Point-of-care testing in the early diagnosis of acute pesticide intoxication: The example of paraquat

Abstract

Acute pesticide intoxication is a common method of suicide globally. This article reviews current diagnostic methods and makes suggestions for future development. In the case of paraquat intoxication, it is characterized by multi-organ failure, causing substantial mortality and morbidity. Early diagnosis may save the life of a paraquat intoxication patient. Conventional paraquat intoxication diagnostic methods, such as symptom review and urine sodium dithionite assay, are time-consuming and impractical in resource-scarce areas where most intoxication cases occur. Several experimental and clinical studies have shown the potential of portable Surface Enhanced Raman Scattering (SERS), paper-based devices, and machine learning for paraquat intoxication diagnosis. Portable SERS and new SERS substrates maintain the sensitivity of SERS while being less costly and more convenient than conventional SERS. Paper-based devices provide the advantages of price and portability. Machine learning algorithms can be implemented as a mobile phone application and facilitate diagnosis in resource-limited areas. Although these methods have not yet met all features of an ideal diagnostic method, the combination and development of these methods offer much promise.

INTRODUCTION

According to a report released by World Health Organization in 2014,¹ acute pesticide intoxication is a common method of suicide globally. For example, in Taiwan, it was reported² that the rate of attempting suicide by ingesting pesticide in 2010 was one-third of that in 1987. Nevertheless, pesticide intoxication remains the third most common method of attempting suicide in 2010, with the age-standardized rate of 2.5 per 100 000, and accounted for 12% of all suicides.³ Our study⁴ also revealed that nearly all patients who had been admitted due to acute pesticide suicide had mental illness, and the major psychiatric diagnoses were depressive disorders (58%), substance use disorders (35%), and adjustment disorder (28%).

Theoretically, pesticides include bactericides, baits, fungicides, herbicides, insecticides, lures, rodenticides, and repellents. Paraquat (PQ) is a widely used bipyridyl contact herbicide with a good safety record when used properly but can cause substantial morbidity and mortality following ingestion. Intentional ingestion of PQ is common in Taiwan because of easy access.⁵ Gramoxone (Syngenta, Taiwan) is available as 24% PQ in the Taiwan market and is used by farmers to control annual and perennial weeds for many kinds of crops. Agricultural workers and their families often have access to PQ in places of work and in their homes, which increases their chance of exposure.

PARAQUAT INTOXICATION

Clinical features

There are three degrees of severity for PQ poisoning.⁶ Mild poisoning causes oral irritation and gastrointestinal upset but eventually results in complete recovery. Moderate to severe poisoning produces acute renal failure and in severe cases acute hepatitis followed by pneumonitis or pulmonary fibrosis, often leading to death in 2–3 weeks. Acute fulminant poisoning results in death within a week due to multisystem organ failure and cardiovascular collapse.

The mechanisms of PQ toxicity involve generation of the superoxide anion, which leads to both the formation of more toxic reactive oxygen species, such as hydrogen peroxide and hydroxyl radicals, and the oxidation of the cellular NADPH (Nicotinamide adenine dinucleotide phosphate), the major source of reducing equivalents for the intracellular reduction in PQ, which results in the disruption of important NADPH-requiring biochemical processes.⁷ The primary cause of mortality in PQ poisoning is respiratory failure due to an oxidative insult to the alveolar epithelium with subsequent obliterating fibrosis or acute respiratory distress syndrome.⁸

Clinical outcomes

A review of the medical literature indicates that most of the published reports of PQ intoxication are from Asian countries. This is an unsurprising finding because PQ herbicide is commonly used in these developing countries. Moreover, the overall mortality rates indicated in the review are high despite intensive resuscitation.^5,9–74 Table I shows the original PQ intoxication report.

TABLE I.

Published original studies (sample size more than 10) of paraquat intoxication from different hospitals.

