Updated Methodology for Projecting US- and State-Level Cancer Counts for the Current Calendar Year: Part II: Evaluation of Incidence and Mortality Projection Methods

Background: The American Cancer Society (ACS) and the NCI collaborate every 5 to 8 years to update the methods for estimating the numbers of new cancer cases and deaths in the current year for the U.S. and individual states. Herein, we compare our current projection methodology with the next generation of statistical models. Methods: A validation study was conducted comparing current projection methods (vector autoregression for incidence; Joinpoint regression for mortality) with the Bayes state-space method and novel Joinpoint algorithms. Incidence data from 1996-2010 were projected to 2014 using two inputs: modeled data and observed data with modeled where observed were missing. For mortality, observed data from 1995 to 2009, 1996 to 2010, 1997 to 2011, and 1998 to 2012, each projected 3 years forward to 2012 to 2015. Projection methods were evaluated using the average absolute relative deviation (AARD) between observed counts (2014 for incidence, 20122015 for mortality) and estimates for 47 cancer sites nationally and 21 sites by state. Results: A novel Joinpoint model provided a good fit for both incidence and mortality, particularly for the most common cancers in the U.S. Notably, the AARD for cancers with cases in 2014 exceeding 49,000 for this model was 3.4%, nearly half that of the current method (6.3%). Conclusions: A data-driven Joinpoint algorithm had versatile performance at the national and state levels and will replace the ACS's current methods. Impact: This methodology provides estimates of cancer data that are not available for the current year, thus continuing to fill an important gap for advocacy, research, and public health planning.

Automated summary

The study compared current cancer incidence and mortality projection methods with next-generation statistical models, finding that a novel Joinpoint model provided a good fit for both types of data, especially for common cancers in the U.S. This data-driven algorithm will replace existing methods and fill an important gap for advocacy, research, and public health planning.