Are Fermented Foods Safe for Human Health? A Narrative Systematic Analysis of Literature

1. Introduction

Fermented foods (FF) constitute a diverse class of foods and beverages whose production relies on the metabolic activity of microorganisms that transform raw substrates into products with distinct sensory, nutritional and functional properties. Fermented foods and beverages have been recently defined as “foods made through desired microbial growth and enzymatic conversions of food components” (Marco et al. 2021). They are ubiquitous across dietary traditions and, in many cultures, represent staples of daily diets. Historically, fermentation offered a robust strategy for preservation by lowering pH, producing antimicrobial metabolites and reducing water activity. Today, the same processes continue to contribute to safety and stability while also shaping organoleptic quality and nutritional value (Macori and Cotter 2018). In recent years, scientific and public interest in fermented foods has accelerated, driven by hypotheses about their capacity to deliver live microbes and fermentation-derived metabolites with potential benefits for human health, including modulation of the gut microbiome, enhancement of bioavailability of nutrients, and generation of bioactive compounds (Balasubramanian et al. 2024; Leeuwendaal et al. 2022; Marco et al. 2017; Künili et al. 2025). At the same time, it remains essential to systematically investigate their safety, recognising that any food, fermented or not, may present hazards if the ingredients, processes or handling are inadequate (Marco et al. 2021). Observational and intervention studies, including broader syntheses, suggest that fermented foods can impact the gut microbiome structure and function, potentially affecting immune and metabolic readouts (Mukherjee et al. 2025). However, causality and generalizability vary by product, context, and study design (Balasubramanian et al. 2024; Marco et al. 2021; Mukherjee et al. 2024; Valentino et al. 2024). The ubiquity and growing popularity of fermented foods underscore the importance of evaluating their safety through transparent, evidence-based approaches.

From a process perspective, lactic acid fermentation, ethanol production, salt addition, drying and ripening steps typically produce combinations of hurdles (low pH, organic acids, ethanol, reduced water activity, bacteriocins) that suppress many pathogens and spoilage organisms (Laranjo, Potes, and Elias 2019; Marco et al. 2017). This underlies the widespread use of fermentation as a preservation strategy across dairy, vegetable, cereals and legumes, meat and fish matrices (Macori and Cotter 2018; Marco et al. 2017). Nevertheless, safety cannot be assumed, since vulnerability can arise at several points, most notably from contaminated raw materials, inadequate process control, cross-contamination after fermentation, or due to the formation and accumulation of certain metabolites under permissive conditions (Barbieri et al. 2019; Laranjo, Potes, and Elias 2019; Marco et al. 2021; Pop et al. 2024; Tian et al. 2022). Furthermore, new risks may arise considering the development of novel FF (e. g. new strains, new matrices, etc…), particularly in light of the shift to plant-based foods as alternative protein sources (Mastrotheodoraki et al. 2026; Jin et al. 2025).

From a Food Safety perspective, microbiological hazards are central to the evaluation of fermented foods because they may enter with raw materials, persist when critical parameters are not adequately controlled, or be introduced post-process despite the protective hurdles created by fermentation (acidification, salt, reduced water activity, and microbial competition) (Laranjo, Potes, and Elias 2019; Marco et al. 2021). At the surveillance level, European foodborne outbreak data are compiled by the European Food Safety Authority (EFSA) (2025). However, the publicly available data present aggregated categories that do not always allow confirmation of the specific food product as the causal vehicle, reporting implicated food groups rather than verified outbreak sources. As such, potential raw materials for fermented foods, and in some cases fermented products themselves, may appear within these categories without permitting an easily definitive attribution, underscoring the need for a systematic appraisal of microbiological risks in this product class.

Regarding chemical risks in fermented foods, they fall into two broad groups: i) contaminants in raw materials, e.g., mycotoxins and pesticide residues in cereals and heavy metals in fish, which (in this case) fermentation may reduce, redistribute or occasionally concentrate, but does not originate or explain their initial presence (Pop et al. 2024; Tian et al. 2022); and, ii) compounds originated during fermentation, such as biogenic amines (e.g., histamine and tyramine produced by decarboxylase-positive microbes in high-protein matrices (Tsai 2005)), and acrylamide (a heat-induced, potentially carcinogenic contaminant on which fermentation can have a positive or negative effect, depending on many factors (Jensen, Hakme, and Feyissa 2025).

This narrative systematic review addresses two core questions: “Are fermented foods safe for human health?” and “Which microbiological and chemical hazards are associated with fermented foods, and what risks do they pose?”. Within PIMENTO—COST Action CA20128 “Promoting Innovation of ferMENTed fOods” Working Group 3, our specific aims were to: i) characterise the occurrence of food hazards in fermented foods when linked to human consumption and reported illness; ii) synthesise the reported risk of exposure to these hazards, considering product type and context of production; and, iii) collate reported foodborne outbreaks attributed to fermented foods, including the extent of resulting disease (Todorovic et al. 2024).

Together, these objectives provide an evidence-based foundation to distinguish hazards arising from contaminated inputs or process deviations from those attributable to fermentation itself, and to inform proportionate risk management and surveillance priorities, particularly in the context of risk-benefit of fermented foods.

2. Methodology

This systematic narrative review was conducted according to a predefined protocol developed within the PIMENTO COST Action CA20128, Working Group 3 (WG3). This protocol followed the guidance of Muka et al. (2020)and the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines (Page et al. 2021) and it was registered on the Open Science Framework: https://osf.io/hsakz/.

