A guide to asymmetries

• Introduction – What are asymmetries?
• What does the literature say?
• How can we use asymmetry data?
• How to interpret the noise of asymmetry
• Monitoring the direction of asymmetry
• Conclusion

Introduction – What are asymmetries?

Inter-limb asymmetry refers to the difference in performance or function between limbs ^[19], and should not be confused with intra-limb asymmetry: differences in muscle performance within the same limb, e.g., quadriceps vs. hamstring. Inter-limb asymmetry has been a popular choice of investigation in recent years, with numerous studies reporting its prevalence in a range of sports including soccer ^{[3, 5]}, cricket ^{[5, 9]}, rugby ^[22] and basketball ^[17]. These studies applied multiple testing modalities to show inter-limb strength differences using the isometric mid-thigh pull ^[14], isometric squat ^{[6, 8]} and isokinetic dynamometry ^{[23, 24]}. Similarly, jump testing with countermovement jumps (CMJ) and drop jumps (DJ) – bilateral and unilateral – have been used to measure inter-limb differences ^{[2, 4, 5, 6, 9, 20]}.

Despite what we know from the research, the prevalence of asymmetry alone does little to inform us whether an athlete might need targeted training interventions to enhance athletic performance or reduce the risk of injury.

What does the literature say?

Literature on this topic has highlighted the population-specific nature of asymmetry ^[20]. Magnitudes of asymmetry vary across different sporting populations due to the different physical demands sports place on their athletes.

However, asymmetry is also task and metric specific: a 10% asymmetry in one test is unlikely to carry over to a different test ^{[6, 7, 8, 11, 15, 20]}. Furthermore, when measuring asymmetry on equipment such as force platforms, practitioners may observe variation across the different output metrics. Recent studies showed that the magnitude of asymmetry varies for different metrics, even within the same test ^{[6, 7]}.

Thus, the task and metric-specific nature of asymmetry is likely to preclude the use of any single threshold in a consistent predictive capacity for injury risk ^[10].

How can we use asymmetry data?

The highly variable and inconsistent nature of asymmetry is something practitioners often gloss over, which inhibits our ability to understand how we can use the data in a meaningful way.

One additional aspect that has received less attention is the “direction of asymmetry” ^[21], which may provide us with important information for interpreting the data. Direction of asymmetry refers to the consistency of the asymmetry value favouring the same limb. For example, if we measure jump height during a unilateral CMJ and the left leg scores 20 cm and the right leg 18 cm, the inter-limb difference is 10% and favours the left side. We might note the same magnitude, however, in a subsequent test, even if leg dominance changes. Whilst this seems obvious (noting that it must favour one side if any amount of asymmetry is present), recent findings have shown just how variable the direction of asymmetry can be ^{[6, 7]}. That further raises the question of how we should interpret such data.

How to interpret the noise of asymmetry

The easiest way to demonstrate this concept is in Table 1, which shows hypothetical asymmetry data for jump height during a unilateral CMJ in 12 athletes. This data is based on a large number of datasets gathered since 2016, so I am confident it is a true representation of asymmetry data. The table includes the raw jump height and coefficient of variation (CV) for each leg. Jump height is the baseline asymmetry value, while CV is the variance in jump height from multiple trials. The table also shows the raw absolute asymmetry, where all values are positive, along with the same value but with limb dominance applied. In that case, a positive value means the asymmetry favour the right limb and a negative value means the asymmetry favours the left limb.

Athlete	Right	CV % (Right)	Left	CV % (Left)	Asymmetry %	Asymmetry % (direction)
Athlete 1	24.4	3.3	22.9	5.1	6.1*	6.1
Athlete 2	27.1	5.7	30.3	7.2	10.6*	-10.6
Athlete 3	23.0	5.1	24.0	4.6	4.2	-4.2
Athlete 4	19.0	8.5	23.6	9.6	19.5*	-19.5
Athlete 5	23.8	3.2	23.9	4.1	0.4	-0.4
Athlete 6	30.2	5.5	27.8	8.6	7.9	7.9
Athlete 7	22.1	2.2	23.1	2.9	4.3*	-4.3
Athlete 8	21.4	3.6	21.1	4.4	1.4	1.4
Athlete 9	27.6	4.9	25.0	6.6	9.4*	9.4
Athlete 10	20.8	10.2	17.9	14.4	13.9	13.9
Athlete 11	26.7	5.0	24.8	6.2	7.1*	7.1
Athlete 12	28.8	9.9	24.6	12.8	14.6*	14.6
Group mean	24.6	5.6	24.1	7.2	8.3	1.8
Standard deviation of the mean	3.5		3.1		5.7	10.2
% of standard deviation relative to the mean	14.2		12.9		68.7	–

Some key aspects stand out from Table 1 relating to asymmetry:

First, although the mean asymmetry value for the group is 8.3%, the individual scores vary massively for each athlete (range = 0.4-19.5%). This large variation is manifested by the standard deviation of 5.7%, which is equal to ~69% of the mean asymmetry score. This can be considered very large compared to the SD for the raw jump height scores, which represents only 13-14% of the mean. Given this high degree of variation within the group, it becomes apparent that group mean data is somewhat limited for asymmetry; and, if we are going to use the data, we must do so on a case-by-case basis.

