Ò

PAIN 153 (2012) 593–601

www.elsevier.com/locate/pain

Compound action potentials recorded in the human spinal cord during

neurostimulation for pain relief
John L. Parker a,b,⇑, Dean M. Karantonis a, Peter S. Single a, Milan Obradovic a, Michael J. Cousins c
Downloaded from http://journals.lww.com/pain by BhDMf5ePHKav1zEoum1tQfN4a+kJLhEZgbsIHo4XMi0hCywCX1AWnYQp/IlQrHD3i3D0OdRyi7TvSFl4Cf3VC1y0abggQZXdgGj2MwlZLeI= on 03/28/2022

a
National Information and Communications Technology Australia, Eveleigh, NSW 2015, Australia
b
Graduate School of Biomedical Engineering, University of New South Wales, Kensington, NSW 2052, Australia
c
Pain Management Research Institute and Kolling Institute, University of Sydney at the Royal North Shore Hospital, St Leonards, NSW 2065, Australia

Sponsorships or competing interests that may be relevant to content are disclosed at the end of this article.

a r t i c l e i n f o a b s t r a c t

Article history: Electrical stimulation of the spinal cord provides effective pain relief to hundreds of thousands of chronic
Received 6 August 2011 neuropathic pain sufferers. The therapy involves implantation of an electrode array into the epidural
Received in revised form 20 November 2011 space of the subject and then stimulation of the dorsal column with electrical pulses. The stimulation
Accepted 21 November 2011
depolarises axons and generates propagating action potentials that interfere with the perception of pain.
Despite the long-term clinical experience with spinal cord stimulation, the mechanism of action is not
understood, and no direct evidence of the properties of neurons being stimulated has been presented.
Keywords:
Here we report novel measurements of evoked compound action potentials from the spinal cords of
Neuromodulation
Neuropathic pain
patients undergoing stimulation for pain relief. The results reveal that Ab sensory nerve fibres are
Physiological measurement recruited at therapeutic stimulation levels and the Ab potential amplitude correlates with the degree
Spinal cord stimulation of coverage of the painful area. Ab-evoked responses are not measurable below a threshold stimulation
level, and their amplitude increases with increasing stimulation current. At high currents, additional late
responses are observed. Our results contribute towards efforts to define the mechanism of spinal cord
stimulation. The minimally invasive recording technique we have developed provides data previously
obtained only through microelectrode techniques in spinal cords of animals. Our observations also allow
the development of systems that use neuronal recording in a feedback loop to control neurostimulation
on a continuous basis and deliver more effective pain relief. This is one of numerous benefits that in vivo
electrophysiological recording can bring to a broad range of neuromodulation therapies.
Ó 2011 International Association for the Study of Pain. Published by Elsevier B.V. All rights reserved.

1. Introduction ameliorates pain through gamma-aminobutyric acid (GABA)-ergic

and adenosine related inhibitory mechanisms [12,17]. They propose
Spinal cord stimulation (SCS) induces tactile paraesthesia, a that electrical stimulation produces orthodromic and antidromic
pleasant tingling sensation, as a result of the stimulation of nerve action potentials [12] and that the antidromic activity regulates
fibres in the dorsal column (DC). The qualitative description of transmission of pain via an interneuron pool and second-order
the induced paraesthesia has been used to hypothesise that Ab fi- wide-dynamic-range neurons (Fig. 1). There is also evidence that
bres are recruited during SCS [17]. The paraesthesia supplants the supraspinal mechanisms may play a role in pain relief [4]; however,
feeling of pain in the body areas innervated by the stimulated fi- the role of ascending activation vs local activation at the segmental
bres. The goal of SCS is to completely cover—in a perceptual level is currently not understood.
sense—the area of pain with paraesthesia because high levels of Evidence for the recruitment of particular fibre types during SCS
coverage are essential for effective pain relief [1]. has previously been restricted to simulated computer models.
SCS was first attempted [23] in the 1960s after research into the Holsheimer [8] concluded that SCS recruits a small number of Ab
gate control theory of pain, where it was observed that nonnoxious fibres (approximately 60), with a diameter between 9.4 and
(eg, touch and vibration) activation of superficial fibres in the DC of 10.7 lm, in the DC (at the T11 segment). Because the DC is inner-
the spinal cord inhibits pain transmission by a gate at the spinal seg- vated by 12 dermatomes at this level, Holsheimer concluded that
mental level [16]. Meyerson and colleagues have postulated that SCS there are only 4–5 fibres per dermatome recruited during SCS
[8]. The relatively small number of fibres (4–5 per dermatome)
with sufficient diameter in the superficial DC has been confirmed
⇑ Corresponding author. Address: NICTA Implant Systems, Level 5, 13 Garden St.,
Eveleigh, NSW 2015, Australia. Tel.: +61 2 9376 2125; fax: +61 2 9376 2031.
in microscopic analysis of human spinal cord sections [5]. Feira-
E-mail address: john.parker@nicta.com.au (J.L. Parker). bend et al. studied the diameter and distribution of fibres in the

