CLINICAL THERAPEUTICs®/VoL. 26, No. 12, 2004 Adherence to a Fixed-Dose Combination of Rosiglitazone Plaleate/Pletformin Hydrochloride in Subjects with Type 2 Diabetes Plellitus: A Retrospective Database Analysis Daniel R. Vanderpoel, PharmD, 1 Mohamed A. Hussein, MSCS, MSPH, 1 Teresa Watson-Heidari, PharmD, 1 and Andrew Perry, MA Hons, MSc 2 ~Outcomes Research Center, Humana Inc., Louisville, Kentuchy, and eGlaxoSmithKline, Philadelphia, Pennsylvania ABSTRACT Background: In 2002, fixed-dose combination therapy (FDCT) with rosiglitazone maleate plus metformin hydrochloride became available for the treatment of type 2 diabetes mellitus (DM-2) in subjects whose disease was uncontrolled on monotherapy with metformin or a thiazolidinedione. FDCT allows a reduced pill burden and a less complex medication regimen. Objective: The objective of this study was to assess changes in medication adherence rates associated with oral hypoglycemic agents in subjects switching from either monotherapy or dual therapy with metformin and/or rosiglitazone to rosiglitazone-metformin FDCT. Methods: In this retrospective database analysis, data were obtained from the pharmacy claims database of a large health benefits company. Prescription claims for subjects aged ___18 years with DM-2 whose disease was uncontrolled on monotherapy with metformin or a thiazolidinedione were analyzed over a 12-month study peri- od (a 6-month preindex period and a 6-month postindex period). Some subjects were receiving monotherapy with either metformin or rosiglitazone during the preindex period and remained on monotherapy throughout the postindex period (Mono/Mono cohort), switched to dual therapy with both agents (Mono/Dual cohort), or switched to FDCT (Mono/FDCT cohort). Some subjects were receiving dual therapy with metformin and rosig- litazone during the preindex period and remained on dual therapy throughout the postindex period (Dual/Dual cohort) or switched to FDCT (Dual/FDCT cohort). A medication possession ratio (MPR)--a proxy measurement of medication adherence--was calculated for each subject for each period. Changes in medication adherence were compared using a general linear model. Results: Overall, data from the records of 16,928 subjects (8499 men, 8429 women; mean [SD] age, 58.12 [11.97] years) were included in this study. There was significantly less reduction in the MPR change for the Mono/FDCT cohort compared with the Mono/Dual cohort (-4.6% vs -12.4%; P < 0.001). There was significant improvement in the mean MPR change for the Dual/FDCT cohort compared with the Dual/Dual cohort (3.5% vs -1.3%; P < 0.005). Conclusions: The results of this retrospective database analysis suggest that rosiglitazone-metformin FDCT yielded significant improvements in medication adherence rates compared with dual therapy regimens. (Clin Ther. 2004;26:2066-2075) Copyright © 2004 Excerpta Medica, Inc. Key words: fixed-dose combination therapy, compliance, adherence, diabetes mellitus, oral antidiabetic medi- cations, oral hypoglycemic agents. Preliminary results of this study were presented in abstract form at the 64th Scientific Sessions of the American Diabetes Association, June 4-8, 2004, Orlando, Florida. Accepted Jor publication October 13, 2004. doi:l 0.1016/j.clinthera.2004.12.018 Printed in the USA. Reproduction in whole or part is not pennitted. 0149 29181041519.00 2066 Copyright © 2004 Excerpts Medics, Inc.
`
`Par Pharm., Inc.
`Exhibit 1034
`Page 001
`
`

