Phase 2 study of palbociclib plus retifanlimab in patients with advanced dedifferentiated liposarcoma Authors: Evan Rosenbaum, Kenneth Seier, Rodrigo Gularte-Mérida, Evan Seffar, Mark A Dickson, Viswatej Avutu, Lauren B Banks, Jason E Chan, Ping Chi, Mrinal M Gounder, Ciara M Kelly, Mary L Keohan, Robert G Maki, Sujana Movva, Damon Reed, Rhoena Desir, Matthew Biniakewitz, Jae-Mun Cho, Michele Duchemin, Joseph P Erinjeri, Robert A Lefkowitz, Andrew Koff, Samuel Singer, William Tap, Li-Xuan Qin, Sandra D’Angelo Published: 2026 ABSTRACT AbstractBackgroundThe cyclin-dependent kinase 4 and 6 (CDK4/6) inhibitor palbociclib delays disease progression in dedifferentiated liposarcoma (DDLPS) by inducing tumor cell quiescence or senescence, though most tumors ultimately progress. Preclinical data suggest CDK4/6 inhibition enhances intratumoral inflammation and may synergize with immune checkpoint inhibitors.MethodsThis non-randomized, open-label, phase 2 study was conducted at Memorial Sloan Kettering Cancer Center. Patients with metastatic or unresectable DDLPS, or those expected to benefit from systemic therapy prior to surgery, were eligible. Patients received palbociclib (125 mg orally, days 1–21 of a 28-day cycle) plus retifanlimab (500 mg intravenous flat dose every 4 weeks). The primary endpoint was to assess the best objective response rate (ORR) by Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1.Results30 patients were treated and evaluable. Median age was 61 years (range 36–81), 67% were male, and 63% were treatment-naive. The median follow-up was 14.8 months. ORR was 20% (95% CI 8% to 39%) and the clinical benefit rate was 53% (95% CI 34% to 72%). Median progression-free survival and overall survival were 4.8 (95% CI 1.71 to 15.7) and 27 months (95% CI 21.7 to not reached (NR)), respectively. Median duration of response was 18.4 months (95% CI 9.4 months to NR). Grade ≥3 treatment-related adverse events occurred in 61% of patients, including one Grade 4 neutropenia. There were no Grade 5 events. Retifanlimab and palbociclib were discontinued due to toxicity in 23% and 17% of patients, respectively. Tumor sequencing identified JUN amplification in 6 of 14 patients with progressive disease versus 2 patients with stable disease or partial response. Patients with progressive disease had a higher median copy number alteration burden compared with those with stable disease or partial response.ConclusionsPalbociclib plus retifanlimab demonstrated deep and durable responses in a subset of patients with advanced DDLPS, with an ORR exceeding that historically observed with either agent alone. Adverse events were generally manageable. Copy number alteration burden and JUN amplification merit further evaluation as potential biomarkers of resistance.Trial registration numberhttps://clinicaltrials.gov/study/NCT04438824.Keywords: Immunotherapy, Combination therapy, Solid tumor WHAT IS ALREADY KNOWN ON THIS TOPIC Preclinical studies demonstrate that cyclin-dependent kinase 4 and 6 inhibition enhances intratumoral inflammation and antitumor immunity when combined with programmed cell death protein 1 blockade, though clinical development in carcinomas was discontinued due to prohibitive toxicity. Preclinical studies demonstrate that cyclin-dependent kinase 4 and 6 inhibition enhances intratumoral inflammation and antitumor immunity when combined with programmed cell death protein 1 blockade, though clinical development in carcinomas was discontinued due to prohibitive toxicity. WHAT THIS STUDY ADDS In patients with dedifferentiated liposarcoma, retifanlimab plus palbociclib achieved higher objective response rates than historical controls of either agent alone, with toxicities that were manageable through supportive care and dose modifications. In patients with dedifferentiated liposarcoma, retifanlimab plus palbociclib achieved higher objective response rates than historical controls of either agent alone, with toxicities that were manageable through supportive care and dose modifications. HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY These findings support further investigation of retifanlimab-palbociclib