Plasma Triglycerides and HDL-C Levels Predict the Development of Diabetic Kidney Disease in Subjects With Type 2 Diabetes: The AMD Annals Initiative

Chronic diabetic kidney disease (DKD) is the major cause of end-stage renal disease worldwide (1). Hyperglycemia and hypertension are the main risk factors for DKD development and progression (2). However, in spite of the achievement of recommended targets for blood glucose and blood pressure, the residual risk for diabetic nephropathy remains high among patients with type 2 diabetes (3,4).

Diabetic dyslipidemia—high triglycerides (TGs) and/or low HDL-cholesterol (HDL-C) levels—may be one of the factors responsible for this high residual risk (4). Interestingly, recent studies demonstrated that intrarenal accumulation of lipids may contribute to glomerular injury (5–7) through the induction of oxidative stress or the release of proinflammatory cytokines and growth factors (8–10). Alterations in synthesis, uptake, or efflux of lipids may be responsible for this accumulation (8,11–13). Thus it was recently demonstrated that the expression in mesangial and tubular cells of key proteins in HDL metabolism, such as ABCA1, ABCG1, and SR-BI, was reduced in diabetic mice with nephropathy, suggesting a contribution of impaired HDL-mediated cellular cholesterol efflux in the development of DKD (14).

Epidemiological studies have demonstrated a link between diabetic dyslipidemia and DKD. High TG and low HDL-C concentrations were associated with DKD in a post hoc analysis of large intervention studies of high-risk patients with diabetes (15–17). The Action in Diabetes and Vascular Disease: Preterax and Diamicron-MR Controlled Evaluation (ADVANCE) Study demonstrated that lower baseline HDL-C levels were a significant and independent predictor of DKD, whereas no association was found with the risk of diabetic retinopathy, suggesting that differences may exist in the pathophysiology of these microvascular complications (15). Furthermore, the hypotriglyceridemic drug fenofibrate has been shown to slow the decline of renal function and to reduce albuminuria (16,17) in patients with type 2 diabetes. Several other, smaller epidemiological studies have pointed to the role of diabetic dyslipidemia in the incidence and progression of DKD (2,18), although with conflicting results (18–20).

A large, international, cross-sectional study of outpatients with diabetes recently demonstrated an independent association of low HDL-C and/or elevated TGs with DKD after controlling for LDL cholesterol (LDL-C) levels and established risk factors for microvascular disease (21). These observations need to be confirmed in large, longitudinal cohort studies of patients with type 2 diabetes.

In Italy, diabetes care is mainly provided by a public network of about 700 diabetes clinics in which teams of specialists provide diagnostic confirmation of, prevention of, and treatment for diabetes and its complications through close follow-up and regular checkups. Since 2004 the Associazione Medici Diabetologi (AMD) Annals Initiative, which involves approximately one-third of all the diabetes outpatient clinics operating within the national health care system, has promoted a continuous improvement effort through the monitoring of a large set of process and outcome indicators, with the aim of examining strengths and limitations of current diabetes care (22–24).

In this study, data were analyzed from a large cohort of subjects with type 2 diabetes without DKD participating in the AMD Annals Initiative over a 4-year follow-up period. The aim of this study was to determine whether high TG and/or low HDL-C plasma concentrations are predictors for the development of DKD and its components after controlling for LDL-C levels and other well-established risk factors such as glycemia and blood pressure.

This was a retrospective observational study of a selected cohort of 15,362 patients with type 2 diabetes with an estimated glomerular filtration rate (eGFR) ≥60 mL/min/1.73 m², normoalbuminuria, and an LDL-C concentration ≤130 mg/dL at baseline, from the database of the Italian Association of Clinical Diabetologists AMD network.

Patients followed up at diabetes centers participating in the Italian AMD initiative. The analysis was performed using a data set of electronic medical records collected between 2004 and 2011. For the purpose of the analysis, we considered only patients who were ≥40 years old and had at least 48 months of follow-up for data on eGFR and albuminuria. The last visit with complete renal data was considered the 4-year evaluation. The baseline visit was selected considering the evaluation performed 48 months before the last visit (range, 42–54 months). In the case of multiple records, the visit closest to 48 months was considered as the baseline visit. All annual visits after the baseline were extracted, if available.

