Introduction

The global burden of cancer presents a formidable public health challenge, ranking as the second leading cause of death worldwide, second only to heart disease.1 Among various cancer types, digestive system cancers (DSCs), including gastric cancer, colorectal cancer, hepatocellular carcinoma, oesophageal cancer, pancreatic cancer and gallbladder cancer, impose a significant burden and account for a substantial proportion of cancer cases globally.2 However, it is equally crucial to acknowledge the substantial health threat posed by cancers occurring outside the digestive system, collectively termed non-DSCs (NDSCs). These encompass a diverse array of cancer types, such as lung, breast, prostate, bladder, ovarian, cervical and kidney cancers, among others.3 The epidemiology of cancer is shaped by a complex interplay of genetic, environmental and lifestyle factors. Among these factors, diet has emerged as a key modifiable risk factor associated with the development of both DSCs and NDSCs.4

As a dietary factor, pickled foods have become an integral part of modern diets, accounting for a significant proportion of the overall food consumption across the globe. These foods are typically defined as those that have undergone preservation techniques such as salting, smoking, curing or the addition of preservatives to extend their shelf life.5 The widespread consumption of pickled foods has raised concerns about their potential impact on cancer, particularly regarding DSCs, including nasopharyngeal carcinoma,6 oesophageal cancer, stomach cancer,7 8 liver cancer9 and colorectal cancer.10 Nonetheless, no comprehensive review or meta-analysis has yet provided complete information on the effect of pickled food consumption on the risk of NDSCs.

Therefore, to comprehensively address this research gap, our objective aims to provide a robust and evidence-based evaluation of the association between pickled food intake and the development of NDSCs by systematically reviewing and synthesising the existing literature. By understanding the potential risks associated with preserved food consumption in relation to NDSCs, this study extends the existing knowledge of dietary factors influencing cancer development and assists in the formulation of preventive strategies.

MethodsInclusion and exclusion criteria

The study protocol was registered with the International Prospective Register of Systematic Reviews Database (PROSPERO) in June 2023 and was updated in July 2024 (CRD42023434186). Observational studies examining the relationship between pickled food consumption and the development of NDSCs were included in this analysis. The NOVA classification is commonly used to define the degree of food processing. This system classifies foods into four categories, ranging from unprocessed or minimally processed items to ultra-processed products, based on the extent of processing.11 Pickled foods are classified as processed foods within the third category, as they undergo modification through the addition of various ingredients while maintaining the fundamental qualities of the original food product. This preservation method involves treating items with salt, sugar, oil, vinegar, alcohol, smoking, drying or other preservatives to prolong their shelf life or improve their taste. The processing of pickled foods may encompass fermentation, as exemplified by sauerkraut and kimchi, or non-fermentation, as demonstrated by pickled cucumbers.12

Our research question was formulated based on the PECOS principle13 (online supplemental table 1). The PECOS elements were as follows: P (participants): all individuals, both children and adults. E (exposure): ‘high consumption’ of pickled food, defined as the highest intake levels reported across studies, including the upper third, fourth or fifth percentiles. This approach allowed for consistent comparative analysis across studies, taking into consideration the inherent variability in dietary patterns across different populations. C (comparison): ‘non-consumption or low consumption’ of pickled food, defined as the lowest intake levels, including the lower third, fourth or fifth percentiles, or no consumption. O (outcome): the risks of various NDSCs, including lung cancer, breast cancer, prostate cancer, kidney cancer, bladder cancer, brain tumours, bone tumours, ovarian cancer, cervical cancer, pancreatic cancer, thyroid cancer, testicular cancer and lymphoma. S (study design): observational studies, including cohort studies, case-control studies and cross-sectional studies. Only English-language articles were included. The exclusion criteria were (1) involved non-human subjects; (2) the absence of primary data analyses (eg, letters, editorials or narrative reviews); or (3) did not provide a clear methodology for data extraction (eg, ORs, relative risks and HRs).

