Objective Faecal immunochemical tests (FITs) are replacing guaiac faecal occult blood tests (gFOBTs) for colorectal cancer (CRC) screening. Incidence of interval colorectal cancer (iCRC) following a negative stool test result is not yet known. We aimed to compare incidence of iCRC following a negative FIT or gFOBT.
Design We searched Ovid Medline, Embase, Cochrane Library, Science Citation Index, PubMed and Google Scholar from inception to 12 December 2017 for citations related to CRC screening based on stool tests. We included studies on FIT or gFOBT iCRC in average-risk screening populations. Main outcome was pooled incidence rate of iCRCs per 100 000 person-years (p-y). Pooled incidence rates were obtained by fitting random-effect Poisson regression models.
Results We identified 7 426 records and included 29 studies. Meta-analyses comprised data of 6 987 825 subjects with a negative test result, in whom 11 932 screen-detected CRCs and 5 548 gFOBT or FIT iCRCs were documented. Median faecal haemoglobin (Hb) positivity cut-off used was 20 (range 10–200) µg Hb/g faeces in the 17 studies that provided FIT results. Pooled incidence rates of iCRC following FIT and gFOBT were 20 (95% CI 14 to 29; I2=99%) and 34 (95% CI 20 to 57; I2=99%) per 100 000 p-y, respectively. Pooled incidence rate ratio of FIT versus gFOBT iCRC was 0.58 (95% CI 0.32 to 1.07; I2=99%) and 0.36 (95% CI 0.17 to 0.75; I2=10%) in sensitivity analysis. For every FIT iCRC, 2.6 screen-detected CRCs were found (ratio 1:2.6); for gFOBT, the ratio between iCRC and screen-detected CRC was 1:1.2. Age below 60 years and the third screening round were significantly associated with a lower iCRC rate.
Conclusion A negative gFOBT result is associated with a higher iCRC incidence than a negative FIT. This supports the use of FIT over gFOBT as CRC screening tool.
- colorectal cancer screening
Statistics from Altmetric.com
Significance of this study
What is already known on this subject?
Stool tests are widely used for colorectal cancer (CRC) screening.
At present, guaiac faecal occult blood testing (gFOBT) is being replaced by faecal immunochemical testing (FIT), which is more sensitive for the detection of CRC as well as its precursors.
However, little is known about the interval colorectal cancer (iCRC) rate following a negative stool test, including gFOBT and FIT, while this is a key quality indicator in screening programmes. Moreover, it is not known how gFOBT and FIT compare with regard to iCRC incidence.
What are the new findings?
Higher incidence rate of iCRC is found after a negative gFOBT compared with FIT, 34 versus 20 iCRCs per 100 000 person-years, respectively.
For every FIT iCRC, 2.6 screen-detected CRCs are found; for gFOBT, the ratio between iCRC and screen-detected CRC is 1:1.2.
How might it impact on clinical practice in the foreseeable future?
The outcome of this meta-analysis helps to adequately inform screenees about the risk of interval cancers after a negative gFOBT or FIT.
A negative gFOBT result is associated with a higher iCRC incidence than a negative FIT result. This favours the use of FIT as CRC screening test over gFOBT. Different faecal haemoglobin positivity thresholds may result in different iCRC incidence rates.
Worldwide, colorectal cancer (CRC) is the third most common cancer in men and the second in women.1 Randomised controlled trials have shown that screening with guaiac faecal occult blood tests (gFOBTs), and subsequent colonoscopy if the result is positive, is associated with a 15%–33% decrease in CRC-related mortality.2–4 Consequently, these stool tests are widely used for CRC screening.5
A cost-effective screening programme is inevitably associated with the occurrence of what are known as interval colorectal cancers (iCRCs)—defined as CRCs detected after a negative screening test and before the next recommended test is due.6 7 The rate of occurrence is strongly related to the sensitivity of a screening test and reflects the quality of a screening programme.8 9 Therefore, the International Agency for Research on Cancer recommends to collect and report data on iCRCs.8 In CRC screening programmes, iCRCs are cases either missed by stool tests or at colonoscopy.7 Prevalence and associated risk factors of post-colonoscopy iCRCs in screening programmes have been described.10 11 However, data on prevalence and associated risk factors of iCRCs following negative occult stool tests are still lacking.
