Website helps predict blood glucose level

I’ve been a T1D for 30 years and used Dexcom for at least 5 years, but I’m about ready to quit using Dexcom. I’ve discovered a website that helps me predict blood sugars. It’s predictbgl.com or managebgl.com (either one) and it’s made a huge difference for me just this month.

For all the money I’ve put into Dexcom (and that my employer has paid for its share of the strips), I have lots of high and low blood sugars. In the few weeks I’ve used predictbgl.com, I’ve averaged about one a week? You can try it out free for a month. Then it’s $60 for 6 months. I’ll warn you that it takes work. You need to log your grams of carbs, protein, and fat, as well as your exercise, insulin doses, blood sugar readings (from finger sticks), and other factors (illness, stress, pain, alcohol, etc.). I’ve discovered that when I have eat out or have homemade soup, I can’t use predictbgl.com, but I use it for every meal I can. I can usually use it for breakfast and lunch, and it’s already given me level blood sugars for those time periods routinely.

You have to start out estimating how many grams of carbs are covered by a unit of insulin at different times of the day, and how many mg/dl a unit of insulin will lower you. There’s no way you’ll guess correctly to start, but there’s a “Coach” feature that suggests what you should change your estimate(s) to. It’s working for me better as I keep revising the estimates.

I’ve developed a spreadsheet that lists grams of carb/protein/fat per 100 grams of foods. predictbgl.com has a food database, but I haven’t tried to use it yet. I weigh my food, add up the grams of carb/protein/fat, and put them into the spreadsheet. The chart warns me when I’m going to go high or low (defined by what I tell it is high or low), and I can eat more or take a shot based on what the chart shows.

My A1C’s over the years have ranged from 5.0 to 6.1 (except for a 6.3 earlier in 2015), but every A1C has included at least a couple hundred low blood sugars that helped average down my high blood sugars. A few weeks of this program has given me real hope that I can keep both highs and lows down to one or two a week over the long term.

The app works best with Apple, but I don’t have that, but I’m getting by (well enough) with my HP computer and Samsung phone. I recommend that you try the website out for a while (as I did) before paying, to make sure that you’ll actually go to the work of weighing your food and working on revising your estimates as necessary. If you really want to avoid low (and high) blood sugars, I haven’t come across a better way. Dexcom doesn’t predict. (You can tell you’re going low, but you don’t know how low, nor do you know how much you should eat to come back up just enough.) And it’s expensive. And I understand that its new device (which I haven’t gotten) has planned transmitter obsolescence, to make sure you can’t milk extra time out of it the way you could with the G4. I have three more sensors to use, but I don’t plan to order more. One sensor costs more (before insurance kicks in) than a whole year of predictbgl.com.

1 Like

I have been looking at this site and app for a few days, actually. In fact, I mentioned it tonight in another thread. To me, it doesn’t replace the CGMS, but it certainly looks interesting. So far, the results have been off for me on both ends of the spectrum, but we’ll see how it goes over a bit longer time. I’m not giving up my Dexcom G4 yet, though :slightly_smiling:

1 Like

First of all, you are doing an outstanding job keeping your BGs in check. I use a scale and spreadsheet for my meals as well. I store my common meals as separate spreadsheets and call them up to see what dose I need.

I checked out the Predict BGL program but was put off by all the data entry and decided not to use it. I can track most of my diabetes metrics by uploading them to Diasend. I can see all my fingerstick readings, CGM data, insulin amounts as well as the carb counts as long as I deliver carb boluses. Diasend even shows me exercise data from my Fitbit.

Now I know that the CGM cannot predict the immediate future but I’ve been able to glean good info by looking back at CGM trends for the last 14 days in what’s called a 14-day standard day. The standard day, if you’re not familiar with it, displays data over one 24-hour period but it overlays and averages the 24 one-hour slices of all 14 days.

So, from this chart I can see, for instance, that I tend to go high one hour after eating dinner. Then, when I eat dinner and see my CGM move up after dinner I can infer from the 14-day chart that I’m likely headed higher. Before my BG is even out of range I can start to take pre-emptive action. I can go for a walk or add insulin or both. This allows me to anticipate informed by the data of a recent 14-day period and be right much more often than I’m wrong.

I’m not trying to convince you to abandon your system. It’s rational and effective for you. And you’re obviously successful with it. I just wanted you to see that there are other ways of using the CGM system to help “predict” BG levels. I like reading posts like yours!

2 Likes

I’ve tried numerous apps and found the data entry to be extremely tedious. As apps go, however this one has (to me) the most intuitive data entry - though I wish it would have skipped all the extra ‘OK’ and ‘Save’ button requirements - pointless extra clicks… Still, easier to use than many others. I’ll continue to play with it for a bit longer, though I think $10/month on top of everything else I’m using (I do not see it replacing other tools yet) is a bit much for me.

