Risk factors associated with false-positive fecal immunochemical test results in colorectal cancer screening.

Introduction
Colorectal cancer (CRC) is one of the most leading causes of tumor-associated mortality worldwide. It is the most commonly diagnosed cancer of the gastrointestinal tract1. However, survival rates of CRC differ significantly between early and advanced CRC patients2. And the development from advanced adenoma (AA) to advanced CRC takes a long time3, suggesting that CRC is a preventable cancer type. Therefore, early detection and removal of precursor lesions could effectively reduce CRC-associated mortality4,5.
Stool-based tests have been recommended as options for population CRC screening. In many regions worldwide, fecal immunochemical test (FIT)-based screening for CRC has been implemented and acquired good effect6–8. Some studies showed that FIT-based screening programs reduced cancer-associated mortality rates by 10%−22% after several rounds of application9–11. Moreover, a recent meta-analysis also showed that people with negative FIT results had lower interval CRC incidence12.
In contrast to the traditional guaiac fecal occult blood test (gFOBT), FIT has higher sensitivity for advanced neoplasia (AN) and does not need dietary restriction13. A recent meta-analysis showed that FIT had high sensitivity (93%) and specificity (91%) for CRC detection. Regarding the detection of AN, the sensitivity and specificity of FIT were 48% and 93%, respectively. The positive predictive value ranged from 28.6% to 62.9%, indicating that a high proportion of FIT results may be false-positive14. Some other studies also reported false-positive FIT rates of 5.8% to 15.7%15–17.
False-positive results of FIT could result in unnecessary colonoscopy examination, which results in patients’ discomfort, extra healthcare costs, and even the risk of complications such as colonic perforation. However, the data about factors associated with the false-positive FIT result is sometimes limited and most of them only considered basic characteristics of patients. Some studies reported that gastroscopy detected more important upper digestive tract findings like gastric cancer in patients with false-positive fecal occult blood test results18–20. Moreover, gastric cancer is another common cause of cancer-associated death and the fifth most frequently diagnosed cancer worldwide in 20181,21. Although the incidence of gastric cancer is decreasing these years, the five-year survival rate remains low, only about 10% in patients with advanced gastric cancer22,23. However, only a few studies focus on the association between false-positive FIT results and upper digestive diseases. And whether people with false-positive FIT results have a higher risk of upper digestive tract diseases or not remains controversial24,25.
Therefore, this study aims to identify the clinical and endoscopic factors associated with false-positive results of FIT for AN and to evaluate whether false-positive FIT results would be indicative of other digestive tract diseases.

Methods

Study population
Individuals scheduled to undergo both gastroscopy and colonoscopy from the outpatient clinics and wards in the Qilu Hospital of Shandong University were prospectively enrolled between April 2020 and April 2021. Only subjects aged from 50 to 75 years were included in this study cohort. The exclusion criteria included the history of CRC; history of surgery of colorectum; history of colorectal disease that may result in the bloody stool such as inflammatory bowel disease, ischemic bowel disease, and the diverticulum with the presence of active bleeding; the current presence of apparently bloody stool; severe diarrhea (more than three times a day) or watery stool and presence of contraindications to endoscopies such as severe cardiopulmonary insufficiency, gestation, and suspected colonic perforation. Patients without a good bowel preparation (Boston Bowel Preparation Scale [BBPS] score ≥ 6) or without complete colonoscopies (reach the cecum) were excluded.
The Research Ethics Committee of Qilu Hospital of Shandong University approved the study protocol [KYLL-2019(KS)−348]. Informed consent was obtained from all participants included in the study. The study protocol conforms to the ethical guidelines of the 1975 Declaration of Helsinki (as revised in Brazil 2013) as reflected in a priori approval by the institution’s human research committee.

Data collection
Participants were firstly invited to complete a questionnaire about their symptoms, life behavior, medical history, and family history of CRC. The items based on other medical literature included age, sex, height, weight, smoking and drinking history, amount of red meat consumption, history of hemorrhoid, the recent usage of antiplatelet drugs such as aspirin.

