cShort term glucose dysregulation following acute poisoning with organophosphorus insecticides but not herbicides, carbamate or pyrethroid insecticides in South Asia

Robert Gifford,1,2* Umesh Chathuranga,2* Thomas Lamb,2* Vasundhara Verma,2* Md Abdus Sattar,3

Adrian Thompson, 1 Sisira Siribaddana,4 Aniruddha Ghose,3 Shareen Forbes,1 Rebecca M Reynolds,1 Michael Eddleston,1,2,5

1 University/BHF Centre for Cardiovascular Science, and 5 Centre for Pesticide Suicide Prevention,University of Edinburgh, UK

2 South Asian Clinical Toxicology Research Collaboration, Faculty of Medicine, University of Peradeniya, Sri Lanka

3 Department of Medicine, Faculty of Medicine, Chittagong Medical College, Chittagong, Bangladesh

4 Department of Medicine, Faculty of Medicine, Rajarata University of Sri Lanka, Sri Lanka

Running Title: Short term hyperglycaemia post-pesticide poisoning

* These authors contributed equally to this work and should be considered co-lead authors

Correspondence: M Eddleston, PTT QMRI E3.22a, 47 Little France Crescent, Edinburgh EH16 4TJ, UK. m.eddleston@ed.ac.uk

AcknowledgementsWe thank the patients, Directors, consultant physicians, medical and nursing staff of Teaching Hospital Anuradhapura and Chittagong Medical College Hospital for their support of this study and the study doctors for their valuable work. June Noble assisted with C-peptide ELISA analyses.

Disclosure statementNone of the authors has any competing financial interest to declare.

ContributionsRG, UC, TL, VV, SS, SF and ME designed and set up this prospective cohort. RG, UC, TL, and ME wrote the first draft of the paper. AT performed laboratory analyses on blood and pesticide samples. RG and RR put the Sri Lanka and Bangladesh data together and performed the statistical analysis. All authors interpreted the data and reviewed and revised the manuscript.

Funding informationThis work was funded by a pilot project grant 4689 from the Novo Nordisk Foundation and Diabetes UK project grant 13/0004682. The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript

Abstract

1

Background: Ingestion of organophosphorus (OP) insecticides is associated with acute

hyperglycaemia. We conducted a prospective study to determine whether glucose dysregulation on

admission associated with ingestion of OP insecticides or other pesticides is sustained to hospital

discharge or to 3-12 months later.

Methods: We recruited participants to two similar studies performed in parallel in Anuradhapura, Sri

Lanka, and Chittagong, Bangladesh, following hospitalisation for OP insecticide, herbicide or other

pesticide self-poisoning. Two-hour 75g oral glucose tolerance testing (OGTT) was performed after

recovery from the acute poisoning, at around the time of discharge. In Sri Lanka, a four time-point

OGTT for area-under-the-curve (AUC), C-peptide and homeostatic modelling of insulin resistance

(HOMA-IR) was undertaken, repeated after one year. In Bangladesh, a two-hour OGTT for glucose was

undertaken and repeated after three months in participants with initial elevated two-hour glucose. We

compared glucose homeostasis by poison group and adjusted findings for age, BMI and sex.

Findings: Seventy-three Sri Lankan and 151 Bangladeshi participants were recruited. We observed

higher mean [SD] fasting (4.91 [0.74] versus 4.66 [0.46] mmol/L, p=0.003) and two-hour glucose (7.94

[2.54] versus 6.71 [1.90] mmol/L, p<0.0001) in OP-poisoned groups than pyrethroid, carbamate,

herbicide or 'other poison' groups at discharge from hospital. In Sri Lanka, HOMA-IR, glucose and C-

peptide AUC were higher in OP than carbamate or herbicide groups. Adjusted analyses remained

significant except for fasting glucose. Follow-up analysis included 92 participants. There was no

significant difference in OGTT results between OP-poisoned and other participants at follow-up (mean

[SD] two-hour fasting glucose 4.67 [0.92] vs 4.82 [0.62], p=0.352; two-hour glucose 6.96 [2.31] mmol/L

vs 6.27 [1.86] mmol/L, p=0.225).

Conclusion: We found in this small prospective study that acute OP insecticide poisoning caused

acute glucose dysregulation that was sustained to hospital discharge but had recovered by three to

twelve months. Acute glucose dysregulation was related to defects in insulin action and secretion. This

study did not address long-term risk of diabetes following acute OP insecticide poisoning, but could

provide a power calculation for such a study

2

.

Keywords

Prevention of diabetes, glucose dysregulation, organophosphorus pesticides, pesticide poisoning,

pancreatic endocrine function, oral glucose tolerance test

3

INTRODUCTION

Self-poisoning with acetylcholinesterase inhibitor (organophosphorus [OP] and carbamate) insecticides

is a major healthcare problem in rural Asia and is recognised by the WHO as one of the three most

important means of suicide worldwide[1-3]. Diabetes is also a major and increasing clinical problem in

South Asia, with an estimated 82 million affected in 2017 (prevalence 8.5%) [4]. The number of patients

with diabetes in 2045 is predicted to be 151 million, a 186% increase. This marked increase reflects

rising obesity, although some have suggested chemical exposure may contribute [5]. Several

plasticisers, pesticides and solvents have been shown to alter glucose homeostasis [6]. The increasing

global production of such chemicals has been associated with a rise in diabetes, leading some to

suggest a causative link, possibly with a pronounced effect in areas like South Asia where diabetes and

pesticide use are especially prevalent [5, 7, 8].

Acute OP and carbamate insecticide poisoning is associated with hyperglycaemia, diabetic

ketoacidosis, and acute pancreatitis [9, 10]. A recent systematic review and meta-analysis of studies in

rats showed a dose-response effect of malathion poisoning on hyperglycaemia [11]. The degree of

hyperglycaemia following acute OP poisoning in humans was predictive of short-term mortality in a

recent retrospective case series (n=184, odds ratio of death = 9.1 [95% CI 1.4 to 60] for admission

glucose of >300 mg/dL [>16.7 mmol/L] compared to a glucose of <140 mg/dL [<7.78 mmol/L], p=0.022)

[12]. There is likely to be overlap between the stress and poisoning responses. Putative

pathophysiologic mechanisms are complex and not completely understood [Box 1].

Chronic exposure to OP insecticides has been associated with increased incidence of type 2 diabetes

in large cohorts in the USA [14, 15], China [16], and more recently India [7]. The proportion of diabetes

that is secondary to chronic pesticide use or other endocrine disrupting chemicals is not known [17].

Whether acute OP insecticide poisoning increases risk of diabetes has only been addressed in limited

retrospective studies demonstrating partial resolution of admission hyperglycaemia by discharge from

hospital [18, 19]. In an assessment at a median of 1.2 months after acute OP insecticide poisoning, Liu

4

et al [20], identified no increased risk of diabetes. However, as pointed out by the authors, its

retrospective design, short duration, and use of random glucose rather than a gold-standard diagnostic

technique, means that such studies are insufficiently sensitive to clearly address this question [Box 2].

Whether changes in glucose tolerance or fasting hyperglycaemia resolve fully after acute OP poisoning

has not been fully addressed.

We conducted a prospective study to assess whether dysregulation of glucose homeostasis in patients

poisoned with acetylcholinesterase inhibitors resolves fully by the time of hospital discharge. We also

looked for signals suggesting continued glucose dysregulation at follow-up over three to twelve months

to provide preliminary data for a large prospective study evaluating diabetes risk in patients surviving

self-poisoning. We used gold-standard techniques to look for evidence of dysregulation. Our

hypotheses were: (1) fasting plasma glucose and 2-hour glucose would be elevated following

organophosphate poisoning compared with other pesticides, and (2) differences would be observed

between groups at follow-up after three months or one year.

METHODS

Study design and participants

We undertook a prospective cohort study in two hospitals situated in areas with high incidences of

pesticide self-poisoning: Teaching Hospital Anuradhapura, Sri Lanka, and Chittagong Medical College

Hospital, Bangladesh, between August 2014 and July 2016. The same investigators planned the study

in both locations but, owing to local resource constraints and differences in health care systems, the

protocol differed. The Sri Lankan study was established first; an opportunity then became apparent to

set up a matched but shorter and simpler study in Bangladesh. To focus available resource on

participants most likely to answer the study question in Bangladesh, with its limited resources, only

those with deranged glucose tolerance at baseline were invited for a follow-up visit.

