1. Introduction
The accurate and detailed dietary assessment in human studies using traditional dietary assessment methods (such as handwritten food records) is labor-intensive and time-consuming both for investigators and study participants. Analysis of dietary compliance—including energy and macro/micronutrient intakes, as well as meal timing—in dietary intervention trials represents an additional challenge. Furthermore, conducting chrononutritional trials investigating the effects of specific meal timings requires the assessment of timely compliance. Recent technological developments—in particular, the wide use of smartphones, tablets, and nutrition applications (apps)—can overcome some of these problems.
Smartphone apps to promote healthier lifestyles and improve health—e.g., dietary assessment apps—have developed drastically. Smartphone technology for dietary assessments demonstrating high acceptability and convenience [
1,
2,
3] is widely used by private individuals, and is of great interest for nutritional and metabolic research. Indeed, in high-income countries, smartphone ownership rates are high. The number of smartphone users in Germany has only grown in recent years, amounting to 66.15 million smartphone owners in 2021 [
4]. Statistical analysis estimated that smartphone ownership was 94.2% in the age group 14–19 years, 95.5% in the age group 20–29 years, 96% in the age group 30–39 years, and 68.2% in the age group 70+ years [
5]. According to the data of the German Obesity Society (DAG), 67% of male and 53% of female Germans are overweight, and one-quarter of these people are obese. Therefore, the potential for using smartphone apps to improve the quality of dietary assessments and support weight management needs to be investigated. An app-based food diary may be more convenient and easier to use than a paper-based diary. In particular, it allows user to complete records at the instant of food intake (e.g., by a scanning a food item’s barcode, or by taking a picture). In this way, errors that may occur due to delayed recording, which is often the case when using a paper-based diary, can be avoided. Smartphone apps also enable rapid transfer of digital data from research participant to researcher. Furthermore, digital solutions allow real-time communication for monitoring the participants’ progress, resulting in an improvement in data quality and dietary compliance in nutritional research.
Further, app-based food tracking might be particularly useful in studies investigating the effects of specific meal timings (often combined with a specific food composition) on study outcomes—so-called chrononutritional studies. In particular, time-restricted eating (TRE)—an approach requiring a shortening of the daily eating window and an adherence to certain meal timings—is of great interest in the scientific community. TRE has been shown to be effective for the prevention and treatment of obesity, diabetes, and other metabolic and non-metabolic disturbances [
6,
7]. Currently, over 200 TRE trials are ongoing according to the ClinicalTrials.gov database (
https://clinicaltrials.gov/ (accessed on 4 July 2022)). Detailed diet compliance assessment is highly important to establish a direct interrelation between trial outcomes and TRE. The potential of smartphone apps to record mealtimes has already been demonstrated by Gill and Panda [
8], who developed a smartphone app to record food-intake events in real-time to analyze food patterns. The results showed that more than half of the study cohort had recording durations of approximately 15 h. Furthermore, the restriction to 10–11 h food intake was associated with weight loss [
8]. However, apart from the uploading of food pictures and notes by the study participants, there was no integrated database for the detection of food composition [
8]. It is becoming clear that real-time recording of food events, combined with accurate recording of food components, could provide a benefit under study conditions. In this regard, dietary apps using a food database appear to be an effective tool to simultaneously track meal timing and food composition in TRE trials. However, despite all of the abovementioned benefits, there are also some limitations when using dietary apps. One important limitation is the questionable accuracy of app-based dietary assessment tools in providing accurate information on the energy and macro/micronutrient intakes, which needs to be investigated further [
3,
9].
