Efficacy and Safety Profile of PD-1 Inhibitors Versus Chemotherapy in the Second-Line Treatment of Advanced Esophageal Squamous Cell Carcinoma: A Systematic Review and Meta-Analysis of Randomized Controlled Trials

Programmed death 1 (PD-1) inhibitors have emerged as the new standard of care for the second-line treatment of advanced esophageal squamous cell carcinoma. There have been lots of research lately concerning the topic. A comprehensive assessment of the efficacy and safety profile between PD-1 inhibitors and chemotherapy is warranted. Hence, we carried out a systematic review and meta-analysis to illustrate this issue. Pubmed, Embase, Cochrane Library, and Embase were searched systematically until May 1, 2022. We extracted data on efficacy and safety and calculated the pooled hazard ratios (HRs) and relative ratios (RRs) with 95% CI using randomized-effect or fixed-effect models. A subgroup analysis was applied to explore the factors modifying the response to PD-1 inhibitors. Ultimately, a total of 5 studies involving 1970 patients were included in our meta-analysis. PD-1 inhibitors group could attain greater overall survival (OS) benefit (HR = 0.73, 95% CI: 0.66–0.81, P < 0.001) and nearly favorable progression-free survival (HR = 0.89, 0.76–1.04, P = 0.13). Treatment-related adverse events (RR = 0.76, 95% CI: 0.64–0.91, P = 0.004) and level 3–5 treatment-related adverse events (RR = 0.40, 95% CI: 0.32–0.49, P < 0.001) were significantly diminished in PD-1 inhibitors groups. Among all modifying factors, programmed death ligand 1 combined positive score was positively associated with the patient’s OS. The analysis suggests that PD-1 inhibitors exhibited better survival outcomes and safety profiles than standard-of-care chemotherapy. High levels of programmed death ligand 1 combined positive scores were associated with an enhanced response to PD-1 immunotherapies concerning OS.

(J Immunother 2023;46:262-270) E sophageal cancer is one of the most prevalent digestive system cancers with high rates of morbidity and mortality. In 2020, esophageal cancer accounted for around 0.6 million new cases and 0.5 million deaths, making it the eighth most often diagnosed and sixth deadliest cancer worldwide. 1 Based on histologic type, esophageal cancer can be further subdivided into 2 subtypes: esophageal squamous cell carcinoma (ESCC) and esophageal adenocarcinoma, with the former accounting for more than 85% of all new cases. [2][3][4] However, esophageal cancer has a distinctive spatial distribution around the globe, notably in cases of ESCC, with variances between nations that exceed 10-fold. 4 Approximately half of the patients with esophageal cancer are from East Asia, and 90% of them have ESCC. 5 In Western nations, esophageal adenocarcinoma accounts for roughly two-thirds of esophageal cancer cases. 6 Although the incidence of ESCC has decreased in recent decades as a result of higher socioeconomic standards and lifestyle changes, it remains an urgent public health concern in China and other developing nations. 5,7 At the time of diagnosis, 60% -70% of patients with ESCC have already reached the locally advanced or metastatic stage due to the disease's insidious signs and rapid progression. 5 They are confronted with a terrible prognosis, with an observed 5-year survival rate of <5% and 22% in regional and distant stages, respectively. 8,9 In systemic therapy, chemotherapy in combination with fluorouracil and platinum-based agents continues to be the first-line drug treatment choice for patients with advanced ESCC, with an efficacy of 40%-60%. 10 However, effective treatment options for metastatic recurrent ESCC are limited after the failure of first-line therapy or chemotherapeutic agent resistance. The previous conventional second-line therapy, that is, single-agent chemotherapy with docetaxel or paclitaxel, was unsatisfactory in terms of therapeutic effect and was associated with substantial neurotoxicity, gastrointestinal damage, and hepatotoxicity. The development of programmed death 1 (PD-1) inhibitors has ushered in a new era in the treatment of patients with advanced ESCC who were refractory to conventional chemotherapies. 11,12 More recently, a number of randomized control trials (RCTs) concerning the second-line therapy of patients with ESCC with immune checkpoint inhibitors have been released. [13][14][15][16][17] Pembrolizumab, Nivolumab, Camrelizumab, and other emerging PD-1 inhibitors have considerableclinical efficacy and safety advantages over traditional chemotherapy regimens, according to published results. In 2021, a meta-analysis examining the therapeutic efficacy and safety of PD-1 inhibitors revealed that second-line PD-1inhibitor drugs significantly increased the overall survival (OS) of patients with advanced ESCC. 18 However, not all of the included research was published in Extenso, and some valuable information was unavailable. Now that these clinical trials have been published as a whole, we have the opportunity to reevaluate the clinical efficacy and safety with greater precision. In addition, it is vital to investigate the demographic that benefits more from PD-1 inhibitors. To address these issues, we conducted a systematic review and meta-analysis of RCTs concerning second-line therapies (PD-1 inhibitors vs standard chemotherapy) for patients with advanced ESCC.