combination therapy in dedifferentiated liposarcoma through a larger validation trial. These findings support further investigation of retifanlimab-palbociclib combination therapy in dedifferentiated liposarcoma through a larger validation trial. Background Dedifferentiated liposarcoma (DDLPS) is a common subtype of soft tissue sarcoma that typically arises in the retroperitoneum or the extremities. While complete surgical resection of primary DDLPS is the upfront treatment of choice, at least 40% of patients have a local recurrence and 15–20% develop distant metastases.1 Patients with unresectable or metastatic disease have a poor prognosis, with an approximate median overall survival (OS) of 16 months.2 Current treatment options for patients with DDLPS show low efficacy. For example, the response rate and median progression-free survival (PFS) after first-line doxorubicin therapy are 8.6% and 7.2 months, respectively.3 Food and Drug Administration (FDA)-approved systemic options in the second line or beyond include the cytotoxic agents trabectedin or eribulin, which have a median PFS of 2.2 and 2 months and a response rate of 9% (inclusive of all liposarcoma histologies—DDLPS specific numbers are not available) and 0%, respectively.4 5 DDLPS is characterized by amplification of chromosome 12q14-q15, including the mouse double minute 2 (MDM2) and cyclin-dependent kinase 4 (CDK4) oncogenes. Early-phase studies of the CDK4 and CDK6 (CDK4/6) inhibitor palbociclib in patients with DDLPS have demonstrated promise in delaying disease progression, with a median PFS of approximately 4 months and a favorable safety profile, though objective responses were achieved in fewer than ≤5% of patients.6 7 Preclinical studies show that CDK4/6 inhibitors can enhance T-cell activation and synergize with programmed cell death protein 1 (PD-1) inhibition through various mechanisms, including de-repression of nuclear factor of activated T cells (NFAT) transcription factors,8 activation of endogenous retroviral elements,9 reduction of regulatory T cells,9 upregulation of cellular senescence,10 and stimulation of antigen processing and presentation.11 Further, CDK4/6 inhibitors promote the formation of memory precursor T cells that can be activated to induce antitumor efficacy by immune checkpoint inhibitors.12,14 A clinical trial of the CDK4/6 inhibitor abemaciclib demonstrated increases in a subset of T cells on treatment.15 As the PD-1 inhibitors nivolumab and pembrolizumab alone lead to objective responses in <10% of patients with DDLPS,16 17 we performed a phase 2 study of palbociclib combined with the PD-1 inhibitor retifanlimab to evaluate whether the combination would have a higher objective response rate (ORR) than historical controls of either agent alone. Methods Patients NCT04438824 enrolled patients aged 18 years or older with unresectable or metastatic DDLPS with or without well-differentiated (WD) liposarcoma. Patients who were thought likely to benefit from systemic therapy prior to surgical resection were eligible. Patients were required to have an Eastern Cooperative Oncology Group (ECOG) performance status of ≤1, measurable disease of DDLPS by Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1, and adequate bone marrow and organ function. Patients with any number of prior therapies were eligible to enroll, except for those who were treated with prior CDK4/6, PD-1, or programmed death-ligand 1 (PD-L1) inhibitors. Complete eligibility criteria are available in the study protocol. Study design and treatment In this single-arm, open-label, single-center phase 2 clinical trial, all patients received palbociclib 125 mg orally once daily for 21 days followed by 7 days off and retifanlimab 500 mg intravenously once every 28 days. Both drugs were administered starting on the first day of each 28-day cycle. A maximum of two dose reductions were permitted for palbociclib (100 mg/day, 75 mg/day once daily). Dose reductions for retifanlimab were not permitted, but dose delays were permitted for select immune-related adverse events (AEs) as described in the protocol. Palbociclib could be interrupted