Longer-term impact of pesticide bans on suicide in South Korea

Abstract

Introduction

Pesticide self-poisoning accounts for 14%–20% of global suicides. South Korea banned paraquat and eight insecticides commonly involved in suicide in 2011–2012. We investigated the longer-term (2013–2019) impact of the pesticide bans on reducing suicides in South Korea.

Methods

Suicide data by sex, age, area and method among people aged 15 years or above were extracted from registered death data (1997–2019). Data for suicide by self-poisoning using different categories of pesticides (herbicides and fungicides; insecticides; rodenticides and other pesticides; and unspecified pesticides) were also extracted. Segmented regression analyses were used to estimate step (level) changes in pesticide suicide rate in 2012 (ie, the intervention period) and 2013–2019 (ie, the post-intervention period) as well as the slope (trend) changes in pesticide suicide trends in 2013–2019 (vs the 2003–2011 pre-intervention period).

Results

The bans on paraquat and eight insecticides (2011–2012) were followed by a 15% reduction in pesticide suicide rates in 2012 and a further 36% reduction in 2013, accompanied by an additional 6% annual decline in trend in 2013–2019 versus 2003–2011, with an estimated 8353 pesticide suicides averted (2012–2019). The reduction was mainly found in males, older people aged 70+ and rural populations. The greatest impact was found in suicides involving herbicides and fungicides (including paraquat) and insecticides across the categories of pesticides. No evidence was found for a shift to other poisoning suicides or an impact on agricultural yields of the pesticide bans.

Conclusion

South Korea’s pesticide bans in 2011–2012 were followed by a sustained decrease in pesticide suicides in the 7-year period (2013–2019) after their implementation. National policies restricting or banning highly hazardous pesticides commonly involved in self-poisoning can prevent many suicide deaths over a sustained period. Additional regulatory efforts could further reduce these unnecessary premature deaths.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• In South Korea, the 2011–2012 paraquat ban was followed by a reduction in pesticide suicide rates in 2013, while the concurrent ban on eight insecticides, as well as their joint longer-term impact on reducing suicide, was not investigated.

WHAT THIS STUDY ADDS

• The ban on paraquat and eight insecticides (2011–2012) was associated with a 15% and 36% immediate reduction in pesticide suicide rates in 2012 and 2013, respectively, followed by an additional 6% annual decline in the trend in 2013–2019, with an estimated 8353 pesticide suicides averted in 2012–2019.

• The absolute reductions were greatest among males, older individuals and rural residents, with the largest decline observed in suicides involving herbicides and fungicides, followed by those involving insecticides.

• No evidence was found for a shift to suicides using other substances or an adverse effect on agricultural yields.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• National policies that restrict or ban highly hazardous pesticides commonly used in self-poisoning can lead to a substantial and sustained reduction in suicide rates.

Introduction

Self-poisoning with pesticides accounts for approximately 14%–20% of all suicides worldwide, with an estimated 110 000–168 000 deaths a year.¹ In 2020, South Korea had the highest suicide rate (24.1 per 100 000 population) among all Organisation of Economic Cooperation and Development countries.² Furthermore, in South Korea, data from earlier years (2006–2010) showed that about one-fifth of all suicides were by pesticide self-poisoning,³ which was higher than that in many other high-income Asian countries such as Japan and Taiwan.^{4 5}

Research has shown that regulations or bans on highly hazardous pesticides (HHPs) are among the most effective6,10 and cost-effective¹¹ approaches to reducing suicides by pesticide poisoning, and overall suicide in countries or regions where pesticide poisoning is among the leading methods of suicide. Recent studies from low-income and middle-income countries such as China,¹² India,^{13 14} Sri Lanka^{15 16} and Mongolia¹⁷ showed that restrictions on specific HHPs, for example, paraquat and endosulfan, were followed by a decline in pesticide suicides12,1517 and overall suicides.^{13 15 17} Similar evidence was found in some high-income countries such as Japan¹⁸ and Taiwan,^{19 20} where policies of restricting access to HHPs such as paraquat were linked to a reduction in pesticide suicides.

