Search strategy

We searched MEDLINE, Embase, and PsycINFO databases for English-language abstracts published between Jan 1, 2017 and Apr 21, 2023. Our search start date was aligned with the end of the search period by the previous systematic review (i.e. Dec 31, 2016). We conducted the original search on April 15, 2023 and updated the search on March 6, 2024 (S1 File). We reviewed manuscripts included in published systematic reviews and manually searched the reference list of eligible studies to identify relevant manuscripts. Additionally, we requested experts examine their personal collections to identify additional articles.

Eligibility criteria

Abstracts documenting the effects of pesticide regulatory changes on suicide by pesticide poisoning were eligible for inclusion. Interventions included any changes in regulation on pesticides, such as changes in the concentration of products or formulations, reintroduction of pesticides that were previously banned, and complete bans on sales, imports, and use of one or more products.

We included interventions conducted at the population-level (national or regional) and excluded interventions at community levels, such as the use of lockable boxes or limiting the ease with which pesticides can be purchased. We excluded initiatives by pesticides manufacturers that were aimed at increasing the safety of their products, e.g. inclusion of emetics in the formulation and training initiatives aimed at pesticide vendors [12]. Studies conducted in selected population groups were excluded, e.g. patients admitted to hospital, data from poison centres, and studies investigating specific age groups or sex.

Manuscripts published in any language were eligible for inclusion if an abstract in English was available. We included ecological, controlled randomised and non-randomised intervention, and natural experimental designs but excluded reviews, except as potential sources of relevant manuscripts.

Screening strategy

BR and LB independently screened titles and abstracts against the eligibility criteria and discussed any disagreements until a consensus was reached (S1 Table). A third independent reviewer was consulted (DK) where there was no agreement. BR and CY independently reviewed full texts and extracted data using a data extraction form, defined a priori. Where the full manuscript was only available in Chinese, CT alone extracted data and conducted the risk of bias assessment.

Risk of bias assessment

Consistent with the previous review, BR and CY independently assessed the risk of bias using a modified version of the risk of bias criteria for interrupted time series studies and classified studies into three categories - high, low or unclear risk of bias (see S1 File) [13]. Studies that ignored secular trends in the pre-intervention period were recorded applying an additional item in the tool.

Data analysis

We conducted a narrative synthesis of the data due to the high level of heterogeneity between studies. Studies varied in terms of interventions, settings, type of regulatory change, type and number of pesticides restricted, and differences in the burden of lethal pesticide poisoning to the study population. Since the proportion of suicides by pesticide ingestion is higher in LMIC, we stratified our results by LMIC and HIC. We marked as ‘not assessed’ where data were missing.

In a post-hoc analysis, we calculated the percentage change in pesticide suicide and overall suicide rates between the pre-intervention year and the last post-intervention year available by obtaining raw data via corresponding authors of time series studies [14–19]. We supplemented these with eligible studies from the previous review [20,21]. We used these data to fit segmented linear regression models using ordinary least squares to allow a graphical representation of pre- and post-intervention suicide trends for studies with at least two years of data before and after the intervention. We analysed, summarized, and plotted data in R version 4.2.1, using the tidyverse and robvis packages [22].

Low- and middle-income countries

All four studies showed a decrease in pesticide suicide rates following changes in regulation limiting access to HHP, even when taking into account secular trends (Fig 3, Table 1) [14,15,19,24]. Changes in trends for pesticide suicide rates were observed shortly after the implementation of the bans (Fig 3). The three studies that reported on overall suicides found a concomitant decrease in suicide rates following the intervention. The only study that did not report on overall suicides provided raw data, which, when plotted, showed a decrease in overall suicides following the intervention (Fig 3) [14]. Data from previously identified studies also showed a decrease in both pesticide and overall suicide rates following a ban on: a) paraquat, dimethoate and fenthion in Sri Lanka between 2008 and 2011, and b) all WHO Class I [26] HHPs in Bangladesh in 2000 (Fig 3) [20,21].

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Graphical representation of changes in pesticide and overall suicide deaths for studies from low- and middle-income countries.

