Angelika Eisele-Metzger
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Angelika Eisele-Metzger has a total of 130 co-authors. Here are the top co-authors by number of shared publications:
- Joerg Johannes Meerpohl
University of Freiburg
8 shared publications - Waldemar Siemens
Institute for Evidence in Medicine, Medical Center - University of Freiburg/Medical Faculty - University of Freiburg, Freiburg, Germany
4 shared publications Christian Haverkamp
University of Freiburg
3 shared publicationsDaniela Schagg
Catholic University of Applied Sciences Freiburg, Freiburg, Germany
3 shared publicationsFelix Heilmeyer
University of Freiburg
3 shared publicationsJoerg J. Meerpohl
Institute for Evidence in Medicine, Medical Center, Faculty of Medicine, University of Freiburg, Freiburg, Germany
3 shared publications
Latest Publications
- 14 August 2026
- Published by Georg Thieme Verlag KG in physioscience
Group therapy in physiotherapy
Group therapy is rarely used in outpatient physiotherapy in Germany, although it is permitted according to the “Heilmittelrichtlinie” and would be preferable to individual therapy if equally suitable. Group therapy could help expand treatment capacities, reduce costs, and promote factors such as motivation and adherence. Whether it is similarly effective to individual therapy for diagnoses that are frequently prescribed, however, remains unclear. To provide an overview of the current evidence on group-based physiotherapy for diagnoses with high prescription rates. An overview of reviews was conducted. A systematic literature search was performed in MEDLINE, CINAHL, EBP, PEDro, OT Seeker, and Google Scholar. Systematic reviews investigating group-based physiotherapy for mobility impairments, back pain, and shoulder lesions were included. Screening and data extraction were conducted according to established standards. The methodological quality of the reviews was assessed using AMSTAR 2. The results were synthesized narratively. Fifteen systematic reviews were included, addressing back pain (n = 7) and mobility impairments (n = 8). No review on shoulder lesions was identified. Some of the reviews reported positive effects of group therapy on functional outcomes and pain compared to no treatment or usual care. Compared to individual therapy, no consistent evidence of superiority or inferiority was found. No direct negative effects of group therapy were reported. The interventions and control groups in the included reviews were mostly heterogeneous. The methodological quality of most reviews is low or critically low. Existing evidence on the effectiveness of group-based physiotherapy is limited and heterogeneous. Most included reviews are of low methodological quality; therefore, no reliable conclusions regarding the effectiveness of group therapy can be drawn. In light of current requirements of the Heilmittelrichtlinie and the high relevance for healthcare provision, there is a need for methodologically rigorous and focused systematic reviews on this topic. If these continue to reveal relevant gaps in the evidence, appropriate clinical studies should be initiated.
- 14 August 2026
- Published by Georg Thieme Verlag KG in physioscience
Group therapy in physiotherapy
Group therapy is rarely used in outpatient physiotherapy in Germany, although it is permitted according to the “Heilmittelrichtlinie” and would be preferable to individual therapy if equally suitable. Group therapy could help expand treatment capacities, reduce costs, and promote factors such as motivation and adherence. Whether it is similarly effective to individual therapy for diagnoses that are frequently prescribed, however, remains unclear. To provide an overview of the current evidence on group-based physiotherapy for diagnoses with high prescription rates. An overview of reviews was conducted. A systematic literature search was performed in MEDLINE, CINAHL, EBP, PEDro, OT Seeker, and Google Scholar. Systematic reviews investigating group-based physiotherapy for mobility impairments, back pain, and shoulder lesions were included. Screening and data extraction were conducted according to established standards. The methodological quality of the reviews was assessed using AMSTAR 2. The results were synthesized narratively. Fifteen systematic reviews were included, addressing back pain (n = 7) and mobility impairments (n = 8). No review on shoulder lesions was identified. Some of the reviews reported positive effects of group therapy on functional outcomes and pain compared to no treatment or usual care. Compared to individual therapy, no consistent evidence of superiority or inferiority was found. No direct negative effects of group therapy were reported. The interventions and control groups in the included reviews were mostly heterogeneous. The methodological quality of most reviews is low or critically low. Existing evidence on the effectiveness of group-based physiotherapy is limited and heterogeneous. Most included reviews are of low methodological quality; therefore, no reliable conclusions regarding the effectiveness of group therapy can be drawn. In light of current requirements of the Heilmittelrichtlinie and the high relevance for healthcare provision, there is a need for methodologically rigorous and focused systematic reviews on this topic. If these continue to reveal relevant gaps in the evidence, appropriate clinical studies should be initiated.
