A bibliometric analysis of immune-related adverse events in cancer patients and a meta-analysis of immune-related adverse events in patients with hepatocellular carcinoma

Data sources and search strategies

We performed bibliometric analyses using the SCI-expanded of WoSCC bibliographic database. To allow for rapid database updates, the literature search was conducted on a single day (September 15, 2022) to avoid bias. In this study, the publication period in this study was set between 1999 and 2022. The search terms were presented as follows: TS = (“Liver Neoplasms” OR “Neoplasms, Hepatic” OR “Neoplasms, Liver” OR “Liver Neoplasm” OR “Neoplasm, Liver” OR “Hepatic Neoplasms” OR “Hepatic Neoplasm” OR “Neoplasm, Hepatic” OR “Cancer of Liver” OR “Hepatocellular Cancer” OR “Cancers, Hepatocellular” OR “Hepatocellular Cancers” OR “Hepatic Cancer” OR “Cancer, Hepatic” OR “Cancers, Hepatic” OR “Hepatic Cancers” OR “Liver Cancer” OR “Cancer, Liver” OR “Cancers, Liver” OR “Liver Cancers” OR “Cancer of the Liver” OR “Cancer, Hepatocellular”) AND TS = (“immune related adverse events” OR “immune related adverse events” OR “irAEs” OR “irAE”). We included only original articles and reviews written in English. Two researchers independently conducted the original data search, and all potential discrepancies were resolved through discussion. Our study ultimately analyzed 2946 articles. The particular screening flowchart is shown in the Figure 1 below.

Figure 1

Flowchart of screening process.

Data collection

Extract the original data from the SCI-expanded database. The information includes title and number of citations, H-index, year of publication, country/region, author, journal, and keywords.

Bibliometric analysis

Bibliometric analysis are extensively to decipher the characteristics of the relevant publications in the specific scientific field. The productivity is generally measured by the number of publications (Np), and the impact was assessed by the number of citations without self-citation (Nc), thus, Np and Nc are the two main aspects represent the level of study.^[10,11] The H-index unifies productivity and impact by finding a threshold that connects Np and Nc. Although the H-index was initially used to assess individual academic success, it has extended to describe the publication output of country/ region, institution, and journals.^[12] In addition, the impact factor (IF) index mainly refers to the latest edition of the Journal Citation Report (JCR), has been widely regarded as indicator of the impact of medical journals.^[13] Local Cited Scores (LCS) are important indicator to evaluate an author and considered to be the Nc of an article.^[14] The Bibliometrix package^[9] in Rstudio (version 2022.02.2) is used for basic bibliometric analysis. To further account for the changes in the number of documents in the year, a fitting polynomial model is used to predict the annual Np. The variable f(x) represents the number of studies per year, and x represents the year of publication. In addition, a bibliometric map was constructed using VOSviewer software to obtain more comprehensive outcome based on co-citation and co-occurrence.^[15] If both items were cited by the third item, they were defined as co-cited. The co-occurrence of keywords measures the most frequently occurring keywords in the same document,^[16] and analysis of keywords can identify research hotspots for irAEs in cancer patients treated with ICIs.

Literature search

We searched PubMed, Embase, and Cochrane databases from database inception to March 22, 2022 (Last search time). There was no language restriction in this meta-analysis. The following search terms were included: (hepatocellular carcinoma) AND (immune checkpoint inhibitor OR PD-1 inhibitor OR PD-L1 inhibitor OR CTLA-4 inhibitor OR pembrolizumab OR nivolumab OR atezolizumab OR durvalumab OR avelumab OR cemiplimab OR ipilimumab OR toripalimab OR sintilimab OR camerlizumab OR tislelizumab) AND (immune-related adverse event OR adverse event). The complete search strategy and results are summarized in Table S1.

Inclusion and exclusion criteria

Studies eligible for inclusion met all the following criteria: (1) participants diagnosed with HCC who were treated with ICIs (PD-1/PD-L1/CTLA-4 inhibitors) alone or in combination with other agents (e.g., other drug or surgery); (2) reporting the incidence of irAEs. Exclusion criteria were as follows: (1) fewer than 5 patients in the ICIs group; (2) studies involving multiple advanced tumors; (3) conference abstracts, basic studies, reviews, letters, case reports, meta-analyses; (4) unpublished relevant data from trials still in progress.

Data extraction

The following pieces of information were extracted from each study: first author, year of publication, study type, sample size, liver function [Child-Pugh classification and Barcelona Clinic Liver Cancer, (BCLC) classification], Eastern Cooperative Oncology Group (ECOG) performance statuses, ICIs agent, dose, duration of follow-up, number of people with any-grade of irAEs, number of people with grade ≥ 3 irAEs, number of patients with treatment discontinuation due to trAEs, number of treatment-related deaths. Severity was graded according to common terminological criteria for adverse reactions (CTCAE). Those described as AEs of special interest and selected AEs suspected to be potential irAEs were also extracted as irAEs in the present study. The data extraction process was performed independently by the two investigators and any disputes were resolved through discussion.

