Short- and Long-Term Effects on Glucose Control of Nonadherence to Insulin Therapy in People With Type 2 Diabetes An In Silico Study

Abstract

Background:

Strict adherence to multiple daily insulin (MDI) therapy is a cornerstone for the achievement of good glucose control in people with advanced type 2 diabetes (T2D). Here, we aim to in silico assess glucose control in T2D subjects with poor adherence to MDI therapy.

Methods:

We tuned the Padova T2D Simulator, originally describing early-stage T2D physiology, around advanced T2D people. One hundred in silico advanced T2D subjects were generated and equipped with optimal MDI therapy: specifically, basal and bolus insulin amounts and injection times were individualized for each subject by applying titration algorithms that iteratively update insulin dose based on glucose deviation from its target. Then, the effect of nonadhering to MDI therapy was assessed using standard glucose control metrics calculated in two 6-month 3-meal/day in silico scenarios: in Scenario 1, subjects received the optimal basal and prandial insulin bolus at each meal; in Scenario 2, subjects received optimal basal insulin and randomly delayed or skipped the prandial insulin bolus in 3 lunches during working days and 1 dinner during weekends.

Results:

A statistically significant degradation was found in all glucose control outcome metrics in Scenario 2 versus Scenario 1: e.g., percent time above 180 mg/dL increased by 22.2% and glucose management index by 0.2%.

Conclusions:

Impaired adherence to MDI therapy in T2D leads to glucose control deteriorations in both short and long terms. Interestingly, short-term hyperglycemia seems being contrasted by residual endogenous insulin secretion, which statistically increased by 3-fold after delayed/skipped insulin boluses compared with optimal ones.

Keywords

computer simulation diabetes control modeling diabetes complications

Introduction

Type 2 diabetes (T2D) affects more than 400 million people worldwide, with a significant financial burden mainly related to hospitalizations and medications used to treat disease complications.¹ T2D is characterized by concomitant insulin resistance and impaired beta-cell function leading to chronic hyperglycemia, which requires different medications depending on the stage of disease progression. In particular, people with advanced T2D must follow a strict schedule of multiple daily injection (MDI) of insulin analogs, similar to patients with type 1 diabetes (T1D), to cover both basal and prandial insulin needs, necessary to guarantee the optimal blood glucose (BG) control. Nevertheless, skipping or delaying prandial insulin bolus is quite common among individuals with diabetes, either since they simply forget it or because they intentionally hide their disease to colleagues and friends due to fear of social stigma. This issue has been widely studied in T1D, especially in adolescents and young adults,^2
-7 showing a 0.8% to 1.5% increased glycated hemoglobin (HbA1c) in subjects failing adherence to MDI therapy. In this regard, it has been shown that emerging technologies, like continuous glucose monitoring, connected insulin pens and artificial pancreas, may help to maximize therapy adherence and consequently improve BG control.^7
-9 However, the impaired adherence to insulin therapy in T2D patients is still poorly investigated^10
-12 and differences between T1D and T2D may exist. This calls for developing tailored solutions to properly predict and treat diabetes complications in target populations.

In order to contribute to this field, here we aim at in silico investigating the effects on glucose control of nonoptimal adherence to MDI therapy in virtual T2D people. Similarly to what was done in T1D,¹³ this is realized by performing an in silico trial using the Padova T2D simulator,¹⁴ ad hoc modified to represent an advanced T2D population under MDI therapy, and by evaluating BG control degradation in both short and long terms.

Methods

The Padova T2D Simulator

The University of Padova T2D Simulator (T2DS)¹⁴ describes glucose, insulin, and C-peptide dynamics in response to specific perturbations, like meal carbohydrate (CHO) intakes and administered treatments. The T2DS consists of a model (Figure 1) describing glucose–insulin–C-peptide interactions and their metabolic fluxes (i.e., intestinal glucose absorption, endogenous glucose production, glucose utilization, and endogenous insulin secretion) and a cohort of 100 in silico early-stage T2D subjects, each one represented by a vector p_j containing the N model parameters:

p_{j} = {[p_{j 1}, p_{j 2}, \dots, p_{j N}]}^{T}, j = 1, 2, \dots, 100

(1)

Each p_j is generated by random extraction from a lognormal joint parameter distribution LN( μ _p , Σ _p ) of model parameters estimated in 204 healthy¹⁵ and 51 early-stage T2D people^16
-18 studied with multiple-tracer experiments.¹⁹ A detailed description of the procedure for subject generation is reported in Visentin et al.¹⁴

Figure 1.

