2025 Volume 48 Issue 3 Pages 234-240
We examined whether the glucose levels and awareness of individuals without diabetes changed after using a sensor-based intermittently scanned continuous glucose monitoring (isCGM) system in their daily lives. Japanese individuals without a diabetes diagnosis wore the isCGM system while maintaining a normal lifestyle during the baseline period. A certified diabetes educator coached them on how to improve their lifestyle based on information from sensor data, food journals, and body composition. The participants subsequently consumed a specific diet, exercised for 2 months, and wore a new sensor after the intervention period. A total of 36 Japanese participants were eligible for analysis in this study. The time above range and the area under the curve did not change between before and after the intervention in overall participants. The visual analogue scale scores significantly increased from before to after the intervention in the overall participants. Stratified analysis was performed by dividing the participants into 18 control participants (glycated hemoglobin level <5.7%) and 18 participants with prediabetes (glycated hemoglobin level 5.7–6.4%). The time in range and the area under the curve significantly increased and decreased after the intervention in participants with prediabetes but not in control participants. The visual analogue scale scores significantly increased from before to after the intervention in both control and prediabetes groups individually. Lifestyle modification, along with the use of an isCGM system, is highly effective at preventing type 2 diabetes mellitus, potentially reducing the individual and public health burdens of diabetes, particularly for individuals with prediabetes.
In 2019, approximately 463 million people worldwide had diabetes, and this number is expected to increase to 700 million by 2045.1) In particular, 1 in 3 people in Asia has diabetes.1) These high prevalence rates signify a risk of type 2 diabetes mellitus, for which effective interventions are needed to prevent its future development. The term prediabetes is increasingly used to describe people with impaired glucose tolerance, impaired fasting glucose or glycated hemoglobin (HbA1c) values, which range between 5.7 and 6.4% (42–50 mmol/mol) in the U.S.A.2) The American Diabetes Association has reported that prediabetes is associated with a heightened risk of cardiovascular disease.3)
In Europe, America,4–6) and Asia,7,8) lifestyle modification is preferred over treatment with metformin for preventing type 2 diabetes mellitus. The Diabetes Prevention Program (DPP) demonstrated that lifestyle intervention can reduce the incidence of type 2 diabetes mellitus by 58% in high-risk adults, and that the annual cost of type 2 diabetes mellitus4–6) is estimated to be $176 billion in direct medical expenses and $69 billion in indirect costs related to absenteeism. However, the DPP has strict goals, including a weight reduction >5%, fat intake <30%, saturated fat intake <10%, fiber intake >15 g/1000 kcal, and exercise >4 h/week,6) and involves frequent visits and coaching for several years. Thus, the program has a high dropout rate of more than 50%.9,10)
Recently, glucose management using capillary samples has been shown to be useful for controlling blood glucose levels in patients with diabetes.11) A novel sensor-based intermittently scanned continuous glucose monitoring (isCGM) system12) can provide information on sensor glucose (SG) by continuously measuring the interstitial fluid glucose concentration of a person.13) By attaching a small sensor patch to the upper arm, without requiring calibration, the system can monitor glucose concentrations for 14 d. Bolinder et al. reported that the use of an isCGM system contributed to a significant reduction in hypoglycemia in patients with type 1 diabetes mellitus.11) Moreover, the system is easy to use and can be widely used in individuals without diabetes.14,15) Nevertheless, no study has investigated the usefulness of an isCGM system for improving blood glucose control and lifestyle choices in healthy individuals.
Thus, in this study, we used the isCGM system in individuals without diabetes and aimed to examine whether SG visualization via the isCGM system can effectively change the lifestyle of Japanese individuals without diabetes and improve their blood sugar levels.
This prospective, nonblinded intervention study was performed at Kinjo Gakuin University. We recruited 48 Japanese individuals without diabetes aged >20 years between 2018 and 2020. Any potentially eligible individuals from Kinjo Gakuin University and the Nagoya City Pharmacist Association were invited to participate in this study. The exclusion criterion for this study was an HbA1c level >6.5% (>51 mmol/mol), any missing data due to stopping the program, and a disconnected isCGM.
