You are on page 1of 10

IJC

International Journal of Cancer

Fecal immunochemical test accuracy in familial risk colorectal


cancer screening
Ines Castro1, Joaqun Cubiella1, Concepcio
n Rivera1, Carmen Gonza lez-Mao2, Pablo Vega1, Santiago Soto1,
Vicent Hernandez , Felipe Iglesias , Mara Teresa Alves , Luis Bujanda5 and Javier Ferna
3 3 4
ndez-Seara1
1
Department of Gastroenterology, Complexo Hospitalario Universitario de Ourense, Ourense, Spain
2
Department of Clinical Analysis, Complexo Hospitalario Universitario de Vigo, Vigo, Spain
3
Department of Gastroenterology, Complexo Hospitalario Universitario de Vigo, Vigo, Spain
4
Research Support Unit, Complexo Hospitalario Universitario de Ourense, Ourense, Spain
5
Biodonostia Institute, Donostia Hospital, University of the Basque Country UPV/EHU, CIBERehd, San Sebastian, Spain

There is little information on fecal immunochemical test (FIT) in familial risk colorectal cancer (CRC) screening. Our study
assesses FIT accuracy, number needed to scope (NNS) and cost to detect a CRC and an advanced neoplasia (AN) in this setting.
We performed a multicentric, prospective, double-blind study of diagnostic tests on individuals with first-degree relatives (FDRs)
with CRC submitted to screening colonoscopy. Two stool samples were collected and fecal hemoglobin in the first sample (FIT1)
and the highest in both samples (FITmax) were determined. Areas under the curve (AUC) for CRC and AN as well as the best FIT1
and FITmax cutoff value for CRC were determined. At this threshold, NNS and the cost per lesion detected were calculated. A
total of 595 individuals were included (one FDR > 60 years, 413; two FDR or one  60 years, 182). AN and CRC were found in
64 (10.8%) and six (1%) patients, respectively. For CRC diagnosis, FIT1 AUC was 0.96 [95% confidence interval (CI): 0.950.98]
and FITmax AUC was 0.95 (95% CI: 0.930.97). For AN diagnosis, FIT1 and FITmax AUC were 0.74 (95% CI: 0.660.82). The best
cutoff point for CRC was 115. At this threshold, the NNS to detect a CRC was 5.67 and 7.67, and the cost per CRC was 1,064e
and 1591.33e on FIT1 and FITmax strategies, respectively. FIT shows high accuracy to detect CRC in familial CRC screening. Per-
forming two tests does not improve diagnostic accuracy, but increases cost and NNS to detect a lesion.

Colorectal cancer (CRC) is the third most common cancer which is the best screening strategy in individuals with a

Early Detection and Diagnosis


worldwide and the second leading cause of cancer deaths in FDR with CRC.7 Clinical practice guidelines recommend
developed countries.1 Several factors have been related to the more aggressive screening than in average risk population,
risk of developing CRC. Age, sex and CRC familial history, based on studies that have shown that endoscopic screening
especially if there are rst-degree relatives (FDR) with CRC, reduces CRC incidence and mortality in individuals with a
are the strongest associated risk factors.2 Furthermore, CRC FDR with CRC.2,8 Screening is recommended from 40 years
risk is directly related to the number of FDRs and inversely or 10 before the youngest proband. However, this approach
related to the age of youngest FDRs.3,4 has an empirical basis and no prospective controlled study
Evidence on the best screening strategy in this population has compared different screening strategies in this popula-
is limited and its quality is low.5,6 At present, it is unclear tion. Furthermore, colonoscopy is an invasive technique that

Key words: colorectal neoplasms, early detection of cancer, adenoma, occult blood, cost-benet analysis
Abbreviations: AN: advanced neoplasia; AUC: area under the curve; CI: confidence interval; CRC: colorectal cancer; FDR: first-degree
relative; FIT: fecal immunochemical test; gFOBT: guaiac fecal occult blood tests; LLR: likelihood ratio; NNS: number needed to scope;
NPV: negative predictive value; PPV: positive predictive value; ROC: receiver-operating characteristic
Conict of interest: Nothing to report
Presented at the 20th United European Gastroenterology Week, Amsterdam 2224 October 2012
Grant sponsor: Conselleria de Sanidade of Xunta de Galicia; Grant number: PS09/74; Grant sponsor: Instituto de Salud Carlos III; Grant
number: PI08/90717; Grant sponsor: Diputacion Foral de Gipuzkoa; Grant number: DFG 07/5; Grant sponsor: EITB-Maratoia; Grant
number: BIO 07/CA/19; Grant sponsors: Academia Medico Quir urgica of Ourense, Asociacion Espa~nola contra el Cancer (Fundacion
Cientca), Obra Social de Kutxa, Departamento de Sanidad del Gobierno Vasco, Accion Transversal contra el Cancer del CIBERehd (2008),
Direccion Xeral de Innovacion e Xestion da Sa
ude P
ublica, Conselleria de Sanidade, Xunta de Galicia
DOI: 10.1002/ijc.28353
History: Received 8 Jan 2013; Accepted 5 June 2013; Online 1 Jul 2013
Correspondence to: Dr. Joaqun Cubiella, Department of Gastroenterology, Complexo Hospitalario Universitario de Ourense, C/ Ramon
Puga 52-54, 32005 Ourense, Spain, Tel.: 134-988385399, Fax: 134-988385399, E-mail: joaquin.cubiella.fernandez@sergas.es

