We investigated the processes underlying glycemic deterioration in type 2 diabetes (T2D).
A total of 732 recently diagnosed patients with T2D from the Innovative Medicines Initiative Diabetes Research on Patient Stratification (IMI DIRECT) study were extensively phenotyped over 3 years, including measures of insulin sensitivity (OGIS), β-cell glucose sensitivity (GS), and insulin clearance (CLIm) from mixed meal tests, liver enzymes, lipid profiles, and baseline regional fat from MRI. The associations between the longitudinal metabolic patterns and HbA1c deterioration, adjusted for changes in BMI and in diabetes medications, were assessed via stepwise multivariable linear and logistic regression.
Faster HbA1c progression was independently associated with faster deterioration of OGIS and GS and increasing CLIm; visceral or liver fat, HDL-cholesterol, and triglycerides had further independent, though weaker, roles (R2 = 0.38). A subgroup of patients with a markedly higher progression rate (fast progressors) was clearly distinguishable considering these variables only (discrimination capacity from area under the receiver operating characteristic = 0.94). The proportion of fast progressors was reduced from 56% to 8–10% in subgroups in which only one trait among OGIS, GS, and CLIm was relatively stable (odds ratios 0.07–0.09). T2D polygenic risk score and baseline pancreatic fat, glucagon-like peptide 1, glucagon, diet, and physical activity did not show an independent role.
Deteriorating insulin sensitivity and β-cell function, increasing insulin clearance, high visceral or liver fat, and worsening of the lipid profile are the crucial factors mediating glycemic deterioration of patients with T2D in the initial phase of the disease. Stabilization of a single trait among insulin sensitivity, β-cell function, and insulin clearance may be relevant to prevent progression.
Introduction
Maintaining glucose levels within appropriate limits in patients with type 2 diabetes (T2D) is a crucial factor to prevent complications. Effective strategies to slow glycemic progression can be supported by understanding the processes underlying deterioration of glucose control.
Few studies have assessed HbA1c trajectories and the possible determinants of glycemic deterioration. An established finding is that β-cell function decline is an important factor (1,2), while contradictory conclusions were drawn for insulin sensitivity (1,3–7). Whether heterogeneous patterns between patients exist in β-cell function and insulin sensitivity decline has not been clarified, an important question for patient stratification and personalized medicine. Other limitations of previous analyses include the incomplete characterization of the metabolic parameters affecting glucose homeostasis (derived using fasting data only [2,4]), the restricted set of traits investigated together, and the lack of potentially relevant measures such as ectopic fat, insulin clearance, or lifestyle. No study has assessed the relationships between the longitudinal trajectories of HbA1c and those of the other metabolic traits.
In this analysis, we have used data from the cohort of recently diagnosed and extensively phenotyped patients with T2D of the Diabetes Research on Patient Stratification (DIRECT) study (8,9) to elucidate the processes underlying glycemic deterioration. Specific features of the DIRECT study are the detailed assessment of the glucose homeostasis parameters and patients all being in the initial phase of the disease. We determined the patterns over a 3-year period of HbA1c, β-cell function, insulin sensitivity, and other relevant laboratory, clinical, and functional parameters and assessed their relevance in the deterioration of glucose control.
Research Design and Methods
Subjects and Protocol
The Innovative Medicines Initiative (IMI) DIRECT project is a multicenter prospective study on northern European adults (8,9) (ClinicalTrials.gov identifier NCT03814915). The present analysis considers the DIRECT cohort of recently diagnosed patients with T2D, who were recruited according to the following criteria: White race, T2D diagnosis according to the American Diabetes Association 2011 criteria (10) not <6 months and not >24 months before baseline examination, previous treatment via lifestyle measures with or without metformin therapy, age between 35 and 74 years, BMI between 20 and 50 kg/m2, estimated glomerular filtration rate >50 mL/min, and HbA1c concentration <7.64% (60.0 mmol/mol) within the previous 3 months. Participants were studied at baseline (month 0) and at months 9, 18, and 36. Subjects with HbA1c available at least in two visits were included in this analysis (N = 750).
All participants provided written informed consent, and the study protocol was approved by the regional research ethics review boards. The research conformed to the ethical principles for medical research involving human participants outlined in the Declaration of Helsinki.
