Drug Persistency Patterns for Patients Treated With Rivastigmine or Donepezil in Usual Care Settings

OBJECTIVES: To compare levels of persistency with 2 cholinesterase (ChE) inhibitors rivastigmine and donepezil for the treatment of Alzheimers' disease (AD) through the use of administrative claims data. METHODS: This retrospective cohort study identified treatment-naive, community-based AD patients having an initial prescription (index event) for rivastigmine or donepezil between June and December 2000, in the United States, from pharmacy claims in a proprietary administrative claims database. Patients were excluded if they received either drug during the 180 days prior to their index prescription or if they did not have continuous plan enrollment during this period and for at least 90 days following the index date. The probability of treatment discontinuation within the first 60 days of treatment was estimated. Time to treatment discontinuation was analyzed for the cohort of patients that remained on therapy for 60 days as well as for subgroups of the cohort reaching either approved or maximum recommended doses of donepezil or rivastigmine. Treatment discontinuation was defined as either a stop of therapy (no prescription refill within 60 days of estimated completion of prior prescription) or a switch to an alternative AD drug. Kaplan-Meier survival and proportional hazard model analyses were performed. Proportion of days covered (PDC) by an AD therapy was also evaluated in each quarter during the first year of follow-up. RESULTS: Of the newly treated AD study population, 30.4% (171/563) of rivastigmine patients and 31.2% (583/1,871) of donepezil patients discontinued treatment within 60 days of starting therapy (P=0.72). For the cohort of patients that remained on therapy 60 days, the mean time to treatment discontinuation was 331 days (95% confidence interval [CI], 307-355) for rivastigmine patients (n=392) versus 337 days (95% CI, 322-352) for donepezil patients (n=1288). The proportion of patients with a PDC e80% after 12 months of follow-up was 23% for the donepezil group and 19% for the rivastigmine group (P=0.34). For the cohort subgroup that reached an approved dose, the mean time to treatment discontinuation was 346 days (95% CI, 318-374) for rivastigmine patients (n=282) versus 338 days (95% CI, 323-353) for donepezil patients (n=1,283). For the cohort subgroup that reached the maximum recommended dose, the mean time to treatment discontinuation was 396 days (95% CI, 343-449) for rivastigmine patients (n=61) versus 364 days (95% CI, 344-384) for donepezil patients (n=712). CONCLUSIONS: Newly treated AD patients in a usual care setting who initiate therapy with either rivastigmine or donepezil have similar levels of persistency with treatment.

Drug Persi stency Pa tterns for Pa t ients Tr ea ted With Riva st i g mine or Donep e z il in Usual Care Sett ing s 26 weeks of treatment with rivastigmine [dose ranging from 1 mg to 6 mg twice a day (BID)] versus ADAS-Cog mean change from baseline of -2.3 ± 0.62 for patients treated with 6-12 mg per day rivastigmine for all 52 weeks). 1 3 Persistence with AD treatment is also critical from an economic standpoint. Previous economic evaluations of ChE inhibitors in the treatment of AD demonstrate the need for long-term, efficacious treatment to achieve the full economic and patient benefits of AD therapy. In most patients, natural p ro g ression of the disease is slow and does not re q u i re significant use of health care resources in early stages. 1 4 , 1 5 One of the major potential economic benefits is a delay in placement of AD patients in nursing homes. An observational follow-up study of the patients in 1 of 3 randomized donepezil clinical trials found that when donepezil was taken at an effective dose (5 mg per day or more) for at least 9 to 12 months, time to first dementiarelated nursing home placement and time to permanent nursing home placement were longer than in the group with minimal use (defined as less than 5 mg per day). 16 However several researchers questioned the findings from this study, including K a r l a w i s h , 1 7 and the AD2000 collaborative group study found no statistically significant delay in nursing home placement with donepezil. 8 There is limited information regarding persistence with AD therapy outside the clinical trial setting; existing studies have focused solely on donepezil and have included small sample sizes. In a retrospective review of pharmacy claims data for 59 AD patients receiving treatment in a usual care setting (i.e., not in a randomized control trial), Roe et al. 18 found that the probability of a new user continuing donepezil at 90 days was 79.7% ± 10.3% and at 180 days was 62.7% ± 12.4%. Approximately 14% of those who continued therapy for at least 180 days showed gaps in treatment of 6 weeks or more. 1 8 The purpose of the current study was to assess drug persistence p a t t e rns for patients treated with either rivastigmine or donepezil in usual care settings, where conditions typically differ considerably from those in controlled clinical trials. There were 2 opposing reasons why a difference in persistence might be expected between rivastigmine and donepezil: the side-effect profile of rivastigmine compared with donepezil might result in lower persistence, and its dual mechanism of action might lead to better efficacy over the long term and thus better persistence.

