Epidemiology and Cancer Prevention
Familial colorectal adenocarcinoma from the Swedish family-cancer database
Article first published online: 10 SEP 2001
Copyright © 2001 Wiley-Liss, Inc.
International Journal of Cancer
Volume 94, Issue 5, pages 743–748, 1 December 2001
How to Cite
Hemminki, K. and Li, X. (2001), Familial colorectal adenocarcinoma from the Swedish family-cancer database. Int. J. Cancer, 94: 743–748. doi: 10.1002/ijc.1533
- Issue published online: 30 OCT 2001
- Article first published online: 10 SEP 2001
- Manuscript Accepted: 22 JUN 2001
- Manuscript Revised: 18 JUN 2001
- Manuscript Received: 21 MAR 2001
- King Gustaf V Jubilee Fund
- The Swedish Cancer Society
- familial risk;
- colorectal cancer;
Familial risks for colorectal (CRC) adenocarcinoma were characterized from the Swedish Family-Cancer Database covering 9.6 million individuals, whose family relationships and cancers were obtained from registered sources, not sensitive to reporting or ascertainment bias. Cancer cases were retrieved from the Swedish Cancer Registry from years 1958–96. Standardized incidence ratios (SIRs) were calculated based on gender-, age-, period- and tumor type specific rates. A total of 4,794 and 67,925 CRCs were recorded in offspring and parents, respectively. For colon and rectal adenocarcinoma, the SIRs in offspring were 2.28 and 1.68 by parental CRC adenocarcinoma, giving attributable proportions of 6.45 and 3.31%, respectively. The SIR of CRC was over 10 when both offspring and parents were diagnosed at a young age. The risk for parental CRC adenocarcinoma was over 100 when 2 or more children were affected. The risk in siblings was also very high when a parent was affected. The familial cancer sites that associated with CRC were those typical of hereditary nonpolyposis colorectal cancer (HNPCC). This is the largest study published on familial CRC and the only one reporting specifically on adenocarcinoma. The data suggest that HNPCC is the largest single disease entity among CRCs, probably accounting for less than 50% of familial CRC. Other familial components appear heterogeneous, characterized by incomplete penetrance, recessive mode of inheritance and few associated tumor sites. © 2001 Wiley-Liss, Inc.
Colorectal cancer (CRC) is one of the most common forms of neoplasia in Western and industrial countries, with 5,000 annual cases (over 10% of all cancers) in Sweden.1 Most CRCs are sporadic and migrant studies have shown that environmental factors are important in CRC.2, 3 Among the environmental factors diet and particularly the intake of vegetables, is assumed to be important.4 Physical activity is another environmental factor linked to the risk of CRC. Insulin and the insulin-like growth factor system are physiological factors modulating the effects of diet and physical activity and the levels of these factors associate with the risk of CRC.5 According to a recent twin study, 60% of the variation in CRC was assigned to random environmental effects and 35% to heritable factors; shared environmental effects accounted for 5% but this effect was not significant.6 Some 10–15% of the CRC burden has been suggested to be the result of a primary genetic factor.7, 8 The most common hereditary CRC disorder is hereditary nonpolyposis CRC (HNPCC) whereas familial adenomatous polyposis (FAP) accounts for minor proportion of the cases.9, 10 In both FAP and HNPCC the lifetime risk of cancer is between 80 and 100%.7, 8 HNPCC accounts for some 3% of CRCs in Finland and the families have an excess of other cancers, particularly at the endometrium, urinary tract, stomach and biliary system.11–14 Family clustering of CRC occurs even when the cases are not part of a defined hereditary syndrome. The history of CRC in first-degree relatives elevates a person's lifetime risk of CRC about 2-fold.15–17 An increased risk of the disease affects especially young people.18–20
In our study we analyze the risk of colorectal adenocarcinoma in first-degree relatives by family history using the nation-wide Swedish Family-Cancer Database.19, 21–23. Although adenocarcinoma is the main histological type of CRC, carcinoids are common among early onset cases and these may obscure familial relationships.24 Our previous studies on CRC have so far focused only on the proportion of HNPCC among familial cases.24, 25 The Database offers unique possibilities for reliable estimation of familial risks, because the data on family relationships and cancers were obtained from registered sources of practically complete coverage.25–27 Other advantages of the present study, compared to the previous ones, are a large population size and adherence specifically to adenocarcinoma. Based on the number of familial parent-offspring pairs, the present study is 2 times large than the largest cohort study15 and 3 or more times larger than other cohort studies18, 28 none of which considered histology.
