JOURNAL OF CLINICAL ONCOLOGY C O R R E S P O N D E N C E Reply to M. Nayan et al We thank Nayan et al 1 for their interest in our recently pub- lished article in Journal of Clinical Oncology on postdiagnosis statin use and prostate cancer mortality. 2 We are grateful for the oppor- tunity to clarify some of the issues raised regarding our study. We acknowledge that several drugs in addition to statins have been suggested to reduce prostate cancer mortality, and we cannot exclude that unmeasured drugs with anticancer effects may have influenced our results. Nayan et al 1 state that our models only considered aspirin and other nonsteroidal anti-inflammatory drug use; however, this is inaccurate. In the multivariable adjusted analyses, we adjusted for an array of drugs that have been suggested to influence cancer mortality including aspirin; nonaspirin non- steroidal anti-inflammatory drugs; antihypertensives, such as b-blockers, calcium channel blockers, and angiotensin-converting enzyme inhibitors; other cardiovascular drugs, such as cardiac glycosides, antiarrhythmic agents, cardiac stimulants, vasodilators, and oral anticoagulants; and insulin and oral antidiabetic drugs. We did not adjust for use of selective serotonin reuptake inhibitors, however only limited research has been undertaken on cancer outcomes associated with selective serotonin reuptake inhibitor use, and the referenced study by Boursi et al 3 was published after the submission of our work. In addition to drug covariates, we included chronic comorbid conditions in the models, and because these conditions typically require drug therapy, adjustment for comorbidity also contributed to adjustment for drug use. We did not consider discontinuation of statin use in our analyses because we regarded the filling of two postdiagnosis statin prescriptions as chronic use throughout the follow-up period; however, of importance, we lagged statin exposure by 1 year. This modification eliminates an important drug use pattern addressed in a recent study of SEER data in the United States 4 , which reported that more than one half of patients who initiated statin use within 6 months after a cancer diagnosis did not fill a statin prescription in their last year of follow-up. As part of end-of-life clinical man- agement, statin therapy is often discontinued, and the introduction of a lag period minimizes the influence of such end-of-life change in exposure. 5 Moreover, the lag period allows for a more mean- ingful latency period, along with the definition of statin use of a minimum of two prescriptions. An additional rationale for our predefined definition of statin exposure without a detailed account of discontinuation is provided by a survey that reported a high persistence of statin use in Denmark. 6 The survey revealed that only 16% of individuals who initiated statin therapy between 2004 and 2010 discontinued treatment. Moreover, discontinuation of statin use would, in any case, attenuate the association between postdiagnosis statin use and prostate cancer mortality. Finally, the lack of prescription data during hospitalization after the start of follow-up was unlikely to induce major exposure misclassification, as patients with prostate and other cancers are rarely hospitalized for longer periods, except as part of end-of-life care; also, patients in Denmark are not provided with drug supplies for more than a few days when discharged from the hospital. Nayan et al 1 are correct to state that we restricted the analysis of cumulative use to the prespecified 5-year analysis conditioned on 5 years’ survival of patients. We divided the cumulative amount of postdiagnosis statin use into tertiles of the total number of defined daily doses of statins filled within 5 years after prostate cancer diagnosis. As we did not observe any evidence of a dose–response relationship according to cumulative amount in this analysis, we did not perform post hoc analyses of cumulative amount; however, with regard to the intensity of statin use, we were more in line with the suggestions by Nayan et al 1 , as we estimated the average dose of statin use (, 1; $ 1 defined daily doses) continuously throughout follow- up in the time-varying analysis (Table 4 2 ). The simulation study by Emilsson et al 4 reporting null associa- tions between statin use initiated within 6 months after cancer diag- nosis and cancer-speci fic or all-cause mortality is intriguing; however, the study had some limitations that impede a comparison with our study. First, the study had a short follow-up of maximum 3 years. As a result of the long survival of the majority of patients with prostate cancer, this may have been insuf ficient to assess the effect of statin use. Second, although we acknowledge that the study by Emilsson et al 4 indeed minimized selection and immortal time bias, statin use was only assessed close to the diagnosis, which may have attenuated associations. In conclusion, despite the inverse association between post- diagnosis statin use and prostate cancer mortality observed in our registry-based study and in several other epidemiologic studies, we agree with Nayan et al that only well-designed, targeted, randomized trials can eventually determine whether prescribing statins as adjuvant prostate cancer therapy should be recommended. Signe Benzon Larsen, Christian Dehlendorff, Charlotte Skriver, and Susanne Oksbjerg Dalton Danish Cancer Society, Copenhagen, Denmark Christina Gade Jespersen Viborg Hospital, Viborg; and Aarhus University Hospital, Aarhus, Denmark Michael Borre Aarhus University Hospital, Aarhus, Denmark Klaus Brasso Copenhagen University Hospital, Copenhagen, Denmark Mette Nørgaard Aarhus University Hospital, Aarhus, Denmark Christoffer Johansen Danish Cancer Society and Copenhagen University Hospital, Copenhagen, Denmark Henrik Toft Sørensen Aarhus University Hospital, Aarhus, Denmark Jesper Hallas University of Southern Denmark, Odense, Denmark Journal of Clinical Oncology, Vol 36, No 6 (February 20), 2018: pp 629-630 © 2017 by American Society of Clinical Oncology 629 VOLUME 36 • NUMBER 6 • FEBRUARY 20, 2018 Downloaded from ascopubs.org by 100.26.137.118 on March 27, 2021 from 100.026.137.118 Copyright © 2021 American Society of Clinical Oncology. All rights reserved.