Development and Analytical Validation of a Smartphone‐Based Quantitative Lateral Flow Immunoassay for Serum Cystatin‐C

ABSTRACT

Background

Cystatin‐C is an important renal function biomarker, but conventional quantification requires centralized laboratory analyzers, limiting timely testing in primary care and resource‐limited settings.

Methods

We developed and validated a quantitative smartphone‐based lateral flow immunoassay (SP‐LFIA) comprising a colorimetric strip, a Redmi Note 12R smartphone, a fixed‐geometry macro‐lens module with integrated LED illumination, and an Android application. The application directly converted background‐corrected test/control‐line (T/C) ratios into concentrations using a batch‐specific third‐order polynomial calibration model. Analytical validation followed relevant Clinical and Laboratory Standards Institute guidelines.

Results

The assay showed a quantitative range of 0.32–8.00 mg/L and a limit of detection of 0.15 mg/L. Intra‐ and inter‐assay coefficients of variation were below 10%. No significant interference was observed from bilirubin, triglycerides, hemoglobin, rheumatoid factor, gentamicin sulfate, norvancomycin hydrochloride, or cisplatin at the tested concentrations. Accelerated testing supported strip performance after storage at 50°C for 28 days. In 120 clinical serum samples, SP‐LFIA agreed closely with the commercial particle‐enhanced turbidimetric immunoassay reference method (R² = 0.9946), with a mean bias of −0.01 mg/L.

Conclusion

The smartphone‐based LFIA provides a reliable, rapid, low‐cost, and practical approach to quantitative Cystatin‐C testing in point‐of‐care settings where serum preparation is available.

We developed a gold nanoparticle‐based lateral flow immunoassay that pairs a fixed‐geometry smartphone reader with an Android application for quantitative serum Cystatin‐C measurement. The 15‐min assay covered 0.32–8.00 mg/L with a 0.15 mg/L detection limit and CVs below 10%. In 120 clinical serum samples, results closely agreed with the commercial PETIA method (R ² = 0.9946; mean bias, −0.01 mg/L), supporting further evaluation for decentralized testing where serum preparation is available.

Abbreviations

AuNP

gold nanoparticle

BSA

bovine serum albumin

CKD

chronic kidney disease

CLSI

Clinical and Laboratory Standards Institute

CV

coefficient of variation

CysC

Cystatin C

ELISA

enzyme‐linked immunosorbent assay

GFR

glomerular filtration rate

IgY

immunoglobulin Y

LFIA

lateral flow immunoassay

LOD

limit of detection

LOQ

limit of quantification

LSPR

Localized Surface Plasmon Resonance

PBS

phosphate buffered saline

PENIA

particle‐enhanced nephelometric immunoassay

PETIA

particle‐enhanced turbidimetric immunoassay

POCT

point‐of‐care testing

1. Introduction

Chronic kidney disease (CKD) represents a major and growing global public health burden, affecting more than 10% of the adult population worldwide and contributing substantially to morbidity, mortality, and healthcare expenditure [1, 2, 3]. Early identification of renal dysfunction is critical because timely intervention can slow disease progression, reduce cardiovascular complications, and improve long‐term outcomes in patients, particularly in the elderly and other high‐risk populations, for example, those with diabetes or cardiovascular disease [4, 5, 6, 7]. However, chronic kidney disease (CKD) often has few or no symptoms during early stages, emphasizing the need for a sensitive accessible biomarker for prompt detection and routine monitoring of renal function.

Cystatin‐C is an endogenous biomarker that can reliably assess renal function, and estimate glomerular filtration rate (GFR). Unlike serum creatinine, Cystatin‐C production is relatively independent of muscle mass, age, sex, and dietary intake, making it particularly useful in risk stratifying in the elderly as well as in patients with diabetes or altered muscle metabolism [2, 6, 7, 8, 9]. Additional evidence, including comparative, clinical, and assay standardization studies, supports the use of Cystatin‐C for GFR assessment [10, 11, 12]. Many clinical studies have demonstrated that elevations of Cystatin‐C are associated with adverse renal and cardiovascular outcomes even when GFR estimates based on creatinine appear to be normal [6, 7, 8, 9]. In addition, longitudinal evidence and clinical guidelines support the use of Cystatin‐C for risk assessment and monitoring in older adults and other high‐risk patient groups [6, 7, 8, 9, 13, 14].

Despite the clinical usefulness of Cystatin‐C measurement and the increased availability of methods to measure Cystatin‐C, routine Cystatin‐C measurement remains greatly limited in many health care settings. Currently, the clinical assays available to measure Cystatin‐C utilize particle‐enhanced turbidimetric or nephelometric immunoassays (PETIA/PENIA) requiring centralized laboratory analyzers, trained personnel, and strict quality control procedures [15, 16, 17, 18]. Consequently, the availability of Cystatin‐C assays is generally limited in primary care clinics, emergency departments, community hospitals, and resource‐limited environments. In addition, the costs associated with laboratory‐based testing, as well as the longer turnaround times to obtain results, severely limit the potential for routine monitoring and screening of large populations for Cystatin‐C.

Alternative analytical methods for the detection of Cystatin‐C have been explored to address these shortcomings, including enzyme‐linked immunosorbent assays (ELISA), electrochemical sensors, and lateral flow immunoassays (LFIA) [19, 20, 21, 22, 23]. LFIAs offer unique advantages over other methods (i.e., ELISA and electrochemical sensors) in their simplicity, fast analysis, and low cost, making them attractive for point‐of‐care testing (POCT) applications. However, these traditional LFIAs are primarily qualitative or semi‐quantitative in nature and typically require visual interpretation of results or a simple card reader for reading. Such methods introduce operator‐dependent variability into the results and make them unsuitable for accurate clinical quantification [22, 23, 24, 25].

Recent advances in smartphone‐based diagnostic technologies have created new opportunities to improve the quantitative capability of colorimetric LFIAs. Smartphones are capable of providing high‐resolution images and have standardized data processing and flexible computer processing capability providing the means to quantitatively analyze signals without specialized instruments [26, 27, 28, 29, 30]. Furthermore, when colorimetric LFIA assays are optimally designed and combined with robust image processing algorithms, smartphone assisted LFIAs exhibit improved analytical performance and reproducibility, while still maintaining the operational simplicity and low‐cost benefits associated with traditional LFIA methods.

