Mikkel Runason Simonsen
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Mikkel Runason Simonsen has a total of 146 co-authors. Here are the top co-authors by number of shared publications:
- Tarec Christoffer El-Galaly
Department of Hematology Aarhus University Hospital Aarhus Denmark
8 shared publications - Ahmed Ludvigsen Al-Mashhadi
Department of Hematology, Aarhus University Hospital, Aarhus, Denmark
5 shared publications Tarec Christoffer El‐Galaly
Department of Clinical Medicine Aarhus University Aarhus Denmark
5 shared publications- Thomas Stauffer Larsen
Odense University Hospital, Odense, Denmark
5 shared publications - Michael Roost Clausen
Department of Hematology, Vejle Hospital, Sygehus Lillebaelt, Vejle, Denmark
4 shared publications - Lasse Hjort Jakobsen
Department of Hematology, Aalborg University Hospital, Aalborg, Denmark
3 shared publications
Latest Publications
- Open Access
- 4 August 2026
- Published by Wiley in Clinical Pharmacology & Therapeutics
Risk of Bias Associated with Varying Sampling Strategies for External Comparator Cohorts for Single‐Arm Trials: Insights from Aggressive Lymphoma
Marketing authorizations for oncology drugs are often based on single‐arm trials (SATs), and real‐world data‐based external comparator cohorts (ECCs) are increasingly explored to contextualize results. In retrospective studies, comparators may be identified by time of diagnosis but included based on time of relapse; such sampling strategies can introduce bias, potentially leading to erroneous efficacy conclusions. Simulated diffuse large B‐cell lymphoma (DLBCL) cohorts were generated using published time‐to‐relapse and post‐relapse survival distributions. Three ECC sampling strategies were evaluated: naïve (screening and inclusion periods coincide), delayed (inclusion period extends beyond screening), and displacement (screening and inclusion periods shifted relative to each other). Different sampling strategies caused significant changes in the time‐to‐relapse distribution in the ECC and, by proxy, the estimated 2‐year post‐relapse overall survival (22.5–52.2%). Naïve sampling consistently resulted in oversampling of early relapses, though bias was reduced with extended inclusion periods; in hypothetical trials, it produced inflated Type I errors (10.5–84.9%) despite α = 0.025. Delayed sampling reduced bias when the inclusion period was sufficiently extended to capture late relapses, yielding unbiased estimates under long delays (Type I error of 2.5% at a 6‐year delay). Displacement sampling depended strongly on the degree of shift: moderate shifts minimized bias, whereas long displacements introduced substantial distortion (Type I error of 2.7% and 0.0% at displacements of 2 and 5, respectively). Real‐world data‐based ECCs in relapsed/refractory oncology are susceptible to selection bias from sampling strategies, particularly through distortions in time‐to‐relapse, leading to erroneous efficacy conclusions; similar bias may occur in non‐oncology settings where time‐to‐treatment initiation is prognostic.
- 4 August 2026
- Published by Wiley in Clinical Pharmacology & Therapeutics
Open AccessRisk of Bias Associated with Varying Sampling Strategies for External Comparator Cohorts for Single‐Arm Trials: Insights from Aggressive Lymphoma
Marketing authorizations for oncology drugs are often based on single‐arm trials (SATs), and real‐world data‐based external comparator cohorts (ECCs) are increasingly explored to contextualize results. In retrospective studies, comparators may be identified by time of diagnosis but included based on time of relapse; such sampling strategies can introduce bias, potentially leading to erroneous efficacy conclusions. Simulated diffuse large B‐cell lymphoma (DLBCL) cohorts were generated using published time‐to‐relapse and post‐relapse survival distributions. Three ECC sampling strategies were evaluated: naïve (screening and inclusion periods coincide), delayed (inclusion period extends beyond screening), and displacement (screening and inclusion periods shifted relative to each other). Different sampling strategies caused significant changes in the time‐to‐relapse distribution in the ECC and, by proxy, the estimated 2‐year post‐relapse overall survival (22.5–52.2%). Naïve sampling consistently resulted in oversampling of early relapses, though bias was reduced with extended inclusion periods; in hypothetical trials, it produced inflated Type I errors (10.5–84.9%) despite α = 0.025. Delayed sampling reduced bias when the inclusion period was sufficiently extended to capture late relapses, yielding unbiased estimates under long delays (Type I error of 2.5% at a 6‐year delay). Displacement sampling depended strongly on the degree of shift: moderate shifts minimized bias, whereas long displacements introduced substantial distortion (Type I error of 2.7% and 0.0% at displacements of 2 and 5, respectively). Real‐world data‐based ECCs in relapsed/refractory oncology are susceptible to selection bias from sampling strategies, particularly through distortions in time‐to‐relapse, leading to erroneous efficacy conclusions; similar bias may occur in non‐oncology settings where time‐to‐treatment initiation is prognostic.
