SEARCH

SEARCH BY CITATION

Keywords:

  • cohort study;
  • breast cancer;
  • risk factors;
  • age at diagnosis

Abstract

  1. Top of page
  2. Abstract
  3. MATERIAL AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

An increasing number of studies indicates that the strength and even direction of association between breast cancer and established risk factors differ according to the woman's age when she develops the disease. This was studied in the setting of a population based cancer registry using a databank with information on age at menarche, parity, age at first birth, oral contraceptive (OC) use, lactation, height and weight. From a cohort of 80.219 women attending population-based cervical and breast cancer screening in Iceland, 1120 cases were identified, aged 26–90 years at diagnosis and 10,537 controls, individually matched to the cases on birth year and age when attending. Information given at last visit before diagnosis was used in the analysis, applying conditional logistic regression. Odds ratios and statistical strength of relationships varied according to age at diagnosis for age at first birth, number of births, duration of lactation, height and weight. The decreased risk associated with young age at first birth and increasing duration of breast feeding became less pronounced with advancing age at diagnosis. A reduced risk associated with an increasing number of births was not detected in women diagnosed under the age of 40. An increased risk associated with giving first birth after 30 years of age was mainly detected in women who had only given 1 birth and were diagnosed under the age of 40 (OR = 7.06 95% CI = 2.16–23.01). A positive association with height and especially with weight was confined to women diagnosed after the age of 55. The results confirm that age at diagnosis should be taken into account when studying the effects of breast cancer risk factors. © 2002 Wiley-Liss, Inc.

The epidemiology of breast cancer has been studied extensively.1 The risk of breast cancer is positively associated with age at first birth and negatively associated with number of births, age at menarche and duration of breast feeding.2 Breast cancer risk has been reported to be positively related with height and body mass index (BMI).3 But studies taking age at diagnosis into account show a more complicated picture. There are several indications that the observed strength and even direction of association between breast cancer and risk factors differ according to age at diagnosis.4–9 Results obtained at 2 different points in time in 2 Icelandic studies using the Cancer Detection Clinic (CDC) cohort did also indicate this. Lower relative risks were observed when the effects of age at first birth were investigated at a later point in time. This was attributed to the higher mean age of the cases in the cohort at the time of the second study.10, 11 Those observations underline the importance of taking into account the age of the group studied, as pointed out by Velentgas et al.9 In the present nested case-control study we investigated the effects of age at menarche, parity, age at first birth, total duration of oral contraceptive (OC) use, height and weight, dividing the cases into 3 groups according to age at diagnosis. This is part of a series of nested case-control studies, using a large Icelandic cohort, where the other studies focus on possible associations between risk factors and somatic p53 mutations and interaction between the presence of germline mutations in the BRCA genes and the effects of other risk factors.

MATERIAL AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIAL AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

The study group has been described in more detail elsewhere.12 At the time of study the CDC cohort consisted of 80,219 women who had attended a nation-wide cervical- and breast cancer-screening program in the years 1979–95 and given answers to questions on reproductive and menstrual factors. The women were at ages 20–81 years when attending. The majority of information in the Databank has been collected in association with the cervical cancer screening, both because the mammography program is relatively new and because the attendance rate for the breast cancer screening has been considerably lower than for the cervical cancer screening.

All Icelandic women diagnosed with first invasive breast cancer in the years 1979–95 were identified in the population based Cancer Registry of Iceland, a total of 1,601 cases. Record linkage identified 85% of those in the CDC databank. Only women who had contributed information to the databank in the years 1979–95 were included. Only answers given before diagnosis of a first breast cancer were used. For women who had attended repeatedly, the data recorded at last applicable visit were used. This resulted in a total of 1,120 cases who fulfilled the inclusion criteria. They comprise 70% of women diagnosed with their first invasive breast cancer in Iceland in the study period. From the CDC databank, we sought 10 controls for each of the 1,120 cases, individually matched on birth year and age when giving information. They had to have been alive at least until the diagnosis-year of their matched case. For the analysis the following variables from the databank were used; age at menarche, age at first birth, number of births, number of children that were breast fed, average number of weeks breast-feeding each child, total duration of oral contraceptive use, height (cm) and weight (kg). All those variables were entered as continuous.