0304-3959/$36.00 Ó 2011 International Association for the Study of Pain. Published by Elsevier B.V. All rights reserved.
doi:10.1016/j.pain.2011.11.023
 Ò
594 J.L. Parker et al. / PAIN 153 (2012) 593–601

Fig. 1. Schematic representation of the spinal cord (A, B) and activation of the Ab neurons via DC stimulation (C). Ab fibres enter the DC and project a descending collateral
branch, the ascending branch travels up the DC towards the brain (C). Electrical stimulation of the ascending Ab fibre results in antidromic activation of the collateral branch,
which is inhibitory to the activity of the wide dynamic range neuron (C). This in turn inhibits transmission of C fibre information to the brain. The electrode is implanted
midline above the DC and selectively stimulates the Ab fibres because they are larger, closer and more sensitive to stimulation than pain fibres that ascend via Lissauer’s tract,
which is further from the midline (B).

superficial DCs and concluded that 0.25% of total fibres have a The choice of stimulating electrode and initial stimulation param-
diameter above 10.7 lm within 300 lm of the DC surface [5]. eters were made during this session with the goal to maximise the
Current theory on the mechanism of pain relief via SCS can be overlap of paraesthesia generation with the painful area. After this
summarised as follows: (1) Stimulation in the DC elicits compound session, the electrodes were connected via custom fabricated con-
action potentials that propagate antidromically into the dorsal nector box to a Tucker-Davis Technologies (Alachua, FL) RZ5 ampli-
horn (where Ab fibres synapse with wide-dynamic-range neurons fier and bioprocessor system and a World Precision Instruments
in inhibitory fashion); (2) orthodromic activity of Ab fibres may (Sarasota, FL) A385 current source. The current source was con-
play a role in pain suppression; and (3) only a small number of fi- nected to the stimulation electrodes chosen during the initial
bres are stimulated in the DC. adjustment phase, and all other electrodes were connected to the
The objective of this study was to establish the feasibility of bioamplifier. The signals recorded are differential measurements
recording spinal cord evoked compound action potentials (ECAPs) with the tip of the electrode of the most rostrally placed lead used
in humans undergoing SCS for pain relief. The measurements yield as the reference. Output from the current source was a charge-bal-
detailed information about the fibres recruited during SCS and pro- anced biphasic waveform with the pulse width for a single phase
vide what is to our knowledge the first direct electrophysiologic and frequency (Table 1). Custom software was used to control
evidence that SCS recruits large-diameter, high-conduction-veloc- the measurement and initiate stimulus. A 3-lead electrocardio-
ity Ab fibres in the DCs of the human spinal cord. gram (ECG) was also recorded by the system on a separate record-
ing channel.
2. Methods System safety was assessed under the human ethics committee
protocol. The maximum stimulation level was set to the maximum
2.1. Patient selection and surgery level obtained for effective pain relief during the routine clinical
adjustment of the system.
As part of a human research ethics committee-approved protocol, Data were recorded continuously during experimental sessions.
permission was given to recruit for this study 5 patients undergoing During each session the amplitude of the stimulus was varied from
routine trial SCS as assessment for implantation of an SCS system. zero until patients reported complete coverage of their pain area
Two standard octopolar (8 electrode) leads were implanted according with paraesthesia or a maximum level of tolerance. The stimulus
to standard clinical procedures under fluoroscopic examination. In all current was adjusted in steps by manual rotation of a potentiome-
patients both the 8-electrode leads were placed midline in the epidu- ter. The size of the current increment was derived from the maxi-
ral space of the spinal cord by fluoroscopic examination, with one 8- mum current level, such that at least 10 levels were used, with an
electrode lead placed superior to the other such that both leads increment of no more than 1 mA. At each stimulus level patients
together formed a linear 16-channel array. This placement was cho- recorded and matched areas of SCS-induced paraesthesia to preex-
sen to maximise the longitudinal distance over which recordings isting areas of pain by colouring body maps. Also, a visual analogue
could be made. The details of the patients, type and origin of pain, scale pain score was recorded before SCS and at each interval
electrical stimulation parameters and neural responses at the percep- where the stimulus level was changed.
tual threshold and at therapeutic levels are detailed in Table 1. Recorded potentials and ECG data were streamed continuously
to a hard disk in the recording system PC. The amplifier output sig-
2.2. Stimulation and recording nals were sampled at 24.4 kHz, after passing through an antialias-
ing filter with a 7.5 kHz cutoff frequency (3 dB corner, first order,
Electrical stimulation parameters were chosen postoperatively 6 dB per octave). The data were stored as time-stamped blocks
by routine adjustment methods with feedback from the patient. and processed off line with custom software (Matlab; Math Works,
 Ò
J.L. Parker et al. / PAIN 153 (2012) 593–601 595

Table 1
Patient information, stimulation inputs, and measured parameters from evoked response recording.a

