RE: Serum Lipids, Lipoproteins, and Risk of Breast Cancer: A Nested Case-Control Study Using Multiple Time Points

In a previous publication in the Journal, we showed that high-density lipoprotein-cholesterol (HDL-C) and apolipoprotein A1 (apoA1) levels were positively associated with breast cancer (BC) risk while non-HDL-C and apolipoprotein B (apoB) levels were negatively associated with BC risk (1). These associations were adjusted for most breast cancer risk factors but not for percent mammographic density (PMD) or alcohol intake as these variables needed additional data extraction (2). Here we report these associations adjusting for both PMD and alcohol intake.

The methods used have been described in our earlier paper (1). This case-control study was nested within the cohort of the Canadian Diet and Breast Cancer Prevention Study, a multicenter randomized controlled trial designed to test whether a reduction in dietary fat intake would reduce the incidence of BC in women with extensive PMD. Subjects provided a nonfasting blood sample at entry to the trial and annually thereafter.

We matched individually case subjects (n = 261) with two control subjects (n = 541) according to age (within one year), date of random assignment (within one year), study center, duration of follow-up (within six months), and the availability of blood samples. The dietary intervention did not have a statistically significant effect on BC incidence (3), and we combined the low-fat dietary intervention and comparison groups. PMD was measured in baseline mammograms using Cumulus software (4), and alcohol intake was assessed from food records collected from all subjects at intervals throughout the trial (average of 3.7 sets of three-day food records per subject).

To take advantage of the multiple blood samples and to adjust for other variables that could change over time, we calculated up to three subaverages of serum lipid measurements for each woman depending on menopausal status at the time of blood collection (1). Subaverages of weight and alcohol (grams/day) were calculated in the same manner.

We examined the association of serum lipid levels with risk of BC using generalized estimating equations analysis in which case-control status was the outcome variable and serum lipid levels (subaverages) were the independent variables. All P values are for two-sided statistical tests. A P value of less than .05 was considered statistically significant.

Table 1 shows selected baseline characteristics of the case and control subjects, which differ slightly from those shown in our previous paper because of missing data on alcohol intake (n = 2) or unavailable mammograms (n = 33).

Table 1.

Selected demographic characteristics, serum lipid variables, alcohol intake and percent mammographic density at baseline

Characteristics	Mean (SD) or %		P*
	Cases Subjects (n = 261)	Control Subjects (n = 541)
Group†
Intervention	51.3	49.4	.60
Comparison	48.7	50.7
Age, y	48.6 (6.3)	48.3 (6.3)	.57
Weight, kg	63.2 (8.8)	61.9 (7.5)	.04
Height, cm	163.5 (6.4)	163.3 (6.2)	.61
Body mass index, kg/m2	23.6 (2.6)	23.3 (2.3)	.09
Smoked ever, % yes	49.8	48.4	.71
Age at menarche, y	12.8 (1.3)	13.0 (1.4)	.08
Parous, % yes	68.2	69.9	.63
Age at birth of first child‡ (n = 178 378), y	26.1 (5.2)	25.4 (4.8)	.15
Number of live births‡ (n = 178 378)	2.17 (0.83)	2.25 (0.84)	.35
Postmenopausal, % yes	31.8	31.8	1.0
HRT used, % yes	28.0	22.2	.07
Family history of breast cancer§, % yes	22.2	17.2	.09
Total cholesterol, mmol/L	5.07 (0.90)	5.11 (0.91)	.58
HDL-C, mmol/L	1.42 (0.41)	1.42 (0.39)	.99
ApoA1, g/L	1.74 (0.28)	1.71 (0.26)	.23
Non-HDL-C‖, mmol/L	3.65 (1.0)	3.69 (0.97)	.61
ApoB, g/L	0.841 (0.207)	0.857 (0.219)	.34
Triglyceride, mmol/L	1.36 (0.78)	1.33 (0.69)	.62
Percent mammographic density	52.2 (15.1)	50.0 (13.7)	.06
Alcohol intake, g/d
Mean (SD)	8.67 (12.06)	6.81 (9.52)	.03
Median (IQR)	4.8 (13.6)	3.10 (10.8)	.03

*P value for case compared with control subjects for two sample t tests for continuous variables and chi-square tests for categorical variables. For alcohol, two sample t tests and Wilcoxon test were used. All statistical tests were two-sided. ApoA1 = apolipoprotein A1; ApoB = apolipoprotein B; HRT = hormone replacement therapy; HDL-C = high-density lipoprotein cholesterol; IQR = interquartile range.

