Assessing temporal eating pattern in free living humans through the myCircadianClock app

Abstract

The quality and quantity of nutrition impact health. However, chrononutrition, the timing, and variation of food intake in relation to the daily sleep-wake cycle are also important contributors to health. This has necessitated an urgent need to measure, analyze, and optimize eating patterns to improve health and manage disease. While written food journals, questionnaires, and 24-hour dietary recalls are acceptable methods to assess the quantity and quality of energy consumption, they are insufficient to capture the timing and day-to-day variation of energy intake. Smartphone applications are novel methods for information-dense real-time food and beverage tracking. Despite the availability of thousands of commercial nutrient apps, they almost always ignore eating patterns, and the raw real-time data is not available to researchers for monitoring and intervening in eating patterns. Our lab developed a smartphone app called myCircadianClock (mCC) and associated software to enable long-term real-time logging that captures temporal components of eating patterns. The mCC app runs on iOS and android operating systems and can be used to track multiple cohorts in parallel studies. The logging burden is decreased by using a timestamped photo and annotation of the food/beverage being logged. Capturing temporal data of consumption in free-living individuals over weeks/months has provided new insights into diverse eating patterns in the real world. This review discusses (1) chrononutrition and the importance of understanding eating patterns, (2) the myCircadianClock app, (3) validation of the mCC app, (4) clinical trials to assess the timing of energy intake, and (5) strengths and limitations of the mCC app.

INTRODUCTION

Accurate assessment of nutrient intake in free-living humans has always been a challenge. This obstacle has only become more complex as a larger proportion of energy intake is consumed from snacks throughout the day [1]. Food and beverages as snacks are more commonly associated with social interactions or enjoyment than with physical needs, and logging quality, quantity, and timing of every ingestion event with high accuracy can be burdensome.

Preclinical and human studies have demonstrated the importance of the timing of food intake and its variation from day-to-day. Variations in meal timing, between weekdays and weekends, and between weeks, can conflict with circadian rhythms. The circadian clock regulates many aspects of metabolism [2, 3] by modulating the timing of the optimal level or function of neuroendocrine regulators, signaling pathways, and metabolic enzymes. The outcome of these coordinated rhythms is a window of time when food is better digested and nutrients are better utilized. As the circadian system cannot rapidly adjust to abrupt changes in mealtimes by >1 h (h) between days–specifically breakfast and dinner–abrupt changes in mealtime can lead to food being ingested when digestion and utilization of nutrients are sub-optimal. If such a mismatch between the ideal circadian eating window and of actual eating pattern continues over days, weeks, or months, it can compromise health [2, 4, 5]. Conversely, consistently consuming all energy intake within an optimal window can reduce disease risk. Thus, tools to capture temporal components of nutrient intake over several weeks are needed.

Smartphone apps offer promising tools to monitor temporal eating patterns. However, most commercial apps ask users to log every food or beverage and portion size to calculate calories and the daily energy intake or trend is shown to the user to guide her/him to manage calories. Some commercial apps ask users to monitor the daily fasting duration. However, none of these commercial apps make their raw data available in real-time for research nor do they allow researchers to monitor and intervene timing of eating. The myCircadianClock (mCC) smartphone app (La Jolla, CA, USA) was developed by our lab to address this unmet research need. Since there is no comparable research app available at the time of preparing this manuscript, we will focus our article on the mCC app as a tool to monitor and intervene in an individual’s eating patterns. This review will discuss [1] chrononutrition and the importance of understanding eating patterns, [2] the myCircadianClock app, [3] validation of the mCC app, [4] clinical trials to assess the timing of energy intake, and [5] strengths and limitations of the mCC app.

CHRONONUTRITION AND THE IMPORTANCE OF UNDERSTANDING EATING PATTERNS

Circadian rhythms, circadian disruption, and health

Circadian is Latin for “about a day”. It is an endogenous time-keeping system that anticipates the optimum behavioral and physiological state at different times of the day or night and optimizes the functions of relevant organs by elevating or suppressing the functions of thousands of genes at the appropriate times of the day. These complex genomic programs result in overt daily rhythms in sleep, activity, and hunger. In other words, just like the circadian clock specifies an optimum window at night when a person is most likely to have restorative sleep, there are optimum windows for eating and physical activities. However, with the change in seasons, the circadian system also adjusts to the changing daylengths and food availability. While the mechanisms that mediated adjustment of the circadian clock to seasonal changes offered an adaptive advantage in the pre-industrial era, prolonged nightly lighting and erratic eating pattern in modern societies can disrupt the circadian clock. Consequently, the genomic program that maintains homeostasis of behavior, physiology, and metabolism is disrupted, which leads to increased risk for chronic diseases.

