Serum Levels of Apolipoproteins and Incident Type 2 Diabetes: A Prospective Cohort Study

A low level of HDL cholesterol (HDL-C) is a known risk factor for type 2 diabetes (T2D) that precedes the onset of the disease (1). Not only HDL content but also various lipoprotein particles have progressively gained attention as new markers for T2D risk independent of established risk factors (2,3). Furthermore, a novel lipoprotein insulin resistance index, which combines six lipoprotein parameters (higher levels of large VLDL particles and small LDL particles, lower levels of large HDL particles, smaller mean LDL and HDL particle size, and larger mean VLDL particle size) was found to be independently associated with incident diabetes (3,4).

Given that lipoproteins are biochemical assemblies, their role in T2D etiology can be more deeply explored by investigating the associations between their main components (e.g., apolipoproteins) and the risk for T2D. Part of the interest in studying apolipoproteins is the reported lack of causality between reduced HDL-C and increased risk of T2D, which raises the question about whether some of the beneficial effects attributed to HDL-C are due to the composition of HDL particles, including apolipoproteins, supporting their role as probable determinants in the association between HDL-C levels and T2D (5).

Few studies have shown that serum levels of apolipoproteins such as apoA1 and apoB are stronger biomarkers of T2D or its complications than traditional lipids (6,7). Higher HDL-C–to–apoA1 and HDL-C–to–apoAII ratios are strongly and independently related to a lower risk of T2D in different populations (8–10). We previously found that apoCIII promoter variants increase T2D risk and the need for insulin treatment in lean subjects (11). Mainly because of the fraction in HDL particles, total serum apoCIII was found to be a key diabetogenic risk factor among Turks (12); however, this prospective study had a short follow-up of 4.4 ± 1.2 years. A comprehensive investigation of the role of the serum apolipoproteinic profile of various apolipoproteins as biomarkers for incident T2D and possible targets for future therapy is still lacking.

We hypothesized that serum levels of various apolipoproteins and the ratios of those apolipoproteins with apoA1 are stronger biomarkers for incident T2D than HDL-C. To this end, we studied the associations of serum HDL-C, apoA1, apoCIII, apoD, and apoE levels as well as the ratios of apoCIII to apoA1, apoD to apoA1, and apoE to apoA1 with incident T2D in the Rotterdam Study, a prospective population-based cohort study.

The Rotterdam Study is a prospective population-based cohort study in Ommoord, a district of Rotterdam, the Netherlands. The design of the Rotterdam Study has been described in detail elsewhere (13). Briefly, in 1989, all residents within the well-defined study area aged ≥55 years were invited to participate; 78% (7,983 of 10,275) agreed. There were no other eligibility criteria to enter the Rotterdam Study except a minimum age and residential area based on postal code. The first examination took place from 1990 to 1993 after which follow-up examinations were conducted every 3–5 years. This study was based on data collected during the third visit (1997–1999). The measures were done based on a case-cohort study design. In this analysis, we only used the random subset of the Rotterdam Study. The Rotterdam Study has been approved by the medical ethics committee according to the Population Screening Act: Rotterdam Study executed by the Ministry of Health, Welfare and Sports of the Netherlands. All participants in the present analysis provided written informed consent to participate and to obtain information from their treating physicians.

During the third visit, fasting blood samples were collected at the research center. Plasma was isolated and immediately put on ice and stored at −80°C. Citrate plasma (200 μL) was sent in July 2008 to Myriad RBM (Austin, TX; www.myriadrbm.com). Fifty biomarkers were quantified by multiplex immunoassay on a custom-designed human multianalyte profile, including four apolipoproteins. The intraassay variability was <4%, and the interassay variability was <13%. The markers investigated in the current study have no standard international calibration reference; therefore, interpretation of the absolute values should be with caution. Because all values in this set of participants have been quantified with the same assay, relative measures such as hazard ratios (HRs) of associations should not have been affected. In this study, we included the apolipoproteins that, besides other lipoproteins, are also constituents of HDL: apoA1, apoCIII, apoD, and apoE. Their measurements were available in grams per liter. We converted values to moles per liter by dividing measurements in grams per liter by molecular weight in grams per mole, as follows: apoA1 (30,778 g/mol), apoCIII (10,852 g/mol), apoD (19,820 g/mol), apoE (36,154 g/mol) (www.myriadrbm.com).

