The majority of stakeholders prioritized adolescents with type 1 diabetes and adults with insulin-requiring Type 2 Diabetes as the populations in greatest need for future research. For each population, rt-CGM was prioritized as the highest method of glucose monitoring for future research, while the method of insulin delivery and outcomes of interest varied by population. When asked to prioritize the final research question within each category, glucose monitoring methods were universally prioritized above insulin delivery methods.

While the stakeholders rated adolescents with type 1 diabetes the highest priority, the original Comparative Effectiveness Review (CER) investigators commented future studies should focus on populations in which diabetes is growing (i.e. elderly individuals with type 1 and type 2 diabetes, insulin-treated type 2 diabetes, minority populations). This difference may be because the stakeholders took a clinical perspective that focused on treatment while the CER investigators took a research perspective that focused on the gaps in data. On the other hand, adherence as outcome was rated high by both stakeholders and the original CER investigators. It is important to note that stakeholders rated the artificial pancreas as the highest priority for the future research, despite the fact that the technology of artificial pancreas is at the developmental stage, is not widely used in practice, and was not included in the original Comparative Effectiveness Review for lack of eligible studies. Nonetheless, this consensus reflected the urge to develop a better and more convenient system for diabetes treatment.

Long-term clinical outcomes were not specifically included for prioritization by the stakeholders; however, stakeholders were given the opportunity to independently identify long-term outcomes for future research. While we feel that long-term clinical outcomes, such as mortality, macrovascular complications, and microvascular complications, are the ultimate goal of interventions for both type 1 and type 2 diabetes populations, such trials would need an extensively long time for followup. The Diabetes Control and Complications Trial showed a significant reduction in microvascular complications after only 6.5 years, and macrovascular complications after 15 years; however, these significant differences were achieved only with a very wide difference in HbA1c of approximately 2% between intervention and placebo groups.⁵ In the United Kingdom Prospective Diabetes Study, it took 10 years for an HbA1c difference of 0.9% to show a reduction in microvascular complications and 20 years to show a reduction in macrovascular complications.¹⁴ Comparative effectiveness studies would require very large number of patients to be followed for many years to show significant micro and macrovascular effects, especially if only small A1c differences are seen, something much too costly to do. Supported by strong rating from the stakeholders, we feel HbA1c is a reasonable surrogate endpoint, and should be used as such when looking at the comparative effectiveness of rt-CGM and sensor-augmented insulin pump, versus other interventions.

Our study had several strengths. First, we used an established approach for consensus building. Second, we invited experts from multiple disciplines as stakeholders including practicing endocrinologists, clinical researchers, and a patient, which increased the generalizability. Third, the stakeholders reached consensus with only one round of survey, which reflected high level of consistency.

Nonetheless, this study has some limitations. First, our study was limited to the scope of the original CER. The original investigators determined the research gaps based on their own findings. Stakeholders did not independently identify research gaps on the basis of populations, interventions, and outcomes, but rather by the limited options that we provided according to our analytic framework. This limitation is offset by the benefit of keeping the study focused on the populations and interventions that were included in our analytic framework. This study did not specifically address the needs for future research in pregnant women because we thought it would require a separate study (with a different group of stakeholders) to adequately determine the research needs for pregnant women having type 1 diabetes, type 2 diabetes, or gestational diabetes. Another limitation is that the complexity of the concepts in this topic may be a barrier for patient stakeholders to contribute. The decision making process associated with prioritizing clinical interventions could potentially be a daunting task for non-clinicians and non-researchers in the field. Clinicians have a level of standardized education and training in the field. The average patient may or may not have the requisite breadth of knowledge and experience to prioritize interventions for the entire field of insulin delivery and glucose monitoring research. Still, in our study, this was minimized by a patient Stakeholder with a longstanding history of diabetes and who has taken an active role in his care throughout his life.