Excessive light at night can directly or indirectly affect sleep and circadian rhythms and upset the neuroendocrine component of metabolic homeostasis [4, 6]. Modern lifestyles also cause another type of circadian disruption. The timing of food relative to daily rhythms in sleep and activity, known as chrononutrition, also significantly affects health. Food, more importantly, the first meal of the day acts as an “entraining cue” for the circadian clocks present in metabolic organs [7]. Consequently, random changes in breakfast time from one day to another day or eating over a long period of time during a 24 h day can cause circadian disruption and upset metabolic homeostasis [8, 9]. Circadian rhythms anticipate breakfast at a habitual time and ensure an optimum level of digestion, absorption, and utilization of nutrients if breakfast occurs at a habitual time [10, 11]. Abrupt changes in breakfast time can cause a breakfast event to occur at a sub-optimal circadian time for breakfast, which may at least partly explain why frequently skipping breakfast can contribute to metabolic disease [12, 13]. Second, spreading eating events over a long window of 24 h day is akin to the central clock being exposed to continuous or prolonged light, which disrupts the circadian clock and also disrupts the metabolic processes that typically occur during overnight fasting; such as fatty acid oxidation [14, 15]. Third, major outputs of the circadian system–physical activity–also appear to have a time of the day-dependent effect on metabolism. Recent studies have shown afternoon exercise can help better control blood glucose among patients with Type 2 Diabetes Mellitus (T2DM) [16, 17]. Altogether, the timing and duration of sleep, the timing of breakfast and window of eating, and the timing of exercise and intensity of exercise can affect metabolism. As the domain of circadian regulation of metabolism is relatively new and most studies are preclinical by nature, we are beginning to witness epidemiological studies connecting timing of sleep, and food affecting health and intervention studies assessing how the timing of eating can be a new tool to reduce the burden of disease [12, 18–21].

Temporal components of eating patterns

The timing of food consumption significantly impacts health [9, 15, 22–24] through at least four different parameters; [1] the time of day (phase), [2] variation in eating time, [3] frequency, and [4] duration of daily feeding and fasting [25].

Phase

The body responds to food, beverages, and medication differently based on the time of day (relative to an individual’s activity/sleep schedule) that it is consumed. This is due to both the current state of nutrient availability and the circadian system coordinating physiology with the anticipated feeding/fasting schedule. For instance, in the middle of the night, in anticipation of fasting, the circadian system increases the production of glucagon and reduces the production of insulin. Nightly sleep hormone melatonin further blunts the release of insulin. Consequently, a late-night snack can cause a much higher rise in blood glucose than it would during the day when the body is anticipating energy intake and insulin is ready for rapid release to regulate blood glucose. Late-night eating has been associated with an increased risk for cardiometabolic diseases [26].

Duration

The duration or window of time (out of the 24 h day) when food and beverages are consumed and conversely, the length of fasting are important as well. Just like there is an interest in optimum duration of sleep for promoting health, recent focus has been on the optimum length of eating- or fasting-window on health outcomes [9, 24, 27–31]. Research suggests a consistent 8–12 h is an optimum eating window leaving 12–16 h of a daily fast [30, 32–46]. Such consistent daily period of the eating-fasting cycle supports a robust circadian rhythm, which in turn maintains optimum metabolism by optimizing nutrient metabolism during the eating window and consistently elevating fatty acid oxidation and tissue repair during the fasting period [22]. Consistent daily fasting has been shown to decrease weight, blood pressure, cholesterol, and improve glucose regulation [29, 30, 34, 35]. Moreover, there is also a positive correlation between the length of the eating window and BMI [35].

Frequency/Distribution

The distribution of energy intake throughout the day has also been shown to affect glucose metabolism and weight. A weight-loss study in women found that larger meals in the morning with smaller dinners significantly improved weight loss and decreased appetite [47]. Many weight loss programs have participants eat 6 meals a day to help regulate blood glucose. However, a recent study with participants that had T2DM found that 3 meals per day (compared to 6 meals per day) led to significant improvements in lowering HbA1c, weight loss, appetite regulation, and a decrease in insulin administration [48].

Variation/Regularity

The regularity of energy intake also plays an important role in health outcomes. The circadian system plays a large role in glucose regulation and metabolism and functions as an anticipatory system to coordinate physiology. For instance, it helps to regulate glucose metabolism by regulating insulin and glucagon. Irregular eating times have been associated with increased cardiometabolic risk [5].

Do we need to log more than meals? And What is a meal?

