A novel strategy to precisely adjust partition coefficient, effectively eliminate matrix interference and achieve accurate and fast gold immunochromatographic assay of ethoxyquin in aquatic products

Abstract

Ethoxyquin (EQ), a common antioxidant in animal feed and aquatic products, poses health risks such as hepatorenal damage and teratogenicity. Conventional detection methods, while sensitive, are often complex and costly. This study introduces a novel pH-dependent extraction strategy to mitigate matrix interference from lipid co-extracts in Gold Immunochromatographic Assay (GICA). This study introduces a pH-dependent extraction strategy leveraging EQ's LogD range, coupled with Plackett-Burman and Box-Behnken experimental designs, to rapidly establish a simplified pretreatment protocol. The method achieved a 10 μg/kg detection limit—below international thresholds—by eliminating lipid matrix interference through pH adjustment. The method's effectiveness was confirmed through validation and comparison with the Chinese national standard method using real samples. This work not only advances the rapid analysis of EQ in food safety monitoring but also provides a versatile framework for designing simplified pretreatment protocols for detecting highly non-polar compounds.

Highlights

• •Developed a rapid GICA method for detecting EQ residues in aquatic products.
• •Discovered lipid coextracts as a new source of matrix interference in GICA detection.
• •Designed extraction and purification strategy by pH-dependent partition coefficients.
• •Quickly established a pretreatment method using Response Surface Methodology (RSM).
• •Achieved a detection limit of 10 μg/kg, and validated with multiple samples.

1. Introduction

Ethoxyquin (6-ethoxy-1,2-dihydro-2,2,4-trimethylquinoline, EQ) is a synthetic antioxidant widely used worldwide, primarily in preservation of animal feed and fruits(Choi et al., 2020; Negreira et al., 2017). However, some toxicological studies suggest that excessive EQ intake may have adverse health effects such as liver and kidney damage(Abou-Hadeed et al., 2021). In vitro studies have also shown that EQ can induce chromosomal aberrations in human lymphocytes(Błaszczyk et al., 2003). So the Joint Meeting on Pesticide Residues (JMPR) reassessed the acceptable daily intake (ADI) in 2008, establishing it at 0.005 mg/kg of body weight. In 2022, the European Commission suspended the use of ethoxyquin as a feed additive in all animal feeds (G/SPS/N/EU/190). Considering the frequently detected EQ residues in aquatic products and other animal-derived products(Merel et al., 2019; Negreira et al., 2017), many countries and organizations have forbidden its use or established maximum residue limits (MRLs) in various foodstuffs, which are usually in the range of 0.5 mg/kg ∼ 5.0 mg/kg(Food and Drug Administration, Department of Health and Human Services, 2024, National Standard of the People 's Republic of China, 2024, The Japan Food Chemical Research Foundation JFCRF, 2024).

Conventional detection methods for EQ residues are mainly based on High-Performance Liquid Chromatography (HPLC)(Berdikova Bohne et al., 2007; Rodríguez-Gómez et al., 2018), Liquid Chromatography-Tandem Mass Spectrometry (LC-MS/MS)(EURL-SRM, 2016; Negreira et al., 2017), and Gas Chromatography-Tandem Mass Spectrometry (GC–MS/MS)(Zhang et al., 2021). These techniques have distinct advantages in terms of sensitivity and accuracy, but unsuitable for on-site rapid screening of large numbers of sampless(Liu et al., 2021). Gold immunochromatographic assay (GICA) and other immunoassays have been widely developed as effective tools for simple and fast analysis of various food hazards, and numerous commercial kits have been successfully and widely applied in different fields(Huang et al., 2016). However, till now, these techniques have hardly been used for EQ. As far as we know, only one Chinese patent (CN104569398A) has used enzyme linked immunosorbent assay (ELISA) to detect EQ in tilapia and shrimp(Luo; et al. Luo et al., 2013). But according to our validation, its recovery rate is far from what was claimed, because the sensitivity of GICA is lower than that of ELISA, the recovery rate does not meet the requirements of GICA. Based on our preliminary work (see relative data in the text), the reason why conventional pretreatment methods cannot be applied to GICA detection may lie in the serious matrix interference from food stuffs, which can lead to significant false negative or false positive results and therefore greatly decrease the sensitivity and accuracy of the immunoassays. But unlike usually observed matrix effects induced by some proteins and irons(Anfossi et al., 2013; Xiaoxiao Wang et al., 2018), here the interference was mainly from nonpolar extracts of animal samples, indicating the significant influence of some lipid compounds, which was seldomly mentioned in previous studies. The unique chemical characteristics of EO makes it very difficult to solve the problem with conventional pretreatment strategies: considering EQ is of strongly nonpolar compound, traditional liquid-liquid extraction (LLE) and purification techniques based on the “like dissolves like” principle seemed very hard to effectively separate these lipid interfering substances which may have similar polarity with EQ(Sun et al., 2024). The partition coefficient of EQ (represented as the logP) is about 3.00 (data derived from MarvinSketch 15.6.29.0 software, calculated using the consensus method), demonstrated excellent solubility in usually exploited nonpolar solvents, so while increasing the polarity of the extraction solvent to eliminate lipid interference, it also reduced the extraction efficiency of EQ. Alternatively, effective purification may be achieved using C₁₈, PSA (primary secondary amine), or other functional materials, but would greatly affect the simplicity and economy of the whole performance and limit its real use for on-site and cost-effective screening. So it seems of great scientific significance to develop new, effective and simple pretreatment strategies for the immunoassay and fast analysis of EQ and other strong nonpolar compounds.

