rickst29
06-24-2006, 05:11 PM
My personal experience over the last 5 weeks (a couple of sensors at the official 3-day use, then one used for 12 days, then one used until it died at day 17):
'Accuracy', treating the OT as 'correct', is so good that I've basically stopped using my Ultra-- the instructions are "confirm your Dexcom reading with a finger-stick test before treating your bG" but I've found the treatment choice indicated by the Dexcom to be 100% reliable over the last 6 weeks.
But remember: I'm now treating the trend, not just the bG. If it's going down sharply from 130, I'm 100% confident in the "going down" part, and need to eat something. The "130" reading is usually 15-20 minutes behind my bG, and even after "extending the curve" in my mind, only accurate to about ±15% (1 standard deviation, 68% of readings within this range).
Although I can "extend the curve" in the graph to make a guess at my "real" bG, I don't bother anymore-- I just "eyeball" the graph, and the current reading, and know what (if anything) to do. Accuracy below bG 80 is a lot worse, maybe ± 30 points, but that hardly matters... you already have your diagnosis, eat sugar tabs NOW. Once in a while, you'll accidentally be pushing your bG up from 110 or so to 140 or so... but that's not a disastrous mistake, just a small one.
There's a special condition where the Dexcom can't be trusted: exercise which **MAKES YOU SWEAT A LOT, RIGHT UNDER THE SENSOR** can send it totally to heck. (It may read 230 while your real bG is 120.) Never trust it at these times, and absolutely never calibrate it under these conditions... you'll mess up the readings for the next 8 hours if you calibrate against a sweat-polluted Sensor before it dries out. (I live in the Nevada Desert, right now the humidity's about 5%, sweat doesn't last. But for most of you, this CAN be a problem... heavy exercise is exactly the time when you most want your CGMS working.)
When I'm sleeping, relatively motionless, my ISF readings have a poorer correlation with bG. Since I've never bothered to wake up merely to 'test against the Ultra', I don't know for sure. But I have found that you MUST NOT sleep with the Sensor squashed against the mattress.
Accuracy is highest shortly after calibration, then declines until the next calibration re-corrects the growing divergence. Brand new Sensors are less accurate in the first 8-16 hours than they are in days 2-3 (or, as I use them, in days 2-16).
The official Clinical study documentation provided on pages 88-102 of the User's Guide (wow, LOTS of graphs and tables) differ from mine in one key respect: The compare Dexcom versus YSI readings from blood draws at the exact same time, while I "extend the curve" and compare against a projected" Dexcom reading for 15-20 minutes later (3-4 data points after the blood draw).
They got ±21%, not using my time-correction adjustment, which is still VERY good. But of course, the percentage error dramatically error increases at low bG, and they had more data points at YSI readings above 121 than below 120. Most "regular" meters also de-emphasize the number of data points taken in the critical sub-70 bG range, IMHO making them look better than they are.
Here, Dexcom's study included 140 YSI-paired readings in the 40-80 range, out of 640 total, a FAR LARGER proportion than studies I have seen for other devices-- particularly the 2005 study of Glucowatch versus 'old' Minimed CGMS versus 'new' Minimed CGMS. Unlike others, Dexcom has the guts to conduct their study in a way which shows how they do in the problem area, the Hypo bGs readings.
As I did, they saw a decline in Accuracy as time goes by after calibration. During hours 9-12, compared to hours 0-3, 5% more of your readings will fail to match One-Touch by a specified percentage 20%, 30%, or 40%. (For example, the percentage of readings which lie within 30% of a One-Touch reading taken at the same instant declines from 79% to 74%.)
In indoor non-exercise conditions, you'll probably match the study results (+/- 21%) if you don't "extend the graph" by 15 minutes before making your comparison. If you "extend the graph", as I do, and use good site locations, and only count on / calibrate Dexcom when the Sensor isn't sweaty, then you can realistically expect ± 15% for actual bG above 100, and ± 25 points for actual bG 50-80.
Almost 288 readings per day, even while you're sleeping, at +/- 15% or 25 points. You can treat the trend, instead of having to work from a couple of "isolated" bG readings. Except for the cost, I think that this is a NO-BRAINER for anyone with Hypo problems or A1c's worse than 6.5.
