Eight healthy active men participated in this study (means ± standard deviation: age, 30 ± 8 years; height, 181 ± 5 cm; body mass, 73 ± 4 kg). Subjects were informed of the experimental protocol and all associated risks prior to giving written informed consent as part of a medical inclusion. All procedures conformed to the Declaration of Helsinki and were approved by the Comité de Protection des Personnes Sud-Est 1, France.

Each subject completed one familiarization session and one experimental session. During the familiarization session, subjects were introduced to all procedures conducted in the experimental session and repeated trials until they performed all tests consistently and as directed. The largest MVC from the familiarization session was used to calculate contraction intensities and the reproducibility of MVC between sessions was verified.

Knee extensor force was measured during voluntary and evoked contractions by a calibrated force transducer (Meiri F2732 200 daN, Celians, Montauban, France) with amplifier that was attached by a non-compliant strap to the right leg immediately proximal to the malleoli of the ankle joint. Subjects were seated upright in a custom-built chair with both hips and right knee at 90° of flexion. The force transducer was fixed to the chair such that force was measured in direct line to the applied force. Electromyographic (EMG) activity of the right knee extensors (RF, vastus lateralis [VL] and vastus medialis [VM]) and flexors (biceps femoris, [BF]) was recorded.

EMG activity was recorded with a pair of self-adhesive surface (10-mm recording diameter) electrodes (Meditrace 100, Covidien, Mansfield, USA) in bipolar configuration with a 30-mm interelectrode distance and the reference on the patella. Low impedance (<5 kΩ) between electrodes was obtained by shaving, gently abrading the skin with sandpaper and then cleaning it with isopropyl alcohol. Signals were analogue-to-digitally converted at a sampling rate of 2000 Hz by PowerLab system (16/30—ML880/P, ADInstruments, Bella Vista, Australia) and octal bio-amplifier (ML138, ADInstruments) with bandpass filter (5–500 Hz) and analyzed offline using Labchart 7 software (ADInstruments).

Single electrical stimuli of 1-ms duration were delivered via constant-current stimulator (DS7A, Digitimer, Welwyn Garden City, Hertfordshire, UK) to the right femoral nerve via a 30-mm diameter surface cathode in the femoral triangle (Meditrace 100, Covidien, Mansfield, USA) and 50 x 90 mm rectangular anode (Durastick Plus, DJO Global, Vista, USA) on the gluteus maximus. Single stimuli were delivered incrementally until plateaus in maximal M-wave (Mmax) and twitch amplitude were reached. Three supramaximal stimuli at 130% of the intensity to produce maximal Mmax and twitch responses (52 ± 9 mA) were delivered at rest.

Single-pulses (0.1-ms rise time; 1-ms duration) were manually delivered by TMS to elicit MEPs and twitches in the right knee extensors. The contralateral motor cortex was stimulated by a magnetic stimulator (Magstim 200², The Magstim Company Ltd, Whitland, UK) with 110-mm double-cone coil (maximum output of 1.4 T) to induce a postero-anterior current. The coil was manually controlled by an experienced investigator throughout the protocol. Subjects wore a cervical collar during all TMS measures to stabilize the head and neck.

Subjects wore a latex swim cap on which lines were drawn between the preauricular points and from nasion to inion to identify the vertex. Every centimeter from 1 cm anterior to 3 cm posterior to the vertex was demarcated along the nasal-inion line and also to 2 cm over the left motor cortex. At each point a stimulus was delivered at 70% maximal stimulator output during brief voluntary contraction of the knee extensors at 10% MVC. Target force was displayed on a screen and subjects were provided with real-time visual feedback during all voluntary contractions throughout the protocol. The coil was positioned at the site evoking the largest VL (39.5 ± 19.2% Mmax), RF (75.9 ± 26.7% Mmax) and VM (45.0 ± 21.3% Mmax) MEP amplitudes and SIT with minimal BF MEP amplitude. This coil position was drawn directly onto the swim cap and used throughout the protocol. Coil position was also verified before the delivery of each stimulus.