In faecal occult blood tests (FOBT)-based CRC screening programmes, gFOBTs have been the most commonly used occult stool tests for years. At present, gFOBT is rapidly replaced by faecal immunochemical testing (FIT).5 FIT detects human-specific globin, whereas gFOBTs react with heme, including consumed non-human heme. FITs are more sensitive for the detection of CRC as well as its precursors than gFOBTs.12 13 Moreover, FITs allow single stool testing, are easier to handle, have higher participation rates and provide quantitative test results, which enables to adjust the positivity cut-off to match available resources.14–16 Despite these advantages of FIT over gFOBT, gFOBT is still being used in several regions.5 Although interval cancer rate is a key quality indicator in screening programmes, data on the incidence rate of FOBT iCRC are limited and no data are available on how these two types of FOBT compare with regard to iCRC.
We therefore performed a systematic literature search and meta-analysis to determine and compare the pooled incidence rates of iCRC following gFOBT and FIT in population-based CRC screening programmes. Second, we assessed if screening-related or patient-related factors are associated with iCRC incidence rate.
Search strategy and selection criteria
We carried out a systematic review and meta-analysis of published trials, including observational and experimental trials or both, according to the preferred reporting items for systematic reviews and meta-analyses (PRISMA)-guidelines.17 We additionally used a checklist containing specifications for reporting of meta-analysis of observational studies in epidemiology (MOOSE).18
Studies were identified through a systematic literature search until 10 May 2016 in the following electronic databases: Medline, Embase, Web of Science, the Cochrane Library, PubMed publisher and Google Scholar. A search update was performed on 12 December 2017. The search strategy was designed and conducted using controlled vocabulary supplemented with keywords and without any restrictions on date or language (online supplementary material 1). The titles and abstracts of identified studies were reviewed by at least two of the authors independently (EW, EHS or EJG). Studies were excluded that did not address the research question, based on the inclusion and exclusion criteria mentioned below. The full texts of the remaining publications were carefully and independently examined by the same authors. In case of disagreement, consensus was reached by consulting a fourth author (MCWS). In addition, the reference lists of the included studies were hand-searched to identify additional, potentially relevant studies (published within 5 years preceding our search).
Supplementary file 1
Studies were included if they reported on CRC occurrence within 1 to 5 years after a negative gFOBT or FIT in average-risk screening populations. Both prospective and retrospective studies were included. Only studies comprising asymptomatic average-risk individuals aged 40 years and above were included, as these were considered representative for a population-based CRC screening programme. Studies were eligible if participants with a positive test were referred for endoscopic confirmation. For the purpose of this systematic review, diagnostic tests accepted as endoscopic confirmation included colonoscopy, or if colonoscopy was not available or contraindicated sigmoidoscopy, CT colonoscopy or double-contrast barium enema. Only full-text articles were included. We did not restrict studies based on language or publication date. If the same screening cohort was described in more than one publication, the one with the most recently updated and most complete data was included.
Accuracy studies in which all participants underwent both the stool test and colonoscopy were excluded. Also excluded were reviews, systematic reviews, editorials and letters to the editor. Lastly, studies in which individuals were referred to endoscopy after two or more consecutive positive tests were excluded.
The primary outcome was the pooled incidence of iCRC during gFOBT and FIT screening per 100 000 person-years (p-y) in an average CRC screening population. Secondary outcomes were the proportional rate between iCRCs a screen-detected CRCs and pooled incidence of iCRCs per subgroup. Subgroups were categorised by means of screening-related and screenee-related factors, including number of screening round, duration of follow-up after a negative stool test, positivity cut-off, gender, age, tumour stage and tumour location.
Screen-detected CRC was defined as a CRC detected by endoscopic conformation after a positive test. Interval colorectal cancers were defined in agreement with the definition of the World Endoscopy Organisation as cancers diagnosed after a negative test and before the next test was due.7 Post-colonoscopy CRCs diagnosed after a negative colonoscopy were not taken into account. If a study did not describe when the next occult blood test was due, we assumed the interval to be 2 years. Proximal CRCs were defined as CRCs located in the caecum, ascending colon, transverse colon or splenic flexure; and distal CRCs as CRCs located in the descending colon, sigmoid colon or rectum. Early CRC was defined as Dukes A, or TNM stage 1.19 In case of quantitative FIT, we converted units for positivity cut-off into micrograms (µg) of haemoglobin (Hb) per gram of stool for each study.20
Study characteristics and data were independently extracted by two investigators (EW and EHS) and recorded on a standardised data extraction form. Any discrepancies were resolved by consensus. The types of data extracted are shown in online supplementary material 2. If data were incomplete, the corresponding author was asked to provide the missing information. Alternatively, we derived data from other publications on the same study cohort. If applicable, data from multiple screening rounds were included for analysis.