This study was conducted to develop and validate a smartphone‐based quantitative colorimetric LFIA for measuring Cystatin‐C in serum. By integrating a conventional gold nanoparticle‐based LFIA strip with a customized smartphone reader and dedicated image analysis software, the proposed platform seeks to overcome the limitations of traditional LFIA systems while retaining their inherent advantages. The analytical performance, clinical applicability, and agreement with an established PETIA reference method were systematically evaluated using human serum samples. This work positions the developed platform as a practical and accessible solution for decentralized renal function assessment and point‐of‐care Cystatin‐C testing.

2. Materials and Methods

2.1. Materials

Chicken egg yolk immunoglobulin Y (IgY) and polyclonal goat anti‐chicken IgY were purchased from Solarbio Biotechnology (Hangzhou, China). Recombinant human Cystatin‐C protein, detection antibody (Cystatin‐C Ab1, CY‐C2401), and capture antibody (Cystatin‐C Ab2, CY‐C2402) were obtained from Chuangye Biotechnology (Shaoxing, China). Casein was purchased from Sigma‐Aldrich (Shanghai, China). Chloroauric acid (HAuCl₄), boric acid, sodium borate, trisodium citrate, sodium carbonate, sodium azide, buffer components and stabilizers (Tween‐20, sucrose, trehalose, BSA, PBS), as well as gentamicin sulfate, norvancomycin hydrochloride, and cisplatin used for drug‐interference testing, were all purchased from Sinopharm Chemical Reagent Co. (Shanghai, China). Nitrocellulose membranes (NC140) were supplied by Sartorius (Göttingen, Germany). We obtained the glass fiber conjugate pads (SB08), sample pads (SM3‐125), and absorbent pads (ABP‐370) from Jinbiao Biotechnology (Shanghai, China).

2.2. Instruments

An HM3035 dispensing instrument (Jinbiao Biotechnology, Shanghai, China) was employed for reagent deposition onto the LFIA substrates. The assembled cards were subsequently cut into 3 mm wide strips using a ZQ3500 guillotine cutter (Jinbiao Biotechnology, Shanghai, China). A Redmi Note 12R Android smartphone (Xiaomi, Beijing, China) served as the imaging and computing device within the portable reader assembly. A fixed‐geometry macro‐lens module (Apexel APL‐MS009, focal length 5.5–7 mm) incorporating LED illumination was mounted on the smartphone camera. The macro lens was fitted directly against the test‐card imaging window, while the built‐in LED ring provided fixed illumination and the reader geometry isolated the imaging region from ambient light and fixed the camera‐to‐strip angle.

2.3. Antibody Adsorption Onto AuNPs

Monodisperse gold nanoparticles (AuNPs) were synthesized using the citrate reduction method. Briefly, 100 mL of 0.01% HAuCl4 solution was heated to boiling, followed by rapid addition of 3 mL of 1% trisodium citrate solution under vigorous stirring. The mixture was boiled for an additional 15 min until a wine‐red color appeared, indicating AuNP formation. The resultant AuNPs exhibited a characteristic localized surface plasmon resonance (LSPR) peak at 525 nm and had an average diameter of approximately 32 nm, as measured by UV–Vis spectroscopy and transmission electron microscopy (TEM), consistent with previous reports [31].

Antibody adsorption onto AuNPs was performed by adjusting two 10 mL aliquots of AuNP colloidal solution (OD 1.0 at 525 nm) to pH 8.75 with 1% (w/v) K₂CO₃. Cystatin‐C Ab1 (150 μg) was added to one aliquot and IgY (120 μg) to the other, followed by gentle mixing and incubation for 20 min at room temperature. Unoccupied surface sites were blocked by adding 4 mL of 1% BSA in 25 mM borate buffer (pH 8.75) to each mixture and incubating for a further 20 min. The suspensions were centrifuged (8000×g, 30 min, 4°C) to collect the Cystatin‐C Ab1–AuNP and IgY–AuNP pellets, which were washed three times with 0.25% BSA in 25 mM borate buffer (pH 8.75). Before storage at 2°C–8°C, the AuNP‐labeled antibody probes were resuspended in 50 mM sodium borate (pH 8.75) containing 1% BSA, 0.2% Tween‐20, 2% sucrose, 2% trehalose, and 0.2% NaN₃.

2.4. Buffer Solutions

The buffer solutions used in this study were as follows: AuNP‐labeled antibody‐probe blocking buffer (25 mM borate buffer, 1% BSA, pH 8.75), AuNP‐labeled antibody‐probe washing buffer (25 mM borate buffer, 0.25% BSA, pH 8.75), probe storage buffer (50 mM sodium borate, 1% BSA, 0.2% Tween‐20, 2% sucrose, 2% trehalose, 0.2% NaN₃, pH 8.75), test/control line dispensing buffer (10 mM PBS, 1% sucrose, pH 8.75), sample pad buffer (0.1 M PBS, 1% BSA, 0.05% Tween‐20, pH 7.4), conjugate pad buffer (0.1 M PBS, 0.05% Tween‐20, 20% sucrose, 1% BSA, pH 7.4), washing buffer (PBS, 0.05% Tween‐20), and assay running buffer (PBS, 0.1% Tween‐20, 1% BSA, pH 7.4).

2.5. LFIA Strip Assembly

LFIA strip assembly involved the integration of four main components: a glass fiber sample pad (SM3‐125), a glass fiber conjugate pad (SB08) preloaded with AuNP‐labeled antibody probes, a nitrocellulose membrane (NC140), and an absorbent pad (ABP‐370). The test line consisted of Cystatin‐C Ab2 (1.2 mg/mL), while the control line contained goat anti‐chicken IgY (1.0 mg/mL), both diluted in 10 mM PBS with 1% sucrose (pH 8.75). Test and control lines were striped onto the nitrocellulose membrane at a density of approximately 1 μL/cm using an HM3035 dispenser. Subsequent drying was carried out at 37°C for 2 h. The sample pad was pretreated with sample pad buffer (0.1 M PBS, 1% BSA, 0.05% Tween‐20, pH 7.4) for 5 min and dried overnight at room temperature. The conjugate pad was immersed in conjugate pad buffer (0.1 M PBS, 0.05% Tween‐20, 20% sucrose, 1% BSA, pH 7.4) and dried overnight; after drying, it was dipped in the optimized mixture of Cystatin‐C Ab1–AuNP and IgY–AuNP probes and dried at 37°C for 1 h. Finally, all components were assembled on an adhesive backing card with 2 mm overlapping between adjacent pads and cut into 3 mm wide strips using a programmable guillotine cutter. The strips were stored at 4°C in sealed aluminum foil pouches with desiccant until use.

