Stefan Buchka

Stefan Buchka

Auto-generated profile, based on publicly available information

Stefan Buchka

Institute for Medical Information Processing, Biometry and Epidemiology, LMU, Munich, Germany

Research Focus & Metrics

Academic Career Length

6Years (2021-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
11 May 2021 in Genome Biology
Verified
Journal Article
Open access
PDF file
SBStefan Buchka
Stefan Buchka
AHAlexander Hapfelmeier
Alexander Hapfelmeier
PGPaul P. Gardner
Paul P. Gardner
RWRory Wilson
Rory Wilson
ABAnne-Laure Boulesteix
Anne-Laure Boulesteix
Most research articles presenting new data analysis methods claim that “the new method performs better than existing methods,” but the veracity of such statements is questionable. Our manuscript discusses and illustrates consequences of the optimistic bias occurring during the evaluation of novel data analysis methods, that is, all biases resulting from, for example, selection of datasets or competing methods, better ability to fix bugs in a preferred method, and selective reporting of method variants. We quantitatively investigate this bias using an example from epigenetic analysis: normalization methods for data generated by the Illumina HumanMethylation450K BeadChip microarray.

Verified
Journal Article
Open access
PDF file
JGJonathan A. Gernert
Jonathan A. Gernert
LBLuise Böhm
Luise Böhm
MSMichaela Starck
Michaela Starck
SBStefan Buchka
Stefan Buchka
TKTania Kümpfel
Tania Kümpfel
Joachim HavlaJoachim Havla
Joachim Havla
The establishment of surrogate markers to detect disability progression in persons with multiple sclerosis (PwMS) is important to improve monitoring of clinical deterioration. Optical coherence tomography (OCT) could be such a tool. However, sufficient longitudinal data of retinal neuroaxonal degeneration as a marker of disease progression exist only for PwMS with a relapsing–remitting course (RRMS) so far. In contrast, longitudinal data of retinal layers in patients with primary-progressive MS (PPMS) are inconsistent, and the association of OCT parameters with ambulatory performance in PwMS has rarely been investigated. We aimed to investigate the relative annual rates of change in retinal layers in PwMS (RRMS and PPMS) compared with healthy controls (HC) using OCT and to evaluate their association with ambulatoryfunctionalscore (AS) worsening in PPMS. A retrospective analysis of a longitudinal OCT dataset of the retinal layers of PwMS and HC from two MS centers in Germany was performed. Walking ability was measured over a standardized distance of 500 m, and changes during the observation period were categorized using the AS and the expanded disability status scale (EDSS). 61 HC with 121 eyes and 119 PwMS (PPMS: 57 patients with 108 eyes; RRMS: 62 patients with 114 eyes) were included. The median follow-up time for PwMS was 3 years. The relative annual change of pRNFL (peripapillary retinal nerve fiber layer) and INL (inner nuclear layer) was significantly different in PwMS compared with HC. RRMS and PPMS subgroups did not differ in the annual atrophy rates. In patients with PPMS, worsening of the AS was significantly associated with increased thinning of the TMV (total macular volume), GCIP (ganglion cell and inner plexiform layer), and ONPL (outer nuclear and outer plexiform layer) (all p-value < 0.05, r > 0.30). For every −0.1% decrease in the TMV, GCIP, and ONPL, the risk of a deterioration in the AS increased by 31% (hazard ratio (HR): 1.309), 11% (HR: 1.112), and 16% (HR: 1.161), respectively. In addition, worsening EDSS in PPMS was significantly associated with the relative annual atrophy rates of pRNFL, TMV, and GCIP (all p-value < 0.05). Disability progression in PPMS can be measured using OCT, and increasing annual atrophy rates of the inner retinal layers are associated with worsening ambulation. OCT is a robust and side-effect-free imaging tool, making it suitable for routine monitoring of PwMS.

15 September 2026 in The Lancet Healthy Longevity
Verified
Journal Article
Open access
CSClaudia D Spies
Claudia D Spies
KSKatrin Schmidt
Katrin Schmidt
JKJörn Kiselev
Jörn Kiselev
SBStefan Buchka
Stefan Buchka
MBManfred Blobner
Manfred Blobner
CZChristian Zöllner
Christian Zöllner

