Sustained Impact of Real-time Continuous Glucose Monitoring in Adults With Type 1 Diabetes on Insulin Pump Therapy: Results After the 24-Month RESCUE Study

Achieving optimal glycemic control remains a challenge for people living with type 1 diabetes, despite rapid advancements in insulin administration technology and better insulin preparations (1). Therefore, the development of strategies and technological tools to lessen the burden of diabetes management with favorable results on metabolic control and quality of life is a continuous endeavor.

In recent years, more and more people living with diabetes have access to real-time continuous glucose monitoring (rtCGM), which has shown both in randomized controlled and observational trials that it improves glucose control and quality of life in people treated with multiple daily insulin injections (MDI) (2–6) or continuous subcutaneous insulin infusion (CSII) (6–12). However, randomized controlled trials are often of short duration (typically 6 months), making it less clear how much of the observed effects are due to heightened motivation of trial participants. In addition, longer-term observational studies often lack sufficient patient numbers to generalize the outcomes to the broader community of people with type 1 diabetes (6,12).

Since September 2014, rtCGM has been reimbursed in Belgium for people with type 1 diabetes who use CSII and are treated in selected specialized diabetes centers. We previously reported 12-month findings from the Reimbursement Study of Continuous Glucose Monitoring in Belgium (RESCUE) study (13). At the start of rtCGM, the RESCUE population had suboptimal glucose control with a high incidence of acute hypoglycemia-related complications, which was also reflected in lower perceived quality of life compared with a general population with type 1 diabetes (14). We reported that 12-month use of rtCGM led to improved glycemic control and fewer hypoglycemia-related hospitalizations and diabetes-related days off from work, with a significant cost reduction. Our aim in the current analysis was to determine whether the observed benefits could be sustained up to 24 months for the total population, with additional specific attention to people with impaired awareness of hypoglycemia (IAH) because they are at higher risk of severe hypoglycemia (15,16).

This multicenter, prospective, observational, cohort study evaluated the impact of nationwide reimbursement of rtCGM systems for adults with type 1 diabetes on CSII therapy. The results from the full 24 months of the study are reported here and consist of the first 12-month period from which the results have been published (13) followed by an additional 12-month extension phase. The study was conducted from September 2014 to March 2019.

The study complied with the Declaration of Helsinki and the International Conference on Harmonization/Good Clinical Practice Guidelines and was approved by the institutional review boards and independent ethics committees of the participating centers. All participants provided informed consent before entering the study.

Originally, a total of 515 adults started in the reimbursement program and were included in the previous analysis (13). Since four centers stopped collecting data after 12 months, only participants of the 13 centers that collected data up to 24 months (n = 441) were included in this analysis. Minimum criteria to receive reimbursement were previously reported (13). Basically, the specialized diabetes centers were free to decide to which adults with type 1 diabetes on CSII they would offer rtCGM reimbursement. Every person who entered the reimbursement program between September 2014 and January 2017 was included, without exception, in the study after informed consent.

The primary end point was evolution over time of HbA_1c between baseline and 24 months after the start of rtCGM reimbursement. Secondary end points were effect of rtCGM on acute diabetes complications (hypoglycemia and/or ketoacidosis), work absenteeism, quality of life, proportion of participants with HbA_1c <7% (<53 mmol/mol), and reasons for discontinuing rtCGM. Additional post hoc analyses examined how many participants reached clinical consensus targets (17) and clinical composite end points (18).

There was no restriction in the devices participants could use in the reimbursement program. Participants were able to switch between insulin pump and glucose sensor brands or switch to newer versions. This led to a shift from low-glucose threshold suspend systems that were used at the start of the study toward low-glucose predictive suspend systems at the end (Supplementary Table 1).

In the Belgian rtCGM reimbursement program, 5 h of specialized education required for the correct application of rtCGM were provided in the first 4 months of the program followed by an additional 1 h of specialized education per 12 months of follow-up, in addition to standard-of-care diabetes education. Specialized education in the rtCGM reimbursement program taught participants how to use rtCGM, including all aspects of monitoring, as well as how to react to the obtained glycemic values.

