Background

Adequate management of chronic disease is difficult. Patients are often required to take one or more medications over the entire lifespan of the disease [1]. Management of chronic disease is further complicated by two patterns of medication non-use: (1) missed medication doses (termed non-adherence in this study) and (2) abrupt discontinuation or substantial medication gap (termed non-persistence or discontinuation in this study) [2]. In developed countries, average adherence to medications for chronic diseases is as low as 50%, while the measure is lower in developing countries due to limited access to healthcare resources [3, 4]. Medication non-use aggravates the burden of chronic diseases and clinical outcomes of patients [4, 5]. Therefore, ensuring adherence and persistence of medications is key to successful management of chronic disease.

Poor adherence and persistence remain a barrier to optimal care for patients with type 2 diabetes mellitus (T2DM) [6,7,8,9]. A systematic review found that only 56.2% in T2DM patients continued treatment one year after treatment initiation [10]. Adherence and persistence to injection drugs are even lower. The persistence rate of insulin glargine in the first year after initiation is below 50% [11]. Suboptimal persistence undermines clinical outcomes, leading to poor glycemic control [12, 13] and increases mortality and comorbidity burden [14, 15]. Moreover, low adherence to antidiabetic medications increases healthcare costs and diminishes quality of life [5, 14, 16].

The causes of low adherence and persistence to T2DM medications are multifactorial [17]. The World Health Organization classified reasons for medication non-use into five categories: patient-related (e.g., age), socioeconomic (e.g., medication costs), condition-related (e.g., presence of complications), health-system-related (e.g., level of continuity of care), and medication-related (e.g., adverse effects) [4]. Similarly, motivations behind medication non-use in T2DM patients on injection therapies are multifaceted. Ineffective communication between patients and providers, inadequate knowledge about medications, and confusing directions for medication use simultaneously undermine treatment processes [18]. Moreover, the classes of antidiabetic medication influence the adherence and persistence to the treatment [1, 19].

The objective of this paper was to review the latest adherence and persistence rates of major antidiabetic medication classes chosen based on their proportional market shares [20]. Moreover, we compared the adherence and persistence rates of individual antidiabetic medications within the same medication class. Moreover, we investigated the potential causes for low adherence and persistence of antidiabetic medications. To this end, we summarized the major findings of the studies published since 2017. The year 2017 was chosen to account for the shift in pharmacological diabetes treatment pattern, as reflected in the medications’ proportional market shares [20], due in large part to the accelerating acceptance and widespread use of such medication classes as sodium glucose co-transporter 2 inhibitors and glucagon-like peptide-1 receptor agonists. Published studies on the topic of adherence and persistence of the selected antidiabetic medications were identified by searching the four databases, i.e., PubMed, Cochrane Library, Google Scholar, and Embase. This information will help guide clinical decisions to optimize treatment adherence and persistence in patients with T2DM while reducing complications and healthcare costs. A precise understanding about the causes of medication non-use will also assist the development of new antidiabetic drugs and delivery devices better equipped to improve adherence and persistence. The key findings of previous studies that analyzed the adherence and persistence of various antidiabetic medications are summarized in Table 1.

Table 1 Summary of previous studies on adherence and persistence of major antidiabetic medications
Full size table

Metformin

Metformin is well-tolerated and economic [21, 22], making it suitable for long-term treatment of T2DM. Furthermore, metformin has demonstrated the highest adherence and persistence rates in antidiabetic medications [2]. However, the adherence and persistence of metformin is still suboptimal [10, 23]. For example, the lowest daily medication possession probability (MPP) of metformin—i.e., the sum of days supplied by prescription fills during follow-up divided by the number of days in follow-up [24]—was only 0.46 [7]. Given that MPP ≥ 0.8 is generally accepted as the cut-off value for good adherence [25], the MPP value of 0.46 is certainly not optimal for a foundation medication like metformin. Likewise, the percentage of metformin users who continued treatment for one year ranged between 62.8 and 73.6% [26, 27]. The share of persistent metformin users declined to 48.5% and 27.7% at the end of second and fifth years of follow-up, respectively [26].

