Efforts to improve glucose control in patients with type 2 diabetes have been limited by the tendency for patients to experience glycemic deterioration during the maintenance phase of treatment. It has previously been reported that insulin use is a predictor of so-called “hyperglycemic relapse” (1). This previous retrospective study was limited by the inability to adjust for medication adherence, as this measure was not available in the database. Here, we reconsider the relationship between insulin use and hyperglycemic relapse in a prospective cohort study after consideration of medication adherence.

We studied 164 patients currently enrolled in a National Institute of Diabetes and Digestive and Kidney Diseases–funded Relapse Prevention Trial. Each patient in this study had documented previous poor glucose control (HbA1c [A1C] >8%), entered the 12-week Diabetes Improvement Program at Vanderbilt University, and completed the program demonstrating improved blood glucose control (A1C <8%). Patients were then contacted for follow-up for the study outcome, time-to-hyperglycemic relapse, defined as a subsequent A1C >8% and a ≥1% rise over nadir A1C.

At baseline, patients were classified as either insulin using or non–insulin using based on self-report during a face-to-face intake. At baseline and during follow-up, patients were asked to quantify how many shots of insulin they missed per week. Adherence to the prescribed regimen was calculated as the average percentage of shots missed per week over the course of the study. Cox proportional hazards models were then constructed to model the effect of adherence on hyperglycemic relapse.

The mean ± SD age of the patients was 55 ± 10.7 years. Approximately 56% were men, 76% Caucasian, and 21% African American. Mean BMI was 34.0 ± 6.9 kg/m2, mean duration of diabetes was 7.1 ± 8.2 years, and mean A1C was 6.70 ± 0.69%. Fifty-five percent were using insulin at baseline.

Mean follow-up of the cohort was 22.2 ± 10.3 months. Thirty-nine patients (24%) experienced hyperglycemic relapse. Lack of adherence to insulin regimen predicted hyperglycemic relapse in a dose-dependent manner. Compared with insulin users missing 0–5% of their shots, those missing 5–10% of their shots had a hazard ratio (HR) of 2.37 (95% CI 1.03–5.47), and those missing >10% of their shots had an HR of 3.14 (1.02–9.68) for hyperglycemic relapse. While insulin use was a univariate predictor of relapse (HR 2.3 [1.16–4.68]), there was no statistically significant difference in relapse rates between adherent insulin users (those missing 0–5% of their shots) and non-insulin users.

To further evaluate the accuracy of adherence for predicting hyperglycemic relapse, we plotted receiver operating characteristic curves using adherence as the classification variable. The area under this receiver operating characteristic curve was 0.69. The optimum cut point for identifying relapse represents ∼8% of shots missed per week or one shot per week for an individual taking two shots a day.

Contrary to previous reports, we find that nonadherence to an insulin regimen, rather than insulin use per se, is the dominant determinant of glycemic deterioration. Individuals missing as little as one insulin shot per week are at high risk for relapse. We have not excluded the possibility that nonadherence to insulin may be a marker for nonadherence to other self-care behaviors that may have also contributed to relapse. Still, strategies to improve adherence to insulin regimens are required to provide sustained glycemic control.

This study was supported in part by National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) Grant R18DK62258 (to T.A.E.), NIDDK Grant T32 DK007563 (to M.J.B.), and a grant from the Diabetes Trust Foundation (to M.J.B.).

1.
Elasy TA, Graber AL, Wolff K, Brown A, Shintani A: Glycemic relapse after an intensive outpatient intervention for type 2 diabetes (Letter).
Diabetes Care
26
:
1645
–1646,
2003