Abstract
Evorpacept blocks the CD47–signal regulatory protein α interaction, enhancing antibody-dependent cellular phagocytosis. In the phase 2 portion of ASPEN-06 (a multicenter, open-label, randomized phase 2/3 study), 127 patients with pretreated, human epidermal growth factor receptor 2 (HER2)-overexpressing, advanced gastric/gastroesophageal junction cancer were randomized to evorpacept plus trastuzumab, ramucirumab and paclitaxel (evo + TRP; n = 63; fresh HER2+ biopsy, n = 22) or TRP alone (n = 64; fresh HER2+ biopsy, n = 26). The primary end point was investigator-assessed objective response rate (ORR). The primary analysis of ORR was designed to evaluate an ORR exceeding the historical 30% benchmark (ramucirumab/paclitaxel) (80% power for intent to treat (ITT) and 50% power for the ITT subpopulation (HER2 overexpression based on a post-trastuzumab ‘fresh’ biopsy)) and an improvement of ≥8.0% and ≥9.7% versus TRP in the ITT and ITT subpopulations, respectively. Key secondary end points included ORR by blinded independent central review, duration of response and progression-free survival by investigator or blinded independent central review, overall survival and safety. Post hoc biomarker analyses, including the assessment of the association of treatment efficacy with retained HER2 status (defined by HER2 positivity on fresh tumor biopsy or amplification in circulating tumor DNA analysis) and CD47 expression in tumor samples, were conducted. The investigator-assessed ORRs were 40.3% (evo + TRP) versus 26.6% (TRP) in the ITT population and 54.8% versus 23.1% in the fresh biopsy HER2+ subgroup. ORR differences (13.7% and 31.7%) exceeded the prespecified thresholds in both the ITT population and fresh biopsy HER2+ subgroup, meeting one of the two primary objectives; however, compared to the historical benchmark (30%), ORRs in the ITT population (40.3%; P = 0.0949, one-sided) and fresh biopsy HER2+ subgroup (54.8%; P = 0.030, one-sided) did not meet the prespecified statistical criterion (one-sided α = 0.025). While hematologic toxicities were more common with evo + TRP, overall safety was similar. In summary, evo + TRP showed encouraging efficacy with manageable safety in advanced gastric/gastroesophageal junction cancer. ClinicalTrials.gov identifier: NCT05002127.
Main
The transmembrane protein CD47, which acts as a major checkpoint in the innate immune system, is overexpressed at different levels in many tumor types and is also expressed on the surface of healthy cells (such as hematopoietic cells)1,2,3,4,5,6. In addition, increased CD47 expression is a negative prognostic factor in multiple cancer types7,8,9,10. A key role of CD47 involves binding to signal regulatory protein-α (SIRPα) on the surface of macrophages, triggering an antiphagocytic signal and preventing the cell from being cleared by the innate immune system1,3,4. Cancer cells can exploit this mechanism by overexpressing CD47 to evade phagocytosis and clearance1,2,4,5,6,11,12. Antibody-dependent cellular phagocytosis (ADCP) of cancer can be promoted through blockade of the CD47–SIRPα interaction, which inhibits this antiphagocytic signal; however, blockade of the CD47–SIRPα interaction alone is insufficient to generate maximum antitumor activity; Fcγ receptor binding by the macrophage is required to provide a prophagocytic signal4.
Evorpacept is a high-affinity engineered fusion protein containing the N-terminal D1 domain of SIRPα genetically linked to a modified, inactive Fc domain from human IgG1 (ref. 13). An inactive Fc domain was incorporated into the design to mitigate off-tumor CD47-targeted ADCP toxicities observed with other anti-CD47 agents, including magrolimab13,14. Combining evorpacept with tumor antigen-specific monoclonal antibodies that possess an active Fc domain (for example, trastuzumab in human epidermal growth factor receptor 2-positive (HER2+) gastric cancer) is hypothesized to improve efficacy through more effective ADCP by reducing the antiphagocytic signal via CD47 blockade, thereby enhancing the antitumor prophagocytic signal induced with the anticancer antibody13,15. In preclinical studies, addition of evorpacept to trastuzumab enhanced ADCP in OE19 cells in a dose-dependent manner and led to statistically significant growth inhibition compared to trastuzumab alone in a gastric cancer mouse model13. In the first-in-human, phase 1 ASPEN-01 study (NCT03013218), evorpacept in combination with trastuzumab, ramucirumab and paclitaxel (TRP) demonstrated promising antitumor activity with an ORR of 72.2% and a favorable safety profile in previously treated gastric cancer15,16. Standard options for treating HER2+ gastric/gastroesophageal junction (GEJ) cancer include trastuzumab in combination with chemotherapy with or without pembrolizumab in the first-line setting and either ramucirumab and paclitaxel (RP) or trastuzumab deruxtecan in the second- or third-line setting17,18,19,20.
Here, we present results from the phase 2 portion of the phase 2/3, randomized ASPEN-06 study, along with HER2 and CD47 post hoc biomarker analyses. The study investigated the efficacy and safety of evorpacept in combination with TRP as second- or third-line treatment for patients with HER2+, advanced gastric/GEJ cancer that had progressed on or after previous anti-HER2 therapy. While RP was a standard-of-care treatment in the second- or third-line setting at initiation of the randomized study, trastuzumab was added to RP to provide a tumor antigen-bound active Fc domain to enable ADCP in the presence of CD47 blockade by evorpacept, and allow assessment of the contribution of evorpacept within a trastuzumab-containing regimen. Given that loss of HER2 expression has been observed in some patients with advanced or recurrent gastric cancer after previous trastuzumab, retained HER2+ disease may be important for treatment selection21. Based on the assumption that maximum ADCP activity of trastuzumab plus evorpacept will occur when their respective targets—HER2 and CD47—are sufficiently expressed, a key exploratory hypothesis was to assess outcomes in patients with retained HER2+ disease and CD47 tumor expression.
Results
Trial design
Between 28 February 2022 and 22 January 2024, 236 patients were screened and 127 were enrolled across 49 academic and community healthcare institutions in ten countries in the open-label, phase 2 portion of the randomized ASPEN-06 study (Fig. 1). The data were collected from 28 February 2022 through 15 May 2025. Key inclusion criteria were HER2+ advanced or metastatic gastric/GEJ adenocarcinoma that had progressed on or after a previous HER2-directed anticancer agent (for example, trastuzumab) and/or fluoropyrimidine-containing/platinum-containing chemotherapy. Key exclusion criteria were symptomatic central nervous system involvement or leptomeningeal disease, previous ramucirumab or any anti-CD47 or anti-SIRPα agent, or systemic anticancer therapy within 4 weeks of starting study treatment. Further information on trial design and eligibility criteria is available in the Methods.
Patients received either evorpacept plus TRP or TRP alone. Evorpacept (30 mg kg−1), trastuzumab (initial dose of 6 mg kg−1; 4 mg kg−1 doses thereafter), and ramucirumab (8 mg kg−1) were administered every 2 weeks; paclitaxel (80 mg m−2) was administered every week for the first 3 weeks of every 28-day cycle. In the protocol-defined primary analysis (data cutoff 24 May 2024), the two primary objectives are (1) to achieve a statistically significant improvement in investigator-assessed objective response rate (ORR) in the evorpacept plus TRP arm in the intent-to-treat (ITT) population and the subgroup with a fresh HER2+ biopsy compared to an assumed historical control of 30%, and (2) to achieve a difference in ORR of ≥8% in the ITT population and ≥9.7% in the fresh biopsy HER2+ subgroup between evorpacept plus TRP and the internal control arm of TRP. Secondary end points include ORR by blinded independent central review (BICR), duration of response (DOR) and progression-free survival (PFS) by investigator or BICR, overall survival (OS) and the frequency and severity of adverse events. Exploratory end points include molecular analysis in tumor biopsy and peripheral blood samples before treatment, and correlation of immunohistochemistry (IHC), response and gene alterations.
