Applied Psychophysiology and Biofeedback, Vol. 30, No.

1, March 2005 (

C 2005)

DOI: 10.1007/s10484-005-2177-8

Biofeedback in the Treatment of Phantom Limb Pain:

A Time-Series Analysis
R. Norman Harden,1,2 Timothy T. Houle,1 Samara Green,1 Thomas A. Remble,1
Stephan R. Weinland,1 Sean Colio,1 Jeffrey Lauzon,1 and Todd Kuiken1

Phantom limb pain (PLP) is a noxious, painful sensation that is perceived to occur in
an amputated limb. It has been reported to occur in up to 85% of amputees. This pilot
study examined the effectiveness of biofeedback in the treatment of nine individuals with
PLP who received up to seven thermal/autogenic biofeedback sessions over the course
of 4–6 weeks. Pain was assessed daily using the visual analog scale (VAS), the sum of
the sensory descriptors, and the sum of the affective descriptors of the McGill short form.
Interrupted time-series analytical models were created for each of the participants, allowing
biofeedback sessions to be modeled as discrete interventions. Analyses of the VAS revealed
that a 20% pain reduction was seen in five of the nine patients in the weeks after session
4, and that at least a 30% pain reduction (range: 25–66%) was seen in six of the seven
patients in the weeks following session 6. Sensory descriptors of pain decreased more than
the affective pain descriptors. These preliminary results provide some support for the use
of biofeedback in the treatment of PLP and indicate the need for further, definitive study.
KEY WORDS: phantom limb pain; post amputee pain; thermal biofeedback; time-series analysis.

INTRODUCTION

Phantom limb pain (PLP) is defined as a painful sensation occurring in a missing

body part (Davis, 1993a, 1993b). PLP often begins within the first week after amputation
and is primarily described as “knifelike, burning, squeezing, and throbbing” (Davis, 1993a,
1993b). Mechanisms for PLP are not well understood, and standard treatments for pain
(i.e., drugs) have proven fairly ineffective (Iacono, Sandyk, Bamford, Awerbuch, & Malone,
1987; Sherman, 1997).
The reported incidence, prevalence and intensity of PLP over the past 30 years is
significant. In a 1983 study, as many as 90% of amputees reported some degree of sensation
associated with a phantom limb (Jensen, Krebs, Nielsen, & Rasmussen, 1983). In the past
10 years, more sharply focused studies have shown a consistent incidence of PLP in
1 Center for Pain Studies, Rehabilitation Institute of Chicago/Northwestern University Medical School, Chicago,
Illinois.
2 Address all correspondence to R. Norman Harden, Center for Pain Studies, Rehabilitation Institute of Chicago,
446 Ontario Street, Suite 1011, Chicago, Illinois 60611; e-mail: nharden@ric.org.

83
1090-0586/05/0300-0083/0

C 2005 Springer Science+Business Media, Inc.
84 Harden, Houle, Green, Remble, Weinland, Colio, Lauzon, and Kuiken

50–59% of amputee patients (Davison & Jones, 1995) (Montoya et al., 1997) (Kooijman,
Dijkstra, Geertzen, Elzinga, & van der Schans, 2000). In the largest study of PLP intensity
to date, Kooijman, Dijkstra, Geertzen, Elzinga, and Schans (2000) evaluated the presence
of pain in 124 amputees in the Netherlands. Sixty-four percent of the Netherlands sample
reported experiencing at least a “moderate” amount of suffering from their pain, with 40%
of the sample reporting that they “always” experienced PLP and another 20% reporting
PLP a “few times a day.” The incidence, frequency, and degree of suffering caused by PLP
underlines the importance of developing effective treatments for this condition, as well as
other types of post-amputation pain.

