Original Article HOW PATIENTS WITH TYPE 1 DIABETES TRANSLATE CONTINUOUS GLUCOSE MONITORING DATA INTO DIABETES MANAGEMENT DECISIONS Jeremy Pettus, MD1; David A. Price, MD, FACE2; Steven V. Edelman, MD1 ABSTRACT Objective: To understand how patients use continuous glucose monitoring (CGM) data in their diabetes management. Methods: We surveyed patients who regularly used CGM (>6 days per week), using 70 questions, many scenario-based. The survey had 6 sections: patient characteristics, general CGM use, hypoglycemia prevention and management, hyperglycemia prevention and management, insulin dosing adjustments (both for incidental hyperglycemia not at meals and at mealtimes), and real-time use versus retrospective analysis. Results: The survey was completed by 222 patients with type 1 diabetes. In response to a glucose of 220 mg/ dL, the average correction dose adjustment based on rate of change arrows varied dramatically. Specifically, when the CGM device showed 2 arrows up (glucose increasing >3 mg/dL/minute), respondents stated they would increase their correction bolus, on average, by 140% (range, 0 to 600%). Conversely, 2 arrows down (glucose decreasing >3 mg/dL/minute) caused respondents to reduce their dose by 42%, with 24% omitting their dose entirely. Furthermore, 59% of respondents stated they would delay a meal in response to rapidly rising glucose, whereas 60% would wait until after a meal to bolus in response to falling
Submitted for publication October 28, 2014 Accepted for publication January 29, 2015 From the 1University of California San Diego, San Diego, California, and 2Dexcom, Inc, San Diego, California. Address correspondence to Dr. Jeremy Pettus, University of California San Diego, 4916 Bayard Street, San Diego, CA 92109. E-mail: [email protected]. Published as a Rapid Electronic Article in Press at http://www.endocrine practice.org on February 25, 2015. DOI: 10.4158/EP14520.OR To purchase reprints of this article, please visit: www.aace.com/reprints. Copyright © 2015 AACE.
glucose levels. With a glucose value of 120 mg/dL and a falling glucose trend, 70% of respondents would prophylactically consume carbohydrates to avoid hypoglycemia. Conclusion: CGM users utilize CGM data to alter multiple aspects of their diabetes care, including insulin dose timing, dose adjustments, and in hypoglycemia prevention. The insulin adjustments are much larger than common recommendations. Additional studies are needed to determine appropriate insulin adjustments based on glucose trend data. (Endocr Pract. 2015;21:613-620) Abbreviations: A1c = hemoglobin A1c; CGM = continuous glucose monitoring; ROC = rate of change; SMBG = selfmonitored blood glucose INTRODUCTION As a result of studies showing improvements in glycemic control in children and adults who regularly use continuous glucose monitoring (CGM) (1-4), the use of realtime CGM is becoming accepted as part of the standard of care in the treatment of patients with type 1 diabetes (5,6). However, the mechanism of action for glycemic benefit from CGM use differs from pharmacologic interventions. The observed benefits are not a direct result of wearing the device but rather the result of behavioral and management changes enabled by the information provided by the CGM devices to the users. The specific interventions patients make in their diabetes management in response to the information provided by CGM devices has not been elucidated. Clinical studies provide respondents with instruction on how to operate their CGM devices; however, few studies provide specific instructions to subjects on how to respond to the real-time glucose information including trend arrows. Exceptions were the DirectNet (7) and Juvenile Diabetes Research Foundation (1) CGM trials. In these studies, subjects were instructed to increase their insulin dose by 10% in response to a rising glucose (1 to
ENDOCRINE PRACTICE Vol 21 No. 6 June 2015 613
614 Using CGM to Make Treatment Decisions, Endocr Pract. 2015;21(No. 6)
2 mg/dL/minute rise) and 20% for more rapidly rising glucose (>2 mg/dL/minute rise), with identical reductions in their dose for down arrows. Although these recommendations generally resulted in improved clinical outcomes, it remains unknown if the recommendations were appropriate, or if patients are making similar adjustments in the real-world setting. Additionally, the recommendations used in these trials have not transitioned into the clinic setting in the form of official guidelines, leaving clinicians without guidance to provide their patients using CGM and often resulting in patients making independent adjustments. Because there are no previous reports that we are aware of that describe in detail how CGM users utilize their devices, a survey was conducted to assess how patients are using CGM and responding to their CGM information in a real-world setting. This report describes and discusses the findings of the survey. METHODS The survey comprised 6 sections: (1) patient characteristics; (2) general CGM use; (3) hypoglycemia prevention and management; (4) hyperglycemia prevention and management; (5) insulin dosing adjustments (both for incidental hyperglycemia not at meals and at mealtimes); and (6) real-time use versus retrospective analysis. In order to contextualize the information, many of the survey questions were framed as clinical scenarios that would be commonly experienced by patients either on multiple daily injections or using an insulin pump. The questions were beta-tested in 20 experienced CGM users and refined repeatedly to assure that the questions were well understood, clear, and unambiguous. Table 1 provides examples of the scenario-based questions. An institutional review board waiver was obtained, and clinical endocrinologists or educators that actively prescribe CGM from across the U.S. were asked to recruit patients. They identified regular Dexcom CGM (Dexcom Inc, San Diego, CA) users (>6 days a week on average) from their practice and provided them a web link to the survey. The on-line survey was available from May 28 to August 26, 2013, using SurveyGizmo (Boulder CO), and included 70 multiple-choice questions. Based on beta testing, it was estimated to take 20 to 30 minutes to complete the survey. Respondents chose from one of three different surveys in which the questions were identical but the order of the sections varied and were provided a $30 gift card for completing the survey. Statistical Methods The survey was descriptive, and no hypothesis testing was performed and no comparative analyses were made. Categorical variables are summarized using counts and percentages. Summary statistics for continuous vari-
ables are summarized using mean and standard deviation. Histograms and other graphical displays are used to illustrate the distribution of the survey responses. SAS software, version 9.3 or later, was used to conduct data conversion and analysis. RESULTS Patient Characteristics The respondents included 222 patients with type 1 diabetes from 22 states across the United States. The mean age of respondents was 46 ± 14 years, the duration of diabetes 22 ± 14 years, 52% were male, and the self-reported hemoglobin A1c (A1c) was 6.9 ± 0.8%. For their method of insulin delivery, 75% used an insulin pump and 25% used multiple daily injections. Education levels varied: 0.5% did not graduate high school, 9% were high school graduates or attended technical or trade school after high school, 14% attended college but did not graduate, 42% were college graduates, and 23% received an advanced degree. Self-perceived math skills were probed, with 6% stating they were below average (compared to most people), 23% believed their math skills were average, and 68% reported their math skills were above or well above average. Participants were not queried about their glycemic awareness status. General CGM Use Seventy-five percent of the respondents had used their CGM device for greater than 1 year. Sensor duration also varied: 43% reported typically wearing sensors for the recommended 7 days (39%) or less (4%), and 57% typically wear sensors for greater than 7 days, including 32% reporting wearing sensors for more than 11 days before changing. Sensors were most commonly worn on the abdomen, followed by the back of the upper arm and upper thigh. The majority of respondents (52%) reported starting CGM because of problems with both high and low glucose, 22% mainly because of problematic hypoglycemia, and 10% because of problematic hyperglycemia. Hypoglycemia Prevention and Management Almost all respondents used a customized low-glucose alert (set between 60 and 100 mg/dL), with only 1% of respondents relying only on the fixed low alarm set at 55 mg/dL that cannot be turned off on Dexcom CGM receivers. Forty-four percent of respondents stated that they would feel symptomatic from hypoglycemia prior to being alerted via their low alarm. Conversely, 33% felt that they were alerted by their CGM prior to symptoms being present. Since starting CGM, 78% reported that the frequency and severity of hypoglycemia had decreased. The majority of respondents (70%) reported waking up at night at least once per week in response to their lowglucose alert, with 72% stating that they rarely or never
