The Xigris affair - "Mystery"

Cumulative 28-day mortality over time

"Mystery" questions that I would like to see answered:

1) Why is the cumulative mortality rate graph for placebo patients so flat (31 - 32%) for the entire duration of the second-half of the trial, considering that the trialists did not ensure that there was optimization of antibiotic therapy among the 99+ treatment centres?

2) Why did the xigris-treated patients' graph climb slightly between June 1999 and September 1999, so that there is no statistical significance difference between xigris-treated and placebo patients by October 1999, and then why it did it tilt downwards so dramatically, and so relenlessly, after that date?

Background information on the PROWESS trial of xigris (recombinant human activated protein C):

The trial was designed to test whether rhAPC would decrease the mortality rate in septic patients. The trial consisted of 1520 patients, and the trial showed that 192 of the 768 (25%) rhAPC-treated patients and 236 of the 752 (31%) placebo-treated patients did not survive 28 days (primary stratified analysis p=0.0071, nonstratified p=0.0057).

In June 1999, a protocol revision was instituted, which had different exclusionary criteria from the original protocol.

Exclusionary criteria included (as worded by the trial investigators):-

• In an effort to increase the discriminatory patient population for this sepsis trial the following exclusion criteria have been added: Patients who have undergone bone marrow, lung, liver, pancreas or small bowel transplantation; patients who are moribund and where death is imminent; patients whose family and/or primary physician have not committed to aggressive management of the patient; and patients with acute clinical pancreatitis without a proven source of infection.
• In an effort to increase the discriminatory patient population for this sepsis trial, exclusion criterion has been further defined along with a requirement that the VCC approve the enrollment of any patient with known or suspected metastatic cancer.
• Exclusion criterion has been added to clarify Window I of the entry criteria. This clarifies the duration of sepsis-induced organ failure which makes a patient ineligible for the trial.

In August 1999, the sponsor introduced a change in the manufacturing of the drug. The original manufactured drug was referred to as Bulk Drug Substance (BDS) BDS2 and the newly manufactured drug BDS2+. According to Lilly & Co., the drug's manufacturer, there was no difference in efficacy between the original and revised formulation of the drug.

A significant difference in mortality was observed between the first and second half of the study as displayed in the above figure. The mortality rates for the 720 patients enrolled under the original protocol were 30% for placebo and 28% for rhAPC. Under the revised protocol, the mortality rates for the 970 enrolled patients were 31% for placebo and 22% for rhAPC. The mortality rates for rhAPC for the original and newly manufactured product were 29% compared to 19%. The implementation of this manufacturing change occurred about half way through the study, at about the same time as the changes in the protocol.

Overall, an analysis of the two protocol versions showed that under the amended version of the protocol there were fewer patients: with malignancies, experiencing non-sepsis deaths, with chronic APACHE II health points, who had life support withdrawn, who were immunocompromised (including patient on chemotherapy and radiotherapy), at a nursing facility prior to entry, and with disabilities and more patients were at home prior to the onset of sepsis.

Commentary, criticism and controversy:

Insightful comments from readers will be included in this section.