Pesticide Exposure or Pesticide Poisoning and the Risk of Depression in Agricultural Populations: A Systematic Review and Meta-Analysis

ABSTRACT Evidence suggests that farm workers represent a high-risk population for the development of depression. The aim of the study was to clarify the association between pesticide exposure/poisoning and depression. The Preferred Reporting Items for Systematic Reviews and Meta-Analysis were conducted in this systematic review and meta-analysis. Relevant studies were included through searching in PubMed, Scopus, Web of Science, ProQuest, CINAHL and pre-print services databases. A total of eight studies were included. A positive but non-significant association was observed between pesticide use and depression (OR = 1.123; 95% CI, 0.932–1.354, p-value = .221) while a significant positive association was observed between pesticide poisoning and depression (OR = 2.942; 95% CI, 1.791–4.831, p-value < .001). The present meta-analysis suggested clearly a significant positive association between pesticide poisoning and depression, confirming the hypothesis that pesticide poisoning is a risk factor of depression. It also revealed positive, yet non-significant association, between pesticide exposure and depression, a finding that enhances recent research but requires to be further supported by future cohort studies, including socioeconomic factors and biomarkers of depression.


Introduction
In the past decades, depression was the fourth most significant among factors that determine the global burden of disease; 1 according to WHO, depression is estimated to represent 7.5% of Disability-Adjusted Life Years (DALYs) 2 and affects more than 264 million individuals worldwide. 3Depression's harmful effects on physical, mental and cognitive functioning increase the suicidal risk and all-cause mortality leading eventually to greater health care services' use and costs. 4isturbances in neurotransmitter functioning and genetic factors have been implicated in the aetiology of anxiety disorders and depression.Depressive symptoms comprise anxiety, mood disorders, feeling of guilt and constant suicidal thoughts; 5 moreover, depression has major effects on perception of reality, comprehension of issues and expression of emotions and affects human as a whole, including physical condition as well as individual experience. 6pidemiological studies have reported higher prevalence of depression in particular occupational groups of individuals.Roberts and Lee found the highest incidence of depression in occupations being related to farming, fishing, and forestry, 7 while Sanne et al. found that male agricultural workers had the highest incidence of depression compared to non-agricultural workers. 8nxiety and depressive symptoms were reported in earlier studies in pesticide-poisoned subjects.Beseler et al. found that depression was significantly associated with a history of pesticide poisoning, but no similar finding was verified for low or high cumulative pesticide exposure. 9imilarly, male farmers who had experienced pesticide poisoning had a higher risk of depression but there was no significant association between cumulative pesticide use and depressive symptoms. 10Povey et al. reported, among other findings, that pesticide poisoning was associated with a higher possibility of depression, but this was not confirmed for low dose chronic pesticide exposure. 11igh level acute poisoning involves inhibition of the enzyme acetylcholinesterase (AChE), causing neurotoxic effects and disturbances in peripheral, autonomic and central nervous system function (the cholinergic crisis) and resulting in a constellation of physical, cognitive, and psychiatric symptoms. 12AChE is used as a biomarker for exposure to organophosphates (OPs). 13Many neurotransmitters, few of which are responsible for mood regulation such as serotonin, are disrupted by OPs, 14 which indicates the link between pesticide exposure and mood disorder, becoming in fact stronger in cases of acute poisoning. 9,15,16Among acute neurotoxic effects are fatigue, blurry vision, nausea, cough, sweating, abdominal cramps and in extreme cases even death; most of these symptoms are expected to be eliminated after exposure ends but it is possible to be continued for an extended period. 17,18fshari et al. reported that educational/behavioral interventions to promote pesticide safety among agricultural workers effectively improved their knowledge and beliefs; however, it was unclear whether those changes would be sustained over time 19 .Moreover, these interventions demonstrated limited effectiveness in changing participants' behaviors and in reducing their exposure to toxic pesticides compared to multifaceted interventions.
Although the farming system is directly reliant upon chemical pesticide applications, studies are limited on the field of the risks associated to cumulative pesticide exposure and its possible chronic effects or to ever use of pesticides, since they focus mostly on acute pesticide poisoning.The objective of the current systematic review and meta-analysis was to: i) define the overall Odds Ratio (OR) of cohort, case-control and cross-sectional studies published to date in order to shed light on the epidemiological relationship among both pesticide exposure and pesticide poisoning and the risk of depression and ii) draw a comparison between them.