Study	Year	Area	Sample sizea	Treatment	Mortality rate (%)	Note
Weng et al.9	2017	Taiwan	222	Hemoperfusion and immunosuppressive therapy	54.1
Gao et al.10	2017	China	138	Hemoperfusion and immunosuppressive therapy	54.3
Hu et al.11	2017	China	103		50.5
Chen et al.12	2017	China	79		48.1
Liu et al.13	2016	China	87	Hemoperfusion	41.4
Fortenberry et al.14	2016	USA	300		8.0%	Data derived from systems
Kim et al.15	2016	Korea	297		65.3
Duan and Wang16	2016	China	146		14.4	Pediatric population
Park et al.17	2016	Korea	263	Consecutive hemoperfusion and hemodialysis versus concurrent hemoperfusion and hemodialysis	81.8 versus 57.9
Li et al.18	2015	China	177		37.9
Mohamed et al.19	2015	Sri Lanka	50		24.0
Zhou et al.20	2015	China	75		52.0	Cases who had computed tomography study
Gong et al.21	2015	China	68	Hemoperfusion, immunosuppressive therapy, antioxidant versus hemoperfusion, immunosuppressive therapy, and antioxidant, Xuebijing	51.2 versus 66.7
Kang et al.22	2015	China	97	Hemoperfusion, and immunosuppressive therapy	42.3	Cases who had computed tomography study
Mohamed et al.23	2015	Sri Lanka	66		25.8
Zhang et al.24	2015	China	205		32.2
Liu et al.25	2015	China	86	Hemoperfusion	75.6
Gao et al.26	2015	China	684	Hemoperfusion, antioxidant versus hemoperfusion, continuous venovenous hemofiltration, and antioxidant	57.4 versus 58.4
Xu et al.27	2015	China	143	Hemoperfusion, immunosuppressive therapy, and antioxidant	46.2
Hong et al.28	2014	Korea	2136	Hemoperfusion, immunosuppressive therapy, and antioxidant	44.0
Lin et al.5	2014	Taiwan	157	Hemoperfusion and immunosuppressive therapy	58.0	Cases who had liaison psychiatry consultation
Lin et al.29	2014	Taiwan	60	Hemoperfusion and immunosuppressive therapy	88.3	Cases who had electrocardiogram study
Ge et al.30	2014	China	23	Hemoperfusion and immunosuppressive therapy	63.6	Pediatric population
Wu et al.31	2014	Taiwan	1811	Hemoperfusion and immunosuppressive therapy	52.0	Data derived from National Health Insurance
Liu et al.32	2013	China	204		72.5
Zhang et al.33	2013	China	78	Hemoperfusion and antioxidant	46.2	Cases who had computed tomography study
Kervegant et al.34	2013	France	34		44.1	Data derived from systems
Yang et al.35	2012	Taiwan	187	Hemoperfusion and immunosuppressive therapy	54.5
Seok et al.36	2012	Korea	41	Hemoperfusion and antioxidant	51.2
Lee et al.37	2012	Korea	272	Hemoperfusion	81.6
Hsu et al.38	2012	Taiwan	207	Hemoperfusion and immunosuppressive therapy	68.6
Shi et al.39	2012	China	85	Hemoperfusion, immunosuppressive therapy, and antioxidant	51.2
Min et al.40	2011	Korea	102	Hemoperfusion and immunosuppressive therapy	76.5
Huang and Zhang41	2011	Korea	138	Hemoperfusion, immunosuppressive therapy, and antioxidant	55.8
Moon and Chun42	2011	Korea	134	Immunosuppressive therapy versus immunosuppressive therapy plus vitamin C	78.1 versus 70.1
Lin et al.43	2011	Taiwan	111	Immunosuppressive therapy versus repeated immunosuppressive therapy	92.0 versus 66.0	Cases with severe poisoning
Kim et al.44	2010	Korea	247		42.9
Roberts et al.45	2011	Sri Lanka	20		30.0
Zhang et al.46	2011	China	25	Hemoperfusion, hemodialysis, and immunosuppressive therapy	40.0
Yen et al.47	2010	Taiwan	16	Hemoperfusion and immunosuppressive therapy	12.5	Cases who had endoscopy study
Koo et al.48	2009	Korea	233	Immunosuppressive therapy, and continuous venovenous hemofiltration	58.4
Gil et al.49	2009	Korea	20		35.0	Cases who had urinary biomarker study
Seok et al.50	2009	Korea	250		58.4
Kim et al.51	2009	Korea	119		14.3	Cases who had computed tomography study
Kim et al.52	2009	Korea	278		58.8
Senarathna et al.69	2009	Sri Lanka	451	Immunosuppressive therapy	61.0
Yang et al.70	2009	Korea	296	Hemoperfusion	57.8
Afzali and Gholyaf 53	2008	Iran	20	Supportive versus immunosuppressive therapy	81.8 versus 33.3
Gil et al.54	2008	Korea	375	Hemoperfusion	70.7
Lin et al.55	2006	Taiwan	23	Steroid versus repeated immunosuppressive therapy	85.7 versus 31.3
Huang et al.71	2006	Taiwan	64		71.9
Hong et al.72	2005	Korea	21	Hemoperfusion	66.7
Huang et al.56,71	2003	Taiwan	58		72.4
Lugo-Vallín et al.57	2003	Venezuela	35	S-carboxymethylcysteine	22.9
Lee et al.58	2002	Korea	602		41.5
Hwang et al.59	2002	Korea	154		43.8
Hong et al.73	2000	Korea	147		44.2
Yamamoto et al.74	2000	Japan	43		72.1
Kao et al.60	1999	Taiwan	13		69.0	Cases who had nuclear pulmonary study
Lin et al.61	1999	Taiwan	60	Supportive versus immunosuppressive therapy	57.1 versus 18.2	Cases with moderate to severe poisoning
Lin et al.62	1996	Taiwan	33	Supportive versus immunosuppressive therapy	70.6 versus 25.0	Cases with moderate to severe poisoning
Ragoucy-Sengler and Pileire63	1996	Guadeloupe	18		66.7
Lin et al.64	1995	Taiwan	21		23.8	Cases who had pulmonary function tests
Tinoco et al.65	1993	Mexico	25		64.0
Im et al.66	1991	Korea	42		76.0
Kaojarern and Ongphiphadhanakul67	1991	Thailand	24		71.0
Wright et al.68	1978	UK	16		44.0

