Back to the title page | E-thesis main page

6. Results

6.1 Associations between known risk factors and breast cancer (I)

Women who had delivered their first child under 30 years of age, had at least two children, or had ever used oral contraceptives had a lower risk of breast cancer than did other women (Figure 6). On the other hand, diagnosed breast cancer in first-degree relatives, other breast diseases earlier in their lifetime, or current smoking increased the risk. The most educated women had a relative risk of 1.4, and women who exercised physically once a week a relative risk of 0.7, but the associations were not statistically significant. No associations were found between breast cancer and factors related to menstrual cycle: age at menarche, age at menopause, or menopausal status.

6.2 Body-size indicators and risk of breast cancer (I)

Height was associated with risk of breast cancer. However, the association was not linear, since only the tallest women, in the fifth quintile, showed an increased risk. The relative risk of breast cancer was 1.8 (95% CI 0.8-4.2) for premenopausal women at least 169 cm tall and 2.3 (95% CI 1.1-4.6) for postmenopausal women at least 166 cm tall compared to those under 160 cm and 156 cm, respectively.

Weight and BMI were not associated with risk of breast cancer. Women whose waist-to-hip ratio (WHR) was high had an increased risk of breast cancer compared to those with a low WHR (Figure 7). The odds ratio was 4.6 (95% CI 2.0-10.7) for premenopausal women with a WHR at least 0.87 compared to women who had the lowest ratio (below 0.79). In postmenopausal women, the odds ratio between the highest (at least 0.89) and lowest quintiles of WHR (below 0.80) was 2.6 (95% CI 1.3-5.1). High body fat percent increased the risk of postmenopausal breast cancer in the highest quintile (OR 2.0, 95% CI 1.0-4.0) compared to the lowest one, whereas weight loss over ten kilograms after the age of 45 modestly decreased the risk (OR 0.6, 95% CI 0.3-1.2) when compared to figures for those who maintained their weight. Weight gain was related to risk of breast cancer neither in premenopausal nor in postmenopausal women.

Figure 6. Known risk and protective factors for breast cancer.

The proportion of ER-positive tumors was lower (47%) in premenopausal than in postmenopausal women (57%). No consistent associations were found between body-size indicators and estrogen receptor status (ER-, ER+ or ER++) of the tumor.

Figure 7. Selected body-size indicators and risk of breast cancer by menopausal status.

6.3 Quality of food frequency questionnaire (II)

The aim of the Validation Study (II) was to assess whether the FFQ used in the Kuopio Breast Cancer Study measured what it was intended to measure. Overall, 29 food groups and 36 nutrients were included in the analyses, but in this section the focus is on the dietary factors (17 food groups and 14 nutrients) that were further examined in Studies III-IV.

Mean food consumption and nutrient intakes in the FFQ were generally lower than in the 14-day diet record (Figure 8). The consumption of sour milk, sugar, soft margarine, coffee, and tea were underestimated by more than 20% in the FFQ compared to figures in the diet records, whereas overestimation was more than 20% for vegetables and low-fat milk. Alcohol was underestimated by more than 20%, whereas overestimation was above 20% for retinol and beta-carotene (Figure 9).

Figure 8. Mean consumption of foods based on the FFQ1 compared to the 14-day diet record.

Figure 9. Mean intake of nutrients based on the FFQ1 compared to the 14-day diet record.

Figure 10 shows the positive association between reproducibility and validity of foods, i.e., good reproducibility was generally related to good validity and vice versa. Reproducibility or validity is good if the value is at least 0.70, acceptable if the value is between 0.50 and 0.69, and it should be at least 0.40 to avoid serious attenuation in the results (Willett and Lenart 1998). According to these limits, both reliability and validity were good (>0.7) for low-fat milk, soft margarine, and coffee, whereas none of these foods was measured extremely poorly (<0.4). The range of reproducibility varied between 0.5 and 0.8; berries, soft margarine, coffee, and most of the dairy products had a correlation over 0.7. The range of validity was quite large, ranging between 0.3 and 0.9. The validity was under 0.4 for oils and meat.

The range of reproducibility was very narrow, between 0.6 and 0.7, for nutrients (Figure 11). Alcohol consumption had a high validity (>0.7), whereas retinol, total fat, and monounsaturated fatty acids had low values (<0.4).

Energy adjustment improved the validity for nutrients more than that for foods, whereas the correction for the attenuation due to within-person variation was similar for nutrients and foods. The energy adjustment improved the correlation coefficient for foods on average 0.02 (between -0.05 and 0.06) and for nutrients on average 0.10 (between -0.04 and 0.24). The correction for the attenuation improved the values for foods on average 0.08 (between 0.04 and 0.12) and for nutrients 0.08 (between 0.06 and 0.10).

When quintile classifications were compared between the FFQ1 and the 14-day diet record, consumption of rye products, sour milk, and poultry were seriously misclassified (Table 4). Total fat, saturated fatty acids, trans-fatty acids, retinol, and vitamin C were the nutrients most often misclassified (Table 5).

