Pediatric Type 1 diabetes insulin pump — CGM-guided overnight basal profile programming and adjustment
Before any fine-tuning can happen, a child starting on insulin pump therapy needs an initial hourly basal rate profile — the slow, continuous trickle of insulin that covers background metabolic needs between meals and overnight. This starting profile is deliberately conservative: a reasonable weight-based estimate that a clinician expects to refine over the following days to weeks using real CGM data, not a final answer.
The starting total daily dose (TDD) for a newly pumped child is usually estimated from prior injection totals or from a weight-based formula (roughly 0.4–0.6 U/kg/day in established pediatric patients, lower during any "honeymoon" residual insulin production phase).
Once TDD is estimated, basal insulin is allocated as 40–50% of that total — lower than the ~50% typical in adult pumping, because children often need proportionally more bolus coverage relative to their smaller, more variable meals. A 22 kg child with an estimated TDD of 21 U/day might start with roughly 9 U of basal insulin spread across 24 hours, an average rate of about 0.375 U/hr.
Rather than a single flat number, most pump templates start with 1–4 broad segments (for example, midnight–6am, 6am–noon, noon–6pm, 6pm–midnight) using clinical judgment about typical circadian variation, refined later into finer hourly segments once real overnight data is available.
A brand-new basal profile is not yet informed by this specific child's actual overnight glucose behavior — it is a population-informed starting estimate. Programming a highly granular, aggressively tuned profile on day one risks locking in guesses that do not match the child's real physiology, and increases the chance of undetected overnight hypoglycemia before the family and care team have CGM trend data to check against.
Instead, the standard workflow is deliberately staged: start flat and conservative, observe fasting/overnight CGM trends without bolus interference, then titrate segment by segment. This mirrors safe pump initiation protocols used across pediatric diabetes centers.
A basal rate that is too aggressive on day one is far riskier in a young child than one that is mildly too low — hypoglycemia unawareness and smaller glycogen reserves make conservative starting doses the safer default, with upward titration guided by data rather than guesswork.
Basal insulin can only be evaluated cleanly when nothing else is influencing glucose — no meals, no correction boluses, no exercise. Overnight fasting periods are the single best window for this: the child is asleep, has typically not eaten for several hours, and any rise or fall in CGM glucose can be attributed almost entirely to a mismatch between the programmed basal rate and true background insulin need.
Each hourly basal segment is judged by what happens to CGM glucose during the hours it governs. If glucose climbs steadily through a segment, the basal rate active during (or just before) that rise is probably too low. If glucose drifts downward, the corresponding segment is probably too high.
Because insulin delivered by a pump has a pharmacodynamic lag — rapid-acting analogs peak roughly 60–90 minutes after infusion and continue acting for 3–4 hours — the segment responsible for a glucose change is usually the one active 1–2 hours before the trend becomes visible on CGM, not the segment during which the trend is observed. This lag is one of the most common sources of titration error when read carelessly.
Multiple nights (typically 3–5) are reviewed together, since a single night can be skewed by residual bolus insulin, an unusually active bedtime, stress, or illness. Consistent patterns across nights are what justify an actual basal change.
Before attributing an overnight trend to the basal rate, other explanations have to be ruled out:
• Residual bolus insulin: a dinner bolus is still acting into the early overnight hours and can mask or mimic a basal problem • Bedtime snack carbohydrates: unaccounted or fast-digesting carbs can cause an early-night rise unrelated to basal • Rebound from an earlier low: overcorrected hypoglycemia can produce a compensatory rise later in the night • Illness, growth spurts, or hormonal surges: temporarily increase insulin need independent of the programmed basal • Sensor lag or compression lows: CGM artifacts (e.g. sleeping on the sensor) can fabricate a false low trend
Only once these are excluded does a persistent, reproducible overnight trend become a genuine candidate for basal segment adjustment in the next stage.
Titration is where the basal profile is actually changed — carefully, one segment at a time, using the trend data gathered overnight. The guiding rule is modest, incremental change: typical adjustments are 10–20% of the current segment rate, applied to the segment active before the observed trend, followed by a re-check period before making the next change.
For a segment where overnight CGM shows a consistent rise: increase the basal rate active roughly 1–2 hours before the rise begins, typically by 10–20%. For a segment showing a consistent fall: decrease the basal rate active before the fall begins by the same margin.
Small, deliberate steps matter for two reasons. First, the pharmacodynamic lag of rapid-acting insulin means the full effect of any change will not be visible until the next full basal-acting cycle — overshooting before that feedback arrives compounds error rather than correcting it. Second, in pediatric patients smaller total doses mean the same percentage change represents a smaller absolute number of units, but the child's smaller size also means the same absolute change has proportionally larger glycemic impact — 20% steps balance responsiveness against overcorrection risk in both directions.
