Association between Oral Chinese Herbal Medicine and Recurrence and Metastasis in Patients with Stages II and III Colorectal Cancer: A Cohort Study in China

Background To evaluate the associations between long-term oral Chinese herbal medicines (CHMs) and recurrence and metastasis (R&M) in patients with stage II and III colorectal cancer (CRC). Furthermore, we aimed to determine the correlation between different syndrome patterns and prognosis and summarized the regularities among CHMs prescriptions, providing reference for clinical practice. Methods An ambispective cohort study was conducted. All CRC patients who sought evaluation and treatment at Xiyuan Hospital and Beijing Cancer Hospital from August 2014 to August 2016 were included. In this study, “whether patients voluntarily take CHMs” was taken as the exposure factor, and the exposure degree was “the duration of CHM use.” Stratification was performed according to the duration of TCM use to determine the relationship with R&M of CRC. The primary outcome was disease-free survival. Patients who had R&M of CRC after taking CHMs for ≥6 months were defined as “worst patients.” R software was used for statistical analysis. The Kaplan–Meier method and Cox regression analysis were used to determine the prognosis. IBM SPSS was used to model a priori association rules; drug use rules were analyzed on this basis. Results A total of 186 patients with stage II and III CRC after radical resection were enrolled. All patients reached the study endpoint by August 2021. The difference in disease-free survival between the two groups was most significant when the cutoff value for CHMs was 18 months (P = 0.0012). Multivariate analysis showed that 18 CHMs were independent protective factors for R&M of CRC (P = 0.001, HR = 0.20, 95% CI = 0.08–0.53). The ratio of Pi (spleen) and Shen (kidney) deficiency in the worst cases was higher than patients without R&M (P = 0.018). Sijunzi and Liuwei Dihuang decoctions were the most frequently used prescriptions in the anti-R&M phase. Conclusion CHMs complying with the “Jianpi Bushen” principle may attenuate the risk of R&M in patients with stage II and III CRC. Pi (spleen) and Shen (kidney) deficiency in patients receiving TCM intervention for the first time within 6 months of radical resection may be associated with a higher CRC R&M rate. Further research is warranted to validate these findings and elucidate underlying biological mechanisms.


Introduction
Colorectal cancer (CRC) is the third most common cancer and the second leading cause of cancer-related deaths [1]. Despite advances in treatment and implementation of screening and surveillance [2], CRC remains a major global health problem with high morbidity and mortality [3].
Patients with nonmetastatic CRC are advised to undergo standard radical surgery with or without adjuvant chemotherapy; however, recurrence and metastasis (R&M) occurs in 30%-50% of such patients receiving optimal treatment, and <20% of patients with advanced CRC survive 5 years after diagnosis [4,5]. Furthermore, despite standard treatment, many CRC patients still have physical symptoms, such as fatigue, insomnia, and fatulence, as well as psychological distress [6], which negatively impact patients' well-being [7]. Tus, prolonging the disease-free survival (DFS) and improving the quality of life are worthy goals for patients with nonmetastatic CRC.
Traditional Chinese medicine (TCM), as a long-standing science and culture, plays an important role in complementary and alternative medicine therapies in China, where TCM is widely accepted by patients and covered by Chinese health insurance. Because of numerous properties, TCM has an increasingly prominent role in various diseases, including CRC [8]. A large cancer center conducted parallel surveys and reported that 83% of patients used TCM, with oral Chinese herbal medicines (CHMs) accounting for 55.8% of the total [9]. CHMs have gained increasing acceptance in the clinic for their proven efcacy in CRC treatment [10]. Our previous study demonstrated that CHMs are associated with a lower R&M rate [11]. We then conducted a multicenter prospective cohort study and reported that using CHMs for 1 year improved survival outcomes in patients with stage II and III CRC [12]. According to the TCM theory, cancer is caused by an imbalance between endogenous physical conditions of the body and exogenous pathogenic factors [12]. Shen (kidney) is the origin of congenital constitution, while Pi (spleen) is the root of acquired constitution. A healthy qi defciency in TCM is mainly manifested by a decline in the autoimmune function caused by innate defciency or acquired hypotrophy, which cannot monitor and annihilate tumor cells. As a result, there is spread and difusion of circulating tumor cells and tumor dormant cells in the body, which eventually leads to tumor progression [13,14]. Patients with nonmetastatic CRC can have tumors excised through conventional Western medicine practices, but the patients may have spleen defciency and residual disease. Tus, several previous studies have opined that CRC patients can be treated with the "Jianpi Bushen" rule [15,16].
TCM emphasizes individuality and places a high value on the holistic view. Syndrome diferentiation refers to diagnosing an illness as a specifc syndrome based on an analysis of the specifc symptoms and physical signs collected through inspection, auscultation and olfaction, inquiry, and palpation, while treatment refers to defning the treatment approach in line with the syndrome diferentiated. Tis view coincides with the principle of precision medicine in modern medicine. But it is difcult to determine a common prescription that is efective. We performed a two-center ambispective cohort study with a maximum follow-up of fve years to provide clinical evidence to show the advantages of TCM "Jianpi Bushen" theory against R&M in patients with stage II and III CRC. In addition, we used the concept of "target population" in enrichment design [17]. Patients using CHMs for more than fve years without R&M were defned as the TCM "target population" in this study. Te core drug groups with a common therapeutic background were extracted using data mining techniques to explore the regularity of CHM prescriptions in the TCM "target population." Tis was achieved to reveal the similarities and diferences in CHMs for CRC and provide a basis for the rational use of CHMs in clinical practice.