Int. J. Cancer: 134, 367375 (2014) V


C 2013 UICC
368 FIT diagnostic accuracy in familial CRC screening

Whats new?
A new study assessed the cost-effectiveness and accuracy of using the fecal immunochemical test, FIT, to detect colorectal
cancer. The test detects the amount of hemoglobin in a stool sample; higher levels indicate the presence of cancer. The
authors administered the test to people who had no symptoms yet, but did have a close relative with colorectal cancer and
had been referred for colonoscopy. They determined that the FIT has a high diagnostic accuracy for familial screening of CRC
and costs less per lesion detected than colonoscopy.

requires sedation, is associated with rare but serious compli- each patient, fecal hemoglobin (ng/ml) in the rst sample
cations (perforation or hemorrhage), requires a substantial (FIT1) and the higher level of the two samples (FITmax)
nancial investment in equipment and expertise and has a were determined.
limited adherence.911 Colonoscopy was performed blind to FIT result by experi-
CRC screening with guaiac fecal occult blood tests enced endoscopists (>200 colonoscopies per year). Bowel
(gFOBTs) has obtained poor results in this population.1214 cleansing, sedation and colonoscopy procedure were per-
However, fecal immunochemical tests (FITs) specic for formed according to the Spanish Guidelines on Quality of
human hemoglobin exhibit a higher sensitivity and specicity Colonoscopy in CRC screening.21 Polyps were categorized as
for the detection of advanced colorectal neoplasia compared non-neoplastic (hyperplastic and inammatory) or neoplastic
to gFOBT.1518 Furthermore, cutoff point can be modied on (adenomas). Adenomas of 10 mm or more in size with vil-
the basis of endoscopic resources available and advanced lous architecture, high-grade dysplasia or intramucosal carci-
colorectal neoplasia prevalence.19 A pilot study performed in noma were classied as advanced adenomas. Invasive cancer
our country shows a high sensitivity (83%) and specicity was considered when malignant cells were observed beyond
(91%) of FIT for detecting advanced adenomas in individuals the muscularis mucosa. AN was dened as advanced ade-
with a CRC familial history.6 noma or invasive cancer. Tumor staging was performed
So far, no data are available on FIT diagnostic accuracy in according to the AJCC classication.22 Patients were classied
familial CRC screening. The aim of our study was to evaluate according to the most advanced lesion.
FIT diagnostic yield for CRC and advanced neoplasia (AN)
in individuals with at least one FDR with CRC and to deter- Sample size calculation
mine diagnostic accuracy and the cost-benet of each strategy The sample size was calculated based on published data on
(one or two samples and different cutoff points).
Early Detection and Diagnosis

FIT sensitivity and specicity for AN in asymptomatic indi-


viduals (67% sensitivity and 91.4% specicity)23 and a esti-
Methods mated AN prevalence of 15% in FDRs with CRC.24 For a
Study design 95% condence level (CI) and 10% accuracy, 567 patients
This multicentric, prospective, double-blind, cross-sectional were required to be included.
study of diagnostic tests was performed in three tertiary hos-
pitals in Spain between January 2010 and December 2011.
Data analysis
The data were included in a specically designed database
Study population (www.coloncruzer.es). Continuous variables were described
Study population consisted of asymptomatic individuals with using means and standard deviation, and categorical variables
at least one FDR with histologically conrmed CRC consecu- by the absolute number and percentage value. Comparisons
tively referred to perform a colonoscopy as a CRC screening to identify differences in FIT1 and FITmax between individu-
method. Exclusion criteria included personal history of CRC, als with CRC, AN, advanced adenomas, nonadvanced adeno-
adenoma or inammatory bowel disease, previous colectomy, mas and no neoplasia were performed using nonparametric
hereditary CRC (Lynchs syndrome, hereditary familial CRC tests (MannWhitney U-test).
type X and polyposis associated to MYH biallelic mutation) To analyze FIT diagnostic accuracy, FIT1 and FITmax
colonoscopy within the past 5 years or nonacceptance of the receiver-operating characteristic (ROC) curves for the diagno-
study. sis of CRC and AN were drawn and the area under the curve
(AUC) was calculated. We determined whether there were
Study interventions statistically signicant differences between FIT1 and FITmax
All participants received an oral explanation and written using the chi-square test of homogeneity of areas. The best
instructions for collecting FIT samples. They collected two FIT1 and FITmax cutoff point for the diagnosis of CRC were
stool samples the week before the colonoscopy. FIT was determined with the Youden index. For each FIT testing
assessed using the automated OCsensorTM (Eiken Chemical, strategy, sensitivity, specicity, positive and negative predic-
Tokyo, Japan), without specic diet or medication restric- tive value (PPV and NPV) and positive and negative likeli-
tions. Samples were processed as previously described.20 In hood ratio (LLR) were calculated for the best cutoff value