'Accuracy', treating the OT as 'correct', is so good that I've basically stopped using my Ultra-- the instructions are "confirm your Dexcom reading with a finger-stick test before treating your bG" but I've found the treatment choice indicated by the Dexcom to be 100% reliable over the last 6 weeks.
But remember: I'm now treating the trend, not just the bG. If it's going down sharply from 130, I'm 100% confident in the "going down" part, and need to eat something. The "130" reading is usually 15-20 minutes behind my bG, and even after "extending the curve" in my mind, only accurate to about ±15% (1 standard deviation, 68% of readings within this range).
Although I can "extend the curve" in the graph to make a guess at my "real" bG, I don't bother anymore-- I just "eyeball" the graph, and the current reading, and know what (if anything) to do. Accuracy below bG 80 is a lot worse, maybe ± 30 points, but that hardly matters... you already have your diagnosis, eat sugar tabs NOW. Once in a while, you'll accidentally be pushing your bG up from 110 or so to 140 or so... but that's not a disastrous mistake, just a small one.
There's a special condition where the Dexcom can't be trusted: exercise which **MAKES YOU SWEAT A LOT, RIGHT UNDER THE SENSOR** can send it totally to heck. (It may read 230 while your real bG is 120.) Never trust it at these times, and absolutely never calibrate it under these conditions... you'll mess up the readings for the next 8 hours if you calibrate against a sweat-polluted Sensor before it dries out. (I live in the Nevada Desert, right now the humidity's about 5%, sweat doesn't last. But for most of you, this CAN be a problem... heavy exercise is exactly the time when you most want your CGMS working.)
When I'm sleeping, relatively motionless, my ISF readings have a poorer correlation with bG. Since I've never bothered to wake up merely to 'test against the Ultra', I don't know for sure. But I have found that you MUST NOT sleep with the Sensor squashed against the mattress.
Accuracy is highest shortly after calibration, then declines until the next calibration re-corrects the growing divergence. Brand new Sensors are less accurate in the first 8-16 hours than they are in days 2-3 (or, as I use them, in days 2-16).
The official Clinical study documentation provided on pages 88-102 of the User's Guide (wow, LOTS of graphs and tables) differ from mine in one key respect: The compare Dexcom versus YSI readings from blood draws at the exact same time, while I "extend the curve" and compare against a projected" Dexcom reading for 15-20 minutes later (3-4 data points after the blood draw).
They got ±21%, not using my time-correction adjustment, which is still VERY good. But of course, the percentage error dramatically error increases at low bG, and they had more data points at YSI readings above 121 than below 120. Most "regular" meters also de-emphasize the number of data points taken in the critical sub-70 bG range, IMHO making them look better than they are.
Here, Dexcom's study included 140 YSI-paired readings in the 40-80 range, out of 640 total, a FAR LARGER proportion than studies I have seen for other devices-- particularly the 2005 study of Glucowatch versus 'old' Minimed CGMS versus 'new' Minimed CGMS. Unlike others, Dexcom has the guts to conduct their study in a way which shows how they do in the problem area, the Hypo bGs readings.
As I did, they saw a decline in Accuracy as time goes by after calibration. During hours 9-12, compared to hours 0-3, 5% more of your readings will fail to match One-Touch by a specified percentage 20%, 30%, or 40%. (For example, the percentage of readings which lie within 30% of a One-Touch reading taken at the same instant declines from 79% to 74%.)
In indoor non-exercise conditions, you'll probably match the study results (+/- 21%) if you don't "extend the graph" by 15 minutes before making your comparison. If you "extend the graph", as I do, and use good site locations, and only count on / calibrate Dexcom when the Sensor isn't sweaty, then you can realistically expect ± 15% for actual bG above 100, and ± 25 points for actual bG 50-80.
Almost 288 readings per day, even while you're sleeping, at +/- 15% or 25 points. You can treat the trend, instead of having to work from a couple of "isolated" bG readings. Except for the cost, I think that this is a NO-BRAINER for anyone with Hypo problems or A1c's worse than 6.5.