Erika Graf

Objectives Structured Early detection of Asymptomatic Liver fibrosis and cirrhosis (SEAL) is a population-based screening programme using non-invasive tests for the early detection of liver fibrosis. This study evaluates the cost implications if the SEAL programme were to be implemented in routine care in Germany. Design This study models cost differences with and without the SEAL screening programme. We regress costs of care on patient characteristics (age, comorbidities, sex, liver diseases, liver cancer and liver fibrosis and cirrhosis (LCI) stage) using statutory health insurance (SHI) data from routine care patients with LCI (n=4177). Based on these results, we predict per-patient costs for the patients newly diagnosed with LCI by SEAL (n=45). Costs with and without screening are estimated using patient age and LCI stage distributions from either SEAL or routine care. Setting SEAL was conducted in two German states. Initial screening was performed by patients’ primary care physicians. Participants Individuals insured by SHI without a prior diagnosis of LCI, eligible for Check-up 35, a general health check-up programme primarily targeting adults aged 35 and older, conducted by primary care physicians. Interventions Screening via aspartate aminotransferase to platelet ratio index in primary care, for further evaluation serological diagnostics and ultrasound examinations in secondary care and specific assessment for definite diagnosis including transient elastography and liver biopsy for selected cases in tertiary care. Primary and secondary outcome measures Primary outcome measures: expected 5-year cost changes for SEAL patients diagnosed with fibrosis or cirrhosis compared to costs without a screening programme. Secondary outcome measures: case mix of leading chronic liver disease and LCI stages among patients diagnosed with advanced fibrosis or cirrhosis in SEAL versus routine care without screening. Results Screening leads to fewer decompensated cases at initial diagnosis (4.6% in SEAL vs 22.8% in routine care) and thus savings in the costs of care within the first years of diagnosis: total expected costs per case were €2175 lower (bias-corrected bootstrap CIs (BCI): €527 to 3734), and LCI-associated costs were reduced by €1218 (BCI: €296 to 2164). Comparing the savings to the additional costs of diagnosis (range: €1575–1726 per detected LCI case) reveals that average changes in costs with screening range from moderate savings to moderate extra costs. Conclusions SEAL liver screening identifies patients in less advanced stages of LCI. If only costs were considered that are directly attributable to LCI, savings within 5 years are unlikely to fully outweigh the costs of screening. However, since this approach might miss additional LCI-related costs, SEAL appears to be cost-neutral compared with routine care when considering total healthcare costs. Registration number The SEAL registration number is DRKS00013460. This study relates to its results.

Background: The advancement of Artificial Intelligence, particularly Large Language Models (LLMs), is rapidly progressing. LLMs, such as OpenAI’s GPT, are becoming vital in scientific and medical processes, including text production, knowledge synthesis, translation, patient communication and data analysis. However, the outcome quality needs to be evaluated to assess the full potential for usage in statistical applications. LLMs show potential for all research areas, including teaching. Integrating LLMs in research, education and medical care poses opportunities and challenges, depending on user competence, experience and attitudes. Objective: This project aims at exploring the use of LLMs in supporting statistical consulting by evaluating the utility, efficiency and satisfaction related to the use of LLMs in statistical consulting from both advisee and consultant perspective. Within this project, we will develop, execute and evaluate a training module for the use of LLMs in statistical consulting. In this context, we aim to identify the strengths, limitations and areas for potential improvement. Furthermore, we will explore experiences, attitudes, fears and current practices regarding the use of LLMs of the staff at the Medical Center and the University of Freiburg. Materials and methods: This multimodal study includes four study parts using qualitative and quantitative methods to gather data. Study part (I) is designed as mixed mode study to explore the use of LLMs in supporting statistical consulting and to evaluate the utility, efficiency and satisfaction related to the use of LLMs. Study part (II) uses a standardized online questionnaire to evaluate the training module. Study part (III) evaluates the consulting sessions using LLMs from advisee perspective. Study part (IV) explores experiences, attitudes, fears and current practices regarding the use of LLMs of the staff at the Medical Center and the University of Freiburg. This study is registered at the Freiburg Registry of Clinical Studies under the ID: FRKS004971.

Background: Dementia detection rate is very low in German primary care, as well as internationally. Amyloid PET imaging can reliably clarify an unclear diagnosis of dementia or an uncertain diagnosis of Alzheimer’s Disease (AD), improving diagnostic confidence and patient management. At present, systematic evidence of such benefits is still lacking. The patient‐related benefits of amyloid PET imaging (ENABLE) study will test whether amyloid PET will have a benefit on patients' functional status. Here, we describe the design of ENABLE, a randomized controlled two‐armed coverage with evidence development study in Germany.Method: Approximately 1,126 patients from 20 German academic memory clinics with a clinical diagnosis of mild‐moderate dementia of unclear etiology or of AD with low diagnostic certainty will be included in the study. Those patients will be randomly assigned to amyloid‐PET or routine diagnostics (no amyloid‐PET) (see Figure). The primary endpoint is change from baseline in the ability to perform instrumental activities of daily living (IADL) at 18 months as measured by the Amsterdam IADL Questionnaire. Secondary endpoints include the change in diagnosis, diagnostic confidence, cognition and clinical outcomes of patients and caregivers up until 24 months. We will apply linear and mixed linear models to assess primary and secondary endpoints.Result: We expect patients' and caregivers' quality of life to be improved after receiving an amyloid PET formal diagnosis, as this will result into timely pharmacological and non‐pharmacological interventions.Conclusion: This study will benefit physicians’ and payers’ decisions on the provision of health care for patients with dementia in Germany.

