dosing for nonprostaglandins was more common than assumed, our findings could have understated nonprostaglandin persistence. Furthermore, our measures of adherence, while accounting for many patient medicationtaking patterns, still do not completely account fbr the impact of product switching and combining, as well as greater number of people being on higher doses in one group. We also relied on the electronically recorded ICD-9 codes and prescriptions to assign our cases and controls, making the assumption that the coding and recording of these was done accurately. Less than 5% of data had missing values, and therefore we did not use any specific techniques for handling missing data. Finally, there may have been selection bias because we were unable to assign patients randomly to the case and control groups. This study included only those patients who did not have a claim tbr any glaucoma medication 12 months befbre the index prescription date. Medication use behavior among these medicationnaive patients may differ from those who have experience with the glaucoma medication therapy. Rates of adverse effects vary across different medications used for glaucoma, which may affect patients' medication use behavior. Our study did not capture differences in adverse-effect profile for different glaucoma medications studied as these data were not available. Furthermore, results do not show the clinical impact of glaucoma medications. However, it is ideal to choose a drug that patients are more adherent to and persistent with; that is, the more adherent they are, the more
109
The Am.ericanJournal of Geriatric Pharmacotherapy M.J.Bhosleet al.
likely the drug is having an impact on patients' IOP compared with not taking it at all. Future studies need to focus on larger sample sizes to make these results more generalizable. These studies would have to be conducted in larger MedicareH M O systems to obtain a larger number of patients in each treatment group category. Additional studies are also needed with longer periods of follow-up to estimate costs associated with prolonged use of latanoprost therapy compared with other treatment options for glaucoma. CONCLUSIONS After adjusting fbr potential confbunders, we found higher medication persistence and adherence rates in glaucoma patients who started with latanoprost therapy as opposed to patients starting treatment with other glaucoma medication. There were no significant additional increases in overall health care costs among patient groups as a result of introduction oflatanoprost therapy, after adjusting for group and time effects, as well as other baseline confbunders. ACKNOWLEDGMENTS This study was funded by a grant from Pfizer Inc., New York, New York. Dr. Reardon is a consultant for Pfizer Inc. REFERENCES 1. Coleman AL. Glaucoma. Lance. 1999;354:1803-1810. 2. Sommer A, Tielsch JM, Katz J, et al. Racial differences in the cause-specific prevalence of blindness in east Baltimore. N E t ~ l J M e d . 1991;325:1412-1417. 3. Rodriguez J, Sanchez R, Munoz B, et al. Causes of blindness and visual impairment in a population-based sample of US Hispanics. Ophthahnology. 2002;109:737-743. 4. Quigley HA, Vitale S. Models of open-angle glaucoma prevalence and incidence in the United States. Invest Ophthahnol Vis Sci. 1997;38:83-91. 5. Infeld DA, O'Shea JG. Glaucoma: Diagnosis and management. Postgrad Med J. 1998 ;74:709-715. 6. Gordon MO, Kass MA. The Ocular Hypertension Treatment Study: Design and baseline description of the participants. Arch Ophthalmol. 1999;117:573-583. 7. Friedman DS, Wolfb RC, O'Colmain BJ, et al, for the Eye Diseases Prevalence Research Group. Prevalence of open-angle glaucoma among adults in the United States. Arch Ophthalmol. 2004;122:532-538. 8. Marquis RE, Whitson IT. Management of glaucoma: Focus on pharmacological therapy. Drugs Aging. 2005; 22:1-21.
I10
9. Fiscella RG. Costs of glaucoma medications. AmJHealth Syst Pharm. 1998 ;55:272-275. 10. Polo V, Larrosa JM, Ferreras A, Honrubia FM. Latanoprost vs combined therapy with timolol plus dorzolamide in open-angle glaucoma: A 24-month study. Ann Ophthahnol. 2005;37:33-36. 11. Xalatan [package insert]. Peapack, NJ: Pharmacia Corporation; 2000. Available at: http://www.fda.gov/ cder/foi/label/2000/2059 7s141bl.pd£ Accessed April 12, 2007. 12. Lumigan [package insert]. Irvine, Calif: Allergan, lnc; 2001. Available at: http://www.fda.gov/cder/foi/label/ 2001/212751bl.pd£ Accessed April 12, 2007. 13. Travatan [package insert]. Fort Worth, Tex: Alcon Laboratories, Inc; 2001. Available at: http://www.fda. gov/cder/foi/label/2001/212571bl.pd£ Accessed April 12, 2007. 14. Xalatan [package insert]. New York, NY: Pfizer Inc.; 2003. Available at: http://www.f~la.gov/cder/foi/ label/2001/20597s231bl.pd£ Accessed April 12, 2007. 15. Xalatan marks 10th amfiversary. Prostaglandin glaucoma medication overcame obstacles. Available at: h t t p : / / www.ophmanagement.com/article.aspx? article- 8664 6. Accessed December 17, 2006. 16. Waknine Y. FDA approvals: Lumigan, cesamet, onmitrope. Medscape Medical News 2006. Available at: h t t p : / / www.medscape.com/viewarticle/54 0937. Accessed December 20, 2006. 17. DiMatteo MR, Giordani PJ, Lepper HS, Croghan TW. Patient adherence and medical treatment outcomes: A meta-analysis. Med Care. 2002 ;40:794-811. 18. Haynes RB, McDonald H, Garg AX, Montague P. Interventions fbr helping patients to fbllow prescriptions for medications. Cochrane Database Syst Rev. 2002;2: CD000011. 19. DiMatteo MR. Variations in patients' adherence to medical reconmlendations: A quantitative review of 50 years of research. Med Care. 2004;42:200-209. 20. Rotchfbrd AP, Murphy KM. Compliance with timolol treatment in glaucoma. Eye. 1998 ;12:234-236. 21. Patel SC, Spaeth GL. Compliance in patients prescribed eyedrops for glaucoma. Ophthalmic Surg. 1995 ;26:233236. 22. Gurwitz JH, Glyml RJ, Monane M, et al. Treatment fbr glaucoma: Adherence by the elderly. AmJPublic Health. 1993;83:711-716. 23. Robin AL, Covert D. Does adjunctive glaucoma therapy affect adherence to the initial primary therapy? Ophthalmology. 2005 ;112:863-868. 24. Shaya FT, Mullins CD, Wong W, Cho J. Discontinuation rates of topical glaucoma medications in a managed care
M.J. Bhosleet al.
