1542
REVIEW ARTICLE
The Efficacy of Traction for Back Pain:
A Systematic Review of Randomized Controlled Trials
Annette A. Harte, BSc, George D. Baxter, DPhil, Jacqueline H. Gracey, DPhil
ABSTRACT. Harte AA, Baxter GD, Gracey JH. The
efficacy of traction for back pain: a systematic review of
randomized controlled trials. Arch Phys Med Rehabil 2003;84:
1542-53.
Objective: To assess the efficacy of traction for patients
with low back pain (LBP) with or without radiating pain,
taking into account the clinical technique or parameters used.
Data Sources: A computer-aided search of MEDLINE, CI-
NAHL, AMED, and the Cochrane Collaboration was con-
ducted for randomized controlled trials (RCTs) in the English
language, from 1966 to December 2001.
Study Selection: RCTs were included if: participants were
over the age of 18 years, with LBP with or without radiating
pain; the intervention group received traction as the main or
sole treatment; the comparison group received sham traction or
another conservative treatment; and the study used 1 of 4
primary outcome measures.
Data Extraction: The study was conducted in 2 strands.
Strand 1 assessed methodologic quality using a specific criteria
list recommended by the Cochrane Back Review Group. The
strength of the evidence was then rated using the Agency for
Health Care Policy and Research system. Strand 2 applied
further inclusion criteria based on recommended clinical pa-
rameters. One reviewer conducted the selection and data ex-
traction.
Data Synthesis: Strand 1: 1 study scored 9 points (maxi-
mum score, 10 points); the other 12 scored between 0 and 3
points, indicating that most were of poor quality. Nine studies
reported negative findings, but only 1 study was of a high
quality. Three studies reported positive findings and 1 study
was inconclusive. Strand 2: only 4 trials having low method-
ologic quality were included, 2 of which reported negative
findings, and 2 positive findings.
Conclusion: The evidence for the use of traction in LBP
remains inconclusive because of the continued lack of meth-
odologic rigor and the limited application of clinical parame-
ters as used in clinical practice. Further trials, which give
attention to these areas, are needed before any firm conclusions
and recommendations may be made.
Key Words: Low back pain; Physical therapy techniques;
Randomized controlled trials; Rehabilitation; Traction.
2003 by the American Congress of Rehabilitation Medi-
cine and the American Academy of Physical Medicine and
Rehabilitation
From the School of Rehabilitation Sciences, University of Ulster, Jordanstown,
Northern Ireland.
Presented in part to Physiotherapy Research Society, April 2002, Keele, UK.
No commercial party having a direct financial interest in the results of the research
supporting this article has or will confer a benefit on the authors or any organization
with which the authors are associated.
Correspondence to Annette A. Harte, Sch of Rehabilitation Sciences, University of
Ulster, Jordanstown BT37 OQB, Ireland, e-mail: Aa.harte@ulster.ac.uk. Reprints are
not available.
0003-9993/03/8410-7831$30.00/0
doi:10.1053/S0003-9993(03)00294-6
Arch Phys Med Rehabil Vol 84, October 2003
PPROXIMATELY 80% OF THE population experiences
A low back pain (LBP) at some time in their lives; 90% will
resolve within 2 to 4 weeks, but 60% to 80% will have
recurrence within 1 year.1 Although back pain is the most
frequently presented disorder of the musculoskeletal system,
little consensus exists about its management, which often in-
cludes referral to physiotherapy where the choice of treatment
modalities is wide and varied. However, the efficacy of many
physiotherapeutic interventions remains questionable.2 One of
the treatment options is traction, which is often combined with
other treatments, for example, manual therapy, exercises, heat,
electrotherapy, and advice. Surveys would indicate the contin-
ued use of traction in the physiotherapy management of LBP:
7% in Southern Ireland and the United Kingdom,3 13.7% in
Northern Ireland,4 7% in the Netherlands,5 21% in the United
States,6 and up to 30% in acute LBP with sciatica in Canada.7
Traction may be applied in many forms: motorized lumbar
traction (traction applied by a motorized pulley), autotraction
(the patient exerts the traction force through a pulling or
pushing action), gravitational traction (traction through a sus-
pension device), or manual traction (forces exerted by the
therapist). In the past, traction has also been applied by use of
a pelvic harness when patients were admitted to the hospital for
bedrest, but this treatment is now apparently obsolete. To apply
the traction force, autotraction and manual traction rely on the
strength of the patient or therapist, gravitational traction on the
weight of the patient, and motorized traction on a motorized
machine. Autotraction, manual traction, and gravitational trac-
tion can be difficult to maintain for a specific period of time
because of fatigue or intolerance to the force or position by the
patient or therapist. Therefore, motorized traction can be more
successfully standardized for repeatability in a trial (which
encompasses the use of a split couch to eliminate friction
between the bed and the body). Research to date has included
all these modes of traction. The variation in treatment modes
may be an additional factor why conclusive results of tractions
effects have remained elusive.
Current Evidence
The evidence to date is conflicting; the UK Royal College of
General Practitioners guidelines8 state that traction does not
appear to be effective for LBP or radiculopathy, which has
discouraged many clinicians from using it. However, these
guidelines were based on the only available systematic review9
of randomized controlled trials (RCTs), which examined the
efficacy of traction for back and neck pain; 17 RCTs were
included in this review, 3 for back and neck pain and 14 for
LBP. The review concluded that because of the poor method-
ologic quality of the studies, it was not possible to formulate a
strong and valid judgment about the efficacy of either lumbar
or cervical traction.
The Cochrane database of abstracts of reviews of effective-
ness10 agreed with this conclusion and highlighted the need for
rigorous RCTs before research-based guidelines could be
agreed for traction. The lack of consensus in this area is further
highlighted by the recent Philadelphia guidelines,11 which state
 LUMBAR TRACTION, Harte
that there is poor evidence to include or exclude lumbar trac-
tion as an intervention for acute or chronic LBP.
No further specific reviews on the effectiveness of traction
have been undertaken since the 1995 review,9 which only
included articles published up to June 1992. However, several
new RCTs have been published, stimulating the need for a new
review focused on traction in LBP to ascertain if conclusions
may now be drawn about the efficacy of traction in LBP.
Clinical Appropriateness
In the past, systematic reviews have concentrated solely on
methodologic quality; however, little attention has been given
to the quality and appropriateness of the intervention, despite
the fact that inadequacies in these areas may also lead to
substantial bias.12 The practitioner implementing evidence-
based practice needs to know that the evidence is based on
findings from high-quality research methods and that the re-
search intervention reflects the treatment protocols that he/she
uses. If the research evidence does not address these 2 factors,
clinicians will not implement the findings. It is difficult to
address clinical appropriateness in research into LBP manage-
ment by physical therapy because treatment tends to be tailored
to fit the patients needs and no 2 patients are identical. Fur-
thermore, treatments are usually complex and administered
within a package of care. Gracey4 showed that, on average,
only 2.3% of patients received advice alone and 1.6% were
treated with the McKenzie regime alone; all others received
treatments that were combined throughout a wide range. In
contrast, research designs typically look at the efficacy of
single treatments in isolation. Indeed, in the present review, all
the studies used traction as the sole intervention. The use of
various combinations of treatment has an obvious implication
for the design of future trials. It is also important to look at the
interventions individually and to examine whether they are
acceptable to clinical practice. To explore this issue, we inves-
tigated whether taking account of the traction treatment param-
eters would alter the conclusions drawn from the available
evidence. The application of optimal treatment weights, length
of the treatment session, frequency of treatment, and length of
the overall traction program are areas that have had limited
research. Therefore, the determination of optimal parameters
can only be drawn from recommendations from expert opin-
ion,13-16 supplemented by limited evidence on the mechanical
and physiologic effects of traction.17-24
Determination of Optimal Treatment Parameters
In deciding what traction weight to apply, one must consider
2 aspects: (1) what weight will overcome friction between the
body and the bed; and (2) what amount of force is required to
exert an effect on the lumbar spine. Judovich25 showed that a
traction force of 26% of the patients body weight was required
to overcome friction. The use of a split tabletop with friction-
free rollers reduces this to a negligible amount. In its absence,
a force in excess of 26% of the body weight must be used
before any effect can be produced at the lumbar spine.
Optimal weights for traction have been investigated by as-