If I had a really regular schedule, I could probably catch on to my trends better. What I’ve liked about Predict BGL (I’ll call it PBGL) is that I can type in what I’m eating and see if it’s going to be enough or if I need to eat more. One other item - I haven’t learned how to make my exercise factor in properly yet, but I’ve made so much improvement in using PBGL that it’s a matter of time until I get the exercise logged properly. Simon (the fellow who developed the program and runs it and emails answers to all my questions) told me how to log the exercise, but I just got the email the other day and have still been too tied up with work to sit down and figure it out.

Regardless of my learning curve, I’ve practically wiped out high and low blood sugars. By the way, when I stop using a CGM, I’ll expect to check my blood sugar at least twice after each meal for a while, but it’s worth it. Life without high and low blood sugars is quite a change for me.

1 Like

That is truly worth the effort! Congrats.

we had an interview w them a while back
http://www.tudiabetes.org/video/live-interview-with-simon-carter-predictbgl/

1 Like

So that’s where I learned about the site!! Thank you. I couldn’t remember where I read about it. I saw a little of the interview, but haven’t finished watching it yet. I understand that many don’t want to keep track of all they eat, but I wish I could get across what a drastic difference it is to go from at least 7 low blood sugars a week to 0-1, and 3 - 5 high blood sugars a week (> 250) to 0-1. I’m thrilled at the difference. This site ought to be promoted more widely. If you want to avoid low (and high) blood sugars, this is the best method I’ve ever seen. Far better (and cheaper) than CGM. The work of recording my eating has also helped my weight.

1 Like

You don’t have to go to town to put in all the data. The minimum is the same as you would use on MDI or using a pump - just BGLs and Carbs.

Note that you don’t HAVE to put in exercise, stress etc - these are optionally things you can add in. You can also factor these things into pump calculations, but it involves a lot more button pushing!

I’ve caught on that I can’t use the website (PBGL) 24/7. Some foods are not easy to calculate, even if I can weigh them. I can almost always calculate breakfast and usually calculate lunch. I don’t bother with PBGL when I go to a restaurant. And once I stop using PBGL in a day, I don’t use it the rest of the day because I won’t be able to include the carryover effect of protein and fat. Which can make things a little trickier at night.

Some days it predicts beautifully; other days it doesn’t. My job has made it harder for me to put the time into using PBGL that I need, but after January I can do better. In the meantime, even if I reset PBGL every time I stick my finger, at least it usually gives me some idea of where my blood sugar will be in a while. I can go to the gym when I see how long I have before expecting my blood sugar to drop under a certain point. I can eat a little more food because the chart shows that I will go low in a couple hours. It’s not a perfect science, but it’s better than anything else I know of for predicting. As I mentioned earlier, Dexcom can tell me that I’m headed low, but it can’t tell me how low I’ll go. Worse, I don’t know if I’ve eaten enough to recover from a low blood sugar.

On the subject of pulling out of low blood sugar, I’ve learned to use 15 grams of corn flakes to raise my blood sugar quickly. Its glycemic load per calorie is about the best I know of for a food that can sit in my desk or car for days at a time. If I have plenty of time to raise my blood sugar before it goes too low, I can eat fruit. But if I have to hurry, give me the corn flakes.

Summing up - It takes work (which gets easier). It’s not perfect. It’s the best thing I know of for avoiding high and low blood sugars. If I had to choose between Dexcom and PBGL, I’d choose PBGL, no contest.

With the incredible accuracy provided by her Dexcom G5, PBGL wouldn’t be worth the extra work for my daughter (or me)…

A couple examples of why Dexcom accuracy is only part of the story: It will not tell you if you’ve eaten too much and your blood sugar will be at 300 in a couple hours. Nor can it tell you at bedtime that three hours later you’ll need to eat. Knowing my blood sugar (by fingerstick) and its trend (by Dexcom) is helpful, but I’ve probably averaged a dozen low blood sugars a week and at least three or four blood sugars over 300 a week even with Dexcom. With PBGL I’m down to a couple a week, and those are pretty much when I am not using PBGL. I can take insulin or eat or exercise better because of the PBGL predictions.

I see prediction (not being caught by surprise) as the key missing element for diabetes. If the chart shows I’m going to be low in 2 hours, I can type in the figures for eating an apple or a sandwich and see if the prediction rises by the amount I want. It is easier to do that because that I like math. It doesn’t help, though, when I’m really busy at work and don’t get the numbers typed in for PBGL. But it’s the best thing I’ve seen yet, although I’m still looking.

