Abstract
OBJECTIVE
To address previous inconsistencies in reports of differential adherence to diabetes medications, we examined medication adherence and evaluated treatment group differences in a substudy of participants in the Glycemia Reduction Approaches in Diabetes: A Comparative Effectiveness Study (GRADE).
RESEARCH DESIGN AND METHODS
GRADE participants (type 2 diabetes duration <10 years, HbA1c 6.8%–8.5%, on metformin alone) were randomly assigned to add insulin glargine, glimepiride, liraglutide, or sitagliptin. Adherence was measured semiannually for 3 years using a validated three-item scale (0–100, lowest to highest adherence) in a substudy (N = 1,739). Analyses included evaluation of adherence over time and testing treatment group differences in adherence and in the association between adherence and primary (HbA1c ≥7.0%) and secondary (HbA1c >7.5%) glycemic outcomes.
RESULTS
Overall mean ± SD adherence (average of participant-level mean ± SD) was high over 3 years of follow-up at 88.7 ± 10.01, on a scale of 0–100, and decreased slightly by 3 years relative to baseline (−2.0± 14.7; P < 0.0001). No intergroup differences were observed until 3 years, when adherence was 5% and 3% higher for the glimepiride and sitagliptin groups, respectively, than for liraglutide (both P < 0.05). Over follow-up and across groups, a 10-point decrease in adherence was associated with 15% and 19% increased risk of reaching primary (HbA1c ≥7.0%) and secondary (HbA1c >7.5%) glycemic outcomes (both P < 0.0001). Lower adherence was somewhat more predictive of the secondary outcome for those assigned to glargine or liraglutide, compared with glimepiride or sitagliptin (each P < 0.05). No other comparisons were significant.
CONCLUSIONS
Medication adherence was consistently high in GRADE. Observed treatment group differences were small and of unclear clinical significance. Overall, lower adherence robustly predicted worsening glycemic control, highlighting the importance of ongoing assessment.
Graphical Abstract
Introduction
Research on medication adherence among adults with type 2 diabetes mellitus (T2DM) has shown that many individuals do not fill initial prescriptions for glucose-lowering medications, prematurely discontinue filling prescriptions over time, and often take fewer doses than prescribed over a given period (1,2). Nonadherence to prescribed diabetes medications has been consistently associated with deterioration in glycemic control, increased risk of hospitalization, microvascular and macrovascular complications, and mortality (3–6). The prevalence of nonadherence may explain the gap between glycemic effects observed in clinical trials versus real-world practice (7). However, adherence often goes unassessed and/or unreported in clinical trials, potentially distorting the interpretation of results (8).
Evidence linking treatment regimen characteristics to medication adherence in T2DM has been inconsistent. Although some studies link more frequent dosing and regimen complexity to nonadherence (9–11), in others no association was found between polypharmacy and medication adherence in T2DM (12), and a positive association was even found between increased number of medications and adherence (13). Beyond the potential effects of regimen complexity and polypharmacy, T2DM research suggests that patients may be differentially adherent depending on the glucose-lowering medication prescribed. Nonadherence has been noted as a particularly important problem for insulin therapy, where evidence suggests lower adherence relative to oral medications (14,15). Evidence links nonadherence to delays in insulin initiation and suggests that patients and health care providers are often reluctant to initiate insulin because of increased perceived burden relative to other treatments (16–19). Persistence with the prescribed duration of treatment and adherence to day-to-day dosing of glucose-lowering medications over time also appear to differ among oral glucose–lowering medications (20). Differences in levels of adherence typically achieved by patients taking different glucose-lowering medications or the extent to which nonadherence places them at risk for poor glycemic outcomes would inform provider-patient decision-making in selecting among treatments.
The Glycemia Reduction Approaches in Diabetes: A Comparative Effectiveness Study (GRADE) was a randomized controlled trial comparing the metabolic effects of four common glucose-lowering medications in combination with metformin in relatively recent-onset T2DM (<10 years since diagnosis) (21–23). Random assignment to treatment options provided an opportunity to evaluate 1) levels of adherence over time and whether any differences were observed among treatments, 2) potential mediators of any observed treatment group differences in adherence, 3) the effect of adherence over time on the risk of reaching glycemic outcomes, and 4) potential differences in the association between nonadherence and glycemic outcomes by treatment regimen. These research aims were examined in the GRADE Emotional Distress Substudy (EDS). We hypothesized that adherence would be lower among participants assigned to insulin glargine. We also hypothesized that the effect of nonadherence on risk for these glycemic outcomes would be strongest for the treatments previously reported in GRADE to be more effective, glargine and liraglutide (23).
Research Design and Methods
Description of GRADE and EDS
GRADE was a multicenter, parallel, unmasked randomized clinical trial with comparison of four commonly used glucose-lowering medications (basal insulin glargine U-100, sulfonylurea glimepiride, GLP-1 agonist liraglutide, or DPP-4 inhibitor sitagliptin) in metformin-treated participants, who were followed for an average of 5 years. Participants were enrolled over the 4-year recruitment period at 36 clinical centers and 9 subsites throughout the U.S. Eligibility criteria included age ≥30 years at time of diagnosis (except for American Indians and Alaska Natives, eligible if age ≥20 years), diagnosis of T2DM within 10 years, HbA1c 6.8%–8.5% (51–69 mmol/mol), treatment with metformin alone (minimum dose of 1,000 mg/day [up to 2,000 mg/day]), and willingness to take a second glucose-lowering medication including injectable medication. Further details of the GRADE protocol have previously been published (22).
EDS began more than halfway through GRADE recruitment, and 1,739 participants were enrolled from 2015 to 2017 (out of 5,047 GRADE participants). All GRADE sites were invited to participate in EDS. A total of 26 centers and 8 subsites across the U.S. obtained institutional review board approval and participated. The GRADE consent form was amended to include information on EDS, and all individuals newly enrolled into GRADE at a participating EDS site were automatically enrolled in EDS. Participants were given additional compensation for EDS procedures. Baseline measures were collected before initiation of the randomized glucose-lowering medication. Details of the EDS protocol have previously been published (21).
Glargine U-100 was initiated at up to 20 units/day, and adjusted to avoid hypoglycemia according to self-monitored blood glucose. Glimepiride was increased from 1–2 mg to a maximum of 8 mg/day in divided doses, with adjustment to avoid hypoglycemia according to self-monitored blood glucose. Liraglutide was initiated at 0.6 mg, with escalation to a maximum of 1.8 mg per day, depending on gastrointestinal (GI) side effects. Sitagliptin was initiated at 100 mg/day, with the dose adjusted according to kidney function. Participants were assessed quarterly and continued their assigned medication regimen until HbA1c level was confirmed to reach >7.5% (>58.5 mmol/mol) (the secondary glycemic outcome). Once this point was reached, additional medications were added (22,23). Participants initially randomized to noninsulin medications had glargine added to their regimen, followed by prandial rapid-acting insulin aspart if, after the addition of glargine, HbA1c levels again reached a confirmed value >7.5%. At this point, metformin was continued and the initially assigned noninsulin medication was discontinued. Participants assigned to the glargine group who reached a confirmed value of HbA1c >7.5% had insulin aspart added to their regimen. The primary and secondary metabolic, microvascular, and cardiovascular outcomes have previously been reported (23,24). Analyses for the current article include all GRADE-EDS participants, including those whose treatment was modified.
