Ashley and Allyson

Team Members

• Ashley Lowber
• Allyson Weiss

Project Video

Link to project video

Project Files

Optimization Files: 

Files for Unloaded Model:

Files for Loaded Model:

Background

It is estimated that, during the latter stages of pregnancy, a woman may have gained up to 11-16 kg of weight¹. Such weight increase can lead to anterior pelvic tilt⁹ or changes in gait kinematics to compensate for more load and altered center of mass⁵. Kinematic changes may lead to increased force in the hip extensors in an attempt to stabilize the torso during gait¹, thus causing overuse injuries in these muscles, which sometimes manifests as lower back pain. One study found that the prevalence of low back pain during pregnancy is greater than 50%, with 22% of participants reporting ongoing low back pain throughout pregnancy¹⁰. According to an article published in 2018, increased strength in gluteus, hamstring, and rectus abdominis muscle groups can reduce anterior pelvic tilt and thus back pain², but this finding is not specific to pregnancy. Low back pain significantly reduces quality of life and functionality in this population, indicating the need for better understanding of the biomechanics of movement during pregnancy and potential solutions to reduce overuse injuries. 

Research Question(s)

How sensitive are changes in key hip muscle forces to changes in gluteus medius and hamstring strengths in loaded vs. unloaded walking?

Methods

Model

We conducted this study using a loaded and unloaded model previously used to study the use of assistive devices to reduce the metabolic cost of walking with a load⁴. The models consist of 39 degrees of freedom, with 8 deemed nonessential for the study⁴. The models were scaled to each subject's size by Dembia, et al. The loaded model included an additional 38kg of weight, split between a backpack and three weight vests⁴. Using ground reaction forces and motion capture data published by Dembia, et al.⁴, we conducted inverse kinematics analysis in OpenSim to estimate joint kinematics.

Figure 1: Representation of the OpenSim model with normal muscle strength, loaded condition⁴.

Static Optimization

Following inverse kinematics analysis, we used external loads data¹¹, joint kinematics data, and each respective model to perform static optimization and estimate lower-limb muscle activity during one gait cycle. Our method minimized the sum of activations squared, where N is the number of muscles in the model: 

The optimization obeyed the following set of constraints: 

Because our model includes only lower-limb muscles, we focused mainly on altered strength of the gluteus and hamstring muscle groups, which have been indicated in previous studies as improving lower back pain with increased strength². To assess the forces resulting from varied gluteus medius and hamstring strengths, we modified the maximum isometric strengths these muscles in the loaded and unloaded models, in isolation. Static optimization was performed on both the loaded and unloaded models for 40%, 60%, 80%, 100%, 110%, 1120%, 130%, 140%, and 150% of initial maximum isometric strength of the following muscles:

• Gluteus medius (1, 2, and 3)
• Biceps Femoris Long Head
• Biceps Femoris Short Head
• Semitendinosus
• Semimembranosus

Following static optimization, we analyzed the resulting forces of the hip extensor muscles (in particular, biceps femoris long head, biceps femoris short head, semitendinosus, semimembranosus, gluteus maximus, and adductor magnus) and the gluteus medius to assess the sensitivity of the changes in each of these muscle forces to varying gluteus medius and hamstring strengths.

Results

Validation

We validated our model by comparing the simulated muscle forces of the gluteus medius, gluteus maximus, and semimembranosus throughout the gait cycle of the unloaded model to those obtained by van der Krogt, et al¹². By comparing these graphs, we saw that our optimization yielded similar patterns in force generated by these muscles (except for the larger peak in the semimembranosus plot towards the end of swing phasefrom the van der Krogt model). The 100% strength conditions in the van der Krogt article (the black line), the force scale is similar to what we found for our baseline for 100% strength graphs, although our optimization does have higher forces than are demonstrated in the article. For the gluteus medius plots, both our result and the van der Krogt results show that less force is generated as strength decreases. (The gluteus maximus and semimembranosus plots in van der Krogt showed weakening of these muscles, while our gluteus maximus and semimembranosus graphs below corresponded to weakening gluteus medius strength; therefore, the trend cannot be compared  directly for these muscles).

Figure 2: Our simulated forces (bottom) vs. van der Krogt, et al. simulated forces over a gait cycle (top). 

