Adherence to and persistence with antidiabetic medications and associations with clinical and economic outcomes in people with type 2 diabetes mellitus: A systematic literature review

Abstract

We designed a systematic literature review to identify available evidence on adherence to and persistence with antidiabetic medication in people with type 2 diabetes (T2D). Electronic screening and congress searches identified real‐world noninterventional studies (published between 2010 and October 2020) reporting estimates of adherence to and persistence with antidiabetic medication in adults with T2D, and associations with glycaemic control, microvascular and/or macrovascular complications, hospitalizations and healthcare costs. Ninety‐two relevant studies were identified, the majority of which were retrospective and reported US data. The proportions of patients considered adherent (median [range] 51.2% [9.4%‐84.3%]) or persistent (median [range] 47.7% [16.9%‐94.0%]) varied widely across studies. Multiple studies reported an association between greater adherence/persistence and greater reductions in glycated haemoglobin levels. Better adherence/persistence was associated with fewer microvascular and/or macrovascular outcomes, although there was little consistency across studies in terms of which outcomes were improved. More adherent and more persistent patients were typically less likely to be hospitalized or to have emergency department visits/admissions and spent fewer days in hospital annually than less adherent/persistent patients. Greater adherence and persistence were generally associated with lower hospitalization costs, higher pharmacy costs and lower or budget‐neutral total healthcare costs compared with lower adherence/persistence. In conclusion, better adherence and persistence in people with T2D is associated with lower rates of microvascular and/or macrovascular outcomes and inpatient hospitalization, and lower or budget‐neutral total healthcare expenditure. Education and treatment strategies to address suboptimal adherence and persistence are needed to improve clinical and economic outcomes.

1. INTRODUCTION

Type 2 diabetes (T2D) is a chronic, progressive disease that has a substantial clinical impact on patients as well as imposing an economic burden on healthcare systems. ¹ T2D is associated with cardiovascular, renal, retinal and neurological complications, and it has been estimated that 50% of people with T2D have early signs of these conditions at diagnosis. ² Complications account for a considerable proportion of the lifetime costs of treating diabetes, ³ and are also linked to reduced health‐related quality of life ⁴ and increased indirect costs from lost workplace productivity. ⁵ The risk of T2D complications is higher in patients with poor glycaemic control. ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹² Various demographic, social, and patient‐ and physician‐related factors contribute to the likelihood of people with T2D achieving glycaemic control, ¹³ , ¹⁴ including the extent to which patients are adherent to and persistent with antidiabetic medication. ¹⁵ , ¹⁶ Although there is evidence that adherence and persistence are associated with improved outcomes, medication‐taking behaviour is not usually considered by decision‐makers and payers alongside clinical benefits and health utility gains when evaluating T2D treatments.

Patients are closely monitored in clinical trials, therefore adherence and persistence during these trials is not representative of medication‐taking behaviour in real‐world settings. ¹⁷ Observational studies must thus be used to estimate adherence and persistence rates and to evaluate the link between medication‐taking behaviour and clinical or economic outcomes. As there is a considerable volume of real‐world evidence in T2D, a systematic review of the literature is a robust way to identify such studies and collate their results. An earlier systematic literature review (SLR) of articles published from 2007 to 2014 found that higher rates of adherence to antidiabetic medication were associated with not only better glycaemic control and fewer hospitalizations but also lower healthcare costs. ¹⁸ During the past few years, however, the use of newer drug classes such as dipeptidyl peptidase‐4 inhibitors, glucagon‐like peptide‐1 receptor agonists (GLP‐1RAs) and sodium‐glucose cotransporter‐2 inhibitors has increased. ¹⁹ Consequently, a new review of the literature is warranted to examine medication‐taking behaviour and its link to outcomes across the current spectrum of available antidiabetic medications.

The present SLR was designed to identify relevant evidence on the patterns of adherence to and persistence with antidiabetic medication in people with T2D, as well as clinical and economic outcomes linked to adherence and persistence, over the period of 2010 to 2020.

2. MATERIALS AND METHODS

2.1. Systematic literature review

Electronic searches were designed to identify real‐world noninterventional studies reporting estimates of adherence to and persistence with antidiabetic medications in people with T2D and associations with clinical and economic outcomes. Journal publications from January 2010 to October 2020 were included in electronic searches, and relevant congress publications from January 2018 to October 2020 were also identified. Systematic searches were conducted in October 2020 in the Medical Literature Analysis and Retrieval System Online (MEDLINE), Embase, Evidence‐Based Medicine Reviews and EconLit. Any databases that were not up to date were also searched via the University of York Centres for Reviews and Dissemination website. The search strategy for MEDLINE is shown in Table S1. The congresses searched are listed in Table S2.

Titles and abstracts were screened in a double‐blind manner by two independent reviewers to determine whether they met the eligibility criteria for inclusion (Table 1). Any disagreements between reviewers were referred to a third reviewer and resolved by consensus. The reference lists of included studies and relevant reviews/editorials were reviewed to identify any further eligible publications that had not been detected in the database searches. All publications meeting the criteria were obtained as full articles and reassessed, and relevant data from publications included after full‐text review were entered into a data extraction table. Quality assessment was carried out on the studies using the critical appraisal tools from the Joanna Briggs Institute. ²⁰

TABLE 1.

