Comparison of the performance of guaiac-based and two immunochemical fecal occult blood tests for identifying advanced colorectal neoplasia in Taiwan

Original article

Comparison of the performance of guaiac-based and twoimmunochemical fecal occult blood tests for identifying

advanced colorectal neoplasia in Taiwan

Chi-Hsing OU,* Fu-Chen KUO,†,‡ Wen-Hung HSU,* ,§ Chien-Yu LU,* Fang-Jung YU,* Chao-Hung KUO,*,¶

Jaw-Yuan WANG,¶,** Ming-Tsang WU,† Jentaie SHIEA,†† Deng-Chyang WU§,¶,‡‡ & Huang-Ming HU*

*Division of Gastroenterology, Department of Internal Medicine, ¶Cancer Center, and **Division ofGastrointestinal and General Surgery, Department of Surgery, Kaohsiung Medical University Hospital,

†Department of Public Health, College of Health Sciences, and §Division of Internal Medicine, KaohsiungMunicipal Hsiao-Kang Hospital, Kaohsiung Medical University, ‡Department of Health Management, I-Shou

University, ††Department of Chemistry, and ‡‡Institute of Biomedical Sciences, National Sun Yat-SenUniversity, Kaohsiung, Taiwan, China

OBJECTIVES: We aimed to evaluate the perfor-mance of fecal occult blood tests (FOBT) for thescreening of colorectal cancer (CRC).

METHODS: We selected participants from a medicalcenter in Taiwan, China from November 2009 to June2011. All participants underwent screening colonos-copy and were asked to provide their stool samples forqualitative immunochemical FOBT (qlFIT), quantita-tive immunochemical FOBT (qnFIT) and the guaiac-based FOBT(GT). The receiver operating characteristiccurve was utilized to determine the optimal cut-offvalue of qnFIT. We measured the detection ability of theFOBT for colonic polypoid lesions.

RESULTS: In all, 699 participants were enrolled inthis study. For qnFIT, we found increased levels of fecalhemoglobin in participants with polyps ≥10 mm

(251.0 ng/mL), villous-containing adenomas (98.7ng/mL) and advanced adenomas (187.9 ng/mL).The optimal cut-off value for qnFIT was 25 ng/mLfor detecting advanced colorectal neoplasms (ACRN).All three FOBT had a similar but low estimate indetecting small and tubular adenomatous polyps.The qnFIT had a better detection ability for largeadenomas (positive likelihood ratio [PLR], 5.6 vs 3.1 vs0.3) and adenomas with villous-components (PLR3.7 vs 3.3 vs 0.3) than qlFIT and GT. For the ACRNgroup, qnFIT also showed the best screeningability with a sensitivity of 56.8%, accuracy of 86.7%and PLR 5.0.

CONCLUSIONS: Both qnFIT and qlFIT performbetter than GT in detecting advanced adenomas andCRC in the Taiwanese population.

KEY WORDS: colorectal neoplasms, early detection of cancer, fecal occult blood test, qualitative research,quantitative.

Correspondence to: Huang-Ming HU, Division of Gastroenterology, Department of Internal Medicine, Kaohsiung Medical University Hospital, 100Tz-You 1st Road, Kaohsiung City, 807 Taiwan, China. Email: [email protected]

Conflict of interest: None.

© 2013 Wiley Publishing Asia Pty Ltd and Chinese Medical Association Shanghai Branch, Chinese Society of Gastroenterology, RenjiHospital Affiliated to Shanghai Jiaotong University School of Medicine

bs_bs_banner

Journal of Digestive Diseases 2013; 14; 474–483 doi: 10.1111/1751-2980.12077

474

INTRODUCTION

Colorectal cancer (CRC) is one of the leading causes ofcancer-related deaths worldwide.1 In Taiwan, China,CRC is the third leading cause of cancer-related mor-tality and the most common among newly diagnosedcancers, according to the cancer incidence reportreleased by the Department of Health of Taiwan. CRChas been shown to arise from precancerous lesions(adenomas) and process through the “adenoma–carcinoma” sequence,2 which can be interrupted bythe early detection and removal of adenomatouspolyps.3 This makes population-based screening anattractive strategy for the prevention of CRC. Colono-scopy is the most accurate screening test and precan-cerous lesions can be removed at the same time.However, the invasive nature of the procedure and itshigh cost limit its acceptability.

The fecal occult blood test (FOBT) is an alternative testfor the population-based screening for CRC. It is non-invasive, easy to be performed and inexpensive, all ofwhich make it more acceptable. Traditional annual orbiennial guaiac-based FOBT (GT) screening has beenshown to reduce the incidence of CRC and CRCmortality.4–7 However, its low sensitivity and accuracylimit its use as a screening test for the detection ofearly CRC and advanced adenomas (AA).8–10 Anotherdisadvantage of GT is that it does not specifically iden-tify the human heme but detects heme and peroxidaseactivity presented in food, including red meat, fruitand vegetables.11 Thus, strict dietary restrictions is nec-essary before collecting stool samples in order toprevent false-positive or false-negative results.

Another stool test, the fecal immunochemical test(FIT), offers a promising alternative for the detectionof CRC. As it detects human hemoglobin (Hb) withspecific antibodies using ELISA, there is no need fordietary restrictions. Another advantage of FIT is thatit is more specific in the detection of blood from thelower gastrointestinal (GI) tract than GT.12,13 Cur-rently, two types of FIT are available. One is the com-mercial test that offers a qualitative interpretation ofstool Hb which can be utilized as an office-basedanalysis. However, its pre-setting cut-off value influ-ences the sensitivity of the test.14 The other is a testthat quantitatively detects the Hb in stool. It hasbeen used as a population-based screening testin several countries and regions including Taiwan(China). An important characteristic of this test is itsability to optimize a cut-off value according to thescreening strategy. In this prospective study, we

aimed to evaluate the performance of the quantita-tive FIT to choose the appropriate cut-off value fordetecting the colonic pre-cancerous lesion, and tocompare the performance characteristics of GT andtwo types of FIT for the detection of colorectal malig-nancy and its precursors in a population undergoingscreening colonoscopy.

