Complete colonoscopy rarely misses cancer

Complete colonoscopy rarely misses cancer Hooi C. Ee, MBBS, PhD, FRACP, James B. Semmens, MSc, PhD, Neville E. Hoffman, MBBS, MD, PhD, FRACP, for the Perth Teaching Hospitals Endoscopy Group Perth, Australia

Background: The assumption that colonoscopy is highly accurate for detecting colorectal cancer was tested by identifying cancer subsequent to colonoscopy in 2 cohorts of patients in which colonoscopy was reported as normal. Methods: A multicenter endoscopy database was used to identify all reportedly normal colonoscopies. One cohort was assessed 5 years after colonoscopy with the use of a populationbased health services-linked database to link patient morbidity, cancer, and mortality data. The second cohort was assessed by identifying patients who had cancer on repeat colonoscopy. Results: Of 1047 patients with normal colonoscopies followed for 5 years or until death if earlier, 5 cancers (0.5%) were detected. This rate was not significantly different from that predicted by Australian statistics (risk = 1.0%, p > 0.1), but significantly lower compared with that for all patients presenting for colonoscopy during the study period (risk = 5.2%, p < 0.001). In another cohort of 8486 patients with reportedly normal colonoscopies, 496 patients underwent repeat colonoscopies during an average follow-up of 3.1 years; cancer was diagnosed at the subsequent procedure in 3 patients (0.6%). Conclusions: The high accuracy of colonoscopy is demonstrated by the low risk of harboring an advanced neoplastic lesion after a normal examination. (Gastrointest Endosc 2002;55:167-71.)

Colonoscopy is regarded as the standard for the diagnosis of colorectal mucosal disease. One of its most important uses is the diagnosis of neoplasia at all stages, from the small benign polyp to the large bleeding and/or obstructing carcinoma. Studies have shown that 4% to 5% of cancers may be missed on a single colonoscopic examination,1,2 comparable with the miss rate for adenomas of 1-cm diameter.3 However, smaller polyps are missed more frequently than larger ones.4,5 Colonoscopy also allows removal of adenomatous polyps, the precursor of invasive carcinoma. This is substantiated by studies showing that colonoscopic polypectomy reduces the incidence of cancer.6-9 In these studies, however, it is unclear how many of the cancers discovered at follow-up colonoscopy were actually lesions missed at the initial examination. Colorectal cancer is a leading cause of cancer death in Westernized societies, and screening proReceived November 29, 2000. For revision April 11, 2001. Accepted October 23, 2001. From the Department of Gastroenterology, Sir Charles Gairdner Hospital, Perth, Western Australia, and the Centre for Health Services Research, Department of Public Health, The University of Western Australia, Perth, Western Australia. Presented in part as a poster at Australian Gastroenterology Week, October 18-21, 2000, Hobart, Australia (J Gastroenterol Hepatol 2000;15:J62). Reprint requests: Hooi C. Ee, MD, Department of Gastroenterology, Sir Charles Gairdner Hospital, Nedlands, Perth, Western Australia 6009. Copyright © 2002 by the American Society for Gastrointestinal Endoscopy 0016-5107/2002/$35.00 + 0 37/1/121190 doi:10.1067/mge.2002.121190 VOLUME 55, NO. 2, 2002

grams have been shown to reduce mortality from this disease.10,11 A consequence of screening for colorectal neoplasia will be a substantial increase in the number of colonoscopies performed. Because colonoscopy is perceived to be highly accurate, a normal examination reassures both doctor and patient that presenting symptoms or a positive screening test are not caused by bowel cancer. To assess whether this strong reliance on colonoscopy is appropriate, patients with a reportedly normal complete colonoscopic examination were followed to determine the number of patients discovered subsequently to have colorectal cancer. Patients were included in this study only if the index colonoscopy was complete to the cecum or ileum and was reported as normal. PATIENTS AND METHODS Two specific databases were used to investigate 2 sets of patient cohort data. A multicenter endoscopic database (Endoscopic Database) was established to record data for endoscopic procedures performed at the 3 metropolitan teaching general hospitals in Perth, Western Australia: Sir Charles Gairdner Hospital (from July 1989), Royal Perth Hospital (from March 1992), and Fremantle Hospital (from November 1994). All procedures were performed by either gastroenterologists or gastroenterology trainees. In addition, the population-based Western Australian Health Services Research Linked Database (WA Linked Database)12,13 was used to ascertain registrations for colorectal cancer from the Cancer Registry for the cohort of patients identified using the Endoscopic Database. Patient records were also linked to deaths recorded in the State Death Registry. GASTROINTESTINAL ENDOSCOPY

167

H Ee, J Semmens, N Hoffman, et al.

