Rapid quantification of artemisinin derivatives in antimalarial drugs with dipstick immunoassays

Abstract

Substandard antimalarial drugs will result in unsatisfied therapeutic efficacy and increase the risk of resistance development. The point-of-care, qualitative, or semi-quantitative dipstick immunoassays cannot differentiate the substandard drugs with confidence. A rapid and quantitative analytical method that can be used under field conditions is needed. Here, three lateral flow immunoassays (LFIAs) based on colloidal gold nanobeads (CGN) as labels were developed for quantification of artemether, dihydroartemisinin and artesunate contents in antimalarial drugs with the aid of a portable optical scanner. Also, time-resolved fluorescent nanobeads (TRFN)-LFIA, coupled with a portable fluorescent lateral flow reader, was developed for quantification of artesunate. Commercial antimalarial drugs were used to validate these LFIAs with comparison to the gold standard high-performance liquid chromatography (HPLC) method. The drug contents estimated with these CGN-LFIAs were in the range of 85.5-109.3% of the contents determined by HPLC with a coefficient of variation (CV) of 4.5-13.0%. The TRFN-LFIA results were in the range of 93.7-108.4% of contents determined by HPLC with a CV of 5.2-8.9%. There were no significant differences between the results of CGN-LFIA and TRFN-LIFA (P = 0.5277, t-test). Both types of LFIAs with portable readers may be used for quantitation of active ingredients in antimalarial drugs and for screening substandard antimalarial drugs in resource-limiting settings.

1. Introduction

The World Health Organization (WHO) estimates that there were 228 million malaria cases in 2018 and most of the cases occurred in Africa [1]. The circulation of substandard antimalarial drugs in the market undermines the efficacy of the treatment and may be one of the contributing factors to resistance development [2, 3]. Almost two decades ago, fake artesunate tablets were widespread in some Southeast Asian countries [2]. More recent surveys in Africa indicated that falsified and substandard antimalarial drugs were also a serious problem. For example, a study conducted on Bioko Island, Equatorial Guinea showed that of the 677 artemisinin-based combination therapy (ACT) drugs collected from 278 outlets, 1.6% were substandard and 7.4% falsified [4]. In Ghana, about 35% of 256 ACTs purchased from known medicine outlets were substandard [5]. In southern Malawi, 7/56 antimalarial drugs were found to be substandard [6]. The situation of falsified and substandard drugs is aggravated by the lack of effective supervision methods in worst-hit areas of malaria, forming a vicious cycle. It is difficult for clinics in these areas to screen substandard drugs due to the lack of simple and effective test methods. Instrumental methods such as high-performance liquid chromatography (HPLC) [7], gas chromatography-mass spectrometry (GC-MS) [8], and LC-MS [9] are often used for quantification of active ingredients in antimalarials. However, these methods are time-consuming and require expensive instrument and professional personnel. Enzyme-linked immunoassay (ELISA) is much less expensive and relatively easy to perform, and we have developed ELISAs to quantify artemether, dihydroartemisinin and artesunate [10–15].

The lateral flow immunoassay (LFIA), based on colloidal gold nanobeads (CGN), has been widely used for testing toxic compounds and diagnosis of infectious diseases at the points of care such as the physician’s examination room and emergency wards in hospitals [16]. CGN-LFIAs or dipsticks have some great advantages such as their user-friendly format, short detection time, and convenient storage conditions. CGN-LFIAs of artemether, dihydroartemisinin, and artesunate have been developed for qualitatively screening falsified drugs and semi-quantitatively detecting substandard drugs. The semi-quantitative method was achieved by serially diluting the samples to the indicator ranges of the dipsticks [11, 12, 15], which provides a rough estimate of the quantity of the testing compound. It has been shown that CGN-LFIAs with the aid of an optical density scanner to read the gray values of the T line and C line can provide quantification of the testing substances with satisfactory accuracies. A CGN-LFIA has been developed for the simultaneous quantification of clenbuterol and ractopamine with recoveries of 94-106% [17]. Another CGN-LFIA for simultaneous and quantitative detection of sulfamethazine, sulfadiazine and sulfaquinoxaline had recoveries ranging from 75% to 82% for egg samples and from 78% to 81% for chicken samples [18].