suming that the effect of traction is achieved by intervertebral
widening or reduction of disk protrusion; however, only the
former has been demonstrated experimentally.7,24 Despite these
studies, it remains unclear what magnitude of force is required
to cause the desired effect in the intact human spine. The
mechanism by which traction may have its effects is not fully
understood, and the neuromodulation of pain, which may re-
quire very low weight, must also be considered as a possible
effect. This notwithstanding, clinical experts recommend using
motorized traction on a friction-free surface and advocate a
1543
wide range of traction weights within their treatment regimes.
Maitland13 suggests 10 to 65kg, with an average of 30 to 45kg;
Cyriax14 suggests 40 to 85kg; Grieve15 suggests 13 to 34kg;
and Hicklings26 suggests 32 to 68kg.
Absent research findings, clinical experts recommend a vari-
ation of treatment times: Cyriax14 suggests 30 to 45 minutes;
Hicklings26 20 to 40 minutes, with an average of 30 minutes;
and Maitland13 and Grieve15 suggest 10 minutes for the initial
treatment and 15 minutes thereafter.
Clinical experts showed more agreement in terms of the
frequency of treatment. Most suggested that acute nerve root
problems require traction on a daily basis13-15 for 5 to 6 days a
week, for at least 2 weeks. Chronic problems may be treated 2
to 3 times weekly for at least 3 weeks because the chronic
problem may be difficult to resolve.13-15
The aim of the present review was to assess the efficacy of
traction for LBP patients with or without radiating pain. The
main objectives of the review were to ascertain if the conclu-
sion of the van der Heijdan9 review is altered in light of the
available evidence since 1992, and to ascertain whether taking
account of recommended clinical parameters for the use of
traction alters the conclusion. The subsidiary objectives were to
assess the relevance of the mode of traction, and to assess the
relevance of traction for acute and chronic LBP.
METHODS
The RCTs published in English were retrieved from elec-
tronic searches of MEDLINE, CINAHL, AMED, Cochrane
Collaboration, and PEDro databases, covering the period 1966
to December 2001. Key words used were: back pain, traction,
lumbar traction, autotraction, manual traction, gravitational
traction, physiotherapy, physical therapy, randomized trial,
clinical trial, and VAX-D. Key words were combined. A hand
search of relevant journals and reference lists was also under-
taken to identify additional references.
To be included in the review, a trial had to meet the follow-
ing criteria: (1) a randomized controlled design; (2) participants
were over the age of 18 years and were treated for LBP with or
without radiating pain, defined as pain localized below the
scapula and above the cleft of the buttocks, with or without
radiation to the lower extremities; (3) at least 1 of the inter-
vention groups had to receive traction as the main or sole
intervention (traction could be in the form of motorized trac-
tion, autotraction, manual traction, or gravitational traction
given continuously or intermittently); (4) the comparison group
had to receive sham traction (placebo traction or low-weight
traction, which is perceived to be ineffective) or another con-
servative treatment modality (trials comparing different types
of traction were excluded); (5) the study had to use at least 1 of
4 primary outcomes2: pain (eg, visual analog scale, numeric
rating system), a global measure of improvement (eg, overall
improvement, subjective improvement of symptoms, propor-
tion of patients recovered), a back painspecific functional
status (eg, Roland-Morris Disability Questionnaire), or generic
functional status or quality of life (eg, the Medical Outcomes
Study 36-Item Short-Form Health Survey, Nottingham Health
Profile, Sickness Impact Profile); and (6) RCTs that included
participants with LBP caused by specific pathologic entities,
such as infection, neoplasm, osteoporosis, rheumatoid arthritis,
or fractures, were excluded.
Strand 1: Methodologic Quality
To assess the methodologic quality of the studies, we used a
specific criteria list (table 1) and a uniform operationalization
of criteria (table 2) according to the Cochrane Back Review
Group recommendations.2 The criteria list, which has already
Arch Phys Med Rehabil Vol 84, October 2003
1544 LUMBAR TRACTION, Harte
Table 1: Criteria List for Assessing Methods Quality2
Patient selection
a) Were the eligibility criteria specified?
b) Treatment allocation
1) Was a method of randomization performed?
2) Was the treatment allocation concealed?
c) Were the groups similar at baseline regarding the most important prognostic indicators?
Interventions
d) Were the index and control interventions explicitly described?
e) Was the care provider blinded to the intervention?
f) Were cointerventions avoided or comparable?
g) Was the compliance acceptable in all groups?
h) Was the patient blinded to the intervention?
Outcome
i) Was the outcome assessor blinded to the intervention?
j) Were the outcome measures relevant?
k) Were adverse effects described?
l) Was the withdrawal/dropout rate described and acceptable?
m) Timing follow-up measurements
1) Was a short-term follow-up measurement performed?
2) Was a long-term follow-up measurement performed?
n) Was the timing of the outcome assessment in both groups comparable?
Statistics
o) Was the sample size for each group described?
p) Did the analysis include an intention-to-treat analysis?
q) Were point estimates and measures of variability presented for the primary outcome?
NOTE. Internal validity criteria: b, e, f, g, h, i, j, l, n, p; descriptive criteria: a, c, d, k, m; statistical criteria: o, q.
been used in a number of reviews of back pain studies, includes
all the criteria of Jadad et al27 and Verhagen et al.28 The list
consists of internal validity criteria, descriptive criteria, and
statistical criteria. Internal validity refers to characteristics of
the study that relate to selection bias (criterion b), performance
bias (criteria e, f, g, h), attrition bias (criteria l, p), and detection
bias (criteria i, j, n). The descriptive criteria (criteria a, c, d, k,
m) refer to the external validity of the trial. The statistical
criteria (criteria o, q) indicate whether calculations can be made
and conclusions drawn independently of the opinion of the
authors. Each criterion was scored yes (Y), no (N), do not
know (DK), and there was no weighting of criteria. The total
score was 19.
The internal validity criteria were used to define the studys
methodologic quality. These criteria total 10 in score. The
results of the total score and the internal validity score were
calculated and are presented for included articles.
Each study was rated for its methodologic quality. The rating
was based on the internal validity score (n10); a trial was
considered to be of high quality if it scored over 50% and of
low quality if it scored below this.29
Overall Outcomes
The outcome of each study was determined by the results of
the primary outcome measures (at least 1 primary outcome). A
study was considered positive if traction was more effective
than the referenced treatment. A study was considered negative
if there was no difference between traction and the referenced
treatment, or if the referenced treatment was more effective
than traction. A study was considered inconclusive if there
were conflicting results between the primary outcome mea-
sures, for example, a positive result on 1 primary outcome and
a negative result on another primary outcome.
Following this, the results were summarized by applying the
US clinical practice guideline for rating acute low back prob-
Arch Phys Med Rehabil Vol 84, October 2003
Y/N/DK
Y/N/DK
Y/N/DK
Y/N/DK
Y/N/DK
Y/N/DK
Y/N/DK
Y/N/DK
Y/N/DK
Y/N/DK
Y/N/DK
Y/N/DK
Y/N/DK
Y/N/DK
Y/N/DK
Y/N/DK
Y/N/DK
Y/N/DK
Y/N/DK
lems in adults.30 This rating system consists of 4 levels of
scientific evidence based on the quality and the outcome of the
studies: (1) strong evidence, generally consistent findings in
multiple high-quality RCTs; (2) moderate evidence, generally
consistent findings in 1 high-quality RCT and 1 or more low-
quality RCT or by generally consistent findings in multiple
low-quality RCTs; (3) limited or conflicting evidence, only 1
RCT (either high or low quality) or inconsistent findings in
multiple RCTs; and (4) no evidence, no RCTs.
Categorization of Studies
Study results were further subcategorized to present findings
for the different types of traction used (ie, motorized traction,
manual traction, autotraction, gravitational traction). Results
were also examined for traction versus placebo or sham trac-
tion, and traction versus another conservative treatment modal-
ity. Where possible, study populations were classified into
acute (6wk), subacute (6 12wk), and chronic (12wk) cat-
egories. This allowed questions to be addressed regarding
tractions effectiveness in acute and chronic LBP and also the
effectiveness of different types of traction used.
Statistical Analysis
Because of the lack of high-quality studies and the hetero-
geneous nature of patients and statistical data, a meta-analysis
was considered inappropriate and a qualitative review was
undertaken.
Strand 2: Clinical Parameter Criteria
To gain inclusion in this section of the review, the studies
from strand 1 had to meet the following inclusion criteria (table
3): (1) use of a friction-free couch or a traction weight that used
at least 26% of body weight (to overcome friction), as well as
a weight or traction force (for therapeutic rationales) given on
 LUMBAR TRACTION, Harte 1545

