As current follow-up modalities for colorectal carcinoma (CRC) have restricted sensitivity, novel diagnostic tools are needed. The presence of CRC changes the endogenous metabolism, resulting in the release of a specific volatile organic compounds (VOC) pattern that can be detected with an electronic nose or Aeonose^TM. To evaluate the use of an electronic nose in the follow-up of CRC, we studied the effect of curative surgery on the VOC pattern recognition using Aeonose^TM. A prospective cohort study was performed, in which 47 patients diagnosed with CRC were included, all of whom underwent curative surgical resection. Breath testing was performed before and after surgery using the Aeonose^TM. A machine learning model was developed by discerning between the 94 pre-and postoperative breath samples. The training model differentiated between the pre-and postoperative CRC breath samples with a sensitivity and specificity of 0.78 (95%CI 0.61–0.90) and 0.73 (95%CI 0.56–0.86), respectively, with an accuracy of 0.76 (95%CI 0.66–0.85), and an area under the curve of 0.79 (95%CI 0.68–0.89). The internal validation of the test set resulted in an accuracy of 0.75 (95%CI 0.51–0.91) and AUC of 0.82 (95%CI 0.61–1). In conclusion, our results suggest that the VOC pattern of CRC patients is altered by curative surgery in a short period, indicating that the exhaled VOCs might be closely related to the presence of CRC. However, to use Aeonose^TM as a potential diagnostic tool in the clinical follow-up of CRC patients, the performance of the models needs to be improved through further large-scale prospective research.

PURPOSE: The Clinical Pharmacogenetics Implementation Consortium (CPIC) recommends screening for four common DPYD variants to prevent severe toxicity in patients with cancer treated with fluoropyrimidines. A 50% starting dose followed by toxicity-based dose titration is advised for patients heterozygous for these variants. In this study, the appropriateness of the CPIC-recommended 5-fluorouracil (5-FU) starting dose was evaluated. PATIENTS AND METHODS: Patients were grouped into four variant categories ( DPYD *2A [c.1905+1G>A], DPYD *13 [c.1679T>G], c.2846A>T [p.D949V], c.1236G>A/HapB3 [p.E412E]) and a DPYD wild-type control group. Uracil loading tests were used for phenotyping. Variant patients started on a 50% reduced 5-FU dose. On the basis of steady-state 5-FU plasma concentrations, dose adjustments were made during cycles 2-4 until an 5-FU target AUC _0-46h of 20-30 mg × h/L was achieved, if tolerated. RESULTS: Twenty-six wild-type controls and 34 DPYD variant patients were included: 16 with c.1236G>A/HapB3, eight with c.1905+1G>A, eight with p.D949V, and two with c.1679T>G. Heterozygous carriers of c.1905+1G>A ( DPYD *2A) and c.1679T>G ( DPYD* 13) displayed significant reduced uracil metabolism. The impact on uracil clearance was highly variable in p.D949V but only minor in c.1236G>A/HapB3 variants. In all, 65% of wild-type controls had 5-FU exposure within target range on a 100% dose (mean, 23.2; IQR, 6.6). In 97% of all variant patients, the 50% reduced dose resulted in 5-FU underexposure, with a median AUC of 10.6 mg × h/L (IQR, 3.2). Dose escalation to 70% or higher was tolerated in most patients, reaching the target AUC in 68% of patients. CONCLUSION: The current CPIC guidelines are overly conservative for c.1236G>A/HapB3 and most p.D949V variants. A 75% starting dose is more appropriate for most c.1236G>A/HapB3 carriers. We recommend 5-FU therapeutic drug monitoring in all patients with DPYD variants to achieve optimal 5-FU exposure.

Ongoing health challenges, such as the increased global burden of chronic disease, are increasingly answered by calls for personalized approaches to healthcare. Genomic medicine, a vital component of these personalization strategies, is applied in risk assessment, prevention, prognostication, and therapeutic targeting. However, several practical, ethical, and technological challenges remain. Across Europe, Personal Health Data Space (PHDS) projects are under development aiming to establish patient-centered, interoperable data ecosystems balancing data access, control, and use for individual citizens to complement the research and commercial focus of the European Health Data Space provisions. The current study explores healthcare users’ and health care professionals’ perspectives on personalized genomic medicine and PHDS solutions, in casu the Personal Genetic Locker (PGL). A mixed-methods design was used, including surveys, interviews, and focus groups. Several meta-themes were generated from the data: (i) participants were interested in genomic information; (ii) participants valued data control, robust infrastructure, and sharing data with non-commercial stakeholders; (iii) autonomy was a central concern for all participants; (iv) institutional and interpersonal trust were highly significant for genomic medicine; and (v) participants encouraged the implementation of PHDSs since PHDSs were thought to promote the use of genomic data and enhance patients’ control over their data. To conclude, we formulated several facilitators to implement genomic medicine in healthcare based on the perspectives of a diverse set of stakeholders.

