- Top of page
- Subjects and methods
Mammographic screening programs for breast cancer have been implemented in many countries and opportunistic mammographies are taken as a diagnostic method. The consequences of the wide application of this technology to age-incidence relationships in breast cancer have not been clarified nor is its effect on familial risk estimation. It was assumed that if screening and diagnostic methods bias familial risk, the highest risk should be noted for sisters diagnosed close in time. Age-specific incidence data were collected from the EUCAN database and from cancer registries of Finland, Norway and Sweden. The Swedish Family-Cancer Database was used to analyse risks for breast cancer among sisters, depending on the time since the first sister was diagnosed with breast cancer. Age-incidence patterns deviated between Germany, with low mammographic coverage, and Sweden, the Netherlands, the UK and France, with variable levels of coverage. The annual age-incidence patterns in Finland, Norway and Sweden changed in concert with the targeted mammographic service. The risk of breast cancer for women with an affected sister, diagnosed between ages 50 to 64 years, was significantly higher within the same or the subsequent year of the sister's diagnosis compared to 5+ years, accounting for 7.3% of all patients. The ordered increase in age-specific incidence of breast cancer in the women targeted by screening studied suggests that mammographic screening is one important factor responsible for the shift of the age of onset for breast cancer towards middle age. However, the effects on the estimation of familial risk are so far small. © 2005 Wiley-Liss, Inc.
Population screening for a cancer may cause major changes in its total and age-specific incidence, which has been witnessed for cervical cancer.1, 2 If the risk factors of a cancer remain constant, an effective screening should reduce disease-specific mortality. However, at least initially, the detection of asymptomatic tumours may cause an increase in incidence in the screened population, “a screening effect”, which may either be due to an earlier diagnosis of tumours (lead time shift) or diagnosis of tumours that would never have been detected (true overdiagnosis).3 In the course of introduction of screening technologies for breast cancer, the issue of overdiagnosis has been of focal interest and the subject of an unsettled debate, which we do not aim to address in the present work.3, 4, 5, 6, 7 On the other hand, less attention has been paid to the changes in age-incidence relationships; data from Sweden and Norway show large increases in age groups that are targeted to screening.6, 8 These changes have taken place at the time when there has been an overall increase in the incidence of breast cancer in all the Western countries. For any cancer for which effective screening methods or methods for diagnosing early or latent tumours are being used, the possible interference with the concept of familial aggregation needs to be considered. Furthermore, age of onset is an important variable in the evaluation of familial risks, which are likely to be modified when large changes in age-incidence data in the whole population are taking place. On breast cancer, no relevant data are available to address the possible effects of diagnostic and screening methods on familial risk. The recent Swedish data on prostate cancer showed shifts in age-incidence data towards younger ages, concomitant with an increasing use of prostate-specific antigen (PSA) screening, and an apparent tendency for a synchronous diagnosis of prostate cancer among brothers.9
In our study, we collected data on age-incidence relationships for breast cancer in various European countries with regard to the coverage of organised mammographic service. More specifically, annual data from the Finnish, Norwegian and Swedish Cancer Registries were used to span the periods of the implementation of the nation-wide mammographic screening programs. We suspect that screening and current diagnostic methods may also affect the detection of familial cancers, which may have implications for recommending women to mutation testing and for the overall risk assessment. Thus, the familial aggregation of breast cancer between sisters was examined as a function of the time after diagnosis of the first breast cancer. The nation-wide Swedish Family-Cancer Database was used. If detection bias were present, familial risks should be highest immediately after the first sister was diagnosed.