Participants were followed from the date of the baseline center visit onward. At baseline and during follow-up, cases of prediabetes and T2D were ascertained through active follow-up of general practitioner records, hospital discharge letters, and glucose measurements from Rotterdam Study visits, which take place approximately every 4 years. Diabetes, prediabetes, and normoglycemia were defined according to the current World Health Organization guidelines. Normoglycemia was defined as a fasting blood glucose level <6.0 mmol/L; prediabetes was defined as a fasting blood glucose of 6.0–7.0 mmol/L or a nonfasting blood glucose of 7.7–11.1 mmol/L (when fasting samples were unavailable); T2D was defined as a fasting blood glucose ≥7.0 mmol/L, a nonfasting blood glucose ≥11.1 mmol/L (when fasting samples were unavailable), or the use of blood glucose–lowering medications (14). Information about the use of blood glucose–lowering medications was derived from both structured home interviews and linkage to pharmacy dispensing records. At baseline, >95% of the Rotterdam Study population was covered by the pharmacies in the study area. All potential events of prediabetes and T2D were independently adjudicated by two study physicians. In case of disagreement, consensus was sought with an endocrinologist. Follow-up data were complete as of 1 January 2012.

Height and weight were measured with the participants standing without shoes and heavy outer garments. BMI was calculated by dividing weight in kilograms by height in meters squared. Waist circumference was measured at the level midway between the lower rib margin and the iliac crest with participants in standing position without heavy outer garments, with emptied pockets, and breathing out gently. Blood pressure was measured at the right-side brachial artery with a random-zero sphygmomanometer and the participant seated, and the mean of two consecutive measurements was used. Insulin, glucose, HDL-C, and triglyceride levels were measured with a cobas 8000 modular analyzer (Roche Diagnostics). The corresponding interassay coefficients of variation are insulin <8%, glucose <1.4%, and lipids <2.1%. Information on medication use, parental diabetes history, and tobacco smoking was collected by trained research assistants using computerized questionnaires during home visits. History of cardiovascular disease (CVD) was defined as a history of coronary heart disease (myocardial infarction, revascularization, coronary artery bypass graft surgery, or percutaneous coronary intervention) and verified from general practitioner medical records. Smoking was classified as current versus noncurrent. Alcohol intake was assessed in grams of ethanol per day as recorded in the questionnaires wherein participants were asked for their average daily consumption.

Linear regression models were used to investigate the age-, sex-, and prevalent diabetes–adjusted association of serum levels of HDL-C, apoA1, apoCIII, apoD, and apoE as well as the ratios of apolipoproteins with apoA1 with fasting glucose and fasting insulin in 971 participants (120 with prevalent diabetes). Markers with a right-skewed distribution were transformed to the natural logarithmic scale (all apolipoproteins, their ratios, HDL-C, triglycerides, fasting glucose, and fasting insulin). We defined the apolipoprotein value as an outlier when the value was 4 SDs higher or lower than the mean of the normal variable. After excluding the outliers, apolipoprotein levels were standardized by dividing the measured value by the SD. To study the apolipoprotein levels jointly, not just individually, we calculated an apolipoproteinic score using β-values from the cross-sectional analysis with glucose levels (Table 2). The score was calculated by the equation −0.013 × apoA1 + 0.016 × apoCIII – 0.009 × apoD + 0.018 × apoE. Among 851 participants free of diabetes, we assessed the association of serum levels of HDL-C, apoA1, apoCIII, apoD, and apoE and the molar ratios of each apolipoprotein with apoA1 as well as the apolipoproteinic score by using Cox proportional hazard models. The first model was adjusted for age and sex. The second model additionally adjusted for BMI. The third model added parental history of diabetes, hypertension, alcohol use, smoking, prevalent CVD, and serum lipid–reducing agents. The fourth model added triglycerides. A multiple imputation procedure was used (five imputations) to impute the missing data for covariates, including weight (0.82%), height (0.82%), smoking (1.03%), triglycerides (2.37%), and prevalent CVD (0.62%). We also assessed linearity of the associations by adding quadratic terms to regression models. All analyses were conducted with SPSS for Windows version 21 software (IBM Corporation, Armonk, NY).

Table 1 summarizes the baseline characteristics of 971 participants. The mean (SD) age at baseline was 73 (7.5) years, and 44.8% were men. The mean (SD) BMI was 26.7 (3.9) kg/m², and 12.6% of the study population used lipid-lowering medications.

Table 2 presents the associations between HDL-C, apoA1, apoCIII, apoD, and apoE as well as the ratios of apolipoproteins with apoA1 and fasting glucose and fasting insulin. ApoCIII was positively associated with fasting glucose. ApoE, apoCIII-to-apoA1 ratio, and apoE-to-apoA1 ratio were positively associated with fasting glucose and fasting insulin. HDL-C, apoA1, and apoD were inversely associated with fasting glucose and fasting insulin.

