The impact of biomedical innovation on cancer mortality in 37 countries, 2003–2017

Abstract

Previous studies have shown that biomedical innovation, as measured by the long-run change in the vintage of MeSH descriptors of PubMed articles, has been the principal cause of declining cancer mortality in the U.S. The effect of biomedical innovation on cancer mortality outside of the U.S. may be different from its effect in the U.S. This study uses similar methods to investigate econometrically the impact that biomedical innovation had on cancer mortality in 37 countries during the period 2003–2017. Estimates based on pooled data by cancer site, country, and year indicate that the age-adjusted cancer mortality rate is significantly inversely related to the vintage of descriptors of PubMed articles about that cancer site 9–20 years (but not 0–8 years) earlier, controlling for the age-adjusted cancer incidence rate. The mortality rate is most strongly inversely related to the vintage of descriptors 16 years earlier. This finding is consistent with evidence from other studies that there is a substantial lag from biomedical innovation to population health. Excluding the U.S. (which accounts for 26 % of the cancer deaths in the 37 countries) from the sample does not have a substantial effect on the estimates. 85 % of the 2003–2017 decline in the cancer mortality rate is estimated to be attributable to the 1987–2001 increase in mean vintage. Only 6 % of the 2003–2017 decline in the cancer mortality rate is estimated to be attributable to the contemporaneous decline in the cancer incidence rate. In addition to estimating models using pooled data for many countries, we estimate models of the cancer mortality rate separately for each country. The estimate of the coefficient of vintage 16 years earlier is negative and significant for 24 of the 37 countries, and negative and marginally significant for 2 other countries; it is not positive and significant for any country. The population-weighted mean of the 37 estimates is very similar to the estimate from the model based on pooled data from many countries.