Familial association of histology specific breast cancers with cancers at other sites
Breast cancer histologies show important differences in their incidence pattern, method of detection and management. Aggregation of breast cancer occurs also in families diagnosed for cancer at sites different from the breast. Therefore, the familial association of histology specific breast cancers with cancers at other sites is of great interest. The nationwide Swedish Family-Cancer Database was used to calculate standardised incidence ratios (SIRs) for breast cancer when parents or sibling were diagnosed with cancer at the most common sites. Significant SIRs were found when parents had breast, ovarian, laryngeal, endometrial, prostate, lung and colon cancers. If women were diagnosed before the age of 50 years, the SIRs were significant when parents were diagnosed with breast, ovarian, and prostate cancers, and leukaemia, and when siblings were diagnosed with squamous cell skin, pancreatic, breast and endometrial cancers. If mothers were diagnosed with breast cancer, histology-specific SIRs were ranked as comedo > tubular > ductal > lobular; SIR for medullary carcinoma was not significant but it was high when mothers presented with ovarian cancer. Other associations were between the upper aerodigestive tract and lobular, colon and comedo, larynx and ductal cancer. Moreover, cervical cancer was associated with comedo and endometrial cancer with the medullary histology. In conclusion, histology-specific breast cancers were associated with specific cancer sites and the strength of the association varied among histologies. © 2004 Wiley-Liss, Inc.
Estimates of familial risk for breast cancer are important for clinical counselling, understanding of cancer aetiology and for identification of cancer-related genes. Only a small proportion of familial risk can be explained by mutations in BRCA1/2 or other known heritable genes.1 The role of BRCA1 mutations in prostate and colon cancer risk remain controversial;2, 3, 4, 5, 6, 7 the risk of pancreatic, gastric, laryngeal, uterine, liver and skin cancer and ocular melanoma has been reported to be increased for BRCA2 mutations.8, 9, 10, 11, 12, 13, 14, 15, 16, 17 The association between BRCA2 mutations and prostate cancer remains debatable.18, 19, 20 Thus, consideration of familial histories for sites different from the breast is of interest from a genetic point of view. For women without detected mutations, family cancer specialists must rely on risk estimates, which include information about the age of diagnosis and the number of affected relatives. Therefore, estimates of familial risk at specific ages of diagnosis are of practical relevance.
The 2 most common histopathological types of the breast cancer are ductal and lobular carcinoma, which account for 70–80% and 5–10% of all invasive breast carcinomas, respectively.21 Rarer types include tubular, comedo, medullary and mucinous carcinomas; a proportion of breast cancers presents with mixed histologies. Breast cancer histologies show important differences in their incidence pattern, method of detection and management. Furthermore, BRCA1/2 mutations have been associated to specific breast cancer histologies, showing somewhat lower frequencies of lobular morphology and BRCA1 related tumours exhibiting a small excess of medullary carcinomas.22, 23 Estimates of familial risks for specific breast cancer histologies are therefore of great interest.
Although a large number of epidemiological studies have been carried out to quantify familial risk for breast cancer, data regarding histology are sparse, especially if the woman's relative (proband) was diagnosed with cancer at sites different from the breast. The nationwide Swedish Family-Cancer Database, which has been updated in 2002, was used to assess the association of early onset breast cancer and histology specific breast cancers with familial cancers at the most common sites. The Database offers unique possibilities for reliable estimation of familial risk because the data on family relationships and cancers are obtained from registered sources of practically complete coverage and free from bias.24
MATERIAL AND METHODS
The Swedish Family-Cancer Database was created in the middle of the 1990s by linking census information, death notifications and the administrative family register at Statistics Sweden to the Swedish Cancer Registry. The Database was updated in the end of 2002 to include more than 10.34 million individuals born in Sweden after 1931 and parents as well as more than 810,000 invasive cancers diagnosed after 1961. Information on both parents for each offspring (daughter) included in the study permitted to organise the Database in 3.27 million families. Because there are some missing links between parents and deceased offspring in the Database before 1991, only breast carcinomas diagnosed between 1991 and 2000 were considered. The study included 20,742 (67,575) cases of invasive breast cancer on 3.25 (2.23) million daughters (mothers).
The Swedish Cancer Registry relies on separate compulsory notifications of cases from clinicians who diagnosed a neoplasm and pathologists/cytologists. A 4-digit diagnostic code according to the 7th revision of the ICD-7 was used until 1992 and thereafter ICD-O-2. In our study, upper aerodigestive tract cancer included lip, mouth, tongue and pharynx; small intestine, colon and rectal cancers included only adenocarcinomas; lung cancer was defined by ICD-7 codes 162–163 and eye cancer included melanoma only. The percentage of histologically or cytologically verified cases of cancer was close to 100%.25 Because cancer histology has been recorded according to the Systematized Nomenclature of Medicine (www.snomed.com) after 1993, the analysis of histology specific breast carcinomas considered diagnoses between 1993 and 2000 and included 17,763 (67,182) cases of invasive breast cancer on 3.23 (2.22) million daughters (mothers).
