C.W. Colwell Jr., S.D. Berkowitz, J.R. Lieberman, P.C. Comp, J.S.
Ginsberg, G. Paiement, J. McElhattan, A.W. Roth, and C.W. Francis
reply:
We appreciate the opportunity to address the questions raised by Dr. Lotke
and colleagues with respect to our paper. Their major concern is whether our
conclusions accurately reflect the EXULT B total knee arthroplasty data.
The conclusion that ximelagatran demonstrates superior efficacy (as defined
by our prespecified end point of total venous thromboembolism and mortality
from all causes) is valid. There was a significant difference in the efficacy
end point between ximelagatran and warfarin. The use of total venous
thromboembolism including venographically detected asymptomatic deep venous
thrombosis as a surrogate for the efficacy of an antithrombotic has been well
accepted by clinicians and regulatory authorities. Venous thromboembolism can
be thought of as a "pyramid," with asymptomatic calf deep venous
thrombosis being the most common (at the bottom of the pyramid) and with fatal
pulmonary embolism being very infrequent (at the top of the pyramid). In
between are symptomatic and asymptomatic calf and proximal deep venous
thrombosis and nonfatal pulmonary embolism. In the EXULT B study, all subgroup
analyses of venous thromboembolism, including asymptomatic calf and proximal
deep venous thrombosis, symptomatic deep venous thrombosis, and pulmonary
embolism, favored ximelagatran. For asymptomatic proximal deep venous
thrombosis, the rates were 3.1% for ximelagatran-treated patients and 3.4% for
warfarin-treated patients, an absolute difference of 0.3%. For symptomatic
deep venous thrombosis and symptomatic pulmonary embolism, the rates were 0.7%
and 0.2%, respectively, for ximelagatran-treated patients and 1.3% and 0.4%,
respectively, for warfarin-treated patients (absolute differences of 0.6% and
0.2%, respectively). When pooled together in the prespecified end point, the
difference in favor of ximelagatran is significant.
Lotke et al. question our analysis of death in the EXULT B study. In our
study, three (0.3%) of 1151 patients in the ximelagatran group died and none
of them had an autopsy. We conservatively categorized them as patients in whom
pulmonary embolism could not be excluded to try to capture and not hide
events, although two of these deaths were listed as myocardial infarction by
the investigator on the basis of electrocardiographic and clinical findings;
one occurred while the patient was receiving ximelagatran and one occurred
during the follow-up period when patient treatment was not controlled. The
third death also occurred during the follow-up period, and the cause was
listed by the investigator as unknown. Studies over the last ten years have
indicated rates of fatal pulmonary embolism following total joint replacement
ranging from 0% to
0.5%1; therefore,
even in the worst-case scenario, a 0.3% rate of fatal pulmonary embolism is
well within the reported range. No confirmed fatal pulmonary emboli occurred
during this study of 2303 patients undergoing total knee arthroplasty. Cardiac
deaths have not usually been reported. It is not biologically plausible for a
drug to be more effective in preventing all of the nonfatal venous
thromboembolism events described above and less effective in preventing fatal
pulmonary embolism. Since the nonfatal events were all verified with objective
diagnostic testing followed by adjudication by an independent central
committee of experts, it is highly likely that the assessment is accurate.
The opinion of Dr. Lotke et al. that the "9.7% diminution in the
incidence of distal deep venous thrombosis in the ximelagatran group [is] a
finding... of dubious importance" is not one shared by investigators
worldwide. No doctor can guess in which patients distal clots will become
clinically important. Ximelagatran did decrease the overall deep venous
thrombosis rate, with most clots being distal; this reflects the findings of
previous thrombophlebitis prevention studies that demonstrated clinically
important reduction in clots distal to the knee, not proximal to the knee. In
a meta-analysis of total knee arthroplasty, the rate of distal deep venous
thrombosis was reported to range from 21% to 71% and the rate of proximal deep
venous thrombosis was reported to range from 1.7% to
12.8%2. Although
proximal deep venous thrombosis is more important with respect to pulmonary
embolism and potential pulmonary embolism deaths than distal deep venous
thrombosis is, distal deep venous thrombosis is associated with a 20%
propagation rate3.
Many of the long-term complications of deep venous thrombosis, such as chronic
venous insufficiency and postthrombotic syndrome, are secondary to distal deep
venous thrombosis. However, in an individual surgeon's practice, they occur
too uncommonly for him or her to note. When long-term sequelae occur, they
rarely come to the attention of the orthopaedic surgeon for care but rather
are treated by other physicians. Therefore, we do consider this reduction in
events to be of clinical importance to both physicians and patients.
The rate of bleeding complications in total knee arthroplasty studies has
ranged from 8.6% to 18.9% and the rate of major bleeding complications has
ranged from 0% to 2.4% with no significant differences being detected between
ximelagatran and
warfarin2. In our
study, the rates of bleeding complications (5.0% for ximelagatran and 3.8% for
warfarin) were below the lower range reported in other studies and the rates
of major bleeding complications were within the range reported in other
studies, with no detectable significant differences between the two drugs. In
the abstract, we stated that "major bleeding was noted in 1% (twelve) of
the ximelagatran-treated patients and in 0.4% (five) of the warfarin-treated
patients (p = 0.09)." In such a large total knee arthroplasty trial
where the number of major bleeding events is so low (and therefore the major
bleeding rates are low), saying that the major bleeding event rate with
ximelagatran is "more than twice that of the warfarin group" is
misleading. For this statement to be accurate, the denominator would have to
be small and the numerator large, which is the opposite of our findings, where
the denominator (the number of patients) was large and the numerator (the
number of bleeds) was small. The major bleeding rates were not significantly
different, and the absolute difference of 0.6% (1.0% compared with 0.4%) would
not be clinically perceptible to physicians.
Dr. Lotke and colleagues asked for more parameters of patient evaluation,
and we concur that this would have been appropriate. For example, once a
bleeding event was documented, it would have been more helpful to clinicians
if we had done a more complete follow-up of all events to their final clinical
outcome. However, as with the previous 45,000 patients who have been studied
worldwide either in a prospective, randomized manner or merely in cohort
studies, this was not fully accomplished.
Multiplying end point rates in both arms as a way to "enlarge"
a clinical trial to provide more precision around the point estimate of the
events is counter to the scientific principles of good clinical trial analysis
and does not provide more certainty around the differences in end points. In
reality, the differences in the end points could either be magnified or
diminished. The suggestion that if a greater number of patients had been
studied, a significantly higher rate of patient-related complications and
adverse events might have occurred may be true, but there is an equal chance
that it would be untrue. We reported what we found in this study with the
point estimates, confidence intervals, and significance included. For a little
more insight, when we combine the numbers in the group dosed with 36 mg in the
EXULT A study4 with
those in the EXULT B study, it provides a >50% increase in patients, but
the percentages of patient-related complications and adverse events do not
change.
In the abstract, we concluded that the new drug "demonstrates
superior efficacy compared with warfarin prophylaxis" on the basis of
the prespecified end points and prespecified statistical analysis. We believe
in the validity of our conclusions and that the facts presented justify
them.