Eligibility criteria

Criteria	Include	Exclude
Population	People with T2D, regardless of age or disease severity	Animal/in vitro studies
		T1D
		Gestational diabetes
		Mixed populations where results for T2D are not reported separately
Intervention	Pharmacological antidiabetic medications	Nonpharmacological interventions (eg, diet‐based interventions, lifestyle changes, guidelines, digital apps)
Outcomes	Estimates of persistency and adherence to antidiabetic medications (including definition of adherence, persistence, discontinuations)	Outcomes not listed in “include” column
	Links between persistency and adherence to clinical and economic outcomes, specifically: Healthcare costs (eg, total costs of care) Hospitalizations HbA1c (glycaemic control) Macrovascular/microvascular short‐term and long‐term outcomes
	Drivers of persistence and adherence (eg, mode of treatment administration, dosing regimen, patient characteristics)
	Statistical/analytical methods
	Data sources
Study design	Studies which have utilized real‐world data to investigate outcomes of interest, including:	Reviews/editorials
	Cohort studies (prospective/retrospective)	RCTs
	Case‐control studies	Nonrandomized experimental studies
	Before‐and‐after studies (observational)	Cross‐sectional studies
	Correlation studies	Case reports/case series
	Longitudinal studies	Animal/in vitro studies
Geography	No restriction	‐
Publication date	Full publications: 2010 onwards (last 10 years) Conference abstracts: 2017 onwards (last 3 years)	Full publications: pre‐2010 Conference abstracts: pre‐2017
Language	English‐language publications or non‐English‐language publications with an English abstract	Non‐English‐language publications without an English abstract

Abbreviations: HbA1c, glycated haemoglobin; RCT, randomized controlled trial; T1D, type 1 diabetes; T2D, type 2 diabetes.

2.2. Outcomes

Relevant outcomes in the SLR were estimates of adherence to and persistence with antidiabetic medications and their associations with glycaemic control, microvascular and macrovascular outcomes, hospitalizations and healthcare costs. We defined adherence as the extent to which a person's antidiabetic medication‐taking behaviour corresponds with recommendations from their healthcare provider. In the studies identified, it was most often measured as proportion of days covered (PDC) by medication, or medication possession ratio (MPR) as detailed below. Persistence, the duration of antidiabetic medication use by a patient, was usually measured as the proportion of patients who remained on treatment for a specified period or as the mean number of days to treatment discontinuation within the observation period.

3. RESULTS

3.1. Search results

In total, 3227 references were included for screening, of which 508 were determined to be relevant for full‐text review (see Figure S1 for the Preferred Reporting Items for Systematic Reviews and Meta‐Analyses [PRISMA] diagram). In total, 255 publications met the inclusion criteria at full‐text review, and an additional eight publications were identified by hand searches.

Only full publications reporting data on the associations between adherence/persistence and the clinical/economic outcomes of interest were prioritized for data extraction and are the focus of this manuscript (n = 92). ²¹ , ²² , ²³ , ²⁴ , ²⁵ , ²⁶ , ²⁷ , ²⁸ , ²⁹ , ³⁰ , ³¹ , ³² , ³³ , ³⁴ , ³⁵ , ³⁶ , ³⁷ , ³⁸ , ³⁹ , ⁴⁰ , ⁴¹ , ⁴² , ⁴³ , ⁴⁴ , ⁴⁵ , ⁴⁶ , ⁴⁷ , ⁴⁸ , ⁴⁹ , ⁵⁰ , ⁵¹ , ⁵² , ⁵³ , ⁵⁴ , ⁵⁵ , ⁵⁶ , ⁵⁷ , ⁵⁸ , ⁵⁹ , ⁶⁰ , ⁶¹ , ⁶² , ⁶³ , ⁶⁴ , ⁶⁵ , ⁶⁶ , ⁶⁷ , ⁶⁸ , ⁶⁹ , ⁷⁰ , ⁷¹ , ⁷² , ⁷³ , ⁷⁴ , ⁷⁵ , ⁷⁶ , ⁷⁷ , ⁷⁸ , ⁷⁹ , ⁸⁰ , ⁸¹ , ⁸² , ⁸³ , ⁸⁴ , ⁸⁵ , ⁸⁶ , ⁸⁷ , ⁸⁸ , ⁸⁹ , ⁹⁰ , ⁹¹ , ⁹² , ⁹³ , ⁹⁴ , ⁹⁵ , ⁹⁶ , ⁹⁷ , ⁹⁸ , ⁹⁹ , ¹⁰⁰ , ¹⁰¹ , ¹⁰² , ¹⁰³ , ¹⁰⁴ , ¹⁰⁵ , ¹⁰⁶ , ¹⁰⁷ , ¹⁰⁸ , ¹⁰⁹ , ¹¹⁰ , ¹¹¹ , ¹¹² The remaining 171 publications (conference abstracts [n = 47] and full publications reporting estimates of adherence/persistence but not associations with clinical/economic outcomes [n = 124]) were excluded.

