Tools and guidance for exposome HIA simulation studies

Introduction

We focused on a simulation study to investigate the influence of typical exposure interventions on a health outcome. Full details of the simulation methodology and results can be found in the main report.

In brief, the simulation study replicated risk of lung cancer due to typical workplace and lifestyle exposures (Respirable Crystalised Silica, Diesel Fumes, and Smoking) experienced in those working in the construction industry.

Using literature and expert consultation we replicated typical lifetime exposure histories for workers  within the construction industry between 1960 and 2060 and simulated their individual risk of Lung Cancer diagnosis. Any currently known exposure trends were assumed to stop in 2020 unless otherwise influenced by our simulated intervention. We then applied a set of exposure interventions typically seen in the construction industry, these focused around reducing the average exposure (i.e. a new technology), reducing the maximum allowed limit of a exposure (i.e. typical policy intervention), and increase in the pension age (i.e. policy with unintended consequences). Interventions were simulated to occur in the calendar year 2010.

We were then looking to observe the influence these health interventions had on annual lung cancer incidence, whilst under varying exposure (i.e. exposome) characteristics. The following table (below) outlines the key characteristics of exposures, and interventions we modified within our simulations.

The results of these interventions are then reported in terms of the time in years since intervention (Year 0 = calendar year 2010). We then report the annual incidence, average age of Lung Cancer Diagnosis, and the number of working years of life lost (relating to expected age of retirement predefined here as 65/70 years old). These are either reported in plots of the annual results themselves, or table of the results by decades since intervention i.e. 0, 10, 20, 30, 40 years post intervention.

Tools

In order to disseminate these results, provide guidance and application to future intervention in similar scenarios, and expansion on the parameters we have produced a set of tools for outside use.

• Excel Spreadsheet Disseminating Results
• Open access - Stata (Statistical Package) Program `ephorhia’
• Open access - Stata (Statistical Package) Full simulation syntax

Discussion

Our goal here was to explore the influence of an exposome with varying characteristics on a health intervention effect. We chose to investigate this within the relatively well studied construction industry, and the link between exposure of workers to ambient contaminants (both work based (silica, and diesel fumes) and general lifestyle i.e. smoking) and lung cancer diagnosis. We hoped to achieve this is such a way as it was a relatable but generalisable to more than just the construction industry setting. These tools generally describe the results we found rather than allow the user to input their own settings. We acknowledge the study is limited in its goal of making these results comparable across industries, exposures, and other health outcomes.

Construction is one of the most widely investigated industries in terms of exposure assessment. Even so, it was very difficult to find accurate estimates of the exposure, across multiple job roles, over number of measurements occurring over long periods of time. We therefore had to make a number of assumptions related to exposure, exposure-outcome relationships. We believe they are as realistic as possible but may be limited in nature. These include assumptions relating to:

• Population average exposure levels
• Trends in population average exposure levels from 1960 to 2020, and the assumption that these trends naturally stop in 2020 unless prompted by further interventions.
• Variance parameters (between and within person variance), and the assumption that they do not change over time
• The assumption that everyone works (within this physically demanding manual job) until pension age if didn’t develop LC or died
• Assumes everyone works in the same job and their exposure only drops in line with population level trends
• Interventions apply to all individuals in the study,
• Interventions apply immediately after implementation
• Assumes a linear-dose response relationship, does greater cumulative exposure increases have greater effects on the increased risk (are you more susceptible after a build up of exposure).
• Does exposure risk differ by age? Possible link to the cumulative exposure – response relationship?
• Does decay risk differ by age? Younger individuals may have more active clearance mechanisms than older participants.
• How exactly does the influence of multiplicative effects influence the annual incidence and the intervention effects
• Study doesn’t account for multiple health events/competing risks (e.g. develop a respiratory illness), partly due to influence these would have on the exposure experienced, and the subsequent exposure outcome definitions we would need to determine.

These tools serve as a summary of the results we have observed. Here they can be explored further by the user, who within the definitions outlined above can see for themselves the influence of an intervention effect under a modifiable exposome. This study aimed to generate a set of results that could be applicable within a variety of scenarios, however to make it relatable we have based this work on a construction site scenario using a set of atmospheric exposures and their influence on a single outcome. This means the results, despite our best intentions, are largely applicable to the scenario and less generalisable that we had hoped. Further work would do well to build on our work to improve this aspect.

This might involve change in scenario’s like: 

• Non-atmospheric ambient exposure, such as shit pattern work, or shorter bursts of work-related stress
• Alternative, more common, or non-binary (i.e. grade of severity) health outcome
• Multiple health outcomes, with competing risks, and their subsequent influence of the exposure levels
• Acute exposure-response relationships rather than long latency periods
• Dose-response relationship
• Multiplicative effects of two, or more, exposures
• Influence of secondary characteristics, such as sex, ethnicity, socio-deprivation on exposure response effects.

These may help further a more generalisable understanding of the influence of the exposome, and the impact of interventions within a wider number of settings.

Additionally, we currently have a number of methods that attempt to predict future burden of disease in a working population. These methods vary from the relatively simple use of Population Attributable Fractions to the more complex Age-Period-Cohort and G-methods. Some inconclusive work has been done to understand the accuracy of these methods in a practical setting when predicting future disease burden. This work we have proposed here would provide be a useful bases to apply these methods and improve our understanding of our ability to accurately predict future disease burden with and without health interventions.