^aThe sample number is the number of PQ intoxication patients.

Clinical therapy

Various treatment modalities have been developed for PQ poisoning: hypo-oxygenation;^75,76 lung radiotherapy;⁷⁷ prolonged extracorporeal detoxification;⁷⁸ antioxidants such as vitamin C;⁴² S-carboxymethylcysteine;⁵⁷ vitamin E⁷⁹ or nitric oxide inhalation;⁸⁰ Xuebijing, a traditional Chinese medicine;⁸¹ and lung transplantation.⁸² Nevertheless, the usefulness of these approaches remains indeterminate.

In previous studies,^61,62 we have demonstrated that an initial pulse therapy of cyclophosphamide (15 mg/kg day) for 2 days and methylprednisolone (1 g/day) for 3 days simultaneously, followed by dexamethasone (20 mg/day) for 14 days may be effective in treating patients with moderate to severe PQ poisoning. In another study,⁵⁵ 23 patients with 50%–90% predictive mortality based on plasma PQ levels were prospectively and randomly assigned to the control and study groups at a proportion of 1:2. The control group received conventional therapy, and the study group received the repeated pulse treatment with long-term steroid therapy. The mortality rate of the control group (85.7%) was higher than that of the study group (31.3%, P = 0.027). The data suggest that anti-inflammatory treatment may reduce the mortality of patients with severe PQ poisoning.⁵⁵

Methylprednisolone pulse therapy, cyclophosphamide, and dexamethasone have been useful therapeutic anti-inflammatory approaches, indicating that PQ's method of action may be severe pulmonary inflammation (rather than lung fibrosis) that leads to lethal hypoxemia. Pulse therapy following dexamethasone treatment can further attenuate severe inflammation from PQ poisoning; its use in our studies improved the survival rate.^55,61,62 In addition, methylprednisolone pulse therapy can suppress superoxide production by neutrophils and macrophages by reducing cytokine and inflammatory mediator release from lymphocytes.⁸³ Repeated methylprednisolone pulse therapy, then, significantly shields patients from further pulmonary free radical damage and subsequent inflammation. Further, steroids in general have been shown to suppress the formation of superoxides in the arachidonic acid cascade and could prove to be useful in future combination approaches.

A standard detoxification protocol^38,43 has been recommended by Cochrane Injuries Group as a beneficial treatment for patients with PQ-induced lung fibrosis.⁸⁴ This protocol includes gastric lavage with large amounts of normal saline, followed by infusion of 1 g/kg activated charcoal and 250 ml of magnesium citrate via a nasogastric tube. Charcoal hemoperfusion is performed for 8 h using a charcoal-containing (Adsorba, Gambro, Germany) dialysis machine (Surdial, Nipro, Japan) if urine PQ is >5 ppm. The extracorporeal technique is repeated if urine PQ remains >5 ppm at 4 h after the first extracorporeal detoxification. Furthermore, all patients receive pulse therapies of cyclophosphamide (15 mg/kg day) for two days and methylprednisolone (1 g/day) for 3 days, simultaneously, followed by dexamethasone (20 mg/day) for 14 days. Steroid pulse therapy is repeated if PaO₂ is <60 mm Hg, whereas cyclophosphamide pulse therapy is repeated if the white cell count is >3000/m³ at 2 weeks after the first cyclophosphamide treatment. Finally, normal-inspired oxygen therapy (FiO₂ 21%) is used for all patients throughout hospitalization.