Patients
Overall, 127 were randomized to evorpacept plus TRP (n = 63) or TRP alone (internal control arm; n = 64) (Fig. 1). At data cutoff for the primary analysis (24 May 2024), the median follow-up duration was 7.8 months. Baseline patient demographics and clinical characteristics were balanced across treatment arms (Table 1). In brief, patients had a median age of 64 years, the majority were male (81.1%) and most were Asian (48.8%) or white (29.9%). The most common primary tumor location was the stomach (72.4%), 105 (82.7%) patients had tumors with immunohistochemistry (IHC) 3+ HER2 overexpression, and 93 (73.2%) patients were receiving the study treatment as their second line of therapy. All patients had received trastuzumab or other HER2-directed therapies as first- or second-line treatment. Eighteen (14.2%) patients had received previous trastuzumab deruxtecan and 27 (21.3%) had received previous anti-programmed cell death protein 1 therapy.
For the post hoc analysis (data cutoff 15 May 2025), 120 (94.5%) patients had plasma samples evaluable for circulating tumor DNA (ctDNA) and amplification of the ERBB2 gene (which encodes HER2) was detected in 86 (67.7%) (Table 1); 47 (37.0%) patients had a fresh HER2+ biopsy. Of the 127 patients enrolled, 95 (74.8%) were identified as having retained HER2+ disease by either a fresh biopsy (n = 47; 37.0%) or ctDNA (n = 86; 67.7%), whereas 32 (25.2%) who were HER2+ on archival tissue did not show detectable ERBB2 gene amplification in ctDNA at baseline. The median time from biopsy to the first dose of study medication was 1.2 months for patients with a fresh biopsy (37.8%) and 12.3 months for those with an archival biopsy (62.2%). Biopsy tissue was available for CD47 expression testing in 120 (94.5%) patients.
Efficacy
In the protocol-defined primary analysis (data cutoff 24 May 2024), ORRs for evorpacept plus TRP were numerically greater than those for the internal control TRP alone: by ≥8% in the ITT population (40.3%, 95% confidence interval (CI) 28.1–49.3 versus 26.6%, 95% CI 16.3–39.1) and by ≥9.7% in the fresh biopsy HER2+ subgroup (54.8%, 95% CI 32.3–66.3 versus 23.1%, 95% CI 9.0–43.6) (Table 2 and Fig. 2a). The differences in ORR between treatment arms (ITT population: 13.7%; fresh biopsy HER2+ subgroup: 31.7%) exceeded the prespecified ORR differences of 8% and 9.7%, respectively, establishing a clinically meaningful contribution of evorpacept and meeting one of the two primary objectives in both the ITT population and fresh biopsy HER2+ subgroup; however, although the observed ORR (40.3%) in the ITT population and ORR (54.8%) in the fresh biopsy HER2+ subgroup exceeded the prespecified historical benchmark of 30% (P = 0.0949, one-sided) and (P = 0.030, one-sided), respectively, this did not satisfy the prespecified statistical success criterion based on a one-sided α of 0.025. Therefore, the primary objective comparing ORR with the historical benchmark was not met. Median DOR by investigator assessment was 15.7 months with evorpacept plus TRP versus 7.6 months with TRP alone in both the ITT population and fresh biopsy HER2+ subgroup (Table 2). Median PFS was 7.5 months versus 7.4 months with evorpacept plus TRP versus TRP alone in the ITT population, and 9.0 months versus 7.4 months in the fresh biopsy HER2+ subgroup (Fig. 2b). OS data were not mature. Similar results were reported for ORR, DOR and PFS assessed by BICR in the ITT population and fresh biopsy HER2+ subgroup (Supplementary Table 1 and Supplementary Figs 1 and 2). Gender-based analyses were also conducted for the primary end point as follows: in male patients, the observed ORRs were 36.4% and 27.1% in the evorpacept plus TRP (n = 55) and TRP arms (n = 48) (Extended Data Table 1); in female patients, ORRs were 50.0% and 25.0% in the evorpacept plus TRP (n = 8) and TRP arms (n = 16) (Extended Data Table 2). Efficacy outcomes for the ITT population and the fresh biopsy HER2+ subgroup in the updated analysis (data cutoff 15 May 2025) were consistent with the primary analysis (Extended Data Table 3, Extended Data Figs. 1 and 2, Supplementary Table 2 and Supplementary Figs. 3 and 4) and OS was similar in both treatment arms (ITT population: hazard ratio (HR) 1.07, 95% CI 0.69–1.66; fresh biopsy HER2+ subgroup: HR 1.05, 95% CI 0.46–2.38) (Extended Data Fig. 3).
Safety
The median (range) duration of evorpacept treatment was 5.9 (0–34) months (data cutoff 15 May 2025). For evorpacept plus TRP versus TRP alone, the median (range) duration of treatment was 5.9 (0–34) months versus 6.0 (0–33) months for trastuzumab, 5.9 (0–31) months versus 4.1 (0–33) months for ramucirumab and 5.4 (0–34) months versus 5.5 (1–22) months for paclitaxel.
The safety population comprised 63 patients in the evorpacept plus TRP arm and 63 in the TRP arm; one patient in the TRP arm who was randomized to treatment did not receive study medication. Treatment-emergent adverse events (TEAEs) of any grade occurred in 100% of patients in both the evorpacept plus TRP and TRP alone arms. The most common TEAEs were neutrophil count decreased/neutropenia (69.8% and 55.6%, respectively), anemia (58.7% and 38.1%) and diarrhea (41.3% and 36.5%) (Extended Data Table 4). Grade ≥3 TEAEs occurred in 57 (90.5%) patients receiving evorpacept plus TRP and 50 (79.4%) receiving TRP alone, most commonly neutrophil count decreased/neutropenia (55.6% and 39.7%, respectively), anemia (23.8% and 17.5%) and white blood cell count decreased/leukopenia (17.5% and 9.5%) (Extended Data Table 4). Febrile neutropenia was reported in two (3.2%) patients in the evorpacept plus TRP arm (grade 3) and five (7.9%) in the TRP arm (three grade 3, two grade 4). Twelve (9.5%) patients experienced grade 5 TEAEs (five in the evorpacept plus TRP arm and seven in the TRP arm). Two further patients in the evorpacept plus TRP arm died due to grade 5 disease progression, one with treatment discontinuation due to disease progression and subsequent death, and one with grade 5 disease progression during active treatment. Three (2.4%) patients experienced grade 5 treatment-related adverse eventss—two in the evorpacept plus TRP arm (grade 5 esophageal perforation in one patient, possibly related to ramucirumab with potential contribution of underlying gastric cancer, but not considered related to evorpacept, trastuzumab, or paclitaxel; grade 5 renal insufficiency, acute respiratory insufficiency, and anasarca in the other patient, possibly related to ramucirumab and trastuzumab with potential contribution by underlying gastric cancer, but not considered related to evorpacept or paclitaxel), and one in the TRP arm, considered related to TRP (grade 5 pneumonia).
In the evorpacept plus TRP arm, TEAEs led to discontinuation of evorpacept in 5 (7.9%), of trastuzumab in 6 (9.5%), of ramucirumab in 7 (11.1%) and of paclitaxel in 13 (20.6%) patients. In the TRP arm, TEAEs led to discontinuation of trastuzumab in 4 (6.3%), ramucirumab in 11 (17.5%) and paclitaxel in 10 (15.9%) patients.
