Scary 4am Low: A Cautionary Tale

I’m a little rattled this morning by the events that played out last night when I should have been sleeping peacefully.  I’ll do my best to convey them here now:

I went to a holiday party last night and both drank and a ate a bit closer to bed time than I like to; it seems the food in my stomach when I’m sleeping digests a bit quicker and I tend to spike a bit higher than I might if I didn’t go to sleep so soon after eating.  I also walked home from the party and was seeing double down arrows on Dex and a BG of 175 which from experience I knew would either level off at 175 once I stopped walking/went to bed, or would continue to raise one I went to bed and super-digestion commenced.

So, I went to bed last night expecting a bit of a BG-coaster but after a comedy of errors, what I got was so much worse.

Here’s what I remember:

Around 4am, I hear M asking me if he should get me some orange juice. I must have been feeling the low at that point because M usually responds only to my squirming and not to the sound of Dex. I also must have been pretty low because when he asked, rather than saying “no” like I normally would do and getting the juice myself, I said “please” and I put my head on the pillow and let the cold sweat drip over my skin.

He brought back a full, tall glass of orange juice (at lease 4x what I would normally use to treat a low) which I drank in two gulps.  He got a second glass that went down just as easily.  I lay there for a couple minutes to try and let the juice absorb when I decided to actually look at Dex.  58 mg/dL.  Not too bad. But 58 wasn’t really a good description of how I felt. I felt sub-zero. So I grabbed my meter and checked.  32.  That’s more like it.

I went into the kitchen and ate a few spoonfuls of pistachio Haagen-Dazs and Nutella.  M came to check that I was ok and laugh and my standing in front of the open freezer with a carton of ice cream and a spoonful of chocolate.  I went back to bed and back to sleep.

In the morning, I assumed I would be sky high but Dex was only reading in the 120’s.  A check with my meter confirmed it.

7 am BG after overnight low.

7 am BG after overnight low.

WTF? Where’s my sky high BG that should have followed my massive 4am consumption?  Looking at Dex confused me even more – this low occurred after a rapid drop in BG at 2am.  I must have awaken to the sound of Dex alerting me to a high and I must have groggily corrected my BG of 185 (as determined by Dex) and corrected with a bolus from my shiny new t:slim pump.  (Now, I know I shouldn’t be correcting from my Dexcom, but I do. And I’ve found it to be reliable enough and even when it’s not the outcome is generally not so far off to be a problem. Generally.) This is the sort of overnight correction I am in the habit of doing, and doing it is barely a blip on my radar and happens mostly on autopilot.  So, I decided to check my pump history to see if this was the case.

t:slim history screen showing a surprise.

t:slim history screen showing a surprise.

Looking at the above image, I was shocked.  I initiated a 15 u bolus at 2:24 in the morning.  Why?  Purely by accident apparently.  The t:slim displays 3 things for every bolus: the time at which it was completed (boluses can be a bit slow, in this case at 2:48 am), the time it began (2:42 am) and what was entered to initiate the bolus.  See in the picture it says “B: Food/BG” then “183 g/NA”?  This means I bolused at 2:42 am for 183 g of carbs rather than a BG of 183 mg/dL!  What should have been 2 u was 15 (max bolus allowed by my pump settings).

How can I make such a stupid mistake? I’ve only been using the t:slim for about a week and I love so many things about it, including that I no longer have to dial in  the units for a bolus.  However, dialing in that number has always provided me with a check that I couldn’t ignore before delivering a dose of insulin. The t:slim asks me 3 times before delivering a dose; it does a good job of shoving information in my face.  What it doesn’t to is engage my brain enough during autopilot to help me see the error of my ways.

