The researchers aimed to conduct a systematic review and meta-analysis related to the development of VTE as a medical condition among patients diagnosed with cancer. The primary research question of this review was:

What is the proportion of venous thromboembolism (VTE) among patients with different types and stages of cancer?

This review was conducted and reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. During the scoping phase, previously published systematic reviews and meta-analyses were consulted to identify evidence gaps, refine cancer-type classifications, and inform the development of the search strategy. However, these secondary sources were not included in the formal selection process, in line with the review’s focus on primary data.

A comprehensive search was conducted on PubMed between 2nd and 12th December 2024 to identify studies reporting quantitative estimates of VTE incidence in cancer patients across different stages. The search strategy was developed through an iterative process, incorporating both free-text terms and Medical Subject Headings (MeSH), and used Boolean operators (AND/OR) to enhance precision and coverage. The final search string was: ((cancer OR tumor OR malignancy OR cancer patients) AND (Venous Thromboembolism OR venous thromboembolism OR VTE OR thrombosis OR thromboembolism) AND (risk OR risk factors OR predictors OR treatment-related risk OR clinical risk factors)).

The search was limited to English-language, full-text studies involving human participants, thereby excluding animal-based research. The date range was restricted to studies published between 2000 and December 2024, capturing developments over the last 25 years. This period was selected for both clinical and methodological reasons. Notably, the 6th edition of the TNM Classification of Malignant Tumors, introduced in 2002 (18), brought standardization to cancer staging, essential for stratifying VTE risk by stage. The early 2000s also marked advances in cancer therapies, increasing awareness of cancer-associated VTE, and the implementation of standardized thromboprophylaxis guidelines. Restricting the review to this timeframe ensured inclusion of studies aligned with modern oncology practices and diagnostic frameworks, thereby enhancing comparability and clinical relevance.

The search strategy primarily targeted studies reporting cancer-associated VTE incidence or proportions. Although the term “risk factors” was included to improve sensitivity and capture observational data, this may have biased retrieval toward analytical studies. To maintain alignment with the review objective, we applied strict eligibility criteria during screening, including only studies reporting VTE incidence or proportion stratified by cancer type or stage.

Three reviewers SK, VS, and, SA reviewed titles, abstracts, and full-text articles, with any discrepancies about study inclusion resolved by discussion among all three authors. Rayyan QCRI (19) was used to identify potentially eligible articles and remove duplicates, if any, to minimize the occurrence of selection bias. This review included studies that specifically investigated cancer-associated VTE in patients of any age or sex. Only observational studies were included, as they are particularly well-suited to capturing real-world incidence patterns of VTE across diverse clinical settings and patient populations, including variations by cancer type, stage, and geographic location (20). To ensure the broadest possible representation, studies published in English were included without geographic restrictions. Studies were excluded if they were case reports, conference abstracts, opinion pieces, editorials, or review articles that did not include original or primary data. Additionally, research with insufficient or incomplete reporting of VTE incidence data was also excluded.

The review protocol was registered in the PROSPERO International Prospective Register of Systematic Reviews (CRD42024616005) and the search results have been depicted in the form of a flow diagram ( Figure 1 ) as recommended by the PRISMA – Preferred Reporting Items for Systematic Reviews and Meta-analysis guidelines.

Initial data extraction was performed by one reviewer, followed by a blinded, independent review by at least one additional team member. Any discrepancies were resolved through consensus discussions among the full review team. While simultaneous dual extraction was not performed, this sequential cross-checking approach ensured rigorous quality control and minimized the risk of bias or errors. The following data were extracted author, title, year of publication, country, objective, sampling, study design, participant characteristics, and quantitative results (type and stage of cancer, VTE rate). Table 1 depicts the basic characteristics of all the selected studies.

The Joanna Briggs Institute (JBI) tools were chosen for assessing the quality of the included studies (JBI Critical Appraisal Tools). JBI offers a range of tools tailored to different study designs, ensuring appropriate evaluation criteria. The JBI tools provide a comprehensive set of criteria for evaluating the methodological quality of studies. These criteria typically cover aspects such as study design, participant selection, intervention description, outcome measurement, statistical analysis, and risk of bias assessment. The JBI tools typically provide response options such as “Yes,” “No,” “Not Applicable (N/A),” and “Unclear” for each assessment criterion. Reviewers selected the most appropriate response based on their evaluation of whether the study adequately addressed the criterion.

