Attributable risks for familial breast cancer by proband status and morphology: A nationwide epidemiologic study from Sweden



Population attributable factions (PAFs) show the proportion of the disease that could be prevented if the cause could be removed. The PAF for familial breast cancer has not been precisely determined. We used the nationwide Swedish Family-Cancer Database on 10.2 million individuals and 190,000 mothers' and 26,000 daughters' breast cancers to calculate familial standardized incidence ratios (SIRs), proportion of cases with a family history and familial PAFs for all invasive and in situ and morphology-specific breast cancers in daughters who were 0–66 years old. The data were calculated by mother only, sister only or both as probands. More than 5,500 familial breast cancers were recorded. The familial SIRs for all invasive breast cancer were 1.79 by breast cancer in the mother only, 2.03 by breast cancer in a sister only and 2.82 by breast cancer in both a mother and sister. The familial PAFs were 3.61, 3.01 and 0.43%, respectively, giving a total PAF of 7.05%. Age-specific risks were shown for the mother and sister history of breast cancer. The PAF values decreased by age when the daughter had a mother history of breast cancer but not when she had a sister history. PAFs did not depend on the morphologic type of breast cancer. The data show that the familial PAF of breast cancer among a 0–66-year-old population of daughters was 7% and independent of the morphologic type. If contribution from the paternal side was allowed for, the PAF would be 11%. © 2002 Wiley-Liss, Inc.

Familial risks in female breast cancer have been the subject of numerous epidemiologic studies. A comprehensive review of the subject, published in 1997, covered 27 case-control and 22 cohort studies and gave a pooled estimate of familial relative risk of 1.8 from an affected daughter, 2.0 from a mother, 2.3 from a sister and 3.6 from amother and sister.1 A recent study reanalyzed 52 epidemiologic studies on familial breast cancer and presented summary risk ratios of 1.80 and 2.93 for 1 and 2 affected first-degree relatives, respectively.2 However, no distinction was made between mother-daughter or sister-sister pairs. Whether the risks between these pairs are equal or unequal are of etiologic and clinical importance and need to be settled. The exact estimation of the family history by the proband status is relevant for the understanding of the genetic mechanisms operating in breast cancer.1, 3–5 Some evidence on recessive effects has been obtained in a recent segregation analysis,6 and a large twin study estimated the hereditary component in breast cancer to be 27% of the twin variation.7 The latter figure was almost 10 times higher than the known genetic causes of breast cancer, implying gaps in the genetic etiology of breast cancer.

A minor technical uncertainty in the published familial risks is that almost all contributing data on relatives were obtained from patients' reports rather than from medical records. For many cancer types, this would be a serious problem but for breast cancer the reporting of family history is considered to be reasonably reliable.8, 9 Two aspects of familial breast cancer are less well established. One pertains to familial risks by morphologic types of breast cancer.10 Only small differences have been found in morphology between BRCA1/2-related and -unrelated breast cancers.11–14 The second aspect is the population attributable fraction (PAF) of breast cancer, which tells how large proportion of all breast cancer can be associated with a family history. PAF depends on the magnitude of the familial risk and the proportion of cases with a family history.15, 16 The previous estimates on familial PAFs in breast cancer among first-degree relatives are based on a number of populations and methods, and they show a wide range from 2.5–10%.3, 17–24

We examine here familial risks in breast cancer using the nationwide Swedish Family-Cancer Database,25 which has been updated in 2001 to include more than 10 million individuals and more than 1 million registered tumors, retrieved from the Swedish Cancer Registry from 1958–1998. The Database offers unique possibilities for reliable estimation of familial risks because the data on family relationships and cancers were obtained from registered sources of practically complete coverage. We want to answer specifically the questions about familial risks and PAFs by each type of familial relationship and breast cancer morphology.


