The Medial-Lateral Pedal Force Component Correlates with Q-Angle during Steady-State Cycling at Different Workloads and Cadences

: Leg movement during cycling is constrained to the pedal/crank path and predominantly occurs in a sagittal plane. Medial-lateral force (FML) applied to the pedals is considered as a waste and does not contribute to the pedaling. The aim of this study was to examine the changes in FML across different cadences and workloads, and to examine the correlation with lateral knee movement (Q-angle). Twenty-two cyclists completed six trials at three workloads (2, 2.5 and 3 W/kg) and three cadences (75, 85, 95 rpm). Forces were recorded from the force pedal mounted to the left side. Absolute and normalized (to the peak total force) FML were compared across conditions and cross-correlation with Q-angle was calculated. Absolute FML was signiﬁcantly different across cadences and workloads ( p < 0.05) with higher absolute FML at higher cadence. There was no signiﬁcant difference in normalized FML across the three cadences. There was a signiﬁcant decrease in normalized FML (~10 N) at higher workloads. Statistically signiﬁcant correlations were found between the FML and Q-angle (R = 0.70–0.77). The results demonstrate the link between the FML and Q-angle in healthy pain-free cyclists during stationary cycling. It has also been observed that smaller normalized magnitude of the FML is present when the force effectiveness is increased.


Introduction
Lower extremity movement during seated cycling is a cyclic motion that predominantly occurs in the sagittal plane. It is also constrained to the pedal/crank path around the bottom bracket. In seated conditions, these limit cyclists to optimize the pedaling technique by applying the force in two directions, vertically and in an anterior-posterior direction. Any medial-lateral force (FML) applied to the pedals during cycling in a seated position is considered as a waste and does not contribute to the pedaling. For these reasons it has often not been measured nor quantified [1,2].
Even though several studies used instrumented pedals that allowed 3-dimensional force measurements [1], some research assumed that the leg movement during cycling is confined to the sagittal plane [3,4]. However, medial-lateral movement of the knee is widely present during cycling and is often linked to overuse injuries [5,6]. As demonstrated by Ruby et al. [7] this movement can be caused by the FML on the pedals and varus/valgus moments. This was further demonstrated by Ericson et al. [6] who found changes in varus and valgus knee loads at different knee positions (knee adduction/abduction) during stationary cycling. It is worth noting, however, that FML at the pedals while cycling at a workload of 225 W and cadence of 90 revolutions per minute (rpm), peaked at approximately 40 N with supporting varus/valgus moment of 1.35 N·m [7]. Bailey et al. [5] showed a close link between previously injured cyclists and knee kinematics in the frontal plane.
They demonstrated that symptomatic cyclists move knees more medially compared to non-symptomatic cyclists, which could consequently result in a higher FML. These studies showed the importance of the multi-dimensional kinematics and dynamics in joint loads assessments. However, it remains unknown how different pedaling conditions (cadence, workload, etc.) affect the magnitude of the FML.
Several studies examining the effectiveness of pedaling and the effects of mechanical constraints (e.g., cadence, seat position, etc.) used only vertical and anterior-posterior components for force computations and neglected or did not measure the FML [2,8]. However, to accurately assess the effect of a mechanical constraint, particularly when the constraint affects the inertial moments of the lower extremity (e.g., cadence, workload, etc.), the FML could play a role in the force parameters calculations [9]. To date, there is little research separately assessing the amplitude of the FML [7,10] and no research that has systematically evaluated the changes in the FML component across different pedaling conditions.
One of the most common measures of knee movement in frontal plane is Q-angle, defined as the angle between the tibia and femur in frontal plane. Pedal FML component could be generated due to excessive range of motion in Q-angle. Although moderate correlations (r 2 = 0.52-0.74) were found between some anatomical characteristics of the foot, Q-angle and lateral knee moments, there is no clear evidence of a cause-and-effect relationship between the FML and Q-angle [11]. Furthermore, the lack of research interest in the FML component means it is unclear how the magnitude and temporal characteristics of the FML change with cadence or workload, or if it correlates with the Q-angle.
It has been previously shown that the risk of a knee injury (e.g., anterior knee pain or patellar tendinitis) is increased if the lower leg is in an abducted position when a knee extensor moment is generated [5]. An abducted lower leg position places the knee medially in respect to the ankle, which can disrupt the knee extensor mechanism [5]. Therefore, to minimize the risk of a knee injury occurrence, one should strive towards smaller FML and less adducted lower leg.
The aim of this study was to examine (1) how FML changes at different cadences and workloads and (2) if frontal plane kinematics of the lower limb will correlate with the temporal characteristics of the FML. We hypothesized that (1) FML will be significantly lower at pedaling conditions that result in higher index of effectiveness and (2) Q-angle of the knee will correlate with the FML.

