Comparison of three different iodine-based bowel regimens for CT colonography
- First Online:
- Cite this article as:
- Campanella, D., Morra, L., Delsanto, S. et al. Eur Radiol (2010) 20: 348. doi:10.1007/s00330-009-1553-9
The aim of this study was to compare the computed tomographic colonography (CTC) image quality and patient acceptance of three iodine-based faecal tagging bowel preparations in 60 patients undergoing the following regimens: a 2-day regimen of meal-time administration of iodine and phospho-soda (GFPH); a 2-day regimen of meal-time mild laxative, followed by iodine administered 2 h before CTC (SD); and a 2-day regimen of meal-time administration of iodine (GF).
Two independent radiologists assessed tagging quality; quantitative measures included the tagged stool density, and computer-aided detection (CAD) false-positive rate.
The GFPH and SD regimens provided better subjective quality than GF (p < 0.001). The latter regimen resulted in a higher proportion of insufficiently tagged segments: the measured average stool density was less than 200 HU in 10.7% in all segments vs 3.6% for SD and <0.5% for GFPH, respectively. Insufficient tagging occurred mostly in the ascending colon and the caecum. The CAD false-positive rate increased following the trend: GFPH < SD < GF (p = 0.00012). GFPH was worse tolerated than SD (p < 0.05).
Considering preparation quality alone, GFPH was the best regimen, but SD provided the best balance between bowel preparation quality and patient acceptability.
KeywordsCT colonography Faecal tagging Bowel preparation Quality assessment Computer aided detection
Computed tomographic colonography (CTC) has been recently considered as a valid option for colorectal cancer (CRC) screening in average risk subjects [1, 2, 3]. Compared with optical colonoscopy (CC), CTC is less invasive, thus providing a better-tolerated test with less side effects but a similar diagnostic accuracy for clinically significant lesions [1, 4]. However, CTC still requires a bowel-cleansing preparation, which is generally prescribed similarly to CC, using a strong and unpleasant purgation [1, 2, 3]. This may be a factor negatively affecting the appeal of the test for patients and their compliance, possibly limiting the efficacy of proposing CTC in screening programs [5, 6]. Oral tagging of fluid and faeces by administering a positive contrast medium has been proposed as an alternative to purgation-only regimens [7, 8]. Faecal tagging (FT) increases CTC performance by reducing both false-positive and false-negative findings, mainly caused by untagged residues, making it possible to visualize polyps “submerged” in the faecal residue [7, 8]. In addition, some authors have suggested that FT may reduce or abolish the need for purgation [9, 10, 11, 12], therefore improving exam tolerance [7, 13, 14].
Different FT regimens have been proposed, using barium- or iodine-based products, or a combination of the two [1, 2, 10, 11, 14, 15]. Barium is effective in tagging faeces [7, 16]; however, barium-based products are insoluble in water, thus the tagged faeces do not mix with the fluid and tend to lie at the bottom of the pools . The non-homogeneous tagging thus obtained may limit the efficacy of digital subtraction software and of computer-aided detection (CAD) algorithms . The administration of hyperosmolar iodine-based regimens provides instead a homogeneous tagging of both the fluid and solid residue, and determines fluid secretion into the colon, thus creating a smooth liquid-to-air interface that improves the performance of digital subtraction software . However, oral iodine administration may cause adverse events, such as nausea, vomiting, skin eruptions [17, 18] and hyperosmolar diarrhea occurs in approximately 10% of cases .
To date there is no consensus on which is the best way to tag the bowel residue, the quantity and timing of contrast-medium administration, and whether and how strong laxatives are also required to obtain a good quality tagging . Moreover, there are several limitations in the ways used to assess the quality of tagging: most methods proposed so far are either subjective [2, 7, 9, 10, 20, 28] or are based on the manual selection of a limited number of regions of interest [14, 16, 21].
The aim of this study was to compare the image quality and the patient acceptance of three iodine-based FT regimens, differing in the total amount of contrast medium administered, administration timing, and additional use of laxatives.
Materials and methods
Viewing software and technical support for the study were provided by im3D, Torino, Italy. Non-consultant authors had full control of the data and information submitted for publication. The study was conducted in accordance with the Helsinki declaration and national legislation. The local ethical committee authorized the use of anonymized CTC datasets for this study.
