The value of conventional views and radiographic magnification in evaluating early rheumatoid arthritis.

744
THE VALUE O F CONVENTIONAL VIEWS AND
RADIOGRAPHIC MAGNIFICATION IN EVALUATING
EARLY RHEUMATOID ARTHRITIS
ROBERT M. HARTLEY, MATTHEW H. LIANG, BARBARA N. WEISSMAN, J. LELAND SOSMAN,
ROBEKT KATZ, and JOHN R. CHARLTON
Fifty-four patients with suspected early rheumatoid arthritis had radiographs taken of their hands and
wrists in 4 views (posteroanterior [PA], oblique, reverse
oblique, and Brewerton) using conventional techniques
and, in the PA view, using radiographic magnification.
The radiographs were “masked” and presented in
random order to 2 radiologists specializing in bone and
joint radiology who interpreted them for malalignment,
erosions, joint space narrowing, and soft tissue swelling.
The PA was the best conventional view for demonstrating malalignment, joint space narrowing, and soft tissue
abnormalities; the Brewerton view was better for detecting erosive disease. Radiographic magnification was
more sensitive than conventional films for evaluating
erosive disease, but otherwise was no better than the
conventional PA view. These results help the physician
choose the radiologic technique or combination of techniques that is most likely to detect specific abnormalities.
From the Robert B. Brigham Multipurpose Arthritis Center
and the Departments of Medicine, Rheumatology and Immunology.
and Radiology, Harvard Medical School, Brigham and Women’s
Hospital, Boston, Massachusetts.
Supported by National Institutes of Health grant AM20580.
Robert M. Hartley, MD, MSc: Milbank Memorial Fund
Scholar; Matthew H. Liang, MD, MPH; Barbara N. Weissman,
MD; J . Leland Sosman, MD; Robert Katz, BA: Department of
Clinical Psychology, Syracuse University, Syracuse, New York;
John R. Charlton, MSc: Department of Community Medicine, St.
Thomas’s Hospital Medical School. London, England.
Address reprint requests to Dr. Robert M. Hartley,
Brigham and Women’s Hospital, 75 Francis Street, Boston, MA
021 IS.
Submitted for publication April 11, 1983; accepted in revised form March 12, 1984.
Arthritis and Rheumatism, Vol. 27, No. 7 (July 1984)
Radiographic imaging is an important tool in the
evaluation of arthritic disorders, particularly in rheumatoid arthritis (RA). Detection of erosions is of
particular diagnostic and prognostic importance. In
the appropriate clinical setting erosions are diagnostic
of RA, and because they indicate structural damage,
suggest the need for aggressive therapy (1-3). The
finding of soft tissue swelling in the proximal interphalangeal joints, metacarpophalartgeal joints, and about
the ulnar styloid also strongly suggests the diagnosis of
RA, even in the absence of erosions (4).
A number of different radiographic techniques
and projections are available for joint imaging. For the
hands and wrists, the posteroanterior (PA), oblique,
reverse oblique, and Brewerton projections are commonly used “conventional views” of the hand. A
“routine” examination may involve any combination
of these projections. Which of these views or combination of views is most efficient in demonstrating the
radiographic abnormalities of RA has not been systematically studied.
Fine-detail radiography using industrial quality,
fine-grained film with optical magnification has been
shown to be useful in the detection of early erosive
disease (5,6) but is not in widespread use. A more
recent development, direct radiographic magnification, offers distinct advantages compared with optical
magnification. In particular, it gives less radiation and
does not require special processing and viewing equipment. Compared with conventional radiography, radiographic magnification gives about 4 times the skin
dose, though the total dose still remains quite low, 4 0
mR (7).
Radiographic magnification is more expensive.
The equipment can only be used for magnification
RADIOGRAPHIC EVALUATION OF RA
Figure 1. Radiographs of hands in the A, posteroanterior. B, oblique, C, reverse oblique, and D, Hrewerton positions.
745
HARTLEY ET AL
746
films, and in addition to this fixed cost, we estimate
that, compared with conventional techniques, the marginal costs of film, processing, and technician time are
about 5 dollars per PA magnification examination of
both hands.