2.1. Research Question

A systematic narrative review of the scientific literature on the safety of fermented foods aimed to answer two specific questions: 1) “Are fermented foods safe for human health?”; and 2) “Which microbiological and chemical hazards are associated with fermented foods, and what risks do they pose?”

2.2. Eligibility Criteria

The eligibility criteria were established based on the PECO framework (Schardt et al. 2007; Mintzker, Blum, and Adler 2023). The population corresponded to the general population, including vulnerable groups (e.g., elderly, children, without any limitation on age, gender, health and physiological conditions (i.e., pregnant women, breastfeeding women, etc). The exposure consisted of the ingestion of any of the fermented foods contained in the PIMENTO search string for fermented foods across the following food groups: dairy, meat and fish, fruits and vegetables, beverages, legumes, cereals and grains. Alcoholic beverages with an alcohol content of more than 1.25% were excluded. Unless specified otherwise, no limits were set for duration or dosage of the ingested fermented food(s). Unless specified otherwise, studies investigating application of fermented foods other than for nutritional purpose (e.g., nasal or topical) were excluded. In addition, studies investigating probiotics were excluded unless the probiotic(s) were added at the beginning of the fermentation process and that there were indications from the literature that the probiotic strain(s) contribute(s) to the fermentation of the food matrix. Exposure, including any possible confounders such as prebiotic fibres or added bioactive compounds, were not included. The outcomes included reporting of (i) the occurrence (levels/magnitude; frequency) of food hazards in the defined fermented foods, when associated with human consumption and reported illness; (ii) the reported associated risk of exposure, taking into account the probability and severity; and (iii) the reported outbreaks that may occur and the extent of the diseases that occur. The comparison (i.e., absence of consumption of the fermented foods), was not considered.

2.3. Literature Search Strategy

Literature search for English written studies was conducted using PubMed, Scopus and Cochrane Library (Cochrane Reviews and Cochrane Trials) databases from 01/01/1970 until 31/12/2024.

A generic search string was developed by the Library of the University of Zurich in the context of PIMENTO project, which considered terms for the search of a broad scope of fermented foods across all food groups, of all types of human studies, as well as of dietary intake (Supplementary Table S1). Specific terms were considered, such as food safety, safety, contaminant, toxic, infection, outbreak.

2.4. Study Selection

Study selection was carried out using CADIMA-Central Access Database for Impact Monitoring and Assessment (https://www.cadima.info).

After deletion of duplicates, a consistency test training was performed in CADIMA before the study selection.

Then, study selection was conducted in two phases: i) title and abstract screening; ii) full text screening, both independently assessed by two reviewers. The inconsistencies found in both phases were solved by simple consensus between the two reviewers and mediated by the project leaders (R.A. and M.L.), whenever necessary.

2.5. Data Extraction

Data were extracted independently by two reviewers and compiled by the project leaders. Standardised forms were used for microbiological and chemical hazards data extraction (Supplementary Table S2).

3. Results

3.1. Scoping Results

Following literature search, papers were retrieved from three databases using the search string presented in Supplementary Table S1. After removal of duplicates, 2828 papers were screened for eligibility based on their title and abstract. A total of 2285 papers were excluded based on the title and abstract screening, while 543 papers were retained for full-text screening. The final selection included 123 studies that were kept for data extraction (Figure 1).

3.2. Microbiological Hazards

The analysis of fermented foods safety incidents or outbreaks reveals several fermented foods associated with serious health risks due to microbiological contamination (Table 1).

Bacteria were the most commonly identified microbiological hazards, with Salmonella, Escherichia coli and other enterobacteria, as well as Campylobacter jejuni and Clostridium botulinum being the most frequently reported (highest number of reported cases), while Listeria monocytogenes shows up as the most lethal.

Concerning toxin-producing pathogens, three B. cereus strains (cereulide and enterotoxins) (Zhou et al. 2014), different Clostridium botulinum strains (toxins types A, B and E) (Bacha et al. 2021; Ganapathiraju et al. 2019; Mehramiri 2017; Pourshafie et al. 1998; Tseng et al. 2009), and one STEC strain (Shiga toxin) have been reported (Ethelberg et al. 2009).

Regarding viruses, only hepatitis E virus and noroviruses have been reported associated to outbreaks (Cho et al. 2016; Guo et al. 2014; Li et al. 2021; Park et al. 2015; Smith et al. 2021). Toxoplasma gondii and Trichinella britovi were the two parasites-that have been reported to be involved in foodborne outbreaks (Cortés-Blanco et al. 2002; da Costa et al. 2020).

Cheese and yogurt are the fermented foods most frequently associated to foodborne outbreaks of microbiological origin, thus being the FF that pose greatest concerns in terms of food safety. Fermented dairy products are also the FF type that have shown the highest prevalence of consumption (Syrpas et al. 2025).

Several epidemiological investigations have highlighted that certain fermented foods have been implicated in severe foodborne outbreaks resulting in a notable number of hospitalizations and, in some cases, fatalities. Among these, fermented dairy products appear most frequently associated with significant public health concerns.

In particular, within the dairy sector, yogurt has been linked to 74 hospitalizations, primarily associated with Salmonella Typhi contamination. However, cheese products have been recurrently implicated in outbreaks involving different pathogenic species. For example, cheese contaminated with C. jejuni caused 44 hospitalizations, while unpasteurized Mexican-style aged cheese (Cotija) was responsible for 36 hospitalizations associated with multidrug-resistant Salmonella Newport. Another serious outbreak was linked to a salad containing cheese, tuna fish and chicken, which resulted in 23 hospitalizations due to L. monocytogenes serotype 4b.

