Correlates of Medication Adherence in the TODAY Cohort of Youth With Type 2 Diabetes

The TODAY (Treatment Options for Type 2 Diabetes in Adolescents and Youth) clinical trial was conducted to evaluate three different treatments for type 2 diabetes in youth (1). The results of the TODAY trial revealed that 45.6% of participants, all with recently diagnosed (<2 years) youth-onset type 2 diabetes, lost glycemic control on randomized treatment usually by the end of the first year (2).

Youth in the TODAY trial were expected to take two pills twice a day. After randomization, the TODAY trial used a predetermined 80% adherence cutoff, which is commonly applied in clinical trials (3), to monitor the adequacy of adherence to study medication during the trial. Rapoff (4) has discussed how adherence (or lack of) can bias clinical trials of promising therapies, especially in pediatric clinical trials. Most pediatric type 1 diabetes studies (5–7) consistently document a correlation between adherence and race, ethnicity, and socioeconomic status, and studies of adults with type 2 diabetes (8,9) have documented that depressed patients are less adherent to their diabetes regimen. There is a dearth of information in the literature regarding adherence to medication in pediatric patients with type 2 diabetes. One report (10) from a study of youth with type 2 diabetes at three clinical sites concluded that “compliance with medications and doctor’s appointments is suboptimal in youth with type 2 diabetes.” The objective of the current analysis was to identify factors that predicted medication adherence and to examine relationships among adherence, glycemic control, and indices of insulin action in the TODAY cohort.

The collaborative study group included 15 clinical centers, a data coordinating center, and central laboratories and reading centers (Supplementary Data). Materials developed and used for the TODAY trial standard diabetes education program and the intensive lifestyle intervention program are available to the public at https://today.bsc.gwu.edu/. The TODAY study design has been reported (11) and is briefly summarized. Between July 2004 and February 2009, 699 youths between the ages of 10 and 17 years with type 2 diabetes were enrolled. To be eligible, all TODAY trial participants had to have received a diagnosis of type 2 diabetes using American Diabetes Association criteria <2 years before the time of randomization; have a BMI in ≥85th percentile; and have an adult caregiver (usually the mother) who agreed to support the youth participating in the study, including accompanying the youth to all visits and helping with diabetes tasks such as medication adherence. Youths had to take ≥80% of their metformin (M) for ≥8 weeks during a run-in period in order to be eligible for randomization. A total of 927 subjects entered the run-in phase, and 699 subjects were randomized and assigned to a treatment group. Eligible participants were randomized to one of the following three treatment arms: 1) M alone (M), 2) M plus rosiglitazone (M+R), and 3) M plus an intensive lifestyle program (M+L). The primary objective of the TODAY trial was to compare the three arms on time to treatment failure (i.e., loss of glycemic control, defined as either an HbA_1c level of ≥8% over a 6-month period or an inability to wean from temporary insulin therapy within 3 months after acute metabolic decompensation). After an average follow-up period of 3.9 years, 319 subjects (45.6%) reached the primary outcome; the M+R arm was superior to the M arm (P = 0.006), and the M+L arm was intermediate but not different from M arm (2).

The dose for all treatment arms was two capsules twice daily. Masked study drug (M or M+R) was provided in 7-day blister packs separated into morning and evening two-pill doses. M and R were encapsulated together. All pills looked, smelled, and tasted the same. Study subjects were instructed to return all blister packs at their regular study visit. HbA_1c values were partially masked to the subject and the investigators. Investigators were not informed of the value, but were notified if the HbA_1c level was on target, stable, rising, or elevated.

For this analysis, adherence was measured while study subjects received randomized treatment (i.e., prior to the primary outcome [loss of glycemic control] or at the end of the study visit). Collection of adherence data (pill counts) and dispensing of the study drug occurred at each study visit (every 2 months in year 1 and then quarterly). Adherence was calculated as the percentage of the prescribed study drug taken, based on pill counts. If pill packs (empty, partial, or full) were not returned at a visit as instructed, then adherence could not be determined and was noted as missing. Adherence could be >100% if the number of pills taken based on empty containers brought to a visit was greater than the prescribed four pills per day. Outlier values were examined, and values >110% were excluded from the analysis.

At each visit, study staff evaluated and discussed adherence with the participant. There was no standardized behavioral intervention to address adherence. Based on adherence barriers identified in discussion with the participant and caregiver, study personnel worked with the participant on strategies to increase medication adherence. Participants could earn “points” for goal attainment, including medication adherence of ≥80%. Participants could earn up to 12 points per month (6 points for 100% medication adherence, 5 points for 90–99% adherence, or 4 points for 80–89% adherence; 3 points for glucose monitoring; 2 points for bringing back blister packs and a logbook to a visit; and 1 point for setting goals at the visit). Subjects who had <80% adherence did not earn points for medication adherence. Accumulated points could be exchanged for incentive items worth up to $150 per year. Adherence was monitored by the study group committees on Procedures Oversight and Retention and Adherence. Clinical centers below a target cutoff of 80% were contacted to address problems and provide support on both participant-specific and site-specific levels.

