Research article 15 October 2026

Early-treatment coagulation activation and radiographic progression during VEGF/VEGFR inhibitor therapy: A multicancer real-world cohort study

Authors

CC BY 4.0 Published: 15 October 2026 DOI: 10.17305/bb.2026.15049

Abstract

Coagulation activation is common in cancer, but its link to treatment durability is unclear. We investigated whether early-treatment coagulation activation correlated with radiographic progression during vascular endothelial growth factor (VEGF)/vascular endothelial growth factor receptor (VEGFR) inhibitor therapy. This single-center, retrospective, multicancer cohort study developed a coagulation activation score (0–2) based on the last available D-dimer (>500 ng/mL) and fibrinogen (>4.0 g/L) measurements taken within 100 days after baseline. Multivariable Cox models calculated hazard ratios (HRs) and 95% confidence intervals (CIs) for subsequent radiographic time to progression among patients who were progression-free at the 100-day landmark. Of 1400 eligible patients, 1089 had evaluable early-treatment measurements, and 909 were included in the landmark analysis. Compared to a score of 0, adjusted HRs were 1.29 (95% CI 1.02–1.62) for score 1 and 1.51 (95% CI 1.07–2.12) for score 2. Score 2 also correlated with lower odds of durable clinical benefit (adjusted odds ratio 0.56, 95% CI 0.38–0.84), while objective response and best tumor shrinkage did not show a consistent gradient. These associations remained significant in time-dependent and timing-restricted analyses. Exploratory subgroup analyses revealed the strongest association in colorectal cancer patients treated with bevacizumab (score 2 vs. 0: HR 2.23, 95% CI 1.21–4.11); however, no association was observed in lung cancer patients treated with bevacizumab. No significant interaction with documented immune checkpoint inhibitor exposure was detected. The score did not enhance discrimination beyond clinical variables and baseline coagulation markers. Early-treatment coagulation activation was associated with reduced treatment durability during VEGF/VEGFR inhibition. Prospective validation in specific cancer–drug contexts is necessary.

Keywords

D-dimer fibrinogen coagulation anti-VEGF VEGFR treatment durability radiographic progression

Full text

Introduction

The VEGF/VEGFR axis regulates tumour angiogenesis, vascular permeability, and the immunosuppressive tumour microenvironment [1,2]. Inhibiting this axis can improve antitumour immunity and has produced clinical benefit when combined with immune checkpoint blockade or cytotoxic therapy in selected cancers [2-5]. Only some patients derive durable benefit. Biomarkers that reliably identify early loss of benefit are still lacking, and many candidate markers perform differently across clinical contexts.

Systemic coagulation activation is common in cancer and is associated with progression, thrombosis, and poor survival [6,7]. D-dimer and fibrinogen are measured routinely in clinical practice and have repeatedly been associated with inferior outcomes in patients receiving immune checkpoint inhibitors [8-14]. These observations suggest that coagulation activation may be more than a surrogate for thrombotic risk. Preclinical studies identify several plausible mechanisms. Thrombin cleaves platelet-bound GARP to release active TGF-beta [15], and thrombin-PAR1 signalling promotes immune evasion [16]. Factor Xa-PAR2 signalling alters myeloid-cell antigen handling and cGAS-STING-IFN induction [17]. Factor X promotes treatment resistance [18], while plasminogen activator inhibitors establish a fibrin-rich, T-cell-excluding microenvironment [19]. Endothelial programmes that promote angiogenesis can also suppress antitumour immunity [1].

Most clinical studies have measured D-dimer or fibrinogen before treatment. A baseline sample can identify patients with an intrinsically poor prognosis, but it cannot show whether coagulation activation persists during therapy or tracks the durability of benefit. On-treatment monitoring may be more informative because treatment can alter VEGF signalling, vascular function, platelet activity, and inflammatory pathways. The available on-treatment studies have produced mixed results, and some found limited value for post-treatment D-dimer measurements [13]. It remains unknown whether early-treatment coagulation activation identifies patients whose tumours initially respond but later progress.

This gap is especially relevant to VEGF/VEGFR-axis inhibitor therapy. Anti-angiogenic treatment can reshape the tumour immune microenvironment [1,2], but reliable biomarkers remain elusive [2]. A dynamic coagulation marker could provide a practical readout of a treatment-resistant microenvironment if it distinguishes durable benefit from initial radiographic shrinkage. We hypothesized that early-treatment coagulation activation would be associated with loss of durable benefit after an initial radiographic response. We evaluated pre-treatment and early-treatment D-dimer and fibrinogen in a multicancer real-world cohort receiving VEGF/VEGFR-axis inhibitor therapy.

Materials and methods

Study design and setting

This retrospective, single-centre, real-world cohort study was conducted at a tertiary cancer centre. Consecutive patients treated with VEGF/VEGFR-axis inhibitor-based therapy between 2 January 2021 and 29 May 2026 were screened. Day 0 was defined as the pre-treatment baseline immediately preceding initiation of VEGF/VEGFR-axis inhibitor therapy. The study followed the STROBE recommendations for observational research [20]. The analysis used de-identified data generated during routine clinical care. The study was approved by the Ethics Committee of the 983rd Hospital of the Joint Logistic Support Force (2025-YK-20-01). The requirement for informed consent was waived because the analysis used retrospective, de-identified data.

Participants

Eligible patients were adults with solid tumours who received bevacizumab or a VEGFR tyrosine kinase inhibitor as part of routine treatment. Patients were required to have a pre-treatment baseline assessment, at least one measurable target lesion, and evaluable follow-up imaging. Patients were excluded if key baseline characteristics, coagulation data, cancer type, drug exposure, or follow-up imaging were unavailable. Of 3745 baseline patients, 1400 met the predefined quality criteria. A total of 1089 had both D-dimer and fibrinogen measured within 100 days after baseline, and 909 remained progression-free at the 100-day landmark and entered the primary analysis (Figure 1).

Cohort selection and timing of early-treatment coagulation assessment.
Figure 1 Cohort selection and timing of early-treatment coagulation assessment. (A) Study flow and treatment timeline. Of 3745 baseline patients, 1400 met predefined quality criteria, 1089 had evaluable early-treatment D-dimer and fibrinogen measurements, and 909 remained progression-free at the 100-day landmark and entered the primary analysis. The early-treatment coagulation activation score (0–2) assigned one point each for D-dimer >500 ng/mL and fibrinogen >4.0 g/L, using the last available measurement of each marker within 100 days after the pre-treatment baseline (day 0). Radiographic TTP was assessed from the 100-day landmark. (B) Overlaid histograms showing the timing of the D-dimer and fibrinogen measurements used to calculate the early-treatment score. The shaded region, bounded by dashed lines, indicates the 61–100-day measurement window used in the timing-restricted sensitivity analysis. Abbreviation: TTP, time to progression.

