Strength training is recommended for participation in physical activities and rehabilitation programs for patients with spinal cord injury (SCI). It is associated with independence in and functional acquisition of daily activities.1, 2, 3 In this population, strength facilitates wheelchair propulsion, pressure relief and transfers.4, 5, 6 It is associated with increased cardiovascular capacity,7, 8 exercise tolerance,8 reduction in shoulder pain2 and improved health.1 During maximal or submaximal aerobic tests, the main limitation is related to strength and local muscle fatigue.4, 9

Strength is reduced by disuse that accompanies the aging process.10 In addition, individuals with SCI are more sedentary9, 11, 12 and less likely to perform any physical activity.13, 14 Consequently, increases in fat percentage and decreases in strength and muscle mass are common characteristics of SCI, especially during the first 3 months after trauma.15, 16, 17 Gorgey and Shepherd reported that atrophy contributes to changes in body composition and to increases in glucose intolerance, type II diabetes, hyperlipidemia, osteoporosis, metabolic syndrome and cardiovascular diseases.18

An accurate assessment of maximum strength is important to determine the workloads that should be used in the design of training programs.19 Many resistance programs are created using percentages of the maximum load. Therefore, correct determination of the relative load is essential for correctly prescribing training. These values comprise one of the main variables used to determine training objectives, such as strength, hypertrophy, power and local muscle endurance.20, 21

One way to determine workload is by using the one-repetition maximum (1RM) test to prescribe workout intensity. The 1RM test is the maximum weight lifted at one time in a controlled manner; it is considered a benchmark in dynamic strength evaluation.19, 22, 23, 24 However, several authors have reported difficulties in test execution and increased risk of injury depending on the population.19, 25, 26, 27 An alternative assessment involves estimating the maximum load through multiple-repetition maximum assessments such as 5RM or 10RM.23, 24, 27 The results found using these tests are variables used for 1RM predictive equations. These equations have been developed for different populations and situations involving trained and untrained individuals,25, 27, 28, 29 both sexes,19, 27, 28, 29 different types of strength machines30, 31 and the elderly.32 Some of the most commonly used 1RM predictive equations are from Epley,33 Lombardi,34 O’Connor et al.,35 Mayhew et al.,36 Brzycki37 and Baechle and Groves.38 Regarding motor impairment, only one study involving individuals with SCI has been conducted.26 Schwingel et al. found a good correlation between the 1RM test and predictive equation of 1RM using the 12-repetition maximum (12RM) test for the bench press and T-bar row machine for Paralympic rowers.26 However, no individual in the sample had a diagnosis of SCI.

The present study aimed (a) to test the cross-validation of current 1RM predictive equations in adults with SCI; (b) to compare the current 1RM predictive equations to a specific developed equation based on the 4- to 12-repetition maximum test (4–12RM) in individuals with SCI; and c) to verify whether relative and absolute 1RM loads are similar between injury levels. Considering correct posture stabilization and positioning adaptations in the test, the hypothesis of this study was that current predictive equations will not be applicable for individuals with SCI and a newly developed one, based on the specific population, will have a positive and significant correlation with the 1RM test.



In total, 45 patients with SCI were consecutively enrolled in the study. They were participants in a rehabilitation program of the Rehabilitation Hospitals and were recruited during the second week of rehabilitation. Before the rehabilitation program, the patients were not participating in physical activities. The data collection period was February 2013 to April 2016.

Patients unable to participate in the rehabilitation program, with a history of metabolic disorders and with a history of cardiovascular, cardiac, or orthopedic surgery that would restrict their ability to execute tests or perform correct exercise biomechanics, were excluded from the selection process. Therefore, the following inclusion criteria were used: male (older than 18 years), diagnosis of traumatic SCI, complete motor lesion (ASIA Impairment Scale [AIS] grade A or B),39, 40 clinical stability and wheelchair use. The International Standards for Neurological Classification of Spinal Cord Injury published by American Spinal Injury Association (ASIA) was used to assess motor level of the individuals with spinal cord injury by a trained physiotherapist.39, 40

