**2. Methods**

This randomized, double-blind, placebo-controlled, multi-center study was conducted at eight centers across China (Table S1) between 31 October 2018 and 29 January 2019. The design of this clinical trial complied with local laws and regulations and was developed in accordance with accepted standards for Good Clinical Practice of Pharmaceutical Products (2003) [18], Provisions of Drug Registration (2007) [19], and the Declaration of Helsinki. Data acquisition and analysis were performed in compliance with protocols approved by the Ethical Committee of Xiamen University and China-Japan Friendship Hospital (\*ethical approval number 2018-97-K69\*) and were prospectively registered on the Chinese Clinical Trial Registry www.chictr.org.cn (ChiCTR1800018611). All subjects signed informed consent forms prior to study initiation.

#### *2.1. Study Patients*

A total of 120 cases were included in the experimental *Houttuynia* group and the control placebo group, respectively, by applying the 1:1 design principle. Patients needed to meet the following inclusion criteria. First, subjects were required to be aged between 18 and 70 years and had to be diagnosed with MGD-related DED. Diagnostic criteria for MGD were according to the "Expert consensus of diagnosis and treatment of meibomian gland dysfunction in China (2017)" [20]. People who had eye symptoms, combined with abnormal eyelid margin and meibomian gland opening, or abnormal meibum secretion, were diagnosed as MGD. Second, diagnostic criteria for DED were according to the "Expert consensus on clinical diagnosis and treatment of dry eye (2013)" [21]. The following examination test criteria were required: 2 mm/5 min < Schirmer I test ≤ 10 mm/5 min, 0 s < TBUT ≤ 10 s, and corneal fluorescence staining point numbers < 10 points. That is, only patients with mild or moderate dry eye were included because severe dry eye generally needs to be combined with other treatments. Third, patients had to not use, or had to have stopped using, artificial tears for more than seven days. We chose just one eye from each subject for observations; if both eyes met the inclusion criteria, the more severely a ffected eye was used, while if the condition in both eyes was identical, the right one was used.

Patients were excluded if they had conditions including additional eye diseases such as macular degeneration, glaucoma, keratitis, and retinal vascular embolism, as well as lacrimal passage obstructions, prominent exophthalmos, and severe corneal decompensation. Patients were also excluded if they had undergone eye surgery less than half a year ago, if they had a history of wearing corneal contact lens up to one week prior to enrollment, or if they had su ffered from other serious primary diseases including those of the cardiovascular, cerebrovascular, liver, kidney, or hematopoietic systems. Patients were excluded from this trial if they had an allergic constitution or were allergic to experimental drug ingredients. Exclusion also took place if patients had participated in other clinical trials within the last three months, or if individuals were unable to cooperate or were mentally ill.

#### *2.2. Study Design*

A double-blind design was used in this study. Thus, the 1:1 ratio between experimental treatment and control groups was used to generate random codes using the statistical software SAS9.4 (SAS Institute Inc., Cary, NC, USA), applying the block randomization method. Subjects were then assigned a medication box labeled with a serial number which contained all medications for the treatment duration. The experimental group used commercially available *Houttuynia* eye drops (specification: 8 mL; national drug approval no. Z20010110; Sichuan Shenghe Pharmaceutical Co. Ltd., China). The exact composition of Houttuynia eye drops was 2-undecanone (C11 H22O) at the concentration of 9.8 μg/mL, and 2-undecanone was the main e ffective component of the Houttuynia

cordata extract [14,15]. Meanwhile, the control group was treated with a placebo agen<sup>t</sup> (Sichuan Shenghe Pharmaceutical Co. Ltd., Chengdu, China), and the placebo was a 0.72% (g/mL) sodium chloride solution with a pH value of 6.7 and osmotic pressure of 237 mOsmol/kg. Patients were also equipped with handheld ultrasonic nebulizers (product number: HL100A, Yuwell, Jiangsu, China) (Figure S1). The medication application method used in this study was atomizing each eye for 20 min per day with the ultrasonic nebulizer held in front of each eye. The atomizing tube was placed 5–10 cm away from the eye, and the patient kept their eyes wide open, staring in all directions intermittently to ensure that the atomizing agen<sup>t</sup> fully touched the conjunctival sac. Patients used 20 mL houttuynia cordata eye drops or placebo at a time. The temperature of an ultrasonic-atomized aerosol is very close to room temperature, making the therapy induce slight irritation. The treatment course was four weeks with visit points at time zero (baseline) and then after one week, two weeks, and four weeks; treatment e fficacy and safety assessments were carried out on-site during these visits. All tests were carried out using the same equipment type. The tester and the statistician were both unaware of patient groupings.

#### *2.3. Outcome Measures*

Primary outcomes evaluated included eye symptom score (using the Chinese Dry Eye Questionnaire, which is a dry eye questionnaire that conforms to the use habits of Chinese patients; the clinical diagnosis of dry eye in Chinese patients showed better diagnostic value than the OSDI questionnaire) [22] (Table S2), meibum quality, and tear break-up time (TBUT). Secondary outcomes evaluated included meibum expressibility, meibomian gland dropout, eyelid margin change, conjunctival congestion, corneal staining, and the Schirmer I test [23].

Sodium fluorescein TBUT was measured using a commercial fluorescein strip which was moistened and applied to the lower eyelid conjunctiva. One minute after application, patients were then asked to blink three times and to hold their eyes open. The time between the eyes open and the appearance of the first dark spot on the precorneal film was then measured using a slit lamp with cobalt blue illumination and a yellow-barrier filter. This was repeated three times before an average for each eye was recorded [21].

