A randomized, double-blind, placebo controlled, multicenter trial was conducted at 8 medical centers in China, with a maintenance chemotherapy plus CHMF group and a maintenance chemotherapy plus placebo group. Participants were enrolled from medical centers with the following eligibility criteria: patients were pathologically or cytologically confirmed of stage IIIa-IV NSCLC (International Union Against Cancer, version 7); efficacy evaluation of the first-line chemotherapy is progression-free, including complete response (CR), partial response (PR) and stable disease (SD); patients were at the age of 18–75 years old; patients had a KPS score of ≥70; patients had an estimated life expectancy of at least 12 weeks; participants had no major organ dysfunction, and had plan for chemotherapy maintenance; and informed consent was obtained from the patients. Patients were not eligible if they had malignant tumors except for NSCLC 5 years prior to study entry, received targeted therapy or other anticancer treatment, or were allergic to chemotherapy drugs.

The whole project was funded by Special Scientific Research for Traditional Chinese Medicine (China). Study of maintenance therapy including chemotherapy, Tyrosine Kinase Inhibitor (TKI) combined with CHMF or CHMF only were conducted synchronously. As one of subprojects, this study protocol was approved by the institutional review boards (IRBs) of the participating medical centers, the independent National Ethics Committee (EC), the Chinese Medicinal Agency, and was registered under the identifier NCT02900742 at the Clinicaltrials.Gov website. After the expire date, patients were followed up by Longhua Hospital affiliated to Shanghai University of TCM until September 21, 2017 which still reviewed by IRB. All participants provided written informed consent before enrolment.

No valid data on survival was available as few studies were designed to evaluate the efficacy of TCM combined with maintenance chemotherapy. The sample size of this study was based on the validity of assumptions and clinical experience of the past studies (Pérol et al., 2012; Wang et al., 2013; Yifen et al., 2014). A two-sided log-rank test with an overall sample size of 114 subjects (57 in the control group and 57 in the treatment group) achieved 80% power at a 0.05 significance level when the median survival time was 6 and 3 months respectively in the treatment group and the control group. The study lasted for 36 months, with subject accrual (entry) occurring in the first time period. The drop-out rate was 10% in the control group.

Eligible patients were randomly assigned 1:1 to receive maintenance chemotherapy with CHMF or with placebo. Central randomization was performed by a clinical research organization (CRO) (Shanghai Clinical Research Center, Shanghai, China http://www.scrcnet.org). Random numbers were automatically generated by computer referring to pre-configured stratified factors (smoking status (smoker, ex-smoker or non-smoker), gender (female, male), age (>70, ≤ 70) and chemotherapeutics (pemetrexed, gemcitabine, and docetaxel) before administration of maintenance chemotherapy. Then the CRO provided the random number to each participating center. The block size and treatment-assignment table were not available to the investigators and patients until the end of the study.

Patients received continuation maintenance chemotherapy (gemcitabine 1,250 mg/m², ivgtt 30 min on days 1 and 8, once every 21 days or pemetrexed 500 mg/m², ivgtt 30 min on day 1, once every 21 days or docetaxel 75 mg/m2, ivgtt 30 min on day 1, once every 21 days) until evidence of disease progression or unacceptable toxicity. Patients in maintenance chemotherapy+CHMF or maintenance chemotherapy+ placebo group received treatment with CHMF or matched placebo respectively. CHMF and placebo granules were administered on the first day after chemotherapy. The medicine was dissolved in 160 mL of warm water for drinking at 9:30 a.m. and 15:00 p.m. every day until the end of chemotherapy or unacceptable toxicity. All procedures were performed under the supervision of clinical research pharmacists. The processed herbs and the granules were prepared in compliance with Good Manufacturing Practice (GMP).

