the relationship between age and incidence of breast cancer population and screening program data

1896

The Relationship Between Age and Incidence of Breast Cancer Population and Screening Program Data

Larry G. Kessler, ScD

Despite extensive study of breast cancer incidence, in- cluding specific studies of the relationship between age and breast cancer incidence, the picture remains confus- ing. This article examines not only the relationship between age and breast cancer, but also trends over time related to this relationship to discern the underlying true age-incidence pattern. The age-incidence curve changes around the menopausal period, most likely due to hormonal changes 10 to 15 years earlier, flattens out in the 40 to 50 year old age range, and then increases as age increases. Recent data showing decreased risk of breast cancer incidence at older ages, e.g., older than 75 years of age, relative to younger ages, are likely an artifact of recent increases in breast cancer screening in the United States. This picture is consistent with increases in screening and with notions of lead time created by increased screening. The increase in screening that has changed the age-incidence relationship may eventually deliver benefits to United States women in terms of mortality deficits, but this is not guaranteed unless screening becomes routine practice and high-quality therapeutic intervention and follow-up occurs as well. Cancer 1992; 69:1896-1903.

Breast cancer has been among the most studied of cancers. Despite this, a full understanding of many aspects of the disease is still elusive. In particular, breast cancer has an unusual age-incidence pattern when viewed with population registry data, and trends in breast cancer incidence over the past few decades in the United States and other countries have been perplexing. In the most recent data from the Surveillance, Epi-

Presented at the American Cancer Society Workshop on Guide- lines and Screening for Breast Cancer, Pasadena, California, October

From the Applied Research Branch, National Cancer Institute, Bethesda, Maryland.

The author thanks Sam Shapiro, Eric Feuer, and Martin Brown for helpful comments on the manuscript. Any remaining errors re- main the author’s responsibility.

Address for reprints: Larry Kessler, ScD, 9000 Rockville Pike, EPN Room 313, Bethesda, MD 20892.

Accepted for publication November 26, 1991.

11-13, 1991.

demiology, and End Results Program (SEER) (Fig. 1, top line), the incidence rate for female breast cancer rises with advancing age until approximately 45 years of age when there is a leveling referred to as Clemmesen’s hook. The curve then rises until its peak at 75 years of age and then begins to decline.’ This is a sharply different pattern than the age-incidence curve in 1973 (Fig. 1, lower line).

These observed incidence rates are the products of the epidemiology of the disease, personal behaviors, and activities of health-care providers. This article will attempt to disentangle various factors that have con- spired to produce the current age-incidence patterns of breast cancer. The following questions will be ad- dressed:

1. What do the population data on incidence and mortality indicate about the relationship of age to the incidence of breast cancer?

2. How has this relationship changed over time (in long-term and recent trends)?

3. What can we infer about the true underlying age- incidence pattern from screening studies?

4. What explains the current pattern of incidence rates by age?

First, population data on incidence and mortality from the United States and other Western nations will be reviewed. Second, long-term and recent trends in breast cancer will be briefly examined, with a focus on age-incidence trends. Third, data from screening programs and studies will be discussed, with a focus on the true underlying incidence of the disease, and this will be tied together with recent population data. Fourth, the relationship between various models of disease pro- gression and screening will be discussed.

Mortality and Incidence Data From The United States

The age-adjusted rates of breast cancer mortality have not changed appreciably for more than 40 years, hover-

Relationship Between Age and Incidence/Kessler 1897

-. RATE PER 100,000

600 I

500 r 400

300

200 i I

100 I J

(5 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85. AGE GROUPS

1973 --Ff 198/

Figure 1. Breast cancer incidence by age. Seer data from 1973 and 1987.

ing around 27 per 100,000, using the 1970 United States standard million for age adjustment. Two distinct trends underlie this apparent consistency. Breast cancer mortality has declined significantly from 7.0 per 100,000 in 1973 to 6.2 per 100,000 in 1988 in women younger than 50 years of age. Although it cannot be firmly established, advances in adjuvant therapy are probably responsible for the bulk of this decline. It is also conceivable that a sharp increase in early detection in the early 1970s, corresponding to the wide media attention given to the breast cancers of two prominent American women and the beginning of the National Cancer Institute/American Cancer Society Breast Cancer Detection Demonstration Program,' led to some subsequent mortality declines. This screening increase

RATE PER 100.000 ~~ 120

90

80

may be connected to changes in women younger than 50 years of age, but for older women, there are strong cohort effects.

