Shlay JC, Chaloner K, Max MB, Flaws B, Reichelderfer P, Wentworth D, Hillman S, Brizz B, Cohn DL, for the Terry Beirn Community Programs for Clinical Research on AIDS . Acupuncture and Amitriptyline for Pain Due to HIV-Related Peripheral NeuropathyA Randomized Controlled Trial. JAMA. 1998;280(18):1590-1595. doi:10.1001/jama.280.18.1590
From the Denver Community Programs for Clinical Research on AIDS, Denver, Colo (Drs Shlay and Cohn); School of Statistics, University of Minnesota, St Paul (Dr Chaloner); National Institute of Dental Research (Dr Max) and National Institute of Allergy and Infectious Diseases (Dr Reichelderfer), National Institutes of Health, Bethesda, Md; Division of Biostatistics, University of Minnesota, Minneapolis (Mss Wentworth and Hillman); and Social and Scientific Systems, Inc, Rockville, Md (Ms Brizz). Mr Flaws is an acupuncturist practicing in Boulder, Colo.
Context.— Peripheral neuropathy is common in persons infected with the human immunodeficiency
virus (HIV) but few data on symptomatic treatment are available.
Objective.— To evaluate the efficacy of a standardized acupuncture regimen (SAR)
and amitriptyline hydrochloride for the relief of pain due to HIV-related
peripheral neuropathy in HIV-infected patients.
Design.— Randomized, placebo-controlled, multicenter clinical trial. Each site
enrolled patients into 1 of the following 3 options: (1) a modified double-blind
2 × 2 factorial design of SAR, amitriptyline, or the combination compared
with placebo, (2) a modified double-blind design of an SAR vs control points,
or (3) a double-blind design of amitriptyline vs placebo.
Setting.— Terry Beirn Community Programs for Clinical Research on AIDS (HIV primary
care providers) in 10 US cities.
Patients.— Patients with HIV-associated, symptomatic, lower-extremity peripheral
neuropathy. Of 250 patients enrolled, 239 were in the acupuncture comparison
(125 in the factorial option and 114 in the SAR option vs control points option),
and 136 patients were in the amitriptyline comparison (125 in the factorial
option and 11 in amitriptyline option vs placebo option).
Interventions.— Standarized acupuncture regimen vs control points, amitriptyline (75
mg/d) vs placebo, or both for 14 weeks.
Main Outcome Measure.— Changes in mean pain scores at 6 and 14 weeks, using a pain scale ranging
from 0.0 (no pain) to 1.75 (extremely intense), recorded daily.
Results.— Patients in all 4 groups showed reduction in mean pain scores at 6 and
14 weeks compared with baseline values. For both the acupuncture and amitriptyline
comparisons, changes in pain score were not significantly different between
the 2 groups. At 6 weeks, the estimated difference in pain reduction for patients
in the SAR group compared with those in the control points group (a negative
value indicates a greater reduction for the "active" treatment) was 0.01 (95%
confidence interval [CI], −0.11 to 0.12; P
=.88) and for patients in the amitriptyline group vs those in the placebo
group was −0.07 (95% CI, −0.22 to 0.08; P=.38).
At 14 weeks, the difference for those in the SAR group compared with those
in the control points group was −0.08 (95% CI, −0.21 to 0.06; P=.26) and for amitriptyline compared with placebo was
0.00 (95% CI, −0.18 to 0.19; P=.99).
Conclusions.— In this study, neither acupuncture nor amitriptyline was more effective
than placebo in relieving pain caused by HIV-related peripheral neuropathy.
PERIPHERAL NEUROPATHIES are diagnosed in 30% to 35% of patients with
human immunodeficiency virus (HIV) and cause pain and dysesthesias.1,2 Symptomatic treatment includes antidepressants,
nonnarcotic and narcotic analgesics, anticonvulsants, and acupuncture.2,3 The use of these treatments is based
on anecdotal4 information and trials in other
We chose to examine the efficacy of 2 commonly used treatments, amitriptyline
hydrochloride and acupuncture, for HIV-related peripheral neuropathy. Amitriptyline
is frequently prescribed for neuropathic pain and has been shown to be an
effective treatment for diabetic, hereditary, toxic, and idiopathic neuropathies.6,7
Although several trials that reported examining acupuncture for chronic
painful conditions claim efficacy,8,9
these studies have methodological limitations, including small sample sizes
and inadequate controls for the nonspecific effects of acupuncture.9- 11 Meta-analyses of studies
of acupuncture for chronic pain show a response rate of approximately 70%
for acupuncture, 50% for "sham" acupuncture (needling points not considered
effective), and 30% for control treatments, such as sham transcutaneous electrical
To evaluate the effect of both a nonstandard and standard medical therapy
for peripheral neuropathy, we performed a multicenter, modified double-blind,
randomized, placebo-controlled study of the separate and combined efficacy
of a standardized acupuncture regimen (SAR) and amitriptyline for the relief
of pain caused by HIV-related peripheral neuropathy.
