Prevalence of the Use of Herbal Medicines among Patients with Cancer: A Systematic Review and Meta-Analysis

Background Although herbal medicines are used by patients with cancer in multiple oncology care settings, the magnitude of herbal medicine use in this context remains unclear. The purpose of this review was to establish the prevalence of herbal medicine use among patients with cancer, across various geographical settings and patient characteristics (age and gender categories). Methods Electronic databases that were searched for data published, from January 2000 to January 2020, were Medline (PubMed), Google Scholar, Embase, and African Index Medicus. Eligible studies reporting prevalence estimates of herbal medicine use amongst cancer patients were pooled using random-effects meta-analyses. Studies were grouped by World Bank region and income groups. Subgroup and meta-regression analyses were performed to explore source of heterogeneity. Results In total, 155 studies with data for 809,065 participants (53.95% female) met the inclusion criteria. Overall, the pooled prevalence of the use of herbal medicine among patients with cancer was 22% (95% confidence interval (CI): 18%–25%), with the highest prevalence estimates for Africa (40%, 95% CI: 23%–58%) and Asia (28%, 95% CI: 21%–35%). The pooled prevalence estimate was higher across low- and middle-income countries (32%, 95% CI: 23%–42%) and lower across high-income countries (17%, 95% CI: 14%–21%). Higher pooled prevalence estimates were found for adult patients with cancer (22%, 95% CI: 19%–26%) compared with children with cancer (18%, 95% CI: 11%–27%) and for female patients (27%, 95% CI: 19%–35%) compared with males (17%, 95% CI: 1%–47%). Conclusion Herbal medicine is used by a large percentage of patients with cancer use. The findings of this review highlight the need for herbal medicine to be integrated in cancer care.


Introduction
Cancer is a major global health problem. In 2018, there were an estimated 9.6 million cancer fatalities and 18.1 million newly diagnosed cases [1]. Current trends indicate that the previously predicted increase to 20 million new cases by 2025 is likely to be surpassed [2]. e overall implications of this high rate of new cases of cancer include increased health, economic, and social costs, which will continue to put a burden on the limited resources and weak healthcare systems in poor countries. As a result, herbal (traditional) medicines use in cancer care may be increased in those countries.
Several studies have identified and documented the clinical (cancer disease) and individual (demographic) factors associated with the usage of herbal medicine in cancer. Numerous factors have been positively correlated with herbal medicine use in cancer, such as young age, high education level, high-income level, ethnicity, female gender, cancer diagnosis, longer survival period since cancer diagnosis, receiving single or multiple cancer chemotherapies, being married, completion of conventional cancer treatment, having certain cancer symptoms, cancer metastasis, and belonging to specific social groups [22][23][24][25][26][27]. However, old age, being a child, having cancer comorbidities, place of residence, and the experience of chemotherapeutic side effects are negatively associated with herbal medicine use in cancer [16,26,28].
Some herbal medicines possess compounds that are pharmacologically active against cancer cells, and preclinical studies have consistently shown that numerous herbal medicines or herbs have antiapoptotic, anti-inflammatory, cell regenerative, and antioxidant effects on cancer cells. However, the clinical evidence concerning the efficacy of most herbal medicines or specific herbs used in cancer is largely inconclusive [29,30]. Clinical studies have reported that the use of herbal medicines in cancer lowered the mortality rate hazard ratio of patients with lung cancer (thereby increasing survival), improved patients' quality of life through reducing cancer symptoms and conventional drug side effects (e.g., nausea and vomiting), and had chemopreventive activity against certain cancers [17,19,31,32].
Conversely, observational studies have indicated that concomitant herbal medicines use with antineoplastic drugs may result in drug to herb interaction (at all pharmacokinetic and pharmacodynamics levels) and adverse side effects or events [9,31,33,34]. Adverse side effects or events range from minor side effects (e.g., gastrointestinal distress and allergy) to severe organ failure (e.g., hepatotoxicity, nephrotoxicity, bone marrow suppression, and respiratory and cardiac failure) [31,32,[35][36][37]. Importantly, the release of antioxidants by herbal medicines is thought to reduce the oxidizing free radicals created by radiotherapy and chemotherapeutic agents against cancer cells, potentially reducing the effectiveness of conventional cancer treatment [24,38]. Similarly, herbs commonly used in cancer such as St. John's wort and grape juice induce cytochrome isoenzymes (especially CYP3A4), which metabolize most conventional anticancer agents, thereby reducing the efficacy of targeted therapies such as tyrosine kinase inhibitors and anticancer hormonal therapies [31,32,38]. St. John's wort, specifically, was found to reduce the levels of plasma irinotecan, docetaxel, and imatinib mesylate antichemotherapeutic agents' concentrations [31]. Additionally, other herbs commonly used in cancer have been found to cause bleeding tendencies following surgery (e.g., ginkgo, garlic), hypoglycemia (e.g., ginseng), and hepatotoxicity (e.g., kava) and possess carcinogenic or negative tumor moderating effects [4,35,39]. In addition, heavy metal contamination in some herbal medicines may alter the pharmacokinetic profile of commonly used conventional cancer treatments [24,38,40].
Despite the widespread herbal medicines use among cancer patients, associated factors, and potential benefits and risks, the overall pooled prevalence of the use of herbal medicines among patients with cancer remains unclear. Previous systematic reviews focused on cancer used qualitative (narrative synthesis) designs and focused on synthesizing primary data on the use of CAM treatment modalities in general [41][42][43][44][45][46][47]. erefore, we conducted a meta-analysis to explore the prevalence of the use of herbal medicine among patients with cancer across various geographical settings and patient characteristics and synthesized the literature on commonly used herbs in cancer. Understanding the prevalence of herbal medicine use by patients with cancer may help inform and guide healthcare policies geared toward integrating herbal medicines use in cancer care. Ultimately, this will help to improve outcomes for patients with cancer, develop wider public health (community) policies around herbal medicine legislation, and promote investment in education and research about herbal medicines used in cancer.

