https://orcid.org/0000-0002-5995-8200
Figures
Abstract
Background
Early detection of pancreatic cancer is essential for improving survival rates. However, noninvasive diagnostic methods are lacking. Novel biomarkers, detectable through liquid biopsy, such as circulating tumor DNA (ctDNA), microRNAs (miRNAs), protein markers, and metabolites, hold promise for early diagnosis.
Methods
A systematic search of PubMed, Embase, Web of Science, and the Cochrane Library was conducted for studies published from January 2014 to May 2024. Studies were included if they evaluated novel biomarkers for early pancreatic cancer detection, reported diagnostic performance metrics (sensitivity, specificity), and had a QUADAS-2 score of ≥3. Data on study characteristics, patient demographics, biomarker types, and diagnostic performance were extracted following PRISMA guidelines. A bivariate random-effects model was used to calculate pooled sensitivity, specificity, positive likelihood ratio (PLR), negative likelihood ratio (NLR), and diagnostic odds ratio (DOR). The area under the summary receiver operating characteristic (SROC) curve assessed overall diagnostic accuracy. The primary outcome was the diagnostic accuracy (sensitivity and specificity) of novel biomarkers in detecting early-stage pancreatic cancer.
Results
A total of 43 studies involving 19,326 participants were included, with 2,749 patients having stage I or II pancreatic cancer. The pooled sensitivities and specificities were as follows:.
miRNA Biomarkers: Sensitivity 0.88 (95% CI 0.79-0.93), Specificity 0.91 (95% CI 0.82-0.95), DOR 72.68 (95% CI 26.64-198.24), AUC 0.95.
Protein Biomarkers: Sensitivity 0.79 (95% CI 0.70-0.86), Specificity 0.88 (95% CI 0.82-0.93), DOR 27.74 (95% CI 14.32-53.76), AUC 0.90.
Metabolite Biomarkers: Sensitivity 0.84 (95% CI 0.73-0.92), Specificity 0.85 (95% CI 0.81-0.88), DOR 31.76 (95% CI 12.38-81.48), AUC 0.90.
ctDNA Biomarkers: Sensitivity 0.65 (95% CI 0.48-0.81), Specificity 0.94 (95% CI 0.88-0.97), DOR 27.73 (95% CI 12.91-59.55), AUC 0.92.
Subgroup analyses showed combining biomarkers with CA19–9 improved diagnostic accuracy. Sensitivity analyses confirmed the robustness of the findings.
Conclusions
Novel biomarkers, particularly miRNAs and protein markers, demonstrate high diagnostic accuracy for early pancreatic cancer detection and have potential for clinical application in improving early diagnosis and patient outcomes.
Systematic review registration
https://www.crd.york.ac.uk/prospero/, Identifier: PROSPERO (CRD42024553633).
Citation: Zheng Z, Lu Z, Yan F, Song Y (2025) The role of novel biomarkers in the early diagnosis of pancreatic cancer: A systematic review and meta-analysis. PLoS One 20(5): e0322720. https://doi.org/10.1371/journal.pone.0322720
Editor: Shuai Ren, Affiliated Hospital of Nanjing University of Chinese Medicine: Jiangsu Province Academy of Traditional Chinese Medicine, CHINA
Received: January 23, 2025; Accepted: March 26, 2025; Published: May 23, 2025
Copyright: © 2025 Zheng et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the paper and its Supporting Information files.
Funding: The author(s) received no specific funding for this work.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Pancreatic cancer is among the most lethal malignancies, exhibiting the lowest survival rates across all cancer types [1]. It is projected to become the second leading cause of cancer-related deaths by 2030 [2]. This dismal mortality rate is primarily attributable to late-stage diagnoses, at which point curative treatments are often unfeasible. Early-stage surgical resection remains the principal potentially curative intervention for pancreatic cancer [3]. However, early-stage disease is frequently asymptomatic or presents with nonspecific symptoms such as fatigue, indigestion, and changes in bowel habits [4], complicating early detection efforts. One of the greatest challenges in the management of PDAC is achieving early detection in high-risk individuals and accurately diagnosing patients who present with suspected symptoms [5]. Consequently, over two-thirds of patients present with regional or distant metastases at diagnosis [6]. The five-year survival rate for patients diagnosed with localized pancreatic cancer is significantly higher than for those with advanced disease stages [7]. Therefore, early detection and diagnosis are critical for improving survival outcomes in pancreatic cancer.
Currently, effective methods for early detection are lacking. Imaging modalities such as computed tomography (CT), magnetic resonance imaging (MRI), and endoscopic ultrasound-guided fine-needle aspiration are limited by cost, invasiveness, and suboptimal sensitivity and specificity, especially in early-stage disease [8]. The most widely used noninvasive biomarker, serum carbohydrate antigen 19–9 (CA19–9), has limited sensitivity and specificity for early pancreatic cancer detection [9,10]. Elevated CA19–9 levels can occur in benign conditions like pancreatitis and other malignancies, leading to false positives [11]. Additionally, approximately 5% to 10% of the population are Lewis antigen-negative and do not produce CA19–9, resulting in false negatives [12]. Research and application of novel biomarkers for the early detection of potentially curable pancreatic cancer remain in their infancy. There is an urgent need for accurate, cost-effective, and efficient noninvasive detection methods capable of diagnosing pancreatic cancer at a resectable stage, which is critical for improving the five-year survival rate of patients with PDAC [13].
Recent advancements have identified novel biomarkers detectable through liquid biopsies, such as circulating tumor DNA (ctDNA), microRNAs (miRNAs), specific protein markers, and metabolite biomarkers [14,15], which show promise in the early diagnosis of pancreatic cancer. These biomarkers offer the potential for noninvasive, sensitive, and specific detection methods. This meta-analysis aims to evaluate the sensitivity and specificity of various novel biomarkers in the early diagnosis of pancreatic cancer, compare their diagnostic performance, and identify the most promising biomarkers for clinical application.
Methods
Protocol and registration
We conducted this systematic review and meta-analysis following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [16]. The study protocol was registered with PROSPERO (registration number CRD42024553633).
Eligibility criteria
We included studies that met the following criteria:
- Study Design: Cohort studies, case-control studies, and diagnostic test accuracy studies.
- Participants: Patients suspected of or diagnosed with pancreatic cancer.
- Biomarkers: Evaluations of novel biomarkers such as ctDNA, miRNAs, or specific protein markers for early diagnosis.
- Outcomes: Reports on diagnostic performance metrics, including sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV).
