Effectiveness of HypoAware, a Brief Partly Web-Based Psychoeducational Intervention for Adults With Type 1 and Insulin-Treated Type 2 Diabetes and Problematic Hypoglycemia: A Cluster Randomized Controlled Trial

Severe hypoglycemia, defined as a low blood glucose level where there is a need of external assistance due to confusion, muscle weakness, or loss of consciousness, is one of the most impactful adverse events associated with insulin treatment in type 1 diabetes (T1D) and type 2 diabetes (T2D) (1). Severe hypoglycemia can lead to significant morbidity, fear of hypoglycemia, reduced quality of life, and even death (1–4), as well as to high societal costs in terms of increased health care consumption and reduced work productivity (2,5). The estimated incidence of severe hypoglycemia ranges from 1.0 to 1.7 episodes/patient/year and occurs in 30–37% of people with T1D (6–8). Longer disease duration increases the risk of severe hypoglycemia to 3.2 episodes/patient/year (9). With progressive insulin deficiency, the frequency of severe hypoglycemia in people with T2D approaches that of T1D (10). In people with diabetes who show hypoglycemia-associated autonomic failure, a defect associated with impaired symptom awareness, the risk of severe hypoglycemia is found to be 25-fold greater (1,11). Recurrent hypoglycemic episodes reduce hormonal counter-regulation, thereby increasing the risk for subsequent hypoglycemia, leading to a vicious cycle of events (11). There is evidence that psychobehavioral factors, like inattention for symptoms of hypoglycemia, misinterpretation, and the postponement of corrective measures, are at least partly responsible for an increased risk of severe hypoglycemia (12). Hypoglycemia-specific structured psychoeducational interventions have been developed, such as Blood Glucose Awareness Training (BGAT), that are aimed at helping patients enhance their understanding of risk factors for hypoglycemia, and improve symptom awareness, as well as adequate and timely corrective behaviors (13). A recent review and meta-analysis (14) showed the effectiveness of psychoeducational programs in terms of improving symptom awareness. However, because of methodological limitations more rigorous trials are needed (S.M.P.A.R., M.d.W., F.J.S., M.W. van Tulder, J.C.F. Ket, W.H.J.J. Cleijne, unpublished observations). Moreover, the use of the Internet can help to lower the burden of such programs on the resources of clinics, health care professionals’ time, and people with diabetes, and thereby enhance opportunities for dissemination (15).

We developed a brief, partly web-based psychoeducational group intervention for people with problematic hypoglycemia called HypoAware (in Dutch: HypoBewust) based on key active ingredients of the evidence-based BGAT, and demonstrated its feasibility and acceptability in a pilot study (15).

This study aims to evaluate the effectiveness of HypoAware in reducing episodes of severe hypoglycemia, and improving hypoglycemia awareness and hypoglycemia-related distress in people with T1D and insulin-treated T2D compared with usual care.

We performed a two-arm, multicenter, cluster randomized controlled trial with follow-up measurements at 2 months (4 weeks postintervention), 4 months, and 6 months. A detailed description of the study protocol is provided elsewhere (16). The study was approved by the Ethics Committee of the VU University Medical Center certified by the Central Committee on Research involving Human Subjects in the Netherlands (NL47354.029.13, registration no. 2014.007). We obtained written informed consent from all participants.

Cluster randomization was performed prior to the recruitment of participants at the level of the participating clinics to avoid contamination within the clinics. Random allocation was performed by two members of the study team, who randomly selected notes from two opaque envelopes, one with the eight names of the clinics and one with four notes with “intervention” and four notes with “control.” Participants were allocated based on the clinic where they were registered.

The study was performed in eight outpatient diabetes clinics in the Netherlands. After baseline measurement, participants were enrolled in HypoAware or received usual care for 6 months, depending on the clinic at which they were registered. Blinding of participants and health care professionals was not possible because of the nature of the intervention.

