Acupuntura e Cefaleia
Click here to load reader
Transcript of Acupuntura e Cefaleia
8/12/2019 Acupuntura e Cefaleia
http://slidepdf.com/reader/full/acupuntura-e-cefaleia 1/9
Journal of Clinical Epidemiology 55 (2002) 77–85
0895-4356/02/$ – see front matter © 2002 Elsevier Science Inc. All rights reserved.
PII: S0895- 4356(01)0 0422-X
Should systematic reviews include non-randomizedand uncontrolled studies?
The case of acupuncture for chronic headache
Klaus Lindea,c,*, Michael Scholzb, Dieter Melchartc
, Stefan N. Willich
a
a
Institute for Social Medicine & Epidemiology, Charité Hospital, Humboldt-Universität, 10098 Berlin, Germany
b
Institute for Medical Statistics & Epidemiology, Technische Universität, Ismaningerstr. 22, 81675 Munich, Germany
c
Centre for Complementary Medicine Research, Department of Internal Medicine II, Technische Universität, Kaiserstr. 9, 80801 Munich, Germany
Received 1 December 2000; received in revised form 28 June 2001; accepted 6 July 2001
Abstract
We aimed to investigate: (1) whether patient and intervention characteristics, design-independent quality aspects, and response rates differ be-
tween randomized and non-randomized trials of acupuncture for chronic headache; (2) whether non-randomized studies provide useful additional in-
formation (regarding long-term effects, prognostic factors, adverse effects, and generalizability); (3) reasons for potential differences in response
rates. Studies including at least five patients and reporting clinical outcome data were identified through searches in Medline, Embase, the Cochrane
Controlled Trials Register, other databases and checking of bibliographies. Twenty-four randomized trials and 35 non-randomized studies (five non-
randomized controlled cohort studies, 10 prospective uncontrolled studies, 10 case series, and 10 cross-sectional surveys) met the inclusion criteria.
Studies were heterogeneous regarding patients, interventions, outcome measurements and results. On average, randomized trials had smaller sample
sizes, met more quality criteria, and had lower response rates (0.59 [95% confidence interval 0.48–0.69] vs. 0.78 [0.72–0.83]). Whether randomized
or not, studies meeting more quality criteria had lower response rates. Non-randomized studies did not have significantly longer follow-up periods,
three included an analysis of prognostic variables, only one reported on adverse effects, and the degree of generalizability was unclear. In the case of
acupuncture for chronic headache, non-randomized studies confirmed the finding of a systematic review of randomized trials that the treatment is
likely to be effective but provided little relevant additional information on long-term effects, prognostic factors, and adverse effects. © 2002
Elsevier Science Inc. All rights reserved.
Keywords: Meta-analysis; Acupuncture; Headache; Migraine; Randomized controlled trial
1. Introduction
Systematic reviews play an increasingly important role
in assessing the existing evidence for health care interven-
tions. Most systematic reviews of prophylactic and thera-
peutic interventions include only randomized controlled tri-
als (RCTs) which are generally considered as the gold
standard for evaluation. However, the medical literature isfull of non-randomized and uncontrolled studies. Should the
results of these studies be dismissed completely? Or should
they be considered as useful additional information for sys-
tematic reviews?
It is obvious that it is highly problematic to draw conclu-
sions on causal effects from non-randomized or even un-
controlled studies. However, in other aspects such studies
can—if planned and performed well—provide valid data.
For example, one problem of RCTs is that they often recruit
only a small and non-representative sample of patients who
are then treated and monitored under conditions which do
not necessarily reflect everyday practice [1,2]. As a conse-
quence, the degree to which the results are generalizable
and externally valid is unclear. Therefore, it would be desir-
able to compare patient characteristics, treatments and
course of disease in RCTs and pragmatic non-randomized
studies. Another problem of RCTs (particularly placebo-
controlled) in chronic conditions is the often limited obser-
vation period due to ethical reasons, feasibility or resources
[3,4]. Cohort studies in which patients receive their pre-
ferred treatment can follow patients more easily over longer
periods. Large-scale non-randomized trials could also be
useful to identify potentially relevant prognostic factors (as
well as factors determining why certain people choose cer-
tain treatments) and to evaluate safety.
Furthermore, if the number or quality of available ran-
domized trials is insufficient to draw conclusions, non-ran-
* Corresponding author. Tel.: 089-33041048; fax: 089-393484.
E-mail address
: [email protected] (K. Linde).
8/12/2019 Acupuntura e Cefaleia
http://slidepdf.com/reader/full/acupuntura-e-cefaleia 2/9
78
K. Linde et al. / Journal of Clinical Epidemiology 55 (2002) 77–85
domized studies might be useful to get a better overview of
what is known so far and to inform future research.
This last point might be particularly relevant in system-
atic reviews of complementary therapies. For example,
there are two systematic reviews (one on acupuncture for
induction of labor [5] and one for cranberries to treat uri-
nary tract infections [6]) limited to RCTs in the Cochrane
Database of Systematic Reviews which did not identify a
single trial meeting the inclusion criteria. Many other re-
views restricted to RCTs contain very few trials. Would
these reviews be more informative if they included non-ran-
domized studies? There is a need for empirical studies in-
vestigating under what circumstances the inclusion of non-
randomized trials is adequate and which problems have to
be taken into account.