Post hoc biomarker analysis
Evaluation of baseline HER2 status
Subgroup analyses indicated that antitumor activity may be more pronounced in patients with retained HER2+ disease compared to the ITT population. Hence, a post hoc biomarker analysis was conducted using mature, final data from the updated analysis (data cutoff 15 May 2025) and in patients with evidence of retained HER2+ disease obtained by fresh biopsy or ERBB2 gene amplification detected in ctDNA. The ORR by investigator with evorpacept plus TRP versus TRP alone was 59.1% versus 24.0% in the fresh biopsy HER2+ subgroup (n = 47), 48.8% versus 25.6% in the ctDNA HER2+ subgroup (n = 86), and 48.9% versus 25.0% in the fresh biopsy or ctDNA HER2+ subgroup (n = 95) (Extended Data Table 3). In all three HER2+ subgroups (fresh biopsy HER2+ subgroup, ctDNA HER2+ subgroup and fresh biopsy or ctDNA HER2+ subgroup), the ORR 95% CIs for evorpacept plus TRP excluded the historical control ORR of 30% (Extended Data Table 3). For the fresh biopsy or ctDNA HER2+ subgroup, median DOR was numerically longer with evorpacept plus TRP versus TRP alone (15.7 versus 9.1 months; HR 0.59, 95% CI 0.24–1.45) (Extended Data Table 3). The HR for PFS numerically favored evorpacept plus TRP compared to TRP alone (0.72, 95% CI 0.44–1.18) (Extended Data Fig. 2), while the HR for OS with evorpacept plus TRP versus TRP alone was 0.95 (95% CI 0.58–1.56) (Extended Data Fig. 3) in the fresh biopsy or ctDNA HER2+ subgroup. Results for DOR, PFS, and OS for the fresh biopsy HER2+ subgroup and the ctDNA HER2+ subgroup were similar to those in the fresh biopsy or ctDNA HER2+ subgroup (Extended Data Table 3 and Extended Data Figs. 2 and 3). ORR, DOR and PFS assessed by BICR in the exploratory subgroups were consistent with the investigator-assessed results (Supplementary Table 2 and Supplementary Figs. 3 and 4).
Evaluation of CD47 expression as a predictive biomarker
Of 127 patients in the ITT population, 90 (70.9%) had retained HER2+ disease by fresh biopsy or ctDNA and were evaluable for CD47 expression. Receiver operating characteristic (ROC) curves and Youden index plots for different staining intensities were generated (Extended Data Figs. 4a,b). The area under the curve for CD47 IHC3+ staining of tumor cells was 0.63 for evorpacept plus TRP versus 0.51 for TRP alone, with a cutoff of 5% IHC3+ identified as a suitable candidate for defining CD47-high based on the maximum vertical distance above the diagonal line in the ROC plot and the maximum Youden index in the evorpacept plus TRP arm in the Youden index plot (Extended Data Figs. 4a,b). Additional analyses of investigator-assessed ORR (logistic regression model), and OS and investigator-assessed PFS (both Cox proportional hazards model) across a range of cutoffs were consistent with this, with CD47 IHC3+ of approximately 5% showing the highest maximum log-likelihood across all three end points (Extended Data Fig. 4c). The 5% IHC3+ threshold of the maximum log-likelihood was independently identified for each end point. Additionally, interaction tests were conducted using a 5% threshold for ORR, PFS and OS. This study was not powered to detect interactions between treatment and subgroups defined by a CD47 threshold; however, unadjusted P values of 0.161, 0.006 and 0.039 were observed for ORR, PFS and OS, respectively, supporting further evaluation of CD47 expression as a predictive biomarker in future studies.
In patients with retained HER2+ disease and ≥5% of tumor cells with membrane CD47 IHC3+ staining (CD47-high; n = 48), there was a further trend for improvement in all efficacy end points with the addition of evorpacept to TRP. The ORR by investigator assessment was higher with evorpacept plus TRP versus TRP alone (63.6% versus 23.1%) (Table 3). Median DOR for evorpacept plus TRP versus TRP alone was 25.5 versus 8.4 months (HR 0.27, 95% CI 0.07–1.06). The HR for PFS was 0.38 (95% CI 0.17–0.84; median PFS 19.5 versus 7.0 months) and for OS was 0.66 (95% CI 0.32–1.37) (Fig. 3). For patients with retained HER2+ disease and CD47 expression <5% (CD47-low; n = 42), there was no evidence that adding evorpacept to TRP provided clinical benefit. The ORR was similar for evorpacept plus TRP and TRP alone (36.4% and 30.0%, respectively) (Table 3). The HR for PFS was 1.48 (95% CI 0.73–3.01; median PFS 5.8 versus 5.5 months) and for OS was 1.74 (95% CI 0.81–3.72; Fig. 3), indicating a trend favoring TRP. Concordance in responder classifications between investigator and BICR assessments exceeded 80% (Supplementary Table 3) and similar results were reported for ORR, DOR, and PFS assessed by BICR (Supplementary Table 4 and Supplementary Fig 4).
To assess whether CD47 expression was a negative prognostic factor for the TRP arm, PFS and OS were modeled using the cut-point of 5% of tumor cells with membrane CD47 IHC3+ staining. There was a trend for shorter PFS (HR 1.42, 95% CI 0.70–2.85) and shorter OS (HR 1.54, 95% CI, 0.73–3.22) in patients with CD47-high (n = 26) versus those with CD47-low (n = 20) tumors, consistent with previous reports.7,8,9,10
There was also no observed benefit of evorpacept in the 32 patients who did not meet the definition of retained HER2+ disease, regardless of CD47 expression. These patients all enrolled with a HER2+ archival biopsy (but were ctDNA-negative and did not provide a fresh biopsy). Of these 32 patients, central assessment of their archival biopsy sample by IHC found that one was HER2-negative, 26 were HER2+, four were IHC2+ but in situ hybridization indeterminate (HER2 status unknown), and one only had a local assessment result. This classification should not be interpreted as indicating confirmed HER2 tissue loss, as post-HER2-directed treatment tissue confirmation was unavailable in these 32 patients. The ORR for these patients was 18.8% for evorpacept plus TRP versus 31.3% for TRP alone (Extended Data Table 5 and Supplementary Table 5). In the subset of patients without evidence of retained HER2+ disease who were CD47-high (n = 15), response rates were similar for evorpacept plus TRP and TRP alone (25.0% versus 20.0%).
Discussion
In this mature analysis of the phase 2 part of the randomized ASPEN-06 study, evorpacept in combination with TRP demonstrated promising efficacy in patients with HER2-overexpressing, advanced or metastatic gastric/GEJ cancer in the second- or third-line setting. In the primary analysis ITT population, the investigator-assessed ORR was 40.3% (95% CI 28.1–49.3) with evorpacept plus TRP and 26.6% (95% CI 16.3–39.1) in the control arm receiving TRP alone.
Consistent with the mechanistic rationale that evorpacept augments the ADCP activity of trastuzumab, post hoc analyses suggest that the clinical benefit of evorpacept plus TRP was strongly enriched in patients whose tumors had both retained HER2+ disease and expressed high levels of CD47. In this subgroup, ORR was 63.6% with evorpacept plus TRP versus 23.1% with TRP alone, median DOR was 25.5 versus 8.4 months, and the HR for PFS (0.38) favored the evorpacept plus TRP arm. Although subgroups were small, there was no evidence that adding evorpacept to TRP treatment was beneficial in patients without evidence of retained HER2+ disease, regardless of CD47 expression levels, or in patients with low CD47 expression despite evidence of retained HER2 positivity.
As evorpacept has the unique design of an inactive Fc domain, prophagocytic FcyR binding is a function of combination therapy with an anticancer antibody. The role of HER2 expression is well characterized for trastuzumab in the first-line treatment of gastric/GEJ cancer17; however, the level of HER2 expression required for clinical benefit is not known in the context of evorpacept-based combination therapy. Nevertheless, diminished efficacy may be expected in patients with loss of HER2 expression resulting from acquired resistance to previous HER2-targeted therapy21. This hypothesis is supported by the greater ORR difference between evorpacept plus TRP and TRP alone in the prespecified subgroup of patients with HER2+ status based on a fresh biopsy, relative to the ITT population. Given the challenges of obtaining fresh biopsy specimens in later line settings, baseline blood samples were collected from all patients to evaluate ERBB2 gene amplification in ctDNA as a more feasible approach for identifying retained HER2 positivity. Analyses of subgroups with ERBB2 gene amplification or a fresh HER2+ biopsy similarly identified patients more likely to benefit from evorpacept plus TRP compared to the overall ITT population.