Searching methods and screening criteria

Two reviewers (JY and PY) conducted a comprehensive search of multiple databases, including PubMed, Embase, Web of Science and the Cochrane Library, from inception to 1 July 2024. Both MeSH terms and text words were used in combination across two primary term blocks: ‘pickled food’ and ‘cancer.’ The full search strategy is described in the supplemental materials (online supplemental table 2). After removing duplicates, the remaining studies were thoroughly screened based on titles, abstracts and full texts. Any discrepancies were resolved through consensus or consultation with a third reviewer (CL). Subsequently, two researchers (HL and LX) extracted data from the included studies using a standard form. The extracted data included the first author’s name, year of publication, study location, age, gender, study sample size, duration of follow-up, pickled food exposure assessment methods, outcome measures for NDSCs and reported risk estimates (crude estimates and adjusted estimates) with 95% CIs. Any discrepancies in data extraction were resolved through discussion and inspection of the original data by a third researcher (CL).

Quality and reporting bias assessment

The National Institutes of Health (NIH) quality assessment tools were used to assess the risk of bias for each study. We documented outcome-specific assessments in online supplemental tables 3 and 4 (available at https://www.nhlbi.nih.gov/health-topics/study-quality-assessment-tools). Each item in the study was rated as ‘yes’ (1 point), ‘no’ (0 point), ‘not reported’ or ‘not applicable’. Items that were adequately described received one point, while those lacking sufficient description or failing to meet quality criteria received zero points. Items lacking clear descriptions were described as ‘not reported’, and those not meeting the assessment criteria were set as ‘not applicable’. The maximum score was 14 points for longitudinal studies and 11 points for observational and cross-sectional studies. Longitudinal studies were classified as high quality (>9 points), medium quality (4–9 points) and low quality (<4 points). Cross-sectional studies were classified as high quality (>7 points), medium quality (3–7 points) and low quality (<3 points). Medium-quality and low-quality studies were considered high-risk publications. Two reviewers (CL and FL) conducted the evaluations, resolving discrepancies through discussion.

Data synthesis and statistical analyses

When more than two studies investigated the same type of cancer, we performed analyses using the DerSimonian–Laird method with a random-effects model in R software (V.3.4.0; R: the R Project for Statistical Computing, Vienna, Austria). The primary analysis focused on comparing the ORs between the highest and lowest categories of pickled food exposure. The definitions of ‘highest’ and ‘lowest’ exposure categories were derived from the exposure classifications provided by the original authors of each included study. For consistency in defining exposure levels, ‘high consumption’ of pickled foods referred to the highest intake levels reported in each study, including the upper third, fourth or fifth percentiles, while ‘low consumption’ corresponded to the lowest intake levels, encompassing the lower third, fourth or fifth percentiles, or no consumption. Online supplemental table 5 provided a summary of the lowest and highest pickled food consumption in each included study. Due to this approach, it is important to note that these categories may exhibit considerable heterogeneity across studies, as they were not standardised in the included literature. Each study’s definition was based on its specific context, population dietary habits and methodological framework.

A random-effects model was employed, and significance was set at p<0.05 (two-tailed). Heterogeneity was quantified using the Q test14 and the I2 score.15 Subgroup analyses were planned based on the subtype of pickled food if sufficient data were available. To assess the robustness of our findings, we conducted sensitivity analyses using two distinct approaches: the leave-one-out method and an analysis restricted to studies reporting crude ORs.

Reporting bias assessment and certainty assessment

A funnel plot and Egger’s test were performed to assess publication bias and small-study effects if more than six studies reported data on the same outcome.16 The certainty of the evidence was evaluated using the Grading of Recommendations, Assessment, Development and Evaluation approach,17 which categorises the certainty into four levels: high, moderate, low or very low. Initially, all outcomes were assigned a low quality due to the observational nature of the included studies (online supplemental table 6). This initial rating could be adjusted based on prespecified criteria. Evidence quality was upgraded for a clear dose-response gradient, a large magnitude of effect (OR≥2 or OR≤0.5) not fully explainable by confounding factors, or if control for plausible confounding did not change the effect estimates significantly. Conversely, the quality was downgraded under several conditions, such as a majority of studies showing a high risk of bias (NIH quality score<4), substantial heterogeneity (I²≥50%) unexplained by sensitivity or subgroup analyses, factors related to the population, intervention or outcomes limited generalisability (indirectness) or the 95% CI for pooled estimates crossing the minimally important difference, highlighting clinical relevance alongside statistical significance.

