The prevalence of thrombocytopenia in patients with acute cancer-associated thrombosis

Venous thromboembolism (VTE) is a common complication in patients with cancer, with thrombosis being the second leading cause of death among outpatients undergoing chemotherapy.^1,2 Because of the unique risk profile of patients with cancer, management of cancer-associated thrombosis (CAT) requires special considerations as compared with VTE in patients without cancer. Thrombocytopenia is also a common complication in patients with cancer, either due to the underlying malignancy or the toxicity of cancer-directed therapy. Clinical decision-making in patients with cancer who develop both thrombosis and thrombocytopenia is challenging, as thrombocytopenia increases the risk of bleeding without conferring protection against thrombosis. Randomized phase 3 anticoagulation trials in CAT often excluded patients with thrombocytopenia.^3-9 Recently released European Hematology Association clinical guidelines on the management of these complex scenarios further emphasize the everyday challenges faced by clinicians in this population.¹⁰ 

Data on the true prevalence of thrombocytopenia in patients with cancer and concomitant thrombosis are extremely limited.¹¹ Thrombocytopenia is common in hematologic malignancies and is often observed following certain cytotoxic chemotherapies in patients with solid tumors.¹² Accordingly, the Flatiron Health Electronic Health Record database of patients with cancer reported a 3-month cumulative incidence of thrombocytopenia of 13% (any grade, platelet count <100 × 10³/μL) in patients with solid tumors. Severe thrombocytopenia (platelet count <50 × 10³/μL) occurred in 6% of patients with solid tumors receiving chemotherapy and in 28% of patients with hematologic malignancies receiving chemotherapy.¹³ However, how frequently thrombocytopenia occurs in the setting of acute CAT is unknown. Although practicing clinicians frequently deal with the challenge of managing anticoagulation in patients with cancer and thrombocytopenia, the true scope of the problem is undefined.

This retrospective cohort study included adult patients (age ≥ 18 years) with cancer at Beth Israel Deaconess Medical Center (BIDMC) from the years 2010 through 2021 who had an index CAT event in any clinical setting (inpatient, outpatient, or emergency department). All study procedures were approved by the Institutional Review Board at BIDMC. The study was conducted according to the Declaration of Helsinki. ICD-9 and ICD-10 codes associated with clinical encounters were used to identify diagnoses of VTE (ICD-9 415, 451-453, and ICD-10 I26, I67, I80, and 182) and cancer (ICD-9 140-239 and ICD-10 C00-D49). The positive predictive value of ICD coding for VTE has previously been validated.^14-22 Eligibility criteria required a new VTE diagnosis (first occurrence of VTE diagnosis during the study period) associated with a first or new occurrence of a cancer diagnosis (within prior 6-month period). The 6-month temporal association of VTE and cancer diagnosis was implemented to exclude patients with a remote history of cancer. Qualifying VTE diagnoses included acute pulmonary embolism (PE), deep vein thrombosis (DVT; divided into proximal, distal, unspecified, and upper extremity DVT), cerebral venous thrombosis (CVT), abdominal thrombosis (portal vein thrombosis, Budd-Chiari Syndrome, inferior vena cava thrombosis, and renal vein thrombosis), and intrathoracic thrombosis (superior vena cava, brachiocephalic, and other unspecified thoracic veins). In instances of multiple entries, the initial occurrence of VTE was used. Chronic, superficial, or unspecified clots were excluded. Cancer type was classified and divided by system. Solid tumors included gastrointestinal (GI), lung/intrathoracic, genitourinary (GU), gynecologic, breast, head and neck, melanoma, nervous system, neuroendocrine (NET), and sarcoma. Hematologic malignancies included lymphoma, multiple myeloma, and myeloid malignancies. Remaining “neoplasm” diagnoses that were classified as benign, nonmelanoma skin cancer, secondary malignancy (metastasis), unspecified behavior, or not otherwise specified were excluded from the study. Those patients without available platelet count data in the 2 weeks before or after the index CAT event were excluded.

