MATERIALS AND METHODS

Subjects

Subjects included in this study were from the Cancer Family Clinic at Karolinska Hospital from 1990 to 1999. A medical history was given by the index patient and all diagnoses in the family were confirmed by medical records, pathological reports, or in very few cases death certificates. The general screening interval was every two years. After polypectomy of at least one adenoma, a new colonoscopy was performed the following year. Data from the colonoscopies were recorded anonymously in Stat View 5.0.1. Families were classified according to family type (HNPCC, HCC, or TCR), or one close relative (OCR) using information available from clinical records (table 1). The information used for classifying families included family history and, if available, data on MSI tests in tumours and mutation screening in mismatch repair genes in affected members. All individuals were divided into risk groups as follows. Tested carriers in HNPCC families (risk group 1) were considered to have a 70% lifetime risk.¹² According to the rules applied to mendelian inheritance, untested first generation members at risk in a HNPCC family have a 35% lifetime risk (risk group 2). Their children who sometimes were under surveillance have a 17% lifetime risk (risk group 3). Subjects in HNPCC families who were under surveillance before testing but tested negative for mutation at the time of the study (risk group 4) were assumed to have the same risk (5%) as the normal population. Obvious gene carriers in families with HCC (risk group 5) have empiric risk values similar to HNPCC families (70%).^1,11,13 First generation (risk group 6) family members accordingly have a lifetime risk of 35% and the second generation (risk group 7) at risk in these families has an estimated lifetime risk of 17%. TCR subjects with a family history of colorectal cancer (risk group 8) have an empirical risk of 10–20%.^1,11,13 OCR subjects with one relative with early age of onset (risk group 9) have a lifetime empiric risk of 20–40%.^1,11,13

A reference group based on three published forensic autopsy studies was used for estimates of adenoma prevalence in the normal population.^14–16 In reference group 1, 185 men and 118 women underwent forensic autopsy. Colonic biopsies were examined for prevalence, size, and degree of dysplasia of identified adenomas.¹⁴ Reference group 2 consisted of 198 men and 167 women who underwent autopsy/forensic autopsy. The study included prevalence, type, and location of adenomas.¹⁵ Reference group 3 comprised autopsies in 370 women and 310 men from areas with various incidences of colorectal cancer. Prevalence of adenomas, size, degree of dysplasia, and site of the adenomas were reported.¹⁶ In a comparison of prevalence in the study group versus the reference population, only individuals less than 54 years of age were included in the analysis. The reason was an uneven distribution of subjects in different age cohorts; in the reference group 75% of the cohort were older than 54 years while only 33% were over 54 years in the study group. Also, it was considered relevant to study the prevalence of adenomas at an early age as this is typical of a predisposition to cancer. Thus in the comparison between the study group and the general population, 338 subjects from the reference group and 204 from the study group were used.

Statistical methods

Differences in prevalence of adenomas between the three family types and the reference population were calculated as relative risks (RR). Differences between the family types were tested by χ² test. Logistic regression was used to test the influence of sex with respect to the risk of developing an adenoma. The number of adenomas among individual subjects with adenomas were compared using the hazard ratio. Age at first adenoma was analysed by ANOVA with two independent factors, sex and family type. The Tukey post hoc test was then used for pairwise comparisons of family types and age at onset of adenomas. Differences between the family types regarding size and degree of dysplasia were compared using the χ² test. Differences in localisation of adenomas and carcinomas in the bowel were analysed using χ² statistics, and corresponding 95% confidence intervals for differences in proportions were calculated.

Genetic testing

MSI tests and genetic testing were performed either as part of previous studies or as part of clinical handling and counselling after 1997, and were not part of this study. The MSI test used established methods and criteria.^17,18 Methods used for mutation screening were denaturant gradient gel electrophoresis, protein truncation test, Southern blot, and direct sequencing.¹⁹

RESULTS

In total, 304 subjects underwent 765 colonoscopies (table 2). Ten colorectal cancers were found in nine individuals before recruitment into the surveillance programme. These tumours have been included in table 2 to obtain correct values for prevalence and mean age at onset of adenomas or cancer in the different family types. History of previous adenomas was unknown, and most individuals had their first colonoscopy through this programme. Four individuals had colorectal cancer detected at their first colonoscopy, and two individuals developed a metachronous colorectal cancer during surveillance.