`

`D.R.Vanderpoel et al. INTRODUCTION According to the Centers for Disease Control and Pre- vention (CDC), the prevalence of diagnosed type 2 diabetes mellitus (DM-2) in the United States more than doubled between 1980 and 2002, from 5.8 million to 13.3 millionJ The direct and indirect costs associated with this disease are estimated to total approximately US $132 billion/y. 2 In addition, the CDC estimates that of Americans born in 2000, 1 in 3 will develop DM-2. 3 Self-care activities associated with diabetes include exercise, diet, blood glucose monitoring, and proper medication use. ~ Substantial attention has been given to enhancing self-care behaviors, ~-r with consid- erable focus on medication adherence. 8-1° Evidence suggests that medication adherence decreases with the increasing complexity of diabetes therapy regi- mens, m-~3 which is important because of the direct relationship found between poor adherence and poor glycemic control, ~<~5 and the inverse relationship found between adherence and health care service use. 1<st Advancements in medication formulations are helping to overcome some of the limitations with multidrug regimens. Fixed-dose combination therapy (FDCT) allows multiple medications, often with complementary mechanisms of action, to be given in a single formu- lation. ~8 In recent years, FDCT was introduced in multiple drug classes and disease states, including DM-2.19-22 FDCT was originally used to allow medi- cations to work synergistically, enabling improved clinical outcomes. ~8 However, due to overall reduc- tions in regimen complexity allowed by FDCT, improvements in medication adherence have also been observed. 19,23 A study by Melikian et a123 showed that glyburide-metformin FDCT increased subjects' medication adherence rates compared with dual therapy with the 2 agents. FDCTs for DM-2 are available in several dosage for- mulations. Therefore, studies of the new FDCTs are needed to validate previous results. In 2002, an FDCT composed of rosiglitazone maleate plus met- formin hydrochloride became available. It is indicat- ed for twice-daily treatment of DM-2 in subjects whose disease is uncontrolled with metformin or rosiglitazone monotherapy or dual therapy with the 2 agents. At launch, it was available in 3 dosage formu- lations: rosiglitazone-metformin 1 mg/500 mg, 2 mg/ 500 mg, and 4 mg/500 mg. It has since become avail- able in 2 additional formulations: 2 mg/1000 mg and 4 mg/1000 mg. The objective of the present study was to assess the changes in medication adherence rates associated with switching between oral hypoglycemic agent (OHA) regimens--specifically, switching from either monotherapy or dual therapy with metformin and/or rosiglitazone to rosiglitazone-metformin FDCT using data from a population of subjects in a health bene- fits company database. MATERIALS AND METHODS Prescription claims data for this retrospective data- base analysis were obtained from the pharmacy claims database of a large health benefits company encompassing -3.5 million covered members. Subjects in the database were enrolled in 1 of 4 health benefit designs: health maintenance organization (HMO), preferred-provider organization (PPO), independent plan, or Medicare risk. The privacy office and statisti- cal department of the health benefits company reviewed and approved the limited data set used in this analysis. Study Population Subjects aged >18 years with active pharmacy bene- fits coverage and who had at least 1 pharmacy claim for rosiglitazone or metformin during the identifica- tion period were included in this analysis. The first study medication claim within the identification peri- od was designated as the index prescription. At least 2 prescription claims for study medications in both the preindex and postindex periods were also required for study inclusion. The index date was defined as the first day of the postindex period. Thus, the index prescrip- tion was included as one of the postindex prescrip- tions. Only data from subjects who maintained contin- uous medication therapy during the study period were included in the analysis. Continuous medication therapy was defined as therapy without a lapse of >60 days between the date of days' supply expiration of any pre- scription fill and the subsequent claim date. Finally, only data from subjects who maintained continuous plan enrollment during the duration of the study peri- od were included. Data from subjects <18 years of age or those with nontraditional pharmacy benefits (eg, a generic-only plan that would limit their ability to obtain the study medications) were excluded. 2067
`
`Par Pharm., Inc.
`Exhibit 1034
`Page 002
`
`

