Differential Influence of LDL Cholesterol and Triglycerides on Lipoprotein(a) Concentrations in Diabetic Patients

We included 191 consecutive diabetic outpatients of Caucasian origin (69 type 1 and 122 type 2 diabetic patients) attending the diabetic unit of a university hospital. The clinical characteristics of these patients are shown in Table 1. Pregnant women, patients with acute or chronic infections, and subjects with severe medical conditions(e.g., malignancy, liver cirrhosis, connective tissue disease, chronic congestive heart failure, and renal failure) were excluded from the study. To avoid the possible transient increase of Lp(a) after starting insulin treatment (21), we also excluded all patients in whom insulin treatment was initiated during the 2 months before the study. Informed written consent was obtained from all the participants. The study was approved by the hospitals' human ethics committee and conducted according to the principles expressed in the Declaration of Helsinki.

For BMI (in kg/m²), measurements of weight and height were taken with the patients wearing light clothing and no shoes. Arterial blood pressure, which was determined by one observer using a standard 12.5-cm cuff mercury sphygmomanometer, was calculated as the mean of three readings taken in the sitting position after at least 10 min rest. Diastolic blood pressure was recorded at the disappearance of the Korotkoff sounds (phase V). Hypertension was diagnosed according to a predetermined blood pressure level(systolic blood pressure >160 mmHg or diastolic blood pressure >90 mmHg)or if the patient was on antihypertensive therapy. Smoking habits were also recorded, and smokers were defined as all subjects currently smoking more than one cigarette per day.

To assess evidence of macroangiopathy, we use the World Health Organization(WHO) protocol, which includes a carefully detailed questionnaire and a 12-lead electrocardiogram(22,23). The presence of diabetic retinopathy was assessed by ophthalmoscopy and fluorescein angiography after midriasis by a trained ophthalmologist unaware of the patient's clinical status. Neuropathy was diagnosed according to criteria suggested by the American Academy of Neurology and the American Diabetes Association (24). The albumin excretion rate (AER) was determined by a double-antibody radioimmunoassay (Diagnostic Products, Los Angeles, CA) and calculated as the mean of the last three 24-h urine samples collected within the previous year.

Metabolic parameters were evaluated in venous blood drawn after an 8-h over-night fast. Blood was collected into EDTA for HbA_lc analysis,into sodium fluoride for the glucose assay, and into plain tubes for both lipid and renal biochemistry analyses. Glucose was determined using a glucose oxidase method on a Hitachi 747 autoanalyzer (Hitachi, Tokyo), and HbA_lc was measured by an ion-exchange high-performance liquid chromatography system (Bio-Rad, Hercules, CA) with a reference range of 4.7-6.6%. Total serum cholesterol and total triglycerides were tested on unfrozen samples by an automated enzymatic method (Boehringer Mannheim,Mannheim, Germany) on a Hitachi 747 analyzer. HDL cholesterol was determined by assaying the cholesterol concentration in the supernatant obtained after precipitating lipoproteins with densities lower than HDL using a mixture of phosphotungstic acid and magnesium chloride(25). LDL cholesterol was calculated following Friedewald's equation [LDL cholesterol = total —HDL cholesterol — (triglycerides/5)] when the serum triglyceride level was <300 mg/dl. When triglyceride values were >300 mg/dl, LDL cholesterol was determined by sequential ultracentrifugation(26). The plasma was adjusted to the d = 1.006 g/ml with sodium bromide solution and ultra-centrifuged at 940,000 rpm for 21 h at 15°C in a Centrikon T-1035 rotor-type TFT 45.6 (Kontron International, Zurich, Switzerland). After centrifugation, the VLDL was carefully removed and LDL cholesterol was calculated as follows: LDL cholesterol = total — HDL — VLDL cholesterol. In addition, LDL cholesterol was corrected by subtracting the cholesterol portion of Lp(a), estimated as 0.3 × Lp(a) concentration(27). Apo AI and apo B-100 were measured by immunonephelometric assay (Behring, Frankfurt, Germany) with a coefficient of variation <5.1%. The patients were stratified by the type of hyperlipidemia according to the Frederickson classification system modified by the WHO (28). Patients with triglycerides between 200 and 1,000 mg/dl exhibited type IV hyperlipidemia,patients with total cholesterol levels >240 mg/dl were assigned type IIa hyperlipidemia, and patients with high levels of both cholesterol and triglycerides were classified as having type IIb hyperlipidemia.

