This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on https://www.jmir.org/, as well as this copyright and license information must be included.
Type 2 diabetes mellitus (T2DM) is a worldwide public health concern. Mobile health management platforms could be a potential way to achieve effective glycemic control.
This study aimed to evaluate the real-world effectiveness of the Lilly Connected Care Program (LCCP) platform in glycemic control among patients with T2DM in China.
This retrospective study included Chinese patients with T2DM (aged ≥18 years) from April 1, 2017, to January 31, 2020, for the LCCP group and from January 1, 2015, to January 31, 2020, for the non-LCCP group. Propensity score matching was used to match the LCCP and non-LCCP groups to reduce confounding, with covariates including age, sex, the duration of diabetes, baseline hemoglobin A1c (HbA1c), and the number of oral antidiabetic medication classes. HbA1c reduction over 4 months, the proportions of patients achieving an HbA1c reduction of ≥0.5% or ≥1%, and the proportions of patients reaching to target HbA1c level of ≤6.5% or <7% were compared between the LCCP and non-LCCP groups. Multivariate linear regression was used to assess factors associated with HbA1c reduction.
A total of 923 patients were included, among whom 303 pairs of patients were well matched after propensity score matching. HbA1c reduction during the 4-month follow-up was significantly larger in the LCCP group than the non-LCCP group (mean 2.21%, SD 2.37% vs mean 1.65%, SD 2.29%;
The LCCP mobile platform was effective in glycemic control among patients with T2DM in China in the real world.
The prevalence of diabetes increased globally over the past decades, and diabetes has become a major public health concern [
Type 2 diabetes mellitus (T2DM) has been reported to be associated with a higher risk of cardiovascular death compared to the population without diabetes [
Mobile health (mHealth) refers to the monitoring and sharing of health information through cell phones or smartphones, handheld tablets, and other wireless devices, which is a potential way to overcome barriers such as distance to care, limited access, and the short supply of qualified diabetes educators [
Here, we conducted a multicenter, retrospective, and observational study that matched patients with T2DM enrolled in the LCCP with those not using the LCCP to evaluate the effectiveness of the LCCP in glycemic control among patients with T2DM in real-world settings. The primary objective was to examine the hemoglobin A1c (HbA1c) reduction relative to baseline in patients with T2DM who actively used the LCCP (the LCCP group) and those not using the LCCP (the non-LCCP group). The secondary objective was to compare the proportions of patients who achieved a target glycemic response, defined as an HbA1c reduction of ≥0.5% or ≥1%, and the proportions of patients who achieved the targeted HbA1c level of <7% or ≤6.5% between the LCCP and non-LCCP groups.
This was a multicenter, retrospective, and observational study conducted in China. Outpatients with T2DM who had inadequate glycemic control with oral antidiabetic medications and received premixed insulin analogue treatment were encouraged by their physicians to register on the LCCP platform without any financial incentives. The LCCP program was initiated in 2017, and our study included patients with T2DM who were recruited in the LCCP group from April 1, 2017, to January 31, 2020. We selected a representative real-world sample of patients with T2DM who had inadequate glycemic control on oral antidiabetic medications and received premixed insulin analogue treatment but were not enrolled in the LCCP as the control group. Patients in the non-LCCP group were those who visited the hospitals in 4 large cities across China, including Chongqing, Dalian, Tianjin, and Xiamen, from January 1, 2015, to January 31, 2020. The reason that the time window for the non-LCCP group identification was 2 years earlier than that for the LCCP group was to ensure an adequate sample size for propensity score matching (PSM). Both the LCCP and non-LCCP groups received their routine treatment for diabetes, which includes lifestyle changes and treatment with oral diabetic medications in addition to premixed insulin according to the treatment guidelines in China. During the study period, the LCCP and treatment landscape for patients with T2DM initiating insulin therapies in China remained largely unchanged.
Deidentified patient-level data were retrieved from WeChat for the LCCP group and from the electronic medical records collected from the hospital information systems in the 4 hospitals for the non-LCCP group. Informed consent was obtained for the LCCP group and waived for the non-LCCP group as data were retrospectively collected. For the LCCP group, after informed consent was obtained, the BG results measured by the BG meter would be uploaded automatically via Bluetooth and collected together with the patient demographics and drug prescription information. For the non-LCCP group, BG data were collected in the laboratories and extracted from the hospital information systems. The date of the first prescription of any premixed insulin analogue was defined as the “index date” in our analyses.