The lowest platelet count in the 2 weeks before or after the index CAT event was collected and classified according to Common Terminology Criteria for Adverse Events (CTCAE v.5.0) grading in relation to the lower limit of normal (LLN): not thrombocytopenic (>LLN), grade 1 (75 × 10³/μL – LLN), grade 2 (50-75 × 10³/μL), grade 3 (25-50 × 10³/μL), and grade 4 (<25 × 10³/μL).²³ The LLN was chosen to be 100 × 10³/μL in this study to exclude clinically insignificant or spurious decreases in platelet count in the 100 × 10³ to 150 × 10³/μL range and to preserve equal intervals of 25 × 10³/μL between grades of thrombocytopenia. Given the retrospective nature of the study, the specific timeframe of 2 weeks before or after the index CAT event was chosen to ensure availability of platelet count data around the time of VTE diagnosis, as the coding date may differ slightly from the clinical event date. The proportion of index CAT events associated with different grades of thrombocytopenia was analyzed according to cancer type (solid vs hematologic). The temporal trend in proportions of thrombocytopenia and cancer types were also analyzed by year (2010-2020) to assess whether there were changes in relative proportion or the number of VTE over time. No formal sample size estimates were performed. Exact binomial confidence intervals were calculated using a 95% confidence level.

A total of 12 198 VTE were identified between October 2010 and June 2021. Excluding duplicate records, ineligible VTE, and absence of qualifying malignancy, the final number of unique patients identified meeting eligibility criteria was 3635 index CAT (Figure 1).

The overall cohort comprised 47% females, with a mean age 65.7 years (standard deviation of 12.8 years). Of the 3635 unique patients with CAT identified, 80% were associated with solid tumors, 18% with hematologic malignancies, and 2% with multiple concurrent cancer diagnoses (Table 1).

Of all unique CAT events identified, 41% were PE (8% with concurrent DVT), 47% DVT, 21% abdominal thrombosis (predominantly portal vein thrombosis), and 2% intrathoracic thrombosis or cerebral sinus thrombosis (Table 1). Notably, PE occurred more often in solid tumors (43%, 95% CI 41%-45%) than in hematologic malignancies (32%, 95% CI 28%-35%). CAT events in hematologic malignancies predominantly comprised DVT (69%, 95% CI 65%-72%) compared with solid tumors (41%, 95% CI 40%-43%). This was in large part due to an increased proportion of upper extremity DVT (34%, 95% CI 30%-38%) in hematologic malignancies vs solid tumors (12%, 95% CI 11%-13%). A high proportion of solid tumor CAT events were abdominal thrombosis (24%, 95% CI 22%-26%), predominantly in gastrointestinal cancers (44%, 95% CI 41%-46%), whereas the proportion was far fewer in hematologic malignancies (7%, 95% CI 5%-9%) (supplemental Table 1).

In the overall cohort, CAT with concurrent thrombocytopenia of any grade (<100 × 10³/μL) was identified in 27% (95% CI 25%-28%) of patients, and grade 3 or 4 thrombocytopenia (<50 × 10³/μL) was observed in 11% (95% CI 10%-13%) of patients. In the cohort with CAT and solid tumors, thrombocytopenia of any grade occurred in 22% (95% CI 21%-24%), and grade 3 or 4 thrombocytopenia occurred in 7% (95% CI 6%-8%) of cases. Gastrointestinal malignancies had the highest prevalence of thrombocytopenia among solid tumor diagnoses (32% any grade and 10% grade 3-4) (Table 2).

In CAT of hematologic malignancies, the corresponding proportions were 47% (95% CI 43%-51%) for any grade thrombocytopenia and 30% (95% CI 27%-34%) for grade 3 or 4 thrombocytopenia. Myeloid malignancies had the highest prevalence of concurrent thrombocytopenia (65% any grade and 50% grade 3-4), with a significant portion (40%) qualifying as grade 4 (Table 2).

