Associations Between Immune-Related Venous Thromboembolism and Efficacy of Immune Checkpoint Inhibitors: A Systematic Review and Meta-Analysis

Abstract

This study aims to summarize the available data and determine if the presence of venous thromboembolism (VTE) immune-related adverse event (irAE) in patients with immune checkpoint inhibitor (ICI) therapy is associated with improved treatment efficacy and clinical outcomes, which in turn was used to help optimize patient selection for anticoagulation therapy and inform rational treatment strategies for overcoming the mechanisms of ICI resistance. PubMed, Embase, Web of Science, and Cochrane Library were searched up to March 18, 2023, for studies assessing the relationship between VTE irAE development during ICI therapy and cancer outcomes. Seven primary articles with a total of 4437 patients were included in the overall survival (OS) meta-analysis. Patients with VTE had a significant increase in overall mortality compared to patients without VTE in adjusted hazard ratios (HRs 1.36, 95% confidence interval [CI] 1.06-1.75, P  =  .02). In the studies where immortal time bias (ITB) was accounted for, patients with VTE irAE also had poor OS than those without. HR and the corresponding 95% CI values in the non-ITB group were 2.53 (1.75-3.66, P < .00001) with low heterogeneity (P  =  .17, I²  =  48%) and 1.21 (1.06-1.37, P  =  .004) in the ITB group with no heterogeneity (P  =  .95, I²  =  0%), respectively. Despite the heterogeneity identified, the evidence does suggest that VTE irAE occurrence could be served as a prognostic indicator, with higher frequencies of occurrence associated with poorer OS. However, the fundamental role of this association with clinical consequences should be further investigated in large cohorts and clinical trials.

Introduction

Venous thromboembolism (VTE), consisting mainly of deep vein thrombosis (DVT) and pulmonary embolism (PE), is a significant contributor to the worldwide disease burden. It has been long established that the presence of cancer is strongly associated with the occurrence of thromboembolic events (TEs). ¹ Patients diagnosed with malignancy are at a significantly elevated risk of VTE, comprising approximately 20% of all cases of this complication and exhibiting a 4 to 9-fold greater likelihood of developing VTE compared to the general population.^2,3 Meanwhile, cancer-associated thrombosis (CAT) is linked to increased morbidity and mortality rates, heightened probability of tumor progression, treatment delays, and augmented healthcare expenditures.^4,5 Consequently, VTE accounts for 9% of mortality in individuals with cancer, positioning it as the second most prevalent cause of death in this population. ⁶ Despite the widely recognized incidence of thrombotic events in individuals with a cancer diagnosis, it is crucial to acknowledge that the likelihood of VTE occurrence is observed with the administration of newer anticancer therapies, such as platinum-based chemotherapy, anti-angiogenesis agents, and more recently, immunotherapy. ⁷

With the advent of immune checkpoint inhibitors (ICIs), the landscape of cancer treatment has changed dramatically, with treatment modality for a wide spectrum of malignancies, such as non-small-cell lung cancer (NSCLC), melanoma, renal cell carcinoma (RCC), head and neck squamous cell carcinoma, and more.^8–10 At present, the impairment of tumoral immune-escape mechanisms is achieved by the targeting of programed cell death protein 1 or its ligand (PD-1/PD-L1) or cytotoxic T-lymphocyte-associated antigen 4 (CTLA-4) through the use of checkpoint inhibitors. ¹¹ Different from conventional chemotherapies and targeted anticancer agents, ICIs enhance the autoimmune system's ability to destroy cancer cells by blocking the expression of negative regulators on cancerous or immune cells. Nevertheless, the administration of ICIs may result in immune-related adverse events (irAEs), which are characterized by atypical reactions of the immune system toward non-targeted organs or systems, such as the skin, thyroid, gastrointestinal tract, or hematological system, including autoimmune hemolytic anemia or thrombocytopenia.^12,13 It should be noted, however, that ICIs can trigger immune responses against non-targeted organs and systems, known as irAEs, such as the skin, thyroid, gastrointestinal tract, or hematological system, including autoimmune hemolytic anemia or thrombocytopenia.^12,13 It is noteworthy that ICI-related cardiovascular toxicities including both cardiac and peripheral vascular diseases should not be neglected. ¹⁴

One of the most common adverse effects associated with ICI therapy is VTE. However, little is currently known regarding the actual incidence of VTE in initial phase II and III clinical trials that culminated in the approval of ICIs.^15–17 It has been reported in several meta-analyses that there is a possibility of VTE in patients receiving ICIs but that the real incidence is less than 5% in patients taking ICIs.^18–20 Several recent meta-analyses started to report concerns for thrombosis and suggested that the real incidence of VTE in patients on ICIs was less than 5%.^18–20 In turn, analysis of existing data has demonstrated that the condition of hypercoagulability may potentially affect the efficacy of immunotherapy and long-term prognosis of cancer patients. More recent pre-clinical studies indicated that coagulation factors, specifically factor X, play a role in cancer immune evasion and might consequently promote resistance to ICIs.^21,22 Multiple other pre-clinical and clinical studies^23,24 showed similar results for the synergistic impact of FXa inhibitors in conjunction with clinical ICI therapy.

Although VTE has been highly correlated with a poorer prognosis in patients with cancer,^25,26 there is a lack of specific evidence concerning the correlation between VTE and clinical outcomes in patients with cancer receiving ICI-based immunotherapy. Recently, it was reported that CAT events were correlated with a poorer prognosis in patients with cancer receiving therapy with ICIs.^27–29 However, the results in a small number of studies indicated that the occurrence of thrombosis exhibited no significant effects on survival of these patients.^30,31 Further, previous meta-analyses on underlying vascular adverse events for patients treated with ICIs primarily focused on the risk of VTE occurrence rather than prognostic impact on clinical outcomes.^18–20 To date, current evidence regarding the relationship between VTE and clinical outcomes in patients with cancer receiving ICI-based immunotherapy is lacking. Therefore, we performed a systematic review and meta-analysis to determine if the presence of VTE irAE in patients with ICI therapy is associated with enhanced treatment effectiveness and clinical outcomes, which in turn was used to help optimize patient selection for anticoagulation (AC) therapy and develop rational treatment strategies for overcoming ICI resistance mechanisms.