The terms meal and snack are difficult to define and thus, have many definitions that are not universally agreed upon. These may be defined by the time of day, size, or type of food [49]. Many studies that included food logging allow participants to determine what constitutes a meal or a snack [50]. One common definition of a meal is an eating occasion that occurs at approximately the same time each day, frequently of a reasonably large size, such as breakfast, lunch, and dinner. This is opposed to a snack that is typically a smaller portion and has a less regular time of consumption [50]. However, in modern society, the regularity of both the time and size of meals has drastically changed. Recent NHANES data has shown that people in the US no longer have 3 meals, but rather a combination of snacks and meals across the day [1]. Moreover, there are large cultural differences in the name, timing, and size of meals. Thus, it is necessary to log all energy intake (regardless of meal/snack classification) rather than only logging meals, especially limited to breakfast, lunch, and dinner, as it oversimplifies the individuals’ eating patterns [51].

THE MYCIRCADIANCLOCK SMARTPHONE APPLICATION (MCC APP)

Background

The basic tenet of a behavioral nutrition intervention study is to (a) monitor the current nutrition habit to screen participants and set baseline benchmark for behavior change, (b) design intervention and monitor whether the participants adhere to the behavioral nutrition recommendation, and (c) make conclusions about feasibility and efficacy of the intervention. Therefore, if we want to change an eating pattern, researchers need methods to monitor and analyze eating patterns and develop interventions to improve chrononutrition in community-dwelling individuals. It is likely that eating patterns are also linked to total energy intake or quality of nutrition. Hence, the method to monitor eating patterns itself or a companion method should collect some aspects of nutrition quality and quantity.

Unlike the wide acceptance of 24-h recall or FFQ for capturing nutrition quality and quantity, there is hardly any standard to capture a daily eating pattern and its day-to-day variation. Food diaries have can capture temporal data, but typically do not record the exact timing of food and/or are not used for long durations. Hence, the mCC app was developed to monitor the temporal aspect of food/beverage/water/medicine in people in the real world. This was inspired by research in animal models of diet-induced obesity that demonstrated that when mice were put on a high-fat diet, they also changed the timing of when they ate (increased eating during their rest phase) [52]. However, if those mice consumed an isocaloric high-fat diet, but were only able to access food at night when they are active (their normal eating time), they did not gain weight [9]. The protocol of restricting food to an 8–12 h window has since become known as time-restricted feeding (TRF) and has been replicated by many labs [53–57].

Preclinical studies [9, 27, 58] inspired research investigating temporal eating patterns in humans to determine (a) when people eat, (b) whether people can adopt a time-restricted eating (TRE) pattern, and (c) what can be the health benefits of TRE. In the first study assessing temporal eating patterns in free-living humans, the mCC app was used to assess the eating pattern in 156 adults in San Diego who did not do any shift work [32]. All participants logged all ingestive events for 3 weeks to determine their daily eating window (time of first energy consumption till the last). Although various approaches can be used to assess the daily window of eating and its day-to-day variation, we adopted the time window in which 95% of eating events occur within 2–3 weeks of food logging as the eating window. This is calculated after collecting at least 2 weeks of data and calculating the 2.5% to 97.5% interval of time in which all food intake occurred. In other words, the eating window is the time interval in which 95% of food and beverage intake likely occurs (Fig. 1). Over 50% of participants had an eating window of 14 h 40 min or more, with less than 10 h fast. This demonstrated that most adults have a long eating window and laid the groundwork for time-restricted eating (TRE) studies in a human population (Fig. 2). To test the feasibility of behavior change to adopt TRE the same study did a TRE intervention with 8 overweight participants. For 16-weeks, participants used the mCC to follow a consistent, self-selected, 10-h eating window. Participants were able to comply with a 10-h TRE intervention, lost a modest amount of body weight, and reported decreased hunger at night, and increased sleep satisfaction and energy levels. At a one-year follow-up, without using the app since the 16-week intervention, the improvements in weight, hunger, sleep, and energy were maintained [32]. This indicated TRE was a feasible and sustainable intervention for improving health outcomes. Since then, the mCC app has been a key research tool in completed and ongoing TRE intervention studies (Table 1) [30, 32, 35, 59].

Fig. 1. 95% eating window compared to a daily eating interval.

Feedograms displaying all caloric food and beverage entries (dots) over a 24-h period for 14 days during baseline (A) and during a TRE intervention (B) for a given participant. The 95% eating window is indicated by the dotted line on the feedogram and the daily eating duration is shown on the right. From Wilkinson et al., 2020.

Fig. 2. Adults have a long eating window.

24-h plots of (A) all food, beverage, water, and medication events and (B) caloric food and beverage events. (C) Percentage of participants in each eating duration bin (h). Data from 156 adults for 21 consecutive days. From Gill and Panda, 2015 [32].

Table 1.

