Microplastic contamination in commercial and traditional dairy products: occurrence, characteristics, and potential risk

Abstract

This study investigates the occurrence, characteristics, and potential health risks of microplastics (MPs) in fifteen widely consumed dairy products from Bangladesh, including industrial and traditional items. Abundances (MPs/kg) ranged from 1643 ± 94 in strawberry yogurt to 5143 ± 120 in mango milk; among solids, industrial yogurt (4616 ± 103) and milk powder (4159 ± 86) were the most contaminated. Fibers dominated (78–90 %), and polyethylene (68–73 %) plus polypropylene (17–20 %) together accounted for over 88 % of polymers. The diversity integrated index (DII) revealed higher heterogeneity in liquids (up to 0.56) than in solids (as low as 0.32). Hierarchical clustering separated industrial products from flavored and traditional ones, and principal component analysis distinguished packaging-derived polymers from process-related polymers. Contamination factor, pollution load index, and Nemerow pollution index classifications indicated moderate to high pollution in most samples. Polymeric hazard index (8.6–19.3) reflected variation in toxic polymer content. Estimated daily intakes varied significantly—children from 2.60 to 75.08 MP/kg.day and adults from 0.78 to 22.52 MP/kg.day—highlighting age-dependent exposure. These findings underscore the pervasive presence of MPs in dairy products, driven by packaging and processing, and call for standardized mitigation strategies to safeguard consumer health.

Graphical abstract

Highlights

• •MP levels ranged from 1643 ± 94 to 5143 ± 120 items/kg in dairy products.
• •Fiber-shaped MPs dominate (78–90 %) in dairy products.
• •Polyethylene and polypropylene make up > 88 % of all MPs.
• •Children's daily dairy intake: up to 75 MP/kg·day.
• •Packaging materials and processing equipment are key MPs sources.

1. Introduction

Over the past few decades, the global proliferation of plastic production and use has led to an unprecedented accumulation of plastic debris in virtually every natural ecosystem, where larger items fragment into microplastics (MPs)—particles smaller than 5 mm—that have now been detected in marine, freshwater, terrestrial, and atmospheric compartments (Basaran et al., 2023; Chakraborty et al., 2024; Lusher et al., 2017; SAPEA, 2018). Initially, research on MPs focused primarily on marine and aquatic environments; however, recent investigations have broadened the scope to include various food matrices, revealing that MPs are not confined to natural ecosystems but also permeate the human food chain (Van Cauwenberghe and Janssen, 2014; Visentin et al., 2024). Among these food groups, dairy products are of particular interest because they are not only a vital source of essential nutrients such as proteins, fats, vitamins, and minerals but are also consumed widely across diverse populations worldwide.

In dairy production, numerous processing steps and materials provide potential avenues for MPs contamination. Dairy commodities such as milk, yogurt, cheese, and butter typically undergo a series of stages—including pasteurization, homogenization, filtration, and packaging—in which contact with plastic-based equipment is inevitable (Visentin et al., 2024; Kutralam-Muniasamy et al., 2020). In many processing facilities, the machinery and components used are manufactured from plastic materials that, under conditions of mechanical wear or chemical degradation, may shed MPs fragments. Furthermore, packaging—in the form of Tetra Pak cartons, high-density polyethylene (HDPE) bottles, and plastic films—is ubiquitous within the dairy industry (Basaran et al., 2023; Kutralam-Muniasamy et al., 2020; Rbaibi Zipak et al., 2024). Not only do these packaging materials have the potential to introduce MPs during the filling and sealing processes, but they may also serve as continuous sources of contamination over the product's shelf life. MPs act not only as physical contaminants but also as carriers of chemical additives like phthalates and bisphenols, which can leach from plastics into dairy products (Santonicola et al., 2019). In addition, the use of contaminated water systems, whether that water is employed for processing, cleaning, or even in livestock husbandry, further exacerbates the risk of MPs ingress into final dairy products. Studies by Diaz-Basantes et al. (2020) and Chakraborty et al. (2024) have documented the occurrence of MPs in various forms of milk, suggesting that contamination may occur at multiple critical points along the dairy supply chain—from production and processing to packaging and storage.

The complexity of the dairy matrix itself adds another layer of difficulty to both the detection and quantification of MPs (Hasan et al., 2022; Dobrzański et al., 2005). Dairy products are inherently heterogeneous, comprising a blend of fats, proteins, carbohydrates, and minerals. These components can interact with MPs particles in ways that may mask or alter their detectable characteristics, thereby complicating efforts to isolate and accurately quantify the contamination (Visentin et al., 2024; Kutralam-Muniasamy et al., 2020). Moreover, the absence of standardized methodologies validated specifically for dairy matrices has resulted in significant variability among reported MPs concentrations. For example, while Buyukunal et al. (2023) reported the detection of MPs in yogurt samples and Zhang et al. (2023) identified their presence in milk powder, differences in sample preparation, detection limits, and identification techniques have led to discrepancies in the reported data.

