MRIGTCNLGE:DT YTM N DEVICES AND SYSTEMS DATA TECHNOLOGIES: EMERGING Care 1

Jan Soupal,ˇ 1 Lenka Petruzelkovˇ a,´ 2 Glycemic Outcomes in Adults George Grunberger,3,4,5 Aneta Haskov´ a,´ 1 Milan Flekac,ˇ 1 Martin Matoulek,1 With T1D Are Impacted More by Ondˇrej Mikes,ˇ 1 Toma´sˇ Pelcl,1 Jan Skrhaˇ Jr.,1 Eva Horova,´ 1 Jan Skrha,ˇ 1 Continuous Christopher G. Parkin,6 Stˇ epˇ an´ Svacina,ˇ 1 Than by Delivery Method: and Martin Prazn´ y´1 3 Years of Follow-Up From the COMISAIR Study https://doi.org/10.2337/dc19-0888

OBJECTIVE This study assessed the clinical impact of four treatment strategies in adults with (T1D): real-time continuous glucose monitoring (rtCGM) with multiple daily insulin injections (rtCGM1MDI), rtCGM with continuous subcuta- neous insulin infusion (rtCGM1CSII), self-monitoring of glucose with MDI (SMBG1MDI), and SMBG with CSII (SMBG1CSII).

RESEARCH DESIGN AND METHODS This 3-year, nonrandomized, prospective, real-world, clinical trial followed 94 par- 1 ticipants with T1D (rtCGM1MDI, n 5 22; rtCGM1CSII, n 5 26; SMBG1MDI, n 5 21; 3rd Department of Internal Medicine, 1st Fac- ulty of Medicine, Charles University, Prague, SMBG1CSII, n 5 25). The main end points were changes in A1C, time in range (70– 180 mg/dL [3.9–10 mmol/L]), time below range (<70 mg/dL [<3.9 mmol/L]), 2Department of Paediatrics, 2nd Faculty of Med- glycemic variability, and incidence of . icine, Charles University, Prague, Czech Republic 3Grunberger Diabetes Institute, Department of Internal Medicine and Center for Molecular RESULTS Medicine & Genetics, Wayne State University At 3 years, the rtCGM groups (rtCGM1MDI and rtCGM1CSII) had significantly lower School of Medicine, Bloomfield Hills, MI 4 A1C (7.0% [53 mmol/mol], P 5 0.0002, and 6.9% [52 mmol/mol], P < 0.0001, Department of Internal Medicine, William Beau- 1 1 mont School of Medicine, Oakland University, respectively), compared with the SMBG CSII and SMBG MDI groups (7.7% Rochester, MI [61 mmol/mol], P 5 0.1.000, and 8.0% [64 mmol/mol], P 5 0.3574, respectively), 5Visiting Professor, 1st Faculty of Medicine with no significant difference between the rtCGM groups. Significant improvements Charles University, Prague, Czech Republic 6 in percentage of time in range were observed only in the rtCGM subgroups CGParkin Communications, Inc., Henderson, NV ˇ (rtCGM1MDI, 48.7–69.0%, P < 0.0001; and rtCGM1CSII, 50.9–72.3%, P < 0.0001) Corresponding author: Jan Soupal, jan.soupal@ seznam.cz and significant reductions in time below range (9.4–5.5%, P 5 0.0287; and 9.0–5.3%, Received 2 May 2019 and accepted 6 August 2019 P 5 0.0325, respectively). Seven severe hypoglycemia episodes occurred: SMBG n 5 n 5 This article contains Supplementary Data online groups, 5; sensor-augmented insulin regimens (SAIR) groups, 2. at http://care.diabetesjournals.org/lookup/suppl/ doi:10.2337/dc19-0888/-/DC1. CONCLUSIONS © 2019 by the American Diabetes Association. rtCGM was superior to SMBG in reducing A1C, hypoglycemia, and other end points in Readers may use this article as long as the work individuals with T1Dregardless of their insulin delivery method. rtCGM1MDI can be is properly cited, the use is educational and not for profit, and the work is not altered. More infor- considered an equivalent but lower-cost alternative to sensor-augmented insulin mation is available at http://www.diabetesjournals pump therapy and superior to treatment with SMBG1MDI or SMBG1CSII therapy. .org/content/license. Diabetes Care Publish Ahead of Print, published online September 17, 2019 2 3-Year Follow-up on COMISAIR Study Diabetes Care