5

Ethical approval was received from Rajarata University of Sri Lanka Research Ethics Committee (May

2014) and Chittagong Medical College Research Ethics Committee (September 2015). Written

informed consent was obtained from each participant, or their relative (for unconscious patients), in

their own language. The study complied with the World Medical Association Declaration of Helsinki. [21]

Criteria for inclusion were age >15 years, confirmed OP, OP and pyrethroid mixture, carbamate, other

pesticide or herbicide poisoning with a named substance (identified from history or the container),

admission to hospital, and recovery to Glasgow Coma Scale (GCS) score of 15 within five days.

Exclusion criteria at baseline were pregnancy, preexisting diagnosis of diabetes, HbA1C >48 mmol/mol

(6.5%), poisoning with more than one substance, inability to give informed consent for participation

within seven days of poisoning, and normal residence beyond practical reach of researchers. Eligible

patients were identified at presentation by emergency department staff or study doctors within 24 hours

of admission. Whether participants stated the substance ingestion was intentional was recorded. All

participants were managed according to local poisoning treatment guidelines.

Anthropometric measurements

In all subjects, weight, height, hip, waist, mid-thigh and mid-upper arm circumference (MUAC) were

measured at the time of the first OGTT; measurements were repeated in Sri Lanka at 1 year follow-up.

Assessment of poisoning severity

At presentation to hospital, measures of poisoning severity were assessed. In Sri Lanka, duration of

atropine treatment, plasma butyrylcholinesterase (BuChE) activity and lactate were recorded at

admission and after eight hours. In Bangladesh, since BuChE measurement was not available,

admission markers of poisoning severity included heart rate (manually for 30 s), blood pressure

(manually using mercury sphygmomanometer) and duration of atropine treatment.

Assessment of glucose tolerance

6

Sri Lanka cohort

Plasma glucose and C peptide were measured in venous blood sampled at the point of admission and

after eight hours. Just prior to hospital discharge, or soon after discharge (n=63, median [IQR] time

since discharge 6 [4 to 12] days), participants attended after an overnight fast for an OGTT with blood

sampling at baseline, 30 min, 60 min and 120 min. After one year, all participants were invited to return

after an overnight fast for repeat HbA1c and OGTT. Samples were initially stored at -20 °C then shifted

to -80 °C storage before analysis. HOMA-IR was calculated using C-peptide and fasting glucose levels

according to the methods of Levy et al.[22]

Bangladesh cohort

Just prior to discharge from hospital, when fully recovered, and after an overnight fast (of at least 8 h),

participants underwent an OGTT with venous blood sampling at baseline and 120 min for measurement

of plasma glucose. HbA1c was measured in the fasting sample. Participants with capillary glucose ≥7.8

mmol/L at 120 min (WHO criterion for diagnosis of impaired glucose tolerance[23]) were identified and

invited to return for repeat OGTT after an overnight fast at three months.

Laboratory assays

For Sri Lankan samples, C-peptide ELISA (Mercodia AB, Uppsala, Sweden), glucose and lactate (YSI

2300 stat plus analyser, YSI Life Sciences, Tunbridge Wells, UK), and HbA1c (Adams A1c HA8180V,

Arkray, Kyoto, Japan) were performed as a single batch analysis in Edinburgh. In Bangladesh, glucose

(Siemens Dimension EXL20, Siemens, Berlin, Germany) and HbA1c (BIORAD-D10 liquid

chromatography by capillary flex piercing electrophoresis, Sebia, Evry, France) were assayed locally. It

was not possible to cross-check assays from each site. Coefficient variations for all assays was 7.2% or

less.

7

World Health Organization plasma values for diagnosis of impaired fasting glucose (IFG; ≥6.1 mmol/L

and <7.0 mmol/L), impaired glucose tolerance (IGT; 2-hour ≥7.8 and <11.1) and diabetes (fasting

glucose ≥7.0 mmol/Land 2-hour glucose ≥11.1 mmol/L) were used [23].

Statistical analysis

The primary hypotheses were that fasting and 2-hour glucose in OP insecticide poisoned patients at

hospital discharge would differ from other forms of pesticide poisoning and that glucose regulation at 3-

12 months would differ from that at hospital discharge. We first tested these hypotheses before

performing an exploratory secondary analysis, adjusting for potential confounders.

Statistical analyses were undertaken using SPSS version 23 for Mac (IBM Corporation, New York,

USA). Non-normally distributed data were transformed where possible using either square root or log

transformation. Baseline demographic and physical characteristics, surrogates of poisoning severity

and OGTT outcomes were compared by poison group using one-way ANOVA for continuous variables,

𝛘2 for categorical variables or independent samples Kruskal-Willis test for non-normally distributed

data. Area-under-the-curve (AUCs) for glucose and C-peptide were calculated using the trapezoidal

rule. Pearson’s correlation was used to investigate the relationship between glucose variables and

demographics, anthropometry and markers of poisoning severity. Further resources are available to

guide the reader on the meaning and interpretation of OGTT AUCs [24, 25].

Variables for participants who were followed up were compared with those who were not using

independent samples t-tests. We used paired t-tests to compare baseline and follow-up anthropometric

and glucose homeostasis data. At both baseline and follow-up, we adjusted for the effect of age, sex

and BMI (selected as the most consistent correlation with glucose) on two-hour glucose and glucose

AUC using linear regression. In a final model we added a marker of severity of poisoning (heart rate in

Bangladesh, admission lactate in Sri Lanka). These markers were selected as they were independent

of poisoning treatment. Statistical significance was set at p<0.05 for primary outcomes (fasting and 2-

8

hour glucose in OP-poisoned compared with other poisoning). Bonferroni adjustment was made for

poison group comparisons according to category of test: demographics: 2 variables p<0.025;

anthropometry: 4 variables p<0.013; poisoning severity: 9 variables in Sri Lanka, 8 variables in

Bangladesh, both p<0.006; glucose variables: 3 in Bangladesh, 8 in Sri Lanka, p<0.017 and 0.006,

respectively.

RESULTS

A total of 224 participants were recruited during their admission for acute pesticide poisoning in both

centre (73 Sri Lanka, 151 Bangladesh; figure 1). Three were then excluded due to an HbA1c >48

mmol/mol (7.8%) at baseline (two Sri Lanka, one Bangladesh; figure 1). In Sri Lanka, three poison

groups were formed: OP (n=29) or carbamate (n=23) insecticides, and herbicides (n=19). In

Bangladesh, due to the more diverse range of poisons, the ingested poisons were categorised into four

groups: OP insecticides (n=70), pyrethroid insecticides (n=40), OP-pyrethroid mixtures (n=17) and

carbamate or other pesticides (n=23). The pyrethroid group was included in Bangladesh as a

comparator cohort because insufficient numbers of herbicide-poisoned patients presented during the

study period. Details of substances ingested are presented in supplementary Table 1.

Clinical characteristics

The demographics, anthropometry, markers of poisoning severity and glucose homeostasis at baseline

are presented in tables 1 and 2. The relatively small size of the studies, plus the variation in poisons

selected by people of certain ages or gender, meant that differences at baseline between groups were

expected.

Sri Lankan cohort: the herbicide poisoned group contained a greater proportion of women and older

people than those in carbamate and OP groups, while the OP group had subjects with significantly

higher BMI (table 1). There was no difference in atropine treatment or hospital stay between OP and

carbamate groups. There was no difference in lactate (on admission or after eight hours) between

9

groups. Random plasma glucose on admission was higher in the OP group than other poison groups,

but this difference had resolved after eight hours (table 2).

Bangladeshi cohort: the OP-pyrethroid group contained a higher proportion of women and had

significantly greater BMI and waist circumference than the other groups but waist:hip ratios did not

differ between groups (table 1). OP and OP/pyrethroid-poisoned patients were treated with atropine

and admitted to hospital for longer than pyrethroid, herbicide or ‘other’ poison groups. There was no

difference in heart rate or blood pressure between groups on admissions.

