Test-positive rate at CT colonography is increased by rectal bleeding and/or unexplained weight loss, unlike other common gastrointestinal symptoms.

Fig. 1 summarizes the patient flowchart in this study. 903 out of the 7361 (12%, 95%CI 11–13%) subjects had at least one clinically significant colonic finding at CTC. Histopathological diagnoses and/or etiology for the confirmed findings are given in Table 1 , showing that the abnormalities were most commonly of mucosal origin. Comparison of CTC to OC and/or surgery was possible for 692/903 (77%; 95%CI 0.074–0.79) of these subjects. CTC findings were confirmed in 637, while 55 patients were false-positive. In addition, repeat CTC, OC and/or surgery was performed within a range of 6–60 months in 1198 subjects with negative CTC findings for the following reasons: 656 patients underwent OC including programmed screening tests in 168, exacerbation of a “non-alarm” gastrointestinal symptom or occurrence of a new symptom in 443, and occurrence of an “alarm” symptom in 48. 527 patients underwent a second CTC including programmed screening tests in 287, new or exacerbated “non alarm symptoms in 216 and “alarm symptoms” in 24. Twelve patients (all in group B) had surgical resection after acute diverticulitis. Five patients were false-negative (2 with >6 mm polyps and 3 with cancers). All controlled cases, yield a 637/642 (99.2%, 95%CI 0.98–0.99) true-positive fraction and a 55/692 (7.95%, 95%CI 0.05–0.09) false-positive fraction. The average number of symptoms per patient was 1.54 ± 0.35 (3800, 1524, 662, 29 and 3 patients had respectively 1, 2, 3, 4 and 5 symptoms). Abdominal pain and constipation were the most frequent, while unexplained weight loss was the least frequent symptom. Groups A, B and C included respectively 1343, 5329 and 689 patients; their demographic characteristics are given in Table 2 . The average age in group C (64.37 ± 14.61 years) was significantly higher than in both groups A (60.96 ± 11.22 years) and B (59.43 ± 14.31 years) ( p  < 0.001). The male/female ratio was lower in group B (0.63) versus group A (1.03) ( p  < 0.001). The rate of E4 findings in the C-RAD reporting system was significantly higher in group C (44/689, 6.4%, 95%CI 0.04–0.08) versus groups A (26/1343, 1.9%, 95%CI 0.01–0.03) and B (64/5329, 1.2%, 95%CI 0.01–0.02) ( p  < 0.001). The regression analysis showed that the group of symptom (B < C), gender (F < M) and age all significantly impacted the test-positive rate at CTC. When pooling patients by age groups ( Table 3 ), the test-positive rate in group C (138/689, 20.0%, 95%CI 0.17–0.23) was significantly higher than in groups A (79/1343, 5.8%, 95%CI 0.04–0.07) and B (420/5329, 7.8%, 95%CI 0.07–0.08) ( p  < 0.001), then, across all ages, except for patients younger than 50 years of age whose test-positive rate in groups C, A and B were respectively (6/96, 6.25%, 95%CI 0.01–0.11), (5/173, 2.9%, 95%CI 0.00–0.05) and (47/1241, 3.8%, 95%CI 0.02–0.04) ( p  = 0.412). Diagnostic values of CTC within the three groups are given in Table 4 . Both the true-positive fraction that was consistently above 97% in all groups ( p  = 0.991) and the false-positive fraction did no differ among all groups ( p  = 0.240).

The authors declare that there is no conflict of interest.

