The MiniMed 670G from Medtronic is an insulin pump coupled with a glucose sensor. It uses a computer program (called an “algorithm”) to automate certain aspects of insulin delivery. I decided to try the 670G partially out of professional interest (everybody and their great aunt have been asking for my opinion on the system), and partially out of personal interest, as my blood glucose control hasn’t been the greatest the past couple of years.
Let me start out by saying this:
Since I started using 670G, my overall blood glucose control is better.
I have to keep reminding myself of this non-inconsequential fact, because every day I find things about this system that I don’t particularly like.
In my opinion, the pump itself leaves a lot to be desired. There are so many features and so many menus and so many safety/confirmation steps that my button thumb is starting to form a blister. The color screen is nice, but not large enough to display everything it needs to display. And the freakin’ clip is upside down. The reservoir connector pokes me in the gut every time I bend over, and I have to unclip it to see the screen and programming menus in the proper orientation.
The “hybrid closed loop” part of the system (what I prefer to call the “semi-automatic feature”) is what makes 670G special. It functions by making adjustments to the BASAL insulin based on data received from the linked glucose sensor. Since previous-generation Medtronic sensors (Sof-Sensor, Enlite) were, shall we say, less than optimal, I went into this highly skeptical. After all, if the sensor isn’t feeding the pump a steady stream of reliable data, what’s the point?
I’ve been pleasantly surprised by the performance of Medtronic’s new Guardian sensor.
In my experience, the Guardian sensor tends to under-estimate glucose values (sometimes significantly) and the overall accuracy still falls short of the Dexcom G5. And it still requires three hands and four different forms of tape to secure it to the skin, along with four to six fingerstick calibrations per day. But the insertion process is quick and painless and the transmitter signal almost never misses a beat. Overall, Guardian represents a major improvement over Medtronic’s previous CGM systems.
The algorithm that determines the basal insulin doses makes adjustments every five minutes based on recent sensor glucose measurements, insulin delivery and meal (carb) entries. The algorithm adapts on a daily basis based on what took place over the previous six days. So there are both reactive (based on what’s already happened) and proactive (predicting what will happen) components to the algorithm.
However, the algorithm is highly conservative in nature, doing everything in its power to keep the user far, far away from anything resembling hypoglycemia. The basal algorithm targets a glucose value of 120 mg/dl, and there are limits to how long the system can deliver zero basal insulin as well as the magnitude and duration of maximum basal delivery. The maximum rate of basal delivery is based on a rate that (the system believes) would cause the glucose to drop below 70 if it was to run for about 8 hours continuously.
Think of it this way: The max basal in auto mode is sort of like not allowing the captain of the ship to turn the rudder too sharply for fear that the ship might sway too much. Of course, when you have a big ship and a small rudder, that isn’t likely to happen. So why the ultra-conservative approach? Most likely because regulatory authorities such as the FDA like it that way. Less risk of hypoglycemia = less chance for negative feedback and legal issues. But it does make the system less responsive than it could be.
For those who normally have a significant difference between their “peak” basal and their “valley” basal, perhaps due to a pronounced dawn phenomenon, the algorithm’s max basal may not be enough to meet peak basal needs.
By the way, bolus doses are NOT automated. Users must still enter their carbs into the pump’s bolus wizard calculator as well as the blood glucose value. But the user’s original sensitivity (correction factor) and target settings are NOT applied. The system calculates correction doses based on a conservative target of 150 mg/dl and a sensitivity/correction factor that it determines on its own.
Despite these shortcomings, I’m seeing less overall variability (more time in-range, fewer extreme highs & lows), and my overnight glucose levels are more stable than they’ve been in years. It’s rare that I don’t wake up within a reasonable range in the morning.
The overnight part makes perfect sense. Because the system adjusts the basal delivery whenever glucose levels are trending up or down, it is able to alter the glucose pattern over the course of many hours and keep things in-range… assuming nothing else interferes.
It reminds me of a large cruise ship. When the ship is moving at high speed, a small rudder will not allow it to change directions quickly enough to avoid things like icebergs (this is what happened to the Titanic).
Not that 670G is going to cause anyone to crash and sink, but there are similarities in terms of what it can and cannot do to self-regulate blood glucose levels. On long stretches of open water, a big cruise ship with a small rudder can get you from port to port very effectively. During the journey, if the ship starts to veer off course, it has enough space and time to get back on course.
670G does the same thing: Over the course of many hours, with nothing affecting glucose levels other than the liver’s usual glucose production, the basal insulin can be adjusted to offset subtle ebbs or flows in glucose levels. Start to rise a bit? The basal delivery increases so that over the next several hours, you’re back on track. Start to drop a bit? The reverse takes place. As a result, you wake up pretty close to the system’s set point of 120 mg/dl most mornings.
