Disease characteristics and treatment patterns of Chinese patients with colorectal cancer: a retrospective study using medical records from China

Background: Colorectal cancer (CRC) is the third most prevalent cancer in China but few large-scale studies were conducted to understand CRC patients. The current study is aimed to gain a real-world perspectives of CRC patients in China. Methods: Using electronic medical records of sampled patients between 2011 and 2016 from 12 hospitals in China, a retrospective cohort study was conducted to describe demographics and disease prognosis of CRC patients, and examine treatment sequences among metastatic CRC (mCRC) patients. Descriptive, comparative and survival analyses were conducted. Results: Among mCRC patients (3,878/8,136, 48%), the fluorouracil, leucovorin, and oxaliplatin (FOLFOX) and other oxaliplatin-based regimens were the most widely-used first-line treatment (42%). Fluorouracil, leucovorin, irinotecan (FOLFIRI) and other irinotecan-based regimens dominated the second-line (40%). There was no a dominated regimen for the third-line. The proportion of patients receiving chemotherapy with targeted biologics increased from less than 20% for the first- and second- lines to 34% for the third-line (p<0.001). The most common sequence from first- to second-line was from FOLFOX and other oxaliplatin-based regimens to FOLFIRI and other irinotecan-based regimens (286/1,200, 24%). Conclusions: Our findings reflected a lack of consensus on the choice of third-line therapy and limited available options in China. It is evident o continue promoting early CRC diagnosis and to increase the accessibility of treatment options for mCRC patients. As the only nationwide large-scale study among CRC and mCRC patients before more biologics became available in China, our results can also be used as the baseline to assess treatment pattern changes before and after more third-line treatment were approved and covered into the National Health Insurance Plan in between 2017 - 2018.

per 100,000 in 2014 [1,2]. Besides age, there has been evidence that a diet high in fats and low in fruits and vegetables increases the risk of developing CRC [3,4]. Because of the rapidly aging population and in the increasing fat intake in China, CRC incidence is expected to continue increasing.
Moreover, the National Central Cancer Registry (NCCR) showed that prognosis of CRC was much poorer in China compared with developed countries [5].
Despite the significant disease burden, there is limited information on CRC patient characteristics and disease patterns in China. The NCCR synthesizes data collected from local registries in China and reports basic statistics such as incidence and mortality by key risk groups at the national level [6].
Population-based studies were usually conducted in particular geographic regions [7][8][9][10], among a specific group of patients [8], or out-of-date [11]. For CRC patients with metastatic colorectal cancer (mCRC), treatments can be complex. As multiple chemotherapeutic and targeted biologic agents emerged, treatment patterns and sequences for mCRC patients have significantly evolved over the past decade [12]. Several studies have reported the complex and changing treatment pattern among mCRC patients in the United States [13- 15,17], Canada [18], and some European countries [19,20].
However, there is a lack of key real-world evidence on the clinical characteristics and treatment patterns of CRC/mCRC patients in China.
Thus, this study was designed to describe baseline characteristics of CRC and mCRC patients, to investigate prognosis in CRC patients, as well as to understand treatment patterns and sequences in mCRC patients using a multi-center oncology database [21]. Findings from this study can be used as evidence to inform clinical management of CRC and mCRC patients.

Data Source
This study analyzed data drawn from a multi-center oncology database, which gathered information from electronic medical records (EMRs) of multiple tertiary hospitals in China. With a large volume of patient-level data of patient sample and a wide geographic coverage, this database provides a platform for conducting a retrospective database study among CRC and mCRC patients. A total of 12 tertiary hospitals from eight provinces were selected across China. Data between January 1, 2011 and September 30, 2016 were extracted, including information on patient baseline characteristics and detailed diagnosis and treatment-related information during each inpatient visit.

