These estimates provide the first standardized baseline for 194 Member States to measure progress toward the Global Initiative for Childhood Cancer (GICC) 2030 target of at least 60% survival. Developed using a common methodology and assumptions, they enable consistent tracking of survival outcomes, assessment of which countries are on course to meet the target, and identification of areas requiring intensified cancer control efforts.
Definition:
5-year net survival is the probability of surviving up to five years after being diagnosed with lymphoid leukaemia. Net survival is the survival attributable exclusively to lymphoid leukaemia and it is estimated by removing the effects of background mortality. Background mortality is the mortality due to causes other than lymphoid leukaemia.
The age distribution of children with lymphoid leukaemia may vary between countries and over time, and net survival varies with age. Therefore, valid comparison of all-ages survival estimates requires age standardisation to correct for these differences. The age standardised estimate is a weighted average of the age-specific estimates and is obtained using a standard set of weights.
Associated terms:
Overall survival, childhood-cancer specific survival
Disaggregation:
WHO region, World Bank income group, Age group (children 0-14 years of age and adolescents 15-19 years of age
Method of measurement
CONCORD-3 estimates for children or adolescents diagnosed with all lymphoid leukaemias combined (ICCC-3 group Ia). These estimates were complemented with those from CONCORD-2 to include survival data covering 1995-1999 for countries included in CONCORD-3 or 1995-2009 for countries not subsequently included in CONCORD-3.
Survival measured directly from high-quality PBCRs offers reliable insights into improvement in childhood cancer care. However, in many countries, PBCRs have not yet achieved sufficient robustness. To address existing data gaps, statistical modelling leveraging associations with covariates can be useful in providing plausible predictions of survival.
Method of estimation:
To predict the 5-year net survival for children (0-14) and adolescents (15-19) diagnosed with lymphoid leukaemia, separate statistical models were developed following three general principles. First, in countries with high-quality survival data from CONCORD, the model should closely align with the available data (accounting for the uncertainty associated with each data point) while filling in gaps and extrapolating to recent time periods. Second, in countries without CONCORD data, the model should provide robust predictions by leveraging the observed relationship between survival data and predictive covariates, to predict the possible level of survival. Third, in countries where survival data had high uncertainty, a balance was required to use both actual survival data and covariate relationships to inform predictions.
Since there are more reported data points with larger sample sizes for children compared to adolescents, separate hierarchical regression models are used to predict 5-year net survival for each age group. These models are identical, except the adolescent age group model uses the predicted 0-14 survival as an additional predictive covariate.
Method of estimation of global and regional aggregates:
Although the models developed offer reasonable predictions based on validation results, several interpretative constraints warrant careful consideration. First, while CONCORD-3 data served as the primary source, CONCORD-2 data were supplemented to maximize data volume for countries or years where data were lacking. Differences between the two, such as changes in morphology definitions and the use of the period approach for net survival estimation, could result in some level of incompatibility and discontinuity. Second, population-based cancer registries are more common in high- and middle-income countries, implying that much of the data informing the modelled relationship were driven by data-rich countries. It is possible that certain presumed relationships might not be completely generalizable to data-sparse countries. Third, in some countries, only subnational survival data were available. In the current analysis, we specify these special incidences; however, no adjustments were made to infer the subnational predictions to the national level. Fourth, our model does not consider relationships with other metrics, such as incidence and mortality. Although a derivative metric, 1-MIR, was used in the model, it does not adequately capture the transitional relationships across the three metrics. Follow-up analyses should consider this transitional structure to generate more coherent prediction. Fifth, the models only considered covariates with sufficient geographical and temporal coverage to allow for estimation and prediction. Other important covariates, such as the availability of chemotherapy, were not incorporated due to the lack of readily available comprehensive data. Finally, the quality of our predictions is largely governed by the quality of the input data. Despite our best efforts to use the most robust model and exploit associations between covariates and lymphoid leukaemia survival, as well as potential spatiotemporal correlation, our predictions for countries with limited or no data are subject to a high level of uncertainty. It is crucial for countries to continue pursuing high-quality cancer registration data and strengthening existing health informatics systems.
If you have any feedback, you are welcome to write it here.
If you need to access the old Global Health Observatory data, you can do it here. But before you leave, please provide us your feedback about our new data portal.