Race-ethnicity was determined by self-report. Participants were categorized as non-Hispanic black, non-Hispanic white, or Hispanic; categories that were too small for separate analysis were combined into Other (7%) and were not included in analyses by race-ethnicity. Household education was the highest education level attained by the parent/guardian; 15 education categories were collapsed into 4 for purposes of analysis. The annual household income of all persons living in the household in the past year was collected by self-report of family members present at the baseline visit; nine categories were collapsed into three for purposes of analysis. Percentage overweight, the recommended outcome for reporting changes in adiposity in youth, was calculated as percentage over median BMI for age and sex (12–14). Health-related quality of life was measured by youth self-report on the generic scale (PedsQL 3.0) with impaired quality of life defined at a cutoff of 71.19 (15). Depressive symptoms were assessed using either the Children’s Depression Inventory for participants <16 years of age or the Beck Depression Inventory II for those ≥16 years of age (16,17). Total scores were calculated for each instrument; a cutoff score ≥13 on the Children’s Depression Inventory and ≥14 on the Beck Depression Inventory II indicated clinically significant depressive symptoms. The TODAY trial primary outcome, time to treatment failure, indicated durability of glycemic control. Three measures of insulin secretion/sensitivity were derived from the oral glucose tolerance test. Insulin sensitivity was calculated as 1/fasting insulin (1/I_F), also called the insulin inverse; this measure correlates strongly with hyperinsulinemic-euglycemic clamp–derived in vivo insulin sensitivity in obese youth with or without type 2 diabetes (18). The insulinogenic index was calculated as the ratio of the incremental insulin and glucose responses from baseline to 30 min (△I₃₀/△G₃₀) (19). This index reflects similar trends in first-phase insulin from the hyperglycemic clamp in obese youth across the glucose tolerance spectrum (18). The oral disposition index (oDI), a measure of β-cell function relative to insulin sensitivity, was calculated as the product of insulin sensitivity multiplied by the insulinogenic index (1/I_F × △I₃₀/△G₃₀); in obese youth as well as adults, the oDI correlates strongly with clamp-derived disposition index (DI), identifies comparable decrements in β-cell function across the glucose tolerance groups (as does the clamp DI), and has analogous predictive power to that of clamp DI for the 2-h glucose concentration of the oral glucose tolerance test (19,20).

Study medication adherence is typically analyzed as a categorical or dichotomous variable; the continuous distribution was not appropriate even with transformation. Therefore, to be practically useful, a cutoff defining “adequate dose” (i.e., adequate to have the intended effect) is identified for monitoring purposes. The TODAY trial used a preplanned 80% cutoff, which is commonly applied in clinical trials (3) to assess the adequacy of adherence during the trial. The effect of adherence on treatment failure was analyzed using time-to-event survival methods, where time was entered as an interval for those who failed, and adherence and the interaction between adherence and treatment group were included in the model. The analysis was repeated for adherence cutoffs at 60%, 70%, and 90% adherence to determine whether there was a threshold effect. Tests of association between adherence and demographic and baseline characteristics used either the χ² test (for categorical characteristics) or ANOVA (for continuous variables). Generalized linear mixed models were used to compare adherence across treatment groups and over time. Study medication adherence cutoffs were applied to mean adherence prior to treatment failure, while the participant was still receiving the assigned study medication.

Measures of insulin secretion/sensitivity (insulin sensitivity, insulinogenic index, and oDI) were analyzed in models including time (visit number), study medication adherence, and their interaction, with a baseline measure of insulin secretion/sensitivity as a covariate. All measures of insulin secretion/sensitivity were log transformed to normalize the distribution. Significance was defined as P < 0.05 with no adjustment for multiple testing; the trial was powered for the primary outcome only.

Of the 699 subjects in the TODAY cohort, 42 were not included in the analysis (8 never returned after the baseline visit, and 34 reached the primary study outcome within the first 5 months of the study). Figure 1 plots the percentage of subjects with adequate study drug adherence (using the 80% cutoff) across time (months 2–48) by treatment group. As seen, overall adherence declined as a function of the number of months in the study (P < 0.0001) but was not different across treatment groups. When the analysis was repeated for adherence cutoffs at 60%, 70%, and 90%, results were similar to 80% (data not shown).

