The manual pancreas

I’ve said it many times - I’m not diabetic, I just have a manual pancreas.


So you produce insulin?

Imo The joke is the so-called artificial pancreas. A pancreas produces insulin using the food that the body it helps keep in perfect balance digests. It doesn’t buy insulin.

With the best tech available commercially today, what we are is poorly informed, half competent syringes.

There is no incentive for any commercial company to produce a true artificial pancreas. In the US there’s resistance to non-invasive glucose monitors that are being tested in Europe because the CGM disposables market would be decimated here. Imagine what would happen if a laboratory could take your t cells and create beta cells that your antibodies could not recognize as foreign.

Eli Lily and Novo Nordisk would lose their biggest cash cows. Every insulin pump and CGM manufacturer would fold those production lines. Hospitals would lose an enormous amount of revenue per patient no longer having diabetes.

Doctors would probably welcome it intellectually, but not financially, as the diseases that are complications of diabetes would also decline.

I’m a disenchanted optimist. I’m a realist. A cure for diabetes will always be at least 10 years away in the United States. I figure in 10 years it might be possible to take a trip to Europe to get cured, but impossible to get out and back in legally. Congress would pass a bill outlawing the importation of genetically modified human beings.


Yep, I am definitely poorly informed and half-competent.


Please understand that I’m not criticizing people but the “high” tech + people combination. To me competence is measurable by performance and consistency.

CGMs are secondary information sources that can’t see the primary source. Pumps deliver insulin to a “dock”, don’t know what has been received by the “customer” inside. we know how to fill the order for insulin but have inaccurate information about what is needed and incomplete control of what is delivered . Half informed, half competent.

An expert employee who isn’t reliable and trustworthy is less valuable than a mediocre one who always gets the job done in time and to spec. You can never turn my back on the former. With the latter, you can go on a vacation and know that when something that they can’t handle comes up, they’ll call you immediately.

Like it or not we are part of the physical loop and the actual controlling algorithm of a pump/loop system. (Imo, we are the least reliable part.) When we outsource part of our control to an pair of unreliable assistants we increase our ignorance about what is happening and lessen our competence at achieving what we would be able to do without them.

Most adult persons with T1D I’ve encountered are fairly well informed how the diabetes management tech they use should work, and how to use it when everything works right.

I consider myself better than average at that. My TIR with pump and CGM is over 90%. I know what to do when I have good data. I’ve had stretches of a week nearly as good as your day, better because I’m flat overnight without a protective or control algorithm. My TIR with just a CGM was over 85%. I know what to do when I have good data. Then I’ll have a day when thing go awry. I have gastroparesis. I get sick. A site that had been working well goes bad with no warning. I get stressed.

All it takes is one failing CGM site to make me poorly informed, and/or a bad infusion site (I’ve had both happen simultaneously) to make me less competent because I’m reacting to, not controlling my BG. On days like those, the best commercial tech becomes less good than MDI and BGMs, and can be much, much worse if I turn my back on it.

My last bad CGM data experience had one stagnate at 165 mg/dL while my BG continued to change. I was uniformed until the next scheduled food bolus hours later when I realized it was goofy. I’ve since set up alarm ranges and TIR measured in minutes in Xdrip+ that alert me when reported BG lingers outside my control range longer than it should.

Pancreases don’t have brains, they react dynamically and immediately to changes. They are purely analog not digital. A system that doesn’t work the same way isn’t an “artificial pancreas”. Control IQ is an automated delivery system. with poor sensors that don’t know what is happening real time , and software that guesses the future. It doesn’t utilize all the information it has that will affect the accuracy of those predictions. Rather than acting instantaneously and accurately to ambient conditions is a highly damped response system.

Basal IQ has the same problems/defects as I do - and more. If blindly trusts CGMs as they intermittently compress and fail, isn’t able to sense site failures, which are more common than occlusions, and less able to sense occlusions than practically anyone who has used a syringe.

The G7 sensor might reduce response time and planned down time, but not the other problems. Non-invasive realtime BGMs could improve response time and accuracy, Pumps might be able to make better decisions. But the problem of delivery to the inside customer will remain. We can become well-informed 3/4 -competent automated syringes.


The only time I rely on CGM is when I am asleep. The rest of the time I use BG tests.

What you are seeing above is not an algorithm (other than nighttime, when I am not eating). The rest of the time, like 16+ hours, it was manual.

Perhaps we are somewhat in agreement about the idea that an algorithm which relies on a CGM will never be as good as what a person can do manually.

1 Like

I think we’re in agreement . if someone is reasonably healthy, highly motivated, has average intelligence and willing to continue to learn, no glorified calculator can outperform them.

Up until about 5 years ago I was getting results similar to what you have, using 10 blood glucose tests per day with MDI and Novolins. Then my health started to degrade.