Participants
The required sample size was calculated with GPower software (version 3.0.10) using a priori power analysis for repeated measures ANOVA. We used the absolute FML data from the pilot trials that included 4 participants. Based on the observed within-subject variation (Cohen's f = 0.2), the alpha level of 0.05 and desired statistical power of 90%, it was determined that a minimum of 19 participants is required. We recruited 16 male and 6 female cyclists ([mean ± SD] age 24 ± 4.8 y, body mass 74.3 ± 11.8 kg, and body height 177.5 ± 5.9 cm) using the means of advertisement on targeted social media (recreational road riders with no history of muscular-skeletal injuries in the past 5 years, riding between 3000 and 8000 km a year). Before the experiment, each participant signed an informed consent document, which was approved by the Slovenian national medical ethics committee (0120-313/2017-3 KME 37/06/17).

Measures and Procedures
After a 10 min warm up (easy cycling at 100 W, preferred cadence), each participant completed six testing trials using three different workloads and three different cadences. Three workloads (LOW, MID, HIGH) were set based on the participant's body mass and were set to 2, 2.5 and 3 W/kg, respectively. These three conditions were performed at a preferred cadence. The three cadence conditions (C75, C85 and C95) were performed at a constant workload of 2 W/kg but with a dedicated cadence of 75, 85 and 95 rpm, respectively. After gradually reaching the required workload and cadence trial lasted 2 min. There was at least 5 min between the trials to avoid fatigue. Trials were performed in a random order and cadence was controlled using a metronome. The three workloads and cadences were selected based on a prior verbal interview with each participant and their performance data.
Each participant brought his/her own road bicycle to the lab, which was mounted on the cycling ergotrainer (RacerMate, Computrainer, Seattle, WA, USA). Calibration of the ergotrainer was performed after the warmup according to the manufacturer's instructions to ensure the most accurate measurement, rear tire pressure was checked prior each test and ambient temperature was controlled continuously. One force pedal (Forped, Cycling Science Ltd., Kranj, Slovenia) was mounted on the left side, which recorded vertical (FV), anterior-posterior (FAP) and medial-lateral (FML) forces at a sampling rate of 1000 Hz. The force pedal was designed for Shimano SPD-SL pedal cleats and this type of the pedal was mounted on the other side. Those participants who did not use a Shimano system, cleats were changed on their shoes matching the position and orientation based on the pedal axis (marks on cleats). The force pedal and the normal pedal had the same axis height and lateral stack.
Active LED markers were positioned on the greater trochanter, lateral condyle, lateral malleolus and two on the force pedal. After volume and orientation calibration following the manufacturer's guidelines for best accuracy and reliability [12], kinematics were recorded with an active kinematics system at a sampling rate of 250 Hz (3D investigator, NDI, Waterloo, ON, Canada). Pedal forces and kinematics data were synchronized through a digital pulse sent from the kinematics system and matched synchronically in the post analysis.

