Keywords
virtual colonoscopy - CT colonography - optic colonoscopy - colon cancer - diagnosis
Palavras-chave
colonoscopia virtual - colonografia por tomografia computadorizada - colonoscopia
ótica - câncer colorretal - diagnóstico
Introduction
The number of cancer cases has been increasing exponentially, which could be attributed
to technological advancements and lifestyle changes. The World Health Organization,
through its International Agency for Research on Cancer, released the latest World
Cancer Report (WCR) in 2014, which clearly showed that cancer is currently a serious
public health problem (Stewart & Wild, 2014). The WCR 2014 is the latest in a series
of 3 reports on the global status of neoplastic diseases. The first, in 2003, estimated
that 5.3 million men and 4.7 million women develop cancer annually. Eleven years thereafter,
as described in the WCR 2014, the number of people living with cancer has increased
to ∼ 14 million per year. In addition, the report found that the number of new cases
would likely increase worldwide by around 70% in just 2 decades.[1]
In Brazil, the number of cancer cases has become alarming. For the 2018 to 2019 biennium,
600,000 new cancer cases were estimated. In Goiás, Brazil, the same document estimated
17,810 new cases, which included all types of cancer.[2] Except for non-melanoma skin cancer, the most prevalent type of cancer in men is
prostate cancer (68,220 new cases), followed by lung cancer (18,740 new cases), and
colorectal cancer (17,380 new cases). In women, the most frequent types are breast
cancer (59,700 new cases), colorectal cancer (18,980 new cases), and cervical cancer
(16,370 new cases).[2] Thus, colorectal cancer has a high incidence in both genders. In addition, this
finding is also observed worldwide; according to the WCR 2014, colorectal cancer accounts
to ∼ 10% of all cancers globally. Moreover, colorectal cancer is the fourth most common
cause of cancer-related death worldwide. Therefore, this type of cancer has not only
a high incidence but also a high mortality rate.[1]
Furthermore, the known etiological association of colorectal cancer with modern sedentary
lifestyle and poor health habits, such as smoking, alcohol intake, and a poor diet,
is also worrisome. A diet rich in calories, animal fat, and meats, together with physical
inactivity, increases the chance of developing colorectal cancer.[3] This could explain the fact that > 65% of new cancer cases occur in developed countries,
where such lifestyle predominates.[4]
There are two main types of colorectal cancer precursor lesions that, if identified
early, may prevent colorectal cancer: conventional adenomatous polyps and serrated
lesions. Adenomas account for 70% of colorectal cancers and are classified as villous,
tubular, and tubulovillous. Presumably, progression of these adenomas to carcinomas
takes on average 10 years in people with moderate risk and 1 to 2 years in those with
familial syndromes, such as Lynch syndrome. By definition, adenomas are dysplastic
alterations, that is, tissue proliferation with loss of differentiation; adenomas
with a high degree of dysplasia are considered advanced and are > 1 cm in size.[5] Serrated lesions account for ∼ 30% of colorectal cancers and are classified as hyperplastic
polyps without oncogenic potential or serrated polyps. They are histologically different,
as they are rarely dysplastic; they have few vessels on their surface and tend to
be flatter lesions. In addition, their distribution is more prevalent in the right
colon and; thus, they are often underdiagnosed, with incomplete colonoscopies or sigmoidoscopies.[5]
Considering the epidemiological and evolutionary characteristics of the disease, early
detection and removal of colorectal cancer precursor lesions could result in better
outcomes and a significant reduction in the incidence and mortality of this type of
cancer. Hence, new detection approaches have been developed, which promise better
outcomes. The American Cancer Society (ACS) published its first screening recommendation
in 1980. Subsequently, recommendations for moderate-risk individuals were provided
in 2008, which were based on a consensus involving the ACS, the American College of
Radiology, and the US Multi-Society Task Force on Colorectal Cancer (representing
the American College of Gastroenterology, the American Gastroenterological Association,
and the American Society for Gastrointestinal Endoscopy). In 2018, new recommendations
were published, reinforcing the idea that screening and early detection of colorectal