The Swedish Family-Cancer Database was initially created in the middle of the 1990s by linking an administrative family register on all Swedish families to the Swedish Cancer Registry.25–27 For each child, there are data on both parents at the time of birth. Each person has been assigned a unique technical identification number (which is different from the national identification number, “personal number”), allowing construction of families, for example, through the mother. The Database includes all persons born in Sweden after 1931 with their biological parents, totaling more than 10.2 million individuals. It was updated in the beginning of 2001 to include cancers from the nationwide Swedish Cancer Registry from 1958–1998.25 The Database is organized in 3.2 million families, with parents and offspring.

The completeness of cancer registration in the 1970s has been estimated to be more than 95% and is now considered to be close to 100%. The percentage of cytologically or histologically verified cases of breast cancers has been close to 100%.28 The Family-Cancer Database has an incomplete linkage from deceased offspring to parents, particularly among those offspring born from 1932–1940. Of a total of 7.0 million offspring, 216,000 have died by the end of follow-up, December 31, 1998. Parental information was missing from 15,600 dead offspring who had a diagnosis of cancer (9.9% of all offspring cancers); most of these were born in the 1930s. This deficit is unlikely to cause any appreciable effect on familial risk estimates in our study but may cause a small reduction in the estimated sibling risks.

The Swedish Cancer Registry is based on compulsory notification of cases.28 A 4-digit diagnostic code according to the 7th revision of the ICD-7 was used until 1992 and thereafter ICD-O-2.

Family history information was collected on all first-degree relatives (parents, siblings and children) but only the mother-daughter, sister-sister and mother-2 daughters relationships were used in our study. All tumor incidence rates were based on the data in the Family-Cancer Database. The risk of morphology-specific breast cancer was calculated for daughters whose mothers presented with any breast cancer and the risk was compared to the rate of morphology-specific breast cancer among all daughters. Follow-up was started at birth, immigration or January 1, 1961, whichever came latest. Follow-up was terminated on diagnosis of first cancer, death, emigration or the closing date of the study, December 31, 1998. Standardized incidence ratios (SIRs) were calculated as the ratio of observed (O) to expected (E) number of cases. The expected numbers were calculated from 5-year-age-, sex-, tumor type-, period- (10-year bands), socioeconomic status- (4 groups) and residential area- (2 groups) specific standard incidence rates for all daughters. Confidence intervals (95% CI) were calculated assuming a Poisson distribution.4

The PAF of cases with a family history of breast cancer was estimated as follows: proportion of cases with a family history × (familial SIR-1)/familial SIR, as defined by Miettinen29 and cited in a textbook as formula 16-21.15 For sisters, the SIRs were calculated only in families of 2 or more sisters, as given earlier.4 The proportion of cases with a sister history was obtained by taking the number of affected sisters in families of 2 or more affected and dividing this by the number of affected sisters in families of at least 1 affected sister. The calculation of SIRs for breast cancer in all families or in families of 2 or more daughters gave essentially similar results. PAF was calculated separately for each mutually exclusive family history.


The Family-Cancer Database covered years 1961–1998 from the Swedish Cancer Registry and included 190,132 and 26,391 breast cancers in mothers and daughters, respectively. A total of 5,507 familial breast cancers were recorded in the 2 generations (mothers and daughters). Among these, 2,156 affected mother-daughter pairs were found in families where no other member presented with breast cancer, with an overall SIR of 1.79 (Table I). The SIR fell from 3.32 among daughters diagnosed from ages 20–24 years to 1.68 among those aged 60–66 years. The proportion of affected daughters with an affected mother was also age-dependent and fell from 11.48% to 6.88%. These figures were used to calculate familial PAFs, which ranged in an age-dependant manner from 8.02% to 2.79%, with an overall PAF of 3.61%. When only a sister was affected, the patterns were quite similar, but the proportion of affected sisters was lower, resulting in a lower PAF value. The overall SIR was 2.03, proportion 5.94% and PAF 3.01%. The smallest group was those who had both an affected mother and sister, constituting 95 triplets. The overall SIR was 2.82, proportion 0.66% and PAF 0.43%. The PAFs for any group are mutually exclusive and they can be summed: the total familial PAF was 7.05%. In Table I and later in Table IV, PAF values are given for all family histories, regardless of the significance of the SIRs.

