244Journal of Pain and Symptom ManagementVol. 28 No. 3 September 2004
Massage Therapy for Symptom Control:
Outcome Study at a Major Cancer Center
Barrie R. Cassileth, PhD and Andrew J. Vickers, PhD
Integrative Medicine Service (B.R.C., A.J.V.) and Biostatistics Service (A.J.V.), Memorial
Sloan-Kettering Cancer Center, New York, New York, USA
Massage is increasingly applied to relieve symptoms in patients with cancer. This practice
is supported by evidence from small randomized trials. No study has examined massage
therapy outcome in a large group of patients. At Memorial Sloan-Kettering Cancer Center,
patients report symptom severity pre- and post-massage therapy using 0–10 rating scales of
pain, fatigue, stress/anxiety, nausea, depression and “other.” Changes in symptom scores
and the modifying effects of patient status (in- or outpatient) and type of massage were
analyzed. Over a three-year period, 1,290 patients were treated. Symptom scores were
reduced by approximately 50%, even for patients reporting high baseline scores.
Outpatients improved about 10% more than inpatients. Benefits persisted, with
outpatients experiencing no return toward baseline scores throughout the duration of 48-
hour follow-up. These data indicate that massage therapy is associated with substantive
improvement in cancer patients’ symptom scores.
2004;28:244–249.?2004 U.S. Cancer Pain Relief Committee. Published by Elsevier Inc.
All rights reserved.
J Pain Symptom Manage
Massage, cancer, pain, anxiety, depression, nausea, fatigue, clinical trials
In the 1940s, Rene A. Spitz learned that
foundling home infants, well fed and warm
but not held or touched, tended to wither away
and die. Of 91 such babies he observed, 27 died
in their first year of life, followed by seven
more in their second year; in other homes, up
to 90 percent died in early infancy. Babies who
survived in the institutions were classified as
hopeless.1,2Harlow’s studies of monkeys that
had been removed from their mothers showed
Address reprint requests to: Barrie R. Cassileth, PhD,
Memorial Sloan-Kettering Cancer Center, 1275 York
Avenue, H13, New York, NY 10021, USA.
Accepted for publication: December 23, 2003.
?2004 U.S. Cancer Pain Relief Committee
Published by Elsevier Inc. All rights reserved.
0885-3924/04/$–see front matter
similar results, plus major dysfunctions as they
developed and when they themselves became
mothers.3His major work, “Touching: The
Human Significance of the Skin,”4clarified a
major message: tactile stimulation is essential
to normal development and even to survival.
The use of human touch as an intervention
against pain and other problems has great
appeal. If effective, it could provide a non-inva-
sive, inexpensive adjunct to the management
of pain and other symptoms experienced by
patients with major chronic illnesses. Data
stances are promising. Massage therapy, de-
fined as manipulation of soft tissue areas of
the body, is offered in clinical settings to assist
Vol. 28 No. 3 September 2004 245Massage for Cancer-Related Symptoms
aches and pains.5It is increasingly used for
mately 20% of U.S. cancer patients seek mas-
sage therapy,6,7and approximately 70% of U.K.
hospices offer it.8Massage is included in treat-
ment guidelines such as those of the National
Comprehensive Cancer Network, which recom-
mends consideration of massage for refractory
Research supports such recommendations.
Several trials suggest that massage can reduce
pain in cancer patients at varying stages of dis-
ized cancer patients were randomized to
massage therapy or to control on a crossover
basis. Pain and anxiety scores fell by approxi-
mately 40% during massage compared with
little or no change during control sessions.12
Massage therapy was superior to control against
anxiety, nausea, fatigue and general well-being
in arandomized study ofpatientsawaitingbone
Massage has been available systematically at
Medicine Service in 1999. Twelve licensed mas-
sage therapists treat inpatients and also provide
massage therapy at our outpatient facility. Pa-
tients who receive integrative medicine ther-
apies record symptom scores before and after
treatment as a routine part of clinical manage-
ment. To avoid bias, staff not associated with
the evaluation, rather than therapists, provide
cards on which patients complete symptom
rating scales. These cards are placed privately
by patients in a closed box. Only research staff
workwith thecompletedcards.Datafrom these
scores, reported here, shed important light on
the management of common and often refrac-
tory symptoms experienced by patients with
Patients at MSKCC may self-refer to massage
therapy or may be referred by a health profes-
sional responsible for their care. About 50% of
inpatient referrals come from MSKCC nurses;
self- and family referrals account for a further
20% and 10%, respectively. Physicians and
other health professionals account for the re-
maining approximately 20%. Outpatient mas-
sage therapy is self-referred.