Background: Dementia detection rate is very low in German primary care, as well as internationally. Amyloid PET imaging can reliably clarify an unclear diagnosis of dementia or an uncertain diagnosis of Alzheimer’s Disease (AD), improving diagnostic confidence and patient management. At present, systematic evidence of such benefits is still lacking. The patient‐related benefits of amyloid PET imaging (ENABLE) study will test whether amyloid PET will have a benefit on patients' functional status. Here, we describe the design of ENABLE, a randomized controlled two‐armed coverage with evidence development study in Germany.Method: Approximately 1,126 patients from 20 German academic memory clinics with a clinical diagnosis of mild‐moderate dementia of unclear etiology or of AD with low diagnostic certainty will be included in the study. Those patients will be randomly assigned to amyloid‐PET or routine diagnostics (no amyloid‐PET) (see Figure). The primary endpoint is change from baseline in the ability to perform instrumental activities of daily living (IADL) at 18 months as measured by the Amsterdam IADL Questionnaire. Secondary endpoints include the change in diagnosis, diagnostic confidence, cognition and clinical outcomes of patients and caregivers up until 24 months. We will apply linear and mixed linear models to assess primary and secondary endpoints.Result: We expect patients' and caregivers' quality of life to be improved after receiving an amyloid PET formal diagnosis, as this will result into timely pharmacological and non‐pharmacological interventions.Conclusion: This study will benefit physicians’ and payers’ decisions on the provision of health care for patients with dementia in Germany.

We prospectively evaluated the effects of stereotactic body radiotherapy (SBRT) on circulating immune cells. Patients with oligo-metastatic and oligo-progressive pulmonary lesions were treated with SBRT with (cSBRT) or without (SBRT group) concurrent systemic treatment (chemotherapy or immune checkpoint blockade) using different fractionation regimes. Immunoprofiling of peripheral blood cells was performed at baseline, during, at the end of SBRT, and at the first and second follow-ups. The study accrued 100 patients (80 with evaluable samples). The proportion of proliferating CD8+ T-cells significantly increased after treatment. This increase remained significant at follow-up in the SBRT group, but not in the cSBRT group and was not detected with doses of >10Gy per fraction indicating that lower doses are necessary to increase proliferating T-cells' frequency. We detected no favorable impact of concurrent systemic treatment on systemic immune responses. The optimal timing of systemic treatment may be post-SBRT to leverage the immune-modulating effects of SBRT.

Background: Liver cirrhosis is a major healthcare problem and the mortality rate is high. During recent years, systemic inflammation has been recognized as a major driver of hepatic decompensation and progression of liver cirrhosis to acute-on-chronic liver failure (ACLF). The aim of the CYTOHEP study is to assess the impact of extracorporeal hemoadsorption with the CytoSorb adsorber on serum bilirubin concentrations, humoral inflammation parameters, liver function parameters, and patient survival in patients with ACLF and acute kidney injury (AKI). Methods: The CYTOHEP study is a prospective, single-center, open-label, three-arm, randomized, controlled intervention trial. Patients with ACLF and AKI stage 3 according to Kidney Disease: Improving Global Outcome (KDIGO) criteria will be randomized into three groups to be treated with (1) continuous renal replacement therapy (CRRT) and CytoSorb, (2) CRRT without CytoSorb, and (3) without both, CRRT and CytoSorb. In the hemoadsorption group, CytoSorb will be used for 72 h. The other groups receive standard of care with early or late initiation of CRRT, respectively. Primary endpoint of the study is serum bilirubin concentration after 72 h, important secondary endpoints are 30-day survival and a panel of inflammatory parameters. Discussion: The CYTOHEP study is designed to evaluate the benefit of extracorporeal hemoadsorption in patients with ACLF. The results of this study will help to better understand the potential role of hemoadsorption for the treatment of ACLF and its impact on bilirubin levels, inflammatory parameters, and survival. Trial registration: ClinicalTrials.gov NCT05019352. Registered on August 24, 2021. Deutsches Register Klinischer Studien (DRKS) DRKS00026082.

Motivated by a clinical prediction problem, a simulation study was performed to compare different approaches for building risk prediction models. Robust prediction models for hospital survival in patients with acute heart failure were to be derived from three highly correlated blood parameters measured up to four times, with predictive ability having explicit priority over interpretability. Methods that relied only on the original predictors were compared with methods using an expanded predictor space including transformations and interactions. Predictors were simulated as transformations and combinations of multivariate normal variables which were fitted to the partly skewed and bimodally distributed original data in such a way that the simulated data mimicked the original covariate structure. Different penalized versions of logistic regression as well as random forests and generalized additive models were investigated using classical logistic regression as a benchmark. Their performance was assessed based on measures of predictive accuracy, model discrimination, and model calibration. Three different scenarios using different subsets of the original data with different numbers of observations and events per variable were investigated. In the investigated setting, where a risk prediction model should be based on a small set of highly correlated and interconnected predictors, Elastic Net and also Ridge logistic regression showed good performance compared to their competitors, while other methods did not lead to substantial improvements or even performed worse than standard logistic regression. Our work demonstrates how simulation studies that mimic relevant features of a specific data set can support the choice of a good modeling strategy.