The effects of basal insulin peglispro vs. insulin glargine on lipoprotein particles by NMR and liver fat content by MRI in patients with diabetes

Patients and studies

The lipid/MRI substudy was conducted as protocol addenda to four Phase 3 clinical studies of BIL compared to glargine in patients with type 1 or type 2 diabetes [13–16]. Three patient cohorts were represented (type 1 diabetes [IMAGINE 1 and IMAGINE 3], type 2 diabetes previously taking insulin (and then randomized to either glargine or BIL, described as the “basal switch” cohort [IMAGINE 5]), and type 2 diabetes insulin naïve [IMAGINE 2]). All four studies were multinational and randomized; IMAGINE 2 and IMAGINE 3 were double-blind. Common features of the Phase 3 studies included intensive insulin adjustment with similar basal (and bolus in type 1 diabetes) insulin adjustment algorithms, patient inclusion and exclusion criteria, no changes in lipid-altering medications up to week 12, and stable doses of background glucose-lowering medications before and during the study. Exclusion criteria included fasting hypertriglyceridemia (defined as serum TGs >400 mg/dL). In addition, if a patient developed elevated fasting TG >600 mg/dL at any time during the trial, the patient was discontinued from the study treatment.

Laboratory methods

For all of the analyses in the lipid substudy, blood was collected after an approximately ≥10-h fast. Lipoprotein particle concentrations and sizes were measured at Liposcience (now Laboratory Corporation of America Holdings, Morrisville, NC, USA) by NMR LipoProfile analysis using the LP2 algorithm as previously described [9]. The liver fat content was measured by MRI as previously described [11]. Global CEC was performed by Vascular Strategies LLC, Plymouth Meeting, PA, USA as previously described [17, 18]. CETP mass (ELISA) and activity (fluorometric assay) were determined by Pacific Biomarkers, Seattle, WA, USA. Apo A1, Apo B and Apo B48 were determined by nephelometry by Siemens Healthcare GmgH, Erlangen, Germany. Apo-B100 was calculated as the difference between Apo-B and Apo-B48. The following assays were performed by Covance, Princeton, NJ, USA: serum free fatty acids (enzymatic), Apo A2 and Apo C3 (turbidometric immunoassay), and adiponectin (Quantikine Human Adiponectin/Acrp30 Immunoassay, R&D Systems, Minneapolis, MN).

Liver fat content (LFC) measurements

LFC was evaluated by MRI. For each subject, all scans were obtained using the same scanning techniques, equipment, and imaging parameters as at baseline. The entire liver was scanned with 3–5 overlapping series in the axial plane. LFC was estimated using six-echo images with spectral model and T2* correction, at sequential alternating out-of-phase and in-phase echo times, as previously described [10]. To ensure consistent interpretation of scans, phantoms containing liquid of varying fat fractions were used for quality control, and study images were centrally evaluated by a qualified vendor (Virtual Scopics Inc. Rochester, NY, USA).

Statistical methods

Statistical software used was SAS version 9.1 or higher (SAS Institute Inc. Cary, NC, USA). Analyses were conducted on all randomized patients in the lipid substudy who took at least one dose of study insulin. A mixed-model repeated measures (MMRM) model was used to analyze continuous variables collected at multiple post-treatment time points with terms for treatment, baseline values of the analysis variables, stratification factors for randomization, week, and treatment by week interactions. Values are presented as least squares mean (LSM) ± standard error (SE) unless otherwise noted. All treatment differences are reported as LSM difference (BIL-glargine) with 95% confidence intervals (CI). For treatment comparison at baseline, an analysis of variance model was used for continuous variables and Fisher’s exact test for categorical outcomes. Spearman’s correlation analyses were performed to assess the relationships between NMR parameters and LFC. To adjust for multiplicity, statistical significance is defined as two-sided p value <0.001. Raw data was analyzed as it was collected without transformation or exclusion of outliers. Missing data was handled through MMRM analysis without explicit imputation.

Patients

Type 1 diabetes—effects of glargine and BIL

Patients treated with glargine had no change from baseline in LDL particle concentration at 52 weeks, either in total LDL or in the large or small LDL subclasses (Fig. 1; Additional file 1: Table S1). In contrast, patients treated with BIL had a significant increase from baseline in total LDL at 52 weeks (Fig. 1; Additional file 1: Table S1). Among the LDL subclasses, large LDL concentrations were not significantly different between treatments at 26 or 52 weeks, although concentrations decreased significantly from baseline with BIL treatment at 26 weeks. However, small LDL concentrations showed significant increases from baseline with BIL treatment at both 26 and 52 weeks (Fig. 1; Additional file 1: Table S1). LDL particle size did not change with glargine treatment, but decreased significantly with BIL treatment from baseline to 26 and 52 weeks (Fig. 2; Additional file 1: Table S1).