In summary, validity was low for oil and meat (Figure 10). Misclassification may also attenuate associations between risk of breast cancer and rye products, sour milk, and poultry (Table 4). Among the nutrients, validity was low for retinol, total fat, and monounsaturated fatty acids (Figure 11), while saturated fatty acids, trans-fatty acids, and vitamin C were most often misclassified (Table 5).

Figure 10. Reproducibility and validity for energy-adjusted foods.

Figure 11. Reproducibility and validity for energy-adjusted nutrients.

Table 4. Cross-classification of food consumption between FFQ1 and diet records.

Food Same or adjacent quintile based on both methods (%) 1st quintile on record but 5th quintile on FFQ (%)
Rye products 66 17
Vegetables 72 3
Fruit 68 7
Berries 69 10
Butter 80 3
Soft margarine 88 4
Oil 64 7
Low-fat milk 82 3
Sour milk 83 17
Cream 62 3
Cheese 72 7
Beef + pork 61 3
Poultry 57 14
Fish 63 10
Coffee 77 3
Tea 85 6
Sugar 78 7

Table 5. Cross-classification of nutrient intake between FFQ1 and diet records.

Nutrient Same or adjacent quintile based on both methods (%) 1st quintile on record but 5th quintile on FFQ (%)
Total fat 59 13
Saturated fatty acids 69 13
Monounsaturated fatty acids 66 3
Polyunsaturated fatty acids 68 3
n-3 fatty acids 64 7
n-6 fatty acids 68 3
Trans-fatty acids 71 17
Fiber 61 7
Retinol 62 17
Beta-carotene 70 7
Vitamin E 59 7
Vitamin C 66 13
Alcohol 93 2

6.4 Food consumption and risk of breast cancer (III)

Associations between foods and risk of breast cancer as well as recall bias due to the threat of breast disease were examined in the Dietary Study (III). Comparison of breast cancer cases and population controls is the classical way of presenting results in case-control studies. The possible recall bias could be assessed by comparing the results in terms of two control groups (population controls and referral controls). Comparisons between the breast cancer cases and the referral controls are considered to result in "unbiased" odds ratios.

Consumption of low-fat milk and poultry was inversely related to the risk of premenopausal breast cancer, while sour milk seemed to increase the risk (Table 6). After the examination of reporting bias, high consumption of milk increased (OR 2.2, 95% CI 1.0-4.9) and poultry decreased (OR 0.4, 95% CI 0.2-0.9) risk of premenopausal breast cancer compared to that in women whose consumption was low.

Butter, oil, cheese, and coffee were related to postmenopausal breast cancer when analyses were carried out in the classical manner between the breast cancer cases and the population controls (Table 6). After the examination of reporting bias, the results showed that high consumption of cream (OR 1.9, 95% CI 1.0-4.0) was associated with increased risk, and high consumption of oil (OR 0.4, 95% CI 0.2-0.8) with decreased risk. There was also suggestive evidence that high consumption of milk and coffee may decrease, and butter increase the risk of postmenopausal breast cancer.

The reporting bias observed explained well the differences in results from use of population controls, referral controls and breast cancer cases. When the food consumption was under-reported or over-reported under the threat of disease, the estimated risk of breast cancer (between the breast cancer cases and referral controls) increased or decreased, respectively.

Table 6. Associations between foods and risk of breast cancer by menopausal status.

Food Cases vs. Population controls Cases vs. Referral controls Reporting "under threat of breast cancer"
Premenopausal women
Rye products - - -
Vegetables - - -
Fruit + berries - - -
Butter - - -
Soft margarine - - -
Oil - - -
Total milk - ­ underestimation
Low-fat milk ¯ - underestimation
Sour milk ­ - (overestimation)
Cream - - -
Cheese - - -
Beef + pork - - -
Poultry ¯ ¯ -
Fish - - -
Coffee - - -
Tea - - underestimation
Sugar (¯) - (underestimation)
Postmenopausal women
Rye products - - -
Vegetables - - -
Fruit + berries - - -
Butter ­ (­) -
Soft margarine - - -
Oil ¯ ¯ -
Total milk (­) (¯) overestimation
Low-fat milk - (¯) (overestimation)
Sour milk (­) - (overestimation)
Cream - ­ underestimation
Cheese ¯ - underestimation
Beef + pork - - -
Poultry - - -
Fish - - -
Coffee ¯ (¯) -
Tea - - -
Sugar - - -

­ = increased risk , p<0.05, ¯ = ecreased risk, p<0.05, ( ) = suggestive effect, - no association.

6.5 Nutrient intake and risk of breast cancer (III)

In premenopausal women, no association between nutrient intake and risk of breast cancer was found in the comparison between the cases and population controls. However, some fatty acids were over-reported under the threat of breast cancer, especially n-3 polyunsaturated fatty acids (Table 7). Thus, after taking the reporting bias into account, a high intake of polyunsaturated fatty acids (OR 0.4, 95% CI 0.2-0.9), n-3 polyunsaturated fatty acids (OR 0.3, 95% CI 0.1-0.6) or n-6 polyunsaturated fatty acids (OR 0.4, 95% CI 0.2-0.8) was related to decreased risk of premenopausal breast cancer compared to risk in women whose intake was low. The association between polyunsaturated fatty acids and breast cancer disappeared when saturated, monounsaturated, and polyunsaturated fatty acids were mutually adjusted in the analyses. There was also some evidence that, under the threat of breast cancer, monounsaturated fatty acids were overestimated. Without overestimation, they may actually have been related to decreased breast cancer risk. This result was, however, not statistically significant. The intakes of total fat, saturated fatty acids, trans-fatty acids, and dietary fiber were not associated with premenopausal breast cancer. Although bias was shown in reporting vitamin intake, only high intake of vitamin E was found to decrease risk of breast cancer (OR 0.5, 95% CI 0.2-1.0).