3.2. Study characteristics

Of the 92 full publications included, ²¹ , ²² , ²³ , ²⁴ , ²⁵ , ²⁶ , ²⁷ , ²⁸ , ²⁹ , ³⁰ , ³¹ , ³² , ³³ , ³⁴ , ³⁵ , ³⁶ , ³⁷ , ³⁸ , ³⁹ , ⁴⁰ , ⁴¹ , ⁴² , ⁴³ , ⁴⁴ , ⁴⁵ , ⁴⁶ , ⁴⁷ , ⁴⁸ , ⁴⁹ , ⁵⁰ , ⁵¹ , ⁵² , ⁵³ , ⁵⁴ , ⁵⁵ , ⁵⁶ , ⁵⁷ , ⁵⁸ , ⁵⁹ , ⁶⁰ , ⁶¹ , ⁶² , ⁶³ , ⁶⁴ , ⁶⁵ , ⁶⁶ , ⁶⁷ , ⁶⁸ , ⁶⁹ , ⁷⁰ , ⁷¹ , ⁷² , ⁷³ , ⁷⁴ , ⁷⁵ , ⁷⁶ , ⁷⁷ , ⁷⁸ , ⁷⁹ , ⁸⁰ , ⁸¹ , ⁸² , ⁸³ , ⁸⁴ , ⁸⁵ , ⁸⁶ , ⁸⁷ , ⁸⁸ , ⁸⁹ , ⁹⁰ , ⁹¹ , ⁹² , ⁹³ , ⁹⁴ , ⁹⁵ , ⁹⁶ , ⁹⁷ , ⁹⁸ , ⁹⁹ , ¹⁰⁰ , ¹⁰¹ , ¹⁰² , ¹⁰³ , ¹⁰⁴ , ¹⁰⁵ , ¹⁰⁶ , ¹⁰⁷ , ¹⁰⁸ , ¹⁰⁹ , ¹¹⁰ , ¹¹¹ , ¹¹² 39 studies were published from 2010 to 2015 and 53 from 2016 to 2020. Most of the studies were retrospective observational cohort studies or database analyses; only 13 of the 92 studies (14%) were prospective. ²¹ , ²³ , ³³ , ⁴⁵ , ⁵⁹ , ⁶⁷ , ⁷⁰ , ⁸⁷ , ⁹² , ⁹³ , ⁹⁷ , ¹⁰² , ¹¹¹ Sixty of the 92 studies (65%) were from the United States. The remaining studies were from Europe (Germany, ²⁴ , ⁹³ Italy, ³⁵ , ⁸¹ Spain, ⁴⁷ Sweden ¹¹² and Switzerland ⁵⁷ ; 7% of studies in total), Asia (China, ²⁹ India, ⁹⁷ Iran, ⁴⁵ Israel, ⁷⁴ , ⁷⁵ Pakistan ⁹² and Taiwan ¹ , ³³ ; 9% of studies in total), Canada (2%) ⁴⁹ , ¹¹¹ and Australia (1%). ⁶⁴ Three studies were multinational (3%). ⁷⁰ , ¹⁰² , ¹⁰⁶ Some studies had additional patient inclusion criteria such as a focus on older ²⁸ , ⁷⁷ , ⁸⁵ , ⁹⁶ or younger patients ²⁷ or US veterans. ⁴¹ , ⁴² , ⁴³ , ⁶² , ⁷⁸ Table S3 shows the number of studies reporting each outcome.

3.3. Estimates of adherence

In total, 71 studies included estimates of adherence (cited in Table S3). There was variation in how adherence was defined: 30 studies used PDC ²² , ²⁸ , ²⁹ , ³⁰ , ³¹ , ³⁴ , ³⁶ , ³⁷ , ³⁸ , ³⁹ , ⁴⁸ , ⁵⁰ , ⁵⁶ , ⁵⁷ , ⁶⁰ , ⁶³ , ⁷¹ , ⁷⁵ , ⁷⁶ , ⁷⁸ , ⁷⁹ , ⁸⁰ , ⁸² , ⁸³ , ⁸⁵ , ⁸⁶ , ⁸⁸ , ⁹⁸ , ¹⁰⁰ , ¹¹⁰ and 26 studies used the MPR. ²³ , ²⁹ , ³² , ³³ , ³⁵ , ⁴¹ , ⁴² , ⁴⁴ , ⁴⁹ , ⁵¹ , ⁵³ , ⁵⁴ , ⁵⁵ , ⁶⁴ , ⁶⁸ , ⁷³ , ⁷⁷ , ⁹⁰ , ⁹¹ , ⁹⁵ , ¹⁰¹ , ¹⁰³ , ¹⁰⁴ , ¹⁰⁶ , ¹⁰⁸ , ¹⁰⁹ MPR is calculated as the total days' supply of treatment in the defined period, divided by the total number of days in the defined period. It is easy to estimate but can lead to overestimation of adherence if patients refill prescriptions early. ¹¹³ PDC is the most commonly used indirect measure of adherence for chronic diseases. ¹¹³ It is a more conservative measure than MPR as it accounts for overlapping days between prescriptions, so it cannot exceed 100% and therefore avoids falsely inflating mean population adherence. A PDC or MPR >80% was the threshold most frequently used to define adherence. Sixteen studies used other methods to define adherence, including self‐reported measures and pill counts. ²¹ , ²⁶ , ⁴⁵ , ⁴⁶ , ⁴⁷ , ⁵⁹ , ⁶⁵ , ⁶⁷ , ⁷⁰ , ⁸⁴ , ⁸⁷ , ⁹² , ⁹³ , ⁹⁴ , ⁹⁷ , ¹¹¹

Study follow‐up duration ranged from 3 months to 10 years; 22 of the 71 studies had a 12‐month follow‐up period. Overall, 28 studies examined adherence to oral antidiabetic medications (OADs), ²³ , ²⁹ , ³³ , ⁴⁴ , ⁴⁵ , ⁴⁸ , ⁴⁹ , ⁵¹ , ⁵³ , ⁵⁴ , ⁵⁵ , ⁵⁶ , ⁵⁷ , ⁵⁹ , ⁶⁰ , ⁶³ , ⁷¹ , ⁷⁸ , ⁸² , ⁸⁴ , ⁸⁵ , ⁸⁶ , ⁹⁰ , ⁹¹ , ⁹² , ⁹⁴ , ⁹⁸ , ¹¹⁰ eight studies examined insulin, ²⁶ , ³² , ³⁹ , ⁷⁰ , ⁷⁷ , ⁸⁸ , ¹⁰⁴ , ¹⁰⁹ seven studies examined injectable GLP‐1RAs ³⁰ , ³⁷ , ³⁸ , ⁷⁹ , ⁸⁰ , ⁸³ , ¹⁰⁶ and one study examined combination therapies. ⁷⁵ Twenty studies reported adherence estimates for multiple classes (OADs, insulins and/or GLP‐1RAs). ²² , ²⁸ , ³¹ , ³⁴ , ³⁵ , ³⁶ , ⁴¹ , ⁴² , ⁵⁰ , ⁶⁴ , ⁶⁸ , ⁷³ , ⁷⁶ , ⁹⁵ , ¹⁰¹ , ¹⁰³ , ¹⁰⁸ Seven studies did not report which antidiabetic medication(s) were included. ²¹ , ⁴⁶ , ⁶⁷ , ⁸⁷ , ⁹³ , ⁹⁷ , ¹⁰⁰

Estimates of adherence varied considerably across studies, from 9.4% to 84.3%, across all medication classes. In general, reported rates of adherence to antidiabetic medications were relatively low: the median adherence across all studies was 51.2%. There was no clear consensus across studies regarding which medication classes were associated with higher adherence.

