Altogether 1678 individuals (1055 women) had colonoscopies during the study period (). The median age at the initial colonoscopy was 41. Individuals from group 4 (fulfilling the Amsterdam criteria) and from group 1 (families with one case diagnosed under age 45, and no other cases) were on average five to eight years younger than from groups 2 and 3. Ninety seven per cent of examinations reached the caecum. The others were either repeated or the patient was given a barium enema.
Characteristics of individuals undergoing surveillance at initial colonoscopy
presents the results of the initial colonoscopies: three quarters were normal, and 9% found only metaplastic polyps. The likelihood of finding an adenoma or cancer increased with age (P < 0.0001). Advanced neoplasia was seen in 3.8% overall, increasing from 2.0% in people younger than 35 to 14.9% in those aged 65 and older (). Cancer was found in five individuals (four from group 4—Amsterdam, and one from group 3—three or more affected individuals, but no cases diagnosed under age 50). The adjusted proportion with advanced neoplasia was highest in group 4 and lowest in individuals from group 1 (χ23 = 16.7, P = 0.0008, ). Individuals from group 1 were also least likely to have simple adenomas (7.8% compared with 12.8-16.5% in the other groups).
Colonoscopic findings (most neoplastic lesion for each individual) at initial surveillance colonoscopy by family risk group and age. Values are numbers (percentages) of patients
Fig 1 Advanced neoplasia and age at initial colonoscopy. The proportion at each age was calculated by using a locally linear smoother.14 This is a more sophisticated version of a running mean
During the study, 1143 individuals (from 740 families) had at least two colonoscopies: 652 had three or more. Altogether 8865 person years elapsed between the initial and the final colonoscopy (3020 in group 4 and 5845 in groups 1-3). The sex and age distribution at first colonoscopy in individuals who had two or more colonoscopies was similar to that in all 1678 individuals, but individuals with strong family histories were more likely to have been rescreened. The median number of years between successive colonoscopies was 3.3 in group 4, 4.6 in group 3, 5.1 in group 2, and 5.1 in group 1.
Incidence of neoplasia on follow-up colonoscopy
When adjusted for age and sex, adenomas were seen on follow-up (colonoscopies after the initial examination) in 26% of group 4, 25% of group 3, 21% of group 2, and 13% of group 1. The adjusted proportions of high risk adenomas and cancer were both greatest in group 4—Amsterdam criteria (5.0% and 1.0%, respectively) compared with 1.7% and 0.1% in groups 1-3—moderate risk (P = 0.005 and P = 0.048, respectively). Advanced neoplasia below the age of 50 on follow-up was most common in group 4 (4.6%). It was 0.5% in individuals from group 3, 0.4% in those from group 2, and 2.2% in those from group 1 (compared with group 4; P = 0.03, P = 0.014, P = 0.32, respectively).
Follow-up findings in individuals at moderate risk were related to initial findings (). Advanced neoplasia on follow-up was most common (12%) in individuals with advanced neoplasia on initial colonoscopy. Seven of the 12 individuals with multiple adenomas on initial colonoscopy had an additional adenoma on follow-up, but none had advanced neoplasia. The incidence of advanced neoplasia during follow-up in people without advanced neoplasia initially depends on family history (P = 0.002) and increases (P = 0.044) with age (). Incidence of advanced neoplasia in people with advanced neoplasia initially was high: 56 per 1000 years in individuals with hereditary non-polyposis colorectal cancer and 28 per 1000 years in individuals with moderate risk ().
Most advanced neoplastic lesion on follow-up surveillance colonoscopy in individuals at moderate familial risk (groups 1-3). Values are numbers (percentages) of patients
Rates of advanced neoplasia (absolute number/person years of follow up) on follow-up surveillance colonoscopy, per 1000 person years
Cancer incidence and mortality
In addition to six cancers detected at initial colonoscopy, 11 subsequent cancers occurred (eight were detected on surveillance), eight of these in individuals with hereditary non-polyposis colorectal cancer. In three cases, the colonoscopic interval was over five years. Six individuals died within 36 months of diagnosis; the other five patients were alive on 31 December 2002 (3.5-15 years after diagnosis).
Analysis of all cause mortality shows that death certification is complete up to December 2002. By that time, five patients had died from colorectal cancer during 13 347 years of follow-up. The death rate due to colorectal cancer in individuals at moderate risk is close to what one would have expected in the general population, but for group 4 (Amsterdam citeria) the observed mortality is nearly five times greater than in the general population. Despite these discouraging observations, our best estimates of the underlying risk in our cohort yields significant reduction in mortality: 81% in moderate-risk and 72% in group 4 ().
Observed number of cases of and deaths from colorectal cancer with expected numbers using the best estimate (plausible range) of risk based on family history
Analysis of the incidence of all cancers (excluding colorectal and endometrial cancer) shows that follow-up was complete up to December 2001. Fifty five cancers were found compared with 48.2 expected, with no decrease in the rate for 2001 compared with 2000.
Excluding cancers detected on the initial colonoscopy (prevalent cancers), eight colorectal cancers were diagnosed during active surveillance in individuals from group 4. That is nine times greater than expected from a “normal” population. By contrast, the one cancer in the group at moderate risk compares to 2.3 expected in a “normal” population. However, the eight cancers in group 4 are still fewer than half the number expected when we use our best estimate of the cancer rates in unscreened relatives of individuals with suspected hereditary non-polyposis colorectal cancer (P = 0.03, ).
shows the cumulative incidence curve as a function of time since first colonoscopy together with the expected curve based on the best estimates. The prevalent cancers detected during screening cause a jump in the observed cumulative incidence at time zero. It is 0.5 years before the curves cross for the moderate risk group and three years for group 4.
Observed cumulative incidence curve as a function of time (in years) since first colonoscopy together with the expected curve based on the best estimates
Analysis including all cancers diagnosed up to 31 December 2003 in individuals first screened before 31 December 2000 yields similar results to the analysis excluding prevalent cancers. Both analyses show a reduction in the incidence of cancer compared with our best estimate of the underlying risk. However, particularly in individuals at moderate risk, the inclusion of prevalent cancers reduces the magnitude of the benefit of colonoscopic surveillance on the incidence of cancer.