Objective Colorectal cancer (CRC) screening programmes based on faecal immunochemical testing for haemoglobin (FIT) typically use a screening interval of 2 years. We aimed to estimate how alternative FIT strategies that use a lower than usual positivity threshold followed by a longer screening interval compare with conventional strategies.
Methods We analysed longitudinal data of 4523 Dutch individuals (50–74 years at baseline) participating in round I of a one-sample FIT screening programme, of which 3427 individuals also participated in round II after 1–3 years. The cohort was followed until 2 years after round II. In both rounds, a cut-off level of ≥50 ng haemoglobin (Hb)/mL buffer (corresponding to 10 µg Hb/g faeces) was used, representing the standard scenario. We determined the cumulative positivity rate (PR) and the numbers of subjects diagnosed with advanced adenomas (N_AdvAd) and early stage CRC (N_earlyCRC) in the cohort over two rounds of screening (standard scenario) and compared it with hypothetical single-round screening with use of a lower cut-off and omission of the second round (alternative scenario).
Results In the standard scenario, the cumulative (ie, round I and II combined) PR, N_AdvAd and N_earlyCRC were 13%, 180% and 26%, respectively. In alternative scenarios using a cut-off level of respectively ≥11 and ≥22 ng/HbmL buffer (corresponding to 2 and 4 µg Hb/g faeces), the PRs were 18% and 13%, the N_AdvAd were 180 and 162 and the N_earlyCRC ranged between 22–27 and 22–26.
Conclusions The diagnostic yield of FIT screening using a lowered positivity threshold in combination with an extended screening interval (up to 5 years) may be similar to conventional FIT strategies. This justifies and motivates further research steps in this direction.
- COLORECTAL CANCER SCREENING
Statistics from Altmetric.com
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.
Significance of this study
What is already known on this subject?
Colorectal cancer (CRC) screening programmes based on faecal immunochemical testing for haemoglobin (FIT) typically use a screening interval of 2 years.
Given its quantitative nature, FIT offers various options for CRC screening, for example, selecting a specific cut-off level and screening interval to optimise the impact of screening.
What are the new findings?
Alternative FIT strategies using a markedly lower cut-off level followed by a longer screening interval (up to 5 years) are estimated to provide a similar diagnostic yield as conventional FIT screening.
These alternative strategies require a similar to higher number of follow-up colonoscopies, and half the number of FITs due to the longer interval.
How might it impact on clinical practice in the foreseeable future?
While saving screening rounds, alternative FIT strategies that use a lower cut-off level in combination with a longer screening interval may provide a similar impact as conventional FIT screening. This justifies and motivates further research steps in this direction.
Colorectal cancer (CRC) is the third most common cancer and cause of cancer-related deaths worldwide, with >1.2 million new cases and >600 000 deaths per year.1 Randomised controlled trials have demonstrated that biennial screening with guaiac-based faecal occult blood testing (gFOBT) reduces CRC mortality by 15%.2
Since the conduct of these trials, a substantial body of evidence has shown that the diagnostic performance of faecal immunochemical testing for haemoglobin (FIT) is superior to gFOBT screening.3–6 Furthermore, FIT screening is associated with higher participation rates than gFOBT5 ,6 and has analytical advantages such as no cross-reactivity with dietary constituents and the option of measuring faecal haemoglobin (Hb) levels quantitatively. Accordingly, the European guidelines for quality assurance in CRC screening recommend screening with FIT in preference to gFOBT.7 Current CRC screening programmes based on FIT are typically designed analogously to gFOBT strategies.8 This means that they use a screening interval of 2 years and a positivity threshold yielding a specificity well above 90%.
However, studies reporting on repeated rounds of FIT screening consistently showed that the diagnostic yield of the first rescreening round is lower as compared with the initial screening round.9–11 We further showed that the diagnostic yield in a second round of FIT screening is the same when using a 1-year, 2-year or 3-year interval.9 When combining these observations with the fact that FIT—given its quantitative nature—offers flexibility to select the positivity threshold, it leads to the idea of considering alternative strategies for FIT screening. The positivity threshold at the initial round could be lowered, thereby increasing sensitivity at the cost of decreasing specificity. Subsequently, the interval to the second round could be significantly increased.