Study Population.
Patients in this ambispective cohort were admitted to the Xiyuan Hospital of China academy of Chinese medical sciences and Beijing Cancer Hospital, Beijing, China between August 2014 and August 2016. Te inclusion criteria were the following: (1) patients between the ages of 18-80 years, with no gender restrictions; (2) the primary, and only, tumor site was CRC, and pathologically diagnosed as adenocarcinoma (pathological wax and sections could be obtained); (3) TNM stages were stage II and III; (4) patients who received standard treatments within 6 months after radical resection (R0 resection), according to the National Comprehensive Cancer Network (NCCN) guidelines [18]. Te exclusion criteria were the following: (1) patients with intestinal obstruction, who cannot take decoction and need intravenous nutrition; (2) patients who are allergic to the known Chinese medicine ingredients used in the study; (3) pregnant and lactating women with mental illness; and (4) those who participated in other clinical studies.

Study Design.
Based on the cumulative use of CHMs for 6 months, we divided patients into an integrated Chinese and Western medicine or Western medicine group [19]. When two consecutive CHM treatments were >6 weeks apart, CHMs were considered discontinued [20]. Patients using CHMs for >5 years without CRC R&M were defned as the TCM "target population." When stratifed according to CHM use, the "worst cases" were defned as patients who had R&M after taking TCM for ≥6 months. Patients without CRC R&M within 5 years were compared with the worst cases. Te regularity of CHM prescriptions in the TCM "target population" was summarized. TCM syndrome pattern criteria were in accordance with the clinical research guidelines for TCM New Drugs [21]. Tere are four types of TCM syndrome patterns: Pi (spleen) Qi defciency; Shen (kidney) Yin defciency; Pi (spleen) and Shen (kidney) defciency; and non-Pi (spleen) and Shen (kidney) defciency.

Follow-Up Strategy and Outcomes.
Patients received follow-up according to NCCN clinical guidelines, with clinical visits every three months for the frst two years, every six months between years 3 and 5, and then yearly after year 5. We recorded TCM syndrome patterns and detailed prescription information of TCM. All physicians participating in the study received training in advance. Patients who occasionally miss clinical visits are followed up by telephone by trained researchers. If the telephone follow-up failed three times, it was defned as lost to follow-up. DFS, defned as the time from intervention to tumor R&M or death from any cause, was considered as the primary end point.