The improvement in treatment response and durability observed with evorpacept plus TRP versus TRP alone supports targeting the CD47–SIRPα axis to enhance the antitumor efficacy of ADCP-based therapies. In addition, findings from the fresh biopsy or ctDNA HER2+ subgroup indicate that retained HER2+ disease is important for trastuzumab’s antitumor activity, including ADCP, in the context of an evorpacept combination, and supports the rationale for combining evorpacept with trastuzumab in additional HER2-overexpressing settings. There is also potential for combining evorpacept with other tumor-targeting antibodies that have ADCP functionality across multiple indications.
Having demonstrated the relevance of the partner anticancer antibody to the combinatorial mechanism of action of evorpacept, further analysis was undertaken to evaluate the relationship between baseline tumor CD47 expression and antitumor activity. CD47 is well documented to be upregulated in tumor tissue as a resistance mechanism to evade immune surveillance1,22,23, and has generally been considered a ubiquitous target that does not require patient selection. More recently, the potential for improved efficacy with the anti-CD47 antibody magrolimab was observed in subsets of patients with high CD47 expression in populations with higher risk myelodysplastic syndrome24 or metastatic non-small cell lung cancer25 but the clinical development of magrolimab was later discontinued. In addition, recent data from preclinical and clinical studies demonstrate that CD47 protein expression varies widely across different tumor types and indicate that the activity of optimized CD47 blockers may be linked to CD47 expression levels26. The post hoc analysis from ASPEN-06 suggests that CD47 protein expression is a potential predictive biomarker for evorpacept efficacy in advanced gastric/GEJ cancer. A candidate cutoff of ≥5% IHC3+ CD47 tumor membrane expression (CD47-high) was identified from a range of cutoffs, and the findings provide preliminary evidence of clinical benefit when adding evorpacept to TRP in CD47-high tumors with evidence of retained HER2+ disease. Hence, a CD47-high cutoff may be a valuable addition to future studies of evorpacept in gastric/GEJ cancer, and the predictive potential of higher CD47 expression should also be considered for other tumor types and anticancer antibody combination regimens; however, CD47 IHC scoring, which was performed and analyzed centrally in the present study, may need to undergo additional validation to enable integration into routine clinical practice. It should be noted that the specific design of evorpacept with an inactive Fc, unlike a traditional anti-CD47 antibody with effector function, allows for higher dosing that may be needed to fully block high tumor levels of CD47 without affecting other CD47 expressing cells, such as hematologic cells. Finally, consistent with previous research and the biological implication of tumor cells co-opting the CD47–SIRPα pathway for immune evasion, higher CD47 protein expression had a trend for worse OS in the TRP alone arm7,8,9,10.
The efficacy of evorpacept in combination with TRP in this study compares favorably with data for RP from previous clinical trials (RAINBOW and DESTINY-Gastric04) in similar second-line gastric/GEJ cancer settings20,27. In the randomized, double-blind, phase 3, RAINBOW study, an ORR of 28% was reported with second-line RP27. Although patients in RAINBOW were enrolled regardless of HER2 status, a similar ORR of 29% was reported with second-line RP in patients with HER2+ metastatic gastric/GEJ cancer in the DESTINY-Gastric 04 trial, in which all patients had retained HER2+ disease confirmed by a biopsy taken after previous HER2-directed therapy20. In the present study, a similar ORR was observed for the internal control arm of TRP alone (26.6%), providing support that the addition of evorpacept was driving the benefit seen in the evorpacept plus TRP arm.
The incidence of TEAEs (any grade or grade ≥3) for evorpacept plus TRP was similar to TRP alone. While hematological TEAEs were somewhat more common with evorpacept plus TRP versus TRP alone, there was no increase in febrile neutropenia and treatment discontinuation rates were similar, indicating that these events were manageable with standard supportive care. To our knowledge, this is the first randomized study to demonstrate both clinical activity of CD47–SIRPα inhibition and a manageable safety profile in solid tumors. The manageable safety and favorable tolerability profiles observed with evorpacept plus TRP are also consistent with previous observations of evorpacept-based regimens15,28. This validates the rationale for designing evorpacept with an inactive Fc domain for improved tolerability alongside the high-affinity CD47-blocking domain to enhance the clinical efficacy with combination partners, and differentiates evorpacept from previous anti-CD47 agents with an active Fc domain11,13,15,29.
Limitations that should be considered when interpreting the data include the open-label design, limited sample size particularly in the subgroups analyzed, relatively low proportion of patients who had undergone fresh biopsy to confirm HER2+ status, exploratory use of ctDNA for defining HER2 status and the post hoc nature of the analyses of CD47 status. With regard to the CD47 analyses, the consistency of benefit seen across efficacy end points with evorpacept plus TRP versus TRP alone, irrespective of the CD47 cutoffs used, provides increased confidence in the finding that CD47 expression is potentially predictive for the benefit of evorpacept; however, the risk of overfitting when a biomarker threshold is derived and evaluated within the same dataset is also acknowledged.
In conclusion, this analysis of the phase 2 part of the ASPEN-06 study suggests that evorpacept in combination with TRP may have clinical efficacy as second- or third-line treatment for patients with advanced gastric/GEJ cancer with retained HER2+ disease. Post hoc analyses suggest that this benefit may be greater in patients with high CD47 expression. This is consistent with the hypothesis that both CD47 expression and retained HER2+ disease would be necessary to drive efficacy through enhanced ADCP with the combination of evorpacept and trastuzumab. Indeed, while acknowledging the limitations of post hoc analyses, the beneficial effect of adding evorpacept to TRP seems to be enhanced in this patient population, and the magnitude of benefit observed suggests the critical importance of studying both expression of CD47 and retained HER2+ disease as predictive factors in future studies. Although this study did not proceed to phase 3 for strategic reasons, further investigation and confirmation of this beneficial signal with evorpacept in larger controlled studies with prospective patient selection based on CD47-high expression and retained HER2+ disease is warranted in patients with advanced gastric/GEJ cancer and other indications, including advanced HER2+ breast cancer. Given the emergence of trastuzumab deruxtecan as second-line therapy in advanced gastric/GEJ cancer, future development of evorpacept may be more relevant in earlier-line settings or in combination with next-generation Fc-competent HER2-directed antibodies such as zanidatamab.
Methods
Study conduct and oversight
The study was conducted in accordance with all relevant local/national regulatory requirements, the International Ethical Guidelines for Biomedical Research Involving Human Subjects, and the Declaration of Helsinki. All research was approved by the relevant institutional review boards or ethics committees at each participating center (Supplementary Table 6). All patients provided written informed consent before study participation and were not compensated. A safety monitoring committee consisting of select investigators and sponsor representatives was implemented for the phase 2 portion of the study. The safety monitoring committee oversaw study conduct until completion of the study analyses, provided safety oversight and conducted a review of planned interim analyses for futility.
Study design
ASPEN-06 (NCT05002127) is an international, multicenter, randomized phase 2/3 study investigating the efficacy and safety of evorpacept in combination with TRP for the treatment of patients with HER2+ advanced or metastatic gastric/GEJ cancer in the second- or third-line setting (Extended Data Fig. 5). In the phase 2, open-label part of the study reported here, 122 patients were planned to be randomized 1:1 to receive evorpacept plus TRP or TRP alone, with a one-arm, two-stage Simon design adopted in the evorpacept plus TRP arm. Randomization was conducted using stochastic minimization based on the variance method30, balancing baseline characteristics: previous use of trastuzumab deruxtecan (yes, no); line of therapy (second or third); geographical region (Asia, non-Asia); type of cancer (gastric, GEJ adenocarcinoma); and HER2 overexpression (IHC3+, IHC2+/in situ hybridization+), together with post-trastuzumab tissue sample (yes, no). The study had two primary objectives: (1) to assess the effect of evorpacept plus TRP on ORR compared to a historical control of RP27; and (2) to assess the contribution of evorpacept to a regimen of TRP, as measured by the difference in ORR between the two randomized treatment arms. The study protocol and statistical analysis plan are available in the Supplementary Information. The study protocol was updated with four amendments. These amendments clarified the study procedures, inclusion and exclusion criteria and patient management; changes did not alter study end points or overall study conduct. Specific details on the amendments are available in Supplementary Table 7.