Parallel to a “conventional” systematic review of the
RCTs of acupuncture for chronic headache [7] we per-
formed a more inclusive analysis to investigate the follow-
ing questions: (1) Do randomized and non-randomized
studies of acupuncture for chronic headache differ in regardto patients, interventions, design-independent quality as-
pects, and response rates? (2) Do non-randomized studies
provide relevant additional information (results on long-
term outcomes, prognostic factors, adverse effects or com-
plications, response rates in representative or well-defined
groups of patients)? (3) In case response rates in random-
ized and non-randomized patients differ, what are possible
explanations?
2. Methods
2.1. Selection criteria
We included reports providing clinical outcome data in
at least five patients with recurrent or chronic headaches
(migraine, tension headache, cluster headache, not precisely
classified chronic or recurrent headaches) who had been
treated with acupuncture (needle insertion at acupuncture
points, pain points, trigger points or other methods of stimu-
lating acupuncture points such as laser acupuncture or elec-
troacupuncture) for prophylaxis of their headaches. We ex-
cluded studies focussing on acute headaches only or solely
reporting physiological outcomes. Studies only available in
Chinese language were not included due to the lack of re-
sources for translation.
2.2. Search strategy
To identify published and unpublished eligible studies we
searched: (1) Medline 1966 to December 1998 (search terms:
acupuncture/acupressure/electroacupuncture and headache/
migraine); (2) Embase 1989 to December 1998 (acupunc-
ture/acupressure and headache/migraine); (3) the Cochrane
Controlled Trials Register (Issue 1, 1999); (4) individual
trial collections and private databases; and (5) bibliogra-
phies of articles obtained. All references identified by the
literature search, and if available their abstracts, were screened
by two reviewers. Full reports of potentially relevant publi-
cations were obtained and checked for eligibility.
2.3. Classification
Included studies were classified into five categories: (1)
Randomized controlled trials; (2) non-randomized con-
trolled cohort studies; (3) uncontrolled prospective studies
(studies apparently based on a preplanned study protocol
and meeting at least one of the following criteria: explicit
in- and exclusion criteria, at least two predefined clinical
outcome measures, predefined treatment intervention); (4)
case series (uncontrolled studies without evidence of a pre-
defined protocol); (5) cross-sectional studies/surveys. The
classification of non-randomized studies was done twice
with a time lag of 5 months by the first author.
2.4. Data extraction
We extracted source details (bibliographic information,
language, country of the first author, whether published in aMedline-listed journal), information on patients (number,
diagnosis, in- and exclusion criteria, use of headache classi-
fication guidelines, setting, mean age, percentage females,
years since onset of disease, resistance to previous treat-
ment), interventions (strategy to choose acupuncture points,
number and duration of sessions), length of follow-up, and
data on response to treatment (definition of response, num-
ber of patients classified as responders after treatment, at
early follow-up, and follow-up of at least 6 months after
treatment; number of patients treated with acupuncture,
time of the earliest available response measurement). Re-
sponse rates for meta-analysis were calculated as number of
responders divided by the number of patients analyzed forthe earliest available time point after treatment. Data from
groups without acupuncture (control groups) were not ana-
lyzed. To assess whether non-randomized studies add rele-
vant information we checked the number of studies: (1)
with at least 6 months follow-up after the last acupuncture
session; (2) which provided outcome data for at least 80%
of treated patients after at least 6 months follow-up; (3)
which systematically investigated prognostic factors; (4)
with data on side effects or complications; (5) with at least
three study centers and a transparent sampling strategy. Se-
lection, extraction and assessments were performed by at
least two reviewers for randomized and non-randomizedcontrolled trials, and by one reviewer for the other studies.
Disagreements were resolved by discussion.
In addition, using a list of 16 items proposed by White
and Ernst [8] we investigated how detailed the acupuncture
treatment was described (results presented in [9]).
2.5. Assessment of quality aspects relevant for headache
The aim of the quality rating was to assess the accuracy
of observation and reporting, as far as possible indepen-
dently from the specific design chosen. Whether random-
ized or not a study with good quality in this respect should
8/12/2019 Acupuntura e Cefaleia
http://slidepdf.com/reader/full/acupuntura-e-cefaleia 3/9
K. Linde et al. / Journal of Clinical Epidemiology 55 (2002) 77–85
79
allow a health care professional treating headache patients
to have an idea of the patients in the study and the course of
symptoms. The criteria were partly derived from the guide-
lines for clinical trials in migraine and tension-type head-
ache from the International Headache Society [10,11]. The
following 10 items could be answered yes, no or unclear:
(1) Sampling method clear and clearly described? (2) Clear
headache diagnosis? (3) Patients characterized (at least age,
sex, duration, severity of symptoms)? (4) At least 4 weeks
baseline period? (5) At least two clinical headache out-
comes? (6) Use of a headache diary? (7) Cointerventions
described? (8) At least 90% of patients included analyzed
after treatment? (9) At least 80% of patients treated ana-
lyzed at early (
6 months) follow-up? (10) At least 80% of
patients treated analyzed at late (
6 months) follow-up? In
case of cross-sectional studies items 9 and 10 were scored
when the survey took place at least 6 months after acupunc-
ture treatment and the response rate was greater than 80%.
2.6. Analyses
Characteristics and fulfillment of the quality criteria
among RCTs and other studies were compared with the chi-
square test and the Mann-Whitney test (no adjustment for
multiple comparisons).