The black dots represent the observed data points. The red solid lines represent the pre- and post-interrupted trend lines for pesticide suicide rates, and the red dashed lines represent the counterfactual trend for pesticide suicides following the intervention. The blue solid lines represent the pre- and post-interrupted trend lines for overall suicide rates, and the blue dashed lines represent the counterfactual trend for overall suicides following the intervention. The vertical dashed line shows the time of the intervention. This figure also includes data extracted from two relevant studies published in the 2017 review (ref). Please note that the scale for the Y-axis is different in each figure, as the age-standardised suicide rates vary considerably from country to country.

https://doi.org/10.1371/journal.pgph.0003785.g003

There was a 60.5% decrease in pesticide suicide rates in China (Table 1) from 2006 to 2018, with a steeper decline observed following the first and the third defined intervention points, when complete bans on five key organophosphorus HHPs and the sale/use of aqueous paraquat formulations were implemented, respectively [19]. The decreasing trend in pesticide suicide rates flattened after the second defined intervention point in 2012, following a ban on a new paraquat parent drug (with added emetic, odorizer and colorant) and aqueous solution (Fig 3). Pesticide rates decreased by 57.2% in areas of low urbanisation and 67.3% in areas of high urbanisation. However, before the three defined intervention points, the pesticide suicide rate in high-urbanisation areas was already declining, and this trend did not exhibit further decreases following the pesticide bans. Conversely, in low-urbanisation areas, the pesticide suicide rate had been increasing, but the trend reversed after the bans on five organophosphorus HHPs and the sale/use of aqueous paraquat in 2008 and 2016 (S2 Table) [19].

The two studies from India covered the same regulatory change at the national level, i.e. an endosulfan ban in May 2011 [14,15]. National pesticide suicide rates decreased by 33.8%-39% in 2014, when accounting for secular trends (Table 1, Fig 3). In Kerala, a state in the southwest of India, pesticide suicide rates were reduced by 21% between 2006 and 2010, following a regional ban on endosulfan in 2005 [15]. However, a slight increase in overall suicides was observed following the regional ban in the same period [15] (Fig 3). Unemployment rates remained relatively stable and gross domestic product per capita increased by approximately 8% per year throughout the study period; neither of which seemed to have affected suicide rates (S2 Table) [15].

The study from Inner Mongolia investigated three separate bans on two insecticides in 2008, 10 organophosphorus insecticides in 2011, and two pesticides (including paraquat) in 2012 [24]. Pesticide suicides decreased by 49%, with a concomitant decrease in overall suicide rates by 33%. However, the study did not take into account secular trends in the analyses. Additionally, the authors report several social policies aimed at improving living standards and reducing poverty levels in Inner Mongolia during the time when the third ban was implemented, which could have affected suicide rates (S2 Table) [24].

High-income countries

All five studies showed a decrease in pesticide suicide rates following bans on HHPs (Fig 4) [16–18,23,25]. Only the study from Taiwan showed a concomitant decrease in overall suicide rates (by 7%) following the complete ban on paraquat in 2020 (Table 1 and Fig 4) [16]. Overall suicide rates (or trends) were not reported for South Korea; however, the authors provided raw data, which, when plotted, showed a reversal from an upwards trend in the pre-intervention period (Fig 4) [18,23].

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Graphical representation of changes in pesticide (in red) and overall (in blue) suicide deaths for studies from high-income countries.

The black dots represent the observed data points. The red solid lines represent the pre- and post-interrupted trend lines for pesticide suicide rates, and the red dashed lines represent the counterfactual trend for pesticide suicides following the intervention. The blue solid lines represent the pre- and post-interrupted trend lines for overall suicide rates, and the blue dashed lines represent the counterfactual trend for overall suicides following the intervention. The vertical dashed line shows the time of the intervention. Please note that the scale for the Y-axis is different in each figure, as the age-standardised suicide rates vary considerably from country to country.

https://doi.org/10.1371/journal.pgph.0003785.g004

The two studies from South Korea investigated the impact of cancelling the re-registration of paraquat in November 2011, which was followed by a complete ban on sales in October 2012 [18,23] (Table 1). Paraquat was responsible for a large proportion of pesticide poisoning deaths (61%) and had a case fatality of 43% in South Korea [27]. After adjusting for several socio-demographic variables, the post-ban period was associated with a 39% lower risk of pesticide suicides compared with the pre-ban period [23] (S3 Table). Cha et al. [18] showed that pesticide suicide rates decreased annually by approximately 28% in 2011–2014. The association between the paraquat ban and pesticide suicide rate was marginal in a Prais–Winsten regression analysis that adjusted for the amount of pesticides sold, the proportion of people involved in farming, unemployment rate, dependency ratio, divorce rate, the 1997 Asian economic crisis, and the 2008 Great Recession [18]. In contrast to the prominent reduction in pesticide suicide rates in rural areas after the paraquat ban, the decreasing trend in urban areas did not change after the ban [18].