Our study inclusion criteria for the LCCP and non-LCCP groups were (1) patients with T2DM aged ≥18 years old on the index date; (2) the first use of premixed insulin analogue was between April 1 2017, and January 31, 2020, for the LCCP group and between January 1, 2015, and January 31, 2020, for the non-LCCP group; (3) had at least one self-reported HbA1c result (“baseline HbA1c”), and the report date (T0) was within 3 weeks after the index date for the LCCP group and between 2 months prior to and 1 month after the index date for the non-LCCP group—if multiple readings were available, the result closest to the index date was used; (4) baseline HbA1c ≥7% and ≤15%; and (5) had at least one self-reported HbA1c result (“follow-up HbA1c”), and the report date (T1) was within 2 to 6 months after the index date—if multiple readings were available, the result closest to 4 months after the index date was used. An additional inclusion criterion for the LCCP group was being an active user on the LCCP platform, defined based on the following criteria: (1) patients who started using the LCCP smart BG-testing device within the first 3 weeks after enrollment and were followed up for 12 weeks; (2) patients who completed ≥3 diabetes education courses per week; and (3) those who had ≥6 self-monitoring BG tests per week during the 12-week follow-up period.
A review of 27 studies summarized that HbA1c reduction ranged from 0.16% to 4.2% in Asian patients with T2DM who initiated premixed insulin analogue, including Humalog premixed insulin [
We first compared the HbA1c reduction between the LCCP and non-LCCP groups. Then, we analyzed the percentages of patients who achieved the glycemic response of an HbA1c reduction of ≥0.5% or ≥1% and the percentages of patients who achieved the target HbA1c levels of <7% or ≤6.5% after 4 months of follow-up in the LCCP and non-LCCP groups, respectively.
PSM was used to reduce confounding. PSM is a widely used approach to accounting for multiple confounders in observational studies. In PSM, a matched population of treated and untreated participants is constructed based on the probabilities of receiving treatment. The underlying factors associated with treatment assignment are balanced through PSM by pairing each treated patient with one or more untreated patients that were roughly equally likely to have received the treatment [
We presented the mean and SD for continuous variables with normal distribution and the median and IQR for continuous variables with skewed distributions. The categorical variables were presented as numerals and percentages. Statistical comparisons between the LCCP and non-LCCP groups were conducted using 2-tailed independent
Multivariable linear regression before matching was used to identify the variables associated with HbA1c reduction. The backward stepwise selection method was used in the multivariable regression models. Variable screening and model adjustment were conducted according to the contribution of covariates in the model. Akaike information criterion value and adjusted
All statistical analyses were performed using R (version 4.0.2; R Foundation for Statistical Computing) statistical software. A 2-sided
The study conformed to the Declaration of Helsinki principles and was approved by the ethics committee of the Second Affiliated Hospital, Chongqing Medical University (ethics number: 2020 Ethics Review(62)); Second Hospital of Dalian Medical University (ethics number: 2021 Ethics Review(155)); Second Hospital of Tianjin Medical University (ethics number: 2021 Ethics Review(028)); and Zhongshan Hospital of Xiamen University (ethics number: 2021 Ethics Review(084)).
A total of 923 adult patients (aged ≥18 years) with T2DM receiving premixed insulin analogue were included in the analyses. Among these patients, 523 were in the LCCP group and 400 were in the non-LCCP group. The baseline characteristics before and after PSM are presented in
Patient characteristics in the Lilly Connected Care Program (LCCP) and non-LCCP groups before and after propensity score matching.
Variable and level | Before matching | After matching | |||||
LCCP (n=523) | Non-LCCP (n=400) | SMDa (%) | LCCP (n=303) | Non-LCCP (n=303) | SMD (%) | ||
Age (years), mean (SD) | 52.32 (11.88) | 57.49 (9.91) | 47.2 | 56.02 (11.02) | 56.19 (10.31) | 1.6 | |
|
0.6 | 0.7 | |||||
Female | 192 (36.7) | 148 (37) | 112 (37) | 111 (36.6) | |||
Male | 331 (63.3) | 252 (63) | 191 (63) | 192 (63.4) | |||
|
60.6 | 7.4 | |||||
0 | 226 (43.2) | 108 (27) | 108 (35.6) | 105 (34.7) | |||
1 | 222 (42.4) | 140 (35) | 123 (40.6) | 120 (39.6) | |||
2 | 66 (12.6) | 104 (26) | 63 (20.8) | 65 (21.5) | |||
≥3 | 9 (1.7) | 48 (12) | 9 (3) | 13 (4.3) | |||
Baseline HbA1cb (%), mean (SD) | 9.82 (1.88) | 9.51 (1.83) | 16.8 | 9.64 (1.80) | 9.61 (1.91) | 1.7 | |
Diabetes duration (years), median (IQR) | 1.01 (0.07-7.65) | 9.02 (2.98-14.02) | 69.3 | 5.02 (0.20-10.63) | 7.16 (2.01-12.07) | 12.6 |
aSMD: standardized mean difference.
bHbA1c: hemoglobin A1c.
