Can a biomarker triage test reduce colonoscopy burden in fecal immunochemical test screening?

For all analyses, we used the Adenoma and Serrated pathway to Colorectal CAncer model which is extensively described elsewhere [8]. In short, the model simulates individuals from age 20 to age 90 or death, whichever comes first. During their life, individuals have the probability of developing up to ten adenomas and/or ten serrated polyps which have certain characteristics such as location, dysplasia and histology. Progression in terms of growth in size (diminutive <6 mm, small 6–9 mm and large ≥10 mm) and development of malignant characteristics is modeled independently for each lesion. When a lesion has progressed to advanced adenoma or arises as a sessile-serrated polyp, it can progress to CRC. We assumed that 85% of CRCs arises via the adenoma-carcinoma pathway and 15% via the serrated pathway [9]. Each year, a tumor may progress to a more advanced stage or may become detected. The natural history model satisfactorily reproduces Dutch colorectal lesion prevalence rates and Dutch CRC incidence and mortality rates in the absence of screening [10,11]. The model is externally validated by replicating several long-term CRC screening trials [12].

As a reference, we simulated the Dutch national CRC screening program. In this strategy, individuals aged 55–75 are biennially invited for FIT-based screening. FIT-positive individuals (≥47 μg hemoglobin per g feces) are referred for diagnostic colonoscopy, that is colonoscopy after a positive screening test, during which all detected lesions are removed with the exception of hyperplastic polyps in the rectosigmoid. Dependent on colonoscopy findings, individuals may be referred to surveillance colonoscopy according to the Dutch surveillance guideline or return to the screening program after 10 years [13].

Adherence to FIT-based screening and colonoscopy was set at 73 and 92% respectively, based on achieved participation rates in the Dutch CRC screening program in 2018 [2]. Test characteristics of the FIT with a cutoff of 47 μg/g (FIT₄₇) were calibrated against the positivity rate and detection rate as reported by two Dutch trials evaluating FIT-based screening [8]. Colonoscopy detection rates for adenomas were based on a systematic review [14]. As the pale color, proximal location and flat appearance hamper visual detection of serrated polyps [15] we assumed a 10% lower colonoscopy detection rate for serrated polyps than for adenomas. Table 1 shows the test characteristics used in the model.

Besides a FIT-based screening strategy, we simulated a FIT plus BM-TT screening strategy in which we assumed that the BM-TT will only be performed in FIT-positive individuals. Thus, the BM-TT can only detect lesions in individuals with fecal hemoglobin levels that are above the cutoff. Furthermore, we assumed that the test can be applied on the same sample as the FIT. This ensures that screening participation is not affected as individuals do not have to undergo additional testing. For the FIT plus BM-TT screening strategy, we first evaluated 24 FIT₄₇ plus BM-TT scenarios differing in sensitivity and specificity of the BM-TT (Table 1). Specificity was set either at 90, 95, 97% or 100%. Sensitivity for late stage CRC varied between 80 and 100%, with 4% increments. Subsequently, we calculated the ratio with which the sensitivity for late stage CRC was changed compared with that of FIT₄₇. For example, when the sensitivity of late stage CRC was set at 92%, the relative increase in sensitivity compared with FIT₄₇ is 8% (92 divided by 85%). Using FIT₄₇ test characteristics as a reference point, sensitivity for early stage CRC, large adenomas and small adenomas was assumed to change with the same ratio as the sensitivity for late stage CRC. Sensitivity for diminutive adenomas and small and large serrated polyps was set at 1 minus the specificity. These 24 scenarios are considered the FIT₄₇ plus low polyp sensitivity BM-TT scenarios.

Furthermore, we simulated 28 FIT₄₇ plus high polyp sensitivity BM-TT scenarios in which the sensitivity for diminutive, small and large polyps of the BM-TT was substantially increased compared with the BM-TT with low polyp sensitivity. In these scenarios, we assumed the sensitivity of the BM-TT for diminutive, small and large polyps and early stage cancer to be respectively 20, 15, 10 and 5% lower (absolute difference) than that for late stage CRC. We also included FIT₄₇ plus BM-TT scenarios in which we assumed that the BM-TT has perfect sensitivity for all polyps and CRC.

