Developing a New Stool Test for Colorectal Cancer Screening, the multitargetFIT (mtFIT)

Abstract

Early detection of colorectal cancer (CRC) through screening is the most effective method to reduce morbidity as well as mortality from this disease. Fecal immunochemical test (FIT)-based screening has shown to be effective, especially in multi-round population-based screening. However, its sensitivity and specificity are suboptimal, leaving room for improvement. In this perspective, the development journey of a new screening test, the multitargetFIT (mtFIT), which outperforms FIT, is described from discovery down to large-scale clinical utility testing and health technology assessment.

Context/Introduction

Colorectal cancer (CRC) is the third most common cancer worldwide and the second leading cause of cancer-related deaths [1]. Early detection through screening is the most effective method to reduce both morbidity as well as mortality due to this disease. The fecal immunochemical test (FIT) detects hemoglobin in stool samples and provides a cost-effective and non-invasive screening option, in particular when applied in the setting of multi-round population-based screening programs. However, its sensitivity and specificity are suboptimal, particularly in detecting advanced adenomas. In this context, the development of the multitarget fecal immunochemical test (mtFIT) is a pivotal step forward in the field of CRC screening.

Unlike the standard FIT, which measures only hemoglobin, mtFIT measures three biomarkers: hemoglobin, calprotectin, and SERPINF2. This multi-biomarker approach significantly improves the test's sensitivity without compromising specificity, in particular by enabling better detection of precursors of colorectal cancer, such as advanced adenomas.

Developing such a diagnostic test is a complex and challenging process (Fig. 1). This perspective aims to provide an overview of this journey, from a researchers’ perspective, highlighting both facilitators and hurdles encountered during the journey (Fig. 1b), while following the principles underlying evaluation of new non-invasive tests for screening journey [2].

Fig. 1.

Biomarker test development roadmap. A The model version of a biomarker test development trajectory. B The real-world experience of the multitarget FIT development trajectory

A) 1 Marker Identification: The initial stage where potential biomarkers are identified through (often) molecular profiling methodologies. 2 Initial Marker Validation: Confirming that identified biomarkers are linked to the disease or condition. 3 Assay Development: Developing a diagnostic test based on validated biomarkers, including design and prototype creation. 4 Large-scale Marker Validation: Testing the assay in the target population to ensure consistent performance. 5 Health Technology Assessment: Analyzing the economic feasibility [7] of the test, considering costs, accessibility, and potential savings. 6 Clinical Trial: Testing the assay in clinical settings with large patient populations to assess its performance in the target population. 7 Test Approval: Regulatory bodies (e.g., FDA, IVDR) evaluate the test’s efficacy, safety, and practicality for commercial use.

B) 1 Marker Identification: The initial stage where potential biomarkers were identified through mass spectrometry-based proteomics. 2 Initial Marker Validation: Confirming that these identified biomarkers were reproducible detected in another sample set. 3 Initial Assay Development (RUO: Research Use Only): developing a diagnostic test based on validated biomarkers, including design and prototype creation. 4 Large-scale Marker Validation: different populations. Testing the assay retrospectively in prospectively collected samples of the target population. 5 Early Health Technology Assessment: Analyzing the economic feasibility of the test, considering costs, accessibility, and potential savings. 6 Assay Production (RUO): Production of a large batch of assays in preparation of the clinical trial. 7 Clinical Trial: Testing the assay prospectively in clinical settings in the target population to assess its performance. 8 Health Technology Assessment: Analyzing the economic feasibility of the test, considering costs, accessibility, and potential savings based on results of the clinical trial. 9 Clinical-grade Assay Development: Redesign a diagnostic test based on validated biomarkers, including possibilities for automated analysis and large-scale production. 10 Clinical Trail: Testing the new clinical-grade assay prospectively in clinical settings in the target population to assess its performance. 11 Technology Assessment: Analyzing the economic feasibility of the test, considering costs, accessibility, and potential savings based on results of the clinical trial. 12 Test Approval: Regulatory bodies (e.g., FDA, IVDR) evaluate the test’s efficacy, safety, and practicality for commercial use. Future steps are indicated in gray.