Among the food categories, fermented meat products, such as salami, accounted for 52 and 21 hospitalizations related to Salmonella Montevideo and Salmonella enterica Typhimurium, respectively.

Deaths have also been reported following the consumption of cheese-based products. Both cheese with tuna fish and chicken salad and Mexican-style cheese were responsible for five deaths each, in both cases due to L. monocytogenes infections. Soft cheese made from raw cow’s milk was linked to two deaths caused by Salmonella Montevideo, while isolated fatal cases involved C. botulinum type A in a generic cheese product and L. monocytogenes serotype 1/2b in Gorgonzola cheese.

However, even though the above identified microbiological risks appear in several FF, from the analysis of the retrieved studies, these risks do not emerge from the fermentation process itself, but they mostly come from the raw materials.

These findings highlight the importance of rigorous hygiene, pasteurization, and monitoring of pathogens such as Listeria monocytogenes, Salmonella spp., and Clostridium botulinum in fermented food production.

3.3. Chemical Hazards

The analysis of fermented food safety incidents or outbreaks suggests various fermented foods associated with relevant health problems caused by chemical hazards, such as mycotoxins, heavy metals, polycyclic aromatic hydrocarbons (PAHs), and acrylamide.

3.3.1. Mycotoxins

Across the extracted records (Table 2), mycotoxin exposure associated with fermented foods clustered around cereal-based products (bread and bakery items) and fermented dairy (yoghurt, kefir, fermented milk, cheese), with additional contributions from coffee. Table 2 summarises, for each food type and toxin, the reported intake and risk metrics used for risk characterisation.

Across cereal-based fermented foods, mycotoxin intakes and, when available, associated risks were described. For bread and bakery, aflatoxin B1 (AFB1) intakes of 0.002–0.02 ng/kg bw/day corresponded to a maximum HQ of 0.0011, a minimum MOE of 21,193, and a maximum LCR of 0.0016 (Hoteit et al. 2024); additional aflatoxins (AFB2, AFG1, AFG2) were each reported at 0.080 ng/kg bw/day without risk characterisation (Sirot, Fremy, and Leblanc 2013; Coppa et al. 2020). Ochratoxin A (OTA) ranged from 0.01–1.18 ng/kg bw/day (max HQ 0.0653; min MOE 12,331) (Hoteit et al. 2024; Sirot, Fremy, and Leblanc 2013). For deoxynivalenol (DON), a worst-case intake of 332.18 ng/kg bw/day yielded HQ 0.0415 (Hoteit et al. 2024; Sirot, Fremy, and Leblanc 2013). Reported intakes for nivalenol (13.29 ng/kg bw/day), zearalenone (0.706 and 8.03 ng/kg bw/day) and fumonisins (FB1 13.49; FB2 3.47 ng/kg bw/day) lacked risk metrics, limiting formal appraisal beyond risk magnitude (Sirot, Fremy, and Leblanc 2013; Coppa et al. 2020).

In dairy, all the available data concerned aflatoxin M1 (AFM1). For cheese, intakes from 0.003–0.06 ng/kg bw/day produced maximum HQs of 0.174–0.33, minimum MOEs of 14,561–127,389, and maximum LCRs up to 0.00038 (Hoteit et al. 2024; Heidari et al. 2024; Farkas et al. 2022) (Behtarin and Movassaghghazani 2024). One study reported at 0.06–0.09 ng/kg bw/day and a minimum MOE of 6482 and a maximum LCR of 0.0005, indicating potential health concern (Massahi et al. 2023). For yoghurt/kefir and fermented milk, AFM1 ranged 0.06–0.07 ng/kg bw/day (max HQ 0.3496) and 0.00026–0.29 ng/kg bw/day (max HQ 0.022; min MOE 40,000), consistent with low concern (Hoteit et al. 2024; Heidari et al. 2024). Coffee contributed very low OTA (0.00037 ng/kg bw/day) with no risk metrics reported in that study (Sirot, Fremy, and Leblanc 2013).

Overall, wherever risk metrics were available, HQs were <1 and MOEs were generally ≥10,000, except for cheese/cream cases, generally suggesting low concern of mycotoxins for most fermented foods assessed. However, the absence of risk metrics particularly for zearalenone, fumonisin and nivalenol constrains definitive risk interpretation (Hoteit et al. 2024; Farkas et al. 2022; Heidari et al. 2024; Sirot, Fremy, and Leblanc 2013; Behtarin and Movassaghghazani 2024; Massahi et al. 2023) (Coppa et al. 2020; Massahi et al. 2024).

3.3.2. Heavy Metals

Table 3 summarises the concentrations in products (µg/kg or mg/kg), estimated dietary intakes (µg/kg bw/day), and risk metrics (THQ/HI/MOE), complemented in a few cases by biomonitoring data. The results are organised by product group (cereal-based breads/bakery; fermented dairy including cheese/cream, yoghurt/kefir/fermented milk; and coffee/chocolate).

