Standard Deviation

I’ve had more exhaustion than usual all summer. The new Minimed Carelink USB device has made it a lot easier to look at my numbers on a regular basis. One thing that stands out is extremly high standard deviation (SD) - over the last two weeks, my SD hit over 100 once and was 79 on average. It’s always been higher than I want - but 50-60 is more typical for me. I know this is a measure of my daily swings from high to low. My suspicions for the swings are usually:

  • I’m just brittle
  • Scarring under the infusion set - inconsistent insulin absorption
  • bad insulin
    The problem with these suspicions is that there isn’t much that I can do abut them, and the insulin is rarely bad. How can I change diet, exercise or medication to improve this number? Any experiments, regimens, or tests you can recommend? What should I be looking for?
    8505-1217275514010.pdf (27.4 KB)

I’m just beginning to look into standard deviation-- so I can’t really recommend what to do! But I’m curious to see the other responses…

My endo is particularly concerned with my SD. She said that that’s a better indicator for your control than your A1C because it considers range, rather than average - although, trust me, she looks at my A1C plenty. My last A1C was 6.9, but my SD was 62. She’d like the former below 6 (wouldn’t we all) and the latter below 40.

What you have to watch more than anything else are post-prandial spikes. You should know what your target BG values are 1 hour and 2 hours after a meal. I check at one hour and try for 140, at 2 hours, 120. That’s a little high, but because I’m getting ready for pregnancy, my diabetes team wants to prepare me for the extreme lows of the first trimester, so we’re starting a little above average. I believe that you’re usually shooting for closer to 120 at 1 hour and 100 at 2 hours. But these are just my targets, of course - not necessarily yours.

To control those spikes after a meal, you should consider the glycemic index of the food you’re eating and how long it will take the food to digest or convert to glucose. You should also play around with bolusing/injecting at a given length of time before a meal and racing the peak of your insulin just ahead of or on top of the peak of your food. I’m still playing with both of these things. I find that a lower carb count at dinner has helped me with post-dinner spikes. A lower insulin-to-carb ratio at breakfast and a higher basal rate from 3-7am also helps me with my mid-morning post-brkfst spikes.

That’s great advice. I don’t have post-prandial targets, but I imagine I’m at about 220 at 1 hour and 150 at 2 hours. I guess I need to do some experiments with the timing of the dose. I think my carb-to-insulin ratio is OK because after 4 hours I’m back to 100, but I can double check that too. Thanks!

I woke up this morning thinking about SD again. And I went to Wikipedia to answer my questions. I haven’t done calculations involving statistics in years and the formula was a beast, so I left Wiki feeling fuzzy. With the help of this standard deviation calculator, here is a real world application to demonstrate why the docs are interested in this number.

Let’s figure the SD for three people’s daily BGs.

Person 1: (85, 100, 102, 105, 120). avg: 102.4, range: 35mg/dl, SD: 11.18
Person 2: (56, 60, 87, 100, 209). avg: 102.4, range: 153mg/dl, SD: 55.78
Person 3: (100, 101, 103, 104, 104). avg: 102.4, range: 4mg/dl, SD: 1.62

These three people have an avg blood sugar of 102.4, so they would likely have the same A1C. But the SD tells us more about who has the tightest control - who has the majority of their numbers nearest their average.

My standard deviation averages around 40-50. I know that my Endo wants it to be lower because it really tells how well you are managing your BG. Suggestions I have would be to make sure your basal rates are set up properly. Bolus for everything that you eat and that means everything!!! Make sure that your bolus is based on correctly counted carbs and portion size. Under bolusing is what gets me at dinner time. Buy a scale that has grams on it for exact measuring. If you like pizza dual wave bolus 60% up front 40% insulin on the back end. When I eat pizza my BG will be fine and then 6 hours later it will shoot up to 250. Now that I know this happens I am better able to control my BG so this does not happen.

I looked at your data. The days that your BG is the highest are the days your carbs are typically the lowest. This tells me that you need to bolus more when you eat and to put the carbs into your pump. I see a lot of priming do you you really change your set that often? Unless it hurts I think you should be able to leave it in for 2-3 days. I use the 6mm, 23in quickset, my average for priming is around 9.5, sometimes 8 or even 10 but never less. You might want to make sure that you are getting all the air bubbles out of your tubing because that can lead to problems. Let me know if any of this helps or if you agree or disagree with what I am saying.


I agree with you, Dave, about the priming. That’s odd to see if more than once every few days.

I take the pump off for showering, and prime 0.4 units to see a bead of insulin before I reconnect. I use the Sure-T metal needle infustion sets, changed every 3 days. In these past few weeks I’ve been going into scar tissue a lot, so I’ve changed early a couple of times. But that’s unusual.

I’m guilty of not bolusing for every single gram of carbohydrate. For example, I’ll have 10g carb as yogurt before I walk the dogs for 40 minutes, and just not bolus at all.

I was curious what the effect of hormones was on high blood sugar. I’m an engineer, so that means an Excel graph! I attached a graph that seems to show that higher progesterone levels correspond with better control, over the last 2.5 months at least.
8507-Hi_BGs_vs_Hormones.bmp (912 KB)

Hard to believe how much worse I’m doing than Person 2!

I checked my standard deviation for a couple weeks. On “good days”, it’s 20-30… on bad days, it’s 50-70.