Analysis
Data analyses were performed using custom written scripts in MATLAB ® . The last minute of the recording was taken into the further analysis. Data from the force pedals were first down-sampled to match the data from the kinematics. All data were filtered using a Butterworth filter, 2nd order with a low cut-off frequency of 10 Hz [13]. All crank cycles from the recording were cut and averaged to form an ensemble average presented on a crank angle scale from the top dead center (TDC, i.e., 0/360 • ) over the bottom dead center (BDC, i.e., 180 • ) back to the TDC. Crank and pedal angles were calculated from the kinematics on the pedal.
The total force (Ftot) was calculated as the vector sum of all three force components. FML was calculated (root mean square) in absolute and relative terms. Normalized FML was expressed as the percentage of the maximal value of Ftot. Kinematic data from the markers on greater trochanter, lateral condyle and lateral malleol were used to calculate Q-angle-the angle between the tibia and femur segments in the frontal plane. This is not a conventional way to calculate the Q-angle, but as this study focuses only on the change in Q-angle throughout the crank cycle, the absolute values are not relevant in this study. Furthermore, the setup used in this study is also the same as used by the majority of cycling-specific 3D kinematics systems (e.g., Retul or STT 3DMA).
The data were presented as mean ± standard deviation using descriptive statistics. Normality of the data was first checked using Shapiro-Wilk's test. One-way repeated measure ANOVAs were used to examine the differences in absolute and relative FML across the three cadences and across the three workloads. Post hoc pairwise tests with a Bonferroni correction for all ANOVA tests were performed if the main effects were found statistically significant (p < 0.05). Additionally, effect sizes for ANOVA (eta-squared) and 95% confidence intervals (CI) for post hoc tests were calculated. Peak cross-correlation coefficients (zero lag) were calculated between the Q-angle and relative FML. All analyses were carried out in SPSS 22 for Windows and in MATLAB ® . The data was complete for all participants and all trials were included in the further analyses.

Results
Cross-correlation coefficients between the Q-angle and relative FML reached values between 0.70 and 0.77 and are illustrated in Figure 1. There were no statistically significant differences in cross-correlation coefficients among the cadences (F(2,42) = 0.642, p = 0.531, across the three cadences and across the three workloads. Post hoc pairwise tests with a Bonferroni correction for all ANOVA tests were performed if the main effects were found statistically significant (p < 0.05). Additionally, effect sizes for ANOVA (eta-squared) and 95% confidence intervals (CI) for post hoc tests were calculated. Peak cross-correlation coefficients (zero lag) were calculated between the Q-angle and relative FML. All analyses were carried out in SPSS 22 for Windows and in MATLAB ® . The data was complete for all participants and all trials were included in the further analyses.

Results
Cross-correlation coefficients between the Q-angle and relative FML reached values between 0.70 and 0.77 and are illustrated in Figure 1. There were no statistically significant differences in cross-correlation coefficients among the cadences (F(2,42) = 0.642, p = 0.531, ŋ 2 = 0.030) and workloads (F(2,42) = 0.458, p = 0.636, ŋ 2 = 0.021). across the three cadences and across the thr Bonferroni correction for all ANOVA tests w statistically significant (p < 0.05). Additionall 95% confidence intervals (CI) for post hoc t coefficients (zero lag) were calculated betwee were carried out in SPSS 22 for Windows an all participants and all trials were included i

Results
Cross-correlation coefficients between t between 0.70 and 0.77 and are illustrated in F differences in cross-correlation coefficients a ŋ 2 = 0.030) and workloads (F(2,42) = 0.458, across the three cadences and across the three workloads. Post hoc pairwise tests w Bonferroni correction for all ANOVA tests were performed if the main effects were fo statistically significant (p < 0.05). Additionally, effect sizes for ANOVA (eta-squared) 95% confidence intervals (CI) for post hoc tests were calculated. Peak cross-correl coefficients (zero lag) were calculated between the Q-angle and relative FML. All ana were carried out in SPSS 22 for Windows and in MATLAB ® . The data was complet all participants and all trials were included in the further analyses.

Results
Cross-correlation coefficients between the Q-angle and relative FML reached va between 0.70 and 0.77 and are illustrated in Figure 1. There were no statistically signif differences in cross-correlation coefficients among the cadences (F(2,42) = 0.642, p = 0 ŋ 2 = 0.030) and workloads (F(2,42) = 0.458, p = 0.636, ŋ 2 = 0.021).  across the three cadences and across the three workloads. Post hoc pairwise tests with Bonferroni correction for all ANOVA tests were performed if the main effects were foun statistically significant (p < 0.05). Additionally, effect sizes for ANOVA (eta-squared) an 95% confidence intervals (CI) for post hoc tests were calculated. Peak cross-correlatio coefficients (zero lag) were calculated between the Q-angle and relative FML. All analys were carried out in SPSS 22 for Windows and in MATLAB ® . The data was complete f all participants and all trials were included in the further analyses.  Bonferroni correction for all ANOV statistically significant (p < 0.05). A 95% confidence intervals (CI) for coefficients (zero lag) were calcula were carried out in SPSS 22 for W all participants and all trials were

