It's 9pm. You've eaten well all day — genuinely well. Protein at lunch, vegetables, no mid-afternoon snack spiral. You're not even hungry. And yet the pull toward the biscuit tin, the chocolate, the cereal straight from the box, is almost overwhelming. You resist. You go to bed. You wake up and wonder what's wrong with you.
Nothing is wrong with you. With PCOS/PMOS, this is one of the most predictable patterns in the condition — and it has a physiological explanation that has absolutely nothing to do with willpower or discipline.
Research consistently shows that women with PCOS/PMOS experience significantly higher rates of food cravings — especially sweet and high-carbohydrate foods — compared to women without the condition. And the evening timing is not coincidental. There are four specific biological mechanisms that converge at 9–10pm, and once you understand them, the craving stops feeling random and starts feeling like information.
The 4 Biological Reasons PCOS/PMOS Cravings Peak at Night
1. Insulin resistance and the evening glucose dip
Insulin resistance is present in up to 70% of PCOS/PMOS cases — it means your cells are less responsive to insulin throughout the day, so blood glucose regulation is already under strain before dinner. If you ate a carbohydrate-heavy lunch or dinner, your body likely produced a larger-than-normal insulin spike to manage the glucose load. Two to three hours later, a rebound drop in blood glucose triggers a neurological "fuel needed now" signal — timed, with unfortunate precision, for somewhere between 9 and 10pm.
This isn't hunger. You ate enough. Your body isn't running low on fuel. What you're feeling is glycaemic instability — and it presents as an urgent, specific craving for fast sugar, which is exactly the thing that will spike glucose rapidly. The craving is technically trying to solve a problem. It's just solving the wrong one.
2. Cortisol rhythm inversion
In a typical circadian rhythm, cortisol peaks at 7–8am and declines steadily through the day, reaching its lowest point in the early hours of the night. Many women with PCOS/PMOS have a flattened or inverted cortisol pattern: lower in the morning (which is why getting started feels so difficult), with a secondary mini-peak in the evening.
That evening cortisol rise drives a glucose release, which triggers an insulin response, which can then produce a dip — and you're back in the craving loop, but now it's cortisol-driven rather than meal-driven. This is also why stress-heavy days reliably produce worse night cravings. The cortisol and sleep connection is one of the most underappreciated factors in PCOS/PMOS symptom severity, and it's particularly relevant here.
3. Dopamine and the reward-deficit state
Insulin resistance doesn't only affect glucose metabolism. It affects dopamine signalling in the brain's reward centres. When insulin can't do its job efficiently, the brain's dopamine pathway becomes less responsive — and dopamine is the anticipation neurotransmitter, not just the pleasure one. It drives the wanting, the seeking, the pull toward something before you've even had it.
When dopamine signalling is blunted, simple carbohydrates — which deliver glucose to the brain rapidly — become uniquely compelling. The brain is seeking a dopamine hit, and sugar is the fastest route available. This is why PCOS/PMOS food noise isn't a matter of wanting sweet things more than other people. It's the brain experiencing a reward-deficit state and taking the most direct route out of it.
4. Luteal phase progesterone drop
In the week before a period, progesterone falls sharply. Progesterone is calming — it converts to allopregnanolone, which binds GABA receptors and produces a mild sedating, anxiety-reducing effect. When progesterone drops, cortisol feels amplified, sleep worsens, and carbohydrate cravings intensify significantly. The body is seeking serotonin precursors, found in carbohydrates, to compensate for the loss of GABAergic calm.
This is why night cravings get dramatically worse in the days before a period — it's not a coincidence or a failure of resolve in a vulnerable moment. It's a specific hormonal mechanism producing a specific behavioural response. And it's entirely predictable, which means it's trackable.
What Your Check-In Data Is Actually Telling You
Once you know the mechanisms, the signals in your daily check-ins start to make sense as a diagnostic rather than just a log.
- Evening cravings + morning energy below 5 → likely glycaemic instability. The glucose dip happened. Check what you ate 2–3 hours before the craving hit, and look for a high-carb meal in that window.
- Evening cravings + stress score above 6 → cortisol evening spike driving it. The craving is stress-mediated, not hunger-mediated. The food isn't the problem.
- Evening cravings clustered on cycle days 24–28 → luteal progesterone crash. Predictable; trackable; not random. This is the most treatable pattern once it's identified.