Here Come the Icebergs
Everyone who lives with diabetes knows about the hour-to-hour, minute-to-minute challenges we must contend with. These are the icebergs – the things the system must navigate around in order to prevent extreme highs and lows.
Essentially, anything that can cause a rapid, abrupt rise or fall in glucose levels represents an iceberg. Remember, we’re dealing with a huge, fast-moving ship with a small rudder.
The pump’s basal adjustment just isn’t powerful or responsive enough to prevent high and low glucose levels when confronted with icebergs.
The icebergs include:
- Food (particularly rapidly-digesting carbohydrates)
- Physical activity (particularly heavy exercise)
- Stress (particularly sudden, unexpected crises)
- Sudden hormone changes (resulting from injuries/trauma, rebounds from lows)
- “Manual” Bolus insulin (from injections or inhaled insulin)
There are also several limitations to the system which hinder its ability to keep glucose levels within range continuously:
- Limits to the automated basal adjustment. As described earlier, the automatic basal adjustments are limited in terms of magnitude and duration. The “maximum basal” is relatively conservative and can only run for a maximum of 4 hours. The “minimum basal” (zero delivery) can only run for a maximum of 2.5 hours. Sometimes, stronger and longer adjustments are truly needed to achieve and maintain in-range glucose levels, these limits can be… well… limiting.
- Sensor reliability. Even when the system is properly calibrated, there will be times when the sensor reports incorrect/inaccurate information to the pump. And the algorithm is only as good as the data being fed into it.
- Absorption issues. Just because a dose of insulin is delivered below the skin does not mean that the insulin reaches the bloodstream as expected and works as expected. This applies to both basal and bolus insulin. Users must change sites frequently/regularly, rotate properly, and troubleshoot effectively.
- Undetected pump problems. There is a long list of things that can go wrong! Air in the tubing, displaced infusion sets, partial occlusions, partially (or completely) spoiled insulin, leakage, and unintended disconnection just to name a few.
- Times out of auto mode. “Auto-mode” is what we call it when the system is automatically adjusting basal insulin based on data fed into it by the glucose sensor. There are several situations in which the system will phase out of auto-mode. There is a transition phase called “safe mode” that initiates under these conditions. Safe mode can run for up to 90 minutes, with a flat basal delivery and no automated adjustment. If the issues noted below are not resolved within 90 minutes, the pump returns to “manual mode”, which means that everything returns to the user’s default settings.
Auto mode is exited if:
- The sensor is in its warm-up phase
- The sensor is not communicating with the pump
- The sensor requires calibration
- There is a significant discrepancy between the sensor and the calibration value
- “Minimum” basal delivery has taken place for 2.5 hours
- “Maximum” basal delivery has taken place for 4 hours
- Glucose is above 300 for more than an hour
- Glucose is above 250 for more than 3 hours
- An occlusion is detected
Despite these “icebergs”, I have managed to achieve better glucose control with 670G than I could achieve on my own using a pump and non-integrated CGM.
Tying the Captain’s Hands
To compound some of the challenges, Medtronic took away certain tools that many consider essential to good self-management. This is like taking away resources that the captain of that big, fast-moving cruise ship would normally use to circumvent the icebergs.
Here are a few of the missing pieces:
- Extended Bolus. When in “auto mode”, users no longer have the option of extending delivery of their boluses. Extending boluses is a useful tool when consuming large portions, slowly-digesting (low-glycemic-index) foods, or meals that take considerable time to consume. It is also essential for anyone with impaired digestion (gastroparesis). Under these conditions, “normal” boluses will peak too early, resulting in hypoglycemia soon after eating followed by a delayed rise. Medtronic may have assumed that the 670G’s basal adjustment would prevent this problem, but it doesn’t. It may reduce the severity of the fall-followed-by-rise pattern, but the pattern still exists. Users must think long and hard about timing their boluses properly to avoid this type of pattern.
- Manual bolus adjustment. Auto mode does not allow users to adjust bolus doses once they have been calculated by the pump, and manual boluses are not an option. Planning to exercise? Tough cookies. Take what we say. Sensor shows a distinct rise or fall at the time of the bolus? Too bad. Take what we say. To alter the dose, users are forced to “lie” to the system by intentionally entering incorrect carb amounts or glucose values.
- Correction doses in “auto mode” are determined by the system, not the user. And the system does not target a glucose of 120 (as does the basal algorithm); it targets 150. Why? Most likely to appease the FDA and get the product to market as soon as possible. The system also does not base correction doses on the “sensitivity” or “correction factor” that the user (and their healthcare team) may know to work, but on a mystery-clad component of the system’s secret algorithm sauce. Despite all that, because the algorithm typically cranks up basal delivery prior to the need for a correction bolus (since the glucose is rising and elevated), the correction boluses sometimes overshoots and leads to lows.