Collected Data
Anthropometric data, HbA1c, blood lipids, and liver enzymes were collected at all visits. A 27-month HbA1c sample was collected in 39 patients. A standardized mixed-meal tolerance test (MMTT) (8) was performed at months 0, 18, and 36 to calculate indices of insulin sensitivity (in fasting conditions, QUICKI [11], and post-MMTT, oral glucose insulin sensitivity [OGIS] [12]), β-cell function (13) (β-cell glucose sensitivity [GS] and rate sensitivity), and insulin clearance (in fasting conditions and post-MMTT [CLIm]). From the baseline visit, we collected glucagon, proinsulin, and glucagon-like peptide 1 (GLP-1), measures of regional fat from MRI (8) (available in 561 participants), of physical activity from accelerometer (8), and of self-reported 24-h nutrient intake (8), and we computed the fatty liver index (14) and a T2D polygenic risk score (15). The whole set of traits considered in this study is described in detail in the Supplementary Data, Supplementary Methods, and Supplementary Table 2.
Assessment of Progression Rates
We computed the progression rates for HbA1c and several traits available at follow-up (Supplementary Table 4). Each trajectory was described with a conditional linear mixed-effect model (16), in which the longitudinal component of the data was described as a proportional function of time, with normally distributed slopes describing individual progression rates. HbA1c progression was adjusted for changes in BMI and diabetes medications, which were recorded at all visits (as dosage and start and end of treatment). The adjustments were assumed to be: 1) proportional to BMI; 2) linearly related to the metformin dose, expressed as percentage of a maximal dose of 3 g; 3) linearly related to the cumulative dose for the other antidiabetic drugs (insulin excluded), expressed as the sum of the percentages of the maximum dose of each drug; and 4) constant under insulin treatment. A proportional effect of delay in HbA1c assay (i.e., of the difference between the time of measurement and the time of sample collection) was also introduced. Medications were considered to be effective if taken at least 30 days before HbA1c measurement. OGIS and QUICKI trajectories were adjusted for changes in BMI. Further details about the conditional linear mixed-effect models are provided in the Supplementary Methods.
Statistical Analysis
Results are presented for participants (N = 732) with GAD <11 units/mL and islet antigen 2 antibodies <7.5 units/mL to exclude other possible forms of diabetes (17). Distributions are described as mean ± SD. Pairwise associations between continuous variables were assessed using the Spearman correlation coefficient; differences between groups were assessed using the Wilcoxon signed-rank test (for two groups) and Kruskal-Wallis test (for three or more groups).
We used stepwise multivariable linear regression to determine the set of variables, as baseline values (Supplementary Table 2) and progression rates (Supplementary Table 4), independently associated with the HbA1c progression rate, with adjustment for center, sex, and age. For baseline variables, both untransformed and transformed values were considered; transformations were logarithmic, or logit when variables were constrained within an interval. The independent variables were included in the regression model when their effects had P < 0.05 and produced an increment in the adjusted R2 value. Two stepwise analyses were performed: one on all participants, excluding MRI variables from the analysis, and one on the subset of participants with MRI data, including these data in the analysis. Standardized coefficients were computed per SD of the underlying data distribution.
Since the distribution of HbA1c progression rates was skewed to the right with a group of patients with high values, we split the subjects into average and fast progressors according to a progression rate threshold (see Results). We used multivariable logistic regression to assess the odds ratios of average versus fast progression, using the independent variables identified in the multiple linear regression analysis of HbA1c progression. The logistic analysis provided values for area under the receiver operating characteristic, sensitivity, specificity, and accuracy to be used as measures of the discrimination capacity of the investigated independent variables over fast versus average progressors. These parameters must not be interpreted as measures of predictive capacity.
Role of the Funding Source
The funders had no role in study design; collection, analysis, and interpretation of data; writing of the report; or the decision to submit the paper for publication. The corresponding author had full access to all data and had final responsibility for the decision to submit for publication.
Results
Subjects’ Baseline Characteristics
At baseline, the participants were aged 62 ± 8 years, were moderately obese (30.4 ± 4.9 kg/m2 BMI), and had an HbA1c of 6.41 ± 0.53% (46.5 ± 5.8 mmol/mol) and fasting glucose of 7.1 ± 1.4 mmol/L (Supplementary Table 2). A total of 34% of the subjects were treated with metformin at baseline; the rest were treatment naive.
