Notes taken during the “Continuous Glucose Monitoring (CGM) in the Management of Diabetes in Pregnancy” session at the American Diabetes Association conference:
Use of CGM in women with T1D
Elisabeth Mathiesen
- Sharing experiences from randomized control trial of using CGM in women with T1D
- 50% of babies with t1d have babies with macrosomia
- 20% still deliver preterm
- 10% have severe neonatal hypoglycemia
- High glucose in the mother induces fetal insulin production, excess growth of all tissues, excess fat deposition, and subsequent complications.
- But with t1d, the additional issue is hypoglycemia– 45% of 108 women with t1d had severe hypoglycemic incident, often early in the pregnancy
- CGM study in Copenhagen, 71 pregnancy mothers, 65% of whom had t1d. Baseline A1c of 7.3%
- Notably, an older CGM– not real-time, but retrospective. 5 – 7 days of use and then review.
- Dramatic improvements– a1c of 5.8 versus 6.3, macrosomia rates of 30% versus 60%
- Next, technology develops. CGMs with alarms (Medtronic). Start with alarm at 3.5mmol/L (63 mg/dL); adjust for individual women.
- What about hyperglycemia alarms? Many women and nurses want them. But a problem: alarm fatigue. Values > 10 mmol/l (180 mg/dL) > 20% of the time. Women turned off the alarms.
- There are challenges to decision support– under-bolusing looks similar to over-bolusing until 60 minutes later. Have to wait to figure out what the proper response is.
- Further, the women saw a mean difference between blood glucose and CGM glucose of 29 mg/dL
- A trial to determine usefulness of CGM in T1D.
- Primary endpoint: large infants for gestational age (>90%)
- Secondary endpoints: preterm babies, neonatal hypoglycemia
- 154 women with singleton pregnancy. Routine pregnancy care. Self-monitored plasma glucose ~4x a day. Dose insulin accordingly
- Wore CGM at 8, 12, 21, 27, 33 weeks
- Use Medtronic Guardian Real Time, because that was the one easily available in Denmark at the time
- All women provided with education and support, as well as with instructions for dosing
- Goals
- Nocturnal: 4 – 6 mmol/l (72-108mg/dL)
- <= 1 mild hypoglycemia per week
- Daytime: 4 – 6 mmol/l pre-prandially, 6 – 7 post-prandially
- Half treatment arm, half controls. 80% of each group is t1d.
- 15% discontinued use after first week.
- 64% used the device per protocol
- Reasons for stopping use
- Numerous disturbing alarms
- Skin irritation
- Sensor inaccuracy
- Hba1c exactly the same between two groups
- Glycemic control
- Average plasma glucose: 6.7 (treatment) versus 6.8 (control) mmol/L
- 16% have hypoglycemic events in both groups
- Endpoints
- Intervention group had higher macrosomia rate. 45% versus 34% for controls.
- Preterm delivery, too, was worse– 29% in treatment, 22% in control
- Subgroup analysis (ie, per-protocol users versus not) show same trend– treatment group is worse!
- Bright sides? 11% versus 19% severe hypoglycemia rates
- What about during delivery?
- Wearing the device before and during delivery
- Main outcome: avoiding fetal hypoglycemia. If glucose has been too high in the womb, the fetus can’t down-regulate its insulin upon birth, and goes low. Sever cases require physician administration of glucose to the newborn
- Results?
- Eight hour plasma glucose is higher in treatment arm– 5.6 versus 5.4
- Slight reduction in neonatal hypoglycemia (34 versus 46% ?)
- Reduction in severe hypoglycemia (11 versus 19%)
- Conclusion: these data do not support the use of CGM in unselected pregnant women, but highlights the need for better sensors.
- Q: What these studies show is the remarkably good care you’re giving to your patients in the control group! Maybe the problem is data overload? And the second problem is data interpretation– that neither the doctor nor the patient knows what to do with the data? I thought I knew what I was doing. I have been working with CGMs for several years.
- Q: The sensor accuracy was a lot worse than we generally see in non-pregnant populations. Do you have any thoughts about why that might be? A: Actually, the data I gave at the beginning was from a non-pregnant population. The reported errors are in the range of 20+ mg/dL. There is a lot of range.
- Q: What about triglycerides? Maybe glucose is not the only parameter. A: You are right, we have to look at other things than glucose. We see a correlation between between the glucose and the size of the baby, and between the glucose the last 8 hours before pregnancy and neonatal hypoglycemia
- Q: From Dexcom. CGM is not a therapy. It is a tool. The fact that CGM didn’t work reflects on how the CGM was used. What did you instruct patients? Were they compliant? A: When we did this study, it was not approved to use CGM without plasma glucose measurements. So we have all the plasma glucose values, and as you see, even my control patients are excellentat their plasma glucose control. They were instructed to look at the value. Followup: So they were not instructed to look at the direction of change? No, they were not.