Marker Identification

The first step in a biomarker test development is the discovery of potential good biomarkers. As a team, we have evaluated several different biomarker types, such as DNA promoter methylation-based, miRNAs, and proteins [3–5], of which the protein biomarkers showed to be more suitable for large-scale use in population-based screening programs. The others imposed constrains that turned out impossible to be solved in terms of cost-effectiveness and/or logistics. As a first step toward a new multiplex protein-based test, we set out to identify novel protein biomarkers that could complement the current FIT. Rather than following the classic approach of first identifying discriminating markers in tissue or cell-line material, followed by validation in the final analyte, we opted for doing the discovery effort directly in the biologic sample used in screening, namely stool, to ensure that identified biomarkers are robust, not degraded in the fecal environment, and therefore with higher chances of success in the following validation steps [6].

Following and optimizing existing protocols [7] we used mass spectrometry-based proteomics to profile stool samples aiming to identify novel protein biomarkers in stool that could outperform or complement hemoglobin in detecting CRC and advanced adenomas.

A major facilitator in this whole development trajectory has been availability of appropriate and well-annotated prospectively collected samples stored in our biobank over many years. These concerned whole-bowel movement stool (whole stool) and/or FIT left-over stool samples that had been collected either from individuals referred to colonoscopy or from individuals participating in FIT-based population screening studies, prior to the scheduled colonoscopy [8–10].

In the first experiment we utilized 22 colonoscopy-controlled stool samples, 12 of which from patients diagnosed with CRC and 10 from individuals without colorectal neoplasia, serving as controls [5]. These samples had been collected in a previous study from symptomatic individuals referred for colonoscopy, where participants were asked to provide a stool sample before undergoing their scheduled colonoscopy. These samples were collected with limited standardization in sample collection. Specifically, these samples were sampled at convenience at home by participants without the use of designated stool collection containers, transported in absence of controlled conditions, and subsequently stored at − 20 °C for an extended period. The goal of this exploratory experiment was to identify biomarkers capable of withstanding such suboptimal handling and storage conditions, thereby indicating their robustness.

The stool samples were filtered for large particles and concentrated. Then, they were fractionated in 10 subsamples based on the molecular weight of the proteins (SDS-PAGE, followed by in-gel digestion), further separated using nanoscale liquid chromatography and analyzed with a LTQ-FT hybrid mass spectrometer (Thermo Fisher). A total of 468 human proteins were identified in this series of samples, of which abundance 93 differed in abundance between CRC vs. control samples (fold change > 0, P ≤ 0.05). As these results confirmed the feasibility of quantifying CRC-specific human proteins in stool samples, the analysis was subsequently extended to a second, larger series of samples.

The second sample set consisted of whole stool samples from 293 individuals from a colonoscopy-controlled referral population. The 291 subjects in this sample set included 79 persons with colorectal cancer (CRC), 40 persons with advanced adenomas, 43 persons with non-advanced adenomas, and 129 persons without colorectal neoplasia (control samples). The samples were processed similarly to the first set, separated using nanoscale liquid chromatography and analyzed by a Q Executive mass spectrometer (Thermo Fisher). In this series, 733 human proteins were identified of which 213 proteins were found to be statistically significantly enriched (fold change > 0, P ≤ 0.05) in CRC samples compared to control samples.

Focusing on the overlap in both series, 29 proteins were statistically significantly enriched in CRC samples after correction for multiple testing (Q ≤ 0.05). As expected, these included hemoglobin subunits alpha 1, beta, and delta (HBA1, HBB, and HBD). In addition, S100 calcium-binding proteins A8 and A9 (S100A8 and S100A9) were present, which are associated with inflammation and have been previously investigated as potential CRC biomarkers. The other proteins on the list are involved in a variety of processes, including immune response (e.g., complement proteins C3, C5, C9, and MPO), iron metabolism (e.g., transferrin and ceruloplasmin), and tissue remodeling (e.g., fibronectin 1 and alpha-2-macroglobulin). Also, other proteins, such as GPI, LDHA, TKT, and TALDO1, are involved in cell metabolism, specifically aerobic glycolysis, which plays an important role in cancer cells.