Baseline data and demographic characteristics by adherence status are shown in Table 1. For each participant, average adherence over time on randomized treatment was dichotomized at the 80% cutoff used during the study to monitor adequate adherence. Participants reporting clinically significant depressive symptoms at baseline were more likely to be in the lower-adherence group. Key characteristics that were not significant included sex, race-ethnicity, determinants of socioeconomic status, and randomized treatment group.

We examined the TODAY trial primary outcome (time-to-failure analysis) with adherence status. The model included the interaction between adherence status and treatment group. The analysis was performed for adherence cutoffs at 60%, 70%, 80%, and 90%. At all four adherence cutoffs, the results by treatment group were similar to the results for 80% adherence reported in the primary outcome article (2) (i.e., the only significant treatment group comparison was M vs. M+R). Failure was not associated with lower adherence to taking medication. Among those subjects who failed therapy, 62.5% had at least an 80% adherence to study medication on average compared with 50.3% in those who did not fail therapy by the end of the study (P = 0.0018).

Figure 2 shows longitudinal data at 6, 24, 36, and 48 months postrandomization and prior to glycemic failure for 1) insulin sensitivity and 2) insulinogenic index by study medication adherence status (cutoff 80%). In this analysis, there was no statistically significant interaction between adherence status and time in the study. On average, ≥80% medication adherence was associated with higher insulin sensitivity (Fig. 2A) (P = 0.0012) and higher oDI (P = 0.0248), but not with insulinogenic index (Fig. 2B) (P = 0.4733). There was a significant trend over time for the insulinogenic index (P = 0.0076) and oDI (P = 0.0307), but not insulin inverse (P = 0.1291). Analysis of medication adherence at other adherence cutoffs revealed similar relationships with insulin secretion/sensitivity.

The TODAY cohort demonstrated deterioration in study medication adherence over time, irrespective of treatment group assignment. Paradoxically, the current analysis found that those who reached the primary outcome actually had higher adherence levels than those who did not reach the primary outcome. Possible explanations include the following: 1) patients who notice high glucose readings were more likely to take their medication and 2) clinicians emphasized medication adherence more strongly and more frequently with participants whose blood glucose levels were higher.

Contrary to expectation, demographic factors (sex, race-ethnicity, household income, and parental educational level) did not predict medication adherence. The lack of correlation with these factors in the TODAY trial may be explained by the limited income and educational range of the families in the TODAY trial. Nearly half of the families in the TODAY trial had an annual income of <$25,000, and, for over half of the families, the highest level of parental education was a high school degree or lower. In addition, our run-in criteria selected for more adherent subjects. All subjects had to have >80% adherence to M therapy for ≥8 weeks before they could be randomized. This may have limited variability in medication adherence postrandomization. It is also possible that selecting for more adherent subjects in the run-in period also selected for subjects with a lower frequency of depressive symptoms.

In the TODAY trial, baseline clinically significant depressive symptoms were more prevalent in the lower-adherence group, suggesting that regular screening for depressive symptoms should be undertaken to identify youth who were at high risk for poor medication adherence. In the TODAY study sample, ∼15% scored at or above the cutoff for clinically significant depressive symptoms (21). These results are similar to those reported in the SEARCH study (22), which found that being female and of older age were risk factors for higher rates of clinically significant depressive symptoms in a pediatric sample of individuals with type 2 diabetes. Studies in adults with type 2 diabetes (23–28) consistently report that depressed patients are less adherent to their diabetes regimen and experience more physical complications of diabetes. Identifying youth who are at risk for poor medication adherence early in the course of disease would make it possible to provide support and, if needed, specific treatment. Although we were not able to determine whether the treatment of depressive symptoms changed adherence over time, our findings support the current guidelines for psychosocial screening in youth with diabetes (29,30). Clinicians may also need to evaluate adherence more carefully in patients with clinically significant depressive symptoms in order to identify and address barriers to adherence.

With regard to the TODAY study primary outcome, we did not find an adherence threshold that predicted loss of glycemic control. Adherence to oral medication was related to higher insulin sensitivity, as expected for the pharmacological mechanisms of action of M+R. However, improved insulin sensitivity was not adequate to compensate for the ongoing decline in β-cell function. These results are consistent with results in adults from the UK Prospective Diabetes Study (UKPDS), in which a continuous decline in β-cell function in adults with type 2 diabetes was seen, irrespective of glucose-lowering treatment (31). Furthermore, the durability of glycemic control in TODAY was not associated with lower adherence to medication (32). Unlike adults with type 2 diabetes, loss of glycemic control in the TODAY trial occurred around the end of the first year in a significant portion of the cohort. During this first year, adherence for the majority of the cohort was ≥80%. This points to the progressive nature of type 2 diabetes seen in approximately half of TODAY youth as well as other factors, including the effects of pubertal hormones. Compared with adult-onset diabetes, youth-onset type 2 diabetes is associated with faster deterioration in glycemic control (33) as well as insulin secretion (34). Other studies (35–38) have reported that youths with type 2 diabetes are also at higher risk for comorbidities (e.g., microalbuminuria and dyslipidemia) earlier in the course of disease progression.

