The effect of unilateral training on contralateral limb strength in young, older, and patient populations: a meta-analysis of cross education

Abstract

Background: Cross education is the contralateral strength gain following unilateral training of the ipsilateral limb. This phenomenon provides an ideal rehabilitation model for acute or chronic rehabilitation; however, previous cross education meta-analyses have been limited to a handful of studies.

Objectives: The present meta-analysis aimed to (1) be as inclusive as possible, (2) compare cross education in young able-bodied, older able-bodied, and patient populations, (3) compare cross education between training modalities, and (4) detail the impact of methodological controls on the quantification of cross education.

Methodology: A review of English literature identified studies that employed unilateral resistance training and reported contralateral strength results. Studies were separated to examine the effect of population, training modality, limb, sex, and familiarization on the magnitude of cross education. The percent strength gain and effect size were calculated for ipsilateral and contralateral limbs.

Results: A total of 96 studies fit the predetermined inclusion criteria and were included in the analysis. The included studies were further divided into 141 units employing separate unilateral training paradigms. These were separated into young, able-bodied (n = 126), older, able-bodied (n = 9), and neuromuscular patients (n = 6). Cross education was an average of 18% (standardized mean difference (SMD) = 0.71) in young, able-bodied participants, 17% (SMD = 0.58) in healthy able-bodied participants, and 29% (SMD = 0.76) in neuromuscular patients.

Conclusion: Cross education was present in young, older, and patient populations and similar between upper and lower limbs and between males and females. Electromyostimulation training was superior to voluntary training paradigms.

Keywords:

• Cross education
• cross-transfer
• unilateral strength training
• electromyostimulation
• contralateral strength transfer
• stroke
• elderly

Background

Cross education is the strength gain that is found in the contralateral limb following a unilateral training program on the homologous limb. Cross education was first reported in 1894 by Scripture et al. [1] who determined that task steadiness and muscular strength could be improved in the contralateral limb following unilateral training. This phenomenon is of great importance for clinical applications and rehabilitation, and requires further mechanistic investigation. Cross education provides a beneficial rehabilitation model for unilateral injuries or disorders; including, acute injuries or immobilization (casting) of a single limb, and neurologic disorders, such as stroke, affecting the body unilaterally.

Previous research has proposed that cross education can be explained by two distinct, but not necessarily mutually exclusive, hypotheses: ‘cross-activation’ and ‘bilateral access’ [2, 3]. The ‘cross-activation’ hypothesis proposes that unilateral activity excites both ipsilateral and contralateral cortical motor areas. With this hypothesis, the unilateral training causes adaptations in both hemispheres, though to a lesser extent in the untrained hemisphere. Alternatively, the ‘bilateral access’ hypothesis suggests that the homologous untrained muscle can access the unilateral adaptations of training through interhemispheric communication from the associated motor areas [2, 3].

Previous meta-analyses and systematic reviews have determined that the average contralateral strength gain from cross education is approximately 8–12% [4–7]. This amount corresponds to approximately 35–60% of the strength increase that is found in the ipsilateral (trained) limb [4, 6, 8]. Manca et al. [7] further separated their estimate of cross education into lower limb (16.4%) and upper limb (9.4%). However, these previous reviews of cross education were limited to 2 [9], 8 [8], 10 [10], 13 [6], 16 [4], and 31 [7] articles. There are several factors that make the review of cross education complicated and limited, including the name discrepancies confounding the search for studies, and the variety of training paradigms. However, the primary reason for the small ‘sample sizes’ of cross education reviews is the stringency of inclusion criteria. The reviews by Munn et al. [6], Carroll et al. [4], Cirer-Sastre et al. [10], and Manca et al. [7], were limited to the analysis of randomized controlled studies. In addition, only studies with full data (means and standard deviations) for each of the ipsilateral experimental, contralateral experimental, and control limbs were included.

