Guide series, Part 2 of 3
The Menstrual Cycle and Type 1 Diabetes, Part 2: AID Systems Across the Cycle
Hybrid closed-loop and advanced hybrid closed-loop systems shift where the work goes; they do not remove it. Part 2 walks through what each named UK system has been measured to do across the cycle, and the AID Algorithm Optimiser ladder that helps you and your team think about settings.
Ask Grace
Want to ask how the AID Optimiser ladder applies to your settings on your specific pump? Ask Grace.
“The algorithm caught most of it”
The arc that comes up in adult AID conversations is the same shape, over and over. She switched to AID eighteen months ago and the change was real. The 3am corrections that used to run her week stopped. The post-meal spikes blunted. The mental load eased. Then the seven days before her period arrived, and the algorithm caught most of it. Not all of it. The late-luteal afternoons still drift to 12 mmol/L (216 mg/dL) on the same lunch the algorithm rides clean for the rest of the month.
The honest answer to “does the system adjust for cycle phase” is: not directly, not yet, in any of the four systems available in the UK. None of the algorithms takes a cycle-phase input. They respond to the glucose they see, and they push more insulin when more is needed. Most of the time, that is enough. In the late luteal week, for some women, it is not.
This is not a system failure. The trials say so. Levy and colleagues (2022, Diabetes Technology and Therapeutics, secondary analysis of the iDCL pivotal Tandem Control-IQ trial) found no significant cycle-phase difference in time in range or insulin delivery in their sixteen-woman cohort. Monroy 2025 (MiniMed 780G prospective) found the system raised insulin delivery in the late luteal phase and preserved time in range, but mean glucose still ran higher in the late luteal phase than in the early follicular phase. Two systems, two consistent findings: AID adapts, the work shifts, and a residual late-luteal signal can still arrive.
If you are the woman on AID reading this: the algorithm doing more for you across the cycle does not mean you should be doing less of the noticing. The data the algorithm cannot ask you for, the cycle phase, is the data you bring to the conversation with your team about whether your settings are right.
What each named UK system has been measured to do
Four AID systems are licensed and in use in the UK adult T1D population: CamAPS FX, Tandem Control-IQ, MiniMed 780G, and Omnipod 5. The cycle-specific evidence is uneven across them. None of the four algorithms takes cycle phase as an input. The differences are in how the algorithm responds to the glucose drift the cycle produces. Tap a system for the cycle-specific read.
CamAPS FX
The Cambridge algorithm, an Android-based hybrid closed-loop running on a phone, paired with a Dexcom CGM and a Dana or Ypsomed pump. CamAPS FX has the strongest evidence base in pregnancy in T1D (AiDAPT, Lee 2023, NEJM). For non-pregnant cycle physiology specifically, CamAPS FX has not yet been the subject of a dedicated published cycle-phase secondary analysis. The algorithm modulates basal strongly and runs a personalised glucose target; the practical observation from CamAPS users is that the late-luteal drift is often partially absorbed but not erased.
Tandem Control-IQ
Algorithm in the t:slim X2 pump, Dexcom CGM. Levy 2022 is the cycle-specific dataset, drawn from the iDCL pivotal trial: in sixteen menstruating women across three or more cycles each, no significant cycle-phase difference was detected in twenty-four-hour mean CGM glucose, time in range, time below 4 mmol/L (72 mg/dL), or weight-based insulin delivery. In this well-controlled cohort, Control-IQ compensated for the cycle effect to the point where the population-average signal disappeared. Two caveats: the cohort entered with relatively well-controlled T1D at baseline, and the sample is small enough that a real-but-modest cycle effect could be hidden inside it.
MiniMed 780G
The Medtronic advanced hybrid closed-loop, paired with a Guardian or Simplera CGM. Monroy 2025 (12 women, 36 cycles): mean glucose was higher in the late luteal phase than in the early follicular phase (7.7 versus 7.3 mmol/L; 139.5 versus 131.5 mg/dL), total daily insulin was higher in the late luteal phase (37.2 versus 33.6 IU), and time in range was preserved at around 83 to 85 per cent in both phases. Mesa 2024 (thirteen women with T1D and recurrent hypoglycaemia, before and after switch from sensor-augmented pump to AHCL) found that on AHCL the cycle-related midfollicular hypoglycaemia signal was reduced and the difference between phases disappeared, while a small residual time-in-range gap remained between the early follicular and late luteal phases. The 780G family adapts; the late-luteal week still arrives.
Omnipod 5
The Insulet algorithm in a tubeless pump, paired with a Dexcom CGM. The cycle-specific evidence base for Omnipod 5 in adult T1D women is the thinnest of the four; the system is newer to the UK adult market and a dedicated cycle-phase secondary analysis has not yet been published. The principles from Levy 2022, Monroy 2025, and Mesa 2024 (algorithm reacts to the glucose it sees, no direct cycle-phase input, residual late-luteal effect possible) are likely to apply, and a published cycle dataset for Omnipod 5 is on the open-evidence queue.
No algorithm takes cycle phase as an input. All four respond to the glucose they see. Where the cohort was well-controlled at baseline, the algorithm has more headroom and the population-average cycle signal can disappear. Where the late-luteal drift is large, a residual signal can persist even with AID. This is the architecture of the evidence rather than a limitation of one system.