Subjects and methods
- Top of page
- Subjects and methods
Pilot studies on mammographic screening were stated in Sweden in the mid-1970s, and in the course of 23 years, a nation-wide mammographic screening was implemented; some 81% of the eligible women have participated.10 Invitation of women to screening has covered age groups of 40 to 74 years in counties with the widest range and from 50 to 69 years in those with the narrowest range; the screening intervals have varied from 18 to 24 months. Also other European countries initiated national mammographic screening schemes in the late 1980s, including Finland, the UK, the Netherlands and France.3, 11 Finland attained a 90% national coverage in the target population of 50- to 59-year-old women in some 5 years and subsequently extended the service to 60 to 64 year olds but with a lower coverage.12 In the Netherlands and UK, about 80% coverage was attained in some 10 years, whereas in France only 50% coverage was attained. The target population has been 50 to 64 year olds in UK, and even older groups in the Netherlands and France.3 A national program was stated in Norway in 1996.6 Only pilot programs have been carried out in Germany. In any country, opportunistic mammographic studies are carried out in clinics, independent of any organised services.
Age-incidence data for female breast cancer was analysed from the EUCAN database, which estimates the incidence of cancer for the year 1995.13 Annual age-specific data for breast cancer were collected from Cancer Registries of Finland, Norway and Sweden.
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.14 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, totalling over 10.5 million individuals. It has been updated in year 2004 to include cancers from the nation-wide Swedish Cancer Registry from years 1958 to 2002. The Database is organised in 3.6 million families, with parents and offspring. Our study included daughters in families with 2 or more sisters. The first women diagnosed with breast cancer in the family were the probands. The sisters of the probands were followed-up from the proband's year of diagnosis to the diagnosis of first cancer, death, emigration or the closing date of the study, December 31, 2002. Standardised incidence ratios (SIRs) were used to compare the incidence of breast cancer among sisters of probands with the incidence of breast cancer in the general population. SIRs were calculated as the ratio of observed (O) to expected (E) number of cases using the cohort method for sibling risk.15 The expected numbers were obtained from age- and calendar year-specific standard incidence rates for the general population. Confidence intervals (95%CI) and p-values (1-sided tests using the SIR for 5+ years as a reference) were calculated assuming a Poisson distribution.
- Top of page
- Subjects and methods
Figure 1 shows the age-specific incidences of breast cancer in year 1995 from 5 European countries. In Germany, 2% of the female population older than 49 years has been reported to have access to mammographic services.3 In the other countries, mammographic screening programs had been in place, with the largest proportion of the target population covered in Sweden and in the Netherlands.
Figure 1. Age-specific incidences of breast cancer in 1995 from The Netherlands, Sweden, France, United Kingdom and Germany; the incidence curves were drawn using data from EUCAN.
Download figure to PowerPoint
Figure 2 shows changes in age-incidence for the 3 Nordic countries. The target population for screening in Finland consisted of 50- to 59-year-old women; the maximum incidence in this period was reached between age 50 and 54 years. Women aged 60 to 64 years have also been included in the screening program but the coverage has not reached the level of the primary target population, yet the maximum incidence shifted towards that age group in 2002. In Norway, 40% of the 50- to 69-year-old women were screened in a 2 year period every 2 years, starting in 1996. The remaining 60% of the age group was covered more gradually. A major upward flux of the incidence took place in the target age groups. In Sweden, the completion of the national scheme lasted for 23 years and the changes were more gradual than in Finland and Norway. However, the most recent age-incidence pattern resembled that of Finland, with the exception of a relatively higher incidence in the 65 to 69 year olds, an age group covered in Sweden but not in Finland.
Figure 2. Age-specific incidences of breast cancer in Finland, Norway and Sweden before, during and after the establishment of national screening programs. The last available incidence data are also presented.
Download figure to PowerPoint
The possible effects of changing incidence trends on familial risk were examined by comparing the incidence of breast cancer among sisters of probands (27,137 women) with the incidence of breast cancer in the general population (1,924,392 women). The data included 817 families with 2 daughters affected by breast cancer, the first one diagnosed between ages 30 and 70 years. The data were stratified by the age at diagnosis of the first sister presenting with breast cancer (Table I). When the first sister was diagnosed before age 50, no systematic effects could be observed but the numbers of cases for single years were small. When the first sister was diagnosed at age 50 to 54 years, the risk for the second sister was 3.08 (95% CIs 1.85–4.82) if both cancers were detected in the same year; this risk was higher than after 5 or more years (p-value = 0.043). Among the 179 women with a sister diagnosed at age 50 to 54 years, 19 patients (10.6%) were diagnosed the same year as their sisters. In the next 2 diagnostic age groups, the highest SIRs were observed 1 year after the first diagnosis, and they were significantly higher than the corresponding SIRs after 5 or more years; the cases detected 1 year after the first diagnosis included 25.3% of the women with a sister diagnosed at age 55 to 64 years. Overall, considering the 50 to 70 year diagnostic groups together, there was a difference of borderline significance in the SIRs between year 1 (SIR 2.29) and year 5+ (1.76).