- Quality Assessment: Studies assessed using the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) tool were included only if they had a rating of “high risk” in at most one domain of bias [17].
Exclusion criteria were:
- Non-human studies.
- Studies lacking complete diagnostic performance data.
- Duplicate publications.
- Studies with excessive heterogeneity in design, patient population, or biomarker detection methods, as identified by QUADAS-2 evaluation.
- Small-scale studies with fewer than 30 participants or conducted exclusively in specific regions or healthcare systems.
Exclusion of Pancreatic Cancer
In the included studies, patients with confirmed pancreatic cancer were excluded based on several criteria. First, any study focusing on patients already diagnosed with pancreatic cancer was excluded, with the diagnosis confirmed through clinical evaluation and medical records. Second, studies that included patients with a history of pancreatic cancer or evidence of malignancy, as determined by clinical assessment, imaging tests (such as CT or MRI), or biopsy, were also excluded. Additionally, studies specifically targeting early-stage pancreatic cancer detection (e.g., stage I and II) excluded patients with advanced-stage cancer (III or IV), which was determined through clinical staging, imaging, and histopathological examination. These procedures were implemented to ensure that all included studies focused on identifying biomarkers for the early detection of pancreatic cancer, while excluding patients with a confirmed diagnosis of the disease.
Search strategy
We performed a comprehensive literature search for studies published from January 2014 to May 2024. The search was conducted on June 1, 2024. Databases searched included PubMed, Embase, Web of Science, and the Cochrane Library. We combined MeSH terms and free-text keywords related to “pancreatic cancer,” “early diagnosis,” and “novel biomarkers” using Boolean operators (AND, OR).The specific search terms were:
- Population: “pancreatic cancer” OR “pancreatic carcinoma”
- Intervention: “biomarker” OR “circulating tumor DNA” OR “ctDNA” OR “microRNA” OR “miRNA” OR “protein marker”
- Outcome: “early diagnosis” OR “early detection” AND “diagnostic accuracy” OR “sensitivity” OR “specificity”
We also screened reference lists of relevant articles to identify additional studies.
The specific search strategy is detailed in the attached table (S1 Table).
Study selection
The search strategy results were stored and managed using the Endnote software. Two independent reviewers screened the titles and abstracts of all retrieved articles. Full-text articles were obtained for studies meeting the inclusion criteria or when eligibility was uncertain. Discrepancies were resolved through discussion or consultation with a third reviewer.
Data extraction
Two reviewers independently extracted data using a standardized form, including:
- Study Characteristics: Author, publication year, country, study design, and sample size.
- Patient Characteristics: Age, sex, and disease stage.
- Biomarker Details: Specific biomarkers evaluated and detection methods used (e.g., PCR, NGS).
- Diagnostic Performance Metrics: Sensitivity, specificity, PPV, NPV, and 95% confidence intervals.
- Quality Assessment: QUADAS-2 scores for patient selection, index test, reference standard, and flow and timing.
Any discrepancies were resolved by consensus or involving a third reviewer.
Missing data were managed by initially contacting the corresponding authors of the original studies to obtain any unavailable or unclear information. In the absence of a response, missing values were estimated from figures or derived through appropriate statistical conversions based on the available data. Studies with substantial missing data that could not be reliably imputed were excluded from the meta-analysis.
Quality assessment
We assessed the methodological quality of the included studies using the QUADAS-2 tool, which evaluates the risk of bias across four domains (patient selection, index test, reference standard, flow and timing) and applicability concerns across three domains (patient selection, index test, reference standard). Two reviewers independently rated the risk of bias and applicability concerns for each domain as “low,” “high,” or “unclear.” Studies were included if at most one bias domain was rated as “high risk” (out of four bias domains). Applicability concerns were documented but did not influence the inclusion decision, as they pertain to generalizability rather than internal validity. Disagreements between reviewers were resolved through discussion or consultation with a third reviewer.
Statistical analysis
We used Stata version 16 for statistical analyses. A bivariate random-effects model calculated pooled estimates of sensitivity, specificity, positive likelihood ratio (PLR), negative likelihood ratio (NLR), and diagnostic odds ratio (DOR) with 95% confidence intervals. We constructed summary receiver operating characteristic (SROC) curves and calculated the area under the curve (AUC) to evaluate overall diagnostic accuracy.
Heterogeneity among studies was assessed using Cochran’s Q test and the I² statistic. An I² value over 50% indicated substantial heterogeneity [18,19]. We explored potential sources of heterogeneity through subgroup analyses based on biomarker type, sample type (blood, urine), and detection methods (e.g., PCR, NGS).
Sensitivity analyses excluded studies with a high risk of bias to assess the robustness of pooled estimates. We evaluated publication bias using funnel plots and Deeks’ funnel plot asymmetry test.
Results reporting
We reported results following PRISMA guidelines. A flow diagram illustrated the study selection process. Tables summarized the characteristics of included studies and quality assessment outcomes. Forest plots presented pooled diagnostic performance metrics. We documented assessments of heterogeneity and publication bias and reported findings from subgroup and sensitivity analyses.
Results
Study selection and characteristics
Our search yielded 1,742 studies, with 502 duplicates removed. After screening titles and abstracts, we selected 222 studies for full-text evaluation. Of these, 178 were excluded for reasons such as not being diagnostic tests, unclear cancer stage definition, lacking specific biomarkers, incomplete diagnostic data, and sample sizes of fewer than 30 participants. An additional two studies were excluded due to insufficient numbers for meta-analysis. Ultimately, 42 studies met the eligibility criteria (Fig 1).
Tables 1 and 2 summarize the included studies, published between 2014 and 2024, involving 19,326 participants. Among them, 2,749 patients had stage I or II pancreatic cancer. All studies used the internationally accepted TNM staging system. Four studies included only stage I patients, three included only stage II patients, and the remaining 36 studies included both stages. The control groups consisted of 16,577 individuals, including patients with chronic pancreatitis, benign pancreatic diseases, and healthy volunteers. Specifically, 28 studies exclusively used healthy volunteers as controls, whereas 15 studies included both healthy volunteers and patients with chronic pancreatitis or benign pancreatic diseases as controls (Tables 1 and 2).
Regarding biomarker types, seven studies investigated ctDNA [20–26], seven focused on miRNAs [27–33], and seven assessed metabolites [34–39] (Table 3). Most studies (n = 27) evaluated more than one biomarker. Sixteen studies focused on a single biomarker. One metabolite study [36] reported on two independent cohorts from China and the United States, treated separately in our analysis. Additionally, 17 studies evaluated protein biomarkers [40–56]. Five studies [57–61] exclusively examined combining novel biomarkers with CA19–9 and were included in subgroup analyses.
