Fax: +45-35257731 or +45-35257734
Development in incidence of breast cancer in non-screened danish women, 1973–2002—a population-based study
Article first published online: 5 DEC 2005
Copyright © 2005 Wiley-Liss, Inc.
International Journal of Cancer
Volume 118, Issue 9, pages 2366–2369, 1 May 2006
How to Cite
Fuglede, N., Langballe, O., Svendsen, A. L., Tjønneland, A., Dalton, S. O. and Johansen, C. (2006), Development in incidence of breast cancer in non-screened danish women, 1973–2002—a population-based study. Int. J. Cancer, 118: 2366–2369. doi: 10.1002/ijc.21654
- Issue published online: 21 FEB 2006
- Article first published online: 5 DEC 2005
- Manuscript Accepted: 22 SEP 2005
- Manuscript Received: 14 FEB 2005
- breast cancer;
- cancer registry;
- invasive ductal cancer;
- invasive lobular cancer;
The authors report on the incidence rates of breast cancer overall and by histology in a population of unscreened women constituting ∼80% of the total population of women in Denmark from 1973–2002, utilizing the files of the nationwide Danish Cancer Registry. The age-specific incidence rates of breast cancer increased throughout the period, and further, marked changes in the age-specific incidence pattern were observed, where the plateau and change of slope around the age of 46–48 in 1973–1981 shifted to around age 64–66 years in 1994–2002. Age-period-cohort modeling indicated that these changes were not attributable to a birth cohort effect. Although lobular breast cancer incidence increased more than ductal breast cancer incidence, this was only observed in the first decade after the introduction of the ICD-O system in Denmark and probably is attributable to this, whereas we observed no disproportionate changes by histology in any age group from 1988–2002. Thus, previous reports of a disproportionate increase in lobular breast cancer could not be confirmed in a non-screened population, whereas important changes over the past decade in the age-specific incidence pattern of breast cancer particular around the time of menopause were indicated. © 2005 Wiley-Liss, Inc.
Breast cancer is the most common female malignancy worldwide, and its global incidence rates are increasing.1, 2 Recent studies from the US and Switzerland3, 4 have reported that the incidence of invasive lobular breast cancer has increased disproportionately compared to the other verified invasive breast cancers. As these trends have also been reported to vary by age,3 these developments might influence the age-specific incidence pattern. Here, we report the development in age-specific breast cancer incidence rates overall and by histology in the Danish female population not exposed to systematic mammography screening, from 1973–2002.
Material and methods
Data have been obtained from the nationwide Danish Cancer Registry comprising information on all cancers diagnosed in Denmark since 1943 and classified according to a modified Danish version of the International Classification of Diseases, Seventh version (ICD-7).5 A total of 73,184 women aged 30–79 years were diagnosed with first primary breast cancer (ICD-7 codes 170.1–170.9) in Denmark, over a 30-year period between 1973 and 2002.
In Denmark, organized screening programs were initiated in Copenhagen municipality, Fyn county and Frederiksberg municipality in 1991, 1993 and 1994, respectively. Altogether, 100,000 women aged 50–69 were covered in the three programs, equivalent to some 20% of the total Danish female population in that age group.6 To analyze incidence rates of breast cancer without possible effects of mammography screening, we excluded these three screening regions from the study population throughout the full study period. The present study population thus included from 74% in 1973 to 81% in 2002 of the total Danish female population aged 30–79 years, with 55,897 cases of breast cancer.
We calculated the age-specific incidence rates of primary invasive breast cancer for 3-year age intervals and for 3-year calendar periods, using the cases from the Danish Cancer Registry, and the population estimates from Statistics Denmark of the defined background female population in the respective age groups and periods as the population at risk.
To investigate whether any change in the pattern of the age-specific incidence rates had occurred, we chose to examine the two separate time periods, 1973–1982 and 1994–2002, respectively, by fitting a model using the variables age, period and birth cohort. We restricted the age interval to include women aged 40–78 years, and constructed a two-way table of age-group by time period, using 3-year intervals. The diagonals of the table corresponded to synthetic overlapping cohorts. Each cohort included women born within a 6-year period, and the same women contributed to 2 adjacent cohorts. We analyzed a total of 13 age groups (40–42 years, 43–45 years,…76–78 years), 6 calendar periods (1973–1975, 1976–1978, 1979–1981 and 1994–1996, 1997–1999, 2000–2002) and 15 birth cohorts (1894–1899, 1897–1902, …,1957–1962), respectively. Note that there were 3 calendar periods in each of our study periods, and 4 calendar periods in between, which were not studied. The oldest 15 birth cohorts contributed to the first study period, whereas the last 15 contributed to the last study period, i.e., the middle 8 birth cohorts contributed to both study periods.
