close

Вход

Забыли?

вход по аккаунту

?

Signal measurement strategiesare they feasible and do they offer any advantage in outcome measurement in osteoarthritis.

код для вставкиСкачать
739
BRIEF REPORT
SIGNAL MEASUREMENT STRATEGIES: ARE T H E Y FEASIBLE AND
DO T H E Y O F F E R ANY ADVANTAGE IN OUTCOME MEASUREMENT
IN OSTEOARTHRITIS?
NICHOLAS BELLAMY, W. WATSON BUCHANAN, CHARLES H. GOLDSMITH. JANE CAMPBELL,
and ERIC DUKU
The applicability of a signal measurement strategy was compared with a traditional method of measuring outcome in osteoarthritis. The signal method detected statistically significant alterations in health status
with small sample sizes and with a relative efficiency
close to or at unity. The prevalence of deterioration in
nonsignal items was low. Signal methods of measurement may provide an alternative approach to outcome
measurement in osteoarthritis clinical trials.
Signal measurement is a methodologic technique in which the measurement of disease is based on
well-defined, individualized targets. Thus, in arthritis,
measurement is restricted to only 1, or a few, selected
joint(s) or symptom(s). This technique has 2 objectives: 1) to tailor the measurement process to the
symptom profile of the individual patient, and 2) to
improve the efficiency of the measurement process by
excluding joints or other items that lack response
__ _ _
From the Department of Medicine. llniversity of Western
Ontario. London, Ontario. and the Dcpartmcnt of Clinical Epidemiology and Biostatistics, McMaster University, Hamilton. Ontario.
Canada.
Supported in part by a grant from the Arthritis Society of
Canada and a grant from Parke-Davis Canada Inc.
Nicholas Bellamy. MD. MSc. FRCP (Edin, Cilas. and C).
PACP: Associate Professor of Medicine, Epidemiology, and Biostatistics. University of Western Ontario; W. Watson Buchanan. MD.
FKCP (Edin. Glas. and C): Professor of Medicine. McMaster
University; Charles H. Goldsmith. PhD: Professor of Clinical Epidemiology and Biostatistics. McMaster University; Jane Campbell.
BA: Research Assistant. University of Western Ontano: Eric Duku.
BSc. Grad Dip, MSc: Department of Clinical Epidemiology and
Biostatistics. McMaster University.
Address reprint requests 10 Nicholas Bellamy. MD, MSc.
Suite 402A, Victoria Hospital, Westminster Tower, 800 Commissioncrs Road East. London, Ontario N6A 4G5, Canada.
Submitted for publication August 22, 1989: accepted in
revised form December 5 . 1989.
Arthritis and Rheumatism, Vol. 33, No. 5 (May 1990)
potential. In spite of the possible advantages of signal
measurement and the success of the technique as
reported by Dixon et a1 ( I ) . it is noteworthy that
neither Ward et al (2) nor Egger et a1 (3) demonstrated
any clear superiority of this technique over more
traditional methods. However, Tugwell et al (4) recently replicated the success of a signal technique in
potentially reducing sample size requirements for clinical trials.
Since all of the above-mentioned studies were
conducted with rheumatoid arthritis patients. the generdlizability of the observations to the measurement of
disease in osteoarthritis (OA) patients is unknown. For
this reason. during the validation of a new outcome
measure for OA clinical trials. designated the Western
Ontario and McMaster Universities (WOMAC) Osteoarthritis Index ( 5 ) . we compared the performance
of a signal method of measurement with that of a
measure based on aggregated items. in each of 3
different dimensions. The reliability, validity, and responsiveness of the WOMAC Index have been documented ( 5 ) .
This report presents a further analysis of data
from that study, comparing signal measurement versus
aggregate techniques of measurement. Specifically, we
wished to examine whether there was any advantage
in using the WOMAC question inventory as a menu for
identifying signal symptoms for each individual in each
of 3 dimensions, compared with administering the
inventory in its entirety. To address this, we identified
(a) the nature and seventy of symptoms in the
WOMAC inventory that were selected as signals, (b)
whether the signal strategy could be successfully applied in outcome measurement in OA of the hip or
knee, (c) the relative efficiency of the 2 techniques, (d)
BRIEF REPORTS
740
whether deterioration in nonsignal items was overlooked using the signal strategy, and (e) the sample
size implications of the 2 techniques.
nonparamctric statistics for paired data, we restricted
our analysis to those patients who both completed the
6-month study and completed all WOMAC questionnaires in their entirety. This reduced the available
sample for analysis to 20 patients for the pain dimension, 27 for the stiffness dimension. and 24 for the
physical function dimension. The remaining questionnaires could not be analyzed using paired statistics
because they were not entirely complete.