Quality assessment
Using the QUADAS-2 tool (Fig 2), we found that 20 studies had a high risk of patient selection bias due to potential case-control designs and unclear enrollment processes. Twelve studies had ambiguous descriptions in this domain. One study had a high risk of bias in the index test domain due to unclear interpretation processes. All studies had a well-defined reference standard, resulting in a low risk of bias for that domain. All studies were rated low risk in the flow and timing domain due to appropriate intervals between index tests and reference standards.
Diagnostic accuracy of protein biomarkers
We included 17 studies assessing protein biomarkers for early pancreatic cancer detection (Fig 3). Using a bivariate random-effects model, the pooled sensitivity was 0.79 (95% CI, 0.70–0.86), and the pooled specificity was 0.88 (95% CI, 0.82–0.93). Forest plots showed substantial variability, with I² values of 83.61% for sensitivity and 92.79% for specificity.
A: Forest plots showing pooled sensitivity (left) and specificity (right) estimates of included studies with 95% confidence intervals (CIs). The red dashed lines indicate the overall pooled sensitivity and specificity. B: Forest plots of positive likelihood ratios (LR+) and negative likelihood ratios (LR-) with pooled estimates. LR+ measures the likelihood of a positive test result in patients with the disease, while LR- assesses the likelihood of a negative test in patients without the disease. C: Forest plots of diagnostic score (log diagnostic odds ratio) and diagnostic odds ratio (DOR) with corresponding 95% CIs. Higher DOR values indicate stronger overall diagnostic accuracy. D: Summary Receiver Operating Characteristic (SROC) curve with prediction and confidence contours, depicting the diagnostic accuracy of biomarkers across studies. The red point represents the summary operating point, and the dotted lines indicate the confidence and prediction intervals. E: Deeks’ funnel plot asymmetry test for publication bias assessment. A non-significant p-value (p > 0.05) suggests no evidence of significant publication bias. F: Fagan’s nomogram demonstrating post-test probabilities based on a pretest probability of 50%. Positive tests increase post-test probability to 87%, while negative tests reduce it to 19%. G: Scatter plot of positive likelihood ratio (PLR) versus negative likelihood ratio (NLR) across studies, illustrating variability in diagnostic accuracy.
The SROC curve had an AUC of 0.90 (95% CI, 0.87–0.93), demonstrating excellent diagnostic accuracy. The pooled DOR was 27.74 (95% CI, 14.32–53.76). Significant heterogeneity was observed, emphasizing the need to investigate variability sources.
The pooled PLR was 6.64 (95% CI, 4.19–10.52), and the pooled NLR was 0.24 (95% CI, 0.17–0.34). Using a pretest probability of 50%, Fagan’s nomogram showed that a positive test increased the posttest probability to 87%, while a negative test reduced it to 19%.
Deeks’ funnel plot asymmetry test yielded a p-value of 0.29, indicating no significant publication bias.
Diagnostic performance of ctDNA biomarkers
Seven studies evaluated ctDNA biomarkers (Fig 4). The pooled sensitivity was 0.65 (95% CI, 0.48–0.81), and the pooled specificity was 0.94 (95% CI, 0.88–0.97). Forest plots revealed significant heterogeneity, with I² values of 94.33% for sensitivity and 85.08% for specificity.
A: Forest plots showing pooled sensitivity (left) and specificity (right) estimates of included studies with 95% confidence intervals (CIs). The red dashed lines indicate the overall pooled sensitivity and specificity. B: Forest plots of positive likelihood ratios (LR+) and negative likelihood ratios (LR-) with pooled estimates. LR+ measures the likelihood of a positive test result in patients with the disease, while LR- assesses the likelihood of a negative test in patients without the disease. C: Forest plots of diagnostic score (log diagnostic odds ratio) and diagnostic odds ratio (DOR) with corresponding 95% CIs. Higher DOR values indicate stronger overall diagnostic accuracy. D: Summary Receiver Operating Characteristic (SROC) curve with prediction and confidence contours, depicting the diagnostic accuracy of biomarkers across studies. The red point represents the summary operating point, and the dotted lines indicate the confidence and prediction intervals. E: Deeks’ funnel plot asymmetry test for publication bias assessment. A non-significant p-value (p > 0.05) suggests no evidence of significant publication bias. F: Fagan’s nomogram showing post-test probabilities for a pretest probability of 50%. A positive test increases the post-test probability to 91%, while a negative test decreases it to 27%. G: Scatter plot of positive likelihood ratio (PLR) versus negative likelihood ratio (NLR) across studies, illustrating variability in diagnostic accuracy.
The SROC curve showed an AUC of 0.92 (95% CI, 0.89–0.94). The pooled DOR was 27.73 (95% CI, 12.91–59.55). The pooled PLR was 10.26 (95% CI, 5.75–18.25), and the pooled NLR was 0.37 (95% CI, 0.22–0.62). Fagan’s nomogram indicated that a positive ctDNA result increased the posttest probability to 91%, while a negative result decreased it to 27%.
Deeks’ funnel plot asymmetry test showed no significant publication bias (p = 0.29).
Diagnostic performance of miRNA biomarkers
Seven studies assessed miRNA biomarkers (Fig 5). The pooled sensitivity was 0.88 (95% CI, 0.79–0.93), and the pooled specificity was 0.91 (95% CI, 0.82–0.95). High heterogeneity was observed, with I² values of 77.36% for sensitivity and 91.27% for specificity.
A: Forest plots showing pooled sensitivity (left) and specificity (right) estimates of included studies with 95% confidence intervals (CIs). The red dashed lines indicate the overall pooled sensitivity and specificity. B: Forest plots of positive likelihood ratios (LR+) and negative likelihood ratios (LR-) with pooled estimates. LR+ measures the likelihood of a positive test result in patients with the disease, while LR- assesses the likelihood of a negative test in patients without the disease. C: Forest plots of diagnostic score (log diagnostic odds ratio) and diagnostic odds ratio (DOR) with corresponding 95% CIs. Higher DOR values indicate stronger overall diagnostic accuracy. D: Summary Receiver Operating Characteristic (SROC) curve with prediction and confidence contours, depicting the diagnostic accuracy of biomarkers across studies. The red point represents the summary operating point, and the dotted lines indicate the confidence and prediction intervals. E: Deeks’ funnel plot asymmetry test for publication bias assessment. A non-significant p-value (p > 0.05) suggests no evidence of significant publication bias. F: Fagan’s nomogram demonstrating post-test probabilities based on a pretest probability of 50%. A positive test result increases the post-test probability to 91%, while a negative test reduces it to 12%. G: Scatter plot of positive likelihood ratio (PLR) versus negative likelihood ratio (NLR) across studies, illustrating variability in diagnostic accuracy.