We modeled the incidence rates as a log-linear function of age, calendar period and cohort, allowing the effect of age to depend on the study period (1973–1981 and 1994–2002). We assumed that the number of cases in age-group a, in period p and cohort c was Poisson distributed with mean value λapc and fitted the following model:
a, age-group; p, period and c, cohort;
napc, person-years at risk in age-group a, period p and cohort c.
It was tested whether any of the two factors, period and cohort, could be excluded.
Since 1978, 95% of breast cancers in Denmark have been histologically verified and classified by the ICD-O. We classified all identified invasive breast cancers diagnosed in 1978–2002 into the following categories consistent with the WHO criteria:7 ductal carcinoma (ICDO-1 code 85003), lobular carcinoma (ICDO-1 code 85203), mixed infiltrating ductal and lobular carcinoma (ICDO-1 code 85223) and other breast cancers, including cancers not verified by histology.
The annual number of cases of breast cancer from 1978 through 2002 in women aged 30–79 years was stratified by histology, and the logarithm of number of cases in each subgroup was graphed as a function of calendar year. We divided the study period into two periods, 1978–1987 and 1988–2002, in an attempt to separately study the effect of the implementation of the ICD-O and the long-term development in the distribution of cases after the classification system was well-established. We focused our attention on the two histological types, ductal and lobular invasive breast cancer, and for each period, separately, we estimated the proportion of ductal cases (pay) as a function of year and age-group, assuming that the number of ductal cases was binominally distributed. Hence, the model fitted was given as follows:
where a is the age-group, y is the year and pay is the proportion of ductal cases in age-group a, in year y out of the total number of ductal and lobular breast cancer cases. SAS version 8.2 was used for the analyses.
Results and discussion
The age-specific breast cancer incidence rates increased steadily over time (Fig. 1). Additionally, changes in the age-specific breast cancer incidence pattern from the beginning to the end of the study period were observed (Fig. 1). During the last 9 years of observation, the plateau and change of slope observed at age46–48 years in the period of 1973–1981 seemed to shift to the 64–66 year age-group in 1994–2002.
The full 3 factor model was fitted and had a deviance of 35 on 28 degrees of freedom. It was not possible to remove the period effect (−2 log Q = 19, df = 3, ˜p = 0.0004), but the cohort effect was not significant (−2 log Q = 20, df = 29, ˜p = 0.48) when adjusted for the other factors. Thus, the cohort effect in the model was negligible and was removed. The effects found can therefore be entirely attributed to age and calendar period. We used the age-group 55–57 years as reference, and the age-specific relative risks obtained by separate study period are plotted in Figure 2.
The estimated relative risk for breast cancer in women diagnosed in 1973–1981 increased steeply until age 46–48, after which there was a level off and gradual change of slope, increasing again from about the age 61–63 years (Fig. 2). In women diagnosed in 1994–2002, the estimated relative risk for breast cancer increased steadily with no apparent inflection until age 64–66 years, after which the risk seemed to level off (Fig. 2).
The logarithm to the annual number of breast cancers by histological type from 1978 to 2002 is plotted in Figure 3. From 1978 to 2002, the number of ductal breast cancer cases increased continuously throughout the study period (Fig. 3), whereas the number of lobular breast cancer cases increased dramatically from 1978 to 1982 and continued to increase but at a much slower rate. In parallel, the number of breast cancer cases grouped as “other” decreased from 1978 to 1982, after which period the number diagnosed annually stabilized. Finally, the number of mixed infiltrating ductal and lobular breast cancer cases increased from the first recorded case in 1995 to 32 in 2002. Age-specific trends in the different histological types were only observed in 1978–1987 (Table I), with a significant disproportional increase in lobular breast cancer compared to ductal in women aged 50 years and older. No disproportionate changes by histology were observed in any age-group, from 1988–2002. Adding the mixed ductal and lobular breast cancers to the lobular breast cancer had no influence on these results.
