Many studies found that persons with a family history of colorectal cancer had an elevated risk for adenomas [4-10], although some failed to demonstrate this association [11, 12]. The only previous study of colorectal cancer among relatives of patients with adenomas that used a cohort analysis was part of the National Polyp Study. The risk for colorectal cancer among parents and siblings of patients with newly diagnosed adenomas participating in the National Polyp Study was found to be nearly 80% higher than the risk among asymptomatic healthy spouses of these patients [13].

We examined whether the risk for colorectal cancer is higher in first-degree relatives of patients with newly diagnosed adenomas than in relatives of endoscopically normal (no previous or newly diagnosed adenoma) controls from the same colonoscopy clinics. Characterization of this risk would provide guidance about whether and when screening should be offered to family members of patients with newly diagnosed adenomas.

Methods

Top Methods Results Discussion Author & Article Info References

Patients

Details of the study procedures and eligibility are described in detail elsewhere [14]. Briefly, the study sample consisted of patients who came to three large colonoscopy practices in New York City between April 1986 and March 1988. Eligible men and women were between 35 and 84 years of age; resided in New York, New Jersey, or Connecticut; spoke English or Spanish; and had undergone colonoscopic examination to at least the splenic flexure (>90% extended to the cecum). Persons who had had cancer, had known familial polyposis coli, or had inflammatory bowel disease were excluded. The colonoscopists completed data sheets indicating the reason for colonoscopy and the clinical findings at the time of colonoscopy. The study pathologist (Dr. Cecilia Fenoglio-Preiser, University of Cincinnati Medical School) reviewed the slides of all suspected neoplastic lesions.

A total of 2988 patients were evaluated, 2443 of whom (81.8%) met the eligibility criteria and were invited to participate in the study by a letter signed by their colonoscopists. Of those who were invited, 2001 (81.9%) agreed to participate in the study. The study participants were interviewed by telephone or by mail questionnaire. Among other data, detailed information on history of colorectal cancer in first-degree relatives (parents, siblings, and children) was collected. Data collectors were blinded to the patients' case–control status. No attempt was made to validate self-reported family history of colorectal cancer by direct interview of family members or by review of medical records.

Of 2001 patients who participated in the study, 303 had newly diagnosed adenoma, and 509 were endoscopically normal controls (no previous or newly diagnosed adenoma). These 812 patients were the baseline case and control groups (probands), respectively, for our study. Participants who did not provide complete information (such as age, cancer history, or vital status) on their first-degree relatives were excluded. Of the probands who provided adequate family history information, 272 were case-patients in whom colorectal adenomas at index colonoscopy were diagnosed for the first time and 462 controls had no history of adenomas and were normal at the index colonoscopy. Patients who had colonoscopy solely because of a family history of colorectal cancer (28 patients with adenoma and 100 controls) were excluded from this analysis. Therefore, our study describes the results for 244 patients with adenomas, 362 controls, and 3727 of their first-degree relatives.

Statistical Analysis

The primary analysis involved a comparison of case-patients who had newly diagnosed adenomas with controls by using a reconstructed cohort design: that is, we assessed the incidence of the outcome among family members assuming that case–control status was the exposure of interest [15]. Many studies examining family history have used this type of cohort analysis [13, 16, 17] because it is less susceptible to the bias in regular case–control analyses that results from the differential age distribution of family members and family size between case and control groups [18]. The Cox proportional-hazards regression model was used to compare the incidence of colorectal cancer among first-degree relatives of patients with adenoma and controls. These analyses were initially adjusted for sex and race of the family members. Because age of the family members (age at diagnosis for relatives who developed cancer and current age or age at death for other relatives) served as the time variable in the Cox model, any differences in the age distribution of family members between cases and controls were also accounted for in these analyses.

Because diagnostic procedures for colorectal cancer have changed over time, year of birth was included in the model to control for the effect of these changes. Furthermore, because familial risk may depend on the specific relationship between the proband and the relative [18], the type of relationship was also included in the models used to analyze all relatives together. However, adjustment for year of birth and type of relationship did not substantially change the effect estimates. To further adjust for proband and family characteristics, separate terms (keeping values the same for all members of a family) for the proband's age, years of education, smoking status, body mass index, and family membership were also included in the Cox model. None of these variables except age of the proband confounded the effect estimates substantially. Thus, the results adjusted for family members' sex and race and probands' age only are presented. The proportional hazards model assumption was examined by plotting the logarithm of the negative logarithm of the estimated product-limit survival function against the logarithm of time and was found to be satisfactory. The risk for other cancers was also examined as a negative control.