I started using CGM to get an idea what was happening with my gastroparesis . (I was not prepared for the unreliability and mediocre accuracy). When I saw an endo the first time my A1C was 6.7, which for me was moderately high . They didn’t understand how I could be doing that on Humulin .

To me that points out that there are some highly educated people that are incapable of getting results as good as a CGM plus a pump algorithm that works poorly to deliver A1C/TIR under 7.5/75%.

The benefit to me using a professional CGM for 2 weeks and a sample for another two was they let me see trends and direction of change which was enough for me to make dietary and timing changes - to get results than their patients weren’t getting after a year with a G6 and T:slim.

imo I was doing “B” work. 75% is barely passing. With a grade like that in college I’d be on probation. The School of Hard Knock has higher standards.

After 3 weeks with a pump and Basal IQ turned off I was getting 90% TIR with no unpredicted low range lows, because I wasn’t relying on the endo to get the numbers for ICR, CF and basal rates correct. Waiting a month for a “educator’s review” or months for an endo appointment made no sense to me.

As you must know determining these things is relatively simple if you eliminate the variables of stress, activity and diet. I could teach it in 15 minutes to anyone., but unless they were more motivated to “stick to the plan” than most dieters, most couldn’t do it for two weeks. imo, Most persons with T1D could.

Anyway after a month I’ve asked the software to be “ungraded” to Control IQ, I knowing the ridiculous effort it takes to keep the stuff working well, and that with the right numbers it can’t do as good as job as I can when I’m able. But a passing job when I’m not as able is better than failing.

I absolutely would not advocate today’s CGM and commercial algorithm pumps except to the infirm or the very young. They are too expensive. I might change if non-invasive CBGMs AND better infusion tech becomes available. By that time, at the rate the FDA moves, there may be true synthetic pancreases available in the EU.


1 Like

If you look at the bell curve of diabetics out there perhaps you will see that the vast majority of insulin users can benefit from CGM with commercial algorithm pump. I don’t have the statistics for you but they are easy to find. I have no problem advocating for expanded access to this tech for everybody. I hope your comments don’t discourage some from giving it a try.


Having done BGMs and syringes, I find that using a pump + CGM takes more effort often enough, at such a large price premium that I stand by what I said. The financial and physical cost/benefit ratio is so poor in comparison to simpler tech that it has to be matched to user need and ability.

Cost “paid for by insurance” is paid for by people. Imagine paying double for a car so a blind person can “drive” and the projected number of blind persons is going to soon be 1/3 or the population.

“Everybody” is what the equipment makers want, persons with T1D and T2D, with one company going so far as to mislead in their commercials about what is required to use one and the benefit.

Despite that, I’d be willing to recommend CGMs because they are consumable products that have a potential educational benefit, and using one can be stopped at any time with no continuing cost. The initial cost of a pump is very high in comparison.

The vast majority of the world would benefit from petrol cars being replaced by EVs immediately, but today the cost of doing it today is prohibitive, the supply so limited that there’s no incentive to make EVs affordable, and the infrastructure isn’t adequate to support “everyone” getting an EV.

"Smart pumps’ are similar. Their infrastructure shortage includes users trained to use them effectively, and in the US, a supply of just 4000 endocrinologists who are just getting used to the idea of the new tech and nor able to handle supporting the users. The surge in CGM adoption stretched waiting times for an appointment from 3 to more than 4 months. Its easier to ramp pump and CGM production, than matching it with trained medical support.

An extreme position on a bell curve of A1Cs has little correlation with personal capability of getting a benefit from pump tech. Pump algorithm’s protective results depend on good input from the user. GIGO.
That argues for the persons who are most capable of using MDI with CGM getting the best outcomes, those with mediocre less, and those with poor getting the least.

Someone with T1D might potentially benefit if they run high or experience lows often, and have wide variations in BG but too often that’s because they can’t/won’t count carbs and/or don’t have accurate numbers for their ICR and CF because of inadequate medical support . Those are problems that a commercial pump can’t solve today, but a CGM might help, by educating the user.

People who need and can afford to have caregivers who are committed to their welfare can benefit the most from the CGM+pump, people who are making serious, committed effort to manage their diabetes but failing, the second most. The benefit to those who are not committed to preserving their own lives is as limited. as the benefit of insulin and healthcare are for those who cannot afford them.

People like me get a marginal benefit at high cost. I’m lucky to be able to afford it. I’d feel guilty about diverting the cost from my family, if I didn’t believe that I’ll need a caregiver too soon and was trying to make that job easy enough that I can afford to pay for it.