Baseline characteristics affect the adherence and persistence of metformin therapy. Higher baseline glycated hemoglobin (HbA1c) was associated with significantly lower rate of discontinuation. A percentage point increase in HbA1c was associated with 30% lower metformin persistence (95% confidence interval [95% CI]: 15–45%) [28]. Older patients were more persistent. For example, one year increase in age was associated with significantly better persistence (odds ratio or OR [95% CI]: 1.02 [1.02–1.02], p < 0.001) [29]. Using lower dose metformin (500 mg as opposed to 1000 mg) was associated with significantly lower discontinuation rate (OR [95% CI] of discontinuation of 500 mg metformin: 0.54 [0.37–0.76], p < 0.01) [28]. Similarly, taking fewer concomitant medication was associated with significantly better persistence (OR [95% CI]: 1.27 [1.20–1.33], p < 0.001) [29]. Patients using extended-release formulation were significantly more persistent than patients using immediate-release formulation (OR [95% CI] of persistence of extended-release formulation: 1.14 [1.10–1.18], p < 0.001) [30]. Table 2 summarizes the factors affecting the adherence and persistence of metformin therapy.

Table 2 Factors affecting adherence and persistence to metformin [19]
Full size table

Sulfonylureas

Sulfonylureas (including chlorpropamide, tolazamide, glipizide, glyburide, and glimepiride) are frequently prescribed to T2DM patients as second-line therapy despite potential hypoglycemic risks [20]. For sulfonylurea users, the proportion of days covered (PDC), or the number of days covered by prescription fills divided by the number of days between the first fill of the medication and the end of the measurement period [24], ranged from 0.62 and 0.72 [31, 32]. The percentage of patients who continued to take sulfonylureas at one year ranged between 50.4 and 68.9% [27, 31]. The longer the treatment period, the lower the persistence. For example, the percentage of patients who continued to take sulfonylureas declined to 51.3%, 47.1%, and 31.6% at the end of two, three and five years of follow-up, respectively [1, 32].

Sulfonylurea users were significantly less persistent than metformin users (hazard ratio or HR [95% CI] of discontinuation of sulfonylureas: 1.2 [1.16–1.24], p < 0.001) [1]. Moreover, the percentage of adherent patients (PDC ≥ 0.8) who used sulfonylurea as an add-on to metformin was significantly lower than that of patients who used sitagliptin as an add-on to metformin (55.9% versus 59.1%, p < 0.001) [32].

Sulfonylureas are commonly associated with hypoglycemia and weight gain [33, 34]. Approximately 20% of patients taking sulfonylureas may experience symptomatic hypoglycemia within six months after treatment initiation [35]. Moreover, sulfonylureas amplified weight gain in the first six months of treatment [36]. Such findings led to a speculation that the adverse events are the major reason for low adherence and persistence of sulfonylureas [33, 37]. However, no difference in treatment adherence was seen between patients who experienced hypoglycemic symptoms and those who did not [38]. Likewise, definitive evidence of the association between weight gain and poor adherence and persistence of sulfonylureas is still lacking.

Sodium glucose co-transporter-2 inhibitors

Sodium glucose cotransporter-2 (SGLT2) inhibitors, e.g., canagliflozin, dapagliflozin, and empagliflozin, are generally well tolerated [39, 40]. The average PDC of SGLT2 inhibitors at one year was between 0.64 and 0.79 [41, 42]. Likewise, the proportion of patients who continued to take SGLT2 inhibitors at one year ranged between 44.3 and 72.1% [42, 43].

