Of those diagnosed with breast cancer during 1973–2000, 40 068 women developed one additional cancer after breast cancer, 3796 women developed two cancers, 351 women developed three cancers, 31 women developed four cancers, and 3 women developed 5 cancers. Of these women, 0.76% were diagnosed by age 29, 6.79% between 30 and 39, 19.72% between 40 and 49, 21.19% between 50 and 59, 22.34% between 60 and 69, and 29.20% at or after age 70. Second cancers were subsequently reported for 11.7% of the patients who were diagnosed with breast cancer before age 50, and 17% of patients diagnosed with breast cancer at or after age 50. Among women with a second tumour diagnosed, the mean time until the second cancer occurred was 6.2 years: 8 years for patients less than 50-years old and 5.7 years for women at or greater than age 50. Among women with a single breast cancer diagnosis, the average time from diagnosis of the breast cancer until death was 6.71 years for women less than 50-years old, and 4.5 years for women at or greater than age 50. 60.28% of the women were still alive at the end of follow-up.

The results of the standardised incidence ratio analysis are found in Table 1. For women diagnosed with breast cancer by age 29, the overall second primary tumour risk is large (SIR=17.3, 95% CI=13.7, 21.6), as is the risk of developing a second breast cancer (SIR=478.5) or stomach cancer (SIR=145.4). As age at primary diagnosis increases, the overall SIR's decrease until the eldest age group. Cancer sites with the highest ratios for women diagnosed with breast cancer before age 50 include bone, breast, leukaemia, lung, and ovary. However, having a breast cancer diagnosis was protective against cancer of the cervix for these age groups. For women who did not develop breast cancer until at least 50 years old, the overall SIR for developing a second cancer by age 59 was 2.7, which is much lower than risks among some of the earlier age groups. Again, the overall ratios decrease with age. Cancer sites with the highest SIR's were breast, colon/rectum, corpus uteri, lung, ovary, skin, and thyroid. Among older women, an initial breast cancer diagnosis was not protective against cervical cancer. Overall, both excess risk and protective effects decreased with age for most of the 14 cancers examined. For example, the SIR for a second breast cancer was greatest for women 20–29 years of age (478.53) and decreased dramatically and consistently after age 30 (Table 1).

Table 2 presents the data comparing the SIR's of a second breast cancer diagnosis with a second non-breast cancer diagnosis within 10-year time intervals. For women diagnosed with breast cancer by age 29, their risk of developing a second primary tumour other than breast within 10 years of their initial diagnosis is large (SIR=8.9, 95% CI=8.27, 9.60), but their risk for developing a second breast cancer, within the same 10 years, is five and a half times greater (SIR=48.4, 95% CI=46.5, 50.3). But after 20 years, these same women actually showed a protective effect against both cancer groups, breast and other, compared to the general population (SIR=0.2, 95% CI=0.2, 0.3; SIR=0.5, 95% CI=0.4, 0.6). For women diagnosed with breast cancer at or after age 70, their risk for developing either a second breast cancer or a second primary cancer, other than breast within 10 years, is virtually the same (SIR=1.84, 95% CI=1.8, 1.9; and SIR=1.6, 95% CI=1.5, 1.7). In general, most women had a more elevated risk of developing a second breast cancer, than a second primary cancer other than breast, but both risks were greater for these women than the general population. The risk reduced with time and age for all groups.

Table 3 summarises data from Kaplan–Meier Curves for time until second primary cancer diagnosis, stratified on age at diagnosis of the breast cancer and limited to the 12.3% of women who were diagnosed with a second primary tumour. The median years of follow-up until the second cancer is diagnosed ranged from 2.3 years for women over 70 developing cervix uteri to 18.08 for women 30–39 developing a second breast cancer. The difference in follow-up time for the age groups can be attributed to (1) a longer at-risk period for women <50 years for developing the second cancer and (2) the increased risk of developing some cancers at an advanced age.

Figure 1 illustrates the median ages at the time of the second cancer diagnosis for each of the 14 cancers. Cervix uteri cancer has the youngest median age with 59 years, while colon/rectum cancer has the oldest median age with 75 years.