Responder proportions were estimated by a generalized
linear mixed model (random effects model) as implemented
in the SAS® Macro GLIMMIX [12]. In particular, we used
a linear logistic regression model with an additional random
intercept term to reflect the heterogeneity between the stud-
ies. GLIMMIX uses restricted pseudo-likelihood (REPL) to
find the parameter estimates of the linear mixed logistic
model we specified. The approach allows for trial-specificand treatment-averaged inference. This estimation proce-
dure was utilized for univariate as well as for multivariate
analysis.
3. Results
From a total of 97 acupuncture studies on headache iden-
tified, 59 met the inclusion criteria. The most frequent rea-
son for exclusion was the mixing of data of a variety of
chronic pain patients without separate reporting of results in
those with chronic headache. Twenty-four of the 59 included
studies were RCTs and 35 studies were non-randomized (5
non-randomized controlled cohort studies, 10 prospective
uncontrolled studies, 10 case series, and 10 cross-sectional
studies). Table 1 provides an overview of basic features of
the included studies which were heterogeneous regarding
patients, interventions, type and timing of outcome mea-
surements.
The median number of patients treated with acupuncture
was significantly smaller in RCTs than in other studies (see
Table 2). A total of 535 patients received acupuncture treat-
ment in the 24 RCTs compared to 2695 in the 35 non-ran-
domized studies. RCTs were performed more often in mi-
graine patients and slightly more often in specialized
outpatient pain departments. Age, sex distribution and the
duration of headache complaints were similar. Only 14
studies (seven RCTs and seven other studies) made state-
ments whether patients had been resistant to previous ther-
apy; in four RCTs and four other studies this had been the
case. Acupuncture sessions lasted significantly longer in the
non-randomized studies while the number of treatment ses-
sions and the proportion of studies using an individualized
treatment strategy was similar.
RCTs met criteria for accuracy of observation and re-
porting much more often than non-randomized studies but
in both groups the number of criteria met covered the full
range indicating that there were very low and good quality
studies in both groups (see Table 3). Ten of the 24 RCTs
and 26 of the 35 non-randomized studies met less than five
quality criteria. RCTs significantly more often described a
clear headache diagnosis, a pre-treatment baseline observa-
tion period, at least two clinically relevant outcome mea-
sures, and the use of headache diaries. Sampling strategies
and cointerventions (such as treatment for headache attacks)were rarely well described regardless of the type of study.
Follow-up was either short or incomplete in most studies as
well. A total of 50 studies presented some sort of responder
data for at least one of the three predefined time-points.
RCTs more often used a clearly defined response criterion.
The pooled response rate was 59% in RCTs [95% confi-
dence interval (CI) 48–69%) and 78% in non-randomized
studies (72–83%; P
0.001).
Ten (42%) RCTs and 14 (40%) non-randomized studies
measured outcome data at least 6 months after completion
of the treatment (see Table 4. However, only two (8%)
RCTs and six non-randomized (17%) studies actually pro-vided data for more than 80% of treated patients. In the two
RCTs and in two non-randomized studies reliable data on
frequency of headache was available from patient diaries.
All these studies suggest that improvements achieved after
treatment last in the majority of patients. The role of prog-
nostic factors was investigated in two (8%) RCTs and three
(9%) non-randomized studies. Due to the small sample sizes
and the exploratory character of the analyses the results can
at best be considered hypothesis generating. Only nine non-
randomized studies (and eight RCTs) gave information
(number of centers, selection process) on how they recruited
their patients. In all these studies acupuncture was provided
within a single practice or unit. Data on adverse effects
(number of patients reporting adverse effects and/or explicit
description of the events observed) was reported in four
(17%) RCTs and one (3%) non-randomized study.
In the univariate analysis diagnosis, sex distribution,
mean age, number and duration of treatment sessions, type
and timing of response measurement as well as complete-
ness of follow-up did not have a significant impact on the
response rates observed (see Table 5). Instead, studies de-
scribing their sampling strategy, a clear headache diagnosis,
having a baseline period and using a headache diary had
significantly lower response rates. Fig. 1 shows that both a
8/12/2019 Acupuntura e Cefaleia
http://slidepdf.com/reader/full/acupuntura-e-cefaleia 4/9
80
K. Linde et al. / Journal of Clinical Epidemiology 55 (2002) 77–85
Table 1
Overview of studies of acupuncture treatment for recurrent headaches
First author (ref) Type of headache Quali ty criteria met Response definition
n
treated with acupuncture
Response rates (%)
After treatment
Follow-up
Early
6 mo.
RCTs
Ahonen [25] Tension 4 Improved at least 2 mo* 12 n.a. 58 n.a.
Baust [26] Migraine 4
50% index reduction 23 61 n.a. n.a.Carlsson [27] Tension 5 — 30 n.a. n.a. n.a.
Ceccherelli [28] Migraine 5
50% index reduction 15 87 n.a. n.a.
Doerr-Proske [29] Migraine 7 — 10 n.a. n.a. n.a.
Dowson [30] Migraine 8
50% frequency reduct. 25 32 n.a. n.a.
Hansen [31] Tension 2
33% index reduction 25 28 n.a. n.a.
Henry [32] Migraine 5 Patient assessment 20 65 n.a. n.a.
Hesse [33] Migraine 8 — 42 n.a. n.a. n.a.