MATERIAL AND METHODS

Study design and participants

Participants who were scheduled for colonoscopyfrom November 2009 to June 2011 were enrolled inthis study. The indication for colonoscopy included aprevious history of colorectal malignancy or polyp, afamily history of CRC or symptoms related to thelower GI tract, such as changes in bowel habits.Asymptomatic participants who asked for a screeningcolonoscopy were also enrolled in this study. Partici-pants who had overt GI bleeding symptoms, a historyof inflammatory bowel disease (IBD), repeated rectalbleeding or other causes of GI bleeding such asangiodysplasia, were excluded from the study. After weobtained their written informed consent, the partici-pants received a study package containing the instru-ments for collecting the stool samples and samplecollection instructions, including how to prevent thesample against the contamination of urine and waterfrom the toilet bowl. No dietary or medication restric-tions were required before the stool sample collection.The participants who provided stool samples andreceived colonoscopic examination were included inthe statistical analysis.

Stool samples and FOBT

The study package contained instruments for threeFOBT: a traditional GT (Hemoccult II test; BeckmanCoulter, Fullerton, CA) and two FIT. The latterincluded a qualitative test (qlFIT) (OC-Light; EikenChemicals Co., Ltd., Tokyo, Japan) and a quantitativetest (qnFIT) (OC-Sensor; Eiken Chemicals Co., Ltd.,Tokyo, Japan). The stool samples for both FIT werecollected by sampling probes that were inserted into atleast six different areas of the stool until the groove ofthe probe was completely covered by the stool accor-ding to the manufacturer’s instructions. The probeswere then immediately inserted into the samplingtubes. For qlFIT, the sampling tube contained thesodium azide solution as a preservative. The stoolsamples were stored in double Ziploc bags (ThaiGriptech, Bankok, Thailand) at 4°C until the

Journal of Digestive Diseases 2013; 14; 474–483 Performance of FOBT for colon neoplasia 475

© 2013 Wiley Publishing Asia Pty Ltd and Chinese Medical Association Shanghai Branch, Chinese Society of Gastroenterology, Renji Hospital Affiliated toShanghai Jiaotong University School of Medicine

participants brought them back when they came toundergo the colonoscopy. In the laboratory thesamples were stored at 4°C and were analyzed withinone week following the manufacturer’s instruc-tions.15,16 For the GT, the stool samples were spreaddirectly onto the filter paper containing guaiac gumthrough the oval spaces on the test kit card. Each cardcontains two guaiac-impregnated windows and thefecal material was applied to two sites on the card, fora total of six windows per participant. A positive resultwas defined as at least one of the six windows turnedblue. For qlFIT, we inserted a strip into the samplingtube from the dip side and read the result at 5 min.The presence of a blue line in the lower and uppercenter of the strip was considered positive. The settingcut-off value of qlFIT was 50 ng/mL. The qnFITsamples were processed by an OC-Micro instrument(Eiken Chemicals Co., Ltd., Tokyo, Japan). ThreeFOBT were analyzed by different technicians whowere blinded to the final results.

Colonoscopy and histological examination

Colonoscopy performed by experienced gastroenter-ologists. Participants who had poor bowel preparationor incomplete examination were excluded, except if itwas caused by the tumor-related obstruction orstenosis. Patients who were diagnosed with IBD werealso excluded from the study. A diagnosis of diverticu-losis or non-bleeding hemorrhoids was categorized asa negative result. If possible, all the neoplasms wereremoved or biopsied for histological examination. Thesize of the lesions (<10 mm or ≥10 mm) and theirlocation (proximal, from the cecum to the splenicflexure; distal, the rest of the colon) were recorded. Allthe biopsied tissues were evaluated by experiencedpathologists who were blinded to the results of bothcolonoscopy and FOBT. The neoplasms were classifiedas hyperplastic polyp, adenomatous polyp oradenocarcinoma. Adenomatous polyps were graded astubular, tubulovillous, villous or serrated polyps. If theparticipants had more than one adenomatous polyp,they were grouped based on the most histologicallyadvanced lesion or the largest polyp. AA were definedas tubulovillous or villous adenomas with a size of≥10 mm or adenomas with high-grade dysplasia. Theadvanced colorectal neoplasm (ACRN) groupincluded participants with AA or CRC.

Statistical analysis

Statistical analyses were performed using SPSS 14.0(SPSS Inc., Chicago, IL, USA). For qnFIT, as the cut-off

value influences the efficacy, we first evaluated thesensitivity, specificity, positive predictive value (PPV),negative predictive value (NPV), accuracy, positivelikelihood ratio (PLR) and 95% confidence interval(CI) according to the different cut-off values (25, 50,75, 100, 125 and 150 ng/mL). The receiver operatingcharacteristic (ROC) curve for determining optimalqnFIT cut-off values was derived by plotting the sensi-tivity versus 1 minus the specificity. The area under theROC curve (AUROC) and Youden’s index (sensitivityplus specificity minus 1) were calculated. The optimalcut-off value of qnFIT was determined based on theone that maximized theYouden’s index. Next, we com-pared the efficacy of the three FOBT in variable catego-ries (size, histology, location and high risk of cancer).Furthermore, we measured interobserver reliability.We compared the mean values by the Mann–Whitneytest for two independent groups and the mean valuesin independent categories by the Kruskal–Wallis testfor three or more independent groups. A two tailed Pvalue of <0.05 was considered statistically significant.

