First-degree relatives of individuals with colorectal cancer are approximately twice as likely to get diagnosed with colorectal cancer or their precursor lesions adenomatous polyps as people with unaffected relatives.1, 2 The risk of colorectal cancer is 4 times higher for people who have more than one such relative compared with those without one.1, 2 A study on twins suggested that about 35% of all colorectal cancer cases can be attributed to heritable factors.3 Thus considering that well-understood familial syndromes such as the familial adenomatous polyposis (FAP) or hereditary nonpolyposis colorectal cancer syndrome (HPNCC) account for less than 5% of all colorectal cancers,4 genetic factors must also be involved in sporadic colorectal carcinogenesis.
There are various ways by which a family history of colorectal cancer may affect colorectal cancer risk. A family history of colorectal cancer may convey a genetic susceptibility that enhances the formation of new lesions or affects the transition from adenomas to carcinomas. Shared behavioral risk factors have also been proposed to underlie part of the association between a family history of colorectal cancer and risk of colorectal neoplasm.
Indications for effects of heritable factors on adenoma growth were found in a study in which small polyps were left in situ for 3 years before removal5: net adenoma growth was observed in 9 out of 14 (64%) patients with a family history of colorectal cancer but only in 22 out of 73 (30%) of the patients without such a family history. Findings from a case-control study on 362 patients suggested a stronger positive association of family history with large adenomas [odds ratio (OR) = 2.1, 95% confidence interval (CI) = 1.3–3.4] than with small adenomas (OR = 1.2, 95% CI = 0.7–2.2).6 Although the same impression can be obtained from smaller studies,7–10 a pooled analyses on 518 adenoma patients observed a stronger association between a family history of colorectal cancer and adenomas with no, mild or moderate dysplasia (OR = 1.5, 95% CI = 1.1–2.1) than with adenomas with severe dysplasia, carcinoma in situ or intramucosal carcinoma (OR = 1.0, 95% CI = 0.6–1.9). However, the difference between these associations was not statistically significant.11
The influence of genetic factors on adenoma multiplicity is clearly visible in patients with inherited polyposis syndromes.4 Responsible genes such as Adenomatous Polyposis Coli (APC) or human MutY homologue (MYH) could have a more subtle influence in sporadic carcinogenesis.4 One study noted a positive association between a sporadic family history of colorectal cancer and adenoma multiplicity.6
We examined associations between family history of colorectal cancer, adenoma multiplicity and adenoma advancement in a large prospective cohort study, which allowed for adjustment for known behavioral risk factors for colorectal neoplasia. Evidence regarding such associations at the population level may yield clues to the search of genetic factors involved in the development of sporadic colorectal adenomas and cancer.
APC, adenomatous polyposis coli; BMI, body mass index; CI, confidence interval; G/d, gram per day; FAP, familial adenomatous polyposis coli; HNPCC, hereditary non polyposis colorectal cancer; HPFS, Health Professionals Follow-Up Study; MET, metabolic equivalent; MYH, MutY homologue; MMR, mismatch repair; OR, odds ratio.
Material and methods
The current study comprises a subset of men of the Health Professionals Follow-Up Study (HPFS): an ongoing prospective study among 51,529 male US health professionals who responded to a mailed questionnaire in 1986 when they were between 40 and 75 years old. The Human Subjects Committee of the Harvard School of Public Health approved the HPFS.
At enrolment in 1986, and every 2 years thereafter, the cohort members were requested to fill out follow-up questionnaires to update information on various risk factors and to identify newly professionally diagnosed cases of various diseases. In 1986, 1990, 1994, 1998 and 2002, participants also completed a semi-quantitative food-frequency questionnaire. Questionnaires were mailed up to 4 times to nonresponders. At the time of the 2002 questionnaire, 37,431 men were alive and still participating.
Only men who completed the 1986 questionnaire and who underwent colonoscopy or sigmoidoscopy between 1986 and 2004 were eligible for the present analyses. Participants who reported to have had colorectal polyps, cancer (except nonmelanoma skin cancer), ulcerative colitis or Crohn's disease before 1986 were excluded. We also excluded cohort members who reported implausible caloric intakes, i.e., lower than 800 kcal/d or greater than 4,200 kcal/d) as well as those who left 70 or more of the items blank on the food-frequency questionnaire. This resulted in a base population of 29,120 men.