SUBJECTS AND METHODS
The Swedish Family-Cancer Database includes all persons born in Sweden after 1934 with their biological parents, totaling over 9.6 million individuals.21 Cancers were retrieved from the nation-wide Swedish Cancer Registry from years 1958–96. The percentage of cytologically or histologically verified cases by site, sex and age at diagnosis for colon cancer was 98% and for rectal cancer was 99%.1 The linkage of families is practically complete in the Family-Cancer Database. Each offspring has been linked to the mother and father using the unique national identification number. For confidentiality reasons, this number has been replaced in the Database by an ad hoc identification number that was used in assembling families. The Family-Cancer Database has a gap among those born between 1935–40 who died between 1960–96. Many of these individuals lack links to parents in the Database and this probably causes a deficit of some cancers and a somewhat inflated risk estimates for fatal cancers. In our present study, the estimates for sibling risks among older individuals may be somewhat decreased because of this deficit. A 4-digit diagnostic code according to the 7th revision of the International Classification of Diseases (ICD-7) was used. The following ICD-7 codes were pooled: “oral” cancer, codes 161 (larynx) and 140–148 (lip, mouth, pharynx), except for code 142 (salivary glands), “lymphoma,” codes 200 (non-Hodgkin's lymphoma), 201 (Hodgkin disease) and 202 (reticulosis); and “leukemia,” codes 204–207 (leukemias), 208 (polycythemia vera) and 209 (myelofibrosis). ICD codes 153.0–154.9 were used for colorectal cancer excepted for code 154.1, which was separately used for anus. Basal cell carcinoma of the skin is not registered in the Cancer Registry. The histological classification of CRC was used to define adenocarcinoma as a pathological anatomic diagnosis (PAD) code 096. Carcinoids included both malignant and benign tumors.1 Group “other” included the remaining tumors, mainly undifferentiated and undefined histologies.
Family history information was collected on all first-degree relatives (parents, siblings and children). Data on siblings was only available on offspring. In most analyses, parents were considered probands and the cancer risk was calculated for offspring. In Table V, however, offspring were probands and in Table VI siblings were probands. Follow-up was terminated on death, emigration, or the closing data of the study, December 31, 1996, whichever came first. Standardized incidence ratios (SIRs) were calculated as the ratio of observed (O) to expected (E) number of cases. The expected numbers were calculated from 5-year-age-, period-, gender- and tumor type-specific standard incidence rates29 Confidence intervals (95%CI) were calculated assuming a Poisson distribution.29 In the calculation of 95%CIs for sibling risks, the dependence between the affected pairs was taken into consideration.30
The population attributable proportion of cases with a family history of CRC was estimated as follows: (familial SIR-1)/familial SIR × proportion of familial cases.31
The Family-Cancer Database covered years 1958–96 from the Swedish Cancer Registry and included 4,794 offspring and 67,925 parents with CRC, comprising 2 main histopathological subtypes (Table I). In offspring (age 0–61 years) 77.6% of all cases were adenocarcinomas and 20.9% were carcinoids; the high proportion of carcinoids is due to the young age of the population.24 Among parents, however, 93.8% of all cases were adenocarcinomas. Among offspring adenocarcinoma, 10.9% had a family history; among parents the proportion was 0.6%.
|Familial cases1||197 (10.4%)||208 (11.4%)||405 (10.9%)||190 (0.5%)||170 (0.6%)||360 (0.6%)|
Age-specific incidence rates of CRC among all men and women, including parents and offspring with and without a family history, are shown in Figure 1. The incidence of CRC increased with age and the male rates were somewhat higher than the female rates. The familial rates were higher than the sporadic rates at all age groups. The relative difference in familial and sporadic rates, however, was maximal in younger age groups.
