Dexcom updates Clarity with AGP report

I first came across the term “ambulatory glucose profile” in a study published in 2013. While other sites charge a fee for the full text article, the site I linked to makes it available, in full, for free. It was originally published in the Journal of Diabetes Science and Technology.

Dexcom Clarity, as of June 3, 2017, has made the AGP a part of its web-based software. I applaud this development. I sure wish, however that they’d let the user customize the targeted BG range. (See correction in the comments.) I see my hypoglycemia threshold at 65 mg/dl (3.6) instead of the fixed 70 mg/dL (3.9). This report uses the upper end of the target range at 180 mg/dL (10 mmol/L) while I set mine much lower.

I know that some people abhor graphs and the number side of diabetes, but I urge you to gain at least a passing awareness of this report as I suspect it is in the process of becoming a standard clinical tool. The next time you see it may be in your endo’s office. The beautiful thing about this one-page report is the the clinician and patient can together look at this data-dense report and quickly extract useful and actionable information in a visually simple way.

While the ambulatory glucose profile is an odd name, the concept is relatively simple. The study suggests that the 14-day period “standard day,” is the ideal amount of time to look at and make treatment decisions and changes going forward. Fewer days may exaggerate the effect of some data outliers while longer periods can obscure significant trends.

A standard day displays in a 24-hour format on the x-axis and blood glucose levels on the y-axis. A 14-day standard day will combine all the data by hour of the day showing the median (50th percentile) as well as other percentile levels determined by individual reports. The AGP report I write about here uses the 10th, 25th, 50th, 75th, and 90th percentiles. Here’s an example from the study cited above.

The green dashed lines represent the 10th percentile (lower line) and 90th percentile (upper line). The dark blue lines mark the 25th percentile (lower line) and 75th percentile (upper line). The bright orange line is the 50th percentile, aka median. The left y-axis is mg/dL and the right y-axis is mmol/L.

Diasend also features the same standardized AGP report.

So, how can we use this report to alter our treatment decisions?

For instance, without dependence on numbers, formulas, or derived indices, clinicians and patients can quickly become skilled at identifying the risk of hypoglycemia. For example, if the 10th percentile curve crosses 70 mg/dL [3.9 mmol/L] or lower, there is moderate risk of hypoglycemia at that time because consistently 10% of the values fall in this range. However, if the 25th percentile curve crosses into hypoglycemia, this implies a marked risk because more than 25% of the glucose values fall in the hypoglycemic range, and consequently this should be addressed before additional therapy is instituted to treat accompanying hyperglycemia as is often seen with significant GV [glycemia variability]. [emphasis added]

In the example above, it’s easily seen that the 10th and 25th percentile dashed lines cross below the 70 mg/dL (3.89 mmol/L) at about 1:00 a.m. while the 10th percentile line crosses again just after 4:00 p.m. The early morning low is the more important of the two to address. This suggest examining any factors that can cause these lows. It might be overdosing for evening snacks or a too-aggressive pump basal rate. The late afternoon propensity to dive low could be exercise related or too much lunch insulin or a too aggressive basal rate.

The beauty of the AGP report is that it can visually call your attention to the time of day that needs action and also suggests corrective factors to consider. The same exercise can be done with analyzing hyperglycemia using the 90th and 75th percentile. The width of the blue section describes the 25th to 75th percentile. It can be quickly understood that squeezing these two lines together by reducing BG variability is highly desirable. It confirmed for me the utility of carb-linited eating.

I’ve also used this graph and the above cited paragraph to persuade a hypo-phobic clinician that my 10th percentile line crossing my hypoglycemia threshold is not a hair-on-fire metabolic emergency, it simply means that there is a “moderate” risk of hypoglycemia. If, however, my 25th percentile line crosses this threshold, then, according to this studies’ authors, I now am at marked risk of hypoglycemia at this time of day.