Patients
Eligible patients were adults (≥18 years old) who had provided informed consent with HER2+ advanced or metastatic gastric/GEJ adenocarcinoma that had progressed on or after a previous HER2-directed anticancer agent (for example, trastuzumab) and/or fluoropyrimidine-containing/platinum-containing chemotherapy. Patients were required to have at least one measurable lesion as defined by Response Evaluation Criteria in Solid Tumors (RECIST) v.1.1, adequate bone marrow function (absolute neutrophil count ≥1,500 mm−3, platelets ≥100,000 mm−3 and hemoglobin ≥9 g dl−1), adequate renal function (estimated creatinine clearance ≥30 ml min−1 (by Cockcroft–Gault equation) and serum creatinine ≤1.5× upper limit of normal (ULN)), adequate liver function (total bilirubin ≤1.5× ULN (≤3.0× ULN if the patient has documented Gilbert syndrome); aspartate transaminase and alanine transaminase ≤3.0× ULN; ≤5.0× ULN if there is liver involvement secondary to tumor), QTcF interval of ≤480 ms (based upon mean value from triplicate ECGs), prothrombin time (international normalized ratio) < 1.5× ULN and a partial thromboplastin time ≤ 5 s above the ULN (unless receiving anticoagulation therapy), resolved acute effects of any previous therapy to baseline severity or grade ≤1 NCI CTCAE v.5.0 except for adverse events not constituting a safety risk by Investigator judgment, available biopsy sample before study entry in the most recent tissue sample used to determine HER2-overexpressing gastric/GEJ cancer, negative serum pregnancy test (for female participants of childbearing potential) at screening, agreement from male and female patients of childbearing potential to use a highly effective method of contraception throughout the study and for at least 120 days after the last dose of evorpacept, at least 7 months after the last dose of trastuzumab (or biosimilars), at least 3 months after the last dose of ramucirumab, or at least 6 months after the last dose of paclitaxel, whichever is later, and an ECOG PS score of 0 or 1.
Patients were excluded if they had symptomatic central nervous system involvement or leptomeningeal disease; previous radiotherapy within 2 weeks of start of study treatment; intolerance to or who have had a severe allergic or anaphylactic reaction to antibodies or infused therapeutic proteins or previous severe allergic or anaphylactic reaction to any of the substances included in the study drug (including excipients); any experimental antibodies or live vaccines in the last 28 days before the first dose of study drug, unless approved by the Sponsor; previously received ramucirumab or any anti-CD47 or anti-SIRPα agent; received systemic anticancer therapy within 4 weeks of starting study treatment; received a blood product transfusion within 14 days of starting study treatment; or a history of autoimmune disorders, autoimmune hemolytic anemia or autoimmune thrombocytopenia or gastrointestinal perforation/fistula; any grade 3–4 gastrointestinal bleeding or history of deep vein thrombosis, pulmonary embolism, arterial thrombosis or any other significant thromboembolism within 3 months before cycle 1 day 1; cirrhosis at a level of Child–Pugh B (or worse) or cirrhosis (any degree) and a history of hepatic encephalopathy or clinically meaningful ascites resulting from cirrhosis; chronic antiplatelet therapy, including dipyridamole or clopidogrel, or similar agent; active, uncontrolled, clinically significant bacterial, fungal or viral infection, including hepatitis B, hepatitis C, SARS-CoV-2, known HIV or AIDS-related illness; any of the following in the previous 6 months: myocardial infarction, unstable angina, coronary/peripheral artery bypass graft/coronary artery stenting, New York Heart Association class II or greater congestive heart failure, cerebrovascular accident or transient ischemic attack; poorly controlled hypertension (>160 mm Hg systolic or >100 mm Hg diastolic for >4 weeks) despite standard medical management within 28 days before the first dose of study drug; any major surgery or serious or nonhealing wound, ulcer or bone fracture within 28 days or minor surgery/subcutaneous venous access device placement within 7 days before enrollment; current active treatment in another interventional therapeutic clinical study; treatment for any other malignancy within the last 3 years before enrollment except for adequately treated non-melanomatous skin cancer, carcinoma in situ, or other early-stage malignancies that have undergone potentially curative therapy; other severe acute or chronic medical or psychiatric condition, including recent (within the past year) or active suicidal ideation or behavior, or laboratory abnormality that may increase the risk associated with study participation or investigational product administration or may interfere with the interpretation of study results and, in the judgment of the Investigator, would make the patient inappropriate for entry into this study; patients who are pregnant or breastfeeding. Further information on eligibility criteria is provided in the Supplementary Information.
Treatments
Eligible patients received either evorpacept plus TRP or TRP alone until disease progression, unacceptable toxicity, study withdrawal, or death. Evorpacept (30 mg kg−1), trastuzumab (initial dose of 6 mg kg−1; 4 mg kg−1 doses thereafter), and ramucirumab (8 mg kg−1) were administered every 2 weeks; paclitaxel (80 mg m−2) was administered every week for the first 3 weeks of every 28-day cycle. All treatments were administered intravenously in the outpatient setting. Patients were observed for at least 2 h after the initial infusion of evorpacept and as clinically indicated thereafter.
No additional anticancer treatments, live vaccines or chronic systemic corticosteroids were allowed during the study. Palliative and supportive care for disease-related symptoms were permitted at the investigator’s discretion. Further information on study drug administration, permitted medications and patient-monitoring are provided in the Supplementary Information.
Biomarkers
For all patients, eligible HER2+ status (per American Society of Clinical Oncology/College of American Pathologists guidelines31) was determined locally (if available) or centrally during screening using a US Food and Drug Administration-approved test for gastric/GEJ cancer on the most recent tissue sample. Subgroup analyses of efficacy are based on the HER2 status recorded in the electronic data capture system before enrollment; however, patients who enrolled with HER2 status determined locally were also retrospectively re-evaluated by central assessment. It was recommended that HER2 status be determined in a biopsy obtained after previous trastuzumab treatment, if safe and feasible. Biopsies from patients whose HER2 overexpression was determined using a post-trastuzumab tissue sample were deemed to be ‘fresh’. As fresh biopsies were only available for a subset of patients, and loss of HER2 positivity has been observed after previous HER2-directed therapy21, amplification of the ERBB2 gene (encoding HER2) in ctDNA was also evaluated. ctDNA was extracted from plasma samples collected on day 1 of cycle 1 before dosing and was assessed for ERBB2 gene amplification utilizing Guardant360 comprehensive genome profiling (Guardant Health). The reportable range for plasma ERBB2 gene amplification detected by the Guardant360 Research Use Only assay in this study is ≥2.18 copies.
CD47 protein expression in the most recent tumor tissue sample was evaluated by IHC (clone EPR21794, Abcam, cat. no. ab218810; lot no. 1058049-2) at IQVIA Laboratories. Samples were manually scored for CD47 expression by a pathologist according to the percent of tumor cells with membrane staining intensities of 0, 1+, 2+ and 3+.
Statistical analyses
With an estimated phase 2 sample size of 61 patients in the evorpacept plus TRP arm, the two-stage Simon design study had 80% power to detect an ORR of >30% at a one-sided type 1 error rate of 2.5% if the true ORR was 47.5% adjusted by the assumed dropout of 5%. Considering a difference of 8% in ORR between evorpacept plus TRP and TRP alone as a clinically meaningful contribution of evorpacept, a sample size of 122 would also provide 68% power to detect this clinically meaningful effect if the true ORR of TRP is 35% at a one-sided type 1 error rate of 18%. For the fresh biopsy HER2+ subgroup, with an estimated sample size of 31 patients per arm, this two-stage Simon design study had 50% power to detect an ORR of >30% at a one-sided type 1 error rate of 2.5% if the true ORR was 47.5% adjusted by the assumed dropout of 5%. For the comparisons to the assumed historical control ORR in the primary analysis, uniformly minimum variance unbiased estimator was applied. The one-sided P value refers to the comparison between evorpacept + TRP and historic control and is adjusted based on actual sample size in Simon’s two-stage design according to Koyama and Chen32. The contribution of evorpacept was assessed similarly in this subgroup: an estimated sample size of 31 patients per arm provided 59% power to detect a clinically meaningful effect (9.7%) if the true ORR of TRP is 35% at a one-sided type 1 error rate of 23%.
