Risk of venous thromboembolism in patients with cancer: a systematic review and meta-analysis

Abstract

Background: People with cancer are known to be at increased risk of venous thromboembolism (VTE), and this risk is believed to vary according to cancer type, stage of disease, and treatment modality. Our purpose was to summarise the existing literature to determine precisely and accurately the absolute risk of VTE in cancer patients, stratified by malignancy site and background risk of VTE.

Methods and findings: We searched the Medline and Embase databases from 1 January 1966 to 14 July 2011 to identify cohort studies comprising people diagnosed with one of eight specified cancer types or where participants were judged to be representative of all people with cancer. For each included study, the number of patients who developed clinically apparent VTE, and the total person-years of follow-up were extracted. Incidence rates of VTE were pooled across studies using the generic inverse variance method. In total, data from 38 individual studies were included. Among average-risk patients, the overall risk of VTE was estimated to be 13 per 1,000 person-years (95% CI, 7 to 23), with the highest risk among patients with cancers of the pancreas, brain, and lung. Among patients judged to be at high risk (due to metastatic disease or receipt of high-risk treatments), the risk of VTE was 68 per 1,000 person-years (95% CI, 48 to 96), with the highest risk among patients with brain cancer (200 per 1,000 person-years; 95% CI, 162 to 247). Our results need to be considered in light of high levels of heterogeneity, which exist due to differences in study population, outcome definition, and average duration of follow-up between studies.

Conclusions: VTE occurs in greater than 1% of cancer patients each year, but this varies widely by cancer type and time since diagnosis. The absolute VTE risks obtained from this review can aid in clinical decision-making about which people with cancer should receive anticoagulant prophylaxis and at what times.

Figure 1. Flow diagram for study selection.

Figure 2. Pooled incidence of venous thromboembolism for overall cancer.

Natural logarithms of the incidence rate are presented on the x-axis. Black diamonds indicate the point estimate (VTE incidence) for each individual study. Horizontal lines indicate the 95% confidence interval surrounding this estimate. Open blue diamonds describe both the point estimate (centre of the diamond) for the pooled VTE incidence for average-risk, high-risk, and all studies (average- and high-risk together), and the 95% confidence interval for this pooled estimate (width of the diamond). Blom (2006) indicates data from .

Figure 3. Pooled incidence of venous thromboembolism for breast cancer.

Natural logarithms of the incidence rate are presented on the x-axis. Symbols as in Figure 2. Blom (2006) indicates data from .

Figure 4. Pooled incidence of venous thromboembolism for lung cancer.

Natural logarithms of the incidence rate are presented on the x-axis. Symbols as in Figure 2. Blom (2006) indicates data from .

Figure 5. Pooled incidence of venous thromboembolism for colorectal cancer.

Natural logarithms of the incidence rate are presented on the x-axis. Symbols as in Figure 2. Blom (2006) indicates data from .

Figure 6. Pooled incidence of venous thromboembolism for prostate cancer.

Natural logarithms of the incidence rate are presented on the x-axis. Symbols as in Figure 2. Blom (2006) indicates data from .

Figure 7. Pooled incidence of venous thromboembolism for brain cancer.

Natural logarithms of the incidence rate are presented on the x-axis. Symbols as in Figure 2. Blom (2006) indicates data from .

Figure 8. Pooled incidence of venous thromboembolism for bone cancer.

Natural logarithms of the incidence rate are presented on the x-axis. Symbols as in Figure 2. Blom (2006) indicates data from .

Figure 9. Pooled incidence of venous thromboembolism for haematological cancer.

Natural logarithms of the incidence rate are presented on the x-axis. The studies reported by Chew et al. (non-Hodgkin lymphoma) and Ku et al. (leukaemia) were both from the California Cancer Registry cohort. However, because they included different subsets of patients and were conducted over different time intervals (Chew, 1993–1995, and Ku, 1993–1999), a decision was made to pool these as separate studies. Symbols as in Figure 2. Blom (2006) indicates data from .

Figure 10. Pooled incidence of venous thromboembolism for pancreatic cancer.

Natural logarithms of the incidence rate are presented on the x-axis. Blom (2006) indicates data from . Blom (2006a) indicates data from .

Figure 11. Relative risks of venous thromboembolism in cancer patients compared with in the general population.

Results for selected cancer types obtained from Cronin-Fenton et al. . IRR, incidence rate ratio.