Patients treated with glargine had no change in VLDL particle concentration over time, either in total VLDL or in the large, medium, or small VLDL subclasses (Fig. 3; Additional file 1: Table S1). Patients treated with BIL had significant increases from baseline in large and medium VLDL subclasses (Fig. 3; Additional file 1: Table S1). Total HDL particle concentration was not significantly different with BIL vs. glargine treatment, but large HDL decreased with BIL (Fig. 4; Additional file 1: Table S1).

Patients treated with glargine had no significant changes in the apolipoproteins measured (Fig. 5; Additional file 1: Table S1). Patients treated with BIL had significant increases from baseline in Apo A2, Apo B100, and Apo C3, with significant treatment differences at 26 and 52 weeks for Apo C3 (Fig. 5; Additional file 1: Table S1). There were no significant changes in CETP or CEC. Adiponectin concentrations decreased significantly from baseline to 52 weeks with both glargine and BIL treatment, with no significant difference between treatments (Fig. 6; Additional file 1: Table S1).

Type 2 diabetes—effects of glargine and BIL

The two cohorts of type 2 diabetes patients were similar in baseline characteristics, with the exception that LFC was numerically higher in the insulin naïve cohort (Table 1). The insulin naïve cohort had a numerically higher concentration of large VLDL compared to the cohort previously taking insulin (basal switch) (Table 1).

Insulin naïve cohort

Total LDL concentrations decreased from baseline with glargine treatment; the change was largely in the small LDL subclass (Fig. 1; Additional file 1: Table S1). Large VLDL decreased from baseline with glargine treatment and was significantly lower compared to the BIL group at 52 weeks (Fig. 3; Additional file 1: Table S1). Compared to baseline, VLDL size decreased while LDL size increased with glargine treatment (Fig. 2; Additional file 1: Table S1). Total HDL concentrations were unchanged with glargine treatment but decreased from baseline with BIL treatment (Fig. 4; Additional file 1: Table S1).

Mean LFC decreased from baseline in patients treated with glargine, while there was no significant change with BIL treatment [LSM difference at 52 weeks: 2.57% (0.94–4.21%); p = 0.002] (Additional file 1: Table S1) [8]. Stronger correlations were found for changes in lipoprotein parameters with changes in LFC in patients taking BIL in this population (Table 3). The strongest correlations were with total LDL, IDL, small LDL, total VLDL, small VLDL, Apo A2, Apo B100, and Apo C3 [all positive; five had correlation coefficients (r) > 0.3].

There were no significant changes from baseline or treatment differences in any measured apolipoproteins in either treatment group at 26 or 52 weeks (Fig. 5; Additional file 1: Table S1). FFAs decreased significantly from baseline in both treatment groups, but this decrease was significantly greater with glargine vs. BIL (Additional file 1: Table S1). There were no significant changes in CETP or CEC. Adiponectin concentrations increased with glargine and decreased with BIL, but these changes were not significant (Fig. 6).

Basal switch cohort

Patients previously on insulin that were randomized to glargine did not have any significant changes from baseline to 52 weeks in total LDL or any LDL subclass during treatment (Fig. 1; Additional file 1: Table S1), or in total VLDL or any VLDL subclass (Fig. 4). Patients assigned to BIL had mean increases from baseline in the large VLDL subclass which were significantly different from the glargine group at 26 weeks (Fig. 3; Additional file 1: Table S1). The glargine group had no changes from baseline in concentrations of HDL or HDL subclass particles. The BIL group had a mean decrease from baseline in total HDL and small HDL particles, but there were no significant differences between treatments at any endpoint and there was a difference at baseline (Fig. 4; Additional file 1: Table S1).

Mean LFC did not change in the glargine group, but increased from baseline in the BIL group [LSM difference at 52 weeks: 5.27% (3.43–7.11%); p < 0.001] (Additional file 1: Table S1) [8]. Some correlations were found for changes in lipoprotein parameters with changes in LFC in patients taking BIL (Table 3). The highest correlations were with small LDL, large VLDL, and Apo B100 (all positive), but none of r values were >0.3.

Patients treated with glargine had no significant change from baseline in any measured apolipoproteins. The BIL group had a significant increase from baseline in Apo C3 at 52 weeks (Fig. 5; Additional file 1: Table S1). There were no significant changes in CETP or CEC. Adiponectin concentrations declined with both BIL and glargine treatment, but the change from baseline was statistically significant only with BIL at 26 weeks (Fig. 6; Additional file 1: Table S1).

