2.1 Search strategy and selection criteria

We searched ClinicalTrials.gov (www.clinicaltrials.gov, CT.gov) and the Chinese Clinical Trial Registry (www.chictr.org.cn, ChiCTR; the main location for registration of Chinese studies) for cancer-related studies registered in mainland China from January 1, 2010 to December 31, 2019. The medical subject headings (MeSH) terms “cancer” OR “tumor” OR “neoplasm” OR “carcinoma” OR “leukemia” OR “lymphoma” were used for the preliminary screening. The location was restricted to mainland China. In this study, IIT was defined as a clinical study initiated by an investigator at a hospital, university, or research institute; thus, studies were included if the applicant's institution in ChiCTR or the responsible party in CT.gov was a hospital, university, or research institute. We excluded duplicates, industry-initiated studies, noncancer-related studies, and studies initiated by sites not on the Chinese mainland. Studies related to cancer supportive care, defined as the best care for cancer pain, cancer-related fatigue, and various adverse effects (e.g., vomiting, leukopenia) experienced during chemotherapy, were also excluded based on the primary endpoint. The final list of cancer-related IITs in the mainland China were included in the analysis.

2.2 Data extraction and assessment

Two of the authors (YC and LM) independently searched CT.gov and ChiCTR with the same search algorithm to confirm the identified studies on April 4, 2020. Two other authors (WJ and DY) extracted the studies from the two registries. Deduplication was performed, both as intra-database deduplication (i.e., deduplication of studies with partially identical characteristics in the same database) and interdatabase deduplication (i.e., deduplication of the same study registered in both databases). Two authors (LM and LY) screened the duplicates in each of the registries according to the registry ID. Studies in both CT.gov and ChiCTR were deduplicated if they were identical in terms of title, leading site, investigator, and enrollment. Studies initiated by industry as well as studies not initiated in the mainland China, were excluded by two authors (LM and LY) according to study sponsor and location. Two authors (WY and LM) assessed the data to determine the cancer type for each study as well as to identify noncancer-related studies, and cancer supportive care studies were also removed at this stage. The above screening and assessments were performed independently by two authors. Next, 24 oncological specialists (volunteers) were divided into three groups to identify detailed interventional measures, including drug type (including cell immunotherapy). One author (YC) was consulted in the case of disagreement.

Tumor types were classified into primary classification by MeSH subject word (C04.588), and specific cancer types (e.g., lung cancer, breast cancer) were classified as the secondary classification. Where the disease could not be classified according to the specified primary classification (e.g., sarcoma and melanoma) or could be classified under a different primary classification according to inclusion criteria, the designation “Others” was used for the primary classification. Less frequently studied cancer types (fewer than five studies) under the primary classification were also classified as Others, and studies of diseases that could be included as multiple cancer types under the same primary classification based on inclusion criteria were classified into the multi-tumor classification (multi) (Supporting Information: Table S1).

Interventional studies involving drugs were classified into four categories according to drug category: cytotoxic therapy, targeted therapy, hormone therapy, and immune checkpoint inhibitor therapy. Further subdivision was not conducted owing to the scarcity and relative complexity of other types of drug. In addition, it was found that some studies classified cellular immunotherapy as a type of drug therapy, so cellular immunotherapy was included in this analysis.

2.3 Data synthesis

Owing to the pattern heterogeneity of CT.gov and ChiCTR, we reassigned the study category (e.g., drug, procedure, radiation), funding type, responsible party, masking, and randomization to ensure common pattern homogeneity. Given that the majority of multicenter studies in both databases had missing site information, the geographic location where studies initiated was identified based on the primary sponsor's address in ChiCTR and the responsible party's address in CT.gov. Indicators such as study type (e.g., interventional, observational) and study design were not reassigned because the original definition and/or study concept differed between the CT.gov and ChiCTR databases. For example, the concept of observational study was quite different between these two registries.

2.4 Statistical analysis

Data processing and statistical analyses were conducted using Excel 365 (Microsoft Corporation). All variables were assessed in a descriptive manner, providing number and percentage for qualitative variables and ranges, medians, and interquartile range (IQR) for quantitative variables. A simple regression model was used to analyze 10-year trends in the number of initiated IITs and the number of drugs. A simple stratification analysis was performed for specific indicators (e.g., drug type, endpoint distribution) by tumor type and study type.

3.1 Results of the search strategy

We retrieved a total of 12,430 cancer-related clinical studies that initially met our search requirements in CT.gov and ChiCTR. A total of 8199 cancer-related IITs from medical or scientific study institutions in the mainland China from 2010 to 2019 were subsequently included in our analysis after verifying the inclusion and exclusion criteria (Figure 1).

