addition, polymorphisms of the apoB, apoC, apoA-IV and lipoprotein lipase LPL genes have been related to
the extent of cholesterol response [14 – 18], though fewer data are available and the findings are less consistent
than those relating to the apoE genotypes. We report here on predictors of plasma total cholesterol change in
a group of 55 individuals during four dietary periods of high or low saturated fat intake. More extensive geno-
typing than had been undertaken previously in a single study has enabled analysis of the extent to which a
range of polymorphisms contribute to variation in cholesterol response to alterations in the nature of
dietary fat in free-living individuals. Particular attention was paid to effects of variation in the cholesterol ester
transfer protein CETP and LPL genes as plasma CET activity and the LPL HindIII polymorphisms were as-
sociated with response of plasma cholesterol to changes in the type of dietary fat in our earlier study [4,18]. It
was also possible to estimate the effect of compliance to dietary advice on variation in response among free
living individuals.
2. Methods
2
.
1
. Participants Sixty two people aged 26 – 64 years, with plasma
cholesterol concentration 5.5 – 7.9 mmoll and triglyce- ride concentration below 3 mmoll, were recruited from
a group of 300 volunteers who responded to advertise- ments in the widely read local daily newspaper. None
were suffering from chronic medical conditions or tak- ing drugs known to influence lipid metabolism. Fifty-
five individuals 23 men and 32 women completed the 20-week study. Four males and three females withdrew
from the study. Their age, baseline total and low-den- sity lipoprotein LDL cholesterol were not significantly
different from those individuals who completed the study. Participants continued their usual activities
throughout the experimental period. Ethical approval was obtained from Southern Regional Health Author-
ity and informed consent gained from each participant.
2
.
2
. Experimental design The study involved a randomised double crossover
trial of two dietary interventions: one high in saturated fatty acids SAFA, the other high in polyunsaturated
fatty acids PUFA. After recruitment, participants recorded their usual diet for a 4-day period. This infor-
mation was used to plan the experimental diets. They were then asked to follow a standard lipid-lowering diet
total and saturated fatty acids providing around 30 and 10 of total energy respectively for 3 weeks
baseline diet. Thereafter they were randomised to one of two groups following the dietary sequences, SAFA –
PUFA – SAFA – PUFA or
PUFA – SAFA – PUFA – SAFA, each dietary phase continuing for 4 weeks
without washout periods. Weight was recorded and a fasting blood sample collected on two occasions, 1 day
apart, during the final week of the baseline and each experimental diet. Participants were also asked to com-
plete a 3-day estimated diet record of all food and drink consumed on the Thursday, Friday and Saturday pre-
ceding the week during which blood samples were taken. The diet records were analysed using the ‘Diet
Cruncher’ programme [19].
3. Experimental diets
Participants’ diets were individually planned on the basis of personal food preferences and usual energy
requirements calculated from the 4-day diet record completed following recruitment into the study. On
both the SAFA and PUFA diets, fat sources relevant to the experimental diets butter and hardened coconut oil
on SAFA and polyunsaturated margarine and safflower oil on diet PUFA were provided free of charge and
added to the relatively low fat baseline diet so that protein provided about 15 energy, carbohydrate 44
and total fat 36, but fat composition differed. In the SAFA diet, 20 of energy came from saturated fatty
acids and 3 from polyunsaturated fatty acids. In the PUFA diet, 10 of energy came from saturated fatty
acids and 10 from polyunsaturated fatty acids. Mo- nounsaturated fatty acids remained constant in both
diets at about 10 total energy. Exchange lists were provided to enable participants to select appropriate
foods. Detailed instructions, menus, and recipes were provided and reinforced during regular interviews and
telephone calls.
4. Dietary compliance
Since the dietary changes involved alteration in in- takes of both saturated and polyunsaturated fatty
acids, dietary compliance was assessed by two methods. Difference in intake of polyunsaturated fatty acids be-
tween PUFA and SAFA was assessed by the absolute change in mole percent of linoleic acid in plasma
triglyceride fatty acids change in plasma triglyceride linoleate. Since changes in blood levels of saturated
fatty acids are less responsive to dietary change and a large proportion of saturated fatty acids are derived
from endogenous sources, alterations in saturated fat content measured as a percentage of total energy were
assessed from the 3-day diet records as the change between the SAFA and PUFA experimental periods
change in SAFA. Both measures were included as continuous measures in the statistical analyses.
5. Laboratory analysis
Blood specimens were centrifuged at 2500 × g for 15 min. Before cholesterol analysis, the plasma was spun
at 8000 × g for 4 min in a bench top Zentifuge 3200, to remove any fibrin that had been formed on storage.