FIT
We used the two-sample FIT in this study. All participants were provided with two-sample FIT kits and verbal instruction about completing FIT at home by themselves. They were asked to collect two stool samples from two consecutive bowel movements according to the instruction procedure. Colonoscopy was performed within one week after samples collection and samples were stored in the 4 °C refrigerators until returned. Received fecal samples, investigator processed FIT samples within 24 h by using the AC-SCREEN® analysis system (FUNOTEC, Dalian, China). The analytical working of the instrument is 10–286 ng/ml, which is equivalent to 10–286 µg/g. In this study, the positive cut-off value was 10 µg/g. The FIT result was regarded as positive once one of the two stool samples was tested positive. Coloscopy results were used as the reference standard. The false-positive FIT result referred to a positive FIT result without AA or CRC detected in subsequent colonoscopy, relatively. Although the FITs might be considered false positive with respect to CRC or its precursors, they are still likely to correctly point to occult bleeding from other potentially clinically relevant lesions and as such should not be considered an entirely false-positive result, as demonstrated in previous studies. Staffs performing the FIT test were all blind to the questionnaire and endoscopy results of participants.

Gastrointestinal endoscopy
Experienced gastroenterologists who performed more than 2,000 colonoscopies at Qilu Hospital performed all endoscopies in this study. All gastroenterologists were blind to the FIT results of participants. Endoscopic lesions detected under endoscopy were biopsied and examined by experienced histopathologists.
All participants were asked to have a low fiber diet and undergo bowel preparation before endoscopy. For colonoscopy, reaching the cecum was considered as complete colonoscopy unless malignant lesions blocked the lumen. Polyps were classified according to the most advanced lesion. The number, location, size, endoscopic and histological characteristics of polyps were all recorded. Inflammatory polyps or hyperplastic polyps were considered normal findings. According to their most advanced histopathologic characteristics, the grade was classified as low and high for polyps with dysplasia. Advanced adenoma (AA) was defined as adenoma with at least one of the following characteristics: the villous component, size ≥ 1 cm, and high-grade dysplasia. Advanced neoplasia (AN) was defined as the presence of AA or CRC. Negative colonoscopy was defined as the absence of AN under coloscopy. Other findings, such as inflammatory bowel disease, isolated colonic ulcers, colorectal inflammation (hyperemia and edema of mucosa under the endoscopy), incidental hemorrhoids, were all recorded in the results.
The number, location, size, endoscopic and histological characteristics of upper gastrointestinal lesions were also recorded. Lesions that were considered to affect the FIT result included gastric cancer, reflux esophagitis, gastric or duodenal ulcers, gastric polyps larger than 10 mm, esophageal varices, and hiatal hernias. Gastritis and duodenitis were not considered as important findings.

Statistical analysis
Participants with missing data were excluded from the analysis. Continuous variables were shown as mean ± SD and compared by student’s test or nonparametric test. Categorical variables were shown as frequency and compared by Fisher’s exact test. The accuracy of FIT for AA and CRC was assessed by sensitivity, specificity, positive prediction value (PPV), negative prediction value (NPV), and corresponding 95% confidence interval (95% CI). Missing data were excluded from analyses (0–1.3% for each variable considered and 1.1% missing excluded from the final model) and complete-case analysis was performed. Logistic regression was used to investigate factors associated with false-positive FIT results. A bivariable logistic regression analysis was firstly used for each factor collected. Then for all factors with a P value smaller than 0.1 in bivariable logistic regression analysis, a multivariate logistic regression analysis was performed to determine final factors and endoscopic findings associated with false-positive FIT results. P value ≤ 0.05 was considered statistically significant. All statistical analysis was performed using R v.4.1.1 (http://www.r-project.org), the forest plot was plotted by forestplot R package.