Although strategies with a lower positivity threshold in combination with a longer interval may have advantages, for instance, regarding the organisational effort, it needs to be considered carefully whether there are disadvantages, for example, regarding the diagnostic yield and the colonoscopy load. We aimed to explore and estimate how such alternative FIT strategies compare with conventional FIT screening in terms of positivity rate and thus colonoscopy demand, detection of advanced adenomas and CRC detection.
To address the research question, we analysed longitudinal data of an ongoing population-based CRC screening study that started in 2006 and aimed to investigate the diagnostic yield and adherence patterns over multiple rounds of FIT screening. As illustrated in figure 1, we focused on the first and the second round of this study and considered the positivity rate and the diagnostic yield of these two rounds cumulatively. We then compared this standard scenario with various alternative scenarios that assumed use of a lower cut-off level in the first round and omission of the second round, resulting in an extended screening interval.
Study design and study population
Details about the design of the study have been described elsewhere.6 ,9 Briefly, demographic data of all individuals aged 50–74 years living in the southwest of The Netherlands were obtained from municipal population registers to identify the target population. This population was screening-naïve since there was no CRC screening programme at the time of recruitment for this study. Random samples were taken from the target population by a computer-generated algorithm (Tenalea, Amsterdam, The Netherlands) to be invited for successive FIT screening rounds. A screening interval of 1, 2 or 3 years was assigned to equally sized groups between the first and the second round, while the screening interval was 2 years for all subjects between the second and the third round. Exclusion criteria were a history of CRC; IBD; an estimated life-expectancy of <5 years; a colonoscopy, sigmoidoscopy or double-contrast barium enema within the previous 3 years and inability to give informed consent. Subjects were no longer invited to subsequent rounds if they tested positive at a prior screening round, if they had become ≥75 years of age, if they had moved out of the region or had died. The occurrence of CRCs in the study population was determined by record linkage with the Dutch Comprehensive Cancer Centre (http://www.iknl.nl). Results of the first, second and third round of the study have been reported previously.6 ,9 ,12
With each screening round, one FIT (OC-Sensor Micro, Eiken Chemical, Tokyo, Japan) was sent by mail to collect a single stool sample of one bowel movement. The test was considered positive when the Hb concentration in the FIT sample was ≥50 ng/mL buffer, which corresponds to 10 µg Hb/g faeces. Applying FIT at this cut-off level is called FIT50 in the following, and analogous terms are used for other cut-off levels. Subjects with a positive FIT50 were scheduled for colonoscopy within 4 weeks. Experienced endoscopists performed all colonoscopies. The maximum reach of the endoscope, adequacy of bowel preparation as well as the characteristics and location of any lesions were recorded. Experienced GI pathologists evaluated all removed lesions. Patients with relevant findings entered a surveillance programme according to the guidelines of the Dutch Society of Gastroenterology, while patients with a negative colonoscopy were considered not to require FIT screening for 10 years.13
For the present analyses, we only included subjects who participated in the first screening round because we required baseline faecal Hb for our approach. The cohort was followed from the first round (baseline), over the second round up to the time point of the third screening round, yielding a follow-up period of 3, 4 or 5 years depending on the interval assigned between the first and the second round.9
In a first step, we considered the standard scenario with two FIT50 rounds and determined the cumulative positivity rate (defined as the proportion of participants with a positive test result), the cumulative number of subjects diagnosed with advanced adenomas as well as the total number of CRCs (detected either as a result of screening or during follow-up as interval CRC). We determined the number of screen-detected CRCs and non-screen-detected CRCs (ie, interval CRCs). We also determined the number of CRCs that were diagnosed at an early stage (Union Internationale Centre le Cancer (UICC) I or II), regardless of whether they were screen-detected or not.
In a second step, we estimated these outcomes for various alternative scenarios which assumed that a cut-off level ≤50 ng/HbmL buffer had been used for FIT in the first round (baseline) and the second round had been skipped. The various alternative scenarios only differed with respect to the cut-off level used to classify baseline Hb levels as positive or negative.