RESULTS

Of the 784 participants scheduled for colonoscopy, 85were subsequently excluded because of a missing colo-noscopy (n = 38), incomplete study or poor bowelpreparation (n = 40) or a diagnosis of IBD (n = 7)(Fig. 1). Finally, 699 participants with a median age of59.5 years (550 patients [78.7%] were over 50 years)were enrolled in this study, in which 308 (44.1%)were males (mean age 59.8 years; median age 60years) and 391 (55.9%) were women (mean age 58.8years; median age 59 years). In all, 697 (99%) under-went a complete examination of the colon. We foundcolorectal polyps in 373 (53.4%) participants, inwhich 201 were found to have hyperplastic polyps,133 had adenomatous polyps and 39 were categorizedas AA. We diagnosed three patients with colon cancer,of whom two had an incomplete colon evaluation dueto tumor-related lumen obstruction or stenosis. InFOBT we received 694 samples for qnFIT, 685 samplesfor qlFIT and 609 samples for GT, respectively. A totalof 593 participants were available for all three testresults.

The characteristics of the colonoscopic findings of the694 participants and their results of the qnFIT analysisare shown in Table 1. We found that the mean fecal Hblevel of qnFIT was similar between the participantswith or without colorectal polyp. However, anincreased fecal Hb levels were found in participantswith polyps ≥10 mm in diameter (P < 0.001), villous-

Journal of Digestive Diseases 2013; 14; 474–483476 C-H Ou et al.


containing adenomas (P = 0.061) or AA (P < 0.001).Participants who had hyperplastic polyps had asimilar stool Hb level to those without colon polyp.No participant had an adenomatous polyp with high-grade dysplasia. The location of the adenomatouspolyps did not influence the Hb levels in the feces(P = 0.475). In three patients diagnosed with CRC thefecal Hb levels were significantly elevated, with a rangefrom 374 to 2714 ng/mL.

In order to optimize the performance of qnFIT, weselected different cut-off values (25, 50, 75, 100, 125and 150 ng/mL) to compare their sensitivity, speci-ficity, PPV, NPV, accuracy and PLR (Table 2). For theadenomatous polyps, these cut-off values did not offergood prediction, with an accuracy of around 74%. TheAUROC for detecting adenomatous polyps by qnFITwas 0.617 (data not shown). When focusing on AA,

the accuracy improved with a range from 87%(25 ng/mL) to 92% (125 ng/mL). The AUROC for thedetection of AA was 0.745 (Fig. 2). According to theYouden’s index, the optimal cut-off value for detectingAA was 25 ng/mL.

We also compared the performances of three differentFOBT (Table 3). We chose 25 ng/mL as the optimalcut-off value of qnFIT. All three FOBT had a relativelylow accuracy and PLR in detecting adenomatouspolyps. This was also applied to the detection of smalladenomatous polyps (<10 mm) and tubular adeno-mas. On the other hand, the detection ability of thethree FOBT differed when they were used to screenpolyps with a malignant potential. The qnFIT gave abetter detection ability for large adenomas (PLR: 5.6for qnFIT vs 3.1 for qlFIT vs 0.3 for GT) and adenomaswith a villous component (PLR: 3.7 for qnFIT vs 3.3for qlFIT vs 0.3 for GT) than qlFIT and GT. For theACRN group, qnFIT gave the best screening ability,with a sensitivity of 56.8% (95% CI 39.5–72.9), anaccuracy of 86.7% (95% CI 83.7–89.3) and a PLR of5.0 (95% CI 3.5–7.2); while the traditional GT had aPPV of 3.8%, an accuracy of 81.3% and a PLR of 0.6only. The performance of qlFIT was better than that ofGT, but its accuracy and PLR were not as good asqnFIT. For six participants who had colorectal serratedadenomas, three had a positive FOBT by qnFIT.

We also assessed the influence of the location of ACRNon the performance of the three FOBT (Table 4). TheGT detected only one case with an ACRN in the proxi-mal colon that was malignant with a sensitivity andPLR of 5.6% and 0.4, respectively. Two cases withACRN in the distal colon were detected by GT (with asensitivity of 10.5% and a PLR of 0.8). The detectionability of the two FIT on ACRN was clearly greater. TheqlFIT achieved a sensitivity of 27.8% and 36.8% andPLR of 2.9 and 4.0, respectively for the prediction ofproximal and distal ACRN. The qnFIT had a sensitivityof 44.4% and a PLR of 3.4 for the proximal ACRN. TheqnFIT offered the best detection ability of these threeFOBT for the ACRN in the distal colon, with a sensi-tivity of 57.9% and a PLR of 4.6.

We also compared the performance of the two FIT inthe detection of ACRN using the cut-off value of 50ng/mL for qnFIT (Table 5). The sensitivity, PPV, accu-racy and PLR were 40.5%, 25.9%, 89.0% and 5.2 inthe qnFIT and 32.4%, 20.0%, 87.7% and 3.8 in theqlFIT, respectively, showing that qnFIT had a betterdetection ability than qlFIT.

Patients scheduled for colonoscopy (N = 784)

Missing colonoscopy (n = 38)

Colonoscopy performed (n = 746)

Excluded (n = 47)

Incomplete colonoscopy or poor bowel

Inflammatory bowel disease (n = 7)

preparation (n = 40)

Patients included in the analysis (n = 699)

OC-sensor (n = 694)

OC-light (n = 685)

Underwent all three FOBT (n = 593)

FOBT

GT (n = 609)

Excluded

Figure 1. Flowchart of the study. FOBT, fecal occult bloodtests; GT, guaiac-based FOBT.



DISCUSSION

Following the discovery of the “adenoma–carcinoma”sequence in the development of CRC2,17 the strategyfor decreasing the incidence of CRC focused on theearly detection of its precursor.18,19 Several techniqueshave been considered as screening tests, includingcolonoscopy, FOBT and a stool DNA test. FOBT, whichutilized the characteristics of the increasing incidenceof blood extravasation from CRC or adenomatouspolyps, has been recommended as the tool forpopulation-based screening.20 Traditional GT has beenshown to be effective in reducing mortality from coloncancer in large randomized controlled trials.4–7

However, because it is not specific to human bloodand has low sensitivity for colorectal neoplasms andAA,8–10 GT is not an ideal test for population-basedscreening, as was shown in this study as it detectedonly three out of 37 cases of ACRN.