cancer should be an effort undertaken by individuals in the population and should
be included in basic health care.[6]
According to the current recommendation, screening should be initiated preferably
at the age of 45 years in those with a moderate risk of colorectal cancer using highly
sensitive fecal tests or structural (visual) examinations. Moderate-risk individuals
are those who have no history of adenomatous polyps or colorectal cancer; no increased
risk for cancer either by family history or hereditary syndromes related to colorectal
cancer, such as familial adenomatous polyposis or Lynch syndrome; no history of pelvic
or abdominal radiation for a previous cancer; and no history of inflammatory bowel
disease. Moreover, screening should be maintained in those aged 75 years who are in
good health and have a life expectancy of > 10 years and should be discontinued in
those aged ≥ 85 years.[6] In addition, according to the ACS recommendation, the following fecal screening
tests should be considered: annual fecal immunochemical test, annual high-sensitivity
guaiac-based fecal occult blood test, and fecal DNA test every 3 years. Among the
structural (visual) examinations, colonoscopy should be performed every 10 years,
computed tomography (CT) colonography every 5 years, and flexible sigmoidoscopy every
5 years. It is further recommended to perform these structural tests when fecal tests
show unfavorable results.[6]
Conventional colonoscopy, or optic colonoscopy, remains the gold standard in evaluating
colorectal cancer precursor lesions as biopsy could be performed when assessing suspicious
alterations in the colonic mucosa. Nevertheless, new approaches have emerged that
have aimed at increasing the number of people that adhere to the screening. Among
these new approaches, CT colonography appears to be a sound choice and is one of the
structural examinations for colorectal cancer screening since 2008.[6]
Computed tomography colonography involves the acquisition of tomographic images that
could be evaluated in two or three dimensions using specific software. The image is
reconstructed to create a view that is similar to the image as seen using the optical
colonoscope. However, it has disadvantages, such as radiation (which is claimed to
be cancer inducing), difficulty in evaluating flat lesions, and the possibility of
colonic perforation, which is common in conventional colonoscopy. Nonetheless, as
an advantage, conventional colonoscopy does not require sedation and allows for extracolonic
findings during the same examination.[6]
Therefore, the present study aimed to evaluate the sensitivity and specificity of
CT colonography in the detection of colorectal polyps through a systematic review
and meta-analysis.
Methods
The paper search and data collection were performed using a search protocol with the
following criteria: subject interest, inclusion criteria, search strategy and data
selection, data analysis, and result presentation and interpretation. Indexed papers
from two available databases (PubMed and Web of Science), including textual descriptions,
were analyzed using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses
(PRISMA) recommendations. We used the following terms to search in PubMed and Web
of Science: screening AND virtual colonoscopy AND colorectal cancer. Only papers in English and published from November 2007 to November 2017 were included
in the present study. References from revision papers and consensus were manually
searched to ensure the inclusion of all relevant papers. No contact with clinical
investigators to verify research in progress was made.
Eligibility Criteria
The inclusion criteria were as follows: (1) study evaluating the screening of a population
with a medium risk of colorectal cancer, that is, those with colon polyps that were
identified and biopsied by colonoscopy (which is the most accurate tool to diagnose
colorectal neoplasia), including false-positive, false-negative, true-positive, and
true-negative values; (2) studies that employed laxative preparation of the colon
(without restriction of the type of laxatives) and that use stool tagging (24 hour
prior to preparation), fluid tagging (during the examination day), and subcutaneous
glucagon, with or without venous contrast; (3) studies that used tomography for distended
intestine due to ambient air or CO2, with images acquired in the supine and prone positions and with at least 16 tomographic
cuts; and (4) studies that obtained 2D and 3D images that were analyzed using suitable
software.