Results
Cross-correlation coefficients between 0.70 and 0.77 and are illus differences in cross-correlation coe ŋ 2 = 0.030) and workloads (F(2,42 There was no statistically significant differences in normalized FML across the three cadences (F(2,42) = 1.44, p = 0.248, Appl. Sci. 2021, 11, x FOR PEER REVIEW across the three cadences and across the three workloads. Pos Bonferroni correction for all ANOVA tests were performed if th statistically significant (p < 0.05). Additionally, effect sizes for A 95% confidence intervals (CI) for post hoc tests were calculat coefficients (zero lag) were calculated between the Q-angle and were carried out in SPSS 22 for Windows and in MATLAB ® . T all participants and all trials were included in the further analy

Results
Cross-correlation coefficients between the Q-angle and rel between 0.70 and 0.77 and are illustrated in Figure 1. There were differences in cross-correlation coefficients among the cadences ŋ 2 = 0.030) and workloads (F(2,42) = 0.458, p = 0.636, ŋ 2 = 0.021 across the three cadences and across the three workloads. Post hoc pairwise tests with a Bonferroni correction for all ANOVA tests were performed if the main effects were found statistically significant (p < 0.05). Additionally, effect sizes for ANOVA (eta-squared) and 95% confidence intervals (CI) for post hoc tests were calculated. Peak cross-correlation coefficients (zero lag) were calculated between the Q-angle and relative FML. All analyses were carried out in SPSS 22 for Windows and in MATLAB ® . The data was complete for all participants and all trials were included in the further analyses.

Discussion
The aim of this study was to test the hypotheses that (1) the FML would correlate with the knee frontal plane kinematics and (2) the FML would differ at different workloads and cadences. The results of the present study support the hypotheses and show a correlation between the Q-angle and the FML and demonstrate the differences in FML across different pedaling conditions. The FML has been often linked to overuse injuries [11]. The risk of anterior knee pain and patellar tendinitis could be increased if the lower leg is in an abducted position when a knee extensor moment is generated [5]. It was demonstrated that cyclists with history of the anterior knee pain display significantly more abduction (knee valgus) during the downstroke and remain in knee abduction during the upstroke [5]. The results of the present study show that a correlation between the temporal characteristics of the FML and QA is present during steady state cycling on a cycling ergometer. The link between the