Clinical trials that used smartphones for temporal nutrient assessment

Application	References	Study Design	Data Analyzed	Assessment	Duration	Participants (n)	Findings
myCircadianClock	Gill and Panda, 2015 (32)	Observational and Single-Arm Pilot Intervention	Time-Stamped photos on app, annotations	Eating window, Calories	3-wk baseline; 16-wk intervention (all days logged)	156 (91F) healthy adults; 8 adults (3F) intervention (≥14 h eating duration)	10-h TRE led to weight loss and improved energy and sleep.
myCircadianClock	Wilkinson and Manoogian et al, 2020 (30)	Single-Arm Pilot Intervention	Time-Stamped photos on app, annotations	Eating window, Calories	2-wk baseline, 12-wk intervention (all days logged)	19 (6F) adults with metabolic syndrome and ≥14 h eating duration	10-h TRE led to weight loss, decreases in blood pressure, cholesterol, HbA1c, and improved sleep.
myCircadianClock	Chow et al., 2020 (35); Malaeb et al., 2020 (85); Lobene et al., 2021 (86); Crose et al., 2021 (87)	Randomized Control Trial	Time-Stamped photos on app, annotations	Eating window	2-wk baseline, 12-wk intervention (all days logged)	20 (17F) adults with overweight and ≥14 h eating duration	8-h TRE: (1) led to weight loss, (2) decreased eating events and caffeine, (3) no adverse effects on bone turnover, (4) improved quality of life
myCircadianClock	Phillips et al., 2021 (59)	Observational and Randomized Control Trial	Time-Stamped photos on app, annotations	Average Eating Duration and NOVA food classification	5-wk baseline/observation; 6-month intervention (all days logged)	213 adults; 54 adults ≥14 h eating duration and one aspect of metabolic syndrome	12-h led to weight loss in the TRE group, but not control or between groups.
myCircadianClock	Prasad et al., 2021 (78)	Observational and Single-Arm Pilot Intervention	Time-Stamped photos on app, annotations	Eating window	2-wk baseline, 90-day intervention (all days logged)	50 (41F) adults (observation), 16 (TRE intervention) overweight or obese, and ≥14 h eating duration	10-h TRE led to weight loss, decreased waist circumference, and decreased blood pressure.
MealLogger	McHill et al, 2017 (90), 2019 (14)	Observational	Time-stamped photos on the app	Eating times, Calories, Macronutrients	7 consecutive days of 30-day study	110, (46F); 106 (45F) 18–22 yrs old (both	Later meal times (relative to melatonin onset) were associated with increased body fat.
Camera Phone, Movisens GmbH	Kosmadopoulos et al., 2020 (89)	Observational	Time-stamped photos from phone, meal annotation, supplement log	Eating times, Calories, Macronutrients	3 of 28–35 days of study logged	31 (6F) police officers	Night shift work delays the timing of food and decreases calorie intake, but did not alter macronutrients.
EasyDietDiary, Camera Phone	Parr et al., 2020 (38)	Single-Arm Pilot Intervention	Text food entries. Food/beverages photos using a phone camera.	Eating times, Calories, Macronutrients	4-wks (all days logged)	19 (10F) adults with T2DM and ≥12 h eating duration	TRE was feasible and did not alter dietary intake, psychological wellbeing, or cognitive function.
Zero	McAllister et al., 2020 (71)	Single-Arm Intervention	Zero smartphone app to document first and last food intake.	First and last Meal; Calories	6-wks (all days logged)	16 male resistance-trained firefighters	TRE decreased advanced oxidation protein products and advance glycated end products in blood.

How it works: the mCC app User Interface

The mCC app is designed to capture the temporal aspects of core behaviors that influence the circadian system. Participants are asked to log ingestion events (food/bev/water/medicine), sleep, and exercise. There is also an option to log health measures including vitals and common blood tests (Fig. 3). The app also has features for guiding participants on how to implement TRE and troubleshoot general or personal barriers to such a behavior change.

Fig. 3.

Data captured on the myCircadianClock smartphone app.

Participants are asked to log their current lifestyle for at least 14 days to capture a baseline assessment. During the baseline period, participants are blinded to their entries once logged. After 14 days, participants can see charts of everything that they logged in the history section, including their current eating window. They are also then able to set goals for a time-restricted eating window, step counts, and sleep times. The mCC app does not automatically recommend an eating interval, rather the research team and the participant decide this interval.

Logging food, beverages, water, and medicine.

There are two ways to log an ingestion event. The first is to take a quick photo and then provide the name of the item and select save (Fig. 4A). The second way to log an item is by annotation only. In this case, the name of the item is logged the same way, but there is no photo and the time is manually entered for when the item was consumed (Fig. 4B). An annotation-only entry can be done for food logged at the time they are consumed or to log items retroactively.

Fig. 4. The myCircadianClock app user interface.