Progression Rates of HbA1c and Other Traits
The individual HbA1c progression rates (Supplementary Fig. 1), adjusted for changes in BMI and in diabetes medications, were on average only slightly positive and mostly distributed close to their median (median, first, and ninth deciles were 0.041, −0.038, and 0.185%/year [0.45, −0.41, and 2.02 mmol mol−1 year−1], respectively). However, the distribution showed a heavy right tail with values up to 0.897%/year (9.8 mmol mol−1 year−1). The adjustment of progression rates for BMI changes implied a standardized coefficient for the BMI effect of 0.37.
All of the other investigated traits had a mean progression rate per year smaller, in absolute value, than 5% of the corresponding baseline average (see Supplementary Table 5 for details). On average, waist circumference, but not BMI, increased very slightly. Insulin sensitivity (as OGIS) and most of the β-cell function parameters decreased. Fasting, but not postmeal, insulin clearance decreased. Total cholesterol did not change, while its fractions showed opposite changes, with HDL increasing and LDL decreasing; fasting triacylglycerol (TG) increased. Creatinine and ALT did not change, while AST and AST/ALT increased.
Several pairwise associations were observed between HbA1c progression rate and laboratory, clinical, and functional parameters (Supplementary Fig. 2). In particular, HbA1c progression rate was clearly associated (P < 0.01) with some baseline traits (positively with BMI, waist circumference, triglycerides, glucagon, and liver and visceral fat; inversely with age, HDL, insulin sensitivity, and β-cell function) and some progression rates (positively with those of triglycerides and liver enzymes; inversely with those of insulin sensitivity, β-cell function, AST/ALT ratio, and HDL).
Several pairwise associations were also observed between the progression rates of the investigated traits (Supplementary Fig. 2B). GS and OGIS progression rates were independent of one another despite HbA1c progression rate being associated with both.
Variables Associated With HbA1c Progression Rate: Multivariable Linear Analysis
In multivariable linear analysis of HbA1c progression rate in all patients, the baseline values and the progression rates of several traits provided an independent contribution (adjusted R2 = 0.38) (Fig. 1A). Faster HbA1c progression was independently associated with lower baseline values and faster deterioration of insulin sensitivity (as OGIS) and β-cell function (mostly as GS), with higher baseline values of MMTT insulin clearance, CLIm, and with its increase (all P values <0.001). Faster HbA1c progression was also independently associated with lower baseline HDL (P < 0.05) or its slower increase (P < 0.001), with a quicker increase of TG (P < 0.001), as well as with higher baseline values of BMI (P < 0.01) and lower baseline values of HbA1c (P < 0.001). The variables with strongest effects were the baseline OGIS value and the progression rates of OGIS, GS, and CLIm (standardized coefficients, in absolute value, between 0.24 and 0.57).
Variables independently associated with HbA1c progression rate from multivariable linear analysis. A: All subjects are included in the analysis (625 with all variables), and MRI measurements are not considered. B: Only subjects with MRI are included in the analysis (374 with all variables), and MRI measurements are taken into consideration. For each variable, the figure shows the standardized coefficients ± 95% CI of the effect. Age and HDL were log-transformed. *P < 0.05; **P < 0.01; ***P < 0.001. bas, baseline value; progr, progression rate; RS, β-cell rate sensitivity.
Variables independently associated with HbA1c progression rate from multivariable linear analysis. A: All subjects are included in the analysis (625 with all variables), and MRI measurements are not considered. B: Only subjects with MRI are included in the analysis (374 with all variables), and MRI measurements are taken into consideration. For each variable, the figure shows the standardized coefficients ± 95% CI of the effect. Age and HDL were log-transformed. *P < 0.05; **P < 0.01; ***P < 0.001. bas, baseline value; progr, progression rate; RS, β-cell rate sensitivity.
In multivariable analysis of the subset of patients with baseline MRI measurements (adjusted R2 = 0.40) (Fig. 1B), baseline visceral fat was positively and independently correlated with HbA1c progression rate; moreover, female sex and younger age independently predicted faster HbA1c progression. The role of the other key metabolic parameters—OGIS, GS, and CLIm—remained similar. Replacing visceral fat with liver fat produced similar results (standardized coefficient equal to 0.15 for visceral fat and to 0.11 for liver fat); when both visceral and liver fat were included in the model, the latter was not independently associated with HbA1c progression.