Incidence rates of iCRC were calculated per 100 000 p-y. The follow-up p-y were calculated as the number of participants with a negative test multiplied by the mean years of follow-up or the number of years for which interval cancers were identified, by using data from the cancer registry.
Pooled incidence rates were obtained by fitting random-effect Poisson regression models. Heterogeneity was quantified by using the inconsistency index (I2) test. Heterogeneity values ranged from 0% (no heterogeneity) to 100%.21 22 I2 greater than 25%, 50% and 75% was defined as indicative of low, moderate and high heterogeneity.22 For studies describing both a gFOBT and FIT study arm, we interpreted each arm as a separate study, ignoring the within-study correlation. We used prediction intervals to calculate the expected incidence of iCRC for new settings similar to the ones included in the meta-analysis.
An incidence rate ratio was used to compare pooled incidence rates of iCRC after FIT and gFOBT. The sensitivity analysis included only studies in which both a FIT and gFOBT study arm was described. This allowed comparison of incidence rate of iCRC between the two test types in the same study population.
Meta-regression analyses served to assess if screening-related and patient-related factors were associated with FOBT iCRCs. For these analyses, gFOBT and FIT studies were pooled together as only few studies reported on these factors. Incidence rate ratios or proportions were used to describe categorical variables. Relative risk was used for continuous outcomes.
A funnel plot was created to assess the presence of publication bias.23 The study quality of observational studies was assessed using the Newcastle-Ottawa criteria of Wells et al.24 Studies were considered as high-quality studies in case of a score of eight or nine out of nine stars according to the Newcastle-Ottawa criteria, absence of selection bias and adequate cohort follow-up. Selection bias was considered to be present if <90% of the total inception cohort was followed. With respect to study follow-up, a minimum of 2 years’ follow-up was required to define a high-quality study. The study quality of randomised trials was assessed using the Cochrane risk of bias tool.25 We performed a post hoc subset analysis with high-quality studies only to assess incidence of iCRC. The quality of evidence was rated by the Grading of Recommendations Assessment, Development and Evaluation (GRADE).26
All analyses were done using R V.2.15.1.
In total, 7 426 records were identified. After removal of duplicates, 3 526 records were screened for eligibility based on title and/or abstract. In total, 452 full-text records were reviewed for eligibility criteria, of which 423 were excluded for various reasons (figure 1). Thus, 29 studies were included for qualitative and quantitative analysis.2 3 6 27–52
Characteristics of the included studies are shown in table 1. Nineteen studies were performed in Europe, seven in Asia and three in North America. Fourteen studies contained data on FIT-related iCRCs, twelve on gFOBT-related iCRCs and three on both gFOBT-related and FIT-related iCRCs. The median faecal Hb positivity cut-off in the 17 studies that provided FIT results was 20 (range 10–200) µg Hb/g faeces. The study quality score of the 27 observational studies ranged from four to eight stars according to the Newcastle-Ottawa criteria (online supplementary table 1). The two randomised controlled trials were both scored as good-quality studies according to the Cochrane risk of bias tool.32 39
Meta-analysis comprised data of 6 987 825 screening participants with a negative FOBT result, ranging from 1 071 to 2 033 526 participants per study (table 1). Of all 29 included studies, total follow-up for participants with a negative screening test was 32 million p-y, with a mean follow-up of 4.0 years. In these studies, 11 932 screen-detected CRCs (range 6 to 2 961) and 5 548 iCRCs (range 0 to 2 047) were documented. For every iCRC, 2.6 screen-detected CRCs were found with FIT. In gFOBT-based studies, the ratio between iCRC and screen-detected CRC was 1:1.2. The forest plot of the ratio of iCRC following a negative stool test compared with screen-detected CRC is shown in online supplementary figure 1. This ratio was 0.19 (95% CI 0.13 to 0.27) for FIT studies compared with 0.36 (95% CI 0.28 to 0.45) for gFOBT studies, p=0.005, I2=99%.