3.4. Estimates of persistence

Estimates of persistence were reported in 31 studies (Table S3). In most studies, persistence was estimated based on the fill time between prescriptions or medication insurance claims. A gap in medication of ≥90 days was used to define discontinuation of medication (non‐persistence) in nine studies, ²⁴ , ²⁹ , ³⁰ , ⁴⁰ , ⁵⁸ , ⁶⁹ , ⁹⁶ , ¹⁰⁶ , ¹⁰⁹ whereas thresholds of ≥30, ²⁵ , ³¹ , ⁵² , ⁸⁹ 45, ⁷⁹ , ⁸⁰ 60 ⁶⁰ , ⁶¹ , ⁹⁰ , ¹⁰⁷ or 120 days ⁷⁴ were used in other studies. Thirteen studies used other definitions for treatment gaps indicating discontinuation, such as a gap exceeding the 90th percentile of the mean duration of prescription fills. ⁷⁷ , ¹⁰⁴

All 31 studies were retrospective. Study duration ranged from 6 months ⁷⁵ , ¹⁰⁷ to 3 years, ⁹⁴ and was 12 months in 17 studies. ²⁴ , ²⁵ , ²⁹ , ³⁰ , ³¹ , ⁴⁰ , ⁵² , ⁵⁸ , ⁶⁰ , ⁶² , ⁶⁹ , ⁷⁷ , ⁸⁰ , ⁹⁰ , ⁹⁶ , ¹⁰⁴ Overall, six studies examined persistence with OADs, ²⁹ , ⁵⁸ , ⁶⁰ , ⁶² , ⁹⁰ , ⁹⁴ 11 studies with insulin, ²⁴ , ²⁵ , ⁴⁰ , ⁵² , ⁶¹ , ⁷⁷ , ⁸⁹ , ⁹⁶ , ¹⁰⁴ , ¹⁰⁷ , ¹⁰⁹ seven studies with injectable GLP‐1RAs ³⁰ , ³⁸ , ⁷⁴ , ⁷⁹ , ⁸⁰ , ⁸¹ , ¹⁰⁶ and four studies with combination therapies. ⁶⁹ , ⁷⁵ , ¹⁰⁵ , ¹¹² Three studies reported persistence estimates for multiple classes of antidiabetic medication. ²⁷ , ³¹ , ⁶⁶

As was the case for adherence, persistence estimates varied widely among studies. The proportion of persistent patients ranged from 16.9% to 94.0% (median: 47.7%) in the studies reporting persistence with the most frequently studied classes of antidiabetic medication (OADs, insulins and GLP‐1RAs).

3.5. Associations between adherence/persistence and clinical and economic outcomes

3.5.1. Glycaemic control

The specific glycaemic outcomes assessed in the studies examining adherence and persistence were overall change in glycated haemoglobin (HbA1c) level, expressed as a percentage, the proportion of patients achieving a target HbA1c or the incidence of hypoglycaemia.

An association between medication adherence and glycaemic control was reported in 42 studies (Table S3), 30 of which investigated OADs or multiple classes of antidiabetic medications. Better adherence to antidiabetic medication was generally associated with improved glycaemic control. A significantly greater decrease in HbA1c, or a lower HbA1c at follow‐up, in more adherent versus less adherent patients was reported by most studies investigating this outcome. ²² , ²⁹ , ³⁰ , ⁴⁴ , ⁴⁷ , ⁷³ , ⁷⁹ , ⁸⁵ In the remaining studies, the HbA1c reduction was nonsignificantly lower in the more adherent patients ⁹⁴ or similar in the two groups. ⁷⁸ The studies also consistently reported a significantly higher likelihood of more adherent patients achieving specific HbA1c targets, such as a ≥1.0% reduction ²⁹ or reduction to <7%, ⁴⁵ than less adherent patients. ²⁹ , ⁴⁵ , ⁵⁶ , ⁷⁸ , ⁸² , ⁸⁴ , ⁸⁵ , ⁹⁰ , ⁹⁴

Four studies investigated the association between adherence to antidiabetic medication and risk of hypoglycaemia. Two of these studies investigated OADs (including sulphonylureas) and found no significant association, ⁷⁸ , ⁹¹ and two studies investigating multiple antidiabetic medication classes found significantly lower rates of acute complications, including hypoglycaemia, in more versus less adherent patients. ²⁸ , ³⁴