Measures
Medication Adherence
Adherence to glucose-lowering medications was assessed every 6 months using a validated three-item instrument (25,26). Participants were instructed to consider “all of the medications you’ve been prescribed as part of the GRADE study, including injections.” The three self-report items ask the following, referencing the last 30 days: 1) “On how many days did you miss at least one dose of any of your diabetes medicines?” (1–30), 2) “How good a job did you do at taking your diabetes medicines in the way you were supposed to?” (“1=Very Poor” to “6=Excellent”), and 3) “How often did you take your diabetes medicines in the way you were supposed to?” (“1=Never” to “6=Always”). A total score for the equally weighted three items was calculated on a scale of 0–100, with higher scores representing better adherence (25,26).
Glycemic Outcomes
The primary glycemic outcome was the first measurement of HbA1c ≥7% (assessed quarterly), confirmed at the next quarterly visit. The secondary glycemic outcome was the first HbA1c >7.5%, confirmed.
Potential Mediators
Follow-up visits were conducted every 3 months, during which the following mediator variables were assessed. HbA1c was measured with a high-performance liquid chromatography method (23,24) and used to determine whether the primary or secondary glycemic outcomes were met at each visit. Hypoglycemia included adjudicated severe hypoglycemia (episodes requiring assistance to treat) or recurrent nonsevere hypoglycemia that first occurred after baseline, regarding which information was collected as follows. Possible episodes of severe hypoglycemia were self-reported and were adjudicated by two reviewers masked to treatment group. In addition, all symptomatic hypoglycemia was self-reported for the last 30 days, and recurrent hypoglycemia was defined as two or more symptomatic hypoglycemia events occurring after baseline. Self-reported GI side effects were defined as weekly or daily occurrences of at least one of nausea, vomiting, bloating, or diarrhea over the previous 30 days. BMI was calculated from height and weight measured in duplicate by certified, trained staff. We also tracked additions to the glucose-lowering medication regimen (per-protocol treatment intensification and those prescribed by nonstudy health care providers) as a binary “treatment intensification” variable at each visit to evaluate whether changes to the medication regimen explained treatment group differences in adherence.
EDS participants completed additional surveys at baseline, and every 6 months for 3 years (21). Beliefs about medicines were captured using two subscales from the Beliefs About Medicines Questionnaire (BMQ) (28): Necessity (5 items) and Concerns (5 items), each scaled to range from 1 to 5, with higher scores indicating stronger necessity and concerns beliefs. Diabetes self-efficacy was assessed with the Perceived Diabetes Self-Management Scale (PDSMS) (29). Scores range from 12 to 40, with higher scores indicating higher self-efficacy. We also included a six-item scale from the Illness Perceptions Questionnaire-Revised, validated for various chronic illnesses, including diabetes (33) to evaluate participants’ perceived ability to influence the course of their diabetes through their own actions; scores range from 6 to 30 with higher scores indicating more perceived control (30). Satisfaction with diabetes treatment was assessed with the Diabetes Treatment Satisfaction Questionnaire (DTSQ) (31). The eight-item Patient Health Questionnaire (PHQ-8) total score was used to measure depressive symptoms, with total scores ranging from 0 to 24. Five somatic (sleep, fatigue, appetite, concentration, and psychomotor changes) and three cognitive-affective (loss of interest, depressed mood, and negative self-feelings) symptom items were also examined separately (32). The 17-item Diabetes Distress Scale (DDS) was used to measure diabetes distress (range 1–6), with higher scores indicating greater diabetes distress (33). Each potential mediator was calculated as the mean value from baseline to month 36 for continuous measures and the mean value across visits from baseline to month 36 for binary measures, representing the proportion of visits at which the event occurred.
Statistical Analyses
All analyses were conducted as intention-to-treat analyses unless otherwise specified. With 1,739 participants, the study had >95% power to detect a 2-point difference in adherence scores using ANOVA with treatment group as the independent variable, assuming an SD of 10 points. Baseline characteristics were summarized by treatment group with use of one-way ANOVA for continuous variables and Pearson χ2 test for categorical variables to assess differences among groups.
Because adherence data were severely skewed and kurtotic, with most participants reporting perfect or near-perfect scores, a log-transformation was applied to adherence data and used as a continuous variable in all analyses. A line graph was used to illustrate the mean medication adherence by treatment group over 36 months. ANOVA was used to assess differences in mean log-transformed medication adherence by treatment group separately at months 6, 12, 24, and 36 (unadjusted). Pairwise comparisons among groups were assessed using unadjusted linear regression. These pairwise comparisons were exploratory and are reported with unadjusted P values and 95% CIs. Unadjusted and adjusted linear generalized estimating equation (GEE) models were used to evaluate the association between treatment group and log-transformed medication adherence over time, with an exchangeable correlation structure to account for within-participant correlation over time. The adjusted GEE model included baseline age, HbA1c, T2DM duration, and study visit as covariates.
Mediation analyses were conducted to assess whether previously described variables mediated (i.e., partially explained/attenuated) the treatment differences in medication adherence at month 36, separately for each potential mediator variable. The main effect between treatment group and log-transformed medication adherence at month 36 was assessed using unadjusted linear regression. The effect of the treatment group on each potential mediator was then assessed using unadjusted linear regression (continuous variables) or logistic regression (binary variables). If the treatment group had a significant effect on the potential mediator (P < 0.05), the effect of the treatment group on log-transformed medication adherence at month 36 was then assessed, with adjustment for the potential mediator as a covariate in the regression model. We compared percent change in the association between log-transformed medication adherence and treatment group, comparing the unadjusted model with the model adjusted for the potential mediator.
For assessment of the marginal association of medication adherence with primary (HbA1c ≥7%) and secondary (HbA1c >7.5%) glycemic outcomes, separate Cox proportional hazards models were fit for each outcome with medication adherence as a time-varying predictor. For assessment of whether medication adherence modified the treatment group effect on each of the glycemic outcomes, separate Cox models were fit for each outcome including the interaction between treatment group and time-varying medication adherence. All Cox models accounted for time-varying covariates by clustering on patient identifier. A significance level of 0.05 was used for all analyses. All analyses were conducted with R, version 4.2.1 (34).
Data and Resource Availability
This article is based on follow-up data and outcome assessments from the 1,739 participants enrolled into EDS. The GRADE and EDS databases were made available in the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) Central Repository in 2025.
Results
Participant Cohort Characteristics
A total of 1,739 GRADE participants were enrolled in EDS. The majority of the participants were male, and average age was 58 years old. At baseline, average diabetes duration was ∼4 years, BMI 34.1 kg/m2, and HbA1c 7.5%. Participants were racially and ethnically diverse. Overall, sample characteristics of EDS and non-EDS GRADE participants were similar, as previously reported (21). Baseline demographic characteristics are summarized in Table 1. Self-reported mean ± SD adherence to metformin monotherapy at baseline was 89.9 ± 11.1 on the 0–100 scale, with a median score of 92.2 (interquartile range 84.4, 100).
Table 1.