The exclusion criteria were as follows: (1) studies that did not compare the results
of CT colonography with those of conventional colonoscopy; (2) studies that evaluated
other variables (e.g., cost-benefit, patient acceptance, and types of laxative); (3)
studies that sought to identify other abnormalities (e.g., intestinal obstruction,
extracolonic findings, postoperative complications); and (4) studies that analyzed
colorectal cancer that is not polyp-related, symptomatic patients (as this study's
objective was to evaluate CT colonography as a screening method), or those with inflammatory
bowel disease.
Study Selection, Data Extraction, and Quality of Evidence
Two researchers performed all the search process from localization to paper selection
independently. Relevant data were extracted and differences resolved by consensus.
Potentially eligible articles were obtained and read in full. A third researcher was
involved to clarify any doubt about the inclusion of a particular paper.
Study quality was assessed using the Grading of Recommendations, Assessment, Development,
and Evaluations (GRADE).[7] The quality of evidence of the studies was classified into four categories: high,
moderate, low, or very low.[8] We also analyzed the influence of possible conflicts of interest and any information
on ethical approval of the studies.[9]
Statistical Analysis
Initially, heterogeneity of the studies, which could strongly affect the results and
was defined as the diversity of the studies, was evaluated using the heterogeneity
χ2 test. The results frequencies of all articles were grouped into a single table, and
diversity was assessed using the heterogeneity χ2 test in 2 × 2 contingency tables to compare the different odds ratios (ORs) with
95% confidence interval.[10]
A p-value > 0.05 based on the heterogeneity χ2 test indicated that the null hypothesis was accepted, thereby confirming that the
studies were homogeneous. Thus, we used the fixed-effect tests, which assumed that
all studies point in the same direction. By contrast, a p-value < 0.05 based on heterogeneity χ2 test indicated diversity and heterogeneity of the studies. In this case, we used
the random-effects test or random testing, such as the DerSimonian-Laird test.[10]
Moreover, global association tests were employed to assess the significance of the
correlation between CT colonography and colorectal cancer diagnosis. To estimate the
efficiency of CT colonography in colorectal cancer diagnosis, the values of the fixed-
and random-effects tests from each study were combined and analyzed using BioEstat
5.0 (BioEstat, Belém, PA, Brazil).[11]
The ORs and 95% confidence intervals of both the fixed-effect tests and random effects
and the weight of the studies (individually and in combination) were calculated to
obtain an estimated combined effect. Studies with increased statistical power, that
is, those with larger populations and greater effect, have greater weight. In addition,
the tests developed forest plot-type graphics. The charts summarize in the same space
all information on the effect and the contribution of each study to the analysis.[10]
Results
Our literature search on PubMed yielded 830 articles, of which 761 were excluded after
reading the title and abstract because of irrelevance. Of the remaining 69 articles,
5 were excluded as they were not written in Portuguese or English, and another 57
were excluded for not meeting the inclusion criteria. A total of 7 articles[12]
[13]
[14]
[15]
[16]
[17]
[18] that had all the necessary data were included in the analysis ([Fig. 1]).
Fig. 1 Flow chart of the studies included in the metanalysis of the sensitivity and specificity
evaluations of virtual colonoscopy.
Most of the studies (85.7%) were performed in Europe, where CT colonography is more
advanced, and only one (14.3%) was performed in the USA ([Table 1]). In addition, only one of the articles was published after the year 2010, indicating
that the progress of the research on CT colonography coincides with the time it was
considered as a screening method. Data of a total of 1,872 patients were obtained,
out of whom 57.2% were male and 42.8% were female. Their ages ranged from 49 to 82
years (mean 59.7 years ± 5.3 years).