The FIT is an alternative test that detects the existenceand the amount of human Hb in stool. Rozen et al.21

reported that Hb loss from adenomas is significantlyassociated with the size and advanced histological fea-tures of the adenoma. In our study, the fecal Hb levelsignificantly increased if the polyp contained a villouscomponent in histology and had a size ≥10 mm in

diameter. Several studies have reported that FIT ismore sensitive than GT.12,22–26 However, its perfor-mance depends on the cut-off values chosen. Cur-rently, the optimal cut-off point of fecal Hb for thedetection of CRC is controversial.23,27–33 Some studieshave concluded that the optimal cut-off point of fecalHb for the detection of CRC is 100 ng/mL.31–33 This isnot necessarily suitable for screening precursor lesionsof CRC. Other studies imply a lower cut-off value leadsto better performance. In this study, we utilized theROC curve and Youden’s index and concluded that25 ng/mL might be a better cut-off value for evaluatingthe existence of ACRN, similar to a study reportedpreviously.34 Moreover, we also found that variablecut-off values, ranging from 25 to 150 ng/mL, offereda similar but low detection rate for tubular adenomas,indicating that FIT is not an acceptable noninvasivetool for screening non-advanced colon polyps.

The qnFIT has several advantages. First, the resultdepends on the cut-off value and it can be adjustedaccording to the variable strategy chosen, such as theintended detection rate and strategy in the screenedpopulation. Second, it can prevent interobservervariations in interpretation.14 Its disadvantage is thatqnFIT needs an instrument that might not be availableeverywhere. Unlike qnFIT, qlFIT is easily performed

Table 1. The fecal hemoglobin (Hb) level by quantitative fecal immunochemical test (qnFIT) according the variablecolonoscopic findings

n (%) Hb level, ng/mL (mean [range]) P value

Participants 694 (100) 41.7 (0–2714)Colonoscopic finding

No polyp 321 (46.3) 28.7 (0–2287)Polyp(s) 370 (53.3) 42.3 (0–2314)Cancer 3 (0.4) 1357.3 (374–2714)

Classification of polyp <0.001Hyperplastic polyp 199 (28.7) 26.9 (0–2158)Non-advanced adenoma 132 (19.0) 22.4 (0–690)Advanced adenoma 39 (5.6) 187.9 (0–2314)ACRN 42 (6.1) 271.5 (0–2714)

Location of adenoma 0.475Proximal 57 (8.2) 45.4 (0–984)Distal 96 (13.8) 49.4 (0–2714)Both 18 (2.6) 164.3 (0–2314)

Size of adenoma <0.001≥10 mm 27 (3.9) 251.0 (0–2714)<10 mm 144 (20.7) 24.4 (0–2158)

Histology of adenoma 0.061Tubular adenoma 139 (20.0) 54.0 (0–2314)Villous or tubulovillous adenoma 26 (3.7) 98.7 (0–790)Serrated adenoma 6 (0.9) 36.7 (0–123)

ACRN, advanced colorectal neoplasm.



and can be utilized in the clinic without the need of aspecial instrument. However, we found that with anadequate cut-off value of 50 ng/mL, qnFIT had ahigher sensitivity and accuracy than qlFIT, indicatingthat qnFIT is a better screening tool for ACRN.

Several studies have noted that the location of thelesion influences the performance of FIT.35,36 Becausethe FIT relies on the antibodies that bind to the globin,the decomposition of the Hb influences the tool’sdetection ability, and it is possible that bleeding fromproximal colon may lead to an underestimation of theHb level.37 Morikawa et al. reported that the sensitivityof qnFIT in the detection of adenomas ≥10 mm in sizein the proximal colon was lower than that of the distallesions (11% vs 25%).36 This is similar to our finding.However, two other studies reached different conclu-sions and found that qnFIT had a similar sensitivityfor both proximal and distal advanced neoplasia.22,38

Further studies are needed to clarify this issue.

In this study we collected six serrated adenoma cases.A serrated colon lesion is considered to be a precursorof CRC and may account for one-third of colon malig-nancies.39 Interestingly, we found that the mean fecalHb level in these cases was lower than that of tubularadenomas. Only qnFIT detected two of these cases.Ta

ble

2.P

erfo

rman

ceo

fq

uan

tita

tive

imm

un

och

emic

alfe

cal

occ

ult

blo

od

test

for

dif

fere

nt

cut-

off

valu

es(N

=69

4)

Sen

siti

vity

(%)

(95%

CI)

Spec

ifici

ty(%

)(9

5%C

I)P

PV

(%)

(95%

CI)

NP

V(%

)(9

5%C

I)A

ccu

racy

(%)

(95%

CI)

PLR

(95%

CI)

Yo

ud

en’s

ind

ex

Ad

eno

ma

≥25

ng/

mL

26.3

(19.

9–33

.6)

89.7

(86.

7–92

.2)

45.5

(35.

4–55

.8)

78.8

(75.

3–82

.0)

74.1

(70.

6–77

.3)

2.5

(1.8

–3.6

)0.

16≥5

0n

g/m

L17

.5(1

2.2–

24.1

)92

.9(9

0.4–

95.0

)44

.8(3

2.6–

57.4

)77

.5(7

4.0–

80.7

)74

.4(7

0.9–

77.6

)2.

5(1

.6–3

.9)

0.10

≥75

ng/

mL

12.9

(8.2

–18.

8)93

.7(9

1.3–

95.6

)40

.0(2

7.0–

54.1

)76

.7(7

3.2–

79.9

)73

.8(7

0.3–

77.0

)2.

0(1

.2–3

.4)

0.07

≥100

ng/

mL

9.4

(5.4

–14.

8)95

.0(9

2.8–

96.7

)38

.1(2

3.6–

54.4

)76

.2(7

2.8–

79.5

)73

.9(7

0.5–

77.2

)1.