Baseline participant characteristics overall and by treatment group
| Overall | Treatment group | |||||
|---|---|---|---|---|---|---|
| Glargine | Glimepiride | Liraglutide | Sitagliptin | P * | ||
| n | 1,739 | 418 | 439 | 450 | 432 | |
| Age (years) | 58.0 ± 10.2 | 57.4 ± 10.1 | 57.6 ± 10.3 | 58.5 ± 10.3 | 58.3 ± 10.1 | 0.317 |
| Diabetes duration (years) | 4.2 ± 2.8 | 4.1 ± 2.8 | 4.4 ± 2.9 | 4.2 ± 2.8 | 4.2 ± 2.8 | 0.606 |
| HbA1c (%) | 7.5 ± 0.5 | 7.5 ± 0.5 | 7.5 ± 0.5 | 7.5 ± 0.5 | 7.5 ± 0.5 | 0.440 |
| BMI | 34.1 ± 6.5 | 34.4 ± 6.3 | 34.0 ± 6.7 | 34.3 ± 6.3 | 33.7 ± 6.7 | 0.391 |
| Medication adherence** | 89.9 ± 11.1 | 89.0 ± 11.6 | 90.7 ± 11.1 | 90.2 ± 10.2 | 89.6 ± 11.3 | 0.140 |
| Sex | 0.130 | |||||
| Male | 1,175 (67.6) | 288 (68.9) | 278 (63.3) | 317 (70.4) | 292 (67.6) | |
| Female | 564 (32.4) | 130 (31.1) | 161 (36.7) | 133 (29.6) | 140 (32.4) | |
| Race and ethnicity | 0.316 | |||||
| Hispanic | 292 (16.8) | 68 (16.3) | 84 (19.1) | 65 (14.4) | 75 (17.4) | |
| Non-Hispanic Black or African American | 317 (18.2) | 79 (18.9) | 90 (20.5) | 82 (18.2) | 66 (15.3) | |
| Non-Hispanic other | 155 (8.9) | 31 (7.4) | 37 (8.4) | 43 (9.6) | 44 (10.2) | |
| Non-Hispanic White | 975 (56.1) | 240 (57.4) | 228 (51.9) | 260 (57.8) | 247 (57.2) | |
| Education | 0.624 | |||||
| College/graduate school | 711 (40.9) | 167 (40.0) | 173 (39.4) | 181 (40.2) | 190 (44.0) | |
| High school/GED or less | 517 (29.7) | 117 (28.0) | 140 (31.9) | 137 (30.4) | 123 (28.5) | |
| Some college | 511 (29.4) | 134 (32.1) | 126 (28.7) | 132 (29.3) | 119 (27.5) | |
| Income (USD) | 0.446 | |||||
| <10K | 93 (6.2) | 24 (6.6) | 20 (5.5) | 24 (6.2) | 25 (6.6) | |
| 10K–20K | 163 (10.9) | 47 (12.8) | 35 (9.6) | 48 (12.4) | 33 (8.7) | |
| 20K–50K | 477 (31.8) | 103 (28.1) | 114 (31.2) | 131 (33.8) | 129 (33.9) | |
| ≥50K | 766 (51.1) | 192 (52.5) | 196 (53.7) | 185 (47.7) | 193 (50.8) | |
| Living situation | 0.437 | |||||
| Alone | 295 (17.0) | 78 (18.7) | 65 (14.8) | 74 (16.4) | 78 (18.1) | |
| With another adult | 1,376 (79.1) | 324 (77.5) | 351 (80.0) | 363 (80.7) | 338 (78.2) | |
| With children only | 68 (3.9) | 16 (3.8) | 23 (5.2) | 13 (2.9) | 16 (3.7) | |
| Employment | 0.869 | |||||
| Currently employed full- or part-time | 974 (56.0) | 232 (55.5) | 249 (56.7) | 243 (54.0) | 250 (57.9) | |
| Currently retired | 459 (26.4) | 110 (26.3) | 111 (25.3) | 123 (27.3) | 115 (26.6) | |
| Other | 306 (17.6) | 76 (18.2) | 79 (18.0) | 84 (18.7) | 67 (15.5) | |
| Medication necessity beliefs | 3.5 ± 0.7 | 3.5 ± 0.7 | 3.6 ± 0.8 | 3.5 ± 0.8 | 3.6 ± 0.7 | 0.028 |
| Medication concerns | 2.6 ± 0.8 | 2.5 ± 0.8 | 2.6 ± 0.9 | 2.6 ± 0.8 | 2.6 ± 0.8 | 0.343 |
| Perceived control of diabetes | 20.8 ± 2.4 | 20.6 ± 2.4 | 20.9 ± 2.6 | 20.8 ± 2.4 | 20.8 ± 2.3 | 0.473 |
| Self-efficacy | 24.6 ± 3.0 | 24.7 ± 3.0 | 24.6 ± 3.0 | 24.6 ± 2.9 | 24.6 ± 3.0 | 0.800 |
| Diabetes distress | 1.7 ± 0.7 | 1.6 ± 0.6 | 1.7 ± 0.9 | 1.7 ± 0.7 | 1.7 ± 0.8 | 0.680 |
| Depressive symptoms total score | 3.4 ± 4.0 | 3.6 ± 4.0 | 3.3 ± 4.0 | 3.6 ± 4.2 | 3.2 ± 3.7 | 0.354 |
| Cognitive-affective symptoms | 0.3 ± 0.5 | 0.3 ± 0.5 | 0.3 ± 0.5 | 0.3 ± 0.5 | 0.3 ± 0.5 | 0.452 |
| Somatic symptoms | 0.5 ± 0.5 | 0.5 ± 0.5 | 0.5 ± 0.5 | 0.5 ± 0.6 | 0.5 ± 0.5 | 0.400 |
| Treatment satisfaction | 5.2 ± 1.2 | 5.1 ± 1.3 | 5.2 ± 1.2 | 5.2 ± 1.1 | 5.2 ± 1.2 | 0.884 |
Data are presented as mean ± SD for continuous variables, and N (%) for categorical variables. Boldface indicates significance at P ≤ 0.05. GED, general equivalency diploma.
*One-way ANOVA (continuous variables) or Pearson χ2 test (categorical variables).
**Adherence on the original 0–100 scale. Medication adherence refers to metformin adherence at the end of run-in.
Medication Adherence Over the Follow-up and Treatment Group Differences
As seen in Supplementary Fig. 1, adherence was high overall across all available visits from baseline to month 36—88.7 ± 10.01 (average of participant-level means ± SD) on the 0–100 scale at baseline—but decreased, −2.0 ± 14.7 (P < 0.0001), across all treatment groups over 3 years of follow-up. This difference on the 0–100 scale increases between months 30 and 36 for the liraglutide group, relative to the other treatment groups (Supplementary Fig. 1).
Measures of adherence at 6, 12, and 24 months did not show statistically significant differences in adherence by treatment group for the overall treatment effects at each time point or for any of the 18 pairwise comparisons (Supplementary Table 1). However, at the 36-month assessment, we did observe a statistically significant overall treatment group difference in adherence (P = 0.03). Table 2 presents estimated percent differences between treatment groups in 36-month adherence on the original scale, showing that adherence was 5.2% higher (2.7-point difference in sample means) for glimepiride (P = 0.003) and 3.5% higher (1.8-point difference in sample means) for sitagliptin (P = 0.050), relative to liraglutide. No other pairwise comparisons were statistically significant. In a supplemental set of analyses using GEE models we examined medication adherence averaged over all available assessments during follow-up by treatment group and found no significant overall treatment group difference (Supplementary Table 2).
