Today, we welcome back Wil Dubois with the second part of his product review of the new Medtronic Minimed 670G, also known as the first Hybrid Closed Loop system ever approved worldwide. This advanced insulin pump-CGM combo received regulatory approval a year ago in September 2016 and has been slowly launching in the U.S. over the past several months. Wil offered his first impressions of the 670G earlier his summer, and has now been on the system long enough to share the Quality of Life impacts he’s experienced.

As I boxed up my Medtronic 670G at the end of my three-month trial, a wave of relief swept over me. This took me by surprise, as I’ve always been a pump lover and had been chomping at the bit to try this new device — the pinnacle of all that we tech-savvy T1s have been clamoring for — since I’d first read about it.

But in the course of using it, I experienced a shift not only in how I felt about the 670G, but about the invasiveness of insulin pumps in general for the first time ever.

To dig into the story of what happened, we have to start at the very beginning with a refresher on the system basics.

Medtronic 670G Basics

Officially known as a “Hybrid Closed Loop” because it only does some of the diabetes thinking for you and leaves the rest in the user’s hands, the 670G is the first of its kind and is by most accounts an early generation of an Artificial Pancreas system.

It has three modes: Auto, Safe, and Manual.

Manual mode: In this mode, the 670G functions as pretty much a garden-variety pump, albeit a nice one. It uses basal rates set by the user and his or her medical team, has variable insulin-to-carb ratios and correction factors, allows for dual and square wave boli and temporary basal rates, and has a low suspend feature that can be used either at a threshold target or in predictive mode, in which the pump will suspend delivery in advance of predicted glucose low, based on CGM readings and insulin on board (IOB).

Virtually none of these features exist in auto mode.

Auto mode: In this “automatic mode,” the system functions completely independently of the user. It uses a mathematical algorithm to control all pump features, and the only two things that you can set are the insulin-to-carb ratio and the duration of insulin action. The rest of your destiny lies in the hands of the math and the Guardian CGM sensor.

One confusion that many people have about the system surrounds its basal delivery. Just erase basal rates from your memory banks, because the 670G doesn’t use them. Instead, it uses “micro boluses,” delivering variable splats of insulin every five minutes — or sometimes, not delivering any insulin at all — in response to feedback from the sensor. As there’s no basal “rate,” there can be no temp rates. The only control you have as a user is to temporarily set the default blood sugar target of the system to 150 mg/dL from the usual 120 mg/dL if you are more active than usual.

Speaking of blood sugar numbers, not only does the Auto mode seek to hold you at 120 mg/dL, it won’t even allow corrections until you are over 150 mg/dL. And speaking of “allowing,” one feature we are all used to that doesn’t exist in Auto mode is the ability to override the pump when we know better. In Auto mode, the 670G won’t permit adjustments. The only way to change the amount of insulin the pump wants to give you is to cancel a bolus and go back and lie to the pump, changing your carb count up or down.

So in Auto, the pump ignores set basal rates and correction factors. It also does not allow for square or dual wave boli. That said, the math driving the system isn’t set in stone. It is an adaptive algorithm, meaning it “learns” and is driven by your most recent week or two of data. But there’s nothing you can do to direct the lesson. It studies your CGM readings and it knows how much insulin it has delivered. All on its own.

Even the user screen in Auto mode tells you almost nothing. It displays only a blue shield with your current sensor glucose number. In this way, the system harkens back to the original Guardian CGM of 2005. (In fairness, there is a button that can lead you to a status screen that displays your micro boli and trend info, but the quick trend look we are used to is not available to us.)

Safe mode: This is the in-between mode that CDE Gary Scheiner jokingly called “a sort of purgatory between Auto mode and Manual mode” in his excellent detailed operational review of the pump. As the 670G system is fully dependent on the sensor, if things go south with said sensor, Auto mode is shut down and you are moved to this technological purgatory where the system delivers insulin based on its memory of what it usually does the time of day you’ve had a failure. It will only do this for an hour and a half. If the problem with a sensor—which could be as simple as a late calibration stick or as complicated as a complete signal loss—isn’t resolved, the pump then drops back into Manual mode and basal delivery resumes based on your settings, so it’s important to have a good set of traditional settings in the pump, even though they won’t be used by the system in Auto mode.