Study Population
Patients with primary diagnosis dates after January 1, 2011 and aged 18 years old and above at primary diagnosis were included. Primary diagnosis dates (baseline) were defined as the first clinical or pathological diagnosis dates recorded in the selected hospital's EMRs, whichever occurred earlier.
If patients were previously diagnosed with CRC outside the selected hospitals, the initial diagnosis dates, if available, were used as the primary diagnosis dates. As a previous phase III clinical trial suggested that the maximum period of the 3 rd line treatment usage was nine months [22], in this study, patients who started third-line chemotherapy after January 1, 2016 were excluded to allow a minimum nine-month observation period for the entire third-line chemotherapy to be documented.
Patients with mCRC, as a subgroup of CRC patients, included those who were classified as TNM stage IV [23] at primary diagnosis, and those whose tumor metastasized before the database lock (September 30, 2016).

Study Variables
Patient demographic and clinical characteristics at primary diagnosis were analyzed for the CRC and the mCRC subgroup. As the first appearance of metastases was not recorded in EMRs, metastases were defined by the adoption of palliative chemotherapy. Recurrence rates were assessed at 1-year, 2-year and 3-year post-index dates and compared by TNM stage and primary tumor site. Disease-free survival was defined as the time between the primary diagnosis date and the first documentation of recurrent of local or regional tumor, or deaths or the last record date in the EMR database, whichever occurred earlier. Treatment patterns of first-, second-and third-line palliative chemotherapy were also analyzed, with regimens and cycles reported. Chemotherapeutic lines were determined by physicians and recorded in EMRs.

Statistics
For continuous variables, mean, standard deviation (SD), median, interquartile range (IQR), minimum (min) and maximum (max) values were presented as appropriate; for categorical variables, number of missing values, frequency distribution and percentage were presented. Missing data were not included in percentage calculations. Chi-square tests were used to compare categorical variables and Kruskal-Wallis tests for continuous variables. Recurrence rates were compared by TNM stage and primary tumor site using Fisher's exact tests instead of Chi-square tests due to a smaller sample size.
Cumulative probabilities of disease-free survival by TNM stage were estimated using the Kaplan-Meier method, and comparisons were performed with the log-rank test. All statistical analyses were performed using STATA version 14.0 (Stata Corporation, Texas, USA). Two-sided tests with a significance level of 0.05 were applied.

Patient Flow
Data of 8,246 CRC patients identified from the selected hospitals during the study period were

Baseline Characteristics for CRC Patients
The mean age of CRC patients was 59 (SD 13) years old and 60% of them were males. Most of the patients (87%) presented in internal medicine departments. The size of primary tumor was only available among 36% (2,926/8,136) patients with a mean of 4.48 centimeters (SD 2.05). The KRAS mutation status testing rate was only 25%. The majority of CRC patients had left-sided primary tumor sites (71%) and were physically well (98%) at baseline with an Eastern Cooperative Oncology Group Performance Status (ECOG PS) scoring of 0 or 1 (Table 1). Of the 8,136 CRC patients, 6,764 (83%) had TNM classification records at primary diagnosis. Among them, 7.2%, 19%, 30% and 44% were at stages I, II, III and IV, respectively. Liver was the most common metastatic site (52%), followed by lung (27%) (Figure 2).

Disease Recurrent Risks for CRC Patients with Radical Surgeries
For patients who underwent radical surgeries, the cumulative recurrence rate at year 1, 2, and 3 was 8.9%, 16% and 30%, respectively. By different baseline TNM stage, an upward trend could be seen from baseline stage I to stage III for all 1-, 2-and 3-year recurrence rates (p<0.001, Table S1).
Recurrence rates did not differ significantly between left-and right-sided primary tumor sites (Table   S2). The log-rank test showed that differences across TNM stage-specific, disease-free survival curves were of statistical significance (p<0.001, Figure 3).

Baseline Characteristics for mCRC Patients
The mean age of mCRC patients at the primary diagnosis date were 57 years old (SD 12). There were more males in mCRC patients (62%) and most (92%) presented in internal medicine departments. The mean size of the primary tumor was 4.52 centimeters (SD 2.09). The KRAS mutation status testing rate was 35%. The majority of CRC patients had left-sided primary tumor sites (69%) and were physically well (98%) at baseline (Table 1).  Figure 4D).

