Haifamed.org.il

CLINICAL ■
Yelena Rozenfeld, MPH; Jacquelyn S. Hunt, PharmD, Craig Plauschinat, PharmD, MPH; and Ken S. Wong, PharmD Objective: To evaluate adherence to oral diabetes
medications (ODMs) in patients with type 2
T ype 2 diabetes is a growing worldwide epidemic, with approx-
imately 20 million diagnosed and undiagnosed persons in the United States.1 Of national concern is the finding that fewer diabetes and the impact of ODM adherence than half of Americans with diabetes have achieved a glyco- Study Design: Retrospective observational study.
sylated hemoglobin (A1C) level of <7%.2 Nonadherence to medication Methods: Medical and pharmacy claims from a
therapy is among several factors contributing to suboptimal glycemic con- managed care plan in Oregon were used to iden- trol. A recent systematic review found significant variation in adherence tify adults with diabetes who newly initiatedODM therapy (n = 2741); a subset of this cohort to oral diabetes medications (ODMs), ranging from 36% to 93%.3 Patient linked to electronic health records was used to nonadherence to medications prescribed for diabetes has been shown to evaluate the relationship between adherence andglycemic control (n = 249). Glycemic control was decrease treatment effectiveness4-6 and increase healthcare costs.7,8 assessed based on most recent glycosylated Although published studies address clinical outcomes, there is a continu- hemoglobin (A1C) measurement within the study period.
ing need to evaluate the association between medication adherence and Results: Mean cohort age was 54 years; 46%
diabetes outcomes. The purposes of this study were to evaluate (1) patient initiated therapy with metformin, 39% with a adherence to ODM and (2) the relationship between adherence and sulfonylurea, and 12% with a thiazolidinedione.
Mean adherence overall was 81%, and 65% of subjects had good adherence (>80%). Increasingage and comorbidity burden were associatedwith higher medication adherence. In the patient subset with A1C measurements, mean baselineA1C was 8%. An inverse relationship existed Setting and Study Population
between ODM adherence and A1C; controlling This retrospective cohort study was conducted by the Providence for baseline A1C and therapy regimen, each 10%increase in ODM adherence was associated with Primary Care Research Network in Oregon within an integrated delivery a 0.1% A1C decrease (P = .0004).
network. After institutional review board approval, medical and pharma- Conclusion: Although most patients were ad-
cy claims were received from the network managed care plan with a 2-tier herent to ODM therapy, adherent patients Managed Care &
were more likely to achieve glycemic control pharmacy benefit design. Eligible patients were identified based on the than nonadherent patients. Greater efforts are Healthcare
Communications, LLC
18 years, a diagnosis of diabetes (International needed to facilitate diabetes self-managementbehaviors to improve patient outcomes.
Classification of Diseases, Ninth Revision code 250.xx), an ODM prescrip- tion (January 31, 2001, through December 31, 2004), and continuous enrollment for 6 months or more. Patients were followed for 12 months from the date of the first ODM prescription fill (the index date). Patients were required to be ODM therapy naive (no fills for any ODM during the 6 months before the index date). ODM agents were grouped by therapeu- tic class, including metformin, sulfonylureas (SUs), thiazolidinediones (TZDs), α-glucosidase inhibitors, and meglitinides. Patients whoseindex diabetes medication was insulin were excluded. Because of small sample size, study results are not presented for patients whose index regi- men consisted of an α-glucosidaseinhibitor or a meglitinide.
In this issue
Variable Definitions
For author information and disclosures,
Medication adherence was calculated see end of text.
www.ajmc.com
for all patients with at least 2 fills of the CLINICAL ■
index ODM, and was defined as the sum of the days supply demographic, clinical, and medication characteristic com- from the index prescription date to the last fill date (exclud- parisons between groups were completed by using t tests, ing days supply that was dispensed at the final prescription analysis of variance, and correlation analysis for evalua- fill), divided by the duration of therapy. Patients using free tion of continuous variables and the χ2 test for categorical combinations of more than 1 ODM were classified according to the first medication that was filled during the study period, Multiple regression was used to evaluate the relationship and study metrics were calculated for that drug. For free com- between adherence and glycemic control by means of a back- bination regimens, patients were considered adherent on a ward elimination process. Several patient-related and specific day if all drugs were considered to be “on hand” on provider-related factors were used as covariates in the model that date. Patients with adherence of less than 80% were clas- for statistical adjustment, including original ODM regimen, baseline A1C, sex, age at index date, CDS, and medication Patients’ complete refill history at baseline was used to cal- burden, as well as provider sex, specialty, and years since med- culate the modified chronic disease score (CDS) to assess ical school graduation. All variables except baseline A1C and patient comorbidity.9 The number of concurrent (non–dia- index ODM regimen were eliminated as nonsignificant pre- betes-related) medications also was assessed as another mea- dictors. The index regimen was categorized as following: (1) metformin, (2) metformin + SU, and (3) SU. All data manip- A subset of study patients who also received care from pri- ulation and statistical analysis were completed using SAS ver- mary care physicians employed within the integrated delivery sion 9.1 (SAS Institute Inc, Cary, NC).
