On the basis of a high correlation, non-HDL-C (non-high-density lipoprotein cholesterol) and apoB (apolipoprotein B) have been suggested to be of equivalent value for clinical practice; however, the strength of this relationship has not been examined in detail in patients with dyslipidaemia. The present study examines the variance of non-HDL-C compared with apoB in 1771 consecutive patients evaluated in a lipid clinic. These patients were divided into normolipidaemic subjects (*n*=407), type I hyperlipoproteinaemia (*n*=16), type IIa (*n*=736) and IIb (*n*=231) hyperlipoproteinaemia, type III hyperlipoproteinaemia (*n*=38), type IV hyperlipoproteinaemia (*n*=509) and type V hyperlipoproteinaemia (*n*=101). The relationship between non-HDL-C and apoB was examined both in terms of correlation and concordance. Correlation was high, but concordance was only moderate in the normolipidaemic subjects and in those with type IIa and type IIb hyperlipoproteinaemia. Correlation and concordance were both low in the subgroups with type III and type V hyperlipoproteinaemia. In those with type IV hyper-lipoproteinaemia, correlation was moderately high (*r*=0.74), but concordance was only fair. In conclusion, our results indicate that there is substantial variance of apoB for given values of non-HDL-C in many dyslipidaemic subjects. It follows that correlation is not adequate as a sole judge of equivalence of laboratory parameters.

## INTRODUCTION

Non-HDL-C [non-HDL (high-density lipoprotein) cholesterol] was introduced by ATP III (Adult Treatment Panel III) as an alternative target of therapy in patients with hypertriglyceridaemia [1]. Non-HDL-C does not require fasting samples and can be calculated with much more methodological precision than LDL-C [LDL (low-density lipoprotein) cholesterol]. Moreover, there is a high correlation between non-HDL-C and apoB (apolipoprotein B). On this basis, non-HDL-C has been stated to be equivalent to apoB and, therefore, to obviate any need to measure apoB [1]. On the other hand, non-HDL-C and apoB do not measure the same thing. Non-HDL-C is the mass of cholesterol in VLDL (very-LDL) and LDL, whereas apoB is the number of these particles. Although the findings from epidemiological studies and clinical trials demonstrate that both non-HDL-C and apoB are superior to LDL-C in predicting the risk of vascular events, the comparison of non-HDL-C and apoB has yielded mixed results, with some studies showing them to be equivalent and others apoB to be superior [2]. It is noteworthy that none have shown non-HDL-C to be superior to apoB [2].

The present study examines the issue from a different perspective. Our objective is to determine the effects, if any, of the conventional disorders of lipoprotein metabolism on the relationship of non-HDL-C and apoB. If non-HDL-C and apoB are, in fact, surrogates for each other, their relationship should be reasonably similar in these different clinical scenarios. If it is not, then extrapolation of a high correlation in an overall group could mask significant pathophysiological discordance in clinically relevant subgroups.

## MATERIALS AND METHODS

### Participants

Results are reported from 1771 consecutive patients who attended the Lipid Clinic of the Laval University Medical Centre in Quebec City between 1995 and 2005. Patients ranged in age from 2 to 86 years. All were instructed to fast for at least 12 h before blood samples were obtained. None were taking any hypolipidaemic medication. The research protocol was approved by the Laval University Medical Centre Ethical Review Committee, and all patients gave written informed consent.

### Characterization of plasma lipids and lipoproteins

Venous blood samples were obtained from an antecubital vein into Vacutainer tubes containing EDTA (1 mg/ml final concentration). Plasma was separated from blood cells by centrifugation at 975 ** g** for 10 min at 4 °C. The CM (chylomicron) fraction was first separated from the VLDL fraction by ultracentrifugation at 274