HDL cholesterol was measured in the supernatant after precipitation of apolipoprotein B apoB-containing
lipoproteins in plasma with phosphotungstic acidmag- nesium chloride solution [20]. Cholesterol concentration
of plasma and lipoprotein fractions were measured, using the Boehringer enzymatic colourimetric method
with a Boehringer CHOD PAP enzymatic colourimetric reagents [21]. Measurement of plasma triglycerides was
conducted using Roche Triglycerides PAP. Coefficient of variation was 1.3 for cholesterol and 2.6 for
triglyceride measurements in the Royal Australasian College of Pathologists’ quality assurance programme.
Plasma LDL cholesterol was calculated using the Friedewald
equation [22].
Apolipoprotein B
and apolipoprotein A-1 levels were measured by an im-
munoturbidimetric method using Boehringer kits co- efficient of variation 5 and 3 respectively. Plasma
cholesteryl ester transfer CET activity was assessed by measuring the transfer of newly synthesised cholesterol
esters from HDL to apoB-containing lipoproteins using an isotopic assay coefficient of variation, 10 [23].
Values from this assay correlate closely with cholesterol ester mass transfer measured by chemical methods [24].
An estimate of small dense LDL levels was obtained by ultracentrifuging EDTA plasma adjusted to d =
1.040 gml for 48 h at 40 000 rpm in a Beckman 50.3 Ti rotor, measuring cholesterol in the d \ 1.040 gml
plasma fraction, subtracting the sum of plasma HDL cholesterol and Lpa cholesterol calculated as 34 of
Lpa mass then converting dense LDL cholesterol to lipoprotein mass using the chemical composition of
dense LDL d = 1.040 – 1.054 gml [25].
Plasma triglyceride fatty acid composition was deter- mined for each of the experimental periods using an
adaptation of the Bligh and Dwyer method [26]. The internal standard employed was C17:0. The triglyceride
fatty acids were separated using a Hewlett Packard HP5890 gas chromatograph GC equipped with a DB
225 megabore column and detected by flame ionisation. The following conditions were maintained during GC
operation: oven temperature, 200°C; detector and injec- tor temperature 250°C; helium gas flow 6.5 mlmin;
split ratio, 5:1. Triglyceride fatty acids were identified by matching retention times with commercial standards
Nu-Check Prep. The precision of the fatty acid analy- sis was determined by repeated analysis of a pooled
plasma sample. The coefficient of variation for triglyce- ride linoleate was 6.4.
5
.
2
.
1
. DNA preparation and analysis DNA was extracted from frozen whole blood, col-
lected after an overnight fast, by the salting out method [27]. Genotype was determined by the polymerase chain
reaction PCR using previously published methods and amplifying oligonucleotides. Taq polymerase was ob-
tained from Bethesda Research Laboratories and reac- tions were performed on a Hybaid Intelligent Heating
Block. ApoCIII – C1100-T genotype was carried out by amplification and the use of allele specific oligonucle-
otides [28]. ApoB signal peptide genotype was deter- mined by direct electrophoresis of the amplified
fragments on a 10 acrylamide gel [29]. CETP geno- type was determined by amplifying a 535 bp fragment
of intron 1 of the CETP gene by PCR followed by Taq I digestion, and separation of fragments on MADGE
polyacrylamide gels [30]. Apolipoprotein E genotyping was carried out by a HhaI digestion, and primers were
the same as previously described [31]. LPL HindIII genotyping was undertaken using restriction enzyme
digestion and separation of the fragments on an agarose gel [32]. The LPL S447X polymorphism in
exon 9 was identified by the introduction of a forced HindfI restriction enzyme site into the PCR product
[33].
5
.
3
. Statistical methods Statistical analysis was performed using the statistical
package STATA. The mean of the measurements made on the 2 consecutive days at the end of each diet period
was taken as the value for that phase of the experiment. Although
the study
involved a
double dietary
crossover, of interest were the three measures of differ- ence in cholesterol and their association with the three
corresponding measures of dietary compliance and other predictors of extent of cholesterol change. The
correlations between difference in cholesterol on the SAFA and PUFA diets and the continuous variables
were computed using the method described by Bland and Altman [34], to adjust for multiple observations for
each person. Means for the change in cholesterol for the different genotypes were also adjusted for multiple
observations using the robust standard error proce- dures provided in STATA [35]. Multiple regression
analysis was used to examine the association between the change in cholesterol and the genotypes of interest
after adjusting for measures of compliance and screen- ing cholesterol. Genetic polymorphisms examined in
Table 5 had previously been shown to be predictive of individual cholesterol response or coded for enzymes
important in lipoprotein metabolism. To be considered for inclusion in the multiple regression model, variables
had to have a P value of less than 0.15 in univariate
analyses. Other metabolic variables for which prior hypotheses existed were also tested in multiple regres-
sion analyses. The observed average difference in total cholesterol between saturated and polyunsaturated fat
periods was compared with the predicted value on the basis of reported dietary change and the metaanalyses
of Clarke and colleagues [36].
6. Results