Prespecified clinical data were collected from a period of 12 months before until 24 months after the start of the reimbursement program. Information about clinical parameters was collected from clinical files at baseline and 4, 8, 12, and 24 months. HbA_1c levels were averaged for prespecified time points: prereimbursement/baseline (before = −12 months until −1 day), 4 months (±2 months), 8 months (±2 months), 12 months (±2 months), and 24 months (±2 months) after start of reimbursement.

Questionnaires (36-Item Short Form Health Survey [19], Problem Areas in Diabetes Short Form [20], and Hypoglycemia Fear Survey [HFS] worry subscale [21]) and standardized diaries (13) were completed at baseline and after 12 and 24 months and scored manually. For patient-reported days off from work, all reasons owing to diabetes were taken into account without differentiating between hypo- and hyperglycemia but excluding consultation appointments with the diabetes team. Patient-reported emergency department admissions and hospitalizations for hypoglycemia (any event) and/or proven ketoacidosis were validated using hospital records in the individual centers.

rtCGM data were collected using the designated diabetes management software from the various manufacturers. Data for the following time points were extracted and averaged: data from entry in the reimbursement program (2 weeks = week 0 until week 2), 4 months (±2 months), 8 months (±2 months), 12 months (±2 months), and 24 months (±2 months) after start of reimbursement. An overview of data completeness is available in Supplementary Table 2.

For data analysis, the full analysis set was used, which comprised all patients who were registered as receiving reimbursement for rtCGM in 1 of the 13 centers. In total, 441 adults were included in the analysis, which gave, even with the observed drop-out rate, enough power (>80%) with a two-sided 5% significance level to detect a mean difference in HbA_1c of 0.3% (4 mmol/mol).

Using a linear mixed model, we evaluated HbA_1c and quality of life, as a function of time, with a random effect of center to handle the correlation between patients of the same center and an unstructured covariance matrix for the five or three repeated measurements within the same patient. By using a linear mixed model, patients with missing data still contributed to the analyses. For evolution of HbA_1c, values at 4, 8, 12, and 24 months were compared with the average value from −12 months until −1 day (before = baseline). For evolution of quality of life, scores on the different questionnaires at 12 and 24 months were compared with the scores at the start of reimbursement. From the multivariable normal distribution implied by the linear mixed model, we derived the relation between baseline HbA_1c and changes in HbA_1c versus baseline. Taking regression to the mean into account, the obtained correlation is not tested versus 0 but versus the correlation, which is already expected purely on the basis of regression to the mean (22). A logistic regression model with generalized estimating equations was used to evaluate the evolution of proportion of participants who reached target HbA_1c (<7% [53 mmol/mol]), who reached clinical consensus targets, who reached composite end points (17,18), with hospitalizations for hypoglycemia or ketoacidosis, with diabetes-related work absenteeism, and with acute hypoglycemic events. Differences in days of diabetes-related work absence, number of hospitalizations for hypoglycemia or ketoacidosis, and acute hypoglycemic events per 100 patient-years were assessed with a negative binomial generalized estimating equation model. Participants who were incapable of working because of disability were excluded.

A Bonferroni-Holm correction was considered for results at 24 months referring to the primary outcome: evolution of HbA_1c for the total population. No adjustment was made for multiple testing of secondary end points.

Post hoc, all analyses were repeated for participants with and without IAH. HbA_1c evolution was also assessed for groups of baseline HbA_1c. The number of participants in these subgroups at baseline and 4, 8, 12, and 24 months is shown in Supplementary Table 3. Differences between the subgroups at different time points were compared using the Mann-Whitney U test for continuous data and with the χ² test for dichotomous data. Statistical analyses were performed using SPSS for Windows software (IBM Corporation, Armonk, NY).