Survival analysis was then constructed stratifying on the women who only developed breast cancer and those women who were diagnosed with breast cancer and a second primary cancer. Survival curves were adjusted for race and grade. In the 20–29 year old category, grade was categorised as an ordinal variable. Figure 2 illustrates survival analysis for women 20–29 years old diagnosed with breast cancer in or after 1988, stratified on multiple tumour status, controlling for race (P=0.621) and grade (P=0.511). The figure shows women with multiple primary cancers had poorer survival than women with only a breast cancer diagnosis over time. Women in this young age group who had metastatic breast cancer have a 15-year survival probability of only 23.23%.

In general, overall survival was poorer for women with multiple primary tumours compared with women with no second tumours. In addition to second tumours, grade of the breast cancer, race, and extent of disease played a role in determining the survival for each age group. A woman with a higher-grade breast tumour followed by a second tumour had much poorer survival than a woman with a low-grade breast tumour followed by a second tumour. This is also true for the extent of disease. Time also acted as a contributing factor to poorer survival.

Changes in both the breast cancer incidence and the mortality rates have been documented in the past 25 years (Young et al, 2001). Evidence shows a fundamental increase in incidence, which is partly attributed to the improved screening and diagnostic techniques as well as to the changes in reproductive patterns, that is, women waiting until later to have children. Mortality rates had been relatively constant from 1973 to 1990. Since 1990, mortality rates have decreased by 2.3% overall. This decrease was more pronounced in women diagnosed with breast cancer before age 50 than with women aged 50 or older and have been attributed to the improvement of the various cancer treatments (Ries et al, 2003). Recent data have shown an overall survival rate for women diagnosed with breast cancer at least 5 years before to be as high as 87% but drops to 77% for the 10-year mark and to 52% by the 20-year mark (Ries et al, 2003). Given the increase in incidence and survival, it is important to assess health risks for breast cancer survivors.

Using a standard method, population-based data were collected from high-quality registries that covered approximately 11% of the US population. The study provided a unique opportunity to study a large number of women, over 335 000, across several years (1973–2000). This is the largest and most comprehensive study to date of cancer among breast cancer survivors. An international study (Chu et al, 1999) that examined the risk of a second primary tumour following an initial breast cancer diagnosis, only assessed risk for two age groups: <50 and ⩾50 years old; this also used the International Association for Research on Cancer definition of a second primary cancer, which is not widely used in the US. Another study examined women developing sarcoma after breast cancer, using the Surveillance, Epidemiology, and End Results (SEER) program data from 1973 to 1997 (Yap et al, 2002). The major drawback in this study was that it included only patients with invasive breast cancer.

This study found that 12.3% of breast cancer survivors were diagnosed with second primary malignant tumours among breast cancer survivors. Excess risks for 13 specific cancers were found, with the highest excess risks for breast. Among women with a second primary tumour, the time to second diagnosis varied by type of tumour and age at diagnosis of the first tumour. The results in this study are comparable to other findings regarding second primary breast cancer (Berstein et al, 2003). As expected, the risk of several cancers, including lung cancer, increased with age.

These results may help to inform patients and clinicians regarding expected quality of life for breast cancer survivors (Cimprich et al, 2002). A greater understanding of the patient's prognosis would assist both the patients and the clinician in treatment options. For example, if a woman understands her risks for developing a second primary cancer, she may choose to alter her treatment to be more aggressive in hopes of minimizing her risk. She may also need to be counselled regarding preventable risk factors, such as weight gain, that have been associated with premature mortality for women with a second breast cancer in other studies (Bernstein et al, 2002).

In this study, breast cancer survivors had the greatest risk for developing breast, bone, colon/rectal, connective tissue (sarcoma), leukaemia, lung, ovarian, and thyroid cancer. There are several possible explanations for these correlations, including genetic mutations. Four possible genetic mutations have been identified which can aid in predicting which women with breast cancer will develop multiple primary tumours. All four of these genetic mutations include breast cancer and other cancers. BRCA1 and BRCA2 are considered to be the breast cancer genes, but BRCA1 is also associated with ovarian cancer (Szabo and King, 1997). BRCA2 is associated with multiple primary tumours including breast, colorectal, ovarian and other cancers. Ovarian cancer, like breast cancer, is an oestrogen-driven cancer. Thus oestrogen is the most obvious link between the two cancers. However, the risk of ovarian cancer is also increased with the presence of the BRCA1 gene mutation (Szabo and King, 1997). Both BRCA1 and BRCA2 may bestow a heightened sensitivity to carcinogenic effects of radiation (Ford et al, 1994; Turner et al, 1999). Recently studies have shown a possible association between lung cancer and BRCA1 and BRCA2 mutation (Bergfeldt et al, 2000).