For each man who reported to have had an adenoma on a follow-up questionnaire for the first time, we asked for permission to request and review relevant medical records. A study investigator, who was blinded to exposure status, reviewed endoscopy and pathology reports. Only when the self-reported diagnosis was confirmed by a histopathological report, case status was assigned. The self-report of a negative endoscopy was reliable; a review of the medical records obtained from a random sample of 200 patients who reported a negative endoscopic result confirmed the absence of adenomas in all cases.
Adenomas in the cecum, ascending colon, hepatic flexure, transverse colon or splenic flexure were classified as being in the proximal colon. Adenomas in the descending or sigmoid colon were classified as being in the distal colon and adenomas in the rectum or at the rectosigmoid junction were classified as rectal.
For diagnoses up to 1990, the number of adenomas, the size of the largest adenoma as assessed endoscopically and pathologically, and location of the most proximal adenoma were recorded. For later diagnoses, number and size of the largest adenoma according to endoscopy and pathology reports were recorded for distal, proximal and rectal adenomas separately. In most years, only the most severe histological subtype of adenoma (in ascending order: tubular, tubulovillous, villous, carcinoma in situ) was registered. From 2002 onward, the most severe histopathological adenoma subtype was recorded for each subsite separately.
Between 1986 and 2004, 3,920 of the participants were diagnosed with adenomas. For classification purposes, we used the adenoma size according to the endoscopy report, but data from the pathology report were used if such information was lacking. Men having at least 1 adenoma ≥ 1 cm or with a (tubulo)villous structure or carcinoma in situ were classified as having advanced adenomas (n = 1,515); men having only adenomas that were smaller than 1 cm and with no mention of a villous or serrated structure were classified as having nonadvanced adenomas (n = 1,518). No information on size and histopathology was present in the pathology and endoscopy reports of 638 adenomas, and the sizes of 231 tubular and 18 serrated adenomas were unknown; these adenomas were not classified as being either advanced or nonadvanced in the main analyses. Because not all men underwent colonoscopy, the classification according to adenoma multiplicity was based on distal and rectal adenoma only (which we collectively describe as “distally located”), under the assumption of a single adenoma where the number of adenomas was not mentioned in the endoscopy and pathology reports (single distally located adenomas: n = 2,003; multiple distally located adenomas: n = 629; no distally located adenomas at all: n = 26,445). The adenoma location was unknown for 43 patients, and these adenoma patients were excluded from the analyses on multiplicity.
Assessment of family history of colorectal cancer
The 1986 questionnaire included questions on the diagnosis of colorectal cancer and corresponding age (before age 50, age 50 to 59, age 60 to 69, age 70+, age unknown) in the father and mother separately. In 1990 and 1992, the men were asked whether any of the siblings had colorectal cancer, and this was also asked for the father and the mother. The information on the presence or absence of colorectal cancer in the family was updated in 1996 when a question on age of first diagnosis was included; the question referred to any parent, sibling and an additional sibling separately and the same categories were used as in 1986. Later questionnaires did not contain a question on family history of colorectal cancer. A report of colorectal cancer in first-degree relatives appears to be reliable.12, 13 Nonetheless, we excluded 250 men (35 of whom were cases) who reported to have a family history of colorectal cancer at only one of the questionnaires but who did not report so on at least two subsequent questionnaires.
The remaining 28,870 men were classified as to whether at least one of their first-degree relatives was known with colorectal cancer and according to age of diagnosis of the youngest affected first-degree relative. For this classification, we used all available information up to 1996 because a family history of colorectal cancer can be regarded as a surrogate or indicator of inherited genetic susceptibility rather than a time-varying factor.
Assessment of other exposure factors
The questionnaires, which have been described in detail elsewhere,14 requested information on age, race, height, weight, physical activity, use of aspirin, smoking history and habits, alcohol consumption and whether the men underwent either colonoscopy or sigmoidoscopy in the past 2 years. The 1996 questionnaire included questions on the numbers of biological brothers and sisters (0, 1, 2, 3, 4, ≥5). The semi-quantitative food frequency questionnaires, which included about 130 items and an open-ended question for unlisted foods, covered more than 90% of the major nutrient intake of participants and inquired after vitamin and mineral supplements.14
Derived nutrients, except alcohol, were adjusted for total energy intake using the residual method.15 The average intake up to the date of diagnosis for cases and the date of last endoscopy for adenoma-free men was calculated to best represent long-term exposure and reduce within-person variation.16 Subsequently, men were grouped into categories according to the exposure factors of interest. Four adenoma patients and 26 controls did not complete questions on covariates that were taken into account in the final models and were excluded, which resulted in a final study population of 28,840 men (including 3,881 adenoma patients).