The analytical approach was structured across three levels to comprehensively evaluate the proportion of VTE among cancer patients. The first level of analysis aimed to compare population-based and hospital-based studies to assess setting-specific differences in venous thromboembolism (VTE) prevalence. A population-based study draws its sample from an entire defined population or community, such as residents of a geographic region or members of a national registry (30). The goal is to obtain a representative snapshot of the population, allowing for prevalence or risk estimates that are broadly generalizable. In this meta-analysis, population-based studies refer to those that utilized large national health registries or administrative databases to evaluate VTE risk among cancer patients. Such studies offer broad, generalizable estimates across both inpatient and outpatient settings (27–29, 31).

In contrast, a hospital- based study selects its sample from patients who visit a specific healthcare facility or group of facilities. While it may provide detailed clinical insights, its findings are less generalizable due to a narrower, treatment-seeking population (30). In the present review, studies included in this category were conducted within clinical settings, such as oncology departments or tertiary care centers, and focused on patients undergoing active cancer treatment. These studies relied on institutional records to document VTE events occurring during hospitalization or therapy, thereby capturing more granular data on treatment-related risks (6, 16, 24). This comparison was designed to capture potential variations attributable to disease severity, clinical monitoring, and treatment exposure, which are typically more pronounced in hospital settings than in population-based cohorts.

At the second level, analyses were performed based on cancer type, and for this, cancers were systematically classified into broad categories according to their anatomical origin. The classification includes gastrointestinal cancers, which have malignancies of the digestive system, such as colorectal, stomach, pancreatic, and gall bladder cancers, and Urogenital Cancers encompassing tumors arising from the genitourinary system, including prostate, kidney, testicular, and ovarian cancer. Also, Hematologic malignancies involve blood-related cancers, such as non-hodgkin lymphoma, myeloma, and leukemia/acute myeloid leukemia. Other tumors were retained as independent classifications due to their varied pathophysiology, these include sarcoma, breast, brain, mesothelium and, and lung cancers (see Supplementary Tables 6 , 7 ).

The third level of analysis was based on cancer stage. Since the included studies employed heterogeneous staging terminologies. While the standard TNM classification system (Tumor size, Node involvement, Metastasis) serves as the gold standard for cancer staging, several studies in this meta-analysis provided only partial TNM data. Therefore, a harmonization process was undertaken to map cancer stages into four standard categories: Stage I (Localized), Stage II (Early Locally Advanced), Stage III (Late Locally Advanced), and Stage IV (Metastatic). The details of this harmonization process are outlined in Supplementary Table 8 ( Appendix ).

Statistical analyses were conducted using R software version 36.0 (R Foundation for Statistical Computing) to evaluate the proportion of VTE in cancer patients. Pooled VTE proportions were calculated along with 95% confidence intervals (CIs) to quantify the estimated risk across different cancer types. Heterogeneity among studies was assessed using the I squared (I²) statistic, with thresholds of 25%, 50%, and 75% indicating low, moderate, and high heterogeneity, respectively. To visually represent the meta-analytic findings, forest plots were generated to illustrate individual study estimates, pooled prevalence rates, and overall heterogeneity. These visualizations were produced using the meta packages in R, facilitating a comprehensive depiction of effect estimates across cancer types. Furthermore, sensitivity analyses were carried out by sequentially excluding individual cancer subtypes to examine their influence on the overall pooled VTE estimates and heterogeneity statistics. This approach enabled the assessment of the robustness of the findings and the identification of cancer types that contribute disproportionately to between-study variability. Statistical significance was set at p < 0.05, and all analyses were conducted in accordance with established meta-analytic guidelines to ensure methodological transparency and reliability.