Baseline characteristics of the 441 participants are shown in Supplementary Table 4. In short, the majority was highly educated and had a long history of type 1 diabetes. On average, they had 6 years of CSII experience at baseline, 56% had hypoglycemia as indication to start rtCGM, and 44% had IAH.

Of 441 adults who started rtCGM in 1 of 13 centers, 87.8% (n = 387) and 81.6% (n = 360) had >12 and 24 months of follow-up, respectively. In total, 81 participants (18.4%) stopped using rtCGM (Supplementary Fig. 1). Participants could have multiple reasons for deciding to stop rtCGM. The most frequent reason for discontinuation was related to the system itself, such as alarm fatigue (n = 27 [6%]). Other reasons were local and/or technical problems (n = 20 [5%]), no apparent benefit for patient and/or physician (n = 20 [5%]), and <70% usage of rtCGM (n = 17 [4%]).

Mean percentage of rtCGM wear time was high throughout the 24 months and remained stable, with 87.1 ± 9.6%, 86.6 ± 8.4%, 86.9 ± 9.5%, and 87.1 ± 10.4% at 4, 8, 12, and 24 months, respectively.

For the total population, HbA_1c was sigcantly lower at 24 months (7.37% [95% CI 7.19–7.55] [57 mmol/mol (55–59)]) compared with baseline (7.64% [7.46–7.82] [60 mmol/mol (58–62)], P < 0.0001), and was stable compared with 12 months (7.34% [7.16–7.52] [57 mmol/mol (55–59)], P = 0.316) (Fig. 1A). A stronger decrease in HbA_1c was observed in participants with higher baseline HbA_1c, although this correlation never exceeded the regression to the mean effect (Fig. 1B). There was no difference in evolution of HbA_1c for participants with and without IAH (Fig. 1C).

The prevalence of acute diabetes complications was lower throughout the study than in the year before. This was already apparent in the 1st year, but was confirmed in the 2nd year. The largest benefit was seen for hypoglycemia-related events for which we gathered data on different levels, from hospitalizations for any hypoglycemic event to receiving glucagon. In some cases, this may have led to a simultaneous decrease in diabetes-related work absenteeism (Table 1).

The decline in hypoglycemia-related events was seen in participants with and without IAH, but those with IAH had higher baseline prevalence and a larger proportion of reduction at follow-up than those with normal hypoglycemia awareness (Fig. 2). Participants with IAH missed, on average, 709 days of work per 100 patient-years in the year before the study, which dropped significantly to 107 days after 24 months (P < 0.0001). For those with normal awareness, this reduced to a lesser extent from 264 days in the year before to 66 days per 100 patient-years at 24 months (P = 0.048).

For the total population, previously observed improvements in general quality of life, as measured by the 36-Item Short Form Health Survey, were sustained throughout the 24-month study. Problem Areas in Diabetes Short Form scores overall decreased by −1.3 points (95% CI −1.7 to −0.9, P < 0.0001), and the worry subscale of HFS was also lower through 24 months of follow-up (18.3 [16.7–19.9] at baseline vs. 14.1 [12.5–15.7] after 24 months, P < 0.0001) (Supplementary Table 5).

Both those with and those without IAH showed improvement in quality of life scores. However, improvement in participants with IAH tended to be higher partly because of the lower perceived quality of life at baseline. This is particularly evident for the HFS worry subscale, for which they had worse baseline scores (20.7 ± 10.8 vs. 16.2 ± 9.5 for IAH vs. non-IAH, respectively, P < 0.0001) and were able to bring them closer to the level of those without IAH during follow-up (Supplementary Table 5).

Because of the observational real-world study design, no blinded glucose measuring period was available. Therefore, we report on the percentage of participants who reached the clinical consensus targets as measured by rtCGM from the first 2 weeks until 24 months onward. For HbA_1c targets, data were available up to 1 year before start.