Use of real-time continuous glucose either CSII (sensor-augmented pump willingness to participate in a 4-day monitoring (rtCGM) has emerged as a [rtCGM1CSII]) or MDI (rtCGM1MDI)– training program at baseline. Exclusion critical component of diabetes self- on glycemic control compared with criteria were as follows: use of rtCGM management for individuals treated the addition of CSII (SMBG1CSII) or within the previous 3 months, ketoaci- with intensive insulin regimens, and it MDI (SMBG1MDI) (19) among 65 indi- dosis within the previous 3 months, is now considered a standard of care for viduals with T1D. At study end, significant concomitant therapy influencing glu- these patients (1–6). A1C reductions from baseline were ob- cose metabolism, pregnant or plan- Recent randomized clinical trials have served in both the SAIR groups ning pregnancy, and demonstrated demonstrated that use of rtCGM results (rtCGM1CSII: 21.1% [212.0 mmol/ nonadherence to current treatment in significant improvements in glycemic mol], P 5 0.0025; rtCGM1MDI: 21.3% regimen. control and hypoglycemia and confers a [214.2 mmol/mol], P 5 0.0034). Although higher quality of life to participants trea- SMBG1CSII use also led to a significant Procedures ted with multiple daily insulin injections A1C reduction (0.5% [5.5 mmol/mol]), Enrolled participants were scheduled (MDIs) compared with traditional self- no significant reductions were seen in for a total of 15 clinic visits (baseline, monitoring of blood glucose (SMBG) the SMBG1MDI group. The increase at week 2, and then at months 1, 3, 6, 9, (7–11). Similar improvements in A1C from baseline in average number of 12, 15, 18, 21, 24, 27, 30, 33, and 36). A and hypoglycemia have also been ob- boluses per day was significantly greater detailed description of the study proce- served in participants using rtCGM with in the rtCGM1CSII and rtCGM1MDI dures was previously published (19). therapy (12,13). Significant groups (3.2 and 2.2, respectively, both At the initial visit, investigators con- reductions in severe hypoglycemia have P , 0.0001) compared with SMBG1CSII firmed eligibility and initiated profes- also been observed in patients with (0.6, P 5 0.08). No increase was seen in sional CGM (iPro2; Medtronic, Northridge, type 1 diabetes (T1D) with problematic the SMBG1MDI group. Importantly, sig- CA) in all participants for 6 days. hypoglycemia who were treated with nificant reductions in percentage of Throughout the study, participants in rtCGM in combination with either MDI time in hypoglycemia (,70 mg/dL [,3.9 the groups not using SAIR had profes- (10) or insulin pump therapy (13). mmol/L]) were observed only in the sional CGM every 3 months. Participants Importantly, a common observation in SAIR groups, from 8 6 4% to 6 6 43%, then attended a structured 4-day train- most rtCGM studies is that glycemic P , 0.01. ing program that addressed basic insulin improvements and other benefits were In the current follow-up study, we administration skills, including timing and dependent upon the persistence of sen- investigated the effects of SAIR interven- dosing of preprandial insulin, prevention sor use (7–15). tions on glycemic control and treatment of hypoglycemia, and theoretical and Although randomized controlled trials persistence among a larger participant practical counting. Par- (RCTs) are recognized as the highest level cohort after 3 years, providing further ticipants were encouraged to use flex- of evidence regarding the efficacy of supportive evidence for the use of rtCGM ible insulin dosing. rtCGM when used within tightly con- in the management of T1D. During training, all treatment mo- trolled settings, our understanding of dalities (rtCGM1MDI, rtCGM1CSII, the real-world use and benefits of rtCGM RESEARCH DESIGN AND METHODS SMBG1MDI, and SMBG1CSII) were in- has been limited. Findings from RCTs The COMISAIR-2 study was the 3-year troduced to participants. In collaboration often fail to reflect actual participant follow-up of the COMISAIR trial (19), with study clinicians, participants se- behaviors and resultant outcomes in which compared the efficacy of the lected their treatment modality accord- real-world clinical practice (16–18). More- long-term use of SAIR regimens among ing to their individual needs and over, there have been few long-term individuals. Participants were recruited preferences. Investigator influence on comparisons to evaluate the efficacy from the participant population treated participant decisions was minimal (6% of rtCGM use in combination with the at the 3rd Department of Internal Med- of cases), and no participant was dis- various insulin delivery methods (e.g., icine, 1st Faculty of Medicine, Charles couraged from using one of the SAIR rtCGM 1 continuous subcutaneous in- University. This report includes results regimens. Participants in the SAIR and sulininfusion[CSII]vs.rtCGM1MDI), and from an additional 29 participants whose CSII groups completed theoretical train- conclusive evidence of rtCGM benefits complete 1-year data were not available ing on the relevant devices, followed by compared with SMBG has been sparse. at the conclusion of the initial COMISAIR treatment initiation and practical train- Because is pri- trial. The study was approved by an ing (including insulin adjustment) with marily dependent on participant behav- independent ethics review board and investigators. ior, different research approaches are conducted in accordance with the Dec- Participants using SAIR were encour- needed to more definitively assess these laration of Helsinki (20). All subjects aged to make self-adjustments to their behavior-based interventions. provided written informed consent be- treatment using rtCGM values (hypergly- We recently reported findings from the fore enrollment. cemia and hypoglycemic alerts and Comparison of Sensor Augmented In- Inclusion criteria were as follows: trends) and to incorporate results of sulin Regimens (COMISAIR) study, a 1- age .18 years, .2 years T1D duration, SMBG into treatment changes. Partici- year, nonrandomized, real-world study A1C 7.0–10% (53–86 mmol/mol), treated pants in non-SAIR groups were encour- that assessed the efficacy of long-term with analog , willingness to use aged to measure their blood glucose at use of sensor-augmented insulin regi- sensors .70% of the time or perform least four times per day. All participants mens (SAIR)–rtCGM combined with SMBG four or more times per day, and were instructed to use only the study care.diabetesjournals.org Soupalˇ and Associates 3