2.6. Software Development

A custom Android application (“Smartphone rapid detection software for Cystatin C”, version 1.0) was developed to control the smartphone‐based reader and perform image analysis. This software is registered with the China National Copyright Administration (Registration No. 2025SR1844772). The application was developed following basic Android app development principles, utilizing Intents for Activity transitions and data transfer to enable user‐App interactions. Fragment and ViewPager components were employed for page reusability.

The application consists of two main modules. The first module is the image acquisition module, which uses the smartphone camera equipped with a macro lens and LED ring light to capture images of LFIA strips. A custom SurfaceView implementation provides real‐time camera preview for photography function, while an OverlayView class extending View uses Canvas to draw a transparent rectangular window on the preview, creating a constrained photography frame that guides users to align the test line within the designated imaging region. Dynamic permissions for camera and file read/write operations are requested at runtime following Android best practices, integrating with SurfaceView to achieve photography preview function.

The second module is the quantification module, which employs the OpenCV library (version 4.8.0, downloaded from https://opencv.org/releases/) for image processing. Captured images are first resized to 50 × 150 pixels using Imgproc.resize(), and then converted to grayscale using Imgproc.cvtColor() with COLOR_BGR2GRAY. The test line, control line, and background regions of interest (ROIs) are identified according to the fixed strip geometry within the reader. The average grayscale values for the test line (I _T ), control line (I _C ), and background (I _B ) are then calculated, and the T/C ratio is computed as:

$$R_{T/C}=\frac{I_B-I_T}{I_B-I_C}$$

A batch‐specific third‐order polynomial calibration model (y = ax ³ + bx ² + cx + d), where y denotes the Cystatin‐C concentration and × the background‐corrected T/C ratio, was embedded in the Android application to convert T/C ratios directly into concentrations. For each test‐strip batch, multi‐point calibration measurements were obtained across the calibration range, and coefficients a–d were re‐estimated and stored in the application as the calibration parameters for that batch. The smartphone application enables users to capture LFIA strip images and automatically obtain the corresponding Cystatin‐C values. Images with obvious positioning errors, severe blur, or incomplete strip insertion were excluded from quantitative analysis.

This batch‐specific calibration strategy transfers the nonlinear multi‐point calibration logic used in automated latex particle‐enhanced turbidimetric immunoassay (PETIA) systems to the smartphone‐based LFIA platform, while the polynomial coefficients for each batch are independently derived from the corresponding LFIA calibration data. The third‐order model accommodates the nonlinear colorimetric T/C response across the analytical range while remaining straightforward to implement in the Android application.

2.7. Assay Procedure

To initiate each test, a mixture of 30 μL serum and 80 μL running buffer (PBS, 0.1% Tween‐20, 1% BSA, pH 7.4) was prepared and deposited onto the sample pad of the LFIA strip. After the introduction of the sample, the LFIA strip was inserted into the smartphone reader so that the test line was aligned with the imaging area of the smartphone camera. Driven by capillary flow, Cystatin‐C in the sample bound to the Cystatin‐C Ab1–AuNP detection probes on the conjugate pad, forming Cystatin‐C Ab1‐AuNP‐Cystatin‐C complexes. These immunocomplexes were specifically captured at the test line by immobilized anti‐Cystatin‐C antibody (Ab2), yielding a visible red band with intensity proportional to Cystatin‐C concentration. Unbound Cystatin‐C Ab1–AuNP probes and IgY‐AuNP continued to migrate, with IgY‐AuNP captured by goat anti‐chicken IgY at the control line, confirming proper flow. After 15 min of development, the strip was imaged under consistent illumination using the smartphone reader. The custom application analyzed the test line intensity and calculated the Cystatin‐C concentration using the batch‐specific third‐order polynomial calibration model applicable to that test‐strip batch. A schematic of the assay workflow and image analysis is presented in Figure 1.

FIGURE 1.

Smartphone‐based LFIA workflow and image quantification. (a) LFIA sandwich assay principle. (b) Serum application, strip development, and smartphone‐reader workflow. (c) Representative smartphone capture and real application interface screens. (d) Fixed‐geometry ROI selection, background‐corrected T/C calculation, batch‐specific third‐order polynomial calibration, and Cystatin‐C concentration output.

2.8. Optimization Procedures

2.8.1. Capture Antibody Concentration Optimization

Capture antibody (Cystatin‐C Ab2) was diluted to 0.4, 0.8, 1.2, and 1.6 mg/mL in 10 mM PBS (pH 8.75) containing 1% sucrose. Each solution was dispensed onto nitrocellulose membranes at 1 μL/cm using an HM3035 dispenser. Membranes were dried at 37°C for 2 h. Test strips were assembled and evaluated with Cystatin‐C standards at 0.5, 2.5, and 8.0 mg/L (n = 3 per concentration). After 15 min development, strips were imaged using the smartphone‐based reader, and T/C ratios were calculated as described in Section 2.6.

2.8.2. Detection Antibody Adsorption Amount Optimization

Cystatin‐C Ab1 at quantities of 50, 100, 150, and 200 μg was added to separate 10 mL aliquots of AuNP solution (OD = 1, pH 8.75). Physical adsorption, blocking, washing, and storage were performed as described in Section 2.3. Probe‐loaded pads were prepared with each antibody amount and assembled into test strips. Evaluation with Cystatin‐C standards (0.5, 2.5, and 8.0 mg/L) and T/C ratio calculation were performed as above (n = 3 per concentration).

2.8.3. BSA Concentration in the Running Buffer Optimization

Running buffers containing 0.5%, 1%, 2%, 3%, and 5% BSA in PBS with 0.1% Tween‐20 (pH 7.4) were prepared. Test strips (n = 5 per condition) were tested with blank (0 mg/L) and 2.5 mg/L Cystatin‐C standards. T/C ratios were determined after 15 min development.

