Introduction

Greenhouses produce the substantial quantity of vegetables and flowers especially in autumn and winter months. However, reports on the health risk associated with work in these operations are scarce (Abell et al. 2000; Restrepo et al. 1990; Sallmén et al. 2003). As the workforce is mostly female, a question of reproductive health effects of greenhouses work conditions is natural. Work in greenhouses is performed in warm microclimate during the most time of the year, involves usually moderately intense or heavy work. The working conditions in greenhouses might involve also indirect exposure to pesticides resulting from contact with flowers and vegetables previously treated with pesticides. While pesticides are extensively used on flowers to protect them against fungi and insects, with vegetables pesticides are used sporadically and main emphasis is on the use of biological agents.

Although pesticides are regularly used in agriculture, relatively little is know about possible adverse health effects, especially reproductive effects, due to the occupational exposure (Hanke and Jurewicz 2004). Several epidemiological studies conducted over the last two decades were focused on reproductive outcomes in populations exposed to pesticides. The results of the analyses showed that employment in agriculture increases the risk of specific morphological abnormalities in sperm, including reduced sperm count per ejaculate and lower percentage of viable sperms (Abell et al. 2000; Ratcliffe et al. 1987). The data on the effect on employment in agriculture on the time to pregnancy revealed a relationship between the decreased fecundability ratio and pesticide exposure (de Cock et al. 1994; Sallmèn et al. 2003). The analyses indicate that parental employment in agriculture could increase the risk of congenital malformations in the offspring, particularly such as orofacial cleft (Nurminen 1995), musculoskeletal (Hemminki et al. 1980) and limb defects (Engel et al. 2000). The data on the effect of occupational exposure to pesticides on birth weight are inconsistent. Although most of epidemiological studies do not reveal a significantly increased risk of small-for-gestational-age birth (SGA), a slower pace of fetal development corresponding to SGA in the population of women exposed to pyrethroids has been recently reported (Hanke et al. 2003).

Maternal work during pregnancy, especially with high work-related physical exertion, is still considered one of the most prevalent risk factors of negative pregnancy outcome (Nurminen et al. 1989; Homer et al. 1990; Koemester et al. 1995; Makowiec-Dąbrowska and Siedlecka 1996; Spinillo et al. 1996; Hanke et al. 1999; Escriba-Aguir et al. 2001). Although in Poland from the legal point of view pregnant women may perform only light work (i.e. with work-related energy expenditure below 700 kcal/shift), recent survey revealed that about 20% of working pregnant women, at least in the first pregnancy trimester, performed jobs with higher energy expenditure (Makowiec-Dabrowska et al. 2003).

The aim of the present study was to investigate whether the work in greenhouse during pregnancy adversely influenced infant birth weight and, if so, which of the two main potential hazards typical of greenhouses environment (heavy physical work and exposure to pesticides) played the major role in this process.

Method

Study population

The list of 14 major greenhouses (each above 5 ha) growing vegetables (cucumbers and tomatoes) was obtained from the Polish Chamber of Horticulture. Nine greenhouses were involved only in this type of production while five, in addition to vegetable growing, were running also some floriculture production (cut flowers and flowerpot flowers).

Between January 2001 and December 2003, 460 women at the age below 45 years, married or who lived with a partner and who had been working for a period of at least 2 years in greenhouses in Poland, were asked to participate in the project. Forty-five years age limit was adopted because we assumed that time span of more than 20 years will negatively influence the reliability of information about working conditions during pregnancy. All women who were actually employed in greenhouses were willing to participate in the study. After interview we excluded 19 women who have never been pregnant. Study group included 1122 pregnancies, which occurred between 1977 and 2001: 386 pregnancies of women who worked in greenhouses, 276 pregnancies of those who worked out of greenhouses, and 460 of women who became pregnant during maternity leave. The definitions of the study population as well as specific examined subgroups are presented in Fig. 1. We classified pregnancies of women working in and out of greenhouses on the basis of the work-related energy expenditure during mother’s work into three groups: A (200–700 kcal/shift—light work); B (701–1000 kcal/shift—moderately intense work); and C (1001–1200 kcal/shift—heavy work). Classification was made for workposts on which women worked for the longest time during pregnancy. Usually it was the work post she worked at the beginning of pregnancy. It has been a common practise in Poland that women works the whole pregnancy on the same position if the pregnancy is without complication. According to 1996 regulations pregnant women were allowed to perform only light work. However this regulation has been slowly enforced, especially in rural regions.

Fig. 1
figure 1

Flow chart of definition of the study population

Full size image

Group A was involved mostly in tasks, which support the work of women taking care of plants or collecting vegetables/flowers. This group comprised women with high-risk pregnancies who had been transferred from work-posts requiring medium or heavy physical work.