**for 45 min using an SW-41 rotor (Beckman Instruments). The VLDL fraction was then isolated by ultracentrifugation using a Beckman 50.3 rotor at a density of 1.006 g/ml, as described by Havel et al. [3]. Heparin and manganese chloride were added to the infranatant, which contains both LDL and HDL, to precipitate the LDL, leaving the HDL in solution [4]. The LDL-C concentration was obtained by differences from the values of cholesterol in the infranatant, measured before and after the precipitation step. The cholesterol and TG [triacylglycerol (triglyceride)] in the plasma and lipoprotein fractions were quantified on an AutoAnalyser RA-500 (Technicon Instruments). The mean recovery of cholesterol in the lipoprotein fractions averaged 95% and ranged from 92–102%. The coefficients of variation of cholesterol and TG determinations were <;2%, as reported previously [5]. HDL-apoA-I and -apoB levels were measured by nephelometry (Dade Behring). The coefficients of variation of the apoB and apoA-I determinations were <;3%.**

*g*### Definition of lipid phenotypes

The lipid phenotypes were defined on the basis of the following criteria, which were age- and sex-corrected: (i) type I: TG >75th percentile for age and gender, CM-TG ≥1 mmol/l and VLDL-TG <;1 mmol/l; (ii) type IIa: LDL-C >75th percentile for age and gender, and TG<75th percentile for age and gender; (iii) type IIb: LDL-C >75th percentile for age and gender, TG >75th percentile for age and gender, and CM-TG <; 1 mmol/l; (iv) type III: TG >75th percentile for age and gender, VLDL-C/TG >0.69 mmol/l plus an apoE2/E2 genotype; (v) type IV: TG >75th percentile for age and gender, LDL-C <;75th percentile, and CM-TG <;1 mmol/l; (vi) type V: TG >75th percentile for age and gender, CM-TG >1 mmol/l and VLDL-TG >1 mmol/l [6,7].

### Statistical analysis

The overall agreement between plasma apoB and non-HDL-C levels was performed using Spearman's rank correlation coefficient. To quantify concordance/discordance, the cohort was divided into quintiles for apoB and non-HDL-C levels. Our *a priori* assessment of clinical agreement (i.e. concordance) was that the two parameters should place a subject within the same quintile, i.e. if the two parameters provided equivalent information, the distribution among quintiles would be the same, in which case the two parameters would be concordant. However, if the rank for the two parameters in any subject fell in two different quintiles, the two parameters would be considered discordant. The κ statistic was used to quantify the overall degree of agreement between the parameters [8]. Statistical analyses were performed using JMP 6.03 software (SAS Institute).

## RESULTS

The results of the lipid and apoB analyses are shown in Table 1. Most of the differences among the groups are a consequence of the diagnostic criteria for each of the dyslipoproteinaemias; however, it is noteworthy that substantial differences in VLDL composition are documented even in the dyslipidaemias not characterized by remnant particles. Thus, in type IV, IIa and IIb, VLDL was cholesterol-enriched compared with the normolipidaemic subjects, and these differences were more pronounced than those in LDL. Moreover, there were major differences in the proportion of VLDL compared with LDL particles. In the normolipidaemic subjects, VLDL constituted 12% of total apoB, 9% in type IIa and 14% in type IIb, whereas they made up 25% in type IV.

. | Type . | . | |||||
---|---|---|---|---|---|---|---|

. | I . | IIa . | IIb . | III . | IV . | V . | Normolipidaemic . |

n | 16 | 736 | 371 | 38 | 509 | 101 | 407 |

Age (years) | 33.5±14.3 | 38.4±19.8 | 38.8±16.7 | 44.2±11 | 45.0±13.7 | 38.9±13.3 | 43.6±19.2 |

Gender (n) (male/female) | 10/6 | 418/318 | 181/190 | 28/10 | 378/131 | 81/20 | 291/116 |

Plasma cholesterol (mmol/l) | 4.88±2.29^{b,c,d,f} | 6.69±1.35^{a,c,d,e,f,g} | 7.06±1.74^{a,b,d,e,f,g} | 8.97±2.67^{a,b,c,e,f,g} | 5.67±1.36^{b,c,d,f,g} | 8.23±4.17^{a,b,c,d,e,g} | 4.85±0.85^{b,c,d,e,f} |