network was identified to evaluate the relationship between medication use and glycemic control. This subset was limited to patients with an A1C test result at study baseline (180 days before the study index date and 60 days subsequent to the The cohort of patients who newly initiated ODM therapy index date). Glycemic control at follow-up was assessed after included 2741 subjects for evaluation of utilization metrics.
a stabilization period subsequent to ODM initiation; it was Mean cohort age was 54 ± 11 years, and 49% were female based on the most recent A1C within the study period, col- (Table). Overall, 84% initiated ODM monotherapy, 15% ini-
lected at least 60 days after the index date. tiated dual combination therapy, and 1% initiated triple com- bination therapy; 41% were prescribed metformin, 33% SUs, Statistical Analysis
9% TZDs, 2% other drugs (including α-glucosidase inhibitors Descriptive statistics were produced for appropriate and meglitinides), and 15% were prescribed various combina- variables. Continuous data were described by means and tion therapies. The mean CDS overall was 2.89 ± 0.99, with standard deviations, and nominal and categorical data a small but significant difference between SU and TZD were described by frequencies and percentages. Unadjusted patients (2.85 vs 2.99, P = .04).
■ Table. Patient Cohort Demographic Characteristics and Drug Utilization Metrics by Index Drug Class*
Metformin
Sulfonylureas
Thiazolidinediones
Total†
Characteristic
(n = 1274)
(n = 1081)
(n = 2741)
Patients with >1 fill of index drug, n (%) *Data are given as mean ± SD unless otherwise indicated.
†Because of small sample size, results for patients whose index drug was an α-glucosidase inhibitor or a meglitinide (n = 49) are not presented byindex drug category in the Table, although they are included in the “Total” column. ‡Adherence was calculated for the patient subset with at least 2 fills of the index drug(s). CDS indicates chronic disease score. Antidiabetic Medication Adherence and Glycemic Control
Mean adherence for the study cohort was 81% and did not medications.3 However, these findings should be interpreted significantly differ by therapeutic class. The effect of patient cautiously, as the studies included diverse populations, various clinical and demographic characteristics on index ODM methods for calculating adherence, and different study dura- adherence was evaluated. Older patients were more likely to tions. Many of these studies also included adherence to be adherent (mean age 56 years vs 52 years, P <.0001), but insulin, which is difficult to estimate based on methodology there was no difference between men and women in adher- that relies on pharmacy claims. Other investigators have ence (P = .61). Adherent patients had a significantly higher found adherence rates similar to ours at 1 or 2 years subse- disease burden, as measured by either CDS (2.99 vs 2.86, quent to new ODM initiation in managed care settings.10,11 P = .0022) or total medication burden (10.3 vs 9.6, P = .022), Using a claims database from a national pharmacy benefits than patients classified as nonadherent. manager, one of the largest studies found mean 1-year adher- In the patient subset with A1C test results (n = 249), mean ence was 79% in a cohort of 79 498 new ODM users.11 As in baseline A1C was 8.0% ± 1.5%. Approximately 44% of this our study, these researchers found similar rates of adherence subset initiated ODM therapy with metformin, 38% with among patients taking metformin and those taking SUs, and SUs, and 12% with combination (metformin plus SU) thera- they did not include currently marketed TZDs because of py. Additionally, these patients were not different from the insufficient data. Other similar ODM studies conducted out- main study population with respect to age, sex, and CDS.
side the United States also have demonstrated consistent For patients initiating index ODM therapy with SU, met- formin, or metformin plus SU (n = 235; TZD patients were Our finding of an association between increasing comor- excluded because of small sample size), multiple regression was bidity and adherence needs to be interpreted with caution used to determine the relationship between adherence and because this association is not reported consistently in the A1C, controlling for therapeutic regimen and baseline A1C.
published literature. Although some studies reported no asso- An inverse relationship was observed between ODM adher- ciation or negative association,15,16 several publications ence and A1C (Figure), in which a 10% increase in index
reported results similar to those in our study.17,18 Furthermore, ODM adherence was associated with a 0.1% decrease in A1C it is conceivable that patients with a higher number of chron- (P = .0004). Index ODM regimen also was significantly asso- ic conditions are better informed about diabetes and its com- ciated with A1C; metformin demonstrated a stronger associa- plications and, therefore, would maintain greater rates of tion (P = .02) than SU (P = .04). ■ Figure. Association Between A1C and Adherence, Adjusted for Baseline
A1C and the ODM Regimen*
Adherence (%)
*Metformin plus a sulfonylurea was used as the reference group for the index ODM regimen.