The study group was stratified according to age at diagnosis of the cases. The following categories were used: <40 years, 40–55 years and >55 years. Those groups were chosen to represent very young or premenopausal, perimenopausal and postmenopausal women respectively. We tested for interaction between each variable that was related to breast cancer risk (entered as a continuous variable) and age at diagnosis as a categorical variable and for menopausal status. Two different cutpoints for age at diagnosis were entered, <40 years vs. older and <56 years vs. older, testing whether the interaction term between those variables would attain statistical significance when added to either of the models.

Conditional multiple logistic regression13 was applied for the multivariate analysis of these matched data, using the statistical package STATA.14

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIAL AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

Table I shows attributes of the three age groups. The 97 cases youngest at diagnosis were also youngest when giving information and they belonged to the latest birth cohorts. It may be noted that among controls the proportion with young age at menarche increases as the groups become younger, but age at menarche has been going down in Iceland.15 A birth cohort effect can also be seen in the pattern of OC use, with a lower percentage of never users in younger groups.

Table I. Cases And Controls Grouped According To Age At Diagnosis: Comparison Of Age And Birth Year, Reproductive And Menstrual Characteristics, Height And Weight
Age at diagnosis Risk factor<40 years (26–39) (97 cases/970 controls)40–55 years (434 cases/4268 controls)>55 years (56–90) (589 cases/5299 controls)
% cases% controls% cases% controls% cases% controls
Age at questionnaire (median, range)31 (20–39)42 (28–59)58 (44–81)
Birth year (median, range)1952 (41–67)1940 (24–55)1924 (03–39)
Age at menarche(n = 96)(n = 969)(n = 433)(n = 4265)(n = 588)(n = 5290)
 <13453232282019
 13233329312726
 >13323439425355
Parous(n = 97)(n = 970)(n = 434)(n = 4268)(n = 589)(n = 5299)
 Yes878892958791
 No131285139
Parous women: age at first birth(n = 84)(n = 857)(n = 399)(n = 4052)(n = 510)(n = 4820)
 <20253526291315
 20–29646265677275
 >29113951510
Parous women: number of births(n = 84)(n = 857)(n = 399)(n = 4052)(n = 510)(n = 4820)
 13121118128
 2363823201915
 3+334166726877
Total duration of OC use (months)(n = 94)(n = 945)(n = 421)(n = 4191)(n = 573)(n = 5212)
 0131130298076
 1–61211111125
 7–59434932311010
 60–1192023151845
 120+136111254
Parous women: total duration of breastfeeding (months)(n = 84)(n = 857)(n = 399)(n = 4052)(n = 510)(n = 4820)
 060.52131
 1–6605751493635
 6–12272730293028
 13–2461314172424
 >241224812
Height (cm)(n = 97)(n = 957)(n = 433)(n = 4260)(n = 587)(n = 5280)
 ≤160131517203340
 161–169585659585650
 170+292924221110
Weight (kg)(n = 97)(n = 953)(n = 433)(n = 4254)(n = 586)(n = 5276)
 ≤60424329282124
 61–79464855575858
 80+11816152118

In Table II the results from the multivariate analysis are shown for 3 groups according to age at diagnosis: <40 years, 40–55 years and >55 years, as well as for the total study group. In the youngest group, 1 year's increase in age at menarche was associated with a 18% reduction in breast cancer risk, whereas the reduction was 8% in the 2 older groups and the association was statistically significant in all age groups. One year's increase in age at first birth was associated with a statistically significantly increased risk of 10% and 5% for women diagnosed younger than 40 years of age and at ages 40–55 years respectively, but increased risk was not detected in women diagnosed over 55 years of age. On the other hand, each additional birth was associated with a statistically significant 8% risk reduction, only for women diagnosed older than 55 years. No association was apparent between risk of breast cancer and total duration of OC use as reported here at last visit. Increasing duration of breast feeding was associated with reduced risk in the youngest group mainly.12 In the group diagnosed after the age of 55 years, each 5 cm increase in height was associated with a 12% increase in risk. A non-significant risk increase of 8% was present in the group diagnosed at the ages 40–55 years, but no association was detected for the women youngest at diagnosis. When pooling the 2 younger groups together, the OR was 1.07 (95% CI = 0.97–1.17). An increase of 10 kg in weight was associated with 7% increase in breast cancer risk after controlling for height in the oldest group (p-value = 0.08). No association was detected in women under 55 years at diagnosis.

