GDFATHER-NEO 试验:新辅助 Nivolumab 加 Visugromab 或对照治疗肌层浸润性膀胱癌结果
Neoadjuvant nivolumab with or without visugromab in muscle-invasive bladder cancer: the phase 2a GDFATHER-NEO trial
在 31 例肌层浸润性膀胱癌患者中,新辅助 Nivolumab 联合 GDF-15 中和抗体 Visugromab 组经本地病理评估的病理完全缓解率为 33%,高于 Nivolumab 联合安慰剂组的 8%;主要终点为病理完全缓解率和影像学反应率。研究评估了该联合方案在未接受过免疫治疗、不适合或拒绝顺铂化疗的肌层浸润性膀胱癌患者中的初步抗肿瘤活性及安全性。
原文明确给出 IIa 期试验中 GDF-15 单抗与 PD-1 单抗联用相比 PD-1 单抗在肌层浸润性膀胱癌中的病理完全缓解率对比。
Main
Bladder cancer is the ninth most commonly diagnosed cancer worldwide, with globally more than 600,000 new cases and approximately 220,000 deaths estimated for 2022 and a male-to-female ratio of about 3:1 (ref. 1). Around 75% of patients present with non-muscle-invasive bladder cancer (NMIBC). Most patients with muscle-invasive (≥T2) bladder cancer (MIBC) present with primary invasive disease, whereas up to 15% had progressed from NMIBC, almost exclusively from high-risk disease2.
The treatment of MIBC is at a watershed moment: novel, more effective systemic therapies are enabling an increasing number of patients to achieve clearance of the bladder tumor and improved survival compared with the therapies that, until recently, were the only standard-of-care options3. This change is the culmination of a 10-year evolution of systemic therapy in MIBC, beginning with immune checkpoint inhibitor (ICI) monotherapy trials4,5, then exploring various combination regimens6, ultimately resulting in the regulatory approval of durvalumab plus gemcitabine–cisplatin, followed by cystectomy and adjuvant durvalumab, and of enfortumab vedotin plus pembrolizumab (EVP), on the basis of pivotal phase 3 trials7,8,9. These advances have been paralleled by profound innovation in the monitoring of treatment response, in particular the assessment of minimal residual disease using circulating tumor DNA or urinary tumor DNA10,11.
ICIs, predominantly antibodies directed against programmed cell death protein 1 (PD-1) or programmed cell death ligand 1 (PD-L1), have entered routine practice for neoadjuvant cancer treatment since 202012. In the neoadjuvant and perioperative setting, more than 30 prospective trials in MIBC have been reported since 201613. These trials sought to increase the rate of pathologic complete response (pCR), an established surrogate endpoint in neoadjuvant MIBC trials14, and to provide options for the estimated ≥50% of patients with MIBC who are ineligible for neoadjuvant cisplatin-based chemotherapy3,13.
Neoadjuvant immune checkpoint blockade can downsize tumors, may render initially unresectable tumors resectable and may enable surgical deescalation with organ preservation12. Two cycles of atezolizumab4 or three cycles of pembrolizumab5 produced pCR rates of 31% and 37%, respectively, and the addition of ipilimumab to nivolumab (N) resulted in a pCR rate of 46%15. Despite this progress, more than half of all patients still do not achieve a pathologic response, opportunities for safe bladder preservation remain rare and effective, well-tolerated options for patients who are ineligible for or decline cisplatin stay limited.
Growth differentiation factor-15 (GDF-15) is a member of the transforming growth factor-β superfamily with established roles in pregnancy and tissue stress response. GDF-15 is overexpressed in many cancers, including bladder cancer, and was identified as a potent local immunosuppressant and T cell repellent and as a determinant of resistance to PD-1 and PD-L1 blockade16,17. Neutralization of GDF-15 is therefore a rational strategy to restore T cell infiltration and antitumor immunity.
In the first-in-human trial GDFATHER-01, the GDF-15-neutralizing antibody visugromab (V) combined with nivolumab showed antitumor activity and a favorable safety profile in patients whose tumors were refractory to or had relapsed on ongoing PD-1 or PD-L1 blockade, including advanced/metastatic urothelial carcinoma (UC)18. The overall response rate (ORR) in UC was 18.5% (95% confidence interval (CI) 6.3–38.1%) with a median duration of response of 28.8 months (95% CI 7.4–39.4 months)19. Here we report the primary results of GDFATHER-NEO, an exploratory phase 2a trial of neoadjuvant nivolumab combined with visugromab or placebo in patients with MIBC.
Results
Patient characteristics, treatment and trial design
Between 26 September 2023 and 26 March 2025, 50 patients were screened and 31 enrolled in GDFATHER-NEO at four centers in Italy. All had newly diagnosed MIBC of clinical stage T2–T4a disease; 19 of 31 (61%) had cT2 N0 M0 tumors. Twenty-two (71%) were male and 9 (29%) were female, consistent with the sex distribution of MIBC. High PD-L1 expression, defined as a locally assessed combined positive score (CPS) ≥10, was balanced between the two groups (8 of 15 (53%) in group A, receiving nivolumab plus placebo, and 9 of 16 (56%) in group B, receiving nivolumab plus visugromab). In total, 6 of 15 patients (40%) in group A and 5 of 16 (31%) in group B were ineligible for cisplatin because of impaired renal function or hearing loss; the remaining patients (9 of 15 (60%) and 11 of 16 (69%), respectively) declined cisplatin-based chemotherapy (Table 1).
Within strata defined by clinical tumor stage and locally assessed PD-L1 expression, patients were allocated sequentially to three 4-weekly cycles of intravenous nivolumab 480 mg plus placebo (group A) or to three cycles of nivolumab 480 mg plus visugromab 20 mg kg−1 infused before nivolumab (group B); patients were blinded to their allocation (Fig. 1b). Four to eight weeks after the last dose, patients were restaged by imaging (that is, cross sectional imaging with thorax–abdomen computed tomography (CT) scan, with or without pelvic magnetic resonance imaging (MRI) scans) and underwent radical cystectomy (RC). Those who refused RC were offered the option to receive a repeat transurethral resection of the bladder tumor (re-TURBT) within a multidisciplinary tumor board discussion and after careful discussion with the patients. Surgery was followed by observation alone or by additional off-protocol therapy at the investigator’s discretion.
a, CONSORT diagram showing the flow of participants in the GDFATHER-NEO trial in which histopathologic response and radiographic response (per RECIST v.1.1) were both assessed to evaluate the efficacy of this multimodal treatment approach. b, Design of the covariate-stratified, parallel-group comparative phase 2a trial with sequential treatment allocation within strata. i.v., intravenous.
The two primary endpoints were the rate of pCR (defined as ypT0 N0 or ypT0 NX, depending on the type of surgery received) and the rate of radiographic response according to Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1, both by local assessment (Fig. 1a). Secondary endpoints comprised major pathologic response (MPR), defined as ypT0/Ta/Tis/T1 N0 after RC (or Nx after re-TURBT), centrally assessed pathologic and radiographic response, time-to-event endpoints including event-free survival (EFS), pharmacokinetics and safety.
Of the 31 patients in the safety analysis population, which was identical to the intention-to-treat population, three patients (two in group A and one in group B) stopped neoadjuvant treatment after the first cycle owing to adverse events (AE), and 28 patients (13 in group A and 15 in group B) received all three planned cycles (Extended Data Table 1). Nine of 31 patients (29%) underwent re-TURBT (3 in group A and 6 in group B), 1 of whom (group B) subsequently underwent RC, and 19 (61%) underwent RC alone (10 in group A and 9 in group B). Three patients did not undergo protocol-defined surgery: two withdrew consent before surgery, and one discontinued neoadjuvant treatment early and underwent RC outside the trial. Adjuvant therapy was given to five patients, all in group A: chemotherapy in four (cisplatin plus gemcitabine in three and carboplatin plus gemcitabine in one) and radiotherapy after TURBT in one.
Protocol deviations mostly concerned missing laboratory assessments and deviations from protocol-defined time windows for participants’ clinical visits. In five participants, retrospective analysis showed that not all eligibility criteria were met; in two participants, investigational medicinal product administration errors occurred leading to the exclusion of one participant from all efficacy analyses; in two participants, re-TURBT had been conducted instead of radical cystectomy before formal introduction of this option with a respective protocol amendment.
Five patients received chemotherapy for recurrent disease: three in group A (carboplatin plus gemcitabine in two and carboplatin followed by carboplatin plus gemcitabine in one) and two in group B (cisplatin plus gemcitabine in both, followed by avelumab in one).
Pathologic and radiographic response
Twenty-eight patients were evaluable for locally assessed pathologic response and 28 for locally assessed radiographic response (Fig. 1a). In group A, one patient withdrew consent before the protocol-defined primary endpoint assessment and one received an incorrect dose; both were excluded from the pathologic and the radiographic efficacy analyses, leaving 13 evaluable patients for each endpoint. In group B, 1 patient who withdrew consent after response imaging but before surgery was excluded from the pathologic efficacy analysis, and a different patient with no lesion suitable for RECIST assessment by local review was excluded from the radiographic efficacy analysis, leaving 15 evaluable patients for each endpoint.