Treatment exposure classification

Anti-VEGF/VEGFR treatment was identified from the curated treatment fields. Documented ICI exposure was identified from medication names and treatment descriptions in baseline admission and last-hospitalization records. The following ICI agents were searched: pembrolizumab, nivolumab, sintilimab, camrelizumab, tislelizumab, toripalimab, serplulimab, adebrelimab, durvalumab, atezolizumab, sugemalimab, penpulimab, zimberelimab, cadonilimab, envafolimab, ipilimumab, and tremelimumab. Chinese generic names and commonly used brand names were also searched. Patients were classified as having anti-VEGF/VEGFR plus ICI when an ICI drug was documented in the treatment record. A stricter definition was used in a sensitivity analysis and required the anti-VEGF/VEGFR drug and ICI to appear in the same treatment sentence. Patients without a documented ICI were classified as anti-VEGF/VEGFR monotherapy. Because these categories were derived from clinical text rather than a complete pharmacy database, they were considered proxy treatment exposures.

Coagulation measurements

Pre-treatment baseline D-dimer and fibrinogen values were obtained from the closest routine measurement before treatment. Early-treatment coagulation activation was assessed using the last available D-dimer and fibrinogen measurements within 100 days after day 0. The two measurements were not required to be obtained on the same date. The score used the last available result for each marker, and the time-dependent model began at the later of the two measurement dates. Continuous values were analysed in their original units, and predefined dichotomous thresholds were D-dimer >500 ng/mL and fibrinogen >4.0 g/L. The early-treatment coagulation activation score was the sum of the two dichotomous variables, yielding scores of 0, 1, or 2. A 61-100-day measurement window was used in a sensitivity analysis to test whether the association depended on measurement timing. Coagulation trajectory was defined by combining baseline and early-treatment scores. Stable low activation was a score of 0 at both time points. Normalized activation was a baseline score >0 and an early score of 0. Newly activated coagulation was a baseline score of 0 and an early score >0. Persistent activation was scores >0 at both time points.

Outcomes

The primary endpoint was subsequent radiographic time to progression (TTP) after a 100-day landmark. Patients who progressed before or at day 100 were excluded from the landmark analysis. Progression was identified from longitudinal radiological assessments using RECIST-style target-lesion measurements and radiology reports [21,22]. A target-lesion increase of at least 20% was considered progression. New lesions were considered progression when reported. Non-target lesions were considered progression only when the radiology report described unequivocal progression. Death without documented radiographic progression was censored at the last evaluable imaging assessment. We consistently used TTP rather than progression-free survival because death data were incomplete. Secondary endpoints were DCB, objective response rate (ORR), deep tumour shrinkage, and early progression. DCB was defined as complete or partial response, or stable disease lasting at least 6 months without progression. ORR was defined as complete or partial response. Deep tumour shrinkage was defined as at least 50% reduction from baseline in the sum of target-lesion diameters. Best tumour shrinkage was analysed as a continuous variable.

Ethical statement

This research was approved by the Ethics Committee of the 983rd Hospital of the Joint Logistic Support Force(2025-YK-20-01). The requirement for informed consent was waived because the study used retrospective, de-identified data.

Statistical analysis

Baseline characteristics were summarized using medians with interquartile ranges for continuous variables and counts with percentages for categorical variables. Kaplan-Meier curves and log-rank tests were used to compare radiographic TTP. Cox proportional hazards models estimated hazard ratios with 95% confidence intervals. The fully adjusted model included age, sex, baseline tumour burden, baseline D-dimer, baseline fibrinogen, cancer type, and anti-VEGF/VEGFR drug class. A separate model additionally adjusted for the timing of the early-treatment coagulation measurement. The proportional hazards assumption was assessed with Schoenfeld residual tests; no evidence of violation was found for the score terms (score 1, P=0.286; score 2, P=0.768) or covariates (all P>0.05). A time-dependent Cox model treated the early-treatment score as an exposure beginning at the later measurement date. Logistic regression estimated the association between early coagulation activation and DCB. Prespecified exploratory subgroups included lung cancer and colorectal cancer, with bevacizumab and VEGFR tyrosine kinase inhibitors analysed separately when sample size permitted. Treatment-context analyses compared anti-VEGF/VEGFR therapy without documented ICI with ICI-containing therapy. Interaction terms between treatment context and the coagulation score were tested in Cox models. Analyses were repeated within bevacizumab and anlotinib strata. A high-confidence cohort required an early-treatment measurement between 61 and 100 days, freedom from progression at the 100-day landmark, and imaging follow-up of at least 6 months unless progression occurred earlier. Because subgroup comparisons were exploratory, no multiplicity correction was applied. Internal validation followed TRIPOD recommendations [23]. Tumour marker reporting followed REMARK recommendations [24].

Results

Cohort characteristics

The analysis included 1400 patients who met the full quality criteria (Table 1). Median age was 61.5 years (IQR 54.6-68.2), and 886 patients (63.3%) were male. Among these patients, 909 entered the 100-day landmark cohort and 491 did not. Compared with excluded patients, landmark-eligible patients were older (median 62.2 vs 60.0 years, P=0.002), more often male (67.0% vs 56.4%, P<0.001), and had greater baseline tumour burden (median 5.9 vs 4.3 cm, P<0.001). Baseline D-dimer and fibrinogen distributions were similar (median 243.2 vs 259.6 ng/mL, P=0.811; 3.74 vs 3.76 g/L, P=0.948). Excluded patients had more progression events and shorter median observed TTP (77.8% and 125 days vs 41.3% and 235 days), consistent with exclusion of early progressors and patients without evaluable early coagulation data. These comparisons are reported in Supplementary Table S1.

Table 1 Baseline and early-treatment characteristics of the 909-patient 100-day landmark cohort, overall and by early-treatment coagulation score
Characteristic Score 0 Score 1 Score 2
Patients, n 427 345 137
Median age, years 61.3 63.6 64.8
Male, n (%) 280 (65.6) 227 (65.8) 102 (74.5)
Median baseline tumour burden, cm 5.4 6.1 7.2
Median baseline D-dimer, ng/mL 189.96 276.48 488.04
Median baseline fibrinogen, g/L 3.54 3.92 4.04
Median early D-dimer, ng/mL 188.03 395.67 953.56
Median early fibrinogen, g/L 3.33 4.12 4.84
Lung cancer, n 212 153 67
Colorectal cancer, n 125 109 32
Bevacizumab, n 270 204 88
Anlotinib, n 133 123 46
Subsequent progression events, n 191 139 45

Early-treatment coagulation activation and subsequent progression

The early-treatment score was 0 in 427 patients, 1 in 345 patients, and 2 in 137 patients. Among patients with documented progression, median time from the landmark to progression was 230, 155, and 130 days for scores 0, 1, and 2, respectively. Numbers of progression events were 191, 139, and 45. Kaplan-Meier curves are shown in Figure 2a. In the unadjusted model, score 1 was associated with an HR of 1.26 (95% CI 1.01-1.57), and score 2 was associated with an HR of 1.27 (95% CI 0.92-1.76) compared with score 0. After full adjustment, the HR was 1.29 (95% CI 1.02-1.62) for score 1 and 1.51 (95% CI 1.07-2.12) for score 2. After additional adjustment for measurement timing, the HRs were 1.33 (95% CI 1.06-1.69) and 1.57 (95% CI 1.12-2.22), respectively. Continuous analysis showed that early-treatment fibrinogen was associated with progression after full adjustment (HR 1.13 per SD, 95% CI 1.02-1.26). Early-treatment D-dimer was not independently associated after adjustment for baseline D-dimer (HR 1.00 per SD, 95% CI 0.88-1.13).