No independent effects were detected for smoking status, family history, T2D polygenic risk score, baseline values of diet, physical activity, pancreatic fat, GLP-1 (total and intact at fasting and total at 60 min), glucagon, and 60-min proinsulin, baseline values and progression rates of AST and ALT.
Further details on the multivariable linear analysis are reported in the Supplementary Results.
Variables Associated With HbA1c Progression Rate: Multivariable Logistic Analysis
The threshold selected to separate the heavy right tail of the distribution of HbA1c progression rates was 0.255%/year (2.79 mmol mol−1 year−1). This threshold split the subjects into average progressors (N = 699), with a progression rate of 0.044 ± 0.076%/year (0.48 ± 0.83 mmol mol−1 year−1), and fast progressors (N = 33), with an ∼10-fold mean progression rate (0.460 ± 0.185%/year or 5.03 ± 2.02 mmol mol−1 year−1) (Fig. 2).
Temporal trajectories or baseline values (bar graphs) of HbA1c and other key traits in fast (red lines) and average (blue lines) progressors. Data are mean ± SE. Simple comparisons between fast and average progressors (Wilcoxon rank-sum test) are shown for baseline values (asterisks at month 0) and progression rates (asterisks at month 18). These comparisons may differ from the results of the multivariable analyses (Figs. 1 and 3). Sex is not included in the figure: males were 42% and 36% in average and fast progressors, respectively (nonsignificant, χ2 test). HbA1c values at 27 months are not displayed, as they were collected in a subgroup of individuals. In average progressors, HbA1c increases from 46.4 ± 0.2 to 46.7 ± 0.3 mmol/mol and in fast progressors, from 48.9 ± 1.21 to 75.7 ± 2.5 mmol/mol. *P < 0.05; **P < 0.01; ***P < 0.001. bas, baseline value; ISRtot, total mixed-meal test insulin secretion; RS, β-cell rate sensitivity.
Temporal trajectories or baseline values (bar graphs) of HbA1c and other key traits in fast (red lines) and average (blue lines) progressors. Data are mean ± SE. Simple comparisons between fast and average progressors (Wilcoxon rank-sum test) are shown for baseline values (asterisks at month 0) and progression rates (asterisks at month 18). These comparisons may differ from the results of the multivariable analyses (Figs. 1 and 3). Sex is not included in the figure: males were 42% and 36% in average and fast progressors, respectively (nonsignificant, χ2 test). HbA1c values at 27 months are not displayed, as they were collected in a subgroup of individuals. In average progressors, HbA1c increases from 46.4 ± 0.2 to 46.7 ± 0.3 mmol/mol and in fast progressors, from 48.9 ± 1.21 to 75.7 ± 2.5 mmol/mol. *P < 0.05; **P < 0.01; ***P < 0.001. bas, baseline value; ISRtot, total mixed-meal test insulin secretion; RS, β-cell rate sensitivity.
We found that the trajectories of most variables independently affecting HbA1c progression as from the linear analysis were clearly different (P < 0.001) in the two groups (Fig. 2): in fast progressors, OGIS and GS strongly declined and TG and CLIm markedly increased. At baseline, fast progressors had lower OGIS (P < 0.05), CLIm (P < 0.01), and HDL (P < 0.001) and higher BMI (P < 0.01).
Logistic analysis substantially confirmed the results of linear regression (Fig. 1), with half of the investigated variables still contributing (P < 0.05) to distinguish average and fast progressors (Fig. 3): fast HbA1c progression independently associated with stronger deterioration and a lower baseline value of OGIS and GS, CLIm increase, and HDL reduction. The discrimination capacity of the logistic model, computed as area under the receiver operating characteristic, was 0.94 (95% CI between 0.86 and 0.98).
Odds ratios ± 95% CI from the multivariable logistic analysis of fast vs. average HbA1c progressors. The independent variables are those identified by multivariable linear analysis of HbA1c progression, excluding MRI variables (N = 625, with 32 fast progressors and 593 average progressors). Age and HDL were log-transformed. Values for sensitivity, specificity, and accuracy were derived via maximization of balanced accuracy. *P < 0.05; **P < 0.01; ***P < 0.001. AUROC, area under the receiver operating characteristic; bas, baseline value; progr, progression rate; RS, β-cell rate sensitivity.
