Low-Density Lipoprotein Values Estimated by Friedewald Equation are Affected by Diabetes Control

Mailing Address: Gustavo Neves de Araujo Ramiro Barcelos 2350, Independência. Postal Code: 90035-003, Porto Alegre, RS – Brazil E-mail: gustavon.araujo@gmail.com Low-Density Lipoprotein Values Estimated by Friedewald Equation are Affected by Diabetes Control Pedro Lima Vieira, Gustavo Neves Araújo, Guilherme Heyden Telo, Luis Felipe Silva Smidt, Mariana Ferreira Jost, Mariana Vargas Furtado, Emilio Hideyuki Moriguchi, Carisi Anne Polanczyk Hospital de Clinicas de Porto Alegre and Universidade Federal do Rio Grande do Sul, Porto Alegre, RS – Brazil


Background
Diagnosis and management of patients with hypercholesterolemia is largely based on LDL cholesterol (LDL-c) levels, considering it the main target of cholesterol lowering therapy.In the early 70s, Friedewald and colleagues 1 created a method to estimate the LDL-c without the use of ultracentrifugation, which was, although gold standard, an expensive and time consuming way of measuring LDL-c.This study represented a milestone in scientific literature, with its findings being used today as a reference for assessing the levels of LDL-c and therapeutic guidance.However, the correlation was not as good in situations where the relationship between VLDL and triglycerides (TG) changes, such as high TG levels (especially above 400 mg/dL) and dysbetalipoproteinemia.
In addition to these situations, more recent studies recommend caution when using Friedewald`s equation (FE) in certain pathological conditions that result in secondary hyperlipidemia (such as diabetes, and kidney disease, for example). 2,3Diabetes not only has a classical pattern of dyslipidemia (increased TG, low HDL-c, and the predominance of particles of LDL-c and small dense), but it also confers a propensity for the development of premature atherosclerosis, probably due to changes in lipoprotein metabolism. 4,5ur goal was to investigate the potential importance of differences between FE and direct measurement of LDL-c in patients with diabetes.

Methods
We conducted a cross-sectional study among patients from an outpatient clinic in a tertiary care university hospital in Southern Brazil.Between 2008 and 2013, 466 consecutive patients with CAD were enrolled at the Ischemic Heart Disease Clinic.All patients had documented CAD, which was defined by the presence of at least one of the following factors: documented history of myocardial infarction, surgical or percutaneous myocardial revascularization and lesion > 50% in at least one coronary artery assessed by angiography, or the presence of angina and positive noninvasive testing of ischemia. 6ood samples were collected by venipuncture into tubes with and without EDTA and sent to the Laboratory of Clinical Pathology for total cholesterol, TG, HDL-c, direct LDL-c and other routine tests The direct measurement of LDL-c was made by the LDL-c Plus kit from Roche Diagnostic, previously standardized in the same laboratory (Roche Diagnostics Brazil).The LDL-c Plus test is a homogeneous enzyme assay for quantitative determination of direct LDL-c in serum, used in automated clinical chemistry analyzers -MODULAR.For quality control purposes, we used the G Precinorm, Precipath HDL/LDL-C, and the curves provided by the device.The measuring/reference range is 3 to 550 mg/dl (0.03 to 5.5 g/l or 0.077 to 14.2 mmol/l).The determination of samples with concentrations of LDL-c > 55 0 mg/dl was made with the re-analysis function.
The result of measured LDL-c in direct mode was only used for research purposes.This study was approved by the Institutional Research and Ethics Committee and informed consent was obtained from all patients.

Statistical analysis
Continuous variables are expressed as mean ± 1 standard deviation (SD).Categorical variables were represented by relative and absolute frequencies.Accuracy was calculated as the percentage of estimates within 30% (P30) of measured LDL.Precision was measured as 1 SD of bias.The agreement between measured and calculated LDL-c was evaluated using Bland-Altman plots, with the calculation of agreement limits (bias ± 2 SD) and CI. 7 Bias was calculated as the mean difference between measured and estimated LDL-c.According to Bland-Altman, 100 individuals are enough to estimate bias and limits of agreement within a 95% CI of about 34% of SD. 8 All data were analyzed using SPSS (version 18.0.0;IBM Company) and P values less than 0.05 were considered significant.

Results
Overall, patients were 63.9 (± 13.3) years of age, 56.8% were women and 46.1% had diabetes.The mean directly measured LDL-c was 106.9 ± 37.1 and the mean Friedewald-estimated LDL-c was 104.7 ± 35.6.The mean total cholesterol and HDL-c were 181.1 (± 42.1) and 48.5 (± 12.3), respectively.The median triglycerides levels were 154.5.Statins were used by 76.5% of the individuals.The mean hemoglobin A1C levels among patients with diabetes were 7.8±1.8%.Complementary data are shown in table 1.
Comparing FE and directly measured LDL-c in patients with different TG levels, those with higher TG levels had greater mean differences and within-group variance (Figure 1).When we divided TG levels into quartiles (TG < 150 mg/dl, 151-200 mg/dl, 201-300 mg/dl and > 300 mg/dl), variance was wider when TG levels were higher.In patients with TG levels above 300 mg/dL, 44% had a difference between FE and directly measured LDL-c greater than 20%.Among patients with TG levels below 150 mg/dL, only 12% had the same difference.Values are reported as mean ± SD, n (%).HBA1C values were only measured in diabetic patients.