I can usually get a good idea if this is the case simply by experience - knowing roughly what I have eaten, what the trend is and how much insulin I have on board. My Vibe/Dexcom will show the trend and IOB on a single screen.

The thought of having to carefully weigh everything and then, to add insult to injury, having to laboriously enter it all onto a smartphone doesn’t fill me with great joy! One of the great advantages of using a Dexcom, is that allied to my experience, I don’t need to carb count with such precision. I can estimate and then correct early (i.e within an hour or two) before my numbers get significantly out of range.

The 640G used with an Enlite apparently does a pretty good job of predicting and then forestalling hypos by cutting basal delivery. With a bit of experience it’s not difficult to do this “manually” with the Dexcom+Vibe

1 Like

I enter almost all of the data on a desktop. It would be a lot harder to enter on a smartphone. So far I haven’t figured out how to get the chart to work on a smartphone, and the chart is the predictor. I don’t use a pump. My experience and knowing roughly what I’ve eaten and A1C’s of 5.7 have not been enough to prevent hundreds of low blood sugars a year. I would rather use a system that so far is helping me avoid low (and high) blood sugars than keep “winging it.” I also like what it does for my weight control.

@keepitsteady and @lisa0005 – Ok, I’m better appreciating what your predict blood glucose level (PBGL) can do. It’s kind of like a manual artificial pancreas algorithm. If your chart software were incorporated into an app and it had data from an insulin pump (both insulin data and carb data) and a CGM, it could automate the PBGL and replace manual data entry.

I think your system is worth a trial. Right now I am personally involved in the process of moving my household with a major lifestyle change so I don’t have the time to devote to this but I am interested. I’ve incorporated many tasks to treat diabetes into my life that most people would consider “too much work.” But I’ve found that after a while habit takes over the chore and I’m enjoying the rewards. As I said I think your system is worth a trial. Thanks for your persistence.

Can you tell me how many pieces of data, on average, that you need to enter into the chart each day and roughly the estimated time it takes to do that?

Entering the data doesn’t take that long, but it depends partly on how you want to do it. For example, do you want to group several foods or enter them one at a time?

I made up a spreadsheet that has a row for, for example, carrots, with calories, carbohydrate grams, protein grams, and fat grams per 100 grams of (raw) carrots. On another sheet I enter the number of grams (for example, 80) of carrots I’m eating, and multiply by the figures on the other sheet. On my list, that’s 33 calories, 7.2 grams carb, 0.8 grams protein. Glycemic index of 71, so I put them as Medium GI (glycemic index).

The tricky part is not just weighing the food; it’s deciding which GI category to use. That’s only for carbs, though. But I’ve tweaked my category to get the chart to match my blood sugar stick (which of course isn’t always accurate). If I ranked a food as Low GI, and my fingerstick showed I was higher than the chart predicted, I may shift that food to Medium GI, realizing the categories can vary (depending on the food itself and the food in combination with other foods).

The more I get used to ranking a food for GI category, the simpler and quicker it is. When the chart disagrees with my glucometer, sometimes I spend more time than others figuring out why. My breakfast is pretty much the same foods day after day, so it takes almost no time to record. If I were you, I’d try the system almost a whole month before buying into it. (There’s a month free trial.) You’ll probably see that some days you think it’s wonderful and others days it’s terrible. Some days you won’t have time for it.

It takes 4-6 BGs per day, but as @keepitsteady says, you can stop and start it as needed. Just try it for a month to get your doses right first. Also - you don’t have to enter food GI factors. The normal GI model is pretty good for most things. But if you want to deep-dive, it’s all there too :slightly_smiling:

4 BGs per day takes about 30 seconds total. And yes, PBGL is looking to automate as much as they can. They already automate fitness/exercise data.

it won’t be worth, it for me, i ues my Dexcom for it,.

You’re missing my point about predicting. When you get up from eating a meal, do you know if you’ve eaten enough that your blood sugar won’t go low before the next meal? Or that it will go so high that you’d better take more insulin? Dexcom won’t predict that. It will only tell you where your blood sugar has been heading the last few minutes or hour or two. When you finish eating, Dexcom has no clue (nor can it give you one) where you will be in 30 minutes or two hours.

Listing how many grams of carbs, fat, and protein you ate and how much insulin you took takes some work, but it is necessary if you’re going to get help in predicting. Sometimes I know to eat a little extra food because when I type in what I ate (or what I’m going to eat), the website warns me that I’ll run low in three hours. Dexcom cannot do that. Or I know from typing in the food that I’m eating too much for my insulin dose, so I either take more insulin or cut back on the food. Dexcom can’t tell me to do that.

You do not use Dexcom to predict two or three hours down the road. You simply guess. The website can make your guess accurate a higher percent of the time.