Multiple primary cancers of the colon, breast and skin (melanoma) as models for polygenic cancers

Authors


Abstract

To assess the role of family history in the development of multiple primary cancer, the Swedish Family-Cancer Database was used to analyze second primary cancer in patients born in 1935 to 1996 with an initial primary cancer of the colon, breast and skin (melanoma) by familial cancer in first-degree relatives. Standardized incidence ratios (SIRs) were calculated from site-, sex- and age-specific rates for all persons (offspring) born in 1935 to 1996. Familial risk (SIR) was calculated for the first and second primary cancers in offspring. A Poisson regression analysis was also performed to assess the risk factors for occurrence of second primary cancer. The familial proportion of multiple primary cancers was 29.0% (9/31) for colon, 16.3% (122/747) for female breast and 14.5% (17/117) for melanoma. Compared with all offspring, patients with family history were at a much higher and significantly increased risk for subsequent primary cancer at colon (SIR = 59.1), skin (SIR = 48.2) and female breast (SIR = 7.9). The corresponding SIRs in patients without family history were 13.8, 10.5 and 5.2 at the three sites. The ratios for incidence of second primary to first primary were highest when diagnosis age was less than 40 years. A Poisson regression analysis showed that family history was one of the major risk factors for occurrence of multiple primary cancers at colon, breast and skin. The high risk of second cancer, even in the absence of family history, would be consistent with a polygenic model of carcinogenesis. © 2001 Wiley-Liss, Inc.

Experimental models and age-incidence relationships in human cancers suggest that most cancers are multistage diseases.1–8 Genes may interact in an unordered or ordered fashion along a polygenic pathway: A→B→C→D→E, and the stages may involve the two alleles of the gene, such as A1 and A2. Cancers almost always have many causes and therefore they are heterogeneous or complex. Thus, in addition to the above pathway, entirely independent or partially intercepting pathways may operate. However, the support for multistage carcinogenesis in humans in vivo is fragmentary. Almost all the known cancer syndromes, including retinoblastoma, Wilms' tumor, Li-Fraumeni syndrome, von Hippel-Landau syndrome, neurofibromatosis 1 and 2, multiple endocrine neoplasia 1 and 2, familial adenomatous polyposis (FAP), hereditary nonpolyposis colorectal cancer (HNPCC) and early-onset breast cancer (BRCA1 and BRCA2) are monogenic, and they conform to a two-stage model in requiring inactivation of the two copies of the tumor suppressor gene.9–11

In the case of monogenic cancers, one defective gene appears to be necessary for complete carcinogenesis. In polygenic cancers, the individual mutant genes are not sufficient to cause cancers, and thus familial risks are only weak because the probability of inheriting exactly the same set of genes diminishes with the number of such genes.12, 13 However, those individuals who have inherited a harmful set of genes would be at high risk for second cancers. The progress in understanding polygenic cancers has been hampered by the lack of model systems for study.

In this nationwide population-based study, we used the Swedish Family-Cancer Database to analyze the risks of developing a second primary cancer at the same site in patients (affected offspring), born after 1934 and diagnosed from 1958 to 1996, with an initial primary cancer of the colon, breast and skin (melanoma) by family history in first-degree relatives. Our previous study on contralateral breast cancer examined risk factors among all women in the Database.14 In the present study, we also calculated familial risk for the first primary cancer at the three sites in offspring based on the same reference rate in all offspring. We hypothesized that if polygenic effects were important in these cancers, the risks of second cancer should exceed those of the first cancer.