In contrast, mortality rates among older women have risen from 88.4 per 100,000 in 1973 to 93.0 per 100,000 in 1988, an increase of 5.2%. One clue to this increase in mortality can be found in incidence trends over the past few decades. Figure 23 shows incidence trends over a long period from the Connecticut Tumor Registry (beginning before 1940) and from 1973 with the SEER program. Especially for women older than 50 years of age, incidence and mortality have risen steadily. What is paradoxic about this particular finding is that effective screening technologies for women older than 50 years of age have achieved wide consensus for many years. Nonetheless, it appears that in this group the fight is being lost. But is it? To answer this question, careful analysis is required.

Trends in any disease reflect the following three factors: rate by age, rate by time, and rate by birth cohort. Any point can be uniquely identified by two of these factors. In Figure 3, incidence rates from the Con- necticut Tumor Registry (the oldest registry in the United States) are plotted by age for successive birth cohorts from 1870 through 1960. On this semi-log plot, the evidence of an increase in incidence over a long period is clear. These same data were fit with age-period-cohort models by Holford et aL4 They concluded that although there were statistically significant effects of all three time-related parameters, the age-cohort model fit the data well, suggesting that the effects seen

-1 120

X

5 0 ~ ~ " I I ' " I ' I ' I 1 " I ' I ' I ' I ~ I ' ' 50 1940 1945 1950 1955 1960 1965 1970 1975 1980 1985

YEAR OF DIAGNOSIS

CTR (Observed) SEER (Observed)

- CTR (Modeled) - SEER (Modeled)

Figure 2. Trends in invasive breast cancer among American women. SEER RATES HAVE CONNECTICUT DATA REMOVED

1898 CANCER Supplement April 1, 2992, Volume 69, No. 7

1000 Rate per 100,000 Comparable Data From Other Countries i- t T

~

- I

100

t I

10 ' I 1 I I I 30 40 50 60 70 80

Age at Dx

1870 + 1880 ++ 1890 + 1900

+ 1910 ++ 1920 4 1930 * 1940

-

Source: CT Tumor Registry

Figure 3. Age-specific breast cancer incidence according to birth cohort.

in this figure are due to age and cohort. The large period effect seen in Figure 2 is not evident because this analysis shows age by 10-year birth cohorts and the period effect is short and recent. It is likely that there are (at least) two different underlying trends producing these unusual patterns.

Evidence of those two separate different trends over time are seen in Figure 2. The longer trend is probably related to cohort factors, as analyzed by Holford et d4, but the recent dramatic inflection since 1982 occurs for all cohorts and is much more likely to be a period effect. A period effect suggests something is occurring at a point in time for the entire population. Holford et aL4 could not see such a change because their analysis ended with data from 1984 and the inflection point (k, the year in which these two trends meet), as estimated empirically by Miller et ~ l . , ~ was 1982. Although screening has been suggested as a possible cause for these changes here and in other co~ntries,~- '~ at least one report by White et ul." suggests that screening cannot explain these recent changes in all age groups. White et al." showed that screening effects are consistent with rises in certain age groups, particularly the group from 55 to 64 years of age, but not in very young or old women. The age-incidence curve is a key to understanding these phenomena.