population. A m J Manag Care. 2002;8(Suppl 10): $271-$277. 25. Nordstrom BL, Friedman DS, Mozaffari E, et al. Persistence and adherence with topical glaucoma therapy. A m J Ophthalmol. 2005 ;140:598-606. 26. Balkrishnan R, Anderson RT. Predictive power of a riskassessment questiom~aire across different disease states: Results in an elderly managed care enrolled population. A m J M a n a g Care. 2001;7:145-153. 27. Choo PW, Rand CS, Inui TS, et al. Validation of patient reports, automated pharmacy records, and pill counts with electronic monitoring of adherence to antihypertensive therapy. Med Care. 1999;37:846-857. 28. Steiner JF, Koepsell TD, Fihn SD, Inui TS. A general method of compliance assessment using centralized pharmacy records. Description and validation. Med Care. 1988;26:814-823. 29. Balkrishnan R, Rajagopalan R, Camacho FT., et al. Predictors of medication adherence and associated health care costs in an older population with type 2 diabetes mellitus: A longitudinal cohort study. Clin Ther. 2003 ;25:2958-2971. 30. Shapiro SS, Wilk MB. An analysis of variance test for normality. Biometrika. 1965 ;52:591-611. 31. Kennedy PE. Estimation with correctly interpreted dummy variables in semilogarithmic equations. A m Econ Rev. 1981;71:801.
The Am.ericanJournal of Geriatric Pharmacotherapy
32. Maiming WG, Mullahy J. Estimating log models: To transform or not to transfbrm? J Health Econ. 2001;20: 461-494. 33. Schwartz GF, Reardon G, Mozafthri E. Persistency with latanoprost or timolol in primary open-angle glaucoma suspects. A m J Ophthalmol. 2004;137(Suppl 1): $13-$16. 34. Reardon G, Schwartz GF, Mozafthri E. Patient persistency with ocular prostaglandin therapy: A population-based, retrospective study. Clin Ther. 2003 ;25:1172-1185. 35. Wilensky J, Fiscella RG, Carlson AM, et al. Measurement of persistence and adherence to regimens of IOP-lowering glaucoma medications using pharmacy claims data. A m J Ophthalmol. 2006 ;141 (Suppl 1):$28-$33. 36. BalkrishnanR, Bond JB, ByerlyWG, etal. Compliance and costs associated with latanoprost therapy. Introduction in a managed care setting. Presented at: Aimual Meeting of American Academy of Ophthalmology; October 2225, 2000; Dallas, Texas. 37. Balkrishnan R, Bond JB, Byerly WG, et al. Medicationrelated predictors of health-related quality of life in glaucoma patients enrolled in a Medicare heakh maintenance organization. A m J Geriatr Pharmacother. 2003; 1:75-81. 38. Balkrishnan R, Christensen DB. A comparison of medication adherence indices to assess long-term inhaled corticosteroid medication use. JAsthma. 2001;38:91-98.
Address correspondence to: Rajesh Balkrishnan, PhD, The Ohio State University College of Pharmacy and School of Public Health, 500 W. 12th Avenue, Columbus, O H 43210. E-mail: [email protected]
III
M.J. Bhosle et al.
Medication Adherence and Health Care Costs with the Introduction of Latanoprost Therapy for Glaucoma in a Medicare Managed Care Population Monali J. Bhosle, MSI; Gregory Reardon, PhD2; Fabian T. Camacho, MS3; Roger T. Anderson, PhD3; and Rajesh Balkrishnan, PhD I tDepartment of Pharmacy Practice and Administration, The Ohio State University, Columbus, Ohio; 21nformagenics Inc., Worthington, Ohio; and 3Department of Public Health Sciences, Wake Forest University School of Medicine, Winston-Salem, North Carolina
ABSTRACT Background: Latanoprost, a prostaglandin inhibitor, is increasingly being used in the therapeutic management of glaucoma. However, there is scant literature examining the cost and outcome ramifications of latanoprost. Objectives: This study examined the medication use behavior (medication-related persistence and adherence rates) and costs associated with the introduction of latanoprost therapy in a treatment-naive older population (aged _>65 years) enrolled in Medicare. Methods: The study employed a retrospective observational cohort design and used administrative claims data fi'om a Medicare health maintenance organization (HMO), which off~red complete coverage to enrollees, including prescription benefits. The case group consisted of patients with glaucoma who began latanoprost therapy. The control group consisted of enrollees with glaucoma who started any therapy other than latanoprost. Both groups were followed up fbr 1 year befbre and after initiation of therapy. Bivariate and multivariate techniques incorporating health care utilization in the year before the start of new therapy were used to determine the study outcomes. Results: The case group comprised 101 patients (mean age, 77.60 years), while the control group included 168 patients (mean age, 77.59 years). There were no significant differences across the 2 groups with respect to age, sex, general health scores on the 12-item Short-Form Health Survey, severity of comorbidity, or the proportion of respondents with perception of worsened health. Introduction of latanoprost therapy was associated with higher medication persistence (hazard ratio, 0.90; 95% CI, 0.68-0.98) and adherence rates (mean [SD], 0.51 [0.26] vs 0.40 [0.25]; P < 0.001) compared with patients starting other glaucoma medication. Furthermore, there were no additional increases in total health care costs in the entire population associated with the introduction of latanoprost therapy, after adjusting fbr group and time effects, as well as other confbunders (mean [SD], $4718.24 [$8982.92] vs $4046.55 [$6505.39]). Conclusions: Latanoprost therapy offered improved medication use behavior in these older adults enrolled in a Medicare HMO. There were no significant additional increases in overall health care costs as a result of introduction of latanoprost therapy, after adjusting for group and time effects, as well as other baseline confbunders in this study cohort. (Am J Geriatr Pharmacother. 2007;5:100-111) Copyright © 2007 Excerpta Medica, Inc. Key words: glaucoma, open-angle glaucoma, medication adherence, elderly. Accepted for publication February 13, 2007. Printed in the USA. Reproduction in whole or part is not permitted.
I00
June 2007
Volume 5 • Number 2
doi: I 0.1016/j.amjophann,2007.05.004 1543-5946/$32.00
Copyright © 2007 Excerpta Medica, Inc.
M.J. Bhosle et al.