Data sources and search strategy
In the present systematic review and meta-analysis, the selected studies were reviewed using the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA); 20 the PRISMA checklist is presented in Web Table S1 (supplementary material).One investigator searched PubMed, Scopus, Web of Science, ProQuest, CINAHL and pre-print services databases from inception up to February 21, 2022.PubMed search terms were (farmers OR "farm workers" OR "agricultural workers" OR "pesticide users" OR "pesticide applicators" OR "pesticide formulators" OR "pesticide mixers") AND (pesticide* OR "pesticide use" OR "pesticide exposure" OR "occupational exposure" OR "organophosphates" OR OPs OR insecticides OR herbicides OR carbamates) AND (depression OR "depressive symptoms").Similar search strategy, but less restrained due to a more limited number of search terms, conducted on the other databases.Systematic review protocol is available in PROSPERO registration number CRD42021249693.

Study selection and eligibility criteria
Given that this systematic review and metaanalysis offer a deep focus in a quite unexplored research area, no studies were excluded based on other criteria than our initial search ones, except from animal studies, case-reports and reviews.Inclusion criteria were the written language being English, the participation of adults subjects, the exposure of interest being general pesticide exposure/use or pesticide poisoning and the outcome being depression.In terms of "pesticide exposure/ use" we included studies concerning "ever/never use of pesticides" and we excluded studies concerning cumulative exposure since the exposure groups' classification (low, medium, high) was different from one study to another, rendering thus the results non-comparable (p.e. each exposure group was defined either by self-report question or by calculating cumulative days of pesticide use, median per study was not the same, etc.).Moreover, references from eligible articles were screened for additional detection of relative articles.

Data extraction and quality assessment
Extracted data included study characteristics such as authors, location, time of data collection, study design, participants, prevalence of depression, type of pesticide exposure, pesticide assessment measures, measuring tools for depression, confounding factors and level of analysis.A separate extract of unadjusted and adjusted effect sizes, where available, was done in order to investigate the impact of potential confounders on the causal relationship between depression and pesticide exposure.
The quality of the included studies was assessed using the Newcastle Ottawa Scale (NOS), which assigns a maximum overall quality score of 10 stars for cohort and case-control studies and of 9 for cross-sectional studies in 3 dimensions: selection (maximum 5 for cohort and case-control studies and maximum 4 for cross-sectional studies), comparability (maximum 2 stars) and exposure for case-control studies or outcome for cohort and cross-sectional studies (maximum 3 stars). 21 total score of 0-3 in NOS indicates poor quality, a score of 4-6 indicates moderate quality and a score of 7-10 indicates high quality.

Statistical analysis
For each study, we calculated the odds ratio (OR) of depression or depressive symptoms in studies with pesticide use or pesticide poisoning.An odds ratio >1 indicates a positive association between pesticide use or pesticide poisoning and depression or depressive symptoms, while an odds ratio <1 indicates a negative association.Also, we calculated the 95% confidence interval (CI) for each odds ratio.We used the I 2 and Hedges Q statistics to assess between-studies heterogeneity.I 2 values higher than 75% indicate high heterogeneity and a p-value < .1 for the Hedges Q statistic indicates statistically significant heterogeneity. 22The heterogeneity between results was very high and thus we applied a random effect model to estimate pooled effects. 23oreover, we estimated the degree of heterogeneity that might be a result of the small number of studies in our meta-analysis. 24For example, literature suggests that with seven studies and no true heterogeneity, I 2 overestimates heterogeneity by an average of 12%, but when the true heterogeneity is 80%, I 2 underestimates heterogeneity by an average of 28%.We considered sample size, gender and countries that studies were conducted as prespecified sources of heterogeneity.We used metaregression to examine the impact of sample size as a continuous variable.Moreover, we transformed sample in a dichotomous variable using the median value to further explore the role of sample size as a potential source of heterogeneity.The Egger's test was used to assess the publication bias with a p-value < .05indicating publication bias. 22We used OpenMeta[Analyst] for the meta-analysis. 25