Eligibility Criteria.
We synthesized hospital-and population-based studies that reported the prevalence of the use of herbal medicines among patients with cancer (Table 1).

Information Sources and Search Strategy.
We searched PubMed, Google Scholar, Embase, and African Index Medicus for articles published from January 2000 to January 2020. e key search terms words that were used to guide the search included the following: "Cancer" OR "Neoplasm" OR "Tumo * " OR "Malignancy" AND "Herbs" OR "Herbal medicine" OR "Herbal material" OR "Herbal preparation" AND "Prevalence" OR "Use" OR "Proportion" OR "Percent * " AND "Observational studies" OR "Cohort" OR "Cross-sectional * " OR "Survey" OR "Cohort" (Supplementary Table 2). In addition, we manually skimmed the references of published review articles and primary studies to obtain any further eligible studies.

Data Extraction: Selection and Coding.
After obtaining relevant studies (A.J.B), two authors (P.P.N and A.M.S) screened the identified articles' abstracts and titles and determined their eligibility for inclusion in this review, independently. Two authors (J.B.A and A.M.S) independently extracted the data using a standard data extraction form or tool created in Microsoft Excel 2016. Before data were extracted, the data extraction tool was pilot tested with 10 studies. Following the findings of the pilot test, improvements to the data extraction tool were made after reaching consensus with all reviewers. e extracted data included the following: (i) methodological or study characteristics, (ii) herbal medicine definitions or terms used, (iii) the focus of study (herbal medicine only or with other CAM modalities), (iii) use of conventional treatment, (iv) gender distribution, (v) participants' average age, (vi) sample size, (vii) proportion or frequency of herbal medicine use, (viii) herbs used in cancer, and (ix) conclusions. Countries were categorized by continent or world regions and according to World Bank economic indicators (Supplementary Table 3). e authors were able to resolve disagreements during data extraction through consensus.

Risk of Bias Assessment.
Using the risk of bias of nonrandomized studies (RoBINS) tool, three authors (J.B.A, A.M.S, and P.P.N) independently assessed and reported the risk of bias in selected studies [48][49][50]. e risk of bias was categorized as high, moderate (unclear), and low across the various categories of bias (participation bias, selection bias, and confounder bias) (Supplementary Tables 4 and 5).

