Occurrence and regulatory compliance of aflatoxin M1 in milk and dairy products on the Serbian market

Sampling

A total of 96 milk and dairy samples were collected and analysed for the presence of AFM1. The analysed samples included milk and various dairy products, namely yoghurt (n=13), sour cream and cream products (n=9), sour milk (n=7), kefir (n=2), cheeses and milk-based spreads (n=22), fluid milk (n=16), and whey (n=1). In addition, seven goat milk and goat dairy products were included. Furthermore, 19 internationally available products from the Serbian market were analysed. These included products from Croatia (yoghurt, n=2; kefir, n=2; cheeses and spreads, n=1; milk; n=2), Slovenia (milk, n=3), Germany (yoghurt, n=1; kefir, n=1; cheeses and spreads, n=4; milk, n=1), and Italy (cheeses, n=2). All imported dairy products available on the Serbian market at the time of sampling were included in the study.

Milk and dairy samples were purchased from supermarkets and local retail markets in Serbia, with the aim of covering the widest possible range of products available to consumers. All samples were collected in their original packaging. Sampling was conducted during the period October-December 2024. For each product, an aggregate sample of 1 L (for liquid products) or 1 kg (for solid products) was collected. Aggregate samples were obtained by combining incremental samples collected from the same lot and were considered representative of the lot. After sampling, all samples were transported to the laboratory of the Institute of Food Technology in Novi Sad and stored at 4 °C until analysis. Sampling procedures followed the principles outlined in Regulation (EC) No 401/2006 (European Commission, 2006).

Chemicals and reagents

The determination of AFM1 was performed using a validated ELISA method in an ISO/IEC 17025:2017 accredited laboratory (ISO, 2017) at the Institute of Food Technology in Novi Sad. Analyses were conducted using a commercial ELISA test kit (I´Screen AFLA M1, Gold Standard Diagnostics, Trieste, Italy). All reagents required for the analysis (AFM1 standards, enzyme conjugate, extraction buffer, washing buffer, developing solution, stop solution and sample diluent) were supplied with the test kit. Methanol, dichloromethane and hexane (Fisher Chemical, USA) were used for sample extraction and preparation. Distilled water was produced using a laboratory water purification system (Crystal EX Laboratory Water Purification System, Adrona, Latvia).

Sample preparation

Milk samples were centrifuged at 3000 × g for 10 min, and the upper fat layer was subsequently removed. A 0.1 mL aliquot of the lower aqueous phase was used directly for analysis. For yoghurt and sour milk, 1 g of sample was extracted with 5 mL of methanol, shaken for 5 min (Griffin Flask Shaker, Griffin and George, Wembley, England), and centrifuged (Boeckel + Co GmbH + Co, Hamburg, Germany) at 5000 × g for 5 min. A 0.25 mL aliquot of the supernatant was evaporated under a nitrogen stream at 40 °C (Reacti-Vap, Thermo Fisher Scientific, Waltham, MA, USA), and the residue was reconstituted in 0.25 mL of sample diluent and vortexed (Velp Scientifica, Usmate, Italy) prior to analysis. For cheese and sour cream, 2 g of sample was extracted with 15 mL of dichloromethane and shaken for 15 min. After filtration, 3.75 mL of the extract was evaporated under nitrogen at 60 °C and reconstituted in 0.75 mL of extraction buffer. Subsequently, 0.75 mL of hexane was added, vortexed, and the upper hexane layer discarded. A 0.050 mL aliquot of the remaining phase was diluted with 0.20 mL of dilution buffer for ELISA analysis.

Aflatoxin M1 analysis in milk and dairy products

All analyses in this study were conducted in the laboratory of the Institute of Food Technology in Novi Sad, University of Novi Sad. AFM1 concentrations were determined using the ELISA method according to the manufacturer’s instructions, as previously described by Kos et al. (2024b). Absorbance was measured at 450 nm using a microplate reader (Thermo Labsystems 354 Multiskan Ascent Microplate Reader Finland). AFM1 concentrations were calculated using the original spreadsheet provided with the ELISA kit (Eurofins Techna; Spreadsheet for I’Screen AFLA M1, IS Afla M1, rev. 6), based on a calibration curve constructed using seven standards (0, 5, 10, 25, 50, 100, and 250 ng/L). Colour development was inversely proportional to the AFM1 concentration in the samples. All analyses were performed in duplicate.

Method validation

The ELISA method used for the determination of AFM1 was previously validated in accordance with European requirements for screening and confirmatory analytical methods established by Regulation (EC) No 2002/657 (European Commission, 2002) and Technical Report 16059 (CEN, 2012), as described by Kos et al. (2024b).