Unloaded Model Results

When hamstring strength was changed in the unloaded model, we saw the largest change in hip muscle force production with 60% reduced strength (Figure 3A). This reduction led to greatly increased adductor magnus and gluteus maximus forces (60% and ~55%, respectively), but minimal change in gluteus medius force production. We hypothesize that the gluteus medius force production remained relatively stable because this muscle primarily acts in a different plane than the hamstring muscles; thus, changes in hamstring strength would not lead to significant compensation by the gluteus medius muscles. 

When gluteus medius strength was altered, on the other hand, we saw the largest changes in biceps femoris short head and semimembranosus muscles (Figure 3B). With increased gluteus medius strength, we saw reduced semimembranosus force production (~35% reduction). Further, a 60% reduction in gluteus medius strength resulted in an approximately 20% increase in total hamstring force production (including all four hamstring muscles). This could be due to the need for the torso to remain stabilized throughout gait, requiring unsuspecting muscles to compensate for weakness. Something interesting to note is that both increased and decreased gluteus medius strengths led to increases in force production by the biceps femoris short head. This could partially be due to the fact that the semimembranosus saw large changes, and thus the biceps femoris may be slightly compensating for those changes. Another, perhaps more probably explanation, is that muscle compensation is complex, and there are many aspects of muscle dynamics to consider when analyzing muscle compensation. 

Figure 3: Graphical representation of the percent change in maximum muscle force for the muscles of interest when hamstring strength (A) and gluteus medius strength (B) is altered in the unloaded model.

In plotting the total hip flexion moment across all conditions, we saw that the kinematics was preserved relatively well with changing gluteus medius strengths, although we do see variation from 20- almost 60% and 60%- 100% of the gait cycle (Figure 4). Looking at the hip flexion moment contribution for the individual muscle/muscle groups, we could see that the adductor magnus and the gluteus medius do not contribute much to the hip flexion moment, especially in comparison to the hamstrings and gluteus maximus (Figures 4 & 5). This is consistent with the relatively small moment arm for the gluteus medius that would activate this type of motion at the hip (Figure 6). The adductor magnus has moments arms that cause both extension and flexion at the hip; given that the net flexion moment is negative across 0-60% of the gait cycle, it makes sense that the adductor magnus would not contribute as much to creating hip extension for this portion of the gait cycle, or to the net flexion moment for the remaining 60-100% of the gait cycle (Figure 6).

Although the gluteus medius does not contribute very much in terms of moment, there is significant variation in the moment it creates in the weaker and stronger states relative to its baseline strength (dotted black line), particularly from 60-100% of the gait cycle. When the gluteus medius is stronger, it creates more extension in swing (which is consistent with the increased percentage of force generation shown in the force plots), and when it is weaker it generates less extension (consistent with the reduced percentage of force generation). When the gluteus medius is weaker, the hamstrings compensate to some extent, which is visible especially in the case where the gluteus medius is at 40% strength (dark blue line). Interestingly, the gluteus maximus actually causes slightly less extension when gluteus medius is weaker, perhaps because the hamstrings generate excess extension moment to preserve kinematics (Figure 4)

Figure 4: Table of total hip flexion moment and moment contributions from the adductor magnus, gluteus maximus, hamstring, and gluteus medius across changing gluteus medius strengths (40%, 80%, 100%, 120%, 140% strength) in unloaded condition

In the unloaded condition with changing hamstring strength, kinematics were not maintained at 40% hamstring strength (which may be the reason why  the adductor magnus and gluteus maximus had a  roughly 55-60% increase in force generation at this level of hamstring strength). With changing hamstring strength, the gluteus medius was very unresponsive, and the adductor magnus added only slightly more extension when the hamstrings were very weak at 40% strength. However, when the hamstrings were weakened, gluteus maximus does act to compensate by increasing its hip extension moment contribution, and with stronger hamstrings, the gluteus maximus decreases its extension moment. In addition, relative to the amount of hip extension contribution that was generated at baseline strength, the hamstrings show a greater increase in extension moment when strengthened compared to the gluteus medius, although the relative increase in force generation is higher in the gluteus medius (Figure 5)

Figure 5: Table of total hip flexion moment and moment contributions from the adductor magnus, gluteus maximus, hamstring, and gluteus medius across changing hamstring strengths (40%, 80%, 100%, 120%, 140% strength) in unloaded condition

Figure 6: Moment arms for the adductor magnus, gluteus maximus, hamstring, and gluteus medius (taken at 100% strength of gluteus medius) in unloaded condition

