The detailed steps of the search and study selection are presented in Fig. 1. The primary search identified a total of 139 records in MEDLINE (46), Scopus (42), Web of Science (30), EMBASE (9), ScienceDirect (10) and grey literature search (2) based on the title screening. After removing duplicate reports, 61 studies were retained. Further screening of the titles and abstracts led to the exclusion of 43 studies and the inclusion of 18 studies. Following a full-text evaluation, 1 additional study was excluded. As a result, 17 studies were deemed eligible for qualitative synthesis, and 12 were selected for the meta-analysis. The characteristics of the selected studies are shown in Table 1. The kappa coefficient of 0.95 revealed the perfect agreement between 2 investigators.

Schematic flowchart of the conducted steps for selecting studies based on the PRISMA statements

The quality assessment of the studies according to the Newcastle-Ottawa Scale is summarised in Table 1.

A sensitivity analysis was conducted to mitigate potential publication bias. This involved systematically excluding each study to evaluate its impact on the overall meta-analytic results, including the main summary estimate and the I² statistic for heterogeneity. The findings were consistent across analyses, demonstrating a relatively low sensitivity to these exclusions and reinforcing the credibility of the meta-analysis results. The illustrated Funnel plot showed us that some studies were out of the expected line and identified potential outliers [10, 17, 18] (Fig. 2). The results of the initial forest plot, which included 17 case-control studies [fixed effects model OR = 2.12, 95%-CI [1.77; 2.54], p-value < 0.00001], had high levels of heterogeneity [I2 = 72%], which also showed there could be some outlier studies (Fig. 3-A). The identified outliers were excluded and the forest plot was illustrated again, as the results were identical [fixed effects model OR = 3.92, 95%-CI [2.98; 5.16], p-value < 0.00001] although with significantly lower levels of heterogeneity [I^2 = 22%] they validated each other (Fig. 3-B).

Funnel plot of included studies for investigation of publication bias

Initial Forest plot of the association between BLV infection and breast cancer risk (A). Final forest plot of the association between BLV infection and breast cancer risk after excluding the outlier studies (B)

The possible parameters associated with the detection rate of BLV were examined in the subgroup analysis (Fig. 4). These parameters include detection method, detection target, sample type, and study location. Although our initial forest plot and funnel plot revealed outliers, these were not excluded from the subgroup analysis to avoid selection bias. The subgroup analysis of the forest plots indicated that the PCR method and the gag gene as the target were not the best choices for these studies, as both demonstrated controversial results compared to others (Fig. 4-A, B). Khalilian et al. reported results that differed from those of other sample type subgroups, likely due to their investigation of BLV in formalin-fixed, paraffin-embedded (FFPE) tissues for cases and blood samples for controls [10] (Fig. 4-C). The interpretation of results would have been more straightforward if the same sample types had been used for both cases and controls. Furthermore, subgroup analysis based on study location did not produce conflicting results (Fig. 4-D). This may be attributed to the exclusion of studies with insufficient data—those with the most contentious outcomes—from the search strategy, preventing their inclusion in the meta-analysis.

Forest plot of the association between BLV infection and risk of breast cancer based on subgroups: detection method (A), detection target (B), sample type (C), and study location (D)

To explore the relationship between BLV and breast cancer, several case-control studies have been conducted across different countries. Summarising the results from these studies can help us better understand regional and ethnic factors that may influence the connection between BLV infection and breast cancer development.

In 2007, Buehring and co-authors published the first research paper reporting the presence of BLV DNA and proteins. The results showed that BLV was detected in 59% of breast cancer cases and 29% of controls. Interestingly, among the breast cancer samples, 69% exhibited BLV proviral DNA in accompanying non-malignant mammary epithelium. This finding suggested that the development of cancer might have been a rare and delayed event within a population of BLV-infected cells in breast tissue. These results provided an encouraging initial step in establishing a causal link between BLV and human breast cancer [26].

In a similar study conducted in 2015, Buehring et al. performed research to investigate the presence of BLV DNA in breast tissue samples. The results demonstrated that BLV DNA was detected in 59% of mammary epithelium samples from American women with breast cancer, which was significantly higher than the 29% found in normal controls. Notably, the frequency of BLV DNA in samples from women with premalignant breast cancer was found to be 38%, falling between the frequencies observed in breast cancer and normal-control samples. These findings support the hypothesis that BLV plays a role in developing cancer [15].

According to a study by Baltzell et al., women diagnosed with breast cancer in Texas were significantly more likely to have BLV DNA in their breast tissue compared to women with benign diagnoses or no history of breast cancer. Women with premalignant breast pathology but no cancer history were found to have an increased risk of having BLV DNA in their breast tissue. The study’s attributable risk of 51.82% suggests that BLV might be responsible for at least half of the breast cancer cases in the studied population. It is worth noting that Texas, where the study subjects are from, is known for its high beef and dairy consumption and thriving cattle industry. 38% of buffy coat cells from the West Coast of the United States have been found to contain BLV. This supports previous findings that have revealed a significant association between BLV DNA in breast tissue and a breast cancer diagnosis [16].

In contrast to previous results, a recent study by Amato et al. found no evidence of BLV DNA in fresh-frozen breast cancer tumours from patients at a hospital in Vermont. This study suggests a low prevalence of BLV in the patient population [27]. However, it is important to note that the reliability and accuracy of these negative results are still to be determined.

Giovanna et al. conducted a case-control study that interestingly showed a higher percentage of BLV DNA detection in the control group. The researchers noted that while the presence of BLV genes in human breast tissue was confirmed, it should be clarified as a possible promoter of malignancy processes in this tissue. The study authors raised the point that their results seemingly contradicted the findings of previous studies [17].