For central review, 27 patients were evaluable for pathologic response and 28 for radiographic response. In addition to the exclusions described above, one patient in group A was not evaluable for central pathologic response because no specimen was available (group A, n = 12; group B, n = 15). For central radiographic response, one patient in group A had no lesion suitable for RECIST assessment, whereas the above-mentioned patient in group B who was considered nonevaluable by local review was considered evaluable for central review (group A, n = 12; group B, n = 16).
Among the 28 patients evaluable for locally assessed pathologic response, pCR was reached by 1 of 13 patients (8%, 95% CI 0–36%) in group A and by 5 of 15 patients (33%, 95% CI 12–62%) in group B (Fig. 2a). An additional two patients in group A and four in group B had residual ypTa, ypTis or ypT1 N0, resulting in MPR rates of 3 of 13 (23%, 95% CI 5–54%) with nivolumab plus placebo and 9 of 15 (60%, 95% CI 32–84%) with nivolumab plus visugromab.
a, Locally assessed histopathologic response (pCR and MPR) per treatment group. b, Centrally assessed histopathologic response (pCR and MPR) per treatment group. c, Correlation of histopathologic response by local (upper line) and central (lower line) assessment with tumor stage (T2 versus T3–T4a) and PD-L1 expression (CPS ≥10 versus CPS <10, central assessment; one participant in the N + P group was not evaluable for central CPS score). d, Locally assessed radiographic response (RECIST v.1.1) per treatment group and its correlation with local histopathologic response. e, Centrally assessed radiographic response (RECIST v.1.1) per treatment group and its correlation with central histopathologic response. f, CT images from one participant illustrative of depth of response achieved after three cycles of neoadjuvant treatment with combined nivolumab plus visugromab. g, Representative haematoxylin and eosin (H&E)-stained sections from participant 19, who achieved an MPR, and participant 12, who achieved a pCR. For each participant, baseline (left) and post-therapy specimens (right) are shown. For participant 19, arrows indicate confluent tumor zones with variable tumor cell types at baseline and T1 tumor after therapy. For participant 12, the arrow indicates tumor cells invading the muscularis, whereas no residual viable tumor was detectable post-therapy. Scale bars, 200 µm. H&E staining was performed once per specimen as part of central pathology review; images are from representative patients with MPR and pCR. ITT, intention to treat; NE, nonevaluable; pNR, pathological nonresponse.
Central pathologic review was largely concordant with local assessment, with pCR in 2 of 12 patients (17%, 95% CI 2–48%) in group A and 7 of 15 (47%, 95% CI 21–73%) in group B, and MPR in 4 of 12 (33%, 95% CI 10–65%) and 9 of 15 (60%, 95% CI 32–84%), respectively, among the 27 evaluable patients (Fig. 2b). Pathologic response rates were higher with nivolumab plus visugromab in each T stage and centrally assessed PD-L1 CPS subgroup, although the subgroups were small (Fig. 2c). Pathologic ypTNM stages for individual patients are provided in Extended Data Table 1.
By local RECIST version 1.1 assessment, two partial responses were recorded in group A (ORR 2 of 13; 15%, 95% CI 2–45%) and seven complete and two partial responses in group B (ORR 9 of 15; 60%, 95% CI 32–84%); the complete response rate was 47% (95% CI 21–73%) (Fig. 2d). Radiographic and pathologic responses were broadly concordant. Central review recorded an ORR of 4 of 12 (33%, 95% CI 10–65%) in group A and of 8 of 16 (50%, 95% CI 25–75%), including two complete responses (13%, 95% CI 2–38%), in group B (Fig. 2e). Baseline and post-treatment CT images from a patient with a radiographic and pathologic complete response by local assessment are shown in Fig. 2f.
Representative haematoxylin and eosin (H&E)-stained sections illustrate the histopathological responses observed after nivolumab plus visugromab treatment (Fig. 2g). While residual viable tumor remained in the participant who achieved MPR (even if only in the form of carcinoma in situ), no residual viable tumor was identified in the post-treatment specimen from the participant who achieved pCR.
Among radiographically evaluable participants, progressive disease as best overall response was recorded in four patients (two in each group). Two of these—one per group—had stopped neoadjuvant treatment early because of AEs (Extended Data Table 1) but were, in line with the statistical analysis plan, imputed as nonresponders.
Event-free survival
At the time of the data cutoff (15 July 2026), the median follow-up was 15.7 months ((interquartile range 15.1–16.5 months).
Event-free survival was defined as the time from cycle 1 day 1 (C1D1) to local (intravesical in patients undergoing re-TURBT) or distant recurrence documented by cross-sectional imaging and/or biopsy, to the start of therapy for progression or recurrence, or to death from any cause. Patients without an event were censored at the start of adjuvant therapy or at last follow-up.
Six events were recorded in total, three in each group. Median EFS was not reached in either group, with protocol-defined follow-up limited to 1 year after surgery (Extended Data Fig. 1; intention-to-treat population, n = 31). No event occurred among patients with a pCR, including the four patients in group B who underwent re-TURBT alone without subsequent RC. All six events occurred either in patients with a pathologic nonresponse or in patients who did not proceed to protocol-defined surgery. Given the small number of events and the short follow-up, these data are descriptive only. Figure 3 displays the core treatment period, surgical procedures, histopathological responses, follow-up including systemic treatments outside of the protocol-defined treatment and potential recurrence of all 31 trial participants in both groups.
Swimmer plots for all 31 trial participants across both treatment arms, depicting all surgical interventions, the local histopathologic response, systemic treatment outside of the trial treatment and disease recurrence occurring during the core trial period and the follow-up period. End-of-trial visit for participant 25 was confirmed post data cutoff (15 July 2026).
Baseline CD8+ T cell state and pathologic response
Baseline intratumoral immune-cell densities and phenotypes did not differ significantly between the two treatment groups (Extended Data Fig. 2), although median densities of total T cells, CD8+ T cells and proliferating CD8+Ki-67+ T cells were numerically higher in group B (Fig. 4a). The proportion of proliferating CD8+ T cells within the total CD8+ T-cell compartment did not significantly differ between the treatment groups (Fig. 4a).
a, Baseline densities of total T cells, total CD8+ T cells and proliferating CD8+ T cells (CD8+Ki67+) and the ratio of proliferating CD8+ T cells to total CD8+ T cells in patients treated with nivolumab plus placebo (N + P) or nivolumab plus visugromab (N + V). b, Baseline densities of total, CD8+ and CD8+Ki67+ T cells stratified by local pathologic response within each treatment group. c, Association of centrally assessed PD-L1 CPS with local pathologic response. The horizontal dotted line denotes the PD-L1 CPS cutoff of 10. Each data point represents one baseline tumor sample from one individual patient. No repeated measurements from the same patient were included within any statistical comparison. Horizontal lines indicate medians. The distribution of individual observations is shown. Statistical comparisons were performed using two-sided Mann–Whitney U tests. n represents individual patients with an evaluable baseline tumor sample and available biomarker measurement. mIF staining and analysis were performed once per patient specimen; each data point represents one patient. mIF, multiplex immunofluorescence; NR, no (pathologic) response.
When T cell densities were analyzed according to pathologic response, patients showing any MPR (including patients with a pCR) to nivolumab plus visugromab had higher baseline densities of total CD8+ T cells and proliferating CD8+Ki-67+ T cells than nonresponders (P = 0.008 and P = 0.003). No such association was observed for pathologic responders versus nonresponders in the nivolumab plus placebo group (Fig. 4b). Responses to the combination occurred across the range of centrally assessed CPS values, and PD-L1 expression did not distinguish pathologic responders from nonresponders (Fig. 4c).
These exploratory data suggest that a preexisting proliferative CD8+ T cell state, rather than PD-L1 status, is associated with response to combined GDF-15 and PD-1 blockade in a rather short-term, 12-week treatment setting.
Baseline circulating total GDF-15 concentrations and intratumoral GDF-15 H scores were comparable between the two groups and were not associated with baseline densities of total T cells, CD8+ T cells or proliferating T cell populations (Extended Data Fig. 3c), nor with pathologic response in either group (Extended Data Fig. 3a,b). Baseline GDF-15 abundance alone therefore did not predict baseline immune cell infiltration or treatment response in this cohort.
GDF-15 blockade has previously been associated with induction of interferon (IFN) signaling and of the IFN-inducible chemokines C–X–C motif chemokine ligand (CXCL)9 and CXCL1018. Circulating cytokine and chemokine concentrations were therefore assessed. Baseline concentrations of CXCL9, CXCL10 and IFN-γ were comparable between groups (Extended Data Fig. 4a). On-treatment CXCL9 and IFN-γ concentrations were frequently below the lower limit of quantification, precluding a robust assessment of treatment-associated changes, whereas longitudinal CXCL10 concentrations varied substantially between patients without a consistent difference between groups (Extended Data Fig. 4b).
