AID and exercise

Major in the majors

Three variables decide almost everything

The EASD/ISPAD 2025 consensus – the anchor

The EASD/ISPAD 2025 position statement (Moser, Zaharieva, Pemberton et al) is the Grade A reference for AID and exercise. It covers announced versus unannounced activity, temporary target increases, and per-system exercise modes for MiniMed 780G, Control-IQ, Omnipod 5, and CamAPS FX. Both lead authors are GNL Scientific Advisers, and the statement is the direct evidence base for the GNL Exercise Planning explorer.

The headline pattern from the consensus: announce planned exercise to the algorithm where possible, raise the target ahead of the session rather than during it, and accept that the post-exercise window is the harder problem rather than the during-exercise window.

What AID can and cannot do during exercise

What it can do

• Reduce or suspend basal insulin in response to falling glucose.
• Raise the algorithm target through an exercise mode or temporary target setting.
• Deliver micro-corrections automatically when glucose rises after exercise.

What it cannot do

• Recall bolus insulin already in subcutaneous tissue.
• Anticipate exercise the user has not announced.
• See the difference between aerobic and anaerobic effort – it sees only the resulting glucose.
• Override the physiology of the 90-minute window after a meal bolus.

The asymmetry is what makes pre-exercise planning still matter on AID. Reactive algorithm work is good. Proactive user input on top of it is better.

Announce strategies and the 30-minute activity feature

Zimmer (2023) reviewed announcement strategies across AID systems. Setting a temporary higher target or enabling exercise mode in advance of planned activity (typically a window of around an hour to ninety minutes ahead) reduces during-exercise hypoglycaemia. The exact ahead-of-time window varies by system and is one of the live calibration questions in the literature.

Turner et al (2024) examined Omnipod 5’s 30-minute activity feature specifically. The broader pattern across AID systems is that announce-ahead works better than reactive correction once exercise is under way.

The manual mode trade-off

Zaharieva (2023) and clinical experience converge on a real trade-off: many people find manual mode more predictable during specific exercise sessions because it removes the continuous background adjustments the algorithm makes. The cost is the loss of automated suspension if glucose drops unexpectedly. Many people use AID as the default and switch to manual or exercise mode only for specific session types where the algorithm’s responsiveness becomes a liability rather than an asset.

This is not a recommendation to switch to manual mode for every exercise session. It is acknowledgement that the algorithm-versus-manual choice is a real one, and that there is no universally correct answer.

Pre-emptive carbs and bolus reductions on AID

Zimmer (2023) and the EASD/ISPAD 2025 consensus both address the pre-exercise meal bolus question. For exercise planned within the 90-minute window after a meal, a meal bolus reduction in combination with the AID temporary target tends to balance hypoglycaemia prevention and post-exercise hyperglycaemia better than relying on the algorithm alone. The exact percentage reduction is the individualisation problem Part 2 covers in detail.

Tagougui (2020) provides the post-meal reductions evidence on AID specifically. The pattern across the literature is that smaller, earlier interventions outperform larger, later corrections.

Real-world AID and exercise – what the registries show

• Joubert (2025) – real-world AID use indicates more variability around activity than pivotal trials reported. Expectations need calibrating accordingly.
• Wang (2024) – paediatric AID and physical activity. Children using AID still see meaningful exercise glucose variation; the algorithm helps but does not flatten the response.
• Seckold (2025) – three exercise types in adolescents on AID. Aerobic, anaerobic and mixed each produce different algorithm response patterns. The user still has to anticipate which type the session will be.
• Moser and Mader (2025) – T1D camp data comparing MiniMed 780G and CamAPS FX (mylife) across multi-day sustained activity. One of the few direct between-system comparisons.

The post-exercise window is the harder problem

Two studies frame this clearly.

• Myette-Côté (2022) – the one-to-two hour window after exercise remains a high-risk period for hypoglycaemia on AID, even when the during-exercise window is well-managed by the algorithm.
• Morrison (2022) – late afternoon vigorous exercise on AID raises post-meal hypoglycaemia risk that evening, but does not necessarily raise overnight hypoglycaemia. The risk has shifted from overnight to post-exercise evening, rather than disappearing.

And looking further forward – Bergford (2023, T1DeXi) – the next-day risk after long sessions is real and underappreciated. Long sessions leave elevated hypo risk into the following day, even after the session itself ends.