Quality assessments

The two investigators independently conducted the quality evaluation process for the included studies. The six randomized controlled trials (RCTs) were evaluated following the Cochrane risk of bias tool.^[17] The evaluation included the following five aspects: sequence generation, allocation concealment, blinding, completeness of outcome data, and other sources of bias. Sixteen cohort studies were evaluated according to the Newscar-Ottawa (NOS) scale,^[18] ranging between zero up to nine stars, which contains the following three main aspects: cohort selection, comparability, and outcomes. Other studies (6 single-arm studies, and 1 pilot trial) were evaluated according to the MINORS scale,^[19] quality was classified as low (0–8), mediate (9–16), high (17–24).

Statistical analysis

The incidence of irAEs was pooled for the included studies in this study and heterogeneity between studies was assessed by Q-test and the I² statistic, and P < 0.10 indicated apparent heterogeneity. Heterogeneity was classified as low (I² < 25%), mediate (I² 25%–75%), and high (I² > 75%). A random-effects model was used if I² > 50 and a fixed-effects model was used if I² < 50. We selected the top 10 incidences of irAE for subgroup analyses, meta-regression, and publication bias test. Subgroup analysis of the incidence and profile of irAEs according to treatment strategy. Multivariate regression analysis was performed to assess risk factors associated with the incidence of irAEs based on treatment strategy (ICIs monotherapy, ICIs combined with other treatments), type of ICIs (anti-PD-1, anti-PD-L1, and anti-CTLA-4), and median patient age (≥ 60, < 60). Publication bias was assessed by funnel plot and Egger’s test. In addition, some irAEs were not subjected to this regression analysis, because of the number of studies less than ten.

All statistical analyses were performed using Stata statistical software version 15.0 (Stata Corp, USA, https://www.stata.com)

Overview

Based on the search strategy, a total of 2946 articles and reviews were retrieved. The total Nc of the retrieved articles was 102,550.26, and the average Nc of each article was 34.81. The H-index for all publications is 140.

Annual trend of publication volume

Figure 2A shows the annual Np associated with irAEs in cancer. Overall, the Np rose from 1 in 1999 to a peak of 749 in 2021, despite fluctuations in the rate of growth. The Figure 2B is a polynomial fitting curve of the trend of Np. Np is significantly correlated with the publication year and the correlation coefficient R2 reaches 0.8552. In addition, as shown in the Figure 3A, Figure 3B, the productivity in this field is led by China (1,269,993) and the United States (2,756,866), and there is a big gap productivity between the two, which means that China still has a lot of space for development. Moreover, the cooperation between China, the United States, and Japan is quite close. Overall, these findings indicate that irAEs in cancer treatment has become a key point of attention and has entered a phase of rapid development.

Figure 2

The number of publications and citation by year (A). Polynomial-fitting curve of the annual trend of publication quantity (R2 = 0.8552) (B).

Figure 3

Network visualization map of publications in the countries/regions (A). Density visualization map of publications in the countries/regions (B).

Analysis of authors

The top 10 authors are listed in Figure 4A according to H-index. Jedd D Wolchok from the United States has the highest H-index, followed by Caroline Robert from France and Yuping Wang from China. In addition, most of these authors are from the United States (4) or France (4), while the other two authors are from China and the United Kingdom, respectively.

Figure 4

Top ten authors analysis according to H-index (A). Top ten paper analysis of local cited scores (B).

Analysis of paper local citations (LCS)

The local citations (LCS) for the top 10 articles is presented in Figure 4B. The paper written by BRAHMER JR in 2018 has the highest LCS (605). This article reviews recommended management strategies and provided guidelines for irAEs in patients treated with ICIs. In addition, other articles have elaborated irAEs from different perspectives, mainly focusing on the recommended management strategies of irAEs, the mechanism of irAEs, the efficacy and safety of ICIs, and the characteristics of each irAEs. Among these literatures, most of them were review and published after 2015, which provided a relatively comprehensive description of irAEs.

Analysis of journals

As shown in the Table 2, JOURNAL FOR IMMUNOTHERAPY OF CANCER (145 articles, IF: 12.469) was the most productive journal and has the highest IF, while FRONTIERS IN ONCOLOGY (105, IF: 5.738) and FRONTIERS IN IMMUNOLOGY (IF: 8.786) ranked second and third according to productivity. About 25% of the papers published in the top 10 journals (713/24.2%), and had high IF (defined as greater than 3.000). Notably, the EUROPEAN JOURNAL OF CANCER (IF = 10.002) and ONCOLOGIST (IF = 5.837) have a higher citations and H-index, but their productivity were low respectively.

