Abstract
GLOBOCAN is the most widely used source of national cancer statistics, yet users routinely treat the difference between two releases as real epidemiological change. We quantified how far estimates move between the 2022 and 2024 cycles and whether this depends on development level. Age-standardised incidence and mortality rates for eight cancers were paired between GLOBOCAN 2022 and 2024 for 185 countries common to both cycles. Agreement was assessed by Bland-Altman analysis and Lin's concordance correlation coefficient (CCC); per-country revision magnitude was related to the Human Development Index (HDI) using rank and robust regression, with partial correlation adjusting for baseline incidence. Pooled concordance was high (CCC 0.97 for incidence, 0.95 for mortality) with negligible mean bias, but limits of agreement were wide and the tail heavy: 62.4%, 40.3%, 28.1% and 13.4% of country-cancer-sex cells were revised by more than 10%, 20%, 30% and 50%. Revision magnitude was inversely related to HDI (Spearman ρ = -0.62 for incidence and -0.62 for mortality; both p < 0.001), falling from 23.3% in low-HDI to 8.1% in very-high-HDI countries and persisted after adjustment for baseline incidence (partial ρ = -0.36); the development gradient is thus only partly explained by lower, more volatile baseline rates. Thyroid cancer showed the largest revisions. These findings provide, to our knowledge, the first quantitative benchmark for interpreting apparent differences between successive GLOBOCAN releases and support more cautious interpretation of single-cycle estimates from lower-HDI settings.