Of the 47,177 patients identified, we excluded those with albuminuria, eGFR ≤60 mL/min/1.73m², or a previous discordant eGFR value (i.e., <60 mL/min/1.73m²) and those with missing data regarding antidiabetic treatment (Supplementary Fig. 1). Furthermore, by study design, subjects with an LDL-C concentration >130 mg/dL were also excluded. A total of 15,362 subjects from 95 diabetes clinics homogeneously distributed throughout the country met the inclusion criteria and were included in the study (Supplementary Fig. 1). The centers involved in the study include about one-third of all the Italian Centers for Diabetes.

When comparing main clinical characteristics of subjects included and excluded from the study, major differences were related to inclusion criteria (baseline eGFR ≥60 mL/min/1.73 m², no albuminuria, and LDL ≤130 mg/dL) and exclusion conditions (most of excluded patients had an eGFR<60 mL/min/1.73 m² or albuminuria at baseline). As expected from the selection criteria, the two groups were different for baseline renal function and lipid profile. The remaining clinical and demographic characteristics were similar at baseline (data not shown).

As already reported (22–24), the analysis of the database is an attempt by the Italian AMD Annals Initiative to identify a set of indicators that can be used in the context of continuous quality improvement. Participating centers adopted the same software systems for everyday management of outpatients, and a specially developed software package allowed us to extract the information we intended to analyze from all the clinical databases (AMD Data File). Moreover, data from all participating centers were collected and centrally analyzed anonymously (22–24).This initiative includes measuring and monitoring HbA_1c, blood pressure (LDL-C), total cholesterol and HDL-C, and TGs. The use of specific classes of drugs (insulin, statins, and two or more antihypertensive agents) was also evaluated. HbA_1c was measured using high-performance liquid chromatography in all participating centers. Since normal ranges for HbA_1c varied among centers, the percentage change with respect to the upper normal value (measured value ÷ upper normal limit) was estimated and multiplied by 6.0 to allow comparisons among the centers. No standardization was applied to TG and HDL-C measurements. For TGs, 85% of the laboratories used the enzymatic-colorimetric method with glycerol-3-phosphate oxidase/phenol + aminophenazone (GPO-PAP); the remaining labs used either enzymatic-colorimetric GPO-PAP with the subtraction of blank (glycerol) or methods using ORTHO instrumentation. Importantly, the normal range and decisional levels are the same with these methods. For HDL-C, no laboratory pretreated the samples; 60% of the laboratories used the enzymatic homogeneous direct method, and the remaining 40% used the same method but different detergents (e.g., polyethylene glycol). Also, for HDL-C, the normal range and decisional levels are the same.

Kidney function was assessed by serum creatinine and urinary albumin excretion measurements. Glomerular filtration rate (GFR) was estimated for each patient using a standardized serum creatinine assay and the Chronic Kidney Disease Epidemiology Collaboration formula (25). Increased urinary albumin excretion was diagnosed and defined as albuminuria if the urinary albumin concentration was >30 mg/L, the urinary albumin excretion rate was >20 µg/min, or the urinary albumin-to-creatinine ratio was >2.5 mg/mmol in men and >3.5 mg/mmol in women.

The primary outcomes were 1) eGFR <60 mL/min/1.73 m²; 2) albuminuria; 3) eGFR <60 mL/min/1.73 m² and albuminuria; and 4) eGFR reduced >30%. The occurrence of prespecified end points was evaluated on a yearly basis over the 4-year study period. Patients were considered to have reached the study end points if, at any annual visit during the study period, they met these criteria.

