Can insulin pumps really be this off when delivering insulin?

I´m reading Stephen Ponders book “Sugar Surfing”, and this quote from page 82 really surprises me and makes me wonder if this can be true.

The quote:
“Did you know that when you use your insulin pump to bolus 1 unit of insulin it can be off by a range of plus or minus 30%-200%?”.

There´s no reference or footnote, and his point is to say that variances you can´t control has impact on your BG.

Anyone who knows anything about this statement?

This surprises me greatly. If this is true, how could we possibly realistically ever manage blood sugar effectively. Perhaps it’s out of context? I could see absorption issues affecting “delivery”, or occlusion issues being an issue, but the actual pump not delivering with accuracy - that’s a shocker to me.

1 Like

I confess I don’t own the book (though everyone says I should!) so I can only assume there’s more context to it. As it stands, there are about eight ways to interpret it–are we talking FDA pump specs, absorption rates…? And based on what data? Seems kinda like the stat about test strip (in)accuracy but I feel like I have a much better handle on what that one means and where it comes from.

The context is as follow:

"Insulin Dosing & Delivery

No matter how you get insulin inside your body, the tools we use are not entirely accurate or precise. There is always acceptable variance in the manufacture of plastic vessels, tubing and syringes. Insulin pumps have an acceptable variance when told to bolus a specific number of units. Did you know that when you use your insulin pump to bolus 1 unit of insulin it can be off by a range of plus or minus 30%-200%? In relation to the powerful effect of a small dose of insulin, even a small variance from shot to shot or pump bolus (including basal delivery) can have meaningful impact on resting bloodsugar.

As to battery operated machinery, do you ever notice a difference in the performance of your electronics as the battery level goes from full to empty? Why would medical devices be immune to the Law of Variability?"

Then he writes about tunneling, the leakage of insulin around an infusion catheter. Next is bubbles in tubed insulin pumps and he recommends visual inspection of the tubing.

End of chapter.

My experience makes me skeptical of this statement. I don’t doubt that there is some range of inaccuracy with pumps but this range of 30% to 200% seems extreme to me.

The larger lesson that I have drawn from his work is that due to mechanical variability as well as site absorption anomalies, we need to be attuned to how variable this system of food, insulin and exercise is. And be able to respond to these situations in a timely and proportionate way.


Hope my comment didn’t seem a criticism of your post, but thanks for the additional context–which doesn’t actually resolve the question(s) does it? One thing that makes it a bit hard to accept, at least from the device side, is that these machines have been developing the ability to deliver smaller and smaller increments over the years–it’s a major advantage they have over injector pens, f’rinstance. And this statement, taken just as read, would seem to be saying that’s a total wash–those increments are getting lost in the noise. Needs a lot more clarification than he gives. Maybe somebody wants to contact him–if it’s causing us confusion, we’re certainly not the only ones.


I don’t believe the MM pump is as inaccurate as the book claims. Perhaps some other brand of pump is what they are referencing?? In any event, I’d say the MM delivery accuracy is acceptably good.

Without further clarification and references to supporting documentation, I think this statement is simply patently absurd. It’s frankly a stupid thing to say because it now makes me approach anything else he might say from a position of doubt.


Here is a quote from the MM 530G System User Guide, page 242, Delivery Accuracy section:

Pumps are programmed with the 1.0 unit per hour rate and then measured for delivery accuracy over a period of 24 hours, in a laboratory setting.
• Testing has confirmed that the delivery accuracy for the maximum/minimum
bolus/basal amounts are within the +/-5 percent of the programmed value.
• All normal boluses will be delivered within a maximum of six minutes.
• Displacement (piston) accuracy: +/-2 percent.

So, as far as the pump itself is concerned, the statement in Ponder’s book is definitely not true. Variability due to bubbles, site, etc., could be higher, but how high - who knows.

Regarding his statement about battery, having background in electrical engineering, I can with high degree of certainty state that pump delivery performance can easily be engineered so that accuracy remains the same for any battery charge within battery life limits. So, stay rested, no need to replace your pump battery before the pump tells you to do so! The statement in the book is way off the mark.

Having said all that, I still highly recommend Ponder’s book. Some details are indeed stated somewhat carelessly, but I would not take that as a reason to discount the basic principles, methods and techniques described.


Here is some information from Animas about dose precision and total flow precision.

Insulin Pump Dose Precision

Precision, Accuracy and Dose Speed

Note: this is written by Animas employees!

Yes, MM pumps reject batteries that will work in other devices, as they are ULTRA conservative about the acceptable battery voltage. When the battery icon shows one bar, the battery is still fully capable of powering the pump to do it’s “thing”, with the only exception being, in the past, communicating with the old BD Link Meters for downloading data to a PC. Just test a recently removed battery from a MM pump and you all might be surprised at how much voltage remains. I keep them to power other devices, rather than toss them in the trash.