Table 2.
Medication adherence across treatment groups at 36 months (n = 1,536)
| Percent difference (95% CI) | P | |
|---|---|---|
| Overall treatment effect (ANOVA) | 0.030 | |
| Pairwise comparisons | ||
| Glimepiride vs. glargine | 2.3 (−1.3, 5.9) | 0.212 |
| Liraglutide vs. glargine | −2.8 (−6.1, 0.6) | 0.108 |
| Sitagliptin vs. glargine | 0.57 (−2.9, 4.2) | 0.751 |
| Glimepiride vs. liraglutide | 5.2 (1.7, 8.8) | 0.003 |
| Sitagliptin vs. liraglutide | 3.5 (0.0004, 7.1000) | 0.050# |
| Sitagliptin vs. glimepiride | −1.6 (−5.0, 1.8) | 0.345 |
Note: the unadjusted linear model estimates the log-transformed medication adherence at the 36-month visit as a function of the treatment groups. The estimated percent difference (%) represents the relative difference in medication adherence between the treatment groups on the original scale, obtained by exponentiating the model coefficients. #The P value for sitagliptin vs. liraglutide is 0.04997, rounded to 0.050. Boldface indicates significance at P ≤ 0.05.
Mediators and Suppressors of Treatment Group Differences at 36 Months
For testing of potential mediators that could account for the observed statistically significant differences in comparison of liraglutide versus glimepiride and sitagliptin at 36 months, we conducted comparisons of glimepiride versus liraglutide and sitagliptin versus liraglutide for each of the prespecified potential mediators over follow-up (Supplementary Table 3). These analyses showed that, compared with the liraglutide group, the glimepiride group had higher HbA1c, reported more hypoglycemia, had fewer GI side effects, reported greater perceived necessity of the assigned treatment regimen, reported stronger perceived control over diabetes, and endorsed fewer total depression symptoms (in particular somatic symptoms on the depression scale) and less sleep disturbance and concentration difficulty (P < 0.05 for all). These analyses also showed that, relative to liraglutide, those assigned to sitagliptin had higher HbA1c, were more likely to meet primary and secondary glycemic outcomes, reported fewer GI side effects, and reported greater perceived necessity of treatment (P < 0.05 for all).
Table 3 presents percent change in the treatment effects on adherence after adjustment for the potential mediator in the model. A significant effect of the mediator combined with an attenuated effect of treatment group on adherence indicates that the mediator variable partially mediated the treatment effect on adherence (35). Based on these criteria, the relatively lower 36-month adherence in the liraglutide versus glimepiride or sitagliptin groups was partially explained by relatively increased GI side effects and decreased perceived necessity of treatment. The treatment effect on adherence for the glimepiride compared with the liraglutide group was additionally partially mediated by somatic depressive symptoms. These differences in 36-month adherence were attenuated on the order of 6%–19% after adjustment for the mediator variable. Importantly, although HbA1c differed between treatments and was associated with adherence at 36 months, it did not attenuate the marginal difference in adherence at 36 months for liraglutide as compared with glimepiride and sitagliptin. Instead, adjustment for HbA1c over the follow-up increased the treatment effect on adherence for glimepiride versus liraglutide by 18% and for sitagliptin versus liraglutide by 32%, indicating effect suppression by HbA1c. Similar suppression effects were found in examining primary and secondary glycemic outcomes as mediators for the difference between sitagliptin and liraglutide. Hypoglycemia also showed a suppression effect in examining the difference between glimepiride and liraglutide. Together, these results show that adjustment for differences in glycemia resulted in larger estimated differences in medication adherence, rather than attenuating differences.
Table 3.
Relative change of estimates of treatment effect on medication adherence at 36 months, after adjustment for potential mediators
| Treatment effect | |||||
|---|---|---|---|---|---|
| Glimepiride vs. liraglutide | Sitagliptin vs. liraglutide | ||||
| Coef. (PC) | P | Coef. (PC) | P | Mediator effect, Wald P | |
| Unadjusted model | 0.0507 | 0.0030 | 0.0342 | 0.050 | — |
| Treatment effect suppressors* | |||||
| HbA1c | 0.0599 (17.96) | 0.0004 | 0.0451 (32.02) | 0.008 | <0.0001 |
| Primary treatment outcome | — | — | 0.0386 (12.90) | 0.027 | 0.0001 |
| Secondary treatment outcome | — | — | 0.0393 (15.03) | 0.023 | <0.0001 |
| Hypoglycemia | 0.0520 (2.42) | 0.0030 | — | — | 0.7480 |
| Treatment effect mediators† | |||||
| GI side effects | 0.0476 (−6.17) | 0.0060 | 0.0308 (−9.88) | 0.077 | 0.0040 |
| Beliefs about medication necessity | 0.0449 (−11.52) | 0.0100 | 0.0278 (−18.76) | 0.111 | 0.0002 |
| PHQ-8 total score | 0.0464 (−8.58) | 0.0070 | — | — | <0.0001 |
| Somatic score | 0.0460 (−9.44) | 0.0080 | — | — | <0.0001 |
| Sleep disturbance | 0.0477 (−6.02) | 0.0060 | — | — | 0.0010 |
| Concentration difficulties | 0.0467 (−8.17) | 0.0070 | — | — | <0.0001 |
Linear regression models with adjustment for each potential mediator separately were used to assess whether the inclusion of the mediator explains the relationship between treatment group and medication adherence. This table only includes potential mediators with significant treatment differences (based on results from Supplementary Table 3). Regression coefficients (Coef.) for the treatment group comparisons are shown first unadjusted (without mediators) and then adjusted for potential mediators one at a time, with the percent change (PC) (% mediation) of the coefficients between the two models. All results are based on unadjusted linear models for log medication adherence at 36 months. —, mediator did not meet the criteria for mediation analysis in this specific pairwise comparison and was therefore not included. Boldface indicates significance at P ≤ 0.05.
*Adjustment for these variables increased the treatment effects on medication adherence in comparison with the marginal model.
†Adjustment for these variables attenuated the treatment effects on medication adherence in comparison with the marginal model, indicating partial mediation.
Medication Adherence Over Follow-up as a Predictor of Glycemic Outcomes and Treatment Group Differences
Across treatment groups, a 10-point increase in adherence score (on the original 0–100 scale) was associated with 13% and 16% decreased risk of reaching primary (HbA1c ≥7.0) and secondary (HbA1c >7.5) glycemic outcomes, respectively (both P < 0.0001) (Supplementary Table 4, first data row). We found no evidence for treatment group effect modification for the primary glycemic outcome (Supplementary Table 4, second data row). For the secondary glycemic outcome, a significant (P = 0.03) interaction was found between adherence and treatment group (Supplementary Table 4, second data row). Increases in adherence were significantly associated with reduced risk for reaching the secondary glycemic outcome for those assigned to glargine or liraglutide, compared with glimepiride or sitagliptin (Supplementary Table 4). As seen in Fig. 1, this differential risk for reaching the secondary glycemic outcome by treatment group was most apparent at lower levels of adherence, where we have relatively fewer cases (see also Supplementary Table 5), leading to wide confidence margins around these estimates. No other treatment group comparisons were significant for the differential association of adherence with the secondary glycemic outcome.