Table 2: Operationalization of the Criteria List for LBP Studies2

a. To score a yes the radiation pattern of back pain and the duration of the disorder must be described appropriately.
b1. A random (unpredictable) assignment sequence. Methods of allocation using date of birth, date of admission, hospital
numbers, or alternation should not be regarded as appropriate.
b2. Assignment generated by an independent person not responsible for determining the eligibility of the patients. This person
has no information about the persons included in the trial and has no influence on the assignment sequence or on the
decision about patient eligibility.
c. To receive a yes, groups must be similar at baseline regarding age, duration of complaints, percentage of patients with
radiating pain, and value of main outcomes.
d. Adequate description of type, modality, application technique, intensity, duration, number of frequency of sessions for both
the index intervention and control interventions, so that others could replicate the treatment.
e. The reviewer determines when enough information about the blinding is given to score a yes.
f. Cointerventions should either be avoided in the trial design or comparable between the index and control group.
g. The reviewer determines when the compliance to the intervention is acceptable, based on the reported intensity, duration,
number, and frequency of sessions for both the index intervention and the control interventions.
h. The reviewer determines when enough information about the blinding is given to score a yes.
i. The reviewer determines (per outcome parameter) when enough information about blinding is given to score a yes.
j. The reviewer determines whether the outcome measures were relevant. For back pain, we recommend using pain, a global
measure of improvement, back-specific functional status, and generic functional status.
k. Each event should be described and correctly attributed to the allocated treatment: if it is explicitly reported that no adverse
effects have occurred, a yes should be scored.
l. Participants included in the study but who did not complete the observation period or were not included in the analysis must
be described. If the percentage of withdrawals and dropouts does not exceed 20% for short-term follow-up and 30% for long-
term follow-up and does not lead to substantial bias, a yes is scored.
m1. Outcome assessment at the end of the intervention period.
m2. Outcome assessment more than 3mo after randomization.
n. Timing of outcome assessment should be identical for all intervention groups and for all important outcome assessments.
o. To be presented for each group at randomization and for most important outcome assessments (ie, in contrast to previous
lists, there is no preset cutoff point to determine whether the sample size is sufficient).
p. All randomized patients are reported/analyzed for the most important moments of effect measurement (minus missing values)
irrespective of noncompliance and cointerventions.
q. Both point estimates and measures of variability should be presented (to be scored for each important parameter separately).
Point estimates are: means, medians, modes, etc. Measures of variability are: standard deviations, 95% confidence in intervals, etc.

a friction-free couch and ranging from 10 to 85kg; (2) length of
treatment times at each session of at least 10 minutes, with an
average time of at least 15 to 30 minutes; and (3) frequency of
treatment with acute nerve root treated daily 5 of 7 days for at
least 2 weeks, and chronic nerve root treated 2 to 3 times
weekly for at least 3 weeks.
RESULTS
Results of Inclusion Procedure
After a search of the databases, we identified 19 RCTs; only
13 of which were included in the final review (table 4). Six
Table 3: Criteria for Strand 2
Clinical Parameters for Traction Application
Criterion 1:
Use of a friction-free couch, or in its absence, a traction weight
that was more than 26% of body weight. In addition to this, a
weight or traction force ranging from 10 to 85kg.
Criterion 2:
Length of treatment times at each session should be a minimum
of 10 minutes with an average time of at least 15 to 30 minutes.
Criterion 3:
Frequency of treatment: acute nerve root treated 5 of 7 days for
a minimum of 2 weeks, chronic LBP treated 2 to 3 times a week
for at least 3 weeks.
articles were excluded; 3 articles31-33 because they compared
different applications of traction, 1 article34 because it included
subjects under the age of 18 years, and 2 articles5,35 because
they were preliminary reports of a main study that was included
in the review.36 Since 1992, 4 articles36-39 fulfilled the selection
criteria and were included in this review.
Of the 13 articles presented, 5 articles36,40-43 were trials with
motorized traction using a split tabletop; other modes of trac-
tion included motorized traction with a plain tabletop
(n3),38,39,44 traction with bedrest (n1),45 autotraction
(n2),46,47 gravitational traction (n1),37 and manual traction
(n1).48
Five of the original articles included in the van der Heijdan
review9 were not included here; 3 articles49-51 because they
were not in the English language, 1 article33 that included
autotraction versus manual traction, and an earlier copy of the
pilot study52 discussed above in its native language. The 3
nonEnglish-language articles were of low methodologic qual-
ity, noted by the original review.9
Strand 1: Methodologic Quality
The quality of the studies ranged from 0 to 9 points on the
internal validity score. Only 1 study36 scored above 5 points,
making it a high-quality study; the other studies ranged from 0
to 3, with many studies scoring 0 or 1 point (table 5). This
reflects the overall low quality of these studies.
Particular areas of methodologic concern were baseline com-
parability (criterion c), blinding of the care provider (criterion
e), lack of comparability of cointerventions (criterion f), lack of

Arch Phys Med Rehabil Vol 84, October 2003

1546 LUMBAR TRACTION, Harte

Table 4: Lumbar Traction Studies (grouped with comparable studies)

Pain Radiation/
Neurologic Acute Subacute Indexed Referenced Results of Method Study
Study Signs Chronic Treatment Treatment Primary Outcomes Score Conclusion