Three formulas (batch number: 1208304, 1208303, and 1207357) were dynamic cycle extracted, then concentrated, spray-dried to make granules by Jiangyin Tian Jiang Pharmaceutical Co. Ltd. (Jiangsu, China). Besides, free of bacteria, mold and Escherichia coli were also monitored by this company before ready to use. The water and alcohol extracts of three formulations were analyzed by high performance liquid chromatography (HPLC) for quality control. Methanol from Yuwang (China) was used for HPLC analysis and LC-MS. Acetonitrile of HPLC grade from J&K (Beijing, China) was used for purity analysis by UPLC. Preparative grade n-hexane, isopropanol, ethyl acetate and methanol were purchased from Shilian (Anhui, China). All preparative reagents were filtered through a 0.22 mm organic membrane. The water used in this study was purified with a Milli-Q water purification system (Millipore, Bedford, MA, USA). Purity analysis of compounds was performed on a Waters ACQUITY UPLC® H-Class system with a PDA detector. The standards (1.00 mg) were prepared with 1 mL methanol, then filter by Nylon membrane filter (0.22 μm). The solution was further analysis by HPLC. As shown in Figure 1, the bioactive components were extracted including Rosmarinic acid (C18H16O8; CAS: 20283-92-5), Gallic acid (C7H6O5; CAS: 149-91-7) and Paris I (C44H70O16; CAS: 50773-41-6) in basic formula, Icariin (C33H40O15; CAS: 489-32-7) Psoralen (C11H6O3; CAS: 66-97-7) Calycosin-7-O-beta-D-glucoside (C22H22O10; CAS: 20633-67-4) and Trigonelline hydrochloride (C7H8ClNO2; CAS: 6138-41-6) in YiQi formula, and Trigonelline hydrochloride (C7H8ClNO2; CAS: 6138-41-6) in YangYin formula.

In conformity with Good Clinical Practice (GCP), in double-blind clinical trials, the investigational drug and placebo should be identical in appearance (color, viscosity, hardness, and other physical properties), smell, packaging, labeling, and other characteristics (Xu et al., 2012; Jiao et al., 2017). It is difficult to define a placebo in Chinese herbal medicine where all-natural substances are potentially therapeutic. Therefore, only food color and artificial flavors included in the raw materials for the placebo. Ingredient including:

Yangyin formula (placebo): Lactose (94.06%): citric yellow pigment (0.15%): sunset yellow pigment (0.09%): caramel pigment (4.1%): citric acid (0.8%): bitterness (0.8%)

Yiqi formula (placebo): Lactose (94.96%): citric yellow pigment (0.2%): sunset yellow pigment (0.04%): caramel pigment (3.3%): bitterness (1.5%)

Basic herbs (placebo): Lactose (92.37%): citric yellow pigment (0.1%): sunset yellow pigment (0.03%): caramel pigment (6.3%): bitterness (1.2%).

The CHMFs were prescribed based on syndrome differentiation. Syndrome differentiation criteria were based on “The Guiding Principles of Clinical Research of New Chinese Medicine (trial)” (China Pharmaceutical Technology Publishing House, 2002) and “Shanghai diagnosis and treatment routine of TCM Syndrome”(Shanghai Municipal Health Bureau edit). Three syndromes of TCM are set as follows: Qi deficiency syndrome is composed of the following symptoms: cough with phlegm, poor appetite, lassitude and weakness, pale and plump tongue; secondary symptoms of spontaneous sweating, loose stool or soft slippery pulse.

Yin deficiency syndrome contains main symptoms of cough with scanty phlegm, dry mouth, reddish tongue; secondary symptoms of night sweating, heartburn and insomnia, low fever, thread and rapid pulse.

Qi and Yin deficiency syndrome contains main symptoms of cough with scanty phlegm, shortness of breath, lassitude and weakness, thirst without the desire to drink; secondary symptoms of spontaneous sweating, night sweating, reddish tongue or tongue with teeth marks, thread and weak pulse. The diagnosis can be given with at least two of the main symptoms and one of the secondary symptoms. Each formula was prescribed by one professor who had worked in our hospital for more than 50 years to facilitate clinical research. Patients who diagnosed with Qi syndrome deficiency, Yin syndrome deficiency received treatment with basic herbs adding YiQi formula and YangYin formula respectively. Patients with both Qi and Yin syndrome deficiency prescribed with basic herbs adding the combination of YiQi and YangYin formula granules. In placebo group, patients have the corresponding placebo granules. The composition of the formulas was shown in Table 1.