Figure 1.
Risk of primary/secondary treatment failure by month 36 across values of medication adherence by treatment group. A: Predicted risk of treatment failure by month 36 across levels of medication adherence on the original scale. Solid lines, point estimates of the relative risk of experiencing primary or secondary treatment failure at month 36; shaded areas, 95% CIs, and dashed lines, their upper and lower bounds. B: Number of participants with adherence data at each study visit by treatment group.
Conclusions
The current findings from the EDS sample of GRADE participants suggest that adherence to assigned glucose-lowering medications was high over time and differed minimally across treatments. The only statistically significant difference emerged after 3 years of treatment and was not consistent with our hypothesis. We expected that adherence would decrease over time and did find a statistically significant decrease across all groups over the follow-up. However, the magnitude of this decrease—approximately 2 points on a scale that ranged from 0 to 100—was of unclear clinical significance. This high level of adherence supports the validity of previously reported trial results on glycemic outcomes from GRADE (23). It also represents a much higher level of sustained adherence than typically seen in cohort studies. For example, a systematic review of 31 cohorts comprising 123,854 individuals showed that adherence to sodium–glucose cotransporter 2 inhibitors decreased markedly over time: 59.5% were adherent at 6 months and 49% were adherent at 1 year (35). This difference underscores the importance of adherence in explaining the gap between outcomes achieved in clinical trials and those observed in clinical practice (7).
We hypothesized that adherence would be lower for participants randomizly assigned to glargine based on prior correlational literature (14–19) but found no evidence to support this expectation. We did find significantly lower levels of adherence for the liraglutide group at 36 months, compared with either the sitagliptin or glimepiride groups. However, the magnitude of these differences is of unclear clinical significance. That said, there is evidence suggesting that nonadherence to liraglutide and other GLP-1 medications is substantial over time (1,36), and it is possible that the differences observed here could increase over longer follow-up.
Examination of potential mediators of observed adherence differences at 36 months suggested that stronger perceived necessity of treatment for the glimepiride and sitagliptin treatment groups in comparison with the liraglutide group accounted for the most substantial portion of the difference in adherence between these groups. More depressive symptoms in the liraglutide group, particularly somatic symptoms, relative to these two other treatment groups also explained a portion of these differences. GI symptoms were also more commonly reported in the liraglutide group compared with glimepiride or sitagliptin and accounted for a portion of the differences between these treatment groups, particularly for the difference between liraglutide and sitagliptin. Although hypoglycemia was more common among those assigned to glimepiride than with liraglutide, adjustment for hypoglycemia enhanced rather than attenuated the difference in medication adherence between the glimepiride and liraglutide groups at 36 months. There was no significant difference in hypoglycemia between the sitagliptin and liraglutide groups. Finally, although HbA1c was significantly higher over the follow-up in the glimepiride and sitagliptin groups as compared with those assigned to liraglutide, consistent with the glycemic outcomes from the overall GRADE cohort (23), adjustment for this effect increased the magnitude of the adherence difference for the liraglutide group relative to glimepiride and sitagliptin by approximately 18% and 32%, respectively. Thus, differences in glycemia did not mediate observed medication adherence differences. Rather, results suggest that had the treatment groups been equal in their effects on glycemia, the expected differences in medication adherence would be significantly larger. We also found no evidence for treatment intensification over time as a mediator of adherence differences at 3 years.
Adherence was a robust predictor of primary and secondary glycemic outcomes for GRADE, and these effects were significant over and above the effects of treatment and despite the overall high levels of adherence. This finding highlights the potential clinical utility of the simple and free-to-use three-item adherence measure used here and in prior studies (25,26). No differences between treatment groups were seen in the association between adherence and primary glycemic outcomes. However, for the secondary glycemic outcome, we found a significant difference between liraglutide and glargine on the one hand and sitagliptin and glimepiride on the other. Risk of reaching the secondary outcome by 3 years was higher for participants reporting low adherence to liraglutide and glargine than for those reporting low adherence to glimepiride or sitagliptin. This suggests that the glycemic consequences of nonadherence may differ between these regimens over time. Further research on the possible differential effects of nonadherence to glucose lowering medications on glycemic outcomes is warranted.
The results of this study should be considered in the context of its design. First, our brief measure of self-reported adherence provides a subjective rating that likely underestimates the true extent of nonadherence. Though it has been validated in HIV/AIDS (25,26) and young adults with youth-onset T2DM (37), a minimum clinically important difference has not been established. Available evidence suggests a linear relationship between adherence scale scores and pill counts, as well as indices of viral load in HIV/AIDS and glycemic control in T2DM (26,37). Although adherence self-reports are often viewed skeptically because of concerns of biases including those related to social desirability and memory, evidence suggests that validity of these measures is comparable with that of more complex and costly alternatives, such as electronic monitoring of pill bottle openings, which are also vulnerable to measurement biases (38–40). Second, levels of adherence were quite high in this study and relatively few participants reported low levels of adherence. While this is a positive from the perspective of interpreting GRADE outcomes (23), lack of variability in adherence may have limited our ability to find relationships. Third, our adherence measure could not distinguish between adherence to the different glucose-lowering medications comprising the participant’s treatment regimen. Previously published data from GRADE indicate that 69% of expected participants initiated rescue glargine and 44% rescue aspart, as indicated by the GRADE protocol (41). Participants not initiating rescue insulin when indicated per protocol would likely not have considered this as nonadherence in their self-reports, leading to an underestimate of the problems with acceptance and adherence to rescue insulin over time. Fourth, enrollment in GRADE required that participants be willing to take injectable medications, treatment was provided at no cost to participants, and substantial support was provided by study staff; thus, our results are not reflective of the usual clinical setting and should be replicated in other studies with examination of real-world adherence to these regimens in other clinical and community samples where nonadherence would be expected to be more prevalent. Fifth, our subsample size, though large, was not a priori powered based on these planned analyses, and the follow-up period was limited to 3 years due to the timing of our substudy start relative to the start of the GRADE recruitment. Finally, because this was a secondary analysis, pairwise post hoc comparisons were conducted using unadjusted P values without correction for multiple testing. This approach is consistent with other secondary analyses from GRADE and preserves interpretability of treatment differences but may increase the risk of type I error.
In conclusion, results of this study document very high levels of adherence in GRADE that robustly predicted glycemic outcomes. These findings highlight the importance of assessing and addressing problems with medication adherence as part of routine T2DM care to ensure maximal glycemic benefits. Where minor treatment group differences were found, they suggested that liraglutide may lead to worse adherence over the relatively long term (3 years after initiation) and that liraglutide and insulin glargine might be less “forgiving” of nonadherence, with respect to glycemic deterioration, as compared with sitagliptin and glimepiride. The extent to which this should change treatment choices is unclear because glargine and liraglutide were also found to have better effects on glycemic outcomes in comparison with the alternatives examined in GRADE (23).
This article contains supplementary material online at https://doi.org/10.2337/figshare.30657905.
Article Information
Acknowledgments. The GRADE Research Group is deeply grateful to their participants, whose loyal dedication made GRADE possible.
