Our remarks will take the form of a brief summary of the safety and efficacy data for each of the three indications sought by the sponsor. Thereafter we will discuss the risk management program proposed by the sponsor. In summary, we find that the data submitted by the sponsor fail to establish the drug’s efficacy for two of the three indications sought and that the company’s proposed risk management strategy is inadequate to optimally reduce the risks, particularly to the liver, associated with use for the third indication.
Prevention of venous thromboembolism (VTE) in patients undergoing knee replacement surgery
The sponsor’s data indicate that in both of the two principal studies submitted for this indication, high dose ximelagatran (36mg twice a day) failed to reduce the incidence of symptomatic proximal or distal deep venous thrombosis (DVT) compared to warfarin. There is an impressive-appearing reduction in the incidence of asymptomatic DVT over the 7-12 day study period (19.8% for ximelagatran vs. 27.4% for warfarin in the study known as Exult A and 21.9% vs. 31.4% in Exult B), but closer inspection reveals that essentially all of the reduction occurred in asymptomatic distal DVTs diagnosed by venography (19.2% vs. 26.7% in Exult A and 21.4% vs. 31.1% in Exult B). As the medical officer concludes, this “is not clinically meaningful.”
In addition, although two drugs (injectable enoxaparin and fondaparinux) are approved for this indication, the sponsor chose to compare ximelagatran to warfarin, which is not approved for this indication. Because warfarin takes 3-5 days to reach therapeutic levels, compared to hours for ximelagatran, “the comparison is unfair,” in the words of the medical officer, particularly because the study lasted only 7-12 days and many warfarin patients were not adequately anticoagulated. Moreover, any convenience advantage over the approved medications conferred by ximelagatran’s being an oral medication is diminished in this short-term, substantially inpatient setting.
Safety issues were of less concern than for longer-term uses of ximelagatran, but serious concerns remain. Major bleeding was more common among high-dose ximelagatran-treated patients (0.9% vs 0.5%; p=0.18; chi-square) as were coronary artery disease events (0.75% for all ximelagatran patients vs. 0.26%; p=0.028). The medical officer described the latter as “unexpected and worrisome.” The rate of alanine aminotransferase (ALT) elevations greater than three times the upper limit of normal (ULN) was 2.1% for high-dose ximelagatran vs. 1.4% for warfarin, but the greater concern is that 0.27% of ximelagatran patients and 0.04% of warfarin patients had similarly elevated ALTs 4-6 weeks after the operation, well after prophylactic therapy had stopped (p=0.09; Fisher’s Exact). There are no longer-term data on potential hepatotoxicity for this indication.
In sum, ximelagatran’s efficacy appears to be limited to asymptomatic distal DVTs, and even then only after comparison to a drug unapproved for this condition. Given concerns about the propensity of the drug to induce bleeding, cardiovascular events and ALT abnormalities, and the existence of other approved medications for this condition, we believe ximelagatran should not be approved for this condition.
Long-term secondary prevention of VTE after standard treatment for an episode of acute VTE
The sponsor has provided only a single 18-month placebo-controlled study (THRIVE III) with a decidedly modest 73% follow-up rate to support this indication. Although ximelagatran (24mg twice a day) failed to reduce all-cause mortality (the death rate was low; 1.1% in the ximelagatran group vs. 1.4% in the placebo group), it showed impressive reductions in both symptomatic VTEs (12 vs. 71) and pulmonary emboli (2 vs. 23). Ximelagatran does indeed appear to be superior to placebo for this condition.
We next discuss the safety data for this and the third indication together, as the medical officer often combined the data from the trials for the two longer-term indications.
For all patients receiving ximelagatran on a long-term basis (>35 days), 7.6% developed ALTs greater than three times the ULN, compared to 1.1% of patients receiving comparators (relative risk = 7.0; 95% confidence interval 5.45-8.99). The increase typically occurred between months one and six of the trial and usually reverted to normal, even with continued drug therapy. It is noteworthy that in trials generally lasting six months (by which time ximelagatran-induced hepatotoxicity was usually apparent), troglitazone (Rezulin) had only a 1.9% incidence of ALTs greater than three times the ULN, compared to 0.6% on placebo. That drug was eventually banned after 94 cases of liver failure, most of them fatal.