In South Korea, a number of pesticide regulations have been implemented since the Pesticide Management Act of 1957 (http://www.rda.go.kr). 60 pesticides were regulated or banned from 1983 to 2019; these included the cancellation of the re-registration of paraquat in November 2011 and the complete ban of paraquat in November 2012.²¹ Paraquat, a commonly used herbicide that is highly lethal when ingested in self-poisoning, was the leading pesticide involved in suicides in South Korea before the paraquat ban.²² Previous studies have shown that the paraquat ban was followed by a reduction in pesticide and overall suicide rates in 2013^{12 23 24} and 2014.²⁴ The reduction in pesticide suicide immediately following the paraquat ban was found to be most marked in males, older people and rural areas.²³ Furthermore, the paraquat ban was not followed by a change in the yield of major crops in the short term.²³

However, the longer-term impact of the South Korean paraquat ban has not been investigated, although findings from a recent time trend analysis suggested a sustained reduction in pesticide suicides after the ban.²⁵ In the meantime, no research has been conducted to examine the effects of the bans on several insecticides that were also implemented in 2011–2012, which included several insecticides commonly involved in pesticide suicides in the country, including endosulfan, dichlorvos and methomyl, which together accounted for 42.9% of insecticide-related deaths in 2011 according to national surveillance data.²⁶ The three insecticides were associated with a relatively high case fatality when ingested; a recent review reported median case fatality ratios of 22.9%, 32.3% and 7.2%, respectively.²⁷ It is unknown whether the effectiveness of bans on these pesticides would be sustained over time and whether a substitution from pesticides banned to other unbanned pesticides or poisons may have occurred and offset the effect of these bans.²⁸

We aimed to investigate the longer-term effect of the 2011–2012 South Korean ban on paraquat and several insecticides on reducing pesticide and non-pesticide poisoning suicide rates and crop yields in the 7 years (2013–2019) following its implementation. We also examined the impact of the pesticide bans by sex, age group and urbanisation level.

Methods

Patient and public involvement

Patients or the public were not involved in the design, conduct, reporting or dissemination plans of our research as the study was based on a secondary analysis of mortality data.

Data for suicide and population

Suicide data for individuals aged 15 years or above (1997–2019) were obtained from Statistics Korea (http://kostat.go.kr). Information on sex, age, area of residence and date of death was extracted. The area of residence was used to identify the urbanisation levels of districts; the metropolis (ie, the capital city Seoul), small-sized and medium-sized cities (ie, cities with >1 million population and Sejong City, a special self-governing city) and rural areas (ie, areas outside urban administrative boundaries) were defined using the governmental administrative divisions based on population sizes and area characteristics.

The underlying causes of death were coded according to the International Classification of Diseases, 10th Revision (ICD-10). Suicide by pesticide poisoning was identified using the ICD-10 code X68; all other suicides were classified as ‘non-pesticide suicides’. Among suicides by pesticide poisoning, ICD-10 T codes were used to identify the specific categories of pesticide agents involved, including insecticides (T60.0–T60.2), herbicides and fungicides (T60.3), rodenticides and other pesticides (T60.4, T60.8) and unspecified pesticides (T60.9). Previous studies from South Korea indicated that the majority of poisoning suicides using herbicides and fungicides were from poisoning using herbicides;²⁹ in a study based on data from an emergency department-based National Injury Surveillance System, paraquat accounted for 87% of all herbicide-related deaths.²⁶

We also extracted data for non-pesticide poisoning suicides using medicinal/biological substances (ICD-10 T36-T50), alcohol, organic solvents and corrosive substances (T51-T57), carbon monoxide (CO) and other gases (T58-T59) and other and unspecified chemicals (T61-T65). Previous studies showed a marked increase in suicide from CO poisoning in South Korea from 2008, and the increase was largely attributed to the use of coal briquettes and barbecue charcoal as sources of CO.^{30 31} We compared trends in pesticide-related and non-pesticide-related poisoning suicides to examine any evidence for a shift from pesticide suicides to non-pesticide poisoning suicides after the pesticide bans.

Mid-year population data by sex, age group and area were obtained from the Korean Statistical Information Service (KOSIS; https://kosis.kr/index/index.do). Data on agricultural crop yields were from the annual Agricultural Production Surveys conducted by Statistics Korea.

Pesticide bans

Information on regulated or banned pesticides, including the active ingredient, date of ban and chemical category, was obtained from the Korea Rural Development Administration (http://www.rda.go.kr) and documents published by the Korea Crop Protection Association³² as well as through internet searches. A total of 39 pesticide products were banned between 2003 and 2019; of these, 19 pesticides were banned in 2011–2019 (online supplemental table 1).