The primary efficacy end point was investigator-assessed ORR as per RECIST v.1.1. Secondary end points included ORR by BICR, DOR and PFS by investigator or BICR, OS and the frequency and severity of adverse events. Exploratory end points included molecular analysis in tumor biopsy and peripheral blood samples before treatment, and correlation of IHC, response and gene alterations.
The primary end point and key efficacy end points were evaluated in two prospectively defined patient populations: the ITT population and a subgroup with HER2 overexpression by a tissue sample obtained at any time following the completion of a previous HER2-directed therapy and before the start of study treatment (fresh biopsy HER2+ subgroup). A subgroup of patients with ERBB2 gene amplification by ctDNA (the ctDNA HER2+ subgroup) and a subgroup with either HER2+ in a fresh biopsy or ERBB2 gene amplification by ctDNA (fresh biopsy or ctDNA HER2+ subgroup) were also analyzed on an exploratory basis. The exploratory evaluation of CD47 as a potential predictive or prognostic biomarker was performed in the fresh biopsy or ctDNA HER2+ subgroup. Safety analyses were performed in the safety population (all enrolled patients who received at least one dose of study drug).
For ORR per investigator or BICR, the proportion of patients with the best confirmed overall response (complete or partial response) was reported along with 95% CIs by treatment arm. Time-to-event outcomes (DOR, PFS and OS) were analyzed using Kaplan–Meier curves and summary statistics (number of events, median and quartiles with 95% CIs) are presented. HRs were calculated from a stratified Cox regression model, with line of therapy and post-trastuzumab tissue sample as stratification factors in the ITT and fresh biopsy HER2+ subgroups, and HRs were calculated from Cox regression models with stratification factors of region (Asia yes or no) and use of previous trastuzumab deruxtecan (yes or no) as covariates in the retained HER2+ population and CD47-high and low subpopulations.
The potential for tumor CD47 expression to serve as a predictive or prognostic biomarker was evaluated by performing ROC analyses for investigator-assessed ORR. Standard ROC curves were generated for CD47 expression at varying intensities (IHC1+2+3+, IHC2+3+ and IHC3+) by treatment33,34. In addition, the Youden index was plotted across all practical cutoffs of CD47 expression to determine their effectiveness with respect to specificity and sensitivity and identify potential cutoff points. The cutoffs identified by ROC and Youden index analyses were further examined for clinical benefit in evorpacept plus TRP compared to TRP alone per maximum likelihood approach using logistic regression models for ORR and Cox proportional hazards models for PFS and OS with the interaction between treatment and CD47 status included; adjustments were also made for region and previous trastuzumab deruxtecan as baseline characteristics as covariates, in which the cutoff value was treated as an unknown parameter in a statistical model. The maximum log-likelihood across the range of CD47 cutoffs via grid search was then refined to the candidate cutoff for classifying CD47-high versus CD47-low35. In addition, the treatment interaction with CD47 status was further tested. The analyses conducted in this subgroup were post hoc and P values were descriptive.
The data were analyzed using three data cutoffs as additional biomarker data became available, and the analyses from the latter two data cutoffs were exploratory. The primary analysis of ORR was conducted using a data cutoff of 24 May 2024, to determine whether enrollment should continue into the phase 3 portion of the study. Updated efficacy and safety results when HER2 ctDNA data became available at a data cutoff of 2 December 2024, have been presented previously.36 This article reports results of the primary analysis for a data cutoff of 24 May 2024, as well as an updated efficacy analysis, safety data and post hoc analyses of HER2 and CD47 biomarker data with a data cutoff of 15 May 2025.
In this study, biological sex of patients was not ascertained; however, gender (self-identified) was collected as a demographic variable and recorded in the electronic case report form. Gender-based treatment effects were evaluated, although the study was not designed to assess gender-based treatment effects, and gender-stratified efficacy analyses were not prespecified.
Protocol deviations
Major protocol deviations occurred and were tracked during the study. The deviation types reported included study procedures, safety, eligibility and informed consent in both treatment arms. Protocol deviations did not affect the safety of study participants or integrity of study results.
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
Data availability
Qualified researchers may request access to study information from phase 2–4 trials sponsored by ALX Oncology that supports the methods and findings reported in this manuscript. This includes access to anonymized participant-level data, as well as other information (for example protocols, clinical study reports, analysis plans), once the product and indication have been approved in the USA, the European Union and other designated markets. Requests for access should be accompanied by a research proposal, including an analysis plan and publication plan. Upon approval, information will be made available following execution of a data-sharing agreement. Requests should be submitted to info@alxoncology.com. The estimated timeframe for response will be within 60 days.
Code availability
The ASPEN-06 trial was conducted using the electronic case report form. Study data were collected and managed using the electronic data capture system called ID-Base/Marvin, v.1.0. Statistical analyses were performed using SAS v.9.4 and R v.4.5.1 (R core team, 2025).
References
Willingham, S. B. et al. The CD47-signal regulatory protein alpha (SIRPa) interaction is a therapeutic target for human solid tumors. Proc. Natl Acad. Sci. USA 109, 6662–6667 (2012).
Petrova, P. S. et al. TTI-621 (SIRPαFc): A CD47-blocking innate immune checkpoint inhibitor with broad antitumor activity and minimal erythrocyte binding. Clin. Cancer Res. 23, 1068–1079 (2017).
Lu, Q. et al. Potential new cancer immunotherapy: anti-CD47-SIRPα antibodies. Onco. Targets Ther. 13, 9323–9331 (2020).
Jiang, Z. et al. Targeting CD47 for cancer immunotherapy. J. Hematol. Oncol. 14, 180 (2021).
Zhou, X., Liu, X. & Huang, L. Macrophage-mediated tumor cell phagocytosis: opportunity for nanomedicine intervention. Adv. Funct. Mater. 31, 2006220 (2021).
Liu, Y. et al. Emerging phagocytosis checkpoints in cancer immunotherapy. Signal Transduct. Target Ther. 8, 104 (2023).
Zhao, H.-J. et al. Prognostic significance of CD47 in human malignancies: a systematic review and meta-analysis. Transl. Cancer Res. 7, 609–621 (2018).
Shi, M. et al. CD47 expression in gastric cancer clinical correlates and association with macrophage infiltration. Cancer Immunol. Immunother. 70, 1831–1840 (2021).
Ye, Y. et al. Survival and clinicopathological significance of CD47 in human solid tumors: an updated systematic reviews and meta-analysis. Cancer Rep. 8, e70296 (2025).
Yoshida, K. et al. CD47 is an adverse prognostic factor and a therapeutic target in gastric cancer. Cancer Med. 4, 1322–1333 (2015).
Advani, R. et al. CD47 blockade by Hu5F9-G4 and rituximab in non-Hodgkin’s lymphoma. N. Engl. J. Med. 379, 1711–1721 (2018).
Liu, J. et al. Pre-clinical development of a humanized anti-cd47 antibody with anti-cancer therapeutic potential. PLoS ONE 10, e0137345 (2015).
Kauder, S. E. et al. ALX148 blocks CD47 and enhances innate and adaptive antitumor immunity with a favorable safety profile. PLoS ONE 13, e0201832 (2018).
Osorio, J. C., Smith, P., Knorr, D. A. & Ravetch, J. V. The antitumor activities of anti-CD47 antibodies require Fc-FcγR interactions. Cancer Cell 41, 2051–2065.e6 (2023).
Lakhani, N. J. et al. Evorpacept alone and in combination with pembrolizumab or trastuzumab in patients with advanced solid tumours (ASPEN-01): a first-in-human, open-label, multicentre, phase 1 dose-escalation and dose-expansion study. Lancet Oncol. 22, 1740–1751 (2021).
Lee, K. W. et al. Evorpacept (ALX148), a CD47 myeloid checkpoint inhibitor, in patients with head and neck squamous cell carcinoma (HNSCC) and with gastric/gastroesophageal cancer (GC); ASPEN-01. J. Immunother. Cancer 9, A530 (2021).
