Continuous Glucose Monitors (CGMs) are a relatively recent development in diabetes control. They are still controversial with many insurance companies, though prevention of even one high-cost ambulance trip to the emergency room should make them cost-effective. But they do have problems.

There are three devices that I am aware of, worldwide, and others in development. The best, according to users, is the Navigator – but it has been withdrawn from the US market. This leaves the Dexcom and the Medtronic sensors. My own experience has been entirely with the Medtronic sensor, but all three work in basically the same way, on interstitial fluid, and all are subject to the same problems of time lag.

Medtronics transmitter (white) and the visible part of the sensor (clear) with a U.S. quarter for scale. I use IV 3000 to hold everything in place.

The sensors available today consist of some kind of reusable transmitter, a receiver, and a disposable needle which injects a small disposable sensor under the skin. The sensor is then connected to the transmitter, which lies against the skin.

The sensors are expensive. The price today at the Medtronic store is $42 a sensor, with an official life of 3 days. (Most people actually manage to use them longer, and in fact I find they are most accurate on days 3-5, though I rarely get one that lasts more than a week.) The big advantage of the Medtronic system is that the Medtronic insulin pump (the kind I have) acts as the receiver, though separate receivers are available. The official life span on the Dexcom sensor is 7 days, though I understand they are correspondingly more expensive. Comments from users of the Dexcom are solicited!

A continuous sensor is extremely useful in that it allows you to see how your blood glucose varies between finger sticks, and also allows you to see whether it is rising, falling or staying reasonably constant. The Medtronic sensor allows you to choose whether to see a 3-hour, a 6-hour, a 12-hour or a 24-hour graph of your blood sugar, or download values about 5 minutes apart to a computer, and I believe the Dexcom sensor is similar. On the negative side, CGM accuracy is highly variable, and it needs regular calibration with a finger-stick meter. There is a further problem (at least with the Medtronic system) with alarms.

I used to eat protein bars for breakfast. They had a good balance of protein, carbohydrate and fat, they took no weighing, and they were consistent day to day. They also produced a considerable spike in blood sugar in the hour after breakfast, easily identified on the CGM graph but missed in finger-stick testing. When I switched to Greek yogurt with fruit, also a good balance of protein and carbohydrate but without the fat, my blood sugar graph leveled out. In fact, the CGM has allowed me to identify many regular meals as causing spikes or delayed rises in blood sugar. (The fat in the bars was not the problem; in fact fat delays the absorption of carbohydrates, but much of the carbohydrate in the bars was fast-absorbing sugars.)

I am not alone in finding trends more useful than absolute numbers. A high blood sugar (within reason) is not really a problem if the blood sugar is decreasing, nor is a moderately low reading with an upward trend. On the other had, a high reading going up or a low reading going lower is cause for immediate finger-stick confirmation and treatment if the reading is confirmed. A low reading at night or if I’m planning to drive is always reason for finger-stick confirmation.

What the sensor actually measures is a slight electrical signal, and this must be calibrated with a finger stick measurement to get the actual blood sugar reading. The calibration constant changes over the life of the sensor, and can drop very fast toward the end of sensor life. Consequently the sensor must be calibrated a minimum of twice a day, preferably when your blood sugar is fairly steady, and may give a lot of false low alarms near the end of its life.

Another problem is that the sensor measures the glucose concentration in the interstitial fluid, that is, the fluid between the cells. While this follows blood glucose, there is a significant time lag. My blood sugar can drop very rapidly, especially during exercise, and as a result CGM measurements during exercise can be much higher than my actual blood sugar. This time lag alone makes me very dubious about using this type of sensor for any closed-loop artificial pancreas.

My biggest complaint with the Medtronic system is the alarms. Ideally, to be most useful the CGM system should wake you up if your blood sugar starts to drop. In fact, the alarm is so low in volume and so high in frequency that I almost never hear it unless I am in a very quiet environment and there is only a layer of cloth between my ears and the pump, not a down comforter. This is partly my hearing, as others will sometimes point out to me that I am beeping. Yes, it does start vibrating eventually if I do not respond, but I feel the vibration only if the pump is tight against my skin. I usually feel it driving (well after I start going low) but often not at night.

Is it worth it? Yes, for me. But it definitely needs some improvements.