Statistical
Analysis. Statistical data were described using the median for continuous variables and percentages (%) for categorical variables. Diferences in the baseline characteristics were grouped by quartiles (Q) of cumulative months of CHM use (Q1, Q2, Q3, and Q4) and assessed using one-way analysis of variance, Kruskal-Wallis H test, χ 2 , or Fisher's tests to determine statistical diferences between the groups. Kaplan-Meier survival analysis was performed using the surv-cutpoint function in the "survminer" R package, which determines the optimal cut point that yielded the minimum P value in the log-rank test between the high-andlow-risk subgroups [22]. Cox proportional hazard ratio (HR) with 95% confdence intervals (CIs) was used to determine the simultaneous impact of other variables potentially associated with each outcome. A P value <0.05 (two-sided) was considered statistically signifcant. Statistical analyses were analyzed using R software (version 3.6.2). Age was a continuous variable, which was converted to a categorical variable using X-tile (version 3.6.1, Yale University) [23]. Regularity of CHM prescriptions through data mining was analyzed by the "Apriori block," which was provided by SPSS 18.0 software for cluster and network analyses [24]. Te study design is illustrated in Figure 1.

Baseline Characteristics.
A total of 200 patients from two hospitals with stage II and III CRC were enrolled in this cohort study between August 2014 and August 2016. Fourteen patients were further excluded from the analysis as they did not conform to the study protocol. A total of 186 patients were enrolled, including 89 integrated Chinese and Western medicine patients and 97 Western medicine patients. 17 patients dropped out during follow-up and 12 patients did not receive regular surveillance in the hospital during the last two years due to COVID-19. Tree patients died due to nontumor causes. Te fowchart of patient recruitment for this study is shown in Figure 2. Te optimum cutof score for age was determined to be 60 years. Te baseline characteristics of participants based on quartiles of cumulative days of CHM use are summarized in Table 1.

Impact of CHM Treatment on Clinical Outcomes. On
August 27, 2021, we collected the fnal data. Te median DFS follow-up is 65.0 (2.0-89.0) months. During the follow-up period, 39 of the 186 patients in this study experienced cancer R&M: the number of patients with R&M was 10 (25.6%), 19 (48.7%), 3 (7.7%), 4 (10.3%), and 3 (7.7%) per year from 1 to 5 years after radical resection, respectively ( Figure 3). Among 39 patients with CRC R&M, there were 21 with rectal cancer and 18 with colon cancer. Tere were eight patients with stage II CRC and 31 patients with stage III CRC ( Figure 4). Kaplan-Meier analysis demonstrated that the integrated Chinese and Western medicine group had a lower incidence of CRC R&M than the Western medicine group (P � 0.078; Figure 5(a)). Using a cutof point of 18 months and the use of CHMs determined by the "survminer" package, patients were divided into low and high exposure groups. Kaplan-Meier curves illustrated that patients with low exposure had a signifcantly worse DFS than those with ≥18 months CHMs (P � 0.0012; Figure 5(b)).

Multivariate Analyses.
We examined the infuence of CHMs on DFS across strata of other factors, including TNM stage, age, sex, family history of CRC, tumor site, and carcinoembryonic antigen (CEA). Te use of CHMs for 18 months was the only factor associated with a signifcant improvement in DFS (HR, 0.23; 95% CI, 0.09-0.58; P � 0.002). No other statistically signifcant interactions were detected. A forest plot of multivariable analyses in these subgroups is shown in Figure 6.    Table 2.

Data Mining of Prescription in the TCM "Target
Population". A total of 13 patients were in the TCM "target population," namely, patients using CHMs for more than fve years without R&M.   Te analysis of association rules identifed 20 highfrequency medicine pair rules and the complicated network for those are visualized in Figure 8. With the cluster analysis, we generated a dendrogram from which three clustering items were extracted with a relative distance of 7.5, among which two kinds of prescription contained more than fve CHMs (Figure 9). Te extracted prescriptions corresponded to three famous Chinese formulae: Sijunzi decoction, Liuwei Dihuang decoction, and Erzhi wan.
In Figure 9 Cluster analysis dendrogram of Chinese herbal medicines in the "target population". Te Y-axis