The overall pooled incidence rate of iCRC following a negative stool test was 26 (95% CI 19 to 36; I2=99%, n=29 studies) per 100 000 p-y (figure 2). Pooled incidence rates of iCRC for FIT and gFOBT were 20 (95% CI 14 to 29; I2=99%) and 34 (95% CI 20 to 57; I2=99%) per 100 000 p-y, respectively (figure 2). The pooled incidence rate ratio between FIT iCRC and gFOBT iCRC was 0.58 (95% CI 0.32 to 1.07, n=29 studies). The GRADE level of evidence was very low (online supplementary table 2). The funnel plot provided no evidence for the presence of publication bias (online supplementary figure 2). The prediction intervals of the incidence rate of FIT and gFOBT iCRC are shown in figure 2.
Subgroup analysis of the studies with high quality established with the Newcastle-Ottawa criteria and Cochrane risk of bias tool yielded an incidence rate of iCRC after FIT of 15 (95% CI 8 to 30, n=7 studies) and after gFOBT of 55 (95% CI 35 to 87, n=8 studies) per 100 000 p-y.
Three studies that described both a FIT and gFOBT arm were included in a sensitivity analysis to compare the incidence rate of iCRC between FIT and gFOBT.32 39 48 The pooled incidence rate ratio between FIT iCRC and gFOBT iCRC was 0.36 (95% CI 0.17 to 0.75, I2=10%). This ratio was classified as high-quality evidence according to the GRADE score (online supplementary table 2).
For meta-regression analyses, data of FIT and gFOBT studies were pooled.
Fifteen out of 29 studies provided data on test-related iCRCs after the first screening round (table 2). Five studies also provided data on iCRC after the second round, four studies after the third and one study after four rounds of screening. After the third round, there was a significant lower risk of iCRC after a negative test compared with the first round (table 2).
Eight out of 29 studies provided data on iCRCs 1 year after a negative test. Six studies provided data on iCRC 2 years after a negative test and two studies 3 years after a negative test (table 2). Compared with 1 year after a negative FOBT, the relative risk of iCRC was 1.25 (95% CI 1.05 to 1.49) after 2 years and 1.19 (95% CI 0.89 to 1.13) after 3 years.
Thirteen out of 17 FIT studies used a single quantitative positivity cut-off, ranging from 0 to 100 µg Hb/g faeces. Association between FIT cut-off and FIT iCRC yielded a relative risk of developing FIT iCRC of 1.00 (0.89–1.13) per 10 µg Hb/g faeces cut-off increase (table 2).
Based on three studies, the iCRC incidence rate ratio between men and women was 1.22 (95% CI 0.94 to 1.57, I2=0%). And based on two out of 29 studies, the iCRC incidence rate ratio of screenees <60 years of age to screenees ≥60 years was 0.25 (95% CI 0.09 to 0.65, I2=62%) (table 3).
Eight out of 29 studies described the location of iCRCs in the colon.6 27 31 33 39 50–52 Based on these studies, iCRCs were located distal from the splenic flexure in 67% (95% CI 64% to 70%, I2=0%) of cases. Six out of 29 studies described tumour stages of iCRCs.6 27 31 33 51 52 These iCRCs were staged as early CRCs in 22% (95% CI 17% to 28%, I2=62%) of cases.
This is the first systematic review and meta-analysis to estimate the pooled incidence rates of iCRCs following negative FOBTs in a CRC screening setting. It showed that iCRCs occur in both FIT-based and gFOBT-based CRC screening. However, the incidence of iCRC is higher after a negative gFOBT than after a negative FIT.
Interval cancer rates reflect the sensitivity of a screening test and quality of a screening programme. International guidelines, therefore, designate the interval cancer rate as an important outcome measure.8 14 Pooled data on incidence of interval cancers following a negative gFOBT and FIT have been awaited for some time.14 53
The findings of this meta-analysis emphasise that screenees should be adequately informed about the risk of CRC after a negative occult blood test. They may mistakenly feel disease-free and fail to respond to CRC symptoms.54 A Swedish study indeed reported a significant delay in CRC diagnosis among those with a false-negative FOBT.55 However, recent evidence showed that both overall and CRC-specific survival rates were better for gFOBT interval cancers than for cancers arising in a non-screened population.6 We found that iCRC accounted for a significant proportion of CRC found in both gFOBT-based and FIT-based screening programmes. In the included gFOBT studies, the total number of CRCs missed by gFOBT almost equalled the number of screen-detected CRCs.
