PD-1/PD-L1 immune checkpoint inhibitors in Hodgkin lymphoma: A meta- and network meta-analysis

Highlights

• •Network meta-analysis identifies the efficacy rankings for various PD-1/PD-L1 inhibitor related therapies in Hodgkin lymphoma.
• •Ranking-based results prioritize combination strategies over monotherapy for maximizing clinical response.
• •Camrelizumab demonstrates notable efficacy specifically in achieving high partial response rates.
• •Integrating immune checkpoint inhibitor with conventional therapy significantly enhances both complete and objective response rates.
• •Two immune checkpoint inhibitors yield high performance for attaining objective response.

Abstract

Background

Hodgkin lymphoma (HL) is a relatively rare lymphoid malignancy that significantly affects the health of adolescents, young adults and the elderly. Even though conventional chemotherapy offers high cure rates, the relapsed/refractory HL remains challenging.

Methods

Pooled single-arm and network meta-analyses were used. Literature search was conducted in PubMed, Embase, Web of Science, the Cochrane Library and Clinical Trials, et al., from the date of establishment dates to August 31, 2025. Studies were screened based on inclusion and exclusion criteria. The primary outcome analyzed was complete response rate (CRR), partial response rate (PRR), and objective response rate (ORR).

Results

11 single-arm trials and five controlled trials, involving 1602 participants and seven ICI-based treatment strategies were included. The meta-analysis revealed an overall CRR of 0.35, a PRR of 0.37, and an ORR of 0.75 across all therapy methods. Of these, one ICI drug in combination with conventional therapy demonstrated the highest CRR and ORR, while camrelizumab achieved the highest PRR. The treatment strategies with the highest CRR in network meta-analysis was the combination of two ICI drugs. The strategies with the highest PRR were camrelizumab, a single ICI drug, while the strategies with the highest ORR was a single ICI drug combined with conventional therapy.

Conclusion

This study suggests that camrelizumab is associated with an improved PRR for HL, whereas, combination strategy of two ICI drugs, and that one of pembrolizumab and camrelizumab combined with conventional chemoradiotherapy are more effective for elevating the CRR and ORR respectively.

Graphical abstract

Introduction

Hodgkin lymphoma (HL) is a relatively rare malignancy originating from the malignant transformation of lymphocytes. Its global incidence is approximately 1.0 per 100,000 individuals, accounting for about 0.4% of all new cancer diagnoses each year [1]. In China, the new HL cases nationwide is about 3600 each year, with an incidence of approximately 0.25 per 100,000. For comparison, the incidence of HL in the United States is 2.5 per 100,000, which is 10 times that of China [2]. The five-year mortality rate of HL stands at approximately 10% globally [3]. Although HL is relatively uncommon, it exhibits a bimodal age distribution, predominantly affecting adolescents and young adults aged 15–35 years, followed by adults over 65 years of age. Notably, elderly patients face a higher risk of mortality, [4] underscoring the significant impact of HL on both younger and older populations.

Currently, front-line treatments for HL have been well-established, including the chemotherapy (CT) regimens like ABVD (doxorubicin [Adriamycin®], bleomycin, vinblastine [Oncovin], and dacarbazine), BEACOPP (bleomycin, etoposide, doxorubicin [Adriamycin®], cyclophosphamide, vincristine [Oncovin], procarbazine, and prednisone), often in combination with radiotherapy (RT), and these regimens have showed a clinical success for achieving cure rates exceeding 90% in early-stage disease and about 80% in advanced stages [4]. Nevertheless, many patients experience therapy-related side effects, resulting in acute adverse reactions such as an increased risk of infection and long-term health problems such as secondary malignancies, sterility, RT-associated hypothyroidism, cardiovascular diseases and bleomycin-associated pulmonary toxicity [[5], [6], [7]]. Moreover, a subset of patients exhibits intrinsic resistance to initial chemotherapy or experiences relapse after treatment. Subsequent therapeutic options, including high-dose chemotherapy (HDCT) and autologous stem cell transplantation (ASCT), often yield limited efficacy [8] and are associated with serious treatment-related toxicities [9,10] even mortality [11,12]. These treatments may also elevate the risks of secondary malignancies, cardiovascular diseases, and pulmonary complications, [13] collectively contributing to impaired quality of life and reduced overall survival.