Taken together, in this neoadjuvant setting, baseline infiltration with proliferating CD8+ T cells appears to correlate with response to nivolumab plus visugromab, whereas baseline GDF-15 abundance or PD-L1 expression seem to be nonpredictive.
Pharmacokinetics
Total visugromab serum concentrations were similar across treated patients and consistent with those observed in the first-in-human GDFATHER-01 phase 1/2 trial (Extended Data Fig. 5). Mean serum concentrations declined between dose administrations but remained above 100 µg ml−1 during the 4-week intervals, consistent with continued GDF-15 neutralization over the whole treatment period.
Safety
The safety analysis population comprised all 31 patients who received at least one dose of trial treatment (Table 2 and Extended Data Table 2). AEs of any grade were recorded in every patient. No treatment-related deaths occurred. One patient in group A died from disease progression, approximately 9 months after RC.
Grade 3–4 treatment-emergent AEs (TEAEs) were reported in 20 of 31 patients (65%): 12 of 15 (80%) in group A and 8 of 16 (50%) in group B (Extended Data Table 2). Treatment-related AEs (TRAEs) of any grade were more frequent with nivolumab plus visugromab (14 of 16 (88%) versus 8 of 15 (53%) for nivolumab plus placebo), driven mainly by asthenia, pruritus and hypothyroidism, whereas grade 3–4 treatment-related events occurred in 5 of 31 patients (16%) and were less frequent with nivolumab plus visugromab (2 of 16, 13%) than in the control group (3 of 15, 20%). These comprised increased liver transaminases in both groups, hypothyroidism and myalgia in group A and interstitial lung disease in group B; the increase in liver enzymes in group B occurred in the context of an immune-mediated hepatic disorder (Table 2). Three serious TRAEs were recorded: immune-mediated hepatic disorder and interstitial lung disease in group B, and hypothyroidism in group A.
At the preferred-term level, asthenia, pruritus and hypothyroidism were numerically more common in group B. Immune-related AEs occurred in 9 of 31 patients (29%): 4 of 15 (27%) in group A and 5 of 16 (31%) in group B; grade 3–4 immune-related events occurred in 5 of 31 patients (16%): 3 of 15 (20%) in group A and 2 of 16 (13%) in group B (Extended Data Table 3). All immune-related events except for one case of pruritus in group B required systemic treatment, with corticosteroids given in seven patients and hormone replacement only in one participant with hypothyroidism. Two patients in group A, and no patients in group B, discontinued systemic treatment because of a TRAE (Table 2).
Surgery was postponed, that is, versus the originally planned date, in five patients in group A and in six patients in group B. One delay in group A was attributed to a non-treatment-related toxicity, and one delay in each group to disease progression; all remaining delays were for logistical reasons only. No postoperative complication attributable to neoadjuvant treatment was recorded.
No relationship was detected between pharmacokinetic exposure and the occurrence of treatment-emergent AEs. One patient in group B had preexisting, transiently detectable ADAs, with no apparent effect on pharmacokinetics or clinical outcome.
Discussion
MIBC treatment has been revolutionized by the advent of enfortumab vedotin, an antibody–drug conjugate, administered in combination with the PD-1 inhibitor pembrolizumab (EVP). The combination establishes a new standard of care, with pCR rates exceeding 50% in both cisplatin-eligible and cisplatin-ineligible populations8,9. Identifying novel agents that are sufficiently potent and safe to be considered for further combinatorial development with EVP is a central focus of current clinical research, with one main focus on integrating well-tolerable immunotherapeutic approaches. In parallel, increasing the proportion of patients who can forgo RC through effective perioperative therapy followed by re-TURBT is an emerging paradigm under investigation in trials such as EV-209 (NCT07475806) and EV-309 (NCT07566156), and deescalation of perioperative therapy by omitting RC or the adjuvant component is a further priority. Enhancing the efficacy of the neoadjuvant component, and thereby the clinical complete response rate, would allow more patients to be considered for such deescalation.
In the exploratory GDFATHER-NEO phase 2a trial, the addition of the GDF-15-neutralizing antibody visugromab to neoadjuvant nivolumab was associated with three- to fourfold higher rates of pathologic and radiographic response than nivolumab alone in patients with MIBC who were ineligible for or who declined cisplatin-based chemotherapy. GDFATHER-NEO is the first trial to evaluate GDF-15 neutralization in the neoadjuvant setting and provides an initial estimate of the effect size of combined GDF-15 and PD-1 blockade in newly diagnosed, checkpoint-inhibitor naive, systemically previously untreated patients with bladder cancer. GDF-15 is a regulator of antitumor immunity20, and its neutralization has been proposed as a means of reversing the immune exclusion that limits the activity of PD-1 and PD-L1 blockade. Transcriptomic analyses of approximately 10,000 tumors from The Cancer Genome Atlas and related datasets have implicated GDF-15-mediated immunosuppression across several tumor types, including urothelial carcinoma18, and notable clinical activity of visugromab plus nivolumab has been reported in patients with urothelial carcinoma, nonsquamous non-small cell lung cancer and hepatocellular carcinoma whose disease was refractory to or had relapsed on ongoing PD-1/-L1 blockade and who had exhausted available treatment options, with observation of deep, long-term responses with the combination in a subgroup of treated patients19.
Patients in the GDFATHER-NEO trial were stratified by clinical tumor stage and PD-L1 expression and were then allocated sequentially to nivolumab plus visugromab or to nivolumab plus matching placebo (Fig. 1b). Locally assessed pCR was 33% with the combination versus 8% with nivolumab alone, MPR was 60% versus 23% and radiographic response was 60% versus 15% (Fig. 2). Central review confirmed the results, with the difference in pCR being even more pronounced (47% versus 17%).
This trial was small and not powered for formal comparison, so these estimates are descriptive only. Yet, the observed magnitude of the differences and their consistency across assessments provide an early indication that combined GDF-15 and PD-1 blockade can potentiate PD-1 blockade substantially in this setting, and support further evaluation of visugromab in combination with checkpoint inhibition in early stage, checkpoint inhibitor-naive disease.
With follow-up limited by protocol to 1 year after surgery, only six EFS events occurred and medians were not reached in either group, so the EFS data cannot be informative about long-term benefit. In line with the diverging pathologic response rates, adjuvant therapy was given to five patients in group A and to none in group B, which might be a potential explanation for the limited number of early events in the monotherapy group and in parallel also be reflective of the treatment effect achieved in group B. No recurrence occurred among the four patients in group B who achieved a pCR after re-TURBT alone; although based on very small numbers, this observation is consistent with the growing interest in response-guided bladder preservation and supports the prospective evaluation of bladder-sparing strategies after visugromab-based combination therapy.
Pathologic responses were most frequent in patients with cT2 tumors and were observed across the range of PD-L1 expression (Fig. 2c). Treatment consisted of three 4-weekly cycles, and surgery was performed 4–8 weeks after the last dose. Whether a longer interval between the last dose and surgery, or additional treatment cycles, could further improve pathologic response or not cannot be determined from these data and merits prospective evaluation.
These results require careful contextualization. Neoadjuvant and perioperative immunotherapy has only recently entered MIBC treatment algorithms3,21. In NIAGARA, the addition of durvalumab to gemcitabine–cisplatin increased the pCR rate from 27.5% to 37.3% and improved EFS7. In cisplatin-eligible patients, EVP increased the pCR rate to 55.8% (KEYNOTE-B15/EV-304)8, and in cisplatin-ineligible or cisplatin-declining patients, the companion phase 3 trial reported a pCR rate of 57.1% (KEYNOTE-905/EV-303), establishing a benchmark and an emerging standard of care in the population studied here9. Checkpoint inhibitor monotherapy trials, including PURE-01 and PANDORE (pembrolizumab), ABACUS (atezolizumab) and AURA (avelumab), have reported pCR rates of 29–37%4,5,22, and combined PD-(L)1 and CTLA-4 blockade has produced pCR rates of 35% (DUTRENEO)23 to 43–46% (NABUCCO)15. The pCR rate of 33% observed with nivolumab plus visugromab lies within the range reported for chemotherapy-free immunotherapy in this setting and was achieved with only three cycles of treatment, supporting the feasibility of chemotherapy-free neoadjuvant approaches in patients who cannot or will not receive cisplatin.
The pCR rate of 8% in the nivolumab plus placebo group is lower than might have been expected. In PrE080724, neoadjuvant nivolumab monotherapy in 12 evaluable patients with cisplatin-ineligible cT2–T4a MIBC produced a pCR rate of 17% and an MPR rate of 25%, close to the 8% pCR and 23% MPR observed by local review, and very similar to the 17% pCR and 33% MPR observed by central review, in the nivolumab plus placebo control group of GDFATHER-NEO. The available data therefore do not provide strong evidence that the control group substantially underperformed; however, the possibility cannot be excluded given the small sample size. Because GDF-15 neutralization is expected to act on a mechanism of resistance that is not specific to a particular PD-1 or PD-L1 backbone, the combination of visugromab with more active regimens (for example, EVP) and other checkpoint inhibitors might yield higher efficacy and warrants formal evaluation.