Long-term follow-up of normal complete colonoscopy

Table 1. Indications for colonoscopy for patient groups

Indication

Cohort 1 % (n)

Bleeding 22 Anemia 16 Diarrhea 16 Pain 15 Follow-up cancer 3 Follow-up polyp 7 Family history colon cancer 8 Radiograph abnormality 5 Other† 8 Total 100

(245) (177) (177) (163) (33) (73) (86) (60) (84) (1098)

All Cohort 2 colonoscopies* % (n) % (n) 23 14 7 11 6 12 19 2 6 100

(115) (68) (36) (53) (32) (59) (95) (9) (29) (496)

22 14 10 11 3 7 5 1 26 99‡

(692) (425) (324) (341) (100) (219) (162) (35) (823) (3121)

Cohort 1 consisted of all patients in the Endoscopic Database who had a complete colonoscopy at which no mucosal abnormality was detected for the period July 1989 to February 1993. Cohort 2 consisted of all patients in the Endoscopic Database who had a normal colonoscopy followed by a subsequent colonoscopy for the period July 1989 to February 1999. *All colonoscopies performed at Sir Charles Gairdner Hospital for the period July 1989 to February 1993. †Includes change in bowel habit and constipation. ‡Percentages do not add up to 100 because of rounding off.

The first cohort consisted of all patients in the Endoscopic Database who had a complete colonoscopy to cecum or ileum with no mucosal abnormality detected during the period July 1989 to February 1993. Only patients in whom the examination was complete to the cecum (determined by identification of ileocecal valve, appendiceal orifice, convergence of taeniae coli, and transillumination in right iliac fossa) or ileum were chosen, inasmuch as a proportion of missed cancers can be attributed to incomplete colonoscopy.1 The cut-off date of February 28, 1993 was chosen to allow 5-year follow-up analysis. This follow-up duration was chosen to ensure that all missed cancers would have eventually presented as advanced disease. The WA Linked Database was used to link patient data in the Western Australian Cancer Registry, Western Australian Hospitals Morbidity Database System, and State Death Registry to identify those patients in whom a diagnosis of colorectal cancer had been made or who had died during the 5 years subsequent to colonoscopy. These databases identified all cancers diagnosed in Western Australia; the only losses would be due to migration out of the state. Estimations from census data supplied by the Australian Bureau of Statistics show that only 1.7% of the population over 18 years of age leave Western Australia permanently each year. In the cohort situation, allowing for deaths and greater residential stability in an older age group with a significant medical history, it was estimated that less than 1% of the patients in this study would be lost to follow-up because of migration.14 The findings for the first cohort were compared with the expected number of colorectal cancers for a 5-year period in an age- and gender-matched population by using the Australian Association of Cancer Registries Database. The database provided colorectal cancer risk figures for 168

GASTROINTESTINAL ENDOSCOPY

specific ages for each gender. The expected number of cancers was determined by summation of the colorectal cancer risk for each individual patient, as specified by the database, for the duration of the follow-up period. This database is published by the Australian Institutes of Health and Welfare15 and is available on the Internet.16 The findings were also compared with the rate of cancer detected for all colonoscopies recorded in the Endoscopy Database at Sir Charles Gairdner Hospital over the same period. Statistical analyses of the risk of cancer in the first cohort compared with the estimated risk of cancer in the age- and gender-matched population, and also with the risk of cancer detection at all colonoscopies, was performed with the chi-square test. The second cohort consisted of all patients identified from the Endoscopic Database undergoing repeat colonoscopy who had previously undergone an index colonoscopy to the cecum or ileum for which no mucosal abnormality was reported. This provided an alternative means of identifying possibly missed lesions and allowed interrogation of the entire Endoscopic Database for entries up to February 1999, including information on patients for whom a 5-year follow-up interval was not possible.