The SROC curve had an AUC of 0.95 (95% CI, 0.93–0.97). The pooled DOR was 72.68 (95% CI, 26.64–198.24). The pooled PLR was 9.57 (95% CI, 4.85–18.88), and the pooled NLR was 0.13 (95% CI, 0.07–0.24). Fagan’s nomogram showed that a positive miRNA test increased the posttest probability to 91%, while a negative result reduced it to 12%.
Deeks’ funnel plot asymmetry test indicated no significant publication bias (p = 0.45).
Diagnostic accuracy of metabolite biomarkers
Seven studies evaluated metabolite biomarkers (Fig 6). The pooled sensitivity was 0.84 (95% CI, 0.73–0.92), and the pooled specificity was 0.85 (95% CI, 0.81–0.88). Significant heterogeneity was present, with I² values of 81.17% for sensitivity and 98.05% for specificity.
A: Forest plots showing pooled sensitivity (left) and specificity (right) estimates of included studies with 95% confidence intervals (CIs). The red dashed lines indicate the overall pooled sensitivity and specificity. B: Forest plots of positive likelihood ratios (LR+) and negative likelihood ratios (LR-) with pooled estimates. LR+ measures the likelihood of a positive test result in patients with the disease, while LR- assesses the likelihood of a negative test in patients without the disease. C: Forest plots of diagnostic score (log diagnostic odds ratio) and diagnostic odds ratio (DOR) with corresponding 95% CIs. Higher DOR values indicate stronger overall diagnostic accuracy. D: Summary Receiver Operating Characteristic (SROC) curve with prediction and confidence contours, depicting the diagnostic accuracy of biomarkers across studies. The red point represents the summary operating point, and the dotted lines indicate the confidence and prediction intervals. E: Deeks’ funnel plot asymmetry test for publication bias assessment. A non-significant p-value (p > 0.05) suggests no evidence of significant publication bias. F: Fagan’s nomogram demonstrating post-test probabilities. Based on a pretest probability of 50%, a positive test result increases the post-test probability to 85%, while a negative test reduces it to 15%. G: Scatter plot of positive likelihood ratio (PLR) versus negative likelihood ratio (NLR) across studies, illustrating variability in diagnostic accuracy.
The SROC curve showed an AUC of 0.90 (95% CI, 0.87–0.93). The pooled DOR was 31.76 (95% CI, 12.38–81.48). The pooled PLR was 5.80 (95% CI, 4.10–8.20), and the pooled NLR was 0.18 (95% CI, 0.10–0.34). Fagan’s nomogram indicated that a positive test increased the posttest probability to 85%, while a negative result reduced it to 15%.
Deeks’ funnel plot asymmetry test suggested potential publication bias (p = 0.04), so these results should be interpreted with caution.
Subgroup analysis results
We conducted subgroup analyses on proteins, ctDNA, miRNAs, and metabolites. When novel biomarkers are combined with CA19–9, the diagnostic performance exceeds what can be achieved by using these biomarkers alone. In this study, we began by conducting a meta-analysis to determine the overall performance of novel biomarkers used in conjunction with CA19–9. We then closely examined the definitions and calculation methods employed in the relevant articles and observed two primary strategies.
The first strategy, known as the “either-positive” approach, calculates sensitivity by classifying a patient’s overall result as positive if either CA19–9 or the novel biomarker is positive. Specificity, on the other hand, is determined by requiring both tests to be negative—referred to as “joint negativity.” In other words, this approach applies an “OR” principle for positivity and an “AND” principle for negativity. By contrast, some studies utilized regression models, scoring systems, or machine learning algorithms (e.g., logistic regression or multi-marker panels) to estimate the combined sensitivity and specificity of a biomarker with CA19–9. Under these methods, an integrated score or a predicted probability serves as a threshold: results above this threshold are classified as positive, and those below are classified as negative. Consequently, in such cases, the reported sensitivity and specificity arise from modeling or receiver operating characteristic (ROC) analysis rather than a simple binary combination. Based on these distinctions, we conducted subgroup analyses for each approach.
Fourteen studies [40–44,47,52,54,56–61] addressed the combination of protein biomarkers with CA19–9 for early pancreatic cancer detection. The pooled analysis showed a sensitivity of 0.90 and a specificity of 0.92. After categorizing the articles by methodology, five [42,43,56,57,62] studies employing the either-positive principle yielded a combined sensitivity of 0.86 and specificity of 0.88 (S1 Fig), whereas nine [40,41,44,47,54,58–61] studies using predictive models or machine learning reported a sensitivity of 0.92 and a specificity of 0.94 (S2 Fig). There was no substantial difference in I² between the two groups. Nevertheless, these results suggest that algorithmic approaches may modestly increase both sensitivity and specificity compared to the simpler either-positive principle.
Another five studies [20,21,23,26,63] investigated ctDNA in combination with CA19–9. Of these, three employed predictive modeling or machine learning, and two adopted the either-positive approach. Because of the limited number of studies, no subgroup meta-analysis was conducted for ctDNA. The pooled results showed a sensitivity of 0.79 and a specificity of 0.95. Although this combination demonstrates relatively high specificity, there is still a need to enhance its sensitivity (Fig 7).
Studies using healthy volunteers as controls showed similar diagnostic performance to those including patients with precancerous conditions or chronic pancreatitis, indicating that these biomarkers may not have the high false-positive rates associated with CA19–9 in certain populations (Fig 8).
Additional analyses specifically examined studies employing ELISA for protein biomarkers and PCR for miRNA biomarkers. Neither the diagnostic accuracy nor the observed heterogeneity demonstrated significant changes within these subgroups (Fig 9).
Sensitivity analysis
We performed sensitivity analyses by excluding studies with high risk of bias and small sample sizes (n ≤ 50). Reanalyzing the data showed no significant changes in pooled estimates. A leave-one-out analysis indicated that no single study unduly influenced the overall results, suggesting robustness (S7 Table).
Discussion
This meta-analysis evaluated the diagnostic accuracy of novel biomarkers—protein biomarkers, ctDNA, miRNAs, and metabolites—for early pancreatic cancer detection. Our findings show that these biomarkers have good diagnostic performance, with miRNA biomarkers exhibiting the highest pooled sensitivity and specificity, followed by protein and metabolite biomarkers. Although ctDNA biomarkers had high specificity, their sensitivity was moderate; however, sensitivity improved after excluding small-sample studies. These results highlight the potential of these biomarkers in enhancing early detection, crucial for improving patient outcomes.