Some researchers have reported that conjugating antibodies to a fluorescent probe could increase the sensitivity of LFIAs [19, 20]. Compared with other fluorescent probes such as quantum dots and fluorescent nanobeads, time-resolved fluorescence (TRFN) using lanthanide chelate as labels has long fluorescence decay time and an exceptionally large Stokes’ shift, thus avoiding interference from background fluorescence and scattered light [21–23]. A TRFN-LFIA has been developed for the quantitative detection of aflatoxin B1 with recoveries ranging from 87.2 to 114.3% [24]. Similarly, a TRFN-LFIA for quantitatively detecting chloramphenicol also showed satisfactory recoveries from chicken meat and shrimps [25]. A TRFN-LFIA has been developed for the simultaneous quantification of chlortetracycline and doxycycline in edible animal tissues with recoveries of 85.8-102.4% and 85.3-101.6%, respectively [26]. In many studies, the reported sensitivities of TRFN-LFIA are much higher than those of CGN-LFIAs [19, 20]. For drug quality control, the need of accuracy is higher than sensitivity due to the milligram grade content in the drug samples. However, few studies investigated which probe of CGN and TRFN performs more accurately.

In the present work, we developed three CGN-LFIAs for quantification of artemether, dihydroartemisinin, and artesunate in ACTs. Meanwhile, we also developed a TRFN-LFIA for artesunate to determine whether it offers higher accuracy than the CGN-LFIA.

2. Materials and methods

2.1. Reagents

Carboxylate-modified TRFN, 1% solids, 200 nm, 580/605 (λ_ex/λ_em), and CGN, 30 nm, were purchased from Shanghai Jieyi Biotechnology. Three haptens of artemether, dihydroartemisinin, and artesunate conjugated with bovine serum albumin (BSA), three monoclonal antibodies (mAbs) against artemether, dihydroartemisinin, and artesunate were produced in our laboratory [11, 12, 15].

1-Ethyl-3-(3-(dimethylamino) propyl) carbodiimide hydrochloride (EDC) and goat anti-mouse IgG were purchased from Sigma (St. Louis, MO, USA). Artemether, dihydroartemisinin and artesunate standards were purchased from the National Institute for the Control of Pharmaceutical and Biological Products (Beijing, China). The nitrocellulose filter membrane (NC membrane), sample pad, conjugate pad, absorbance pad, support plate, the portable optical scanner, and the fluorescence lateral flow reader were obtained from Shanghai Jinbiao Biotechnology Co., Ltd. (Shanghai, China). The other chemicals and organic solvents used were of analytical grade and purchased from Sinopharm Chemical Reagent (Beijing, China). The commercial antimalarial drugs used for detection were collected from Africa and Southeast Asia.

2.2. Preparation of the CGN-mAb probe

To label the mAb against artesunate with CGN, the pH of CGN solution was adjusted to 7.5 using 0.1 M potassium borate solution (pH 8.3). Then 40 μL of mAb against artesunate (1 mg/mL, dissolved in water) were added to 1 mL of CGN solution, and the mixture was gently stirred for labeling. Antibody binding to CGN is the result of hydrophobic interaction, coordinate bond formation, and electrovalent bond formation (Fig. 1B–i). After 10 min, 50 μL of 10% BSA were added and stirred to block excess sites for another 10 min. The mixture was centrifuged at 9,600 rpm for 20 min, and the resulting precipitate was resuspended in 600 μL of 0.01 M phosphate buffer (PB, pH 7.5). Finally, the resuspended solution was released in the conjugate pad to saturation and dried for 3 h. All steps were operated at room temperature.

Fig. 1.

A. Assembly diagram of LFIA. B. Schematic synthesis of the CGN-mAb probe (i) and the TRFN-mAb probe (ii). C. Portable optical scanner reading the gray values of the T line and C line in a CGN-LFIA. D. Portable lateral flow fluorescence reader reading the fluorescent intensity of T line and C line in a TRFN-LFIA.