Materials and Methods

Reporting and Protocol Registration

According to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines ³² (http://www.prisma-statement.org/), we identified relevant studies for systematic review and meta-analysis. This systematic review is registered at Prospero (http://www.crd.york.ac.uk/PROSPERO/) with registry number CRD42023408003.

Search Strategy

We performed an unrestricted literature search of the PubMed/Medline, Embase, Web of Science, and Cochrane Library databases for relevant, English-written articles from inception to March 18, 2023. Search terms included MesH, EMTREE, and Web of Science headings and free-text terms covering variations in terminology for neoplasms, ICIs, VTE, and survival terms. Each database was individually searched by title, abstract, keyword, and subject heading for relevant search terms. We combined the aforementioned four groups of search terms, namely, (1) neoplasms, (2) ICIs, (3) VTE, and (4) survival. A full description of the search terms and strategy is provided in Supplemental Tables 1 and 2. Besides, the reference lists of all retrieved articles were scrutinized to detect any potentially relevant studies.

Selection Criteria

A comprehensive search of all full-text studies and abstracts covering VTE events and survival outcomes in patients with solid tumors treated with ICIs was conducted. The inclusion criteria for the selected studies were delineated as follows: (1) patients: adult patients who were histologically confirmed malignant solid tumors; (2) intervention methods: all patients received immunotherapy with ICIs, including PD-1 or PD-L1 inhibitors or CTLA-4 inhibitors, as either first-line, second-line, or multiple lines of therapy; (3) comparison factor: we compared clinical outcomes such as overall survival (OS) or PFS in patients receiving immunotherapy between the cases and controls. The cases would refer to patients who were diagnosed with VTE while undergoing ICI treatment within 6 months after the initiation of immunotherapy, while the controls would represent patients without VTE. Both symptomatic VTE and incidental VTE detected incidentally were confirmed by objective imaging such as complete compression venous ultrasonography (CUS) for DVT and computed tomography pulmonary angiography (CTPA) or ventilation/perfusion lung scan for PE; (4) outcome: the analyzed clinical outcomes included the progression-free survival (PFS), OS, objective response rate (ORR), and disease control rate (DCR). The effect measure of time-to-event outcomes (such as OS or PFS) was measured by hazard ratio (HR) with the corresponding 95% confidence intervals (CIs); and (5) study design: we included the types of observational studies including prospective/retrospective cohort studies, as well as both population-based and hospital-based case–control studies in the systematic review. Studies that were potentially eligible were selected for full-text review.

As a rule of thumb, the following criteria were excluded: (1) conference abstracts, case reports, reviews, editorial letters, summaries of meetings or comments, meta-analysis, or in vitro and animal studies; (2) insufficient OS or PFS outcomes information provided; (3) duplicate publications; (4) ongoing studies with results not published at the time of the literature search; (5) studies that did not provide OS or PFS outcome data specific to VTE irAE; and (6) studies only reporting Kaplan–Meier curves without directly reporting HR or 95% CIs. A kappa coefficient value was calculated for article screening to evaluate inter-reviewer agreement. ³³

Data Extraction and Management

The identified records were subjected to duplicate removal through the use of EndNote Reference Manager software (version X9, Thomson Reuters, New York, NY, USA). Two investigators (HML and HL) independently screened and assessed the eligibility of identified publications in an unblinded standardized manner. The study selection process involved an initial review of titles and abstracts for eligibility, followed by full-text assessment of potentially relevant articles. Data extraction of the included studies was performed independently by two reviewers (HML and HL) using a pre-piloted standardized data extraction form. Any disagreements were resolved through consensus and adjudicated by a third reviewer (TL), if necessary. Figure 1 shows the PRISMA flow diagram of our review.

Figure 1.

Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram of the study selection process.

Abbreviations: VTE, venous thromboembolism; AT, arterial thrombosis; HR, hazard ratio; CIs, confidence intervals; OS, overall survival; PFS, progression-free survival.

From each of the eligible studies, the following data were collected: (1) general data (study design, first author's name, year of publication, and country), (2) characteristics of trial participants (numbers of patients, types of VTE, cancer histotype, and VTE incidence), (3) type of intervention (type of immunotherapy, treatment line, and duration), (4) study quality characteristics (whether immortal time bias [ITB] was accounted for), and (5) type of outcome measures (PFS, OS, and DCR). A pooled HR and 95% CI were calculated from each study to estimate the prognostic role of VTE in patients with solid tumors treated with ICIs.

Quality Assessment

Two of the authors (HML and HL) independently evaluated the methodological quality of the included studies using a modified version of the Newcastle–Ottawa Scale (NOS), ³⁴ which assigns a maximum of 4 points for selection, 2 points for comparability, and 3 points for outcomes. Studies scoring ≤3 was considered low quality, while a study with a score of ≥7 was regarded as a high-quality study. Each study was assigned a score from 0 to 9 points, and higher NOS points indicated higher literature quality. Discrepancies were resolved via further discussions among the coauthors (HML, HL, and TL).

Statistical Analysis

All of the statistical analyses were performed using RevMan 5.3 software (Cochrane Collaboration, Oxford, UK) and STATA software version 14.0 (StataCorp, College Station, TX, USA). The effect size of PFS and OS was pooled through HR and its 95% CI, and DCR data were evaluated by odds ratio (OR) with 95% CI. The heterogeneity of the included studies was assessed using the χ²-based Cochran Q test and I² statistic (A P-value of Cochran's Q test < 0.10 or I² statistic >50% indicated significant heterogeneity). The study conducted subgroup and sensitivity analyses to identify the sources of heterogeneity, with negligible heterogeneity leading to the implementation of a fixed-effects model. The random-effects model was applied for the pooled data if moderate or high level heterogeneity exists. Additionally, subgroup analyses were performed to examine the impact of patient and study quality characteristics on the development of VTT irAE during ICI therapy and potential cancer treatment outcome. Qualitative evaluation of publication bias was conducted through graphical visualization of funnel plots, while quantitative evaluation was performed using Begg's and Egger's tests. Additionally, to assess the consistency of the primary results, sensitivity analysis was performed for the overall summary effects by sequentially removing each study and re-running the analysis. A two-sided P-value less than .05 was considered to be statistically significant.