The two studies from Taiwan evaluated the impact of the ban on the import and production of paraquat in February 2018 and the ban on the sale and use of paraquat in February 2020 (Table 1) [16,17]. The proportion of suicides due to pesticide poisoning was 12.1% (2011–2017), with paraquat responsible for 5% of all suicides in Taiwan before the bans [17]. Both studies showed a fewer-than-expected pesticide suicide rate following the 2018–2020 paraquat ban; 37% fewer-than-expected number of pesticide suicides in 2019, mainly due to a 58% fewer-than-expected suicides by paraquat ingestion [17], and 44% fewer-than-expected number of pesticide suicides in 2020 [16]. There was no difference in overall suicide rates up to 2019 but when data from 2020 was included in the model, they found a concomitant 7% fewer-than-expected overall suicides in 2020 [16]. The decrease in pesticide suicides was more marked among, rural residents, males, and middle-aged to older individuals compared to their counterparts [16,17].

The study from Japan assessed the impact of restrictions on sales and use of a 20% liquid formulation of paraquat, which accounted for 82.6% of deaths due to pesticide poisoning, and the subsequent introduction of a 4.3% paraquat ion with 4·1% diquat ion combination product in 1986 to replace the more toxic 20% liquid paraquat formulation [25]. Overall, there was a reduction of pesticide suicides by 91.9%, from 8.6% of all suicides in 1985 to 0.7% in 2019 (Table 1). However, the study did not account for secular trends or assess changes in other factors concomitant to the intervention that might have affected suicide rates.

Summary of studies reported by Gunnell et al. [ 10 ]

In the previous systematic review, Gunnell et al. [10] described 27 studies where changes in regulation for HHP had taken place in 16 countries, which included data up to 2016 [8,20,21,28–51]. Five countries were LMIC (Sri Lanka, Bangladesh, Colombia, Jordan, and India) and 11 were HIC (Greece, South Korea, Taiwan, Denmark, UK, Finland, Germany, Hungary, Ireland, Japan, and USA).

Implementation of regulatory changes limiting access of HHPs led to a reduction in pesticide suicide in most countries. In Greece, there were no changes in pesticide suicide rates following the withdrawal of parathion and monocrotophos in Crete in 2003 [34]. Reductions in rates were observed in several European countries, such as Denmark [39], Finland [42], Germany [43], Hungary [44,45], Ireland [40,46], and England (but not Scotland) [40,41], as well as Japan [47] and the USA [49–51]. However, the picture was mixed in Ireland and in England, where there was an initial increase in pesticide suicide rates followed by a reduction in rates.

Overall suicide either decreased in three countries (i.e. Sri Lanka [8,20,31–33], Bangladesh [21], and Colombia [28]) or remained the same in one country (i.e. India [30]) following regulatory changes in all LMIC that reported on suicide rates. Results from HIC were mixed, with data from some countries showing an increase in overall suicides (i.e. Greece [34], Taiwan [37,38], UK) [40,41], Ireland [40,46], Finland [42], and USA [49–51]).

However, only five studies used appropriate time series analyses to take into account secular trends prior to the implementation of regulatory changes in access to HHPs [8,20,21,28,36]. These studies were conducted in four LMIC counties (i.e. Sri Lanka, Bangladesh, Colombia) and in one HIC (i.e. South Korea). All but the study from Colombia [28] reported decreases in pesticide suicide and overall suicide rates following bans on HHPs.

Two studies published after 2017 were included in the previous systematic review of the same topic [20,21]. These studies were unpublished at the time of the publication of the review; however, the authors had access to the findings from these and therefore decided to include them in their systematic review. Since they were published within our study period and conducted time series analyses, we used the raw data from the two studies to produce graphs representing the pre- and post-intervention periods (Fig 3).

In Sri Lanka, a phased ban on dimenthoate and fention was implemented between 2008 and 2011, and a phased ban on paraquat was implemented between 2009 and 2012. Pesticide suicide rates dropped by 50% and overall suicide rates by 21% between 2011 and 2015 [20]. Using negative binomial regression, they found that the pesticide and overall suicide rates after the intervention were lower than what would be expected should the previous trends continue [20]. The proportion of suicides due to pesticides decreased from 38% in 2000 to 30% in 2015.

In Bangladesh, organophosphorus insecticide HHPs were fully banned at the end of 2000, when all WHO class Ia and Ib pesticides were banned [21]. These included dichlorvos, dicrotophos, disulfoton, ethyl parathion, methyl parathion, mercury compounds, and monocrotophos phosphamidon. Using negative binomial regression to account for secular trends, they reported a decrease in pesticide suicides by 65.1% and in overall suicide rates by 25.0% between 1996 and 2014, lower than what would have been expected should the bans have not occurred [21]. The proportion of suicides due to pesticides decreased from 44.7% to 20.9% during the same period.