Study ProtocolFigure 1 shows the details of the study schedule. Upon screening and enrollment, the HbA1c levels and physical characteristics of individuals were obtained and analyzed using the A1c Gear and T-SCAN PLUS (Kobe Medicare, Kobe, Japan). Participants without diabetes wore the isCGM system sensor (Freestyle Libre Pro, Abbott Diabetes Care, Witney, Oxon, U.K.) while engaging in their regular lifestyle for 14 d to obtain baseline information; the participants were blinded to all SG data. After 2 d of wearing the sensor, a 75-g oral glucose tolerance test (OGTT) was conducted. Data on meals and diet were recorded while the participants were wearing the sensors. Data from sensors, participants’ food journals, and body composition measurements were analyzed. These data were then provided to the certified diabetes educators (CDEs), who subsequently coached the participants. Briefly, the participants were advised to adopt a healthy lifestyle, which included: (1) avoiding a single-item diet and carbohydrate intake, (2) regulating their alcohol consumption, (3) eating carbohydrate intake last in a meal, and (4) exercising at 15–20 min after the meal every day. We also encouraged the participants to self-monitor their dietary habits. The participants were then placed on a specific diet and exercise regimen for 2 months. After the intervention, the participants wore the new sensor for 14 d, and another OGTT was subsequently conducted. The participants also answered a questionnaire before and after the intervention period.
isCGM, intermittently scanned continuous glucose monitoring; GMI, glucose management indicator; TIR, time in range; TBR, time below range; TAR, time above range; SG, sensor glucose; OGTT, oral glucose tolerance test.
All baseline participant data were analyzed. Stratified analysis was then performed by dividing the participants into control participants (individuals with HbA1c levels <5.7%; <42 mmol/mol) and participants with prediabetes (individuals with HbA1c levels 5.7–6.4%; 42–50 mmol/mol)2) based on calculations using the A1c Gear (Sanwa Kagaku Kenkyusho, Nagoya, Japan).
Data CollectionInterfluid SG values were obtained every 15 min using an isCGM system sensor. Information on body composition (weight, body mass index [BMI], body fat mass, percent body fat, total body water [TBW], abdominal obesity), HbA1c levels, SG values, time in range (TIR), time below range (TBR), time above range (TAR), and glucose management indicator (GMI) was collected before (baseline) and after the intervention. The area under the curve (AUC) of the OGTT results using the interfluid SG values was calculated based on the trapezoidal rule.
Questionnaire SurveyQuestionnaire surveys were conducted regarding lifestyle, which included questions about (1) the amount in a meal, (2) the amount of carbohydrates consumed, (3) the frequency of having only 1 meal a day, (4) the frequency of overeating carbohydrates, (5) turn of eating carbohydrates, (6) the frequency of eating snacks, (7) the amount of exercise, (8) the frequency of exercise, (9) the amount of liquid consumed, and (10) the frequency of drinking. The visual analogue scale (VAS) was used, with a 100-mm-long line indicating positive (questions 5, 7, and 8) or negative scores. The 100-mm point of the line reflects “many” in each question. We evaluated each score and the total score (0–1000) before and after the intervention.
Statistical AnalysesStudent’s t-test was used to compare age between groups. Chi-squared tests were used for gender, alcohol consumption habits, smoking habits, family history of diabetes, and hypertension/hyperlipidemia between groups. Paired t-tests were used to analyze differences in body composition, GMI, TIR, TBR, TAR, SG and AUC before and after the intervention period. The Wilcoxon signed-rank sum test was used to analyze the VAS scores. A p-value <0.05 was considered to indicate statistical significance.
Ethical ConsiderationsThis study was conducted in accordance with the principles of the Declaration of Helsinki. This study was approved by the Ethics Committee of Kinjo Gakuin University and was performed in accordance with the Good Clinical Practice Guidelines (H18016). Written informed consent was obtained from all participants.
We recruited 48 Japanese individuals. Eight individuals were missing data due to stopping the program early. Four individuals could not be analyzed because the isCGMs were removed. Thus, the dropout rate for this program was 25.0% (12/48 individuals). A total of 36 Japanese participants were included in the analysis for this study. Table 1 shows the baseline characteristics of the participants. Overall, the 36 participants (21 males and 15 females) had a mean age of 50.7 ± 11.0 years. The rates of alcohol consumption habits, smoking habits, family history of diabetes, and hypertension/hyperlipidemia were 38.9% (14/36 individuals), 33.3 (12/36 individuals), 11.1% (4/36 individuals), and 22.2% (8/36 individuals), respectively. The participants were subsequently divided into 18 control participants and 18 participants with prediabetes. The 18 control participants (8 males and 10 females) had a mean age of 48.4 ± 10.4 years, whereas the 18 participants with prediabetes (13 males and 5 females) had a mean age of 53.1 ± 11.8 years. No significant differences were observed in gender, age, alcohol consumption, smoking habits or family history of diabetes between the control and prediabetes groups.
