Peak Torque Prediction Using Handgrip and Strength Predictors in Men and Women With Motor Complete Spinal Cord Injury

Abstract

Objectives:

To establish predictive equations for peak torque of muscle groups with totally and partially preserved innervation in individuals with motor complete spinal cord injury (SCI), based on hand dynamometry and strength predictor variables.

Methods:

The cross-sectional study conducted at a rehabilitation hospital consecutively recruited 108 men and women with SCI. All participants performed maximum peak torque tests for shoulder abduction/adduction (isokinetic), trunk flexion/extension (isometric), and handgrip strength testing (hand dynamometer) to establish predictive peak torque equations. The primary outcomes were peak torque variables. Handgrip strength, age, injury level, time since injury, age at injury, body mass, height, body mass index, and physical activity level were the secondary outcomes used as strength predictor variables.

Results:

Handgrip strength was a predictor variable for shoulder abduction/adduction peak torque. The best predictive models for shoulder abduction/adduction peak torque exhibited R² = 0.57 and R² = 0.60, respectively (p ≤ .05). Injury level showed the highest significant predictive capacity for trunk flexion/extension peak torque models (R² = 0.38 and R² = 0.29; p ≤ .05).

Conclusion:

Shoulder abduction/adduction peak torque predictive equations may be an alternative for use in an accessible strength tool (hand dynamometry) to evaluate training and rehabilitation programs. Trunk flexion/extension peak torque equations exhibited moderate correlations and high standard error of the estimates and should be used with caution.

Introduction

Muscle strength is deemed essential for individuals with spinal cord injury (SCI),¹ and it is a fundamental physical capacity trained during and after rehabilitation.^2,3 Reduced strength has been associated with lower physical function^4–6 and functional performance and with the severity of the SCI.^7–9 Strength training is also essential to sport performance¹⁰ and reduces the risk of injury.¹¹ Muscle strength may play an important role in improving performance in adaptive sports.¹² Adapted sports represent a suitable and efficient option for achieving various objectives during the rehabilitation process (e.g., balance, strength, wheelchair ability, and social interaction).^13,14 As such, the functional potential and strength of individuals with SCI affect their performance.¹⁵ An accurate assessment of muscle strength is essential to determine the workloads to be applied in training and rehabilitation programs.¹⁶

Isokinetic dynamometry, a valid and reliable tool for strength and peak torque assessment,^8,17 has been adopted to evaluate the strength of individuals with SCI in clinical and research settings.^{8,9,11,18–24} Isokinetic peak torque is commonly used to assess muscle performance because of its reproducibility and association with clinical manifestations.²⁵ Isokinetic dynamometry is often used to evaluate muscle function and provides detailed and accurate information on torque, position, and limb velocity in different populations.²⁶ Moreover, isokinetic devices provide sensitive results for muscle groups with totally and partially preserved innervation⁵ (e.g., trunk muscles), which may affect optimal functional independence.^5,27–29 In 2020, a study established peak torque cutoff points in muscle groups with totally (shoulder abduction and adduction) and partially preserved innervation (trunk flexion and extension) for functional independence and wheelchair skills in men with SCI.⁵ Furthermore, peak torque may be related to better Paralympic sports performance.²⁵ Thus, isokinetic dynamometry is an important tool to predict functional independence in activities of daily living and sports performance in individuals with SCI.

Although isokinetic dynamometry is considered the gold standard in muscle strength evaluation^30,31 and can detect peak torque differences in high and low paraplegia,⁵ the instrument is not accessible to all rehabilitation or sports centers and requires a considerable amount of time to test populations with impaired mobility.³² Thus, predictive peak torque equations using a practical strength assessment tool associated with strength predictor variables are clinically more accessible and efficient. Grip strength measurement using a hand dynamometer is simple, objective, and easy to apply in individuals with SCI⁸ and may, therefore, be an alternative to predict the peak torque for muscle groups with totally and partially preserved innervation that has been previously established as cutoff points for functional independence and wheelchair skills. Moreover, isometric strength tests are simple to administer, involve minimal injury risk, and can detect subtle changes in strength.^33–35 To the best of our knowledge, only one study has used isokinetic dynamometry and handgrip strength to assess independence and quality of life in SCI.³⁶ However, the authors did not correlate the two strength measures.

