Reading in the 50's while 80/90's is pretty much in range....
No. Dexcom accuracy starts REALLY going to heck under 60 mg/dL (e.g., might say "58" when you're only "48", might blow a "below 55!" alarm when you're really at 66 mg/dL) -- but it should not EVER be sounding "below 55!" alarms when you're in the 90s. (Well, OK, the mathematical possibility of such reports from a properly working Sensor *is* well above 0%, but it's not at all frequent. Under proper conditions, with a properly operating Dexcom, I'll SWAG each "below 55!" to have less than a 5% chance of being sung-out so badly off the mark, when "true" bG is actually above 90 mg/dL.)
There’s a number of posts in this Thread about calibration, and I have minor disagreements. For your consideration, here’s my own opinions/rules. (There’s six of them; as Helmut says there’s LOTS of considerations in doing a really good job.) Remember, I am not a licensed medical professional of any kind, and these are entirely unqualified personal opinions from a LAY PERSON: You use them entirely at your own risk, unless explicitly approved by a member of your QUALIFIED Medical Treatment, or by Dexcom. Dexcom in fact disagrees with me on some of my opinions!
(1) Never, ever calibrate during a period when bG is RISING, even a little bit. Dexcom readings are much more “late” on rising bG than falling bG. (Calibration during a time of “moderately” falling bG, a fall rate of less than 1 mg/dL per minute, is OK.)
detail: This is because the two delays of “First, higher bG soaks into ISF”; and “Then, glucose crosses the Dexcom membrane in sufficient quantity to show the change within it’s glucose oxidase” are additive, making Dexcom readings both “late” AND “later”. But in falling bG, insulin drives glucose across cellular membranes FIRST, and mostly from ISF. Then the drop in bG, “backfilling” from blood into into ISF, begins happening after ISF glucose had already fallen to a lower level. In this case, the delay across the Dexcom Sensor’s surface and the the delay between ISF and bG glucose help the Dexcom to look more like bG, with less “delay”. (The two delays are counter-balancing each other.) The “success” of this counter-balancing effect varies with each person, the rate of bG drop, and the “freshness” of ISF concentrations surrounding your site. (If you’re exercising, with a high heart rate, ISF is getting “refreshed” to match blood more quickly.) For me, in normal daytime activity, they’re almost exactly a tie-- moderately falling bG matches has the Dexcom readings matching my One-Touch fingerstick with no delay at all, with nearly the same variance as steady-state readings. Lucky me! But YMMV, And even I have “late” Dexcom readings on falls when I’m sleeping or otherwise immobile; my ISF becomes more “stale”.
(Whether ISF is a better indication than bG for monitoring cellular glucose starvation is an interesting question; I think it that it IS. So, in falling bG, it isn’t that Dexcom is “right” in almost matching the fingerstick value-- at the cellar level, you’re hurting. The almost-matching bG and Dexcom values are BOTH late in reporting the problem.
(2) Never, ever calibrate below 75-80 mg/dL.Just like other glucose-oxidase based measurement devices (including nearly all fingerstick strips and meters), Dexcom just doesn’t have a high enough raw data value to work from, accuracy becomes much worse. It’s much smarter to give it calibration values dispersed over the range 80-120 mg/dL and let it estimate lower values methematically-- entering “bad” data really screws it up, and even though the fingerstick is done carefully, Dexcom’s internal loss of accuracy makes the data "BAD!"for calibration purposes.
Think for a moment about the fingerstick accuracy standard "20%, or 20 mg/dL for bG readings below 100 mg/dL": At 100 mg/dL, that's a standard deviation of 20%, but at 50 mg/dL, that's a standard deviation of 40%. At 20 mg/dL, the standard deviation is as wide as the value of the reading! CGMS devices have the same challenge, but with the extra challenges of "it's not really capillary blood; and "ISF becomes more stale or fresh for many reasons; and most of all, "The Sensor isn't fresh out of the bottle, it's already polluted by previous sampling!" Because of these challenges, CMGS accuracy at low-normal and Hypo readings falls of much more badly than fingersticks. (3) Be sure to enter calibrations across your full range of "typical values". For optimal results, IMO, you want to have your recent calibration set include a reading from "normal/slighty high' bG (100-110 mg/dL); a "slightly low" reading (just above 80 mg/dL, a bit low but NOT breaking my rule #3); and a "high reading", right around the value of your High Alarm setpoint.