Early-treatment coagulation activation and subsequent radiographic outcomes after the 100-day landmark.
Figure 2 Early-treatment coagulation activation and subsequent radiographic outcomes after the 100-day landmark. (A) Kaplan–Meier curves for radiographic time to progression according to early-treatment coagulation score (0, 1, or 2). Shaded bands represent 95% CIs. (B) Forest plot of adjusted hazard ratios for subsequent radiographic time to progression comparing score 1 versus 0 and score 2 versus 0 in the primary fully adjusted model and in the timing-adjusted sensitivity model. Horizontal lines indicate 95% CIs. (C) Proportions of patients with durable clinical benefit, objective response, and deep response according to early-treatment coagulation score. (D) Distribution of best tumour shrinkage by early-treatment coagulation score. Individual points represent patients, vertical bars indicate interquartile ranges, and circles indicate medians; the dashed horizontal line marks no change from baseline. Abbreviations: CI, confidence interval; DCB, durable clinical benefit; HR, hazard ratio; ORR, objective response rate; TTP, time to progression.

Coagulation trajectory

Among patients who progressed, median TTP after the landmark was 7.7 months for stable low activation, 5.8 months for normalized activation, 6.2 months for newly activated coagulation, and 3.8 months for persistent activation (Figure 3a-c).

Coagulation trajectories and sensitivity analyses of radiographic progression.
Figure 3 Coagulation trajectories and sensitivity analyses of radiographic progression. (A) Kaplan–Meier curves for radiographic TTP from the 100-day landmark according to coagulation trajectories between baseline and early treatment: stable low (score 0 at both assessments), normalized (baseline score >0 and early-treatment score 0), newly activated (baseline score 0 and early-treatment score >0), and persistent activation (score >0 at both assessments). Shaded bands indicate 95% CIs. (B) Fully adjusted HRs from Cox models for persistent versus stable-low activation, each one-point increase in the time-varying early-treatment score, and scores 1 and 2 versus 0 in the sensitivity analysis restricted to measurements obtained 61–100 days after baseline. Points indicate HRs, horizontal lines indicate 95% CIs, and the dashed vertical line marks HR = 1. (C) Median observed TTP from the 100-day landmark by coagulation trajectory. These medians were calculated only among patients with documented progression and are not Kaplan–Meier estimates for the full trajectory groups. Abbreviations: CI, confidence interval; HR, hazard ratio; TTP, time to progression.

Time-dependent and timing-restricted analyses

The early-treatment score was modelled as a time-varying exposure beginning at its measurement. Each one-point increase was associated with an HR of 1.23 (95% CI 1.07-1.41) after full adjustment. In the 61-100-day measurement window, 587 patients and 244 progression events were available. Fully adjusted hazard ratios were 1.57 (95% CI 1.17-2.11) for score 1 and 1.85 (95% CI 1.19-2.87) for score 2 (Table 2).

Table 2 Main models for subsequent radiographic progression after the 100-day landmark
Exposure Model HR (95% CI) p value
Early score 1 vs 0 Fully adjusted 1.29 (1.02-1.62) 0.030
Early score 2 vs 0 Fully adjusted 1.51 (1.07-2.12) 0.018
Early score 1 vs 0 Adjusted for measurement timing 1.33 (1.06-1.69) 0.016
Early score 2 vs 0 Adjusted for measurement timing 1.57 (1.12-2.22) 0.010
Persistent activation vs stable low Fully adjusted trajectory model 1.47 (1.05-2.05) 0.025
Time-varying early score, per 1 point Fully adjusted time-dependent model 1.23 (1.07-1.41) 0.003

Treatment durability and tumour shrinkage

Early-treatment coagulation activation was associated with lower DCB. Compared with score 0, score 2 was associated with an adjusted odds ratio of 0.56 (95% CI 0.38-0.84) for DCB. Early D-dimer elevation was associated with an adjusted OR of 0.67 (95% CI 0.48-0.93), and early fibrinogen elevation was associated with an adjusted OR of 0.67 (95% CI 0.49-0.92). Initial radiographic response did not show a consistent gradient. ORR was 41.2% for score 0, 46.1% for score 1, and 38.0% for score 2. Deep tumour shrinkage was observed in 28.3%, 31.0%, and 28.5% of patients, respectively. Median best tumour shrinkage was -20.0%, -20.6%, and -17.3% (Figure 2c,d).

Cancer type and drug heterogeneity

The strongest association was in colorectal cancer treated with bevacizumab. In this subgroup of 234 patients with 101 progression events, score 2 versus score 0 was associated with an HR of 2.23 (95% CI 1.21-4.11). Score 1 versus score 0 showed a weaker and non-significant association (HR 1.26, 95% CI 0.82-1.95) (Figure 4a). In the overall colorectal cancer subgroup, score 2 was associated with an HR of 2.12 (95% CI 1.17-3.85). The association did not replicate in lung cancer treated with bevacizumab. Among 269 patients with 114 events, score 2 versus score 0 had an HR of 0.98 (95% CI 0.53-1.83). Lung cancer treated with anlotinib showed a non-significant association (HR 0.94, 95% CI 0.37-2.44). DCB rates across score groups in selected strata are shown in Figure 4a,b. The oesophageal and gynaecological subgroups were limited by sample size and were considered exploratory (Figure 4).

Cancer- and drug-specific associations between early-treatment coagulation activation and radiographic outcomes.
Figure 4 Cancer- and drug-specific associations between early-treatment coagulation activation and radiographic outcomes. (A) Forest plot of unadjusted HRs for radiographic TTP after the 100-day landmark, comparing early-treatment coagulation scores 1 and 2 with score 0 across exploratory cancer–drug strata. Blue circles and red squares represent comparisons of score 1 and score 2 with score 0, respectively; horizontal lines indicate 95% CIs. The dashed vertical line marks HR = 1. “All” denotes all included drug groups within each cancer type. (B) DCB rates by early-treatment coagulation score (0, 1, or 2) in CRC treated with bev and lung cancer treated with bev or anlo. Values above the bars indicate the percentage of patients achieving DCB. Abbreviations: anlo, anlotinib; bev, bevacizumab; CI, confidence interval; CRC, colorectal cancer; DCB, durable clinical benefit; HR, hazard ratio; TTP, time to progression.