Discussion
In this ambispective cohort study, 186 patients with stage II and III CRC were followed up for fve years. Furthermore, we found a positive relationship between CHMs and the risk of R&M in stage II and III CRC. We further observed that CHMs intervention for 18 months (as the cutof value) was associated with the lowest risk of clinical outcome. Moreover, these associations were independent of other disease factors. In this study by comparing the TCM syndrome patterns of patients without CRC R&M within 5 years with the worst cases, it was found that the syndrome patterns of Pi (spleen) and Shen (kidney) defciency were more common in the worst cases. Due to the limited number of patients, the exact relationship between TCM syndrome type and CRC R&M warrants further study by expanding the sample size. Our fndings on the distribution of syndromes in CRC patients were similar to a study based on 760 CRC patients from two cancer centers in China [26]. Tis is the frst study to summarize the prescription regularity of CHMs by patients to prevent CRC R&M, which is an important preference for TCM clinical practice. TCM theory holds that "Te region where pathogenic factors invade must be defcient of Qi". Pi (spleen) qi defciency is a pathologic change characterized by qi defciency combined with impaired transporting and transforming function of the spleen [27]. Patients' gastrointestinal function needs to be restored after radical surgery, and chemotherapy-induced nausea and vomiting are common side efects. Dizziness, weariness, indigestion, abdominal distension, lassitude, anorexia, diarrhea, and other Pi (spleen) qi-defcient disorders are common in CRC patients. Shen (kidney) yin defciency is a pathologic change characterized by a defciency of yin to nourish the kidney, leading to defciencyfre or defciency-heat, which is characterized by dizziness, forgetfulness, tinnitus, backache, and lack of libido, among other symptoms [27]. In addition, because cancer is a chronic wasting disease, CRC patients are predisposed to manifest Shen (kidney) yin defciency syndrome. A total of 180 patients with curative rectal cancer operations were administered a questionnaire, most of whom experienced sexual dysfunction [28].
In TCM prescriptions, herbs are generally used in combination as "formulas". Based on classical prescriptions, adding, subtracting, or changing medicine can achieve the therapeutic efect for diferent symptoms. In this study, we used data mining to reveal the underlying CHMs prescription regularity in patients with stage II and III CRC. We found that three classical and well-known prescriptions: Sijunzi decoction, Liuwei Dihuang decoction, and Erzhi wan were the most frequently used. Te frst two prescriptions were frst recorded in the Song dynasty, while Erzhi wan was recorded in the Ming dynasty. Sijunzi decoction is the basic     [30]. A study based on network pharmacology investigation revealed that the mechanism of Sijunzi decoction and Liuwei Dihuang decoction refers to the CRC pathway [31]. Te drug pair, Ecliptae Herba and Ligustri Lucidi fructus, which is also a traditional Chinese prescription called Erzhi wan, has been widely used to replenish the liver and kidney, nourish yin, and stop bleeding. Te modern pharmacological study of Erzhi on CRC needs further exploration.

Evidence-Based Complementary and Alternative Medicine
Several limitations to this study should also be considered. Our results may have been confounded by factors that were not captured in our clinical study, such as molecular genetics, physical activity, and dietary habits [32]. During the long-term CHM treatment, it is likely that patients were not fully compliant with prescriptions. Furthermore, due to the COVID-19 pandemic, several patients were unable to follow-up as scheduled, resulting in the clinical outcome being missed in evaluation. All the limitations mentioned  Evidence-Based Complementary and Alternative Medicine above would have tended to overestimate the true associations between CHMs and R&M. Moreover, due to inadequate ambispective studies, patient information may be incomplete and data collection and follow-up may be biased.

Conclusions
Te results of this ambispective cohort study suggested that CHMs may be associated with a lower risk of R&M in patients with stage II and III CRC. Tese fndings are consistent with our previous studies. We further found that the efect on prolonging DFS was the greatest when the intervention using CHMs lasted for 18 months in patients.
Our study attempted to further reveal the role of TCM in improving clinical outcomes by observing visceral pattern identifcation and concluding the regularity of CHMs prescriptions through data mining in patients undergoing fve years of treatment for CRC prevention. It was demonstrated that "Jianpi Bushen" was the main method of treatment for patients with stage II and III CRC. Further research is required to confrm the generalizability of our fndings.

Data Availability
Te datasets analyzed during the current study are available from the corresponding authors on reasonable request.

Ethical Approval
Tis study was approved by the Ethics Committee of Xiyuan Hospital, China Academy of Chinese Medical Sciences (2019XLA044-5) and Beijing Cancer Hospital (2019KT104-GZ01) and was performed in accordance with the Declaration of Helsinki.

Consent
Written informed consent was obtained from all individual patients included in the study.