Primary analysis: fasting and 2 hour glucose at hospital discharge and follow-up

In both cohorts, fasting and two-hour OGTT glucose concentrations were higher in OP (including OP-

pyrethroid) poisoned groups than others (mean [SD] baseline fasting glucose 4.91 [0.744] versus 4.66

[0.466] mmol/L, p=0.003; mean [SD] 2-hour glucose 7.94 [2.54] versus 6.71 [1.90] mmol/L, p<0.0001).

Forty-six participants were included in the follow-up analysis in each location. A comparison of subjects

who were and who were not followed up is shown in supplementary table 2. In Sri Lanka, there were no

significant differences in anthropometry, BP or other factors between groups. As expected, in

Bangladesh (where only patients with deranged glucose tolerance at discharge were followed up) there

were some differences between participant groups (supplementary table 2).

There was no significant difference between groups in fasting or two-hour glucose in either location at

follow up (table 3). Mean [SD] follow-up fasting glucose was 4.67 [0.92] mmol/L after OP poisoning

versus 4.82 [0.62] mmol/L after other forms of poisoning (p=0.352), while at follow up, two-hour glucose

was 6.96 [2.31] mmol/L after OP poisoning versus 6.27 [1.86] mmol/L after other forms of poisoning

(p=0.225). No significant difference in glucose homeostasis was noted in the Sri Lankan cohort

subanalysis (figure 2 and table 3). HbA1c was lower at follow-up in the OP and carbamate insecticide

groups but not the herbicide group (table 3).

10

Exploratory secondary analysis: associations between poisoning characteristics and glucose

homeostasis.

As expected, we found several OGTT variables at around the time of hospital discharge to be

correlated with surrogates of severity of illness (eg. duration of atropine treatment or hospital stay,

haemodynamics on admission, and/or BuChE activity; table 4), even after Bonferroni adjustment (see

Methods – Statstical analysis). There was no association between duration of admission and the fasting

or two-hour OGTT glucose values (r = 0.055, p=0.555; r = 0.121, p=0.195, respectively). No

correlations were significant after Bonferroni adjustment (see Methods – statistical analyses).

Sri Lankan cohort: Duration of atropine treatment correlated positively with two-hour OGTT glucose

and glucose AUC (table 4). BuChE activity on admission and after eight hours was inversely correlated

with all markers of hyperglycaemia, in particular glucose AUC (r = -0.374, p=0.002; and r = -0.441,

p<0.001, respectively). Serum lactate after eight hours demonstrated an inverse trend with fasting

OGTT glucose but positively with two-hour OGTT glucose and glucose AUC (r = -0.311, p=0.009; r =

0.262, p=0.029; and r = 0.260, p=0.030, respectively). Serum lactate on admission correlated with two-

hour glucose but not with fasting glucose, glucose AUC or HOMA-IR (table 4).

We observed significantly greater two-hour OGTT glucose and AUCs for glucose and C-peptide in the

OP group than carbamate or herbicide groups at baseline (table 2 and figure 2). Fasting OGTT glucose

was non-significantly higher in the OP group and HOMA-IR in OP and carbamate groups. After

adjustment for age, sex and BMI, fasting glucose remained insignificant and HOMA-IR became

significant, while two-hour OGTT glucose and AUCs for glucose and C-peptide remained significant.

Additional adjustment for admission lactate concentrations caused two-hour glucose to become

insignificant but did not affect other findings.

11

Bangladeshi cohort: Fasting and two-hour OGTT glucose concentrations at the time of discharge

correlated positively with markers of increasingly severe toxicity during the admission (durations of

atropine treatment and hospital stay, admission heart rate, blood pressure; table 4). We observed

significantly greater fasting and two-hour OGTT glucose concentrations in the OP and OP/pyrethroid

groups than pyrethroid or ‘other poison’ groups. This was most pronounced in the OP-pyrethroid group

(table 2 and figure 2). After adjustment for age, sex and BMI, the association with fasting glucose was

not significant, while two-hour OGTT glucose remained significant. After additional adjustment for

admission heart rate, as a marker of poisoning severity, the fasting glucose remained insignificant and

the 2-hour glucose significant.

Both cohorts. After adjustment for age, BMI and sex the elevated fasting and two hour glucose in all

OP-poisoned participants compared with others persisted (p=0.003 and p<0.0001, respectively)

Clinical criteria for pre-diabetes and diabetes at follow up

In Sri Lanka, the number of participants with plasma glucose consistent with IGT were 3 (19%), 4 (24%)

and 1 (8%) in OP, carbamate and herbicide groups, respectively. Only one participant met plasma

glucose criteria for diabetes (OP group, 6%) and one for IFG (herbicide group, 8%).

In Bangladesh, the numbers of participants with plasma glucose consistent with IGT were 3 (16%), 4

(44%), 2 (22%) and 0 (0%) for OP, OP/pyrethroid, pyrethroid and ‘other’ groups, respectively. The

numbers meeting criteria for diabetes were 2 (8%), 2 (22%), 1 (11%), and 0 (0%) participants in OP,

OP/pyrethroid, pyrethroid and ‘other’ groups, respectively. One OP-poisoned patient (4%) and one

OP/pyrethroid-poisoned patient (11%) met criteria for IFG.

DISCUSSION

This is the first prospective study assessing the longitudinal glycaemic response to acute pesticide

poisoning using the standard OGTT. While we were unable to replicate the protocol in Bangladesh and

12

Sri Lanka due to infrastructure and resource constraints, our primary finding was consistent across both

sites. Severe poisoning from OP or combined OP-pyrethroid pesticides, and not from carbamates,

pyrethroid or herbicides, was associated with fasting and two-hour hyperglycaemia around the time of

discharge after adjustment for BMI, age and sex. The deranged glucose tolerance had resolved by

three months and one year in the Bangladeshi and Sri Lankan cohorts, respectively, with no differences

between OP and control groups after adjustment for BMI, age and sex.

In our exploratory secondary analysis, we found that adjusting for multiple confounders diminished the

observed effect size. However, given each covariate adjusted for acted on primary outcomes in the

same direction, this did not refute our primary finding.

We observed acute hyperglycaemia following poisoning from OP insecticides, which appeared to be

the result of defects in insulin action in combination with a defect in insulin secretion, as has been

reported by others [7, 26, 27]. Organophosphorus poisoning did not appear to increase the risk of

developing diabetes over 3 to 12 months. This supports the findings of retrospective studies of serum

or urine glucose at admission or shortly afterwards [18-20].

The degree of acute hyperglycaemia following OP poisoning was independent of markers of illness

severity. We postulate that this observation was distinct from ‘stress hyperglycaemia’, which typically

resolves following recovery from critical illness but is associated with increased long-term risk of type 2

diabetes [13]. Acutely elevated HbA1c was recently reported in a retrospective study of stress

hyperglycaemia, implying participants who develop this condition may be at higher risk of developing

type 2 diabetes before the acute illness [28]. While HbA1c did not differ between groups at baseline or

follow-up in our study, it was modestly lower at one year follow-up in the OP and carbamate insecticide

groups.

13

The mechanisms underpinning hyperglycaemia may differ between acute and chronic exposure. In a

mouse model, Velmurugan et al found that glucose intolerance following chronic OP exposure was not

due to anticholinesterase activity, but was instead an effect of OP degradation by intestinal microbiota,

producing short chain fatty acids which induce hepatic gluconeogenesis [29].

Limited evidence is available of the effect of pyrethroid poisoning on diabetes [30]. Pyrethroid poisoning

alone was not associated with hyperglycaemia in an OGTT at hospital discharge in this study (although

others have noted hyperglycaemia on admission to hospital in pyrethroid poisoning that was associated

with complications) [31]. However, we found a greater effect on glucose regulation from combined

OP/pyrethroid poisoning than OP poisoning alone. This may reflect higher concentrations of OP within

combination pesticides than stand-alone products, as was noted in a recent review of poisoning

presentation severity in India [32].

