Fewer players are still at the CD47 table.
Four months ago, another CD47 player dropped out. In Pfizer's Q1 2026 earnings and latest pipeline update, a CD47 drug, maplirpacept, was removed from the development list. This means that all of Pfizer's CD47 assets, acquired from Trillium for $2.3 billion, are now out.
Going forward, it's possible that more players will exit the field. A review of the latest semi-annual reports of listed pharmaceutical companies on the Hong Kong stock exchange reveals that some CD47 molecules that were once heavily promoted are now being mentioned less frequently in financial reports and promotional materials.
At present, for pharmaceutical companies that are still holding on, whether CD47 is still worth pursuing and how to move forward is a question that must be considered.
Fortresses Fall One by One
Over the past decade, the CD47 field has been continuously iterating and upgrading, but no molecule has yet to clearly prove itself in clinical settings.
Most notable is Gilead's magrolimab.
In 2020, Gilead acquired FortySeven for approximately $4.9 billion, bringing magrolimab into its portfolio, which was then the world's most advanced CD47 antibody. It had also shown a high overall response rate of up to 91% in early trials for MDS and AML when combined with azacitidine, generating significant expectations.
In terms of design, magrolimab uses an IgG4-type antibody, which reduces binding to Fc receptors, decreases non-specific activation, and utilizes a dose-escalation administration strategy, aiming to address the anemia issue caused by the high expression of CD47 on normal red blood cells.
However, after actually entering clinical research, it was found that the optimization of magrolimab was far from sufficient. The ENHANCE-3 study showed that in AML patients who had not previously received intensive chemotherapy and were treated with azacitidine + venetoclax, the addition of magrolimab did not improve overall survival and even increased the risk of death.
Based on safety risks and poor efficacy, Gilead submitted an application to discontinue magrolimab.
This is just the beginning, with 2025 being the darkest moment for CD47. The last strongholds of CD47, which have garnered the most attention overseas, have fallen one after another.
First in April, ALX Oncology announced that its CD47 inhibitor Evorpacept failed to meet primary endpoint goals in both Phase II trials, ASPEN-03 and ASPEN-04, when combined with Keytruda. Previously, the company had already terminated Evorpacept clinical trials for MDS and AML.
In June, Pfizer announced that the development of maplirpacept, a CD47-targeting drug it acquired, would be terminated due to difficulties in patient enrollment in its Phase II study for the treatment of relapsed/refractory diffuse large B-cell lymphoma, after another CD47 product, ontorpacept, had previously been discontinued from its pipeline.
The collapse of these "strongholds" once left the CD47 track in a state of confusion.
Nubia's 'Don't Eat Me' Slogan Not as Simple as it Sounds
The core dilemma of the CD47 target lies in the balance between toxicity and efficacy, which is difficult to achieve, and this is also the common reason for the failure of these heavyweight products.
Many molecules are being upgraded based on the CD47 dilemma, but the effect is not ideal.
For instance, Pfizer's maplirpacept attempted to reduce toxicity through SIRPα-IgG4 fusion protein, but the overly weakened binding affinity resulted in insufficient efficacy, leading to the adoption of high-dose administration.
In clinical settings, the maximum doses of ontopacept and maplirpacept were 2 mg/kg and 24 mg/kg, respectively. Although maplirpacept resulted in fewer ≥3-grade adverse events than ontopacept in clinical trials, it still led to a 30% incidence of thrombocytopenia.
The strategy of completely inactivating Fc is also not viable. ALX Oncology's evorpacept is a typical example, which retains the high-affinity CD47 binding structure while adopting an inactive Fc, with the goal of reducing Fc-mediated effects related to normal red blood cells, thereby improving the safety of traditional CD47 antibodies.
However, the safer design did not lead to better efficacy. The study of Evorpacept in combination with K drug for the treatment of head and neck squamous cell carcinoma also failed to meet the primary endpoint.
Arch Oncology's AO-176 is also a CD47 antibody targeting the IgG2 Fc region that does not activate ADCC but has direct cytotoxic effects, with minimal impact on hematological parameters in non-human primates and no observed transient anemia. However, interim data from the Phase I/II clinical trial of AO-176 showed that 33% of patients experienced thrombocytopenia, including 22% with Grade 3 transient thrombocytopenia, and 22% experienced anemia.
It is clear that under the "don't eat me" mechanism, the balance line between efficacy and toxicity has always been an insurmountable obstacle for CD47 drugs. As batch after batch of optimized molecules fail, the clinical value of CD47 is also repeatedly called into question.
Retreat or Advance
After Pfizer and other contenders fell one after another, the crucial question facing all the remaining players is whether to retreat or press on.
Currently, a number of players are still actively exploring CD47, and some positive signals have indeed emerged.
Evorpacept is a typical example of combination therapy. Because its Fc segment is inactivated, Evorpacept has almost no single-agent activity (ADCC, ADC, CDC and other effects) and cannot trigger macrophages to phagocytose tumors. Therefore, combination therapy is necessary, relying on other monoclonal antibodies or ADC drugs used in combination to provide the "eat me" activation signal and further exert immune effects.