To plot the prevalence of adenomas against age on a logarithmic scale, we determined whether an adenoma was present at the first endoscopy that was registered after study entry, and calculated the prevalence across strata of age. Linear regression, weighted by the inverse of the variance of the estimated proportion, was used to estimate the slopes of the resulting curves. In sub-analyses, we excluded the men who reported having undergone endoscopy before entering the study and those who underwent their first endoscopy for reasons other than routine screening.
Multinomial logistic regression was used to compare the distributions of a family history of colorectal cancer among patients having advanced adenoma and patients having nonadvanced adenoma to evaluate etiological heterogeneity (case–case analyses). Within the same multinomial logistic regression model,17 we compared both categories of patients with the people who did not report adenomas during the follow-up period. As a sensitivity analysis, we checked whether the associations remained similar when studying the group with single adenoma only. Whether the association between family history and advanced vs. nonadvanced adenomas depended on the location of the adenoma was also explored in the group of men with adenoma at only 1 location. We did so by fitting a model containing 2 indicator variables for location, the family history indicator variable and 2 cross-product terms of each of the indicator variables for location and family history. The p-value for evaluating heterogeneity of the odds ratios was obtained by comparing a model with the two cross-product terms and a model without any such term using a likelihood ratio test.
Multinomial logistic regression was also used to compare the distributions of a family history of colorectal cancer among patients with multiple and single distally located adenomas, and men who did not report any adenomas. Data from all adenoma patients were used when studying the association between the number of adenomas in the distal colon or rectum, but data from patients with at least 1 proximal adenoma were used when studying the association between a family history of colorectal cancer and the number of proximal adenomas because patients with at least 1 proximal adenoma must have undergone colonoscopy, which is a requirement for the diagnosis of proximal adenomas. However, the latter was not the case for the entire study population.
The main models included total energy intake (quintiles), age (in 5-year age groups), history of endoscopy prior to study entry (yes/no), routine screening vs. other indications for any endoscopy, aspirin use (>2 times/week vs. ≤2 times/week), use of multivitamins (current, former, never), smoking (never, quit ≤ 10 years ago, quit >10 years ago, current, missing), consumption of red meat (quintiles), alcohol (0, 0 to <10, 10 to <20, 20 to <30, ≥30 g/d), intake of folate (quintiles), calcium (quintiles), BMI (quintiles), and physical activity (quintiles), in addition to a family history of colorectal cancer. As these presumed risk factors did not appreciably affect the risk estimates corresponding to a family history of colorectal cancer, we did not check whether inclusion of additional dietary and lifestyle factors affected the risk estimates. Because the number of siblings did not importantly affect the risk estimates and because clinical strategies are unlikely to depend on the number of siblings by itself, we have only adjusted the models referring to affected siblings for the number of siblings (0, 1, 2, 3, 4, ≥5).
Additional sensitivity analyses evaluated whether the case–case associations depended on indication of endoscopy (screening vs. complaints), race and age (continuous term) by including cross-products term in the model and evaluating them using likelihood ratio tests.
All reported p-values are two-sided. p-values < 0.05 were considered statistically significant. The analyses were performed using SAS version 9.1 (SAS Institute, Cary, NC).
Description of the study population
Up to 1986, 9.0% of the men reported a family history of colorectal cancer. This figure increased to 11.3% in 1990, 12.9% in 1992 and 14.9% in 1996, which most likely reflects ageing of the family members of the men.
One type of relative was affected for 92.3% of the 4,286 men with a family history of colorectal cancer (father: n = 1,610; mother: n = 1,353; sibling: n = 663; at least one of the parents, but unknown who: n = 328), whereas 7.2% had 2 types of affected relatives (father + mother: n = 132; father + at least 1 sibling: n = 78; mother + at least 1 sibling : n = 90; at least one of the parents + at least 1 sibling: n = 10) and the mother, father and at least 1 sibling of 22 men (0.5%) were affected. Multiple siblings of 42 men were diagnosed with colorectal cancer.