The partition coefficient (logP) plays a significant role in guiding the extraction process by indicating the distribution of a compound between oil and water. In principle for some compounds, such a distribution may vary under difference pH conditions and then further represented logD, which provides a more accurate understanding of how the compound behaves during various extractions(Sun et al., 2024; Trtić-Petrović et al., 2010). It is interesting to assume that: if the target with significantly varied logD within certain conditions, the precise pH adjustment may effectively enhance its nonpolar difference with the matrix components and therefore achieve both satisfactory extraction and purification. Though the pH adjustment has been used to improve the the pretreatment efficiency in various analysis(Campone et al., 2012), such a strategy has not be proposed and verified, that based on precisely adjust the partition coefficient of analyte and co-exiting components to eliminate the matrix interference. Here the secondary amine on the quinoline ring of EQ, which acts as a weak organic base(Aoki et al., 2010), has a pKa value of 5.15 and exhibits a broad range of logD values (based on MarvinSketch 15.6.29.0 software). So under alkaline conditions, it should exist in a molecular form with a higher logD, enhancing its solubility in organic solvents. By lowering the solution pH, the logD value of EQ significantly decreased, and would lead to ionization and improved solubility in aqueous solutions, which should significantly help o separate EQ from lipid co-extracts.

Based on the proposal mentioned above, in this study a novel and simple pretreatment strategy by pH adjustment of logD was suggested, and its efficiency to eliminate nonpolar matrix interference was well validated. Then the GICA was successfully established for rapid, sensitive and accurate detection of EQ in aquatic products. This provides an effective tool for on-site screening of EQ, more importantly, it offers a new way of thinking to develop pretreatment techniques for separation of target analytes from matrix interference, which would greatly help to improve the efficiency of fast analysis for various food hazards.

2. Materials and methods

2.1. Materials

The nitrocellulose membrane (CN140) was purchased from Sartorius Group (Göttingen, Germany), the PVC sheets, fiberglass membrane, and absorbent paper (CH27) were purchased from Shanghai Jinbiao Biotechnology Co., Ltd. (Shanghai, China), EQ antigens and antibodies were from Jiangnan University, and goat anti-mouse IgG was purchased from Beijing Solabao Technology Co., Ltd. (Beijing, China). HPLC-MS grade acetonitrile (ACN), methanol (MeOH) and formic acid were supplied by Fisher (Hampton, USA). Analytical-grade n-hexane, acetone, hydrochloric acid, sodium hydroxide, anhydrous magnesium sulfate, and anhydrous sodium sulfate were purchased from Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China). Analytical-grade chloroform was purchased from Shanghai Lingfeng Chemical Reagent Co., Ltd. (Shanghai, China). The bicinchoninic acid assay (BCA) protein assay kit was purchased from Beijing Solarbio Science & Technology Co., Ltd. (Beijing, China). EQ standard was purchased from Shanghai Yuanye Biotechnology Co., Ltd. (Shanghai, China) with a Certified Concentration of 1004 mg/L and an uncertainty of 31 mg/L. Ultrapure water was produced using a Milli-Q Millipore Water System (Millipore Corp., Bedford, MA, USA).

The large yellow croaker (Larimichthys Crocea), bass (Lateolabrax Japonicas), japanese eel (Anguilla Japonica), swimming crab (Portunus Trituberculatus), and atlantic salmon (Salmo Salar) were purchased from the market in Qingdao, China, additionally, some large yellow croaker samples were obtained from Ningde market, China. All samples were dead and chilled when purchased. The head, tail and bone of the fish sample were removed, edible parts such as muscle and skin were homogenized. The crab was deshelled to remove the edible parts and then homogenized uniformly. All samples were labeled and stored at −20 °C for use (Chinese National Standard GB/T 30891–2014). Each sample was tested with the Chinese national standard method (Chinese National Standard GB 23200.89–2016) to determine whether the presence of EQ in the samples.