2.8.4. Tween‐20 Concentration in the Running Buffer Optimization

Running buffers containing 0.01%, 0.05%, 0.1%, 0.2%, and 0.5% Tween‐20 (with 1% BSA in PBS, pH 7.4) were prepared. Test strips (n = 10 per condition) were tested with 2.5 mg/L Cystatin‐C standard. T/C ratios and coefficients of variation (CV%) were calculated.

2.8.5. Assay Time Optimization

Test strips were tested with Cystatin‐C standards at 0.25, 0.5, 1.0, 2.0, 4.0, and 8.0 mg/L. Images were captured at 1‐min intervals from immediately after sample application (0 min) to 18 min (n = 3 per time point), and T/C ratios were plotted against time for each concentration.

2.9. Validation of the Developed LFIA

After establishing the assay design, a comprehensive analytical validation was conducted, encompassing evaluations of precision, linearity, dynamic range, limit of detection (LOD), limit of quantitation (LOQ), interference testing, comparison with a reference method, and accelerated stability assessment.

2.9.1. Precision

The precision of the Cystatin‐C measurement assay was evaluated in accordance with the CLSI EP15‐A2 guidelines [32]. Precision was evaluated across a range of Cystatin‐C concentrations using serum specimens at 0.75 (low), 3.00 (medium), and 6.00 (high) mg/L. Each specimen was tested 20 times in a single day to estimate within‐run precision (repeatability), while each specimen was tested a total of 12 separate days to estimate between‐run precision (intermediate precision). For each concentration level, the mean, standard deviation, and coefficient of variation (CV) were calculated.

2.9.2. Linearity and Sensitivity

To determine both the linearity of the assay performance and the sensitivity of the assay, Cystatin‐C standard solutions covering the expected clinical concentration range (0–8.00 mg/L) were generated and analyzed using the LFIA. The test data were analyzed in accordance with the CLSI EP6‐A guidelines [33]. A calibration curve was generated using the batch‐specific third‐order polynomial model applicable to the test‐strip batch and integrated in the smartphone application, and the quantitative working range of the dose–response relationship was identified. For sensitivity assessment, the LOD was defined as the mean blank signal plus three standard deviations, based on 10 replicate blank measurements, while the LOQ was determined as the lowest concentration yielding a coefficient of variation ≤ 20%, in accordance with CLSI EP17‐A2 guidelines [34].

2.9.3. Interference Testing

Following CLSI EP7‐A2 guidelines [35], potential interference from endogenous and exogenous substances was systematically examined. Moderate‐level Cystatin‐C samples (~2.50 mg/L) were spiked with common endogenous interferents, including bilirubin (0.23 mg/mL), triglycerides (12 mg/mL), hemoglobin (6 mg/mL), and rheumatoid factor (710 IU/mL), or with representative nephrotoxic drugs, including gentamicin sulfate (30 mg/L), norvancomycin hydrochloride (100 mg/L), and cisplatin (20 mg/L), and then analyzed using the LFIA. Results from spiked samples were compared against unspiked controls, with recoveries within ±15% of the control value defined as the absence of significant interference. Interferent‐only samples without added Cystatin‐C were also analyzed to check whether the interferents generated false‐positive signals.

2.9.4. Method Comparison

A total of 120 residual clinical serum specimens collected during routine diagnostic procedures were obtained from the Second Affiliated Hospital of Wenzhou Medical University, Wenzhou, China. The study was conducted in accordance with the Declaration of Helsinki and approved by the Ethics Committee of the Second Affiliated Hospital of Wenzhou Medical University (Approval No. 2026‐K‐51‐01; approved on 30 January 2026). The study was registered in the National Medical Research Registration and Filing Information System (registration No. MR‐33‐26‐014646; registered on 25 February 2026). Because the study used residual clinical specimens collected during routine diagnostic procedures and involved no additional procedures or direct contact with patients, the Ethics Committee granted a waiver of informed consent. The specimens used for analytical validation were selected to cover a broad clinically relevant range of Cystatin‐C concentrations. PETIA‐measured Cystatin‐C concentrations ranged from 0.45 to 7.97 mg/L. A method comparison between the developed LFIA and a commercial particle‐enhanced turbidimetric immunoassay (PETIA) reference method for Cystatin‐C measurement was then performed in accordance with CLSI EP9‐A2 guidelines [36]. The reference method employed a Cystatin C Assay Kit based on latex particle‐enhanced turbidimetric immunoassay (PETIA) from Zhejiang Kuaye Biotechnology Co. Ltd., China, and all measurements were carried out according to the manufacturer's instructions. The LFIA result for each sample was compared with the corresponding PETIA result to evaluate the agreement between the two methods. Correlation was assessed using linear regression (Pearson's R ²), and agreement was further evaluated with Bland–Altman analysis to determine the mean bias and 95% limits of agreement.

2.9.5. Accelerated Stability

Accelerated stability testing was performed to estimate the potential shelf‐life of the LFIA strips. Batches were stored at 21°C, 37°C, and 50°C for up to 28 days, with samples removed at defined time points (Weeks 0, 1, 2, 3, and 4) for evaluation. At each interval, strips were tested using Cystatin‐C–positive samples at three concentration levels (1.00, 2.50, and 5.00 mg/L), and the resulting T‐line signal intensities were compared with those obtained from fresh strips stored at 2°C–8°C. Stability was defined as acceptable when the signal recovery remained within 85%–115% of the initial value. The accelerated study was used to provide a preliminary assessment of strip robustness under stressed storage conditions; real‐time stability studies are required to establish the shelf‐life under routine ambient storage conditions.

2.10. Application of the Developed LFIA to Clinical Serum Samples

The developed smartphone‐based LFIA was applied to the determination of Cystatin‐C in human serum samples. For each sample, 30 μL of serum was mixed with 80 μL of running buffer and applied to the sample pad of the LFIA strip. After 15 min of development, the strip was inserted into the smartphone reader, and the Cystatin‐C concentration was automatically calculated by the custom application using the batch‐specific calibration model applicable to that test‐strip batch. Each serum sample was also analyzed using the reference particle‐enhanced turbidimetric immunoassay (PETIA) for method comparison. All measurements were performed according to the optimized assay conditions described above.