The procedures of labeling mAb against artemether and dihydroartemisinin with CGN were the same as above, the parameters were as follows: 10 μL of mAb against artemether (1 mg/mL, dissolved in water) were labeled with 1 mL of CGN solution, and finally resuspended in 200 μL of PB; 20 μL of mAb against dihydroartemisinin (1 mg/mL, dissolved in water) were labeled with 1 mL of CGN solution, and finally resuspended in 250 μL of PB.

2.3. Preparation of the TRFN-mAb probe

One hundred μL of TRFN was suspended in 400 μL of 0.05 M boric acid buffer (BB, pH 8) and then 20 μL of EDC (10 mg/mL, dissolved in BB) was added and reacted on a shaking table for 20 min. The TRFN and mAb were conjugated via chemical bonds (Fig. 1B–ii). After centrifugation at 10,000 rpm for 10 min, the precipitate was suspended in BB. Next, 100 μL of mAb (1 mg/mL, dissolved in water) were added for labeling with shaking for 2 h. Then, 55 μL of 10% BSA (dissolved in BB) was added to block excess sites for 1 h. The mixture was centrifuged and resuspended in BB twice and then dispersed in 9 mL of BB. Finally, the dispersed solution was released in TRFN conjugate pad to saturation and dried for 3 h. All steps were operated at room temperature and sonication was applied at all steps to ensure that the TRFN suspension was uniform.

2.4. Assembly of LFIAs

As illustrated in Fig. 1A, the CGN-LFIA dipstick consists of five parts: the support plate at the bottom, the NC membrane, the absorption pad, the sample pad, and the conjugate pad. These parts were assembled with two adjacent parts overlapping about 2 mm. The test line (T line) included artesunate-BSA, artemether-BSA, or dihydroartemisinin-BSA, all at a concentration of 1 mg/mL (dissolved in water), while the control line (C line) included goat anti-mouse IgG (1 mg/mL, dissolved in water). Both lines were formed by spraying the solutions onto the NC membrane at the volume of 1 μL/cm using a dispenser platform at an interval of 5 mm. The plate was cut to 3 mm width after the lines dried. Finally, the dipsticks were stored in plastic containers with desiccant at 4 °C.

The procedure of assembly for the TRFN-LFIA was the same as for CGN-LFIA and the width of dipsticks was 4 mm.

2.5. Sample preparation

Commercial drugs (10 samples for artesunate, artemether, and dihydroartemisinin each) collected from Africa and Southeast Asia were used to validate the assays. For drug extraction, a drug tablet was put into a 10 mL tube and pounded to powder, to which 5 mL of acetonitrile were added. After sonicating for 20 min, the mixture was shaken for 4 min on a vortex shaker. After standing for 2 h, it was centrifuged at 4,000 rpm for 20 min, and the supernatant was collected and diluted with acetonitrile to 2 mg/mL based on the drug content on the label. Finally, it was filtered through a 0.22 μm Millipore filter as the stock drug solution for HPLC. The concentrations of stock solutions are shown in Table 1 and Table 2.

Table 1.

Comparison of the analysis results for artemether contents by CGN-LFIA and reference HPLC.

Commercial names	SM/DCa	Lot No.	Region b	Expected results (mg/mL)	CGN-LFIA results (mg/mL)	HPLC results (mg/mL)
ARTRINE	LIC Pharma	908	Cot d’lovire	4	3.86±0.20	3.99±0.01
SUPA ARTE	AA Medical Products Ltd.	408/410	Myanmar	4	3.50±0.17	3.91±0.03
ARTE+	Poly Gold Co., Ltd.	310, R-705	Myanmar	4	3.28±0.25	3.59±0.07
PAI COTOKIN	PAI Brothers’ Int’l Trading Co., Ltd.	1903AA7314	Myanmar	4	3.23±0.42	3.05±0.01
ARTEMETHER TABLERTS	Kunming Pharmaceutical Corp	unknown	Myanmar	10	9.24±0.52	9.85±0.21
Artemether Capsules	KPC Group	120224	Myanmar	8	6.12±0.36	6.50±0.01
Coartem	Beijing Novartis Pharma Ltd.	04-3275	Myanmar	4	3.51±0.31	3.4±0.02
Artefan20/120	Ajanta pharma Ltd.	P0251C	Myanmar	4	3.09±0.35	3.44±0.01
Artemether Injection	Kunming Phannaceutical Corp	GYZZ H10900011	Myanmar	8	5.12±0.41	5.76±0.07
MALARTE-AL SS	Global Pharma Healthcare Pvt Ltd., Chennai -India	1612AA3690	Myanmar	4	3.29±0.36	3.70±0.06

Each sample was assayed in triplicate by each method, and the results are shown as mean ± standard deviation.