Three variationsof massage therapy are avail-
able to patients at MSKCC: standard (“Swed-
ish”) massage,5light touch massage, and foot
massage. A specific type of massage is requested
for almost all referrals. Patients receive the re-
quested type of massage approximately three
cases are requests for regular massage in which
either light touch or foot massage was given,
typically because the practitioner felt that a
weak or late-stage patient could not tolerate
a regular massage or because the patient was
too ill to move into a comfortable position to
receive standard massage therapy.
Massage sessions average 20 minutes in
length for inpatients and 60 minutes for outpa-
tients. Before and 5–15 minutes after massage
therapy, patients are given a 5 ? 8 inch card
with numerical rating scales for common symp-
toms: pain, fatigue, stress/anxiety, nausea, de-
pression, and “other.” Patients rate each on a
0 (“Not at all bothersome”) to 10 (“Extremely
Data from symptom cards for massage ther-
apy from April, 2000, when use of the rating
scales was initiated, to March, 2003 were ana-
lyzed. Combinations of interventions, such as
were excluded. Comparisons between different
types of massage or symptoms were conducted
by ANCOVA of the presenting symptom, with
baseline score as a covariate. The presenting
symptom was defined as that with the highest
baseline score. If more than one symptom was
scored equally high, the presenting symptom
depression, anxiety, nausea, fatigue, other. The
main analyses concern the initial episode of
care. This ensures that each patient is included
in the analysis only once. Analyses were con-
ducted by AV using Stata 7 statistical software
(Stata Corp., College Station, Texas).
Ethical approval for this retrospective review
of clinical data was given by the MSKCC IRB.
Cards were returned for 3,609 episodes of
care; post-therapy data were available for 3,359
(93%). Of these, 2,465 (73%) involved care of
an MSKCC patient, with smaller numbers for
cancer patients from other hospitals (94, 3%),
246 Vol. 28 No. 3 September 2004 Cassileth and Vickers
Types of Massage Therapy Received
961 (74)329 (26)1290 (100)
family members (78, 2%), staff (345, 10%) and
members of the public (377, 11%). Data re-
ported below reflect the initial episode of
care for the 1,290 different MSKCC patients
who provided post-treatment data.
As indicated in Table 1, the most commonly
administered touch therapies were standard
(Swedish) massage or foot massage, with far
fewer patients receiving light touch massage.
touch therapy during the same session, i.e.,
some foot massage and some Swedish mas-
sage during a single session. Foot massage
was predominantly used for inpatients; stan-
dard and light touch massage was more equally
common presenting symptom was anxiety (397,
31%), followed by pain (366, 28%) and fatigue
(312, 24%). Fewer than 10% of patients
reported greatest distress from depression,
nausea or another symptom.
The immediate effects of massage therapy
on symptoms are shown in Table 2. Although
major improvements in symptom scores are ap-
reduced by a mean of 54% (95% C.I. 52%,
56%)—Table 2 may actually underestimate
massage effects. Patients did not necessarily ex-
Improvements in Symptom Scores Following Massage Therapy
SymptomnBaseline Post-treatmentChange Improvement
Figures are given as mean (standard deviation).
aDefined as the symptom with the highest score at baseline.
number of zero or near zero scores. A patient
presenting with depression, for example, may
have reported no pain or nausea. For such
a patient, no improvement in pain and nausea
would be possible, thus diluting the apparent
effects of treatment on these symptoms at the
Accordingly, Table 3 includes only data for
symptoms rated four or higher at baseline, the
traditional threshold for considering a symp-
tom of at least “moderate” severity. The strong-
changes for fatigue, although a 43% reduction
in fatigue is clinically relevant. There was no
evidence of an attenuation of effect at high
baseline scores. For example, an approximate
45% improvement in pain scores was seen even
in the 244 patients with baseline scores of
seven or above, and in the case of anxiety, im-
provements were always close to 60%, regard-
less of baseline score.
reported symptom scores 0.56 points lower
(95% C.I. 0.27, 0.85; P ? 0.0002) than inpa-
tients, equivalent to an approximate 10%
greater improvement. Effects by type of mas-
sage are shown in Table 4. Adjusting for in- or
outpatient and baseline score, patients receiv-
ing Swedish and light touch massage had supe-
rior outcomes to those receiving foot massage
(0.32 points; 95% C.I. 0.03, 0.60 P ? 0.03). Pa-
tients receiving Swedish or light touch massage
had an average 58% improvement in severity of
their presenting symptom compared to a 50%
improvement in patients receiving foot mas-
sage. There was no significant difference be-
tween Swedish and light touch massage (0.41
points better response for light touch; 95% C.I.
