New version 7/5-2003: LDL-cholesterol changed.

Participants: Morten Lindberg, Gustav Mikkelsen and Arne Åsberg
Laboratory of medical biochemistry
Trondheim University Hospital
Trondheim, Norway

Date of report: April 15, 2003

Revised: April 28, 2003; May 7, 2003

N=1448

Age 18-90, median 46 years

Median body mass index: 23.0

Fasting for 1-21, median 11 hours (the distribution is bimodal)

N=1285

Age 18-89, median 47 years

Median body mass index: 24.7

Fasting for 1-20, median 11 hours (the distribution is bimodal)

N=2

The distribution of the cholesterol values for 1448 women and 1285 men is shown below.

The distribution is approximately gaussian. A linear regression model with cholesterol as the dependent variable and age, sex, body mass index, tobacco use (converted to a score of 1, 2 or 3), alcohol use (converted to a score of 1, 2 or 3) and hours since the last meal as the predictor variables showed that only age, sex, and body mass index are statistically significant predictor variables of the cholesterol concentration.

The univariate effect of body mass index is illustrated below (the least squares regression line is also drawn).

We do not think that partitioning according to body mass index is appropriate.

Therefore, we considered only age and sex as partitioning criteria.

Given below are plots of cholesterol concentration against age for women and men (the least squares regression line is drawn in each plot).

Women and men

< 50 years: 3.2 – 6.8 mmol/L

> 50 years: 4.1 – 7.9 mmol/L

The differences between women and men are very small, so we agree with NORIP that partitioning by sex is not necessary.

We propose partitioning by age at 30 and 50 years. Partitioning by 50 years only leads to 5.34 % of the 674 persons younger than 30 years having cholesterol below the lower limit of 3.2 mmol/L, and 0.59 % having cholesterol above the higher limit of 6.8 mmol/L.

Using cholesterol data from the 674 persons below 30 years of age, 843 persons aged 30-49 and 1216 persons older than 49 years in the database, we found the reference intervals:

The differences between countries are relatively small, and probably not clinically significant.

Generally, the proposed limits are not very different from those given in Olesen H, ed. Kompendium i Laboratoriemedicin, Amtsrådsforeningen I Danmark, 1988, and in Stakkestad J, Åsberg A, ed. Brukerhåndbok i klinisk kjemi, Haugesund, Akademisk Fagforlag, 2002. Sex differences are given in the former reference, but not in the latter.

N=1379

Age 18-90, median 46 years

Median body mass index: 23.0

Fasting for 1-21, median 11 hours (the distribution is bimodal)

N=1222

Age 18-89, median 47 years

Median body mass index: 24.7

Fasting for 1-20, median 11 hours (the distribution is bimodal)

N=2

The distribution of the HDL-cholesterol values for 1379 women and 1222 men is shown below.

The distribution is not gaussian. However, the distribution of log₁₀(HDL-cholesterol) is approximately gaussian. A linear regression model with log₁₀(HDL-cholesterol) as the dependent variable and age, sex, body mass index, tobacco use (converted to a score of 1, 2 or 3), alcohol use (converted to a score of 1, 2 or 3) and hours since the last meal as the predictor variables showed that all these variables are statistically significant predictor variables of the HDL-cholesterol concentration. The effects of sex, age, and body mass index had the highest t-values.

The univariate effect of body mass index is illustrated below (the least squares regression line is also drawn).

We do not think that partitioning according to body mass index is appropriate.

The univariate effect of alcohol is illustrated below.

We do not think that partitioning according to alcohol use is appropriate.

The effect of hours since the last meal, and the effect of tobacco use are not clinically significant.

Therefore, we considered only age and sex as partitioning criteria.

Given below are plots of HDL-cholesterol concentration against age for women and men (the least squares regression line is drawn in each plot).

Women

1.03 – 2.60 mmol/L

Men

0.85 – 2.12 mmol/L

We agree with NORIP that the effect of age is relatively small. One might consider partitioning men at 30 years in addition to 50 years, because the higher limit is a bit too high for men younger than 30 years; however, this is of minor clinical significance. We also agree with NORIP that partitioning by sex is necessary. Using HDL-cholesterol data from the 1379 women and the 1222 men in the database, we found the reference intervals

which are the same as those proposed by NORIP.

The differences between countries are relatively small, and probably not clinically significant.

