Predictive value of TCM tongue characteristics for chemotherapy-induced myelosuppression in patients with lung cancer

Abstract

This study aimed to investigate the clinical predictors, including traditional Chinese medicine tongue characteristics and other clinical parameters for chemotherapy-induced myelosuppression (CIM), and then to develop a clinical prediction model and construct a nomogram. A total of 103 patients with lung cancer were prospectively enrolled in this study. All of them were scheduled to receive first-line chemotherapy regimens. Participants were randomly assigned to either the training group (n = 52) or the test group (n = 51). Tongue characteristics and clinical parameters were collected before the start of chemotherapy, and then the incidence of myelosuppression was assessed after treatment. We used univariate logistic regression analysis to identify the risk predictors for assessing the incidence of CIM. Moreover, we developed a predictive model and a nomogram using multivariate logistic regression analysis. Finally, we evaluated the predictive performance of the model by examining the area under the curve value of the receiver operating characteristic, calibration curve, and decision curve analysis. As a result, a total of 3 independent predictors were found to be associated with the CIM in multivariate regression analysis: the fat tongue (OR = 3.67), Karnofsky performance status score (OR = 0.11), and the number of high-toxic drugs in chemotherapy regimens (OR = 4.78). Then a model was constructed using these 3 predictors and it exhibited a robust predictive performance with an area under the curve of 0.82 and the consistent calibration curves. Besides, the decision curve analysis results suggested that applying this predictive model can result in more net clinical benefit for patients. We established a traditional Chinese medicine prediction model based on the tongue characteristics and clinical parameters, which could serve as a useful tool for assessing the risk of CIM.

1. Introduction

Lung cancer is the second most common malignancy, with 2,206,771 new cases and 1,796,144 new deaths being reported globally in 2020 alone.^[1] Lung cancer usually has no obvious symptoms in its early stages, so most patients are already at an advanced stage when they are diagnosed, and the opportunity for radical treatment is already missed, which makes the overall 5-year survival rate of advanced lung cancer still low. With the continuous expansion of molecular biology, lung cancer has entered an era of blossoming, with various types of target drugs emerging, and immunotherapy has become a hot topic nowadays. However, chemotherapeutic drugs still have an unshakable position in the treatment of lung cancer due to their broad spectrum of anticancer, target-independent effects, and precise efficacy, thus most lung cancer patients still need to receive standardized chemotherapy.

Though cytotoxic chemotherapy exhibits good effectiveness in cancer treatment and is widely used in clinical practice, it also can lead to considerable side reactions that seriously impact the health and life quality of patients. Myelosuppression is one of the most common adverse reactions during chemotherapy, accompanied by clinical manifestations such as anemia, thrombocytopenia, and neutropenia.^[2–4] It is a dose-limiting toxic effect, and this complication can be alleviated by reducing the dose and frequency of drug use, but this will reduce the efficacy of chemotherapy and even harm patient prognosis.^[5] Thus, it is necessary to identify risk predictors for the development of myelosuppression and to implement appropriate interventions to reduce its incidence before the initiation of chemotherapy.

Tongue diagnosis is one of the important components of traditional Chinese medicine (TCM) diagnosis, which plays an important role in clinical practice by observing different tongue features to determine the internal health condition of the human body.^[6] The tongue features have been reported previously in both acute and chronic diseases such as COVID-19, cancer, diabetes, and kidney diseases,^[7–12] simultaneously, it was used for disease prediction as well.^[13–15] However, the relationship between tongue features and chemotherapy-induced myelosuppression (CIM) has been neglected, and to the best of our knowledge, there were no study has yet performed on that. The change of tongue color and shape reflected a large quantity of physiological and pathological information. Given that, our study aimed to explore the relationship between CIM and tongue features, and then propose an innovative, noninvasive prediction approach based on tongue features and clinical characteristics, which may provide potential value for screening high-risk patients and preventing CIM.

2. Materials and methods

2.1. Patient population and data collection

This is a prospective study from June 2019 to March 2023. Data on patients with lung cancer who will receive chemotherapy were collected at the China-Japan Friendship Hospital. The study was approved by the Institutional Ethics Committee of the China-Japan Friendship Hospital. Inform consent was obtained from all patients for their data to be used for research. Patients did not receive financial compensation.