Similar trends in cancer incidence have been examined in detail in the Nordic countries, Iceland, Denmark, and S~eden ' ,~ , '~ , '~ . Generally, these patterns mirror those in the United States, with stable mortality over a long period of time, rising incidence beginning no later than 1960, and, where data are available into the 1980s, a sharper rise in incidence rates.14 There is a slight differ- ence in the pattern in Denmark, where incidence rates did not begin to climb until 1960. Analyses of the Swed- ish and the Iceland data suggested to those researchers that screening was the most likely explanation for the recent rise in incidence rates. A careful examination of the European data with an eye on long-term trends shows that these trends are also consistent with cohort effects in these countries.

Models of Age and Incidence of Breast Cancer

To extract an age curve that represents the true effect of age on breast cancer incidence, Moolgavkar ef all5 compared data from several Western countries and Ja- pan. The age-incidence parameters from these models are shown in Figures 4 and 5. These are not incidence rates, they are parameters from an age-period-cohort model of a series of incidence data. As noticed by Mool- gavkar et al.,14 it is not the exact magnitude of the rates, but the shape of the age-incidence curve that is of para- mount interest. This first plot shows the age-incidence parameters for four populations (Iceland, Denmark, Osaka, and Connecticut). The parameters in Figure 4 are plotted on a linear scale. Clemmesen's hook is evident in all four data sets, although it is slightly more pronounced in the data of the Western countries. Inci- dence rates continue to climb throughout the age range examined; however, in the Japanese data, the climb with increasing age appears almost nil in comparison to Western women. The semi-log plot in Figure 5 contains the same data, but the change in slope from younger than 45 years of age (approximately) to a lesser rate of increase is more obvious. Rates are still increasing, but at a slower rate. As in the registry-based data on which the models were based, there is a sharp increase in the premenopausal period, a slower increase from approximately 50 years of age and older, and the highest rates at older ages. These figures are similar to those from the Connecticut data just shown, which are unadjusted incidence rates. The conclusions of Moolgavkar's analysis are that the extracted age-incidence curves are strik- ingly similar between countries and that many differ- ences across countries and across time within countries and across time within countries can be explained by adjustment for cohort effects or temporal (period) ef-

1899 Relationship Between Age and IncidencelKessler

AGE PARAMETERS 350

300

250

200

150

100

10 20 30 40 50 60 70 80 90

AGE

ICELAND + DENMARK - +++ CONNECTICUT + OSAKA

Figure 4. Age parameters derived from Moolgavkar et al.

fects or both. Regarding age by incidence, the increase in incidence that dominates high rates among older women appears to begm at approximately 50 years of age rather than earlier,

An alternative modeling approach to understand the age-incidence relationship is provided by Manton and Stallard.'6,'7 They estimated the underlying relationship between age and breast cancer within the con- text of a two-disease model applied to mortality data, while trying to infer the underlying incidence pattern. Their first efforts" modeled mortality rates from 1 year. More recent work, especially applied to older women, used the same model but permitted a cohort effect within the model. Briefly, they term the two types of breast cancer as early disease and late disease, and a set of parameters from their recent modeling efforts are plotted to show how the estimated underlying mortality rates change over age (Fig. 6). In this work, early disease refers to the type of breast cancer that is more prevalent among younger women. The rates of early and late disease are estimated from assuming a paramet- ric mixture model of the two diseases. There is no sug- gestion by Manton and Stallard that the data can iden- tify which women have which type of disease. Al- though these rates are predicted mortality rates, the parameterization of their model suggests an underlying incidence function that is theoretically similar to those shown here. It is this incidence function on which the

modeling efforts are based. The bottom line with the boxes as points depicts the so-called early disease, the line with circles depicts late disease. The top line is the sum of the two, representing all breast cancer mortality. As in the models of Moolgavkar and others, Clemme- sen's hook is demonstrated, showing it not to be an artifact of data plotted at any single point in time. In addition, these models show the continuing increase of disease by age.

By relying on registry data, incidence data are derived from women who appear symptomatically in gen- eral, and this is only one version of the true incidence of disease. The models of Manton and Stallard'6,'7 have a similar disadvantage in attempting to uncover the true underlying pattern of disease because women who die of other causes or women who are cured of breast cancer (even if not cured forever, but whose lives are lengthened by treatment long enough so that they die of other diseases) do not appear in their statistics.