More serious hepatotoxicity (ALT > three times the ULN and bilirubin > twice the ULN) was observed in 37 patients taking ximelagatran compared to five taking comparators (0.53% vs. 0.08%; relative risk = 6.64; 95% confidence interval 2.61-16.87). Nine of the 37 died and three of these were judged by the medical officer to be related to ximelagatran, a rate of one case per 2316 patients (mean treatment of 357 days). This rate will likely be higher in clinical practice as patients excluded from the clinical trials are treated with ximelagatran. Importantly, 4.3% of patients had ALTs greater than five times the ULN (0.4% for comparators) and 1.5% had ALTs greater than ten times the ULN (0.1% for comparators).
Patients receiving ximelagatran for secondary prevention of VTE had similar incidences of major bleeding events as did those in the placebo group. For the atrial fibrillation indication (see below), the ximelagatran bleeding rate was lower than for the warfarin comparators, but did not reach statistical significance. For both indications, coronary artery adverse events were somewhat higher for ximelagatran than comparators (the same comparison was statistically significant for post-knee replacement prophylaxis), a finding the medical officer again describes as “worrisome.”
In sum, while the drug appears to be effective for this indication, the risks are significant. Below we discuss how the sponsor’s proposed risk management strategy is inadequate.
Prevention of stroke and other thromboembolic complications associated with atrial fibrillation
The sponsor submitted two non-inferiority trials comparing ximelagatran (36mg twice a day) to warfarin, one double-blinded (SPORTIF V) and the other open-label (SPORTIF III). These studies are both seriously flawed because the manufacturer pre-specified a non-inferiority margin of 2%. This means that if the annual incidence of stroke/systemic embolic events in the ximelagatran group was under 5.1% (2% + the 3.1% the company assumed as the rate in the warfarin group), the drug could be considered non-inferior to warfarin. As the medical officer puts it, ximelagatran could be “only half as effective as warfarin and still be considered “non-inferior.” The medical officer asserts that this 2% margin “was not agreed to by the Agency” and, moreover, that “the margin chosen was too liberal.”
The two studies, despite being similarly designed, had divergent results: in the open-label study ximelagatran fared better than warfarin (annual rate 1.64% vs. 2.29%; relative risk = 0.71; 95% confidence interval: 0.48-1.06), but in the double-blinded study ximelagatran fared worse (1.61% vs. 1.16%; relative risk = 1.39; 95% confidence interval: 0.91-2.12). As the statistical reviewer states, “In general, the results from double blind studies are more reliable for many reasons.” He then goes further: “Unless the clinical judgment is that a loss of 2% of the effect of warfarin is clinically acceptable, in my opinion [ximelagatran] has not been demonstrated to be noninferior to warfarin.”
In view of the significant problems with the trial design and the toxicity associated with long-term use described above, approval of ximelagatran for this indication is not warranted.
The sponsor’s risk management program
The sponsor has proposed an ALT-monitoring program (RiskMAP) that is similar to that ultimately adopted in the clinical trials: baseline testing, followed by monthly testing for at least six months, with triggers for more frequent monitoring or discontinuation with specific ALT levels above the ULN. Cases of severe liver injury, including one fatal one, were observed even with the relatively high level of 63% compliance with this algorithm. Outside of the artificial clinical trials setting, compliance is likely to be lower. For example, by just the third month of therapy, fewer than 5% of troglitazone patients had received the recommended liver enzyme monitoring, despite numerous reminders from the FDA. FDA’s Office of Drug Safety concludes, “we do not agree that the sponsor’s proposed RiskMAP [program] is adequate … Currently, the proposed monitoring plan provides no guarantee of safeguarding the patient from developing a rapid onset and life-threatening reaction.”
The risk management program also fails to address the risk of delayed hepatotoxicity associated with prophylactic use after surgery and makes no attempt to reduce the risks of bleeding or coronary artery disease.
A number of approaches that go beyond the RiskMap approach have been put forth for consideration by FDA personnel and should be implemented. These include a black box warning, a mandatory patient registry for long-term users that would be linked to performance, patient-physician agreements, and restrictions on promotion, distribution and packaging.
 This figure is consistent with the one case per 2000 patients predicted by Hyman Zimmerman’s model in which 10% of patients with this more serious hepatotoxicity are predicted to develop liver failure, transplant or death.
 This would occur if the event rate in the warfarin rate was lower than the predicted 3.1% (actually under 2%), as proved to be the case. The medical officer states that the “derivation [of the 3.1% figure] from referenced historical trials is unclear.”
 Graham DJ, Drinkard CR, Shatin D, Tsong Y, Burgess MJ. Liver enzyme monitoring in patients treated with troglitazone. Journal of the American Medical Association 2001;286:831-3.