Bang, Y.-J. et al. Trastuzumab in combination with chemotherapy versus chemotherapy alone for treatment of HER2-positive advanced gastric or gastro-oesophageal junction cancer (ToGA): a phase 3, open-label, randomised controlled trial. Lancet 376, 687–697 (2010).
Janjigian, Y. Y. et al. Pembrolizumab plus trastuzumab and chemotherapy for HER2-positive gastric or gastro-oesophageal junction adenocarcinoma: interim analyses from the phase 3 KEYNOTE-811 randomized placebo-controlled trial. Lancet 402, 2197–2208 (2023).
Scheck, M. K., Hofheinz, R. D. & Lorenzen, S. HER2-positive gastric cancer and antibody treatment: state of the art and future developments. Cancers 16, 1336 (2024).
Shitara, K. et al. Trastuzumab deruxtecan or ramucirumab plus paclitaxel in gastric cancer. N. Engl. J. Med. 24, 336–348 (2025).
Saeki, H. et al. Re-evaluation of HER2 status in patients with HER2-positive advanced or recurrent gastric cancer refractory to trastuzumab (KSCC1604). Eur. J. Cancer 105, 41–49 (2018).
Oldenborg, P. A. CD47: a cell surface glycoprotein which regulates multiple functions of hematopoietic cells in health and disease. ISRN Hematol. 2013, 614619 (2013).
Chao, M. P. et al. Anti-CD47 antibody synergizes with rituximab to promote phagocytosis and eradicate non-Hodgkin lymphoma. Cell 142, 699–713 (2010).
Johnson, L. D. S. & Zhang, Y. Impact of mutational landscape and CD47 levels in the magrolimab + azacitidine vs placebo + azacitidine trial in patients with untreated higher risk myelodysplastic syndrome. Blood 144, 4604 (2024).
Van Buuren, N. et al. 184 Biomarker analysis of magrolimab plus docetaxel in patients with 2L+ mNSCLC from ELEVATE-Lung and UC, a phase 2 multicohort study. J. Immuno.Ther. Cancer 12, A212 (2024).
Chen, C., Lu, F., Huang, H. & Pan, Y. Translating CD47-targeted therapy in gastrointestinal cancers: Insights from preclinical to clinical studies. iScience 27, 111478 (2024).
Wilke, H. et al. Ramucirumab plus paclitaxel versus placebo plus paclitaxel in patients with previously treated advanced gastric or gastro-oesophageal junction adenocarcinoma (RAINBOW): a double-blind, randomised phase 3 trial. Lancet Oncol. 15, 1224–1235 (2014).
Kim, T. M. et al. Evorpacept plus rituximab for the treatment of relapsed or refractory non-Hodgkin lymphoma: results from the phase I ASPEN-01 study. Haematologica 110, 2102–2112 (2025).
Ansell, S. M. et al. Phase I study of the CD47 blocker TTI-621 in patients with relapsed or refractory hematologic malignancies. Clin. Cancer Res. 27, 2190–2199 (2021).
Pocock, S. J. & Simon, R. Sequential treatment assignment with balancing for prognostic factors in the controlled clinical trial. Biometrics 31, 103–115 (1975).
Bartley, A. N. et al. HER2 testing and clinical decision making in gastroesophageal adenocarcinoma: guideline from the College of American Pathologists, American Society for Clinical Pathology, and the American Society of Clinical Oncology. J. Clin. Oncol. 35, 446–464 (2017).
Koyama, T. & Chen, H. Proper inference from Simon’s two-stage designs. Statist. Med. 27, 3145–3154 (2008).
Polley, M. C. & Dignam, J. J. Statistical considerations in the evaluation of continuous biomarkers. J. Nucl. Med. 62, 605–611 (2021).
Emancipator, K. et al. Comparing programmed death ligand 1 scores for predicting pembrolizumab efficacy in head and neck cancer. Mod. Pathol. 34, 532–541 (2021).
Hui, J. & Guo, W. Optimal biomarker cutoff identification and validation. In Statistics in Biosciences Vol. 14, 352–362 (Springer, 2022).
Shitara, K. et al. Final analysis of the randomized phase 2 part of the ASPEN-06 study: a phase 2/3 study of evorpacept (ALX148), a CD47 myeloid checkpoint inhibitor, in patients with HER2-overexpressing gastric/gastroesophageal cancer (GC). J. Clin. Oncol. 43, A332 (2025).
Acknowledgements
The authors thank the patients who participated in the ASPEN-06 study, the investigators and the study-site personnel. Funding for this trial was provided by ALX Oncology Inc, USA, which participated in study design and conduct, data collection and analysis and preparation of the manuscript. The authors received no specific funding for this work. Under the direction of the authors, medical writing support was provided by T. Williamson of Twist Medical with funding from ALX Oncology, in accordance with Good Publication Practice (GPP 2022) guidelines.
Funding
This study was sponsored by ALX Oncology Inc, USA (no grant number), which participated in study design and conduct, data collection and analysis and preparation of the manuscript.
Author information
Authors and Affiliations
Contributions
K.S., Z.W., E.V.C., J.T. and K.-W.L. are principal investigators and contributed to study conception, design, data acquisition and quality control. K.S. had access to the raw data. C.C., C.D.L.F., J.L., S.Y.R. and Y.K.K. are principal investigators and contributed to data acquisition and quality control. A.F., C.M. and D.B. contributed to biomarker data acquisition and analysis. S.R., J.P., P.F. and A.C.T. contributed to clinical development, study conception and study design. D.B. and C.M. contributed to statistical analyses. All authors were involved in the drafting and/or critical revision of the manuscript and approved the final draft for submission.
Corresponding author
Ethics declarations
Competing interests
K.S. reports receiving personal fees for consulting and advisory roles from Bristol-Myers Squibb, Takeda, Ono Pharmaceutical, Novartis, Daiichi Sankyo, Amgen, Boehringer Ingelheim, Merck Pharmaceutical, Astellas, Guardant Health Japan, Janssen, AstraZeneca, Zymeworks Biopharmaceuticals, ALX Oncology, Bayer, GlaxoSmithKline K.K., HEALIOS K.K., Moderna and Arcus Biosciences; receiving honoraria from Bristol-Myers Squibb, Ono Pharmaceutical, Janssen, Eli Lilly, Astellas and AstraZeneca; and receiving research funding (all to institution) from Astellas, Ono Pharmaceutical, Daiichi Sankyo, Taiho Pharmaceutical, Chugai, Merck Pharmaceutical, Amgen, Eisai, PRA Health Sciences, Syneos Health, AstraZeneca, PPD-SNBL K.K. and TORAY. Z.W. reports consulting for Abbvie, Amgen, Alligator, AstraZeneca, Daiichi, Merck, BMS, Novartis and Revolution Medicine. J.T. reports personal financial interest in the form of scientific consultancy roles for Accent Therapeutics, Alentis Therapeutics, AstraZeneca, Boehringer Ingelheim, Bristol-Myers Squibb, Carina Biotech, Cartography Biosciences, Chugai, Daiichi Sankyo, F. Hoffmann-La Roche, Genentech, Ipsen Innovation SAS, Johnson & Johnson/Janssen, Kayak Therapeutics, Larkspur Biosciences, Lilly, Marengo Therapeutics, Menarini, Merus, MSD, Novartis, One-carbon Therapeutics, Ono Pharma USA, Peptomyc, Pfizer, Pierre Fabre, Quantro Therapeutics, Scandion Oncology, Scorpion Therapeutics, Servier, Sotio Biotech, Syntelios AG, Taiho, Takeda Oncology, Theriva Biologics and Tolremo Therapeutics; and stocks from Adualys Therapeutics, Alentis Therapeutics, Oniria Therapeutics, 1TRIALSP and Pangaea Oncology. E.V.C. reports participation in advisory boards for Abbvie, Agenus, ALX, Amgen, Arcus Biosciences, Astellas, AstraZeneca, Bayer, Beigene, Bexon Clinical, Biontech, Boehringer Ingelheim, Bristol-Myers Squibb, Cantargia, Daiichi, Debiopharma, Elmedix, Eisai, Fosum, Galapagos, GSK, Ipsen, Iteos, Lilly, Merck