CM-C (mmol/l) | 3.36±2.90 | 0.4 | 0 | 0 | 0 | 4.75±3.39 | 0 |

VLDL-C (mmol/l) | 0.23±0.13^{c,d,e,f} | 0.45±0.25^{c,d,e,f} | 1.27±0.82^{a,b,d,e,f,g} | 5.59±2.94^{a,b,c,e,f,g} | 2.13±1.44^{a,b,c,d,f,g} | 1.66±1.20^{a,b,c,d,e,g} | 0.54±0.29^{c,d,e,f} |

LDL-C (mmol/l) | 0.73±0.58^{b,c,d,e,f,g} | 5.03±1.35^{a,c,d,e,f,g} | 4.81±1.49^{a,b,d,e,f,g} | 2.31±0.65^{a,b,c,f,g} | 2.68±0.83^{a,b,c,f,g} | 1.56±1.19^{a,b,c,d,e,g} | 3.15±0.67^{a,b,c,d,e,f} |

HDL-C (mmol/l) | 0.41±0.17^{b,c,d,e,f,g} | 1.21±0.32^{a,c,d,e,f,g} | 0.97±0.23^{a,b,e,f,g} | 0.91±0.24^{a,b,f,g} | 0.84±0.23^{a,b,c,f,g} | 0.61±0.18^{a,b,c,d,e,g} | 1.16±0.38^{a,b,c,d,e,f} |

Non-HDL-C (mmol/l) | 4.47±2.32^{b,c,d,f} | 5.47±1.37^{a,c,d,e,f,g} | 6.09±1.75^{a,b,d,e,f,g} | 8.06±2.76^{a,b,c,e,g} | 4.84±1.36^{b,c,d,f,g} | 7.62±4.18^{a,b,c,e,g} | 3.69±0.78^{b,c,d,e,f} |

Plasma TG (mmol/l) | 19.04±12.95^{b,c,d,e,g} | 1.20±0.44^{a,c,d,e,f} | 2.84±1.43^{a,b,d,e,f,g} | 6.13±3.78^{a,b,c,e,f,g} | 4.67±2.98^{a,b,c,d,f,g} | 17.35±14.74^{b,c,d,e,g} | 1.38±0.52^{a,c,d,e,f} |

CM-TG (mmol/l) | 17.77±12.95^{e} | 0.27 | 0.62±0.20 | 0.97 | 0.75±0.20^{a,f} | 13.34±14.50^{e} | 0 |

VLDL-TG (mmol/l) | 0.55±0.26^{c,d,e,f} | 0.66±0.38^{c,d,e,f,g} | 2.09±1.27^{a,b,d,e,f,g} | 5.18±3.69^{a,b,c,e,f,g} | 3.91±2.84^{a,b,c,d,f,g} | 2.92±1.88^{a,b,c,d,e,g} | 0.88±0.46^{b,c,d,e,f} |

LDL-TG (mmol/l) | 0.35±0.12^{c,f,g} | 0.32±0.11^{c,d,e,f,g} | 0.48±0.23^{a,b,d,e,g} | 0.42±0.11^{b,c,f,g} | 0.40±0.15^{b,c,f,g} | 0.50±0.23^{a,b,d,e,g} | 0.27±0.09^{a,b,c,d,e,f} |

HDL-TG (mmol/l) | 0.34±0.09^{b,c,d,f,g} | 0.21±0.05^{a,c,d,e,f,g} | 0.27±0.07^{a,b,d,e,f,g} | 0.39±0.09^{a,b,c,e,f,g} | 0.33±0.11^{b,c,d,f,g} | 0.48±0.21^{a,b,c,d,e,g} | 0.23±0.05^{a,b,c,d,e,f} |

Plasma apoB (g/l) | 0.48±0.16^{b,c,d,e,f,g} | 1.36±0.29^{a,c,d,e,f,g} | 1.53±0.36^{a,b,d,e,f,g} | 1.04±0.21^{a,b,c,e} | 1.18±0.23^{a,b,c,d,f,g} | 0.96±0.40^{a,b,c,e} | 0.99±0.21^{a,b,c,e} |