Scheme of the Padova T2D simulator¹⁴ integrating models of subcutaneous fast-/long-acting insulin pharmacokinetics, CGM, and SMBG.

The in silico early-stage T2D population is not appropriate for the proposed aim, since these patients, due to their moderately impaired insulin function, are not treated with exogenous insulin for BG control. Therefore, the T2DS needs to be integrated with an in silico insulin-dependent advanced T2D population and models of exogenous insulin delivery to enable MDI therapy. Of note, the T2DS already incorporates models of both CGM²⁰ and self-monitoring of blood glucose (SMBG).²¹

Insulin Therapy Incorporation

To simulate MDI therapy, two pharmacokinetic (PK) models describing subcutaneous administration of both basal long-acting insulin degludec (iDeg) and prandial fast-acting insulin lispro have been incorporated into the T2DS. These models have been developed and validated against clinical data in previous studies.^22,23 The PK models are described by the following mass balance equations:

{\begin{array}{l} {\dot{I}}_{i D e g 1} (t) = - k_{d 1} \cdot I_{i D e g 1} (t) + F \cdot D o s e_{i D e g} δ (t) & I_{i D e g 1} (0) = 0 \\ {\dot{I}}_{i D e g 2} (t) = - k_{d 2} \cdot I_{i D e g 2} (t) + k_{d 1} \cdot I_{i D e g 1} (t) & I_{i D e g 2} (0) = 0 \\ {\dot{I}}_{i D e g 3} (t) = - k_{a} \cdot I_{i D e g 3} (t) + k_{d 2} \cdot I_{i D e g 2} (t) & I_{i D e g 3} (0) = 0 \end{array}

(2)

{\begin{array}{l} {\dot{I}}_{l i s 1} (t) = - (k_{a 1} + k_{d}) \cdot I_{l i s 1} (t) + & {\dot{I}}_{l i s 1} (0) = 0 \\ + D o s e_{l i s} \cdot δ (t - τ) \\ {\dot{I}}_{l i s 2} (t) = - k_{a 2} \cdot I_{l i s 2} (t) + k_{d} \cdot I_{l i s 1} (t) & {\dot{I}}_{l i s 2} (0) = 0 \end{array}

(3)

R a_{I} (t) = k_{a} \cdot I_{i D e g 3} (t) + k_{a 1} \cdot I_{l i s 1} (t) + k_{a 2} \cdot I_{l i s 2} (t)

(4)

where Dose_iDeg (pmol/kg) is the iDeg dose, $δ$ is the unitary impulse response to an instantaneous bolus, F (dimensionless) is its bioavailability, k_d₁ and k_d₂ (min)^-1 are molecular conversion rates, k_a (min)^-1 is the iDeg absorption rate, Dose_lis (pmol/kg/) is the lispro dose, τ (min) is the delay in lispro appearance, k_d (min)^-1 is the rate diffusion, k_a₁ and k_a₂ (min)^-1 are the rates of lispro absorption, and Ra_I (mg/kg/min) is the total exogenous insulin rate of appearance entering in the insulin subsystem of Figure 1.

Insulin-Dependent Advanced T2D Population

The T2DS has been tailored to an insulin-dependent advanced T2D population, by adopting a recently proposed population cloning technique.²⁴ In particular, the T2DS model has been identified on average plasma glucose, insulin, and C-peptide time courses data of insulin-naïve T2D patients (ie, subjects who are initiated to insulin therapy) collected during a mixed meal tolerance test (MMTT),²⁵ and fasting concentrations of the same metabolites and hormones measured in insulin-treated T2D patients.²⁶ It is worth underlining that data available from Holst et al.²⁵ were related to insulin-naïve subjects, i.e., who just started insulin therapy, while the present work focuses on individuals with uncontrolled T2D who already are under insulin therapy and likely exhibit even more impaired insulin function. Thus, we considered fasting data coming from Uzunlulu et al.²⁶ to maximize the compatibility between the tailored T2DS and the target population object of study. Given the complexity of T2DS model, it was impossible to reliably estimate the entire set of model parameters from the sole average plasma glucose, insulin, and C-peptide data. Thus, only key metabolic parameters describing insulin sensitivity (V_mx, k_p₃), insulin secretion ( Φ _b , Φ _s , Φ _d ) and hepatic extraction (HE_b) were estimated from MMTT data, while the remaining parameters were fixed to population values. To note, insulin sensitivity, secretion, and hepatic extraction are known to be the key parameters that mostly characterize T2D pathophysiology and differ among populations, as already found in previous analysis.²⁷ Model parameters were estimated using a Maximum a Posteriori estimator, using μ _p , and Σ _p as a priori information (a similar approach was previously developed for in silico cloning T1D individuals).²⁸

The insulin PK parameters obtained from the literature,^22,23 the parameter set obtained by identifying the model on the average data of Holst et al.²⁵ and the fasting concentrations reported in Uzunlulu et al.²⁶ were used to update the distribution mean vector (µˆ $μ$ _p). However, the availability of population average data only precluded the possibility to run individual model identifications; therefore, the covariance matrix was kept as the original Σ _p . Thus, the updated joint parameter distribution LN (µˆ $μ$ p, Σ _p ) was used for generating a new in silico cohort of 100 insulin-dependent advanced T2D adults treated with basal/bolus insulin.

