A PRISMA-guided systematic review of musculoskeletal modelling approaches in lower-limb cycling biomechanics

This PRISMA-guided systematic review of 28 studies on lower-limb cycling musculoskeletal modelling reveals significant inconsistencies in reporting, validation, and participant diversity, highlighting the urgent need for standardized practices, transparent open science, and more rigorous, inclusive research to improve reproducibility and clinical relevance.

The Gist

Imagine you are trying to figure out exactly how a bicycle works, but you can't take it apart, and you can't see the gears turning inside the rider's legs. You can see the wheels spinning and the rider moving, but the "engine" (the muscles and bones) is hidden under the skin.

This paper is a systematic review (a giant, organized search) of 28 different studies that tried to build digital twins of cyclists to see what's happening inside those hidden engines.

Here is the breakdown of what they found, using some everyday analogies:

1. The Goal: Looking Under the Hood

Scientists use computer models to simulate cycling. Think of these models like flight simulators for cyclists.

• Real life: You can measure how fast a bike goes or how hard a rider pushes the pedal.
• The Simulation: You can see the invisible stuff: how much force a specific muscle is pulling, how much pressure is on a knee joint, or how the brain tells the leg to move.
• The Problem: The authors wanted to know: Are these flight simulators accurate? Are they built the same way by everyone? And are they being used to solve real problems?

2. What They Found: The "Flight Simulator" Chaos

After looking at 28 studies, the authors found that the field is a bit like a group of people trying to build a car engine, but everyone is using different blueprints, different tools, and different parts.

• The "Who" Problem (The Male Bias):
  Almost all the digital cyclists in these studies were young, fit men.
  • Analogy: Imagine if every car safety test was done only with a 25-year-old male crash-test dummy. You wouldn't know if the airbags work for a grandmother or a child. The models are great for young men, but we don't know if they work for women, older people, or people with injuries.
• The "Blueprint" Problem (Inconsistent Models):
  Some scientists built a simple 2D stick-figure bike. Others built a complex 3D robot with 286 muscles.
  • Analogy: It's like one person drawing a map of a city with just the main highways, while another draws every single alleyway and tree. Because they didn't agree on the rules, it's hard to compare their results.
• The "Black Box" Problem (Secret Recipes):
  Most scientists built their models but didn't share the code or the blueprints.
  • Analogy: Imagine a chef making a delicious soup but refusing to tell anyone the recipe or let them taste it. If you can't see how they made it, you can't check if it's good or if you can make it yourself. Only 4 out of 28 studies shared their "recipes" (code/models).

3. How They Tested It: The "Guess and Check" Game

The researchers looked at how these models were validated (checked for accuracy).

• The Missing Data: Most studies checked if the movement looked right (did the leg bend the right way?), but very few checked if the internal forces were right (did the muscle actually pull that hard?).
  • Analogy: It's like checking if a car drives down the street correctly, but never opening the hood to see if the engine is actually running or if the tires are inflated.
• The "Easy Button": Most models used a simple rule: "The body tries to use the least amount of energy possible."
  • Analogy: It's like assuming a driver always takes the shortest, easiest route home. But sometimes, people take detours, speed up, or get tired. The models are too simple to capture the messy reality of human movement.

4. The Good News

Despite the messiness, these simulations are doing great things:

• Bike Fitting: They help figure out the perfect seat height and handlebar position to stop knee pain.
• Rehabilitation: They help design better ways to help people with spinal cord injuries or muscle weakness learn to pedal again.
• Performance: They help athletes understand how to pedal more efficiently.

5. The Bottom Line: What Needs to Change?

The authors conclude that while these "digital cyclists" are powerful tools, the field needs to grow up.

• Diversity: We need models of women, older people, and people with different body types, not just young men.
• Transparency: Scientists need to share their code and blueprints so others can check their work.
• Realism: The models need to be tested against real internal data, not just guesses.

In short: We have built some amazing digital bicycles, but right now, they are mostly built for a very specific type of rider, using secret recipes, and we aren't 100% sure they work for everyone. The goal is to make these simulations open, diverse, and accurate enough to help everyone ride better and safer.

Technical Summary

1. Problem Statement

Cycling is a critical activity for sports performance, rehabilitation, and public health. While experimental methods (e.g., motion capture, force plates, EMG) provide data on observable kinematics and kinetics, they cannot directly measure internal biomechanical variables such as muscle forces, joint loading, and metabolic cost. Musculoskeletal (MSK) simulations are increasingly used to estimate these internal variables and explore "what-if" scenarios.

However, the field suffers from a lack of standardization. It remains unclear how consistently MSK simulations are applied, validated, and reported in cycling biomechanics. There is a need to synthesize current practices regarding:

• Modelling choices: Anatomical scope, dimensionality, and control strategies.
• Validation: How models are verified against experimental or literature data.
• Reproducibility: The availability of code, models, and detailed reporting.
• Demographics: The representativeness of participant cohorts (e.g., sex, age, clinical status).

2. Methodology

The study followed the PRISMA 2020 guidelines for systematic reviews.