- Open Access
- 19 June 2026
- Published by Elsevier in Journal of Clinical Epidemiology
- Vol. 198, 112390
Cochrane Evaluation of (Semi-) Automated Review Methods (CESAR): Protocol for an adaptive platform study within reviews
- 19 June 2026
- Published by Elsevier in Journal of Clinical Epidemiology
- Vol. 198, 112390
Open AccessCochrane Evaluation of (Semi-) Automated Review Methods (CESAR): Protocol for an adaptive platform study within reviews
- Open Access
- 8 June 2026
- Published by Cambridge University Press (CUP) in Research Synthesis Methods
- Vol. 17 (5) , 1-3
Response to: Five methodological considerations for validating LLMs in risk of bias assessment
- 8 June 2026
- Published by Cambridge University Press (CUP) in Research Synthesis Methods
- Vol. 17 (5) , 1-3
Open AccessResponse to: Five methodological considerations for validating LLMs in risk of bias assessment
- Open Access
- 12 March 2025
- Published by Cambridge University Press (CUP) in Research Synthesis Methods
- Vol. 16 (3) , 491-508
Exploring the potential of Claude 2 for risk of bias assessment: Using a large language model to assess randomized controlled trials with RoB 2
Systematic reviews are essential for evidence-based health care, but conducting them is time- and resource-consuming. To date, efforts have been made to accelerate and (semi-)automate various steps of systematic reviews through the use of artificial intelligence (AI) and the emergence of large language models (LLMs) promises further opportunities. One crucial but complex task within systematic review conduct is assessing the risk of bias (RoB) of included studies. Therefore, the aim of this study was to test the LLM Claude 2 for RoB assessment of 100 randomized controlled trials, published in English language from 2013 onwards, using the revised Cochrane risk of bias tool (‘RoB 2’; involving judgements for five specific domains and an overall judgement). We assessed the agreement of RoB judgements by Claude with human judgements published in Cochrane reviews. The observed agreement between Claude and Cochrane authors ranged from 41% for the overall judgement to 71% for domain 4 (‘outcome measurement’). Cohen’s κ was lowest for domain 5 (‘selective reporting’; 0.10 (95% confidence interval (CI): −0.10–0.31)) and highest for domain 3 (‘missing data’; 0.31 (95% CI: 0.10–0.52)), indicating slight to fair agreement. Fair agreement was found for the overall judgement (Cohen’s κ: 0.22 (95% CI: 0.06–0.38)). Sensitivity analyses using alternative prompting techniques or the more recent version Claude 3 did not result in substantial changes. Currently, Claude’s RoB 2 judgements cannot replace human RoB assessment. However, the potential of LLMs to support RoB assessment should be further explored.
- 12 March 2025
- Published by Cambridge University Press (CUP) in Research Synthesis Methods
- Vol. 16 (3) , 491-508
Open AccessExploring the potential of Claude 2 for risk of bias assessment: Using a large language model to assess randomized controlled trials with RoB 2
Systematic reviews are essential for evidence-based health care, but conducting them is time- and resource-consuming. To date, efforts have been made to accelerate and (semi-)automate various steps of systematic reviews through the use of artificial intelligence (AI) and the emergence of large language models (LLMs) promises further opportunities. One crucial but complex task within systematic review conduct is assessing the risk of bias (RoB) of included studies. Therefore, the aim of this study was to test the LLM Claude 2 for RoB assessment of 100 randomized controlled trials, published in English language from 2013 onwards, using the revised Cochrane risk of bias tool (‘RoB 2’; involving judgements for five specific domains and an overall judgement). We assessed the agreement of RoB judgements by Claude with human judgements published in Cochrane reviews. The observed agreement between Claude and Cochrane authors ranged from 41% for the overall judgement to 71% for domain 4 (‘outcome measurement’). Cohen’s κ was lowest for domain 5 (‘selective reporting’; 0.10 (95% confidence interval (CI): −0.10–0.31)) and highest for domain 3 (‘missing data’; 0.31 (95% CI: 0.10–0.52)), indicating slight to fair agreement. Fair agreement was found for the overall judgement (Cohen’s κ: 0.22 (95% CI: 0.06–0.38)). Sensitivity analyses using alternative prompting techniques or the more recent version Claude 3 did not result in substantial changes. Currently, Claude’s RoB 2 judgements cannot replace human RoB assessment. However, the potential of LLMs to support RoB assessment should be further explored.
