• health surveillance
• public health
• cancer
• respiratory

When the global burden of disease (GBD) methodology was introduced some 30 years ago, it represented a significant step toward providing systematic, comparable disease surveillance data on a worldwide scale. Since that time, the GBD approach has led to valuable insights about the relative importance of various risk factors and causes of death and how they have changed over time. As a result, data from GBD reports have become a common source of supporting data for the background sections of research papers, grant proposals and policy documents.

The three related papers on the burden of occupational disease and injury in this issue of the journal1–3 are likely to have similar impacts. These papers based on 2016 GBD data present a welcome updated estimate of the overall burden of illness and injury from occupational exposures1 and more detailed burden assessments for cancer2 and non-malignant, non-infectious respiratory diseases.3

A notable finding from this work is that an estimated 1.5 million deaths worldwide in 2016 were attributable to occupational exposures. While the corresponding attributable proportion, 2.8% of all deaths, may look like a small part of the public health picture, contributing, perhaps, to an impression that occupational exposures are not very important anymore, other GBD data4 indicate that, in fact, occupational exposures account for more deaths than a number of well-known risk factors that tend to receive far more attention from the public and policy makers, including low physical activity, low dietary fibre, high red and processed meat consumption, drug use and unsafe sex. Another finding of broad interest is the observation that while the estimated rates of mortality and disability-adjusted life years (DALYs) attributable to occupational risk factors declined over 26 years of GBD data, the absolute numbers increased, apparently due to growth of employment in exposed occupations.1

For cancer, too, these new GBD estimates2 show that the population impact of occupational exposures appears to outrank that of many prominent risk factors, including ambient and household air pollution, high body mass index, secondhand tobacco smoke outside the workplace and every individual dietary component evaluated in recent GBD reports.4 As for the overall burden of occupational exposures, the estimated numbers of deaths and DALYs attributed to occupational carcinogens have increased over time, even while the rates decreased. However, the data suggest that these trends have not been uniform across regions or among risk factors. Notably, declining rates are seen in high Sociodemographic Index (SDI) countries, while increasing rates are reported for countries for the middle and lower SDI ranks.2

The burden of non-malignant respiratory disease from occupational exposures3 may be even larger in relative terms, but these diseases have few other established risk factors for comparison. For this group of diseases, too, the data indicate modest increases in absolute numbers of deaths and DALYs, despite declining rates.3

The main findings of these papers lead to several observations with mixed implications for past and future occupational health efforts. On the one hand, declining rates suggest that progress has been made toward reducing the risks of major occupational diseases and injuries. These data are most encouraging with respect to asbestos—the single most important occupational carcinogen—for which the attributable death rate decreased by 14% from 1990 to 2016.2 On the other hand, despite this important progress, the total burden of work-related illness and injury in terms of absolute numbers of deaths and DALYs remains stubbornly high and even appears to be increasing for some outcomes, including cancer. This conclusion is troubling from a public health perspective, as it suggests that reductions in risk have been offset by increases in the numbers of exposed workers, resulting in a greater burden of death and disability. It also appears that the progress that has been made has not been equitably distributed. Taken together, these observations suggest that efforts to reduce the incidence of occupational disease have had some notable successes, particularly in higher-income countries, yet occupational risk factors remain important, with much work still to be done to reverse the growth of the absolute burden of work-related disease and injury on a global basis.

The ability to support inferences like these is one of the most valuable contributions of the GBD data. However, as useful as the data can be and as familiar as they have become in the public health realm, it is important to recognise that these estimates of outwardly simple metrics are ultimately built on incomplete data through a complex process requiring numerous assumptions. The three papers in this issue illustrate some of the key limitations of the GBD approach, as well as its strengths. While the authors acknowledge the main limitations of their data and methods, a few key points are worth noting.

One methodological limitation with particular significance in the context of occupational health stems from the absence of comprehensive, comparable quantitative data on exposure levels. Lacking such data, the authors assigned exposure status for most agents to industry or occupation categories based on limited published data—mostly from high-income countries—and applied relative risks for exposed–unexposed comparisons from available studies and meta-analyses. Exposure estimates for carcinogens incorporated data from high-income countries to estimate the prevalence of ‘high’ and ‘low’ exposure, with a larger proportion of workers assumed to be highly exposed in lower-income countries.2 A similar approach was used for noise.1 These assumptions are reasonable enough given the scarcity of data for most agents in most countries, yet there is considerable room for measurement error. For example, if exposures in middle-income and low-income countries are quantitatively higher than in the studies from which the data were drawn, the true number of attributable cases, and thus both the rate and the absolute burden, could be higher than estimated.

In addition, the estimates for cancer are based on only 14 of 47 currently recognised carcinogens with evidence of occupational exposure.5 This decision was taken in part because of a lack of exposure data for many of those agents.2 However, although based on exposure rather than outcomes, it effectively excludes several types of cancer, notably cancer of the bladder, that are known to be associated with established occupational carcinogens.5 Including other occupational carcinogens, particularly those with high relative risks or large numbers of exposed workers, such as welding fumes,6 could lead to higher burden estimates. And, as the authors note,1 including other occupational agents with widespread exposure but less solid evidence of causality, such as pesticides, shiftwork, or stress could also result in a higher estimated burden.

Another question of interest is the extent to which population growth in itself, rather than shifting employment toward exposed jobs, possibly by outsourcing production from higher-cost to lower-cost regions, accounts for the increasing burden of many work-related conditions despite declining risks. Analyses of other GBD data4 have attempted to separate the effects of population growth and changes in risk factor distributions, and this could presumably be done for occupational exposures, as well.

A final feature of the GBD methodology that may not be readily apparent is that the burden estimates presented in these papers are inferences about the effects of past exposures on current mortality and morbidity. While these essentially retrospective inferences are logically valid and informative, they do not directly address pressing questions about the burden that will arise from current exposures or the benefits that could be achieved through future interventions. An approach to this kind of predictive estimation has been illustrated using national data on the burden of occupational cancer.7 Estimates of the future global burden from occupational exposures would help inform decisions about how to attack the large remaining burden going forward.