Compared with the year before rtCGM reimbursement, more participants were able to obtain HbA_1c below the target level of 7% (53 mmol/mol) (Table 2). More participants could attain the targets of time spent <70 mg/dL (<3.9 mmol/L) and <54 mg/dL (<3.0 mmol/L) after 24 months (Table 2). Of those who did not reach these hypoglycemia consensus targets in the first 2 weeks, 53.8% and 48.4% did reach the targets for time <54 mg/dL (<3.0 mmol/L) and <70 mg/dL (<3.9 mmol/L) after 24 months, respectively.

The proportion of participants who reached consensus targets of time in range (TIR) and time >250 mg/dL (>13.9 mmol/L) did not significantly change with even a trend toward a small reduction during follow-up. However, a significant reduction was observed in the number of participants who spent <25% of time >180 mg/dL (>10.0 mmol/L) (Table 2).

The number of participants who reached the combined end points of HbA_1c <7% (<53 mmol/mol) with <1% of time spent <54 mg/dL (<3.0 mmol/L) and HbA_1c <7% (<53 mmol/mol) without severe hypoglycemic episodes more than doubled during the study. This was not observed for the combined end points with TIR (Table 2).

Both participants with and participants without IAH benefited from rtCGM, with increased proportions in both groups achieving the predefined targets for hypoglycemia throughout 24 months of follow-up. While there were no differences within or between groups for targets of TIR and hyperglycemia, more participants with normal awareness reached consensus targets for hypoglycemia than those with IAH (Supplementary Table 6).

This study tried to provide more insight into how people with type 1 diabetes use advanced technology to manage their diabetes and how this influences daily life in the long run. To our knowledge, the RESCUE study is the largest and one of the longest prospective real-world cohort studies to assess clinical and patient-reported outcome measures after initiation of rtCGM reimbursement in the long term. As reported here, rtCGM use by adults with type 1 diabetes on CSII therapy followed in 13 specialized centers was associated with 24 months of sustained improvements in HbA_1c; quality of life, especially for fear of hypoglycemia; and acute hypoglycemic events.

Although the clinical benefits of rtCGM have been demonstrated in numerous randomized controlled trials (2–5,7–9), the trials often lacked sufficient length to inform us about the long-term sustainability and clinical impact of rtCGM use. Moreover, stringent inclusion and exclusion criteria often preclude high-risk populations from participation. Therefore, real-world evidence is needed to evaluate the long-term effectiveness of already-approved technology (23,24). To our knowledge, RESCUE is the largest prospective real-world study where adults with type 1 diabetes were followed for 2 years while using rtCGM, which allowed us to distinguish study effects from sustained benefits. Only two other prospective observational studies were of longer duration. First, the prospective Comparison of Sensor-Augmented Insulin Regimens (COMISAIR) study followed 94 people divided into four groups (rtCGM + CSII, rtCGM + MDI, capillary tests + CSII, and capillary tests + MDI) and lasted 3 years (6). The investigators showed that rtCGM combined with CSII or MDI were equally superior to capillary finger stick tests with CSII or MDI regarding HbA_1c and time spent in hypoglycemia. Second, a study by Gómez et al. (12) prospectively followed 111 adults with type 1 diabetes starting sensor-augmented pump therapy because of hypoglycemia. Mean follow-up time was 47 months, with only 50 patients followed for >40 months. This poorly controlled population could achieve an HbA_1c reduction of −1.7% (−19 mmol/mol) together with a reduction in severe hypoglycemic events. In addition to these studies, our study provided an association between rtCGM use in a large population and the long-term sustainability of its benefits regarding clinical and patient-reported outcome measures within the context of real-world diabetes self-management and sufficient diabetes education (25).

