Results
Patient Demographics and Baseline Clinical Characteristics
A total of 4,804 patients with T2DM were included. Of these, 2,257 initiated insulin glargine by pen device, 1,915 initiated insulin glargine by vial-and-syringe, and 632 initiated insulin detemir by pen device. Most subjects were male, with 74.6% aged 40 to 64 years (Table 1). On average, subjects were using two OADs at baseline and had a mean baseline A1C of 9.37%, exceeding targets typically recommended by the American Diabetes Association. Detailed information within each treatment group can be found in the original studies. There were no significant differences in baseline population characteristics among the 3 studies.
Treatment Persistence with Insulin Treatment
The average insulin treatment persistence rate over the 1-year follow-up period was 65.0% for all patients, with the highest persistence rate seen among patients initiating insulin glargine by disposable pen (mean persistence of 70.6%), which was significantly higher than those who initiated insulin glargine by vial-and-syringe (62.9%), and those who initiated insulin detemir by disposable pen (51.1%; Table 2). On average, patients persisted with insulin treatment for 308 days, and similarly to the persistence rate findings, patients initiating insulin glargine by disposable pen showed the longest persistence times (320 days) compared with those who initiated insulin glargine by vial-and-syringe (305 days) and those who initiated insulin detemir by disposable pen (279 days; Table 2 and Fig. 1). Sensitivity analyses using 75th percentiles revealed similar results (Table 2). Regression analyses controlling for confounding factors also confirmed these results in showing that insulin glargine users were 50 to 60% more likely to be persistent during 1-year follow-up and had a 30% lower risk for early discontinuation (Fig. 2).
(Enlarge Image)
Figure 1.
Kaplan-Meier curve on time to discontinuation of insulin treatment (90th percentile).
Factors Associated with Treatment Persistence and Time to Treatment Discontinuation
Logistic regression identified baseline factors associated with 1-year treatment persistence (Fig. 2A). Significantly higher persistence was associated with the following factors: older age, initiation with insulin glargine using disposable pen or vial-and-syringe, or a baseline prescription for exenatide or sitagliptin. Persistence was not significantly influenced by gender, baseline A1C, or CCI.
(Enlarge Image)
Figure 2.
Predictors of treatment persistence and time to discontinuation. Baseline factors that (A) predict persistence with insulin treatment and (B) predict time to discontinuation of insulin treatment for those initiating insulin with insulin glargine or insulin detemir. CCI= Charlson comorbidity index; SU= sulfonylurea; TZD= thiazolidinedione.
Cox regression analysis of time to treatment discontinuation identified similar factors. Patients discontinued insulin treatment significantly earlier if they received treatment with insulin detemir by pen compared with insulin glargine by pen device or vial-and-syringe (Fig. 2B). As with treatment persistence, patients who were older, initiated insulin glargine by disposable pen or vial-and-syringe, or were prescribed exenatide or sitagliptin discontinued medication at later times (Fig. 2B).
Association of Treatment Persistence and Outcomes
There were significant correlations between treatment persistence and 1-year follow-up outcomes. Those patients who persisted with insulin treatment during the 1-year follow-up period had lower A1C levels at follow-up and showed a greater reduction in A1C levels from baseline than those who did not, with similar numbers of hypoglycemic events in both groups (Fig. 3). There were also significant correlations between more treatment persistent days and lower A1C levels at follow-up (R2 = −0.05191, P = .0249), as well as greater A1C reductions (R2 = −0.07818, P = .0025).
(Enlarge Image)
Figure 3.
Association between 1-year treatment persistence with insulin and health care utilization and costs. The mean number of health care utilizations (A) and mean health care costs (B) during follow-up for those who were either persistent or nonpersistent over the 1-year follow-up.
Those who persisted with treatment for the 1-year follow-up period also exhibited a lower number of health care utilizations and had higher pharmacy costs but similar overall health care costs (Fig. 4), owing to lower medical costs. There were significant correlations between the number of treatment persistent days and the total number of hospitalizations (R 2 = −0.05011, P = .0005), the number of diabetes-related hospitalizations (R 2 = −0.03428, P = .0175), and pharmacy costs (R 2 = 0.11601, P <.0001).
(Enlarge Image)
Figure 4.
Association between 1-year treatment persistence with insulin and clinical outcomes. A1C value at follow-up (A), change in A1C from baseline (B), and the mean number of hypoglycemic events during follow-up (C) for those who were either persistent or nonpersistent over the 1-year follow-up.
A similar pattern was observed for the duration of insulin treatment and 1-year follow-up outcomes; longer duration of insulin treatment was associated with a greater A1C reduction from baseline, lower hospitalization rates, and higher pharmacy costs (Fig. 5).
(Enlarge Image)
Figure 5.
Distribution of 1-year follow-up treatment persistence duration and (A) mean change in A1C from baseline, (B) any hypoglycemia-related event, (C) any hospitalization, and (D) mean pharmacy costs.