Coxhead et al42 Sciatic Not clearly Motorized traction (split No treatment, VAS (100 to 100) no 1 No conclusive
symptoms, stated top) alone or with manipulation, significant difference evidence that any
radiation at manipulation, exercises, or between groups at 4wk. of the 4 individual
least to exercises, corset in corset alone or in GSE: no significant treatments was
buttock crease any combination of 1, any combination. difference between effective, but it is
2, 3, or 4 treatments. Factorial design 16 groups at 4wk, 4mo, possible that each
Similar treatment groups in total. 16mo. conferred some
duration time, greater All groups received Return to work: no benefit,
number of treatments, SWD and advice significant difference at particularly
less time per (n166) 4wk. manipulation.
treatment. Daily for More significant results
first week then with (P.01) with greater
decreasing frequency. number of treatments.
All groups received
SWD and advice
(n166)
Lidstrom and Pain referred 1mo to 1y Motorized split tabletop. Group 1: hot packs Measured at end of 0 Statistically
Zachrisson43 below the (acute- Intermittent pelvic and rest (time of treatment 4wk. significant
knee with chronic mix) traction (4s pull, 2s treatment session GSE: noticeable improvement for
limited rest, Fowler position). average time of the improvement, status the group
straight-leg 20min each treatment other treatments) quo, worse. Significant receiving traction
raising session. Traction (n21). improvement (P.01) and exercises.
weight: subjects Group 2: hot packs, between the traction
weight in kg converted massage, group and group 1 and
to pounds (eg, 54kg mobilization and group 2.
body weight, traction strengthening
weight of 54lb). exercises (15min)
Isometric exercises: (n22).
abdominal and hip
extensors. 10
treatments over 1mo
(n21).
Matthews et al40 Back pain plus 013 wk (acute- Traction friction-free Infrared, 15min, 3 4 in first 2wk and then at 1 Traction hastens
nerve root subacute) couch: weight at least weekly, 23 wk 1, 3, 6 and 12mo. recovery by an
pain 45kg (to relieve pain), period. GSE: 6-point score. amount of
30min, daily, Both groups given borderline
maximum of a 3-wk posture advice, statistical
period. painkillers, and significant (P.05
Both groups given corset (n60). 0.1) vs control.
posture advice,
painkillers, and corset
(n85).
Sherry et al39 All had leg pain 0.2530y Intermittent traction TENS for 30min daily Measured before, 3 Traction gave a
(subacute- 22.542.75kg (5095lb), for 20d then once immediately after, and at positive result
chronic) 30min, 15 cycles of weekly for 4wk. No 6mo. VAS: significant (P.0001).
decompression. 5 parameters given. improvement with
per week for 4wk then traction. 50%
once weekly for 4wk. improvement was set as
being significant. 68.4%
improved with VAX-D.
Werners et al38 LBP and pain 5y to 10y? Motorized traction with Interferential: Six 10- ODI: both groups 3 Traction and IFT is
radiating into (chronic) massage (all in the min sessions over improved significantly, ineffective. Cannot
leg bed). Six 10-min 1421d (n74). but no significant rule out positive
sessions over 1421d difference between effect of massage.
(n73). groups.
VAS: no significant
difference between
groups GSE (3mo).
Ljunggren et al48 Radiating pain 930wk Manual traction (to Isometric exercises: RMDQ: no significant 2 The 40%
and (subacute- achieve pain relief abdominal, back, hip difference between improvement in
neurologic chronic) 300N) 10min once a and thigh. 20min groups. both groups
signs L5-S1. day or in a few cases daily for 57d VAS (10-cm line): no conforms to the
Disk herniation 5min twice daily for (n26). significant difference placebo effect.
on myleogram 57d (n24). between groups.
GSE: no significant
difference before and
after treatment (57d)
Beurskens et al36 Nonspecific LBP 6wk Motorized (split top) Sham traction 3%20% RMDQ, VAS, 7-point 9 Does not support
with or (subacute- traction 27%50% body weight. 12 patient-rated the claim that
without chronic) body weight. 12 treatment sessions improvement, work traction is
radiation treatment sessions of of 20min over 5-wk absence, 512wk and 6mo. efficacious for
20min over 5-wk period (n74). Both groups improved but patients with LBP.
period (n77). no significant difference
between groups.

Arch Phys Med Rehabil Vol 84, October 2003

 LUMBAR TRACTION, Harte 1547

Table 4: (Contd) Lumbar Traction Studies (grouped with comparable studies)

Pain Radiation/
Neurologic Acute Subacute Indexed Referenced Results of Method Study
Study Signs Chronic Treatment Treatment Primary Outcomes Score Conclusion

Matthews and Sciatic or 1.546wk Traction: plain couch. Traction: plain couch, 3wk and 6wk from start of 0 The lack of statistical
Hicklings44 femoral nerve (acute- Traction weight: traction weight did trial. Dropouts too great significance may
distribution chronic) 3658.5kg (80135lb), not exceed 9kg by 3mo to use results. well be due to the
radiating 30min daily, 5d/wk for (20lb) (amount Results at 6wk: overall small sample
distal to knee 3 consecutive weeks insufficient to pain reduction 28.8% in number.
with or (n13). overcome friction). 7 traction group and
without back crossed over into 18.9% in the control
pain treatment group group (NS).
(n14).
Pal et al45 Back pain and Median value Patients admitted Patients admitted Measured at admission, 1 No significant
sciatica not group A: 2d; hospital; traction as hospital; traction as weekly in hospital and improvement with
defined group B: 56d part of bedrest; 5.5 part of bedrest; 1.4 follow-up at OPD clinic. traction.
8.2kg depending on 1.8kg (n15). Significant improvement
body weight (n24). in pain score 1wk
(P.01), at 2wk (P.001).
No significant difference
between groups.
Weber et al46 (Spina- All had radiating Not stated Autotraction: 7-part Sham autotraction. Measured pain rating 1 No significant
trac) pain, frame 4070kPa force, Same protocol as before and after improvement with
neurologic 1012s traction with traction group treatment (no time span autotraction.
symptoms 1012s rest for 20min (n23). given). No significant
and signs once a day. Maximum difference between
corresponding 8 sessions (n21). groups.
to L5-S1 root
Weber41 All had radiating Not stated Friction-free traction Simulated traction. Immediately before and 3 Traction therapy
pain, couch. Force of up to 7kPa after treatment (57d). does not alter the
neurologic Force: 13 body weight, (just enough to Pain radiating. No course of radicular
symptoms 5s pull, 5s rest for tighten harness) significant difference symptoms and
and signs 20min, once daily for (n35). between groups. signs due to a
corresponding 57d (n37). herniated IVD.
to L5-S1 root
Moret et al37 Back pain with Not recorded Vertical traction: 45min Bedrest 12wk (n8). RMDQ and pain rating: 3 No conclusion can
pain radiating 4 day or 30min 6 2wk. be drawn
to below the per day over 12wk GSE: 3wk post-treatment. regarding the
knee with period including Higher mean reduction of effect of vertical
neurologic bedrest (n8). leg pain in traction traction or lumbar
signs group (NS). radicular
No change in back pain. syndrome.
RMDQ difference of 2.2
(95% CI, 1.57 to 5.97).
Larsson et al47 Lumbago- 2wk and Autotraction, angle of Corset same as Before and at 1, 3wk after 0 There were clear
sciatica with 312mo pull could be altered traction group. starting treatment; pain differences
or without (acute- to get alleviation of Same advice. intensity and a between the 2
neurologic subacute) pain. Side lying or Bedrest for 1wk then questionnaire at 3mo. groups. It appear
deficit supine. Treatment 1h mobilize (n41). Significant results. that the patients
3 in 1wk. Corset Completely recovered at who did not
fitted to maintain 1wk (P.05), 3wk (NS). become pain free
lumber lordosis after Completely recovered or from back or leg
traction. Advice and free from leg pain at pain did not
equalization of leg 1wk (P.001) and at 3wk achieve any
length. Bedrest for (P.005). lasting alleviation
1wk then mobilize Completely recovered or of pain.
(n41). pain free from pain in
back or leg at 1wk
(P.001) and 3wk
(P.01).
After 3mo, no significant
difference between
groups (persisting and
free from symptoms).

Abbreviations: CI, confidence interval; GSE, global subjective outcome by patient; IFT, interferential therapy; IVD, intervertebral disk; NS, not significant; ODI, Oswestry
Disability Index; OPD, outpatient department; RMDQ, Roland-Morris Disability Questionnaire; SWD, short-wave diathermy; TENS, transcutaneous electric nerve stimulation; VAS,
visual analog scale; VAX-D, vertebral axial decompression.

discussion about compliance to treatment (criterion g), lack of
patient blinding (criterion h), lack of blinding of the assessor
(criterion i), lack of relevant outcome measures (criterion j),
description of adverse effects (criterion k), timing of outcome
assessment not described (criterion n), no intention to treat
(criterion p), and lack of statistical data (criterion q). Articles
typically met the following criteria: explicit description of
interventions (criterion d), description of dropouts (criterion l),
and short-term (and to a lesser extent long-term) follow-up
(criterion m).
The quality of the articles published since 1992 was better
than the studies published earlier. The median of the total
methodologic score (n19) for the articles published since
1992 was 9.5 (range, 8 17), and for articles before this it was