ResultsStudy selection

This meta-analysis was reported in accordance with the Meta-analysis of Observational Studies in Epidemiology reporting guideline18 (online supplemental table 7). The study selection process is summarised in figure 1. Initially, 3465 records were identified through searches. After assessing 98 full-text articles, a total of 51 observational studies were included in the study, consisting of 36 case-control studies, 14 cohort studies and 1 cross-sectional study.19–66

Figure 1

Literature search flow chart. Flowchart of the literature search and selection for systematic review and meta-analysis.

Study characteristics

The meta-analysis encompassed a total population of 2 518 507 individuals, with ages ranging from 18 to 90 years (online supplemental table 8). The quantitative analysis investigated breast cancer (n=23), prostate cancer (n=14), lung cancer (n=13), lymphoma (n=6), bladder cancer (n=9), kidney cancer (n=4), brain cancer (n=4), thyroid cancer (n=2), cervical cancer (n=2) and leukaemia (n=2). The sample sizes of the included studies ranged from 135 to 567 169 participants. The studies were conducted in various countries, with 8 from the United States, 10 from China, 9 from Uruguay, 3 from Japan, 2 from Canada, 2 from Spain, 3 from Sweden, 2 from the Netherlands, 1 from Poland, 1 from Indonesia, 1 from Italy, 1 from France, 1 from South Korea, 1 from Argentina, 1 from Greece (assuming Athens refers to Greece), 1 from Iceland, 1 from Australia, 1 from Serbia and 1 from Norway. Among all the included studies, 20 studies included both men and women participants, 17 studies focused on women, 10 studies focused on men, and four studies did not report gender proportions. Online supplemental table 9 provided comprehensive details on the confounding variables considered in the most adjusted model for each study, with a median (range) number of 10 (1–16) variables.

Quality assessment and risk of bias

According to the NIH quality assessment tool, 36 studies (71%) received a ‘high quality’ rating, while 15 studies (29%) were rated ‘medium quality’. None of the studies received a ‘low quality’ rating (online supplemental tables 3 and 4).

In the bias analysis of the meta-analysis, the funnel plot for lung cancer studies appeared symmetrical, indicating no significant publication bias. However, for breast cancer, prostate cancer, lymphoma and bladder cancer, the funnel plots exhibited varying degrees of asymmetry (online supplemental figures 1-4), suggesting the presence of publication bias. Egger’s test results further supported these observations. For lung cancer, Egger’s test showed no significant bias (p=0.881) (online supplemental figure 5). In contrast, significant bias was detected for breast cancer (p=0.006), prostate cancer (p<0.001), lymphoma (p=0.005) and bladder cancer (p=0.015).

Pickled food was associated with an increased risk of breast, prostate, lymphoma, bladder and kidney cancers

The meta-analysis revealed that the consumption of pickled foods is associated with an increased risk of several types of cancer (table 1). Breast cancer was investigated in a total of 37 706 cases among 1 095 935 participants from 23 studies, yielding an OR of 1.22 (95% CI: 1.07 to 1.39, I2=85.1%, p<0.01), with a very low certainty of evidence. Prostate cancer, with 14 studies from 28 398 cases among 818 562 participants, resulted in an OR of 1.38 (95% CI: 1.18 to 1.60, I2=75.9%, p<0.01), also showing a very low certainty. Lymphoma was analysed in six studies, including 5977 cases among 1 020 030 participants, and showed an OR of 1.12 (95% CI: 1.01 to 1.25, I2=55.8%, p=0.05) with very low certainty. Bladder cancer, with nine studies from 6267 cases among 702 161 participants, had an OR of 1.60 (95% CI: 1.23 to 2.07, I2=85.1%, p<0.0001) with very low certainty. Kidney cancer, from three studies involving 3279 cases among 611 050 participants, had an OR of 1.28 (95% CI: 1.13 to 1.45, I2=0%, p=0.56) and low certainty.