We repeated these 52 scenarios assuming a lower FIT cutoff, that is 15 μg/g. Lowering the FIT cutoff leads to a less stringent initial selection of individuals. This may increase the added value of a BM-TT if sensitivity is sufficiently high because less advanced neoplasia is missed by the primary screening instrument.

To account for the burden of screening and CRC treatment, we calculated quality-adjusted lifeyears (QALYs). For healthy individuals, we assumed a utility of 0.84 [16]. As stool-based testing is considered as noninvasive, we assumed that this test does not lead to utility losses. Furthermore, the study of Niv et al. [17] showed that colonoscopy did not affect quality of life. Therefore, we assumed no utility loss due to colonoscopy in the base-case analysis. For complications due to colonoscopy, we assumed a disutility of 0.34 during 2 weeks [18]. Table 2 shows disutilities related to screening, surveillance and CRC treatment.

Costs were determined from a healthcare payer perspective and included costs related to screening, surveillance and CRC treatment (Table 2). Costs were converted to 2019 Euros using the consumer price index for that year [24]. As costs of the BM-TT are unknown, we varied the costs using the costs of colonoscopy as a reference since the purpose of the BM-TT is to reduce the number of unnecessary colonoscopies. That is, we set the costs of the BM-TT to 10, 25, 50, 100, 150 and 200% of the costs of colonoscopy.

For each scenario, we simulated a cohort of 20,000,000 individuals and calculated the number of CRC cases and deaths, colonoscopy burden, lifeyears, QALYs, costs and net monetary benefit. Colonoscopy burden was indicated by the number of colonoscopies (both diagnostic and surveillance) as well as the number of negative diagnostic colonoscopies for each scenario. Net monetary benefit (NMB) was calculated as (total QALYs x threshold) – total cost, using the Dutch gross domestic product per capita (€35,916) as threshold [24]. We applied a 3% discount rate to the effects and costs from age 50 onwards [25].

All FIT plus BM-TT scenarios were compared with the reference strategy consisting of FIT₄₇-based screening with respect to CRC incidence and mortality rates, colonoscopy burden, lifeyears and QALYs. Furthermore, we calculated incremental QALYs per averted colonoscopy to simultaneously consider both screening effectiveness and colonoscopy burden as well as the incremental NMB to gain insight in the cost–effectiveness of FIT plus BM-TT scenarios compared with FIT-screening only.

In one-way sensitivity analyses, we set the utility loss due to colonoscopy at 0.34 during 2 days, based on previous cost–effectiveness analyses [18]. Furthermore, we evaluated scenarios in which we assumed that the high polyp sensitivity BM-TT only has high sensitivity for large and small adenomas and large serrated polyps (Table 1). Thus, sensitivity for diminutive adenomas and small serrated polyps was not increased.

Table 3 shows the summary results for the base-case analysis whereas a more detailed overview is provided in the Supplementary Material. When a BM-TT with low polyp sensitivity was added to FIT₄₇-screening, the number of required colonoscopies decreased to 150–228 in the lifetime of a cohort of 1000 individuals (58–72% decrease compared with FIT₄₇-screening only). The number of diagnostic colonoscopies in which no polyps were detected ranged between 0 and 11 (89–100% decrease). This decrease in colonoscopy burden was slightly lower for the FIT₁₅ plus low polyp sensitivity BM-TT scenarios.

For the FIT₄₇ plus high polyp sensitivity BM-TT scenarios, the number of lifetime colonoscopies per 1000 individuals varied between 416 and 467 (14–23% decrease) whereas the number of negative diagnostic colonoscopies ranged between 11 and 25 (75–89% decrease). The FIT₁₅ plus high polyp sensitivity BM-TT scenarios led to slightly smaller decreases in colonoscopy burden.