Arch Phys Med Rehabil Vol 84, October 2003

1548 LUMBAR TRACTION, Harte

Table 5: Methods Scores

Total Internal High/Low Outcome

Study a b1 b2 c d e f g h i j k l m1 m2 n o p q Score Validity Quality of Study

Beurskens et al36* Y Y Y Y Y N Y N Y Y Y N Y Y Y Y Y Y Y 17 9 High Negative

Moret et al37* N Y Y N Y N Y ? N N Y Y Y Y N ? Y NA Y 10 3 Low Inconclusive
Tendency
to Positive
Sherry et al39* Y Y ? ? N N Y ? N Y N N Y Y Y ? Y N Y 9 3 Low Positive
Werners et al38* N Y N N Y N ? ? N N Y N Y Y Y ? N Y Y 8 3 Low Negative
Matthews et al40 Y Y ? N Y N N ? N Y N N N Y Y ? Y N N 7 1 Low Negative
Weber41 N N ? ? Y N ? Y Y Y N Y N Y N ? N N N 6 3 Low Negative
Ljunggren et al48 Y N ? ? Y N ? ? N Y N N Y Y N ? N N N 5 2 Low Negative
Coxhead et al42 Y N ? N N N ? ? N ? N N Y Y Y ? N Y N 5 1 Low Negative
Larsson et al47 N N ? Y N N ? ? N ? N N Y Y Y ? Y N N 5 0 Low Positive
Pal et al45 N N ? N N N ? Y ? ? N N Y Y ? ? Y N N 4 1 Low Negative
Matthews and Hickling44 Y N ? ? Y N N N ? ? N N N Y Y ? N N N 4 0 Low Negative
Lidstrom and Zachrisson43 Y Y ? N Y N ? ? N N N N ? Y N ? N N N 4 0 Low Positive
Weber et al46 (Spina-trac) N ? ? ? N N ? ? Y ? N N N Y N ? N N N 2 1 Low Negative

NOTE. The highlighted criteria refer to internal validity (b, e, f, g, h, i, j, l, n, p,) and give the internal validity score. A cutoff of 5 points
distinguishes between high- and low-quality articles.
Abbreviation: NA, not applicable.
*Studies after 1992.

5 (range, 27). The internal validity (n10) scores showed a
median of 3 (range, 39) and 1 (range, 0 3), respectively.
Outcomes. Nine studies36,38,40-42,44-46,48 reported negative
findings; however, only 1 study36 was of high quality (scoring
9 points), with the others ranging from 0 to 3 points. Three
low-quality studies39,43,47 reported positive findings. One pilot
study37 had inconclusive findings, but showed a tendency to-
ward a positive result; the author suggested that this was
possibly because of the small sample numbers in the trial. The
overall outcome showed a trend toward traction being ineffec-
tive, however, because of the low quality of the various studies
research methods, no definitive conclusions are possible.
Placebo comparisons. Six studies compared traction with
sham traction (table 6). Sham traction is a low-weight or
placebo traction that the given researcher considers to be inef-
fective. Three studies used motorized traction (2 on a split
tabletop,36,41 1 on a plain tabletop44), 1 used autotraction,46 1
used gravitational traction,37 and 1 used traction as part of bed-
rest.45 Only the study by Beurskens et al36 was of a high
quality, and all gave negative results except for the inconclu-
sive result of Moret et al.37
Conservative treatment comparisons. Seven studies com-
pared traction with other conservative treatments (tables 6, 7);
3 studies used motorized traction with a split tabletop,40,42,43 3
used motorized traction without a split tabletop,38,39,40 1 used
manual traction,48 and 1 used autotraction.47
Traction was compared with no treatment, manipulation,
exercises, corset, interferential, transcutaneous electric nerve
stimulation (TENS), heat, and massage. All of these studies
had low methodologic quality scores (range, 0 3) and only 2
studies39,43 reported a positive effect of motorized traction
when compared with a treatment package that included hot
packs, massage, mobilization and strengthening exercises, and
TENS, respectively.
Motorized traction. Motorized traction is the only form of
traction that can be adequately standardized in respect to
weight applied. Seven trials used this type of traction (table 8).
Two trials36,41 compared motorized traction with sham traction,
and both gave negative results; however, only the Beurskens
study36 was of high quality. The other 5 studies38-40,42,43 com-
pared motorized traction with conservative treatment, but all
were of low quality (range, 0 3). Three studies38,40,42 reported
negative findings while 2 trials39,43 reported positive findings.
Classification of chronicity. One of the major problems
identified was the use of heterogeneous populations. In partic-
ular (where specified), the time from onset of symptoms to

Table 6: Lumbar Traction Compared With Placebo or Sham Traction

Control Group: Placebo Quality of Study

Study Intervention Group: Traction or Sham Traction Methodology Conclusion

Beurskens et al36 Motorized split tabletop 30% Motorized split tabletop 3% High Negative
50% body weight (n77) 27% body weight (n74)
Matthews and Hicklings44 Traction plain couch 3658.5kg Traction plain couch under Low Negative
(80130lb) (n13) 9kg (20lb) (n14)
Pal et al45 Traction as part of bedrest Traction as part of bedrest Low Negative
5.58.2kg (n24) 1.41.8kg (n15)
Weber et al46 (Spina-trac) Autotraction Autotraction frames blocked Low Negative
no traction force
Weber41 Motorized split tabletop Simulated traction force 7kPa Low Negative
13 body weight (n37) (n35)
Moret et al37 Vertical traction and bedrest Bedrest only Low Inconclusive

Arch Phys Med Rehabil Vol 84, October 2003

 LUMBAR TRACTION, Harte 1549

Table 7: Lumbar Traction Compared With Conservative Treatment

Intervention Group: Traction Quality of Study

Study as Main/Sole Modality Conservative Treatment Methodology Conclusion

Coxhead et al42 Motorized split tabletop alone or combined No treatment, manipulation, exercises, or Low Negative
with manipulation, exercises, or corset in corset (n166)
any combination of 14 modalities
(n166)
Lidstrom and Motorized split tabletop (n21) Groups 1: hot packs and rest (n21) Low Positive
Zachrisson43 Group 2: hot packs, massage, mobilization,
and strengthening exercises (n22)
Matthews et al40 Motorized split tabletop at least 45kg Infrared Low Negative
Werners et al38 Motorized massage Interferential Low Negative
Sherry et al39 Motorized traction VAX-D TENS Low Positive
Larsson et al47 Autotraction and corset Corset Low Negative
Ljunggren et al48 Manual traction Isometric exercise Low Negative

inclusion in the study was wide and variable (see table 4). Five
studies37,41,42,45,46 provided no clear indication of time from
onset; therefore, these articles could not be included in this
subsection of the review. Of the remaining 8 articles, 2 in-
cluded acute and subacute patients,40,47 both of which had
low-quality scores (1, 0, respectively); 1 gave a negative re-
sult40 and the other a positive result.47 Three studies36,39,48
included subacute and chronic patients, 2 of which were of low
quality39,48 and 1 was of high quality36; and 2 gave a negative
result36,48 while the other gave a positive result.39 One article38
included chronic patients, but was of low quality and gave a
negative result. The remaining 2 articles43,44 included the whole
spectrum of acute and chronic patients; both were of low
quality and 1 gave a positive result while the other reported a
negative result. Because of the diverse populations included in
these studies, it is difficult to draw any conclusions.
Strand 2: Clinical Criteria
When the criteria for the clinical parameters of traction were
applied to the 13 articles, only 4 low-quality studies (internal
validity score, 0 3) met the selection criteria39,40,43,44 (table 9).
Two studies40,44 reported negative findings and 2 studies39,43
reported positive findings.
The only high-quality study36 was excluded from this selec-
tion because its sham traction group had received a clinically
significant traction weight.
Four other studies37,41,47,48 failed to meet the criteria because
they only gave traction for 1 week. Two studies42,47 gave no
indication of the weights used. One study45 used a light weight
without a friction-free surface, and 1 trial38 used weights within
the clinical parameters, but it was unclear whether the bed was
friction free.
DISCUSSION
The aim of this review was to ascertain if the evidence for
the use of traction for persons with LBP had altered since the
original 1995 review. The evidence was rated on methodologic
quality but further inclusion criteria were selected based on
appropriate clinical parameters for the application of traction.
Only 1 study36 was of high methodologic quality, but it was
excluded when clinical parameters were applied.
Strand 1: Methodologic Quality
The use of methodologic quality scoring systems is still
subject to debate, and many different types are used.28 In their
review, van der Heijdan et al9 used a system that included a
16-point criteria list, subdivided into a 45-point checklist;
points in the latter checklist were weighted according to their
influence on validity and precision. However, weighting of a
system may not be any better, and only gives more points over
which disagreement may occur.12 The scoring system used in
the present review was recommended by the Cochrane Back