MPs inherently absorb heavy metals, persistent organic pollutants (POPs), and endocrine-disrupting chemicals from their surroundings, and when ingested, these pollutant-laden MPs can mechanically irritate the gastrointestinal tract and facilitate the uptake of toxic compounds (Crawford and Quinn, 2017; SAPEA, 2018). Research by Cox et al. (2019) has demonstrated that MP particles, particularly those at the nanoscale, can penetrate the intestinal barrier and accumulate within internal tissues, thereby raising serious concerns about systemic exposure and chronic health effects. These effects may include the triggering of inflammatory responses, the generation of reactive oxygen species (ROS), and the onset of oxidative stress, which have been implicated in the development of chronic conditions such as cardiovascular disease, neurodegenerative disorders, and even certain types of cancer, as further discussed by Winiarska et al. (2024).

Given the extensive global consumption of dairy products by people across all age groups and the essential nutritional role these products play in human diets, understanding the implications of MPs contamination is of paramount importance for public health. Without comprehensive data on the presence, concentration, and characteristics of MPs in dairy products, consumers remain unaware of their potential exposure and associated risks. Despite the evident risks, current research efforts have predominantly concentrated on a limited range of dairy products—primarily liquid milk, yogurt, and milk powder—leaving many other forms of commonly consumed dairy items largely unexamined. Diaz-Basantes et al. (2020), Chakraborty et al. (2024), Buyukunal et al. (2023), Zhang et al. (2023), Cox et al. (2019), and Winiarska et al. (2024) have all contributed valuable evidence that the dairy supply chain is vulnerable to MP intrusion and that, if left unchecked, these contaminants could have serious implications for consumer health. In light of these challenges, the present study aims to investigate the prevalence and characteristics of MPs across a wide range of dairy products—including ultra-high temperature processing (UHT) milk, flavored milks (vanilla, chocolate, mango), various yogurts (strawberry, savory), buttermilk, milk powder, traditional curds (clay pot curd, cup curd), and locally make dairy-based sweets (Kalojam— a black sweet dumpling made from milk solids that is deep fried and soaked in sugar, and Roshogolla — spongy cottage-cheese dumpling soaked in sweet sugar syrup), as well as chocolate and vanilla ice cream. The study also seeks to assess the pollution load and polymer-related risks, and to estimate the oral ingestion of MPs through dairy product consumption among children and adults.

2. Materials and methods

2.1. Sample collection

To investigate the presence of MPs in dairy products, samples were chosen based on their widespread domestic usage and availability across various retail platforms, including supermarkets, local markets, and general stores in Bangladesh. Between November 2024 and January 2025, we collected fifteen dairy samples from a range of commercial and traditional outlets across Bangladesh's South-West region to capture diverse product sources. The collected samples were categorized into two primary groups: (a) liquid samples: UHT milk, vanilla milk, chocolate milk, mango milk, strawberry yogurt, savory yogurt, and buttermilk, and (b) solid samples: milk powder, yogurt, clay pot curd, cup curd, Kalojam, Roshogolla, chocolate ice cream, and vanilla ice cream. Given the potential influence of packaging on MP contamination, material composition was carefully documented (Table S1). To maintain sample integrity, all dairy products were stored at 4 ⁰C and carefully handled to ensure analysis before their expiration dates, thereby minimizing degradation and external contamination.

2.2. Sample preparation

To extract MPs from the dairy products, a hot alkaline digestion method was applied (Visentin et al., 2024). Briefly, 10 g of each sample was treated with 10 % KOH solution in a beaker, mixed with a sterilized glass rod to ensure uniform distribution, and allowed to react at room temperature for 24 h to break down organic components. The mixture was then digested on a hot plate at 55–60 ⁰C for 4–5 h to achieve sufficient decomposition for MPs isolation. However, for chocolate milkshake and chocolate ice cream samples, a density separation step after the digestion step was required to facilitate efficient MPs isolation. This process involved the use of an oversaturated sodium chloride (6 M NaCl) solution, which assisted in separating MPs from the organic matrix.