Additionally, it has long been recognized that age,^36,37 age at the SCI event,³⁷ level and severity of injury,^38,39 and time since injury^36,38 are predictors used in SCI studies. However, these variables are not influenced by exercise, whereas muscle strength and power,^6,39 physical activity level,³⁶ and body composition⁶ are directly affected by interventions such as strength training. All these outcomes might be used as predictors to establish predictive equations in SCI individuals.

Thus, the present study aimed to establish predictive peak torque equations for muscle groups with totally and partially preserved innervation in individuals with SCI, based on hand dynamometry and strength predictor variables. We hypothesize that the predictive equations using handgrip and strength predictors will exhibit a positive and significant association with the peak torque of muscle groups with totally preserved innervation and a significantly lower correlation with partially preserved innervation.

Methods

Ethics approval

The study was approved by the institutional Research Ethics Committee (protocol n. 4.140.630). All participants were outpatients and provided written informed consent.

Participants

One hundred and eight men and women with SCI were consecutively recruited from the rehabilitation program of a network center of rehabilitation hospitals until the calculated sample size was achieved. Data were collected from July 2018 to February 2020.

Inclusion criteria were (1) diagnosed with traumatic SCI paraplegia, (2) time since injury of at least 6 months, and (3) complete motor lesion (American Spinal Injury Association Impairment Scale A or B).⁴⁰ Participants were excluded if they had a history of metabolic disorders or cardiovascular, cardiac, or orthopedic surgery that would hamper their exercise performance (Table 1).

Table 1.

Participant demographic data [presented as median (25th and 75th percentiles)] and spinal cord injury levels and etiology [expressed as absolute values (frequency)]

Demographic and spinal cord injury data	Total (n = 108)	Female (n = 20)	Male (n = 88)
Age, years	31.7 (24.9–40.0)	36.6 (27.9–40.9)	30.5 (24.3–40.0)
TSI, months	45.8 (19.1–134.8)	96.0 (20.4–249.2)	43.9 (19.1–93.7)
Age at injury, years	22.8 (19.3–32.5)	21.4 (18.2–31.8)	23.1 (19.3–33.7)
Body mass, kg	68.0 (56.9–77.9)	59.8 (54.3–70.2)	69.1 (58.8–78.3)*
Height, cm	169.8 (166.0–174.1)	163.7 (159.8–167.0)	170.3 (167.0–175.9)*
BMI, kg/m2	23.6 (20.3–26.2)	21.7 (20.4–26.1)	23.7 (20.2–26.6)
PAS	2.0 (2.0–3.0)	2.0 (2.5–4.0)	2.0 (2.0–3.0)
SCI injury level
HP	52 (48.1%)	9 (45.0%)	43 (48.9%)
MP	42 (38.9%)	8 (40.0%)	34 (38.6%)
LP	14 (13.0%)	3 (15.0%)	11 (12.5%)
Etiology
Auto accident	23 (21.3%)	7 (35.0%)	16 (18.2%)
Diving	1 (0.9%)	—	1 (1.1%)
Falls	12 (11.1%)	2 (10.0%)	10 (11.4%)
Gunshot wounds	62 (57.4%)	11 (55.0%)	51 (58.0%)
Hit by falling object	2 (1.9%)	—	2 (2.3%)
Knife wound	1 (0.9%)	—	1 (1.1%)
Motorcycle accident	7 (6.5%)	—	7 (8.0%)

Note: BMI = body mass index; HP = high paraplegia (T1 to T4); LP = low paraplegia (T12 to L3); MP = medium paraplegia (T5 to T11); PAS = physical activity scale; TSI = time since injury.

^*Significant difference in relation to the female group (p ≤ .05).