Biomarker Panels

Since the goal was to identify proteins which could increase sensitivity for CRC and its high-risk precursors, without jeopardizing specificity, we then assessed which proteins would be complementary and could potentially outperform hemoglobin in discriminating CRC from control samples. Both logistic regression and classification and regression tree (CART) analysis were used to select biomarker panels. Logistic regression optimized AUC in ROC analysis, while CART produced a decision tree. We used both methods because, on the one hand, the standard method allows us to select multivariable panels, is easy to interpret, and provides a linear predictor where you can easily determine a cut-off to optimize sensitivity and specificity. CART, on the other hand, is admittedly less common and offers less flexibility in adjusting cut-offs, but it is a natural way to incorporate interactions between biomarkers within a decision tree structure. This tree is easy to interpret as it provides insight into combined biomarker ranges where the probability of AN is high versus low. This dual approach ensured robust, consistent findings, highlighted overlapping and unique biomarkers, and provided a comprehensive list for validation.

Several combinations of four proteins outperformed hemoglobin alone in discriminating between CRC and control samples at a fixed high specificity of 95%, essential in screening. While these analyses yielded multiple panels of proteins, there was significant overlap in the proteins selected, suggesting their potential as biomarkers. The optimal panel size appeared to be three proteins and bigger panels did not further increase performance.

The sensitivity of hemoglobin for detecting CRC at 95% specificity, as measured by mass spectrometry in this study, was 43%. In contrast, protein panels identified in the study reached sensitivities of 80% for detecting CRC and 45% for detecting advanced adenomas at that 95% specificity. The higher sensitivities achieved by the protein panels compared to hemoglobin in this study confirmed that these proteins, especially when combined with hemoglobin, could possibly improve CRC screening.

Initial Marker Validation

Our first goal was to develop an assay which was compatible with population-based CRC screening program logistics. This implied changing from whole stool samples to small stool samples that are compatible with transport via a mailbox. Moreover, we moved from mass spectrometry-based measurements to, an antibody-based test, like FIT. These changes also decreased the cost per test. Therefore, we conducted a validation study on an independent series of FIT samples using an antibody-based assay to determine if the candidate biomarkers identified by mass spectrometry could also be quantified in small stool sample volumes. In this study, 72FIT samples were used. These samples came from a colonoscopy-controlled symptomatic population and contained 14 patients with CRC, 16 patients with advanced adenoma, 18 patients with non-advanced adenomas and 24 individuals without colorectal neoplasia (control samples) [5].

We used an off-the-shelf antibody-based assay for four proteins: A2M, MPO, RBP4, and adiponectin. It is important to note that the choice for these four proteins was limited by antibody availability rather than focusing on the best-performing biomarker panels.

Of the four off-the-shelf assays, A2M, MPO, and adiponectin were present at significantly greater concentrations in the FIT samples from patients with CRC or advanced neoplasia than in control samples (P < 0.001 and P < 0.01, respectively). This validation experiment demonstrated that the results from the mass spectrometry-based analysis could be validated in an independent series of FIT samples using antibody-based tests [5]

Initial Assay Development (Research Use Only)

The next step in our development path was to perform a large-scale validation of best-performing antibody panels using biobanked FIT samples with known colonoscopy outcome [11]. To this end, we developed antibody-based assays for the top candidate biomarkers proteins previously identified [11]. The goal was to develop a multitarget FIT (mtFIT) using antibody-based assays that would be compatible with the current practice of FIT in the population-based screening program. First, we selected ten proteins out of the initial 29. This selection was based on complementarity, performance in multivariate analysis and biologic considerations. The ten protein biomarkers selected were as follows: a-2-macroglobulin, calprotectin, C3 complement, hemoglobin, haptoglobin, hemopexin, lactotransferrin, myeloperoxidase, retinol-binding protein 4, and serpin family F member 2.

The research-use-only (RUO) antibody-based assays were developed using Meso-Scale Discovery (MSD) technology (MSD, Rockville, MD, USA) since this allowed for multiplex protein biomarker detection and quantification, making it possible to measure multiple biomarkers in a single assay. Due to technical reasons, assay development was only successful for nine proteins, leaving out retinol-binding protein 4.