Using CGM to Make Treatment Decisions, Endocr Pract. 2015;21(No. 6) 615
Table 1 Sample Survey Questions You are awoken at night to a low-glucose alert. You look at your CGM and see that you are at 70 mg/dL with an angled down arrow. You react by (select all that apply) (choices): I do not use a low-glucose alert setting; I confirm the glucose with a blood glucose test on my meter most of the time; I treat my glucose with food or drink, without performing a blood glucose test on my meter most of the time; I wait until the alert buzzes again before I react most of the time; I turn the alert off for the night most of the time. Imagine it has been 4 hours since your last dose of meal insulin and your CGM receiver shows a value of 220 mg/ dL (matching your fingerstick BG of 220 mg/dL) with trend arrow and trend graph flat (straight across). If you were NOT planning on eating or exercising, what dose of insulin would you give yourself to bring your glucose down to around 120 mg/dL? (choose the closest value). Imagine it has been 4 hours since your last dose of meal insulin and your CGM receiver shows a value of 220 mg/dL (matching your fingerstick BG of 220 mg/dL) with 1 arrow (trend arrow) straight up. If you were NOT planning on eating or exercising, how many additional units would you give, compared to when your trend arrow was flat to bring your glucose down to around 120 mg/dL? (choose the closest value). Assume it is lunch time and your glucose is 110 mg/dL. You have 2 arrows (trend arrows) pointing straight up. You are eating a meal with 50 grams of carbohydrates and your usual fat and protein. How much would you increase your meal insulin, compared to if your trend arrow was flat? (choose the closest value). If my glucose was 120 mg/dL, the earliest I usually take immediate action to avoid hypoglycemia if my trend arrow is (Choices): arrow angled down; 1 arrow straight down; 2 arrows straight down; I don’t ever respond to the trend arrows; I don’t respond to the trend arrows at a glucose of 120 mg/dL. Since starting CGM, the blood glucose values I am targeting, during the day (compared to before I was on CGM) are now (Choices): higher; lower; have not changed; I’m not sure. Abbreviations: BG = blood glucose; CGM = continuous glucose monitoring. slept through their alert and 19% stating that their spouse/ partner was typically alerted first. Half the respondents stated they would confirm the CGM low glucose with a fingerstick blood glucose when alerted to a low at night, whereas half stated they would treat the low glucose without confirming. Forty-two percent (42%) of respondents stated that their CGM device alerted somebody around them to respond to their hypoglycemia alarm when they themselves were unable to respond at least one time in the last 6 months. To prevent hypoglycemia, 70% would prophylactically consume carbohydrates in response to a displayed glucose of 120 mg/dL with a decreasing trend (angled or downward rate of change [ROC] arrow). Hyperglycemia Prevention and Management Since starting CGM, 59% of respondents stated that their A1c had decreased by at least 0.5%, whereas 4.5% stated that it had increased. The remainder indicated no change or they were unsure. Almost all respondents (98%) used high-glucose alerts. Among 70% of respondents, the high alert setting was 200 mg/dL or less. The majority of respondents (66%) stated that they woke up at least once per week at night in response to their hyperglycemia alert. When alert-
ed at night to a high glucose, the majority (79%) stated they would respond by taking an extra injection or bolus (if there was 1 trend arrow up) on the majority of occasions, 9% stated they would occasionally bolus, and 12% stated they would not respond. When alerted to a nocturnal high glucose, 66% would usually or always confirm with a fingerstick blood glucose measurement, 23% would confirm less than half the time, and 11% would never do a confirmatory fingerstick. Sixty-five percent of respondents reported that since starting CGM, the number of injections or boluses of shortacting insulin they took per day had increased, 15% felt the number had decreased, and the remainder indicated no change. However, respondents were not asked if their total daily dose of insulin had changed. With a glucose of 130 mg/dL and a rapid rise (2 arrows up), 22% stated they would take additional insulin even if it was only 1 hour since their previous insulin dose. Others (35%) would dose only if it was more than 2 hours since their dose, and 15% would wait until the glucose value was high. Only 19% would not dose based on high rates of glucose rise. Respondents were not asked if they frequently had low glucose within a few hours of a correction bolus. Since starting CGM, many respondents adjusted their glucose targets. During the day, 57% lowered, 35% did
616 Using CGM to Make Treatment Decisions, Endocr Pract. 2015;21(No. 6)
not change, and 4% raised their glucose targets. During the night, 38% lowered, 42% did not change, and 14% raised their glucose targets. The remainder did not recall their past targets. Insulin Dosing Adjustments for Incidental Hyperglycemia, Unrelated to a Meal Respondents were provided a scenario in which it had been 4 hours since taking any insulin or eating and their CGM device showed a glucose value of 220 mg/dL (confirmed by self-monitored blood glucose [SMBG]) with a flat ROC arrow. They were asked how much insulin they would take to correct their glucose to 120 mg/dL if they were not planning on eating or exercising. Then, respondents were asked the same question with varying ROC arrows (1 arrow up, 2 arrows up, 1 arrow down, 2 arrows down) and asked how the insulin dose would change compared to when the trend arrow was flat. The percentages of increase or decrease of insulin dose were calculated from their typical correctional dose at the flat CGM arrow and included in Figure 1. These results were also averaged and are summarized in Table 2.
With a flat arrow indicating stable glucose values, the average correction dose was 2.8 units, indicating an average correctional factor of 35.7 mg/dL/unit. In response to 1 or 2 arrows up, respondents stated they would increase their typical correctional dose by 111% and 140%, respectively. However, the range of responses varied from 0 to 600%. In response to 1 or 2 arrows down, respondents stated they would decrease their average correctional dose by 40 and 42%, respectively. The responses varied from 0 to 100%, indicating that some patients would not adjust their bolus and others would omit it completely. Of note, in the more extreme situation of 2 arrows down, 24% of patients would not bolus at all in response to a glucose of 220 mg/dL. Although the range of responses varied significantly, the average correction dose (adjusted for the average correction dose of 2.8 units at stable glucose) ranged from 1.6 to 6.7 units, depending on the ROC arrow. Insulin Dosing Adjustments at Mealtime Respondents were provided a scenario in which their CGM showed a glucose value of 110 mg/dL and they were
Fig. 1. Impact of the direction and rate of glucose change on a correctional insulin dose at hyperglycemia (220 mg/dL) unrelated to a meal or prior insulin dose. Left panels indicate the percentage of respondents who increased their insulin dosages 0 to 400% when 1 UP arrow (A) or 2 UP arrows (C) were displayed. Right panels indicate the percentage of respondents who decreased their insulin dosages 0 to 100% when 1 DOWN arrow (B) or 2 DOWN arrows (D) were displayed.