Treatment context and coagulation-associated risk

Documented ICI exposure was present in 546 of 1400 patients. Another 166 patients had an ICI name but no clear treatment-action phrase and were not classified as exposed. The stricter same-sentence definition identified 442 patients with concurrent anti-VEGF/VEGFR and ICI treatment. The comparator comprised 854 patients without documented ICI. Treatment-context models used the 100-day landmark cohort, not the full treatment cohort. The model for anti-VEGF/VEGFR therapy without documented ICI included 556 patients and 230 events; the ICI-containing model included 353 patients and 145 events; the same-sentence concurrent model included 295 patients and 123 events. The adjusted HR per one-point increase was 1.26 (95% CI 1.03-1.54) in anti-VEGF/VEGFR therapy without documented ICI and 1.21 (95% CI 0.94-1.56) in ICI-containing therapy. The treatment-context interaction was not significant (P=0.596). With the same-sentence definition, the HRs were 1.26 (95% CI 1.03-1.54) and 1.16 (95% CI 0.88-1.54), with an interaction P of 0.361 (Figure 5a; Table 3). In the high-confidence cohort, the association remained significant in anti-VEGF/VEGFR therapy without documented ICI (HR 1.53, 95% CI 1.18-1.99, P=0.001) and in ICI-containing therapy (HR 1.42, 95% CI 1.03-1.96, P=0.034). The association was borderline in the same-sentence concurrent group (HR 1.37, 95% CI 0.97-1.95, P=0.075). Within drug strata, model sizes were 441 patients and 195 events for bevacizumab without documented ICI, 121 patients and 41 events for bevacizumab plus ICI, 105 patients and 28 events for anlotinib without documented ICI, and 197 patients and 85 events for anlotinib plus ICI. Treatment-by-score interaction tests within drug strata were not significant (bevacizumab, P=0.200; anlotinib, P=0.923) (Figure 5b). The exploratory ICI-only text cohort was not used for the primary analysis. Median progression times among progressors by score and treatment context are shown in Figure 5c.

Treatment-context analyses of early-treatment coagulation activation and subsequent radiographic progression.
Figure 5 Treatment-context analyses of early-treatment coagulation activation and subsequent radiographic progression. (A) Adjusted HRs per one-point increase in the early-treatment coagulation score for radiographic TTP after the 100-day landmark across treatment contexts and high-confidence subgroups. “Monotherapy” denotes anti-VEGF/VEGFR therapy without documented ICI exposure, rather than confirmed single-agent treatment. Concurrent therapy required both drugs to be documented in the same treatment sentence. High-confidence analyses required coagulation assessment at 61–100 days, freedom from progression at day 100, and ≥6 months of imaging follow-up unless progression occurred earlier. (B) Corresponding adjusted HRs within bevacizumab and anlotinib treatment contexts. In both forest plots, blue circles represent therapy without documented ICI and red squares represent ICI-containing therapy; horizontal lines indicate 95% CIs, and dashed vertical lines mark HR = 1. Treatment-by-score interaction tests were not significant overall (p = 0.596) or within the bevacizumab (p = 0.200) and anlotinib (p = 0.923) strata. (C) Median observed radiographic TTP, in days, from the 100-day landmark by coagulation score and treatment context among patients with documented progression. Labels indicate medians and numbers of progressors (n); these medians are not Kaplan–Meier estimates for the full treatment groups. Abbreviations: CI, confidence interval; HR, hazard ratio; ICI, immune checkpoint inhibitor; TTP, time to progression; VEGF, vascular endothelial growth factor; VEGFR, vascular endothelial growth factor receptor.
Table 3 Treatment-context analyses in the 100-day landmark cohort. Model sample sizes and progression event counts are reported separately for each treatment context
Treatment cohort Full treatment cohort, n Landmark model, n Progression events, n HR per 1-point score increase (95% CI) Interaction p
Anti-VEGF/VEGFR therapy without documented ICI 854 556 230 1.26 (1.03-1.54) Reference
Any documented ICI-containing therapy 546 353 145 1.21 (0.94-1.56) 0.596
Concurrent anti-VEGF/VEGFR plus ICI 442 295 123 1.16 (0.88-1.54) 0.361
Bevacizumab without documented ICI 685 441 195 1.30 (1.05-1.62) Reference
Bevacizumab plus ICI 202 121 41 1.10 (0.67-1.80) 0.200
Anlotinib without documented ICI 147 105 28 1.26 (0.71-2.25) Reference
Anlotinib plus ICI 284 197 85 1.40 (1.01-1.94) 0.923
High-confidence therapy without documented ICI Not applicable 219 141 1.53 (1.18-1.99) Reference
High-confidence ICI-containing therapy Not applicable 160 103 1.42 (1.03-1.96) Not estimated

Abbreviations: CI, confidence interval; HR, hazard ratio; ICI, immune checkpoint inhibitor; VEGF, vascular endothelial growth factor; VEGFR, vascular endothelial growth factor receptor.

Internal validation

Five-fold cross-validation showed that clinical variables alone yielded a C-index of 0.590. Adding baseline coagulation increased the C-index to 0.597, and adding the early-treatment score produced a C-index of 0.596. The trajectory model yielded a C-index of 0.600. The addition of the early score did not improve discrimination beyond the clinical plus baseline coagulation model (delta C-index -0.001, 95% bootstrap interval -0.011 to 0.008). Dynamic coagulation markers therefore provided prognostic information but did not form a stand-alone prediction model (Figure S1).

Discussion

Early-treatment coagulation activation was associated with shorter radiographic TTP and lower DCB after VEGF/VEGFR-axis inhibitor therapy. The association was strongest for persistent activation and was similar in therapy without documented ICI and in ICI-containing therapy. ICI co-treatment did not significantly modify the association in interaction tests. Objective response and best tumour shrinkage did not follow the same pattern as subsequent progression, and these analyses cannot establish that coagulation activation causes resistance. The results support monitoring coagulation during treatment rather than relying on a pre-treatment sample alone.

Treatment context limits how the marker should be interpreted. Early-treatment coagulation activation remained associated with progression in both anti-VEGF/VEGFR therapy without documented ICI and ICI-containing therapy, although the point estimate was slightly stronger in the former. In the high-confidence cohort, the association was significant in both groups (therapy without documented ICI, HR 1.53; ICI-containing therapy, HR 1.42). ICI co-treatment therefore neither clearly increased nor removed the coagulation-associated risk. D-dimer and fibrinogen are unlikely to be treatment-specific pharmacodynamic markers of immune checkpoint inhibition. A broader interpretation is that coagulation activation may mark a resistance phenotype that occurs across different VEGF/VEGFR-directed regimens.

Drug-specific results differed by treatment context. The association was present in bevacizumab therapy without documented ICI but not in bevacizumab plus ICI, and in anlotinib plus ICI but not anlotinib therapy without documented ICI. Formal interaction tests within drug strata were not significant, so these differences remain hypothesis-generating. Possible explanations include differences in vascular normalization, platelet activation, concurrent chemotherapy, cancer mix, and treatment line. Bevacizumab targets circulating VEGF-A, whereas VEGFR tyrosine kinase inhibitors act on endothelial and tumour-cell signalling; their effects on the tumour microenvironment and coagulation system are unlikely to be identical [1-5]. The marker should therefore be assessed within defined drug and cancer contexts rather than as a universal biomarker.

The distinction between baseline and early-treatment coagulation activation is clinically important. Previous studies have linked high pre-treatment D-dimer or fibrinogen to inferior outcomes [8-14]. A prospective multicenter trial in metastatic renal cell carcinoma treated with nivolumab also evaluated fibrinogen and supports the relevance of coagulation markers [25]. These studies established the prognostic value of coagulation markers but could not distinguish patients whose activation resolves during treatment from those whose activation persists. In this cohort, baseline activation alone was not consistently associated with progression after adjustment. Persistent activation carried the strongest signal, and baseline-activated patients whose coagulation normalized did not show a persistent excess risk. This pattern defines a plausible monitoring interval, but it does not show that coagulation activation is directly modifiable or causal.