SUBJECTS AND METHODS

The Swedish Family-Cancer Database, updated in 1999, was formed from the Second Generation Register maintained by Statistics Sweden and linked by the individually unique national registration number to the Swedish Cancer Register at the National Board of Health and Welfare. The Database includes all persons (offspring) born in Sweden after 1934 with their biological parents, totaling 6,425,681 offspring (3,295,282 males and 3,130,399 females).15, 16 Since 1958, all new cases of cancer in Sweden have been reported to the Swedish Cancer Register by compulsory reporting from clinicians who diagnose a neoplasm and the pathologists/cytologists, who must report separately any diagnosis of cancer made on pathological and cytological specimens. The Family Cancer Database has a gap among those born between 1935 and 1940 who died between 1960 and 1997. Many of these individuals lack links to parents in the Database, and this probably causes a deficit of some cancers and somewhat inflated risk estimates for fatal cancers. The site of cancer is registered based on a four-digit diagnostic code according to the 7th revision of the International Classification of Diseases (ICD-7).

All offspring with an initial primary cancer of the colon, breast (female) and skin (melanoma) diagnosed between January 1958 and December 1996 were retrieved from the Database. The patients were scored for a second primary cancer at the same site. Cases of second cancers were extracted from the Database if the diagnosis date of the first and second cancers differed by at least 1 month. Also, topology, i.e., the organ site, had to be different (other section of colon, contralateral breast or different part of body for melanoma) from the first cancer, as specified by the four-digit ICD code. Separately, data are given for those second cancers that were diagnosed ≥1 year after the first cancer. Familial cases were defined as the patients with an affected parent or affected sib with the same type of cancer.

The standardized incidence ratio (SIR) was used to estimate the risk for a new primary cancer of the colon and breast and melanoma in offspring. SIRs were calculated by dividing observed cases by expected cases based on site-, sex- and age-specific incidence rate of first primary cancer at the three sites in all offspring. For first primary cancer, person-years at risk were accumulated for each offspring beginning with the date of birth or January 1, 1958, whichever came last, and ending with the date of diagnosis of a first primary cancer, date of death, date of emigration or December 31, 1996, whichever came first. For second primary cancer, person-years at risk were accumulated for each patient beginning with the date of diagnosis of the first primary cancer and ending with the date of diagnosis of a second primary cancer, date of death, date of emigration or December 31, 1996, whichever came first. The expected number of second cancers was obtained by assuming that these persons experienced the same cancer incidence as prevailed in all offspring. The 95% confidence intervals (95% CIs) for SIR and incidence ratio were calculated assuming that the cases followed a Poisson distribution.17 A Poisson regression analysis was performed by considering a number of explanatory variables in the Genmod procedure of the SAS program.

RESULTS

Table I reports the numbers and familial proportions of first and subsequent primary cancers at colon, breast and skin (melanoma). Among first primary cancers, the proportion of familial cases was 7.6% for colon, 11.4% for female breast and 3.8% for melanoma, whereas among multiple primaries, the familial proportion was 29.0% for colon, 16.3% for female breast and 14.5% for melanoma. The proportion was 40.0% for colon, 17.3% for breast and 16.3% for melanoma among those whose second diagnosis was ≥1 year after the first cancer.

Table I. Number and Familial Proportion of First and Second Primary Cancers
Cancer siteFirst primary cancerSecond primary cancer1
Total casesFamilial cases2Familial proportionTotal casesFamilial casesFamilial proportion
  • 1

    Subsequent cancers were included if first and second diagnoses differed by ≥1 month and topology. Figures in parentheses denote second cancers diagnosed ≥1 year after the first cancers.

  • 2

    Patients had a parent or sib affected with the same type cancer.

  • 3

    Female breast cancer.

Colon3,2472477.6%31 (15)9 (6)29.0% (40.0%)
Breast317,4871,99711.4%747 (560)122 (97)16.3% (17.3%)
Melanoma80753033.8%117 (98)17 (16)14.5% (16.3%)

Table II presents SIRs for familial first primary cancer in offspring at the three sites and SIRs for subsequent primaries by family history. All SIRs were calculated based on the same reference rate, the site-, sex- and age-specific incidence for all offspring. To estimate the effect of familial factors on the occurrence of second cancer, we also calculated the ratio (C/B ratio) of SIR for second cancer with family history to SIR for second cancer without family history. At all three sites, the patients, even without affected first-degree relatives, were at much higher risk for the occurrence of a second primary at the particular site than those with an affected first-degree relative for an initial primary cancer. However, the C/B ratios were not much different from the SIRs for familial cancer (first cancer). When ≥1 year was allowed between the first and the second cancer, the SIRs were somewhat decreased; the largest decrease was for sporadic colon cancer.