2.2. Preparation of GICA test strips

In this study, a competitive method was used to prepare the GICA test strips. The process parameters were established for the amount of antigen, the amount of antibody labeling, and the stability after drying(C. Zhou et al., 2014). Specifically, the NC membrane was applied to the coating area of the PVC plate, and EQ antigen and goat anti-mouse secondary antibody were applied to the NC membrane using a dot blotting instrument (HGS101, AUTOKUN Technology, Hangzhou, China) at 1 μL/cm, respectively, as the test line (T line) and control line (C line). After drying at 37 °C for a certain period, the membrane was sealed and stored. Subsequently, the sample pad and absorbent pad were sequentially attached to the PVC board. Using a cutting machine (D6M, Hangan Electronic Technology Co., Ltd., Shanghai, China), the assembled large cards were cut into 3 mm wide test strips, which were then placed into the card cover and pressed tightly to complete the assembly of the test strips. The finished strips were stored in aluminum foil bags with desiccants for sealed storage. Colloidal gold and antibodies were evenly mixed under a specific pH. Then, 5 % bovine serum protein solution was added to achieve a final concentration of 1 %, and 3 % polyethylene glycol (MW 20000) was incorporated to reach a final concentration of 0.05 %. Next, 5 μL of the gold - labeled antibody solution was added to each well and dried at 37 °C for 0.5 h to complete the preparation of the gold - labeled microwell.

2.3. GICA procedure and result determination

The GICA utilizes the principle of indirect competition, and the detection procedure was as follows(Jia et al., 2024): 150 μL of the test sample solution was pipetted into the gold-labeled microwell, then blown with a small pipette until the red substance within the well is completely dissolved. After waiting for the reaction for 3 min, all the solution in the gold-labeled microwell was aspirated and added to the sample well, and timing begins after the addition of the sample. The test results should be read at 8 min. Results were determined using two methods: visual inspection and reading with an instrument (HG-8, Hugo Scientific Instruments Co., Ltd., Shanghai, China).

• A.Negative result (−): The color intensity of the T line was equal to or greater than that of the C line, with a T/C ratio greater than 1, indicating that the EQ concentration was below the detection limit.
• B.Positive result (+): The color intensity of the T line was less than that of the C line, with a T/C ratio less than 1, indicating that the EQ concentration was above the detection limit.
• C.The control line should always have been visible, clear, and consistent; otherwise, the result was considered invalid.

As a semi-quantitative method, GICA provides a relatively accurate value for detection. Given its application scenarios, GICA serves as a rapid screening tool during on-site testing, where positive results still require confirmation of the actual content using conventional laboratory instruments (HPLC or LC-MS/MS). This study focused on addressing the severe matrix interference that critically impacts its practical use, aiming to enhance the applicability of GICA for ethoxyquin detection. Therefore in order to consider the application, this study was as close as possible to the operation of the field users, reducing the process of making the standard curve, and only using the T/C value as the basis for judgment.

2.4. Sample pretreatment

The method was established based on the China Entry-Exit Inspection and Quarantine Industry Standard (SN/T 3856–2014), with modifications and simplifications to meet the requirements of on-site testing. First, accurately weigh 2.0 g of the sample into a 50 mL polypropylene centrifuge tube. Add 1 % vitamin C, followed by the EQ standard, and incubate in the dark for 10 min. Next, add 1.0 g of anhydrous sodium carbonate and 1.0 g of anhydrous sodium sulfate, then add 10 mL of n-hexane and vortex to mix. Centrifuge at 4000 rpm for 5 min and transfer the upper n-hexane layer to another 15 mL centrifuge tube. Add 4 mL of 1 M hydrochloric acid to acidify the mixture, allow the phases to separate, and then remove the n-hexane layer. To the lower aqueous phase, add 1 mL of 2 M sodium hydroxide and extract twice with 2 mL of n-hexane. Combine the n-hexane extracts, concentrate under nitrogen at 40 °C, and re-dissolve in 1.0 mL of solvent (1 % PBS and 1 % Triton-100 solution).

2.5. HPLC-MS/MS instrumentation and conditions

The HPLC-MS/MS method referenced the research of Wang et al., and modifications were made accordingly(M. Wang et al., 2023). The analysis of target veterinary drug residues was conducted using an Agilent Technologies spectrometer (1290 Infinity/6410B, Santa Clara, California, USA). The column temperature was set at 30 °C, and the injection volume was 5 μL. Isocratic elution was performed with mobile phase A (0.1 % formic acid in water) and B (0.1 % formic acid in acetonitrile) in a ratio of 30:70, at a flow rate of 0.35 mL/min.