PD-1/PD-L1 immune checkpoint inhibitors (ICI) represent a novel class of anticancer agents that function by competitively binding to PD-1/PD-L1, preventing tumor cells from escaping immune surveillance, and restoring T-cell mediated tumor suppression [14]. Owing to their targeted mechanism and potentially more favorable toxicity profile compared with conventional chemotherapy, ICI have been incorporated into clinical practice for a range of malignancies, including advanced melanoma, non-small cell lung cancer, and renal cell carcinoma [15]. Since the initial FDA approval of nivolumab for HL in 2016, [16] several PD-1/PD-L1 inhibitors have been authorized for the treatment of relapsed or refractory HL, [17] demonstrating particularly promising outcomes in this setting. In parallel, multiple therapeutic strategies incorporating ICI have been developed, spanning monotherapy, combined chemotherapy, and regimens involving other targeted agents [18]. A growing body of clinical trials and observational studies has sought to evaluate the efficacy of these ICI-based regimens in HL.

However, due to the relatively low incidence of HL, most clinical investigations on PD-1/PD-L1 inhibitors in this disease are constrained by small sample sizes. Furthermore, related studies indicate that the efficacy of PD-1/PD-L1 ICI exhibits considerable heterogeneity among individuals and their overall effectiveness at the population level has not been fully established.(T. S [19]; T. S. Y [[20], [21], [22], [23], [24]]) The available evidence consists largely of retrospective and single-arm studies, with a notable scarcity of prospective randomized controlled trials (RCT) that directly compare different immunotherapy-based strategies. Consequently, there is a lack of robust cross-comparative evidence regarding the relative efficacy among various ICI-containing regimens, leaving clinicians without high-level evidence to guide optimal treatment selection.

In this study, we conducted a comprehensive electronic literature search across multiple databases to identify all prospective clinical trials investigating immunotherapy in HL. By synthesizing the available evidence, we performed a systematic literature review and meta-analysis to thoroughly evaluate the efficacy of different immunotherapeutic approaches in HL. Furthermore, a network meta-analysis (NMA) was employed to compare the relative benefits of various treatment strategies. This study aims to provide clinicians with direct and quantitative evidence to support the selection of optimal immunotherapy regimens for patients with HL.

Materials and methods

Search strategy and selection criteria

For English-language literature, the PubMed, Embase, Web of Science (WOS) and the Cochrane Library were retrieved, and for Chinese one, the China National Knowledge Infrastructure (CNKI), Wanfang Data and VIP databases were searched, all from the establishment date through August 31, 2025. A combination of search terms was used, including “lymphoma”, “immune checkpoint inhibitors”, “pembrolizumab”, “nivolumab”, “camrelizumab”, “sintilimab”, “toripalimab”, “tislelizumab”, “atezolizumab”, “durvalumab”, “avelumab”, “ipilimumab” and “randomized controlled trial”.

Studies were deemed eligible for inclusion as: published or unpublished phase I, II, or III clinical trials; those included patients with histologically or cytologically confirmed HL who had received at least one ICI treatment and at least one clinical assessing of ICI-based treatment strategy, the assessing outcome included complete response rate (CRR), partial response rate (PRR) and objective response rate (ORR). These outcome measures were defined according to the Response Evaluation Criteria in Solid Tumours (RECIST) guideline version 1.1. Complete response referred to the disappearance of all target lesions, Partial response referred to a decrease of at least a 30% in the sum of diameters of target lesions, taking as reference the baseline sum of the diameters, and objective response represented the sum of complete response and partial response [25]. The exclusion criteria encompassed trials wherein tumor size constitutes the primary measure, or trials involving minors.

The following data were extracted: study characteristics (first author name, publication year, the study phase, registered number, and the sample size); demographic information (gender and age); treatment methods, and the assessing outcome (CRR, PRR and ORR). The outcome assessed by a blind independent review committee based on the intention-to-treat principle was prioritized, as well as those reported at different follow-up time points within a single trial. The literature screening and data extraction were conducted independently, and any discrepancies were resolved through discussion with senior investigators.

Risk of bias evaluation

The risk of bias was evaluated using the following evaluation tools: Cochrane Risk of Bias tool 2.0 for RCT(J. A. C [26]) and ROBINS-I for single-arm studies.(J. A [27]) Each literature was independently evaluated by two researchers, and consensus was reached through discussion with the senior investigator in case there were disagreements.

Data analysis

The R software (version 4.3.2) was used for data analysis, with the “meta” package utilized for single-arm meta-analysis and the “gemtc” package employed for NMA. The primary outcomes were the CRR, PRR, and ORR of the patients after treatment.