Yeah, been there. Doing that. If nothing else a flashlight can still be useful for a long time using batteries which the pump will no longer tolerate. :smirk:

1 Like

@DrBB: I didn´t take your comment as criticism. I just wanted to give some context for all the readers of this post, even though it doesn´t resolve anything.

I agree with you that this really seems to counteract the ability for pumps to deliver smaller and smaller increments of insulin. That´s why this statement from an MD made me wonder.

I am thinking about sending him an e-mail. I think he should clearify things for further editions.

1 Like

Even if he doesn’t do that, it would still be nice to find out what the heck he truly meant. Some examples, documented in some way if possible, would be helpful. As you point out, he’s not just some Joe or Josephina Troll. When he says something there are greater expectations of credibility.

Hi Siri,

Yes, everything we wrote in the book was thoroughly researched and/or clearly an opinion. In the case of insulin pump dosing variability, you can go to this link to see our source:

Thanks to the person who tweeted this so I could be made aware of the confusion.


Kevin McMahon
co-author: Sugar Surfing


Thanks @Kevin_McMahon. It’s a long (!) article and I haven’t read to the bottom yet but did come across this paragraph:

Currently approved insulin infusion pumps specify basal rates of 0.05–30 U/h, and at 1 U/h the accuracy of delivery is typically within 5% of the programmed dose. These approved pumps also specify bolus doses of 0.05–30 U, and for a bolus of 1 U, the accuracy of delivery (when reported) ranges from ±30% to ±200% of the specified dose.

I’m not sure I have this right but it seems like they’re saying the basal deliveries are more accurate than the bolus ones. If so, maybe that’s because the basal is a series of deliveries over the hour so they have a chance to average out to within 5%, whereas the bolus is usually a larger dose, all in one go. I’m guessing the ±30% to ±200% variance tracks with the size of the dose–i.e., a small amount off in absolute terms is proportionally larger to a tiny dose than a large one. Because I’m pretty sure I’d notice if a 6U bolus was actually giving me 12 U of insulin.

Gotta say I’m still pretty confused though. Maybe there’s more further down in the article…


“There are three kinds of lies: lies, damned lies, and statistics.”


Thanks for identifying the relevent paragraph in the cited article. In fact the citation refers to a report on a conference on Diabetes Technology. It is NOT a peer-reviewed scientific papere. The paragraph quoted by DrBB and presumalbly the source of the suggestion in Sugar Surfing re: the accuracy of bolus delivery is actually a paraphrasing of a presentation by Alan Stevens of the FDA. The data underlying the contention that error rates can be up to plus/minus 200% are cited. as “accuracy of delivery (when reported)” - there is no specific citation to a peer-reviewed publication or even an internal FDA report. We therefore have to question where these numbers come from.

TBH, if the FDA (or any of the other regulatory agencies) were identifying errors of this size in bolus doses (I exclude rare pump fault/failure events) the devices would simply not be approved. Kevin McMahon (co-author of Sugar Surfing) is to be commended for responding) but he perhaps needs to check out the underlying sources of the data and its potential reliability (or otherwise). UK government agencies when assessing medical evidence assign grades of A to D where A would be from an RCT or similar and D would be anecdotal. I would suggest that the figures on Bolus delivery errors are closer to D than to A.

It is worth pointing out also that the very next next sentence in this paragraph, discussing delivery errors resulting from occlusions states “it can take as muc as 24 and 60 hours to detect an occlusion”. Any pump user would know that in the real world this statement is at best of limited relevence.



@Kevin_McMahon: Thanks for clearing things up and contributing to the discussion. I´ve skimmed through the article you´re linking too and guess the paragraph @DrBB is quoting is the one you refer to in your book.

And like others I´m still confused.

  1. Does “accuracy of delivery (when reported) ranges from ±30% to ±200% of the specified dose” really mean that when - for instance - bolusing 5 U I could get anything from 3,5 U to 6,5 U in best case (±30%) or anything from 0 U to 10 U in worst case (±200%)?

  2. Like @jjm335 writes, the citation you refer to in your book is a report on a conference on Diabetes Technology and not a peer-reviewed scientific paper. Do you know which paper/papers the numbers come from?

  3. If I have understood this right, I agree with @jjm335 errors of this size in bolus doses, the insulin pumps would simply not be approved by a regulatory agency.

I would really appreciate it if you could clear things up some more.

1 Like

The error, as far as I understand, is defined as (100%)*(delivered - intended)/intended. So, in your example, the worst-case ±200% error would imply that the dose delivered could actually be as high as 15U (+200% of 5 is 10) or as low as -5U (-200% of 5 is -10). This is just to illustrate how completely wrong and nonsensical that statement about pump accuracy is, Pumps are not perfect by any means, but they are certainly not that bad either. On another note, if we could figure out how to suck out 5U (or any) insulin, it would be far easier to sugar surf :wink: Sorry for the digression, @Kevin_McMahon - I love the book - it’s just that this particular statement needs to be corrected or just deleted. I also suggest deleting the statement about “battery operated machinery” which in the context of insulin pumps makes no sense either. We have more than enough variability to deal with, no need to invent more stuff.