26 December 2025 in PLOS ONE
Verified
Journal Article
Open access
SBStefan Buchka
Stefan Buchka
JHJoachim Havla
Joachim Havla
BÖBegüm Irmak Ön
Begüm Irmak Ön
RRRaphael Rehms
Raphael Rehms
UMUlrich Mansmann
Ulrich Mansmann
Background: Individual-level surrogacy (ILS) assesses how well a surrogate endpoint predicts treatment effect at the individual level. The paper discusses mutual information (MI) and the likelihood reduction factor (LRF) as ways to quantify ILS. It also reassesses ILS statements for people with relapsing-remitting multiple sclerosis (RRMS) by using T2 MRI lesions (T2L) as surrogates for disability and disease activity. ILS is often reported using inadequate concepts, e.g. subgroup analysis, correlation, odds ratios, sensitivity/specificity, or metrics based on Prentice’s criteria. Methods: LRF assesses ILS quality by determining prediction quality through shared information (MI) between surrogate(s) and clinical endpoint(s). A simulation study validates LRF as a measure of ILS quality. Individual-level data from ten randomized controlled trials (n = 5673) provided longitudinal information on T2L, T2 MRI lesion volumes (T2V), future disability progression (EDSS), and relapses. LRFs for different scenarios were calculated. Results were compared to those obtained by methods commonly applied in RRMS literature. Results: Simulations confirmed the robustness of LRF as a reliable ILS quality measure. Two of ten trials showed weak ILS between T2V and EDSS (LRF = 0.21, CI95%: 0.16–0.26; LRF = 0.28, CI95%: 0.23–0.34). Other LRFs were below 0.2. A method commonly used in the MS literature also showed no strong ILS. Conclusion: LRF is an important measure to quantify ILS and prediction quality. But it is rarely applied in RRMS research. LRF did not reveal robust surrogacy patterns when applied to data of ten clinical trials. Existing surrogacy claims should be reassessed since ILS assessments in the MS literature may have major limitations.

1 January 2025 in Therapeutic Advances in Neurological Disorders
Verified
Journal Article
Open access
SBStefan Buchka
Stefan Buchka
AHAlexander Hapfelmeier
Alexander Hapfelmeier
JKJan Kirschke
Jan Kirschke
VSVerena Steuerwald
Verena Steuerwald
MNMarkus Naumann
Markus Naumann
PGProVal-MS Study Group
ProVal-MS Study Group
Background: The course of relapsing-remitting multiple sclerosis (RRMS), frequently preceded by the clinically isolated syndrome (CIS), is variable and challenging to predict. Given many treatment options available, prognostic algorithms are gaining importance in informing initial treatment decisions. However, to date, only a few externally validated exists. External validation, which involves the application of a model to independent data, is essential. Privacy-preserving federated analyses of individual-level data facilitate external validation using clinical datasets that are typically difficult to access. Objectives: Using data from the ProVal-MS study to externally validate the multiple sclerosis treatment decision score (MS-TDS), a predictive algorithm for early RRMS and CIS. The MS-TDS predicts the probability of the occurrence of at least one new or enlarging T2 lesion within 6–24 months following the onset of the disease and supports choosing between initiating platform treatment or a ‘wait-and-see’ approach. A secondary objective is to demonstrate the feasibility of privacy-preserving federated concepts within the Data Integration for Future Medicine (DIFUTURE) consortium. Design: Prospective, multicentric, non-interventional cohort study (ProVal-MS) within DIFUTURE. Methods: The calibrated MS-TDS was evaluated using the area under the receiver operating characteristic curve (AUROC) and the Brier score in both pooled and distributed settings. A decision curve analysis (DCA) was used to evaluate the net benefit of treatment decisions made by the MS-TDS in comparison to those made by treating neurologists. Results: Of the 271 individuals diagnosed with CIS or early RRMS, 202 (78.2%) received platform treatment, while 59 (21.8%) did not receive treatment. The AUROC was 0.561 (95% CI: 0.492–0.630) in the pooled analysis and 0.567 (95% CI: 0.496–0.634) in the distributed analysis. DCA demonstrated a net benefit that was commensurate with that achieved by decisions made by experienced neurologists. Conclusion: The external validation of the MS-TDS demonstrated low, non-significant predictive performance; however, it may serve as a useful complement, particularly for less-experienced neurologists. The distributed validation was found to be both feasible and compliant with data protection regulations.

21 May 2026
Verified
Book Chapter
Open access
PPPeter Pallaoro
Peter Pallaoro
SBSandra Bilger
Sandra Bilger
MBMartin Boeker
Martin Boeker
MBMarkus Budeus
Markus Budeus
VDVera Dehmelt
Vera Dehmelt
SBStefan Buchka
Stefan Buchka
Data quality assessment in medical research is crucial if the findings are to be data-driven. This is especially true if data is collected in clinical routine and originates from multicenter studies with complex, multi-step data processing pipelines that aim to harmonize and integrate data across all participating centers for data evaluation and analysis. The multicentric, non-interventional, and prospective ProVal-MS cohort study was established to externally validate a treatment decision score for adults with relapsing multiple sclerosis. Hence, a data quality management strategy was necessary to ensure the reliability of the findings. The evaluation and analysis of the implemented data quality assessment strategy is the primary focus of this work. It involves identifying data quality issues, assessing whether they can be resolved, and exploring and implementing potential improvements. A questionnaire was used to identify, evaluate, and analyze the data quality issues encountered, along with the time required to resolve them. We found that with an earlier DQ assessment, up to 700 (70%) working hours could be saved. As a consequence, we plan to assess data quality earlier in the processing pipeline to reduce the time required to detect and resolve data quality issues.