Results

Search Results and Study Characteristics

Initially, a total of 478 records were retrieved by searching the PubMed, Embase, Web of Science, and Cochrane Library databases, of which 111 references were excluded because of duplicate publications. The remaining 367 articles were excluded by reading the titles and abstracts of the articles based on the inclusion and exclusion criteria, including 74 conference abstracts, 26 case reports, 127 reviews/meta-analysis, 5 editorial letters, and 108 clearly irrelevant references. The full text of 25 articles was assessed for inclusion, and 13 eligible studies were finally included in the present systematic review.^{27–31,35–42} Following this, we excluded articles that did not provide OS or PFS outcome data specific to VTE irAE^27,39,41,42 and those that did not show HR or CIs of OS and/or PFS.^38,40 Finally, 7 primary articles with a total of 4437 patients were included in the meta-analysis.^{28–31,35–37} The literature search process is shown in the PRISMA flow diagram (Figure 1). An assessment of inter-reviewer reliability using Cohen's Kappa statistic yielded good agreement (Kappa  =  0.83).

The sample sizes for the included 13 studies ranged from 133 to 1686 participants (median n  =  552), with the percentage of male ranging from 43.1 to 72.0%. Of the 13 studies, 11 were prospective observational studies, whereas the remaining 2 were retrospective cohort analyses. Of these studies, 3 evaluated both PFS and OS,^28,29,31 and null of the studies just assessed the PFS. Regarding the type of tumor, the most common site of origin in this review was lung cancer (56.4% of cases, of which 33.6% were explicitly NSCLC), followed by melanoma (19.9%) and genitourinary cancer (6.9%). As for the blockade of ICIs, PD-1 inhibitors were the most frequently prescribed, while mixed/combined therapy (PD-1/PD-L1 inhibitor with a CTLA-4 inhibitor) was only utilized in 6 studies. The range of publication year for included studies was from 2019 to 2023. The studies were conducted in 9 different countries, of which the majority came from the USA (44%) followed by Spain (22%). The median follow-up durations varied among the included studies, ranging from 5.6 to 37.8 months, and the incidence of VTE/AE irAE was between 3% and 24%. Table 1 provides a summary of the key characteristics of all eligible studies.

Table 1.

Systematic Review of the Studies Evaluating the Correlation Between the Development of Immune-Related Venous Thromboembolism (VTE) and Clinical Outcomes Secondary to Immune Checkpoint Inhibitors (ICIs).

