The Menstrual Cycle and Type 1 Diabetes, Part 1: What the Body Is Doing

It is not in your head

The pattern that comes up over and over in DAFNE and BERTIE clinics is the same arc. Three weeks of the cycle, things mostly behave. Then the seven days before the period arrive and the same lunch you have been eating for years runs to 14 mmol/L (252 mg/dL) when last week it would have stayed under 9 mmol/L (162 mg/dL). The CGM line bends in a shape women now recognise before they have checked the calendar. The correction barely shifts it. The next correction barely shifts it. By morning the body has handed back what it took. Until next month.

Women who raise this with their diabetes team often get a polite “investigate” written in the notes and the conversation moves on. If this is the pattern in your data, you are not wrong. The work you have already done, the noticing, the trying to ride it out, the asking the team, is the right work. The reason it has not landed is not the question; it is that the question has lived for decades on the edge of the literature, in studies of six, twelve, fifteen women. Only in the last few years have the cohorts grown large enough for the pattern to be undeniable.

Tatulashvili and colleagues (2022, Journal of Clinical Endocrinology and Metabolism) measured ambulatory glucose by cycle phase across hundreds of person-cycles on flash CGM. Time in range fell in the luteal phase compared with the follicular phase, with both mean glucose and time-above-range higher in the second half. The same direction has held in every primary study large enough to ask the question, from the small clamp studies of the 1990s through to the AID-era secondary analyses (Gamarra and Trimboli 2023 systematic review).

It does not happen to every woman; somewhere between forty and sixty per cent of women with T1D notice the change month to month. For the women who do, the cycle is not noise to be filtered out. It is signal.

What is happening across the four phases

The cycle is conventionally split into four phases, anchored to ovulation. Day numbers are approximate, every body runs its own clock, and the routing here is the median pattern that has emerged from the larger CGM cohorts. The hormonal arc is the same in everyone with regular cycles; the glucose response to it is heterogeneous.

Menstrual phase, days 1 to 5

Oestrogen and progesterone are at their lowest. The womb lining is being shed. For many women with T1D, this phase is the most stable of the cycle: in one of the larger real-world datasets, time in range was higher in the menstrual phase than in any other (Syno menstrual cycle analysis 2026, GNL real-world dataset). The pattern most women describe is that everything they have been holding through the luteal week loosens in the first day or two of bleeding, and the glucose responds to ordinary bolus and ordinary basal in the way it used to.

Follicular phase, days 6 to 13

Oestrogen rises, progesterone stays low. The ovary is preparing the egg for ovulation. Follicular-phase glucose is variable across women: in some cohorts time in range remains good; in others it dips below the menstrual-phase peak. Insulin sensitivity is broadly intact, and the practical observation from women tracking their own data is that the same dose works the same way most of the week. Where it does not, the cause is more often life (sleep, stress, illness, a hard exercise day) than physiology.

Ovulatory phase, around day 14

Oestrogen peaks and then falls; luteinising hormone surges and triggers ovulation; progesterone begins its rise. This is a brief, busy hormonal moment, and it shows up in the glucose data as the highest variability and the highest hypoglycaemia risk of the cycle. In the same Syno dataset, ovulatory-phase coefficient of variation was the highest of any phase, and time below range was elevated. Many women notice an unexplained low in the day or two around ovulation; the data supports the experience.

Luteal phase, days 15 to 28

Progesterone rises through the early luteal phase, peaks in the mid-luteal phase, then falls in the late luteal phase as the cycle resets. This is where the strongest signal sits. Brown and colleagues (2015, Journal of Diabetes Science and Technology) showed that the high blood glucose index rose significantly across the cycle, peaking in the early luteal phase and returning to baseline by late luteal, and that nocturnal insulin sensitivity dropped from the early follicular to the early-mid-luteal and late-luteal phases. Tatulashvili 2022 confirmed the same direction in a larger flash-CGM cohort. The luteal phase is when most of the pattern lives.

