Beyond Platinum, ICIs in Metastatic Cervical Cancer: A Systematic Review

Simple Summary Approaches beyond first-line chemotherapy to treat advanced cervical cancer (CC) are currently limited. Immune checkpoint inhibitors (ICIs) are showing high efficacy, thus remodeling the therapeutic scenario of many solid tumors. With our systematic review, we aimed to summarize the latest clinical trials using ICIs in CC. Our systematic review managed to demonstrate that ICIs might represent an appealing strategy for advanced CC, with 2 out of 3 patients responding to ICIs without further concerns about safety. PD-L1 status might be an indicator of response; however, the search for new predictive biomarkers is mandatory. Further studies are needed for appropriate patient selection and a tailored approach. Abstract Background: Cervical cancer (CC) constitutes the fourth most common tumor among the female population. Therapeutic approaches to advanced CC are limited, with dismal results in terms of survival, mainly after progression to platinum-based regimens. Immune checkpoint inhibitors (ICIs) are remodeling the therapeutic scenario of many solid tumors. The role of ICIs in CC should be addressed. Therefore, we systematically reviewed the latest clinical trials employing ICIs in advanced CC to assess which ICIs have been employed and how ICIs might meet the need for new therapeutic options in terms of efficacy and safety. Methods: The review was conducted following the PRISMA guidelines. The following efficacy outcomes were specifically collected: overall response rate (ORR), disease control rate (DCR), progression-free survival (PFS), and overall survival (OS); for safety: type, number, and grade of adverse events (AEs). Results: A total of 17 studies were analyzed. Anti-PD1 (pembrolizumab, nivolumab, cemiplimab, balstilimab, and tislelizumab), anti-PD-L1 (atezolizumab), and anti-CTLA-4 (ipilimumab, zalifrelimab) agents were employed both as single agents or combinations. Overall ORR ranged from 0% to 65.9%. ORR ranged from 5.9% to 69.6% in PD-L1-positive patients and from 0% to 50% in PD-L1-negative patients. DCR was 30.6–94.1%. mPFS ranged from 2 to 10.4 months. mOS ranged from 8 months to not reached. PD-L1 status did not impact survival. A total of 33.9% to 100% of patients experienced AEs. Conclusion: Immunotherapy represents an appealing strategy for patients with advanced CC, as 2 out of 3 patients seem to respond to ICIs. PD-L1 status might be an indicator of response without impacting survival.


Introduction
With an incidence of 15.6 per 100,000 inhabitants per year, cervical cancer (CC) represents the fourth most common cancer among the female population, as well as the fourth cause of cancer-related death worldwide, bearing a mortality rate of 8.8 deaths per 100.000 inhabitants per year [1]. The median age at diagnosis is 49. A dual peak of incidence of CC is registered among the 40-64 and the 65-74 age subgroups, respectively, with 1.8 and 2.4 cases per 100,000 inhabitants per year [1,2]. The 5-year relative survival is 66.3%, as CC is often diagnosed at an early stage on account of early human papillomavirus (HPV) infection detection, thus presenting localized in 44% of cases or spread to regional lymph nodes in 36% of cases. However, 16% of CC cases are diagnosed at the metastatic stage, with a 5-year relative survival dropping to only 17.6% [3].

Treatment Options in Advanced CC
Surgery or definitive radiotherapy are considered the primary treatments for earlystage disease, while concurrent platinum-based chemotherapy and radiotherapy (CTRT) represent the standard of care in the locally advanced disease setting. RT or CTRT is also feasible for recurrences after surgery without previous adjuvant RT. Instead, pelvic exenteration remains the only therapeutic option for women with central pelvic recurrence after RT. Pharmacological approaches to patients with distant or loco-regional recurrences, not eligible for surgery or RT, are currently limited [4,5].
In the metastatic setting, platinum-based chemotherapy plus bevacizumab is used as the first choice, with a median overall survival (mOS) of 17 months. However, therapeutic options after progression to first-line therapy are limited, and survival is dismal in this stage, with less than one year of OS [6][7][8][9]. Thus, the search for new therapeutic approaches is an unmet need for advanced CC.