- Evening cravings + sleep quality below 5 → the bidirectional loop. Poor sleep elevates ghrelin (the hunger hormone) and lowers leptin (the satiety signal) by the following evening, making the next night's cravings measurably worse. One bad night compounds into the next.
- Evening cravings absent on days you walked 7,000+ steps → exercise and insulin resistance are directly connected. On high-step days, insulin sensitivity improves and glucose is more stable through the evening. The craving doesn't arrive because the dip didn't happen.
The core insight: you can't suppress these cravings with willpower because they're not hunger — they're metabolic signals. But once you know which mechanism is driving yours on any given night, you can address it directly rather than trying to overrule it.
The 3 Patterns Worth Looking For Over 14+ Days
Fourteen days of consistent check-ins produces a pattern that a single evening of self-awareness can't. Here's what's worth investigating:
The food-timing pattern
Log what you ate for dinner and when. Then log craving intensity at 9–10pm, on a simple 1–5 scale. After 14 days, look for whether high-carb dinners consistently predict stronger cravings 2–3 hours later. If that correlation holds, the mechanism is glycaemic instability, and shifting to a protein-first dinner structure may break the cycle. The data tells you whether that change is working within two weeks — you don't have to guess.
The cycle-phase overlay
If you're tracking cycle day, check whether your worst craving nights cluster in the last 7 days of your cycle. If they do, this is the luteal progesterone drop — a hormonal pattern, not a food problem. Magnesium glycinate supplementation, complex carbohydrates (rather than refined sugar), and a slightly earlier bedtime during this window are evidence-based responses to this specific mechanism. The key is identifying that this is the driver before trying an intervention — because the intervention for luteal cravings is different from the intervention for insulin-driven cravings.
The stress-craving correlation
Use your morning stress data alongside your evening craving scores. Rate your stress each morning. Then check whether your evening cravings are strongest on days when morning stress was high. If there's a clear correlation across 14+ days, cortisol inversion is the driver — and the intervention is earlier in the day (a morning walk, a protein breakfast, not skipping meals) rather than anything at 9pm.
Atlas's pattern engine surfaces these correlations automatically at Premium tier. You don't have to calculate the correlations yourself or remember to look — they appear in your pattern summary when the data supports them.
What to Actually Do With This
The interventions below are deliberately mechanism-specific. The generic advice — "eat more protein," "cut sugar," "go to bed earlier" — isn't wrong, but it's imprecise. Imprecise interventions produce inconsistent results, and inconsistent results feel like failure. The point is to match the intervention to the mechanism.
If it's glycaemic instability: prioritise protein before carbohydrates at dinner — eating protein first reduces the insulin spike from the meal. Add 10 minutes of walking after dinner; post-meal movement is an efficient insulin sensitiser. A small fat and protein snack (a handful of almonds, a little cheese) at around 8pm can stabilise blood glucose before the dip would otherwise hit.
If it's luteal-phase-driven: don't fight it with restriction — that strategy reliably fails in the luteal window. Instead, switch from ultra-refined carbohydrates (biscuits, sweets, white bread) to complex ones: oat cakes, a banana, dark chocolate at 70%+. The brain needs serotonin precursors; you're just choosing a better delivery vehicle. The craving is appropriate. The food choice is what changes.
If it's cortisol-driven: the craving is a stress symptom, not a food problem — and trying to solve it at 9pm with dietary choices is the wrong point of intervention. A cortisol reset earlier in the day (a 10-minute walk, a protein breakfast, no meal-skipping) reduces the evening mini-spike before it drives the craving. This one requires patience because the intervention and the symptom are hours apart.
If it's sleep-debt compounding: prioritise one night of earlier sleep. The leptin and ghrelin reset takes 24–48 hours, but it works faster than any dietary change. One night doesn't fix a pattern, but it breaks the compounding loop — and sometimes that's all you need to interrupt the cycle long enough to regain some stability.
The common thread is that none of these interventions is "eat less" or "try harder." They're all upstream resets. Once you know which mechanism is running your cravings, the intervention almost picks itself.
Start Seeing Your Pattern
Night sugar cravings with PCOS/PMOS aren't a mystery — they're a signal with a source. The pattern engine in Atlas correlates your evening craving scores with your meal timing, stress scores, sleep quality, and cycle day to show you which mechanism is driving yours — and whether your interventions are working.
Start Atlas on a monthly plan. The pattern engine does the correlation work — you just have to check in for 30 seconds a day. momentum.madethis.app