Standard Medical Treatment in Post-Amputation Pain

Standard medical treatments for PLP have been relatively ineffective, and the need
for better treatments has been consistently noted (Sherman, 1997). At the time of this
report, no pharmacologic agent has been shown to be consistently effective in the treatment
of PLP. In 1987, Iacono, Linford, Pharm, and Sandyk reviewed the literature on opioids,
b-blockers, anticonvulsants, tricyclics, neuroleptic agents, and neurosurgical interventions.
They concluded: “Unfortunately, partial relief or failure is often the result of even the most
diligent medical management” (Iacono, Linford, & Sandyk, 1987). In a recent review of all
English trials published between 1966 and 1999, Halbert, Crotty, and Cameron reviewed
12 trials, which included the use of epidurals, nerve blocks, calcitonin, and transcutaneous
electrical nerve stimulation. The authors found only three trials that reported modest and
temporary decrease of amputee pain. Two positive trials examined the treatment of acute
phantom pain with preoperative (‘pre-emptive’), intraoperative, and early (at <2 weeks)
postoperative interventions utilizing epidural anesthesia (Nikolajsen, Ilkjaer, Christensen,
Kroner, & Jensen, 1997) and a regional nerve block (Pinzur, Garla, Pluth, & Vrbos, 1996).
Another ‘positive’ study examined the treatment of chronic phantom pain with a late (at
>2 weeks) postoperative intervention utilizing galvanic stimulation using a metal threaded
sock (Farabloc Development Corp., Coquitlam, BC, Canada)(Conine, Hershler, Alexander,
& Crisp, 1993). Halbert, Crotty and Cameron emphasized the importance of ongoing
research to develop more effective treatments for PLP (Halbert, Crotty, & Cameron, 2002).
No differences between gabapentin and placebo were observed for change in VAS ratings,
mood, sleep interferences, or activities of daily living (Bone, Critchley, & Buggy, 2002).

Psychophysiological Factors in Post-Amputation Pain

The current lack of understanding of psychophysiological mechanisms may be the pri-

mary reason why ‘standard’ treatments (i.e., medications) have been relatively ineffective,
but new imaging and quantitative psychophysical techniques are beginning to elucidate the
mechanisms of PLP. There is a growing body of evidence that suggests PLP is related to sym-
pathetic nervous system activity (Katz, 1997b). Possible etiological mechanisms in PLP are
related to sympathetic over-activity with primary vasoconstriction, and ischemia of patho-
logically altered structures in the residual limb, (such as neuroma), or direct noradrenergic
stimulation of neuroma (Devor, 1997). It is possible that other pathologic tissue in the
residual limb (such as pathologic receptors, or up regulation of ‘normal’ receptors) may
Biofeedback in the Treatment of Phantom Limb Pain 85

become more sensitive to local or circulating catecholamine levels (Harden et al., 2004).
Several studies suggest a complex relationship between residual limb temperature and po-
tential PLP involvement (Sherman, Bruno, & Sherman, 1987). Katz’s study of 28 amputees
revealed evidence of the relationship between sympathetic arousal and pain intensity (Katz,
1997b). Further, in his review of the literature, Katz found that medications and surgeries
designed to reduce or stop sympathetic activity produced significant, although temporary,
reduction in pain (Katz, 1997a). Since these studies indicate that vasoconstriction and a
relative decrease in temperature in the residual limb are physiologic markers of autonomic
disturbance, these signs may also serve as physiologic targets for treatment.

