CamAPS FX FAQ

The algorithm

What does CamAPS FX use to run?

CamAPS FX runs on body weight and total daily insulin dose. It does not use preset basal rates, insulin sensitivity factor, insulin-to-carb ratio, or active insulin time as user-defined settings.

How does CamAPS adapt over time?

Adaptation happens at three levels: daily (long-term insulin need), hourly (circadian rhythm), and post-meal (learned over approximately five days of use).

How is CamAPS different from other systems such as 780G and Control-IQ?

CamAPS relies on basal-heavy corrections, scaling its internal basal delivery up to 500–700% of its calculated base rate. Systems such as 780G and Control-IQ cap basal at around 250% and rely instead on auto-boluses to correct glucose.

What about Active Insulin Time?

Active Insulin Time is not a user setting in CamAPS FX. The system estimates it dynamically within each loop cycle.

Getting started — the first two weeks

What is the guidance for the first 14 days?

During the first 14 days, the general principle is to avoid using Boost and Ease-off unless safety requires it. This allows the algorithm time to learn individual patterns without interference from manual overrides.

What Personal Glucose Target is worth exploring at the start?

The most commonly used starting range is 5.0–6.0 mmol/L (100–120 mg/dL). On average, a lower target tends to produce higher time in range but also increases the likelihood of hypoglycaemia. Night-time is generally the safest window to explore tighter targets. For young children, higher daytime targets are often used because of variable activity levels — this is worth discussing with the care team.

Advanced features

When is Boost typically useful?

Boost is generally used during illness or known glucose-raising events. Using it frequently tends to blunt the algorithm’s learning over time, so most people find it works best used sparingly.

When is Ease-off typically useful?

Ease-off is commonly used around sporadic exercise, illness, or stress. It can be scheduled in advance, which many parents find helpful for planning around children’s activity. For exercise, start Ease Off 90 minutes before the session begins — not at the start of the session. Starting earlier gives the algorithm time to reduce delivery before movement begins.

What does Add Meal do?

Add Meal functions like an extended bolus and is often helpful for high-fat or high-protein meals. The system only delivers the additional insulin if it determines it is needed based on glucose trends.

Exercise

How do overnight hypos after evening training tend to be managed?

A common approach is to raise the overnight Personal Glucose Target. Snacking at bedtime tends to be more useful when glucose is trending down with insulin on board, rather than as a routine strategy. If a pattern repeats, it is worth reviewing the approach. For more on this, see the AID and exercise guide.

Meals and bolus strategies

Is exact carb counting necessary?

Not always. Trial evidence suggests that simplified meal announcement is broadly non-inferior to exact carb counting when using CamAPS FX — though individual responses vary.

How do high-fat meals such as pizza tend to be handled?

A commonly used approach is entering approximately 60% of carbs as a normal entry and 40% as a slowly absorbed meal. Add Meal can be useful if a late rise appears. Boost is generally held in reserve unless necessary. Individual response varies considerably — CGM data is the most useful feedback tool here.

Do carb-estimation apps such as Snaq help?

A randomised controlled trial (n=44, three weeks) found that Snaq was associated with +6.6% time in range and a mean glucose reduction of 0.54 mmol/L, with lower estimation error compared with patient estimates or CalorieMama. Consistency of estimation was not significantly better. Real-world uptake was low at approximately 1.6 uses per day, with 19% of advice followed.

Sensor considerations

Dexcom or Libre — what are the differences?

Both are factory-calibrated. Dexcom G6/G7 allows optional calibration if drift occurs and has a 10-day wear period but a larger footprint. Libre 3 does not offer a calibration option and has a 15-day wear period with a smaller sensor.

Is inserting sensors early worth exploring?

If you experience significant sensor inaccuracy in the first hours of wear, inserting the sensor 24–36 hours before activation is worth exploring. The idea is that this allows initial inflammation to settle before the sensor goes live.

Connectivity and troubleshooting

Why does Auto mode sometimes turn off?

A common reason is families toggling the system off under the impression that insulin delivery stops when glucose is low. Technical causes include the phone being more than 6 metres from the pump, Bluetooth cache issues, and infusion set problems.

What troubleshooting routine tends to work well?

Restarting the phone at set changes, clearing the Bluetooth cache, and keeping the phone within 6 metres of the pump resolve most recurring connectivity issues.

GNL resources

• Tips and tricks for CamAPS FX
• Podcast episode: Exercise and CamAPS FX
• CGM series
• AID systems guide
• Bolus strategies for challenging meals

References

• Alwan H et al. Real-World Evidence Analysis of a Hybrid Closed-Loop System. J Diabetes Sci Technol. 2025;19(2):385–389. PMID: 37421250
• Baumgartner M et al. Carbohydrate Estimation Accuracy of Two Commercially Available Smartphone Applications. J Diabetes Sci Technol. 2024. PMID: 39058316
• Moser O et al. Exercise and CGM in type 1 diabetes (position statement). Diabetologia. 2020. PMID: 33047169
• Moser O et al. The use of automated insulin delivery around physical activity in T1D. Diabetologia. 2025;68(2):255–280. PMID: 39653802
• Tecce N et al. AI-Powered Carbohydrate Counting for Type 1 Diabetes. Diabetes Care. 2025;48(8):e97–e98. PMID: 40397829

Questions still open

• Algorithm transparency: CamAPS ignores presets, but clinicians lack clear published guidance on its internal caps compared with 780G and Control-IQ.
• Boost and Ease-off overuse: how often does frequent use meaningfully blunt algorithm learning in real-world settings?
• Meal handling: should AID systems auto-detect fat and protein effect rather than relying on a manual slowly-absorbed-meal entry?
• App integration: Snaq showed +6.6% time in range, but uptake was low. Whether embedding AI tools inside AID workflows improves adherence remains an open question.