Our results align with studies reporting high diagnostic accuracy of miRNAs in pancreatic cancer detection [64–67]. The high sensitivity and specificity of miRNA biomarkers may be due to their stability in bodily fluids and role in gene regulation related to carcinogenesis [68]. Protein biomarkers also showed strong diagnostic performance, consistent with research emphasizing their potential in cancer diagnosis [69,70]. The moderate sensitivity but high specificity of ctDNA biomarkers reflect challenges in detecting low levels of tumor-derived DNA in early-stage disease [71]. Metabolite biomarkers demonstrated good diagnostic accuracy, supporting findings that metabolic changes indicate tumor presence and progression [72].
Unlike previous analyses focusing on all pancreatic cancer patients, our study specifically analyzed stage I and II patients. These biomarkers showed stronger diagnostic performance in early-stage diagnosis compared to overall pancreatic cancer populations.
The high sensitivity and specificity of miRNA and protein biomarkers suggest they could be effective noninvasive tools for early detection. miRNAs may reflect early molecular changes in carcinogenesis, and protein biomarkers can be detected through established assays like ELISA, offering practical clinical advantages. The high specificity of ctDNA biomarkers indicates potential for confirming diagnoses, though combining them with other biomarkers or imaging may enhance effectiveness. Metabolite biomarkers provide another detection avenue but may require standardized platforms for consistency.
Incorporating these novel biomarkers into clinical practice could significantly improve early detection rates, increasing eligibility for curative surgery. Combining them with the traditional CA19–9 assay could enhance diagnostic accuracy, as our subgroup analyses showed improved performance with combined use. This approach could mitigate individual biomarker limitations and offer a more comprehensive diagnostic tool. In addition, our study performed a regrouped subgroup analysis of the literature on combining the novel biomarkers with CA19–9. Some studies used an “either-positive” principle—treating a test result as positive if either CA19–9 or the novel biomarker was positive—an approach that tends to increase sensitivity in clinical screening, while applying a “joint negativity principle” to enhance specificity. Other studies employed regression equations, machine learning algorithms, or multi-marker predictive models to integrate multiple variables into a single composite score, using a specified threshold to determine positivity or negativity. The subgroup analysis showed that, for protein biomarkers combined with CA19–9, the overall sensitivity and specificity under the either-positive principle (0.86 and 0.88, respectively) were slightly lower than those achieved by predictive modeling or machine learning methods 0.92 and 0.94, respectively). These findings underscore the importance of algorithmic optimization for improving diagnostic accuracy. They also point to the potential necessity and feasibility of incorporating additional biomarkers or refining detection techniques and algorithms to further enhance diagnostic performance. Our analyses also indicated that diagnostic performance was unaffected by control group composition, suggesting these biomarkers may avoid CA19–9’s high false-positive rates in certain populations.
A major strength of this meta-analysis is the comprehensive evaluation across numerous studies, providing a reliable assessment of diagnostic performance. We used strict inclusion criteria and quality assessments, which enhanced the robustness of our findings. The statistical approach we employed was a random-effects meta-analysis model, which accounts for between-study variability by assuming that the true effect sizes may differ across studies. This model was chosen to accommodate the heterogeneity observed in study designs, patient populations, biomarkers, and detection methods. By using this model, we were able to calculate pooled estimates of sensitivity, specificity, positive likelihood ratio (PLR), negative likelihood ratio (NLR), and diagnostic odds ratio (DOR) with 95% confidence intervals. Additionally, we constructed summary receiver operating characteristic (SROC) curves to assess overall diagnostic accuracy.
However, significant heterogeneity existed, likely due to biomarker diversity and differences in study design, patient populations, cancer staging, control groups, and detection methods. Potential publication bias, especially in metabolite studies, suggests caution and may relate to sample size. Sensitivity analyses did not show significant heterogeneity associated with very large or small sample sizes. Despite efforts to explore heterogeneity sources, residual variability remains. Methodological quality varied among studies, which may have influenced the pooled estimates.
Future research should focus on standardizing detection and validation methods through large, multicenter prospective studies. Developing standardized protocols for sample handling is crucial for clinical translation. Although biomarkers were pooled by type to estimate the overall diagnostic accuracy of each class (miRNAs, proteins, ctDNA), we acknowledge that individual biomarkers within each category may exhibit different cut-off values, sensitivities, and specificities. This variability can limit the generalizability of the pooled estimates. Therefore, future studies should prioritize the validation of specific biomarker combinations rather than isolated markers, and standardize detection methods and cut-off values within each biomarker type to improve diagnostic accuracy and consistency. Furthermore, exploring the combined use of biomarkers and integrating them with imaging techniques could enhance diagnostic performance. Finally, investigating the cost-effectiveness and feasibility of routine screening, especially for high-risk populations, is essential for practical clinical application.
Conclusion
This meta-analysis highlights the promising diagnostic performance of novel biomarkers, particularly miRNA and protein biomarkers, in early pancreatic cancer detection. Despite challenges from heterogeneity and methodological differences, these biomarkers offer potential for improving early diagnosis and patient prognosis. Further investigation through well-designed prospective studies is warranted to integrate these biomarkers into clinical practice.
Supporting information
S1 file. Supporting information.
S1 Fig. Sensitivity with CA19–9 + Protein Biomarker (Either-Positive).S2 Fig. Sensitivity with CA19–9 + Protein Biomarker (Predictive/ Machine Learning). S1 Table. Detailed search strategy for the systematic review.S2 Table. QUADAS-2 review items.S3 Table. QUADAS-2 review results.S4 Table. Characteristics and diagnostic outcomes of studies using protein or ctDNA combined with CA19–9 biomarkers.S5 Table. Characteristics and diagnostic outcomes of studies in healthy control populations.S6 Table. Diagnostic accuracy of protein biomarkers and miRNA biomarkers using specific detection methods for early pancreatic cancer.S7 Table. Results of leave-one-out analysis.
