1. Introduction

There are several components of diabetes management including nutrition therapy, insulin therapy, and other glucose-lowering medications. Although dietary interventions may not eliminate or reverse the need for medication and insulin, this lifestyle modification could reduce the need for these therapies in certain scenarios for patients with diabetes (PWD) (including T1D and T2D) ^[1][2]. One important aspect of dietary interventions that aim to prevent T2D incidence or control complications and delay progression in PWD, has focused on carefully monitoring or restricting carbohydrate intake to improve glycemic control ^[1]. When combined with supervision from healthcare professionals (registered dietitian or Certified Diabetes Care and Education Specialist), carbohydrate counting has been used as the gold standard for determining insulin dosing for PWD to maintain glycemic control ^[3][4]. Yet, carbohydrate counting is challenging; miss-estimation is common ^[5][6][7], and unfortunately, most PWD in the U.S. do not have access to multi-disciplinary care. Other challenges to carbohydrate counting include the time burden required for recording and calculating food intake and maintaining a stable weight within the dietary flexibility that carbohydrate counting allows ^[8]. Optimal dosing of insulin based on carbohydrate counting may also be difficult to achieve. 

There are several components of diabetes management including nutrition therapy, insulin therapy, and other glucose-lowering medications. Although dietary interventions may not eliminate or reverse the need for medication and insulin, this lifestyle modification could reduce the need for these therapies in certain scenarios for patients with diabetes (PWD) (including T1D and T2D) [3,4]. One important aspect of dietary interventions that aim to prevent T2D incidence or control complications and delay progression in PWD, has focused on carefully monitoring or restricting carbohydrate intake to improve glycemic control [3]. When combined with supervision from healthcare professionals (registered dietitian or Certified Diabetes Care and Education Specialist), carbohydrate counting has been used as the gold standard for determining insulin dosing for PWD to maintain glycemic control [8,9]. Yet, carbohydrate counting is challenging; miss-estimation is common [10,11,12], and unfortunately, most PWD in the U.S. do not have access to multi-disciplinary care. Other challenges to carbohydrate counting include the time burden required for recording and calculating food intake and maintaining a stable weight within the dietary flexibility that carbohydrate counting allows [7]. Optimal dosing of insulin based on carbohydrate counting may also be difficult to achieve. 

Evidence shows that mobile image-based dietary assessment tools (MIDATs) are more accurate than traditional dietary recall for collecting information on food intake and show a potential to improve disease outcomes ^[9][10]. Mobile-based interventions including assessing dietary intake and monitoring lifestyle factors and receiving feedback have led to decreased glycated hemoglobin (HbA1c) in PWD and better weight control for obesity prevention as a result of improved lifestyle factors compared to those not receiving an intervention ^[11]. Some examples of mobile technologies for assessing dietary or energy intake are camera-scan-sensor technologies, remote food photography methods, and real-time food recognition systems ^[10]. These existing MIDATs, built for research or for self-monitoring dietary intake, still have several limitations including their ability to accurately identify and quantify the macro/micronutrient content of foods and beverages from a simple picture as well as their ability to capture eating behaviors (eating time and frequency) and dietary intake variability ^[10]. Nevertheless, a MIDAT that captures detailed dietary information designed for diabetes management, specifically focusing on improving glycemic control for PWD and connecting to insulin delivery mechanisms, has the potential to achieve personalized, real-time insulin dosing based on users’ dietary behaviors. Currently, however, a gap exists in determining what information should be captured to more comprehensively and accurately link dietary behaviors to individualized health outcome data for monitoring dietary intake, real-time insulin dosing modifications and tailoring weight management strategies. The current review will discuss the relevant dietary factors linked to glycemic control that could be incorporated and recorded in a MIDAT, but currently are not, to make it a better tool for managing glycemic control among those on insulin therapy.

Evidence shows that mobile image-based dietary assessment tools (MIDATs) are more accurate than traditional dietary recall for collecting information on food intake and show a potential to improve disease outcomes [20,24]. Mobile-based interventions including assessing dietary intake and monitoring lifestyle factors and receiving feedback have led to decreased glycated hemoglobin (HbA1c) in PWD and better weight control for obesity prevention as a result of improved lifestyle factors compared to those not receiving an intervention [22]. Some examples of mobile technologies for assessing dietary or energy intake are camera-scan-sensor technologies, remote food photography methods, and real-time food recognition systems [24]. These existing MIDATs, built for research or for self-monitoring dietary intake, still have several limitations including their ability to accurately identify and quantify the macro/micronutrient content of foods and beverages from a simple picture as well as their ability to capture eating behaviors (eating time and frequency) and dietary intake variability [24]. Nevertheless, a MIDAT that captures detailed dietary information designed for diabetes management, specifically focusing on improving glycemic control for PWD and connecting to insulin delivery mechanisms, has the potential to achieve personalized, real-time insulin dosing based on users’ dietary behaviors. Currently, however, a gap exists in determining what information should be captured to more comprehensively and accurately link dietary behaviors to individualized health outcome data for monitoring dietary intake, real-time insulin dosing modifications and tailoring weight management strategies. The current review will discuss the relevant dietary factors linked to glycemic control that could be incorporated and recorded in a MIDAT, but currently are not, to make it a better tool for managing glycemic control among those on insulin therapy.