Author/Year	Study Design	Country	Sample Size (M/F)	Disease(s)	ICI type	Treatment Line	ITB	Follow-up (median [IQR])	VTE/AE Incidence (%, 95% CI)	Outcomes (VTE/AT irAE vs no VTE/AT irAE)
										OS	PFS	DCR
Bar et al 2019 38 a	Retrospective	Israel	1215 (717/498)	NSCLC (42%), melanoma (35.5%)	PEMBRO (45.2%), NIVO (38.7%), IPI (3.2%)	Not reported	Yes	12 mo	6 mo: 2.6 12 mo: 3.0	3 vs 14 mo HR 3.01 (P < .0001)	Not reported	Not reported
Nichetti et al 2019 39 a	Prospective	Italy	217 (136/81)	NSCLC	Anti-PD-1 (69.6%), anti-PD-L1 (26.7%), anti-PD-L1  +  anti-CTLA-4 (3.7%)	Line<2 (25.4%) Line ≥2 (74.6%)	No	37.8 (22.6-43.9)mo	13.8	Adjusted HR 2.93 (P  =  .0006)	Not reported	Not reported
Moik et al 2021 28	Retrospective	Austria	672 (402/270)	Melanoma (30.4%), NSCLC (24.1%), RCC (11.0%)	NIVO (42.0%), PEMBRO (40.0%), IPI (6.7%), IPI plus NIVO (6.0%), ATEZO (4.5%), AVELU (0.9%)	Line 2 (12.5%) Line 3 (2.8%)	No	23.1 mo	6 mo: 5.0 (3.4-6.9) Overall: 12.9 (8.2-18.5)	11.6 vs 25.5 mo HR 3.09 (P < .001)	1.7 vs 6.7 mo HR 3.63 (P < .001)	OR 0.84 95% CI 0.43-1.63
Kewan et al 2021 30	Retrospective	USA	552 (359/193)	Lung (47.3%), genitourinary (23.2%), melanoma (17.2%)	NIVO (62%), PEMBRO (26.6%), IPI (14.5%), ATEZO (12%)	Not reported	Yes	12 mo	12.1	12.8 vs 13.4 mo HR 1.044 (P  =  .791)	Not reported	Not reported
Deschênes-Simard et al 2021 31	Retrospective	Canada	593 (332/271)	NSCLC	Anti-PD-1 (91.9%), with NIVO (53.5%)	Line 1 (29.3%) Line ≥2 (70.7%)	Yes	12.7 (4.9-22.7) mo	9.9 (7.5-12.3)	15.3 vs 18.9 mo Adjusted HR 1.10 (P  =  .57)	3.4 vs 5.5 mo Adjusted HR1.27 (P  =  .17)	1.05 (0.61-1.80) P  =  .86
Gutierrez-Sainz et al 2021 35	Retrospective	Spain	229 (146/83)	Lung (48.0%), melanoma (23.6%) RCC (11.8%)	PEMBRO (40.2%), NIVO plus IPI (8.4%)	Not reported	Yes	9.8 mo	7 (4-10)	3.53 vs 18.76 mo Adjusted HR 1.33 (P  =  .44)	Not reported	Not reported
Guven et al 2021 36	Retrospective	Turkey	133 (86/47)	RCC (26.3%), melanoma (24.1%)	NIVO (46.7%), IPI (20%), PEMBRO (20%), AVELU (13.3%)	Line ≤2 (48.1%) Line ≥3 (51.9%)	Yes	10.1 (5.8-18.5) mo	11.3	15.8 vs 24.9 mo Adjusted HR 1.208 (P  =  .616)	Not reported	Not reported
Roopkumar et al 2021 37	Retrospective	USA	1686 (1014/672)	Lung (49.6%), melanoma (13.2%)	NIVO (20%), PEMBRO (40%), ATEZO (13.3%), DURVA (6.6%), IPI plus NIVO (20%)	Not reported	Yes	438 (7-1971) days	6 mo: 7.1 12 mo: 10.9 Overall: 24	365 vs 453 days Adjusted HR 1.22 (P < .008)	Not reported	Not reported
Sussman et al 2021 27 a	Retrospective	USA	228 (154/74)	Melanoma	PEMBRO (38.7%), IPI plus NIVO (29.4%), IPI (20%), NIVO (12.3%)	Not reported	Yes	27.3 (9.3-38.9) mo	6 mo: 8.0 (4.9-12.0) 12 mo: 12.9 (8.9-17.7)	20 mo vs NR, adjusted HR 2.27 (P  =  .002)	Not reported
Alma et al 2022 40 a	Retrospective	France	481 (327/154)	Lung	NIVO (53%), PEMBRO (27%), DURVA (11%), ATEZO (9%)	The median number of ICI cycles administered was 7 (IQR: 3-14, range: 1-100)	Yes	9.8 (4.18-18.8) mo	9.8	42.5 vs 86.8 mo (P  =  .006)	Not reported	Not reported
Cánovas et al 2022 41 a	Retrospective	Spain	956 (616/340)	Lung (70%), melanoma (30%)	PEMBRO (41%), NIVO (27%), ATEZO (11%), DURVA (4%), IPI plus NIVO (2%)	Line 1 (52%) Line 2 (30%) Line ≥3 (18%)	Yes	Lung: 14 mo Melanoma: 17 mo	6.3	Lung: 12 vs 19 mo (P  =  .0049) Melanoma: 10 vs 29 mo (P  =  .034)	Not reported	Not reported
Sheng et al 2022 42 a	Retrospective	USA	279 (201/78)	Urothelial cancer	PEMBRO (57%), ATEZO (40%), NIVO (3%)	Median 5 (range 1-59)	No	5.6 (0.3-51.6) mo	13	14.9 vs 24.8 mo (P  =  .0014) Adjusted HR 2.296 (P  =  .0004)	Not reported	Not reported
Bjørnhart et al 2023 29	Prospective	Denmark	572 (274/298)	NSCLC	PEMBRO (77%), NIVO (20%), ATEZO (2%)	Line 1 (49%) Line ≥2 (51%)	No	16.5 (6.7-35.6) mo	Cumulative VTE events at 1, 3, 6, and 12 months were 7.5%, 9.6%, 13.0%, and 14.4%	Cohort A: 12.1 vs 23.1mo Cohort B: 6.5 vs 14.5 mo Adjusted HR 2.12 (P < .0001)	4.8 vs 7.0 mo HR 1.31, adjusted HR NS	OR 0.99, 95% CI 0.50-1.93

Abbreviations: ICIs, immune checkpoint inhibitors; VTE, venous thromboembolic events; AT, arterial thrombosis; M/F, man/female; PE, pulmonary embolism; IQR, interquartile range; MI, myocardial infarction; mo, months; NSCLC, non-small-cell lung cancer; RCC, renal cell carcinoma; PD1, programmed cell death 1; PD-L1, programmed death ligand 1; ATEZO, atezolizumab; PEMBRO, pembrolizumab; IPI, ipilimumab; DURVA, durvalumab; AVELU, avelumab; NIVO, nivolumab; ITB, immortal time bias; NR, not reached; NS, not significant; HR, hazards ratio; OR, odds ratio; CI, confidence intervals; OS, overall survival; PFS, progression-free survival; DCR, disease control rate.

^a The studies did not report OS or PFS outcome data specific to VTE irAE.

Systematic Review

Data Review According to Outcome

A total of 7813 participants were included in the OS analysis (13 studies), and 1837 participants were included in the PFS and DCR analysis (3 studies). Of the total 13 included studies, 9 (69.2%) studies investigating OS showed that the occurrence of VTE/AE irAE was a prognostic marker, as a higher frequency of occurrence was correlated with a poorer OS outcome,^{27–29,38–42} whereas 4 (30.8%) studies showed that ICIs-related VTE/AE occurrence was not independently associated with shorter OS.^30,31,35,36 Additionally, 2 studies reported poor correlations between VTE/AE irAE and PFS as the primary study endpoint,^29,31 while only 1 study showed that VTE/AE irAE could be served as unfavorable prognostic factors for PFS. ²⁸

Data Review According to Type of Thrombotic Event

A total of 13 studies assessed TE after immunotherapy. Of these, 7 studies evaluated exclusively VTE irAE,^{28–31,35–37} and 6 studies evaluated mixed arterial and venous thrombotic complications in general but also provided details on VTE irAE.^27,28–42 Furthermore, 3^28,29,37 of the 7 studies that specifically assessed VTE irAE were positive, meaning they observed an excellent correlation between ICIs-related VTE and poor clinical outcomes, and 4 were negative.^30,31,35,36 As for studies that reported on both arterial and VTE/AE irAEs, all of the studies were positive,^27,28–42 meaning that they were significantly associated with a worse prognosis.