We used a 2×2 factorial design to determine whether SAR, amitriptyline,
or the combination was more effective than placebo. The SAR consisted of acupuncture
points chosen by the study acupuncturists and several consultants to be effective
for peripheral neuropathic pain. This regimen was compared with control points
that were not "true" points defined by any standard acupuncture text14 (Figure 1).
We compared the efficacy of amitriptyline with placebo capsules of identical
appearance. Enrollment in the factorial design began in May 1993, but patients
at some sites were reluctant to be randomized to receive amitriptyline and
some clinicians were unwilling to provide amitriptyline to their patients
because it was a commonly abused drug in their communities. The study design
was modified in March 1995 so that sites could choose only 1 of 3 options.
Each site could (1) continue to enroll into the factorial design (factorial
option), (2) enroll into a single-factor design of SAR vs control points
(acupuncture option), or (3) enroll into a single-factor design of amitriptyline
vs placebo (amitriptyline option) (Figure
Randomization schedules were prepared using random blocks stratified
by unit. Patients were randomized to treatment by the study units by telephoning
the Statistical Center at the University of Minnesota, Minneapolis. The unit
pharmacists were the only people unblinded to the placebo vs amitriptyline
assignment, and the acupuncturists were the only people unblinded to the SAR
vs control points assignments. The pain diaries and the assessments of pain
relief were collected by study staff who were blinded to the treatment assignments.
Patients were recruited from 11 units of the Terry Beirn Community Programs
for Clinical Research on AIDS, an organization sponsored by the National Institutes
of Health, which conducts clinical trials in primary care settings. The study
was approved by each institutional review board. All participants gave written
informed consent. To be eligible, participants had to be aged 13 years or
older; have documented HIV infection; have symptoms of HIV-related lower extremity
peripheral neuropathy, diagnosed by a physician based on history and clinical
examination; and have completed a baseline pain diary prior to randomization.
Antiretroviral therapy was allowed and dosages of analgesic medication or
herbal therapies used at randomization were maintained or reduced. The initiation
of new treatments during the study was discouraged but allowed when necessary.
Patients were excluded if they were being treated for an acute opportunistic
infection or malignancy except nonsystemic Kaposi sarcoma, were pregnant,
or had taken a tricyclic antidepressant or monoamine oxidase inhibitor 2 weeks
For the acupuncture comparison, patients were randomly assigned to receive
SAR or control points twice weekly during a 6-week induction phase, followed
by weekly treatment during an 8-week maintenance phase. This SAR was based
on a Chinese theory that peripheral neuropathy caused by diabetes and HIV-related
peripheral neuropathy have similar mechanisms. The SAR included spleen points
9, 7, and 6, with the additional supplemental points of Ba Feng (M-LE-8) for
complaints of pain or numbness in the toes, Ran Gu (kidney 2) for complaints
of pain or numbness in the soles, and Tai Ki (kidney 3) for complaints of
pain or numbness in the heel (Figure 1).14 The control points were located on the back of the
leg (Figure 1). For the SAR and
control points, acupuncture needles were inserted to a specified depth. Each
location was manipulated both superiorly and inferiorly. Then the needles
were reinserted into the specified point. After 10 to 15 minutes, the needles
were remanipulated and replaced into the original location for another 5
to 10 minutes. The depth of insertion was between 1.28 to 2.54 cm (0.5 to
1.0 in) for spleen point 9, 2.54 to 3.81 cm (1.0 to 1.5 in) for spleen point
7, and 1.5 to 3.05 cm (0.6 to 1.2 in) for spleen point 6. For the control
points, insertion was less than 1.28 cm (0.5 in). Study acupuncturists received
standardized training in the technique. In addition, a videotape of the acupuncture
and the control treatment was provided to each of the acupuncturists in
the study. To maintain blinding and to determine the need for supplemental
points, the acupuncturists asked all patients a series of standard questions,
irrespective of treatment arm. For those in the SAR group, spleen points
9, 7, and 6 were always used. Supplemental acupuncture points were used only
if the patient answered "yes" to the corresponding question. The control
points consisted of only 3 specified points.