Using CGM to Make Treatment Decisions, Endocr Pract. 2015;21(No. 6) 617
Table 2 Comparison of Recommended Insulin Dosage Adjustments vs. Participant Self-Reported Correctional Insulin Dose Adjustments in Response to Rate of Change Arrows Recommended insulin dosage adjustmenta,b
Participants’ reported insulin dosage adjustment
Constant: glucose is steady – not increasing/decreasing more than 1 md/dL per minute
No change
NA
Slowly rising: glucose is rising 1-2 mg/dL per minute
10% higher
Rising: glucose is rising 2-3 mg/dL per minute
20% higher
111% higher
Rapidly rising: glucose is rising more than 3 mg/dL per minute
20% higher
140% higher
Indication
10% lower
Falling: glucose is falling 2-3 mg/dL per minute
20% lower
40% lower
Rapidly falling: glucose is falling more than 3 mg/dL per minute
20% lower
42% lower
Slowly falling: glucose is falling 1-2 mg/ dL per minute
Abbreviation: NA = not applicable. aDiabetes Research In Children Network (DirecNet) Study Group. Pediatr Diabetes. 2008;9:142-147. bThe Juvenile Diabetes Research Foundation Continuous Glucose Monitoring Study Group. Diabetes Technol Ther. 2008;10:310-321.
planning to eat 50 grams of carbohydrates. They were asked how much insulin they would take for that meal when the trend arrow was flat. In the subsequent questions, respondents were asked how the meal dose would change if the glucose and meal were the same but there were 2 trend arrows up or 2 arrows down. Based on these data, a percent increase or decrease was calculated from their typical meal insulin dose. In response to a flat arrow, the average bolus for a 50-gram carbohydrate meal was 4.7 units, indicating an average insulin to carbohydrate ratio of 1:10.6. As shown in Figure 2, in response to 2 arrows up, respondents would increase their typical mealtime dose by an average 81%. Conversely, in response to 2 arrows down, respondents would decrease their dose by an average 46%. In addition to increasing their bolus at mealtime in response to 2 arrows up, 59% of respondents stated
that they would allow more time between their insulin bolus and starting their meal. Conversely, in response to 2 arrows down, 60% stated that they would take their insulin bolus after their meal. Therefore, both the insulin timing and the insulin dose amount were affected by ROC or arrows. In response to a question that asked how they would determine the actual changes in their insulin dose based on ROC arrows, most respondents (61%) stated they would usually estimate the dose adjustment based mainly on their past experiences; only 14% stated they would use a certain percent increase or decrease in their insulin dose based on the ROC arrow (e.g., 20% increase for 1 arrow up), and 14% stated they would take a fixed dose of insulin based on the ROC arrow (e.g., extra 2 units for 1 arrow up). Only 11% of respondents stated they did not use the ROC arrows to adjust their dose.
618 Using CGM to Make Treatment Decisions, Endocr Pract. 2015;21(No. 6)
Fig. 2. Impact of the direction and rate of glucose change on a mealtime insulin dose at euglycemia (110 mg/dL). Left panel (A) indicates the percentage of respondents who increased their insulin dosages 0 to 400% when 2 UP arrows were displayed. Right panel (B) indicates the percentage of respondents who decreased their insulin dosages 0 to 100% when 2 DOWN arrows were displayed.
Real-Time Use Versus Retrospective Analysis CGM device downloads were performed sporadically; 40% never downloaded, 17% rarely did so, and 15% stated they only downloaded before a clinician office visit. During clinician office visits, 19% of respondents stated that their CGM data were never or rarely downloaded for review. When data were reviewed, 44% reported the time spent on the review was usually less than 5 minutes. The majority of respondents rated the real-time trend information (51%) or the real-time low- and high-glucose alerts (30%) as the most important information gleaned from their CGM device. Conversely, only 3.6% of respondents reported detecting retrospective patterns or problems from CGM downloads either themselves or with their healthcare provider as the most helpful component of CGM. However, 15% thought that the real-time receiver information and downloads were equally important. DISCUSSION The survey was conducted to gain insight into what regular CGM users are doing in their daily lives to translate their CGM information into improvements in their diabetes management. Our findings revealed multiple areas in which patients alter their daily treatment decisions in real time. Notably, patients are frequently being alerted and awoken at night to hyperglycemia and hypoglycemia, perceive greater benefit from the real-time information provided compared to retrospective data analysis, and use the ROC arrows or trend graphs to adjust insulin timing and dosage. The changes that patients are making in their insulin dosage based on ROC arrows are variable, often quite dramatic, and much larger than previous recommendations suggest. Also of note is the lack of retrospective CGM analysis reported by study participants, similar to data reported in the type 1 diabetes exchange diabetic registry (8).