VLDL-apoB (g/l) | 0.07±0.04^{b,c,d,e,f,g} | 0.12±0.06^{a,c,d,e,f} | 0.21±0.09^{a.b.d.e.f.g} | 0.44±0.18^{a,b,c,e,f,g} | 0.29±0.13^{a,b,c,d,f,g} | 0.26±0.17^{a,b,c,d,e,g} | 0.12±0.05^{a,c,d,e,f} |

LDL-apoB (g/l) | 0.41±0.15^{b,c,d,e,f,g} | 1.24±0.28^{a,c,d,e,f,g} | 1.32±0.32^{a,b,d,e,f,g} | 0.59±0.15^{a,b,c,e,f,g} | 0.89±0.22^{a,b,c,d,f} | 0.70±0.32^{a,b,c,d,e,g} | 0.87±0.19^{a,b,c,d,f} |

HDL-apoA-I (g/l) | 0.80±0.15^{b,c,d,e,f,g} | 1.25±0.23^{a,c,e,f,g} | 1.19±0.20^{a,b,e,f,g} | 1.25±0.24^{a,e,f} | 1.13±0.21^{a,b,c,d,f,g} | 1.04±0.27^{a,b,c,d,e,g} | 1.22±0.25^{a,b,c,e,f} |

VLDL-C/VLDL-apoB (g/l) | 4.24±3.01^{d,e,f} | 3.91±2.22^{c,d,e,f,g} | 6.09±2.06^{b,d,e,f,g} | 12.37±3.13^{a,b,c,e,f,g} | 7.13±2.14^{a,b,c,d,f,g} | 8.84±12.11^{a,b,c,d,e,g} | 4.80±4.25^{b,c,d,e,f} |

LDL-C/LDL-apoB (g/l) | 1.75±0.88^{b,c,d,e,f,g} | 4.04±0.36^{a,c,e,f,g} | 3.65±0.38^{a,b,d,e,f} | 3.92±0.67^{a,c,e,f,g} | 3.01±0.51^{a,b,c,d,f,g} | 2.00±0.80^{a,b,c,d,e,g} | 3.64±0.3^{9a,b,d,e,f} |

VLDL-TG/VLDL-apoB (g/l) | 10.40±6.85^{f} | 5.86±3.29^{c,d,e,f,g} | 10.18±3.78^{b,e,f,g} | 11.08±4.46^{b,f,g} | 13.09±4.91^{b,c,f,g} | 19.11±35.98^{a,b,c,d,e,g} | 7.97±8.93^{b,c,d,e,f} |

LDL-TG/LDL-apoB (g/l) | 0.85±0.28^{b,c,d,e,f,g} | 0.26±0.07^{a,c,d,e,f,g} | 0.35±0.11^{a,b,d,e,f,g} | 0.74±0.21^{a,b,c,e,g} | 0.47±0.17^{a,b,c,d,f,g} | 0.77±0.32^{a,b,c,e,f,g} | 0.31±0.10^{a,b,c,d,e,f} |

. | Type . | . | |||||
---|---|---|---|---|---|---|---|

. | I . | IIa . | IIb . | III . | IV . | V . | Normolipidaemic . |

n | 16 | 736 | 371 | 38 | 509 | 101 | 407 |

Age (years) | 33.5±14.3 | 38.4±19.8 | 38.8±16.7 | 44.2±11 | 45.0±13.7 | 38.9±13.3 | 43.6±19.2 |

Gender (n) (male/female) | 10/6 | 418/318 | 181/190 | 28/10 | 378/131 | 81/20 | 291/116 |

Plasma cholesterol (mmol/l) | 4.88±2.29^{b,c,d,f} | 6.69±1.35^{a,c,d,e,f,g} | 7.06±1.74^{a,b,d,e,f,g} | 8.97±2.67^{a,b,c,e,f,g} | 5.67±1.36^{b,c,d,f,g} | 8.23±4.17^{a,b,c,d,e,g} | 4.85±0.85^{b,c,d,e,f} |