Incidence as Estimated by a Screening Program

An alternative approach to uncovering the true age-incidence relationship is the examination of data from screening programs or trials. It was stated previously that incident cases documented in registries are a com- posite of women at all stages of disease and who have had undetected disease for varying lengths of time. For

AGE PARAMETERS 1000

d

0.1 1 1 ~ 1 1

10 20 30 40 50 60 70 80 90

AGE

+ DENMARK ICELAND

CONNECTICUT OSAKA

Figure 5 . Age parameters derived from Moolgavkar et al.


250

c a,

4

0 0 0 150

0 0

$ zoo

r 100

L Q) Q

a, 50 t, a u1

0 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 95 90 95 100

Age Figure 6. Breast cancer mortality derived from Manton and

_.,Early Disease-Late Disease - A 1 I Breast CA Stallard.

example, cases reported at 60 years of age include women who had undiagnosed disease at perhaps 59, 58, or 55 years of age. The conventionally reported incidence rate at 60 years of age represents newly diag- nosed cases at that age. Screening programs may be able to provide a more accurate picture of the growth of disease by first, excluding all prevalent cases and second, by systematically detecting disease growth at subsequent screens.

Comparing age-incidence curves from different studies can be difficult. For example, do cases reported include both in situ and invasive disease? In addition, different age groups may be reported, the modes of early detection may differ, and interval cases may not be included in reported data. The author will try to compare several screening studies with these differ- ences in mind.

In Figure 7, the age-incidence relationship after screening is shown using data from four different sources. The first bars in the graph are from the large Swedish trial (W-E), and the subsequent sets are from Health Insurance Plan of New York Breast Screening Project (HIP), the Breast Cancer Detection Demonstra- tion Project (BCDDP), and the National Breast Screen- ing Study of Canada.5,’8,’9 Despite substantial differ- ences in technology used in the studies mentioned, there is some considerable similarity among the results. The lowest rates are provided by women 40 to 44 years of age, and this increases until sometime between 50 and 60 years of age when there is a leveling of the rates. The only two populations with large numbers older than 69 years of age are W-E and BCDDP, and they suggest a possible decline in new cancers at older ages.

This is in contrast to all the modeling efforts based on mortality data or on incidence registries.

An extension of this crude analysis was undertaken by Gail et ul.,’’ and the results are shown in Figure 8. Invasive only and in situ plus invasive disease are shown, and the patterns are similar. Screening increases dramatically until 50 years of age. After a peak in the 60 to 64 years of age group, there is a modest decrease at older ages. Although BCDDP is not a ran- domized trial and concern has been expressed about the generalizability of the data from that study, these incidence rates should not be affected strongly by any selec- tivity bias. The similarity to the HIP trial and the W-E data suggest that we are looking at the true underlying incidence of the disease. However, it is possible that the interscreening intervals are set sufficiently close together so that slow-growing disease perhaps among certain age groups will not come forth until later, yet exists at subdetectable levels during the screening. This is another way of saying that lead time increases with age, and this has been supported by data from the screening studies of Nijmegen” and the W-E.”

Recent Trends in Incidence

Different views of these recent trends provide insight into the age-incidence relationship. First, the recent trends in breast cancer incidence in contrast to the long- term trend bear discussion. Miller et al. and Kessler et ui.5*6 concluded that this trend was consistent with increases in mammographic screening. Second, the analyses of White et ai.” were consistent with these findings, but there remains some question in certain age groups. In analyses of data from a large US Health Maintenance

Relationship Between Age and Incidence/Kessler

RATE PER 1,000 WOMEN

W-E (SWEDEN) HIP BCDDP

DATA SOURCE

40-44 rn 45-49 1 50-54

0 60-64 65-69 70-74

1901

I NBSS (CANADA)