INTRODUCTION
Glaucoma, sometimes referred to as a "silent blinder," is the second leading cause of blindness overall and the leading cause among blacks and Hispanics in the United States. 1-3 Glaucoma is usually asymptomatic until late in its course. As many as 50% of glaucoma patients in developed countries are unaware that they have the disease. 4 The incidence of glaucoma is predicted to be between 1.1% and 3% of the US population, with the incidence increasing to --6% in patients aged _>65 years, s The major risk factors for glaucoma are age _>65 years, family history of glaucoma, certain systemic diseases (eg, hypel"cension, diabetes mellitus, hyperthyroidism), myopia, and black descent. 6 The most commonly occurring fbrm of the disease in western populations is primary open-angle glaucoma, an increase in imraocular pressure (IOP) with resulting damage to the optic nerve, deterioration of visual fields, and eventual loss of vision. Open-angle glaucoma affects >2 million Americans, a finding that is expected to rise to >3 million by 2020 considering the rapid aging of the US population. 7 Treatment fbr glaucoma is usually focused on IOP reduction with medications, as this is believed to prevent or arrest deterioration of the optic nerve and visual fields by reducing aqueous humor production or increasing aqueous outflow. Laser and surgical procedures are mostly reserved for patients who do not respond to drug therapy. Some of the factors that could impact achieving optimal IOP in a patient may include effectiveness of a medication, disease progression, and patient adherence, s,6 In recent years, prostaglandins have emerged as the mainstay of treatment fbr open-angle glaucoma, and latanoprost is fi'equcntly used as primary therapy. .8,9 Latanoprost belongs to the category of prostaglandin analogues and reduces IOP by increasing the uveoscleral flow without affecting the aqueous flow. Latanoprost was introduced in the US market in late 1996 as an antiglaucoma medication and has been widely used as an adjunctive and often primary therapy, possibly due to its QD administration schedule and effective IOP-lowering e f f e c t s . 9 The drug reduces IOP by 30% in patients with open-angle glaucoma and is well tolerated in most patients. 9 In a randomized study of 35 patients with open-angle glaucoma, Polo et al 1° compared the effect on IOP at the end of 2 weeks and 3 months of latanoprost (n = 18) versus timolol plus dorzolamide (n = 17). In this patient cohort, latanoprost (0.005%), administered QD, showed sig*Trademark: Xalatan ® (Pfizer Inc., New York, New York).
The Am.erican Journal of Geriatric Pharmacotherapy
nificantly better long-term reduction (at the end of 3 months of therapy) in IOP than combined therapy with timolol plus dorzolamide (timolol 0.5%, BID; dorzolamide 2%, BID). Similar to latanoprost, other prostaglandin analogues (eg, bimatoprost, travoprost [both approved in 2001]) effectively manage glaucoma by lowering IOP levels, s All these prostaglandins offer a convenient QD dosage schedule. Common adverse effects of these medications include increased iris pigmentation s and hyperemia, predominantly in predisposed eyes. Previous clinical trials indicated hyperemia rates ranging f?om 5% to 15% for latanoprost, 11 15% to 45% for bimatoprost, 12 and 35% to 50% for travoprost. ~3 Latanoprost was the only prostaglandin analogue approved by the US Food and Drug Administration (FDA) as a first-line agent for open-angle glaucoma and ocular hypertension when the present study was conducted ~4 and was generally regarded as a first-line agent by glaucoma spedalists soon after its introduction in 1996. is Bimatoprost was approved by the FDA in 2006 as a first-line therapy fbr elevated IOP-assodatcd openangle glaucoma or ocular hypertension. 16 Medication use behavior, as determined by medicationrelated persistence and adherence among glaucoma patients, is an important issue. Persistence to medication refers to the duration of continuous medication use as prescribed, 17 and adherence refers to the extent to which a person's behavior (eg, following prescribed medication regimen and a diet, and/or executing lifestyle changes) corresponds with advice/recommendations fi'om health care professionals, is Both adherence and persistence to pharmacotherapy arc important in treating chronic conditions, as patients' nonadherence to the treatment regimen may result in poor outcomes, including treatment failure, resistance to the therapy, medication overdose, disease progression, otherwise preventable hospitalizations, and unnecessary medical expenditures. 19 Overall, patients' fhilure to follow the prescribed medication regimen for their chronic condition may cost as much as $300 billion to the US health c a r e s y s t e m s . 19 The lack of overt symptoms early in the stage of disease may tend to decrease the medication-related persistence/ adherence among glaucoma patients. 4 The estimated medication nonadhcrence rate ranged between 23% and 59% fbr glaucoma patients, 2°-22 and the rate was found to be lower (23%) in Medicaid-enrolled patients older than 65 years. 22 In one retrospective study of glaucoma patients enrolled in a large national health care provider obtained through a third-party anonymous source (PharMetrics Inc., Watertown,
101
The American Journal of Geriatric Pharmacotherapy
M.J. Bhosle et al.
Massachusetts), 23 glaucoma pharmacotherapy requiring >2 administrations per day was one of the important predictors of nonadherence. In this population of 1784 patients who had received at least 2 latanoprost prescription refills while on no other glaucoma medication and in whom a second glaucoma medication was then added to their latanoprost regimen, mean (SD) refill intervals fbr latanoprost were 40.6 (21.8) days before the addition of a second drug and 4Z4 (24.4) days after the addition of a second drug. Compared with other therapies, the use of latanoprost (with its QD administration schedule and lower systemic effects) may be associated with improvements in medication use behavior. 24 In a study by Nordstrom et al, 2s patients continuously enrolled for at least 1 year before the initial glaucoma diagnosis date (UnitedHeakhCare) were identified from the Ingenix Research Database between July 1, 1995, and December 31, 2001. In this retrospective claims data analysis, the study cohort consisted of patients who were newly diagnosed with open-angle (diagnosed glaucoma, n = 3623) or strspect (n = 1677) glaucoma and who received at least one dispensing of a topical ocular hypotensivc medication. Prostaglandin analogues had higher medication-related persistence compared with J]-blockcrs (hazard ratios fbr discontinuation, 0.40 [95% CI, 0.35-0.44] fbr diagnosed glaucoma patients and 0.44 [95% CI, 0.37-0.52] fbr patients with suspect glaucoma). However, none of these studies reported medication use behavior associated with the introduction of latanoprost therapy in patients aged >65 years. Improved medication-related persistence and adherence are associated with improved therapy outcomes and lower utilization of health care services, which in turn reduce treatment-related and overall health-related costs in patients with chronic diseases. 2°-23 However, previous research has not tested whether the frequency of dosing and tolerability of therapy, as effects isolated from product or therapy class selection, are related to adherence or persistence. The primary objective of the current study was to compare the medication use behavior (persistence and adherence) and costs impact in Medicare-aged glaucoma patients who were newly treated with latanoprost versus other glaucoma medications. Because of its acceptance as a first-line agent and as the mostprescribed alternative to traditional standard therapies, 8-1° including the [3-blocker class, and newer agents such as dorzolamide/timolol, we chose to compare the use of latanoprost with other ocular hypotensive agents.