In multigrain, wholemeal (or whole wheat), rye, and white breads, several metallic elements are present, including manganese (Mn), copper (Cu), nickel (Ni), lead (Pb), arsenic (As), chromium (Cr), cobalt (Co), cadmium (Cd), and mercury (Hg). The average daily intake of these metals is primarily dominated by manganese at 44.0 µg/kg body weight and copper at 6.62 µg/kg body weight, while other metals contribute at lower levels, such as nickel at 0.69 µg/kg, lead at 0.15 µg/kg, and arsenic at 0.06 µg/kg. Concentrations of these metals in bakery products can vary significantly; for example, manganese levels range from 19,084 to 28.151 µg/kg, copper from 3.139 to 4.377 µg/kg, nickel from 260 to 710 µg/kg, and lead from 20.1 to 56.3 µg/kg. While individual Target Hazard Quotients (THQs) are less than 1, the cumulative Hazard Index (HI) reaches between 1.37 and 1.71. This suggests a potential cumulative effect, even though single-metal assessments seem acceptable. Additional analyses of bread products indicate lead and nickel contributions with a reported Margin of Exposure (MOE) of 18.7. Furthermore, methylmercury is being investigated in separate datasets. Together, these findings highlight the importance of product formulation (such as whole grain or bran inclusions), and the specific types of metals present in determining the risk outcomes associated with these bakery products.

In Neisi, Farhadi, Cheraghian, et al. (2024), Generalized Additive Models (GAM) were employed for advanced exposure assessment in healthy and patient groups, providing probabilistic insights into the exposure-risk relationships between heavy metals in food and cardiovascular diseases by statistically comparing the mean concentrations of heavy metals in urine between the patient group and the healthy group. GAM analysis revealed that the levels of Sr and Cd in bread exceeded Iran’s standard limit (Neisi, Farhadi, Cheraghian, et al. 2024). There was a significant correlation (p-value < 0.05) between urinary heavy metal levels in individuals with heart disease and those in the healthy group, excluding Ni.

A detailed profile of sheep cheese reveals several concerning elements, with Target Hazard Quotients (THQs) calculated for both preschool children and adults. The most notable finding is arsenic (As), with THQ values of 11.06 for preschool children and 3.48 for adults, reflecting a mean product level of 0.73 ± 0.29 mg/kg. Antimony (Sb) also presented significant values, with a THQ of 2.5 for preschool children and 0.79 for adults, and a mean level of 0.22 ± 0.15 mg/kg. Other elements such as barium (Ba), chromium (Cr), copper (Cu), iron (Fe), lead (Pb), lithium (Li), molybdenum (Mo), nickel (Ni), and strontium (Sr) showed THQs well below 1 across all groups. These findings highlight that cheese type and origin are crucial factors, with exposure in young children being the most critical concern (Almášiová et al. 2024).

In processed dairy products, titanium (Ti) was detected at 0.12 ± 0.17 mg/kg, although risk levels were not reported (Rompelberg et al. 2016). Specific entries indicate lead and nickel in sour cream (Pb 1.98 µg/kg, Ni 48 µg/kg), which have a relatively high MOE of 1111.6, suggesting low concern. Conversely, cheese had lead levels of 52.24 µg/kg and nickel at 48 µg/kg, with an MOE of 57.4. Cheese curd had even higher lead levels at 183.28 µg/kg and nickel at 135 µg/kg, yielding an MOE of 2.15, indicating the narrowest margin of safety among dairy products concerning lead exposure (Pipoyan et al. 2023). Most cheese and cream varieties pose low concern, but certain products, like sheep cheese and specific curds, exhibit elevated lead and arsenic levels that necessitate targeted sourcing controls and verification testing, particularly for items marketed to children.

In the realm of fermented milk, a wide array of metals (including mercury (Hg), lead (Pb), cadmium (Cd), chromium (Cr), arsenic (As), zinc (Zn), copper (Cu), iron (Fe), manganese (Mn), and nickel (Ni)) showed ranges such as Pb from 0 to 35.01 µg/kg, Cd from 0 to 305 µg/kg, Cr from 0 to 433 µg/kg, and As from 0 to 17 µg/kg. All assessed metals had a THQ of less than 1, and the Hazard Index (HI) was below 1 overall, indicating a low risk under typical intake scenarios [Ref 1755]. For the traditional Armenian fermented milk Matsun, the estimated intakes of lead (0.84 µg/kg) and nickel (19 µg/kg) resulted in a high MOE value of 1513.8, indicating low concern (Pipoyan et al. 2023).

Yoghurt showed titanium levels of 0.47 ± 0.46 mg/kg, but risk characterization was not provided (Bolliger, van Zijl, and Louw 1992).

For coffee, various surveys report cadmium levels ranging from 0.002 to 0.100 mg/kg, copper from 0.507 to 10.527 mg/kg, lead from 0.033 to 0.695 mg/kg, nickel from 0.952 to 1.511 mg/kg, and zinc from 4.472 to 26.109 mg/kg (Rahimi et al. 2024). A separate dataset estimates lead intake at 0.1398 µg/kg body weight per day (with nickel at 0.1031 µg/kg body weight per day), which presents a relatively tighter margin of exposure compared to dairy products (Pipoyan et al. 2023). In individual case reports, chocolate was identified as a potential source of arsenic. In one report, arsenic biomonitoring in exposed individuals detected concentrations of 0.3–0.4 mL/L in blood and 5.3–5.8 mL/L in urine, and high maternal exposure was linked to foetal mortality (Konstantinidis et al. 2023).

While direct comparisons across studies are limited due to context and unit variations, these findings suggest that coffee (for lead exposure) and certain cocoa/chocolate products (for arsenic exposure) can be significant contributors to specific consumption patterns.

The heterogeneity in sampling frames, origins, and analytical methods (and occasional non-reported risk fields) restricts strict cross-study comparability. Furthermore, biomonitoring entries (e.g., arsenic levels in chocolate consumers) reflect specific contexts that may not be generalized. Future studies should aim to harmonize consumption scenarios, incorporate distributional intake reporting, and prioritize the investigation of speciation (e.g., inorganic arsenic) where feasible.