The inclusion criteria were primary lung cancer diagnosed by histopathology; about to undergo the first chemotherapy and the treatment time exceeding 1 chemotherapy cycle; older than 18 years; no previous use of any drugs to prevent myelosuppression; the Karnofsky Performance Status (KPS) score is not <60 and the expected survival time is more than 3 months; the patient has no contraindications to chemotherapy. The patients were excluded if they had had a blood system disease or mental and psychological illness. Besides, patients accepting immunotherapy or targeted therapy are excluded as well.

Demographic and clinicopathological data were collected, including age, sex, KPS score, body mass index (BMI), history of smoking and drinking, tumor node metastasis stage, tumor pathological type, bone metastasis, and the number of high-toxic drugs in chemotherapy regimens (NHDCR). The toxicity of chemotherapeutic drugs is graded into high and low by myelotoxicity.^[16–18] Myelosuppression after chemotherapy is more likely due to high-toxic drugs, so we only focused on the number of high-toxic drugs. The detailed information on chemotherapy regimens involved in the current study is shown in Table S1, Supplemental Digital Content, http://links.lww.com/MD/M83.

Patients are categorized according to the presence or absence of myelosuppression within 30 days after the first chemotherapy session. The criteria for myelosuppression are as follows: white blood cells (WBC) < 4 × 10⁹/L, neutrophils < 2 × 10⁹/L, platelets (PLT) < 100 × 10⁹/L, hemoglobin (HGB) < 110 g/L. The diagnosis of myelosuppression was considered when 1 of these 4 indicators occurred, if more than 1 indicator occurred, the most serious one was selected.^[19]

2.2. Tongue image collection and tongue feature analysis

The tongue images were collected by applying the DS01-B tongue diagnosis instrument (Daosh Co., Shanghai, China) (Figure S1, Supplemental Digital Content, http://links.lww.com/MD/M82). This instrument is equipped with a high-definition camera, which can capture tongue images. With a stable light source to simulate the natural light, a confined space can be created between the instrument and the subject, which offers a stable shooting environment for collecting tongue images. A computer was equipped in the DS01-B as well, in which the image processing software was used to photograph the images of the tongue of all the participants. To avoid the influence of factors such as eating or drinking. The participants were required to rinse their mouths before the tongue images were collected and all tongue images were photographed from 7 to 8 am before breakfast.

After the tongue image collection was finished, 2 experienced TCM experts independently analyzed the tongue features including tongue shape (fat, median, lean, crack, and tooth-marked) (Figure S2A–E, Supplemental Digital Content, http://links.lww.com/MD/M85); tongue color (whitish, reddish, redness, and purple) (Figure S2F–I, Supplemental Digital Content, http://links.lww.com/MD/M85); tongue-coating thickness (thick, thin, and no fur) (Figure S2J–L, Supplemental Digital Content, http://links.lww.com/MD/M85); tongue-coating color (whitish, yellowish, and black) (Figure S2M–O, Supplemental Digital Content, http://links.lww.com/MD/M85). Conflicts on tongue image analysis were resolved by a consensus with the third experts.

2.3. Statistical analysis

R 3.6.3 was used for all statistical analysis. The continuous variables are presented as mean ± SD and medians with interquartile ranges (IQRs) for normal and nonnormal data, respectively. Categorical variables are presented as n (%). Participants were grouped into training and validation cohorts randomly by R “Caret” package.

Patients were randomly divided into training and test sets according to 1:1. Univariate logistic regression analysis was performed to explore the potential risk factors associated with myelosuppression. Then the variables with P < .05 aforementioned were selected into a multivariate logistic regression analysis to identify the independent risk factors and to build a nomogram model for predicting the CIM in the training set, then the model was validated in the test set. The receiver operating characteristic (ROC) curve and calibration curve were utilized to estimate the predictive performance of this risk model. Finally, the model was assessed by decision curve analysis (DCA) to detect whether using this model could yield closer associations with clinical net benefit than single independent risk factors. In addition, we divided all patients into subgroups according to clinical characteristics such as clinical stage, chemotherapy regimens, age, gender, etc. We then performed stratified logistic regression analysis to study whether the nomogram model maintained predictive power in different cohorts.