Overall | Control | Prediabetes | p-Value | |
---|---|---|---|---|
Participants | 36 | 18 | 18 | |
Gender | ||||
Male | 21 (58.3%) | 8 (44.4%) | 13 (72.2%) | 0.176 |
Female | 15 (41.7%) | 10 (55.6%) | 5 (27.8%) | |
Age (years) | 50.7 ± 11.0 | 48.4 ± 10.4 | 53.1 ± 11.8 | 0.216 |
Alcohol consumption habits | 14 (38.9%) | 6 (33.3%) | 8 (44.4%) | 0.733 |
Smoking habits | 12 (33.3%) | 6 (33.3%) | 6 (33.3%) | 1.000 |
Family history of diabetes | 4 (11.1%) | 1 (5.56%) | 3 (16.7%) | 0.603 |
Hypertension/hyperlipidemia | 8 (22.2%) | 4 (22.2%) | 4 (22.2%) | 1.000 |
The table shows the number of participants, gender, alcohol consumption habits, smoking habits, family history of diabetes, hypertension or hyperlipidemia, and average ± standard deviation of age for the overall participants, control participants, and participants with prediabetes. The p-values were used to compare age (Student’s t-test) and gender, alcohol consumption habits, smoking habits, family history of diabetes, and hypertension/hyperlipidemia (chi-squared test) between control participants and participants with prediabetes.
Table 2 shows the physical characteristics of the participants. Overall, participants had HbA1c levels of 5.7 ± 0.4% (39 ± 4.4 mmol/mol) at baseline. After the intervention, the TAR significantly decreased for overall participants. No significant changes in HbA1c, body composition (weight, BMI, body fat mass, percent body fat, TBW, and abdominal obesity), GMI, TIR, TBR, and SG values after 2 months of intervention were observed in overall participants.
Before | After | p-Value | |
---|---|---|---|
(A) Overall | |||
HbA1c (%) | 5.7 ± 0.4 | 5.7 ± 0.4 | 0.939 |
Weight (kg) | 63.9 ± 11.2 | 63.6 ± 10.8 | 0.176 |
BMI (kg/m2) | 23.1 ± 3.3 | 23.0 ± 3.2 | 0.156 |
Body fat mass (kg) | 14.8 ± 4.4 | 14.4 ± 4.9 | 0.304 |
Percent body fat (%) | 23.0 ± 4.5 | 23.0 ± 5.1 | 0.967 |
TBW (L) | 35.3 ± 6.2 | 34.9 ± 5.8 | 0.320 |
Abdominal obesity (%) | 0.4 ± 0.0 | 0.4 ± 0.0 | 0.849 |
GMI (%) | 5.2 ± 0.5 | 5.1 ± 0.4 | 0.246 |
TIR (%) | 78.7 ± 16.0 | 81.6 ± 13.1 | 0.234 |
TBR (%) | 7.7 ± 13.0 | 5.7 ± 9.9 | 0.360 |
TAR (%) | 8.5 ± 6.3 | 6.9 ± 6.1 | 0.020 |
SG values (mg/dL) | 101.2 ± 13.6 | 100.3 ± 10.9 | 0.577 |
(B) Control | |||
HbA1c (%) | 5.4 ± 0.2 | 5.5 ± 0.3 | 0.100 |
Weight (kg) | 60.2 ± 11.9 | 60.2 ± 11.9 | 0.946 |
BMI (kg/m2) | 21.8 ± 2.9 | 21.4 ± 2.9 | 0.855 |
Body fat mass (kg) | 13.1 ± 3.5 | 13.2 ± 3.7 | 0.774 |
Percent body fat (%) | 22.0 ± 4.0 | 22.1 ± 4.5 | 0.743 |
TBW (L) | 33.9 ± 7.1 | 33.8 ± 7.1 | 0.818 |
Abdominal obesity (%) | 0.4 ± 0.0 | 0.4 ± 0.0 | 0.923 |
GMI (%) | 4.9 ± 0.3 | 4.9 ± 0.3 | 0.374 |
TIR (%) | 81.3 ± 17.8 | 83.5 ± 15.1 | 0.989 |
TBR (%) | 8.49 ± 13.7 | 7.60 ± 12.5 | 0.821 |
TAR (%) | 4.9 ± 2.9 | 3.7 ± 2.1 | 0.112 |
SG values (mg/dL) | 95.1 ± 9.4 | 95.1 ± 7.1 | 1.000 |
(C) Prediabetes | |||
HbA1c (%) | 5.9 ± 0.3 | 5.9 ± 0.3 | 0.180 |