In this study, we defined the pre-intervention (2003–2011), intervention (2012) and post-intervention (2013–2019) periods based on the dates when the major pesticides involved in suicides were banned in South Korea. These major pesticides included paraquat, which was banned in 2011–2012, and certain insecticides banned in December 2011. The re-registration of paraquat was cancelled on 23 November 2011, and its use was fully banned from November 2012. On 6 December 2011, a total of eight insecticides were banned, including six insecticides classified by the WHO as class I hazard pesticides (dichlorvos, methidathion, methomyl, ethyl p-nitrophenyl benzenethiophosphonate [EPN], monocrotophos and omethoate) and two WHO class II pesticides (endosulfan and benfuracarb); of note, these eight insecticides were the major causes of deaths involving insecticides.²⁶ We thus considered the year 2012 as the year when major pesticides involved in suicides in South Korea were banned and modelled it as the intervention period in the interrupted time-series analysis. In April of this year (2012), three pesticides (tolylfluanid, tebuconazole+tolylfluanid and dichlofluanid) were additionally banned. During the post-intervention period (2013–2019), a total of seven pesticides were further banned (online supplemental table 1).

Analyses

We calculated annual age-standardised suicide rates (1997–2019) in individuals aged 15 years or above using age-specific weights derived from the 2000 WHO World Standard Population. We also calculated suicide rates by (i) method (pesticide vs non-pesticide); (ii) pesticide agent (herbicides and fungicides vs insecticides vs rodenticides and other pesticides vs unspecified); and (iii) non-pesticide chemical or gas (medicinal/biological substances vs alcohol, organic solvents and corrosive substances vs CO and other gases vs other and unspecified chemicals). We plotted the annual suicide rates and examined the trends graphically.

We conducted interrupted time-series analyses using segmented regression to assess the impact of pesticide bans implemented in 2011–2012 (ie, the intervention), as described above, on reducing pesticide suicide rates in South Korea after the bans, relative to those expected based on suicide rates and trends during the pre-intervention period (2003–2011). The year 2003 was chosen as the starting year of the pre-intervention period as a previous study from South Korea, which applied the joinpoint regression analysis, revealed a consistent downward trend in pesticide suicide from 2003, up to the time when the bans of major pesticides involved in suicides were implemented in 2011–2012.²³

Poisson regression models were fitted to the annual number of pesticide suicides, with the logarithm of population size as the offset. Segmented regression analyses were used to estimate the following:

1. The step (level) changes in pesticide suicide rate in 2012 (ie, the intervention period) and 2013 (ie, the start year of the post-intervention period), relative to the expected rates in the absence of the intervention.

2. The slope (trend) changes in pesticide suicide trends in 2013–2019 (vs the 2003–2011 pre-intervention period).

This approach allowed us to examine both the immediate effects in 2012 and 2013 (ie, the step changes in suicide incidence comparing the observed and expected number of suicide) following the pesticide bans and the longer-term effect on trends in pesticide suicide rates (ie, the slope change in suicide trends between 2003–2011 and 2013–2019), which capture the sustained impact of these regulatory measures on HHPs. Specifically, the step change for 2012 was estimated as the difference between the observed rate and the expected value based on the 2003–2011 pre-intervention trend. In contrast, the step change for 2013 was estimated as the difference between the observed rate in 2013 and the expected value projected from the same pre-intervention trend, extended beyond the observed value in 2012, that is, the additional level change given the pre-intervention trend and the step change in 2012.

Rate ratios (RRs) for step and slope changes in suicide and their 95% CIs were estimated. We included year, sex and age group in the models to adjust for time trends as well as changes in population structure. Stratified analyses were conducted by sex, age group (15–49, 50–59, 60–69 and 70+years), area (metropolis, city and rural areas) and the category of pesticide agents (herbicides and fungicides; insecticides; rodenticides and other pesticides; and unspecified). Differences in RRs among subgroups were examined by including appropriate interaction terms in the regression models. The likelihood ratio test was used to examine the statistical evidence for the difference between the models with and without the interaction terms. The number of suicides averted was calculated as the difference between the expected number of suicides (E), estimated based on the pre-intervention trends, and the observed number of suicides (O) during 2012–2019 (ie, E-O).

Two sensitivity analyses were conducted. The first sensitivity analysis defined the pre-intervention period as 2004–2011, as the pesticide suicide rate peaked in the year 2004 (figure 1A). In the second sensitivity analysis, the post-intervention period was defined as 2012–2019, without treating the year 2012 as a separate intervention period. All analyses were performed using Stata 19.0 (StataCorp, College Station, TX, USA).