Identification and selection of studies
The initial search of all databases retrieved 3.653 potentially relevant studies.After excluding duplicates and screening of title and abstract, the number of pertinent publications was reduced to 262 studies.We excluded reviews, case reports, animal studies, publications that did not study depression as an outcome and pesticide exposure or pesticide poisoning as a determinant and publications that were not casecontrol, cohort or cross-sectional studies.In addition, we excluded studies that have not calculated OR as a measure of association between exposure and outcome and have reported instead, rates of depression among exposed and non-exposed groups or cholinesterase activity levels through blood samples and pesticide metabolites through urine samples.
Eleven studies were included in our qualitative synthesis, from which eight were included in our quantitative synthesis (meta-analysis) and three were excluded due to missing data.The included studies in our meta-analysis comprise a total of five cross-sectional studies, two cohort studies, and one nested case-control study that met our inclusion criteria.The flow diagram of systematic review is shown in Figure 1.
The overall OR for the relation between pesticide poisoning and depression was 2.942 (95% CI = 1.791 to 4.831, p-value < .001)(Figure 3).The heterogeneity between results was very high (I 2 = 78.8%,p-value for the Hedges Q statistic <0.001).Considering that the number of studies in this meta-analysis was small (n = 7), I 2 can range from 66.8% (when the true heterogeneity is 0%) to 100% (when the true heterogeneity is 100%).Thus, there was substantial to considerable heterogeneity in both cases.A leave-one-out sensitivity analysis showed that no single study had a disproportional effect on the overall OR, which varied between 2.445 (95% CI = 1.596 to 3.746, p-value < .001)and 3.426 (95% CI = 2.174 to 5.399, p-value < .001)(Supplementary Figure S2).Subgroup analysis identified a difference regarding countries.In particular, the pooled OR in studies that were conducted in the USA was 2.643 (95% CI = 1.875 to 3.725, p-value < .001,I 2 = 17.8%), while in studies that were conducted in the United Kingdom and Korea was 3.914 (95% CI = 1.083 to 14.145, p-value = .037,I 2 = 91.5%).Considering gender in the subgroup analysis, we found that the pooled OR in studies included both males and females was 7.303 (95% CI = 4.122 to 12.940, p-value < .001,I 2 = 0%), while in studies included only males was 1.948 (95% CI = 1.267 to 2.997, p-value = .002,I 2 = 60.1%), and in studies included only females was 2.699 (95% CI = 1.117 to 6.522, p-value = .027,I 2 = 62.2%).OR was independent of the sample size (p-value = .8).The pooled OR in studies with small number of participants was 4.475 (95% CI = 1.765 to 11.345, p-value = .002,I 2 = 78.4%),while in studies with high number of participants was 2.243 (95% CI = 1.326 to 3.795, p-value = .003,I 2 = 72.7%).P-value for Egger's test implied potential publication bias (p-value < .05).