Participants were stratified into three groups for analysis: tetraplegia group (TP; C6 to C8), high paraplegia group (HP; T1–T6), and low paraplegia group (LP; T7–L2). The first group (TP) had upper limb impairment, including reduced hand grip strength. Therefore, the assistance equipment was used to fix the hand press over the bar. The division between TP and paraplegia groups was in accordance with the ASIA tetraplegia classification.39, 40 The TP and HP groups had cardiovascular dysfunctions due to autonomic nervous system alterations and trunk instability mainly due to innervations of the abdomen (intercostal nerves T7 to T11). All groups had reduced or absent strength in the lower limbs.

Sample demographics are presented in Table 1.

Table 1 Subjects demographics and descriptive statistic of strength variables for tetraplegia, high paraplegia and low paraplegia groups


Before testing, participants were informed about all procedures and were instructed regarding the execution techniques.

Body composition assessment

On the first day of testing, participants underwent a body composition assessment. During this evaluation, body mass (BM), lean body mass (LBM), height (cm), skinfold sum (∑SF) and body fat percentage using the skinfold protocol were measured.41 Brachial biceps, brachial triceps and subscapular and suprailiac sites were used in the Durnin and Womersley skinfold equation for body fat percentage prediction.16, 41, 42, 43, 44 Pectoral, mid-axillary, abdominal, thigh and leg sites were also measured to calculate the skinfold sum.

Adaptation and familiarization

To adapt to weight training and to familiarize participants with the test, four exercise sessions were performed 2 weeks before the assessments.25, 45 Each session was performed with a minimum interval of 48 h (Figure 1).

Figure 1
figure 1

Protocol for 1RM predictive equation based on 4–12RM test.1RM, one-repetition maximum test; 4–12RM, four to twelve repetition test; Rep: repetitions.

During the first week, two sessions of two series of 12–15 repetitions to maximum with intervals of 1–2 min were performed for neuromuscular adaption. The exercises were performed using the following cable machines: bench press, peck deck, pull down, seated row and triceps pushdown. Dumbbells were used only during biceps exercises.

During the second week, two sessions of three series of 6–10 repetitions to maximum of bench press with a bar with intervals of 2–3 min were performed for test familiarization. Seated row, triceps pushdown and dumbbell biceps exercises were retained.

Assistance equipments such as bandages and neoprene strips were used for trunk and leg stabilization of all participants at the bench. For individuals with tetraplegia, assistance equipment was used to fix the hand press over the bar. The same adaptations were used during the familiarization phase and maximum strength tests.

Maximum strength tests

Maximum strength tests (1RM and 4–12RM) were executed using bench press exercises on a bench that was 26 cm wide and 123 cm long. This exercise is considered the best isolated assessment to predict total dynamic strength,22 upper limb strength19 and loads for tests and exercises.19 The barbell is 3.1 kg and 1.84 cm, and weights range from 0.5 to 20 kg. Patients were instructed to refrain from eating or smoking for 3 h before the tests, to not perform strenuous exercise for 6 h before the tests and to empty their bladder before the tests.

Bench press exercises were performed in the supine position with the participant’s feet on the ground. The hips and legs of the participants were stabilized with straps. Each repetition had four phases as follows: (1) extended elbows and hands holding the bar; (2) elbow flexion and horizontal shoulder extension (eccentric phase for approximately 2 s); (3) light touch of the barbell at the mesosternal point; and (4) elbow extension and horizontal shoulder flexion (concentric phase). During the first repetition, two physical educators put the barbell in the participant’s hands. Grip width was measured with elbows at 90 degrees and arms parallel to the ground. The mesosternal point was marked before execution, and no physical support was allowed during valid repetitions.

The maximum strength tests were assessed by the same tester and were performed randomly with 48- and 72-h intervals to avoid accumulated fatigue25, 28, 29 (Figure 1).

