Does menstrual cycle affect blood sugar? What CGM data shows
How the menstrual cycle affects blood sugar: in a 209-woman CGM study, glucose swings peaked late in the luteal phase while average glucose stayed flat.
- Research & Methods
- Glucose
Short answer: It can, though mostly through swings rather than levels. In a 209-woman continuous glucose monitor (CGM) study, the menstrual cycle did not move average blood sugar in women without diabetes. It changed how much glucose swung around that average, with the biggest swings in the late luteal phase, the stretch of days before a period starts.
That finding comes from a preprint by Aharoni and colleagues at the Weizmann Institute, posted on medRxiv on 15 October 2026 (Aharoni et al., 2026, doi:10.64898/2026.10.14.26365513). It has not been peer reviewed yet, and it studied women without diabetes, so it says nothing direct about type 1 or type 2 diabetes management.
This article is educational and is not medical advice. If you manage diabetes or have symptoms of high or low blood sugar, talk with a clinician.
Mean glucose vs glycemic variability
Two different questions hide inside "does my cycle affect my blood sugar?"
The first is about level. Is your glucose higher on some days of the cycle than others? Mean glucose and the glucose management indicator (GMI), a CGM-based estimate of HbA1c, answer that one.
The second is about movement. How much does glucose rise after meals and fall again, and how wide is the gap between a day's highs and lows? Researchers call this glycemic variability. Common measures include the standard deviation (SD), the coefficient of variation (CV, the SD divided by the mean), and MAGE, the mean amplitude of glycemic excursions.
A person can have a flat average with bigger daily swings, or the reverse. Older studies of the menstrual cycle mostly relied on occasional blood draws, sometimes one measurement per phase, which can catch a level but cannot see swings. The Weizmann team argues that this is one reason earlier results disagreed with each other.
Late-luteal peak in a 209-woman CGM cohort (Aharoni et al. 2026 preprint)
The study drew on the Human Phenotype Project, a large Israeli research cohort. After screening, 209 naturally cycling women remained. They were not using hormonal contraception or hormone therapy and had no diagnosed diabetes. Their average age was 43.2 years and average BMI 24.3. Each wore a FreeStyle Libre Pro iQ sensor for up to 14 days while logging meals in an app in real time.
To compare women with different cycle lengths, the researchers stretched every cycle onto a 0 to 100% scale. Day one of the period is 0%, estimated ovulation sits at 50%, and the next period starts at 100%. They then fit smooth cyclic curves to 36 daily CGM metrics, adjusting for age and BMI.
What they found:
- Average glucose did not change. Mean glucose and GMI showed no significant cycle pattern (both q = 0.126).
- Six variability measures did. CV, MAGE, SD, glucose range, mean absolute glucose change (MAG) and a within-series SD all followed the same wave. They were lowest at about 38 to 41% of the cycle, just before estimated ovulation, and highest at about 80 to 85%, in the late luteal phase. A smaller secondary peak appeared around 20 to 25%.
- The swing was modest. From low point to peak, the change was about 6 to 10% of the cycle average. For CV that was 1.15 percentage points, and for the daily glucose range 5.22 mg/dL.
- Men showed no pattern. As a check against analysis artifacts, the team gave 209 age- and BMI-matched men random "pseudo-cycles". None of the 36 metrics was cyclic in the men.
The effect lived mostly in the daytime. Hourly CV was 11.25% in the luteal phase against 10.45% in the follicular phase during the day, while overnight values were close (7.53% vs 7.22%).
Food explained part of the picture. Across 3,815 meals, glucose fell back from its post-meal peak 11.5% faster in the luteal phase. Peaks and total excursion (iAUC) were only slightly higher and not statistically significant. Breakfast was where the gap opened: luteal breakfasts had a 15.2% larger peak rise, 20.4% larger iAUC and 22.5% faster descent. When the researchers compared breakfasts of similar carbohydrate, protein and fat content within the same woman, iAUC was 42.5% higher and descent 53.2% faster in the luteal phase.
A faster descent sounds like a good thing, but the authors read it as the other side of a bigger spike: a higher peak brings a steeper fall back to baseline. Pre-meal glucose, daily calories and carbohydrate intake did not differ between phases, and the variability wave held up after adjusting for carbohydrates and for logged exercise. That argues against "people just eat more before their period" as the whole explanation.
Use the chart to place a cycle day on the study's timeline, or compare meal responses between phases.
Glucose swings across the menstrual cycle
From a CGM study of 209 women without diabetes (preprint). Place a cycle day on the study's timeline, or compare meal responses in the two halves of the cycle.