- Top Cited PapersOpen Access
- 26 February 2025
- Published by Elsevier in Journal of Clinical Epidemiology
- Vol. 181, 111746
Large language models for conducting systematic reviews: on the rise, but not yet ready for use—a scoping review
- 26 February 2025
- Published by Elsevier in Journal of Clinical Epidemiology
- Vol. 181, 111746
Top Cited PapersOpen AccessLarge language models for conducting systematic reviews: on the rise, but not yet ready for use—a scoping review
- Open Access
Large language models for conducting systematic reviews: on the rise, but not yet ready for use – a scoping review
Background Machine learning (ML) promises versatile help in the creation of systematic reviews (SRs). Recently, further developments in the form of large language models (LLMs) and their application in SR conduct attracted attention. Objective To provide an overview of ML and specifically LLM applications in SR conduct in health research. Study design We systematically searched MEDLINE, Web of Science, IEEEXplore, ACM Digital Library, Europe PMC (preprints), Google Scholar, and conducted an additional hand search (last search: 26 February 2024). We included scientific articles in English or German, published from April 2021 onwards, building upon the results of a mapping review with a related research question. Two reviewers independently screened studies for eligibility; after piloting, one reviewer extracted data, checked by another. Results Our database search yielded 8054 hits, and we identified 33 articles from our hand search. Of the 196 included reports, 159 described more traditional ML techniques, 37 focused on LLMs. LLM approaches covered 10 of 13 defined SR steps, most frequently literature search (n=15, 41%), study selection (n=14, 38%), and data extraction (n=11, 30%). The mostly recurring LLM was GPT (n=33, 89%). Validation studies were predominant (n=21, 57%). In half of the studies, authors evaluated LLM use as promising (n=20, 54%), one quarter as neutral (n=9, 24%) and one fifth as non-promising (n=8, 22%). Conclusions Although LLMs show promise in supporting SR creation, fully established or validated applications are often lacking. The rapid increase in research on LLMs for evidence synthesis production highlights their growing relevance. HIGHLIGHTS Machine learning (ML) offers promising support for systematic review (SR) creation. GPT was the most commonly used large language model (LLM) to support SR production. LLM application included 10 of 13 defined SR steps, most often literature search. Validation studies predominated, but fully established LLM applications are rare. LLM research for SR conduct is surging, highlighting the increasing relevance.
Open AccessLarge language models for conducting systematic reviews: on the rise, but not yet ready for use – a scoping review
Background Machine learning (ML) promises versatile help in the creation of systematic reviews (SRs). Recently, further developments in the form of large language models (LLMs) and their application in SR conduct attracted attention. Objective To provide an overview of ML and specifically LLM applications in SR conduct in health research. Study design We systematically searched MEDLINE, Web of Science, IEEEXplore, ACM Digital Library, Europe PMC (preprints), Google Scholar, and conducted an additional hand search (last search: 26 February 2024). We included scientific articles in English or German, published from April 2021 onwards, building upon the results of a mapping review with a related research question. Two reviewers independently screened studies for eligibility; after piloting, one reviewer extracted data, checked by another. Results Our database search yielded 8054 hits, and we identified 33 articles from our hand search. Of the 196 included reports, 159 described more traditional ML techniques, 37 focused on LLMs. LLM approaches covered 10 of 13 defined SR steps, most frequently literature search (n=15, 41%), study selection (n=14, 38%), and data extraction (n=11, 30%). The mostly recurring LLM was GPT (n=33, 89%). Validation studies were predominant (n=21, 57%). In half of the studies, authors evaluated LLM use as promising (n=20, 54%), one quarter as neutral (n=9, 24%) and one fifth as non-promising (n=8, 22%). Conclusions Although LLMs show promise in supporting SR creation, fully established or validated applications are often lacking. The rapid increase in research on LLMs for evidence synthesis production highlights their growing relevance. HIGHLIGHTS Machine learning (ML) offers promising support for systematic review (SR) creation. GPT was the most commonly used large language model (LLM) to support SR production. LLM application included 10 of 13 defined SR steps, most often literature search. Validation studies predominated, but fully established LLM applications are rare. LLM research for SR conduct is surging, highlighting the increasing relevance.