Clinical diagnosis

Immediate treatment requires immediate diagnosis. Diagnosis of PQ poisoning is based on clinical history, physical examination, and laboratory examination and is confirmed by urine (sodium dithionite reaction) and blood (spectrophotometry, Hitachi, Tokyo, Japan) tests.^8,85 The sodium dithionite test is based on the reduction of PQ by sodium thionite under alkaline conditions to form stable, blue-colored radical ions. A strong navy or dark blue generally indicates significant PQ ingestion and often forebodes a poor prognosis.

POINT-OF-CARE TESTING IN THE EARLY DIAGNOSIS OF PARAQUAT INTOXICATION

The current clinical standard for PQ poisoning diagnosis is doubly burdened by slow methodology and clinical scarcity in more remote geographic regions. Detection based on symptoms depends heavily on physician experience, and a urine sodium dithionite assay lacks precision in quantifying PQ poisoning. Examining the serum/plasma PQ level offers greater precision but is very time-consuming.

In the past decade, various methods have been investigated and proposed to improve speed and reduce cost while providing great sensitivity and specificity. Many of these approaches rely on bulky instrumentation such as mass spectrometers and gas chromatographs. We too wish to focus on speed, cost, sensitivity, and specificity but are particularly interested in doing so while providing greater portability. Among the advances made and reported in the literature, Zhu et al.⁸⁶ have developed a portable Raman spectrometer for on-site PQ poisoning diagnosis, and Kuan and his colleagues⁸⁷ have proposed a fast and portable paper-based PQ diagnostic device. In a novel development, Chen et al.¹² applied machine learning, in a manner similar to a support vector machine (SVM), to PQ diagnosis. Furthermore, while enzyme-linked immunosorbent assay (ELISA) for PQ diagnosis typically requires specific equipment, a promising paper-based ELISA approach has been developed for on-site PQ diagnosis.⁸⁸ This article reviews the development and application of PQ poisoning diagnostics. The advantageous approaches described herein may offer promising alternative solutions to improving mortality among patients poisoned with PQ.

Surface-enhanced Raman spectroscopy (SERS)

Raman spectroscopy can provide fingerprint specificity as surface-enhanced Raman spectroscopy (SERS), an approach with single-molecule sensitivity [Fig. 1(a)].⁸⁹ SERS can provide fingerprint specificity and trace-level sensitivity, making it an immensely valuable and increasingly popular tool for detecting small quantity targets.⁹⁰ Moreover, SERS is a speedy detection tool capable of completing single trials within seconds.^86,91 For the past few years, SERS has been adopted in various areas, from forensic science^92,93 and art analysis^94,95 to geoscience,⁹⁶ biomedical applications,^97–99 and PQ detection.^100–103 By applying SERS, Gao et al.¹⁰⁰ and Tang et al.¹⁰¹ achieved a PQ limit of detection (LOD) in a water sample of below 2 × 10⁻⁹M and 0.1 mg/l, respectively. Fang et al.¹⁰³ were able to detect the PQ residue concentration on a fruit peel as low as 0.1 mg/l through SERS. Using a different SERS substrate, Dao et al.¹⁰² achieved an LOD of only 0.01 mg/l.

FIG. 1.

Paraquat detection by SERS. (a) The mechanism of SERS. Yellow is the SERS substrate. Blue circles are the sample. (b) SERS spectra of paraquat in water (a), plasma (b), and urine (c) (50 μg/L (10⁻⁶ g/L)) with 100 mM of NaCl. Reproduced with permission from Zhu et al., RSC Adv. 6(65), 59919–59926 (2016). Copyright 2016 Royal Society of Chemistry.⁸⁶ (c) SERS spectra of the urine (a) and plasma (b) of the paraquat poisoned patient and of saline solution. Reproduced with permission from Zhu et al., RSC Adv. 6(65), 59919–59926 (2016). Copyright 2016 Royal Society of Chemistry.⁸⁶

Even though SERS demonstrates extreme sensitivity to PQ in aqueous samples with LOD as low as 2 μg l⁻¹ in the case of 100 mM NaCl as an aggregation reagent,⁸⁶ the practical use of SERS as a PQ poisoning diagnostic tool will require overcoming several obstacles. First, its ponderous size hinders it from being a first-choice, on-site detection tool.⁹⁶ Second, severe interference from plentiful salt, protein, and other biological sample components containing nitrogen and sulfur blocks PQ SERS signal characteristics, discouraging clinical employment.^86,104 Applying the water sample PQ detection methodology and conditions to PQ detection in human plasma or urine encounters LOD higher than 50 μg l⁻¹.⁸⁶ Third, reproducibility of the SERS signal is highly dependent on the consistency and stability of the SERS substrate and the measurement environment [Figs. 1(b) and 1(c)].^86,91