Odds ratios ± 95% CI from the multivariable logistic analysis of fast vs. average HbA1c progressors. The independent variables are those identified by multivariable linear analysis of HbA1c progression, excluding MRI variables (N = 625, with 32 fast progressors and 593 average progressors). Age and HDL were log-transformed. Values for sensitivity, specificity, and accuracy were derived via maximization of balanced accuracy. *P < 0.05; **P < 0.01; ***P < 0.001. AUROC, area under the receiver operating characteristic; bas, baseline value; progr, progression rate; RS, β-cell rate sensitivity.
Similar outcomes were obtained using lower HbA1c progression rate thresholds, which resulted in larger numbers of patients classified as fast progressors (Supplementary Results and Supplementary Figs. 1 and 3).
At baseline, the percentage of patients treated with metformin were not different between fast progressors (39.4% [95% CI 24.7–56.3%]) and average progressors (33.9% [30.5–37.5%]; P = 0.64). At the last visit, the percentage of patients treated with any diabetes medication was somewhat higher in fast progressors, as expected (P = 0.048; details provided in the Supplementary Results). Only seven average progressors were on insulin at the last visit.
Impact of Stable OGIS, GS, or CLIm on Proportion of Fast HbA1c Progressors
Because HbA1c progression was associated with worsening of three main factors—OGIS, GS, and CLIm—we have evaluated the possible importance of maintaining one of these key traits relatively stable in order to avoid fast progression. For this purpose, we considered each trait as deteriorating if its progression rate fell within its worst tertile (the bottom tertile for OGIS and GS and the top tertile for CLIm) and as stable if it fell in the other two tertiles. We examined the subgroups of patients in which none or only one of these key traits was relatively stable (Table 1).
Proportion of fast HbA1c progressors with different combinations of stable/deteriorating conditions for GS, OGIS, and CLIm progression rates
Condition* . | Average progressors (N) . | Fast progressors (N) . | Fast progressors (%) (95% CI) . | Odds ratio (95% CI) . | P value† . | ||
---|---|---|---|---|---|---|---|
GS . | OGIS . | CLIm . | |||||
Deteriorating | Deteriorating | Stable | 47 | 5 | 9.6 (4.2–20.6) | 0.09 (0.02–0.32) | 2E−4 |
Deteriorating | Stable | Deteriorating | 56 | 6 | 9.7 (4.5–19.5) | 0.09 (0.02–0.30) | 8E−5 |
Stable | Deteriorating | Deteriorating | 34 | 3 | 8.1 (2.8–21.3) | 0.07 (0.02–0.32) | 4E−4 |
Deteriorating | Deteriorating | Deteriorating | 8 | 10 | 55.6 (33.7–75.4) | — | — |
Condition* . | Average progressors (N) . | Fast progressors (N) . | Fast progressors (%) (95% CI) . | Odds ratio (95% CI) . | P value† . | ||
---|---|---|---|---|---|---|---|
GS . | OGIS . | CLIm . | |||||
Deteriorating | Deteriorating | Stable | 47 | 5 | 9.6 (4.2–20.6) | 0.09 (0.02–0.32) | 2E−4 |
Deteriorating | Stable | Deteriorating | 56 | 6 | 9.7 (4.5–19.5) | 0.09 (0.02–0.30) | 8E−5 |
Stable | Deteriorating | Deteriorating | 34 | 3 | 8.1 (2.8–21.3) | 0.07 (0.02–0.32) | 4E−4 |
Deteriorating | Deteriorating | Deteriorating | 8 | 10 | 55.6 (33.7–75.4) | — | — |
The progression rate thresholds dividing stable and deteriorating traits for OGIS, GS, and CLIm are −16.68 mL min−1 m−2 year−1, −4.07 pmol min−1 m−2 mmol−1 L year−1, and 0.0184 L min−1 m−2 year−1, respectively.
Two-sided χ2 test (α = 0.05), with Yates continuity correction, on the proportion of fast progressors in the row compared with the same proportion in the last row.