Several mechanisms connect coagulation proteases to immune evasion. Thrombin releases active TGF-beta from platelet-bound GARP and promotes an immunosuppressive microenvironment [15]. Thrombin-PAR1 signalling suppresses CD8 T-cell-mediated tumour control [16]. Factor Xa-PAR2 signalling impairs myeloid-cell antigen uptake and reduces cGAS-STING-IFN activation, whereas FXa inhibition can synergize with immune checkpoint blockade [17]. Factor X promotes treatment resistance [18], and plasminogen activator inhibitors establish a fibrin-rich microenvironment that retains suppressive macrophages and excludes cytotoxic T cells [19]. These pathways provide a plausible way for persistent coagulation activation to coexist with initial radiographic control while shortening the duration of benefit [26-28].

Initial shrinkage did not predict durable benefit. In this cohort, ORR and best tumour shrinkage were similar across early coagulation score groups, whereas DCB and subsequent progression differed. A first-scan response can therefore be present even when benefit will be short-lived. Early identification of non-durable benefit is an important goal in immunotherapy research, and multimodal approaches that combine ctDNA, tumour burden, and early radiographic change can predict durable benefit [29,30]. A coagulation-based marker would be less specific and less mechanistically direct than circulating tumour DNA. Its low cost and routine availability could nevertheless make it useful for initial stratification or as an adjunct when molecular monitoring is unavailable.

We treated the ICI-only analysis as exploratory. Text records identified 307 patients without documented anti-VEGF/VEGFR therapy, and 273 had imaging and baseline coagulation data. In this group, the early-treatment coagulation score showed a strong association with progression. This cohort was defined by the absence of treatment text rather than by confirmed pharmacy records. Because the source cohort was assembled around VEGF/VEGFR treatment, unrecognized anti-VEGF/VEGFR exposure is plausible. This result remains hypothesis-generating and does not define the primary clinical message.

Alternative explanations are possible. D-dimer and fibrinogen are nonspecific acute-phase and coagulation markers. Elevated values may reflect venous thromboembolism, infection, inflammation, liver dysfunction, recent surgery, or anticoagulant therapy. Anticoagulation, thrombotic events, infection, and C-reactive protein were not systematically captured. The association may therefore partly reflect systemic illness or undetected thrombosis rather than a tumour-specific immune-resistant state. Adjustment for baseline tumour burden, baseline coagulation, cancer type, and drug class reduced but did not eliminate residual confounding. A prospective study should measure anticoagulation exposure, thrombotic events, inflammatory markers, platelet counts, and liver function.

The primary analysis only included patients who remained progression-free at the 100-day landmark, thereby excluding patients with early treatment failure and potentially more aggressive disease. This selection may limit generalizability to the overall treated population. Treatment exposure was extracted from clinical text and could not fully distinguish concurrent from sequential ICI use. The stricter same-sentence definition reduced misclassification but also reduced sample size. Imaging endpoints were derived from routine reports and longitudinal target-lesion measurements rather than independent blinded central review. The cohort was from one centre with unequal cancer and drug distributions, and treatment lines and concurrent chemotherapy were heterogeneous. Treatment line, concurrent chemotherapy, anticoagulation, thrombosis, infection, inflammatory markers, platelet counts, and liver function were not systematically captured. Internal validation showed only limited incremental discrimination; the marker is not a stand-alone prediction model or a clinically validated tool.

The early-treatment coagulation score should be tested prospectively with protocol-defined D-dimer and fibrinogen measurements and manually verified treatment dates. Prospective studies must record anticoagulation status, venous thromboembolism, infection, inflammatory markers, liver function, and platelet counts. Treatment-context analyses should be prespecified within bevacizumab and VEGFR tyrosine kinase inhibitor strata. Mechanistic studies should examine associations between dynamic coagulation markers and tissue factor, FXa-PAR2, thrombin-GARP-TGF-beta, fibrin-rich extracellular matrix, myeloid-cell activation, and T-cell exclusion. Machine learning and large language models may help integrate multimodal biomarker data, but their outputs require validation in clinical datasets [31]. Clinical utility should be assessed with decision-curve analysis before early coagulation activation is used to guide imaging surveillance or treatment intensification. Anticoagulation should not be treated as a routine intervention without direct evidence. The present study cannot distinguish whether coagulation activation is causal, compensatory, or a marker of aggressive disease biology.

Conclusion

In a multicancer real-world cohort receiving VEGF/VEGFR-axis inhibitor therapy, early-treatment coagulation activation was associated with non-durable radiographic benefit and a higher risk of subsequent progression. Persistent activation showed the strongest association, independent of baseline tumour burden and coagulation, and the pattern was similar in therapy without documented ICI and in ICI-containing therapy. ICI co-treatment did not significantly modify the association in interaction tests. Objective response and best tumour shrinkage did not follow the same pattern as subsequent progression. Dynamic D-dimer and fibrinogen monitoring should be evaluated prospectively in defined cancer-drug contexts.

Supplemental data

Table S1 Comparison of landmark-eligible and excluded patients
Characteristic Landmark-eligible patients Excluded patients p value
Patients, n 909 491
Median age, years 62.2 60.0 0.002
Male, n (%) 609 (67.0) 277 (56.4) <0.001
Median baseline tumour burden, cm 5.9 4.3 <0.001
Median baseline D-dimer, ng/mL 243.2 259.6 0.811
Median baseline fibrinogen, g/L 3.74 3.76 0.948
Progression events, n (%) 375 (41.3) 382 (77.8) <0.001
Median observed TTP, days 235 125 <0.001

Patients were excluded from the 909-patient landmark cohort if they progressed before or at day 100 or lacked evaluable early-treatment coagulation data. The excluded group therefore included a higher proportion of early progressors and patients with limited coagulation follow-up. Abbreviations: TTP, time to progression.

Internal validation and incremental discrimination of prognostic models for radiographic progression.
Figure S1 Internal validation and incremental discrimination of prognostic models for radiographic progression. (A) Five-fold cross-validated C-indices for radiographic TTP using clinical variables alone or combined with baseline coagulation, baseline coagulation plus the early-treatment score, cumulative coagulation burden, or coagulation trajectory. Values beside the bars indicate C-indices. (B) Time-dependent AUCs at 180, 365, and 730 days after the 100-day landmark for the clinical, clinical plus baseline coagulation, clinical plus baseline coagulation and early-treatment score, and clinical plus trajectory models. (C) Differences in C-index (ΔC-index) between the indicated models and their respective reference models. Points represent ΔC-index estimates, and horizontal lines indicate 95% bootstrap CIs. The dashed vertical line marks ΔC-index = 0; positive values indicate improved discrimination relative to the reference model. Abbreviations: AUC, area under the receiver operating characteristic curve; CI, confidence interval; C-index, concordance index; TTP, time to progression.

