Article Text

PDF

Estimation of the familial relative risk of cancer by site from a French population based family study on colorectal cancer (CCREF study)
  1. N Andrieu1,
  2. G Launoy2,
  3. R Guillois3,
  4. C Ory-Paoletti4,
  5. M Gignoux2
  1. 1Inserm EMI00-06, Tour Evry 2, 523 Place des Terrasses de l’Agora, 91034 Evry Cedex, France, and Service de Biostatistiques, Institut Curie, 26 rue d’Ulm, 75248 Paris Cedex 5, France
  2. 2Inserm ERI3-Faculté de Médecine, CHU de Caen, 14033 Caen, France
  3. 3Inserm EMI00-06, Tour Evry 2, 523 Place des Terrasses de l’Agora, 91034 Evry Cedex, France
  4. 4Unité Inserm xu-521-Institut Gustave Roussy, 39, rue Camille Desmoulins, 94805 Villejuif Cedex, France
  1. Correspondence to:
    Dr N Andrieu
    Inserm Emi 00-06/Service de Biostatistiques, Institut Curie, 26, rue d’Ulm, 75248 Paris Cedex, France; nadine.andrieucurie.net

Abstract

Background: Colorectal cancer (CRC) is the second most common cause of death from cancer in France. A family history of CRC increases an individual’s risk of developing CRC. Family history has been suggested to have a greater impact on proximal than distal tumours.

Aim: We estimated the familial risk of CRC and other cancers, and examined how risk varies according to localisation of the tumour in the colorectal tract.

Subjects: We recorded all cases of CRC diagnosed between 1993 and 1998 in the region served by the Calvados Cancer Registry. A trained interviewer asked all participants about their family history of cancer.

Statistical methods: Familial risk was estimated from a cohort analysis of the relatives of the CRC cases. The expected numbers of cancers were calculated from Calvados incidence rates. Familial relative risks were calculated using standardised incidence ratios.

Results: Our findings showed that colon cancer had a stronger familial/genetic component (relative risk (RR) 1.47) than rectal cancer (RR 0.98). The familial/genetic component appeared stronger for proximal colon cancer than for distal colon cancer only among women (RR 2.24 v RR 1.45). CRC appeared to be positively associated with leukaemia (RR 1.77), stomach cancer (RR 1.32), and testicular cancer (RR 3.13), and negatively associated with urinary bladder cancer (RR 0.57) within families. The cancer spectrum associated with CRC among younger participants included prostate (RR 1.93), uterus (RR 2.49), and thyroid (RR 3.85) cancers.

Conclusion: If our results are confirmed, follow up guidelines for patients with a family history of CRC should depend on the sex and tumour site of affected relatives to avoid needless invasive screening.

  • CRC, colorectal cancer
  • RR, relative risk
  • SIR, standardised incidence ratio
  • colorectal cancer
  • family study
  • relative risk
  • French population

Statistics from Altmetric.com

Colorectal cancer (CRC) is the second most common cause of death from cancer in France. An estimated 36 000 cases were diagnosed in 2000 in France.1 As with many cancers, a family history of CRC increases an individual’s risk of developing the disease. The risk of CRC associated with a family history of CRC has been estimated mainly from studies of first degree relatives; a family history of CRC is usually estimated to double the risk.2–4 Other studies have concentrated on more selected populations (that is, high risk families). These have proved that some syndromes, including less than 5% of CRC cases, are inherited.5 Some researchers have suggested that family history has a greater impact on proximal than distal tumours.6,7 Indeed, there are numerous developmental, biological, and molecular differences between the proximal and distal colon and rectum that suggest differing susceptibilities to neoplastic transformation.8–10

A population based family study, the CCREF (Calvados ColoREctal Family) study, has been carried out in Calvados, France.11 The main aim of this study was to determine the role of inherited factors in the transmission of CRC. In this paper, we estimated the familial risk of CRC and other cancers and examined how these risks varied according to localisation of the tumour in the colorectal tract.

MATERIALS AND METHODS

We recorded all cases of CRC diagnosed in the region covered by the Calvados Cancer Registry between September 1993 and December 1998. Information on tumour histology was collected for all CRC cases. Patient details were obtained from the pathology laboratories in the Calvados region. Patients’ general practitioners were contacted to obtain permission to contact the patient. The Calvados Research Ethics Committee approved the study design.