Figure 1. Annual age-standardised rates (per 100 000) of (A) overall, pesticide and non-pesticide suicides (2000–2019) and (B) poisoning suicide by causative agent (2003–2019) in South Korea. Note: In (A) and (B), the area highlighted in yellow indicates the years (2011–2012) when paraquat and several insecticides, which were the main insecticides involved in suicides, were banned. In (A), the vertical dashed line indicates the year (2003) when the defined pre-intervention period began, and the arrows indicate the years when national pesticide regulations were implemented; refer to online supplemental table 1 for the details of the numbered pesticide products banned. In (B), the International Classification of Diseases, Tenth Revision (ICD-10) codes for the causative agents used in poisoning are provided in the brackets. CO, carbon monoxide.

We used Joinpoint regression models to examine temporal changes in agricultural yields from 2000 to 2019. Models were fitted allowing for a maximum of three joinpoints, in accordance with established guidelines,³³ and identified joinpoints were evaluated in relation to the timing of the pesticide bans. Joinpoint regression was conducted using the Joinpoint Trend Analysis Software (https://surveillance.cancer.gov/joinpoint/).

Results

The age-standardised rates of suicide by pesticide poisoning increased with the rate reaching a peak at 9.2 per 100 000 in 2004 (figure 1A); the rate gradually declined afterwards and showed a further reduction following the implementations of the bans on multiple pesticide products in 2011–2013. Non-pesticide suicide rates showed an increase in 2000–2011, followed by a reduction in 2011–2017 and some level-offs in 2017–2019.

Figure 1B shows trends in age-standardised rates of poisoning suicides using different solid, liquid or gaseous substances in 2003–2019. Among suicides by pesticide self-poisoning, those involving herbicides and fungicides accounted for the majority, followed by those using unspecified pesticides, insecticides, rodenticides and other pesticides. Overall, pesticide suicide rates using all categories of pesticides decreased over the study period, while the marked decrease in suicides involving herbicides and fungicides after 2011 accounted for most of the reduction in pesticide suicides.

Poisoning suicides using CO and other gases increased markedly from 2008 and levelled off from 2014 (figure 1B). Of note, the marked increase in CO and other gassing suicides commenced in 2008, 3 years before the paraquat and insecticides bans from 2011, and the upward trend did not become more marked following the pesticide bans. The rate of suicide by medicine poisoning was relatively low and showed a stable upward trend during the study period. The rates of suicide by poisoning using alcohol, corrosive substances and other substances were minimal. The rates of poisoning suicide using other and unspecified substances decreased over time, particularly during 2003–2007.

Figure 2 presents the annual age-standardised pesticide suicide rates in South Korea from 2003 to 2019, along with the regression line of best fit based on interrupted time-series analyses. Table 1 summarises the results of the interrupted time-series analysis, both overall and stratified by sex, age group, region and category of pesticide agents. During the pre-intervention period (2003–2011), the pesticide suicide rate declined by 7% annually (RR for slope=0.93; 95% CI 0.92 to 0.94). In the intervention year (2012), the pesticide suicide rates showed a 15% reduction compared with the expected value based on the pre-intervention trend (step change RR=0.85; 95% CI 0.78 to 0.94). In the following year (2013), there was a further 36% (step change RR=0.64; 95% CI 0.56 to 0.72) reduction in pesticide suicide rates. The slope change during the post-intervention period (2013–2019) was 0.94 (95% CI 0.91 to 0.98), indicating an additional 6% annual decrease relative to the pre-intervention trend.

Figure 2. Trends in annual age-standardised pesticide suicide rates in South Korea, 2003–2019. Note: ‘Observed’: annual age-standardised pesticide suicide rates per 100 000. The area highlighted in yellow indicates the years (2011–2012) when paraquat and several insecticides, which were the main insecticides involved in suicides, were banned. EPN: ethyl p-nitrophenyl benzenethiophosphonate.

Table 1. Suicide rate ratios (RRs) and their 95% confidence intervals (CIs) for the step and slope changes in pesticide suicides during the intervention period (2012) and post-intervention period (2013–2019), relative to the pre-intervention trend (2003–2011), by sex, age, area and the category of pesticide agents.