Large-Scale Marker Validation

Using the antibody-based assays we next performed a large-scale validation of best-performing antibody panels using biobanked left-over FIT samples. These were left-over samples from two previous Dutch study populations, who collected their samples in OC-Sensor collection devices (Eiken Chemical). Samples were stored at − 80 °C and hemoglobin levels were measured using the OC-Sensor DIANA analyzer (Eiken Chemical). After analysis, samples were re-stored at − 80 °C. The first series consisted of samples from 1038 participants in the primary colonoscopy arm of a screening trial comparing primary colonoscopy vs CT colonography (COlonoscopy or COlonography for Screening; COCOS) [8, 12]. These pre-colonoscopy samples were collected in 2009–2010. The second series included 246 participants from a colonoscopy-controlled symptomatic population, with samples collected between 2003 and 2014 before colonoscopy [13], except for nine CRC samples collected at least two weeks post-colonoscopy but before surgery. Ultimately, a series of 1284 FIT samples (control n = 769, non-advanced adenoma n = 250, advanced adenoma n = 135, non-advanced serrated polyp n = 53, advanced serrated polyp n = 30, and CRC n = 47) were analyzed. All 1284 participants had provided a single FIT sample that was used for both FIT and testing for the new protein markers.

All nine protein biomarkers demonstrated significantly higher concentrations (P < 0.001) in samples from individuals with colorectal cancer (CRC) compared to those without colorectal neoplasia. Eight out of the nine protein biomarkers exhibited significantly higher concentrations in samples from individuals with advanced adenoma (AA) compared to those without colorectal neoplasia. Samples from individuals with advanced serrated polyps (ASP) showed protein biomarker concentrations similar to those from individuals without colorectal neoplasia [11].

Next, because of superior predictive performance, de Classification and Regression Tree (CART) analysis was applied to identify the optimal combination of three protein biomarkers to have the best diagnostic performance for detecting advanced neoplasia (AN), yielding hemoglobin, calprotectin, and serpin family F member 2 (serpinF2). This combination was then named multitarget FIT (mtFIT). At an equal specificity of 96.6%, the cross-validated sensitivity for AN of mtFIT was 42.9% vs 37.3% for FIT. The observed increased AN sensitivity was due to an increase in sensitivity for AA (37.8% for mtFIT versus 28.1% for FIT), as the sensitivity for CRC was not significantly different (78.7% for mtFIT versus 80.9% for FIT), and the sensitivity for ASP were equal for both tests (10.0%). Again, we confirmed meaningful differences in protein levels between cases and controls using antibody-based assays in FIT samples [11].

Early Health Technology Assessment

Since the goal of developing a new test is to have a method holding potential for real-world application, a health technology assessment (HTA) needs to be performed at this early stage. This is not only important to evaluate the potential clinical effectiveness, but also to obtain an estimate of the potential cost-effectiveness of the new test compared to FIT, as this would be critical for justifying further clinical development. Such analysis was performed for mtFIT to evaluate its impact and cost-effectiveness.

This early HTA, using the well-established and validated ASCCA (Adenoma and Serrated pathway to Colorectal Cancer) microsimulation model [14], involved a comparison of the projected long-term health outcomes and cost-effectiveness of mtFIT versus FIT in the context of the current Dutch FIT-based national screening program. Based on the cross-validated study data, mtFIT-based screening, compared to FIT-based screening, was projected to result in a further 12% CRC incidence reduction and an 8% CRC mortality reduction. The maximum cost per test, at which mtFIT-based screening would still be cost-effective compared to FIT-based screening, was estimated at 59 euro. Based on these results, we concluded that there was a solid basis to proceed with the development of the mtFIT and a large-scale, prospective screening trial was designed to further validate the test’s performance in a real-world setting.

Assay Production for the Clinical Trial

In preparation for the clinical trial, production of a large batch of mtFIT assays for hemoglobin, calprotectin, and serpin family F member 2 (serpinF2) was assigned to Meso-Scale Discovery (MSD, Rockville, MD, USA).

