Introduction

Baker’s asthma is one of the most frequent occupational asthma (OA) in industrialized countries (Jeebhay et al. 2019; Page et al. 2010). The prevalence is estimated to be from 4 to 13% (Jeebhay et al. 2019). In France, flour was still the most frequent causative agent of OA during the period 2008–2014 (Iwatsubo et al. 2016). Baker’s asthma has seen the list of its causative agents expand over the years with the evolution of baking techniques, with the use of cereal flour, fungal enzymes, other ingredients like white egg proteins but also the presence of storage mites, arthropods and molds (Quirce and Diaz-Perales 2013).

Bronchial hyperresponsiveness (BHR) is considered the physiological hallmark of asthma, whether occupational or not (Malo et al. 2015; Lipinska-Ojrzanowska et al. 2020). Thus, it is a common hypothesis that BHR is a stage on the pathway to asthma (Malo et al. 2015; Radon et al. 2016). A longitudinal follow-up of the BHR status in populations exposed to asthmogens seems therefore to be relevant.

Apart in asthma, BHR is also observed in other lung diseases, such as chronic obstructive pulmonary disease (Borak and Lefkowitz 2016). BHR was shown to be a predictor of irreversible loss of lung function in asthmatics with or without pulmonary symptoms (Borak and Lefkowitz 2016) but also in asymptomatic individuals from a Swiss population-based cohort (Brutsche et al. 2006) and in followed-up firefighters (Aldrich et al. 2016). However, BHR was not predictive of lung function loss in a recent relatively short-term study among aluminum smelter workers (Abramson et al. 2020).

An association of occupational flour dust exposure with a decreased lung function has been found in flour mill (Bagheri et al. 2013; Demeke and Haile 2018; Lagiso et al. 2020; Mohammadien et al. 2013; Zamani et al. 2019) or bakery workers (Fahim and El-Prince 2013; Moghaddasi et al. 2014; Said et al. 2017), as recently reviewed by Zamani (Zamani et al. 2019). In these studies, the occupational flour dust exposure was solely represented by the exposure status (Fahim and El-Prince 2013; Lagiso et al. 2020; Moghaddasi et al. 2014; Mohammadien et al. 2013), the duration of exposure (Demeke and Haile 2018; Said et al. 2017), or by the atmospheric flour dust concentration at the time of the data collection (Bagheri et al. 2013; Zamani et al. 2019). Different spirometric parameters were affected depending on the study (Zamani et al. 2019). All these studies were cross-sectional, with the exception of the two-year follow-up study by Zamani et al. Indeed, many of the published longitudinal studies in bakers concerned rather apprentices (Acouetey et al. 2013; de Meer et al. 2003; De Zotti and Bovenzi 2000; Demange et al. 2018; Gautrin et al. 2000; Skjold et al. 2008; Tossa et al. 2010; Walusiak et al. 2002). These studies brought knowledge about the time course of the work-related respiratory symptoms or diseases at the beginning of exposure. The highest incidence of these symptoms seems to be in the first 6 to 12 months of employment (Radon et al. 2016). The recognised risk factors are atopy (Skjold et al. 2008), pre-existing sensitization to occupational allergens (de Meer et al. 2003) and BHR (de Meer et al. 2003; Demange et al. 2018; Radon et al. 2016; Skjold et al. 2008). We identified four longitudinal studies in bakers at work (Cullinan et al. 2001; Florentin et al. 2014; Kim et al. 2013; Remen et al. 2013). None of these studies were focused on BHR and lung function evolution, except for Kim et al., who focused on BHR changes after 6 month job relocation in a small population of 12 bakers.

The present longitudinal study aimed at describing the natural history of BHR and FEV1 decline by identifying their respective predictors in a population of bakers and a group of non-exposed controls, the latter providing a baseline for the exploration of the dose–response relationships.

Materials and methods

Study design and inclusion criteria

The study design was a longitudinal follow-up of the workforces at 4 industrial bakeries, with visits in 1992, 1994, 1996, 2001, 2003 and 2005 (n = 110) and of a non-exposed group of blue-collar workers (n = 38) at a salt-packing factory with visits in 1993, 2002, 2004 and 2006. Each visit comprised a clinical examination, a standardized respiratory questionnaire, occupational allergen skin prick tests (SPT), pulmonary function tests and a methacholine bronchial challenge test, all carried out at the workplace. The cross-sectional analysis of the initial data collected in 1992 was previously published (Bohadana et al. 1994). The baker jobs were croissant making, dough mixing, special breads making, general bread making, control and packaging, deliveries, cleaning, oven handling. The non-exposed salt-packing factory workers were employed in production, maintenance, or quality control departments. All workers who volunteered to participate were included. However, workers with an acute respiratory infection were not included in the visit. The non-exposed subjects were included if they had not been occupationally exposed to flour dusts or to another respiratory health hazard. As the aim of this paper was to determine long-term predictors, we included in the analyses all subjects who had attended at least two visits. Figure 1 presents the flowchart of the participation of the subjects in the study.