The inconsistent terminology and the unintentional examination of cross education using the contralateral limb as a ‘control limb’ for unilateral training has confounded the analysis of the field. Cross education of strength has been referred to by many names including cross-transfer, cross-over, or contralateral training. Similarly, the cross education of skill following unilateral practice is typically referred to as interlateral transfer of learning, bilateral transfer, or intermanual transfer. These studies generally focus on single session practice, rather than training, and the transfer of a skill, rather than strength. Although widely studied, the practice paradigms and the outcome measurements of the cross education of skill vary drastically across studies making them extremely difficult to quantitatively compare. Therefore, this meta-analysis focuses solely on the cross education of strength.

Lastly, variability in training paradigms makes it difficult to compare cross education between studies. There is a considerable variation in the duration (number of sessions), volume (contractions per session), intensity (load), and modality (type of contraction or stimuli) of unilateral training. The reviews by Carroll et al. [4], Munn et al. [6], and Manca et al. [7] limited their analyses to studies employing training intensities greater than 50% maximal strength for a minimum of 2 weeks. Most notably, the previous meta-analyses included only isometric, isokinetic, and dynamic training [4, 6, 7, 10], specifically excluding ‘alternative’ training via electromyostimulation (EMS), transcranial magnetic stimulation, vibration, or acupuncture.

The present analysis prioritized inclusivity over selectivity to capture the greatest overview of the field. A review of literature was undertaken to include as many ‘contralateral strength transfer’ studies as possible, including studies that unintentionally examined cross education by using an untrained contralateral limb as a control for unilateral training. The present analysis included studies using ‘alternative’ training, specifically EMS training (or neuromuscular electrical stimulation (NMES)), since previous meta-analysis have not previously included ‘non-traditional’ forms of strength training. In order to advance the use of cross education for rehabilitation purposes, the analysis was not limited to healthy populations as long as strength was assessed pre and post intervention.

Methods

Definitions

For the purpose of this analysis the term study will refer to an article as referenced. The term unit will refer to a training paradigm within a study, while the term limb will be the designated trained, untrained, or control limb of a participant. For example, one study may have two units within it where one unit was assigned to one type of training (e.g. eccentric training, elbow flexion training, low frequency training, etc.) and another unit was assigned to a separate training paradigm (e.g. concentric training, knee flexion training, high frequency training, etc.).

Literature search

The included studies were collected from an ongoing review of cross education and unilateral training literature. Studies were identified using Google Scholar, PubMed, and Research Gate using the search terms: cross education, cross-transfer, interlimb transfer, and contralateral strength gain. The reference list of each identified study was examined to include previously noted cross education studies not identified in the database search. In addition, studies using unilateral training were identified using search terms including: unilateral strength training, dominant AND non-dominant control limb and were examined for the unintentional observation of cross education where the contralateral limb was designated as a control limb.

Inclusion criteria

The selection of inclusion criteria was designed to be as inclusive as possible for the broadest review possible.

Population. All ages, sexes, and abilities were included in the present review. Units were separated into three groups: (1) young able-bodied (young) participants (< 50 years of age), (2) older able-bodied (older) participants (>50 years of age), and (3) neuromuscular disorder (patient) populations.

Training paradigm. All training types aimed at improving strength were included in the present study, including EMS training which has been previously excluded from cross education meta-analyses. Training modalities (contraction types) were separated into the following categories: isometric, isokinetic, dynamic (including isotonic), EMS, or ‘other’. If two types of voluntary contractions were performed for training, then the unit was placed in the ‘other’ category. The EMS category consists of stimulation alone or superimposed on a voluntary contraction. Any training intensity (load) was included as long as it was greater than 0% maximal strength (i.e. the intention was strength gain, rather than endurance gain). The criteria for number of sessions was >5 sessions to include training stimuli rather than mechanistic examinations.

Outcomes. Studies were included if strength was measured and reported in any manner including: pretraining and posttraining means, mean gain, or percent gain. Studies were further separated into units only where separate training paradigms were employed, rather than separate outcomes. Where one training unit had multiple outcomes, the single outcome that was homologous to the training modality (i.e. closest in contraction type, joint angle, speed of contraction, etc.) was selected, with the exception of EMS, vibration, or electroacupuncture training, where a voluntary contraction was selected. When multiple contraction types were used for training, as well as testing, the contraction type used most in training was selected as the outcome measure.