Lipoprotein subclasses in type 1 diabetes

Glargine treatment had little effect on LDL particles; BIL treatment led to increases from baseline in total LDL, small LDL, and Apo B100, but a decrease in large LDL. Glargine did not affect VLDL particles; BIL treatment led to increases from baseline in medium and large VLDL (similar in diameter range to the VLDL1 subfraction [30]), and an increase in Apo C3. Medium HDL was (nominally) lower with glargine, but increased with BIL treatment. These changes are consistent with the increase in serum TGs noted with BIL treatment in the Phase 2/3 clinical program [8].

In a cross-sectional study, patients with type 1 diabetes compared to healthy controls had lower medium VLDL, lower small HDL, and larger HDL size, with higher concentrations of large HDL [21]. Whether these differences were due to insulin treatment is not known. The effects of intensive insulin therapy on NMR lipoproteins were evaluated during the Diabetes Complications and Control Trial (DCCT) [24]. The results showed that intensive insulin therapy was associated with larger LDL diameter and lower levels of small LDL and small HDL.

The effects of these differences on coronary artery disease (CAD) risk were studied by Erbey et al. [20] in the Pittsburgh Epidemiology Study of Diabetes Complications (EDC). In a large cohort of patients with type 1 diabetes (n = 337), higher concentrations of small, dense LDL were associated with higher cholesterol, TGs, total LDL and lower HDL-C, and an increased risk of CAD. Soedamah et al. [22] also studied subjects with type 1 diabetes from the EDC cohort in a nested case (CAD, n = 59) control study (non-CAD, n = 59). In univariate analyses, lipid mass and particle concentrations of all three VLDL subclasses, and small and medium LDL were higher in cases than controls, while large HDL concentration was lower. Medium HDL was higher in patients with CAD and in the multivariate model was associated with CAD. Lyons et al. [23] evaluated lipoprotein subclasses and relationships with carotid intima media thickness (CIMT) in DCCT/EDIC participants. In analyses that adjusted for multiple variables, LDL particle concentrations were associated with internal CIMT in both sexes; LDL-C and Apo B were also associated with CIMT.

In our study, patients with type 1 diabetes treated with glargine did not have significant changes in any lipoprotein subclasses, presumably because the study treatment was similar to the patients’ pre-study treatment. In patients treated with BIL, lipoprotein changes were suggestive of an increased risk for CVD (although the magnitude of the change in CVD risk is uncertain). There were too few cardiovascular events in the BIL program to make assessments of any relationships with lipoprotein subclasses [31].

Lipoprotein subclasses in type 2 diabetes

In the type 2 diabetes insulin naïve cohort, glargine use was associated with nominal decreases from baseline in total and small LDL and large VLDL; these changes were not seen with BIL treatment. In contrast, patients with type 2 diabetes previously on insulin that were randomized to glargine (basal switch cohort) did not have these changes, suggesting effects of prior insulin treatment. Randomization to BIL treatment was associated with increased large VLDL in the basal switch cohort, which was concordant with the observed changes in TGs [8].

Several small, short term studies evaluated insulin effects on lipid subfractions in type 2 diabetes, which showed that insulin treatment was associated with lower concentrations of TG, VLDL, small LDL, and small HDL, as well as increased activity of adipose tissue lipoprotein lipase (LPL) and CETP, and decreased activity of hepatic lipase [26–29, 32]. These changes are consistent with the known effects of insulin in lowering circulating TG [2], and the subsequent effects of TG levels on lipoproteins [33].

Although the effects of conventional insulin on serum TGs have been previously reported [2], data for the effects of insulin on VLDL particles are limited. In the overall BIL program, TGs increased from baseline when patients who were previous taking insulin were switched to BIL, but not when insulin naïve patients were treated with BIL [8]. In the current study, changes in large VLDL most closely paralleled these observations (as did, to a lesser extent, VLDL size), suggesting that the observed differences between glargine and BIL hinged on changes in the large VLDL subclass. Glargine has a greater peripheral insulin effect than BIL and thus would be expected to increase peripheral lipogenesis, while reducing lipolysis [4]. This may be mediated via increased fatty acid uptake subsequent to increased LPL activity [25]. BIL would have less effect on peripheral lipogenesis, and would allow increased peripheral lipolysis compared to glargine, which would likely result in increased FFA levels and hepatic VLDL production. The studies noted above did not report surrogate imaging markers of CVD to assess effects of lipid subfractions on CVD risk. In the BIL Phase 3 program, there were too few clinical CV events in patients with type 2 diabetes to assess any relationship of these lipid fractions with clinical disease [31].