Closed-Loop in Type 1 Diabetes in Pregnancy
Helen Murphy
- As a comment on the last talk, it is too hard to assimilate all the data all the time in real live, so we need the modelers and the computational people to come to our aid in interpreting the data.
- In this talk, addressing the promise of the closed loop/artificial pancreas in pregnancy
- Some phase 1 trial data about initial tests of closed-loop (CL) in pregnancy
- Speculate as to whether or not closed-loop is feasible in clinical limitation in the immediate future
- Who really needs a closed-loop? Will we really be out of a job in a few months if it works well? The benefits and burdens.
- Review of the closed-loop: glucose-responsive insulin treatment. Feedback loop. In this study, it’s a hybrid system. Patient maintains some degree of control.
- Control algorithm modulates basal insulin delivery.
- Reactive– proportional derivatives. Not good for pregnancy for a number of reasons.
- Adaptive/predictive– model to predict movement and modulate insulin delivery every 15 minutes accordingly. –> Insulin delivery is being adjusted much more frequently than patients can do in real life.
- Notably, to achieve normoglycemia, we need lots of variability in insulin delivery all the time.
- Insulin 3 or 4 x daily gives an ugly CGM trace. Not enough, certainly not during pregnancy.
- Difference in CGM glucose profiles stratified according to macrosomia of infant shows that there mothers of macrosomia babies ran lower overnight, then ran high after breakfast and through the rest of the day.
- There is no doubt that women with macrosomic babies have a different glucose profile than women without macromesic babies.
- Pilot study–
- Can CL be used safely to adjust doses overnigh in pregnancy?
- Can CGM accuracy be trusted?
- Is CL treatment feasible?
- CL took control for fast overnight, and then for breakfast until noon. With 65g of carbs, including juice
- Sample patient: insulin goes up and down according to CGM curve. Keeps glucose close to range. Seriously. Like magic. Oh how I want one!
- Periods of 3 – 4 hours overnight where the woman is given no insulin, and she doesn’t need it. Then kicks up to prevent dawn phenomenon.
- CGM slightly overestimates dawn phenom, but CL still able to outperform people, hedging its bets and keeping patient in range
- Not perfect but no episodes of hypoglycemia, and relatively little hyperglycemia. Surprising especially given the high carb meal.
- Median difference between CGM and blood glucose is ~10%. Certainly good enough for closed loop insulin delivery.
- Range is decent overnight. But after high carb breakfast, almost impossible to keep glucose in range.
- Conclusions
- Sensor was certainly good enough to continue using for the closed-loop
- More work is needed to improved post-prandial control
- More work needed for non-sedentary conditions
- The physiology of meal digestion. Glucose in early and late pregnancy having an effect on glucose up to six hours after dinner.
- Breakfast is faster. Half of glucose reflected in BG within 1hr (rather than 2hrs at night). No insulin fast enough right now to catch that.
- Post-prandial glucose disposal at night is particularly sluggish in late pregnancy. Glucose appears quickly, but is disposed of slowly.
- What complicates matters is gestational changes in the absorption rates of insulin. Using Aspart in early pregnancy, time to max effect is 50+- 10 min. But 80+-30min in late pregnancy –> Slower and more variable.
- Fully automated closed-loop is not feasible at this time, because we don’t have insulin fast enough to administer once BG is already going up. We will need to administer it early.
- Also, increase rates of glucose disposal in late pregnancy, ie with exercise
- Second study: CGM-augmented pump therapy versus CL with high carb meals followed by vigorous exercise after meals for 20 minutes.
- Does the exercise help prevent post-prandial spikes?
- Can this supplementation allow us to bridge the gap to closing the loop at night?
- Algorithm + exercise achieved 95% time in target overnight. Well controlled women on pump arm had 100%, but had higher hypoglycemic index
- Compared sensor accuracy during rest and during strenuous walking. Sensor accuracy is much better at rest. Walking accuracy was much lower. Yet algorithm was able to adjust and still achieve 80% time in target during the exercise.
- Who needs a closed loop?
- Sample patient who is amazing. One time above target in a week. She doesn’t need a closed loop; she will outperform it.
- But what about for the rest of us?
- We looked at this with 336 people using CL at home, unsupervised. Glucose control is far superior to sensor-augmented pump treatment.
- Patient response is overwhelmingly positive– I haven’t slept so well in years. It has transformed my life.
- Drawbacks– size, the implants, the alarms are annoying, the battery life isn’t very good. Adolescents gripe :) Still big, bulky, prototypes.
- What do we do in pregnancy? Do we accept the limits of the technology and the algorithms and get to work improving them? The view we’re taking in our center is, Let’s start the dialogue. This is the beginning. The CL does better than most patients can do at night on their own currently.