In the randomized Phase II ASPEN-06 trial, evorpacept was combined with trastuzumab, ramucirumab, and paclitaxel to treat HER2-positive gastric cancer/gastroesophageal junction cancer that had previously received anti-HER2 treatment.
In the 2025 updated data, among confirmed HER2-positive patients, the ORR reached 48.9% versus 24.5% in the control arm, with mDOR of 15.7 months versus 9.1 months. Still, the FDA held back from any strong endorsement—after all, Enhertu looms as a formidable hurdle in this indication.
There are also many domestic players holding on.
Kangfang Bio's AK117 has also been optimized, positioning it as a new generation of CD47 antibodies, addressing issues such as red blood cell agglutination faced by traditional CD47 antibodies through molecular design.
In 2026, EHA announced the AK117-206 study, a randomized, double-blind, placebo-controlled Phase II trial, in which 60 newly diagnosed AML patients who were not suitable for intensive chemotherapy received AK117 (or placebo) + azacitidine + venetoclax. The data showed an ORR of 80.0% in the AK117 group, compared to 66.7% in the control group. The CRc rate was 56.7% versus 53.3% in the placebo group, with a duration of CRc of 10.4 months versus 6.5 months, demonstrating a significant advantage.
However, the details are still worth noting. The duration of CRc in the placebo group was 6.5 months, but in the clinical data for standard AML treatment (VIALE-A), the duration was 14.7 months, with the current control group being significantly weaker than in the past. Although this is based on differences in patient baseline conditions, the data still raises questions about the value of AK117.
Safety issues are also not to be ignored. In the study, the incidence rates of ≥3-grade treatment-emergent adverse events (TEAEs) in the two treatment groups were 96.7% and 93.3%, respectively, with 43.3% of patients in both groups reporting ≥3-grade anemia. This suggests that AK117 did not significantly exacerbate anemia, but also failed to eliminate the inherent hematologic toxicity burden of CD47-class drugs.
Kangfang Bio is currently conducting three Phase III clinical trials for the combination of AK117 and AK112, making it one of the fastest-moving players in the CD47 field.
In terms of progress, another comparable one is Yimeng Bio's IMM01, an IgG1 Fc fusion protein that reduces its binding affinity to red blood cells by introducing mutations in the D1 domain and removing glycosylation sites.
IMM01 in combination with azacitidine for the treatment of CMML showed a Phase II clinical ORR of 72.7% (compared to 37.5% in the control group), with a CR rate of 27.3%, significantly higher than the 8.3% in the control group.
On July 16, 2026, Eucure Biotech announced that the Phase III clinical trial of IMM01 in combination with azacitidine for the treatment of newly diagnosed CMML has completed enrollment of all 173 patients, with progress comparable to that of AK117.
The efforts of these adherents may bring some new hope to the CD47 field, or perhaps it will be another disappointment.
Don't Be Fooled by the 'Next-Gen PD-1' Narrative
The future of CD47 may be bright, but the road ahead is bound to be winding. So far, CD47 has taught us the most important lesson - not to be obscured by the "next-generation PD-1" narrative.
Following the success of PD-1, the industry has come to realize that tumor immunotherapy is not just about attacking, but also about applying the "brakes", which is also an important approach.
Research has found that like T cells, macrophages also have a "brake". As a result, the CD47 "don't eat me" signal has been developed into a new tumor treatment immune checkpoint. For a time, CD47 was seen as one of the most promising immune targets after PD-1.
But in hindsight, the immune system was still too complex.
Firstly, CD47 is not a simple immune "switch". Existing research indicates that CD47 does not merely transmit a "don't eat me" signal through SIRPα, but rather is widely present on the surface of human cells and interacts with multiple receptors and signaling pathways, including SIRPα, integrin, VEGFR2, CD36, and FAS.
Furthermore, a summary of multiple studies found that CD47 drugs induced tumor immune suppression in humans, which was far less effective than the antitumor activity they exhibited in animal models such as mice.
A recent 2026 study in Science also revealed the complexity of CD47 in hematological tumors.
A research team used human AML cells and human macrophages to conduct a genome-wide CRISPR screen, and found that CD47 had a pronounced inhibitory effect on phagocytosis by mouse macrophages, but this effect was highly limited in human macrophages.
Further research found that CD43 on the surface of AML cells, as well as its O-glycosylation and sialylation, is an important mechanism limiting the phagocytosis of human macrophages. This suggests that in the human system, the control of the "don't eat me" signal of macrophages may not be limited to just the CD47-SIRPα pathway.
Various studies have shown that CD47 is far more complex than previously imagined, and in the past, the unknown aspects of it were obscured by the narrative of being the "next-generation PD-1".
This may be the most perilous moment in innovative drug development, not when a target fails, but when the industry is overly optimistic about a target, to the point of ignoring the risks that it inherently possesses.