Do insulin pumps and continuous glucose monitors actually improve outcomes?

Nearly every day in my practice, a patient with diabetes asks me whether he or she should switch from multiple daily insulin injections to an insulin pump.  I often have a discussion with patients about whether or not they should be using a CGM (continuous glucose monitor) to help monitor blood glucose instead of just using SMBG (self-monitoring of blood glucose).  As an endocrinologist, it is very important to be able to advise patients about specifically what these new technologies have to offer them.  Do they decrease mortality?  Do they decrease long-term diabetes complications?  Do they improve glycemic control?  Do they improve quality of life for patients?  Do they lower costs?  All new medical technologies need to undergo a rigorous evaluation and testing with these types of questions in mind.  This is critical not just so that I can be honest and helpful to my patients, but also from the overall perspective of the healthcare system.

In that vein, Yeh et al recently published a meta-analysis in the Annals of Internal Medicine called “Comparative Effectiveness and Safety of Methods of Insulin Delivery and Glucose Monitoring for Diabetes Mellitus: A Systematic Review and Meta-analysis.”

This meta-analysis, funded by AHRQ, looked at the differences between:

  1. MDI vs CSII (multiple daily injections vs continuous subcutaneous insulin infusion)
  2. Type 1 vs type 2 diabetes
  3. SMBG (self-monitoring of blood glucose) vs rt-CGM (real-time continuous glucose monitoring)

What types of studies did they include in their meta-analysis?

  • Studies of adults, adolescents, or children with type 1 or type 2 diabetes mellitus
  • Studies from 1966-2012
  • 19 studies comparing CSII with MDI (>3 injections per day of either basal/bolus insulin or NPH/regular)
  • 10 studies comparing CGM with SMBG (>3 fingersticks per day)
  • 4 studies comparing SAP (Sensor-augmented pump) use with MDI + SMBG

* Studies were excluded if regular insulin was used in the CSII (pump) group (they felt this to be a weakness of prior analyses)

Here is the key data table:

A few things pop out from this table:

  1. Overall, they assessed the strength of evidence as relatively weak.
  2. In children and adolescents, CSII showed no difference in clinical outcomes from MDI.  CSII was better in terms of quality-of-life.
  3. In adults with type 1 diabetes, CSII led to more symptomatic hypoglycemia, but better hemoglobin A1c and quality-of-life.
  4. There were no differences between CSII and MDI in adults with type 2 diabetes.
  5. CGM, whether with an insulin pump or not, led to a benefit in glycemic control without any difference in hypoglycemia.

Some concerns and words of caution when interpreting these results:

  • Meta analyses can always suffer from publication bias.  That is, studies are much more likely to be published if they show positive results.  So it is possible that studies have been done that generated results that would have shown no difference between the two methods being studied, but these may never have been published and thus cannot be included in the meta-analysis.
  • These studies all had durations of 12-52 weeks.  There were no studies reporting on long-term outcomes like micro or macrovascular disease.
  • 24 of the articles (approximately 2/3) were supported by pharmaceutical companies

What does this mean?

According to this meta-analysis, CGMs did improve glycemic control.  Insulin pumps did not appear to have a significant effect on clinical outcomes, but did positively effect quality of life.  Remember that the studies included were all between 12 and 52 weeks, so one major limitation is that any longer-term effects would not be teased out.

While some may discount the quality of life improvements seen with the pump as being less important than clinical outcomes, I caution people from doing so.  In a condition as omnipresent as diabetes, maintaining good quality of life for the patient is critical and a very important goal.

In the end, the decision about whether or not to use one of these devices comes down to a conversation with the patient and their family, based on their personal preferences and what each device might offer them in terms of benefits and harms.  This meta-analysis adds some more information to that conversation.

Finally, this meta-analysis shows that we simply need more data to study so that more concrete conclusions can be drawn.

Physician empathy associated with lower rates of diabetes complications


I’ll admit that I’ve not yet read this journal article to form my own conclusions, but I found the headline interesting nonetheless.  Here is a link to the NY Times article and a link to the original journal article in the Journal of Academic Medicine.  It makes intuitive sense… if your physician has a higher level of empathy, you are more likely to form a positive treatment relationship, and the patient is thus more likely to find meaningful and useful treatment recommendations from that relationship, and will end up with fewer acute metabolic complications.  Definitely adding this paper to my journal article reading list.


3 Thoughts About The V-Go Insulin Delivery Device

I had a chance this week to spend some hands-on time learning about the new V-Go insulin delivery device from Valeritas.  Valeritas’ website states that “The V-Go is engineered to simplify basalbolus insulin therapy for the millions of people suffering from Type 2 diabetes.”


The facts:

– This is the first disposable insulin-delivery device that will give basal-bolus insulin.

– It is mechanical, containing no electronics.

– It is designed to be worn for 24 hours before it needs to be refilled with insulin.

– The needle is a 30 gauge needle that stays in the user while the V-Go is in use.

– The V-Go comes in 3 “sizes”: one that delivers 20 units of basal insulin over 24 hours, one that delivers 30 units, and one that delivers 40 units.  Each device is also capable of giving bolus insulin in 2 unit increments up to 36 total units of bolus insulin per 24 hour period.

Three thoughts about the V-Go:

1) Convenient: This device appears well-built and relatively easy to use.  It is about the size of an Omnipod, and because it is placed on the user for 24 hours straight, allows her to leave insulin at home when going out for the day.  The device needs to be refilled every 24 hours with rapid-acting insulin.  So, the user need only take fingerstick/testing supplies out with her when going to work, running errands, or going out to meals.  It is also somewhat elegant that there are no electronics in the device.