Nice idea, but not true; the pancreas does not know what non-fully-developed-T1s have eaten until a significant time until after they eat. Depending on the responsiveness of the NFDT1 the result will be a blood sugar rise to a level somewhere in the range 150-250mg/dL. Unfortunately there have been very few, if more than one, studies of NFDT1 responses over the time period - such studies require CGMs if real world conditions are maintained. This seems to be the major and I suspect only study:

So careful reading is required to understand all of that and to avoid making stupid comments based on averages. Here is an informative picture which shows averages:

(I’m quoting women because I just replicated this test on one subject, my wife; HbA1c 5.2, baseline fasting BG 100mg/dL, 2 hour post-prandial BG typically around 100mg/dL, one hour post prandial between around140mg/dL and 250mg/dL, 2-3 hour post prandial often hypo.) So the interesting thing is the variation. If you read the full article it turns out that a significant number of people are going over 200mg/dL after one hour:

There is also a distressing number of people who are going hypo. The data refers to 148 people with one, single, 6 day test. My wife’s recent experiences are that the highs and lows are debilitating; we don’t notice that because we are used to them, but my wife is not and they “knock her out” (in her own words). Count hyper lines, count the hypo lines, divide by 148 and that is the percentage of people who, over a 6 day period, become incapacitated, “knocked out.”

No, the non-diabetic “automatic” pancreas does not work. We can do better. Scientists can do a lot more research on NFDT1s (i.e. everyone who isn’t an FDT1) and most likely explain the high rate of automobile accidents, along with a lot else.


I moved this discussion into its own thread and am open to be better topic name.


I changed it, @eric’s aphorism seems to sum it all up.


Thanks for posting this article. It’s extremely interesting, and helps relieve some of my anxiety surrounding going “high” or out of range after a meal. My husband (non-D) always says, “You’re doing the best that you can.” Not a very satisfying comment, because my perception is, I’m not doing good enough!

Is your conclusion that all the subjects that had “high” postprandial BG are headed to T2D?

But you are a great sugar-surfer!!


Absolutely not; no such conclusion can be drawn based on one set of data over 6 days for each subject (888 meals total). Any conclusions like this would require long term studies and those studies would have to include the whole non-diabetic population, not just the approximately 30% of the subjects who met the “pre-diabetes” criterion (see “Table 1, clinical characteristics”).

The problem with all of this stuff is that “pre-diabetes” doesn’t have anything to do with T2 or T1 diabetes; rather it is a term describing a BG behavior on eating and fasting, quoting the paper:

criteria according to the American Diabetes Association (fasting glucose ≥100 mg/dL and/or glycated hemoglobin ≥5.7%)

So that criterion, well, pair of criteria, manages to capture around 1 in 3 of the subjects; presumably it would extend to 32% of non-diabetic 50 year old Spanish women and 28.8% of 46 year old Spanish men in NW Spain. Is that really useful?

“Pre-diabetic” is a term doctors use when they encounter someone with a fasting BG 100mg/dL or more or an HbA1c of 5.7% or more. It doesn’t mean anything else. The study states:

Those women [* * *] who had prediabetes presented a statistically higher postprandial glycemic response

So think about that… prediabetes means a higher HbA1c and/or a higher fasting blood glucose and it is correlated with a higher postprandial glucose. I find that statement weird; it seems obvious to me that a group of people identified based on higher average blood sugar are expected to have higher blood sugar after they eat. What would be startling is if there was the opposite correlation, but the authors of the paper didn’t test for that:

In females, those who are in a prediabetes situation had a significantly higher postprandial glucose levels. The same fact was observed in the group of males, but without reaching statistical significance.

The study did not seem to look at the variation in response for single subjects; it used data for 6 meals, one per day for each subject but doesn’t seem to compare response variation within those 6 meals with response between different subjects. This is relevant because I observed that my wife’s response varies enormously with regard to peak, peak time and duration with time of day and exercise, the latter was not monitored in the study and the former was factored out by using only one meal (“dinner” I think; probably around 9PM).

My wife’s major excursions are happening in response to relatively small amounts of carbs eaten after getting up; so previously fasting and eating something like a ‘mini bagel’ with 21g of sugar/starch. Larger meals seem (mostly) “softer” and exercise, including just walking around, seems to have a massive effect.

As the authors say:

Future studies are necessary to study in depth how gender affects the postprandial glycemic responses. Increase knowledge of glucose response to meals can contribute to better management of diseases related to glucose metabolism.

1 Like

Here is a one-on-one BG challenge test from a few years ago. Both subjects drank a sugar soft drink at time=0 minutes, about 45 grams of pure sugar in there. All of the numbers are from the same meter (and a really good meter too!)

A diabetic vs a non-diabetic. The non-diabetic was a healthy young runner in his 20’s. The diabetic is just some old guy.