Table I. SIRs, Familial Proportions and PAFs For 0–66-Year-Old Daughters1
AgeOSIR95% CIProportion (%)PAF (%)PAF A+B+C
  • 1

    O, observed cases; CI, confidence interval.

  • 2

    Bold type, 95% CI does not include 1.00.

Mother only (PAF A)
Sister only (PAF B)
Mother and sister (PAF C)

The data on SIRs are shown in Figure 1. As an additional graph, we show data on daughters whose mothers were diagnosed before age 67 years, to make the data fully comparable to those for sisters, because both of them were also diagnosed before age 67 years. The truncation of the maternal age to 66 years produced a graph that was between sister-sister and mother-daughter graphs. The overall SIR in this group was 1.88. The PAFs in the 3 mutually exclusive groups are shown in Figure 2. Whereas the PAFs from an affected mother decrease with age, the values from a sister remain stable.

Figure 1.

Age-specific SIR of invasive breast cancer by the proband status in the 0–66-year-old population of daughters in the Swedish Family-Cancer Database. As an additional group, SIRs are shown for daughters whose mothers were diagnosed before age 67 years.

Figure 2.

Age-specific PAF values by the proband status. Note that only significant PAF values from Table I are shown.

To respond to the clinical counseling situation, age-specific SIRs in daughters' age groups at diagnosis are shown by probands' age groups in Tables II and III. These data were only standardized for period, which is the reason for the small differences between the overall SIRs and those in Table I. Both a mother's age of onset and a daughter's age influenced the daughter's risk of breast cancer, which is shown in Table II in families where a mother and only 1 daughter were affected. The high familial risks were at young ages: SIR 6.09 for 0–39-year-old daughters by mothers in the same age group. The SIR declined systematically by age, and only one SIR was higher than 2.00 when either the mother or the daughter were more than 50 years. A similar analysis between sister pairs is shown in Table III. The trend was similar to the previous Table, and the marginal SIRs declined systematically from 2.36/2.37 to 1.77/1.79, however with one high SIR of 4.03 in the oldest possible diagnostic group, apparently a chance increase.

Table II. Age-Specific SIR For Breast Cancer by Mother History of Breast Cancer1
Age at diagnosisMother's age at diagnosis
  • 1

    Bold type: 95% CI does not include 1.00. Period of follow-up: 1961–98.

 95% CI4.11–8.702.13–3.341.54–2.321.38–2.040.97–1.611.70–2.09
 95% CI1.26–4.091.81–2.611.76–2.311.49–1.891.46–1.831.68–1.91
 95% CI0.00–5.201.01–1.981.28–1.881.48–1.951.50–1.851.53–1.77
 95% CI0.33–5.230.32–1.701.24–2.501.39–2.241.37–2.00
 95% CI2.78–5.151.88–2.431.65–1.991.56–1.831.51–1.741.68–1.82
Table III. Age-Specific SIR for Breast Cancer by Sister History of Breast Cancer1
Age at diagnosisSister's age at diagnosis
  • 1

    Bold type: 95% CI does not include 1.00. Period of follow-up: 1961–98.

 95% CI1.68–7.021.22–4.280.95–3.940.18–5.811.39–4.01
 95% CI1.22–4.261.07–2.981.28–3.570.67–3.701.26–3.10
 95% CI0.95–3.891.29–3.570.89–2.580.62–3.161.14–2.85
 95% CI0.18–5.680.73–3.910.61–3.111.46–9.980.95–3.32
 95% CI1.38–3.981.27–3.121.14–2.840.93–3.281.28–2.94