Pesticide exposure or pesticide poisoning and depression
There are only few studies shedding light on the impact of pesticide use among agricultural population and even so, they focus on acute pesticide poisoning effects rather than possible chronic effects.][34][35] Due to controversial results of the studies conducted, meta-analysis is of significant importance in order to reveal trends that may not be noticeable in each epidemiological study.To the present, this is the first meta-analysis comprising data of both pesticide exposure and pesticide poisoning to investigate the association between pesticide exposure and depression as well as pesticide poisoning and depression.
The results of our meta-analysis showed a positive but non-significant association between pesticide exposure and depression (OR = 1.123; 95% CI: 0.932-1.354)and a significant positive association between pesticide poisoning and depression (OR = 2.942; 95% CI: 1.791-4.831)with very high heterogeneity between results.All studies included in our meta-analysis confirmed a positive association between both pesticide exposure and pesticide poisoning and depression.Moreover, regarding pesticide poisoning, seven out of eight selected studies showed statistically significant results while regarding pesticide exposure this was the case for only one out of four studies.
These results provide further evidence of previous systematic reviews concerning the risks associated with general pesticide exposure, apart from pesticide poisoning events.Freire & Koifman presented statistically significant increased risks between pesticide poisoning and depression or other psychiatric disorders (OR ranging from 2.08 to 5.95) while lower risks were found for chronic pesticide exposure. 36Updated evidence in Cancino et al. review showed a clear association between pesticide exposure and incidence of depressive symptoms. 37Zanchi et al. also demonstrated a positive association between occupational exposure to pesticides and depression while pesticide poisoning was associated with higher risks of depression compared with chronic pesticide exposure; moreover, severe and multiple pesticide poisoning cases increased as well the risks estimates for depression compared with milder cases. 38ur results are consistent with several studies.Compared to other occupational groups, pesticide applicators being intoxicated appeared to have higher odds of experiencing mental disorders. 39eidy et al. reported that exposed migrant farmworkers who had experienced acute poisoning with diagnosed cholinesterase inhibition felt anxious or depressed more often than non-exposed controls. 40ther studies reported significant mood disorders in subjects who had experienced acute poisoning from 3 to >10 years before. 41,42Wesseling et al. found that neuropsychiatric symptoms on the Questionnaire-16 and the BSI were significantly increased among the organophosphate mildpoisoned. 43n a cohort of farm residents in Colorado, depression was significantly associated, after adjustments, with pesticide poisoning in 3 years of follow-up (OR = 2.00; 95% CI: 0.91-4.39)and was primarily due to significant associations with the psychological situations of being bothered by things (OR = 3.29; 95% CI: 1.95-5.55)and feeling everything was an effort (OR = 1.93; 95% CI: 1.14-3.27). 15The authors found that participants who reported past pesticide poisoning at enrollment were twice as likely to report depression during follow-up compared to those who have not experienced a pesticide poisoning.In the same study, depression was positively associated with cumulative days of use of any pesticide, pesticide poisoning and ever experiencing a high pesticide exposure event among the first two case groups but not among the third.Beseler et al. also showed that pesticide poisoning was more strongly associated with depression (OR = 2.57; 95% CI: 1.74-3.79)than intermediate cumulative exposure (OR = 1.07; 95% CI: 0.87-1.31)or high cumulative exposure (OR = 1.11; 95% CI: 0.87-1.42)or an unusual High Pesticide Exposure Event (HPEE) (OR = 1.65; 95% CI: 1.33-2.05). 16Campos et al. found a positive correlation between pesticide poisoning, defined as feeling ill after application of pesticides, and self-reported depression (OR = 2.62; 95% CI: 1.63-4.21). 44o date, there are only few studies indicating that cumulative pesticide exposure is also associated with depression in the absence of an acute poisoning.Although most of the above studies suggest an exposure of high intensity, Salvi et al. found that during the exposure period to OPs, psychiatric diagnoses were higher than expected in tobacco farmers, particularly anxiety disorders and depression. 45Kamel et al. also detected in the AHS that neurologic symptoms examined among pesticide applicators, such as feeling depressed, indifferent or withdrawn during the last 12 months, were associated with chronic pesticide exposure.Stronger associations were found for cumulative use of insecticides (OR = 1.64; 95% CI, 1.36-1.97),especially OPs and organochlorines, even after excluding individuals with a history of pesticide poisoning. 46ur results are also consistent with past research reporting associations of depression with individual or occupational pesticide exposure/poisoning.Parron et al. showed that environmental exposure to pesticides affect the human health by increasing the incidence of certain neurological disorders at the level of the general population. 47Chen et al. reported that household pesticide exposure was associated with a 50% higher risk of depressive symptoms in the general population that attenuated light recreational physical activity. 48Wu et al. suggested that exposure to OPs may increase the risk of depression in young and middle-aged male populations. 49Associations between a biomarker of pesticide exposure and children's depression symptoms were observed, particularly among girls during early adolescence. 50Moreover, adolescent depressive symptoms were elevated in homes where cypermethrin was detected, 51 while limiting exposure to pesticides might complement existing efforts to curb the rise in depression in adolescents. 52Abdel Rasoul et al. reported that children applying pesticides had lower AChE activity than non-applicators and showed significantly more neurological symptoms. 53Rohlman et al. detected significantly elevated metabolite levels specific to chlorpyrifos exposure in adolescent pesticide applicators compared to control group. 54