Participants were stratified into three subgroups to characterize the sample: high paraplegia (HP; T1 to T4), medium paraplegia (MP; T5 to T11), and low paraplegia (LP; T12 to L3).^41,42

Procedures

Participants were advised of the procedures, given instructions, and submitted to clinical assessments. Peak torque and strength assessments were conducted after a 24- to 72-hour interval.

Clinical assessment

Body mass was measured and height calculated based on Rufino et al.,⁴³ who determined height by multiplying the half-arm span by 2 and dividing the result by 1.03 for women and 1.06 for men.

Physical activity scale

Physical activity level was estimated based on hours of participation in sports and exercises, according to Janssen et al.⁴⁴: (1) sedentary (0 hours per week), (2) moderately active (1 to 3 hours per week), (3) active (3 to 6 hours per week), and (4) very active/athlete (more than 6 hours per week).

Peak torque and strength assessments

Four maximum peak torque and strength tests were carried out in a fixed order, using an isokinetic dynamometer (Biodex System 4) and hand dynamometer (Saehan Corporation, Masan, Korea), with a 5-minute rest between tests, as follows: (1) concentric shoulder abduction/adduction at 60°/s; (2) maximum voluntary isometric contraction (MVIC) of trunk flexion; (3) MVIC of trunk extension, and (4) maximum handgrip strength. The peak torque of these muscle groups has been previously established as cutoff points for functional independence and wheelchair skills. The same assessor conducted all the tests and gave standardized verbal encouragement.

Concentric shoulder abduction and adduction: Participants were allowed to hold the handle with their noninvolved limb to increase trunk stabilization (Figures 1A and 1B). Familiarization consisted of two sets of 10 submaximal concentric and reciprocal repetitions at 60°/s and level 2 on the perceived exertion scale for resistance exercises (OMNI-RES), with a 1-minute rest between sets.²⁴ This was followed by shoulder abduction and adduction assessments to obtain the peak torque for each movement (five maximal concentric and reciprocal contractions at 60°/s).

Figure 1.

Illustration of the position adopted for shoulder abduction and adduction (A and B) and trunk flexion (C) and extension (D) in maximum strength testing on an isokinetic dynamometer. (A) Initial shoulder abduction position in the isokinetic test, at an angle of 15º. (B) Initial shoulder adduction position in the isokinetic test, at an angle of 90º. (C) Trunk flexion position in the isometric test, at a hip angle of 55º. (D) Initial trunk extension position in the isometric test, at a hip angle of 105º.

MVIC of trunk flexion and extension: In a pilot study, 80° trunk flexion was the position of balance for individuals with SCI in the sitting position without assistance. Given the reduced trunk muscle strength in individuals with a high-level injury, tests were performed at 55° and 105° in the sagittal plane for hip flexion and extension, respectively. The participants were unbalanced in the direction of the movement being tested and stabilized by straps and belts on the back support to allow the detection of low torque values (Figures 1C and 1D).

Peak torque assessment consisted of two submaximal (level 2 on the OMNI-RES)²⁴ familiarization sets and two 6-second MVIC sets separated by 1-minute intervals. For the MVIC sets, participants were instructed to execute a continuous maximum contraction and avoid bouncing movements. The highest peak torque between the two MVIC sets was used for analysis. Participants performed the trunk flexion test with their arms relaxed; for trunk extension, they were instructed to rest their hands on the belts (Figures 1C and 1D).

Handgrip strength assessment: Handgrip strength was measured using a hand dynamometer (Saehan Corporation) with participants seated, shoulder adducted, elbow flexed to 90°, wrist in a neutral position, and the distal phalanges of the hand resting on the dynamometer handle. Participants then exerted maximum handgrip strength three times for each hand, alternating between hands and resting for 1 minute between attempts. After each attempt, the participant was informed about the result to motivate the next effort. The final result was the sum of the highest measures for the left and right sides.^45,46

Main outcomes

The primary outcomes were isokinetic (shoulder abduction/adduction) and isometric (trunk flexion/extension) peak torque. Handgrip strength, age, injury level (HP, MP, and LP), time since injury, age at injury, body mass, height, body mass index, and physical activity level were the secondary outcomes used as strength predictor variables.