Multiplex immunofluorescence identified a response-associated immune phenotype in the combination group, characterized by higher baseline densities of total CD8+ and proliferating CD8+Ki-67+ T cells (Fig. 4b). Responses occurred across the range of centrally assessed PD-L1 CPS values, suggesting that PD-L1 CPS alone may be insufficient to identify responders. Baseline densities of total T cells, CD8+ T cells and CD8+Ki-67+ T cells were numerically higher in group B, but between-group differences were not significant, the distributions overlapped substantially and the proliferating fraction of CD8+ T cells was comparable between groups (Fig. 4a). Restricting the comparison to the range of baseline CD8+ T cell density common to both groups (≤300 cells mm−2), pCR was observed in 4 of 11 patients in group B and 1 of 12 in group A (Fig. 4a), suggesting that the greater antitumor activity of the combination was not explained by baseline immune infiltration alone. These observations are compatible with a model in which GDF-15 constrains the activity of preexisting intratumoral CD8+ T cells and in which GDF-15 neutralization enables a more effective antitumor response. The small sample size and the exploratory nature of these analyses mean that this interpretation remains hypothesis-generating.
Neither circulating GDF-15 nor intratumoral pro-GDF-15 levels were associated with pathologic response or immune-cell infiltration. These findings contrast with previous reports that linked elevated baseline circulating GDF-15 concentrations to inferior outcomes for PD-1/PD-L1 blockade, as observed in metastatic melanoma16 and advanced NSCLC25. They also differ in part from observations made in the GDFATHER-01, first-in-human phase 1/2 trial, where higher baseline serum and intratumoral GDF-15 levels were associated with reduced CD8+ T cell infiltration and where GDF-15 blockade induced circulating CXCL9 and CXCL1018. Notably, median baseline circulating GDF-15 concentrations and intratumoral pro-GDF-15 abundance in GDFATHER-NEO were substantially lower than those reported in the metastatic melanoma and advanced NSCLC cohorts of GDFATHER-01. Cross-trial comparisons should therefore be interpreted with caution in light of given differences in GDF-15 abundance, disease type/setting, treatment duration difference, baseline immune composition and patient status (treatment naive versus advanced, relapsed/refractory). The pharmacodynamic significance of GDF-15-related biomarkers may therefore be context dependent and heavily impacted by the factors described above.
Given the association of clinical response to combination treatment in GDFATHER-NEO with baseline infiltration of CD8+ T cells and proliferating CD8+ T cells, functional immune states may be more informative specifically in early-stage disease with short-term treatment than static measurements of PD-L1 or GDF-15. However, this association requires further investigation and validation in independent cohorts to define the specific cellular programs involved and establish how GDF-15 reshapes the tumor immune microenvironment and influences responses to PD-1 blockade. The high pCR rate in the combination arm limited the availability of tumor-containing on-treatment biopsies, precluding a robust assessment of the intratumoral effects of GDF-15 blockade. Further translational studies will be needed to define how GDF-15 shapes antitumor immunity across disease settings and to identify biomarkers that most accurately capture biologically relevant GDF-15 activity.
The combination of nivolumab and visugromab was well tolerated, and the pattern of AEs was broadly similar in the two groups. No treatment-related deaths occurred, no patient discontinued the nivolumab plus visugromab combination because of toxicity—as opposed to two patients in the nivolumab control arm—and no delay for surgery or postoperative complication was attributed to neoadjuvant treatment (Table 2). No safety signal specific to visugromab emerged.
This exploratory trial has important limitations. With just 15 and 16 patients allocated per group, all estimates and the comparisons between groups are descriptive rather than inferential, despite a balanced distribution of key patient and tumor characteristics through participant stratification. Treatment was allocated sequentially within strata rather than by randomization26; sequential allocation does not provide concealment on the sponsor side and leaves the comparison susceptible to selection and period effects, although patients were blinded to treatment received. The allocation sequence was generated and applied centrally and was not disclosed to investigators, so that the next assignment within a stratum could not be anticipated at the time of enrollment. Once a patient had been allocated, investigators and sponsor staff were aware of the assigned treatment; patients were not. Formal randomization would, in addition, have protected against imbalance in unobserved confounders. Local and central pathologic review and central imaging review were performed in a blinded fashion. Finally, the small number of patients within each stratum limits the interpretation of the subgroup analyses, the exploratory biomarker analyses were not adjusted for multiple comparisons and the trial was conducted at only four centers in a single country.
In previously reported neoadjuvant series, pathologic nodal involvement has been uncommon or absent27,28 among patients achieving ypT<2 disease, and given the exploratory nature of the trial, the protocol predefined reporting of ypT0 responses from radical cystectomy and re-TURBT together. We note that one patient in group B who underwent re-TURBT (showing remaining local disease) followed by radical cystectomy was found to have N1 disease and was classified as a nonresponder, which illustrates the residual uncertainty attached to nodal status in patients not undergoing lymph node dissection. We recognize that this complicates comparison with other neoadjuvant trials in MIBC, but decided not to insist on RC and organ loss for endpoint assessment purposes. Patient reluctance to undergo cystectomy has increasingly affected recent neoadjuvant trials, reflecting both the greater efficacy of contemporary neoadjuvant regimens, with a larger proportion of patients achieving a complete response, and advances in staging, in the quality of transurethral resection and in the tolerability of these regimens relative to platinum-based chemotherapy. Together, these developments have made patients more willing to consider bladder preservation, even though the long-term oncologic safety of that approach has not been established.
Accounting for these limitations, the 33% pCR rate (47% in the central assessment) obtained with nivolumab plus visugromab is promising and within the range of the pCR (that is ypT0) proportions reported by the plethora of most active neoadjuvant studies using ICIs6. Radiographic and pathologic assessments were directionally concordant, but the agreement was imperfect: the centrally assessed radiographic complete response rate (13%) was considerably lower than the centrally assessed pCR rate (47%) in the combination group, whereas local assessment produced the opposite relationship. Imaging response alone therefore cannot be regarded as a surrogate for pathologic response in this setting, and its performance is likely to depend on center experience and on the imaging modality used.
Additional limitations include the relatively short duration of follow-up and the use of a protocol-defined efficacy-evaluable population for assessment of the pCR endpoint, both of which should be considered when interpreting the results and comparing them with other neoadjuvant MIBC trials.
In summary, the GDFATHER-NEO trial provides the first evidence that combined GDF-15 and PD-1 blockade may potentiate anti-PD-1 activity substantially in patients with previously untreated, checkpoint inhibitor-naive cancer, here in neoadjuvant treatment of MIBC. The combination with a Q4W schedule appears feasible and well tolerated. This tolerability may allow further combination into existing first-line immunotherapy schemes such as enfortumab vedotin plus pembrolizumab in neoadjuvant treatment of MIBC and treatment of advanced/metastatic urothelial cancer. Randomized phase 2b trials of visugromab are recruiting in treatment-naive advanced disease (NSCLC, NCT07098988) and in checkpoint inhibitor-refractory disease (NSCLC, NCT07246863 and hepatocellular carcinoma, NCT07219459) and will further inform on the ultimate clinical significance of GDF-15 blockade in these settings. Larger randomized trials in the neoadjuvant setting with long-term follow-up should be able to establish whether GDF-15 neutralization can improve bladder preservation, cure rates and overall survival in localized, immunoresponsive tumors such as MIBC.
Methods
Trial design and oversight
GDFATHER-NEO (NCT06059547; EudraCT 2022-002816-22; EU CT 2024-513957-55-00) is an exploratory, stratified, participant-blinded, parallel-group phase 2a trial of neoadjuvant PD-1 blockade with or without GDF-15 neutralization in patients with MIBC who are ineligible for or who decline cisplatin-based chemotherapy. Before allocation, patients were stratified by clinical tumor stage (T2 versus T3–T4a) and by PD-L1 expression on tumor and immune cells, determined by local pathology as a combined positive score (CPS ≥10 versus <10), into one of four strata: T2 and CPS <10, T2 and CPS ≥10, T3–T4a and CPS <10, and T3–T4a and CPS ≥10 (Fig. 1b).
Within each stratum, patients were allocated 1:1 to nivolumab plus visugromab (group B) or to nivolumab plus placebo (group A). Allocation was performed centrally and sequentially within each stratum (A–B–A–B, …) until at least 12 efficacy-evaluable patients had been allocated to each group. Sequential allocation with blinding of patients, rather than randomization, was chosen because of the small size of the individual strata, to prevent imbalance between treatment arms, and because this was the first evaluation of combined GDF-15 and PD-1 blockade in patients with early-stage disease treated with curative intent. Investigators and sponsor staff were aware of treatment allocation; patients were not. Local pathologists and the independent central pathology and imaging reviewers assessed responses without knowledge of treatment allocation.