Following digestion, the heated dairy product samples were filtered through GF/B glass microfiber filters (nominal retention ∼1.0 μm) using a vacuum-pump filtration system. The filters containing extracted MPs were then dried in an oven at 40 ⁰C for 24 h to remove residual moisture. Finally, the dried filters were stored in glass Petri dishes to preserve sample integrity until further analysis.

2.3. MP identification

The extracted MPs were categorized into two size ranges: 0.1–5 mm and 0.01–0.1 mm, to enhance data justification, based on previous studies and available laboratory facilities. MP particles were visually identified under a microscope (OPTIKA, Italy) by adjusting the magnification range between 10x and 100x. They were then quantified based on their color, size, and shape.

To determine the chemical composition of the MPs, Attenuated Total Reflection (ATR) coupled with Fourier Transform Infrared Spectroscopy (FTIR) spectroscopy (PerkinElmer Spotlight 400) was used. The stored spectrum ranged from 4000 to 550 cm¹, and 16 co-scans were performed at a resolution of 4 cm¹ for each assessment. To prevent measurement errors, background cleansing was conducted before each measurement. Finally, the spectrum obtained through ATR-FTIR was matched (>70 % similarity) to the reference database to confirm the specific polymer type.

2.4. Safety and contamination control

All the procedures involving reagent filtration, reagent handling, glass washing, sample digestion, and sample filtration were carried out in a dedicated clean room. Deionized water used in preparing the reagent has been filtered through GF/B to remove MP contamination. The digestion protocol was set up, adapted from Visentin et al. (2024). Throughout the study, cotton clothing and nitrile gloves were worn to minimize contamination.

Blanks were prepared along with each sample and according to the same protocol, including digestion, filtration, and detection steps. In the analyzed blank filters, a maximum of 4–5 MPs contaminants were found. Blank filters were processed with each batch of samples, and blank correction was applied to individual samples by subtracting the exact number of MPs detected on the corresponding blank. The highest blank count was 5 MPs, while the lowest sample (strawberry yogurt), meaning the blank contributed only ∼0.30 % of the sample.

2.5. MP-induced pollution indices

To assess MP pollution in dairy samples, this study adopted several indicators such as contamination factor (CF), pollution load index (PLI), Nemerow pollution index (NPI), and polymer hazard index (PHI). These indices are instrumental in assessing potential health risks and guiding mitigation measures. CF is used to determine the pollution level in individual samples by comparing the MP concentration measured to a background value, which is presented in Eq. (1) (Binelli et al., 2025).

$$\text{CF}=\frac{\text{Apx}}{\text{Bpx}}\dots \dots \dots \dots$$ (1)

Where Apx is measured MP particles, and Bpx is the background value MP concentration; and CF values were classified as low (CF < 1), moderate (1 < CF < 3), considerable (3 < CF < 6), and very high (CF > 6).

Dairy products serve as the environmental media for MP pollution. To characterize the overall MP pollution load, first, the concentration factor for each sample is determined by Eq. (2), and then the PLI (Eq. (3)) is calculated based on the concentration factor (CFi) (Tomlinson et al., 1980).

$$\text{CFi}=\frac{Ci}{Co}\dots \dots \dots \dots$$ (2)

$$\text{PLI}=\sqrt{{\text{CF}}_{\mathrm{i}}}\dots \dots \dots \dots$$ (3)

Where Ci is the MP concentration in a sample and C₀ is the minimum average MP content in processed foods, 1.68 n/kg (Fadare et al., 2021). Criteria for the danger categories of MP pollution load, as defined by Xu et al. (2018) and outlined by Lin et al. (2022), are applied to interpret the PLI values. The PLI categorizes risk levels as follows: PLI >10 (Risk I), >20 (Risk II), >30 (Risk III), and >40 (Risk IV).

The NPI (Eq. (4)) uses the level of widespread contamination to determine the quality class. Since various toxicants may influence each sampling site, this method offers a rational explanation of the toxicants present at each sampling point taken together. The NPI provides an integrated measure of contamination using both the average CF and the maximum CF observed in the sample.

$$\text{NPI}=\sqrt{\frac{\mathrm{C}{\mathrm{F}}^2+\mathrm{C}{\mathrm{F}}_{\max }^2}{2}}\dots \dots \dots \dots$$ (4)

The average of the contamination factors for the MPs under study is called CF, whereas the maximum CF for an MP in a sample is called CF_max (Chakraborty et al., 2024; Proshad et al., 2022). The NPI indicates contamination levels as follows: NPI <0.7 (safe), 0.7–1.0 (warning), 1.0–2.0 (slightly polluted), 2.0–3.0 (moderately polluted), and >3.0 (highly polluted).