■ ■ Methods Cohort Construction
This was a retrospective cohort study utilizing longitudinal, integrated medical and pharmacy claims data from the PharMetrics Anonymous Patient-Centric Database. The PharMetrics database includes patient-level medical and pharm a c e u t i c a l claims histories of more than 27 million managed care patients belonging to 60 national and regional health plans in the United States. This database is representative of the national commerc i a l l y insured population on a variety of demographic measures, including g e o g r a p h y, age, gender, and pro d u c t type. This database also includes elderly patients (aged 65 years and older) who are enrolled in Medicare managed care plans (e.g., Medicare + Choice, now known as Medicare Advantage) that provide full medical benefits, including prescription drugs.
The database covered the period January 1, 2000, through December 31, 2001. The study population was selected for observation based on evidence of pharmacologic treatment for AD. Included were treatment-naive AD patients aged >40 years who received their index (i.e., initial) prescription for either rivastigmine or donepezil during the 6-month period starting on or after June 1, 2000, when both products were available and ending on or before December 31, 2000. Treatment-naive AD patients were defined as those who had not received a prescription for any AD medication in the 6 months prior to their index AD prescription. Newly treated AD patients were excluded if they did not have continuous enrollment in their health plan during the 6-month preindex period and for at least 90 days following the index date,. Newly treated AD patients were followed until either disenrollment from their health plan or the end of the study period (i.e., December 31, 2001). Survival analysis techniques were used to adjust for any differences in follow-up between the 2 treatments.
Per the product information label, the dosages of donepezil shown to be effective in controlled clinical trials are 5 mg or 10 mg daily (QD), and treatment with 10 mg should not be contemplated until patients have been on a daily dose of 5 mg for 4 to 6 weeks. For rivastigmine, the dosages shown to be effective in controlled clinical trials are 3 mg to 6 mg BID. The starting dose is 1.5 mg BID. If this dose is well tolerated after a minimum of 2 weeks of treatment, the dose may be increased to 3 mg BID. Subsequent increases to 4.5 mg BID or 6 mg BID should be attempted after a minimum of 2 weeks at the pre v i o u s dose. Given these titration schedules, patients taking these drugs should reach their maintenance dose within 60 days following the initiation of therapy. Thus the study evaluated the distribution of dosing level for patients at 60 days following the initiation of AD therapy. The study also evaluated the maximum dose level achieved at any point during the study period.

Study Definitions
Two approaches were utilized to determine persistence with AD treatment. With the first approach, patients were classified as remaining persistent with their index AD medication as long as they did not stop their index AD therapy or switch to a diff e re n t AD medication. A stop of AD treatment was defined as a patient not receiving a refill for the index AD medication within 60 days after exhausting the drug supply from the prior prescription. 19 The supply extension period of 60 days was selected based on the typical 30-day supply for rivastigmine and donepezil prescriptions, and assuming a 33% adherence Drug Persi stency Pa tterns for Pa t ients Tr ea ted With Riva st i g mine or Donep e z il in Usual Care Sett ing s rate (30 days/33% = 90 days total [30 days supplied + 60 additional d a y s ] ) . 1 9 As recommended by Dezii, 1 9 the 60-day supply extension was also used for AD prescriptions, with days supply exceeding 30 days (< 5% of all AD prescriptions had days supplied amounts > 45 days, including 90-day mail-order prescriptions). Switching was defined as a patient, at any time following the index date, filling a prescription for any AD medication other than the index medication, regardless of continued index drug use. "Time to treatment discontinuation" (i.e., the rate at which patients discontinue therapy over time), was calculated as the number of days from the index pre s c r i p t i o n fill date to either (1) the date of index medication supply exhaustion preceding a stop of treatment or (2) the date of the fill of a nonindex AD medication (i.e., switch), whichever occurred first.
Some patients may intend to persist with therapy and fill a prescription but end up not using any of the medication. Thus, a sensitivity analysis was conducted based on a more conserv a t i v e definition of a stop of treatment. The stop date was changed from "date of end of days supply of last prescription + supply extension days" to "fill date of last prescription." Because the reasons for switching or stopping treatment might be diff e rent in the early period of ChE-inhibitor treatment (e.g., side effects) compared with the later treatment period (e.g., lack of efficacy), the time to treatment discontinuation approach included 2 analytic components. First, the study estimated the probability of treatment discontinuation in the first 60 days for the entire study population, by index treatment g roup. Second, the study evaluated time to treatment discontinuation for the cohort of patients (designated the "study cohort") that remained on its index AD medication for at least 60 days. This was considered a reasonable minimum period of time under usual care conditions to observe either (1) titration or (2) dropout because of intolerance to side effects.
For the purpose of subgroup analyses, the study cohort was also classified based on whether or not patients reached the following dosing levels (per product prescribing information) for their index AD medication prior to either treatment discontinuation or the end of the study period: 1. approved dose (6-12 mg per day of rivastigmine or 5-10 mg per day of donepezil) 2. maximum recommended dose (12 mg per day rivastigmine or 10 mg per day donepezil). The dosing level was determined from the pharmacy claims data by first taking the strength of dose into consideration (e.g., a 1.5 mg pill versus a 6 mg pill) and then dividing the quantity of drug dispensed by the days of drug supplied to determine the daily dose.
The second approach used to determine treatment persistence was the pro p o rtion of days covered (PDC). 2 0 The number of days supplied from each filled prescription was used to calculate the proportion of days on which a patient had an AD medication available in a given time interval. This analysis was conducted on the entire study population. Per Benner et al. (2002), 20 the cohort was divided into 3 groups: adherent patients were defined as those with a PDC ≥ 80%; partially adherent patients were those having a PDC of 20% to 79%; and patients with a PDC <20% were considered n o n a d h e re n t . N o n a d h e rence or part i a l adherence in a given time interval were considered as suboptimal p e r s i s t e n c e. The PDC was evaluated at 3-month interv a l s (1-3, 4-6, 7-9, 10-12 months) following the index date for patients who had complete follow-up data at each time point.