References

  1. Zhang L, Xu J, Zhou S, Yao F, Zhang R, You W, et al. Endothelial DGKG promotes tumor angiogenesis and immune evasion in hepatocellular carcinoma. J Hepatol. 2024;80(1):82–98. https://doi.org/10.1016/j.jhep.2023.10.006
  2. Garcia J, Hurwitz HI, Sandler AB, Miles D, Coleman RL, Deurloo R, et al. Bevacizumab (Avastin) in cancer treatment: a review of 15 years of clinical experience and future outlook. Cancer Treat Rev. 2020;86:102017. https://doi.org/10.1016/j.ctrv.2020.102017
  3. Socinski MA, Nishio M, Jotte RM, Cappuzzo F, Orlandi F, Stroyakovskiy D, et al. IMpower150 final overall survival analyses for atezolizumab plus bevacizumab and chemotherapy in first-line metastatic nonsquamous NSCLC. J Thorac Oncol. 2021;16(11):1909–24. https://doi.org/10.1016/j.jtho.2021.07.009
  4. Park S, Kim TM, Han JY, Lee GW, Shim BY, Lee YG, et al. Phase III, randomized study of atezolizumab plus bevacizumab and chemotherapy in patients with EGFR- or ALK-mutated non-small-cell lung cancer (ATTLAS, KCSG-LU19-04). J Clin Oncol. 2024;42(11):1241–51. https://doi.org/10.1200/JCO.23.01891
  5. Wang F, Jin Y, Wang M, Luo HY, Fang WJ, Wang YN, et al. Combined anti-PD-1, HDAC inhibitor and anti-VEGF for MSS/pMMR colorectal cancer: a randomized phase 2 trial. Nat Med. 2024;30(4):1035–43. https://doi.org/10.1038/s41591-024-02813-1
  6. Falanga A, Marchetti M. Hemostatic biomarkers in cancer progression. Thromb Res. 2018;164:S54–61. https://doi.org/10.1016/j.thromres.2018.01.017
  7. Tieken C, Versteeg HH. Anticoagulants versus cancer. Thromb Res. 2016;140:S148–53. https://doi.org/10.1016/S0049-3848(16)30114-1
  8. Wu Y, Liu X, Li H, Wang W, Ye L, Zhou Y, et al. D-dimer levels predict the treatment efficacy and prognosis of esophageal squamous cell carcinoma treated with PD-1/PD-L1 inhibitors. Int J Biol Markers. 2024;39(3):209–16. https://doi.org/10.1177/03936155241262045
  9. Pang B, Wang J, Wang J, Zhang J, Ren X, Han Y. Association of baseline plasma D-dimer and platelets with progression-free survival in patients with stage IV non-small cell lung cancer treated with anti-PD-1 antibody. Transl Lung Cancer Res. 2024;13(11):3106–21. https://doi.org/10.21037/tlcr-24-763
  10. Li X, Lu D, Zhang Z, Zhang Y, Wang J, Hu Y. Prognostic value of plasma D-dimer levels in advanced non-small cell lung cancer patients treated with immune checkpoint inhibitors: a retrospective study. J Thorac Dis. 2022;14(10):4125–35. https://doi.org/10.21037/jtd-22-1363
  11. Chen C, Yin H, Zhang Y, Chen H, Xu J, Ren L. Plasma D-dimer and interleukin-6 are associated with treatment response and progression-free survival in advanced NSCLC patients on anti-PD-1 therapy. Cancer Med. 2023;12(15):15831–40. https://doi.org/10.1002/cam4.6222
  12. Xu L, Li Y, Wang K, Liu C, Liu R, Zhang W. D-dimer predicts the response of patients with gastric cancer to first-line immunotherapy combined with chemotherapy. J Gastrointest Oncol. 2025;16(3):899–908. https://doi.org/10.21037/jgo-24-824
  13. Geidel G, Adam L, Bansch S, Fekade N, Degenhardt S, Runger A, et al. Baseline D-dimers as predictive marker of efficacy of cemiplimab treatment in cutaneous squamous cell carcinoma. Eur J Cancer. 2025;226:115639. https://doi.org/10.1016/j.ejca.2025.115639
  14. Ho K, Matsuda S, Booka E, Soneda W, Okui J, Hoshino S, et al. Plasma fibrinogen predicts response to immune checkpoint inhibitor by inflammatory tumor microenvironment in esophageal cancer. Cancer Med. 2026;15(1):e71548. https://doi.org/10.1002/cam4.71548
  15. Metelli A, Wu BX, Riesenberg B, Guglietta S, Huck JD, Mills C, et al. Thrombin contributes to cancer immune evasion via proteolysis of platelet-bound GARP to activate LTGF-beta. Sci Transl Med. 2020;12(525):eaay4860. https://doi.org/10.1126/scitranslmed.aay4860
  16. Yang Y, Stang A, Schweickert PG, Lanman NA, Paul EN, Monia BP, et al. Thrombin signaling promotes pancreatic adenocarcinoma through PAR-1-dependent immune evasion. Cancer Res. 2019;79(13):3417–30. https://doi.org/10.1158/0008-5472.CAN-18-3206
  17. Wilgenbus P, Pott J, Pagel S, Witzler C, Royce J, Marini F, et al. Coagulation proteases modulate nucleic acid uptake and cGAS–STING–IFN induction in the tumor microenvironment. JCI Insight. 2025;10(17):e190311. https://doi.org/10.1172/jci.insight.190311
  18. Cali B, Troiani M, Bressan S, Attanasio G, Merler S, Moscarda V, et al. Coagulation factor X promotes resistance to androgen-deprivation therapy in prostate cancer. Cancer Cell. 2024;42(10):1676–92.e11. https://doi.org/10.1016/j.ccell.2024.08.018
  19. Falcomata C, Nielsen SR, Schaefer MM, Singh B, Tepper A, Chhamalwan D, et al. Plasminogen activator inhibitors orchestrate the immunosuppressive tumor microenvironment in pancreatic cancer. bioRxiv. 2025. https://doi.org/10.1101/2025.06.11.659098
  20. Vandenbroucke JP, von Elm E, Altman DG, Gotzsche PC, Mulrow CD, Pocock SJ, et al. Strengthening the Reporting of Observational Studies in Epidemiology (STROBE): explanation and elaboration. Ann Intern Med. 2007;147(8):W163–94. https://doi.org/10.7326/0003-4819-147-8-200710160-00010-w1
  21. Eisenhauer EA, Therasse P, Bogaerts J, Schwartz LH, Sargent D, Ford R, et al. New response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1). Eur J Cancer. 2009;45(2):228–47. https://doi.org/10.1016/j.ejca.2008.10.026
  22. Seymour L, Bogaerts J, Perrone A, Adam J, Barlesi F, Baselga J, et al. iRECIST: guidelines for response criteria for use in trials testing immunotherapeutics. Lancet Oncol. 2017;18(3):e143–52. https://doi.org/10.1016/S1470-2045(17)30074-8
  23. Collins GS, Reitsma JB, Altman DG, Moons KGM. Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): the TRIPOD statement. Ann Intern Med. 2015;162(1):55–63. https://doi.org/10.7326/M14-0697
  24. Altman DG, McShane LM, Sauerbrei W, Taube SE. Reporting recommendations for tumor marker prognostic studies (REMARK): explanation and elaboration. BMC Med. 2012;10:51. https://doi.org/10.1186/1741-7015-10-51
  25. Tsimafeyeu I, Musaeva G, Utyashev I, Zakurdaeva K, Gerk I, Olshanskaya A, et al. Fibrinogen levels in patients with metastatic renal cell carcinoma treated with nivolumab: results of a multicenter prospective trial. Kidney Cancer. 2023;7(1):115–21. https://doi.org/10.3233/KCA-230007
  26. Ruf W, Graf C. Coagulation signaling and cancer immunotherapy. Thromb Res. 2020;191:S106–11. https://doi.org/10.1016/S0049-3848(20)30406-0
  27. Galmiche A, Saidak Z, Ghiringhelli F. Targeting coagulation to unlock antitumor immunity? Oncoimmunology. 2022;11(1):2045696. https://doi.org/10.1080/2162402X.2022.2045696
  28. Bell HN, Zou W. Beyond the barrier: unraveling the mechanisms of immunotherapy resistance. Annu Rev Immunol. 2024;42(1):521–50. https://doi.org/10.1146/annurev-immunol-101819-024752
  29. Ai X, Jia B, He Z, Zhang J, Zhuo M, Zhao J, et al. Noninvasive early identification of durable clinical benefit from immune checkpoint inhibition: a prospective multicenter study (NCT04566432). Signal Transduct Target Ther. 2024;9(1):350. https://doi.org/10.1038/s41392-024-02060-3
  30. Nabet BY, Esfahani MS, Moding EJ, Hamilton EG, Chabon JJ, Rizvi H, et al. Noninvasive early identification of therapeutic benefit from immune checkpoint inhibition. Cell. 2020;183(2):363–76.e13. https://doi.org/10.1016/j.cell.2020.09.001
  31. Guo SB, Liu DY, Fang XJ, Meng Y, Zhou ZZ, Li J, et al. Current concerns and future directions of large language model ChatGPT in medicine: a machine-learning-driven global-scale bibliometric analysis. Int J Surg. 2026;112(2):2805–22. https://doi.org/10.1097/JS9.0000000000003668