- Open Access
- 26 July 2026
- Published by Wiley in Cancer Medicine
- Vol. 15 (7) , e72116
Machine Learning for Prediction of High‐Risk Infections in Patients With Cancer
Purpose: Infectious complications in patients with cancer are contributors to hospitalisations, treatment disruption, and mortality. This study aimed to develop machine learning (ML) models for risk stratification of infection‐related hospitalisations (IRHs) among patients with cancer. Methods: Adult patients diagnosed with and treated for lymphoma, multiple myeloma (MM), chronic lymphocytic leukaemia (CLL), colorectal, or lung cancer from 2013 to 2023 in the North Denmark Region were included. Data from national cancer registries and patient health records were used. Serious infections were defined as sepsis, positive blood culture, intensive care admission, or death during hospitalisation. Models were developed using multiple ML methodologies across three different data setups: ML —using data up to hospital admission; ML 48 — using data up to 48 h after admission; and ML 48,reduced —a pruned version of ML 48 with fewer features. Data were split into training, validation and held‐out test sets. Results: Among 9874 patients 2661 IRHs occurred, including 691 (21.5%) serious infections. On the held‐out test set, ML achieved the best performance (ROC‐AUC: 0.79) versus ML 48 (0.76), with ML 48 having higher specificity at matched sensitivity (50.9% vs 44.9% for sensitivity ~85%). ML 48,reduced reached a ROC‐AUC of 0.75 (specificity 46.5% at sensitivity ~85%). Subgroup analysis showed the best performance for lymphoma (ROC‐AUC: 0.86) and lowest for lung cancer (ROC‐AUC: 0.73). The model performed better in patients with ECOG score 0–1 compared to ≥ 2 (ROC‐AUC 0.88 vs. 0.77). SHAP analysis identified previous blood cultures, carbamide, and neutrophil counts as key predictive features across all models. Conclusion: The ML models showed moderate performance in risk stratification of IRHs in patients with cancer. Such models could provide decision support in hospitalisation decisions and discharge.
- 26 July 2026
- Published by Wiley in Cancer Medicine
- Vol. 15 (7) , e72116
Open AccessMachine Learning for Prediction of High‐Risk Infections in Patients With Cancer
Purpose: Infectious complications in patients with cancer are contributors to hospitalisations, treatment disruption, and mortality. This study aimed to develop machine learning (ML) models for risk stratification of infection‐related hospitalisations (IRHs) among patients with cancer. Methods: Adult patients diagnosed with and treated for lymphoma, multiple myeloma (MM), chronic lymphocytic leukaemia (CLL), colorectal, or lung cancer from 2013 to 2023 in the North Denmark Region were included. Data from national cancer registries and patient health records were used. Serious infections were defined as sepsis, positive blood culture, intensive care admission, or death during hospitalisation. Models were developed using multiple ML methodologies across three different data setups: ML —using data up to hospital admission; ML 48 — using data up to 48 h after admission; and ML 48,reduced —a pruned version of ML 48 with fewer features. Data were split into training, validation and held‐out test sets. Results: Among 9874 patients 2661 IRHs occurred, including 691 (21.5%) serious infections. On the held‐out test set, ML achieved the best performance (ROC‐AUC: 0.79) versus ML 48 (0.76), with ML 48 having higher specificity at matched sensitivity (50.9% vs 44.9% for sensitivity ~85%). ML 48,reduced reached a ROC‐AUC of 0.75 (specificity 46.5% at sensitivity ~85%). Subgroup analysis showed the best performance for lymphoma (ROC‐AUC: 0.86) and lowest for lung cancer (ROC‐AUC: 0.73). The model performed better in patients with ECOG score 0–1 compared to ≥ 2 (ROC‐AUC 0.88 vs. 0.77). SHAP analysis identified previous blood cultures, carbamide, and neutrophil counts as key predictive features across all models. Conclusion: The ML models showed moderate performance in risk stratification of IRHs in patients with cancer. Such models could provide decision support in hospitalisation decisions and discharge.