We found that the proportion of fast progressors was 56% in the patient subgroup in whom GS, OGIS, and CLIm were all deteriorating and decreased to 8–10% in the subgroups in which a single trait—GS, OGIS, or CLIm—was stable. All proportions were different from 0 at 90% confidence level, stressing that fast progression did not imply quick changes for each of the three considered traits. All differences in proportions (one stable trait vs. none) had P < 0.001 and were associated to odds ratio for fast versus average progression <0.1 (Table 1); thus, relatively stable progression rate of one single trait among GS, OGIS, and CLIm was strongly associated to reduced glycemic deterioration.
Conclusions
Leveraging on the detailed participant characterization of the DIRECT study, we have been able to elucidate the processes underlying glycemic deterioration in patients with T2D in the initial phase of the disease. We found that HbA1c deterioration was independently associated with: 1) a decrease in insulin sensitivity; 2) a decrease in β-cell function (primarily β-cell glucose sensitivity); 3) an increase in insulin clearance; and 4) lower values of insulin sensitivity and glucose sensitivity and higher values of insulin clearance at baseline. Further variables independently associated with faster HbA1c progression were declining HDL, increasing TG, and high baseline visceral or liver fat.
The variables identified by multivariable linear analysis also explained the rapid HbA1c deterioration detected in a subset of patients (identified as fast progressors), the strongest predicting variables of the multivariable linear model being significant also with logistic analysis. Clear differences were evident between fast and average HbA1c progressors (Fig. 2), consistent with the associations derived from the multivariable linear analysis. The high discrimination capacity of the logistic analysis suggests that the selected variables capture the most relevant pathophysiological factors underlying glycemic deterioration.
The independent associations with HbA1c progression of several variables, in particular the progression rates of insulin sensitivity, β-cell function, and insulin clearance, and the existence of fast HbA1c progressors with relatively stable conditions for any of these three traits (Table 1) indicate that: 1) the processes of glycemic deterioration are heterogeneous in this population of patients with T2D; and 2) fast progression does not imply quick deterioration of a specific trait (e.g., insulin sensitivity or β-cell function).
The dichotomous analysis shows that the odds for fast versus average progression are substantially reduced when glucose sensitivity, insulin sensitivity, or insulin clearance is relatively stable. Although these findings do not demonstrate causality, they suggest that preventing either high degradation rates of glucose sensitivity or insulin sensitivity, or high increase rates of insulin clearance, may be an effective strategy to slow down glycemic deterioration in the initial phase of the disease. This reemphasizes the importance of lifestyle interventions aiming at controlling insulin resistance, as preventing deterioration of the other traits currently appears more difficult.
This study also shows that insulin resistance plays a major role in glycemic deterioration in these patients with T2D. In particular, we show associations of glycemic deterioration with baseline insulin sensitivity and its longitudinal change that the Belfast Diet Study (1), UK Prospective Diabetes Study (UKPDS) (4,18), and A Diabetes Outcome Progression Trial (ADOPT) (6) could not identify, possibly due to differences in subject selection or to the use of post-MMTT versus fasting insulin sensitivity indices. We also demonstrate that the associations between glycemic deterioration and insulin sensitivity are independent from both the baseline value and the progression rate of the β-cell function and that insulin resistance progresses independently from β-cell glucose sensitivity. Since, in our analysis, both HbA1c and insulin sensitivity trajectories were adjusted for BMI changes and BMI did not increase on average, we can conclude that worsening of insulin resistance in T2D and the associated glycemic deterioration are partly independent from BMI changes. Whether the observed average increases in TG and AST (for which progression rates were inversely correlated with OGIS progression rate) have a role in insulin sensitivity deterioration (19) and whether this is mediated by ectopic fat accumulation (20) deserve further study.
UKPDS 25 and 26 (4,18), the Belfast Diet Study (1), and the ADOPT study (6) identified baseline HOMA of β-cell function as a predictor of glycemic deterioration (insulin requirement within 6 years for UKPDS, time of failure to dietary therapy for the Belfast Diet Study, and monotherapy failure before 4 years for ADOPT). Our study confirms the role of β-cell dysfunction as driver of glycemic deterioration using a dynamic β-cell function assessment based on a glucose challenge, rather than on fasting data only. We show that both baseline β-cell dysfunction (especially β-cell glucose sensitivity) and its deterioration over time are independently associated with HbA1c worsening. Moreover, we demonstrate that patients with limited or absent deterioration in β-cell function have considerably lower odds of rapid glycemic deterioration.