Int. J. Cancer: 134, 367375 (2014) V


C 2013 UICC
Castro et al. 369

and for preestablished positivity thresholds (50, 75, 100, 150 Table 1. Baseline characteristics of individuals included in the study
and 200 ng/ml). Sensitivity and specicity for AN at the best Characteristics1 n 5 595
CRC detection cutoff point were compared with the rest of Age (years) 54.83 6 10.49
thresholds using McNemar test.25
Sex
Three additional analyses were performed. First, we
Male 245 (41.2%)
dened FIT sensitivity and specicity at the best cutoff point
according to AN location (distal to splenic exure and proxi- Female 350 (58.8%)
mal to splenic exure) and we compared them with McNe- Familial risk
mars test. Second, we determined if there were differences in 1 FDR > 60 years 413 (69.4%)
sensitivity and specicity for detecting AN at the best cutoff 1 FDR < 60 years 126 (21.2%)
point according to familial risk group (one FDR > 60 years,
2 FDR 52 (8.7%)
two FDR or one  60 years). Finally, we determined the pro-
Previous colonoscopy (yes) 121 (20.3%)
portion of hyperplastic polyps detected at the optimal thresh-
old as well as the ndings in the individuals with a false- Bowel cleansing
positive FIT result for AN. Excellent 226 (38.0%)
Fair 286 (48.1%)
Cost-benefit analysis Regular 34 (5.7%)
Finally, for each positivity threshold and strategy, the number Poor 37 (6.2%)
of subjects needed to scope (NNS) to detect a CRC or an AN Inadequate 12 (2%)
and the direct cost per lesion detected were determined. The
Complete colonoscopy (yes) 579 (97.3%)
direct costs were calculated on the basis of published colono-
Non-neoplastic polyps (yes) 89 (14.9%)
scopy costs in Spain (colonoscopy, 70e and colonoscopy with
biopsy, 140e)26 and FIT determination cost (3.2e). Hyperplastic (yes) 81 (13.6%)
The EPIDAT 3.1 software (Direccion Xeral de Innovacion Inflammatory (yes) 10 (1.7%)
e Xestion da Sa ude Publica, Santiago de Compostela, Spain) Any neoplasia (yes) 208 (35%)
was used to perform sample size calculation and ROC curve Nonadvanced adenoma (yes) 144 (24.2%)
drawings and comparisons. Statistical analyses were per-
Advanced adenoma (yes)2 58 (9.7%)
formed using the SPSS statistical software, version 15.0
Colorectal cancer (yes) 6 (1%)
(SPSS, Chicago, IL). A p-value <0.05 was considered statisti-

Early Detection and Diagnosis


cally signicant. Tumor stage3
Stage I 4 (66.6%)

Other aspects Stage II 1 (16.6%)


The study was approved by the Galician Clinical Research Stage III 1 (16.6%)
Ethics Committee (Codes 2009/123 and 2009/179) under res- Location
olutions dated May 28, 2009 and September 10, 2009. Rectum 2 (33.3%)
Patients clinical histories were accessed for study purposes in
Sigmoid colon 3 (50%)
accordance with the research protocols laid down by clinical
Transverse colon 1(16.6%)
documentation departments. Patients provided written
informed consent. Finally, QADAS quality assessment tool Advanced neoplasia (yes)4 64 (10.8%)
for diagnostic tests, STARD statement and STROBE checklist Distal to splenic flexure (yes) 49 (8.2%)
for cross-sectional study were used to write this original Proximal to splenic flexure (yes) 20 (3.4%)
article.2729
Qualitative variables are expressed as absolute numbers and percent-
age. Quantitative variables are expressed as mean and standard
Results deviation.
1
Individuals were classified according to the most advanced lesion.
Baseline characteristics 2
Advanced adenoma: adenoma > 1 cm in size, or with high-grade dys-
We evaluated 880 asymptomatic individuals with a CRC fam- plasia, or with villous component (>25%).
3
ily history referred for colonoscopy. Forty-four declined to According to AJCC classification.
4
Advanced neoplasia: advanced adenoma or cancer.
participate in the study and 201 individuals had exclusion Abbreviation: FDR: first-degree relative.
criteria: no FDR with CRC (160), personal history of CRC
(7), colonoscopy within last 5 years (31) and older than 80
years (3). A total of 635 individuals gave their written teristics of the study cohort as well as the number of FDRs
informed consent but 24 did not perform colonoscopy and with CRC.
16 did not return FIT. Finally, 595 (93.7%) subjects com- Colonoscopy detected a neoplastic lesion in 208 (35%)
pleted the protocol. Table 1 shows the demographic charac- individuals, in 58 (9.7%) an advanced adenoma and in six

Int. J. Cancer: 134, 367375 (2014) V


C 2013 UICC
370 FIT diagnostic accuracy in familial CRC screening

Table 2. Fecal hemoglobin concentration (ng/ml) according to the most advanced lesion
Strategy Colorectal cancer Advanced adenoma1 p2 Nonadvanced adenoma p3 No neoplasia p4
FIT15 444.5 6 472.3 406.53 6 8.33.7 0.05 25.4 6 123.6 <0.001 13.5 6 124.2 <0.001
FITmax6 656 6 983 487.1 6 910.8 0.09 37.6 6 149.6 <0.001 29.8 6 188.3 <0.001

Values expressed as mean 6 standard deviation. Differences with p < 0.05 are considered statistically significant.
1
Advanced adenoma: adenoma > 1 cm in size, or with high-grade dysplasia, or with villous component (>25%).
2
Significance of the fecal hemoglobin concentration differences between individuals with colorectal cancer and advanced adenomas in MannWhit-
ney U-test.
3
Significance of the fecal hemoglobin concentration differences between individuals with colorectal cancer, advanced adenomas and nonadvanced
adenomas in MannWhitney U-test.
4
Significance of the fecal hemoglobin concentration differences between individuals with colorectal cancer, advanced adenomas and no neoplasia
in MannWhitney U-test.
5
Fecal hemoglobin concentration in the first sample.
6
Higher fecal hemoglobin concentration in the two samples.
Abbreviation: FIT: fecal immunochemical test.