About this article

Correspondence

*Correspondence to Xin Zhou: [email protected] and Shuo Sun: [email protected].

Equal contribution

Equally contributed to this work: Yingda Liu and Bai Liu.

Article history

Received: 2026-09-19 · Accepted: 2026-10-10

Funding

This research project is funded by the Internal Research Program of the 983rd Research Institute of the Chinese Logistics Support Force, with project number 983YN25F0010.

Data availability

The datasets generated during the current study are available from the corresponding author on reasonable request.

Conflict of interest

Authors declare no conflicts of interest.

Artificial intelligence (AI) use statement

During the preparation of this work the authors used DeepSeek-V4-Flash in order to assist with language polishing, grammar correction, and sentence structure optimization. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication.

Downloads

Download data is not yet available.

References

Zhang L, Xu J, Zhou S, Yao F, Zhang R, You W, et al. Endothelial DGKG promotes tumor angiogenesis and immune evasion in hepatocellular carcinoma. J Hepatol. 2024;80(1):82–98. https://doi.org/10.1016/j.jhep.2023.10.006 DOI: https://doi.org/10.1016/j.jhep.2023.10.006

Garcia J, Hurwitz HI, Sandler AB, Miles D, Coleman RL, Deurloo R, et al. Bevacizumab (Avastin) in cancer treatment: a review of 15 years of clinical experience and future outlook. Cancer Treat Rev. 2020;86:102017. https://doi.org/10.1016/j.ctrv.2020.102017 DOI: https://doi.org/10.1016/j.ctrv.2020.102017

Socinski MA, Nishio M, Jotte RM, Cappuzzo F, Orlandi F, Stroyakovskiy D, et al. IMpower150 final overall survival analyses for atezolizumab plus bevacizumab and chemotherapy in first-line metastatic nonsquamous NSCLC. J Thorac Oncol. 2021;16(11):1909–24. https://doi.org/10.1016/j.jtho.2021.07.009 DOI: https://doi.org/10.1016/j.jtho.2021.07.009

Park S, Kim TM, Han JY, Lee GW, Shim BY, Lee YG, et al. Phase III, randomized study of atezolizumab plus bevacizumab and chemotherapy in patients with EGFR- or ALK-mutated non-small-cell lung cancer (ATTLAS, KCSG-LU19-04). J Clin Oncol. 2024;42(11):1241–51. https://doi.org/10.1200/JCO.23.01891 DOI: https://doi.org/10.1200/JCO.23.01891

Wang F, Jin Y, Wang M, Luo HY, Fang WJ, Wang YN, et al. Combined anti-PD-1, HDAC inhibitor and anti-VEGF for MSS/pMMR colorectal cancer: a randomized phase 2 trial. Nat Med. 2024;30(4):1035–43. https://doi.org/10.1038/s41591-024-02813-1 DOI: https://doi.org/10.1038/s41591-024-02813-1

Falanga A, Marchetti M. Hemostatic biomarkers in cancer progression. Thromb Res. 2018;164:S54–61. https://doi.org/10.1016/j.thromres.2018.01.017 DOI: https://doi.org/10.1016/j.thromres.2018.01.017

Tieken C, Versteeg HH. Anticoagulants versus cancer. Thromb Res. 2016;140:S148–53. https://doi.org/10.1016/S0049-3848(16)30114-1 DOI: https://doi.org/10.1016/S0049-3848(16)30114-1

Wu Y, Liu X, Li H, Wang W, Ye L, Zhou Y, et al. D-dimer levels predict the treatment efficacy and prognosis of esophageal squamous cell carcinoma treated with PD-1/PD-L1 inhibitors. Int J Biol Markers. 2024;39(3):209–16. https://doi.org/10.1177/03936155241262045 DOI: https://doi.org/10.1177/03936155241262045

Pang B, Wang J, Wang J, Zhang J, Ren X, Han Y. Association of baseline plasma D-dimer and platelets with progression-free survival in patients with stage IV non-small cell lung cancer treated with anti-PD-1 antibody. Transl Lung Cancer Res. 2024;13(11):3106–21. https://doi.org/10.21037/tlcr-24-763 DOI: https://doi.org/10.21037/tlcr-24-763

Li X, Lu D, Zhang Z, Zhang Y, Wang J, Hu Y. Prognostic value of plasma D-dimer levels in advanced non-small cell lung cancer patients treated with immune checkpoint inhibitors: a retrospective study. J Thorac Dis. 2022;14(10):4125–35. https://doi.org/10.21037/jtd-22-1363 DOI: https://doi.org/10.21037/jtd-22-1363

Chen C, Yin H, Zhang Y, Chen H, Xu J, Ren L. Plasma D-dimer and interleukin-6 are associated with treatment response and progression-free survival in advanced NSCLC patients on anti-PD-1 therapy. Cancer Med. 2023;12(15):15831–40. https://doi.org/10.1002/cam4.6222 DOI: https://doi.org/10.1002/cam4.6222

Xu L, Li Y, Wang K, Liu C, Liu R, Zhang W. D-dimer predicts the response of patients with gastric cancer to first-line immunotherapy combined with chemotherapy. J Gastrointest Oncol. 2025;16(3):899–908. https://doi.org/10.21037/jgo-24-824 DOI: https://doi.org/10.21037/jgo-24-824