A trained interviewer asked all participants about their family history of cancer. The family history questionnaire included details about spouses and all first and second degree relatives, including their date of birth, age at death if deceased, history of cancer, age at diagnosis, and place of care. Information on all cancer cases in family members was verified through the local cancer registers, and by checking the records of medical care centres, general practitioners, or specialists.

Statistical methods

Familial cancer risks were estimated from a cohort analysis of the relatives of the CRC cases. At risk patients entered the cohort on 1 January 1970, or at birth if they were born after the 1 January 1970. They were censored on diagnosis of CRC if affected, on death if deceased, or when the family history questionnaire was completed in other cases.

Relatives whose date of birth was unknown were assigned a fictitious date of birth according to their position in the family. The date of birth remained unknown for approximately 20% of relatives. This missing information concerned mainly grandparents, who accounted for very few person years (≈0.6% of total person years). The expected number of cases of cancers was calculated from Calvados age, sex, and period specific incidence rates. Four periods were available for CRC: 1978–1982, 1983–1987, 1988–1992, and 1992–1995. Only the first three periods were available for the other cancers. Familial relative risks (RR) were calculated using standardised incidence ratios (SIR). Two sided 95% confidence intervals (CI) for relative risk estimates, heterogeneity, and trend tests are based on the Poisson distribution.12

RESULTS

Between September 1993 and December 1998, 1351 cases of CRC were diagnosed in the Calvados region. Thirty four of these patients had already deceased when the information concerning CRC diagnosis was received. Thus 1317 general practitioners were contacted to obtain permission to contact their patients. The patient had since deceased in 94 cases (7.1%), 137 refused to allow us to contact their patients (10.4%), and 1085 gave permission to contact their patients (82.5%). The major reason for general practitioner refusal was the patient’s worsening health or, less frequently, the patient’s mental state. Of the 1085 patients for whom permission was obtained, 36 had deceased between the general practitioner giving permission and the interviewer contacting them (3.3%), 282 refused to participate (26.0%), and 767 agreed to participate (70.7%). Thus 767 of 1351 incident cases (56.8%) were included in the CCREF study. The local research ethics committee did not allow us to ask CRC patients the reasons for their refusal to participate in the study, and thus causes of CRC patient refusal could not be studied. One CRC patient was excluded because of lack of information about his family.

The 766 participating patients (or probands) belonged to 761 independent families. Five families contained two probands. The number of relatives according to sex, and mean age (at interview if alive or age at death if deceased) was determined for each type of relationship (table 1). The mean number of people per family was 14.9 (SEM 5.6). There were an average of 7.7 (SEM 3.6) first degree relatives and 6.0 (SEM 4.4) second degree relatives. We also calculated the number of person years per relationship category (table 1), corresponding to the follow up data from 1 January 1970.

Table 1

 Distribution of relatives according to their relationship with the proband

There was no difference in tumour localisation between participating and non-participating patients (table 2). Approximately half (50.8%) of the tumours in female probands were localised in the proximal colon (caecum-ascending and transverse) compared with 40.6% in male probands (data not shown).

Table 2

 Localisation of tumours in participating and non-participating probands

Declared cancer cases were systematically verified. When the precise intestinal site was known (colon or rectum), CRC was confirmed in 79.1% of cases in all degree relatives and in 87.5% of cases in first degree relatives. When the precise site was unknown, the rate of confirmed CRC decreased to 49.8% in all degree relatives and to 62.5% in first degree relatives (table 3). CRC cases were included in this study if they were diagnosed after 1 January 1970. Of the 107 cases considered to be CRC (colon-rectum and intestine SAI), 78% were confirmed (data not shown).

Table 3

 Distribution of tumours by localisation among the 761 families

For other cancer sites, the rate of confirmed cases of cancer among relatives differed according to cancer site. For example, the rate of histologically confirmed cancer was very low for the stomach and pancreas (33.3% and 30%, respectively), medium for the uterus (precise and unknown sites), breast, and prostate (49.4%, 51.4%, and 56% respectively), approximately 70% for the kidney and urinary bladder, and between 80% and 92% for the ovaries, the uterus (precise site), and thyroid. All melanomas and testicular cancers were confirmed by pathological records. For all sites, nearly 100% agreement was found between case reports and pathological reports (data not shown).