(1%) a CRC (Table 1). Overall, AN was detected in 64 sensitivity for AN detection ranged between 34.38 and 50%
(10.8%) patients. No differences were found in AN preva- and specicity between 92.66 and 98.72%. In only one deter-
lence according to the familial risk group: 9.9% in individuals mination strategy, statistically signicant differences in speci-
with one FDR > 60 years and 12.6% in individuals with two city between 115, 50 (p < 0.001) and 75 ng/ml (p 5 0.008)
FDRs or one FDR  60 years (p 5 0.32). Colonoscopy was were detected. In FITmax strategy, statistically signicant dif-
completed in 579 (97.3%) and with adequate bowel cleansing ferences in specicity were also found between 115, 50 (p <
in 512 (86.1%) subjects. 0.001) and 75 ng/ml (p < 0.001).
At a 115 ng/ml threshold, FIT was positive in 49% of the
Fecal hemoglobin concentration individuals with a distal AN and in 20% of the individuals
Patients with an AN had a fecal hemoglobin concentration with a proximal AN, reaching this statistically signicant dif-
statistically higher than those without an AN in both FIT1 ference (p < 0.001). On the contrary, we found no differen-
(422.5 6 814.1 ng/ml, 16.7 6 123.8 ng/ml; p < 0.001) and ces in specicity according to AN location (distal: 98.2%,
FITmax (518.3 6 921.5 ng/ml, 31.9 6 178.2 ng/ml; p < proximal: 94.8%; p 5 0.18). According to familial risk group
0.001). Fecal hemoglobin concentration was signicantly FIT sensitivity at a 115 ng/ml cutoff value did not differ
Early Detection and Diagnosis

higher among individuals with CRC or advanced adenomas between individuals with one FDR > 60 years and individu-
when compared to subjects with nonadvanced adenomas or als with two FDRs or one FDR  60 years (34.8 and 9.1%; p
without neoplasia. In contrast, no signicant differences were 5 0.21). We did not nd any differences in specicity at the
found between individuals with CRC and advanced adeno- same cutoff point either (93.3 and 92.4%; p 5 0.72). Finally,
mas as shown in Table 2. hyperplastic polyps were found in 26.5% of the individuals
with FIT  115 ng/ml and in 12.8% with FIT < 115 ng/ml
(p 5 0.03). Endoscopic ndings in individuals with a false-
FIT diagnostic accuracy analysis positive result for AN (9/34) at the optimal cutoff point were
Diagnostic accuracy was analyzed using ROC curves (Fig. 1). as follows: nonadvanced adenomas (5), hyperplastic polyps
For CRC diagnosis, FIT1 AUC was 0.96 (95% CI: 0.950.98) (3), inammatory polyps (1), diverticular disease (1) and
and FITmax AUC was 0.95 (95% CI: 0.930.97). This differ- hemorrhoids (2).
ence was statistically signicant in the chi-square homogene-
ity area test (p 5 0.025). For AN diagnosis, FIT1 AUC was
0.74 (95% CI: 0.660.82) and FITmax was 0.74 (95% CI: Cost-benefit analysis
0.660.82) without statistically signicant differences (p 5 When direct colonoscopy screening was analyzed, the NNS
0.86). The best cutoff value for CRC detection was 115 ng/ml to detect a CRC and an AN were 99.17 and 9.6, respectively.
both for FIT1 and FITmax. The NNS to detect a CRC or an AN decreased from the low-
Table 3 describes sensitivity, specicity, PPV, NPV, LLR est positivity threshold to the best cutoff value, and then rose
positive and negative to detect a CRC at different FIT1 and again. At 115 ng/ml, in the FIT1 strategy, the NNS to detect
FITmax positivity thresholds. As shown, both strategies show a CRC and an AN were 5.67 and 1.36. At the same cutoff
a high sensitivity and specicity for CRC, although FITmax point, in FITmax strategy, the NNS to detect a CRC and an
decreases specicity and PPV, without increasing the sensitiv- AN increased to 7.67 and 1.64, respectively, as shown in
ity or NPV. Table 4 describes sensitivity, specicity, PPV, Table 5.
NPV, LLR positive and negative to detect an AN at different Cost-benet analysis is displayed in Table 5. The cost per
FIT1 and FITmax cutoff points. Depending on the number CRC and AN detected in the direct colonoscopy screening
of determinations and the positivity threshold cutoff used strategy were 9998.33e and 967.58e, respectively. In contrast,

Int. J. Cancer: 134, 367375 (2014) V


C 2013 UICC
Castro et al. 371

detecting AN. Furthermore, FIT is cost-effective for CRC


screening.15,23,30 In recent years, several published studies
have analyzed FIT diagnostic accuracy in familial risk CRC
screening, although none of them were specically designed
to assess it.1619,23,3035 Our study conrms previous data
showing a high sensitivity and specicity for CRC at a 100
ng/ml positive threshold, reaching 100 and 90%, respec-
tively. However, diagnostic accuracy for AN was lower in
our study. Probably that was due to the inclusion of symp-
tomatic patients as in Levis study (sensitivity 74% and spec-
icity 93%).36 Our study also conrms high FIT specicity
for AN at a 50 ng/ml cutoff point determined in Terhars
study. Instead, they found a lower sensitivity probably
owing to a smaller sample size and inclusion of patients
with recent colonoscopy.37 Furthermore, in a pilot study
performed in Spain, FIT at a 50 ng/ml cutoff point showed
a high sensitivity (83%) and specicity (91%) for detecting
advanced adenomas in individuals with a family history of
CRC.6
Our main contribution is that we have specically eval-
uated FIT diagnostic yield for CRC and AN in asymptomatic
population with FDRs with CRC. The information we pro-
vide is relevant because the level of evidence on the best
strategy for CRC screening in this population is low. Based
on an increased risk of developing CRC and colonic neopla-
sia compared to the general population, more aggressive
screening strategies have been recommended.2,8,38 High FIT
Figure 1. Receiver-operating characteristics curves and area under diagnostic accuracy for CRC together with an adequate diag-
the curve of fecal immunochemical test for colorectal cancer (a) nostic performance for AN makes FIT a proper screening
and advanced neoplasia (b). 1Advanced neoplasia: advanced