Statistical Analysis
All statistical and descriptive analyses were performed using SAS version 8.02 (Cary, North Carolina). Descriptive statistics, including means (±SD) for continuous data and relative fre q u e n c i e s for categorical data, were compared using t tests and chi-square d tests, respectively. Survival analysis methods were used to evaluate time to treatment discontinuation. Patients were followed until the first occurrence of one of the following events: discontinuation of index therapy, switch to diff e rent AD medication, disenrollment from health plan (censored), or end of study period (censore d ) . 2 1 The primary reason to use survival analysis to measure drug persistence is that the analysis does not require follow-up to be identical for all patients. Because of potential selection bias associated with a re t rospective design, Cox p ro p o rtional hazards re g ression models were used to evaluate the risk of treatment discontinuation after adjusting for patients' baseline characteristics, including patient age (at index date) and gender, preindex period utilization (i.e., hospitalizations, office visits, prescription drugs), and level of comorbidity in the preindex period (including the use of a modified version of the Charlson Comorbidity Index 22 developed by re s e a rchers in a D a rtmouth-Manitoba collaboration 2 3 ). The cumulative pro b a b i l i t i e s of continuing therapy were plotted as a function of time using the Kaplan-Meier product-limit method (which results in graphical survival curves depicting the rate of treatment discontinuation over time). The a priori threshold level for statistical significance was P ≤ 0.05.