One frightening side note, however, is that when the pump drops back into Manual, the low suspend features are disabled, and have to be turned back on manually. I find this a huge safety issue. If the pump drops out of Auto in the middle of the night, its best protective suite of features is off the table.

Rockin’ the Buffet: The 670G at Its Best

For me, the 670G really shone when it came to meals. During the time I was on the pump, my post-meal excursions were few, far less extreme, and far shorter than I traditionally witness. Even though I had no dual or square wave boli to deploy, the fact that the pump could constantly add insulin following a meal to suppress glucose surges gave me amazing postprandial control. In fact, I was able to “safely” eat a wide variety of foods that I typically forswear. Pasta, at one time a guaranteed blood sugar disaster for me, became commonplace in my diet with no blood sugar penalty, which had a negative unintended consequence: I put on 20 pounds in three months.

Meanwhile, another place the pump did a splendid job was in the control of overnight sugars, at least to its own definition of control (recall that it’s happy if you are 150 mg/dL). Still, on the pump my morning numbers were consistently in that range, not something I can say about other therapies I’ve used—pumps or pens—where I often saw a pretty wide range of dawn blood sugar readings.

And there’s still more good news: I had my A1C test just days after I the end of my 670G trial, and I saw an improvement on the system, although the number is still nothing to brag about: 7.0%. Still, given the system’s targets, this was about as good as I expected.

But it wasn’t paradise. In fact, there was more than one snake in the garden…

Out of Control, Without Controls

Oddly, for me at least, the 670G was absolutely worst at the one thing it’s supposed to be absolutely best at: Preventing lows.

In fact, its theoretical ability to improve safety from lows helped fast track the pump through the FDA. Wait… I shouldn’t say theoretical. The pump’s capability is well documented. It just didn’t play out for me, as I suffered low blood sugars nearly every afternoon. The lows struck even if I took the time to enter a higher temporary target blood sugar.

So how could that possibly be?

The lows were well outside the active insulin range of my noon meals, so they had to be triggered by the micro boluses delivered in the early afternoon. I suspect that this is a problem unique to me, as my afternoon basal rates on every pump I’ve even used are virtually nil. But still, I was surprised that the adaptive algorithm didn’t, well, adapt to my needs.

Not that it didn’t try. As I began to coast low each afternoon, the pump would shut down the micro-splats of insulin (giving me a daily alarm that no insulin had been delivered for two and a half hours and requiring a fingerstick to keep working) but it was always too little, too late. By the end of the afternoon I’d be scrapping 70 mg/dL.

To compensate, I took to under-bolusing lunches by entering falsely low carb counts—remember, no overrides allowed—so I’d run artificially high in the early afternoon. Of course, then the system spent the afternoon throwing insulin at me (now giving me an alarm that the maximum Auto mode delivery had been pumping for four hours and requiring a fingerstick to keep working).

If I under-bloused enough, the micro boluses couldn’t drive me down quite to a low by the end of the day, so it worked, but it was sloppy. And it was the first brick in a wall of distrust in the system, which is pretty much death for a relationship that requires a high level of trust to stick with. By the end of my trial, jittery about lows (I was having the most I’ve had for years) I was frequently second-guessing the pump at all times of day, and found myself often canceling a bolus that looked “too big” and going back and entering a lower carb count.

Needless to say, my control got worse and worse as the trial went on. But my fear of lows overrode my common sense.

So what did my doc and my trainer think was causing this? Well, let me fill you in there…

Training? What Training?

Going in to the trial, I made the decision NOT to monkey with the settings on my own.

First, I wanted to report accurately on the typical experience for our readers; and second, I wasn’t sure my knowledge of pumps would really be all that helpful with something so revolutionary. Put another way, I didn’t want to put my foot in it, so I relied on the pump trainer.

As it turned out, I was the first real, live person my pump trainer started on the 670G. She was fully trained by Medtronic Corp HQ, of course, but had no field experience, nor did she have diabetes. We had one session to cover the basics, another a couple of weeks later to engage the Auto mode, and she met me for one follow-up at my doctor’s office. She followed my data uploads—when I could get the damn thing to upload (it gave me fits every time), and emailed recommended changes, which I followed. My IC ratios got frighteningly aggressive—a type 2-like 1:9 for lunch up from my historic 1:15, but I was told this is a common experience with the system’s approach to handling insulin. When I emailed her about the lows, she responded that she didn’t see any that concerned her.