Data are given as means ± SDs; categorical variables are described as frequencies and percentages. The main analysis aimed to evaluate the association between baseline HDL-C and TGs with renal outcomes during the study period. To take into account potential variability among diabetes centers participating in the study, a mixed logistic regression model with diabetes clinics fitted as a random effect was used for each renal outcome to estimate odds ratios (ORs) with their 95% CIs. Univariate analysis was corrected for baseline eGFR, whereas multivariate models adjusted for sex, age, duration of diabetes, BMI, eGFR, HbA_1c, LDL-C, blood pressure, presence of retinopathy, smoking status, and pharmacological treatment (lipid-lowering medications, statins, fibrates, antihypertensive drugs, ACE inhibitors or angiotensin II receptor antagonists, aspirin, and antidiabetic therapy). Multivariate models were fitted including a missing indicator variable (only for duration of diabetes, BMI, and smoking status in the case of missing values). Data were analyses using STATA software version 14 (StataCorp, College Station, TX). P values <0.05 were considered statistically significant.

Table 1 shows the baseline clinical characteristics of participants with type 2 diabetes stratified by HDL-C and TG values. As expected from the study design, renal function was preserved at baseline, with a mean eGFR of 87 mL/min/1.73 m². Overall, clinical data for subjects with or without low HDL-C concentrations (<40 mg/dL in males and <50 mg/dL in females) showed only small but significant differences. Subjects with low HDL-C concentrations at baseline had a higher BMI (30.4 vs. 28.7 kg/m²), slightly worse glucose control with higher HbA_1c, and more HbA_1c values above target. Subjects in the low HDL-C group were more frequently female and younger, with a shorter known diabetes duration. As for lipid profile, this group also had a larger percentage of subjects with high TG levels, a smaller percentage of subjects with out-of-target LDL-C levels, and more use of lipid-lowering medications (49% vs. 44%). Also, the percentages of the group who were smokers and who were receiving antihypertensive treatments were larger in the low HDL-C group.

Similar findings were observed when comparing the baseline characteristics of subjects with normal and high TG values (Table 1). Subjects with high TGs had a higher BMI, worse glucose control, lower HDL-C levels, a higher percentage of patients taking lipid-lowering and antihypertensive medications, and a larger percentage of smokers compared with subjects with normal TGs. No differences were noted in basal use of aspirin when stratifying the study population according to TG/HDL-C levels.

For antidiabetic therapy, subjects with low HDL-C or high TG values were more frequently treated with oral hypoglycemic agents, with or without insulin, and less frequently treated with insulin alone or diet.

At the end of the 4 years of follow-up, among 15,362 study subjects, 1,962 developed low eGFR values (12.8%), 1,167 (7.6%) showed eGFR reduced >30%, 3,570 (23.2%) developed albuminuria, and 614 (4.0%) developed both low eGFR and albuminuria (Supplementary Table 1). As shown in Fig. 1, all renal outcomes were significantly worse in subjects with high TG (Fig. 1A) and/or low HDL-C (Fig. 1B) values than in subjects with baseline lipid values in the normal range.

Table 2 shows baseline characteristics according to the development of DKD outcomes. Subjects developing low eGFR within 4 years of follow-up (n = 1,962) were more frequently female and older, with a longer diabetes duration at baseline. They also had higher baseline blood pressure and HbA_1c. Baseline creatinine levels were higher and eGFR lower in those who developed low eGFR at follow-up. For lipid profile, the low eGFR group showed higher TGs and included a large percentage of subjects with TGs above and HDL-C levels below targets compared with the higher eGFR group, whereas LDL-C control was better in this group.

Similar findings were noted when comparing baseline variables of patients who developed albuminuria (n = 3,570) at follow-up versus those who remained normoalbuminuric (Table 2), with the exception of the percentage of smokers, which was larger among those developing albuminuria but not those developing low eGFR.

The percentage of subjects with retinopathy at baseline was significantly larger in subjects developing either low eGFR (23% vs. 19%) and/or albuminuria (24% vs. 18%).

Antihypertensive drugs and aspirin were more frequently used by subjects developing DKD, whereas between-group differences in the use of lipid-lowering medications were not statistically significant. For hypoglycemic treatment, those developing low eGFR were treated more frequently with insulin, either alone or in combination with oral hypoglycemic agents.