- Open Access
- 19 June 2026
- Published by Wiley in European Journal of Haematology
- Vol. 117 (4) , 1061-1071
Risk Factors of Blood Culture Positivity and Mortality in Bloodstream Infections: A Population‐Based Study of Patients With B‐Cell Malignancies
Background: Bloodstream infections remain a major cause of mortality among cancer patients. Pathogen surveillance and risk stratification tools are key to improved management. This study describes trends in causative pathogens and explores risk factors for positive blood cultures (PBCs) and post‐infection mortality in cancer patients. Methods: We performed a cohort study of patients diagnosed with a B‐cell malignancy in the North Denmark Region between 2013 and 2022, linking cancer registry data to health records. Causative pathogens were described. LASSO regression was used to select variables associated with PBCs and post‐infection mortality. Odds ratios were estimated by univariate analyses. Results: Among 1592 patients with B‐cell malignancies, 341 patients (21.4%) had at least one PBC, yielding 484 PBCs. We observed a shift from Gram‐positive to Gram‐negative dominance, driven by an increased frequency of the most common pathogen Escherichia coli (24.8%). Patients with neutropenia, elevated CRP, sepsis, hypotension, or kidney failure had a higher risk of a PBC. Advanced age, kidney failure, and sepsis were all associated with 30‐ and 90‐day post‐infection mortality. Conclusion: Routinely available parameters obtained at time of admission were associated with blood culture positivity and mortality. Our findings support the rationale for developing predictive models to individualise infection management.
- 19 June 2026
- Published by Wiley in European Journal of Haematology
- Vol. 117 (4) , 1061-1071
Open AccessRisk Factors of Blood Culture Positivity and Mortality in Bloodstream Infections: A Population‐Based Study of Patients With B‐Cell Malignancies
Background: Bloodstream infections remain a major cause of mortality among cancer patients. Pathogen surveillance and risk stratification tools are key to improved management. This study describes trends in causative pathogens and explores risk factors for positive blood cultures (PBCs) and post‐infection mortality in cancer patients. Methods: We performed a cohort study of patients diagnosed with a B‐cell malignancy in the North Denmark Region between 2013 and 2022, linking cancer registry data to health records. Causative pathogens were described. LASSO regression was used to select variables associated with PBCs and post‐infection mortality. Odds ratios were estimated by univariate analyses. Results: Among 1592 patients with B‐cell malignancies, 341 patients (21.4%) had at least one PBC, yielding 484 PBCs. We observed a shift from Gram‐positive to Gram‐negative dominance, driven by an increased frequency of the most common pathogen Escherichia coli (24.8%). Patients with neutropenia, elevated CRP, sepsis, hypotension, or kidney failure had a higher risk of a PBC. Advanced age, kidney failure, and sepsis were all associated with 30‐ and 90‐day post‐infection mortality. Conclusion: Routinely available parameters obtained at time of admission were associated with blood culture positivity and mortality. Our findings support the rationale for developing predictive models to individualise infection management.