Fortunately, solutions are proposed to address the above-listed problems. First of all, demand in SERS portability encourages the development of portable SERS. Portable SERS has been successfully used in geological research,⁹⁰ food safety tests,⁹⁹ and wide-ranging chemical and biological agent detection, such as methyl parathion detection and bacillus spore detection.^104–106 Using portable SERS and silver colloidal substrates, Cowcher et al.⁹¹ established an on-site Bacillus bacterial spore identification method with an LOD of 29.9 nM, which is far below the anthrax infection level. Furthermore, using newly developed metallic nanoparticles as SERS substrates diminishes the effect of biological matrix interference and enhances SERS signal reproducibility.¹⁰⁷ A silver film-over-nanosphere (AgFON) substrate, for example, enabled detection of the Ricin B chain in human blood with an LOD of 1 μg/ml.¹⁰⁸ Adopting portable SERS, Zhu et al.⁸⁶ proposed a rapid and sensitive on-site PQ detection method in a biological sample. Combining pinhole shell-isolated nanoparticle-enhanced Raman spectroscopy (SHINERS) with the addition of iodide lowers the LOD of PQ in human plasma or urine samples to 1 μg l⁻¹, which is far beneath the concentration in a PQ poisoning patient sample. Furthermore, every measurement using this approach takes less than 10 s, and the whole process can be completed within 1 min without any tedious sample pretreatment. Beyond that, the stability of using pinhole shell-isolated nanoparticles as SERS substrates expands the PQ SERS signal duration from less than 6 h in plasma and 5 min in urine (bare Au nanoparticle as a SERS substrate) to at least 12 h in plasma and 20 min in urine, which is crucial when it comes to practical application.⁸⁶ Even though this method lacks sufficient clinical trials, it is still promising.

Despite their many advantages, SERS substrates can be costly because they require microfabrication and even nanofabrication to attain the metallic nanostructure.¹⁰⁹ The production of less expensive and more easily manufactured SERS substrates is necessary for the development of a practical SERS-based PQ diagnostic tool. Even though no inexpensive SERS substrate has yet been implemented to target PQ diagnosis, PQ diagnostic tool development may benefit from research into inexpensive SERS substrates in general. By inkjet printing the sensitive SERS substrate onto paper, Yu and White¹⁰⁹ leveraged ease-of-manufacture and reduced the cost of each array to 2 cents without sacrificing SERS sensitivity. Because it is easy to make, paper-based SERS substrates can be readily produced, saving the cost of SERS substrate storage.¹⁰⁹ Detection of low concentration targets, such as Rhodamine 6 G and p-aminothiophenol, has been carried out using a paper-based SERS substrate.¹¹⁰ Also, a paper-based SERS substrate integrated with microfluidic channels has been used to determine the glucose level in blood samples.¹¹¹ Commercial tape is another available option for the production of inexpensive and easily manufactured SERS substrates. In one example of this, colloidal gold nanoparticles were applied to the surface of commercial tape, and a simple “paste and peel off” process allowed the tape-based SERS tool to quickly and inexpensively detect pesticide residues on fruits and vegetables.¹¹² Considering the availability of inexpensive SERS substrate alternatives, suitable and feasible portable SERS PQ diagnostic tools are imminent [Fig. 3(a)].

FIG. 3.

Proposed on-site paraquat detection system based on SERS, ELISA, and machine learning. (a) The combination of the cheap SERS substrate, such as inkjet-printed silver nanoparticles on chromatography paper,¹⁰⁹ and the portable SERS device creates a possibility for a portable paraquat detection system. Reproduced with permission from Yu and White, Anal. Chem. 82(23), 9626–9630 (2010). Copyright 2010 American Chemical Society. (b) Incorporating paper-based ELISA⁸⁸ and mobile ELISA reader¹²⁰ yields a feasible scheme for a portable paraquat detection system. Reproduced with permission from Cheng et al., Angew. Chem. Int. Ed. Engl. 49(28), 4771–4774 (2010). Copyright 2010 Wiley VCH Verlag GmbH & Co, KGaA and Berg et al., ACS Nano 9(8), 7857–7866 (2015). Copyright 2015 American Chemical Society. (c) Machine learning approach in mobile phones that can diagnose paraquat poisoning from the blood routine test can serve as a point-of-care paraquat detection device.