^aSM/DC – Stated Manufacturer/Distributing Company.

^bAll samples were collected in different regions of Myanmar and Africa.

^cAt the time of the assay, all drags collected from Myamnar have passed the expiration dates.

Table 2.

Comparison of the analysis results for dihydroartemisinin contents by CGN-LFIA and reference HPLC.

Commercial names	SM/DC	Lot NO.	Region a	Expected results (mg/mL)	CGN-LFIA results (mg/mL)	HPLC results (mg/mL)
Malacur	Ciron Drugs & Pharmaceuticals Pvt. Ltd.	9E13120	Cot d’lovire	8	7.71±0.45	7.9±0.03
- DUO-CONTE CXIN	Zhejiang Holley Nanhu Pharmaceutical Co., Ltd., China	141117	Myanmar	8	7.61±0.38	7.04±0.07
		140329	Myanmar	8	6.55±0.35	6.96±0.02
		130635	Myanmar	8	6.13±0.71	5.61±0.01
		unknown	Myanmar	8	6.50±0.58	7.08±0.05
D-ARTEPP	Guilin Pharmaceutical Co., Ltd., China	SQ140102	Myanmar	8	6.81±0.45	6.40±0.03
		SQ130801	Myanmar	8	6.38±0.29	6.80±0.02
		SQ130401	Myanmar	8	6.01±0.46	6.59±0.01
DARPLEX	Kunming Pharmaceutical Co., Ltd., China	12 EA	Myanmar	8	5.55±0.36	6.12±0.02
SHUANG QING QING HAO SU PAI KUI PIAN	Chongqing Hualiyankangzhiyao Co., Ltd., China	130421	Myanmar	8	7.15±0.49	7.61±0.03

Each sample was assayed in triplicate by each method and the results are shown as mean ± standard deviation.

^aAll samples were collected in different regions of Myanmar or Africa.

^bAt the time of the assay, all drugs collected from Myanmar have passed the expiration dates.

Given that the purpose of the quantification method is for drug testing under field conditions, we further evaluated whether ethanol could be used to replace acetonitrile for drug extraction. For the above procedure, the three artemisinin drugs were mixed with 100%, 95%, or 75% ethanol, and the mixtures were shaken for 30 min. The contents of the artemisinin derivatives were determined by HPLC and their recovery rates were compared with that from acetonitrile.

2.6. Detection of the test compounds with LFIAs

For the CGN-LFIAs, the dipsticks were fixed in the plastic housings and incubated at 30 °C. Fifty μL of the serial dilutions of the drugs (artesunate: 0, 62.5, 125, 250, 500, 1000 and 2000 ng/mL; artemether: 0, 4.69, 9.38, 18.75, 37.5 and 75 ng/mL; and dihydroartemisinin: 0, 12.5, 25, 50, 100 and 200 ng/mL) were added dropwise into the sample pads of the respective CGN-LFIAs in triplicate. The dipsticks were incubated for 15 min and read by using a portable optical scanner to obtain the gray values of the T and C lines (Fig. 1C). A standard sigmoidal curve was plotted according to the values of T/C, and the concentration of the drug sample was calculated.

The TRFN-LFIA procedure was the same as for CGN-LFIA using the artesunate solutions at 0, 125, 250, 500, 1000, and 2000 ng/mL. To each sample pad, 100 μL of the artesunate solution were added and the assay was done in triplicate. After 15 min, the dipsticks were observed with naked eyes under a UV light and then read by a portable fluorescent lateral flow reader (Fig. 1D). The fluorescent values of the T and C lines were used to construct the standard sigmoidal curves.