RESULTS Colon cancer identified from population-based database (Cohort 1) Between July 1989 and February 1993, 4056 colonoscopies were recorded in the Endoscopic Database, and of these, 1098 (men, 40%; women, 60%) were performed to the cecum or ileum without detection of any mucosal abnormality. The most common indications for which these normal colonoscopies were as follows: bleeding (22%), anemia (16%), diarrhea (16%), and abdominal pain (15%) (Table 1). For the patients who underwent the 1098 normal examinations, 5-year follow-up was possible in 1047 by using the WA Linked Database. In this cohort, 5 patients (0.5%) were found to have colorectal cancer subsequent to the index colonoscopy. The interval between the normal examination and cancer diagnosis ranged from 3.5 to 5.2 years and the cancers were located in either the sigmoid colon or rectum (Table 2). The negative predictive value of colonoscopy for this cohort was calculated to be 99.5%. This result was supported by data from the Australian Association of Cancer Registries Database,15 which estimates the risk of colon cancer for an individual in the general population. The estimated number of colon cancers in an age- and gender-matched group of 1047 individuals from the general population was 10.55 cancers (1.0%). This latter estimate supports the actual risk of colon cancer (0.5%, 5/1047) in the present study cohort (chi-square = 2.00, p > 0.1). Although the risk of colorectal cancer for the follow-up period in this study cohort was not significantly different from that of the age- and genderVOLUME 55, NO. 2, 2002

Long-term follow-up of normal complete colonoscopy

H Ee, J Semmens, N Hoffman, et al.

Table 2. Patients from Cohort 1 subsequently found to have cancer recorded in Western Australian Cancer Registry Patient Age (y) 1 2

63 50

3 4 5

79 45 80

Indication Diarrhea Polyp on radiograph Bleeding F/u Ca F/u Ca, pain, bleeding

Interval to cancer (y)

Cancer location

5.2 3.7

Mid rectum Rectum

3.7 3.5 3.8

Rectosigmoid Sigmoid at 18 cm Rectum

Patient 3 is same as patient 1 in Table 3. F/u Ca, Follow-up colonoscopy for previous cancer.

matched general population, the majority of patients who undergo colonoscopy have symptoms and are not representative of the general population. A more appropriate group for comparison with the first cohort is all patients presenting for colonoscopy. The Endoscopic Database recorded 163 cancers diagnosed at 3121 colonoscopies (men, 42%; women, 58%) performed at Sir Charles Gairdner Hospital (5.2%, 163/3121) for the same period (July 1989 until February 1993). This is an order of magnitude higher than the risk found in the 5-year follow-up of patients after a normal colonoscopy (0.5%, 5/1047; chi-square = 45.9, p < 0.001). The most common indications for the 3121 colonoscopies were bleeding (22%), anemia (14%), and abdominal pain (11%) (Table 1). Colon cancer identified at repeat colonoscopy (Cohort 2) To further assess the implications of a normal colonoscopy with respect to subsequent colon cancer, the outcome of a repeat colonoscopy was determined in a cohort of patients in which no mucosal abnormality was identified on the first examination. This was another way of using the Endoscopic Database to identify possibly missed lesions at initial colonoscopy, and allowed interrogation of the entire Endoscopic Database up to February 28, 1999; including information on those patients in whom a 5year follow-up was not possible. A total of 26,058 colonoscopies were performed in 21,504 patients. Of these examinations, 8486 reached the cecum or ileum and no mucosal abnormality was identified. A search for follow-up colonoscopies in relation to the 8486 index examinations revealed that 595 repeat colonoscopies were performed in a cohort of 496 patients (men, 40%; women, 60%). The most common indications for repeat colonoscopy were as follows: bleeding (23%), family history of colorectal cancer (19%), anemia (14%), history of colonic adenoma (12%), and abdominal pain (11%) (Table 1). The VOLUME 55, NO. 2, 2002

Table 3. Patients found to have cancer on repeat colonoscopy after previously normal colonoscopy Patient Age (y) 1 2 3

79 80 40

Indication

Interval to cancer (y)

Cancer location

Bleeding F/u Ca Family hx Ca

3.7 1.8 4.2

Rectosigmoid Sigmoid at 20 cm Sigmoid

Patient 1 is same as patient 3 in Table 2. F/u Ca, Follow-up colonoscopy for previous cancer; Family hx Ca, family history of colorectal cancer.