Table II. –Multivariate Analysis Of The Effects Of Risk Factors According To Age At Diagnosis1
Risk factorsAll ages n = 1120<40 years n = 9740–55 years n = 434>55 years n = 589
Odds ratio (95% CI)p valueOdds ratio (95% CI)p valueOdds ratio (95% CI)p valueOdds ratio (95% CI)p value
  • 1

    Odds ratios are given per unit change of the variables.

Age at menarche0.91 (0.87–0.96)<0.0010.82 (0.68–0.99)0.0400.92 (0.84–0.99)0.0430.92 (0.87–0.99)0.019
Parous (yes/no)0.96 (0.74–1.25)0.7731.02 (0.43–2.45)0.9530.84 (0.52–1.36)0.4751.00 (0.71–1.41)0.992
Age at first birth1.03 (1.01–1.05)<0.0011.10 (1.02–1.18)0.0091.05 (1.02–1.08)<0.0011.01 (0.99–1.04)0.194
Number of births0.92 (0.88–0.97)0.0041.12 (0.82–1.54)0.4670.93 (0.84–1.03)0.1470.92 (0.86–0.98)0.008
OC use (incr. 12 weeks)1.00 (0.99–1.01)0.9441.00 (0.97–1.03)0.9451.00 (0.98–1.02)0.9281.00 (0.98–1.03)0.682
Lactation (incr.6 mo)0.95 (0.91–0.99)0.0310.76 (0.59–0.99)0.0440.94 (0.86–1.03)0.1960.96 (0.91–1.01)0.134
Height (incr. 5 cm)1.09 (1.03–1.16)0.0040.99 (0.79–1.22)0.9081.08 (0.98–1.20)0.1111.12 (1.03–1.22)0.011
Weight (incr. 10 kg)1.03 (0.97–1.09)0.2751.05 (0.84–1.31)0.6480.98 (0.89–1.07)0.6211.07 (0.99–1.16)0.078

Only 2 models testing for interaction between either age at diagnosis or menopausal status and the risk factors in Table II resulted in statistically significant or nearly significant interaction factors. Those were the models including an interaction term between breast feeding and age at diagnosis (<40 vs. 40, p-value of 0.08)12 and an interaction term between weight and age at diagnosis (divided into <56 and 56 years and older), where the p-value was 0.07.

In the multivariate analysis in Table III, with stratification according to both number of births and age at first birth, an increased risk for women giving first birth after the age of 30 years when compared to nullipara, was mainly detected in women with only 1 birth (OR 1.62, 95% CI = 1.12–2.34). Table IV shows that the observed risk increase in this category was mainly explained by the group diagnosed under the age of 40, where the OR was 7.06 (95% CI = 2.16–23.01). When confining the analysis to women diagnosed 40 years or older the OR in this category was 1.40 (95% CI = 0.95–2.07) (data not shown).

Table III. Effects Of Age At First Birth And Number Of Births For Parous Women, As Compared To Nulliparous Women, For The Total Study Group1
Number of birthsAge at first birthNumber of casesNumber of controlsOdds ratio (95% CI)p
  • 1

    Multivariate analysis, taking into account the effects of age at menarche, lactation, height and weight. 1120 cases.