Sample size. The inclusion criterion for unit sample size was ≥3 to get an appropriate mean and standard deviation for effect size calculation. No control group was required for inclusion in the analysis.

Analysis

Effect size. Where means and standard deviations were reported effect size was calculated for each limb within a unit using The Cochrane Collaboration Review Manager (RevMan V.5.3) [11]. The standardized mean difference (SMD) and 95% confidence intervals were calculated using inverse variance as the statistical method, and random effects as the analysis model. Statistical significance (Z-score) was calculated in RevMan to determine if the effect size is greater than null. Where standard error (SE) was reported it was converted to standard deviation (SD) using the following formula including group sample size (n): $$SD=\mathrm{ }SE\mathrm{ }\times \mathrm{ }\sqrt{n}$$

The effect size was calculated where possible for the experimental limbs (trained and untrained) and the control limb(s). If both limbs of the control group were measured (dominant and non-dominant) then each limb was separately used as a control for the experimental limb. If only one control limb was tested then it was included as the control for both the trained and untrained experimental limbs.

Percent gain. Where means were reported the percent gain of the trained and/or untrained limb was calculated according to the following formula: $$\mathrm{\%}\mathrm{ }\mathit{\text{Gain}}=\frac{\mathit{\text{Post}}-Pre}{Pre}\mathrm{ }\times \mathrm{ }100$$

If only percent gain was reported but not pretraining or posttraining mean values then the percent gain was included as reported.

Cross-body transfer. The magnitude of cross-body transfer was calculated to determine how much of the training effect was transferred to the untrained limb. The calculation was conducted for each unit as follows: $$\mathit{\text{Cross}}-\mathit{\text{body}}\mathrm{ }\mathit{\text{Transfer}}=\frac{\mathit{\text{Untrained}}\text{ \% }\mathit{\text{Gain}}}{\mathit{\text{Trained}}\text{ \% }\mathit{\text{Gain}}}\mathrm{ }\times \mathrm{ }100$$

Comparisons. Independent sample t-tests were performed using SAS 9.4 (SAS Institute Inc., Cary, NC, USA) with a 0.05 significance level. The magnitude of percent gain in the untrained (cross education) limb and the trained limb was examined between (1) upper versus lower limb, (2) males versus females, and (3) familiarized versus non-familiarized units. The upper limb training consisted of elbow flexion, wrist flexion and extension, and handgrip exercises amongst others. The lower limb training consisted primarily of knee extension and flexion, and secondarily plantar flexion and dorsiflexion exercises. The effect of sex was examined from units that were composed of only males or only females. Finally, familiarization was taken as reported and included anything from a familiarization contraction or testing procedures familiarization to an entire familiarization session.

Results

Study and unit characteristics

A total of 113 studies were identified and 96 studies were included in the analysis (Figure 1). The 17 excluded studies did not fit the following criteria: no strength measure (4 studies), no strength data reported for untrained limb (4 studies), no pretest data (4 studies), less than 5 training sessions (2 studies), and less than 3 participants (3 studies). The remaining 96 studies included a total of 141 units. Of those, 126 units (from 87 studies [12–97]) included young, able-bodied participants with a median age of 23 years and a median sample size of 11 (range 3–342) participants. Nine units (from 8 studies [13, 27, 72, 92, 93, 98–100]) included older, able-bodied participants with a median age of 69 years and a median sample size of 11 (range 6–14).

Figure 1. Flow diagram of the identification and review process.

The remaining 6 units (from 6 studies [101–106]) were conducted using neuromuscular patient populations with a median sample size of 10 (range 5–21) participants. The neuromuscular disorder breakdown is as follows: stroke patients (three studies), patients with various neuromuscular disorders (one study), multiple sclerosis (one study), and osteoarthritis patients (one study).