Table 3 shows univariate and multivariate associations of low HDL-C and high TGs with renal outcomes at follow-up. Overall, in this population of subjects with LDL-C concentrations <130 mg/dL, having atherogenic dyslipidemia at baseline was a significant risk factor for DKD and was associated with all the examined renal outcomes.

In particular, a TG concentration ≥150 mg/dL increased the risk of low eGFR by 26% and of a reduced eGFR by 29%; it increased the risk of albuminuria by 19% and of developing both low eGFR and albuminuria by 35%. These associations were only attenuated by multivariate adjustment.

Low HDL-C concentrations (<40 mg/dL in men, <50 mg/dL in women) were associated with a 27% increased risk of low GFR and a 28% increased risk of reduced GFR; the risk of developing albuminuria associated with low HDL concentrations was 24% and that of developing both low GFR and albuminuria was 44%. These associations were still significant and only attenuated by multivariate adjustment.

When examined as continuous variables, each 50 mg/dL increase in TG values augmented the risk of low eGFR by 10% and of eGFR reduction by 8%, whereas it increased the risk of albuminuria by 6% and the risk of developing one abnormality by 13%. Each 10 mg/dL increase in HDL-C level decreased the risk of developing low eGFR or albuminuria by 9%, of reduced eGFR by 9%, and of developing one abnormality by 12%. All these associations were only attenuated by multivariate adjustment (Table 3).

As shown in Fig. 2, low HDL-C and/or high TG levels significantly increased the risk of developing renal outcomes after factoring for sex, age, blood pressure, glucose control, and LDL-C levels. The risk of developing DKD associated with high TG/low HDL-C values was attenuated in subjects with at-target values of the other major risk factors, that is, blood pressure, HbA_1c, and LDL-C levels (Supplementary Table 2). In this well-controlled group, the risk of developing albuminuria associated with low HDL-C levels remained significant (P = 0.018)(Supplementary Table 2).

DKD is a chronic and harmful complication of type 2 diabetes. The epidemiology and natural history of DKD have changed in the past three decades, mostly as a result of better diagnostic and treatment tools. In particular, therapeutic progress has led to a larger number of subjects reaching the recommended targets for blood glucose and blood pressure. In spite of the better control of known risk factors, the residual risk for DKD is still high; thus the identification of other modifiable risk factors in addition to hyperglycemia and hypertension is urgently needed.

Our data clearly indicate the main features of diabetic dyslipidemia—high TG and/or low HDL-C levels—as important risk factors for the development of DKD. Indeed, in a large population of outpatients with type 2 diabetes and controlled LDL-C levels, low HDL-C and high TG levels were independent risk factors for the development and progression of renal disease.

In our cohort of >15,000 subjects without DKD at baseline, 32% developed DKD after 4 years of follow-up. This incidence is slightly higher than that observed in a previous study (26) but comparable to the results of the ADVANCE Study (15). In the latter study, during a median follow-up period of 5 years, 32% of participants developed new or worsening microvascular disease and one-third (28%) experienced a renal event. This high incidence may also depend on a selection bias, since subjects participating in our study are routinely followed by specialist outpatient clinics; thus it is possible that our patients are more comparable to the population at high risk for cardiovascular disease (CVD) in the ADVANCE study than to the subjects with diabetes not referred to the diabetes centers. However, it is important to note that there are relevant differences between patients participating in the ADVANCE Study, who had diabetes and were at high risk for macrovascular events, and some of whom had established DKD at baseline, and those in our cohort, including DKD-free patients routinely screened at diabetes centers all over Italy, who were not (because of the observational nature of our study) selected for CVD risk and/or for therapies; thus our study is representative of common clinical practice.

In keeping with the results of the ADVANCE Study (15), we found that albuminuria was the most frequent renal event observed within the 4 years of follow-up; indeed, 23.2% of subjects in our cohort developed albuminuria, 19% developed low eGFR values or a reduced eGFR, and 4% developed both renal abnormalities.