Lp(a) was measured by enzyme-linked immunosorbent assay using a monoclonal Lp(a) antibody technique (Macra Terumo, Newark, DE). Our intra- and interassay coefficients of variation were 4.6 and 5.3%, respectively. Lp(a) phenotypes were determined by SDS-PAGE under reducing conditions followed by immunoblotting using a method slightly modified from that described by Utermann et al. (29). The sizes of apo(a) bands in samples were estimated by comparison with mobility of an apo(a) band in a standard loaded gel in an immediately adjacent lane. According to their relative mobilities compared with apo B-100, apo(a)patterns were categorized into phenotypes F (faster than apo B-100); B(similar to apo B-100); and S1, S2, S3, and S4 (all slower than apo B-100). They were then categorized into the respective double-band phenotypes. The phenotype was defined as null when the blots showed no bands at all.

We used the χ² test for categorical variables and Student's t test for continuous variables to verify the absence of statistical differences. Given their skewed distribution, Lp(a) and triglyceride results were displayed both as median and range. Both Lp(a) and triglyceride results were logarithmically transformed when a parametric test was used. As previously reported, Lp(a) phenotypes were classified by size into three groups: small (phenotypes F, B, S1, and S2), big (phenotypes S3 and S4) and null (no bands)(10,30). When the patient had a double band, the smaller band was used to express the phenotype in the analysis. Analysis of variance was used to evaluate both the differences of Lp(a) levels among apo(a) phenotypes and the differences of Lp(a) concentrations among the types of hyperlipidemia. The correlations between Lp(a) levels and the quantitative variables (i.e., age, BMI, AER,HbA_lc, total cholesterol, HDL cholesterol, LDL cholesterol,triglycerides, apo A1, and apo B were studied by linear regression analysis. To further explore the variables independently associated with Lp(a)concentrations, a stepwise multiple regression analysis was performed. P values (two-tailed) <0.05 were considered statistically significant. The data were analyzed with the Statistical Package for the Social Sciences software.

The clinical characteristics of diabetic patients and their lipoprotein profiles according to Lp(a) phenotype are detailed in Table 2. As shown, we did not observe differences in the lipid parameters among Lp(a) phenotype groups. In a univariate analysis, total cholesterol (r = 0.29, P <0.001), LDL cholesterol (r = 0.30, P = 0.001), apo B(r = 0.24, P = 0.02), and AER (r = 0.27, P= 0.002) correlated with Lp(a) serum levels. These correlations remained at significant levels when types 1 and 2 diabetic patients were studied separately. No significant correlation was detected between Lp(a) and the other variables analyzed. After correcting LDL cholesterol by Lp(a)concentration, the correlation between LDL cholesterol and Lp(a) was lower,but it persisted significantly (r = 0.23, P = 0.01). In the multiple regression analysis, with the exception of AER, LDL cholesterol(positively) and triglycerides (negatively) were independently related to Lp(a) concentration both in the whole group and when the analysis was performed separately according to the type of diabetes (Tables 3 and 4). LDL cholesterol and triglycerides explained the 6.6 and 7.8% of Lp(a) variation, respectively, in the first model (Table 3). After correcting LDL cholesterol by Lp(a) concentration, the two lipid variables explained the 3.8 and 6.4% variation, respectively, in the second model(Table 3). In addition, Lp(a)serum levels were different depending on the phenotype of hyperlipoproteinemia(Table 5). Lp(a) serum concentrations were higher in diabetic patients with hypercholesterolemia(type IIa hyperlipoproteinemia) than in patients with hypertriglyceridemia(type IV hyperlipoproteinemia). The overall distribution of Lp(a) phenotypes according to the type of hyperlipoproteinemia was similar among the groups(Table 5). Because of the relatively high frequency of null phenotypes, a separate analysis was performed to exclude patients heterozygous for the null allele. Thus, when we considered only double-band phenotypes, the frequency of small and big phenotypes was similar for each of the lipidemia categories(Table 5).

The present study indicates that LDL cholesterol(positively) and triglycerides (negatively) are independently related to the Lp(a) concentration both in the whole group of diabetic patients and according to the type of diabetes. LDL cholesterol and triglyceride values explain the 10.2% of Lp(a) variation in our unselected diabetic patients. Therefore, it is not surprising that we also observed differences in Lp(a) serum levels depending on the type of hyperlipoproteinemia. Thus, the highest Lp(a) levels were found in diabetic patients with type IIa hyperlipoproteinemia, and the lowest concentrations were detected in patients with type IV hyperlipoproteinemia. We did not find differences in the distribution of Lp(a)phenotypes among the groups of hyperlipoproteinemia, thus avoiding a misinterpretation of the results caused by a possible bias in the selection of the patients.

