Temporal trends (1970–2019) of costs (in 2017-equivalent US$ millions) a considering the actual distribution of the mean amounts provided for each decade in the conservative subset and b using model predictions (i.e. OLS: ordinary least-squares; GAM: generalized additive model; linear regression, quadratic regression, MARS: multiple adaptive regression splines) and quantile regressions. We considered models calibrated and fitted with at least 75% of cost data completeness from the dataset. We log10-tranformed cost estimates using information from the cost estimate per year 2017 USD exchange rate column in the conservative subset).