2.11. Statistical Analysis

Statistical evaluation of the obtained results was performed using the Statistica package (13.3). Quantitative data are presented as mean ± standard deviation (SD). Linear regression analysis was used to evaluate the relationship between the smartphone‐based LFIA and the reference PETIA method. Method agreement was further assessed using Bland–Altman analysis, and the mean bias with 95% limits of agreement was calculated. A Wilcoxon matched‐pairs signed‐rank test was used to assess whether paired measurements differed significantly between the two methods. A p value < 0.05 was considered statistically significant.

3. Results

3.1. Optimization of Assay Parameters

To establish the assay conditions that yielded optimal analytical performance, the principal parameters affecting the smartphone‐based lateral flow immunoassay were systematically optimized using the T/C ratio as the response metric. The T/C ratio was selected because it minimizes the influence of background fluctuation and inter‐strip variation, thereby providing a more robust basis for condition screening. Following the optimization strategy commonly adopted in lateral flow assay development, the concentration of capture antibody on the test line, the adsorption amount of detection antibody on AuNPs, the BSA content in the running buffer, the Tween‐20 concentration in the running buffer, and the assay development time were sequentially evaluated. The optimized conditions are summarized in Figure 2, Tables 1 and 2.

FIGURE 2.

Optimization of capture‐antibody concentration and detection‐antibody adsorption amount for the smartphone‐based LFIA. (a) T/C ratios at four Cystatin‐C Ab2 concentrations (0.4, 0.8, 1.2, and 1.6 mg/mL). (b) T/C ratios obtained with Cystatin‐C Ab1 adsorption amounts of 50, 100, 150, and 200 μg per 10 mL AuNP solution. Cystatin‐C standards were 0.5, 2.5, and 8.0 mg/L. Data are mean ± SD (n = 3).

TABLE 1.

Effect of BSA concentration on T/C ratio.

BSA concentration (%)	Blank sample (0 mg/L)	2.5 mg/L standard
0.5	0.05 ± 0.02	0.48 ± 0.04
1.0	0.02 ± 0.01	0.62 ± 0.03
2.0	0.02 ± 0.01	0.55 ± 0.03
3.0	0.01 ± 0.01	0.47 ± 0.04
5.0	0.01 ± 0.01	0.37 ± 0.05

TABLE 2.

Effect of Tween‐20 concentration on T/C ratio and reproducibility.

Tween‐20 concentration (%)	T/C ratio (0.5 mg/L)	CV% (n = 10)
0.01	0.12 ± 0.02	15.4
0.05	0.15 ± 0.02	9.7
0.10	0.18 ± 0.01	5.2
0.20	0.16 ± 0.01	8.2
0.50	0.13 ± 0.02	12.1

3.1.1. Capture Antibody Concentration on the Test Line

The concentration of capture antibody immobilized on the nitrocellulose membrane plays a critical role in determining assay sensitivity. As shown in Figure 2A, the T/C ratio increased progressively as the concentration of Cystatin‐C Ab2 increased from 0.4 to 1.2 mg/mL for all tested standards. When the concentration was further increased to 1.6 mg/mL, no further significant enhancement in signal was observed. Accordingly, 1.2 mg/mL was selected as the optimal capture antibody concentration, as it provided strong signal intensity without unnecessary reagent consumption.

3.1.2. Detection Antibody Adsorption Amount

The amount of detection antibody adsorbed onto AuNPs was subsequently optimized because it directly affects probe activity and signal generation. As shown in Figure 2B, the T/C ratio increased markedly as the amount of Cystatin‐C Ab1 increased from 50 to 150 μg per 10 mL AuNP solution. The highest responses were obtained at 150 μg, whereas a decline in signal was observed when the adsorption amount was further increased to 200 μg. For the low‐concentration standard (0.5 mg/L), the 150 μg condition also produced the strongest response, indicating improved sensitivity in the low‐concentration range. Therefore, 150 μg per 10 mL AuNP solution was selected as the optimal adsorption amount for subsequent experiments. Functional formation of the AuNP‐labeled antibody probe was supported by the dose‐dependent responses shown in Figure 2B.

3.1.3. BSA Concentration in the Running Buffer

The concentration of BSA in the running buffer was optimized to achieve an appropriate balance between suppression of nonspecific binding and maintenance of specific signal generation. As summarized in Table 1, the highest T/C ratio for the 2.5 mg/L Cystatin‐C standard was obtained at 1% BSA (0.62 ± 0.03), while the blank sample remained low (0.02 ± 0.01). At 0.5% BSA, the blank T/C ratio increased to 0.05 ± 0.02, suggesting insufficient blocking. In contrast, further increasing the BSA concentration to 2%–5% resulted in a progressive decline in T/C ratio for the positive sample. Based on these results, 1% BSA was selected as the optimal concentration in the running buffer. Thus, although 0.5% BSA retained a positive signal, 1% BSA provided a better signal‐to‐background balance.

3.1.4. Tween‐20 Concentration in the Running Buffer

Tween‐20 concentration was further optimized because surfactant content strongly affects sample wetting, flow uniformity, and assay reproducibility. As summarized in Table 2, the highest T/C ratio for the 0.5 mg/L standard was obtained at 0.10% Tween‐20 (0.18 ± 0.01), which also yielded the lowest coefficient of variation (5.2%). At 0.01% Tween‐20, the T/C ratio was lower and the CV was markedly higher, suggesting poor wetting and less uniform migration. At concentrations above 0.10%, the T/C ratio gradually decreased. Therefore, 0.10% Tween‐20 was selected as the optimal surfactant concentration for the running buffer.

3.1.5. Optimal Assay Time

The development of the test line signal was monitored over time (0–18 min) across a range of Cystatin‐C concentrations. For high concentrations (> 2.00 mg/L), the test line became visible within 8 min. Signal intensity increased progressively and reached a plateau before 15 min for all concentrations tested, with no significant change observed thereafter. Therefore, 15 min was established as the optimal assay time, ensuring complete sample migration and signal stabilization while maintaining rapid analysis (Figure 3).

FIGURE 3.

Assay‐time optimization across six Cystatin‐C concentrations. (a) 8.00, 4.00, and 2.00 mg/L. (b) 1.00, 0.50, and 0.25 mg/L. Data are mean ± SD.