We also compared the proportion of patients who achieved the glycemic response of an HbA1c reduction of ≥0.5% or ≥1%. Overall, 209 (69%) out of 303 participants in the LCCP group had an HbA1c reduction ≥1% at follow-up, which was significantly higher than that in the non-LCCP group (174/303, 57.4%;
Results showed that the LCCP group had a larger proportion of patients reaching the target HbA1c levels of <7% and ≤6.5% than the non-LCCP group, and the difference was statistically significant for the target HbA1c level of ≤6.5% (88/303, 29% vs 61/303, 20.1%;
To further investigate the variables associated with HbA1c reduction, a multivariate linear regression model was used (adjusted
Comparison of hemoglobin A1c (HbA1c) reduction between the Lilly Connected Care Program (LCCP) and non-LCCP groups after matching.
Group | LCCP (n=303) | Non-LCCP (n=303) | |
Follow-up HbA1c (%), mean (SD) | 7.43 (1.86) | 7.95 (1.85) | <.001 |
HbA1c reduction (%)a, mean (SD) | 2.21 (2.37) | 1.65 (2.29) | .003 |
HbA1c reduction ≥0.5%, n (%) | 229 (75.6) | 206 (68) | .04 |
HbA1c reduction ≥1%, n (%) | 209 (69) | 174 (57.4) | .003 |
Target HbA1c ≤6.5%, n (%) | 88 (29) | 61 (20.1) | .01 |
Target HbA1c <7%, n (%) | 128 (42.2) | 109 (36) | .11 |
aHbA1c reduction is calculated as baseline HbA1c minus follow-up HbA1c.
Variables associated with hemoglobin A1c (HbA1c) reduction.
Variables | Level (comparator vs reference) | Estimate (SE) | Statistic | |
LCCPa participation | LCCP vs non-LCCP | 0.492 (0.159) | 3.101 | .002 |
Age | N/Ab | –0.021 (0.006) | –3.236 | .001 |
Sex | Male vs female | 0.185 (0.140) | 1.317 | .19 |
Baseline HbA1c | N/A | 0.800 (0.037) | 21.431 | <.001 |
Duration of diabetes | Long vs shortc | –0.393 (0.155) | –2.537 | .01 |
Baseline dose of premixed insulin analogue | N/A | –0.011 (0.005) | –2.235 | .03 |
aLCCP: Lilly Connected Care Program.
bN/A: not applicable.
cLong is defined as a duration of diabetes ≥5 years, and short is defined as a duration of diabetes <5 years.
This was the first retrospective observational study demonstrating the effectiveness of the LCCP in glycemic control among patients with T2DM in China using real-world patients as comparators. We found that patients with T2DM in the LCCP group had a larger HbA1c reduction compared to those in the non-LCCP group (mean 2.21%, SD 2.37% vs mean 1.65%, SD 2.29%;
Diabetes education is critical for improving patients’ self-management [
Prior studies have investigated the use of the LCCP and illustrated its effectiveness in glycemic control through a single-arm design [
All the aforementioned studies were conducted in LCCP participants without comparing to patients not using the LCCP. In our study, we included patients not using the LCCP as the control group to provide further evidence of the effectiveness of the LCCP. Prior studies also illustrated the behavior improvement among LCCP users, such as the increased number of diabetes education courses taken and higher frequency of self-monitoring of BG [
Besides LCCP participation, other variables associated with HbA1c reduction were identified, such as age, sex, baseline HbA1c, and the duration of diabetes. Prior studies reported that women and younger patients were more willing to be involved in health app use than men and older adults [
Several limitations of this study are noted as follows. First, the follow-up period in this study was approximately 4 months, which was relatively short to observe the outcomes in diabetes management. Therefore, the long-term effectiveness of the LCCP education platform shall be further investigated. Second, because of the retrospective study design, some variables including residence, socioeconomic status, educational level, and diabetes-related complications were not collected or controlled for, which could potentially lead to confounding. Thus, randomized control trials to further evaluate the effectiveness and future studies with a larger sample size and longer follow-up period are needed. Third, recall bias might occur because HbA1c levels in the LCCP group were self-reported by the patients.
This real-world study found that the LCCP platform was effective in improving glycemic control among patients with T2DM in China. The LCCP platform provides an effective way to enhance glycemic control, patient education, and self-management in patients with T2DM, especially during the COVID-19 pandemic.
Comparison of hemoglobin A1c change between the Lilly Connected Care Program (LCCP) and the non-LCCP groups after full matching and inverse-probability of weighting analysis.
blood glucose
fasting blood glucose
hemoglobin A1c
Lilly Connected Care Program
mobile health
propensity score matching
type 2 diabetes mellitus
The authors thank all the patients who participated in this study. The authors would also like to thank Happy Life Tech (HLT), Yidu Tech Inc, for operation support.
This study was funded by Eli Lilly and Company.
DL, XM, and YY designed the study. DL, BS, YC, XS, YD, XM, and YY were involved in the data collection. XM was involved in the statistical analysis. All authors have critically reviewed the manuscript.
YD, XM, and YY are employees and shareholders of Eli Lilly and Company. The other authors have no conflicts of interest to declare.