Follow-up was carried out at 2, 4, and 6 months to assess the clinical response and tolerance. Tumor imaging assessments (MRI or CT) were performed at screening, at the end of C2, and then once every two cycles until the end-of-treatment visit. Objective tumor response was measured using RECIST version 1.1 and assessed by the investigator and by a blinded independent review committee (IRC) and Blood chemistry, hematology, concomitant medications, and adverse events were assessed on day 1 of each 21-day cycle. Patient-reported outcomes were assessed at day 1 (baseline), and every 2 cycles, on day 1 of the subsequent cycles, at the end-of-treatment visit, and at the post-treatment follow-up visit, using the lung cancer symptom scale (LCSS). Patients who had progressive disease continue to be assessed every 3 months.

The primary endpoint of the study was progression-free survival (PFS), defined as the time interval between the date of enrolment and the date of disease progression or death by any cause. The secondary endpoints were LCSS, Karnofsky Performance Status (KPS) score and AEs of the regimen. Adverse events including toxicity and side effects were reported according to Common Terminology Criteria for Adverse Events V3.0 (CTC AE) issued by National Cancer Institute (NCI) (https://ctep.cancer.gov/protocolDevelopment/electronic_applications/docs/ctcaev3.pdf). All of the unexpected responses possibly or definitely related to the research were reported. In case of serious adverse events (SAEs), experimental treatment was stopped immediately and appropriate treatment was provided. The type and frequency of adverse events were reported for each group.

Statistical analysis was performed using the SPSS software (version 18). For variable data (baseline), a chi-square test was used. Descriptive statistics were used to summarize patient characteristics by treatment group. Patients that had received at least one of the intended therapies were included in efficacy analysis. PFS was compared between the maintenance chemotherapy plus CHMF group and the maintenance chemotherapy plus placebo group using the log-rank test. Analysis of PFS was performed in all patients who had at least one post-baseline imaging assessment (modified intention-to-treat population). The KPS score “improvement and stability rates” after treatment, which was analyzed using a chi-square test, was categorized into improved (decrease in score ≥ 10), stable (score not changed) or worse (increase in score ≥ 10). Ordered hierarchical data (NCI-CTC graded AEs) were analyzed using a rank sum test. LCSS analysis was conducted on the QoL population, defined as the population of patients with a completed LCSS assessment (one question from the LCSS) at baseline and at least once during the study. The time to worsening of symptoms (TWS) was measured from the date of randomization to the date of a first clinically meaningful worsening for each of the LCSS items. Worsening of symptoms was defined as a change of one-half standard deviation, determined from the scores of corresponding baseline items (Boye et al., 2016). TWS was censored at the date of the last LCSS assessment for patients with unknown LCSS status or lost to follow-up and was analyzed using the Kaplan-Meier estimator. HRs were estimated by unadjusted Cox regression analysis, with assigned treatment as the only covariate. The mean maximum improvement from the baseline score was calculated for each of the LCSS items. Results were considered statistically significant if P ≤ 0.05.

Between July 22, 2013, and April 7, 2016, 71 patients were finally enrolled. As TKIs and CHMF alone as maintenance therapy after first line treatment were popular, it was hard to enroll patients who chose to receive chemotherapy. In April 2016, the interim analysis was conducted by the Data and Safety Monitoring Board which composed of biostatisticians from a clinical research organization (CRO) and clinicians. The result showed a significance improvement of PFS in maintenance chemotherapy plus CHMF compared with maintenance chemotherapy plus placebo [HR (95%) = 0.50 (0.28–0.92), P = 0.025]. So, the enrollment was ended before reaching the target number. The Data and Safety Monitoring Board kept the results confidential, and follow-up was conducted by our investigators until September 21, 2017. Of 71 patients, 35 were assigned to receive maintenance chemotherapy with CHMF and 36 to receive maintenance chemotherapy with placebo. During face-to-face interviews, the study protocol and objective were explained to patients in the hospital. Three patients from the placebo group and one from the CHMF group refused chemotherapy immediately after randomization. Three patients in the placebo group withdrew for personal reasons. One patient in the CHMF group became ineligible for the study. These patients were excluded from analysis. At the time of data cutoff (September 21, 2017), 6% (2/33) of patients in the CHMF group but no patients in the placebo group were still receiving treatment. Sixty-one patients discontinued the study, including 31 assigned to receive maintenance chemotherapies plus CHMF and 30 assigned to receive maintenance chemotherapy plus placebo (Figure 2). The most common reason for study discontinuation was disease progression, occurring in 84.8% (28/33) of patients who were assigned to receive maintenance chemotherapy plus CHMF and in 80% (24/30) of patients who were assigned to receive maintenance chemotherapy plus placebo. In the CHMF population lost to follow-up, 1 patient switched to targeted therapy after two treatment cycles and 2 patients refused treatment before progression. In the placebo population lost to follow-up, 1 patient have a serious adverse event after one treatment cycle, 4 patients refused treatment before progression, and 1 patient died before progression. These patients were included in the intention-to-treat efficacy analysis as shown in Figure 2. Baseline characteristics and stratification factors were well balanced in the two groups as shown in Table 2.