This is what you see when you initiate a bolus with the t:slim

This is what you see when you initiate a bolus with the t:slim

The bolus screen clearly indicates where to enter a BG and where to enter carbs.  However, once you click, the next screen is nearly identical for each:

BG or carb keypad

BG in carb windowBG in BG window Keypad; BG in carb window; BG in BG window

Once the numbers are entered, it displays a calculated dose then you hit “next”, confirm twice, and done.  The bolus initiates and the pump vibrates quickly upon completion.   With auto-pilot initiated, I entered and confirmed and was asleep long before the vibration.  End of story.

So.  What have I learned?

  1. The obnoxious dial-a-bolus method used in all other pumps may have irked me most day,s but it did provide me with a mental check that probably would have prevented this from happening.
  2. Actually using my meter to confirm my Dex BG would probably have taken my brain far enough out of autopilot to notice my mistake and prevent this from happening.
  3. For some reason, my first instinct when I go to calculate a bolus is to click on the left most box first, and when my brain is in BG mode instead of carb mode, there is potentially a problem.

I started with the t:slim about a week ago.  While I am still smitten by it, I’m realizing I need to approach this change with a bit more caution that I have been.  I’ve had Animas’ pumps for about 12 years and I can use one with my eyes closed.  I’m going to have to spend some time focusing more on D and destroy my old auto pilot (or at least reprogram it).

The impact of 20%

There’s been a lot of discussion in the DOC about BG test strips.  There’s the strip safely campaign focused on making sure all test strips for sale in the US are subjected to the same rigorous quality control measures to ensure optimal accuracy; many discussions over whether or not 20% error allowed by the FDA is good enough, and countless user experiences.

So, from where does this problem stem? And how big of a problem is it really?  The FDA has imposed an ISO standard on glucose monitoring. The standard is this:

ISO 15197 specifies that

> or =95% of the BG results shall fall within +/-15 mg/dL of the reference method at BG concentrations <75 mg/dL and within +/-20% at BG concentrations > or =75 mg/dL

This means that for BGs less than 75 mg/dL, our meters spit out results that are up to 15% off from reality 95% of the time, and when our BGs are over 75, they can be up to 20% off from reality.  Compliance to this standard is often shown via a Clarke error grid, like the one below (for freestyle test strips).

Clarke error grid for Freestyle Lite test strips. From

As you’ll notice in the picture, the black dots are all within the region labeled A which is the 15-20% bracket.  You’ll also notice that while the 20% range seems huge for larger BG values, this window is actually quite small for lower BG values.  Also, for any given plasma glucose value, there is a normal distribution of BG values from 0% (completely accurate) to ±20% (with up to 5% of numbers being more than 20% inaccurate). I think it helps to visualize wha this really means by looking at this generic normal distribution:

Normal Distribution: For our purposes, imagine the horizontal axis labeled with -20% on the left where it says “-2σ” and +20% in place of “2σ” on the right. The height of the peak (centered at 0%, “µ”) is determined by the number of BG values that are of by a given percentage.  Notice that 68.2% of your BG values will be in that dark blue region with errors far less than 20%.

Imagine that you kept a running list of the error of every one of your BG meter readings compared to your real plasma blood glucose.  Then you graphed this data with the %error on the horizontal axis and the number of reading per % error on the vertical axis, you’d see a distribution very much like the one in the picture.  Replace the “2σ” and “-2σ” with +20% and -20% and the “µ” with 0% on the horizontal. (I haven’t yet found reliable data to label the y axis with, but it would likely vary from brand to brand and I think it might be reported on that little piece of paper we always throw away when we open a new box of strips.) The light blue tail regions represent the 5% of numbers that are outside of the 15-20% BG allowance and the other 95% are in the two darker blue regions with most being concentrated nearer to the middle which represents your actual BG level.

With that in mind, let’s now consider what effect the FDA regulation has in practice. What I’m really interested in is how this error translates to errors in my diabetes regimen and the impact on my BG/A1C.   Below, I’ll walk you through my calculation of insulin doses delivered based on meter reading and the resulting corrected BGs.  To simplify, I calculated using a 20% error for all BG values from 40-400 mg/dL with a target range of 70-120 mg/dL.