Early Detection and Diagnosis


method in this population. However, FIT as a screening
adenomas (adenoma > 1 cm in size, with high-grade dysplasia, or
with villous component > 25%) or colorectal cancer. AUC: area method must be evaluated in prospective controlled studies
under the curve; 95% CI: 95% confidence interval. [Color figure designed to determine CRC mortality reduction in the long
can be viewed in the online issue, which is available at term.
wileyonlinelibrary.com.] Another contribution is that we analyzed in this popula-
tion different FIT strategies: one or two tests and different
in the optimal cutoff point, cost per CRC detected was positive thresholds not only to evaluate diagnostic accuracy
reduced between 89.3 and 84.1% and cost per AN detected but also endoscopic resources required and cost per lesion
was reduced between 73.65 and 64.76%, depending on the detected. Thus, analyzing two samples does not improve
number of FIT determinations. Finally, using two FIT deter- diagnostic accuracy and, instead, increases the costs by aug-
minations increased cost per CRC detected between 46.3 and menting the number of colonoscopies needed to detect a
53.2%, and cost per AN detected between 28.86 and 36.87% CRC or an AN. Cost-benet analysis on familial CRC screen-
when compared to only one FIT determination. ing is scarce and mainly based on colononoscopy.26,39 Our
cost-benet analysis may allow health authorities to dene
Discussion the recommended strategy for this cohort. However, our
Our study has determined that FIT has a high diagnostic cost-benet analysis must be put in context. For example, as
accuracy for CRC detection at a single round in familial CRC long as the diagnostic process expense mainly depends on
screening and the optimal cutoff point is 115 ng/ml. Also, colonoscopy, if colonoscopy cost increases, the reduction in
one round detects 39% of AN at this positive threshold with the cost per CRC or AN is also increased.
one FIT determination. Moreover, the performance of two Our study has several limitations. The main, common to
determinations does not increase the diagnostic accuracy and most of FIT diagnostic accuracy studies, is low CRC preva-
is associated with an increase in number of patients needed lence.17,19,30,32,33 Besides, as in the study by Gimeno et al.,
to scope and cost per lesion detected. most of the subjects included had only one FDR with CRC
Many studies have evaluated diagnostic accuracy of semi- older than 60 years.6 Thus, FIT as a screening CRC method
quantitative FIT and have demonstrated that FIT is highly should be specically evaluated in populations with higher
sensitive for CRC and has an adequate performance for CRC risk; i.e., two FDRs with CRC.3,4 Finally, as our study

Int. J. Cancer: 134, 367375 (2014) V


C 2013 UICC
Early Detection and Diagnosis

372

Table 3. Performance characteristics of FIT1 and FITmax for colorectal cancer detection at different positive thresholds
Individuals
Hemoglobin with Positive Negative Positive Negative
concentration FIT a positive predictive predictive likelihood likelihood
2 2
(ng/ml) strategy result1 Sensititvity Specificity value2 value2 ratio2 ratio2
3
50 FIT1 53 (8.90%) 100 (51.6898.45) 92.02 (89.4694.02) 11.32 (4.6923.72) 100 (99.1299.98) 12.53 (9.5316.49)
FITmax4 71 (11.93%) 100 (51.6898.45) 88.96 (86.0891.32) 8.45 (3.4818.11) 100 (99.0999.98) 9.06 (7.2011.40)
75 FIT13 44 (7.39%) 100 (51.6898.45) 93.55 (91.1795.34) 13.64 (5.6728.05) 100 (99.1499.98) 15.50 (11.4021.08)
FITmax4 61 (10.25%) 100 (51.6898.45) 90.66 (87.9592.83) 9.84 (4.0620.85) 100 (99.1199.98) 10.71 (8.3313.77)
100 FIT13 37 (6.21%) 100 (51.6898.45) 94.74 (92.5396.34) 16.22 (6.7732.68) 100 (99.1599.98) 19 (13.4926.76)
FITmax4 51 (8.57%) 100 (51.6898.45) 92.36 (89.8394.31) 11.76 (4.8724.56) 100 (99.1399.98) 13.09 (9.8817.33)
3
115 FIT1 34 (5.71%) 100 (51.6898.45) 95.25 (93.1296.76) 17.65 (7.3935.17) 100 (99.1599.98) 21.04 (14.6530.20)
FITmax4 46 (7.73%) 100 (51.6898.45) 93.21 (90.7995.05) 13.04 (5.4226.95) 100 (99.1399.98) 14.73 (10.9219.86)
3
150 FIT1 31 (5.21%) 83.33 (36.4899.12) 95.59 (93.5197.04) 16.13 (6.0934.47) 99.82 (98.8699.99) 18.88 (11.2431.72) 0.17 (0.031.04)
FITmax4 42 (7.05%) 83.33 (36.4899.12) 93.72 (91.3695.48) 11.90 (4.4726.43) 99.82 (98.8399.99) 13.27 (8.2521.33) 0.18 (0.031.06)
3
200 FIT1 29 (4.87%) 66.67 (24.1194.00) 95.76 (93.7197.18) 13.79 (4.5132.57) 99.65 (98.5999.94) 15.71 (7.9331.11) 0.35 (0.111.08)
FITmax4 39 (6.55%) 66.67 (24.1194.00) 94.06 (91.7595.77) 10.26 (3.3425.16) 99.64 (98.5699.94) 11.22 (5.8521.50) 0.35 (0.111.10)
1
Values are expressed as absolute numbers and percentage.
2
Values are expressed as percentage and its 95% confidence interval.
3
Fecal hemoglobin concentration in the first sample.
4
Higher fecal hemoglobin concentration of the two samples.
Abbreviation: FIT: fecal immunochemical test.