Patient Sex Cause and area of Character of pain Type At percept threshold At comfort level Conduction Estimated
no. and pain of velocity, fibre
Ab Stimulus No. of Ab Stimulus
age, pain ms 1 diameter,
amplitude, parameters dermatomes amplitude, current,
y lm
lV (current, pulse with lV mA
width, rate) paraesthesia
1 F, FBSS; sacral, Burning, throbbing N 17.4 5 mA, 80 ls, 10 150 10.0 49 8.1
65 buttock, lower stabbing 40 Hz
limb
2 F, CPRS I due to Tingling, stinging, N 18.9 4.3 mA, 120 ls, 5 36 4.6 65 10.8
38 ankle/knee burning, throbbing 40 Hz
injury; left leg
below knee
3 F, FBSS; lower back, Buzzing, tingling N 0.0b 0.93 mA, 8 310 2.6 52 8.6
32 bilateral leg hyperalgesia to 120 ls, 60 Hz
touch
4 F, Disc prolapse due Dragging, pulling, N 14.7 9 mA, 200 ls, 8 54 11.5 59 9.8
64 to lifting injury; ‘‘electric sensations’’ 60 Hz
lower back, right radiating into leg
leg burning
a
The percept threshold represents the point at which the patient first perceives the paraesthesia as the stimulation current is increased. The current level and the Ab
amplitude at this point are recorded along with the pulse width and stimulation frequency. The comfort level is defined as the maximum stimulus level obtained before a
report of an uncomfortable stimulation. This is the stimulation strength where the individual received the maximum pain relief benefit, and the stimulation current and Ab
evoked amplitudes are recorded. Conduction velocities were determined from the negative peak position as illustrated in Fig. 6. FBSS = failed back surgery syndrome;
CPRS = complex regional pain syndrome; N = neuropathic.
b
No discernible Ab response was measured in patient 3 at percept threshold.

Natick, MA). Postprocessing consisted of averaging data frames and

interpolation. A data frame consisted of a digitised waveform from
the period delimited by the start of a stimulus until 1 ms before the
next stimulus. The ECG affected the quality of the ECAP recording,
and its detection was based on its magnitude exceeding a thresh-
old value. This threshold was adjusted to the minimum value re-
quired to ensure reliable detection and minimise the number of
falsely positive detections. The remaining ECG-free data were
ensemble averaged across approximately 80 frames (depending
on the number of ECG-affected frames that were discarded) to im-
prove the signal-to-noise ratio. This averaged waveform was then
interpolated by a factor of 5 and passed through a 3 kHz-cutoff
low-pass filter (finite impulse response (FIR) order 51); the result
was subsequently used for display and ECAP propagation analysis.

3. Results

3.1. ECAP recording characteristics

The electrical response from the electrodes contains compo-

nents induced by the stimulus current and components that are
neurologic in origin. The stimulus-induced component is com-
monly referred to as electrode artefact, and its properties depend
on stimulation parameters, electrode materials and the measure-
ment amplifier [15]. Electrode artefact can completely obscure
the recorded neural response; a large body of literature describes Fig. 2. Measurements obtained on electrodes in the spinal cord for patient 1 after a
techniques to address this issue [2,3,19]. In this study, the bioam- biphasic stimulus pulse. The end of the stimulus phase is indicated by the dashed
lines. (A) The measurement artefact (F1) is negative after the positive second phase
plifier system recorded artefact as an exponentially decaying signal
of the stimulus. (B) The artefact (F1) is positive after a negative second phase for the
that occurred at the end of the stimulus. Fig. 2 illustrates the arte- stimulus. The series of waveform features indicated by F2 are propagating ECAPs.
fact for a biphasic stimulus pulse of width 120 ls per phase. The
recorded signal during the stimulation period (which appeared be-
fore 0.5 ms) was the response from the amplifier saturating. the polarity of the stimulus. The neurologic responses are generated
To interpret the recordings it is essential to understand the con- from the first cathodic pulse and have the same sign irrespective of
tribution to the signal from the artefact and that from the neurologic the sign of the last phase of the stimulation. Secondly, the artefact
response. In these experiments the artefact has a number of features decays exponentially and decays well below the neurologic compo-
that allow it to be easily differentiated from a neurologic response. nents 1 ms after stimulus. A measurement on an electrode suffi-
Firstly, the artefact has the opposite sign of the last phase of the bi- ciently far from the stimulating electrode generates a neurologic
phasic stimuli on the stimulating electrode closest to the measure- response that is much larger than the magnitude of the artefact
ment electrode. The recordings in Fig. 2 show the effect of reversing present. Fig. 3 illustrates responses recorded from patient 3.
 Ò
596 J.L. Parker et al. / PAIN 153 (2012) 593–601

Fig. 3. The measurement artefact and evoked responses recorded for patient 4 at a
range of stimulation currents (pulse width of 120 ls). The neural responses, which
are magnified in the inset, are recorded on a single electrode 56 mm from the
stimulation electrodes. The stimulus artefact appears before 0.5 ms. Below thresh-
old, there is no neural response (eg, at 16 mA), whereas above threshold, there is a
neural response of increasing amplitude (at 19 mA and above).