Table I. Age And Calendar Period Standardized Incidence Ratios For Sisters Of Women Affected By Breast Cancer (SIR)1
|Age at diagnosis of first breast cancer (years)||Time since diagnosis of first sister (years)||Observed number of cases||%||SIR||95% CI2||p-value3|
- Top of page
- Subjects and methods
The present data show the large shifts in the age-incidence patterns of breast cancer in countries where national mammographic screening has been offered. Although we have no direct evidence that the observed changes are due to mammographic screening programs, the international comparisons and time-dependent changes in the targeted age groups indirectly implicate mammographic screening. Of course, patterns of risk factors for breast cancer, such as increasing use of hormonal replacement therapy (HRT), early menarche, late first pregnancy and low parity, are also changing but they would not act in such an orchestrated fashion on the screened populations. HRT use has increased rapidly in the European countries but not exceptionally in the Nordic countries;16 furthermore, the risk of HRT, based on randomized trials, appears to be too small to explain the large shifts in age-incidence patterns.17 Among the European countries analysed, Germany has been covered only by pilot mammographic schemes, assumed to reach no more that 2% of women aged 50 or more years.3 The age-incidence pattern for Germany was exactly like that of Finland, Norway and Sweden in the prescreening period. For other European countries, the Netherlands, UK and France, the age-incidence curves were consistent with a certain level of mammographic penetration.3, 11 In the Nordic counties there is no noticeable decrease in incidence at high ages, which would be predicted by an effective screening. However, mammography may have eliminated an increase in age groups over 70 years, which could have been expected, because of the long-lasting increasing incidence trends for breast cancer. The present ecological data are no strong argument for or against beneficial effects of mammographic screening. However, they indicate that a drastic remodelling of the age-incidence relationships has taken place and is yet to continue. Undoubtedly, opportunistic screening will have similar effects, which will be noted in all developed countries when the proportion of screened middle-aged women will increase. Breast cancer will definitely become a disease of middle aged women. In Finland, 20% of breast cancers were diagnosed before age 50 years in 2002 (www.cancer.fi).
Clustering of cancer in family members may be an indication of a heritable disease. Family history is the basis for clinical genetic counselling, for which the numbers and types of cancers and their age of onset are recorded.18, 19 Uncontrollable variables, such as family size add uncertainty to clinical risk estimation. We suspected that screening and current diagnostic methods could also influence the detection of familial cancers through a lead time bias, i.e., an early diagnosis of breast cancer in a sister of a patient. It can be assumed that sisters of a newly diagnosed sister actively seek screening and other medical advice, and thus their diagnosis will be affected by the lead time bias. In our study, some evidence was found to support the existence of such a bias in the groups when the first sister was diagnosed between ages 50 to 64 years, the primary target groups for mammographic screening. The second sister had a significantly higher risk to be diagnosed in the same or the subsequent year compared to 5+ years. However, the differences in SIRs were not very large (the largest difference in SIR was 3.08 at year 0 and 1.87 after 5 or more years) and the bias affected only 7.3% of the women [(19+26+15)/817] with an affected sister. We conclude that up to now the screening and diagnostic methods have not appreciably distorted familial risks for breast cancer. Moreover, the results tone down any hasty conclusions about the unfortunate diagnoses of 2 sisters almost synchronously. They have a familial disease but the risk may not be as large as the 2 closely timed diagnoses might suggest.