The major tasks for group B were preparation of ground for plants (mostly tomatoes and cucumbers), setting, quilting and cultivation of young plants and preparation of seedlings for flowerpot flowers cultivation. Group C was involved in picking up vegetables, culture liquidation and greenhouse cleaning as well cultivation of the upper and bottom part of a plants. The heavy work on flowerpot flowers cultivation consisted of setting and cultivating plants, watering plants and preparation plants for shipment.

We excluded from the analyses of birth weight small group of pregnancies in the women performing light physical work in greenhouses (200–700 kcal/shift; group A), as we were informed by greenhouses administrators that this group comprised women with high-risk pregnancies who had been transferred from work-posts requiring heavy physical work. During their pregnancy the physician informed them that their pregnancy is at risk and they were able to perform only light work.

Control group consisted of women working in greenhouses at the time of survey, who during one or more pregnancies performed light work out of greenhouse (e.g. administrative jobs, work in stores and clothes sewing). This group was selected out of 276 pregnancies of women who performed work out of greenhouses. Classification to this category was based on the opinion of expert from Department of Work Physiology and Ergonomics of the Nofer Institute of Occupational Medicine.

Information about application of pesticides in 1997–2001 was received from persons responsible for chemical protection in each examined greenhouse. Trade names of the pesticides, names and amounts of active ingredients, type of cultivation and its area were abstracted from pesticide application registers run by each greenhouse operator. The data was obtained for 252 pregnancies (jobs in greenhouse). Based on the Pan American Pesticide Database (Orme and Kegley 2000–2004) (http: www.pesticideinfo.org) we identified pesticides, which were classified as reproductive and developmental (RD) toxins. Out of 196 pesticides classified as RD in that base, 17 were applied in the examined greenhouses during 1997–2001(Table 1). The reference group constituted of women working out of greenhouse with not exposed to RD pesticides.

Table 1 The list of active ingredients of the RD pesticides used between 1977 and 2001 in Polish greenhouses
Full size table

A questionnaire survey to collect demographic information, anthropometric characteristics (weight and height), job history, information about the duration of employment in greenhouse, type of job, reproductive history [spontaneous abortions, live births, the number and duration of pregnancy (less than 37 weeks and more than 37 weeks) birth weight and sex of each infant] and smoking was performed by specially trained interviewers. The questionnaire also included queries about the job characteristics (type of cultivation, tasks and duration of work) and protective outfit used during work.

Work-related energy expenditure was estimated during different field operations in randomly selected groups of women currently working in greenhouses involved in the study. The basis of estimation was the measurements of pulmonary ventilation and Datta–Ramanathan equation for convert these values into energy expenditure (Datta and Ramanatan 1969). The level of heaviness of work performed in greenhouses by particular mother was calculated on the basis of the estimation of work-related energy expenditure during work operations and description of work performed during pregnancy. Assessment of work-related physical exertion for a mother working out of greenhouses during pregnancy was based on job specification, using the catalogue of occupational characteristics prepared by experts from the Department of Work Physiology and Ergonomics of the Nofer Institute of Occupational Medicine, Lodz.

The retrospective history was recorded to establish the pregnancies of each woman and the outcome of each pregnancy. Such outcomes reported by each interviewer were recorded according to the following definitions: (a) full-term baby—fetus born alive above 37th week of gestation, (b) preterm delivery—fetus born alive below 37th week of gestation and (c) spontaneous abortion—fetal loss till the end of 22 weeks of gestation. A malformed child was defined as one with some structural or functional defect as detected by the neonatologist.

Infant birth weight was abstracted out of Child Health Card, which was issued for every child identified in our project and contained data about child physical development till the time of adolescence. Women were inquired about amount and timing of smoking during each pregnancy.

Statistical analysis

The t-test was used to assess differences in mean birth weight calculated for pregnancy of women with selected characteristics. Logistic regression model was applied to evaluate the impact of pesticide exposure and intensity of work-related physical exertion on birth weight. All regression models were fitted using STATA 6 software. The model included also variables known to affect birth weight: maternal weight, age, smoking during pregnancy, education and the place of residence. Models considered inside correlation between individual observations for the same woman (Watier et al. 1997).

Results

Study population

Pregnant women who performed jobs classified according to work-related energy expenditure as heavy (group C) moderate (group B) and light (group A) compared to controls (light work out of greenhouse) had significantly more often only primary or only vocational education. The differences were statistically significant for groups B and C (Table 2).

Table 2 Study participants by level of workload
Full size table

The number of the women living in rural areas was slightly higher in the greenhouse workers than in the control, while controls were more often inhabitants of medium-sized towns. Significant differences were observed only for group B.