3. Results

3.1. Demographic and clinical characteristics

A total of 103 patients were included in the current study and the characteristics of the patients in the training and test sets were presented in Table 1. For the entire cohort, 75.7% (78 of 103) were males and 24.3% (25 of 103) were females. As shown in Table S2, Supplemental Digital Content, http://links.lww.com/MD/M84, the average age was 63.54 ± 8.51 years old; the average BMI was 24.04 ± 3.6 kg/m². The median KPS score was 80 (IQR, 70–90). Nearly half of the patients had a smoking history (53.4%, 55 of 103), and few of them had a drinking history (35.9%, 37 of 103). In terms of the primary site of the tumor, 42.7% (44 of 103) was in the left lung while 57.3% (59 of 103) was in the right lung. In total, 45.6% of all tumors were squamous cell carcinoma, 25.2% of them were adenocarcinoma and 29.2% of them were small cell carcinoma. Most tumors were advanced, with 24.3% (25 of 103) being stage III tumors and 58.3% (60 of 103) being stage IV tumors. Bone metastasis was observed in 25.2% (26 of 103) of the patients. In total, 34% (35 of 103) of the patients had 2 high-toxic drugs included in their chemotherapeutic regimen, and 49.5% (51 of 103) of the patients had 1 high-toxic drug while 16.5% (17 of 103) had no high-toxic drugs in their chemotherapeutic regimen.

Table 1.

Patient characteristics for the training and test set.

Characteristics	Training set (n = 52) No. (%)	Test set (n = 51) No. (%)	χ 2	P value
Gender
Male	41 (78.8%)	37 (72.5%)	0.555	.456
Female	11 (21.2%)	14 (27.5%)
Age
<60	14 (26.9%)	14 (27.5%)	0.004	.952
≥60	38 (73.1%)	37 (72.5%)
KPS score
≤70	20 (38.5%)	21 (41.2%)	0.079	.778
>70	32 (61.5%)	30 (58.8%)
BMI (x ± s)	24.10 ± 3.61	23.97 ± 3.59	0.064	.826
Smoking
Yes	26 (50.0%)	22 (43.1%)	0.487	.485
No	26 (50.0%)	29 (56.9%)
Drinking
Yes	34 (65.4%)	32 (62.7%)	0.078	.780
No	18 (34.6%)	19 (37.3%)
Pathology type
Squamous cell carcinoma	12 (23.1%)	14 (27.5%)	1.720	.423
Adenocarcinoma	27 (51.9%)	20 (39.2%)
Small cell carcinoma	13 (25.0%)	17 (33.3%)
TNM stage
I–II	11 (21.2%)	7 (13.7%)	0.985	.321
III–IV	41 (78.8%)	44 (86.3%)
Tumor site
Left	22 (42.3%)	22 (43.1%)	0.007	.932
Right	30 (57.7%)	29 (56.9%)
Bone metastasis
Yes	12 (23.1%)	14 (27.5%)	0.261	.609
No	40 (76.9%)	37 (71.5%)
NHDCR
0	10 (19.2%)	7 (13.7%)	5.566	.062
1	30 (57.7%)	21 (41.2%)
2	12 (23.1%)	23 (45.1%)

BMI = body mass index, KPS = Karnofsky performance status, NHDCR = number of high-toxic drugs in chemotherapy regimens, TNM = tumor node metastasis.

3.2. The distribution of tongue features

Among the 103 patients, the most frequent tongue shape was the median tongue (43.7%) and the most common tongue color was reddish (40.8%). As for the tongue coating, the most common color of tongue coating was yellowish (50.5%) and the most frequent thickness for the tongue coating was thick (60.2%). The black tongue coating was rare during clinical diagnosis and therefore not found in the current study (Table 2).

Table 2.

The tongue characteristics of patients.

Characteristics	n (%)
Tongue color
Whitish	29 (28.2%)
Reddish	42 (40.8%)
Redness	23 (22.3%)
Purple	9 (8.7%)
Tongue shape
Median	45 (43.7%)
Fat	30 (29.1%)
Lean	6 (5.8%)
Tooth-marked	14 (13.6%)
Crack	20 (19.4%)
Tongue-coating color
Whitish	51 (49.5%)
Yellowish	52 (50.5%)
Dye	0
Tongue-coating thickness
Thin	35 (34.0%)
Thick	62 (60.2%)
Desquamative	6 (5.8%)
No tongue coating	0

3.3. Model specifications and predictors of CIM

Tongue characteristics, as well as other clinical parameters, were selected as candidate variables for the prediction model. The univariate logistic regression analysis identified that fat tongue was associated with CIM risk (Table 3). Similarly, the univariate logistic regression analysis identified the following 2 clinical variables that had an association with CIM risk: KPS score, and NHDCR (Table 4). On multivariable analysis, fat tongue (OR, 3.67; 95% CI, 1.29–10.43; P = .015), KPS score (OR, 0.11; 95% CI, 0.04–0.31; P < .001), and NHDCR was 2 (OR, 4.78; 95% CI, 1.06–21.51; P = .041) were each independently associated with CIM risk (Table 5).

