Eleanor Hayes-Larson1, Crystal Shaw1,2, Sarah F Ackley3, Scott C Zimmerman3, M Maria Glymour3, Rebecca E Graff3, John S Witte3, Lindsay C Kobayashi4, Elizabeth Rose Mayeda1. 1. Department of Epidemiology, UCLA Fielding School of Public Health, Los Angeles, California, USA. 2. Department of Biostatistics, UCLA Fielding School of Public Health, Los Angeles, California, USA. 3. Department of Epidemiology and Biostatistics, University of California, San Francisco, San Francisco, California, USA. 4. Department of Epidemiology, University of Michigan School of Public Health, Ann Arbor, Michigan, USA.
Abstract
BACKGROUND: Cancer is inversely associated with dementia. Using simulations, we examined whether this inverse association may be explained by dementia diagnosis timing, including death before dementia diagnosis and differential diagnosis patterns by cancer history. METHODS: We used multistate Markov simulation models to generate cohorts 65 years of age and free of cancer and dementia at baseline; follow-up for incident cancer (all cancers, breast, prostate, and lung cancer), dementia, dementia diagnosis among those with dementia, and death occurred monthly over 30 years. Models specified no true effect of cancer on dementia, and used age-specific transition rates calibrated to U.S. population and cohort data. We varied the average lapse between dementia onset and diagnosis, including nondifferential and differential delays by cancer history, and examined observed incidence rate ratios (IRRs) for the effect of cancer on dementia diagnosis. RESULTS: Nondifferential dementia diagnosis delay introduced minimal bias (IRRs = 0.98-1.02) for all cancer, breast, and prostate models and substantial bias (IRR = 0.78) in lung cancer models. For the differential dementia diagnosis delay model of all cancer types combined, simulation scenarios with ≥20% lower dementia diagnosis rate (additional 4.5-month delay) in those with cancer history versus without yielded results consistent with literature estimates. Longer dementia diagnosis delays in those with cancer and higher mortality in those with cancer and dementia yielded more bias. CONCLUSIONS: Delays in dementia diagnosis may play a role in the inverse cancer-dementia relationship, especially for more fatal cancers, but moderate differential delays in those with cancer were needed to fully explain the literature-reported IRRs.
BACKGROUND: Cancer is inversely associated with dementia. Using simulations, we examined whether this inverse association may be explained by dementia diagnosis timing, including death before dementia diagnosis and differential diagnosis patterns by cancer history. METHODS: We used multistate Markov simulation models to generate cohorts 65 years of age and free of cancer and dementia at baseline; follow-up for incident cancer (all cancers, breast, prostate, and lung cancer), dementia, dementia diagnosis among those with dementia, and death occurred monthly over 30 years. Models specified no true effect of cancer on dementia, and used age-specific transition rates calibrated to U.S. population and cohort data. We varied the average lapse between dementia onset and diagnosis, including nondifferential and differential delays by cancer history, and examined observed incidence rate ratios (IRRs) for the effect of cancer on dementia diagnosis. RESULTS: Nondifferential dementia diagnosis delay introduced minimal bias (IRRs = 0.98-1.02) for all cancer, breast, and prostate models and substantial bias (IRR = 0.78) in lung cancer models. For the differential dementia diagnosis delay model of all cancer types combined, simulation scenarios with ≥20% lower dementia diagnosis rate (additional 4.5-month delay) in those with cancer history versus without yielded results consistent with literature estimates. Longer dementia diagnosis delays in those with cancer and higher mortality in those with cancer and dementia yielded more bias. CONCLUSIONS: Delays in dementia diagnosis may play a role in the inverse cancer-dementia relationship, especially for more fatal cancers, but moderate differential delays in those with cancer were needed to fully explain the literature-reported IRRs.
Authors: Elizabeth R Mayeda; M Maria Glymour; Charles P Quesenberry; Julene K Johnson; Eliseo J Pérez-Stable; Rachel A Whitmer Journal: Alzheimers Dement Date: 2017-02-05 Impact factor: 21.566
Authors: Monica Ospina-Romero; M Maria Glymour; Eleanor Hayes-Larson; Elizabeth Rose Mayeda; Rebecca E Graff; Willa D Brenowitz; Sarah F Ackley; John S Witte; Lindsay C Kobayashi Journal: JAMA Netw Open Date: 2020-11-02