DOI:10.2214/AJR.08.1362
AJR 2008; 191:W191
© American Roentgen Ray Society
Reply
Elaine M. Caoili and
Jon A. Jacobson
University of Michigan Medical Center Ann Arbor, MI
WEB—This is a Web exclusive article.
We appreciate the insightful comments of Drs. Liao and Lin
[1] regarding the McNemar test
with respect to our article, "MDCT and Radiography of Wrist Fractures:
Radiographic Sensitivity and Fracture Patterns"
[2]. As the authors have
stated, the McNemar test is a nonparametric test best used for matched data
when nominal data (yes vs no, correct vs incorrect) are available. Readers in
our study were asked to determine whether the imaging findings showed a
fracture, no fracture, or equivocal findings for fracture. As Liao and Lin
have suggested, there are always alternative statistical tests that could be
used to analyze any data. A more elegant and perhaps even more appropriate
test would have been a statistical test for mixed models that would have taken
into account the number of readers and their responses as well as the repeated
measures taken for each patient in our study.
Nevertheless, we performed the McNemar test, as suggested, comparing
radiographic reports and CT findings. The results for the asymptotic test are
scaphoid (p = 0.003), lunate (p = 0.32), triquetrum
(p = 0.41), hamate (p = 0.71), and capitate (p =
0.65). The results for the exact test are similar to those in the letter from
Liao and Lin [1]. None of that
data would have changed from the original manuscript. The data that would have
changed according to the McNemar asymptotic test are trapezium (p =
1.0), ulna (p = 0.16), and metacarpal (p = 0.15). Results
for the exact test are similar to the letter. However, note that no discordant
pairs were found in the data for both pisiform and radius; thus, the McNemar
test (either asymptotic or exact) could not be applied. In addition, the data
for the trapezoid were unsuitable for the McNemar test as well; no fractures
of the trapezoid were identified on radiography, but three of 61 CT scans did
show a fracture.
Again, we thank Liao and Lin for their discussion; we appreciate that it is
often difficult to reproduce results without the original data.
References
- Liao YY, Lin YM. McNemar test is preferred for comparison of
diagnostic techniques. (letter) AJR 2008;191
: W190[Free Full Text]
- Welling RD, Jacobson JA, Jamadar DA, Chong S, Caoili EM, Jebson
PJL. MDCT and radiography of wrist fractures: radiographic sensitivity and
fracture patterns. AJR 2008;190
: 10–16[Abstract/Free Full Text]
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