Spoiler alert! "Test Subject 1" was me. :grinning:


This is even more true when looping and algorithms that deliver the insulin automatically are relying on the bad CGM data (roller coaster rides anyone?). But I wouldn’t trade the tech for anything. Prior to the tech, I hardly slept for the fear that my 2yo was dying while asleep. Knowing his BGs (even if only in the ballpark) and knowing loop is making its best attempt at correcting, help me sleep better at nights. 6 years in and we couldn’t be better. A1c constantly between 5.5 and 6.0 with 1% low/ severe low and 10% high…we’ll take it!


That’s one thing that worries me; it is very hard (but certainly possible) to come up with algorithms which are robust in the presence of bad, but temporarily bad, data. The fear is the roller coaster; feedback, in the audio sense of when someone picks up the microphone and the hall fills with a high pitched bell that sounds on every syllable.

Have you seen this happening? It should happen, it would happen if I wrote the algorithm because I don’t know how to make it stable. I suspect the big three are being very cautious because of that possibility, it’s certainly Lawyeria (cf Liberia, not the legal software).

This is one of the advantages of FOSS loop code; we can chose to do something which, while not actually stupid, would certainly transport us to Lawyeria if we did it to someone else. It’s also one of the challenges - producing an adequate insulin pump based on unreliable information is akin to many other engineering problems but more dangerous.


I’m getting tired of your condescending nit-picking, but not too tired to respond. You are wrong.

The comment is missing a comma before the last word, but is absolutely true. It’s your understanding, not mine that’s flawed. What I stated is not what you believe it said.

My standard of perfection is natural and organic because those are automatic manifestations of the laws of physics. I appreciate and admire biology more than I appreciate or admire technology - and I’m an engineer. To me an artificial mechanism will never be as good as a natural organ. A prosthesis isn’t an artificial organ, it’s a substitute that lets us get by when a natural organ fails.

Your perfection would define the operation of a human body as imperfect because it doesn’t work like a robot following directions moment by moment. That’s backwards, distorted values and thinking.

The pancreas is not digital. It doesn’t work like a CGM, a pump and a crude computer because it doesn’t need to. It doesn’t need external power and insulin. It doesn’t need to monitor blood glucose or an algorithm. It follows laws of physics without needing to understand anything. It did so for countless millennia before humans discovered fire.

The healthy pancreas is the standard of insulin provision. How it works is the benchmark for measuring perfection.

I didn’t say that it reacted instantly to blood glucose levels. It grows and repairs itself using food provided by the body and its reflexive response to its nutrient and hormonal environment combined with signals from the autonomic nervous system helps keep the body in perfect balance. It works in symbiosis with the rest of the body. A body in balance doesn’t mean that an individual aspect like pulse, respiration rate, or blood glucose level is at a near constant level, it means that the body as a whole is in balance.

Balance means that you can stay conscious, function, survive in generally good health and can adapt to changes. If you start working physically harder than your normal, and do it for a while, the body will adjust to the new normal. If you experience a sudden shock, your adrenaline level will spike quickly because being able to react and run or fight quickly is directly linked to physical survival. There is a mechanism to store a large surplus glucose in stable form and release it later, so blood glucose level does not need to react as quickly as adrenaline. Other hormones have different initiators, regulators, response times and delivery cycles.

None of those things describe the “artificial pancreas”. It’s an external sub-idiot level storage and delivery system that uses poor data to do one thing by rote. If it worked like it’s supposed to do, it would make as little sense as a pacemaker that prevented the heart from beating faster or slower than a moderate pace, no matter the need.

An AP isn’t even as close to the pancreases function as scuba equipment is to breathing. Scuba with no computing reacts dynamically to the actual environment and demand to deliver air. AP’s algorithms guess what you might need.

What we know about the pancreas is like knowing Leonardo DaVinci was a painter and sculptor. Those were only two kinds of things he produced- and we are still discovering how he did what he did. No one knows where his ideas and motivation came from. There have been imitators of DaVinci’s painting and sculpting, but never an artificial DaVinci that could equal his body of work. We barely understand the DaVinci or the pancreas.

I don’t pretend to. If you believe you do, you are mistaken.

An AP is a crude hack compared to the real organ. I’ll use one because I have few available alternatives that today that work as well at imitating one of the functions of the healthy organ I was born with. Transplants couldn’t cover 1% of persons with T1D, cloned beta cells are in shorter supply.

I guess I fall into that category then with my Tandem pump. The algorithm has allowed me to stop micromanaging my diabetes (unless I choose to) and I don’t because my A1c is now the best it has been in decades. Thanks algorithm!


Wouldn’t the non-diabetic have that sugar response (the one in Orange?) The orange doesn’t go very high. If it had been me without medication, I would be up into 250 or more pretty quick after drinking that much sugar.

I kept looking on this thread for an answer to this one but couldn’t find it.

Certainly the thread devolved into ad hominem disagreements after this last post so @eric may have been discouraged from continuing.

Can you post up the answer though??