Fig. 1
figure 1

Flowchart of the study. *Drop-out unrelated to health

Full size image

Respiratory health status

Respiratory symptoms were recorded at each medical examination by the same experienced interviewer using the European Coal and Steel Community questionnaire (Bohadana et al. 1994; Minette 1989). The questionnaire included detailed smoking habits, history of respiratory diseases and childhood allergic diseases such as asthma and eczema. Clinical atopy was defined as having childhood allergic diseases (asthma or eczema). At each visit, we abstracted respiratory symptoms defined as a wheezing chest or a runny nose in the absence of any cold.

Spirometry was carried out by the same 3 experienced technicians using the same electronic spirometer (Spiro-Analyzer ST 300, Fukuda Sangyo Co. Tokyo, Japan) in compliance with the European Respiratory Society recommendations (Quanjer et al. 1993). The observed/predicted FEV1 ratio was used for description, using the reference equations from Quanjer (1993).

We used 3 additive doses of methacholine delivered by a nebulizer (Mediprom FDC 88—Paris, France), according to an abbreviated version of the methacholine bronchial challenge (MBC) test (Gardner 1979). This corresponds to a maximal inhalated dose of  10.5 µmol, i.e., 2 100 µg. The first dose of 0.5 µmol is equivalent to one puff from an aerosolized solution with a concentration of 0.36 mg/ml, using a nebulizer delivering 0.14 ml/minute during 2 min. The second dose of 2.5 µmol is equivalent to one puff from an aerosolized solution with a concentration of 1.79 mg/ml. The third and final dose of 7.5 µmol is equivalent to one puff from an aerosolized solution with a concentration of 5.36 mg/ml.

The challenge test was discontinued either after the inhalation of the last dose of methacholine or if the FEV1 fell by 20% or more below the baseline value obtained prior to MBC testing. In that case, subjects were classified as having a positive MBC test, i.e., having bronchial hyperresponsiveness (denoted BHR in the sequel).

Skin prick tests to occupational allergens

Three SPTs were performed using wheat flour, rye flour and alpha amylase, respectively (Stallergenes Laboratories, Fresnes, France). A 9% codeine phosphate solution was used as a positive control and the 50% glycerinated diluent as a negative control. A positive SPT was defined as a wheal diameter at least 3 mm larger than those obtained with the negative controls after 20 min. Sensitization to occupational allergens was defined at each visit as a positive reaction to at least one of the three allergens tested.

Exposure assessment

In each bakery, some full-shift dust samples of the inhalable fraction were obtained in order to assess the exposure of each job assignment. Closed-face filter holders were used (Millipore, Massachusetts, USA), housing pre-weighed 37-mm diameter polyvinyl chloride filters (Pall Gelman Sciences, Champs sur Marne) connected to portable battery-operated vacuum pumps (Escort, MSA, Châtillon sur Chalaronne) sampling at airflow rates of 1 L/min. The PVC filters were weighed twice in the laboratory during their preparation before sampling and again in the same laboratory after sampling. These weighings were done in a controlled and constant atmosphere for heat and humidity after a stay of at least twelve hours in the room. Before the first weighing, the filters were treated with Triton to eliminate electrostatic charges. The pumps were calibrated before and after sampling. Generally, the samples were taken on workers with the filter holder fixed near the breathing zone. In exceptional cases, stationary sampling was used for team leaders in charge of administrative tasks in a room isolated from the workshop. They were reserved for workers present in this room in a static position and who did not intervene directly in the manufacturing process. The pumps and sample holders were placed a few tens of centimeters away from their upper respiratory tracts. This measurement of the airborne exposure could thus be considered as representative of their individual exposure.

For a given job, the exposure level was calculated as the mean value of the dust concentration measurements obtained for this job. As these measurements could not be considered an exact measurement of the chronic exposure, these exposure means were categorized into three groups, < 1 mg/m3, 1-5 mg/m3, > 5 mg/m3 which were considered reliable by the experienced industrial hygienist (MH) who was part of the research group. This exposure level was assigned to all workers of the job. Exposure measurements were repeated several times over the entire study period, at least each time the working conditions (new equipment, automation of a production line, etc.) changed.