Table II. Risks for First Primary Familial Cancer and For Familial and Sporadic Secondary Cancers1
Cancer siteFirst primary cancerFollow-up timeSecond primary cancerRatio (C/B)(95% CI)
Familial (A)Sporadic (B)Familial (C)
NSIR(95% CI)NSIR(95% CI)NSIR(95% CI)
  • 1

    All standardized incidence ratios (SIRs) were calculated based on site-, sex- and age-specific incidence of the first primary cancer in all offspring. 95% CI, 95% confidence interval.

Colon2472.2(1.9–2.5)≥1 mo2213.8(8.7–21.0)959.1(26.8–113)4.3(2.0–9.3)
≥1 yr96.1(2.8–11.6)642.3(15.2–92.8)7.0(2.5–19.6)
Breast1,9971.9(1.8–2.0)≥1 mo6255.2(4.8–5.6)1227.9(6.5–9.4)1.5(1.2–1.8)
≥1 yr4634.2(3.8–4.6)976.8(5.5–8.3)1.6(1.3–2.0)
Melanoma3032.8(2.5–3.1)≥1 mo10010.5(8.6–12.8)1748.2(28.0–77.4)4.6(2.7–7.6)
≥1 yr829.2(7.3–11.4)1648.1(27.4–78.2)5.2(3.1–9.0)

Table III shows SIRs of second events after an initial primary cancer at the three sites by family history and follow-up interval. The patients were at significantly increased risk of developing a subsequent primary at the three sites through all follow-up periods. The occurrence of second events showed the highest SIR within the first year of follow-up regardless of family history. The patients with a family history were at a higher risk of developing second cancer than those without a family history in all follow-up periods.

Table III. Risk for Second Primary Cancer by Family History and Follow-up Time1
Second cancer siteFollow-up timeSporadicFamilial
NSIR(95% CI)NSIR(95% CI)
  • 1

    All SIRs were calculated based on site-, sex- and age-specific incidence of the first primary cancer in all offspring.

Colon1–11 mo13117(61.9–200)3287(54.0–848)
1–9 yr44.2(1.1–11.0)330.3(5.7–89.6)
10–38 yr59.3(2.9–22.0)370.3(13.2–208)
Breast1–11 mo16216.4(14.0–19.2)2520.5(13.2–30.2)
1–9 yr3974.5(4.1–4.9)796.8(5.4–8.5)
10–38 yr663.3(2.6–4.3)187.4(4.4–11.7)
Melanoma1–11 mo1833.8(20.0–53.5)151.1(0.0–293)
1–9 yr6611.1(8.6–14.1)1147.6(23.6–85.5)
10–38 yr165.9(3.3–9.5)553.8(17.0–126)

Table IV and Figures 1 to 3 show the age-specific incidence of first and second primary cancers and relative risk (RR) by age group. For all three sites, the ratios (B/A) for incidence of second primary to first primary were highest when the diagnosis age was under 40 years and decreased with increasing diagnosis age (from 44.1 to 17.6 for colon, from 68.6 to 3.0 for breast and from 36.8 to 8.0 for melanoma). The ratios (D/C) for incidence of familial second primary to sporadic second primary were highest at 40 to 44 years for colon (10.7; Fig. 1, using all second colon cancers instead of familial second colon cancers because of the few familial second colon cancers), at 25 to 29 years for breast (7.6; Fig. 2) and at 30 to 34 years for melanoma (25.5; Fig. 3).