Table 1

Summary of the results for the meta-analysis

Pickled food was not significantly associated with an increase in risk for lung, brain, thyroid, cervical cancer and leukaemia

However, the consumption of pickled food was not associated with an increased risk of other types of cancer. Lung cancer, with 13 studies from 27 598 cases among 1 611 091 participants, had an OR of 1.19 (95% CI: 0.95 to 1.48, I2=87.3%, p<0.01), also showing very low certainty. Brain cancer was analysed in four studies, including 3237 cases among 596 361 participants, showing an OR of 0.97 (95% CI: 0.84 to 1.12, I2=52.9%, p=0.10) with very low certainty. Thyroid cancer was investigated in two studies from 618 cases among 567 855 participants, revealing an OR of 1.04 (95% CI: 0.80 to 1.36, I2=7.3%, p=0.30) and very low certainty. Cervical cancer was explored in two studies involving 150 cases among 567 334 participants, achieving an OR of 1.38 (95% CI: 0.71 to 2.66, I2=27.6%, p=0.24) with very low certainty. Finally, leukaemia was investigated in two studies, including 2106 cases among 591 940 participants, showing an OR of 1.12 (95% CI: 0.57 to 2.22, I2=93.5%, p<0.01) and very low certainty (online supplemental figure 6). The evidence quality for these cancers was uniformly very low, highlighting the need for further rigorous research to draw more definitive conclusions.

Subgroup analyses and sensitivity analyses of pickled foods on cancer risk

According to the NOVA classification of foods, pickled and preserved foods can be systematically divided into six categories: pickled food, preserved food, salted food, cured food, processed food and fermented food. Subgroup analyses were conducted to examine the relationship between specific types of pickled food consumption and the risk of NDSCs (table 2). For breast cancer, processed meat was associated with an increased risk (OR: 1.30, 95% CI: 1.16 to 1.47, I²=81.20%, p<0.01), whereas fermented food showed no significant risk increase (OR: 0.82, 95% CI: 0.65 to 1.04, I²=28.30%, p=0.24). Salted food displayed a neutral effect (OR: 1.00, 95% CI: 0.98 to 1.03, I²=1.90%, p<0.01). Processed meat was also associated with an increased risk of prostate cancer (OR: 1.33, 95% CI: 1.12 to 1.58, I²=77.80%, p<0.01), while salted food showed a stronger association (OR: 1.93, 95% CI: 1.47 to 2.53, I²=0.00%, p=0.43). Notably, pickled vegetables showed a high-risk ratio (OR: 4.19, 95% CI: 0.81 to 21.78, I²=90.20%, p<0.01), although the wide CI indicated significant heterogeneity. The risk of bladder cancer was elevated with processed meat (OR: 1.37, 95% CI: 1.13 to 1.65, I²=62.40%, p=0.03) and even more so with salted food (OR: 2.28, 95% CI: 1.78 to 2.92, I²=0.00%, p=0.82). An elevated risk of lung cancer was also observed with processed meat (OR: 1.43, 95% CI: 1.14 to 1.80, I²=82.20%, p<0.01) and salted meat (OR: 1.59, 95% CI: 1.08 to 2.35, I²=75.10%, p<0.01) (online supplemental figure 7). The findings highlight the potential carcinogenic risks associated with specific types of preserved foods and underscore the importance of nuanced dietary assessments in cancer epidemiology.