A rigorous selection process was followed in accordance with PRISMA guidelines. Initially, 5,397 records were identified through database searches. After removing 158 duplicates, 5,239 unique records were screened for relevance. During the title and abstract screening phase, 506 records were excluded for the following reasons: lack of original data (e.g., literature reviews, meta-analyses, systematic reviews), case series, animal-based studies, or surveys. Studies that did not report VTE occurrence stratified by cancer type or stage, or those in which VTE was not the primary outcome in cancer patients, were also excluded at this stage. This resulted in 4,733 records being considered for further evaluation. Subsequently, 4,706 records were excluded for not meeting the inclusion criteria. The remaining 27 full-text articles were assessed for eligibility. Of these, 13 were excluded after full-text review, resulting in 14 studies that fulfilled all inclusion criteria and were included in the final review. We included observational studies that reported the incidence or proportion of VTE in cancer patients, provided the data were stratified by cancer type or stage. Cross-sectional studies were excluded only if they did not specifically investigate cancer-associated VTE or failed to present stratified data relevant to our research objective. To minimize ambiguity, studies were not excluded solely based on their use of the term “prevalence” to describe VTE outcomes.

The 14 studies included in this systematic review span multiple continents and employ various study designs to examine venous thromboembolism (VTE) in cancer patients. A summary of the included studies is presented in Table 1 . These studies were conducted in a diverse range of countries, including Japan (21), Austria (31), Greece (22), the United States (6, 24), Saudi Arabia (23), Sweden (27, 29), Ireland (25), Canada (26), Italy (32), and the Netherlands (33), thereby contributing to a globally representative dataset. The geographical distribution of these studies is illustrated in Figure 2 , where the grading reflects the number of studies reported per country, based on the number of VTE cases.

The studies employed two principal research designs: hospital-based and population-based ( Table 1 ). Of the 14 studies, 10 followed a hospital-based design (6, 21–26, 32, 33), focusing on patient-level data within clinical or tertiary care settings. The remaining four were population-based studies that utilized national or regional healthcare registries (27–29, 31).

The JBI critical appraisal tool was employed to evaluate the risk of bias in the included studies. The assessment was conducted based on a set of methodological criteria outlined in the tool. Based on the JBI cohort checklist, the risk of bias assessment across 13 included cohort studies reveals a generally favorable methodological quality, especially in domains concerning population selection and exposure assessment. Most studies received a “Yes” rating in the early domains (D1–D5), suggesting that participants were appropriately selected, exposures were reliably measured, and confounders were identified. However, issues arise in later domains related to follow-up and outcome management. Several studies, such as Grilz et al. (2024) and Mandala et al. (2009), demonstrated high risk of bias, with “No” ratings particularly in domains assessing follow-up completeness and statistical adjustments (D9 and D10). A few studies, including Diab et al. (2010) and Balabanova et al. (2022), had multiple “Unclear” ratings in the follow-up and outcome domains (D6–D10), indicating potential concerns about reporting transparency or study design limitations. In contrast, Bloma et al. (2005) and Hemelrijck et al. (2010) showed consistently low risk across almost all domains. Overall, while the majority of the studies were methodologically robust in their design and reporting, select studies with unclear or high-risk ratings in key areas should be interpreted with caution during evidence synthesis.

The risk of bias assessment for the case-control study by Marks et al. (2014), based on the JBI checklist, reveals moderate risk. The study performed well in consistent exposure measurement (D3, D5), identification of confounding factors (D6), and use of appropriate statistical analysis (D10). However, concerns arise due to unclear group comparability (D1), exposure assessment (D4), and strategies for managing confounding (D7). Notably, outcome assessment (D8) was inadequate (“No”). These limitations suggest potential bias, warranting cautious interpretation of the findings ( Figure 3 ).

Among 7,23,898 cancer patients in population-based studies ( Supplementary Table 4 ), the VTE proportion was 2.9%, whereas among 6,24,242 cancer patients in hospital-based studies ( Supplementary Table 2 ), the VTE proportion was 4.1% (see Supplementary Table 2 ). The higher VTE proportion in hospital-based studies suggests an increased risk associated with hospitalization, disease severity, and intensive treatments. This difference was statistically significant as per the Chi-square test, χ²(1, N = 1,348,140) = 1452.22, p <.001, indicating a strong association between study setting and VTE occurrence (see Supplementary Table 3 ). Fisher’s exact test and likelihood ratio also supported the significance (p < 0.001).

The pooled analysis across cancer stages revealed a non-linear trend in VTE proportions. Specifically, the estimated VTE proportions were 6% for both Stage I and Stage II, increasing to 10% in Stage III and slightly declining to 9% in Stage IV. These findings suggest a higher thrombotic burden in more advanced cancer stages, particularly in Stage III and IV disease. The overall pooled VTE proportion across all stages was 7% (95% CI: 6%–10%), with substantial heterogeneity (I² = 87.6%, τ² = 0.0837, p < 0.0001), as illustrated in the forest plot ( Supplementary Figure 12 ).