Table 1—Baseline characteristics Characteristic rtCGM 1 MDI (n 5 22) rtCGM 1 CSII (n 5 26) SMBG 1 CSII (n 5 25) SMBG 1 MDI (n 5 21) P value Male (%) 59 50 48 52 0.89 Age (years) 32.6 6 11.5 32.3 6 9.9 33 6 9.3 35 6 15 0.95 Duration of diabetes (years) 13.7 6 9.8 14.6 6 7.8 13.4 6 8.4 13.5 6 8.8 0.86 A1C (mmol/mol) 66.6 6 10.0 66.5 6 10.2 67.3 6 9676 8.6 0.95 A1C (%) 8.2 6 0.9 8.2 6 0.9 8.3 6 0.8 8.3 6 0.8 0.92 Mean sensor glucose (mmol/L) 10.5 6 1.4 10.3 6 1.5 10.4 6 1.6 10.4 6 1.3 0.89 BMI (kg/m2)266 4256 4256 3256 3 0.91 Body weight (kg) 76.6 6 14 72.5 6 15 74 6 11 73.7 6 13 0.96 The total daily dose of insulin (units) 48.1 6 15 46.2 6 11.5 46.7 6 11.4 48.8 6 13.5 0.93 Relative proportion of bolus insulin (%) 48.7 6 3.9 48.7 6 4 50.1 6 4.4 50 6 4.4 0.61 No. of boluses/day (n) 3.9 6 0.9 3.8 6 0.8 3.8 6 0.9 3.8 6 0.7 0.99 Frequency of SMBG/day (n) 3.7 6 1 3.7 6 1.2 3.8 6 1.1 3.6 6 1 0.95 Values are presented as mean 6 SD. blood glucose meter provided to them participants at baseline and every percentage of time ,70 mg/dL (,3.9 for all SMBG measurements taken dur- 3 months in SMBG participants. All par- mmol/L), rtCGM usage (SAIR groups), ing this trial. ticipants were provided with a personal change in average number of boluses At each clinic visit, participants were blood glucose meter (OneTouch [Life- per day, and incidence of hypoglycemia. screened for adverse events, sensor in- Scan, Milpitas, CA] or CONTOUR LINK sertion sites were inspected (SAIR partic- [ Diabetes Care, Basel, Switzer- Measures ipants), and data from all rtCGM systems, land]), which was used for diabetes A1C values were measured at the base- insulin pumps, and blood glucose meters self-management purposes and calibra- line and then every 3 months until were downloaded for analysis. tion of rtCGM. We highlighted to par- study end. A1C was analyzed by a high- ticipants the importance of regular performance liquid chromatography Glucose Monitoring Devices downloading and review of the data method on a Variant II analyzer (Bio- Participants in the CSII group wore one from rtCGM devices and insulin pumps. Rad, Hercules, CA). The normal refer- of two types of insulin pumps: MiniMed A bolus calculator was set for all partic- ence range of A1C in our laboratory is – – Paradigm Veo (Medtronic) and Animas ipants with insulin pumps. 4.0 6.0% (20 42 mmol/mol). Initially, Vibe (Animas Corporation, West Chester, all patients were monitored by profes- PA). Participants in the rtCGM1CSII sub- Outcomes sional CGM for 6 days. Then, throughout groupusedeithertheMiniMedParadigm The primary end point was the difference the study, subjects in the groups not Veo System with Enlite sensors (Med- in A1C between the groups after 3 years using rtCGM were assessed by pro- tronic) or Animas Vibe system with Dex- of follow-up. Secondary end points were fessional CGM for 6 days every 3 ComG4 sensors (Dexcom, San Diego, CA). as follows: change in glycemic variability months. The subgroup of participants who se- (expressed as the total SD of blood Severe hypoglycemia was defined as lected rtCGM1MDI therapy used a glucose, average daily glucose from an episode requiring assistance from DexCom G4 rtCGM system. The iPro2 CGM, and percentage of time spent in another person or neurological recovery was used for glucose monitoring in all range 70–180 mg/dL [3.9–10.0 mmol/L]), in response to restoration of plasma glucose to normal. Ketoacidosis was de- fined as an episode of (.252 mg/dL [.14 mmol/L]) with low serum bicarbonate (,15 mmol/L), low pH (,7.3), or both together with either ketonemia or ketonuria that required treatment in a health care facility.