Insulin Therapy Individualization

Both iDeg and lispro doses were individualized for each in silico subject, in terms of amount and time of administration. The procedure is described as follows.

Long-acting insulin individualization

Subject-specific basal iDeg dose amount and injection time were determined by a 2-step procedure. Each subject underwent two 6-month trials, with 3 meal/day (carbohydrate amounts: 58.2±22.5 g, 77.7±27.0 g, 83.9±32.3 g, at breakfast, lunch, and dinner, respectively) and iDeg injected at morning or evening, respectively. iDeg amount was individualized following the American Diabetes Association (ADA) standard guidelines,²⁹ starting from 10 U and optimally adjusting the dose toward a fasting plasma glucose (FPG) target of [80-130] mg/dL. In particular, daily basal iDeg dose (D) adjustment was done every 3 days based on the following titration rule:

D_{k + 1} = {\begin{array}{l} D_{k} - 4 U & i f F P G < 56 m g / d L \\ D_{k} - 2 U & i f F P G \in [56 - 80) m g / d L \\ D_{k} + 2 U & i f F P G \in (130 - 180) m g / d L \\ D_{k} + 4 U & i f F P G \in [180 - 270) m g / d L \\ D_{k} + 6 U & i f F P G \geq 270 m g / d L \end{array}

(5)

where FPG was calculated as mean of the last 3-day values.

Then, iDeg injection time was determined for each subject following the scheme illustrated in Figure 2, i.e., it was chosen between morning and evening as the one causing the minimum percent time in hypoglycemia (BG < 70 mg/dL) and the maximum percent time in euglycemia (BG in target [70-180] mg/dL).

Figure 2.

Decision tree describing the method for selecting the optimal iDeg injection time.

Fast-acting insulin individualization

Individual prandial fast-acting insulin lispro was determined based on clinical guidelines.³⁰ Specifically, each virtual subject underwent a 6-month 3-meal/day trial, receiving its individualized daily basal insulin (TDI_basal) determined in the previous step. Mealtime insulin lispro dose was initiated at the 20% of TDI_basal and adjusted every 3 days based on either 2-hr postmeal glucose (with target range of [120-160] mg/dL) or values before the following meal or at bedtime (with target range of [90-130] mg/dL). In particular, insulin adjustments at breakfast, lunch and dinner were done by varying the respective insulin-to-carb ratio (CR) as follows:

C R_{k + 1} = {\begin{array}{l} C R_{k} + \frac{C R_{m a x} - C R_{k}}{2}; C R_{m i n} = C R_{k} \\ i f G_{1} < 120 m g / d L o r G_{2} < 90 m g / d L \\ C R_{k} - \frac{C R_{k} - C R_{m i n}}{2}; C R_{m a x} = C R_{k} \\ i f G_{1} > 160 m g / d L o r G_{2} > 130 m g / d L \end{array}

(6)

where CR_max and CR_min are initiated to 200 and 1 g/U, respectively; G₁ is the 2-hr postmeal glucose; and G₂ is the glucose value before the following meal or at bedtime. Finally, for each meal, the amount of insulin lispro (Bolus) was calculated from the estimated carbohydrate content ( $\hat{C H O}$ ) as

B o l u s = \frac{\hat{C H O}}{C R}

(7)

In Silico Trial

An in silico trial has been performed using the T2DS incorporating the advanced T2D patients under MDI therapy, to evaluate the impact of nonadherence to insulin therapy on BG control. Specifically, the in silico advanced T2D patients underwent two 6-month scenarios, with realistic meal administrations³¹: three meals per day were randomly generated for each subject, assuming mealtime uniformly distributed in the intervals 06:30 am to 08:00 am (breakfast), 11:30 am to 01:00 pm (lunch), 06:30 pm to 08:00 pm (dinner); meal amounts were randomly sampled from a uniform distribution with mean and SD derived from data published in Brazeau et al.³¹ (58.2 ± 22.5 g for breakfast, 77.7 ± 27.0 g for lunch, 83.9 ± 32.3 g for dinner).

