Vertical ground reaction force magnitudes and rates
not positively correlated with prospective running injury
1Kiernan,
Neuromechanics Research Core
1University
D., 1,2Shim, J. K., 1Miller, R. H.
of Maryland, College Park, MD, USA 2Kyung Hee University, Yong-in, South Korea
Results
Research Question
Do vertical ground reaction forces (VGRF) measured at
baseline predict prospective injury development?
•  Hypothesis: VGRF theorized to cause injury; therefore,
baseline VGRF should positively correlate with
prospective injury
0
Recruitment: n = 44, 22 male, age 30±10
1
2
3
4 PCs accounted for >90% of VGRF variance
PC2 positively correlated with number of (r = 0.31, p < 0.01),
and pain caused by (r = 0.25, p < 0.05), injuries
•  High PC scores à low magnitudes and rates
•  Corroborates negative rate-injury correlation
Methods
4
Traditional metrics show no positive VGRF-injury correlations
•  Rate negatively correlated with number of (r = -0.31, p <
0.01), and pain caused by (r = -0.26, p = 0.01), injuries
Baseline biomechanics: 10 left and 10 right
stances used to calculate VGRF magnitude and
rate
A
A
5
B
A
Table 1: Correlations and coefficients of determination – r (r2) – between ipsilateral VGRF (magnitude, rate, PCs)
and injuries (number, pain, days missed). * indicates significance at p < 0.05.
Magnitude
Rate
PC1
PC2
PC3
PC4
3
VGRF
1A
2.5
7
VGRF (BW)
6
2
1.5
8
1
2A
0.5
0
9
0
40
50
60
70
80
90
100
Figure 1: (A) At heel-strike VGRF rapidly and stereotypically rises to an impact peak.
(B) Representative VGRF curve normalized to stance time. Magnitude calculated as
highest value in impact peak (1). Rate calculated as the slope between 20-80% of
stance onset (VGRF > 15 N) and impact peak (2).
1 0
Principal Component Model (PCM) used to
objectively identify waveform features explaining
>90% of VGRF variance
VGRF@t2
VGRF@t2
PC2: Direction
of the second
largest variance
1.5
1 6
1 4
2
1 5
PC1: Direction
of the largest
variance
1 8
VGRF@t1
Figure 2: Principle component model. Variance-based coordinate system transformation that
transforms VGRF data from 101 time points to 101 uncorrelated principal components (PCs).
Prospective injury: 26 weekly internet surveys
•  Number of injuries
•  Location of each injury
•  Pain caused by each injury
•  Days of running missed due to each injury
A
A
B
A
0
10
20
30
40
50
60
70
80
90
100
2 2
2 1
2 3
Figure 2: Survey questions.
(A) For each injury, participants could click on
the figure up to 6 times to indicate injury
location.
(B) Pain was subjectively rated by moving the
slider bar from 0 or “no pain” to 10 or “worst
pain imaginable”
2 4
Figure 3: Principal component 2 (18.9% of VGRF variance). Interpreted as capturing the VGRF impact peak and
loading rate, as well as acting as ‘difference operator’ on the size of impact peak relative to overall VGRF peak.
High PC scores were associated with lower impact peak magnitudes and rates of loading but higher second peak
magnitudes. High Pc scores were also associated with more, and more painful injuries
Discussion
Contrary to theory, VGRF magnitude and rate did not predict
injury
•  In fact, rate negatively correlated with injury
•  Consistent with earlier prospective work
PCM results suggest method useful for objectively identifying
biomechanical waveform features that may predict injury
VGRF may, however, be insufficient to identify at-risk runners
2 5
2 6
0
Stance (%)
2 0
1 9
VGRF@t1
Mean
Mean + 2SD (high injury)
Mean - 2SD (low injury)
2.5
0.5
e n d
Days Missed
-0.11 (0.01)
-0.12 (0.01)
-0.07 (0.01)
0.14 (0.02)
-0.06 (0.00)
0.02 (0.00)
3
VGRF (BW)
1 1
30
Pain
-0.16 (0.03)
-0.26 (0.07) *
-0.04 (0.00)
0.25 (0.07) *
-0.04 (0.00)
0.02 (0.00)
Stance (%)
1 3
1 2
20
1
1 7
Time (weeks)
10
Number
-0.19 (0.03)
-0.31 (0.09) *
-0.12 (0.01)
0.31 (0.10) *
-0.06 (0.00)
0.06 (0.00)
Analyses: Pearson correlations between
ipsilateral VGRF (magnitude, rate, PCs) and
injury (number, pain, days missed)
Future work will examine efficacy of other predictors
•  Internal loading of injury-prone structures is the most
direct cause of injury; therefore, modelled internal loads
should predict injury
References
Acknowledgements
Contact
1. Lopes AD, et al. Sports Med 42, 891-905, 2012.
2. Zadpoor AA & Nikooyan AA. Clinical Biomech 26, 23-28, 2011.
3. Deluzio KJ, et al. Hum Mov Sci 16, 201-217, 1997.
4. Nigg BM, et al. J Appl Biomech 11, 407-432, 1995.
Support for this research was provided by the University of Maryland
Kinesiology Graduate Research Initiative Fund to Dovin Kiernan.
Dovin Kiernan was also supported by an NSERC CGS-M scholarship
Dovin Kiernan
Neuromechanics Research Core
[email protected]
http://sph.umd.edu/neuromechanics