Table I illustrates that dietary and lifestyle characteristics were comparable for men with and without a family history of colorectal cancer. Men with a family history of colorectal cancer, however, were more likely to have undergone endoscopy before study entry, but they were less likely to have been examined for routine screening. A similar pattern was observed for the characteristics at study entry (not shown).
Table I. Age-Adjusted Characteristics of the Men of the Health Professionals Follow-Up Study According to the Presence or Absence of a Family History of Colorectal Cancer1
Family history of colorectal cancer, n = 4,286
No family history of colorectal cancer, n = 24,554
Updated variables are used for time-varying exposures (see method section). Mean values are presented. Missing values are excluded.
Met-h: metabolic equivalent task hours: the ratio of the metabolic rate during activity to the resting metabolic rate.18
Of the study participants, 13.5% had at least one adenoma during the follow-up. A total of 2,193 (71.3%) cases had only adenomas that were classified as tubular, 22 (0.7%) cases had only serrated adenomas, 633 (20.6%) had at least one tubulovillous adenoma but no villous adenoma or carcinoma in situ, 167 (5.4%) had at least 1 villous adenoma but no carcinoma in situ and 60 (2.0%) men had carcinoma in situ. The histopathology was unknown for the remaining 806 adenoma patients. A total of 1,203 (35.1%) adenomas were 10 mm in diameter or larger and 2,227 were smaller than 10 mm. Size was unknown for 451 adenoma patients and 148 of them also belonged to the group without information on histopathology. A total of 709 adenoma patients had adenomas in the rectum, 2,052 men had adenomas in the distal colon and 1,799 men had adenomas in the proximal colon. No information on location was available for 43 adenoma patients. A total of 622 men had multiple adenomas in the distal colon or rectum combined and 500 men were registered with multiple adenomas in the proximal colon.
Men with either tubulovillous or villous adenomas were more likely to have large adenomas than men with tubular adenomas (OR for tubulovillous vs. tubular = 8.04, 95% CI = 6.52–9.92; OR for villous vs. tubular = 10.0, 95% CI = 7.15–14.0). Men with adenomas in the distal colon were 2.29 (95% CI = 1.95–2.68) times more likely and men with adenomas in the rectum were 2.01 (95% CI = 1.64–2.46) times more likely to have advanced adenomas than men with proximal adenomas.
Men with multiple adenomas in the distal colon or rectum were 1.55 (95% CI = 1.26–1.91) times more likely to be diagnosed with advanced adenomas than patients with single adenomas, which was also the case when similar patients but without any proximal adenomas were studied (OR = 1.62, 95% CI = 1.27–2.07).
Family history and prevalence of adenomas
Adenomas occurred more frequently among men who underwent a first endoscopy at an older age than among men who were younger at first endoscopy (Fig. 1a). Adenomas were more common in men with a family history of colorectal cancer than in men without such a history in all age categories, but the relatively higher prevalence in men with a family history was notably visible in the younger age categories. Apparent straight lines could be drawn independently in this plot for men with and without a family history of colorectal cancer, which suggests that adenoma prevalence increases as a power function of age. The slope of the weighted regression line was 1.8 for men with a family history of colorectal cancer and 2.9 for men without (p = 0.039), which suggests that one of the rate-limiting steps in adenoma development has already occurred in men with a family history of colorectal cancer.19 Therefore, the association between family history and adenomas appears stronger in the younger age categories. However, this difference could at least partly be attributed to the youngest age group. Although the difference in slopes was still visible after exclusion of people who underwent an endoscopy prior to age 50, the difference in slopes was smaller and not longer statistically significant (slope among those with a family history of colorectal cancer: 1.4; slope among those without such a history: 1.9; p = 0.38). A similar plot was obtained when we restricted the analyses to men who underwent their first endoscopy because of routine screening (not shown).