Biofeedback in the Treatment of PLP

Biofeedback has become a common therapy used to successfully treat many chronic
pain conditions. Such biofeedback modalities include EMG biofeedback; skin tempera-
ture feedback; and muscle relaxation procedures (Arena & Schwartz, 2003; Schwartz &
Schwartz, 2003), which help reduce muscle tension and increase blood flow (which has
been used for PLP) (Belleggia & Birbaumer, 2001); The use of biofeedback to treat PLP is
not novel; several prior studies have yielded positive results (Belleggia & Birbaumer, 2001;
Dougherty, 1980; Sherman, Gall, & Gormly, 1979).
Sherman, Gall, and Gormly (1979) have reported that different descriptions (symptom
sets) of PLP vary in response to different forms of biofeedback therapy. Recently, Sherman
and collaborators have been attempting to match behavioral and medical treatments of PLP
with underlying physiological correlates (e.g., amputees with burning pain receive thermal
biofeedback; patients with cramping pain receive EMG biofeedback). Unfortunately, all
but one of these studies are case reports (the exception being Sherman et al., 1979) with
widely varying methods and materials. The effectiveness of specific biofeedback procedures
matched by pain description (e.g., burning or cramping) in this population largely remains
unproven.
The present study examined the effectiveness of thermal biofeedback training in treat-
ing PLP. Thermal biofeedback may be particularly well suited to reduce some of the com-
plex physiological factors that affect PLP by mediating net regional sympathetic arousal.
Thermal biofeedback allows an individual to monitor peripheral temperature and thereby
indirectly monitor and modify sympathetic activity (Freedman, 1991). If sympathetic ner-
vous system activity is indeed causally related to PLP, gaining control over such activity
should result in decreased pain. Increased control over sympathetic activity is hypothesized
to lead to decreased pain in the weeks following treatment.

METHOD

Participants

To be eligible for the study, patients had to report PLP in excess of phantom sensations
(mean VAS pain at baseline = 38.2, SD = 21.4). Nine outpatients (five males, four females)
being treated for PLP participated in this pilot study. Patients’ average age was 57.6 years
(SD = 11.0). Five patients had lower extremity amputations and four had upper extremity
86 Harden, Houle, Green, Remble, Weinland, Colio, Lauzon, and Kuiken

amputations. Four subjects had bilateral amputations; two below the knee, and two below
the elbow. Six of the patients used prosthetics. All patients provided informed consent
for treatment with biofeedback. All patients were instructed to maintain their current
medication regimens without alteration throughout the course of the study. Thus, the study
represents biofeedback as an add-on intervention to our pharmacologic standard of care.

Measures and Materials

Subjects completed a twice-daily pain diary. The diary consisted of the McGill Pain
Questionnaire Short Form (MPQ-SF) (Melzack, 1987), the Beck Depression Inventory
(BDI) (Beck, Steer, Beck, & Newman, 1993; Beck, Steer, & Brown, 1993), and the trait
portion of the State-Trait Anxiety Inventory (STAI) (Spielberger, Gorsuch, & Lushene,
1970).

McGill Pain Questionnaire: Short Form

The complete MPQ-SF was administered at each clinic visit to assess weekly changes
in overall pain intensity.

Beck Depression Inventory (BDI)

The BDI was used to assess depressive symptoms at baseline and at study termination.
This measure was used in secondary analyses to determine whether topiramate or its
pain relief improved depressive symptoms, and whether the effectiveness of topiramate is
affected by the extent of depressive symptoms.

Trait Form of the State-Trait Anxiety Inventory (STAI)

The STAI, which assesses anxiety related symptoms, was used at baseline and study
termination for secondary analyses similar to those described for the BDI.
Temperature was monitored on the Thought Technology Biograph using ProComp 2.0
software.

Procedure

During the initial visit, subjects were informed about the study and provided consent.
Patients completed a general health questionnaire and questions or issues concerning treat-
ment were discussed. Participants were educated about the rationale of the study and no
deception was used concerning the nature of the treatment. Finally, the subjects were given
the pain diaries with instructions to complete the diaries at the same time each morning
and each evening. There was an average time of 8.7 days between the initial visit and first
biofeedback session (range 4–20 days).
Biofeedback consisted of three treatment phases: treatment-training (sessions 1–3),
warming (sessions 4–6), and follow-up (session 7). A script provided standard introduction
Biofeedback in the Treatment of Phantom Limb Pain 87

to biofeedback, rationale, and instruction on relaxation training and thermal biofeedback. It