This survey highlights some areas in which patients are using CGM to achieve a lower A1c. High-glucose alerts commonly wake up CGM users. As nocturnal hyperglycemia typically goes unrecognized with SMBG, sleeping represents a potentially large hyperglycemic burden. The ability to be alerted and to respond to nocturnal hyperglycemia explains some of the glycemic improvement. The majority of respondents also stated that they were now taking more insulin boluses or injections per day since starting on CGM. Additionally, most respondents reported adjusting their insulin timing relative to a meal, based on the ROC, and many users lowered their glycemic targets since starting on CGM. Thus, the ability of CGM devices to limit nocturnal hyperglycemia, increase frequency of boluses, provide for more rapid correction of hyperglycemia, adjust insulin timing, and lower glucose targets are likely mechanisms of CGM use attributing to better glycemic control and lower A1c values. The most striking observation was the frequency and degree of modifications that respondents were making to their insulin dosing based on their ROC or trend arrows. Insulin-using patients are taught to use standard insulin sensitivity or “correction factors” as part of their rapidacting insulin dose calculation. However, when based on SMBG alone, this standard correction factor does not take into account the patient’s glucose trend. Even in individuals who use CGM, a published recommendation suggests a modest 10 to 20% increase or decrease in their dose adjustments depending on the direction of the trend arrows (9). In this survey, respondents reported making significantly larger adjustments to their insulin dose. For example, a patient who would normally take 3 units to correct for a glucose value of 220 mg/dL based on a flat or horizontal glucose trend would increase that dose to 7.2 units (a 140% increase) in response to 2 arrows up. Thus, in that situation, the majority of the bolus would
Using CGM to Make Treatment Decisions, Endocr Pract. 2015;21(No. 6) 619
be derived from the ROC and would not be considered if a user was dependent only on a snapshot glucose value provided by SMBG. The large adjustments that respondents reported making here were unexpected but may be appropriate. In a hypothetical patient using a correction factor of 1 unit: 35 mg/dL, a correction dose of 3 units would be taken for a glucose of 225 mg/dL with a horizontal or flat trend arrow to lower the glucose to 120 mg/dL. If the glucose was rising at a sustained rate of 3 mg/dL/minute, in 30 minutes the glucose would rise 90 mg/dL to 315 mg/dL; in 60 minutes, the glucose would rise 180 mg/dL to 405 mg/dL. In this way, patients may not be calculating their correction dose based on what the current glucose value is, but rather, what they perceive it will be when the insulin bolus is acting or peaking. This is similar for rapidly falling glucose. Even without further corrective insulin, in 30 minutes the glucose would fall from 225 to 130 mg/dL. Clearly, the recommendations of a 10 or 20% reduction to a dose would have minimal glycemic impact in preventing significant hypoglycemia. In addition to lowering A1c, use of CGM devices has proven to help patients “stay in range,” with many studies showing reduction in hypoglycemia (1,2,10). In line with previous studies, the most of the respondents in our study reported a decrease in hypoglycemic events since starting on their CGM device. There were 3 key actions patients took in response to their CGM data that could lessen hypoglycemia. First, most respondents reduced or eliminated their insulin dose in response to ROC arrows for both correction insulin doses and mealtime doses. Second, the majority of respondents were prophylactically treating themselves to prevent hypoglycemia rather than waiting for hypoglycemia to occur. Both of these situations suggest that patients are using their CGM devices to make more proactive choices with their diabetes management to avoid hypoglycemia entirely. Finally, and perhaps most importantly, was the response to nocturnal hypoglycemia alerts. Most respondents stated that they were awoken by their CGM alerts in response to low-glucose alerts at least once per week. As nocturnal hypoglycemia represents a potentially deadly phenomenon, the frequent nocturnal CGM alerts reported highlights the potentially lifesaving aspect of CGM alerts. Further, the ability of CGM devices to alert not only the patient but also friends and family as reported in the survey represents another potential area in which the CGM alerts may reduce the duration and severity of hypoglycemia. The strength of this survey is the large number of diverse respondents that came from across the U.S. Many of the questions were asked as detailed clinical scenarios, which are reflective of real-life decisions people using CGM face day-to-day. However, there are several limitations, and it is unknown whether the results are generalizable. The data came from regular, experienced, and
provider-identified successful CGM users. Respondents in this survey reported confidence in the accuracy of the glucose readings, which obviously had an impact on their decision making, especially the large adjustments made to their insulin doses. Patients who do not trust the accuracy of their CGM data may be less willing to make similar adjustments based on their CGM data (11) and/or use their CGM device frequently (12). The data was all selfreported, and the survey can only report what patients state they are doing and not necessarily what they are actually doing or what providers should recommend. This survey did not determine if the insulin dose adjustments made by the participants resulted in improved glucose control in the immediate time frame following any dose change; however, the respondents did report overall reductions in hypoglycemia as well as an improved A1c. CONCLUSION In summary, the current study is the first report describing how patients are using their CGM devices to modify their diabetes management. The extent and magnitude of these adjustments had not been previously quantified; the insulin dose modifications are far greater than commonly recommended. It is clear from this survey that patients place a tremendous amount of importance on ROC information when determining insulin doses and depend on CGM alerts for their safety. Research is needed to determine appropriate dose adjustments based on ROC, with the goal of translating this research into practical guidelines or formulas that patients could use. In this way, ROC correctional formulas could become the next big piece of information to provide patients to be used in addition to carbohydrate ratio and correction factors to more accurately determine insulin boluses. ACKNOWLEDGMENT We thank the study participants and consultants with the SiGMa group for enrolling participants and reviewing the manuscript: Gregg Gerety, MD, Albany, NY; Damon Tanton, MD, Celebration, FL; Nicholas B. Argento, MD, Columbia, MD; Kimberly Bourne, MD, Orlando, FL; Laura Akright, MD, Schertz, TX; Joe Henske, MD, Glen Ellyn, IL; Rachel Malish, ACNS-BC, CDE, Austin, TX; Heather Lamar, RD, CDE, Las Vegas, NV; Michael Harris, MD, Beverly Hills, CA; Tomas Walker, PhD, DNP, APN, CDE, Henderson, NV; John Purcell, MD, Ponte Vedra Beach, FL; Firas Akhrass, MD, San Antonio, TX; Melissa Young, MD, Freehold, NJ; Chris Sadler, MA, PA, CDE, La Jolla, CA; Jerome Fischer, MD, San Antonio, TX; Traci Wells, MD, MSN, Beachwood, OH; Rosemarie Lajara, MD, Plano, TX. We also thank Kiley Hill for assistance with survey planning and execution, Katherine Nakamura for statistical analysis, Farah Bowman for assistance with
620 Using CGM to Make Treatment Decisions, Endocr Pract. 2015;21(No. 6)
SurveyGizmo, and Christopher G. Parkin for editorial assistance in the preparation of this manuscript. Funding for the study was provided by Dexcom Inc, San Diego, CA. The study concept, protocol development, study implementation, study interpretation, and manuscript development were done by Drs. Pettus, Edelman, and Price. All authors read and approved the final manuscript. DISCLOSURE Dr. Edelman has served as a consultant and advisory board member for and has received research funding from Dexcom Inc. Dr. Pettus has served as a consultant for Dexcom Inc. Dr. Price is an employee of Dexcom Inc. REFERENCES 1. Tamborlane WV, Beck RW, Bode BW, et al. Juvenile Diabetes Research Foundation Continuous Glucose Monitoring Study Group. Continuous glucose monitoring and intensive treatment of type 1 diabetes. N Engl J Med. 2008;359:1464-1476. 2. Battelino T, Conget I, Olsen B, et al. The use and efficacy of continuous glucose monitoring in type 1 diabetes treated with insulin pump therapy: a randomised controlled trial. Diabetologia. 2012;55:3155-3162. 3. Riveline JP, Jollois FX, Messaoudi N, et al. Insulin-pump use in everyday practice: data from an exhaustive regional registry in France. Diabetes Metab. 2008;34:132-139. 4. Bergenstal RM, Tamborlane WV, Ahmann A, et al. Effectiveness of sensor-augmented insulin-pump therapy in type 1 diabetes. N Engl J Med. 2010;363:311-320.
5. Blevins TC, Bode BW, Garg SK, et al. Statement by the American Association of Clinical Endocrinologists Consensus Panel on continuous glucose monitoring. Endocr Pract. 2010;16:730-745. 6. Klonoff DC, Buckingham B, Christiansen JS, et al. Continuous glucose monitoring: an Endocrine Society Clinical Practice Guideline. J Clin Endocrinol Metab. 2011;96:29682979. 7. Diabetes Research In Children Network (DirecNet) Study Group. Use of the DirecNet Applied Treatment Algorithm (DATA) for diabetes management with a real-time continuous glucose monitor (the FreeStyle Navigator). Pediatr Diabetes. 2008;9:142-147. 8. Wong JC, Foster NC, Maahs DM, et al. Real-time continuous glucose monitoring among participants in the T1D Exchange clinic registry. Diabetes Care. 2014;37:2702-2709. 9. Hirsch IB. Clinical review: Realistic expectations and practical use of continuous glucose monitoring for the endocrinologist. J Clin Endocrinol Metab. 2009;94:2232-2238. 10. Garg S, Zisser H, Schwartz S, et al. Improvement in glycemic excursions with a transcutaneous, real-time continuous glucose sensor: a randomized controlled trial. Diabetes Care. 2006;29:44-50. 11. Polonsky WH, Hesseler D. What are the quality of liferelated benefits and losses associated with real-time continuous glucose monitoring? A survey of current users. Diabetes Technol Ther. 2013;15:295-301. 12. Chamberlain J, Dopita D, Gilgen E. Persistence of continuous glucose monitoring use in a community setting 1 year after purchase. Clin Diabetes. 2013;31:106-109.