Discussion
The aim of this study was to test the hypotheses that (1) the FML would correlate with the knee frontal plane kinematics and (2) the FML would differ at different workloads and cadences. The results of the present study support the hypotheses and show a correlation between the Q-angle and the FML and demonstrate the differences in FML across different pedaling conditions. The FML has been often linked to overuse injuries [11]. The risk of anterior knee pain and patellar tendinitis could be increased if the lower leg is in an abducted position when a knee extensor moment is generated [5]. It was demonstrated that cyclists with history of the anterior knee pain display significantly more abduction (knee valgus) during the downstroke and remain in knee abduction during the upstroke [5]. The results of the present study show that a correlation between the temporal characteristics of the FML and QA is present during steady state cycling on a cycling ergometer. The link between the FML and frontal plane knee kinematics was previously demonstrated [11] but was limited to only specific data points (e.g., peak) and not across the entire force/motion pattern.
Knee movement in the frontal plane plays a significant role in the FML production. However, it cannot be the only mechanism and other factors remain to be tested. It was also observed that certain amount of inter-individual variability exists (Figure 3) when the correlation between the knee movement and FML was drawn (correlation coefficients ranging between 0.57 and 0.97). Therefore, future research should also focus on factors associated with FML, with some emphasis on inter-individual difference. One of the factors that could play a role in the FML production could be the lateral stability/control of the hip. In particular, the activity of the gluteus medius as an important muscle for providing stability around the hip region, which is then directly transferred to the pedals over the knee and ankle joints.
Appl. Sci. 2021, 11, x FOR PEER REVIEW 6 of 8 FML and frontal plane knee kinematics was previously demonstrated [11] but was limited to only specific data points (e.g., peak) and not across the entire force/motion pattern. Knee movement in the frontal plane plays a significant role in the FML production. However, it cannot be the only mechanism and other factors remain to be tested. It was also observed that certain amount of inter-individual variability exists (Figure 3) when the correlation between the knee movement and FML was drawn (correlation coefficients ranging between 0.57 and 0.97). Therefore, future research should also focus on factors associated with FML, with some emphasis on inter-individual difference. One of the factors that could play a role in the FML production could be the lateral stability/control of the hip. In particular, the activity of the gluteus medius as an important muscle for providing stability around the hip region, which is then directly transferred to the pedals over the knee and ankle joints.  Root mean square of the absolute and relative FML was found to be different across the three tested workloads. Moreover, a post analysis revealed that the absolute FML was found greater, whereas the relative FML exhibited lower values during the HIGH compared to LOW cycling conditions. Higher workload at the same cadence can only be achieved by applying more force to the pedals. It is therefore not surprising that the absolute FML was also increased. On the other hand, a reduction in relative FML during HIGH compared to LOW correlates with an increase in force effectiveness [8]. Increased force effectiveness observed during higher workloads has been often reported in the literature [14] and differences in IE among workloads were also observed in the present study. Results of this study suggest that the relative amount of the FML is reduced when pedaling technique is more effective.
Even though the majority of previous biomechanical studies examining force effectiveness during cycling neglected FML component [1], it clearly plays a significant role in the calculations of the Ftot applied to the pedals, and consequently the IE. That itself would not be problematic if the relative contribution of the FML was always proportional to the Ftot applied to the pedals. However, results of this study showed that the magnitude of the FML is affected by cadence and workload. Hence, future studies using force effectiveness as a biomechanical parameter should use a three-dimensional force sensor.
It was interesting to observe that there were no systematic differences in Ftot when compared among the three tested cadence conditions. In theory, with the same pedaling technique (i.e., same force effectiveness) and workload, the Ftot should be larger at lower cadences. This was not the case in the present study with the only possible explanation being that there was a larger proportion of the ineffective force during the higher cadences, which consequently lowered the force effectiveness [10]. One of the components contributing to the ineffective force is also the FML, which was seen to increase in the C95 compared to the C75 condition. Henceforth, a possible mechanism for the decrease in force effectiveness at higher cadences is partially also the increase in the FML.
One of the limitations of this study was that only one side of the subjects' pedal forces and kinematics was recorded. It has been demonstrated recently that substantial contra-lateral differences can be present even in highly trained cyclists [13]. Even though these differences are not directly related to performance, they could provide a better understanding on the FML component. Another limitation is that only pedal forces were recorded, without consideration of moments, which could provide an insight into the stability of the foot and the amount of rotation present at the pedal. Both of those moments are dynamically changing under isometric conditions (due to the fixation at the pedal) and as such can affect the load on the joints of the lower extremity.
This study has a practical value for clinicians and sport scientists who work in the field of cycling biomechanics optimization. Results showed a correlation between the FML and Q-angle. Therefore, even with limited technology (i.e., force pedals), one could assess only Q-angle by means of video analysis. By minimizing the change in Q-angle (e.g., cadence) it can be expected to reduce the FML and consequentially minimize the valgus/varus moments in the knee joint.

Conclusions
The results of the present study imply that the FML component in healthy young recreational riders with no knee pain is an important factor that needs to be considered in the assessment of pedaling technique. Furthermore, lateral knee movement is related to the FML and should be taken into an account in a process of cycling biomechanics optimization. The results of this study confirm the conclusions from previous studies on cycling biomechanics [7] that a complete description of the knee loading requires the information of the three-dimensional loading at the pedal.  Institutional Review Board Statement: Before the experiment, each participant signed an informed consent document, which was approved by the Slovenian national medical ethics committee.
Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.