All responses in this study were made on 5point Likert scales (0 = none, I = mild, 2 = moderate,
3 = severe, and 4 = extreme). For pain and stiffness,
these scales required responses to questions about
severity, while for physical function, the questions
were phrased in terms of degree of difficulty experienced. Aggregate values for each dimension were
calculated by totaling the scores across all component
items. Signal values for each dimension were represented by the score of the individual item selected.
To examine the possibility that patients tended
to focus on signals pertaining to the aspects of their
physical condition that were most severely affected,
the relative ranking of the signal measurement among
other items in the same dimension was determined.
Values were tied for some items; that is, signal values
had the same severity score as nonsignal items. Such
occurrences are indicated in the tables. In instances
where signal severity scores were lower than scores
for some nonsignal items, the rank is specified, regardless of any ties in higher-ranked items (for ranking
rules. see Table 1).
PATIENTS AND METHODS
Thirty patients with primary OA of the hip or
knee were interviewed prior to total joint replacement
surgery and at 6 weeks, 3 months, and 6 months after
surgery. The criteria used for patient selection, as well
as the disease and demographic profiles of the study
group, have been reported previously ( 5 ) .
At the initial interview, each patient was shown
the item inventory of the WOMAC Index and asked to
select 1 item from each of the 3 dimensions that was of
importance to him or her as the focus for further
measurement during the study. Specifically, patients
were asked to select 1 pain item, 1 stiffness item, and
1 physical function item that was most important to
them, i.e., that they most hoped the treatment they
were about to receive would improve. However, in
order to compare signal and aggregate strategies at
each of the assessment points, the full WOMAC Index
was self-administered throughout the study, thus obtaining serial measurements on each of the components of its 3 dimensions.
The present analysis was confined to a comparison of baseline results versus results at 6 months,
because these represent the extremes of the response
for both signal and aggregate strategies. Since we
wished to compare the scores of the same individuals
at these 2 time points using both parametric and
Table 1.
Pain dimension signals'
No. of patients
selecting item
as a signal
~~~~~~
~
Signal score
at baselinc
Item ranking+
Meant
Scorc at
baseline
Change from baseline
to 6 months ( P )
r-test
Wilcoxon test
SD (parametric) (nonparametric)
SD
Meant8
2.8
1.3
2.1
1.2
<0.001
2.8
0.8
2.6
0.9
<O.OOI
3.3
2.0
0.6
ND
1.9
1.4
1.4
<O.OOI
4.0
ND
2.4
2.9
10.5
0.9
4.4
2.9
10.5
1.0
0.9
<0.001
<0.001
<0.001
~
Individual item
Pain while walking on a Rat surface
Pain while walking up or down stairs
5
10
Pain at night while in bed
Pain while sitting or lying
Pain while standing upright
Signal (n = 20)
Aggregate (n = 20)
3
1
I
NA
NA
3.
10,
IT,.
IF.
in.
in.
~(i,
17. I F
1%. I .
2. 11, 1. I . I
3F
in
NA
NA
in
in
* ND = not determined (due to lack of sufficient numbers); NA = not applicable.
t Item rank for each patient who selected item as a signal (possible number of ranks
=
1.6
4.4
0.002
0.001
<0.001
0.001
0.004
<0.001
<0.001
<0.001
5).
$ Except for the aggregate, possible scores could range from 0 (not affected) to 4 (most severe).
5 Derived by aggregating all scores at baseline for each individual item.
T Tied values. Note: In the case of a tied pair, there is a discontinuity of 1 category in the subsequent ranking. Thus, if there is a tied pair in
third rank, then the 5 ranks are expressed as follows: I . 2, 3Y. 5 .
BRIEF REPORTS
Each of the 3 dimensions of WOMAC was
analyzed separately using the Wilcoxon matched pairs
signed rank test and Student’s paired 1-test (6). The
SPSS-X software program was used to calculate the
test statistics (7). P values less than 0.05 were considered significant, and no correction was made for
multiple comparisons. We elected to use both parametric and nonparametric techniques. since parametric techniques may be applicable for certain ordinallevel data. However. our data were generally not
normally distributed, and we believe the use of the
nonparametric technique provides a more conservative estimate of statistical significance (6); results of 2
previous studies using the WOMAC Index support this
contention (5,8).