It demonstrates that the artefact has a positive sign and is decaying

exponentially, while the neurologic responses (Fig. 3, inset) produce
a negative peak appearing shortly after 1.5 ms. Thirdly, the neuro-
logic response has a threshold; at low stimulation currents no neu-
rologic response is measured. The response only appears after a
stimulation threshold current is reached (19 mA in Fig. 3). The arte-
fact component does not display such threshold behaviour, but
rather grows in magnitude from very small currents. Fourthly, the
measured neurologic responses propagate with time whereas the
artefact does not. The responses appear on electrodes in sequence
delayed by the propagation of the signal. Fig. 4 shows the propagat-
ing responses in the caudal direction for all subjects from whom a
response was obtained. The data reported here includes the elec-
trode artefact and no subtraction scheme has been implemented
to remove it. The artefact is sufficiently small at the measurement
intervals where the ECAPs are recorded so as not to affect conclu-
sions drawn from the data.
The ECAP morphology consists of a positive P1 peak followed by
a negative N1 and then another positive P2 peak (Fig. 4e). This
shape is characteristic of the potentials generated from depolaris- Fig. 4. ECAP measurements for 4 patients from whom an Ab response recording
ing nerve fibres with a relatively narrow distribution of diameters. was obtained (A–D). ECAPs appear in the plots from approximately 0.8 to 3 ms with
respect to the time axis. The measurement electrodes (labelled from 1 as the most
Axonal depolarization is primarily driven by the second-order dif- rostral to 16 as the most caudal) are caudal to the stimulating electrodes in all cases
ference of the nodal (nodes of Ranvier) field potentials (referred to and the stimulation current was set to the current, which achieved best subjective
as the activating function), where the stimulus elicits this change pain relief as reported by the patients. Note the presence of the stimulation artefact
in membrane potentials [7]. The resulting compound action poten- before 0.5 ms. (E) Measurement of the evoked response in patient 1 on all caudal
electrodes at a relatively high current amplitude, clearly presenting the ECAP
tial is the sum of the single fibre action potentials (SFAPs). The
morphology, with P1, N1 and P2 peaks indicated by the arrows.
SFAP characteristic triphasic shape is the result of the dipoles gen-
erated by ionic flows and change in membrane potentials [20,26].
An ECAP’s morphology is also affected by the slight differences in
SFAP velocities and the filtering effect of the measurement elec- Stimulation-induced paraesthesia was achieved in patient 5, in
trode. This electrode is 4 mm long and samples the electric field whom no evoked potentials were recorded. However, this was
produced by the travelling SFAP population. achieved with only one bipolar stimulus electrode pair, and mov-
ECAPs were successfully recorded in 4 of the 5 patients tested. ing the stimulus to a distal electrode pair where recording was ini-
Fig. 5b shows typical neural responses from subject 1 at a fixed stim- tially attempted failed to evoke paraesthesia at the therapeutic
ulus current level, measured on 7 successive electrodes (Fig. 5a, a–f) current levels. This indicates that the electrodes were placed in a
ascending the spinal cord and 4 successive electrodes (Fig. 5a, l–o) region where a local pathology prevented efficient electrical cou-
descending the spinal cord. The response shifts in time and de- pling (both for stimulation and recording) to the spinal cord.
creases in amplitude as the ECAP propagates along the spinal cord.
Present in the evoked response waveform is a first positive (P1), first 3.2. Conduction velocity
negative (N1) and second negative (P2) series of peaks, as described
earlier. The P1 peak is only observable after the signal has propa- The average conduction velocity of the responding fibres can be
gated a sufficient distance away from the stimulation electrode determined from the difference in timing between the N1 peaks
(Fig. 5a, i), as it is masked by the stimulus artefact. and the distance travelled determined from X-ray (Fig. 5). Fig. 6
 Ò
J.L. Parker et al. / PAIN 153 (2012) 593–601 597

Fig. 6. The negative peak (N1; Fig. 4e) time position of the ECAPs plotted against
the propagation distance in the spinal cord. The distance was estimated from the
fluoroscope image in Fig. 5a and the known inter-electrode spacing. The different
series represent the stimulation currents: 10, 7 and 5 mA. ‘‘Ortho’’ and ‘‘anti’’
indicate the points measured in the orthodromic and antidromic directions,
respectively.

discomfort from the patients. Fig. 7a displays an example of this

late response starting 5 to 10 ms after stimulus, which in this case
is only present in the data taken from electrodes caudal to the
stimulating electrode. Note that the ECAP was still present when
late responses were measured (Fig. 7b). Late responses were also
observed in a further 2 of the 5 patients and were present on elec-
trodes caudal and rostral to the stimulating site.