Paper-based diagnostic device

The great capacity for using paper as an SERS substrate for PQ poisoning diagnosis lends itself to colorimetric PQ detection methods.⁸⁷ Due to its low-price, prevalence, portability, rapidness, compatibility with current technologies, high surface area to volume ratio and suitability for commercial mass production, paper has been applied to various types of biologically relevant applications.^113–115 Many paper-based analytical devices using photoelectrochemical or colorimetric assay platforms are proposed to quantify pesticide residue in food or water. Nevertheless, while most of them focus on organophosphorus compound or carbamate compound detection, few targets PQ detection.^116–118 Among these few efforts targeting PQ detection, even fewer efforts deal with PQ detection in human samples. In 2016, Kuan et al.⁸⁷ introduced a colorimetric paper-based analytical device for PQ detection in human serum. They fabricated 96-well plates as the format for their paper-based analytical PQ detection device by wax printing technology and coated it with reagents, including sodium dithionate, ascorbic acid, and sodium hydroxide. Since the blue radical ion is present when sodium dithionite or ascorbic acid is in alkaline solutions, the existence of PQ can be observed through visible color changes, which means that there is no need for any other external bulky and expensive instrument to read the result [Fig. 2(a)]. Moreover, conventionally, time required to report PQ poisoning by using spectrophotometry to detect PQ is 24 h. This device curtails the PQ detection time to only 10 min, granting physicians performing medical treatment as soon as possible and maybe increasing the survival rate. The LODs of this device for sodium dithionite and ascorbic acid assays in normal human serum are 13.80 and 3.86 ppm, respectively, which are relatively not ideal compared to the traditional method [Fig. 2(b)]. However, being a prototype, this colorimetric paper-based analytical device for PQ detection has a lot of room for optimization and improvement. Hence, despite the LODs, this device remains a promising PQ poisoning option.

FIG. 2.

A colorimetric paper-based analytical device for paraquat detection. (a) The mechanism of the colorimetric paper-based paraquat detection device. Detection regents were applied to each paper well using a pipet. The detection point turns blue if paraquat is present. Reproduced with permission from Kuan et al., Biomicrofluidics 10(3), 034118 (2016). Copyright 2016 AIP Publishing.⁸⁷ (b) Different detection regent, a sodium dithionite assay (left) and an ascorbic acid assay (right), using normal human serum. Calibration curve created at several concentrations—0, 5, 10, 25, and 50 ppm. (N = 10; mean intensity ± S.D.) Reproduced with permission from Kuan et al., Biomicrofluidics 10(3), 034118 (2016). Copyright 2016 AIP Publishing.⁸⁷

Even though Enzyme-Linked Immunosorbent Assay (ELISA) for PQ detection¹¹⁹ has not yet been implemented on the paper, from our perspective, a portable and cheap paper-based ELISA PQ diagnosis tool seems to be possible with the advent of paper-based ELISA⁸⁸ and devices.¹²⁰ ELISA for PQ detection dates back to the 1980s.^121,122 Using ELISA, Tomita et al.¹²² described a method that can detect PQ in patients with the PQ concentration ranging from 1.56–100 ng/ml in 1988. With improvement and optimization, nowadays, the ELISA method can detect the PQ concentration in the human urine sample as low as 2 ng/ml⁻¹.¹¹⁹ Nevertheless, conventional ELISA takes more than an hour for incubation and blocking steps. Besides, to quantify the result, a plate reader typically costs $20 000 which is too expensive for resource limited areas where PQ poisoning occurs most frequently.⁸⁸ Paper-based ELISA (P-ELISA) integrates the sensitivity and specificity of ELISA with the portability, low price, and ease-of-use of paper-based platforms being highly desirable, especially for point-of-care and remote region diagnosis.⁸⁸ Paper-based ELISA (P-ELISA) shortens the normally lengthy ELISA procedure to less than an hour, and the results can be measured using an inexpensive desktop scanner. The sensitivity of P-ELISA has been improved, and P-ELISA has been widely used for decades, despite initial sensitivity concerns. Examples of P-ELISA include the research of Liu et al.¹²³ who detected hepatitis B surface antigenin serum by P-ELISA and Shih et al.¹²⁴ who identified Escherichia coli by P-ELISA. In addition to P-ELISA helping improving portability and cost of PQ diagnosis, introduction of a cellphone-based hand-held microplate reader eliminates the burdens of cumbersome and expensive instrumentation.¹²⁰

Hand-held microplate readers use a 3D-printed opto-mechanical attachment and light-emitting-diode (LED) array to hold and illuminate a 96-well plate. The LED light transmitted through each well is gathered by optical fibers, and the signal is translated using a custom-designed application that provides results within 1 min. The accuracy of this cartridge is greater than 90%.¹²⁰ Although this device is not currently designed for reading P-ELISA, it offers a point to develop one. Combining the advantages of P-ELISA with a portable ELISA reader moves us a step closer to developing an inexpensive, fast, viable, and portable point-of-care PQ poisoning diagnosis tool [Fig. 3(b)].