3.3.3. Polycyclic Aromatic Hydrocarbons (PAH)s

Across cereal-based fermented foods, PAH intakes and, when available, risks were described exclusively for bread (Table 4). Benzo[a]pyrene (BaP) represented the main quantified compound, with reported mean daily intakes ranging from 48.1 ± 20.4 ng/day to 4 × 10⁻⁵–1.02 µg/kg bw/day. Carcinogenic risk values indicated that bread consumers were exposed to moderate or high carcinogenic risk levels through the intake of PAHs (Asadi Touranlou et al. 2024). Grouped PAH fractions (PAH4, PAH8, PAH16) were also reported but lacked quantitative intake or risk data. Overall, available results suggest that, although bread constitutes a relevant contributor to PAH dietary exposure, the limited number of studies and incomplete reporting of risk metrics prevent a comprehensive risk assessment.

3.3.4. Acrylamide

Acrylamide, which has recently been the subject of intense research due to its association with increased carcinogenic, neurotoxic and genotoxic risks from exposure to heat-treated foods, contributes to the risk of fermented foods from the consumption of bread, coffee, and tea. Table 5 summarizes the reported acrylamide intake from fermented foods (bread, brewed coffee, tea).

Bellicha et al. (2022) reported that the main contributors to acrylamide intake were coffee (mean ± SD: 10.4 ± 13.6 μg/d), potato fries and chips (mean ± SD: 10.1 ± 16.3 μg/d), pastries and cakes (mean ± SD: 3.1 ± 3.3 μg/d), and bread (mean ± SD: 2.1 ± 1.6 μg/d). Overall, food groups that can be considered fermented foods (coffee, bread, pastries, and cakes) contributed approximately 56% of total intake. A borderline significant positive association was observed between dietary acrylamide exposure and breast cancer risk overall. The association was observed more specifically in premenopausal females. However, data on the association between acrylamide intake from fermented foods (bread, pastries, coffee) and breast cancer are not specifically available in this study. Similar findings were observed in a Polish study (Mojska et al. 2010): the main sources of dietary acrylamide are bread (45%), French fries and potato crisps (23%), and roasted coffee (19%). Acrylamide exposure from bread varies between 0.16 and 0.25 µg/kg body weight/day, and from coffee approximately 0.08 µg/kg body weight/day for total age group. The observations are similar to other European countries. The risk characterisation of acrylamide intake from fermented foods could not be extracted from these data.

Data from a Japanese study (Kawahara et al. 2018) revealed that coffee, along with potato and vegetable cooked at high temperature, contribute significantly to long term dietary acrylamide exposure. The calculated MOE values for the median and 95th percentile of probable long-term average dietary acrylamide exposure using the non-carcinogenic endpoint were 2792 and 1648, which indicates that neoplastic risk cannot be excluded. Separate data for fermented foods cannot be extracted from this study.

According to the study by McCullough et al. (2019), based on the nutrition cohort of 151,337 participants, dietary acrylamide is not associated with renal cell cancer (RCC) risk. As the main contributors of acrylamide in American populations were identified French fries (23% of intake), coffee (15%) and bread (10%). Acrylamide intake was not associated with RCC among men, women, or both combined after adjusting for age and sex or in multivariable analyses.

3.3.5. Other Chemical Hazards

The analysed studies of various hazardous organic and inorganic compounds identified in fermented foods reported the presence of multiple contaminants, including dioxins, bisphenol A, bromates, nitrosamines, alkaloids, ultraviolet filters, furan, and trace elements such as arsenic, among others. Table 6 summarises, for each food type and hazardous organic and inorganic compounds, the reported intake and risk metrics used for risk characterisation.

In dairy products, including cheese, yogurt, and dairy desserts, several dioxin-like compounds were detected, with estimated average daily intakes ranging from 7.7 to 233 g/day. Although some authors have suggested that fermentation processes may contribute to the degradation of dioxins, Loutfy et al. (2006) observed that, in the case of cheese, concentrations of dioxins and dioxin-like contaminants exceeded the acceptable limits. Risk characterization indicated a potential health concern, as exposures remained above the tolerable daily intake (TDI) established for these compounds. Similarly, bisphenol A (BPA) was detected in cheese and yogurt at concentrations ranging from 521 to 640 ng/mL. Although the calculated hazard quotient (HQ) suggested a need for consideration of non-carcinogenic risks. Among other hazardous organic compounds detected in dairy products are antimicrobial residues. In the study by Alenezi et al. (2024), residues of ampicillin, tetracycline, oxytetracycline, and amoxicillin were found at concentrations ranging from 2.44 to 220.3 µg/L. However, the overall risk associated with these antibiotic residues in dairy products was considered low.

In bread and bakery products, potassium bromate (KBrO₃) was frequently identified as the main contaminant. Concentrations ranged from 0.0000282 to 49.19 mg/kg/day depending on the product type, with the highest values observed in wheat bread. The corresponding hazard ratios (HRs) varied widely, ranging from 0.166 to over 295,000. A significantly higher risk was observed for consumption of bread from Cameroon (Ncheuveu Nkwatoh, Fon, and Navti 2023) than for the studied products from Ghana (Ayembilla et al. 2024).