For each scenario, each subject received the individual basal insulin (iDeg) injected at breakfast or bedtime, as described in Section 2.4, while meal insulin (lispro) bolus was determined from subject’s CR and meal $\hat{C H O}$ by equation (7). For what concerns meal insulin bolus, two different levels of adherence were assumed: (1) for all the simulated meals, each subject received the optimal insulin bolus at mealtime (Scenario 1); (2) among the 100 in silico subjects, 65 randomly injected the insulin bolus with an arbitrary delay (between 30 and 60 min) with respect to meal ingestion in 3 lunches during working days and 1 dinner during weekends, while 35 randomly missed the meal insulin bolus in 3 lunches during working days and 1 dinner during weekends. In particular, in case of delayed boluses, the insulin amount was determined by applying a reduction with respect to the optimal bolus, which linearly increased with the delay, from 0% (i.e., no delay) to 100% (delay of 60 min), to reduce the risk of postprandial hypoglycemia. Of note, Scenario 2 was designed according to Cherñavvsky et al.,⁷ mimicking what occurs in real life: the patient occasionally delays/skips the lunch insulin bolus because sometimes she/he forgets or she/he works full-time and intentionally delivers insulin bolus once back to office instead of at the restaurant; similarly, this can occasionally happen during weekend dinners. It is also worth noting that the 6-month length ensures that, for each subject, all the possible delayed and skipped insulin bolus combinations are explored at least once.

Glucose Control Metrics

Glucose control was assessed in terms of simulated CGM using standardized control outcome metrics for clinical care,^29,32 i.e., average glucose (Mean, mg/dL), glucose standard deviation (SD, mg/dL), glucose coefficient of variation (CV, %), percent time CGM-glucose in ranges [70-180] (T_t,70-180, %), <70 (T_b,70, %), <54 (T_b,54, %), >180 (T_a,180, %), >250 mg/dL (T_a,250, %), and glucose management index (GMI, %).³³

Glucose control metrics were calculated based on the 6-hour CGM data following each meal affected by delayed or skipped insulin bolus. Values were reported as mean ± standard deviation (SD) for normally distributed variables, otherwise as median [25th-75th] percentiles. Statistical comparison was done by Student’s t-test for normally distributed variables, or Wilcoxon sign rank test, otherwise, with significant p-value set at 0.05.

Results

Insulin-Dependent Advanced T2D Population

The in silico population of insulin-dependent advanced T2D subjects was characterized by a mean ± standard deviation (SD) age of 55 ± 18 year, body weight of 87 ± 15 kg, body mass index of 31 ± 5 kg/m², basal glucose of 241 ± 31 mg/dL, basal insulin of 33 ± 13 pmol/L, basal C-peptide of 435 ± 133 pmol/L. These baseline characteristics values are in agreement with the clinical study²⁶ and reflect those of T2D patients who start MDI therapy for BG control to compensate insufficient beta-cell function.^29,34 It is important to point out that the resulting basal glucose is expected to be much higher than that reported in both studies.^25,26 Indeed, subjects in Holst et al.²⁵ are insulin-naïve T2D; moreover, it is expected that their fasting glucose is decreased following basal insulin titration. Similarly, subjects studied in Uzunlulu et al.²⁶ are patients with T2D who had been treated with insulin as a monotherapy or as a component of combination therapy; hence, their lower blood glucose (~130 mg/dL) is the result of insulin treatment. Conversely, the fasting glucose reported here is that of individuals with advanced T2D in absence of exogenous insulin (i.e., before starting basal/bolus therapy). Simulated plasma glucose, insulin, and C-peptide curves obtained by replicating the MMTT described in Holst et al.²⁵ are shown in Figure 3. Concentration peaks occur at (mean ± SD) 116 ± 13 min, 124 ± 23 min, and 154 ± 23 min, for glucose, insulin, and C-peptide, respectively, similarly to clinical observations (about 120 min).²⁵

Figure 3.

Simulated glucose (left), insulin (center), and C-peptide (right panel) concentrations in advanced T2D subjects in uncontrolled conditions (i.e., without exogenous insulin administration), obtained during a 480-min meal test with 96 g of carbohydrates given at t = 60 min, mimicking the clinical study.²⁵

Basal and Bolus Insulin Titration

The adaptation of basal (iDeg) and bolus (lispro) insulin dose following the titration periods are shown in Figure 4. After 6 months, mean ± SD FPG was 124 ± 10 mg/dL (baseline change of -110 ± 49 mg/dL), as a result of a total daily insulin (TDI) of 1.32 ± 0.70 U/kg/day distributed among TDI_basal of 0.82 ± 0.48 U/kg/day (62% ± 14% of TDI) and TDI_bolus of 0.49 ± 0.31 U/kg/day (38% ± 14% of TDI). To note, FPG values are in agreement with ADA standards of care,²⁹ and the TDI_basal-to-TDI_bolus ratio well matches that reported in the real study.³⁴

Figure 4.