One-repetition maximum test

Before the 1RM test, participants performed a warm-up of 5–10 repetitions with 50% of the perceived maximum load. After 1 min of rest, 3 to 5 repetitions with 70% of the perceived maximum load were performed.22, 23, 24 Perceived maximum load was estimated based on researcher and participant perceptions regarding the four sessions of adaption and familiarization performed previously.

After the warm-up, the participant rested for 2 min, the load was increased and the exercise was performed. After a 5-min interval, the load was increased or decreased to allow only one repetition. The maximum number of attempts during the same session was five according to the procedure described in the literature22, 23, 24 (Figure 1).

Four- to twelve-repetition maximum test

The same warm-up protocol for the 1RM test was performed. The initial load used was ~80–90% of the maximum perceived load.22, 23, 24 The participants were instructed to perform the most possible lifts until concentric movement failure. The number of repetitions was supposed to be at least 4 and at most 12. Otherwise, a new attempt was performed after 5 min of rest. The load was decreased if the repetition number was less than 4 or increased if it was more than 12. The maximum number of attempts during the same session was five according to the procedure described in the literature.22, 23, 24 The load and repetitions were inserted at the following current predictive equations:

  • Eq1: Epley B33: 1RM=(0.0333 × load) × reps × load

  • Eq2: Lombardi P34: 1RM=(reps1) × load

  • Eq3: O'Connor R et al.35: 1RM=load x [1+(0.025 x reps)]

  • Eq4: Mayhew J et al36: 1RM=100 × load/(52.2+41.9 × exp (−0.055 × reps))

  • Eq5: Brzycki M37: 1RM=(load × (1.0278−(0.0278 × reps))1

  • Eq6: Baechle R and Groves E38: 1RM=load × ((0.0375 × reps)+0.978)

Statistical analysis

A sample of 45 patients was required to perform linear multiple regression analysis (fixed model) considering an effect size of 0.35 (average), α of 5% and power (1−β) of 90% with three predictors.

To assess the cross-validation, the intraclass correlation coefficient (ICC) with Bland–Altman plot was used to compare the current predictive equations with 1RM test. The ICC was classified based on Cicchetti standards as follows: below 0.40—level of clinical significance is poor; 0.40 to 0.59—fair; 0.60–0,74—good; and 0.75–1.00—excellent.46 Confidence intervals of 95% (95% CI) was used between comparisons.

To compare groups, one-way analysis of variance (ANOVA) with the Bonferroni post hoc test (P<0.016) was used when normal distribution was detected. The significance level was set at 0.016 to a more conservative analysis (0.05 divided by three comparison two by two). The Kruskal–Wallis test and Mann–Whitney post hoc test were used to compare non-normal variables.

Stepwise multiple regression analysis was used to generate specific equations for predicting 1RM load in individuals with SCI. These regression analyses maintain or remove the predictor variables for the best equation models. Three clusters of three variables were established. Each cluster was created as follows: cluster 1 (weight for 4–12RM), cluster 2 (repetitions for 4–12RM), cluster 3 (time since injury, body mass, body mass index, ∑SF, injury level, time since injury and age). Each variable from cluster 3 was entered into one stepwise regression with three predictors to determine 1RM load. Weight and repetitions for 4–12RM were maintained at all generated equations because these variables were reported at most of previous prediction equations. Injury level was considered as a dummy variable, indicating which group (TP, HP and LP) was used for a particular observation. The accuracy of the 1RM prediction was quantified using R-square and the standard error of the estimate.

The SPSS (version 22.0; IBM SPSS Statistics, Armonk, NY, USA) statistical package was used for the data processing. In the absence of multiple comparisons, statistical significance was set at P<0.016 (two-tailed).

Statement of ethics

We certify that all applicable institutional and governmental regulations concerning the ethical use of human volunteers were followed during the course of this research. The study was approved by the ethics committee of the Sarah Network of Rehabilitation Hospitals (number 13326213.7.0000.0022), and all patients provided written informed consent to participate in the study.


Descriptive results

All anthropometric and clinical variables, except time since injury, demonstrated no significant difference between groups. For the HP and LP groups, time since injury was statistically significantly different than that for the TP group (Table 1).