The trial was sponsored by CatalYm GmbH (Munich, Germany). The protocol and its amendments were approved by the competent ethics committee of the coordinating center (Comitato Etico Territoriale Lombardia 1) and by the Italian regulatory authority (Agencia Italiana del Farmaco), and the trial was conducted in accordance with the principles of the Declaration of Helsinki, the guidelines for Good Clinical Practice of the International Council for Harmonisation and applicable local law. All patients provided written informed consent before any trial-specific procedure and received no compensation for participation.
The clinical cutoff date was 15 July 2026 for the analyses reported here and was supplemented by central review, bioanalytical and translational data. Data management, biometrics, on-site monitoring and bioanalytical and translational analyses were performed by, or under the oversight of, the sponsor. Independent central review of surgical specimens (pathologic response) and of imaging (radiographic response) was performed by Discovery Life Sciences Biomarker Services GmbH and by Median Technologies, respectively.
Participants
Patients aged ≥18 years with newly diagnosed cT2–T4a N0 M0 MIBC who were ineligible for cisplatin-based chemotherapy according to the modified Galsky criteria29 or who declined such therapy, and who were considered fit to undergo radical cystectomy after 12 weeks of neoadjuvant therapy, were eligible. Clinical staging by CT was mandatory; positron emission tomography (PET)–CT and multiparametric MRI (mpMRI) were optional. Archival tumor material from a TURBT performed within 3 months before the start of treatment had to be available as a paraffin block. Only patients with an Eastern Cooperative Oncology Group performance status of 0 or 1 were eligible.
Further inclusion criteria were adequate bone marrow function (haemoglobin ≥10 g dl−1, platelet count ≥100 × 109 l−1, leukocyte count ≥2.5 × 109 l−1), adequate hepatic function (aspartate and alanine aminotransferase ≤2× the upper limit of normal (ULN); bilirubin ≤1.5 × ULN), adequate renal function (serum creatinine <1.5× ULN and/or creatinine clearance ≥50 ml min−1) and coagulation parameters within the normal range. Patients with type 2 diabetes mellitus had to discontinue metformin, which alters circulating GDF-15 concentrations, and switch to an alternative antidiabetic agent before the start of treatment.
Patients who had received prior systemic anticancer therapy, prior radiotherapy to the bladder tumor or a partial cystectomy, and patients who declined radical cystectomy at the time of trial entry were excluded. Patients with major tissue injury, clinically significant cardiovascular disease, active autoimmune disease, chronic high-dose systemic corticosteroid treatment or known or expected hypersensitivity to visugromab or nivolumab were also excluded. Full eligibility criteria are provided in the trial protocol (Supplementary Information). Sex was recorded as documented in the medical record and was used for descriptive purposes only; because of the small sample size, no sex- or gender-based analyses were prespecified, and none were performed.
Surgical procedures
A TURBT performed according to local practice within 3 months before the start of treatment was required for every patient, for histological characterization and staging. The resulting paraffin-embedded tumor material had to be made available for the additional histological and molecular analyses of the trial, in particular for PD-L1 CPS scoring; residual material was returned to the providing pathology department on request.
The trial comprised a screening phase of up to 4 weeks, beginning on the day written informed consent was obtained, and a core phase of 19 weeks, beginning on the day of the first administration of trial treatment. The core phase comprised the three 4-week immunotherapy cycles and a window of 4–8 weeks after the last neoadjuvant dose during which surgery was to be performed. Radical cystectomy or re-TURBT was planned at the time of trial entry and provided the surgical specimen (paraffin block) required for the assessment of pathologic response.
After completion of neoadjuvant treatment, and at the discretion of the investigator and the patient, one of the following procedures was performed, in line with the recommendations of the European Association of Urology: RC with pelvic lymph node dissection (PLND); or maximal re-TURBT, with adjuvant chemoradiotherapy or adjuvant nivolumab to be considered as clinically appropriate.
Provision of tumor material for the assessment of pathologic response and of immune cell infiltration was mandatory irrespective of whether a radical or a bladder-sparing approach was chosen. Surgical procedures followed each center’s standard practice. Patients who declined, or who were unsuitable for, RC or re-TURBT after neoadjuvant treatment were managed according to European Association of Urology guidelines, which recommend chemoradiotherapy as the principal organ-sparing option. An end-of-treatment (EoT) visit was scheduled 21 days (±2 days) after completion of the surgical procedure.
Neoadjuvant treatment and procedures
Visugromab (CTL-002) is an investigational, humanized, hinge-stabilized immunoglobulin G4 (IgG4) monoclonal antibody that binds GDF-15 with high affinity. It was administered intravenously over approximately 60 min every 4 weeks (Q4W) at 20 mg kg−1. Pharmacokinetic–pharmacodynamic modeling of data from the first-in-human trial indicated that, in combination with nivolumab, complete neutralization of circulating GDF-15 is achieved with 10 mg kg−1 every 2 weeks or with 20 mg kg−1 every 3–4 weeks.
Nivolumab, a fully human IgG4 monoclonal antibody directed against PD-1, was administered in accordance with its summary of product characteristics30 and local guidelines. The 480 mg Q4W schedule was selected so that both agents could be given on the same 4-weekly visit schedule (Fig. 1b).
In group B, nivolumab and visugromab were given on day 1 of each of the three cycles, with visugromab administered first, and followed by a 30-min observation period before the nivolumab infusion. In group A, patients received placebo (drug-free infusion bags identical in appearance to the visugromab bags) before nivolumab, on the same schedule.
AEs were graded according to the National Cancer Institute Common Terminology Criteria for Adverse Events version 5.031 and were managed according to toxicity-specific algorithms defined in the protocol. Guidance on optional premedication before visugromab and on dose interruption of nivolumab and visugromab in the event of immune-related AEs was provided to investigators in the protocol (Supplementary Information).
Clinical assessments and outcomes
The primary objective was to assess the antitumor activity of visugromab in combination with nivolumab. Secondary objectives were to assess the safety and tolerability of the combination, its pharmacokinetics and pharmacodynamics, the development of antidrug antibodies (ADAs), and clinical outcomes in the neoadjuvant setting. Collection of molecular and genomic profiling data from peripheral blood and tumor tissue was an exploratory objective. Optional 18F-fluorodeoxyglucose PET–CT and mpMRI were permitted for exploratory response assessment.
The two primary endpoints were the rate of pCR, determined by blinded local pathologists, and the rate of radiographic response according to RECIST version 1.132, determined by investigators. Secondary endpoints were AEs and serious AEs (SAEs), ADAs, delay of surgery, the rate of major pathologic response (MPR), event-free survival (EFS), time to relapse and overall survival (OS), standard pharmacokinetic parameters, baseline serum GDF-15 concentrations, and pCR, MPR and radiographic response as determined by independent central review.
pCR was defined as the absence of any residual viable tumor in the RC or re-TURBT specimen. Where a re-TURBT was followed by an RC, the pathologic evaluation of the RC specimen was used. Residual non-muscle-invasive disease (ypTa or ypT1) and residual carcinoma in situ were not considered a pCR. MPR was defined as ypT0/Ta/Tis/T1 N0 M0 after RC or ypT0/Ta/Tis/T1 NX M0 after re-TURBT, and therefore includes pCR. Radiographic response rate was defined as the proportion of patients with a complete or partial response according to RECIST version 1.1. Four protocol amendments took place. One implemented clarifications requested by local authorities (Italy) before the start of patient enrolment (removal of need for mandatory hospitalization after first administration of study medication in cycle 1 and extension of screening period to 28 days). A second amendment introduced re-TURBT as a surgical alternative to RC together with the assessment of ADAs as an additional secondary endpoint. A third amendment allowed inclusion of participants with stable atrial fibrillation. Post-enrollment, by the fourth amendment, central pathological review (for pCR) and central radiographic review (for ORR) were formally established as secondary endpoints (complementary central review had been mentioned in early protocol versions already) in line with clarifications on definitions of pCR and MPR.
A safety follow-up visit was scheduled 90 days (±3 days) after the last neoadjuvant dose to capture all AEs and SAEs occurring after neoadjuvant treatment, including postoperative complications. Longer-term outcomes were captured at follow-up visits for up to 12 months after RC or re-TURBT.
Exploratory biomarker endpoints, serum visugromab and ADA measurements
For the assessment of total visugromab, serum samples were collected 30 min before infusion and immediately before the end of infusion during cycles 1–3, on days 2 and 8 of cycle 1 and at the EoT visit. For the assessment of ADAs, serum samples were collected 30 min before infusion on day 1 of cycles 1–3 and at the EoT visit. Serum was isolated using standard procedures, cryopreserved at −80°C and analyzed using validated electrochemiluminescence assays for total visugromab and for ADA detection.