Table IV. Incidence Rates and Relative Risks of Colon and Breast Cancer and Melanoma
Cancer siteAge at diagnosis (yr)Incidence rate (cases/100,000 person-years)RR (B/A)(95% CI)Incidence rate (cases/100,000 person-years)RR (D/C)(95% CI)
First primary cancer, all (A)Second primary cancer, all (B)Second primary cancer, sporadic (C)Second primary cancer, familial (D)
Colon<402.089.344.1(20.8–64.5)93.80.00.0
40–496.6160.024.1(13.9–31.8)121.5556.34.6(1.4–14.9)
50–6115.6275.117.6(10.7–22.7)209.4889.74.2(1.5–12.2)
Breast<4012.8875.368.6(57.4–75.0)794.91416.61.8(1.2–2.7)
40–49110.3874.67.9(7.2–8.3)825.51238.01.5(1.2–2.0)
50–61197.4588.43.0(2.6–3.2)577.0679.51.2(0.8–1.7)
Melanoma<405.0185.136.8(27.6–42.5)144.91116.37.7(4.0–14.8)
40–4916.9176.110.4(7.7–12.2)174.2229.61.3(0.3–5.5)
50–6123.5188.38.0(5.4–9.8)172.5630.33.7(1.1–12.2)
Figure 1.

Age-specific incidence of first primary colon cancer in offspring and all second primary colon cancers and sporadic second primary colon cancers in colon cancer patients.

Figure 2.

Age-specific incidence of first primary breast cancer in female offspring and familial second primary breast cancers and sporadic second primary breast cancers in female breast cancer patients.

Figure 3.

Age-specific incidence of first primary melanoma in offspring and familial second primary melanomas and sporadic second primary melanomas in melanoma patients.

Table V gives the explanatory variables and the RRs for second primary cancer at the three sites in respect to the reference category (in bold face) based on a Poisson regression model. Residence and occupation showed no effect on the occurrence of second primaries at any of the three sites. Sex showed a borderline significant effect on second colon cancer (RR = 2.0; 95% CI = 1.0 to 4.2) and second melanoma (RR = 1.4; 95% CI = 1.0 to 2.0). Although age and period of first diagnosis showed no effect on second colon cancer and melanoma, for second breast cancer, there was a systematic decrease in RR from early to late age of onset and an increase in RR from period before 1975 to the period after 1985. For all three sites, family history and follow-up time showed a significant effect on the occurrence of second primaries, but, compared with 1 to 9 years of follow-up, 10 to 38 years of follow-up showed a significant increased risk only for colon cancer.

Table V. Relative Risks for Second Primary Cancers of The Colon, Breast and Skin (Melanoma) by Poisson Regression Model1
VariableValueColon cancerFemale breast cancerMelanoma
NRR(95% CI)NRR(95% CI)NRR(95% CI)
  • 1

    Bold face shows the reference category.

SexMale192.0(1.0–4.2)581.4(1.0–2.0)
Female121.0747591.0
ResidenceOther70.6(0.3–1.4)2591.0(0.9–1.1)481.1(0.8–1.6)
Metropolitan241.04881.0691.0
OccupationFarmer61.6(0.6–4.3)981.2(0.9–1.6)110.9(0.5–1.8)
Worker101.1(0.4–2.3)2181.1(0.8–1.2)371.2(0.8–1.9)
Other20.4(0.1–2.0)1011.1(0.9–1.4)181.1(0.6–1.9)
Professional131.03301.0511.0
Age at first diagnosis (yr)<40131.02781.0731.0
40–4981.1(0.4–3.0)3610.7(0.6–0.8)290.8(0.5–1.2)
50–61103.1(0.9–10.0)1080.5(0.4–0.6)151.4(0.8–2.6)
Period of first diagnosis1958–197420.3(0.1–2.04)220.4(0.2–0.6)150.6(0.3–1.2)
1975–1984121.7(0.6–5.21)1970.7(0.5–0.8)350.7(0.4–1.1)
1985–1996171.05281.0671.0
Year after first diagnosis<11618.8(7.6–46.7)1873.4(2.9–4.1)1925(1.5–4.1)
1–971.04761.0771.0
10–3883.7(1.1–11.5)841.0(0.8–1.3)210.8(0.5–1.4)
Family historyYes95.0(2.3–11.0)6251.5(1.2–1.8)174.2(2.5–7.0)
No221.01221.01001.0