was read to each patient during visit one of session one only. During each of the first three
treatment-training sessions, biofeedback sensors were attached and a 10-minute baseline
was recorded. Next, a 20-minute treatment period began with the biofeedback therapist
reading 20 autogenic phrases to the patient in a slow and rhythmic manner. Subjects were
instructed to repeat the phrases they liked best while imagining sensations of warmth and
heaviness in their unaffected extremities. During this stage, the patients received feedback
from a healthy extremity. Patients were then instructed to practice the technique at home at
least once a day.
During treatment-training sessions 2 and 3, the patient’s progress was reviewed and
any questions regarding the warming techniques were discussed. Following the ten-minute
baseline period, the standard 20 autogenic phrases were read to the patient at the beginning
of the 20-minute treatment period as in session one, but beginning in session 2 the patient
was instructed to gradually move focus from the healthy limb to the residual limb, and was
given the feedback received from both a healthy digit and the residual limb.
Warming sessions (4–6) included a 10-minute baseline thermal recording focused on
subject feedback from the residual limb only. In these sessions, the autogenic phrases were
not read, allowing the patient to self-direct warming via their newly learned individual
repertoire of phrases. The patient was left alone to allow self-initiated warming techniques
for 20 minutes. Following this phase, the patient was instructed to finish the technique, and
any problems encountered were discussed.
Session 7 was a follow-up session. Follow-up biofeedback was similar to warming
sessions 4, 5 and 6, except that equivalent levels of feedback were not provided. During the
final session, problems with techniques were discussed along with a therapeutic directive to
foster patient’s perceived ability to successfully employ the autogenic techniques learned
throughout treatment (Bandura, 1977). This is equivalent to a total ‘dose’ (actual time on
the biofeedback equipment, receiving feedback) of 140 minutes.

Biofeedback Technique

Thermistors were attached to consistent multiple sites throughout the treatment. A

thermistor was fixed to the end of the residual limb in an unscarred area. Under circum-
stances of a bilateral amputation, the primary thermistor was attached to the same specific
residual limb in order to maintain consistency throughout treatment. A second thermistor
was attached to the corresponding site on the contralateral limb. A final thermistor was
attached to an extremity of a healthy limb.

Data Analysis

Descriptors of the MPQ-SF were separately summed to create a sensory and affective
score for each diary entry. Morning and afternoon diary entries were averaged to create
a ratings score for affective, sensory, and VAS pain. This data reduction technique was
conducted to eliminate diurnal variation effects from the analysis. Missing values were
replaced using adjacent point interpolation. Reported compliance with the diaries was
excellent, with less than 1% of the data missing.
88 Harden, Houle, Green, Remble, Weinland, Colio, Lauzon, and Kuiken

Analyses were carried out using the SPSS 11.5 for Windows package (SPSS Inc.,
Chicago, IL). To identify the comparative effects of biofeedback treatment over time, the
VAS, sensory, and affective scores of the participant’s daily diary were modeled using a
time-series regression analysis. Psychometric data collected at each session were analyzed
using a repeated measures analysis of variance.
For each subject, unique interrupted time-series regression models were used to model
pain reporting over time. A separate model was created for affective, sensory, and VAS
pain scales. Treatment was divided into discrete time periods beginning on the day of
the biofeedback session and including all of the days prior to the next session. Thus,
each patient’s treatment was divided into sections (treatment periods) corresponding to
the biofeedback sessions. Treatment periods were modeled as discrete interventions and
were dummy coded to represent abrupt, permanent impacts on pain reporting. The dummy
vectors for each treatment period were coded such that they contained a 1 for observations
corresponding to that treatment period and a 0 for all else. Each of the regression weights
then corresponded to changes in pain from baseline associated with that treatment period.
Use of the dummy vectors allowed for statistical evaluation of intra-individual changes in
pain biofeedback session.
When observations are collected repeatedly over time, subsequent observations tend
to be more related than observations more temporally distant. This process has been called
autocorrelation, and is the correlation of the series with itself. Autocorrelation in a series
can lead to the violation of statistical assumptions. To ensure the validity of the inferential
statistics, a first-order autoregressive error process was included to correct for autocorre-
lation in the residuals (Box, Jenkins, & Reinsel, 1994). To ensure that the residuals were
‘white noise’ (were not correlated with time; a necessary assumption in regression analysis),
the corrected models, also including the session parameters, were then evaluated using the
modified Box-Pierce Q∗ . In all models, a first-order autoregressive parameter successfully
reduced the residuals of the model to white noise, and thus the significance values of the
models can be interpreted conventionally.
Finally, mixed models were used to aggregate the individual parameters of the time-
series regression models for global comparisons by treatment session. The use of these
models allowed for modeling the group’s changes throughout treatment. In addition, each
participant’s individual error structure for specification was uniquely modeled (i.e. the
temporal relationship between treatment periods).

