Cancer statistics: a comparison between World Health Organization (WHO) and Global Burden of Disease (GBD)

Abstract

This study aims to define comparability of 2016 statistics between Global Health Estimates (GHE) and Global Health Data Exchange (GHDx) registries for disability-adjusted life years (DALYs) and mortality of the 25 most frequent worldwide malignancies. An excellent correlation can be found between the two registries for both cancer DALYs (r = 0.988) and mortality (r = 0.993). Cancer-related DALYs are substantially equivalent (mean bias, −1.9%; P = 0.603), while cancer mortality is modestly but significantly overestimated in GDHx (mean bias, 9.1%; P = 0.004). These results suggest that cancer DALYs estimate appear almost overlapping between GHE and GHDx registries, while cancer mortality is slightly overestimated in GDHx.

Introduction

Accessing official health-related repositories such as the Global Health Estimates (GHE) maintained by the World Health Organization (WHO),¹ or the Global Health Data Exchange (GHDx) derived from the Global Burden of Disease (GBD) Study,² is perhaps the most rapid, efficient and reliable strategy for obtaining recent statistics on worldwide epidemiologic burden for the vast majority of human diseases, and cancer makes no exception to this rule.³

The selection of the most appropriate source of cancer statistics is one of the most important aspects that challenge the minds of policymakers, healthcare administrators, epidemiologists, clinicians, scientists and even patients.⁴ Notably, the WHO has earlier emphasized that the previous GBD 2010 data were similar in some areas to WHO estimates, while substantial differences could be noted for many other conditions, thus concluding that more detailed information on data and methods used were necessary before the WHO could endorse GBD conclusions.⁵ The recent release of the new 2016 GHE data would hence make it important to verify the magnitude of potential differences still existing between these two registries. Therefore, the aim of this study is to objectively define the degree of comparability between 2016 cancer statistics contained in the GHE and GHDx registries.

Methods

The GHE database provides a large and comparable set of epidemiologic information (healthcare impact and mortality) obtained from recent vital registration data for all countries submitting information to the WHO, as well as information obtained from WHO Departments, collaborating United Nations (UN) Agencies and expert advisory groups and academic collaborators. In the minority of countries without usable vital registration data, statistics of GBD study were also included. The final number of sources and countries varies across the different diseased conditions.

The GHDx registry provides a standardized analytical approach for estimating epidemiologic information (incidence, prevalence, healthcare impact, mortality and so forth) on over 328 diseases and injuries from 195 different countries within 21 regions and 7 super-regions.⁶ A total number of 18 792 sources of epidemiological surveillance data (country-years of disease reporting) were analyzed for obtaining 2016 data, which were then combined using a Bayesian meta-regression tool. As specifically regards malignancies, the epidemiologic burden of different cancers was assessed by estimating the association between mortality and incidence, considering the impact on survival of access to, and quality of, treatment for cancer sites.

Comparability between the two registries was verified by extracting raw data (in absolute values) of disability-adjusted life year (DALY; as measure of burden of diseases, injuries and risk factors) and mortality of the 25 most frequent malignancies around the world in the two sexes, which were hence separately considered. The paired raw data for each cancer and either gender obtained from the two registries were imported into an Excel file (Microsoft, Redmond, WA, USA) and analyzed with Analyse-it (Analyse-it Software Ltd, Leeds, UK). The comparison between DALYs and mortality data contained in GHE and GHDx registries was carried out with Spearman’s correlation, Deming’s fit and Bland & Altman plot analysis. The study was performed in accordance with the Declaration of Helsinki and under the terms of relevant local legislation.

Results

The result of the correlation of DALYs and deaths for 25 most frequent worldwide malignancies retrieved from GHE and GDHx registries for the year 2016 in the two separate sexes is shown in figure 1. An excellent correlation can be found between the two registries, for both DALYs (r = 0.988; 95% CI, 0.979–0.994; P < 0.001) and mortality (r = 0.993; 95% CI, 0.988–0.996; P < 0.001). The Deming’s fit is (GDHx) = 0.95 × (GHE) + 0.03 for DALYs and (GDHx) = 1.07 × (GHE) for mortality, respectively. Bland and Altman plot analysis confirms that cancer-related DALYs are substantially equivalent between the two registries (mean bias, −1.9%; 95% CI, −9.3 to 5.5%; P = 0.603), while cancer mortality data are modestly but significantly overestimated in the GDHx compared with the GHE registry (mean bias, 9.1%; 95% CI, 3.1–15.2%; P = 0.004). In 82% of cases, GDHx mortality values are higher than those in the GDHx database. The higher biases can be observed for breast cancer deaths in men (+3.0-fold), larynx cancer deaths in women (+1.84-fold), non-melanoma skin cancer deaths in men (+1.31-fold), Hodgkin lymphoma deaths in men (+1.27-fold) and lip/oral cavity cancer deaths in women (+1.21-fold), whereas testicular cancer mortality is overestimated in the GHE register (+1.38-fold).

Discussion

Obtaining accurate epidemiologic information is almost unavoidable for informing policy dialogue, measuring progress in health care, as well as for developing effective healthcare interventions. Although the WHO and the Institute for Health Metrics and Evaluation (IHME), which coordinates the GBD Study, have recently agreed to establish a narrow partnership, the two organizations are still maintaining two separate health-related repositories, the GHE and the GDHx. This is an understandable source of uncertainty for policymakers, healthcare administrators, epidemiologists, clinicians, scientists and patients, who may remain undecided on which database shall be used.

Despite the differences between GHE and GDHx, the results of our analysis attest that the DALYs and mortality data for the 25 most frequent worldwide cancers contained in these two large health-related repositories are overall comparable. The bias of DALYs estimates between the two sources appears almost negligible, so that the two registries can perhaps be used almost interchangeably for assessing cancer-related healthcare burden. A modest bias was instead observed for cancer mortality estimates, whereby the values contained in the GDHx repository are on average ∼9% higher than those contained in the GHE registry. More specifically, 82% of mortality estimates for cancers in both genders are higher in GDHx than in the GHE. Nevertheless, this is not really surprising considering that the two registries use different data sources, analyses and covariates for modeling epidemiologic data.

We can hence conclude that the estimation of cancer-related healthcare burden can be indifferently performed in GHE or GDHx, while cancer-related mortality shall be uniformed on a single database.

Conflicts of interest: None declared.

References