December 20, 2006

Northfield Laboratories has announced preliminary results from its pivotal Phase III trial assessing the safety and efficacy of PolyHeme, its human hemoglobin-based oxygen-carrying red blood cell substitute, in the treatment of severely injured and bleeding patients when blood is needed but not immediately available. However, because of discrepancies in the initial data, the database will be unlocked and corrected prior to finalizing the statistical analyses.

Northfield received the initial draft top-line data from its contract research organization (CRO) on last week. During its review and interpretation of these results, two discrepancies in the dates of death for patients in the study, in which mortality is the primary endpoint, were identified. The CRO was notified and agreed that these data points were inaccurate. It is necessary to unlock the study database to make the corrections.

To ensure absolute accuracy before the database is relocked, Northfield has requested verification by the CRO of all critical variables and has also initiated an independent verification. FDA has been notified, but the study results will not be submitted until final. Because of the interest in the study, however, Northfield has made the decision to report the preliminary results as provided.

The primary efficacy endpoint of the study is a dual superiority and noninferiority test of day 30 mortality in the modified intent-to-treat population. In order to achieve the superiority endpoint, the upper limit of the confidence interval (CI) surrounding the observed difference in mortality between the treatment and control groups needed to be less than zero; for the noninferiority endpoint, the upper limit of the CI needed to be 7 percent or less. The noninferiority boundary is based on the potential to provide a benefit in situations where transfusion of blood is indicated but not available.

The preliminary results received by Northfield indicate that in the primary modified intent-to-treat population (those who were randomized and received some treatment) the upper limit of the CI exceeded the 7 percent threshold by 0.3 percent. However, in the prespecified per-protocol population (those who both received the correct treatment and did not otherwise violate the protocol), the upper limit of the CI was below the threshold at 5.8 percent.