The use of fasting vs. non-fasting triglyceride concentration for estimating the prevalence of high LDL-cholesterol and metabolic syndrome in population surveys

For practical reasons fasting blood samples are difficult to obtain in large population studies and thus such studies commonly comprise non-fasting or 'semi-fasting' blood samples [1‐5]. For example, in a long series of population-based studies in Finland since the early 1970s, a four hour fasting has been requested. Fasting is not a necessary prerequisite for many analytes, but for serum triglyceride (Tg) it is important [2, 6]. Fasting serum Tg values are needed for cardiovascular disease risk evaluation [7, 8] for the calculation of low-density lipoprotein cholesterol (LDL-C) concentration and for the classification of subjects into those with and without metabolic syndrome [9‐11]. However, non-fasting Tg has emerged recently as an increasingly important risk factor [12, 13]. Also estimating trends in these risk factors has become of utmost importance in epidemiology [14].

Recently, we derived correction factors whereby insufficiently fasted/postprandial serum Tg values can be converted into "corrected" fasting values [15]. In view of the potential error resulting from using non-fasting serum Tg values in the calculation of LDL-C and consequently in the classification of subjects into the metabolic syndrome, we transformed non-fasting data to fasting data using our recently published factors [15].

The aim of the present study was to assess how much a corrected non-fasting serum Tg deviates from the true fasting value and whether the correction for non-fasting serum Tg is worthwhile for calculation of LDL-C or for classifying subjects into the metabolic syndrome in health examination surveys.

The National FINRISK 2007 Study (FR07) is the eight consecutive population-based risk factor survey in Finland. The surveys have been carried out with 5-year intervals since 1972 and the sample sizes have varied from 6500 to 13 500 men and women, depending on the survey year. The total sample size of the FR07 study was 9957 persons in the age range of 25-74 years. It was a random sample, drawn from the population register and stratified according to sex, 10-year age group and five geographical areas. Of the 9957 invited persons, 6247 (62.7%) took part in both the health examination and blood sampling [3].

During the year 2007 the same subjects participated in health examinations twice. In January-April, subjects were requested to fast for at least 4 hours before phlebotomy. The participants of the surveys were asked the time in full hours since their last meal by a study nurse. For the present analysis (non-fasting FR07, visit 1) eligible were those who reported having fasted ≥2 to ≤8 hours (1979 men) and ≥2 to ≤7 hours (2303 women) based on our earlier work on effects of fasting on Tg values [15]. In the substudy carried out in May-June the same participants were requested to fast for 10 hours (true-fasting FR07, visit 2).

In addition to the two groups described above, we identified a separate group drawn from the FR07 sample comprising 552 subjects who fulfilled the following fasting criteria: ≥8 hours for men and ≥7 hours for women in visit 1 and who had also participated in the true-fasting study, i.e., visit 2. These subjects are referred to as the Reference Group and they are not included in either of the above parent population samples. Individuals with serum Tg values >10 mmol/L were excluded from all analyses.

In addition to reporting the results on all men and women, we defined categories of healthy men (n = 824) and women (n = 994) as those who had a BMI ≤35; reported an alcohol consumption below the 90^th percentile in self-reported questionnaire; had no diagnosed cardiovascular disease (CVD), diabetes or cancer; had no medication for hypercholesterolemia; and had normal blood pressure. The results are also presented for subgroups with severe obesity BMI>35 (men n = 101, and women n = 183).

Subjects were classified into a high LDL-C group by using the cut-off value ≥3.00 mmol/L calculated by the Friedewald formula [16]. Subjects were classified as having the metabolic syndrome according to three different definitions [9‐11].

Figure 1 shows the Altman-Bland plot indicating a 5.9% (95% CI: 4.8-7.0) bias between the difference of corrected FR07 and true-fasting data for serum Tg. Also, illustrated by the figure, the magnitude of the bias does not seem to depend on the Tg concentration. Figure 2 shows the distribution for raw data (all) (visit 1) before and after correction for non-fasting serum Tg concentrations of the subjects (n = 4282) and the true-fasting data (visit 2) of the same subjects.

