Long-term Health Benets to Newly Diagnosed Type-2 Diabetes from Short-term Continuous-subcutaneous Insulin Injection Therapy

Background Continuous-subcutaneous insulin injection (CSII) therapy to type 2 diabetes mellitus (T2DM) patients generated short-term health benets. Our aims were to investigate long-term health benets of CSII monotherapy, in combination with metformin and pioglitazone, or with sitagliptin. Methods In this randomized clinical trial, patients were treated for around 90 days and were monitored for one year. Demographic and laboratory data were analysed using the UKPDS_OM2 program to estimate 20-year health benets. Multiple linear regression model was used to identify factors associated with changes of each health benet. ischaemic heart no in among the three benets were lower for older for females in and ulcer and in blindness on The three protocols produced the same benets. at three time points post-treatment, eleven health indicators were simulated using the United Kingdom Prospective Diabetes Study Outcomes Model II (UKPDS_OM2). Algorithms for the model were developed based on 30 years of follow-up data from the United Kingdom Perspective Diabetes Study [16]. The eleven health indicators were: an aggregation of life expectancy, quality-adjusted life expectancy (QALE), cumulative risk of all-causes of death and of eight essential diabetes-related complications, ischaemic heart disease [IHD], myocardial infarction [MI], heart failure [HF], stroke, amputation, renal failure [RF], blindness and ulcer. Values of health indicators at pre-treatment simulated by using values of parameters – demographic and laboratory data – collected Values of health at post-treatments were simulated by synthetically using values of parameters at three time-points: end of in-hospital treatment, 3-month follow-up and 12-month follow-up. In parameter setting of the simulation model, it was necessary to assume changes parameter from the multiple linear regression analyses show that prolongation of either life expectancy or total QALE among the three groups were not statistically signicant, after the demographic characteristics and baseline indicators were adjusted. This model also shows that extensions of life expectancy were age-dependent. The second (aged 41 to 46 years), third (aged 47 to 53 years), and fourth (aged 54 to 65 years) quartile age groups had less extension by 0.15, 0.28 and 0.61 years less, respectively, than the rst quartile (aged 29 to 40 years) age group. The changes of total QALE from pre-treatment to post-treatment were similar, with 0.18, 0.29 and 0.63 years shorter in the second, third and fourth quartile age groups, respectively. In addition, patients who were smoker gained less extension of total QALE (by 0.13 year) compared to those who were non-smokers or ex-smokers (Table 4). failure; IHD, ischaemic heart disease; LDL-c, low-density lipoprotein cholesterol; MI, myocardial infarction; PVD, peripheral vascular disease; QALE, quality-adjusted life expectancy; RF, renal failure; SBP, systolic blood pressure; T2DM, type 2 diabetes mellitus; UKPDS_OM2, United Kingdom Prospective Diabetes Study Outcomes Model II.


Abstract Background
Continuous-subcutaneous insulin injection (CSII) therapy to type 2 diabetes mellitus (T2DM) patients generated short-term health bene ts. Our aims were to investigate long-term health bene ts of CSII monotherapy, in combination with metformin and pioglitazone, or with sitagliptin.

Methods
In this randomized clinical trial, patients were treated for around 90 days and were monitored for one year. Demographic and laboratory data were analysed using the UKPDS_OM2 program to estimate 20-year health bene ts. Multiple linear regression model was used to identify factors associated with changes of each health bene t.

Results
For the 134 treated patients, most health bene t indicators were improved signi cantly, except for renal failure. For example, life expectancies increased by 0.41±0.48 year and quality-adjusted life expectancy (QALE) by 0.45±0.46 year (p<0.001). Reductions in 20-year risk were: amputation by 70.6%, ulcer 66.7%, blindness 57.1%, stroke 45.5%, myocardial infarction 43.5%, all-causes of death 20.5%, ischaemic heart disease 6.7%, with heart failure < 0.1%. However, no difference in bene ts was found among the three therapeutic protocols. Health bene ts were lower for older patients, for females in amputation and ulcer risk, and for smokers in blindness risk.

Conclusions
Short-term CSII therapy produced signi cant and multiple long-term health bene ts (based on simulated risk analyses) to T2DM patients and bene ts were modi ed by age, sex and smoking factors. The three therapeutic protocols produced the same bene ts.