Prevalence and number of adenomas

The RR value of developing an adenoma before the age of 54 years in all risk groups (except tested non-carriers) was 2.6 compared with the general population (p=0.001) (table 3). Sex had no influence on these results. RR was even higher (4.5) for tested gene carriers in HNPCC families and also statistically significantly higher for the other two family types (table 3). No difference was demonstrated (p>0.5) between the three family types for adenoma prevalence but a difference in the number of adenomas between family types was observed (fig 1). Individuals with adenomas from TCR families had more adenomas (3.8) than HCC subjects (2.4) and HNPCC families (1.2). The hazard ratio between TCR and HNPCC was 3.3 (CI 2.2–4.6) and between HCC and HNPCC 2.0 (CI 1.3–2.9). The hazard ratio between TCR and HCC was 1.6 (CI 1.1–2.18).

Cancer risk in adenomas

The prevalence in each risk group varied in relation to the estimated cancer risk in the various risk groups in HNPCC and HCC (table 2). The high prevalence of adenomas in the relatively low risk TCR families was unexpected. To obtain a relative value of cancer risk which could be used for comparisons between different risk groups, prevalence was related to the estimated cancer risk (table 4). The risk values (prevalence/estimated cancer risk) were highest in HNPCC, and lowest in TCR and the normal population. To obtain a relative cancer risk in each adenoma, the risk values were divided by the number of adenomas per individual (from fig 1) (table 4). Mean age of the families was similar (table 2), and the risk of cancer was the same when only subjects less than 54 years of age were compared (data not shown).

Localisation of adenomas and carcinomas

Adenomas were located throughout the colon and rectum, as depicted in fig 2. While adenomas were evenly distributed in HNPCC, HCC, and in the general population, TCR adenomas seemed to be mostly proximal (fig 2).

To determine if there was a difference in cancer risk depending on location, a comparison was made between adenoma location and carcinoma location in the different family types (fig 2). Localisation of cancers among the total number of family relatives was obtained from investigation of all 111 families in the study. In total, there were 81 HNPCC cancers, 52 HCC cancers, and 51 TCR cancers. To determine the localisation of sporadic cancer in the general population, three published reports^20–22 were used (fig 2). There was a clear difference in percentage of distal outcome between adenomas and carcinomas in families with TCR (p<0.001) but also in HCC (p<0.05) and the normal population (p<0.001) (fig 2).

Mean age at first observed adenoma

Mean age at identification of the first adenoma was 43 years in HNPCC, 50 years in HCC, and 52 years in TCR (fig 3). There was a statistically significant difference between age at first adenoma in HNPCC and TCR (p=0.006). The difference between HNPCC and HCC was as expected (p=0.02) as this was defined by the inclusion criteria.

Sex and age

Overall, there was a difference (p<0.05) between the prevalence of adenomas in men (34%) and women (24%) (tables 2, 3). Men and women showed a systematic difference in mean age at first adenoma (fig 3). There was no interaction effects; the sex difference was assumed to be constant over family type and the difference between family types was assumed to be constant over sex.

Histopathology

In total, 152 tubular adenomas, 18 tubulovillous adenomas, two villous adenomas, 18 serrated adenomas, and five unclassified adenomas were removed. The vast majority (88%) were less than 5 mm. In four of the adenomas estimation of size was not done. Size and degree of dysplasia in adenomas from the second or later colonoscopies were compared. Thirteen adenomas were >5 mm in size but only three showed a high degree of dysplasia (table 5). In total, eight subjects had 11 adenomas with high dysplasia, and six of those eight subjects were from HNPCC families (table 6). There was a statistically significant difference between the number of adenomas with high degree dysplasia in the HNPCC compared with the other two family types (χ²=6.7, p<0.01).