Data Review According to Primary Tumor

Ten articles included a study population of NSCLC/lung cancer.^{28–31,35,37–41} Three studies evaluated populations containing only NSCLC (n  =  1382),^29,31,39 including only 1 prospective cohort study. ²⁹ OS and PFS were used as independent prognostic indicators in 2 of the 3 studies.^29,31 Of these, 2 (66.7%) studies showed that a high percentage of VTE irAE incidence was associated with a poorer prognosis,^29,39 whereas both studies reported that ICIs-related VTE could not serve as unfavorable prognostic factors for PFS.^29,31

Eight articles included a study population of melanoma.^{27,28,30,35–38,41} Of these 8 studies, 5 had positive findings that VTE/AE irAEs were significant factors associated with survival,^{27,28,37,38,41} while the remaining 3 had negative associations with OS.^30,35,36 Only 1 study evaluated participants exclusively with a diagnosis of melanoma. ²⁷ Findings from this study indicated that VTE/AE irAEs were closely correlated with poor OS. Additionally, 5 articles included urinary tract participants as part of their cohort.^{28,30,35,36,42} Of those, 3 (60%) studies exclusively evaluated participants with RCC (n  =  136).^28,35,36 One was associated with a shorter OS ²⁸ while the remaining were negative studies.^35,36 Besides that, 1 study evaluated participants exclusively with a diagnosis of urothelial cancer (n  =  279), ⁴² and this study was associated with positive findings.

Data Review According to Type of ICI Agent

Among the studies included, all studies treated patients with different PD-1 inhibitors. Of those, nivolumab was the most-involved anti-PD-1 antibody, either alone (53.8%) or in combination with ipilimumab (38.5%). Four studies included patients treated with PD-1/PD-L1 inhibitor monotherapy,^29,31,40,42 and 6 evaluated patients treated with anti-PD-1/PD-L1 and anti-CTLA4 combination therapy.^{27,28,35,37,39,41} Among the studies that administered PD-1/PD-L1 inhibitor monotherapy, 3 were positive, meaning they observed an excellent correlation between ICIs-related VTE and poor OS,^29,40,42 and only 1 observed a shorter OS/PFS that did not reach statistical significance. ³¹ Of those that included anti-PD-1/PD-L1 and anti-CTLA4 combination therapy, 5 were positive^{27,28,37,39,41} and only 1 was negative. ³⁵

Data Review According to ITB

Four of the 13 studies addressed the confounding impact of ITB on the results.^28,29,39,42 Of those, 3 of these 4 studies employed Cox regression analysis with time-dependent co-variables to control for potential ITB,^29,39,42 while 2 utilized Landmark analyses of survival as an approach to minimize ITB.^28,29 One study performed both Landmark and Cox regression models with time-varying covariates as part of survival analyses, ²⁹ whereas the other applied both Landmark and multi-state model treating thrombotic events as time-dependent co-variables. ²⁸ Of the studies that is account for ITB, all were positive, meaning they observed an excellent correlation between ICIs-related VTE/AE and poor OS.^28,29,39,42

Meta-Analysis

The Effect of VTE on Clinical Outcome

OS

Seven studies, comprising a total of 4437 participants, met the inclusion criteria for the OS meta-analysis.^{28,29,30,31,35–37} The meta-analysis of the included studies showed that the VTE irAE occurrence served as a prognostic indicator, with higher frequencies of occurrence associated with poorer survival results (HR 1.47, 95% CI 1.10-1.98, P  =  .01) (Figure 2A). The analysis showed a significant high degree of heterogeneity among all included studies (Cochran’s Q statistic, P < .0001, I²  =  79%). Further, the meta-analysis of OS was based on 1717 patients in 3 studies for univariate Cox proportional hazards regression analysis and 5 studies including a total of 3213 patients for multivariate HR analysis. HR and the corresponding 95% CI values according to univariate analysis were 1.32 (0.92-1.89, P  =  .13) with moderate heterogeneity (P  =  .04, I²  =  69%) and 1.36 (1.06-1.75, P  =  .02) according to multivariate analysis with moderate level heterogeneity (P  =  .06, I²  =  55%), respectively (Figure 2B and C). The multivariate results indicated that patients that developed VTE irAE had poorer OS than those without. The occurrence of VTE irAE was associated with an increased risk of death.

Figure 2.

Forest plots of pooled hazard ratios of venous thromboembolism (VTE) on overall survival in cancer patients treated with immune checkpoint inhibitors. (A) Overall survival in univariate and multivariate analysis. (B) Overall survival in the univariate Cox regression analysis. (C) Overall survival with multivariable-adjusted hazard ratios (HRs). The large diamond at the bottle of the plot indicates the overall summary estimate (width of the diamond represents the 95% confidence interval [CI]); the size of the squares indicates the weight of individual studies in the pooled analysis. Abbreviations: irVTE, immune-related venous thromboembolism.

PFS

Only 3 studies (n  =  1837) met the inclusion criteria to be included in the PFS meta-analysis.^28,29,31 ICIs-related VTE occurrence was not independently associated with shorter PFS (HR 1.82, 95% CI 0.93-3.55, P  =  .08) (Figure 3A), and highly significant heterogeneity was shown between studies (Cochran's Q statistic, P < .0001, I²  =  89%). In terms of the univariate analysis, only 2 studies that included a total of 1165 patients were available for our meta-analysis.^29,31 No significant heterogeneity existed for the meta-analysis of PFS in univariate analysis (P  =  .87, I²  =  0%), and the fixed model effect was used to synthesize the data. The combined HRs for PFS were 1.35 (95% CI: 1.04-1.71, P  =  .02) (Figure 3B), suggesting that there was a strong correlation between a higher frequency of VTE irAE occurrence and a poorer PFS. Given the limited number of included studies, we did not perform subgroup analyses for PFS.

Figure 3.

Forest plots of pooled hazard ratios of venous thromboembolism (VTE) on progression-free survival in cancer patients treated with immune checkpoint inhibitors. (A) Progression-free survival in univariate and multivariate analyses. (B) Progression-free survival in the univariate Cox regression analysis. (C) Progression-free survival with multivariable-adjusted hazard ratios (HRs). The large diamond at the bottle of the plot indicates the overall summary estimate (width of the diamond represents the 95% confidence interval [CI]); the size of the squares indicates the weight of individual studies in the pooled analysis. Abbreviations: irVTE, immune-related venous thromboembolism.