Four methods of determining stimulus intensity were investigated in the following order: 1) RMT: Beginning at 30% of maximal stimulator output and increasing by 5% increments to 80%, subjects received 10 stimuli at each stimulus intensity with the knee extensors completely relaxed. Stimuli were delivered at 10-s intervals. 2) AMT/stimulus–response curve at 10% MVC: Subjects performed brief voluntary contractions (~2-3 s) of the knee extensors with TMS delivered 10 times at 20, 25, 30, 35 and then 40% of maximal stimulator output. Subjects then performed brief contractions with TMS delivered 4 consecutive times at each of the following randomly-ordered stimulus intensities: 50, 60, 70 and 80% of maximal stimulator output. All stimuli were delivered at 15-s intervals. 3) Stimulus–response curve at 20% MVC: Subjects performed brief contractions (~2-3 s) of the knee extensors with TMS delivered 4 consecutive times at each of the following randomly-ordered stimulus intensities: 20, 30, 40, 50, 60, 70 and 80% maximal stimulator output. Stimuli were delivered at 15-s intervals. 4) Stimulus–response curve at 50% MVC: Similar to the stimulus–response curve at 20% MVC except that stimuli were delivered at 20-s intervals. During voluntary contractions, TMS was always delivered once the subject had contracted to the appropriate force level and the force had stabilized 32 and 10 min rest was provided between each of the four methods.

Peak-to-peak MEP and Mmax amplitudes were measured offline for each individual response. Individual MEP and Mmax amplitudes were then averaged and MEP amplitudes were normalized to Mmax amplitudes evoked in relaxed muscle. Data collected from a similar group of subjects in our laboratory indicated Mmax amplitudes were similar at rest and at the contraction intensities employed in this study (i.e. up to 50% MVC, unpublished observations, 2012). RMT was determined as the lowest stimulus intensity producing at least 5 MEPs of at least 0.05 mV from 10 stimuli. RMT was also determined from 6 and 8 stimuli (minimum of 3 and 4 MEPs, respectively). Stimulus intensities of 120 and 130% RMT were determined for comparison with methods used in other lower-limb studies 9 11 13 14 27 30 33 . AMT was determined by visual identification of MEPs above background EMG from contractions at 10% MVC 34 and corresponded to the lowest stimulus intensity producing MEPs in at least half the contractions. Classically, fixed thresholds are used to determine the presence of a MEP (i.e. 0.2 mV at 10% MVC 12 ); however, the large variability in background EMG activity for the three measured quadriceps muscles rendered this method impractical. AMT was also determined from 6 and 8 stimuli (minimum of 3 and 4 MEPs, respectively). The stimulus intensity of 120% AMT was determined for comparison because of its use in other lower-limb studies 10 12 . Stimulus–response curves at 10, 20 and 50% MVC were used to determine stimulus intensity by identifying the minimum stimulus intensity to evoke maximal MEP amplitude (i.e. the lowest intensity resulting in an increase of less than 5% MEP amplitude at higher stimulus intensities). Individual MEPs from a typical stimulus–response curve at 20% MVC for one subject are presented in Figure 1. Antagonist MEP amplitude was examined to verify that this stimulus intensity did not elicit increased TMS-induced coactivation. For the 10% MVC stimulus–response curve, only the first 4 stimuli at 20, 30 and 40% maximal stimulator output were considered. Where a plateau was not reached, MEP amplitude at 80% maximal stimulator output was compared to the estimated maximal MEP amplitude from Boltzmann modeling (see next paragraph). If mean MEP amplitude was greater or equal to the maximal modeled MEP amplitude, 80% was accepted as being part of the plateau and selected as the appropriate stimulus intensity. Otherwise, a plateau was determined to not have occurred and the data was excluded from analyses.

MEP amplitude from stimulus–response curves were modeled with a Boltzmann sigmoidal function 35 using the equation:

MEPmax S = MEPmax 1 + exp S 50 - S k

where MEPmax is the estimated maximal MEP amplitude, S is the stimulus intensity, S50 is the stimulus intensity required to produce a response equal to half MEPmax and k is the slope parameter (inversely proportional to maximal function steepness). To eliminate the effects of background EMG in the modeling process, an amplitude of 0 mV was assigned to all responses in which there was no discernible MEP.

Statistical analyses were performed with Statistica (version 8, Tulsa, USA). The Shapiro-Wilk test was used to verify data normality. One-way repeated measures analyses of variance (ANOVA) were used to evaluate the method of stimulus determination (120 and 130% RMT, 120% AMT and stimulus–response curves), any difference between muscles and the effect of contraction intensity on Boltzmann parameters. One-way repeated measures ANOVA were also used to compare AMT and RMT determined from 6, 8 and 10 stimuli. When ANOVA revealed significant interactions, the Newman-Keuls post-hoc test was used to identify differences. The effect size as determined from the ANOVA were calculated as ω² to reduce potential bias associated with the small sample size 36 . Dependent t-tests were used to compare Boltzmann and linear relationships for the coefficient of determination of MEP amplitude. Statistical significance was set at p < 0.05. All data are expressed as means ± standard deviation except Figure 2 where values are expressed as means ± standard error of the mean.