- 12 May 2026
- Published by Wiley in British Journal of Haematology
- Vol. 209 (1) , 291-295
Population‐based survival after mantle cell lymphoma in Norway and Denmark: Limited impact of earlier BTKi access
- 12 May 2026
- Published by Wiley in British Journal of Haematology
- Vol. 209 (1) , 291-295
Population‐based survival after mantle cell lymphoma in Norway and Denmark: Limited impact of earlier BTKi access
- Open Access
- 17 April 2026
- Published by Wiley in British Journal of Haematology
- Vol. 208 (6) , 2104-2112
Improved survival with fludarabine‐based therapies in mixed phenotype acute leukaemia: A population‐based study using the WHO 2022 classification
Summary: Mixed phenotype acute leukaemia (MPAL) is a rare subtype of acute leukaemia possessing significant therapeutic challenges, as no standardized, evidence‐based treatment regimen has been defined. In this nationwide study, we aimed to assess the effect of an acute lymphoid leukaemia (ALL)‐like regimen; an acute myeloid leukaemia (AML)‐like regimen; and a hybrid regimen on complete remission (CR), overall survival (OS) and event‐free survival (EFS). Patients were identified through the Danish National Pathology Registry and validated according to the 2022 WHO classification. OS was estimated using the Kaplan–Meier estimator, and differences in OS were assessed with the log‐rank test. Inverse probability weighting was used to balance compared groups with respect to age and calendar year. Among the 43 intensively treated WHO 2022 MPAL patients, 25 (58.1%) received ALL regimens, 10 (23.3%) received AML regimens and 8 (18.6%) received hybrid regimens. CR rates by treatment regimen were highest for hybrid regimen (87.5% (95% confidence interval (CI): 47.3%–99.7)), followed by ALL regimen (72% (95% CI: 50.6%–87.5%)) and AML regimen (40% (95% CI: 12.2%–73.8%)). Treatment with hybrid regimens consisting of a fludarabine‐based approach (fludarabine, cytarabine, G‐CSF and idarubicin [FLAG‐IDA], fludarabine, cytarabine, G‐CSF and mitoxantrone [Mito‐FLAG] or fludarabine, etoposide, G‐CSF, mitoxantrone and cytarabine [FLEGMA]) was associated with improved OS and EFS compared to a classic daunorubicin–cytarabine AML regimen ( p = 0.02 and p = 0.002 respectively).
- 17 April 2026
- Published by Wiley in British Journal of Haematology
- Vol. 208 (6) , 2104-2112
Open AccessImproved survival with fludarabine‐based therapies in mixed phenotype acute leukaemia: A population‐based study using the WHO 2022 classification
Summary: Mixed phenotype acute leukaemia (MPAL) is a rare subtype of acute leukaemia possessing significant therapeutic challenges, as no standardized, evidence‐based treatment regimen has been defined. In this nationwide study, we aimed to assess the effect of an acute lymphoid leukaemia (ALL)‐like regimen; an acute myeloid leukaemia (AML)‐like regimen; and a hybrid regimen on complete remission (CR), overall survival (OS) and event‐free survival (EFS). Patients were identified through the Danish National Pathology Registry and validated according to the 2022 WHO classification. OS was estimated using the Kaplan–Meier estimator, and differences in OS were assessed with the log‐rank test. Inverse probability weighting was used to balance compared groups with respect to age and calendar year. Among the 43 intensively treated WHO 2022 MPAL patients, 25 (58.1%) received ALL regimens, 10 (23.3%) received AML regimens and 8 (18.6%) received hybrid regimens. CR rates by treatment regimen were highest for hybrid regimen (87.5% (95% confidence interval (CI): 47.3%–99.7)), followed by ALL regimen (72% (95% CI: 50.6%–87.5%)) and AML regimen (40% (95% CI: 12.2%–73.8%)). Treatment with hybrid regimens consisting of a fludarabine‐based approach (fludarabine, cytarabine, G‐CSF and idarubicin [FLAG‐IDA], fludarabine, cytarabine, G‐CSF and mitoxantrone [Mito‐FLAG] or fludarabine, etoposide, G‐CSF, mitoxantrone and cytarabine [FLEGMA]) was associated with improved OS and EFS compared to a classic daunorubicin–cytarabine AML regimen ( p = 0.02 and p = 0.002 respectively).