Progression of simulated basal insulin dose (iDeg, a) and bolus insulin at breakfast (Bolus_B, b), lunch (Bolus_L, c) and dinner (Bolus_D, d), obtained during the 6-month period. In silico data are reported as mean (thick line) ± standard deviation (shaded area).

Optimal Versus Impaired Adherence to MDI Therapy

In silico CGM time courses obtained with optimal (Scenario 1) versus impaired adherence to MDI therapy (Scenario 2) are shown in Figure 5. It can be noted that both CGM average and variability mostly differ among scenarios after lunch during working days and after dinner during weekends. In particular, Scenario 2 leads to more elevated glucose levels toward the hyperglycemic range.

Figure 5.

Comparison of weekly BG time courses obtained by simulating optimal therapy (Scenario 1, red line and shaded area) and delay/skip bolus in 65/35 ratio (Scenario 2, blue line and shaded area) in 100 in silico adults with T2D. Weekly BG are meant as average of 28 weeks. Data are reported as mean (continuous thick line) ± standard deviation (shaded area).

This result is confirmed by the distribution of CGM metrics (Table 1). Failing in adhering to MDI therapy resulted in a degradation of BG control, with statistically significant increase of glucose mean and variability (quantified by SD and CV), a reduction of T_t,70-180 by 38% (median percent relative change with respect to Scenario 1), and an increase of both T_a,180 and T_a,250 by 57% and 258%, respectively. However, the reduction of both T_b,70 and T_b,54 was statistically significant, but the associated risk of late postprandial hypoglycemia was not clinically relevant. The glucose control degradation also is noticeable in the long term, proven by the statistically significant (P < .001) increase of GMI.

Table 1.

Distribution of Glucose Control Outcome Metrics Obtained Under Optimal Therapy (Scenario 1) or Delay/Skip Bolus in 65/35 Ratio (Scenario 2).

	Scenario 1	Scenario 2
Mean (mg/dL)	169.0 [154.3-181.5]	202.1 [189.5-215.7]*
SD (mg/dL)	24.3 [20.2-28.1]	32.6 [28.7-37.8]*
CV (%)	14.6 [12.7-15.8]	16.0 [14.6-18.0]*
T_t,70-180 (%)	61.0 [49.1-73.0]	34.3 [27.6-43.3]*
T_b,70 (%)	0.6 [0.3-1.3]	0.3 [0.1-0.6]*
T_b,54 (%)	0.1 [0.0-0.4]	0.1 [0.0-0.2]*
T_a,180 (%)	38.7 [25.0-49.9]	65.3 [56.7-71.9]*
T_a,250 (%)	3.7 [1.9-9.2]	19.3 [10.9-29.1]*
GMI (%)	7.1 [6.8-7.3]	7.3 [7.1-7.6]*

All metrics were calculated based on the individual CGM profiles during the 6 hours following each meal with delay/skip insulin bolus, except for GMI, which was calculated on a 28-week basis. Values are reported as mean ± SD for normally distributed variables, or as median [25th-75th] percentiles, otherwise. Statistical comparison was done using paired Student t-test for normally distributed values, Wilcoxon signed-rank test otherwise. Significance level was set at P = .05 (asterisk and bold highlights a statistically significant difference).

Abbreviations: SD, standard deviation; CV, coefficient of variation; T_t,70-180, time in target range [70-180] mg/dL; T_b,70, time below 70 mg/dL; T_b,54, time below 54 mg/dL; T_a,180, time above 180 mg/dL; T_a,250, time above 250 mg/dL; GMI, glucose management index; CGM, continuous glucose monitoring.

Discussion

This work evaluates in silico the impact of a nonadherence to MDI therapy on glucose control in people with advanced T2D. To achieve this aim, we exploited an in silico cloning technique to make the Padova T2D simulator (T2DS) able to well reproduce glucose regulation in insulin-dependent people with advanced T2D. To simulate MDI therapy, the T2DS was also equipped with PK models of both basal (iDeg) and prandial (lispro) insulins, and specific titration algorithms were implemented to determine, for each in silico subject, the optimal individual therapy parameters, i.e., insulin injection time and dose amount.

