Send Orders for Reprints to reprints@benthamscience.ae Reviews on Recent Clinical Trials, 2015, 10, 00-00 1 1574-8871/15 $58.00+.00 © 2015 Bentham Science Publishers Changes in and Impact of the Death Review Process in the Prostate, Lung, Colorectal and Ovarian (PLCO) Cancer Screening Trial Anthony B. Miller a,* , Ronald Feld b , Robert Fontana c , John K. Gohagan d , Ismail Jatoi e , Walter Law- rence Jr. f , Amy Miller g , Philip C. Prorok h , Ashwani Rajput i , Morris Sherman j , Gilbert Welch k , Pat- rick Wright l , Susan Yurgalevitch g , Peter Albertsen m a Dalla Lana School of Public Health, University of Toronto, Toronto, ON, Canada; b The Princess Margaret Cancer Centre, University Avenue, Toronto, ON, Canada; c Mayo Medical School, Roches- ter, MN, USA; d Office of Cancer Prevention, National Institutes of Health, Bethesda, MD, USA; e University of Texas Health Science Center, San Antonio, TX, USA; f Virginia Commonwealth University, Richmond, VA, USA; g Westat, Rockville, MD, USA; h Division of Cancer Prevention, Na- tional Cancer Institute, Bethesda, MD, USA; i Department of Surgery, University of New Mexico, NM, USA; j Toronto General Hospital, Toronto, ON, Canada; k Dartmouth Institute for Health Policy & Clinical Practice, Hanover, NH, USA; l Information Management Services, Inc., Rockville, MD, USA; m University of Connecticut, Farmington, CT, USA Abstract: Death review was conducted for the Prostate, Lung, Colorectal and Ovarian (PLCO) cancer screening trial to avoid the biases associated with causes of death entered on death certificates. An algorithm selected deaths for review. Records on diagnosis and terminal illness were perused in the coordinating cen- ter and by the chair of the death review committee (DRC). Identifying information and randomization arm was removed. Three reviewers independently determined the cause of death. Disagreement was resolved at a meeting of the DRC. This process was subsequently simplified. The cause of death was determined by one DRC member and compared to the death certificate. With agreement the case was finalized. When discordant, the records were sent to a second DRC mem- ber. If the reviewers agreed, the case was finalized. If not, a third member reviewed. If two of the three reviewers agreed, the case was sent back to the discordant reviewer. If the reviewer remained discordant the case was resolved by a confer- ence call. Of the 4728 death reviews that were completed, the DRC confirmed the death certificate underlying cause for over 90%. Between 5% and 13% of the certified deaths were regarded as indirect causes of death, associated with the treatment of the ascertained cancer; differential for prostate cancer, 11% in the intervention arm and 6% in the control. Without review, between 1% and 6% of the deaths that occurred would not have been assigned to the relevant PLCO cancer. The DRC completed 76% of those requiring review before the process ceased. Keywords: Colorectal cancer, death review, death certificates, lung cancer, ovarian cancer, prostate cancer, randomized screen- ing trials. 1. INTRODUCTION Determining the cause of death accurately is essential in a randomized screening trial such as the Prostate, Lung, Co- lorectal and Ovarian (PLCO) cancer screening trial as cause- specific mortality is the endpoint for the trial. It is known that there are uncertainties over the accuracy of the underly- ing cause of death entered on many death certificates, in part because some physicians only enter the precipitating rather than the underlying cause of death, and in part, because there may be genuine uncertainty over the true cause of a patient’s terminal illness and death [1, 2]. It has been suggested that cause-specific mortality may not be the best endpoint for randomized screening trials, and * Address correspondence to this author at the Dalla Lana School of Public Health, University of Toronto, Toronto, Ontario, M5T 3M7, Canada; Tel/Fax: 1-416-487-5825/1-416-978-7735; E-mail: ab.miller@sympatico.ca. that all-cause mortality is preferable [3]. This was largely to overcome the biases recognized to potentially affect the un- derlying cause entered on death certificates, sticking diagno- sis [4] and slippery linkage biases [3]. Sticking diagnosis is the fact that an individual’s cancer diagnosis influences the cause of death entered on the death certificate, even though the individual may not have died from events originating with the cancer or its treatment. Slippery linkage is the fail- ure to recognize that death was the result of the cancer or its treatment, but another cause - e.g. a heart attack after sur- gery, was entered on the death certificate. To date, no screen- ing trial has used all-cause mortality as the primary endpoint, and none were powered adequately to enable them to do so. Further, in a trial such as PLCO, with 3 screening tests of- fered to each participant allocated to screening, there would be difficulties in interpreting the effect of changes in all- cause mortality. To overcome these issues of misclassifica- tion and biases if death certificates were relied upon for clas- Anthony B. Miller