- Open Access
- 19 November 2025
- Published by Springer Nature in Blood Cancer Journal
- Vol. 15 (1) , 199
Opportunities and risks associated with external comparator cohorts in clinical drug development in hematology
- 19 November 2025
- Published by Springer Nature in Blood Cancer Journal
- Vol. 15 (1) , 199
Open AccessOpportunities and risks associated with external comparator cohorts in clinical drug development in hematology
- Open Access
- 22 October 2025
- Published by Springer Nature in BMC Medical Research Methodology
- Vol. 25 (1) , 1-10
A-calibration: assessment of prediction models for survival data under censoring
Background: Evaluating the performance of predictive models for survival is essential before they can be trusted for real-world applications and decision making. While good measures such as the C-index are available for model discrimination, the toolbox for model calibration is much more limited in the time-to-event setting. The method of D-calibration was therefore an important contribution that yields a single numeric value for calibration across the available follow-up time. D-calibration consists of performing a Pearson’s goodness-of-fit test on transformed survival times. Censored survival times are handled using an imputation approach which however tends to yield a conservative test and loss of power. Methods: In this paper, we introduce A-calibration based on Akritas’s goodness-of-fit test which is designed specifically for censored time-to-event data. Through theoretical arguments, simulations, and a case study, we compare A- and D-calibration as measures of calibration. In the simulation study, the power of each test to reject a false null-hypothesis was assessed for varying censoring mechanisms (memoryless, uniform and zero censoring), censoring rates, and parameter values of the predictive model considered. Results: The simulation study demonstrated that A-calibration had similar or superior power to D-calibration in all considered cases, and that D-calibration, unlike A-calibration, was particularly sensitive to censoring. Conclusions: Advantages of A-calibration compared to D-calibration have been demonstrated through theoretical considerations, a simulation study, and a case study, while no disadvantages relative to D-calibration were identified.
- 22 October 2025
- Published by Springer Nature in BMC Medical Research Methodology
- Vol. 25 (1) , 1-10
Open AccessA-calibration: assessment of prediction models for survival data under censoring
Background: Evaluating the performance of predictive models for survival is essential before they can be trusted for real-world applications and decision making. While good measures such as the C-index are available for model discrimination, the toolbox for model calibration is much more limited in the time-to-event setting. The method of D-calibration was therefore an important contribution that yields a single numeric value for calibration across the available follow-up time. D-calibration consists of performing a Pearson’s goodness-of-fit test on transformed survival times. Censored survival times are handled using an imputation approach which however tends to yield a conservative test and loss of power. Methods: In this paper, we introduce A-calibration based on Akritas’s goodness-of-fit test which is designed specifically for censored time-to-event data. Through theoretical arguments, simulations, and a case study, we compare A- and D-calibration as measures of calibration. In the simulation study, the power of each test to reject a false null-hypothesis was assessed for varying censoring mechanisms (memoryless, uniform and zero censoring), censoring rates, and parameter values of the predictive model considered. Results: The simulation study demonstrated that A-calibration had similar or superior power to D-calibration in all considered cases, and that D-calibration, unlike A-calibration, was particularly sensitive to censoring. Conclusions: Advantages of A-calibration compared to D-calibration have been demonstrated through theoretical considerations, a simulation study, and a case study, while no disadvantages relative to D-calibration were identified.
- Open Access
- 3 September 2025
- Published by Elsevier in Clinical Lymphoma Myeloma and Leukemia
- Vol. 26 (1) , e62-e76
International Prognostic Models as Tools for Selection of Higher-risk Trial-eligible Patients With Diffuse Large B-Cell lymphoma
- 3 September 2025
- Published by Elsevier in Clinical Lymphoma Myeloma and Leukemia
- Vol. 26 (1) , e62-e76
Open AccessInternational Prognostic Models as Tools for Selection of Higher-risk Trial-eligible Patients With Diffuse Large B-Cell lymphoma