■ ■ Results
The study population comprised 2,434 AD patients newly tre a t e d with either rivastigmine (n = 563) or donepezil (n = 1,871) who met all study eligibility criteria. Of this population, 30.4% (171/563) of rivastigmine patients and 31.2% (583/1,871) of donepezil patients discontinued treatment within 60 days of the index date (P = 0.72) ( Figure 1). There were no statistically significant differences in demographic characteristics (i.e., age, gender, comorbidity) or preindex period utilization between patients who discontinued treatment and those who continued treatment after 60 days.
The study cohort comprised the remaining patients who did not discontinue treatment within 60 days ( Figure 1, Table 1). T h e re was a statistically significant (P<0.001) diff e re n c e between the 2 index treatment groups in terms of mean follow-up time after the index date (due to donepezil being first to market in the United States), but this difference did not impact the analysis since survival analysis methods were used to account for differential follow-up and the similarity in the range of follow-up times between the 2 groups.
The study cohort used a considerable amount of health care services in the preindex period (e.g., >15 office visits and 30 prescription fills in a 6-month period) ( Table 1). Levels of p reindex re s o u rce utilization and comorbidity were comparable between the 2 treatment groups (Table 1). Table 2 presents the distribution of the study cohort' s daily dosing level at 60 days, and the maximum dose achieved at any point in the study period, following the initiation of AD therapy (based on daily dose calculation for the prescription refill with fill date closest to, but not sooner than, days 60 and 90 postindex date). Since the minimum dose of donepezil available on the market is 5 mg, it is not surprising that by day 60, 98% of donepezil-treated patients were taking doses shown to be effective in controlled clinical trials. It should be noted that both available strengths of donepezil, the 5 mg and 10 mg tablets, are considered to be clinically effective doses when taken once daily; however 2% of donepezil patients had prescription claims for the 5 mg tablet but with twice the number of days supplied as quantity dispensed. Therefore, based on our calculation for daily dose, these patients were not classified as receiving a minimum daily dose of 5 mg of donepezil.
Approximately one third of the rivastigmine-treated patients who had persisted with therapy for at least 60 days were not receiving a dose shown to be effective in controlled clinical trials    . Adjusting for age, gend e r, p reindex utilization, and concomitant disease did not signific a n t l y change the hazard ratio. The mean times to treatment discontinuation are provided in Table 3. The unadjusted Kaplan-Meier curves for time to treatment discontinuation are shown in Figure 2. The curves are flat for the first 60 days because, by definition, the study cohort could not discontinue or switch t reatment during that time. The sensitivity analysis (i.e., changing the calculation for date of treatment discontinuation) did not significantly change the results. The mean time to treatment discontinuation was 318 days (95% CI, 293-343) for rivastigmine patients as compared with 315 days (95% CI, 301-329) for donepezil patients.

Distribution of Daily Dose for Study Population at 60 Days after Initiation of Therapy, and Maximum Dose at Any Point in Study Period
Analyses were repeated for the subgroup of patients in each index treatment group that reached an approved dose during the study period (282/392 rivastigmine patients; 1,283/1,288 donepezil patients). Similar to the results for the full study c o h o rt, the likelihood of treatment discontinuation for rivastigmine patients reaching an approved dose was not statistically significantly different from that of the corresponding donepezil subgroup (RR=0.89; P=0.24; 95% CI, 0.73-1.08). Adjusting for age, gender, preindex utilization, and concomitant disease did not significantly change the hazard ratio. The mean times to treatment discontinuation for this subgroup are provided in Table 3. The unadjusted Kaplan-Meier curves for time to t reatment discontinuation for this subgroup are shown in Figure 3.
Analyses were also repeated for the subgroup of patients in each index treatment group who reached the maximum recommended dose during the study period (61/392 [15.6%] rivastigmine patients; 712/1,288 [55.3%] donepezil patients). T h e re was a trend toward a reduced likelihood of treatment discontinuation for this subgroup of rivastigmine patients (RR = 0.68, 95% CI 0.43 -1.06) relative to the corre s p o n d i n g donepezil subgroup, but the diff e rence was not statistically significant (P=0.09). Adjusting for age, gender, preindex utilization, and concomitant disease did not significantly change the hazard ratio. The mean times to treatment discontinuation for this   Table 3. The unadjusted Kaplan-Meier c u rves for time to treatment discontinuation for this subgroup are shown in Figure 4. The results of the PDC analysis (Table 4) indicated that the p ro p o rtion of patients persisting with AD therapy (i.e., maintaining a PDC ≥ 80%) declined over time. In the first 3-month interval following the index date, a significantly higher percentage (P< 0.01) of donepezil patients (44%) versus rivastigmine patients (35%) maintained a PDC ≥ 80%. However, there were no statistically significant diff e rences for the 3-6 month, 7-9 month, and 10-12 month intervals following the index date. The mean PDC at 12 months was 33% in both treatment groups, and the proportion of patients with a PDC ≥ 80% after 12 months of follow-up was 23% for the donepezil group and 19% for the rivastigmine group (P = 0.34).