Because the diverse risks of recurrent and severe hypoglycemia have been identified in the literature (26), it is important that hypoglycemia is prevented through the use of rtCGM. In the unique Belgian rtCGM reimbursement system (13), the diabetes teams, independently of each other or of predefined criteria, selected a population with a high prevalence of hypoglycemia-related acute complications, which is now included as a main indication for rtCGM reimbursement by other countries (27) and is acknowledged by the international community as one of the most important factors contributing to why people should use continuous glucose sensors (18). Our results show that the number of clinically severe hypoglycemic events can be markedly reduced by use of rtCGM. Importantly, the improvement in HbA_1c indicates that hypoglycemia reduction was not achieved at the expense of a deterioration of overall glycemic control. Together with findings from other studies addressing the use of rtCGM in adults prone to hypoglycemia, this indicates that rtCGM can effectively address problematic hypoglycemia in people treated by MDI as well as by CSII (4,7,9,12).

In the RESCUE population, 44% of participants had IAH in varying degrees. This is two to three times more than what has been described in the type 1 diabetes community (14,16). It was apparent from frequencies of hypoglycemia-related hospitalizations and severe hypoglycemic events that participants with IAH had a higher risk of developing such acute complications, something that has already been described previously by others (16,28,29). Previous studies have not found evidence that the use of rtCGM could improve hypoglycemia awareness (4,7). One study suggested that improvement in IAH can be achieved through structured education and frequent contact irrespective of the treatment modality or use of rtCGM (30). The effect of this structured education could even be maintained when people return to standard care, switch from CSII to MDI or vice versa, and do not wear their sensor for a sufficient amount of time (31). Therefore, the best option to effectively manage people with IAH is to implement a combination of rtCGM (with or without CSII per preference) and structured education with frequent follow-up contacts (32).

Our study reports on the proportion of people treated by rtCGM and CSII achieving the consensus targets for glycemic control (17) in real life. rtCGM and sensor-augmented pumps focus primarily on hypoglycemia avoidance, which helped in reaching the consensus targets for hypoglycemia in ∼70% of the RESCUE population. Of note, our population gradually transitioned to devices with more advanced algorithms as they were introduced into the market during the duration of the study, which could have led to some being able to further prevent hypoglycemia. On the other hand, reaching targets for TIR and hyperglycemia proved to be more difficult, with barely 30% achieving the recommended levels. We even observed a small trend toward fewer participants achieving targets for TIR and hyperglycemia, an observation that has been previously described in studies with sensor-augmented pumps with the low-glucose predictive suspend feature (33,34). A possible reason for this finding may be attributed to how the patient manages a predictive insulin pump suspension, namely the consumption of carbohydrates in addition to insulin suspension to correct for a future hypoglycemic episode (34).

We incorporated quality-of-life questionnaires, which are powerful tools to inform other patients, clinicians, and policymakers (35). Management of type 1 diabetes is a daily task with a considerable burden on quality of daily living. An important driver of this burden is hypoglycemia because it can have a negative impact on relationships, sleep quality, employment, and body image as a result of heightened levels of stress and anxiety (36). We provide further evidence that hypoglycemia has debilitating effects on quality of life, as shown by the overall lower perceived health status at baseline of participants with IAH. Nevertheless, the use of technology that helps to identify and prevent hypoglycemia, in this case rtCGM, has proven to be a vital component to normalize daily life for these people, which has also been demonstrated in previous studies (2,4,7,9,11,31).

This study has limitations. First, the RESCUE study originally was a collaboration among 17 centers with a total of 515 participants. However, contribution of data was voluntary, and four centers found the workload too high to keep collecting the data, which contributed to the total drop-out rate of 30%. Therefore, for the complete analysis up to 24 months, we only used data of the 13 centers that contributed until the end of the study. Second, since RESCUE was a nonrandomized observational trial, it is possible that factors other than rtCGM use could affect the studied outcome measures. For example, it is possible that diabetes education that was provided when starting rtCGM sparked the motivation of participants to get their diabetes back on track, apart from rtCGM use. However, this peak in motivation is known to fade after some time (37). Nevertheless, we observed a sustained benefit even after 2 years, which contributes to the rationale that rtCGM instigates altered behavior. Finally, because of the real-world design, a variety of sensor-insulin pump combinations with different suspend features and levels of accuracy were used. This has to be taken into account, in particular when interpreting the hypoglycemia outcomes.