Heydenreich [34] Migraine 3 Unclear 20 n.a. n.a. 80
Heydenreich [35] Migraine 3 Unclear 50 n.a. n.a. 80
Johansson [36] Tension 0 — 17 n.a. n.a. n.a.
Kubiena [37] Migraine 4
33% index reduction 15 75 n.a. n.a.
Lehmann [38] Migraine 3 — 21 n.a. n.a. n.a.
Lenhard [39] Migraine 7
33% frequency reduct. 16 n.a. 56 n.a.
Loh [40] Various 3 Patient assessment 23 35 n.a. n.a.
Okazaki [41] Migraine 1 Unclear 20 75 n.a. n.a.
Pintov [42] Migraine 7 — 11 n.a. n.a. n.a.
Tavola [43] Tension 10
50% index reduction 15 53 n.a. 40Vincent [44] Migraine 9
33% frequency reduct. 16 47 n.a. n.a.
Vincent [45] Tension 9
50% index reduction 14 64 43 n.a.
Weinschütz [46] Migraine 5 Single case statistics 20 50 65 n.a.
Weinschütz [47] Migraine 5 Single case statistics 20 45 75 n.a.
White [48] Tension 5 — 5 n.a. n.a. n.a.
Non-randomized controlled cohort studies
Borglum-J. [49] Unclear 8 Unclear 32 79 n.a. n.a.
Guccione [50] Migraine 1 — 24 n.a. n.a. n.a.
Sold-Darseff [51] Migraine 1 Unclear 30 93 n.a. n.a.
Turk [52] Migraine 1 Unclear 100 86 n.a. n.a.
Zibin [53] Migraine 0 Unclear 340 99 n.a. n.a.
Uncontrolled prospective studies
Baischer [54] Migraine 10
33% index reduction 31 70 n.a. 58
Boivie [55] Migraine 10 Unclear 30 n.a. 56 56
Börglum-J. [56] Tension 6 Unclear 21 67 57 n.a.
Börglum-J. [57] Tension 5 Unclear 16 63 n.a. n.a.
Ceccherelli [58] Various 4
50% index reduction 12 100 n.a. n.a.Ceccherelli [59] Various 6
40% index reduction 92 76 n.a. n.a.
Kim [60] Migraine 3 Meeting defined criteria* 25 n.a. 80 n.a.
Major [61] Migraine 8
50% frequency reduct. 33 57 n.a. n.a.
Tolksdorf [62] Various 5
50% index reduction 70 59 n.a. 38
Zwölfer [63] Tension 3 Patient assessment 20 80 n.a. n.a.
Case series
Batra [64] Migraine 3 At least 50% improvem.* 20 n.a. 65 n.a.
Cheng [65] Various 2 No headache at all 33 91 n.a. n.a.
Fischer [66] Unclear 2
50% symptom reduct.* 189 75 n.a. n.a.
Heydenreich [67] Migraine 0 At least 50% improvem.* 320 n.a. n.a. 88
Heydenreich [68] Migraine 3 At least 50% improvem.* 296 n.a. n.a. 76
Laitinen [69] Migraine 5 Meeting defined criteria* 39 92 n.a. 46
Lawrence [70] Various 1 Alleviation symptoms* 96 92 n.a. n.a.
Spoerel [71] Various 2 Severity & freq. reduct.* 38 n.a. 84 n.a.
Stux [72] Migraine 0 Unclear 50 88 n.a. n.a.
Yamauchi [73] Various 2 Unclear 7 71 n.a. n.a.
SurveysCarlsson [74] Various 4 Patient assessment 79 57 n.a. 23
Chen [75] Unclear 1 Patient assessment 44 n.a. n.a. 68
Haug [76] Unclear 1 Patient assessment 49 n.a. n.a. 84
Junilla [77] Unclear 2
30% symptom reduct.* 115 n.a. n.a. 77
Junilla [78] Various 2 Patient assessment 69 81 n.a. n.a.
Kubiena [79] Migraine 1 Patient assessment 118 n.a. 62 n.a.
Maxion [80] Migraine 1 Patient assessment 13 15 n.a. n.a.
McKibbin [81] Various 4 — 107 n.a. n.a. n.a.
Pöntinen [82] Migraine 2 Patient assessment 107 n.a. n.a. 57
Shifman [83] Various 0 Patient assessment 33 58 n.a. n.a.
n.a.
not available; *operationalization unclear.
Patient assessments performed with various categorizations (without use of headache diaries).
8/12/2019 Acupuntura e Cefaleia
http://slidepdf.com/reader/full/acupuntura-e-cefaleia 5/9
K. Linde et al. / Journal of Clinical Epidemiology 55 (2002) 77–85
81
randomized trial design and a higher number of quality cri-
teria met were associated with lower response rates. We
were unable to specify an adequate model for multiple re-
gression analysis. Apart from design (randomized or not)
other factors had little consistent effects.
4. Discussion
Our results show that in the case of acupuncture for
chronic headache: (1) RCTs and non-randomized studies
differ in regard to patients included, interventions provided,
and response rates reported; (2) RCTs have in average bet-
ter accuracy of observation and reporting, but a relevantproportion of RCTs has very low and some prospective un-
controlled trials have acceptable accuracy of observation
and reporting; (3) the available non-randomized studies pro-
vide only limited useful additional information in regard to
long-term response and prognostic factors, and no addi-
tional information regarding adverse effects and generaliz-
ability; (4) better study design (randomized vs. not random-
ized) and better accuracy of observation and reporting
(regardless whether a study is randomized or not) are asso-
ciated with lower response rates.