Our results agree with those reported by Rainwater(31) in the nondiabetic population. This author studied the extent of variation in Lp(a) not predicted by the LPA genotype using siblings genetically identical by descent at the LPA locus and demonstrated that the Lp(a) concentration was directly correlated with cholesterol and negatively correlated with triglycerides. The two lipid measures explained 5.3% of the total variation in Lp(a) concentrations, and this percentage was reduced to 4% when a correction of LDL cholesterol by Lp(a) concentration was performed. In our study, LDL cholesterol and triglycerides explain 14.4% of Lp(a) variation, but this figure decreases to 10.2% after correcting LDL cholesterol by Lp(a)concentration. The greater influence of both LDL cholesterol and triglyceride in our study could be attributed to the higher frequency of lipid disturbances in diabetic patients compared with the nondiabetic population.

A positive correlation between Lp(a) concentration and LDL cholesterol has been reported previously(15,16,17,18),but the precise mechanism underlying this association is unknown. Although Lp(a) can be catabolized by the LDL receptor, the affinity of Lp(a) to the LDL receptor is considerably lower than that of LDL(32,33,34). Thus, it seems that this common catabolic pathway is not the reason for the association between LDL cholesterol and Lp(a). There is strong evidence that Lp(a) levels are more dependent on the rate of synthesis of apo(a) than on its catabolic rate (35). It is possible that the rate of apo B secretion from the liver could be the link between Lp(a) and LDL cholesterol. However, Morrisett et al.(36) demonstrated that apo B in LDL is synthesized at a rate approximately four times greater than that of the apo B used to produce Lp(a), which suggests that two different pools of apo B are available for lipoprotein production. On the other hand, this relationship between LDL cholesterol and Lp(a) might be observed because LDL cholesterol calculated by the Friedewald formula also included cholesterol contained in the Lp(a) particle. In fact, it has been reported in the general population that the correlation of Lp(a) with LDL cholesterol levels disappeared when the contribution of Lp(a) to LDL cholesterol was considered(15,16). In the diabetic patients included in the present study, Lp(a) concentrations were significantly correlated with LDL cholesterol even after subtracting the estimated contribution from Lp(a). In addition, we found that corrected LDL cholesterol was an independent variable explaining 3.8% of the total variation in Lp(a) concentrations.

We have also observed an inverse and independent association between triglycerides and Lp(a) that explains the 6.4% of Lp(a) serum variation,suggesting that plasma triglycerides could play an important role in Lp(a)metabolism. This finding has been previously reported in the nondiabetic population(15,31,37,38,39,40,41),but the mechanism involved in the association remains poorly understood. The inverse relationship between Lp(a) and triglycerides may be observed because apo(a) is also present in triglyceride-rich particles (TRPs)(42,43,44,45). Apo(a)-containing TRPs, in parallel with the chylomicron remnants, would be rapidly taken up by the liver through the remnant-receptor pathway. Thus, the lower levels of Lp(a) in patients with elevated triglycerides could be a result of a more accelerated clearance of the TRP apo(a) compared with the slower apo(a) catabolism in the LDL density range(42,44). Low concentrations of Lp(a) have been detected in patients with lipoprotein lipase deficiency (43), and it is tempting to speculate that in diabetic patients with functional deficiency of lipoprotein lipase, the higher fraction of apo(a) linked to TRPs might explain the higher catabolic rate of Lp(a).

Hypertriglyceridemia is associated with a risk of coronary heart disease,but the association is not as strong as it is for LDL, and the relationship of triglycerides to atherosclerosis has been a source of confusion. This is partly because when statistical adjustment is made for the effect of HDL, the independent effect of plasma triglycerides becomes weaker or disappears(46,47). Our results suggest that the inverse relationship that also exists between plasma triglycerides and Lp(a) concentrations is another confounding factor that must be considered in future studies addressing this issue.

In conclusion, we observed that LDL cholesterol (positively) and triglycerides (negatively) were independently related to Lp(a) concentration in diabetic patients. These results indicate that serum Lp(a) concentrations depend on lipid profiles and suggest that treatment of diabetic dyslipemia may also affect Lp(a) concentrations.