We used a quasi-experimental research design comprised of a single-arm pre-post intervention. Participants were not paid for their participation, but because the program was free, they took part in the program at no cost. The University of Michigan Institutional Review Board (IRB) ruled that analyses of these previously collected and de-identified data were not subject to IRB regulation.

We recruited participants to this trial in three phases. The first phase recruited a sample of convenience following the launch of the Low-Carb Program (November 14, 2015-November 14, 2016), whereby 105,950 adults with type 2 diabetes between the ages of 18 and 99 years signed up to participate in the program. Participants could live anywhere in the world. To have a broad applicability to a nonclinical trial setting, the only de facto exclusion criterion was the inability to understand English. Second, upon sign-up, the program prompted individuals to complete an initial baseline survey; 19,646 of 105,950 (18.54%) did so. Of those, 7809 people had complete baseline datasets including weight, a recent HbA_1c result (taken within 4 months), and medication use. Third, we used GraphPad Random Generator Software to randomly select a subset of 1000 participants to be followed for 12 months, thus enabling us to select participants for no other reason than that they were randomly selected by the software. Therefore, we did not include all the 7809 patients to follow over a year, but instead followed a random subsample of 1000 (see Figure 1).

The Low-Carb Program is a completely automated, structured 10-week health intervention for adults with type 2 diabetes. Participants are given access to nutrition-focused modules, with a new module available each week over the course of 10 weeks. The modules are designed to help participants gradually reduce their total carbohydrate intake to less than 130 grams per day to meet their self-selected goals. The program encourages participants to make behavior changes based on “action points” or behavior change goals at the end of each module. These goals are supported with resources that are available to download, including information sheets, recipes, and suggested food substitution ideas. The Low-Carb Program online platform also includes digital tools for submitting self-monitoring data on a number of different variables including blood glucose levels, blood pressure, mood, sleep, food intake, and body weight. Weekly automated feedback is provided to users based on their use of the program through email notifications, and participants are notified when the next week’s module has been opened. Lessons are taught through videos, written content, or podcasts of varying lengths (approximately 3 to 12 minutes long).

The first 2 weeks of the program contain an explanation of the physiology of type 2 diabetes and the role of diet, including a description of how a low-carbohydrate diet can help manage postprandial blood glucose levels and weight. The subsequent week’s modules explore strategies to reduce dietary sources of sugar, in particular, high-starch foods, such as bread, pasta, and rice. Participants are encouraged to make portion control and carbohydrate restriction decisions based on visual plate representations. In place of carbohydrate-rich foods, an increased intake of green vegetables, low-glycemic index fruits (eg, blueberries, strawberries, and raspberries) and fats (eg, from olive oil, butter, eggs, nuts, and full-fat dairy) are advocated. The program stresses the importance of regular contact with the participants’ health care providers for adjustments in medications in weeks 1, 2, and 10. After the 10 weeks of modules have been opened, participants continue to have access to the education content as well as the ability to continue to track their health (glycemic control, weight) and access support from the discussion board. See Table 1 for a list of the weekly topics.

Much of the content of the Low-Carb Program is based on an in-person, nurse- and physician-led, low-carbohydrate training program conducted in a primary health care setting [17]. For example, the dietary recommendations reflect an understanding of the glycemic index, a relative ranking of carbohydrates in foods according to how they affect blood glucose levels. A meal of pure glucose (the index food) has a score of 100, boiled potatoes are scored at 96, cornflakes at 93, and brown bread at 74, all of which are higher than table sugar at 63 [18]. This kind of information helps participants understand that both sugary and starchy foods increase blood glucose, and it also explains why the UK’s National Institute for Health and Care Excellence advises physicians to “encourage high-fiber, low-glycemic index sources of carbohydrate in the diet” for type 2 diabetes [19]. Based on this, the program suggests a reduction in all sugary foods and replacing starchy foods, such as potato or rice, with green leafy vegetables, healthy fats, and some protein.

The content and strategies used in the program build off prior research and theory. For example, evidence suggests that goal setting can act as an effective behavior change strategy used to improve adherence to lifestyle intervention programs in obesity management programs [20]. Therefore, the program encourages participants to select a goal at the beginning of the program (eg, to lose weight, reduce medication dependency, or make healthier choices for their whole family). Participants are also prompted to consider how their health would benefit from attaining their goal. Throughout the program, participants are periodically prompted to consider how close they are to attaining their goal.

The program further reinforces behavior change through integrated tracking whereby program users are encouraged to track their health data including mood, food intake, blood glucose levels, weight, sleep, and HbA_1c. According to the Control Theory of behavior change, monitoring goal progress—that is, evaluating one’s ongoing performance relative to the standard—and responding accordingly is critical to goal attainment [21]. Recent findings suggest that program interventions that elevate the frequency of progress monitoring are likely to induce behavior change [22].

In addition, prior studies demonstrate that peer support may improve blood glucose control [23,24], peer-based support may be as effective for weight loss as coach-based support [25], and that online discussion boards can be supportive for weight loss [26]. Therefore, the program encourages social support by matching new participants of the program to a “buddy,” a previous graduate of the program, based on similar demographics including age, gender, and their self-selected goal. Participants are encouraged to interact with that buddy and peers on the program’s moderated online discussion board.