The treatment cycles were summarized in Figure 3A. In the overall population, 45.45% of patients in the CHMF group and 20.0% of patients in the placebo group remained on study treatment for at least 6 cycles. In addition, at the end of cycle 10, 27.21% of patients receiving maintenance chemotherapy with CHMF remained on study treatment, compared to 10.0% of those receiving maintenance chemotherapy with placebo. In the overall population, no patient in the two groups experienced dose delays due to AEs.

After a median follow-up of 4.18 months (IQR 2.95~9.00), the 3-, 6-, and 12-month progression-free survival was 78.8 vs. 71.5%, 51.5 vs. 27.9%, and 21.2 vs. 8% in the CHMF and placebo groups. Median PFS (Figure 3B) from randomization was 6.43 months (3.27–11.80) in the CHMF group and 3.73 months (2.07–4.60) in the placebo group (HR 0.55, 95% CI 0.28–0.88, P = 0.019).

The KPS scores “improvement and stability rates” were 93.3% in the CHMF group vs. 70% in the placebo group (P = 0.047, 95% CI 0.03–0.935).

In the QoL population as shown in Figure 4, more improvements were observed in the LCSS scores from baseline in the CHMF group than in the placebo group. In the CHMF group, after each of the 3 cycles of chemotherapy, median decrease in score from baseline was seen for fatigue (HR = 0.28, 0.07–0.88, P = 0.03) and interference with daily activities (HR = 0.40, 0.11–0.93, P = 0.04), indicating improvement. Improvement was also seen in the dyspnea score (HR = 0.38, 0.11–0.87, P = 0.03) after 2 cycles of chemotherapy. No change or worsening in the LCSS scores was seen for all of the other symptoms, including overall symptoms, pain, and overall QoL.

Among the 63 patients that received maintenance chemotherapy, one patient refused chemotherapy for drug nephrotic syndrome. The most common grade 1 or 2 toxicities occurring in >15% of patients (Table 3) were fatigue, loss of appetite, dry mouth, leucopenia, nausea, vomiting and neutropenia. After two cycles of treatment (Figure 5), the most common grade 1 or 2 toxicities were asthenia or fatigue (17 of 29 [58.7%] patients in the CHMF group vs. 14 of 20 [70%] patients in the placebo group, P = 0.42), loss of appetite (7 of 29 [24.1%] patients in the CHMF group vs. 12 of 20 [60%] patients in the placebo group, P = 0.011) and dry mouth (5 [17.2%] vs. 5 [25.0%]). One grade 3 serious adverse events (SAEs) were reported in the CHMF group and three in the placebo group. However, after four treatment cycles, the incidence of symptoms increased as follows: asthenia or fatigue (13 of 20 [65.0%] patients in the CHMF group vs5 of 7 [71.4%] patients in the placebo group, P = 0.76), loss of appetite (4 of 20 [20.0%] patients in the CHMF group vs. 6 of 7 [85.7%] patients in the placebo group, P = 0.004), and dry mouth (1 [5.0%] vs. 4 [57.1%], P = 0.011). However, three grade-3 SAEs were reported in the placebo group.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.