A parallel randomised controlled trial of the Hypoglycaemia Awareness Restoration Programme for adults with type 1 diabetes and problematic hypoglycaemia despite optimised self-care (HARPdoc)

Abstract

Impaired awareness of hypoglycaemia (IAH) is a major risk for severe hypoglycaemia in insulin treatment of type 1 diabetes (T1D). To explore the hypothesis that unhelpful health beliefs create barriers to regaining awareness, we conducted a multi-centre, randomised, parallel, two-arm trial (ClinicalTrials.gov NCT02940873) in adults with T1D and treatment-resistant IAH and severe hypoglycaemia, with blinded analysis of 12-month recall of severe hypoglycaemia at 12 and/or 24 months the primary outcome. Secondary outcomes included cognitive and emotional measures. Adults with T1D, IAH and severe hypoglycaemia despite structured education in insulin adjustment, +/- diabetes technologies, were randomised to the "Hypoglycaemia Awareness Restoration Programme despite optimised self-care" (HARPdoc, n = 49), a psychoeducation programme uniquely focussing on changing cognitive barriers to avoiding hypoglycaemia, or the evidence-based "Blood Glucose Awareness Training" (BGAT, n = 50), both delivered over six weeks. Median [IQR] severe hypoglycaemia at baseline was 5[2-12] per patient/year, 1[0-5] at 12 months and 0[0-2] at 24 months, with no superiority for HARPdoc (HARPdoc vs BGAT incident rate ratios [95% CI] 1.25[0.51, 3.09], p = 0.62 and 1.26[0.48, 3.35], p = 0.64 respectively), nor for changes in hypoglycaemia awareness scores or fear. Compared to BGAT, HARPdoc significantly reduced endorsement of unhelpful cognitions (Estimated Mean Difference for Attitudes to Awareness scores at 24 months, -2.07 [-3.37,-0.560], p = 0.01) and reduced scores for diabetes distress (-6.70[-12.50,-0.89], p = 0.02); depression (-1.86[-3.30, -0.43], p = 0.01) and anxiety (-1.89[-3.32, -0.47], p = 0.01). Despite positive impact on cognitive barriers around hypoglycaemia avoidance and on diabetes-related and general emotional distress scores, HARPdoc was not more effective than BGAT at reducing severe hypoglycaemia.

Conflict of interest statement

S.A.A. has served on Advisory Boards for Novo Nordisk and Medtronic and has spoken at educational meetings supported by NovoNordisk and Sanofi in the past year and is a co-investigator on the EU IMI HypoRESOLVE programme. P.C. has received personal fees from Abbott, Dexcom. Insulet, Medtronic, Novo Nordisk, Lilly and Sanofi. S.H. has served on Advisory Boards and Consulted with Eli Lilly, NovoNordisk, Zealand Pharma and served on speaker panels for NovoNordisk. He is a co-investigator on the EU IMI HypoRESOLVE programme. E.T. is a consultant to Medtronic. A.B. declares that he has received honoraria from Astra Zeneca and Sanofi for speaking at educational events and sponsorship from Lilly and Janssen to attend conferences. L.G.F. is on the Advisory Group for the EU IMI HypoRESOLVE programme. She is also the General Manager of HFS-Global LLC, a business developed in partnership with the University of Virginia to license use of the HFS-II to for-profit entities. No licensing fees were charged for the use of HFS-II in this study. N.S. is the director of the London Safety and Training Solutions Ltd, which offers training in patient safety, implementation solutions and human factors to health care organisations and the pharmaceutical industry. The other authors have no disclosures.

© 2022. The Author(s).

Figures

Fig. 1. CONSORT diagram for the HARPdoc RCT, showing participant flow through the study.

Participant numbers at each stage are shown in the figure. * Patients judged as potential participants for the trial, including some (n = 329) identified through remote review of electronic patient records by researchers followed by “cold call” and others identified at clinical team meetings. ¥ Attendance of at least the first 3 days of allocated treatment (HARPdoc or BGAT) plus at least one 1:1 session for HARPdoc was considered as “received allocated treatment”.

Fig. 2. A box plot to present the summary statistics of the number of SH events in the previous 12 months (primary endpoint) by treatment arm and time point.

The grey box represents the 25–75% interquartile range, the median is represented by the white line in each box and the whiskers cover the minimum to the maximum values, to the limit of 1.5× the IQR. Individual values for rates outside this limit have been excluded for clarity and are presented in Table 5. A three-level random intercept negative binomial model has been used, using a missing at random assumption (MAR) and a significance level of 2.5% (two sided). The N for each group is baseline: 49 and 50, 12 months 42 and 45 and 24 months 41 and 42 HARPdoc and BGAT respectively. Source data are available on request to SAA or IB, as described in the text.

Fig. 3. Standardised effect sizes of secondary endpoints A2A, PAID and HADS.

Values below zero indicate better outcomes in HARPdoc. The data are presented as estimated mean differences and 95% confidence intervals, with open triangles representing means of 12 month data; closed triangles representing means of 24-month data and bars representing 95% confidence intervals. Analysis is by a three-level random intercept linear regression regression model. A significance level of 5% (two sided) has been used. The standardised effect sizes for other secondary endpoints were not significant and are shown, together with numerical data for the depicted parameters, and the numbers providing data at each time point in each intervention, in Tables 4 and 6. Source data are available on request to S.A.A. or I.B., as described in the text.