for up to 3 weeks, while retifanlimab could be interrupted for up to 12 weeks. Patients were evaluated radiographically every 8 weeks until week 48 and then every 12 weeks thereafter. The study was conducted in accordance with FDA regulations and Good Clinical Practice guidelines. Molecular profiling We identified pre-existing clinical tumor biopsies in the Memorial Sloan Kettering Cancer Center Biobank prior to the trial from 28 of the 30 trial patients. These were sequenced using MSK-IMPACT.18 Sequence processing was carried out as described in Zehir et al.19 Five patients had two samples, the union of both profiles is reported (P14, P15, P16, P28, and P37). Patients were stratified based on best ORR. Their mutational and copy number landscapes were compared to identify enriched alterations for any one best ORR class. In addition, known genomic metrics representative of genomic instability (eg, Fraction of Genome Altered), mutation burden, and microsatellite instability were also compared across best ORR categories. Two patients (P24 and P30) with clinical disease progression prior to their scan were included within the progressive disease (PD) best ORR category for comparisons. Endpoints and statistical analysis The primary endpoint was to assess the best ORR by RECIST 1.1 by 48 weeks after initiating treatment. A one-stage design was used to evaluate for an improvement in best ORR from 5% to 25%. The design called for 30 eligible and evaluable patients. If ≥4 patients responded (≥13%), this treatment combination would be considered promising. The design provided 96% power with a type 1 error rate of 6%. Secondary endpoints included estimating the PFS, OS, duration of response among responders, and clinical benefit rate (defined as the proportion of patients with a best response of complete response (CR), partial response (PR), or stable disease (SD)), and describing the safety of the treatment combination. PFS was defined as the time from treatment initiation to the earlier of disease progression or death from any cause. Patients who were alive and progression-free were censored at the time of their last follow-up visit. OS was defined as the time from treatment initiation until death from any cause. The Kaplan-Meier method was used to evaluate time-to-event endpoints. AEs were recorded and categorized according to National Cancer Institute Common Terminology Criteria for Adverse Events V.5.0. Results Patient characteristics and demographics A total of 30 patients were treated between October 2021 and May 2024. Patient characteristics are shown in table 1. The median age was 61 years (range 36–81), 67% of patients were male, and 90% had a primary retroperitoneal site of disease. 25 patients (83.3%) had an ECOG performance status of 0; the remainder had a performance status of 1. 13 patients (43%) had distant metastases, such as in the lung, liver, or soft tissue. 11 patients (37%) had at least 1 line of prior therapy, 9 of whom received an anthracycline-based regimen. Table 1Patient characteristicsTotal (n=30)Age (years) Median (IQR)61.5 (50.9–68.4) Range36.8–81.5Sex Female10 (33.3) Male20 (66.7)Race Asian3 (10.0) White27 (90.0)Ethnicity Hispanic or Latino2 (6.9) Not Hispanic or Latino27 (93.1) Unknown1 (3.3)Primary disease site Retroperitoneum27 (90.0) Lower extremity3 (10.0)Distant metastases at diagnosis No17 (56.7) Yes13 (43.3)ECOG performance status 025 (83.3) 15 (16.7)Prior number of systemic therapies 019 (63.3) 16 (20.0) 23 (10.0) 31 (03.3) 41 (03.3)Prior systemic therapy received Anthracycline-based9 (30) Gemcitabine and docetaxel5 (16.7) Eribulin1 (3.3) Pazopanib1 (3.3) Brigimadlin1 (3.3)Data presented as n (%) unless otherwise indicated.ECOG, Eastern Cooperative Group. Patient characteristics Data presented as n (%) unless otherwise indicated. ECOG, Eastern Cooperative Group. Efficacy At the time of data cut-off in March 2025, the median follow-up among survivors was 14.8 months. 4 patients (13%) remained on treatment, 3 (10%) on active follow-up without disease progression (2 stopped treatment due to toxicity and 1 refused further therapy), and 23 (77%) discontinued therapy due to progression. 