Another novel finding is the strong and independent association between HbA1c progression and insulin clearance during the MMTT, CLIm. To our knowledge, this is the first study examining insulin clearance trajectories after T2D onset. We found that higher baseline CLIm and faster CLIm increase over time independently associate with faster HbA1c progression. This is consistent with the glucose homeostasis mechanisms, as higher CLIm reduces the average insulin levels. Notably, we found a positive correlation between insulin sensitivity and insulin clearance, considering both the baseline values of the two traits, in agreement with previous findings (21), and their progression rates (Supplementary Fig. 2). However, on average, in spite of a decrease in insulin sensitivity, insulin clearance did not decrease. These findings show that, while in subjects with prediabetes, insulin clearance reduction may be a way to mitigate the effects of insulin resistance (22), in patients with T2D, this compensation appears present but impaired and contributing to glycemic deterioration. The reasons underlying these results remain elusive. The lack of decrease in insulin clearance may be explained by the decrease of total MMTT insulin secretion and consequent desaturation of insulin utilization (23) only in fast progressors, as in average progressors, total insulin secretion slightly increased (Fig. 2). Whether hepatic or extrahepatic mechanisms underlie these findings cannot be determined from this study and deserves further investigation.
Our results on TG and HDL effects were partially anticipated by a study of the Genetics of Diabetes Audit and Research in Tayside Scotland (GoDARTS) (24), in which the outcome was the risk of progression to insulin treatment. The study identified baseline TG and HDL (besides BMI, sex, and age, year, and HbA1c at diagnosis) as independent determinants. A later study on the same data (25), investigating the baseline determinants of HbA1c progression rate over ∼9 years, confirmed an independent effect of HDL (together with age, BMI, and year at diagnosis) but not of TG. The 5-year Fenofibrate Intervention and Event Lowering in Diabetes (FIELD) study in patients with T2D on lifestyle measures only revealed that the HDL effect on initiation of oral hypoglycemic agents survives the adjustment for HOMA of insulin resistance (26). Compared with previous studies (24–26), our analysis includes the progression rates of plasma lipid components and baseline MRI assessment of regional fat. We show that baseline HDL and BMI and the progression rates of TG and HDL are associated with HbA1c progression, even after accounting for the effects of the three main determinants of glucose homeostasis (i.e., insulin sensitivity, β-cell function, and insulin clearance). In the subset of participants with MRI data, baseline visceral fat or liver fat was independently correlated with HbA1c progression rate, a further novel observation. These findings suggest that additional lipid-dependent factors contribute to HbA1c deterioration, possible candidates being fat accumulation in the viscera (with excessive supply of fatty acids to the liver [27]), liver fat and consequent hepatic insulin resistance (28), or glucose overproduction (29). The role of visceral/liver fat supports interventions to reduce ectopic fat as a possible way for slowing future glycemic progression.
Previous studies have reported an inverse correlation between baseline age and HbA1c progression (1,4,6,24,25,30). In our analysis, baseline age does not have a clear independent role in the multivariable model, most likely because the age range is relatively narrow relative to other studies or because the stronger predictors of HbA1c progression are correlated with age. The latter explanation would suggest that the age univariate effect on glycemic deterioration is indirect. We do not find a clear sex effect in glycemic deterioration, in agreement with most previous studies (1,4,6,24,25).
In the multivariable model, baseline HbA1c was independently and inversely correlated with HbA1c progression rate, in contrast with previous findings (1,4,6,24,30). However, baseline HbA1c was not significant in the logistic model. The most likely explanation of this finding is regression to the mean: indeed, a random decrease in baseline HbA1c can produce a higher estimate of HbA1c progression rate, particularly when the follow-up period is not long, as in our study. Tight glycemic control, an inclusion criterion, may have enhanced this effect.
This study does not find a relevant role of other variables often associated with glucose control. In particular, we did not find an effect of smoking status (reported in the General Practice Research Database [30]), T2D polygenic risk score (in agreement with GoDARTS [24]), and baseline values of diet, physical activity, pancreatic fat, GLP-1, and glucagon. Several of these variables were not associated with HbA1c progression rate even in simple correlation analysis (Supplementary Fig. 2). The lack of association for pancreatic fat is particularly relevant and contributes to the ongoing discussion on the role of pancreas fat in T2D management (31).