Statistical analysis

Sample size was calculated based on a fixed linear multiple regression model, considering an effect size of 0.15, α of 5%, power (1 - β) of 90%, and four predictors out of 9 possible variables (handgrip strength, sex, age, time since injury, age at injury, body mass, height, physical activity level, and SCI level), resulting in a minimum sample of 108 individuals with SCI.

The Kolmogorov-Smirnov normality test was used to analyze data distribution. Descriptive data were expressed as median and interquartile (25th and 75th percentiles) for the nonparametric outcomes, and the Mann-Whitney U test was used to compare groups (men and women). The chi-square test was applied to compare the frequency proportions of SCI level and etiology between male and female participants.

A linear multiple regression model was used to generate four equations for predicting isokinetic (shoulder abduction/adduction) and isometric (trunk flexion/extension) peak torque. To avoid collinearity, Spearman’s test was applied to correlate the strength predictor variables (handgrip strength, age, injury level, time since injury, age at injury, body mass, height, body mass index, and physical activity level). The correlation matrix was analyzed, and variables that exhibited a high or very high correlation were considered collinear.³⁹ Correlation coefficient values were classified as very weak (<0.20), weak (0.20 to 0.39), moderate (0.40 to 0.69), high (0.70 to 0.90), and very high (>0.90).⁴⁷ SCI level and sex were considered dummy variables, indicating which injury level or sex were used for specific observations. The accuracy of peak torque predictions was quantified using R² and the standard error of the estimate (SEE).

The outlier labeling rule was used to detect outliers and discrepancies.⁴⁸ Outlier values were calculated by the difference between the 25th and 75th percentiles, multiplied by a factor of 2.2. The result was then subtracted from the 25th percentile and added to the 75th percentile.

The IBM SPSS Statistics package (version 22.0; SPSS Inc, Armonk, NY), and G^*Power statistical power analysis software (version 3.1.9.2; Universität Kiel, Germany) were used. Statistical significance was set at 5% (p ≤ .05, two-tailed).

Results

Participant characteristics

There were no dropouts in this study. The men were heavier (69.1 kg vs. 59.8 kg; p ≤ .05) and taller (170.3 cm vs. 163.7 cm; p ≤ .05) than the women. There were no significant differences in SCI level and etiology distribution between male and female participants (Table 1).

Peak torque and strength comparisons

Handgrip strength, shoulder abduction/adduction peak torque, and trunk flexion/extension peak torque were significantly higher in males when compared to females (Table 2).

Table 2.

Descriptive data for handgrip and isokinetic and isometric peak torque for male and female participants

Muscle strength outcomes	Total (n = 108)	Female (n = 20)	Male (n = 88)
Handgrip, kgf	100.5 (78.0–112.0)	60.5 (54.0–73.5)	103.0 (89.0–115.5)*
Shoulder abduction	61.1 (50.7–72.2)	44.9 (38.9–52.5)	63.8 (57.3–75.6)*
Shoulder adduction	87.8 (70.5–101.3)	62.3 (51.6–64.5)	92.0 (78.8–106.3)*
Trunk flexion, N.m	48.4 (37.4–62.6)	40.7 (33.0–47.9)	52.6 (38.4–64.3)*
Trunk extension, N.m	42.3 (33.6–57.9)	38.2 (30.6–44.3)	43.3 (34.6–60.9)*

Note: Muscle strength variables are expressed as median (25th and 75th percentiles).

^*Significant difference in relation to the female group (p ≤ .05).

Isokinetic and isometric peak torque prediction

All the regression equations were significant; however, trunk flexion/extension equations showed moderate correlations (R = 0.61 and R = 0.54; R² = 0.38 and R² = 0.29, respectively; p ≤ .05) and shoulder abduction/adduction high correlations (R = 0.75 and R = 0.77; R² = 0.57 and R² = 0.60, respectively; p ≤ .05). The predictor variables age, time since injury, and age at injury were removed by stepwise regression analysis in all equations (Table 3).