Trial registration
The clinical trial was registered in ClinicalTrials.gov on November 15, 2011, with the registration number: NCT01471808. Background Type 2 Diabetes Mellitus (T2DM) is a common non-communicable disease around the world and is characterized by two major features: insulin resistance and insu cient insulin secretion from pancreatic beta-cells [1]. Furthermore, progressive deteriorations of beta-cell often lead to fasting and postprandial hyperglycaemia [1], which accelerates beta-cell dysfunction [2] and mortality. Therefore, insulin supplement therapy is often applied to these patients.
In addition to CSII therapy alone, combined therapies with other anti-hyperglycaemic medicines have shown additional short-term bene ts. For example, the combined therapy patients often required less total daily insulin doses [8][9][10][11][12], achieved their glycaemic goals sooner [10], experienced less hypoglycaemia [10,12] and nocturnal hypoglycaemia [13], and showed reduced glycaemic variability [8, 14] -a risk factor to vascular complications of diabetes [15] -than those with the CSII monotherapy. However, long-term health bene ts from these therapeutic protocols need to be determined systematically.
All patients were randomly assigned to one of the three in-hospital treatment groups: short-term continuous subcutaneous insulin infusion (CSII) monotherapy using rapid-acting insulin analogues (Humalog or NovoRapid), CSII plus metformin (Gehuazhi® 0.5 tid) and pioglitazone (actos® 30mg qd) (CSII-MP), and CSII plus sitagliptin (JANUVIA® 100mg qd) group (CSII-S). The CSII treatments for the three groups were the same, with an initial insulin dosage 0.5-0.7IU/kg/d. In 2 to 3 day after initiation of CSII treatments for the three groups of patients, when the glycaemic targets (fasting plasma glucose ≤ 6.0 mmol/l and postprandial blood glucose ≤ 8.0 mmol/l) were achieved and then maintained for two weeks, CSII treatments would be withdrawn. After the two-week CSII treatments, all patients were monitored at the 3rd and 12th months as outpatients. For two groups of patients, the three drugs were administered at the same time with CSII, but continued for 3 months from CSII initiation.

Data collection
Before initiation of treatment, data were collected from all patients: demographics, behavioural characteristics (age, sex, duration of diabetes, smoking status), anthropometric indices (height), and some additional conditions (peripheral vascular disease, atrial brillation, albuminuria). In addition, during pre-and post-therapy (end of in-hospital treatment, 3rd and 12th months), laboratory data were conducted: blood indicators Values of health indicators at pre-treatment were simulated by using values of parameters -demographic and laboratory data -collected at baseline ( Table 1). Values of health indicators at post-treatments were simulated by synthetically using values of parameters at three time-points: end of in-hospital treatment, 3-month follow-up and 12-month follow-up. In parameter setting of the simulation model, it was necessary to assume changes for each parameter over time. Parameters in continuous types were raised by 1.5% annually, while binary ones were not changed [17]. Only patients who had completed all laboratory parameters for the 12 months were included into the simulation model and in this report. Certain missing values at 3-month were inputted with ve-fold Multiple Imputation by Chained Equations via the "mice" package of R based on all candidate predictors and outcomes [18]. Variables which had high correlations (Coe cient of determination R square > 0.8) with others were inputted through the Bayesian linear regression method, whereas the rests were inputted via the Random Forest Imputations method. Patterns of missing data were investigated by the R function -"densityplot" -for inspecting the imputations and graphical representations [18].

Statistical Analyses
In description of data distribution, continuous variables which conformed approximately to normal distributions were presented as mean ± standard deviations, while others were presented as medians and internal-quartiles. Categorical variables were presented in terms of quantities and percentages.
Differences for all parameters at baseline among the three therapy groups were compared. The Chi-square test or Fisher exact test was used in categorical variables, and the paired t test or Wilcoxon's test was used for continuous variables.
The multiple linear regression model was used to identify effects of key factors on change extent for each health indicator. In model tting for each health indicator, each dependent variable was the difference as calculated by subtracting the pre-treatment from the post-treatment simulation values. The independent variables included therapy strategies, sex, age, peripheral vascular disease (PVD), baseline life expectancy, baseline total QALE, baseline SBP and baseline HR. Baseline life expectancy and baseline total QALE were for adjustment, and other variables were also used to identify their effects. All analyses were performed using R 4.0.3. A two-sided P value of 0.05 or less was considered to be signi cant.