9(1

.0–3

.4)

0.04

≥125

ng/

mL

8.2

(4.5

–13.

4)95

.4(9

3.3–

97.0

)36

.8(2

1.8–

54.0

)76

.1(7

2.6–

79.3

)73

.9(7

0.5–

77.2

)1.

8(0

.9–3

.4)

0.04

≥150

ng/

mL

7.0

(3.7

–11.

9)96

.9(9

3.9–

97.5

)36

.4(2

0.4–

54.9

)76

.0(7

2.5–

79.2

)74

.1(7

0.6–

77.3

)1.

8(0

.9–3

.5)

0.04

Ad

van

ced

Ad

eno

ma

≥25

ng/

mL

56.4

(39.

6–72

.2)

88.2

(85.

5–90

.6)

22.2

(14.

5–31

.7)

97.1

(95.

5–98

.3)

86.5

(83.

7–88

.9)

4.8

(3.4

–6.8

)0.

45≥5

0n

g/m

L38

.5(2

3.4–

55.4

)92

.1(8

9.7–

94.0

)22

.4(1

3.1–

34.2

)96

.2(9

4.4–

97.5

)89

.1(8

6.5–

91.3

)4.

8(3

.0–7

.8)

0.31

≥75

ng/

mL

35.9

(21.

2–52

.8)

93.7

(91.

6–95

.5)

25.5

(14.

7–39

.0)

96.1

(94.

3–97

.5)

90.5

(88.

1–92

.6)

5.7

(3.4

–9.6

)0.

30≥1

00n

g/m

L30

.8(1

7.0–

47.6

)95

.4(9

3.5–

96.9

)28

.6(1

5.7–

44.6

)95

.9(9

4.0–

97.3

)91

.8(8

9.5–

93.7

)6.

7(3

.7–1

2.1)

0.26

≥125

ng/

mL

25.6

(13.

0–42

.1)

95.7

(93.

9–97

.1)

26.3

(13.

4–43

.1)

95.6

(93.

7–97

.0)

92.1

(89.

8–94

.0)

6.0

(3.1

–11.

4)0.

21≥1

50n

g/m

L20

.5(9

.3–3

6.5)

96.2

(94.

4–97

.5)

24.2

(11.

1–42

.3)

95.3

(93.

4–96

.8)

91.9

(89.

7–93

.9)

5.4

(2.6

–11.

1)0.

17

CI,

con

fid

ence

inte

rval

;NP

V,n

egat

ive

pre

dic

tive

valu

e;P

LR,p

osi

tive

like

lih

oo

dra

tio

;PP

V,p

osi

tive

pre

dic

tive

valu

e.Yo

ud

en’s

ind

ex=

sen

siti

vity

+sp

ecifi

city

-1.

1.0

1.0

0.8

0.8

0.6

0.6

0.4

0.4

1-Specificity

0.2

0.2

0.0

0.0

Sensitiv

ity

Figure 2. The receiver operating characteristic curve fordetermining the optimal cut-off value of quantitativeimmunochemical fecal occult blood test in detection ofadvanced adenoma.



Tab

le3.

Co

mp

aris

on

of

the

per

form

ance

of

the

thre

efe

cal

occ

ult

blo

od

test

sin

vari

ou

sco

lon

osc

op

icfi

nd

ings

(N=

593)

Sen

siti

vity

(%)

(95%

CI)

Spec

ifici

ty(%

)(9

5%C

I)P

PV

(%)

(95%

CI)

NP

V(%

)(9

5%C

I)A

ccu

racy

(%)

(95%

CI)

PLR

(95%

CI)

Ad

eno

ma

qn

FIT

29.3

(212

.2–3

7.3)

91.0

(87.

9–93

.5)

52.4

(41.

2–63

.4)

79.2

(75.

4–82

.6)

75.4

(71.

7–78

.8)

3.2

(2.2

–4.8

)q

lFIT

16.7

(11.

1–23

.6)

92.1

(89.

2–94

.4)

41.7

(29.

1–55

.1)

76.6

(72.

7–80

.1)

73.0

(69.

3–76

.6)

2.1

(1.3

–3.4

)G

T10

.0(5

.7–1

6.0)

85.3

(81.

7–88

.5)

18.8

(10.

9–29

.0)

73.7

(69.

7–77

.5)

66.3

(62.

3–70

.1)

0.7

(0.4

–1.2

)A

CR

Nq

nFI

T56

.8(3

9.5–

72.9

)88

.7(8

5.7–

91.2

)25

.0(1

6.2–

35.6

)96

.9(9

5.0–

98.2

)86

.7(8

3.7–

89.3

)5.

0(3

.5–7

.2)

qlF

IT32

.4(1

8.0–

49.8

)91

.4(8

8.7–

93.6

)20

.0(1

0.8–

32.3

)95

.3(9

3.2–

96.9

)87

.7(8

4.8–

90.2

)3.

8(2

.2–6

.4)

GT

8.1

(1.7

–21.

9)86

.2(8

3.0–

88.9

)3.

8(0

.8–1

0.6)

93.4

(90.

9–95

.4)

81.3

(77.

9–84

.3)

0.6

(0.2

–1.8

)Sm

all

aden

om

a(<

10m

m)

qn

FIT

22.2

(15.

3–30

.5)

88.0

(84.

7–90

.8)

33.3

(23.

4–44

.5)

80.8

(77.

1–84

.1)

74.0

(70.

3–77

.5)

1.9

(1.2

–2.8

)q

lFIT

14.3

(8.7

–21.

6)91

.0(8

8.0–

93.4

)30

.0(1

8.9–

43.2

)79

.7(7

6.1–

83.1

)74

.7(7

1.0–

78.2

)1.

6(0

.9–2

.7)

GT

11.1

(6.2

–17.

9)85

.9(8

2.4–

88.9

)17

.5(9

.9–2

7.6)

78.2

(74.