Table 3.

Stepwise regression (R and R² values) and standard error estimate (SEE) for isokinetic (shoulder abduction and adduction) and isometric (trunk flexion and extension) peak torque, standardized beta weights (β), confidence intervals (95% CI) for unstandardized beta coefficients, and probability values (p) of the predictor variables

Predictor variables		Shoulder abduction			Shoulder adduction			Trunk flexion			Trunk extension
R		0.75*			0.77*			0.61*			0.54*
R 2		0.57			0.60			0.38			0.29
SEE (SEE %)		11.1 (17.7)			14.3 (12.8)			16.7 (21.5)			18.2 (38.2)
	β	95% CI	p	β	95% CI	p	β	95% CI	p	β	95% CI	p
Handgrip	0.36	0/13 to 0.38	<.01	0.27	0.08 to 0.42	<.05	0.09	−0.05 to 0.32	.54	0.03	−0.17 to 0.22	.80
Body mass	0.35	0.24 to 0.53	<.01	0.15	0.02 to 0.40	<.05	—	—	—	—	—	—
Sex	−0.22	−16.83 to −1.84	<.05	−0.30	−26.20 to −7.51	<.05	−0.16	−20.05 to 2.71	.13	−0.19	−22.35 to 1.57	.88
Injury level	—	—	—	—	—	—	0.54	2.21 to 3.98	<.01	0.39	1.23 to 3.10	<.01
Height	—	—	—	0.27	0.33 to 1.20	<.05	—	—	—	—	—	—
PAS	0.27	2.52 to 7.31	<.01	—	—	—	—	—	—	—	—	—
BMI	—	—	—	—	—	—	0.09	−0.29 to 1.05	.26	0.30	0.55 to 1.95	<.05

Note: BMI = body mass index; PAS = physical activity scale.

^*Statistical significance at p≤ .05.

Handgrip strength was statistically significant in shoulder abduction (β = 0.36, 95% CI = 0.13–0.38, p ≤ .01) and adduction (β = 0.27, 95% CI = 0.08–0.42, p ≤ .05) predictive equations, but not for predictive trunk equations (p > .05). Body mass and sex were also significant predictors included in shoulder abduction/adduction regression models and injury level, sex, and body mass index for trunk flexion/extension regression models (Table 3).

The equation models are presented below (eAppendix):

where BM (body mass), BMI (body mass index), HG (handgrip), HT (height), IL (injury level), PAS (physical activity scale), S-ABD (shoulder abduction, in N.m), S-ADD (shoulder adduction, in N.m), T-EXT (trunk extension, in N.m), and T-FLX (trunk flexion, in N.m).

Discussion

Four significant predictive equations were established for peak torque of muscle groups with totally and partially preserved innervation in individuals with SCI, based on hand dynamometry and strength predictor variables. However, handgrip strength was not statistically significant in predictive equations for trunk flexion/extension. Handgrip strength is a simple, accessible tool that provides information regarding the association between strength and peak torque and can be used in clinical research and rehabilitation assessment.

Our findings indicated that handgrip strength is a significant predictor for shoulder abduction/adduction peak torque. The models explained 57% and 60% of shoulder abduction/adduction peak torque, respectively. Handgrip strength has been associated with overall upper and lower body strength in sport performance⁴⁹ and with strength in one-repetition maximum tests of upper and lower limbs in women.⁵⁰ Moreover, a recent study demonstrated that isometric tests provide insight into the force production capability for dynamic performance capabilities.³⁵ For individuals with SCI, research focuses predominantly on the impaired handgrip strength of individuals with tetraplegia.^8,51,52 However, one study described time since injury as a predictor of handgrip strength in individuals with paraplegia.³⁶ Although handgrip strength was a dependent variable in the present study, time since injury was excluded as a predictor in the equation models.