In fermented meat products, notable contaminants included N-nitrosodimethylamine (NDMA) in salt-fermented pollock, with estimated exposures of 0.002–0.090 g/kg bw/day. Monte Carlo-based cancer risk simulations (mean = 1.0 × 10⁻⁵; 95th percentile = 3.73 × 10⁻⁵) indicated a low but measurable carcinogenic potential associated with frequent consumption (Sun et al. 2023). Additionally, elevated urinary iodine concentrations (530–2124 mg/L) were observed in consumers of chorizo, suggesting a possible association with thyroid hormone dysregulation due to iodide or thyroid-active compounds (AOR = 1.88) (Conrey et al. 2008).

In several studies (Chen, Huang, and Wu 2024; De Mul et al. 2008; Fanaike et al. 2019; Scholl, Huybrechts, et al. 2012; Scholl, Humblet, et al. 2012), different types of foods, including fermented products, were analysed collectively, rather than as separate categories. Consequently, the detected contaminants reflected the combined diversity of the sampled products rather than the specific composition of fermented foods. Reported substances included benzophenone-type ultraviolet filters (BP, BP-3, PBZ, 4-MBP), furan, and pyrrolizidine alkaloids, all occurring at levels generally below toxicological concern for adults. However, for children, some MOE values for furan exposure approached the 100–10,000 range, suggesting a potential concern for sensitive groups (Scholl, Huybrechts, et al. 2012).

Coffee was mentioned in a single study as a source of trigonelline, a natural alkaloid with an estimated daily intake of approximately 13.8 g/day, which was not associated with any toxic or adverse health effects (Konstantinidis et al. 2023).

Two additional studies included products that could potentially be categorized as fermented, although this could not be confirmed based on the available descriptions. The study by Han et al. (2024) investigated homemade sausages and reported the presence of sodium nitrite, with a very low MOE. The reported dose of 3.5 g/kg was extremely high compared with intake standards (maximum 125 mg/kg in the final product). Similarly, in the study of Cvetković et al. (2019) on millet-based products, atropine and scopolamine were detected at concentrations of 0.23 ng/kg bw/day.

For several other studies no quantitative data on intake or risk characterization were provided. Nevertheless, the analysed fermented products can be considered in the context of their possible contamination with various compounds, such as verotoxin in yogurt (Morgan et al. 1993), caffeine in coffee (Adeleye et al. 2023), or phthalates in bread (Adeleye et al. 2023).

4. Discussion

This narrative systematic analysis was designed to answer two complementary questions: “Are fermented foods safe for humans?” and “What are the main microbiological and chemical hazards posed by fermented foods and the associated risks?”.

Fermentation is widely recognised as a food-processing strategy that can enhance safety and stability through technological approaches such as acidification, salt addition, reduced water activity, production of antimicrobial metabolites, and microbial competition. These properties can inhibit or suppress many pathogenic and spoilage organisms, contributing to safer foods and extended shelf-life (Chai and Chen 2025). Fermented foods are considered an established diet component with a generally favourable safety profile, despite safety depending on ingredient quality and manufacturing conditions rather than on fermentation alone, as empirically evidenced by the collected data: the microbiological hazards and outbreaks identified were rarely interpretable as being caused by fermentation itself; instead, they aligned with entry via the substrate (e.g., contaminated milk, meat, cereals, vegetables, fish) or with manufacturing practices and post-processing contamination. This perspective is particularly important for food products for which there is no final step to decrease the associated risk (e.g., many cheeses and cured meats), as the microbiological quality of the production environment and raw materials becomes a decisive factor in determining risk.

Across the retrieved literature, bacteria were the most frequently reported microbiological hazards, with recurrent reporting of Salmonella spp., pathogenic Escherichia coli and other Enterobacteriaceae, Campylobacter jejuni, and Clostridium botulinum. Listeria monocytogenes appeared less frequently but was associated with the most severe outcomes, consistent with its high case fatality in vulnerable groups. The product group most often implicated was fermented dairy, especially cheeses and yoghurt, followed by selected fermented meats, with sporadic entries in vegetable, cereal-based, fish, and mixed dishes containing fermented ingredients. Despite considerable heterogeneity in the context, a pattern was identified: hazards tend to enter early (raw milk or insufficiently controlled ingredients), persist when fermentation parameters are insufficient to suppress them, or be introduced after fermentation (e.g., slicing/handling, mixed dishes). This pattern is also consistent with the role of starter cultures as a safety tool, considering that the well-selected cultures can reduce opportunities for pathogen growth, but they cannot compensate for heavily contaminated substrates or poor hygienic conditions (Skowron et al. 2022). A key implication is the heightened relevance of production contexts. This is especially true in traditional or household fermentations. In these contexts, sanitation and temperature control are more variable (Cocolin et al. 2016). In such contexts, safety margins can narrow substantially, increasing the likelihood that pathogens or toxin-producing organisms present in the substrate will survive or proliferate. However, insufficient evidence was found to establish a clear picture of the relationship between production contexts and the associated risk of fermented foods.

Across the studies retrieved in our systematic analysis, chemical hazards were also reported across diverse fermented foods, namely mycotoxins, especially in cereal-based foods and dairy products, heavy metals, with product- and geography-specific drivers, process-related contaminants, e.g., PAHs and acrylamide in baked/smoked matrices, and a smaller set of other compounds, such as furan, UV filters and nitrosamines in specific contexts. The chemical hazards reported in fermented foods can be classified within two broad, policy-relevant categories: i) hazards primarily driven by the food matrix and upstream contamination and ii) hazards that may be influenced by processing choices, including fermentation-related pathways. This distinction is critical for understanding whether fermented foods are “safe”: the data indicate that most chemical hazards are not inherently created by fermentation but rather reflect contamination present in raw materials and primary production environments, with fermentation sometimes modifying, rather than initiating, exposure. In fact, fermentation and downstream processing can change concentrations by degradation, binding, redistribution, or concentration via moisture loss, so a matrix-based hazard can still be relevant at the consumption level (Assohoun et al. 2013).

