Details our eyes cannot see: Challenges for the analysis of body position during bicycle fitting

Introduction

Bicycle fitting is an interactive and detailed process to adjust bicycle components to the cyclist’s posture. Primary outcomes of this process are to improve comfort and reduce injury risk (Priego Quesada et al., 2019). In competitive cycling, athletes aim to improve their performance via changes in the body position on the bicycle (Fintelman et al., 2014). Since the 90s, bicycle fitting methods have attempted that the body position on the bicycle is consistent between cyclists. One example is measuring the knee flexion angle to ensure that the saddle is not excessively high or low, which has been considered for a long time a potential strategy to prevent lower extremity injuries (Holmes et al., 1994). However, multiple methods have been utilised to determine the position of a cyclist on the bicycle, i.e., manual goniometers, inclinometers, two-dimensional video footage and three-dimensional motion analysis (Carpes et al., 2009; Fonda et al., 2014; Holliday et al., 2017; Priego Quesada et al., 2017). Various challenges impact the consistency of bicycle fitting, which results in variations of fittings for the same cyclist (Braeckevelt et al., 2019). One of the challenges involves the lack of normative data to determine an optimal position on the bicycle, particularly differentiating those with and without injuries or illnesses (Bini & Flores Bini, 2018; Van Hoof et al., 2012). On top of that, discrepancies between methods and differences in goals of cyclists (e.g., racing vs. commuting) should determine intended angles during bicycle fitting (Bini & Hume, 2016; Holliday et al., 2017; Millour et al., 2019a, Millour, et al., 2019b).

To present some recommendations on the suitability of each method to determine the body position on the bicycle, this editorial paper discusses strengths and limitations of various methods utilised in bicycle fitting. In addition, directions for future studies to increase consistency in bicycle fitting will be proposed.

Bike fitting should focus on the body position: influencing factors

Historically, bicycle fitting has utilised equations based on anthropometrical data with the assumption that the bicycle should be proportional to the cyclists’ body dimensions (De Vey Mestdagh, 1998). Although this assumption is logical, studies proposed that looking at joint angles would be a better approach to ensure that injury risk could be reduced (Holmes et al., 1994) and performance maximised (Peveler, 2008). Since then, the bicycle fitting industry transitioned largely to the assessment of the body position on the bicycle via movement analysis (i.e., kinematics). This transition also triggered a debate between static and dynamic measures as suitable methods to determine the body position on the bicycle. The challenge although has been to determine the ‘ideal’ body position on the bicycle for different cycling disciplines and exercising at different intensities. Finding an ‘ideal’ fitting is difficult because the body position is sensitive to changes in exercise conditions such as intensity (Bini et al., 2019; Holliday et al., 2019), cadence (Bini, Rossato et al., 2010) and fatigue state (Bini, Diefenthaeler et al., 2010). Therefore, when assessing cyclists, the decision in terms of intensity, cadence and test duration is critical. However, there is limited evidence proposing different target angles when cyclists pedal at low vs. high intensity (Holliday et al., 2019), requiring larger cohort studies to develop comprehensive normative data. Therefore, whenever possible, cyclists should be assessed during prolonged tests to analyse the body position during simulated racing or training. This approach though may not be suitable for most clinical and practical settings due to time constraints.

Moreover, bicycle fitting would be different depending on different factors such as discipline (Bini et al., 2014a), gender (Encarnación-Martínez et al., 2021) and training level (García-López et al., 2016). A clear example of differences in bicycle fitting is the greater drop handlebars used by professional cyclists compared to recreational cyclists (García-López et al., 2016). Another example involves the use of bicycles for rehabilitation, which requires further guidelines when patients progress through the various stages of the rehabilitation process (e.g., improvements in the knee range of motion). Although some recommendations exist for knee flexion angles in order to optimise posture on the bicycle (Bini et al., 2011; Ferrer-Roca et al., 2014; Millour et al., 2019a), studies indicated that joint reaction forces during pedalling are not largely sensitive to small changes in the bicycle configuration (i.e., 2–3 cm; Bini, 2021; Bini & Hume, 2014; Menard et al., 2020). Moreover, there are no reference angular data for other joints with scientific evidence, which limits bicycle fitting based on angular kinematics.

Static and dynamic bike fitting: different ways to obtain the same goal?