Persons were eligible for this study if they were ≥18 years of age, were treated for T1D or T2D in an outpatient setting, performed at least three multiple daily insulin injections a day or used continuous subcutaneous insulin infusion (pump therapy), and had experienced one or more episodes of severe hypoglycemia in the past 2 years and/or had a heightened risk of severe hypoglycemia operationalized by subjective impaired awareness of hypoglycemia (no or significantly reduced autonomic symptoms when the blood glucose level drops below 4 mmol/L (72 mg/dL), as defined by Gold et al. [17]). Furthermore, patients had to have access to the Internet, and be willing and able to attend the three group meetings.

Patients were excluded from participation when they met one or more of the following exclusion criteria: serious medical comorbidity (e.g., cancer or dialysis); a major psychiatric disorder (e.g., bipolar depression or schizophrenia); established cognitive impairment; substance abuse; pregnancy; and insufficient Dutch language skills or visual impairments.

HypoAware is an adapted version of Dutch BGAT (18), and was developed in close collaboration with diabetes experts with experience in BGAT and persons with diabetes. It consists of three group sessions of 2.5 h over 4 weeks, combined with two user-friendly online modules in the weeks between meetings. Groups consist of eight participants and are led by two trained diabetes professionals. HypoAware is guided by the biopsychobehavioral model of risk of severe hypoglycemia by Gonder-Frederick et al. (12), and aims to improve symptom recognition, risk awareness, preventive and problem-solving strategies, and coping with (the risk of) hypoglycemia. Topics covered in the group meetings and online are presented in Supplementary Table 1. Group sessions are highly interactive and aimed at patient empowerment, in which patient responsibility and informed decision-making is encouraged (when appropriate). Partners or significant others are invited to the third session. More information about the development and feasibility of HypoAware is published elsewhere (15).

Participants in the care as usual (CAU) group had access to comprehensive diabetes care as normally provided by their diabetes team (also available to the participants of the intervention group). After 6 months of follow-up, our study team trained the diabetes care teams from the control group, and participants from the control group received the intervention, if they so wished.

All measures mentioned below, unless otherwise mentioned, were administered web-based at baseline, and at 2, 4, and 6 months of follow-up. Clinical data were collected from the medical files.

• Episodes of severe hypoglycemia, defined as “a hypoglycemic event serious enough to require the help of another person” (self-report). We asked about the number of events occurring over the past 6 months at baseline, and over the past 2 months at all follow-up measurements.

• Hypoglycemia awareness was evaluated using the validated Gold score (17). The Gold score consists of one item on a Likert scale, ranging from 1 (“always aware”) to 7 (“always unaware”). A score of ≥4 indicates impaired hypoglycemia awareness.

• Fear of hypoglycemia was measured using the revised Hypoglycemia Fear Survey (HFS-II) with 33 items on a 0–4 Likert scale (19). Higher values indicate more worries (18 items) or hypoglycemia-related avoidance behaviors (15 items).

• Episodes of nonsevere hypoglycemia in the past week, defined as “a hypoglycemic event not serious enough to require the help of another person.” We asked participants to count the number of blood glucose measurements in the past week that were <4 mmol/L (<72 mg/dL), and asked them how many times they had felt low and treated their blood glucose level with carbohydrates without confirming the level with a test.

• Diabetes-related distress was assessed with the short-form Problem Areas In Diabetes scale (PAID-5) (20). This scale consists of five items on a Likert scale (0–4). A higher score means more diabetes-related distress. A score of ≥8 indicates elevated diabetes-related distress. We added one item from the PAID-20, “Worrying about low blood glucose reactions” with scores of 3 and 4 (“somewhat serious problem” and “serious problem,” respectively), indicating high hypoglycemia-related distress.

• Anxiety and depression were assessed with the Hospital Anxiety and Depression Scale (HADS) (21). Higher scores indicate more symptom burden of anxiety and depression. Scores for the subscales for anxiety and depression of <8 are considered to be normal, scores between 8 and 10 indicate borderline anxiety or depression, and scores between 11 and 21 indicate elevated anxiety or depression.

• Diabetes-specific self-efficacy was evaluated with the Confidence In Diabetes Self-care (CIDS) scale with 21 items (Likert scale 1–5) (22). Sum scores are transformed to 0–100. Higher scores indicate more overall diabetes-specific self-efficacy.