2.7. HPLC analysis

The artesunate, artemether and dihydroartemisinin standard samples were prepared at gradient concentrations of 0.3125, 0.625, 1.25, 2.5, and 5 mg/mL. The HPLC system consists of a Waters 600E multi-solvent delivery system and a Waters 2487 dual λ absorbance detector (Milford, MA, USA). The working conditions of artesunate and dihydroartemisinin were as follows [12, 15]. A C18 reverse-phase column (250×64.6 mm, 5 mm particle size, Thermo, Vantaa, Finland) was adopted. The mobile phase consisted of acetonitrile-ultrapure water (60:40, v/v), the flow rate was 1 mL/min, the injection volume was 10 μL and the wavelength of the detector was 216 nm. The retention time for artesunate was 9.5 min, and for dihydroartemisinin was 12 min and 17 min for the two enantiomers. The working condition of artemether was as follows [11]. A C18 reverse-phase column was used as above, the mobile phase consisted of acetonitrile-0.5% acetic acid (70:30, v/v), the flow rate was 1 mL/min, the injection volume was 10 μL and the wavelength of the detector was 210 nm. The retention time for artemether was 17.7 min.

2.8. Statistical analysis

The correlation between LFIA and HPLC results was analyzed using the Bland-Altman test in MedCalc software. It takes the average value of two methods as X axis, the difference value as Y axis, the value of bias adding and subtracting 1.96 times the standard deviation (STD) as the limits of agreement, with one drug plotted as one point [27]. The comparison between CGN-LFIA and TRFN-LFIA results was conducted by the t-test using the MedCalc software for independent samples [28]. Statistical significance was considered at P < 0.05.

3. Results and discussion

3.1. Establishment of standard curves using the LFIAs and portable readers

Using serial dilutions of the standard drugs, we tested the performance of the CGN-LFIAs and TFRN-LFIAs (Fig. 2). The LFIAs were developed based on the principle of competitive immunoassay. The more targets in the sample, the less labeled mAb probes bound to the antigens coated on the T line. Thus, for each drug tested, the color intensity of the T line in the CGN-LFIAs decreased as the concentration of the drug increased (Fig. 2A–C). For the TRFN-LFIAs, the fluorescence intensity of the T line decreased as the concentration of the drug increased (Fig. 2D). The color or fluorescence intensities of the C and T lines in the CGN-LFIAs and TRFN-LFIAs were read with a portable optical or fluorescence scanner/reader, respectively, and the intensity values were used to make standard curves.

Fig. 2.

A-C. Images of CGN-LFIAs for artemether (A), dihydroartemisinin (B) and artesunate (C) showing the C line (upper) and the T line (lower) tested with different concentrations of the respective compounds. D. Images of artesunate TRFN-LFIA tested with different concentrations of artesunate, and viewed with the aid of a UV light.

The standard sigmoidal curves of artemether, dihydroartemisinin and artesunate CGN-LFIAs based on the T/C ratios of the gray values of T and C lines were fitted into four-parameter logarithmic equations (R²=0.997, 0.999, 0.999, respectively). Based on the standard curves, the working range of artemether, i.e. the 20-80% inhibitory concentration (IC₂₀ – IC₅₀), was 5.55–60.29 ng/mL, and the 50% inhibitory concentration (IC₅₀) was 20.07 ng/mL (Fig. 3A). The working range of dihydroartemisinin was 17.38–81.09 ng/mL with IC₅₀ of 34.90 ng/mL (Fig. 3B), and the working range of artesunate was 86.16–1964.42 ng/mL with IC₅₀ of 377.27 ng/mL (Fig. 3C).

Fig. 3.

Standard curves of artemether (A), dihydroartemisinin (B), and artesunate (C) detected by CGN-LFIA, as well as artesunate (D) detected by TRFN-LFIA. IC₂₀, IC₅₀ and IC₈₀ are indicated in the graphs. The working range is defined as IC₂₀ – IC₅₀.

Similarly, the standard curve of artesunate TRFN-LFIA was fitted into a four-parameter logarithmic equation (R²=0.997) (Fig. 3D). The working range was 116.27–1073.25 ng/mL, and the IC₅₀ value was 340.07 ng/mL, which was slightly lower than that of CGN-LFIA.