Notably, after controlling for LDL-C levels and numerous confounders, the incidence and progression of these renal abnormalities were independently associated with TG and/or HDL-C levels outside the recommended targets at baseline.

The adjusted risk for developing any renal event associated with lower HDL-C levels or higher TG levels was between 19% and 44%. Our data also show that there was no threshold in the association between dyslipidemia and DKD risk, since the risk showed a linear trend with increasing TG or decreasing HDL-C values.

Although the risk associated with high TGs and low HDL-C cannot be directly compared, and since all analyses indicated an independent prognostic value of both parameters, our data also suggest that lower HDL-C levels may be more strongly associated with DKD risk, especially when albuminuria occurs; thus a TG concentration ≥150 mg/dL increased the risk of albuminuria by 19% and of developing low eGFR or albuminuria by 35%, whereas that associated with low HDL-C levels was 24% for albuminuria and 44% for both abnormalities.

These associations were partly attenuated by multivariate adjustment, especially in the subgroup analysis, when subjects who were at target for blood pressure, glucose control, and LDL-C were considered. This observation points to the major role played by classical risk factors, which are, as recommended by current guidelines, the major targets for DKD prevention. On the other hand, since the association of diabetic dyslipidemia with DKD risk remained significant after controlling for numerous confounders—including BMI, smoking habit, use of drugs, and blood pressure and glucose control—an independent role of lipid fractions on kidney function is still plausible, as supported by several pieces of experimental evidence.

Thus, DKD may share some common features with atherosclerosis; HDL particles may have a protective role in both processes, including antioxidant and anti-inflammatory properties. Furthermore, recent evidence suggests that the atheroprotective role of HDL-C is not solely limited to its circulating concentrations, but rather depends on the qualitative properties of different HDL particles (27). These protective properties of HDL particles seem to be impaired by diabetes, as recently demonstrated in women with type 2 diabetes who had a dysfunctional HDL subpopulation distribution when compared with women without diabetes, in spite of similar HDL-C concentrations (28); indeed, this altered HDL subpopulation profile was associated with higher levels of inflammatory markers, which may contribute to the high CVD risk observed in women with diabetes (29,30). In this regard, it was recently reported that in patients with diabetes and nephropathy, increased serum concentration of advanced glycation end products was associated with impairment of the antioxidative capacity of HDL particles (31).

On the other hand, dyslipidemia per se is not sufficient to initiate kidney damage since individuals without diabetes but with elevated cholesterol or TG levels rarely develop kidney disease; accordingly, it is plausible that the metabolic derangement typical of diabetes (hyperglycemia, insulin resistance) facilitates the lipotoxic effects on the microvascular bed and is necessary for DKD to develop.

High TGs and low HDL-C are two clinical components of the metabolic syndrome and may be consequences of the underlying insulin resistance. Indeed, a growing body of evidence supports a pathogenic role of insulin resistance in kidney dysfunction through mechanisms involving glomerular hyperfiltration and increased vascular permeability caused by hyperinsulinemia, subclinical inflammation, or podocyte abnormalities (32,33). These findings, mostly deriving from experimental studies, are supported by gene-association studies and interventional studies of the effect of insulin sensitizers on DKD progression (32).

The association between dyslipidemia and microvascular disease is also supported by recent epidemiological studies. In the ADVANCE Study, the only longitudinal study with a number of participants and follow-up duration comparable with those in our study, the risk of developing renal events associated with lower HDL-C levels was 19%, which is similar to our findings (15).

Notably, a large, cross-sectional, multicenter study recently reported that in a population of subjects with diabetes and controlled LDL-C levels, TG and HDL-C levels were significantly and independently associated with diabetic microvascular disease, especially kidney disease, without any difference among different geographic regions (21).

Our results are in keeping with these cross-sectional findings, demonstrating the independent role of dyslipidemia across a wide set of covariates and confounders, starting with a population with similar baseline characteristics. Furthermore, it is important to note that both a study by Sacks et al. (21) and ours enrolled patients who had controlled LDL-C levels, allowing better dissection of the role of a high TG/low HDL-C phenotype on DKD.