Study design and patient population
Patient demographics by preparation group (GFPH phospho-soda and iodine regimen, GF iodine solution only, SD same-day faecal tagging, M male, F female, symp subjects with alarm symptoms (i.e. positive FOBT, blood in faeces, refractory iron deficient anemia, abdominal pain, alternating bowel, constipation or diarrhea), asymp asymptomatic subjects with increased CRC risk for family or personal history enrolled in a prospective clinical trial)
Mean age ± standard deviation
Indication to CTC, symp:asymp
54.8 ± 9.1 (range 39–76)
68.3 ± 9.4 (range 49–89)
66.4 ± 9.4 (range 40–83)
Bowel regimens and CTC image protocol
- Regimen 1
Phospho-soda and iodine. An iodine-based solution was prepared by diluting a commercially available solution of 100 ml sodium diatrizoate and meglumine diatrizoate (iodine concentration 370 mg/ml; total iodine load 37 g; Gastrografin, Bayer Schering Pharma, Milan, Italy) in 1.5 l water. Subjects were asked to ingest 250 ml of the solution at each of the three main meals starting at lunchtime 2 days before CTC; the last dose was administered 2 h before CTC. At 8 am and 18 pm of the day before CTC, patients were asked to ingest a total of two 45 ml phospho-soda vials (Fosfo-Soda Fleet, Bergamon, Rome, Italy).
- Regimen 2
Mild purgation with same-day faecal tagging (SD). Patients were requested to ingest one sachet of a Macrogol 3350-based mild laxative (Movicol, Norgine Italia, Milan, Italy) at each of the three main meals, starting 2 days before CTC. On the day of CTC, patients ingested a 50-ml dose of sodium diatrizoate and meglumine diatrizoate solution (total iodine load 18.5 g) diluted in 500 ml water 2 h before CTC. To avoid artefacts caused by the high concentration of the residual contrast material in the gastric lumen, patients were also asked to drink an additional 500 ml water.
- Regimen 3
An iodine-based solution was prepared according to the patient’s body weight, by diluting 2 ml/kg of sodium diatrizoate and meglumine diatrizoate in 2 l water, up to a maximum of 150 ml (maximum total iodine load 55.5 g). The solution was administered in approximately 330 ml doses at each of the three main meals, starting from breakfast time 2 days before CTC and ending at dinnertime of the day before. Laxatives were not administered.
The same CTC technique was used in all cases. Patients were placed onto the CT table and a small flexible rectal catheter was positioned. Immediately before imaging, pneumocolon was obtained by insufflating CO2 by means of an automatic device, until maximum patient tolerance was reached. N-butyl-scopolamine (Buscopan; Boehringer Ingelheim, Italy) was administered intravenously if colon distension was insufficient and/or if the patient complained of pain during insufflation, which occurred in approximately 20–30% of patients in each group. CTC was carried out in all cases with a 16-slice image CT system (LightSpeed 16; GE Healthcare, Milwaukee, Wis., USA) in the supine and prone positions with the following imaging protocol: 120 kVp, ≤50 effective mAs, section thickness 2.5 mm or less and reconstruction interval 1.25 mm.
Assessment of preparation quality
Qualitative assessment of CTC preparation. Scores were assigned by two experienced CTC readers. When no fluid or solid residues were visible, the segment was labelled as empty. Ratings were combined as follows: segments labelled as empty by both readers were assigned the highest quality score, whereas segments labeled as non-empty by both readers were assigned the average of the two ratings; segments that were classified as non-empty by one reader were assigned the available score, rather than being classified as empty. Average per-segment scores of 2 or less were considered of non-diagnostic quality
Untagged fluid and/or untagged solid residues
Partially untagged fluid and/or partially untagged solid faeces that limit colonic wall assessment
Partially untagged fluid and/or partially untagged solid faeces that do not limit colonic wall assessment
The number and percentage of insufficiently tagged segments was evaluated within each preparation. Tagging was considered insufficient if the MSD value was below 200 HU. In the literature, density thresholds between 150 and 300 HU have been proposed to identify sufficiently tagged stool [9, 22]; an intermediate value was therefore selected as a cut-off for our study.
Relative standard deviation (RSD) was employed as a measure of tagging homogeneity within each segment. Only sufficiently tagged segments (i.e. those with MSD value greater or equal to 200 HU) were included in the assessment of the RSD. In fact, if the average intensity is too low, colonic mucosa cannot be distinguished regardless of the standard deviation; homogeneity of truly tagged material can thus be more accurately assessed if poorly tagged segments are excluded. We also used CAD to quantitatively assess the bowel preparation. Based on the assumption that CAD algorithms show a higher number of candidates when tagging is inhomogeneous , we hypothesized that the number of CAD prompts is related to preparation quality and have used this experimental measure to compare the three regimens. A CAD system (im3D, Torino, Italy; commercially available world-wide, although currently not available in the USA) was used to assess the number of false-positive prompts for each patient. The employed CAD system is based on the following steps: electronic cleansing, automatic segmentation of the cleansed colon, polyp candidate segmentation through curvature-based features and discrimination between true polyps and false alarms. A more detailed description of the CAD algorithm is available elsewhere . The employed polyp candidate segmentation and classification algorithm was developed on non-FT cases, hence its performance is greatly dependent on the quality of the electronically cleansed images and on the number of untagged residues. Measures were taken by setting the CAD’s operating point to a sensitivity of 87% for polyps measuring at least 6 mm.