For the amitriptyline comparison, the patients were randomized to receive
a 14-week course of either amitriptyline or placebo capsules by mouth once
a day. They were instructed to take them between 1 to 2 hours before bedtime.
An initial daily dose of 25 mg of amitriptyline hydrochloride was increased
every 2 to 3 days until a maximum dosage of 75 mg/d was reached.15,16
The placebo capsules were identical in appearance and taste to the active
capsules. Patients were followed up for the 14-week study period and for adverse
event monitoring for an additonal 8 weeks after the study treatment had discontinued.
Results were monitored by the HIV Therapeutic Trials Data Safety and
Monitoring Board of the National Institute of Allergy and Infectious Diseases,
National Institutes of Health, Bethesda, Md. Data monitoring used the Lan–De
Mets method17 as a guideline for early stopping
to account for increased type I error probability by examining the data before
the designed study end.
Patients rated their pain in a diary once daily, choosing from the Gracely
scale of 13 words that describe the intensity.18
The scale ranges from no pain (0.0), weak (0.45), mild (0.74), moderate (1.09),
strong (1.36), to extremely intense (1.75). The words had been assigned magnitudes
on the basis of ratio-scaling procedures that demonstrated internal consistency,
reliability, and objectivity.18 The scale has
distinguished active from control interventions in experimental and clinical
At the end of both the induction and maintenance phases, patients reported
their global pain relief (complete, a lot, moderate, slight, none, or worse)
after they were asked the following question: "Since the beginning of the
study, how would you rate the relief of pain and/or discomfort in your legs
and feet?" A study physician, trained in neurologic examination, tested the
patient at randomization and at 14 weeks. A neurologic summary score was computed
as an average of 3 separate scores for muscle strength, sensory ability, and
reflex. Each physician who performed the neurologic assessment reviewed a
videotape that detailed how the examination was to be completed. The patients
also completed a self-administered, 39-item, quality-of-life assessment tool.20 The complete tool, consisting of 11 different dimensions,
was administered at baseline and 14 weeks, and the dimension corresponding
to physical functioning was also administered at 6 weeks. To assess the effectiveness
of the blinding, all patients were asked to guess their treatment assignments
at 14 weeks. Patients were monitored for grade 4 adverse events and death.
Adverse experiences occurring within 8 weeks of study treatment were graded
on a 5-point severity scale (grade 5 corresponding to death) according to
a standardized toxicity scale. Any grade 4 or 5 event was reportable irrespective
of presumed relationship to study treatment.
Comparison of treatment groups for the primary end point of change in
pain, as measured by the pain diary, used a linear model with baseline characteristics,
clinical unit, and option (factorial or single factor) as covariates. If the
average weekly pain score for the sixth week of treatment was present, it
was used. If it was missing, the closest weekly average within the 6-week
visit window of 4 to 10 weeks was used. Similarly, this was done for the 14-week
end point and the visit window of 11 to 16 weeks. A linear model repeated
measures analysis of the weekly pain averages was also performed, with the
same explanatory variables.21 Estimates of
the difference between SAR and control points were calculated for each of
the 14 weeks. The global pain relief rating was analyzed using a log-linear
model, with likelihood ratio tests for differences among treatment groups,
which were adjusted for option.22
We verified that results from the 3 treatments could be pooled by checking
that the interaction term between acupuncture and amitriptyline in the factorial
option and the option by treatment interaction were nonsignificant.
Secondary outcomes were the permanent discontinuation of study treatments,
changes in quality of life, and changes in neurologic summary scores, which
were analyzed similarly to the primary end point. All analyses were on an
intent-to-treat basis. The evaluation of the blinding compared the patients'
guesses of the therapy received with the treatment group. Using a log-linear
model, we adjusted for option and for whether the patient reported moderate
or more pain relief with the 14-week global pain relief rating.
For the original 2×2 factorial design, a sample size of 260 patients
was calculated to provide a 90% power of detecting a mean difference between
treatments of 0.20 (half the difference between "moderate" and "mild" pain)
on the Gracely pain intensity scale using a type I error of .05 (2-sided).
After the study design was modified, sample size requirements were estimated
at 260 per group. In February 1997, the monitoring board recommended closing
the study because it concluded that the results were definitive for both acupuncture
and amitriptyline comparisons.
From May 1993 to February 1997, 250 patients were enrolled. Of those,
239 were in the acupuncture comparison (125 in the factorial option and 114
randomized to SAR or control points), and 136 were in the amitriptyline comparison
(125 from the factorial option and 11 randomized to either active or placebo
amitriptyline) (Figure 2). Baseline
characteristics (Table 1) were
similar in the active and control groups for both comparisons.