Only read this if you're interested in all the nasty mathematical and chemical issues behind my recommendation; if you don't LIKE math, it could hurt your eyeballs. ;) Think of CGMS calibration curves as a straight-line graph. (In fact, Minimed's microcode clearly is a straight-line graph, but because Dexcom treats their "ISIG-like" raw data value as proprietary data, I don't know. I have seen enough Dexcom raw data to know that the behavior is vastly different than Minimed's, with much higher "background" levels and, overall, a much smaller "slope" for a given bG change. It might truly be a curve, and not just a straight line; I've don't have enough raw data numbers to say. And even if I DID have an adequate data set to amke this determination, I would respect their choice of keeping the raw data private.) Now back to the gory details: Obviously, if you enter your "startup" values, for example, "105" and "103" mg/dL, your CGMS has good calibration data for readings right around that value--
But the “slope” of the curve? It’s got absolutely nothing to work from, it’s an almost total guess! In order to define a line, as Minimed would, or ANY kind of “curve” at all, you need to enter at least one good calibration value at a significantly different bG.
When my initial entries for a “fresh startup” are that “good” (grrrrr, hardly ever happens), I actually eat two sugar tabs to FORCE my bG up to about 120-130, in order to add a “high bG” calibration value right away. (Glucose tabs because they work fast and consistently, with no unpredictable “tail” of rising bG after about 30 minutes.) Remember rule #1 first, and be absolutely sure that the rise is Finished before entering a value!) Then I’d correct aggressively, enough insulin to push my bG down to 70 mg/dL at +3 hours, right after making that entry.
(My motto is “insulin works slow, backfilling works fast”. While wearing a CGMS Sensor you trust, it’s smart to correct aggressively-- if you absolutely, positively won’t fail to keep your eye on your readings. Just temporarily raise the adjustable low alarm setpoint, so that you catch the intended fall with plenty of time for “backfill” to reach your bloodstream well before your ISF and actually reach Hypo levels. (In this case, I’d set the low alarm to it’s max value, 100 mg/dL.) Obviously, having targeted 70 mg/dL at only +3 hours, and “risking” a sub-50 reading before the correction were to finish, l would be eating considerable “backfill” UNLESS the following hours ended up with VERY unexpected behavior.
Don’t do these aggressive corrections unless you are totally, absolutely certain to catch the drop at a realtively high bG reading, with plenty of time left to digest “backfill” before going hypo! If my initial calibration had been on the high side, I’d correct and wait for the drop to occur AND level off to less than 1 mg/dL, with no “backfill” food on board, before doing an entry. Typically, 4-5 hours later.
For me, a “perfect calibration set” has a recent number from the 80-85 mg/dL range, a number around 110-115, and a number around 140. (which is actually well above my “high alarm” setpoint of 120 mg/dL, but my Dexcom is set to blow a lot of High Alarms. 140 readings aren’t at all ucommon for me.)For Jan’s child, a “perfect calibration set” will need a number or two well above that, at least 180-200 mg/dL. And MAYBE even a bit more than that, although extreme numbers (250+) should not IMO ever be used: such extreme numbers distort the curve a lot, and in such a bad Hyperglycemic situation, you should be using a sequence of fingerstick tests anyway.
If you think about it, this already implied my next Rule. But for clarity, I’ll state it anyway:
(4) It’s not the RAW NUMBER of calibrations per day; it’s the quality of the “recent calibration set”.