Results

Characteristics and endoscopy results of participants
A total of 929 participants was prospectively enrolled (Fig. 1). Table 1 described the overall baseline characteristics of included participants according FIT results. The mean age of participants was 59 ± 6 years. And the number of male and female was comparable. Most participants did not have a history of smoking and drinking. Only a few participants (4.1%) had a family history of CRC. Table 2 showed the major colonoscopic and gastroscopic findings by FIT status among included individuals. The prevalence of AN was 7% (65 of 929 participants). Of these, 55 participants were diagnosed as AA and 10 as CRC. Common non-advanced neoplasia findings included the presence of ≥ 3 non-advanced adenomas (21.6%) and colorectal diverticulum (8.6%). The FIT positive group had a higher proportion of AN than the FIT negative group (P < 0.001). As to gastroscopy results, common findings included gastric polyps (33%) and atrophic gastritis (57.8%). The FIT positive group had a higher proportion of gastric cancer than the FIT negative group (P = 0.014). The detailed information of FIT values stratified by endoscopic outcomes were shown in Supplementary Table 1.

Diagnostic accuracy of FIT for the detection of AN
With a cut-off value of 10 µg/g, the positive rate of FIT was 16.03% and the false-positive rate was 11.63%. The overall sensitivity, specificity, PPV, NPV of FIT for detecting AN were 64.6% (95% CI, 51.7%−75.8%), 87.6% (95% CI, 85.2%−89.7%), 28.2% (95% CI, 21.3%−36.2%), and 97.1% (95% CI, 95.5%−98.1%), respectively. For AA, the corresponding values were 58.2% (95% CI, 44.1%−71.1%), 86.6% (95% CI, 84.1%−88.8%), 21.5% (95% CI, 15.4%−29.1%), and 97.1% (95% CI, 95.5%−98.1%). And for CRC, these values were 100.0% (95% CI, 65.6%−100.00%), 84.9% (95% CI, 82.4%−87.1%), 6.7% (95% CI, 3.4%−12.3%), and 100.00% (95% CI, 99.4%−100.00%). In men, the sensitivity, specificity, PPV, NPV of FIT for detecting AN were 67.3% (95% CI, 52.3%−79.6%), 84.6% (95% CI, 80.7%−87.8%), 34.0% (95% CI, 24.9%−44.4%), and 95.6% (95% CI, 92.9%−97.4%), respectively. In women, the sensitivity, specificity, PPV, NPV of FIT for detecting AN were 56.3% (95% CI, 30.6%−79.2%), 90.4% (95% CI, 87.2%−92.9%), 17.3% (95% CI, 8.7%−30.8%), and 98.3% (95% CI, 96.4%−99.3%), respectively. Supplementary Table 2 showed diagnostic accuracy of FIT for AN in different subgroups.

Risk factors associated with false-positive FIT results
Participants with a negative coloscopy (n = 864) were included in the logistic regression analysis to find out factors associated with the false-positive result of FIT. Supplementary Table 3 described the baseline characteristics of participants with a negative coloscopy according FIT results. The results of bivariable and multivariable logistic regression were shown in Table 3. In bivariable logistic regression analysis, clinical characteristics such as male (OR = 1.72; 95% CI, 1.14–2.61; P = 0.010), current drinker (OR = 1.88; 95% CI, 1.03–3.28; P = 0.031), current smoking (OR = 1.77; 95% CI, 1.01–2.99; P = 0.040) and more consumption of meat (OR = 2.20; 95% CI, 1.14–4.09; P = 0.015), endoscopic findings such as colorectal inflammation (OR = 2.29; 95% CI, 1.08–4.52; P = 0.022), presence of three or more non-advanced adenomas (OR = 1.76; 95% CI, 1.12–2.71; P = 0.012), hiatal hernia (OR = 3.64; 95% CI, 0.96–11.76; P = 0.038) and gastric cancer (OR = 9.76; 95% CI, 2.12–50.14; P = 0.003) were all associated with false-positive results. Several factors were independently associated with false-positive FIT results after adjusting confounding factors in multivariable logistic regression analysis. Among clinical characteristics, male (OR = 1.61; 95% CI, 1.05–2.48; P = 0.030) was independently associated with false-positive FIT results. However, the association between consumption amount of meat and false-positive results was attenuated to nonsignificant after adjustment. As to the colonoscopy findings, colorectal inflammation (OR = 2.99; 95% CI, 1.38–6.04; P = 0.003) and presence of three or more non-advanced adenomas (OR = 1.78; 95% CI, 1.11–2.82; P = 0.015) were found associated with false-positive FIT results for AN. Among upper digestive findings, gastric cancer (OR = 11.33; 95% CI, 2.40–59.52; P = 0.002) was independently associated with the false-positive FIT (Table 3 and Fig. 2). However, the power might be limited due to small case numbers and the wide confidence interval.