The PHI (Eq. (5)) evaluates the potential risk posed by different types of MPs polymers present in dairy samples.

$$\text{PHIx}=\mathrm{\Sigma }\frac{\text{Tpx}}{\text{Apx}}\times \text{Qy}\dots \dots \dots \dots$$ (5)

In this equation, Tpx represents the amount of a specific MP polymer in the sample, and Qy is the polymer risk score as defined by Lithner et al. (2011). The PHI categorizes as: PHI <10 (Low I), 10–100 (Moderate II), 101–1000 (Considerable III), and 1001–10000 (High IV).

To assess the risk associated with MP contamination in dairy products, a multilayer feed-forward artificial neural network (ANN) was employed as a predictive modeling tool. MP prevalence data—along with contextual variables—served as the input layer, while quantitative risk metrics (CF, PLI, and PHI) were defined as the output layer. The ANN was trained using MATLAB R2023a, implementing the Levenberg–Marquardt algorithm for input–output fitting, regression, and curve optimization (Chakraborty et al., 2024). Each neuron's output was computed using the activation function applied to the weighted sum of inputs plus a bias term (Eq. (6)):

$$a=f\sum _{i=1}^n\left(wixi+b\right)\dots \dots \dots \dots \dots \text{.}$$ (6)

Where a is the neuron's output, f is the activation function, wᵢ is the weight for input i, xᵢ is input i, and b is the bias.

To ensure comparability among heterogeneous features, input data were normalized using min–max scaling (Eq. (7)):

$$x^{\prime }=\frac{x-\min \left(x\right)}{\mathit{max}\left(x\right)-\mathit{min}\left(x\right)}\dots \dots \dots \dots \dots \dots \dots \dots$$ (7)

Model performance was evaluated using Mean Squared Error (MSE) and the coefficient of determination (R²) (Eqs. (8) and (9)). MSE quantified the average squared difference between predicted and observed values, while R² indicated the proportion of variance in the target explained by the model:

$$MSE=\frac{1}{n}\sum _{i=1}^n{\left(yᵢ-ŷᵢ\right)}^2\dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots \dots$$ (8)

$$R^2=1-\frac{{\sum }_{i=1}^n{\left(yᵢ-ŷᵢ\right)}^2}{{\sum }_{i=1}^n{\left(yᵢ-ỹ\right)}^2}\dots \dots \dots \dots \dots \dots \dots \dots \dots$$ (9)

Here, n is the number of observations, $yᵢ$ is the actual (target) value, $ŷᵢ$ is the predicted value, and $ỹ$ is the mean of actual values.

2.6. Estimated daily intake assessment

The dietary exposure to MPs was estimated by calculating the estimated daily intake (EDI in MPs/(kg.day)) for two population groups—children and adults— presented in Eq. (10) (Ling et al., 2024).

$$\text{EDI}=\frac{\mathrm{\rho }\times \mathrm{U}}{\text{BW}}\dots \dots \dots \dots$$ (10)

Here, MPs' concentration per processed food item is represented by ρ (MPs/kg). U is the rate of ingestion per person (kg/day). In this case, U stood for the recommended daily consumption of various processed foods. The average body weight (BW) is 50 kg for adults and 15 kg for children (Ghosh et al., 2020). According to Dairy Industries International (2020), milk is consumed at a rate of 0.219 kg/day/person. Sweet intake is reported as 0.0179 kg/day/person based on Helgi Library (2020) data. Dhar et al. (2021) noted that ice cream is consumed at 0.03004 kg/day/person.

2.7. Diversity index

The diversity of MPs’ physical characteristics, including size, shape, color, and polymers in the dairy products sample, was evaluated using the Simpson diversity index (SDI) (Eq. (11)) and the diversity integrated index (DII) (Eq. (12)). The SDI is calculated as:

$$\text{SDI}=1-\sum {\left(\frac{\text{Pi}}{\mathrm{P}}\right)}^2\dots \dots \dots \dots$$ (11)

Where Pi is the number of entities in the ith MP's property, such as shape, size, color, and polymer, and P is the total number of MPs. The DII integrates the diversity of all examined properties and is calculated as:

$$\text{DII}=\sqrt[4]{\text{Size}\text{SDI}\times \text{Shape}\text{SDI}\times \text{color}\text{SDI}\times \text{polymer}\text{SDI}}\dots \dots$$ (12)

The SDI and DII have a range of 0–1, where the assessment values near one indicate high diversity (McLaughlin et al., 2016).