Meta-analyses were performed to calculate pooled rates of the primary outcomes. The odds ratios (OR) and their 95% confidence intervals (CI) were calculated. Heterogeneity was assessed by the I² statistics and the Cochran's Q-test (tau-squared and p values). Methodological framework for model selection based on data heterogeneity was systematically applied to both the single-arm and network meta-analyses. Specifically, when statistical heterogeneity was large (I²>50%), a random-effects model (REM) was used, otherwise, a fixed-effects model (FEM) was applied [28].

NMA employed Markov chain Monte Carlo simulation to compare any two treatment strategies by simultaneously synthesizing direct and indirect evidence. The framework of frequency counting method was utilized due to its capacity for computational efficiency and its independence from subjective priori assumptions. The 95% CI for each outcome were reported as summary statistics of the estimates. Treatment rankings were determined based on the probability of superiority, summarized using Surfaces Under the Cumulative RAnking (SUCRA) values, ranging from 0 (the least efficacious treatment) to 1 (the most efficacious treatment).

Results

Literature search, selection and characteristics of the included studies

The literature screening process (Fig. 1): A total of 835 relevant studies were retrieved with 244 duplicate records; 591 articles were initially screened based on their summaries. And, the second screening of the 101 articles, involving a thorough reading of the full texts, 85 papers were further eliminated, leaving a total of 16 articles to be analyzed. The meta-analysis included 16 studies (1602 participants) for the CRR, 14 studies (1501 participants) for the PRR, and 14 studies (1501 participants) for the ORR, covering seven ICI-based treatment strategies (conventional therapy, one ICI drug with conventional therapy, pembrolizumab, camrelizumab, nivolumab, avelumab, and two ICI drugs). Of the included studies, 14 were phase I/II clinical trials, while one was a phase III trial. The number of participants ranged from 10 to 304, with a median age of 26–38 years. Approximately 46% of participants were female (Table 1).

Fig. 1.

Flowchart of literature search and selection. A total of 835 relevant studies were retrieved from 7 literature databases, with 244 duplicated records eliminated firstly. The remaining 591 studies were screened based on their Abstract and 490 studies were removed. The remaining 101 studies were further screened with a thorough reading of the full texts and 85 studies were eliminated. A total of 16 articles were finally included for the follow-up analysis, with all of them for conventional meta-analysis and 5 of them for NMA.

Table 1.

Baseline characteristics and interventions of the included studies.

Study (First Name, Year)	Registered Number	Phase	NO.	Female, %	Median Age, year	Treatment Strategy	Treatment Line	Complete response	Partial response	Objective response
[31,31]	NCT02684292	II	Arm1: 151 Arm2: 153	Arm1: 44% Arm2: 41%	Arm1: 36 Arm2: 35	Arm1: Pembrolizumab Arm2: Brentuximab vedotin	≥2	Arm1: 37 Arm2: 37	Arm1: 62 Arm2: 46	Arm1: 99 Arm2: 83
[19,53]	NCT02603419	I	31	NR	38	Avelumab	NR	2	15	17
[54,54]	NCT02362997	II	30	47%	33	Pembrolizumab	≤3	25	NR	NR
[55,55]	NCT02453594	II	210	46%	35	Pembrolizumab	≥1	58	92	150
[29](Y [29])	NCT02961101, NCT03250962	II	Arm1: 42 Arm2: 19	Arm1: 37% Arm2: 40%	Arm1: 28 Arm2: 26	Arm1: Camrelizumab Arm2: Decitabine + camrelizumab	≥2	Arm1: 33 Arm2: 6	Arm1: 7 Arm2: 11	Arm1: 40 Arm2: 17
[56,56]	NCT01896999	I/II	21	48%	33	Brentuximab vedotin + Ipilimumab	≥2	12	4	16
[57,57]	NCT02181738	II	100	NR	NR	Nivolumab	≥2	17	56	73
Q. Wang, 2023 [58]	NR	NR	Arm1: 34 Arm2: 50	Arm1: 29% Arm2: 40%	Arm1: 35 Arm2: 35	Arm1: Camrelizumab Arm2: Epirubicin-AVD	2	Arm1: 3 Arm2: 2	Arm1: 25 Arm2: 21	Arm1: 28 Arm2: 23
[59](S [59])	NCT01592370	I	31	NR	NR	Nivolumab + Ipilimumab	≥2	6	17	23
R. Dada, 2018 [60]	NR	II	10	50%	26	Nivolumab	≥3	7	1	8
[61](S. M [61])	NCT02181713	II	243	42%	34	Nivolumab	≥2	52	121	173
[32,32]	NCT01896999	II	Arm1: 61 Arm2: 57	49%	34	Arm1: Brentuximab vedotin + nivolumab Arm2: Brentuximab vedotin + nivolumab + ipilimumab	≥1	Arm1: 37 Arm2: 38	Arm1: 17 Arm2: 12	Arm1: 54 Arm2: 50
[62,62]	NCT01953692	1b	31	42%	32	pembrolizumab	≥2	6	12	18
[63,63]	NCT02572167	I/II	91	56%	34	Brentuximab vedotin + Nivolumab	≥1	61	16	77
[22,64]	NCT03343665	I/II	Arm1: 50 Arm2: 116	Arm1: 66% Arm2: 52%	Arm1: 36 Arm2: 38	Arm1: Nivolumab 40 mg Arm2: Nivolumab 3mg/kg	≥1	Arm1: 19 Arm2: 40	Arm1: 14 Arm2: 38	Arm1: 33 Arm2: 78
P [30](P [30])	NCT04044222	III	Arm1: 34 Arm2: 37	NR	NR	Arm1: sintilimab + ifosfamide, carboplatin, and etoposide (ICE) Arm2: placebo + ifosfamide, carboplatin, and etoposide (ICE)	1	Arm1: 21 Arm2: 12	NR	NR