There was a difference in adherence and persistence among individual SGLT2 inhibitors. Canagliflozin was associated with significantly higher adherence and persistence than dapagliflozin (OR [95% CI] of adherence for canagliflozin compared with dapagliflozin: 1.29 [1.02–1.72]; HR [95% CI] of discontinuation of dapagliflozin: 1.28 [1.15–1.42]; both p < 0.001) [41]. Treatment with canagliflozin is suspected to be associated with increased risks of lower extremity amputation and skeletal fractures, besides other adverse events commonly ascribed to SGLT2 inhibitors [44]. However, the impact of these adverse effects on the adherence and persistence of canagliflozin has not been fully investigated, warranting further studies. Similarly, empagliflozin users were significantly more adherent and persistent than dapagliflozin users (OR [95% CI] of adherence and persistence of empagliflozin: 1.39 [1.29–1.51] and 1.14 [1.06–1.22], respectively; both p < 0.01) [42]. No study has directly compared the adherence and persistence rates between canagliflozin and empagliflozin.

The persistence rate of SGLT2 inhibitors was comparable to, and in some case even higher than, well-persisted antidiabetic medication classes (i.e., metformin and dipeptidyl peptidase-4 inhibitors). For example, SGLT2 inhibitor users were as persistent as metformin users (HR [95% CI] of discontinuation of SGLT2 inhibitors: 1.04 [0.93–1.17], p = 0.458) [2]. Likewise, over 50% of SGLT2 inhibitor users and metformin users remained persistent into the second year of treatment, while the majority of patients treated with other medication classes discontinued [2]. Furthermore, the discontinuation rate (i.e., having a prescription gap of > 90 days) of canagliflozin users was 12% less (p < 0.001) than that of sitagliptin (a dipeptidyl peptidase-4 inhibitor) users [45]

SGLT2 inhibitors block glucose reabsorption in the renal proximal tubules of the kidneys [41, 46, 47]. Their distinct mechanism of action entails a distinct set of adverse events, such as orthostatic hypotension, ketoacidosis, and most notably, genitourinary tract infections [48, 49]. The unique safety profile of SGLT2 inhibitors led to a distinct factor for persistence. For example, higher estimated glomerular filtration rate (eGFR) was associated with significantly lower persistence rate (the beta coefficient [standard error] for discontinuation per one-unit increase in eGFR: 0.01 [± 0.00], p < 0.001) [49]. The increased likelihood of genitourinary tract infections due to the hyperfiltration of urinary glucose excretion was proposed as a reason for the association [49]. Similar conclusions were made by other studies [48, 50].

On the other hand, the factors for persistence commonly observed in other antidiabetic medications also applied to SGLT2 inhibitors. For example, younger age [41], female gender [41, 49], higher baseline anxiety [41], higher baseline HbA1c [49], and baseline insulin use [41, 51] were associated with significantly higher discontinuation rates in SGLT2 inhibitor users. In contrast, lower starting dose [41], perceived feeling of improved clinical outcome [41, 51], and taking fewer number of concomitant medications [49, 51] were associated with significantly lower discontinuation rates in SGLT2 inhibitor users.

Dipeptidyl peptidase-4 inhibitors

As of July 2021, four dipeptidyl peptidase-4 (DPP4) inhibitors (sitagliptin, saxagliptin, linagliptin, and alogliptin) are in clinical use in the US [52]. Other DPP4 inhibitors that are being prescribed worldwide include anagliptin, tenegliptin, vildagliptin, omarigliptin, and trelagliptin [53]. DPP4 inhibitors prevent the degradation of glucagon like peptide-1 (GLP1) and stimulate postprandial insulin secretion in a glucose-dependent manner [54]. DPP4 inhibitors are generally well tolerated and pose little risk of hypoglycemia and weight gain [55, 56]. This safety profile leads to good adherence and persistence in DPP4 inhibitor users. The mean PDC at one year of DPP4 inhibitors ranged from 0.67 to 0.77 [56, 57]. The percentage of patients who remained persistent to DPP4 inhibitors at one year was between 56.7 and 78.8% [27, 57]. The median persistence of DPP4 inhibitors was approximately seventeen months [2, 58].