Due to the large quantity, i.e., 15,000, of assays custom-made lots of the antibodies were prepared specifically for this study and a bridging analysis was needed to verify reliability and consistency with the earlier results. The bridging analysis showed that the performance of the new lot of the SULFO-TAG-conjugated (detection) antibodies for SerpinF2 and hemoglobin was comparable to the performance of the original lot of detection antibodies, while the new lot of the calprotectin detection antibody showed good performance but observed lower protein signals. Therefore, we adapted the calprotectin cut-off from 251 µg/g feces to 59.45 µg/g feces [15].

Clinical Trial

To prospectively validate the performance of mtFIT in a real-world screening setting, a screening trial (ClinicalTrials.gov, NCT05314309) was designed [16]. The primary outcome of the trial was the relative detection rate of AN for mtFIT versus FIT. Secondary outcomes were relative detection rates of mtFIT versus FIT for CRC, advanced adenoma, and advanced serrated polyp individually as well as the predicted long-term effects on CRC incidence, mortality, and costs for programmatic mtFIT screening versus FIT-based screening.

Trial participants were recruited through the national Dutch FIT-based CRC screening program, leveraging its infrastructure, logistics and collaboration with the organizations managing its daily operations. This has turned out to be a crucial facilitator for this trial. Invitations to participate were sent out to individuals aged 55–75 years eligible for the screening program. A representative sample of 35,786 individuals from the South-West region of the Netherlands, encompassing both rural and urban areas, were invited. Of these, 15,283 (42.7%) consented to participate. Data from 13,187 participants were available for analysis, representing 86.3% of those who initially consented to participate.

All participants received two collection tubes, one for FIT and one for mtFIT, along with instructions to collect samples from the same bowel movement. FIT samples were analyzed using a fully automated clinical chemistry analyzer, following standard procedures of the Dutch national CRC screening program. A cut-off of 47 µg hemoglobin (Hb) per gram of feces, according to the standard in the Dutch screening program, was used to classify FIT results as positive or negative. Unlike previous prospective studies in the Dutch CRC screening program (e.g., comparing FOBgold to OC-sensor), the ministry of Health, Welfare and Sport following advice from the Dutch Health Council this time did not allow us to apply a lower cut-off of 15 µg hemoglobin (Hb) per gram of feces, which would have allowed for a direct comparison of mtFIT and FIT at equally high positivity rates (also see below). mtFIT samples were analyzed with the custom-made RUO multiplex antibody-based test to measure hemoglobin, calprotectin, and serpinF2. Duplicate measurements were conducted with strict quality controls, and the CART-based decision tree algorithm was applied to determine positivity or negativity of the mtFIT results.

Participants with positive results from either FIT, mtFIT, or both were referred for colonoscopy. In total 538 individuals (4.1%) tested positive for FIT and 1201 individuals (9.1%) tested positive for mtFIT. To account for these different positivity rates between the tests, analysis of the trial results was performed considering three scenarios.

In scenario 1, the performance of mtFIT and FIT was evaluated on the basis of outcomes directly observed in the study, implying that the positivity rate of mtFIT was expected to be higher than the positivity rate of FIT, inherent to the imposed 47 µg Hb/g feces FIT cut-off.

In scenario 2, the mtFIT and FIT were compared at an equally low positivity rate, which was the positivity rate observed for FIT using the 47 µg Hb/g feces cut-off in the study. Researchers adjusted the cut-off levels in the mtFIT algorithm to achieve the same low positivity rate as FIT. This comparison was important to determine whether mtFIT could achieve a higher detection rate for advanced neoplasia while maintaining a similar number of colonoscopy referrals compared to FIT.

In scenario 3 mtFIT and FIT were compared at an equally high positivity rate, equivalent to the positivity rate observed for mtFIT with its pre-established cut-offs. Because the study could not lower the FIT cut-off below 47 µg Hb/g feces, data from previous screening studies was used to impute the detection rates for FIT at a higher positivity rate [12, 17]. This comparison was conducted to investigate if mtFIT maintained its superior detection rate for advanced neoplasia compared to FIT even at higher positivity rates.