Outcome measures

The training characteristics are presented in Table 1 for each of the groups. The results of effect size and percent gain for the number of units that fit each criterion are reported for the untrained (cross education) limb in Table 2 and for the trained limb in Table 3. Forest plots are presented for the untrained limb in Figure 2 for the young group and Figure 3 for the older (A) and patient (B) groups.

Figure 2. Forest plot of standardized mean difference (SMD) for each young unit included in the analysis for the untrained (cross education) limb. Light grey lines indicate cutoff values for small (0.2), moderate (0.5), and large (0.8) effect sizes.

Figure 3. Forest plot of standardized mean difference (SMD) for each older (A) and patient (B) unit included in the analysis for the untrained (cross education) limb. Light grey lines indicate cutoff values for small (0.2), moderate (0.5), and large (0.8) effect sizes. DF: dorsiflexion; KE: knee extension; MS: multiple sclerosis; OA: osteoarthritis.

Table 1. Median and range of training characteristics.

Training characteristic	Young	Older	Patients
	Median (range)
Number of training sessions	21 (6–84)	27 (6–36)	18 (16–27)
Training volume (sets x reps)	30 (3–250)	34 (20–40)	33 (24–42)
Training intensity (% maximum)	100 (10–100)	100 (70–100)	100 (80–100)

Table 2. Effect size (standardized mean difference), percent gain, and controlled percent gain for the untrained (contralateral) limb.

	N (units)	Effect size (95% CI)	N (units)	% gain	Cross-body transfer
Young	86	0.71 (0.60, 0.83)**	126	18	70%
Isometric	24	0.73 (0.56, 0.89)**	37	15	65%
Isokinetic	23	0.61 (0.41, 0.80)**	31	20	70%
Dynamic	27	0.65 (0.43, 0.86)**	41	18	71%
EMS	6	1.57 (0.81, 2.33)**	10	27	77%
Other	6	0.46 (0.18, 0.74)**	7	12	80%
Older	6	0.58 (0.22, 0.94)**	9	15	48%
Patients	4	0.76 (0.21, 1.31)**	6	29	77%
Control limb	38	0.10 (–0.04, 0.23)	48	2.2	–

CI: confidence interval; EMS: electromyostimulation. , **p < 0.001, for effect size only. Cross-body transfer is the amount of strength gain transferred from the ipsilateral limb to the contralateral limb.

Table 3. Effect size (standardized mean difference), percent gain, and controlled percent gain for the trained (ipsilateral) limb.

	N (units)	Effect size (95% CI)	N (units)	% gain
Young	81	1.11 (0.96, 1.25)**	123	29
Isometric	25	1.11 (0.90, 1.32)**	37	25
Isokinetic	20	0.95 (0.79, 1.11)**	30	31
Dynamic	23	1.18 (0.89, 1.47)**	39	31
EMS	7	1.87 (0.93, 2.82)**	10	35
Other	6	0.53 (0.24, 0.81)**	7	15
Elderly	6	1.44 (1.00, 1.87)**	9	31
Patients	2	0.56 (0.11, 1.01)*	4	29
Control limb	32	0.13 (–0.02. 0.27)	39	3.0

CI: confidence interval; EMS: electromyostimulation. *p < 0.05, ** p < 0.001, for effect size only.

The average percent gain (above baseline strength) in the untrained contralateral limb of young participants following unilateral training in the ipsilateral limb was 18%, as calculated from 126 units. A review of 86 units with adequate cross education data (means and standard deviations of the untrained limb) resulted in an effect size of 0.71 (95% CI: 0.60–0.83, p < 0.001). The amount of cross education was similar amongst different training modalities with the exception of EMS training. EMS training was employed in 10 units, which demonstrated an average strength gain of 27%. Six units reported enough data to calculate effect size which was large [107] at 1.57 (95% CI: 0.81–2.33, p < 0.001). This is greater than the small effect size of 0.10 (95% CI: –0.04–0.23, p = 0.16) in the control limb, which corresponded to a mean 2.2% gain.