https://doi.org/10.1371/journal.pone.0322720.s001
(PDF)
References
- 1. Siegel RL, Miller KD, Jemal A. Cancer statistics, 2020. CA Cancer J Clin. 2020;70(1):7–30. pmid:31912902
- 2. Rahib L, Smith BD, Aizenberg R, Rosenzweig AB, Fleshman JM, Matrisian LM, et al. Projecting cancer incidence and deaths to 2030: the unexpected burden of thyroid, liver, and pancreas cancers in the United States. Cancer Res. 2014;74(11):2913–21. pmid:24840647
- 3. Hidalgo M. Pancreatic cancer. N Engl J Med. 2010;362(17):1605–17. pmid:20427809
- 4. Walter FM, Mills K, Mendonça SC, Abel GA, Basu B, Carroll N, et al. Symptoms and patient factors associated with diagnostic intervals for pancreatic cancer (SYMPTOM pancreatic study): a prospective cohort study. Lancet Gastroenterol Hepatol. 2016;1(4):298–306. pmid:28404200
- 5. Ren S, Qian L-C, Cao Y-Y, Daniels MJ, Song L-N, Tian Y, et al. Computed tomography-based radiomics diagnostic approach for differential diagnosis between early- and late-stage pancreatic ductal adenocarcinoma. World J Gastrointest Oncol. 2024;16(4):1256–67. pmid:38660647
- 6. Park W, Chawla A, O’Reilly EM. Pancreatic Cancer: A Review. JAMA. 2021;326(9):851–62. pmid:34547082
- 7. Siegel RL, Giaquinto AN, Jemal A. Cancer statistics, 2024. CA Cancer J Clin. 2024;74(1):12–49. pmid:38230766
- 8. Singhi AD, Koay EJ, Chari ST, Maitra A. Early Detection of Pancreatic Cancer: Opportunities and Challenges. Gastroenterology. 2019;156(7):2024–40. pmid:30721664
- 9. Zhao B, Zhao B, Chen F. Diagnostic value of serum carbohydrate antigen 19-9 in pancreatic cancer: a systematic review and meta-analysis. Eur J Gastroenterol Hepatol. 2022;34(9):891–904. pmid:35913776
- 10. Rosty C, Goggins M. Early detection of pancreatic carcinoma. Hematol Oncol Clin North Am. 2002;16(1):37–52. pmid:12063828
- 11. Azizian A, Rühlmann F, Krause T, Bernhardt M, Jo P, König A, et al. CA19-9 for detecting recurrence of pancreatic cancer. Sci Rep. 2020;10(1):1332. pmid:31992753
- 12. Luo G, Fan Z, Cheng H, Jin K, Guo M, Lu Y, et al. New observations on the utility of CA19-9 as a biomarker in Lewis negative patients with pancreatic cancer. Pancreatology. 2018;18(8):971–6. pmid:30131287
- 13. Cao Y-Y, Guo K, Zhao R, Li Y, Lv X-J, Lu Z-P, et al. Untargeted metabolomics characterization of the resectable pancreatic ductal adenocarcinoma. Digit Health. 2023;9:20552076231179007. pmid:37312938
- 14. Zhao R, Ren S, Li C, Guo K, Lu Z, Tian L, et al. Biomarkers for pancreatic cancer based on tissue and serum metabolomics analysis in a multicenter study. Cancer Med. 2023;12(4):5158–71. pmid:36161527
- 15. Cao Y-Y, Lv X-J, Li H, Qian L-C, Si H-P, Li Y, et al. Serum High-Density Lipoprotein Cholesterol Concentrations in Pancreatic Ductal Adenocarcinoma and Its Association With Histological Grade in a Chinese Population. Cancer Control. 2025;32:10732748251316602. pmid:39927839
- 16. Page MJ, Moher D, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, et al. PRISMA 2020 explanation and elaboration: updated guidance and exemplars for reporting systematic reviews. BMJ. 2021;372:n160. pmid:33781993
- 17. Whiting PF, Rutjes AWS, Westwood ME, Mallett S, Deeks JJ, Reitsma JB, et al. QUADAS-2: a revised tool for the quality assessment of diagnostic accuracy studies. Ann Intern Med. 2011;155(8):529–36. pmid:22007046
- 18. Higgins JPT, Thompson SG. Quantifying heterogeneity in a meta-analysis. Stat Med. 2002;21(11):1539–58. pmid:12111919
- 19. Higgins JPT, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta-analyses. BMJ. 2003;327(7414):557–60. pmid:12958120
- 20. Cohen JD, Javed AA, Thoburn C, Wong F, Tie J, Gibbs P, et al. Combined circulating tumor DNA and protein biomarker-based liquid biopsy for the earlier detection of pancreatic cancers. Proc Natl Acad Sci U S A. 2017;114(38):10202–7. pmid:28874546
- 21. Bauden M, Pamart D, Ansari D, Herzog M, Eccleston M, Micallef J, et al. Circulating nucleosomes as epigenetic biomarkers in pancreatic cancer. Clin Epigenetics. 2015;7:106. pmid:26451166
- 22. Eissa MAL, Lerner L, Abdelfatah E, Shankar N, Canner JK, Hasan NM, et al. Promoter methylation of ADAMTS1 and BNC1 as potential biomarkers for early detection of pancreatic cancer in blood. Clin Epigenetics. 2019;11(1):59. pmid:30953539
- 23. Fujimoto Y, Suehiro Y, Kaino S, Suenaga S, Tsuyama T, Matsui H, et al. Combination of CA19-9 and Blood Free-Circulating Methylated RUNX3 May Be Useful to Diagnose Stage I Pancreatic Cancer. Oncology. 2021;99(4):234–9. pmid:33440396
- 24. Henriksen SD, Madsen PH, Larsen AC, Johansen MB, Drewes AM, Pedersen IS, et al. Cell-free DNA promoter hypermethylation in plasma as a diagnostic marker for pancreatic adenocarcinoma. Clin Epigenetics. 2016;8:117. pmid:27891190
- 25. Majumder S, Taylor WR, Foote PH, Berger CK, Wu CW, Mahoney DW, et al. High Detection Rates of Pancreatic Cancer Across Stages by Plasma Assay of Novel Methylated DNA Markers and CA19-9. Clin Cancer Res. 2021;27(9):2523–32. pmid:33593879
- 26. Wu H, Guo S, Liu X, Li Y, Su Z, He Q, et al. Noninvasive detection of pancreatic ductal adenocarcinoma using the methylation signature of circulating tumour DNA. BMC Med. 2022;20(1):458. pmid:36434648