A1C indicates glycosylated hemoglobin; ODM, oral diabetes medication.
CLINICAL ■
Although this study provides valuable informa- Take-away Points
tion about diabetes management in a real-world In a managed care setting, this retrospective study demonstrated that setting, several limitations are important to consid- although most patients were adherent with prescribed oral diabetes medica-tion (ODM), about 35% of patients were classified as nonadherent.
er. Many variables were unavailable for study After adjusting for baseline glycosylated hemoglobin (A1C) and therapy inclusion, including the severity and duration of regimen, each 10% increase in adherence with ODM was associated with a type 2 diabetes, patient racial/ethnic background, body mass index, information on patient dietary When lab data are not available (as is common in managed care data- bases), pharmacy claims may serve to identify patients for targeted outcomes/ habits, and frequency of patient glucose self-moni- toring. We studied patients from a managed care Greater efforts are needed to improve medication adherence and patient plan in Oregon; because this study did not include a geographically diverse population, it may be dif- ficult to generalize findings to the US population.
Because our study included patients enrolled in a adherence despite their greater medication burden and managed care organization, study findings may not be applica- ble to patients without health insurance. Medication adher- Several studies of the impact of adherence on glycemic ence was measured by patient prescription refills and not drug control have been conducted, primarily using cohorts derived consumed by the patient; however, other studies have shown from managed care databases with limited geographic repre- that these are highly correlated and have supported the use of sentation. One study of 677 patients from Michigan diagnosed pharmacy claims data for the evaluation of these metrics.17,21 with diabetes, hypercholesterolemia, and hypertension found Despite these limitations, this study provides valuable informa- a 10% increase in nonadherence to metformin and statins tion in support of the literature that suggests a relationship related to a 0.14% increase in A1C.19 Schectman et al studied between adherence and glycemic control in a real-world set- 810 patients with type 2 diabetes from a clinic serving a low- ting, and underscores the importance of medication adherence income population in central Virginia and found that for each 10% increase in adherence, A1C decreased by 0.16%.5 Another cross-sectional study of 301 patients from 6 regional practice sites, which used the patient questionnaire–based Morisky score to evaluate adherence, found that better adher- Overall, study findings indicated good adherence with ence was associated with a 10% lower A1C, adjusted for ODMs in this managed care population with diabetes. An covariates.6 Finally, a study of 2995 patients from a managed association was found between ODM adherence and glycemic care organization in the southeastern United States who were control, such that each 10% increase in ODM adherence was prescribed metformin or SU found significantly higher mean associated with a 0.1% decrease in A1C. These findings high- adherence for patients who reached the goal A1C of <7% light the importance of medication adherence for attaining (77% and 82%, respectively) than for patients who did not glycemic control, thus reducing the incidence of diabetes (62% and 72%, respectively).4 Taken together, these results complications. Initiatives targeting improved medication support the findings of the current study, and it is noteworthy adherence in patients with type 2 diabetes are important to that despite population differences, the studies consistently demonstrated a similar relationship between adherence and Acknowledgments
We acknowledge Parker Pettus, MS, and K. Arnold Chan, MD, ScD, of The United Kingdom Prospective Diabetes Study demon- the Channing Laboratory, Brigham & Women’s Hospital and Harvard Medical strated that reducing A1C in patients with diabetes by 1% School for the development of the SAS code used to calculate the chronic dis- ease score in this study. We also thank James Slater, PharmD, Providence decreased the risk of any diabetes-related complications by Health Plan, for guidance and support in the execution of the study.
21%, death related to diabetes by 21%, and the incidence of Author Affiliations: From Providence Physician Division, Beaverton, Ore
myocardial infarction by 14%.20 Because any improvement in (YR, JSH); Novartis Pharmaceuticals Corporation (CP), East Hanover, NJ; and Novartis Pharmaceuticals Corporation (KSW), San Marino, Calif.
glycemic control across the diabetic range is likely to reduce Funding Source: This study was funded by Novartis Pharmaceuticals
the risk of diabetic complications,20 improved medication adherence to facilitate achievement of glycemic control has Author Disclosure: Drs Plauschinat and Wong are employees of Novartis
Pharmaceuticals Corporation, which provided funding for this study. Dr important consequences for long-term outcomes in patients Plauschinat also reports owning stock in Novartis Pharmaceuticals Corporation. The other authors (YR, JSH) report no relationship or financial Antidiabetic Medication Adherence and Glycemic Control
interest with any entity that would pose a conflict of interest with the subject 9. Clark DO, M Von Korff, Saunders K, Baluch WM, Simon GE. A chron-
ic disease score with empirically derived weights. Med Care. 1995;33:783-795.
Authorship Information: Concept and design (YR, JSH, CP, KSW);
acquisition of data (YR, JSH); analysis and interpretation of data (YR, KSW); 10. Venturini F, Nichol MB, Sung JCY, Bailey KL, Cody M, McCombs JS.
Compliance with sulfonylureas in a health maintenance organization: a
drafting of the manuscript (YR, JSH, CP, KSW); critical revision of the man- pharmacy-record based study. Ann Pharmacother. 1999;33:281-288.
uscript for important intellectual content (YR, JSH, CP, KSW); statistical 11. Boccuzzi S, Wogen J, Fox J, Sung J, Shah A, Kim J. Utilization of
analysis (YR); obtaining funding (JSH, KSW); administrative, technical, or oral hypoglycemic agents in a drug-insured US population. Diabetes logistic support (JSH, KSW); and supervision (JSH).
Address correspondence to: Yelena Rozenfeld, MPH, Providence
12. Morningstar BA, Sketris IS, Kephart GC, Sclar DA. Variation in
Physician Division, 3601 Murray Blvd, Ste 45, Beaverton, OR 97005. E-mail: pharmacy prescription refill adherence measures by type of oral anti- hyperglycaemic drug therapy in seniors in Nova Scotia, Canada. J ClinPharm Ther. 2002;27:213-220.
13. Evans JM, Donnan PT, Morris AD. Adherence to oral hypoglycemics
prior to insulin therapy in type 2 diabetes. Diabet Med. 2002;19:685-688.
14. Spoelstra JA, Stolk RP, Heerdink ER, et al. Refill compliance in type
1. National Institutes of Health. Fact Sheet: Type 2 Diabetes. www.nih.
2 diabetes mellitus: a predictor of switching to insulin therapy? gov/about/researchresultsforthepublic/Type2Diabetes.pdf. Accessed Pharmacoepidemiol Drug Saf. 2003;12:121-127.
15. Wogen J, Kreilick CA, Livornese RC, et al. Patient adherence with
2. Fan T, Koro CE, Fedder DO, Bowlin SJ. Ethnic disparities and trends
amlodipine, lisinopril or valsartan therapy in a usual-care setting. J in glycemic control among adults with type 2 diabetes in the U.S. from Manag Care Pharm. 2003;9:424-429.
1988 to 2002. Diabetes Care. 2006;29:1924-1925.
16. Catalan VS, Couture JA, LeLorier J. Predictors of persistence of use
3. Cramer JA. A systematic review of adherence with medications for
of the novel antidiabetic agent Acarbose. Arch Intern Med. diabetes. Diabetes Care. 2004;27:1218-1224.
4. Lawrence D, Ragucci KR, Long LB, Parris BS, Helfer LA. Relationship
17. Christensen DB, Williams B, Goldberg HI, et al. Comparison of pre-
of oral antihyperglycemic (sulfonylurea or metformin) medication scription and medical records in reflecting patient antihypertensive adherence and hemoglobin A1C goal attainment for HMO patients drug therapy. Pharmacotherapy. 1994;28:99-104.
enrolled in a diabetes disease management program. J Manag Care 18. Pedan A, Varasteh LT, Schneeweiss S. Analysis of factors associat-
ed with statin adherence in a hierarchical model considering physician, 5. Schectman JM, Nadkarni MM, Voss JD. The association between
pharmacy, patient, and prescription characteristics. J Manag Care diabetes metabolic control and drug adherence in an indigent popula- tion. Diabetes Care. 2002;25:1015-1021.
19. Pladevall M, Williams LK, Potts LA, et al. Clinical outcomes and
6. Krapek K, King K, Warren SS, et al. Medication adherence and asso-
adherence to medications measured by claims data in patients with ciated hemoglobin A1C in type 2 diabetes. Ann Pharmacother. 2004;38: diabetes. Diabetes Care. 2004;27:2800-2805.
20. Stratton IM, Adler AI, Neil HA, et al. Association of glycaemia with
7. Lee WC, Balu S, Cobden D, Joshi AV, Pashos CL. Prevalence and eco-
macrovascular and microvascular complications of type 2 diabetes nomic consequences of medication adherence in diabetes: a systemat- (UKPDS 35): prospective observational study. BMJ. 2000;321:405-412.
ic literature review. Manag Care Interface. 2006;19:31-41.
21. Choo PW, Rand CS, Inui TS, et al. Validation of patient reports,
8. Sokol MC, McGuigan KA, Verbrugge RR, Epstein RS. Impact of med-
automated pharmacy records, and pill counts with electronic monitor- ication adherence on hospitalization risk and healthcare cost. Med ing of adherence to antihypertensive therapy. Med Care. 1999;37:

Source: http://www.haifamed.org.il/pictures/files/AJMC_08feb_Rozenfeld_71to75.pdf