3.4. Postural changes

Fig. 8 shows measurements of ECAPs and the late response dur-

ing a series of postural changes with a constant stimulation cur-
rent. Initially the patient was in a sitting position (back at 80°
to horizontal). Current amplitude at this point was at the patient
Fig. 5. Fluoroscope image of the electrode placement in the epidural space (A). The comfort limit (32 mA), and the late response was present during
electrodes are labelled a through o, and responses recorded from those electrodes this time. When patients then leaned forwards (100° to horizon-
are similarly labelled in (B). The tip electrode of the rostral array (ie, adjacent to a) tal), they reported a ‘‘weakening’’ of the paraesthesia, which was
was used as the common electrode for measurement. Spinal cord ECAPs are shown
accompanied by a decrease in ECAP amplitude and increase in late
for patient 1 (B). The propagating action potentials measured on each electrode in
the orthodromic and antidromic direction are plotted on separate axes. The response magnitude (seen at 410 s). Finally the patient was moved
stimulus cathode is labelled i, and its 2 most adjacent electrodes formed the anode. to a supine position (0° to horizontal). During this postural
The stimulus was a biphasic current pulse of width 120 ls and amplitude 10 mA. change, which occurred over a period of approximately 40 s (from
420 s), the ECAPs grew substantially in magnitude and the mor-
phology of the late response altered (a relative decrease in ampli-
plots the N1 peak time positions against the distances from the tude but a lengthening of period). In this new supine position the
stimulation electrodes for a number of stimulation current ampli- patient reported ‘‘cramping’’ and found it difficult to raise the
tudes (5, 7, and 10 mA). In this case, the orthodromic velocity was knees when asked to do so (at 530 s). At this point the late re-
48.6 ms 1, and antidromic velocity 46.6 ms 1, as calculated from a sponse was observed to increase in amplitude significantly.
linear fit to the complete data set. The nerve conduction velocity Movement-induced changes in perceptual response as a result
determined in this way for all the patients is detailed in Table 1. of SCS parameters are a common problem [11,22]. It is the result
From Fig. 6, it is clear that the conduction velocity is almost the of changing separation between electrode and spinal cord, which
same for either direction across the range of currents, despite dou- in turn changes the efficiency of electrical stimulation. The proxim-
bling the delivered current. The conduction velocity measured at ity of the electrode to the DC changes the recruitment efficiency,
10% above threshold is no faster than the conduction velocity mea- and indeed this is clearly observed in Fig. 8: the posture change
sured at the therapeutic levels. Given that low currents would be from sitting (at time 420 s) to lying (at 460 s) results in approxi-
expected to produce responses from low threshold fibres (which mately a threefold increase in Ab amplitude.
are larger in diameter and hence have a faster conduction velocity) As mentioned, the presence of the late responses was often
[7], the results indicate that a relatively narrow distribution of fibre accompanied by patient reports of discomfort. The discomfort was
sizes responded to the stimuli during our studies. not reported as painful but was described as a ‘‘tightness’’ and as
‘‘too much,’’ indicating the overall stimulation level was too high.
3.3. Late responses Patient feedback and neural recordings associated with a series of
postural changes (Fig. 8) provide interesting data on the neurophys-
At high stimulation amplitudes—that is, at currents close to the iologic changes taking place during such movements. As described,
patient comfort limit—a response later than the ECAP (‘‘late’’ re- one patient reported ‘‘cramping’’ and found difficulty raising the
sponse) was measured; this was often accompanied by reports of knees from a supine position when being stimulated at relatively
 Ò
598 J.L. Parker et al. / PAIN 153 (2012) 593–601

Fig. 7. Measurement of the evoked response in patient 1 at a stimulus magnitude that was uncomfortable for the patient. The data are shown at 2 different scales for
illustrative purposes. Present is a ‘‘late’’ response (6–17 ms) seen clearly in the antidromic direction, but which is barely present in the orthodromic direction (A). Focussing on
a smaller time frame from (A), ECAPs are present from 0.8 to 3 ms, in both the orthodromic and antidromic directions (B).

Fig. 8. Amplitude of the ‘‘fast’’ Ab response and ‘‘late’’ response plotted during postural changes in patient 1 (A). The stimulation current level was held constant as the patient
moved from a sitting position (position 1, at 350 s), to sitting forwards (position 2, at 410 s), moving to a supine position (position 3, at 460 s) and finally attempting to raise
the legs (position 4, at 535 s). The amplitude of the fast response (voltage of peak P2 minus voltage of peak N1) is plotted (red line), and amplitude of the late response in the
window 5 to 15 ms after stimulus is plotted (blue line). (B) Spinal cord responses recorded at specific postural positions corresponding to data in (A). Each waveform has been
averaged over several seconds and was used to calculate the amplitudes displayed in (A). The 4 postures examined are indicated in (B); current amplitude (I) is provided in
mA.

high current amplitude. Such a report is consistent with the recruit- 3.5. Paraesthesia mapping
ment of muscle afferents in the dorsal roots accompanied by muscle
activity. The late response recorded at 7 ms (Fig. 7a) may represent Fig. 9a shows an example of the amplitude growth of the
the electromyographic (muscle) response and if so this explains evoked response waveform with increasing stimulus current. The
why the signal is not propagating. The cause of this late response amplitude of the ECAP—determined by the difference between
is likely to originate from either activation of the nociceptive reflex the N1 and P2 peaks—varied with the stimulus current applied,
arc, or activation of the muscle afferents of the dorsal roots, both of and was linear over the range of measurement (Fig. 9b). The stim-
which trigger a muscle response. However, further work is required ulation intensity changes the area of paraesthesia as recorded by
to define its source and mechanism. The clinical protocol restricted the subject on a body map of the patient’s pain area (Fig. 9c). Per-
the stimulation range to below the therapeutic level defined in the ception threshold, the point at which the subject first feels the sen-
programming phase of the procedure. Under such restrictions, the sation accompanying the stimulation, corresponded to the first
late responses were measured when the patients changed posture detectable evoked response in 3 of 4 patients, while in patient 3
and without adjustment of the stimulus amplitude. there was no discernible response at this point (Table 1).
 Ò
J.L. Parker et al. / PAIN 153 (2012) 593–601 599