3–81

.7)

70.0

(66.

1–73

.7)

0.8

(0.5

–1.4

)La

rge

aden

om

a(≥

10m

m)

qn

FIT

66.7

(44.

7–84

.4)

88.1

(85.

1–90

.6)

19.1

(11.

3–29

.1)

98.4

(96.

9–99

.3)

87.2

(84.

2–89

.8)

5.6

(3.9

–8.0

)q

lFIT

29.2

(12.

6–51

.1)

90.7

(88.

0–92

.9)

11.7

(4.8

–22.

6)96

.8(9

4.9–

98.1

)88

.2(8

5.3–

90.7

)3.

1(1

.6–6

.1)

GT

4.2

(0.1

–21.

1)86

.1(8

3.0–

88.9

)1.

3(0

–6.8

)95

.5(9

3.4–

97.1

)82

.8(7

9.5–

85.8

)0.

3(0

–2.1

)Tu

bu

lar

aden

om

aq

nFI

T24

.8(1

7.4–

33.5

)88

.6(8

5.3–

91.3

)35

.7(2

5.6–

46.9

)82

.1(7

8.5–

85.4

)75

.6(7

1.9–

79.0

)2.

2(1

.5–3

.2)

qlF

IT14

.9(9

.1–2

2.5)

91.1

(88.

2–93

.5)

30.0

(18.

9–43

.2)

80.7

(77.

1–83

.9)

75.6

(71.

9–79

.0)

1.7

(1.0

–2.8

)G

T11

.6(6

.5–1

8.7)

86.0

(82.

6–89

.0)

12.5

(9.9

–27.

6)79

.1(7

5.4–

82.6

)70

.8(6

7.0–

74.5

)0.

8(0

.5–1

.4)

Vil

lou

so

rtu

bu

lovi

llo

us

aden

om

aq

nFI

T47

.8(2

6.8–

69.4

)87

.2(8

4.2–

89.8

)13

.1(6

.7–2

2.2)

97.6

(95.

9–98

.8)

85.7

(82.

6–88

.4)

3.7

(2.3

–6.0

)q

lFIT

30.4

(13.

2–52

.9)

90.7

(88.

0–93

.0)

11.7

(4.8

–22.

6)97

.0(9

5.2–

98.3

)88

.4(8

5.5–

90.8

)3.

3(1

.7–6

.4)

GT

4.3

(0.1

–22.

0)86

.1(8

3.0–

88.9

)1.

3(0

–6.8

)95

.7(9

3.6–

97.3

)83

.0(7

9.7–

85.9

)0.

3(0

–2.2

)Se

rrat

edad

eno

ma

qn

FIT

50.0

(11.

8–88

.2)

86.2

(83.

1–88

.9)

3.6

(0.7

–10.

1)99

.4(9

8.3–

99.9

)85

.8(8

2.8–

88.5

)3.

6(1

.6–8

.3)

qlF

IT0

(0–4

5.9)

89.8

(87.

0–92

.1)

0(0

–6.0

)98

.9(9

7.6–

99.6

)88

.9(8

6.1–

91.3

)0

(NaN

)G

T0

(0–4

5.9)

86.4

(83.

3–89

.0)

0(0

–4.5

)98

.8(9

7.5–

99.6

)85

.5(8

2.4–

88.2

)0

(NaN

)

The

cut-

off

hem

ogl

ob

inva

lue

of

the

qu

anti

tati

veim

mu

no

chem

ical

feca

lo

ccu

ltb

loo

dte

st(q

nFI

T)is

25n

g/m

L.A

CR

N,a

dva

nce

dco

lore

ctal

neo

pla

sm;C

I,co

nfi

den

cein

terv

al;G

T,gu

aiac

-bas

edfe

cal

occ

ult

blo

od

test

;qlF

IT,q

ual

itat

ive

imm

un

och

emic

alfe

cal

occ

ult

blo

od

test

;NP

V,n

egat

ive

pre

-d

icti

veva

lue;

PLR

,po

siti

veli

keli

ho

od

rati

o;P

PV,

po

siti

vep

red

icti

veva

lue.



The detection ability of the FIT in this type of lesionneeds further research.

Our study had several limitations. First, the enrolledpopulation was not an average-risk population assome patients were symptomatic. Therefore, wecannot extrapolate our findings to the performance ofFIT for screening average-risk populations in Taiwan.Second, the PPV of GT was lower than those reportedin earlier reports in an average-risk population.14 Thelack of dietary restriction could contribute to the lowPPV for GT. Our study samples from medical centerscould be different from that found in the generalpopulation, suggesting that the positive rates may beoverestimated in our study. Third, although we usedcolonoscopy as the gold standard, the missing colo-noscopies should be considered. The colonoscopicmissing rate depends on the endoscopist’s experienceand adequate bowel preparation, and the influence ofthe latter led to the exclusion of some participantsfrom the study.

In conclusion, we found that both quantitative andqualitative FIT perform better than GT in detecting AAand CRC in the Taiwanese population. In addition,

the qnFIT permits the choice of an appropriate cut-offvalue, thus offering flexibility for variable screeningstrategies and the patients’ clinical characteristics.

ACKNOWLEDGMENT

This study was supported by part by a grant fromthe Excellence for Cancer Research Center Grant,DOH100-TD-C-111-002, Department of Health,Executive Yuan, Taiwan and China Kaohsiung MedicalUniversity Hospital (KMUH97-7G44, KMUH97-7G45, KMUH97-7G46, 10005).

REFERENCES

1 Parkin DM, Bray F, Ferlay J, Pisani P. Global cancerstatistics, 2002. CA Cancer J Clin 2005; 55: 74–108.

2 Fearon ER, Vogelstein B. A genetic model for colorectaltumorigenesis. Cell 1990; 61: 759–67.

3 Arnold CN, Goel A, Blum HE, Boland CR. Molecularpathogenesis of colorectal cancer: implications formolecular diagnosis. Cancer 2005; 104: 2035–47.