Int. J. Cancer: 134, 367375 (2014) V


FIT diagnostic accuracy in familial CRC screening

C 2013 UICC
Table 4. Performance characteristics of FIT1 and FITmax for advanced neoplasia1 detection at different thresholds
Individuals
Hemoglobin with Positive Negative Positive Negative
concentration FIT a positive predictive predictive likelihood likelihood
(ng/ml) strategy result2 Sensititvity3 p4 Specificity3 p5 value3 value3 ratio3 ratio3
6
50 FIT1 53 (8.90%) 46.88 (34.4559.67) 0.06 95.67 (93.4797.17) <0.001 56.60 (42.3669.90) 93.73 (91.2695.55) 10.82 (6.7117.44) 0.56 (0.440.70)
7
FITmax 71 (11.93%) 50 (37.3762.63) 0.12 92.66 (90.0194.66) <0.001 45.07 (33.4057.28) 93.89 (91.4095.72) 6.81 (4.6110.04) 0.54 (0.420.69)
75 FIT16 44 (7.39%) 42.19 (30.1655.15) 0.5 96.80 (94.8298.06) 0.008 61.36 (45.5175.25) 93.28 (90.7895.17) 13.18 (7.6122.81) 0.60 (0.480.74)
7
FITmax 61 (10.25%) 46.88 (34.4559.67) 0.5 94.16 (91.7295.94) <0.001 49.18 (36.2962.17) 93.63 (91.1395.49) 8.03 (5.2212.34) 0.56 (0.450.71)
100 FIT16 37 (6.21%) 40.63 (28.7553.62) 1 97.93 (96.2198.91) 0.5 70.27 (52.8383.56) 93.19 (90.6995.08) 19.61 (10.1837.77) 0.61 (0.490.74)
7
FITmax 51 (8.57%) 43.75 (31.5856.67) 1 95.67 (93.4797.17) 0.63 54.90 (40.45) 93.38 (90.8795.26) 10.10 (6.2116.43) 0.59 (0.470.73)
115 FIT16 34 (5.71%) 39.06 (27.3652.08) 98.31 (96.6999.17) 73.53 (55.3586.49) 93.05 (90.5494.95) 23.05 (11.2647.18) 0.62 (0.510.75)
FITmax7 46 (7.73%) 43.75 (31.5856.67) 96.61 (45.3974.54) 60.87 (45.3974.54) 93.44 (90.9595.30) 12.91 (7.5821.98) 0.58 (0.470.72)
150 FIT16 31 (5.21%) 35.94 (96.9399.30) 0.5 98.49 (96.9399.30) 1 74.19 (55.0787.46) 92.73 (90.1994.67) 23.85 (11.1451.08) 0.64 (0.540.78)
FITmax7 42 (7.05%) 42.19 (30.1655.15) 1 97.18 (95.2898.35) 0.25 64.29 (47.9978.00) 93.31 (90.8195.18) 14.93 (8.4026.55) 0.59 (0.480.73)
200 FIT16 29 (4.87%) 34.38 (23.2547.39) 0.25 98.72 (97.2799.44) 0.5 75.86 (56.0788.98) 92.81 (90.3294.71) 26.96 (11.9960.61) 0.66 (0.560.79)
FITmax7 39 (6.55%) 39.06 (27.3652.08) 0.25 97.36 (95.5198.49) 0.12 64.10 (47.1578.32) 92.99 (90.4594.90) 14.82 (8.1327.01) 0.63 (0.510.76)
1
Advanced neoplasia: advanced adenomas (adenoma > 1 cm in size, with high-grade dysplasia, or with villous component > 25%) or colorectal cancer.
2
Values are expressed as absolute numbers and percentage.
3
Values are expressed as percentage and its 95% confidence interval.
4
Significance of the sensitivity differences when compared to the optimal cutoff point in McNemar test. Differences with p < 0.05 are considered statistically significant.
5
Significance of the specificity differences when compared to the optimal cutoff point in McNemar test. Differences with p < 0.05 are considered statistically significant.
6
Fecal hemoglobin concentration in the first sample.
7
Higher fecal hemoglobin concentration of the two samples.
Abbreviation: FIT: fecal immunochemical test.

Early Detection and Diagnosis


374 FIT diagnostic accuracy in familial CRC screening

Table 5. Number of colonoscopies needed to detect one lesion and cost per lesion (e) according to positivity threshold and FIT testing
strategy
Cost per lesion
Number need to scope detected (e)
Positivity
threshold Cost increment
Lesion (ng/ml) FIT1 FITmax FIT1 FITmax (%)
0 99.17 9998.33
50 8.83 11.83 1390.67 2034.67 46.30
75 7.33 10.17 1239.00 1859.67 50.09
100 6.17 8.33 1134.00 1684.67 48.55
CRC
115 5.67 7.67 1064.00 1591.33 49.56
150 6.20 8.40 1206.80 1825.60 51.27
200 7.00 9.75 1421.00 2177.00 53.20
0 9.60 967.58
50 1.83 2.29 287.72 393.81 36.87
75 1.63 2.03 275.33 371.93 35.08
Advanced 100 1.42 1.79 267.69 361.00 34.85
neoplasia1
115 1.36 1.64 255.36 341.00 33.53
150 1.35 1.56 262.35 338.07 28.86
200 1.33 1.63 270.67 362.83 34.04
1
Advanced neoplasia: advanced adenomas (adenoma > 1 cm in size, with high-grade dysplasia, or with villous component > 25%) or colorectal
cancer.
Abbreviations: FIT: fecal immunochemical test; CRC: colorectal cancer.
Early Detection and Diagnosis