Statistical Analysis The basic characteristics of each group were analyzed using nonparametric tests (Kruskal-Wallis and ANOVA). The data of repeated measurements (obtained every 3 months) such as the mean glu- cose levels, time in/below target range, and glycemic variability were compared using a linear mixed-effects model. P Figure 1—Change in A1C from baseline by study group. SAP, sensor-augmented pump. values ,0.05 were considered statistically 4 3-Year Follow-up on COMISAIR Study Diabetes Care

Table 2—Significance of A1C change over 36 months Month Month Month Month Month Month Month Month Month Month Month Month 3 6 9 12 15 18 21 24 27 30 33 36 rtCGM1MDI 0.0017 0.0006 ,0.0001 ,0.0001 ,0.0001 0.0003 ,0.0001 ,0.0001 0.0001 0.0004 0.0006 0.0002 rtCGM1CSII ,0.0001 ,0.0001 ,0.0001 ,0.0001 ,0.0001 ,0.0001 ,0.0001 ,0.0001 ,0.0001 ,0.0001 0.0001 ,0.0001 SMBG1MDI 1.0000 1.0000 1.0000 0.3914 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 SMBG1CSII 1.0000 1.0000 1.0000 0.0056 0.0413 0.2539 0.1756 0.2096 0.7677 0.2954 1.0000 0.3574

significant. Analyses were conducted using changes in each study group at all study rtCGM1MDI versus SMBG1MDI, 14.21% the R statistical package, version 3.1.1. Data visits. (95% CI 6.45 to 222, P 5 0.0007); are expressed as mean 6 SD values. Forty-eight percent (n 5 23) of SAIR rtCGM1MDI versus SMBG1CSII, 11.13% participants achieved ,7.0% A1C at (95% CI 4.46–17.81, P 5 0.0016); RESULTS 3 years (rtCGM1MDI, 43% [n 5 9]; rtCGM1CSII versus SMBG1MDI, 17.58% Baseline Characteristics and rtCGM1CSII, 56% [n 5 14]) compared (95% CI, 10.9–24.27, P , 0.0001); and Adherence with 9% (n 5 2) of SMBG1CSII and 16% rtCGM1CSII versus SMBG1CSII, 14.5% A total of 94 participants were enrolled (n 5 3) of SMBG 1MDI participants. (95% CI 8.82–20.19, P , 0.0001). No in the study; 88 completed all study Between-group comparisons of A1C differences were seen between the visits. Among the six participants who changes showed significant differences rtCGM1MDI and rtCGM1CSII groups. discontinued the study, two SMBG1CSII between the SAIR and SMBG groups at Significant reductions in percentage of participants and one rtCGM1CSII partic- 3 years, favoring use of rtCGM (Table 3). time below range (,70 mg/dL [,3.9 ipant withdrew for personal reasons, No significant differences between the mmol/L]) were seen in the rtCGM1MDI one SMBG1CSII participant decided SAIR subgroups or SMBG subgroups were (P 5 0.0387) and rtCGM1CSII (P 5 to initiate rtCGM after 1 year, one observed. 0.0235) groups but not the SMBG1CSII rtCGM1MDI initiated rtCGM1CSII, Significant differences between the (0.4847) or SMBG1MDI (P 5 1.000) and one SMBG1MDI participant died rtCGM1MDI group and SMBG groups groups (Fig. 2). due to breast cancer. Baseline character- were observed beginning at month 6, Insulin Boluses istics were similar in the four study whereas the differences between the At study end, the average number of 1 groups (Table 1). rtCGM CSII group and SMBG groups boluses per day was lower in both SMBG All SAIR participants wore their sen- were observed beginning at month 3. groups in comparison with the rtCGM sors .70% of the time. No significant Average Sensor Glucose groups (6.9 6 1.9 vs. 4.5 6 1.1, P , changes in total insulin dose or body Significant differences in improvements 0.0001). A higher frequency of boluses weight were observed in any of the study in average sensor glucose were seen in wasseeninparticipantswithSMBG1CSII groups. the rtCGM1MDI and rtCGM1CSII versus the self-reported boluses in the groups but not in the SMBG1CSII or SMBG1MDI group (4.9 6 1.2 vs. 4.1 6 Primary and Secondary End Points SMBG1MDI groups (Table 3). No signif- 0.8, P 5 0.02). No significant difference Change in A1C icant between-group differences within between rtCGM1CSII and rtCGM1MDI 1 At 3 years, the rtCGM MDI and the SAIR or SMBG subgroups were wasobserved (7.1 6 1.9vs. 6.66 1.9,P 5 1 fi rtCGM CSII groups had signi cantly observed. 0.4) (Supplementary Table 2). lower A1C (7.0% [53 mmol/mol], P 5 0.0002, and 6.9% [52 mmol/mol], P , Glycemic Variability rtCGM Use fi 0.0001, respectively), compared with the Signi cant differences in glycemic vari- Mean percentage use of rtCGM in the 1 SMBG1MDI and SMBG1CSII groups ability were observed between rtCGM SAIR groups was high throughout the 1 1 (8.0% [64 mmol/mol], P 5 0.3574, and MDI versus SMBG MDI, rtCGM CSII study period, with slight but notable 1 1 1 7.7% [61 mmol/mol], P 5 0.1.000, re- versus SMBG MDI, and SMBG CSII increases from year 1 (rtCGM MDI, 1 6 1 6 spectively). No significant differences in versus SMBG MDI (Table 3). No signif- 85.7 9%; rtCGM CSII, 86.7 10%) 1 6 A1C between the rtCGM1MDI and icant differences were seen between and year 3 (rtCGM MDI, 88.0 8%; 1 1 fi 1 6 fi rtCGM1CSII groups (P 5 0.61) or rtCGM MDI and rtCGM CSII. Signi - rtCGM CSII, 87.0 8%). No signi cant SMBG1MDI and SMBG1CSII (P 5 0.69) cant improvements in time in range differences between the subgroups were were observed. and time spent in hypoglycemia were observed (Supplementary Table 2). 1 Significant reductions in A1C were observed at 3 years in the rtCGM SMBG Use 1 1 seen in the rtCGM1MDI and rtCGM MDI, rtCGM CSII, and SMBG CSII The average frequency of fingerstick 1 groups at all follow-up visits throughout groups but not the SMBG MDI group tests performed per day was lower in the 3-year study period (Fig. 1 and Table (Fig. 2). the SAIR group compared with the SMBG 2). Significant A1C reductions were seen Time in Range group (3.0 6 0.9 vs. 3.8 6 1.2, P 5 0.001). in the SMBG1CSII group only at month Improvements in time in range (70–180 It is important to note that the rtCGM 12 (P 5 0.183); no significant reductions mg/dL [3.9–10.0 mmol/L]) among SAIR devices required twice daily calibra- were seen in the SMBG1MDI group. subgroups were significantly greater tion with fingerstick testing. Within Supplementary Table 1 presents A1C than observed in the SMBG subgroups: the SAIR group, daily SMBG frequency care.diabetesjournals.org Soupalˇ and Associates 5 SMBG rtCGM rtCGM rtCGM rtCGM al 3 Table td rusDfeec 5 CI 95% Difference groups Study rtCGM was significantly lower among Dexcom G4 sensor users (n 5 32) compared with 1 1 1 1 1 1 Medtronic Enlite sensor users (n 5 14): SIv.SMBG vs. CSII D s rtCGM vs. MDI SIv.SMBG vs. CSII D s SMBG vs. MDI SIv.SMBG vs. CSII D s SMBG vs. MDI —