Submitted for publication October 28, 2014 Accepted for publication January 29, 2015 From the 1University of California San Diego, San Diego, California, and 2Dexcom, Inc, San Diego, California. Address correspondence to Dr. Jeremy Pettus, University of California San Diego, 4916 Bayard Street, San Diego, CA 92109. E-mail: [email protected]. Published as a Rapid Electronic Article in Press at http://www.endocrine practice.org on February 25, 2015. DOI: 10.4158/EP14520.OR To purchase reprints of this article, please visit: www.aace.com/reprints. Copyright © 2015 AACE.
glucose levels. With a glucose value of 120 mg/dL and a falling glucose trend, 70% of respondents would prophylactically consume carbohydrates to avoid hypoglycemia. Conclusion: CGM users utilize CGM data to alter multiple aspects of their diabetes care, including insulin dose timing, dose adjustments, and in hypoglycemia prevention. The insulin adjustments are much larger than common recommendations. Additional studies are needed to determine appropriate insulin adjustments based on glucose trend data. (Endocr Pract. 2015;21:613-620) Abbreviations: A1c = hemoglobin A1c; CGM = continuous glucose monitoring; ROC = rate of change; SMBG = selfmonitored blood glucose INTRODUCTION As a result of studies showing improvements in glycemic control in children and adults who regularly use continuous glucose monitoring (CGM) (1-4), the use of realtime CGM is becoming accepted as part of the standard of care in the treatment of patients with type 1 diabetes (5,6). However, the mechanism of action for glycemic benefit from CGM use differs from pharmacologic interventions. The observed benefits are not a direct result of wearing the device but rather the result of behavioral and management changes enabled by the information provided by the CGM devices to the users. The specific interventions patients make in their diabetes management in response to the information provided by CGM devices has not been elucidated. Clinical studies provide respondents with instruction on how to operate their CGM devices; however, few studies provide specific instructions to subjects on how to respond to the real-time glucose information including trend arrows. Exceptions were the DirectNet (7) and Juvenile Diabetes Research Foundation (1) CGM trials. In these studies, subjects were instructed to increase their insulin dose by 10% in response to a rising glucose (1 to
ENDOCRINE PRACTICE Vol 21 No. 6 June 2015 613
614 Using CGM to Make Treatment Decisions, Endocr Pract. 2015;21(No. 6)
2 mg/dL/minute rise) and 20% for more rapidly rising glucose (>2 mg/dL/minute rise), with identical reductions in their dose for down arrows. Although these recommendations generally resulted in improved clinical outcomes, it remains unknown if the recommendations were appropriate, or if patients are making similar adjustments in the real-world setting. Additionally, the recommendations used in these trials have not transitioned into the clinic setting in the form of official guidelines, leaving clinicians without guidance to provide their patients using CGM and often resulting in patients making independent adjustments. Because there are no previous reports that we are aware of that describe in detail how CGM users utilize their devices, a survey was conducted to assess how patients are using CGM and responding to their CGM information in a real-world setting. This report describes and discusses the findings of the survey. METHODS The survey comprised 6 sections: (1) patient characteristics; (2) general CGM use; (3) hypoglycemia prevention and management; (4) hyperglycemia prevention and management; (5) insulin dosing adjustments (both for incidental hyperglycemia not at meals and at mealtimes); and (6) real-time use versus retrospective analysis. In order to contextualize the information, many of the survey questions were framed as clinical scenarios that would be commonly experienced by patients either on multiple daily injections or using an insulin pump. The questions were beta-tested in 20 experienced CGM users and refined repeatedly to assure that the questions were well understood, clear, and unambiguous. Table 1 provides examples of the scenario-based questions. An institutional review board waiver was obtained, and clinical endocrinologists or educators that actively prescribe CGM from across the U.S. were asked to recruit patients. They identified regular Dexcom CGM (Dexcom Inc, San Diego, CA) users (>6 days a week on average) from their practice and provided them a web link to the survey. The on-line survey was available from May 28 to August 26, 2013, using SurveyGizmo (Boulder CO), and included 70 multiple-choice questions. Based on beta testing, it was estimated to take 20 to 30 minutes to complete the survey. Respondents chose from one of three different surveys in which the questions were identical but the order of the sections varied and were provided a $30 gift card for completing the survey. Statistical Methods The survey was descriptive, and no hypothesis testing was performed and no comparative analyses were made. Categorical variables are summarized using counts and percentages. Summary statistics for continuous vari-
ables are summarized using mean and standard deviation. Histograms and other graphical displays are used to illustrate the distribution of the survey responses. SAS software, version 9.3 or later, was used to conduct data conversion and analysis. RESULTS Patient Characteristics The respondents included 222 patients with type 1 diabetes from 22 states across the United States. The mean age of respondents was 46 ± 14 years, the duration of diabetes 22 ± 14 years, 52% were male, and the self-reported hemoglobin A1c (A1c) was 6.9 ± 0.8%. For their method of insulin delivery, 75% used an insulin pump and 25% used multiple daily injections. Education levels varied: 0.5% did not graduate high school, 9% were high school graduates or attended technical or trade school after high school, 14% attended college but did not graduate, 42% were college graduates, and 23% received an advanced degree. Self-perceived math skills were probed, with 6% stating they were below average (compared to most people), 23% believed their math skills were average, and 68% reported their math skills were above or well above average. Participants were not queried about their glycemic awareness status. General CGM Use Seventy-five percent of the respondents had used their CGM device for greater than 1 year. Sensor duration also varied: 43% reported typically wearing sensors for the recommended 7 days (39%) or less (4%), and 57% typically wear sensors for greater than 7 days, including 32% reporting wearing sensors for more than 11 days before changing. Sensors were most commonly worn on the abdomen, followed by the back of the upper arm and upper thigh. The majority of respondents (52%) reported starting CGM because of problems with both high and low glucose, 22% mainly because of problematic hypoglycemia, and 10% because of problematic hyperglycemia. Hypoglycemia Prevention and Management Almost all respondents used a customized low-glucose alert (set between 60 and 100 mg/dL), with only 1% of respondents relying only on the fixed low alarm set at 55 mg/dL that cannot be turned off on Dexcom CGM receivers. Forty-four percent of respondents stated that they would feel symptomatic from hypoglycemia prior to being alerted via their low alarm. Conversely, 33% felt that they were alerted by their CGM prior to symptoms being present. Since starting CGM, 78% reported that the frequency and severity of hypoglycemia had decreased. The majority of respondents (70%) reported waking up at night at least once per week in response to their lowglucose alert, with 72% stating that they rarely or never