PATIENTS AND METHODS
The study's report followed the "Preferred Reporting Items for Systematic Reviews and Meta-Analyses" guidelines. 19 We prospectively registered the protocol in PROSPERO (www.crd.york.ac.uk/prospero/), the international prospective register of systematic reviews, with a registration number (CRD42022326180). As a meta -analysis, the study does not require the approval of an ethics committee or Institutional Review Board.

Search Strategy and Study Selection
To identify relevant articles, we performed an information search on the following 3 databases, Cochrane Library, PubMed, and EMBASE, dating from the inception of the databases to May 1, 2022, with no language restrictions. The keywords utilized in our search strategy were "(esophageal OR esophagus OR oesophageal) AND (carcinoma OR cancer) AND (squamous) AND (metastatic OR advanced) AND (second-line) AND (programmed cell death protein 1 OR PD-1 OR immune checkpoint inhibitors) AND (overall survival OR progression-free survival OR objective response rate OR disease control rate OR treatment-related adverse events)". Details search strategy in Pubmed is presented in Supplemental Table 1 (Supplemental Digital Content 1, http:// links.lww.com/JIT/A776).
Publications retrieved through the database search were collected and screened to exclude duplicate publications with considerable care. Two investigators (Z.J. and M.Z.) independently performed the initial review based on the titles and abstracts of the article to identify potential eligible reports. To evaluate the quality of the research and the reliability of the results, we excluded conference abstracts or meeting reports because they contain only the abstract and cannot provide sufficient data. Potentially relevant reports were subjected to a fulltext review, and the relevance of the reports was confirmed during the data extraction process.

Inclusion and Exclusion Criteria
Studies were included in the meta-analysis if they met the following criteria: (1) patients with advanced, metastatic, or recurrent ESCC who progressed on or were intolerant to previous first-line chemotherapy; (2) single PD-1 inhibitors were administered to the experimental group and standard chemotherapy regimens were delivered to the control group; (3) the study reported reliable clinical efficacy and safety indicators, such as OS, progression-free survival (PFS), objective response rate (ORR), disease control rate (DCR), and treatment-related adverse events (TRAEs); and (4) randomized control studies; The exclusion criteria were as follows: (1) unpublished papers, such as conferences or meeting reports; (2) the outcome index was ambiguous or could not be merged; and (3) non-RCTs.

Data Extraction
The 2 aforementioned investigators independently extracted the data on characteristics and results including the year of publication, trial ID, study design and phase, period of enrollment, treatment arms, sample size, study quality, and the quality of evidence. Any discrepancies between reviewers were resolved by consensus.

Quality Assessment
The risk of bias assessment for the RCTs would be conducted by one researcher and verified by the second using the Cochrane Risk of Bias tool (https://methods. cochrane.org/bias/resources/cochrane-risk-bias-tool). Trials were assessed based on the following domains: (1) bias arising from the randomization process and the timing of identification and recruitment of participants in relation to the timing of the randomization; (2) bias due to deviations from the intended intervention; (3) bias due to missing outcome data; (4) bias due to the measurement of the outcome; and (5) bias due to the selection of the reported result.