ewe-ru ifrne nAC vrg esrguoe n lcmcvraiiy(D t3years 3 at (SD) variability glycemic and glucose, sensor average A1C, in differences Between-group 2.7 6 0.6 vs. 3.7 6 0.8, P , 0.001 (Supplementary Table 2).

Adverse Events 1 1 1 1 1 1 MDI CSII MDI Seven severe episodes of hypoglycemia CSII MDI SI0.12 CSII were reported during the 3-year study period: two within the SMBG1CSII group, three in the SMBG1MDI group,

( 1 ( ( ( ( ( 2 2 2 2 2 one within the rtCGM CSII group 2 2 2 2 2 2 0.09 08 ( 10.8) 0.78 0.99 0.87 .)( 9.8) 0.9 .)( 8.5) .)( 1.0) .)( 9.5) 1.3) (which occurred when the participant was not wearing the sensor), and one within the rtCGM1MDI group. Three

1,%(mlml vrg esrguoe gd mo/)Gyei aiblt,m/L(mmol/L) mg/dL variability, Glycemic (mmol/L) mg/dL glucose, sensor Average (mmol/mol) % A1C, episodes of ketoacidosis occurred: one 2 2 2 2 2 2 2 2 t .50.69 0.35 to 2

2 1 ( in the SMBG CSII group, one in the .3to 1.23 .5to 1.45 .8to 1.38 2 2 42to 14.2 34to 13.4 58to 15.8 52to 15.2 .5t .9001 1.98 0.0011 0.59 to 0.35 0.52 . to 1.3 . o38 ( 3.8) to 5.7 . o6.4) to 3.8 SMBG1MDI group, and one in the rtCGM1CSII group; all cases were ad- 2 2 2 2 2 2 2 2 0.5 .30.0011 0.33 0.52 .50.0016 0.35 judicated. Four allergic reactions to sen- 5.5) .)( 3.6) 5.7) .)( 3.8) sor wear occurred but did not result in study discontinuation. No infections re- quiring assistance were reported during , , P 0.0001 0.0001 the 3-year study period. au ifrne9%CI 95% Difference value

CONCLUSIONS To our knowledge, this is the first pro- spective, real-world, 3-year study to si- ( ( 2 2 2 2 2 2 2 2 2 01)( (0.11) 2 .8 ( 1.28) 23.0 .7 ( 1.17) .2 ( 1.62) 21.1 29.2 .5 ( 0.35) .1 ( 1.51) 27.0 multaneously compare four different . 119.08 6.2 treatment strategies based on different combinations of glucose monitoring sys- tems and insulin delivery methods. As 2 2 2 2 2 reported here, use of rtCGM among 2 2 2 2 2 26 o115.84 to 42.66 34 to 33.48 2 .2to 1.82 .7to 2.37 .5to 2.35 04 o1. 0.75 14.4 to 10.44 27 o91 0.0009 9.18 to 32.76 23t 22 0.0007 12.24 to 42.3 .6to 1.86 .6t .7 ( 0.37) to 1.06