We calculated HOMA-IR using C-peptide rather than insulin due to greater stability of the former at -20

°C. Elevated C-peptide concentrations suggest insulin secretion is present in OP and carbamate

poisoned participants at baseline. In these subjects, the presence of acute hyperglycaemia indicates

that insulin secretion is not adequate, but a raised HOMA-IR also signifies insulin resistance. This

supports prior findings from animal models that the hyperglycaemia caused by acute OP poisoning is

mediated by insulin resistance (reviewed in Lasram et al. [9]). While significantly higher baseline

HOMA-IR was noted in OP and carbamate groups than the herbicide group (after adjustment for age,

sex, BMI and lactate), no differences in HOMA-IR were observed at follow-up. Thus, it appears that an

‘unmasking’ of the insulin secretory defect occurs acutely due to an OP poisoning insult, such that OP

poisoning acts as an acute pancreatic stress test. The resolution of the hyperglycaemia is likely

secondary to the insulin secretory defect subsequently resolving.

Herbicide poisoning was not associated with glucose dysregulation, consistent with the negative

association with diabetes demonstrated by the Agricultural Health Study [15].

14

The prevalence of diabetes in Bangladesh and Sri Lanka is high (8.0% and 7.9%, respectively [33]). A

small number of participants met biochemical criteria for diabetes. Further testing and/or history-taking

would be required to confirm or refute a diagnosis. The sample size was too small to test the

associations of pesticide poisoning with IFG, IGT or type 2 diabetes.

Strengths of this study include the consistent findings in two different locations with high prevalence of

both pesticide poisoning and type 2 diabetes. We observed similar acute adverse effects of poisoning

on glucose homeostasis at baseline and resolution at follow-up despite differing follow-up protocols in

terms of subject risk of diabetes (inviting high-risk subjects in Bangladesh vs all patients in Sri Lanka),

and the timing of follow-up after the acute event (three vs 12 months). While it was not possible to

follow the Bangladeshi cohort for 12 months, we anticipate that had the Bangadeshi cohort been done

so, we would have seen the same degree of resolution as observed in Sri Lanka.

Limitations include the relatively short follow-up in this pilot study. Longer studies are required. There

was a variable loss to follow-up (19.3% in Bangladesh, 35.2% in Sri Lanka). It was not possible to

cross-check assays between both sites, however we do not think this impeded the within-site group

comparison or use of HbA1c to exclude pre-existing diabetes at each site. The unavoidable differences

in protocol between countries limits comparisons between datasets. However, importantly, the clinically

relevant outcomes of fasting and 2-hour glucose and HbA1c (as well as markers of illness severity)

from two settings were consistent, which enhances the robustness and applicability of our conclusions.

The study was limited by the exclusion of more severe poisoning (GCS <15 by day 5, imposed for

ethical reasons). Severely poisoned patients could have demonstrated higher stress hyperglycaemia

and glucose intolerance at baseline, which may have persisted at 3 months, and possibly at 1 year.

We did not collect data on occupation. It is possible that chronic OP exposure among farm workers

could have confounded our results. However, given the different mechanisms which are now

15

understood (Box 1), chronic OP exposure would be likely to have caused pre-existing diabetes,

sustained IFG and/or IGT in these patients, whereas we observed resolution of the initial insult to

glucose dysregulation, making confounding by chronic OP exposure unlikely.

In conclusion, consistent findings from animal models and human case series of acute exposure (and

from larger studies of chronic exposure) suggest an association between OP insecticide poisoning and

type 2 diabetes risk. This study did not find that acute OP-induced glucose dysregulation persisted for

3-12 months. A larger prospective study of longer duration is required to assess any association

between OP insecticide poisoning and diabetes; this study provides preliminary data for sample size

estimation for such a study. We have shown that follow-up of individuals following acute poisoning for

assessment of glucose homeostasis is feasible, even up to a year after the acute event.

16

References

1. World Health Organization. Preventing suicide: A global imperative. 2014; [cited Available from: http://www.who.int/mental_health/suicide-prevention/world_report_2014/en/. 2. Mashreky SR, Rahman F, Rahman A. Suicide kills more than 10,000 people every year in Bangladesh. Arch Suicide Res. 2013;17:387-396.3. Eddleston M, Phillips MR. Self poisoning with pesticides. In: editors. BMJ. 2004. p. 42-44. 4. International Diabetes Federation. IDF Diabetes Atlas, 8th edn. 2017; [cited Dec 2017]. Available from: http://www.diabetesatlas.org. 5. Neel BA, Sargis RM. The paradox of progress: environmental disruption of metabolism and the diabetes epidemic. Diabetes. 2011;60:1838-1848.6. Bergman A, Heindel JJ, Jobling S, et al. World Health Organization State of the Science of Endocrine Disrupting Chemicals. 2012; [cited Available from: http://http://www.unep.org/pdf/WHO_HSE_PHE_IHE_2013.1_eng.pdf. 7. Velmurugan G, Ramprasath T, Gilles M, et al. Gut Microbiota, Endocrine-Disrupting Chemicals, and the Diabetes Epidemic. Trends Endocrinol Metab. 2017;28:612-625.8. Nadal A, Quesada I, Tudurí E, et al. Endocrine-disrupting chemicals and the regulation of energy balance. Nature Reviews Endocrinology. 2017;9. Lasram MM, Dhouib IB, Annabi A, et al. A review on the molecular mechanisms involved in insulin resistance induced by organophosphorus pesticides. Toxicology. 2014;322:1-13.10. Swaminathan K. Pesticides and human diabetes: a link worth exploring? Diabet Med. 2013;30:1268-1271.11. Ramirez-Vargas M, Flores-Alfaro E, Uriostegui-Acosta M, et al. Effects of exposure to malathion on blood glucose concentration: a meta-analysis. Environmental science and pollution research international. 2017;12. Moon JM, Chun BJ, Cho YS. Hyperglycemia at presentation is associated with in hospital mortality in non-diabetic patient with organophosphate poisoning. Clin Toxicol (Phila). 2016;54:252-258.13. Ali Abdelhamid Y, Kar P, Finnis ME, et al. Stress hyperglycaemia in critically ill patients and the subsequent risk of diabetes: a systematic review and meta-analysis. Critical Care. 2016;20:301.14. Saldana TM, Basso O, Hoppin JA, et al. Pesticide exposure and self-reported gestational diabetes mellitus in the Agricultural Health Study. Diabetes Care. 2007;30:529-534.15. Montgomery MP, Kamel F, Saldana TM, et al. Incident diabetes and pesticide exposure among licensed pesticide applicators: Agricultural Health Study, 1993-2003. Am J Epidemiol. 2008;167:1235-1246.16. Huang X, Zhang C, Hu R, et al. Association between occupational exposures to pesticides with heterogeneous chemical structures and farmer health in China. Sci Rep. 2016;6:25190.17. Gifford R, Siribaddana S, Forbes S, et al. Endocrine-disrupting chemicals and the diabetes epidemic in countries in the WHO South-East Asia region. Lancet Diabetes Endocrinol. 2015;3:925-927.18. Panda S, Nanda R, Mangaraj M, et al. Glycemic status in organophosphorus poisoning. Journal of Nepal Health Research Council. 2016;19. Shobha TR, Prakash O. Glycosuria in organophosphate and carbamate poisoning. J Assoc Physicians India. 2000;48:1197-1199.20. Liu S-H, Lin J-L, Shen H-L, et al. Acute large-dose exposure to organophosphates in patients with and without diabetes mellitus: analysis of mortality rate and new-onset diabetes mellitus. Environmental Health. 2014;13:11.21. World Medical Association Declaration of Helsinki: ethical principles for medical research involving human subjects. Jama. 2013;310:2191-2194.22. Levy JC, Matthews DR, Hermans MP. Correct homeostasis model assessment (HOMA) evaluation uses the computer program. Diabetes Care. 1998;21:2191-2192.23. Alberti KG, Zimmet PZ. Definition, diagnosis and classification of diabetes mellitus and its complications. Part 1: diagnosis and classification of diabetes mellitus provisional report of a WHO consultation. Diabet Med. 1998;15:539-553.24. Sakaguchi K, Takeda K, Maeda M, et al. Glucose area under the curve during oral glucose tolerance test as an index of glucose intolerance. Diabetology International. 2016;7:53-58.25. Tschritter O, Fritsche A, Shirkavand F, et al. Assessing the shape of the glucose curve during an oral glucose tolerance test. Diabetes Care. 2003;26:1026-1033.