The content of this manuscript is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Duality of Interest. J.S.G. reports participation on a data safety monitoring or advisory board for Vanderbilt University, outside of the submitted work. N.M.B. reports participation on a data safety monitoring board for Fast Antibiotic Susceptibility Testing for Gram Negative Bacteremia Trial (FAST), outside of the submitted work. D.J.W. reports participation on a data safety monitoring board for Novo Nordisk Semaglutide cardiOvascular oUtcomes trial (SOUL) and Evaluate Renal Function with Semaglutide Once Weekly (FLOW) trials, outside of the submitted work. C.J.H. reports support from grants 2PO1 AG003949 (National Institute on Aging), P30DK111022P (NIDDK), P30DK111022F (NIDDK), 4-SRA-2021-1071-M-B (JDRF), and 4-SRA-2022-1187-M-B (JDRF) outside of the submitted work. C.P. reports support from National Center for Complementary and Integrative Health and American Diabetes Association grants, outside of the submitted work. No other potential conflicts of interest relevant to this article were reported.
Author Contributions. All authors affirmed that authorship is merited based on the International Committee of Medical Journal Editors (ICMJE) authorship criteria. J.S.G., H.K.-S., D.J.W., C.J.H., V.L., and A.L.C. contributed to the conception and design of the research. D.J.W., H.P., B.F., and V.L. contributed to the acquisition of data. H.W. and D.U. contributed to the statistical analysis of data. J.S.G., H.W., N.M.B., D.U., H.K.-S., M.R.G., D.J.W., C.J.H., G.C.-R., C.P., and E.A.W. contributed to the interpretation of data and results. J.S.G., H.K.-S., and E.A.W. contributed to the acquisition of funding. J.S.G., H.K.-S., M.R.G., and B.F. contributed to the supervision and management of research. J.S.G., H.W., and H.K.-S. drafted the manuscript. All authors critically reviewed the manuscript for important intellectual content and approved the final version to be published. J.S.G. and N.M.B. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Prior Presentation. Parts of this work were presented at the 85th Scientific Sessions of the American Diabetes Association, Chicago, IL, 20–23 June 2025.
Handling Editors. The journal editor responsible for overseeing the review of the manuscript was Stephen S. Rich.
Funding Statement
GRADE is supported by a grant from the NIDDK of the National Institutes of Health under award no. U01DK098246. The planning of GRADE was supported by a U34 planning grant from the NIDDK (U34-DK-088043). The American Diabetes Association supported the initial planning meeting for the U34 proposal. The National Heart, Lung, and Blood Institute and the Centers for Disease Control and Prevention also provided funding support. The Department of Veterans Affairs provided resources and facilities. Additional support was provided by National Institutes of Health grants P30 DK017047, P30 DK020541-44, P30 DK020572, P30 DK072476, P30 DK079626, P30 DK092926, U54 GM104940, UL1 TR000439, UL1 TR000445, UL1 TR001108, UL1 TR001409, UL1 TR001449, UL1 TR002243, UL1 TR002345, UL1 TR002378, UL1 TR002489, UL1 TR002529, UL1 TR002535, UL1 TR002537, and UL1 TR002548. Educational materials have been provided by the National Diabetes Education Program. Material support in the form of donated medications and supplies has been provided by Becton, Dickinson and Company; Bristol-Myers Squibb; Merck; Novo Nordisk; Roche Diagnostics; and Sanofi. EDS was supported by a grant from the NIDDK of the National Institutes of Health under award no. R01 DK104845. Additional support was provided by National Institutes of Health grant P30 DK111022.
Footnotes
Clinical trial reg. no. NCT01794143, clinicaltrials.gov
*A list of members of the GRADE Research Group can be found in supplementary material online.
This article is part of a special article collection available at https://diabetesjournals.org/collection/2066/reports-from-the-grade-study.
Contributor Information
Jeffrey S. Gonzalez, Email: grademail@bsc.gwu.edu.
GRADE Research Group*:
J.P. Crandall, M.D. McKee, S. Behringer-Massera, J. Brown-Friday, E. Xhori, K. Ballentine-Cargill, S. Duran, H. Estrella, S. Gonzalez de la torre, J. Lukin, L.S. Phillips, E. Burgess, D. Olson, M. Rhee, P. Wilson, T.S. Raines, J. Boers, J. Costello, M. Maher-Albertelli, R. Mungara, L. Savoye, C.A. White, C. Gullett, L. Holloway, F. Morehead, S. Person, M. Sibymon, S. Tanukonda, C. Adams, A. Ross, A. Balasubramanyam, R. Gaba, E. Gonzalez Hattery, A. Ideozu, J. Jimenez, G. Montes, C. Wright, F. Ismail-Beigi, C. Falck-Ytter, L. Sayyed Kassem, A. Sood, M. Tiktin, T. Kulow, C. Newman, K.A. Stancil, B. Cramer, J. Iacoboni, M.V. Kononets, C. Sandsers, L. Tucker, A. Werner, A. Maxwell, G. McPhee, C. Patel, L. Colosimo, A. Krol, R. Goland, J. Pring, L. Alfano, P. Kringas, C. Hausheer, J. Tejada, K. Gumpel, A. Kirpitch, H. Schneier, C.N. Mariash, K.J. Mather, H.M. Ismail, A. Lteif, M. Mullen, T. Hamilton, N. Patel, G. Riera, M. Jackson, V. Pirics, D. Aguillar, D. Howard, S. Hurt, R. Bergenstal, A. Carlson, T. Martens, M. Johnson, R. Hill, J. Hyatt, C. Jensen, M. Madden, D. Martin, H. Willis, W. Konerza, S. Yang, K. Kleeberger, R. Passi, S. Fortmann, M. Herson, K. Mularski, H. Glauber, J. Prihoda, B. Ash, C. Carlson, P.A. Ramey, E. Schield, B. Torgrimson-Ojerio, K. Arnold, B. Kauffman, E. Panos, S. Sahnow, K. Bays, K. Berame, J. Cook, D. Ghioni, J. Gluth, K. Schell, J. Criscola, C. Friason, S. Jones, S. Nazarov, D. Wexler, M.E. Larkin, J. Meigs, B. Chambers, A. Dushkin, G. Rocchio, M. Yepes, B. Steiner, H. Dulin, M. Cayford, K. Chu, A. DeManbey, M. Hillard, K. Martin, N. Thangthaeng, L. Gurry, R. Kochis, E. Raymond, V. Ripley, C. Stevens, J. Park, V. Aroda, A. Ghazi, M. Magee, A. Ressing, A. Loveland, M. Hamm, M. Hurtado, A. Kuhn, J. Leger, L. Manandhar, F. Mwicigi, O. Sanchez, T. Young, R. Garg, V. Lagari-Libhaber, H.J. Florez, W.M. Valencia, J. Marks, S. Casula, L. Oropesa-Gonzalez, L. Hue, A. Cuadot, R. Nieto-Martinez, A.K. Riccio