3.2 Timing trends of initiated IITs

The number of cancer-related IIT registrations has increased significantly over time (see Supporting Information: Videos S1 and S2), with an average annual increase of 31.8% (Figure 2). Studies in 2018 and 2019 accounted for 46.4% (3805/8199) of the total, of which the number in 2019 was 11.4 times (2116/185) that in 2010. Registrations on ChiCTR have surpassed those on CT.gov since 2018, with a 45.5% increase (313/688) in 2018 compared with CT.gov and a significant increase of 77.8% from the previous year in 2018 alone.

The primary classification of cancer-related IITs included 11 major categories, with pediatric tumors and gynecological tumors as two independent categories (Figure 3a). Of the IITs in which cancer classification was identified, digestive-tract tumors were most commonly studied (38.9%, 3193/8199), followed by thoracic tumors (15.8%, 1296/8199) and hematological tumors (10.9%, 891/8199).

The top 10 cancer types in the secondary classification of cancer-related IITs are shown in Figure 3b. Of these, lung cancer (15.5%, 1267/8199) was the most frequently studied cancer type, followed by breast cancer (9.5%, 783/8199), colorectal cancer (9.2%, 752/8199), and hepatic cancer (9.1%, 746/8199). In addition, 5 of the top 10 cancer types were digestive tumors, namely colorectal cancer, liver cancer, gastric cancer, esophageal cancer, and pancreatic cancer, accounting for 33.9% (2779/8199).

As shown in Figure 3c, the distribution of patients by age showed that, of the 303 IITs in which only adult patients (aged ≥18, <65 years) were included, patients with breast cancer (21.5%, 65/303), leukemia (14.5%, 44/303), cervical cancer (8.6%, 26/303), and nasopharyngeal cancer (8.6%, 26/303) were the most common; in 44 IITs in which only elderly patients (≥65 years) were included, digestive tumors were the most studied (52.3%, 23/44). In the primary classification, there were a total of 52 IITs on pediatric tumors, most of which were on hematological tumors, with IITs on leukemia accounting for 48.1% (25/52) and lymphoma accounting for 11.5% (6/52) of studies. With the exception of studies on pediatric tumors, 798 studies included patients aged 0–17 years, with the highest proportion of 59.4% (19/32) in bone tumors, followed by 51.2% (186/363) in leukemia, and 25.9% (91/352) in lymphoma.

We integrated the phase, randomization, masking, and enrollment of cancer-related IITs on the two registration websites (Supporting Information: Table S2). Phase 0 (namely pilot studies) was unique to ChiCTR, which mainly included diagnostic studies (n = 99), interventional studies (n = 333), and observational studies (n = 216). All 56 early phase 1 studies came from CT.gov, and were mainly focused on drugs (n = 27) and biological agents (n = 18). There were 3898 phase 1 to phase 4 studies, among which phase 2 accounted for the majority, at 39.0% (1521/3898). Most phase 3 studies adopted randomized methods (92.5%, 577/624), whereas the use of blinded settings was generally low (17.1%, 107/624). The proportion of randomized and double-blind design was only 8.0%, of which phase 3 was the highest at 16.0%. A notable gap between the average and median of recruited patients in phase 4 was attributed to the fact that one study recruited 200,000 patients. There were 437 clinical studies with sample sizes ≥1000, of which 29.1% (127/437) were interventional studies and 51.5% (225/437) were observational studies. Among the studies with large sample sizes, the top five cancer-related clinical studies were focused on colorectal cancer (n = 62), lung cancer (n = 60), breast cancer (n = 54), gastric cancer (n = 37), and hepatic cancer (n = 36). The differences in other classifications of study design between CT.gov and ChiCTR are shown in Supporting Information: Tables S3 and S4.

3.3 Drug classification of interventional cancer-related IITs

A total of 5215 interventional studies were classified according to specific interventional measures. Cancer-related IITs were mainly focused on drugs (44.2%, 2305/5215), procedures (14.9%, 775/5215), and combination treatment (21.0%, 1093/5215). Behavioral intervention and diet therapy, as supplement treatment, accounted for a relatively small proportion of studies: 0.9% (49/5215) and 1.1% (59/5215), respectively. In addition, 12 cancer-related IITs adopted gene therapy as an intervention.

As shown in Figure 4 and Supporting Information: Table S5, the overall development of interventional studies on different drug types (including combination administration) over the past decade appeared to increase year on year, but this trend was not significant for hormone therapy. Cytotoxic drug therapy, as conventional treatment, was the focus of the largest number of studies, and cytotoxic drugs in combination with targeted drugs were the most common studies among combination therapeutic strategies. In 2016, studies on immune checkpoint inhibitors began to appear, with a significant year-on-year growth trend for combination studies with cytotoxic, targeted, and hormone agents. For immune cell therapy, a total of 440 studies have been carried out in the past decade.