HPV Infection in CC
Persistent HPV infection is commonly known as the cause of nearly all CC cases, with HPV-16 alone responsible for over 50% of all CCs globally, particularly among the Caucasian population [10][11][12][13]. Despite progress in early HPV detection and extensive vaccination programs, CC still holds one of the highest burdens of disease globally, notably in lowincome countries, thus having a significant impact on women's health worldwide [1][2][3]. HPV infection determines the production of E6 and E7 proteins, with an inhibitory role for the onco-suppressors p53 and Rb [14]. It has been widely demonstrated that HPV infection is responsible for a specific immune response, as an HPV16 E2-and E6-targeted T-helper immune response has been shown in healthy subjects, which might be crucial for controlling HPV infections. HPV can boost the immune response, recruiting E6-and E7-specific T cells, but this mechanism seems lacking in CC patients. Hence, an impaired CD4 + T-cell immunity against E2 and E6 antigens has been seen among CC patients, mainly lacking Interferon (IFN)-gamma and Interleukin (IL)-5 production, if compared with healthy subjects [10]. Activating the immune response against the viral infection represents an attractive approach for therapies targeting the immune system, such as immune checkpoint inhibitors (ICIs).

ICIs and PD-L1 in CC and Aims of the Systematic Review
During the last 10 years, ICIs have modified the therapeutic landscape of many solid tumors, and their application in gynecological malignancies has been intensively investigated [15,16]. Removing the brake pedal by inhibiting negative immune checkpoints such as Programmed Death 1 (PD1), PD-Ligand 1 (PD-L1), and Cytotoxic T-lymphocyteassociated protein 4 (CTLA4), ICIs could produce a robust antitumor activity [15][16][17].
Compared to endometrial and ovarian cancer, CC has a higher rate of PD-L1, as up to 80% of squamous and around 65% of adenocarcinomas are PD-L1-positive, and CD8 + T cells express PD1 [18,19]. However, while the increased expression of PD-L1 has been associated with poorer prognosis or lower OS in other tumor subtypes, this is not the case for CC [20]. PD-L1 has already been addressed as potentially associated with a better ICIs response in CC patients. Additionally, other factors could justify a suitable response to ICIs: a high tumor mutational burden (TMB) and, therefore, a high neo-antigens load that can stimulate immune activation; also, around 8-10% of CC carry a deficit of mismatch-repair genes leading to microsatellite instability (MSI) [21].
ICIs trials have also been conducted in CC. However, there is currently a gap in knowledge regarding the role of ICIs in the treatment scenario of advanced CC patients. We hereby systematically reviewed the latest clinical trials regarding the use of ICIs for CC treatment to address which agents have been employed and assess how ICIs might meet the need for new therapeutic options, notably in the advanced or recurrent CC setting, in terms of response rate and survival, and, finally, the safety profile. To our knowledge, this is the first systematic review analyzing the use of ICIs in the advanced CC setting.

Protocol Registration
We registered the protocol for this systematic review with PROSPERO (CRD42022314512).

Search Strategy and Data Extraction
This systematic review was carried out following the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) statements [22]. Two authors (MFPM and BAM) independently performed a literature search of the databases PubMed, EMBASE, and Cochrane Central Register of Controlled Trials, on March 31, 2022. The search terms ("cervical neoplasms" OR ("cervical" AND "neoplasms") OR "cervical cancer" OR ("cervical" AND "cancer") OR "cervix cancer" OR ("cervix" AND "cancer") AND (''immune checkpoint inhibitors" OR "ICIs" OR "avelumab" OR "nivolumab" OR "atezolizumab" OR "pembrolizumab" OR "durvalumab" OR "cemiplimab" OR "tremelimumab" OR "ipilimumab" or "dostarlimab" OR "balstilimab" OR "camrelizumab") were used. An additional search for conference abstracts from the American Association of Clinical Oncology (ASCO), European Society of Medical Oncology (ESMO), Society of Gynecologic Oncology (SGO) was performed. Article citations were manually checked for additional references.

Inclusion and Exclusion Criteria, Population, Intervention, and Outcomes
We included phase I-IV clinical trials reporting efficacy and safety data of ICIs (single agents or combinations) in metastatic/recurrent CC patients, written in the English language. From multicohort trials, the number and data of CC patients were identified. Metaanalyses, reviews, case reports, correspondences, personal opinions, and in vitro/animal studies were excluded. For the selected studies, the following data were collected: trial name, first author, year of publication, phase, number of treated patients, administered drugs and dosage, and primary and secondary endpoints. We specifically addressed the following efficacy outcomes: overall response rate (ORR), disease control rate (DCR), progression-free survival (PFS), and overall survival (OS) for safety, number, and grade of treatment-related adverse events (AEs) (Supplementary Table S1).