When interpreting our results, a number of limitations
have to be kept in mind. While extractions and assessments
for the RCTs were performed by at least two reviewers the
non-randomized studies were analyzed by a single reviewer.Although each study was read at least two times it cannot be
Table 2
Comparison of randomized and non-randomized studies of acupuncture for chronic headache: Patient and intervention characteristics
RCTs (
n
24) Other studies (
n
35) P-value
Patient characteristics
N treated with acupuncture (median, range) 20 (5–100) 39 (7–340)
0.001
Diagnosis
Migraine 16 (67%) 16 (46%)
Tension-type headache 7 (29%) 3 (8%)various/unclear 1 (4%) 16 (46%) 0.002
Mean age (median, range) 40 (32–46; mi: 7) 40.5 (22–53, mi: 19) 0.515
% proportion female (median, range) 80 (57–100, mi: 3) 78 (33–100, mi: 10) 0.667
Onset of headache
5 years ago 18 (94%, mi: 6) 18 (100%, mi: 17) 0.866
Setting mi: 5 mi: 6
Specialized outpatient dpt. etc. 17 (89%) 20 (69%)
Practice 2 (11%) 9 (31%) 0.161
Intervention characteristics
Number of acupuncture sessions (median; range) 8 (4–14; mi: 3) 10 (1–30; mi: 11) 0.731
Duration of sessions in minutes (median; range) 15 (1–30; mi: 7) 22.5 (10–45; mi: 10) 0.001
Intervention individualized 9 (43%, mi: 3) 14 (48%, mi: 6) 0.631
mi
number of studies with missing information.
Table 3
Comparison of randomized and non-randomized studies of acupuncture for chronic headache: accuracy of observation and reporting, outcome characteristics
and response rates
RCTs (
n
24) Other studies (
n
35) P-value
Accuracy of observation and reporting
Sampling described 8 (33%) 9 (26%) 0.559
Clear headache diagnosis 16 (67%) 8 (23%) 0.001
Patients characterized 17 (71%) 17 (49%) 0.112
At least 4 wks baseline 15 (63%) 5 (14%)
0.001
Cointerventions described 7 (29%) 10 (29%) 1.000
At least two outcomes 20 (83%) 18 (51%) 0.014
Headache diary used 18 (75%) 8 (23%)
0.001
Complete after treatment 11 (46%) 19 (54%) 0.601
Complete at early follow-up 6 (25%) 10 (29%) 1.000
Complete at late follow-up 2 ( 8%) 6 (17%) 0.453
Median number (range) of items met 5 (0–10) 2 (0–10) 0.008
At least 5 quality crit. met 14 (58%) 9 (26%) 0.016
Subjective quality weak 6 (25%) 29 (83%)
0.001
Outcome characteristics
Response data available 16 (67%) 33 (94%) 0.024
Measurement within 3 months after treatment 16 (89%) 21 (64%) 0.634
Complete at response measurement 10 (56%) 21 (64%) 0.765
Response rates (95% confidence intervals) 0.59 (0.48–0.69) 0.78 (0.72–0.83)
0.001
mi
number of trials with missing information.
8/12/2019 Acupuntura e Cefaleia
http://slidepdf.com/reader/full/acupuntura-e-cefaleia 6/9
82
K. Linde et al. / Journal of Clinical Epidemiology 55 (2002) 77–85
ruled out that coding errors occurred or that some details
would have been interpreted differently by a second re-
viewer. The list for assessing the accuracy of observation
and reporting was not formally validated. It can only be
considered as a trial to operationalize quality aspects whichgo beyond methodological criteria covered in commonly
used scales [13–15]. We did multiple statistical testing;
therefore, the significant P-values reported have to be inter-
preted with caution.
A number of empirical analyses have addressed the ques-
tion whether effect size estimates from RCTs and non-ran-
domized controlled studies differ (e.g., [16–18] for newer
original work and [19,20] for reviews of older studies). The
findings of these studies are variable. Among randomized
trials there is evidence that trials with better quality yield
less optimistic results. Particularly, adequate concealment
of randomization and double-blinding seem to be associatedwith less optimistic outcomes [21–23]. We are not aware of
studies comparing characteristics, design-independent qual-
ity aspects and outcomes of RCTs and non-randomized (in-
cluding uncontrolled) studies.
We believe that the current discussion of quality of clini-
cal trials over-emphasizes methodological aspects and disre-
gards clinical aspects. This might have pragmatic reasons as
methodological criteria of good quality are valid in most in-
stances while clinical quality criteria strongly differ among
conditions and interventions. Nevertheless, in our opinion
the quality of a trial on a complex treatment for chronic
headache cannot be assessed in a valid manner by checking
only whether there are the words randomized, double-blind
and a description of drop-outs and withdrawals in the text.
Ten of the 24 RCTs met less than half of our criteria for ac-
curacy of observation and reporting which seem to us quite
basic. This indicates that a relevant proportion of the RCTs
reviewed are unlikely to provide clinically useful results.