2.8. Statistical data analysis

The Origin 2018 (Origin Lab Co., Northampton, MA, USA) was used to plot the figures, and Excel was used to do statistical analysis. Python libraries, including pandas, matplotlib, seaborn, and scipy, were used via PyCharm Community Edition 2024.3 to generate heatmaps with hierarchical cluster analysis. The Shapiro-Wilk test (p > 0.05) validated a normal distribution of the data. A one-way ANOVA was conducted to evaluate MP levels among dairy samples, with statistical significance established at p < 0.05.

3. Results and discussion

3.1. Abundance and characteristics (size, shape, color, and polymer) of MPs in dairy products

MP abundance varied across dairy products (Fig. 1A, Table S1). Among the liquid dairy samples, products in TeFtra Pak cartons generally display higher measured values, with mango milk leading at 5143 ± 120.28 MP/kg, followed by vanilla milk at 4333 ± 202.15 MP/kg, and chocolate milk at 2743 ± 159.37 MP/kg (Fig. 1A). UHT milk and strawberry yogurt show lower values of 2644 ± 157.78 and 1643 ± 94.16, respectively. In contrast, the HDPE bottled products—savory yogurt and buttermilk—show moderate abundance levels of 3416 ± 84.98 MP/kg and 1709 ± 55.58 MP/kg, respectively. Among the solid dairy samples, industrial yogurt packaged in plastic shows the highest abundance (4616 ± 102.74 MP/kg), closely followed by industrial milk powder in Tetra Pak cartons (4159 ± 85.76 MP/kg) (Fig. 1A). Locally produced cup curd in plastic tubs ranks next (3933 ± 81.65 MP/kg), while clay pot curd—housed in traditional clay vessels—yields a moderate abundance (2699 ± 47.14 MP/kg) (Fig. 1A). Traditional sweets—Roshogolla (2453 ± 58.87 MP/kg) and Kalojam (2186 ± 41.09 MP/kg)—fall in the mid-range, and industrial ice creams in plastic containers register the lowest abundances: chocolate (2306.33 ± 115.85 MP/kg) and vanilla (2033 ± 81.65 MP/kg).

Fig. 1.

MP's (A) abundance, (B)size, (C) shape, (D)color, and (E) polymer types in dairy product samples.

These levels align with 53 ± 15 to 430 ± 30 MP/kg in milk powder and 96 ± 2 to 246 ± 4 MP/L in liquid milk in Bangladesh (Chakraborty et al., 2024), 164–427 MP/L in Indian branded milk (Kiruba et al., 2022), 4145–6800 MP/L in Swiss milk (Da Costa Filho et al., 2021), 84–128 MP/L in Turkish raw milk (Rbaibi Zipak et al., 2024), and 10–100 MP/kg in Chinese powdered milk (Zhang et al., 2023). Contamination pathways span collection, processing, filtration, storage, transport, packaging, use of plastic farming tools, and airborne MPs from milking machinery (Kutralam-Muniasamy et al., 2020; Diaz-Basantes et al., 2020; Caramia and Guerriero, 2010; Lopes and Stamford, 1997), underscoring the urgency for targeted mitigation strategies. Airborne MPs contamination also acts as an additive (post-production) source for food and dairy, introducing particles—especially fibers—onto exposed products and processing surfaces during handling, storage, and retail display (Gürmeriç and Basaran, 2025). Fibers dominate many food and dairy MP profiles because textiles, wipes, and indoor dust continuously shed microfibers that remain suspended or are resuspended by human activity and air currents, facilitating deposition onto open vats, cutting boards, and aging rooms (Visentin et al., 2025). Mechanistically, fibers’ high aspect ratio and low settling velocity favor transport within indoor environments and increase the likelihood of entanglement with food surfaces and equipment, while clothing and textile handling in processing areas amplify airborne fiber loads (Gürmeriç and Basaran, 2025; Sarojini et al., 2025).