In conclusion, rtCGM led to improvements in hypoglycemia-related glucose control over 24 months in people with type 1 diabetes on an insulin pump. The sustained higher prevalence of people achieving the consensus targets for hypoglycemia, with fewer hypoglycemia-related acute episodes and diabetes-related days off from work, could have been an important factor in the improvement of quality of life, especially fear of hypoglycemia.

Acknowledgments. The authors thank the data nurses, the local investigators, and their teams for monitoring the patients, completing the case reporting files, and collecting data.

Funding. S.C. received a doctoral grant for strategic basic research and P.G. received a grant for a clinical PhD fellowship, both from Fonds Wetenschappelijk Onderzoek.

Duality of Interest. S.C. received travel grants from Medtronic and Roche. C.D.B. reports consulting fees and honoraria for speaking for Abbott, AstraZeneca, Boehringer Ingelheim, A. Menarini Diagnostics, Eli Lilly, Medtronic, Novo Nordisk, and Roche. F.N. reports consulting fees and honoraria for speaking from Abbott, AstraZeneca, Boehringer Ingelheim, Eli Lilly, Johnson & Johnson, Medtronic, Merck Sharp & Dohme, Novo Nordisk, Roche, and Sanofi. K.S. reports consulting fees and honoraria for speaking from Abbott, AstraZeneca, Boehringer Ingelheim, Eli Lilly, Medtronic, and Merck Sharp & Dohme. E.W. reports consulting fees and honoraria for speaking from AstraZeneca, Boehringer Ingelheim, Eli Lilly, Medtronic, Merck Sharp & Dohme, Novo Nordisk, and Roche. C.V. reports consulting and speaking fees from Medtronic, Boehringer Ingelheim, AstraZeneca, and Sanofi. C.M. serves or has served on the advisory panels for Novo Nordisk, Sanofi, Merck Sharp & Dohme, Eli Lilly, Novartis, AstraZeneca, Boehringer Ingelheim, Hanmi Pharmaceuticals, Roche, Medtronic, ActoBio Therapeutics, Pfizer, Dianax, and Union Chimique Belge. Financial compensation for these activities has been received by KU Leuven. KU Leuven has received research support for C.M. from Medtronic, Novo Nordisk, Sanofi, Merck Sharp & Dohme, Eli Lilly, Roche, Abbott, ActoBio Therapeutics, and Novartis. C.M. serves or has served on the speakers’ bureau for Novo Nordisk, Sanofi, Merck Sharp & Dohme, Eli Lilly, Boehringer Ingelheim, AstraZeneca, and Novartis. Financial compensation for these activities has been received by KU Leuven. P.G. serves or has served on the advisory panels for Novo Nordisk, Sanofi, Boehringer Ingelheim, Janssen Pharmaceuticals, Roche, Medtronic, and Bayer. Financial compensation for these activities has been received by KU Leuven. P.G. serves or has served on the speakers’ bureau for Merck Sharp & Dohme, Boehringer Ingelheim, Bayer, Medtronic, Abbott, and Roche. Financial compensation for these activities has been received by KU Leuven. KU Leuven received for P.G. nonfinancial support for travel from Sanofi, A. Menarini Diagnostics, Medtronic, and Roche. All disclosures were unrelated to the present work. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. S.C. collected and analyzed the data, performed statistical analyses, discussed and wrote the manuscript, and made figures and tables. C.D.B., F.N., R.P.R., I.L., A.M., D.S., K.S., M.S., E.W., Y.T., C.V., and B.K. collected and discussed the data and edited the manuscript. C.M. and P.G. designed the study, analyzed and discussed the data, and wrote the manuscript. S.C. and P.G. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at the 13th International Conference on Advanced Technologies & Treatments for Diabetes, Madrid, Spain, 19–22 February 2020.