*NR: not reported. 

Risk of bias evaluation

In the 11 single-arm studies, four were assessed as having some, while seven were considered as having high risk of bias, primarily due to issues in the classification of interventions. Among the 5 controlled trials, two were judged as low risk, and three as having some concerns of the risk of bias, with the main source of bias arising from the measurement of outcomes (Supplementary Figs. S1-S4). Publication bias was assessed using funnel plots for CRR, PRR, and ORR. The visual inspections of these plots are available in the Supplementary Figs. S5–S7.

Meta-analysis of multiple therapies in HL patients

Fig. 2 showed that the pooled CRR for all therapy strategies was 0.35 (95% CI, 0.24–0.48; REM). And for the specific treatments, a combination of one ICI drug and conventional therapy achieved the highest CRR at 0.66 (95% CI, 0.60–0.71; FEM), while a combination of two ICI drugs showed a moderate effect at 0.42 (95% CI, 0.14–0.76; REM) (Fig. 2).

Fig. 2.

Pooled CRR for therapy strategies. Each panel illustrates the pooled CRR for a specific treatment strategy. Summary estimates are provided using both fixed-effects and random-effects models. Horizontal lines represent the 95% CI. Statistical results for heterogeneity testing (including I² statistic and p-value) are presented for each analysis.

For PRR, the overall estimate was 0.37 (95% CI, 0.30–0.45) based on the REM for all therapy strategies. Camrelizumab demonstrated the most favorable PRR (0.68, 95% CI 0.54–0.79; FEM), followed by avelumab (0.48, 95%CI 0.30–0.67; single study). Pembrolizumab (0.42, 95% CI 0.38–0.47; FEM) and nivolumab (0.41, 95%CI 0.27–0.56; REM) also achieved relatively high PRR (Fig. 3).

Fig. 3.

Pooled PRR for therapy strategies. Each panel illustrates the pooled PRR for a specific treatment strategy. Summary estimates are provided using both fixed-effects and random-effects models. Horizontal lines represent the 95% CI. Statistical results for heterogeneity testing (including I² statistic and p-value) are presented for each analysis.

Regarding ORR, the pooled estimate for all therapy strategies was 0.75 (95% CI, 0.67–0.81; REM). Combining one ICI drug with conventional therapy yielded the highest ORR (0.87, 95% CI, 0.82–0.91; FEM). Furthermore, camrelizumab (0.85, 95% CI 0.73–0.92; FEM) and two ICI drugs (0.83, 95% CI 0.72–0.90; REM) exhibited significant treatment efficacy (Fig. 4).

Fig. 4.

Pooled ORR for therapy strategies. Each panel illustrates the pooled ORR for a specific treatment strategy. Summary estimates are provided using both fixed-effects and random-effects models. Horizontal lines represent the 95% CI. Statistical results for heterogeneity testing (including I² statistic and p-value) are presented for each analysis.

NMA and SUCRA rankings for HL patients

From the overall studies, five controlled trials (n = 638) evaluating five different treatment strategies (conventional therapy, one ICI drug with conventional therapy, pembrolizumab, camrelizumab, and two ICI drugs) were selected for the NMA. Of the 638 patients analyzed, treatment distribution was as follows: 240 received conventional therapy alone, 137 received one immunotherapy agent combined with conventional therapy, 151 received pembrolizumab, 53 received camrelizumab, and 57 received dual immunotherapy of pembrolizumab and camrelizumab.