Relative efficiency (RE) for the parametric analysis was calculated using the method reported by
Liang et a1 (9). where RE (signal versus aggregate) =
(Isignal/taggregatc)*. For the nonparametric tests, RE (signal versus aggregate) = (Zsigna,/Zaggregarc)’).
Nonsignal
deterioration was determined by comparing presurgery and postsurgery scores for each item not identified as a signal in the full WOMAC inventory and by
noting the frequency and magnitude of any deterioration. Sample size requirements, based on matched
pairs analysis, were calculated for both the signal and
aggregate strategies for each dimension, using a parametric technique. Calculations were based on the assumptions that P values less than 0.05 were significant.
the power of the test was 90%. and the difference to be
detected could be in either a positive or a negative
direction, i.e., a 2-tailed test of the null hypothesis. For
each strategy, the sample size formula used was as
follows: n matched pairs of observations = ([Zo,os +
Z , , , , ] C ~ D ) ~where
.
u = the standard deviation of
differences and D = the decimal difference (from
baseline) to be detected (i.e.. 0.25 of mean) (6).
RESULTS
Pain. Each of the 5 items included in the pain
dimension was selected as a signal by 1 or more
patients (Table 1). Measures of pain under conditions
of activity were more frequently selected as signals
than those under conditions of passivity: Pain observed when walking up or down stairs was the most
frequently selected signal. The mean scores at baseline
for signal items indicated that patients selected items
for which the severity was intermediate, and which
could therefore potentially show response (i.e., could
either improve or deteriorate). Six of 20 patients (30%)
741
selected as signals items for which they scored the
pain as extreme (i.e., a score of 4): none selected
signals scored as 0. The pain signal was usually ranked
highest in severity. though it was often tied in severity
with other nonsignal items.
Using both the signal and the aggregate strategies, there was statistically significant improvement at
6 months postsurgery versus baseline (P < 0.001),
irrespective of the type of analysis ( i x . , parametric or
nonparametric). The relative efficiency (signal versus
aggregate) was 1.00 for nonparametric analysis and
1.30 for parametric analysis. When individual items
were analyzed, all 5 pain items detected statistically
significant improvement in the pain level ( P 5 0.004).
Clinical deterioration in those items not selected as
signals (nonsignal deterioration) occurred on 4 occasions (i.e., 5% of item selections) in 3 patients. The
magnitude of the deterioration was as follows: mean
1.00, SD 0, range 0. Sample size requirements were
lower for the signal strategy (n = 17) than for the
aggregate strategy (n = 30).
Stiffness. Patients designated both stiffness
items as signals, with morning stiffness being selected
slightly more frequently than “gelling” (Table 2). The
mean scores at baseline suggested that patients were
selecting potentially responsive signals. Two of 27
patients (7%) selected items for which the seventy was
rated as extreme; in both cases, this was the morning
stiffness item. No patient selected as a signal an item
for which the severity was scored as 0. The stiffness,
signal was often ranked highest, except in 2 instances.
although it was tied in severity with the other nonsignal item.
Use of both the signal and the aggregate strategies enabled detection of statistically significant improvement ( P < 0.001), regardless of the type of
analysis. The techniques used were of similar relative
efficiency (signal versus aggregate) for parametric
analysis (RE = 1.OO); the signal technique was slightly
less efficient by nonparametric analysis (RE = 0.94).
When the individual items were analyzed, both stiffness items detected statistically significant improvement ( P < 0.001). Clinical deterioration in nonsignal
items occurred in only 2 situations in 2 patients (i.e.,
7% of item selections). The magnitude of the deterioration was as follows: mean 1.00, SD 0. range 0.
Although the required sample sizes for the 2 strategies
were quite similar, that for the aggregate strategy (n =
33) was lower than that for the signal strategy (n = 26).
Physical function. Eleven of the 18 physical
function items in the WOMAC inventory were se-
742
Table 2.
BRIEF REPORTS
Stiffness dimension signals*
Signal score
at baseline
No.. of patients
selecting item
as a signal
Individual item
Morning stiffness
"Gelling"
Signal (n = 27)
Aggregate (n = 27)
16
11
NA
NA
Change from baseline
to 6 months (P)
Score at
baseline
Item rankingt
Meant
SD
Meant8
SD
r-test
(parametric)
Wilcoxon test
(nonparametric)
I , I . I , IW. iw, in
19. in. 1. 1, 17
1, 18. 1. 1, 16
in, I . 19. I . 1. IF
in. in. IR. 2, 2
NA
NA
2.5
1 .o
2.4
0.9
<0.001
<0.001
2.1
0.7
2.0
0.8
<0.001
<0.001
2.3
4.4
0.9
1.6
2.3
4.4
0.9
1.6
<0.001
<0.001
<0.001
<0.001
* Morning stifhess = stiffness on first awakening; "gelling" = stiffness after sitting, lying. or resting later in the day: NA
t Item rank for each patient who selected item a s a signal (possible number of ranks = 2).
t Except for the aggregate, possible scores could range from 0 (not affected) to 4 (most severe).