Immediately before performing CTC, all patients were routinely interviewed in order to assess their tolerance to bowel preparation. For the assessment, we used an 11-point visual analogical scale (VAS), described elsewhere , where a value of 0 corresponded to no discomfort and 10 to a situation of extreme discomfort giving a strong negative impact on the patients’ daily activities. Intermediate values corresponded to intermediate values of tolerance.
Qualitative parameters were compared using the Kruskal-Wallis one-way analysis of variance on a per-segment and per-patient basis [25, 26]. At first, the three regimens were compared to verify whether they were significantly different; if differences were found, then multiple comparisons across pairs were performed in order to identify the sources of the observed differences. On a per-segment basis, the null hypothesis of identical medians for all regimens was tested against the alternative hypothesis that the medians were not equal.
The Kruskall-Wallis test was also applied to each regimen to verify if the tagging quality was homogeneous throughout the colon. Here the null hypothesis was of identical median for all segments, against the hypothesis that the medians were not equal. A correction factor for ties was applied to all tests. As multiple tests were performed on the same set of data, we applied the Bonferroni principle to correct for multiple comparisons. When significant differences were observed, pair-wise multiple comparisons were performed to identify underlying differences, at the α level of 0.05 after correcting for multiple comparisons, using the Tukey test adapted to non-parametric tests.
Inter-rater agreement was analysed employing the weighted κ coefficient with linear weights, on a per-segment basis. As the number of segments evaluated was high compared with the number of rating categories, asymptotic confidence intervals were calculated. Although interpretation of κ depends on the number of categories and their relative prevalence, values above 0.61 generally indicate a substantial agreement .
The Kruskal-Wallis analysis of variance was also employed to evaluate differences in MSD values, analogous to the qualitative ratings, as well as to test for within-regimen differences in the IDR. Finally, the Jonchkeere-Terpstra test was applied to evaluate CAD results . The null hypothesis that the median numbers of prompts were all equal was tested against the alternative hypothesis that a trend exists between groups so that the median number of prompts is higher for GFPH than for SD, and for SD than for GF. This hypothesis was supported by preliminary data analyses and by our clinical experience. A p level of 0.05 or lower was considered statistically significant. All distributions were characterized either by showing box plots or by indicating the median and the inter-quartile range (IQR).
Distribution of qualitative ratings across segments and preparations. The numbers are the median (interquartile ranges are expressed within parentheses)
The inter-observer per-segment agreement was κ = 0.62 (C.I. 95%: 0.58–0.66).
The median IDR values were 147 HU (IQR: 133 HU), 168 HU (IQR: 167 HU) and 373 HU (IQR: 495) respectively for GFPH, SD and GF. The IDR differences between the regimens were significant (p < 0.001). GF was found to have a greater IDR value than both SD and GFPH (p < 0.05); no differences were observed between GFPH and SD (p > 0.05).