Statistical Analyses
All data were gathered and preserved in Microsoft Excel spreadsheets (Microsoft Excel software, Microsoft). Then we performed the meta-analyses using R software, version 4.1.2 (R Foundation for Statistical Computing), and the meta package in R. 20 A 2-sided P value <0.05 was considered statistically significant.
We initially analyzed the clinical efficacy and safety of PD-1 inhibitors versus chemotherapies in the secondline therapy of patients with ESCC. Hazard ratios (HRs) for OS and PFS with the relative 95% CI and incidence of ORR, DCR, and TRAEs (any grade TRAEs and grade 3-5 TRAEs) were respectively extracted from each study. To assess the heterogeneity of the results from individual studies, the Higgins I 2 statistic was calculated, and I 2 > 50% was regarded as a threshold indicating significant heterogeneity between studies. Summary HRs and relative ratios (RRs) were calculated with random or fixed-effect models depending on the heterogeneity of included studies. To confirm the robustness of the results, sensitivity analysis was conducted by omitting one study by turn (Supplemental Fig. 1, Supplemental Digital Content 2, http:// links.lww.com/JIT/A777). The publication bias could be evaluated through a funnel plot, Harbord rank test, and Egger linear regression test. In our study, however, publication bias has not been quantitatively assessed through insufficient RCTs.
Then as for the main clinical efficacy endpoint OS, a subgroup analysis was applied to explore the factors modifying the response to PD-1 inhibitors. Nine subgroups were available for the analyses, including sex, combined positive score (CPS) of programmed death ligand 1 (PD-L1) expression, tumor proportion score (TPS) of PD-L1 expression, Eastern Cooperative Oncology Group performance score, region, race, smoking status, age, organ metastasis, and lymph node metastasis. Because the subgroups within each trial were independent, we could conduct the interaction effect utilizing the forming formulas: log HRAB = log HRAC-log HRBC and SE 2 (log HRAB) = SE 2 (log HRAC)+SE 2 (log HRBC). 21 Then we summarized the HRs in different studies and analyzed the differences in subgroups.
Considering the differences in the participants included in each study and our inability to obtain individual patient data, we did not perform an indirect comparison of the differences in efficacy and safety among PD-1 inhibitors.

Search Results and Study Characteristics
The literature search originally yielded 96 identified articles from the 3 following electronic databases (PubMed, EMBASE, and Cochrane Central). There were 60 records left after the duplicate was discarded. After the screening of titles and abstracts, 18 articles were further scrutinized by a full-text screen. We ultimately conceal 13 articles for different reasons (ie, conferences or meeting posters, reviews, letters, comments, and subgroups of included RCTs) based on the inclusion and exclusion criteria. Five trials were finally included in the quantitative meta-analysis. The flow diagram of the search and selection process is shown in Figure 1.
The 5 included articles (one phase 2 and four phase 3) were all designed as RCTs with low bias. The sample size varied from 190 to 512, with a total of 1970 subjects. Among them, 987 received PD-1 inhibitors including Pembrolizumab, Nivolumab, Camrelizumab, Tirelizumab, and Sintilimab, and the other 983 received standard chemotherapy including paclitaxel, irinotecan, and docetaxel. The total number of males in the study subjects was 1705 (86.5%). Table 1 presents the study characteristics of these 5 studies.

Quality Assessment
The majority of RCTs were of low risk of bias in different domains of the Cochrane Risk of Bias tool (Fig. 2). In KEYNOTE-181, the methodological section is not sufficiently clear in its description of the randomization process, and our final score is "some concerns". All 5 RCTs were considered to be an acceptable risk of bias (Fig. 2).
Afterward, we performed a sensitivity analysis to ascertain whether the individual studies significantly affect the overall results. The results indicated that individual studies had little influence on the final results and demonstrated that the analysis was relatively robust and reliable (Supplemental Fig. 1, Supplemental Digital Content 2, http://links.lww.com/JIT/A777). Due to the comparatively small number of included studies, we did not draw a funnel plot to demonstrate publication bias.