?0.11, 0.13; P ? 0.12).
Vol. 28 No. 3 September 2004247Massage for Cancer-Related Symptoms
Improvements in Symptom Scores After Massage Therapy
Symptomn Baseline Post-treatment ChangeImprovement
Data include only baseline scores of four or higher. Figures are given as mean (standard deviation).
aDefined as the symptom with the highest score at baseline.
Similar effects appear for additional massage
therapy interventions. Percent improvement in
presenting symptoms for episodes two to five
for the same individuals are, respectively: 53%
(n ? 450; 95% C.I. 50%, 56%); 58% (n ? 203;
95% C.I. 54%, 63%); 56% (n ? 118; 95% C.I.
49%, 62%) and 61% (n ? 73, 95% C.I. 53%,
69%). Indeed, in a general linear model ad-
justed for baseline score and clustered by pa-
tient, the coefficient for each treatment was
negative and statistically significant (P ? 0.001),
suggesting that the effects of massage therapy
probably increase for each additional treatment.
We attempted to follow about one in four
patients (83 outpatients and 247 inpatients)
for up to two days to obtain data in addition to
their immediate post-treatment scores. Inpa-
tients and outpatients were assessed typically
two to five hours after treatment; outpatients
were additionally assessed approximately 24
and 48 hours after treatment. Data were ob-
tients (96%). These patients versus those we
did not attempt to follow beyond 48 hours
received similar therapies: 43%, 53%, 9% vs.
48%, 47%, 8% received regular, foot or light
massage, respectively. The two groups also had
comparable immediate responses to therapy
(53% vs. 54% improvement, P ? 0.6) and simi-
lar proportions of inpatients (78% v. 73%).
Baseline scores were slightly lower in patients
followed longer (6.3 vs. 6.7), suggesting that
patients followed beyond 48 hours were reason-
Differences in Effect by Type of Massage Therapy
Type of massagen Baseline Post-treatment Change Improvement
Figures are given as mean (standard deviation).
ably representative of the whole sample. As
shown in Table 5, the effects of touch therapy
forinpatients didnot persistin thelonger term.
Two to five hours after treatment, scores were
approximately 0.5 points higher than immedi-
ately after treatment. This suggests that inpa-
tient severity scores returned to baseline within
a day or so.
The effects of massage therapy lasted longer
for outpatients.Indeed, there is noevidencethat
symptom scores regress toward baseline values
massage therapy concerning the time course
of symptom improvement.
This is the largest study of massage for cancer
patients yet reported. Such studies typically
note sample sizes of six,142315and 54.13We
found no study of massage therapy with a
sample size greater than 100. This may reflect
that massage therapy services have been avail-
able only rarely until recently, at major cancer
centers. Our first conclusion, therefore, is that
implementationof ahigh volumemassage ther-
apy service is feasible at a major cancer center.
Major, clinically relevant, immediate im-
provements in symptom scores were reported
following massage therapy. Given the observa-
tional nature of this study, we cannot make con-
clusionsabout thecause ofthis effect. However,
248 Vol. 28 No. 3 September 2004 Cassileth and Vickers
Symptom Scores for Longer-Term Follow-Up
Later same day
n ? 74: 5.8 (2.2)
n ? 73: 2.7 (2.1)
n ? 53: 2.9 (2.3)
n ? 49: 2.7 (2.2)
n ? 38: 2.6 (2.4)
Later same day
n ? 237: 6.2 (2.4)
n ? 237: 3.1 (2.7)
n ? 237: 3.7 (2.9)
Figures are given as mean (standard deviation).
it is notable that the size of the effects found
are highly similar to those reported in prior
randomized trials of massage therapy in
cancer patients. For example, Grealish et al.
reported that pain and anxiety scores improved
during treatment from approximately 2.5 to 1.5
(40%) and from 5.4 to 3.2 (41%), respectively,
with no change in controls.12The comparable
figures reported by Stephenson et al. are 2.9 to
1.4 (53%) and 4.7 to 2.4 (50%), again, with
marginal change in controls (15). This suggests
that the results reported in randomized trials
can be achieved in the clinical setting.