The conclusion of our systematic review of RCTs [7]
was that the available evidence suggests that acupuncture is
effective in migraine and tension-type headache but that
further trials with better methodology and larger sample
sizes are urgently needed. The non-randomized studies in-
cluded in the analysis presented here reinforce the impres-
sion of effectiveness. However, the amount of relevant new
information they add is disappointing. The differences be-
tween RCTs and non-randomized studies suggest that pa-
tients and interventions in RCTs are not representative for
routine acupuncture practice. But the sampling strategies in
non-randomized studies—when described at all—did not
allow assessment of generalizability either. Both RCTs and
non-randomized studies almost completely neglected the
question of safety. The majority of non-randomized studies
had very low quality and the potential advantages of using a
non-randomized design were rarely ever used. If we would
have included the non-randomized studies in our original
systematic review [7] the conclusions regarding implica-
tions for practice would not have been different. The con-
clusions regarding future research are obvious: apart from
better RCTs better non-randomized studies with long-term
follow-up and data about safety are needed.
Including non-randomized studies increases the workload
of a systematic review considerably. As efficient search strat-
egies to identify non-randomized studies have not been de-
veloped yet the search has to be very broad. Both during
screening for eligibility and when coding included studies
Table 4
Comparison of randomized and non-randomized studies of acupuncture for
chronic headache: information on long-term results, prognostic factors,
generalizability, and safety
RCTs
(
n
24)
Other studies
(
n
35) P-value
At least 6 months follow-up after
treatment 10 (42%) 14 (40%) 1.000Outcome data for at least 80% of
included patients at follow-up
(at least 6 months) 2 (8%) 6 (17%) 0.453
Analysis of prognostic factors 2 (8%) 3 (9%) 1.000
At least three study centers and
transparent sampling strategy – –
Data on adverse effects 4 (17%) 1 (3%) 0.148
Table 5
Pooled response rates (random effects model) in all studies (both RCTs and
non-randomized) meeting or not meeting defined criteria (univariate mixed
logistic regression model)
Variable
Response rates
(95% confidence intervals)
Criterion not met vs. criterion met P-value
Clinical featureOnly migraine patients 0.72 (0.63–0.79) vs. 0.73 (0.65–0.81) 0.7543
Intervention feature
10 sessions of 15
minutes 0.76 (0.63–0.86) vs. 0.69 (0.60–0.78) 0.3414
Accuracy of observation
and reporting
Sampling described 0.76 (0.71–0.82) vs. 0.62 (0.50–0.72) 0.0138
Clear headache
diagnosis 0.78 (0.72–0.83) vs. 0.62 (0.52–0.72) 0.0037
Patients characterized 0.78 (0.70–0.84) vs. 0.68 (0.60–0.75) 0.0581
At least 4 wks baseline 0.77 (0.72–0.82) vs. 0.61 (0.49–0.72) 0.0099
Cointerventions
described 0.74 (0.67–0.80) vs. 0.69 (0.57–0.79) 0.4695
At least two outcomes 0.77 (0.68–0.83) vs. 0.70 (0.62–0.77) 0.2256
Headache diary used 0.78 (0.73–0.83) vs. 0.61 (0.51–0.70) 0.0015Complete after
treatment 0.75 (0.68–0.82) vs. 0.69 (0.60–0.77) 0.2208
Complete at early
follow-up 0.75 (0.68–0.80) vs. 0.66 (0.54–0.77) 0.1735
Complete at late
follow-up 0.73 (0.66–0.79) vs. 0.72 (0.58–0.82) 0.8816
At least 5 quality crit.
met 0.77 (0.71–0.82) vs. 0.63 (0.53–0.73) 0.0186
Outcome features
Response criterion
50% reduction 0.74 (0.68–0.80) vs. 0.70 (0.57–0.80) 0.4966
Early response
measurement 0.71 (0.64–0.77) vs. 0.79 (0.66–0.88) 0.2383
Complete at response
measurement 0.75 (0.66–0.82) vs. 0.71 (0.63–0.78) 0.4425
8/12/2019 Acupuntura e Cefaleia
http://slidepdf.com/reader/full/acupuntura-e-cefaleia 7/9
8/12/2019 Acupuntura e Cefaleia
http://slidepdf.com/reader/full/acupuntura-e-cefaleia 8/9
84
K. Linde et al. / Journal of Clinical Epidemiology 55 (2002) 77–85
[27] Carlsson J, Fahlcrantz A, Augustinsson LE. Muscle tenderness in ten-
sion headache treated with acupuncture or physiotherapy. Cephalal-
gia 1990;10:131–41.
[28] Ceccherelli F, Altafini L, Rossato M, et al. Trattamento agopunturale
dell’emicrania senz’aura. Studio in doppio cieco vs placebo. Proceed-
ings XV Congresso Nazionale, AISD Associazione Italiana per lo
Studio del Dolore, S. Margherita Ligure, 1992. pp. 310–8.
[29] Doerr-Proske H, Wittchen HU. Ein muskel- und gefä·orientiertes
Entspannungsprogramm (SEP) zur Behandlung chronischer Migräne-patienten: Eine randomisierte klinische Vergleichsstudie. Ztschr Psy-
chosom Med 1985;31:2472–66.
[30] Dowson DI, Lewith GT, Machin D. The effects of acupuncture ver-
sus placebo in the treatment of headache. Pain 1985;21:35–42.
[31] Hansen E, Hansen JH. Acupuncture treatment of chronic tension
headache: a controlled cross-over trial. Cephalalgia 1985;5:137–42.