Smoking during pregnancy was more prevalent in all groups of the greenhouse workers as compared with controls. In all groups the most prevalent was category of women smoking during three pregnancy trimesters. The highest rate of women who had abstained from smoking during any period of pregnancy was noted in group B, while the lowest rates were recorded for groups A and C.

No significant differences in an average height and weight were found when each group of the greenhouse workers were compared with controls. Each of the three groups of women who worked during pregnancy in greenhouses was significantly older than controls. However, they have also slightly higher parity as judged by second and higher order pregnancies.

Most of the pregnancies of women working out of greenhouses (81%) occurred in 1977–1990, while less than 2/3 of pregnancies of women who worked in greenhouses occurred at that time period. The difference between group C and controls, and group B and controls were statistically significant.

Pregnancy outcome and job performed during pregnancy

Compared with controls, group A had the highest rates of spontaneous abortions. Similar trend was observed for preterm delivery (Table 3). We were not surprised as, acording to the information obtained from the women and administrator, we considered group A as highly selected, including mostly women with high-risk pregnancies.

Table 3 The level of workload (jobs in greenhouses and out of greenhouse) and pregnancy outcomes.
Full size table

Women in group B, and specifically in group C, had higher rates of spontaneous abortions, preterm delivery, LBW, LBW at term (indicator of SGA) and birth defects than controls; however, none of them reached level of statistical significance. In general, women performing any jobs in greenhouse had lower mean birth weight than those involved in light work out of greenhouses (controls). Nevertheless, the decrement was statistically significant only for group performing heavy work in greenhouses (p = 0.05) (Table 3).

Use of pesticides in greenhouses at the time pregnant women worked in greenhouses

Information about use of 17 active ingredients of pesticides classified in the PAN Pesticide Database as reproductive and developmental toxins (RD) was obtained for period of 1977–2001. Information was available for 252 (65.3%) pregnancies of women who worked during pregnancy in greenhouse. At least one of these active ingredients was used between 1977 and 2003 during 85.7% pregnancies. The most often used active ingredients of pesticides were: mancozeb (25.18%), fenbutatin-oxide (11.88 %) and triforine (11.35%). Pesticides classified as RD were more often used in 1977–1990 (91.4%) than in 1997–2001 (72.41%) (Table 1). Pregnancies in group B (89.9%) and C (85.5%) had more often potential exposure to RD pesticides.

Use of pesticides and infant birth weight

Mothers employed in greenhouse at work posts where RD pesticides had been applied delivered infants with slightly lower birth weight (3293 g) than those employed at out of greenhouses (3322 g) (p = 0.56) (Table 4). As the workload may seriously confound such observations, we run similar comparisons separately for women in groups B and C. Women in group B potentially exposed to RD pesticides had slightly higher mean infants birth weight than controls (p = 0.38) while those at group C almost exactly the same.

Table 4 Application of RD pesticides and birth weight by workload classes
Full size table

Multivariate analyses of factors affecting birth weight

To investigate the possible association between level of energy expenditure during one shift (8 h) and the infant birth weight, a linear model was developed (Table 5). The model considered inside correlation between individual observations for the same woman (Watier et al. 1997). This model enables estimation of the effects of such variables as heaviness of work, application of RD pesticides while controlling for the main cofounders (weight of the women before pregnancy, smoking during pregnancy, mother’s education and place of residence). Children of women performing jobs in greenhouses with exposure to RD pesticides had similar birth weight to reference group (Table 5). After controlling for several potential cofounders, we noted a significant effect of heavy physical work during pregnancy on infant birth weight. Smoking in the second trimester of pregnancy significantly reduced, while living in a big city significantly increased, the mean birth weight.

Table 5 Linear regression model of birth weight of level of workload and other selected variables. (N = 270)
Full size table

Discussion

The mean birth weight of infants whose mothers worked in greenhouse during pregnancy and performed heavy work (>1000 kcal/shift) was 177 g lower than that of those whose mothers worked out of greenhouse and performed light work (<700 kcal/shift). The respective estimate for moderate energy expenditure during work (100–700 kcal/shift) was 97 g.

Reproductive hazards have been a particular concern among scientists who have studied the greenhouse operations. The 1990 Restrepo report on Colombian’s cut flower industry found moderate increase in the prevalence of abortion, prematurity and congenital malformations among children conceived after either parent started working in greenhouse (Restrepo et al. 1990).

Greenhouse work also may interfere with conception. The study of 1767 female members of the Danish Gardens Trade Union by Abell and colleagues has revealed that female workers in flower greenhouse have reduced fecundability (Abell et al. 2000). Abell et al. (2000) also studied semen among ornamental-flower greenhouse workers. The median sperm concentration was 40% lower among men with more than 10 years of greenhouse experience than among men with less than 5 years’ experience.