Genant and his colleagues compared the image
quality of optical and radiographic magnification techniques and showed that for thin objects such as the
hahd, the former provides somewhat better images
(6,8,9). Both magnification techniques produce images
superior to those of conventional radiography. In a
recent study using optical magnification to examine 25
patients, Genant provided evidence suggesting that
this technique is more sensitive than conventional
radiologic methods for detecting erosions and soft
tissue swelling (10). However, the clinical utility of
radiographic magnification has had little evaluation. In
one study, 2 radiologists were asked to comment upon
the diagnostic information available in magnification
and conventional films of 212 patients (7). Magnification was thought to provide substantially more “helpful” information than codventional radiography in
arthritic disorders. However, in only 5% of the cases
was the additional iriformation considered “essential”
for diagnosis. In another study, Ishigaki fouhd that
50% of small bone defects noted in 28 patients with
gout were found only by using a 6x radiographic
maghification view (I 1).
In this stlldy we sought to determine which of 4
conventional views is most sensitive in detecting the
abnormalities associated with early rheumatoid arthritis, and whether radiographic magnification provided
additional sensitivity.
PATIENTS AND METHODS
All patients were seen ih a 96-bed unit specializing in
the care of rheumatologic and orthopedic patients. Criteria
for admission to the study were: diagnosis of probable or
possible rheumatoid arthritis, joint symptoms for less than 2
years, and no gross erosions on routine PA hand film.
Patients who met these criteria were examined radiagraphically using 5 different techniques: PA, Brewerton, oblique,
and reverse oblique views were taken of both hands and
wrists using standard technology (the “conventional” films,
Figures 1A-D), and a PA radiographic magnification film
was taken of the right hand and wrist, using an RSI microfocus tube with 0.09-mm focal spot. Patients were positioned
with hands palm down on a plexiglass platform. The targetfilm distance and object-film distance were set to achieve 2x
magnification. Trimax 2 screens and XUD film were used
with standard processing.
Each film was coded, masked, and presented in
random order to 2 radiologists specializing in bone and joint
radiology, both of whom have more than 10 years of
specialty experience. Because we wished to study the use of
these techniques in everyday reading conditions, the radiologists were not permitted to confer on their observations nor
to use standard comparison films. Each joint was inspected
for malalignment, erosions, joint space narrowing, and soft
tissue swelling. Twenty-three joints on each hand and wrist
were evaluated for the first 3 abnormalities, and 24 for the
fourth. Each joint was described as “normal,” “questionably abnormal,” or “definitely abnormal.” Computer outputs were checked against transcripts of the original dictations to insure accuracy of data entry.
Logistic multiple regression was used in the analysis
of the data for 2 reasons (12). First, observations ofjoints on
the hands of the same patient cannot be truly independent,
and therefore, a model that adjusts for this lack of independence is required. Second, we wished to assess the effect of
each technique upon the detection of positive finding: but
knew that differences between right and left hands, patients,
and particularly between observers might obscure this effect. Logistic multiple regression can be used to adjust for
these factors that are of lesser ihterest so that the effects of
radiographic view and magnification can be more clearly
evaluated. The basic model used for analysis expressed the
number of positive findings as a function of 4 independent
variables: hand (left or right), radiologist ( 1 or 2), patient (1
through 54), and technique (the 4 named above). The “standardized logits” used in the figures express the proportions
of all joints found to be positive after transformation re-
Table 1. Numbers of questionably abnormal (QA) and definitely abnormal (DA) joints according to radiologic technique and abnormality as
read by radiologist 1
Abnormality
Erosions
View
Posteroantetior
Oblique
Reverse oblique
Brewerton
Radiographic
magnification*
Alignment
Joint space narrowing
Soft tissue swelling
QA
DA
QA
DA
QA
DA
QA
DA
I79
149
I27
I24
39
36
24
55
52
50
31
16
9
5
10
I79
I46
89
125
175
I28
67
97
200
204
I42
152
I25
126
52
77
84
78
23
25
87
70
171
80
14
* Since radiographic magnification views were taken only of the right
hand, there were half as many joints examined per film
Total number
of films
54
54
41
54
54
RADIOGRAPHIC EVALUATION OF RA
747
Table 2. Numbers of questionably abnormal (QA) and definitely abnormal (DA)joints according to radiologic technique and abnormality as
read bv radioloaist 2
Abnormality
Erosions
View
Posteroanterior
Oblique
Reverse oblique
Brewerton
Radiographic
magnification *
~~
Alignment
Soft tissue swelling
Joint space narrowing
QA
DA
QA
DA
QA
DA
QA
DA
Total number
of films
29
39
39
51
13
II
28
33
8
I
3
0
26
13
7
9
75
101
24
IS
I19
88
43
88
57
52
30
39
13
41
26
40
54
54
54
54
I
20
30
89
51
so
54
23
12
~~
~~
* Since radiographic magnification views were taken only of the right hand, there were half as many joints examined per film.
quired to comply with the assumptions of linear modeling.