3.4 Geographic distribution of studies

Although cancer-related IITs were carried out in 143 cities of 30 provinces/municipalities in mainland China, a geographic trend toward study location in economically developed cities (Shanghai, Beijing, and Guangdong) was apparent; these studies represented half of the total (51.4%, 4212/8199), while fewer than 10 studies took place in Qinghai, Ningxia, and Hainan, with smaller populations (Table 1). Cancer-related IITs in each province were mainly concentrated in the capital city (see Supporting Information: Videos S3 and S4). Studies carried out in 19 capital cities (excluding municipalities) accounted for more than 80% of all studies in their respective province, of which IITs in four provinces were carried out only in medical and scientific institutions in the provincial capital.

High regional incidence of specific tumor types had a strong correlation with the number of clinical studies performed in that region. Nasopharyngeal cancer occurs with a high incidence rate in Guangdong and Guangxi provinces, and studies on this cancer type accounted for 44.9% (133/296) and 9.1% (27/296) of the total nasopharyngeal cancer studies performed nationwide, ranking first and second, respectively. In Henan Province, which has a high incidence of esophageal cancer, 204 IITs were performed in the past decade, of which 57 were studies on esophageal cancer, accounting for 27.9% (57/204) and 12.6% (57/453) of the total clinical esophageal cancer studies in Henan Province and nationwide, respectively.

3.5 Funding type

Cancer-related IITs were divided into nine categories according to funding category (Figure 5). For 114 studies, more than one funding source was reported. Additionally, 14.3% (1176/8199) of IITs received government funding, with the average funding rate fluctuating by around 13% over 10 years.

Although over 30% of funding for cancer-related IITs came from only four provinces, all these provinces carried out less than 50 studies. However, in Guangdong province, which was among the top three provinces for number of IITs performed, only 8.1% of cancer-related IITs received government fundings, while the majority of studies (35.7%; 474/1328) received funds from colleges and universities, twice the national average. Across all studies, the majority were funded by hospitals, accounting for 33.4% (2739/8199) in total; the hospital funding rate exceeding 40% in seven provinces. Additionally, 18.7% (1534/8199) of studies required funds raised by the investigators. Few studies (2.9%, 237/8199) were financed by social funds (such as nonprofit foundation), and 8.8% (722/8199) of studies were funded by industry.

3.6 Chronological shifts in IITs characteristics

The characteristics of IITs of the periods from 2010 to 2014 (n = 1630) and from 2015 to 2019 (n = 6569) are listed in Table 2. Compared with trials in 2010–2014, a higher proportion of trials in 2015–2019 were registered before the first participants enrolled (64.9% [4264 of 6569] vs. 40.2% [656 of 1630], p < 0.001) and had a data monitoring committee (DMC)(44.5% [2921 of 6569] vs. 40.0% [652 of 1630], p = 0.001).

The proportion of randomized trials in 2015–2019 was lower than that in 2010–2014 (51.5% [1151 of 2234] vs. 62.7% [534 of 852], p < 0.001), whereas the proportion of blinded trials was comparable between the two periods. With regard to study phase, the proportion of trials in phase 3 decreased from 19.2% (199 of 1036) in 2010–2014 to 14.8% (425 of 2862) in 2015–2019 (p = 0.001), with no significant difference in other phases between the two periods. In terms of age selection, older adults were less likely to be enrolled in 2015–2019 (87.9% [1432 of 1630] vs. 82.7% [5342 of 6569], p < 0.001).

Differences in study intervention were also apparent. The proportion of trials involving cytotoxic drug therapy decreased from 48.9% (579 of 1184) in 2010–2014 to 34.8% (1403 of 4031) in 2015–2019 (p < 0.001), whereas an increase was observed in the proportion of trials involving targeted therapy (14.2% [168 of 1184] in 2010–2014 vs. 17.8% [716 of 4031] in 2015–2019, p = 0.004), immune checkpoint inhibitor therapy (0% [0 of 1184] vs. 6.6% [265 of 4031], p < 0.001), and immune cell therapy (4.5% [53 of 1184] vs. 9.6% [387 of 4031], p < 0.001).

With respect to study funding, the proportion of funded trials decreased from 56.3% (917 of 1630) in 2010–2014 to 36.4% (2394 of 6596) in 2015–2019 (p < 0.001), especially for phase 3 trials (20.0% [183 of 917] vs. 15.2% [382 of 2394], p = 0.006). However, the funding for phase 1 and phase 1/2 increased from 8.7% (80 of 917) in 2010–2014 to 12.2% (291 of 2394) in 2015–2019 (p = 0.005), and from 7.3% (67 of 917) to 11.9% (285 of 2394) (p < 0.001), respectively, while funding for other phases did not change substantially between the two periods.