The BUTCH trial was a double-blind, placebo-controlled, randomized clinical trial conducted in 38 hospitals in China, and registered as No. ChiCTR2100053112 in the Chinese clinical trial registry (http://www.chictr.org.cn). The steering committee was responsible for the design and conduct of the trial and for the analysis of the trial results. The data and safety monitoring board (DSMB) oversaw the trial and performed regular assessments of safety outcomes. Trial centers and investigators are listed in the Online Supplementary Materials (OSM) Table S1.

Details of the trial rationale, design, and methods are provided in the published protocol [20]. The trial protocol was approved by the institutional review boards at the Air Force Medical Center, Chinese People’s Liberation Army, and the trial adhered to the principles of the Declaration of Helsinki and the guidelines for Good Clinical Practice. All patients or legally authorized representatives provided written informed consent before enrollment.

Participants were eligible if they were aged between 35 and 85 years; had ≥ 70% stenosis or occlusion in the unilateral ICA or horizontal segment (M1) of MCA; had cerebral hypoperfusion in the ipsilateral MCA territory; had experienced no transient ischemic attacks (TIA) or ischemic strokes within 2 weeks; provided signed informed consent from either the patient or a surrogate. Exclusion criteria included the following: (1) ≥ 50% stenosis or occlusion in contralateral ICA or M1 of MCA; (2) ≥ 50% stenosis or occlusion in bilateral carotid arteries or an innominate artery; (3) preference to carry out CAS, CEA, or other revascularization surgeries; (4) ≥ 10 mm infarction in both MCA territories on CT or MRI; (5) other cerebral diseases (such as infection, degeneration, demyelination, tumor, or trauma) which influence cerebral perfusion; (6) received NBP or vasodilatation therapy within 30 days; (7) pregnant or lactating women; (8) severe cardiac, pulmonary, alimentary, or neoplastic diseases, and life expectancy ≤ 6 months; (9) serum creatinine ≥ 140 μmol/L or transaminase ≥ three times the upper limit of normal value; (10) cerebral stenosis caused by any disease other than atherosclerosis, such as arterial dissection, vasculitis, or moyamoya disease; (11) allergy to NBP or Apium graveolens; (12) participation in other clinical studies within 3 months; (13) other situations that are not suitable for inclusion.

Eligible patients were randomly assigned in a 1:1 ratio to the NBP group or the placebo group. Every drug kit was labeled with sequential numbers corresponding to a computer-generated randomization list. Successive patients were enrolled into any participating site with competitive principles and distributed with the kits from the lowest number to highest number by the drug administrators. All drug kits had an identical appearance and similar smell because the placebo was designed as a low and ineffective dose rather than a blank one. The patients, investigators, and drug administrators were blinded to the drug allocation.

The patients in the NBP group received 200 mg oral NBP three times daily. Because NBP has a plant odor, the placebo was designed as a low and ineffective dose rather than a blank one, in order to ensure an effective double-blind in this trial. Based upon previous study, no effect level (NOEL) is defined as the maximum recommended therapeutic dose (MRTD)/10 [21]. Since the MRTD for NBP is a dose of 200 mg, 20 mg is assumed to be the NOEL. Indeed, the 20 mg NBP is indistinguishable from the 200 mg dose by odor. Therefore, the patients in placebo group received 20 mg oral NBP three times daily. Both groups underwent 4 weeks of treatment while receiving standard long-term medical care according to the guidelines [1]. However, any drug resulting in volume expansion, vasodilatation, or induced hypertension was prohibited in this trial. The second cerebral perfusion was assessed at 12 weeks since a previous study suggested that NBP had a persistent effect on improving cerebral hypoperfusion in the patients with cerebral artery stenosis after its discontinuation [19]. Cerebral ischemic events including TIA and ischemic stroke in stenotic MCA and in other territory and drug adverse reactions were also recorded in the follow-up.

Cerebral artery stenosis was assessed by magnetic resonance angiography (MRA), computed tomography angiography (CTA), or digital subtraction angiography (DSA). The extracranial artery stenotic degree was calculated according to the North American Symptomatic Carotid Endarterectomy Trial (NASCET) and the intracranial artery stenotic degree was calculated according to the Warfarin-Aspirin Symptomatic Intracranial Disease (WASID) guidelines.

Cerebral perfusion was assessed by computed tomography perfusion (CTP). CTP was required to be carried out in the same equipment, using the same software, and by the same imaging technician within every center. The regions of interest (ROIs) was drawn in both stenotic and mirror MCA territories. In drawing ROIs, the regions with cerebral infarctions, calcifications, large vessel passages, and ventricles were avoided as much as possible. The shape, location, and size of ROIs must be as consistent as possible between baseline CTP and follow-up CTP in the same patient. The cerebral perfusion parameters included CBF, cerebral blood volume (CBV), mean transit time (MTT) and time to peak (TTP). The ratios of parameters in stenotic territory to those in mirror territory were used for the cerebral perfusion normalization and data analysis. These ratios were termed relative CBF (rCBF), relative CBV (rCBV), relative MTT (rMTT), and relative TTP (rTTP), respectively. Considering that CBF and CBV decrease and MTT and TTP increase in cerebral hypoperfusion, the rCBF was calculated by (CBFstenosis/CBFmirror) × 100%, rCBV by (CBVstenosis/CBVmirror) × 100%, rMTT by (MTTmirror/MTTstenosis) × 100%, and rTTP by (TTPmirror/TTPstenosis) × 100%. In normal subjects, the variability in left- and right-sided cerebral perfusion is not over 10% [22]. Therefore, the cerebral hypoperfusion was defined as rCBF ≤ 90% in this trial.