RESULTS

Table I presents the baseline pain values of each subject along with the percentage
change in pain corresponding with each treatment period on the individual level. Eight
out of nine patients in the study showed statistically significant improvement after some
biofeedback sessions, and overall improvement increased as treatments increased. Four
out of nine patients displayed statistically significant changes in pain at some point dur-
ing treatment (p < .05). After session 4, five out of nine patients reported experiencing
at least a 20% reduction in pain, and by session 6, the 7 patients still enrolled reported
an average of 39.3% reduction in pain (range 25–66% reduction). One patient (#2) re-
ported statistically significant increases in pain in several treatment periods (p < .05; see
Table I).
Biofeedback in the Treatment of Phantom Limb Pain 89

Table I. VAS Pain Score at Baseline with Percentage Change for Each Subject by Treatment Period
Session
Subject BL 1 2 3 4 5 6 7

1 46 −13 −1 2 −10 −27

2 62 4 −1 6† 5 11†
3B 42 28 −17 −40∗ −26 −81∗∗ −58∗∗ −71∗∗
4 78 −1 −23∗∗∗ 2 −13∗ −40∗∗∗ −25∗∗∗
5 19 −33 −5 5 −21 −1 −66
6 24 8 −29 −15 −46 −16 −39
7 25 −41 −11 −40∗ −45∗∗ −33 −34 −15
8B 36 1 5 27 14 0 −33∗∗
9 12 25 3 43 −38 −54 −52

Note. B patient had substantial “hot-burning” pain at baseline; BL = baseline.

∗ p < .05 decrease in pain.
∗∗ p < .01 decrease in pain.
∗∗∗ p < .001 decrease in pain.
† p < .05 increase in pain.

Figure 1 displays the aggregated mean and median reduction in VAS pain as a function
of treatment period. The median pain response was quite similar to the mean response,
indicating that the observed treatment effects were not largely influenced by extreme outliers
(see Table I for individual responses). Beginning after the fourth treatment session (p = .06)
and continuing through the remainder of treatment, statistically significant decreases in pain
as compared with baseline were observed (p s < .01). Further, the effects due to treatment
increased with subsequent sessions as after sessions 5–7, patients reported less pain than
after sessions 1–3 (p < .01). (See Fig. 1).
The subject sample reported low levels of anxiety and depressive symptoms even
at baseline (STAI: M = 34.4, SD = 11.6; BDI: M = 8.75, SD = 7.1). Although slight
decreases were seen in both the STAI and BDI over treatment, observed changes were not
statistically significant (Time: p s > .25).
Figure 2 displays the aggregated mean and median reduction in MPQ-Sensory pain as
a function of treatment period. Beginning after the third treatment session (p = .051) and
continuing through the remainder of treatment, statistically significant decreases in sensory
pain as compared with baseline were observed (p s < .05). The median pain response was
quite similar to the mean response, indicating that the observed treatment effects were not
largely influenced by extreme outliers.

Fig. 1. Average percentage change in VAS by treatment week, ∗ p ≤ .01.

90 Harden, Houle, Green, Remble, Weinland, Colio, Lauzon, and Kuiken

Fig. 2. Average percentage change in MPQ-sensory by treatment

week, ∗ p ≤ .05.