DCR

Three studies^28,29,31 (n  =  1837) included in the meta-analysis reported the data on the objective response analysis. The pooled meta-analysis showed that DCR did not significantly differ between the irVTE and no irVTE groups (OR 0.97, 95% CI 0.68-1.38, P  =  .86). The result had no heterogeneity corresponding to the I² value (P  =  .88, I²  =  0%), and the fixed-effects model was adopted (Figure 4). Similar results were obtained by analyzing the same parameters with a random-effect model. Then, we conducted sensitivity analysis to further explore heterogeneity of included studies, which proves that our conclusion was relatively stable and reliable (Supplemental Figure S6C). In summary, the DCR rate was not significantly lower in patients with irVTE than those without. There were no significant differences in efficacy outcomes found between the 2 groups.

Figure 4.

Forest plots of pooled odds ratios of venous thromboembolism (VTE) on disease control rate in cancer patients treated with immune checkpoint inhibitors. The large diamond at the bottle of the plot indicates the overall summary estimate (width of the diamond represents the 95% confidence interval [CI]); the size of the squares indicates the weight of individual studies in the pooled analysis. Abbreviations: irVTE, immune-related venous thromboembolism.

Subgroup Analyses

Next, a subgroup analysis was conducted according to the patient and study quality characteristics, and the pooled results are shown in Table 2. Only OS of the studies with multivariable-adjusted HRs were eligible due to low number of studies. Regarding patient characteristics (cancer type, ICI agent type, and length of follow-up), the test for subgroup differences demonstrated that there was no statistically significant subgroup effect (P  =  .91, .68, .41, all I²  =  0%, respectively) (Table 2; Supplemental Figure S1-3). Subgroup analyses showed that no factors among patients’ baseline characteristics would affect the difference in the OS between the irVTE and non-irVTE groups.

Table 2.

Subgroup Analyses of the Association Between Immune-Related Venous Thromboembolic Event Development and Overall Survival.

Subgroup Analyses for OS	N	Pooled HR (95% CI)	Heterogeneity		Reference
			I2	P
Overall	3885	1.58 (1.12-2.22)	81%	<.0001	28,29, 31, 35–37
Patient characteristics
Cancer type
NSCLC	1165	1.52 (0.80-2.90)	85%	.009	29, 31
Mixed tumor types	2720	1.60 (0.93-2.75)	84%	.0004	28, 35–37
Type of ICI agent
Single agent immunotherapy	1298	1.45 (0.89-2.34)	72%	.03	29, 31, 36
≥2 immunotherapy agents	2587	1.72 (0.88-3.38)	89%	.001	28, 35, 37
Length of follow-up (mo)
<12	362	1.27 (0.75-2.14)	0%	.86	35,36
≥12	3523	0.69 (1.10-2.59)	88%	<.0001	28,29, 31, 37
Study quality characteristics
Number of participants
<500	362	1.27 (0.75-2.14)	0%	.86	35,36
≥500	3523	1.69 (1.10-2.59)	88%	<.0001	28,29, 31, 37
Study design
Prospective	572	2.12 (1.49-3.03)	—	—	29
Retrospective	3313	1.47 (1.00-2.17)	80%	.0006	28, 31, 35–37
Model
Multi-state model	672	3.09 (2.07-4.60)	—	—	28
Cox regression model	3213	1.36 (1.06-1.75)	55%	.06	29, 31, 35–37
Accounting for ITB
Yes	1244	2.53 (1.75-3.66)	48%	.17	28,29
No	2641	1.21 (1.06-1.37)	0%	.95	31, 35–37

Abbreviations: ICIs, immune checkpoint inhibitors; VTE, venous thromboembolic events; OS, overall survival; HR, hazards ratio; CI, confidence intervals; N, number of patients included; mo, months; NSCLC, non-small cell lung cancer; ITB, immortal time bias.

Meanwhile, regarding study quality characteristics, we performed subgroup analysis of OS based on the number of participants and whether ITB was accounted for. For the remaining study design and model factors, there were too few studies to conduct meaningful subgroup meta-analyses. Subgroup meta-analyses in the number of participants showed opposite findings (Table 2), implying that patients who developed irVTE appeared to have shorter OS than individuals that did not in the large sample size subgroup (HR 1.69, 95% CI 1.10-2.59, P < .0001). However, stratified analysis by number of participants did not reveal a material difference between the large sample size subgroup (≥500) and the small sample size subgroup (<500) (P  =  .41, I²  =  0%) (Supplemental Figure S4). Moreover, we performed subgroup analysis according to whether ITB was accounted for, in order to investigate whether the occurrence of VTE irAE (time-dependent variable) is associated with clinical outcomes of patients with solid tumors treated with ICIs. The results showed that ICIs-related VTE occurrence was independently associated with shorter OS in the ITB and non-ITB groups. HR and the corresponding 95% CI values in the ITB group were 2.53 (1.75-3.66, P < .00001) with low heterogeneity (P  =  .17, I²  =  48%) and 1.21 (1.06-1.37, P  =  .004) in the non-ITB group with no heterogeneity (P  =  .95, I²  =  0%), respectively (Table 2; Supplemental Figure S5). Otherwise, the test for subgroup differences suggested that there was a considerable difference between subgroups (P  =  .0002, I²  =  92.8%) (Supplemental Figure S5), which suggested that whether ITB was accounted for may be one of the potential sources of heterogeneity.

Sensitivity Analyses and Publication Bias

As a measure of the methodological quality of the studies involved, we utilized the NOS. The specific features extracted from each study are detailed in Table 3. Four studies were rated as high quality and 3 as moderate quality. The majority of studies performed multivariate analyses, but only 33% accounted for ITB.

Table 3.

Quality Assessment of the Studies Included in the Meta-Analysis by the Newcastle–Ottawa Scale (NOS).