Table 8: Trials Using Motorized Traction

Intervention Group: Traction Quality of Study

Study as Main/Sole Modality Control Group Methodology Conclusion

Coxhead et al42 Motorized split tabletop alone or combined No treatment, manipulation, exercises, or Low Negative
with manipulation, exercises, or corset in corset (n166)
any combination of 14 modalities
(n166)
Lidstrom and Motorized split tabletop (n21) Groups 1: hot packs and rest (n21) Low Positive
Zachrisson43 Group 2: hot packs, massage, mobilization,
and strengthening exercises (n22)
Matthews et al40 Motorized split tabletop, at least 45kg Infrared Low Negative
Beurskens et al36 Motorized split tabletop 30%50% body Motorized split tabletop 3%27% body High Negative
weight (n77) weight (n74)
Weber41 Motorized split tabletop 13 body weight Simulated traction force 7kPa (n35) Low Negative
(n37)
Werners et al38 Motorized traction with massage Interferential Low Negative
Sherry et al39 Motorized traction VAX-D TENS Low Positive

Arch Phys Med Rehabil Vol 84, October 2003

1550 LUMBAR TRACTION, Harte

Table 9: Clinical Parameter Criteria for Strand 2 Inclusion

Friction-Free Couch or Traction Treatment: Acute Nerve Root Daily

Weight Over 26% Body Weight Length of Treatment Time: 57d for Minimum 2wk;
Plus Traction Force After Minimum 10min, Chronic LBP 23 Times Selected or
Study This of 585kg Average Time of 1530min Weekly for Minimum 3wk Not Selected

Lidstrom and Zachrisson43 Friction-free motorized; weights 20min (Y) 10 treatments over 1mo (Y) Selected
12 body weight (Y)

Weber41
Matthews and Hicklings44
Larsson et al47
Coxhead et al42
Weber et al46 (Spina-trac)
Pal et al45
Matthews et al40
Ljunggren et al48
Beurskens et al36
Moret et al37
Werners et al38
Sherry et al39
Friction-free motorized; 13 body 20min (Y) Once daily for 57d (N) Not selected
weight (Y)
Not friction-free motorized, but 30min (Y) Daily 5d/wk for 3wk (Y) Selected
weights high (Y)
Not friction free, autotraction; 1h (Y) 3 times in 1wk (N) Not selected
weight unknown (N)
Friction-free motorized, no No information (N) Daily for 1wk then 3 times Not selected
information on weights used weekly for another 3wk (Y)
(N)
Forces unclear; autotraction (N) 20min (Y) Once a day for 4d; if no change Not selected
stop treatment; if improving, 8
treatments (N)
Traction on bedrest, but not Continuous while on Continuous (Y) Not selected
friction free; light weights (N) bedrest (Y)
Friction-free motorized with 30min (Y) Daily maximum of 3wk (Y) Selected
minimum of 45kg (Y)
Not friction free, manual traction; 5min twice daily or 57d treatment period (N) Not selected
weight 300N (N) once for 10min (N)
Friction-free motorized; both 20min (Y) 12 times in 5wk (Y) Not selected
groups received traction
weights within the limits
stated (N)
Friction-free gravitational 30min (Y) 6 times/d for 12wk (N) Not selected
traction; 12 body weight (Y)
1020kg not sure if friction-free 10min (N) 6 sessions over 1421d (Y) Not selected
motorized traction (N)
No information on bed, but 30min (Y) 5 times/wk for 4wk then once a Selected
weights high enough to week for 4wk (Y)
overcome friction (Y)

Abbreviations: N, failed to meet criteria, Y, meets criteria. All 3 must be fulfilled to meet the inclusion criteria.

Review Group.2 It consists of a 19-point criteria checklist, with
a subclassification of 10 internal validity points that were used
to give the methodologic quality score. The editorial board of
the Cochrane Back Review Group recommended2 that a min-
imum list of internal validity criteria should be included in any
chosen system: treatment allocation, withdrawal and dropout
rate, blinded patients, outcome assessors blinded, and an inten-
tion-to-treat analysis should be used in the statistical analysis.
The criteria used in the present review went further and also
included blinded care provider, cointerventions avoided or
comparable, relevant outcome measures, and comparable tim-
ing of outcome measures between the groups. Beyond this, the
Cochrane Back Review Group suggests that 1 scoring system
should be used for all LBP research, so that comparison can be
clearly made between reviews.
Our results highlight the continuing problems with the meth-
odologic quality of LBP research. Only 1 high-quality study
was found36 and the remaining studies were of low quality, thus
limiting the conclusions available from this review. The 4
studies36-39 reported after 1992 were of a higher quality (scor-
ing 39 points) and had improved in areas such as description
of randomization, cointerventions, withdrawal and dropouts,
the use of relevant outcome measures, and an intention-to-treat
analysis. Areas that still need improvement are blinding of the
assessor (only 2 articles36,37 clearly stated they were blinded),
compliance, timing of outcome measures, and blinding of the
patients and the care provider. The latter 2 criteria offer a
particular challenge to research on physical interventions. They
are often difficult to achieve, and may highlight a problem
inherent in using this type of scoring system in therapeutic
trials. The pre-1992 articles showed a very low quality of
methodologic rigor (scoring 0 3 points, with only 1 article
scoring 3 points), reflecting the findings of the van der Heijdan
review.9 One main difference is notable, Matthews et al40
scored 52 points in the van der Heijdan review,9 placing it just
within the category of a high-quality trial. In the present
review, it was similarly placed on the ranking of total meth-
odologic scores, but on the internal validity score it scored only
1 point, which resulted in it being placed in the low-quality
category.
Study population. To achieve a homogenous group, the
patients must be clearly defined in terms of the nature and the
duration of the condition. Only 7 articles36,39,40,43,44,47,48 clearly
stated the type of back pain patients (radiation, leg pain) and
length of time from onset. If the group is of a heterogeneous
nature it may include those patients unlikely to respond to
traction, which dilutes the effects of the intervention. This
problem can be overcome to some extent by the randomization
of a large group of subjects, therefore preventing bias by
distributing known and unknown prognostic factors evenly
between the groups. However, only 1 trial36 adequately de-
scribed a random concealed allocation and many of the pre-