	Pre-intervention (2003–2011)		Intervention period (2012)		Post-intervention (2013–2019)				Estimated number of suicides averted (2012–2019)*
	Slope		Step change		Step change		Slope change
	RR	(95% CI)	RR	(95% CI)	RR	(95% CI)	RR	(95% CI)
Total	0.93	(0.92 to 0.94)	0.85	(0.78 to 0.94)	0.64	(0.56 to 0.72)	0.94	(0.91 to 0.98)	8353
Sex
Male	0.93	(0.91 to 0.95)	0.84	(0.66 to 1.09)	0.64	(0.49 to 0.84)	0.95	(0.89 to 1.01)	5753
Female	0.93	(0.90 to 0.95)	0.87	(0.68 to 1.12)	0.63	(0.48 to 0.83)	0.93	(0.87 to 0.99)	2598
Age group
15–49	0.87	(0.86 to 0.88)	0.87	(0.76 to 1.00)	0.57	(0.49 to 0.65)	0.89	(0.85 to 0.94)	967
50–59	0.92	(0.90 to 0.94)	0.85	(0.78 to 0.93)	0.66	(0.59 to 0.75)	0.92	(0.89 to 0.96)	1296
60–69	0.95	(0.93 to 0.96)	0.75	(0.69 to 0.81)	0.61	(0.54 to 0.69)	0.92	(0.89 to 0.95)	2497
70+	0.98	(0.97 to 1.00)	0.81	(0.76 to 0.87)	0.64	(0.56 to 0.73)	0.91	(0.88 to 0.95)	6231
Area
Metropolis	0.88	(0.85 to 0.90)	0.96	(0.73 to 1.28)	0.67	(0.48 to 0.93)	0.99	(0.92 to 1.07)	146
City	0.90	(0.89 to 0.91)	0.97	(0.83 to 1.14)	0.60	(0.50 to 0.72)	0.99	(0.95 to 1.03)	821
Rural	0.94	(0.93 to 0.95)	0.83	(0.75 to 0.91)	0.64	(0.56 to 0.73)	0.93	(0.89 to 0.96)	7619
Category of pesticide agents†
Herbicides and fungicides	0.95	(0.94 to 0.96)	0.79	(0.71 to 0.88)	0.54	(0.47 to 0.63)	0.89	(0.85 to 0.93)	8534
Insecticides	0.98	(0.96 to 1.00)	0.85	(0.72 to 1.00)	0.70	(0.59 to 0.84)	0.93	(0.89 to 0.97)	1394
Rodenticides and other pesticides	0.82	(0.79 to 0.85)	0.95	(0.61 to 1.49)	1.46	(0.92 to 2.33)	1.03	(0.95 to 1.13)	−50
Unspecified pesticides	0.86	(0.84 to 0.87)	1.06	(0.91 to 1.24)	0.91	(0.77 to 1.08)	1.07	(1.04 to 1.11)	−290

95% CIs of RRs that do not include one (i.e., p < 0.05) are highlighted in bold.

^*The number of suicides averted was calculated as the difference between the expected number of suicides (E), estimated based on pre-intervention trends, and the observed number of suicides (O) during 2012–2019 (ie, E-O). The estimates of sub-groups would not sum up to the total number (n=8353) as they were estimated from separate models in individual sub-groups with different pre-intervention trends.

^†Category of pesticide agents (International Classification of Disease, 10th Revision (ICD-10), code): herbicides and fungicides (ICD-10 T60.3), insecticides (T60.0-T60.2), rodenticides and other pesticides (T60.4, T60.8) and unspecified pesticides (T60.9).

CI, confidence interval; RR, rate ratio.

There was statistical evidence for the following sex, age and urban-rural differences (all p for interaction<0.001), although the differences were mostly small, with overlapping 95% CIs of the RR estimates. The change in pesticide suicide trends before and after the intervention was slightly greater among females (RR=0.93 vs 0.95). The relative step change in 2012 was more pronounced in older groups aged 60–69 and 70+years (RR=0.75–0.81 vs 0.85–0.87), while in 2013, it was slightly greater in the youngest group aged 15–49 years (RR=0.57 vs 0.61-0.66); the change in trends before and after the 2011–2012 pesticide bans was greater in the youngest group (RR=0.89 vs 0.91-0.92). Regarding rural-urban differences, the step change in 2012 (RR=0.83 vs 0.96-0.97) and the slope change comparing 2013–2019 with 2003–2011 (RR=0.93 vs 0.99) were most pronounced among people living in rural areas, while the step change in 2013 was most marked among people living in cities (RR=0.60 vs 0.64-0.67).

Reductions in all step and slope changes in suicides involving herbicides, fungicides and insecticides (ie, the major pesticides affected by the bans) were most pronounced across the categories of pesticide agents (all p for interaction <0.001). In 2012, suicide rates involving herbicides and fungicides declined by 21% relative to the expected based on the pre-intervention trend (2003–2011) and decreased by a further 46% in 2013. This was followed by an 11% additional annual decrease during the post-intervention period (2013–2019). Similarly, suicide rates involving insecticides were 15% and 30% lower than expected in 2012 and 2013, respectively, with a subsequent 7% additional annual decline during 2013–2019 relative to the pre-intervention trend. In contrast, suicides involving unspecified pesticides showed no reductions and even a slightly upward trend from 2013 to 2019. Sensitivity analyses, using alternative pre-intervention (2004–2011; online supplemental table 2) or post-intervention periods (2012–2019; online supplemental table 3), showed a consistent pattern.