The higher incidence of iCRCs after a negative gFOBT compared with FIT in sensitivity analysis is likely due to the higher sensitivity of FIT for the detection of Hb. Best evidence suggests that the most used gFOBT probably has an effective cut-off of around 150 µg of Hb/g of faeces, whereas the accurate detection level of most FITs lies at 5–10 µg Hb/g of faeces.13 56 57 The incidence of iCRC as primary outcome did not significantly differ between gFOBT and FIT. This may have been due to excessive study bias as shown by subgroup analyses.
We found that older age was associated with a higher iCRC incidence after a negative test. Indeed, the elderly have a higher risk of CRC and its precursors.58–60 Further, the risk of a FOBT-related iCRC was not significantly different between men versus women. This implies, in view of the fact that FOBT screening detects more CRCs in men,61 that the ratio of screen-detected CRCs versus interval cancers is less favourable in women than in men. Furthermore, we found lower risks of FOBT-related iCRC with every screening round compared with the first round. A possible explanation for this finding is that with every screening round, more CRCs are detected and therefore changes of missing CRC with the stool test decline as well. Lastly, use of a higher positivity cut-off did not result in higher incidence of FIT interval cancers. This finding needs to be confirmed when more data become available. Moreover, studies included in our analyses used low cut-offs, which might be the reason that this association was not found.
For quality assessment of CRC screening, it is recommended to monitor the iCRC incidence.53 Various measures can be used for this purpose. The incidence of FOBT iCRC can be calculated as the ratio of iCRCs versus (1) participants with a negative test, (2) person-years follow-up in those with a negative test, (3) total CRCs (detected and missed) and (4) CRCs expected without screening. In our study, we assessed iCRC per person-years, reflecting the absolute number of iCRC cases in CRC screening populations over time. However, when calculating incidence rates on a programme level, participation rates should also be taken into account. Secondary outcome in our study was the relative rate of iCRC versus screen-detected CRC, which is an indirect measure of test sensitivity. Previously published data revealed a higher test sensitivity for FIT compared with gFOBT.62 63
Although this comprehensive meta-analysis is based on a large number of person-years, the point estimates of our calculated pooled incidence rates of iCRC should be interpreted cautiously. First, high statistical heterogeneity among studies was shown. We assessed the robustness of conclusions concerning the effect sizes of real interest in our meta-analysis as substantial statistical heterogeneity was observed in the overall pooled data. Statistical heterogeneity represents the approximate proportion of total variability in point estimates that can be attributed to heterogeneity in underlying incidence rates. To explain the observed heterogeneity of the incidence rates between studies, we performed subgroup analyses. A potential important source of heterogeneity are differences between populations screened in terms of gender distribution, age distribution and number of performed screening rounds. These were all identified in the subgroup analyses as factors that partly explained the observed heterogeneity. Furthermore, we performed sensitivity analyses by only including studies directly comparing gFOBT and FIT, limiting the influence of factors introducing heterogeneity, to directly estimate the difference in incidence rates of both tests. This analysis resulted in a higher gFOBT iCRC incidence compared with FIT. The marked inconsistency among the included trials in incidence rate ratios for iCRC (I 2 =99%) was substantially reduced (I 2 =10%) when differences between populations were taken into account. Another important factor potentially introducing heterogeneity between studies was study quality. Additional sensitivity analyses of high-quality studies only showed a significant higher gFOBT iCRC incidence than FIT iCRC incidence. Second, test kits may be discrepant in terms of cancer detectability and the resultant future risk of iCRC. Test reliability, stability and the ability to detect invasive cancer or advanced adenoma of different kits have been compared in several previous studies.64 65 Further stratifying analysis in our study to correct for differences in test kits was not feasible. Only few of the included studies reported on iCRCs within specific subgroups; therefore, limited analyses could be done for gFOBT and FIT separately.
In conclusion, interval cancers occur in both gFOBT-based and FIT-based CRC screening programmes. The latter is associated with a lower incidence of iCRC, which further supports the use of FIT over gFOBT.
Contributors EW, EHS and MCWS conceptualised and designed the study. Responsible for data acquisition were EW and EHS. Data analyses were done by EW and DN. Interpretation of data was done by EW and EHS with contribution of all co-authors. The manuscript was drafted by EW and EHS. All co-authors provided critical revision for important intellectual content. Statistical analyses were done by DN. Supervised execution of the study was done by MCWS. All authors read and approved the final manuscript.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None declared.
Patient consent Not required.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.