Ankle strength is an important factor for ambulation and balance [1, 2] and should therefore be closely monitored. A novel dynamometer has been developed in order to accurately measure both ankle dorsi-flexion and plantar-flexion strength in children and adults with neuromuscular disorders. Normative data were established and predictive models computed for both children and adults. The device was shown to have a good level of reliability with low relative standard errors of measurement between test and retest sessions. The feasibility to use this novel dynamometer on patients was confirmed on a small sample of patients affected with calpainopathy.
Feasibility of the measurements
The device was both easy to use and transport. Its size makes it usable for children, as young as five years old, as well as for elderly people as old as eighty years of age. The device allows one to measure torque values up to 200 N⋅m with a very good stabilization of the joint even for ankle plantar-flexion which is a function difficult to measure due to the extent of its strength, even for children . A few subjects complained of discomforts but none refused to perform the measurements. The reluctance of some to perform at their maximal strength on the first trial may be partly explained by the apprehension of the subjects to hurt themselves.
We found that the right side was significantly stronger than the left side independently of the laterality determined by hand dominance. The hand dominance does not seem to be a good indicator for leg dominance. Indeed the recommendation given in the literature is to establish foot preference by asking the subject which leg he or she would choose to kick a ball . This method was applied by several authors who found that there was no significant differences in force between the dominant and the non-dominant side in dorsi-flexion strength [20, 32] or between the left and right leg . However we found that the right side was significantly stronger than the left side; this result is comparable to that obtained by other authors [25, 33].
In adults, the dorsi-flexion torque values found in this study are comparable to those found in other studies where the ankle joint was flexed at 90° [18, 20, 34]. In other studies, where the ankle joint was not chosen at 90°, dorsi-flexion and plantar-flexion torque values may differ slightly [13, 14, 27]. These differences observed between different studies may be attributed to differences in the population tested and to the test protocol itself especially the positioning of the subject. Indeed, ankle plantar-flexion torque has been demonstrated to be greater when the knee was extended rather than flexed to 90° [21–23]. Moreover it was shown that ankle plantar-flexion torque is maximal when the ankle joint is dorsiflexed by 15°  and ankle dorsi-flexion torque is maximal when the ankle joint is plantarflexed by 10-15° [16, 19]. In children, results for dorsi-flexion and plantar-flexion torque values in our study differ slightly from other studies [12, 24] probably due to discrepancies in test protocols already mentioned above and due to difference in stature that can have a great impact on strength.
Torque values found in our study are therefore probably lower than the maximum value possible when both the ankle and knee position are optimized. However, the seated position was chosen over other positions in order to minimize the possibilities of co-contractions and to enable as many subjects as possible to perform the measurements even in the presence of contractures.
Most authors have used age, weight, sex and sometimes height as predictive variables for ankle strength [12, 32, 34, 35]. In this study we found that height was the best predictor of strength for both children and adults and that other variables added little to the model. Using height to predict strength in children may enable one to assess muscle weakness in children with muscle disabilities associated with growth disorders. If a child is small for his age, prediction will make it possible to establish the part of muscle weakness that is due to disease and not growth retardation. Eek et al.  considered age, weight and sex as the predicting variables of ankle dorsiflexors strength in children and found determination coefficients comparable to ours. They indicated that sex was included in the model when the age was over 13. As Hogrel et al. , we found that when height was taken into account, no gender effect was seen in the strength of children. In adults, the model predicting dorsi-flexion torque in our study has similar determination coefficients compared to other models [32, 34, 35]. The quality of our plantar flexion model could not be assessed either in children or adults as no other study computing this model could be found. The difference in determination coefficients between dorsi-flexion and plantar-flexion models found in our study can be interpreted as a higher variability in plantar-flexion strength rather than dorsi-flexion strength among adults. Even though our model only explains a small amount of plantar-flexion variability, it may be useful in evaluating strength deficit for this function as, to our knowledge, no other model has been established.
The significant positive differences between retest and test sessions suggest that there might be a minor learning effect in performing maximal voluntary contractions for ankle movements. This phenomenon may also be explained by the apprehension of the subject to hurt himself during the first trial as was previously mentioned or by the fact that this function is not performed isometrically in everyday life and can therefore be difficult to perform at first. This increase in strength during the retest session on plantar-flexion strength measurements has been observed by several authors [14, 15]. Standard error of measurements and ICCs found in our study are similar or higher to those found in other studies assessing reliability in dorsi-flexion or plantar-flexion strength [14, 25, 34]. Due to the possible learning effect in assessing ankle strength, a training session should be recommended in clinical trials where there is a follow up of patient strength over time. A training session would also be useful in evaluating the smallest detectable differences that is relevant to the patient population studied. Indeed the smallest detectable difference found in our study was established in healthy subjects and may not be a good criterion for detecting changes in subjects with much lower strength values as it can be noticed in our results (see Figure 3) that the difference between test and retest session is higher for higher strength values. Moreover, the variations in subsequent measurements depend on motivation, skill, learning effect and several individual variables that are complex to analyse. This can affect estimates of reliability.
Feasibility in a small patient group
The dynamometer was applied without particular difficulties to nine patients with limb girdle muscular dystrophy. Values of maximal torque as low as 1.0 N⋅m could be recorded. Using the predictive equations established in this study, all subjects were shown to have a significant strength deficit compared to the normal population. Therefore these quantitative strength measurements relative to the normal population highlight the strength deficit of these patients in a quantitative way. The device could therefore be used in clinical trials for the follow up of different pathologies in which leg muscles are affected, even severely. Note that most of the patients present a strength that lies under the smallest detectable difference, which may be of limited pertinence in this case.
Not all subjects had a retest session, meaning that norms and predictive equations were established on test values which might not be representative of the maximum values as a learning effect was demonstrated on a subgroup of patients.