All neuroimaging data were either uploaded through the website or collected by hard disk and stored in the Trial Imaging Center. Two independent neuroradiologists monitored all the imaging data collected from each center, and confirmed that the baseline imaging data met inclusion and exclusion criteria and that the follow-up CTP is valid. Any disagreements were decided by the third neuroradiologist.

The grade of CBF change for treatment response was classified into amelioration, stabilization, and deterioration according to the ratio of rCBF at 12 weeks to baseline rCBF. The variability in repeated scans vary ± 10%, and the intra-patient and intra-observer differences are in the range of 10% [22, 23]. A previous study used 10% change as the threshold of cerebral perfusion impairment [24]. Therefore, CBF amelioration was defined as rCBFafter/rCBFbefore −1 ≥ 10%, and CBF deterioration was defined as rCBFafter/rCBFbefore −1 ≤ −10%. Furthermore, CBF stabilization was defined as −10% < rCBFafter/rCBFbefore −1<10%. The primary efficacy outcome was the percentage of patients achieving CBF amelioration at 12 weeks.

The comparable grades of CBV, MTT, and TTP changes were classified into amelioration, stabilization, and deterioration according to the ratio of values at 12 weeks to baseline. The secondary outcomes included the percentages of patients achieving CBV/MTT/TTP amelioration. In addition, the absolute values of rCBF, rCBV, rMTT, and rTTP change from baseline to 12 weeks in each patient were included in the secondary outcomes. The secondary outcomes also included the incidence of cerebral ischemic events in stenotic MCA territory and in any territory at 12 weeks, which were confirmed by CT or MRI. The safety outcomes included the incidence of all adverse events and serious adverse events (SAEs) during 12 weeks of follow-up.

A previous study showed that standard medical treatment achieved CBF improvement in approximately 30% patients with MCA stenosis [25]. NBP treatment resulted in 15% higher percentage of CBF amelioration than placebo in patients with carotid stenosis and cerebral hypoperfusion [19]. Assuming a power of 90%, α level of 0.05 (two-sided), and an overall dropout rate of 20%, the final target sample size estimate was 480 patients (240 patients in each treatment group).

The analysis plan was prespecified. The intention-to-treat population was defined as all randomized patients and the modified intention-to-treat population included the patients who completed follow-up CTP evaluations. The primary efficacy outcome and all CTP-based outcomes were analyzed in the modified intention-to-treat population. Analysis of cerebral ischemic events and safety outcomes was performed in the intention-to-treat population. In the sensitivity analysis, missing data of the primary outcome in the intention-to-treat population were handled using ratio imputation and multiple imputation. Ratio imputation indicated that missing data of the grade of CBF change was predicted by the ratio of the CBF amelioration, stabilization, and deterioration in patients with valid CBF data. The subgroup analyses of the primary outcome were done on the basis of age (< 60 or ≥ 60 years), sex, hypertension, diabetes, hypercholesterolemia, overweight and obesity (body mass index ≥ 24 kg/m2), taking statin drugs, previous cerebral ischemic events including TIA and ischemic stroke, stenotic location (ICA or MCA), and stenotic degree (severe or occlusion). No interim analyses were performed in this study. An independent DSMB reviewed the safety regularly and decided whether the study should continue.

Continuous variables were represented as means with standard deviations for normally distributed data or medians with interquartile ranges (IQRs) for non-normally distributed parameters. Categorical variables were presented as counts with corresponding percentages. Statistical comparisons between two groups were performed using chi-square tests or Fisher’s exact tests for categorical CTP response outcomes, while Student’s t-tests or the Wilcoxon Rank-Sum Test‌ were applied for continuous CTP response outcomes. The incidences of cerebral ischemic events and safety outcomes were similarly compared between two groups using appropriate categorical statistical tests. The categorical CTP response was compared using the log-binomial model and adjusting relevant covariates, and the corresponding treatment effects were presented as risk ratio (RR) with 95% confidence intervals (CIs). The Hodges-Lehmann Test was used to calculate the median difference with 95% CI for continuous CTP response. The new cerebral ischemic events and safety outcomes were compared using the log-binomial model and presenting RRs and 95% CIs. A two-sided p ≤ 0.05 was considered statistically significant. All statistical analyses were implemented using SAS version 9.4 (SAS Institute, Cary, NC, USA) and R version 4.5.0.