4 Mandel JS, Church TR, Bond JH et al. The effect of fecaloccult-blood screening on the incidence of colorectal cancer.N Engl J Med 2000; 343: 1603–7.

5 Hardcastle JD, Chamberlain JO, Robinson MH et al.Randomised controlled trial of faecal-occult-blood screeningfor colorectal cancer. Lancet 1996; 348: 1472–7.

Table 4. Comparison of the performance of the three fecal occult blood tests in the detection of advanced colorectalneoplasm in different locations

Sensitivity (%)(95% CI)

Specificity (%)(95% CI)

PPV (%)(95% CI)

NPV (%)(95% CI)

Accuracy (%)(95% CI) PLR (95% CI)

Proximal ACRN (n = 18)qnFIT 44.4 (21.5–69.2) 86.8 (83.7–89.4) 9.5 (4.2–17.9) 98.0 (96.4–99.1) 85.5 (82.4–88.2) 3.4 (1.9–5.9)qlFIT 27.8 (9.7–53.5) 90.4 (87.7–92.7) 8.3 (2.8–18.4) 97.6 (95.9–98.7) 88.5 (85.7–91.0) 2.9 (1.3–6.4)GT 5.6 (0.1–27.3) 86.3 (83.2–89.0) 1.3 (0–6.8) 96.7 (94.8–98.1) 83.6 (80.4–86.5) 0.4 (0.1–2.7)

Distal ACRN (n = 19)qnFIT 57.9 (33.5–79.8) 87.3 (84.3–89.9) 13.1 (6.7–22.2) 98.4 (96.9–99.3) 86.3 (83.3–89.0) 4.6 (2.9–7.1)qlFIT 36.8 (16.3–61.6) 90.8 (88.1–93.0) 11.7 (4.8–22.6) 97.8 (96.1–98.8) 89.0 (86.2–91.4) 4.0 (2.1–7.6)GT 10.5 (1.3–33.1) 86.4 (83.3–89.1) 2.5 (0.3–8.7) 96.7 (94.8–98.1) 84.0 (80.8–86.8) 0.8 (0.2–2.9)

The cut-off hemoglobin value of the qnFIT is 25 ng/mL.CI, confidence interval; GT, guaiac-based fecal occult blood test; qlFIT, qualitative immunochemical fecal occult blood test; qnFIT, quanti-tative immunochemical fecal occult blood test; NPV, negative predictive value; PLR, positive likelihood ratio; PPV, positive predictivevalue.

Table 5. The comparison of the two immunochemical fecal occult blood tests (N = 593) for advanced colorectal neoplasm(ACRN) using the cut-off value of 50 ng/mL for qnFIT

Sensitivity (%)(95% CI)

Specificity (%)(95% CI)

PPV (%)(95% CI)

NPV (%)(95% CI)

Accuracy (%)(95% CI) PLR (95% CI)

ACRNqnFIT 40.5 (24.8–57.9) 92.3 (89.7–94.4) 25.9 (15.3–39.0) 95.9 (93.8–97.4) 89.0 (86.2–91.4) 5.2 (3.2–8.5)qlFIT 32.4 (18.0–49.8) 91.4 (88.7–93.6) 20.0 (10.8–32.3) 95.3 (93.2–96.9) 87.7 (84.8–90.2) 3.8 (2.2–6.4)

qlFIT, qualitative immunochemical fecal occult blood test; qnFIT, quantitative immunochemical fecal occult blood test; NPV, negative pre-dictive value; PLR, positive likelihood ratio; PPV, positive predictive value.



6 Kronborg O, Fenger C, Olsen J, Jorgensen OD,Sondergaard O. Randomised study of screening forcolorectal cancer with faecal-occult-blood test. Lancet 1996;348: 1467–71.

7 Mandel JS, Bond JH, Church TR et al. Reducing mortalityfrom colorectal cancer by screening for fecal occult blood.Minnesota Colon Cancer Control Study. N Engl J Med 1993;328: 1365–71.

8 Imperiale TF, Ransohoff DF, Itzkowitz SH, Turnbull BA,Ross ME; Colorectal Cancer Study Group. Fecal DNA versusfecal occult blood for colorectal-cancer screening in anaverage-risk population. N Engl J Med 2004; 351: 2704–14.

9 Rozen P, Levi Z, Hazazi R et al. Quantitative colonoscopicevaluation of relative efficiencies of an immunochemicalfaecal occult blood test and a sensitive guaiac test fordetecting significant colorectal neoplasms. Aliment PharmacolTher 2009; 29: 450–7.

10 Nakama H, Yamamoto M, Kamijo N et al. Colonoscopicevaluation of immunochemical fecal occult blood test fordetection of colorectal neoplasia. Hepatogastroenterology1999; 46: 228–31.

11 Macrae FA, St John DJ, Caligiore P, Taylor LS, Legge JW.Optimal dietary conditions for hemoccult testing.Gastroenterology 1982; 82: 899–903.

12 Zhu MM, Xu XT, Nie F, Tong JL, Xiao SD, Ran ZH.Comparison of immunochemical and guaiac-based fecaloccult blood test in screening and surveillance for advancedcolorectal neoplasms: a meta-analysis. J Dig Dis 2010; 11:148–60.

13 Whitlock EP, Lin JS, Liles E, Beil TL, Fu R. Screening forcolorectal cancer: a targeted, updated systematic review forthe U.S. Preventive Services Task Force. Ann Intern Med2008; 149: 638–58.

14 van Dam L, Kuipers EJ, van Leerdam ME. Performanceimprovements of stool-based screening tests. Best Pract ResClin Gastroenterol 2010; 24: 479–92.

15 Vilkin A, Rozen P, Levi Z et al. Performance characteristicsand evaluation of an automated-developed and quantitative,immunochemical, fecal occult blood screening test. Am JGastroenterol 2005; 100: 2519–25.