Data Analysis and Synthesis.
e metaprop command in Stata software (version 12) was used to analyze the data. Freeman-Tukey double arcsine transformation was used to "stabilize the raw proportions" [51]. e pooled prevalence estimates and their confidence intervals (CI) were computed using the DerSimonian and Laird random-effects model (DL) and Wald method, respectively, based on "the transformed values and their variances" [51]. We inspected the forest plots for heterogeneity and then quantified this using chi-square tests and the I 2 statistic. Because of significant heterogeneity (>50%), we explored the possible modifying effects of a number of study-level variables on the overall pooled prevalence of herbal medicine use among patients with cancer based on subgroup and meta-regression analyses [52,53]. Modifying variables included the following: (i) year of publication (before or after 2010), (ii) study focus (herbal medicine alone vs. herbal medicine with other CAM), (iii) data collection method (researcher-administered vs. self-administered vs. document or record reviews), (iv) country income level (low and middle vs. high income), (v) study setting (hospital vs. community), (vi) study population (adults vs. children), (vii) cancer type (breast vs. prostate vs. hematological vs. others), (viii) region (continent), (ix) World Bank subregion, (x) study design (crosssectional vs. cohort), and (xi) study country. We evaluated publication bias by inspecting the funnel plot for asymmetry and confirmed it using Egger's regression test [54]. We also reported the weighted pooled prevalence estimates and their 95% CIs. Data related to herbs used by patients with cancer were summarized and described.

Study Selection Flow.
In total, 6414 studies were retrieved from various search engines and databases (Figure 1). After eliminating duplicates, 4882 articles were selected for critical screening of the titles and abstracts. e final meta-analysis included 153 full-text articles that met the inclusion criteria. Eighty-six (86) articles were included in the qualitative synthesis of commonly used herbs by patients with cancer.  ). e most common data collection method was self-administered interviews (52.90%), and the most common cancer type was breast cancer (18.71%).

Overall Subgroup and Meta-Regression Analyses.
We conducted subgroup and meta-regression analyses to investigate the influence of various patient and study characteristics on the overall observed prevalence estimates to explore the heterogeneity observed among the included studies. e subgroup analyses indicated that geographical region (P ≤ 0.001), subregion (P ≤ 0.001), income group (P ≤ 0.001), study focus or type (P ≤ 0.001), study country (P ≤ 0.001), and study design (P ≤ 0.001) had statistically significant moderating effects on the overall pooled prevalence of herbal medicine usage in cancer. Conversely, in the bivariate meta-regression analysis, only three variables were related to the overall pooled prevalence of usage of herbal medicine by cancer patients. Visual inspection of the scatter plot showed that studies that investigated herbal medicine in conjunction with other CAM modalities in cancer (20%, 95% CI: 16-23%) were more than twice less likely to report a higher pooled prevalence than studies that focused on herbal medicine use alone (48%, 95% CI: 35%-61%; β � −1.47, 95% CI: −2.18 to −0.76; P ≤ 0.001; Figure 3(a)). In addition, studies from high-income countries were less likely to report a high pooled prevalence of herbal medicine use in cancer compared to those from low-and middle-income countries (β � −0.803, 95% CI: −1.23 to −0.38; P ≤ 0.001; Figure 3(b)). However, there was a moderate positive relationship between subregion and pooled prevalence of herbal medicine use in cancer, with certain subregions being more likely than others to report a high prevalence of the use of herbal medicine in cancer (β � 0.134, 95% CI: 0.005-0.264; P � 0.043).

Pooled Prevalence of Herbal Medicine Use by Adult
e subgroup analyses indicated that study country (P ≤ 0.001), study focus (P ≤ 0.001), study region (P ≤ 0.001), study design (P ≤ 0.001), income group (P � 0.02), and study subregion (P ≤ 0.001) were statistically significant moderators of the pooled prevalence of herbal medicine usage by adults with cancer. However, bivariate meta-regression revealed that only study focus and income group had negative relationships with the pooled prevalence. Studies that focused on herbal medicine along with other CAM modalities (21%, 95% CI: 15%-27%; β � −1.506, 95% CI: −2.25 to −0.76; P ≤ 0.001) and those from high-income countries (β � −0.75, 95% CI: −1.23 to −0.27; P � 0.002) were less likely to report a high pooled prevalence of herbal medicine use in adults with cancer than studies that focused on herbal medicine alone (49%, 95% CI: 34%-64%) and were from low-and middle-income countries.
e subgroup analysis showed that the study focus (P ≤ 0.001) and study region (P ≤ 0.001) had a statistically significant moderating effect on the overall pooled prevalence of the use of herbal medicine among female patients with cancer. However, in the meta-regression analysis, only study focus had a weak negative relationship with the pooled prevalence of herbal medicine use in cancer. Studies that focused on herbal medicine along with other CAM modalities (25%, 95% CI: 17%-33%; β � −1.634, 95% CI: −3.49-0.220; P � 0.082; Figure 7) were less likely to have a high pooled prevalence than studies that focused on herbal medicine use in cancer alone (63%, 95% CI: 55%-71%).      reported by patients with cancer are listed in Supplementary  Table 6.