• Search Strategy: A comprehensive search was conducted on August 1, 2024, across four databases: Scopus, PubMed, IEEE Xplore, and Web of Science.
• Timeframe: Studies published between January 2010 and July 2024.
• Inclusion Criteria:
  • Peer-reviewed English-language journal articles.
  • Application of MSK simulations to lower-limb cycling.
  • Focus on internal biomechanics (muscle forces, joint loading).
  • Applications in performance, rehabilitation, or injury.
• Exclusion Criteria:
  • Non-peer-reviewed work (theses, reviews).
  • Studies without MSK models (e.g., inverse kinematics only, EMG-only).
  • Non-cycling tasks or upper-body focus.
• Data Extraction: Reviewers independently extracted data on study aims, participant characteristics, modelling frameworks (software, DOF, muscle count), simulation methods (optimization, control), cost functions, and validation strategies.
• Final Sample: 28 studies met all inclusion criteria out of an initial pool of 1,006 records.

3. Key Contributions

This review provides the first systematic synthesis specifically focused on musculoskeletal simulation in cycling, distinguishing itself from previous reviews that focused on experimental pedalling biomechanics or general predictive modelling.

• Taxonomy of Practices: It categorizes the diverse methodological landscape, identifying trends in model complexity, control strategies, and validation approaches.
• Identification of Gaps: It highlights critical deficiencies in reporting transparency, participant diversity, and validation rigor.
• Reproducibility Assessment: It quantifies the lack of open-source code and model sharing, despite the prevalence of open-source software platforms.
• Demographic Analysis: It exposes a severe gender and clinical bias in the current literature.

4. Key Results

A. Study Aims and Applications

The 28 studies primarily focused on four areas:

1. Equipment/Configuration Optimization (53.6%): Analyzing saddle position, crank length, Q-factor, and bike geometry.
2. Neuromuscular Coordination (64.3%): Investigating muscle activation strategies and contractile behavior.
3. Rehabilitation/FES (32.1%): Optimizing Functional Electrical Stimulation (FES) and assistive devices for clinical populations.
4. Model Development: Creating or validating computational frameworks.

B. Participant Demographics

• Total Participants: 272 individuals across all studies.
• Sex Bias: Overwhelmingly male. Of the 246 participants with reported sex, 72.8% were male and only 17.6% were female.
• Clinical Representation: Minimal. Only 4 participants had spinal cord injuries (SCI) and 13 had knee osteoarthritis.
• Age/Ability: Predominantly young adults (24–30 years) and recreational/trained cyclists. Professional cyclists were rare (5.1%).

C. Modelling Frameworks

• Anatomical Scope: Highly variable. Ranged from single-leg models (14.3%) to full-body models with arms (14.3%) and without arms (42.9%).
• Dimensionality: 60.7% were 3D models; 39.3% were 2D.
• Complexity: Muscle-tendon unit (MTU) counts varied wildly from 4 to 286. Degrees of Freedom (DOF) were often unreported (approx. 50% of studies) or ambiguous regarding whether they referred to the base model or the closed-chain cycling system.
• Software: OpenSim was dominant (57.1%), followed by custom software (25.0%) and AnyBody (14.3%).
• Reproducibility: Only 14.3% (4/28) of studies made models or code publicly available.

D. Experimental Data and Validation

• Data Collection: 53.6% collected new data; 28.6% used existing datasets; 17.9% were simulation-only.
• Inputs: All studies using data reported kinematics. Kinetic data (forces/torques) were reported in 73.9%. Neuromuscular data (EMG) were rare (26.1%), and physiological/metabolic data were extremely rare (4.3%).
• Validation: 75% of studies performed some comparison. However, validation was often superficial:
  • Most compared kinematics or pedal forces.
  • Only 66.7% validated muscle activation timing.
  • No study validated simulated joint loads or resistance profiles against experimental measurements.
  • Many relied on literature values rather than direct experimental validation.

E. Computational Approaches

• Methods: Static Optimization (SO) was most common (32.1%), followed by Forward Dynamics (FD) and Optimal Control (OC) (17.9% each).
• Cost Functions: Among studies with explicit cost functions, 77.3% minimized muscle activity/effort. Very few incorporated metabolic energy, fatigue, or multi-objective formulations.
• Sensitivity Analysis: While 27/28 studies varied parameters (e.g., geometry, workload), these were mostly exploratory rather than formal sensitivity analyses. Uncertainty quantification was absent.

5. Significance and Future Directions

Significance:
The review concludes that while MSK simulations in cycling are technically diverse, the field is methodologically inconsistent and lacks reproducibility. The heavy reliance on young male cohorts limits the generalizability of findings to women, older adults, and clinical populations. Furthermore, the lack of rigorous validation (especially for internal loads) and the scarcity of open code hinder the translation of these models into clinical practice or equipment design.

Recommendations for Future Research:

1. Demographic Inclusivity: Future studies must include balanced sex representation and diverse clinical populations (e.g., SCI, OA, elderly).
2. Standardized Reporting: Authors must explicitly report model details (DOF, muscle lists, scaling procedures) and distinguish between base models and cycling-specific adaptations.
3. Enhanced Validation: Move beyond kinematic validation to include kinetic, EMG, and physiological validation. Internal load predictions require stronger evidence.
4. Open Science: Mandatory sharing of models, code, and data to facilitate reproducibility and verification.
5. Advanced Methodologies: Shift from static optimization toward fully predictive optimal control and real-time estimation pipelines. Incorporate fatigue and metabolic cost into cost functions.
6. Rigorous Sensitivity Analysis: Implement formal uncertainty quantification to understand how model assumptions affect outcomes.

This review serves as a foundational roadmap for transitioning cycling biomechanics from descriptive, inverse analyses toward robust, predictive, and clinically applicable musculoskeletal simulations.