Table 1
Absolute frequency and relative percentage by gender; total number of participants;
minimum, maximum, and mean age; and standard deviation
|
Author
|
Year
|
Local
|
Gender
|
Total
|
Age
|
|
M
|
F (%)
|
F
|
F (%)
|
Min
|
Max
|
Mean
|
SD
|
|
Cornett
|
2008
|
USA
|
81
|
50.9
|
78
|
49.1
|
159
|
49
|
82
|
59.3
|
3
|
|
Graser
|
2009
|
Germany
|
171
|
55.0
|
140
|
45.0
|
311
|
50
|
81
|
60.5
|
7
|
|
Liedenbaum
|
2009
|
Netherlands
|
187
|
61.9
|
115
|
38.1
|
302
|
–
|
–
|
61.0
|
6
|
|
Sali
|
2010
|
Italy
|
30
|
61.2
|
19
|
38.8
|
49
|
–
|
–
|
60.5
|
–
|
|
Liedenbaum
|
2010
|
Netherlands
|
187
|
61.9
|
115
|
38.1
|
302
|
50
|
75
|
–
|
–
|
|
Heresbach
|
2011
|
France
|
131
|
51.8
|
122
|
48.2
|
253
|
50
|
67
|
57.2
|
–
|
|
Lefere
|
2013
|
Portugal
|
–
|
–
|
–
|
–
|
496
|
50
|
75
|
–
|
–
|
|
Combined
|
|
|
787
|
57.2
|
589
|
42.8
|
1872
|
49
|
82
|
59.7
|
5.3
|
Abbreviations: F, female; M, male; Max, maximum; Min, minimum; SD, standard deviation.
The estimated sensitivity of CT colonography for all studies was 88.4% (46.3–95.7%,
coefficient of variation [CV] = 28.5%); ([Fig. 2]). When grouped by lesion size, the data showed that the sensitivity was 82.5% (62.0–99.9%,
CV = 25.1%) for lesions up to 9 mm and 90.2% (64.0–100.0%, CV = 7.4%) for lesions > 9 mm.
Of the 7 articles, only 1 did not show data on polyps > 10 mm ([Table 2]).
Fig. 2 Sensitivity and specificity of virtual colonoscopy grouped by lesion size.
Table 2
Sensitivity and specificity of virtual colonoscopy grouped by lesion size
|
Author
|
Year
|
6–9 mm
|
> 9 mm
|
|
Sens.
|
95% CI
|
Spec.
|
95% CI
|
Sens.
|
95% CI
|
Spec.
|
95% CI
|
|
Cornett
|
2008
|
38.6
|
–
|
–
|
–
|
83.0
|
–
|
–
|
–
|
|
Graser
|
2009
|
91.3
|
79.2–97.6
|
93.1
|
89.3–95.9
|
92.0
|
74.0–99.0
|
97.9
|
95.4–99.2
|
|
Liedenbaum
|
2009
|
91.0
|
85.0–91.0
|
69.0
|
60.0–89.0
|
82.0
|
74.0–89.0
|
86.0
|
80.0–93.0
|
|
Sali
|
2010
|
95.5
|
77.2–99.9
|
51.9
|
32.0–71.3
|
–
|
–
|
–
|
–
|
|
Liedenbaum
|
2010
|
75.0
|
69.0–81.0
|
–
|
–
|
92.0
|
88.0–96.0
|
–
|
–
|
|
Heresbach
|
2011
|
88.0
|
62.0–98.0
|
91.0
|
76.0–98.0
|
92.0
|
64.0–100.0
|
97.0
|
86.0–100.0
|
|
Lefere
|
2013
|
98.1
|
88.6–99.9
|
91.0
|
87.8–93.4
|
100.0
|
84.0–100.0
|
98.1
|
96.3–99.0
|
|
Combined
|
|
82.5
|
62.0–99.9
|
79.2
|
32.0–98.0
|
90.2
|
64.0–100.0
|
94.7
|
80.0–100.0
|
Abbreviations: Sens., sensitivity; Spec., specificity; CI, confidence interval.
Regarding specificity ([Fig. 2]), CT colonography had an overall specificity of 73.6% (47.4–100.0%, CV = 37.5%).