Geidel G, Adam L, Bansch S, Fekade N, Degenhardt S, Runger A, et al. Baseline D-dimers as predictive marker of efficacy of cemiplimab treatment in cutaneous squamous cell carcinoma. Eur J Cancer. 2025;226:115639. https://doi.org/10.1016/j.ejca.2025.115639 DOI: https://doi.org/10.1016/j.ejca.2025.115639

Ho K, Matsuda S, Booka E, Soneda W, Okui J, Hoshino S, et al. Plasma fibrinogen predicts response to immune checkpoint inhibitor by inflammatory tumor microenvironment in esophageal cancer. Cancer Med. 2026;15(1):e71548. https://doi.org/10.1002/cam4.71548 DOI: https://doi.org/10.1002/cam4.71548

Metelli A, Wu BX, Riesenberg B, Guglietta S, Huck JD, Mills C, et al. Thrombin contributes to cancer immune evasion via proteolysis of platelet-bound GARP to activate LTGF-beta. Sci Transl Med. 2020;12(525):eaay4860. https://doi.org/10.1126/scitranslmed.aay4860 DOI: https://doi.org/10.1126/scitranslmed.aay4860

Yang Y, Stang A, Schweickert PG, Lanman NA, Paul EN, Monia BP, et al. Thrombin signaling promotes pancreatic adenocarcinoma through PAR-1-dependent immune evasion. Cancer Res. 2019;79(13):3417–30. https://doi.org/10.1158/0008-5472.CAN-18-3206 DOI: https://doi.org/10.1158/0008-5472.CAN-18-3206

Wilgenbus P, Pott J, Pagel S, Witzler C, Royce J, Marini F, et al. Coagulation proteases modulate nucleic acid uptake and cGAS–STING–IFN induction in the tumor microenvironment. JCI Insight. 2025;10(17):e190311. https://doi.org/10.1172/jci.insight.190311 DOI: https://doi.org/10.1172/jci.insight.190311

Cali B, Troiani M, Bressan S, Attanasio G, Merler S, Moscarda V, et al. Coagulation factor X promotes resistance to androgen-deprivation therapy in prostate cancer. Cancer Cell. 2024;42(10):1676–92.e11. https://doi.org/10.1016/j.ccell.2024.08.018 DOI: https://doi.org/10.1016/j.ccell.2024.08.018

Falcomata C, Nielsen SR, Schaefer MM, Singh B, Tepper A, Chhamalwan D, et al. Plasminogen activator inhibitors orchestrate the immunosuppressive tumor microenvironment in pancreatic cancer. bioRxiv. 2025. https://doi.org/10.1101/2025.06.11.659098 DOI: https://doi.org/10.1101/2025.06.11.659098

Vandenbroucke JP, von Elm E, Altman DG, Gotzsche PC, Mulrow CD, Pocock SJ, et al. Strengthening the Reporting of Observational Studies in Epidemiology (STROBE): explanation and elaboration. Ann Intern Med. 2007;147(8):W163–94. https://doi.org/10.7326/0003-4819-147-8-200710160-00010-w1 DOI: https://doi.org/10.7326/0003-4819-147-8-200710160-00010-w1

Eisenhauer EA, Therasse P, Bogaerts J, Schwartz LH, Sargent D, Ford R, et al. New response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1). Eur J Cancer. 2009;45(2):228–47. https://doi.org/10.1016/j.ejca.2008.10.026 DOI: https://doi.org/10.1016/j.ejca.2008.10.026

Seymour L, Bogaerts J, Perrone A, Adam J, Barlesi F, Baselga J, et al. iRECIST: guidelines for response criteria for use in trials testing immunotherapeutics. Lancet Oncol. 2017;18(3):e143–52. https://doi.org/10.1016/S1470-2045(17)30074-8 DOI: https://doi.org/10.1016/S1470-2045(17)30074-8

Collins GS, Reitsma JB, Altman DG, Moons KGM. Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): the TRIPOD statement. Ann Intern Med. 2015;162(1):55–63. https://doi.org/10.7326/M14-0697 DOI: https://doi.org/10.7326/M14-0697

Altman DG, McShane LM, Sauerbrei W, Taube SE. Reporting recommendations for tumor marker prognostic studies (REMARK): explanation and elaboration. BMC Med. 2012;10:51. https://doi.org/10.1186/1741-7015-10-51 DOI: https://doi.org/10.1186/1741-7015-10-51

Tsimafeyeu I, Musaeva G, Utyashev I, Zakurdaeva K, Gerk I, Olshanskaya A, et al. Fibrinogen levels in patients with metastatic renal cell carcinoma treated with nivolumab: results of a multicenter prospective trial. Kidney Cancer. 2023;7(1):115–21. https://doi.org/10.3233/KCA-230007 DOI: https://doi.org/10.3233/KCA-230007

Ruf W, Graf C. Coagulation signaling and cancer immunotherapy. Thromb Res. 2020;191:S106–11. https://doi.org/10.1016/S0049-3848(20)30406-0 DOI: https://doi.org/10.1016/S0049-3848(20)30406-0

Galmiche A, Saidak Z, Ghiringhelli F. Targeting coagulation to unlock antitumor immunity? Oncoimmunology. 2022;11(1):2045696. https://doi.org/10.1080/2162402X.2022.2045696 DOI: https://doi.org/10.1080/2162402X.2022.2045696

Bell HN, Zou W. Beyond the barrier: unraveling the mechanisms of immunotherapy resistance. Annu Rev Immunol. 2024;42(1):521–50. https://doi.org/10.1146/annurev-immunol-101819-024752 DOI: https://doi.org/10.1146/annurev-immunol-101819-024752

Ai X, Jia B, He Z, Zhang J, Zhuo M, Zhao J, et al. Noninvasive early identification of durable clinical benefit from immune checkpoint inhibition: a prospective multicenter study (NCT04566432). Signal Transduct Target Ther. 2024;9(1):350. https://doi.org/10.1038/s41392-024-02060-3 DOI: https://doi.org/10.1038/s41392-024-02060-3

Nabet BY, Esfahani MS, Moding EJ, Hamilton EG, Chabon JJ, Rizvi H, et al. Noninvasive early identification of therapeutic benefit from immune checkpoint inhibition. Cell. 2020;183(2):363–76.e13. https://doi.org/10.1016/j.cell.2020.09.001 DOI: https://doi.org/10.1016/j.cell.2020.09.001

Guo SB, Liu DY, Fang XJ, Meng Y, Zhou ZZ, Li J, et al. Current concerns and future directions of large language model ChatGPT in medicine: a machine-learning-driven global-scale bibliometric analysis. Int J Surg. 2026;112(2):2805–22. https://doi.org/10.1097/JS9.0000000000003668 DOI: https://doi.org/10.1097/JS9.0000000000003668

Article files

Additional Files

Publication details

Citation
Advance online
Language
EN
License
CC BY 4.0
Section
Research article

How to cite (Vancouver)

Liu Y, Liu B, Liu S, Zhang T, Lv T, Han S, et al. Early-treatment coagulation activation and radiographic progression during VEGF/VEGFR inhibitor therapy: A multicancer real-world cohort study. Biomol Biomed. 2026. https://doi.org/10.17305/bb.2026.15049

Share

Related articles