|Age group (years)||1978–1987||1988–2002|
|β1||95% CI||p-value||β1||95% CI||p-value|
|30–39||0.0025||(−0.0049 to 0.010)||0.51||0.0021||(−0.00096 to 0.0051)||0.18|
|40–49||−0.0016||(−0.0066 to 0.0035)||0.54||−0.0010||(−0.0030 to 0.0011)||0.34|
|50–59||−0.010||(−0.014 to −0.0058)||<0.001||−0.0002||(−0.002 to 0.0017)||0.85|
|60–69||−0.0064||(−0.011 to −0.0020)||0.004||0.0004||(−0.0016 to 0.0023)||0.72|
|70–80||−0.013||(−0.017 to −0.0087)||<0.001||−0.0011||(−0.0034 to 0.0012)||0.33|
Our data demonstrate a marked increase in incidence rates of first primary breast cancer over the last 30 years. Also, we observed a recent shift and change in the age-specific pattern previously described in Denmark.8, 9 The age-specific incidence curve for the past decade (1994–2002) showed a steady increase with no inflection at the time of menopause and did not seem attributable to a birth cohort effect. The inflection on the age-specific incidence curve at the time of menopause, commonly denoted Clemmesen's Hook10 observed in the earlier part of the study period has been a well-described characteristic for age-specific breast cancer risk. To our knowledge, this change in age-incidence pattern has not been reported before, although observations from the SEER population in the United States, where a continuous increase in age-specific incidence rates from 1990–1997 was observed, with only a slight change in level at age 45–50 years, might indicate similar changes.1, 11
We included only about 80% of the total Danish female population not enrolled in systematic mammography screening programs in our study population, thus excluding that introduction of screening can explain the observed changes over the past decades in breast cancer incidence we present here. Increased use of clinical mammography in terms of diagnostic mammography and in opportunistic screening of asymptomatic women may of course attribute to the recent changes. In general, however, use of diagnostic mammography has been relatively limited in Denmark and is much lower than in other European countries.12 A recent Danish study showed that very few women aged 50–69 years seek opportunistic screening, even in the 9 counties with a positive policy towards opportunistic screening,6 suggesting that other factors attribute to the change in development in age-specific rates.
The disproportionate increase in lobular breast cancer reported in a couple of recent studies3, 4 coupled with a higher increase in lobular breast cancer incidence among women aged 50 years or above compared to women below 50 years of age3 could also be hypothesized to attribute to such an age-specific incidence curve development as observed in the present data. Recent studies suggest that the use of hormone replacement therapy is more strongly related to lobular breast cancer than to ductal breast cancer risk.3, 4, 13, 14 We did, however, only observe a disproportional increase in lobular breast cancer in the first decade of available histological data, probably reflecting the introduction of the ICD-O system in Denmark in 1978. This interpretation is further supported by the parallel decrease in the breast cancers classified as other over the same period. Over the past 15 years studied, the increase in lobular breast cancer was no different to that of ductal breast cancer and we found no suggestion of any age-specific increase in lobular breast cancer compared to ductal cancers.
Although similar criteria have been used to classify breast cancers across studies,3, 4 misclassification due to introduction of major changes in diagnostic and pathologic criteria of lobular breast cancer cannot be ruled out. In line with our findings, a decrease in the rates of breast cancer grouped as other, subsequent to introduction of a broadened definition of the lobular carcinoma, was observed in a smaller scale Swiss study.4 However, the authors state that misclassification could not account for the continuing increase of lobular cancers observed 15 years or more after this change in criteria.4 Another large scale study using SEER data from the United States offered no information as to the exact timing of criteria introduction/revision in relation to their study period.3
Differences between the study populations may account for the different findings, including the obvious differences in screening exposure or distributions of factors such as obesity,15 alcohol,16, 17 reproductive factors18 and use of hormone replacement therapy13, 19, 20, 21 but cannot be evaluated further because of the descriptive character of our data.
In conclusion, we could not confirm earlier findings of a disproportionate increase in lobular compared to ductal breast cancer. Our observation of a possible recent change in the age-specific risk for breast cancer specifically around the time of and after menopause in a population not exposed to screening is to our knowledge not described before and warrants replication and further clarification in epidemiologic studies.
- 7TavassoliFA, DevileeP, eds. Pathology and genetics of tumors of the breast and female genital organs. Lyon: IARC Press, 2003.