Mycotoxins represent one of the clearest examples of hazards primarily driven by upstream contamination rather than fermentation biology. In the broader literature, mycotoxins in fermented products are largely attributable to fungal contamination of ingredients before processing, including during storage, and can persist because many mycotoxins are relatively heat-stable and may survive standard processing conditions (Tian et al. 2022). This framing aligns with the pattern observed in the obtained results, in which mycotoxin-related findings were concentrated in cereal-based products (where grain quality and storage are decisive) and in dairy, where aflatoxin M1 reflects carry-over from contaminated animal feed into milk and derived products (Pop et al. 2024). It is also important to mention that the fermentation can sometimes reduce mycotoxin concentrations through microbial binding or biotransformation, but these effects are variable and depend on strain, matrix, and process parameters, and, therefore, fermentation should not be assumed to be uniformly protective (Tian et al. 2022). From a risk management perspective, it is clear that the most effective strategy is to ensure the safety of raw materials (through good agricultural practices, storage control, supplier verification and targeted monitoring), rather than relying on fermentation to remediate contaminated ingredients (Loi et al. 2023). This supports that many chemical hazards associated with fermented foods reflect the food system upstream rather than an inherent hazard generated by fermentation.

The literature indicates that metal concentrations in dairy and other fermented matrices are primarily determined by upstream and processing-chain factors, particularly environmental contamination of raw materials and potential contributions from contact materials, rather than by microbial fermentation processes (Boudebbouz et al. 2021; Yan et al. 2022). Metals emerged in the obtained results as a highly context-dependent hazard class, with risk characterisation varying substantially by product type, geographic origin, and population subgroup (e.g., differences driven by consumption patterns and body weight assumptions in children and adults). This variability is consistent with broader evidence indicating that metal occurrence in milk and dairy products is shaped primarily by upstream environmental determinants, including contamination of soil, water, and feed, together with potential concentration effects during processing and differences in local production systems, thereby motivating structured monitoring and risk assessment across supply chains (Boudebbouz et al. 2021; Yan et al. 2022).

For the substances considered process-derived contaminants, the fermentation of the food is secondary to the contribution of high-temperature processing and/or smoking within the overall manufacturing chain. PAHs exemplify this distinction since they arise primarily from incomplete combustion and smoke deposition, and their occurrence in foods is therefore governed mainly by smoking-related variables (e.g., wood type, smoke-generation temperature and airflow, smoking duration, distance to the heat source, fat content and casing, and equipment design and maintenance), rather than by microbial fermentation pathways (Cheng et al. 2023; Ledesma, Rendueles, and Díaz 2016). Thus, PAHs reported in fermented products (e.g., smoked fermented meat products) should be interpreted as reflecting the smoking unit operation and its exposure conditions, not as hazards generated by fermentation. This is consistent with the previous published evidence showing that PAH profiles and loads in thermally processed meat products are largely driven by processing parameters and smoking technology (Onopiuk et al. 2021).

Acrylamide provides a parallel example of a contaminant determined primarily by thermal processing rather than fermentation. Acrylamide formation is dominated by Maillard chemistry during baking or roasting, shaped by precursor availability (particularly free asparagine and reducing sugars), temperature and time profiles and moisture conditions. Accordingly, it is best conceptualised as a thermal-processing hazard (Seyedzadeh-Hashemi et al. 2026). Nonetheless, fermentation can modulate acrylamide formation indirectly, most notably in sourdough systems, by altering precursor pools and dough pH, thus serving as a potential mitigation strategy within an otherwise heat-driven pathway (Ameur et al. 2024; Seyedzadeh-Hashemi et al. 2026). Thus, the evidence indicates that when PAHs and acrylamide are reported in products classified as fermented, their occurrence should be interpreted primarily as a function of co-occurring unit operations in the manufacturing chain (e.g., smoking, baking, roasting) and their associated process conditions, rather than being attributed to microbial fermentation as the causal determinant.

Another example is Furan, a recognised heat-induced processing contaminant, reported especially in coffee and heat-treated foods. While coffee may be consumed in fermented-product contexts (e.g., fermented coffee processing), furan formation is primarily driven by roasting and thermal degradation of precursors rather than fermentation (Batool et al. 2023). As reflected in the obtained results, this again supports the broader conclusion that several hazards associated with fermented foods are, in practice, indicators of thermal processing and product-specific manufacturing chains. Nitrosamines are relevant to specific product categories, most notably cured meats, where nitrite and/or nitrate use, processing, and storage conditions influence formation pathways. Regulatory and scientific bodies have highlighted nitrosamines as genotoxic carcinogens of concern in foods, supporting the need for continued mitigation and surveillance in relevant processed meat chains (Chain et al. 2023). Recent literature has also summarised formation mechanisms and control strategies, framing nitrosamines as a hazard shaped by formulation and processing conditions rather than by fermentation as a microbial transformation alone (Rot et al. 2025).

A small number of studies addressed compounds such as UV filters in specific contexts. These are best interpreted as emerging contaminants potentially linked to packaging materials, environmental contamination, or complex supply-chain exposures rather than fermentation itself. Evidence highlights that migration and contamination depend on packaging composition, food properties (fat content, acidity), time-temperature, and regulatory compliance, underscoring the importance of considering packaging as part of the overall risk system (Rot et al. 2025; González-López et al. 2023). Given the emerging nature of these hazards and the fragmented reporting landscape, a lack of retrieved findings should not be interpreted as the absence of risk, but rather as potentially reflecting limited data availability and under-investigation at present.