One subject did not reach a plateau in MEP amplitude in RF with the 10% MVC stimulus–response curve and was thus excluded from all relevant analyses.

Neither AMT nor RMT were different whether determination occurred with the first 6, 8 or 10 responses at each stimulus intensity for any muscle (p > 0.05). Therefore all subsequent analyses were conducted based upon AMT and RMT determined from 10 stimuli at each stimulus intensity. Selected TMS intensity determined by RMT, AMT and stimulus–response curves are presented in Figure 2. Stimulus intensities determined from RMT (120 and 130%) and stimulus–response curves at 10% MVC were higher than the intensity determined by stimulus–response curve at 50% MVC (VL: F(5,35) = 8.54, p < 0.001, ω² = 0.48; RF: F(5,30) = 8.13, p < 0.001, ω² = 0.50; VM: F(5,35) = 7.69, p < 0.001, ω² = 0.45). Stimulus intensity at 120% AMT was lower than stimulus intensity determined from stimulus–response curves at both 10 and 20% MVC (p < 0.05). Table 1 presents the selected stimulus intensities from the stimulus–response curves as a percentage of the stimulus intensity to elicit both RMT and AMT to contextualize the differences between these methods. There was also no difference in selected intensity between muscles for any method (RMT: F(2,14) = 2.62, p = 0.11, ω² = 0.16; AMT: F(2,14) = 1.21, p = 0.33, ω² = 0.02; 10% MVC: F(2,12) = 1.00, p = 0.40, ω² = 0.00; 20% MVC: F(2,14) = 1.15, p = 0.35, ω² = 0.02; 50% MVC: F(2,14) = 0.778, p = 0.48, ω² = 0) nor difference in normalized MEP amplitude at the selected stimulus intensity between 10, 20 and 50% MVC stimulus–response curves (VL: F(2,14) = 3.23, p = 0.07, ω² = 0.21; RF: F(2,12) = 2.48, p = 0.13, ω² = 0.16; VM: F(2,14) = 2.81, p = 0.09, ω² = 0.18) (Table 2,). Raw BF MEP amplitudes at the selected stimulus intensities were 0.51 ± 0.54, 0.53 ± 0.41 and 0.53 ± 0.41 mV for VL, 0.42 ± 0.47, 0.43 ± 0.41 and 0.54 ± 0.31 mV for RF and 0.40 ± 0.45, 0.45 ± 0.40 and 0.59 ± 0.42 mV for VM for 10, 20 and 50% MVC stimulus response curves, respectively. At the stimulus intensity selected by AMT determined during contractions at 10% MVC, raw BF MEP amplitudes were 0.30 ± 0.41, 0.28 ± 0.42 and 0.29 ± 0.42 mV for VL, RF and VM, respectively. A single stimulus–response curve at 50% MVC is presented in Figure 3.

AMT: active motor threshold; MVC: maximal voluntary force; RMT: resting motor threshold. For rectus femoris, values from the stimulus–response curve at 10% MVC are n = 7. Values are expressed as means ± standard deviation and (range).

MVC: maximal voluntary force. For rectus femoris, values from the stimulus–response curve at 10% MVC are n = 7. Normalized motor-evoked potential amplitudes are expressed as means  ±  standard deviation and (range).

In VL, RF and VM, Mmax amplitudes were 16.2 ± 4.1 mV, 7.4 ± 1.8 mV and 17.0 ± 6.7 mV, respectively. Central drive as indicated by RMS · Mmax^-1 for VL (0.0046 ± 0.0014), RF (0.0039 ± 0.0007) and VM (0.0053 ± 0.0025) at 10% MVC and VL (0.0088 ± 0.0024), RF (0.0086 ± 0.0019) and VM (0.0100 ± 0.0039) at 20% MVC were similar (F(2,14) = 1.32, p = 0.30, ω² = 0.05, and F(2,14) = 0.660, p = 0.53, ω² = 0, respectively). At 50% MVC, RMS · Mmax^-1 for RF (0.0376 ± 0.0160) was greater than for both VL (0.0237 ± 0.0094) and VM (0.0264 ± 0.0115) (F(2,14) = 8.36, p = 0.004, ω² = 0.00).