At 0 the logit equals a proportion of 0.5, while negative
numbers are less and positive numbers are more than a
proportion of 0.5.
Separate analyses were conducted for each abnormality, counting either questionably abnormal or definitely
abnormal joints. The goodness of fit of the logit models to the
data was assessed by examining deviance after fitting the
model (deviance is -2 times log-likelihood and is distributed
asymtotically as chi-square) (13). The estimates provided by
the model for the various radiologic techniques were compared statistically by using 2-tailed [-tests (12). Unless
otherwise stated, all significance tests were conducted at the
1% level, thereby allowing for multiple testing. Also determined was a measure of observer agreement for each
technique. This was calculated as the sum of the numbers of
joints which the radiologists agreed were normal, questionably abnormal, o r definitely abnormal, divided by the number of joints read.
were women whose mean age was 55.4 years (range
28-79); the 11 men also had a mean age of 55.4 years
(range 36-71).
Comparison of radiologists. Tables I and 2 present the 2 radiologists' findings using each technique.
Radiologist 1 consistently recorded more findings,
questionable and definite, regardless of the technique
or abnormality being evaluated. Interobserver agree-
;I
b Qucrtiomblc Erosions
s'-2.5
RESULTS
-3.0
Forty-one patients had complete sets of 5 films
read by both radiologists; 13 had 1 film missing (the
reverse oblique). Analyses were also performed using
only the 41 patients with complete sets of films; the
conclusions were unchanged. Forty-three patients
.3.5
01
-2.5
;-.5
-3.0
$4.0
.-1.5
PA
0.
no
I"
Deviances for various models*
Full
Abnormality
Alignment
Erosions
Joint space narrowing
Soft tissue swelling
Degrees of freedom
9 -4.0
s'
Without
technique
NO
variables
296
625
952
782
890
* Distributed asymtotically as x2. All comparisons between models
for a particular abnormality significant at P < 0.001.
t Full = radiologist + radiologic view + hand + patient; without
technique = radiologist + hand + patient.
;
-3.5
-5.0
84 I
1,204
3,412
2,026
945
I
I -3.0
j -4.5
.s
242
568
900
139
881
.Y
..
-3.5
Modell
g: Definilc loin1 Space Narrowing
-2.0 I
-3.5
Table 3.
I
L
c: Qucrliorvblc loin1 Space Narrowing
01
I0
.
-4.0
.A
"M
I
1
0.
10
I.
IH
I
Figure 2. Comparison of sensitivities of techniques. Sensitivities of
each technique in detecting each of the radiographic abnormalities
expressed as standardized logits of the proportion of joints found
positive. (Taller bars indicate greater sensitivity.) PA = posteroanterior; OB = oblique; RO = reverse oblique; BK = Brewerton; RM
= radiographic magnification. **P< 0.001 compared with PA view.
' P < 0.01 compared with Brewerton view (examined for definite
erosions only).
HARTLEY ET AL
748
7 1
DIP
PIP
MCP
WRIST
Figure 3. Comparison of sensitivities of radiographic techniques at
different joints. Percent of joints demonstrating dcfinite erosions
according to location and technique. DIP = distal interphalangeal;
PIP = proximal intcrphalangeal; MCP = metacarpophalangeal;
wrist = midcarpal, radiocarpal, and ulnocarpal: PA = posteroantcrior; OB = oblique; RO = reverse oblique; BR = Brewerton; RM =
radiographic magnification.
ment ranged from 81 to 92% on the 5 views for the 4
abnormalities. To some extent, the low prevalence of
abnormalities masks the degree of interobserver variation in recording questionable and definite findings.
Detailed analysis of this variation will be the subject of
a future report.
Comparison of radiologic views. Linear modeling techniques were employed to analyze the ability of
the various techniques to detect each of the 4 radiographic abnormalities. Table 3 presents the chisquares (x’) and the degrees of freedom (df) for the
models. The “full” model analyzed the number of
findings on each film as a function of 4 independent
variables: radiologist, hand (left or right), patient, and
technique. The “without technique” model did not
include this last variable. By comparing the x’ of the
“full” models used for our analysis with their respective df, it may be seen that the models adequately
describe the data (the x2 are about the same or less
than their df). Comparing the “full” and “without
technique” models shows that the “technique” parameter is highly significant in the additional variation
it explains. This indicates that significant differences
exist among the radiologic techniques in the number of
findings noted on them. Results are presented only for
definite findings; similar values were also obtained for
readings scored “questionable.”