Our institutional review board approved the study and authorized this retrospective patient data analysis without further consent. We searched our hospital records for all subjects who completed a CTC procedure between June 2003 and August 2010. This search yielded 9122 subjects (3822 males, 5300 females, average age: 60.11 ± 13.75 years, range: 18–96 years). Indications for CTC included screening and direct referral ( n  = 8573), secondary referral after incomplete OC ( n  = 285), and Double Contrast Barium Enema (DCBE) referral change ( n  = 264). This referral change was justified by the non-superiority of DCBE over CTC for colonic lesions in several studies [16] , [17] . Written informed consent was given by all subjects prior to procedures. 1761 subjects with a familial or personal history of polyps or colorectal cancer, genetic conditions, inflammatory bowel disease, who were at increased- or high-risk for colorectal cancer [2] were excluded. The remaining 7361 subjects, with average-risk [18] for CRC (general population) (3073 males, 4288 females, average age: 60.3 ± 13.9 years, range 18–96 years) were evaluated. Their clinical status with regard to the presence of the following gastrointestinal symptoms, prior to CTC was retrieved from the referral forms and/or gathered by patient's anamnesis and all other available patient data, including: (i) abdominal pain, (ii) constipation, (iii) diarrhea, (iv) irregular bowel movement, (v) bloating, (vi) melena, (vii) rectal bleeding, and (viii) unexplained weight loss. We retrospectively assigned the subjects to three main groups, according to the purported clinical importance of these symptoms regarding the level of specificity for CRC [5] : group A included the asymptomatic subjects; group B, the patients with one or more nonspecific symptom(s) (i–vii) in the absence of an established “alarm” symptom (vii and viii), who were assigned to group C. All patients underwent the same standardized procedure that consisted into three steps including patient preparation, scanning and data interpretation. The preparation involved two steps including cathartic colonic cleansing and residual fluid tagging. For patients in good general condition, colonic cleansing was achieved by a one-day clear liquid diet, one bottle of sodium phosphate preparation (Fleet-Phospho-soda ® , Wolfs, Zwijndrecht, Belgium) and 4 tablets of bisacodyl (Dulcolax ® , Boehringer Ingelheim, Ingelheim, Germany). For frail patients, cleansing consisted into 2 days of low-residue diet combined to 8 g of magnesium–sulphate on the examination day's morning, in addition to 2 tablets of bisacodyl and 100 ml of contrast agent (Gastrografin ® , Schering AG, Berlin, Germany) twice a day. In patients with renal insufficiency, cardiac failure or severe hypertension, preparation consisted in 3 days of low-residue diet with 2l of Moviprep ® (Norgine, Heverlee, Belgium) (propylene-glycol + ascorbic acid) and 4 tablets of bisacodyl the day before the study. Residual fluid tagging was obtained by ingestion of 100 ml Gastrografin ® the evening before the procedure and total colonic residual fluid volume was reduced by using a suppository of bisacodyl approximately 2 h before examination, except for patients who underwent CTC after incomplete OC. These patients drank 100 ml of Gastrografin and inserted a suppository of bisacodyl 1 h before the procedure. Before data acquisition, an iv injection of 20 mg/1 ml of Buscopan ® (butylhyoscinbromid – Boehringer Ingelheim, Bruxelles, Belgium) was performed and a rectal cannula was inserted for colonic distension with an automatic carbon dioxide insufflator VMX-1010A (Vimap technologies™, Girona, Spain). A 32-row (GE Lightspeed VCT™, GE Healthcare, Milwaukee, WI) until 09/2010, then a 64-row (GE Discovery CT750 HD™, GE Healthcare, Milwaukee, WI) multislice scanners were used for image acquisitions. Parameters consisted into 1.2 mm-thick slices with a 0.625 mm reconstruction interval, using a 50 mA s low-dose protocols with variable kV, adjusted to body-density for dose reduction, supplemented since 2010 by an adaptive statistical iterative reconstruction algorithm (ASIR) (GE Healthcare, Milwaukee, WI). Two acquisitions were performed: the first in supine position and the second, either in prone position, or right decubitus for unfit and obese patients. Immediate review of the images was performed by a radiologist in all cases. In 897 patients (10%), a third acquisition was ordered because of a segmental collapse preventing confident analysis. Reading was performed offline on a workstation (Advantage Windows, GE Healthcare, Milwaukee, WI) with a software (Colon VCAR) allowing filet-view, supplemented by “computer aided diagnosis” (CAD) assistance from January 2009, and electronic cleansing from June 2010. Reconstruction algorithms, image display preferences and reading principles used for interpretation are described elsewhere [19] . We used C-RAD reporting classification for all findings [20] . Each finding was assigned to both a colonic segment and a distance to the anal margin. Clinically significant colonic findings were defined as either ≥6 mm polyps, masses or others requiring work-up or treatment [20] . Clinical files, and reports were searched for repeat CTC, OC and surgical procedures after the initial CTC, when applicable. A reviewer was requested to match CTC and the reference standard findings, using available location by segment, and/or distance from the anal margin. We did not attempt to match using the size criteria because of known size discrepancies between OC and CTC polyp measurement [21] . Using an “intention to treat” algorithm, per-patient CTC diagnostic values for the diagnosis of clinically significant colonic findings were calculated. Patients with at least one matched finding were considered true-positive while patients with CTC findings unmatching the reference standard were considered as false-positive. Those with no finding at CTC and at least one positive finding on the reference standards were considered false-negative. For patients with several colonic findings, two clinicians having access to all available data were requested to determine in consensus, the most significant with regard to CRC. For example, an individual with both a >6 mm colonic polyp and a non-neoplastic colonic mass accounted, for only the first. In addition, they were requested to establish a potential correspondence between the clinical symptoms and CTC colonic and significant extra-colonic findings (i.e.: E4 grade of the CTC Reporting and Data System, unrelated to a colonic disease). Percentages are given with their 95% Confidence Intervals (CI). Continuous variables are compared using analysis of variance, and two-tailed t -tests are used for direct comparison between two groups. Pearson chi-square tests are used to compare proportions and percentages. We used a regression analysis to evaluate the impact of the group of symptoms, age and gender on the test-positive rate. A p -value of less than 0.05 denotes a statistical significance.