I found the below AGP interesting. It was derived from data gathered from a small set of gluco-normal people. I find it interesting that even in the non-diabetic, the 10th percentile of blood glucose trends below 70 mg/dL from about 4:00 to 8:00 a.m. Pointing out this fact to my then endocrinologist made her realize that she needs to allow me this same excursion without her duty to warn me!

Congrats to Dexcom’s continued improvement of it’s Clarity program!

Edited to add the link to the full study for free. See first paragraph.


Thank you for the explanation of this AGP graph and how it can be used to better control! My endocrinologist has been using this graph since it became available in Diasend a year or more ago, but he has never really explained it in the way you have presented here, which has brought a clarity and significance to the information and what to keep an eye out for. Knowing the significance of the 10th and 25th percentiles is very helpful. In Diasend, this same information is also available under CGM > Statistics, but it does not include the 10th and 90th percentiles, only 25th and 75th. I find the text-based information easier to access, and so I’ll likely contact Diasend and ask if the same information could be included in the statistics as is portrayed in the AGP graph, since at the moment it includes all the same information except that one bit. Does that study you cite happen to mention what the most helpful timeframe to look at is when using the AGP information?

1 Like

The study makes a case that the latest 14-day period as the clearest, most actionable period, if I understand your question. The previous oft-repeated advice was to wait until you had a three-day trend to make any treatment changes but this tool allows for say lows on Monday, Wednesday, and Friday at a similar time to accumulate enough statistical heft in a 14-day run to warrant examination.

1 Like

The Settings page on Clarity DOES allow the user to set the high and low thresholds for the AGP graph.

1 Like

Thank-you! I looked for it for a fair amount of time but did not scroll far enough. Nice catch! The study I cited made reference to clinicians being able to customize thresholds. I’m glad Dexcom allows this.

Thank you, this is very useful. I wonder if diabetes educator centres are teaching this stuff. I had been taught the same thing about a three-day pattern before changing doses. This was useful in the days of fixed insulin regimens and a regimented diet, but not so much today with the complexities of pumps, MDI, and CGM. I will attempt to use the AGP graph with the information you’ve provided via this research to better my control, especially as it relates to factors above and beyond carbohydrates.

1 Like

I suspect not, but perhaps it will come to their attention soon. When I read this study back in 2013, I thought it would soon move into clinical practice. Here we are four years later and it has yet to reach critical mass. With Diasend and Clarity using it, maybe the more progressive clinicians will find it useful and word can spread. The one-page visual, even with limited patient numeracy, makes it a hands-down winner in a time-starved clinician’s practice. You can lead a horse to water…

Thank you for the explanation. I saw this on clarity but it is new to me. It’s like what I tried to do with the 14 day data graph but this is a much better presentation.

Thanks @Terry4, @Thas, and @Jen for this info and your comments. I spent a while today downloading my D-devices but didn’t go into Clarity since it is done automatically. I probably wouldn’t have checked out the AGP report anyway without your guidance.

I currently find myself downloading to 3 platforms (not including Clarity) and find that nothing is quite perfect for my needs. Right now Tidepool is the only platform that will combine my Tandem pump, my Freestyle meters, and my Dexcom G5 Mobile. Tidepool has also recently allowed users to give permission for our data to be used by researchers and for that reason I will be diligent about regularly downloading to that site. I am in the process of writing a blogpost about Tidepool and will share that once I publish it.

I still use Diasend because it gives me the best logbook printout to give to Medicare. One thing I like about the Diasend logbook is the ability to see it with and without manual entries (meaning numbers from my Dexcom that I enter into my pump for boluses). Because Diasend does not sync with my Dexcom G5 Mobile, it is less useful to me than it used to be. I called Dexcom to ask about that and they indicated that it is doubtful that synchronization with Diasend will occur. Of course if Medicare ever ends up reimbursing the G5 and really requires me to use the receiver, I can download the receiver to Diasend.