17

26. Song Y, Chou EL, Baecker A, et al. Endocrine-disrupting chemicals, risk of type 2 diabetes, and diabetes-related metabolic traits: A systematic review and meta-analysis. J Diabetes. 2015;27. Swaminathan K, Thangavel G. Pesticides and human diabetes: a pilot project to explore a possible link. Practical Diabetes. 2015;32:111-113.28. Du YT, Kar P, Abdelhamid YA, et al. Glycated haemoglobin is increased in critically ill patients with stress hyperglycaemia: implications for risk of diabetes in survivors of critical illness. Diabetes Res Clin Pract. 2017;29. Velmurugan G, Ramprasath T, Swaminathan K, et al. Gut microbial degradation of organophosphate insecticides-induces glucose intolerance via gluconeogenesis. Genome Biol. 2017;18:8.30. Xiao X, Clark JM, Park Y. Potential contribution of insecticide exposure and development of obesity and type 2 diabetes. Food Chem Toxicol. 2017;105:456-474.31. Kim D, Moon J, Chun B. The initial hyperglycemia in acute type II pyrethroid poisoning. J Korean Med Sci. 2015;30:365-370.32. Iyyadurai R, Peter JV, Immanuel S, et al. Organophosphate-pyrethroid combination pesticides may be associated with increased toxicity in human poisoning compared to either pesticide alone. Clinical Toxicology. 2014;52:538-541.33. WHO. Diabetes country profiles. 2016; [cited July]. Available from: http://www.who.int/diabetes/country-profiles/en/#B. 34. Marik PE, Bellomo R. Stress hyperglycemia: an essential survival response! Critical Care. 2013;17:305-305.35. Rahimi R, Abdollahi M. A review on the mechanisms involved in hyperglycemia induced by organophosphorus pesticides. Pesticide Biochemistry and Physiology. 2007;88:115-121.36. Velmurugan G, Ramprasath T, Swaminathan K, et al. Gut microbial degradation of organophosphate insecticides-induces glucose intolerance via gluconeogenesis. Genome Biol. 2017;18:37. Lukowicz C, Ellero-Simatos S, Regnier M, et al. Metabolic Effects of a Chronic Dietary Exposure to a Low-Dose Pesticide Cocktail in Mice: Sexual Dimorphism and Role of the Constitutive Androstane Receptor. Environ Health Perspect. 2018;126:067007.

18

Box 1. Poisoning and stress hyperglycaemia

Stress hyperglycaemia might be considered a purposeful process mobilising glucose to survive an

existential threat or challenge. The mechanisms are complex but result from activation of short-term

survival modalities: the sympathetic-adrenal medulla (SAM) system and hypothalamic-pituitary-adrenal

(HPA) axis [34]. The presence of stress hyperglycaemia in critical illness is associated with increased long

term risk of type 2 diabetes [13].

Organophosphorus insecticides similarly cause SAM and HPA axis activation, but there are additional

direct effects of the OP compounds which contribute to hyperglycaemia [10]. These are pleiotropic at the

cellular level, inducing cellular damage and oxidative stress [9]. However, it appears difficult to distinguish

such primary toxic effects from hyperglycaemia secondary to SAM and HPA axis activation. Ultimately,

hyperglycaemia occurs through both peripheral and central mechanisms [20]. Peripherally, glucose

production is increased (increased glycogenolysis and reduced glycogenesis in the liver and skeletal

muscle) and glucose uptake reduced (impaired glycolysis and accelerated lipolysis in peripheral tissues)

[35]. Central effects relate to pancreatic cellular damage and oxidative stress, reduced insulin secretion,

compounded by high adrenaline concentrations [9].

Recently, the mechanism by which chronic OP exposure could cause hyperglycaemia, and more

importantly increase the long-term risk of diabetes, has been elucidated and is distinct from acute OP

poisoning. Chronic microbial OP degradation leads to hepatic acetate accumulation with oxidative stress,

the downstream effect of gluconeogenesis and hyperglycaemia contributing to adiposity [36]. This may

occur at tolerable daily intake levels of OPs and may exhibit preponderance for females than males [37].

19

Box 2. Gold standard assessment of glycaemic control

Previous studies examining the effects of acute pesticide poisoning on glycaemic control have used

suboptimal tests, such as random plasma glucose concentrations. Gold-standard techniques for diagnosis

of diabetes include a 75g oral glucose tolerance test over 2 hours after an overnight fast (OGTT) or

glycated haemoglobin A1c (HbA1c). Fasting glucose levels are taken as a marker of background insulin

secretion, while glucose tolerance measures the effect of prandial insulin secretion. Insulin resistance can

be measured from fasting glucose and insulin or C-peptide concentrations, via a homeostatic modelling of

insulin resistance (HOMA-IR) [22].

20

Appendices:

Supplementary table 1 Pesticides ingested by the recruited participants by history

Supplementary table 2. Comparison of followed up vs non-followed up participants

21

Table 1. Clinical characteristics of participants at baseline

BANGLADESH SRI LANKA

All

(n=150)

OP

(n=70)

OP and

pyrethroid

(n=17)

Pyrethroid

(n=40)

Other

(n=23)

Unadjusted

p=value

All

(n=71)

Organophosphat

e

(n=29)

Carbamate

(n=23)

Herbicide

(n=19)

p=value

Demographics

Female

(n) %

58

(38.4%)

20

(28.6%)

12 (70.6%) 16

(40.0%)

10

(41.7%)

0.014 26

(37.0%)

10

(34.4%)

5

(21.7%)

11

(57.9%)

0.030

Age (years)

median (IQR)

22.0

(19.0-

30.0)

22.0 (18-

28.5)

29.0 (21.5-

40.0)

24.5 (20-

33.8)

21.0

(18.3-

27.5)

0.218 32.0

(23.0-45)

33.0

(23.0-45.0)

40.0

(30.0-55.0)

25.0

(22.0-30.0)

0.004

Intentional ingestion

(n) %

116

(77.5%)

57

(81.4%)

12 (70.6%) 30 (75.0%) 17

(75.0%)

0.708 73

(100%)

30

(100%)

23

(100%)

20

(100%)

-

Anthropometry

BMI kg/m2

(mean, SD)

20.0

(3.2)

19.2 (2.6) 22.0 (3.6) 20.7

(3.8)

19.5

(3.0)

0.003 21.1

(5.68)

22.5 (6.21) 19.4 (6.39) 21.1 (2.89) 0.055

Waist (cm) mean (SD) 71.6

(8.8)

70.4 (7.8) 74.7 (7.1) 73.0 (10.7) 70.8

(8.5)

0.019 75.2

(9.20)

77.7 (9.31) 73.3 (9.20) 75.1 (9.20) 0.335

Upper arm (cm) mean

(SD)

24.9

(3.0)

24.3 (2.5) 26.3 (3.1) 25.7 (3.1) 24.5

(3.3)

0.154 25.6

(3.34)

25.8 (3.43) 24.4 (2.82) 26.8 (3.47) 0.162

Mid thigh (cm) 39.3 38.4 (3.6) 40.0 (3.4) 41.0 (8.2) 38.6 0.063 43.1 44.0 (6.41) 40.4 (8.00) 45.2 (6.85) 0.103

22

mean (SD) (5.3) (3.8) (7.25)

Waist: hip ratio

mean (SD)

0.86

(0.10)

0.85 (0.12) 0.87 (0.05) 0.86 (0.08) 0.86

(0.06)

0.731 0.89

(0.87)

0.92 (0.04) 0.90 (0.07) 0.84 (0.04) 0.543

Indices of poisoning severity

Duration to presentation

(hours)

median (IQR)

4.0

(2.3-5.7)

4.0

(2.6-5.8)

5.0

(2.5-13.5)

4.0

(2.0-5.0)

4.5

(2.3-6.0)

0.360 2.0

(2.0-3.0)

2.0

(1.00-3.0)

2.0

(1.75-3.0)

2.0

(2.0-3.0)

0.661

Duration of atropine

treatment (days)

median (IQR)

2.0

(0.5-3.8)

3.0

(2.0-4.0)

2.8

(1.3-4.0)

0.5

(0.1-1.4)

0.5

(0.0-2.4)

<0.0001 2.00

(0.00-

4.50)

2.00 (1.00-5.00) 3.00 (1.75-

8.00)

Not

applicable‡

0.109

Admission BuChE

activity (U/L) median

(IQR)