Veliz, M. Gutt, Y.J. Kendal, B. Veciana, S.H. Hox, H. Petrovitch, M. Matwichyna, V. Jenkins, L. Broadwater, R.R. Ishii, N.O. Bermudez, D.S. Hsia, W.T. Cefalu, F.L. Greenway, C. Waguespack, E. King, G. Fry, A. Dragg, B. Gildersleeve, J. Arceneaux, N. Haynes, A. Thomassie, M. Pavlionis, B. Bourgeois, C. Hazlett, J. Krakoff, J.M. Curtis, T. Killean, M. Khalid, E. Joshevama, E. Diaz, D. Martin, K. Tsingine, T. Karshner, M.A. Banerji, P. August, M. Lee, D. Lorber, N.M. Brown, D.H. Josephson, L.L. Thomas, M. Tsovian, A. Cherian, M.H. Jacobson, M.M. Mishko, M.S. Kirkman, J.B. Buse, J. Diner, J. Dostou, S. Machineni, L. Young, K. Bergamo, A. Goley, J. Kerr, J.F. Largay, S. Guarda, J. Cuffee, D. Culmer, R. Fraser, H. Almeida, S. Coffer, E. Debnam, L. Kiker, S. Morton, K. Josey, G. Fuller, W.T. Garvey, A.L. Cherrington, D. Dyer, M.C.R. Lawson, O. Griffith, A. Agne, S. McCullars, R.M. Cohen, J. Craig, M.C. Rogge, K. Burton, K. Kersey, C. Wilson, S. Lipp, M.B. Vonder Meulen, C. Adkins, T. Onadeko, N. Rasouli, C. Baker, E. Schroeder, M. Razzaghi, C. Lyon, R. Penaloza, C. Underkofler, R. Lorch, S. Douglass, S. Steiner, W.I. Sivitz, E. Cline, L.K. Knosp, J. McConnell, T. Lowe, W.H. Herman, R. Pop-Busui, M.H. Tan, C. Martin, A. Waltje, A. Katona, L. Goodhall, R. Eggleston, S. Kuo, S. Bojescu, S. Bule, N. Kessler, E. LaSalle, K. Whitley, E.R. Seaquist, A. Bantle, T. Harindhanavudhi, A. Kumar, B. Redmon, J. Bantle, M. Coe, M. Mech, A. Taddese, L. Lysne, S. Smith, C. Desouza, L. Kuechenmeister, V. Shivaswamy, S. Burbach, M.G. Rodriguez, K. Seipel, A. Alfred, A.L. Morales, J. Eggert, G. Lord, W. Taylor, R. Tillson, D.S. Schade, A. Adolphe, M. Burge, E. Duran-Valdez, J. Martinez, A. Bancroft, S. Kunkel, F. Ali Jamaleddin Ahmad, D. Hernandez McGinnis, B. Pucchetti, E. Scripsick, A. Zamorano, R.A. DeFronzo, E. Cersosimo, M. Abdul-Ghani, C. Triplitt, D. Juarez, R.I. Garza, H. Verastiqui, K. Wright, C. Puckett, P. Raskin, C. Rhee, S. Abraham, L.F. Jordan, S. Sao, L. Morton, O. Smith, L. Osornio Walker, L. Schnurr-Breen, R. Ayala, R.B. Kreymer, D. Sturgess, K.M. Utzschneider, S.E. Kahn, L. Alarcon-Casas Wright, E.J. Boyko, E.C. Tsai, D.L. Trence, S. Trikudanathan, B.N. Fattaleh, B.K. Montgomery, K.M. Atkinson, A. Kozedub, T. Concepcion, C. Moak, N. Prikhodko, S. Rhothisen, W. Tamborlane, A. Camp, B. Gulanski, S.E. Inzucchi, K. Pham, M. Alguard, P. Gatcomb, K. Lessard, M. Perez, L. Iannone, E. Magenheimer, A. Montosa, J. Fradkin, H.B. Burch, A.A. Bremer, D.M. Nathan, J.M. Lachin, H. Krause-Steinrauf, N. Younes, I. Bebu, N. Butera, C.J. Buys, A. Fagan, Y. Gao, A. Ghosh, M.R. Gramzinski, S.D. Hall, E. Kazemi, E. Legowski, H. Liu, C. Suratt, M. Tripputi, A. Arey, M. Backman, J. Bethepu, C. Lund, P. Mangat Dhaliwal, P. McGee, E. Mesimer, L. Ngo, M. Steffes, J. Seegmiller, A. Saenger, V. Arends, D. Gabrielson, T. Conner, S. Warren, J. Day, J. Huminik, A. Scrymgeour, I. Abdouch, G. Bahtiyar, P. Brantley, F.E. Broyles, G. Canaris, P. Copeland, J.J. Craine, W.L. Fein, A. Gliwa, L. Hope, M.S. Lee, R. Meiners, V. Meiners, H. O’Neal, J.E. Park, A. Sacerdote, E. Sledge Jr., L. Soni, J. Steppel-Reznik, A. Turchin, S. Golden, J. Gonzalez, A. Naik, and E. Walker
Supporting information
References
- 1. Lassen MCH, Johansen ND, Modin D, et al. Adherence to glucagon-like peptide-1 receptor agonist treatment in type 2 diabetes mellitus: a nationwide registry study. Diabetes Obes Metab 2024;26:5239–5250 [DOI] [PubMed] [Google Scholar]
- 2. Polonsky WH, Henry RR. Poor medication adherence in type 2 diabetes: recognizing the scope of the problem and its key contributors. Patient Prefer Adherence 2016;10:1299–1307 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Bonafede MMK, Kalsekar A, Pawaskar M, et al. A retrospective database analysis of insulin use patterns in insulin-naïve patients with type 2 diabetes initiating basal insulin or mixtures. Patient Prefer Adherence 2010;4:147–156 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Khunti K, Seidu S, Kunutsor S, Davies M. Association between adherence to pharmacotherapy and outcomes in type 2 diabetes: a meta-analysis. Diabetes Care 2017;40:1588–1596 [DOI] [PubMed] [Google Scholar]
- 5. Nguyen D, Liang X, Nguyen S, et al. Association between insulin adherence and hemoglobin A1c: a systematic review and meta-analysis. J Manag Care Spec Pharm 2024;30:1167–1177 [Google Scholar]
- 6. Simpson SH, Lin M, Eurich DT. Medication adherence affects risk of new diabetes complications: a cohort study. Ann Pharmacother 2016;50:741–746 [DOI] [PubMed] [Google Scholar]
- 7. Edelman SV, Polonsky WH. Type 2 diabetes in the real world: the elusive nature of glycemic control. Diabetes Care 2017;40:1425–1432 [DOI] [PubMed] [Google Scholar]
- 8. Gossec L, Tubach F, Dougados M, Ravaud P. Reporting of adherence to medication in recent randomized controlled trials of 6 chronic diseases: a systematic literature review. Am J Med Sci 2007;334:248–254 [DOI] [PubMed] [Google Scholar]
- 9. Assawasuwannakit P, Braund R, Duffull SB. A model-based meta-analysis of the influence of factors that impact adherence to medications. J Clin Pharm Ther 2015;40:24–31 [DOI] [PubMed] [Google Scholar]
- 10. Claxton AJ, Cramer J, Pierce C. A systematic review of the associations between dose regimens and medication compliance. Clin Ther 2001;23:1296–1310 [DOI] [PubMed] [Google Scholar]
- 11. Leslie KH, McCowan C, Pell JP. Adherence to cardiovascular medication: a review of systematic reviews. J Public Health (Oxf) 2019;41:E84–E94 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Grant RW, Devita NG, Singer DE, Meigs JB. . Polypharmacy and medication adherence in patients with type 2 diabetes. Diabetes Care 2003;26:1408–1412 [DOI] [PubMed] [Google Scholar]
- 13. Watanabe JH, Bounthavong M, Chen T, Ney JP. . Association of polypharmacy and statin new-user adherence in a Veterans Health Administration population: a retrospective cohort study. Ann Pharmacother 2013;47:1253–1259 [DOI] [PubMed] [Google Scholar]
- 14. Davies MJ, Gagliardino JJ, Gray LJ, Khunti K, Mohan V, Hughes R. Real-world factors affecting adherence to insulin therapy in patients with type 1 or type 2 diabetes mellitus: a systematic review. Diabet Med 2013;30:512–524 [DOI] [PubMed] [Google Scholar]