Transitivity assumption analysis: For the three outcomes examined, only the CRR formed a fully connected network, enabling simultaneous direct and indirect comparisons. Within the CRR network, the transitivity assumption-essential for valid indirect comparisons-was evaluated by testing the consistency between direct and indirect evidence. The heterogeneity assessed within designs was moderate (I² = 47.5%), and the Cochran's Q-test for inconsistency yielded a non-significant p-value of 0.17, indicating no statistically detectable inconsistency in the network. This result supported the underlying transitivity assumption and suggested acceptable coherence between direct and indirect evidence for the CRR outcome. Based on this data consistency and relatively low heterogeneity, an FEM was applied to estimate the pooled comparative estimates for the CRR network. Conversely, for the PRR and ORR networks, because the available evidence failed to construct closed loops, local network heterogeneity and inconsistency parameters could not be mathematically derived. Therefore, REM was applied for the PRR and ORR to account for any potential variations across the included clinical trials.

CRR between treatment strategies (Fig. 5a): The combination of two ICI drugs yielded an OR of 5.76 (95%CI: 1.79–18.53, FEM), while one ICI drug combined with conventional therapy resulted in an OR of 4.44 (95% CI: 1.79–18.53, FEM). By contrast, there was no significant difference in CRR between pembrolizumab and camrelizumab and traditional therapy, with OR of 1.02 (95% CI: 0.60–1.72, FEM) and 0.86 (95% CI: 0.26–2.83, FEM) respectively. The combination of two ICI drugs, as well as one ICI drug combined with conventional therapy, demonstrated superior outcomes compared to traditional treatment alone.

Fig. 5.

Comparisons of each outcome in the NMA. Network plots illustrate direct and indirect comparisons for each outcome. Circular nodes represent therapy strategies. Lines represent the direct comparison, with thick proportional to the OR between two therapies. In tables, each odd ratio represents the column strategy relative to the row strategy. A: Conventional therapy; B: One ICI drug with conventional therapy; C: One ICI drug (Pembrolizumab); D: One ICI drug (Camrelizumab); E: Two ICI drugs.

PRR between treatment strategies (Fig. 5b): pembrolizumab showed an OR of 1.62 (95% CI: 1.01–2.60, REM) and camrelizumab showed an OR of 3.84 (95% CI: 1.49–9.88, REM). In contrast, neither the dual immunotherapy combination (OR = 0.39, 95% CI: 0.07–2.24, REM) nor the single ICI drug plus conventional therapy (OR = 0.56, 95% CI: 0.12–2.61, REM) showed a statistically significant improvement in partial response compared to traditional treatment. Thus, both pembrolizumab and camrelizumab performed significantly better than conventional treatment.

ORR between treatment strategies (Fig. 5c): Significant improvements in the ORR were observed with pembrolizumab (OR = 1.61, 95% CI: 1.01–2.55, REM), camrelizumab (OR = 5.48, 95% CI: 1.93–15.54, REM) and the strategies combining one ICI drug with conventional therapy (OR = 12.89, 95% CI: 1.30–127.46, REM); all of this outperformed traditional treatment.

Probability of superiority: In Table S1, consistency was observed between the Bayesian ranking profiles-illustrated by the SUCRA values for each treatment strategy across outcomes-and the results based on OR estimates. The two ICI combinations demonstrated the highest ranking for complete response (SUCRA=0.88) and the second highest for objective response (SUCRA=0.81), yet the lowest for partial response (SUCRA=0.12). The combination of one ICI with conventional therapy achieved the highest rank for objective response (SUCRA=0.82) and the second highest for complete response (SUCRA=0.82). For partial response, camrelizumab and pembrolizumab were ranked first and second, with SUCRA values of 0.98 and 0.68, respectively. Taken together, the combination of a single ICI with conventional therapy emerges with the most consistently high ranking across the key efficacy outcomes.

In summary, these findings of NMA collectively indicate that the efficacy profiles of the various ICI-based strategies differ markedly across outcomes. Two ICI drugs and one ICI drug combined with conventional therapy were most effective for achieving complete response. In contrast, for partial response and objective response, camrelizumab and pembrolizumab monotherapies, as well as one ICI drug combined with conventional therapy, showed significant benefits over conventional treatment.