NK NK NK NK NK 79.8

(21.2-

170)

15.0 (4.60-87.2) 124.6(43.1-

192.2)

163.2 (59.6-

238.4)

<0.0001

BuChE activity 8h post

admission(U/L), median

(IQR)

NK NK NK NK NK 62.8

(18.2-

200)

14.5(6.5-38.2) 103.7(52.5-

198.8)

192.1(54.0-

341.0)

<0.0001

Duration of admission

(days) median (IQR)

3.0 (2.0-

5.0)

4.5 (3.0-

6.0)

5.0 (3.0-6.5) 2 (2.0-3.0) 2.5 (2.0-

4.75)

<0.0001 4.0 (2.0-

6.0)

5.0 (2.5-6.5) 5.0 (3.0-

9.0)7.0)

2.0 (2.0-2.0) 0.002

Duration to Baseline

OGTT from date of

ingestion (days) median

(IQR)

3.0

(2.0-5.0)

5.0

(3.0-6.0)

5.0

(3.0-6.0)

2.0

(2.0-3.0)

2.0

(2.0-

4.75)

<0.0001 10.0

(6.0-

15.0)

9.0 (7.0-15.0) 12.0 (9.0-

17.0)

7.5 (4.0-

13.0)

0.038

Plasma lactate on NK NK NK NK NK - 1.70 1.85 (1.53-2.35) 1.65(1.30- 1.50(1.00- 0.130

23

admission (mmol/L)

median (IQR)

(1.30-

2.43)

2.58) 2.20)

Plasma lactate 8h after

admission (mmol/L)

median (IQR)

NK NK NK NK NK - 1.60

(0.13-

2.10)

1.60(1.40-2.18) 1.70(1.20-

2.10)

1.50(1.00-

2.00)

0.338

Heartrate on admission

(min-1) mean (SD)

84.3

(14.7)

83.7(14.0) 89.1 (15.2) 83.8 (15.1) 83.2

(15.9)

0.553 NK NK NK NK -

Systolic BP on admission

(mmHg) mean (SD)

107

(13.6)

107 (11.8) 109 (10.8) 110 (17.0) 107

(13.6)

0.133 NK NK NK NK -

*Participants with herbicide poisoning were not treated with atropine

Abbreviations: AUC, area under the curve; BMI, body mass index; BP, blood pressure; BuChE, Butyrylcholinesterase; HOMA IR, homeostatic

modelling assessment of insulin resistance 2; HR, heart rate; IQR, interquartile range; MUAC, mid-upper arm circumference; NK, not known; OGTT,

two hour oral glucose tolerance test; OP, organophosphorus insecticide; SD, standard deviation. P values two tailed refer to one-way ANOVA or Chi

squared (normally distributed data) or independent samples Kruskal–Wallis test (non-normally distributed data).

24

Table 2. Glucose variables during the hospital admission and baseline assessment

BANGLADESH SRI LANKA

All

(n=150)

Organophosphat

e

(n=70)

Organophosphate

and pyrethroid

(n=17)

Pyrethroid

(n=40)

Other

(n=23)

Unadjuste

d p=value

Adjusted

p=value

†Model 1

‡model 2

All

(n=71)

Organophosphate

(n=29)

Carbamate

(n=23)

Herbicide

(n=19)

p=value Adjusted p-

value

*Model 1

†model 2

Admission glucose

(mmol/L) mean (SD)

5.63

(2.03)

6.28 (2.67) 5.32 (1.46) 5.02

(1.07)

0.072 0.006*

0.034†

Glucose eight hours

after admission

(mmol/L) mean (SD)

5.38

(1.50)

5.34 (1.69) 5.57 (1.70 5.21

(0.86)

0.738 0.332*

0.751†

Fasting glucose

(mmol/L) mean (SD)

4.69

(0.63)

4.74 (0.54) 5.09 (1.18) 4.47 (0.34) 4.56

(0.56)

0.002 0.055†

0.083‡

7.00

(5.40-

8.20)

5.19 (0.76) 4.83 (0.38) 4.92

(0.54)

0.083 0.178*

0.117†

HbA1c (mmol/mol;

%) mean (SD)

35 (7);

5.4

(0.3)

35 (4); 5.3 (0.4) 37 (4); 5.6 (0.3) 35 (3); 5.4

(0.3)

35 (3)

5.4

(0.3)

0.104 0.717†

0.898‡

36 (5);

5.4

(0.5)

37 (5); 5.5 (0.4) 36 (5); 5.4

(0.5)

35 (5); 5.3

(0.5)

0.310 0.378*

0.191†

Two-hour plasma

glucose (mmol/L)

mean (SD)

7.42

(2.25)

7.56 (2.26) 9.24 (2.48) 6.94 (2.13) 6.38

(1.18)

<0.0001 0.014†

0.016‡

7.06

(2.49)

8.11 (0.76) 6.34 (2.01) 6.72

(1.94)

0.043 0.027*

0.092†

HOMA IR median

(IQR)

-NK - NK - NK - NK - - NK NK 0.770

(0.5495

-1.09)

0.820 (0.440-1.65) 0.910

(0.610-

1.31)

0.690

(0.460-

0.820)

0.102 0.021*

0.019†

AUC Glucose

(mmolmin/L) mean

- NK - NK NK - - NK - NK - NK NK 917

(210)

996 (255) 887 (149) 838 (160) 0.024 0.011*

0.039†

25

(SD)

AUC C-peptide

(nmolmin/l) mean

(SD)

- NK - NK NK - - NK - NK - NK NK 177

(118-

254)

233 (135-303) 186 (122-

243)

132 (86.8-

192)

0.007 0.003*

0.003†

*Model 1 – adjusted for Age, Sex and BMI

†Model 2 – adjusted for Age, Sex, BMI and marker of illness severity (Bangladesh: admission heart rate, Sri Lanka: admission lactate)

AUC, area under the curve; BMI, body mass index; BP, blood pressure; BuChE, Butyrylcholinesterase; HbA1c, glycosylated haemoglobin A1; HOMA

IR, homeostatic modelling assessment of insulin resistance 2; IQR, interquartile range; MUAC, NK, not knownmid-upper arm circumference; OGTT,

two hour oral glucose tolerance test; OP, organophosphorus insecticide; SD, standard deviation; Other, other pesticide ( carbamates, abamectins,

avermectin, emermectin, conazoles, zinc phosphide, copper oxychloride) P values refer to one-way ANOVA (normally distributed data) or

independent samples Kruskal–Wallis test (non-normally distributed data).

26

Table 3: Anthropometric and glucose variables at follow-up.

BANGLADESH (followed up at three months) n =46 SRI LANKA (followed up at one year) n = 46

All

(n=46)

Organophosphat

e

(n=25)

Organophosphate

and pyrethroid

(n=9)

Prethroid

(n=9)

Other

(n=3)

Unadjusted

p value

Adjusted p

value*

All

(n=46)

Organophoshate

(n=16)

Carbamate

(n=17)

Herbicide

(n=13)

Unadjusted

p value

Adjusted

p value*

Follow-up

interval (days)

median (IQR)

85.0

(74.0-

91.2)

84.0 (73.0-91.0) 85.0 (80.5-91.0) 87.0

(72.0-

102)

91.0

(-)

0.597 365

(355-

379)

363 (350-379) 368 (357-

397)

365 (363-

368)

0.822

BMI (g/km2)

median

(IQR)

Baseline 20.2

(18.0-

23.3)

18.35 (17.1-22.0) 20.4 (18.3-13.5) 23.4

(21.8-

27.9)

19.15

(18.5-

19.8)

0.025 19.6

(17.8-

22.1)

20.4 (18.4 –

23.0)

17.9 (15.2-

21.7)

20.6

(18.9-

22.8)

0.588

Follow-

up

20.7

(18.7-

23.9)

20.9 (19.5-24.3) 19.9 (18.0-

24.0)

21.0

(18.8-

23.1)

0.957

P value 0.042 0.068 0.172 0.343

Waist (cm),

mean (SD)

Baseline 73.7

(8.96)

71.7 (9.24) 74.8 (6.13) 78.9

(9.23)

72.0

(10.8)

0.224 75.2

(9.78)

79.1 (10.2) 73.6 (9.59) 72.5

(8.68)