- 15. Skriver LKL, Nielsen MW, Walther S, Nørlev JD, Hangaard S. Factors associated with adherence or nonadherence to insulin therapy among adults with type 2 diabetes mellitus: a scoping review. J Diabetes Complications 2023;37:108596. [DOI] [PubMed] [Google Scholar]
- 16. Krall J, Gabbay R, Zickmund S, Hamm ME, Williams KR, Siminerio L. Current perspectives on psychological insulin resistance: primary care provider and patient views. Diabetes Technol Ther 2015;17:268–274 [DOI] [PubMed] [Google Scholar]
- 17. Polonsky WH, Fisher L, Guzman S, Villa-Caballero L, Edelman SV. Psychological insulin resistance in patients with type 2 diabetes: the scope of the problem. Diabetes Care 2005;28:2543–2545 [DOI] [PubMed] [Google Scholar]
- 18. Sanz-Pastor AG, Enríquez AJ, Bao AS, Ampudia-Blasco FJ. Current barriers to initiating insulin therapy in individuals with type 2 diabetes. Front Endocrinol (Lausanne) 2024;15:1366368 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Snoek FJ, Fisher L, Polonsky WH, et al. Overcoming psychological insulin resistance: a practical guide for healthcare professionals. Prim Care Diabetes 2021;15:619–621 [DOI] [PubMed] [Google Scholar]
- 20. McGovern A, Hinton W, Calderara S, Munro N, Whyte M, de Lusignan S. A class comparison of medication persistence in people with type 2 diabetes: a retrospective observational study. Diabetes Ther 2018;9:229–242 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Cherrington AL, Krause-Steinrauf H, Bebu I, et al.; GRADE Research Group . Study of emotional distress in a comparative effectiveness trial of diabetes treatments: rationale and design. Contemp Clin Trials 2021;107:106366. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Nathan DM, Buse JB, Kahn SE, et al.; GRADE Study Research Group . Rationale and design of the Glycemia Reduction Approaches in Diabetes: A Comparative Effectiveness Study (GRADE). Diabetes Care 2013;36:2254–2261 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. GRADE Study Research Group; Nathan DM, Lachin JM, Balasubramanyam A, et al. Glycemia reduction in type 2 diabetes - glycemic outcomes. N Engl J Med 2022;387:1063–1074 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Nathan DM, Lachin JM, Bebu I, et al. . Glycemia reduction in type 2 diabetes- microvascular and cardiovascular outcomes. N Engl J Med 2022;387:1075–1088 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Wilson IB, Lee Y, Michaud J, Fowler FJ Jr, Rogers WH. Validation of a new three-item self-report measure for medication adherence. AIDS Behav 2016;20:2700–2708 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Wilson IB, Tie Y, Padilla M, Rogers WH, Beer L. Performance of a short, self-report adherence scale in a probability sample of persons using HIV antiretroviral therapy in the United States. Aids 2020;34:2239–2247 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. GRADE Study Research Group; Nathan DM, Lachin JM, Bebu I, et al. Glycemia reduction in type 2 diabetes - microvascular and cardiovascular outcomes. N Engl J Med 2022;387:1075–1088 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Horne R, Weinman J, Hankins M. The beliefs about medicines questionnaire: the development and evaluation of a new method for assessing the cognitive representation of medication. Psychol Health 1999;14:1–24 [Google Scholar]
- 29. Wallston KA, Rothman RL, Cherrington A. Psychometric properties of the Perceived Diabetes Self-Management Scale (PDSMS). J Behav Med 2007;30:395–401 [DOI] [PubMed] [Google Scholar]
- 30. Moss-Morris R, Weinman J, Petrie K, Horne R, Cameron L, Buick D. The revised Illness Perception Questionnaire (IPQ-R). Psychol Health 2002;17:1–16 [Google Scholar]
- 31. Bradley C, Plowright R, Stewart J, Valentine J, Witthaus E. The Diabetes Treatment Satisfaction Questionnaire change version (DTSQc) evaluated in insulin glargine trials shows greater responsiveness to improvements than the original DTSQ. Health Qual Life Outcomes 2007;5:57. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Kroenke K, Strine TW, Spitzer RL, Williams JBW, Berry JT, Mokdad AH. The PHQ-8 as a measure of current depression in the general population. J Affect Disord 2009;114:163–173 [DOI] [PubMed] [Google Scholar]
- 33. Polonsky WH, Fisher L, Earles J, et al. Assessing psychosocial distress in diabetes: development of the Diabetes Distress Scale. Diabetes Care 2005;28:626–631 [DOI] [PubMed] [Google Scholar]
- 34. R Core Team . R: A Language and Environment for Statistical Computing. Vienna, Austria, R Foundation for Statistical Computing, 2023 [Google Scholar]
- 35. Ofori-Asenso R, Sahle BW, Chin KL, et al. Poor adherence and persistence to sodium glucose co-transporter 2 inhibitors in real-world settings: evidence from a systematic review and meta-analysis. Diabetes Metab Res Rev 2021;37:e3350. [DOI] [PubMed] [Google Scholar]
- 36. Weiss T, Carr RD, Pal S, et al. Real-world adherence and discontinuation of glucagon-like peptide-1 receptor agonists therapy in type 2 diabetes mellitus patients in the United States. Patient Prefer Adherence 2020;14:2337–2345 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Trief PM, Burke B, Kalichman S, et al. Brief report: validity of self-report measures of medication adherence in young adults with youth-onset type 2 diabetes. Diabetes Res Clin Pract 2025;224:112201. [DOI] [PubMed] [Google Scholar]
- 38. Gonzalez JS, Schneider HE. Methodological issues in the assessment of diabetes treatment adherence. Curr Diab Rep 2011;11:472–479 [DOI] [PubMed] [Google Scholar]
- 39. Stirratt MJ, Dunbar-Jacob J, Crane HM, et al. Self-report measures of medication adherence behavior: recommendations on optimal use. Transl Behav Med 2015;5:470–482 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Gonzalez JS, Schneider HE, Wexler DJ, et al. Validity of medication adherence self-reports in adults with type 2 diabetes. Diabetes Care 2013;36:831–837 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. Hollander PA, Krause-Steinrauf H, Butera NM, et al.; GRADE Research Group . The use of rescue insulin in the Glycemia Reduction Approaches in Diabetes: A Comparative Effectiveness Study (GRADE). Diabetes Care 2024;47:638–645 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.