• Health-related quality of life was measured using the EuroQol 5-dimensional 5-level version (EQ-5D-5 L) and the visual analog scale (EQ-5D VAS) (23). Scores of the EQ-5D-5 L lie between 0 and 1, with higher scores indicating a better health-related quality of life. The scores of the EQ5D-VAS lie between 0 and 100, with a higher score indicating a higher subjective health status.

• The most recently determined glycosylated hemoglobin (HbA_1c) was collected from the medical file before enrollment in the study and around the time of the 6-month follow-up.

Other clinical data and demographics were only collected at baseline, as follows: type of diabetes, treatment regimen, diabetes duration, diabetes-related complications, lifetime severe hypoglycemic events, and comorbidities.

We assessed the number of group meetings attended via contact with the trainers after every group meeting and the number of online modules completed via the back end of the online environment. A minimum of two of three meetings and two of two modules are considered essential to obtain benefit from the intervention.

Power calculation was based on the European BGAT intervention study (24). We aimed to detect a difference of 0.94 episodes of severe hypoglycemia/6 months (1.07 ± 2.85 for the control group vs. 0.13 ± 0.33 for the BGAT group at the 6-month follow-up) between the two groups, with three follow-up measurements with an assumed correlation of 0.6, the α set at 0.05, a power of 80%, and adjusting for clustering within eight diabetes teams with an intraclass correlation of 0.01, as described by Twisk (25). Based on the power calculation, we needed to include 124 participants.

Differences between study participants and dropouts, the median number of events of severe hypoglycemia during the study, the percentage of persons affected with one or more events of hypoglycemia during the study, and an HbA_1c level between baseline and the 6-month follow-up were compared with Student t tests, Mann-Whitney U test, and χ² test with SPSS Statistics for Windows version 22.0 (IBM, Armonk, NY).

Primary and secondary outcome data were analyzed based on the intention-to-treat principle. Because we did not find between-clinic differences, it was not necessary to use multilevel modeling. Because of the repeated measurements, statistical analysis was performed using generalized estimation equations (GEEs) with Stata Statistical Software for Windows release 11.2 (Stata corporation, College Station, Texas). The GEE is a longitudinal technique, which takes into account that observations can be correlated. We used a linear model for continuous outcomes, a logistic model for dichotomous outcomes, and a Poisson or negative binomial model for count outcomes. In case of skewed continuous data, linear GEE analysis was performed after log transformation. Besides the crude analysis (only adjusted for the baseline value of the particular outcome), two adjusted analyses were performed; one adjusted for demographics (sex, age, and education level) and one adjusted for diabetes-related characteristics (episodes of severe hypoglycemia in the previous 2 years, diabetes duration, diabetes treatment method, HbA_1c level, and current use of real-time continuous glucose monitoring [RT-CGM]). A P value of ≤0.05 was considered to be statistically significant.

The baseline measurement was completed by 137 participants (Table 1). We have no record of the total number of patients reached by the local diabetes care teams. Participants were on average 52 ± 13 years old and almost evenly distributed across sex (46% female). A total of 88% of all participants had type 1 diabetes (n = 121), the mean HbA_1c level was 60.7 ± 11.6 mmol/mol (7.7 ± 1.1%), the mean duration of diabetes was 26 ± 13.6 years, and 42% of participants had one or more diabetes complications (n = 57). Following the criteria of Gold et al. (17), 76% of participants had impaired hypoglycemia awareness (n = 104); patients reported a median of two episodes of severe hypoglycemia in the past 6 months (interquartile range [IQR] 0–5). Besides the number of nonsevere hypoglycemic episodes (P = 0.011), there were no further statistically significant differences after clinical randomization at baseline.

Participant flow is depicted in the CONSORT diagram (Fig. 1). Nineteen participants dropped out of the study at the 6-month follow-up (14%; n = 13 intervention group, n = 6 control group). Dropouts were younger (46 vs. 53 years old, P = 0.018), had a lower level of education (47 vs. 25%, P = 0.024), had a shorter duration of diabetes (19 vs. 27 years, P = 0.015), had a higher HbA_1c level (66 vs. 60 mmol/L, P = 0.032), and were more often female (68 vs. 42%, P = 0.034) than participants who remained in the study.