Positivity rates of the various alternative scenarios could be assessed directly by determining the proportion of the cohort whose baseline Hb levels were equal or above the respective cut-off level. The number of subjects diagnosed with advanced adenomas at the alternative scenarios could not be assessed directly given that only those with Hb levels ≥50 ng/mL buffer underwent colonoscopy at baseline. Therefore, we used an indirect approach to estimate this number as described in table 1. The approach made use of the results of a diagnostic study that was conducted in a similar study population in the same Dutch region and determined test performance characteristics of OC-Sensor Micro (Eiken Chemical) at various cut-off levels.14 In this study, all included subjects underwent both FIT and colonoscopy.
To determine the number of screen-detected CRCs for the alternative scenarios, we considered persons who were diagnosed with CRC until the end of the follow-up period (screen-detected or interval CRC) and assessed whether their baseline Hb level was equal or above the cut-off level of the respective scenario. We also determined the number of early stage CRCs (UICC I or II) for the alternative scenarios, taking into account uncertainty regarding stage progression as far as relevant. Accordingly, we distinguished between CRCs that were definitely or possibly detected at an early stage. The assumptions and criteria we used for classifying CRCs as screen-detected and early stage in the alternative scenarios are described in detail in the online supplementary material.
We also determined the number of FITs and follow-up colonoscopies for the standard scenario and the various alternative scenarios.
Overall, 4523 subjects (48% male) participated in the first FIT50 round. The mean age (SD) was 60.5 (6.6) years. In the second round, 742 subjects (16%) were no longer eligible and 354 subjects (8%) were eligible, but did not participate. This left 3427 (76%) participants in the second FIT50 round (91% of those being eligible). A flow chart of the cohort from baseline to the end of the follow-up period is provided in the online supplementary material. A total of 36 CRCs were diagnosed in the cohort; 22 with first-round FIT50, 6 with second-round FIT50 and 8 during intervals. Figure 2 provides information on these CRCs, including whether they were screen-detected with FIT50 or interval CRCs, the respective baseline FIT level, the stage at diagnosis and the time between baseline and diagnosis.
Table 2 shows the outcomes of interest for the standard scenario and the various alternative scenarios. In the standard scenario (ie, the two FIT50 rounds), the cumulative positivity rate was 13%. Advanced adenomas were detected in 180 subjects, the majority of which (70%) were diagnosed in the first FIT50 round.9 Overall, 26 CRCs (72% of all CRCs) were diagnosed at an early stage, of which 22 were screen-detected.
Regarding the alternative scenarios, table 2 shows the results for selected cut-off levels ranging from 11 to 50 ng/mL. Using FIT11 for the alternative single initial screening round yielded a similar number of subjects with advanced adenomas as the standard scenario, while the positivity rate was 5% higher (ie, 18% vs 13%; p<0.0001). The number of CRCs diagnosed at an early stage was estimated to range between 22 and 27. The lower estimate (ie, 22) only included CRCs that would definitely have been diagnosed at an early stage, while the upper estimate (ie, 27) also included those that would possibly have been diagnosed at an early stage (see Methods section and online supplementary material). A single-round FIT22 scenario yielded a similar positivity rate as the two-round standard scenario, while the number of subjects diagnosed with advanced adenomas was estimated to be 10% lower. The number of early stage CRCs was estimated to range between 22 and 26. For FIT36, the positivity rate decreased to 10% and the number of subjects diagnosed with advanced adenomas was 25% lower as compared with the standard scenario. The number of early stage CRCs was estimated to range between 21 and 25. The online supplementary material provides information regarding the overlap in subjects diagnosed with advanced adenomas between the standard and the alternative scenarios.
There was a trade-off between a higher number of FITs in the standard versus a higher number of colonoscopies in the alternative scenarios when the latter used cut-off levels <22 ng/mL. For alternative scenarios with cut-off levels ≥22 mg/mL, both the number of FITs and colonoscopies were lower than in the standard scenario (table 2).