Loaded Model Results

When we look at the results from the loaded model (Figure 7), we see the most sensitive muscle for both alterations in hamstring and gluteus medius strengths being the adductor magnus. While this muscle is often thought of as part of the hip adduction muscle group, it also plays a significant role in hip extension¹³. The gluteus maximus also showed some sensitivity to variable hamstring strength, with approximately 25% reduction in force production in response to 40% strength increase in the hamstring muscles. This is an expected trend, as the gluteus maximus is a hip extensor and thus would need to compensate for other increased hip extensor forces to maintain the same kinematics. Something interesting to note with variable gluteus medius strength is decreased gluteus maximus peak force production for both increases and decreases in gluteus medius strength. These two counterintuitive cases are another demonstration of the complexity of muscle compensation. 

Figure 7: Graphical representation of the percent change in maximum muscle force for the muscles of interest when hamstring strength (A) and gluteus medius strength (B) is altered in the loaded model.

Kinematics were maintained more closely in the loaded model, although there is variation across gluteus medius strength condition from roughly 60-70% of the gait cycle. In the loaded case, we still see that the adductor magnus is very insensitive to the changes in the gluteus medius (given that there is essentially complete overlap across curves for the changing strength conditions). When we strengthened the gluteus medius under load, it created a greater extension moment similar to what was observed in the unloaded condition, and the hamstring moment and gluteus maximus moment slightly decreased early in gait (this is possibly due to the fact that the gluteus medius does pull more weight as an extensor). However, the gluteus medius has a larger extension moment from around 40-55% gait cycle, which is the portion of the gait cycle where the hamstring moment arm decreases.  When the gluteus medius is weakened, the gluteus medius creates less hip extension, and the hamstrings and gluteus maximus increase their moment contributions in the earlier portion of gait. 

Figure 8: Table of total hip flexion moment and moment contributions from the adductor magnus, gluteus maximus, hamstring, and gluteus medius across changing gluteus medius strengths (40%, 80%, 100%, 120%, 140% strength) in loaded condition

The most consistent kinematics preservation was achieved under load with changing hamstring strengths (Fig. 9). In this case, the adductor magnus and gluteus medius had very low sensitivity to changing hamstring strength; unsurprisingly, hamstring moment is most sensitive to changes in hamstring strength, although gluteus maximus does produce differential moments earlier on in the gait cycle. With stronger hamstrings, the hamstrings create more extension in the early and late phases of the walking gait cycle; when they are weaker, they produce less extension in these portions of the gait cycle. Where the hamstring extension moment increases early in gate, the gluteus maximus extension decreases somewhat, which is consistent with the moment arms shown for these muscles throughout the gait cycle, given that they act as extensors during the entirety of the cycle with the hamstring moment arm greater than the gluteus maximus (Figure 10).

Figure 9: Table of total hip flexion moment and moment contributions from the adductor magnus, gluteus maximus, hamstring, and gluteus medius across changing hamstring strengths (40%, 80%, 100%, 120%, 140% strength) in loaded condition

Figure 10: Moment arms for the adductor magnus, gluteus maximus, hamstring, and gluteus medius (taken at 100% strength of gluteus medius) in loaded condition

Conclusion

Under load, our optimization showed peak extension moments that were far smaller than than in unloaded walking, which could be due to the limitation with our optimization model. We found that altering gluteus medius strength in loaded and unloaded conditions does not significantly affect hip extensor force production. The adductor magnus sees large relative changes in force, but produces insignificant changes in moment for changes in gluteus medius and hamstring strengths during loaded walking. Under no-load walking, strengthening the hamstrings allows the hamstrings to have much greater increase in extension moment compared to the gluteus medius when gluteus medius is strengthened. This difference is less visible under loading. However, this does lead us to believe that altering the gluteus medius strength may not be the most effective approach for people walking under increased loads as a strategy to decrease overuse injuries; instead, strengthening hamstrings may prove to be more effective. 