Table 3.

Univariate logistic regression analysis of tongue characteristics.

Characteristics	Feature	OR	95% CI	P value
Tongue color
Whitish	No	1	Reference	—
	Yes	1.63	0.68–3.87	.274
Reddish	No	1	Reference	—
	Yes	1.46	0.66–3.28	.351
Redness	No	1	Reference	—
	Yes	0.59	0.21–1.55	.300
Purple	No	1	Reference	—
	Yes	0.17	0.01–0.97	.099
Tongue shape
Fat	No	1	Reference	—
	Yes	3.26	1.37–8.05	.009
Lean	No	1	Reference	—
	Yes	0.74	0.13–4.26	.739
Tooth-marked	No	1	Reference	—
	Yes	3.21	0.99–10.39	.052
Crack	No	1	Reference	—
	Yes	1.01	0.36–2.71	.984
Tongue-coating color
Whitish	No	1	Reference	—
	Yes	1.83	0.83–4.12	.138
Yellowish	No	1	Reference	—
	Yes	0.55	0.24–1.21	.138
Tongue-coating thickness
Thin	No	1	Reference	—
	Yes	0.58	0.24–1.35	.215
Thick	No	1	Reference	—
	Yes	1.77	0.79–4.13	.174
Desquamative	No	1	Reference	—
	Yes	0.74	0.10–4.00	.739

The “—” indicates that the value is not available; CI = confidence interval, OR = odds ratio.

Table 4.

Univariate logistic regression analysis of clinical characteristics.

Characteristics	OR	95% CI	P value
Gender
Female	1	Reference	—
Male	1.26	0.51–3.13	.623
Age
<60	1	Reference	—
≥60	1.57	0.64–4.07	.334
Smoking
No	1	Reference	—
Yes	1.17	0.97–1.42	.099
Drinking
No	1	Reference	—
Yes	1.20	0.98–1.46	.074
BMI
<18.5	1	Reference	—
18.5–24	1.36	0.23–8.17	.739
≥24	1.33	0.22–7.98	.753
KPS score
≤70	1	Reference	—
>70	0.65	0.54–0.77	.001
Pathological stage
I–II	1	Reference	—
III–IV	1.40	0.48–4.09	.538
Pathology
Squamous cell carcinoma	1	Reference	—
Adenocarcinoma	0.50	0.18–1.34	.165
Small cell carcinoma	1.17	0.41–3.34	.774
Tumor site
Left	1	Reference	—
Right	0.66	0.30–1.47	.313
Bone metastasis
No	1	Reference	—
Yes	1.42	0.58-3.48	.455
NHDCR
0	1	Reference	—
1	1.77	0.50–6.25	.373
2	3.86	1.05–14.21	.042

The “—” indicates that the value is not available; BMI = body mass index, CI = confidence interval, KPS = Karnofsky performance status, NHDCR = number of high-toxic drugs in chemotherapy regimens, OR = odds ratio.

Table 5.

Multivariate logistic regression analysis of the potential risk factors.

Characteristics	OR	95% CI	P value
Fat tongue
No	1	Reference	—
Yes	3.67	1.29–10.43	.015
KPS score
≤70	1	Reference	—
>70	0.11	0.04–0.31	<.001
NHDCR
0	1	Reference	—
1	1.63	0.40–6.73	.500
2	4.78	1.06–21.51	.041

The “—” indicates that the value is not available; CI = confidence interval, KPS = Karnofsky performance status, NHDCR = number of high-toxic drugs in chemotherapy regimens, OR = odds ratio.

3.4. Model performance and nomogram

In the training set, the CIM risk prediction model was developed utilizing the aforementioned 3 factors (fat tongue, KPS score, and NHDCR) (Fig. 1). The model includes 3 independent factors, namely, fat tongue (Yes or No), KPS score (≤70 or >70), and NHDCR (0, 1, or 3), followed by developing a nomogram to illustrate the model. The higher total points summed by the points for each factor in the nomogram were associated with a higher probability of CIM. For instance, a patient with a fat tongue, a KPS score exceeding 70, and a NHDCR were 2 would have approximately 138 points (60 for a fat tongue, 0 for a KPS score, and 78 for NHDCR). For this patient, the predicted probability of occurrence for CIM was about 55%.