- 27. Debernardi S, Massat NJ, Radon TP, Sangaralingam A, Banissi A, Ennis DP, et al. Noninvasive urinary miRNA biomarkers for early detection of pancreatic adenocarcinoma. Am J Cancer Res. 2015;5(11):3455–66. pmid:26807325
- 28. Masterson AN, Chowdhury NN, Fang Y, Yip-Schneider MT, Hati S, Gupta P, et al. Amplification-Free, High-Throughput Nanoplasmonic Quantification of Circulating MicroRNAs in Unprocessed Plasma Microsamples for Earlier Pancreatic Cancer Detection. ACS Sens. 2023;8(3):1085–100. pmid:36853001
- 29. Yu Y, Tong Y, Zhong A, Wang Y, Lu R, Guo L. Identification of Serum microRNA-25 as a novel biomarker for pancreatic cancer. Medicine (Baltimore). 2020;99(52):e23863. pmid:33350781
- 30. Nakamura K, Zhu Z, Roy S, Jun E, Han H, Munoz RM, et al. An Exosome-based Transcriptomic Signature for Noninvasive, Early Detection of Patients With Pancreatic Ductal Adenocarcinoma: A Multicenter Cohort Study. Gastroenterology. 2022;163(5):1252-1266.e2. pmid:35850192
- 31. Huang J, Gao G, Ge Y, Liu J, Cui H, Zheng R, et al. Development of a Serum-Based MicroRNA Signature for Early Detection of Pancreatic Cancer: A Multicenter Cohort Study. Dig Dis Sci. 2024;69(4):1263–73. pmid:38451429
- 32. Schultz NA, Dehlendorff C, Jensen BV, Bjerregaard JK, Nielsen KR, Bojesen SE, et al. MicroRNA biomarkers in whole blood for detection of pancreatic cancer. JAMA. 2014;311(4):392–404. pmid:24449318
- 33. PubMed pmid:WOS:000422130000003. Ganepola GA, Rutledge JR, Suman P, Yiengpruksawan A, Chang DH. Novel blood-based microRNA biomarker panel for early diagnosis of pancreatic cancer. World J Gastrointest Oncol. 2014;6(1):22–33. pmid:24578785
- 34. Wolrab D, Jirásko R, Cífková E, Höring M, Mei D, Chocholoušková M, et al. Lipidomic profiling of human serum enables detection of pancreatic cancer. Nat Commun. 2022;13(1):124. pmid:35013261
- 35. Fukutake N, Ueno M, Hiraoka N, Shimada K, Shiraishi K, Saruki N, et al. A Novel Multivariate Index for Pancreatic Cancer Detection Based On the Plasma Free Amino Acid Profile. PLoS One. 2015;10(7):e0132223. pmid:26133769
- 36. Xie G, Lu L, Qiu Y, Ni Q, Zhang W, Gao Y-T, et al. Plasma metabolite biomarkers for the detection of pancreatic cancer. J Proteome Res. 2015;14(2):1195–202. pmid:25429707
- 37. Zhang Y, Qiu L, Wang Y, Qin X, Li Z. High-throughput and high-sensitivity quantitative analysis of serum unsaturated fatty acids by chip-based nanoelectrospray ionization-Fourier transform ion cyclotron resonance mass spectrometry: early stage diagnostic biomarkers of pancreatic cancer. Analyst. 2014;139(7):1697–706. pmid:24551873
- 38. Mayerle J, Kalthoff H, Reszka R, Kamlage B, Peter E, Schniewind B, et al. Metabolic biomarker signature to differentiate pancreatic ductal adenocarcinoma from chronic pancreatitis. Gut. 2018;67(1):128–37. pmid:28108468
- 39. Hirata Y, Kobayashi T, Nishiumi S, Yamanaka K, Nakagawa T, Fujigaki S, et al. Identification of highly sensitive biomarkers that can aid the early detection of pancreatic cancer using GC/MS/MS-based targeted metabolomics. Clin Chim Acta. 2017;468:98–104. pmid:28215548
- 40. Aronsson L, Andersson R, Bauden M, Andersson B, Bygott T, Ansari D. High-density and targeted glycoproteomic profiling of serum proteins in pancreatic cancer and intraductal papillary mucinous neoplasm. Scand J Gastroenterol. 2018;53(12):1597–603. pmid:30509115
- 41. Brand RE, Persson J, Bratlie SO, Chung DC, Katona BW, Carrato A, et al. Detection of Early-Stage Pancreatic Ductal Adenocarcinoma From Blood Samples: Results of a Multiplex Biomarker Signature Validation Study. Clin Transl Gastroenterol. 2022;13(3):e00468. pmid:35166713
- 42. Han S, Zhou X, Sui X, He C, Cai M, Ma J, et al. Serum dickkopf-1 is a novel serological biomarker for the diagnosis and prognosis of pancreatic cancer. Oncotarget. 2015;6(23):19907–17. pmid:26101916
- 43. Kashiro A, Kobayashi M, Oh T, Miyamoto M, Atsumi J, Nagashima K, et al. Clinical development of a blood biomarker using apolipoprotein-A2 isoforms for early detection of pancreatic cancer. J Gastroenterol. 2024;59(3):263–78. pmid:38261000
- 44. Lee MJ, Na K, Shin H, Kim C-Y, Cho J-Y, Kang CM, et al. Early Diagnostic Ability of Human Complement Factor B in Pancreatic Cancer Is Partly Linked to Its Potential Tumor-Promoting Role. J Proteome Res. 2021;20(12):5315–28. pmid:34766501
- 45. Li T, Xia J, Yun H, Sun G, Shen Y, Wang P, et al. A novel autoantibody signatures for enhanced clinical diagnosis of pancreatic ductal adenocarcinoma. Cancer Cell Int. 2023;23(1):273. pmid:37974212
- 46. Matsunaga T, Ohtsuka T, Asano K, Kimura H, Ohuchida K, Kitada H, et al. S100P in Duodenal Fluid Is a Useful Diagnostic Marker for Pancreatic Ductal Adenocarcinoma. Pancreas. 2017;46(10):1288–95. pmid:28984789
- 47. Mawaribuchi S, Shimomura O, Oda T, Hiemori K, Shimizu K, Yamase K, et al. rBC2LCN-reactive SERPINA3 is a glycobiomarker candidate for pancreatic ductal adenocarcinoma. Glycobiology. 2023;33(4):342–52. pmid:36728830
- 48. Mohamed AA, Soliman H, Ismail M, Ziada D, Farid TM, Aref AM, et al. Evaluation of circulating ADH and MIC-1 as diagnostic markers in Egyptian patients with pancreatic cancer. Pancreatology. 2015;15(1):34–9. pmid:25464937