Figure 3 displays the aggregated mean and median reduction in MPQ-Affective pain
as a function of treatment period. Affective pain did not decrease throughout treatment
as no significant differences were observed between any treatment periods (p > .05).
The mean pain response differed considerably from the median response, reflecting both
the variation in response to treatment and the low amount of observed variance in this
measure.
Pain, as measured by the VAS, decreased throughout treatment. The nine patient mean
VAS score at baseline was M = 38.2 (SD = 21.4). Average pain decreased during most
treatment periods and was M = 37.8 (SD = 23.4) for period 1, M = 34.3 (SD = 18.6)
for period 2, M = 37.3 (SD = 23.5) for period 3, M = 32.9 (SD = 22.9) for period 4,
M = 28.2 (SD = 20.3) for period 5, M = 20.5 (SD = 18.1) for period 6, and M = 16.8
(SD = 6.3) for period 7.
Previous research has suggested a superior response to biofeedback in those amputees
with “burning” pain (Sherman, Gall, & Gormly, 1979), but thermal biofeedback did not
appear to decrease pain in amputees with “burning” pain in this study. The baseline diaries
of subject #3 and subject #8 identified them as subjects with a predominantly “hot-burning”
presentation, so those subjects are noted with a “B” in Table I. These two subjects reported
comparatively more burning pain than the other subjects in the analysis and also reported
more burning quality to their pain in conjunction with the other sensory descriptors of the
MPQ (although all subjects reported experiencing more than one type of sensory quality
to their pain). As can be seen in Table I, subject #3 was one of the better responders
to the biofeedback, but interestingly, subject #8, who reported the highest amount of

Fig. 3. Average percentage change in MPQ-affective by treatment

week.
Biofeedback in the Treatment of Phantom Limb Pain 91

hot-burning sensations, reported less benefit than the average participant in the study. The
small subsample precludes meaningful group comparisons, but this study did not show a
superior response from these two participants.

DISCUSSION

This pilot study supports the hypothesis that thermal biofeedback training may have
an important role to play in the management of PLP. After receiving six sessions of training
(and a ‘dose’ of 140 minutes of biofeedback), patients reported an average of 39% decrease
in daily pain, with three patients reporting a greater than 50% reduction. Changes in
the sensory descriptors of pain (i.e. “sharp, burning” etc.) were mirrored by changes in
the VAS. The treatment appeared to be broadly effective in this small sample, with only one
patient reporting a significant increase in pain. Furthermore, it appears that all the subjects
that responded had not plateaued, and it is likely that more treatment (a higher ‘dose’) may
well lead to greater improvement.
This pilot study was not designed or powered to investigate PLP mechanisms, but there
are multiple possible explanations for these results. For instance, the results of this study
may indicate that patients were able to (centrally) alter the perceived painful sensations
in their residual or phantom limbs, or it may be indicative of physiologic change in the
periphery, such as improved blood flow and/or decreased sympathetic output. It is very
likely that the impact is situated at multiple levels of the neuro-axis; afferent (sensory
processing), efferent (autonomic) and central. Another logical explanation for the observed
decrease in pain, and one that has received a great deal of attention in the headache
literature, is through a possible mechanism for induced cognitive changes. Biofeedback
facilitates cognitive changes in patients by demonstrating that physiological processes are
under their control (i.e. increased internal control) and that they are in fact able to influence
these physiological processes (i.e. enhancing self-efficacy) (Rokicki et al., 1997). Thus,
biofeedback may simply be a vehicle to shift a patient’s focus from external to internal
methods of pain coping, making perceived pain more tolerable.
Future studies should assess changes in perceived control and self efficacy in the
treatment of PLP, with similar statistical methods used in the present study to identify
if cognitive changes predict decreases in pain. Future studies could monitor, record and
correlate the changing physiological states induced by biofeedback (i.e. ability to control
hand temperature) with time-series regression methodology. This would identify whether
changes in physiology predict or predate decreases in pain and sympathetic arousal. In
this pilot the therapist reported (anecdotally) that ‘all’ subjects achieved a degree of
control over the physiologic variable of interest (temperature of the residual limb), this
is a critical variable that must be formally monitored and recorded in future, definitive
studies. We can and should utilize more sophisticated psychophysical and psychometric
methodologies.
The results from the McGill short form pain questionnaire indicated that affective
descriptors of pain did not decrease as much as sensory descriptors. This is likely a ‘floor
effect’, since the nine patients in this study all initially reported low levels of affective pain
(M = 1.8 out of a possible score of 12) which did not fluctuate greatly during the course of
treatment. Changes that occurred during this study in the affective descriptors of pain were
neither trending nor significant, and did not contain a great deal of daily variance. This
92 Harden, Houle, Green, Remble, Weinland, Colio, Lauzon, and Kuiken