0.139

Follow-

up

85.3

(9.14)

89.0 (9.57) 82.6 (7.72) 84.15

(0.42)

0.109

P value <0.0001 <0.0001 0.003 <0.0001

MUAC (cm)

mean, SD)

Baseline 25.6

(2.98)

24.8 (2.93) 26.2 (2.64) 27.4

(3.22)

25.2

(3.03)

0.147 25.6

(3.21)

25.6 (3.59) 24.8 (2.99) 26.5

(2.96)

0.332

Follow-

up

26.9

(3.12)

27.2 (3.89) 25.9 (2.18) 26.9

(3.13)

0.192

27

P value 0.002 0.114 0.114 0.011

Mid-thigh

(cm) mean,

SD))

Baseline 39.2

(4.29)

38.1 (4.10) 39.2 (2.35) 42.4

(5.88)

39.18

(4.37)

0.116 43.4

(5.32))

43.8 (5.32) 41.1 (8.69) 45.9

(6.29)

0.185

Follow-

up

44.7

(7.19)

45.6 (8.75) 42.7 (5.74) 46.2

(7.19)

0.348

P value 0.188 0.25 0.47 0.75

Fasting

plasma

glucose

(mmol/L)

mean (SD)

Baseline 4.97

(0.82)

4.94 (0.54) 5.47 (1.53) 4.69

(0.39)

4.70

(0.47)

0.185 0.756 4.91

(0.53)

5.09 (0.14) 4.23 (0.36) 4.91

(0.18)

0.146 0.276

Follow-

up

4.60

(0.72)

4.45 (0.70) 5.02 (0.87) 4.73

(0.56)

4.30

(0.36)

0.165 0.891 4.86

(0.87)

4.81 (1.20) 4.78 (0.61) 5.04

(0.69)

0.689 0.387

P value 0.016 0.0025 0.742 0.886 0.470 0.763 0.344 0.793 0.616

Two-hour

glucose

(mmol/L)

mean (SD)

Baseline 9.83

(1.68)

9.84 (1.41) 10.1 (2.41) 9.83

(1.74)

8.63

(1.15)

0.254 0.373 7.06

(2.48)

8.11 (0.77) 6.34 (0.49) 6.72

(0.55)

0.102 0.340

Follow-

up

7.05

(2.39)

6.50 (2.20) 8.72 (2.83) 7.16

(2.27)

6.23

(1.04)

0.106 0.879 6.23

(1.77)

6.66 (1.77) 5.94 (1.82) 6.07

(1.76)

0.484 0.375

P value <0.0001 <0.0001 0.378 0.190 0.0223 0.025 0.074 0.410 0.319

HbA1c

(mmol/mol)

mean (SD)

Baseline 36 (4);

5.4

(0.4)

36 (4); 5.4 (0.4) 38 (4); 5.6 (0.4) 36 (4); 5.4

(0.3)

35 (4);

5.1

(0.3)

0.152 0.408 36 (6);

5.4

(0.6)

37 (5); 5.5 (0.4) 37 (5); 5.5

(0.5)

33 (8); 5.1

(0.7)

0.119 0.211

Follow-

up

33 (5);

5.1

(0.5)

33 (5); 5.2 (0.5) 32 (5); 5.1

(0.5)

33 (4); 5.2

(0.4)

0.759 0.605

P value 0.01 0.026 0.007 0.655

28

HOMA IR,

median (IQR)

Baseline 0.79

(0.49-

1.04)

0.90 (0.44-1.32) 0.91 (0.60-

1.20)

0.69

(0.46-

0.80)

0.207 0.300

Follow-

up

0.71

(0.42 –

1.04)

0.65 (0.43-0.99) 0.46 (0.43-

1.03)

0.96

(0.47-

1.24)

0.428 0.256

P value 0.209 0.273 0.066 0.19

AUC Glucose

(mmolmin/L)

mean (SD)

Baseline 919

(196)

1010 (248) 875 (155) 870 (140) 0.091 0.259

Follow-

up

864

(205)

864 (201) 876 (236) 850 (181) 0.944 0.893

P value 0.142 0.064 0.984 0.698

AUC C-

peptide

(nmolmin/L)

median (IQR)

Baseline 185

(118-

246)

222 (117-330) 171 (119-

249)

183 (93.7-

209)

0.176 0.116

Follow-

up

132

(55.9-

188)

121 (63.6-188) 98.8 (48.6-

183)

150 (54.1-

200)

0.692 0.438

P value <0.0001 0.002 0.002 0.459

* Model for adjusted group comparison included BMI, age and sex

Abbreviations: AUC, area under the curve; BMI, body mass index; BP, blood pressure; BuChE, Butyrylcholinesterase; HbA1c, glycosylated

haemoglobin A1; HOMA IR, homeostatic modelling assessment of insulin resistance 2; IQR, interquartile range; MUAC, mid-upper arm

circumference; OGTT, two hour oral glucose tolerance test; OP, organophosphorus insecticide; Other, other pesticide (carbamate, abamectin,

29

avermectin); SD, standard deviation. P values between groups refer to one-way ANOVA. P values within groups (baseline versus follow-up) refer to

paired samples t tests.

Table 4. Pearson’s correlations between baseline OGTT glucose variables and markers of illness severity

Bangladesh Sri LankaFasting glucose

2-hour glucose

Fasting glucose

2-hour glucose

Glucose AUC

HOMA-IR

Duration of admission

0.244 (p=0.003)*

0.215 (p=0.008)*

0.019 (p=0.876)

0.097 (p=0.421)

0.017 (p=0.885)

0.071 (p=0.554)

Duration of atropine treatment

0.233 (p=0.008)*

0.307 (p<0.001)*

-0.123 (p=0.425)

0.312 (p=0.039)

0.308 (p=0.042)

0.154 (p=0.348)

Admission HR 0.210 (p=0.010)*

0.385 (p<0.001)*

Admission SBP 0.234 (p=0.005)*

0.287 (p<0.001)*

BuChE on admission

-0.251 (p=0.041)

-0.291 (p=0.017)

-0.374 (p=0.002)*

-0.329 (p=0.007*)

BuChE after 8 hours

-0.279 (p=0.021)

-0.358 (p=0.003)*

r = -0.441, (p<0.001)*

-0.295 (p=0.015)

Lactate on admission

-0.014 (p=0.909)

0.305, (p=0.010)

0.260 (p=0.030)

0.003 (p=0.980)

Lactate after 8 hours

-0.311 (p=0.009)

0.264 (p=0.029)

0.156 (p=0.201)

0.100 (p=0.413)

Abbreviations: AUC, area under the curve; BMI, body mass index; BP, blood pressure; BuChE, Butyrylcholinesterase; HOMA IR, homeostatic modelling assessment of insulin resistance 2; HR heart, rate; IQR, interquartile range; OGTT, two hour oral glucose tolerance test; OP, organophosphorus insecticide; SD, standard deviation. After Bonferroni adjustment, singificance was set at p<0.017 for Bangladesh and p<0.008 for Sri Lanka. *denotes significance.

30

31

Figure 1.

32

Figure 2.

33

Figure captions

Figure 1. Recruitment and follow-up of patients in the study. BuChE, butyrylcholinesterase; HbA1c,

glycated haemoglobin OP; OGTT, oral glucose tolerance test; organophosphate insecticide

Figure 2. Results of the 75g oral glucose tolerance test (OGTT) at baseline in Bangladesh (figure 2A) and

Sri Lanka (figure 2B), and after follow-up in Bangladesh (3 months, figure 2C) and in Sri Lanka (1 year,

figure 2D). Organophosphate (filled circle, dotted line), organophosphate-pyrethroid (filled square),

pyrethroid (triangle pointing upwards), herbicide (empty circle, solid line), carbamate (empty square,

dashed line) other (triangle pointing downwards), *** p<0.0001, * p<0.05, NS p>0.05

34

Supplementary table 1 Pesticides ingested by the recruited participants by history

Bangladesh Sri LankaPesticide Number Pesticide Number

OP insecticides MalathionChlorpyrifosDimethoatePhenthoate DiazinonQuinalphos

212015653

(n=70)

ProfenofosChlorpyrifosDiazinonQuinalphos

Unknown organophosphate

10542

9(n=30)

Pyrethroids CypermethrinLambda cyhalothrinFenvaleratePermethrin

201541

(n=41)

OP & pyrethroid combinations

Chlorpyrifos & cypermethrinProfenofos & cypermethrin

152

(n=17)Carbamates Carbofuran

Carbosulfan33

(n=6)

CarbofuranCarbosulfan

1013

(n=23)Herbicides Glyphosate 1

(n=1)GlyphosateMCPAPropanil

1721

(n=20)Others Abermectins/emermectin

Conazoles86

35

Zinc phosphideCopper oxichloride

21

(n=17)

Abbreviations: IQR, interquartile range; MCPA, 2-methyl-4-chlorophenoxyacetic acid; OP, organophosphate insecticide; SD, standard deviation.

