Dc. Diffin et al., EFFECT OF ANATOMIC DISTRIBUTION OF PULMONARY EMBOLI ON INTEROBSERVER AGREEMENT IN THE INTERPRETATION OF PULMONARY ANGIOGRAPHY, American journal of roentgenology, 171(4), 1998, pp. 1085-1089
OBJECTIVE, This study examines the anatomic distribution of emboli on
pulmonary angiography and attempts to determine the relationship of ve
ssel size to interobserver agreement, two factors having important imp
lications in comparing pulmonary angiography with cross-sectional imag
ing for pulmonary embolism. MATERIALS AND METHODS. One hundred twenty-
five consecutive pulmonary angiograms were reviewed retrospectively by
three interventional radiologists. Initial interpretations were recor
ded and compared to determine interobserver agreement on a per-patient
and per-embolus basis. Discordant interpretations were reviewed by al
l radiologists for a consensus interpretation. RESULTS. Unanimous per-
patient agreement occurred in 91% (114/125) of initial interpretations
. The largest artery containing acute pulmonary embolism was segmental
or larger in 24 patients (83% of patients with acute positive finding
s, 19% of all patients) and subsegmental in only five patients (17% an
d 4%, respectively). On a per-patient basis, initial interobserver agr
eement averaged 45% and unanimous consensus agreement was achieved for
79% of patients having isolated subsegmental pulmonary embolism. Cons
ensus readings altered initial per-patient interpretations for 30% of
patients having only subsegmental pulmonary embolism; per-embolus inte
rpretations were altered for 37% of all subsegmental emboli. CONCLUSIO
N. Subsegmental emboli occurring as isolated findings are relatively r
are. Approximately one third of subsegmental emboli and one third of p
atients having isolated subsegmental emboli may be initially misdiagno
sed on pulmonary angiography. Objections to cross-sectional imaging fo
r pulmonary embolism based on the inability to detect subsegmental pul
monary embolism when compared with pulmonary angiography should be ree
xamined with this data in mind.