Standardized incidence ratios
Follow-up started for daughters at birth, immigration date or first year of the study (1991 or 1993), whichever came latest. Follow-up terminated at diagnosis of invasive/benign breast cancer, death, emigration date, December 31, 2000 or the age of diagnosis as specified in the study. The 1.78 million daughters who became mothers were treated independently, first as daughters and then as mothers.
Standardized incidence ratio (SIR) was used to estimate familial relative risk. SIR based on only parents diagnosed (SIRparent), only full siblings diagnosed (SIRsibling) or both parents and full siblings diagnosed (SIRparent&sibling) as probands. For prostate and testis cancers, probands were males; for breast, cervical, endometrial, ovarian and other female genital cancers, probands were females and for the rest of the cancer sites, probands were both males and females. SIR was calculated as the ratio of the observed to the expected number of cases. The expected numbers were computed from age (5-year bands), parity (6 groups from “any parturition” to “more than 5 parturitions”), socioeconomic status (6 groups), age at birth (5 groups, 5 years bands between “before age of 20” and “after age of 35”) and residential area (4 groups) standardised incidence rates. Ninety-five percent confidence intervals (CI) were calculated assuming a Poisson distribution;26p-values were adjusted for multiplicity by using Bonferroni's method.27 SIRsibling based on daughters with at least 1 full sibling at risk and was calculated by the cohort method as described elsewhere.28
Table I presents the number of observed and expected cases of invasive breast cancer, SIRparent, SIRsibling and the corresponding 95% confidence intervals for daughters when their relatives were diagnosed with cancer at different sites. Significant p-values after adjustment for multiple comparisons by Bonferroni's method are also shown. SIRparent&sibling was significant (different from 1) for breast and “all” cancer sites and is presented as a footnote to Table I. Significant increases in SIRparent were found for breast, ovarian, laryngeal, endometrial, prostate, “all”, lung and colon cancers, with a range from 1.62 to 1.07. Estimated SIRsibling was significant for bone, breast, squamous cell skin and “all” cancers, with a range from 3.54 to 1.19.
Table I. Risk for Breast Cancer in Offspring by Familial Cancer1
|Upper aerodigestive tract||175||0.99||0.85||1.15||25||1.17||0.75||1.72|
|Other female genital||39||0.88||0.62||1.20||4||1.08||0.28||2.78|
|Skin, squamous cell||357||0.98||0.88||1.08||34||1.53||1.06||2.13|
Table II shows the estimated risks after taking the age of diagnosis into consideration. Results are presented for the sites that were significant in any of the comparisons of the present study. If daughters were diagnosed before the age of 50 years, the SIRparent was significant for breast, ovarian and prostate cancers and leukaemia and the SIRsibling was significant for squamous cell skin, pancreatic, breast and endometrial cancers. The SIRmother&sister was significant for breast cancer. If the relative was diagnosed before age 50 years, the SIRmother was significant for breast, ovarian and endometrial cancers. SIRmother for breast cancer was particularly increased. SIRsibling was not significant for any site. If both the case and the proband were diagnosed before the age of 50 years, SIRmother for breast and ovarian cancers increased significantly. SIRsister was significant for breast and ovarian cancers.
Table II. Risk for Breast Cancer after Restriction of Age of Diagnosis1
|Cases diagnosed before age of 50|
|Upper aerodigestive tract||67||0.94||0.73||1.20||4||0.58||0.15||1.51|
|Skin, squamous cell||146||1.06||0.90||1.25||21||3.003||1.86||4.60|
|Proband diagnosed before age of 50|
|Upper aerodigestive tract||5||0.59||0.18||1.38||3||0.62||0.12||1.82|
|Prostate||1||0.45||0.00||2.59|| || || || |
|Skin, squamous cell||9||1.79||0.81||3.42||10||1.18||0.56||2.18|
|Both case and proband diagnosed before age of 50|
|Upper aerodigestive tract||4||0.70||0.18||1.82||1||0.31||0.00||1.76|
|Prostate||1||0.62||0.00||3.53|| || || || |
|Skin, squamous cell||8||2.20||0.94||4.36||8||1.36||0.58||2.68|
The number of daughters diagnosed before the age of 69 years was 12,182 (ductal), 2,456 (lobular), 1,670 (tubular), 860 (comedo), 292 (medullary) and 243 (mucinous). The mean age of onset of the carcinomas was 43.6 (medullary), 48.5 (comedo), 51.7 (ductal), 52.7 (comedo), 53.1 (tubular) and 53.5 (lobular). Morbus paget, papillary carcinoma and cistosarcoma phyllodes were also analysed but the SIRs were not significant for any site.