Subgroup Analysis to Explore Factors Modifying the Clinical Efficacy
In the subgroup analysis, no significant difference was observed in subgroups including age, sex, smoking status, race, region, Eastern Cooperative Oncology Group status, organ metastasis, and lymph node metastasis. However, the results demonstrated the lower level of PD-L1 expression was associated with a more diminished response to PD-1 inhibitors.

DISCUSSIONS
In this meta-analysis consisting of 5 RCTs involving 1910 participants, we compared the clinical efficacy and safety profile of PD-1 inhibitors versus chemotherapy for patients with ESCC in the second-line settings. Furthermore, the subsequent subgroup analysis and interaction effect analysis explored the factors modifying the response to PD-1 inhibitors. The results of our study showed that PD-1 inhibitors could provide superior OS benefits, improved remission rates, and safety profiles. This means that PD-1 inhibitors have an advantage over chemotherapy in both short and long-term outcomes in patients to some content. We also discovered that patients with higher levels of PD-L1 expression responded better to treatment with PD-1 inhibitors with regard to OS.
In terms of clinical efficacy, our meta-analysis provided reliable evidence that PD-1 inhibitors could provide superior beneficial effects in OS and ORR in contrast to traditional  chemotherapy in ESCC. The result was in concordance with the previous meta-analysis conducted in 2021. 18,22 Previous research has proved that tumor cells could upregulate PD-L1 expression to inhibit the function of T lymphocytes, thus evading persistent inflammatory or autoimmune reactions. PD-1 inhibitors can block the binding of PD-L1 to PD-1, suppress the proliferation of PD-1 + cancer cells, interrupt the immune escape, and enable the T lymphocytes to eliminate cancer cells. 23 PD-1 inhibitor therapies have shown promising therapeutic effects in patients with metastatic esophageal cancer in many clinical trials in first-line or second-line settings. 24,25 However, while significant OS gains were observed in the immune checkpoint-suppressed group, PFS was not significantly improved. The phenomena were attempted to be explained in terms of lagged effects of immunotherapy and pseudo-progression. 26 Compared with traditional cytotoxic chemotherapies, immunotherapy may require additional time to attain measurable or sustained clinical effects. 27 In the initial phase of immune therapies, the preexisting tumor may not have reduced directly in size but has further expanded into new lesions. The radiographic enlargement of the tumor after the treatment caused a misimpression of disease progression according to the RECIST criteria and was regarded as indicative of ineffective therapy. 28 However, the subsequently decreased tumor burden demonstrated the effectiveness of immunotherapeutic agents. Actually, an immune checkpoint inhibitor might enhance the accumulation and infiltration of inflammatory cells, thus inducing the expansion of tumor lesions, and resulting in misclassification of treated patients as disease progression and reduction in PFS. The phenomenon also called "pseudo-progression," has been described in the previous case report of immune checkpoint inhibitor treatment in digestive system tumors and has been estimated to have an incidence of 6.0% in solid tumors. [28][29][30] The included 5 RCTs, all used the traditional efficacy evaluation criteria for solid tumors (RECIST v1.1) and probably underestimated the PFS in PD-1 inhibitor groups.
According to the TRAEs and grade 3-5 TRAEs results, PD-1 inhibitors might have better drug safety and tolerability compared with standard chemotherapy. In the second-line treatment of ESCC, the past single-agent chemotherapy includes paclitaxel, docetaxel, and irinotecan. Paclitaxel and docetaxel both belong to cytotoxic antitumor drugs and exert their anticancer effects by stabilizing microtubules, blocking the cell cycle, causing abnormal mitosis or termination, and thus restricting the process of tumor cell division. 31 Irinotecan can bind to the topoisomerase I-DNA complex and prevent relegation of the DNA strand and eventually leads to double-strand DNA breakage and cell death. 