Even then, left with a relative percent difference in asymmetry, it isn’t entirely clear what that value means, regardless of how big or small it is. Thus, I have also provided the CV value for each athlete, gathered from three trials on each limb ^[25]. The CV represents the variance in the test and is a measure of absolute reliability. Previous research ^[16] showed that, in order for an asymmetry to be considered “real,” the between-limb difference (asymmetry) must be greater than the intra-limb variance (CV). Simply put, for each athlete’s asymmetry score to be meaningful, it must be greater than both the left and right CV values. This enables us to differentiate between the signal (asymmetry) and the noise (CV). In addition, although anecdotal, previous literature has suggested that this is a fair comparison to make, given both metrics are percent values ^{[6, 7]}.

Taking this notion forward, if we look at each player’s asymmetry score relative to the CV, it’s clear that some show a real inter-limb difference and some do not (an asterisk indicates the asymmetry scores that are real). Although participant 10, for example, shows asymmetry of 13.9%, what is probably more concerning is the CV scores for jump height are actually higher. In instances like this, practitioners should use this type of information to determine why there were such variations in the raw scores – did the player not warm-up properly? are they carrying a small injury? are they not motivated, problems at home, etc. – especially when previous literature has indicated that we should be aiming for CV values of 5-10% in our test protocols ^{[12, 25]}.

As a final point to note, for those athletes who exhibit both a real asymmetry and acceptable CV scores, we don’t necessarily have to immediately target the imbalance in training. However, if we monitor this data over time and see that the asymmetry remains “real” at each test session and the same limb is consistently underperforming, then perhaps we are in a position to consider whether specific interventions are necessary to improve the weaker limb’s capacity. 

Monitoring the direction of asymmetry

Recent studies have shown that the direction of asymmetry may be just as variable as the magnitude, if not more so.

Bishop et al. used a single test session to measure inter-limb asymmetry from the unilateral isometric squat, CMJ and standing broad jump tests in recreational sport athletes ^[6]. Peak force was measured across all tests and the consistency of asymmetry favouring the same limb between tests was also calculated using the Kappa coefficient. This method looks at the proportion of tests in which asymmetry favours the same limb between tests, after removing any agreement by chance ^{[26, 27]}. Results showed poor levels of agreement (Kappa values < 0), indicating that limb dominance varied substantially between tasks, supporting the notion of task-specificity.

A more recent study investigated how consistently asymmetry favoured the same limb for common metrics, this time between test sessions. Bishop et al. used a test-retest design with 28 athletes measuring asymmetry for the unilateral isometric squat, CMJ and DJ tests ^[6]. Kappa coefficients investigating the consistency of asymmetry favouring the same side ranged from fair to substantial in the isometric squat (Kappa = 0.29-0.64), substantial in the CMJ (Kappa = 0.64-0.66) and fair to moderate in the DJ (Kappa = 0.36-0.56). Similar results emerged in a number of test-retest ^{[6, 11, 13]} and seasonal variation ^{[1, 9]} study designs.

Collectively, these findings indicate that the direction of asymmetry is highly variable and also varies across tasks, just like the magnitude. Therefore, there is arguably less of an explanation for between-limb differences in healthy athletes, assuming they have not been injured recently and have maintained their training consistently over time.

In addition, the mean value at the group level disguises much of the inherent variability associated with asymmetry, as we see in Table 1. The final column shows individual asymmetry scores, but this time with directional dominance. Limb dominance clearly varies from person to person, by virtue of both positive and negative values.

Understanding and monitoring this information is key, as it highlights that between-limb deficits are not the same for all athletes. Figure 1 shows this graphically, with scores above 0 exhibiting right limb dominance and scores below 0 exhibiting left limb dominance.

Although Figure 1 only shows data for a single test session, the stark variation in both the magnitude and direction of asymmetry is clear. If we now add data for a second hypothetical test session (Figure 2), some interesting results become apparent.

Athletes 3 and 7 from Figure 2 illustrate the fluctuating nature of limb dominance in this unilateral CMJ test. For both athletes, while a reduction in the magnitude of asymmetry from 4.2 to 3.6 (athlete 3) and 4.3 to 3.2 (athlete 7) may intuitively seem to be output “heading in the right direction,” such small changes are likely to be less important than shifts in asymmetry of 7.5-8% from one limb to the other.

Seasonal variation studies on asymmetry have reported this type of data over three test sessions at pre-, mid- and end of season ^{[1, 9]}. Those showed that it is not uncommon for the direction of asymmetry to return to the original limb during the third test session. As such, these data likely reflect “fluctuations in performance variability.” In contrast, some athletes exhibit more consistent limb dominance characteristics (e.g., athletes 9-12 in Figure 2). Practitioners should determine whether results like this are a by-product of competing in their sport over time ^[18] or whether such between-limb deficits pose a potential risk to injury or performance.

Conclusion

Practitioners need to understand the associated noise of inter-limb asymmetry, which is virtually guaranteed, relative to a more stable performance metric like jump height; and take account of the direction of asymmetry, which has built a growing interest in the last 4-5 years and provides greater context to the ratio statistic of asymmetry. 

By doing so, we are more likely to look for meaningful trends over time and not fall into the trap of acting on data too soon.