Serum cholesterol during 27 years: Assessment of systematic error and affecting factors and their role in interpreting population trends

https://doi.org/10.1016/j.cca.2006.10.021Get rights and content

Abstract

Background

The aim was to assess which factors cause a systematic error in serum total cholesterol measurements and how bias can influence the interpretation of serum cholesterol changes of the Finnish population.

Methods

Data on precision and accuracy during 27 years for serum total cholesterol were documented from participation in 438 rounds of five different external quality assessment (EQA) programs.

Results

The mean annual accuracy (bias) of the cholesterol assay using the results from all EQAs during 1978–2004 was − 0.74% (95%CI − 0.88 to − 0.60). An exceptionally large deviation in bias coincided with the introduction of a new serum calibrator lot. New methods or instrumentation had only a minor impact on serum cholesterol bias. The mean serum cholesterol bias during the latest five population studies in 1982–2002 was −0.10% (95%CI −0.60 to 0.40) but comparison of the bias between the last study (CDC EQA in 2002) and the four previous ones (WHO EQA) showed a net difference of 3.32% (p < 0.001). Correcting the mean serum cholesterol of men with respect to WHO and CDC EQA bias changed the interpretation for the last two survey years from an increase of 1.8% to no change.

Conclusions

It is necessary to participate in EQA programs, which include target values measured by the CDC cholesterol reference method and then to perform bias corrections on the mean cholesterol values of the populations.

Introduction

In the estimation of the population-attributable risk of coronary heart disease mortality, blood lipids play a central role [1]. Longitudinal risk assessment is done by regular monitoring of serum lipids in population health surveys spanning several decades [2]. Knowing the systematic error of serum cholesterol for each time point is important for minimizing the error in risk calculations and for interpreting changes in serum cholesterol on a population level.

Laboratories usually confirm and adjust the analytical performance of their methods by participating in external quality assessment (EQA) programs. Activity in this field for lipid measurements started in the early 1970s in clinical chemistry in the Nordic countries [3], [4] as well as in other countries and organizations, e.g. WHO [5] and Center for Disease Control (CDC, Atlanta) [6].

There are many variables that may cause problems in maintaining satisfactory precision and accuracy of methods from year to year. New methods and analysers replace old ones and these changes may affect the analytical bias [7]. There are also many other factors that add to the measurement uncertainty in assays. These include method of calibration, use of external and internal quality assessment materials, calibration and control materials. The reference sample materials (fresh, frozen, lyophilized) may vary between different EQA surveys. Recently it has been shown that lyophilized sera gave 6% lower bias results for cholesterol compared with fresh–frozen sera [8]. The major problem is that during manufacturing these lyophilised materials, changes occur in lipoproteins and other serum components causing a different reactivity in chemical reactions compared to that observed for native serum samples. This artificial matrix effect is well documented in the literature [7], [9], [10], [11]. Because most of the commercial calibrators are also lyophilized sera, erroneous conclusions concerning the accuracy of lipid measurements may be drawn [8]. In a clinical setting a higher systematic error can be allowed compared to that necessary in epidemiologic research.

Serum cholesterol and other cardiovascular risk factor levels in the Finnish population have been followed over a 30-year period. Risk factor measurements began with the North Karelia Project [12] in the province of North Karelia and in the reference area Kuopio. Since 1972 risk factors have been monitored regularly every 5 years in six different areas from a random sample of about 10,000 Finnish adults.

In the present work we show how we have succeeded in maintaining the accuracy of the serum cholesterol assay by participating regularly in several different EQA programs during a quarter of a century. We evaluated the effect of modifications of the methods, types of calibrations and implementation of new analysers on bias. In addition, to estimate the trueness of mean cholesterol values in large Finnish population health surveys, we have calculated the bias correction for mean cholesterol values of the Finnish population during Finrisk health survey years.