3.2. Method Validation

3.2.1. Precision of the LFIA

The precision of the smartphone‐based LFIA was evaluated at three clinically relevant Cystatin‐C concentrations (low: ~0.75 mg/L, medium: ~3.00 mg/L, high: ~6.00 mg/L). The results, summarized in Table 3, demonstrated excellent repeatability and intermediate precision. The within‐run coefficients of variation (CVs, n = 20) were 7.40%, 5.12%, and 3.87% for the low, medium, and high levels, respectively. The between‐run CVs (n = 12) were 9.56%, 6.79%, and 5.02%, respectively. All CV values were well below the 15% acceptance threshold, confirming the high reproducibility of the assay. The raw data for within‐run and between‐run precision are provided in Tables S1 and S2.

TABLE 3.

Reproducibility of the smartphone‐based LFIA for Cystatin‐C.

Cystatin‐C levels (mg/L)	Within‐run (n = 20)		Between‐run (n = 12)
	Average (mg/L)	CV (%)	Average (mg/L)	CV (%)
High (~6.00)	6.22	3.87	6.19	5.02
Medium (~3.00)	3.03	5.12	2.94	6.79
Low (~0.75)	0.74	7.40	0.78	9.56

3.2.2. Linearity, LOD, and LOQ

Cystatin‐C standard solutions were used to establish the quantitative capability of the assay. The smartphone application provided the calibration curve using the embedded batch‐specific third‐order polynomial model, which provided a dynamic range from 0.32 to 8.00 mg/L. The linear regression analysis of the comparison between measured and expected values yielded a correlation coefficient (R ²) for the Cystatin‐C standard concentration versus raw data concentration of 0.9994, demonstrating excellent linearity for the smartphone‐based LFIA (Figure 4). The LOD was calculated based on the mean of the 10 blank samples +3 SDs to be 0.15 mg/L. The limit of quantification (LOQ), defined as the lowest concentration measurable with a CV ≤ 20%, was established at 0.32 mg/L, where the observed CV was approximately 10%.

FIGURE 4.

Calibration curve for the developed LFIA using serum calibrators. Each concentration was measured in triplicate; data are mean ± SD (n = 3), and error bars represent SD.

3.2.3. Interference Testing

Potential interference from common endogenous substances and three representative nephrotoxic drugs was evaluated. Using the unspiked control (2.57 ± 0.14 mg/L) as the reference, recoveries for the seven interferent‐spiked samples ranged from 85% to 112%, all within the predefined 85%–115% acceptance interval (Table 4). At a nominal Cystatin‐C concentration of 2.50 mg/L, gentamicin sulfate (30 mg/L), norvancomycin hydrochloride (100 mg/L), and cisplatin (20 mg/L) yielded LFIA results of 2.56 ± 0.10, 2.49 ± 0.14, and 2.58 ± 0.15 mg/L, corresponding to recoveries of 100%, 97%, and 100%, respectively. Interferent‐only responses were below the 0.15 mg/L LOD for all seven tested substances (Table 4), indicating no measurable false‐positive signal under the tested conditions. The raw data for interference testing are provided in Table S3.

TABLE 4.

Effects of common endogenous interferents and representative nephrotoxic drugs on Cystatin‐C measurement by the smartphone‐based LFIA.

Interferent	Concentration of interferent	Cystatin‐C (mg/L)	LFIA assay result (mg/L)	Recovery (%)	Interferent‐only LFIA response (mg/L)
Control	—	2.50	2.57 ± 0.14	103	Not applicable
Bilirubin	0.23 mg/mL	2.50	2.86 ± 0.15	111	< 0.15 (< LOD)
Triglycerides	12 mg/mL	2.50	2.87 ± 0.16	112	< 0.15 (< LOD)
Hemoglobin	6 mg/mL	2.50	2.87 ± 0.11	112	< 0.15 (< LOD)
Rheumatoid Factor	710 IU/mL	2.50	2.19 ± 0.13	85	< 0.15 (< LOD)
Gentamicin sulfate	30 mg/L	2.50	2.56 ± 0.10	100	< 0.15 (< LOD)
Norvancomycin hydrochloride	100 mg/L	2.50	2.49 ± 0.14	97	< 0.15 (< LOD)
Cisplatin	20 mg/L	2.50	2.58 ± 0.15	100	< 0.15 (< LOD)

Note: For interferent‐spiked samples, recoveries were calculated relative to the unspiked control; the control‐row recovery (103%) was calculated relative to the nominal Cystatin‐C concentration of 2.50 mg/L. Interferent‐only samples contained the stated interferent without added Cystatin‐C; responses below 0.15 mg/L are reported as< LOD.

3.2.4. Accelerated Stability Testing

The stability of LFIA strips was assessed at 21°C, 37°C, and 50°C for an extended period of 28 days. The results of the accelerated stability testing show that the test signal values for low (1.00 mg/L), medium (2.50 mg/L), and high (5.00 mg/L) concentration results remained between 85% to 115% of their initial values after being stored for 28 days at each temperature (Figure 5). Therefore, there was no significant deterioration of analytical performance for the LFIA strips during the accelerated storage testing. These findings strongly suggest that LFIA strips have good storage stability. However, to determine product shelf life and confirm long‐term stability under normal handling/storing conditions, real‐time stability studies are necessary. A long‐term 2°C–8°C refrigerated‐control study was not included and should be added in future shelf‐life validation.

FIGURE 5.

Accelerated stability testing of LFIA strips. Color region indicates data within 85%–115% across levels.

3.3. Application of the Method and Statistical Analysis of the Results

The Cystatin‐C measurement performance of the smartphone‐based LFIA was evaluated using 120 clinical serum samples spanning a broad range of Cystatin‐C concentrations. The results were compared with those obtained using the commercially available PETIA reference method to assess the practical utility of the proposed method under clinical conditions.

Each serum sample was tested using the optimized lateral flow immunoassay (LFIA) method. The sample concentration of Cystatin‐C was automatically calculated with respect to the calibration curve by a custom smartphone application. All samples were also tested with the manufacturer‐authorized PETIA method. After both analytical techniques were complete, paired results were compared statistically through linear regression analysis.

Linear regression analysis indicated a very strong correlation between LFIA and PETIA results. As illustrated in Figure 6, with PETIA used as X and LFIA as Y, the regression equation was Y = 0.9889X + 0.0385, with an R ² value of 0.9946. This indicates a very strong linear association between the LFIA and PETIA methods across the evaluated Cystatin‐C concentration range.

FIGURE 6.