Risk of Bias
Two reviewers independently assessed the risk of bias. In case of disagreement, a third reviewer would have been consulted. The Risk Of Bias In Non-randomized Studies of Interventions (ROBINS-I) tool was used to assess the risk of bias, including eight domains: confounding bias; selection bias; classification bias; deviation from intended interventions bias; missing data; measure outcome bias; selection of the reported results; overall bias [23].

Results
A total of 124 studies were identified via electronic research. A total of 116 studies were eligible after duplicate removal and screening based solely on title and abstract analyses. A total of 6 studies were written in languages other than English; 30 case reports, reviews, correspondences, personal opinions, or commentaries were removed; the complete text was not available in 1 study, while 31 were removed for focusing on different topics after applying inclusion and exclusion criteria. Therefore, a total of 17 studies were included in our review ( Figure 1).
of Interventions (ROBINS-I) tool was used to assess the risk of bias, including eight domains: confounding bias; selection bias; classification bias; deviation from intended interventions bias; missing data; measure outcome bias; selection of the reported results; overall bias [23].

Results
A total of 124 studies were identified via electronic research. A total of 116 studies were eligible after duplicate removal and screening based solely on title and abstract analyses. A total of 6 studies were written in languages other than English; 30 case reports, reviews, correspondences, personal opinions, or commentaries were removed; the complete text was not available in 1 study, while 31 were removed for focusing on different topics after applying inclusion and exclusion criteria. Therefore, a total of 17 studies were included in our review ( Figure 1).

Figure 1. PRISMA flow diagram for selection process.
A total of 124 studies were identified via electronic research. A total of 116 studies were eligible after duplicate removal, and 85 after title and abstract analyses. After checking all the inclusion and exclusion criteria, 17 studies were included in our review.
A total of 8 studies reported DCR, ranging from 30.6% to 94.1%. A total of 13 studies reported mPFS that ranged from 2 to 10.4 months. mOS was reported by 14 studies and ranged from 8 months to NR (Figure 3).  A total of 33.9% to 100% of patients developed AEs, of which up to 81.8% were over grade 3 (≥G3). G3 AEs were identified as severe or medically significant but not immediately life-threatening AEs, according to the Common Terminology Criteria for Adverse Events (CTCAE) definition [42]. In combination, a higher toxicity rate was reported.  A total of 33.9% to 100% of patients developed AEs, of which up to 81.8% were over grade 3 (≥G3). G3 AEs were identified as severe or medically significant but not immediately life-threatening AEs, according to the Common Terminology Criteria for Adverse Events (CTCAE) definition [42]. In combination, a higher toxicity rate was reported. Table 2 resumes the main results of the included studies. No serious risk of bias emerged (Supplementary Figure S1).

Pembrolizumab
Pembrolizumab binds to the PD-1 receptor, blocking immune-suppressing ligands PD-L1 and PD-L2 from interacting with PD-1, and represents one of the most employed ICIs in the daily clinical practice of many solid tumors. Pembrolizumab plays a key role in CC, as after KEYNOTE-158 and KEYNOTE-826 it was approved for PD-L1-positive CC patients as a single agent after chemotherapy progression, and plus chemotherapy in first line [43][44][45]. A total of 4 studies using pembrolizumab were included in our systematic review [24][25][26][27]. Patients were not selected for PD-L1, or HPV, except from one study that included only HPV 16/18 + patients [25].
A total of 617 women with not pre-treated advanced CC were randomized 1:1 in the phase III double-blind KEYNOTE-826 trial (NCT03635567) to receive pembrolizumab 200 mg or placebo (PBO) q3w for up to 35 cycles, added to paclitaxel (175 mg/m2) and investigator's choice between cisplatin (50 mg/m2) or carboplatin (5 mg/mL/min). Patients could also receive bevacizumab 15 mg/kg q3w. OS and PFS by RECIST were co-primary endpoints, while DoR, ORR, and PFS rate at 12 months were secondary endpoints. Three populations were analyzed: PD-L1-positive patients with CPS ≥ 10, PD-L1-positive with CPS 1-10, and all comers. The first interim analysis was pre-planned in the PD-L1-positive patients (defined as CPS ≥ 1). A total of 88.6% of patients in the pembrolizumab group and 89% in the PBO group had PD-L1-positive cancers. Bevacizumab was administered to 63.6% of patients in the pembrolizumab arm and 62.5% in the PBO arm. The results showed that PFS was significantly longer in the pembrolizumab arm compared to PBO, achieving an mPFS of 10.4 months (95% CI, 9.1-12.  49.1%, respectively). A total of 14 possible treatment-related deaths were reported in both arms; in the pembrolizumab group, a slightly higher number of ≥G3 TRAEs (81.8% vs. 75.1%) and more immune-related AEs (irAEs-33.9% vs. 15.2%) were observed, with one death potentially due to an irAE [27].
In conclusion, these data demonstrate that pembrolizumab has efficacy in CC patients, with the maximum advantage in the case of PD-L1 positivity, without particular safety concerns. At the moment, the use in pre-treated patients as monotherapy, and combined with chemotherapy in naïve patients, is justified by these results, which led to the approvals by the FDA in October 2021, and the European Medical Agency (EMA) in April 2022, of pembrolizumab plus chemotherapy combination for the first-line treatment, and pembrolizumab monotherapy for pre-treated patients, in case of CC with PD-L1 positivity defined as CPS ≥ 1 [43,44]. Only one study specifically focused on HPV + patients, which seemed to achieve reasonable disease control with ICIs, especially when HPV and PD-L1 positivity are detected together.