Sensitivity analysis

To evaluate the statistical robustness of our models and to prohibit potential outlier effects, we conducted comprehensive, multi-layered leave-one-out sensitivity analyses for all three clinical endpoints, including the network and the single-arm meta-analysis. For the NMA framework of the CRR endpoint, when sequentially omitting individual studies among the primary datasets, the mathematical model successfully achieved statistical convergence across three occasions (exclude [29], Q. Wang, 2023, and P [30]); however, excluding trials from the remaining two trials led to network disconnection or non-convergence (exclude [31] and [32]). Among the three converged models, the overall treatment hierarchy demonstrated high stability, with two ICI drugs consistently maintaining the status of optimal treatment (SUCRA range: 0.859–0.935), followed by an ICI drug combined with conventional therapy (SUCRA range: 0.715–0.809), whereas conventional therapy alone performed poorly across all scenarios (SUCRA range: 0.072–0.169). These results indicate that our network conclusions of CRR endpoint are highly robust and are not influenced by bias in individual trials. Meanwhile, for the PRR and ORR network models, sequentially excluding controlled trials led to the network disconnection or non-convergence, making it impossible to obtain relevant sensitivity analysis results (Figs. S7-S10).

Conversely, for the single-arm descriptive meta-analysis, across all three clinical endpoints, the overall combined baseline rates remained exceptionally stable upon the sequential omission of any single trial, confirming the global robustness of the baseline synthesis. To define these variations objectively, a relative change of 15% or greater in the subgroup point estimates was utilized as the quantitative threshold for high-influence points. Specifically, our tracking identified that six independent studies contributed to visible adjustments within the CRR subgroups upon omission, and two studies introduced a notable shift within the PRR arm, whereas the ORR subgroup estimates demonstrated a profound absolute immunity, with no individual historical study causing any meaningful fluctuation to its treatment modalities upon exclusion. Details of all layers are shown in the Supplementary (Table S2-S4).

Discussion

The meta-analysis revealed an overall CRR of 0.35 (95% CI, 0.24–0.48), a PRR of 0.37 (95% CI, 0.30–0.45), and an ORR of 0.75 (95% CI, 0.67–0.81) across all therapeutic strategies. Among these, one ICI combined with conventional therapy demonstrated the highest CRR (0.66, 95% CI, 0.60–0.71) and ORR (0.87, 95% CI, 0.82–0.91), whereas camrelizumab achieved the highest PRR (0.68, 95% CI, 0.54–0.79).

In the NMA, the treatment strategy associated with the highest CRR was the combination of two ICI. The highest PRR was observed with camrelizumab monotherapy, while the highest ORR was associated with a single ICI combined with conventional treatment.

This study found that camrelizumab, as well as pembrolizumab, achieved a high PRR but a relatively low CRR when used to treat certain subtypes of HL. This phenomenon is not unique to camrelizumab or to this specific malignancy; rather, it represents a common challenge encountered with PD-1 inhibitors across various B-cell and T-cell lymphomas. For example, in the KEYNOTE-170 study, pembrolizumab achieved a PRR of up to 32% in patients with relapsed or refractory primary mediastinal large B-cell lymphoma (PMBCL), but the CRR was only 13% [33]. Similarly, the CheckMate 139 study evaluating nivolumab in diffuse large B-cell lymphoma (DLBCL) demonstrated limited efficacy, with a CRR of only 3%.(S. M [34]) This “high PRR/low CRR” pattern likely reflects a fundamental mismatch between the mechanism of action of PD-1 inhibitors and the complex tumor microenvironment (TME) of HL. PD-1 inhibitors can effectively reverse T-cell exhaustion, inducing tumor regression in regions with sufficient immune cell infiltration, thereby resulting in relatively high PRRs. However, their efficacy is limited by several factors, including the prevalent “immune desert” phenotype observed in HL, [35] the compensatory upregulation of multiple immunosuppressive pathways (such as TIM-3 and LAG-3), and tumor antigen heterogeneity [36]. Collectively, these factors hinder the immune system’s ability to achieve complete tumor eradication, making sustained increases in CRR difficult.