The aforementioned findings were supported by analyses in which we took variation in modifiable risk factors into account. Men with a family history of colorectal cancer were 1.75 times more likely to get colorectal adenomas than men without an affected first-degree relative (95% CI = 1.60–1.91); this association was similar among men who underwent all registered endoscopies for routine screening (OR = 1.76, 95% CI = 1.60–1.94) and men who underwent at least one endoscopy for another reason (OR = 1.71, 95% CI = 1.41–2.08; pheterogeneity = 0.81). The associations were similar for men with Southern European (OR = 1.83, 95% CI = 1.53–2.17), Northern European (OR = 1.67, 95% CI = 1.51–1.85) or other ethnic background (OR = 2.38, 95% CI = 1.70–3.34; pheterogeneity = 0.63). As also visible in the figure, the strength of the association was stronger for men who were diagnosed at young age than for men who were diagnosed at older age (for those aged ≤55 years: OR = 2.37, 95% CI = 1.85–3.02; for those aged >55 years: OR = 1.67, 95% CI = 1.52–1.83; pheterogeneity = 0.048).
The overall risk was similar for individuals with an affected sibling (OR = 1.69, 95% CI = 1.42–2.02) and for individuals with an affected parent (OR = 1.65, 95% CI = 1.50–1.81). Men with multiple affected relatives (OR = 2.36, 95% CI = 1.84–3.04) were more likely to be diagnosed than men with only 1 affected relative (OR = 1.75, 95% CI = 1.60–1.92).
Family history of colorectal cancer and advanced and nonadvanced adenomas
Figure 1b shows that advanced adenomas were more frequently found than nonadvanced adenomas from age 60 onward in those with and without a family history of colorectal cancer, but the picture was less clear at younger ages. The slopes of the curves referring to advanced and nonadvanced adenomas were similar among men with and without a family history of colorectal cancer. In both groups, the proportion of men with advanced and with nonadvanced adenomas increased with age.
After adjustment for presumed risk factors, a family history of colorectal cancer was similarly associated with advanced and nonadvanced adenomas when the entire study population (Table II) as well as when the subgroup of men with only 1 adenoma was studied (advanced vs. nonadvanced: OR = 0.89, 95% CI = 0.71–1.13; advanced vs. not known with adenomas: OR = 1.42, 95% CI = 1.19–1.71; nonadvanced vs. not known with adenomas: OR = 1.59, 95% CI = 1.36–1.86). We conducted sensitivity analyses to check whether these associations remained similar when men having adenomas at one location only were studied. These analyses suggested different associations according to adenoma location (pheterogeneity = 0.023): a family history of colorectal cancer tended to be more strongly associated with nonadvanced than with advanced adenomas in the distal colon. This also applied to rectal adenomas, but the smaller number of cases resulted in a wide CI for the case–case analysis and an association with family history could only be detected for advanced rectal adenomas. Proximal advanced adenomas, however, were more strongly associated with a family history than were proximal nonadvanced adenomas. When we restricted the case subgroup to those with only 1 proximal adenoma (advanced: n = 206, nonadvanced: n = 480), however, this difference was less convincing (advanced vs. nonadvanced: OR = 1.18, 95% CI = 0.79–1.76; advanced vs. not known with adenomas: OR = 1.79, 95% CI = 1.28–2.49; nonadvanced vs. not known with adenomas: OR = 1.51, 95% CI = 1.20–1.91). Neither the main case–case analysis nor the sensitivity case–case analyses, which compared the associations of a family history of colorectal cancer with the occurrence of advanced and with the occurrence of nonadvanced adenomas, showed statistically significant differences between the associations.
Table II. Risk of Advanced and Nonadvanced Colorectal Adenomas According to Location or Strength and Type of Family History of Colorectal cancer (CRC)1
The multinomial logistic regression models were adjusted for age (in 5-yr age groups), history of endoscopy prior to study entry (yes/no), routine screening versus other indications for any endoscopy, aspirin use (>2 times/wk vs. ≤ 2 times/wk), use of multivitamins (current, former, never), smoking (never, quit ≤ 10 yr ago, quit>10 yr ago, current, missing) consumption of red meat (quintiles), alcohol (0, 0–<10, 10–<20, 20–<30, ≥30 g/d); intake of folate (quintiles), calcium (quintiles), BMI (quintiles), physical activity (quintiles), total energy intake (quintiles).
Because of missing values, not all totals add up to 1,496, 1,507 or 24,959.
Men with a family history of CRC / without family history of CRC.
Restricted to men who had only adenomas at one location. Similar conclusions were obtained when we focused on men who had at least one adenoma at the studied location.
This test was based on evaluating an ordinal variable for the age of diagnosis among cases only (4 = before age 50, 3 = age 50–59, 2 = age 60–69, 1 =≥ age 70).
Categories are not mutually exclusive. Estimates are mutually adjusted.