Fig. 9. ECAPs recorded at different current levels (A) and correlated with the percentage of overlap of paraesthesia and pain (C). Responses are displayed for patient 1 with a
pulse width of 80 ls for a current level ranging from 16 mA (1) to 30 mA (8) in 2 mA increments (A). Perception threshold (ECAP l) was attained at a stimulation current level
of 16 mA. Complete coverage of painful areas with paraesthesia was obtained at 28 mA (ECAP 7). ECAP amplitudes (voltage of peak P2 minus voltage of peak N1) for the entire
range of current amplitudes are plotted in (B).

4. Discussion SFAPs) and its characteristics reveal a great deal about the proper-
ties of the responding fibre, including the conduction velocity,
4.1. Spinal cord recording stimulation threshold and response amplitude which can be deter-
mined directly, and the fibre diameter which can be inferred.
Somatosensory-evoked potentials to peripheral stimulation are
routinely measured at the scalp for diagnostic purposes [6,18,21] 4.2. Ab fibre conduction velocity
and epidural monitoring of spinal cord potentials in response to
peripheral nerve stimulation during spinal scoliosis surgery The conduction velocity measurements demonstrate that SCS
[9,13,24] and other procedures for estimating spinal cord integrity recruits Ab fibres in the DCs, thus confirming a long-held theory
are well established. For example, Maruyama et al. report the use [8,12]. Ab fibres are large, thickly myelinated fibres that carry
of epidural spinal cord potential recordings made at the cervical touch and pressure information, with conduction velocities typi-
(C5–7) spinal level in response to stimulation at the cauda equina cally in the range of 30 to 70 ms 1 [25]. In comparison, Ad neurons
[14]. Their recordings show ECAP amplitudes of 5 lV at a stimulus are thinly myelinated fibres responsible for sharp pain and have
intensity 2 times that of threshold, and 13 lV at 32 times thresh- much slower conduction velocities (typically 12 to 30 ms 1) [25],
old. Our measurements differ from these techniques in that we while C fibres have even slower conduction velocities (0.5 to
have measured the evoked response in the human spinal cord 1.2 ms 1) [25]. Results from the 4 patients tested revealed a range
within 14 mm of the stimulation, allowing local neural activity to of conduction velocities from 49 to 65 ms 1, indicating that the Ab
be studied. Further, we have made ECAP recordings in the order fibres of the DC are generating the recorded ECAPs.
of hundreds of microvolts at stimulus intensities less than twice
threshold. This level of proximity and sensitivity in neuronal 4.3. Ab fibre recruitment and paraesthesia
recording has not been presented previously with epidural elec-
trodes but rather has necessitated microelectrode recording di- The relationship between the induced paraesthesia and the Ab
rectly from the spinal cord [29]. amplitude of patient 1 is shown in Fig. 9c. The ECAP amplitude in-
Electrodes placed in the epidural space are separated from the creases with increasing current level (Fig. 9b), as does the area of
spinal cord by a thickness of intervening cerebrospinal fluid (CSF) paraesthesia coverage. The area of coverage and hence the effec-
and spinal meninges of as much as 6 mm. CSF is the preferred con- tiveness of the pain relief is not constant with current amplitude
ductive path, and so there are several possible mechanisms for because of postural variations but is positively correlated with
activation of the DC fibres: direct electrical stimulation of the ECAP amplitude for a given posture. Substantial overlap of the area
superficial layer of the DC, or the stimulation of adjacent spinal of paraesthesia with the area of pain experience is essential for
nerve root fibres that are bathed in the CSF. Regardless of the pre- adequate pain relief during SCS. The magnitude of Ab ECAPs is pro-
cise site of stimulation, the measurements taken are from the portional to the level of Ab recruitment. Ab recruitment level de-
propagating ECAP in the DC. This ECAP represents the summation fines the area and magnitude of the paraesthesia coverage as it is
of the responses from a large number of single fibres (sum of the this population of neurons which generates both the paraesthesias
 Ò
600 J.L. Parker et al. / PAIN 153 (2012) 593–601