Fewer studies (n = 20) investigated the association between persistence and glycaemic control, including six studies examining injectable GLP‐1RAs, ³⁰ , ³⁸ , ⁷⁴ , ⁷⁹ , ⁸¹ , ¹⁰⁶ six examining combination therapies ²⁷ , ⁶⁶ , ⁶⁹ , ⁷⁵ , ¹⁰⁵ , ¹¹² and four each examining insulin ⁴⁰ , ⁹⁶ , ¹⁰⁷ , ¹⁰⁹ and OADs. ⁵⁸ , ⁶⁰ , ⁶² , ⁹⁴ Persistence with medication was also generally associated with better glycaemic control. All but one ¹⁰⁶ of the 14 studies examining GLP‐1RAs, insulin or OADs reported a greater reduction in HbA1c, a greater proportion of patients achieving a target HbA1c or a trend for better outcomes in persistent than in nonpersistent patients. ³⁰ , ³⁸ , ⁴⁰ , ⁵⁸ , ⁶⁰ , ⁶² , ⁷⁴ , ⁷⁹ , ⁸¹ , ⁹⁴ , ⁹⁶ , ¹⁰⁷ , ¹⁰⁹ However, the results were more heterogeneous in the studies investigating combination therapies, ²⁷ , ⁶⁶ , ⁶⁹ , ⁷⁵ , ¹⁰⁵ , ¹¹² with only two studies clearly demonstrating superior glycaemic control in persistent compared with nonpersistent patients. ⁶⁹ , ¹¹² Lin et al, ⁶⁹ in a study of patients receiving GLP‐1RAs and basal insulin, also reported that medication persistence was linked to lower rates of hypoglycaemia. Figure 1 summarizes the findings from studies investigating the association between adherence or persistence and change in HbA1c.

FIGURE 1.

Change in HbA1c level from baseline by A, adherence to and B, persistence with antidiabetic medication in studies reporting this outcome. Bars with an asterisk indicate statistically significant results for adherent/persistent versus nonadherent/nonpersistent patients (P < 0.05). Min et al ⁷⁸ and Reynolds et al ⁹⁴ did not report P values for adherent versus nonadherent patients. Eliasson et al ¹¹² and Melzer‐Cohen et al ⁷⁵ did not report P values for persistent versus nonpersistent patients. For adherence, studies were included if they reported change in HbA1c (follow‐up times varied, as indicated for each study); for persistence, studies were included if they reported change in HbA1c over 6, 12 or 24 months. ^aData shown are means, except for Min et al ⁷⁸ which are median. HbA1c, glycated haemoglobin

3.5.2. Microvascular and macrovascular outcomes

Nine studies investigated the associations between adherence to antidiabetic medications and microvascular and/or macrovascular outcomes. ²² , ⁴⁶ , ⁴⁸ , ⁵⁰ , ⁶³ , ⁹⁷ , ⁹⁸ , ⁹⁹ , ¹⁰⁸ Five of these examined OADs ⁴⁸ , ⁵⁰ , ⁶³ , ⁹⁸ , ⁹⁹ ; medication class was mixed or not specified in the remaining studies. ²² , ⁴⁶ , ⁹⁷ , ¹⁰⁸ Most studies were retrospective, and four of these studies had >54 000 participants. ⁴⁸ , ⁵⁰ , ⁶³ , ⁹⁹ The two prospective studies were small, with <250 patients each. ⁹⁷ , ⁹⁸ Follow‐up ranged from 3 months ⁹⁷ to 10 years, ⁶³ and was ≥5 years in four studies. ⁴⁶ , ⁴⁸ , ⁶³ , ¹⁰⁸

The microvascular outcomes examined included peripheral vascular disease, retinopathy, nephropathy, renal events, neuropathy and amputations/ulcers. The macrovascular outcomes included cerebrovascular disease, stroke, transient ischaemic attack, ischaemic heart disease, myocardial infarction and angina (Table 2). Six studies examined both microvascular and macrovascular outcomes. ²² , ⁴⁶ , ⁴⁸ , ⁵⁰ , ⁹⁸ , ⁹⁹

TABLE 2.

Findings from studies reporting associations between adherence or persistence and microvascular and/or macrovascular outcomes ^a

					Statistically significant associations		No significant associations found
Study	Country, design	Treatment	Number of patients	Follow‐up	Microvascular outcomes	Macrovascular outcomes	Microvascular outcomes	Macrovascular outcomes
Adherence
An and Nichol 22	United States, retrospective	OADs (biguanides, SUs, TZDs and/or insulin‐sensitizing agents) and hypertension medication	2334	33 months	Any microvascular outcome (from renal failure and diabetic retinopathy)	Any macrovascular outcome (from MI and stroke)
Fukuda and Mizobe 46	Japan, retrospective	NR	11 331	96 months	Retinopathy Nephropathy Neuropathy			IHD Cerebrovascular disease Chronic arterial occlusion
Gatwood et al 48	United States, retrospective	OADs (not specified)	159 032	5 years	Retinopathy	Stroke MI	Neuropathy	TIA Angina
Gibson et al 50	United States, retrospective	SUs, meglitinides, biguanides, TZDs or AGIs	55 356 (OADs only)	18 months	Amputations/ulcers Renal events Neuropathy Retinopathy	MI	PVD	Cerebrovascular disease
		SUs, meglitinides, biguanides, TZDs or AGIs ± insulin	96 734 (OADs ± insulin)		Amputations/ulcers Renal events Neuropathy Retinopathy	MI Cerebrovascular disease	PVD
Kim et al 63	Korea, retrospective	Biguanide, SUs and others	65 067	10 years	NR	Cerebrovascular disease	NR	MI
Samu et al 97	India, prospective	NR	86	3 months	Neuropathy (diabetic foot)	NR
Sattler et al 98	United States, prospective	OADs (not specified)	243	12 months	Rate of “no (microvascular or macrovascular) complication” was higher in adherent than nonadherent patients		PVD Nephropathy Neuropathy	Cerebrovascular complications Cardiovascular complications
Simpson et al 99	United States, retrospective	OADs (not specified)	54 505	2.3 years (mean)		Any macrovascular outcome (from MI, stroke, heart failure, angina, CABG and angioplasty)	Any microvascular outcome (nephropathy, neuropathy, PVD or retinopathy)
Yu et al 108	United States, retrospective	Insulin and/or OADs	4708	12‐90 months	Any microvascular outcome (from diabetic foot, neuropathy, retinopathy and nephropathy)	NR	NR	NR
Persistence
Iglay et al 58	United States, retrospective	SUs (first, second and third generation)	104 082	12 months	PVD	Stroke TIA CHF MI IHD
Kalirai et al 61	United States, retrospective	Insulin detemir or insulin glargine	23 645	24 months	Nephropathy Neuropathy		Retinopathy	CVD Cerebrovascular disease

Abbreviations: AGI, alpha glucosidase inhibitor; CABG, coronary artery bypass graft; CHF, chronic heart failure; CVD, cardiovascular disease; IHD, ischaemic heart disease; MI, myocardial infarction; NR, not reported; OAD, oral antidiabetic drug; PVD, peripheral vascular disease; SU, sulphonylurea; TIA, transient ischaemic attack; TZD, thiazolidinedione.