Machine learning diagnosis

In addition to developing physical PQ poisoning diagnostic tools, there are some researchers working on machine learning approaches. Because of the progress in computing power and the availability of commercial software packages, such as MATLAB toolboxes, machine learning is garnering increasing interest in several research fields including disease diagnosis.¹²⁵ The problem point in disease diagnosis is a classifying problem, i.e., classifying if a patient tests positive for a particular disease. Hence, analysis using a support vector machine (SVM), which can perform efficient non-linear classification, improves accuracy among machine learning algorithms built to analyze biomedical signals.^126,127 Some SVM-based disease diagnoses have achieved classification accuracies beyond 95%.^128–133 SVM has been proposed to diagnose breast cancer,^128–131 hepatitis,¹³² thyroid disease,¹³³ Parkinson's disease,¹³⁴ diabetes mellitus,¹³⁵ schizophrenia,¹³⁶ bipolar disorder,¹³⁶ and PQ poisoning.^12,137,138

Instead of performing diagnosis based on analyzing patient plasma samples using a gas chromatography-mass spectrometer,¹³⁷ an SVM for PQ poisoning diagnosis, developed by Chen et al.,¹² uses information obtained from a blood routine test (BRT), one of the most common tests in a hospital, to diagnose PQ poisoning. This blood routine test, also known as the complete blood count, covers indexes like white blood cell, red blood cell, and blood platelet counts. The correlation between PQ poisoning and blood routine test indexes was made clear by using a PQ poisoning diagnosis SVM on 79 patients and healthy controls. Two blood routine test indexes, white blood cell and absolute value of neutrophilic granulocyte, were identified as the two features most correlated with PQ poisoning. However, this SVM only had an accuracy of 70.5%, a sensitivity of 57.3%, and a specificity of 82.9% (Table II) The accuracy, sensitivity, and specificity of this SVM for PQ poisoning diagnosis may be improved through various feature selection techniques, much like what has been accomplished for SVM-based breast cancer diagnosis.^128–131 Furthermore, SVM can be implemented as a mobile phone application. This SVM for PQ poisoning diagnosis and the blood routine test are promising PQ diagnostic tools [Fig. 3(c)].

TABLE II.

Average classification performance results of the support vector machine for paraquat diagnosis. BRT, blood routine test; PQC, paraquat concentration.

Data sources	Accuracy (%)	Sensitivity (%)	Specificity (%)
Data with PQC index	82.7 ± 13.8	66.7 ± 18.1	94.6 ± 26.1
Data with BRT index	70.5 ± 22.3	57.3 ± 40.1	82.9 ± 30.1
Data with PQC + BRT index	71.3 ± 23.8	63.7 ± 50.2	78.7 ± 44.5

Although SVM for PQ poisoning diagnosis based on the blood routine test index is inexpensive, rapid, and simple, it is burdened with some potential shortcomings. The “black box” character of the SVM model may lead to false discovery.¹²⁵ For example, an SVM model that produced highly accurate results in one dataset is not guaranteed to perform as accurately in other datasets. Also, the choice and number of attributes may cause overfitting, especially for small studies.¹²⁵ Hence, in addition to optimizing its performance, this SVM for PQ poisoning diagnosis requires additional clinical trials and validations.

POINT-OF-CARE TESTING IN THE EARLY DIAGNOSIS OF OTHER PESTICIDE INTOXICATION

Intentional intoxication with organophosphate insecticide is another serious public health problem in developing countries.¹³⁹ Previous analysis of medical records⁴ revealed that organophosphate accounted for most common type (76.8%) of acute pesticide exposure at Chang Gung Memorial Hospital.

Chronologically, the three clinical syndromes after acute organophosphate intoxication include (1) acute cholinergic crisis due to acetylcholinesterase suppression, (2) intermediate syndrome (0.5–7 days) that has an unclear underlying mechanism, and (3) delayed polyneuropathy (6–21 days) explained by the inhibition of neuropathy target esterase. Acute cholinergic crisis¹⁴⁰ comprises signs and symptoms resulting from hyperstimulation of muscarinic receptors in the parasympathetic system (e.g., bradycardia, bronchospasm, bronchorrhea, hypotension, diarrhea, vomiting, miosis, lachrymation, salivation, and urination), nicotinic receptors in the sympathetic system (e.g., hypertension, tachycardia, mydriasis, and sweating), nicotinic receptors at the neuromuscular junction (e.g., muscle weakness, paralysis, and fasciculations), and both central muscarinic and nicotinic receptors in the central nervous system (e.g., confusion, agitation, coma, and respiratory failure). Our study¹³⁹ indicated that all patients suffered from acute cholinergic crisis after intentional ingestion of organophosphate pesticide and the mortality rate was 15.2%.