The difference between the associations with advanced and nonadvanced adenomas did not seem to depend on race (pheterogeneity = 0.65), having had screening endoscopy or not (pheterogeneity = 0.32), or age (pheterogeneity = 0.75). Among the men with adenomas, the distribution of adenoma size seemed comparable among those with (median = 6 mm, 25th percentile = 4 mm, 75th percentile = 10 mm) and without a family history of colorectal cancer (median = 7 mm, 25th percentile = 4 mm, 75th percentile = 10 mm), although a formal statistical test (the Wilcoxon two-sample test) suggested that men without a family history of colorectal cancer were more likely to have larger adenomas (p = 0.038). This tendency was also visible when we studied adenomas in the rectum, distal colon and proximal colon separately, but these differences were not statistically significant.
Family history of colorectal cancer and adenoma multiplicity
Figure 1c shows the distribution of single and multiple adenomas in the distal colon and rectum combined according to age groups and presence or absence of a family history of colorectal cancer. Across all age groups and both among men with and without a family history of colorectal cancer, single distally located adenomas were more prevalent than multiple distally located adenomas. The slope representing the prevalence of single distally located adenomas among men with a family history of colorectal cancer was almost horizontal (suggesting only 1 rate limiting step), whereas the slope for men without a family history of colorectal cancer was slightly positive, but this was largely due to the younger age groups. A similar pattern was observed regarding multiple distally located adenomas, although the slopes were positive and the parallelism more obvious.
The prevalence of multiple, but also single distally located adenomas was higher among men with a family history of colorectal cancer, which is also reflected in the odds ratios presented in Table III. The association with family history was stronger for multiple than for single distally located adenomas. The difference in associations between family history and multiple distally located adenomas, and family history and single distally located adenomas, increased in relation to the number of affected relatives. However, the strength of the association did not depend on the age of diagnosis of the family members. The difference between the associations with multiple and single distally located adenomas did not depend on race (pheterogeneity = 0.80), having had screening endoscopy or not (pheterogeneity = 0.67) and age (pheterogeneity = 0.21; see also Fig. 1c). The case–case analysis and the 2 case-control comparisons on family history of colorectal cancer and adenoma multiplicity were very similar when the distal colon and rectum were studied separately (estimates not shown). Associations in the same direction were seen among men only known with nonadvanced adenomas (multiple vs. single distally located: OR = 1.33, 95% CI = 0.90–1.96; multiple distal vs. no distally located adenomas: OR = 2.20, 95% CI = 1.56–3.10; single distally located vs. no distally located adenomas: OR = 1.66, 95% CI = 1.37–2.00) and men known with at least 1 advanced adenoma (multiple vs. single distally located: OR = 1.40, 95% CI = 1.02–1.91; multiple distally located vs. no distally located adenomas: OR = 1.90, 95% CI = 1.46–2.47; single distally located vs. no distally located adenomas: OR = 1.36, 95% CI = 1.13–1.63).
Table III. Risk of Multiple Distally Located Versus Single Distally Located Colorectal Adenomas According to Strength and Type of Family History of Colorectal Cancer (CRC)1
The association between family history and adenoma multiplicity was also visible when we studied the number of adenomas. Of the 1,985 men with one adenoma in the distal colon or rectum, 382 (19.2%) had a family history; this also applied to 96 of the 440 (21.8%) men with two distally located adenomas, 32 of the 124 (25.8%) men with three, 13 of the 33 (39.4%) men with 4 and 11 of the 25 (44.0%) of the men with 5 or more distally located adenomas.
A similar pattern was visible when the number of proximal adenomas was studied. Of the 1,293 men with 1 adenoma in the proximal colon, 275 (21.3%) had a family history; this applied to 84 of the 313 (26.8%) men with 2 adenomas in the proximal colon, 29 of the 115 (25.2%) men with 3, 13 of the 36 (36.1%) men with 4 and 11 of the 36 (30.6%) of the men with 5 or more proximal adenomas.