The method demonstrated satisfactory sensitivity, precision, and trueness for milk and dairy matrices. Limits of quantification provided by the manufacturer were 0.005 µg/kg for milk, 0.025 µg/kg for fermented dairy products, and 0.037 µg/kg for cheese. Method performance was evaluated using milk powder reference material (04370, FAPAS, York, UK), as well as naturally contaminated and spiked dairy matrices. In addition, reference materials were routinely analysed within each analytical batch to ensure the validity of the obtained results. Precision parameters, expressed as relative standard deviations for repeatability and reproducibility, complied with the acceptance criterion of <20 %, while trueness and recovery values ranged between 80 % and 120 %. The suitability of the reported LOQ levels was additionally confirmed by spiking blank samples, which demonstrated acceptable recovery and repeatability. In addition, the performance of the ELISA method is regularly verified through participation in proficiency testing schemes.

Statistical analysis

The collected data were systematized in Microsoft Excel (v. 2019) and analysed through descriptive statistical analysis performed with Tibco Statistica (v. 14). Results were expressed as mean values, standard deviation, minimum and maximum concentrations, as well as the percentage of contaminated samples.

Differences in concentration levels (µg/kg) across product groups and origin were evaluated using a two-way analysis of variance (ANOVA), with product group (milk, fermented dairy products, soft cheeses) and origin (Serbian/domestic vs. imported) included as fixed factors, along with their interaction term (product group × origin). Given the positively skewed distribution of concentration values, the dependent variable was log₁₀-transformed prior to analysis to improve normality and homoscedasticity assumptions. The normality of residuals was evaluated using the Shapiro–Wilk test, while the homogeneity of variance was assessed using Levene’s test. No major violations of ANOVA assumptions were observed after log₁₀ transformation (p > 0.05). When a statistically significant interaction was observed, simple main effects were examined by comparing Serbian and imported products within each product group. Adjustment for multiple comparisons was performed using the Holm correction method.

Occurrence and regulatory compliance

The analysis of milk and dairy products collected in 2024 confirms the persistence of AFM1 contamination in Serbia, consistent with a long-standing trend reported over the past decade. AFM1 was detected in all analysed categories of milk and dairy products (Table 1). The highest mean concentration was observed in cheeses and spreads (0.226 µg/kg), followed by kefir (0.156 µg/kg), sour cream (0.120 µg/kg), yoghurt (0.083 µg/kg), sour milk (0.066 µg/kg) and milk (0.058 µg/kg). The highest individual concentration was detected in a cheese and spread sample (0.573 µg/kg), while the lowest concentrations were observed among imported products, particularly imported milk samples (0.006 µg/kg).

Table 1. Prevalence and concentration of AFM1 (µg/kg) in milk and dairy products

n - number of samples; SD - standard deviation; n.a. - not applicable (n=1); positive samples refer to samples with AFM1 concentrations above the limit of quantification (LOQ)

According to the current Serbian regulation, which sets the ML of AFM1 in milk at 0.25 µg/kg, all analysed milk samples complied with the national regulatory limit, and no exceedances were recorded. However, when the same results were evaluated against the stricter EU ML for AFM1 in milk (0.05 µg/kg), 10 out of 16 milk samples (63 %) exceeded the EU threshold. This substantial discrepancy between Serbian and EU regulatory standards suggests that, despite compliance with national legislation, consumers may still be exposed to AFM1 at levels considered unacceptable under EU criteria, potentially posing a public health concern, particularly for children and high dairy consumers. The proportion of dairy samples exceeding 0.05 µg/kg is presented in Figure 1.

Figure 1. Percentage of milk and dairy product samples with AFM1 concentrations above 0.05 µg/kg, corresponding to the European Union maximum limit for AFM1 in milk. Product categories with a very small number of samples, such as kefir, should be interpreted with caution

Compared with previously published Serbian data, the mean AFM1 concentration in milk observed in 2024 indicates a marked reduction relative to the severe contamination episodes reported during and immediately after the 2013 aflatoxin crisis. The widespread aflatoxin contamination of maize in 2012-2013 triggered a national food safety crisis in Serbia, underscoring the vulnerability of the dairy supply chain to mycotoxin contamination and resulting in elevated AFM1 levels in milk, significant economic impacts, and regulatory responses (Kos et al., 2024a). During that period, AFM1 contamination was detected in nearly all milk samples, with mean concentrations around 0.13 µg/kg and approximately 70 % of samples above the EU limit of 0.05 µg/kg (Škrbić et al., 2014, Torović, 2015). Similarly high contamination levels were reported by Kos et al. (2014) in 2013, with 86 % of samples exceeding the EU limit and concentrations reaching up to 1.2 µg/kg, resulting in elevated dietary exposure, particularly among children. This trend was further confirmed by a systematic review of Serbian milk from 2015-2016, which reported AFM1 detection in 70-85 % of raw milk and up to 98 % of heat-processed milk samples, while exceedances of the Serbian limit were rare, compared with substantially higher exceedances of the EU limit (17-30 %), a pattern consistent with the present study (Milićević et al., 2017).