A Home screen, live camera to quickly take a photo of what will be logged. B Annotation page (logging the name of items). C Exercise logging home screen. D Sleep logging home screen. E Health logging home screen. F Intake history page for a given week. G Activity and Sleep history page.

Logging exercise and capturing activity.

Exercise can also be logged using a timer or logged retrospectively. Participants will enter the time, name, and intensity of the exercise (Fig. 4C). Exercise is displayed on the home screen daily tracker as a red bar. Step counts are captured from Google Fit or Apple Health Kit (if participants give permissions). If a participant sets a daily step count goal, their daily progress step counts appear on the home screen in real-time (Fig. 4A). The mCC app also captures activity counts from the phone directly, which is displayed on the history page.

Logging sleep.

Participants can log sleep in two ways: [1] starting a timer when they go to sleep and then ending the timer when they wake up or [2] logging their sleep time retroactively. In both cases, once the sleep interval is entered, they are asked to rate if they feel rested (Fig. 4D). If they select that they don’t feel rested, then they can write why or they can select one of 3 common responses including “insufficient sleep,” “woke up once or more”, and “difficulty falling asleep”. The time of sleep onset and wake is displayed on the home screen daily tracker as a small blue moon icon (Fig. 4A).

Logging health measures.

Health measures such as vitals and lab work (common blood tests) can be entered on the health page (Fig. 4E). Participants select the name of the health measures from a list, then provide the date, and if they were fasting when the measurement was taken. Entries can be viewed on the history page.

History pages.

Participants can review everything that they have logged on the history pages. There are 3 different tabs: Intake, Activity, and Health (Figs. 4F, 4G). On the intake page (chart view), participants can see all of the food, beverage, water, and medication entries for the past weeks, with their actual eating window and their goal eating window displayed at the bottom (Fig. 4F).

Data security.

The mCC app is HIPAA compliant and keeps all participant data secured. All data from the app is double-encrypted and stored in AWS cloud servers (Fig. 5). Each coordinator portal and backend database is also associated with a HIPAA-compliant email account (Fig. 5). Researchers can use the protected account to communicate with participants as appropriate based on the study protocol.

Fig. 5. The myCircadianClock digital platform.

Participants Data (orange arrows; food/beverages, water, medication, sleep, exercise, health measures, and survey responses) logged on the mCC app is sent to the cloud-based mCC server. Study-specific data can be accessed by the research team (navy) in real-time on the mCC dashboard. The research team can also send notifications, surveys, and educational material to participants on the mCC app via the mCC server. Participant data can be downloaded for data integration and analysis by the research team at any time.

VALIDATION OF THE MCC APP

Five methods have validated data from the mCC app. First, the number of ingestion events captured through the app is comparable or surpasses the number of meal records found in NHANES data that rely on 24 h recall [60]. In the first study using the app, the researchers found the number of ingestion events/day varied widely between participants, with 4.22 ± 0.1 (mean+SEM) for the bottom decile to 15.52 ± 0.34 for the top decile, of which 3.33 ± 0.07 and 10.55 ± 0.24, respectively, were caloric food/beverage events [32]. Second, the fraction of time participants forgot to log food/beverages that they consumed was calculated from the mCC data; a push notification was sent to their phones at a random time (during their reported wake times) 1–2 times per day that asked if they had consumed an item within the last 30 mins. Participants gave a yes or no response [32]. The response was timestamped and compared to their entries on the app. This was used to estimate a false negative rate (when food/beverage was consumed but not logged) of 10.34%. This validation method is limited in that it is only validating logging at random times throughout the day and would not ensure that all entries were captured. Such an objective measure of false negatives in a community setting without complicated methods is a unique strength of the app platform. The more widely used method of 24-h recall does not calculate the number of times or the amount of food and beverages a person might have failed to report, so it is impossible to compare the false negative rate from the mCC app with more widely used methods. Third, a conservative estimation of energy content in food images and food descriptions found the participants were consuming more than their calculated energy requirements [32], which implies the users are not under-reporting their energy intake. Fourth, real-time logging of energy intake with the app found longer eating windows than participants self-reported on questionnaires; highlighting the perils of questionnaire-based self-reports of eating patterns. In questionnaire or telephone interviews when participants were asked about their typical eating pattern, the vast majority of them responded with an eating window of 12 h or less. However, objective measures of eating window from their self-reported ingestion events through the mCC app were significantly longer, because the app captures day-to-day changes in eating pattern (including time of first or last food/beverage and late-night snacks/beverages). Participants logged late-night snacks and even food or beverages consumed when they had disrupted sleep. Only < 10% of participants had an eating window of ≤ 12 h. Fifth, data captured from the mCC app is consistent with known circadian behavioral changes during weekdays and weekends. It is well known that a large number of non-shift workers in the western world wake up late on the weekend [61]. The mCC app also found a significant delay in breakfast time with 40% of participants delaying breakfast by 1 h or longer and 25% by 2.18 h [32].