Discussion
It has become increasingly common to base healthcare decision-making on information obtained from evidencebased medicine. Previously, this information was obtained from systematic reviews of interventional studies. However, this evidence is now also being acquired from systematic reviews of observational studies. e present review investigated the prevalence of the herbal medicine use among patients with cancer to inform and guide the development of healthcare policies concerning integrating herbal medicine in clinical cancer care. is review suggested that a large estimated percentage of cancer patients use herbal medicine, especially during conventional treatment.
e overall pooled prevalence of herbal medicine usage by patients with cancer was 22% (95% CI: 18%-25%), which means approximately one in five patients with cancer used herbal medicine(s) following a cancer diagnosis. is finding was consistent with the literature, where herbal medicine was reported as the leading form of CAM used in cancer [4-6, 8, 11, 12, 114]. is review also found that Africa and Asia had the highest pooled prevalence of the usage herbal medicine in cancer, with the lowest prevalence recorded in Oceania. Similarly, a larger percentage of patients with cancer from low-and middleincome countries used herbal medicine compared with those from high-income countries. is trend was repeated across specific subpopulations of children, adults, and female patients with cancer. e variation in prevalence across regions may be explained by variances in geographical characteristics (i.e., conditions that make some herbs easily available), cultural beliefs and attitudes, and liberalized or low regulation of herbal medicines [18-20, 31, 79, 131, 139, 162]. Conversely, the high herbal medicine usage in low-and middle-income countries might possibly be because of the low income levels, which may mean that patients with cancer are unable to pay for conventional cancer care (financial constraints) and or due to deeply rooted cultural practices related or favorable to use of herbal medicines. For example, as shown in this study, Asian countries such as South Korea and Taiwan, despite having the conditions and economic power to receive high-quality conventional therapies, patients from these countries still continue to use herbal medicine while accepting conventional therapies. Above all, high-income countries (of North America and Europe) where most studies included in this review were conducted do not possess a specialized or deeply ingrained traditional medicine use culture compared to countries of Asia and Africa.
e findings of this review suggested that compared to children (with cancer), adult patients with cancer were more likely to use herbal medicine, although this difference was not statistically significant. is variance in prevalence may be related to adults having more freedom to use herbal medicine than children who generally depend on their parents to access such products [106]. However, given the physiological nature of children (i.e., immaturity of organs such as the liver) and the potential risks of herbal medicine for children as reported in previous studies, parents need to make informed decisions based on evidence-based information before administering herbal medicine to their children to protect children from possible harmful effects [37].
We also found that more female patients with cancer compared to their male counterparts used herbal medicine. However, the pooled prevalence of the use of herbal medicine by male patients with cancer was from studies conducted among patients with prostate cancer; therefore, that reported prevalence best represents herbal medicine usage among patients with prostate cancer. Similarly, most studies that focused on female patients with cancer included patients with breast cancer, although the use of herbal medicine was higher in breast cancer than prostate cancer. ese gender-based findings concur with previous literature, where the use of herbal medicine in cancer was related to being female, with women more likely to use herbal medicine as a primary mode of healthcare than men [22,23,138].
Finally, this review revealed several herbs commonly used in cancer, some with proven evidence of beneficial effects (anticancer effects) and others with potential risks (harmful side effects and drug interactions) to patients. ose with possible detrimental effects to patients included garlic, ginseng, kava, and St. John's wort [31,32,35]. Given that most of the studies that reported use of those herbs were conducted among cancer patients who were receiving conventional cancer therapies, clinicians (oncologists) should ask about herbal medicine use during their routine care of such patients.