Subsequent monitoring studies demonstrated that, although average AFM1 concentrations declined after the crisis, contamination remained frequent. Between 2013 and 2018, AFM1 was detected in approximately 70-85% of milk samples in Serbia, with 22-31 % exceeding the EU limit (Jajić et al., 2018; Milićević et al., 2019). Importantly, AFM1 occurrence was consistently higher in raw milk than in processed dairy products, where exceedances were rare (<2%) and mean concentrations generally remained below 0.02 µg/kg, indicating effective control measures at the processing stage (Djekic et al., 2020).

More recent large-scale studies confirm the persistence of AFM1 contamination in Serbia. Udovicki et al. (2023) reported AFM1 detection in approximately 78 % of milk samples collected between 2015 and 2022, while Krstović et al. (2025) documented AFM1 presence in nearly 70 % of raw milk samples during 2021–2025, with pronounced seasonal and interannual variability. These findings are supported by data from 2022 showing widespread aflatoxin B1 contamination of maize and high prevalence of AFM1 in milk, highlighting the strong link between climatic conditions, feed contamination and subsequent AFM1 transfer into milk (Kos et al., 2024b). Overall, while average AFM1 concentrations were lower than those reported during the 2013 aflatoxin crisis, the high prevalence of contamination and the substantial proportion of samples exceeding the EU MLs indicate that AFM1 remains a relevant food safety concern. Serbia is formally committed to harmonising its food safety legislation with EU standards. However, the multi-year inconsistencies in MLs for AFM1, persisting for over a decade, reflect a continuous challenge in controlling contamination, which is clearly confirmed by the high prevalence and widespread occurrence observed in the present study.

Differences according to product group and origin

To further evaluate whether the observed regulatory asymmetry was reflected in measurable differences between domestically produced and imported dairy products, AFM1 concentrations were analysed according to product group (milk, fermented dairy products, soft cheeses) and origin (Serbian vs. Imported). A two-way analysis of variance (ANOVA) was performed on log₁₀-transformed concentration values. The results are presented in Table 2.

Table 2. Results of the two-way ANOVA

The two-way ANOVA revealed no statistically significant main effect of product group (F(2,81)=4.73, p=0.063), while origin (F(1,81)=134.82, p<0.001) and the interaction between product group and origin were statistically significant (F(2,81)=13.49, p<0.001). The presence of a significant interaction indicates that differences between Serbian and imported products varied across product categories; therefore, interpretation focused on the analysis of simple effects. Due to the significant interaction, additional analyses were conducted to examine differences between product groups separately for Serbian and imported samples. Although the interaction term was statistically significant, post hoc comparisons performed within each origin did not reveal statistically significant differences between product groups (all p > 0.05). This finding suggests that the interaction was primarily driven by differences between Serbian and imported products within product categories, rather than by pronounced differences among product groups within the same origin. Subsequently, comparisons between Serbian and imported products were performed within each product group using t-tests on log₁₀-transformed concentrations, with Holm adjustment for multiple comparisons (Table 3).

Table 3. Comparison of Serbian vs. imported products within each product group

Statistically significant differences between origins were observed in all three product categories. AFM1 concentrations were significantly higher in Serbian compared with imported products in milk (p_adj=0.004), fermented dairy products (p_adj<0.001), and soft cheeses (p_adj<0.001). For visual confirmation of the ANOVA findings and simple effects analysis, the distribution of concentrations (µg/kg) within each product group is presented separately according to origin (Serbian vs. Imported).

Figure 2. Distribution of AFM1 concentrations (µg/kg) by product group and origin (Serbian vs. Imported products), presented on a log₁₀ scale. Boxes represent the interquartile range, horizontal lines indicate median values, and points indicate individual observations

The graphical representation clearly demonstrates that concentrations in imported samples were generally lower and clustered at the lower end of the scale across all product groups, whereas Serbian samples exhibited higher values and greater variability, particularly among fermented products and soft cheeses (Figure 2). The consistently lower AFM1 concentrations observed in imported products may reflect stricter feed quality control systems and closer alignment with EU regulatory requirements, suggesting that current preventive measures in domestic production remain insufficient to consistently minimize contamination. This pattern is consistent with the results of the simple effects analysis (Table 3), confirming statistically significant differences between Serbian and imported products within each category. Notably, the higher AFM1 concentrations observed in fermented dairy products and cheeses in the present study are consistent with observations that AFM1 binds strongly to casein and is concentrated during coagulation and fermentation processes (Škrbić et al., 2015, Goncalves et al., 2021, Jauković et al., 2024). Numerous studies report higher contamination levels in cheeses and fermented products compared with raw or fluid milk (Škrbić et al., 2015; Sarmast et al., 2021; Abi Rizk et al., 2026).