For non-pesticide agents, suicide rates involving CO and other gases were 44% lower than expected in 2012, followed by a further 37% downward deviation in trend during 2013–2019, compared with the pre-intervention trend (online supplemental table 4). By contrast, the rates of suicide involving other and unspecified substances showed a positive 36% and 56% step change in 2012 and 2013, respectively, and showed a slight 4% upward change in the slope of the trend thereafter; however, the absolute rates remained low and declined during the post-intervention period (figure 1B). No step or slope changes were found for trends in suicides involving medicine or alcohol, organic solvents and corrosive substances.

There were an estimated 8353 fewer pesticide suicides (ie, the estimated number of suicides averted) in 2012–2019 (table 1). The largest reductions were found in males (5753 fewer suicides), the oldest group aged 70+ (6231 fewer suicides) and people living in rural areas (7619 fewer suicides) (table 1). The overall reduction in pesticide suicides was mainly attributable to a substantial decline in suicides involving herbicides (including paraquat) and fungicides, accounting for an estimated 8534 fewer suicides. In comparison, the contribution from insecticide-related suicides was smaller, with an estimated 1394 fewer suicides. Of note, the combined reduced number of pesticide suicides involving herbicides (mainly paraquat) and fungicides and insecticides markedly outnumbered the combined higher-than-expected number of pesticide suicides involving rodenticides and other pesticides (50 more suicides) and unspecified pesticides (290 more suicides).

Joinpoint regression analyses of agricultural yields from 2000 to 2019 did not identify changes temporally aligned with the 2011–2012 bans on paraquat and other insecticides (figure 3). Yields of food crops showed a continuous decline, while yields of special crops remained stable over the study period, and vegetable yields increased steadily. For fruits, the analysis identified a shift from a declining to an increasing trend in 2011, followed by a return to a declining trend from 2015. These patterns do not indicate an adverse impact of the 2011–2012 pesticide bans on crop yields.

Figure 3. Joinpoint regression results in agricultural crop yields in South Korea, 2000–2019. Note: The arrows (with numbers) indicate the years when national pesticide regulations were implemented; refer to online supplemental table 1 for the details of the numbered pesticide products banned. The area highlighted in yellow indicates the years (2011–2012) when paraquat and several insecticides, which were the main insecticides involved in suicides, were banned. The symbols represent the observed yields, and the lines represent the fitted trends using joinpoint regression models.

Discussion

Following the 2011–2012 bans on paraquat and certain insecticide pesticides, the rate of suicide from pesticide poisoning in South Korea showed a 15% (2012) and 36% (2013) reduction relative to the expected values, followed by an additional 6% annual decline in trend in pesticide suicides in 2013–2019 versus 2003–2011. Overall, there were an estimated 8353 suicides averted in 2012–2019, with a greater reduction in the number of pesticide suicides in males, older people aged 70+ and people living in rural areas. The greatest reduction in pesticide suicides was found for those involving herbicides and fungicides (including paraquat), followed by those involving insecticides. There was no evidence of a shift from suicides by pesticide poisoning to those using non-pesticide poisoning methods. The pesticide bans were not associated with obvious changes in major crop yields. These findings underscore the longer-term effectiveness of pesticide bans in reducing suicide rates, with little evidence of substitution to suicides using other substances and no indication of adverse impacts on crop production.