Section snippets

External quality assessment

Since 1978 the Laboratory of Analytical Biochemistry at KTL has taken part in five different EQAs, altogether 438 rounds. Quality assessment samples have been received from the WHO Lipid Reference Program (Prague) two–six times a year in 1978–1997 (three concentration levels), from the ShortTerm A surveys of the External Quality Assessment Scheme (LabQ1, Labquality Ltd, Helsinki) twelve times a year in 1978–2004, from the Lipids and Lipoprotein surveys (LabQ2, Labquality Ltd, Helsinki) twice a

Changes in the method

In the 1970s total serum cholesterol was determined by the modified Liebermann–Burchard (LB) method [14]. The automated cholesterol method was developed for use on a Technicon AutoAnalyzer II system (Technicon Instruments Corporation, Tarrytown, NY) [15]. The enzymatic CHOD cholesterol method had been under development using also AutoAnalyzer II (AAII) equipment and tested for some years before finally taking into routine use in 1980 (Table 2). Using fresh sera the mean serum cholesterol

External quality assessment

Efforts to eliminate systematic errors of especially lipid methods by EQA in clinical chemistry laboratories in Scandinavia have a long tradition [16], [17], [18]. The comparability of results for many methods between clinical chemistry laboratories was assured by founding of the EQA organization (Labquality Oy) in the 1970s in Finland. Without this national external quality program it would have been impossible to trace reliably changes in the serum cholesterol levels of the Finnish population

Conclusions

The systematic error of the cholesterol methods has had only a minor effect on estimating and interpreting changes in serum cholesterol in our population surveys. To be able to detect ever smaller changes in serum cholesterol on a population level we conclude from our experience that it is necessary to participate in EQA programs which include target values measured by the CDC cholesterol reference method and then to perform bias corrections on the mean cholesterol values of the populations.

References (25)

  • L.M. Thienpont et al.

    Trueness verification in European external quality assessment schemes: time to care about the quality of the samples

    Scand J Clin Lab Invest

    (2003)
  • G.L. Myers et al.

    Matrix effects in biological reference materials used in standardization of cholesterol measurements

    Fresenius J Anal Chem

    (1990)
  • Cited by (33)

    • 40-Year CHD Mortality Trends and the Role of Risk Factors in Mortality Decline: The North Karelia Project Experience

      2016, Global Heart
      Citation Excerpt :

      Due to changes in laboratory technology during the 40-year period, several methods were used for determining serum cholesterol levels. Methods, instruments, and reagents for cholesterol measurement and the quality analysis data have been described elsewhere [17]. Systematic measurement errors due to changes in laboratory methods and reagents in different study years have been corrected on the basis of the quality control analyses [13].

    • Prevalence and clinical correlates of familial hypercholesterolemia founder mutations in the general population

      2015, Atherosclerosis
      Citation Excerpt :

      All included studies were carried out in accordance with the Declaration of Helsinki and had approvals from local ethics committees as described previously [15]. Serum samples obtained from participants at baseline during each cohort study were analyzed for total cholesterol (n = 28,443) [14,16], LDL cholesterol (n = 15,381) [14], high-density lipoprotein (HDL) cholesterol (n = 28,443) [14,17], and triglycerides (n = 28,442) [14,18] using standard enzymatic assays as described previously. LDL cholesterol was also measured using the Friedewald equation (n = 27,846) [19].

    • What have we learnt about high-density lipoprotein cholesterol measurements during 32 years? Experiences in Finland 1980-2012

      2013, Clinica Chimica Acta
      Citation Excerpt :

      Since 1978 the Laboratory of Analytical Biochemistry at the National Institute for Health and Welfare (THL) has participated in several different EQAs. Details of the EQA programs have been described previously [13]. In this study for HDL-C we used results from the WHO Lipid Reference Program (Prague) in which we have participated 2–6 times annually in 1978–1997 (3 concentration levels); the Lipids and Lipoprotein surveys (Labquality, Helsinki), twice a year in 1989–2012 (2–3 concentration levels); and CDC's Lipid Standardization Program (Atlanta), twice a year in 2001, 2002 and 2004–2006 (3–4 concentration levels).

    • Cardiovascular prevention guidelines in daily practice: a comparison of EUROASPIRE I, II, and III surveys in eight European countries

      2009, The Lancet
      Citation Excerpt :

      To standardise measurements, the laboratory has taken part in the Lipid Standardization Program organised by the US Centers for Disease Control and Prevention (CDC) Atlanta, GA, USA, and External Quality Assessment Schemes organised by Labquality, Helsinki, Finland. The data from the CDC external quality assessment programme showed that there was no systematic error (bias) in the cholesterol method over three decades.17 On the basis of these observations, the individual cholesterol concentrations from EUROASPIRE I were corrected by a multiplication factor of 1·13 for comparative purposes.

    View all citing articles on Scopus
    View full text