information may reveal a particular bias about this small sample: the patients participating
in treatment did not appear to have a large emotional component to their pain experience
(i.e. describing the pain as “punishing-cruel”), and they reported low levels of depressive
and anxiety symptoms, even at baseline. Thus, the relative lack of changes in affective pain
reporting must be interpreted with an eye toward the small sample size and makeup of this
particular group.
The sample size of this pilot limits the generalizability of the results. The use of time-
series analysis and hierarchical linear modeling provided robust statistical power which
allowed statistical differences to be detected with only nine participants, but the stability of
the results cannot be ascertained from a sample this small even though the observed effect
sizes were quite large. Further study with larger samples is clearly warranted to elucidate
which factors affect the treatment outcome. It will also be important to asses the durability
of any positive change.
The current study adds to the body of literature suggesting that thermal biofeedback
may be a viable treatment option for patients with PLP. The change processes induced
with biofeedback need further study, including direct comparisons with other cognitive-
behavioral approaches, like relaxation training. A definitive study designed to answer some
of these outstanding questions should include more objective measures of physiologic
change, as correlated with the psychometrics.

REFERENCES

Arena, J. G., & Schwartz, M. S. (2003). Psychophysiological assessment and biofeedback baselines: A primer.
In M. S. Schwartz & F. Andrasik (Eds.), Biofeedback: A practitioner’s guide (3rd ed., pp. 128–158). New
York: Guilford.
Bandura, A. (1977). Self-efficacy: Toward a unifying theory of behavioral change. Psychological Review, 84(2),
191–215.
Beck, A. T., Steer, R. A., Beck, J. S., & Newman, C. F. (1993). Hopelessness, depression, suicidal ideation, and
clinical diagnosis of depression. Suicide Life Threat Behavior, 23(2), 139–145.
Beck, A. T., Steer, R. A., & Brown, G. (1993). Dysfunctional attitudes and suicidal ideation in psychiatric
outpatients. Suicide Life Threat Behavior, 23(1), 11–20.
Belleggia, G., & Birbaumer, N. (2001). Treatment of phantom limb pain with combined EMG and thermal
biofeedback: A case report. Applied Psychophysiology and Biofeedback, 26(2), 141–146.
Bone, M., Critchley, P., & Buggy, D. J. (2002). Gabapentin in postamputation phantom limb pain: A randomized,
double-blind, placebo-controlled, cross-over study. Regional Anesthesia and Pain Medicine, 27(5), 481–
486.
Box, G., Jenkins, G., & Reinsel, G. (1994). Time series analysis, forecasting and control (3rd ed.). San Francisco:
Holden-Day.
Conine, T.,Hershler, C., Alexander, S., & Crisp, R. (1993). The efficacy of Farabloc in the treatment of phantom
limb pain. Canadian Journal of Rehabilitation, 6, 155–161.
Davis, R. W. (1993a). Phantom sensation, phantom pain, and stump pain. Archives of Physical Medicine and
Rehabilitation, 74(1), 79–91.
Davis, R. W. (1993b). Successful treatment for phantom pain. Orthopedics, 16(6), 691–695.
Davison, M., & Jones, B. (1995). A quantitative analysis of extreme choice. Journal of Experimental Analysis of
Behavior, 64, 147–162.
Devor, M. (1997). Phantom pain as an expression of referred and neuropathic pain. In R. Sherman (Ed.), Phantom
pain (pp. 33–58). New York: Plenum.
Dougherty, J. (1980). Relief of phantom limb pain after EMG biofeedback-assisted relaxation: A case report.
Behaviour Research and Therapy, 18(4), 355–357.
Freedman, R. R. (1991). Physiological mechanisms of temperature biofeedback. Biofeedback and Self Regulation,
16(2), 95–115.
Halbert, J., Crotty, M., & Cameron, I. D. (2002). Evidence for the optimal management of acute and chronic
phantom pain: A systematic review. Clinical Journal of Pain, 18(2), 84–92.
Biofeedback in the Treatment of Phantom Limb Pain 93