The FDDB Extender app (FDDB Internetportale GmbH, Bremen, Germany) is a smartphone dietary assessment app that is widely used by private users trying to reduce or maintain their body weight. By recording all consumed food items using barcodes or manual search in the database, along with their amounts and the timing of meals, the app provides users with information on their energy as well as macronutrient and micronutrient intakes based on the associated FDDB database. The app indicates daily required calories, which are calculated constantly based on the personal body weight goals. Additionally, the FDDB Extender app allows registration of exercise and calories burned while exercising, which are automatically included in the total summary of energy requirements. This aims at leading users to consciously follow and adapt their eating pattern to achieve their goals in a pleasant and motivating way. In addition to timely records of meal-intake events, the FDDB app also categorizes the meal type (i.e., breakfast, snack, lunch, or dinner), making it a potentially helpful tool for chrononutritional studies. However, it is currently unknown whether the FDDB app is appropriate for use in research, i.e., whether it provides accurate information on the energy and macro/micronutrient intakes comparable with professional dietary assessment software. To the best of our knowledge, to date, only one paper published in 2017 has addressed this question for the FDDB app, comparing the energy contents of 13 selected foods in the FDDB database with the German Food Database (Bundeslebensmittelschlüssel, BLS) [
9]. However, no studies conducted on the FDDB validation thus far have been based on the analysis of long-term food records and addressed macro- and micronutrient intakes.
Therefore, the objective of this study was to evaluate the validity of the FDDB smartphone app and food database compared with PRODI® version 6.5 (Nutri-Science GmbH, Freiburg, Germany)—a professional software platform widely used in medical practice and nutritional research in Germany. The study was conducted in the context of a trial on intermittent fasting comparing the effects of early versus late daily timeframes of food restriction.
4. Discussion
Our results showed that the FDDB app and database enabled an automatic recording of the timing and duration of the daily eating window, providing an effective assessment of timely compliance in our chrononutritional trial, whereas the meal timing analysis using the PRODI® software required time-intensive manual processing of daily food records. Furthermore, FDDB showed a good agreement with the PRODI® software in energy and macronutrient intake data. On the other hand, the FDDB data on most micronutrients and saturated/unsaturated fat intake were inaccurate. In sum, the FDDB app and database seem to allow a more accurate, quicker, and less sophisticated analysis of food timing than conventional approaches, with energy intake and macronutrient composition seeming reliable, making it a potentially attractive tool for nutritional and especially chrononutritional research.
In the present work, we found a good agreement between these two methods for energy intake during the run-in phase and during the eTRE and lTRE interventions, and the total energy intake values obtained in both analyses showed a high correlation coefficient. The mean individual difference was good, with −31.5 kcal (LOA −277 kcal, 214 kcal) and only one outlier. This finding is in contrast to the data of Holzmann et al. [
9], who showed a large discrepancy of up to 30.6% in the energy contents of several food items between FDDB and BLS data. This discrepancy might be explained by the fact that only 13 food items were assessed in [
7], whereas our research investigated long-term dietary records (for a total of 42 days) in terms of the dietary intervention study.
Furthermore, our analysis showed a high correlation of macronutrient intakes for carbohydrates, protein, and fat, with only minor differences between the two tools, as a percentage of daily energy intake and in grams (2.3–7.3%). In agreement with this, the Bland–Altman graphs indicated that individual differences between the two methods were within acceptable ranges, in compliance with the 95% level of agreement, with a 1.96 standard deviation and few outliers. Thus, our data provided a clear statement that both dietary software platforms show almost identical data for macronutrients and energy levels when long-term nutritional diaries are analyzed. Thus, for energy content and macronutrients, FDDB provides data comparable to the established professional software platform PRODI
®, and could be used effectively in nutritional research. In particular, the FDDB app and database were used in a recently published trial to compare energy and macronutrient intakes during 8 weeks of TRE vs. continuous energy restriction intervention and 6 weeks of follow-up in 42 overweight subjects [
13]. Notably, in this study, most of the study participants stated that the documentation was not an additional stress factor in their everyday life [
13]. Two dietary intervention trials—one TRE trial (NCT04351672) [
10] and one plant-based nutrition trial (NCT03901183)—using the FDDB app to monitor energy and macronutrient intakes as well as meal timings, are currently ongoing.