Linear correlation between Cystatin‐C concentrations measured by the smartphone‐based LFIA and the PETIA reference method using 120 human serum samples. PETIA measurements are plotted on the x‐axis and LFIA measurements on the y‐axis. The solid line represents the linear regression fit (Y = 0.9889X + 0.0385, R ² = 0.9946), indicating a very strong linear association across the evaluated concentration range.

To further assess method agreement, Bland–Altman analysis was performed (Figure 7). The mean bias, calculated as PETIA minus LFIA, was −0.01 mg/L, and the 95% limits of agreement ranged from −0.34 to 0.32 mg/L. The percentage‐difference distribution relative to PETIA across the measured concentration range is shown in Figure 7b. The paired measurements did not differ significantly between the two methods (Wilcoxon matched‐pairs signed‐rank test, p = 0.2961). Together, these analyses support close agreement between the two methods within the evaluated clinical interval.

FIGURE 7.

Agreement between LFIA and PETIA measurements in 120 human serum samples. (a) Bland–Altman differences (PETIA − LFIA) plotted against the method mean, showing the mean bias (solid line) and 95% limits of agreement (dashed lines); mean bias, −0.01 mg/L; limits, −0.34 to 0.32 mg/L. (b) Percentage difference relative to PETIA across the measured Cystatin‐C range (0.45–7.97 mg/L).

4. Discussion

Cystatin‐C has been identified as an accurate and sensitive biomarker for assessment of renal function and to identify early indications of renal impairment due to its relative independence to muscle mass, age, and dietary factors relative to creatinine‐based markers. Accurate and timely cystatin C measurements are therefore essential for screening and risk stratification in patients with early chronic kidney disease (CKD) [6, 7, 8, 9]. However, traditional methods for quantifying Cystatin‐C have depended primarily on automated immunoassays conducted in centralized laboratories (particle‐enhanced turbidimetric and nephelometric methods) that require trained personnel, automated analyzers, and extensive turnaround times for completion [15, 16, 17, 18], which severely limits accessibility to these immunoassays in community healthcare settings, emergency rooms, and resource‐poor environments.

In this investigation, we developed a smartphone‐based LFIA for quantitative analysis of serum Cystatin‐C levels. A key methodological feature of the platform is the transfer of the nonlinear multi‐point calibration logic used in automated latex particle‐enhanced turbidimetric immunoassay (PETIA) systems to smartphone‐based LFIA. By re‐estimating the third‐order polynomial coefficients for each test‐strip batch and storing the resulting batch‐specific model in the Android application, the system directly converts background‐corrected T/C ratios into concentrations. Results support the conclusion that the proposed method had comparable analytical performance to laboratory‐based immunoassays but better characteristics with regards to speed, portability, and simplicity of use. By integrating a conventional colorimetric LFIA strip with a customized smartphone reader and a dedicated image analysis algorithm, the platform effectively bridges the gap between centralized laboratory testing and decentralized point‐of‐care diagnostics [21, 22, 23, 29, 37].

The analytical sensitivity and dynamic range of the assay analysis meet clinical requirements for quantifying renal function. The limit of detection is 0.15 mg/L, which can reliably identify CKD below the normal reference range in healthy adults [38]. Furthermore, the assay covers a span of 0.32–8.00 mg/L and eliminates or reduces the need for dilution of samples, simplifying routine patient management procedures. These performance characteristics support the suitability of the assay for both screening and follow‐up applications in clinical practice.

Interference testing further demonstrated recoveries of 85%–112% for four endogenous interferents and three representative nephrotoxic drugs at the evaluated challenge concentrations. Gentamicin sulfate, norvancomycin hydrochloride, and cisplatin yielded recoveries of 100%, 97%, and 100%, respectively, and the corresponding interferent‐only samples generated responses below the LOD. These findings support assay robustness under the tested in vitro conditions but should not be generalized to other nephrotoxic agents, metabolites, concentrations, or clinical exposure scenarios.

In the method comparison involving 120 clinical serum samples, the smartphone‐based LFIA showed excellent agreement with the commercial PETIA reference method. Linear regression demonstrated a near‐unity slope (Y = 0.9889X + 0.0385, where X denotes PETIA and Y denotes LFIA) and a high coefficient of determination (R ² = 0.9946). Bland–Altman analysis yielded a mean bias of −0.01 mg/L and 95% limits of agreement from −0.34 to 0.32 mg/L. The Wilcoxon matched‐pairs signed‐rank test showed no statistically significant difference between paired measurements (p = 0.2961), further supporting close agreement across the evaluated concentration range [36]. These findings illustrate that the LFIA is comparable to established laboratory methods, suggesting potential for decentralized analysis of Cystatin‐C. However, broader multicenter validation is still warranted before routine clinical implementation.

Comparative performance is summarized in Table 5. The present study suggests that the smartphone‐based colorimetric LFIA has an excellent balance of sensitivity, dynamic range, and practical accessibility. The limit of detection (0.15 mg/L) is lower than that of Zhang et al. [0.98 mg/L] and comparable with that of Bikkarolla et al. (0.18 mg/L) [22, 38]. A major distinguishing factor of this work is that the image analysis system is integrated into the smartphone, therefore eliminating subjectivity in measurement interpretation compared to previous LFIA studies [22, 38, 39].

TABLE 5.

Performance comparison of cystatin‐C detection methods between this work and literature.

Method	LOD (mg/L)	LOQ (mg/L)	Linear range (mg/L)	Reader/Imaging device
Smartphone‐based colorimetric LFIA (This work)	0.15	0.32	0.32–8.00	Smartphone camera (with macro lens)
Colorimetric LFIA (Zhang et al. [22])	0.98	—	0.1–2.5	Visual color intensity (no smartphone)
Fluorescent LFIA (Natarajan et al. [29])	0.023	0.029	0.023–32	Fluorescence reader required
Colorimetric LFIA (Bikkarolla et al. [38])	0.18	0.39	0.39–7.50	Visual (mobile phone optional)

Recent LFIA developments have used lanthanide reporters [37], two‐dimensional magnetic fluorescent nanofilms [40], and magnetic nanozymes [41, 42] to improve sensitivity, target breadth, or multiplexing. Although these approaches offer strong analytical gains, they generally add specialized optical readout, magnetic‐enrichment or catalytic‐development steps, or more complex probe fabrication. In contrast, the present work prioritizes quantitative colorimetric readout from a conventional AuNP strip using a low‐cost smartphone reader.