Results
Among the 262 recruited patients in our clinical trial, 134 completed the one-year follow-up laboratory tests. Therefore, the latter were the study subjects for this report. Distribution of these subjects in the three treatment groups were: 42, 48 and 44 in CSII, CSII-MP and CSII-S, respectively. Their characteristics at baseline are summarized in Table 1. There was no signi cant difference in these characteristics among the three groups, except the pre-treatment SBP (P = 0.035) and HR (P = 0.028).
For all patients, simulation analyses of the collected data indicate that the treatments signi cantly increased life expectancy by 0.41 ± 0.48 year and QALE by 0.45 ± 0.46 year (all p < 0.001). Based on univariate analyses, there were no signi cant changes in the two indicators among the three treatment groups (Table 2). Total QALE: total quality-adjusted life expectancy.
From simulated risk analyses on the collected data, the nine health indicators at baseline (pre-treatment) for all patients were classi ed into four levels: the highest risk level, from 37% and 48%, included all causes of death and MI; the second level, from 10-18%, included amputation, IHD and stroke; the third level, from 4-7%, included blindness, ulcer and HF; and the lowest level included only RF with risk less than 3%.
After treatment, changes in simulation values of 20 years cumulative risks for all-causes of death, MI, amputation, IHD, stroke, blindness, ulcer and HF were all statistically signi cant (p < 0.001), except RF ( Table 3). The most reductions were observed for MI, amputation and all-causes of death 20%, 12% and 8%, respectively. Reductions for stroke, ulcer and blindness were around 4% each. Comparing results from the three treatment protocols, patients in CSII-MP group showed larger reductions for the 20 years cumulative risks in MI, amputation and all-causes of death (22%, 13% and 9%, respectively) than the other two groups (Table 3). On the other hand, the CSII group had fewer reductions for the same three health indicators: 17%, 10% and 7% respectively. However, there were no signi cant differences among results from the three protocols, except the 20 years cumulative risks for IHD by 1% in the CSII-MP group (Table 3).
Results from the multiple linear regression analyses show that prolongation of either life expectancy or total QALE among the three groups were not statistically signi cant, after the demographic characteristics and baseline indicators were adjusted. This model also shows that extensions of life expectancy were age-dependent. The second (aged 41 to 46 years), third (aged 47 to 53 years), and fourth (aged 54 to 65 years) quartile age groups had less extension by 0.15, 0.28 and 0.61 years less, respectively, than the rst quartile (aged 29 to 40 years) age group. The changes of total QALE from pre-treatment to post-treatment were similar, with 0.18, 0.29 and 0.63 years shorter in the second, third and fourth quartile age groups, respectively. In addition, patients who were smoker gained less extension of total QALE (by 0.13 year) compared to those who were non-smokers or ex-smokers (Table 4).  PVD, peripheral vascular disease; Total QALE: total quality-adjusted life expectancy; SBP, systolic blood pressure; HR, heart rate.
After the demographic characters and main baseline indicators were adjusted, the multiple linear regression analyses indicate that reductions of the 20 years cumulative risks in all-causes of death and the eight diabetic complications were not statistically signi cant among the three therapy groups (Table 5). Compared to patients aged 29 to 40 years, the 41 to 46 years patients achieved smaller decline of cumulative risks in amputation, MI and ulcer, with 7%, 6% and 1% less, respectively; the 47 to 53 years patients achieved 10% and 2% less reduction of cumulative risks in MI and IHD, respectively; the 54 to 65 years patients gained achieved 23%, 9%, 8% and 4% less decline in cumulative risks for MI, amputation, all-causes of death and stroke, respectively. Female patients achieved 8% and 2% less reduction of cumulative risks for amputation and ulcer, respectively. Those who had pre-treatment PVD, in comparison to those without, achieved 2% less reduction in stroke risk, but 10%, 4% and 2% more reduction in risks for amputation, all-causes of death and ulcer, respectively. Current smokers achieved 1% less reduction in blindness risk than non-smokers and ex- smokers. An increase of 1 bpm in baseline HR was correlated with 0.1% of more reduction in all-causes of death risk. An increase of 1 bpm of mmHg in baseline SBP was correlated with 0.1% more reduction in stroke risk.  PVD, peripheral vascular disease; Total QALE: total quality-adjusted life expectancy; SBP, systolic blood pressure; HR, heart rate.