The average percent gain in the untrained limb of older participants following unilateral training was 15%, as calculated from 9 units. A review of 6 units with adequate cross education data resulted in an effect size of 0.58 (95% CI: 0.22–0.94, p < 0.01). The modes of training included: dynamic (5), isokinetic (2), isometric (1), and resistance tubing (1). The amount of cross education in the Patients subgroup was a 29% strength gain (calculated from 6 units), which corresponded to a large effect size of 0.76 (95% CI: 0.21–1.31, p < 0.01, calculated from 4 units). Five of the studies employed strength training (resistive exercises) of the less-affected limb, one study [102] employed kicking and tracking movements of the less-affected limb while secured to a tilt-table.

The influence of limb, sex, and task familiarization had no influence on the percent gain of the untrained or trained limb, or the cross-body transfer, as presented in Table 4.

Table 4. The number of units that fall within each category: sex of the unit, the usage of familiarization, the limb involved, and the presence of a control group from the able-bodied participants.

	Number of units	Number of participants	% gain (untrained)	% gain (trained)	Cross-body transfer
Overall: able-bodied	135	2362	29	18	68%
Limb trained
Lower	68	791	18	28	68%
Upper	67	1571	17	30	69%
Sex of unit
Males only	65	997	16	26	65%
Females only	27	747	17	33	54%
Both sexes	34	506
Unknown sex	9	112
Familiarized
Yes	67	1387	18	29
No	68	975	17	29
Control group (Y/N)	79/56

Note: the presence of a control group (‘Y’) does not indicate the reporting of control results. *significant difference in % gain or cross-body transfer between unit categories (lower vs upper; males vs females; yes vs no familiarization), p < 0.05.

Discussion

The primary aim of the current meta-analysis was to prioritize inclusivity for the largest systematic analysis of cross education. Secondarily, this meta-analysis aimed to further cross education within the rehabilitation field by quantifying the presence of cross education in young and older able-bodied participants, as well as in patient populations. By carefully identifying the crucial inclusion criteria and reducing inclusion selectivity this meta-analysis was able to include data from 96 studies with 141 units of training groups.

The cross education gain was an 18% increase from baseline strength in young, able-bodied adults; a 15% increase in older, able-bodied participants, and a 29% increase in a patient population consisting of poststroke, neuromuscular disorders, and osteoarthritis patients. The values of cross education are higher than the previous and most widely cited estimates of 8% by Carroll et al. [4], and Munn et al. [6], and higher than the recent estimate of 12% reported by Manca et al. [7] The cross-body transfer to the untrained limb ranged from 52% to 80% of the ipsilateral training effect.

The separation of training modalities allowed for the analysis of cross education and training adaptation from different contraction types with sufficient sample sizes and statistical power. This identified the advanced capabilities of EMS training producing a cross education effect of 27%, of which previous meta-analyses excluded [4, 6, 7, 10]. Compared to cross education produced by isokinetic (20%), dynamic (18%), and isometric (15%) voluntary contractions, it is evident that EMS training produces a superior transfer of strength. The logistical ease of EMS training for varying populations and the associated voluntary strength gains, make it an ideal modality for cross education in rehabilitation settings. Additionally, EMS training provides a viable alternative for patients (e.g. osteoarthritis) where pain or joint stiffness are limiting factors in conventional strength training protocols [108].

The rehabilitative benefits of cross education are present, both as a strength gain and a prevention of strength loss. Andrushko et al. [109] detailed the preventative effects (sparing of muscle atrophy) of unilateral limb training during a period of contralateral limb immobilization. Alternatively, the present meta-analysis has demonstrated the presence of a strength gain in the contralateral (more-affected) limb of patient populations, following unilateral training of the less-affected limb. Dragert and Zehr [101] reported significant improvements in the timed-up-and-go (TUG) test following unilateral dorsiflexion training poststroke, and small but non-significant improvements in the modified Ashworth and Berg balance tests. Similarly, Kim et al. [102] demonstrated significant increases in gait velocity, cadence, stride length, symmetry, and double support periods following unilateral kicking movements of the less-affected limb, poststroke. Manca et al. [103] compared functional gains following direct versus contralateral training of the more-affected versus less-affected limb, respectively. Significant improvements in timed walking tests were seen in both groups. However, the direct training group had larger effects as well as significant improvements on the TUG test, for which contralateral training group did not. Taken together, the contralateral strength gains of cross education are promising for the rehabilitation of functional movements, specifically when the more-affected limb is unable to perform strength training.