Advanced platforms such as the fluorescent LFIA described by Natarajan et al. [29] may be superior, with limits of detection in the range of 0.023 mg/L and extended linear range; however, have limited scalability and routine application in decentralized and/or limited‐resource situations due to the equipment costs of fluorescence readers [26]. In contrast, the present study used colorimetric gold nanoparticles for labeling in order to maintain simplicity and reduce costs. The limit of detection is also clinically relevant for CKD screening and routine measurements of renal function.

Potential sources of variability in measurement results exist with colorimetric LFIA systems. Variability in the equalization of membrane properties, efficiency of antibody immobilization, stability of AuNP‐labeled antibody probes, and sample matrix effects (e.g., hemolysis or lipemia) may introduce variability in measurement results [21, 22, 23]. Standardized strip production and consistent illumination and imaging through specialized smartphone readers will minimize these sources of variability. The batch‐specific third‐order polynomial calibration model, along with standardized acquisition of images and analysis of grayscale values using defined regions of interest, will minimize variability in measurement associated with uneven color distribution along the test line and minimize interpretation errors by users in measurement results [28, 30]. However, improper sample selection or handling will produce variability in measurement results, thus highlighting the need for standardized operating procedures for future clinical use.

From a methodological and engineering standpoint, the proposed smartphone‐based LFIA represents an equilibrium between analytical performance and system complexity. As compared to fluorescence or lanthanide‐based quantitative LFIA systems that are dependent on specialized optical instrumentation [29, 37, 40, 41, 42], the present method retains most of the inherent advantages of gold nanoparticle‐based methods, such as low cost, robustness, or ease of manufacture, while producing reliable quantitative output from smartphone‐assisted evaluation of images. Such a balance will facilitate greater implementation on a large scale and provide viability for decentralized applications in community healthcare.

Future research should focus on validating the current LFIA with specific groups of patients. In particular, further assessment of feasibility using patients suffering from chronic kidney disease, with diabetic nephropathy or the elderly, is warranted because small changes in Cystatin‐C levels have significant prognostic value [6, 7, 8, 9, 13, 14]. Longitudinal studies that look at disease progression and response to therapy would further enhance the clinical utility of the LFIA. Additionally, integration of the LFIA with telemedicine platforms and electronic records will facilitate remote monitoring of Cystatin‐C levels, data‐driven clinical decision making for patients, and the broader use of decentralized testing for renal function. Such use would require validated serum‐preparation workflows and device‐standardized image acquisition before home‐based implementation.

This study has multiple limitations even though it has shown promise with the analytical comparison and clinical comparison of the LFIA system. The first limitation is that this study was done using samples from one site and should be confirmed with a larger number of sites before broad clinical implementation of the method. Secondly, this study did not examine the performance consistency of the smartphone system across multiple smartphone brands/models or operators. Ambient light and imaging angle were controlled by the reader design: the macro lens was fitted directly against the test‐card imaging window, and the built‐in LED provided constant illumination within the fixed imaging geometry. The third limitation is that the estimated shelf life of the LFIA device was based upon accelerated stability testing of gold nanoparticles and has yet to be confirmed with long‐term real‐time studies. All three of these considerations need to be elucidated in future studies before wider use of the LFIA technology can occur in a clinical setting. In addition, ζ‐potential and antibody adsorption/loading efficiency were not evaluated. Although no significant analytical interference was observed for gentamicin sulfate, norvancomycin hydrochloride, or cisplatin at the tested concentrations, only three representative nephrotoxic drugs were examined; additional drugs should be evaluated in future studies.

5. Conclusions

In this study, a smartphone‐based quantitative colorimetric lateral flow immunoassay was successfully developed and validated for the determination of Cystatin‐C in human serum. By integrating a conventional gold nanoparticle‐based LFIA strip with a customized smartphone reader and dedicated image analysis software, the proposed platform enables reliable quantitative analysis while maintaining the inherent advantages of simplicity, low cost, and rapid operation.

The results show that the LFIA had a low limit of detection, a broad quantitative range, good precision, no significant interference from the tested endogenous substances or representative nephrotoxic drugs at the evaluated concentrations, and close agreement with the commercial PETIA method in serum specimens. These results for the smartphone‐enhanced LFIA provide accurate and reproducible Cystatin‐C measurements with the operational advantages of ease of use, low‐cost testing, and rapid analysis. Overall, this work highlights the potential of smartphone‐assisted colorimetric LFIA as a practical platform for decentralized and point‐of‐care Cystatin‐C testing. Further multicenter clinical validation and real‐time stability studies are warranted before routine clinical implementation.

Author Contributions

Ye Lian and Ruiying Zheng: conceptualization. Ye Lian, Ruiying Zheng, and Chenqi Yang: methodology. Ye Lian, Chenqi Yang, and Yueyang Luo: software. Ye Lian, Ruiying Zheng, and Na Zhang: validation. Na Zhang and Yueyang Luo: formal analysis. Ye Lian, Ruiying Zheng and Chenqi Yang: investigation. Guojun Lian and Baoqing Li: resources. Ye Lian and Ruiying Zheng: data curation. Ye Lian and Ruiying Zheng: writing – original draft preparation. Guojun Lian, Baoqing Li and Na Zhang: writing – review and editing. Guojun Lian and Baoqing Li: supervision. Guojun Lian and Baoqing Li: funding acquisition. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by the National College Students Innovation and Entrepreneurship Training Program, grant number 202516405001S.

Ethics Statement

The study was conducted in accordance with the Declaration of Helsinki and approved by the Ethics Committee of the Second Affiliated Hospital of Wenzhou Medical University (Approval No. 2026‐K‐51‐01; approved on 30 January 2026). The study was registered in the National Medical Research Registration and Filing Information System (registration No. MR‐33‐26‐014646; registered on 25 February 2026).

Consent

The requirement for informed consent was waived by the Ethics Committee because the study used residual clinical specimens collected during routine diagnostic procedures and involved no additional procedures or direct contact with patients.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: Raw data for the coefficient of variation for the within‐run assay.

Table S2: Raw data for the coefficient of variation for the between‐run assay.

Table S3: Raw data for interference testing.

Data Availability Statement

The data presented in this study are available on request from the corresponding author.