6 of 30 patients had a confirmed PR, yielding a best ORR of 20% (95% CI 8% to 39%). Of the remaining patients, 10 (33%) had a best response of SD and 14 (47%) had PD. Two patients were not evaluable by RECIST because of clinical progression prior to the first radiographic assessment. The clinical benefit rate was 53% (95% CI 34% to 72%). 12 patients (40%) had a reduction in the sum of target lesions’ diameters (figure 1A). Most PRs were deep and durable, with the shortest lasting 9.4 months and the longest continuing through data cut-off (figure 1B). 11 patients (37%) remained progression-free at 6 months (figure 1C). Nine patients underwent surgical resection after progression on palbociclib plus retifanlimab, three of whom had R0 resections and no evidence of disease at the time of data cut-off. The best response by RECIST among those who underwent surgery was PD (n=5), SD (n=3), and PR (n=1). The median PFS and OS were 4.8 (95% CI 1.7 to 15.7) and 27 months (95% CI 21.7 to not reached (NR)), respectively (figure 2A,B). Among the six patients who achieved a PR, the median duration of response was 18.4 months (95% CI 9.4 months to NR) (figure 2C). The median time to first objective response was 15.8 weeks (range 7.6–40.9). Responses to combined palbociclib and retifanlimab. (A) Waterfall plot illustrating the maximum reduction in the sum of longest diameters of target lesions per RECIST V.1.1. Patients who were not evaluable for objective response assessment by RECIST are denoted by an “X”. (B) Spider plot showing the change in the sum of longest target tumor diameters over time per patient. (C) Swimmer’s plot of time on treatment for each patient. NE, not evaluable; PD, progressive disease; PR, partial response; RECIST, Response Evaluation Criteria in Solid Tumors; SD, stable disease. Survival on combined palbociclib and retifanlimab. Kaplan-Meier curves showing (A) progression-free survival, (B) overall survival, and (C) duration of response among responders. Safety Across the cohort, all patients had at least one treatment-related AE (TRAE) (table 2). Hematologic toxicities, including leukopenia (27 patients, 90%), neutropenia (22 patients, 73%), thrombocytopenia (n=20, 67%), and anemia (n=18, 60%), were the most common AEs of any grade. Grade 3 leukopenia and neutropenia were observed in 8 (27%) and 13 (43%) patients, respectively. Grade 4 hematologic AEs were rare; only one patient experienced Grade 4 neutropenia. One patient experienced an episode of febrile neutropenia (3.3%). Non-hematologic AEs included fatigue (13 patients, 43%), rash (n=8, 27%), pruritus (n=7, 23%), and anorexia (n=6, 20%). Laboratory abnormalities included increases in alanine aminotransferase (ALT) and aspartate aminotransferases (AST; 9 patients, 30% each), increased alkaline phosphatase (n=8, 27%), increased creatinine (8, 27%), increased lipase (n=7, 23%), and hypophosphatemia (n=6, 20%). All Grade 3 or higher TRAEs are reported in online supplemental table S1. The most common (occurring in greater than or equal to 10% of patients) Grade 3 or higher non-hematologic AEs were increased ALT (4 patients, 13%) and increased AST (n=3, 10%). Table 2Treatment-related adverse events occurring in 10% of patients. Data presented as n (%)Grade Adverse event 1 2 3 4 Any grade Leukopenia4 (13)15 (50)8 (27)027 (90)Neutropenia1 (3)7 (23)13 (43)1 (3)22 (73)Thrombocytopenia18 (60)2 (7)0020 (67)Anemia11 (37)7 (23)0018 (60)Fatigue11 (37)2 (7)0013 (43)Alanine aminotransferase increased4 (13)1 (3)4 (13)09 (30)Aspartate aminotransferase increased3 (10)3 (10)3 (10)09 (30)Creatinine increased7 (23)1 (3)008 (27)Alkaline phosphatase increased4 (13)3 (10)1 (3)08 (27)Rash4 (13)4 (13)008 (27)Pruritus6 (20)1 (3)007 (23)Lipase increased3 (10)3 (10)1 (3)07 (23)Hypophosphatemia6 (20)0006 (20)Anorexia3 (10)3 (10)006 (20)Mucositis oral3 (10)2 (7)005 (17)Headache4 (13)0004 (13)Myalgia4 (13)0004 (13)Nausea4 (13)0004 (13)Cough3 (10)1 (3)004 (13)Diarrhea2 (7)02 (7)04 (13)Serum amylase increased1 (3)3 (10)004 (13)Hepatitis03 (10)1 (3)04 (13)Constipation3 (10)0003 (10)Dry skin3 (10)0003 (10)Pain in extremity3 (10)0003 (10)Fever2 (7)1 (3)003 (10)Gastroesophageal reflux2 (7)1 (3)003 (10) Treatment-related adverse events occurring in 10% of patients. Data presented as n (%) Palbociclib dose reductions due to toxicity occurred in seven patients: five for neutropenia, one for unacceptable fatigue, and one for infection (unrelated to neutropenia). Six patients developed AEs necessitating discontinuation of retifanlimab and four developed AEs necessitating discontinuation of palbociclib (n=4). Four of the six who discontinued retifanlimab experienced Grade 3 elevations in AST/ALT consistent with hepatitis and one developed Grade 2 AST/ALT elevation (hepatitis) and subsequently Grade 3 increased creatinine (nephritis). The final patient discontinued both retifanlimab and palbociclib due to Grade 3 cognitive disturbance of unclear etiology, which improved after cessation of treatment and did not require specific intervention. This patient remains in active follow-up with a sustained PR. One patient with a sustained PR who withdrew from protocol therapy developed Grade 3 colitis nearly 6 months after the last retifanlimab dose while on active follow-up. Among patients who discontinued therapy due to toxicity, the best responses were PR in three and SD in three. Two of three PRs persisted for 10 and 14 months, respectively, until disease progression, while the third PR was ongoing at the time of data cut-off. All toxicities presumed to be immune-related, except for an AE of immune-related myasthenia gravis, resolved fully with corticosteroid treatment. Three patients with hepatitis were rechallenged with palbociclib monotherapy after resolution; one experienced recurrent transaminase elevation and discontinued treatment. Genomic profiling via MSK-IMPACT MSK-IMPACT molecular profiling was obtained at an average of 15 months prior to the study treatment initiation (range 0–101 months), with four patients profiled after treatment initiation (median 8 months, range 1–15 months). As expected, all patients had amplification of chromosome 12q, including the MDM2 and CDK4 genes. Across all patients we identified 18 additional recurrent CNAs that span key oncogenic signaling pathways (figure 3A). The most altered pathways were receptor tyrosine kinase signaling (10 of 28 patients, NTRK1, FGFR1/2, PTPRT), transcription regulation of various pathways (JUN, GLI1, RUNX1), DNA damage repair (TOP1), p53 (MDM2, TP53, PDM1D), cell cycle regulation (CDK4), and epigenetic regulation (ATRX, DNMT1) (online supplemental table S2). Tumor mutation burden was low across tumors (median 1 mutation per tumor, ranging from 0 to 4), as expected in DDLPS. The fraction of genome altered ranged from 0 to 0.8, with PD patients being marginally higher than those with PR as best response (+0.07, p≥0.05) (figure 3B). In addition, the number of copy number alterations (number of unbalanced segments) was highest in patients with a best response of PD compared with PR or SD (p≥0.05; figure 3C). JUN amplification—the most common alteration within the transcription regulation pathway—occurred in eight patients and was enriched in the PD group (p=0.39) with 6/14 PD versus 1/6, and 1/8 patients in the PR and SD groups, respectively. In addition, TERT and ATRX were exclusively altered in PD and SD groups. TERT showed amplifications in four and one patients in the PD and SD groups, respectively, and ATRX was altered (two loss-of-function mutations, one missense, and one deep deletion) in three and one patient in the PD and SD groups, respectively. Genomic landscape and copy number burden stratified by best RECIST response to palbociclib plus retifanlimab. (A) Oncoprint summarizing recurrent somatic alterations identified by MSK-IMPACT in pretreatment tumor samples (n=28), grouped by best RECIST V.1.1 response (PR, SD, PD). (B) Fraction of genome altered stratified by best RECIST response. (C) Total number of copy number alterations (unbalanced segments) stratified by best RECIST response. PD, progressive disease; PR, partial response; SD, stable disease. Discussion This open-label phase 2 study of palbociclib and retifanlimab in patients with unresectable or metastatic DDLPS met its