- Open Access
Exploring the potential of Claude 2 for risk of bias assessment: Using a large language model to assess randomized controlled trials with RoB 2
Systematic reviews are essential for evidence based healthcare, but conducting them is time and resource consuming. To date, efforts have been made to accelerate and (semi-) automate various steps of systematic reviews through the use of artificial intelligence and the emergence of large language models (LLMs) promises further opportunities. One crucial but complex task within systematic review conduct is assessing the risk of bias of included studies. Therefore, the aim of this study was to test the LLM Claude 2 for risk of bias assessment of 100 randomized controlled trials using the revised Cochrane risk of bias tool (“RoB 2”; involving judgements for five specific domains and an overall judgement). We assessed the agreement of risk of bias judgements by Claude with human judgements published in Cochrane Reviews. The observed agreement between Claude and Cochrane authors ranged from 41% for the overall judgement to 71% for domain 4 (“outcome measurement”). Cohen’s κ was lowest for domain 5 (“selective reporting”; 0.10 (95% confidence interval (CI): −0.10-0.31)) and highest for domain 3 (“missing data”; 0.31 (95% CI: 0.10-0.52)), indicating slight to fair agreement. Fair agreement was found for the overall judgement (Cohen’s κ: 0.22 (95% CI: 0.06-0.38)). Sensitivity analyses using alternative prompting techniques or the more recent version Claude 3 did not result in substantial changes. Currently, Claude’s RoB 2 judgements cannot replace human risk of bias assessment. However, the potential of LLMs to support risk of bias assessment should be further explored.
Open AccessExploring the potential of Claude 2 for risk of bias assessment: Using a large language model to assess randomized controlled trials with RoB 2
Systematic reviews are essential for evidence based healthcare, but conducting them is time and resource consuming. To date, efforts have been made to accelerate and (semi-) automate various steps of systematic reviews through the use of artificial intelligence and the emergence of large language models (LLMs) promises further opportunities. One crucial but complex task within systematic review conduct is assessing the risk of bias of included studies. Therefore, the aim of this study was to test the LLM Claude 2 for risk of bias assessment of 100 randomized controlled trials using the revised Cochrane risk of bias tool (“RoB 2”; involving judgements for five specific domains and an overall judgement). We assessed the agreement of risk of bias judgements by Claude with human judgements published in Cochrane Reviews. The observed agreement between Claude and Cochrane authors ranged from 41% for the overall judgement to 71% for domain 4 (“outcome measurement”). Cohen’s κ was lowest for domain 5 (“selective reporting”; 0.10 (95% confidence interval (CI): −0.10-0.31)) and highest for domain 3 (“missing data”; 0.31 (95% CI: 0.10-0.52)), indicating slight to fair agreement. Fair agreement was found for the overall judgement (Cohen’s κ: 0.22 (95% CI: 0.06-0.38)). Sensitivity analyses using alternative prompting techniques or the more recent version Claude 3 did not result in substantial changes. Currently, Claude’s RoB 2 judgements cannot replace human risk of bias assessment. However, the potential of LLMs to support risk of bias assessment should be further explored.