Adiponectin

There is extensive literature describing adiponectin as an adipokine that is associated with insulin sensitivity and vascular wall anti-inflammatory activity in type 2 diabetes, and showing that lower levels of adiponectin are associated with insulin resistance in type 2 diabetes [34–38]. However, there are only limited data in patients with type 1 diabetes. In the current study, the type 1 diabetes cohort had much higher mean adiponectin levels than the type 2 diabetes cohorts at baseline. Other investigators have reported that adiponectin levels are higher in patients with type 1 diabetes compared to non-diabetes controls [39–43]. The observations from the current study that adiponectin levels decreased from baseline with both BIL and glargine treatment are also concordant with literature showing that adiponectin levels decline with intensification of insulin treatment [42–44]. In summary, the current data confirm that patients with type 1 diabetes may have high adiponectin levels, but whether lower levels of adiponectin are associated with insulin resistance or clinical cardiovascular disease cannot be confirmed.

Associations with liver fat content

Increased LFC has been associated with increased risk for CVD [45–47]. Whether this is the result of “common soil” with other markers of insulin resistance is unclear. However, the data from the current study characterizes some of the associations between LFC and the NMR profile of lipoproteins and apolipoproteins.

Liver fat content was assessed at baseline and after insulin treatment in these diabetes cohorts [12]. In correlations of baseline LFC with baseline values of lipoproteins and other soluble biomarkers, we found that large VLDL, VLDL size, Apo C3, and small LDL had the strongest and most consistent positive correlations with LFC, with r values of 0.216‒0.460 across the diabetes cohorts. There were also positive correlations with small HDL and Apo B100, and a negative correlation with adiponectin. These parameters are all associated with features of insulin resistance.

These observations support the concept that LFC is a function of insulin resistance not only in type 2 diabetes, but also in type 1 diabetes. The negative baseline correlations of LFC with adiponectin, an insulin-sensitizing and anti-steatotic adipokine, are consistent with previous observations that lower levels of adiponectin are associated with higher LFC [48]. Of note, even in the type 1 diabetes cohort that had a much higher baseline adiponectin compared to the type 2 diabetes cohorts, the negative baseline relationship of adiponectin to LFC was present. These data suggest that even with lower LFC in type 1 diabetes than in type 2 diabetes, some of the same mechanisms of insulin resistance may be operative.

Because glargine treatment reduced LFC in insulin naïve patients, and patients randomized to BIL who had previously been on insulin had increases in LFC, we examined the correlations between changes in the various lipid parameters, change in LFC, and treatment (Table 3). In patients with type 1 diabetes treated with glargine, there was very little absolute change in LFC and few major correlations of LFC change with lipoprotein changes, except for a direct association with large VLDL. In insulin naïve patients with type 2 diabetes treated with glargine, there was a decrease in LFC that correlated positively with changes in VLDL measures; this effect was even stronger in patients with type 2 diabetes previously treated with insulin, where there was only a small decrease in LFC with glargine.

These observations suggest that no single lipoprotein subclass, apolipoprotein or even adiponectin is a major driver of change in LFC in these insulin-treated patients with diabetes, although large VLDL, VLDL size, Apo B100, and Apo C3 may be related in some scenarios.

Strengths and limitations

The strengths of this study include the carefully characterized subgroup of patients from a large Phase 3 program in which three cohorts of patients with different diabetes types were studied [5]. There was uniform collection of serial lipid profiles and LFC, measures of glycemic efficacy, insulin dosing and adverse event profiles. The MRI data for LFC and the lipid and lipoprotein measures were all collected in standardized fashion across the substudy cohort. Background glucose- and lipid-lowering medications were stable before randomization and during the study and did not differ between treatment groups.

Limitations of this study include: (1) The substudy cohort was not representative of the whole Phase 3 cohort in terms of randomization and outcomes, as the ability to perform NMR/MRI was limited to certain study sites. (2) The study cohort is too small to do extensive multivariable analyses of whether such things as sex, glycemic control, background glucose-lowering medications, lipid-lowering medications, or changes in body weight affected lipids and lipoproteins or LFC. (3) Although elevated TG (and VLDL TG) and nonalcoholic fatty liver have been associated with hepatic insulin resistance [49, 50], we did not obtain insulin levels needed to calculate insulin resistance because of limited ability to interpret plasma insulin concentrations in insulin-treated patients. (4) Although we did multiplicity adjustments, the results for individual lipoprotein parameters should be interpreted with caution. (5) Although in patients treated with BIL lipoprotein particle and apoprotein changes were suggestive of increased CVD risk (although the magnitude of the change in CVD risk is uncertain), analyses related to clinical CVD risk or events were not feasible in this study. (6) Individual studies did not capture all of the details of all potential variables that might affect lipid values, including menopausal status and hormone replacement therapy.