Fisrt, let’s look at the range of meter readings you might expect compared to your real BG.  Keep in mind that these lines contain 95% of all readings, with at least 68% of those readings concentrated closer to the dotted line than either solid line.

Meter Readings 20% accuracy

For every real BG value, your meter is calibrated to correctly identify that value with up to a 20% error. On the horizontal X axis are the real values and the green and red lines indicate the possible range of values returned by your meter.

Using my insulin sensitivity factor, I then calculated the dose of insulin my pump would deliver in response to the meter readings shown above.  Again, remember that numbers are concentrated more closely to the dotted line than either solid line.

Insulin Dose

The error in the meter readings translate to an error in correction dosing. The dashed line shows how much insulin is needed to correct a given BG value to the nearest target (omitting insulin to reach the bottom of the range or dossing to drop to the top of the range). The red and blue lines correspond to correcting a meter value that is 20% above or 20% below the actual value.

Assuming no other influences (ie spherical diabetes in a vacuum), the next graph shows how my BG will respond to getting either the min or max insulin correction.  These represent REAL BG values resulting from inaccurately dosed corrections caused by the 20% error in meter readings. The space between the lines contains 95% of all possible outcomes.

Corrected BG

The range of correction doses lead to a range of corrected BG values. The largest discrepancies occurring alongside the largest real BG values.

And lastly, I converted the above graph into percentages (because I like them).  You’ll see the maximum percent over the target range (red, from under correcting a high or over correcting a low) and under the target range (orange, from over correcting a high or under correcting a low).  This isn’t the % of resulting BG values outside of the range, just the % off from the target of a single BG resulting from inaccuracy in meter readings.  And again, the space between the lines contains 95% of all possible outcomes.

The % above or below the target range of the corrected BG.

The % above or below the target range of the corrected BG.

I’m going to say it one more time because I think it is very important.  These pictures represent the extremes – 95% of BG values are 20% accurate or better, with 68% of the values concentrated closer to reality.  This then means that all of my calculations represent ranges of data with a similar distribution.

So, there it is.  In pictures. The direct impact of the 20% FDA-approved error on our BG values (in a vacuum).

And now for my opinion on the subject: Personally, I feel like it’s just not that big of a deal.  Since I am more likely to treat a low with a standard regimen of 15-30g of carbs and protein, the under correction there is mostly irrelevant.The possibility of over or under correcting a high over 240 is very real but since I’m already in the habit of double checking higher numbers (making sure there’s no rogue sugar on my finger tips) I don’t see that as a huge problem either.  Plus the introduction of a second BG reading is enough to assure me that I’m comfortably in the 68% of readings that are far less than 20% off and not on the outskirts.

Here’s a copy of the excel sheet I used to generate these pictures.  Have fun  if you like that sort of thing and please let me know if you disagree with any of my calculations (or my opinion).

Here are a few other articles and opinions on the 20% issue:

About meters being outside of the FDA limits

Opinion and lots of facts about the FDA and meter accuracy

What the FDA regulations mean to diabetics – blog

This is goiong to sound stupid…

Recently, my insurance company shifted my OneTouch Ultra test strips to the top tier of my prescription coverage (meaning that they went from being $60 for a 3 month supply to $150!) and left me with Freestyle Lite strips as my only affordable option.

At first I was angry.  I use an Animas/OneTouch Ping and I often bolus from the remote, which used OneTouch strips exclusively.  I wined and complained and managed to get a free meter (or two) out of the deal, but ultimately, I have had to make the switch to using Freestyle.

It has been a couple months now and I have gotten used to dialing in my BG for correction boluses again.  I have even found the Freestyle meter to have many redeeming qualities; the flashlight that illuminates the test strip is nothing short of the best idea ever, the strips (which fit 50 to a vial as opposed to OneTouch’s 25) barely need any blood at all, and the strips have a cute little butterfly on them.