The decrease in colonoscopy burden came at the cost of a decrease in screening effectiveness for all FIT₄₇ plus low polyp sensitivity BM-TT scenarios and the majority of the FIT₄₇ plus high polyp sensitivity BM-TT scenarios (Table 3 & Supplementary Material). Depending on the test characteristics of the low polyp sensitivity BM-TT that was added to FIT₄₇-screening, the number of lifetime CRC deaths ranged between 17.1 and 19.1 in a cohort of 1000 individuals. This equals to a 27–41% increase in CRC deaths compared with FIT₄₇-screening only. FIT₄₇-screening led to 16,330 QALYs in a cohort of 1000 individuals. This decreased to 16,304–16,313 QALYs when a BM-TT with low polyp sensitivity was added to FIT₄₇-screening.

For the FIT₄₇ plus high polyp sensitivity BM-TT scenarios, the increase in lifetime CRC deaths was 0–4.4% whereas QALYs decreased with 0–3 per 1000 individuals. Thus, adding a BM-TT with high polyp sensitivity to FIT₄₇-screening led to a smaller reduction in colonoscopy burden than adding a BM-TT with low polyp sensitivity. On the other hand, screening effectiveness only slightly decreased. In fact, when the BM-TT had perfect sensitivity for polyps and CRC, screening effectiveness did not decrease.

Also for the FIT₁₅ plus low sensitivity BM-TT scenarios, screening effectiveness decreased compared with FIT₄₇-screening only. In contrast, the FIT₁₅ plus high polyp sensitivity BM-TT scenarios led to a reduction in colonoscopy burden as well as an increase in screening effectiveness compared with FIT₄₇-screening only. Depending on BM-TT test characteristics, CRC mortality decreased with 2.2–6.6% whereas QALYs increased with 1–5 per 1000 individuals.

To simultaneously consider colonoscopy burden and screening effectiveness, we calculated incremental QALYs per averted colonoscopy. For the FIT₄₇ plus low polyp sensitivity BM-TT scenarios, each averted colonoscopy led to a 0.055–0.066 QALY decrease in screening effectiveness. For the FIT₄₇ plus high polyp sensitivity BM-TT scenarios, incremental QALYs per averted colonoscopy ranged between -0.024 and 0.002 QALYs compared with FIT₄₇-screening only. For the FIT₁₅ plus low polyp sensitivity BM-TT, incremental QALYs per averted colonoscopy were comparable to the FIT₄₇ plus low polyp sensitivity BM-TT. In contrast, incremental QALYs per averted colonoscopy were 0.023–0.644 QALYs for the FIT₁₅ plus high polyp sensitivity BM-TT. Note that this reflects an increase in screening effectiveness per averted colonoscopy.

Figure 1 shows the incremental NMB for the FIT plus triage test scenarios compared with FIT₄₇-screening only for different costs of the BM-TT. When the costs for the BM-TT are varied between 77 and 1543 Euros, that is between 10 and 200% the costs of colonoscopy, all FIT₁₅ plus high polyp sensitivity BM-TT scenarios and some of the FIT₄₇ plus high polyp sensitivity BM-TT scenarios led to a positive incremental NMB, indicating that these scenarios are considered cost effective at a willingness-to-pay threshold of €35,916. All scenarios in which a low-polyp sensitivity BM-TT is added to FIT-screening led to a negative incremental NMB, indicating that these scenarios are not cost effective compared with FIT₄₇-screening only.

The Supplementary Material shows the results of the sensitivity analyses. Repeating the base-case analysis assuming that colonoscopy led to a utility loss did not impact the discounted QALYs per 1000 individuals. The FIT plus high polyp sensitivity BM-TT scenarios in which the BM-TT only had high sensitivity for small and large adenomas and large serrated polyps led to a greater reduction in colonoscopy burden compared with the FIT plus high polyp sensitivity BM-TT scenarios in which the BM-TT has high sensitivity for polyps of all sizes. On the other hand, screening effectiveness was more decreased. When comparing these scenarios in terms of incremental QALYs per averted colonoscopy, figures were more favourable for the FIT plus high polyp sensitivity scenarios in which the BM-TT had high sensitivity for polyps of all sizes.