3.2. Validation of the CGN-LFIAs

We used commercial drugs containing artemether, dihydroartemisinin, and artesunate, 10 each, which were collected in the past, to determine the performance of the CGN-LFIAs for quantifying these drugs. For each of the drugs, we also used HPLC as the gold standard to determine the actual contents of the active indredients in these drugs. All assays were performed in triplicate and the results are presented in Tables 1–3. The artemether contents in the 10 drugs estimated using the artemether CGN-LFIA were 88.9–106.0% of the contents determined by HPLC with a coefficient of variation (CV) of 4.9–13.0% (Table 1). The contents in the 10 drugs estimated using the dihydroartemisinin CGN-LFIA were 90.7–109.3% of the values determined by HPLC, with a CV of 4.5–11.6%. For artesunate drugs, the estimated contents based on the artesunate CGN-LFIA were 85.5–108.6% of the contents determined by HPLC with a CV of 5.0–11.8% (Table 3). To determine the consistency of the dipsticks, three batches of the CGN-dipsticks of artemether, dihydroartemisinin, and artesunate were used to quantify the active ingredients of drug samples. The inter-assay CVs for artemether, dihydroartemisinin, and artesunate were 4.5-12.7%, 3.9-9.3%, and 6.0-13.4% respectively (Table S1–S3).

Table 3.

Comparison of the quantification results for artesunate contents by CGN-LFIA, TRFN-LFIA and reference HPLC.

Drug names	SM/DC	Lot NO.	Regiona	Expected concentration (mg/mL)	TRFN-LFIA results (mg/mL)	CGN-LFIA results (mg/mL)	HPLC results (mg/mL)
Artequine CHILD	Acino Pharma AG	1850138	Cot d’lovire	2	2.05±0.12	2.15±0.18	2.02±0.01
Artequine ADULT	Acino Pharma AG	1950190	Cot d’lovire	2	2.13±0.13	1.85±0.20	2.00±0.02
Artesunate	MEDIPLANTEX Liberty Group Trading Ltd.	210514	Myanmar	2	1.63±0.11	1.89±0.20	1.74±0.01
		216214	Myanmar	2	1.68±0.15	1.61±0.08	1.72±0.01
		216114	Myanmar	2	1.81 ±0.11	1.62±0.13	1.67±0.01
		264613	Myanmar	2	1.79±0.13	1.47±0.12	1.72±0.02
		264513	Myanmar	2	1.65±0.09	1.68±0.16	1.64±0.02
		264713	Myanmar	2	1.92±0.12	1.70±0.20	1.83±0.02
Traphosunat	M/S Sandar Myaing Co.	02/090412	Myanmar	2	1.65±0.09	1.77±0.14	1.65±0.02
Artesunate	Guilin Pharmoceutical Co., Ltd.	40502	Myanmar	2	1.61±0.11	1.64±0.18	1.70±0.04

Each sample was assayed in triplicate by each method and the results are shown as mean ± standard deviation.

^aAll samples were collected in different regions of Myanmar and Africa.

^bAt the time of the assay, all drugs collected from Myanmar have passed the expiration dates.

The CGN-LFIA and HPLC results were further compared using the Bland-Altman test (Fig. 4). For the artemether drugs (Table 1), the bias was −0.30, and all points were within two limits of agreement (±1.96 STDs) (−0.83, 0.24), indicating a good agreement between artemether CGN-LFIA and HPLC (Fig. 4A). Similarly, for the dihydroartemisinin drugs (Table 2), the Bland-Altman bias plot for the CGN-LFIA and HPLC showed a bias of −0.17 with all points being within ±1.96 STDs (−1.11, 0.77) (Fig. 4B). For artesunate drugs, the comparison also identified good agreement between the CGN-LFIA and HPLC methods (Fig. 4C).

Fig. 4.

Bland-Altman bias plots. A. Artemether CGN-LFIA and HPLC for ten artemether samples. B. Dihydroartemisinin CGN-LFIA and HPLC for ten dihydroartemisinin samples. C. Artesunate CGN-LFIA and HPLC for ten artesunate samples. D. Artesunate TRFN-LFIA and HPLC for ten artesunate samples. The x-axis shows the mean of the two methods, while the y-axis shows the difference of the two methods. Concentrations are expressed as mg/mL. The solid line represents the bias between the two methods, and the dashed lines represent the bias ±1.96 STD.