Estimate of familial CRC risk according to tumour localisation of proband

The risk of developing CRC associated with having a family history of CRC was 1.54 (95% CI 1.26–1.86) (table 4). The familial CRC risk appeared to be greater, but not significantly so, for first degree relatives than for second degree relatives (RR 1.71 (95% CI 1.35–2.13) and RR 1.22 (95% CI 0.82–1.76), respectively) (data not shown). The risk of familial CRC was slightly, but not significantly (p for trend ∼0.10), associated with tumour localisation among relatives with CRC (both known and unknown sites). The risk decreased with distance from the ileum; from the caecum-ascending colon (RR 2.04 (95% CI 1.24–3.15)), transverse colon (RR 1.57 (95% CI 0.78–2.82)), descending-sigmoid colon (RR 1.68 (95% CI 1.19–2.30)) to the rectosigmoid junction and rectum (RR 1.26 (95% CI 0.89–1.75)). However, this decrease was only significant (p for trend = 0.03) among relatives with colon cancer; the risk of developing colon cancer decreased from 2.37 (95% CI 1.26–4.05) when the proband’s tumour was located in the caecum-ascending section, to 2.03 (95% CI 0.87–3.99) in the transverse section, 1.43 (95% CI 0.85–2.26) in the descending-sigmoid section, and 1.13 (95% CI 0.67–1.79) in the rectosigmoid junction-rectum section (table 4). The risk of rectal cancer was not increased by a family history of CRC. Indeed, SIR was 0.98 (95% CI 0.67–1.40) and no trend was found according to localisation of the proband’s tumour.

Table 4

 Relative risk of developing colorectal cancer (CRC) among relatives of the 761 families according to tumour localisation

We estimated the familial risk of CRC according to the sex of affected relative (table 5) and age of the proband at diagnosis (table 6).

Table 5

 Relative risk of colorectal cancer (CRC) among relatives of the 761 families according to tumour localisation and relative’s sex

Table 6

 Relative risk of colorectal cancer (CRC) among relatives of the 761 families according to the proband’s age at diagnosis and tumour localisation

The estimated risk did not differ according to sex (SIR 1.55 (95% CI 1.16–2.03) among female relatives; SIR 1.53 (95% CI 1.15–2.00) among male relatives). The familial CRC risk decreased with distance from the ileum only among female relatives. Indeed, a stronger trend was observed among female relatives (p for trend = 0.07) than among males and this difference was more significant when the familial risk of developing colon cancer was considered (p = 0.04 among female relatives and NS among male relatives). For both female and male relatives, the risk of rectal cancer was not increased by a family history of CRC and no trend was observed according to localisation of the proband’s tumour. There was a slight, but non-significant, increase in the risk of rectal cancer among females when the proband’s tumour was located in the proximal section. Similar familial risks were found according to the proband’s sex. Indeed, there was a significant trend in colon familial risk (p = 0.025) according to localisation of the proband’s tumour when the proband was a woman (data not shown).

There was a slight difference between the estimated familial risk of CRC among relatives of probands aged 60 years or less at the time of diagnosis and those of probands aged more than 60 years at the time of diagnosis (1.81 (95% CI 1.28–2.48) and 1.43 (95% CI 1.11–1.80), respectively) (table 6). Among relatives of probands aged over 60 years, there was no trend in familial risk according to localisation of the proband’s tumour. Conversely, the familial risk of CRC (p = 0.023) and colon cancer (p = 0.03) decreased with distance from the ileum among relatives of probands aged less than 60 years at diagnosis. For relatives of both younger and older probands, the risk of rectal cancer was not increased by having a family history of CRC (0.95 and 1.00, respectively) and no trend was found according to localisation of the proband’s tumour. When relatives were subdivided according to localisation of the proband’s tumour by both age and sex of the proband, the number of subjects in each category was too small to provide sufficient precision. However, the familial risk of CRC decreased significantly (p = 0.03) with distance from the ileum among female relatives of younger probands (2.82, 2.52, and 0.71 respectively for proximal, distal, and rectosigmoid junction-rectum sections). The risk of CRC among female relatives of older probands with distal cancer was 2.00 (95% CI 1.03–3.49). When the six families fulfilling the Amsterdam criteria13,14 were excluded, the results were unchanged.

Estimate of the risk of cancer at other sites according to tumour localisation of the proband

We estimated the risk of cancer at 21 other sites among relatives according to localisation of the proband’s tumour. Among 21 cancer sites tested, only four SIRs (19%) were significant based on 95% CI (table 7).