For lesions up to 9 mm, the specificity was 79.2% (32.0–98.0%, CV = 22.9%); for lesions > 9 mm,
the specificity was 94.7% (80.0–100.0%, CV = 6.2%). No differences in sensitivity
according to lesion size were found (p = 0.0958); however, the specificity was higher for lesions > 9 mm (p < 0.0001) ([Table 2]).
No statistically significant difference in sensitivity was observed (p = 0.0958) when grouped by lesion size; an increase in lesion size did not increase
the sensitivity of the test ([Table 3], [Fig. 3]).
Table 3
Calculation of the odds ratio, 95% confidence interval, and weight of each study and
of the studies combined for the sensitivity of virtual colonoscopy grouped by lesion
size
|
Author
|
Year
|
Sensitivity
|
OR
|
95% CI
|
Weight
|
|
≤ 9 mm
|
> 9 mm
|
Lower
|
Upper
|
|
Cornett
|
2008
|
38.6
|
83.0
|
7.744
|
6.289
|
9.537
|
88.62
|
|
Graser
|
2009
|
91.3
|
92.0
|
1.095
|
0.798
|
1.503
|
38.41
|
|
Liedenbaum
|
2009
|
91.0
|
82.0
|
0.452
|
0.345
|
0.592
|
52.10
|
|
Liedenbaum
|
2010
|
75.0
|
92.0
|
3.817
|
2.917
|
4.995
|
53.10
|
|
Heresbach
|
2011
|
88.0
|
92.0
|
1.565
|
1.163
|
2.106
|
43.59
|
|
Lefere
|
2013
|
98.1
|
100.0
|
39.755
|
2.345
|
659.371
|
0.49
|
|
Combined
|
|
80.3
|
90.2
|
2.372
|
0.858
|
6.553
|
p
= 0.0958
|
Abbreviations: CI, confidence interval; OR, odds ratio.
Fig. 3 Forest plot showing the odds ratios (OR) and 95% confidence intervals of each study
and of the studies combined for the sensitivity of virtual colonoscopy.
However, an increase in specificity (p < 0.0001) when grouped by lesion size was found. The chance of diagnosing lesions > 9 mm
was 3 times greater than the chance of diagnosing lesions ≤ 9 mm ([Table 4], [Fig. 4]).
Table 4
Calculation of the odds ratio, 95% confidence interval, and weight of each study and
of the studies combined for the specificity of virtual colonoscopy grouped by lesion
size
|
Author
|
Year
|
Specificity
|
OR
|
95% CI
|
Weight
|
|
≤ 9 mm
|
> 9 mm
|
Lower
|
Upper
|
|
Graser
|
2008
|
91.3
|
92.0
|
3.455
|
2.103
|
5.677
|
15.58
|
|
Liedenbaum
|
2009
|
91.0
|
82.0
|
2.760
|
2.208
|
3.450
|
77.04
|
|
Heresbach
|
2011
|
88.0
|
92.0
|
3.198
|
2.095
|
4.882
|
21.47
|
|
Lefere
|
2013
|
98.1
|
100.0
|
5.106
|
3.088
|
8.444
|
15.18
|
|
Combined
|
|
92.1
|
91.5
|
3.171
|
2.671
|
3.765
|
P
< 0.0001
|
Abbreviations: CI, confidence interval; OR, odds ratio.
Fig. 4 Forest plot showing the odds ratios and 95% confidence intervals of each study and
of the studies combined for the specificity of virtual colonoscopy.