102
MATERIALS AND METHODS Study Population The target population induded older adults (aged _>65 years) with primary open-angle glaucoma enrolled in a Medicare health maintenance organization (HMO) in the southeastern United States. This managed care plan was operational in late 1996 and was the sole provider of medical care to its enrollees ("lock-in" risk-benefit plan). On enrollment in the plan, the enrollee was mailed a questionnaire that induded questions about demographics, clinical conditions, health service utilization in the year before joining the plan, lifestyle, depression, and quality of life. Since its inception in 1996, the response rate for this questionnaire has been >80%. 26 A comprehensive range of preventive, diagnostic, and treatment services is provided by plan providers that includes all necessary office visits. Prescription drugs are covered according to a 2- or 3-tiered open formulary. For prescription medications prescribed by a plan physician and obtained at a plan pharmacy, patients pay a $6 generic or $12 name brand copayment per prescription unit or refill. The 12-item Short-Form Health Survey (SF-12) was used to assess general health status. The SF-12 is a self administered questionnaire that evaluates symptoms, flmctioning, and quality of lifE. A higher score indicates better quality of life in patients, and 0 indicates worst possible quality of life (eg, near-death or worse-than-death experience, especially in elderly patients with chronic conditions and comorbidities). Enrollment in the plan was tracked fi'om 1996 until December 2002. During this period, there were a total of 14,654 patients aged _>65 years. The mean (SD) age of this cohort was 74.69 (6.81) years. Eligible patients were required to have a previous medical claim coded for open-angle glaucoma during the study period (International Classification of Diseases, Ninth Revision [ICD-9] code 365.1), and at least one prescription fill for a glaucoma medication during the study period.
Study Design The study had a retrospective observational cohort design. The study compared medication persistence, medication adherence, and health care costs among 2 patient cohorts starting a new glaucoma medication. The study covered the period from January 2000 to December 2002 (36 months). The earliest and specific glaucoma medication dispensed in calendar year 2001 was identified as the index agent. The time frame extended from 1 year before (preindex) to i year after
M.J. Bhosle et al.
(postindcx) the start date (index date) of latanoprost index therapy (comprising the case group) and the same time intervals for the non-latanoprost index medications (collectively comprising the control group). To assure that the patient was new m the index agent, patients in both the case cohort and control cohorts were required to not have a previotts fill for the specific index agent for a period of 1 year before the index date. Patients aged <65 years old and those who had a prescription claim fbr any open-angle glaucoma prescription in the prcindex period were excluded fi'om the study. M e a s u r e m e n t and O u t c o m e s
Prescription refill patterns were used to derive measures of adherence and persistence under the assumption that a prescription filled was a prescription taken. Pharmacy records have been shown to be valid measures of cumulative exposure and gaps in medication supply. 27 We performed random checks on pharmacy claims to compare days of prescription supply against the quantity of medication dispensed to determine if there was a reasonable fit with the figure entered, which we found true in all cases examined. We further confirmed days of supply for each medication based on bottle size using the methodology followed by Nordstrom ctal. 2s The days' supply calculations were based on the fbllowing assumptions: for most nonprostaglandins, up to 5 mL were assumed to provide 60 days of therapy; up to 10 mL were assumed to provide 90 days of therapy; and >10 mL were assumed to provide 120 days of therapy. The drugs other than prostaglandins are used QD (eg, timolol gel, levobunolol 0.5%) or BID (eg, timolol solution, levobunolol 0.25%) depending on the product formulation and the ophthalmologist's/manufacturer's recommended dosing. The prostaglandins are indicated for QD use. Because the nonprostaglandin products prescribed for study patients were generally BID fbrmulations, we adjusted the expected days' formulation for prostaglandins to last longer, though not twice as long, as the nonprostaglandins. Accordingly, for prostaglandins, we assumed that: up to 2.5 mL provided 60 days of therapy, up to 5 mL provided 90 days of therapy, and > 5 mL provided 120 days of therapy. Working Definition of Persistence Rate Persistence to the new therapy was calculated with survival analysis techniques. We obtained the time to discontinuation of initial therapy by finding the number of days from the starting date to the end of the last dispensing (ie, last fill date plus days' supply)
The Am.erican Journal of Geriatric Pharmacotherapy
of the index agent befbre a gap in prescription supply. In the analysis, discontinuation of therapy (discontinuation date) was defined as no further index drug refill 60 days (if dispensed 1 bottle) or 120 days after the last prescription fill (if dispensed >1 bottle). Patients who reached their study end date without discontinuing the initially prescribed therapy were censored as of the end date. The Kaplan-Meier plots showed the time to discontinuation of therapy for the 2 study cohorts (case vs control) as well as fbr latanoprost and the individual medication classes formed by subcategorizing the control group as [8-blockers, combination dorzolamidc/ timolol, and other medication class (Table I). Working Definition of Adherence Rate An index of glaucoma medication utilization for the postindex year was computed for each patient based on the Med-Total approach by Stcincr ct al. 2s The Medication Possession Ratio (MPR) was calculated as the total nmnber of days of glauconla prescription supply dispensed in this period divided by the number of days between the index date and the dispensing date of the last prescription in this period. The number of days a person was in the hospital was subtracted from the denominator because any drug taken during this time may have been dispensed by the hospital; inpatient pharmacy claims were not available for observation. As suggested by Stciner et al, MPRs >1 or <0 were truncated to i and 0, respectively. Health Care Costs The study examined costs associated with total health care utilization that included all medical and pharmacy claims, as opposed m examining glaucomarelated utilization and costs. This was done to avoid problems inherent in attributing all billings for health care service utilization to a specific condition due to miscoding of diagnosis. 29 Patients were fbllowcd up for complete health care service utilization (ie, hospitalizations, emergency dcpamncnt visit, outpatient physician visits, utilization of glaucoma medication) and costs 1 year before and 1 year after the index date. Actual costs paid by the plan m providers for medical, hospital, and pharmacy services were used m estimate total costs for each of these years. Statistical Analysis Descriptive (frequencies and percentages) analyses of baseline characteristics and univariate (paired t tests and Z2) analyses were conducted comparing demographics and study outcome variables across groups.
103
The Am.erican Journal of Geriatric Pharmacotherapy
M . f Bhosle et al.
Table I. Other glaucoma medications used in the study (control group comprised patients who started any therapy other than latanoprost*; case group consisted of patients who started with latanoprost therapy [2.5 mL, a 60 days' supply]).
