Dottorato in MALATTIE INFETTIVE, MICROBIOLOGIA E SANITA' PUBBLICA

Titolo della tesi: The role of genomic biomarkers and simulation models in cancer prevention and early cancer detection. Case studies on colorectal and lung cancers.

This thesis has both a clinical and a methodological purpose. The first purpose is to show the critical impact of genomic biomarkers in cancer prevention nowadays, which is not related only to the Mendelian genetics (e.g., BRCA variants or Lynch syndrome). The second is to exemplify when the discrete event simulation model is the appropriate epidemiological methodology to answer a research question in cancer prevention. These purposes are satisfied through the detailed analyses of three case studies, summarized below. Identification of biomarkers involved in multiple pathways of cancer growth can facilitate the rational development of prevention strategies. Potential uses of biomarkers in public health include, among others: early detection of asymptomatic patients (e.g., colorectal cancer screening – first case study); aiding in the diagnosis of patients at higher risk of cancer (e.g., patients with indeterminate pulmonary nodules – second case study); surveillance of individuals known to be at high risk of cancer (e.g., carriers of cancer-related variants such as BRCA1/2), and surrogate endpoint markers for primary prevention strategies such as chemoprevention or other therapeutic strategies (e.g., cetuximab in patients with mutated K-ras colorectal cancer – third case study). In several cases, analytic modeling methodologies, such as discrete event simulation models, have an essential role in generating clinical evidence for cancer prevention, especially when other study designs (i.e., clinical trials, cohort and case-control studies) are not feasible. The case studies explored in depth when the discrete event simulation model is the appropriate epidemiological methodology (i) to generate evidence on biomarker-related strategies for cancer prevention (first and second case studies), and (ii) to guide the correct planning of clinical trial designs for biomarker-driven therapies (third case study). In the first case study, we analyzed the application of a blood test that uses circulating, cell-free methylated Septin9 (mSEPT9) DNA in the screening of colorectal cancer. Current recommendations of The US Preventive Services Task Force (USPSTF) on colorectal cancer (CRC) screening strategies are based on models that assume 100% adherence. Since adherence can have a large effect on screening outcomes, we aimed to compare the effectiveness of CRC screening strategies under reported adherence rates at the population level. We developed and validated a microsimulation model to assess the effectiveness of colonoscopy (COL), flexible sigmoidoscopy (FS), high-sensitivity guaiac fecal occult blood-test (HS-gFOBT), fecal immunochemical test (FIT), multitarget stool DNA test (FIT-DNA), computed tomography colonography (CTC), and methylated SEPT9 DNA test (SEPT9) in terms of CRC incidence and mortality, incremental life-years gained (LYG), number of colonoscopies, and adverse events for men and women 50 years or older over their lifetime. We assessed outcomes under 100% adherence rates and reported adherence rates. We also performed sensitivity analyses to evaluate the impact of varying adherence levels on CRC outcomes. Assuming 100% adherence, FIT-DNA, FIT, HSgFOBT, and SEPT9 averted 42-45 CRC cases and 25-26 CRC deaths, COL 46 cases and 26 deaths, CTC 39 cases and 23 deaths, FS 32 cases and 19 deaths per 1,000 individuals. Assuming reported adherence, SEPT9 averted 37 CRC cases and 23 CRC deaths, COL 34 cases and 20 deaths, FIT-DNA, FIT, CTC and HS-gFOBT 16-25 cases and 10-16 deaths per 1000 individuals. LYG reflected the effectiveness of each strategy in reducing CRC cases and deaths. Adverse events were more common for COL (3.7 per 1000 screened) and annual SEPT9 (3.4 per 1000 screened), and proportional to the number of colonoscopies. Among the screening strategies recommended by USPSTF, colonoscopy results in the largest benefit when we account for adherence. Adherence rates higher than 65-70% would be required for any stool or blood-based screening modality to match the benefits of colonoscopy. Nowadays, we have evidence that blood-based screening modality can reach that adherence rate. In the second case study we explored the use of a bronchial gene-expression classifier (BGC) that combines information from 24 genomic biomarkers and several risk factors (i.e., sex, age and smoking). Nowadays, bronchoscopy is the safest procedure for lung cancer diagnosis when an invasive evaluation is required after a suspicious result at an imaging procedure. However, its sensitivity is relatively low, especially for small and peripheral