I download my pump and meters to the Tandem t:connect site, but it really has no value for me because it does not include my CGM data. But hopefully when the t:slim X2 updates its software to include G5, t:connect will include it.

For those using Nightscout or xdrip, they also have this report.


I like that feature in Diasend where I could upload my Animas Ping pump, my fingerstick meter, and my Dex CGM. Diasend also automatically uploaded my Fitbit but only gave a total number of steps and didn’t sync them across each day. If I input my boluses as carb boluses then Diasend also post total daily carbs. The fingerstick values seemed redundant to the CGM to me. But it was interesting to see the high correlation between the CGM and fingerstick averages.

When I decided to stop using the Ping pump, I lost some meaningful data with the insulin dosing and related carbs. I adopted the Loop protocol last November so the effectiveness of the continuously variable basal rates is better looked at through the lens of the ultimate BG line. So, I’m only monitoring the BG line now and as long as its good, I don’t find the underlying carbs and insulin informative is needed. But that’s just me; ydmv, of course. From the BG line I can make many educated guesses about what to adjust.

I love this from the website.

We believe that you own your own diabetes data. It’s your disease, so it’s your data. If you would like to donate your data to an anonymous database, we’ll make that easy. If you’d like to share data with your doctor, we’ll make that easy, too.

I look forward to reading your take on things.

How do you find your spring migration north this year? There was more than one time this last winter when I wish I was in the Southwest.

I’ve been following your Medicare/CGM comments closely. I’m still using the G4 system as it’s covered under my supplemental insurance. I think Medicare will be cheaper in the long run but I’ll stick with my current insurance/CGM set up through the end of this year. I’m looking forward to the G6 next year.

Sorry for the slightly off-topic detour. I continue to be amazed at what and how Tidepool is helping the patient cause.

The Elevator Pitch

We are building the “Mozilla of Diabetes Platforms.” We will disrupt the current model of diabetes technology and care by delivering a full, end-to-end software stack using an open development model. Our first application, “blip,” enables more accessible and intuitive visualizations and more fluid patient/doctor communication and insights. Other applications include remote monitoring, smart meal bolusing and “Endo in the Sky” recommendations. Our platform will also enable improvements in Artificial Pancreas development, helping researchers to focus on algorithms, trials and effectiveness; we’ll handle the middleware, cloud monitoring and visualization software.

The project will be developed in the open on GitHub. The platform and code will be available for free to non-profits and non-commercial researchers. Within three years we will generate self-sustaining revenue through licensing fees to commercial devices makers; from the other side, patients and clinics will demand that their devices work with our platform. We may also generate revenue from payer subsidies or clinic subscriptions. We will be agile and pivot as necessary to figure this out.

It’s kind of a unique blend between a not-for-profit and a commercial company. I’m curious to see if striving to do good will spin off unexpected profits. I certainly think people and companies may find a social good bottom line is a sufficient reward to motivate the enterprise.

Slight drift - My recent gyration via Liberty Medical to get my G5 supplies (yes, it happened “once”, but no more) was nuts. I kept sending them BG logs (Diasend) over and over to no avail.

Finally they told me the only way Medicare would accept the log would be the “hand written” log on a form they provided. So I took my Diasend logs and wrote all the BG readings in on their forms, scanned them, and sent them back and they were happy. They shipped my transmitter, receiver and one box of sensors.

The next day they backed out of the CGM/Medicare deal.

So at least I have a transmitter for 3-months. I’ve been using Dexcom for well over seven years.

This is a bad rollout for sure.

Interesting because Liberty accepted my logbook from Diasend…

Thanks for this update @Terry4, the AGP report is indeed a nice improvement for Clarity.

In the top part of the AGP page there is a summary of glucose statistics, including (in tiny font) reference ranges calculated from population without diabetes, which I find interesting and somewhat surprising. For example, they give SD of 10-26 mg/dL for non-D people, the upper end of which is higher than what I expected. Unfortunately, they do not say where these non-D reference ranges are coming from. Where did you locate that non-D AGP graph you posted?