Discussion
A total of 929 participants was prospectively included in this study. With a cut-off value of 10 µg/g, the positive rate of FIT was 16.0%. For detecting AN, the sensitivity and specificity were 64.6% and 87.6%, respectively. Then 864 participants with a negative coloscopy were included in logistic regression analysis to determine factors associated with false-positive FIT results. Among them, 107 had false-positive FIT results. Male was found associated with false-positive FIT results. In addition, endoscopic findings including colorectal inflammation, presence of ≥ 3 non-advanced adenomas, and gastric cancer also contributed to false-positive FIT results.
A previous study showed that about a quarter of 687 patients with AN had discordant result in 2-sample FIT screening26. Therefore, 2-sample FIT was used in this study to yield more detection of AN. Several studies reported that FIT’s sensitivity and specificity for AN ranged from 38.0% to 68.8% and 85.8% to 93.6%, respectively27–31. Our data showed comparable diagnostic accuracy to these studies with a 10 µg/g positive cut-off value. However, the low PPV for AN in these studies suggested that a considerable proportion of positive FIT results were false-positive. Therefore, factors associated with false-positive FIT results should be further investigated.
Some studies had investigated the factors associated with false-positive FIT results16,17,32. A Germany colorectal cancer screening study and a Netherlands study reported a statistically significant association between male and false-positive FIT results17,32. In addition, the study from Germany reported a significant association between red meat consumption in bivariable logistic regression, while the association was then attenuated to be nonsignificant in multivariable logistic regression analysis32. This result was also shown in this study, in agreement with the detection specificity of FIT for human hemoglobin33. FIT represents an advancement over the guaiac method for detecting hemoglobin in the stool, utilizing antibodies specific for human hemoglobin instead of the more non-specific peroxidase reaction, characterized by a minimal likelihood of being affected by diet34,35. Some studies reported the association between aspirin use and false-positive FIT results32,36. However, recent aspirin use was not associated in this study. One reason may be that all participants were asked to stop using aspirin at least 7 days before endoscopy. Thus the effect of aspirin use on FIT results was attenuated. In another large-scale study from Hong Kong including 5539 asymptomatic participants, older age, smoking, and polypoid adenoma were associated with false-positive results16. In this study, smoking and drinking were all not found associated with false-positive FIT results after adjustment.
As to the colonoscopy findings, colorectal inflammation and the presence of ≥ 3 non-advanced adenomas were found to increase the risk of having false-positive results in this study. Colorectal inflammation presents as hyperemia and edema of mucosa under the endoscopy, which may result in a small amount of blood loss and lead to false-positive FIT results. We found that the presence of ≥ 3 non-advanced adenomas was associated with false-positive results of FIT. A study from the Netherlands made the same conclusion by logistic regression analysis37. One plausible explanation may be that lesions may have a wider surface area in contact with feces with the increase of numbers. The concentration of fecal hemoglobin may increase, resulting in false-positive results. Hemorrhoids were commonly thought to be a source of blood loss. A retrospective study from Korea found they were associated with false-positive FIT results, which was different from our findings15,38. A probable reason was that we had excluded participants with obvious bleeding of the digestive tract. Thus, most hemorrhoids detected in our study were not in activity and did not bleed when collecting samples.