The HP and LP groups did not differ significantly regarding 1RM (kg), 4–12RM (kg), 4–12RM (repetitions), or relative strength (1RM divided by body mass) (Table 1). All these strength variables, except 4–12RM (repetitions) for the LP group, differed significantly between the TP group and the HP and LP groups (Table 1).

No injuries occurred during adaptation, familiarization and maximum strength tests.

Current predictive equation analysis

There were no significant differences between 1RM test and the current predictive equations (Table 2).

Table 2 One-repetition maximum test (1RM) results and the current predictive equations based on multiple maximum repetitions (kg)

ICC values were significant and were classified as excellent for all five current predictive equation, for all groups (Table 3). Additionally, the Bland–Altman plot demonstrated that both the difference between the means and the intervals around these differences (±1.96 times the s.d.) were small for all groups (Table 3).

Table 3 Bland and Altman method and intraclass correlation coefficient (ICC) comparing 1RM test and current predictive equations for groups

The pdrestimated 1RM results for the tetraplegia group (Tables 2 and 3 and Figure 2). For the low paraplegia group, the predictive equation results were overestimated (Tables 2 and 3 and Figure 2). In Figure 2, there are points plotted over and under Bland–Altman’s range limits for TP and LP, respectively.

Figure 2
figure 2

Bland and Altman method comparing 1RM test and current predictive equations for groups.1RM: one-repetition maximum test; EQ: equation; HP: high paraplegia group; LP: low paraplegia group; TP: tetraplegia group. A full colour version of this figure is available at the Spinal Cord journal online.

For total group, the predictive equation of Brzycki (Eq5)37 had the lowest absolute mean difference compared to 1RM test (1RM vs Eq5=0.4 kg), but the highest interval range around the differences (Δ1RM vs Eq5=16.8 kg) (Table 3 and Figure 2). O’Connor et al.35 predictive equation (Eq3) presented the lowest interval range around the differences but the highest mean difference (Δ1RM vs Eq3=12.7 and 1RM vs Eq3=1.8 kg, respectively) (Table 3 and Figure 2). The predictive equation of Lombardi (Eq2)34 presented the second best associated Bland–Altman results (0.5 kg and 12.8 kg for mean difference and interval range around the differences, respectively) (Table 3 and Figure 2).

Specific prediction equation

The two equation models created for 1RM demonstrated a high R-square (0.972 and 0.973, P<0.01, respectively) and low standard errors of the estimate for the measured 1RM (2.88 kg or 5.4% and 2.90 kg or 5.5% for models 1 and 2, respectively) (Table 4). The first model used only 4–12RM weight and repetitions; the second model also included the injury level (Table 4). For both equations, 4–12RM weight was the best predictor (0.98 and 0.99, P<0.01, for models 1 and 2, respectively) (Table 4). Therefore, we opted for an equation with fewer predictors as follows because it is clinically easier to use:

Table 4 R-square, adjusted R2 values, 95% CI and standard error estimate for 1RM test, weights (β) and probability values (p) for significant predictors of 4–12RM (kg), 4–12RM (repetitions) and injury level


Contrary to our initial hypothesis, current 1RM predictive equations can be used in individuals with spinal cord injuries to assess strength. The predictive equation of Lombardi34 presented the best associated cross-validity results. Additionally, an equation to predict 1RM strength based on the 4–12RM test for individuals with different levels of SCI was developed. The created predictive equation should be tested in order to verify whether it presents better accuracy than the current ones. The equation had a significant correlation with the 1RM test, and load was a better predictor than the number of repetitions.

Load is also the main predictor of other prediction equations. Whisemant et al.47 tested the validity of 11 prediction equations for 1RM and stated that increased equation precision is proportional to fewer repetitions and higher weight. Therefore, during the first attempt of the 4–12RM test, the researcher should estimate the initial load at 4 repetitions. This procedure will reduce error in the difference between the estimated result and the result obtained during the 1RM test.