The Results were as follows:

Scenario 1: Based on observed outcomes in the study, mtFIT detected nearly twice the number of participants with advanced neoplasia compared to FIT (299 vs. 159 individuals, or detection rates of 2.27% vs. 1.21%). This means that for every 1000 people screened, mtFIT would identify 10.6 more cases of advanced neoplasia than FIT. The positive predictive value (PPV) for advanced neoplasia was 24.9% and 29.6% for mtFIT and FIT, respectively (Table 1). MtFIT also showed higher detection rates for colorectal cancer, advanced adenomas, and advanced serrated polyps individually. This scenario highlights the potential for meaningful improvements in detection when using mtFIT compared to the current Dutch screening program using FIT with a 47 µg Hb/g feces cut-off.

Table 1.

Results of the prospective screening trial

Scenario 2: Equally low positivity rates (both tests at 3.6%). At the same low positivity rate, mtFIT identified advanced neoplasia in 178 participants, while FIT identified AN in 159 participants. This translates to detection rates of 1.36% for mtFIT and 1.22% for FIT. The relative detection rate of mtFIT compared to FIT was 1.12, which means mtFIT detected 12% more cases of advanced neoplasia. The PPV for advanced neoplasia was 37.6% for mtFIT and 33.8% for FIT (Table 1). Importantly, mtFIT showed a statistically significant 17% increase in the detection rate of advanced adenomas.

Scenario 3: Equally high positivity rates (both tests at 8.4%): Due to the limitations of the study design, a direct comparison at the same high positivity rate was not possible (more on this in discussion paragraph). Therefore, data from previous CRC screening studies were used to impute the performance of FIT at a higher positivity rate.

With the imputed data, this analysis indicated a 15% increase in the detection rate of advanced neoplasia (including colorectal cancer and advanced adenoma only) with mtFIT compared to FIT. The PPV for advanced neoplasia was 22.1% and 19.9% for mtFIT and FIT, respectively (Table 1). Specifically for advanced adenomas, mtFIT demonstrated a 19% increase in the detection rate. These results highlight that also at a higher positivity rate for both tests, mtFIT still offers an advantage in detecting advanced lesions, particularly advanced adenomas compared to FIT.

The overall findings consistently showed better AN detection rates for mtFIT than the standard FIT. This improved detection, particularly for advanced adenomas, has the potential to translate into meaningful reductions in colorectal cancer incidence and mortality over time.

Health Technology Assessment

To evaluate the long-term impact and cost-effectiveness of using mtFIT compared to the standard FIT in CRC screening, the results of the trial were also submitted for a health technology assessment. To estimate long-term effects, the well-established and validated ASCCA (Adenoma and Serrated pathway to Colorectal Cancer) microsimulation model was employed [14]. Data from the mtFIT trial were incorporated into the model and used to project the long-term impact of an mtFIT-based screening program on CRC incidence, CRC mortality, quality-adjusted life years lived, and costs compared to FIT screening. Given that cost of a commercial mtFIT test is currently unknown, the aim of this analysis was to determine the maximum mtFIT test costs at which an mtFIT-based screening program remained cost-effective compared with FIT-based screening.

Results were predicted for scenarios one and two, but not for scenario three due to missing data on serrated lesions in previous screening studies, which prevented imputing predicted detection rates (Table 1). Since the ASCCA model incorporates the serrated pathway, reliable results could not be produced without this input.

In the first scenario with the different positivity rates for mtFIT and FIT, the modeling predicted that mtFIT-based screening could lead to an extra 21% reduction in CRC incidence and an extra 18% reduction in mortality compared to FIT screening. For this scenario, the model predicted that at mtFIT costs of up to €148 mtFIT-based screening remains cost-effective compared with FIT­based screening, as the cost-effectiveness ratio remains below the WTP threshold.

In scenario 2, at equally low positivity rates, screening with mtFIT compared with FIT was predicted to reduce colorectal cancer incidence by 5% and CRC mortality by 4%. In this analysis, the maximum cost of the mtFIT test was estimated to be €57 to ensure an incremental cost-effectiveness ratio of mtFIT screening compared with FIT screening below the willingness­to­pay threshold.