SAR vs Control Points.— The change in pain was not significantly different between the 2 groups
at either 6 or 14 weeks (Table 2).
Both groups showed improvement in pain from an average intensity of "moderate"
to "mild" (Figure 3). The estimated
difference of the SAR group compared with the control points group was 0.01
at 6 weeks (95% confidence interval [CI], −0.11 to 0.12; P =.88) and −0.08 at 14 weeks (95% CI, −0.21 to 0.06; P=.26). At 6 weeks, the SAR group had less pain relief
than patients in the control points group by 0.01 U and at 14 weeks, the SAR
group had 0.08 U more relief than patients in the control points group. Repeated
measures analyses of weekly pain averages during the entire 14-week period
gave weekly effects, which were small and nonsignificant (P values ranging from .10 to .94).
There were no significant differences in the quality of life, neurologic
summary score (Table 2), number
of grade 4 adverse events, deaths, or discontinuations. By 14 weeks, 20%
of patients randomized to the SAR group and 25% of those randomized to control
points group had discontinued treatment. Three patients assigned to the SAR
option and 10 assigned to the control points experienced a grade 4 adverse
The difference in the global pain relief rating between the 2 groups
was not significant at 6 weeks (P=.65). However,
at 14 weeks, there was a nominally significant difference (P=.03) with a slightly higher proportion of patients in the SAR group
reporting moderate or more pain relief than those in the control points group
(Table 3). However, after adjustment
for multiple comparisons, the result is not significant.
Amitriptyline vs Placebo.— The change in pain score at 6 and 14 weeks was not significantly different
between the active and placebo groups (Table 2). As with the SAR vs control points comparison, both
groups showed improvement over time (Figure
3). The estimated difference of amitriptyline compared with placebo
was −0.07 at 6 weeks (95% CI, −0.22 to 0.08; P=.38) and 0.00 at 14 weeks (95% CI, −0.18 to 0.19; P =.99). That is, at 6 weeks, patients taking amitriptyline had more
pain relief by 0.07 U than those taking placebo and there was no difference
at 14 weeks. Repeated measures analyses of weekly pain averages indicated
that the largest beneficial effect was at week 3 (P=.05),
but after adjusting for multiple comparisons, the result was not statistically
There were no statistically significant differences in quality of life,
neurologic summary scores (Table 2),
number of grade 4 adverse events, or deaths. Six patients assigned to the
amitriptyline and 2 assigned to placebo options experienced grade 4 adverse
events (P=.20). By 14 weeks, 35% of patients randomized
to either the amitriptyline or placebo groups had discontinued drug treatment.
The difference in the global pain relief rating between the 2 groups was not
significant at 6 weeks (P=.68) or 14 weeks (P=.81) (Table 3).
Factorial Option.— The test for interaction in change of pain between the 2 factors was
not significant at either 6 or 14 weeks (P=.17 and P =.31, respectively). There was no significant difference
in the change in pain among the 4 groups at either 6 or 14 weeks (P=.37 and P =.64, respectively). All study
groups in the factorial option showed improvement in pain.
Figure 2 shows the number
of patients providing pain diary data and global pain relief ratings at 6
and 14 weeks. To examine the sensitivity of the conclusions to missing data,
the analyses were repeated using 2 common methods to impute missing data.
The first assumes that the patients' missing data indicated no change in
their pain from baseline; the second uses the last value of the weekly pain
reported to calculate the end point. Under both methods to impute the missing
pain diary data, the results of the study did not reach statistical significance
for either comparison at either 6 or 14 weeks.
For the acupuncture comparison, although the patients' guesses and the
treatment assignments were not independent (P=.007,
data not shown), there was a strong association between the guess and the
global pain relief rating. Those reporting moderate or more relief at 14 weeks
tended to guess that they received the SAR. After adjusting for option and
the reported relief being moderate or more, the patients' guesses and the
treatment assignments were not independent (P=.02),
but the association was small. This differed in the amitriptyline comparison,
in which a large proportion of patients correctly guessed the study treatment,
irrespective of their level of pain relief (P<.001)
The main findings of this study show that treatment with this SAR had
little or no effect on HIV-related peripheral neuropathy compared with the
control points. Similarly, amitriptyline, as commonly used, was not significantly
more effective than placebo (Table 2
and Figure 3). All treatment groups
improved during the study period by the amount hypothesized in the design,
suggesting that the modest decline in pain scores in all groups was either
attributable to a placebo effect or patients entered the study at times of
symptomatic flares and improved spontaneously thereafter.