55-59

Figure 7. Age and incidence screening studies. HIP and BCDDP second screens only

Organization with similiar incidence trends, Glass and Hooverz3 concluded screening was not responsible for that increase. The recent trends in SEER data also can be depicted by examining the dramatic change in the age-incidence curve (Fig. 1). In 1973, the incidence rates were typical of those shown earlier (and modeled by many as previously described), rising steeply, albeit at low levels, until approximately 40 or 45 years of age, then flattening out briefly (Clemmesen's hook), and rising again through the last observed age group. By 1987, a tremendous increase above what had been previously shown typified the data. This large increase began at approximately 55 years of age, peaked at 75 years of age, and then declined.

It is natural to conclude that such a large rise in such a short period cannot be due to changes in the epidemi-

300

I

100

n- " 20-2425-2930-3435-3940-4445-4950-5455-5960-6465-6970-7475-79

AGE GROUP

INVASIVE ONLY INVASIVE AND IN-SITU

Source Gail et a l , JNCI. 1991

Figure 8. Age-specific incidence rates derived from BCDOP data.

ology of disease. A 40% increase in disease incidence in women older than 50 years of age within less than 10 years cannot be easily explained. At least two different analyses of available data pointed to screening playing a role in this increase. In analyzing an increase in breast cancer incidence in metropolitan Atlanta, Liff et al." showed a corresponding increase in the number of mammograms per 100,000 women taking place simulta- neously. From a completely different vantage point, Kessler et d6 used the explosive growth in the number of mammography machines in the United States in a model attempting to explain the increase in breast cancer rates in all of the SEER program. Dawson and Thompsonz4 and surveys in six different communities in the United Statesz5 showed an increase in the use of breast screening using data from the 1987 National Health Interview Survey in comparison with previous estimation of screening described by Howard.26 With all this evidence of the spread of mammography from a variety of sources, screening is a logical explanation for the different age-incidence curves between 1973 and 1987. However, there remains the following dilemma: the largest incidence rate increases are seen in women 60 to 75 years of age, but the limited evidence concern- ing increases in screening appear to be most dramatic among younger women (from 40 to 60 years of age).

This is perplexing, but one explanation may be the interaction of lead time with age. Feuer and Wunz7 suggested that recent trends in screening can explain the bulk of the increases in incidence, if lead time is related to age, as has been suggested by a number of investiga- tors. If lead time is assumed to be 5 years for women


older than 70 years of age and approximately 2 years for women from 40 to 59 years of age, then Feuer and WunZ7 concluded that recent screening for cancer that is modest among older women and larger among younger women can still conspire to provide the current incidence picture. Their calculations showed that the increase in incidence is generally consistent with data on increases in mammography use for women 40 to 59 years of age and older than 60 years of age. They do not claim that all the increase is due to screening; they sub- tract off the long-term trend that still goes largely unex- plained. However, the recent period effect, as shown in Miller et ~ l . , ~ is consistent with a screening effect.

Discussion

Research on the relationship of age to cancer is critically related to the quest for the underlying causes of the disease. Breast cancer has been a challenge for many reasons. A number of risk factors have been shown to affect breast cancer, especially reproductive factors that tend to occur at specific times in the chronologic life cycle of a woman. In addition, studies of screening have proven the efficacy of mammography and physical examination in reducing mortality from breast cancer, which in turn has led to an increase in screening. Na- tional and international changes in risk factors related to breast cancer (such as fertility and diet) and recent dramatic changes in screening have provided a complex picture of the incidence rates of breast cancer, especially the age-incidence relationship.

The age-incidence curve changes around the menopausal period, most likely due to reproductive changes 10 to 15 years earlier such as ~hi ldbir th .~~-~ ' The latent effect of hormonal changes to women in their reproductive period and the high prevalence of surgical in- terventions to end menstruation affect women in their 40s and later. The rapid increase in breast disease between the age of 20 and the mid 40s reflects the increase in hormonal activity. This increase in rates decreases with advancing age.