Arch Phys Med Rehabil Vol 84, October 2003

 LUMBAR TRACTION, Harte
1992 articles had small numbers. Sample size is also important
to precision in the estimation of effect: a small sample group
may present difficulties in detecting a statistically significant
effect (low statistical power).
Blinding. Blinding of the patient in this type of trial is
fraught with problems. Attempts have been made to blind
patients by using low-weight or sham traction as the placebo
treatment,36,41,44,45 but this strategy often resulted in low
weights being applied. Because the mechanism by which trac-
tion may have an effect on the lumbar spine is not fully
understood, it cannot be assumed that sham traction with low
weights will not have an effect.35 Krause et al53 proposed that
pain modulation through increasing non-nociceptive input
might be an effect of traction that would require very low
weights.
Blinding the patient and care provider are difficult to
achieve, but blinding the assessor should be included at all
times; only 5 studies clearly blinded the assessor.36,39,40,41,48
This factor must be improved to prevent biased results, partic-
ularly where the study cannot blind the care provider and/or the
patient.
Withdrawal and dropout rates. Withdrawal and dropout
rates should be recorded in a study and should not exceed 20%
in the short term or 30% in the long term,29 otherwise these
numbers may lead to substantial bias. Half (50%) of the articles
did not adequately describe the dropout rate or exceeded these
figures. An intention-to-treat analysis may statistically address
those patients who have dropped out, but it is also important to
understand why they did not complete treatment. Only 3 arti-
cles36,38,42 used an intention-to-treat analysis.
Cointerventions. Comparability of cointerventions must
be considered in a study; otherwise results may be inaccurate in
their conclusions. Only 3 studies36,37,39 described and con-
trolled these cointerventions adequately. The study by Werners
et al38 did not attempt to look at cointerventions, even though
they state in their discussion that it is possible that patients
were attending more than one clinic simultaneously.
Outcome measures. An outcome measure needs to be re-
liable, valid, and sensitive enough to measure small but clini-
cally relevant changes.53 Without this feature, the studys re-
sults can be inaccurate and misleading. Much of the early
research (pre-1992) used objective clinical outcome measures
such as range of movement, straight-leg raising, and muscle
power; however, many of these have not been shown to be
reliable or valid measures of effect.54 The emphasis is now
toward the patients perception of change, measured through
questionnaires relating to disability, quality of life, and the
patients ability to cope. The post-1992 studies showed the use
of the primary outcomes as recommended by the Cochrane
Back Review Group,2 and thus moved away from objective
clinical signs toward the currently recommended outcomes.
Subclassification of studies. A secondary aim of the
present review was to look at the relevance of the different
modes of traction and the effect of traction for acute, subacute,
and chronic LBP. In particular, we were interested in looking
at the studies using motorized traction, because it is the only
type of traction that can be adequately standardized and is
commonly used in clinical practice. Because of the consistently
low quality of the trials, no further conclusions could be drawn
about the effects of the mode of traction. Likewise, on consid-
ering the poor specification of eligibility criteria, we could
draw no firm conclusions about the effectiveness of traction for
acute and subacute LBP. The only high-quality article36 con-
cluded that traction is ineffective for subacute and chronic
LBP, but serious problems with the clinical parameters they
used render this conclusion unacceptable.
1551
Overall conclusion based on methodologic quality. On
applying the Agency for Health Care Policy and Research
(AHCPR) rating30 to these studies, we found moderate evi-
dence that traction is ineffective in the treatment of nonspecific
LBP; that is, 1 high-quality study and several low-level studies.
Strand 2: Clinical Parameter Criteria
Because inappropriate treatment procedures or inadequate
treatment doses may lead to serious performance bias in a
trial,12 trials must be examined for these criteria as well as for
methodologic quality. A trial may be of a high methodologic
quality but if assessment reveals that its treatment procedures
are of poor quality, that weakness will affect the strength of the
overall conclusion. The study by Beurskens et al36 was scored
as a high-quality study, but, on applying the clinical parameter
criteria to it, we found it to be lacking. This study used
low-weight or sham traction as the control group and weights
ranged from 3% to 20% of body weight; however, the traction
weights recommended clinically ranged from 10 to 85kg.
Therefore, it is likely that some of the control group may have
received traction weights at clinically relevant levels; conse-
quently, this discrepancy excluded the only high-quality study
from this section of the review. Other trials were excluded for
applying traction for a time span that was unlikely to affect
change (1wk),37,41,46-48 for not giving any indication of the
weight applied,42,46,47 or using light weights on a nonfriction-
free surface.38,45 One article failed to meet any of the criteria,48
4 articles38,42,46,47 failed to meet 2 of the 3 criteria, and 4
articles36,37,41,45 failed to meet 1 of the criteria. This finding
shows the need for an emphasis to be placed on the clinical
application of the interventions used in a study.
Overall conclusion of strand 2. On applying the AHCPR
rating,30 we found limited evidence that traction is ineffective
for the managing of nonspecific LBP: findings in multiple
RCTs were conflicting. This section of the review highlights
the importance of addressing both the methodologic quality
and the clinical appropriateness of the intervention.
Physical therapy is characterized by diverse combinations of
treatments, which include physical agents (heat, cold, electro-
therapy) accompanied by exercise and manual therapy. During
a course of treatment, there is often a move from passive
treatments (heat, cold, electrotherapy, manual therapy) toward
a more active form of therapy in the later stages.6 In the present
review, the studies typically investigated the effects of traction
given as the sole treatment (in a limited number of cases, back
care advice and analgesics were given). This approach does not
reflect how therapists manage LBP. This factor must be ad-
dressed in the development of future studies. A need obviously
exists for a high-quality pragmatic RCT design in this area of
physiotherapy. A pragmatic trial evaluates a treatment policy
rather than the treatment itself; it is not concerned with how the
treatment works but whether it works in clinical practice (ef-
fectiveness), and it allows for variations between patients that
occur in real clinical situations.55 Research in physical therapy
must move to a new height and away from the traditional RCT,
but in doing so it must strive to produce a study of high
methodologic quality.
Limitations of This Review
A primary limitation of the present review was that articles
were limited to the English language. Research on traction has
traditionally taken place in the Netherlands and Germany and
therefore we may have excluded some suitable studies.
Arch Phys Med Rehabil Vol 84, October 2003
1552 LUMBAR TRACTION, Harte
CONCLUSION
This review showed that the evidence (on methodologic
quality) for the effectiveness of traction in the management of
LBP has not changed despite the addition of 4 new studies
since the last review in 1995. The studies reviewed were
generally of poor quality, and it was not possible to make
recommendations for the use of traction based on the results of
these trials. When the appropriateness of the clinical interven-
tion was addressed, the level of evidence was reduced from
moderate to limited evidence. This highlights the importance of
paying close attention to the clinical appropriateness of the
intervention in addition to methodologic quality, and may, if
used in future research, lead to wider implementation of re-
search findings.
Recommendations for Further Study
The study by Beurskens et al36 was the only high-quality
article reviewed and it addresses the subacute to chronic LBP
patient population. This study shows that a high-quality study
in this area can be achieved,56 but the flaws we describe in this
article need to be overcome (traction parameters, traction as the
sole intervention). One group of patients that has not been
addressed by a high-quality study are the patients with acute
radiculopathy. Many researchers have attempted to study this
area, but their studies have failed because of the use of poor
inclusion criteria and other methodologic problems.
Classification of LBP is a problem, but radicular pain is
perhaps easier to classify than many other, less distinct groups.
In designing a study, one must give attention not only to