Among the chemical hazards identified, biogenic amines (BAs) represent the clearest example of a hazard that can be linked directly to the fermentation ecosystem, namely, the presence and predominance of decarboxylase-positive microorganisms, the availability of free amino acid precursors, and process conditions that permit microbial growth and enzymatic activity (Hazards 2011; Gardini et al. 2016). BAs such as histamine, tyramine, putrescine and cadaverine are formed through microbial amino acid decarboxylation, and their accumulation is therefore highly matrix-dependent, being most relevant in protein-rich foods (e.g., cheeses, fermented meats, fish sauces) where precursor amino acids are abundant and where ripening and storage can provide time for amine production (Gardini et al. 2016; Barbieri et al. 2019). Importantly, the BA hazard is not an inherent or inevitable consequence of fermentation, but rather a function of controllable ecological and technological factors. Hygienic quality of the raw materials and processing environment (which influences the initial load and diversity of BA-producing microorganisms), the choice of starter cultures (including avoidance of decarboxylase-positive strains and, where appropriate, use of strains with amine-degrading capacity), and the management of critical parameters such as temperature, salt concentration, pH and acidification kinetics and oxygen availability, which collectively determine whether BA-formers can proliferate and express decarboxylase activity (Gardini et al. 2016; Hazards 2011). Accordingly, BA formation is best framed as a predictable and manageable chemical risk within fermented foods, for which risk reduction can be achieved through validated processing design and verification, reinforcing that hazards attributed to fermented foods frequently reflect upstream or process-control determinants rather than fermentation as an intrinsically hazardous step (Barbieri et al. 2019; Hazards 2011).

Taken together, the evidence gathered in the present systematic review supports a coherent and practically useful interpretation of safety in fermented foods. The hazard landscape is largely structured by where hazards enter the chain and which unit operations dominate exposure, rather than by fermentation as an intrinsically hazardous technology. Across food categories, the most frequently reported chemical risks are consistent with i) upstream contaminants that originate in primary production and raw materials (e.g., mycotoxins and metals), and ii) process-derived contaminants associated with smoking or high-temperature steps that may co-occur with fermentation (e.g., PAHs, acrylamide, furan). In contrast, the subset of hazards most plausibly linked to fermentation is comparatively narrow, with BAs representing the most emblematic example, driven by the presence of decarboxylase-positive microorganisms, precursor availability and permissive processing conditions. This aligns with recent evidence emphasising both the long history of safe consumption and the need to interpret hazards through a product- and process-specific lens, rather than attributing risk generically to fermentation (Marco et al. 2021; Pop et al. 2024). Additionally, production context is a major determinant of both microbiological and chemical hazard profiles, yet it is insufficiently captured in the published literature in a way that supports structured comparison. The current evidence is most often reported by food category, while information on whether the product was produced in a household, artisanal, or industrial setting is either absent, inconsistently described, or not operationalised in risk characterisation. This matters because the same fermented food category can embody markedly different safety envelopes depending on context. Household fermentation may involve higher variability in hygiene, temperature control, salt application, and fermentation kinetics, artisanal production can be highly heterogeneous, sometimes combining traditional practices with variable degrees of standardisation and industrial processes typically rely on defined starters, validated time, temperature, salt and pH targets, and environmental monitoring schemes. These contextual differences plausibly influence not only pathogen entry and persistence (i.e., through raw-material handling, sanitation, cross-contamination, and post-process exposure), but also the formation and accumulation of certain fermentation-linked metabolites such as BAs, which are strongly shaped by microbial ecology and process conditions (Barbieri et al. 2019; Gardini et al. 2016; Laranjo, Potes, and Elias 2019). Consequently, future research would benefit from moving beyond categorisation by food type alone and adopting a standardised reporting framework that systematically captures production setting, critical control parameters, starter culture use (including strain-level characterisation where relevant), and post-fermentation handling conditions. Such harmonisation would improve comparability across studies and enable a more rigorous differentiation between hazards attributable to contaminated inputs, process deviations, and fermentation ecology. The same approach would also strengthen the interpretability of risk metrics by anchoring them to transparent intake scenarios and population subgroups and by clarifying whether reported values reflect typical manufacturing practices or exceptional conditions. This is particularly relevant for hazards that are demonstrably upstream-driven, such as metals and mycotoxins, where supply-chain context and environmental determinants dominate concentrations in the finished product, and where relying on downstream processing to offset contaminated inputs is unlikely to be a robust safety strategy. Finally, the increasing uptake of genomic and high-resolution analytical methods offers a timely opportunity to improve attribution and verification in complex fermentation environments, supporting more precise identification of hazards, routes of contamination, and the effectiveness of controls (Hazards et al. 2019).

Overall, these considerations reinforce the rationale that fermented foods are best viewed as generally safe when produced under appropriate controls, with risk concentrated in identifiable scenarios, such as contaminated raw materials, inadequate management of critical parameters, and variable hygiene and temperature control in less standardised settings.

5. Conclusions

These results illustrate the complex balance required to assess dietary exposure to heavy metals in fermented foods. While levels of some metals approach the thresholds of concern, the overall risk indices, supported by biomonitoring evidence, indicate risks that can be managed with sensible dietary choices. This analysis highlights the importance of continued monitoring, refining risk assessments, and considering vulnerable populations in food safety and regulatory frameworks.