Why some women see this clearly and others do not

The forty-to-sixty per cent figure that runs through the literature is the share of women with T1D who notice a measurable cycle-driven glucose change. The other forty to sixty per cent see no clear pattern. This is not a discipline gap or a tracking gap; it is a real biological heterogeneity, and the larger systematic reviews treat it as one of the central facts of the topic.

If you do not see the pattern in your own data, you are not missing it. If you do, you are not imagining it. The same applies cycle to cycle: the intra-patient consistency between cycles is moderate, not perfect, and a phase-driven pattern that is clear in one cycle can soften or shift the next month if life moves. The right inference from the population-average data is “is this a pattern for me, in my cycles, in my data?”, not “this will happen to you”.

The work you do tracking your cycle alongside your CGM is the work that resolves the question for your own body. The data is honest. The cycle, where it is signal, will show up.

What about my AID system, does it handle this?

The short answer is partly. The longer answer is the whole of Part 2. Briefly: in a Tandem Control-IQ secondary analysis (Levy 2022, Diabetes Technology and Therapeutics) no significant phase-driven change in time in range or insulin delivery was detected in sixteen menstruating women; in a MiniMed 780G prospective study (Monroy 2025) the system increased insulin delivery in the late luteal phase and time in range was preserved, although mean glucose still ran higher in the late luteal phase than in the early follicular phase. AID systems shift the work; they do not erase the cycle.

If you are on AID and the late-luteal week still surprises you, that experience is consistent with the data. Part 2 walks through what each named system can and cannot adjust for, the AID Algorithm Optimiser ladder reviewed by the manufacturer medical leads, and the conversation to take to your team about settings.

Tracking your own pattern

The fastest way to see whether the cycle is signal in your own data is to overlay the two records. Your CGM already keeps the glucose log. A cycle-tracking app, a paper diary, or the period-tracking field in your AID system records the cycle. Two or three cycles is enough to see whether a luteal-phase drift, an ovulatory dip, or a menstrual-phase loosening is showing up for you.

Look first for a higher mean glucose and lower time in range in the late luteal phase (the seven days before your period), versus the menstrual or early-follicular phase. Look for an ovulatory window of higher variability and a slightly higher chance of unexplained lows. Look for a loosening of the pattern in the first day or two of bleeding. Then look at whether the pattern is reproducible across two or three cycles, or whether it shifts. Both are normal.

What the data cannot tell you is what to change. That is the conversation with your care team, and the next section is the script.

The conversation with your diabetes care team

The reason this conversation has historically gone nowhere is not that the team is dismissive. It is that the cycle question has lived in the gap between the diabetologist (who sees the glucose) and the gynaecologist (who sees the cycle), and neither has been routinely trained on the joint pattern. The literature has caught up in the last few years; the clinic conversation is catching up more slowly.

Bring your data. Two or three cycles of CGM with the cycle dates marked, ideally as a printed AGP overlay or a screenshot, gives the team something concrete to read with you. Name the phase: “my time in range drops in the seven days before my period” lands more cleanly than “I have a problem with my cycle”. Ask specifically: the conversations that move are the ones that ask “would it be reasonable to look at a cycle-phase basal pattern?” or “would it be reasonable to adjust my insulin sensitivity in the luteal week?”, not the ones that ask “is this normal?”. If the first answer is dismissive, ask in writing. A clinic letter or secure message with the data attached creates a record. Ask again at the next appointment.

Your diabetes care team is the first conversation. Personalised dose adjustments are theirs to make with you, on your data, in your context. This guide does not give doses, and no online resource should. What it can give you is the language to ask, and the literature to anchor the ask.

Evidence backbone, Part 1

Site delivery is voice-first; the underlying source layer is research-grade and lives in the GNL Grace wiki. Tap any source for the citation detail.

The full grade map (per-source grade, per-claim chain, residual gaps) sits at gnl-grace/wiki/evidence-grades/menstrual-evidence.md. The concept page Grace’s RAG draws from is gnl-grace/wiki/concepts/menstrual-cycle-and-t1d.md.