The quantitative nature of FIT as opposed to gFOBT offers new options for CRC screening that are yet to be fully explored. In this study, we explored FIT strategies that use a lower than usual positivity threshold in combination with an extended screening interval (up to 5 years), based on the analysis of data from repeated FIT screening. While such strategies would save screening rounds, our results suggest that they likely do not markedly differ from conventional strategies with respect to diagnostic yield and cumulative positivity rate, with some trade-off depending on the respective cut-off level. There were scenarios with similar lesion detection and a higher number of follow-up colonoscopies, and scenarios with slightly lower lesion detection and a similar number of follow-up colonoscopies. The number of FITs was, as a matter of course, halved compared with the standard scenario due to the longer interval.
The next question is what motivates to further pursue this research, that is, what may be advantages of such screening strategies? An obvious advantage is the reduction in the number of screening rounds, which saves efforts and costs related to the organisation of each screening round and may also reduce the burden to the screenees. For example, offering screening from age 50 to 70 years requires 11 rounds when a 2-year interval is used, but only 5 rounds when a 5-year interval is used. Furthermore, the increased likelihood of detecting lesions at one single round going along with the higher per-test sensitivity (resulting from the lower positivity threshold) would be particularly advantageous for subjects who participate in screening on an irregular basis. For example, a person with advanced adenomas who participates in screening only once has a 50% chance of being detected by screening when FIT11 is used, but only a 35% chance when FIT50 is used. In this study, this potential advantage was not apparent because the proportion of subjects who participated in both FIT50 rounds was very high. However, it is expected to be relevant in settings with less favourable patterns of longitudinal adherence. This may occur when there is no or a suboptimal invitation system, as it is often the case in countries with a decentralised health system. For example, a retrospective cohort analysis from the USA evaluating adherence to repeated yearly FOBT showed that among 395 000 subjects who received exclusively FOBT and no other screening tool, about 40% were tested only once during a 5-year study period.15 If strategies with a longer screening interval are used, the design needs to make sure that non-participants are reinvited after a reasonable time frame and not several years later when invitation to the next screening round is due.
Colonoscopy capacity needs to be taken into account when discussing the implications of alternative FIT strategies that use a lower positivity threshold. Our results suggest that the total demand for workup colonoscopy could be rather similar to conventional strategies, with some variation depending on the respective cut-off level. From a programme perspective, the higher number of positive tests per invited birth cohort is more or less compensated for by the lower number of birth cohorts that is invited per round due to the longer interval. It is important to note that this compensation is achieved during the steady state of an established screening programme. When a programme based on such alternative FIT strategies is started, the time taken for complete roll-out may need to be adapted in settings with limited colonoscopy capacity.
To our knowledge, there are no other studies with a similar approach. We previously used microsimulation modelling to assess cost-effectiveness of various FIT screening strategies over a period of 30 years, varying the cut-off level (50–200 ng/mL), the screening interval (1–3 years) and the age range to which screening is offered.16 Screening at a cut-off level of 50 ng/mL was found to be more cost-effective than at higher cut-off levels, which held true for the range of explored intervals. This indirectly supports our finding that screening at a lower cut-off level in combination with a longer interval may not be disadvantageous. The current study adds to these findings by lowering the cut-off level together with extending the screening interval in ranges not previously explored.
Empirical evidence on FIT screening with intervals of 3 or more years is limited. The European guidelines for quality assurance in CRC screening and diagnosis recommend that the screening interval for FIT should not exceed 3 years,7 referring to three case–control studies from Japan.17–19 These studies determined the risk of developing CRC and of dying from CRC according to FIT screening history However, the subgroup analyses that focused on the optimal screening interval had methodological issues regarding sample size and confounder adjustment as detailed in the online supplementary material. Apart from that, the positivity rate was 2.4% in these studies indicating that a high cut-off level was used. This is in contrast to our approach that compensated for the longer interval by lowering the cut-off level.
Experimental evidence for varying the screening interval between 1, 2 and 3 years without varying the cut-off level has recently been reported from the study that the present analysis was based on.9 The detection rate of advanced neoplasia in the second round did not vary between groups assigned to the different intervals. Although the comparison was not powered to detect small differences in detection rate between groups, these findings question the intuitive thinking that less frequent FIT screening (at a constant cut-off level) inevitably decreases the cumulative diagnostic yield. With respect to our analysis, it suggests that the screening interval could be extended to 5 years when the skipped round is compensated by use of a lower cut-off level in the former round.
