Question 1: What is the best method to assess the occurrence of occupational asthma?
Theoretically, two epidemiologic approaches can be used to assess the occurrence of occupational asthma in a given population:
1. Assessment of the occurrence of occupational asthma per se.
2. Assessment of adult-onset asthma and the attributable fractions and population attributable fractions due to specific occupational exposures.
The former approach involves identifying individuals with diagnosed occupational asthma in a specified population, whereas the latter gives an estimate of the occurrence of occupational asthma at the population level, but does not identify individuals with occupational asthma.
Assessment of the occurrence of occupational asthma per se
For assessing the occurrence of occupational asthma in a given population per se one needs an estimate of the numerator representing the cases of occupational asthma and the denominator representing the population at risk which produces the cases, expressed in person-time.
For this calculation the numerator should be formed of verified cases of occupational asthma. Whether these should be incident (new) cases of occupational asthma or whether prevalent cases could be used depends on the purpose
of the estimation. Incident cases are more suitable than prevalent cases when assessing effects of occupational exposures on the etiology of asthma and for predicting future trends of public health burden from occupational asthma. Assessment of attributable fractions and burden of disease is usually based on incident data. However, prevalent cases of asthma may be relevant when assessing total burden to public health from current cases, i.e. when assessing burden from increased symptoms and health care utilization from asthma, because both prevalent and incident cases contribute to the burden due to illness, disability, health care costs and other consequences of asthma. However, it is not always easy to decide whether a case of occupational asthma is incident or prevalent. This difficulty is illustrated when considering a person who has had childhood asthma with a long intermittent period without asthma and recurrence of the disease in relation to a specific occupational exposure. In the opinion of the authors, this subject should be counted as having incident occupational asthma, if the disease would not have recurred in the absence of the specific workplace exposure. This view is based on definition of causality using counterfactual statements [9
]. According to counterfactual reasoning, the statement that John's occupational exposure caused his asthma is equivalent to saying that had John not experienced the occupational exposure he would not have developed asthma. Accepting this we see that for causality in adulthood it is irrelevant whether or not John had asthma in childhood from which he had recovered. More relevant is that John is without asthma when encountering the exposure of interest. Similar reasoning can be applied to groups of individuals and probabilities of developing asthma among exposed and unexposed.
The choice of the right denominator is a difficult task and is also dependent on the purpose of the assessment or study question to be addressed. For assessing the incidence of occupational asthma, the denominator should be person-years at risk of getting occupational asthma in the population for which the occurrence is assessed. This is easily calculated if we have a specific study population followed for answering the question on occurrence of occupational asthma, but needs more consideration when using existing population registries. At the population level, the right population for assessing incidence of occupational asthma is adult population that has ever been at work, as occupational asthma may be detected even after the person has quit his/her job, although in such a case the relevant time-period at risk may be limited to a few years. Sometimes only those in certain 'high-risk' occupational groups are included as being at risk, but as new causes of occupational asthma are constantly identified even in workforces that have traditionally not been considered as high risk-occupations, this approach may not be valid. When assessing occupational asthma incidence for certain occupational groups and in relation to specific exposures, people ever exposed to those specific exposures or working in those specific occupations form the relevant denominator by the same logic.
When assessing the prevalence of occupational asthma, the total adult population at certain point in time (for estimating prevalence) and certain time period (for estimating period prevalence) can be used as the denominator, as the purpose of such assessment is usually related to current burden to society from ill-health and heath care burden from all the existing cases.
The issues related to the accuracy and comparability of this type of assessment of occurrence of occupational asthma include questions on
• How to define occupational asthma?
• How should occupational asthma be verified?
• What is the coverage of the identification system for occupational asthma?
• What is the access to (occupational) health services?
• What are the workers' compensation practices?
• How does the whole social security system influence all these?
The definition of occupational asthma varies from country to country and so do the identification and diagnostic procedures. Countries with well developed occupational health care systems tend to have a broader coverage and this is likely to lead to higher estimates of occurrence. Well-functioning workers' compensation system may enhance detection of and reduce ill-health from occupational asthma, but on the other hand the compensation system may influence the diagnostic procedures and decisions so that cases with occupational asthma that would not be compensated might not be diagnosed at all. An example of this could be irritant-induced asthma for which the diagnostic procedures are less standardized than for hypersensitivity-type of occupational asthma and it may go undetected if the compensation system requires very specific diagnostic tests, such as specific bronchial inhalation challenges [10
]. On the other hand, if there are poor compensation and social security systems for those who develop occupational asthma, diseased workers may not seek medical help and continue working under exposure until the point where their asthma has become severe and causes severe disability.