Figure 1.

The nomogram for predicting the CIM risk in patients with lung cancer. CIM = chemotherapy-induced myelosuppression.

To further assess the discriminative ability of the model, the area under curve (AUC) value of the nomogram model was 0.82 (95% CI: 0.71–0.93) which was higher than all individual factors, demonstrating a good predictive ability (Fig. 2A). The calibration curve indicated that the actual and predicted probability of CIM had a good consistency (Fig. 2B). The practical clinical value of the nomogram was assessed using decision curve analysis (DCA) as well. The results showed that the application of the nomogram will bring more net benefit with the threshold probability between 0 and 0.55 (Fig. 2C). Similarly, in the test set, the AUC of the model was 0.82 (95% CI: 0.71–0.94), and the calibration curve and DCA curve showed a good performance as well (Fig. 2D–F).

Figure 2.

Comparison and validation of the nomogram. (A) ROC curves of the nomogram, fat tongue, KPS score, and NHDCR in the training set. (B) Calibration curves for nomogram in the training set. (C) DCA for the nomogram, fat tongue, KPS score, and NHDCR in the training set. (D) ROC curves of the nomogram, fat tongue, KPS score, and NHDCR in the test set. (E) Calibration curves for nomogram in the test set. (F) DCA for the nomogram, fat tongue, KPS score, and NHDCR in the test set. DCA = decision curve analysis, KPS = Karnofsky performance status, NHDCR = number of high-toxic drugs in chemotherapy regimens, ROC = receiver operating characteristic.

Furthermore, we conducted stratified analysis for the nomogram model based on clinical characteristics such as clinical stage, chemotherapy regimens, age, and gender to explore the model’s robustness. The study results indicate that patients with a high probability predicted by the model are at a greater risk of developing CIM, which demonstrates the model’s robustness (Table 6). Particularly, the nomogram model had poor predictive performance in subgroups tumor node metastasis stage (I + II), gemcitabine + platinum, and irinotecan + platinum, which may be related to the small sample size in this subgroup.

Table 6.

Univariate logistic regression analysis of the nomogram model in stratified cohort.

Characteristics	OR	95% CI	P value
Gender
Male	11.67	3.93–34.67	<.001
Female	8.1	1.23–53.20	.029
Age
<60	29.75	3.47–255.02	.002
≥60	7.43	2.65–20.85	<.001
Smoking
Yes	20.31	3.76–109.89	<.001
No	9.07	2.58–31.89	<.001
Drinking
Yes	9.07	2.80–29.32	<.001
No	14	2.81–69.76	.001
TNM stage
I–II	25,636	0–Inf	.996
III–IV	7.74	2.91–20.58	<.001
Chemotherapy regimen
Pemetrexed + platinum	6.38	1.35–30.14	.019
Paclitaxel + platinum	22.5	2.60–194.51	.005
Etoposide + platinum	25	1.80–346.70	.016
Gemcitabine + platinum	21,272	0–Inf	.996
Irinotecan + platinum	2.33	0.11–50.99	.590

CI = confidence interval, Inf = infinite, OR = odds ratio, TNM = tumor node metastasis.

4. Discussion

Myelosuppression characterized by leukopenia, neutropenia, anemia, and thrombocytopenia, is a common complication following chemotherapy. Severe myelosuppression can result in treatment delay, and some can even be life-threatening, which adds to the burden borne by patients with cancer and their families. Thus, exploring the risk factors of CIM before treatment begins is of great significance for carrying out some preventive interventions to reduce the incidence of myelosuppression. In the current prospective cohort study, we investigated the risk factors of myelosuppression in patients with lung cancer to guide the clinical treatment and management.

To the best of our knowledge, this is the first attempt to apply TCM tongue diagnosis to the survey of myelosuppression in patients with lung cancer. In this study, by establishing the prediction model, we identified 3 risk factors including fat tongue, KPS score, and NHDCR which were associated with the occurrence of myelosuppression.