Age-specific SIRs of CRC in offspring by parental CRC are shown in Table II. Paternal and maternal cancer sites were also analyzed separately but because there were no large differences in SIRs for the common sites, the data were combined for both parents. Overall, the SIR was 1.64 (95%CI 1.47–1.81) and the highest SIR, 12.37 (95%CI 7.74–18.07), was among offspring diagnosed at ages before 40, when parents' diagnostic age was 40–49. There was another high SIR of 11.54 (95% CI 3.00–25.61) at diagnostic age <40 for both offspring and parents.
|Parent's age at diagnosis||Offspring's age at diagnosis|
|O||SIR||95% CI||O||SIR||95% CI||O||SIR||95% CI||O||SIR||95% CI|
|70 and older||19||1.16||0.70||1.74||68||1.59||1.23||1.99||72||1.07||0.84||1.33||159||1.26||1.07||1.46|
SIR for CRC in offspring by parental cancer is shown in Table III. The number of familial CRC cases differ somewhat between Tables II and III because in the former CRC is analyzed combined but in the latter colon and rectum are separated among probands. The small difference in SIRs for CRC between Tables II and III is due to the higher familial risk between concordant (colon–colon, rectum–rectum) sites than between discordant sites (colon–rectum).25 The offspring were at a particular risk of developing CRC when parents presented with colon (SIR 2.18), rectal (SIR 1.70), endometrial (SIR 1.82), nervous system cancer (SIR 1.32) and bone tumors (SIR 3.79). A separate analysis was carried out using only those parents as probands who were diagnosed before age 62 years, to make them comparable in age to the offspring generation. This resulted in an increase in the SIR for the sites listed above. Additionally, CRC was increased from parental myeloma (SIR 2.50). For offspring diagnosed before age 40 years, the increased risk was even more pronounced by parental colon, rectal, endometrial and nervous system cancers. A borderline significant risk was also associated with parental ovarian cancer (SIR 2.08).
|Parent's cancer||All offspring||All offspring, parents diagnosed <62 years||Offspring diagnosed <40 years, parents diagnosed <62 years|
|O||SIR||95% CI||O||SIR||95% CI||O||SIR||95% CI|
|Upper aerodigestive tract||32||0.83||0.57||1.14||11||1.06||0.53||1.78||3||1.03||0.19||2.52|
Using the familial SIRs of Table III and the proportions of offspring with an affected parent, we can calculate the population attributable proportions for family history. For colon cancer the attributable proportion was 4.57% (i.e., [2.18-1]/2.18 × 274/3247; the last term is the number of offspring with an affected parent, divided by the number of all offspring). The attributable proportion for colon adenocarcinoma was 6.45%, for rectal cancer 3.25% and for rectal adenocarcinoma 3.31%.
Table IV presents SIRs for offspring cancers by parental CRC and endometrial cancer, considering double primaries. With 2 CRCs (1st and 2nd primary) in parents, mother with CRC and endometrial cancer and both father and mother with CRC resulted in a markedly increased risk of colon cancer in offspring. It is noteworthy that the SIRs were highest when mother had a CRC and an endometrial cancer. CRC and endometrial cancer in mother also lead to an increased risk for rectal, male genital (2 cases) and eye cancer (3 cases). Endocrine gland cancer in offspring was in excess when a parent had 2 CRCs; kidney cancer was increased when both father and mother presented with CRC.
|Offspring cancer||Parents with 2 colorectal adenocarcinomas||Mother with colorectal adenocarcinoma and endometrial cancer||Both father and mother with colorectal adenocarcinoma|
|O||E||SIR||95% CI||O||E||SIR||95% CI||O||E||SIR||95% CI|
|Upper aerodigestive tract||3||1.76||1.71||0.32||4.18||2||1.03||1.94||0.18||5.56|
|Other female genitals||1||0.33||3.06||0.00||11.99||1||0.27||3.72||0.00||14.58||1||0.18||5.57||0.00||21.85|
|Other male genitals||2||0.21||9.53||0.90||27.30||2||0.17||11.57||1.09||33.17|
SIRs for parental adenocarcinoma are shown in Table V by the number of affected offspring. The SIR for parents diagnosed before age 50 years was very high, 117.84 (adenocarcinoma, SIR 129.18) when 2 or more children had CRC. The risk leveled off when the parental diagnostic age was 50 years or more.