In liquid dairy products, 57 % of MPs measured 0.1–5 mm and 43 % < 0.1 mm, whereas solids showed 61 % and 39 %, respectively (Fig. 1B; F = 4.36, F-Crit = 4.19, P = 0.04). Literature reports dominant fragments at 20–250 μm and similar size distributions across milk matrices (Kiruba et al., 2022; Chakraborty et al., 2024; Kutralam-Muniasamy et al., 2020; Rbaibi Zipak et al., 2024; Basaran et al., 2023; Diaz-Basantes et al., 2020), while particles <10 μm pose oxidative-stress and gut–liver risks (Visalli et al., 2021). Fibers dominated both liquids (78 %) and solids (90 %), with lower proportions of films (16 %/8 %), pellets (4.8 %/1 %), and fragments (1.7 %/1 %) (Fig. 1C; F = 37.39, F-Crit = 2.88, P = 7.17E-11), driven by textile shedding and tubing wear (Bharathi Dileepan et al., 2025) and influencing ingestion/retention (Kadac-Czapska et al., 2023; Lee et al., 2023). Across seven colors, liquids were 38 % transparent, 35 % black, 13 % red, 6 % blue, 5 % yellow, 2 % green, 0 % violet; solids were 39 % black, 29 % transparent, 11 % red, 10 % blue, 4 % green, 4 % yellow, 2 % violet (Fig. 1D; F = 10.01, F-Crit = 2.23, P = 6.6E-08). FTIR (Fig. S1) identified PE, PP, PS, PA, and NY-6, with PE > PP > PS > PA > NY-6 in both matrices (Fig. 1E; F = 40.37, F-Crit = 2.53, P = 9.31E-16); characteristic peaks confirmed polymer types and traced contamination to packaging and processing equipment (Banica et al., 2024; Pilevar et al., 2019; Diaz-Basantes et al., 2020; Da Costa Filho et al., 2021; Kiruba et al., 2022; Zhang et al., 2023; Caramia and Guerriero, 2010).

Fig. S2, an alluvial diagram, provides a visual map of how different MP characteristics—abundance, size, shape, color, and polymer type—are interrelated within dairy products. It starts with dairy product types (liquid and solid) and traces connections across multiple MP features, highlighting dominant traits like fiber shapes, transparent and black color, and common polymers (e.g., PE, PP).

3.2. Diversity and distribution of MPs in dairy samples

The study evaluated the SDI and DII for the liquid and solid dairy products based on size, shape, color, and polymer composition (Table S2). In liquids, DII peaked at 0.56 for UHT milk and 0.55 for savory yogurt, followed by 0.50 in chocolate milk and 0.48 in buttermilk, with vanilla and mango milks falling below 0.45; one-way ANOVA confirmed significant SDI differences for liquids (F = 10.135, F crit = 3.008, P = 0.00016) and solids (F = 57.231, F crit = 2.946, P = 4.59E-12). High heterogeneity in liquids mirrors intense UHT treatment, homogenization, and aseptic filling that accelerate polymer fragmentation (SDI-Shape ≥0.52) and ink/coating flaking (Kutralam-Muniasamy et al., 2020; Buyukunal et al., 2023). Liquid SDI-Color (0.59–0.72) and polymer diversity (0.46–0.62) implicate PE, PP, PET, PA, and PS ingress (Banica et al., 2024; Basaran et al., 2024). Solids show lower DII (0.32–0.45) with very narrow shape portfolios (SDI-Shape ≤0.23 in yogurt, cup curd, Kalojam, ice cream; near zero in clay-pot curd) but high color diversity (0.63–0.79). Milk powder is an outlier (DII = 0.45; balanced SDIs) due to spray-dryer atomization dislodging multi-polymer fragments (Visentin et al., 2024; Bintsis and Papademas, 2022). The richer MP diversity in liquids suggests broader chemical exposures linked to cumulative oxidative stress and endocrine disruption (Kadac-Czapska et al., 2024a; Talaie et al., 2025).

3.3. MPs' source identification based on multivariate analysis

Fig. 2A’s dendrogram and heat map cluster dairy products into three contamination groups—industrial (UHT milk, curd, milk powder), processed/flavored (ice cream, chocolate milk), and traditional (buttermilk, Kalojam, Roshogolla)—based on MP size, shape, color, and polymer type, indicating common sources from standardized packaging lines, mixing/freezing processes, and regional preparation techniques. Shape traits (fragments, fibers, films) group by shared formation pathways (Smith et al., 2018), color clusters reflect packaging dyes (Kutralam-Muniasamy et al., 2020), and polymer/size clusters mirror manufacturing usage and environmental fragmentation (Cox et al., 2019). Principal component analysis (Fig. 2B) on five polymers yielded three components explaining 81.1 % of variance: PC1 (34.1 %) with strong positive loadings for PE, PS, PA and a negative loading for NY-6—distinguishing packaging-derived MPs from textile/equipment sources (Banica et al., 2024; Bintsis and Papademas, 2022); PC2 (30.7 %) contrasting negative PP against positive PS (with PA and NY-6 contributions), separating hose/glove inputs from PS debris (Kutralam-Muniasamy et al., 2020; Buyukunal et al., 2023); and PC3 (16.3 %) dominated by PA (with moderate NY-6 and negative PS), pointing to adhesives, labels, and industrial membranes as sources (Peng et al., 2023). The PCA not only consolidates the complex variance inherent in the original dataset but also segregates MP sources. Packaging-derived contaminants (primarily PE, PS, and PA) appear distinctly from those related to processing and textural sources (such as PP and NY-6). This multifaceted contamination profile is consistent with previous studies that have demonstrated similar patterns in dairy products, where various stages of production, packaging, and post-processing contribute differentially to the MPs burden (Kadac-Czapska et al., 2024b; Talaie A et al., 2025). Together, these multivariate approaches effectively trace MPs back to packaging and processing origins.