Tongue diagnosis plays a pivotal role in TCM for thousands of years. It is a unique TCM method used to discriminate physiological functions and pathological conditions by observing the changes in the tongue features. The tongue features like tongue shape, color, and tongue coating can reflect the internal health status of the body (e.g., Organs, Qi, Blood, Cold, Heat), thus it can be a good option for disease diagnosis.^[20] The previous study has reported the connection between tongue features and diseases. For instance, Hsu et al^[21] revealed the particular tongue features of patients with breast cancer, such as lean tongue shape, whitish tongue color, thick tongue coating, yellow tongue coating, and cracked tongue shape, which could enhance the role of TCM in integrated medical care for patients. Lee et al^[22] demonstrated that people with metabolic syndrome have white tongue coating, which represents a metabolic disorder. Also, tongue features were widely used for disease diagnosis. Zhang et al^[23] developed a diagnostic method for diabetes by utilizing tongue features, it has good predictive accuracy. Song et al^[10] found that compared with normal people, patients with early esophageal cancer possess special tongue color, so they constructed a diagnostic model which can be utilized to screen for early esophageal cancer owing to its convenient and noninvasive nature.

TCM theory holds that myelosuppression is associated with the deficiency of Qi and Blood, and the chemotherapeutic drugs can debilitate the spleen, the source of Qi and Blood. In this study, we revealed that fat tongue was one of the risk factors for myelosuppression. According to the TCM theory, tongue features can reflect the inside situation of the body, so the fat tongue is connected with spleen weakness, which indirectly indicates that patients with fat tongue have an obstacle in generating Qi and Blood, leading to a higher risk of myelosuppression.

Performance status (PS) is a reliable indicator of the patient’s general condition and quality of life (Qol). KPS is one of the most commonly used scales to assess a patient’s functional status as well as Qol, and it is an important factor for predicting prognosis as well as determining the choice of treatment or the intensity of palliative treatment in cancer patients.^[24] The higher KPS score, the better quality of life for patients and the more suitable for taking in chemotherapy. The current study indicated that the lower KPS score, the more possible occurrence of myelosuppression after chemotherapy. A previous study pointed out that patients with lower baseline Qol scores have a higher risk of adverse drug reactions.^[25] Due to the side effects, chemotherapy also can impair the Qol and then decline the KPS score.^[26] Moreover, some studies revealed that the Qol score (KPS or ECOG) can predict the risk of myelosuppression, with a lower Qol score indicating a higher possibility of CIM.^[27,28] Therefore, before the beginning of chemotherapeutic treatment, the functional status should not be neglected exceptionally for these patients with low KPS scores.

Generally, clinical experience suggests that the higher the toxicity and the greater the number of chemotherapeutic drugs, the higher the likelihood of myelosuppression. The correlation between a single drug or chemotherapy regimen and the occurrence of myelosuppression has been reported previously, however, the actual chemotherapy regimens used in clinical practice are complex and varied, so the correlation between most chemotherapy drugs or regimens and the occurrence of myelosuppression remains largely unknown. In this study, chemotherapeutic drugs were classified according to toxicity classes, and the results showed that utilizing of exceeding 2 high-toxic drugs was a risk factor for myelosuppression, suggesting that when more chemotherapy regimens are available, the use of regimens with less than 2 highly toxic drug classes may reduce the occurrence of myelosuppression, and this more individualized treatment approach may bring greater benefits to patients in terms of both treatment and prevention of toxic effects. This more individualized approach may provide greater clinical benefit to patients, both in terms of treatment and prevention of toxicities.

Furthermore, to enable the model to achieve a more reliable and valuable prediction efficacy in clinical settings, the conjoint nomogram was developed for assessing individual CIM risk with satisfactory discrimination. The calibration curve and DCA curve demonstrated that the nomogram had good potential to be applied in clinical practice.

In conclusion, fat tongue, KPS score, and toxicity of chemotherapeutic regimens were found to be good predictors of occurring myelosuppression. Therefore, targeted preventive interventions such as low-toxicity chemotherapy regimens, combined with herbal conditioning and endurance training should be adopted for high-risk patients to reduce the occurrence of myelosuppression. The patients involved in the current study were only admitted by our hospital, and the number of patients was small. Thus, a large cohort from the multicenter might be needed to verify our results in the future. Further research is also needed to get a better understanding of how to choose a personalized treatment regimen according to the patient’s physical status and tongue characteristics.

Author contributions

Conceptualization: Jian-Rong Sun, Li-Qun Jia.

Data curation: Jian-Rong Sun, Kai-Xuan Li.

Investigation: Yan-Ni Lou, Rong Huang.

Methodology: Rong Huang.

Project administration: Li-Qun Jia.

Software: Jian-Rong Sun, Yan-Ni Lou.

Visualization: Jian-Rong Sun.

Writing—original draft: Jian-Rong Sun.

Supplementary Material