Statistics

According to the main objective of this study, we studied the respective determinants of BHR and respiratory symptoms and the determinants of the evolution of FEV1 during the study follow-up.

Firstly, we studied whether BHR, respiratory symptoms and their joint occurrence were associated with gender, cumulative smoking, clinical atopy, sensitization to occupational allergens and exposure to flour dust (level and duration) using a mixed logistic regression with the subject as a random factor to account for the longitudinal nature of the data. Secondly, the repeated FEV1 measurements were modeled using linear mixed models with the subject as random effect. Several models were considered, including cumulative smoking, pre-visit and concurrent BHR, sensitization to occupational allergens, duration and exposure level, as independent variables. All analyses were adjusted for the technician performing the spirometry. Age, gender and height were taken into account by subtracting the predicted FEV1 from the observed values.

The analysis was carried out with STATA statistical software (STATA, College Station, TX, USA).

Results

One hundred and ten bakers and 38 non-exposed subjects attended the inclusion visit and at least one follow-up visit, representing 483 visits. That is 3.3 visits per subject with a mean duration of follow-up of 6 years among the bakers and 8 years among the non-exposed subjects. Table 1 gives a basic description of subjects at inclusion, from inclusion to last follow-up and at the last visit. At inclusion, clinical atopy was less often reported by bakers (11%) than non-exposed workers (21%), whereas sensitization to occupational allergens was more often observed in bakers (20% versus 16%). At the last visit, this prevalence was 30% in bakers versus 10% in non-exposed workers. At inclusion, the exposed workers had been working as bakers for 10.6 years on average, and the most frequent category of exposure level to inhalable dust was between 1 to 5 mg/m3 at inclusion and at the last visit.

Table 1 Basic description of population study
Full size table

Table 2 describes the health outcomes of interest according to the exposure status, at inclusion and the last visit. At inclusion, the prevalence of BHR was 24% in bakers and increased to reach 36% at the last visit. In non-exposed workers, this prevalence increased slightly between these two visits (8 and 11%, respectively). Moreover, in bakers at both visits, the prevalence of BHR increased with duration of exposure. Such a pattern was not observed with the exposure level. The prevalence of respiratory symptoms was similar at inclusion in exposed and non-exposed workers (38% versus 34%, respectively), while it was higher in bakers at the last visit (52% versus 40%, respectively). For this outcome, no pattern with duration of exposure or mean level of exposure was observed. The observed FEV1 was close to the predicted value in bakers both at inclusion and the last visit and slightly higher in non-exposed workers. In bakers, at both visits, the FEV1 in percent predicted decreased with duration of exposure. As for BHR prevalence, no pattern was observed with the exposure level.

Table 2 Description of the health outcomes
Full size table

The probability of having BHR (Table 3) doubled when the duration of exposure increased by ten years (adjusted OR = 2.2, 95% CI = [1.0;5.0]). Female gender and the number of pack-years for ever-smokers also significantly increased the probability of having BHR. The probability of reporting respiratory symptoms was multiplied by almost 4 when the duration of exposure increased by ten years (adjusted OR = 3.6, 95% CI = [1.4;9.4]). Clinical atopy was significantly associated with respiratory symptoms (adjusted OR = 16.5, 95% CI = [1.7;157.1]). When considering having BHR and at least one respiratory symptom simultaneously, duration of exposure and clinical atopy were significant predictors. The exposure level had no statistical significant effect on any of the three outcomes considered, neither did a history of sensitization to occupational allergens.

Table 3 Models for the health outcomes according to a mixed logistic regression (OR [95%CI])
Full size table

Table 4 presents the results of two multiple linear regression analyses of ΔFEV1, the difference between observed FEV1 and its predicted value (Quanjer et al. 1993) according to gender, height and age. In the first model, ΔFEV1 decreased by almost 200 ml per ten years of exposure (adjusted β = -193.5, 95%CI = [− 268.0;− 119.1]). BHR at a preceding visit was significantly (although borderline) associated with ΔFEV1 even after adjustment for the significant effect of BHR at the current visit. Similarly to the results for symptoms and BHR, there was no effect of the exposure level. Finally, as expected, ΔFEV1 decreased significantly with smoking (by 6 ml per pack year). The second model excluding the (non-significant) exposure level and including the history of occupational sensitization confirmed the relationship with duration of exposure, the association with BHR at the current visit and cumulative smoking. Having had BHR at a preceding visit was now borderline non-significant. A history of occupational sensitization had no statistically significant effect on ΔFEV1. However, when excluding duration of exposure, history of occupational sensitization predicted a statistically significant decrease in ΔFEV1 (data not shown). For all outcomes, we tested interactions between the different independent variables. None was statistically significant.

