Can an app show how food affects blood sugar without a sensor?

An app without a sensor cannot measure blood sugar. What it can do is estimate how a meal is likely to affect it, using what research has shown about food and the body. Whether that estimate is useful depends on what you want to know.

What happens to blood sugar after a meal?

When a meal contains carbohydrate, glucose from digestion enters the blood, rises for a while, and is brought back down as insulin helps the body’s tissues take it up. In a study of 153 healthy, non-obese people aged 7 to 80 who wore blinded continuous glucose monitors for up to 10 days, glucose spent a median 96% of the time between 70 and 140 mg/dL (3.9 to 7.8 mmol/L) and about 30 minutes a day above that range (Shah et al., 2019). A rise after eating is normal. What varies is its size and shape.

What makes the rise bigger or smaller?

These are the factors with the clearest evidence. Most of the studies are small, so they show direction more reliably than size.

• How much carbohydrate, and what kind. The glycemic index compares the rise in blood glucose after a food with the rise after the same amount of carbohydrate taken as pure glucose (Jenkins et al., 1981). The 2021 international tables list values for more than 4,000 items and show that some food groups vary a lot: breads, breakfast cereals and rice come in high-, medium- and low-GI versions (Atkinson et al., 2021). In 1,002 healthy UK adults, the meal’s macronutrient content explained about 15% of the variation in glucose responses, more than person-specific factors such as the gut microbiome (about 6%) (Berry et al., 2020). What is on the plate matters, but it leaves most of the differences between responses unexplained.
• Fiber, depending on the kind. In the original glycemic index study, a food’s total fiber content did not predict its glucose rise (Jenkins et al., 1981); the type of fiber seems to matter more than the amount. A meta-analysis of 103 short feeding-trial comparisons (538 participants) found that adding oat beta-glucan to a carbohydrate meal lowered the glucose rise by about 23% on average, but the size depended on the dose and only the higher-molecular-weight forms had a clear effect (Zurbau et al., 2021). In a small study of 15 healthy people, an evening meal of barley-kernel bread, or of white bread with added barley fiber and resistant starch, improved glucose tolerance at a standard breakfast the next morning (Nilsson et al., 2008), so one meal can shape the response to the next.
• Fat and protein. Across 62 foods tested in small groups of healthy volunteers, foods with more fat or protein tended to give smaller glucose rises (Jenkins et al., 1981). Both slow how quickly the stomach empties. In 6 men with type 2 diabetes, 30 ml of olive oil taken 30 minutes before a potato meal slowed stomach emptying and delayed the glucose rise and its peak (Gentilcore et al., 2006). In 8 people with type 2 diabetes, 55 g of whey protein before or with a potato meal reduced the rise, through slower stomach emptying and a larger insulin response (Ma et al., 2009). These were small studies with large doses, so the size of the effect in an ordinary meal, or in other people, is less certain.
• Time of day. In a tightly controlled laboratory study of healthy adults, the same meal produced a glucose response about 17% higher at 8 pm than at 8 am, an effect the authors traced to the body clock itself, separate from the daily cycle of sleep and meals (Morris et al., 2015). In 20 healthy volunteers given identical meals, the rise was smaller at breakfast than at lunch or dinner (Saad et al., 2012). There is more on this in why the same meal hits differently at 8 am and 8 pm.
• Activity. Contracting muscle moves glucose transporters to its surface through signals set off by the contraction itself (Richter and Hargreaves, 2013). A meta-analysis of 7 one-day crossover trials, mostly in adults with overweight, found that short light-walking breaks from sitting lowered glucose and insulin after meals, and lowered them more than standing breaks did (Buffey et al., 2022). In a trial in 41 adults with type 2 diabetes, advice to walk for 10 minutes after each main meal gave a post-meal rise about 12% smaller than advice to walk 30 minutes once a day with no set timing, and about 22% smaller after the evening meal (Reynolds et al., 2016).
• Sleep. After six nights of four hours in bed, 11 healthy young men handled glucose less well than after six nights with 12 hours in bed (Spiegel et al., 1999). This was a short laboratory study, and how much ordinary short nights matter for one person is less clear.