Table 2

Subgroup analysis of pickled food consumption and non-digestive system cancers

In the sensitivity analyses, the leave-one-out analysis corroborated the results of the main meta-analysis, showing no significant heterogeneity observed across studies. This consistency enhances the reliability of our meta-analysis findings (online supplemental figure 8). For the analysis limited to studies that reported crude ORs (online supplemental figure 9), the results were as follows: breast cancer (n=5) showed an OR of 1.39 (95% CI: 1.01 to 1.91), lung cancer (n=3) presented an OR of 0.96 (95% CI: 0.49 to 1.87), both of which aligned with the overall meta-analysis results, indicating a consistent effect across different study designs and reporting standards. However, prostate cancer (n=2) demonstrated an OR of 1.97 (95% CI: 0.58 to 6.73), which was consistent in direction with the main meta-analysis results but exhibited a much wider CI. These findings suggest potential variability in effect estimates when analysing only crude ORs and underscore the need for cautious interpretation, particularly when fewer studies are included in a subgroup analysis.

Discussion

This meta-analysis comprehensively assessed the association between pickled food consumption and the risk of various cancers, including a total of 2 518 507 individuals across diverse international studies. Our findings demonstrate a significant association between the consumption of pickled food and an increased risk of NDSCs, including breast, prostate, lymphoma, bladder and kidney cancers. Notably, processed and salted foods were associated with a higher risk of cancer, particularly prostate and bladder cancers. In contrast, pickled food consumption showed no significant effect on the risk of lung, brain, thyroid and cervical cancers and leukaemia. The results highlight the potential specificity of dietary risks associated with different cancer types.

The global diversity of the studies included in this meta-analysis, spanning multiple continents, provides a comprehensive view of the impact of pickled food consumption on cancer risk across different populations and cultural dietary habits. Notably, the studies were predominantly conducted in the United States and China, with 10 studies from each country, followed by 9 studies from Uruguay. This distribution highlights significant contributions from both Western and Eastern countries to the research field.67 The wide geographic spread is crucial, as it incorporates a variety of genetic backgrounds, environmental exposures and dietary patterns, which are essential for understanding the generalised effects of pickled food on cancer risk. Moreover, a substantial number of studies originated from Asia, particularly China and Japan, is significant given the higher prevalence of pickled food consumption in these regions compared with Western countries.68 This could potentially explain variations in cancer incidences linked to specific dietary practices. In addition, the high number of studies from Uruguay suggests a growing interest and recognition in emerging research hubs about the importance of diet in cancer epidemiology.

Previous research has reported mixed results concerning the association between pickled food consumption and the risk of DSC.7 9 10 The present meta-analysis highlights a significant variation in the impact of pickled food on the risk of NDSCs. Notably, pickled food was significantly associated with a higher risk of developing breast, prostate, lymphoma, bladder and kidney cancers. Some of the inconsistencies observed across different research findings may be attributed to earlier studies not differentiating between the types of pickled food and their specific processing methods. Compared with these studies, our analysis employed a more nuanced approach, considering various types of pickling processes and their unique risks. Moreover, our findings concur with the associations noted in several large-scale cohort studies that suggested similar associations, particularly regarding prostate and bladder cancers. Furthermore, the robustness of our results is supported by the leave-one-out sensitivity analysis, which indicated no significant heterogeneity and confirmed the consistency of our findings across different studies. The subgroup analysis, particularly for prostate cancer, highlighted the variability in risk estimates, suggesting that the type of pickled food and its specific preservation method could differentially impact cancer risk.

The observed increase in cancer risk may be attributed to several factors inherent to pickled food. These foods often contain high levels of nitrates and nitrites, which can be converted into carcinogenic nitrosamines during the pickling process.69 Additionally, the high salt content typical of these foods can lead to an increased risk of certain cancers by inducing hypernatraemia, which affects cellular metabolism and could contribute to DNA damage or inflammation. Furthermore, the preservatives and chemicals used in pickling, such as benzoates and other additives, may also play a role in carcinogenesis.70