Figures 2a-h present the results separately for
each radiologic abnormality and the 2 scorings, questionably or definitely abnormal. For most of the abnor-
A
Figure 4. Two views of rheumatoid erosions. A, The posteroantcrior view shows erosion of the proximal phalanx of the index finger and a
small cyst-like lucency in the proximal phalanx of the third finger (3). B, The magnification view shows the erosion of the index proximal
phalanx (arrow) to advantage.
749
RADIOGRAPHIC EVALUATION OF RA
magnification were the superior views for detection of
erosions about the metacarpophalangeal and proximal
interphalangeal joints. This point is demonstrated in
Figures 4 and 5 .
DISCUSSION
Figure 5. The Brewerton view of the hand shown in Figures 4A and
B (note order of fingers reversed) shows an erosion of the proximal
phalanx of thc third finger (3) that was not demonstrated on the
other views (arrow). A probable erosion of the third metacarpal
head, not previously evident, is also present in this view.
malities, the relative sensitivities of the techniques are
similar for both questionable and definite findings.
Comparing conventional techniques, more
questionable and definite alignment, soft tissue, and
joint space abnormalities were noted on the standard
PA film than on any of the other 3. In many instances,
these differences were statistically significant despite
the relatively small number of total findings. However,
for erosive disease the PA was 1 of the 2 least sensitive
views. The Brewerton view, which was the least
sensitive for 2 of the other 3 abnormalities, was
significantly better than the PA in the detection of
erosions (Figure 20.
Compared with the conventional films in the
detection of alignment, joint space, and soft tissue
abnormalities, the magnification technique showed
results most like the standard PA. However, it was
much better than the PA in detecting definite erosive
changes and also was significantly better than the
Brewerton (Figure 20. Figure 3 compares the sensitivities of the techniques at different joints. When compared with the PA, the Brewerton and radiographic
The methodology of this study enabled separate
analyses to be made of the sensitivities of each view
and of radiographic magnification in detecting the 4
abnormalities that are typical of early RA. Comparison
of the PA magnification film with the conventional PA
film isolates the effect of magnification while the
effects of the different views can be determined by
comparison of the conventional films. Our results
show that the usefulness of each of the conventional
views depends upon the location and type of radiographic abnormality being sought. They also demonstrate that radiographic magnification has the ability to
provide additional information, though not about all
types of abnormalities.
For the detection of malalignment, joint space
narrowing, and soft tissue swelling, the PA is the best
conventional view. The others-the oblique, reverse
oblique, and particularly the Brewerton-diminish the
radiologist’s ability to properly view fat planes and
joint spaces, thereby making these views less able to
demonstrate these particular abnormalities. Radiographic magnification added little further information.
Therefore, for malalignment, joint space narrowing,
and soft tissue swelling, the view, rather than the use
of magnification, determines sensitivity.
Our findings suggest that different radiologic
abnormalities, even those occurring in the same anatomic areas, may be best evaluated by different views.
In this study, the PA was one of the least sensitive
views in the detection of erosions. The Brewerton and
reverse oblique, the 2 least sensitive in evaluating the
other 3 abnormalities, performed better than the PA in
the detection of erosive disease. Similarly, Norgaard
has argued that the reverse oblique is more sensitive in
detecting erosions (14,15), although others have disputed that claim (16,17). Our findings would also
indicate that the PA may not be the single best
conventional film for detecting erosions. The anatomic
distribution of erosions noted on the Brewerton suggests that, compared with the PA, it enables better
visualization of the joints most classically involved in
RA .
In contrast to the other 3 abnormalities, both
750
the particular view and the use of magnification appear
important in determining sensitivity to erosive disease.
Our magnification film was a PA view and was more
sensitive than all the conventional films, including the
conventional PA, in detecting erosions. Using optical
magnification technique, Genant similarly demonstrated it to be more sensitive than conventional PA and
oblique films in detecting erosive disease (10). The use
of magnification, therefore, more than offset the disadvantage of the PA view which, as indicated by comparison of the conventional views, was not the optimal
view for detecting erosions. It is quite possible that
magnification films taken in Brewerton or reverse
oblique views may perform even better than our PA
magnification film in detecting erosions.
Two further comments should be made. First,
in comparative studies such as this, no independent
verification of the findings exists. Since it is not known
how many patients presenting with possible or probable RA may have abnormalities, especially on a magnification film, the use of control films of “normal”
subjects may aid in ascertaining whether the falsepositive rate is likely to be high. However, this control
still would not allow for the calculation of indices of
test performance. In this study, we have assumed that
findings scored as questionably or definitely abnormal
reflect the ability of specific techniques relative to one
another to demonstrate the abnormalities sought. Second, we provide evidence that particular techniques
provide more information. Whether diagnosis or clinical management would be altered by this information
has not been determined and deserves further study.