Olaya-Galán et al. conducted an observational case-control study in which the researchers used Nested PCR, In-situ PCR, and immunohistochemistry to detect BLV in blood and breast tissues. The results showed that BLV was more prevalent in the cases group (61.3%) compared to the controls (48.2%). The study confirmed a statistically significant association between BLV and breast cancer, with the virus found in both the blood and breast tissues of participants. Therefore, BLV was identified as an intermediate risk factor for breast cancer in Colombia [18].

Schwingel et al. investigated the presence of the BLV genome in breast cancer tissues in south Brazil. The results showed that BLV DNA was more common in breast cancer tissue (30.5%) compared to healthy breast tissue (13.9%). The researchers suggested that the link between BLV and breast cancer is stronger than the links to lifestyle and reproductive history. They also noted that dairy products are more commonly consumed in south Brazil than in comparison to other regions of the country. Based on these findings, the study suggests that BLV may be a potential factor that increases the risk of breast cancer in women [19].

A study conducted by Delaramina et al. amplified BLV proviral genes from breast tumour samples and healthy control samples from women. The results showed a positivity rate of 95.9% in tumour samples and 59% in healthy tissue samples. This evidence confirms the presence of the BLV genome in the breast tissues of women in the state of Minas Gerais. It indicates a statistically significant positive association between BLV infection and breast cancer within this population [20].

In a study by Buehring and colleagues, it was revealed that 80% of women who had been diagnosed with breast cancer had BLV DNA in their breast tissue, compared to only 41% of women with no history of breast cancer. The results also showed that 60.4% of women who tested positive for BLV and had not been diagnosed with breast cancer later developed the disease, whilst only 14.6% of BLV-negative women developed breast cancer. Hence, it suggests a possible temporal relationship between BLV infection and the subsequent development of cancer. In 74.2% of breast cancer patients, BLV infection was present years before the diagnosis, indicating that BLV might play a role in the development or acceleration of breast cancer [21].

A study conducted by Lawson and Glenn found that BLV was present in 78% of 23 benign breast specimens and 91% of 22 subsequent breast cancers in the same patients. The presence of BLV was confirmed by sequencing the products of standard PCR. As the prevalence of this virus is so high in both benign and later breast cancer cells, it is likely to be also present in other virus-positive benign and cancerous cells [22].

Khalilian et al. conducted a study involving 400 samples, comprising 200 breast cancer-suspected tissue samples and 200 blood samples from women without breast cancer, collected from two hospitals in Qom Province, Iran. Of the breast cancer-suspected samples, 172 were confirmed malignant. Using Nested PCR, the study detected BLV tax and gag genes in 30% and 8% of the malignant tissue samples, respectively. Additionally, 16.5% of the blood samples from women without breast cancer tested positive for BLV. The authors proposed that the Nested PCR technique could help establish a connection between human breast cancer and BLV infection in cattle [10].

A recent study conducted by Dabaghi et al. in 2022 investigated the presence of BLV in breast tissue and blood samples. The study used the Nested PCR method to detect BLV infection and showed that 13% of the blood samples and 8% of the breast paraffin tissue samples were infected with BLV. Although there was a notable relationship between BLV infection and breast cancer in the studied population’s paraffin tissue samples, more blood samples tested positive for this virus. Therefore, blood samples are preferable for detecting this virus in patients [23].

A study conducted in China by Zhang et al. had findings that contradicted the previous dominant beliefs. The researchers stated that there was no association between breast cancer and BLV. However, it is crucial to be cautious in interpreting these findings. The Chinese scientists used a commercial BLV testing kit designed for cows on human blood samples, which could have affected the results. As Buehring and colleagues pointed out, these commercial kits are not intended for testing human sera [28]. Some human sera could yield negative results because the final detection step involves a labelled antibody to bovine rather than human immunoglobulin [29].

A study by Saito et al. reported the absence of BLV DNA in Japanese human cell lines. The authors examined DNA extracted from 145 cell lines but did not detect BLV DNA. This raises questions about the optimality of the protocol as previous research has shown the presence of BLV DNA in some proportion, not 0%. Saito et al. pointed out that a potential flaw was using a PCR method designed for sheep cell lines, not humans. It was mentioned earlier that Buehring et al. raised concerns about the study by Zhang et al., where a bovine ELISA was used on human samples [28, 29]. Additionally, Saito et al. admitted that they only used two breast cancer cell lines [30].

Similarly, Yamanaka et al. used PCR to examine the presence of BLV proviral DNA in human blood and breast cancer tissue samples and found all the samples yielded negative results [31].

A recent study by Khan et al. examined the presence of BLV in human breast tissue through Nested PCR by amplifying tax and gag genes. The study results showed that BLV genes were positive in 26.8% of the samples from breast cancer patients, while only 10% of the samples without cancer were positive. Therefore, the study suggests that there may be a relationship between the presence of the BLV gene and breast cancer [24].

The results of a study by Khasawneh et al. showed that BLV was detected in 18.4% of the breast cancer samples and none of the control samples tested positive for BLV. The study also investigated the relationship between BLV and breast cancer molecular subtypes, finding the most positive cases in luminal A and luminal B patients. However, the correlation with the HER2 subtype was not statistically significant due to a small sample size. It was observed that larger tumours were more frequently associated with metastatic tissue in sentinel lymph nodes. Most patients had tumours smaller than 5 cm, with a notable prevalence of grades 2 and 3, which are indicative of a poorer prognosis. However, no significant correlation with BLV DNA was identified, likely due to the small sample size analysed [25 ].