and the pain relief. Therefore, having determined the relationship mechanism of SCS is via Ab fibre recruitment, which confirms the
between paraesthesia coverage and magnitude of evoked potential postulated theory. The presence of a longer, nonpropagating re-
(Fig. 9), it may be possible to use only the evoked response ampli- sponse at higher potentials is accompanied by uncomfortable par-
tude as a quantitative neurophysiologic measure associated with aesthesia. The in vivo measurement and characterisation of the
pain relief experienced by SCS implant recipients. ECAP yields new insights into the mechanisms of SCS and provides
a unique tool to characterise the fibres responding to the stimula-
4.4. Estimation of fibre number and size tion. The Ab ECAP is a measure of the level of neural recruitment
and provides a potential feedback parameter to optimise the oper-
The amplitude of the Ab responses provides a direct indication ation of neuromodulation function on a continuous basis.
of the number of fibres responding to the stimulus, which is di- Data from 5 patients are insufficient to draw conclusions about
rectly related to the area of coverage of the induced paraesthesia. the variability in ECAP responses from individual to individual or
Coverage of the complete area of the leg requires at least 5 dermat- more interestingly from individuals who experience neuropathic
omes, which according to Holsheimer’s model [8] would involve pain and normal individuals. For instance, differences in conduc-
approximately 25 fibres in the DC surface. The contribution to tion velocity and sensitivity may be expected if results obtained
the ECAP from a single fibre is likely to be small (less than 1 lV) in animal pain models translate to humans [10]. The measurement
[27]. The magnitude of the Ab response we measured for SCS cov- technique may also provide a useful guide to programming of
erage varied with each patient (Table 1) and for a number of pa- stimulation parameters of neuromodulation systems. It has the po-
tients the magnitude of their ECAPs was considerably larger than tential to determine the most sensitive combinations of electrodes
predicted, indicating many more fibres are recruited than previ- to use for stimulation and the recruitment curves could give a
ously predicted by modelling. practical indication of the available range of currents for control
Nerve fibre conduction velocity is linearly related to the fibre of the pain relieving paraesthesia generation. Finally, it is intrigu-
diameter [28] (approximately 6 ms 1 per lm). From our measure- ing to contemplate the use of the ECAP recordings in patient selec-
ments for the first subject the fibre diameter estimated for tion for SCS. A significant number of patients trialling SCS fail to
48.6 ms 1 velocity is 8.1 lm (Table 1), which is smaller than pre- achieve satisfactory pain relief, and there may be significant differ-
dicted by simulation models [8]. The measurement of the velocity ences in the responses measured in this group from the responding
is made over a relatively long distance (42 mm) and is linear over group. This would allow patient selection at the time of the trial
the range of measurements (Fig. 6). The Ab fibres branch on entry procedure, allowing a move to immediate full implantation and
into the DCs, and as they do so, they become smaller and slower significantly reducing the overall procedure cost. Data and out-
conducting. It is possible that recruitment occurs before branching comes from a large population of individuals are required to eluci-
and that the measurements are indicative of the fibres having date any or all of the systematic differences postulated above.
reached their final diameter as they propagate up and down the
cord. Approximately 8.7% of the fibres in DC have a diameter great-
Conflict of interest statement
er than 8.1 lm [5], which from our data would yield more than
1000 fibres of suitable diameter in the superficial DC. Further stud-
The authors report no conflict of interest.
ies are required to resolve these uncertainties.

4.5. Overstimulation and side effects Acknowledgements

Patients treated with SCS systems often report shocks and We thank Peter Ayre and Linda Critchley for assistance in prep-
uncomfortable paraesthesia induced by movement [11,22]. These aration of the human study protocol and execution of the experi-
side effects were observed during the clinical recording sessions ments; and James Laird and Mark Bickerstaff for assistance in
as a ‘‘late’’ response (Fig. 7). The precise origin of the late response collection and analysis of data.
cannot be determined from the data alone. The responses dis-
played in Fig. 7 do not propagate; rather, they occur at the same Appendix A. Supplementary data
time after stimulation for all the electrodes measured. If they were
propagating along the spinal cord then a small shift in the onset Supplementary data associated with this article can be found, in
time would be expected that is proportional to the velocity of tra- the online version, at doi:10.1016/j.pain.2011.11.023.
vel and the distance between the electrodes. This is not the case,
and so the signal measured is likely to arise outside the spinal cord. References
The therapeutic range of stimulation current may lie between
[1] Alo KM, Holsheimer J. New trends in neuromodulation for the management of
the threshold current at which Ab ECAPs are observed and the cur-
neuropathic pain. Neurosurgery 2002;50:690.
rent which induces the late response. The relationship between [2] Blum RA, Ross JD, Brown EA, DeWeerth SP. An integrated system for
late response current thresholds and Ab recruitment thresholds simultaneous, multichannel neuronal stimulation and recording. IEEE Trans
varied between patients. The late response threshold indicates Biomed Circuits Syst 2007;54:2608–18.
[3] Brown EA, Ross JD, Blum RA, Yoonkey N, Wheeler BC, DeWeerth SP. Stimulus-
the sensitivity of muscle afferents or nontargeted fibres, and this artifact elimination in a multi-electrode system. IEEE Trans Biomed Circuits
may be influenced by electrode placement, anatomy or electrode Syst 2008;2:10–21.
design as well as by other factors. Understanding the influences [4] El-Khoury C, Hawwa N, Baliki M, Atweh SF, Jabbur SJ, Saadé NE. Attenuation of
neuropathic pain by segmental and supraspinal activation of the dorsal
of these factors via measurement of the evoked responses will al- column system in awake rats. Neuroscience 2002;112:541–53.
low improvement in stimulation systems and surgical methods. [5] Feirabend HKP, Choufoer H, Ploeger S, Holsheimer J, van Gool JD. Morphometry
of human superficial dorsal and dorsolateral column fibres: significance to
spinal cord stimulation. Brain 2002;125:1137–49.
4.6. Conclusions [6] Grundy BL, Villani RM. Evoked potentials: intraoperative and ICU
monitoring. New York: Springer-Verlag; 1988.
In summary, we have presented novel measurements of ECAPs [7] Holsheimer J. Principles of neurostimulation. In: Simpson BA, editor. Electrical
stimulation and the relief of pain. Pain research and clinical management, vol.
in the human spinal cord in patients undergoing SCS for treatment
15. Amsterdam: Elsevier Science; 2003. p. 17–36.
of chronic pain. Larger populations of fibres are responding to the [8] Holsheimer J. Which neuronal elements are activated directly by spinal cord
stimulus than previously suggested by modelling. A key step in the stimulation? Neuromodulation 2002;5:25–31.
 Ò
J.L. Parker et al. / PAIN 153 (2012) 593–601 601