Clinical diagnosis of organophosphate intoxication depends on (1) clinical symptoms and history of organophosphate pesticide exposure, (2) reaction of clinical symptoms to the anticholinergic agent, and (3) suppression of the level of acetylcholinesterase in blood.¹⁴¹ Organophosphate/carbamate pesticide strongly binds with acetylcholinesterase family compounds in human blood. A decrement in blood cholinesterase activity can be used as a biomarker to diagnose organophosphate/carbamate intoxication and to monitor response to anticholinergic medication.¹⁴² Nevertheless, there are several dozens of these organophosphate/carbamate compounds currently registered for use by the Environmental Protection Agency in the United States. Therefore, the clinical diagnosis of organophosphate/carbamate depends on the demonstration of depression of blood acetylcholinesterase activity but not counts on the measurement of individual organophosphate/carbamate concentrations in blood.

In a study,¹¹⁸ we have developed a paper-based, metabolic assay designed for the rapid, semi-quantitative measurement of organophosphate intoxication. The paper-based platforms, including point-of-care devices and 96-well plates, provided semi-quantitative information regarding the concentration of acetylcholinesterase (not individual organophosphate/carbamate compounds). In this analysis, acetylcholinesterase catalyzes the hydrolysis of acetylcholine into choline and acetic acid. Choline is then oxidized by choline oxidase to betaine and hydrogen peroxide, and 3,3′,5,5′-tetramethylbenzidine reacts with hydrogen peroxide to develop an intense blue color that can be recorded as an output signal.¹¹⁸ The paper-based 96-well-plate was used to measure the level of organophosphate poisoning in 3 patients. The results were comparable to those obtained using the hospital conventional enzymatic method currently considering the gold standard.¹¹⁸

In a study, Kim et al.¹⁴³ examined the relationship between blood lactate levels and mortality from glyphosate surfactant intoxication. Of the 232 patients, lactate was significantly higher in non-survivors than in survivors (6.5 ± 3.1 mmol/l versus 3.3 ± 2.2 mmol/l, p < 0.001), and elevated lactate was significantly associated with 30-day mortality. Notably, instead of measuring the blood glyphosate surfactant level, the lactate levels were measured using a GEM Premier 3000^® blood gas analyzer as the point-of-care testing (IL Headquarters, Bedford, MA, USA) immediately after arrival at the emergency department.¹⁴³

Few data are available in the literature on the clinical application of point-of-care testing after acute pesticide intoxication although a point-of-care testing for detection of blood protein adducts resulting from exposure to organophosphate nerve agents has been described.¹⁴⁴

SUMMARY AND PERSPECTIVES

Pesticide intoxication remains a major contributor to the global burden of suicide.¹⁴⁵ Despite intensive efforts, PQ poisoning is common and produces high mortality in developing countries. While mild poisoning patients may fully recover, severe poisoning patients may face permanent injure or even death.

Proper instantaneous treatment, such as anti-inflammatory treatments, to reduce mortality depends heavily on immediate diagnosis. Current clinical PQ diagnosis methods are slow, and clinical facilities are scarce. Various point-of-care PQ poisoning diagnostics have been proposed to address these problems. Retaining SERS's sensitivity and speed, portable SERS and the use of novel, promising, portable, and inexpensive SERS substrates eliminate the issues of bulkiness and cost associated with conventional SERS. Moreover, paper-based PQ poisoning diagnostic devices, such as those mentioned in the work of Kuan et al.,⁸⁷ greatly reduce cost and diagnosis time. Additionally, machine learning algorithms are available that may be leveraged to diagnose PQ poisoning from a common blood routine test.

An ideal point-of-care PQ poisoning diagnosis method should be sensitive, fast, inexpensive, and portable. Even though none of the above methods fits all these features, the incorporation of different methods and the development of new technology may take us one step closer to achieving more ideal point-of-care diagnosis methods. Furthermore, in order to assess the potential medical and economical benefit and cost of introducing these PQ intoxication methods, a careful cost-benefit analysis should be performed.¹⁴⁶

In addition to PQ diagnosis, such point-of-care approaches can be used for rapid identification of ingested pesticides and theoretically can expedite rescue at the emergency department.¹⁴⁶ In summary, the point-of-care approaches have the potential to play a main role in revolutionizing the diagnosis, initiation, and monitoring of treatment of pesticide intoxication.