interval between the index colonoscopy and subsequent examinations ranged from 3 days to 8.8 years (mean 3.1 years). Colorectal cancers were found in 3 of the 496 patients (risk = 0.6%), and similar to Cohort 1, the cancers were located in the sigmoid colon or rectum (Table 3). The negative predictive value of colonoscopy for this cohort was calculated to be 99.4%. Details of the 3 cases are shown in Table 3. One patient was also included in the first cohort (Patient 1 in Table 3 is the same as Patient 3 in Table 2). Patient 2 had an anastomotic recurrence of cancer, as opposed to a previously undiagnosed cancer. DISCUSSION This study confirms that colonoscopy is extremely accurate in the detection of colorectal cancer. The present study evaluated outcomes for patients who previously had undergone a complete colonoscopy that was reported to be normal. This has not been previously reported; its importance lies in the increasing demand in Australia and many countries for screening programs for colorectal cancer.10,11 The rationale for such programs is to increase the detection of early stage colorectal neoplasms and hence reduce morbidity and mortality from this malignancy.10,11 For patients in the first cohort the risk of a diagnosis of colorectal cancer within 5 years of a reportedly normal colonoscopy was only 0.5% (5/1047). Thus, a normal complete colonoscopy had a negative predictive value for colorectal cancer of 99.5% in this cohort. Of the 5 patients with cancer identified subsequently, the shortest interval between the initial colonoscopy and the eventual detection of cancer was 3.5 years. Because it is unlikely that an invasive cancer would escape detection for 3.5 years, it is our belief that no frank cancer was missed in this colonoscopy series. Rather, a missed premalignant adenoma is the most likely explanation for subsequent cancer in this interval. The longest interval to detection of cancer in this series was 5.2 years, which again suggests that the missed lesion was a premalignant adenoma. However, the possibility that the entire sequence from normal mucosa to invasive carcinoma occurred within this interval GASTROINTESTINAL ENDOSCOPY

169

H Ee, J Semmens, N Hoffman, et al.

cannot be excluded. Nevertheless, all subsequent cases of cancer were most likely because of missed neoplastic polyps that had not as yet transformed to invasive carcinoma at the time of the index procedure. However, the value of colonoscopy in relation to colorectal neoplasia is not merely the detection of frankly malignant lesions, but also the detection and removal of premalignant lesions. The risk of subsequently detecting colorectal cancer in patients in whom a previous colonoscopy was normal (0.5%, 5/1047) was similar to the estimated risk of colorectal cancer in an age- and gender-matched group in the general population (1.0%, 10.55/1047; chisquare = 2.00, p > 0.1). Although this can be misinterpreted to mean that colonoscopy makes no difference with regard to the likelihood of subsequently having cancer develop, all 1047 patients in the cohort undergoing colonoscopy had symptoms or indications for which colorectal cancer had to be excluded. Indeed, based on data from the Endoscopic Database for the same period, the overall rate of cancer detection for all colonoscopies was 5.2%, a markedly greater frequency than the 0.5% risk of cancer during the follow-up period after a normal colonoscopy in the first cohort (chisquare = 45.9, p < 0.001). Thus, a normal colonoscopy lowers the likelihood of an advanced neoplastic lesion from a pretest probability of 5.2% to a posttest probability of 0.5%. An expanded database was used in the second part of the present study to identify a cohort of 496 patients who had a repeat colonoscopy and who had previously undergone a complete colonoscopy that was normal. The advantage of this strategy was that a long follow-up period was unnecessary and it thus enabled examination of the entire Endoscopy Database until February 28, 1999. Only 3 cancers (0.6%, 3/496) were detected on repeat colonoscopy in this cohort of patients, which again suggests that the rate of an initial examination falsely negative for cancer is low (negative predictive value = 99.4%). The shortest interval to cancer detection after a normal examination was 1.8 years (Patient 2), but this proved to be an anastomotic recurrence of a previously resected cancer. In this case, it is likely that no abnormality was present at the initial postsurgical colonoscopy and a strong case can be made for exclusion of this patient from the analysis. One of the other 2 patients (Patient 3, 4.2-year interval to cancer) had a family history consistent with hereditary nonpolyposis colorectal cancer as defined by the “Amsterdam” criteria.17 Specifically, colorectal cancer was diagnosed in his mother at 50 years of age (the first of 2 metachronous cancers), his maternal uncle at 50 years of age, maternal grandfather at 45 years of age, and maternal great-aunt during her 170