Further analysis revealed that, for patients who respond poorly to monotherapy (i.e., those who fail to achieve complete or partial remission), combined treatment strategies offer clear advantages. Our NMA results reveal a highly interesting pattern. Specifically, Camrelizumab monotherapy achieved a highest ranking for PRR (SUCRA = 0.98) but plummeted to the lowest tier for CRR (SUCRA = 0.20). This statistical polarization is consistent with the current consensus in tumor immunotherapy. When immune checkpoint inhibition initiates an antitumor immune response but remains insufficient, combined strategies can enhance efficacy through complementary mechanisms. For instance, combining two ICIs—such as a CTLA-4 inhibitor with a PD-1 inhibitor—targets different phases of T-cell activation, resulting in stronger immune priming and effector function. This approach has demonstrated substantial efficacy in solid tumors such as melanoma [37]. In the context of lymphoma, however, combining PD-1 inhibitors with conventional chemoradiotherapy appears to be more clinically applicable. Chemotherapy and radiotherapy can directly induce tumor cell death and promote the release of tumor antigens through immunogenic cell death, thereby reshaping the tumor microenvironment. This process enhances T-cell infiltration and facilitates the conversion of “cold” tumors into “hot” tumors, [38] creating favorable conditions for subsequent PD-1 inhibitor activity. In addition, combining PD-1 inhibitors with targeted agents, such as lenalidomide, shows considerable promise in lymphoma treatment. Lenalidomide not only exerts direct antitumor effects but also enhances T-cell and natural killer cell (NK) function, resulting in synergistic antitumor activity when combined with immune checkpoint blockade [39]. Therefore, future therapeutic strategies for HL should prioritize the exploration and optimization of rational combined regimens. Multi-targeted and multi-modal synergistic approaches may help overcome the limitations of monotherapy and ultimately improve patient outcomes.

The NMA applied in this study represents an advanced approach to evidence synthesis. Its principal advantage lies in its ability to integrate both direct and indirect comparative evidence across multiple interventions for the same disease within a unified statistical framework [40]. In this study, NMA enabled us to address the current lack of randomized controlled trials directly comparing PD-1/PD-L1 inhibitors in HL. By constructing an evidence network that linked various treatment strategies, including immunotherapy monotherapies and combined regimens, we were able to perform quantitative comparisons across trials. This approach maximizes the use of available clinical data and improves the precision of treatment effect estimates. Importantly, NMA also allows for the estimation of ranking probabilities for all interventions, identifying the likelihood of each treatment being the most effective option, thereby providing intuitive and quantitative support for clinical decision-making that traditional pairwise meta-analyses cannot offer [41]. Although closed loops could not be formed for some secondary outcomes, limiting inconsistency assessment, a well-connected evidence network was successfully established for the primary outcome of complete remission rate. Through this network, we were able to integrate both direct and indirect evidence, allowing for more comprehensive and reliable comparative assessments of therapies such as camrelizumab and pembrolizumab than would be possible using direct comparisons alone. The design, conduct, and reporting of this study adhered strictly to the PRISMA-NMA guidelines to ensure methodological rigor and transparency [42].

Compared to the two existing meta-analyses [43,44] related to PD-1/PD-L1 inhibitors in HL that we could obtain so far, this study added new insights and distinguished itself in the following aspects. First, the two previous meta-analyses both focused on relapsed / refractory HL and one of them [43] further focused on low-dose PD-1 monotherapy, which differed from this study regarding all types of HL and all dosage of PD-1/ PD-L1 mono- or combination therapy. Second, this study retrieved a wider range of literature databases (including three Chinese literature databases considering the large number of HL patients in China), leading to a larger cohort size compared to the two meta-analyses mentioned above (1602 vs 1440 and 84). Third, from the methodological perspective, this study included NMA in addition to conventional meta-analysis, which simultaneously integrated all direct and indirect evidence to provide a comprehensive performance analysis of PD-1/PD-L1 in HL. Lastly, from the perspective of analysis results, this study showed consistency with the two previous meta-analyses in some aspects also provided some novelties. For example, Sun et al. [44] found that pembrolizumab combination therapy reached significantly higher ORR than pembrolizumab monotherapy, which was confirmed in this study. We also added that, not only for pembrolizumab but for several ICIs such as camrelizumab, ipilimumab and nivolumab, an elevating ORR had been observed for one ICI combined with no matter another ICI or conventional chemoradiotherapy compared with the ICI monotherapy. Further, the results of NMA rankings also supported that ICI combination strategies outperformed ICI monotherapy and conventional treatment strategies. Moreover, the efficacy of combination strategies including ICI with other types of cancer treatment is also worth to explore. For example, a large multi-center study [45] has confirmed that combining ICI with tyrosine kinase inhibitors (targeted anti-cancer drugs) showed higher overall and progression-free survival compared to dual ICI combination therapy (nivolumab + ipilimumab) in patients with intermediate-risk renal cell carcinoma. Although this study lacked research related to the safety of ICIs which had been discussed in the two meta-analyses mentioned above, as we set the main focus on the efficacy and merely included prospective clinical trials in this study which were not enough for safety analysis (for example, retrospective or observational studies should be included too), the issues of safety for ICI treatment are not to be ignored. There are existing studies indicating that adverse events were associated with ICI treatment in cancer patients. For example, Jia et al. [46] systematically elucidated the characteristics of adverse events for lung cancer patients after receiving ICI treatment. Vitale et al. also explored multiple types of adverse events such as infusion-related reactions [47], malnutrition [48,49], hyponatremia [50], digestive and endocrine disorders [51] that related to ICI treatment among cancer patients in their series of studies. Future clinical studies on the safety of ICI treatment in patients with HL will remain a key focus.