Generally, the proposed limits are markedly higher than those given in Olesen H, ed. Kompendium i Laboratoriemedicin, Amtsrådsforeningen I Danmark, 1988, and somewhat higher than the limits given in Stakkestad J, Åsberg A, ed. Brukerhåndbok i klinisk kjemi, Haugesund, Akademisk Fagforlag, 2002.

N=1374

Age 18-90, median 46 years

Median body mass index: 23.0

Fasting for 1-21, median 11 hours (the distribution is bimodal)

N=1208

Age 18-89, median 47 years

Median body mass index: 24.7

Fasting for 1-20, median 11 hours (the distribution is bimodal)

N=2

The distribution of the LDL-chlesterol values for 1374 women and 1208 men is shown below.

The distribution is not far from gaussian, and more gaussian than the distribution of log₁₀(LDL-cholesterol). A linear regression model with LDL-cholesterol as the dependent variable and age, sex, body mass index, tobacco use (converted to a score of 1, 2 or 3), alcohol use (converted to a score of 1, 2 or 3) and hours since the last meal as the predictor variables showed that only age, sex, and body mass index are statistically significant predictor variables of the LDL-cholesterol concentration.

The univariate effect of body mass index is illustrated below (the least squares regression line is also drawn).

We do not think that partitioning according to body mass index is appropriate.

Therefore, we considered only age and sex as partitioning criteria.

Given below are plots of LDL-cholesterol concentration against age for women and men (the least squares regression line is drawn in each plot).

Women and men

< 50 years: 1.3 – 4.6 mmol/L

> 50 years: 2.0 – 5.4 mmol/L

The differences between women and men are relatively small, so we agree with NORIP that partitioning by sex is not necessary.

We propose partitioning by age at 30 and 50 years. Partitioning by 50 years only leads to 4.22 % of the 640 persons younger than 30 years having LDL-cholesterol below the lower limit of 1.3 mmol/L, and 0.63 % having LDL-cholesterol above the higher limit of 4.6 mmol/L.

Using LDL-cholesterol data from the 640 persons below 30 years of age, 790 persons aged 30-49 and 1152 persons older than 49 years in the database, we found the reference intervals:

The differences between countries are relatively small and may not be clinically significant.

Generally, the proposed limits are somewhat lower than those given in Olesen H, ed. Kompendium i Laboratoriemedicin, Amtsrådsforeningen I Danmark, 1988, and in Stakkestad J, Åsberg A, ed. Brukerhåndbok i klinisk kjemi, Haugesund, Akademisk Fagforlag, 2002.

N=645

Age 18-90, median 48 years

Median body mass index: 23.3

Fasting for 12-21, median 13 hours

N=556

Age 18-85, median 50 years

Median body mass index: 24.8

Fasting for 12-20, median 13 hours

N=2

The distribution of the triglyceride values for 645 women and 556 men is shown below.

The distribution is clearly not gaussian. However, the distribution of log₁₀(triglycerides) is approximately gaussian. A linear regression model with log₁₀(triglycerides) as the dependent variable and age, sex, body mass index, tobacco use (converted to a score of 1, 2 or 3), alcohol use (converted to a score of 1, 2 or 3) and hours since the last meal as the predictor variables showed that age, sex, body mass index, and tobacco use are statistically significant predictor variables of the triglyceride concentration.

The univariate effect of body mass index is illustrated below (the least squares regression line is also drawn).

We do not think that partitioning according to body mass index is appropriate.

The effect of tobacco use is slight: The median triglycerides is 0.98 mmol/L, 0.94 mmol/L and 1,05 mmol/L, respectively, in the in groups smoking 0, 1-5, and more than 5 cigarettes per day.

Therefore, we considered only age and sex as partitioning criteria.

Given below are plots of triglyceride concentration against age for women and men (the least squares regression line is drawn in each plot).

Women and men

0.47 – 2.58 mmol/L

We agree with NORIP that partitioning by sex and age is unnecessary. Using triglyceride data from the 1203 persons in the database (included the two individuals with unknown sex), we found the reference interval:

which is almost the same as the interval proposed by NORIP.

The differences between countries are very small at the lower limit, and somewhat greater at the higher limit (Denmark showing the highest value and Sweden the lowest); these differences are probably not clinically significant.

Generally, the higher limit is lower than the higher limits for men given in Olesen H, ed. Kompendium i Laboratoriemedicin, Amtsrådsforeningen I Danmark, 1988, where sex differences are distinct. However, the NORIP limits are not far from those given in Stakkestad J, Åsberg A, ed. Brukerhåndbok i klinisk kjemi, Haugesund, Akademisk Fagforlag, 2002.