Medication
No. of Patients Who Started with the Medication
Bottle Size (No. of Patients Who Started on the Bottle Size)
Usual Dosage
Carbonic anhydrase inhibitor Dorzolamide hydrochloride ophthalmic solution Brinzolamide ophthalmic suspension
16 I0 6
5 mL (n - 5), 10 mL (n - 5) 5 mL (n - 4), 10 mL (n - 2)
TID TID
t-Blockers Timolol maleate Timolol maleate (gel-forming solution) Levobunolol hydrochloride ophthalmic solution Metipranolol ophthalmic solution Carteolol hydrochloride
69 39 7 21 I I
5 mL (n = 39) 5 mL (n - 7) 5 mL (n = 9), I0 mL (n = 9), 15 mL (n = 3) I0 mL (n = I) I0 mL (n -- I)
BID QD BID BID BID
Other prostaglandin analogues Bimatoprost ophthalmic solution Travoprost solution
21 II I0
2.5 mL (n - 7), 5 mL (n - 4) 2.5 mL (n - 10)
QD QD
ot-Agonists Apraclonidine solution Brimonidine tartrate ophthalmic solution Propine solution (dipivefrin hydrochloride) Combination dorzolamide/timolol Dorzolamide hydrochloride/timolol maleate ophthalmic solution Unoprostone isopropyl ophthalmic solution
6 I 5 0
5 mL (n - I) 5mL(n- I),10mL(n-3),15mL(n-
I)
TID TID BID
55 55
I
5 mL (n = 12), 10 mL (n = 43)
BID
5mL(n=
BID
I)
*Days' supply assumptionsfor the control gpoup were as follows: up to 5 mL 60 days;up to I0 mL, 90 days;and >10 mL 120 days.
Survival and multiple regressions were used to determine the medication-related persistence, adherence, and cost impact of latanoprost therapy compared with other glaucoma medications. Survival techniques were used to analyze predictors of medication-related persistence. Cox proportional hazards models were constructed fbr latanoprost versus other glaucoma medication groups (model 1) and for individual medication class using [3-blockers as the reference group (model 2). Other variables included in the models were patients' demographics (eg, age, sex). All refills of the index prostaglandin were tracked during the postindex year. Persistence was identified as the duration of continuous treatment with the index medication. 25 However, within the regression model, patients who experienced the introduction of a new ocular hypotensive in the postindex year were classified as such with the dichotomous variable "combination therapy." If a patient had switched f?om the index agent to a new agent, this would have been indicated
104
in the model as a discontinuation of the index agent, rather than a "switch" per se. The fbllowing regression model fbr adherence was used for each of the 2 outcomes: outcome = f (demographic and clinical confounders, treatment group indicator, year indicator, year x group interaction), where outcomes were annual M P R scores and total annual health care costs. The year indicator and year x group interaction variables were included only in the cost regression, as adherence was measured the postindex year only. The year x group interaction represents the relative cost impact of one therapy compared with another in the cost regression. The regression controlled for demographic and clinical confounders, including age, sex, total number of prescription medications during the 2-year period, and propensity for high health care utilization in the preindex year. Because most patients did not have other comorbid conditions listed, we used their health care utilization in the preindex year as a risk stratifier.
M.J. Bhosle et al.
The dependent variable in the cost impact regression was annual health care costs reimbursed by Medicare. The distribution of health care costs was skewed (as determined by the Shapiro-Wilk test3°). To avoid potential estimation problems in the model such as heteroscedasticity, the natural logarithm of cost was used as the adjusted dependent variable for the analyses. Parameter estimates obtained from regressing log-transformed costs on covariates were interpreted by using the antilog of model estimates. 31,32 Multiple linear regressions using the postindex year data were employed to obtain the differences in treatment adherence between the 2 groups fbr both sets of comparisons. This regression adjusted for the same set of confounders described earlier. The dependent variable was adherence to the glaucoma medication regimen in the postindex year. The regressions also checked for the presence of multicollinearity (linear relation between predictor variables) and autocorrdation (correlation between sequential values [in this case, yearly values] of variables). Correlation coefficients (R 2) between the predictor variables were not greater than 0.3, indicating that multicollinearity effects were not significant. Results of the Durbin-Watson test performed on the regressions were <2 fbr all the regressions, indicating that there was no auto correlation problem within the data. All statistical analyses were conducted using the Stata 9.0 software (StataCorp LP, College Station, Texas).
The American Journal of Geriatric Pharmacotherapy
Table II. Baseline characteristics of the study population.
Parameter
Case Group ( n - 100)
Contr-olGroup ( n - 168)
Age, y Mean (SD) Range
77.60 (6.54) 65-98
77.59 (6.72) 65-96
Male sex, %
30
35
59.50 (21.31) 25-I00
62.36 (21.98) 0-I 00
3.02 (2.58) 0-I 2
3.09 (2.36) 0-I 2
Proportion of respondents perceiving worse health Mean (SD) Range
0.037 (0.19) 0-1
0.075 (0.26) 0-1
Total no. of prescription medications* Mean (SD) Range
36.98 (28.41) 0-140
31.75 (24.53) 0-149
0.80
0.75
SF-12 general health scores Mean (SD) Range Charlson's comorbidity severity index score Mean (SD) Range
Proportion on monotherapy
SF-12 = 12-item Short-Form Health Sur~ey. *P < 0.05;meantotal number of prescriptionstaken by patients in caseand control groups reflecLsa mix day supplies.
RESULTS
The study sample represented -0.02% of the elderly population (n = 14,546) in the health care plan. Table II compares the adjusted descriptive characteristics of the 2 study populations. The case group comprised 101 patients, while the control group included 168 patients. One outlier patient in the case group who had total health care costs of $338,000 in the postindex year was dropped f~om this model since such high costs usually represent end-of-life care. There were no significant differences across the 2 groups with respect to age, sex, general health scores on the SF-12, severity of comorbidity, or the proportion of respondents with perception of worsened heakh. A total number of mean prescription medications during the 2-year period were higher for patients in the case group compared with patients in the control group (36.98 vs 31.75; P < 0.05). Bivariate tests indicated no significant differences in the health care costs in both the years between the 2 comparison groups (cases vs controis) (Table III).