Exposure measurement
Self-reports of the exposure history were used in all our selected studies.Past research showed a weak association of farmers' self-reported acute health symptoms with AChE depression, [55][56][57][58] a finding that inclines that using self-reported symptoms may not reflect a reliable indicator of organophosphate exposure taking into account the different pesticide types used in low concentration for multiple crops.Nevertheless, there are also non-persistent pesticides that reflect only recent exposure; in this sense using a self-reported questionnaire could overcome this particular limitation. 349][30] The effects on the central nervous system by cumulative low-dose pesticide exposure is yet to be established, necessitating additional research to arrive at reliable conclusions in this area. 36Under this notion, the distinction between low-dose exposure and high-dose pesticide events or pesticide poisoning is important in order to establish causal relationship between pesticide exposure and the manifestation of mental health issues.

Case definition
Difficulties are encountered in assessing the presence and the range of depression since the methodologies used in our selected studies are greatly ranged from a single self-reported question such as "Has a doctor ever diagnosed you with depression . . .?" 9,15,[28][29][30] to structured questionnaires and diagnostic methods. 10,11,15,26,27,31he first could probably affect sensitivity and specificity in measuring depression and subsequently lead to misclassification, while the latter could enhance quality of studies 36 .It is noteworthy that farmers were more likely to report symptoms of depression when questioned in validated depression scales but when additional structured clinical interviews (SCID) were conducted by a physician, no association was detected. 12oreover, these questionnaires provide an initial screening of depressive symptoms while subsequent physician assessment and confirmatory tests are required to diagnose depression. 37n the contrary, physicians have difficulties in recognizing depressive symptoms in their patients and they tend to identify cases of major depression more efficiently than cases of dysthymic disorder or intermittent episodes of depression. 59,60Thus, focusing only on depression diagnosis, treatment or hospitalization could lead to a loss of participants with depressive symptoms who did not seek any treatment, while using standardized scales may detect a broader range of depression types. 9

Subgroup analysis
The subgroup analysis of the present study identified a difference regarding gender in meta-analysis for pesticide use and depression and a difference regarding countries in meta-analysis for pesticide poisoning and depression.More specific, pooled OR for the relation between pesticide exposure and depression in studies including both males and females was greater than those including only females and pooled OR for the relation between pesticide poisoning and depression in studies conducted in the USA was greater than those conducted in the UK and in Korea.
Combination of age and socioeconomic status could explain in some extent the heterogeneity by countries.In studies conducted in the UK and in Korea, mean age of participants ranged from 62 to 70 years, while in studies conducted in the USA, mean age ranged from 29 to 47 years.The elderly in rural Korean areas have noted higher depression scores than in urban areas, due to differences in physical/ psychological factors, socioeconomic status, and access to medical resources. 61In addition, previous studies reported that low socioeconomic condition is associated with elevated risk of depression among elderly in Korea. 10Other potential explanations may be the inadequate safety training for using pesticides in Korea, since there was no such mandatory law, and also the poor availability of mental health care services in rural areas. 10ender heterogeneity is probably due to the fact that females are not themselves licensed applicators and therefore devote less time in mixing and applying pesticides than males. 27Another explanation could be that females tend to respect more safety practices when using pesticides than males.