My average was also lower on the good days. For me, the two seem to be very related.

I agree this is an important measure (though I have never had a doctor ask about it!)… but I think that your standard deviation with vary HUGELY by how often and when you test.

For example, if I don’t test until two hours after I eat (which I often don’t), then I miss the spike, but my A1c records that… and my highs usually last longer than my lows (because I feel my lows but I don’t feel my highs).

So it seems that standard deviation and average should be considered TOGETHER. But usually what my endo did was look at my A1c and the number of more severe hypoglymic episodes (under 60). This gave her an indicator of my control and whether these lows were making my A1c look better than my control actually was. I guess this is a way of getting at the same concept??

I’m thankful for this discussion though-- I will watch my SD more carefully! I guess the SD tells a lot more if you have a CGMS and it is not so dependent upon when you test…

Do you use the super bolus technique? I do and I prime (“Fill Cannula” for Cozmo users) daily.

Try experimenting with the timing of the insulin. I give my bolus 15-20 minutes before eating, as long as I am not below 80 or have a lot of active insulin. Also, I have tried using dual wave boluses over 30-60 minutes and waiting even longer to eat… 30-45 minutes. This works well for me, but I know that there are many differing opinions and bolusing techniques!

When would you do this: “dual wave boluses over 30-60 minutes and waiting even longer to eat… 30-45 minutes”? That sounds very gutsy to me.

You’re not alone! I had A1c between 7.5 and 8.5 for my first four years of diabetes. I thought that my control was “OK”-- not horrible. But seeing numbers above 250 and even 400 was not rare. I have learned that careful monitoring, carb counting, bolusing and correcting can make those numbers rare… and I feel a lot better! Small changes can make a big difference (like waiting 15 minutes to eat…).

I do it anytime. Not only with high fat meals.

The reasoning (which not everyone agrees with-- and I’m not even sure that I’m right— and I wish that I had a CGMS so I could have a better idea):

I bolus for a dual wave bolus over 30-60 minutes (30 if I’m higher, 60 if I’m lower). This means that the effectiveness of the insulin is spread out-- so I won’t drop 60 points in an hour. Then I check my blood sugar and eat when I am under 100. If I don’t feel like checking (honesty here that I don’t always check), then I wait 30 minutes.

I check my blood sugar often after eating and I can eat a 50g meal without going over 120.

I don’t do this all the time (maybe only twice a week) because it takes planning, but as I try to tighten my control, I want to use it more often.

It is gutsy because you risk lows with this method. SO you have to be willing (and have enough test strips) to check 4-5 times when you are first using it. I find using the dual wave bolus helps me to prevent such sudden swings in my blood sugar that I get from giving a big bolus all at once.

Just the thoughts of a type 1 trying to figure out what the heck is going on… disagreement is welcome!!


Can you explain why I would want to go through all of that instead of just increasing my bolus amount up front and then setting a temp basal for a certain amount of time to compensate for the extra bolus? I may not really be understanding this technique so please explain.

My Dr. mentioned doing this as well. She said exactly what you wrote that you will have less spike and it will be more moderate. There are so many things to watch and think about but I should try this method more often myself especially since I have CGM. Thanks for reminding me of this technique. I will test it on my lunch today!!


You are exactly right, the effect is the same as the method you described. The reason for the super bolus technique is so that the basal insulin that you gave is NOT recorded as “insulin on board”. If you use your method, then when you correct an hour later, then the pump considers the basal that you gave upfront to be a bolus. BUT it is still basal insulin just given earlier and your correction bolus should not subtract out that part.

Maybe it doesn’t make much of a difference… but that’s the reason.

I’ve never checked my St. Dev.(Sx) but it always bothered me that the “all knowing” A1C is really just based on a mean and the mean is what my Endo looks at when I visit…not even a median. The mean doesn’t really tell you that much. What a great idea to use the Sx!
I’m on the Omnipod and I have the software but haven’t bothered with it yet because it is generally considered useless and they are working on new software that is supposed to be great.
I just used Excel to calculate my Sx over the past couple of months and it’s about 25. On good days it’s as low as 18 and on a bad as high as 35. I’ve never discussed this with my Endo. Exactly what does everyone’s Endo suggest is a “good” Sx? Does anyone have any idea what the Sx might be for an individual without diabetes? I usually push my Endo for this kind of info instead of the standard….”it’s good enough for you” answer.
I agree with David that a good scale and accurate carb counting are necessary for good control.
I also do an extended bolus (aka dual wave/combo) often. I think the success of this method is largely understanding your own metabolism and accurately recording everything you eat so you can check for patterns.
As for hormones….yes, they affect everything. I usually get slightly higher numbers at different times of my monthly cycle. There are a few conversations about that on here. The name of one is “bc” and bg control. I’m not sure how to link to it but you can do a search.
Have you checked your basals lately? These can need adjustment over time/hormones/age too. I would start there and then when your confident with them, adjust your carb:insulin ratio and experiment with extended boluses.

As far as standard deviations go I think my Endo told me that 30 and under is what she was looking for but I could be totally wrong. Lower is better that I do know! Infact I was just looking at my info on Carelink, using a finger stick my sdev is 37 over the last 2 weeks. My CGM (which I believe more) says that it is actually 30.