Item		Moik et al 2021	Kewan et al 2021	Deschênes-Simard et al 2021	Gutierrez-Sainz et al 2021	Guven et al 2021	Roopkumar et al 2021	Bjørnhart et al 2023
A	Selection
	Representativeness of the exposed cohort	★	★	★	★	★	★	★
	Selection of the non-exposed cohort	★	★	★	★	★	★	★
	Ascertainment of exposure	★	★	★	★	★	★	★
	Demonstration that outcome of interest was not present at start of study	★	★	★	★	★	★	★
B	Comparability
	Comparability of cohorts on the basis of the design or analysis a	—	—	★	★	★	★	★
C	Outcome
	Assessment of outcome	★	★	★	★	★	★	★
	Was follow-up long enough for outcomes to occur b	★	—	—	★	—	★	★
	Adequacy of follow-up of cohorts c	—	★	★	—	—	★	★
	Total	6	6	7	7	6	8	8

^a None of the included studies adequately controlled confounders.

^b If individual studies specified follow-up duration, we assessed long enough for outcomes to occur.

^c If individual studies lost more than 30% of the included patients or did not specify those lost, we assessed as inadequate.

There was no significant effect of leaving 1 study out on the pooled HR for immune-related VTE and ICI efficacy as assessed by leave-one-out sensitivity analyses (pooled HR of OS ranged from 1.29 to 1.58, and no 95% CI contained 1) (Supplemental Figure S6A). Also, sensitivity analyses of both PFS and DCR did not indicate that any individual study altered the results (Supplemental Figure S6B and S6C); that is, no single study affected the pooled HR/OR or 95% CI. In addition, visual inspection of Begg’s funnel plot of OS did not suggest substantial asymmetry, indicating that no publication bias existed (P  =  .072). Neither PFS nor DCR meta-analyses tested funnel plot asymmetry due to the relatively low number of studies included. No statistical significant publication bias for OS, PFS, and DCR was also shown by the Egger regression asymmetry test (P  =  .461, .728, and .580, respectively). Other than that, for small-study effects and publication bias, publication bias analysis of OS by Egger’s test was not observed in the any of the subgroup meta-analyses (P  =  .294 − .895).

Discussion

Currently, the decision regarding whether the development of VTE irAE is associated with the ICI therapeutic efficacy and survival outcomes remains controversial. However, previous systematic reviews and meta-analyses on underlying vascular adverse events for patients treated with ICIs primarily focused on the risk of VTE occurrence rather than the prognostic impact on clinical outcomes.^18–20 To the best of our knowledge, this is the first systematic review and meta-analysis exclusively dedicated to the assessment of the potential correlations between the development of VTE irAE and clinical outcomes. Meanwhile, the results of this review may provide important references for clinical oncologists to prevent thrombotic events during ICI therapy and enhance the clinical outcome of targeted immunotherapy regimens. The results of our meta-analysis indicate that there was a strong correlation between a higher frequency of VTE irAE occurrence and a poorer OS (adjusted HR 1.36; 95% CI 1.06-1.75; P  =  .02), while the development of VTE irAE was not independently associated with shorter PFS (adjusted HR 2.14; 95% CI 0.76-5.98; P  =  .15). In addition, the DCR did not show significant ICI efficacy outcomes between the irVTE and no irVTE groups (OR 0.97, 95% CI 0.68-1.38; P  =  .86).

However, the exact mechanisms of the association between the development of VTE irAE and poor prognosis have not been thoroughly elucidated to date. First, it is well known that hypercoagulability is commonly associated with high aggressiveness, increased metastasis, and tumor progression. ⁴³ The promoting effects of the hemostatic system to tumor progression was dependent not only on conventional prothrombotic functions but also on PAR activation both in tumor and stromal cells.^44,45 Additionally, of interest, recent evidences suggest that VTE has been increasingly reported as an underappreciated adverse event associated with immunotherapy with ICIs, with a cumulative incidence of 5% to 8% at 6 months and over 10% at 12 months.^28,30,37,46 The existing data reveals contradictory findings regarding the potential association between ICIs and an elevated risk of VTE in comparison to conventional chemotherapy.^31,38 The idea that stimulated CD8+ T cells induce monocytes to express tissue factor (TF), thus worsening the thrombosis, has been proposed as a likely mechanism. ⁴⁷ Furthermore, the activation of neutrophil extracellular traps (NETs) by tumor-infiltrating myeloid-derived suppressor cells (MDSCs) can be triggered by cytokines such as IL-8, ultimately promoting procoagulant activity through platelet activation. ³⁷ Hence, a VTE event is an increasingly common and potentially fatal complication in patients with ICIs, making it the second leading cause of death in cancer patients. ⁶ Moreover, in addition to the theory of coagulation in tumor progression, the immune functions of the hemostatic system has been proposed as a possible explanation for the association. Various positive regulators of the blood coagulation pathways, including platelets, the TF pathway, and proteolytic signaling involving protease-activated receptors (PARs), are important regulators of both innate and adaptive immune responses.^48,49

More importantly, recent evidence suggests that extravascular coagulation may contribute to tumor immune evasion and promote resistance to ICI treatment.^21–24 In this regard, pre-clinical data suggest that thrombin promotes tumor immune evasion via proteolysis of platelet-bound glycoprotein A repetitions predominant (GARP) to activate receptor for latent transforming growth factor-β1 (LTGF-β1). ²² Beyond this, recent studies conducted on immune-competent mice have revealed novel mechanisms that loss or pharmacological blockade of the macrophage factor Xa (FXa)-PAR2 signaling pathway could enhance anti-tumor immunity. This is achieved by promoting the accumulation of tumor-killing cytotoxic T cells (CTL) in the tumor microenvironment (TME), while simultaneously preventing the recruitment of regulatory T cells (Treg) and macrophage polarization toward an immunosuppressive phenotype. ²⁴ This hypothesis was further supported by the clinical finding that concomitant treatment with anticoagulants and immunotherapy resulted in the synergistic inhibition of tumor growth and significantly increased OS. More specifically, recent clinical observations suggest that anticoagulant therapy with FXa inhibitor (FXa-i) rivaroxaban may have a dual benefit in cancer patients by preventing VTE and enhancing the efficacy of ICI therapy through the restoration of anti-tumor immunity.^21,23 Thus, the synergistic effect of AC in combination with ICIs is already possible with currently available data, indicating that the targeting relationship of rivaroxaban and FXa-PAR2 signaling pathway may partially account for the relationship between irVTE and prognosis.