Strengths and limitations

This is among the first studies investigating the longer-term effectiveness of bans on multiple pesticides, including paraquat and some insecticides commonly involved in suicide, on reducing suicide rates. There are some limitations to the study. First, the ecological associations between pesticide bans and suicide rates could not be directly inferred at the individual level or interpreted as causal. The observed reduction in pesticide suicides could be due to other factors that occurred during the same periods. For example, several national suicide prevention initiatives were introduced in South Korea around the same period as the pesticide bans, most notably the enactment of the Suicide Prevention Act in 2012 and the revision of media reporting guidelines on suicide in 2013, which have been suggested to attenuate the impact of celebrity suicides on population suicide rates.³⁴ However, these initiatives were broad in scope, did not specifically target pesticide self-poisonings and were therefore unlikely to account for reductions in pesticide suicides. By contrast, the decreases observed in our study were markedly method-specific and pesticide-specific, with the largest reductions occurring in suicides involving herbicides and insecticides directly affected by the regulatory bans. This pattern makes it unlikely that the observed reductions in pesticide suicides can be fully attributed to other non-specific suicide prevention measures. Second, the large reduction in suicide rates in the category of poisonings using herbicides and fungicides may not all involve paraquat, and there is no specific ICD code for paraquat poisoning. However, previous reports indicated that paraquat was the most commonly used agent in episodes of pesticide self-poisoning in South Korea³⁵ and accounted for the majority of suicide deaths by herbicide and insecticide poisoning before the paraquat ban.²⁶ Third, the use of ICD-10 codes to identify pesticide suicides may lead to an underestimation, as suicides were likely to be underreported and misclassified. However, if the coding practices have remained consistent over time, any underestimation would be similar across the pre- and post-intervention periods. Furthermore, the accuracy of cause-of-death data has improved markedly in South Korea since 2000 by linking registered death data to several other administrative datasets.³⁶ It has been reported that the potential underestimation of pesticide suicides would not substantially affect the estimated trends in South Korea.³⁷ Lastly, the interrupted time-series analysis was based on annual data, resulting in a limited number of data points,³⁸ which may constrain the precision of trend estimation and subgroup comparisons.

Comparison with previous studies

We found 15% and 36% reductions in pesticide suicide rates in 2012 and 2013, respectively, following the 2011–2012 pesticide bans and an additional 6% annual decline in the trend during the post-intervention period (2013–2019); this suggests that the short-term effect of the pesticide bans on reducing the pesticide suicide rate in 2013 reported by Cha et al²³ was sustained over the following 6-year period (2014–2019). In line with the bans on paraquat and eight major insecticides from 2011, the reduction in pesticide suicide rates in 2013 was primarily attributable to the reduction in suicides involving herbicides (mainly paraquat) and fungicides, which were 46% lower than expected, and suicides involving insecticides, which were 30% lower than expected. The decline in post-intervention pesticide suicide trends was also mainly attributable to the reductions in suicides involving herbicides and fungicides (11% annual decrease, equating to 8534 fewer suicides during 2012–2019) and insecticides (7% annual decrease, equating to 1394 fewer suicides during 2012–2019). By contrast, since organophosphorus or pyrethroid insecticides have a relatively lower case fatality and account for a smaller proportion of poisoning cases compared with paraquat,³⁵ the overall number of deaths by insecticide poisoning averted was smaller than that from herbicides (primarily paraquat) and fungicide poisoning. Furthermore, in a survey conducted after the ban on paraquat, some South Korean farmers reported that they were still able to purchase paraquat from local markets, suggesting an incomplete implementation of the ban.³⁹ To prevent self-poisonings associated with these banned pesticide products, active and intensive intervention efforts are necessary. These could include a monitoring system for illegal and black-market transactions of prohibited pesticides, as well as collection programmes for banned pesticides stored in households.

The longer-term effect of pesticide bans on reducing pesticide suicide aligns with findings from Sri Lanka, which reported a significant decrease in pesticide suicide rates from 2011 to 2015 following the bans on paraquat, dimethoate and fenthion implemented between 2008 and 2011.¹⁵ Similarly, Bangladesh experienced a marked reduction in pesticide suicides (2001–2014) after banning all WHO Class I pesticides (eg, dichlorvos) in 2000.⁴⁰ Of note, WHO classifications of pesticide toxicity, which are based primarily on laboratory LD₅₀ (median lethal dose, i.e., the dose that causes death in 50% of the test animals) testing in rats, may not accurately reflect real-world fatality following pesticide ingestion, due to differences in human susceptibility and the complex formulations of marketed pesticides that do not correspond to doses tested under laboratory conditions.⁴¹ The sustained decline in pesticide-related suicides in South Korea, Sri Lanka and Bangladesh is likely to be attributable to the fact that the banned pesticides accounted for a large share of hospital presentations and fatalities associated with pesticide poisoning in these countries. In South Korea, paraquat had the highest case fatality among pesticides and was involved in the majority of pesticide self-poisoning deaths.^{26 29} Organophosphorus insecticides were the second most frequently used pesticides involved in pesticide self-poisoning before the bans in 2011.²² In Sri Lanka, self-poisoning cases involving organophosphorus insecticides accounted for one-third of all pesticide self-poisoning hospital presentations, with dimethoate (case fatality 21%) and fenthion (15%) being the most lethal.⁴² In Bangladesh, a nationally representative study indicated that almost half of suicides were due to pesticides.⁴³ The most frequently implicated pesticides include organophosphates, carbamates and herbicides such as paraquat.⁴⁴ Preventing access to these highly lethal pesticides can significantly lower suicide rates, as individuals may turn to less lethal alternatives or have their suicidal impulse pass without engaging in suicidal behaviour due to unavailable means.^{6 7}