■ ■ Discussion
Treatment with ChE inhibitors has been proven to slow the decline in cognitive and functional abilities for patients with AD, [5][6][7][8] and persistence with therapy is critical for long-term effectiveness. Reasons for lack of persistence with ChE therapy are numerous. Patients or physicians may perceive that the therapy is no longer effective, patients may become intolerant of side effects, or the dosing system may be inconvenient. The p resent study assessed persistence with patients' initial AD therapy by evaluating longitudinal, patient-level prescription claims data from real-world practice settings. The results indicated that levels of persistence with treatment were similar for newly tre a t e d AD patients receiving either rivastigmine or donepezil in a real-world setting.
These persistence results from real-world data are in contrast to those from a recent randomized open-label study that dire c t l y c o m p a red treatment outcomes and persistency after 12 weeks for patients receiving either rivastigmine or donepezil. 24 The results of this short-term trial indicated that both groups showed comparable cognitive improvements, but more donepezil-tre a t e d patients (89.3%) completed the study than rivastigmine-treated patients (69.1%). 2 4 Given the diff e rences in study design between the published clinical trial by Wilkinson et al. and the current study, it is difficult to compare the persistence findings.
The short -t e rm trial re s u l t s 2 4 also indicated that fewer r i v a s t i g m i n e -t reated patients (60%) than donepezil-tre a t e d patients (98.2%) reached the maximum recommended dose at some point in the 12-week study. In comparison, the real-world results from the current study indicated that only 11% of r i v a s t i g m i n e -t reated patients and 55% of d o n e p e z i l -t re a t e d patients received the maximum re c o m m e n d e d dose at some point in the 12-month follow-up period.
Delays in cognitive decline associated with persistence with AD therapy may have important personal, social, and economic implications. For example, a recent study has demonstrated that when donepezil was taken for at least 9 to 12 months,  Drug Persi stency Pa tterns for Pa t ients Tr ea ted With Riva st i g mine or Donep e z il in Usual Care Sett ing s delays to nursing home placement were longer than when it was taken for a shorter time period. 16 In our study, similar rates of persistence were shown for both rivastigmine (42%) and donepezil (35%) at 9 months. A database study has shown that, in a large managed Medicare plan, AD patients receiving donepezil for ≥ 270 days had health care costs that were $4,921 lower than control AD patients not being treated, while costs for those receiving donepezil for less than 270 days were only $3,579 lower than untreated control AD patients. 25 It should be noted that a recently published study (AD2000) calls into question the long-term (2-year) clinical benefit of donepezil. 8 Over a 3-year period, the study enrolled patients who were referred to memory clinics with a diagnosis of AD and were not already taking a ChE inhibitor. After a 12-week pretreatment period, 486 patients were randomized to receive 5 mg per day of donepezil, 10 mg of donepezil, or placebo. The results indicated that the groups taking donepezil showed minor improvements in cognitive and functional abilities, which persisted over 2 years. There were no statistically significant d i ff e rences in the time to nursing home placement or pro g re s s i o n to disability among the 3 treatment groups. In addition, there were no significant differences between donepezil and placebo for behavioral and psychological changes, psychological well-being of the primary care g i v e r, and death from AD. The longt e rm effectiveness data for rivastigmine are not yet available.
The division of the patient population into those who stop treatment within 60 days and those who continue for longer allows us to estimate the impact of acute side effects separately from the impact of long-term side effects and efficacy on persistence with therapy. These estimates have the potential to give a more accurate picture of patient persistence than a single measure.
The 2 methodological approaches utilized in the study to examine treatment persistence provided dissimilar re s u l t s . Based on time to treatment discontinuation, approximately 50% of AD patients were persistent with AD therapy after 9 months of treatment. In contrast, the proportion of patients deemed persistent after 9 months of treatment based on the PDC approach was 32%. The difference may be due to the less restrictive definition of persistence used in the time to treatment discontinuation approach (i.e., allows for a supply extension of 60 days). The advantage of the PDC is its ease of calculation and explanation. The advantage of the survival analysis approach is that it can be used to determine the percentage of patients who a re persistent on each day of follow-up, thereby revealing subtle trends.
The persistence findings from this study are based on AD prescription drug fills that were recorded in an administrative claims database. There are 2 principal benefits associated with the use of this type of data source. First, claims databases offer access to large populations of patients that are typical of routine clinical practice, including patients that might be excluded from controlled clinical trials. Second, these data permit the conduct of longer-term analyses than can generally be conducted with clinical trial data.

Limitations
It is important to recognize some study limitations related to the use of claims data. First, the prescription refill rates estimated from the claims data serve as a proxy for persistence with therapy; we cannot be sure that patients actually took the medications that they received. However, studies in other disease areas have shown that pharmacy dispensing records correlate well with patient drug exposure. 26,27 Second, claims data contain no direct clinical information, especially as related to treatment side effects and the study cohort' s level of AD disease severity. It is possible that treatment side effects may explain, in part, why some rivastigmine patients were not receiving a dose shown to be effective in controlled clinical trials (≥ 6 mg daily). However, other factors could also be involved, including the dosage instructions given by physicians to AD patients. A concern about the impact of varying disease severity across treatment groups was offset to some extent in the study by the inclusion of only "newly tre a t e d " AD patients.
Third, it is possible that the prescription records contained