GASTROINTESTINAL ENDOSCOPY

Long-term follow-up of normal complete colonoscopy

fifth decade of life. It has been suggested that in hereditary nonpolyposis colorectal cancer, the interval for the transition from normal mucosa to carcinoma may be substantially contracted.18 Thus, 2 of the 3 patients found to have cancer on repeat colonoscopy in this cohort had disorders in which cancer can develop rapidly. This suggests that there may well have been no detectable lesion at the initial examination. Moreover, these findings reaffirm the high sensitivity of colonoscopy for detection of cancer and significant premalignant lesions. Although this current study in which follow-up was obtained in patients who had undergone a reportedly normal colonoscopy differs in design from 2 previous studies in which patients were followed after colonoscopic polypectomy, the low miss rates for cancer at initial colonoscopy are similar in all of these studies.7,8 In the National Polyp Study in the United States, 5 cancers were found on repeat colonoscopy in 1210 patients (0.4%, 5/1210) followed for a mean of approximately 6 years.8 All 5 cancers were malignant polyps, with 3 of these detected at 3 years and 2 found at 6 and 7 years, respectively, of follow-up. Similarly, in a follow-up of a larger group of patients in a study in Japan, 31 cases of colorectal cancer were identified in 6715 patients (0.5%, 31/6715) over a mean followup of 6 years after initial colonoscopy.7 Although colonoscopy is assumed to be accurate for the diagnosis of colorectal cancer, it is nevertheless imperfect. Haseman et al.1 found that in 47 of 941 cases (5.0%) of colorectal cancer diagnosed at various hospitals in Indiana, a colonoscopy had been performed within the prior 3 years. In 20 (42.6%) of these cases, the colonoscopy had not reached the tumor, and in the remainder the tumors (thought likely to have been present at the earlier examinations) were not seen. To avoid complicating the results of the present study with missed diagnoses of cancer because of inability to reach the proximal colon, only patients in whom the entire colon had been examined were included. Similarly, Gorski et al.2 reviewed the records of 557 patients who underwent surgery for colorectal cancer and identified 22 (3.9%) who had undergone 1 or more colonoscopies between 4 to 59 months before the diagnosis of cancer. A further 7 patients with carcinoma in situ were also found to have undergone an earlier examination. Although both of these studies show the fallibility of colonoscopy in cancer diagnosis, the results cannot be compared directly with those of the present study because of differences in designs. Remarkably, in the current study all subsequently detected cancers were located in either the sigmoid colon or rectum. This differs from the findings in the study of Haseman et al.1 in which 9 of 27 missed canVOLUME 55, NO. 2, 2002

Long-term follow-up of normal complete colonoscopy

cers were located in the sigmoid colon or rectum, provided those areas had been examined. Likewise, in the study of Gorski et al.,2 1 of 7 in situ cancers and 4 of 22 invasive cancers were located in the sigmoid colon or rectum. Although it is intriguing to speculate on the reasons for differences in location of missed cancers, it is important to note that the number of cases in the present study is small. Nonetheless, our findings underscore the need for vigilance when examining the sigmoid colon and rectum. Colonoscopy is more likely to miss premalignant adenomas than cancers because adenomas are usually much smaller than cancers. Studies in which serial colonoscopies were performed within short intervals found miss rates for adenoma of 15% to 27% for lesions less than 6 mm in diameter, 12% to 13% for lesions 6 to 9 mm, and 6% to 7% for lesions greater than 9 mm.4,5 However, the missed adenoma 10 mm or less in diameter has less clinical significance because the likelihood that an adenoma in this size range harbors carcinoma, or will progress to invasive carcinoma, is low.19 This study confirms that colonoscopy is highly accurate for the detection of colorectal cancer. Irrespective of presenting symptoms, when a colonoscopy is normal the risk that an advanced colorectal neoplasm was present was small (0.5%). The colonoscopies in this study were performed by gastroenterologists and gastroenterology trainees within teaching hospitals. Some investigators have found that procedures performed by gastroenterologists have a higher sensitivity for detecting polyps20 and cancer,1 compared with those performed by nongastroenterologists. This is likely because of the adequacy of training and the frequent performance of colonoscopy. The high sensitivity of colonoscopy for detection of colorectal cancer in the present study is especially relevant to the issue of increasing demand for screening programs for colorectal cancer in large populations. By implication, the adequacy of training in colonoscopy must be assured and standards for the delivery of this care maintained.

H Ee, J Semmens, N Hoffman, et al.

4.

5.

6.

7.

8.

9.

10.

11.

12.

13.

14.

15.

16.

17.

ACKNOWLEDGEMENT The Perth Teaching Hospitals Endoscopy Group consists of the Endoscopy Units of Sir Charles Gairdner, Royal Perth and Fremantle Hospitals. REFERENCES 1. Haseman JH, Lemmel GT, Rahmani EY, Rex DK. Failure of colonoscopy to detect colorectal cancer: evaluation of 47 cases in 20 hospitals. Gastrointest Endosc 1997;45:451-5. 2. Gorski TF, Rosen L, Riether R, Stasik J, Khubchandani I. Colorectal cancer after surveillance colonoscopy: falsenegative examination or fast growth? Dis Colon Rectum 1999;42:877-80. 3. Hixson LJ, Fennerty MB, Sampliner RE, Garewal HS. VOLUME 55, NO. 2, 2002

18.

