Accurate carbohydrate (CHO) counting to achieve satisfactory glucose control in type 1 diabetes (T1D) remains challenging in practice, and thus novel approaches are still needed (1,2). GoCARB is a computer vision-based application installed in a smartphone device that provides users with CHO content estimations from photos taken of plated meals (3). We present the results of a pilot prospective randomized controlled crossover study (NCT02546063) evaluating the effects of GoCARB on postprandial and overall glucose control in individuals with T1D using sensor-augmented insulin pump (SAP) therapy.

One week of GoCARB use was compared with conventional methods to estimate meal CHO content in 20 adults with T1D using SAP therapy (mean age 35 ± 14 years, BMI 25.5 ± 3.8 kg/m2, HbA1c 7.5 ± 0.6% [58.7 ± 5.9 mmol/mol], duration of diabetes 17 ± 10 years, duration of SAP use 2.7 ± 1.7 years, 13 men and 7 women). Prerandomization, each participant underwent a 2-week familiarization with the training version of the GoCARB software (without automated CHO output).

A detailed description of the system has been presented previously (4). In brief, the application requires users to take two subsequent photos of a plated meal with a reference card placed next to it. The application then generates the CHO content based on volumetric assessment of food components on the plate, referenced against a nutritional database. During the GoCARB intervention period, the application provided automated CHO suggestion. Individual meal insulin bolusing was at the patient’s own discretion. A real-time continuous glucose monitoring (CGM) device was worn during both study periods. Analysis was per intention-to-treat.

CGM and insulin end points are shown in Table 1. GoCARB significantly reduced proportion of time spent hyperglycemic (>12 mmol/L, P = 0.039) and glycemic variability (measured as standard deviation of sensor glucose, P = 0.007), with similar time spent hypoglycemic (P = 0.58). Total daily insulin dose and number of insulin boluses per day were comparable (P = 0.14 and P = 0.12, respectively).

Table 1

CGM and insulin end points

GoCARB group (n = 20)Control group (n = 20)P value
% Time hyperglycemic (>12 mmol/L) 15.0 ± 2.0 18.2 ± 2.1 0.039 
% Time hypoglycemic (<3.5 mmol/L) 2.3 ± 0.8 2.6 ± 0.7 0.58 
% Time in target (3.9–10 mmol/L) 65.9 ± 2.7 63.2 ± 2.8 0.19 
180-min postprandial iAUC (mmol/L/min) 205.9 ± 29.3 269.9 ± 39.8 0.13 
Mean glucose (mmol/L) 8.7 ± 0.3 8.9 ± 0.3 0.15 
Glucose standard deviation (mmol/L) 3.0 ± 0.1 3.2 ± 0.2 0.007 
Daily bolus insulin (units/24 h) 27.5 ± 2.3 30.0 ± 2.3 0.11 
Number of boluses (n/24 h) 6.8 ± 0.4 7.3 ± 0.5 0.12 
Total daily insulin (units/24 h) 47.5 ± 3.2 50.0 ± 3.2 0.14 
GoCARB group (n = 20)Control group (n = 20)P value
% Time hyperglycemic (>12 mmol/L) 15.0 ± 2.0 18.2 ± 2.1 0.039 
% Time hypoglycemic (<3.5 mmol/L) 2.3 ± 0.8 2.6 ± 0.7 0.58 
% Time in target (3.9–10 mmol/L) 65.9 ± 2.7 63.2 ± 2.8 0.19 
180-min postprandial iAUC (mmol/L/min) 205.9 ± 29.3 269.9 ± 39.8 0.13 
Mean glucose (mmol/L) 8.7 ± 0.3 8.9 ± 0.3 0.15 
Glucose standard deviation (mmol/L) 3.0 ± 0.1 3.2 ± 0.2 0.007 
Daily bolus insulin (units/24 h) 27.5 ± 2.3 30.0 ± 2.3 0.11 
Number of boluses (n/24 h) 6.8 ± 0.4 7.3 ± 0.5 0.12 
Total daily insulin (units/24 h) 47.5 ± 3.2 50.0 ± 3.2 0.14 

Data presented as mean ± SEM. iAUC, incremental area under the sensor glucose curve.

We showed for the first time in a randomized crossover design study the feasibility of a novel automated CHO-estimation approach, with improvements in both proportion of time spent hyperglycemic and glucose variability as measured by sensor glucose. Automation of CHO estimation may, therefore, improve glucose control and help support daily T1D self-management. Of note, postprandial glucose excursions did not significantly differ in the current study. This may be partly due to the relatively short duration and limited sample size of this pilot study. Also, the postprandial period defined a priori in this study may not have accounted for the delayed glucose excursions caused by other macronutrients, such fat and protein (5). Post hoc analysis revealed that use of the application was most frequent at lunch and dinner. This may have been related to the restriction of GoCARB use to plated meals. The study consisted of T1D patients who were trained in CHO counting, thereby limiting its generalizability to the wider population. Longer and larger studies are still needed, inclusive of a broader T1D population, to validate our findings and assess the impact of GoCARB on the burden of self-management and quality of life.

Acknowledgments. The authors thank the participants for their time, effort, and commitment.

Funding. Support was provided by European Union Seventh Framework Programme (FP7-PEOPLE-2011-IAPP) under grant agreement no. 286408.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. L.B., J.D., M.A., M.L., D.R., P.D., S.M., and C.S. contributed to conception and design of the study. L.B., J.D., and T.Z. contributed to acquisition of data. L.B., J.D., and C.T.N. contributed to data analysis. L.B., J.D., M.A., S.M., and C.S. were involved in drafting the manuscript. L.B., J.D., C.T.N., M.A., M.L., D.R., G.R., T.Z., S.M., and C.S. were involved in revising the manuscript critically for important intellectual content. All authors approved the final version of the manuscript. C.S. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Clinical trial reg. no. NCT02546063, clinicaltrials.gov.

1.
Mehta
SN
,
Quinn
N
,
Volkening
LK
,
Laffel
LM
.
Impact of carbohydrate counting on glycemic control in children with type 1 diabetes
.
Diabetes Care
2009
;
32
:
1014
1016
2.
Brazeau
AS
,
Mircescu
H
,
Desjardins
K
, et al
.
Carbohydrate counting accuracy and blood glucose variability in adults with type 1 diabetes
.
Diabetes Res Clin Pract
2013
;
99
:
19
23
3.
Rhyner
D
,
Loher
H
,
Dehais
J
, et al
.
Carbohydrate estimation by a mobile phone-based system versus self-estimations of individuals with type 1 diabetes mellitus: a comparative study
.
J Med Internet Res
2016
;
18
:
e101
4.
Anthimopoulos
M
,
Dehais
J
,
Shevchik
S
, et al
.
Computer vision-based carbohydrate estimation for type 1 patients with diabetes using smartphones
.
J Diabetes Sci Technol
2015
;
9
:
507
515
5.
Bell
KJ
,
Smart
CE
,
Steil
GM
,
Brand-Miller
JC
,
King
B
,
Wolpert
HA
.
Impact of fat, protein, and glycemic index on postprandial glucose control in type 1 diabetes: implications for intensive diabetes management in the continuous glucose monitoring era
.
Diabetes Care
2015
;
38
:
1008
1015
Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at http://www.diabetesjournals.org/content/license.