[9] Jones SJ, Edgar MA, Ransford AO. Sensory nerve conduction in the human [20] Plonsey R. Action potential sources and their volume conductor fields. Proc
spinal cord: epidural recordings made during scoliosis surgery. J Neurol IEEE 1997;65:601–11.
Neurosurg Psychiatry 1982;45:446. [21] Saidman LJ, Smith NT. Monitoring in anesthesia. London: Butterworth–
[10] Kohama I, Ishikawa K, Kocsis JD. Synaptic reorganization in the substantia Heinemann; 1984.
gelatinosa after peripheral nerve neuroma formation: aberrant innervation of [22] Schade CM, Schultz D, Tamayo N, Iyer S, Panken E. Automatic adaptation of
lamina II neurons by Abeta afferents. J Neurosci 2000;20:1538–49. neurostimulation therapy in response to changes in patient position: results of
[11] Kuechmann C, Valine T, Wolfe D. Could automatic position-adaptive the Posture Responsive Spinal Cord Stimulation (PRS) research study. Pain
stimulation be useful in spinal cord stimulation? Eur J Pain 2009;13:S243. Physician 2011;14:407–17.
[12] Linderoth B, Meyerson BA. Dorsal column stimulation: modulation of [23] Shealy CN, Mortimer JT, Reswick JB. Electrical inhibition of pain by stimulation
somatosensory and autonomic function. Semin Neurosci 1995;7:263–77. of the dorsal columns. Anesth Analg 1967;46:489–91.
[13] Machida M, Weinstein SL, Yamada T, Kimura J. Spinal cord monitoring: [24] Shimoji K, Willis WD. Evoked spinal cord potentials: an illustrated guide to
electrophysiological measures of sensory and motor function during spinal physiology, pharmacology, and recording techniques. New York: Springer;
surgery. Spine 1985;10:407. 2006.
[14] Maruyama Y, Shimoji K, Shimizu H, Kuribayashi H, Fujioka H. Human spinal [25] Strichartz GR, Pastijn E, Sugimoto K. Neural physiology and local anesthetic
cord potentials evoked by different sources of stimulation and conduction action. In: Cousins MJ, Carr DB, Horlocker TT, Bridenbaugh PO, editors. Cousins
velocities along the cord. J Neurophysiol 1982;48:1098. and Bridenbaugh’s neural blockade in clinical anesthesia and pain medicine.
[15] Mayer S, Geddes LA, Bourland JD, Ogborn L. Electrode recovery potential. Ann 4th ed. Philadelphia: Lipincott, Williams, and Wilkins; 2008. p. 26–47.
Biomed Eng 1992;20:385–94. [26] Struijk JJ. The extracellular potential of a myelinated nerve fiber in an
[16] Melzack R, Wall PD. Pain mechanisms: a new theory. Science 1965;150:971–9. unbounded medium and in nerve cuff models. Biophys J 1997;72:2457–69.
[17] Meyerson B, Linderoth B. Mode of action of spinal cord stimulation in [27] Struijk LNS, Akay M, Struijk JJ. The single nerve fiber action potential and the
neuropathic pain. J Pain Symptom Manage 2006;31:S6–S12. filter bank—a modeling approach. IEEE Trans Biomed Eng 2008;55:372–5.
[18] Nuwer MR, Dawson EG, Carlson LG, Kanim LEA, Sherman JE. Somatosensory [28] Waxman SG. Determinants of conduction velocity in myelinated nerve fibers.
evoked potential spinal cord monitoring reduces neurologic deficits after Muscle Nerve 1980;3:141–50.
scoliosis surgery: results of a large multicenter survey. Electroencephalogr Clin [29] Woolf CJ. Evidence for a central component of post-injury pain
Neurophysiol 1995;96:6–11. hypersensitivity. Nature 1983;306:686–8.
[19] O’Keeffe DT, Lyons GM, Donnelly AE, Byrne CA. Stimulus artifact removal using
a software-based two-stage peak detection algorithm. J Neurosci Methods
2001;109:137–45.