The regression analysis for persistence and adherence rates included data from the postindex year only. Starting with latanoprost therapy was associated with a slightly better persistence rate compared with starting with other medications (hazard ratio of discontinuation, 0.90; 95% CI, 0.68-0.98) (Figure 1, Table IV [model 1]). Furthermore, patients' discontinuation rate of initially prescribed medication was lower for latanoprost compared with ]3-blockers (hazard ratio of discontinuation, 0.92; 95% CI, 0.69-0.96) (Figure 2, Table IV [model 2]). Persistence rate to combination dorzohmide/tinlolol was higher compared with the IB-blockers, as indicated by a hazard ratio of discontinuation for combination dorzolamide/timolol compared with 13-blockers of 0.89 (95% CI, 0.84-0.98). It is noteworthy that the Kaplan-Meier curves are flat for the first 60 days because patients could not discontinue their therapy during that time. The covariate "total nmnber of prescription medications" had a significantly lower risk of discontinuation in models 1
105
The Am.erican Journal of Geriatric Pharmacotherapy
M.J. Bhosle et al.
and 2; each additional prescription used by the patient was associated with a 1% and 4% decrease in probability of discontinuation, respectively. Table V displays the results of the comparison of adherence rates and health care costs between the case group and the comrol group using multiple regression analysis. The mean adherence rate for the case group was significantly higher than that of the control group after adjusting fbr other variables ([3 = 0.057; P < 0.01). Introduction of latanoprost therapy was not associated with a significant increase in costs over other glaucoma medications. Sensitivity
Table III. Study outcomes (unadjusted). Case Group (n - I00)
Parameter
Analysis
The appropriateness of the specification of the dependent variable for the cost impact analysis was examined. This was done by observing both logged (natural logarithm of annual health care costs as a dependent variable) and unlogged (annual health care costs as a dependent variable) spedfications of the annual health care costs in the regression model used for comparing total health care costs between the case and control groups. Both variations of the spedfication produced an identical set of significant predictor variables. The direction of effect of the significant variables remained
Control Group (n - 168)
Total health care costs in preindex year, $ Mean (SD) Range
3749.97 (6892.24) 3097.51 (4983.15) 0-44,739.38 0-37,190.37
Total health care costs in postindex year; $ Mean (SD) Range
4718.24 (8982.92) 4046.55 (6505.39) 144.51-73,779.71 0-45,974.35
Persistence rate to new medication therapy (postindex year) Mean (SD) Range
0.75 (0.32) 0-1
0.68 (0.36) 0-1
Adherence rate to new medication therapy (postindex year-)* Mean (SD) Range
0.51 (0.26) 0-1
0.40 (0.25) 0-1
*P < 0.001.
Latanoprost . . . . OGM 1.00 ..........................
4-J
.............
0.75 -
.., |--%
L
'
OO
!
0.50 -
O O O-
,. . . . , "I
o O_
i L--'I
0.25 . . . . . . ,.,. . . . . . . . . . . . . . .
~ ...........................
,o
0-I
I
I
I
I
0
100
200
300
400
Days After Glaucoma Treatment Initiation Figure I. Proportion of patients who remained on the initially prescribed glaucoma medication therapy. Note: KaplanMeier curve adjusted for age, sex, total number of prescription medications, and combination therapy shows time to discontinuation after starting the new therapy for the 2 study cohorts (latanoprost group vs other glaucoma medication [OGM] group).
106
M.J. Bhosle et al.
The American Journal of Geriatric Pharmacotherapy
Table IV. Adjusted Cox proportional models for time to discontinuation of the initially prescribed glaucoma medication therapy.* Values are given as hazard ratio (95% CI). Model I (Latanoprost vs OGMs)
Model 2 (Based on Medication Class)
0.90 (0.68-0.98)t Reference category -
0.92 (0.69-0.96)# Reference category
1.01 (0.68-1.51) 1.00 (0.99-1.00) I. 16 (0.87-1.54)
0.89 (0.84-0.98)t 1.95 (I,37-1.77)# 1.00 (0.67-1.49) 1.00 (0.99-1.00) I. 13 (0.85-1.51)
0.99 (0.97-0,99)*
0.96 (0,88-0.99)#
0.86 (0.64-1.15)
0.88 (0.66-1.18)
Variable Latanoprost OGMs 13-Blockers Combination clorzolam (dorzolam ide/timolol) Other medication class# Age Age squared§ Male sex Total no, of prescription medications Combination therapy (>1 glaucoma medication in postindex group)
OGMs - other glaucoma medications. *Per~sistencewas calculated using only postindex year dat~ tP < 0.05. Other medication classFor model 2 included carbonic anhydr~se inhibitors,other pros~aglandin analogues, 0~-agonis~s,and unoprostone isopr~)pylophthalmic solution. §Age squared = age. age..
. . . . j3-Blockers Latanoprost Combination dorzolamide/timolol .... Other
1.00 -
f~
. . . . . . .
0.75 -
. . . .
I
e°m
O| O c-
0.50 -
. . . . . .
! I
O
O O-
La _ _
~
L-------- ..._1
! . . . .
o O_
i-~_~--~___~__ ~_,,~.-~~
7
I ' !
0.25 -
"-I I__ . . . . . . . . . .
I
- I
_ _
-,
I. . . . . . . .
L
L. L° L. . . . . L.
OI
I
0
100
I 200
I 300
I 400
Days after Glaucoma Treatment Initiation Figure 2. Proportion of patients who remained on the initially prescribed glaucoma medication therapy (by medication class). Note: Kaplan-Meier curve adjusted for age, sex, total number of prescription medications, and combination therapy shows time to discontinuation after starting the new therapy (latanoprost, J3-blockers, combination dorzolamide/timolol, other medications).
107
The American Journal of Geriatric Pharmacotherapy
M.J. Bhosle et al.
Table V. Comparison of adherence rates and total health care costs between case and control groups.*
Dependent Variables (13 [SE-])
Predictor Variables
Adherence Ratet
Age Age squared Male sex
0.05 (0.04)
-0.14 (0.18)
-0.00033 (0.0003)
0.0009 (0.001)
-0.055 (0.04)
0.32 (0. II)
High health care costs (>$4000) in preindex year Latanoprost
Natural Log of Total Health Care Costs
NA
2.21 (0.13)¢
0.057 (0.03)t
0.14 (0.17)
Year
NA
-0.62 (0.15)¢
Latanoprost group-year interaction
NA
0.085 (0.23)
Total no. of prescription medications
--0.003 (0.007)~
Combination therapy (>1 glaucoma medication in the postindex group)
Constant
0.009 (0.002)~
0.21 (0.03)
0.064 (0.16)
-1.90 (1.63)
12.14 (1.68)
0.052
0.25
Adjusted R2
NA = not applicable. *Reference groups (omitted categories) include:control, female sex,costs _~$4000/yearin preindex yea~,patients with no combination therapy,and postindex ye~ t Adherence was calculated usingonly postindex year data. ~:P < 0.01. §P < 0.05.
the same. The logged specification produced more conservative estimates in the magnitude of estimation; for example, in the logged specification model, high health care costs in the prcindcx year were associated with increased annual health care costs (2.21 [0.13]; P < 0.001). The sensitivity analysis using unlogged specification indicated similar direction of effects in the association ($955Z35 [$1055.25]; P < 0.01). DISCUSSION
The results of this study highlight the importance of the introduction of latanoprost therapy in improving medication-related adherence among patients aged _>65 years with glaucoma. In this study, after adjusting for the effects of patient characteristics, patients who started with latanoprost had higher medication adherence rates compared with patients starting with
108
other glaucoma medications. Patients' time to discontinuation of initially prescribed glaucoma medication therapy was lower for the latanoprost group when compared with other glaucoma medications, thus indicating higher persistence for the latanoprost group. Further categorization of the control group indicated that latanoprost and combination dorzolamide/timolol had better persistence rates compared with ]3-blockers. In the regression analysis, a higher number of total prescription medications was associated with a lower medication adherence rate. The negative association between total number of prescriptions and patients' medication adherence has been reported in the literature. 29 Patients in the latanoprost group had a higher number of total prescription medications compared with patients in the control group; their adherence and persistence rates were higher.