Our approach used lower than usual positivity thresholds for FIT. Generally, this is a way to bridge the difference between FOBT screening and primary colonoscopy screening, where all screenees are offered colonoscopy irrespective of any blood in stool. As shown in our analysis, the overall demand for colonoscopy of such strategies may be kept at the same level as for conventional FIT screening if the screening interval is extended. Diagnostic studies suggest that when lowering the positivity threshold to the Hb levels that we used in our analysis, the increase in sensitivity relative to the decrease in specificity is similar to higher cut-off levels.14 Accordingly, we expect similar conclusions if our approach is applied to data from programmes that used higher cut-off levels. In other words, lowering the positivity threshold (although not to the levels we used in our analysis) and extending the interval could also be an option for settings with a lower colonoscopy capacity. For some FIT products, analytical imprecision (ie, variability between measurements repeated under similar conditions) has been shown to vary according to the cut-off level.20 Thus, if low cut-off levels are used for routine screening, the precision profiles need to be optimised accordingly. The recent optimisation of buffers of several FITs has already contributed to this issue and allows use of lower cut-off levels.
There are strengths and limitations to this study that should be noted. We developed an approach that was directly linked to empirical longitudinal data of FIT screening. This avoided the need for assumptions regarding conditional independence of sequential FIT testing, which plays a key role in the context of exploring longer screening intervals. There were uncertainties regarding stage progression for some CRCs, which we tried to address in a systematic and transparent way. The same applies to the number of adenomas detected in the alternative scenarios, which we estimated by combining the findings of the first FIT50 round with the results of a study comparing FIT and colonoscopy in all subjects screened.14 The latter was conducted in a similar setting and showed a similar detection rate for FIT50 regarding advanced adenomas (3.1% vs 2.8% in the first FIT50 round). The sample size of the diagnostic study was moderate, but the course of the receiver operating characteristic (ROC) curve at lower cut-off levels was confirmed by other diagnostic studies on quantitative FIT.21
There is uncertainty whether the clinical benefit of detecting advanced adenomas that bleed less versus detecting those that bleed more is similar. It may be that the latter are more likely to progress to lesions that would be symptom-detected at an early stage anyway (and would thus not require screen-detection), but it may also be that the level of bleeding is positively associated with the progressive potential. As possible limitations, it should also be noted that not all CRCs testing positive with FIT at baseline may actually have been detected at workup colonoscopy and that de novo CRCs could have occurred that were not detectable at baseline.
Although imperfect, our approach can be considered as a step that helps deciding whether or not trials that directly compare conventional and alternative FIT strategies with an extended screening interval are justified. Further preliminary evidence from other databases on repeated FIT screening that are analysed in the same way would be of value: first, for the purpose of validation given the inherently low number of CRCs detected after the initial screening round in our study. Second, studies that used a higher cut-off level at the initial round would provide insights regarding generalisability to other settings. With respect to the time frame, our analysis was focused on skipping the FIT round that follows the initial screening. It will be interesting to conduct similar analyses for subsequent screening rounds to explore alternative FIT strategies from a longer-term perspective. Finally, as the diagnostic yield of FIT differs by age and sex, increased sample sizes would allow for subgroup analyses according to these factors.
In conclusion, our findings suggest that the diagnostic yield of alternative FIT strategies using a lower than usual positivity threshold in combination with an extended screening interval (up to 5 years) may be similar to conventional strategies. This justifies and motivates further research steps in this direction given that such alternative strategies could present interesting options for CRC screening, either generally or in particular settings.
Contributors UH, EJG and EJK conceived the idea for this analysis. The data used for this analysis stem from a study that was designed and conducted by EJK. UH and EJG collaborated on the data analysis. UH drafted the manuscript. EJG, IL-V, MCWS and EJK critically reviewed the manuscript and gave important intellectual input. All authors approved the final version of the manuscript.
Competing interests None declared.
Patient consent Obtained.
Ethics approval Dutch Ministry of Health.
Provenance and peer review Not commissioned; externally peer reviewed.