Fiber, salt, and sugar exhibited high correlation coefficients and moderate differences of 10–14% between the two tools, showing that, for these nutrients, FDDB also provides data comparable to the professional software platform PRODI®. However, saturated fat and cholesterol intakes were strongly underreported with the FDDB database (−40.3% and −61.1%, respectively), showing low correlation coefficients. Moreover, alcohol intake showed 4.5-fold higher values with FDDB than with PRODI®. However, the subjects were asked not to drink alcohol for the whole study duration, so that the amounts of alcohol detected by FDDB (but not by PRODI®) would mainly come from the food (such as apple juice or similar), which might explain the high discrepancy. Furthermore, data on vitamins A, D, E, B1, B2, B6, and B12 were markedly different between the FDDB and PRODI® tools. Only vitamin C demonstrated a difference between the two tools below 10%, and showed the highest correlation coefficient for the software comparison (r = 0.697). A possible reason for is that the information on vitamin C content is often explicitly listed on products, giving FDDB app users the ability to enter detailed information about the contents of ascorbic acid present in industrialized products. In contrast, listing the contents of other vitamins in processed foods is not considered to be as relevant. The same applies to minerals. Indeed, for minerals (i.e., potassium, magnesium, calcium, iron, phosphorus, copper, zinc, chloride, iodide, manganese, sulfur), the intakes were mostly underreported in FDDB compared to PRODI® data. As mentioned previously, disclosure of micronutrient contents on the nutrient tables of processed foods is not mandatory. Therefore, it is likely that no registration of the correct amounts of minerals and vitamins is present on the FDDB extender smartphone application. A lot of the presented data in the FDDB database are collected by users or from other databases, making them broad but at the same time imprecise, as not all food products available on the market are labeled with complete nutritional information. In contrast, all food products registered on PRODI® have been analyzed with regard to all micronutrients, and present precise data on them. Taken together, our data suggest that the FDDB app does not present reliable data on saturated fat, cholesterol, vitamin, and mineral contents; thus, the application would not be suitable for dietary assessment in trials focused on the influence of these nutrients.
In recent last years, a range of studies have evaluated the feasibility and accuracy of smartphone apps for dietary assessment, demonstrating significant interest of the scientific community in this topic. One of the first studies in this field, conducted in the UK, evaluated a smartphone app named My Meal Mate (MMM), initially developed to facilitate weight loss, by comparing a digital 7-day food diary to 24-hour dietary recalls [
1]. The study showed a good agreement in energy intake between the methods (a difference of 218 kJ/d, with LOAs ranging from −2434 kJ to 2022 kJ), and found the app to be useful for the assessment of group means of carbohydrate, fat, and protein intakes. Two other studies comparing smartphone apps for dietary assessment in epidemiological research with 24-hour dietary recalls also showed reliable results on the energy and macronutrient intakes at the group level [
14,
15]. Similarly, two clinical trials that compared mean macronutrient data between professional software platforms for the evaluation of dietary records and electronic diet recording showed no significant deviations of the macronutrients, and high correlations between both methods [
16,
17]. Another study compared a tablet app for assessing dietary intakes with the measured food intake/food waste method in military personnel, and provided satisfactory results in the assessment of energy, macronutrient, and selected micronutrient intakes, whereas some micronutrients showed large deviations [
18]. Notably, a smartphone image-based dietary assessment app tested among Canadian adults showed large deviations from 3-day food diaries in a range of nutrients, including total energy, protein, carbohydrates, fat, saturated fatty acids, and iron [
19], suggesting that the accuracy of smartphone app data has to be validated before utilization in dietary trials.
Common advantages and limitations in using dietary apps for research purposes in general, and the FDDB app in particular, must be discussed. In nutritional and chrononutritional trials, the accurate analysis of caloric intake, food composition, and timing plays an essential role in the assessment of dietary compliance and of the physiological and metabolic changes in response to the modification of nutritional patterns. To this end, using digital food records in food-tracking apps might be more advantageous compared to traditional methods, such as paper-based diaries or 24-hour recalls. Indeed, the usage of smartphone apps such as the FDDB app helps to avoid a time-intensive transfer of handwritten protocol data into food database software by a dietician, and allows records to be completed directly during the meal. Obstacles such as poor readability of handwritten food diaries or adding food to the diary afterwards could be avoided through documentation via smartphone apps.
Furthermore, the number of food items in the database must be considered to decide on the method of dietary analysis in the clinical trial. Notably, the FDDB database contains about 531,350 food items (as of 19 July 2022), with a daily increase of hundreds of new foods, and is therefore more extensive than the PRODI® database. This makes food searches easier for both the recording by study participants and the further analysis by researchers. Another advantage of using the FDDB Extender app is the opportunity to create recipes for subjects to cook at home, and of a simple down-calculation to one portion with proportional listing of the individual food items.