16 Rozen P, Waked A, Vilkin A, Levi Z, Niv Y. Evaluation of adesk top instrument for the automated development andimmunochemical quantification of fecal occult blood. MedSci Monit 2006; 12: MT27–32.

17 Fearon ER. Molecular genetics of colorectal cancer. Annu RevPathol 2011; 6: 479–507.

18 Winawer SJ, Zauber AG, Ho MN et al.; The National PolypStudy Workgroup. Prevention of colorectal cancer bycolonoscopic polypectomy. N Engl J Med 1993; 329:1977–81.

19 Citarda F, Tomaselli G, Capocaccia R, Barcherini S,Crespi M; Italian Multicentre Study Group. Efficacy instandard clinical practice of colonoscopic polypectomy inreducing colorectal cancer incidence. Gut 2001; 48: 812–5.

20 Levin B, Lieberman DA, McFarland B et al; American CancerSociety Colorectal Cancer Advisory Group; US Multi-SocietyTask Force; American College of Radiology Colon CancerCommittee. Screening and surveillance for the earlydetection of colorectal cancer and adenomatous polyps,2008: a joint guideline from the American Cancer Society,the US Multi-Society Task Force on Colorectal Cancer, andthe American College of Radiology. Gastroenterology2008; 134: 1570–95.

21 Rozen P, Levi Z, Hazazi R et al. Identification of colorectaladenomas by a quantitative immunochemical faecal occultblood screening test depends on adenoma characteristics,

development threshold used and number of testsperformed. Aliment Pharmacol Ther 2009; 29: 906–17.

22 Guittet L, Bouvier V, Mariotte N et al. Comparison of aguaiac and an immunochemical faecal occult blood test forthe detection of colonic lesions according to lesion type andlocation. Br J Cancer 2009; 100: 1230–5.

23 Park DI, Ryu S, Kim YH et al. Comparison of guaiac-basedand quantitative immunochemical fecal occult blood testingin a population at average risk undergoing colorectal cancerscreening. Am J Gastroenterol 2010; 105: 2017–25.

24 Hundt S, Haug U, Brenner H. Comparative evaluation ofimmunochemical fecal occult blood tests for colorectaladenoma detection. Ann Intern Med 2009; 150:162–9.

25 Yoshinaga M, Motomura S, Takeda H, Yanagisawa Z,Ikeda K. Evaluation of the sensitivity of animmunochemical fecal occult blood test for colorectalneoplasia. Am J Gastroenterol 1995; 90: 1076–9.

26 Fattah AS, Nakama H, Kamijo N, Fujimori K, Zhang B.Colorectal adenomatous polyps detected by immuno-chemical occult blood screening. Hepatogastroenterology 1998;45: 712–6.

27 Nakama H, Zhang B, Zhang X. Evaluation of the optimumcut-off point in immunochemical occult blood testing inscreening for colorectal cancer. Eur J Cancer 2001; 37:398–401.

28 Terhaar sive Droste JS, Oort FA, van der Hulst RW et al.Higher fecal immunochemical test cutoff levels: lowerpositivity rates but still acceptable detection rates forearly-stage colorectal cancers. Cancer Epidemiol BiomarkersPrev 2011; 20: 272–80.

29 Hol L, Wilschut JA, van Ballegooijen M et al. Screening forcolorectal cancer: random comparison of guaiac andimmunochemical faecal occult blood testing at differentcut-off levels. Br J Cancer 2009; 100: 1103–10.

30 Wilschut JA, Hol L, Dekker E et al. Cost-effectivenessanalysis of a quantitative immunochemical test forcolorectal cancer screening. Gastroenterology 2011; 141:1648–55.e1.

31 Chen LS, Liao CS, Chang SH, Lai HC, Chen TH.Cost-effectiveness analysis for determining optimal cut-offof immunochemical faecal occult blood test for population-based colorectal cancer screening (KCIS 16). J Med Screen2007; 14: 191–9.

32 Guittet L, Bouvier V, Mariotte N et al. Comparison of aguaiac based and an immunochemical faecal occult bloodtest in screening for colorectal cancer in a general averagerisk population. Gut 2007; 56: 210–4.

33 van Rossum LG, van Rijn AF, Laheij RJ et al. Randomcomparison of guaiac and immunochemical fecal occultblood tests for colorectal cancer in a screening population.Gastroenterology 2008; 135: 82–90.

34 Omata F, Shintani A, Isozaki M, Masuda K, Fujita Y,Fukui T. Diagnostic performance of quantitative fecalimmunochemical test and multivariate prediction model forcolorectal neoplasms in asymptomatic individuals. Eur JGastroenterol Hepatol 2011; 23: 1036–41.

35 Haug U, Kuntz KM, Knudsen AB, Hundt S, Brenner H.Sensitivity of immunochemical faecal occult blood testingfor detecting left- vs right-sided colorectal neoplasia. Br JCancer 2011; 104: 1779–85.

36 Morikawa T, Kato J, Yamaji Y, Wada R, Mitsushima T,Shiratori Y. A comparison of the immunochemical fecaloccult blood test and total colonoscopy in theasymptomatic population. Gastroenterology 2005; 129:422–8.



37 Segnan N. European Guidelines for Quality Assurance inColorectal Cancer Screening and Diagnosis, 1st edn.Luxembourg: EU, 2010.

38 de Wijkerslooth TR, Stoop EM, Bossuyt PM et al.Immunochemical fecal occult blood testing is equallysensitive for proximal and distal advanced neoplasia. Am JGastroenterol 2012; 107: 1570–8.

39 Rex DK, Ahnen DJ, Baron JA et al. Serrated lesions of thecolorectum: review and recommendations from an expertpanel. Am J Gastroenterol 2012; 107: 1315–29; quiz 1314,1330.



Documents

Comparison of the performance of guaiac-based and two immunochemical fecal occult blood tests for identifying advanced colorectal neoplasia in Taiwan