^a Where there are discrepancies reported int results in the same article depending on the statistical model, the key findings as reported by authors in the article abstract are shown.

In general, adherence to antidiabetic medication was associated with lower rates of both microvascular and macrovascular complications compared with nonadherence, but there was substantial heterogeneity across the study results (Table 2). For example, the four largest studies, all of which assessed adherence to OADs, found that adherent patients have significantly lower rates of some outcomes but not others, with no consensus on which outcome category (microvascular or macrovascular) was significantly linked to adherence. ⁴⁸ , ⁵⁰ , ⁶³ , ⁹⁹ The medications that patients received were not fully reported in each study, and therefore no inferences between outcomes and medication class can be made using these data.

Two large US studies that examined the association between persistence and microvascular and macrovascular outcomes were identified (Table 2). Iglay et al ⁵⁸ included 104 082 patients followed up for 1 year and found lower rates in persistent versus nonpersistent patients for all cardiovascular, cerebrovascular and peripheral vascular outcomes examined. Kalirai et al ⁶¹ studied 23 645 patients and reported significantly lower rates of nephropathy and neuropathy after insulin initiation in persistent versus nonpersistent patients, but no significant difference in the rates of retinopathy, cardiovascular disease (CVD) or cerebrovascular disease. ⁶¹

3.5.3. Hospitalizations

The association between medication adherence and rates of hospitalization was reported in 18 studies, for patients receiving insulin, ²⁶ , ⁸⁸ OADs, ³³ , ⁴⁸ , ⁵⁴ , ⁵⁵ , ⁵⁷ , ⁷¹ , ⁹⁴ injectable GLP‐1RAs ⁸³ or multiple treatment classes, ²⁸ , ³⁴ , ³⁶ , ⁴⁷ , ⁵⁰ , ¹⁰¹ or with treatment not specified. ⁹³ , ¹⁰⁰ The specific outcomes investigated included hospital inpatient admissions, emergency department (ED) visits and admissions, and outpatient visits. Follow‐up duration ranged from 6 months to 7 years, with a median of 3 years. Most studies included >5000 patients; six studies had >90 000 patients (Table S4). ²⁸ , ³⁴ , ⁴⁸ , ⁵⁰ , ¹⁰⁰ , ¹⁰¹

Patients who were more adherent to antidiabetic medications were generally less likely to be hospitalized (and/or were less likely to have ED visits or admissions, or spent fewer days in hospital annually) than less adherent patients in the majority of studies across all treatment classes (Table S4). ²⁶ , ²⁸ , ³³ , ³⁴ , ⁴⁸ , ⁵⁰ , ⁵⁴ , ⁵⁷ , ⁷¹ , ⁸³ , ⁸⁸ , ⁹⁴ , ¹⁰⁰ , ¹⁰¹ The association between adherence and outpatient visits was more complex. Some studies found no significant association between adherence and outpatient visits, ⁴⁷ , ⁹³ whereas others found that adherence was linked to having more outpatient visits (Table S4). ³⁶ , ⁵⁰ , ⁵⁵ Among the latter studies, two were designed to investigate whether the rate of outpatient visits influenced adherence, and concluded that more frequent outpatient visits led to better adherence to antidiabetic medication; however, these studies did not report rates of hospitalization or other interactions with the healthcare system. ³⁶ , ⁵⁵ The third study reported that greater adherence was linked to more outpatient visits, but fewer ED visits and hospitalizations. ⁵⁰

Eleven studies reported data on the association between persistence and the rate of hospitalization (including ED visits and/or ED admissions) in patients receiving insulin, ⁴⁰ , ⁵² , ⁶¹ , ⁷⁷ , ⁸⁹ , ¹⁰⁴ , ¹⁰⁷ injectable GLP‐1RAs, ⁷⁴ , ¹⁰⁶ OADs ⁹⁴ or combination therapy. ⁶⁹ Follow‐up duration ranged from 6 months to 3 years, and sample sizes ranged from 534 to 23 645 patients (Table S5). Interruption or discontinuation of therapy was associated with an increased rate of hospitalization and longer hospital stays in most studies, ⁴⁰ , ⁵² , ⁶⁹ , ⁷⁷ , ⁸⁹ , ⁹⁴ , ¹⁰⁴ , ¹⁰⁷ but not in all studies. ⁷⁴ , ¹⁰⁶ However, four ⁵² , ⁶¹ , ⁶⁹ , ⁷⁷ of the five studies that reported data on outpatient visits found no significant association with persistence (Table S5).

3.5.4. Healthcare costs

Associations between adherence and healthcare costs in people with T2D were reported in 20 studies that investigated OADs, ²³ , ²⁹ , ³³ , ⁴⁹ , ⁵³ , ⁵⁴ insulin, ²⁶ , ³² , ³⁹ , ⁸⁸ injectable GLP‐1RAs ³⁰ or multiple treatment classes, ²⁸ , ³⁴ , ³⁵ , ⁴¹ , ⁵⁰ , ⁶⁸ , ⁷² , ⁷⁶ or that did not specify the treatment. ⁴⁶ Thirteen studies were from the United States, with the remaining studies from Canada, ⁴⁹ Italy, ³⁵ Japan, ⁴⁶ Korea ²³ , ⁵⁴ and Taiwan. ³³ , ⁶⁸ Seven studies had a follow‐up duration of ≤1 year and nine studies had ≥3 years' follow‐up. Sample size varied widely, from 301 to 0.74 million patients, with a median of 17 982 patients.