was only designed to evaluate FIT as a diagnostic test, we data, obtained accuracy is 12.64%, close to the expected in
cannot determine which is the best screening strategy in this the initial study design.
population. In conclusion, our study shows that FIT has an elevated
Moreover, our study aimed to reect routine conditions in diagnostic accuracy for CRC detection in CRC familial risk
clinical practice. For this reason, we have included incom- screening. Two FIT tests do not improve diagnostic accuracy
plete colonoscopies and with insufcient cleansing. However, for CRC or AN, although sensitivity for AN is increased at
it should be noted that our endoscopic examinations are the expense of increasing the NNS and cost per lesion
mostly adequate to the quality indicators described on the detected.
Spanish clinical practice guideline on quality of colonoscopy
on CRC screening.21 In fact, our adenoma detection rate,
35%, is clearly superior to the adenoma detection rate dened
Acknowledgements
CIBERehd is funded by the Instituto de Salud Carlos III. The study sponsors
as adequate in colonoscopy-based CRC screening pro- had no involvement in the collection, analysis and interpretation of data; in
grams.21,40 Finally, our study is limited by the fact that we the writing of the manuscript or in the decision to submit the manuscript for
did not achieve the expected sensitivity and prevalence of publication.
AN in sample size calculation. However, based on actual

References
1. Ferlay J, Shin H-R, Bray F, et al. Estimates of 3. Baglietto L, Jenkins MA, Severi G, et al. Measures lance in individuals with a family history of
worldwide burden of cancer in 2008: GLOBO- of familial aggregation depend on denition of colorectal cancer: 16 year, prospective, follow-up
CAN 2008. Int J Cancer 2010;127:2893917. family history: meta-analysis for colorectal cancer. study. Br Med J 2005;331:1047.
2. Levin B, Lieberman D, McFarland B, et al. J Clin Epidemiol 2006;59:11424. 6. Gimeno-Garca AZ, Quintero E, Nicolas-Perez D,
Screening and surveillance for the early detection 4. Butterworth AS, Higgins JPT, Pharoah P. Relative et al. Screening for familial colorectal cancer with
of colorectal cancer and adenomatous polyps, and absolute risk of colorectal cancer for individ- a sensitive immunochemical fecal occult blood
2008: a joint guideline from the American Cancer uals with a family history: a meta-analysis. Eur J test: a pilot study. Eur J Gastroenterol Hepatol
Society, the US Multi-Society Task Force on Cancer 2006;42:21627. 2009;21:10627.
Colorectal Cancer, and the American College of 5. Dove-Edwin I, Sasieni P, Adams J, et al. Preven- 7. Winawer S, Fletcher R, Rex D, et al. Colorectal
Radiology. Gastroenterology 2008;134:157095. tion of colorectal cancer by colonoscopic surveil- cancer screening and surveillance: clinical guide-