While full determination of the cost of mtFIT awaits further details of a final clinical-grade mtFIT assay, the identified cost thresholds appear achievable for antibody-based assays in general. Comparisons with existing programs and anticipated cost reductions through technological advancements support the feasibility of these thresholds. These findings reinforce the potential of mtFIT as a practical and impactful tool for colorectal cancer screening.

Clinical-Grade Assay Development and Further Steps Toward Implementation

A prospective longitudinal screening trial is planned to further validate the findings of the initial trial and confirm the long-term benefits of mtFIT screening, using the clinical-grade assay. This trial will be crucial for providing robust evidence of mtFIT’s effectiveness in a real-world screening setting and addressing the limitations of the previous study, particularly the inability to directly compare mtFIT and FIT at the same high positivity rate.

The results of this trial will again be used in a final analysis for long-term impact and cost-effectiveness of an mtFIT-based screening approach. This will provide the remaining essential evidence for mtFIT-based screening the program and is crucial to approval inform policy decisions regarding its adoption.

Discussion

The development of the multitargetFIT (mtFIT) has successfully progressed through key steps toward achieving improved accuracy in detecting AN compared to the standard FIT used in CRC screening.

Proteomics analysis using mass spectrometry was employed to examine stool samples from individuals with and without CRC, leading to the identification of a panel of protein biomarkers that could complement or outperform hemoglobin in detecting AN. Based on these findings, an antibody-based research-use-only mtFIT was developed to measure a combination of three proteins: hemoglobin, calprotectin, and serpinF2. Clinical validation studies in the Netherlands population, including a large-scale population-based screening trial, demonstrated that this test improved AN detection.

Model-based health technology assessment incorporating the trials’ findings further predicted that mtFIT-based screening could be cost-effective compared to FIT-based screening, reducing CRC incidence and mortality at acceptable incremental costs. These results provide evidence that mtFIT testing can bring a significant advancement in CRC screening, offering enhanced sensitivity for detecting precancerous lesions and the potential to lower the disease burden further.

However, in this process we have encountered several challenges, some of which are highlighted below.

Assay Development Is Not a Linear Process

One of the challenges we encountered in test development is the significant cost and complexity involved in creating a diagnostic test. As researchers, we chose to pursue a high-quality test that we could outsource due to our limited resources for developing it ourselves (academic setting) in a standardized and reproducible way. This approach resulted in two rounds of development, one for the research-use-only test and a future one for the clinical-grade test. The clinical-grade test will be conducted on a different platform to ensure it is automated guarantying high-throughput performance, essential in programmatic population screening. Ideally, we would have used the same platform from the outset, but funding and capacity constraints prevented it. This relates to the well-known ‘valley of death’ in the translation of research discoveries into clinical applications, particularly in the development of biomarker tests. Many promising biomarker candidates often fail to make the transition from early-stage research (the “bench”) to validated, clinically useful tests (the “bedside”) [18].

Trial Design Suboptimal Due to Restrictions by the Dutch Government

Upon advice of the Health Council, the Ministry of Health and Welfare and Sport of the Netherlands imposed a significant constraint on the mtFIT trial by mandating a FIT cut-off of 47 µg Hb/g feces instead of the proposed 15 µg Hb/g feces. This decision, while stemming from regulatory considerations, had substantial implications for the study’s design and the interpretation of its results. It made a direct comparison of both tests at different thresholds, ensuring equal specificity at different positivity rates, impossible. The decision was driven by the choice to adhere to established protocols within the Dutch national CRC screening program, as well as to avoid additional colonoscopies. However, this led to an opposite effect. A new calculation of the sample size resulted in an increase from 10,000 participants to 13,000 participants and the number of colonoscopies from 900 to 960.

Despite a formal appeal procedure, this restriction stayed in place, not only restricting the interpretation of the results of the trial but also introducing new challenges in financing and timespan of the ongoing work.

