Fifteen Years of Model-Based Glycaemic Control in Preterm Neonates: Safety, Performance, and Evolution of Insulin Sensitivity and Cohort Characteristics

Imagine a tiny, fragile world inside a hospital's Neonatal Intensive Care Unit (NICU), where the newest arrivals are so small they fit in the palm of a hand. These pre-term babies are like high-performance race cars that haven't had their engines fully tuned yet; their bodies are still learning how to handle fuel. One of the trickiest fuels for them is sugar (glucose). Too much sugar, and their developing brains and organs can get damaged; too little, and they can crash into a dangerous low. For decades, doctors have tried to keep this sugar level in a "Goldilocks zone"—not too hot, not too cold, just right. But because these babies are so small and their bodies change so fast, guessing the right amount of sugar medicine (insulin) is like trying to hit a moving target while riding a unicycle. If you guess wrong, you risk a sugar crash.

To solve this, scientists started using "digital twins." Think of these as video game avatars of the real baby, built from math and biology. Every time a nurse checks the baby's blood sugar, the computer updates the avatar, predicts how the baby's body will react to different doses of insulin over the next few hours, and suggests the safest, most effective dose. It's like having a super-smart co-pilot that knows the baby's engine better than anyone else. This study looks at a 15-year journey of using this digital co-pilot in a specific hospital in New Zealand, asking a simple but crucial question: Did this smart system work well over time, and did the babies themselves change in a way that made the job harder?

---

The 15-Year Flight Log: A Smart Pilot vs. A Changing Crew

This paper is a long-term report card on a computerized system called STAR-GRYPHON (a fancy name for a system designed to prevent both high and low blood sugar in newborns). The researchers looked back at 15 years of data from the Christchurch Women's Hospital NICU, comparing three different eras of care:

1. The "Old School" Era (2005–2008): Doctors used a paper chart and a sliding scale to guess insulin doses.
2. The "First Gen" Computer Era (2009–2012): They switched to the first version of the digital twin system (STAR-1).
3. The "Pro" Era (2013–2023): They used the upgraded, smarter digital twin (STAR-GRYPHON).

The goal was to see if the computer kept the babies' blood sugar in the safe zone (between 4.0 and 8.0 mmol/L) without causing dangerous lows, and to see if the system's performance changed as the years went by.

The Good News: The Computer is a Safe Co-Pilot

The study found that the computerized system was a huge improvement over the old paper charts.

• Safety First: The digital system was much safer. In the old days, about 2.1% of blood sugar readings were dangerously low (below 4.0 mmol/L). With the digital twin, that number dropped to a tiny 0.6% to 0.9%.
• Better Control: The system kept more babies in the "Goldilocks zone." In the early computer days, about 70% of the time was spent in the safe range. By the time they reached the 2013–2014 period, that jumped to 77.3%.
• Fewer Crashes: Severe low blood sugar (below 2.6 mmol/L) became very rare. In the old paper days, there was 1 severe crash for every 25 babies. With the new system, it was only 1 crash for every 33 babies.

The computer didn't just guess; it learned. It used a "digital twin" to simulate what would happen if they gave a little more or a little less insulin, choosing the option that kept the baby safe while avoiding high sugar.

The Twist: The Babies Changed, Making the Job Harder

Here is where the story gets interesting. Even though the computer system stayed exactly the same for the last 13 years, the results started to get slightly worse after 2015. The percentage of time babies spent in the safe sugar zone dipped, and there was a bit more high sugar (hyperglycemia).

Why? The paper suggests the problem wasn't the computer; it was the crew. The babies being treated in the later years were different.

• Smaller and Tougher: The babies in the later years (2021–2023) were significantly smaller. Their average birth weight dropped from about 932 grams in the 2013–2014 group to just 654 grams in the 2021–2023 group.
• The "Sticky" Engine: The most important finding was about Insulin Sensitivity (SI). Imagine insulin as a key that unlocks the door to let sugar into the cells. In the early years, the babies' bodies were very sensitive to the key; a small turn opened the door wide. But over time, the babies became "resistant." Their locks got rusty. The paper shows that insulin sensitivity dropped dramatically—by as much as 85% in some comparisons between the earliest and latest groups.

Because the babies became more resistant, the doctors had to give much higher doses of insulin to get the same effect. In fact, the average insulin dose went up from 0.03 U/kg/hr in the early days to 0.07 U/kg/hr in the later years.

The Saturation Wall

The paper points out a tricky limit. Even though the doctors kept turning up the insulin dose, the sugar levels didn't always come down as expected. It seems the babies hit a "saturation wall."

Think of it like trying to fill a bucket with a hose. If the bucket has a hole (resistance), you turn the hose up. But eventually, even if you turn the hose to maximum, the water can't get in any faster because the hole is too small or the bucket is full. The study suggests that in these tiny babies, the body might stop responding to insulin once the dose hits around 0.2 U/kg/hr. When this happens, giving more insulin doesn't lower the sugar; it just increases the risk of a crash later when the baby's body suddenly becomes sensitive again.

The Bottom Line

The paper concludes that the STAR-GRYPHON system is safe and effective, but it is facing a new challenge. The babies in the NICU are surviving at smaller and smaller sizes, and their bodies are becoming more resistant to insulin. This creates a difficult trade-off: to keep the sugar in the safe zone, doctors might need to give huge doses of insulin, which risks a crash, or they might have to restrict the sugar the babies eat (nutrition), which risks the baby not growing.

The study doesn't say the system is broken; it says the system is doing its best against a changing opponent. The "digital twin" is still the best tool they have, but the doctors now know that the babies they are treating are fundamentally different—and more difficult to manage—than the ones they treated 15 years ago. The future of care might require not just better computers, but a new conversation about how much sugar these tiny, super-resistant babies should be eating to stay safe.