`

`CLINICAL THERAPEUTICS @ Subjects were stratified into 1 of 5 therapy cohorts based on study medication use (Table I), as follows: monotherapy with either rosiglitazone or metfor- min throughout the study period (Mono/Mono); monotherapy with either agent in the preindex peri- od and dual therapy with both agents in the post- index period (Mono/Dual); monotherapy with either agent in the preindex period and rosiglitazone- metformin FDCT in the postindex period (Mono/ FDCT); dual therapy throughout the study period (Dual/Dual); dual therapy in the preindex period and rosiglitazone-metformin FDCT in the postindex peri- od (Dual/FDCT). Study Period Definitions All study data were obtained from a 22-month period from May 1, 2002, to February 29, 2004. Prescription claims for each subject were analyzed over a 12-month study period. The study period included a 6-month preindex and a 6-month postin- dex time period. The study's identification period was defined as the 10 months between November 1, 2002, and August 31, 2003. During this period, all subjects had an index date assigned to them. In sub- jects who did not switch to FDCT, the index date was defined as the first prescription fill date for a study medication during the identification period; in sub- jects who switched to FDCT, the index date was defined as the first fill date for FDCT during the iden- tification period. Adherence Rate Definitions Using the prescription claims database, a medication possession ratio (MPR) was calculated for each subject in the 6-month preindex and 6-month postindex peri- ods. The MPR is a proxy measurement of medication adherence, with a scale of 0% to 100%, in which high- er values indicate higher medication adherence. Due to variations in days' supply, values >100% were possible with the MPR adherence rate calculation. Thus, all val- ues >100% were truncated to 100% for the purpose of analysis. MPR was calculated as follows: MPR = Total days' supply obtained Date of last claim - Date of first claim'~ + Days supply of last claim Y For subjects using dual therapy, a dual therapy MPR (DTMPR) was calculated, as follows: DTMPR = (Total days' supply obtained)/2 Date of last claim - Date of first claim~ + Days supply of last claim J Statistical Analysis The primary focus of this analysis was to compare changes in medication adherence between monother- apy, dual therapy, and FDCT regimens. Specifically, comparisons of interest included the adherence rates of subjects switching from monotherapy to dual therapy versus those switching from monotherapy to FDCT (Mono/Dual vs Mono/FDCT), and subjects remaining on dual therapy versus those switch- ing from dual therapy to FDCT (Dual/Dual vs Dual/ FDCT). To compare changes in adherence, an equivalent form of analysis of covariance (ANCOVA) was per- formed using a general linear model. 2~ The model's outcome variable (ie, change in adherence) was cal- culated for each subject, as follows: MPR Change = Postindex MPR - Preindex MPR Table I. Cohort descriptions, determined by drug therapy. Cohort Preindex Therapy Postindex Therapy Mono/Mono Mono/Dual Nono/FDCT Dual/Dual Dual/FDCT Mefformin or rosiglitazone Mefformin or rosiglitazone Mefformin or rosiglitazone Mefformin and rosiglitazone Mefformin and rosiglitazone Mefformin or rosiglitazone Mefformin and rosiglitazone Rosiglitazone/metformin Mefformin and rosiglitazone Rosiglitazone/metformin FDCT = fixed-dose combination therap;z 2068
`
`Par Pharm., Inc.
`Exhibit 1034
`Page 003
`
`