Table 2 shows the crude and adjusted results of the GEE analyses, using 499 data points (9% of participants were missing data points; n = 250 intervention group, n = 249 control group) from four measurements in 6 months (Fig. 1).

Taking into account the four time points, and adjusting for baseline numbers and demographic and diabetes-related variables, the GEE analyses showed a nonsignificant 33% reduction in episodes of severe hypoglycemia in the HypoAware group compared with the CAU group (relative risk [RR] 0.67 [95% CI 0.39–1.16], P = 0.150). Looking at the proportion of participants with impaired awareness versus those with no impaired awareness, as defined by Gold et al. (17), HypoAware resulted in significantly reduced odds of impaired awareness (odds ratio [OR] 0.38 [95% CI 0.15–0.95], P = 0.038). Participants in the HypoAware group experienced a significant drop of 30% in distress concerning hypoglycemia, as measured with the PAID item (RR 0.70 [95% CI 0.56–0.88], P = 0.002) and a borderline significant 20% reduction in worries, as measured with the Worry scale of the HFS-II (RR 0.80 [95% CI 0.64–1.01], P = 0.059).

Comparing the median number of severe hypoglycemic events during the course of the study (T1–T3, covering 6 months) with an independent-samples Mann-Whitney U test, we found a significant difference between the groups, as follows: a median of 2.5 severe hypoglycemic events/6 months (IQR 1–10) in the CAU condition versus 1 severe hypoglycemic event/6 months (IQR 0–6.5) in the HypoAware group (P = 0.030). The groups also differed in the proportion of participants affected by one or more events of severe hypoglycemia during the study, with 79% for the CAU group and 53% for the intervention group (Pearson χ² test, P = 0.003).

There were 18% more episodes of nonsevere hypoglycemia reported in the HypoAware group relative to the CAU group, but this was not significant (RR 1.18 [95% CI 0.90–1.53], P = 0.231).

Change in HbA_1c levels between baseline and the 6-month follow-up did not differ significantly between the groups (mean difference −2.3 mmol/mol [95% CI −5.1 to −0.5], P = 0.101). HbA_1c concentration dropped 2.4 mmol/mol in the control group versus 0.1 mmol/mol in the intervention group. We found no significant differences in diabetes distress (PAID), CIDS scale score, symptoms of anxiety and depression (HADS), and health-related quality of life (EQ5-D) between the HypoAware and CAU groups (Table 2).

In terms of participant involvement, 61 of 71 participants in the intervention group completed all three meetings (86%), and 7 participants completed two of three meetings (10%). Sixty-four of 71 participants completed both Internet modules (90%).

This study demonstrated that participants who received treatment with HypoAware showed better hypoglycemia-related and psychological outcomes than participants in the CAU group, although findings were not all statistically significant. The results corroborate earlier BGAT findings (13), although we reduced the number of sessions by more than half, and substituted a fair amount of content by Internet modules. Moreover, we observed good program adherence, with 96% participating in two or more sessions (of three), and 90% completing the two online sessions.

The improvement in hypoglycemia awareness could be considered the most direct effect of the intervention. We saw 38% odds of reduced impaired hypoglycemia awareness in the HypoAware group compared with the control group. Improved symptom awareness is key to preventing severe hypoglycemia in the absence of RT-CGM and is probably responsible for the finding that participants receiving treatment with HypoAware reported 18% more events of nonsevere hypoglycemia than participants receiving CAU. Importantly, an increase in nonsevere hypoglycemia events in the intervention group did not translate into more worries. In contrast, we found a trend toward lower hypoglycemia-related distress in the intervention group, and that HbA_1c levels did not deteriorate. These findings suggest improved self-management following the HypoAware intervention.