2) Not enough insulin for some patients: Though this device promises convenience, the amount of insulin that can be delivered is too little for some people with type 2 diabetes.  The most that can be delivered is 40 units of basal insulin and 12 units of bolus insulin per meal (totaling 36 units of bolus insulin a day).

3) Inability to titrate: The device comes in 3 sizes, each delivering a set basal amount of insulin over 24 hours.  These are either 20, 30, or 40 units.  While this may be useful for a patient who has already been on a steady dose of basal insulin, it does not give much flexibility to titrate doses.  Once a patient has paid her co-pay at the pharmacy and picked up a month’s supply, she would have to pay again to switch to a different basal rate.

I think this device represents an interesting start towards a market that will likely increase rapidly in the next few years of devices intended to deliver insulin to people with type 2 diabetes.  Given the above limitations, I think it’s overall usefulness will prove limited, but I look forward to seeing what future iterations and generations of devices will look like.

What do you think?  Would you want to try this (either patients with type 2 diabetes or providers caring for patients with type 2 diabetes)?

Quantified Self and Diabetes… the Perfect Match

Take a few minutes and watch this video of Jana Beck from a Quantified Self meeting as she explains how she took data from her Dexcom CGM (continuous glucose monitor) and created her own data visualizations (Thank you to Russ Cucina for sharing this video with me).  I think that the visualizations she created are very cool and obviously helped her in her journey to try out a new method of managing her diabetes with a low carbohydrate diet.

Beyond this particular video, I am also very excited about the Quantified Self movement and its overlap with diabetes.  Type 1 diabetes is a disease that requires patients to monitor their physiologic status on a frequent and routine basis, from food intake to activity levels to glucose levels.  Many people in the Quantified Self movement are voluntarily doing much of the same thing.  While these people never have to have the same concerns of a person with diabetes that, if they feel like “slacking off” for a day, something might go horribly wrong, they are at least starting to develop some empathy and interest.  There is a fantastic synergy here, introducing a new cadre of talented, engaged, and enthusiastic people to the field of diabetes technology.  This is happening whether or not they realize it!  This entire group of people are trying to monitor their every action and learn from the data visualizations… precisely the thing we try to help people with type 1 diabetes do!  I’ve not yet been to one of the QS MeetUp events in San Francisco, but plan to go sometime soon.  I am convinced that advancements in diabetes care will come from the QS movement, whether intentional or accidental.

Tip: If you already know about diabetes and CGM devices, you can skip to minute 6 and start there.  The visualizations start at about minute 8 of the video.

A recent reminder of the limits of diabetes technology

As excited as I am, and many people are, about making major strides in improving the capabilities of diabetes technology, we have to keep in mind the limitations of technology.  There is often a time and a place where innovations can improve the daily life and “workflow” of a person with diabetes, but there are situations where face-to-face human encounters will never be replaced.  I had a recent reminder of this.

At one of the locations where I work, I do regular video-chat appointments with people with diabetes who live in remote areas.  At one such visit recently, I spoke with a patient, Gloria, who was using hundreds of units of basal insulin a day and hundreds of units of bolus insulin a day.  Gloria lamented the fact that this amount used to control her diabetes well but no longer was doing the trick and her hemoglobin A1c was now over 14%.  There are many considerations in a situation like this, including ruling out an infection as the cause of her high insulin doses “no longer working,” and after thinking this through, I was tempted to proclaim her situation due to worsening diabetes and insulin resistance.  I was ready to switch Gloria over to U500 insulin (five times the concentration of “normal” U100 insulin… think about the new forms of laundry detergent where you need less fluid to do the job).  This is a somewhat drastic step because of the risks of confusion on the part of the patient but especially the rest of the medical community over her insulin doses, since U500 is relatively infrequently used.

Before switching Gloria over to U500, I asked her to make the multi-hour drive to come see us in clinic, in person, face-to-face.  I wanted to make sure she was actually using her insulin properly before ramping up her dose.  She met with our excellent diabetes educator, who discovered that, in fact, Gloria was making several errors in how she was injecting her insulin.  Over the next two weeks, without any changes in insulin dose, Gloria’s blood sugars came down from 300s and 400s to 100s and 200s, and we avoided needing to switch to U500.

Sitting together with the patient and diabetes educator, watching the patient go through the end-to-end process of injecting her insulin was not something we could do over a video-chat from hundreds of miles away.  There was nothing that could have replaced this in-person, and personal, interaction.  Even as a technology proponent, I smile at that simple fact, because it is the human interaction that led me and many others to become physicians.  When designing new technologies, or thinking about how to implement existing ones into practice, it is important to remember that human interaction cannot be replaced in every situation, nor should we strive to do so.  Technology should serve to augment and enhance these interactions.

A fantastic rant about diabetes technology from Scott Hanselman

This fantastic rant about the frustrating state of diabetes technology from Scott Hanselman, a type 1 diabetic, has been making its way around the blogosphere and a few of my email chains.  In his blog post, he decries the slow pace at which diabetes technology is moving, showing an example of a program he wrote for his PalmPilot in 1998 that was able to give him in-depth analysis of his blood sugar management.  He correctly points out some of the major technological issues that people with diabetes still suffer from today, including less-than-optimal accuracy of blood sugar readings, a lack of standards and interoperability, and a lack of useful wireless technology.

Scott is dead-on in the most critical respect here: The typical workflow that a type 1 diabetic still has to endure to acquire his or her glucose values, transmit/download the values, collate values from different devices, and analyze the values is entirely too cumbersome, slow, and inefficient.  The current diabetes technology industry has done little to solve this.