|Home | About | Journals | Submit | Contact Us | Français|
Despite regular colonoscopy, interval colorectal cancer (CRC) may occur. Long-term studies examining CRC rates in patients with previous colonoscopy are lacking.
We examined the rate of interval CRC in the Polyp Prevention Trial-Continued Follow-up Study (PPT-CFS), an observational study of participants which began after the Polyp Prevention Trial (PPT) ended.
A national U.S. community-based polyp prevention trial
Medical records of CRC were collected, reviewed, and abstracted in a standardized fashion.
Among 2,079 PPT participants, 1,297 (62.4%) agreed to participate in the PPT-CFS. They were followed for a median of 6.2 years after 4.3 years of median follow up in the main PPT. Nine cases of CRC were diagnosed over 7,626 person years of observation (PYO) for an incidence rate of 1.2/1000 PYO. The ratio of CRCs observed compared to that expected by SEER was 0.64 (95% CI 0.28–1.06). Including all CRCs (N=22) since the beginning of the PPT trial, the observed compared to expected rate by SEER was 0.74 (95% CI 0.47–1.05). Of patients who developed CRC in the PPT-CFS, 78% had a history of an advanced adenoma compared to only 43% among patients who remained cancer-free (p=0.04).
A relatively small number of interval cancers were detected.
Despite frequent colonoscopy during the PPT trial, in the years subsequent to the trial, there was a persistent ongoing risk for cancer. Subjects with a history of advanced adenoma are at increased risk for subsequent cancer and should be followed closely with continued surveillance.
Screening can reduce mortality to colorectal cancer (CRC) (1–3). Evidence suggests that colonoscopy screening can also reduce the incidence of colorectal cancer, presumably due to the removal of adenomatous polyps, the precursor lesion to CRC (4–6). After initial colonoscopy, patients with adenomas are followed with surveillance colonoscopy, to identify and remove subsequent adenomas before they progress to CRC. However, colonoscopy is not perfect, and interval cancers diagnosed between surveillance examinations may occur. Several studies have reported interval cancer incidence rates ranging from 1.7–2.4 cancers per 1000 person years of observation (PYO) (7–9). Multiple explanations for interval cancers have been proposed, including: missed lesions during the initial colonoscopy, incomplete adenoma removal, development of new lesions, and failed detection of cancer despite biopsy sampling (7;9;10). In one study, the rate of cancer detection at the first follow-up interval was 3.79/1000 PYO compared to only 0.96/1000 PYO for the second follow up interval, suggesting that many of the cancers diagnosed at the first interval were missed (7). Studies reporting on interval cancer rates have generally followed subjects for a limited time frame of only a few years.
The Polyp Prevention Trial (PPT) was a four-year multicenter, randomized controlled trial designed to examine the effect of a low fat, high fiber, high fruit and vegetable diet on the recurrence of colorectal adenomas (11). In the trial, 13 cases of interval CRC were identified over 5810 person years of observation, for a rate of 2.2/1000 PYO (9). During the main phase of the trial, subjects underwent frequent colonoscopy including at study entry and at one and four years after enrollment, with an average of 3.1 colonoscopy exams during the trial. After completion of the PPT study, patients were given the option to enroll in the Polyp Prevention Trial-Continued Follow-up Study (PPT-CFS) (12). In the PPT-CFS, patients agreed to passive follow-up and consented for review of their health records and results of subsequent colonoscopy for an additional four or more year follow-up period.
The aim of this study was to determine the rate of interval CRC subsequent to the main phase of the PPT and to characterize and examine the circumstances surrounding the diagnosis of cancer. A report of the utilization of colonoscopy and the adenoma yield during the PPT-CFS is available (13). This report concerns the cancer diagnoses. In particular, because subjects had undergone frequent colonoscopy as part of the PPT trial, we wanted to evaluate how frequent colonoscopy examinations (mean 3.1 exams over a median 4.3 years), albeit within a limited time frame, might impact the rate of subsequent colorectal cancer over a more prolonged observation period.
The PPT enrolled 2079 patients who had a baseline enrollment colonoscopy (T0) with at least one adenomatous polyp but no history of CRC. Details of the study population have been reported (11). Subjects were randomized to a dietary intervention to adopt a low fat, high fiber, fruits and vegetable diet versus their usual diet. A repeat colonoscopy was performed in one to two years (T1) to detect and remove lesions which may have been missed at the baseline colonoscopy. Subjects underwent a repeat colonoscopy in four years (T4) to examine the impact of dietary change on adenoma recurrence. Eligibility for the study required that subjects be 35 years and older and have one or more colorectal adenomas removed during a qualifying colonoscopy within 6 months prior to randomization. Exclusion criteria were: history of CRC, surgical resection of adenomas, bowel resection, polyposis syndrome, inflammatory bowel disease, weight greater than 150% of recommended weight, lipid lowering drug use or a condition that would limit compliance with the diet protocol. The study was approved by the institutional review board of all participating centers.