Weight (kg) | 67.6 ± 9.4 | 66.9 ± 8.6 | 0.105 |
BMI (kg/m2) | 24.4 ± 3.3 | 24.2 ± 3.1 | 0.121 |
Body fat mass (kg) | 16.4 ± 4.7 | 15.7 ± 5.7 | 0.233 |
Percent body fat (%) | 24.1 ± 5.0 | 24.0 ± 5.5 | 0.832 |
TBW (L) | 36.8 ± 4.9 | 36.0 ± 4.0 | 0.341 |
Abdominal obesity (%) | 0.4 ± 0.0 | 0.4 ± 0.0 | 0.783 |
GMI (%) | 5.4 ± 0.5 | 5.3 ± 0.4 | 0.477 |
TIR (%) | 76.2 ± 13.9 | 81.9 ± 11.1 | 0.029 |
TBR (%) | 7.0 ± 12.7 | 3.8 ± 6.3 | 0.159 |
TAR (%) | 12.1 ± 6.9 | 10.0 ± 7.1 | 0.087 |
SG values (mg/dL) | 107.3 ± 14.6 | 105.6 ± 11.6 | 0.474 |
The table shows the mean ± standard deviation of HbA1c, weight, BMI, body fat mass, percent body fat, TBW, abdominal obesity, GMI, TIR, TBR, TAR, and SG values over 2 months before and after intervention in overall participants (n = 36), control (n = 18), and prediabetes (n = 18) participants. p-Values were compared before and after the intervention (paired t-test). BMI, body mass index; TBW, total body water; GMI, glucose management indicator; TIR, time in range; TBR, time below range; TAR, time above range; SG, sensor glucose.
The control and prediabetes participants had HbA1c levels of 5.4 ± 0.2% (39 ± 2.2 mmol/mol) and 5.9 ± 0.3% (44 ± 3.3 mmol/mol), respectively, at baseline. The TIR significantly increased after the intervention for participants with prediabetes (p = 0.029) but not for control participants. No significant differences in HbA1c, body composition, GMI, TAR, TBR, and SG values before and after 2 months of intervention were observed in the control or prediabetes participants.
SG Values and AUCs of the OGTT before and after the InterventionFigure 2 shows the isCGM system results of the 75 g OGTT before and after the intervention. SG values at 30 and 45 min after the OGTT were significantly lower in the postintervention period than before the intervention in overall participants (Fig. 2A; p < 0.05). The AUC did not change after the intervention for overall participants (Fig. 2B).
(A, C, E) Interfluid sensor glucose values and (B, D, F) AUC for the 75 g OGTT before and after the intervention using the isCGM sensor. The AUC was calculated based on the trapezoidal rule. The values indicate the means ± standard deviations. (A, B) Overall participants (n = 36). (C, D) Control participants (n = 18). (E, F) Participants with prediabetes (n = 18). *p < 0.05, **p < 0.01, before vs. after (paired t-test). AUC, area under the curve; OGTT, oral glucose tolerance test.
SG values 15–90 min after the OGTT were significantly lower in the postintervention period than before the intervention in participants with prediabetes (Fig. 2E; 30 and 45 min, p < 0.01; 15, 60, 75, and 90 min, p < 0.05) but not in control participants (Fig. 2C). The AUC also significantly decreased after the intervention for participants with prediabetes (Fig. 2F; p < 0.001) but not for control participants (Fig. 2D).