Static measurements to determine joint angles in cyclists have been based on the assumptions that 1) cyclists remain relatively static on the bicycle and produce movement consistently and 2) angles can be measured using simple tools such as manual goniometers. Research has suggested that a relatively static posture on the bicycle is challenged by increases in exercise intensity (Bini et al., 2012) and the presence of muscle fatigue (Bini et al., 2012; Peveler et al., 2012) as well as cyclists’ level of experience (Chapman et al., 2009). Measurements of angles taken statically and dynamically are often different because movement creates angular momentum, which is not observed in static poses (Bini & Hume, 2016; Holliday et al., 2017; Millour et al., 2019a). Large differences in hip (5 ± 1°), knee (8 ± 2°) and ankle (9 ± 2°) angles were observed between static and dynamic methods at 6 o’clock (Bini & Hume, 2016). Similarities were observed for the 3 o’clock crank position between static and dynamic methods, suggesting that this could be an alternative to the traditional 6 o’clock crank position (Bini & Hume, 2016). However, limited reference data have been produced in this alternative position of the crank cycle (Bini & Hume, 2016; Bini et al., 2014a), which complicates its current application during bicycle fitting assessment.

It is critical to acknowledge that some studies measured the knee angle at a fixed point of the crank cycle (e.g., 3 o’clock and 6 o’clock, Bini, 2020; Bini & Hume, 2016; Holliday et al., 2019), whilst others looked at minimum flexion (Encarnación-Martínez et al., 2021; Millour et al., 2019a), which should result in different angles. However, it is unclear if differences in angles between the 6 o’clock and the minimum flexion are clinically meaningful. This means that further comparison of bicycles fitted using both methods is required to clarify if they lead to different outcomes. Whilst the benefit of looking at a fixed point of the crank cycle is that tracking of the whole movement is not required, the advantage in determining the minimum flexion is that the analysis can be automated. Regardless of the position of the crank, static methods are potentially limited as the central nervous system may not be able to fully replicate the joint position due to differences in proprioceptive input (e.g., lack of angular velocity, Fuentes & Bastian, 2009).

Although static and dynamic methods are different, strategies to overcome these differences have been proposed. One example is to consider that an offset in angles is consistently observed when cyclists move their legs (e.g., ~8° of knee flexion, Bini & Hume, 2016; Millour et al., 2019a; Swart & Holliday, 2019). The most important limitation of this option is the assumption that all cyclists differ by 8°, which is not true because the confidence interval for the differences between methods range from 6.2 to 10° (Bini & Hume, 2016). Therefore, static and dynamic methods do not seem interchangeable (Holliday et al., 2017) and guidelines including consistency in methods are required for each method with validation studies demonstrating which of these methods are more accurate in determining the body position on the bicycle. It is apparent that, at this stage, dynamic methods can be more suitable, particularly if exercise intensity during bicycle fitting can approximate training loads (Holliday et al., 2019). However, future research is needed to produce normative data for ranges of exercise intensities to better inform dynamical bicycle fitting. Another element of concern is that many clinicians and bicycle fitters may be limited to use static measurement tools, which hopefully should not limit their capability to perform bicycle fitting.

Does sagittal plane analysis reflect the true movement of cyclists?

Although cyclists move their segments largely in the sagittal plane, rotations in the transverse plane and mediolateral movements are a part of the pedalling motion (Bini et al., 2016; Carpes et al., 2009). This element has implications in determining the influence of hip adductions/abductions and inter-pedal distance in the frontal plane (i.e., Q-factor). Furthermore, the methods to determine joint angles are different when data are obtained using two- or three-dimensional data (Fonda et al., 2014; Umberger & Martin, 2001). One example is the assumption that anatomical landmarks provide an accurate representation of joint centres, which is normally a key principle of two-dimensional analysis. Also, ‘out-of-plane’ movement (i.e., mediolateral) is observed in cycling but cannot be captured by two-dimensional analysis. These lead to differences between three- and two-dimensional kinematics ranging between 2° and 10° (Fonda et al., 2014; Umberger & Martin, 2001). This difference adds an important confounder to the use of simple methods such as sagittal plane video recording during bicycle fitting. Therefore, when assessing two-dimensional derived joint angles, it is assumed that limitations will exist in relation to the gold-standard assessment (i.e., three-dimensional). In terms of bicycle fitting normative data, it is also unclear what are the implications when using two- vs. three-dimensional data. It is also important to note that commercially available devices need extended validity and reliability analysis before informing bicycle fitting, but preliminary results are promising (FitzGibbon et al., 2017; Scoz et al., 2021).

One additional difference between two- and three-dimensional analysis involves the measurement of absolute vs. relative angles. When conducting three-dimensional analysis, a static trial is normally used to determine individualised angles with angular data presented relative to the individual’s standing posture. For two-dimensional analysis in cycling, this principle has been initially explored by Priego Quesada et al. (2017), who observed a mean bias of ~11° [8–13° confidence interval for differences] in knee flexion when standing posture was taken into account. This finding indicates that cyclists’ posture should be considered when determining joint angles on the bicycle, regardless of the use of two- or three-dimensional data. However, evidence in this topic is limited, particularly when proposing ranges of joint angles to bicycle fitting or when predictive equations are developed to determine bicycle fitting based on joint angles. Moreover, most studies did not state clearly if they used relative or absolute angles in their calculations (Bini et al., 2014a, 2014b; Peveler et al., 2012), making it difficult to compare results obtained from relative angles (Priego Quesada et al., 2017). A future research consensus is necessary to improve obtaining scientific evidence from the results of different studies, including adherence to ISB recommendations for kinematic modelling (Wu & Cavanagh, 1995).