After the completion of the PPT study, subjects were offered the option to continue in the observational PPT-CFS. Participants agreed to provide information to the researchers via annual health and lifestyle questionnaires which included questions about hospitalization, colonoscopy and cancer diagnoses. Subsequent colonoscopy was performed at the discretion of the clinicians caring for the patients. Colonoscopy and pathology records were obtained and the indications and outcomes for the procedures were abstracted in a standardized manner. Colorectal cancer cases were also identified from the annual questionnaires. Original medical records were obtained and reviewed for these cases including procedure, radiology, operative, and histopathology reports for information regarding the diagnosis, size, location, and stage of cancer.
The baseline characteristics of subjects who participated in PPT-CFS were compared to those who declined participation using the Wilcoxon rank-sum test and chi-square to compare means and proportions of exposure variables. Advanced adenomas were defined as adenomas ≥ 10 millimeters, or with tubulovillous or villous histology, or with high-grade dysplasia (11).
For this study, all patients with colorectal cancer diagnosed after enrollment in the PPT-CFS were considered interval cancers. Medical records were obtained and reviewed for these cases, including procedure notes, operative reports and histopathology findings. Based upon the patients’ records, cancer stage was determined according to American Joint Committee on Cancer Criteria (14).
The incidence rate was calculated based upon the number of subsequent CRC cases and the total person years of observation. For subjects who consented to participate, observation for the PPT-CFS began on the day following the year 4 colonoscopy of the main PPT trial, which was what determined whether the patient had an adenoma recurrence. Subjects’ person years of follow-up accrued until the latest date at which their CRC status was known, either the planned study end date, their date of death, or the date of submission of their last health and lifestyle questionnaire. Expected numbers of cancers in the general population were calculated by applying Surveillance, Epidemiology and End Results (SEER) rates for each gender, five year age group, and four year calendar period to the corresponding PYO in PPT-CFS. We then calculated a standardized incidence ratio comparing the CRC rate in the PPT-CFS to the expected value. Additionally, we combined data from a previous study evaluating interval cancers found during the PPT study period (9) with the PPT-CFS data and compared this to expected SEER CRC rates. The PYO for the main PPT study component was calculated as the time from PPT enrollment to the time of death, loss to follow up, diagnosis of cancer, or completion of the final endpoint of the main PPT trial, whichever came first, and it included the entire PPT cohort. Confidence intervals for the observed to expected ratio were calculated assuming a Poisson distribution for cancers in PPT; SEER rates were assumed to be fixed. Cancers diagnosed during the PPT-CFS were analyzed according to an algorithm developed by Pabby et al (9) and grouped into one of four potential etiologies: failed biopsy detection, incomplete removal, missed cancer and new cancer. A fifth category, cancer of indeterminate cause, was created for our analysis. This group included cases of cancer with advanced features (size 2cm and stage III or IV) which were diagnosed greater than 60 months from the last colonoscopy, a scenario not encountered by Pabby et al due to a shorter duration of follow-up. Advanced cancer cases with > 60 months between colonoscopy could represent either missed or new cancer; hence, they were considered indeterminate.
Of 2079 patients enrolled in the PPT, 1905 completed the study. Of those that completed the PPT, 1297 (68%) agreed to participate in the PPT-CFS and were followed for a median of 6.2 years. PPT-CFS participants compared to PPT subjects who did not participate in the PPT-CFS were more likely male (67 vs. 59%, p<.001), more educated (78% vs. 68% post high school education, p<.001), and to have a family history of CRC (28 vs. 24%, p=.03) (Table 1). During the PPT trial, PPT-CFS participants underwent more colonoscopies compared to non-participants (3.26 vs. 3.14, P value 0.02). The two groups had no significant differences in their likelihood of having an advanced adenoma during the PPT. Average age upon PPT-CFS study entry was 65.3 ± 9.6 years. A total of 11% of subjects were lost to follow up and had their observation period censored prior to the designed endpoint. These were subjects who were presumed alive but did not submit a final health status questionnaire.
During the PPT-CFS, an additional nine cases of CRC were diagnosed over 7626 person years of observation for a rate of 1.2 cases/1000 PYO. Based on SEER CRC incidence rates, 14.2 cancers were expected in an age and gender matched population. The ratio of cancers observed in the PPT-CFS study compared to that expected by SEER was 0.64 (95% CI 0.28–1.06).