Analysis of research hotspots

A total of 5951 keywords were extracted from the 2496 articles and reviews. A density map was generated for keywords with the co-occurrence greater than 100 times, which includes 43 keywords in the map. As shown in Figure 5A, nivolumab was the most frequent keyword, with 1145 co-occurrences, followed by immune checkpoint inhibitor (1166), immune related adverse event (1113).

Figure 5

Density visualization map of high frequency keyword (A). Network visualization map of co-occurrence and clustering analysis of the frequent 52 Keywords (B).

Clustering analysis and a network map were performed for co-occurrence keywords by VOSviewer (Figure 5B). Keywords were extracted from titles and abstracts of the 2496 papers. The cumulative frequency of the keywords was calculated and the threshold was set to 40.00% because the keywords with high frequency can accurately reveal the main topic of a field. After calculation, a network map for keywords appearing more than 100 times was generated. There were 43 nodes and links in the network map, and the 43 keywords with high-frequency formed 4 clusters. Cluster 1 was the largest cluster included 21 keywords, mainly related to mechanism of ICIs. Cluster 2 contained 14 keywords, reflected the main focus in clinical trials. Cluster 3 comprised of 4 keywords, mainly related to prognosisof cancer. Cluster 4 included 4 keywords mainly about adverse event and management in the cancer patients. The top frequent occurrences of keywords were “nivolumab,” “immune checkpoint inhibitor”, “immune related adverse event”, and “immune related adverse event” has a strong relationship with almost all of elements, suggesting that the researches related to ICIs in cancer mainly focused on irAEs.

Analysis of research hotspots based on types of cancer

ICIs, as a new treatment in this field, have been widely used in various tumors since their advent, which has changed the current treatment mode to a certain extent. However, while ICIs bring significant survival benefits to patients, immunotoxicity to various organs has also become an unavoidable problem in clinical practice. Based on the above bibliometric analysis, in order to further analyze the research progress of irAEs in various solid tumors, keywords of the same type of tumors were combined (e.g., In term melanoma and advanced melanoma merged) and selected the top 10 tumor-related keywords, as shown in the Table 3 below.

After further retrieval of relevant articles on ICIs in HCC and irAEs in HCC, as shown Figure 6A, Figure 6B, the research on ICIs in HCC is superior to that on irAEs in HCC in terms of publication time and number of publications. Since 2017, the number of publications on irAEs in HCC has been increasing year by year. Combined with the above bibliometrics results and the world liver cancer epidemiology data, we found that the study of irAEs in HCC is of great significance. Therefore, we further conducted a meta-analysis on the incidence of irAEs in patients with HCC.

Figure 6

The number of publications by year in ICIs and hepatocellular carcinoma (A). The number of publication by year in immune-related adverse events and hepatocellular carcinoma (B).

Literature search

Our search strategy yielded 2381 studies, of which 694 studies were removed due to duplication. 1658 studies were excluded according to the exclusion criteria. Finally, 29 articles including 3066 HCC patients were eligible for the present meta-analysis. The study selection is shown in Figure 7A.

Figure 7

Flow chart of the included studies (A). The incidence of each type of immune-related adverse events (B).

Study characteristics

A total of 29 studies were included in our study which contained 16 cohort studies,^{[20,21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35]} six RCTs,^{[36, 37, 38, 39, 40, 41]} one pilot study,^[42] and six single-arm studies.^{[43, 44, 45, 46, 47, 48]} Thirty-five cohorts of 3066 patients were included, with seven cohorts included in the monotherapy cohort and 28 cohorts included in the combination therapy cohort. There were 17 studies involving anti-PD-1 antibody, 10 studies involving anti-PD-L1 antibody, 1 study involving ant-CTLA-4 antibody, and 1 study involving anti-PD-1 with anti-CTLA-4 antibodies. The primary characteristics of the eligible studies are presented in Table 4. Majority of participants were identified as Child-Pugh classes A or B and ECOG performance status of 0–2, and most were at BCLC stage B or C. In addition, most patients had background liver disease, which may increase the incidence of irAEs.