Results

Top Methods Results Discussion Author & Article Info References

Characteristics of the study groups are shown in Table 1. Patients with adenomas were more likely than those who had never had adenomas to be male, older, and white. Patients with adenomas were less likely to have had colonoscopy done for gastrointestinal symptoms (gastrointestinal bleeding, abdominal pain, and changes in bowel habits) [19] but were more likely to have had it done for other reasons (general screening; nonspecific gastrointestinal symptoms except abdominal pain, changes in bowel habits, or gastrointestinal bleeding; unexplained weight loss; and other nonspecific symptoms).

First-degree relatives of patients with newly diagnosed adenoma, estimated from the Cox model, had a relative risk for colorectal cancer of 1.74 (95% CI, 1.24 to 2.45) compared with relatives of controls (Table 2). No increased risk was found for other cancers among family members of patients with newly diagnosed adenoma. No heterogeneity was found in the relative risks calculated for probands with one adenoma compared with multiple adenomas, right-sided compared with left-sided adenomas, or small ( 5 mm) compared with large (>5 mm) adenomas (results not shown).

The effect of age at diagnosis of adenoma in probands on the risk for colorectal cancer among their family members is shown in Table 3. The risk for colorectal cancer among family members was inversely associated with the age at which adenomas were diagnosed in probands. Compared with family members of probands who were older than 60 years of age when adenoma was diagnosed, family members of probands who were 50 years of age or younger at diagnosis had more than four times the risk for colorectal cancer.

Because very few members of the offspring generation had colorectal cancer, the risk among children could not be examined separately. The risk for colorectal cancer was similar among parents (Table 4) and siblings (Table 5). For family members of patients with adenomas, we examined the effect of having an affected family member on the risk for colorectal cancer. Among parents, the risk was nearly three times higher if they had affected siblings (Table 4). Similarly, among siblings, the risk was more than five times higher if they had affected parents (Table 5).

Discussion

Top Methods Results Discussion Author & Article Info References

We found an elevated risk for colorectal cancer among first-degree relatives of patients with colorectal adenomas. This relative risk (1.74) is similar to that reported by Winawer and colleagues (1.78) [13] for family members of patients with newly diagnosed adenomas but is somewhat lower than the estimates in previous studies that used a case–control comparison [4-8]. It has been shown that the relative risk in relation to family history estimated from case–control comparisons is generally an overestimate [18].

Several previous studies of colorectal cancer among family members of patients with colorectal cancer found relative risks from 1.5 to 6; again, estimates were somewhat lower in cohort studies [20, 21] than in case–control studies [8, 12, 22, 23]. Elevated risk for colorectal cancer among family members of patients with colorectal cancer and patients with colorectal adenoma confirms the common hereditary predisposition of colorectal adenomas and cancers proposed in the Utah studies [2, 3].

We, as did Winawer and colleagues [13], found that family members of patients who were younger when adenomas were diagnosed were at higher risk for colorectal cancer than were patients who were older at diagnosis. This risk was particularly high among family members of patients who were 50 years of age or younger when adenoma was diagnosed. We also found a higher risk for colorectal cancer among parents or siblings of patients with adenomas who had other family members with colorectal cancer than among those who did not have other affected family members. The risk for colorectal cancer was similar among parents and siblings of patients with adenomas, a finding that Winawer and colleagues could not examine directly because they compared the incidence among relatives of adenoma patients with the incidence among spouses of adenoma patients [13]. Other studies of colorectal cancer among family members of patients with colorectal cancer, however, found a higher risk among siblings than among parents [7, 8, 22, 24, 25].

We did not find an increased risk for colorectal cancer among family members of patients with right-sided adenomas compared with family members of patients with left-sided adenomas. This finding was described among family members of patients with colorectal cancer [22, 26]. In addition, we did not find a higher risk among family members of patients with larger adenomas compared with family members of patients with smaller adenomas; this finding was presented in a recent case–control study [27].