The erosion of medication adherence seen in the TODAY trial is similar to the results of medication adherence deterioration over time in pediatric type 1 diabetes studies (39,40). Literature reviews of clinical trials and clinical practice (41–43) report that rates of adherence for adolescents with chronic illnesses vary, depending on the disease, the complexity of the treatment regimen, and the adherence measures used. However, there is a consensus across studies (4,44) that rates of adherence to medication are generally <50%, especially for adolescents. As incentives were used to improve medication adherence in the TODAY trial, the observed erosion of adherence over time may also relate in part to an initial response to incentives with subsequent habituation.

Limitations of the current report include the use of pill counts for measuring medication adherence and lack of adherence data when visits were missed. Pill counts as a measure of medication adherence often result in an overestimation of the actual number of pills taken. Rapoff (4) has written that this is especially true in pediatric patients. Our criterion of ≥80% adherence for ≥8 weeks prior to randomization may also have selected for a more adherent cohort than is typical for youths with type 2 diabetes. Participants received coaching to improve medication adherence and could earn incentives for better adherence to medication, providing greater opportunity to improve adherence than would be present in general clinical care. Data were not kept on the rate of incentives attained by subjects in the TODAY trial. Although our cohort was relatively homogeneous with regard to demographic factors, the cohort is representative of the pediatric population with diagnosed type 2 diabetes. Although we assessed for clinically significant depressive symptoms, we did not track treatment for identified clinically significant depressive symptoms. The SEARCH and TODAY studies showed a similar sex (female predominance) and ethnic/racial distribution (majority Hispanic or African American), family income (majority making <$50,000), BMI (z-score 2.1), family history of diabetes (>70% in both), and C-peptide level (∼3.5 in both cohorts) (45–47).

In conclusion, medication adherence showed similar declines over time in all three treatment groups and was not related to race-ethnicity or socioeconomic status in this cohort of primarily minority youth characterized by low household income and low parental education levels. The only participant characteristic that was related to low medication adherence was the presence of baseline clinically significant depressive symptoms. We found that no cutoff of medication adherence in the TODAY trial was related to time to treatment failure. Although medication adherence was associated with better insulin sensitivity, it could not compensate for the progressive decline in β-cell function. These results support the literature that type 2 diabetes in many youths runs a progressive course. For patients whose glycemic control is deteriorating while receiving therapy with M, the assessment of adherence and barriers to adherence must be addressed. However, in youths who are taking most of their medication, near 100% adherence is still unlikely to maintain glycemic control, and, thus, clinicians should consider intensification of therapy (pharmacologic and/or nonpharmacologic) early in the course of the disease.

Acknowledgments. The TODAY Study Group thanks the following companies for donations in support of the study: Becton, Dickinson and Company; Bristol-Myers Squibb; Eli Lilly and Company; GlaxoSmithKline; LifeScan, Inc.; Pfizer; and Sanofi. The authors also thank the American Indian partners associated with the clinical center located at the University of Oklahoma Health Sciences Center, including members of the Absentee Shawnee Tribe, Cherokee Nation, Chickasaw Nation, Choctaw Nation of Oklahoma, and Oklahoma City Area Indian Health Service, for their participation and guidance.

The opinions expressed in this article are those of the authors and do not necessarily reflect the views of the respective Tribal and Indian Health Service Institution Review Boards or their members.

Funding. This work was completed with funding from the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) and the National Institutes of Health Office of the Director through grants U01-DK-61212, U01-DK-61230, U01-DK-61239, U01-DK-61242, and U01-DK-61254.

The NIDDK project office was involved in all aspects of the study, including design and conduct; collection, management, analysis, and interpretation of the data; review and approval of the manuscript; and decision to submit the manuscript for publication.

The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Duality of Interest. L.L.K. is a consultant to Takeda Phamaceuticals. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. L.L.K., B.J.A., S.V.M., and K.H. researched data, contributed to the discussion, wrote the manuscript, and reviewed and edited the manuscript. R.I. researched the data and reviewed and edited the manuscript. T.L.C., A.W., and K.J.N. researched the data, contributed to the discussion, and reviewed and edited the manuscript. L.A.H. researched the data and contributed to the discussion. K.H. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Some of the data presented were reported in a presentation entitled “Study Medication Adherence and Outcomes in the TODAY Cohort of Youth With Type 2 Diabetes” by S.V.M. to the International Society for Pediatric and Adolescent Diabetes, Toronto, Ontario, Canada, 3 September 2014.