- 49. Nam H, Hong S-S, Jung KH, Kang S, Park MS, Kang S, et al. A Serum Marker for Early Pancreatic Cancer With a Possible Link to Diabetes. J Natl Cancer Inst. 2022;114(2):228–34. pmid:34613397
- 50. Radon TP, Massat NJ, Jones R, Alrawashdeh W, Dumartin L, Ennis D, et al. Identification of a Three-Biomarker Panel in Urine for Early Detection of Pancreatic Adenocarcinoma. Clin Cancer Res. 2015;21(15):3512–21. pmid:26240291
- 51. Sato Y, Kobayashi T, Nishiumi S, Okada A, Fujita T, Sanuki T, et al. Prospective Study Using Plasma Apolipoprotein A2-Isoforms to Screen for High-Risk Status of Pancreatic Cancer. Cancers (Basel). 2020;12(9):2625. pmid:32937962
- 52. Wang X, Li Y, Tian H, Qi J, Li M, Fu C, et al. Macrophage inhibitory cytokine 1 (MIC-1/GDF15) as a novel diagnostic serum biomarker in pancreatic ductal adenocarcinoma. BMC Cancer. 2014;14:578. pmid:25106741
- 53. Wen Y-R, Lin X-W, Zhou Y-W, Xu L, Zhang J-L, Chen C-Y, et al. N-glycan biosignatures as a potential diagnostic biomarker for early-stage pancreatic cancer. World J Gastrointest Oncol. 2024;16(3):659–69. pmid:38577461
- 54. Yamada K, Higashi K, Nagahori H, Saito K. Circulating natural antibodies against 3’-sialyllactose complement the diagnostic performance of CA19-9 for the early detection of pancreatic ductal adenocarcinoma. Cancer Biomark. 2020;27(1):121–8. pmid:31771041
- 55. Yu J, Ploner A, Kordes M, Löhr M, Nilsson M, de Maturana MEL, et al. Plasma protein biomarkers for early detection of pancreatic ductal adenocarcinoma. Int J Cancer. 2021;148(8):2048–58. pmid:33411965
- 56. Zhou C-Y, Dong Y-P, Sun X, Sui X, Zhu H, Zhao Y-Q, et al. High levels of serum glypican-1 indicate poor prognosis in pancreatic ductal adenocarcinoma. Cancer Med. 2018;7(11):5525–33. pmid:30358133
- 57. Haab B, Qian L, Staal B, Jain M, Fahrmann J, Worthington C, et al. Abstract A014: Combinations of previously reported biomarkers achieve improved sensitivity and specificity of detection of early stage pancreatic ductal adenocarcinoma. Cancer Research. 2024;84(2_Supplement):A014–A014.
- 58. Kim H, Kang KN, Shin YS, Byun Y, Han Y, Kwon W, et al. Biomarker Panel for the Diagnosis of Pancreatic Ductal Adenocarcinoma. Cancers (Basel). 2020;12(6):1443. pmid:32492943
- 59. Firpo MA, Boucher KM, Bleicher J, Khanderao GD, Rosati A, Poruk KE, et al. Multianalyte Serum Biomarker Panel for Early Detection of Pancreatic Adenocarcinoma. JCO Clin Cancer Inform. 2023;7:e2200160. pmid:36913644
- 60. Lee DH, Yoon W, Lee A, Han Y, Byun Y, Kang JS, et al. Multi-biomarker panel prediction model for diagnosis of pancreatic cancer. J Hepato-Biliary-Pancreatic Sci. 2023;30(1):122–32. PubMed pmid:WOS:000657132200001.
- 61. Zhang P, Zou M, Wen X, Gu F, Li J, Liu G, et al. Development of serum parameters panels for the early detection of pancreatic cancer. Int J Cancer. 2014;134(11):2646–55. pmid:24615168
- 62. Wang J, Raimondo M, Guha S, Chen J, Diao L, Dong X, et al. Circulating microRNAs in Pancreatic Juice as Candidate Biomarkers of Pancreatic Cancer. J Cancer. 2014;5(8):696–705. pmid:25258651
- 63. Majumder S, Chatterjee A, Taylor WR, Foote PH, Mohiuddin SAA, Berger CK, et al. Sa301 Feasibility of detecting cancer in intraductal papillary mucinous neoplasms using plasma methylated dna markers. Gastroenterology. 2021;160(6):S-476.
- 64. Li X, Gao P, Wang Y, Wang X. Blood-Derived microRNAs for Pancreatic Cancer Diagnosis: A Narrative Review and Meta-Analysis. Front Physiol. 2018;9:685. pmid:29922178
- 65. Li Z-Z, Shen L-F, Li Y-Y, Chen P, Chen L-Z. Clinical utility of microRNA-378 as early diagnostic biomarker of human cancers: a meta-analysis of diagnostic test. Oncotarget. 2016;7(36):58569–78. pmid:27448977
- 66. Sun X, Zhou X, Zhang Y, Zhu X, Liu H. Systematic Review and Meta-Analysis of Diagnostic Accuracy of miRNAs in Patients with Pancreatic Cancer. Dis Markers. 2018;2018:6292396. pmid:29887920
- 67. Wei L, Yao K, Gan S, Suo Z. Clinical utilization of serum- or plasma-based miRNAs as early detection biomarkers for pancreatic cancer: A meta-analysis up to now. Medicine (Baltimore). 2018;97(35):e12132. pmid:30170450
- 68. Kathuria-Prakash N, Dave P, Garcia L, Brown P, Drakaki A. MicroRNAs in Genitourinary Malignancies: An Exciting Frontier of Cancer Diagnostics and Therapeutics. Int J Mol Sci. 2024;25(17):9499. pmid:39273446
- 69. Chen Y-Z, Liu D, Zhao Y-X, Wang H-T, Gao Y, Chen Y. Diagnostic performance of serum macrophage inhibitory cytokine-1 in pancreatic cancer: a meta-analysis and meta-regression analysis. DNA Cell Biol. 2014;33(6):370–7. pmid:24592997
- 70. Hu H, Zhang Q, Huang C, Shen Y, Chen X, Shi X, et al. Diagnostic value of S100P for pancreatic cancer: a meta-analysis. Tumour Biol. 2014;35(10):9479–85. pmid:25123266
- 71. Vlataki K, Antonouli S, Kalyvioti C, Lampri E, Kamina S, Mauri D, et al. Circulating Tumor DNA in the Management of Early-Stage Breast Cancer. Cells. 2023;12(12):1573. pmid:37371043
- 72. Hanahan D, Weinberg RA. Hallmarks of cancer: the next generation. Cell. 2011;144(5):646–74. pmid:21376230