For the acupuncture comparison, the results were strengthened by 2 methodological
features of the trial. First, the sample size of approximately 120 patients
per treatment group is many times larger than those in previously published
trials of acupuncture,9 and the CIs were narrow,
making it unlikely that a large positive treatment effect was missed by chance.
Second, the control points appeared reasonably effective in preserving the
blinding (Table 4). Many of
the study clinicians and, presumably, the study participants were favorably
disposed toward acupuncture. If patients were able to guess their treatment
better than randomly, the resulting placebo effects would be expected to
bias the result in favor of this SAR,10,12,24
thus making our finding of a similar effect even more convincing.
We cannot completely rule out the possibility that the SAR had a modest
and delayed analgesic effect, in view of the nominally significant result
of SAR compared with control points on the global pain relief rating at 14
weeks, although this was not seen at 6 weeks. This is unlikely, however, in
view of the finding of no significant difference in the pain diary scores.
Our study was designed with a sample size that provided sufficient power to
detect even a small difference between the SAR and control points. The CIs
at both 6 and 14 weeks rule out any clinically meaningful beneficial effects
of SAR based on the primary end point of the pain diary scores.
One possible explanation for the lack of efficacy of the SAR is that
we chose the wrong "active points." Consensus on the SAR was reached by 8
acupuncturists before protocol implementation. Another explanation is that
the use of nonclassical points as a control provided a real effect and was
not an inert control. There is evidence from animal and human studies that
acupuncture at either classical or nonclassical locations may have analgesic
mechanisms such as the release of endogenous opioids27
or activation of other brain and spinal cord pathways that reduce pain.28
There is controversy over what constitutes an acceptable control group
for acupuncture studies.8,29 It
is possible that the novelty of an experience like acupuncture may generate
a placebo analgesic effect quite apart from specific effects produced by needling
specific points.30 Unless the study includes
a "sham" acupuncture group as a control, such nonspecific effects may bias
toward a result in favor of the active intervention.
The SAR chosen for this study differs from the practice of most acupuncturists,
who treat patients with individualized regimens.31
We chose to study standardized points to test the hypothesis that these specific
points promote analgesia for chronic foot and leg pain13
and because such a study is easier to blind and replicate. If the acupuncturists
had used individualized treatment, the results would not be generalizable
to other acupuncturists, and the treatment, if efficacious, could not be used
by other practitioners. Our approach enabled us to derive a conclusion about
these acupuncture points but not about individualized treatments.
Amitriptyline is used in the treatment of HIV-related peripheral neuropathy32 but was not effective in this study. The lack of
efficacy at 14 weeks was confirmed by the analysis of the secondary end points.
Although the 6-week CI did not completely rule out the beneficial effect of
0.20 that the study was designed to detect, there was no supporting evidence
of beneficial effect from any of the secondary end points. In addition, another
study in HIV-related peripheral neuropathy agrees with our findings.33 The indication that the blinding was not maintained
also confirms the lack of efficacy because unblinding tends to bias toward
a hypothesized active intervention.24,34
It is possible that a higher dose of amitriptyline would have resulted
in a larger treatment effect. We chose this dose based on common clinical
practice and on the only 2 published prospective randomized dose-response
studies of tricyclic antidepressants used for chronic pain.15,16
No previously controlled trials of amitriptyline in neuropathic pain
have followed up patients for longer than 8 weeks.33,35
Clinical trials of amitriptyline for neuropathies of diabetic and nondiabetic
etiologies have shown larger, short-term, clinically meaningful effects.6,7,19 Mechanisms for this
include facilitation of the analgesic action of norepinephrine and serotonin
released by endogenous analgesic systems16,19
and the blockade of sodium channels in peripheral sprouts from damaged nerves.36 Presumably, the neuropathological features of the
HIV-associated distal axonal neuropathy generate painful discharges resistant
to the analgesic actions of tricyclic antidepressants.37,38
In conclusion, this is the largest reported randomized, placebo-controlled,
clinical trial of symptomatic treatment for HIV-related peripheral neuropathy.
Overall, our results indicate that neither this SAR given over 14 weeks nor
amitriptyline hydrochloride, 75 mg/d, was effective in relieving pain and
neither therapy can be recommended for the treatment of HIV-related peripheral
neuropathy. Additional clinical trials are needed because there are no effective
treatments for this chronic debilitating condition.39