In conclusion, asymptomatic subjects, patients with “non-alarm” gastrointestinal symptoms and even young patients with “alarm” symptoms for CRC have nearly similarly low test-positive rates at CTC. Overall, the rate of colonic and extra-colonic findings increased with age, male gender and the presence of “alarm” symptoms for CRC. The high diagnostic value of CTC in all patient groups makes it the examination of choice in low-yield patients; who include asymptomatic subjects and those with “non-alarm” gastrointestinal symptoms.

In a recent meta-analysis evaluating the value of various symptoms for CRC in primary care, Jellema et al. found large heterogeneities in the sensitivity and specificity of most symptoms [6] . Rectal bleeding and unexplained weight loss have consistently higher specificity than the others, according to another recent meta-analysis [5] . Similarly to studies using OC, the test-positive rate at CTC in our study was higher in patients with these “alarm” symptoms than in asymptomatic subjects and those with minor gastrointestinal symptoms ( p  < 0.05). Although the average age was also significantly higher in the “alarm” symptoms group ( p  < 0.05), age was not a key determinant, since similar differences were observed in all age groups above 50 years. In terms of CRC diagnostic recommendations, a test-positive rate around 20% makes OC the procedure of choice in patients with “alarm” symptoms, owing to the high resection-rate. An exception to this rule, though requiring stronger evidence, may be patients aged less than 50 years. Their findings prevalence was 6/96 (6.25%; 95%CI 0.01–0.11), i.e.: higher than in age-matched patients, but comparable to the overall prevalence in the other groups. Due to the low specificity of the remaining gastrointestinal symptoms (alone or in combination), diagnostic recommendations currently fails to avoid OC facilities congestion by procedures yielding low rates of resection [9] , [10] , [11] , [22] , [23] . Exception made of melena and hemorrhoids, the “non-alarm” symptoms evaluated in this study are commonly described in patients with irritable bowel syndrome (IBS) which is a functional bowel disorder affecting 7–15% of the general population in the USA, women being twice more often afflicted than men [24] , [25] . Its diagnosis nicely illustrates the dichotomy between the daily management of gastrointestinal symptoms and evidence-based diagnostic recommendations. Indeed, the diagnosis of IBS is based exclusively on clinical criteria [26] , but, up to 75% of clinicians believe that it should be ascertained by exclusion of organic disease [27] . For this purpose, approximately half of the patients with known or suspected IBS have undergone at least one diagnostic OC procedure [28] . Lastly, it has been estimated that 25% of the OC procedures performed in the USA are for patients with “non-alarm” gastrointestinal symptoms, although the actual risk of CRC in these patients is not higher than in asymptomatic individuals in several studies reporting very close ranges of test-positive rate at OC among IBS (range 1.9–9.3%) and asymptomatic individuals (range 4.5–12.1%) [29] , [30] . Several reasons could be advocated for this dichotomy, including physicians and patient's anxiety, symptoms severity and the lack of evidence that other diagnostic procedures may be as accurate or cost-effective than OC. To our knowledge, our study is the first evaluating colonic findings according to gastrointestinal symptoms using CTC. This was possible as, since 2002, CTC has progressively become a routine procedure in our institution, acknowledged both by our gastroenterologists, general practitioners and patients. The ranges of test-positive rate in asymptomatic individuals and those with “non-alarm” symptoms were low and comparable to those reported using OC. In both groups, the test-positive rate increased roughly linearly with aging ( p  < 0.05). Given their relatively low test-positive rate, patients with “non-alarm” symptoms may indeed benefit from accurate noninvasive diagnostic procedures. In our series of 5329 patients with “non-alarm” symptoms, 4088 were older than 50 years of age: it may thus be emphasized that these symptoms were an incentive to comply with screening guidelines and should therefore be carefully sought. Although patients under 50 years of age and suspected of IBS should not undergo any colonic exploration, many of them unfortunately do in most institutions, including ours. In our series, this malpractice applies to 1241 patients (23% referred by gastroenterologists) for whom CTC has a low test-positive rate in good correlation with other series, but represent the most comprehensive and the less harmful option as compared to OC, the standard of reference. In this setting, our study confirmed that CTC has a high true-positive fraction and a low false-positive fraction for polyps and CRC in all groups [15] , [31] clearly advocating for its recommendation in patients