Implications and Recommendations.
Regardless of variation in the level of herbal medicine regulatory frameworks in different countries across the world, the high percentage of the usage of herbal medicine reported by this study calls for some form of integration of herbal medicine into cancer care. Healthcare providers must be at the center of this integration. e lack of sufficient clinical evidence should not be a deterrent to this integration, although health practitioners at all levels of patient care should routinely ask about, offer, and document evidence-based advice to patients with cancer on the safety and possible benefits of herbs and herb-drug interactions. Routine discussion of these issues during cancer screening, treatment, and follow-up may help to improve patient care outcomes. However, to equip health workers with evolving evidence on herbal medicines used in cancer, health educators need to continue incorporating knowledge about herbal medicines in oncological care training curricula and also develop programs geared toward understanding, evaluating, and validating herbal medicine use in cancer. In the short-term, health managers could develop short courses or refresher training for in-service healthcare workers on herbal medicines used in cancer to improve their knowledge on this subject.
In addition, policy makers at national governmental and international levels, such as drug authorities and health Evidence-Based Complementary and Alternative Medicine ministries, should incorporate and update new evidence regarding herbal medicine into oncology treatment guidelines, standard operating procedures, patient charts or electronic medical records, and pharmacopeias.
is will assist healthcare workers to document herbal medicine practices in clinical care, which will subsequently promote clinical research on herbal medicine use in cancer.
As evidence regarding herbal medicine continues to evolve, policy makers in countries that regulate herbal medicine as dietary supplements or do not regulate herbal medicine at all need to update, review, or change their herbal medicine regulatory frameworks (either entirely or on a case-by-case basis) to protect patients with cancer from possible harmful effects posed by some herbs. In addition, as the media and other informal sources of information on herbal medicine are responsible for the high use of herbal medicine by cancer patients, oncology care centers and policy makers could create official websites or other media platforms with authentic and updated information on commonly or locally available herbal medicines to counter the misinformation from other sources. ese platforms may be communicated to patients during routine cancer care. Importantly, these platforms should encourage patients to always seek advice regarding their specific circumstances from a qualified healthcare professional.
Successful integration of herbal medicine into cancer care either as an alternative form of medicine or alongside cancer medicine requires further high-quality multidisciplinary research on herbal medicines used in cancer, which requires research funding. erefore, policy makers need to advocate, fundraise, and allocate resources for cancer research concerning herbal medicine use. For example, the lack of funding for research on herbal medicines may explain the relatively few published studies on herbal medicine in cancer, especially in Africa and South America, as observed in this study. In addition, it is important to note that in this study, we were unable to conduct a meta-analysis of the usage of herbal medicine in cancer across many other patients' characteristics due to inconsistent study variables. erefore, it is necessary to develop a standardized survey tool customizable to various patients with cancer and settings to measure herbal medicine use in cancer. Such a tool will allow comprehensive systematic reviews to be conducted on this subject. It is also necessary to conduct more herbal medicine-specific observational research in cancer to obtain extensive statistics on the extent of herbal medicine usage in cancer across the world.

Limitations and Strengths of is Study.
In terms of the quality of included studies, we rated the majority of studies as having a moderate to high risk of bias, which might have led to over-or underestimation of the pooled prevalence of herbal medicine use reported in this study. is was expected given that nearly all studies used cross-sectional designs. However, the majority of the included studies had moderately large sample sizes and high response rates. Second, there was a high proportion of heterogeneity (between-study heterogeneity) associated with the estimated pooled prevalence(s) reported in this study. However, this was minimized through estimating the pooled prevalence using a random-effects model and performing extensive subgroup and meta-regression analyses. ird, we acknowledge the limited number of studies from sub-Saharan Africa, and the pooled estimate from the few available studies might be overestimated. However, this provides an opportunity for further research on usage of herbal medicine among patients with cancer in Africa. Fourth, we only included studies published in the English language (thus missing studies published in other languages, particularly from the Francophone or Portuguese speaking countries) and did not include grey literature (dissertations or conference abstracts), which might have affected the outcomes (pooled prevalence rates) reported. Nonetheless, based on examination of the funnel plots and use of Egger's test of funnel plot asymmetry, no evidence of small-study effects (publication bias) was observed (found) across a sample of primary studies included in this study; therefore, the results of this review are unlikely to reflect bias. e majority of the included primary studies were prospective, and sufficient time was invested in these studies, making their results somewhat reliable. Finally, this study provides a strong point of reference for future studies, as it is one of the first reviews to be conducted on the prevalence of the use of herbal medicine amongst cancer patients.

Conclusion
is systematic review shows that a large percentage of patients with cancer use herbal medicine, especially those from low-and middle-income countries. In addition, larger percentages of adult patients with cancer (compared with children) and female patients with cancer (compared with males) used herbal medicine. In summary, there is need for additional epidemiological investigations exploring herbal medicine integration into cancer care especially for low-and middle-income countries.

Data Availability
e data used to support the findings of this study are included within the article and will be available online after publication of the article.

Supplementary Materials
Supplementary data to this article are submitted together with the manuscript and are available online. (Supplementary Materials)