Interestingly, we found that some upper gastrointestinal conditions including gastric cancer and hiatal hernia were also associated with false-positive FIT results in bivariable logistic regression analysis. After correcting confounding factors, gastric cancer was independently associated with false-positive FIT results. There had been some researches on the relation between upper gastrointestinal conditions and FIT results and on whether patients with a false-positive FIT result should receive an examination of the upper digestive tract or not18–20,24,39–41. However, they did not correct other confounding factors, and the results were controversial. A study from Korea had investigated the incidence of gastric and esophageal cancer after a positive FIT without colorectal cancer under colonoscopy. It showed that gastric and esophageal cancer risk was higher in FIT +/CRC- patients than FIT- patients (0.68% and 0.38% within one year, respectively), suggesting a relation between upper gastrointestinal cancer and positive FIT results. Another study from Korea retrospectively reviewed data of 340 participants with positive FIT results. All participants underwent both gastroscopy and coloscopy. Results showed that 3 (1.2%) gastric cancers were found in 243 patients with false-positive results. Because the rate of gastric cancer was higher than the incidence of gastric cancer among the general population in Korea, the authors concluded that upper gastrointestinal examination is mandatory in patients with false-positive FIT results20. However, a recent study from the Netherlands found no difference in the incidence of upper gastrointestinal cancers between patients with positive FIT results and negative results in the past 3 years40. Another study reported low PPV of FIT for gastric cancer24. Thus, routine upper gastrointestinal examination among these patients was not recommended in these studies. Conflicting results may be caused by the different incidence of gastric cancer between European and Eastern Asian countries, dietary differences, study designs and sample sizes39,42. In this study, we prospectively enrolled participants with both gastroscopy and colonoscopy. After correcting all other confounding factors, it enabled us to analyze the statistically significant association between upper gastrointestinal diseases and false-positive FIT results. Our results showed that gastric cancer was independently associated with false-positive FIT results. However, given the limited cases, our results may not justify routine upper endoscopy investigation for all false positive patients. In addition to FIT results, the use of other clinical factors for predicting gastric cancer risk may allow the selection of subjects for gastric cancer evaluation, such as older age, symptoms, high FIT concentration, high-incidence regions43,44. In addition, we found that peptic ulcer was not related to false-positive results. In contrast, a study in Singapore retrospectively collected coloscopy results of patients with positive FIT results. In 202 patients with subsequent gastroduodenoscopy, 16 (7.8%) patients had peptic ulcers. The authors regarded it as the source of occult bleeding18. One reason for this discordant result was that peptic ulcers detected in our study were of small size, and most of them were in the healing stage.
Our study prospectively enrolled 929 participants. All participants underwent gastroscopy and coloscopy after completing the FIT examination. However, there are some limitations. Firstly, generalizability to true population-based screening is limited, since this was a hospital-based cohort including symptomatic patients with selection bias. Secondly, all individuals stopped aspirin before colonoscopy and lesions of the small intestinal may be ignored, which limited the analysis of these influencing factors. Besides, inter-operator variability, and the positivity rule of FIT (one of two samples at a low cutoff) potentially inflating false positives. Lastly, this study adopted a relatively low positivity cut-off (10 μg/g). Applying a higher cutoff level will lower sensitivity while increasing specificity, thus influencing the proportion of false‐negative and false‐positive screenees. However, previous meta-analysis showed that differences in cutoff levels between studies might have little impact on the conclusions by sensitivity analysis45. Future large-scale studies are needed to elucidate potential reasons for apparently false-positive FIT results in a routine CRC screening setting.

Conclusions
In conclusion, our study suggested that male, colorectal inflammation, ≥ 3 non-advanced adenomas, and gastric cancer were associated with false-positive FIT results. Although the FIT results may be false-positive for the detection of AN, it may be potentially indicative of other gastrointestinal conditions such as gastric cancer. Other clinical risk factors should be considered when deciding whether to perform upper endoscopy assessment in individuals with false positive FIT results. Further studies should focus on the reasons underlying the false-positive FIT results and implications for further medical intervention.

Supplementary Information