The non-D graph is part of Figure 3 in the cited study.

Here’s some additional writing about the the statistics for gluco-normals.

Note that below each data component or panel in the statistical summary section is a reference range derived from a normal reference population (mean ± 2SD).2 Although patients with diabetes (particularly T1DM) are not expected to achieve completely normal glucose values, this gives a frame of reference. Acceptable or desirable values for various subsets of T1DM points (e.g., toddlers, adolescents, adults, elderly, those with hypoglycemic unawareness) can be established over time. Right now, those caring for toddlers (<6 years old), children (<13 years old), and adolescents (13–20 years old) can change target ranges as desired and look at the derived HbA1c (based on CGM data) to see if these patients are reaching the ADA61 or International Society for Pediatric and Adolescent Diabetes63 suggested age-specific targets.

Footnote 2:

  1. Mazze RS, Strock E, Wesley D, Borgman S, Morgan B, Bergenstal R, Cuddihy R. Characterizing glucose exposure for individuals with normal glucose tolerance using continuous glucose monitoring and ambulatory glucose profile analysis. Diabetes Technol Ther. 2008;10(3):149–59. [PubMed]
1 Like

I was just checking out this new report, @Terry4. Thanks for bringing it to my attention. I really like the story it tells.

There are days that diabetes is so unkind and unfair and they seem to consume our brain power exponentially more than the more calm days. My perception becomes that Caleb’s control is out of control. These Dexcom reports are so calming. They reassure me that those days are the exception, not the norm and when he’s actually 80% in range on a consistent basis, we’re winning. A much different picture than I have without this reporting!

1 Like

An 80% TIR is a winner, and not easily done! I assume that an occasional bad day for Caleb is the source of your worry. We all have bad BG days. Stuff happens and we cannot anticipate everything and always respond appropriately.

I was reminded of how easy it is to get off track with a diabetes regimen. I target changing my infusion site every 72 hours (three days). I set up a reminder system with my Google calendar. Yet, on a recent travel visit with relatives, I overlooked two sensors in a row and let them go to four days. Degrading BG lines finally got my attention.

I can fully understand the inherent distractions in any teenager’s life. Diabetes is harder for them to control simply due to that factor alone. Add in growth spurts and burgeoning hormones and it seems harder still.

Since starting to Loop, diabetes has kicked our butts. We’ve had to learn a lot of new things, and we’ve made mistakes. He was also sick - strep throat followed by a cold. His strep was treated w a liquid antibiotic - something I’ve avoided since he was wee, but the pharmacy was out of pills and Dad did the pickup. His insulin resistance became something I’ve never seen before - just ugly. It came on slowly - I was on watch for the first couple of days of medicine and he was fine so I forgot that might be an issue. I attributed the resistance to illness and then hormones, until it dropped a few days after finishing the medicine. <----that was hell on earth particularly while trying to figure out Loop settings and not realizing we were fighting factors way out of our control.

He leaves for Italy in three weeks. He’s open Looping during the day and we’re getting our confidence back with settings. Closed looping at night is working well, but we need to test more traditional Italian cuisine before he leaves.

It’s been a rough few months and I’m counting the days until July 10th - when his :airplane:️ lands in the USA.

Great question. I can’t say I know the answer for sure, but I think it has something to do with artificial sweetener. When they came home with it, I called the pharmacist to find out sugar content. I was told it couldn’t be quantified, but it was very little. I was also told there were other flavor/sweetener ingredients that were sugar feee although non-specified. Caleb has always shown crazy results from certain artificial sweeteners - sorbitol drives his blood sugar through the roof and it does not easily correct. I’m scarred from a skinny cow experience we had years ago. This was much like that.

I will never ever allow him to have liquid meds again. Illness usually means we are fighting lows. I’m confident the drugs were the the cause of his insulin needs nearly doubling during that time.