.8t 0.8) to 0.58 adults with T1D treated with MDI or CSII therapy was associated with 3 years – .60.33 6.66 2 2 2 2 2 of sustained improvements from base- .1 ( 0.51) 0.88) .8 ( 0.68) 0.7) 12.6 line in A1C, glycemic variability, and time in range, with significantly greater reduc- tions in time spent below range , , , , ( 70 mg/dL [ 3.9 mmol/L]); both P 0.0001 0.0001 au ifrne9%CI 95% Difference value time in range and time below range are now emerging as important metrics of glycemic control. Importantly, we ob- served comparable improvements in 1 1 ( ( ( 2 2 both the rtCGM CSII and rtCGM MDI 2 2 2 2 2 2 2 2 2 2 12.06 12.24 .4 ( 0.24) 0.18 .2( 4.32 .5 ( 0.25) .7 ( 0.67) .8 ( 0.68) .2 ( 0.42) .1 ( 0.01) .6( 7.56

. ( 4.5 groups, suggesting equivalent efficacy regardless of the insulin delivery method used. Although similar improvements in 2 2 2 2 2 2 2 2 2 glycemic control have been shown in 2 2 2 .9to 0.59 .9to 0.59 .2to 1.02 .8to 0.78 85 o55 0.0006 5.58 to 18.54 83 o59 0.0003 5.94 to 18.36 06 o19 0.17 1.98 to 10.62 06 o14 0.14 1.44 to 10.62 40 o12 0.02 1.26 to 14.04 .3t 0.31) to 1.03 .5t 0.32) to 0.35 . o57 0.95 5.76 to 6.3 previous RCTs (7–13), our findings dem- onstrate the long-term sustainability of 2 2 2 2 rtCGM use, its clinical benefits, and its 0.11) 0.08) 0.33) 0.07) implications regarding medication ad- herence within the context of real-world diabetes self-management. The consis-

P tently high percentage of time that par- value ticipants wore their sensors during the 3-year study period suggests that rtCGM was perceived to be a valuable tool in 6 3-Year Follow-up on COMISAIR Study Diabetes Care

Figure 2—Changes in percentage of time in range and time in hypoglycemia. SAP, sensor-augmented pump.

their self-management regimens, and Additionally, an increasing number of changes in A1C between the study sub- it may also explain the significant in- payers and regulatory agencies are rec- groups were comparable, which suggests crease in the number of daily boluses ognizing the inherent limitations of RCTs that device differences did not impact observed in the SAIR groups; no in providing real-world evidence (RWE) our findings. Additionally, with the ex- changes in daily bolusing were seen about the efficacy of medications and use ception of patients with insulin pumps in the SMBG groups. Additionally this of medical devices in clinical practice. As (CGM1CSII and SMBG1 CSII groups), all persistence in CGM use correlates with such, they are now focusing on RWE to bolusing data gathered from the other the increased number of participants inform their decisions. For example, both study groups were self-reported. Al- getting to the goal, suggesting the the U.S. Food and Drug Administration though it is possible that participants perceived value translated into im- and European Medicines Agency are may have overreported their bolusing proved clinical outcomes. asking manufacturers to provide RWE frequency, given the higher number of From a clinical perspective, the glyce- in combination with RCT findings when boluses within the rtCGM groups, which mic improvements observed among evaluating both the short- and long-term appear to correlate with better glycemic rtCGM users will likely lead to significant safety and effectiveness of new drug outcomes versus SMBG groups, we be- reductions in long-term complications and submissions, par- lieve the impact of overreporting was (21). However, our findings also have ticularly in the assessment of medical minimal. important implications for payers. As devices in real-world clinical practice Importantly, our findings demonstrate reported by Gilmer et al. (22), a 1.0% (23–26). that the use of rtCGM with MDI can reduction in A1C from 8.0 to 7.0% is The study has notable limitations. Be- be considered an equivalent but more associated with ;$820 in savings over cause this was a nonrandomized study, it is cost-effective treatment alternative to 3 years in adults with diabetes but possible that there were some unmeasured sensor-augmented insulin pumps for without heart disease and hyperten- factors that could impact our findings. many individuals with T1D. For exam- sion; the savings are even greater when For example, it is possible that the more ple, in a recent analysis of the Multiple one or both of these comorbidities are motivated study participants may have Daily Injections and Continuous Glucose present. selected to use rtCGM. Although one would Monitoring in Diabetes (DIAMOND) trial In addition to the long duration of expect motivated participants to achieve (8), Skandari and colleagues (27) found assessment, another strength is the use greater improvements than participants that among rtCGM1CSII participants, the of a real-world study design. Although who are less motivated, we observed no total per-person 28-week costs were the efficacy and clinical utility of rtCGM significant between-group differences in $8,272 vs. $5,623 among rtCGM1MDI have been demonstrated in numerous motivation. Because all subjects were will- users; the difference was primarily RCTs (7–13), they do not necessarily ing to participate in a “Dose Adjustment for attributed to CSII use. The increasing reflect the behaviors and clinical re- Normal Eating (DAFNE)-like” 4-day training focus on reducing costs while improv- sponses of participants in real life be- program, motivation likely only had a minor ing outcomes may impact reimburse- cause RCTs strictly control the setting and impact on results, if any. Moreover, if we ment decisions regarding current and delivery of interventions to minimize the had not allowed participants to choose the future sensor-augmented insulin pump effect of external factors on outcomes regimens that met their individual needs systems. (16–18). Nor do they inform us about the and preferences, we would have likely In conclusion, in individuals with T1D long-term sustainability and clinical im- seen a much higher discontinuation rate, with suboptimal glycemic control, use of pact of rtCGM use beyond the defined which would have resulted in a gradual rtCGM was superior to SMBG in reducing study durations. In our study, we allowed loss in our ability to describe differences A1C, hypoglycemia, and the other end participants to choose the insulin/ between study groups. Another potential points regardless of the insulin delivery monitoring option that met their indi- limitation is that different types of insulin method used; both methods provided vidual needs, which reflects real-life pumps and rtCGM systems were used comparable glycemic benefits. Our find- decision-making in most clinical practices. in this study. However, as reported, ings may provide guidance to clinicians care.diabetesjournals.org Soupalˇ and Associates 7