with “non-alarm” gastrointestinal symptoms. Considering the recommendations guidelines for CRC, the main disadvantage of CTC, namely the inability to perform polypectomy, may raise cost-effectiveness issues. In our study, at least 420 additional OC were performed in the group of patients with “non-alarm” symptoms, while approximately 5000 potential diagnostic OC procedures were replaced by less expensive CTC procedures, putting obviously the cost-effectiveness balance in favor of a primary diagnostic procedure by CTC in our institution. In addition, offering a same-day OC after positive CTC scenario currently prevent the need for a second bowel preparation in most institutions including ours [32] . The ability to detect clinically significant extra-colonic abnormalities (E4) may also impact CTC cost-effectiveness. A total number of 134 individuals had E4 findings, which represent an asset for CTC. Interestingly, the rate of E4 findings distribution per group was quite similar to the test-positive rate distribution, prevailing thus in older patients. This indicates that E4 findings rate are not either increased by nonspecific gastrointestinal symptoms. Lastly, the concern of exposure to ionizing radiation during CTC, has dropped significantly with the latest generation of scanners [33] . In our institution, the combined dose for supine and prone acquisitions of an average-sized subject is approximately 3 mSv (i.e.: levels comparable to the annual environmental scatter doses). Some limitations may be applied to this study and its conclusions, the first of which is its retrospective nature that resulted into several patients lost to follow-up and clinical data missing. Histopathological data were also missing in several patient from mis-retrieved polypectomies that are, however, not uncommon [33] . Reported symptoms were not based on a checklist of all gastrointestinal symptoms. The same design forced us to pool symptoms in two groups, resulting into potential masking of the effect of one or several underrepresented individual symptom(s) on the CTC test-positive rate. Moreover, age of symptoms onset, their intensity and duration lacked, preventing a better detailed relationship between symptoms and test-positive rate. Reported diagnostic values of CTC, though comparable with previously reported studies, are subject to verification bias, since OC was mostly performed when CTC was positive. In addition, most of the unverified CTC positive cases were small polyps, which probably decreased artificially the false-positive rate. Lastly, patients with at least one gastrointestinal symptom largely outnumbered asymptomatic subjects in our study, with 6929 out of the 7361 average-risk subjects for CRC having at least one symptom. This selection bias was caused by a progressive referral replacement of DCBE by CTC during the study period.

Colorectal cancer (CRC) is the second cause of cancer-related death [1] and generally results from the transformation of clinically silent adenomas [2] that are sought by screening tests [3] . Persistence or sudden occurrence of various abdominal symptoms is often considered an indication to search or rule out colonic abnormalities, including CRC or precancerous polyps [4] . Literature suggests that the use of optical colonoscopy (OC) is warranted only for subjects with rectal bleeding and unexplained weight loss [5] , whereas the other symptoms’ specificity remain questionable [6] , [7] , [8] . Meanwhile, the current diagnosis guidelines for individuals with average-risk for CRC only apply if there is no gastrointestinal symptom or complain [2] , raising potentially important concerns. Indeed, as long as all symptoms are considered equivalent in terms of diagnostic yield, individuals with nonspecific gastrointestinal symptoms are evaluated, when needed, by OC, causing potential congestion of the facilities by low resection-rate procedures [9] , [10] , [11] . Second, patient compliance to current CRC screening guidelines is low. Almost 50% of asymptomatic subjects 50 years of age and older escape screening programs over a period of 10 years [12] , while subjects with nonspecific gastrointestinal symptoms agree to undergo colonic explorations, for reassurance in a greater percentage [7] . Computed tomography (CT) colonography (CTC) has emerged over the past decade as an accurate and less invasive alternative to OC in series of symptomatic patients [13] , [14] . Similarly good results were obtained in series of asymptomatic subjects [15] . To our knowledge, there are little data evaluating the test-positive rate according to gastrointestinal symptoms at CTC in the literature. This has implications for risk-stratification and potentially impacts CRC screening recommendation. We therefore evaluate in this study, the distribution of clinically significant colonic findings and extra-colonic at CTC, according to symptoms and age through a review of a 7-year experience in a single non-academic center.