lesions. We assessed benefits and costs of introducing this BGC to improve the performance of bronchoscopy and the overall diagnostic process for early detection of lung cancer. We used a discrete-event simulation to compare clinical and economic outcomes of two different strategies with the standard practice in former and current smokers with indeterminate nodules: (i) location-based strategy - integrated the BGC to the bronchoscopy indication; (ii) simplified strategy - extended use of bronchoscopy plus BGC also on small and peripheral lesions. Outcomes modeled were rate of invasive procedures, quality-adjusted life-years (QALYs), costs, and incremental cost-effectiveness ratios. Compared with the standard practice, the location-based strategy (i) reduced absolute rate of invasive procedures by 3.3% without increasing costs at the current BGC market price. It resulted in savings when the BGC price was less than $3,000. The simplified strategy (ii) reduced absolute rate of invasive procedures by 10% and improved quality-adjusted life expectancy, producing an incremental cost-effectiveness ratio of $10,109 per QALY. In patients with indeterminate nodules, both BGC strategies reduced unnecessary invasive procedures at high risk of adverse events. Moreover, compared with the standard practice, the simplified use of BGC for central and peripheral lesions resulted in larger QALYs gains at acceptable cost. The location-based is cost-saving if the price of classifier declines. In the third case study, we assessed the choice of using “enrichment” clinical trials to evaluate the efficacy of cetuximab in the pivotal trial for the Food and Drug Administration (FDA)-approval. In 2004, cetuximab (Erbitux) was FDA-approved for patients with EGFR-positive metastatic colorectal cancer (mCRC) based on a pivotal trial restricted to biomarker-positives. However, subsequent studies showed that EGFR-negatives also benefitted from cetuximab, while patients with mutated K-ras tumors do not. In 2009, EGFR testing of patients with mCRC was formally discouraged. We assessed clinical and cost consequences of approving cetuximab without clinical evidence about biomarker-negative patients. We performed an analysis on patients with chemotherapy-resistant mCRC, and compared three strategies with best supportive care (reference strategy): (i) “treat all”: cetuximab to all patients regardless of biomarker status; (ii) “treat EGFR+ only”: cetuximab to patients with EGFR-expressing tumors; (iii) “treat wild type K-ras only”: cetuximab restricted to patients with wild-type K-ras tumors. We calculated the costs and life-years gained if the “treat all” strategy were adopted from 2004 to 2009. Treating wild type K-ras only was the most cost-effective strategy ($85,809/life-year). Treating all resulted in $119,893/life-year, while treating EGFR+ only yielded $120,277/life-year. From 2004-2009, treating all would have resulted in 33,640 life-years gained and $3,940 million cost, while treating only EGFR+ produced 21,420 life-years gained and $2,420 million cost. Indicating cetuximab for EGFR-positives between 2004 and 2009, instead of all, led to life-years lost without reducing costs per life-year gained. Testing biomarker-negative patients in the pivotal trial would have produced valuable information that could have improved health outcomes. Based on the findings from the case studies, we concluded that biomarker-driven modalities for cancer prevention can have impactful clinical implications. Specifically, blood-based modality with mSEPT9 for colorectal cancer screening can match the benefits of colonoscopy and be recommended in clinical practice (adherence rate for this modality is higher than 70%). In the second case study, the use of the BGC to investigate central and peripheral lesions resulted in larger clinical benefits at acceptable costs compared to bronchoscopy. Moreover, its use would be cost saving if the market price of this device would be less than $3,000. In the last case study, we showed that the appropriate use of a biomarker-stratified trial design would have improved health outcomes in patients with metastatic colorectal cancer in the years 2004-2009. Additionally, we concluded that simulation models can be effectively used as epidemiological instrument to answer several research questions, for example: to timely appraise long-term outcomes (first case study); to assess factors not easily addressable in clinical trials, such as real-world adherence (first case study); to assess and to directly compare multiple screening modalities (first and second case study); to evaluate economic outcomes (second case study); to assess the target market price of the genomic application (second case study); to explore limitations of original study design (third case study).