The medians (25^th, 75^th percentiles) of the non-fasting and corrected Tg data were 1.18 (0.87, 1.72) mmol/L (P < 0.0001) and 1.06 (0.78, 1.52) mmol/L (P < 0.0001) compared to the median of the true-fasting value, 1.00 (0.75, 1.38) mmol/L, Additional file 1. The median serum Tg concentrations of the non-fasting data differed significantly from true-fasting values for all groups, Additional file 1. After correction the corrected medians differed significantly from the true fasting medians for all subjects, all men, all women and healthy men, Additional file 1. The within sub-group difference between corrected and true-fasting median serum Tg was largest among all men 10.2% (P < 0.0001). In the reference group, in which all subjects had fasted for at least 7 hours at both visits (n = 552), the difference between the two visits was less than 2%, Additional file 2.

In population health examination surveys standardization and long-term quality assessment of risk factor measurements are essential. Especially so in monitoring trends and tracking of risk factors [3, 5, 19]. Among the major risk factors for CVD, serum Tg has emerged recently as an increasingly important risk factor [7, 12, 13, 20, 21]. In both diagnostics and risk assessment fasting serum Tg concentration is required either for the calculation of LDL-C or classification of subjects into the metabolic syndrome. In large population-based health examination surveys true fasting for all subjects is often difficult to achieve [1‐5, 22] and usefulness of undefined post prandial serum Tg data is doubtful. In a previous study, we calculated factors by which non-fasting serum Tg values can be converted into fasting values when the time from the last meal is known [15].

In the present study we wanted to test whether the statistical correction of non-fasting serum Tg into fasting values is worthwhile in calculating LDL-C and categorizing subjects into the metabolic syndrome. Our data showed that using incomplete fasting serum Tg concentration to calculate LDL-C concentration results in misclassification of 5.1% of subjects as having an LDL-C below 3.00 mmol/L compared to true fasting values. Correction of post prandial serum Tg narrowed misclassification to 1.6%, increasing the number of subjects at risk by some tens of thousands at the population level in Finland. Likewise, categorizing subjects for the metabolic syndrome by the MS-ATP-2005 [11] criteria using corrected non-fasting Tg instead of the non-fasted values, decreased the proportion of misclassification and resulted in virtually the same prevalence as the use of true-fasting serum Tg, the difference being 2.1 percentage units. According to MS-IDF-2005 [10] and MS-IDFTF-2009 [9] criteria the differences were very similar.

The total intraindividual biological variation of fasting serum Tg in the reference group was 17.1%, which is much narrower than in other similar studies [23] but the total biological variation in our original samples (21.9%) was of the same order of magnitude as found by others, 21.7 - 29.9% [18], having a time frame of months between the phlebotomies. The small effects of non-fasting vs. fasting serum Tg on misclassification into high LDL-C or metabolic syndrome should be compared to the intraindividual biological variation of serum Tg. Because of the high biological variation of serum Tg several measurements of Tg per subject are needed for reliably estimating the CVD risk [7, 23].

In addition to its use as a risk factor, serum Tg concentration is used widely to calculate LDL-C with the Friedewald formula [16]. Because of limitations of the formula [24], the new recommendation from the National Cholesterol Education Program (NCEP) encourages direct measurement of LDL-C. Increasing concerns have, however, been raised about the specificity of different direct methods for measuring the cholesterol concentration from various lipoprotein fractions. The newly published data showed [25] that all evaluated direct LDL-C methods failed to meet the NCEP's total error goals when analysing samples from patients with cardiovascular disease or dyslipidemia. Recent data have proved that the number of LDL particles is a more reliable indicator for cardiovascular risk assessment than the cholesterol concentration in lipoprotein particles [26]. Because there is only one apolipoprotein B (apoB) molecule per one LDL particle, measurement of apoB gives an estimate of the number of lipoprotein particles. Therefore, it has been suggested that, to improve CVD risk assessment, apoB measurement should be adopted in addition to LDL-C to national guidelines [27]. Implementation of such a measure would benefit greatly large surveys, since measurement of apoB does not require a fasting blood sample.

Our data based on a random population sample of 4282 individuals examined twice within a 3 month interval showed that measuring serum Tg after an incomplete fast causes a relatively small misclassification of the prevalence of people with high LDL-C or metabolic syndrome.