To our knowledge, our study is the first that systematically examined whether a family history of colorectal cancer influences adenoma multiplicity in a large number of asymptomatic and symptomatic men. Previous studies have reported on the prevalence of advanced adenomas in men with and without a family history of colorectal cancer, but these studies were small7–10 and only 2 of them studied these associations systematically.6, 11
We observed that a family history of colorectal cancer was more strongly associated with multiple distally located adenomas than with single distally located adenomas. A similar indication for proximal adenomas was found. A family history of colorectal cancer was not differently associated with risk of advanced and nonadvanced adenomas in the entire population of cases, although potential differences according to subsite were found. The data tended somewhat toward a stronger positive association with nonadvanced adenomas than with advanced adenomas in the distal colon and rectum, but toward a stronger association with advanced than with nonadvanced proximal adenomas.
By adjusting the associations for presumed modifiable risk factors for colorectal cancer, we largely excluded the explanation that shared environmental factors underlie the observed associations, which is in line with a study that compared incidence rates of colorectal cancer between siblings and spouses, and between parents and their offspring.20
Not all participants underwent colonoscopy, which may have led to some bias in the analyses on advanced and nonadvanced adenomas. However, since most polypectomies take place during colonoscopy as recommended,21 the impact of bias due to incomplete bowel examinations may be limited with regard to the case–case analyses. Nonetheless, some misclassification of case status has inevitably occurred. As for any adenoma study, some adenomas will have been missed at endoscopy, which are more likely to be small.22 The miss rate will probably not affect the case–case analyses importantly because endoscopists are likely to examine the colon thoroughly once an adenoma has been diagnosed. Provided endoscopists do not conduct the examinations more thoroughly when a patient with a family history of colorectal cancer presents, the effect of missed adenomas is likely to be largely diluted by the much larger number of truly adenoma-free men in the comparisons vs. men not known with adenomas.
Drawbacks of our study are that different physicians performed the endoscopies and that the classification of size and histopathological characteristics are based on judgment of different community pathologists. It can be argued that the classification of advanced adenomas should incorporate dysplasia as an important determinant of colorectal cancer risk,23 but we decided not to do so because the consistency of the classification into high-grade and low-grade dysplasia has been shown to be poor when different community pathologists are involved.24, 25 The fair agreement of classification of histopathological types24–26 between pathologists in combination with the strong association between size, histopathological characteristics and dysplasia27–29 further supports our classification. However, we did not categorize 23% of the adenomas as advanced or nonadvanced adenomas. As physicians may be less inclined to report observations deemed clinically unimportant than those deemed clinically important, we expect that the noncategorized adenomas are mostly nonadvanced adenomas. Reassuringly, our conclusions regarding the case–case analyses are robust given that the OR for advanced vs. nonadvanced adenomas was 0.93 when we treated the nonclassified adenomas as if they were nonadvanced adenomas and 1.05 when we treated them as advanced adenomas, which were both not statistically significant.
In line with previous studies, we observed that the younger a person was diagnosed with any type of adenoma, the stronger was the observed association with family history of colorectal cancer. A stronger association was also found among men with young affected family members; hence, the etiology of adenomas and cancer occurring early in life appears to be more strongly determined by heritable factors than for those occurring later. In particular adenomas at younger age may have occurred among men who belong to a family with heritable colorectal cancer syndromes, but we are confident that our findings apply to sporadic carcinogenesis as only a few patients were diagnosed with a large number of adenomas and because fewer than 3% were diagnosed before age 50, while the mean age at last endoscopy was 66 years old. Patients with hereditary colorectal cancer syndromes develop the disease at relatively young age, but colorectal cancer patients with a family history also tend to get the disease 10 years earlier than patients without such a history.30 Advanced and nonadvanced adenomas also occurred at a younger age among men with a family history of colorectal cancer in our study. This could point toward family members being more susceptible to the occurrence of mutations.
The most striking observation of this study was that patients with a family history of adenomas were more likely to have multiple adenomas at diagnosis, which our study design allowed to demonstrate most clearly for distally located adenomas. This corresponds with findings from a small study in which the number of aberrant crypt foci was higher in patients with a family history of sporadic colorectal cancer than in patients without,31 and a case-control study also suggested that people with affected relatives were more likely to have multiple adenomas.6 A similar indication was found in a study among 992 patients, but this finding has to be interpreted with caution as it is unclear whether the analysis was adjusted for age.32 Perhaps the same susceptibility genes33 that modify the severity of the familial adenomatous polyposis coli (FAP) syndrome or other polyposis syndromes may also determine the association between a family history and adenoma multiplicity in sporadic adenoma patients. Cancer genes that were differently expressed in macroscopically normal rectosigmoid mucosa in individuals with a sporadic family history of colon cancer compared with individuals without such history34 could also underlie such an association. Although we could not detect differences in the association of family history of colorectal cancer with adenoma multiplicity according to adenoma location, future studies should tease out the role of family history of colorectal cancer in determining adenoma location further as adenomas at different locations may be determined by different genes or even different mutations within the same gene. Such has already been found for hereditary syndromes: HNPCC, which is caused by mutations in mismatch repair genes, is predominantly associated with proximal tumors, whereas FAP, which is caused by mutations in the APC gene, is more frequently associated with distally located lesions.35 Phenotypic variations, including in location, have also been shown for FAP patients with different type of mutations in the APC gene.36 It is possible that multiplicity of so-called sporadic adenomas may also be caused by different mutational patterns that result in distinct manifestations of the disease, e.g., in adenomas occurring at different locations.