CM-C (mmol/l) | 3.36±2.90 | 0.4 | 0 | 0 | 0 | 4.75±3.39 | 0 |

VLDL-C (mmol/l) | 0.23±0.13^{c,d,e,f} | 0.45±0.25^{c,d,e,f} | 1.27±0.82^{a,b,d,e,f,g} | 5.59±2.94^{a,b,c,e,f,g} | 2.13±1.44^{a,b,c,d,f,g} | 1.66±1.20^{a,b,c,d,e,g} | 0.54±0.29^{c,d,e,f} |

LDL-C (mmol/l) | 0.73±0.58^{b,c,d,e,f,g} | 5.03±1.35^{a,c,d,e,f,g} | 4.81±1.49^{a,b,d,e,f,g} | 2.31±0.65^{a,b,c,f,g} | 2.68±0.83^{a,b,c,f,g} | 1.56±1.19^{a,b,c,d,e,g} | 3.15±0.67^{a,b,c,d,e,f} |

HDL-C (mmol/l) | 0.41±0.17^{b,c,d,e,f,g} | 1.21±0.32^{a,c,d,e,f,g} | 0.97±0.23^{a,b,e,f,g} | 0.91±0.24^{a,b,f,g} | 0.84±0.23^{a,b,c,f,g} | 0.61±0.18^{a,b,c,d,e,g} | 1.16±0.38^{a,b,c,d,e,f} |

Non-HDL-C (mmol/l) | 4.47±2.32^{b,c,d,f} | 5.47±1.37^{a,c,d,e,f,g} | 6.09±1.75^{a,b,d,e,f,g} | 8.06±2.76^{a,b,c,e,g} | 4.84±1.36^{b,c,d,f,g} | 7.62±4.18^{a,b,c,e,g} | 3.69±0.78^{b,c,d,e,f} |

Plasma TG (mmol/l) | 19.04±12.95^{b,c,d,e,g} | 1.20±0.44^{a,c,d,e,f} | 2.84±1.43^{a,b,d,e,f,g} | 6.13±3.78^{a,b,c,e,f,g} | 4.67±2.98^{a,b,c,d,f,g} | 17.35±14.74^{b,c,d,e,g} | 1.38±0.52^{a,c,d,e,f} |

CM-TG (mmol/l) | 17.77±12.95^{e} | 0.27 | 0.62±0.20 | 0.97 | 0.75±0.20^{a,f} | 13.34±14.50^{e} | 0 |

VLDL-TG (mmol/l) | 0.55±0.26^{c,d,e,f} | 0.66±0.38^{c,d,e,f,g} | 2.09±1.27^{a,b,d,e,f,g} | 5.18±3.69^{a,b,c,e,f,g} | 3.91±2.84^{a,b,c,d,f,g} | 2.92±1.88^{a,b,c,d,e,g} | 0.88±0.46^{b,c,d,e,f} |

LDL-TG (mmol/l) | 0.35±0.12^{c,f,g} | 0.32±0.11^{c,d,e,f,g} | 0.48±0.23^{a,b,d,e,g} | 0.42±0.11^{b,c,f,g} | 0.40±0.15^{b,c,f,g} | 0.50±0.23^{a,b,d,e,g} | 0.27±0.09^{a,b,c,d,e,f} |

HDL-TG (mmol/l) | 0.34±0.09^{b,c,d,f,g} | 0.21±0.05^{a,c,d,e,f,g} | 0.27±0.07^{a,b,d,e,f,g} | 0.39±0.09^{a,b,c,e,f,g} | 0.33±0.11^{b,c,d,f,g} | 0.48±0.21^{a,b,c,d,e,g} | 0.23±0.05^{a,b,c,d,e,f} |

Plasma apoB (g/l) | 0.48±0.16^{b,c,d,e,f,g} | 1.36±0.29^{a,c,d,e,f,g} | 1.53±0.36^{a,b,d,e,f,g} | 1.04±0.21^{a,b,c,e} | 1.18±0.23^{a,b,c,d,f,g} | 0.96±0.40^{a,b,c,e} | 0.99±0.21^{a,b,c,e} |

