The impact of the dose adjustment for normal eating (DAFNE) structured education programme on health outcomes and healthcare costs for people with type 1 diabetes in Ireland

Abstract

The dose adjustment for normal eating (DAFNE) structured education programme has been shown to be clinically and cost-effective in enhancing self-management and improving health outcomes for people with type 1 diabetes.^1-4 Like many jurisdictions worldwide, the National Clinical Guideline for type 1 diabetes mellitus in Ireland, which was first published in 2018 and recently updated in 2024, calls for the provision of the DAFNE programme as a core element of clinical practice.⁵ Notably, however, this recommendation emerged in the absence of Irish-specific data on clinical and cost-effectiveness and instead was informed by and dependent upon the generalisability of the well-established international evidence base.^1-4 While this is a justifiable assumption, the evidence base supporting the clinical recommendation to implement DAFNE in the Irish context is worthy of further interrogation. To this end, we undertook an opportunistic, secondary analysis of historical data collected via a cluster randomised controlled trial (RCT) of people with type 1 diabetes in Ireland.^6-8 Notwithstanding the RCT study design, we do not provide definitive, causal evidence given sample size and data limitations. Instead, our results may be interpreted as independent associations between DAFNE completion and health outcomes and healthcare costs. Taken together, our findings that DAFNE was associated with lower distress and better quality of life at no additional healthcare cost are supportive of its inclusion in the Irish clinical guideline.

Data from the cluster RCT were previously used to evaluate the clinical and cost-effectiveness of group follow-up versus individual follow-up for people with type 1 diabetes who completed the DAFNE programme.^{7, 8} Notably, the RCT also recruited a third arm of participants who did not complete DAFNE. The study authors initially planned to include this non-DAFNE arm as a third comparator in the RCT. However, the study failed to recruit a sufficient number of participants in the non-DAFNE arm to justify its inclusion. The study recruited participants from nine hospital centres between October 2006 and February 2009. Inclusion and exclusion criteria are reported elsewhere.⁶ In total, 437 individuals were collectively recruited by the centres randomised to deliver DAFNE, while 57 individuals were recruited by the centres randomised to the non-DAFNE arm. Nonetheless, baseline and follow-up data were collected for all 494 recruited participants: DAFNE and non-DAFNE. At baseline, the mean age was 41 years [standard deviation (SD): 12] in the DAFNE arm and 42 years (SD: 14) in the non-DAFNE arm. The proportion of females was 54% for DAFNE and 44% for non-DAFNE. The mean duration of diabetes was 16 years (SD: 11) for DAFNE and 17 years (SD: 12) for non-DAFNE. Mean HbA1c was 8.31% or 67 mmol/mol (SD: 1.35) for the DAFNE arm and 8.30% or 67 mmol/mol (SD: 1.27) for the non-DAFNE arm.

For the purposes of our statistical analyses, complete case data for a range of health outcomes and healthcare cost variables collected in the RCT at baseline and 12 months were utilised. Notably, given the dependence on postal questionnaires for data collection, missing data were a significant issue at follow-up. Further, the level of missing data was greater for the non-DAFNE arm than for the DAFNE arm. Multilevel multivariable regression analysis was conducted for each dependent variable, which comprised the change in the variable between baseline and 12 months. The independent variables included were DAFNE status, age, gender, duration of illness and baseline HBA1c level. This pragmatic choice was informed by Akaike information criterion statistics, with supplementary analyses to test robustness.

The full set of summary statistics and regression analysis results are presented in Table 1. In the Problem Areas in Diabetes (PAID)⁹ analysis, DAFNE was associated with lower levels of diabetes distress (6.68: p = 0.039). In the Diabetes-Specific Quality of Life Scale (DSQOLS)¹⁰ analysis, DAFNE was associated with better disease-specific quality of life (11.51: p = 0.000). In the EQ-5D-3L¹¹ analyses, DAFNE was associated with increases in index scores (0.055: p = 0.045) and visual analogue scale (VAS) scores (5.19: p = 0.018), indicating better general quality of life. Finally, in the healthcare cost analysis, DAFNE did not have a statistically significant impact.

These findings notwithstanding, the study had a number of limitations. First, given the sample size of the non-DAFNE arm, questions may be raised over its representativeness of the true population. Second, given the degree of missing data, the greater level of missingness for the non-DAFNE arm relative to the DAFNE arm, and complete case analysis approach adopted, the regression coefficients presented may well be biased upwards or downwards. Third, as the data is over 10 years old, the findings are reflective of the pre-pandemic population and care pathway for type 1 diabetes, which has since benefited from the emergence of new drugs and technologies. In light of these limitations, it is prudent to be cautious in the interpretation of our findings and their generalisability to contemporary national and international settings.

The authors declare that there are no conflicts of interest to declare.