Discussion
Colorectal cancer has a high global and national incidence and high mortality rates,
and it is the fourth leading cause of cancer-related deaths. Thus, studies that could
further our knowledge on accurate diagnostic methods are necessary to reduce negative
outcomes.[4] Fecal occult blood screening is a method currently used by the public health care
system. However, despite having a relatively good sensitivity, it has a considerable
number of false positives. Currently, colonoscopy is almost always the selected diagnostic
test and is recommended only for high-risk patients or those with positive fecal occult
blood tests, according to the public health care system.[2]
History of colorectal cancer reveals that > 80% of the cases result from the progression
of an adenomatous polyp. Additionally, the adenoma-carcinoma progression occurs at
a slow rate. On average, the progression of benign lesions to malignant lesions occurs
in a period of 10 years, which is an ideal period for detection.[2] Hence, more precise and faster diagnostic methods are beneficial. The present study
aimed at evaluating the sensitivity of CT colonography as a diagnostic test and at
understanding whether its use is advantageous or not for the reduction of the incidence
and consequences of colorectal cancer worldwide.
Since its implementation in the 1990s, CT colonography has become a viable alternative
to conventional colonoscopy (or optic colonoscopy) for the screening of colorectal
cancer. It has the following advantages: examination without sedation; lower risk
of complications compared with conventional colonoscopy, such as intestinal perforation
during the examination; and better patient tolerance, thereby increasing patient adherence
to the test. Moreover, with CT colonography, detecting extracolonic alterations in
the abdominal or pelvic cavity is possible, which could in turn lead to new treatment
approaches depending on the attending physician.[19] Our study only analyzed the sensitivity and specificity of CT colonography and did
not consider the risk of complications or patient acceptance.
The comparative analysis in our study showed that the sensitivity of CT colonography
is similar to that of conventional colonoscopy for adenomas up to 9 mm (75–93%) and
those > 9 mm (89–98%). However, the specificity of optic colonoscopy (94%) is statistically
higher than that of CT colonography for polyps below < 9 mm. For larger polyps, the
specificity of CT colonography and that of conventional colonoscopy are statistically
close.[6] Thus, for conventional colonoscopy, polyp size has less influence on screening efficiency
as it relies on direct visualization of the lesion, which minimizes the interference
of fluids, feces, or intestinal anatomy alterations.
Our study was in agreement with other similar studies. For example, a previous large
metanalysis showed a sensitivity CT colonography ranging from 86 to 96% for adenomas > 6 mm.[20] Thus, CT colonography is a reliable option when optic colonoscopy is not available
or for patients who would not prefer the conventional method. Confidence in this method,
even for small lesions, is attributed to the slow development of the lesions. The
growth rate of adenomas is ∼ 3 to 4% per year, which indicates low chances of developing
significantly during the recommended screening interval.
Nonetheless, if a polyp is found and if it is impossible to exclude a potential malignancy
during CT colonography, conventional colonoscopy must be performed for histopathological
analysis and to determine the best treatment strategy.[21] This process could be performed on the same day. Therefore, hospital centers with
both radiological and endoscopic departments could guarantee full patient evaluation.
This study has limitations. The number of studies included in our metanalysis is small,
which could be because we use only two databases and only one database provided articles
that met our eligibility criteria. In addition, our restriction to include only articles
in Portuguese and English prevented us from including relevant studies in other languages.
Finally, during the development of our metanalysis, we noticed that most studies opted
for unspecific or poorly defined eligibility criteria, which resulted in the exclusion
of numerous articles.
Conclusion
The present metanalysis reports the sensitivity and specificity of CT colonography
in the detection of colorectal polyps. Overall, CT colonography has a sensitivity
and specificity of 76.7% and 73.3%, respectively. In addition, the specificity is
higher for lesions > 9 mm. Therefore, CT colonography has better specificity for larger
lesions.
Most of the studies analyzed in our study were conducted before 2010, which is about
a decade after the development of the technique using CT colonography and in agreement
with the period in which CT colonography started being indicated as a screening method
by the European and American guidelines. Therefore, further studies aimed at analyzing
the technique after further technological advancements are warranted, which could
lead to the development of more modern devices.
Highlights
-
Virtual colonoscopy has better specificity for larger lesions.
-
Virtual colonoscopy has become a viable alternative to conventional colonoscopy.
-
The sensitivity of virtual colonoscopy is similar to that of conventional colonoscopy.