Nivolumab
Nivolumab, another historical anti-PD1 agent, was used as monotherapy in pre-treated patients. Of note, even if most patients were PD-L1-positive, PD-L1 status was not an inclusion criterion for the selected studies [28,29,31]. We also reviewed the studies of nivolumab combinations, as the association with the anti-CTLA4 ipilimumab was the first double ICIs combination used in CC patients [30].

Nivolumab as Single Agent
Nivolumab was tested as a single agent at the flat dose of 240 mg q2w in one phase I/II study and 3 mg/kg q2w in one phase II trial.
Together, these data show that nivolumab is a reliable option in pre-treated CC patients, with higher efficacy in the case of PD-L1 positivity. Differently, no prognostic role of PD-L1 is evidenced with the combination of nivolumab and ipilimumab.
A total of 161 women with advanced pre-treated CC were enrolled to receive balstilimab 3 mg/kg q2w (24 months maximum) in the NCT03104699 phase II trial, whose primary endpoint was ORR, while DoR and DCR were secondary endpoints. A total of Responses were durable, with an mDOR of 15.4 months (95% CI, 5.7 months-NR). ORR among PD-L1 + patients was 20.0% (95% CI, 12.9-29.7%) and 7.9% among PD-L1-negative women. Results were independent of histology or previous treatment with bevacizumab. TRAEs were experienced by 71.4% of patients, the most common being asthenia (23%) and diarrhea (12.4%). ≥G3 TRAEs occurred in 11.8% of patients [32].
In the NCT03495882 phase II trial, 143 patients with pre-treated advanced CC received the combination of balstilimab 3 mg/kg q2w with zalifrelimab 1 mg/kg q6w (up to 2 years). A total of 55% of patients were PD-L1-positive (defined as CPS ≥ 1), 25% PD-L1-negative. ORR-the primary endpoint was 22%, ranging from 11% among PD-L1-negative patients to 27% among PD-L1-positive patients. mDOR was NR. A total of 35% irAES were detected, of whom 10.5% were ≥G3. A total of 10% of patients discontinued the treatment, and two deaths were recorded [33].
These data show that balstilimab is a newer ICI that could be further investigated in CC patients. Similarly to previous trials, there is a slight tendency for higher responses in PD-L1-positive than PD-L1-negative patients, also in the combination studies.
The results of these trials confirm that cemiplimab could be used in pre-treated CC patients, also with rare histologic subtypes, with a potential predictive role of PD-L1 for cemiplimab versus chemotherapy. No safety concerns, but dismal efficacy emerged from the combination of cemiplimab with RT.
In conclusion, these data constitute a rationale for combining ICIs with agents having different mechanisms of action, such as TKIs and chemotherapy.

Tislelizumab
The anti-PD1 tislelizumab 200 mg was administered with anlotinib 10 mg qd for 14 days q3w in a phase II trial. A total of 32 patients were enrolled, while 17 were evaluated as they received at least four cycles of treatment. An ORR of 35.3% emerged (95% CI, 17.3-58.7%), and DCR was 94.1% (95% CI, 73-98.9%). mPFS and OS were NR. All patients experienced G1/2 TRAEs, while only 0.06% experienced G3 TRAEs [38]. The combination of these two agents seems, therefore, effective and safe for CC patients.