The AID Algorithm Optimiser ladder, and how to think about settings
The GNL AID Algorithm Optimiser is a five-level ladder for thinking about algorithm strength across CamAPS FX, MiniMed 780G, Tandem Control-IQ, Tandem Mobi (CIQ+), and Insulet Omnipod 5: how short the active insulin time is, how low the glucose target is, how strongly the algorithm modulates basal.
The ladder, and the key drivers and settings adjusted across each level, has been reviewed and refined with input from CamAPS, MiniMed, Tandem and Insulet global medical leads. Their input has shaped which levers are exposed at each level and how they are described. However, the levels themselves have not been validated directly against any manufacturer’s internal simulator or proprietary dataset, so this is not a manufacturer endorsement of the GNL ladder. The Optimiser remains a Grade D educational suggestion layer built on a Grade A and B evidence base (peer-reviewed RCTs, registry data, pharmacokinetic studies, and the GNL real-world dataset analysis).
The Optimiser is educational. It carries a deliberate, declared subjective bias toward the importance of insulin-on-board (IOB) visibility and the IOB-vs-algorithm-strength trade-off, because that trade-off is the dominant safety driver in real-world use and is under-represented in stock manufacturer guidance. Any deviation from manufacturer-recommended starting settings is a clinical decision made with your diabetes care team.
Why it matters in the cycle conversation: where the data shows a real luteal-week drift on your current settings, the team conversation can usefully include “is there a level of algorithm strength that would suit my late-luteal week better, with safety guard rails?”. This is the team’s call on your data; the Optimiser is the shared vocabulary. The figures shown are population-average estimates, not personalised; your own correction factor and insulin-to-carb ratio, set with your team, are what apply to you.
When to bring the cycle conversation to your team
The teams who handle this conversation well are doing so in increasing numbers, and the literature has finally caught up enough to support them. The conversation lands more cleanly when you bring three things.
First, two or three cycles of overlay data: your CGM record with the cycle dates marked, ideally as a printed AGP or a screenshot from your AID app. The team can read the pattern with you in two minutes. Second, the named question. “I notice my time in range drops by ten to fifteen percentage points in the seven days before my period; can we look at whether my settings could absorb more of that?” lands more cleanly than “is this normal?”. Third, the named system. The settings that move are system-specific: on CamAPS FX, the personalised glucose target and the algorithm-strength setting; on MiniMed 780G, the active insulin time and the glucose target; on Control-IQ, the personal profile and exercise mode use; on Omnipod 5, the basal programme and adaptive basal. Any change is the team’s call on your data.
If the first answer is “we do not really adjust for that”, ask in writing. Send the question by clinic letter or secure message, with the data attached. The literature is now strong enough that the cycle pattern is no longer a fringe observation, and a recorded ask gives the team a reason to come back to it. Ask again at the next appointment.
The GNL position
AID is a real and meaningful improvement in T1D care, including for women navigating the cycle effect. The evidence base is consistent on this. AID does not erase the cycle, and it should not be expected to. Where it shifts the work is in the noticing, the timing, and the negotiation with your team about settings, not in the absence of work.
The system you are on is rarely the bottleneck. The bottleneck is more often that the cycle conversation is new to many adult diabetes teams, the published evidence has only matured in the last three to five years, and the workflow for reading cycle-overlaid CGM data has not yet landed in routine clinic. Your data, brought specifically and asked specifically, is what shifts that.
Evidence backbone, Part 2
Voice-first delivery; research-grade source layer. Tap any source for the citation detail.
Levy 2022, Tandem Control-IQ across the cycle
Levy CJ et al. Insulin Delivery and Glucose Variability Throughout the Menstrual Cycle on Closed Loop Control. Diabetes Technol Ther. doi:10.1089/dia.2021.0431. iDCL secondary analysis, sixteen menstruating women.
Monroy 2025, MiniMed 780G across the cycle (780MENS)
Monroy G et al. 780MENS Prospective Study, MiniMed 780G across the menstrual cycle. Diabetes Technol Ther. Twelve women, thirty-six consecutive cycles.
Mesa 2024, AHCL across the cycle in women prone to hypoglycaemia
Mesa A et al. AHCL Across the Menstrual Cycle in Women With T1D Prone to Hypoglycaemia. Diabetes Technol Ther. Thirteen women, before and after switch from sensor-augmented pump to AHCL.
Lee 2023, AiDAPT (CamAPS FX in pregnancy)
Lee TTM et al. AiDAPT, Automated Insulin Delivery in Pregnancy Complicated by T1D. NEJM 389(17):1566-1578. doi:10.1056/NEJMoa2303911. Multicentre RCT of CamAPS FX, nine UK NHS sites, one hundred and twenty-four pregnant women.
Brown 2015, cycle-physiology baseline
Brown SA et al. J Diabetes Sci Technol 9(6):1192-1199. The pre-AID cycle-physiology baseline against which the AID-era data is compared.
Tatulashvili 2022, the population reference
Tatulashvili S et al. J Clin Endocrinol Metab 107(10):2793-2800. doi:10.1210/clinem/dgac443. Largest contemporary cycle-phase ambulatory CGM cohort; the population reference for what AID has to absorb.
AID Algorithm Optimiser positioning policy
Reviewed by manufacturer global medical leads, not endorsed by any of them. Grade D educational synthesis on a Grade A and B evidence base. Full positioning at gnl-grace/wiki/policies/aid-optimiser-positioning.md. The Optimiser audience is open to all registered GNL users; the disclaimer architecture (population-average framing, “not a medical device” attestation, correction-dose framing, care-team referral) is the safety gate.
Part 2 of 3
AID Systems Across the Cycle