|Offspring colorectal cancer||Parental age at diagnosis|
|O||E||SIR||95% CI||O||E||SIR||95% CI||O||E||SIR||95% CI|
|No colorectal cancer||5224||5255.69||0.99||0.97||1.02||57209||57349.87||1.00||0.99||1.01||62433||62605.57||1.00||0.99||1.01|
|One child colorectal cancer||33||7.26||4.54||3.13||6.23||349||212.78||1.64||1.47||1.82||382||220.04||1.74||1.57||1.91|
|Two or more children colorectal cancer||6||0.05||117.84||42.41||231.00||6||1.51||3.98||1.43||7.81||12||1.56||7.70||3.96||12.68|
|Any children colorectal cancer||39||7.31||5.33||3.79||7.14||355||214.28||1.66||1.49||1.83||394||221.60||1.78||1.61||1.96|
Sibling risks were calculated separately for siblings of CRC probands when a sib was diagnosed before age 40 or later (Table VI). Siblings could be identified only in the offspring generation. The SIR for CRC in siblings diagnosed at any age was 5.45 (adenocarcinoma SIR 6.49), but when a parent was also affected the SIR was 12.90. If the sib was diagnosed before age 40, the SIR was 36.36 in multicase families. For rectal cancer the risk at any age was 3.29 (adenocarcinoma SIR 3.37) but it was 9.09 for rectal adenocarcinoma diagnosed before age 40 years; in multicase families the SIR was 100.00. An interesting difference between colon and rectal adenocarcinoma was in families where a parent lacked CRC: SIR was 2.18 for colon but only 0.64 for rectum. The risk of endometrial cancer was 1.92 (borderline significance); notably, no other site was in excess.
|Cancer sites||Sibling age at diagnosis|
|O||E||SIR||95% CI||O||E||SIR||95% CI||O||E||SIR||95% CI|
|Upper aerodigestive tract||1||0.90||1.11||0.00||6.14||5||3.15||1.59||0.35||4.65||6||4.05||1.48||0.37||4.12|
|Parents without CRC||4||1.02||3.92||0.72||12.31||21||10.45||2.01||0.88||4.19||25||11.46||2.18||1.00||4.41|
|Parents with CRC||4||0.11||36.36||6.69||114.17||12||1.14||10.53||3.83||24.50||16||1.24||12.90||5.20||28.28|
|Parents without CRC||5||7.36||0.68||0.15||1.99||5||7.80||0.64||0.14||1.88|
|Parents with CRC||5||0.05||100.00||22.31||292.55||3||0.80||3.75||0.50||13.00||8||0.85||9.41||2.84||24.13|
|Other female genitals||2||0.53||3.74||0.25||15.16||2||0.76||2.64||0.18||10.72|
We analyzed data from the nationwide Swedish Family-Cancer Database, which is unique both by its size and population-based structure. The main limitation is the age truncation due to the registers used, i.e., the offspring population was born after 1934 and the Cancer Registry was started in 1958. The follow-up was until the end of 1996, giving 61 years as the maximal age in the offspring generation. Another limitation is that in 1 birth cohort (years 1935–40) of diseased offspring, linkage to parents is incomplete. None of these limitations, however, should cause bias to the results. The age of the parental population is not limited and the systematic effects of diagnostic age on familial risk can be confirmed by parental age (Table II). Nor are there population screening programs that could potentially bias the study, except that FAP has been recognized as a syndrome for a long time and the screening in family members has probably extensively reduced the risk of invasive CRC.25 We have observed clustering of cancers typical of FAP among in situ colon cancers only and believe that very little of the familial clustering observed in our present study is due to FAP. Family screening for HNPCC has been started later, probably with minimal repercussions to our study.
The present overall familial risk of 2.18 and 1.68 for colon and rectal adenocarcinoma, respectively, agrees with the estimates from other population-based studies, considering the age distribution of 0–61 years in the present population and the fact that the previous studies did not focus on a specific histology.15, 18 The population attributable proportion for a family history of 6.45% and 3.31% in colon and rectal adenocarcinoma are also consistent with the 7% estimate from a U.S. study on first-degree relatives because the inclusion of all first-degree relatives gives a higher attributable proportion.18 The risk for offspring CRC was over 10 when they and their parents were diagnosed at a young age, <40 for offspring and <50 for parents.
