Longitudinal study of new and prevalent use of self-monitoring of blood glucose

Abstract

Objective: We sought to assess longitudinal association between self-monitoring of blood glucose (SMBG) and glycemic control in diabetic patients from an integrated health plan (Kaiser Permanente Northern California).

Research design and methods: Longitudinal analyses of glycemic control among 1) 16,091 patients initiating SMBG (new-user cohort) and 2) 15,347 ongoing users of SMBG (prevalent-user cohort). SMBG frequency was based on pharmacy use (number of blood glucose test strips dispensed), and glycemic control was based on HbA(1c) (A1C). In the new-user cohort, ANCOVA models (pre- and posttest design) were used to assess the effect of initiating SMBG. In the prevalent-user cohort, repeated-measure, mixed-effects models with random-intercept and time-dependent covariates were used to assess changes in SMBG and A1C. All models were stratified by therapy (no medications, oral agents only, or insulin) and adjusted for baseline A1C, sociodemographics, insulin injection frequency, comorbidity index, medication adherence, smoking status, health care use, and provider specialty.

Results: Greater SMBG practice frequency among new users was associated with a graded decrease in A1C (relative to nonusers) regardless of diabetes therapy (P < 0.0001). Changes in SMBG frequency among prevalent users were associated with an inverse graded change in A1C only among pharmacologically treated patients (P < 0.0001).

Conclusions: These observational findings are consistent with short-term benefits of initiating SMBG practice for all patients but continuing benefits only for pharmacologically treated patients. Differences in effectiveness between new versus prevalent users of SMBG have implications for guideline development and interpretation of observational outcomes data.

Figure 1

Adjusted dose-responsive change in A1C associated with SMBG initiation among patients not previously using SMBG and not treated pharmacologically (n = 7,872), treated with an oral agent only (n = 5,546), and patients treated with insulin (n = 840). Patients-switching-therapy changes were excluded. Therapy changes were excluded. Models adjusted for age, sex, insulin injection frequency (insulin model only), comorbidity index, oral medication refill adherence (OHA model only), appointment keeping, inpatient and outpatient utilization, smoking status, type of primary care provider, socioeconomic status indicators, timing of A1C test, and baseline A1C.