For the PPT period, 13 cases of CRC were diagnosed and 15.4 cancers were expected based upon the SEER incidence rates. When the data from the PPT period is combined with that of the PPT-CFS, a total of 22 cancer cases were diagnosed compared to 29.6 cancers expected for an standardized incidence ratio of 0.74 (95% CI 0.47–1.05).
Of the new CRC cases diagnosed during the PPT-CFS, 78% (7/9) were diagnosed in men, and 44% (4/9) had a family history of CRC (Table 2). During the PPT, these patients had a mean of 3.3 adenomas (range 1–7) and the mean size of their largest adenoma was 8.2 mm (range 4–12mm). Four patients had polyps with tubulovillous or villous histology and one had high-grade dysplasia. Seven of the nine patients had a previous history of advanced adenomas; however, 44% (4/9) had no adenomas present on the final colonoscopy at the completion of the PPT. Risk factors were compared between patients in the PPT-CFS who subsequently developed CRC and those who were cancer-free (Table 3). History of an advanced adenoma was the only factor significantly different between the two groups (78% vs. 43%, p value 0.04).
The nine patients subsequently diagnosed with CRC had all undergone at least 3 colonoscopies prior to diagnosis since beginning the PPT, with a mean of 3.9 colonoscopies over a period ranging from 80–131 months (Table 4). The average interval between the last colonoscopy and cancer diagnosis was 41 months (range 11–83 months). Mean age at CRC diagnosis was 71 years (range 52–82). Eight of the nine cancers were located proximal to the splenic flexure: one in the cecum, two in the ascending colon, five in the transverse colon and one in the descending colon. The average tumor size was 2.9 cm. At diagnosis, the majority of the cases were either UICC/AJCC stage I or II; four cases were stage I, three stage II, one stage III and one stage IV.
Various clinical circumstances led to the colonoscopy that diagnosed CRC. In two patients, cancer was detected upon colonoscopy follow-up of a lesion with high-grade dysplasia one year prior. In the remaining seven cases, cancer was diagnosed during a colonoscopy performed for evaluation of symptoms (one case presented with a change in bowel habit, one with rectal bleeding, two with anemia and three with abdominal pain).
The etiology of each case of CRC was assessed based upon the published algorithm (9). Four cases (#2, 4, 5, 9) were detected in colonic segments where a polyp was found previously. Two of these cases were determined to be due to incomplete removal of a previous adenoma and the other two cases to failed biopsy detection of a neoplastic lesion. Of the five cases that were detected at previously polyp-free segments, one case was missed, two cases were newly developed cancers and two cases were due to indeterminate reasons.
Nine participants developed colorectal cancer during the PPT follow-up study for a rate of 1.2 cases/1000 PYO, despite undergoing a mean of more than 3 colonoscopy procedures during the PPT trial. This interval cancer rate was 64% of that expected by SEER and when cancers detected in the main PPT trial and the follow-up study were combined, the overall rate was 74% of that expected by SEER. The absolute rates were not statistically significantly different from that expected based on SEER data. Our study included a relatively prolonged >10 year median follow up. Previous studies of interval cancer were limited to a shorter follow-up of three to six years (7–9;15;16). Our results confirm the need for continued colonoscopy surveillance in at risk subjects due to ongoing risk of colorectal cancer and especially among those with a history of advanced adenoma.
Our data are consistent with studies which demonstrate a continued risk of CRC despite previous colonoscopy. In a review of data from 3 multi-center adenoma prevention trials, 19 cases of colorectal cancer were diagnosed among 2915 patients with an overall incidence of 1.74 per 1000 PYO and an incidence ratio of 0.98 (95% CI, 0.63–1.54) when compared to SEER (7). A similar incidence was found in the VA cooperative study where among 1171 subjects with neoplasia at baseline colonoscopy, 1.7 cases of CRC per 1000 PYO were diagnosed in follow-up over a 5.5 year period (8). It should be noted when comparing the CRC incidence ratio in patients under surveillance colonoscopy compared to those in SEER, that surveillance studies include patients with a previous history of adenoma who are at higher risk for subsequent neoplasia. However, they presumably are at lower risk for CRC due to having undergone colonoscopy.
