Discussion
This cross-sectional study found that after adjustment for major known confounders, JEM-determined workplace exposures to airborne asthmagens were not associated with poor asthma control in patients referred to a regional NHS severe and difficult-to-treat asthma service. Application of the OAsJEM indicated that patients were likely to be exposed at high-intensity or medium-intensity to airborne airway irritants or sensitisers in the workplace, in 29–39% of cases. The main strength of this study is the availability of a whole baseline data set for a large regional NHS SA service, although with missing data. Besides any unidentified errors in transcribing data that may have occurred, the potential for information bias from clinical variables remains low since data had been entered at the point of clinical assessment using objective measurements (eg, blood count, BMI), via medical records (eg, comorbid diagnoses) or using standardised tools (eg, ACQ7 score). We have tried to understand the effect of confounding by demographic factors (eg, deprivation, gender, age) through models built using multivariate logistic regression.
It is then necessary to consider the limitations of the study. The UK SA registry does not currently report data on patients’ work and workplace exposures, nor on rates of diagnosed work-related disease.19 The Birmingham clinical database, which feeds into the UK SA registry, does retain some data on employment (yes/no) and current occupation (open comment), though this is insufficient to determine whether patients are genuinely exposed at the time of clinical review and database entry (eg, long-term sickness absence), and it is assumed for this study that all are present at work and neither working from home nor absent. Moreover, there are no data on the presence of workplace exposure control measures that may be in place to mitigate any risk that is assumed by the JEM, so prevalence estimates do not take these into account. JEMs are used to overcome the recall bias associated with self-reporting of exposures for population-based studies, and in general, do not allow for individual variations in the work process and workers’ tasks. Determining exposures using OAsJEM involves matching a patient’s job with the most appropriate code in the occupational classification (here, ISCO-88), and then, a second step of expert attribution according to the specific context of an individual worker. However, as data had not been collected specifically for this study, given open comments were often brief (eg, teacher, nursery nurse, secretary) meaning no further interpretation was possible, or at least had to be assumed and thus may have led to an information bias. There were only a small number of included patients with normal asthma control (mean ACQ7 score=2.9, SD=1.2; proportion with ACQ7>1.5=86%), reducing the likelihood of observing small effect sizes. Moreover, although rates of missing data were on the whole low, the proportion of missing data for ACQ7 was relatively higher (26%) and can be attributed to failure on the part of healthcare professionals to carry out the assessment at the baseline visit, for some individuals.
It was not possible to adjust for the type or degree of treatment prescribed at baseline, due to the extent of missing treatment data (eg, presence and dose of inhaled corticosteroids missing in 35% of cases within the whole database). Most patients had been referred as difficult-to-treat cases from secondary care asthma services (86%), so it seems unlikely that patients would have been without inhaled corticosteroids at all. The direction of any relationship between variables with significant associations is unclear, due to the retrospective and cross-sectional design and cannot be assumed to be causative-based solely on this study. The given occupational data do not take into account the consequences of previous employment, which may have been difficult to navigate for a worker and necessitated job change. Finally, there may be a healthy worker effect20 whereby more severely affected patients with work-related disease have left their employment by the time of assessment, reducing the overall work effect on asthma control within the study; this is an inevitable limitation of cross-sectional study design in working populations.
Published work suggests, at least indirectly, that a proportion of patients with SA may also have WRA. For instance, Vandenplas et al21 established a multinational European cohort of consecutive patients with OA (n=997), all with sensitiser-induced OA confirmed using specific inhalation challenge testing. 16% of included patients met their definition of SA, which comprised the need for high-level treatment (ie, Global Initiative for Asthma (GINA) treatment step 4–5), plus ≥1 of (1) ‘poor symptom control’ defined by the use of short-acting beta agonists ≥1 per day, (2) ≥2 severe exacerbations in the previous year or (3) airflow obstruction. WEA is common and has a prevalence similar in magnitude to OA, occurring in a median of 22% of adults with asthma.6 Data on health outcome are limited to a small number of cross-sectional studies but these show that WEA is associated with lower health-related quality of life (comprising mood, breathlessness, social disruption and health concern) than non-WRA, and greater self-rated disease severity, treatment use and work-related stress.22 23
This study indicates a significant proportion of patients with SA may be exposed to airway sensitisers and irritants, health, social care and education account for much (22%) of this work. Population studies and data from reporting schemes indicate that those in public-facing roles particularly, are exposed to a range of agents with both chlorine-based and quaternary ammonium-based cleaning agents and disinfectants featuring prominently; in addition, we have seen similar exposures in office workers with OA, traditionally considered to be low risk for WRA.24 25 A recent study from the French NutriNet-Santé cohort, observed significant associations between occupational exposures to sensitisers, irritants, cleaning products and disinfectants and uncontrolled adult-onset asthma, when compared with both normal participants and those with controlled asthma.26 It is perhaps then a surprise that a similar result was not observed in the current study, and differences may be explained at least in part, by the limitations of using a specialist clinical data set as outlined above. It would be unwise to dismiss outright the role of workplace exposures in SA, given their estimated prevalence and we advocate using both the SA disease registry and prospective longitudinal study of asthma outcome, that would variously take into account work-relatedness of symptoms, the latency of inhaled exposure, mitigating factors (eg, engineering controls, respiratory protective equipment) and job changes. Broadening the national registry data requirement at baseline and follow-up to include current job role and level of exposure, would give important insights into the contribution and impact of workplace exposures on SA.