ICIs Targeting PD-L1
We subsequently searched for trials using anti-PD-L1 agents in CC patients; only atezolizumab studies were found [39,40].
These data do not seem to support the use of anti-PD-L1 agents in combination with bevacizumab in CC patients.

ICIs Targeting CTLA-4
As in other tumor subtypes, we reviewed the inhibition of the CTLA4 pathway with single agents and found a unique trial with ipilimumab. As monotherapy, ipilimumab was administered to 42 patients in a phase I/II NCT01693783 study at the dosage of 10 mg/kg q3w for four doses, followed by maintenance with four cycles q12w. A total of 1 PR and 10 SDs were recorded, thus resulting in an ORR of 2.9%. mPFS was 2.5 months (95% CI, 2.1-3.2 mos), and mOS was 8.5 months (95% CI, 3.6 mos-NR). Results were not influenced by PD-L1 expression. G3 TRAEs were reported in four patients (9.5%), with three having severe colitis [41]. Therefore, the inhibition of CTLA-4 with single-agent strategies seems less effective than targeting the PD1/PD-L1 pathway in CC patients.

Summary of Systematic Review Results
Our systematic review confirms the benefit of response and survival in patients diagnosed with advanced CC receiving an ICI. Up to 2 out of 3 patients respond to ICIs. However, results are very heterogeneous due to the designs of the trials, administered agents and combinations, and selection criteria; therefore, a large range, from 0% to 65.9%, of ORR is found. In patients not selected for PD-L1, ORR ranges from 0% to 26.3% with single-agent ICIs, reaching 38.4% with dual ICIs association, 65.9% with the combination of ICIs and chemotherapy, 55.6% with ICIs plus TKIs. In PD-L1-positive patients, ORR ranges from 5.9% to 33% with single-agent ICIs, from 27% to 36% with dual ICIs, reaching 68.1% and 69% when ICIs are combined with chemotherapy and TKIs, respectively. In PD-L1negative women, ORR is 0-16.7% to single-agent ICIs, 11-35.8% to double ICIs, and reaches 50% after ICIs + TKIs. Therefore, the response rate is driven mainly by PD-L1-positive patients, but PD-L1-negative ones are less represented in the studies, as expected, given the high rate of PD-L1 positivity found in CC.
As already known in other tumor subtypes, ICIs impact survival in advanced CC patients, ranging from 8 to over 21 months, with most studies that have not reached the mOS yet.
The safety profile is manageable with single agents and in line with other ICIs studies in the combination setting.

PD-L1 and Predictive Biomarkers for ICIs Response
KEYNOTE-158 was the only study to include only PD-L1-positive patients [26]. There is no uniformity regarding the method and cut-off used to detect PD-L1 positivity in the different clinical trials ( Table 1). Most of them measured the combined positive score (CPS), defined as the number of PD-L1 staining cells divided by the total number of viable tumor cells multiplied by 100. In two studies, the tumor-proportional score (TPS)-defined as the percentage of tumor cells expressing PD-L1, was used. Most studies defined 1% as a cut-off to define PD-L1 positivity. The staining platforms used for PD-L1 detection varied between the studies, with 28-8 (Dako), 22C3, and SP263 (Ventana) antibodies mainly used. Moreover, scores could be at higher risk of inaccuracy when a low cut-off is considered, such as in the CC case.
Although the regulatory approvals of ICIs are based on PD-L1 expression, challenges remain, including variable expression, different antibodies, and staining platforms, and the lack of an unequivocal scoring system, that still now raises the question of whether PD-L1 is the unique suitable biomarker to predict response to ICIs in advanced CC patients. It was previously demonstrated that higher PD-L1 expression and CD8 + T cells infiltration predicted a better response to chemo-and radiotherapy, but also ICIs [48][49][50][51][52]. More recently, it has been shown that ICIs administration increased T and B lymphocytes and natural killer (NK) cells infiltration in the tumor microenvironment (TME) in a directly proportional way with ICIs response [53]. A recent analysis found two different clusters of TME could be found in CC patients. In the first cluster, the immune-suppressive TME, a high infiltration of myeloid-derived suppressor cells (MDSCs), macrophages, and Tregs was found. On the other hand, a high infiltration of activated T cells and NK cells was associated with an immune-responsive TME: once again, this subtype seemed to respond better to ICIs [21,53,54].
As PD-L1 does not appear as an entirely consistent and reproducible biomarker, other elements with a prognostic and predictive role should be investigated, such as TMB or MS status, which have also been studied in CC, to build a multi-marker classification eventually.