To demonstrate some issues affecting the estimates based on this assessment approach, consider a comparison of registry data from two regions. The other data come from Finland, a high-income country with high quality and coverage of occupational health services and mandatory workers' compensation for employees. The other data come from West Midlands in UK, also a high-income country but with larger socio-economic differences and a substantially smaller proportion of the workforce with appropriate occupational health services. The population size of these two areas is quite similar: 5 326 314 inhabitants in Finland in 2009 and 5 267 308 inhabitants in West Midlands in 2002 [11
]. The working-age population (defined as 20-64 years old adults for the purposes of these calculations) in West Midlands was slightly larger (3 061 210) than in Finland (2 839 686). The number of new cases of occupational asthma reported to the Finnish Register of Occupational Diseases between 1983 and 2002 [13
] is presented in Figure and that reported to the SHIELD Register in West Midlands during 1980 to 2002 [14
] is presented in Figure . The beginning of the follow-up period showed an increasing trend in both countries, but the numbers seemed to stabilize around 1989, perhaps due to more standardized diagnostic procedures and reporting practices. The number of new cases in Finland during this period was between 270 and 400 per year, and that in West Midlands 70-140 annually. Calculated based on the mean number of new cases per year during 1989-2002, the average incidence rate of occupational asthma was approximately 0.10 per 1000 person-years in Finland and 0.03 per 1000 person-years in West Midlands. For comparison, the incidence rate of adult-onset asthma in the working age population was in Finland in the late 1990s 0.9 per 1000 person-years estimated in a population-based incident case-control study [16
]. The substantially larger incidence of occupational asthma in Finland is surprising in the light of a slightly smaller working-age population and West Midlands being an area with more traditional industry. The factors that may explain the observed difference are likely to be related to the occupational health services coverage and different reporting systems and practices. In Finland the reporting of new cases to the registry is mandatory, while in West-Midlands it is voluntary and based mainly on one occupational lung disease specialist centre. In Finland the diagnostic criteria were stricter requiring usually a specific inhalation challenge, while in West Midlands evidence from serial PEF measurements was accepted as a confirmatory test. This fact could be expected to decrease the rate of diagnosed cases in Finland compared to West Midlands, but as the results show an opposite trend, this further emphasizes the differences in coverage of health care services, compensation systems and reporting systems.
Figure 2 New cases of occupational asthma in Finland between 1983 and 2002 reported to the Finnish Register of Occupational Diseases . The colour indicates the causal agent. The population of Finland in 2002 was 5 326 314, 20-64 yr population 2 839 686, employed (more ...)
Figure 3 New cases of occupational asthma in West-Midlands, UK, between 1980 and 2002 reported to the SHIELD register [14,15]. The population of West-Midlands in 2001 was 5 267 308, 20-64 yr population 3 061 210, employed population 2 511 000.
This example demonstrates that even if certain factors that influence the figures in registries could be better standardized by international consensus agreements, for example the definition of occupational asthma and the diagnostic criteria, many factors affecting the results are difficult to take into account, e.g. access to occupational health care and compensation practices as well as reporting systems [17
]. Thus, national or regional registries that receive their data from the routine health care practices provide useful information for assessing trends in occupational asthma over time at national or regional level, but they do not provide very useful data for international comparisons. The absolute values are highly influenced by the health care and social security systems, so these figures may not be of value for calculating public health burden from occupational asthma.
Assessment based on the occurrence of adult-onset asthma and attributable fractions and population attributable fractions due to specific occupational exposures
For assessing the occurrence of occupational asthma and its public health impact based on attributable fractions and population attributable fractions due to specific occupational exposures, the following estimates are needed:
1. An effect estimate for a specific or all occupational exposures, in the form of incidence rate ratio (IRR), and
2. An estimate of attributable fraction (AF) calculated based on this, and
3. An estimate of population attributable fraction (PAF) calculated based on AF and the prevalence of occupational exposure(s) of interest in the population for which the assessment is made (Pe).
Incident rate ratio gives an estimate of the risk of developing asthma in relation to the exposure of interest and can be calculated according to the formula given in Table .
Measures of effect and burden of disease
Odds ratio (OR) can be used to substitute IRR with the following considerations. ORs from incident case-control studies with density sampling of controls are unbiased estimates of IRR, whereas ORs from cohort studies and prevalent case-control studies are reasonable approximates for IRRs when the prevalence of adult asthma is < 10%, which is true in most parts of the world.
Attributable fraction can be calculated by the formulas shown in Table . Attributable fraction assesses the impact of exposure by measuring its contribution to the total incidence under exposure [18
], so it is used to estimate the proportion of exposed cases for whom the disease could be attributed to the exposure of interest. This fraction is sometimes interpreted as the probability that exposure caused the case and called the etiologic fraction. However, such an interpretation should be applied with caution, remembering that this approach applies to a population rather than to an individual case. When investigating a potential case of occupational asthma in a specialist clinical center, it should be remembered that the referral process usually works towards increasing the probability that exposure caused the case. For example, detecting work-related pattern of asthma symptoms increases the probability that occupational exposure caused the disease in this specific case and also increases the likelihood of referral to the specialist center for further investigations.