VLDL-apoB (g/l) | 0.07±0.04^{b,c,d,e,f,g} | 0.12±0.06^{a,c,d,e,f} | 0.21±0.09^{a.b.d.e.f.g} | 0.44±0.18^{a,b,c,e,f,g} | 0.29±0.13^{a,b,c,d,f,g} | 0.26±0.17^{a,b,c,d,e,g} | 0.12±0.05^{a,c,d,e,f} |

LDL-apoB (g/l) | 0.41±0.15^{b,c,d,e,f,g} | 1.24±0.28^{a,c,d,e,f,g} | 1.32±0.32^{a,b,d,e,f,g} | 0.59±0.15^{a,b,c,e,f,g} | 0.89±0.22^{a,b,c,d,f} | 0.70±0.32^{a,b,c,d,e,g} | 0.87±0.19^{a,b,c,d,f} |

HDL-apoA-I (g/l) | 0.80±0.15^{b,c,d,e,f,g} | 1.25±0.23^{a,c,e,f,g} | 1.19±0.20^{a,b,e,f,g} | 1.25±0.24^{a,e,f} | 1.13±0.21^{a,b,c,d,f,g} | 1.04±0.27^{a,b,c,d,e,g} | 1.22±0.25^{a,b,c,e,f} |

VLDL-C/VLDL-apoB (g/l) | 4.24±3.01^{d,e,f} | 3.91±2.22^{c,d,e,f,g} | 6.09±2.06^{b,d,e,f,g} | 12.37±3.13^{a,b,c,e,f,g} | 7.13±2.14^{a,b,c,d,f,g} | 8.84±12.11^{a,b,c,d,e,g} | 4.80±4.25^{b,c,d,e,f} |

LDL-C/LDL-apoB (g/l) | 1.75±0.88^{b,c,d,e,f,g} | 4.04±0.36^{a,c,e,f,g} | 3.65±0.38^{a,b,d,e,f} | 3.92±0.67^{a,c,e,f,g} | 3.01±0.51^{a,b,c,d,f,g} | 2.00±0.80^{a,b,c,d,e,g} | 3.64±0.3^{9a,b,d,e,f} |

VLDL-TG/VLDL-apoB (g/l) | 10.40±6.85^{f} | 5.86±3.29^{c,d,e,f,g} | 10.18±3.78^{b,e,f,g} | 11.08±4.46^{b,f,g} | 13.09±4.91^{b,c,f,g} | 19.11±35.98^{a,b,c,d,e,g} | 7.97±8.93^{b,c,d,e,f} |

LDL-TG/LDL-apoB (g/l) | 0.85±0.28^{b,c,d,e,f,g} | 0.26±0.07^{a,c,d,e,f,g} | 0.35±0.11^{a,b,d,e,f,g} | 0.74±0.21^{a,b,c,e,g} | 0.47±0.17^{a,b,c,d,f,g} | 0.77±0.32^{a,b,c,e,f,g} | 0.31±0.10^{a,b,c,d,e,f} |

The correspondence between non-HDL-C and apoB can be expressed both as correlation and concordance. The overall relationship between non-HDL-C and apoB is shown in Figure 1. A strong positive correlation (*r*=0.78) was evident. Nevertheless, there were a substantial number of outliers, and this resulted in the fact that the concordance between the two was only moderate. Accordingly, we separated the overall group into the different lipid phenotypes. The relationships for the normolipidaemic subjects and those with type IIa and type IIb are shown in Figure 2. In each case, the correlation was high (>0.9), whereas concordance was only moderate (Table 2).