Fig. 2.

(A) Heatmap with hierarchical clustering analysis, and (B) principal component analysis.

3.4. Assessment of MPs contamination index in dairy products

CF values in liquid samples ranged from 1 to 3.13 (mean = 1.88 ± 0.73), while solid samples ranged from 1 to 2.27 (mean = 1.50 ± 0.47) (Fig. 3A). These moderate CF levels indicate consistent MP exposure during processing, packaging, or storage, aligning with earlier findings on food-chain plastic contact (Zhang et al., 2023; Rillig, 2012).

Fig. 3.

(A) CF, (B) NPI, (C) PLI, and (D) PHI assessment results in dairy product samples.

NPI differentiated pollution intensity: most liquid samples fall within the moderately polluted category, with one sample classified as highly polluted. Solid samples span slightly to moderately polluted levels (Fig. 3B). This contrast likely stems from longer exposure durations and enhanced retention in solids, fostering greater MP accumulation (Lusher et al., 2017).

PLI values for liquids varied between 31.27 and 55.32, and solids between 34.78 and 52.42, placing them in risk categories III and IV (Fig. 3C). Specifically, vanilla and chocolate milks were in risk IV, whereas various dairy desserts and frozen items were in risk III. These elevated PLI scores reflect contributions from packaging materials, ultra-high temperature processing, and storage conditions that promote MP shedding (Bouwmeester et al., 2015). By contrast, conventional dairy fats like sour cream and butter exhibit substantially lower PLI values (Banica et al., 2024).

PHI incorporated polymer toxicity: liquids exhibited values from 11.37 to 15.28, solids from 8.63 to 19.27 (Fig. 3D). Vanilla and UHT milks scored highest (15.28 and 14.55) due to hazardous polymers (PA, NY-6), whereas mango milk's lower PHI reflected predominance of low-risk PE and PP despite high MP counts (Galloway et al., 2017). In solids, vanilla ice cream and cup curd showed high PHI linked to PA and NY-6; chocolate ice cream and Roshogolla had moderate PHI with mainly PS (Rochman et al., 2013). These results underscore that toxic risk depends on polymer type as well as MP quantity.

An artificial neural network using tan-sigmoidal activations accurately predicted CF, NPI, and PHI. Optimal architectures (1:3:1 for CF, 4:3:1 for NPI, 5:5:1 for PHI) yielded mean squared errors down to 10⁻⁶ and correlation coefficients near unity (Fig. 4A, Table S3), with predicted versus observed values closely aligned (Fig. 4B), demonstrating robust potential for forecasting MP-related environmental risks in dairy products.

Fig. 4.

(A) ANN–based topology for evaluating MP pollution indices: (a) CF; (b) NPI; (c) PHI; and (B) ANN training and validation: (a) CF; (b) NPI; (c) PHI.

3.5. Health risk evaluation of MP exposure through dairy consumption

The compiled EDI data reveal clear differences in MP exposure across dairy products and between age groups (Fig. 5). In liquid dairy samples, Mango Milk exhibits the highest exposure for both children (75.08 MP/kg.day; 27407.04 MP/kg.year) and adults (22.52 MP/kg.day; 8222.11 MP/kg.year) (Fig. 5A and B). This high exposure is likely linked to the use of artificial flavorings, colorants, and specific packaging materials that can shed particulate matter during processing. In contrast, Strawberry Yogurt shows the lowest exposure in liquids (23.98 MP/kg.day for children and 7.19 MP/kg.day for adults), possibly because it contains fewer synthetic additives and benefits from fermentation processes that may reduce MP presence. One-way ANOVA revealed a statistically significant difference in microplastic exposure between adults and children for liquid samples (F = 17.704, F(crit) = 4.747, P = 0.001), but no significant difference for solid samples (F = 3.761, F(crit) = 4.600, P = 0.072).

Fig. 5.

Estimated daily intake (EDI) assessment results in the dairy product samples (A) daily and (B) yearly.