Limitations

Our study has several limitations which should not be overlooked when interpreting our results. First, we assumed tendon strain would be negligible in muscle force calculation, allowing us to model the tendon as rigid. Second, our results only represent one subject's walking patterns from one trial, limiting the generalizability of these results. We also allowed all muscles in the model to activate independently, which may result in compensatory muscle coordinations that are physiologically difficult. Fourth, due to the nature of static optimization, we neglected muscle activation and deactivation dynamics in our simulations. Beyond the nature of static optimization, our optimization algorithm experienced a few hiccups during computation, particularly when optimizing for the unloaded model. On several optimization steps (4-20, depending on the model, out of 24), the algorithm warned us that it had found a local minima but the constraints we defined were not being met. Because of these challenges, our results should be assessed and interpreted with caution. Finally, and perhaps most importantly, while we aimed to study the kinematics and dynamics of pregnancy, the weight distribution of the loaded model was not representative of a pregnant model. Weight was added to the models via backpacks and weight belts, rather than concentrated on the anterior side of the torso. Further, the weight carried by the subjects was 38kg, while the weight typically gained during pregnancy ranges from 11kg to 16kg. Despite these limitations, this study provides a good starting point for the study of the effects of gluteus medius and hamstring strengths on hip extensor force production during walking with and without a load.

Future Work

This project is baseline work for studying effect of changing gluteus medius and hamstring strengths on muscle coordination strategies during walking. This study uses a model with additional load placed on the back of the skeleton, more closely representing added load of a backpack, for example. In order to more directly study the effects of gluteus medius and hamstring strengths on hip extensor activation patterns and force generation, future work should assess muscle coordination strategies under similar load and walking conditions using a model with added weight concentrated on the front of the skeleton. Future work could also explore the effect of varied hip extensor strength on lower-back muscle activation in a model with added weight concentrated anteriorly in order to assess the impact of hip extensor strength on lower back muscle overuse injuries in pregnant women.

Acknowledgments

We could not have completed this project without the help and support of Carmichael Ong, Nick Bianco, and the rest of the ME 485 teaching team.

References

(1) Branco, M., Santos-Rocha, R., & Vieira, F. (2014). Biomechanics of gait during pregnancy. The Scientific World Journal, 2014, 1–5. https://doi.org/10.1155/2014/527940.

(2) Crichton-Stuart (2018). Exercises to fix anterior pelvic tilt. Medical News Today.

(3) Cui, W., et al. (2019). Effects of Toe-Out and Toe-In Gaits on Lower-Extremity Kinematics, Dynamics, and Electromyography. Applied Sciences, 9(23), 5245. https://doi.org/10.3390/app9235245.  

(4) Dembia, C. L., Silder, A., Uchida, T. K., Hicks, J. L., & Delp, S. L. (2017). Simulating ideal assistive devices to reduce the metabolic cost of walking with heavy loads. PLoS ONE, 12(7). https://doi.org/10.1371/journal.pone.0180320.

(5) Foti, T., Davids, J.R., & Bagley, A. (2000). A Biomechanical Analysis of Gait During Pregnancy. The Journal of Bone & Joint Surgery, 82(5), 625.

(6) Huang, T. P., & Kuo, A. D. (2013). Mechanics and energetics of load carriage during human walking. Journal of Experimental Biology. https://doi.org/10.1242/jeb.091587.

(7) Manyozo, S. (2019). Low back pain during pregnancy: Prevalence, risk factors and association with daily activities among pregnant women in Urban Blantyre, Malawi. Malawi Medical Journal, 31(1), 71. https://doi.org/10.4314/mmj.v31i1.12.

(8) Noon, M.L. & Hoch, A.Z. (2012). Challenges of the Pregnant Athlete and Low Back Pain. Current Sports Medicine Reports, 11(1), 43-48. DOI: 10.1249/JSR.0b013e31824330b6.

(9) Norino, S., et al. (2019). Pelvic alignment changes during the perinatal period. PLoS One, 14(10). DOI: 10.1371/journal.pone.0223776. 

(10) Ostgaard, H.C., Andersson, G.B.J., Karlsson, K. (1991). Prevalence of Back Pain in Pregnancy. Spine, 16(5), 549-552. 

(11) Silder, A., Delp, S. L., & Besier, T. (2013). Men and women adopt similar walking mechanics and muscle activation patterns during load carriage. Journal of Biomechanics, 46(14), 2522–2528. https://doi.org/10.1016/j.jbiomech.2013.06.020.

(12) van der Krogt, M.M., Delp, S., Schwartz, M.H. (2012). How robust is human gait to muscle weakness? Gait & Posture. doi:10.1016/j.gaitpost.2012.01.017.

(13) Adductor Magnus Muscle (2016). Yoganatomy.