Multiple primary cancers, including acute adult leukaemia, have been associated with a family history of breast cancer (Rauscher et al, 2002). P53, known as Li–Fraumeni Syndrome, is associated with the development of cancer at a young age and is associated with breast cancer, sarcoma, brain cancer, and leukaemia (Lee, 1980). Women with P53 mutation have a lifetime probability of almost 100% for developing at least one of these cancers. Lastly the genetic mutation PTEN or Cowden Disease increases the risk for both breast and thyroid cancer.

Breast cancer survivors appeared to have protection from developing cervical cancer. Low socioeconomic status is a risk factor for cervical cancer while high socioeconomic status is a risk factor for breast cancer. However, the causal relationship underlying these associations is not well understood.

Breast cancer was associated with increased risk of connective tissue cancer for several of the age groups. Two possible explanations are (1) many women with breast cancer undergo radiation, which increases a woman's risk for connective tissue cancer in the surrounding areas and (2) some genetic mutations, including Li–Fraumeni Syndrome are linked with both breast cancer and connective tissue cancer (Malkin et al, 1990; Rubino et al, 2003).

Radiation may also increase the risk of lung cancer (Matesich and Shapiro, 2003; Roychoudhuri et al, 2004). During the 1950s, most of the upper body was exposed to radiation as the treatment for breast cancer. This exposure included the thyroid, lungs, breasts, and chest wall. This excess risk should therefore decrease as the cohort exposed to high doses of radiation doses ages. As well as radiation, chemotherapy as a treatment option for breast cancer may be a risk factor for developing leukaemia as a second primary cancer.

Among women who developed a second tumour, the time to cancer varied by age. Risk for cancer generally increases with age. This is most evident for the ‘older age' cancers such as colorectal, in which median years to second diagnosis decreased from 18 for the 20–29 year group to less than four for the 70+ year group. Other age-dependent cancers were ovarian (median years ranging from 11 to less than 4) and pancreas (median years ranging from 15 to about 4). In the past, clinicians used as a general rule that if a woman is cancer free for 5 years after the initial diagnosis, then she is cured of cancer. This analysis shows that a sizeable proportion of women who developed a second tumour were cancer-free 5 years after their initial diagnosis. With the exception of the oldest group, the median time to diagnosis exceeded 5 years for the women who had a second tumour. For women who were diagnosed with breast cancer between the ages of 30 and 39, the median number of years until they were diagnosed with any second tumour was 11.42 years. This long lapse for the 30–39 year old age group is primarily due to the fact that the median number of years until a second breast cancer diagnosis was 18.08 years. It appears that the women in this age group are treated very aggressively and tend to have mastectomies performed on them. The only second cancer where the median number of years was less than 5 was cancer of the cervix uteri, but this estimate is probably unrealistic because of the small number of women who developed cancer of the cervix uteri after breast cancer. On the other hand, connective tissue cancer had a median number of years until diagnosis of over 10 years, except for the 70+ age group.

This study has several limitations. Although population-based cancer registries provide the opportunity to study these rare events, the registries lack some important variables such as smoking status, specific doses of treatment, and genetic make-up. Also, quality of data used for classifying multiple primary tumours is not under the control of the registries and may vary across site and over time. Even studies that used data sets known to be accurate found errors that affected the interpretation (Bergfeldt et al, 2000). Finally, missing and incomplete data may introduce measurement error. In this study, 5.7% of women known to have been diagnosed with a second tumour were excluded from the incidence risk analysis because of unknown diagnosis. If the distribution of the tumour sites in these women is substantially different from the distribution of known tumour sites, results for some of the more rare tumours may be biased. However, the number of unknown tumours is unlikely to have affected the risks from the more frequent tumours. Whatever bias, if present, is likely to be small, as there are striking similarities in results from published studies of selected second tumours (Yap et al, 2002).

In conclusion, it is clear that some women who survive breast cancer are at increased risk of additional cancers. Excess risks attenuate with increasing age at first diagnosis and with increasing years since first diagnosis. Excess risk also varies by cancer type, consistent with existing causal hypotheses regarding genetics mutations and risk of aggressive radiation therapy. More research should focus on isolating the factors that may predispose a breast cancer survivor to the development of subsequent cancers.