Information about family history of colorectal cancer was provided by the probands. The validity of the self-reports is of major concern for our results. Self-reported family history of colorectal cancer has been validated with medical records in an Australian study [28], in which self-reporting had a sensitivity of 87% and a specificity of 97% for indicating colorectal cancer among first-degree relatives. Moreover, patients with adenoma and controls were found to be similar in reporting their family members' cancer diagnosis [28]. All of the participants had symptoms and underwent colonoscopic procedures; therefore, they were less likely to differentially report cancers among family members. Furthermore, the risk for other cancer among family members examined in this study was closer to null, a finding that further supports the validity of self-reports in our study.

Like the patients in a previous colonoscopy-based study [13], patients in our study who underwent colonoscopy solely because of their family history were excluded to avoid referral bias. Because more controls than case-patients were excluded, inclusion of those patients in this analysis would have attenuated the effect estimates. This exclusion was based on an a priori decision by the authors because controls who do not have adenomas would be more likely to have been referred than patients with adenomas who were referred for reasons other than adenoma-related symptoms (such as family history). In fact, the prevalence of family history among study participants is similar to the prevalence observed among participants in previous colonoscopy-based studies [4, 13].

The underlying genetic basis of the shared familial risk for colorectal adenomas and cancers is not clear. We excluded persons with a family history of familial polyposis coli but not those with a family history of hereditary nonpolyposis colorectal cancer, which may account for up to 10% of colorectal cancers [29]. The mismatch-repair genes (hMSH2, hMLH1, hPMS1, and hPMS2) responsible for hereditary nonpolyposis colorectal cancer have been found in both familial [30] and sporadic colorectal cancer [31]. Replication errors, the phenotype of these genetic defects, have also been shown in adenomas, suggesting a possible role of mismatch-repair genes in the adenoma-adenocarcinoma sequence [32]. Future studies must examine these genetic defects as well as other molecular and genetic markers that may be predictive of the risk for colorectal neoplasms among family members of patients with adenoma.

To our knowledge, ours is the first study that used a cohort analysis to compare the risk for colorectal cancer among first-degree relatives of patients with adenomas with the risk among first-degree relatives of endoscopically normal controls. The only previous study in which cohort analysis was used [13] compared the risk in parents and siblings of patients who had adenomas with the risk in spouses of a sample of patients (normal or with adenoma or cancer) who were referred for colonoscopy. One potential problem of comparing the incidence of colorectal cancer among parents and siblings with that among spouses is that parents would have a different age distribution and would be from different generations than the spouses. Although the authors adjusted for these factors in their analyses, the possibility of bias caused by differences in age, generation, and type of relationship with probands was probably greater in their study than in ours. It may be argued that our inclusion of endoscopically normal controls who had undergone colonoscopy in the same clinics as the patients with adenomas may have resulted in control and case-patient groups that were very similar for many characteristics, including family history. However, it was reassuring that our study and that of Winawer and colleagues (studies that used two different types of controls) found almost identical risk estimates.

Although the risk that we found was modest, the fact that our results confirm the findings of a previous study reinforces the current guidelines for screening of family members of patients with colorectal adenomas. The guidelines suggest that first-degree relatives of patients with adenomas should undergo screening measures (yearly fecal occult blood test followed by colonoscopy if results of the fecal occult blood test are positive or double-contrast barium enema with flexible sigmoidoscopy) starting at 40 years of age [33]. Future studies must determine whether more intense screening measures, such as use of colonoscopy as the primary screening test [34], are beneficial in preventing colorectal cancer in this high-risk population and examine the cost-effectiveness of these measures. Future studies also must examine the magnitude of risk associated with various characteristics of patients with adenomas to identify family members of those patients who are at the highest risk, such as family members of patients who are 50 years of age or younger when adenoma is diagnosed or who have living affected family members [13].

In conclusion, we found an elevated risk for colorectal cancer among first-degree relatives of patients with colorectal adenomas. This risk was higher for relatives of patients with early-onset adenomas and for relatives of patients with affected family members. Our findings, in conjunction with findings from previous studies, support a common hereditary or familial predisposition to colorectal adenomas and colorectal cancers and have implications for current colorectal cancer screening guidelines.

Dr. Neugut: Division of Oncology, Columbia-Presbyterian Medical Center, 630 West 168th Street, New York, NY 10032.

Drs. Forde and Treat: Department of Surgery, Columbia-Presbyterian Medical Center, 630 West 168th Street, New York, NY 10032.

Dr. Waye: 650 Park Avenue, New York, NY 10021.