32 The past single-agent chemotherapy also similarly acts on normal cells, especially in blood cells, stomach and bowel, and hair follicles, resulting in low blood counts, nausea, diarrhea, and hair loss. Immunotherapy for the treatment of cancers can significantly lower the incidence of the aforementioned adverse effects, as has been also established with other tumors. 33 Comparatively speaking, PD-1 inhibitors might have the potential to induce an excessively upregulated immune system and disrupts the "immune balance," thus resulting in a unique set of side effects through unbridled inflammation termed immune-related adverse events, resulting in cardiovascular, endocrine, dermal, hematological, neurological, pulmonary, gastrointestinal, musculoskeletal, and ocular damage 34,35 These special complications present a challenge to oncologists to evaluate the appropriateness of referrals and management regimens for patients.
In the subgroup analysis, we concluded that PD-L1 CPS was positively associated with the patient's OS. The expression status of PD-L1 in tumors has been utilized as a biomarker to predict the clinical efficacy in several malignancies. 36,37 Nevertheless, there are 2 major established methods to evaluate positive PD-L1 cells. 38 TPS is defined as the number of positive tumor cells divided by the total number of viable tumor cells multiplied by 100%. CPS is described as the ratio of positive tumor cells, lymphocytes, and macrophages to the total number of viable tumor cells. 39 Theoretically, CPS might provide a more precise depiction of the tumor microenvironment and describes the extent of PD-L1 status more accurately. Our analysis results seemed to also confirm that CPS was more predictive of the additional benefit of PD-1 inhibitors compared with traditional chemotherapy in the second-line treatment of patients with ESCC. Further studies are still needed to explore the role of PD-L1 expression in immunotherapy.
In addition, we discovered that, although not statistically significant, women with advanced ESCC probably acquired a greater survival benefit with PD-1 inhibitors. The conclusion contradicted the findings of previous investigations on solid tumors. A meta-analysis included more than 11,000 patients with different kinds of solid tumors (mostly lung carcinoma and melanoma) and ultimately concluded that the immune checkpoint inhibitors might decrease the risk of death more in males compared with females. 40 However, another two studies concerning the immunotherapy of advanced cancers reported no statistically significant association between OS and participants' sexes. 41,42 Regarding the sex factor, it cannot be ignored that the imbalance in the sex ratio of participants contributed to the challenge of determining the interaction between sex and efficacy. In the 5 included RCTs concerning esophagus carcinoma, the ratio of male to female was 8-1, a disparate ratio. Future clinical designs should ensure sufficient female subjects to more comprehensively assess sex differences in clinical efficacy.
There are also some limitations that should be noted. Firstly, the standard chemotherapy regimen for the control group in the different studies varied slightly; for instance, Paclitaxel was used for Keynote 181, whereas Docetaxel was used for Attraction-3. This possibly affected our results to a certain extent. Secondly, the expression status of PD-L1 was determined using several antibodies, 28-8 assay in ATTRACTION-3, SP263 assay in RATIONALE-302, 22C3 assay in KEYNOTE-181 and ORIENT-2, and 6E8 assay in ESCORT, which might have an influence on the evaluation of PD-L1expression. Thirdly, this study was conducted at the trial level, and the individual patient data are unavailable, which may diminish the accuracy of the result. Fourth, due to the insufficient number of randomized controlled trials included, we did not apply meta-regression to explore the sources of heterogeneity but applied subgroup analysis directly. Despite these limitations, the meta-analysis provided a thorough summary of the known RCTs on the second-line treatment of ESCC.

CONCLUSIONS
Our analysis suggested that PD-1 inhibitors exhibited better survival outcomes and safety profiles than standardof-care chemotherapy in the second-line settings of patients with ESCC. The subsequent subgroup analysis revealed that the PD-L1 CPS was positively associated with OS. More studies with larger sample sizes and longterm follow-ups are needed to evaluate the safety and efficacy of PD-1 inhibitors in ESCC.