primary endpoint, with 6 of 30 patients achieving an objective response, yielding a best ORR of 20%. A subset of patients derived significant clinical benefit, as evidenced by the duration of response of 18.4 months and large decreases in tumor burden by RECIST. After a median follow-up of 14.8 months among survivors, 23% of patients remained on study. While cross-trial comparisons have significant limitations, the 20% ORR and 4.8-month median PFS on palbociclib and retifanlimab reported in this study compare favorably to historical outcomes with palbociclib monotherapy. The first prospective study of palbociclib alone in patients with WD or DDLPS was a phase 2 study of 200 mg palbociclib daily for 14 days of a 21-day cycle. In that trial, 19 of 29 evaluable patients (66%) were progression-free at 12 weeks and the median PFS was 17.9 weeks. One patient (3.4%) achieved an objective response. Notably, this study included five patients with well-differentiated liposarcoma, which grows slower compared with DDLPS.6 A subsequent phase 2 study of palbociclib 125 mg daily for 21 days of a 28-day cycle in 59 evaluable patients with WD/DDLPS (22% of whom had WD disease only) yielded similar results: a PFS rate at 12 weeks of 57.2% and a median PFS of 17.9 weeks. One patient (1.7%) achieved a PR to palbociclib.7 This study’s outcomes also match or exceed those of trials of PD-1 inhibition with or without inhibition of cytotoxic T-lymphocyte associated protein-4 in patients with DDLPS. The phase 2 SARC028 trial evaluated pembrolizumab alone and included 34 patients with DDLPS. Three of these patients (8.8%) had a PR to treatment. The median PFS among all patients with liposarcoma (n=40, including an additional four patients with pleomorphic liposarcoma and two with myxoid/round cell liposarcoma) was 8 weeks.16 The expansion phase of a randomized phase 2 study of nivolumab with or without ipilimumab in select soft tissue sarcomas enrolled 29 patients with DDLPS to receive either nivolumab (n=15) or the combination (n=14). The response rates to nivolumab alone and nivolumab plus ipilimumab were 6.7% (1 of 15) and 14.3% (2 of 14), with a median PFS of 4.6 and 5.5 months, respectively. Of note, four patients (13.8%) had WD disease on histologic assessment. A separate phase 2 study of durvalumab plus tremilimumab enrolled six patients with liposarcoma, including three with dedifferentiated, two with WD, and one with pleomorphic disease. The 12-week PFS rate in this subset was 17%, among whom there were no responses.20 The response rate and PFS reported here also improve on other recently tested and approved agents in DDLPS. A recently reported randomized phase 3 trial of the MDM2 inhibitor brigimadlin versus doxorubicin in treatment-naïve patients with advanced or metastatic DDLPS failed to meet its primary endpoint of PFS by blinded independent central review. Although the median PFS of 7.2 months in the doxorubicin arm was longer than anticipated, the ORR to doxorubicin was only 8.6%.3 Currently available second-line agents also have low response rates and lead to brief PFS on average. One such chemotherapeutic, eribulin, had significantly longer survival rates compared with dacarbazine (2.9 vs 1.7 months and 15.6 vs 8.4 months for PFS and OS, respectively) in patients with advanced liposarcoma. However, a subset analysis in patients with DDLPS found no significant difference in PFS and no objective responses.21 Similarly, trabectedin led to a significantly longer median PFS compared with dacarbazine (3.0 vs 1.5 months) in patients with liposarcoma. PRs were observed in 9% of patients with liposarcoma, although a subset analysis in patients with DDLPS was not performed.22 The toxicities observed in this study are similar to those in previous studies of CDK4/6 plus PD-1 inhibition. The phase 2 NEWFLAME study in patients with hormone receptor (HR)-positive breast cancer administered nivolumab, abemaciclib, and letrozole or fulvestrant. Of 17 treated patients, 6 (35.3%) had Grade 3 hepatotoxicity leading to treatment discontinuation and 1 patient died from immune-mediated interstitial lung disease (ILD). The study