<?xml version="1.0" encoding="utf-8"?>
<recommendedContent xmlns="http://api.mspoke.com">
    <recommendedItem id="20100101_19_200"
                     title="Debate Surges on Composite Endpoints"
                     score="-0.003"
                     href="http://www.medpagetoday.com/PublicHealthPolicy/ClinicalTrials/tb/18046?impressionId=1265807328858"
                     
      &lt;p&gt;Composite endpoints can obscure the real findings of clinical trials, two researchers charged in a &lt;em&gt;JAMA&lt;/em&gt; commentary this week, but others who had led trials using such outcomes defended the practice.&lt;/p&gt;
&lt;p&gt;Composite endpoints  --  where a study&apos;s main outcome is a combination of two or more different types of events, such as death and nonfatal myocardial infarction  --  can serve useful purposes, George Tomlinson, PhD, and Allan S. Detsky, MD, PhD, both of the University of Toronto, wrote in the Jan. 20 &lt;em&gt;Journal of the American Medical Association&lt;/em&gt;.&lt;/p&gt;
&lt;p&gt;But the benefits of composite endpoints come with a price, they argued  --  confusion to physicians and patients.&lt;/p&gt;
&lt;p&gt;&quot;Clinicians want to know if all events in the composite outcome are affected equally by the intervention,&quot; Tomlinson and Detsky wrote.&lt;/p&gt;
&lt;p&gt;Physicians can usually find results for the endpoint&apos;s individual components, they acknowledged, but it may &quot;result in some confusion, because the component relative risks may have broad confidence intervals and differ widely, at times even extending in opposite directions.&quot;&lt;/p&gt;
&lt;p&gt;Moreover, the Toronto researchers argued, if readers must examine results of the individual components of the composite endpoint to grasp the study&apos;s clinical implications, it defeats the composite endpoint&apos;s original purpose.&lt;/p&gt;
&lt;p&gt;&quot;While [readers] were enticed by a trial performed according to rigorous principles and based on the primary composite outcome, once the results have been reported they find that their interest has been redirected to individual outcomes of questionable importance,&quot; Tomlinson and Detsky wrote.&lt;/p&gt;
&lt;p&gt;They acknowledged that composites may sometimes make clinical sense, or are necessary because no single outcome is a natural primary endpoint by itself. Another practical rationale is to reduce the number of patients necessary in a study to detect a significant treatment effect.&lt;/p&gt;
&lt;p&gt;For example, if an outcome is expected to occur at a 5% annual rate and the trial is planned to last five years, more than 2,500 patients are needed to establish a hazard ratio of 0.75 with &lt;em&gt;P&lt;/em&gt;&amp;lt;0.05, Tomlinson and Detsky noted.&lt;/p&gt;
&lt;p&gt;But if several outcomes can be combined into a composite endpoint that has an annual rate of 20%, fewer than 800 patients will provide adequate power.&lt;/p&gt;
&lt;p&gt;That&apos;s generally fine when the individual component events occur at approximately equal rates, are of similar seriousness, and change in the same way with treatment, but that is frequently not the case, Tomlinson and Detsky contended.&lt;/p&gt;
&lt;p&gt;Steven Nissen, MD, a Cleveland Clinic cardiologist, agreed in a phone interview with &lt;em&gt;MedPage Today&lt;/em&gt;.&lt;/p&gt;
&lt;p&gt;Noting that composites of death, heart attack, and stroke are common in cardiovascular therapy trials, &quot;one would argue that all three of those outcomes are fairly grave [and] involve death or permanent injury,&quot; he said, and thus can be appropriate to combine into a single outcome.&lt;/p&gt;
&lt;p&gt;But, he added, &quot;what if the combination of endpoints is illogical, where you&apos;re combining grave endpoints with endpoints that are much less serious.&quot; In that case, the composite is much more difficult to interpret and may actually mislead readers about the study&apos;s true findings, Nissen suggested.&lt;/p&gt;
&lt;p&gt;&quot;Composite endpoints are a necessary evil, but they have to be thought through very carefully,&quot; he said.&lt;/p&gt;
&lt;p&gt;One trial with composites of serious and not-so-serious outcomes was reported last month at the American Society of Hematology meeting.&lt;/p&gt;