when discussing treatment/monitoring https://www.niceorguk/guidance/ng17. Accessed hypoglycaemia (IN CONTROL): a randomised, options with their participants. 2 February 2018 open-label, crossover trial. Lancet Diabetes En- 2. American Diabetes Association. 6. Glycemic tar- docrinol 2016;4:893–902 gets: Standards of Medical Care in Diabetesd2018. 14. Hirsch IB, Abelseth J, Bode BW, et al. Sensor- Diabetes Care 2018;41(Suppl. 1):S55–S64 augmented insulin pump therapy: results of the ’ Acknowledgments. The authors thank Nad a 3. Borot S, Benhamou PY, Atlan C, et al.; Societ´ e´ first randomized treat-to-target study. Diabetes ıˇ Pavl´ckova,´ Eva Kabrtov´ a,´ and Gabriela Bastrova´ francophone du diabete` (SFD); Societ´ efrançaise´ Technol Ther 2008;10:377–383 (all from 3rd Department of Internal Medicine, d’endocrinologie (SFE); Evaluation´ dans le dia- 15. Tamborlane WV, Beck RW, Bode BW, et al.; 1st Faculty of Medicine, Charles University, bete` des implants actifs Group (EVADIAC). Juvenile Diabetes Research Foundation Continu- Prague, Czech Republic) for technical assistance. Practical implementation, education and inter- ous Glucose Monitoring Study Group. Continuous Funding. This study was initiated, designed, and pretation guidelines for continuous glucose mon- glucose monitoring and intensive treatment of performed by the investigators and supported by itoring: a French position statement. Diabetes type 1 diabetes. N Engl J Med 2008;359:1464– the Agency for Healthcare Research of the Czech Metab 2018;44:61–72 1476 Republic (grant 15-26705A) and Ministry of 4. Danne T, Nimri R, Battelino T, et al. International 16. Edelman SV, Polonsky WH. Health of the Czech Republic (RVO-VFN64165). consensus on use of continuous glucose monitor- in the real world: the elusive nature of glycemic Duality of Interest. Dexcom, Inc., provided – – ˇ ing. Diabetes Care 2017;40:1631 1640 control. Diabetes Care 2017;40:1425 1432 funding for manuscript development. J.So. has 5. Fonseca VA, Grunberger G, Anhalt H, et al.; 17. van Onzenoort HA, Menger FE, Neef C, et al. received speaker honoraria and has consulted Consensus Conference Writing Committee. Con- Participation in a clinical trial enhances adherence for Abbott, Dexcom, Eli Lilly, Medtronic, Novo tinuous glucose monitoring: a consensus con- and persistence to treatment: a retrospective Nordisk, and Roche. L.P. has received speaker ference of the American Association of Clinical cohort study. Hypertension 2011;58:573–578 honoraria and has consulted for Medtronic and Endocrinologists and American College of Endo- 18. Rothwell PM. External validity of rando- Dexcom. G.G. has received speaker honoraria crinology. Endocr Pract 2016;22:1008–1021 mised controlled trials: “to whom do the results and has consulted for AstraZeneca, Boehringer 6. Peters AL, Ahmann AJ, Battelino T, et al. of this trial apply?”. Lancet 2005;365:82–93 Ingelheim, Eli Lilly, Medtronic, Novo Nordisk, and Diabetes technology-continuous subcutaneous 19. Soupalˇ J, Petruzelkovˇ a´ L, Flekacˇ M, et al. fi Sano . M.F. has received speaker honoraria and insulin infusion therapy and continuous glucose Comparison of different treatment modalities for has consulted for AstraZeneca, Medtronic, monitoring in adults: an Endocrine Society clin- type 1 diabetes, including sensor-augmented Merck Sharp & Dohme, Boehringer Ingelheim, ical practice guideline. J Clin Endocrinol Metab insulin regimens, in 52 weeks of follow-up: fi Eli Lilly, Novo Nordisk, and Sano . M.M. has 2016;101:3922–3937 a COMISAIR study. Diabetes Technol Ther received speaker honoraria for Eli Lilly and Novo – ˇ 7. Lind M, Polonsky W, Hirsch IB, et al. Continuous 2016;18:532 538 Nordisk. J.Sk. Jr. has received speaker honoraria glucose monitoring vs conventional therapy for gly- 20. World Medical Association Declaration of and has consulted for Boehringer Ingelheim, Eli cemic control in adults with type 1 diabetes treated Helsinki: recommendations guiding physicians in fi ˇ Lilly, Novo Nordisk, and Sano .J.Sk. has received with multiple daily insulin injections: the GOLD biomedical research involving human subjects. speaker honoraria and has consulted for Astra- randomized clinical trial. JAMA 2017;317:379–387 JAMA 1997;277:925–926 Zeneca, Merck Sharp & Dohme, Boehringer 8. Beck RW, Riddlesworth