Formalin-fixed, paraffin-embedded (FFPE) tumor tissue obtained at screening (TURBT) and at post-treatment RC or re-TURBT was sectioned at 4 μm. After histological confirmation of tumor content on H&E-stained sections, multiplex immunofluorescence was performed using a validated 6 + 1-plex OPAL assay (Akoya Biosciences) on a Leica Bond Rx automated staining platform (BOND-RX software version 7.0.3.216, Leica Biosystems). The panel comprised CD4 (F7.2.38, Cell Marque), CD8 (C8/144B, Agilent), Ki-67 (MIB-1, Agilent), FoxP3 (SP97, Abcam), PD-L1 (SP263, Roche) and GDF-15 (HPA011191, Sigma-Aldrich). Chromogenic immunohistochemistry for pan-cytokeratin (AE1/AE3, Agilent) was subsequently performed on the same slide using a Ventana Discovery Ultra platform (software version 12.5 or higher) to enable segmentation of tumor compartments.
Whole-slide multispectral fluorescence imaging was performed at 20× magnification using an Akoya PhenoImager HT/Vectra Polaris platform (software version 1.0.13), followed by bright-field scanning after pan-cytokeratin staining. Pathologist-assisted digital image analysis was performed using validated Visiopharm algorithms (version 2021.09.2.11085, Visiopharm). Tumor, stromal and adjacent non-neoplastic tissue compartments were segmented on the basis of pan-cytokeratin staining, and predefined immune cell and biomarker-positive populations were quantified as cell densities (cells mm−2) and as relative proportions within each compartment. Staining quality was assessed using on-slide and internal tissue controls, and markers that failed predefined quality criteria were excluded from downstream analyses.
Visiopharm assigns each detected cell to a single, mutually exclusive phenotype; cells co-expressing several markers (for example, CD8+Ki-67+) are therefore not included in the corresponding single-positive populations. To derive total marker-positive cell densities, composite variables were reconstructed by summing all mutually exclusive phenotypes containing the marker of interest. All reconstructed densities were calculated within the same tissue compartment and are expressed as cells per square millimeter.
All assays were performed according to predefined analytical procedures and validation criteria established for exploratory profiling of the tumor microenvironment and of GDF-15 biology.
Evaluation of intratumoral pro-GDF-15 expression in human FFPE samples
Intratumoral pro-GDF-15 expression was determined using a rabbit polyclonal anti-GDF-15 antibody (HPA011191, Sigma-Aldrich) as the primary antibody, with automated staining on a Ventana BenchMark Ultra platform (Roche; software versions 12.3.1 and 12.5.4). Binding of the anti-GDF-15 antibody was visualized using a secondary horseradish peroxidase-conjugated antibody (Ventana OptiView Universal DAB Detection Kit, 760-700; Roche Diagnostics) and a precipitating enzyme reaction product. Slides were evaluated by a board-certified pathologist who scored cytoplasmic staining of tumor cells and, where applicable, membranous staining. Cytoplasmic staining was assigned to four intensity categories, from 0 (no staining) to 3+ (intense staining), and the percentage of stained cells in each category was recorded. The H score, which reflects tumor GDF-15 protein levels, was calculated by summing the products of the percentage of cells in each category and the corresponding intensity value, as follows:
$$\begin{array}{l}{\rm{H}}\; {\rm{score}}=[1\times ({\rm{percentage}}\; {\rm{of}}\; {\rm{intensity}}1)]\\\qquad\qquad+[2\times ({\rm{percentage}}\; {\rm{of}}\; {\rm{intensity}}2)]\\\qquad\qquad+[3\times ({\rm{percentage}}\; {\rm{of}}\; {\rm{intensity}}3)]\end{array}$$
Central evaluation of PD-L1 protein expression in human FFPE samples
PD-L1 expression was assessed centrally by immunohistochemistry using the SP263 assay (790-4905; Roche Diagnostics) according to the manufacturer’s protocol, with a rabbit IgG monoclonal antibody (790-4795; Roche Diagnostics) as isotype control. Staining was evaluated by a board-certified pathologist, and the tumor proportion score (TPS) and CPS were determined as previously described.
Measurement of cytokine and chemokine levels in patient samples
Serum samples for cytokine and chemokine analysis were collected 30 min before infusion on day 1 of cycles 1–3 and at the EoT visit. CXCL9 (MIG) and CXCL10 (IP-10) concentrations were measured using validated ELISAs (Human CXCL9/MIG Quantikine ELISA Kit, DCX900; Human CXCL10/IP-10 Quantikine ELISA Kit, DIP100; R&D Systems). IFN-γ concentrations were determined using a validated multiplex immunoassay (V-PLEX Plus Proinflammatory Panel 1 (human) Kit, K15049G; Meso Scale Discovery). For graphical presentation and statistical analysis, concentrations below the LLOQ were assigned a value of LLOQ/2. All assays were performed in compliance with Good Laboratory Practice.
Statistical analysis
This exploratory, stratified, participant-blinded, parallel-group phase 2a trial was descriptive in design. No formal sample size calculation was performed, and no hypothesis was tested; the trial was not powered to detect a difference between treatment groups, and all comparisons between groups are exploratory. Continuous variables are summarized by number of observations, mean, median, minimum and maximum; categorical variables are summarized by counts and percentages. Percentages are calculated using the number of evaluable patients in the relevant analysis population as the denominator. Presented confidence intervals are 95% exact binomial (Clopper–Pearson) confidence intervals.
The safety analysis population comprised all patients who received at least one dose of trial treatment; it was identical to the intention-to-treat population. Safety data are summarized by treatment group and visit. The efficacy analysis population comprised all patients who received at least one infusion and underwent RC or re-TURBT, unless surgery had been precluded by confirmed disease progression or by discontinuation for toxicity. Separate efficacy analysis populations were defined for locally and for centrally assessed pathologic and radiographic response, as detailed in Fig. 1a. In case of early discontinuation due to confirmed disease progression or toxicity, participants were considered as nonresponders for the primary and secondary endpoint analyses of histopathological and radiographic response.
Two-group comparisons of continuous variables were made using the two-sided Mann–Whitney U test. Because of the small sample sizes and the potentially nonnormal distributions of biomarker measurements, no assumption of normality was made, and comparisons were performed using this nonparametric test. No adjustment was made for multiple comparisons, and no covariates were included in the exploratory biomarker comparisons. Exact P values are reported in the corresponding figures and figure legends; values exceeding the software reporting limit are shown as P > 0.9999. All P values are reported descriptively, and no inferential claim is made, given the exploratory design and the small sample size. Analyses were performed using GraphPad Prism version 11.0.0 (GraphPad Software).
Time-to-event endpoints were summarized using the Kaplan–Meier method. Response rates are presented as the number and proportion of responders in each treatment group.
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
Data availability
Individual participant data underlying this Article are held by the sponsor. In line with Regulation (EU) No. 536/2014 and with the requirements for transparent reporting of clinical trials in WHO primary registries, aggregated results will be posted to the trial registries and a full clinical trial report will be prepared. Additional source data files supporting the findings of this trial are provided as Supplementary Information. Requests for access to de-identified individual participant data may be submitted to the sponsor at info@catalym.com and will be reviewed on the basis of a methodologically sound proposal, subject to the terms of the informed consent obtained and to the General Data Protection Regulation; approved requests require a data access agreement. Source data are provided with this paper.
Code availability
No custom code or mathematical algorithm was specifically developed for this study. Software and packages used in the analyses can be found in the Methods or in the Reporting Summary.
References
Filho, A. M., Briganti, A., Jemal, A. & Bray, F. Bladder cancer incidence and mortality: a global overview and recent trends. Eur. Urol. 89, 426–436 (2026).
Powles, T. et al. ESMO Guidelines Committee. Bladder cancer: ESMO clinical practice guideline for diagnosis, treatment and follow-up. Ann. Oncol. 33, 244–258 (2022).
Galsky, M. D., Ballas, L. & Kamat, A. M. Muscle-invasive bladder cancer: a watershed moment. J. Clin. Oncol. 43, 3041–3051 (2025).
Powles, T. et al. Clinical efficacy and biomarker analysis of neoadjuvant atezolizumab in operable urothelial carcinoma in the ABACUS trial. Nat. Med. 25, 1706–1714 (2019).
Necchi, A. et al. Updated results of PURE-01 with preliminary activity of neoadjuvant pembrolizumab in patients with muscle-invasive bladder carcinoma with variant histologies. Eur. Urol. 77, 439–446 (2020).
Necchi, A. et al. Immunotherapy in the treatment of localized genitourinary cancers. JAMA Oncol 9, 1447–1454 (2023).
Powles, T. et al. Perioperative durvalumab with neoadjuvant chemotherapy in operable bladder cancer. N. Engl. J. Med. 391, 1773–1786 (2024).
Galsky, M. D. et al. Enfortumab vedotin and pembrolizumab in cisplatin-eligible bladder cancer. N. Engl. J. Med. 395, 338–348 (2026).
Vulsteke, C. et al. Perioperative enfortumab vedotin and pembrolizumab in bladder cancer. N. Engl. J. Med. 394, 1257–1269 (2026).
Powles, T. et al. ctDNA-guided adjuvant atezolizumab in muscle-invasive bladder cancer. N. Engl. J. Med. 393, 2395–2408 (2025).