Incidence of irAE

Of the 29 included studies, a total of 6 studies reported overall any-grade irAEs and 7 studies reported overall grade ≥ 3 irAEs, with incidences of 61.0% (95% CI 38.5%–81.3%, I² = 98.02%; Figure S1), 13.2% (95% CI 7.9%–19.6%, I² = 87.11%; Figure S2). Eighteen studies reported trAEs leading to treatment discontinuation, with an incidence of 10.7% (95% CI 6.3%–16.0%, I² = 89.08%; Figure S3). Deaths due to trAEs were reported in 17 studies, with an incidence of 3.1% (95% CI 0.8%–6.3%, I² = 86.74%; Figure S4). The incidence of each type of irAEs is shown in the Figure 7B: the most common any-grade irAEs were RCCEP (44.8%), fatigue (28.9%), hyperbilirubinemia (28.8%), elevated AST (25.0%), proteinuria (23.2%), elevated alanine aminotransferase (ALT) (22.9%), hand–foot skin reaction (22.1%), hypertension (21.8%), thrombocytopenia (19.7%), and fever (19.4%). The most common grade ≥ 3 irAEs were elevated AST (6.2%), hypertension (5.3%), elevated ALT (4.3%), thrombocytopenia (4.0%), hyperbilirubinemia (3.9%), proteinuria (3.4%), increased γ-glutamyltransferase (3.2%), neutropenia (2.6%), gastrointestinal bleeding (2.5%), and hand–foot skin reaction (2.3%).

Organ-specific irAEs

With regards to organ-specific irAEs, we summarized the incidence of irAEs in HCC patients treated with ICIs involving the skin, gastrointestinal tract, endocrine, liver, kidney, hematological system, lung, and others. The most common any-grade irAEs involved the skin (13.4%– 44.8%). The grade ≥ 3 irAEs mostly occurred in the liver (3.2%–6.2%). The detailed information is presented (Figure S5-74)

Skin

There were four categories of skin-related irAEs, including pruritus, rash, hand-foot skin reaction, and RCCEP. RCCEP was the most common of the any-grade irAEs, with an incidence of 44.8% (95% CI 22.2%-68.7%, I² = 95.14%), and hand-foot skin reaction was the most common of the grade ≥ 3 irAEs, with an incidence of 2.3% (95% CI 0.7%–1.4%, I² = 69.22%).

Gastrointestinal tract

A total of six classes of irAE were associated with the gastrointestinal system, including nausea and vomiting, gastrointestinal bleeding, colitis, diarrhea, abdominal pain, and constipation. Of these, abdominal pain was the most common of the any-grade irAEs, with an incidence of 18.7% (95% CI 12.7%–25.4%, I² = 81.94%), and gastrointestinal bleeding was the most common of the grade ≥ 3 irAEs, with an incidence of 2.5% (95% CI 1.3%–3.9%, I² = 0%).

Endocrine

There were six classes of irAE that related to endocrine, including hypothyroidism, hyperthyroidism, thyroiditis chronic, hypophysitis, adrenal insufficiency, and hyperglycemia. Of these, hypothyroidism was the most common of the any-grade irAEs, with an incidence of 13.2% (95% CI 9.9%–16.8%, I² = 69.05%), and adrenal insufficiency and hyperglycaemia were the most common of the grade ≥ 3 irAEs, both with an incidence of 0.6% (95% CI 0%–1.6%, I² = 25.74%) (95% CI 0.1%–1.3%, I² = 0%).

Liver

There were four classes of liver-related irAE, including hyperbilirubinemia, elevated AST, elevated ALT, and increased γ-glutamyltransferase. Of these, hyperbilirubinemia was the most common 28.8% (95% CI 17.0%–42.2%, I² = 92,49%) of any-grade irAEs, and elevated AST was the most common of grade ≥ 3 irAEs with an incidence of 6.2% (95% CI 4.0%–8.9%, I² = 73.19%).

Kidney

There were two categories of kidney-related irAE, including proteinuria and nephritis. Of these, proteinuria was the most common in both any-grade irAEs and grade ≥ 3 irAEs, with an incidence of 23.2% (95% CI 17.4–29.6%, I² = 86.62%) and 3.4% (95% CI 2.5%–4.4%, I² = 33.24%), respectively.

Hematological system

There were four categories of hematological-related irAE, including: thrombocytopenia, leucopenia, neutropenia, and anemia. Thrombocytopenia was the most common irAE in both any grade and grade ≥ 3 irAE, with incidences of 19.7% (95% CI 10.9%–30.2%, I² = 93.39%) and 4.0% (95% CI,2.8%–5.2%, I² = 41.39%), respectively.

Lung

Two categories of lung-related irAE including: cough and pneumonia were summarized. Of these, any-grade irAEs were most common in cough, with an incidence of 12.0% (95% CI 6.2%–19.2%, I² = 66.60%), and grade ≥ 3 irAEs were most common in pneumonia, with an incidence of 1.6% (95% CI 0.9%–2.5%, I² = 0%).

Other

Other irAEs included hypertension, infusion reactions, fever, decreased appetite, mouth ulcers, and fatigue. Of the any-grade irAEs, fatigue was the most common irAE, with an incidence of 28.9% (95% CI 22.8%–35.3%, I² = 86.23%), and of the grade ≥ 3 irAEs, hypertension was the most common irAE, with an incidence of 5.3% (95% CI 2.6%–8.6%, I² = 85.47%).