`

`D.R.Vanderpoel et al. Within this model, a multivariate slope test was performed to ensure equality of the slopes across stra- ta. Adjusted least squares mean of the change in adherence was calculated for each of the cohorts of interest. Pairwise comparisons were performed using the Tukey test to determine whether differences were significant between the cohorts of interest. All hypothesis testing was performed at a significance level of 0.05. Several factors were controlled for within the multi- variate model, including age, gender, insulin use, non- study OHA use, total pill burden at the index date, and chronic disease score (CDS) (defined later) from the preindex period. Insulin and nonstudy OHA use were determined from the prescription claims data. Subjects were flagged if they had claims for any of these medication types. The total pill burden calcula- tion included all oral nonstudy medications and was calculated for each study subject at the index date. The CDS score is a metric that uses age, gender, and medication history obtained from pharmacy claims data to calculate a risk-adjustment score. 25,26 Preindex CDS scores were calculated for all study subjects. Specifically, the CDS score by Clark et a125 was calculated, which includes 3 risk-adjustment metrics, including total costs, outpatient costs, and primary care visits. To test for differences in demographic characteris- tics across the therapy cohorts, an unbalanced analy- sis of variance, chi-square, and t tests were used. All analyses for this study were conducted using SAS ver- sion 8.2 (SAS Institute Inc., Cary, North Carolina). Multivariate analyses were performed using PROC GLM (SAS Institute Inc.). 2~ RESULTS Of 178,288 subjects with DM-2 identified using the database, data from the 16,928 who met the inclu- sion criteria were used in this analysis. As shown in Table II, the Mono/Mono cohort was the largest group, consisting of 14,291 (84.4%) subjects. Gender was evenly distributed in the overall population, with data from 8499 (50.2%) men and 8429 (49.8%) women included. The mean (SD) age was 58.12 (11.97) years (range, 19-99 years). Mean (SD) total pill burden at the index date was 4.53 (4.77) pills, with a medi- an of 3.00 pills (range, 0-56). Of the total sample, 7956 (47.0%) had HMO coverage, 5222 (30.9%) had PPO coverage, 3722 (22.0%) had Medicare risk bene- fits, and 28 (0.2%) subjects had individual plan bene- fits. A total of 2700 (16.0%) subjects were concur- rently using insulin therapy, and 10,045 (59.3%) subjects were concurrently using nonstudy OHAs. The total study sample's predicted mean (SD) CDS total cost calculation was US $3211.81 ($1581.98). The mean (SD) predicted CDS outpatient cost was $1504.45 ($664.65), and the mean (SD) predicted number of CDS primary care visits was 3.20 (0.94) (range, 0.82-8.09 visits). Analysis of Covariance The overall result of the multivariate model was statistically significant (F = 52.73; P < 0.001), indicat- ing that at least 1 of the independent variables was different from 0. Further investigation revealed that the results of 3 of 9 independent variables compared in the model were statistically significant. Specifically, the therapy cohorts variable was significant (F = 125.75; P < 0.001), indicating that a statistically significant difference in adherence change existed among the 5 study cohorts. In addition, gender and total pill burden at the index date were statistically significant variables (F = 8.48, P < 0.004; and F = 4.61, P < 0.032, respectively). Monotherapy Of the 13,371 subjects using a monotherapy regi- men before the index date, 14,291 (91.8%) remained on monotherapy (Mono/Mono), 931 (6.0%) switched to dual therapy (Mono/Dual), and 349 (2.2%) switched to FDCT (Mono/FDCT) after the index date (Table II). The mean (SD) age of the monother- apy cohorts were 38.30 (12.17), 36.87 (11.31), and 34.91 (11.33) years, respectively. A significantly lower proportion of subjects had Medicare risk bene- fits in the Mono/Dual cohort (147 subjects [13.8%]; P < 0.001) and the Mono/FDCT cohort (34 subjects [13.3%]; P < 0.003) compared with the Mono/Mono cohort (3394 subjects [23.7%]). In addition, the pro- portion of subjects with nonstudy OHA use was sig- nificantly lower in the Mono/Mono cohort (8234 subjects [37.8%]; P < 0.001) and Mono/FDCT cohort (201 subjects [37.6%]; P < 0.001) compared with the Mono/Dual cohort (673 subjects [72.3%]). The CDS scores were each significantly lower for the Mono/FDCT cohort compared with the Mono/Mono 2069
`
`Par Pharm., Inc.
`Exhibit 1034
`Page 004
`
`