The effects of massage were smaller and less
persistent for inpatients. Thereare two possible
explanations. First, inpatients are more subject
may undergo procedures or have medication
changed. Inpatientsalso tend to receive shorter
massage treatments in less comfortable settings
Fig. 1. Time course of treatment effects for outpa-
tients: symptom scores are given as means with stan-
than do our outpatients. The relationship be-
tween the length of massage treatment and
the size and duration of effects is worthy of
We found that Swedish and light touch mas-
sage were superior to foot massage, even after
controlling for baseline severity and location of
treatment. It may be that the effects of touch to
the body are more profound than touch given
only to the feet. However, it is also possible that
the apparently lesser effects of foot massage
reflect a case mix inadequately captured by
baseline symptom scores. For example, weak or
cachectic patients often receive massage just to
the feet rather than the whole body. Set against
such an explanation is that such patients also
receive light touch massage.
In conclusion, massage therapies apparently
lead to large, immediate improvements in
symptoms scores in cancer patients, even those
with very high baseline scores indicating sub-
stantial levels of pain, anxiety, or other symp-
benefit across the total of 48 hours studied. We
plan a prospective controlled trial for longer
periods of time to determine the duration of
achieves major reductions in cancer patients’
pain, fatigue, nausea, anxiety and depression.
Massage therapy appears to be an uncommonly
non-invasive and inexpensive means of symp-
tom control for patients with serious chronic
illness. It is non-invasive, inexpensive, comfort-
ing, free of side effects and greatly appreciated
by recipients. This non-randomized study sug-
gests that it is also markedly effective.
Anne Seidler collated and entered the data
for this article. The authors also thank the
MSKCC massage therapists, headed by Wendy
Miner, LMT and the staff of the MSKCC Bend-
heim Integrative Medicine Center.
of psychiatric conditions inearly childhood. The Psy-
choanalytic Study of the Child, Vol. 1. New York:
International Universities Press. 1945;1:53–74.
Vol. 28 No. 3 September 2004249Massage for Cancer-Related Symptoms
2. Spitz RA, Wolf KM. Anaclitic depression: an in-
childhood. The Psychoanalytic Study of the Child,
Vol. 2. New York: International Universities Press.
3. Harlow H. A variable-temperature surrogate
mother for studying attachment in infant monkeys.
Behav Res Meth 1973;5(3):269–272.
4. Montagu A. Touching: The human significance
of skin. New York: Harper & Row, 1986.
5. Vickers A, Zollman C. ABC of complementary
medicine. Massage therapies. BMJ 1999;319(7219):
6. Ashikaga T, Bosompra K, O’Brien P, et al. Use
cancer patients: prevalence, patterns and communi-
cation with physicians. Support Care Cancer 2002;
7. Bernstein BJ, Grasso T. Prevalence of comple-
mentary and alternative medicine use in cancer pa-
tients. Oncology 2001;15(10):1267–1272.
8. Vickers A. Massage and aromatherapy: a guide
for health professionals. London: Chapman and
9. GrossmanSA, Benedetti C,Payne R, etal. NCCN
practice guidelines for cancer pain. Oncology 1999;
10. Weinrich SP, Weinrich MC. The effect of mas-
sage on pain in cancer patients. Appl Nurs Res 1990;
11. Ferrell-Torry AT, Glick OJ. The use of therapeu-
tic massage as a nursing intervention to modify anxi-
ety and the perception of cancer pain. Cancer Nurs
12. Grealish L, Lomasney A, Whiteman B. Foot mas-
with cancer. Cancer Nurs 2000;23(3):237–243.
13. Ahles TA, Tope DM, Pinkson B, et al. Massage
therapy for patients undergoing autologous bone
marrow transplantation. J Pain Symptom Manage
14. Sims S. Slow stroke back massage for cancer pa-
tients. Nursing Times 1986;82(47):47–50.
15. Stephenson NL, Weinrich SP, Tavakoli AS. The
effects of foot reflexology on anxiety and pain in
patients with breast and lung cancer. Oncol Nurs