[32] Henry P, Baille H, Dartigues F, Jogeix M. Traitement de la maladie
par acupuncture: étude controlée. Premières Rencontres Médecines
Alternatives 1986;12/13:209–16.
[33] Hesse J, Mögelvang B, Simonsen H. Acupuncture versus meto-
proplol in migraine prophylaxis: a randomized trial of trigger point
inactivation. J Intern Med 1994;235:451–6.
[34] Heydenreich A, Thiessen M. Effektivitätsvergleich zwischen medika-
mentöser Therapie, invasiver und nichtinvasiver Akupunktur bei der
Migräne (Eine prospektive randomisierte Studie mit Fremdauswer-tung). Ztschr ärztl Fortbild 1989;83:877–9.
[35] Heydenreich A. Punktförmige transkutane elektrische Nervenstimula-
tion (PuTENS) im einfachen Placebovergleich bei der Migräne (Eine
prospektive randomisierte Studie). Ztschr ärztl Fortbild 1989;83:881–3.
[36] Johansson V, Kosic S, Lindahl O, Lindwall L, Tibbling L. Effect of
acupuncture in tension headache and brainstem reflexes. In: Bonica
JJ, Albe-Fessard D, editors. Advances in pain research and therapy,
vol. 1. New York: Raven Press, 1976. pp. 839–41.
[37] Kubiena G, Nissel H, Porenta G, Veitl M, Wessely P. Akupunktur bei
Migräne. Follow-up Studie. Dtsche Ztschr Akup 1992;35:140–8.
[38] Lehmann V, Banzhaf E, Kunze E, et al. Randomisierte klinisch kon-
trollierte Studie der Effizienz der Akupunktur im Vergleich zur Elek-
troakupunktur sowie zur Pharmakotherapie mit Propranolol bei Pa-
tienten mit häufiger Migräne. Dtsche Ztschr Akup 1991;34:27–30.
[39] Lenhard L, Walte PME. Acupuncure in the prophylactic treatment of
migraine headaches: pilot study. N Z Med J 1983;96:663–6.
[40] Loh L, Nathan PW, Schott GD, Zilkha KJ. Acupuncture versus medi-
cal treatment for migraine and muscle tension headaches. J Neurol
Neurosurg Psychiat 1984;47:333–7.
[41] Okazaki K, Sadove MS, Kim SI, Lee MH, Cheng D. Ryodoraku ther-
apy for migraine headache. Am J Chin Med 1975;3;61–70.
[42] Pintov S, Lahat E, Alstein M, Vogel Z, Barg J. Acupuncture and the
opioid system: implications in management of migraine. Pediatr Neu-
rol 1997;17:129–33.
[43] Tavola T, Gala C, Conte G, Invernizzi G. Traditional Chinese acupunc-
ture in tension-type headache: a controlled study. Pain 1992;48:325–9.
[44] Vincent CA. A controlled trial of the treatment of migraine by acu-
puncture. Clin J Pain 1989;5:305–12.
[45] Vincent CA. The treatment of tension headache by acupuncture: a
controlled single case design with time series analysis. J Psychosom
Res 1990;34:553–61.
[46] Weinschütz T, Lindner V, Niederberger U, Schreiber J, Soyka D.
Möglichkeiten und Grenzen der Akupunktur in einer neurologischen
Schmerzambulanz: Reevaluierung der klinischen Wirkung am
Beispiel der Migränebehandlung. Schimrigk K: Verhandlungen der
Deutschen Gesellschaft für Neurologie, vol. 7. Heidelberg: Springer
Verlag, 1993. pp. 533–4.
[47] Weinschütz T, Niederberger U, Johnsen S, Schreiber J, Kropp P. Zur
neuroregulativen Wirkung der Akupunktur bei Kopfschmerzpa-
tienten. Dtsche Ztschr Akup 1994;37:106–17.
[48] White AR, Eddleston C, Hardie R, Resch KL, Ernst E. A pilot study
of acupuncture for tension headache, using a novel placebo. Acup
Med 1996;14:11–5.
[49] Börglum-Jensen L, Melsen B, Jensen SB. Effect of acupuncture on
headache measured by reduction in number of attacks and use of
drugs. Scand J Dent Res 1979;87:373–80.
[50] Guccione G, Gentile F, Pullara C, Sparacia B, Liguori A, Sparacia A.
Trattamento dell’emicrania senza aura con agopuntura basata sulla
differenziazione delle sindromi vs terapia farmacologica e agopunu-
tra standardizzata. In: Gruppo italiano di ricerca in agopuntura. La
ricerca in agopuntura in Italia. Roma: Edizioni Paracelso, 1996. pp.
175–7.[51] Sold-Darseff J, Leydhecker W. Akupunktur bei Schmerzen im Kopf-
bereich sowie bei Blepharospasmus ohne organische Ursache. Klin
Monatsbl Augenheilkd 1986;189:167–9.
[52] Turk Z, Moser I. Heilung des chronischen Kopfschmerzes durch Ak-
upunktur. Erfahrungsheilkunde 1990;39:724–6.
[53] Zibin D. Treating vascular migraine with deep puncture at fengchi: a
report of 240 cases. Int J Clin Acup 1994;5:455–8.
[54] Baischer W. Acupuncture in migraine: long-term outcome and pre-
dicting factors. Headache 1995;35:472–4.