In spite of the unique extensive phenotyping of our study and the consistent results, a significant limitation is the relatively short follow-up period (3 years). The accuracy of the estimated HbA1c progression rate over this time frame may be limited, and in a longer time period, the factors contributing to progression may differ. In this study, we could not assess the changes over time of relevant variables such as regional fat by MRI, diet, and physical activity. MRI measurements were available only for a subset of subjects. Insulin sensitivity was not derived from the gold-standard euglycemic clamp. As the cohort included only patients of White race, our findings are not generalizable to other racial/ethnic groups. Causal relationships could not be inferred from our regression analyses. The study of the mechanisms underlying the deterioration of the factors affecting HbA1c progression, an important aspect to envisage optimal treatment strategies, also requires further investigation.
In summary, based on the extensively phenotyped cohort of White European patients with diabetes of the DIRECT study, we identified decreasing insulin sensitivity, deteriorating β-cell function, increasing insulin clearance, high liver or visceral fat, and worsening of the lipid profile as the most important factors independently associated with HbA1c deterioration in the early phase of the disease. We also showed that patients with a relatively stable value over time of at least one of insulin sensitivity, β-cell glucose sensitivity, or insulin clearance have considerably reduced odds of fast HbA1c increase. This study contributes to the understanding of the factors underlying diabetes progression, elucidating the processes that might be targeted for personalized treatments.
M.I.M. and A.M. are currently affiliated with Genentech, South San Francisco, CA.
This article contains supplementary material online at https://doi.org/10.2337/figshare.13222205.
Article Information
Acknowledgments. The authors thank the participants across all IMI DIRECT study centers for the contributions to the study. The authors also thank the staff involved in the design, implementation, and conduction of the study.
Funding. The work leading to this publication has received support from the Innovative Medicines Initiative Joint Undertaking under grant agreement 115317 (DIRECT), resources of which are composed of financial contribution from the European Union’s Seventh Framework Programme (FP7/2007-2013) and European Federation of Pharmaceutical Industries and Associations companies’ in-kind contribution. Information on the project can be found at https://www.direct-diabetes.org/.
Duality of Interest. C.J. reports grants from Innovative Medicines Initiative European Union during the conduct of the study and personal fees from Novo Nordisk, Sanofi, AstraZeneca, and Boehringer Ingelheim outside the submitted work. L.M.’t.H. reports grants from IMI Joint Undertaking during the conduct of the study. J.W.B. reports grants from IMI European Union during the conduct of the study. P.B.M. is an employee of Sanofi Deutschland GmbH. M.R. is an employee of Novo Nordisk A/S. S.B. reports personal fees from Intomics A/S and Proscion A/S outside the submitted work. As of January 2020, A.M. is an employee of Genentech and a holder of Roche stock. H.R. is an employee and shareholder of Sanofi. M.I.M. reports employment and stock from Genentech; grants and personal fees from Merck, Novo Nordisk, Eli Lilly and Company, and Pfizer; grants from Roche, Servier, Sanofi Aventis, AbbVie Inc., AstraZeneca, Boehringer Ingelheim, Janssen, and Takeda outside the submitted work. J.M.S. reports grants from KTH during the conduct of the study. P.W.F. reports research funding from Boehringer Ingelheim, Eli Lilly and Company, Janssen, Novo Nordisk A/S, Sanofi Aventis, and Servier; consulting fees from Eli Lilly and Company, Novo Nordisk, and Zoe Global Ltd; and stock options in Zoe Global Ltd. No other potential conflicts of interest relevant to this article were reported.
The funders had no role in study design; collection, analysis, and interpretation of data; writing of the report; or the decision to submit the paper for publication.
Author Contributions. R.B. and A.M. designed the analysis, analyzed the data, and wrote the manuscript. R.B., C.J., A.G.J., M.W., E.R.P., and A.M. interpreted the results. E.R.P. and A.M. supervised the analysis. C.J., A.G.J., A.K., M.W., and E.R.P. reviewed the manuscript. All authors were involved in the DIRECT study at different levels and were essential for the production, release, and management of the data analyzed in this article. R.B. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Prior Presentation. This work was presented at the 55th Annual Meeting of the European Association for the Study of Diabetes, Barcelona, Spain, 16–20 September 2019.