So what about my doc? Mine was the first prescription my endo wrote for a 670G. She had a short training on it as well, but I got the impression that it was more like an executive briefing. Rather than being a resource for me, she was eagerly looking forward to my thoughts on the device.

So I was largely on my own. Of course, to be fair, no one other than the folks involved in the clinical trials has any field experience with this device yet either—a possible reason there’s so much confusion about what it can and can’t do.

What about the Help Line? Were they a helpful resource? I did call Medtronic’s Help Line, but only to get assistance with uploading the pump to CareLink. They were friendly enough, but the call dragged on for an hour and a half with the tech mainly having me repeatedly de-install and re-install Java. In the end I had to switch to a different computer, although they could never figure out why. This is typical of my past experiences with MedT’s Help Line, so when things went south on me with the low blood sugars, I didn’t call in again.

In hindsight, I think the reason why was that I was too worn out by the 670G system to think clearly, as it’s the worst nag you can imagine.

Wake Up and Talk to Me

During my trial I was plagued with alarms, many of which I’m yet to understand. I carefully read the phone book-sized manual and fed the sensor the required number of fingersticks, but other factors lurking deep within the algorithm apparently trigger the need for feedback. Way too frequently, at 2am, 3am, or 4am the pump would wake me up, demanding a fingerstick to continue operation.

By the end of the first month, I was feeling frazzled.

By the end of the second, I was exhausted.

By the end of the third month, I was wiped out.

Decision Point

And before I knew it, the trial period was up and I had to decide: Was I going to stay on the 670G? Well, from the start of this story you know I didn’t. And it wasn’t any one thing. Here was my thinking:

  • To use a system that gives you so little control, you have to be able to trust it 100%, and I was never able to develop that level of trust.
  • The acres of tape that held the sensor and transmitter in place gave me terrible rashes.
  • Previously a pump lover, I developed a hatred for having something attached to me 24/7. I don’t know why. Maybe because it disrupted my sleep so much that I began to project anger on it.

But on top of all of that, in deciding whether or not to purchase one for myself, I had to ask myself: What’s in the pipeline? What will the next 4 years bring? Committing to a pump means committing to a relationship that lasts almost as long as the typical marriage. I just couldn’t bring myself to make that kind of commitment.

So how’d those hard-to-get sensors play out over my trial? Well, first off, I had no supply issues, as I received all three boxes for the trial from the get-go. I reported scary-good results with my first few, but after that I saw a drop in accuracy against the calibrations.

Still, the sensor stability was good overall, especially in the 150-200 mg/dl range. For me anyway, they were less accurate at lower numbers, and most mornings the 670G’s integrated CGM thought I was 15-20 points lower than the Bayer Contour fingersticks used to calibrate it. Playing around with the number of calibration fingersticks didn’t change this trend, but still, other than the one event previously reported, I didn’t have any readings that were wildly off.

A Good Choice… for Others

So I’m back to insulin pens, for now.

While I was happy to be rid of the 670G, with its alarms, its lows, its nasty tape, its emotionally painful uploads—I was terribly bummed. I wanted it to work for me. And in many ways, it did.

But now that I’ve caught up on my sleep, what do I think about the system in hindsight? I think that for anyone going on an insulin pump for the first time it would be foolish to choose any other pump. After all, the 670G has some amazing capabilities as an automated system, but it can also be used as a garden-variety pump, and a damn fine one. Given that, and given our increasingly paltry options, I think that if you are starting a pump for the first time, the 670G is a fine and very logical choice.

But what if you are currently pumping with a traditional pump, should you switch or should you wait?

That’s the tough call. I suspect most people will do better on this system than on other pumps, especially if they can just “let go” and let the system run the show, but I don’t think it will be the only game in town for long. So really, it comes down to how much you’d rather gamble on the future compared to how much you need extra help today.

Editor’s Note: We thank you for sharing your honest opinion as always, Wil. While Your Diabetes May Vary (and opinions too), it’s certainly helpful to learn about your real-world experiences.