Another increase in incidence, which often con- fuses the age by incidence picture, is the temporal trend in the disease. As Holford et d4 have demonstrated using United States data (principally from Connecticut) there are notable cohort effects appearing in the 20th century. Although a variety of explanations for these effects have been offered, especially fertility timing changes, no explanations completely suffice for these complex patters. Analysis of Connecticut data by Hahn and Moolgavkar3' suggested that changes in fertility are inconsistent with changes in disease incidence rates and trends.

Although registry -reported incidence of disease dramatically increases with age, data from screening studies challenge the notion that underlying incidence rates continue to increase with age. Available data from a series of screening programs show a stabilizing of incidence rates (detected by screening) after 50 years of age. This is in sharp contrast to models of disease based on registry data. This flattening out of the incidence curve may be in some contrast with the often-cited multistage theory of carcinogenesis based on work by Armi- tage and However, if slower-growing tumors among older persons can be accommodated in the math- ematic and biologic theories of this disease as Manton and StallardI6 have implied, then there may be no dilemma.

Feuer and WunZ7 showed that combining the results of incidence rates from screening studies with notions of lead time is consistent with the current incidence picture, showing the role that screening is playing in the United States. If screening is responsible for these trends, then this may be good news. If these early detection activities are effective in finding tumors earlier than has been the case and if appropriate therapy is delivered, then we can expect decreases in mortality over the next half-decade, at first slowly, and then more rapidly. However, we cannot be assured of such an out- come if screening is not maintained and if both the screening and the follow-up of screen-positive women are not at the highest standards.

References

1. Ries LAG, Hankey BF, Miller BA, Hartman AM, Edwards BK. Cancer Statistics Review 1973-1988. National Cancer Institute.

2. Baker LH. Breast cancer detection demonstration project: Five year summary report. C A 1982; 32:194-225.

3. Miller BA, Feuer EJ, Hankey BF. The increasing incidence of breast cancer since 1982: Of early detection. Cancer Causes and Control 1991; 2:67-74.

4. Holford TR, Roush GC, McKay LA. Trends in female breast cancer in Connecticut and the United States. Clin Epidemioi

5. Miller BA, Baines CJ, To T, Wall C. The Canadian National Breast Screening Study. In: Miller AB, Chamberlain J, Day NE, Hakama M, Prorok PC, eds. Cancer Screening. New York: Cam- bridge University Press, 1991; 45-55.

6. Kessler LG, Feuer EJ, Brown ML. Projections of the breast cancer burden to U.S. women: 1990-2000. Prev Med 1991; 20:l-13.

7. Forrest PA. Screening and breast cancer incidence. ] Natl Cancer rnst 1990; 82:1525-1526.

8. Ranstam J, Janzon L, Olsson, H. Rising incidence of breast cancer among young women in Sweden. Br ] Cancer 1990;

9. Sigurdsson K, Adalsteinsson S, Ragnarsson J. Trends in cervical and breast cancer in Iceland: A statistical evaluation of trends in incidence and mortality for the period 1955-1989, their relation

NlH Pub. NO. 91-2789,1991; 11. 43.

1991; 44:29-39.

61 : 120-122.

Relationship Between Age and Incidence/Kessler 1903

10.

11.

12.

13.

14.

15.

16.

17.

18

19

20

21

to screening and prediction to the year 2000. Znt ] Cancer 1991;

Liff JM, Sung JFC, Chow WH, Greenbery RS, Flanders WD. Does increased detection account for the rising incidence of breast cancer? A m ]Pub l i c Health 1991; 81:462-465. White E, Lee CY, Kristal AR. An evaluation of the increase in breast cancer incidence in relation to mammography use. ] NatL Cancer Inst 1990; 82:1546-1552. Caygill CPJ, Hill MJ. Trends in European breast cancer incidence and possible etiology. Tumori 1991; 77:126-129. Ewertz M, Cartensen, B. Trends in breast cancer incidence and mortality in Denmark, 1943-1982. Int ] Cancer 1988; 41:46-51. Davis DL, Hoe1 D, Fox J, Lopez, A. Epidemiology: International trends in cancer mortality in France, West Germany, Italy, Ja- pan, England and Wales, and the USA. Lancet 1990; 336:474- 481. Moolgavkar SH, Stevens RG, Lee JAH. Effect of age on incidence of breast cancer in females. ] Nut i Cancer Znst 1979; 62