Int. J. Cancer: 134, 367375 (2014) V


C 2013 UICC
Castro et al. 375

lines and rationale-Update based on new evi- blood testing in screening for colorectal cancer. 31. Oort F a, Turenhout ST van, Coupe VMH, et al.
dence. Gastroenterology 2003;124:54460. Eur J Cancer 2001;37:398401. Double sampling of a faecal immunochemical test
8. Atkin WS, Edwards R, Kralj-Hans I, et al. Once- 20. Vilkin A, Rozen P, Levi Z, et al. Performance is not superior to single sampling for detection of
only exible sigmoidoscopy screening in preven- characteristics and evaluation of an automated- colorectal neoplasia: a colonoscopy controlled
tion of colorectal cancer: a multicentre rando- developed and quantitative, immunochemical, prospective cohort study. BMC Cancer 2011;11:
mised controlled trial. Lancet 2010;375:162433. fecal occult blood screening test. Am J Gastroen- 434.
9. Bujanda L, Sarasqueta C, Zubiaurre L, et al. Low terol 2005;100:251925. 32. Morikawa T, Kato J, Yamaji Y, et al. A compari-
adherence to colonoscopy in the screening of 21. Jover R, Herraiz M, Alarcon O, et al. Clinical son of the immunochemical fecal occult blood
rst-degree relatives of patients with colorectal practice guidelines: quality of colonoscopy in test and total colonoscopy in the asymptomatic
cancer. Gut 2007;56:171418. colorectal cancer screening. Endoscopy 2012;44: population. Gastroenterology 2005;129:4228.
10. Cottet V, Pariente A, Nalet B, et al. Low compli- 44451. 33. Haug U, Hundt S, Brenner H. Quantitative
ance with colonoscopic screening in rst-degree 22. OConnell JB, Maggard MA, Ko CY. Colon can- immunochemical fecal occult blood testing for
relatives of patients with large adenomas. Aliment cer survival rates with the new American Joint colorectal adenoma detection: evaluation in the
Pharmacol Ther 2006;24:1019. Committee on Cancer sixth edition staging. J target population of screening and comparison
11. Codori AM, Petersen GM, Miglioretti DL, et al. Natl Cancer Inst 2004;96:14205. with qualitative tests. Am J Gastroenterol 2010;
Health beliefs and endoscopic screening for colo- 23. Levi Z, Rozen P, Hazazi R, et al. A quantitative 105:68290.
rectal cancer: potential for cancer prevention. immunochemical fecal occult blood test for colo- 34. Nakama H, Yamamoto M, Kamijo N, et al. Colo-
Prev Med 2001;33:12836. rectal neoplasia. Ann Intern Med 2007;146:244 noscopic evaluation of immunochemical fecal
12. Krinstinsson J, Nygaard K, Aadland E, et al. 55. occult blood test for detection of colorectal neo-
Screening of rst degree relatives of patients 24. Ferrandez A, Thomson C, Blas JM. Riesgo de plasia. Hepatogastroenterology 1999;46:22831.
operated for colorectal cancer: evaluation of fecal neoplasia en familiares de primer grado de 35. Nakama H, Zhang B, Fattah a S. A cost-effective
calprotectin vs hemoccult II. Digestion 2001; 64: pacientes con cancer colorrectal. Resultados de analysis of the optimum number of stool speci-
104110. un programa prospectivo y regional de cribado mens collected for immunochemical occult blood
13. Houlston S. Murday V, Harocopos C, et al. mediante colonoscopia en poblaci on de riesgo. screening for colorectal cancer. Eur J Cancer
Screening and genetic counselling for relatives of Gastroenterol Hepatol 2008;31:1889. 2000;36:64750.
patients with colorectal cancer in a family cancer 25. Trajman A, Luiz RR. McNemar chi2 test revis- 36. Levi Z, Rozen P, Hazazi R, et al. Can quantica-
clinic. Br Med J 1990;301:3668. ited: comparing sensitivity and specicity of diag- tion of faecal occult blood predetermine the need
14. Stephenson BM, Murday VA, Quirke P, et al. nostic examinations. Scand J Clin Lab Invest for colonoscopy in patients at risk for non-
Feasibility of family based screening for colorectal 2008;68:7780. syndromic familial colorectal cancer? Aliment
neoplasia: experience in one general surgical 26. Ladabaum U, Ferrandez A, Lanas A. Cost-effec- Pharmacol Ther 2006; 24:147581.
practice. Gut 1991;34:96100. tiveness of colorectal cancer screening in high- 37. Terhaar Sive Droste JS, Turenhout ST van, et al.
15. Rozen P, Levi Z, Hazazi R, et al. Quantitative risk Spanish patients: use of a validated model to Faecal immunochemical test accuracy in patients
colonoscopic evaluation of relative efciencies of inform public policy. Cancer Epidemiol Bio- referred for surveillance colonoscopy: a multi-
an immunochemical faecal occult blood test and markers Prev 2010;19:276576. centre cohort study. BMC Gastroenterol 2012;12:
a sensitive guaiac test for detecting signicant 27. Whiting P, Rutjes A, Dinnes J, et al. Methods for 94.
colorectal neoplasms. Aliment Pharmacol Ther assessing the quality of diagnostic accuracy stud- 38. Grupo de trabajo de la gua de practica clnica de
2009;29:4507. ies. Health Technol Assess 2004;8:1234. prevenci on del cancer colorrectal. Actualizacion
16. Greenberg PD, Bertario L, Gnauck R, et al. A 28. Elm E von, Altman DG, Egger M, et al. The 2009. Prevenci on del Cancer Colorrectal. Gua de
prospective multicenter evaluation of new fecal Strengthening the Reporting of Observational practica clnica. Barcelona: Asociaci
on Espa~nola

Early Detection and Diagnosis


occult blood tests in patients undergoing colono- Studies in Epidemiology (STROBE) statement: de Gastroenterologa, Sociedad Espa~ nola de
scopy. Am J Gastroenterol 2000; 95:13318. guidelines for reporting observational studies. J Medicina de Familia y Comunitaria, y Centro
17. Park DI, Ryu S, Kim Y-H, et al. Comparison Clin Epidemiol 2008;61:3449. Cochrane Iberoamericano; 2009. Programa de
of Guaiac-Based and Quantitative Immuno- 29. Bossuyt PM, Reitsma JB, Bruns DE, et al. Stand- Elaboraci on de Guas de Practica.
chemical Fecal Occult Blood Testing in a Pop- ards for Reporting of Diagnostic Accuracy. 39. Ramsey SD, Wilschut J, Boer R, et al. A
ulation at Average Risk Undergoing Colorectal Toward complete and accurate reporting of stud- decision-analytic evaluation of the cost-
Cancer Screening. Am J Gastroenterol 2010;105: ies of diagnostic accuracy: the STARD initiative. effectiveness of family historybased colorectal
2017-25. Standards for Reporting of Diagnostic Accuracy. cancer screening programs. Am J Gastroenterol
18. Hundt S, Haug U, Brenner H. Comparative Eval- Ann Intern Med 2003;138:40-4. 2010;105:18619.
uation of Immunochemical Fecal Occult Blood 30. Rozen P, Comaneshter D, Levi Z, et al. Cumula- 40. Kaminski MF, Regula J, Kraszewska E, et al.
Tests. Ann Intern Med 2007; 150:1629. tive evaluation of a quantitative immunochemical Quality indicators for colonoscopy and the risk of
19. Nakama H, Zhang B, Zhang X. Evaluation of the fecal occult blood test to determine its optimal interval cancer. N Engl J Med 2010;362:
optimum cut-off point in immunochemical occult clinical use. Cancer 2010;116:211525. 1795803.

Int. J. Cancer: 134, 367375 (2014) V


C 2013 UICC
Copyright of International Journal of Cancer is the property of John Wiley & Sons, Inc. and
its content may not be copied or emailed to multiple sites or posted to a listserv without the
copyright holder's express written permission. However, users may print, download, or email
articles for individual use.

You might also like