In summary, multiple lines of methods were used to validate both the timestamp of ingestion events and the total quantity of calories consumed from data collected in the mCC app. There is a general concern that the act of logging food may reduce food intake or observing the timestamp of food intake may change eating patterns. However, in a recently concluded RCT to evaluate the effect of TRE on weight loss among participants with obesity, subjects assigned to the control group did not change their eating pattern or calorie intake even though they logged their food and beverages for several weeks during the study and the control cohort lost significantly less weight than the TRE cohort [35]. This result further validates that the use of the mCC app itself does not significantly change ingestive behavior.

Scope for future validation and their limitations

One new form of validation is also ongoing: validating entries by comparing them with continuous glucose monitors (CGM) and actigraphy data. Data from CGMs and actigraphy are currently being used to assess missed entries on the mCC app (unpublished). Participants wore CGMs, actigraphy watches, and logged all energy consumption on the mCC app in parallel for 14 continuous days. Data had high temporal resolution with the CGM capturing glucose concentration in interstitial fluid every 15 mins and the actigraphy watches assessing activity and light exposure every 30 s. These two data streams are aligned with data from the mCC app to allow for the validation of app entries. Increases in glucose (as measured by CGM) that could not be accounted for by a food/beverage entry, or activity (ie. small glucose increases after waking or with activity) while the participant was awake, are deemed a missed entry. The number of daily missed entries can be used to validate the accuracy of logging caloric food and beverage consumption on the mCC app. This validation method is limited in that glucose levels from the CGM may increase due to physiological changes (cortisol peak, activity, etc) rather than energy consumption, which could result in a false positive. This is partially corrected by the addition of actigraphy data, which may explain activity-induced changes in interstitial glucose. A false negative is also possible as some low glycemic foods (ex. nuts) may not cause a notable spike in glucose. Thus, there may be energy intake that was not logged and not detectable on the CGM.

It should be noted that there is no current “gold-standard reference” method to assess the timing of energy intake. Methods, such as video surveillance and hand and jaw sensors have been used to passively assess ingestion events. However, these methods require participants to wear monitors on their jaw and wrist or wear a live camera and have only been tested in controlled environments (lab or apartment) [62, 63]. Validating food entries with doubly labeled water or BEE can only estimate the number of calories logged, but do not take into account when they were consumed. The mCC app timestamps all photos of entries which allows for an unbiased assessment of temporal energy intake.

CLINICAL TRIALS ASSESSING TEMPORAL EATING PATTERN

To date, there have been at least 39 clinical trials assessing the effects of TRE in humans. They are mostly a mix of pilot studies with single-arm, randomized cross-over trials, and randomized control trials design [24, 29, 30, 32–46, 59, 64–84]. The number of participants ranges from 8–116 and the eating/fasting time varies from 4–12 h eating window and 20–12 h fasting. Some of them have a specific time of day that the eating window must occur whereas others allow participants to self-determine their eating window. Although some were able to monitor all food intake by providing food in a lab setting [29], most studies did not monitor temporal eating patterns before the intervention or during the study. These studies asked patients to follow a specific eating window and asked participants whether they complied or not. This is akin to running a caloric restriction intervention study in which the habitual energy intake of participants is not monitored at baseline nor during the intervention. Only 8 of the studies tracked temporal eating patterns in free-living participants throughout the study, 1 used EasyDietDiary and photos [38], 2 used written food diaries [34, 74], and 5 used the mCC app [30, 32, 35, 59, 78].

A recent RCT with 116 participants used a texting-based app that asked participants if they had adhered to their eating window, but did not have them track food [70]. This study was conducted remotely (with half of the participants enrolling online and never directly interacting with the research team) and had a low response rate to daily questions about TRE adherence. This study found a modest weight loss in the TRE group, that was not significantly different from that achieved in the control arm. Given no data was presented about the habitual eating pattern, and objective measurement of eating pattern and there was a lack of training on how to adopt a behavior intervention, it is not surprising that the study did not find a significant effect of TRE on most of the health outcomes relative to the control group. The finding from this study also highlights the need for objective measurement of eating patterns both before and during TRE intervention, and behavioral guidance to adhere to TRE intervention.

The myCircadianClock app to assess temporal eating patterns

The utility of a research tool to measure temporal ingestive behavior is also reflected in the studies in which it was used and the study outcomes (Table 1).