Discussion
Previous studies indicate that short-term CSII treatment of newly diagnosed T2DM patients improved beta-cell function and glycaemic control with long-term drug-free remission [3,4]. In addition, glycaemic remission after therapy was associated with lower red blood cell distribution width at baseline [19], higher decrement of total daily insulin dose during CSII therapy [20], lower fasting plasma glucose [21] and elevated 1,5-anhydroglucitol [22]. However, long-term health bene ts from such short-term CSII therapy have not been well-characterized. Therefore, our investigation provides new information to ll the information gap.
From our clinical trial, all three CSII therapeutic protocols generated similar long-term health bene ts to the T2DM patients, with no signi cant differences among them. This indicates that addition of the three oral anti-hyperglycaemic medicines to CSII therapy did not enhance e cacy compared to CSII monotherapy. Analyses of effects for all patients indicate that their life expectancy and total QALE were signi cantly extended. In addition, the cumulative risks of all-causes of death and diabetic vascular complications: MI, stroke, amputation, blindness, and ulcer, were reduced signi cantly. In support of our observations, patients achieved glycaemic control and beta-cell function in less time, and longer glycaemic remission rates using intensive insulin therapy compared to oral therapies [3]. Other reports indicate that certain antidiabetic effects of CSII therapy were still effective 3 months after termination of the therapy [5,6].
Our data indicate e cacy of the CSII therapy (either alone or in combination with the oral medications). However, the therapy did not generate equal e cacy among the monitored complications of diabetes. The therapy was highly effective in both absolute risk reduction and ratio of risk reduction for amputation, MI and all-causes of death. Changes in the absolute risks for stroke, blindness and ulcer were small, due to the small cumulative risks at baseline. It was a pleasant observation that the therapy was estimated to prevent half of the patients from suffering each disease. On the other hand, the therapy had limited effects on preventing IHD and HF. Although changes in their risk were statistically signi cant, both the absolute reduction and the ratio of risk reduction were too small to be clinically relevant. Therefore, from the perspective of population bene ts, the CSII therapy generated signi cant bene ts to T2DM patients in MI and amputation, followed by stroke. Although ulcer and blindness had high rates of changes, their risks were low. Furthermore, the therapy generated little to no bene ts to IHD, HF, and RF.
From 3 months after the therapy, changes in the variations of core parameters (HbA1c, HDL-c, LDL-c, etc.) were observed, but there were no statistically signi cant changes among the parameters and health indicators, as well as among the three therapy groups. A similar observation indicates diminutions of beta-cell functions which were improved via intensive insulin treatment, whether the patients were in the sitagliptin or placebo groups [23]. Furthermore, protective effects of sitagliptin on beta-cell functions were not noticeable [23]. Other reports, however, indicate some bene ts from intensive insulin therapy with oral anti-hyperglycaemic agents: less hypoglycaemia [10,12,13], lower glycaemic variability [8,14], and less usage of daily insulin [8][9][10][11][12]. For these reasons, adding oral agents to CSII monotherapy may still be useful, even though our monitored end points did not show additional long-term health bene ts.
Our results also show that patients who were younger achieved more long-term health bene ts than the elders from the three therapeutic protocols, although there was no signi cant difference among them. Our observations are supported by a report on longer drug-free glycaemic remission after short-term CSII [5]. Other important associations were identi ed by our study: the cumulative risks for ulcer, amputation and all-causes of death were reduced more in patients with pre-treatment PVD, while that of stroke in the same patients was on the contrary. Female patients received fewer bene ts in amputation and ulcer. Obviously, all observed associations have practical values in improving the management of patients' health. Further re nement of our observations may contribute to achieving personalized health management. For example, better algorithms for classi cation of personal needs should identify more precisely the roles of various in uences and factors on health bene ts.
Despite our effort to conduct this clinical trial vigorously, there are limitations in this study. First, the long-term health bene ts were simulated based on the UKPDS_OM2 program. Despite our effort to adjust ethnicity effects (using Asian-Indians as representatives), there are still potential bias as shown in a report using a German population for estimating risks for death, MI and stroke [17]. Nevertheless, our risk estimation was a conservative one and the estimation is still useful for China because such an effort has not been performed in China previously. Therefore, our results should stimulate additional investigations in China and around the world. Second, out of the 262 recruited subjects, our investigation included only 134 who had completed the 12-month monitoring activities. However, the characteristics of the 134 individuals were similar to those of excluded subjects. It was our decision that inclusion of the excluded subjects who had incomplete data would certainly compromise our estimation of health bene ts.

Conclusions
CSII therapy alone generated positive and signi cant long-term health bene ts to T2DM patients. Based on our estimation criteria, addition of other anti-diabetic medication to CSII did not generate more health bene ts. Availability of data and materials The datasets used and analysed during the current study are available from the corresponding author on reasonable request.

Funding
This work was supported by the National Key R&D Program of China (grant numbers: 2018YFC1314100) and the Science and Technology Program of Guangzhou (grant numbers: 202002020053).

Authorship
All named authors meet the International Committee of Medical Journal Editors criteria for authorship for this article, take responsibility for the integrity of the work as a whole, and have given their approval for this version to be published.

Authors' Contributions
Wanjun Zhang contributed to the study by designing and writing the manuscript. Weijian Ke collected the original data and provided expert advice on results interpretation. Qiao Bian, Huijie Guo and Dantong Zheng de ned the analytic strategy and analysed the data. Qun He, Liehua Liu supported data collection, statistical modelling, and data analysis. Yanbing Li and Yinghua Xia provided expert advice about study design, data analysis and results interpretation.