Table 7

 Relative risk of developing cancer at selected sites among relatives according to localisation of the proband’s colorectal cancer (CRC)

The risks of testicular cancer (3.13 (95% CI 1.14–6.80)) and leukaemia (1.77 (95% CI 1.01–2.88)) were elevated among relatives. The risk of stomach cancer was slightly, but not significantly, elevated among relatives of probands with CRC (1.32 (95% CI 0.94–1.81)). The risk of urinary bladder cancer (0.57 (95% CI 0.30–0.98)) was significantly decreased among relatives of CRC probands. Probands with proximal and distal colon cancers were grouped together as there was no difference in familial risks of developing cancer at any site (data not shown). Even though the subdivision of our population according to whether the proband had proximal or distal CRC led to small categories, the point estimates for a given cancer site were similar in both groups. Thus the risk of cancer at different sites is shown according to whether the proband had colon or rectal cancer. There were no significant differences in estimates except for urinary bladder cancer, which for the heterogeneity test led to significance (p = 0.052). The risk of developing bladder cancer was lower when the proband had rectal cancer (0.21 (95% CI 0.02–0.77)) than when the proband had colon cancer (0.85 (95% CI 0.43–1.53)). The risk differed significantly according to the proband’s age at CRC diagnosis for prostate (p = 0.002), uterus (p = 0.005), and thyroid (p = 0.04) cancer. The risk of developing a cancer at these sites was greater when the proband was less than 50 years at CRC diagnosis than when they were older (prostate 1.93 v 0.65; uterus 2.49 v 0.86; thyroid 3.85 v 0.73). There was no difference in SIR according to the proband’s age for other sites such as the stomach, urinary bladder, or breast. When the three families that fulfilled the Amsterdam criteria II13 were excluded (that is, criteria including extracolonic cancers: endometrium, small bowel, ureter, or renal pelvis), the association with cancer of the uterus among younger subjects was weaker but still significant. For all other associations, exclusion of these families did not change the results (data not shown).

DISCUSSION

We have shown here that colon cancer had a stronger familial/genetic component than rectal cancer. Our findings provide evidence that a family history of colorectal cancers is associated only with colon cancer and, more precisely, that a family history of colon cancer is associated with colon cancer. Although this observation may be affected by the fact that rectal cancer represents approximately 30% of all colorectal cancers, implying that we might have no power to detect a familial/genetic effect in rectal cancer, the familial risk point estimates were small (that is, between 0.7 and 1.3) except among women and younger patients where a family history of rectal cancer might increase the risk of developing proximal cancer. The familial/genetic component appeared stronger (although not significantly) for proximal colon cancer than for distal colon cancer, but only among women.

CRC appeared to be positively associated with leukaemia, stomach cancer, and testicular cancer, and negatively associated with urinary bladder cancer within families. The cancer spectrum associated with CRC among younger participants included prostate, uterus, and thyroid cancers, which are all hormone dependent tumours.

The results of this study are in agreement with previous reports of the relationship between family history and risk of colorectal cancer by site.3,15–25 All of these studies found a difference between familial risk associated with colon and rectal cancer, with a higher familial risk associated with the colon than with the rectum. The amplitude of the familial risk associated with the rectum varies from study to study but most found that RR was increased by approximately 50% whereas that associated with colon cancer was generally found to be increased by approximately 100%. However, Planck et al. found that the risk of developing rectal cancer was multiplied by 10 among children of women with rectal carcinoma diagnosed before the age of 50 years, even though the confidence interval was large.25 In common with us, Fuchs and colleagues26 found no increase in familial risk associated with rectal cancer. However, these groups provided contradictory results concerning the familial risk according to tumour localisation. Indeed, some groups found that the familial risk was higher for distal tumours,3,26 others found that it was higher for proximal tumours,6,7 and others found no difference.9,18–20,22,27 This discrepancy may be due to lack of power, as most of these studies included a small number of proximal tumours. Alternatively, it may be due to residual heterogeneity in the familial risk even after subdivision by localisation. Indeed, in our study, the familial risk appeared stronger for the proximal colon than for the distal colon only among women. Molecular differences have been identified between proximal and distal tumours, suggesting that two different genetic pathways are responsible for the development of these tumours. Microsatellite instability, which was initially described in hereditary non-polyposis colorectal cancer, is more frequent in proximal than distal tumours whereas chromosomal instability involving common allelic losses in tumours such as APC on 5q, p53 on 17q, are more frequent in distal and rectal tumours than in proximal tumours.6,28,29 Thus the nature and amplitude of the genetic factors involved in the initiation or/and progression of the tumour may differ between localisations.