`

`CLINICAL THERAPEUTICS ® Table II. Baseline characteristics of the study population.* Mono/Mono Mono/Dual Hono/FDCT Dual/Dual DuaI/FDCT Population Characteristic (n = 14,29 I) (n = 93 I) (n = 349) (n = 1230) (n = 127) (N = 16,928) Gender;, no. (%) Hale 7257 (50.8) 450 (48.3) 158 (45.3) 500 (40.7) 64 (50.4) 8429 (49.8) Female 7034 (49.2) 481 (51.7) 191 (54.7) 730 (59.4) 63 (49.6) 8499 (50.2) Age, y Mean (SD) 58.50 (12.17) 56.87 (I 1.31) 54.91 (I 1.35) 56.00 (9.72) 53.69 (10.58) 58.12 (I 1.97) Median 58.00 56.00 54.00 56.00 53.00 58.00 Range 19 99 19 90 26 89 19 89 32 87 19@9 Total pill burden Mean (SD) 4.56 (4.84) 4.31 (4.36) 3.45 (3.27) 4.73 (4.67) 4.49 (3.61) 4.53 (4.77) Median 3.00 3.00 2.00 4.00 3.00 3.00 Range 0 56 0 33 (>22 0 38 0 20 (~56 Insurance type, no. (%) HMO 6,607 (46.2) 463 (49.7) 172 (49.3) 657 (53.4) 57 (44.9) 7,956 (47.0) PPO 4,263 (29.9) 320 (34.4) 123 (35.2) 462 (37.6) 54 (42.5) 5,222 (30.8) Medicare 3,394 (23.7) 147 (15.8) 54 (15.5) III (9.0) 16 (12.6) 3,722 (22.0) Independent 27 (0.2) I (0.1) 0 (0.0) 0 (0.0) 0 (0.0) 28 (0.2) Insulin use, no. (%) Yes 12,049 (84.3) 762 (81.8) 289 (82.8) 1,026 (83.4) 102 (80.3) 14,228 (84.1) No 2,242 (15.7) 169 (18.2) 60 (17.2) 204 (16.6) 25 (19.7) 2,700 (16.0) Nonstudy OHA use, no. (%) Yes 8,254 (57.8) 675 (72.5) 201 (57.6) 837 (68.0) 78 (61.4) 10,045 (59.3) No 6,037 (42.2) 256 (27.5) 148 (42.4) 393 (32.0) 49 (38.6) 6,883 (40.7) FDCT = fixed-dose combination therapy; HMO = health maintenance organization; PPO = preferred@rovider organization; OHA = oral hypoglycemic agent. *See Figure I for a description of each study cohort. and Mono/Dual cohorts: predicted total cost, US $2870.94 versus $3232.79 (P < 0.001) and $2870.94 versus $3212.24 (P < 0.002), respectively; predict- ed outpatient cost, $1390.72 versus $1506.68 (P < 0.005) and $1390.72 versus $1533.39 (P < 0.003), respectively; and predicted primary care visits, 3.06 versus 3.21 visits (P < 0.01) and 3.06 versus 3.24 vis- its (P < 0.007), respectively. As illustrated in Figure 1, the Mono/Dual cohort exhibited a slightly lower preindex MPR (0.80 [0.19]) compared with the Mono/Mono cohort (0.90 [0.13]) and the MonofFDCT cohort (0.87 [0.15]). In compar- ing the differences in postindex and preindex MPRs, declines in the mean MPR changes for the Mono/Mono and MonofFDCT cohorts were only slight (-1.5% and -4.6%, respectively), whereas a substantial decline was found in the Mono/Dual therapy cohort (-12.4%). As shown in Figure 2, the mean number of study medica- tion refills increased slightly from the preindex to the postindex period in the Mono/Mono cohort (from 4.60 [1.34] to 5.47 [1.51]) and the MonofTDCT cohort (from 4.11 [1.38] to 5.27 [1.62]) cohort, whereas a dramatic decrease from 5.74 [2.25] to 4.08 [2.82] was observed in the Mono/Dual cohort. A direct relation- ship between the changes in MPR and number of refills does not exist due to the wide variations in days' sup- ply obtained per refill. The pairwise comparisons in the multivariate analysis confirmed statistically significant between- cohort differences in change in adherence. As illus- trated in Figure 3, the analysis revealed a statistically significant difference in mean MPR change between the Mono/Dual and the Mono/FDCT cohorts (7.8%; 95% CI, 5.3%-10.4%; P < 0.001). Although a direct comparison between the Mono/iVlono and Mono/FDCT cohorts requires cau- 2070
`
`Par Pharm., Inc.
`Exhibit 1034
`Page 005
`
`