Day 24 of a 28-day cycle sits at about 82% on the study's normalized scale (luteal half). That falls in the study's "biggest swings" window (80–85%). In the study, six glycemic variability measures were lowest at 38–41% and highest at 80–85%, while mean glucose did not change across the cycle.
What it shows: the cycle changed how much glucose moved, not where it sat on average. Day-to-day swings were smallest just before estimated ovulation and largest late in the luteal phase, and the meal effect showed up mostly at breakfast. The day mapping uses the study's own calendar method (ovulation assumed 14 days before the next period), so it is an estimate, and the windows describe the group average, not any one person.
Data table
| Metric | Swing | 95% CI |
|---|---|---|
| CV (Coefficient of variation) | 1.15 percentage points | 0.76–1.85 |
| MAGE (Mean amplitude of glycemic excursions) | 3.11 mg/dL | 1.64–5.13 |
| SD (Standard deviation) | 1.05 mg/dL | 0.6–1.75 |
| Range (Glucose range (max minus min)) | 5.22 mg/dL | 2.94–8.42 |
| MAG (Mean absolute glucose change) | 0.76 mg/dL per hour | 0.44–1.21 |
| SDwsh (SD within series (hourly)) | 0.29 mg/dL | 0.18–0.5 |
| Meal set | Peak rise | iAUC | Descent rate |
|---|---|---|---|
| All meals (3,815) | +4.1% (q = 0.187) | +5.3% (q = 0.187) | +11.5% (q = 0.015) |
| Breakfasts | +15.2% (q = 0.016) | +20.4% (q = 0.016) | +22.5% (q = 0.016) |
| Matched breakfasts, same woman | +32.5% (q < 0.001, n = 78) | +42.5% (q = 0.0011, n = 78) | +53.2% (q < 0.001, n = 72) |
| Time of day | Luteal | Follicular |
|---|---|---|
| Daytime | 11.25% | 10.45% |
| Overnight | 7.53% | 7.22% |
What about low or high blood sugar during a period?
Searches for "low blood sugar during period, not diabetic" and the high-sugar version are common. This study offers limited help. Time below 70 mg/dL and time above 140 mg/dL were among the 36 metrics tested, and neither was one of the six that passed the study's confirmation criteria. The variability peak sits in the late luteal phase, before bleeding starts, not during the period itself. The swings involved are a few mg/dL in people whose glucose stayed in a healthy range. Shakiness, sweating, confusion or very high readings are a reason to check with a clinician, not something to put down to your cycle.
Why calendar-estimated ovulation is a limitation
The 50% mark on that timeline is an estimate. The team rebuilt each cycle day from self-reported period dates and assumed ovulation fell 14 days before the next period. Nobody measured ovulation directly, and the authors list this as a main limitation: some days will have been assigned to the wrong phase, which can blur or shift the timing of the peaks.
They did run two checks. In 139 women with blood metabolomics, an estrogen metabolite peaked before estimated ovulation and a progesterone metabolite peaked after it, as expected at the group level. Results also held when the luteal phase was set anywhere from 12 to 16 days. Neither check confirms ovulation in any single woman.
There are other limits. Participants were mostly midlife women, with BMI between 18 and 35 and regular cycles. Sensors read about every 15 minutes, meals and exercise were self-reported, and the design was observational, so it shows association rather than cause.
A smaller study approached the problem differently. Lin and colleagues followed 49 participants aged 18 to 30 for an average of 79.3 days, with a Dexcom G6 CGM, a Fitbit Sense and daily urine hormone tests that marked each phase (Lin et al., 2023, doi:10.1038/s41746-023-00884-x). They looked at daily median glucose and did find a biphasic level change: lowest in the late follicular phase (5.8 mmol/L) and highest in the luteal phase (6.1 mmol/L). Both studies place the low point near late follicular and the high point in the luteal phase. They disagree on whether the average itself moves, and younger participants, hormone-defined phases and a different glucose summary are all plausible reasons.
What this means for CGM and wearable metabolic features
The Weizmann authors suggest that cycle phase may deserve a place in how CGM data from premenopausal women are read. A 14-day sensor covers only part of a cycle, so two wears that start at different points can look different for reasons that have nothing to do with diet or health. CGMs are also used more and more over the counter, often with reference ranges drawn from mixed groups of men and women.