Among 851 participants free of diabetes, 110 cases of incident diabetes were identified during a median follow-up of 13.5 years (10 diabetes cases per 1,000 person-years). After adjustment for age, sex, BMI, parental history of diabetes, hypertension, alcohol use, smoking, prevalent CVD, and serum lipid–reducing agents (Table 3, model 3), high serum HDL-C levels remained associated with a lower risk for T2D, whereas high serum levels of apoCIII, apoE, apoCIII-to-apoA1 ratio, apoE-to-apoA1 ratio, and apolipoproteinic score remained associated with a higher risk for T2D. Only apoCIII (per 1 SD naturally log-transformed HR 1.42 [95% CI 1.03, 1.96]) and apoCIII-to-apoA1 ratio (per 1 SD naturally log-transformed HR 1.56 [95% CI 1.04, 2.36]) survived the adjustment for triglycerides in the last model. All these adjustments slightly changed the effect estimates of the association between serum apoCIII levels and apoCIII-to-apoA1 ratio and incident T2D from the age- and sex-adjusted model (per 1 SD naturally log-transformed HR 1.49 and 1.64, respectively) to the final model (per 1 SD naturally log-transformed, HR 1.42 and 1.56, respectively). We assessed linearity of the associations by adding quadratic terms to regression models; none were significant (P > 0.1).

In this prospective population-based cohort study, we investigated the associations of serum levels of apolipoproteins (apoA1, apoCIII, apoD, apoE), HDL-C, and the ratios of the apolipoproteins with apoA1 and their apolipoproteinic score with the risk of T2D. After adjustment for age, sex, BMI, parental history of diabetes, hypertension, alcohol use, smoking, prevalent CVD, and serum lipid–reducing agents, high serum HDL-C levels remained associated with a low risk for T2D, whereas high serum levels of apoCIII and apoE, the apoCIII-to-apoA1 and apoE-to-apoA1 ratios, and the apolipoproteinic score remained associated with a high risk for T2D. Only apoCIII and apoCIII-to-apoA1 ratio survived the adjustment for triglycerides, being identified as stronger T2D risk markers than HDL-C.

ApoA1 is the major proteinic component and main indictor of the antiatherogenic HDL (15,16). Thus, we calculated molar ratios of serum apolipoproteins, which have in common the presence in HDL besides other lipoproteins, to investigate how the variability of the serum apolipoproteinic flux could be associated with the risk for T2D and whether this variability could be a better risk marker than HDL-C.

Few previous studies have shown that serum levels of apolipoproteins better estimate the risk of T2D or its complications than traditional lipids (6,16). A higher ratio of HDL-C to apoA1 and apoA1 levels have been strongly and independently related to a lower risk of incident T2D in various populations (8–10). In agreement with these studies, we found that serum apoA1 levels are associated with T2D risk. We extended the list, however, to a larger set of apolipoproteins and included serum levels of apoCIII, apoD, and apoE. The results indicate that apoCIII is a better predictor of T2D than apoA1. Associations of serum levels of apoCIII with the risk of T2D have been previously investigated prospectively (12), but with a shorter follow-up of 4.4 ± 1.2 years based on logistic regression, whereas we performed survival analysis with available incident diabetes date during a median follow-up of 13.5 years.

The studied apolipoproteins are found in combination with HDL, LDL, VLDL, or chylomicrons (17). The variation in apolipoprotein levels, therefore, represents the variation in both apolipoprotein flux and the levels of lipids they bind. Even interventions that modify apolipoprotein levels affect lipid levels simultaneously. For instance, gemfibrozil increases apoA1 and apoE concentrations as well as HDL-C levels (18). Because dyslipidemia starts before the diagnosis of T2D (19), the association of apolipoproteins and T2D could be confounded by lipid levels. To disentangle the effect of apolipoproteins from HDL, LDL, VLDL, or chylomicrons, we used molar ratios of serum apolipoproteins levels to apoA1 and adjusted their associations for lipid levels. ApoCIII was the only apolipoprotein that remained significant after adjustments, and its ratio with apoA1 strongly predicted T2D. This finding may indicate that total flux of apoCIII and its ratio with the main indictor of HDL (apoA1) has a stronger impact on T2D etiology than HDL-C itself.

Genetically elevated levels of apoCIII have been related to a higher risk of diabetes. We have previously showed the associations of apoCIII promoter variants with the risk of T2D in lean healthy subjects (11). Hokanson et al. (20) found an apoCIII haplotype to be associated with type 1 diabetes. These studies may indicate that the association we report between apoCIII and T2D is likely causal. The mechanisms through which apoCIII may cause T2D remain unclear; however, Åvall et al. (21) found that apoCIII links islet insulin resistance to β-cell failure in diabetes, which may indicate that apoCIII is mainly working through β-cell dysfunction rather than through insulin resistance.