However, there is one thing that continues to bother me about Freestyle:  After each BG test, I grab the strip with my teeth to pull it out of the meter and suck off any excess blood.  (Does this make me a vampire?)  Then I drop both the strip and the meter back into the little owl coin purse from which they came.

The problem is this: Freestyle test strips taste funny.  There.  I said it.  They have a weird chemical taste that OneTouch is lacking and I hate it.  I hate it more than I love the strip illuminating flashlight.  More than the cute little butterfly adorning each strip.  More than the custom skins I can decorate my meter with.

I hate the taste of Freestyle  strips.

Hello, Goodbye

Apidra week from Hell - normally these range bars are short and close(ish) to the green happy-zone. 😦

First a little background as to why I decided to try out Apidra:
I’ve mentioned before that I participated in Joslin’s Why WAIT? program. It was accomplished through a combination of increased exercise, decreased food intake, and the addition of Symlin to my D-regimen. These things combined to cut my total daily dose of insulin in half, and then some. Originally the team had wanted to switch me from Humalog to Apidra because it peaks faster and would theoretically have made blousing with Symlin a bit easier. But, I had 10 vials of Humalog in my fridge, no real job, and switching just wasn’t an option.

Fast forward to last month, about 4 or 5 months post Why WAIT?, to my endo appointment. I approached my endo about switching to Apidra since my humalog supply had dwindled, and, although she wasn’t sure the benefit would be significant, she didn’t see any reason not to try it. So, I walked away with a prescription for a three months supply of Apidra.

After some trouble getting things straightened out with my new online pharmacy, I finally received the Apidra two weeks ago and began using it last week, on Feb 28 at 7pm. And then it began.

Day 1: Hello Apidra. My BG is running high all day; I wake up with a BG of 120, eat nothing and by the time I’m at work an hour later my BG has jumped to over 250. I do correction bolus after correction bolus to no avail and spend the entire work day over 250, even after eating a carb-free salad for lunch. Promptly at 5pm, when I normally head home from work, all of the boluses finally catch up to me and I drop almost instantly to about 55. I correct and make it home with a BG hovering around 100. Dinner is chicken and brown rice. My BG stays flat until bedtime but as soon as I fell asleep, Dex begins to wail. I probably test and correct 5 times overnight and my BG stays flat around 200 all night until an hour or so before dawn when it levels off around 100.

Day 2: Site change in AM. Repeat of day one.

Day 3: After the last 2 days of BG hell , I increase all rates and ratios by about 10% (except the post work/before bed ones) and shorten the IOB time on my pump so I can bolus more aggressively. I also add Symlin to most meals. My BGs level off a bit but they still sit well above my happy zone all day and require frequent corrections that seem to have no effect. I go low overnight.

Day 4: It is like Day 3 never happened and is a repeat of Day 1, again. (Only I have a low-carb, high fiber dinner that send my BG up to over 400!) In the middle of the night I have a low that requires 3 juice boxes and I still wake up the next day with a BG of only 75.

Day 5: Increase all rates again. No change. At this point I realized that I should be monitoring for ketones and luckily find none.

Day 6: Realize that I can’t fix this problem on my own and call Joslin for help. I upload my pump and Dex data from the last week and send it after work. At this point the 5+ days of highs have given me a terrible headache that I can’t quite shake. I officially hate Apidra. However, having a carb-free dinner has allowed me to sleep an entire night without any complaints from Dex. (Although he caught a low before bedtime that I corrected with ice cream perfectly – flat at 100 all night!!)

Day 7: After talking with the nurse educator and experiencing my 8th BG over 400 in less than a week (which is a number I haven’t hit in years!), we decide that I will switch back to Humalog immediately after work today.  Good bye Apidra.  You will not be missed.

So, unless this was some sort of ill-timed, un-related BG surge (which I doubt), everything should be straightened out soon.  And I can’t wait.