According to the latest clinical guidelines for HL treatment from the National Comprehensive Cancer Network (NCCN) [52], current combination therapies with PD-1/PD-L1 ICI (e.g., nivolumab‑doxorubicin [Adriamycin®], vinblastine [Oncovin], and dacarbazine [AVD]) have been incorporated as a preferred strategy for advanced classic HL (Category 1, based on high-level evidence with NCCN panel consensus ≥85%). On the other hand, nivolumab monotherapy is recommended in the guidelines for classic HL in specific situations (e.g., relapsed/refractory HL after failure of autologous stem cell transplantation [ASCT] or brentuximab vedotin [BV] therapy), which is a Category 2A recommendation (based on lower-level evidence with NCCN panel consensus ≥85%). The results of this study showed consistency with the recommendations above that PD-1/PD-L1 monotherapy strategies (e.g., pembrolizumab, camrelizumab) have an advantage over conventional treatments in improving PRR (N = 1025) and combining PD-1/PD-L1 inhibitors with conventional chemoradiotherapy improved CRR and ORR in HL patients (N = 249). Additionally, this study revealed a therapeutic approach not yet mentioned in the guidelines: the combination of two PD-1/PD-L1 inhibitors may hold promising prospects for HL treatment (N = 88). From a clinical perspective, we believe this study offer three valuable perspectives that extend beyond current recommendations of curing HL in guidelines: 1) ICI monotherapy strategies retain value in guideline-specified relapsed or refractory settings, where PRR is a clinically meaningful endpoint; 2) Clinical decision-making may need to shift from monotherapy to the combination of multiple drugs or treatment modalities when CRR/ORR dominates PRR for curable HL; 3) Although not involved in this study yet, we believe that it deserves dedicated prospective study that sequential treatment, regardless of conventional chemoradiotherapy, ICI monotherapy or combination strategies, is needed to optimize the benefit-risk ratio, especially when the trade-off between remission depth and treatment burden is central.

Despite the clinical insights provided by this NMA, several limitations should be acknowledged. First, the relatively small number of RCT included reflects the scarcity of high-quality clinical evidence regarding PD-1 inhibitors in HL, which may reduce statistical power and lead to less precise estimates for certain comparisons. Second, from a methodological perspective, evidence networks for outcomes beyond the primary endpoint of CRR—such as PRR and ORR—failed to form closed loops. This limitation was primarily attributable to heterogeneity and inconsistency in outcome reporting across studies, resulting in insufficient shared comparators for indirect comparisons. Consequently, the full potential of NMA could not be realized for these secondary endpoints. Third, the findings of camrelizumab in this study were based on two trials conducted on Chinese populations, which limited the generalizability of these findings. Future clinical trials based on larger and more diverse cohorts (potentially through global collaborations) were warranted to further confirm the performance of camrelizumab. Lastly, limited by the included types of clinical trials, this study lacked analysis results regarding the safety of ICI treatment for HL patients, which will be a key focus of our future search by incorporating a wider variety of original studies.

Conclusions

In conclusion, this study suggests that in HL, PD-1/PD-L1 ICI monotherapy is associated with an improved PRR, whereas combined strategies-either with conventional chemoradiotherapy or with other ICI-are more effective for elevating the CRR and ORR.

Consent

Not applicable.

Ethical approval

Not applicable.

CRediT authorship contribution statement

Tingxi Zhu: Writing – original draft, Formal analysis. Zhongyu Liu: Writing – original draft, Data curation. Huan Sun: Writing – review & editing, Methodology. Tianchi Lin: Data curation. Zhe Lu: Data curation. Xiaoyan Yang: Writing – review & editing, Supervision, Methodology, Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability

All the data of the current study were available at reasonable request to the corresponding authors.