A comparison between individual DPP4 inhibitors showed that sitagliptin and saxagliptin demonstrated similar adherence and persistence rates [56]. On the other hand, sitagliptin and saxagliptin were associated with significantly better adherence and persistence than linagliptin (OR [95% CI] of adherence for sitagliptin and saxagliptin: 1.40 [1.25–1.57] and 1.46 [1.29–1.66], respectively; HR [95% CI] of discontinuation for sitagliptin and saxagliptin: 0.88 [0.82–0.94] and 0.85 [0.79–0.91], respectively; all p < 0.001) [56].

The persistence of once-daily (QD) DPP4 inhibitors was comparable to that of twice-daily (BID) DPP4 inhibitors (HR [95% CI] of discontinuation for BID regimen: 1.022 [0.994–1.050], p = 0.1187) [59]. Similarly, the difference between the adherence rates of BID regimen and QD regimen was not significant (OR [95% CI] of adherence for BID regimen: 0.945 [0.780–1.145], p = 0.5636) [59]. On the other hand, QD regimen showed significantly higher adherence and persistence rates than once-weekly (QW) regimen (HR [95% CI] of discontinuation for QW regimens: 1.699 [1.585–1.822], p < 0.0001; OR [95% CI] of adherence for QW regimen: 0.029 [0.024–0.036], p < 0.0001) [59]. However, there are also conflicting results. In a prospective study, adherence was improved by 0.1 point (p = 0.03) on the Diabetes Treatment Satisfactions Questionnaire (DTSQ) scale after patients switched from QD DPP4 inhibitors (i.e., sitagliptin, vildagliptin, alogliptin, linagliptin, and teneligliptin) to QW trelagliptin [60].

The persistence of DPP4 inhibitors is lower than that of metformin (HR [95% CI] of discontinuation for DPP4 inhibitors: 1.43 [1.38–1.49], p < 0.001) [2]. However, DPP4 inhibitors are well adhered and persisted in T2DM patients with impaired kidney functions at all renal impairment stages [58]. Similarly, the adherence of DPP4 inhibitors in patients with chronic kidney disease (CKD) was significantly higher (OR [95% CI] of adherence for DPP4 inhibitors: 1.41 [1.25–1.59], p < 0.01) than pioglitazone, a well-established second-line therapy for T2DM patients with CKD [57].

The speculation that regimen complexity or the severity of diabetes may not necessarily lower adherence to DPP4 inhibitors [23] was supported by several studies [57,58,59]. Common reasons for discontinuing DPP4 inhibitors were inadequate glycemic control and intolerance, as with other antidiabetic medications [61].

Insulin

Insulin is key to improving the glycemic outcome in many T2DM patients. However, the adherence and persistence rates of insulin treatment are generally suboptimal. For example, one-year persistence rates of insulin treatment, measured as the percentage of patients who remained on therapy, were only 20–66.8% [62, 63]. Similarly, adherence to insulin was low; 58.5% of T2DM patients on insulin therapy scored poor adherence (scoring below 6) on the Morisky-Green Questionnaire [64]. The adherence to insulin treatment was inversely related to treatment period. The proportion of adherent patients (i.e., those with PDC ≥ 0.8) on basal insulins dropped to 33.8% three years after treatment initiation [65]. Generally, basal insulins demonstrated better persistence than rapid-acting and short-acting insulins [2]. The major findings of previous studies that analyzed the adherence and persistence to insulin therapy are summarized in Table 1.

Simple reminding has been insufficient to improve the adherence and persistence of insulin treatment. In a recent randomized clinical trial, individualized interventions such as quarterly educational mailings, telephone consultation by pharmacists, and text reminders did not improve insulin persistence despite increasing intensity of the interventions [66]. Multifaceted approaches are crucial to adequately address insulin non-adherence and non-persistence. Table 3 summarizes the factors affecting adherence and persistence to insulin therapy.