All data were acquired using an Agilent G6410B triple quadrupole mass spectrometer with an ESI source, operating in multiple reaction monitoring (MRM) mode. The mass spectrometer was run in positive electrospray ionization (ESI⁺) mode. The capillary voltage for the ESI⁺ source was set to +4000 V. The gas temperature and flow rate were 350 °C and 3 L/min, respectively, with a nebulizer pressure of 30 psi. MassHunter QuanAnalysis B.09.00 software was used for batch processing of the collected data. MRM parameters were presented in Supplementary Table 1.

2.6. Determination of components

The detection of protein content was conducted using the BCA method(Reichelt et al., 2016). Considering the potential interference from lipid co-extracts in the complex solution, which can cause turbidity and may have an impact on the results, the samples were centrifuged at 15,000 rpm at 4 °C to remove the upper layer of water-insoluble materials, with only the clear lower phase being analyzed. The determination of lipid content was conducted using the chloroform-methanol method, with modifications made to the method described by Wang et al.(Xiaoxiao Wang et al., 2018). The resuspended solution was extracted thoroughly with 15 mL of a chloroform-methanol mixture (volume ratio 2:1). Then, 5 mL of 0.5 % zinc acetate was added, and the mixture was thoroughly blended and allowed to stand for 16 h. After allowing the mixture to settle and separate into layers, exactly 8 mL of the lower chloroform layer was carefully removed and transferred to another clean container. This was then dried to a constant weight at 80 °C.

3. Results and discussion

3.1. Preparation and investigation of GICA test strips

The microwell gold immunochromatography was used in this study, where analytes and gold - labeled antibodies fully mix in microwell before detection. This method enhances stability and sensitivity. The reaction of the sample in the microwell can avoid the problem of insufficient reaction between the analyte and the gold-labeled antibody and the competitiveness of the coating antigen, and also avoid environmental interference(Hua et al., 2021). In the preparation, the amount of reagent material can be reduced to reduce the cost.

Using the T/C value as the parameter, the influence of antigen ratio, and antibody labeling amount and drying time on the strip was investigated, and the main preparation conditions was optimized, which were summarized as the following: antigen concentration of 0.273 mg/mL, antibody labeling amount of 3 mg/mL, and drying at 47 °C for 4 days (Supplementary Table 2–4).

Using standard solutions, the detection ability of the GICA test strip for EQ was evaluated. When the T/C value of 1.0 was set as the threshold for detecting a positive result, the lowest detection limit was calculated as 3 μg/L, and the RSD was less than 7.5 % at concentration levels ranging from 0 to 10 μg/L (Supplementary Table 5). The specificity of the GICA test strips was evaluated, including four compounds with quinolone ring structure analogs (ethoxyquinoline dimer, enrofloxacin, norfloxacin, and ofloxacin) and five lipophilic antioxidant analogs (ascorbyl palmitate, butylated hydroxyanisole, butylated hydroxytoluene, propyl gallate, and tert-butylhydroquinone). The results were all negative, with no cross-reactivity, indicating good specificity (Supplementary Table 6). All these indicated satisfactory efficiency of the prepared strip and could fully meet relative requirements.

3.2. The investigation of matrix effects

Using the bass and large yellow croaker as samples, the efficiency of the two representative pretreatment methods widely applied for detection of EQ (based on HPLC-MS/MS analysis) was investigated. Though there is significance in solvents and operation procedures, these two techniques both exploited conventional liquid-liquid extraction based on the “like dissolves like” principle, and their efficiency has been fully verified to be adopted into some standard regulations (for example GB 23200.89–2016). However, when the techniques were applied in the GICA test, very poor results were observed: large yellow croaker samples consistently yielded false-negative results, even at an spiking concentration of 100 μg/kg (Table 1),which was about 30 times of the detection limit (3 μg/L) in standard solutions. The real EQ contents in the pretreated solutions were further measured by the HPLC-MS/MS method, and the EQ extraction recoveries of bass and large yellow croaker by the two pretreatment methods were calculated to be higher than 51 % and 79 %, respectively (Supplementary Table 7), which clearly indicated that, the observed poor sensitivity, “false negative results” and the significant difference in the two samples, should be mainly due to the matrix interference, other than the insufficient extraction.

Table 1.

The results of gold immunochromatographic assay detection using common methods (added concentration: 20 μg/kg).

Method	Matrix	T/C	Result
	Reagent Blank	2.535	−
Method A	Large yellow croaker- Blank 1	2.005	−
	Large yellow croaker- Blank 2	1.927	−
	Large yellow croaker- ADD 1	1.297	−
	Large yellow croaker- ADD 2	1.438	−
	Large yellow croaker- ADD 3	1.353	−
	Bass- Blank 1	1.406	−
	Bass- Blank 2	1.651	−
	Bass- ADD 1	0.638	+
	Bass- ADD 2	0.404	+
	Bass- ADD 3	0.203	+
Method B	Large yellow croaker- ADD 1	1.568	−
	Large yellow croaker- ADD 2	1.718	−
	Large yellow croaker- ADD 1	1.905	−
	Bass- ADD 1	0.000	+
	Bass- ADD 2	0.140	+
	Bass- ADD 3	0.372	+

Note: ADD means the addition rate recovery experiment, (−) is negative, (+) is positive.