- Open Access
- 9 January 2024
- Published by Wiley in Clinical Physiology and Functional Imaging
- Vol. 44 (3) , 187-204
Is musculoskeletal pain associated with increased muscle stiffness? Evidence map and critical appraisal of muscle measurements using shear wave elastography
Introduction and Aims: Approximately 21% of the world's population suffers from musculoskeletal conditions, often associated with sensations of stiff muscles. Targeted therapy requires knowing whether typically involved muscles are objectively stiffer compared to asymptomatic individuals. Muscle stiffness is quantified using ultrasound shear wave elastography (SWE). Publications on SWE‐based comparisons of muscle stiffness between individuals with and without musculoskeletal pain are increasing rapidly. This work reviewed and mapped the existing evidence regarding objectively measured muscle stiffness in musculoskeletal pain conditions and surveyed current methods of applying SWE to measure muscle stiffness.Methods: A systematic search was conducted in PubMed and CINAHL using the keywords “muscle stiffness”, “shear wave elastography”, “pain”, “asymptomatic controls” and synonyms. The search was supplemented by a hand search using Google Scholar. Included articles were critically appraised with the AXIS tool, supplemented by items related to SWE methods. Results were visually mapped and narratively described.Results: Thirty of 137 identified articles were included. High‐quality evidence was missing. The results comprise studies reporting lower stiffness in symptomatic participants, no differences between groups and higher stiffness in symptomatic individuals. Results differed between pain conditions and muscles, and also between studies that examined the same muscle(s) and pathology. The methods of the application of SWE were inconsistent and the reporting was often incomplete.Conclusions: Existing evidence regarding the objective stiffness of muscles in musculoskeletal pain conditions is conflicting. Methodological differences may explain most of the inconsistencies between findings. Methodological standards for SWE measurements of muscles are urgently required.
- 9 January 2024
- Published by Wiley in Clinical Physiology and Functional Imaging
- Vol. 44 (3) , 187-204
Open AccessIs musculoskeletal pain associated with increased muscle stiffness? Evidence map and critical appraisal of muscle measurements using shear wave elastography
Introduction and Aims: Approximately 21% of the world's population suffers from musculoskeletal conditions, often associated with sensations of stiff muscles. Targeted therapy requires knowing whether typically involved muscles are objectively stiffer compared to asymptomatic individuals. Muscle stiffness is quantified using ultrasound shear wave elastography (SWE). Publications on SWE‐based comparisons of muscle stiffness between individuals with and without musculoskeletal pain are increasing rapidly. This work reviewed and mapped the existing evidence regarding objectively measured muscle stiffness in musculoskeletal pain conditions and surveyed current methods of applying SWE to measure muscle stiffness.Methods: A systematic search was conducted in PubMed and CINAHL using the keywords “muscle stiffness”, “shear wave elastography”, “pain”, “asymptomatic controls” and synonyms. The search was supplemented by a hand search using Google Scholar. Included articles were critically appraised with the AXIS tool, supplemented by items related to SWE methods. Results were visually mapped and narratively described.Results: Thirty of 137 identified articles were included. High‐quality evidence was missing. The results comprise studies reporting lower stiffness in symptomatic participants, no differences between groups and higher stiffness in symptomatic individuals. Results differed between pain conditions and muscles, and also between studies that examined the same muscle(s) and pathology. The methods of the application of SWE were inconsistent and the reporting was often incomplete.Conclusions: Existing evidence regarding the objective stiffness of muscles in musculoskeletal pain conditions is conflicting. Methodological differences may explain most of the inconsistencies between findings. Methodological standards for SWE measurements of muscles are urgently required.