01278081
1<2014871.21 (0.66–2.23)0.530
20–22221840.88 (0.53–1.45)0.606
23–29453580.86 (0.59–1.26)0.452
30+532401.62 (1.12–2.34)0.010
2<20243010.55 (0.34–0.89)0.015
20–22545300.74 (0.52–1.04)0.086
23–291027930.91 (0.68–1.22)0.522
30+412381.21 (0.81–1.80)0.348
3+<2015117670.62 (0.47–0.83)0.001
20–2221126090.61 (0.47–0.79)0.000
23–2924723680.77 (0.59–0.99)0.043
30+282470.84 (0.53–1.31)0.439
Table IV. Effects Of Age At First Birth And Number Of Births For Parous Women, As Compared To Nulliparous Women, For Women Diagonsed <40 Years Of Age1
Number of birthsAge at first birthNumber of casesNumber of controlsOdds ratio (95% CI)p
  • 1

    Multivariate analysis, taking into account the effects of age at menarche, lactation, height and weight. 97 cases.

0131131
1<203390.87 (0.23–3.33)0.842
20–228561.68 (0.63–4.49)0.299
23–297711.13 (0.41–3.12)0.805
30+8147.06 (2.16–23.01)0.001
2<2010961.24 (0.49–3.17)0.646
20–22101111.15 (0.44–2.98)0.769
23–2991110.99 (0.35–2.76)0.979
30+1101.56 (0.16–14.97)0.700
3+<2081630.66 (0.22–1.96)0.639
20–22111311.28 (0.45–3.66)0.639
23–299552.62 (0.86–7.93)0.089
30+

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIAL AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

The results presented here support the notion that age at diagnosis should be taken into account when studying effects of breast cancer risk factors. Both odds ratios and statistical strength of relationships varied considerably according to age at diagnosis for age at first birth, number of births, duration of lactation, height and weight. The positive association with age at first birth was not present in the oldest group and the negative association with total duration of breast feeding was only statistically significant in the youngest group,12 whereas the negative association with an increasing number of births was only significant in the oldest group. Furthermore, the increased risk associated with giving first birth after 30 years of age was mainly detected in uniparous women who were younger than 40 years at diagnosis. An inverse association between breast cancer risk and age at menarche was strongest in the youngest age group as others have found,7 but a positive association with height and especially with weight was apparent only after the age of 55.

The diminishing effects of age at first birth with age and the lack of protection from increasing number of pregnancies in the youngest groups relate to previous findings of several researchers indicating that after each childbirth, women experience a transient increase in breast cancer risk,16–20 lasting 3–4 years after each pregnancy.21, 22 This transient risk has been postulated to account for lack of observed protective effects of parity in women while they are at childbearing ages, that is in women diagnosed under the age of 45 or 40.4, 23 This provides at least part of the explanation of the effect of age at first birth on breast cancer risk among young women22 and is in accordance with the present findings that the relative risk associated with giving first birth after 30 years of age was by far the highest in women who were younger than 40 years at diagnosis and uniparous. The latter being in accordance with studies reporting that the highest transient increase in risk is found among women who have few children.17, 21, 22

A positive association with height was statistically significant only in the group diagnosed after the age of 55 years but a non-significant association was observed in the group diagnosed at ages 40–55 years. An earlier Icelandic study using a different cohort, also reported a positive association that was confined to women diagnosed at ages over 54 years, but the number of younger cases was small (91 women).24 A recent pooled analysis of cohort studies found a positive association in postmenopausal women and a less clear relationship in premenopausal women.25 A mechanism has been suggested by which affluent diet during childhood and adolescence may influence both height and the risk of breast cancer.3

The positive association between increasing weight and breast cancer, after controlling for height, was present only at postmenopausal ages. This is in accordance with what most other researchers have found and has been attributed to increased aromatase activity in adipose tissue and thus increased levels of estrogens, that make a bigger impact in postmenopausal women than in women at fertile ages.26, 27 No effects were detected for the groups younger than 55 years at diagnosis, but as presented in a metaanalysis28 and a recent review article,26 several investigators have reported a negative association between breast cancer and body mass index at those younger ages.