In lots of previous studies, some proteins and irons in aquatic products have been fully verified able to significantly interfere to GICA and other immunoassays, especially the co-existing proteins were believed as the main source of the matrix interference in most cases. While in this study, n-hexane was used as the extraction solvent, and the protein content in the co-extractants (<0.60 mg/mL) was far lower than the levels reported in other literature that can cause interference(Xiaoxiao Wang et al., 2018). In addition, the protein content in bass extracts was almost twice that of large yellow croaker, but demonstrated much little interference to the GICA performance (Table 2). Moreover, the protein-induced interference, which was assumed by the possible non-specific interaction with IgG reagents, usually lead to “false positive results”, other than the “false negative results” observed here. All these allowed us to suggest that, proteins were not the primary source of interference, and some other matrix components should play important roles.

Table 2.

The results of protein (n = 3) and lipid co-extracts(n = 2) content in the solution to be measured by different pretreatment methods.

Method	Matrix	Protein Content (mg/mL)	Lipid Co-extracts Content (g)
Method A	Large yellow croaker	0.338 ± 0.017	0.589 ± 0.038
	Bass	0.600 ± 0.017	0.128 ± 0.006
Method B	Large yellow croaker	0.262 ± 0.022	0.326 ± 0.043
	Bass	0.502 ± 0.091	0.125 ± 0.003
This work	Large yellow croaker	0.164 ± 0.026	0.120 ± 0.006
	Bass	0.140 ± 0.008	0.119 ± 0.010

The detection value of protein in blank reagent was 0.0712 mg / mL, and the c lipid co-extracts in large yellow croaker meat was 0.645 g, 0.171 g in bass meat.

Considering the strong nonpolar extraction conditions, the influence of some lipid co-extracts on the GICA could be reasonably proposed, though till now it has not been reported in previous studies. With both two pretreatments, there were significant lipid co-extract remained in the prepared solutions, even reach a revery rate of 90 % in comparison to the original fish samples. The total lipid co-extract content in large yellow croaker was measured approximately 3–5 times that of that of bass, fitting well with the interference observed. When the lipid co-extracts isolated from the large yellow croaker were added to EQ standard at various concentrations, it resulted in a noticeable increase in the T/C values of the GICA test, indicating significant interference as observed before(Table 3). A Kendall correlation analysis was performed between the amount of lipid added and the average T/C value, yielding a Kendall correlation coefficient of 0.867 with a p-value of 0.017 (<0.05), which revealed a strong positive correlation between the two variables with no statistically significant difference(The python code was provided in supplementary materials)(Smirnova et al., 2022). Ultimately, it was determined that lipid co-extracts were the main interfering substance. On the one hand, lipids may cover or encapsulate the target compound or affect the reaction environment, resulting in a reduction of the available target compound for reaction and causing interference. On the other hand, lipids may interact with the antibody surface, leading to a decrease in the affinity of the antibody for the target molecule(Xiaoyan Wang, 2024). Based all these results, some lipid co-extracts could be fully confirmed as an important source of matrix interference to GICA, and such a new finding would be much valuable to further develop and improve the GICA for various food hazards, especially those with similar and strong polarity with EQ. Detailed information is still very limited, such as the accurate components, their action mechanism and the relation with extraction conditions, and we will continue to clarify them in the future work.

Table 3.

Gradient increase of lipid interference on gold immunochromatographic assay(n = 3).

Amount of lipid extract added (g)	Average of T/C	RSD (%)
0	0.342	2.9
0.1	0.868	2.0
0.2	0.932	10.7
0.3	1.025	3.9
0.4	1.014	2.2
0.5	1.119	4.8

In the 10 μg/kg EQ standard solution, lipid extracts obtained from method A were added incrementally, with the amounts being 0, 0.1, 0.2, 0.3, 0.4, and 0.5 g, respectively.

3.3. Design and evaluation of the new LLE strategy based on the adjustment of the partition coefficient

As mentioned in the introduction, in principle the secondary amine on the quinoline ring of EQ allows it to exhibit a broad range of logD values under different pH conditions. When the pH drops below 2, the logD of EQ would decreased to transfer into ionization state and improve the solubility in aqueous solutions. Then, adjusting the pH above 8 can make the logD quickly rise, turning EQ into the molecular state for re-extraction by n-hexane (Fig. 1).

Fig. 1.

Variations in logD and pKa Values of Ethoxyquin.