Treatment Sequences for mCRC Patients with Palliative Chemotherapy
In total, 1,200 patients had records on both first-and second-line of treatments, of whom 404 patients had records on third-line of treatments. In this study, the mean age of CRC and mCRC patients diagnosed during 2011 to 2016 was 59 and 57 years old, respectively. For both CRC and mCRC patients, there were more male CRC patients than females. These findings are not surprising compared to previous published reviews and regional studies in China [7][8][9]. It has been reported that estrogen could prevent CRC [24], which may explain the relatively small proportion of female patients.
In this study, approximately 30% of patients presented with TNM stage III and about 40% presented with TNM stage IV at diagnosis. The percentage of patients diagnosed with TNM stage III is generally in line with previous findings which reported a range of 30%-40%; while, the percentage of patients identified with stage IV is larger than that reported in previous studies which is ranged 20%-33% [25,26]. The inconsistence on the percentage of patients diagnosed with stage IV across studies may be due to the differences in study samples, e.g. the current study included higher proportion of patients aged 60 years and over than other studies. It is also possible that the early detection of colorectal cancer in China is not as prevalent as in other countries. A decrease in the proportion of patients with advanced stage CRC may be expected if an early detection program can be implemented [25].
Most patients in this study did not have their KRAS mutation status tested. Although KRAS mutation analysis may provide additional useful information on risk stratification in colorectal cancer, the predictive value of KRAS mutation for non-response to chemotherapy is still questioning [24]. Previous studies found that the expression of KRAS is associated with recurrence, survival and benefit of adjuvant chemotherapy [27]. Future studies on the predictive value of KRAs in Chinese CRC patients are required.
As expected, the proportion of right-sided CRC (22%) found in this study was higher than that reported in the 1990s (15%) and 1980s (11%) [11]. A similar rightward shift in the primary tumor site of CRC has also been reported in North America, the United Kingdom, Japan and Northern Ireland [28][29][30][31][32], and it is associated with aging [33]. Our study has also found that among patients after radical surgeries, the primary tumor site was not associated with disease-free survival. This is supported by the finding from a most recent study which included 4,426 Chinese patients with stage I, II and III CRC [34]. Patient at earlier stages had a significant improvement in survival. This finding underlines the importance of early diagnosis and increased awareness of CRC in China [35,36].
This study had several limitations. Although we had a wide geographic coverage, all selected hospitals were tertiary hospitals located in large cities. In the three-tier healthcare system of China, patients with severe diseases, such as cancer, would mostly be referred to tertiary hospitals in large cities, but nevertheless those who cannot afford such treatments may present elsewhere. Therefore, sampling bias may still present a barrier to understanding the complete picture of CRC and mCRC patients in China. In addition, if patients visited hospitals outside the selected ones during the observation period, those records would not have been captured and therefore could not be analyzed in this study. Moreover, we utilized a multi-center database specialized in oncology, however the study population was a sample of the whole CRC patient population seen for care in the 12 hospitals.
There is no evidence that the sampling process and data collection may introduce significant selection bias. Some key time points, such as the appearance of metastasis, were not recorded in EMRs and were estimated using best proxies. Finally, some key information might not be available or

Ethics Approval and consent to participate
De-identification of personal information was performed at the data entry stage in the multi-center oncology database. Ethics approvals were obtained for this study from Beijing Cancer Hospital Ethics Committee, Sun Yat-sen University Cancer Center Ethics Committee and Xijing Hospital Ethics Committee. For hospitals that did not require ethics approval as a mandatory, a form of 'Confidential and non-use Agreement' was assigned. The informed consent was waived by the IRBs from all hospitals.

Consent for publication
Not applicable

Availability of data and materials
The data underlying the study is from EMR databases in multiple tertiary hospitals in China. It is not publicly available, and restrictions apply to the availability of the data. Disease-free survival curves from radical surgeries by TNM stage at diagnosis Treatment patterns of palliative chemotherapy for mCRC patients Figure 5 Treatment sequences of palliative chemotherapy for mCRC patients

Supplementary Files
This is a list of supplementary files associated with this preprint. Click to download.