Sustainable Funding for a Biomarker Development Trajectory

Clearly research funding is an important facilitator to perform long-term research projects. It is however often subject to changes and long-term funding stability can be difficult to secure, as grants and funding sources may fluctuate or be discontinued. Here we managed to secure funding through multiple rounds of grants, and we were fortunate to be successful for over 10 years to continue this research (see Fig. 2). However, this was a challenge, since we had to deal with budget changes (inflation, prolonged timelines, increase in study inclusions, COVID-19 period, etc.). Obtaining funding has been a continuous challenge, full details of which are beyond the scope of the present paper.

Fig. 2.

Facilitating factors for development of a new biomarker test

Biobanks (Blue Arrow): Represents repositories of biologic samples used to support research. Includes collection, storage, and management of samples with associated data. Infrastructure & Financing (Yellow Arrow): Signifies the systems and funding mechanisms needed to support research and development activities. Engaged Stakeholders (Orange Arrow): Highlights the involvement of individuals or groups with a vested interest in the research outcomes, fostering collaboration, and alignment of goals. Team Science (Green Arrow): Represents a collaborative approach that integrates expertise from multiple disciplines to solve complex scientific problems.

Besides the normal challenges in obtaining funding, we experienced that getting nearer to a product for clinical implementation resulted in less opportunities for obtaining (charity-driven) grants. While it is reasonable to expect industry to finance these final steps, the diagnostic market differs significantly from the pharmaceutical market. Despite guiding 70–80% of medical decisions, diagnostics are undervalued, seen as enablers rather than primary drivers of health outcomes. This perception results in less focus on innovation funding. Additionally, the diagnostic market is viewed as lower risk and requiring smaller investments than pharma, but it also offers lower returns and faces more challenging reimbursement pathways.

A step we took to mitigate this challenge was to establish a spin-off company specifically for the further development of mtFIT (i.e., CRCbioscreen).

Next to these challenges several facilitators play a crucial role in advancing the progress of the mtFIT development by addressing key areas essential for success (see Fig. 2). The biobank provided a centralized repository of high-quality biologic samples and associated data, ensuring that there was access to standardized and well-characterized materials. This accelerated the initial validation studies. There were many stakeholders that were engaged and contributed toward the success of this work. Their input on the study design, will make it more likely that these findings will be adopted in clinical practice. Finally, this work is a clear example of team science in which interdisciplinary and multi-center collaborations, brought together diverse expertise’s to tackle the challenges we encountered and to drive this research project toward meaningful results.

In summary, in this perspective we have shown a successful example of the development of a new test for screening until clinical validation in real-world setting, following the steps of the biomarker translational path with different evaluating moments as recommended by the guiding principles of new non-invasive screening tests.

Biographies

Meike de Wit

Evelien Dekker

Manon Spaander

Monique van Leerdam

Veerle Coupe

Gerrit Meijer

Beatriz Carvalho

Author Contributions

Meike de Wit wrote the manuscript with input from all authors. All authors critically reviewed the final version of the manuscript.

Data Availability

No datasets were generated or analyzed during the current study.

Declarations

Competing interests

MdW is co-founder, stockholder and board member (COO) of CRCbioscreen BV. BC, MdW, VC and GM have several patents pending and/or issued. GM is co-founder, stockholder and board member (CSO) of CRCbioscreen BV, he has a research collaboration with CZ Health Insurances (cash matching to ZonMW grant) and he has research collaborations with Exact Sciences, Sysmex, Sentinel Ch. SpA, Personal Genome Diagnostics (PGDX), DELFi Diagnostics and Hartwig Medical Foundation; these companies provide materials, equipment and/or sample/genomic analyses. ED has endoscopic equipment on loan of FujiFilm, receive a research grant from FujiFilm. Has received honorarium for consultancy from FujiFilm, Olympus, InterVenn, and Ambu, and speakers' fees from Olympus, GI Supply, Norgine, IPSEN, PAION and FujiFilm. ED is also Chair CRC Screening Committee of World Endoscopy Organisation, Chair Dutch Postpolypectomy surveillance Guideline, Member Post-polypectomy surveillance Guideline of ESGE. MS has received research support from Medtronic, Boston Scientific, Sentinel and Sysmex.