Necchi, A. et al. Gemcitabine intravesical system plus cetrelimab or cetrelimab alone as neoadjuvant therapy in muscle-invasive bladder cancer: SunRISe-4 primary analysis and biomarker results. J. Clin. Oncol. 44, 586–597 (2026).
Awada, G. et al. The rapidly evolving paradigm of neoadjuvant immunotherapy across cancer types. Nat. Cancer 6, 967–987 (2025).
Suartz, C. V. et al. Neoadjuvant immunotherapy in bladder cancer: ushering in a new era of treatment—a systematic review of current evidence. Eur. Urol. Open Sci. 79, 45–59 (2025).
Chang, E. et al. Refining neoadjuvant therapy clinical trial design for muscle-invasive bladder cancer before cystectomy: a joint US Food and Drug Administration and Bladder Cancer Advocacy Network workshop. Nat. Rev. Urol. 19, 37–46 (2022).
van Dijk, N. et al. Preoperative ipilimumab plus nivolumab in locoregionally advanced urothelial cancer: the NABUCCO trial. Nat. Med. 26, 1839–1844 (2020).
Haake, M. et al. Tumor-derived GDF-15 blocks LFA-1 dependent T cell recruitment and suppresses responses to anti-PD-1 treatment. Nat. Commun. 14, 4253 (2023).
Melero, I., Klar, K. & Leo, E. GDF-15 blockade: a multi-directional approach to potentiate cancer immunotherapy and alleviate cancer cachexia. Clin. Transl. Med. 15, e70280 (2025).
Melero, I. et al. Neutralizing GDF-15 can overcome anti-PD-1 and anti-PD-L1 resistance in solid tumours. Nature 637, 1218–1227 (2025).
Melero, I. et al. Long-term follow-up of a phase 1/2 trial of anti-GDF-15 antibody visugromab plus anti-PD-1 antibody nivolumab in anti-PD-1/-L1 relapsed/refractory solid tumors. J. Hematol. Oncol. 19, 64 (2026).
Wischhusen, J., Melero, I. & Fridman, W. H. Growth/differentiation factor-15 (GDF-15): from biomarker to novel targetable immune checkpoint. Front. Immunol. 11, 951 (2020).
Molina-Cerrillo, J. et al. Revisiting neoadjuvant chemotherapy in cisplatin-eligible muscle-invasive bladder cancer. Eur. Urol. Oncol. 9, 168–180 (2026).
Blanc, J. et al. Avelumab-based neoadjuvant therapy in patients with muscle-invasive bladder cancer (AURA Oncodistinct-004): a phase 2 multicenter clinical trial. J. Immunother. Cancer. 13, e012045 (2025).
Grande, E. et al. DUTRENEO trial: a randomized phase II trial of DUrvalumab and TREmelimumab versus chemotherapy as a NEOadjuvant approach to muscle-invasive urothelial bladder cancer (MIBC) patients (pts) prospectively selected by an interferon (INF)-gamma immune signature. J. Clin. Oncol. 38, 5012 (2020).
Grivas, P. et al. PrECOG PrE0807: a phase 1b feasibility trial of neoadjuvant nivolumab without and with lirilumab in patients with muscle-invasive bladder cancer ineligible for or refusing cisplatin-based neoadjuvant chemotherapy. Eur. Urol. Oncol. 7, 914–922 (2024).
Hong, G. et al. Plasma GDF15 levels associated with circulating immune cells predict the efficacy of PD-1/PD-L1 inhibitor treatment and prognosis in patients with advanced non-small cell lung cancer. J. Cancer Res. Clin. Oncol. 149, 159–171 (2023).
Schulz, K. F. et al. ‘Allocation concealment’: the evolution and adoption of a methodological term. J. R. Soc. Med. 111, 216–224 (2018).
Necchi, A. et al. Pembrolizumab as Neoadjuvant Therapy Before Radical Cystectomy in Patients with Muscle-Invasive Urothelial Bladder Carcinoma (PURE-01): an open-label, single-arm, phase II study. J. Clin. Oncol. 36, 3353–3360 (2018).
Mercinelli, C. et al. First results of NURE-combo: a phase II study of neoadjuvant nivolumab and nab-paclitaxel, followed by postsurgical adjuvant nivolumab, for muscle-invasive bladder cancer. J. Clin. Oncol. 42, 4196–4205 (2024).
Jiang, D. M. et al. Defining cisplatin eligibility in patients with muscle-invasive bladder cancer. Nat. Rev. Urol. 18, 104–114 (2021).
Opdivo SmPC (marketing authorisation holder: Bristol-Myers Squibb Pharma EEIG). European Medicines Agency https://www.ema.europa.eu/en/medicines/human/EPAR/opdivo#product-info (2026).
US Department of Health and Human Services. Common Terminology Criteria for Adverse Events (CTCAE) v5.0. U.S. National Cancer Institute, Division of Cancer Therapy and Diagnosis https://ctep.cancer.gov/protocolDevelopment/electronic_applications/docs/CTCAE_v5_Quick_Reference_5x7.pdf (2017).
Eisenhauer, E. A. et al. New response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1). Eur. J. Cancer 45, 228–247 (2009).
Acknowledgements
The investigators thank the participants and the staff members at the trial sites who managed and cared for them; the members of the data monitoring committee; F. Montorsi and A. Briganti (both Milan) for contributing patients for this trial; G. Brambilla (Milan) for imaging response review; the staff of Precision for Medicine (Edinburgh, UK) and of Metronomia (Munich, Germany) involved in data collection and analyses; the staff of Discovery Life Sciences (Kassel, Germany) for performing the central histopathological assessment of the tumor tissue specimens; M. Foster, consultant pathologist, for expert histopathological review; and M. Hartmann from Ecc-Oncology (Trier, Germany) for medical writing. The funder of the trial provided nivolumab free of cost and provided third-party funding for trial coordination, documentation, monitoring and data management. Medical writing and editing support, funded by CatalYm GmbH, was provided by Markus Hartmann, PhD, European Consulting & Contracting in Oncology, Trier, Germany.
Funding
The trial was sponsored and funded by CatalYm GmbH (Munich, Germany), which financially supported all trial-related procedures and provided the investigational medicinal products. No specific trial-related grants supported this work except for J.W.
Author information
Author notes
These authors jointly supervised this work: Eugen Leo, Felix S. Lichtenegger.
Authors and Affiliations
Department of Medical Oncology, IRCCS San Raffaele Hospital, Comprehensive Cancer Center, Milan, Italy
Andrea Necchi, Chiara Mercinelli, Brigida A. Maiorano, Michela Piacentini & Antonio Cigliola
Vita-Salute San Raffaele University, Milan, Italy
Andrea Necchi, Chiara Mercinelli & Antonio Cigliola
Department of Oncology, Fondazione Policlinico Universitario Agostino Gemelli, Rome, Italy
Roberto Iacovelli, Chiara Ciccarese & Chiara Ligato
Translational Medicine and Surgery, Università Cattolica del Sacro Cuore, Rome, Italy
Roberto Iacovelli, Chiara Ciccarese & Chiara Ligato
Department of Medical Oncology, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy
Patrizia Giannatempo & Alessandro Rametta
Division of Urology, Department of Surgical Sciences, San Giovanni Battista Hospital, Turin, Italy
Paolo Gontero & Francesco Soria
University of Turin, Turin, Italy
Paolo Gontero
CatalYm GmbH, Munich, Germany
Christine Schuberth-Wagner, Bernhard Rössler, Michael Metterlein, Stefan Ries, Thorsten Ross, Aasim Amin, Tobias Dreker, Frank Hermann, Lena Lemke, Kathrin Klar, Olga Mitina, Ana Kamova, Petra Fettes, Alexandra Auckenthaler, Sujata Rao, Eugen Leo & Felix S. Lichtenegger
Department of Gynecology, University Hospital Würzburg, Würzburg, Germany
Joerg Wischhusen
Authors
- Andrea Necchi
- Chiara Mercinelli
- Brigida A. Maiorano
- Michela Piacentini
- Antonio Cigliola
- Roberto Iacovelli
- Chiara Ciccarese
- Chiara Ligato
- Patrizia Giannatempo
- Alessandro Rametta
- Paolo Gontero
- Francesco Soria
- Christine Schuberth-Wagner
- Joerg Wischhusen
- Bernhard Rössler
- Michael Metterlein
- Stefan Ries
- Thorsten Ross
- Aasim Amin
- Tobias Dreker
- Frank Hermann
- Lena Lemke
- Kathrin Klar
- Olga Mitina
- Ana Kamova
- Petra Fettes
- Alexandra Auckenthaler
- Sujata Rao
- Eugen Leo
- Felix S. Lichtenegger
Contributions
A.N., E.L., F.S.L., F.H. and J.W. conceptualized the trial. A.N., C.M., B.A.M., M.P., A.C., R.I., C.C., C.L., P. Giannatempo, A.R., P. Gontero., F.S., A. Amin, M.M., L.L., O.M., A.K., P.F., A. Auckenthaler and F.S.L. were responsible for data curation; A.N. had access to the clinical trial’s raw data. T.R., K.K., A. Amin, T.D., L.L., O.M., A.K., P.F., A. Auckenthaler, S. Rao, E.L. and F.S.L. performed the formal analysis. A.N., C.M., B.A.M., M.P., A.C., R.I., C.C., C.L., P. Giannatempo, A.R., P. Gontero. and F.S. conducted the clinical investigations. E.L., F.S.L., F.H., T.R., K.K. and P.F. were responsible for the methodology. B.R., O.M., A.K., A. Auckenthaler and P.F. were responsible for project administration. E.L. and C.S.W. were responsible for funding acquisition. A.N., E.L., F.S.L., S. Rao, S. Ries, K.K. and P.F. supervised the project. A. Amin, T.D., L.L., K.K., T.R., E.L. and F.S.L. performed validations. T.R., K.K., A. Amin, T.D., L.L., O.M., A.K., E.L. and F.S.L. generated visualizations. All authors participated in the writing or critical review and editing of the paper. All authors approved the paper for submission for publication.