methodologic quality but also to the clinical appropriateness of
the intervention. To reflect clinical practice, we need to under-
stand the clinical parameters and treatment regimes being used
by clinicians, and these must be addressed before a trial can be
designed to look further at the effectiveness of traction for LBP
with or without radiating pain.
References
1. Hides JA, Richardson CA, Jull GA. Multifidus muscle recovery is
not automatic after resolution of acute, first episodes low back
pain. Spine 1996;21:2763-9.
2. van Tulder MW, Assendelft WJ, Koes BW, Bouter LM. Method
guidelines for systematic reviews in the Cochrane Collaboration
Back Review Group for Spinal Disorders. Spine 1997;22:2323-30.
3. Foster NE, Thompson KA, Baxter GD, Allen JM. Management of
nonspecific low back pain by physiotherapists in Britain and
Ireland. Spine 1999;24:1332-42.
4. Gracey J. Physiotherapy management of low back pain. A survey
of current practice in Northern Ireland [PhD dissertation]. Jordans-
town, Northern Ireland: Univ Ulster; 2001.
5. van der Heijdan GJ, Beurskens AJ, Dirx MJ, Bouter LM, Linde-
man E. Efficacy of lumbar traction: a randomised clinical trial.
Physiotherapy 1995;81:29-35.
6. Jette AM, Delitto A. Physical therapy choices for musculoskeletal
improvements. Phys Ther 1997;77:145-54.
7. Li LC, Bombardier C. Physical therapy management of low back
pain: an exploratory survey of therapist approaches. Phys Ther
2001;81:1018-27.
8. Waddell G, McIntosh A, Hutchinson A, Feder G, Lewis M. Low
back pain evidence review. London: Royal Coll General Practi-
tioners; 1999.
9. van der Heijdan GJ, Beurskens AJ, Koes BW, Assendelft WJ, de
Vet HC, Bouter LM. The efficacy of traction for back and neck
pain: a systematic, blinded review of randomised clinical trial
methods. Phys Ther 1995;75:93-104.
10. The Database of Abstracts of Reviews of Effectiveness. York
(UK): Cochrane Library; 1997. CRD Database No. DARE-
978020.
Arch Phys Med Rehabil Vol 84, October 2003
11. Philadelphia Panel evidence-based clinical practice guidelines on
selected rehabilitation interventions for low back pain [see com-
ments]. Phys Ther 2001;81:1641-74.
12. Bjordal JM, Greve G. What may alter the conclusion of reviews?
Phys Ther Rev 1998;3:121-32.
13. Maitland GD. Vertebral manipulation. 8th ed. Edinburgh:
Churchill Livingstone; 2001.
14. Cyriax J. Textbook of orthopaedic medicine. Vol 1, 8th ed. Lon-
don: Bailliere Tindall; 1982.
15. Grieve GP. Common vertebral joint problems. Edinburgh:
Churchill Livingstone; 1981.
16. Saunders HD. Use of spinal traction in the treatment of neck and
back conditions. Clin Orthop 1983;Oct(179):31-7.
17. Colachis SC, Strohm BR. Effects of traction on separation of
lumbar vertebrae. Arch Phys Med Rehabil 1969;44:251-8.
18. Gillstrom P, Ericson K, Hindmarsh T. Autotraction in lumbar disc
herniation. A myelographic study before and after treatment. Arch
Orthop Trauma Surg 1985;104:207-10.
19. Gupta RC, Ramarao SV. Epidurography in reduction of lumbar
disc prolapse by traction. Arch Phys Med Rehabil 1978;59:322-7.
20. Hood CJ, Hart D, Smith HG, Davis H. Comparison of electro-
myographic activity in normal lumbar sacrospinalis musculature
during continuous and intermittent pelvic traction. J Orthop Sports
Phys Ther 1981;2:137-41.
21. Judovich B, Nobel GR. Traction therapy, a study of resistance
forces. Am J Surg 1957;93:108-14.
22. Lee RY, Evans JH. Loads in the lumbar spine during traction
therapy. Aust J Physiother 2001;47:102-8.
23. Onel D, Tuzlaci M, Sari H, Demir K. 1989 Computed tomo-
graphic investigation of the effect of traction on lumbar disc
herniations. Spine 1989;14:82-90.
24. Twomey LT. Sustained lumbar traction. An experimental study of
long spine segments. Spine 1985;10:146-9.
25. Judovich BD. Lumbar traction therapy: elimination of physical
factors that prevent lumbar stretch. JAMA 1955;159:549-50.
26. Hicklings J. Spinal traction technique. Physiotherapy 1972;58:58-
63.
27. Jadad AR, Moore A, Carroll D, et al. Assessing the quality of
reports of randomised clinical trials: is blinding necessary? Con-
trol Clin Trials 1996;17:1-12.
28. Verhagen AP, de Brie RA, Lenssen AF, et al. Quality assessment
of trials: a comparison of three criteria lists. Phys Ther Rev
2000;5:49-58.
29. van Tulder MW, Koes B, Bouter LM. Conservative treatment of
acute and chronic nonspecific low back pain. A systematic review
of randomized controlled trials of the most common interventions.
Spine 2000;22:2128-5.
30. Bigos S, Bowyer O, Braen G, et al. Acute low back problems in
adults. Clinical practice guideline no. 14. Rockville: Agency for
Health Care Policy and Research, Public Health Service, US
Department of Health and Human Services; 1994. AHCPR Pub-
lication No. 95-0642.
31. Tesio L, Merio A. Autotraction versus passive traction: an open
controlled study in lumbar disc herniation. Arch Phys Med Reha-
bil 1993;74:871-6.
32. Guvenol K, Tuzun C, Peken O, Goktay Y. A comparison of
inverted spinal traction and conventional traction in the treatment
of lumbar disc herniation. Physiother Theory Pract 2000;16:151-
60.
33. Ljunggren EA, Weber H, Larsen S. Autotraction versus manual
traction in patients with prolapsed lumbar intervertebral discs.
Scand J Rehabil Med 1984;16:117-24.
34. Sweetman BJ, Heinrich I, Anderson JA. A randomised controlled
trial of exercises, shortwave diathermy, and traction for low back
pain, with evidence of diagnosis-related response to treatment.
J Orthop Rheumatol 1993;6:159-66.
35. Beurskens AJ, de Vet HC, Koke AJ, et al. Efficacy of traction for
non-specific low back pain: a randomised clinical trial. Lancet
1995;346:1596-600.
36. Beurskens AJ, de Vet HC, Koke AJ, et al. The efficacy of traction
for non-specific low back pain 12 week and 6 month results of a
randomized clinical trial. Spine 1997;22:2756-62.
 LUMBAR TRACTION, Harte
37. Moret NC, van der Strap M, Hagmeijer R, Molenaar A, Koes BW.
Design and feasibility of a randomised clinical trial to evaluate the
effect of vertical traction in patients with a lumbar radicular
syndrome. Man Ther 1998;3:203-11.
38. Werners R, Pynsent PB, Bulstrode CJ. 1999 Randomized trial
comparing interferential therapy with motorised lumbar traction
and massage in the management of low back pain in a primary
care setting. Spine 1999;24:1579-84.
39. Sherry E, Kitchener P, Smart R. A prospective randomized con-
trolled study of VAX-D and TENS for the treatment of chronic
low back pain. Neurol Res 2001;23:780-4.
40. Matthews JA, Mills SB, Jenkins VM, et al. Back pain and sciatica:
controlled trials of manipulation, traction, sclerosant and epidural
injections. Br J Rheumatol 1987;26:416-23.
41. Weber H. Traction therapy in sciatica due to disc prolapse (does
traction treatment have any positive effect on patients suffering
from sciatica caused by disc prolapse?). J Oslo City Hosp 1973;
23:161-76.
42. Coxhead CE, Inskip H, Meade TW, North WR, Troup JD. Mul-
ticentre trial of physiotherapy in the management of sciatic symp-
toms. Lancet 1981;1:1065-68.
43. Lidstrom A, Zachrisson M. Physical therapy on low back pain and
sciatica. Scand J Rehabil Med 1979;2:37-42.
44. Matthews JA, Hicklings J. Lumbar traction: a double blind con-
trolled study for sciatica. Rheumatol Rehabil 1975;14:222-5.
45. Pal P, Mangion P, Hossian MA, Diffey LA. Controlled trial of
continuous lumbar traction in the treatment of back pain and
sciatica. Br J Rheumatol 1986;25:181-3.
46. Weber H, Ljunggren AE, Walker L. Traction therapy in patients
with herniated lumbar intervertebral discs. J Oslo City Hosp
1984;34:61-70.
1553
47. Larsson U, Choler U, Lidstrom A, et al. Autotraction for treatment
of lumbago-sciatica. Acta Orthop Scand 1980;51:791-8.
48. Ljunggren E, Walker L, Weber H, Amundsen T. Manual traction
versus isometric exercises in patients with herniated intervertebral
lumbar discs. Physiother Theory Pract 1992;8:207-13.
49. Walker L, Svenkerud T, Weber H. Traksjonsbehandling ved lum-
bago-ischias: en kontrollert undersolske med Spina-trac. Fysiot-
erapeuten 1982;49:161-3.
50. Bilhaug O. Autotraksjon for ischialgpasienter: en kontrollert sam-
menlikning mellom effekten av Auto-traksjon-B og isometriske
ovelser ad modum Hume endall og enkins. Fysioterapeuten 1978;
48:377-9.
51. Reust P, Chantraine A, Vischer TL. [Treatment of lumbar sciatica
with or without neurological deficit using mechanical traction. A
double-blind study] [French]. Schweiz Med Wochenschr 1988;
118:271-4.
52. van der Heijdan GJ, Bouter LM, Terpstra-Lindeman E. De effec-
tiviteit van tractie bij lage rugklachten: de resultaten van een
pilotstudy. Ned T Fysiother 1991;101:37-43.
53. Krause M, Refshauge KM, Dessen M, Boland R. Lumbar spine
traction: evaluation of effects and recommended application for
treatment. Man Ther 2000;5:72-81.
54. Foster NE. Researching low back pain? An overview of the
pitfalls. Phys Ther Rev 1998;3:9-17.
55. Wakefield A. Evidence-based physiotherapy: the case for prag-
matic randomised controlled trials. Physiotherapy 2000;86:
394-6.
56. Koes B. Commentary: high force traction did not improve the
clinical course of sub-acute and chronic non-specific low back
pain. Aust Physiother 1998;44:213-4.
Arch Phys Med Rehabil Vol 84, October 2003