3.3. Validation and comparison of the artesunate TRFN-LFIA

The 10 artesunate drugs were further evaluated by using the newly developed artesunate TRFN-LFIA. Compared to the HPLC-determined drug contents, the measurement using TRFN-LFIA was 93.7–108.4% of HPLC with the CV of 5.2–8.9% (Table 3). Comparison by the Bland-Altman test showed that the bias of the two methods was 0.02, and all points were within 1.96 STDs (−0.15, 0.19), indicating good agreement between the two methods (Fig. 4D). We also tested three batches of the TRFN- dipsticks for quantifying the artesunate contents in the drug samples, and the inter-assay CVs ranged from 4.9 to 10.6%. Compared with the artesunate CGN-LFIA method, the artesunate TRFN-LFIA method showed a lower bias and STD, suggesting a higher correlation between TRFN-LEIA and HPLC than between CGN-LFIA and HPLC. However, there was no significant difference between the results of the two types of immunoassays (P = 0.5277, t-test). It is noteworthy that some studies reported a much higher sensitivity of the TRFN-LFIA method than the CGN-LFIA [19, 20], whereas others did not find remarkable superiority of the TRFN-LFIA method [29]. Such discrepancies may be due to many factors including the molecules to be detected and the mAbs used, which may vary in affinity and avidity. We also speculate that the slightly higher correlation between artesunate TRFN-LFIA and HPLC results may be due to the different principles of digital image processing for the two portable scanners. The TRFN-LFIA scanner was more sensitive to the changes in the fluorescence intensity than the CGN-LFIA scanner to the changes in the gray-scale image intensity. Of note, the price of the TRFN-LFIA scanner was nearly twice as much as the CGN-LFIA scanner.

3.4. Potential use of the quantitative LFIAs for detecting substandard drugs

Substandard drugs have different definitions, and may refer to less than 80, 85 or 90% of the labeled content [2, 4, 6]. If 80% of the labeled content was the evaluation criterion to define substandard, using the maximum detection deviation of 85.5% of all the LFIAs developed in this study, the drug content below 68.4% (80% multiplying by 85.5%) of the labeled content should be identified as substandard by our LFIAs. However, those drugs with contents falling between 68.4% and 80.0% would have the possibility to be overlooked.

3.5. Extraction of drugs with ethanol

While acetonitrile is traditionally used to extract artemisinin family drugs given their low solubility in water, acetonitrile is not environmentally safe and the extraction procedure is difficult to be performed in remote endemic sites. Thus, we tested ethanol as the solvent for extraction of artemisinin derivatives, given it is cheaper, safer and more environmentally friendly than acetonitrile. We used three concentrations (100%, 95% and 75%) of ethanol for extracting artemether, dihydroartemisinin and artemether and compared the drug recovery rates with that from acetonitrile extraction. As shown in Table S4, 100% ethanol had almost the same recovery rates as that with acetonitrile for all the drugs tested, whereas 95% ethanol had > 90% recovery rates for the drugs tested. In contrast, 75% ethanol only had 81-87% recovery rates for the three drugs. This result showed that under field conditions, pure ethanol can replace acetonitrile as the solvent for extraction of artemisinin family drugs.

4. Conclusions

Two types of LFIAs labeled with CGN and TRFN have been developed for the quantification of artemether, dihydroartemisinin and artesunate in antimalarial drugs with the aid of portable readers. All LFIAs performed well and the results were within 85.5–109.3% of results determined by HPLC. Further improvement of these dipsticks may lead to products deployable for quantification of artemisinin derivatives in ACTs and screening substandard antimalarial drugs in resource-poor areas.

Highlights.

• Four lateral flow immunoassays (LFIAs) were developed for quantifying artemisinins.

• Three LFIAs were based on colloidal gold nanobeads for a portable optical scanner.

• One LFIA was based on fluorescent nanobeads for a portable fluorescent reader.

• All LFIAs quantified contents of artemisinin drugs within 86-109% of HPLC values.

• Both types have potentials for on-site screening of substandard antimalarial drugs.