WASHINGTON _ Almost a year and a half ago, when Centocor Inc.halted the Centocor: HA-1A Efficacy in Septic Shock (CHESS) trialdue to excess deaths in patients receiving HA-1A, the company'smarket capitalization dropped $440 million in a single day and thenews sent shock waves through the entire biotechnology industry.The results, although not announced in detail back in early 1993, raisedthe ugly possibility that a drug intended to benefit patients had harmedthem. Data from this infamous clinical trial appears in the July 1 issueof the Annals of Internal Medicine.It is somewhat unusual for researchers to publish "negative" results(i.e., results from an agent that did not prove effective). An editorialaccompanying the article praised its authors for "presenting data fromthis essentially negative study that until now was widely discussed onlyin the press."HA-1A is a monoclonal antibody which was developed to targetendotoxin, a component found in the walls of gram-negative bacteria.Other types of bacteria, such as gram-positive bacteria, don't containendotoxin.The critical complication of targeting patients with gram-negativeinfections is that doctors cannot determine the source of infectionbefore treating a patient. Clinical trials targeting gram-negative patientsthus must enroll a much broader population of septic patients in orderto "hit" the target.(This fact did not go unnoticed by Wall Street _ a drug which wouldhave to be widely administered even though almost half of therecipients stood no chance to benefit could potentially generate hugerevenues.)In the CHESS trial, HA-1A was given to a total of 2,199 patients withseptic shock. Only 28 percent of those enrolled fell into the "targetpopulation," patients with confirmed gram-negative bacteremia.Among those patients, 293 received placebo and 328 received HA-1A.The 14-day mortality rate was 32 percent (95 patients) in the placebogroup and 33 percent (101 patients) in the HA-1A group. Thedifference was not statistically significant, with a p-value of 0.864 (a p-value must be below 0.05 to establish a statistically significantdifference between two agents).The CHESS trial's designers estimated they would need to enrollbetween 3,400 to 7,500 patients in order to get 1,500 target populationpatients, the amount needed to prove efficacy. But the CHESS trial washalted well before that point.The reason had to do with what happened to the non-target populationpatients. Malvern, Pa.-based Centocor had to prove that HA-1A did notcause any harm in the large number of patients who received it whohad no chance to benefit. And it was here that the explosive resultsended HA-1A's development as a treatment for sepsis.A total of 1,578 CHESS trial patients did not have gram-negativebacteremia. Of those, 793 received placebo and 785 received HA-1A.The mortality rate was 37 percent (292 patients) in the placebo groupand 41 percent (318 patients) in the HA-1A group. The difference hada p-value of 0.073.Although the p-value of 0.073 is above the standard threshold ofstatistical significance (0.05), it is below the trial's built-in "stoppingcriteria" of 0.1 for the non-gram-negative patient population. Thequestion burning in the minds of many researchers is why the trial wasdesigned with such a stringent stopping rule.Company representatives weren't available to comment due to the July4 holiday. In the Annals article, the authors note only that "the level ofsignificance used to stop the trial was greater than that usedconventionally in significance tests of differences between treatmentgroups in a clinical trial."Several researchers and statisticians suggested that the FDA requiredthe company to use the stringent rule due to its concerns about thesafety of HA-1A. A study in a canine model of septic shock done byresearchers at the National Institutes of Health and published last yearin the Journal of the American Medical Association showed that HA-1A killed more dogs than a non-specific control antibody or a placebo.In addition, questions about the reproducibility of early preclinical andclinical studies of HA-1A have been raised in numerous peer-reviewedjournals. As a result, the FDA may have been particularly watchful forany suggestion of harm.Tom Fleming, chairman of the Biostatistics Department at theUniversity of Washington in Seattle and a member of the FDA'sAdvisory Committee on Vaccines and Related Products, told BioWorldthat the notion of an FDA edict to Centocor was speculation. He addedthat Centocor should not have agreed to such a stringent stopping rule."They (Centocor) set themselves up to have a high chance of stoppingthe trial with negative results," he said. "The trial, as designed, did notprove convincingly that HA-1A was harmful, in fact the trial provednothing. These are what I would call equivocal results."Fleming said that it was proper to scrutinize HA-1A's effects in thenon-target patient population, but that the statistical methods used wereflawed and don't clearly establish whether the drug is harmful orneutral in those patients. Thus the CHESS data provide an inconclusiveepitaph for HA-1A, a drug which cost Centocor approximately $200million to develop.The Annals editorial offered this post-mortem on HA-1A and its one-time rival, Xoma Corp.'s E5: "The monoclonal antibodies werehastened into long, expensive clinical trials without an adequatepublished scientific record or, in the absence of a clinically relevant invitro or in vivo assay, a clearly formulated and demonstrablemechanism of action. Thus, the lack of data sufficient for licensure ofthese monoclonal antibodies for the treatment of sepsis should not havebeen unexpected." n
-- Lisa Piercey Washington Editor
(c) 1997 American Health Consultants. All rights reserved.