### Correlation between plasma apoB and non-HDL-C levels in the whole cohort

### Correlation between plasma apoB and non-HDL-C levels in subjects with normal phenotype, type IIa and type IIb

Type . | κ Statistic* . | Standard error . | Agreement† . |
---|---|---|---|

Normal | 0.53 | 0.03 | Moderate |

I | − | − | − |

IIa | 0.56 | 0.02 | Moderate |

IIb | 0.50 | 0.03 | Moderate |

III | −0.01 | 0.02 | Absent |

IV | 0.38 | 0.03 | Fair |

V | 0.08 | 0.04 | Slight |

Type . | κ Statistic* . | Standard error . | Agreement† . |
---|---|---|---|

Normal | 0.53 | 0.03 | Moderate |

I | − | − | − |

IIa | 0.56 | 0.02 | Moderate |

IIb | 0.50 | 0.03 | Moderate |

III | −0.01 | 0.02 | Absent |

IV | 0.38 | 0.03 | Fair |

V | 0.08 | 0.04 | Slight |

The results for the patients with type I, III and V are shown in Figure 3 and summarized in Table 2. Marked variance between non-HDL-C and apoB was evident. The findings in type IV are shown in Figure 4 and Table 2. Once again, there was considerable variance in the relationship between non-HDL-C and apoB. Correlation was moderately high (*r*=0.74), but concordance was only fair. Within the group of patients with type IV hyperlipoproteinaemia, agreement related to the plasma TG levels. Thus for those with levels <;3.0 mmol/l (*n*=167), correlation was high (*r*=0.9) and concordance was moderate (κ=0.41), whereas, for those with levels >3.0 mmol/l (*n*=342), correlation was lower (*r*=0.65) and concordance only fair (κ=0.35). The differences between those with mild compared with moderate hypertriglyceridaemia are easy to appreciate in Figure 5. In those with mild hypertriglyceridaemia (TG elevated for age and gender but <;3 mmol/l), there appeared to be a simple linear relationship between apoB and non-HDL-C. In contrast, in those with moderate hypertriglyceridaemia (TG <;3.0 mmol/l), the relationship was more complex and two subgroups appeared to be present. In one subgroup, which encompassed the majority of the subjects, there was a linear relationship, although with considerable scatter, between non-HDL-C and apoB. There was a second subgroup present, in which the subjects were characterized by a non-HDL-C that appears to be disproportionately elevated compared with apoB.

### Correlation between plasma apoB and non-HDL-C levels in subjects with type I, type III and type V

### Correlation between plasma apoB and non-HDL-C levels in subjects with type IV

### Correlation between plasma apoB and non-HDL-C levels in subjects with TG <; or ≥3 mmol/l

## DISCUSSION

Our present findings confirm that, overall, there is a strong positive correlation between non-HDL-C and apoB; however, correlation is only one measure of agreement. To be clinically equivalent, two parameters must not only be highly correlated, but they must be highly concordant. In this case, for the overall group, the two parameters were only moderately concordant. Correlation relates the degree of change in one parameter to the degree of change in another. Concordance, on the other hand, expresses the variance in one parameter for any given value of the other. Moderate concordance means that, for any given value of one parameter, there is a relatively wide range of values of the other. Diagnosis and therapy of dyslipidaemia is triggered by specific values of the parameter selected. If two values are only moderately concordant, clinical decisions will be determined by which parameter has been selected. Thus what is measured will often govern therapy.

Our overall finding is consistent with previous results [9,10]; however, the present study is the first to examine the strength of the relationship between non-HDL-C and apoB among different dyslipidaemic phenotypes. Very high correlation was evident in normolipidaemic individuals and those with type IIa and IIb. However, even in these groups, concordance was only moderate. In the other dyslipoproteinaemias, correlation was moderately high in type III, but concordance was low, whereas, in type I and type V, correlation and concordance were both low. In type I, non-HDL-C did not differ from the normolipidaemic group, but apoB was half the level. The highest non-HDL-C values were in type III and type V; however, the average apoB in these two groups did not differ significantly from the normal group. In these three groups, therefore, there is extreme discordance between non-HDL-C and apoB. It might be argued that types I, III and V are sufficiently unusual that they are not important exceptions to the rule. Not so with type IV, which is one of the most common of the clinical dyslipoproteinemias and the most common in patients with vascular disease [11]. In type IV, correlation is only moderate and concordance is low. Make type IV an exception and there is no rule. It is true that discordance in type IV is more pronounced in the subgroup with the higher TG levels, but clinical rules must hold in the more trying cases and not just the simple ones. Thus overall and among the specific dyslipoproteinemias, there is such substantial variance between non-HDL-C and apoB that the two are not equivalent for clinical practice.