Table III presents SIRparent for ductal, lobular, tubular, comedo, medullary and mucinous carcinomas because only a few SIRsibling were significant. For daughters diagnosed with ductal carcinoma, SIRparent was significant for breast, laryngeal, ovarian, lung and prostate cancers, with a range from 1.62 to 1.09. The SIRsibling was significant for breast (2.00) and squamous cell skin (1.87) cancer (data not shown). If daughters were diagnosed with lobular carcinoma, SIRparent was significant for breast and upper aerodigestive tract cancers and SIRsister was significant for breast (2.13). When daughters were diagnosed with tubular carcinoma, only SIRs for breast cancer were significant. The SIRsister was 2.44. For daughters diagnosed with comedo carcinoma, the SIRmother was significant for ovarian, breast, cervical and colon cancers and they ranged from 2.33 to 1.46. For daughters diagnosed with medullary carcinoma, SIRparent was significant for ovarian and endometrial cancers. If daughters were diagnosed with mucinous carcinoma, SIRparent was significant for breast cancer.
Table III. Risk for Histology Specific Breast Cancer in Daughters of Affected Parents1
|Upper aerodigestive tract||83||0.93||0.74||1.15||28||1.52||1.01||2.20||14||1.12||0.61||1.88||3||0.49||0.09||1.44||1||0.46||0.00||2.63|| || || || |
|Bone||8||1.08||0.46||2.15||1||0.66||0.00||3.79|| || || || ||1||1.94||0.00||11.13|| || || || ||1||7.13||0.00||40.89|
If mothers were diagnosed with breast cancer, significant SIRmother could be ordered as mucinous (2.01) > comedo (1.94) > tubular (1.63) > ductal (1.62) > lobular (1.55), but differences among histologies were not significant. SIRmother for medullary breast carcinoma was not significant. If sisters were diagnosed with breast cancer, significant SIRsister could be ordered as tubular (2.44) > lobular (2.13) > ductal (2.00), but differences were not significant. If mothers were diagnosed with ovarian cancer, significant SIRmother could be ranked as medullary (3.97) > comedo (2.33) > ductal (1.37). The SIRmother for medullary carcinoma was significantly higher than the SIRmother for any histology.
Most studies of the familial association of cancer are based on interview data; for breast cancer such data appear to be reasonably reliable with probably little need for adjustment for ascertainment bias.29, 30 In contrast, our study relied entirely on registered data of complete coverage. Other important features of the present study were the large number of cases analysed, the standardisation for parity and age of first birth, and the exact definition of the proband status.
To deal with the issue of multiple comparisons, p-values were adjusted by using Bonferroni's method. This was a conservative analysis and some established associations, such as breast and ovarian cancer when cases and proband were diagnosed before the age of 50 (Table II), were no longer significant after the multiplicity adjustment. Significant findings at the 95% confidence level were examined for biological plausibility and consistency with previous literature, as discussed later.
SIRs were lower than results based on previous versions of the Swedish Family-Cancer Database31 due to standardisation and ageing of the offspring. A joint analysis of benign and invasive breast carcinomas (results not shown) suggests that the widespread use of screening mammography in the 1990s may also have resulted in lower familial risks for invasive breast cancer. SIRparent for breast, ovarian, endometrial and prostate cancers were significant in the present study and in the earlier report,31 and we found additional significant SIRparent for laryngeal, lung and colon cancers. The familial association of breast cancer with laryngeal cancers may be attributed in part to alcohol consumption.32 For colon cancer, the association could rest on oestrogen hormones, which may be involved in malignant colorectal tumours as shown in epidemiological and clinical studies.33
In addition to breast cancer, SIRsibling was significant for skin squamous cell and bone cancers. Three of the 6 cases related to a sibling history for bone cancer were diagnosed before an age of 50 years. Moreover, if only daughters diagnosed before age of 50 were considered, SIRparent for leukaemia and SIRsibling for pancreatic and skin squamous cell cancers were significant, the association of breast and skin was significant even after the adjustment for multiple comparisons. The clustering of early onset breast cancer with bone cancer and leukaemia may be attributed, at least to some extend, to Li-Fraumeni syndrome.34
Whereas SIRmother relied on mothers of any age, SIRsister based on offspring with at least 1 full sibling at risk, all of whom were less than 69 years of age. Thus, SIRmother and SIRsister are not comparable. Nevertheless, SIRmother was higher than SIRsister after restriction of proband's age of diagnosis in Table II. As discussed in several studies,35, 36 the unequal number of parturitions and the different calendar year of diagnosis of mothers and sisters could result in biased risks. If age of diagnosis was restricted to 50 years, 97% of the daughters were parous and the median calendar year of diagnosis of mothers was 1971 and that of sisters 1996. Therefore, in order to avoid confounding effects, SIRs were calculated for women diagnosed in the 1990s who gave birth twice, before being diagnosed with breast cancer. SIRmother was 1.99 (0.19–7.33) and SIRsister was 1.99 (1.03–3.76) for these women. These data could suggest that early onset breast cancer is mainly due to dominant effects. However, the data are uninformative of the causes in late onset breast cancer. Since information about proband's parity and calendar year of diagnosis is usually available, these confounding effects should be taken into account in future studies about the familial association of early onset breast cancer.