Table IV shows familial SIRs, proportions and PAFs by morphologic types in daughters with invasive and in situ breast cancer. Because specific morphology has been recorded since 1993, we counted daughters' breast cancer during 1993–1998. Familial breast cancer was defined when breast cancer was diagnosed in mother or sister irrespective of its time of occurrence. The total number of morphology-specific breast cancers in daughters was 13,084 first invasive and 1,736 in situ breast cancers. Familial risk was 1.72 for all invasive breast cancer by all mothers; the SIR was identical when a sister was the proband (SIR 1.84). However, when both mother and sister were affected, the SIR was 3.11. The PAFs ranged from 0.46 for mother and sister history to 3.37 for mother history only, adding up to 6.20. There were limited differences in familial risks by the morphologic type. By mother history, the range in SIRs was from 1.61 for the mixed group “other” type to 2.08 for comedo cancer. The comedo type also showed the highest PAF of 5.04 for the mother history. For a sister history, only ductal and lobular cancers showed a significant increase of 1.82 and 2.16, respectively. The total PAF value was 6.08 for ductal and 7.04 for lobular cancer.

Table IV. SIR By Family History for Morphologic Types of Breast Cancer in Daughters1
Morphologic typeFamily historyOSIR95% CIProportion (%)PAF (%)
  • 1

    Bold type: 95% CI does not include 1.00. Period of follow-up: 1993–98.

Mother and sister293.211.52–6.530.690.47
Total PAF6.08
Mother and sister63.250.83–10.080.740.51
Total PAF7.04
Mother and sister10.730.00–5.950.170.00
Total PAF6.82
Mother and sister33.520.47–14.750.710.51
Total PAF6.97
Mother and sister103.771.27–9.840.820.60
Total PAF5.62
Mother and sister493.111.63–5.830.670.46
Total PAF6.20
In situ
Mother and sister53.800.85–12.650.890.66
Total PAF9.03
Mother and sister39.911.32–41.502.262.03
Total PAF8.06
Mother and sister26.980.47–36.281.611.38
Total PAF6.55
Mother and sister13.980.00–32.230.850.64
Total PAF11.80
Mother and sister115.101.79–12.961.180.95
Total PAF8.93

Familial in situ breast cancer in daughters showed somewhat higher familial risks than invasive breast cancer when assessed by a family history of invasive breast cancer (Table IV). In all in situ breast cancer, the familial risk was 1.94 by mothers and 2.19 by sisters, with respective PAFs of 4.49 and 3.49; the overall combined PAF was 8.93. Lobular in situ breast cancer had an SIR of 9.91 when both a mother and a sister were affected.


The present data were based on the Swedish Family-Cancer Database, which has an almost complete nationwide coverage of families and cancer cases and which has been used in some 100 cancer studies so far.25–27 The derived estimates for familial risk should be free of bias and apply to a 0–66-year-old population. Also, the size of our study tells about its power: the number of familial cases is some 3–4 times larger than that from other published cohorts. The limitation is that only nuclear families were available for study and there was no possibility to assess breast cancer risks from second-degree relatives, as discussed later. A special aspect of our study is that we carried out the analysis by 3 mutually exclusive proband groups. For clinical purposes, it is very important to exactly define the proband's status: the doctor is seeing a sister or a daughter of a proband, not her first-degree relative. Unfortunately, the term first-degree relative has been widely used in genetic epidemiology of cancer, showing unfamiliarity with the clinical situation and an a priori disregard of genetic mechanisms, such as recessive or sex-limited inheritance, as discussed later. In the same token, we may point out that the term first-degree relative is not unfortunate in the case of breast cancer, because the present data show that the risks from a mother and from a sister are very close to each other and the results should settle the issue at least for invasive breast cancer. For in situ breast cancer, the difference was larger between the mother and sister history, but this may be an effect of breast cancer screening.30 A sister diagnosed with breast cancer is most likely of a larger concern for an equally old woman than a diagnosed mother, and, consequently, medical advice will preferentially be sought after a sister history. The small proportion of deceased offspring with cancer that cannot be linked to their parents may cause a small underestimation of the sibling risks but it does not change the conclusion. Most of these unlinked cases were born in the 1930s, and analysis by birth cohorts show similar risks from mother and sister throughout (data not shown). It is likely that the previous studies showing higher breast cancer risks from sisters than from mothers are due to problems of obtaining representative data from a single compared to 2 generations.