Even though FDDB can be used as an application for smartphones, it is also possible to log in to user accounts via computer, which makes it easy for the study staff to monitor and evaluate the nutritional data. By visiting the participant’s account, the dietician can check the amount of food intake, its composition, and its timing, and in this way monitor dietary compliance almost in “real time”. A monitoring system that runs parallel to the study helps to avoid protocol errors. If needed, participants can be contacted immediately to receive support, thereby maintaining the compliance. Notably, analysis of dietary records using the PRODI® software is not possible in “real time”, but only after the collection of records, excluding the possibility of rapid feedback and food behavior correction of study participants. Moreover, the meal timing analysis in the PRODI® software can be performed only via the time-intensive manual processing of daily food records. Furthermore, using the FDDB smartphone app also allows study staff to monitor possible weight fluctuations and rapidly detect possible causes, e.g., deviations in eating patterns or usual calorie intake. This is essential for both isocaloric and weight loss trials. Therefore, our findings show that the FDDB app can potentially be used for such specific aims as research monitoring and, if needed, for the quick correction of timely compliance, caloric intake, and macronutrient intake.
Using dietary apps for research purposes also has several limitations. In particular, some smartphone-app-based dietary assessment tools are based on photographic images of meals, where the manual interpretation and decoding of images remain labor-intensive, and might provide an imprecise information on the caloric intake and nutrient composition [
3]. For this reason, most of the currently available dietary assessment apps are similar to conventional tools, i.e., food diaries and 24-hour recalls, representing a form of self-reporting. A well-known limitation in all self-reported dietary assessments is conscious or unconscious under-reporting, which can result in serious observation and interpretation bias [
20]. Nevertheless, using smartphone apps directly during the meal might markedly reduce the under-reporting and associated misinterpretation [
1,
21]. Moreover, the usage of apps—especially those that show an individual overview of optimal caloric and macronutrient intake—may lead to changes in eating behavior and weight loss even at baseline, resulting in data bias. This effect has already been discussed by Gill and Panda [
8], who used the app “myCircadianClock” (mCC) with photo documentation to record eating times and patterns. To avoid changes in eating behavior, uploaded food images and their timestamps were immediately deleted from the participants’ smartphones after being transmitted to a server [
8]. Wilkinson et al. [
22] highlighted the possible problem of changes in behavioral patterns, and referred to the difficulty of tracking consistent use of the mCC app in a chrononutritional study.
Finally, using a smartphone app for food recording could be challenging, especially for the older generation, while younger generations might benefit from it. For this reason, using an app correctly could be more difficult to learn than handwritten logging, as already reported in a previous randomized controlled trial [
23]. Therefore, alternatives—e.g., handwritten documentation—should be provided. Furthermore, detailed instruction in the use of the app by a dietician must be given to each participant. In this respect, the development of instructions in the form of an online presentation might be a promising tool to train subjects in advance, which could also be effectively used under the present conditions of the COVID-19 pandemic.
We also have to mention a possible limitation of our research concerning the methodology of the meal recording. Indeed, six subjects used a paper-based protocol that was transferred in parallel into FDDB and into PRODI® for subsequent analysis. In contrast, four subjects entered their diets directly into the FDDB app, which means that, for these individuals, meal records were first exported from FDDB in the form of an Excel table, and then transferred to PRODI® for analysis. Notably, the main aim of our study was to compare the results of the dietary record analysis using the FDDB vs. PRODI® databases; therefore, inconsistency in the meal recording would not have a strong effect on these data. The flexibility in methods of dietary recording allowed us to achieve higher compliance with the study protocols in our trial, because older subjects mostly preferred handwritten protocols. Nevertheless, as mentioned above, the use of digital protocols via app can affect the quality of food recording. Therefore, in future studies, it could be feasible to collect a paper-based diary simultaneously with a digital FDDB diary in the same subjects, and to verify in this way whether the different methods of recording meals affect the results.
Finally, due to the relatively small number of subjects and the high number of tests in this study, we must acknowledge that chance findings cannot be ruled out. From this reason, and to confirm the generalizability of our results, larger studies and replication studies in other populations are needed.