A similar trend is observed in solid dairy products. Industrially processed yogurt has the highest EDI—67.39 MP/kg.day (24600.42 MP/kg.year) for children and 20.21 MP/kg.day (7380.12 MP/kg.year) for adults (Fig. 5A and B). On the other hand, Kalojam—which is produced through minimal processing and contains fewer synthetic additives—has the lowest EDI values in solids, at 2.60 MP/kg.day (952.29 MP/kg.year) for children and 0.78 MP/kg.day (285.68 MP/kg.year) for adults (Fig. 5A and B).

Comparative data highlight that industrial and traditional dairy products also differ in their MPs EDIs. For example, conventional butter has reported EDI values ranging from 0.268 to 1.071 MPs/(kg·day) for children and 0.279 to 16.071 MPs/(kg·day) for adults, whereas organic butter exhibited slightly higher values, from 0.446 to 0.714 MPs/(kg·day) for children and 6.696 to 10.714 MPs/(kg·day) for adults (Banica et al., 2024).

The observed variations in EDIs across different dairy products suggest that exposure—and consequently, potential health risk—varies considerably based on processing methods, packaging materials, and product composition.

4. Limitations of the study

Sampling was restricted to fifteen dairy products from a single region and season, which may not reflect broader geographic or temporal variability in MP contamination. Visual microscopy and ATR-FTIR identification carry inherent detection limits and may misclassify polymers at low concentrations or in complex matrices. Despite clean-room protocols and blanks, cross-contamination cannot be entirely ruled out, potentially skewing abundance estimates. The study was designed as an occurrence and exposure assessment rather than formal health-risk assessment such as hazard quotient (HQ) or hazard index (HI).EDI estimated ingestion quantities without considering bioavailability, additive leaching, or in vivo toxicodynamics. Multivariate analyses based on relative abundances rather than absolute fluxes limit extrapolation to environmental loading. Future work should widen the sampling scope, integrate nanoscale detection techniques, and incorporate toxicity assays to define exposure and health risks better.

5. Conclusion

This study provides the first comprehensive assessment of MPs contamination across a broad spectrum of industrial and traditional dairy products in Bangladesh. Every sample—liquid and solid—harbored MPs at levels that raise food-safety and public-health concerns. Flavored milks in Tetra Pak cartons and industrial yogurt emerged as the most heavily contaminated, with MP loads exceeding 5000 MP/kg. Diversity indices highlighted greater MP heterogeneity in liquid matrices (DII up to 0.56), while solids retained narrower contamination profiles. Multivariate analyses (hierarchical clustering and PCA) further disentangled packaging-derived signatures from processing-related inputs. Pollution-evaluation indices (CF, PLI, NPI, PHI) consistently classified most dairy products as moderately to highly polluted, and estimated daily intakes (up to 75 MP/kg.day in children) underscored substantial age-dependent exposure. Future research should quantify the release of chemical additives from dairy-associated MPs under gastrointestinal conditions and evaluate the long-term health impacts of chronic MP ingestion. By targeting both packaging innovation and processing safeguards, stakeholders can reduce MP ingress into dairy foods and better protect consumer health. Based on the findings, promoting biodegradable or low-shedding packaging materials, upgrading high-wear food-processing components to minimize polymer abrasion, and implementing routine microplastic monitoring at critical processing stages can enhance the practical relevance and safety of food products (Kaseke et al., 2023).Mitigation of microplastic contamination in dairy requires packaging reform (e.g., biodegradable or low-leaching alternatives), improved filtration techniques during processing, replacement of plastic-based equipment and regulatory standards like establish standardized monitoring protocols, permissible limits for MPs in dairy products to ensure consumer safety (Gürmeriç and Basaran et al., 2025).

Credit author statement

Sadia Sultana Mitu: Conceptualization, Methodology, Investigation, Formal analysis, Data curation, Validation, Visualization, Writing – original draft.

Gopal Chandra Ghosh: Conceptualization, Methodology, Resources, Supervision, Writing – original draft, Writing – review and editing.

Tapos Kumar Chakraborty, Samina Zaman: Conceptualization, Methodology, Writing – review and editing.

Seiya Hanamoto, Yongkui Yang, Marfiah Ab.Wahid,: Writing–review and editing.

Ismail M Rahman, Md Hasibuzzaman, Jarin Tasnim Asha, Sojib Islam, Danisha Sultana, Mahfuz Ahmmed: Methodology, Data curation, Validation, Visualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Handling Editor: Dr. Quancai Sun

Appendix A. Supplementary data

The following is the Supplementary data to this article:

Data availability

Data will be made available on request.