DISCUSSION

The increased occurrence of second primary cancers after an initial primary could result from 1) intensive medical surveillance after first diagnosis, 2) therapy-induced exposure to X-rays and carcinogens,18 and 3) shared environmental, hereditary and immunological factors between the first and the second cancers. As 98% of new primary cancers are verified histopatholocally or cytologically in the Swedish Cancer Registry, it is unlikely that intensive medical surveillance is causing diagnostic misclassification. A study has also confirmed that the Swedish Cancer Registry is a reliable source for studying multiple malignant tumors.19 However, the diagnosis of second cancer may be arrived at earlier, causing an increase in incidence during the first year of follow-up and a deficit later. In the present study, we included the patients whose first and second diagnoses differed by ≥1 month and by anatomic site, ensuring that they were new primaries. We also did a separate analysis in the patients with ≥1 year of follow-up. As the results agree, we presented most data based on the patients with ≥1 month of follow-up and different topology between the first and second diagnoses.

A great number of studies have shown that there is a hereditary component in colon and breast cancer and melanoma, and several genes and familial cancer syndromes have been identified for these cancers.10, 16, 20–29 However, very few population-based studies have been conducted to assess the role of family history in the development of multiple primary cancers at these sites. Most of the studies have been carried out in affected families that do show multiple primaries, including in colon cancer carriers of APC and mismatch repair gene mutations,10, 26 in breast cancer BRCA1- and BRCA2-related families30–33 and in melanoma P16INK4a mutation carriers.34–37 We have recently described familial and other effects on contralateral breast cancer in all women born between 1900 and 1959 from this Database.14

The present data on offspring showed that at all three sites, the familial proportion of patients with multiple primaries was much larger than that among all patients (Table I). The difference was 3.8-fold for color cancer and melanoma but only 1.4-fold for breast cancer (from 11.4% to 16.3%). However, it is noteworthy that by far the largest proportion of second cancer patients lacked family history. Patients with family history were at a much higher risk for the second primary cancer through all follow-up periods than those without family history (Tables II, III). Moreover, family history showed a significant effect on the occurrence of second primaries based on the Poisson regression analysis after adjustment for sex, residence, occupation, age and period of first diagnosis and follow-up intervals (Table V). These data suggest that the familial factor ranked as an important contributor to the development of multiple primary cancers at these sites.

Age-specific incidence ratio of first and second cancers revealed that compared with all offspring, cancer patients with earlier age of onset ran a higher relative risk for occurrence of primary cancer at the three sites. Also, the ratio of familial and sporadic second primary cancers suggested that family history had an age-specific effect on occurrence of multiple primaries (Table IV, Figs. 1–3). It is noteworthy that all the subjects in the present study were relatively young (aged 0 to 61 years), partially explaining the strong familial effects observed.

Interestingly, compared with familial first cancers, the risk for even a sporadic second cancer was very high (Table II). The increase in risk of the second cancers was largest during the first year of follow-up. Intensive medical surveillance after the first diagnosis and treatment-related exposure to carcinogens may partly explain the observed excess risk. However, there was no large difference in RR between 1 and 9 years and between 10 and 38 years after first diagnosis for breast cancer and melanoma, excluding any large effects by treatment at these two sites. Another explanation is that cancer patients are a subgroup at a high risk and that second cancer signals inherited or acquired susceptibility to cancer.

Cancer is attributed to the accumulation of genetic alterations in cells.3, 5, 7, 23 The familial risks observed here suggest that an inherited susceptibility predisposes a small proportion of cancer patients to second cancer. The high risk for second primary cancer in the patients without a family history would be consistent with a polygenic model and probably with a variable degree of environmental modification.

Ancillary