&lt;p&gt;Presented by Jeffrey Carson, MD, of the University of Medicine and Dentistry of New Jersey in New Brunswick, N.J., it tested different postoperative blood transfusion volumes. (See &lt;a href=&quot;http://www.medpagetoday.com/MeetingCoverage/ASHHematology/17418&quot; mce_href=&quot;http://www.medpagetoday.com/MeetingCoverage/ASHHematology/17418&quot; target=&quot;_blank&quot;&gt;ASH: Lower Threshold for Post-op Transfusion Proves Safe&lt;/a&gt;)&lt;/p&gt;
&lt;p&gt;Its primary outcome was a combination of death and walking ability, and a secondary endpoint was a composite of death, myocardial infarction, infection, congestive heart failure, stroke, and venous thromboembolism.&lt;/p&gt;
&lt;p&gt;In a recent phone interview with &lt;em&gt;MedPage Today&lt;/em&gt;, Carson said composites are frequently chosen to reduce the necessary sample size, but in his study the rationale was more about finding an outcome that best reflected the clinical issue.&lt;/p&gt;
&lt;p&gt;Including death alongside less serious outcomes such as inability to walk or infections was intended to capture the possibility that a transfusion regime might improve the lesser outcome but increase mortality.&lt;/p&gt;
&lt;p&gt;&quot;You wouldn&apos;t want to declare that, well, it improves your chances of walking and not consider its impact on death,&quot; Carson explained.&lt;/p&gt;
&lt;p&gt;Another study with a composite endpoint was a 2006 study of rosiglitazone (Avandia) called DREAM. Its primary endpoint combined death with incidence of new-onset diabetes. (See &lt;a href=&quot;http://www.medpagetoday.com/Endocrinology/Diabetes/4115&quot; mce_href=&quot;http://www.medpagetoday.com/Endocrinology/Diabetes/4115&quot; target=&quot;_blank&quot;&gt;EASD: Avandia Prevents Progression to Diabetes in High-Risk Patients&lt;/a&gt;)&lt;/p&gt;
&lt;p&gt;Tomlinson and Detsky held up DREAM in the very first paragraph of their &lt;em&gt;JAMA&lt;/em&gt; commentary as an example of a questionable composite endpoint.&lt;/p&gt;
&lt;p&gt;&quot;Death and diabetes are quite far apart in the spectrum of severity,&quot; they wrote, suggesting that clinicians would find the outcome  --  a 60% reduction in the two events  --  hard to interpret.&lt;/p&gt;
&lt;p&gt;&quot;Two questions arise,&quot; they wrote. &quot;Was there a 60% reduction in both death and diabetes? Are the two outcomes just as likely to occur?&quot;&lt;/p&gt;
&lt;p&gt;Nissen agreed that DREAM well illustrated the problems that can arise from composite endpoints.&lt;/p&gt;
&lt;p&gt;&quot;I do not like the situation where endpoints that have a great deal of difference in gravity and seriousness are combined,&quot; he told &lt;em&gt;MedPage Today&lt;/em&gt;. &quot;This was a great example.&quot;&lt;/p&gt;
&lt;p&gt;But the lead investigator on DREAM defended the composite outcome in a phone interview with &lt;em&gt;MedPage Today&lt;/em&gt;, insisting that sample size or other cost-based rationales never entered into the decision to use it.&lt;/p&gt;
&lt;p&gt;&quot;I&apos;ll quote from the paper,&quot; said Hertzel Gerstein, MD, of McMaster University in Hamilton, Ontario. &quot;It was to account for the possibility that diabetes might develop at a different rate in individuals who died than in individuals who survived. . . . It had nothing to do with any suggestion that the drug might prevent death. In fact, what we did in the DREAM trial, it was explicitly designed to [enroll] people at low risk of having serious outcomes including death.&quot;&lt;/p&gt;
&lt;p&gt;Gerstein continued, &quot;It was designed that way in order to be careful that we did not overestimate the benefit of the drug and provide the most conservative estimate of the benefit of the drug on diabetes prevention.&quot;&lt;/p&gt;
&lt;p&gt;Nissen, however, argued that composites chosen for legitimate scientific reasons are subject to misinterpretation when the results are published or submitted to regulators.&lt;/p&gt;
&lt;p&gt;Consider the composite of death, myocardial infarction, stroke, or hospitalization for unstable angina or revascularization, a common endpoint in registration trials for cardiovascular drugs, he said. Very often it&apos;s the hospitalizations that dominate the composite outcome, as they are far more common than the more serious events.&lt;/p&gt;