Changes in the VAS Score for Consciousness before and after InterventionVAS scores before and after the intervention were significantly greater in the overall participants (489 vs. 566, p < 0.01), control participants (492 vs. 566, p < 0.01), and participants with prediabetes (491 vs. 570, p < 0.01). Table 3 shows the significant improvements in overall participants for each VAS score among the following items after the intervention: (2) the amount of carbohydrates consumed, (5) turn of eating carbohydrates, (7) the amount of exercise, (8) the frequency of exercise, and (10) the frequency of drinking. (2) The amount of carbohydrates consumed, (3) the frequency of having only 1 meal a day, (5) turn of eating carbohydrates, and (10) the frequency of drinking were improved in control participants. Conversely, improvements in (5) turn of eating carbohydrates, (7) the amount of exercise, and (8) the frequency of exercise were observed in the prediabetic participants.
No. | (A) Overall | (B) Control | (C) Prediabetes | ||||||
---|---|---|---|---|---|---|---|---|---|
(n = 36) | (n = 18) | (n = 18) | |||||||
Before | After | p-Value | Before | After | p-Value | Before | After | p-Value | |
(1) | 55.0 (49.0–73.0) | 53.0 (47.0–67.0) | 0.374 | 54.0 (48.5–62.5) | 53.0 (45.8–59.9) | 0.358 | 63.5 (50.0–75.4) | 56.0 (48.0–66.9) | 0.508 |
(2) | 66.0 (49.0–77.0) | 52.0 (37.0–62.0) | 0.002 | 71.8 (53.0–82.0) | 52.0 (37.0–64.3) | <0.01 | 65.5 (44.3–71.3) | 51.5 (34.8–60.5) | 0.058 |
(3) | 55.0 (28.0–73.0) | 42.5 (26.0–64.0) | 0.114 | 56.0 (46.8–76.0) | 45.0 (34.5–56.0) | 0.039 | 45.0 (24.3–66.8) | 39.5 (13.3–72.8) | 0.586 |
(4) | 52.0 (28.0–65.0) | 42.5 (29.0–54.0) | 0.114 | 56.0 (27.3–75.5) | 43.3 (36.0–56.5) | 0.367 | 50.0 (30.3–59.3) | 31.0 (14.8–52.8) | 0.065 |
(5) | 60.0 (48.0–78.0) | 80.0 (53.0–87.0) | 0.012 | 72.8 (49.9–79.0) | 80.5 (47.5–84.3) | 0.030 | 52.0 (47.0–76.5) | 81.5 (70.5–86.8) | 0.015 |
(6) | 30.0 (15.0–73.0) | 32.0 (11.0–52.0) | 0.124 | 27.0 (15.0–51.8) | 36.3 (11.8-51.3) | 0.345 | 55.0 (15.0–82.1) | 29.0 (10.8–64.8) | 0.069 |
(7) | 18.0 (9.0–40.0) | 37.0 (15.0–54.0) | 0.002 | 24.5 (13.0–46.0) | 28.8 (14.3-57.3) | 0.140 | 17.0 (5.0–37.6) | 44.0 (24.1–62.8) | 0.001 |
(8) | 18.0 (10.0–39.0) | 30.0 (12.0–56.0) | 0.002 | 23.0 (11.5–38.3) | 28.5 (11.8-54.8) | 0.325 | 17.0 (5.0–37.6) | 42.0 (24.1–62.8) | 0.003 |
(9) | 28.0 (4.0–59.0) | 17.0 (2.0–58.0) | 0.087 | 30.0 (7.0–61.0) | 16.0 (2.0-69.1) | 0.127 | 27.0 (4.0–71.8) | 17.5 (2.6–57.8) | 0.477 |
(10) | 30.0 (5.0–75.0) | 18.0 (2.0–76.0) | 0.010 | 30.0 (6.0–74.5) | 17.0 (2.8-76.0) | 0.011 | 33.5 (2.8–81.8) | 18.5 (6.5–19.6) | 0.069 |
The table shows the median (interquartile range). No. represents the question number on the questionnaire about (1) the amount in a meal, (2) the amount of carbohydrates consumed, (3) the frequency of only 1 meal a day, (4) the frequency of overeating carbohydrates, (5) turn of eating carbohydrates, (6) the frequency of eating snacks, (7) the amount of exercise, (8) the frequency of exercise, (9) the amount of liquid consumed, and (10) the frequency of drinking. The median scores before and after the intervention were reported for overall participants (n = 36), control participants (n = 18), and participants with prediabetes (n = 18). p-Values were compared before and after the intervention.