There is hardly any literature on familial risks in breast cancer by morphology but because we have dealt with this issue elsewhere, we refer the discussion to this source.13, 14 Suffice to state that, within statistical limits, all the common morphologic types showed equally high familial risks. It will be of interest to see how different morphologic types perform in gene expression systems.31–33 On the other hand, we reported new data on age-specific familial risks on breast cancer when only a mother or a sister was affected (Tables II, III). Previously, such a definition of the proband status has rarely been applied.17 These data should be useful for a clinical situation and define the daughter's risk at various ages by the diagnosis age of the mother.

Familial risk can be due to shared genes or environment or due to chance. By comparing various family relationships, we have estimated that for most cancers the familial risks are mainly due to heritable causes.34 The exceptions are cancers with strong environmental factors, such as lung and cervical cancer. The data on PAFs will not distinguish the causes of familial clustering of cancer. The present data on PAFs were given by the exact proband status, which were mutually exclusive, and could be summed. However, it is essential to point out that the PAFs refer to a 0–66-year-old population of daughters. In the literature, different ways of presenting PAF can be found and hence a large variation in the given values from 2.5–19%.19–22 The terms first-degree or second-degree relative appear imprecise in this context because of difficulties in defining the ages and numbers of the relatives, similar to problems experienced in defining syndromes such as hereditary nonpolyposis colorectal cancer by the number of those affected in the family.3, 35 These difficulties were illustrated in Figure 2 that showed a sharply declining PAF by age for a mother history, whereas for the sister history, the PAFs were relatively stable throughout the age-range studied. Another problem in a gender-limited cancer is that the effects to daughters could be transmitted by the paternal grandmother, on whom information is lacking.3 Technically, this issue could be handled by considering in the PAF calculation that both the maternal and paternal lineage can contribute equally to the risk in daughters, as discussed below.

In the present population of 0–66-year-old daughters, the mother history, without affected sisters, accounted for a PAF of 3.61%, the sister history, without an affected mother, 3.01%, and the mother and sister history 0.43%; the total PAF was 7.05% (Table I). A genetic interpretation of these results would be that the mother history could be due to dominant heritable effects, sister history due to recessive effects or effects transmitted from the paternal grandmother and mother and sister history due to high-penetrant dominant effects. However, a more plausible interpretation may be that the first 2 histories show low-penetrant dominant effects, particularly because the SIRs were similar. In an attempt to consider the contribution of both parental lineages, the proportion of cases with a maternal family history could be multiplied by 2. This would double the PAFs of mother only (from 3.61% to 7.22%) and mother and sister (from 0.43% to 0.86%), giving a total PAF of 11.09%. Of course, environmental effects would modify the risks. As further results, we showed that the PAF values did not depend on the morphologic type of breast cancer; the total PAFs for specific morphologies ranged from 6.08–7.04. For in situ breast cancers, the PAFs were somewhat higher because the familial SIRs were higher.

The PAF values give an estimate on the heritable effects for breast cancer when only nuclear families can be studied for single-locus or additive effects.5, 36 In this population of daughters, the total PAF was thus 7.05% and 11.09% if 3 generations could be considered. These figures are considerably higher than the currently known genetic causes of breast cancer, mainly the BRCA1/2 and ATM genes.37–39 The efforts to identify new high-risk susceptibility genes have not been rewarding; many recent linkage studies on breast cancer appear to be difficult to reproduce, probably for a number of reasons.40–42 On the other hand, the present PAFs are lower than the 27% heritable proportion found in a twin study.7 The result from the twin study was based on a multifactorial threshold model, the assumptions of which have been discussed.5, 36 However, these results may not be incompatible if nonadditive polygenic effects operate in breast cancer. Age-incidence relationships in contralateral breast cancer and breast cancers among twins are unique in that the incidence rates do not increase by age, which may be an indication of monogenic or polygenic effects on a small susceptible population.43, 44 Further progress in the understanding of the heritable basis of breast requires coordinated efforts of genetic epidemiology and molecular genetics.


The Family-Cancer Database was created by linking registers maintained at Statistics Sweden and the Swedish Cancer Registry.