Sensitivity analysis demonstrated that omission of individual stages marginally altered the pooled estimate, ranging from 6.76% to 8.24%, without substantially reducing heterogeneity. For instance, excluding Stage II yielded the highest pooled estimate (8.24%, I² = 78.4%), while excluding Stage III resulted in the lowest (6.76%, I² = 84.7%). These results underscore the robustness of the pooled estimate and highlight Stage III and IV as critical contributors to the elevated VTE risk in cancer progression ( Supplementary Table 11 ).

While this meta-analysis offers valuable insights, several limitations must be acknowledged. First, notable heterogeneity in study design (hospital- vs. population-based) and VTE diagnostic methods may have influenced the pooled estimates. Second, the search strategy included broad terms such as “risk factors” but omitted outcome-specific terms like “incidence” or “proportion,” which may have reduced precision and biased retrieval toward analytical studies. Although only studies reporting VTE incidence or proportion were included, these issues may have affected search comprehensiveness. Additionally, the lack of MeSH in the search may have impacted precision and reproducibility. Citation chaining was also not systematically applied; while relevant reviews were consulted during the scoping phase, their references were not exhaustively screened, potentially limiting the retrieval of additional eligible studies. Substantial heterogeneity remained across pooled estimates, even after stratification by cancer type, stage, and study setting. This likely reflects the influence of unmeasured or inconsistently reported variables such as treatment regimens, thromboprophylaxis practices, and patient-level risk factors. Due to limited reporting, further meta-regression or subgroup analyses were not feasible. Future updates should refine the search strategy by incorporating outcome-specific and MeSH terms, applying citation chaining, and emphasizing the need for standardized reporting in primary studies to support more robust meta-analytic modelling.

Cancer staging harmonization across studies was limited by inconsistent reporting of TNM components. While the TNM system incorporates tumor size (T), nodal involvement (N), and metastasis (M), several included studies reported only partial staging information. In such cases, we conservatively assigned stage groupings using available data, following accepted clinical conventions, as detailed in Supplementary Table 8 . Although this approach enabled retention of stage as a key stratifying variable, we acknowledge that it introduces potential for staging misclassification. Future studies and meta-analyses would benefit from standardized and complete TNM reporting to support more accurate stage-based comparisons.

Due to the small number of included studies and variability in risk of bias assessment, we were unable to conduct a formal sensitivity analysis excluding high-risk studies or apply quality-weighted models. This limits our ability to quantitatively assess the influence of lower-quality studies such as Grilz et al. (2024) on pooled estimates. Future meta-analyses with larger and more homogeneous datasets may consider integrating quality-based weighting or subgroup exclusion to enhance validity.

Patient demographics, including age, genetic predisposition, and lifestyle factors, were not uniformly accounted for, which may affect generalizability. Additionally, important clinical variables such as cancer-related surgery, chemotherapy, and radiotherapy were inconsistently reported across studies. These treatment-related exposures may have acted as confounding factors, potentially influencing the observed VTE proportions and contributing to between-study variability. Also, as authors, we acknowledge the importance and interpretive utility of the Number Needed to Treat (NNT) in clinical research. The NNT offers a straightforward and intuitive measure of the effectiveness of a healthcare intervention by estimating the number of patients who need to receive a particular treatment for one additional patient to benefit (42). However, NNT is derived from absolute risk reduction (ARR), which presupposes the existence of a control or comparison group, typically in randomized controlled trials (RCTs). Given that the included studies in this review were predominantly observational and not designed as interventional trials, the requisite conditions for calculating NNT were not met. As such, we were unable to compute NNT values within the current meta-analytic framework. Furthermore, the literature search was restricted to a single database (PubMed), which may have limited the comprehensiveness of the evidence base. Future reviews should consider incorporating additional databases to enhance the breadth and representativeness of the included studies. While efforts were made to avoid duplication, two population-based Swedish prostate cancer studies may involve overlapping cohorts. This potential duplication could have inflated the pooled population denominators and VTE event counts, thereby subtly biasing the overall estimates.