Some studies have tested the validity of the prediction equations in other populations.26, 28, 31, 32, 48, 49 Knutzen et al.32 tested four estimation equations in older men and women and verified that Brzycki equation was the one that presented the highest correlation coefficient for bench press (r=0.89). Nascimento et al.48 analyzed the validation of Brzycki equation and suggested that the equation can be performed to assess 1RM in the bench press. Menêses et al.,49 otherwise, found that O’Connor equation produced the best estimation of 1RM for young adults with weight training experience.

In our study, all equations (Epley,33 Lombardi,34 O’Connor et al.,35 Mayhew et al.36 Brzycki37 and Baechle and Groves38) demonstrated the accurate and reliable data concerning bench press 1RM results in individuals with SCI. However, Lombardi34 has provided the best validity performance among them. Among the studies that have tested the validity of prediction equations,26, 28, 32, 48, 49 only Schwingel et al.26 used Lombardi’s prediction equation. The authors tested the predictive equation with nine individuals with motor disabilities;26 however, none of them had SCI. The results showed that all predictive equations underestimated 1RM of bench press, and Lombardi’s prediction equation had the highest mean difference (δ=18.1 kg).26 The tested subjects were international-level paralympic rowers, and the sample characteristics could have contributed for this difference. Nevertheless, the authors concluded that all tested equations were accurate.26

Although the current equations presented good accuracy, a new equation was created with the specific SCI population in order to be compared in future studies. All assessed groups had impairment or absence of leg strength and, therefore, reduced posture stabilization. Balance is worse for individuals with a higher injury level, such as the TP and HP groups, because of the reduced strength or absence of strength in the abdominal muscles.39, 40 Individuals with injury higher than the C7 myotome have impairment of triceps muscle strength; therefore, their bench press maximum load was reduced. In addition, the impaired trunk balance, reduced upper limb strength and absence of leg strength39, 40 could alter posture stabilization and correct positioning, and, consequently, could influence the test results. Therefore, further studies must test the specific predictive equation of the present study with another data set of individuals with SCI in order to confirm the internal validation of the results. Moreover, it would be important to determine the cross-validity of such a prediction equation compared to a gold standard (e.g., dynamometer isokinetic strength test). It also should be tested with female subjects with different SCI levels.

The present study sample had individuals who were not participating in physical activities before rehabilitation program. Even with a novice strength training population, there were no injuries reported by the patients, during any training sessions and assessments. Some authors have reported an increased risk of injury,19, 25, 26, 27 although their studies did not demonstrate incidence of injuries during tests' procedure. Therefore, the 1RM test, performed correctly, can be considered a safe strength assessment for individuals with SCI.

Some studies have revealed that generalized prediction equations are not applicable to specific populations;50, 51 however, the present study results showed that for individuals with SCI, the tested 1RM prediction equations are accurate for bench press. It is important to note that the pectoralis major, brachial triceps and deltoids are the primary muscles used during bench press exercises. This movement pattern has limitations when generalizing other tests of movement or functional independence.

Study limitations

The weight range (0.5–20 kg) was not sensitive enough to adequately adjust the load for individuals with higher lesion levels (TP). Lower weights should be used for more precise 1RM testing results for this specific group. Other study limitation was that the same tester assessed both 1RM and 4–12RM, i.e., the tester was not blinded of the maximum strength tests.


This study concluded that all 1RM predictive equations used in this study are accurate to assess individuals with SCI at the bench press exercise. However, the predictive equation of Lombardi presented the best-associated cross-validity results. A specific prediction equation to determine 1RM based on 4–12RM test in individuals with spinal cord injury was also proposed. Results of 1RM have practical value for allied health professionals in assessing and prescribing strength training programs. It is important to have an easy, non-expensive, safe and practical strength test for spinal cord injured individuals for clinical use during a rehabilitation program. Future studies could be conducted to determine the cross-validity of the elaborated prediction equation comparing it to a gold standard (for example, dynamometer isokinetic strength test).

Data archiving

There were no data to deposite.