We characterized familial CRC in many ways, some of which are novel. The extracolonic sites that associated with CRC were endometrium, nervous system, ovary (early onset cancer) and kidney (when both parents had CRC). All these cancers are manifestations in HNPCC.8, 11, 32 Double primary cancers at colorectum and endometrium have previously been pointed out as likely indicators of HNPCC.33 We observed an SIR of 4.87 when a parented presented with 2 CRCs and of 8.44 when a mother presented with a CRC and endometrial cancer (Table IV). Other sites that were increased in some analyses were bone, male genitals, eye, endocrine glands and bone marrow (myeloma). None of these, however, showed a consistent pattern of increase and the association to CRC adenocarcinoma or, possibly to HNPCC, remains tentative. The lack of associations to sites such as breast and prostate that have been noted in earlier studies is notable.34
Some of the present results suggest underlying genetic mechanisms. All common cancers have many genetic and environmental causes, i.e., they are heterogeneous. The 35% heritable component for CRC in twin studies suggests that, in addition to HNPCC, other heritable factors underlie susceptibility to CRC.6, 35 A twin study makes no assumptions about the mode of inheritance and both a recessive and polygenic mode can be accommodated, in addition to the dominant mode. Familial clustering may be defined between parents and offspring, indicative of a dominant mode of inheritance, between siblings without affected parents, indicative of a recessive inheritance, or among all first-degree relatives, implying a mixed mode. Two recent studies from the Swedish Family-Cancer Database suggest that in colon and rectal cancer, a recessive mode can be discerned, in addition to a dominant mode,30, 36 In our present study, the risk of parental CRC adenocarcinoma was vastly increased to an SIR of 129 when 2 or more offspring presented with a concordant cancer (Table V). The risk, however, was almost solely focused on those diagnosed before age 50 years and was probably largely due to HNPCC. Cancer risks to siblings were also relevant to this point (Table VI). From a CRC proband, an increased risk of sibling rectal cancer was only observed when a parent was also affected, i.e., an HNPCC-like dominant effect. The risk of colon cancer, however, was 2.18 in the large number of families where neither parent was affected, suggesting an incomplete penetrance or a recessive-type of an effect.
Finally, is there familial CRC other than HNPCC and other known rarer syndromes? We have recently estimated epidemiologically from this Database that from 20–50% of familial CRC can be ascribed to HNPCC among offspring aged 0–61 years.25 The estimate was based on the parent-offspring relationships only and resorted to the comparison of familial risks between anatomic sites in CRC, as HNPCC has preferentially a proximal localization and between CRC and endometrium. The assumption was that the association between CRC and endometrial cancer was only due to HNPCC and that the risk of endometrial cancer was somewhat higher than that of CRC.8, 37 The present data on offspring CRC risk by parental CRC and endometrial cancer would also be consistent with the estimation that some 50% of familial CRC adenocarcinoma was due to HNPCC and the proportion could be even higher among those diagnosed before age 40 (Table III). When the parental CRC risk was analyzed by offspring cancer, the association to endometrial cancer was small (SIR 1.21) compared to colon and rectal adenocarcinoma (SIRs 1.96 and 1.94, Table IV), suggesting a far lower attributable proportion of HNPCC at higher age. The other relevant new data relate to sibling risks (Table VI). Particularly in colon cancer, the majority of the affected siblings presented with a relatively late onset disease and without an affected parent, thus suggesting involvement of low-penetrance or recessive genes unrelated to HNPCC. These non-HNPCC familial CRCs will probably be difficult to characterize genetically, because they may be heterogeneous with low penetrance, recessive mode and no strong association to other types of cancers.
- 1Epidemiology Centre. Cancer incidence in Sweden 1997. Stockholm: The National Board of Health and Welfare, 1999.
- 2IARC. Cancer: causes, occurrence and control. Lyon: IARC, 1990.
- 4World Cancer Research Fund. Food, nutrition and the prevention of cancer: a global perspective: Washington, DC: American Institute of Cancer Research, 1997.
- 291994. Statistical methods in cancer research. Volume IV. Descriptive epidemiology. IARC Sci Publ 1994;128: 1–302., , .
- 31Modern epidemiology. 2nd ed. Philadelphia: Lippincott-Raven, 1998., .
- 32Hereditary nonpolyposis colorectal cancer. In: VogelsteinB, KinzlerKW, eds. The genetic basis of human cancer. New York: McGraw-Hill, 1998, p 333–46..