The methodological issues related to the accuracy and comparability of assessment based on attributable fraction include the following questions:
• Is the estimate of exposure used for assessing the effect valid?
• Is the effect estimate valid?
When assessing attributable fraction, the estimate of exposure should be accurate in terms of giving a valid effect estimate, but it does not need to be representative of the entire population. The latter is needed when assessing the population attributable fraction. However, the effect is likely to vary according to the quantity of exposure, so assessment of the exposure distribution is preferable over dichotomous exposure assessment (which just compares all exposed to the unexposed). Having valid quantitative exposure estimate would make it possible to estimate the risk function according to different exposure levels, giving a better understanding of the risk for different quantities of exposure. To ensure a valid effect estimate, this should be based on a high quality study or on meta-analysis if enough high quality data is available to carry out such analysis. High quality study should be based on incident cases and be free of any majors biases and confounding and it would be valuable if a quantitative assessment according to the exposure level had been conducted. Assessment of attributable fraction is comparable between populations that have similar exposure (both qualitatively and quantitatively), so this assessment is less dependent on country-specific health care and insurance systems.
Assessment of population attributable fraction (PAF) requires an estimate of the prevalence of the occupational exposure(s) of interest for the population of interest (Pe) and an estimate of AF or IRR. The formulas for calculation of PAF are also given in Table . Population attributable fraction is defined as the reduction in incidence that would be achieved if the population were entirely unexposed [18
]. In public health it is interpreted as the proportion of cases in the population that could be prevented if exposure would be reduced to zero [19
The final step in assessing occurrence of occupational asthma based on estimating population attributable fraction due to specific occupational exposure(s) is calculating Excess burden of disease (EBD). Excess number of cases attributable to specific occupational exposures can be calculated by the formula shown in Table .
The issues related to the accuracy and comparability of the assessment of occurrence of occupational asthma based on population attributable fraction and excess burden of disease include the following questions:
• Is the estimate of exposure valid and representative of the population?
• Is the effect estimate valid?
• Is the estimate of incidence of asthma valid for the population of interest?
The estimate of exposure should be valid in relation to the effect estimate used and it should also be representative of the exposure, preferably exposure distribution, of the population for which the assessment is made. Again, the effect estimate should be from a high quality study or meta-analysis if there is enough high quality data available. And again, a risk function according to different exposure levels would be preferable over a risk ratio related to dichotomous exposure. The comparability of values given by this method of assessment for different populations depends on the similarity of exposures, both qualitatively and quantitatively. When assessing excess burden of disease another issue of importance is how to get a valid estimate of the incidence of asthma for the population(s) for which the assessment is made. This may be attained from a high-quality study or in some countries from existing registries. As an example of the latter, in Finland the National Institute of Social Security keeps a register on all cases of asthma who receive special reimbursement for asthma medications. Asthmatics usually receive this right and are registered about six months after their diagnosis is made, so the registry reflects rather well the incidence of asthma. However, it is not always possible to be sure that the case registered is new and that the case has asthma rather than severe COPD, so for a high quality study the medical records of the cases should be checked.
In summary, the method based on assessment of population attributable fraction and excess burden of disease provides good and comparable estimates of excess incidence (and public health impact) due to specific occupational exposures and specific occupational groups. This approach is not much affected by country-specific practices, such as occupational health access and workers compensation system, as is the assessment based on identifying cases diagnosed with occupational asthma per se.
However, the assessment of burden of disease from occupational exposures overall is accompanied by further methodological issues. The major one is the decision on which occupational groups or occupational exposures should be included in the exposed category and which occupations should form the unexposed reference category. Usually administrative employees are used as the reference group, as they do not have any heavy industrial-type of exposures, but more recent research has identified factors in office-type indoor spaces that seem to be determinants of adult-onset asthma [21
], so this may not be an optimal reference group. Including all those not currently at work in the reference category may be especially problematic, as some people may be unemployed because of previous work-related symptoms. If all other occupations than administrative work is included in the 'exposed', this will include many jobs that do not have any exposures that would be expected to induce asthma. Thus, this would dilute the effects of exposures with clear asthmagenic effect. If the decision about including occupations in the exposed category is based a priori
on studies reporting an IRR or OR > 1, it should be remembered that the choice of the reference category influences the risk estimate for the specific occupational groups. If a more restrictive criteria would be applied to include only occupations for which the lower 95% confidence interval of IRR or OR is > 1, one should remember that the sample size of the original study influences the width of the confidence interval, and thus the significance of the findings. In addition, it is difficult to foresee where new asthmagenic exposures turn up, as new chemicals are introduced into the working life all the time.