In spite of the potential etiological heterogeneity of sporadic adenoma multiplicity, the stronger association between family history of colorectal cancer and multiplicity than the one with advanced adenoma stage for at least distally located adenomas in our study suggests that the genetic components involved in sporadic colorectal carcinogenesis mostly drives the occurrence of key mutations rather than enhancing the growth signals in prevalent distally located adenomas in the gut. Indeed, in a cross-sectional study a family history of colorectal cancer was associated with an approximately 2-fold higher recurrence rate within 3 years, while no difference in distribution of family history could be detected between patients with and without adenomas at study entry.37 These observations suggest that genetic factors may play a more important role in the earlier stages of the adenoma–carcinoma sequence than in later stages. It remains to be determined if this also applies to proximal adenomas, as cancers in the proximal colon develop via different pathways than do distally located cancers.35
However, it cannot be precluded that heritable factors stimulate growth of distally located or proximal minuscule adenomas rather than enhance the occurrence of new lesions. The results from our study suggest that genetic factors may be less likely to stimulate adenoma growth of at least the majority of distally located adenomas importantly because we did not observe a stronger association with advanced than with nonadvanced adenomas in the distal colon and rectum; this observation was in line with data from a pooled analysis that compared adenomas according to degree of dysplasia.11 On the other hand, a study in which small adenomas were left in situ for up to 3 years, patients with a first-degree family member with sporadic colorectal cancer were more likely to have adenomas that showed net growth than patients without such a family history,5 though this was a small study. In addition, relatively small studies7–10 as well as a larger case-control study6 suggested that larger adenomas were more common among those with relatives with colorectal cancer. In our study, a less steep curve was visible among men with a family history of colorectal cancer diagnosed before age 55 in the graphs depicting adenoma prevalence, which supports the existence of a hereditary subgroup of adenomas that develops faster than the majority. In HNPCC kindred, the ratio of adenomas to cancer is smaller than in the general population, while colorectal cancer is common and occurs at an early age.38 This suggests that HNPCC adenomas, which are more likely to be proximally located,35 pass quicker through the adenoma–carcinoma sequence than do other adenomas, and this may be the case for other, yet unidentified adenoma subgroups. Likewise, a group of carcinomas may exist with similar properties, which may or may not originate from aggressive adenomas. Nonetheless, in the general population, in which distally located adenomas are more common, a family history of colorectal cancer seems to play a more important role in the early stages of adenoma development than in later stages. Considering that some adenomas might never turn into a more advanced lesion because they regress in size,39 we cannot conclude that all adenomas have the potential to develop into a carcinoma provided necessary growth conditions will be fulfilled. The selection process that determines the transition of nonadvanced to advanced distally located adenomas, however, does not seem to be substantially influenced by heritable factors strongly related to a family history of colorectal cancer.
In conclusion, our data suggest that adenoma multiplicity seems to have a hereditary basis in sporadic colorectal carcinogenesis, but we could not confirm a role of hereditary factors in adenoma advancement in at least the distal colon and rectum. In addition to confirming our results and aiming to identify the underlying genetic factors, future studies could assess the age distribution of the relatives to be able to construct a family risk index that allows studying underlying genetic susceptibility more precisely.
The HPFS is supported by NCI Research Grant CA 55075. Dr. Petra Wark's visit to the Harvard School of Public Health was made possible with support from the Dutch Cancer Society, and she was further supported by the Netherlands Organisation for Health Research and Development and Cancer Research UK. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Cancer Institute or the National Institutes of Health.