Table 4 Models for ΔFEV1 (Observed − predicted according to gender, height and age) in ml according to a multiple linear mixed analysis (β coefficient [95%CI])
Full size table

Discussion

In our study, BHR, respiratory symptoms and their simultaneous occurrence depended on duration of exposure, which confirms, for BHR, the cross-sectional results previously published on the inclusion data (Bohadana et al. 1994). Analyzing respiratory symptoms, odds ratios for clinical atopy were high and statistically significant. FEV1 significantly decreased with duration of exposure and BHR at a preceding visit. This result persisted when accounting for the effect of BHR at the current visit. No effect of quantitative exposure level was observed.

In our study, the odds ratio of clinical atopy when analyzing BHR was high although not significant. This is consistent with previous analyses of data of a cohort of apprentice bakers and pastry-makers, in which atopy or degree of sensitization was associated with incidence of BHR (Tossa et al. 2010; Wild et al. 2017). Duration of follow-up in that study was shorter (18 months) than in the present study, and study populations were younger (mean age about 18 years) with no previous occupational exposure to flour. A possible explanation of the less clear-cut effect of atopy could be that atopy has an effect on BHR in the first years of occupational exposure but no later effect on the onset of BHR. Indeed, the workers in our study had finished their apprenticeships since an average of 11 years at inclusion. However, it is noticeable that the prevalence of atopy was lower among bakers than in non-exposed subjects in our study. This suggests a selection effect with an early drop-out from the job of the subjects with atopy previous to the recruitment into the study. Such an effect was shown earlier in earlier studies among pastry-making or hairdressing apprentices (Iwatsubo et al. 2003; Monso et al. 2000). It could also be a healthy hire effect, by which subjects with childhood allergic diseases avoid high risk occupations, as suggested by results from a British cohort (Butland et al. 2011). As in the Butland study, we used a definition of atopy based on self-reported childhood allergic diseases, which is different from the definition used in the former cited studies (Monso et al. 2000; Tossa et al. 2010; Wild et al. 2017), based on specific IgE or SPT as in the standardized definition. This definition is the recommended one (Johansson et al. 2001), unfortunately we could not perform any of these investigations and used instead a definition based on a questionnaire. However, it has been suggested that a questionnaire may capture a more consistent chronic picture of atopy (Hoppin et al. 2011) than these biochemical measures. Both prevalences of IgE sensitization or allergen skin test reactivity in (not elderly) adults have been found to decrease over 20 (Amaral et al. 2016) and 10 (Warm et al. 2012) years of follow-up. However, self-reporting symptoms have other defaults such as recall bias.

In our study, we failed to observe a statistically significant effect of occupational sensitization on BHR, respiratory symptoms or ∆FEV1, although in the literature, sensitization to occupational allergens has been observed to be a risk factor for work-related symptoms in bakers (Droste et al. 2005; Harris-Roberts et al. 2009; Hur et al. 2008; Jacobs et al. 2008; Mbatchou Ngahane et al. 2014; Olivieri et al. 2013) or for exercise-induced bronchoconstriction (Minov et al. 2013). Moreover, high levels of flour-specific IgE have been found to be good predictors of a positive specific inhalation challenge test (van Kampen et al. 2008). A cross-sectional study in bakers did not observe a relationship between occupational sensitization and BHR (Storaas et al. 2007). The authors hypothesized a non-IgE mediated mechanism causing BHR and airway symptoms based on their former works (Storaas et al. 2005). Concerning our study, we could only test three occupational allergens (wheat flour, rye flour and alpha amylase) and might have missed other relevant ones. As mentioned in the introduction, the use of a large variety of additives has developed during the last decades (Quirce and Diaz-Perales 2013). However, this does not mean there was no effect of occupational sensitization in our study. Indeed, when no exposure variables were present in the model (data not shown), FEV1 was significantly lower among workers with occupational sensitization. The fact that occupational sensitization was only a predictor of a decline in FEV1 when duration of exposure was not included might mean that exposure predicts this decline and sensitization independently, but that the latter is not an intermediate in the causal relation between exposure and decline in FEV1. In our study, the prevalence of BHR and symptoms depended on gender and atopic background. Similar to other studies, a higher prevalence of BHR has been observed in women than in men (Brutsche et al. 2006) and in atopics than in non-atopics in a previous study in bakers (Pavlovic et al. 2001).