However, I now have 5 bottles of Apidra that I don’t know what to do with.  does anyone know of a place to “sell” (It would be great to get some part of the $60 copay back) or donate unopened bottle of insulin?

The science behind the glucose test strip

There’s a lot of talk in the DOC about making a more accurate BG test strip. We want a test strip that is inexpensive with as little error as possible. That’s not too much to ask. Or is it? I decided to look into how current BG strip technology works and do a few basic calculations to see just how big of a difference there is between blood with a BG of 40 verses that with a BG of 300 mg/dL.

Test strips contain two chemicals; an enzyme called “glucose oxidase” which is transforms glucose into gluconolactone, and a “mediator” (typically ferricyanide) which participates in the reaction and produces ferrocene and releases electrons.  The electrons make up the current that the meter then interprets as a BG value.  When there is a higher concentration of glucose in the blood there will be a higher concentration of electrons produced which will cause an increase in the current sent across the electrode on the strip, and a higher BG reading is obtained.

Other compounds which are often found in the blood also transform the ferricyanide into ferrocene and release electrons.  (Such as acetomenophen, vitamin C, and uric acid.)  To prevent this from skewing the test results, each test strip has two working electrodes (plus a reference).  One working electrode has both of the chemicals I described above, and the other has only the mediator but no enzyme.  Thus both electrodes measure the current that results from acetomenophen, vitamine C, and uric acid, but only the electrode with the enzyme also measures the current produced by the glucose reaction.  This allows the meter to subtract out the current that results from these other compounds in the blood.

Current meters can measure as little as 36,000 molecules of glucose in 30uL of whole blood (uL = microliter, most strips use about 3uL).  A single gram of glucose contains around 3,400,000,000,000,000,000,000 molecules (that’s 3.4×10^21).  After thinking about all of this, I began to wonder just how different 40 mg/dL blood is from 300 mg/dL blood.

I decided the best way to compare would be to figure out what mass percent these numbers translate to.  I found online that the approximate density of blood is 1.05 g/mL which is equal to 1.05 mg/uL (which means that 1uL of blood has a mass of approximately 1.05mg).

A BG reading of 95mg/dL can easily be converted to mg/uL by dividing by 100,000.  So,  95mg/dL = 0.00095 mg/uL (I chose this BG value since it is within range of a nondiabetic person’s average BG thus the density of blood given online should be most correct at this level.) This means that in 1 uL of blood, there are 0.00095mg of glucose  (or 3.23×10^15 molecules).

To get a mass% (the percent of the total mass of the whole blood that is glucose) just divide the mass of glucose in 1uL by the mass of 1uL of blood: 0.00095/1.05 = .000904 or 0.0904%.

Using the same math, a BG of 40 mg/dL = 0.038% and has 1.36×10^15 molecules per uL, and a BG of 300 mg/dL = 0.285% with 1.02×10^16 molecules in 1uL.  These differences are actually quite miniscule if you think about it.

[It is important to note that outside of the range of a non-diabetic blood sugar level, the density is going to change.  At elevated BGs, the actual density might be higher, making the number calculated using the above rational somewhat elevated, and the low BG levels will appear somewhat lower than the actual value. However, with these tiny percentages, the change in density is very likely minimal and can thus be disregarded.]

In my experience as a chemist, I have found that analytical equipment varies in price depending on the level of accuracy.  For example, for every additional decimal place that a balance is capable of measuring, the price of that balance increases exponentially.  We are essentially asking these companies to produce a product that measures the concentration of the glucose in our blood to an additional decimal place.  I would be willing to bet that the technology already exists, but the cost is well outside of what any of us can afford to pay.

Before our meters become more accurate, one of two things needs to happen.  The materials needed for a more accurate test strip using current technology need to get cheaper, or someone needs to figure out an entirely new way of measuring glucose (or at least some new chemical coatings for the electrodes).

So, that’s the way it is.  Now, feel free to form whatever opinion you’d like 😉