No effects were found of total duration of OC use as reported here at last visit to the clinic. A previous Icelandic study focusing on OC use at young ages and therefore using information given at the first visit to the CDC and only from birth cohorts who had the possibility of exposure to OCs from an early age, indicated a dose-response relationship between breast cancer risk and duration of OC use in young users.29

The present study group was stratified according to age at diagnosis with cutting points <40, 40–55 and >55 years at diagnosis, where the oldest group represented women diagnosed at postmenopausal ages. The data were also analyzed with stratification according to menopausal status, giving similar results except that this did not allow for investigating separately women diagnosed under the age of 40. This age group, however, is of special interest. Women diagnosed at young ages are etiologically somewhat different from older cases because of a high prevalence of patients with family history30, 31 and presence of germ line mutations in the BRCA genes.32, 33 The latter has been demonstrated in Iceland, where the prevailing Icelandic BRCA2 founder mutation (999del5)34 was found in 24% of unselected women with breast cancer diagnosed below the age of 40 and in a much lower proportion of older patients.35 As there are indications from 2 recent studies that pregnancies may not have the same protective effects on BRCA mutation carriers as on noncarriers,36, 37 which is supported by our own preliminary data, this might be expected to be reflected when studying the youngest patients.

In our study the cases and controls were matched on age when responding and on birth year. This may have somewhat increased the precision of the study by eliminating variation due to birth cohort effects and age when responding, both factors being associated with all of the risk factors being studied. The information was gathered on the average several years before the diagnosis of the cases. For women still at reproductive ages this may have resulted in non-differential misclassification, thus lowering the power of the study. Relative risk estimates do vary considerably between studies that focus on the effects of risk factors according to age. One explanation of this is the play of chance, often there is little power to detect small effects on risk.

When linking the CDC cohort with the Cancer Registry, 85% of cases diagnosed in the study period were identified. The remaining 15% were mainly women who were borne early in the 20th century.12 This was to be expected, because it has been shown previously that women born in the first decades of the 20th century are not well represented in the CDC databank, whereas in younger birth cohorts up to 97% of women have contributed information.15 The observed associations between risk factors and breast cancer are not likely to be biased even though 15% of all cases diagnosed in the study period were not included in the databank. This is because the information for both cases and controls came from the same source and was collected in the same way for all respondents.

Current understanding of the biological effects of reproductive factors, has come both from epidemiological and laboratory studies. Russo et al.38–41 have investigated the effects of pregnancies on the differentiation of epithelial cells in the mammary gland, both in animal models and in human breast tissue. They have shown that during the lifespan of a female, the breast tissue is constantly changing due to complicated developmental changes, with parous women having greater number of differentiated glandular structures in the breast than nullipara. Furthermore, that the undifferentiated stem cells in the breast are susceptible to the effects of carcinogens whereas the differentiated tissue is less susceptible.

The results of the present study underline the complexity of breast cancer etiology. Above have been mentioned three explanations for different effects of risk factors according to age. The first relates to short term effects of the particular risk factor, as for the postulated transient adverse effects of pregnancies. The second being that the exposure does not become relevant until after a certain age has been reached, as with the proposed mechanism of estrogen production in adipose tissue in overweight women. The third explanation is that the special effects that are observed in young women may relate to the presence of subgroups who are genetically predisposed to breast cancer and are therefore likely to be diagnosed at young ages. Those women may react in a special way to the effects of other breast cancer risk factors. The present study confirms that there are subgroups of women with respect to effects of responses to known risk factors. To be able to better understand the etiology of breast cancer, it is of importance to identify and study those subgroups in more detail.

Acknowledgements

  1. Top of page
  2. Abstract
  3. MATERIAL AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

The authors acknowledge Dr. K. Sigurdsson director of the Cancer Detection Clinic of the Icelandic Cancer Society for providing the data and professors T. Hakulinen and H.M. Ögmundsdottir for helpful comments on the article.

REFERENCES

  1. Top of page
  2. Abstract
  3. MATERIAL AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES