Refill adherence and persistence to lipid‐lowering medicines in patients with type 2 diabetes: A nation‐wide register‐based study

Abstract Purpose This study aimed to describe and compare refill adherence and persistence to lipid‐lowering medicines in patients with type 2 diabetes by previous cardiovascular disease (CVD). Methods We followed 97 595 patients (58% men; 23% with previous CVD) who were 18 years of age or older when initiating lipid‐lowering medicines in 2007–2010 until first fill of multi‐dose dispensed medicines, death, or 3 years. Using personal identity numbers, we linked individuals' data from the Swedish Prescribed Drug Register, the Swedish National Diabetes Register, the National Patient Register, the Cause of Death Register, and the Longitudinal Integration Database for Health Insurance and Labour Market Studies. We assessed refill adherence using the medication possession ratio (MPR) and the maximum gap method, and measured persistence from initiation to discontinuation of treatment or until 3 years after initiation. We analyzed differences in refill adherence and persistence by previous CVD in multiple regression models, adjusted for socioeconomic status, concurrent medicines, and clinical characteristics. Results The mean age of the study population was 64 years, 80% were born in Sweden, and 56% filled prescriptions for diabetes medicines. Mean MPR was 71%, 39% were adherent according to the maximum gap method, and mean persistence was 758 days. Patients with previous CVD showed higher MPR (3%) and lower risk for discontinuing treatment (12%) compared with patients without previous CVD (P < 0.0001). Conclusions Patients with previous CVD were more likely to be adherent to treatment and had lower risk for discontinuation compared with patients without previous CVD.


| Study population
In the Swedish Prescribed Drug Register (SPDR), we identified patients aged ≥18 years and registered in the National Diabetes Register (NDR) with type 2 diabetes, who initiated use of lipid-lowering medicines between 1 January 2007 and 31 December 2010 (the index period).
Our study distinguished between NDR-registered patients with type 1 and type 2 diabetes by applying the epidemiological definition of type 2 diabetes. Such individuals receive treatment with diet and/or other glucose-lowering medicines than insulin, or experience onset of diabetes at age ≥40 years and receive insulin treatment and/or other glucose-lowering medicines. [16][17][18][19][20] To identify incident users of lipid-lowering medicines, we established a washout period encompassing the 366 days preceding the first day of filled prescription (the index date). Our study excluded patients who (1) filled either extemporaneously prepared prescriptions for lipid-lowering medicines that lacked information about package size, or bile acid sequestrants more frequently prescribed for indications other than hyperlipidemia; 21 or (2) used a combination of different lipid-lowering substances or different strengths of the same substance ( Figure 1). Combination therapy comprised prescriptions for (1) more than 1 substance or multiple strengths of the same substance filled on the same date, or (2) a previously filled substance/strength that was filled again within 45 days after finishing the previous supply of that substance/strength and filling another substance/strength during the gap. Multiple lipid-lowering substances in the same unit (eg, tablet) were considered monotherapy. We followed all patients until the first fill date of multi-dose dispensed medicines (because these were automatically dispensed even if the patient never redeemed the medicines), death or 3 years after the index date, whichever occurred first.

| Data sources
Patients' unique Swedish personal identity number allowed us to link information from the SPDR, the NDR, the National Patient Register, and the Cause of Death Register (all administered by the Swedish National Board of Health and Welfare) as well as the Longitudinal Integration Database for Health Insurance and Labour Market Studies (LISA) (managed by Statistics Sweden). The regional ethics review board approved our study .
Filled prescriptions were collected from the SPDR, which individualizes its data on all prescriptions filled since 1 July 2005, 22 including information about age, sex, type of medicine, package size, date of dispensing, and free text dosage instructions from the prescriber. We gathered clinical data from the NDR, which has maintained nation-wide records on diabetes care in Sweden since 1996. 23 We obtained individual data on socioeconomic status from the LISA database 24 and received data on CVD and cancer diagnoses fromThe National Patient Register. 25 The Cause of Death Register provided data on date of death. 26

| Estimation of days' supply and overall observation period
We used SPDR data to estimate the number of days with medicines on hand and the overall observation period. To

| Refill adherence and persistence
We based our adherence estimates on the medication possession ratio (MPR) 9,10 and the maximum gap method. [27][28][29] To calculate MPR, which represents the proportion of days with medicines on hand during the observation period, we divided the total days' supply by the total number of observation days. In this study, MPR is a continuous variable. To compare our results with studies that categorize MPR exceeding 80% as adherent behavior, we divided MPR into quintiles. 11 The maximum gap method identifies gaps between filled prescriptions, allowing patients to be without medicines on hand for a predefined time period without defining them as non-adherent. 11 We defined a gap as ≥45 days between 2 prescriptions. Thus, patients with no gaps were adherent. We based the cutoff for gap length on the Swedish reimbursement system, which allows patients to fill a maximum of 3-month's supply per refill, 30 the most common practice for lipid-lowering medicines. We also estimated the mean number of gaps and mean number of days within gaps for non-adherent patients.
We defined persistence as the duration between initiation and discontinuation of treatment 9,11 and discontinuation as a gap of ≥180 days between 2 filled prescriptions (representing 2 refills within the reimbursement scheme). The discontinuation date was the last day with medicines on hand before the first discontinuation gap. We estimated persistence from the index date to the discontinuation date or the end of the observation period, whichever occurred first. To estimate the annual discontinuation rate, we divided the number of patients who discontinued treatment during each year by the total number of patients who were persistent at the start of each year.

| Sensitivity analyses
Using the maximum gap method, we estimated the stability of the 45-day cutoff by assessing 2 alternative gap lengths (30 and 90 days). Furthermore, Censored patients did not have the same possibility to fulfill the 180-day discontinuation gap cutoff; thus, to estimate the impact of immortal time bias, we assessed alternative lengths of the discontinuation gap cutoff (90 and 135 days) in patients who were censored during the observation period.

| Covariates
Potential confounders of refill adherence and discontinuation of treatment included (1)

| Statistical analyses
We analyzed differences in refill adherence and persistence according to previous CVD (no previous CVD considered the reference) in 3 multivariable regression models based on the potential confounders' character.  Table 3.
We used multiple linear regression to analyze differences in MPR, and Cox proportional hazard regression and Kaplan-Meier to analyze differences in discontinuation. Difference in gap occurrence were analyzed with logistic regression adjusted for all potential confounders in the fully adjusted model.
Data management and statistical analyses were performed using SAS Software Version 9.4 (SAS Institute, Cary NC).

| Refill adherence
Mean MPR in the total study population was 70.9% (Table 2). 76.3% for patients with previous CVD, and 69.3% for patients without previous CVD. Adjusted for potential confounders, the difference in

| Persistence
In the total study population, 72.5% of patients were persistent for at least 1 year and 56.3% were persistent for 3 years (

P-value
Insulin and other glucoselowering medicines   with previous CVD, censoring resulting from filled prescriptions for multi-dose dispensed medicines and death was 9.2% and 8.4%, respectively, compared with 2.1% and 2.4%, respectively, in patients without previous CVD.
Adjusted for potential confounders ( CVD compared with patients without previous CVD (Figure 2). of the population remained persistent after 10 years of follow-up.

| Sensitivity analysis
Altogether, these findings suggest that the first year with lipid-lowering medicines is crucial for the continuation of treatment, providing valuable information for health care providers to consider when treating patients with moderate to high CVD risk.
Additionally, patients who filled prescriptions for other glucoselowering and/or antihypertensive medicines had a higher refill adherence and longer persistence to lipid-lowering medicines. This is consistent with an earlier study that showed higher adherence to statins with increasing number of concurrent medicines (to a certain threshold). 37 That finding is positive because the treatment approach for diabetes may involve multiple medicines. Furthermore, patients born in Africa or the Americas were significantly less adherent and more likely to discontinue treatment. Although these patients represent only 2% of the total study population, the difference in adherence and persistence compared with patients born in Sweden could be due to misunderstandings and language difficulties between patient and health care provider 38,39 that will require early discussion between patient and provider to facilitate adequate use of medicines, including sufficient adherence to treatment.

| Strengths and limitations
The most important strengths of this study are its national coverage of patients and its use of data from national registers. Such data provide a great advantage for studying refill adherence instead of prescription adherence that lacks information about whether prescriptions are filled. However, we cannot assure that our patients actually ingested the filled medicines. Another limitation is the reduction of patients included in the fully adjusted model. NDR coverage of patients was 50%-80% in 2007-2010, thus limiting the availability of data on clinical characteristics because not all covariates were measured annually.
Nevertheless, our fully adjusted model included nearly 20 000 patients.
Altogether 9 out of 10 users of lipid-lowering medicines with type 2 diabetes were included in the final study population after applying the exclusion criteria. The study population was limited to patients with type 2 diabetes who filled prescriptions for lipid-lowering medicines in monotherapy. Thus, patients using combination of lipid-lowering medication were excluded, corresponding to 3.7% of all new users of lipid-lowering medicines with type 2 diabetes. Combination of lipidlowering substances is recommended only in patients with several risk factors of CVD and who do not reach target valued for LDL-cholesterol with monotherapy. We also excluded patients with multi-dose dispensed medicines which may limit the generalizability of our findings somewhat in the oldest part of the population as the majority of multi-dose dispensed medicines is issued to patients ≥65 years of age. Thus, our study population might be younger and have less comorbidity than the general population of type 2 diabetes patients in Sweden.

| CONCLUSIONS
In this nation-wide study of patients with type 2 diabetes, refill adherence to lipid-lowering medicines was 71%. Almost 75% of patients were persistent for 1 year, and 56% were persistent 3 years. Patients with previous CVD showed higher refill adherence and longer persistence compared with patients without previous CVD. This information is valuable and important for consideration by health care providers who treat patients with type 2 diabetes and moderate to high risk for CVD.

ETHICS STATEMENT
The regional Ethical Review Board at the University of Gothenburg approved the study (#563-12).

ACKNOWLEDGEMENT
This study was funded by a grant from the Swedish Research Council for Health, Working Life and Welfare (#2013-0521). The authors would like to thank Karen Williams for the linguistic revision.

CONFLICT OF INTEREST STATEMENT
Karolina Andersson Sundell is employed by AstraZeneca. However, the views expressed in this study are her own and not those of AstraZeneca. The remaining authors declare no conflict of interest.