For CAT events with PE, thrombocytopenia of any grade occurred in 19% of cases, and grade 3 to 4 thrombocytopenia occurred in 8% of cases. CAT events with DVT had higher rates of thrombocytopenia, with 28% of any grade and 14% of grade 3 to 4. Abdominal thrombosis was also commonly associated with thrombocytopenia (41% any grade and 14% grade 3-4).

Temporal trends were also analyzed within the study period. The mix of cancer types during this period also remained similar, with 77% to 82% in solid tumors, 16% to 21% in hematologic malignancies, and 2% to 4% with multiple malignancies (Figure 2A). The proportion of PE associated with CAT cases over the study period remained stable, averaging 41% with a range of 37% to 49%. The proportion of DVT averaged 47% and ranged from 40% to 54%.

The proportion of CAT with thrombocytopenia remained fairly stable across the 10-year study period, with normal platelet count in 70% to 75% of CAT events, any grade thrombocytopenia in 25% to 30% of events, and grade 3 to 4 thrombocytopenia in 8% to 14% of events (Figure 2B).

Thrombocytopenia is known to frequently occur in the setting of CAT, but to the best of our knowledge, this is the first study providing point estimates of the prevalence of concurrent thrombocytopenia and CAT. Nearly 1 in 4 patients with solid tumor diagnosis and 1 in 2 with hematologic malignancies were diagnosed with CAT and thrombocytopenia. Notably, severe thrombocytopenia (<50 × 10³/μL) was present in 30% of all patients with hematologic malignancy and a diagnosis of new VTE.

The distribution of VTE in the cohort differed between the solid tumor and hematologic malignancy groups. Among the solid tumor cohort, the distribution between DVT and PE was fairly even, except in the gastrointestinal group, who frequently were diagnosed with intra-abdominal DVT. Concurrent thrombocytopenia and intra-abdominal thrombosis were particularly common (every 2.5 cases). This high rate of concurrent thrombocytopenia and abdominal thrombosis is likely a reflection of the myelosuppressive regimens used to manage advanced gastrointestinal malignancies such as pancreatic cancer.^24-27 Cirrhosis is also associated with an increased risk for portal vein thrombosis and often leads to thrombocytopenia. There was a higher prevalence of DVT in the hematologic malignancy group compared with the solid tumor group (69% vs 41%), driven largely by higher rates of upper extremity DVT (34% vs 12%), which is attributed to the frequent use of catheter thrombosis in hematologic malignancy.^28,29 

Severe thrombocytopenia is commonly observed in patients with hematologic malignancies. Chemotherapy-induced thrombocytopenia in diffuse large B-cell lymphoma patients receiving multiagent chemotherapy is associated with grade 3 or 4 thrombocytopenia in ∼7% of patients on clinical trials.^30-32 In our cohort with lymphoma, grade 3 or 4 thrombocytopenia was present in 25% of patients with acute VTE. Similarly, in clinical trials for myeloma, the incidence of thrombocytopenia (any grade) is between 6% and 15%, whereas in our cohort, 41% of patients had thrombocytopenia.^33-35 These data suggest that thrombocytopenia may be more common among patients with VTE compared with those without VTE. Thrombocytopenia has been linked to VTE in cancer via direct activation of platelets.³⁶ 

As the management of malignancies is transitioning to more targeted therapies rather than conventional cytotoxic regimens (that are often more myelosuppressive), we analyzed the trends in VTE and thrombocytopenia over time.³⁷ Interestingly, the prevalence of thrombocytopenia with CAT has remained largely unchanged over the last decade. Similarly, the distribution of cancer diagnosis and the prevalence of severe thrombocytopenia was fairly consistent over the timeframe assessed. There was a transition from ICD-9 to ICD-10 during the study period (starting late 2015), and there were numerically more VTE diagnosed following the transition. However, we did not observe an impact on tumor type, VTE type, or the prevalence of concurrent thrombocytopenia. The increase in VTE diagnosis in cancer cohorts is consistent with the findings of a recent Danish population-based study demonstrating a continual rise in VTE diagnosis over the decade.³⁸ We note that in 2020 there was a sudden decrease in CAT diagnosis, which is likely due to the COVID-19 pandemic and changes in health care use and diagnosis of nonrespiratory illnesses.^32,39 