Meibomian gland functionality was assessed by applying digital pressure to the central (nasal, temporal) third of the lower/upper lids in order to determine MGD extent and severity (i.e., expressibility and meibum quality). Scoring values for meibum quality were as follows: 0, clear (normal); 1, cloudy; 2, cloudy with particles; and 3, inspissated (like toothpaste). Values were recorded for the highest grade encountered from any expressed glands, encompassing a range between 0 and 3. Similarly, meibum expressibility scoring criteria were as follows: 0, all glands expressed; 1, 3–4 glands expressed; 2, 1–2 glands expressed; and 3, no glands expressed. Meibum expressibility was scored based on five glands from the nasal side, while values for the middle and temporal sides of the lower eyelid were also recorded. Total scores for these three parts were recorded at the same time, across a range between 0 and 9 [23].

The eyelid margin change scoring criteria used follow standards set by the international MGD consensus group, while the presence or absence of lid abnormalities were scored via irregularity of the lid margin, lid margin vascular engorgemen<sup>t</sup> including plugging of the meibomian orifices, and anterior or retroplacement of the MCJ. These were either scored as present (1) or absent (0), across a range between 0 and 4 [23].

Meibomian gland dropout scoring criteria refer to the loss of acinar tissue detected by meibography. Non-contact meibography using a standard infrared video security camera was examined in patients. Scoring values for meibomian gland dropout were as follows: 0, no gland dropout; 1, between 1% and 33%; 2, between 34% and 66%; and 3, ≥67% dropout. The percentage of partial, or total, gland dropout of the lower lid was recorded across a range between 0 and 3 [23].

The MGD staging was determined according to the score of meibum quality, eyelid margin change, meibomian gland expressibility (middle meibomian gland expression score), and dropout. The higher these index scores were, the more serious the MGD was [20].

The conjunctival congestion scoring criteria used here follow the classification outlined by the Cornea and Contact Lens Research Unit, as follows: 0, no hyperemia; 1, mild hyperemia; 2, moderate hyperemia; and 3, severe hyperemia [24].

The corneal staining scoring criteria involved fluorescein staining that was examined two minutes subsequent to the instillation of sodium fluorescein at the slit lamp using cobalt blue illumination and a yellow-barrier filter. The 12-points method was used to assess the staining degree. Each of the four quadrants of the cornea on a scale ranged between 0 and 3: 0, no staining; 1, between one and 30 dots stained; 2, >30 dots stained without fusion; and 3, corneal diffuse or coalescent punctate staining, corneal filaments, and an ulcer. Scores for the four quadrants were then summed to attain a total score for each eye across a range between 0 and 12 [21].

The Schirmer I test was performed on unanesthetized patients by placing a strip in each eye and recording the wetted length (mm) after five minutes [21].

Eye tolerance, irritant symptoms, and treatment satisfaction values were also incorporated in this analysis.

The safety evaluations included adverse events (AEs), routine eye examinations (including monitoring of uncorrected/corrected visual acuity and intraocular pressure (IOP)), vital signs, and laboratory indicators (including blood and urine routine, liver and kidney function) which were selected according to the requirements of each hospital and the condition of subjects. Artificial tears, anti-inflammatory eye drops, and physical therapies such as eyelid cleaning, the use of hot compresses, meibomian gland massages, acupuncture, and silicone eye masks were prohibited. The comorbidity of patients and the use of any drug or treatment for the comorbidity was recorded, including the drug name (or other treatment), dosage, number, and time of use. If a prohibited drug or treatment was used, the subjects were required to discontinue the study.

#### *2.4. Statistical Methods*

A descriptive quantitative statistical analysis was performed encompassing the number of cases, mean, standard error of the mean (SEM), and 95% confidence intervals (CIs). A group *t*-test or a Wilcoxon rank-sum test was then used to make comparisons between groups and covariance analysis was also performed. In the case of qualitative data, a descriptive statistical analysis was carried out on the number of cases in various categories and their percentages. Thus, the chi-squared test alongside the Fisher exact test was used to compare enumerated data between the different treatment groups, while a Wilcoxon rank-sum test was used for comparison amongs<sup>t</sup> treatment groups. A CMH-x<sup>2</sup> test was used when the center or other factors were considered, while the relationship between a reduction in eye symptom score after four weeks of treatment and baseline scores was analyzed using Spearman's correlation coefficient. The reduction in eyelid margin change score vs. the baseline was also examined using Spearman's correlation coefficient, following the intention-to-treat principle (ITT). The latest observation data were carried forward to the final test result in each case to determine missing values for the main efficacy evaluation index (the last observation carried forward (LOCF) method). The post hoc subgroups were defined according to baseline severity of symptoms (eye symptoms score ≥15 vs. <15) and severity of signs (meibum quality score 0–3, middle meibomian gland expression score 0–3, eyelid margin change score 0–4, conjunctival congestion score 0–2, corneal staining score 0, 1, >1). Tests of interaction were used to evaluate whether the effect of atomization treatment with Houttuynia eye drops and placebo differed among subgroups. The software packages SAS9.4 (SAS Institute Inc., Cary, NC, USA) and GraphPad Prism version 8.2.1 (GraphPad Software, San Diego, CA, USA) were used for all statistical calculations. All tests were two-sided, and *p* < 0.05 was considered significant.