M.J. Bhosle et al.
Previous retrospective cohort studies have compared the medication persistence rate in patients (aged <65 years) treated with latanoprost versus 13-blockers, brimonidine, or carbonic anhydrase inhibitors. These studies found that patients treated initially with latanoprost monotherapy remained on therapy significantly (P < 0.001 to P < 0.05) longer than those treated with comparator drugs. As reported by Schwartz et al, 33 in their retrospective cohort study of a managed care database, glaucoma suspect patients treated with timolol monotherapy were 39% (95% CI, 1.21-1.58) more likely to discontinue and 27% (95 % CI, 1.13-1.43) more likely to change therapy (P < 0.001 for both comparisons) compared with latanoprost monotherapy. A study comparing persistency rates across topical prostaglandins in >4000 patients followed up for 15 months fbund that glaucoma patients treated with latanoprost were significantly (P < 0.001) more likely to continue therapy than patients receiving either bimatoprost or travoprost. 34 Wilensky et al 3S compared persistency rates across topical prostaglandins using an algorithm to correct for inconsistent data reported in quantity supplied and days' supply field of database used (IMS Health LifeLink database wifla pharmacy claims). This study reported significantly higher (P _< 0.05) mean number of days adherent to therapy for bimatoprost-treated patients (291.2 days) compared with latanoprost-treated patients (281.0 days) but not significantly different from travoprost-treated patients (287.0 days). In the current study, introduction of latanoprost therapy was not associated with a significant increase in total health care costs compared with other glaucoma medications. These results are consistent with the previous pilot study (N = 78) conducted by Balkrishnan et a136 in glaucoma patients. This study reported no significant additional increases in total health care charges in the study population associated with the introduction of latanoprost therapy, after adjusting for group and time effects, as well as other baseline confounders. The current study, however, should be considered an initial exploration of these issues, and caution is required when interpreting these findings, due to a number of study limitations. First, the observational study design does not permit causal inference of results. In spite of confounder adjustment, the study results may be subject to some issues of treatment group selection. Also, the measure of adherence depended on pharmacy records. Even though this method has been shown to be reliable, 27 it cannot guard against
The Am.erican Journal of Geriatric Pharmacotherapy
instances of undetected adherence, such as the case when a hospitalized patient is given a generous supply of medication at time of discharge, and no record of the medication exists in the patient's pharmacy data. As our previous research with topical medications has highlighted the measurement issues associated with assessing retrospective medication use with retrospective data, we have used mukiple measures (both persistence and adherence), which have been used previously to describe medication use behavior. 37,38 Additionally, we continued the days' supply recorded in the data set by calculating days of medication supply based on bottle size of each glaucoma medication. Our algorithm fbr estimating persistence on medications other than prostaglandins assumed that most such prescriptions were dosed BID. There was a greater incidence of BID formulations of nonprostaglandins in the actual number of prescriptions dispensed. Due to the reliance of this study on a claims database, we could not observe how often ophthalmologists actually directed the patient to dose these agents. If QD dosing for nonprostaglandins was more common than assumed, our findings could have understated nonprostaglandin persistence. Furthermore, our measures of adherence, while accounting for many patient medicationtaking patterns, still do not completely account fbr the impact of product switching and combining, as well as greater number of people being on higher doses in one group. We also relied on the electronically recorded ICD-9 codes and prescriptions to assign our cases and controls, making the assumption that the coding and recording of these was done accurately. Less than 5% of data had missing values, and therefore we did not use any specific techniques for handling missing data. Finally, there may have been selection bias because we were unable to assign patients randomly to the case and control groups. This study included only those patients who did not have a claim tbr any glaucoma medication 12 months befbre the index prescription date. Medication use behavior among these medicationnaive patients may differ from those who have experience with the glaucoma medication therapy. Rates of adverse effects vary across different medications used for glaucoma, which may affect patients' medication use behavior. Our study did not capture differences in adverse-effect profile for different glaucoma medications studied as these data were not available. Furthermore, results do not show the clinical impact of glaucoma medications. However, it is ideal to choose a drug that patients are more adherent to and persistent with; that is, the more adherent they are, the more
109
The Am.ericanJournal of Geriatric Pharmacotherapy M.J.Bhosleet al.
likely the drug is having an impact on patients' IOP compared with not taking it at all. Future studies need to focus on larger sample sizes to make these results more generalizable. These studies would have to be conducted in larger MedicareH M O systems to obtain a larger number of patients in each treatment group category. Additional studies are also needed with longer periods of follow-up to estimate costs associated with prolonged use of latanoprost therapy compared with other treatment options for glaucoma. CONCLUSIONS After adjusting fbr potential confbunders, we found higher medication persistence and adherence rates in glaucoma patients who started with latanoprost therapy as opposed to patients starting treatment with other glaucoma medication. There were no significant additional increases in overall health care costs among patient groups as a result of introduction oflatanoprost therapy, after adjusting for group and time effects, as well as other baseline confbunders. ACKNOWLEDGMENTS This study was funded by a grant from Pfizer Inc., New York, New York. Dr. Reardon is a consultant for Pfizer Inc. REFERENCES 1. Coleman AL. Glaucoma. Lance. 1999;354:1803-1810. 2. Sommer A, Tielsch JM, Katz J, et al. Racial differences in the cause-specific prevalence of blindness in east Baltimore. N E t ~ l J M e d . 1991;325:1412-1417. 3. Rodriguez J, Sanchez R, Munoz B, et al. Causes of blindness and visual impairment in a population-based sample of US Hispanics. Ophthahnology. 2002;109:737-743. 4. Quigley HA, Vitale S. Models of open-angle glaucoma prevalence and incidence in the United States. Invest Ophthahnol Vis Sci. 1997;38:83-91. 5. Infeld DA, O'Shea JG. Glaucoma: Diagnosis and management. Postgrad Med J. 1998 ;74:709-715. 6. Gordon MO, Kass MA. The Ocular Hypertension Treatment Study: Design and baseline description of the participants. Arch Ophthalmol. 1999;117:573-583. 7. Friedman DS, Wolfb RC, O'Colmain BJ, et al, for the Eye Diseases Prevalence Research Group. Prevalence of open-angle glaucoma among adults in the United States. Arch Ophthalmol. 2004;122:532-538. 8. Marquis RE, Whitson IT. Management of glaucoma: Focus on pharmacological therapy. Drugs Aging. 2005; 22:1-21.