&lt;p&gt;&quot;When these companies go to the FDA, they often ask for a label related to the composite outcome. &apos;This drug is approved to reduce the risk of death, heart attack, stroke, and hospitalization for revascularization.&apos; Is that a good regulatory decision or a bad regulatory decision?&quot; Nissen asked.&lt;/p&gt;
&lt;p&gt;Jeffrey Carson said such composites are frequently criticized, but they can be reported in such a way as to minimize the chance of misinterpretation.&lt;/p&gt;
&lt;p&gt;&quot;It&apos;s likely that readmission to the hospital is the predominant reason for an event [in the composite] in studies of that sort,&quot; he said. &quot;You shouldn&apos;t say that it affects mortality, and you shouldn&apos;t say it affects myocardial infarction. What you should say is that it looks like the predominant effect here is on readmissions.&quot;&lt;/p&gt;
&lt;p&gt;Tomlinson and Detsky suggested that one way around this problem would be for authors and readers to assign weights to the various components of a composite outcome to reflect their clinical importance, &quot;similar to the way quality of life is measured.&quot;&lt;/p&gt;
&lt;p&gt;Alternatively, they wrote, when a composite outcome is driven by effects on the most numerous but least severe component, it should be understood to have shown an &quot;effect on surrogate outcomes and not definitive ones.&quot;&lt;/p&gt;

    </recommendedItem>
    <recommendedItem id="20090101_19_3650"
                     title="Gabapentin Studies for Off-Label Uses Cooked? (CME/CE)"
                     score="-0.005"
                     href="http://www.medpagetoday.com/Neurology/PainManagement/tb/16952?impressionId=1265807328858"
                     
      Primary outcome measures were apparently a moving target in many company-sponsored studies of gabapentin (Neurontin) for off-label uses such as neuropathic pain and bipolar disorder, researchers have suggested.&lt;br&gt;
&lt;br&gt;Eight of 12 studies of gabapentin published over a 15-year span reported primary outcomes that differed from those specified in the original trial protocols, said Kay Dickersin, PhD, of Johns Hopkins University, and colleagues.&lt;br&gt;
&lt;br&gt;The researchers reviewed internal documents from Pfizer and its Parke-Davis unit on 20 gabapentin trials, which became public as a result of litigation against the firms. Eight of the studies were never published.&lt;br&gt;
&lt;br&gt;&quot;All the changes that took place between what was specified in the protocol, what was known before publication (as presented in the internal company research reports), and what was reported to the public led to a more favorable presentation in the medical literature of gabapentin&apos;s efficacy for unapproved indications,&quot; Dickersin and colleagues said.&lt;p&gt;&lt;/p&gt;
&lt;p&gt;Their study is the second this week to report that trial outcomes often change between original design and published report.&lt;/p&gt;
&lt;p&gt;On Monday, a study in &lt;em&gt;Annals of Family Medicine&lt;/em&gt; found that the primary outcomes of nearly a third of randomized trials, as published in five top journals, differed from those previously indicated in clinical trial registry listings. (See &lt;a href=&quot;http://www.medpagetoday.com/PublicHealthPolicy/ClinicalTrials/16895&quot; mce_href=&quot;http://www.medpagetoday.com/PublicHealthPolicy/ClinicalTrials/16895&quot; target=&quot;_blank&quot;&gt;Primary Outcomes Often Altered Prior to Trial Reporting&lt;/a&gt;)&lt;/p&gt;
&lt;p&gt;The gabapentin trials examined the drug for prevention of migraine and treatment of bipolar disorders, neuropathic pain, and nociceptive pain. None of these are currently approved indications for gabapentin, but the drug is approved for postherpetic neuralgia and is believed to be widely prescribed off-label for other types of pain.&lt;/p&gt;
&lt;p&gt;In fact, the litigation leading to the availability of the trial documents charged that Pfizer and Parke-Davis had improperly marketed gabapentin for such unapproved indications.&lt;/p&gt;
&lt;p&gt;The documents were either pretrial study protocols or standardized internal reports prepared at the study&apos;s end. Dickersin and colleagues compared the primary outcomes described in these documents with those stated in publications.&lt;/p&gt;