`

`D.R.Vanderpoel et al. rv, o_ '21 • Preinclex [] Postinclex 0.90 0.89 0.87 0.80 0.83 0.84 0.83 n To 0.82 I I I I Mono/Mono Mono/Dual Mono/FDCT Dual/Dual Dual/FDCT Study Cohort Figure I. Medication possession ratios (MPRs) in the preindex and postindex periods. Mono/Mono = subjects receiving monotherapy with metformin or rosiglitazone throughout the study (n = 14,291); Mono/Dual = subjects receiv- ing monotherapy with either agent in the preindex period and dual therapy with both agents in the postindex period (n = 931); Mono/FDCT = subjects receiving monotherapy with either agent in the preindex period and rosiglitazone/metformin fixed-dose combination therapy (FDCT) in the postindex period (n = 349); Dual/Dual = subjects receiving dual therapy with both agents throughout the study (n = 1230); Dual/FDCT = subjects receiv- ing dual therapy with both agents in the preindex period and FDCT in the postindex period (n = 127). rv, .m 6 Z 5.47 5.74 • Preinclex [] Postinclex 5.27 5.24 5.20 I I I I Mono/Mono Mono/Dual Mono/FDCT Dual/Dual Dual/FDCT Study Cohort Figure 2. Number of study medication refills in the preindex and postindex periods. See Figure I for a description of each study cohort. To ease comparability, the total number of study medication refills was divided by 2 for all dual therapy cohorts. 2071
`
`Par Pharm., Inc.
`Exhibit 1034
`Page 006
`
`

`

`CLINICAL THERAPEUTICS @ • Mono/Dual [] Mono/FDCT "¢ -3 o_ IE .E "<1 -6 o • - -9- -12- -15 -4.6% - 12.4% Figure 3. Changes in medication possession ratio (MPR) in subjects receiving monotherapy in the preindex period. See Figure I for a description of each study cohort. P < 0.001 between cohorts. tion due to the clinical differences in disease progres- sion, the mean MPR change was compared in the Mono/Mono and Mono/FDCT cohorts to determine whether a similar adherence rate could be achieved with FDCT as with monotherapy. A statistically sig- nificant difference was detected (3.1%; 95% CI, 1.0%- 5.0%; P < 0.002). Dual Therapy Of the 1357 subjects on a dual therapy regimen in the preindex period, 1230 (90.6%) subjects remained on dual therapy (Dual/Dual), whereas 127 (9.4%) switched to FDCT (Dual/FDCT) in the postindex peri- od (Table TT). The proportion of men was significantly higher in the Dual/Dual cohort compared with the Dual/FDCT cohort (730 [59.4%] vs 63 [49.6%] sub- jects, respectively; P < 0.034). The mean (SD) age was significantly higher in the Dual/Dual cohort compared with the Dual/FDCT cohort (56.00 [9.72] vs 53.69 [10.58] years, respectively; P < 0.02). The 2 cohorts had no statistically significant differences in total pill burden, insulin use rate, nonstudy OHA use rate (Table TT), and predicted CDS scores at the index date. As shown in Figure 1, the Dual/Dual cohort exhib- ited a slightly higher mean (SD) MPR (0.84 [0.15]) in the preindex period compared with the Dual/FDCT cohort (0.79 [0.18]). Figure 2 shows that the mean (SD) number of study medication refills increased slightly for both the Dual/Dual (from 4.56 [1.42] to 5.24 [1.49]) and Dual/FDCT (3.87 [1.48] to 5.20 [1.69]) cohorts from the preindex to the postindex period. A slight decline in MPR change (-1.3%) was observed in the Dual/Dual cohort; in contrast, a sub- stantial increase of 3.5% was observed in the Dual/ FDCT cohort (Figure 4). Adjusting for all covariates, a statistically significant difference of mean change in adherence rate was detected between the Dual/Dual and Dual/FDCT cohorts (4.8%; 95% CI, 1.0%-8.6%; P < 0.005). DISCUSSION Previous studies have investigated the relationship between drug regimen complexity and adher- ence. 11-13 However, there has been limited work examining the relationship between FDCT regimens and adherence in DM-2. 23 Additional studies of FDCT are needed to validate the results of completed adherence studies. Therefore, the main objective of the present study was to compare the change in medi- cation adherence as a result of a change in drug regi- men complexity. Within the present study, the Mono/FDCT cohort had a significant improvement in adherence com- pared with the Mono/Dual cohort (P < 0.001). One • Dual/Dual [] DuaI/FDCT ,¢ o_ IE c "<1 o IE 4 I 3- 2- I- 0ii -I i -2- I -I.3% 3.5% Figure 4. Changes in medication possession ratio (MPR) in subjects receiving dual therapy in the preindex period. See Figure I for a description of each study cohort. P < 0.005 between cohorts. 2072
`
`Par Pharm., Inc.
`Exhibit 1034
`Page 007
`
`