I10
9. Fiscella RG. Costs of glaucoma medications. AmJHealth Syst Pharm. 1998 ;55:272-275. 10. Polo V, Larrosa JM, Ferreras A, Honrubia FM. Latanoprost vs combined therapy with timolol plus dorzolamide in open-angle glaucoma: A 24-month study. Ann Ophthahnol. 2005;37:33-36. 11. Xalatan [package insert]. Peapack, NJ: Pharmacia Corporation; 2000. Available at: http://www.fda.gov/ cder/foi/label/2000/2059 7s141bl.pd£ Accessed April 12, 2007. 12. Lumigan [package insert]. Irvine, Calif: Allergan, lnc; 2001. Available at: http://www.fda.gov/cder/foi/label/ 2001/212751bl.pd£ Accessed April 12, 2007. 13. Travatan [package insert]. Fort Worth, Tex: Alcon Laboratories, Inc; 2001. Available at: http://www.fda. gov/cder/foi/label/2001/212571bl.pd£ Accessed April 12, 2007. 14. Xalatan [package insert]. New York, NY: Pfizer Inc.; 2003. Available at: http://www.f~la.gov/cder/foi/ label/2001/20597s231bl.pd£ Accessed April 12, 2007. 15. Xalatan marks 10th amfiversary. Prostaglandin glaucoma medication overcame obstacles. Available at: h t t p : / / www.ophmanagement.com/article.aspx? article- 8664 6. Accessed December 17, 2006. 16. Waknine Y. FDA approvals: Lumigan, cesamet, onmitrope. Medscape Medical News 2006. Available at: h t t p : / / www.medscape.com/viewarticle/54 0937. Accessed December 20, 2006. 17. DiMatteo MR, Giordani PJ, Lepper HS, Croghan TW. Patient adherence and medical treatment outcomes: A meta-analysis. Med Care. 2002 ;40:794-811. 18. Haynes RB, McDonald H, Garg AX, Montague P. Interventions fbr helping patients to fbllow prescriptions for medications. Cochrane Database Syst Rev. 2002;2: CD000011. 19. DiMatteo MR. Variations in patients' adherence to medical reconmlendations: A quantitative review of 50 years of research. Med Care. 2004;42:200-209. 20. Rotchfbrd AP, Murphy KM. Compliance with timolol treatment in glaucoma. Eye. 1998 ;12:234-236. 21. Patel SC, Spaeth GL. Compliance in patients prescribed eyedrops for glaucoma. Ophthalmic Surg. 1995 ;26:233236. 22. Gurwitz JH, Glyml RJ, Monane M, et al. Treatment fbr glaucoma: Adherence by the elderly. AmJPublic Health. 1993;83:711-716. 23. Robin AL, Covert D. Does adjunctive glaucoma therapy affect adherence to the initial primary therapy? Ophthalmology. 2005 ;112:863-868. 24. Shaya FT, Mullins CD, Wong W, Cho J. Discontinuation rates of topical glaucoma medications in a managed care
M.J. Bhosleet al.
population. A m J Manag Care. 2002;8(Suppl 10): $271-$277. 25. Nordstrom BL, Friedman DS, Mozaffari E, et al. Persistence and adherence with topical glaucoma therapy. A m J Ophthalmol. 2005 ;140:598-606. 26. Balkrishnan R, Anderson RT. Predictive power of a riskassessment questiom~aire across different disease states: Results in an elderly managed care enrolled population. A m J M a n a g Care. 2001;7:145-153. 27. Choo PW, Rand CS, Inui TS, et al. Validation of patient reports, automated pharmacy records, and pill counts with electronic monitoring of adherence to antihypertensive therapy. Med Care. 1999;37:846-857. 28. Steiner JF, Koepsell TD, Fihn SD, Inui TS. A general method of compliance assessment using centralized pharmacy records. Description and validation. Med Care. 1988;26:814-823. 29. Balkrishnan R, Rajagopalan R, Camacho FT., et al. Predictors of medication adherence and associated health care costs in an older population with type 2 diabetes mellitus: A longitudinal cohort study. Clin Ther. 2003 ;25:2958-2971. 30. Shapiro SS, Wilk MB. An analysis of variance test for normality. Biometrika. 1965 ;52:591-611. 31. Kennedy PE. Estimation with correctly interpreted dummy variables in semilogarithmic equations. A m Econ Rev. 1981;71:801.
The Am.ericanJournal of Geriatric Pharmacotherapy
32. Maiming WG, Mullahy J. Estimating log models: To transform or not to transfbrm? J Health Econ. 2001;20: 461-494. 33. Schwartz GF, Reardon G, Mozafthri E. Persistency with latanoprost or timolol in primary open-angle glaucoma suspects. A m J Ophthalmol. 2004;137(Suppl 1): $13-$16. 34. Reardon G, Schwartz GF, Mozafthri E. Patient persistency with ocular prostaglandin therapy: A population-based, retrospective study. Clin Ther. 2003 ;25:1172-1185. 35. Wilensky J, Fiscella RG, Carlson AM, et al. Measurement of persistence and adherence to regimens of IOP-lowering glaucoma medications using pharmacy claims data. A m J Ophthalmol. 2006 ;141 (Suppl 1):$28-$33. 36. BalkrishnanR, Bond JB, ByerlyWG, etal. Compliance and costs associated with latanoprost therapy. Introduction in a managed care setting. Presented at: Aimual Meeting of American Academy of Ophthalmology; October 2225, 2000; Dallas, Texas. 37. Balkrishnan R, Bond JB, Byerly WG, et al. Medicationrelated predictors of health-related quality of life in glaucoma patients enrolled in a Medicare heakh maintenance organization. A m J Geriatr Pharmacother. 2003; 1:75-81. 38. Balkrishnan R, Christensen DB. A comparison of medication adherence indices to assess long-term inhaled corticosteroid medication use. JAsthma. 2001;38:91-98.
Address correspondence to: Rajesh Balkrishnan, PhD, The Ohio State University College of Pharmacy and School of Public Health, 500 W. 12th Avenue, Columbus, O H 43210. E-mail: [email protected]
III