Figure 2 Bland and Altmann plots of average between measured and estimated LDL vs. difference between them in the 186 patients with diabetes (A) and 280 patients without diabetes (B) (P = 0.005 and 0.248, respectively).
Bias and concordance superior (CSL) and inferior limit (CIL) described above.

Figure 3
Bland and Altmann plots of average between measured and estimated LDL vs. difference between them in the 56 patients with uncontrolled diabetes (A) and 95 patients with controlled diabetes (B) (P = 0.005 and 0.248, respectively).Bias and concordance superior (CSL) and inferior limit (CIL) described above.

A B A B
Vieira et al.

Original Article
Among patients with diabetes, the FE systematically underestimated measured LDL-c (Figure 2).Bias was 3.7 (p=0.005) and 1.1 mg/dl (p=0.248),accuracy P30 (95% CI) was 86% and 93%, and precision was 80 and 79 in patients with or without diabetes, respectively.A non-significant p value suggests that there is no difference between tests.When we compared patients with diabetes (Figure 3), bias was 5 mg/dl (p=0.016) and 1.93 mg/dl (p=0.179),accuracy P30 (95% CI) was 83% and 88%, and precision was 78 and 77 in subjects with hemoglobin A1C above 8 mg/dl versus below cutoff point, respectively.Bias was similar in patients without diabetes vs. patients with diabetes and HbA1C < 8% (1.1 and 1.93 mg/dl), although it was greater (5.0 mg/dl) among patients with diabetes and HbA1C >8%.There was no statistically significant difference between mean TG levels in patients with HBA1C below and above 8 mg/dl (155 and 164 mg/dl respectively, p=0.402).

Discussion
3][14][15] Dosing levels of LDL-c is extremely important, since the diagnosis and management of dyslipidemia is largely based on their concentrations.However, the gold standard method for its measurement (beta-quantification by ultracentrifugation) requires expensive instrumentation not available in routine clinical laboratories, and it is laborious and time consuming.
Although there are several technical disadvantages of the FE -the need for fasting, relatively high imprecision secondary to accumulated variation of the other 3 measures, and recognized limitations in certain patientsthe method is routinely used in clinical practice, and will only be replaced if other methods show clear advantages in performance, cost-effectiveness, or other financial benefits.
The FE is widely applied and performs remarkably well for most patients.However, differences between FE and directly measured LDL-c may arise in some conditions, such as higher TG levels.A recent study 16 has shown that FE tends to underestimate LDL-c most when accuracy is most crucial, especially if TG levels are above 150 mg/dl.Our study also showed that high TG levels makes FE underestimate LDL-c values.There is a significant proportion of patients with significant difference (> 20%) in test results, and this number exponentially increases with TG levels.Particularly in the presence of TG levels above 150 mg/dl, LDL-c underestimation is clinically relevant and may result in undertreatment.
A previous study 2 suggested that calculation of LDL-c by FE may be inaccurate for assessment of cardiovascular risk in patients with type II diabetes and may not be appropriate for management of lipoprotein abnormalities in those patients.In their study, the FE overestimated by > 10% the actual LDL-c concentration in 39% of patients with diabetes and underestimated the true value in 13% of patients, with only 48% accuracy.Hirany et al. 17 suggested that direct LDL-c assay could be more reliable, accurate, rapid and cost-effective method than the FE for LDL-c determination in individuals with diabetes.
On the other hand, Whiting et al. 18 found that FE was adequate for the measurement of LDL-c in diabetes of both type I and II, with a 68% FE accuracy for subjects with type II diabetes.In our study, FE was inaccurate among overall individuals with diabetes.However, when we divided patients with diabetes into good and poor disease control (HbA1C below and above 8 mg/dL, respectively), the first group behaved as if it did not have diabetes, with a very good correlation between FE and measured LDL-c.Median TG levels were similar in both groups.

Study limitations
We could not perform a multivariate analysis including TG values among patients with poor controlled diabetes, both due to the limited sample size among groups, and because TG levels are mathematically linked with variables in the study (FE).Although chances of losing statistical power is high (poor controlled diabetes usually is associated with higher TG levels), the message of our study is that patients with poor metabolic control should have their LDL-c measured instead of estimated.

Conclusion
Even though there is a good correlation between FE and directly measured LDL-c in general population, the first tends to underestimate LDL-c levels in individuals with high TG levels and poorly controlled diabetes.Considering this high-risk subset of patients, underestimation of LDL-c levels could result in undertreatment of dyslipidemia and predispose these patients to acute events.Patients with diabetes with a good disease control have good correlation between calculated and measured LDL-c, and performing FE in these patients is acceptable.
Most studies to date relating LDL-c levels and cardiovascular disease used FE to estimate LDL-c.Thus, future LDL-guided therapeutic studies may consider using direct LDL-c measurement to determine ideal targets.
. The total cholesterol, TG and HDL-c were measured in accordance with the routine Clinical Pathology Laboratory of our hospital.The LDL-c was calculated by FE ([LDL-c] = [CT] -([HDL-c] + [VLDL-c]), where the concentration of VLDL-c is estimated by the serum concentration of TG [VLDL-c] = [TG]/5).

Figure 1
Figure 1Percentage of patients with difference of estimated and measured LDL-c > 20%.