Using CGM to Make Treatment Decisions, Endocr Pract. 2015;21(No. 6) 615
Table 1 Sample Survey Questions You are awoken at night to a low-glucose alert. You look at your CGM and see that you are at 70 mg/dL with an angled down arrow. You react by (select all that apply) (choices): I do not use a low-glucose alert setting; I confirm the glucose with a blood glucose test on my meter most of the time; I treat my glucose with food or drink, without performing a blood glucose test on my meter most of the time; I wait until the alert buzzes again before I react most of the time; I turn the alert off for the night most of the time. Imagine it has been 4 hours since your last dose of meal insulin and your CGM receiver shows a value of 220 mg/ dL (matching your fingerstick BG of 220 mg/dL) with trend arrow and trend graph flat (straight across). If you were NOT planning on eating or exercising, what dose of insulin would you give yourself to bring your glucose down to around 120 mg/dL? (choose the closest value). Imagine it has been 4 hours since your last dose of meal insulin and your CGM receiver shows a value of 220 mg/dL (matching your fingerstick BG of 220 mg/dL) with 1 arrow (trend arrow) straight up. If you were NOT planning on eating or exercising, how many additional units would you give, compared to when your trend arrow was flat to bring your glucose down to around 120 mg/dL? (choose the closest value). Assume it is lunch time and your glucose is 110 mg/dL. You have 2 arrows (trend arrows) pointing straight up. You are eating a meal with 50 grams of carbohydrates and your usual fat and protein. How much would you increase your meal insulin, compared to if your trend arrow was flat? (choose the closest value). If my glucose was 120 mg/dL, the earliest I usually take immediate action to avoid hypoglycemia if my trend arrow is (Choices): arrow angled down; 1 arrow straight down; 2 arrows straight down; I don’t ever respond to the trend arrows; I don’t respond to the trend arrows at a glucose of 120 mg/dL. Since starting CGM, the blood glucose values I am targeting, during the day (compared to before I was on CGM) are now (Choices): higher; lower; have not changed; I’m not sure. Abbreviations: BG = blood glucose; CGM = continuous glucose monitoring. slept through their alert and 19% stating that their spouse/ partner was typically alerted first. Half the respondents stated they would confirm the CGM low glucose with a fingerstick blood glucose when alerted to a low at night, whereas half stated they would treat the low glucose without confirming. Forty-two percent (42%) of respondents stated that their CGM device alerted somebody around them to respond to their hypoglycemia alarm when they themselves were unable to respond at least one time in the last 6 months. To prevent hypoglycemia, 70% would prophylactically consume carbohydrates in response to a displayed glucose of 120 mg/dL with a decreasing trend (angled or downward rate of change [ROC] arrow). Hyperglycemia Prevention and Management Since starting CGM, 59% of respondents stated that their A1c had decreased by at least 0.5%, whereas 4.5% stated that it had increased. The remainder indicated no change or they were unsure. Almost all respondents (98%) used high-glucose alerts. Among 70% of respondents, the high alert setting was 200 mg/dL or less. The majority of respondents (66%) stated that they woke up at least once per week at night in response to their hyperglycemia alert. When alert-
ed at night to a high glucose, the majority (79%) stated they would respond by taking an extra injection or bolus (if there was 1 trend arrow up) on the majority of occasions, 9% stated they would occasionally bolus, and 12% stated they would not respond. When alerted to a nocturnal high glucose, 66% would usually or always confirm with a fingerstick blood glucose measurement, 23% would confirm less than half the time, and 11% would never do a confirmatory fingerstick. Sixty-five percent of respondents reported that since starting CGM, the number of injections or boluses of shortacting insulin they took per day had increased, 15% felt the number had decreased, and the remainder indicated no change. However, respondents were not asked if their total daily dose of insulin had changed. With a glucose of 130 mg/dL and a rapid rise (2 arrows up), 22% stated they would take additional insulin even if it was only 1 hour since their previous insulin dose. Others (35%) would dose only if it was more than 2 hours since their dose, and 15% would wait until the glucose value was high. Only 19% would not dose based on high rates of glucose rise. Respondents were not asked if they frequently had low glucose within a few hours of a correction bolus. Since starting CGM, many respondents adjusted their glucose targets. During the day, 57% lowered, 35% did
616 Using CGM to Make Treatment Decisions, Endocr Pract. 2015;21(No. 6)
not change, and 4% raised their glucose targets. During the night, 38% lowered, 42% did not change, and 14% raised their glucose targets. The remainder did not recall their past targets. Insulin Dosing Adjustments for Incidental Hyperglycemia, Unrelated to a Meal Respondents were provided a scenario in which it had been 4 hours since taking any insulin or eating and their CGM device showed a glucose value of 220 mg/dL (confirmed by self-monitored blood glucose [SMBG]) with a flat ROC arrow. They were asked how much insulin they would take to correct their glucose to 120 mg/dL if they were not planning on eating or exercising. Then, respondents were asked the same question with varying ROC arrows (1 arrow up, 2 arrows up, 1 arrow down, 2 arrows down) and asked how the insulin dose would change compared to when the trend arrow was flat. The percentages of increase or decrease of insulin dose were calculated from their typical correctional dose at the flat CGM arrow and included in Figure 1. These results were also averaged and are summarized in Table 2.
With a flat arrow indicating stable glucose values, the average correction dose was 2.8 units, indicating an average correctional factor of 35.7 mg/dL/unit. In response to 1 or 2 arrows up, respondents stated they would increase their typical correctional dose by 111% and 140%, respectively. However, the range of responses varied from 0 to 600%. In response to 1 or 2 arrows down, respondents stated they would decrease their average correctional dose by 40 and 42%, respectively. The responses varied from 0 to 100%, indicating that some patients would not adjust their bolus and others would omit it completely. Of note, in the more extreme situation of 2 arrows down, 24% of patients would not bolus at all in response to a glucose of 220 mg/dL. Although the range of responses varied significantly, the average correction dose (adjusted for the average correction dose of 2.8 units at stable glucose) ranged from 1.6 to 6.7 units, depending on the ROC arrow. Insulin Dosing Adjustments at Mealtime Respondents were provided a scenario in which their CGM showed a glucose value of 110 mg/dL and they were
Fig. 1. Impact of the direction and rate of glucose change on a correctional insulin dose at hyperglycemia (220 mg/dL) unrelated to a meal or prior insulin dose. Left panels indicate the percentage of respondents who increased their insulin dosages 0 to 400% when 1 UP arrow (A) or 2 UP arrows (C) were displayed. Right panels indicate the percentage of respondents who decreased their insulin dosages 0 to 100% when 1 DOWN arrow (B) or 2 DOWN arrows (D) were displayed.