In postmenopausal women, retinol increased and polyunsaturated fatty acids decreased risk of breast cancer in the comparison between the cases and population controls (Table 7). The association between polyunsaturated fatty acids and risk of breast cancer disappeared when saturated, monounsaturated, and polyunsaturated fatty acids were mutually adjusted in the analyses. High intake of beta-carotene decreased risk when the cases were compared to the referral controls (OR 0.5, 95% CI 0.2-1.0).

Table 7. Association between nutrient intake and risk of breast cancer by menopausal status.

Nutrient Cases vs. Population controls Cases vs. Referral controls Reporting "under threat of breast cancer"
Premenopausal women
Total fat - - -
Saturated fatty acids - - -
Monounsaturated fatty acids - - -
Polyunsaturated fatty acids - ¯1 (overestimation)
n-3 fatty acids - ¯ overestimation
n-6 fatty acids - ¯ (overestimation)
Trans fatty acids - - -
Dietary fiber - - -
Retinol - - -
Retinol with supplements - - underestimation
Beta-carotene - - -
Beta-carotene with supplements - - underestimation
Vitamin E - ¯ overestimation
Vitamin E with supplements - - -
Vitamin C - - -
Vitamin C with supplements - - -
Postmenopausal women
Polyunsaturated fatty acids ¯1 - -
Retinol ­ - -
Retinol with supplements - - -
Beta-carotene - - -
Beta-carotene with supplements - ¯ -

­ = increased risk, p<0.05, ¯ = decreased risk, p<0.05, ( ) = suggestive effect, - no association.

1 No association when saturated, monounsaturated and polyunsaturated fatty acids were mutually adjusted in the analyses.

6.6 Comparison of different methods of assessing alcohol consumption (IV)

Reported alcohol consumption was somewhat higher in the interview-based AQ than in the self-administered FFQ. The mean consumption of alcohol was 28 g/week for the premenopausal cases and 24 g/week for the premenopausal population controls based on the AQ, while the figures were 15 g and 14 g for the postmenopausal women. Based on the FFQ, the difference in alcohol consumption between the cases and population controls was statistically significant (p=0.03) only for postmenopausal women (9 g/week for cases and 14 g/week for population controls). About 30% of premenopausal and 60% of postmenopausal women were classified as abstainers.

The Pearson correlation coefficient between the AQ and the FFQ was 0.60 for the current alcohol consumption in all women, being high for the premenopausal women (0.80) and quite low for the postmenopausal women (0.40). No noteworthy difference, however, was found when current alcohol consumption was categorized into quintiles by the two alcohol measurement methods. In all, 64% of premenopausal and 70% of postmenopausal women were categorized in the same quintile. The percentages decreased to 59% and 57%, respectively, when only alcohol drinkers were compared. Reported lifetime non-drinking and ex-drinking levels were less accurate in postmenopausal than in premenopausal women. Ten postmenopausal women were maximally misclassified, but none of the premenopausal women. The overall proportion of ex-drinkers was quite low, 4% for premenopausal and 5% for postmenopausal women.

Current alcohol consumption was correlated with lifetime alcohol consumption in all women (r=0.74), but not so much with alcohol consumption at the age of first use (r=0.48) or with alcohol consumption before age 30 (r=0.42).

6.7 Current and past alcohol consumption and risk of breast cancer (IV)

Current alcohol consumption, based on the AQ, was not associated with risk of premenopausal or postmenopausal breast cancer. The odds ratios were around 1.0 for all alcohol consumption categories, with one exception for premenopausal ex-drinkers (OR 1.4, 95% CI 0.3-6.2). The results of the FFQ were comparable to those observed in the AQ.

Cumulative lifetime alcohol consumption was not associated with risk of breast cancer in premenopausal or postmenopausal women. In addition, alcohol consumption at early ages (before age 30) did not increase the risk.

6.8 Toenail selenium concentration and risk of breast cancer (V)

The mean toenail selenium concentration was 0.80 mg/kg in premenopausal cases and 0.84 mg/kg in the population controls, whereas the concentrations were 0.77 mg/kg and 0.80 mg/kg, respectively, in postmenopausal women. The relative risk of breast cancer between the highest and the lowest quintile of toenail selenium concentration was 1.1 (95% CI 0.4-3.2) for premenopausal women and 0.7 (95% CI 0.3-1.5) for postmenopausal women. Intakes of retinol, beta-carotene, vitamin E, and vitamin C did not modify these associations.


Back to the title page | E-thesis main page