Corresponding author
Correspondence to Andrea Necchi.
Ethics declarations
Competing interests
J.W. is a co-founder of the biotech company CatalYm GmbH, owner of GDF-15 related patents and stock owner of CatalYm GmbH. C.S.W., B.R., M.M., S. Ries, T.R., K.K., A. Amin, T.D., F.H., L.L., O.M., A.K., P.F., A. Auckenthaler, S. Rao., E.L. and F.S.L. are current or former employees of CatalYm GmbH and/or hold stock options of CatalYm GmbH. J.W. has received research funding or consulting and/or advisory fees from CatalYm GmbH. C.S.W., J.W., K.K., M.M., F.H. and E.L. are inventors on patents related to visugromab and anti-GDF-15 treatment. A.N., C.M., B.A.M., M.P., A.C., R.I., C.C., C.L., P. Giannatempo, A.R., P. Gontero and F.S. were clinical investigators; investigator fees were paid to their institutions. Several coauthors declare potentially competing interests (grants or contracts and/or royalties or licenses, stock or stock options, consulting fees, honoraria, travel or meeting support, or advisory or data safety monitoring board participation) apart from their relationship with CatalYm GmbH: A.N. (grants and consulting fees from AstraZeneca, Bristol-Myers Squibb, Merck Sharp & Dohme and Gilead Sciences; consulting fees from Johnson & Johnson, Samsung Bioepis, Daiichi Sankyo and Pfizer; meeting/travel support from Gilead, Johnson & Johnson and Merck Sharp & Dohme); C.M. (honoraria from Astellas and AstraZeneca; travel/meeting support from Bayer, Ipsen and Johnson & Johnson); B.A.M. (research funding from Merck Serono; consulting/advisory roles for Astellas, AstraZeneca, Bayer, Johnson & Johnson, Ipsen, Merck Serono and Merck Sharp & Dohme; meeting/travel support from Ipsen, Janssen, Novartis, Pfizer); A.C. (honoraria from Astellas and Merck Sharp & Dohme; meeting/travel support from Ipsen and Merck, Sharp & Dohme; advisory or safety monitoring board activities for Johnson & Johnson); R.I. (consulting fees and honoraria from Astellas, Bristol-Myers Squibb, Johnson & Johnson and Merck Sharp & Dohme; advisory or safety monitoring board activities for Astellas, Bristol-Myers Squibb, Johnson & Johnson, Merck Sharp & Dohme and Pfizer); C.C. (advisory board roles and/or meeting/travel support from Astellas, AstraZeneca, Bristol-Myers Squibb, Eisai, Ipsen, Janssen, Merck Sharp & Dohme, Novartis, Pfizer); P. Giannatempo (consulting fees, honoraria and travel/meeting support from Astellas, AstraZeneca, Johnson & Johnson, Merck and Merck Sharp & Dohme); P. Gontero (research grants and meeting/travel support from Intuitive Surgical and CRT Foundation; Consulting fees from AstraZeneca, Merck Sharp & Dohme, Pfizer and Photocure; honoraria from AstraZeneca, Ferring, Medac and Merck); F.S. (consulting fees from Accord, AstraZeneca, Cepheid, Ferring, Johnson & Johnson, Medac, Merck Sharp & Dohme, Pfizer and Photocure; honoraria from AstraZeneca, Johnson & Johnson and Medac; advisory or safety monitoring board activities for Accord, AstraZeneca, Cepheid, Ferring, Johnson & Johnson, Medac, Merck Sharp & Dohme and Pfizer); J.W. (research grants (Transcan-3) from the German Ministry for Education & Research; inventor royalties from patents owned by the Julius-Maximilians-University Würzburg); F.H. (consulting fees, co-ownership and stock options of Avenor Therapeutics; employment at Thermosome until March 2026); L.L. (stock options held from Novartis); and P.F. (stock options from Amgen and BioNTech). M.P., C.L. and A.R. declared no competing interests.
Peer review
Peer review information
Nature Medicine thanks the anonymous reviewers for their contribution to the peer review of this work. Primary Handling Editor: Ulrike Harjes, 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 Kaplan Meier estimates for event-free survival (EFS).
Depicted is the ITT population for groups A (nivolumab plus placebo) and B (nivolumab plus visugromab).
Extended Data Fig. 2 Baseline multiplex immunofluorescence characterization of additional immune and tumor-cell populations.
Baseline densities of CD4+ T cells, regulatory T cells (CD4+FoxP3+), proliferating CD4+ T cells (CD4+Ki67+), proliferating regulatory T cells (CD4+FoxP3+Ki67+), total panCK+ tumor cells, proliferating tumor cells (panCK+Ki67+), total GDF15+ cells, and GDF15+ tumor cells in tumors from the monotherapy (nivolumab plus placebo, N + P) and combination therapy (nivolumab plus visugromab, N + V) groups. Each data point represents one patient. Horizontal lines indicate medians. The distribution of individual observations is shown. Statistical comparisons were performed using two-sided Mann–Whitney U tests, as described for Fig. 4. mIF staining and analysis were performed once per patient specimen; each data point represents one patient.
Extended Data Fig. 3 Baseline circulating and intratumoral GDF-15 levels across treatment and pathologic response groups and their association with T-cell infiltration.
(a) Baseline serum GDF-15 concentrations and intratumoral GDF-15 H-scores in the monotherapy and combination therapy groups. (b) Baseline serum GDF-15 concentrations and intratumoral GDF-15 H-scores stratified by pathologic response within each treatment group. (c) Baseline densities of total T cells, total CD8+ T cells, proliferating CD4/CD8 T cells (CD4+Ki67+/CD8+Ki67+), and proliferating CD8+ T cells (CD8+Ki67+) stratified by median serum GDF-15 concentration (top) or intratumoral GDF-15 H-score (bottom), with data pooled across treatment groups. Each data point represents one patient. Horizontal lines indicate medians. The distribution of individual observations is shown. Statistical comparisons were performed using two-sided Mann–Whitney U tests, as described for fig. 4. mIF staining and analysis were performed once per patient specimen; each data point represents one patient.
Extended Data Fig. 4 Baseline circulating cytokines and chemokines and longitudinal CXCL-10 dynamics.
(a) Baseline circulating concentrations of CXCL9, CXCL10 and IFN-γ in the nivolumab plus placebo (N + P) and nivolumab plus visugromab (N + V) groups. Each data point represents one baseline serum sample from one individual patient. The horizontal dotted line indicates the lower limit of quantification (LLOQ). Horizontal lines indicate medians, and the distribution of individual observations is shown. Baseline comparisons between treatment groups were performed using two-sided Mann–Whitney U tests. (b) Longitudinal circulating CXCL10 concentrations measured at baseline (cycle 1 day 1, C1D1), cycle 2 day 1 (C2D1), cycle 3 day 1 (C3D1) and end of treatment (EoT). Data points represent individual patient measurements, and lines connect measurements obtained from the same patient at different time points. Panel b is presented descriptively; no formal statistical comparison of longitudinal changes was performed. Values below the LLOQ were assigned a value of LLOQ/2. n represents individual patients with an available measurement at the indicated time point. mIF staining and analysis were performed once per patient specimen; each data point represents one patient.
Extended Data Fig. 5 Visugromab serum concentrations over time.
Observed total visugromab serum concentrations during treatment and at end-of-therapy (EoT). Individual observations and geometric mean concentrations are shown.
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
Necchi, A., Mercinelli, C., Maiorano, B.A. et al. Neoadjuvant nivolumab with or without visugromab in muscle-invasive bladder cancer: the phase 2a GDFATHER-NEO trial. Nat Med (2026). https://doi.org/10.1038/s41591-026-04740-9
Received: 07 September 2026
Accepted: 06 October 2026
Published: 24 October 2026
Version of record: 24 October 2026
DOI: https://doi.org/10.1038/s41591-026-04740-9
来源:Nature 健康护理主题 · nature.com