The first use of the app was in 2015 which found that the median daily eating interval of the 156 adults in the study was 14.75 h. A 10-h TRE intervention in 8 of those participants led to weight loss, improved sleep, and higher energy levels [32]. Since then, the mCC was updated to improve the app quality, esthetics, and function including the health tracking features. More recent studies have built off of the original study to understand the health effects of TRE. So far three studies have been published that assess 10-h TRE in adults with metabolic syndrome [30] and 8-h [35] and 10-h TRE [78] in adults with overweight.

The mCC app was next used by Wilkinson and Manoogian et al., 2020 [30] to assess the health impact of 10-h TRE for 12-weeks in individuals with metabolic syndrome in a single-arm study. 19 adults (13 M, 6 F) completed the study. Participants used the app with 95.3% and 85.6% logging adherence (logged at least 2 entries a minimum of 5 h apart in a given day) during baseline and intervention respectively. During baseline, participants had an eating window of 15.13 h. During the intervention, the mean eating window was decreased by 4.35 h (28.75%), resulting in a 10.78 h eating interval. Food/beverage items more than 1 h outside of the selected TRE interval were small (7% of days logged). Significant improvements were seen in body weight, waist circumference, blood pressure, LDL cholesterol, non-HDL cholesterol, HbA1c, and restfulness following sleep (Table 1). There were no changes in physical activity. This was especially exciting given that 16 of the 19 participants were already on medication to treat at least one aspect of metabolic syndrome. Statin and antihypertensive drugs were the most common with 79% and 63% (respectively) of participants using the drugs.

The first RCT using the mCC app was Chow et al. 2020, who examined the health impact of 8-h TRE in participants with overweight or obesity [35]. Participants had an average baseline eating interval of 15.20 h during baseline. During the 12-week intervention, the non-TRE group had a 15.10 h eating interval and the TRE group had a 9.90 h eating interval. Although the average eating window was greater than the 8-h TRE goal, participants were able to decrease their eating window by over 5 h. Compared to non-TRE, TRE decreased the number of eating events, weight, lean mass, and visceral fat (P ≤ 0.05). Compared to baseline, the TRE group showed a significant reduction in the number of daily eating occasions (−21.9%), and reduced weight, fat mass, lean mass, and visceral fat (Table 1). Participants also had decreased number of eating events and caffeine intake [85], no adverse effects in bone turnover [86], and improved quality of life [87]. As in Wilkinson and Manoogian et al., 2020, there were no changes in physical activity [30].

The mCC app was also used to assess the eating pattern of 213 participants in Switzerland [59]. Participants had an average eating window of 15.48 h. Of the 213 participants, 54 were randomized to either a 12-h TRE intervention or Standard Dietary Advice (SDA). Small but significant decreases in weight, BMI, and waist circumference were seen in the TRE group (12.50 h eating window), but there was no significant change in weight in the SDA group or between groups [59].

Prasad et al., 2021 used the mCC app to observe eating patterns of 50 adults in New York City and implemented a 10-h TRE intervention in 16 participants who had overweight or obesity and an eating window of ≥ 14 h [78]. The 50 participants in the observation phase had a 14.53 h eating window, consistent with previous studies in San Diego, CA (15.13 h and 14.75 h) [30, 32], Minneapolis, MN (15.20 h) [35], and Lausanne, Switzerland (15.48 h) [59]. The 16 participants in the TRE intervention reduced their eating window to 11.90 h and had significant decreases in body weight, waist circumference, and systolic blood pressure [78] (Table 1).

All five of the studies found that the eating window of free-living adults is around 15 h. They also demonstrated that TRE is feasible and has the potential to be a powerful intervention to both treat and prevent cardiometabolic disorders. Larger randomized control trials are currently underway to better understand the effects of TRE in various populations around the world. There are currently six ongoing clinical trials that are using the myCircadianClock app. These studies are investigating the health impacts of TRE in firefighters that work 24-h shifts (NCT03533023, San Diego, USA), adults with metabolic syndrome (NCT04057339, San Diego, USA; NCT04328233, Toruń, Poland), adults with overweight and obesity, and/or prediabetic (NCT03590158, Adelaide, Australia; NCT03956290, New York, USA, NCT04259632, Minnesota, USA).

Other methods of assessing temporal eating patterns

Clinical trials that have studied temporal aspects of eating patterns have used a variety of tools aside from the mCC app (Table 1). Some have been in-lab studies that controlled food timing by providing food and watching the participants eat [29]. Studies outside of the lab have used food journals (written or online), surveys, cameras, and smartphones. Studies that have used smartphones frequently used the camera function to capture images of food with a timestamp. Sometimes having participants capture photos at the beginning and end of an eating event to correctly adjust for food that was on the plate but not eaten. Others used third-party apps to either log or annotate photo entries.

Gupta et al., 2017 [88] used the camera feature on smartphones to capture all caloric food/beverage items that were consumed for 21 days, in 94 healthy participants. This was meant as a feasibility study to assess eating patterns in India. As in the Gill and Panda, 2015 study, they found that over 50% of participants ate for over 15 h per day.