- The only allowable adjustment to the basal algorithm is a temporary target of 150 rather than 120. Presumably, this is for exercise. Trouble is, it takes a couple of hours for this adjustment to actually get the glucose up to 150 (so it has to be set a few hours in advance of exercise), and it isn’t sufficient to prevent hypoglycemia with longer and more intense forms of physical activity.
- Temp Basal. User-set temporary basal adjustments are not an option in “auto mode” since the system adjusts the basal on its own. And there are limits to how much (and for how long) the system will raise/lower basal delivery. If the user knows that their basal needs are going to be significantly higher (or lower) for a period of time, it makes sense to be able to inform the system before glucose levels start trending out of range. For example, having a high-fat restaurant dinner always causes my glucose levels to rise throughout the night. I know that I can temporarily raise my basal rate to deal with this, but the 670G doesn’t allow for this kind of input. Instead, it waits until the glucose levels start rising (usually several hours after the meal), and then, after four hours of delivering what it considers my “maximal basal,” it alarms and shifts out of auto mode and into “safe mode” where the basal delivery is even lower than what it had been delivering.
A creative captain can figure out ways around daily challenges. Users have the option of switching in and out of auto-mode if they want to have access to things like temp basal settings, square/dual boluses, and self-adjustment of bolus doses. I often switch out of auto-mode during the daytime for this reason. One can adjust the timing and amount of boluses doses when delayed food absorption is anticipated. Delivery can be suspended temporarily and carbs can be consumed before/during exercise to keep from dropping low.
Trusting the 670G system, combined with some creative self-management techniques, has led to improved glucose management.
The Ideal Passenger
So who can benefit most from 670G? When functioning properly, it pushes the user towards glucose values in the 130-160 range, and A1c’s in the high 6s to low 7s. If this represents a nice improvement for you, then go for it. If it represents a step backwards, or is not consistent with your personal goals, then you might want to pass.
Trust me, passengers on this cruise ship don’t get to lie around on lounge chairs and be waited on hand & foot. This is a working cruise! Expect to spend a fair amount of time in the kitchens and engine rooms. You’ll need to manage/maintain both the glucose sensor and the pump. Plan to perform fingersticks four to eight times a day for sensor calibrations and to keep the auto-mode feature running smoothly. Expect more alarms and alerts than usual. You’ll need to plan well-ahead to prevent lows (and highs, on occasion) when exercising. And you’ll still need to manage things the old-fashioned way when out of auto-mode (typically 10-20% of the time), and that means making sure your basal and bolus settings are properly fine-tuned.
It also takes a shrewd captain to navigate this ship. Using the 670G and its full array of automated delivery features requires a great deal of training, aptitude, and attention to detail. Compared to traditional pump therapy (with or without an independent continuous glucose monitor), 670G can be quite complex. Training typically requires three separate sessions: one to learn the pump, one to learn the sensor, and one to learn the automated basal delivery system (auto-mode). One must understand all of the conditions that are necessary for auto-mode to activate. There is even an “auto-mode readiness” status screen that is nine items long! You’ll need to know when/why the system may go out of auto-mode, and how to get back in. And then there’s “safe mode” – a sort of purgatory between auto-mode and manual mode in which the pump is delivering basal insulin but not self-adjusting it until an auto-mode problem is fixed. I’m still trying to wrap my brain around that one.
People who maintain a structured and predictable lifestyle also tend to succeed with 670G. There is much less chance of a cruise ship mishap in the Caribbean than in the Arctic Circle because of the lack of icebergs. Someone whose life includes frequent/inconsistent meals & snacks, random physical activity, changing work shifts, varied sleep schedules or significant stress is going to struggle in auto-mode. By contrast, those who eat consistent/well-spaced meals (with minimal snacks), work out regularly (or not at all) and lead generally relaxed and predictable lives tend to stay in-range most of the time in auto-mode. And for that matter, those whose basal requirements include a significant peak & valley (based on previous pump use & basal fine-tuning) may struggle a bit due to the limited basal variations that can take place in auto-mode. Those with relatively stable basal requirements throughout the day & night are more likely to experience success in auto-mode.
One other virtue that can’t be understated is PATIENCE. Don’t expect your diabetes worries to vanish the moment the box arrives from Medtronic. The company is struggling to meet sensor supply demand and furnish certified trainers, so you may have to wait a while to get started. Once you get started on the pump, you’ll likely need to wait for a second round of training to get set up on the sensor. And the system’s algorithm will require several days in “manual mode” before it has enough data for you to switch over to auto-mode (which requires yet another training session). Even then, the system continues to self-analyze data in order to improve its performance, so it might take several weeks before it reaches peak performance. Bottom line: patient people ultimately reap the benefits of the system.