=
not applicable.
5 Derived by aggregating all scores at baseline for each individual item.
Tied values. See Table 1 for explanation of ranking rules.
items for which the degree of difficulty was rated as
extreme (i.e., a score of 4); none selected a signal
scored as 0. Rankings of the physical function signals
selected vaned from first to sixteenth, but they were
usually ranked first, second, or third in severity.
lected as signals (Table 3). Difficulty ascending stairs
was the most frequently selected signal. The mean
scores at baseline for signal items suggested that
patients were selecting items for which there was
potential response. Six of 24 patients (25%) selected
Table 3.
Physical function dimension signals'
No. of patients
selecting item
as a signal
Individual item
Descending stairs
Ascending stairs
Rising from sitting
Standing
Bending to floor
Walking on flat surface
Getting idout of car
Going shopping
Putting on socks
Rising from bed
Taking off socks
Lying in bed
Getting idout of bath
Sitting
Getting odoff toilet
Heavy domestic duties
Light domestic duties
Getting,onloff a bus
Signal (n = 74)
Aggregate (n = 24)
2
5
0
0
2
3
3
3
1
1
0
1
1
0
0
2
0
0
NA
NA
Signal score
at baseline
Change from baseline
to 6 months (P)
Score at
baseline
Item rankingt
Meant
SD
Meant§
SD
in, in
29, 29, l(1. 68, 18
NA
NA
61. 5T
IF. Ilq, 16Ti
3n, 3n. in
7(, 18, 18
1%
13F
NA
86
5n
NA
NA
in, 3n
NA
NA
NA
NA
3.5
3.2
NA
NA
2.0
3.3
2.7
2.7
4.0
2.0
NA
3.0
2.0
NA
NA
3.5
NA
NA
3.0
42.8
0.7
0.8
NA
NA
0.0
0.6
0.6
0.6
ND
ND
NA
ND
ND
NA
NA
0.7
NA
NA
0.8
12.8
2.7
2.9
2.8
2.4
2.4
2.0
2.8
2.9
2.5
2.2
2.0
1.5
2.8
1.6
2.1
3.3
1.8
2.1
3.0
42.8
1 .o
1.0
0.9
1.1
I .3
1.1
1.1
0.8
1.4
1.2
1.4
1.3
1.2
1.1
1.3
1.2
I .2
1.7
0.8
12.8
* NA = not applicable; N D = not determined (due to lack of sufficient numbers).
+ Item rank for each patient who selected item as a signal (possible number of ranks = 18).
t Except for the aggregate, possible scores could range from 0 (not affected) to 4 (most difficult).
§ Derived by aggregating all scores at bascline for each individual item.
TI Tied values. See Table 1 for explanation of ranking rules.
r-test
(parametric)
Wilcoxon test
(nonparametric)
10.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
0.001
<O.M)l
<0.001
<0.001
<0.001
<0.001
0.001
<0.001
0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
0.003
<O.OOl
<0.001
<O.M)I
0.001
0.002
<0.001
0.001
<0.001
0.002
<0.001
0.009
<0.001
<0.001
BRIEF REPORTS
Both signal and aggregate strategies detected
statistically significant improvement ( P < 0.001), irrespective of the type of analysis. The signal strategy
was slightly more efficient by parametric analysis (RE
= 1.02) but slightly less efficient by nonparametric
analysis (RE = 0.96). When individual items were
analyzed, all physical function items detected statistically significant improvement, regardless of the type of
analysis ( P 5 0.002). Clinical deterioration in nonsignal items occurred on 27 occasions in 7 patients 6.e..
7% of item selections). The magnitude of the deterioration was as follows: mean 1.56, SD 0.70, range 3.
The sample size requirements were lower for the
signal strategy (n = 12) than for the aggregate strategy
(n = 16).