HPV Role in ICIs Response
Notwithstanding the predominant role of HPV in CC development, the trials we reviewed did not attribute a central role to HPV for ICIs response, except for the phase II NCT03444376 trial, which specifically included HPV-16-or -18-positive patients treated with pembrolizumab plus a vaccine [25]. Results were satisfactory in terms of response rate and survival but not particularly different from studies not focusing on HPV-positive patients. As CC is a paradigmatic example of an HPV-dependent neoplasm, vaccines seem feasible and effective, combined with ICIs, to reciprocally potentiate efficacy and overcome resistance. Oncogenes E6 and E7 represent ideal targets for CC therapeutic vaccines [55][56][57][58][59][60][61][62][63][64][65][66].
It was previously found that, despite HPV-specific T cells infiltrating the primary and metastatic sites, though not E6-or E7-specific, CC cannot be eradicated, thus suggesting the existence of an immunosuppressive tumor TME [14]. Antigen-presenting cells (APCs) themselves present HPV antigens in a tolerogenic way, activating immune-escaping rather than immune-activating pathways [67,68]. Indeed, the PD-1/PD-L1 axis supports tumor mechanisms for escaping immune response by down-regulating tumor-directed T cells. Therefore, PD-L1 might also be considered a marker of an advancing HPV infection [69]. Even if in a small sample size, the results of NCT03444376 were more satisfactory if HPVpositive patients also expressed PD-L1 [25].

Future Perspectives
It is of great importance that ICIs efficacy is not counterbalanced by increasing toxicity, also considering the combination setting. Therefore, the upfront combination of chemo-and immunotherapy appears to be a compelling strategy for these patients. New combinations will provide further options for expanding ICIs efficacy and treatment options in CC (Table 3).  Interesting associations might be those with RT, as radiations induce the production of neo-antigens that boost the immune system and increase CD4 + and CD8 + infiltration [51,70]. Effectively, in the locally advanced setting, ICIs are being explored in addition to definitive CTRT [71]. Used sequentially after chemoradiation, ipilimumab induced a 1-year PFS of 81% and 1-year OS of 90% in the phase I GOG-9929 trial [72,73]. A total of 52 patients with locally advanced CC were randomized to pembrolizumab (3 doses) after or concurrently with the CTRT regimen in a phase II trial. Safety was the primary endpoint of the analyses: 4 G1 AEs and 23 G3 AEs were reported [74]. However, the phase III CALLA study did not meet its primary endpoint of improving PFS with durvalumab added to CTRT versus CTRT alone [75]. Definitive results of pembrolizumab plus CTRT are expected from ENGOT-cx11/KEYNOTE-A18 in a high-risk CC population [76]. Finally, the triple combination of ICIs, chemotherapy, and anti-angiogenics, as well as the combination of anti-angiogenics and TKIs, have a strong rationale for efficacy and are being explored by several studies [77]. Sequencing strategies after progression to ICIs are needed: combo immunotherapy and novel agents, such as antibody-drug conjugates, could be employed.

Limitations of the Systematic Review
Our systematic review has some limitations. First and foremost, the included trials are heterogeneous in terms of treated patients and biomarkers selection. Moreover, studies are ongoing, with survival data yet to be completed. Furthermore, there is a small number of high-quality studies, such as randomized trials. A quantitative meta-analysis could not be performed, as most of the available trials were not designed in a comparative manner. Therefore, the conclusions about the efficacy and safety of ICIs in CC that can be drawn from our analysis are only descriptive. We are confident that a longer follow-up and a larger number of randomized trials would better clarify the real effect of ICIs on the survival of advanced CC patients.

Conclusions
Immunotherapy represents an appealing strategy for patients with advanced CC. A subset of patients had a benefit from ICIs with long-lasting responses even in a chemorefractory setting. Moreover, the combination of chemotherapy and immunotherapy seems to be an effective first-line treatment with acceptable toxicity. Longer follow-ups could confirm these results. Further research is needed for an appropriate patient selection and a tailored approach.