36

Supplementary table 2. Comparison of followed up vs non-followed up participants

BANGLADESH SRI LANKA

All(n=150)

Followed up (n=46)

Not followed up (n=104)

Comparison followed up participants

vs not followed up participants

All(n=71)

Followed up (n=46)

Not followed up (n-25)

Comparison followed up

participants vs not followed up

participants

Demographics Female (N) %

58(38.4%)

20 (43.5%) 38(36.6%) 0.446 73 (37.0%) 16 (34.7%) 10 (28.6%) 0.8827

Age median (IQR)

22.0 (19.0-30.0)

26.0 (20.0-36.5) 22 (19-28) 0.026 (Mann Whitney U)

32.0(23.0-45)

30.5 (22.3-44.5) 30.5 (34.0-37.5)

0.221

Intentional ingestion(n) %

116 (77.5%) 32 (69.6%) 84 (80.8%) 0.131 73(100%)

46 (100%) 25 (100%) -

Anthropometry BMI kg/m2

(mean, SD)20.0 (3.2) 20.3 (4.88) 19.6 (2.94) 0.296 21.1 (5.68) 20.8 (5.01) 22.9 (6.77) 0.274

Waist (cm) mean (SD) 71.6(8.8)

25.1 (4.86) 24.6 (2.87) 0.452 75.2 (9.20) 73.3 (8.72) 76.4 (9.13) 0.454

Upper arm (cm) mean (SD)

24.9 (3.0) 72.2 (14.2) 70.5 (8.60) 0.392 25.6 (3.34) 25.4 (3.26) 25.8 (3.60) 0.704

Mid thigh (cm) mean (SD)

39.3 (5.3) 38.3 (7.31) 39.3 (5.79) 0.413 43.1 (7.25) 43.4 (7.12) 42.8 (1.53) 0.745

Waist hip ratioMean (SD)

0.86 (0.07) 0.88 (0.07) 0.85 (0.06) 0.057 0.89 (0.87) 0.90 (0.28) 0.88 (0.16) 0.713

Indices of poisoning severity

Duration to presentation (hours)

median (IQR)

4.0 (2.3-5.7)

4.0 (2.0-4.0) 4.0 (2.0-5.5) 0.402 2.0(2.0-3.0)

2.0 (2.0-3.0) 2.0 (1.3-3.0) 0.771

Duration of atropine treatment (days) median (IQR)

2.0 (0.5-3.8)

3.0 (2.0-4.0) 2.0 (0.50-3.0) 0.001 2.0 (0.0-4.5)

2.0 (0.0-48) 1.0 (0.0-4.8) 0.977

Admission BuChE activity (U/L) median

(IQR)

NK NK NK - 79.8 (21.2-170)

112 (48.9-181) 102 (14.5-213) 0.182

37

BuChE activity 8h post admission(U/L),

median (IQR)

NK NK NK - 62.8 (18.2-200)

112 (48.8-181) 103 (14.5-213) 0.129

Duration to Baseline OGTT from date of

ingestion (days) median (IQR)

3.0(2.0-5.0)

4.5 (3.0-6.0) 3.0 (2.0-5.0) 0.003 10.0 (6.0-15.0)

10.0 (8.0-15.0) 9.0 (5.0-14.0) 0.878

Duration of admission (days) median (IQR)

3.0 (2.0-5.0) 4.5 (3.0-6.0) 3.0 (2.0-5.0) 0.073

Plasma lactate on admission (mmol/L)

median (IQR)

NK NK NK - 1.70 (1.30-2.43)

1.65 (1.30-2.20) 2.05 (1.10-2.57)

0.679

Plasma lactate 8h after admission (mmol/L)

median (IQR)

NK NK NK - 1.60 (.125-2.10)

1.60 (1.33-2.20) 1.65 (1.05-3.45)

0.199

Heartrate on admission (min-1) mean

(SD)

84.3 (14.7) 91.6 (15.1) 81.5 (13.5) 0.001 NK -

Systolic BP on admission (mmHg)

mean (SD)

107 (13.6) 113 (16.5) 105 (11.4) 0.003 NK -

Glucose homeostasis Fasting glucose

(mmol/L) mean (SD)

4.69 (0.63) 4.98 (0.84) 4.56 (0.472) 0.002 7.00 (5.40-8.20)

4.92 (0.53) 5.15 (0.74) 0.122

HbA1c (mmol/mol; %)

mean (SD)

35 (4); 5.4 (0.6)

36 (4); 5.4 (0.6) 35 (3); 5.4 (0.5) 0.184 36 (5); 5.4 (0.7)

36 (6); 5.4 (0.8) 35 (4); 5.4 (0.6)

0.736

Two-hour plasma glucose (mmol/L)

mean (SD)

7.42 (2.25) 9.81 (1.72) 6.41 (1.60) <0.0001 7.06 (2.49) 7.07 (2.49) 7.42 (2.61) 0.570

HOMA IR median (IQR)

- - - - 0.77 (0.55-1.09)

0.73 (0.45-0.96) 0.77 (0.50-1.17)

0.568

AUC Glucose (mmolmin/L)

mean (SD)

- - - - 917 (210) 925 (202) 906 (229) 0.728

AUC C-peptide (nmolmin/l)

median (IQR)

177 (118-254)

182 (116-251) 158 (120-276) 0.840

38

Abbreviations: AUC, area under the curve; BMI, body mass index; BP, blood pressure; BuChE, Butyrylcholinesterase; HbA1c, glycosylated haemoglobin A1; HOMA IR, homeostatic modelling assessment of insulin resistance 2; IQR, interquartile range; OGTT, two hour oral glucose tolerance test; OP, organophosphorus insecticide; SD, standard deviation. P values are for independent samples t test (normally distributed data) or independent samples Mann Whitney U tests (non-normally distributed data).

39

Supplementary table 3. Pearson’s correlations between baseline OGTT glucose variables and markers of illness severity

Bangladesh Sri LankaFasted glucose

2-hour glucose

Fasted glucose

2-hour glucose

Glucose AUC

HOMA-IR

Duration of admission

0.244 (p=0.003)

0.215 (p=0.008)

0.019 (p=0.876)

0.097 (p=0.421)

0.017 (p=0.885)

0.071 (p=0.554)

Duration of atropine treatment

0.233 (p=0.008)

0.307 (p<0.001)

-0.123 (p=0.425)

0.312 (p=0.039)

0.308 (p=0.042)

0.154 (p=0.348)

Admission HR 0.210 (p=0.010)

0.385 (p<0.001)

Admission SBP 0.234 (p=0.005)

0.287 (p<0.001)

BuChE on admission

-0.251 (p=0.041)

-0.291 (p=0.017)

-0.374 (p=0.002)

-0.329 (p=0.007)

BuChE after 8 hours

-0.279 (p=0.021)

-0.358 (p=0.003)

r = -0.441, (p<0.001)

-0.295 (p=0.015)

Lactate on admission

-0.014 (p=0.909)

0.305, (p=0.010)

0.260 (p=0.030)

0.003 (p=0.980)

Lactate after 8 hours

-0.311 (p=0.009)

0.264 (p=0.029)

0.156 (p=0.201)

0.100 (p=0.413)

Abbreviations: AUC, area under the curve; BMI, body mass index; BP, blood pressure; BuChE, Butyrylcholinesterase; HOMA IR, homeostatic modelling assessment of insulin resistance 2; HR heart, rate; IQR, interquartile range; OGTT, two hour oral glucose tolerance test; OP, organophosphorus insecticide; SD, standard deviation.

40