Some indications that exposure to pesticide during pregnancy may affect infant birthweight came from studies of farmers. Polish female farmers exposed to pesticides in the first and second trimesters of pregnancy had infants with birth weight lower by 100 g than that of infants born to nonexposed women. The results were adjusted for pregnancy duration and other factors, including smoking, which might influence birth weight (Dąbrowski et al. 2003). In another study conducted in the same area, maternal exposure to synthetic pyrethroids in the first or the second trimester was associated with decreased birth weight (p = 0.02) (Hanke et al. 2003).

In the reported study of women working in greenhouses, no negative effect of pesticide (both classified as RD and no RD) exposure was observed. The possible explanation is that in vegetables greenhouses level of exposure is low, as operators rather relay on biological methods of crops protections.

Heavy physical effort at work during pregnancy constitutes a risk factor of SGA. Such findings were reported by some investigators (Nurminen et al. 1989; Hanke et al. 1999) but were not confirmed by others (Fortier et al. 1995). Heavy work is thought to reduce the blood volume available to the fetus and, consequently, the amount of oxygen and nutrients (Morris et al. 1956). A negative correlation was found between physical workload and growth retardation in the newborns of workingwomen in Finland (Nurminen et al. 1989). The adjusted OR calculated for mothers performing work with a moderate physical load in the third trimester who gave birth to SGA babies was 2.4, compared with that for mothers at sedentary work. A study conducted in Poland demonstrated an increased risk of SGA for moderate energy expenditure during work (Makowiec-Dąbrowska and Siedlecka 1996). Another study conducted in Poland also found an excessive risk of SGA in the group reporting heavy physical effort at work (Hanke et al. 1999). Mazurkewich et al. (2000) also noted that heavy physical work during pregnancy was the risk factor for preterm delivery (OR= 1.22; 95% CI 1.16–1.29).

In our study no relationship was found between energy expenditure during work and prematurity. Most probable explanation was that women performing medium and heavy physical work who were at risk of preterm delivery were identified by obstetricians and moved to light work. The observed decrement in birth weight was probably due slower pace of fetal development corresponding to the SGA effect. However, due to lack of exact estimates of pregnancy duration, such hypothesis could not be directly tested and the analyses utilized LBW at term as a surrogate of SGA did not confirmed it.

Potential confounding factors

The examined population included the homogenous group of women currently working in greenhouses. The comparisons were made between subgroups of pregnancies reported in this population. We did adjust for the well-recognised factors that affect birth weight, such as maternal weight, age, smoking, place of residence and mother’s education. We have not controlled for parity and year the pregnancy took place as the first factor did not differentiate examined subgroups and the second was not found to be related to any of examined poor pregnancy outcomes.

Reliability of pesticide exposure assessment

Our study involved a reproductive assessment of the relationship between pesticide use in greenhouses and adverse pregnancy outcomes and, therefore, was subject to some limitations typical of this study design. However, we have undertaken several steps to obtain valid assessments. Information about application of pesticides between 1997 and 2001 was received from persons responsible for chemical protection in each examined greenhouse. Pesticide trade names, specifications and amounts of the active ingredients, type of cultivation and its area were abstracted from pesticide application registers run by each greenhouse operator.

Although biological monitoring seems to be the optimal method to assess individual exposure to pesticides, such an approach is possible only in prospective cohort studies. This method has not as yet been applied in the retrospective assessment, although it has been strongly recommended for this kind of study (Copper et al. 2001). On the other hand, the results of the US National Cancer Institute project indicate that the data obtained via a standarized questionnaire may be reasonable indicators of occupational exposure if biomarker data are not available (Blair and Zahm 1992; Hernandez-Valero et al. 2001). Epidemiological studies on environmental and occupational risk factors of poor pregnancy outcomes and childhood late health outcomes will continue to utilise the case-control approach. Further progress in clarifying the relationship between prenatal exposure to pesticides and the health outcomes under study will heavily rely on success in developing valid methods for retrospective risk assessment of parental occupational and residential exposure to pesticides in the critical time-window.

Conclusion

No negative effect on birth weight of exposure to pesticide, including those classified as reproductive and developmental toxins, was observed. The exposure level in greenhouses producing mostly vegetables, although not measured, seems to be low due to introduction of biological protection of the plants.

In Poland, despite the fact that new legislation aimed at the protection of pregnant women and their foetuses has recently come into force, almost every tenth woman reports heavy physical work at her job during the first trimester. We presume that the fear of losing the job may keep a woman from revealing pregnancy to her employer (Makowiec-Dąbrowska et al. 2003).

Our results indicating that infants of mothers with heavy work during pregnancy had lower mean birth weight than infants whose mothers had light work out of greenhouse should be treated as strong argument for monitoring effectiveness of enforcement of existing legislation both on local as well as national level.