Despite substantial heterogeneity across studies, greater adherence was generally associated with lower inpatient admission costs but higher pharmacy costs. ²⁶ , ²⁸ , ²⁹ , ³⁰ , ³³ , ³⁴ , ³⁹ , ⁴¹ , ⁴⁹ , ⁷² , ⁷⁶ , ⁸⁸ Most studies found that total healthcare expenditure in adherent patients was lower than ²⁸ , ³² , ³⁹ , ⁴⁹ , ⁵³ , ⁵⁴ , ⁷² , ⁸⁸ or similar to ²³ , ²⁶ , ³⁴ , ⁷⁶ that in nonadherent patients (Figure 2), but in two studies adherence was associated with higher total costs than nonadherence. ³⁰ , ³³

FIGURE 2.

Healthcare costs by A, adherence to and B, persistence with antidiabetic medication in selected relevant studies. Bars with an asterisk indicate statistically significant results for adherent/persistent versus nonadherent/nonpersistent patients (P < 0.05). Buysman et al, ²⁹ Gentil et al ⁴⁹ and Hansen et al ⁵³ did not report P values for adherent versus nonadherent patients. Kalirai et al ⁶¹ and Lin et al ⁶⁹ did not report P values for persistent versus nonpersistent patients. ^aData from the subset of patients with depression/anxiety. ^bData from the subset of patients without depression/anxiety. ^cData from the first year of treatment. ^dData from the second year of treatment. CAD, Canadian dollar; ED, emergency department; FU, follow‐up; NR, not reported; USD, United States dollars

In total, 11 studies reported data on healthcare costs associated with persistence on insulin, ²⁴ , ²⁵ , ⁴⁰ , ⁵² , ⁶¹ , ⁷⁷ , ⁸⁹ , ⁹⁶ , ¹⁰⁷ injectable GLP‐1RAs ³⁰ and/or combination therapies. ⁶⁹ The study duration was relatively short for studying persistence: 6 months ¹⁰⁷ or 12 months ²⁴ , ²⁵ , ³⁰ , ⁴⁰ , ⁵² , ⁶⁹ , ⁷⁷ , ⁹⁶ in most studies and 24 months ⁶¹ , ⁸⁹ in two studies. Persistence was typically associated with higher pharmacy costs ²⁴ , ²⁵ , ²⁹ , ³⁰ , ⁴⁰ , ⁵² , ⁶¹ , ⁶⁹ , ⁷⁷ , ⁸⁹ , ¹⁰⁷ but lower healthcare costs, including acute care costs, inpatient and outpatient visits and ED visits. ²⁴ , ²⁵ , ⁵² , ⁶¹ , ⁶⁹ , ⁸⁹ Overall, total healthcare expenditure for persistent patients across studies was typically lower than or similar to that for nonpersistent patients (Figure 2).

4. DISCUSSION

Observational studies are recognized by payers and other stakeholders as an important means of obtaining data on medication‐taking behaviour, which cannot be assessed in clinical trials. ¹⁷ We reviewed the available evidence from observational studies on adherence to and persistence with antidiabetic medication in people with T2D, and how these relate to clinical and economic outcomes. Despite heterogeneity across studies in terms of antidiabetic medications used, length of follow‐up, geography, patients' clinical and demographic characteristics and the specific outcomes examined, some findings were consistent. Overall rates of adherence and persistence in people with T2D are suboptimal, as previously reported, ¹¹⁴ but better adherence and persistence are associated with clinical benefits, including improved glycaemic control, fewer hospitalizations and ED visits, and lower incidences of microvascular and macrovascular complications. Adherence and persistence were linked to lower rates of some microvascular and/or macrovascular outcomes but not others, which may be attributable in part to disparities in medications used, study setting and design. Overall, several outcomes that predict disability and absenteeism in people with T2D, including myocardial infarction, stroke, peripheral neuropathy, retinopathy and diabetic foot, ¹¹⁵ were associated with worse adherence and/or persistence in at least some of the studies identified in this review, highlighting the relevance of these outcomes in treatment decision‐making.

As a chronic condition affecting multiple organ systems, T2D is associated with substantial and rising healthcare costs. The International Diabetes Federation estimates that the worldwide health expenditure due to diabetes in adults has increased threefold in the past 15 years, from $232 billion in 2007 to $760 billion in 2019, of which 50% is attributable to managing diabetes complications ¹¹⁵ ; therefore, the influence of adherence and persistence on healthcare costs in T2D is a pertinent area for study. In the present SLR, we found an association between better adherence and persistence and either lower or similar total healthcare costs, compared with worse adherence and persistence. Cost estimates varied widely across the studies identified, which was likely to be owing to disparities in study variables and location. Generally, both adherence and persistence were associated with higher pharmacy costs that were offset by lower hospitalization costs, resulting in lower or budget‐neutral total healthcare expenditure for adherent/persistent patients. An association between adherence and reduced total health expenditure has been reported for several other chronic conditions. ¹¹⁶ Notably, more than two‐thirds of the studies reporting cost data were from North America, and most of the remainder were from Asia, with only one study from a European country (Italy). Further evidence is therefore needed, in particular from Europe, on the associations between adherence and persistence and healthcare costs.