Why does the same meal affect people differently?

Large studies with continuous glucose monitors found that identical meals produce very different responses from one person to the next. In an Israeli cohort of 800 people and 46,898 meals, the response to the same meal was highly variable (Zeevi et al., 2015). In 1,002 healthy UK adults, twins among them, the variation between people in glucose response to identical test meals was large, with a coefficient of variation of 68% (Berry et al., 2020).

Predicting one person’s response is hard even with a lot of data. The Zeevi model was built on a week of sensor readings per person and also used blood tests, body measurements, the gut microbiome, diet and activity. The Berry model reached a correlation of 0.77 for glucose using similarly rich data. In 327 US adults without diabetes, a personalized model predicted responses better than calorie or carbohydrate counts alone (R of 0.62 against 0.34 and 0.40), after six days of sensor data (Mendes-Soares et al., 2019). The Berry and Mendes-Soares papers both declare authors who work for, or consult for, companies that sell personalized nutrition, which is worth keeping in mind.

For any estimate made without a sensor, this sets a clear limit. It can describe the typical response to a meal, adjusted for the things it knows about, but it has no direct way to see how one person’s body differs from the average for that meal.

What does a continuous glucose monitor measure, and what are its limits?

A continuous glucose monitor (CGM) is a small sensor worn on the body that reads glucose in the fluid under the skin every few minutes. It gives real readings from the wearer’s own body, which no estimate can do.

It has limits of its own:

• A short delay. Glucose reaches the fluid under the skin a few minutes after it changes in the blood. In 8 healthy adults after an overnight fast, the delay was about 5 to 6 minutes, which the authors judged small enough not to undermine sensor accuracy in the fasting state (Basu et al., 2013). A lag of a few minutes matters little when glucose is steady and more when it is changing quickly, as after a meal.
• Device disagreement. When 16 adults without diabetes wore two different CGMs at once during 28 days on a research ward, the two devices ranked the same 760 meals quite differently (a mean rank correlation of 0.43) (Howard, Guo and Hall, 2020). The authors concluded that choosing meals by one sensor’s ranking may be premature. Meal-by-meal comparisons from a single sensor carry noise.
• Readings, not reasons. A sensor shows what happened, not why. Explaining a curve still needs a record of what was eaten and of how the person slept and moved.

What is an estimate useful for, and what is it not?

An estimate is useful for:

• seeing patterns across days, such as which meals tend to give a larger estimated rise;
• comparing two versions of a meal, for example with more fiber or with protein first;
• seeing how the same meal looks earlier or later in the day;
• seeing the estimated effect of a short walk after eating;
• learning how carbohydrate, fiber, fat, protein, sleep and activity tend to work together.

It is not useful for:

• diagnosing anything;
• any decision about medication, including insulin;
• exact numbers, targets or thresholds;
• knowing whether one person’s response is high or low compared with other people.

How does IQFood do it?

IQFood takes the estimate route, and since we make it, read this paragraph with that in mind. From the meals you log (their carbohydrates, fiber, protein and fat) and short daily check-ins about sleep, activity, workouts and stress, it estimates how glucose, insulin, ketones, energy and fat-burn are likely to move through the day, and draws the next hours ahead. With your permission it reads steps, sleep and workouts from Apple Health (iPhone) or Health Connect (Android). It does not connect to glucose sensors and does not measure anything, and its curves can differ from real physiology. The method, its limits and the 45 sources behind it are on the science page.

When is it worth talking to a professional?

A doctor or registered dietitian is the right person when there are symptoms after eating that worry you, a condition that affects blood sugar, medicine that lowers it, a pregnancy, or a plan to use sensor readings to make health decisions. Questions about a blood test or a sensor reading belong with the clinician who ordered it. No app curve, estimated or measured, replaces that conversation.

This article is educational and is not medical advice. IQFood is not a medical device and does not diagnose, treat, cure or prevent any condition.