Confounders
All our selected studies examined the association between pesticide exposure and depression after adjustment for age.Additionally, five studies considered smoking habit as a confounder, while three studies eliminated gender and education level, and two studies eliminated alcohol use.It is noteworthy that one single study included economic status as potential confounder, and none included any social factors.Considering all above factors as well as socioeconomic factors in future research, when measuring relationships between pesticide exposure and depression, is an issue of major importance.

Limitations
Among the highlights of the present meta-analyses is that, to our knowledge, they are the first comprehensive meta-analyses to investigate a possible relationship on one hand between pesticide exposure and the risk of depression and on the other between pesticide poisoning and the risk of depression.However, our study has several limitations.First, it was not possible to investigate a dose-response relationship between different groups of pesticide exposure (low, median, high) and depression or even provide a cut-off point.Second, high heterogeneity was observed among the original data, decreasing thus the reliability level of the present study; we performed random effects model and subgroup analyses to overcome this certain limitation.Third, we did not investigate specific compounds or pesticide categories in our subgroup analyses.It is possible that some of the heterogeneity might be due to the different types of co-exposure to other chemical compounds that farmers were exposed to.Although retrospective self-reported pesticide experience is the most commonly used method for measuring pesticide exposure in cross-sectional and case-control studies, it may lead to a potential recall bias. 36Finally, since we identified publication bias in our meta-analysis, it is probable that studies without a statistically significant association between pesticide exposure and depression have not been published or even submitted for publication.

Conclusions
The present systematic review and meta-analysis provides the most recent evidence on the positive association between both general pesticide exposure and pesticide poisoning and depression.Future research should improve exposure and outcome measurements by including biological biomarkers and specific pesticide-specific biomarkers of depression as well as controlling for socioeconomic confounding factors. 37Given that different chemical classes may interact with the central nervous system through various mechanisms, the association between mixed pesticide use and depressive symptoms should be further examined.Prospective cohort studies, aiming at both large samples and low-dose pesticide exposure and using detailed exposure evaluation methods, should be conducted.Imposing strict surveillance measures in agricultural companies, implementing safety training programs and monitoring farmers' mental health are essential policy strategies to safeguard the mental health of the exposed agricultural workers.

Figure 1 .
Figure 1.Flow diagram of systematic review of literature concerning pesticide exposure or pesticide poisoning and depression risk.

Figure 2 .
Figure 2. Forest plot for the relation between pesticide use and depression or depressive symptoms.Black solid line denotes the 95% confidence interval for odds ratio in each study, while the red dotted line denotes the pooled odds ratio from meta-analysis.

Figure 3 .
Figure 3. Forest plot for the relation between pesticide poisoning and depression or depressive symptoms.Black solid line denotes the 95% confidence interval for odds ratio in each study, while the red dotted line denotes the pooled odds ratio from meta-analysis.

Table 1 .
Characteristics of studies on pesticide exposure or pesticide poisoning and depression.
a Multivariable analysis refers to studies where confounders were eliminated, while univariate analysis refers to studies where confounders were not eliminated.Abbreviations: AHS=Agricultural Health Study, BDI-II=Beck Depression Inventory-Second Edition, BSI=Brief Symptom Inventory, CES-D=Center for Epidemiologic Studies Depression Scale, CFFHHS=Colorado Farm Family Health and Hazard Surveillance, GDS=Geriatric Depression Scale, HPEE=High Pesticide Exposure Event, OR=Odds Ratio, PD=Parkinson Disease, PHQ=Patient Health Questionnaire, PPE=Personal Protective Equipment, PRE-B=Pre-Enrollment Both, PRE-E=Pre-Enrollment, POD= Prevalence Of Depression, POST=Post Enrollment, MSA=Mutualité Sociale Agricole.