Evaluation of the 4Kscore in EPIC: Reply

We write in response to the letter from Prof. Hans Lilja and Dr. Andrew Vickers with regard to our recent publication in Cancer Epidemiology, Biomarkers & Prevention entitled “Evaluation of the 4Kscore Test in Relation to Subsequent Risk of Aggressive Prostate Cancer in the European Prospective Investigation into Cancer and Nutrition” (1).This project began in 2011 following discussions among Profs. Lilja, Travis, and Key to measure prostate-related proteins in EPIC: microseminoprotein β (MSMB), prostate-specific antigen (PSA) and subforms, and human kallikrein 2 (hK2). Although our initial interest was MSMB, Prof. Lilja suggested including PSA, PSA subforms, and hK2.A pilot study in Prof. Lilja’s laboratory assessed the use of citrated plasma for this project. Only after Profs. Lilja, Travis, and Key all agreed that the samples were suitable did Prof. Lilja’s team subsequently complete all assays (2013–2015). The findings on MSMB, which included the use of total PSA measurements, were published in 2019 with Prof. Lilja as cosenior author (2).Once all data were generated, Prof. Lilja suggested that they would also be suitable to evaluate the discriminative performance of an emerging predictive score for prostate cancer, the 4Kscore. As AUC is a rank-order statistic and very few analyte values were below our lower limits of detection, lower values in our plasma samples—akin to a monotonic transformation—were not likely to influence the relative ranking of participants or the validity of our discrimination analysis. We designed the analysis plan with Prof. Lilja and Dr. Vickers, incorporating their specifications for PSA thresholds and novel matching strategies. We used code provided by Vickers’ team, implemented their requested updates throughout, and shared outputs at every step.At Prof. Lilja’s request, and to increase statistical power, we also agreed in 2016 to include preexisting data from the Malmö Diet and Cancer study on the 4Kscore. These data were not originally collected as a 1:1 matched case–control study and were therefore rematched to conform to our study’s design; the results were materially unchanged. We note that our results are also materially similar to those reported in the Multiethnic Cohort Study (3), which is one of few prospective studies of the 4Kscore in men with an unrestricted range of analytes at recruitment.As noted in our correspondence with Prof. Lilja and Dr. Vickers over the course of this project, we sent 1 blinded quality-control (QC) aliquot for every 10 samples (approximately 405 QC aliquots in this study). Additionally, at Prof. Lilja’s request in 2017, we excluded 32 samples with values below the advised lower limit of detection together with their matched partners (32 pairs; 64 men). Their letter cites a discrepancy of 674 between men measured and analyzed. The QCs plus the specified pair exclusions account for 469 of these; the remaining 205 are samples with analyte measures without a complete 1:1 matched partner. All of this information was shared with Prof. Lilja and Dr. Vickers during the collaboration.As disclosed in their letter, Prof. Lilja and Dr. Vickers hold patents relating to the use of kallikrein markers included in the 4Kscore with assignments to OPKO entities from 2013. We were not made aware of these financial relationships during initial measurement planning or when the initial analysis plan was agreed; we first became aware of this in 2015.After implementing multiple rounds of requested feedback from collaborators Prof. Lilja and Dr. Vickers between 2015 and 2022, and with no material changes in results, we proposed to move forward to publication. Prof. Lilja and Dr. Vickers remained dissatisfied with the results, but between 2022 and 2024 they did not provide any further specific requests or suggestions for updated analyses to draw deeper insights.We therefore elected to proceed with submission of our findings. Profs. Lilja and Manjer and Dr. Vickers requested removal from the manuscript, along with removal of Malmö Diet and Cancer study data, which did not alter any conclusions about the performance of the 4Kscore.G.K. Fensom reports grants from Cancer Research UK during the conduct of the study. R.C. Travis reports grants from Cancer Research UK during the conduct of the study. No disclosures were reported by the other authors.The authors would like to express their gratitude to the late Dr. Domenico Palli for his contribution to the wider collaboration and the development of this work.