The absolute number of severe hypoglycemia events over the course of our study period of 6 months was significantly different between the two groups. When adjusting for diabetes-related and demographic variables, the 33% risk reduction for severe hypoglycemia in the HypoAware group was not statistically significant. If we translate our findings in terms of numbers needed to treat, 0.7 patient needs to be exposed to HypoAware in order to prevent one episode of severe hypoglycemia, or, as stated otherwise, 11 events of severe hypoglycemia can be prevented by one delivery of HypoAware. It is important to note that the observed reductions in severe hypoglycemia were achieved in a sample of patients who were not all seriously affected by severe hypoglycemia (although most had impaired awareness) with access to ongoing comprehensive outpatient diabetes care.

Interestingly, almost a quarter (24%) of patients in the control group used RT-CGM, compared with 8% in the HypoAware group. We have no explanation for this difference, but CGM was not an exclusion criterion and, indeed, is available in the Netherlands for patients with persistent, problematic diabetes regulation. We added RT-CGM use as a confounder in the main analyses and can speculate that the effects of HypoAware are more pronounced in a patient population without RT-CGM use because RT-CGM has been found to be effective in preventing severe hypoglycemia (26). More research is needed to examine the role of RT-CGM and the interaction with HypoAware in a larger sample of patients with problematic hypoglycemia. Yeoh et al. (14) recently proposed an algorithm for offering psychoeducation combined with pump therapy and/or CGM to patients experiencing recurrent hypoglycemic events, but the evidence for using such tailored strategies is still lacking.

Some limitations and strengths of our study deserve to be mentioned.

For the assessment of hypoglycemia, we relied on self-report rather than objective glycemic data. There is, however, evidence to support the reliability of self-reported severe hypoglycemia (27). The use of CGM to quantify the time spent in hypoglycemia in future studies is recommended.

It was not possible to blind patients and caregivers to the randomization because of the nature of the study; however, since both groups received treatment with HypoAware (the control group as an incentive after the study period of 6 months), we can assume similar expectancies in both groups.

A follow-up period of 6 months is a relatively short period of time to demonstrate a reduction in severe hypoglycemia, and longer-term studies are therefore warranted. Based on previous BGAT research, we can expect the effects to be maintained for at least 1 year after the intervention (28).

The strengths of this study include the multicenter cluster randomized design, lowering the risk of contamination; and the testing of the intervention in a real-life setting as an add-on to usual care in a mixed sample of patients with type 1 and type 2 diabetes who were at risk for severe hypoglycemia, strengthening external validity. The highly structured nature of the intervention (with similar effects observed across clinics) allows for replication studies. Alongside this randomized controlled trial, we plan to perform an economic evaluation. The results of this evaluation will be presented separately.

In conclusion, HypoAware is a brief, “blended” group and web-based psychoeducational intervention that helps patients with problematic hypoglycemia lower their risk of severe hypoglycemia, improve their hypoglycemia awareness, and reduce their worries about hypoglycemia. Further dissemination and testing of the intervention in different care settings are recommended.

Acknowledgments. The authors thank Nienke Veldhuijzen, Department of Epidemiology & Biostatistics, VU University Medical Center, the Netherlands, for her help with the sample size calculation. The authors also thank the trainers from the following participating centers: Zaans Medisch Centrum, Bernhoven, Slingeland Ziekenhuis, Jeroen Bosch Ziekenhuis, Spaarne Gasthuis, Franciscus Gasthuis & Vlietland, Gelre Ziekenhuizen, and VU Medisch Centrum, and all their participants.

Funding. The study and manuscript preparation are funded by ZonMw (837001406), the Dutch Organisation for Health Research and Development. The authors have received unrestricted cofunding from Novo Nordisk, Agis Achmea (a Dutch health care insurer), and VU University Medical Center.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. S.M.P.A.R. constructed the design of the study, executed the study, and drafted the manuscript. M.d.W. and F.J.S. developed the study, constructed the design of the study, and revised the manuscript. J.W.T. participated in the design of the study, performed the statistical analyses, and revised the manuscript. The final manuscript was read and approved by all authors. F.J.S. 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. Parts of this study were presented in abstract form at the 76th Scientific Sessions of the American Diabetes Association, New Orleans, LA, 10–14 June 2016.