DISCUSSION
The principal objective of evaluative research is
to detect clinically important and statistically significant alterations in health status. This objective can be
most readily achieved by the application of highly
responsive instruments to probe aspects of disease
that are of defined importance. We have developed
and reported on such an instrument ( 5 . 8 ) , which is of
potential value in the self-assessment of patients with
OA of the hip or knee who have had surgical or
pharmacologic intervention. To further enhance the
statistical efficiency of the WOMAC Index and more
closely tailor the measurement process to the unique
symptom profile of the individual patient, we investigated the relative merits of using a single question
from each of the 3 dimensions in the WOMAC inventory (signal technique).
To be considered an ideal alternative, the signal
strategy would have to satisfy the following requirements: 1) Patients would differ in their selection of
signal items. such that a single fixed item would not
adequately convey the nature of all patients' symptomatology. 2) Patients would avoid signal items that
lack response potential. This would certainly include
the avoidance of items given a severity score of 0 and
would likely entail the use of few, if any, items given a
severity score of 4. 3) The signal strategy would be
capable of detesting statistically significant ( P 5 0.05)
alterations in health status with conventional sample
sizes. 4) Relativie efficiency would be greater than
unity for the technique to be judged more efficient than
the aggregate technique. As a result of increased RE,
sample size requirements would be lower for the signal
strategy than fsr the aggregate strategy. 5 ) The signal
743
technique would adequately capture the patients'
symptoms, such that deterioration occurring among
items not selected as signals would not be overlooked.
Of course, an index meeting all of these criteria
is unusual. However, a signal strategy meeting most of
these criteria might still provide a useful alternative to
more traditional forms of measurement. Several common themes emerge from the present study. The
selection of the majority of items in the WOMAC
inventory suggests that different patients place importance on different symptoms. The use of only a single
fixed item is. therefore. deemed inappropriate and
justifies the use of either a signal or aggregate approach to measurement. Indeed, the patient global
assessment, which is often recommended and used in
clinical studies, may itself be the result of a process in
which the patient selects, weights, and aggregates a
limited number of symptoms into an overall score (i.e.,
a signal strategy). However, the issue of selection
cannot be decided merely by adopting a policy of
selecting the signal by the most severely affected item
in each WOMAC dimension since, as indicated in
Tables 1-3, 71% of the first-placed rankings were tied.
It is also important to note that not all items
selected as signals were ranked first in severity, and
some of those selected were tied in rank. It is of
substantial advantage, then, to allow the patient to
designate which item is most important and will therefore be the principal object of observation and therapy
for that patient. The success of the signal strategy in
the present study is likely due to the fact that the vast
majority of patients selected potentially responsive
items as signals: Very few selected items for which the
severity was scored as 4, and none selected items
scored as 0. However, as indicated by the rankings,
patients tended to select more severely affected items
as signals. In all 3 dimensions, the signal strategy was
responsive to change, demonstrating P values 50.002
in spite of the relatively small sample sizes used (n =
20-27). It should be noted that such high levels of
significance may be a reflection of the potency of the
surgical intervention. However, we have also observed similar levels of significance in another validation study of the WOMAC, in the context of a doubleblind, randomized controlled trial of 2 nonsteroidal
antiinflammatory drugs (8). We do not regard the
observed success in this study as being generalizable
to other subscales containing components that do not
have strong potential for improvement or deterioration.
The pain, stiffness, and physical function dimensions of
BRIEF REPORTS
744
WOMAC contain only items that have been evaluated
and found to have high response potential (5,8).
Relative efficiency is a crude, albeit convenient,
method of comparing the effect size of different instruments. As such, its value may be affected in different
studies by the severity of involvement for individual
items and the frequency with which they are selected
as signals. The utility of the RE as a measure of
statistical economy of one instrument over another is
uncertain and merits further evaluation. Nevertheless,
the 3 dimensions included in the final WOMAC Index
showed RE values close to unity, especially when the
preferred nonparametric comparisons were considered. It is possible that the greater response of signal
variables is explained by the effects of regression to
the mean or a limit effect. We think it is unlikely that
the response is merely a statistical aberration due to
repeated random sampling (i.e.. regression to the
mean), since the Index as a whole also reflected the
benefits of surgery. However, the further elucidation
of a regression effect would require a randomized,
"placebo"-controlled study, a design that poses major
practical and ethical problems in the surgical setting. A
limit effect is also unlikely, since in many cases,
patients selected signals that were either tied at rank 1
or of a lower rank. Although many of the signals
selected were items for which seventy was rated as
extreme, we believe that the fact that not all were
makes the signal measure different from the simple
selection of items scored as extreme. even though the
numbers may not be that different.