Inherent to retrospective analyses, we recognize there are limitations to this study. Inclusion into the VTE cohort required a VTE diagnosis within 6 months of a new cancer diagnosis. The purpose of this requirement was to better ensure attribution of VTE relative to the underlying cancer diagnosis or cancer-directed therapy rather than include VTE many years after early cancer diagnosis. However, we acknowledge that the true incidence of CAT with thrombocytopenia may be underestimated, as a population of patients with more advanced disease many years after their initial diagnosis were excluded. This study was retrospective and cross-sectional in nature, focusing on thrombocytopenia at the time of the index CAT event; however, thrombocytopenia further removed in time and duration of thrombocytopenia are also of clinical relevance. We speculate that it is likely that the incidence of thrombocytopenia in the 3- to 6-month period following VTE is higher than the already high baseline coincidence of thrombocytopenia and VTE in this cohort. Future analyses of these aspects may provide further insight into the evolving thrombotic vs bleeding risk profile of CAT. The database is limited to a single institution such that there may be favored chemotherapeutic regimens that could affect the co-occurrence of thrombocytopenia and CAT. However, we believe that the relative proportion of VTE diagnoses in our thrombocytopenia cohort is generally applicable. In prior studies defining the risk and overall incidence of VTE according to solid tumor diagnosis, the greatest number of events were GI, followed by lung, breast, GU, and gynecologic malignancies.^40-42 Similarly, in our cohort, the highest proportion were GI, followed by lung, GU, and gynecologic malignancies. Although chemotherapeutic regimens have evolved over the last decade, the relative prevalence of thrombocytopenia and CAT was consistent over an extended period across tumor diagnosis. The use of ICD coding as the basis for establishing diagnoses raises the possibility of mislabeling. The accuracy of ICD coding has been studied and validated in several other studies for both VTE and various cancer subtypes, with positive predictive values between 75% to 90%.^14-20 What is less well established is the sensitivity of ICD coding for VTE, as we cannot exclude the possibility of an underestimation due to the absence of coding, as is suggested by the increase in VTE diagnoses after changing to ICD-10 coding. Nonetheless, even with the increased diagnosis of CAT with ICD-10 implementation, the relative proportion of patients with thrombocytopenia remained stable. We also acknowledge that some VTE events or platelet counts may have been diagnosed or measured outside our hospital medical record system.

This study highlights the high coprevalence of thrombocytopenia and CAT. Approximately 1 in 5 patients with solid tumors and 1 in 2 patients with hematologic malignancies with a diagnosis of VTE have concurrent thrombocytopenia. These point estimates serve to better define the scope of the problem and serve as justification for additional clinical trials addressing appropriate anticoagulation for acute CAT with thrombocytopenia.

The authors thank George Silva of the BIDMC InSIGHT Core for his collaboration and support in database query design and data extraction. The visual abstract was created on BioRender.com.

Contribution: C.H., R.P., and J.I.Z. were involved in the conception and design of the study, and revised the manuscript; C.H. performed the primary data collection, analysis, and drafted the manuscript.

Conflict-of-interest disclosure: C.H. has received consulting fees from Anthos Therapeutics and was previously employed at Baxter/Baxalta. J.I.Z. received research funding from Incyte and Quercegen; provided consultancy services to Sanofi, CSL, and Parexel; and reports honoraria from/advisory board participation with Pfizer/Bristol Myers Squibb, Portola, and Daiichi. R.P. declares no competing financial interests.

Correspondence: Jeffrey I. Zwicker, Memorial Sloan Kettering Cancer Institute, 1275 York Ave, New York, NY 10065; e-mail: zwickerj@mskcc.org.