With the exception of melanomas and testicular cancer, the declared cancers were not all confirmed histologically, leading to a possible bias in risks if cancer was misreported. However, this is unlikely. Indeed, a recent study showed that cancer affected probands reported their family history of cancer with a high positive predictive value. Some cancers, such as female breast cancer, prostate cancer, and leukaemia, were reported accurately for first, second, and third degree relatives, with a positive predictive value exceeding 70%.30 Family members may be more aware of cancers in younger relatives than in older relatives because deaths due to cancer at a young age are more memorable than those at an older age, which may be considered as being due to old age. This may lead to underreporting of cancers in older relatives. In our study, non-affected family members were not checked because of the lack of a national French registry and the limited time period covered by the Calvados registry. Therefore, underreporting of cancers among older relatives might have affected the relationship between CRC and cancers that typically occur later in live, such as prostate cancer. However, Ziogas and Anton-Culver also showed that the negative predictive value was more than 97% for all cancer sites, suggesting that the number of underreported cancers was low even among older relatives.30

Most of the observed association with uterine cancer may be due to the three families prone to hereditary non-polyposis colorectal cancer that met the Amsterdam criteria II.13 However, when these families were excluded from the analyses, a residual association was found in our study, as in other studies.3,31,32 This is the first time that familial associations have been found between colorectal cancer and leukaemia, testicular cancer, and thyroid cancer in members of the younger proband’s family. In agreement with our results on families of younger probands, the risk of CRC was significantly increased by a family history of prostate cancer in two of three studies,2,3,32 but only among women in the study of Slattery and Kerber.3 A bias in prostate cancer detection cannot be excluded in our study or in others if awareness of a family history of CRC led the general practitioner to perform a rectal examination, which generally includes prostate examination, among male relatives. However, the CRC cases in our study were incident cases and therefore all cases of prostate cancer were reported retrospectively and were unlikely to be correlated with the proband’s diagnosis. Finally, the familial risk of CRC was similar among families of younger probands and female probands in our study. Surprisingly, we found no increase in familial risk of breast cancer associated with a family history of CRC. However, only three of seven studies that looked for an association with breast cancer2,3,31–35 found a significantly increased risk.2,3,34

Together with the elevated familial risk associated with proximal cancer, the increased familial risks associated with testicular and prostate cancer agree with the hypothesis that the increased risk associated with a family history among women and younger subjects is partially due to the influence of endogenous hormones, which are genetically controlled.

Precise and unbiased estimates of the risk of developing CRC are required to draw up relevant guidelines. Most current guidelines were determined using substantially high relative risks of CRC associated with being a first degree CRC relative, without taking into account any potential variation in this relative risk. Indeed, the American Cancer Society recommended full colon examination beginning at age 40 years (or 10 years before the age of the proband if younger than 50 years) if either colon or rectal cancer were diagnosed in a first degree relative aged less than 60 years.36 However, if our results concerning the lack of familial risk associated with rectal cancer are confirmed, some invasive screening may be prevented and screening might depend on the sex of the affected relative and tumour localisation. The association between CRC and a family history at any other sites, such as testis, prostate, or thyroid, has to be confirmed before it can be considered as a criterion for screening recommendations. Differences in familial risks may improve screening strategies. Indeed, a family history of CRC is recognised as a common risk factor for CRC and most screening strategies incorporate mainly family history.36

Acknowledgments

This study was made possible by a fruitful collaboration between general practitioners, anatomopathologists (Drs J Chasles, E De Ranieri, P Dupin, F Galateau, B Gallet, J Jehan, A-M Mandard, D Panthou, F Petit, M Rollet and P Rousselot), gastroenterologists, radiotherapists, radiologists, surgeons, and general practitioners working in Calvados. This study was financed by la Fondation de France, la SNFGE, la Ligue Nationale Contre le Cancer, les ligues départementales: du Calvados, de l’Orne, de la Manche, des Yvelines, de l’Essonne et des Hauts-de-Seine, l’Institut Gustave Roussy, l’ARC, la MGEN, and l’Inserm.

We would like to thank S Marlière, C Pain, S Iacobelli, C Bégon, and Mrs Ozanne for their technical help, Dr B Guillois for his help with blood sample collection, and Alex Edelman and associates for correcting our English.

REFERENCES

View Abstract

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.