Manton KG, Stallard E. A two-disease model of female breast cancer: Mortality in 1969 among white females in the United States. ] Nut1 Cancer Znst 1980; 64:9-16. Manton KG, Stallard E. Demographics of breast cancer in older women. Presented at the Forum for Breast Cancer Screening, 1990. Sturbridge, MA, July 31. Fagerberg G, Baldetorp L, Grontoft 0, Lundstrom B, Manson JC, Nordenskjold 8. Effects of repeated mammographic screening on breast cancer stage distribution. Oncology 1985; 24:465-473. Smart f. The role of mammography in the prevention of mortality from breast cancer. In: DeVita VT, Hellman S, Rosenberg S, eds. Cancer Prevention. Philadelphia: JB Lippincott, 1989. Gail MH, Brinton LA, Byar DP et al. Projecting individualized probabilities of developing breast cancer for white females who are being examined annually. ] Nat l Cancer Znst 1989; 81:l-8. Peeters PHM, Verbeek ALM, Hendriks JHCL, van Bon, MJM. Screening for breast cancer in Nijmegen: Report of 6 screening rounds, 197551986, Znt ] Cancer 1989; 43226-230.

48~523-528.

493-501.

22.

23.

24.

25.

26.

27.

28.

29.

30.

31.

32.

33.

Tabar L, Faberberg G, Day NE, Holmberg, L. What is the opti- mum interval between mammographic screening examina- tions?: An Analysis based on the latest results of the Swedish two-county breast cancer screening trial. Br ] Cancer 1987;

Glass AG, Hoover RN. Rising incidence of breast cancer: Rela- tionship to stage and receptor status. ] Nat l Cancer Inst 1990;

Dawson DA, Thompson GB. Breast cancer risk factors and screening; United States, 1987. National Center for Health Sta- tistics. Vital Health Statistics 1990; 172(10):22. National Cancer Institute (NCI) Breast Cancer Screening Con- sortium. Mammography: A missed clinical opportunity? Results of the National Cancer Institute (NCI) Breast Cancer Screening Consortium and National Health Interview Survey Studies.

Howard J. Using mammography for cancer control: An unreal- ized potential. C A 1987; 3733-48. Feuer EJ, Wun LM. How much of the recent rise in breast cancer incidence can be explained by increases in mammography utili- zation: A dynamic population model approach. ] Natl Cancer Znst (submitted for review). Brinton LA. Menopause and the risk of breast cancer. Ann N Y Acad Sci 1990; 592:357-362. Brinton LA, Hoover R, Fraumeni JF. Reproductive factors in the aetiology of breast cancer. Br ] Cancer 1983; 47:757-762. Brinton LA, Schairer C, Hoover RN, Fraumeni JF. Menstrual factors and risk of breast cancer. Cancer Invest 1988; 6:245-254. Hahn RA, Moolgavkar SH. Nuiliparity, decade of first birth, and breast cancer in Connecticut cohorts 1855-1945: An ecological study. A m 1 Public Health 1989; 79:1503-1507. Armitage P, Doll R. The age distribution of cancer and multistage theory of carcinogenesis. BY ] Cancer 1954; 8:l-12. Moolgavkar SH. The multistage theory of carcinogenesis and the age distribution of cancer in man. ] Natl Cancer Znst 1978;

55:547-551.

82~693-696.

]AMA 1990; 264:54-58.

61149-52.

the relationship between age and incidence of breast cancer population and screening program data

Documents