`

`D.R.Vanderpoel et al. possible explanation for this improvement is the decrease in regimen complexity. One would expect the Mono/FDCT cohort to have an adherence rate decline similar to that of the Mono/Mono cohort because neither had a change in regimen complex- ity. However, a statistically significant difference was found between these 2 cohorts (P < 0.002). This dif- ference could be partially due to illness perception factors (ie, heightened awareness of disease and/or associated treatments), as identified in previous studies, 27,28 which could not be controlled for in the multivariate model of a retrospective database analysis. In addition, the rosiglitazone/metformin FDCT dosage options--2 mg/1000 mg and 4 mg/ 1000 mg--which were available only after Novem- ber 2003, may also be expected to influence adher- ence rates in subjects who require higher doses of metformin. In contrast, when comparing the Dual/Dual cohort with the Dual/FDCT cohort, an overall change in adherence from a decline to an increase was observed (P < 0.005). This result may be of interest because the reduction in regimen complexity may be associated with a reverse in the decline in adherence. However, another potential cause of the change in adherence could be the temporary effect of guidance given by the physician when adding a new medication to the existing regimen. In the present study, the total pill burden compared by means and medians was similar, yet large differ- ences were apparent when comparing the range in pill burden. The Mono/Mono cohort had the greatest range in pill burden compared with all of the other cohorts. However, the overall effect of this large range on the ANCOVA results was minimal due to the large sample size of the Mono/Mono cohort. The present study shares a similar design with pre- viously completed retrospective adherence studies. However, its methods differ considerably. One major difference is the use of change in MPR, rather than MPR at study end, as the outcome measure. Change in MPR provides a normally distributed outcome, which is an important assumption for the statistical methods used. In addition, change in MPR provides a directional sign that explains whether medication adherence is increasing or decreasing. A second important difference is the choice of statistical meth- ods. Cohorts of unequal size are common in retro- spective database analyses; therefore, a general linear model was used because of its robustness to hetero- geneity of variance compared with traditional ANCOVA. 2~ This study inherits the limitations of claims data- base research. > Specifically, risks of overcoding and undercoding--as well as miscoding--must be acknowledged. In addition, underreported pharmacy claims can occur due to physician sampling, cash- paying customers, and Medicare recipients meeting maximum allowable benefits. Several factors were not investigated in this analy- sis, limiting the ability to provide further insight into medication adherence. Specifically, economic issues, including member expenditures, insurance type, and safety/tolerability issues, were not investigated. Differences in the distribution of insurance type across the cohorts were noted. However, further stud- ies are needed to determine the overall effect of insur- ance type on medication use. Likewise, the effect of differences in member out-of-pocket expenditures and/or medication tolerability on drug use were not within the scope of this study. Other limitations included the use of proxy adher- ence measurements. MPR was chosen for this analy- sis due to its ability to provide a simple measure of drug use frequency over a period of time. Compared with the adherence rate calculation (ie, total days' supply obtained divided by total days in the study period), the MPR calculation is affected solely by the gaps in medication refills that allow for simplified interpretation. However, limitations exist with the MPR calculation, most notably, the need for continu- ous therapy with multiple prescriptions for an accu- rate calculation. This requirement can cause an inflat- ed adherence rate due to the exclusion of data from subjects with poor medication adherence. CONCLUSIONS Based on the results of this analysis of a health bene- fits company population, rosiglitazone-metformin FDCT yielded significant improvements in medica- tion adherence rates compared with dual therapy reg- imens. Additional studies are needed to determine the sustainability of the differences in medication adherence rates over a longer follow-up period, and the clinical implications of improved medication adherence. 2073
`
`Par Pharm., Inc.
`Exhibit 1034
`Page 008
`
`