Table 3 Factors affecting adherence and persistence to insulin therapy
Full size table

Glucagon-Like peptide-1 receptor agonists

Glucagon-like peptide-1 receptor agonists (GLP-1RA) improve glycemic control and cardiovascular factors, reduce body weight, and rarely induce hypoglycemia [67]. GLP-1RA agents are preferred second-line treatment options for T2DM patients with cardiovascular comorbidities [68]. Furthermore, GLP-1RA agents are recommended as the first injectable medication before insulin [69]. As of 2021, nine formulations of injectable GLP-1RA agents have been approved worldwide (Table 4). Oral semaglutide (brand name: Rybelsus®) was the first oral formulation of GLP-1RA approved by the US Food and Drug Administration for the treatment of T2DM [70].

Table 4 List of injectable GLP-1RA agents currently in clinical use worldwide
Full size table

The average PDC of injectable GLP-1RA agents at six months ranged from 0.61 to 0.76 [71, 72]. The proportion of patients who continued treatment with injectable GLP-1RA at six months was between 32.1 and 74.0% [73, 74]. The adherence and persistence of oral semaglutide remain to be seen.

Injectable GLP-1RA agents differ in dosing regimens, need for dose titration and reconstitution, and administration device features [75]. These differences led to differences in adherence and persistence rates among individual GLP-1RA agents. Dulaglutide showed significantly higher persistence than other GLP-1RA agents (HR [95% CI] of discontinuation compared with dulaglutide: 2.5 [2.1–3.0] for exenatide QW, 1.6 [1.5–1.8] for liraglutide, 1.4 [1.3–1.5] for semaglutide, and 2.8 [2.3–3.3] for lixisenatide; all p < 0.001) [69, 75]. Similarly, dulaglutide was associated with significantly higher adherence than other GLP-1RA agents (OR [95% CI] of adherence compared with dulaglutide: 0.63 [0.55–0.73] for albiglutide, 0.32 [0.28–0.37] for exenatide BID, 0.48 [0.43–0.53] for exenatide QW, and 0.65 [0.59–0.71] for liraglutide; all p < 0.05) [76].

In general, GLP-1RA agents with QW regimen demonstrated significantly better adherence and persistence than GLP-1RA agents with QD or BID regimen [67, 72, 74, 76]. In terms of delivery method, GLP-1RA agents using simple delivery systems (single-use pen or auto-injector device) had significantly higher adherence and persistence than GLP-1RA using multi-use pen or syringe [67, 71, 75,76,77]. Furthermore, experiencing early response (defined as improvements in HbA1c and body weight within six months after treatment initiation) was associated with significantly higher adherence and persistence in GLP-1RA users [73]. Other factors for the adherence and persistence of GLP-1RA agents are summarized in Table 5.

Table 5 Factors affecting the adherence and persistence to GLP-1RA
Full size table

Thiazolidinediones

Thiazolidinediones (TZDs), including pioglitazone, are agonists of peroxisome proliferator-activated receptor-γ (PPAR-γ) used in the treatment of T2DM. TZDs reduce plasma glucose by directly activating PPAR-γ and improving insulin sensitivity [78, 79]. Furthermore, TZDs, along with DPP4 inhibitors, are the major treatment option for T2DM patients with impaired renal functions [57]. Despite their therapeutic benefits, few studies published since 2017 analyzed the adherence and persistence of TZDs due partly to the diminished market share (in case of rosiglitazone) or withdrawal from the market (in case of troglitazone) [20, 80,81,82]. Still, the adherence of TZDs that are still being prescribed (e.g., pioglitazone) has been reported to range from 0.36 (measured in the daily MPP) to 0.72 (measured in PDC) [7, 57]. The proportion of patients who remained persistent with TZDs for one-year varied from 46.3 to 75.6% (Table 1) [1, 57].