We identified only few studies exploring the longitudinal decline of FEV1 according to BHR (Abramson et al. 2020; Aldrich et al. 2016; Brutsche et al. 2006). A population-based study (Brutsche et al. 2006) found a relationship between FEV1 and BHR but did not separate past and concurrent BHR as determinants of FEV1. In new-employed aluminum smelter workers, Abramson et al. did not find any predictive value of BHR at start of employment concerning the decline of lung function (Abramson et al. 2020). The median of follow-up was 4 years, which might be too short to be able to observe a detectable functional decline. A study with a longer follow-up in firefighters (Aldrich et al. 2016) found that BHR at 12-year follow-up was associated with an estimated 15.4 ml/y greater FEV1 decline than experienced in case of no BHR at follow-up. In our study, the sum of the pre-visit and current BHR effects corresponds to a 180 ml approximated FEV1 loss during 13 years of follow-up, equivalent to a 14 ml/year loss. This is roughly in-line with Aldrich et al.’s results (Aldrich et al. 2016), although the context of exposure is very different: a massive exposure to inorganic dust and gases for the World Trade Center’s collapse-exposed firefighters and a chronic exposure to mostly organic dust for the bakers. However, the analyses in firefighters were adjusted on the changes in FEV1 shortly before and after the 9/11/2001, controlling for the effect of this massive exposure. In our study, duration of exposure predicted an additional decrease in near 20 ml/y in FEV1, independently of the effect of BHR. In Aldrich et al.’s study, the exposure after the World Trade Center’s collapse was not taken into account, 76% of the participants being retired at the end of follow-up (Aldrich et al. 2016).

In our study, we did not observe any dose–response relationships using the quantitative exposure. As the exposure estimation was based on on-going exposure measurements and assessments by industrial hygienists, it is unlikely that it was due to exposure misclassification. A more likely mechanism would be that the high-exposed tasks are only performed by workers, with potentially low sensitization, who withstand these high exposure levels. It could be that workers themselves move to lower-exposed tasks or are more careful, e.g., wearing protective equipment from the appearance of the first work-related symptoms.

Our study has several strengths. First, we followed-up the population over 13 years with the same study staff using the same protocol. Second, for ten bakers (9%) only, the reason for drop-out was unknown, and two bakers only dropped out for reasons related to occupational health. This loss to follow-up rate is quite low when compared to studies on apprentices (e.g., 20% in (Tossa et al. 2010)). This could be due to the fact that all the data collection was done at the workplace. The corresponding drawback was that we had to use an abbreviated version of the MBC test for ethical reason. Subjects lost to follow-up in our study were older, had a shorter duration of employment but had a greater FEV1 than the other bakers (data not shown). However, no statistical difference between them and the other bakers was observed. Third, an attrition bias during the follow-up seems unlikely as no difference in health respiratory status was found at the last visit.

The main weakness of this study is the relatively low sample size, due to small-medium size of the working resources in that activity sector in France. In particular, the low number of subjects in the non-exposed group might be considered a problem. One must, however, bear in mind that the results of the present study do not rely on an exposed vs. non-exposed comparison. The non-exposed population is taken as a baseline exposure category in our search of dose–response analyses, in the present paper with duration of exposure and levels of exposure. Thus, the inclusion of the non-exposed subjects strengthened the power of detecting such a dose–response. Moreover, the inclusion allowed to better take into account confounders. Indeed (data not shown), repeating the analyses without the non-exposed confirmed the effect of the duration of exposure. On the other hand, some confounding factors (pack-years, history of BHR) were no longer significant.

A second possible weakness is the absence of systematic individual exposure measurements which led us to use a semi-quantitative exposure levels in three categories in whose reliability we could trust. The sketchy individual exposure measurements would have been unreliable. Moreover, the measurements were obtained using a 1L/min flowrate, which was up to date at the time of the measurements but is lower than the flowrates (2.5–3L/min) which are recommended nowadays. The precision of these suboptimal measurements was, however, deemed sufficient for the classification in the three exposure categories used.

Finally, a selection effect with an avoidance of the job or an early drop-out from the job of the subjects with atopy cannot be excluded, as discussed above. Despite this healthy worker selection effect, we could observe detrimental effects of exposure on bronchial responsiveness and on FEV1.

Conclusion

In flour-exposed industrial bakers, length of exposure and smoking are long-term determinants of BHR and of the decrease in FEV1. BHR at a preceding visit predicted lower FEV1 even when accounting for the effect of BHR at the current visit. Prevention is still needed for these workers.