This is the first study to demonstrate that lifestyle intervention, along with the use of an isCGM system in healthy patients, particularly those with prediabetes, promotes patients’ glycemic management.
The previous DPP had a high dropout rate of more than 50%.9,10) The National DPP of the U.K. followed up with the participants through 44 weeks, with a dropout rate of 68.1%.9) Previous reports from the DPP have shown an overweight prevalence >5.6% (41 mmol/mol) (BMI >25 kg/m2) among patients. The goal was to achieve a minimum of 7% weight loss or weight maintenance and a minimum of 150 min of weekly physical activity within the first 24 weeks. In the present study, the dropout rate was low at 25.0% in the 2-month program. Despite the lack of an exact comparison regarding the dropout rate due to the different study periods, participants in this study might have been proactive based on blood glucose fluctuations, without setting strict goals. This is also supported by the change in awareness indicated by the questionnaire survey. Our study suggests that visualizing SG using the isCGM system may effectively change lifestyle factors in individuals without diabetes. On the contrary, the dropout rate might be slightly lower because of several factors: the participants were affiliated with specific organizations, and the isCGM was provided free of charge. Some individuals may not like wearing the isCGM, considering it less invasive than SMBG but still a sophisticated instrument.
The HbA1c level and body composition of participants did not significantly change after wearing the isCGM system, which may be attributable to the intervention and education periods being too short to alter the physical characteristics. The number of instructional sessions should have been increased, such as DPPs,6,7) and the compliance rate should have been measured. Moreover, the average BMI was <25 kg/m2 in participants with prediabetes before the intervention. TAR during the 2-week period improved at the second measurement in the overall participants. On the other hand, TIR improved only in participants with prediabetes. This finding suggests that the intervention decreased many high- and low-blood glucose spikes. In particular, our results showed that the 1-h SG values and AUC from the OGTT results improved from the first to the second measurement in participants with prediabetes. Cross-sectional studies have indicated that individuals with increased 1-h plasma glucose levels have an increased risk of developing metabolic syndrome and cardiovascular diseases.16–18) Thus, our program may help prevent future cardiovascular diseases. The usefulness of the isCGM could not be fully confirmed by comparing a group instructed with the isCGM to a group instructed without it. Therefore, further studies are needed.
The participants exhibited changes in awareness while wearing the isCGM system, suggesting that the behavioural changes observed were in response to visualizing SG without a diagnosis of diabetes; the awareness of participants changed in response to wearing the isCGM system.10) In particular, CDEs’ instruction to overall participants about carbohydrates, exercise, and drinking was effective. Instructions on carbohydrates, the frequency of only 1 meal a day, and drinking were effective for the control group. Moreover, instructions on carbohydrates and exercise were effective for participants with prediabetes. The visualization of SGs via the isCGM system and instruction from CDEs could be used to facilitate behavioural changes. Instruction on lifestyle associated with SG fluctuations may help change attitudes and prevent the future risk of diabetes onset. However, the survey should have been based on a reliable questionnaire to clearly assess the usefulness of the intervention by increasing individual awareness.
In this study, 22.2% of the participants had hypertension or hyperlipidemia; however, no significant difference was observed in these parameters between the control and prediabetes participants. In addition, the participants were not considered to have any medical conditions such as gastrointestinal disorders, gastrointestinal surgery or inflammatory diseases.
Our study showed that lifestyle modification, along with the use of an isCGM system, is a highly effective means of facilitating behavioural changes to prevent type 2 diabetes mellitus, particularly in individuals with prediabetes. Hence, using the isCGM system may help delay or prevent the development of complications, thereby substantially reducing the individual and public health burdens of diabetes. However, further long-term studies are needed to confirm these findings.
This study was supported by a Grant-in-Aid for Early Career Scientists (No. 18K17338) from Japan’s Ministry of Education, Culture, Sports, Science and Technology (MEXT). We would like to thank Mr. Noda, Mr. Suwa, and the staff members of Kinjo Gakuin University who were involved in this study.
A.T.-G., K.A., R.S., and M.Y. contributed to the conception and design of the research. A.T.-G., S.F., and M.Y. contributed to the data collection. A.T.-G. contributed to the analysis, interpretation of data, and the writing of the manuscript.
R.S. is affiliated with a department endowed by Medtronic. The other authors declare no conflict of interest.