was terminated early due to safety concerns.23 Similarly, a phase 1b study of pembrolizumab plus abemaciclib with or without anastrazole in patients with metastatic HR-positive breast cancer identified 16 of 54 patients (30%) with Grade 3 increases in AST and/or ALT and 7 (13%) with ILD or pneumonitis across the two cohorts of the study, including two deaths from ILD.24 Finally, a phase 1b study of pembrolizumab plus abemaciclib in 50 patients with advanced non-small cell lung cancer also identified unexpectedly high increases in AST and pneumonitis (32% of patients each).25 Although palbociclib plus retifanlimab induced similar degrees of elevated hepatic transaminases, there were no cases of pneumonitis and no Grade 5 events. All hepatic AEs in our study were fully reversible with corticosteroids. Ideally, patients most likely to benefit from palbociclib plus retifanlimab could be identified prospectively through predictive biomarkers, thereby maximizing benefit while minimizing toxicity and avoiding ineffective treatment in non-responders. To this end, paired baseline and on-treatment tumor and blood samples were collected in this study. Tumor specimens are undergoing single-cell RNA sequencing and immunohistochemistry for candidate biomarkers, including CDH1826 and ANGPTL4,15 while peripheral blood is being profiled in parallel. These analyses are ongoing, and results will be reported in a subsequent manuscript. An exploratory analysis of clinical-grade targeted exome sequencing performed on tumor tissue identified a potential molecular correlate of resistance. JUN amplification was enriched among patients with PD (6/14 PD vs 1/6 PR and 1/8 SD). Amplification of 1p32, including JUN, leads to interference of adipocytic differentiation27 and to promotion of DDLPS cell migration and invasion28; knockdown of c-Jun inhibits DDLPS cell proliferation in vitro and in vivo.28 29 In multi-omics analyses of DDLPS samples, JUN amplification was associated with a poor prognosis and its expression significantly correlated with somatic copy number burden.30 31 In our cohort, this enrichment among non-responders raises the possibility that JUN amplification may confer resistance to palbociclib plus PD-1 blockade. While limited by small sample size, these findings suggest that alteration of JUN warrants further investigation as a candidate biomarker of resistance to immune checkpoint blockade-based therapy in DDLPS rather than solely a marker of poor prognosis. In conclusion, this phase 2 study of palbociclib plus retifanlimab in patients with dedifferentiated liposarcoma met its primary endpoint. A subset of patients had a sustained reduction in tumor burden and remain either on treatment or on active surveillance without disease progression. AEs were generally manageable with supportive care and dose modifications. While a randomized study is needed to confirm that the combination is superior to either agent alone, these results are encouraging compared with historical controls and compared with current FDA-approved treatment options for patients with DDLPS who have progressed on prior lines of therapy. Supplementary material Footnotes Funding: This study was supported in part by the Geoffrey Beene Foundation, Cycle for Survival, and Food and Drug Administration (FDA) R01 FD007528. Memorial Sloan Kettering Cancer Center (MSK) is supported by the National Institutes of Health/National Cancer (NCI) Institute Cancer Center Support Grant P30 CA008748. Funding and retifanlimab for the clinical trial in this manuscript were provided by Incyte Corporation. Provenance and peer review: Not commissioned; externally peer reviewed. Patient consent for publication: Not applicable. Ethics approval: This study involves human participants and was approved by Institutional Review Board at Memorial Sloan Kettering Cancer Center IRB Number: 20-062 A(17). Participants gave informed consent to participate in the study before taking part. Data availability statement Data are available upon reasonable request. References Associated Data Supplementary Materials Data Availability Statement Data are available upon reasonable request.