Due to export requirements and the obligation to comply with EU legislation, Serbian dairy products intended for export must meet the EU MLs for AFM1 (0.05 µg/kg). Consequently, milk exceeding the EU threshold but remaining below the national ML of 0.25 µg/kg may legally remain on the domestic market. This regulatory asymmetry potentially creates a disparity between products intended for export and those available for domestic consumption. As an EU candidate country, Serbia is formally committed to harmonising its food and feed safety framework with EU legislation. However, European Commission progress reports under Chapter 12 - Food Safety, Veterinary and Phytosanitary Policy (European Commission, 2025) have repeatedly highlighted gaps in legislative alignment, enforcement of risk-based controls, and institutional capacity, including laboratory resources and inter-institutional coordination. In this context, the long-standing divergence between Serbian and EU MLs for AFM1 reflects not only regulatory differences but also structural challenges in achieving full implementation of EU-compliant food safety systems.

In a global context, the prevalence and distribution of AFM1 observed in the present study are consistent with international findings reported across diverse dairy production systems. For example, high prevalence has been reported in Bangladesh, where AFM1 was detected in 70-100 % of raw, pasteurized and UHT milk samples, with up to 97 % of samples exceeding the EU MLs in some surveys (Tarannum et al., 2020; Sumon et al., 2021). Similarly elevated contamination has been documented in several African countries, including Ghana, where AFM1 was detected in 55.8-94.6 % of raw milk samples, with concentrations reaching 1606.8 ng/L and exceedance rates up to 100 % (Kortei et al., 2022; Nuhu et al., 2025), as well as in Nigeria, where AFM1 concentrations in goat milk reached up to 3108 ng/L (Akinyemi et al., 2022). In contrast, countries with well-established monitoring systems and strict feed control measures are generally associated with lower AFM1 prevalence and fewer exceedances of regulatory limits. For instance, studies from Spain and Italy reported no detectable AFM1 or exceedance rates below 1% in milk samples (Flores-Flores and González-Peñas, 2018; Roila et al., 2021; Rodríguez-Cañás et al., 2024). Studies from neighbouring countries in Southeast Europe report heterogeneous contamination patterns, with a pronounced regional peak during the 2013 crisis followed by partial improvement in some countries, while sporadic exceedances and seasonal peaks persist, underscoring the importance of feed management and monitoring intensity rather than agro-climatic factors alone (Bilandžić et al., 2016; Kesić et al., 2021; Ilievska et al., 2022). Recent Croatian monitoring data indicate that approximately 5-15 % of raw milk samples exceeded the EU MLs (Bilandžić et al., 2025). Moreover, comparative analyses have shown a substantially higher proportion of AFM1-positive samples in dairy products of Croatian origin (55.1 %) compared with imported products (21.9 %), which may reflect differences in regional feed quality and production practices, although the observed variation could also be influenced by sampling characteristics and product comparability (Kovač Tomas et al., 2025).

This persistent occurrence of AFM1 in Serbian milk, particularly in samples exceeding the EU MLs, likely reflects the underlying AFB1 contamination in maize (Kos et al., 2013; Janić Hajnal et al., 2017; Janić Hajnal et al., 2020; Kos et al., 2020; Kos et al., 2024b). While official data for 2024 are not yet available, the extreme heat and drought during the summer months likely promoted AFB1 formation, which subsequently contributed to the AFM1 levels observed in milk and dairy products analysed in this study. These findings support growing evidence that climate variability, particularly elevated temperatures and drought conditions, may increasingly contribute to aflatoxin contamination in feed and consequently to AFM1 occurrence in dairy products across Southeast Europe. Persistent AFM1 contamination in Serbian milk may represent an early indicator of a broader regional trend, as rising temperatures, altered precipitation patterns, and increasing climate variability are expected to intensify mycotoxin risk across the Balkan region (Kos et al., 2023).

Certain limitations of the present study should also be considered. Some dairy products were grouped into broader categories for statistical analysis, which may have masked potential differences among individual product types. In addition, several categories, particularly kefir and whey, included a limited number of samples. Nevertheless, despite the small number of analysed samples, the obtained findings are consistent with previous reports. Milićević et al. (2021) also reported the occurrence of AFM1 in whey samples from Serbia, indicating that AFM1 may persist in whey as a consequence of toxin transfer during dairy processing. Similarly, the occurrence of AFM1 in kefir and other fermented dairy products has been documented in previous studies (Khorshidi et al., 2022), supporting the observation that AFM1 can persist in different fermented dairy matrices. Furthermore, AFM1 occurrence was not correlated with feed contamination data, and confirmatory chromatographic analysis was not performed. Future research should address these limitations through the inclusion of a larger number of representative samples, feed contamination analysis, and confirmatory LC-MS/MS analysis.