(Based on MarvinSketch software)

For fabrication of the new pretreatment system, the influence of extraction solvent was at first investigated. Acetonitrile could effectively precipitate proteins and reduce the co-extraction of lipid interferences(Perestrelo et al., 2019). However, it did not yield satisfactory recovery rates (less than 10 %), possibly due to inefficient transfer of the compound between phases during acid-base adjustments. n-hexane was usually exploited for EQ, and here it also demonstrated satisfactory extraction efficiency (greater than 50 %), therefore was chosen to mix with anhydrous sodium carbonate and anhydrous sodium sulfate as the final extraction solvents, anhydrous sodium carbonate was used to adjust the pH, while anhydrous sodium sulfate was used to remove water. The solution's pH was rapidly adjusted between 1 and 11 using hydrochloric acid or sodium hydroxide since there were no buffering salts in the solution, which aligned well with the proposed EQ distribution behavior (Fig. 1). Then the Plackett-Burman experimental design was used to quickly identify the critical factors in the pretreatment process, the steepest ascent method was applied to determine the central conditions, and finally, Box-Behnken response surface methodology was employed to optimize the pretreatment method.

3.3.1. Plackett-Burman experimental

The Plackett-Burman experimental design could effectively captures the interactions between factors and provides an economical and efficient approach to screening important variables affecting the experiment(Wu et al., 2024). Here selected factors included: sample weight, Vc ratio, masses of anhydrous sodium carbonate and sodium sulfate, extraction solvent volume, hydrochloric acid volume, extraction agent volume, nitrogen blowing temperature, and reconstitution volume. Additionally, virtual factors were incorporated to account for experimental errors. The levels and variables for the Plackett-Burman design are presented in supplementary table 8.

Based on the results of the Plackett-Burman design (Table 4). The trial data were analyzed using analysis of variance (ANOVA), and the Lenth method was applied to identify significant effects(Lenth, 1989). A semi-normal probability plot and Pareto plot of the standardized effects of the factors were generated (Supplementary Fig. 2), and additional details were derived from the supplementary table 9. The P-value of the model was 0.0225 (<0.05), which confirmed well its significance and reliability. Both figures indicated that C (anhydrous sodium carbonate), A(sample weight), and F(hydrochloric acid volume) were the main influencing factors. The positive correlation with sample weight was likely due to the pre-treatment method's capacity to accommodate more sample extraction, leading to higher drug concentration in the solvent and more stable extraction. Sodium carbonate showed a negative correlation. When 2 g sodium carbonate was added, it exceeded its solubility, resulting in excess sodium carbonate that may have interfered with the extraction process by covering the surface of the sample. The volume of hydrochloric acid affected the pH but may also have limited the aqueous phase solution, preventing the complete transfer of ethoxyquin from the organic phase to the aqueous phase.

Table 4.

The results of the Plackett-Burman design.

NO.	A	B	C	D	E	F	G	H	I	Recovery rate
	Sample weight	VC ratio	Anhydrous sodium carbonate	Anhydrous sodium sulfate	Extract volume	Hydrochloric acid volume	Secondary extraction volume	Nitrogen blowing temperature	Redissolved volume
	g	%	g	g	mL	mL	mL	°C	mL	%
1	3.0	2.0	1.0	1.0	10.0	4.0	1.0	40	1.0	128.31
2	2.0	1.0	1.0	2.0	10.0	4.0	2.0	30	1.0	61.83
3	3.0	2.0	2.0	1.0	10.0	2.0	2.0	30	1.0	45.64
4	2.0	2.0	2.0	1.0	15.0	4.0	2.0	30	0.5	20.96
5	3.0	1.0	2.0	2.0	10.0	4.0	2.0	40	0.5	67.49
6	3.0	2.0	1.0	2.0	15.0	4.0	1.0	30	0.5	83.19
7	2.0	1.0	1.0	1.0	10.0	2.0	1.0	30	0.5	46.49
8	3.0	1.0	2.0	2.0	15.0	2.0	1.0	30	1.0	22.10
9	3.0	1.0	1.0	1.0	15.0	2.0	2.0	40	0.5	85.31
10	2.0	2.0	1.0	2.0	15.0	2.0	2.0	40	1.0	53.09
11	2.0	1.0	2.0	1.0	15.0	4.0	1.0	40	1.0	23.08
12	2.0	2.0	2.0	2.0	10.0	2.0	1.0	40	0.5	12.99

3.3.2. Results of the steepest ascent experiment

Based on the Plackett-Burman experimental design, the factors with a significant impact on the response value were identified. The step lengths were determined according to the effect size of each factor, and the direction of ascent (positive or negative) was selected to identify the maximum response value. In the steepest ascent experiment, the non-significant factors were held constant(Wei et al., 2023). However, the coefficients of the corresponding factors in the regression equation were large, which did not align with the actual sample pretreatment process. Therefore, custom step lengths were defined as the following: anhydrous sodium carbonate, which had a negative effect, was assigned a step length of −0.5 g, while sample weight and hydrochloric acid volume, which had positive effects, were assigned step lengths of +0.5 g and + 1 mL, respectively. The design and results of the steepest ascent experiment were presented in Table 5. Since the outcomes of experiments No. 3 and No. 4 were essentially similar, experimental treatment No. 3 was selected as the center point for the Box-Behnken response surface experiment.