DISCUSSION

In the following discussion of the results, we will use the RR of adenomas as an estimate of tumour initiation rate in individuals. When the actual cancer risk in an individual is higher than the increase in initiation rate, this is considered to depend on an increased tumour progression rate in adenomas.

A higher initiation rate in all risk groups

There is much discussion as to whether the prevalence and frequency of adenomas in individuals with an increased risk of colorectal cancer are higher than in the general population.^23–26 In this study, gene carriers in HNPCC had an RR of 4.5 compared with the general population of presenting with an adenoma before the age of 54 years. All risk groups combined had an RR of 2.6. The increased risk of adenomas in all groups compared with the general population indicates that in all families there is an increased initiation rate that explains, at least in part, the increased risk of colorectal cancer. The different numbers of adenomas in the family types suggests that the initiation rate is increased most in TCR families and increased least in HNPCC.

Autopsy studies to assess the number of adenomas in the general population are not optimal. However, the autopsy studies chosen for this study were specifically designed to give adenoma values representing the normal population. Furthermore, some authors found autopsies to be more reliable than colonoscopies in detecting adenomas <10 mm.²⁷ Other studies have shown a missrate of 15–27% for detecting adenomas <5 mm using colonoscopies.^28,29 However, as most of the study objects were included in a surveillance programme, a minute adenoma would show up at the next screening if missed at the previous one, thus giving correct values for prevalence in the study group.

A higher rate of progression in the high risk syndromes

The excess in cancer risk in adenomas in HNPCC and HCC compared with TCR and the general population in this study suggested that apart from increased initiation there is also an increased rate in tumour progression. As the genes for HCC are still unknown, we estimated the penetrance as equal to that of HNPCC. If this is an overestimation, the cancer risk in each adenoma in HCC is also overestimated and could be even lower. In particular, HNPCC subjects have an up to eightfold greater cancer risk in each adenoma compared with TCR subjects and the normal population (table 4). This fits well with an increased mutation rate in tumours because of deficient mismatch repair in this syndrome. The shift from proximal adenomas to distal carcinomas among the latter two groups and also in the general population suggests that there is a difference in cancer risk in adenomas depending on the location in the colorectum (fig 2). Thus distal adenomas seem to be associated with a much higher risk of cancer in all patients except HNPCC where the risk of cancer seems equally high in proximal and distal adenomas.

Differences in mean age between syndromes

Mean age at first observed adenoma was 44 years in HNPCC which is in accordance with previous studies (fig 1). The higher penetrance for men in HNPCC found in this study (table 3) confirmed previous reports.³⁰ A higher penetrance for men was also suggested in HNPCC and HCC in this study, while TCR subjects were not informative in this respect depending on different mean ages among men and women.

In summary, there was a clear overrepresentation of adenomas in all three family types. This seems to justify regular colonoscopy surveillance for prevention in these patients. The data support a proposed model (table 7) with increased rates of both initiation and tumour progression in HNPCC. In HNPCC it appears that there is a rapid transformation from adenoma to carcinoma as there was a higher grade of dysplasia in HNPCC adenomas unrelated to size. HNPCC also displayed the highest risk of cancer in each adenoma, regardless of location (table 4). In HCC families, an increased initiation rate as well as an increased progression rate were also found although the cancer risk in each adenoma was lower in proximal adenomas and the progression from adenoma to carcinoma is likely to be slower than in HNPCC. TCR families seem to have the highest increase in initiation rate as they had the highest number of adenomas. In TCR there seems to be a low cancer risk in proximal adenomas and a relatively higher cancer risk in distal adenomas. The differences between the three family types (HNPCC, HCC, and TCR) with regard to age at onset, prevalence, location, size and degree of dysplasia, and cancer risk of adenomas have been used to revise surveillance protocols in Sweden. The guidelines from the Swedish National Oncogenetic Counsil now recommend the following: in HNPCC, regular colonoscopy every 1–2 years; in HCC, regular colonoscopy every 3–5 years; and in TCR, colonoscopy (or alternating colonoscopy/sigmoidoscopy) every 3–5 years.