Sharp & Dohme, Merck KGaA, Microbial Machines, Mirati, Novartis, Nordic, Novocure, Pierre Fabre, Pfizer, Sanofi, Seattle Genetics, Servier, Simcere, Takeda, Taiho and Trishula. C.C. reports receiving personal fees for consulting and advisory roles from MSD Oncology, Merck Serono, Daiichi Sankyo, AstraZeneca, Bristol-Myers Squibb, Jazz Pharmaceuticals, Gilead, Astellas, Pierre Fabre, Servier, and Amgen; and receiving honoraria from Servier, Amgen, Merck Serono, Bristol-Myers Squibb, Daiichi Sankyo, Merck Sharp & Dohme, Merck Serono and AstraZeneca. C.D.L.F. reports receiving personal fees for consulting and advisory roles from Agenus, Amgen, Astellas, AstraZeneca, Bayer, Beigene, Bristol-Myers Squibb, Daichi Sankyo, Eisai, Gilead, Ipsen, Jazz Pharmaceuticals, Lilly, Merck, MSD, Natera, Pierre Fabre Oncologie, Roche, Servier and Summit Therapeutics. J.L. reports nonfinancial interests from the following roles: steering committee member for AstraZeneca and Seattle Genetics; Principal Investigator for AstraZeneca, BMS, Daiichi Sankyo, Leaptherapeutics and Merck MSD; Project Lead for Genome and Company, Oncxerna and Samsung Bioepis; advisory role for Mirati Therapeutics; AP council member for ASCO; and KSMO member. P.F., A.F., J.P., A.C.T., C.M. and S.R. were employed by ALX Oncology at the time of the study and report stock and other ownership interests at ALX Oncology. D.B. was consulting with ALX Oncology, Elevation Oncology, Avidicure, Stork Therapeutics, Firefly Bio and Boxer Capital; is employed part-time by Intelligencia; and reports stock and other ownership interests in ALX Oncology, Enliven Therapeutics and Intelligencia. J.P. and S.R. were on the board of directors at ALX Oncology at the time of the study. J.P. also reports a leadership or fiduciary role at Tallac Therapeutics. K.-W.L. reports research support paid to his institution from ALX Oncology, AstraZeneca, Ono Pharmaceutical, Merck Sharp & Dohme, Merck KGaA, Roche, Pfizer, BeiGene, Leap Therapeutics, Zymeworks, Astellas, MacroGenics, Amgen, Seagen, Bolt Therapeutics, Trishula Therapeutics, Oncologie, Pharmacyclics, MedPacto, Green Cross Corp, ABL Bio, Y-Biologics, Daiichi Sankyo, Taiho Pharmaceutical, InventisBio, Elevar Therapeutics, MetaFines, Idience, Genome & Company and Exelixis; personal fees/honoraria from Ono Pharmaceutical, Boryung, Daiichi Sankyo, Astellas and Sanofi-Aventis; and participation on a data Ssafety monitoring board or advisory board for ALX Oncology and MetaFines. S.Y.R. and Y.-K.K. declare no competing interests.
Peer review
Peer review information
Nature Medicine thanks Steven Maron, Ilyas Sahin and Qingzhao Yu for their contribution to the peer review of this work. Primary Handling Editors: Ulrike Harjes and Saheli Sadanand, in collaboration with the Nature Medicine team.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Extended data
Extended Data Fig. 1 Updated analysis of ORRs assessed by the investigator in the ITT population in the (a) evorpacept plus TRP and (b) TRP alone treatment arms. Waterfall plots are of the best objective responses.
Data cutoff: 15 May 2025. CI, confidence interval; Evo, evorpacept; F, fresh biopsy; ID, identification number; ITT, intent-to-treat; ORR, objective response rate; P, positive circulating tumor DNA; RECIST v1.1, Response Evaluation Criteria in Solid Tumors version 1.1; TRP, trastuzumab, ramucirumab, and paclitaxel.
Extended Data Fig. 2 Kaplan–Meier analyses of PFS assessed by investigator in the (a) ITT population, (b) the fresh biopsy HER2+ subgroup, (c) the ctDNA HER2+ subgroup, and (d) the fresh biopsy or ctDNA HER2+ subgroup in the updated analysis.
Data cutoff: 15 May 2025. CI, confidence interval; ctDNA, circulating tumor DNA; Evo, evorpacept; HER2+, human epidermal growth factor receptor 2-positive; HR, hazard ratio; ITT, intent-to-treat; m, median; PFS, progression-free survival; TRP, trastuzumab, ramucirumab, and paclitaxel.
Extended Data Fig. 3 Kaplan–Meier analyses of OS in the (a) ITT population, (b) the fresh biopsy HER2+ subgroup, (c) the ctDNA HER2+ subgroup, and (d) the fresh biopsy or ctDNA HER2+ subgroup in the updated analysis.
Data cutoff: 15 May 2025. CI, confidence interval; ctDNA, circulating tumor DNA; Evo, evorpacept; HER2+, human epidermal growth factor receptor 2-positive; HR, hazard ratio; ITT, intent-to-treat; m, median; OS, overall survival; TRP, trastuzumab, ramucirumab, and paclitaxel.
Extended Data Fig. 4 CD47 expression at different staining intensity thresholds (IHC1+2+3+, IHC2+3+, and IHC3+) based on confirmed ORR by investigator assessment shown in (a) ROC curves and (b) Youden Index plots; and (c) CD47 expression as a predictive or prognostic biomarker evaluated using maximum log-likelihood by CD47 IHC3+ cutoffs from 1 to 30 for ORR by investigator assessment,a PFS by investigator assessment,b and OSb.
aLogistic regression adjustment was used. bCox proportional hazards models were used. All models are adjusted for region and prior trastuzumab deruxtecan as baseline characteristics. AUC, area under the curve; Evo, evorpacept; IHC, immunohistochemistry; ORR, objective response rate; OS, overall survival; PFS, progression-free survival; ROC, receiver operating characteristic; TRP, trastuzumab, ramucirumab, and paclitaxel; YI, Youden Index.
Extended Data Fig. 5 Study design.
aTo reduce the risk of accidental imbalances, randomization was conducted using stochastic minimization based on the variance method2, balancing the following (prognostic) baseline characteristics: previous use of trastuzumab deruxtecan (yes, no); line of therapy (second or third); geographical region (Asia, non-Asia); type of cancer (gastric, GEJ adenocarcinoma); and immunohistochemistry (IHC3+, IHC2+/in situ hybridization+), together with post-trastuzumab tissue sample (yes, no). +, positive; 2 L, second line; 3 L, third line; DOR, duration of response; GEJ, gastroesophageal junction; HER2, human epidermal growth factor receptor 2; ITT, intent-to-treat; ORR, objective response rate; OS, overall survival; PFS, progression-free survival; R, randomization; SIRPα, receptor signal regulatory protein alpha; TRP, trastuzumab, ramucirumab, and paclitaxel. 1. Wilke, H., Muro, K., Van Cutsem, E., et al. Ramucirumab plus paclitaxel versus placebo plus paclitaxel in patients with previously treated advanced gastric or gastroesophageal junction adenocarcinoma (RAINBOW): a double-blind, randomized phase 3 trial. Lancet Oncol 15, 1224–1235 (2014). 2. Pocock, S. J. & Simon, R. Sequential treatment assignment with balancing for prognostic factors in the controlled clinical trial. Biometrics 31, 103–115 (1975).
Supplementary information
Supplementary Information (download PDF )
Supplementary Tables 1–7, Figs. 1–4, Study protocol and Statistical analysis plan
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/
About this article
Cite this article
Shitara, K., Wainberg, Z., Tabernero, J. et al. Evorpacept plus trastuzumab, ramucirumab and paclitaxel in HER2-positive gastric cancer: a randomized phase 2 trial. Nat Med (2026). https://doi.org/10.1038/s41591-026-04700-3
Received:
Accepted:
Published:
Version of record:
DOI: https://doi.org/10.1038/s41591-026-04700-3