With respect to sample size requirements, we
have demonstrated that the signal strategy may be less
demanding. With the exception of the stiffness dimension, sample size requirements were lower with the
signal strategy than with the aggregate strategy. This,
however, does not indicate that the signal strategy is
necessarily the technique of choice. If, indeed, comprehensiveness is the key objective, then regardless of
the slightly larger sample size requirement. the aggregate strategy remains the preferred technique. Indeed,
investigators are frequently faced with the dilemma of
selecting between simple and complex measures, each
having different sample size requirements. The consequence of applying standard measures by signal or
traditional techniques has been reported for only a
small number of musculoskeletal indices. We anticipate that the present data will allow potential users to
select the preferred mode of administration for future
studies using the WOMAC Index. Deterioration in
items not selected as signals occurred in each of the 3
subscales. However, these deteriorations were infrequent (9%) and of a low order of magnitude (mean
1.47; SD 0.74).
These data indicate that, while the signal strategy does not fulfill all of the ideal criteria mentioned
above, from a practical standpoint, such a strategy
may nevertheless provide a useful alternative to aggregate techniques of measurement, at least with respect
to the WOMAC Index. In particular, it allows (a) a
broader-based item selection than the use of a single
fixed item, (b) the detection of statistically significant
alterations in health status, (c) higher levels of RE than
the majority of either the aggregate or individual-item
approaches to outcome measurement, and (d) reduced
sample size requirements for studies using WOMAC
as the principal outcome measure. It is limited by (a)
the tendency of patients to select more severely affected items and, in some instances, items of extreme
seventy lacking response potential, and (b) its inability
to detect nonsignal deterioration. The generalizability
of our observations to future applications of WOMAC
or, indeed, to signal methods of administration of
other health status instruments, is, by necessity, limited. We do, however, agree with Tugwell and colleagues (4,lO) that the signal technique merits further
evaluation. since in some circumstances, it may facilitate attainment of high levels of statistical significance
and clinical relevance.
REFERENCES
1. Dixon JS, Hill J , Bird HA: Measurement of proximal
2.
3.
4.
5.
inter-phalangeal joint circumference in rheumatoid arthritis: one joint or ten? Br J Rheumatol 26:123-125,
1987
Ward JR, Niethammer TA, Egger MJ: Can we just
measure signal joints? Should ring size and waking time
be analyzed only in selected patients. Controversies in
the Clinical Evaluation of Analgesic-Anti-lnflamatory-Antirheumatic Drugs. Edited by H E Paulus, GE
Ehrlich, E Lindenlaub. New York, FK Schattauer Verlag, 1981
Egger MJ. Huth DA, Ward JR, Reading JC, Williams
HJ: Reduced joint count indices in the evaluation of
rheumatoid arthritis. Arthritis Rheum 28:61M19, 1985
Tugwell P, Bombardier C, Buchanan WW, Goldsmith
CH, Grace E, Hanna B: The MACTAR patient preference disability questionnaire: an individualized functional priority approach for assessing improvement in
physical disability in clinical trials in rheumatoid arthritis. J Rheumatol 14:446451, 1987
Bellamy N , Buchanan WW, Goldsmith CH. Campbell J,
Stitt L: Validation study of WOMAC: a health status
BRIEF REPORTS
instrument for measuring clinically-important patientrelevant outcomes following total hip or knee arthroplasty
in osteoarthritis. J W h o p Rheumatol 1:95-108, 1988
6. Colton T: Statistics In Medicine, Boston, Little, Brown
and Co., 1974
7. The SPSS-X Users Guide. Third edition. Chicago, SPSS
Inc., 1988
8. Bellahy N , Buchanan WW, Goldsmith CH, Campbell J,
Stitt LW: Validation study of WOMAC: a health status
instrument for measuring clinically important patient
745
relevant outcomes to antirheumatic drug therapy in
patients with osteoarthritis of the hip or knee. J Rheumatol 15:183>1840. 1988
9. Liang MH, Larson MG, Cullen KE. Schwartz JA:
Comparative measurement efficiency and sensitivity of
five health status instruments for arthritis research.
Arthritis Rheum 28:542-547. 1985
10. Tugwell 1’. Bombardier C: Practical issues in clinical
trials as applied to current state of the art in rheurnatology trials. Br J Rheumatol 22 (suppl):9-13, 1983
Документ
Категория
Без категории
Просмотров
2
Размер файла
557 Кб
Теги
measurements, outcomes, signali, osteoarthritis, feasible, offer, strategiesare, advantages
1/--страниц
Пожаловаться на содержимое документа