Assessment of peak shoulder abduction/adduction torque is essential in individuals with SCI and even more important for those who engage in sports. Both shoulder abduction and adduction are essential for wheelchair propulsion and activities of daily living.^11,36,53 Moreover, the shoulder joint is subject to constant loads during wheelchair propulsion,^15,54,55 and wheelchair athletes exhibit a higher rate of shoulder injury than sedentary individuals due to muscle imbalances.^53,56,57 Surface electromyography and kinematic analysis showed high demands of upper extremity weight-bearing activities in the shoulder girdle due to the weight distribution of the wheelchair user.¹⁵ As such, shoulder abduction/adduction and peak torque assessments are necessary to continually verify muscle imbalances and progression after strength training.

Recently, cutoff points for wheelchair ability used shoulder abduction/adduction peak torque as predictors.⁵ A cutoff point of 2.35 N.m/kg is required for the sum of shoulder adduction and abduction peak torque to achieve 111.9 seconds in the Adapted Manual Wheelchair Circuit (AMWC).⁵ In addition, 0.97 N.m/kg and 0.96 N.m/kg are cutoff points of the shoulder abduction to 18 seconds (performance score) and 300 m (3-min test) of the AMWC.⁵ Moreover, shoulder adduction strength can be considered a predictor in preventing shoulder pain in individuals with paraplegia.¹⁵ Thus, significant equation models for abduction/adduction peak torque based on handgrip strength make it possible to estimate peak torque and confirm hand dynamometry as a useful assessment tool. In addition to handgrip strength, body mass and sex were also significant predictors used in the equations. A previous study identified age as a predictor variable for peak shoulder adduction torque in 52 men with paraplegia.³⁶ The sample size used here (108) and inclusion of both men and women demonstrate that sex is a decisive predictor variable in the equation models.

Although the equation models showed moderate and significant correlations for peak trunk flexion/extension torque, handgrip strength was a nonsignificant predictor variable. It was expected that handgrip assessment would contribute to the equation models in terms of responsiveness at each injury level. The present study used a large sample with approximately 52% of medium and low paraplegia, and the trunk flexion/extension peak torque might correlate with handgrip strength in the same injury levels. Therefore, injury level was considered a dummy variable and the only predictor in the equation model for partially preserved innervated muscle groups. Previous studies have demonstrated the importance of injury level and its association with functional independence,⁵ one-repetition maximum in the bench press,⁵⁸ peak torque,^9,20 and power.²⁰ The established predictive equations explained only 38.0% and 29.0% of trunk flexion/extension, respectively, and should, therefore, be used with caution given the high SEE values of 21.5% and 38.2%, respectively. Future studies might investigate handgrip and trunk flexion/extension correlations using injury levels under T5.

Insofar as handgrip strength cannot clearly be recommended to assess general functional performance for all kinds of exercise programs in older adults,⁵⁹ the results obtained here should be interpreted with caution for all strength assessments, especially when associated with functional independence. Although cutoff points have been established using shoulder abduction/adduction and trunk flexion/extension, the equations exhibited SEE values ranging from 12.8% to 38.2%.⁵ The magnitude of the error when comparing directly to functional independence scales might be higher, considering the error of the peak torque cutoff points. Thus, we recommend using predictive equations for shoulder abduction/adduction peak torque based on handgrip strength to assess training or rehabilitation programs. Further research to investigate the predictive capacity of handgrip strength in functional independence and wheelchair ability may help to establish cutoff points for clinical practice.

Study limitations

The height and the PAS were significant predictors included in the equations. However, the height was estimated by the half-arm span, and the PAS does not measure the intensity of the training. Therefore, the real influence of both predictors should be confirmed using a specific assessment of the height and a better sensitivity physical activity scale.

Conclusion

In the present study, handgrip strength was a significant predictor only for peak shoulder abduction/adduction torque. Four significant predictive equations were established for shoulder abduction/adduction and trunk flexion/extension peak torque based on handgrip strength and strength predictor variables. Predictive equations for peak shoulder abduction/adduction torque may be an alternative for use in an accessible strength tool (hand dynamometry) to evaluate training and rehabilitation programs. Peak torque equations for trunk flexion/extension showed moderate correlations and high SEE values and should be used with caution.