This was a large SLR, including studies from all geographical regions and examining a broad range of outcomes. Although the outcomes examined are not independent from each other—for example, reduced rates of complications result in lower healthcare costs—this is nonetheless a comprehensive overview of the impact of adherence and persistence. Further SLRs could be used to capture additional outcomes linked to suboptimal adherence and persistence: previous studies have reported increased absenteeism, ¹¹⁷ more days of short‐term disability ¹¹⁷ and greater mortality. ⁶³ , ¹¹⁸ Although most studies reported the antidiabetic medication class used, discrepancies across studies in patient characteristics, study methodology and duration did not enable direct comparisons to be made between medication classes for any of the outcomes. Approximately half of the studies identified in the SLR were published between 2010 and 2015, meaning that more recently approved antidiabetic medications were not included in most of these studies. Further good‐quality observational studies are needed to systematically compare adherence and persistence across different drug classes and drugs with different modes and frequency of administration and different treatment benefits for complications such as CVD.

Several limitations of the studies identified in this review should be noted, particularly the fact that many were retrospective analyses using healthcare and claims databases. Such studies are inherently vulnerable to the effects of confounding, whereby certain factors, such as education, lifestyle and other sociodemographic variables, could influence both adherence to medication and health outcomes. Inertia in treatment decision‐making may also confound the relationship between medication‐taking behaviour and outcomes: adherence to or persistence with antidiabetic medication that has not been optimized is unlikely to be reflected in clinical benefit. Furthermore, PDC and MPR are useful proxies for adherence, but may not always accurately reflect actual medication‐taking behaviour. Finally, although comparing different antidiabetic medication classes is of great clinical interest, the substantial inter‐study heterogeneity in patient characteristics and methods used to estimate adherence/persistence in this SLR did not enable meaningful comparisons across drug classes.

The present SLR highlights the benefits that can be achieved by using therapeutic approaches that improve adherence and persistence as well as clinical outcomes. Across various diseases, higher compliance, a close correlate of adherence, has been reported for dosing regimens that require less frequent administration. ¹¹⁹ To illustrate, a study included in this SLR examining two populations receiving GLP‐1RAs in Germany and the United Kingdom found that twice‐daily exenatide was associated with a 30% to 40% greater likelihood of treatment discontinuation than once‐daily liraglutide. ¹⁰⁶ Furthermore, a recent meta‐analysis of seven studies investigating 75 159 people with T2D reported an 11% lower risk of nonadherence with once‐weekly versus once‐daily injectable GLP‐1RAs. ¹²⁰ Persistence is also favourably influenced by less frequent dosing regimens: in a real‐world, United States‐based study, the use of once‐weekly injectable GLP‐1RAs was associated with better persistence and adherence than daily regimens in propensity score‐matched cohorts. ¹²¹ Evidence from populations with other chronic diseases such as osteoporosis ¹²² and CVD ¹²³ also indicates that lower dosing frequency predicts better adherence and/or persistence. The use of medications early in the treatment pathway that are linked to symptomatic benefit in addition to adherence and persistence may provide an ongoing positive effect on health outcomes; however, to achieve sustained improvements, the use of treatment regimens that enhance adherence and persistence should also be considered as part of wider, holistic treatment strategies for people with T2D. As indicated by the studies identified in this review, many of which were carried out in primary care databases, routine management of patients with T2D is increasingly delivered in a primary care setting. ¹²⁴ Consequently, it is vital that primary care physicians receive education in strategies to maximize adherence and persistence, in communicating the benefits of this to patients, and in understanding and addressing reasons for poor adherence or persistence. Other approaches to maximizing the likelihood of adherence and persistence that may be applicable in primary care include the use of personalized digital technologies. ¹²⁵

In this SLR of studies published between 2010 and 2020, greater adherence to and persistence with antidiabetic medication in adults with T2D was typically associated with better clinical and economic outcomes. These findings suggest that the clinical benefits of adherence and persistence for patients are likely to be reflected in positive impacts for payers, healthcare systems and society. Further investigation of the factors that determine medication‐taking behaviour should be used to identify barriers to optimal adherence and persistence in people with T2D.

CONFLICT OF INTEREST

M.E. has received honoraria from AstraZeneca, Boehringer Ingelheim and Novo Nordisk. S.E. and M.F. are employees of Novo Nordisk A/S. J.F. was an employee of Novo Nordisk A/S at the time of the review, and is now an employee of Ferring Pharmaceuticals A/S. P.H. is an employee of Mtech Access, funded by Novo Nordisk A/S to carry out the SLR. W.P. has served as a consultant for Eli Lilly, Novo Nordisk and Sanofi.

AUTHOR CONTRIBUTIONS

Mads Faurby, João Diogo Da Rocha Fernandes and Pollyanna Hudson: designed the SLR. All authors contributed to the interpretation of data and critical review of the manuscript and approved the final version for submission.

PEER REVIEW

The peer review history for this article is available at https://publons.com/publon/10.1111/dom.14603.

Supporting information

Table S1. Search strategy for MEDLINE

Table S2. List of congresses searched

Table S3. Number of studies reporting each outcome

Table S4. Summary of studies reporting the associations between adherence to antidiabetic medication and inpatient admissions, ED visits/admissions and outpatient claims in patients with T2D

Table S5. Summary of studies reporting the associations between persistence with antidiabetic medication and inpatient admissions, ED visits/admissions and outpatient claims in patients with T2D

Figure S1. Preferred Reporting Items for Systematic Reviews and Meta‐Analyses (PRISMA) diagram.

Evans M, Engberg S, Faurby M, Fernandes JDDR, Hudson P, Polonsky W. Adherence to and persistence with antidiabetic medications and associations with clinical and economic outcomes in people with type 2 diabetes mellitus: A systematic literature review. Diabetes Obes Metab. 2022;24(3):377‐390. doi: 10.1111/dom.14603

DATA AVAILABILITY STATEMENT

‘All data analyzed during this systematic literature review are published elsewhere, and relevant collated data are included in this article [and/or] its supplementary material files. Further enquiries can be directed to the corresponding author’