Notably, results to date from studies investigating VTE irAE and clinical outcomes produced mixed results pointing toward potential heterogeneity with regard to study design, quality, and accounting for ITB. Further stratification analysis revealed no statistically significant differences in OS between subgroups with tumor type (NSCLC vs mixed tumor types) and ICI agent type (monotherapy vs combination-agent immunotherapies). The results concerning OS from stratification factor subgroup analyses were consistent with those of the primary analysis,^30,31,35,36 suggesting that the effects of VTE irAE on clinical outcomes may not be potentially dependent on the primary tumor and ICI agent types. However, 3 studies evaluated populations containing only NSCLC (n  =  1382).^29,31,39 Of these, 2 (66.7%) studies showed that the increased incidence of VTE irAE correlated with a poorer OS, serving as a prognostic indicator,^29,39 whereas both studies reported that ICIs-related VTE could not serve as unfavorable prognostic factors for PFS.^29,31 Moreover, findings from only 1 study which evaluated participants who had melanoma diagnoses only indicated that VTE/AE irAEs were also closely correlated with poor OS. ²⁷ Of course, it is not excluded that the unbalanced and inadequate number of studies in some subgroups may also have contributed to the identified heterogeneity presented here. Therefore, more research and evidence based on a specific cancer type will be needed in the future to further explore the association between the development of VTE irAE and survival outcomes.

Of note, it is important to underline that CAT is a time-dependent event. When CAT is assessed as a constant variable, a bias referred to as “ITB” arises, which significantly impacts studies investigating the correlation between a time-varying variable and clinical outcomes. ⁵⁰ Typically, this bias is commonly associated with a limited follow-up period in a specific cohort where the outcome of interest cannot manifest. Thus, there was a potential confounding effect of ITB on survival endpoints that cannot be discounted. However, it is important to note that only a limited number of studies examined in this review have focused on the correlation between irVTE and clinical outcomes in patients undergoing ICI treatment, with particular attention to ITB in evaluating the significance of this relationship.^28,29,39,42 Based on our analysis, the results of ITB subgroup analysis, especially, showed that ICIs-related VTE occurrence was independently associated with shorter OS in the non-ITB and ITB groups, with HR of 2.53 and 1.21, respectively. Otherwise, the test for subgroup differences suggested that there was a considerable difference between subgroups (P  =  .0002, I²  =  92.8%), suggesting that the effects of VTE irAE on clinical outcomes may be potentially dependent on whether ITB was accounted for. Consistent with our findings, more recent studies by Carmona-Bayonas et al ⁵¹ have indicated that the importance of accounting for time-varying effects should not be overlooked in the analysis of cancer-associated VTE. As such, the implementation of flexible analyses utilizing multi-state models represents a promising approach for the evaluation of cancer-associated VTE. Taken together, all of the above indicated that the prevention and the identification of ITB are essential for obtaining unbiased estimates of effect in the current era of precision medicine in oncology. Future observational studies should take the methodological issue into account, such as time-dependent exposure as a time-dependent variable, ⁵² or extended Cox model with time-varying covariates, ⁵³ or using Landmark analysis. ⁵⁴

Based on compelling evidence derived from systematic review, cohort studies, and case–control studies, it has been postulated that certain patients experiencing irAEs may exhibit an enhanced immune response and improved response to ICIs, particularly in relation to skin or endocrine irAEs.^55,56 However, previous investigations have not specifically explored the association between the occurrence of immune-related VTE and the clinical efficacy in patients receiving ICI therapy.^18–20 Therefore, in the present systematic review and meta-analysis, an accurate magnitude of the predictive effect of irVTE on ICI efficacy was obtained for the first time. The available evidence indicates that the emergence of VTE as an irAE is linked to the anti-tumor effects of ICIs, thereby rendering VTE a promising biomarker for predicting long-term oncologic outcomes in patients undergoing ICI therapy. However, the current data on immune-related VTE do not support the routine use of anticoagulants in patients receiving immunotherapy in the absence of established indications for concomitant AC.^46,57,58 Further large-scale prospective studies are necessary to elucidate the potential of coagulation cascade inhibition in augmenting ICI response and its subsequent implications.

Some limitations of the current study must be considered. Firstly, the meta-analysis included studies with different designs (eg, prospective and retrospective). The preponderance of data collected was from retrospective studies, which may engender sundry biases such as information, selection, and potential outcome measurement biases. Secondly, the included studies varied in terms of histological and ICI agent types, tumor staging, follow-up duration, study population, design, and model. The existing of confounding factors among the included studies varied considerably, which contributed to the significant heterogeneity (I²>50%). Although not unexpected, the presence of significant heterogeneity in the combined results may compromise their reliability, as VTE/AE irAEs were not the primary outcomes of interest in several of the studies included. Additionally, selective analysis reporting resulted in the exclusion of positive studies from our meta-analysis due to insufficient data (ie, HR and CI). However, due to the small number of included clinical trials with diverse outcome measurements, we could not draw definitive results that are not significant for PFS and DCR but significant for OS. In general, these biases may potentially lead to an underestimation of the association between VTE/AE irAEs and clinical outcomes. Therefore, additional prospective clinical investigations that document the favorable and unfavorable correlations between VTE/AE irAE and ICI therapy outcomes are imperative to authenticate our results.

Conclusions

Given the increasing and expanding use of ICIs in clinical practice, it is imperative that clinicians should be aware of the potential associated immune-related VTE events. Although the current reporting deficiencies and biases remain within the scientific literature, evidence does exist supporting that VTE irAE occurrence could serve as a prognostic indicator, with higher frequencies of occurrence associated with poorer OS, while the development of VTE irAEs was not independently associated with shorter PFS. Nevertheless, additional prospective clinical investigations are necessary to substantiate the association between VTE irAEs and clinical consequences, particularly in diverse primary cancer locations and types of ICIs, as well as the prevention and the identification of ITB.