Table 5.

Design and Results of Steepest Ascent Experiment(n = 3).

Number	step size	Sample weight (g)	Anhydrous sodium carbonate (g)	Hydrochloric acid volume (mL)	Rcovery rate (%)
1	0	2.0	2.0	2.0	25.35
2	0 + △1	2.5	1.5	3.0	44.36
3	0 + △2	3.0	1.0	4.0	64.98
4	0 + △3	3.5	0.5	5.0	71.09
5	0 + △4	4.0	0.0	6.0	41.90

3.3.3. Analysis of the box-Behnken experimental design

Based on the steepest ascent experiment, a Box-Behnken experiment was conducted using experiment No.3 as the center point. A 3-factor, 3-level experiment was designed according to the Box-Behnken method, with recovery rate as the response variable (Table 6) (J. Zhou et al., 2018).

Table 6.

Design and Results of Box-Behnken Experiment.

No.	A Sample weight (g)	B Anhydrous sodium carbonate (g)	C Hydrochloric acid volume (mL)	Rcovery rate (%)
1	3.0	1.0	4.0	70.19
2	3.0	0.5	5.0	66.29
3	3.0	1.5	3.0	54.20
4	2.5	0.5	4.0	63.03
5	3.0	0.5	3.0	59.03
6	3.5	1.5	4.0	62.97
7	3.0	1.5	5.0	59.30
8	3.5	0.5	4.0	64.09
9	2.5	1.0	3.0	63.81
10	3.0	1.0	4.0	73.94
11	3.0	1.0	4.0	75.77
12	2.5	1.5	4.0	62.96
13	3.0	1.0	4.0	74.19
14	3.5	1.0	3.0	64.58
15	2.5	1.0	5.0	73.15
16	3.5	1.0	5.0	67.10
17	3.0	1.0	4.0	74.02

The variance analysis of the regression equation (Supplementary Table 10) indicated that the model had a high degree of significance and fit, influenced by the independent variables rather than random fluctuations. According to the results of the variance analysis, the factors affecting the recovery rate in order of significance were: hydrochloric acid volume > anhydrous sodium carbonate > sample weight. The response surface analysis describing the interactions between these two factors was shown in Supplementary Figs. 3 A-F.

During method development, significant recovery rate differences emerged due to varying condition combinations, with some yielding low recoveries. This is common in response surface analysis, as preprocessing changes greatly impact recovery, and extreme condition combinations can heavily interfere with drug extraction. However, this is integral to model computation and crucial for developing a good model. Similar results have also been observed in other studies using this approach(Wei et al., 2023). Using the established model, the optimal conditions were predicted by the Design Expert software as: sample weight of 2.804 g, anhydrous sodium carbonate amount of 0.952 g, and hydrochloric acid volume of 4.374 mL, which were expected to achieve the highest recovery rate. A recovery experiment was conducted using the optimized conditions and validated by mass spectrometry (n = 6). The final recovery rate was calculated as 78.52 %, though slightly higher than the predicted value of 74.37 % from the response surface method, it still fell within the predicted range (0.935) and essentially close to the theoretical value(Wu et al., 2024). The newly established method demonstrates a strong lipid removal effect (Table 2). These results demonstrate that the model reasonably reflects the actual sample pretreatment recovery rate, confirming that the optimized method is feasible and applicable.

3.4. Method validation

In accordance with the European Union Directive 2002/657(Commission, European, 2002) and the Food and Drug Administration (FDA) guidelines for bioanalytical method validation(U.S. Department of Health and Human Services, Food and Drug Administration, CDER, CVM, 2001), the semi-quantitative GICA method used in this study was validated. The validation focused on the method's sensitivity, accuracy, and precision in large yellow croaker, as well as its performance across various aquatic products. Finally, the method was applied to real sample analysis and compared with the Chinese National Standard (GB 23200.89–2016).