Boltzmann curves from a typical subject are presented in Figure 4. Boltzmann curves provided a significantly better fit for the relationship between MEP amplitude and stimulator intensity than a linear relationship for stimulus–response curves at 10, 20 and 50% MVC for all muscles (p < 0.05). As contraction intensity increased, S50 decreased in all muscles (VL: F(2,14) = 33.1, p < 0.001, ω² = 0.79; RF: F(2,14) = 55.6, p < 0.001, ω² = 0.87; VM: F(2,14) = 32.5, p < 0.001, ω² = 0.79). Few differences were observed in MEPmax · Mmax^-1 (only RF lower at 10% MVC; VL: F(2,14) = 1.88, p = 0.19, ω² = 0.09; RF: F(2,14) = 3.88, p = 0.046, ω² = 0.25; VM: F(2,14) = 2.40, p = 0.13, ω² = 0.14) and k (only VL lower at 10% MVC; VL: F(2,14) = 7.50, p = 0.006, ω² = 0.43; RF: F(2,14) = 1.62, p = 0.23, ω² = 0.07; VM: F(2,14) = 0.911, p = 0.42, ω² = 0). Results from modeling the stimulus–response curve data with the Boltzmann equation are presented in Table 3.

MEPmax · Mmax^-1: maximal motor-evoked potential amplitude (MEPmax) normalized to maximal M-wave amplitude, MVC: maximal voluntary force, S50: stimulus intensity to evoke motor-evoked potentials half the amplitude of MEPmax (as % maximal stimulator output), k: slope parameter (inversely proportional to maximal function steepness), r²: coefficient of determination. Values are expressed as means ± standard deviation. Significantly different from 50% maximal voluntary force (MVC), * (p < 0.05) and ** (p < 0.01); significantly different from 20% MVC, # (p < 0.05) and ## (p < 0.01); significantly different from linear relationship, † (p < 0.05) and ‡ (p < 0.01).

Studies determining stimulus intensity during contractions have often used normalized MEP amplitude or area of a given size as criteria 7 8 19 20 . For example, Sidhu et al. 20 selected an intensity that produced the largest RF MEP with the stipulations that this must be at least 50% Mmax and that antagonist BF MEP amplitude be less than 10% raw RF MEP amplitude. In VL and VM in the present study, only 2 of 8 and 3 of 8 subjects, respectively, satisfied the requirement that MEP amplitude be ≥50% Mmax. In the case where several quadriceps muscles are examined, the latter criterion is ambiguous. BF may often be greater than 10% raw MEP amplitude in at least one knee extensor muscle and for one subject in the present study this was the case in all muscles at almost all stimulus intensities evaluated and also at almost all coil positions evaluated in the determination of optimal coil position.

There was no difference in selected stimulus intensity between muscles as determined by any method. This suggests that one muscle could be used as a surrogate for other quadriceps muscles. RF alone has frequently been used to determine quadriceps stimulus intensity 6 8 19 20 . When RF is normalized, MEP amplitude is larger than for either VL or VM due to consistently smaller Mmax in the RF and little differences in raw MEP amplitude. The presentation of normalized RF MEP amplitudes instead of VL and VM may give the impression of eliciting greater corticospinal drive to the quadriceps muscles. In the present study, this was not due to a greater RF contribution since RMS · Mmax^-1 was only greater than that of VL and VM at 50% MVC and normalized RF MEPs are larger than VL and VM at all contraction intensities. RF may not be an ideal surrogate because of the difficulty in recording clear M waves in this muscle. Furthermore, RF is the sole biarticular muscle of the quadriceps femoris, and thus, may not be representative of the muscle group.

An important limitation to the protocol is that it was not conducted on a second day to investigate the day-to-day variability of the methods employed in this study. Further investigations are required to establish whether the different methods employed to evaluate TMS parameters with fatigue are reproducible on different days. The present study also used a maximal response in contracting muscle as a reference point to evaluate multiple fatigue-related TMS parameters since this provides important insights into the manifestation and development of fatigue; however, recent studies suggest that TMS responses elicited by a submaximal stimulus intensity may also further understanding of fatigue mechanisms 1 2 3 . This reinforces the necessity of selecting an appropriate method to determine TMS intensity directly related to the parameters being investigated. In the context of the evaluation of cortical voluntary activation and corticospinal excitability with fatigue, maximal responses as investigated in this study are pertinent. In other research and diagnostic areas employing TMS, this may not be the case, and methods such as RMT and AMT may be the methods of choice for determining optimal stimulus intensity. Further studies must also determine the specific relevance of TMS-induced maximal and submaximal responses in both healthy and clinical populations in the context of fatigue, including the manner in which this affects measures of cortical voluntary activation and both excitatory and inhibitory mechanisms.