†Random assignment group.

‡Among parous.

§At least one first-degree relative diagnosed with breast cancer.

‖Calculated as the difference between total cholesterol and HDL-C.

Table 2 shows the associations of lipids and lipoproteins with risk of BC after adjustment for other risk factors shown in the table footnote, and before and after adjustment for PMD and alcohol. Alcohol intake (P = .01) and PMD (P = .004) were both positively associated with risk of BC. HDL-C, apoA1, and non-HDL-C were statistically significantly associated with BC risk before but not after adjustment for PMD and alcohol. ApoB was statistically significantly and inversely associated with BC risk before (P = .007) and after adjustment for both PMD and alcohol (P = .03).

Table 2.

Association of serum lipids, alcohol intake, and baseline PMD with risk of breast cancer

							With PMD			With PMD and alcohol
	Original*			Subset*,†			Subset‡			Subset§
	(279 case subjects, 558 control subjects)			(261 case subjects, 541 control subjects)			(261 case subjects, 541 control subjects)			(261 case subjects, 541 control subjects)
Variable	Beta	SE	P	Beta	SE	P	Beta	SE	P	Beta	SE	P
Alcohol, g, and										0.0219	0.0086	.01
Baseline PMD (no lipids), %							0.0191	0.0063	.003	0.0185	0.0064	.004
HDL-C, mmol/L	0.4074	0.2048	.05	0.4096	0.2105	.05	0.3247	0.2107	.12	0.1817	0.2188	.41
Alcohol, g										0.0196	0.0089	.03
Baseline PMD, %							0.0178	0.0063	.005	0.0178	0.0063	.005
ApoA1, g/L	0.7136	0.3181	.02	0.7425	0.3283	.02	0.6556	0.3250	.04	0.4467	0.3285	.17
Alcohol, g										0.0183	0.0088	.04
Baseline PMD, %							0.0180	0.0063	.004	0.0179	0.0063	.005
Non HDL-C, mmol/L	−0.1919	0.0874	.03	−0.1908	0.0875	.03	−0.1629	0.0893	.07	−0.1359	0.0898	.13
Alcohol, g										0.0203	0.0086	.02
Baseline PMD, %							0.0180	0.0063	.005	0.0177	0.0064	.006
ApoB, g/L	−1.0099	0.384	.01	−1.0332	0.3851	.007	−0.9178	0.3869	.02	−0.8236	0.3900	.03
Alcohol, g										0.0202	0.0087	.02
Baseline PMD, %							0.0178	0.0063	.005	0.0174	0.0064	.006

*Generalized estimating equations adjusted for random assignment group (intervention, comparison), parity at baseline (parous, nonparous), if smoked ever at baseline (yes, no), if had first-degree relatives with breast cancer at baseline (yes, no), study site, age at menarche (years), age at birth of first child (years), number of live births, subaverage weight (kg), subaverage age (years), date of random assignment, menopausal status, and HRT use (three categories). ApoA1 = apolipoprotein A1; ApoB = apolipoprotein B; HDL-C = high density lipoprotein cholesterol; PMD = percent mammographic density.

†Subset with both alcohol and baseline PMD measurements available.

‡As in *, with addition of PMD.

§As in *, with the addition of PMD and alcohol.

The previously reported inverse association of ApoB with BC risk is not because of confounding by alcohol or PMD. Further investigation of the relationship between serum lipids and BC risk, including the effects of long-term statin use, appears to be warranted (5).

Note

Clinical Trial registration: Clinicaltrials.gov. Identifier: NCT00148057.