The in silico trial was designed to recreate a typical weekly routine, assuming T2D patients deviating from the perfect adherence to MDI therapy by delaying or skipping some meal insulin boluses. Such a situation is rather probable, and can be due to both lifestyle factors and psychosocial barriers (e.g., patients miss insulin injection at mealtime due to low attention, fear of embarrassment, and social stigma).³⁵ The resulting poor adherence translated into a BG control deterioration with prolonged percent time in hyperglycemia. This was particularly evident in the postprandial phases following delayed or skipped insulin meal bolus, while it was not clinically meaningful (even if still statistically significant) when calculated throughout the 24-hr time interval (data not shown). However, this is expected since the fasting overnight period is far from meal intakes, and thus less affected by nonadherence.

As mentioned in the introduction, nonadherence to MDI therapy has been widely studied in T1D, and the increased time in hyperglycemia is known occurring in this population as well. However, we evaluated whether impaired adherence has a similar or different impact on glucose control in the two populations, by performing the same in silico trial (i.e., both Scenarios 1 and 2) in the virtual T1D population of the UVA/Padova T1D Simulator.³⁶ Similarly to what was found in T2D, the glucose control outcome distribution revealed a degradation of glucose control toward increased hyerglycemic ranges and high average glucose (by means of GMI). Interestingly, the increase of time in hyperglycemia was significantly higher in T1D than T2D: percent relative changes of T_a,180 were 123 [59-274] in T1D versus 57 [23-148] in T2D (P < .001), percent relative changes of T_a,250 were 422 [216-1502] in T1D versus 258 [123-914] in T2D (P = .032). The lower increase of time in hyperglycemia observed in T2D population can be potentially explained by the T2D residual beta-cell function, that could compensate, even if minimally, for the inadequate exogenous insulin delivery. Indeed, by looking at the available in silico C-peptide concentrations in T2D (Figure 6), it can be observed that C-peptide secretion is increased in Scenario 2 versus Scenario 1, especially after working days lunches and during weekend dinners, with area under the C-peptide curve growing from 11.8 [3.6-32.8] 10³ pmol min/L to 46.5 [26.2-77.0] 10³ pmol min/L (P < .001) and maximum C-peptide concentration increasing from 134 [40-315] pmol/L to 410 [243-642] pmol/L (P < .001). However, glucose control degradation in the long term was statistically similar, with GMI increasing by 0.2% [0.1%-0.3%] in both T1D and T2D populations.

Figure 6.

Comparison of weekly C-peptide time courses obtained by simulating optimal therapy (Scenario 1, red line and shaded area) and delay/skip bolus in 65/35 ratio (Scenario 2, blue line and shaded area) in 100 in silico adults with T2D. Weekly C-peptide are meant as average of 28 weeks. Data are reported as mean (continuous thick line) ± standard deviation (shaded area).

The differences observed between the two populations are in line with what was found in a prospective study analyzing people with T1D and T2D³⁷ and, implicitly, underline the importance of running in silico trials with the most appropriate population, to obtain reliable results for the target group. Indeed, a simulator able to well describe the dynamics of a specific population represents a reliable framework for supporting the development of targeted therapies and predictive algorithms to improve BG control and minimize diabetes complications.

Despite the positive results, this study presents some limitations. First, the usability of T2DS tailoring methodology relies on data availability. This approach has been established to cope with the lack of individual glucose tracer data for the target population, which precludes the possibility of building a population-specific covariance matrix employed for subject generation. This implies that the covariance matrix is assumed to be the same in both early- and advanced-stage T2D. Nevertheless, this limitation does not preclude the simulator reliability, as recently presented in Visentin et al.²⁴: specifically, it was shown that the tailoring methodology allows the T2DS to well reproduce both average and variability outcomes observed in a real insulin-naïve T2D population.

It is also worth pointing out that the in silico trial assumed a precise model of meal distribution,³¹ as well as insulin dosing strategies based on well-established guidelines.^29,30 This was done to mimic, as much as possible, what occurs in real life. Nevertheless, future works will investigate the effect of nonadherence to insulin therapy under different conditions and alternative (e.g., preventive) dosing strategies.