In this study, we compared the quality of three iodine-based CTC preparations. The best results in terms of bowel cleanliness have been obtained using GFPH, a regimen requiring the administration of phospho-soda and a fractioned dose of a hyperosmolar iodine solution. A similar regimen, with the addition of barium, has been previously used as bowel preparation for CTC in the USA [1, 2]. GFPH provided a homogeneous pattern of faecal tagging, with a total administered volume of iodine smaller than other protocols , but it rated as the worst accepted among the three regimens which we have evaluated. This may be explained by the presence of a strong laxative, as patient acceptance has been shown to decrease by increasing the laxative strength [11, 21]. Furthermore, phospho-soda is now being used less frequently, as possible kidney damage has been reported following its administration, prompting a warning from the FDA in December 2008 to make the drug available to patients only under medical prescription. Other laxatives, such as magnesium citrate or PEG, may be used in alternative. SD proved instead to be a regimen with the best balance in terms of quality of bowel preparation and tolerability. Inhomogeneous faecal tagging was present in only 10% of cases, mostly in the sigmoid colon and rectum, while insufficient fluid tagging (e.g. density below 200 HU) was reported in only 4% of cases, again mainly in the distal colon and rectum. An insufficient tagging could be overcome by increasing the time interval between oral administration and scanning, and possibly also by increasing the total iodine load. SD has several advantages over other faecal tagging regimens, explaining its good tolerability. First, the iodine solution can be administered in a protected environment, making it possible to deal promptly with any severe adverse reactions, though we have never experienced them and their incidence is rarely reported [17, 18]. Second, it requires only a small iodine load. Third, the proposed association with a Macrogol-based mild laxative is well tolerated. No side effects were reported in a recent study on 132 asymptomatic patients after administration of this mild laxative . Finally, the patients do not need to collect the contrast medium in advance, as it is dispensed to them when they come to hospital for CTC. The SD regimen seems, therefore, to fulfill the criteria for providing an optimal balance between tolerance and safety, which are key factors for increasing compliance to screening programs. However, compliance to CTC-based screening programs has never been assessed, thus the effect of different bowel preparation regimens on the subject’s compliance still remains unknown. The iodine-based GF regimen resulted in an unsatisfactory faecal tagging, especially in the proximal and distal colon segments. This may have probably been due to an inappropriate timing of contrast administration.
There are several limitations to this study, which are discussed below.
The three regimens evaluated in this study are by no means comprehensive for all possible faecal tagging bowel preparations for CTC, and several other combinations may be explored, with variable iodine and/or barium content, agent osmolarity, administration timing and type of laxative. Experimenting with a large number of different regimens would, however, entail large-scale testing on a multi-centre basis. However, our results provide important information for best focusing on the search for the optimal bowel preparation for CTC. (1) Preparation regimens should always include administration of contrast material on the morning of test, to guarantee adequate faecal tagging in the proximal colon. Similar recommendations have been reported previously . (2) A mild laxative is sufficient to obtain a good quality faecal tagging. Administering a strong laxative, while effective, reduces overall CTC tolerance and its use should be probably limited to subjects with constipation or in selected clinical cases. Conversely, if laxatives are not added to the preparation, faecal tagging may be inadequate. (3) The total iodine content of preparation regimens depends on the administration schedule. Less than half the total quantity of iodine has been used for the SD scheme with respect to the other two fractioned regimens in order to obtain comparable results, probably because in the latter regimens the contrast medium has been partially evacuated before the CTC examination. Finally, administering lower volume of contrast material may reduce adverse events such as diarrhoea.
This is a retrospective study, but to minimize this possible bias we have randomly selected cases in order to reproduce conditions similar to a randomized study. We did not investigate the degree of side effects for the three regimens, and events such as diarrhea were not quantified. As an attempt to estimate this, however, we have correlated the tolerance with the different intrinsic characteristics of the regimens and their laxative strength.
We did not evaluate sensitivity and specificity in detecting colorectal polyps in relation to the different regimens. Despite the large number of CTC exams evaluated, there were not enough polyps in each group for providing the study with a sufficient statistical power for this purpose. The quantitative density measurement which we used was semi-automatic, considering that heterogeneous and low-intensity tagged pools can only be segmented manually. Therefore, our method is not entirely reproducible and is time consuming. However, to our knowledge this is the first time that measurements have been extended to the entire colon and not limited to a few randomly selected slices [14, 21]. If fully automated density measurements were to become available in the future, it would be possible to design personalized FT regimens by taking into account physical parameters such as body weight and size or the individual patient’s bowel habits. To obtain a fully automated assessment, we have tested our dataset with a CAD system. However, even these results may not be reproducible with other CAD algorithms and operating points. Quantitative and qualitative ratings could not be compared because they measure different aspects of preparation: the faecal tagging quality in the latter and the density of the tagged faeces the former. However, the proportion of unsatisfactory quality were not significantly different on a per-segment basis.
Finally, we did not consider barium tagging agents in our study. Although it may be relevant to compare barium and iodine agents, possibly in a randomized study, our main focus was on timing, contrast medium load and use of laxatives.
In conclusion, meal-time administration of a mild laxative for 2 days, followed by a hyperosmolar iodine tagging agent orally administered 2 h before the CTC study, provides the best balance between quality of bowel preparation and patient acceptability, compared with other iodine-based regimens differing in administration timing, iodine load, and laxative strength and use. Further studies should include the quantitative evaluation of the preparation quality, but fully automated methods should be developed to guarantee reproducibility.
We thank im3D S.p.A. (Torino, Italy) for providing viewing software, technical support and statistical consultancy for this study.