For people building wearable metabolic features, the lesson is about which statistic to track. A feature that only watches mean glucose would have missed every cycle effect in this dataset. The signal was in variability, meal responses and time of day, and cycle-aware features in general need those dynamics rather than single averages. Our post on HRV across the menstrual cycle covers the heart-rate side, and two open research datasets pair wearables with cycle labels: the mcPHASES menstrual wearable dataset and a wristband heart-rate and ovulation dataset from Soochow.
None of this means a watch can read your glucose. No smartwatch measures blood sugar through the skin today; our glucose monitor watch guide explains what is and isn't available, and the state of PPG glucose research covers why optical methods are hard. For the broader link between blood sugar and the heart, see PPG, diabetes and cardiovascular risk. The preprint itself has a record in the ChatPPG research hub, and the research hub tracks newer wearable papers.
References
- Aharoni S, Godneva A, Shaulitch R, et al. Menstrual cycle-related dynamics of glycemic variability in women without diabetes. medRxiv (preprint, not peer reviewed). Posted 2026-10-15. doi:10.64898/2026.10.14.26365513
- Lin G, Siddiqui R, Lin Z, et al. Blood glucose variance measured by continuous glucose monitors across the menstrual cycle. npj Digital Medicine. 2023;6:140. doi:10.1038/s41746-023-00884-x
Frequently Asked Questions
- Does your period raise blood sugar?
- Not on average, in the largest CGM study so far. Among 209 naturally cycling women without diabetes, mean glucose and the glucose management indicator did not vary across the cycle. What changed was how much glucose moved: day-to-day swings were largest late in the luteal phase, the days before a period starts. A smaller 2023 study that tracked hormones did find slightly higher median glucose in the luteal phase, so the question is not fully settled.
- Can the luteal phase cause glucose swings if you're not diabetic?
- In a 2026 preprint of 209 women without diabetes, six measures of glycemic variability peaked at 80 to 85% of the way through the cycle, which is the late luteal phase. The peak-to-low swing was about 6 to 10% of the cycle average. Breakfast spikes were 15.2% larger in the luteal phase. These are small, normal-range changes, and the study is not yet peer reviewed. This is educational information, not medical advice.
- Is a CGM useful across the menstrual cycle?
- It can show patterns a single finger-prick or lab test misses, which is why researchers used CGMs here. The study authors suggest cycle phase may need to be considered when CGM data from premenopausal women are interpreted, because a 14-day sensor covers only part of a cycle. If you have diabetes or symptoms of high or low blood sugar, talk with a clinician; this study did not include people with diabetes.
- Does average glucose change with the menstrual cycle?
- The 2026 Human Phenotype Project preprint found no significant change in mean glucose or GMI across the cycle (both q = 0.126). A 2023 npj Digital Medicine study of 49 younger participants, with phases set by daily urine hormone tests, found daily median glucose lowest in the late follicular phase (5.8 mmol/L) and highest in the luteal phase (6.1 mmol/L). Different ages, methods and phase definitions likely explain part of the gap.
- Can wearables detect cycle phase from glucose?
- Not reliably yet. The cycle effects in these studies are group averages and small compared with differences between people, and the larger study estimated ovulation from calendar dates rather than measuring it. Researchers have suggested glucose could become one input among several, alongside signals such as temperature and pulse rate, but no validated glucose-based cycle tracker exists.
Related articles
Glucose monitor watch: what works today and what the research shows
No glucose monitor watch can read blood sugar through the skin yet. Here's what the FDA says, what research shows, and how to get glucose on your wrist.
Apple Watch hypertension notifications: how PPG spots possible high blood pressure
Apple Watch hypertension notifications read 30 days of PPG data. FDA filings show 41.2% sensitivity and 92.3% specificity. Here's what an alert means.
The Cardiac Cycle in the PPG Signal: Systole, Diastole, and What Each Phase Reveals
Every PPG waveform is a compressed diary of one heartbeat. The rising edge, the peak, the notch, and the falling tail each correspond to distinct mech...
Samsung Galaxy Watch blood pressure: how it works and drifts
Galaxy Watch blood pressure needs a cuff every 28 days. Data from 1,714 users show readings shift about 6 mmHg per recalibration, more with age.
Bilateral Ring PPG: What a 1,810-Person Study Shows About HR, BP, and Left–Right Differences
A large-scale bilateral ring PPG study measured both hands simultaneously in 1,810 participants. The results reveal where ring sensors excel (heart rate), where they show modest utility (age), and where limitations remain (blood pressure).
PPG Data Augmentation
Data augmentation for PPG models is not as simple as flipping images or cropping photos. The physiological constraints of cardiac signals define hard ...
Devices in this article
Research records with the datasets, papers and published validation results for each device.