Conclusion

In this article, we showed that impaired adherence to MDI therapy impacts glucose control in people with T2D on both short- and long-term scenarios, resulting in increased percent time spent in hyperglycemic ranges and higher GMI. Such findings were similar to what also occurs in T1D patients, however revealing that the residual endogenous insulin secretion in T2D contributes to mitigate the BG control degradation toward hyperglycemia due to the incorrect exogenous insulin administration. Despite this minimal physiological countermeasure, it is worth underlining that adhering to insulin schedule as much as possible is fundamental to guarantee the best glucose control management. In this direction, the proposed simulations as well as the basal-bolus T2DS environment can be further used as test bench for assessing the potential contribution of cutting-edge technologies to increase patients’ compliance and adherence to therapy, and to determine clinical differences of the impact of such technologies between people with T1D and T2D. This population-centric approach may help to minimize diabetes complications and optimize BG control.

Footnotes

Abbreviations

ADA, American Diabetes Association; B, breakfast; BG, blood glucose; CGM, continuous glucose monitoring; CHO, carbohydrates; CR, carbohydrate-to-insulin ratio; D, dinner; GMI, glucose management index; iDeg, insulin degludec; L, lunch; MMTT, mixed meal tolerance test; PK, pharmacokinetics; SD, standard deviation; SMBG, self-monitoring of blood glucose; T1D, type 1 diabetes; T2D, type 2 diabetes; UVA, University of Virginia.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was funded by Becton, Dickinson and Company.

ORCID iDs

Roberto Visentin

Claudio Cobelli

Chiara Dalla Man

References

International Diabetes Federation. IDF Diabetes Atlas. http://www.diabetesatlas.org/. Accessed January 4, 2024.

Hoffman

. Adolescent adherence in type 1 diabetes. Compr Ther. 2002;28(2):128-133.

Burdick

Chase

Slover

, et al. Missed insulin meal boluses and elevated hemoglobin a1c levels in children receiving insulin pump therapy. Pediatrics. 2004;113(3, pt 1):e221-e224.

Datye

Boyle

Simmons

, et al. Timing of meal insulin and its relation to adherence to therapy in type 1 diabetes. J Diabetes Sci Technol. 2018;12(2):349-355.

Clements

Foster

Maahs

, et al. Hemoglobin A1c (HbA1c) changes over time among adolescent and young adult participants in the T1D exchange clinic registry. Pediatr Diabetes. 2016;17(5):327-336.

Jaser

Datye

. Frequency of missed insulin boluses in type 1 diabetes and its impact on diabetes control. Diabetes Technol Ther. 2016;18(6):341-342.

Cherñavvsky

DeBoer

Keith-Hynes

, et al. Use of an artificial pancreas among adolescents for a missed snack bolus and an underestimated meal bolus. Pediatr Diabetes. 2016;17(1):28-35.

Vallis

Ryan

Berard

, et al. How continuous glucose monitoring can motivate self-management: can motivation follow behaviour? Can J Diabetes. 2023;47(5):435-444.

Steenkamp

Eby

Gulati

Liao

. Adherence and persistence to insulin therapy in people with diabetes: impact of connected insulin pen delivery ecosystem. J Diabetes Sci Technol. 2022;16(4):995-1002.

10.

Imai

Hardie

Georgiou

. Adherence to guideline-recommended HbA1c testing frequency and better outcomes in patients with type 2 diabetes: a 5-year retrospective cohort study in Australian general practice. BMJ Qual Saf. 2021;30(9):706-714.

11.

Unni

Gupta

Sternbach

. Trends of self-reported nonadherence among type 2 diabetes medication users in the United States across three years using the self-reported Medication Adherence Reasons Scale. Nutr Metab Cardiovasc Dis. 2022;32(1):151-159.

12.

Shah

Levy

. Emerging technologies for the management of type 2 diabetes mellitus. J Diabetes. 2021;13(9):713-724.

13.

Randløv

Poulsen

. How much do forgotten insulin injections matter to hemoglobin A1c in people with diabetes? A simulation study. J Diabetes Sci Technol. 2008;2(2):229-235.

14.

Visentin

Cobelli

Dalla Man

. The Padova type 2 diabetes simulator from triple-tracer single-meal studies: in silico trials also possible in rare but not-so-rare individuals. Diabetes Technol Ther. 2020;22(12):892-903.

15.

Basu

Dalla Man

Campioni

, et al. Effects of age and sex on postprandial glucose metabolism differences in glucose turnover, insulin secretion, insulin action, and hepatic insulin extraction. Diabetes. 2006;55(7):2001-2014.

16.

Bock

Dalla Man

Campioni

, et al. Pathogenesis of pre-diabetes: mechanisms of fasting and postprandial hyperglycemia in people with impaired fasting glucose and/or impaired glucose tolerance. Diabetes. 2006;55(12):3536-3549.

17.

Vella

Bock

Giesler

, et al. Effects of dipeptidyl peptidase-4 inhibition on gastrointestinal function, meal appearance, and glucose metabolism in type 2 diabetes. Diabetes. 2007;56(5):1475-1480.