T, Ruedy K, et al.; 21. Nathan DM, Genuth S, Lachin J, et al.; Di- Ingelheim, Eli Lilly, Novartis, Novo Nordisk, DIAMOND Study Group. Effect of continuous abetes Control and Complications Trial Research fi and Sano . C.G.P. has received consulting fees glucose monitoring on glycemic control in adults Group. The effect of intensive treatment of from Dexcom, Diasome, Onduo, Proteus, Roche ˇ with type 1 diabetes using insulin injections: the diabetes on the development and progression Diabetes Care, and Senseonics. S.S. has received DIAMOND randomized clinical trial. JAMA 2017; of long-term complications in insulin-dependent speaker honoraria and has consulted for Astra- 317:371–378 diabetes mellitus. N Engl J Med 1993;329:977– Zeneca, Berlin-Chemie, Boehringer Ingelheim, 9. Beck RW, Riddlesworth TD, Ruedy K, et al.; 986 Novartis, Eli Lilly, and Novo Nordisk. M.P. has DIAMOND Study Group. Continuous glucose 22. Gilmer TP, O’Connor PJ, Rush WA, et al. received speaker honoraria and has consulted monitoring versus usual care in patients with Predictors of health care costs in adults with for Abbott, AstraZeneca, Boehringer Ingelheim, type 2 diabetes receiving multiple daily insulin diabetes. Diabetes Care 2005;28:59–64 Dexcom, Eli Lilly, Novartis, Novo Nordisk, Med- injections: a randomized trial. Ann Intern Med 23. Khosla S, White R, Medina J, et al. Real world fi tronic, Sano , Takeda, and Roche. No other 2017;167:365–374 evidence (RWE) - a disruptive innovation or the fl potential con icts of interest relevant to this 10. Heinemann L, Freckmann G, Ehrmann D, quiet evolution of medical evidence generation? article were reported. et al. Real-time continuous glucose monitoring F1000Res 2018;7:111 Author Contributions. J.So.ˇ was responsible ˇ ˇ in adults with type 1 diabetes and impaired 24. Food and Drug Administration (FDA). Use of for the protocol design. J.So., L.P., G.G., A.H., J.Sk. hypoglycaemia awareness or severe hypoglycae- real-world evidence to support regulatory Jr., E.H., J.Sk.,ˇ C.G.P., S.S.,ˇ and M. P. wrote and ˇ mia treated with multiple daily insulin injections decision-making for medical devices [Internet]. revised the manuscript. J.So., A.H., M.F., M.M., (HypoDE): a multicentre, randomised controlled Available from https://www.fda.gov/downloads/ O.M., T.P., and M.P. were responsible for study trial. Lancet 2018;391:1367–1377 medicaldevices/deviceregulationand-guidance/ implementation and administration. J.So.,ˇ A.H., ˇ 11. Polonsky WH, Hessler D, Ruedy KJ, Beck RW; guidancedocuments/ucm513027.pdf. Accessed M.F., C.G.P., and M.P. reviewed the data. J.So. is DIAMOND Study Group. The impact of contin- 5 April 2017; 30 March 2019 the guarantor of this work and, as such, had full uous glucose monitoring on markers of quality of 25. Resnic FS, Matheny ME. Medical devices in access to all the data in the study and takes life in adults with type 1 diabetes: further findings the real world. N Engl J Med 2018;378:595–597 responsibility for the integrity of the data and the from the DIAMOND randomized clinical trial. 26. Katkade VB, Sanders KN, Zou KH. Real world accuracy of the data analysis. Diabetes Care 2017;40:736–741 data: an opportunity to supplement existing Prior Presentation. This work will be presented 12. Pickup JC, Freeman SC, Sutton AJ. Glycaemic evidence for the use of long-established med- at the 55th Annual Meeting of the European control in type 1 diabetes during real time icines in health care decision making. J Multi- Association for the Study of Diabetes, Barcelona, continuous glucose monitoring compared with discip Healthc 2018;11:295–304 – , 16 20 September 2019. self monitoring of blood glucose: meta-analysis 27. Wan W, Skandari MR, Minc A, et al. Cost- of randomised controlled trials using individual effectiveness of initiating an insulin pump in T1D References patient data. BMJ 2011;343:d3805 adults using continuous glucose monitoring com- 1. National Institute for Health and Care Excel- 13. van Beers CA, DeVries JH, Kleijer SJ, et al. pared with multiple daily insulin injections: the lence. Type 1 diabetes in adults: diagnosis and Continuous glucose monitoring for patients with DIAMOND Randomized Trial. Med Decis Making management [Internet], 2015. Available from type 1 diabetes and impaired awareness of 2018;38:942–953