IMO, Anyone who says things like “You should never calibrate, except when asked for”, or “Dexcom gets confused by too many calibrations”, before having been told (or seen from your data upload, or graph) what the recent calibration values actually were, is making a recommendation without adequate background information. (And therefore, very likely to be making a BAD recommendation.)
There is only ONE way in which extra calibrations become a bad thing-- that’s when new, redundant readings push the “weight” of other readings down via being “old” – and those deeply-discounted bG readings were at non-redundant bG levels, necessary to set a good curve.
I recommend that you want to have at least two “high quality” calibrations during the preceding 12 hours, separated by at least 30 mg/dL from each other. (“High Quality” requires proper handwashing and fingerstick technique, PLUS not taken within 15 minutes rising from bed, PLUS not taken during rising bG.) That’s not merely twice the rate of Dexcom’s default schedule, it adds the “significantly different” requirement necessary to set a proper “curve”, rather than merely provide redundant information about a pretty-much identical bG- you might need even more than 2 fingerpokes per twelve hours, in order to get a non-redundant reading on board.
This leads directly to my rule 5 (like Rule #4, an obvious corollary from the preceding Rule):
(4) Start-Up days should usually be given have EVEN MORE pokes.
Since start-up bG readings are taken within just a few minutes of each other, they’re gonna be at pretty much the same reading: The real reason why Dexcom asks for two, I think, is to insist on making you give yourself a sort of “double-check” on your technique and strips; if they’re AREN’T close, something’s wrong. I feel that it’s always wrong to leave Dexcom running for twelve hours without getting that third, different data point into the initial calibration calculation.
But in the case of a true FIRST TIME start-up, (which is probably a majority of start-ups for everyone here except me), you also flying blind due to the actual break-in behavior of a newly-injected Sensor. First day accuracy is always bad, and needs even more calibrations.
Why? Several reasons. First, for almost everyone, there is at least a small reaction to the chemical presence of the Sensor; after the initial response, this "stabilizes" over many hours. Second, for absolutely everyone, unless you're dead ;) , there is a reaction to the physical damage of the violent "punch-in" of the wire. This also calms down" over time. Your body cells are probably also reacting to ultra-tiny manufacturing variances (scratches) on the wire surface itself. (Please do understand that I'm talking about scratches which are MUCH smaller than the intended holes for ISF contact; most of these are within the EDGES of those same holes. Rather quickly, in my theory, your body clogs up these scratches in a strong reaction which then subsides, leaving holes at sizes which behave more consistently after the counterattack is finished. Finally, the internal Glucose Oxidase reagent has a "break-in" period too, consisting of both "initial fill" and surface property issues-- this "break-in" is NOT completely finished in only two hours.
My last Rule might surprise many of you, until I explain it (in terms of the previous Rules): (6) Never, EVER plug in a One-Touch-Ultra "calibration cable". Why? it's obvious, actually: Your meter is full of readings taken during rising bG; and readings taken during too-fast falling bG; and maybe an occasional "that was weird, I'll wash my hand and check it again... yep, I sure messed up on that first try!"
Dexcom has NO IDEA which readings are "good" for calibration, and which readings are "bad" for calibration. And remember-- every redundant calibration, taken within about 24 hours of it's almost identical predecessor, helps to push down the priority of every non-redundant predecessor. If there's a whole lot of these nearly identical bG readings within the last day, good for you (YAY, you're a flat-liner! Unlike an ER pt., that's a good thing when you're wearing Dexcom.)
Having too many recent and redundant calibration entires in the receiver will cause even a BRAND-NEW entry with better variance to be discounted; there’s simply too many, everything gets discounted. When you control bG data entry, as I recommend, you simply skip entering 3rth 4th and etc. “nearly-redundant” values which you’ve obtained via fingerstick in the last 12 hours. If you plug in the meter, though, you’re in a world of hurt, calibration-wise, for values far from the “redundant” fingerstick bGs.
Throw those cables away, I say. Even if you’re insurance provides OTU strips and meters, throw out the "One Touch Ultra (or Ultra-II) Calibration Cables. Always enter calibration values from the menu.