Breast cancer histologies differ in many aspects. A disproportionate increase of lobular breast cancer compared to other histologies has been shown in recent studies37, 38, 39 and can be probably attributed to the use of combined hormone replacement therapy in the last 10 years. The mean age at diagnosis of medullary carcinoma has been reported to be significantly lower than the age at diagnosis of other histologies,40 which is in agreement with the results of our study. Also calcifications, firmness and opacity of breast tissue vary with histopathology,41e.g., lobular carcinoma cells tend to infiltrate into the surrounding breast cancer tissue as single cells and often elicit minimal stromal response, thereby resulting in tumours that are difficult to palpate and more likely to be missed by mammographic screening. The association between method of detection and breast cancer histology was assessed by Newcomer et al., who showed that lobular and medullary carcinomas are more often detected by clinical breast examination than other histologies.42 The management of breast cancer depends on the histology of the carcinoma as well. Endocrine therapy is guided by the presence and degree of expression of oestrogen and progesterone receptors, and lobular carcinomas are more likely to be oestrogen and progesterone receptor positive. On the other hand, mastectomy may be preferred for lobular carcinomas because it is known that these tumours are more likely to be bilateral and to recur compared to other breast carcinomas. In addition to these differences, the incidence of ductal and lobular carcinomas increased with increasing ages of diagnosis, while the incidences of tubular, comedo, medullary and mucinous carcinomas reached a plateau between 50 and 55, 45 and 50, 35 and 40 and 45 and 50 years, respectively (results not shown). The association of early onset cancers, e.g., medullary breast and ovarian cancers, has probably a genetic basis, whereas the association of cancers diagnosed later in life, e.g. lobular and upper aerodigestive tract cancers, could also rest on environmental effects.
When mothers and daughters were affected with ductal cancer, the SIRmother was 1.64 (1.38–1.94). If mothers and daughters were affected with histologies different from ductal cancer, the SIRmother was 1.75 (1.61–1.91) (results not shown). Because histology has been recorded after 1993, the number of concordant breast cancers in the Database was low (5 lobular cancers, 3 tubular cancers, 1 comedo cancer and 1 medullary cancer). To what degree inherited factors influence histology remains somewhat open.43 We have been unable to locate any literature on the association of histology specific breast cancers with cancers at sites different from the breast. Data on familial breast cancer risks by histology specific breast cancer based on previous versions of the Swedish Family-Cancer Database44, 45 show the same trends as in the present study. SIRmother for breast cancer were ranked in the 3 studies as comedo > tubular > ductal > lobular > medullary (SIRmother for medullary carcinoma was not significant). SIRsister for lobular carcinoma was higher than SIRsister for ductal carcinoma in the 3 studies, but we found the highest risk for tubular carcinoma, whereas SIRsister for tubular carcinoma was placed between lobular and ductal risks previously. Since the present estimates include 5 new sisters diagnosed with lobular and 8 with tubular carcinoma between 1998 and 2000, we consider the last ranking more accurate.
Histology specific breast cancers were associated with specific cancer sites and the degree of association varied among histologies. We found interesting results in particular for medullary carcinoma. SIRmother for medullary breast cancer was not significant when mothers had breast cancer but it was 3.97 (2.04, 6.96) when mothers had ovarian cancer, i.e., significantly higher than SIRmother for ductal, lobular and tubular carcinomas in daughters of mothers with ovarian cancer. These data likely reflect the association among medullary cancer, ovarian cancer, BRCA1 mutations and early onset breast cancer. Other associations were between the upper aerodigestive tract and lobular, colon and comedo, and larynx and ductal cancer; also lung and prostate cancers associated with ductal cancer but the SIRs were not different from other breast cancer histologies. Moreover, cervical cancer associated with the comedo and endometrial cancer with the medullary histology. However, as we have no way of excluding chance association, familial risks for specific histologies of breast cancer should be analysed in future studies. The critical question that needs to be settled is to what extent the familial risks of histology specific familial risks in breast cancer differ and to what extent the causes are heritable or environmental.