&lt;p&gt;Nine of the 20 studies were published as full-length articles in medical journals. Data from three others appeared in shorter forms: a letter to the editor, a conference abstract, and as part of a pooled analysis of multiple trials.&lt;/p&gt;
&lt;p&gt;In these 12 trials, there were 21 primary outcomes identified in the study protocols or internal reports. Six of these were not reported at all in the publications, and four others were demoted to secondary outcomes, Dickersin and colleagues found. The other 11 were reported without change.&lt;/p&gt;
&lt;p&gt;The published studies also included 12 primary outcomes that were not mentioned in the internal documents, and five others that were upgraded from secondary outcomes.&lt;/p&gt;
&lt;p&gt;Altogether, eight of the 12 published studies involved some discrepancy in primary outcomes between the publications and the internal documents, Dickersin and colleagues said.&lt;/p&gt;
&lt;p&gt;In the four trials for which there was full agreement, three reported a statistically significant finding in gabapentin&apos;s favor for the primary outcomes.&lt;/p&gt;
&lt;p&gt;In the eight trials with disagreement, five reported favorable results for the new primary outcomes. Of those five, four were published as full-length articles, as were all three of those with published primary outcomes, the same as described in the internal documents.&lt;/p&gt;
&lt;p&gt;All but one of the unpublished studies found either insignificant or only marginally significant results for their primary outcomes, Dickersin and colleagues indicated.&lt;/p&gt;
&lt;p&gt;In an interview, Dickersin said the findings undercut at least some of the justification for off-label gabapentin in pain indications and bipolar disorders.&lt;/p&gt;
&lt;p&gt;&quot;Because the outcomes we examined were fiddled with to some extent, we don&apos;t know what the outcome was for the primary outcome that was originally set,&quot; she said.&lt;/p&gt;
&lt;p&gt;&quot;If I were a physician, or a patient, I would be thinking, what is the true evidence base here?&quot;&lt;/p&gt;
&lt;p&gt;She added that the altered outcomes and unpublished studies had the potential to skew results of systematic reviews, such as those produced by the Cochrane Collaboration.&lt;/p&gt;
&lt;p&gt;A 2005 Cochrane review, based in part on the studies Dickersin and colleagues examined, had found that gabapentin was effective for acute and chronic pain.&lt;/p&gt;
&lt;p&gt;&quot;[It] should now be updated with the inclusion of unpublished information made available through litigation,&quot; she and her colleagues wrote.&lt;/p&gt;
&lt;p&gt;They also recommended that registration of clinical trials on sites such as Clinicaltrials.gov should be mandatory, and the listings should include full study protocols and amendments.&lt;/p&gt;
&lt;p&gt;&quot;The word transparency is now one that&apos;s on everybody&apos;s lips, whether we&apos;re talking about healthcare reform or research evidence,&quot; Dickersin said.&lt;/p&gt;
&lt;p&gt;She added that attention of regulators and journal editors should not exclusively focus on studies of new drugs and devices.&lt;/p&gt;
&lt;p&gt;&quot;We have to start thinking more about off-label drug usage, and are we really getting the information we need to make good decisions,&quot; she said.&lt;/p&gt;
&lt;div style=&quot;float:left;border-style:solid;border-width:1px;border-color:#8dabbc;font-family:arial;font-size:12px;background-color:#DBE9F2;padding:5px;&quot;&gt;&lt;p&gt;No external funding for the study was reported.&lt;/p&gt;&lt;p&gt;Dickersin reported having served as a paid expert witness in the litigation related to the gabapentin trials; funds from this effort were donated to Johns Hopkins to support academic scholarship in the area of reporting biases. One co-author received fees from plaintiffs&apos; lawyers in the suit.&lt;/p&gt;&lt;/div&gt;&lt;div style=&quot;clear:both;&quot;&gt;&lt;/div&gt;
    </recommendedItem>
    <recommendedItem id="20090101_4_558"
                     title="ACC: Major Clinical Trials Get Double Exposure by Journals and Meetings"
                     score="-0.005"
                     href="