Background: The role of radiological staging and surveillance imaging is under debate for T1 colorectal cancer (CRC) as the risk of distant metastases is low and imaging may lead to the detection of incidental findings.Objective: The aim of this study was to evaluate the yield of radiological staging and surveillance imaging for T1 CRC.Methods: In this retrospective multicenter cohort study, all patients of 10 Dutch hospitals with histologically proven T1 CRC who underwent radiological staging in the period 2000–2014 were included. Clinical characteristics, pathological, endoscopic, surgical and imaging reports at baseline and during follow‐up were recorded and analyzed. Patients were classified as high‐risk T1 CRC if at least one of the histological risk factors (lymphovascular invasion, poor tumor differentiation, deep submucosal invasion or positive resection margins) was present and as low‐risk when all risk factors were absent.Results: Of the 628 included patients, 3 (0.5%) had synchronous distant metastases, 13 (2.1%) malignant incidental findings and 129 (20.5%) benign incidental findings at baseline staging. Radiological surveillance was performed among 336 (53.5%) patients. The 5‐year cumulative incidence of distant recurrence, malignant and benign incidental findings were 2.4% (95% confidence interval (CI): 1.1%–5.4%), 2.5% (95% CI: 0.6%–10.4%) and 18.3% (95% CI: 13.4%–24.7%), respectively. No distant metastatic events occurred among low‐risk T1 CRC patients.Conclusion: The risk of synchronous distant metastases and distant recurrence in T1 CRC is low, while there is a substantial risk of detecting incidental findings. Radiological staging seems unnecessary prior to local excision of suspected T1 CRC and after local excision of low‐risk T1 CRC. Radiological surveillance should not be performed in patients with low‐risk T1 CRC.

If Colorectal cancer (CRC) is detected and treated early, the survival rate is high. This is one of the reasons that population-based screening programs for the early detection of CRC using the faecal immunochemical test (FIT) started worldwide. These programs compete with regular colonoscopy programs and increase the waiting time for symptomatic patients. However, the literature has shown that the correlation between intestinal complaints and the gain of colonoscopy is poor. The aim of this study is to assess the diagnostic utility of symptoms for the yield (CRC) of colonoscopy and to compare this with the diagnostic utility of FIT when offered to symptomatic patients. Methods: We performed a systematic review search for CRC as an outcome of colonoscopy in referred symptomatic patients and separately for CRC as an outcome in symptomatic patients with a positive FIT. We searched systematically for clinical trials or observational studies in databases, followed by hand-searching of reference lists. We used random Meta-Disc to evaluate the diagnostic performance, using the exploration of heterogeneity with a variety of test statistics and by computing the pooled estimates. Results: We included 35 studies, with almost 5 million symptomatic patients. In addition, we included nine prospective studies with a positive FIT in symptomatic patients, with more than 5000 patients. Significant heterogeneity was found for every symptom and the outcome of colonoscopy in the effect size of sensitivity, specificity, positive likelihood ratio, negative likelihood ratio and diagnostic odds ratio. In a random effect model, the pooled sensitivity of colonoscopy in symptomatic patients was very low (25%). However, the pooled sensitivity in symptomatic patients with a positive FIT was 83% and the pooled specificity 77%. A total of 75 symptomatic patients (1.4%) had a false-negative FIT. Conclusion: Adding FIT in symptomatic patients seems useful for predicting CRC as an outcome of colonoscopy. FIT seems a potential tool for an improved triage of colonoscopy in symptomatic patients.

PURPOSE: The use of real-world data (RWD) in oncology is becoming increasingly important for clinical decision making and tailoring treatment. Despite the significant success of targeted therapy and immunotherapy in advanced melanoma, substantial variability in clinical responses to these treatments emphasizes the need for personalized approaches to therapy. MATERIALS AND METHODS: In this pilot study, 239 patients with melanoma were included to predict the response to both targeted therapies and immunotherapies. We used machine learning (ML) to incorporate RWD and applied explainable artificial intelligence (XAI) to explain the individual predictions. RESULTS: We developed, validated, and compared four ML models to evaluate 2-year survival using RWD. Our research showed encouraging outcomes, achieving an AUC of more than 80% and an estimated accuracy of over 74% across the four ML models. The random forest model exhibited the highest performance in predicting 2-year survival with an AUC of 0.85. Local interpretable model-agnostic explanations was used to explain individual predictions and provide trust and insights into the clinical implications of the ML model. CONCLUSION: With this proof-of-concept, we integrated RWD into predictive modeling using ML techniques to predict clinical outcomes and explore their potential implications for clinical decision making. The potential of XAI was demonstrated to enhance trust and improve the usability of the model in clinical settings. Further research, including foundation modeling and generative AI, will likely increase the predictive power of prognostic and predictive ML models in advanced melanoma.