2. Information a MIDAT to Manage and Improve Diabetes should Capture

2.1. Evaluating the Role of Time of Eating Occasion on Biomarkers of Glycemic Control in Patients with Diabetes (PWD)

3.1. Evaluating the Role of Time of Eating Occasion on Biomarkers of Glycemic Control in Patients with Diabetes (PWD)

Time of eating is a potentially important factor to consider in developing an image-based dietary assessment tool designed for optimizing metabolic control. Specifically, breakfast consumption/skipping, late eating, and frequency of meals may be relevant factors to glycemic control. Evidence shows that breakfast skipping is related to increased risk of obesity, insulin insensitivity, and higher risk of T2D compared with eating breakfast ^[12][13][14]. Findings on late eating/snacking behaviors showed negative effects on glycemic control, which could be due to the impact of late eating on energy intake and postprandial (PP) blood glucose response ^[15]. Meal frequency might be associated with neurological and hormonal responses that affect blood insulin and glucose levels and infrequent meal consumption was associated with improved glycemic control ^[16][17]. Quantifying the timing of dietary intake holds promise to inform the development of a sensitive and targeted dietary analysis system for the management of diabetes and should be considered in a MIDAT. In addition, MIDAT capability to send mobile notifications to prompt and change eating behaviors in terms of time of eating could help improve glycemic control.

Time of eating is a potentially important factor to consider in developing an image-based dietary assessment tool designed for optimizing metabolic control. Specifically, breakfast consumption/skipping, late eating, and frequency of meals may be relevant factors to glycemic control. Evidence shows that breakfast skipping is related to increased risk of obesity, insulin insensitivity, and higher risk of T2D compared with eating breakfast [25,26,29]. Findings on late eating/snacking behaviors showed negative effects on glycemic control, which could be due to the impact of late eating on energy intake and PP blood glucose response [34]. Meal frequency might be associated with neurological and hormonal responses that affect blood insulin and glucose levels and infrequent meal consumption was associated with improved glycemic control [37,39]. Quantifying the timing of dietary intake holds promise to inform the development of a sensitive and targeted dietary analysis system for the management of diabetes and should be considered in a MIDAT. In addition, MIDAT capability to send mobile notifications to prompt and change eating behaviors in terms of time of eating could help improve glycemic control.

2.2. Determining How the Macronutrient Composition of Meals Effects Glycemic Control in PWD

3.2. Determining How the Macronutrient Composition of Meals Effects Glycemic Control in PWD

There is evidence that macronutrient components including fat, protein, and dietary fiber as well as other factors such as the glycemic index of a food and food order may impact glycemic control and thus should be considered and further investigated to determine optimal insulin dosing to improve diabetes management. Some but not all evidence suggests that dietary fat and protein could delay and increase glycemic response, therefore increasing insulin needs ^{[18][19][20][21]}. There are inconsistent findings regarding the effect of dietary fiber intake on glycemic control in patients with T1D ^{[22][23][24][25][26]}, while some evidence shows fiber might be beneficial for PP blood glucose parameters among those with T2D ^{[27][28][29][30]}. Reviewed studies show that carbohydrate-last food order may improve PP glucose profile ^{[4][31][32][33][34]}. Additionally, studies investigating the link between glycemic index and glycemic response provided evidence suggesting an improved glycemic response, specifically fasting blood glucose and HbA1c, associated with consumption of low glycemic index foods ^[35][36][37]. Nevertheless, evidence has shown significant interpersonal variations in glucose response by the various dietary components, indicating a need for individualized nutrition therapy for diabetes management. Understanding how the macronutrient breakdown of the meal and other factors including glycemic index and food order affect glycemic response is essential for the development of a MIDAT that can capture these components and link information to insulin dosing requirements to improve diabetes management and control.

There is evidence that macronutrient components including fat, protein, and dietary fiber as well as other factors such as the glycemic index of a food and food order may impact glycemic control and thus should be considered and further investigated to determine optimal insulin dosing to improve diabetes management. Some but not all evidence suggests that dietary fat and protein could delay and increase glycemic response, therefore increasing insulin needs [40,41,42,44]. There are inconsistent findings regarding the effect of dietary fiber intake on glycemic control in patients with T1D [48,59,60,61,62], while some evidence shows fiber might be beneficial for PP blood glucose parameters among those with T2D [47,51,63,64]. Reviewed studies show that carbohydrate-last food order may improve PP glucose profile [9,46,52,53,68]. Additionally, studies investigating the link between glycemic index and glycemic response provided evidence suggesting an improved glycemic response, specifically fasting blood glucose and HbA1c, associated with consumption of low glycemic index foods [49,50,67]. Nevertheless, evidence has shown significant interpersonal variations in glucose response by the various dietary components, indicating a need for individualized nutrition therapy for diabetes management. Understanding how the macronutrient breakdown of the meal and other factors including glycemic index and food order affect glycemic response is essential for the development of a MIDAT that can capture these components and link information to insulin dosing requirements to improve diabetes management and control.

Time of eating, macronutrient content and macronutrient order of meals are important aspects to be captured in a MIDAT to enable personalized, real-time diabetes management.