19.

20.

Prospective blinded trial of the colonoscopic miss-rate of large colorectal polyps. Gastrointest Endosc 1991;37:125-7. Bensen S, Mott LA, Dain B, Rothstein R, Baron J. The colonoscopic miss rate and true one-year recurrence of colorectal neoplastic polyps. Polyp Prevention Study Group. Am J Gastroenterol 1999;94:194-9. Rex DK, Cutler CS, Lemmel GT, Rahmani EY, Clark DW, Helper DJ, et al. Colonoscopic miss rates of adenomas determined by back-to-back colonoscopies. Gastroenterology 1997;112:24-8. Muller AD, Sonnenberg A. Prevention of colorectal cancer by flexible endoscopy and polypectomy. A case-control study of 32,702 veterans. Ann Intern Med 1995;123:904-10. Nozaki R, Takagi K, Takano M, Miyata M. Clinical investigation of colorectal cancer detected by follow-up colonoscopy after endoscopic polypectomy. Dis Colon Rectum 1997;40:S16-22. Winawer SJ, Zauber AG, Ho MN, O’Brien MJ, Gottlieb LS, Steinberg SS, et al. Prevention of colorectal cancer by colonoscopic polypectomy. New Engl J Med 1993;329:1977-81. Thiis-Evensen E, Hoff GS, Sauar J, Langmark F, Majak BM, Vatn MH. Population-based surveillance by colonoscopy: effect on the incidence of colorectal cancer. Telemark Polyp Study I. Scand J Gastroenterol 1999;34:414-20. Winawer SJ, Fletcher RH, Miller L, Godlee F, Stolar MH, Mulrow CD, et al. Colorectal cancer screening: clinical guidelines and rationale [published errata appear in Gastroenterology 1997 Mar;1121060 and 1998 Mar;114:625]. Gastroenterology 1997;112:594-642. Australian Health Technology Advisory Committee (AHTAC). Colorectal cancer screening. Canberra: Australian Government Publishing Service; 1997. Holman CD, Bass AJ, Rouse IL, Hobbs MS. Population-based linkage of health records in Western Australia: development of a health services research linked database. Aust N Z J Public Health 1999;23:453-9. Semmens JB, Lawrence-Brown MM, Fletcher DR, Rouse IL, Holman CD. The Quality of Surgical Care Project: a model to evaluate surgical outcomes in Western Australia using population-based record linkage. Aust N Z J Surg 1998;68:397-403. Semmens JB, Platell C, Threlfall TJ, Holman CD. A population-based study of the incidence, mortality and outcomes in patients following surgery for colorectal cancer in Western Australia. Aust N Z J Surg 2000;70:11-8. Australian Institute of Health and Welfare (AIHW), Australian Association of Cancer Registries (AACR). Cancer in Australia 1996: incidence and mortality data for 1996 and selected data for 1997 and 1998. Cancer Series. Vol CAN7. Canberra: Australian Institute of Health and Welfare (AIHW); 1999. Australian Institute of Health and Welfare (AIHW). Cancer in Australia 1996: incidence and mortality data for 1996 and selected data for 1997 and 1998. Available at: http://www. aihw.gov.au/publications/health/ca96/ca96.pdf. Accessed 1999. Lynch HT, Smyrk TC, Watson P, Lanspa SJ, Lynch JF, Lynch PM, et al. Genetics, natural history, tumor spectrum, and pathology of hereditary nonpolyposis colorectal cancer: an updated review. Gastroenterology 1993;104:1535-49. Lynch HT, Smyrk T, Lynch JF. Overview of natural history, pathology, molecular genetics and management of IINPCC (Lynch syndrome). Int J Cancer 1996;69:38-43. Itzkowitz SH, Kim YS. Colonic polyps and polyposis syndromes. In: Feldman M, Scharschmidt BF, Sleisenger MH, editors. Sleisenger and Fordtran’s gastrointestinal and liver disease: pathophysiology/diagnosis/management. Vol 2. Philadelphia: WB Saunders; 1998. p. 1865-905. Rex DK, Rahmnani EY, Haseman JH, Lemmel GT, Kaster S, Buckley JS. Relative sensitivity of colonoscopy and barium enema for detection of colorectal cancer in clinical practice. Gastroenterology 1997;112:17-23. GASTROINTESTINAL ENDOSCOPY

171