Using CGM to Make Treatment Decisions, Endocr Pract. 2015;21(No. 6) 617
Table 2 Comparison of Recommended Insulin Dosage Adjustments vs. Participant Self-Reported Correctional Insulin Dose Adjustments in Response to Rate of Change Arrows Recommended insulin dosage adjustmenta,b
Participants’ reported insulin dosage adjustment
Constant: glucose is steady – not increasing/decreasing more than 1 md/dL per minute
No change
NA
Slowly rising: glucose is rising 1-2 mg/dL per minute
10% higher
Rising: glucose is rising 2-3 mg/dL per minute
20% higher
111% higher
Rapidly rising: glucose is rising more than 3 mg/dL per minute
20% higher
140% higher
Indication
10% lower
Falling: glucose is falling 2-3 mg/dL per minute
20% lower
40% lower
Rapidly falling: glucose is falling more than 3 mg/dL per minute
20% lower
42% lower
Slowly falling: glucose is falling 1-2 mg/ dL per minute
Abbreviation: NA = not applicable. aDiabetes Research In Children Network (DirecNet) Study Group. Pediatr Diabetes. 2008;9:142-147. bThe Juvenile Diabetes Research Foundation Continuous Glucose Monitoring Study Group. Diabetes Technol Ther. 2008;10:310-321.
planning to eat 50 grams of carbohydrates. They were asked how much insulin they would take for that meal when the trend arrow was flat. In the subsequent questions, respondents were asked how the meal dose would change if the glucose and meal were the same but there were 2 trend arrows up or 2 arrows down. Based on these data, a percent increase or decrease was calculated from their typical meal insulin dose. In response to a flat arrow, the average bolus for a 50-gram carbohydrate meal was 4.7 units, indicating an average insulin to carbohydrate ratio of 1:10.6. As shown in Figure 2, in response to 2 arrows up, respondents would increase their typical mealtime dose by an average 81%. Conversely, in response to 2 arrows down, respondents would decrease their dose by an average 46%. In addition to increasing their bolus at mealtime in response to 2 arrows up, 59% of respondents stated
that they would allow more time between their insulin bolus and starting their meal. Conversely, in response to 2 arrows down, 60% stated that they would take their insulin bolus after their meal. Therefore, both the insulin timing and the insulin dose amount were affected by ROC or arrows. In response to a question that asked how they would determine the actual changes in their insulin dose based on ROC arrows, most respondents (61%) stated they would usually estimate the dose adjustment based mainly on their past experiences; only 14% stated they would use a certain percent increase or decrease in their insulin dose based on the ROC arrow (e.g., 20% increase for 1 arrow up), and 14% stated they would take a fixed dose of insulin based on the ROC arrow (e.g., extra 2 units for 1 arrow up). Only 11% of respondents stated they did not use the ROC arrows to adjust their dose.
618 Using CGM to Make Treatment Decisions, Endocr Pract. 2015;21(No. 6)
Fig. 2. Impact of the direction and rate of glucose change on a mealtime insulin dose at euglycemia (110 mg/dL). Left panel (A) indicates the percentage of respondents who increased their insulin dosages 0 to 400% when 2 UP arrows were displayed. Right panel (B) indicates the percentage of respondents who decreased their insulin dosages 0 to 100% when 2 DOWN arrows were displayed.
Real-Time Use Versus Retrospective Analysis CGM device downloads were performed sporadically; 40% never downloaded, 17% rarely did so, and 15% stated they only downloaded before a clinician office visit. During clinician office visits, 19% of respondents stated that their CGM data were never or rarely downloaded for review. When data were reviewed, 44% reported the time spent on the review was usually less than 5 minutes. The majority of respondents rated the real-time trend information (51%) or the real-time low- and high-glucose alerts (30%) as the most important information gleaned from their CGM device. Conversely, only 3.6% of respondents reported detecting retrospective patterns or problems from CGM downloads either themselves or with their healthcare provider as the most helpful component of CGM. However, 15% thought that the real-time receiver information and downloads were equally important. DISCUSSION The survey was conducted to gain insight into what regular CGM users are doing in their daily lives to translate their CGM information into improvements in their diabetes management. Our findings revealed multiple areas in which patients alter their daily treatment decisions in real time. Notably, patients are frequently being alerted and awoken at night to hyperglycemia and hypoglycemia, perceive greater benefit from the real-time information provided compared to retrospective data analysis, and use the ROC arrows or trend graphs to adjust insulin timing and dosage. The changes that patients are making in their insulin dosage based on ROC arrows are variable, often quite dramatic, and much larger than previous recommendations suggest. Also of note is the lack of retrospective CGM analysis reported by study participants, similar to data reported in the type 1 diabetes exchange diabetic registry (8).
This survey highlights some areas in which patients are using CGM to achieve a lower A1c. High-glucose alerts commonly wake up CGM users. As nocturnal hyperglycemia typically goes unrecognized with SMBG, sleeping represents a potentially large hyperglycemic burden. The ability to be alerted and to respond to nocturnal hyperglycemia explains some of the glycemic improvement. The majority of respondents also stated that they were now taking more insulin boluses or injections per day since starting on CGM. Additionally, most respondents reported adjusting their insulin timing relative to a meal, based on the ROC, and many users lowered their glycemic targets since starting on CGM. Thus, the ability of CGM devices to limit nocturnal hyperglycemia, increase frequency of boluses, provide for more rapid correction of hyperglycemia, adjust insulin timing, and lower glucose targets are likely mechanisms of CGM use attributing to better glycemic control and lower A1c values. The most striking observation was the frequency and degree of modifications that respondents were making to their insulin dosing based on their ROC or trend arrows. Insulin-using patients are taught to use standard insulin sensitivity or “correction factors” as part of their rapidacting insulin dose calculation. However, when based on SMBG alone, this standard correction factor does not take into account the patient’s glucose trend. Even in individuals who use CGM, a published recommendation suggests a modest 10 to 20% increase or decrease in their dose adjustments depending on the direction of the trend arrows (9). In this survey, respondents reported making significantly larger adjustments to their insulin dose. For example, a patient who would normally take 3 units to correct for a glucose value of 220 mg/dL based on a flat or horizontal glucose trend would increase that dose to 7.2 units (a 140% increase) in response to 2 arrows up. Thus, in that situation, the majority of the bolus would
Using CGM to Make Treatment Decisions, Endocr Pract. 2015;21(No. 6) 619