18.

Basu

Dalla Man

Basu

Toffolo

Cobelli

Rizza

. Effects of type 2 diabetes on insulin secretion, insulin action, glucose effectiveness, and postprandial glucose metabolism. Diabetes Care. 2009;32(5):866-872.

19.

Basu

Di Camillo

Toffolo

, et al. Use of a novel triple-tracer approach to assess postprandial glucose metabolism. Am J Physiol Endocrinol Metab. 2003;284(1):E55-69.

20.

Facchinetti

Del Favero

Sparacino

Cobelli

. Model of glucose sensor error components: identification and assessment for new Dexcom G4 generation devices. Med Biol Eng Comput. 2015;53(12):1259-1269.

21.

Vettoretti

Facchinetti

Sparacino

Cobelli

. A model of self-monitoring blood glucose measurement error. J Diabetes Sci Technol. 2017;11(4):724-735.

22.

Schiavon

Visentin

Giegerich

, et al. In silico head-to-head comparison of insulin glargine 300 u/ml and insulin degludec 100 u/ml in type 1 diabetes. Diabetes Technol Ther. 2020;22(8):553-561.

23.

Schiavon

Dalla Man

Cobelli

. Modeling subcutaneous absorption of fast-acting insulin in type 1 diabetes. IEEE Trans Biomed Eng. 2018;65(9):2079-2086.

24.

Visentin

Schiavon

Bonet

Riz

Wagenhuber

Dalla Man

. Tailoring the Padova Type 2 Diabetes Simulator for Treatment Guidance in Target Populations. IEEE Trans Biomed Eng. In press. doi: 10.1109/TBME.2024.3352153.

25.

Holst

Buse

Rodbard

, et al. IDegLira improves both fasting and postprandial glucose control as demonstrated using continuous glucose monitoring and a standardized meal test. J Diabetes Sci Technol. 2015;10(2):389-397.

26.

Uzunlulu

Oguz

Arslan Bahadir

Erbakan

Vural Keskinler

Alpaslan Mesci

. C-peptide concentrations in patients with type 2 diabetes treated with insulin. Diabetes Metab Syndr. 2019;13(6):3099-3104.

27.

Møller

Dalla Man

Overgaard

, et al. Ethnic differences in insulin sensitivity, β-cell function, and hepatic extraction between Japanese and Caucasians: a minimal model analysis. J Clin Endocrinol Metab. 2014;99(11):4273-4280.

28.

Visentin

Dalla Man

Cobelli

. One-day Bayesian cloning of type 1 diabetes subjects: toward a single-Day UVA/Padova type 1 diabetes simulator. IEEE Trans Biomed Eng. 2016;63(11):2416-2424.

29.

Glycemic targets: standards of medical care in diabetes—2020. Diabetes Care. 2020;43(suppl 1):S66-S76.

30.

Riddle

Bolli

Ziemen

, et al. New insulin glargine 300 Units/mL versus glargine 100 Units/mL in people with type 2 diabetes using basal and mealtime insulin: glucose control and hypoglycemia in a 6-month randomized controlled trial (EDITION 1). Diabetes Care. 2014;37(10):2755-2762.

31.

Brazeau

Mircescu

Desjardins

, et al. Carbohydrate counting accuracy and blood glucose variability in adults with type 1 diabetes. Diabetes Res Clin Pract. 2013;99(1):19-23.

32.

Battelino

Danne

Bergenstal

, et al. Clinical targets for continuous glucose monitoring data interpretation: recommendations from the international consensus on time in range. Diabetes Care. 2019;42(8):1593-1603.

33.

Bergenstal

Beck

, et al. Glucose Management Indicator (GMI): a new term for estimating A1C from continuous glucose monitoring. Diabetes Care. 2018;41:2275-2280.

34.

Wallia

Molitch

. Insulin therapy for type 2 diabetes mellitus. JAMA. 2014;311(22):2315.

35.

Polonsky

Henry

. Poor medication adherence in type 2 diabetes: recognizing the scope of the problem and its key contributors. Patient Prefer Adherence. 2016;10:1299-1307.

36.

Visentin

Campos-Náñez

Schiavon

, et al. The UVA/Padova type 1 diabetes simulator goes from single meal to single day. J Diabetes Sci Technol. 2018;12(2):273-281.

37.

Sieber

Wienbarg

Mahoney

Haak

. 528-P: different glycemic responses in people with T1 or T2DM after use of an insulin injection supporting app. Diabetes. 2023;72(suppl 1):528.