On top of traditional video-based methods, new technology involving inertial measurement units (i.e., IMUs, Bini & Hume, 2020) and marker-less methods (Bini et al., 2021) has been introduced to assess cyclists’ movement and body position on the bicycle. These methods require further validation in terms of key outcomes utilised for bicycle fitting to expand the range of normative data. To provide a summary of validity and reliability on the various methods to determine the body position on the bicycle, Table 1 highlights these elements. It is important to note that validity and reliability (intra- and inter-sessions) should be further explored for each joint from the cited studies when utilising each method for assessment of posture on the bicycle. For example, in the case of the intra-session reliability of 3D optoelectronic MOCAP, the only study available obtained the 3D coordinates using three cameras to calculate angles only in the sagittal plane (Encarnación-Martínez et al., 2021), which suggests further work needed in this element.

Table 1. Summary of main methods used to determine the body position on the bicycle with available data on their validity and intra- and inter-session reliability. Interpretation of validity and reliability is provided as stated by authors from each cited study. Different joint angles are assessed in each study.

Methods	Validity	Intra-session reliability	Inter-session reliability
Manual goniometer	Poor to moderate (ICC = 0.33–0.73)^1	2.7–4.8° (low typical error)^1	Moderate to good^1
Inclinometer	Poor to moderate (ICC = 0.25–0.72)^1	2.6–4.5° (low typical error)^1	Moderate to good^1
Electronic goniometer	Underestimated knee flexion angles^2	Good^2	Not tested
2D video analysis with markers	~6-10° of difference to 3D optoelectronic MOCAP^3	No significant difference between trials (1.7 ± 8.4°)^2	Moderate to excellent reliability (ICC = 0.77–0.98)^4
2D video marker-less	3–12° in relation to 2D video analysis with markers (significant difference)^5	Not tested	Not tested
3D optoelectronic MOCAP	Considered gold standard	Good (1.1 ± 4.1° of difference) (ICC = 0.97–0.99)^6	Moderate to good^1
3D IMU motion analysis	ICC = 0.33–1.00 for hip, knee and ankle^7	Good to excellent (ICC = 0.84–0.99)^8	Poor to excellent (ICC = 0.16–0.97)^8
3D bicycle fitting specific systems	Moderate (ICC = 0.52 for knee flexion)^9, and high (ICC = 0.70–0.90)^10	Not tested	Not tested

¹Holliday et al. (2017), without ICC data for inter-session; ²Fonda et al. (2014), ³Umberger and Martin (2001), ⁴Burnie et al. (2020), ⁵Bini et al. (2021), ⁶Encarnación-Martínez et al. (2021), ⁷Poitras et al. (2019), ⁸Bini and Hume (2020), ⁹FitzGibbon et al. (2017) and ¹⁰Scoz et al. (2021).

How to fit it all? Summary and some recommendations

Bicycle fitting is popular and largely beneficial to a range of cyclists, but consistency in methods to assess posture of cyclists on their bicycles is lacking. This is critical because decisions on if and how to change bicycle fitting are largely dependent on the validity and reliability of assessment outcomes. Therefore, we propose that

1. Whenever bicycle fitters are limited in terms of technology, static methods can be used to provide initial data (Swart & Holliday, 2019), taking into account cyclists with inter-limb asymmetry in length (i.e., both limbs should be assessed).

2. If video-recording resources are accessible (i.e., two-dimensional), bicycle fitting should involve movement analysis and measurements of angles, in line with existing data (Bini & Hume, 2016; Millour et al., 2019a; Swart & Holliday, 2019).

3. Posture of cyclists out of the bicycle should be considered in the fitting with the purpose of individualising the analysis of the body position on the bicycle (Holliday & Swart, 2021).

4. For dynamic analysis, cyclists should exercise at intensity and cadence similar to their training profile (Bini et al., 2012; Holliday et al., 2019; Peveler et al., 2012).

5. Whenever available, three-dimensional motion analysis is preferred in order to increase validity of data obtained from different planes of motion (Fonda et al., 2014; Holliday et al., 2017; Umberger & Martin, 2001).

6. The implementation of novel technology (i.e., IMUs and marker-less) for bike fit requires caution particularly when small changes in the body position are sought.

7. Future research should inform a consensus on key elements for assessment of the body position on the bicycle and bicycle fitting, including data on joint moments and power for a range of populations (e.g., injured and those with chronic illness).

Disclosure statement

No potential conflict of interest was reported by the author(s).