- Open Access
- 30 August 2021
- Published by Springer Nature in Systematic Reviews
- Vol. 10 (1) , 1-9
Work-related interventions for preventing back pain—protocol for a systematic review and network meta-analysis
Background: Back pain is a widespread health problem that accounts for substantial disability and high costs. The workplace is considered to critically affect the occurrence and persistence of back pain and therefore offers an important opportunity for preventive interventions. Various work-related intervention strategies including both single- and multicomponent interventions have been developed and evaluated so far. To determine their effectiveness, a method of analysis is needed that particularly meets the challenges of the multidimensionality and diversity of these interventions. This planned systematic review and network meta-analysis aims to compare the effects of different work-related interventions for preventing non-specific back pain in people within a formal employment-related context. Methods: We will search the following databases: CENTRAL, MEDLINE, Web of Science, CINAHL, PsycINFO, PEDro, SPORTDiscus, and Academic Search Premier from their inception onwards, as well as additional sources. Randomized controlled trials (RCTs) and cluster-RCTs will be considered if they (1) include people within a formal employment-related context, (2) include people without back pain or mixed samples (i.e., people with and without back pain), (3) compare one or more work-related preventive intervention(s) to a control condition, and (4) assess non-specific back pain (incidence or/and pain intensity), ability to work (numbers of participants or/and numbers of days absent from work), intervention-related adverse events or/and self-reported satisfaction with the intervention. Random-effects pairwise meta-analyses and frequentist network meta-analyses will be conducted where appropriate. We will calculate summary effect sizes for each comparison of interventions and rank interventions according to their P scores. If feasible, we will conduct additional component network meta-analyses. We plan to conduct subgroup analyses for job exposure, intervention duration, baseline back pain, different localizations of back pain, and gender. Risk of bias will be assessed using RoB 2 and the certainty of the evidence will be rated using the GRADE approach. Discussion: This systematic review aims to identify work-related intervention strategies as well as components within work-related interventions that are effective for preventing back pain. We expect the results to provide guidance for selecting the most promising interventions and foster the purposeful use of resources. Additionally, they may inform the development and implementation of work-related interventions as well as the design of future research in this field. Trial registration: PROSPERO CRD42021232469
- 30 August 2021
- Published by Springer Nature in Systematic Reviews
- Vol. 10 (1) , 1-9
Open AccessWork-related interventions for preventing back pain—protocol for a systematic review and network meta-analysis
Background: Back pain is a widespread health problem that accounts for substantial disability and high costs. The workplace is considered to critically affect the occurrence and persistence of back pain and therefore offers an important opportunity for preventive interventions. Various work-related intervention strategies including both single- and multicomponent interventions have been developed and evaluated so far. To determine their effectiveness, a method of analysis is needed that particularly meets the challenges of the multidimensionality and diversity of these interventions. This planned systematic review and network meta-analysis aims to compare the effects of different work-related interventions for preventing non-specific back pain in people within a formal employment-related context. Methods: We will search the following databases: CENTRAL, MEDLINE, Web of Science, CINAHL, PsycINFO, PEDro, SPORTDiscus, and Academic Search Premier from their inception onwards, as well as additional sources. Randomized controlled trials (RCTs) and cluster-RCTs will be considered if they (1) include people within a formal employment-related context, (2) include people without back pain or mixed samples (i.e., people with and without back pain), (3) compare one or more work-related preventive intervention(s) to a control condition, and (4) assess non-specific back pain (incidence or/and pain intensity), ability to work (numbers of participants or/and numbers of days absent from work), intervention-related adverse events or/and self-reported satisfaction with the intervention. Random-effects pairwise meta-analyses and frequentist network meta-analyses will be conducted where appropriate. We will calculate summary effect sizes for each comparison of interventions and rank interventions according to their P scores. If feasible, we will conduct additional component network meta-analyses. We plan to conduct subgroup analyses for job exposure, intervention duration, baseline back pain, different localizations of back pain, and gender. Risk of bias will be assessed using RoB 2 and the certainty of the evidence will be rated using the GRADE approach. Discussion: This systematic review aims to identify work-related intervention strategies as well as components within work-related interventions that are effective for preventing back pain. We expect the results to provide guidance for selecting the most promising interventions and foster the purposeful use of resources. Additionally, they may inform the development and implementation of work-related interventions as well as the design of future research in this field. Trial registration: PROSPERO CRD42021232469
- 14 January 2021
- Published by Taylor & Francis in Physiotherapy Theory and Practice
- Vol. 38 (10) , 1426-1437
Supporting patient adherence to physical activity and exercise: evaluation of a behavior change counseling training program for physiotherapists
Physiotherapists promote physical activity and exercise, yet patient adherence is often poor. To support patient adherence, behavior change techniques (BCTs) should be applied. These are insufficie...
- 14 January 2021
- Published by Taylor & Francis in Physiotherapy Theory and Practice
- Vol. 38 (10) , 1426-1437
Supporting patient adherence to physical activity and exercise: evaluation of a behavior change counseling training program for physiotherapists
Physiotherapists promote physical activity and exercise, yet patient adherence is often poor. To support patient adherence, behavior change techniques (BCTs) should be applied. These are insufficie...