The typical dyslipidemia of T2D is characterized by high triglyceride and low HDL-C levels (22). ApoCIII is an atherogenic protein found on HDL, VLDL, and LDL and leads to hypertriglyceridemia through three actions that can impair plasma lipoprotein metabolism. First, apoCIII inhibits VLDL clearance, which not only results in hypertriglyceridemia but also increases the formation of small dense LDL, known as lipoprotein insulin resistance index components (4,23). Second, apoCIII at high concentrations inhibits the action of lipoprotein lipase to hydrolyze lipoprotein triglyceride in vitro, but this is not yet shown in humans (24). Third, apoCIII assists in the formation of triglyceride-rich VLDL in liver cells, stimulating the secretion of VLDL (25). Furthermore, antisense inhibition of apoCIII lowers apoCIII and triglycerides in hypertriglyceridemia (26). All these findings provide evidence that triglycerides are downstream of apoCIII, which is in line with the current results. When adjusted for triglycerides, the associations between apoCIII and T2D are weaker but remain significant, suggesting that other pathways might be involved.

Previous studies suggest that apolipoproteins in lipoproteins or plasma may predict coronary heart disease better (27–30) and that their HDL fraction seems to be the most important (31). This observation might also be true for T2D. Total serum apoCIII, mainly because of the fraction in HDL particles, was found to be a key diabetogenic risk factor in a study among Turks (12). However, further investigation is needed to better explore the role of apolipoproteins in T2D etiology and to compare the importance of their HDL fraction with the non-HDL fraction.

The current findings have important clinical implications. They emphasize the importance of the variability in the total serum apolipoproteinic flux rather than the HDL-C level itself. Patients with T2D have defective HDL particles with altered composition in addition to low HDL-C levels (32,33). Thus, HDL components, such as apolipoproteins, might be good candidate biomarkers to evaluate the risk of T2D and even potential targets for prevention or treatment of the disease if found to be casual (34,35).

This study has certain strengths. To our knowledge, this prospective, population-based cohort study is the first to investigate the association between human serum levels of apoCIII, apoD, and apoE and their ratios to apoA1 with incident T2D during a long-term follow-up. We used data from a well-characterized prospective cohort study, which allowed us to correct for a wide range of covariates. We are also aware of the limitations of the study. The population is ≥55 years of age and mainly comprises individuals of European ancestry. Thus, generalization of the results to younger age-groups and other ethnicities should be done with caution. Reference values for apolipoproteins depend on choice of assay technique; therefore, comparison of the values should take into account the chosen technique (36). We do not have available measurements of apolipoprotein amounts in every lipoprotein particle separately and specifically in HDL (only total serum levels), so we could not study the diabetes risk apolipoproteinic profile within HDL or other lipoprotein particles compounded by the studied apolipoproteins. The studied apolipoproteins all have in common their presence in HDL, but only serum apoA1 and apoD are totally found in HDL. Chylomicrons, secreted from the intestinal enterocyte, also contain apoA1, which is quickly transferred to HDL in the bloodstream (15,16).

In conclusion, serum levels of apoCIII and the apoCIII-to-apoA1 ratio are associated with risk of T2D independent of known risk factors and are a stronger biomarker than HDL-C level. Aside from the observed association between HDL-C and T2D, the findings highlight the role of variation in composition of HDL particles and other lipoproteins in relation to the disease. Therapeutic approaches for T2D should aim for the normalization of both quantity and composition of HDL particles and other lipoproteins. More data are needed to determine the importance of levels of apoCIII in specific lipoproteins for T2D risk assessment and management and to elucidate the interaction between triglycerides and apoCIII in relation to risk of T2D.

Acknowledgments. The dedication, commitment, and contribution of inhabitants, general practitioners, and pharmacists of the Ommoord district to the Rotterdam Study are gratefully acknowledged. The authors thank Layal Chaker, Jolande Verkroost-van Heemst, and Ke-Xin Wen of Erasmus Medical Center for their invaluable contribution to the collection of the diabetes data.

Funding. The Rotterdam Study is supported by the Erasmus Medical Center and Erasmus University Rotterdam; Netherlands Organization for Scientific Research (NWO); Netherlands Organisation for Health Research and Development; Research Institute for Diseases in the Elderly; Netherlands Genomics Initiative; Ministry of Education, Culture and Science; Ministry of Health, Welfare and Sports; European Commission (Directorate-General XII); and Municipality of Rotterdam. M.K. is supported by an NWO Veni grant (916.16.079). A.D. is supported by an NWO Veni grant (916.12.154) and Erasmus University Rotterdam Fellowship.

None of the funders had a role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; or preparation, review, or approval of the manuscript.

Duality of Interest. O.H.F. works for ErasmusAGE, a center for aging research across the life course funded by Nestlé Nutrition (Nestec Ltd.), Metagenics Inc., and AXA. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. A.B. ran the analysis and wrote the manuscript. S.L. contributed to the analysis. M.A.I. and A.H. designed the study. O.H.F. designed the study and provided resources. E.J.G.S., M.K., and A.D. designed the study and critically revised the manuscript. All authors read and approved the manuscript. A.D. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.