At baseline, an online survey asked participants to report on their type of diabetes, year of diagnosis, their most recent HbA_1c test result and date, current medications (medication name, dose, and regimen), age, gender, socioeconomic status (based on household income), and presence of comorbid chronic illnesses. At 12 months, participants were again asked to report on their current HbA_1c, weight, and medications.

Analyses were performed using the SPSS version 21.0 (SPSS Inc, Chicago, IL, USA). We examined the difference in characteristics from baseline to 12-month follow-up using paired t tests. The primary outcome was change in HbA_1c and body weight (kg, percent of initial body weight). The secondary outcome was change in need for diabetes medication. We stratified our cohort into three groups according to baseline glycemic control as defined by baseline HbA_1c: (1) elevated baseline HbA_1c greater than or equal to 7.5%, (2) slightly elevated baseline HbA_1c 6.5% to 7.4%, or (3) normal baseline HbA_1c less than 6.5%. Outcomes were also analyzed within strata based on participant’s Low-Carb Program completion (ie, completers: engaged with all 10 of the Low-Carb Program weekly modules; n=528), partial completers (engaged with 4-9 modules; n=144), or noncompleters (engaged with ≤3 modules; n=328).

Some of our results took into account the entire sample, regardless of follow-up information or lesson completion. For participants who did not report their outcomes at 12 months, we followed the highly conservative approach of assuming that they did not improve at all (last observation carried forward), by imputing their baseline values as their outcome values. For example, participants who did not comply with reporting a 12-month outcome were treated as having no change in the outcome variable, and thus were not counted as having any HbA_1c or weight improvement.

At baseline, mean HbA_1c was 7.8% (SD 2.1%), mean weight was 89.6 kg (SD 23.1), and mean age was 56.1 years (SD 15.7) years. More than half of participants were female (59.3%, 593/1000), 90.4% (904/1000) were white, all were from the United Kingdom, and more than one-third had comorbid hypertension (39.7%, 397/1000) or hypercholesterolemia (35.0%, 350/1000). At baseline, participants were taking a mean of 1.21 (SD 1.01) hypoglycemic medications. See Table 2 for details.

Of the 1000 baseline participants, 708 (70.80%) reported outcomes at 12 months, 528 (52.80%) completed all lessons, and 672 (67.20%) completed at least 40% of the lessons. For the remaining 292 people lost to follow-up, the last recorded data point was carried forward. Of 447 people with elevated HbA_1c (≥7.5%) at baseline, 247 (55.3%) reported outcomes at 12 months and 191 (42.7%) completed all lessons. Of 296 people with slightly elevated HbA_1c (6.5%-7.5%) at baseline, 238 (80.4%) had outcomes at 12 months and 182 (61.4%) completed all lessons. Of 257 people with a normal baseline HbA_1c level (HbA_1c <6.5%) who began the study, 223 (86.8%) had outcomes at 12 months and 155 (60.3%) completed all lessons (see Figure 1 for the participant flowchart of the study).

Considering all participants pooled across baseline HbA_1c, those who completed the Low-Carb Program showed a statistically significant change in HbA_1c of –1.17% (SD 1.43; t₅₂₇=18.724, P<.001). Partial completers showed a statistically significant change in HbA_1c of –0.6% (SD 1.69; t₁₄₃=4.276, P<.001) and noncompleters showed a nonsignificant HbA_1c change of only –0.16% (SD 1.13; t₃₂₈=2.54, P=.01). Results stratified by baseline HbA_1c are presented in Table 3, and results for just Low-Carb Program completers are presented in Figure 2.

Considering all baseline HbA_1c groups combined, Low-Carb Program completers (n=528) showed a significant reduction in weight, with a mean body weight change of –7.45 kg (SD 12.63) or –7.0% (SD 12.81%; t₅₂₇=13.551, P<.001). Partial completers (n=144) showed a reduction in weight, with a mean body weight change of –2.13 kg (SD 16.40) or –1.1% (SD 25.42%); however, this weight change was not statistically significant (t₁₄₃=1.563, P=.12). Noncompleters (n=328) did not have a statistically significant change in weight, with mean change of –0.35 kg (SD 10.13) or 0.7% (SD 13.41%; t₃₂₇=0.625, P=.53). Results, stratified by baseline HbA_1c, are presented in Table 4, and results for just Low-Carb Program completers are presented in Figure 3.

The majority of participants (714/1000, 71.40%) were prescribed at least one hypoglycemic medication at baseline. At 1 year, of those originally prescribed medications, 289/714 (40.4%) individuals were able to stop one or more hypoglycemic medications. Of the 743 participants who started with an HbA_1c, equal to or above the type 2 diabetes threshold of 6.5%, 195 (26.2%) reduced their HbA_1c to below the threshold while taking no glucose-lowering medications or just metformin.

For participants who completed the program, the proportion prescribed hypoglycemic medications changed significantly between baseline and follow-up for metformin (χ²₂₄=146.5, P<.05) and other hypoglycemic medications (all hypoglycemic medications other than metformin and insulin: χ²₂₄=73.8, P<.05). However, there was no significant change in being prescribed insulin (χ²₂₄=34.1, P=.08; see Figure 4).