To succeed in “auto mode” one must:
- Have conservative glucose management goals
- Put in a certain amount of work
- Possess some technical aptitude
- Lead a relatively structured life
- Be patient in terms of expectations
Tips for Insulin Pump Users
To get the most out of a cruise, it helps to know someone who knows cruises. They’ll guide you on things like avoiding “inside” cabins (unless you enjoy claustrophobia), hitting restaurants during non-peak hours, and snagging poolside lounge chairs early. Here are a few insider tips to help you get the most from 670G:
- Change the sensors weekly. The Guardian sensors are designed for seven days of use, and you’ll do well by abiding to that. Unlike Dexcom sensors (which almost beg to be reused), Guardian sensors often start to deteriorate after the first week of use. And since the sensor is directing your basal insulin delivery, it just isn’t worth the risk.
- Call on your Medtronic trainers and clinical specialists for ongoing assistance. They have insight-a-plenty to help minimize the system’s downsides, shorten your learning curve, and ultimately extract the most possible benefit.
- Regular use of 670G has been shown to reduce the frequency and severity of hypoglycemia, but don’t count on the system to fix your lows once they have occurred. Cutting back (or turning off) basal insulin will help to raise the blood sugar, but it won’t do so quickly. It usually takes a few hours for this to occur. By that time, you could get into an accident, have a seizure, or worse. Use rapid-acting carbs to treat your lows. And when you treat your lows, do so conservatively. Since low blood sugars will usually coincide with a marked reduction in basal delivery by the pump, you won’t need as much carb as usual to bring your lows back up to normal.
- Pay attention to the timing of your boluses. Glucose levels are likely to spike just as much in auto-mode as they do in manual mode (or with any other system) if the boluses are given too late. For most meals and snacks, it is still necessary to bolus 15-20 minutes in advance. However, without an extended (dual/square) bolus option, bolusing for slowly-consumed, high-fat or low-glycemic-index foods, the user will need to delay or “split” their boluses in order prevent a post-meal drop and post-post-meal rise.
- Plan ahead when exercising. The only auto-mode feature that can help at all is the “temporary target” of 150. If you want to use this, it will need to be set at least 1-2 hours prior to the workout. Setting it at the onset of exercise will do no good since it takes a couple of hours for basal changes to influence glucose levels (remember: big ship, tiny rudder). And if you expect a delayed drop following your heavier workouts, keep the temporary target of 150 in place for several hours post-exercise.
- Download your data, and work with your healthcare team to analyze the results. Carelink Personal and Carelink Pro will let you see where improvements are taking place and where adjustments may needed. The two variables that you have control over in auto-mode are the carb ratios and active insulin time. Carelink reports will help you to optimize both of these settings. For people who did not fine-tune their basal settings through a series of fasting tests prior to using 670G, it is likely that your basal delivery will be reduced when in auto-mode. This, in turn, usually means that carb ratios will need to become more aggressive – something to watch for.
Issues With the Entire Fleet
Interestingly, I don’t believe the limitations inherent to 670G are limited to this particular system, although shortcomings to 670Gs algorithm and elimination of certain self-management tools may amplify the challenges. ANY hybrid closed-loop system that relies on basal adjustment alone is going to struggle to avoid the daily icebergs. My hope is that future iterations of the Medtronic hybrid closed-loop system and those developed by Medtronic’s competitors (Insulet, Tandem, Bigfoot and Beta-Bionics to name a few) will offer improvements such as:
- Ability for the user to customize how aggressive or conservative the algorithm functions
- Allowing the user to implement traditional advanced features such as extended bolus delivery, temporary basal rates, and the ability to override the pump’s bolus calculations
- Simplified and more logical pump programming, with fewer confirmation steps and overlapping menus
- Improved sensor accuracy with reduced calibration requirements
- Incorporation of faster-acting insulin, such as Novo Nordisk’s Fiasp. This will essentially increase the size of the ship’s rudder and allow it to do a better job of avoiding the icebergs
It would be nice to have a system that performs well in long stretches of open water AND navigates the icebergs deftly, and does so with minimal inconvenience to the user. Nevertheless, we have some very good options already. Unless you’re a subject under the United Healthcare dictatorship, you have CHOICES. You can pursue the 670G right away, deal with some of its shortcomings but enjoy the advantages of semi-automated delivery right away, and upgrade to its next-generation (new and improved, to some extent) system in a couple of years. Or you can go with another pump company, enjoy a pump that may have fewer drawbacks, utilize a separate CGM, and wait a couple of years for them to catch up to (or bypass) Medtronic on the hybrid closed-loop front.
The 670G represents an important step towards fully automating glucose control. However, it is important to put it in the proper context and set expectations at an appropriate level. Despite the system’s many limitations, it will improve the quality of life for many people affected by insulin-dependent diabetes.
And that’s saying something.