The obvious question is whether non-HDL-C or apoB relates more closely to the risk of vascular disease. If one is clearly superior to the other, then all difficulty disappears. If apoB and LDL-C are compared, the answer is straightforward. Many prospective epidemiological studies and clinical trials have reported that apoB is superior to LDL-C as an index of the risk of vascular disease (for a review, see [2]). Superiority has been demonstrated both by conventional regression analyses as well as by ROC (receiver operating characteristic) curve analyses. Unfortunately, the answer is not as simple and neat when apoB and non-HDL-C are weighed against each other. Although a number of studies have shown apoB to be superior to non-HDL-C, others have shown them to be equivalent, particularly when results of ROC curves are taken as decisive [2,12,13]. Importantly, none have shown non-HDL-C to be superior to apoB. The weight of epidemiological evidence therefore favours apoB.

Furthermore, the results of the present study indicate that assessing the value of highly correlated parameters by summary results may be overly simplistic, no matter how sophisticated the statistical test. Almost always, patient groups are metabolically and clinically heterogeneous, not homogeneous, in composition. Disease may alter either plasma lipoprotein number or composition and, depending on which occurs and to what extent, the tightness of the relationship between non-HDL-C and apoB may or may not be altered. That is, in fact, what our findings demonstrate. In normolipidaemic patients and those with type IIa and type IIb hyperlipoproteinaemia, there is high correlation between non-HDL-C and apoB. In all of these disorders, VLDL constitutes only a small minority of the total number of apoB lipoprotein particles, whereas, in the others, in all of whom correlation was much lower or non-existent, the lipoprotein makeup of the *d*<1.006 g/ml was much more complex.

Thus, in type I hyperlipoproteinaemia, CM particles that are rich in TG but relatively depleted in cholesterol compared with normal VLDL are characteristic, whereas, in type III hyperlipoproteinaemia, cholesterol-rich remnant VLDL and CM cholesterol-rich particles are hallmarks of the disorder, but, in type V, an unpredictable assortment of CMs and VLDL particles are present. Moreover, in all three, the proportion of LDL particles of the total number of apoB particles is much less than in the normolipidaemic subjects or those with either form of type II hyperlipoproteinaemia. Thus marked variance between apoB and non-HDL-C in types I, III and V would be anticipated and that is what we found.

That correlation between non-HDL-C and apoB is low in type IV but not in type IIb is surprising; however, on average, of the total apoB, VLDL particles make up a smaller proportion of the total in type IIB than type IV, indicating that conversion of VLDL into LDL particles is less effective in type IV than in type IIB. Less effective conversion would be expected to result in more remnants with longer half-lives and the finding of a higher VLDL-C/apoB ratio in type IV than type IIB is further evidence in favour of this concept. Our findings demonstrate that type IV hyperlipoproteinaemia is heterogeneous. In those with mild hypertriglyceridaemia, although concordance is low, there is, overall, a clear relationship between non-HDL-C and apoB. With higher TG levels, not only is variance increased, but there appears to be a number of subjects that distinctly differ from the rest. These subjects have a non-HDL-C level that is disproportionately high compared with apoB. This finding raises the possibility that at least a subgroup of subjects with type IV hyperlipoproteinaemia might have a substantial number of remnant lipoproteins, and that this subgroup might represent an intermediate and perhaps clinically discrete form of type III hyperlipoproteinaemia. Further studies are required to test this hypothesis. In the interim, the results of the present study establish that important differences exist in the relationship between apoB and non-HDL-C in the different dyslipoproteinaemias.

## Abbreviations

We are grateful to the patients for their generous collaboration and invaluable contribution. J.-C.H. is recipient of a Doctoral Research Award from the Heart and Stroke Foundation of Canada (HSFC). P.C. is recipient of a Research Award from the Fonds de la Recherche en Santé du Québec (FRSQ).