Our findings highlight the critical role of targeted pesticide bans as a public health intervention to reduce suicide rates, and their effect could be more marked in certain demographic groups or areas. Our data showed a larger impact of pesticide bans on reducing pesticide suicides among men, older people and individuals living in rural areas, which could be attributed to pesticide poisoning being a common method of suicide in these populations in South Korea.^{3 43} The sex and age-specific patterns in pesticide suicide were similar to those reported in Taiwan,¹⁹ where males, the older population and people living in rural areas also have the largest burden of pesticide suicides. Although males and older adults accounted for a larger absolute number of pesticide suicides averted, the post-intervention reductions were slightly greater on the relative scale among females and the youngest age group. This pattern likely reflects differences in baseline risk and access to pesticides across subgroups. Among populations such as females and younger individuals, who had lower baseline exposure to agricultural pesticides and more limited access to pesticides stored in households, proportional changes could be larger following pesticide bans, even though the absolute number of events prevented was smaller than in their counterparts. In contrast, rural areas experienced substantial reductions on both absolute and relative scales, indicating that pesticides represented a major and less substitutable means of suicide in these settings. Accordingly, restricting access resulted not only in large decreases in the number of pesticide suicides but also in pronounced proportional declines.

Our findings indicated that the reduction in pesticide suicides following the pesticide bans was accompanied by a decline in overall suicide rates, indicating limited evidence of method substitution. There was no evidence of a shift from pesticide suicide to fatal self-poisoning using medicines, alcohol, organic solvents or corrosive substances after the pesticide bans. During the post-intervention period, although suicides involving other and unspecified pesticides occurred at rates higher than expected, the absolute rates in this category declined and remained low. These findings indicate that, even if some small degree of substitution to other substances occurred, it did not offset the effect of the pesticide bans on reducing pesticide suicide rates. Although there was an upward trend in suicides from poisoning using CO and other gases, this increase commenced in 2008 before the pesticide bans (2011) and the upward trend did not intensify after the pesticide bans. Furthermore, the demographic profile of CO poisoning deaths, which occurred more frequently in younger or middle-aged urban residents,⁴⁵ differed markedly from that of pesticide self-poisoning, which was concentrated among older adults in rural areas.¹ This distinction suggests that the observed reductions in pesticide suicides are unlikely to be explained by a shift to suicide by CO poisoning. In a recent study from India, an increase in suicides by hanging in males was found following a national ban on endosulfan in 2011.¹³ By contrast, our results showed no increase in all non-pesticide suicides during the study period. Furthermore, a recent time trend analysis from South Korea using data up to 2022 showed no increases in hanging suicide following the 2011-2012 pesticide bans.²⁵ Future research is needed to investigate whether individuals who contemplated suicide by pesticide self-poisoning shifted to CO poisoning or other suicide methods following the pesticide bans in South Korea in the longer term.

Our results provided no evidence to support the concern that restricting access to highly lethal pesticides may negatively impact agricultural productivity. There was no decline in the yields of any crop types following the pesticide bans. This aligns with findings from multiple studies in other countries such as Sri Lanka,^{15 46} India,^{14 47} Bangladesh,⁴⁰ Japan¹⁸ and Taiwan.¹⁹

The continuous reduction in deaths from pesticide self-poisoning noted in 2019 was still accompanied by a total of 782 pesticide suicides in that year. Further studies to identify the pesticides responsible for most of these deaths are required, followed by regulations to further prevent these deaths. There is an urgent need to regulate, restrict or ban HHPs commonly involved in self-poisoning to prevent premature and avoidable deaths.

Conclusion

South Korea’s bans on paraquat and several major insecticides in 2011–2012 were followed by a substantial and sustained reduction in pesticide suicides in the 7-year period (2013–2019) after the bans, without an increase in the number of suicides involving other poisoning substances. With an estimated 8000 suicides averted, the reduction in pesticide suicides was more marked in males, the elderly and people living in rural areas than in their counterparts. National policies restricting or banning HHPs could prevent many deaths from pesticide poisoning over a sustained period following the implementation of such policies.

Data availability statement

Data are available in a public, open access repository.