The commonly held belief that fermented foods possess preventive health benefits primarily stems from their well-documented roles in promoting gut microbiome health. These foods are rich in probiotics, which are thought to enhance gut flora diversity and functionality, potentially leading to improved immune responses and reduced inflammation.71–73 Given these properties, it is often hypothesised that fermented foods could contribute to reducing the risk of various chronic diseases, including cancers, particularly those outside the digestive system.74 However, the results of our meta-analysis present a more nuanced picture, specifically in the context of NDSCs such as breast cancer. Our analysis included data on the consumption of fermented food and showed no significant protective effects. Several factors could potentially explain why our study did not find a clear protective effect of fermented foods against cancer. First, the health impacts of fermented food can be influenced by the variability in fermentation processes, including the strains of bacteria used, fermentation conditions and the food’s matrix. Such differences in fermentation practices can lead to inconsistent health outcomes. Additionally, individuals who consume fermented food often lead healthier lifestyles overall, representing a potential confounding factor.75 76 Lastly, the impact of fermented food can vary by genetic and regional dietary differences, which can affect how these foods interact with the body’s metabolic and immune processes.

Our study underscores the importance of dietary assessments in cancer risk evaluation. Public health guidelines might need to consider more detailed recommendations regarding the consumption of pickled foods. Future studies need to consider the complex interrelationships between diet, socioeconomic status, smoking habits, genetic factors and populations, as well as underlying biological mechanisms and variations across different groups. Understanding the interplay between the multiple factors is crucial for exploring the association between dietary factors and cancer risk. Besides, well-designed cohort studies and randomised controlled trials with detailed dietary assessments are needed to better understand the relationship between fermented food consumption and cancer risk. Therefore, integrating these variables into research designs will allow for a more comprehensive understanding of the influence of lifestyle, environmental and genetic factors on cancer risk.

This meta-analysis has several notable strengths, but it also has some limitations that warrant consideration. One of the primary strengths lies in its substantial sample size and the rigorous adjustment for numerous important dietary and lifestyle factors, which enhance the reliability and validity of the findings. However, it is essential to acknowledge certain limitations. First, while we have assessed the association between pickled food consumption and the overall risk of non-gastrointestinal cancers, caution must be exercised in interpreting these results due to the presence of publication bias. The variation in the quality of individual studies, with 29% rated as medium quality, suggests potential inconsistencies in study methodologies that could affect the overall analysis. Although none of the studies were rated as low quality, the presence of medium-quality studies indicates that some findings might be influenced by methodological weaknesses. Moreover, the analysis is also limited by the inherent nature of observational studies, which are susceptible to residual confounding despite adjustments. Therefore, the observed associations might be affected by unmeasured factors. Furthermore, the substantial heterogeneity observed in some cancer types, such as lung and bladder cancers, with I² values exceeding 85%, indicates that the effects may vary significantly across different populations and conditions, complicating the interpretation of pooled estimates. Additionally, the variation in reporting standards and the use of different measures of association across studies (eg, crude ORs vs adjusted ORs) might introduce discrepancies that could impact the synthesis of the data. While the subgroup analyses attempt to address this by analysing studies based on similarity in reporting, they still reflect an inherent challenge in combining diverse studies into a coherent meta-analytical framework.

A notable limitation is the absence of dose-response analysis, which restricts the ability to understand the relationship between the quantity of pickled food consumed and cancer risk. The primary reason for not conducting a dose-response analysis is the variability in measurement methods and reporting standards across studies, which complicates efforts to perform an effective analysis of dose-response relationships. Different studies used diverse approaches to quantify food intake, leading to inconsistencies that challenge the integration of data for such a detailed analysis. However, to facilitate a more comprehensive understanding and to aid further research, we have documented the dose and corresponding effect sizes reported by individual studies in online supplemental table 5. Furthermore, due to the insufficient number of studies providing detailed demographic characteristics, subgroup analyses based on population characteristics were not conducted, which could have provided deeper insights into the influence of different demographic factors on the association between pickled food consumption and cancer risk. This represents a significant gap in the current literature and should be addressed in future research to better understand the impact of these characteristics on the observed associations.