A common error is changing several adjacent segments simultaneously after seeing one bad night. Because each segment's insulin effect extends into the following 3–4 hours, adjacent segments overlap in their glycemic influence — increasing segment 2am and segment 4am together, for example, can produce a compounded overnight low that neither adjustment alone would have caused.
The safer approach changes one segment (or occasionally two well-separated segments) per titration pass, then waits 2–3 nights to observe the isolated effect before touching a neighboring segment. This keeps cause and effect traceable, which matters enormously when the "sensor" is a sleeping child and the feedback loop is inherently slow.
Overlapping-segment overcorrection is one of the most frequently cited causes of iatrogenic nocturnal hypoglycemia during pump titration in pediatric patients — the fix is procedural discipline (one segment, one change, one waiting period) rather than a more complex algorithm.
Once the base overnight profile is reasonably stable, two predictable, physiologically distinct patterns still need dedicated basal segments: the early-morning dawn phenomenon driven by counter-regulatory hormones, and the day-to-day difference in insulin sensitivity between sedentary school days and more active weekends.
Between roughly 4am and 8am, a nocturnal surge in growth hormone, cortisol, and catecholamines increases hepatic glucose output and reduces peripheral insulin sensitivity — a normal physiological process that is exaggerated in children and adolescents during growth spurts and puberty. Without a dedicated adjustment, this produces a predictable glucose rise in the last hours of sleep that a flat overnight basal rate cannot cover without causing earlier-night lows.
The standard fix is a distinct, elevated basal segment covering roughly 4am–8am, increased 20–40% above the pre-dawn overnight rate. This is functionally similar to a scheduled temporary basal increase, but programmed as a permanent segment in the base profile because the pattern recurs nightly rather than being an occasional event.
Physical activity dramatically increases insulin sensitivity, sometimes for 12–24 hours after exercise ends. A school day with mostly sedentary classroom time has different background insulin needs than a weekend or holiday with unstructured, higher physical activity — the same fixed basal rate applied to both days will run too high on active days (risking exercise-related lows) or too low on sedentary days.
Many pediatric patients therefore run two parallel basal profiles — one for school/sedentary days and one for weekend/active days — switched manually or by scheduled pump profile, with the weekend profile typically reduced 10–25% relative to the weekday profile, most noticeably in the afternoon and evening segments following typical activity windows.
| Product | Indication | Trial Design | Key Result |
|---|---|---|---|
| 4am–8am (dawn) | Both profiles | Elevated segment for growth hormone / cortisol surge | Same on both days — physiology does not follow the calendar |
| 8am–3pm (school hours) | Weekday profile | Standard rate; mostly seated classroom activity | No activity-related reduction needed |
| 8am–3pm (weekend) | Weekend profile | Reduced 10–20%; unstructured play, sports, outdoor activity | Lowers hypoglycemia risk during active hours |
| 3pm–9pm (after school / evening) | Both, weighted differently | Weekday: after-school activity partial reduction; weekend: full-day activity carries over | Captures lingering post-exercise sensitivity |
The final stage confirms that the titrated basal profile actually holds glucose steady overnight across the full 70–180 mg/dL target range on a clean validation night, then transitions that validated profile into a hybrid closed-loop system — not as a replacement for the profile, but as the baseline the algorithm continuously modulates around.
A validation night is a final fasting/overnight CGM check performed after all segment titrations are complete, with no bolus or meal interference. Success criteria are stricter than during the earlier trend-collection stage: glucose should remain within the 70–180 mg/dL target range for essentially the entire night, with no segment showing a sustained rise or fall pattern.
If validation fails in one segment, that segment returns to the titration stage for another small adjustment and another re-check — validation is a gate, not a one-time event, and it is repeated after any adjustment until the full night holds steady.
Hybrid closed-loop (automated insulin delivery) systems use a control algorithm to continuously raise or lower insulin delivery in small increments based on real-time CGM readings, compared against the pump's programmed basal rate as a reference point. This means a poorly titrated underlying basal profile does not disappear once closed-loop is activated — the algorithm simply has to work harder, modulating further from a wrong baseline, which shows up as more frequent and larger corrections, more time spent outside target, and less predictable behavior when the algorithm temporarily reverts to manual basal delivery (for example, during sensor warm-up or system errors).
A well-validated basal profile gives the closed-loop algorithm a stable starting point to modulate from in small increments, rather than a distorted baseline it must constantly fight against — meaningfully improving both time-in-range and the predictability of the system's behavior during any fallback to manual delivery.
Closed-loop automation is a modulation layer on top of a validated basal profile, not a substitute for one — the titration work in Stages 1–5 remains the foundation the algorithm depends on, and is revisited whenever growth, puberty, or activity patterns shift the child's underlying insulin needs.