be derived from the ROC and would not be considered if a user was dependent only on a snapshot glucose value provided by SMBG. The large adjustments that respondents reported making here were unexpected but may be appropriate. In a hypothetical patient using a correction factor of 1 unit: 35 mg/dL, a correction dose of 3 units would be taken for a glucose of 225 mg/dL with a horizontal or flat trend arrow to lower the glucose to 120 mg/dL. If the glucose was rising at a sustained rate of 3 mg/dL/minute, in 30 minutes the glucose would rise 90 mg/dL to 315 mg/dL; in 60 minutes, the glucose would rise 180 mg/dL to 405 mg/dL. In this way, patients may not be calculating their correction dose based on what the current glucose value is, but rather, what they perceive it will be when the insulin bolus is acting or peaking. This is similar for rapidly falling glucose. Even without further corrective insulin, in 30 minutes the glucose would fall from 225 to 130 mg/dL. Clearly, the recommendations of a 10 or 20% reduction to a dose would have minimal glycemic impact in preventing significant hypoglycemia. In addition to lowering A1c, use of CGM devices has proven to help patients “stay in range,” with many studies showing reduction in hypoglycemia (1,2,10). In line with previous studies, the most of the respondents in our study reported a decrease in hypoglycemic events since starting on their CGM device. There were 3 key actions patients took in response to their CGM data that could lessen hypoglycemia. First, most respondents reduced or eliminated their insulin dose in response to ROC arrows for both correction insulin doses and mealtime doses. Second, the majority of respondents were prophylactically treating themselves to prevent hypoglycemia rather than waiting for hypoglycemia to occur. Both of these situations suggest that patients are using their CGM devices to make more proactive choices with their diabetes management to avoid hypoglycemia entirely. Finally, and perhaps most importantly, was the response to nocturnal hypoglycemia alerts. Most respondents stated that they were awoken by their CGM alerts in response to low-glucose alerts at least once per week. As nocturnal hypoglycemia represents a potentially deadly phenomenon, the frequent nocturnal CGM alerts reported highlights the potentially lifesaving aspect of CGM alerts. Further, the ability of CGM devices to alert not only the patient but also friends and family as reported in the survey represents another potential area in which the CGM alerts may reduce the duration and severity of hypoglycemia. The strength of this survey is the large number of diverse respondents that came from across the U.S. Many of the questions were asked as detailed clinical scenarios, which are reflective of real-life decisions people using CGM face day-to-day. However, there are several limitations, and it is unknown whether the results are generalizable. The data came from regular, experienced, and
provider-identified successful CGM users. Respondents in this survey reported confidence in the accuracy of the glucose readings, which obviously had an impact on their decision making, especially the large adjustments made to their insulin doses. Patients who do not trust the accuracy of their CGM data may be less willing to make similar adjustments based on their CGM data (11) and/or use their CGM device frequently (12). The data was all selfreported, and the survey can only report what patients state they are doing and not necessarily what they are actually doing or what providers should recommend. This survey did not determine if the insulin dose adjustments made by the participants resulted in improved glucose control in the immediate time frame following any dose change; however, the respondents did report overall reductions in hypoglycemia as well as an improved A1c. CONCLUSION In summary, the current study is the first report describing how patients are using their CGM devices to modify their diabetes management. The extent and magnitude of these adjustments had not been previously quantified; the insulin dose modifications are far greater than commonly recommended. It is clear from this survey that patients place a tremendous amount of importance on ROC information when determining insulin doses and depend on CGM alerts for their safety. Research is needed to determine appropriate dose adjustments based on ROC, with the goal of translating this research into practical guidelines or formulas that patients could use. In this way, ROC correctional formulas could become the next big piece of information to provide patients to be used in addition to carbohydrate ratio and correction factors to more accurately determine insulin boluses. ACKNOWLEDGMENT We thank the study participants and consultants with the SiGMa group for enrolling participants and reviewing the manuscript: Gregg Gerety, MD, Albany, NY; Damon Tanton, MD, Celebration, FL; Nicholas B. Argento, MD, Columbia, MD; Kimberly Bourne, MD, Orlando, FL; Laura Akright, MD, Schertz, TX; Joe Henske, MD, Glen Ellyn, IL; Rachel Malish, ACNS-BC, CDE, Austin, TX; Heather Lamar, RD, CDE, Las Vegas, NV; Michael Harris, MD, Beverly Hills, CA; Tomas Walker, PhD, DNP, APN, CDE, Henderson, NV; John Purcell, MD, Ponte Vedra Beach, FL; Firas Akhrass, MD, San Antonio, TX; Melissa Young, MD, Freehold, NJ; Chris Sadler, MA, PA, CDE, La Jolla, CA; Jerome Fischer, MD, San Antonio, TX; Traci Wells, MD, MSN, Beachwood, OH; Rosemarie Lajara, MD, Plano, TX. We also thank Kiley Hill for assistance with survey planning and execution, Katherine Nakamura for statistical analysis, Farah Bowman for assistance with
620 Using CGM to Make Treatment Decisions, Endocr Pract. 2015;21(No. 6)
SurveyGizmo, and Christopher G. Parkin for editorial assistance in the preparation of this manuscript. Funding for the study was provided by Dexcom Inc, San Diego, CA. The study concept, protocol development, study implementation, study interpretation, and manuscript development were done by Drs. Pettus, Edelman, and Price. All authors read and approved the final manuscript. DISCLOSURE Dr. Edelman has served as a consultant and advisory board member for and has received research funding from Dexcom Inc. Dr. Pettus has served as a consultant for Dexcom Inc. Dr. Price is an employee of Dexcom Inc. REFERENCES 1. Tamborlane WV, Beck RW, Bode BW, et al. Juvenile Diabetes Research Foundation Continuous Glucose Monitoring Study Group. Continuous glucose monitoring and intensive treatment of type 1 diabetes. N Engl J Med. 2008;359:1464-1476. 2. Battelino T, Conget I, Olsen B, et al. The use and efficacy of continuous glucose monitoring in type 1 diabetes treated with insulin pump therapy: a randomised controlled trial. Diabetologia. 2012;55:3155-3162. 3. Riveline JP, Jollois FX, Messaoudi N, et al. Insulin-pump use in everyday practice: data from an exhaustive regional registry in France. Diabetes Metab. 2008;34:132-139. 4. Bergenstal RM, Tamborlane WV, Ahmann A, et al. Effectiveness of sensor-augmented insulin-pump therapy in type 1 diabetes. N Engl J Med. 2010;363:311-320.
5. Blevins TC, Bode BW, Garg SK, et al. Statement by the American Association of Clinical Endocrinologists Consensus Panel on continuous glucose monitoring. Endocr Pract. 2010;16:730-745. 6. Klonoff DC, Buckingham B, Christiansen JS, et al. Continuous glucose monitoring: an Endocrine Society Clinical Practice Guideline. J Clin Endocrinol Metab. 2011;96:29682979. 7. Diabetes Research In Children Network (DirecNet) Study Group. Use of the DirecNet Applied Treatment Algorithm (DATA) for diabetes management with a real-time continuous glucose monitor (the FreeStyle Navigator). Pediatr Diabetes. 2008;9:142-147. 8. Wong JC, Foster NC, Maahs DM, et al. Real-time continuous glucose monitoring among participants in the T1D Exchange clinic registry. Diabetes Care. 2014;37:2702-2709. 9. Hirsch IB. Clinical review: Realistic expectations and practical use of continuous glucose monitoring for the endocrinologist. J Clin Endocrinol Metab. 2009;94:2232-2238. 10. Garg S, Zisser H, Schwartz S, et al. Improvement in glycemic excursions with a transcutaneous, real-time continuous glucose sensor: a randomized controlled trial. Diabetes Care. 2006;29:44-50. 11. Polonsky WH, Hesseler D. What are the quality of liferelated benefits and losses associated with real-time continuous glucose monitoring? A survey of current users. Diabetes Technol Ther. 2013;15:295-301. 12. Chamberlain J, Dopita D, Gilgen E. Persistence of continuous glucose monitoring use in a community setting 1 year after purchase. Clin Diabetes. 2013;31:106-109.