Kosmadopoulos et al., 2020 [89] provided Samsung phones to participants to assess the eating times of police officers working morning, late, or night shifts over 28–35 days. Participants were asked to take photos of all food and beverages before an eating event, and another photo if they did not finish what was photographed. Participants then classified the entry as “breakfast,” “lunch,” “dinner,” “snack,” or “beverage only” using the movi-sensXS smartphone application (Movisens GmbH, Karlsruhe, Germany). Participants were also asked to complete a chart describing the quantity and type of food supplements they consumed such as vitamins. Registered dieticians analyzed the entries for macronutrient compositions and calorie content. Using this method, they were able to determine how macronutrient intake varied based on shift schedule and time of day. They also found that the daily eating window as calculated as the time lapse between breakfast and the last meal was longer during night shifts (13.9 h ± 3.1 h) compared to rest days (11.3 h ± 1.8 h, mean ± SD) [89]. However, this approach to calculating the average eating window does not take into account the day-to-day shift in the eating window (if any), which is better captured by calculating the 95% eating interval from several days of data (Fig. 4).

The third-party app, MealLogger (Wellness Foundry, New York, NY, USA), is a photo food journal app that has been used in multiple research studies to assess temporal eating patterns [14, 90]. Similar to mCC, it has users take a photo (timestamped) and name the items in the photo. McHill et al., 2017, and 2019 used Meallogger to assess the eating pattern of young adults over 7 days [14, 90]. Participants were also asked to have a standard item in all photos for reference. The eating duration was not calculated. Later meal timing relative to melatonin onset was associated with increased body fat [90]. Eating events were reported across the day (at all circadian phases relative to melatonin onset), with a peak around 7 pm. Individuals with overweight had 8% more of their daily calories in the evening [14].

Parr et al., 2020 gave 19 participants with T2DM the choice of paper food journals or a smartphone app EasyDietDiary (Xyris Software, Brisbane, Australia), a third-party smartphone app to document the timing of written food entries. They also had participants use their phone’s camera to photograph all energy intake [38]. In this feasibility study, they reported having 74 [24] % compliance to a 9-h TRE intervention with no significant change in dietary intake between participants in the TRE group and habitual eaters (Table 1).

McAllister et al., 2020 implemented a 10-h TRE intervention in firefighters using the third-party smartphone app Zero (Big Sky Health, Inc., San Francisco, CA) for participants to enter the time for first and last dietary intake each day [71]. Participants also had the option to use a paper form instead of Zero. MyFitnessPal, or a written food log, was also used to document dietary intake at the beginning and end of the intervention. At the end of a 6-week intervention, the 16 male resistance-trained firefighters significantly decreased advanced oxidation protein products and advance glycated end products (Table 1).

All of these apps are based on the thematic basis of the mCC app in that it used time-stamped photo entries. However, the apps do not allow for personalized communication with participants, provide a visual representation of the eating window or entries for the participants, nor do they capture sleep, exercise, or health information. Importantly, as these are both third-party apps, they are not customizable to individual studies.

STRENGTHS AND LIMITATIONS OF THE MCC APP

Strengths

The mCC app has multiple strengths as a research tool to capture temporal eating patterns. First, the mCC app provides the ability to track temporal lifestyle patterns (including eating patterns) in real-time, which has otherwise been lacking. This allows for both the user and the research team to monitor eating patterns in real-time rather than retrospective analysis. Second, the logging burden is also decreased as caloric information and portion size are not logged. Third, user engagement can be customized and automated with scheduled push notifications and surveys as well as personalized messages. Fourth, the app is scalable, which allows for data capture from individuals anywhere in the world on iPhone or Androids. Finally, because the mCC app was created and is maintained by a research team, researchers have full access to all participant data, and data is never shared with commercial entities.

Limitations

Although smartphone apps are powerful tools, they still have limitations. The main limitation to determining temporal lifestyle is that the app is dependent on participants entering everything they consume, and when they sleep and exercise. The necessity of manual logging innately leaves room for inaccuracy most likely caused by a lack of entries. Because the mCC app does not require calorie or portion size estimates when logging, it is not able to accurately assess the timing of calorie or macronutrient intake.

CONCLUSION

Capturing nutrient intake in free-living humans is an ongoing challenge. However modern technologies, specifically smartphones, are enabling new ways to capture nutrient intake. Smartphone apps such as myCircadianClock may be helpful research tools to assess temporal eating patterns, exercise, sleep, and health information for many consecutive days. The information gained from these tools will provide much-needed insight into nutrition and health. Tracking temporal components of nutrient intake can also serve as a helpful insight into patient care.