Through the addition of EQ standard solutions for testing, when the addition concentration was 10 μg/kg, a good T/C value differentiation could still be achieved, allowing for accuracy determination. Therefore, the detection limit of this method was 10 μg/kg. This detection limit is the same as that of Method A (Chinese National Standard, LC-MS/MS detection) and is much lower than the EQ detection limits set by various countries for aquatic products. When validation was carried out at three different addition concentrations of 10.0 μg/kg, 20.0 μg/kg, and 100.0 μg/kg (low, medium, and high), good validation results were obtained for all three concentration levels, with RSD values below 19.0 %. Moreover, when the EQ concentration in the sample was higher than 100.0 μg/kg, the T line of the GICA test strip showed a disappearance result. Subsequently, the effectiveness of this method in other types of aquatic products was verified. In bass, swimming crab, Atlantic salmon, and Japanese eel, a detection limit of 10 μg/kg could be achieved, with accurate determinations and no occurrence of false negatives or false positives. The overall RSD value was below 15.5% (Table 7). Nine actual samples were simultaneously tested using the developed method and Method A. At a detection line of 10 μg/kg, all results were accurately identified without any misjudgment (Table 8). Therefore, this method was validated to be effective for the rapid detection of EQ in high-fat or matrix-interfered aquatic products, serving as a reliable alternative for on-site testing. In this study, a rapid detection method of EQ in aquatic products based on GICA was successfully developed. Compared with the patented method of Luo et al., the pretreatment method improved the recovery rate and matched the GICA detection well under the same premise of removing lipid interference(Luo; et al., 2013).

Table 7.

Validation Results of GICA(n = 3).

Matrix	Added concentration (μg/kg)	Result	Average T/C	RSD(%)
Large yellow croaker	0.0	−	1.188	4.1
	10.0	+	0.570	11.1
	20.0	+	0.203	19.0
	100.0	+	0.000	0.0
Bass	0.0	−	1.213	8.8
	10.0	+	0.674	7.6
Swimming crab	0.0	−	1.208	3.2
	10.0	+	0.585	15.5
Atlantic salmon	0.0	−	1.307	1.8
	10.0	+	0.330	4.5
Japanese eel	0.0	−	1.372	7.7
	10.0	+	0.620	14.4

Note: (−) is negative, (+) is positive.

Table 8.

Real Sample Analysis (n = 2).

Samples	Source	This work (T/C value)	Method A (μg/kg)
Large yellow croaker-1	Ningde	1.410	<LOD
Large yellow croaker-2		1.360	<LOD
Large yellow croaker-3		0.661	27.14
Large yellow croaker-4		0.834	12.95
Large yellow croaker-5	Qingdao	1.188	<LOD
Bass		1.213	<LOD
Japanese eel		1.208	<LOD
Swimming crab		1.307	<LOD
Atlantic salmon		1.372	<LOD

Note: LOD indicates that the detection value was below the limit of detection (10 μg/kg) of Method A (GB 23200.89–2016).

4. Conclusion

In this study, a novel on-site screening method compatible with GICA was successfully developed for the rapid detection of EQ in aquatic products. By leveraging the pH-dependent logD properties of EQ—exhibiting high hydrophobicity under alkaline conditions and reduced hydrophobicity under acidic conditions—a tailored extraction and purification strategy was designed to effectively separate EQ from lipid co-extracts, thereby eliminating matrix interference that has compromised GICA accuracy. The pretreatment protocol, involving alkaline n-hexane extraction followed by pH adjustment, was systematically optimized using Plackett-Burman experimental design, the steepest ascent method, and Box-Behnken response surface analysis. This approach achieved an EQ recovery rate of 78.52 %, closely aligning with model predictions and validating the robustness of the optimization process. The method demonstrated a detection limit of 10 μg/kg, comparable to conventional LC-MS/MS performance and well below international regulatory thresholds, while offering superior simplicity, cost-effectiveness, and field applicability. The method demonstrated good detection performance at low, medium, and high spike levels, with RSD values less than 19.0 %. Validation in various aquatic matrices confirmed the reliability of the method for high - fat or complex samples. This work not only established a practical GICA solution for EQ monitoring for the first time, but also provided a general framework for designing pretreatment strategies for highly non-polar pollutants, which can advance the development of rapid, on-site detection and analysis tools and enhance food safety monitoring capabilities. .

CRediT authorship contribution statement

Qifan Sun: Writing – original draft, Validation, Investigation, Formal analysis, Data curation. Lin Zhang: Validation, Data curation. Shaoen Zhang: Resources, Funding acquisition. Qingzhou Chen: Resources, Funding acquisition. Jinjun Ying: Resources, Data curation. Hong Lin: Supervision, Resources, Funding acquisition. Jianxin Sui: Supervision, Resources, Funding acquisition. Kaiqiang Wang: Supervision, Project administration, Funding acquisition. Xiudan Wang: Supervision, Resources, Funding acquisition. Limin Cao: Writing – review & editing, Supervision, Project administration, Funding acquisition, Formal analysis.

Declaration of competing interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Appendix A. Supplementary data

Supplementary material

Data availability

Data will be made available on request.