For the practicing physician having to decide
what radiographs to order, this study helps clarify the
options. If radiographic magnification is either
unavailable or the added radiation exposure thought
inappropriate, then the PA is clearly the single best
screen for the 4 abnormalities we have studied. The
addition of a Brewerton, or possibly a reverse oblique,
may enhance the ability to detect erosive disease,
though at the cost of additional radiation and expense.
If only erosive disease is of interest, then we have
found the Brewerton to be the single best view.
The decision to order radiographic magnification instead of a conventional film depends upon the
importance of detecting subtle erosive changes. In
specific situations, such as evaluating whether to use
possibly toxic drugs, the detection of such erosions
may be quite important and thereby justify the use of
radiographic magnification. However, another radio-
HARTLEY ET AL
graphic sign of destructive disease, joint space narrowing, was several times more common than erosions in
our patients and was detected equally well by both PA
and image magnification techniques. Soft tissue swelling and juxtaarticular demineralization were better
detected by image magnification but are findings that,
by themselves, rarely affect therapeutic decisions.
Therefore, weighing the benefits of radiographic magnification against its added cost and radiation exposure
would, in many clinical situations, favor the continued
use of conventional techniques.
ACKNOWLEDGMENTS
We are indebted to Holly Fossel, Karen Cullen,
Mary Scamman, and Holley Eaton for data entry, to Evelyn
Cone for secretarial assistance, and to Kristin Mortimer for
editorial comments.
REFERENCES
1. Brook A, Corbett M: Radiographic changes in early
rheumatoid disease. Ann Rheum Dis 36:71-73, 1977
2. Brook A, Fleming A, Corbett M: Relationship of radiological change to clinical outcome in rheumatoid arthritis. Ann Rheum Dis 36:274-275, 1977
3. Martel W, Hayes JT, Duff IF: The pattern of bone
erosion in the hand and wrist in rheumatoid arthritis.
Radiology 84:204-214, 1965
4. Weissman BN, Sosman JL: The radiology of rheumatoid
arthritis. Orthop Clin North Am 6:653-674, 1975
5. Mall JC, Genant HK, Silcox DC, McCarty DJ: The
efficacy of fine-detail radiography in the evaluation of
patients with rheumatoid arthritis. Radiology 112:37-42,
1974
6. Genant HK, Doi K: High-resolution skeletal radiography: image quality and clinical applications. Curr Probl
Diagn Radiol 7:2-62, 1978
7. Genant HK, Doi K, Mall JC, Sickles EA: Direct radiographic magnification for skeletal radiology. Radiology
123:47-55, 1977
8. Genant HK, Doi K , Mall JC: Optical versus radiographic magnification for fine-detail skeletal radiography.
Invest Radiol 10:160-172, 1975
9. Doi K, Genant HK, Rossmann K: Comparison of image
quality obtained with optical and radiographic magnification techniques in fine-detailed skeletal radiography:
effect of object thickness. Radiology 118:189-195, 1976
10. Genant HK: Methods of assessing radiographic change
in rheumatoid arthritis. Am J Med 75:35-46, 1983
1 1 . Ishigaki T: First metatarsal-phalangeal joint of gout:
macroroentgenographic examination in 6 times magnification. Nippon Acta Rad 335339454, 1973
RADIOGRAPHIC EVALUATION OF RA
75 1
12. Cox DR: Analysis of Binary Data. London, Methuen
and Company, 1970
13. Nelder JA, Wedderburn RWM: Generalized linear models. J R Statist Soc 135:370-384, 1972
cal changes in rheumatoid polyarthritis. Br J Radio1
53:63-73, 1980
16. Allander E, Brekkan A, Idbohrn H , Thorsteinsson J,
Olafsson 0, Sigfussort N , Sievers K: Is Norgaard’s
radiological sign for early rheumatoid arthritis reliable?
Scand J Rheumatol 2:161-166, 1973
17. De Smet AA, Martin NL, Fritz SL, Lindsley HB:
Radiographic projections for the diaghosis of arthritis of
the hands and wrists. Radiology 139577-581, 1981
14. Norgaard F: Earliest roentgenological changes in
polyarthritis of the rheumatoid type: rheumatoid arthritis. Radiology 85:325-329, 1965
15. Norgaard F: A follow-up study of the earliest radiologi-