Harden, R. N., Rudin, N. J., Bruehl, S., Kee, W., Devang, D. K., Kooch, J., et al. (2004). Increased systemic
catecholamines in complex regional pain syndrome and relationship to psychological factors: A pilot study.
Anesthesia and Analgesia, 99(5), 1478–1485.
Iacono, R. P., Linford, J., & Sandyk, R. (1987). Pain management after lower extremity amputation. Neurosurgery,
20(3), 496–500.
Iacono, R. P., Sandyk, R., Bamford, C. R., Awerbuch, G., & Malone, J. M. (1987). Post-amputation phantom pain
and autonomous stump movements. Functional Neurology, 2(3), 343–348.
Jensen, T. S., Krebs, B., Nielsen, J., & Rasmussen, P. (1983). Phantom limb, phantom pain and stump pain in
amputees during the first 6 months following limb amputation. Pain, 17(3), 243–256.
Katz, J. (1997a). Appendix I. Literature review. In R. A. Sherman (Ed.), Phantom pain (pp. 171–180). New York:
Plenum.
Katz, J. (1997b). The role of the sympathetic nervous system in phantom pain. In R. Sherman, M. Devor, D. Jones,
J. Katz & J. Marbach (Eds.), Phantom pain (pp. 63–88). New York, London: Plenum.
Kooijman, C. M., Dijkstra, P. U., Geertzen, J. H., Elzinga, A., & van der Schans, C. P. (2000). Phantom pain and
phantom sensations in upper limb amputees: An epidemiological study. Pain, 87(1), 33–41.
Melzack, R. (1987). The short form of the McGill Pain Questionnaire. Pain, 30, 191–197.
Montoya, P., Larbig, W., Grulke, N., Flor, H., Taub, E., & Birbaumer, N. (1997). The relationship of phantom
limb pain to other phantom limb phenomena in upper extremity amputees. Pain, 72(1–2), 87–93.
Nikolajsen, L., Ilkjaer, S., Christensen, J. H., Kroner, K., & Jensen, T. S. (1997). Randomised trial of epidural
bupivacaine and morphine in prevention of stump and phantom pain in lower-limb amputation. Lancet,
350(9088), 1353–1357.
Pinzur, M. S., Garla, P. G., Pluth, T., & Vrbos, L. (1996). Continuous postoperative infusion of a regional anesthetic
after an amputation of the lower extremity. A randomized clinical trial. The Journal of Bone and Joint Surgery
American, 78(10), 1501–1505.
Rokicki, L. A., Holroyd, K. A., France, C. R., Lipchik, G. L., France, J. L., & Kvaal, S. A. (1997). Change
mechanisms associated with combined relaxation/EMG biofeedback training for chronic tension headache.
Applied Psychophysiology and Biofeedback, 22(1), 21–41.
Schwartz, N. M., & Schwartz, M. S. (2003). Definitions of biofeedback and applied physiology. In M. S. Schwartz
(Ed.), Biofeedback: A practitioner’s guide (3d ed., pp. 27–42). New York: Guilford.
Sherman, R. A. (1997). History of treatment attempts. In K. Heerman-Do (Ed.), Phantom pain (pp. 143–148).
New York: Plenum.
Sherman, R. A., Bruno, G. M., & Sherman, C. (1987). Psychological factors influencing chronic phantom limb
pain: An analysis of the literature. Pain, 28, 285–295.
Sherman, R. A., Gall, N., & Gormly, J. (1979). Treatment of phantom limb pain with muscular relaxation training
to disrupt the pain-anxiety-tension cycle. Pain, 6(1), 47–55.
Spielberger, C. D., Gorsuch, R. L., & Lushene, R. E. (1970). Manual for the State-Trait Anxiety Inventory. Palo
Alto: Consulting Psychologists.