[55] Boivie J, Brattberg G. Are there long lasting effects on migraine
headache after one series of acupuncture treatments? Am J Chin Med
1987;15:69–75.
[56] Börglum-Jensen L, Jensen SB. Effect of acupuncture on tension
headache and urinary catecholamine excretion. Scand J Dent Res
1982;90:397–403.[57] Börglum-Jensen L, Tallgren A, Troest T, Jensen SB. Effect of
acupuncture on myogenic headache. Scand J Dent Res 1977;85:
456–70.
[58] Ceccherelli F, Cavallo C, Ambrosio F, Altopiedi A, Moro-Boscolo E,
Giron GP. Psychologische Untersuchung bei Patienten mit chronis-
chem Schmerz, die mit Akupunkturreflextherapie behandelt wurden.
Dtsch Ztschr Akup 1983;5:105–12.
[59] Ceccherelli F, Ambrosio F, Altopiedi A, et al. Trattamento agopun-
turale delle cefalee. Min Riflessoter Laserter 1985;2:45–59.
[60] Kim KC, Yount RA. The effect of acupuncture on migraine head-
ache. Am J Chin Med 1974;2:407–11.
[61] Major P, Standnes B. Akupunkturbehandling ved migrene. Tidskr
Nor Laegeforen 1983;103:686–9.
[62] Tolksdorf W, Klimm J, Klimczyk K, Penninger M. Die Akupunktur
zur Behandlung chronischer Kopfschmerzen. Eine prospektive Studie
zum Einfluss der Depression auf den Behandlungserfolg. Dtsche
Ztschr Akup 1988;31:104–9.
[63] Zwölfer W, Kaiser B, Cartellieri M, Spacek A. Behandlung des Spannung-
skopfschmerzes mit Akupunktur. Dtsche Ztschr Akup 1994;37:129–32.
[64] Batra YK. Acupuncture in the treatment of migraine. Am J Acup
1986;14:135–7.
[65] Cheng ACK. The treatment of headaches employing acupuncture.
Am J Chin Med 1975;3:181–5.
[66] Fischer MV, Behr A, von Reumont J. Acupuncture: a therapeutic
concept in the treatment of painful conditions and functional disor-
ders. Report on 971 cases. Acupuncture & Electro-Therapeutics Res
Int J 1984;9:11–29.
[67] Heydenreich A. Die punktförmige transkutane elektrische Nerven-
stimulation in der Migränetherapie. Psychiat Neurol med Psychol
1988;40:717–23.
[68] Heydenreich A. Die punktförmige transkutane elektrische Nerven-
stimulation bei funktionellen vertebrogenen Störungen und bei der
Migräne. Ztschr Ges Inn Med 1988;43:651–3.
[69] Laitinen J. Acupuncture for migraine prophylaxis: a prospective clini-
cal study with six months follow-up. Am J Chin Med 1975;3:271–4.
[70] Lawrence RM. Acupuncture for the treatment of headache. A prelim-
inary report. Am J Acup 1974;2:105–8.
[71] Spoerel WE, Verkey M, Leung CY. Acupuncture in chronic pain. Am
J Chin Med 1976;4:267–79.
[72] Stux G. Treatment of migraine with acupuncture and moxibustion.
Acup Med 1987;4(2):6.
[73] Yamauchi N. The results of therapeutic acupuncture in a pain clinic.
Can Anaesth Soc J 1976;23:196–206.
8/12/2019 Acupuntura e Cefaleia
http://slidepdf.com/reader/full/acupuntura-e-cefaleia 9/9
K. Linde et al. / Journal of Clinical Epidemiology 55 (2002) 77–85
85
[74] Carlsson CPO, Sjölund BH. Acupuncture and subtypes of chronic
pain: assessment of long-term results. Clin J Pain 1994;10:290–5.
[75] Chen GS, Hwang YC. Therapeutic effect of acupuncture for chronic
pain. Am J Chin Med 1977;5:45–61.
[76] Haug HU, Robben H. Die Akupunktur als Objekt allgemeinmediz-
inischer Forschung. Ztschr Allg Med 1986;62:607–12.
[77] Junnila SYT. Acupuncture for long-term treatment of headache in a
national health center. Am J Acup 1986;14:351–3.
[78] Junnila SYT. Long-term treatment of chronic pain with acupuncture.Part I. Acupuncture & Electro-Therapeutics Res Int J 1987;12:23–36.
[79] Kubiena G. Akupunktur bei Migräne – eine Dreijahresstudie mit den
Migränepatient(inn)en der Ambulanz des Ludwig-Boltzmann-Institutes
für Akupunktur in Wien 1976-1978. Dtsch Ztschr Akup 1985;1:4–14.
[80] Maxion H, Klein HJ. Zur Pathogenese und Behandlung der Migräne.
Münch med Wschr 1978;120:315–8.
[81] McKibbin LS, Downie R. Treatment of migraine headaches using au-
ricular acupuncture techniques. Laser Ther 1993;5:23–8.
[82] Pöntinen PJ, Salmela TM. Acupuncture treatment of migraine. In: Si-
cuteri F, editor. Headache new vistas. Florence: Biomedical Press,
1977. pp. 251–6.[83] Shifman AC. The clinical response of 328 private patients to acu-
puncture therapy. Am J Chin Med 1975;3:165–79.