`

`CLINICAL THERAPEUTICS ® ACKNOWLEDGHENTS The authors gratefully acknowledge contributions from Brenda Miller, MAT (data programming) and Susmita Utukuri, PharmD (editorial assistance). REFERENCES 1. Centers for Disease Control and Prevention (CDC). Data & trends: Diabetes surveillance system, preva- lence of diabetes [CDC Web site]. Available at: http:// www. cdc. gov/diab e t e s/s t atis tics/p r ev/national/ figpersons.htm. Accessed July 12, 2004. 2. Centers for Disease Control and Prevention (CDC). National diabetes fact sheet: National estimates on dia- betes [CDC Web site]. Available at: http://wwwcdc. gov/diabetes/pubs/estimates.htm. Accessed July 12, 2004. 3. Centers for Disease Control and Prevention (CDC). Diabetes: Disabling, deadly, and on the rise [CDC Web site]. Available at: http://www.cdc.gov/nccdphp/ aag/aag_ddt.htm. Accessed July 12, 2004. 4. Polonsky WH, Earles J, Smith S, et al. Integrating med- ical management with diabetes self-management train- ing: A randomized control trial of the Diabetes Out- patient Intensive Treatment program. Diabetes Care. 2003;26:3048-3053. 5. Cox RH, Carpenter JP, Bruce FA, et al. Characteristics of low-income African-American and Caucasian adults that are important in self-management of type 2 dia- betes. J Community Health. 2004;29:155-170. 6. Keyserling TC, Samuel-Hodge CD, Ammerman AS, et al. A randomized trial of an intervention to improve self-care behaviors of African-American women with type 2 diabetes: Impact on physical activity. Diabetes Care. 2002;25:1576-1583. 7. Jones H, Edwards L, Vallis TM, et al, for the Diabetes Stages of Change (DISC) Study Group. Changes in dia- betes self-care behaviors make a difference in glycemic control: The Diabetes Stages of Change (DISC) stu@ Diabetes Care. 2003;26:732-737. 8. Grant RW, Devita NG, Singer DE, Meigs JB. Poly- pharmacy and medication adherence in patients with type 2 diabetes. Diabetes Care. 2003;26:1408-1412. 9. Cramer JA. A systematic review of adherence with med- ications for diabetes. Diabetes Care. 2004;27:1218-1224. 10. Walker EA, Usher JA. Understanding and enhancing adherence in adults with diabetes. Curr Diab Rep. 2003;3:141-148. 11. Dailey G, Kim MS, Lian JE Patient compliance and persistence with antihyperglycemic drug regimens: Evaluation of a Medicaid patient population with type 2 diabetes mellitus. Clin Ther. 2001;23:1311-1320. 12. Dailey G, Kim MS, kianJE Patient compliance and per- sistence with anti-hyperglycemic therapy: Evaluation of a population of type 2 diabetic patients. ] Int Med Res. 2002;30:71-79. 13. Donnan PT, MacDonald TM, Morris AD. Adherence to prescribed oral hypoglycaemic medication in a popu- lation of patients with type 2 diabetes: A retrospective cohort study Diabet Med. 2002;19:279-284. 14. Schectman JM, Nadkarni MM, Voss JD. The associa- tion between diabetes metabolic control and drug adherence in an indigent population. Diabetes Care. 2002 ;25:1015-1021. 15. Guillausseau PJ. Influence of oral antidiabetic drugs compliance on metabolic control in type 2 diabetes. A survey in general practice. Diabetes Metab. 2003;29: 79-81. 16. 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. 17. Hepke KL, Martus MT, Share DA. Costs and utilization associated with pharmaceutical adherence m a diabet- ic population. Am ] Manag Care. 2004; 10:144-151. 18. Sica DA. Rationale for fixed-dose combinations in the treatment of hypertension: The cycle repeats. Drugs. 2002 ;62:443-462. 19. Taylor AA, Shoheiber O. Adherence to antihyper- tensive therapy with fixed-dose amlodipine besylate/ benazepril HC1 versus comparable component-based therap)~ Congest Heart Fail. 2003;9:324-332. 20. Nelson HS. Advair: Co