There were numerous methodological deficiencies that were identified by previous meta-analyses including the need for control group data [6] and the lack of familiarization [4]. Both of these methodological controls are instituted for the purpose of minimizing ‘quick jumps in strength’ that would over-estimate the magnitude of cross education. The present meta-analysis included 48 control units for the cross education limb reporting an average strength gain of 2.2% (median: 2.1%, range: –6%–11%). Therefore, the inclusion of a control group is important to account for the over-estimation of cross education due to extraneous factors such as task familiarization.

It has been shown that task familiarity and familiarization contractions can increase force approximately 3–11% within a single session [110–113]. Carroll et al. [4] estimated that the effect of familiarization on the overestimation of cross education was approximately 4%. Therefore, it is surprising that there was no significant difference in the strength gain between groups that were familiarized and those that were not. It was hypothesized that a lack of familiarization would overestimate the magnitude of the cross education and training strength gain. The likely reason for the absence of difference in the strength gain is the lack of reporting in the majority of studies as to what was considered to be ‘familiarization’. Since most studies neglected to detail the method of familiarization, any study which noted that its participants were ‘familiarized’, be it a demonstration, a single test contraction, or an entire session, was included in the ‘familiarized’ group.

The large number of units included in the present meta-analysis allowed for the comparison of cross education between upper and lower limbs and between sexes in 135 units of able-bodied participants. Manca et al. [7], separated 31 studies into upper and lower limb training finding a larger magnitude of cross education in the lower limb (16.4%) compared to the upper limb (9.4%). However, the present meta-analysis found no significant difference between cross education in the lower (18%) and upper (17%) limbs. Similarly, there was no significant difference (p = 0.60) in the magnitude of cross education between males (16%) and females (17%), However, comparison between sexes in the trained limb revealed slightly larger (p = 0.06) training adaptations in females (33%) compared to males (26%). This resulted in a slightly larger (p = 0.17) cross-body transfer of strength in males (65% transfer) compared to females (54% transfer).

To date, many studies have assumed an equality between sexes in the magnitude of cross education, often citing the review by Zhou [8], which does not compare sexes. In the literature, only two studies [43, 100] included sex comparisons following unilateral training. Both studies also found significant differences between sexes in the magnitude of the training adaptation, but no difference in the magnitude of cross education. This indicates that there is a difference in the amount of transfer (or ratio between trained and untrained limbs) between the sexes, however previous literature is conflicting. Hubal et al. [43] found a significantly higher strength cross-body transfer ratio in females (21%) compared to males (16%). Alternatively, Tracy et al. [100] found a significantly lower strength transfer ratio in females (32% transfer) compared to males (36% transfer).

Conclusion

A review of 141 unilateral training units resulted in a cross education strength gain of 18% in young adults, 15% in older adults, and 29% in a patient population, which is higher than previous estimates [4, 6, 7] of 8% to 12%. The cross education effect was accompanied by a significant moderate to large effect size in each population. The average cross-body transfer ranged from 48% to 77% slightly higher that previous estimates of 35–60% [4, 6]. The present analysis identified: the presence of cross education in young and older able-bodied participants as well as patient populations; the efficacy of EMS training over voluntary modalities; and the equivalence in cross education between upper and lower limbs as well as in males and females. The 15–29% magnitude of cross education is promising for the use of unilateral training in rehabilitation.

Disclosure Statement

No potential conflict of interest was reported by the authors.

Additional information

Notes on contributors

David A. Gabriel

Lara A. Green recently completed her Ph.D. in health biosciences at Brock University examining the phenomenon of cross education. David A. Gabriel completed his Ph.D. in biomechanics at McGill University in 1995. He worked as a post-doctoral fellow in orthopedic biomechanics at the Mayo Clinic until 1997. He is currently a professor at Brock University.