In our study, the majority of subjects who developed CRC had a previous advanced adenoma during the PPT trial. This was the only factor significantly different between patients who developed CRC and those who did not. The association of increased CRC risk in patients with advanced adenomas has been previously observed (17). In our population, subjects appear to remain at increased risk for CRC even with a negative interval colonoscopy. Four of the nine patients who developed CRC had no polyps found at the final T4 colonoscopy at PPT completion. For example, case #7 was a 44 year old female at enrollment with an advanced adenoma at T0/T1 but no polyps at T4. She was diagnosed with CRC 73 months after T4 colonoscopy or 10 years after her initial advanced adenoma. She did manifest an advanced adenoma at a young age. In contrast, however, cases #6 and #8 had no prior history of advanced adenomas and were polyp-free at T4 but developed CRC. The risk of cancer is certainly not exclusive to subjects with prior advanced adenomas, but ongoing risk is prominent in that subgroup.
Some patients had received multiple colonoscopies over a relatively short time interval, yet they were diagnosed with CRC. Four patients had undergone at least three colonoscopies over approximately eight years. Four of the nine patients had a colonoscopy that did not diagnose their cancer two years previously. Concerns about the quality of colonoscopy have increased amid evidence of a significant miss rate during colonoscopy. In tandem colonoscopy studies, miss rates of 0–6% were observed for adenomas ≥10 mm but the miss rate increased markedly for lesions 5–10 mm, ranging from 12–13%, and in lesions less than 5 mm, miss rates of 16–27% were noted (18;19). In studies of CT colonography, the colonoscopy miss rate for adenomas ≥10 mm has been reported at 12–17% (20–22). Several factors that adversely affect colonoscopy performance have been identified, including rapid withdrawal time, poor bowel prep, and performance by a non-gastroenterologist (23–26). It is difficult to assess the quality of colonoscopy performance due to a lack of standard colonoscopy reporting. To address this issue, the National Colorectal Cancer Roundtable task force recently developed a standardized colonoscopy reporting system to facilitate auditing of procedures and quality improvement (27).
Two recent case-control studies of colonoscopy suggest that colonoscopy is not as effective in preventing mortality (28) or incidence of right sided cancer (29). One could ask how our results contribute to this concern, given that we do not demonstrate a statistically significant decrease in CRC compared to SEER, despite subjects undergoing frequent colonoscopy. It must be emphasized that this study is not a population based study of colonoscopy effectiveness in that all patients began with adenomatous polyps. Approximately 18% had a history of adenomas in the previous 5 years to enrollment, and over 37% had an advanced adenoma at enrollment (11). Thus, this cohort is at higher risk for CRC than the general population followed in SEER. Furthermore, although 8 of the 9 cancers in the PPT-CFS were proximal to the splenic flexure, 5 of the 8 were potentially a consequence of suboptimal colonoscopy quality due to one missed lesion, two failed biopsy detections, and two incomplete removals, whereas the others were deemed new or of indeterminate cause
A recent investigation of surveillance colonoscopy found that subjects were more likely to develop recurrent adenomas in the same colonic segment, suggesting that particular attention be paid to where a previous adenoma has been removed (30). Our data further supports recognition of the site of a previous polyp and close observation of this colonic segment, especially among subjects with an advanced adenoma.
Limitations of our study should be acknowledged. Only a small number of cases of CRC were identified, limiting statistical analysis. It was difficult to determine whether a cancer occurred at the site of a previously identified pre-malignant lesion or whether it merely occurred in the same colonic segment, because of limited descriptive information on adenoma location from colonoscopy reports. Only 68% of subjects originally enrolled in the PPT trial participated in the PPT-CFS and this could have led to a biased sample. When characteristics of the PPT-CFS participants and non-participants were compared, despite some statistical differences, the two populations do not appear to be at a clinically significantly different risk for colorectal cancer (Table 1). Finally, the occurrence of colorectal cancer relied on patient self-report. However, additional cases identified via a more comprehensive search would only increase the rate, which was already substantial.
In Pabby et al’s (9) original algorithm analyzing interval cancers, it was assumed that advanced cancers (≥2 cm in size and stage III or IV) detected > 30 months after the last colonoscopy at a previously polyp-free segment were missed lesions. The longest potential interval between colonoscopy in the PPT trial was 48 months, an interval presumed unlikely to be long enough for a new cancer to develop advanced features. In this study, the longest interval from colonoscopy to CRC diagnosis was 83 months, a time period long enough to develop a new cancer with advanced features. Thus, in this cohort advanced cancers detected years later were categorized as indeterminate.
In conclusion, despite frequent colonoscopy during the PPT trial, there was a persistent ongoing risk for cancer in the years subsequent to the trial. Subjects with a history of advanced adenoma are at increased risk for subsequent cancer and should be followed closely with continued surveillance.
The authors acknowledge Westat, Inc for their work as coordinating center for the PPT-CFS.
Conflict of interest: There is none to disclose
Financial Disclosures: None