The impact of atherogenic dyslipidemia on microvascular disease is not limited to type 2 diabetes. In children with type 1 diabetes, high HDL-C levels and good glycemic control were favorable prognostic factors for regression of microalbuminuria during long-term treatment with ACE inhibitors (34). Furthermore, in a cross-sectional analysis of a subset of the Diabetes Control and Complications Trial (DCCT)/Epidemiology of Diabetes Interventions and Complications Cohort (EDIC) Study cohort (35), albuminuria was associated with specific HDL subclasses. Finally, a recent report from the Finnish Diabetic Nephropathy (FinnDiane) Study (36) showed that TGs and cholesterol content in the VLDL particles were associated with incident albuminuria and its progression.

Conversely, the association of dyslipidemia with DKD was not confirmed by other studies (18–20). The reasons for these conflicting results may depend on several factors, including a different genetic background and/or gene–diet interactions, which, although not specifically evaluated in our study, may mitigate the power of the association between lipids and renal outcomes.

Also, sex may be a factor that potentially influences the association of DKD risk with dyslipidemia. It has been demonstrated that sex differences exist in the prevalence of DKD clinical manifestations (24,37): women with diabetes show more GFR reduction, whereas albuminuria occurs more frequently in men. A single-center observational study found that lower HDL-C levels were associated with the progression of DKD in men but not in women (38). When we tested this hypothesis, however, we did not find any sex difference in DKD risk. Similarly, no difference was found when the study population was stratified according to age, blood pressure values, and glucose control, nor in the subgroup of subjects with at-target LDL-C concentrations (<100 mg/dL), further demonstrating the strength of the association between dyslipidemia and DKD. This was particularly true for low HDL levels that were significantly associated with albuminuria risk, even when only patients at target for blood glucose, blood pressure, and LDL-C levels were considered (n > 1,600).

Our study has several strengths and limitations. The strengths of this study include the duration of observation, the large number of patients, and the strict inclusion criteria, which allowed only subjects with controlled LDL-C levels and with repeated measurements of renal function within the normal range to be included.

The limitations include the lack of centralized measurements and standardization of laboratory parameters; also, the observational nature of our study and the lack of information on duration of use of hypoglycemic and hypolipidemic drugs are other important issues to be considered when addressing cause-and-effect relationships between dyslipidemia and microvascular outcomes.

International guidelines recommend maintaining blood glucose and blood pressure levels within the target limits to avoid or delay DKD. Despite improvements in blood glucose and blood pressure control as a result of these guidelines, many patients still develop DKD, and the residual risk for this complication remains high.

Our data clearly indicate that both high TG and low HDL-C are independent risk factors for DKD development. These results may have important therapeutic implications; indeed, in the Fenofibrate Intervention and Event Lowering in Diabetes (FIELD) and Action to Control Cardiovascular Risk in Diabetes (ACCORD) trials, fenofibrate treatment was associated with a reduction in albuminuria (39). Also, a dose-dependent effect of omega-3 fatty acids on DKD in subjects with hypertrygliceridemia (40) was recently reported. To date, there are no data available for treatments that increase HDL-C levels. Only large, long-term interventional studies will clarify whether lipid-lowering medications decreasing TG levels and/or increasing HDL-C levels are effective in reducing DKD risk among patients with type 2 diabetes.

Acknowledgments. The authors thank Dr. F.M. Sacks (Department of Nutrition, Harvard School of Public Health, Boston, MA) for reading the manuscript and providing useful suggestions. The authors thank all of the centers participating in the AMD Annals Initiative (a complete list is provided in the Supplementary Data).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. G.T.R. and P.F. researched data, wrote the first draft of the manuscript, and edited the manuscript. S.D.C., P.G., and R.P. researched data and contributed to the discussion. F.V., A.P., A.C., S.G., C.G., and D.C. critically revised the manuscript and contributed to the discussion. G.T.R. 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.