The safety issues surrounding TZDs significantly lowered adherence and persistence. For example, after the FDA issued a safety warning in June 2011 for pioglitazone and its possible link to bladder cancer, the discontinuation rate of pioglitazone increased significantly from 36.3% in the year 2010 to 41.0% in the year 2011 (p < 0.01) [57].

Conclusion

This article provided a comprehensive review of the adherence and persistence of major antidiabetic medications, i.e., metformin, sulfonylureas, SGLT2 inhibitors, DPP4 inhibitors, insulins, GLP-1RA agents, and TZDs. The adherence and persistence of major antidiabetic medications were not optimal, given that PDC ≥ 0.8 and the proportion of persistent patients ≥ 80% are generally recognized as optimum [25].

Most studies reported adherence in PDC and defined patients with PDC ≥ 0.8 as adherent. Persistence was predominantly defined as continuous treatment without a prescription gap of more than 90 days. In most studies, the proportion of patients who remained persistent was used as the measure of the drug’s persistence rate.

The highest rate of adherence and persistence was consistently observed in metformin users. Second to metformin were SGLT2 inhibitors. Injectable therapies such as insulin and GLP-1RA agents trailed low on the adherence and persistence rates. To the best of our knowledge, no studies published since the year 2017 analyzed the adherence and persistence of TZDs independently.

Most studies pointed out that the prevalence and severity of adverse events is associated with low medication adherence and persistence. Baseline characteristics (e.g., baseline HbA1c level), socioeconomic factors (e.g., medication costs and insurance status), demographic information (e.g., age, gender, or ethnicity), and comorbidity profiles also had significant impacts on adherence and persistence in T2DM patients.

It is important to note that reports on the adherence and persistence rates varied, depending on study design, data source, and patient sample. Using different definitions of adherence (e.g., daily MPP or PDC) and of persistence (e.g., continuous treatment without a prescription gap of over 60 days or 90 days) may have also contributed to the variations in the findings. In the similar vein, primary adherence (i.e., the rate at which patients fill prescriptions for the first time after treatment initiation) is critical for timely treatment of both acute and chronic conditions [83]. Despite its clinical significance, we found that the primary adherence of antidiabetic medications has not been extensively covered in the literature. Moreover, most studies measured adherence and persistence over one year after the initiation of antidiabetic drugs. Because these measures are inversely related to the duration of follow-up, using different observation periods may have led to different results. Lastly, most studies relied on electronic health records (EHR) and claims data to analyze the adherence and persistence. Thus, it should be acknowledged that a purely claims-based study may have underestimated adherence and persistence by leaving out the patients who paid out of pocket. On the contrary, a purely EHR-based study may have overestimated the adherence and persistence because there are significantly more provider attempts to prescribe a drug than there are patients voluntarily taking the drug in the long term [7].

This review article aimed to address the following points. First, this review article offers a concise summary of the adherence and persistence rates of antidiabetic medications that comprise of the most of the proportional market share worldwide. Secondly, the findings summarized in this article may help guide the clinicians and the patients to make informed treatment decisions. Study results from randomized clinical trials confirm the effectiveness and safety of a medication. However, the results may not always translate into everyday clinical benefit, as there are diverse factors at work, including adherence and persistence. In this light, knowing the treatment-related factors and patient-level information affecting the behavior of medication (e.g., treatment adherence and persistence) use may equip the medical community to identify the beneficiaries of the drug’s effectiveness with higher granularity. Lastly, information contained in this article may provide insight into the paths for the development of new antidiabetic medications or injection devices that are more amenable to fostering adherence and persistence.

Because the treatment of diabetes relies heavily on antidiabetic medications, along with lifestyle modification, dietary interventions, and other medications for comorbid conditions, ensuring adherence and persistence is key to adequately managing the disease. Moreover, adherence and persistence are clinically important phenomena with implications for research and clinical practice. Therefore, clinicians should be aware of the challenges concerning adherence and persistence to antidiabetic medications. It is imperative that they pay attention to how the multifactorial nature of medication non-use undermines patients’ quality of life and clinical outcomes.