Is hypoglycemia fear independently associated with health-related quality of life?

Detailed methodology for this cross-sectional observational research study was previously published [16]. The study combined patient-reported data collected through mail surveys and retrospective analysis of administrative claims data (medical data, pharmacy data, and enrollment information) from a large US health plan (the Life Sciences Research Database). The study protocol was approved by a central institutional review board (IRB) prior to study initiation. And due to the fact that the patients in our study were identified from administrative claims data, the study concept was also approved by the large US health plan prior to commencement of the study. Participants were adult enrollees in the health plan with a diagnosis of type 2 diabetes and ≥2 pharmacy claims for an oral antidiabetic medication (OAD) with or without insulin identified in the administrative claims data from December 1, 2008, to November 30, 2009.

For the survey, patients (N = 3999) were contacted directly by mail. The patient survey assessed demographic variables, age at time of type 2 diabetes diagnosis, family history of diabetes, and smoking status. Patients were asked about their current OAD, duration of treatment, and modifications to their treatment within the previous 6 months. Medication adherence was assessed using the Morisky self-report medication adherence 4-item questionnaire [17].

Fear of hypoglycemia was evaluated using the HFS [18]. This validated assessment lists 15 behaviors that patients may engage in to avoid low blood sugar, with potential consequences, and 18 items related to low blood sugar about which patients with diabetes may be concerned. The recall period was 6 months. A total score, Behavior subscale score (sum of the 15 behavioral items), and Worry subscale score (sum of the 18 worry items) were generated. In addition, patients were asked whether they had changed their medication or were afraid to change medications because of hypoglycemia.

Health-related quality of life was assessed using 2 instruments. Overall health status was assessed using the EQ-5D, a standardized measure that rates 5 dimensions of health (mobility, self-care, usual activities, pain/discomfort, anxiety/depression) across 3 levels (no problems, some/moderate problems, and extreme problems) [19],[20]. The index value (ie, health utility) between 0 (death) and 1 (perfect health) was calculated using the value sets (weights) from a US population sample. The other instrument was the SF-12 [21], which can produce an MCS and a PCS. Summary scores range from 0 to 100, with a higher score indicating a better quality of life in mental and physical components. Hypoglycemia was defined based on American Diabetes Association symptoms [22] and blood glucose readings at the time that symptoms occurred.

The survey packet we used in this study contained a pre-paid cash incentive of $10 to the patients, along with the invitation letter printed on letterhead, the IRB-approved consent form, the survey instrument printed in booklet form, and a postage-paid business reply envelope. We sent reminders to the patients 2 and 4 weeks following the initial survey to increase our response rate.

Survey variables (eg, demographics and clinical characteristics) were analyzed descriptively, and bivariate comparisons of demographic characteristics and outcome measures were performed using appropriate tests (eg, t-test or chi-square test) based on the distribution of the measure. Multivariate analyses of HFS, SF-12, and EQ-5D outcomes were conducted using ordinary least squares. Specific predictors (ie, explanatory variables) to be included were selected primarily based on clinical rationale. Associations among symptomatic hypoglycemia, fear of hypoglycemia, and HRQOL were modeled in 2 ordinary least-squares regressions, controlling for sociodemographics, illness characteristics, and treatment factors [16]. Model 1 of the HRQOL = f (symptomatic hypoglycemia) tested the associations between symptomatic hypoglycemia and HRQOL, and based on previously published results [16], these were expected to be statistically significant. Model 2, which was the primary assessment of the study hypothesis, added 1 additional variable of hypoglycemia fear and tested the impact that fear of hypoglycemia had on health status independently of the impact of hypoglycemia symptoms. Following standard procedure, regression diagnostics were performed to assess goodness of fit and violations of model assumptions (eg, multicollinearity, heteroskedasticity) for each model. Any violations of the model were noted and appropriate corrections made to the data. The fitted and observed data were examined to uncover outliers, their effect on the analysis, and possible misspecification of the initial equation. SAS 9.2 software (SAS Institute, Inc., Cary, NC) was used for statistical analyses [23]. The significance level was set at 0.05 (2-tailed).

Because different classes of antidiabetic drugs may be more likely to cause hypoglycemia [24]-[27], the association between fear for hypoglycemia and quality of life might also be different across different patient subgroups of antidiabetic medication users. Due to this concern, the same statistical analysis procedures were conducted on two patient subgroups of antidiabetic medication users: insulin users and sulfonylurea users.

Of 3999 patients contacted, 813 responded with a complete survey (response rate, 20.3%). Of these, 578 (71.1%) reported ≥1 hypoglycemia symptom within the previous 3 months. Among the population characteristics assessed, only OAD group and sex showed a statistically significant difference between patients with and without symptomatic hypoglycemia (Table 1). Patients in the OAD plus insulin group were more likely to report ≥1 hypoglycemia symptom (83.3% [244/293]) compared with those in the sulfonylurea without insulin group (69.9% [172/246]) or the nonsulfonylurea OAD without insulin group (59.1% [162/274]; P < 0.001). Significantly more women (77.1% [263/341]) than men (66.7% [315/472]) reported ≥1 hypoglycemia symptom.

Symptomatic hypoglycemia was associated with more fear of hypoglycemia (higher total HFS score) and worse HRQOL (lower SF-12 MCS, SF-12 PCS, and EQ-5D scores; each, P < 0.05; Figure 1). Table 2 presents the associations between the HFS and various factors. The presence of symptomatic hypoglycemia was associated with greater fear of hypoglycemia and had the strongest association (regression coefficient >10 points of HFS total score, P < 0.0001). Other factors also showed statistically significant associations (Table 2). Higher income was associated with less fear of hypoglycemia (P < 0.0001), as was white racial categorization compared with nonwhite racial categorizations (P < 0.0001). Hispanic ethnicity was associated with greater fear of hypoglycemia (P = 0.03). Longer duration with a diagnosis of diabetes and better self-reported adherence were also associated with greater fear of hypoglycemia (each P < 0.0001).

The ordinary least-squares model of HRQOL using symptomatic hypoglycemia only and controlling for demographics and illness characteristics (Model 1, Table 3) yielded regression coefficients for symptomatic hypoglycemia on HRQOL scores. The associations between symptomatic hypoglycemia and EQ-5D score and SF-12 MCS score were statistically significant (each P < 0.05). In Model 2, which included the additional variable of HFS, the HFS score was significantly associated with scores on all 3 HRQOL outcomes (each P < 0.0001), whereas symptomatic hypoglycemia showed a statistically significant association only for the SF-12 MCS (P = 0.022). The fully specified regression models for Table 3 are provided in Additional file 1: Tables S1.1-S1.6.

Subgroup analyses were conducted for insulin users (Table 4) and sulfonylurea users (Table 5). For insulin users (n = 293), Model 1 showed that the symptomatic hypoglycemia was not associated with 3 HRQOL scores (each P > 0.05). Model 2 showed that the HFS score was still significantly associated with scores on all 3 HRQOL outcomes (each P < 0.05), whereas symptomatic hypoglycemia showed no statistically significant association with any of the 3 HRQOL outcomes (each P > 0.05). For sulfonylurea users (n = 246), Model 1 showed that the symptomatic hypoglycemia was associated with EQ-5D scores and SF-12 MCS scores (each P < 0.05). Model 2 showed that the HFS score was significantly associated with scores on all 3 HRQOL outcomes (each P < 0.05), whereas symptomatic hypoglycemia was only significantly associated with EQ-5D scores and SF-12 MCS scores (each P < 0.05). The fully specified regression models for Table 4 and Table 5 are provided in Additional file 1: Tables S2.0-S3.5.