Clinical efficacy of portable, cost-effective manual thermal pulsation for obstructive meibomian gland dysfunction and factors associated with therapeutic efficacy Authors: Xiaoqi Cheng, Hangyu Huang, Zixin Fan, Heyun Wang, Xiaochun Zheng, Lina Dong, Ming Li, Guoming Zhang, Shengli Mi Published: 2026 ABSTRACT AbstractObjectiveTo evaluate the efficacy and safety of portable, cost-effective manual thermal pulsation as a single therapy for meibomian gland dysfunction (MGD) and to compare the microbiota and lipid composition in subjects with different therapeutic efficacy.Methods and analysisIn this single-centre, randomised, positive-controlled clinical trial, 84 subjects were randomly assigned (1:1) to receive a single treatment of manual (self-developed device) or automatic (Lipiflow) thermal pulsation and followed up for 4 weeks, 76 of which were finally analysed. Efficacy was evaluated by Δ Meibomian Gland Score (MGS) (Δ variable=parameter at follow-up−parameter at baseline), Δ non-invasive break-up time (NIBUT), Δ standard patient evaluation of eye dryness (SPEED), Δ Meibomian Gland Expressibility Score (MES) and Δ corneal fluorescein staining (CFS), with safety investigated. Subjects who underwent manual thermal pulsation were further divided into two subgroups based on whether ΔMGS ≥−2, meibum microbiome and lipid metabolomics in the subgroups were analysed.ResultsΔMGS, ΔNIBUT, ΔSPEED, ΔMES and ΔCFS were not significantly different between manual and automatic thermal pulsation groups in 4 weeks (both p>0.05), indicating non-inferior efficacy of manual thermal pulsation. No serious ocular or device-related adverse events were reported. A more complex meibum microorganism diversity and 46 differentially expressed lipids (such as triglycerides, diglycerides, ceramides and oxidised species) were identified in those with a relatively weak therapeutic efficacy.ConclusionsPortable, cost-effective manual thermal pulsation demonstrated non-inferiority to automatic thermal pulsation for MGD during the 4-week follow-up period, with therapeutic efficacy associated with meibum micro-organism and lipid components. WHAT IS ALREADY KNOWN ON THIS TOPIC Meibomian gland dysfunction (MGD) results from abnormal secretion of meibum due to blockage of terminal ducts and glandular secretions, causing symptoms such as ocular dryness, tingling, foreign body sensation and fluctuating vision. Meibomian gland dysfunction (MGD) results from abnormal secretion of meibum due to blockage of terminal ducts and glandular secretions, causing symptoms such as ocular dryness, tingling, foreign body sensation and fluctuating vision. WHAT THIS STUDY ADDS To overcome the spatial limitations of current thermal pulsation systems, thereby transcending a key barrier to accessible treatment, we developed a novel, portable cost-effective device for manual thermal pulsation and confirmed its non-inferiority compared with automatic thermal pulsation. To overcome the spatial limitations of current thermal pulsation systems, thereby transcending a key barrier to accessible treatment, we developed a novel, portable cost-effective device for manual thermal pulsation and confirmed its non-inferiority compared with automatic thermal pulsation. The observed variations in therapeutic efficacy may be linked to interindividual differences in the meibum microbiome and lipid composition. The observed variations in therapeutic efficacy may be linked to interindividual differences in the meibum microbiome and lipid composition. HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY Beyond demonstrating efficacy and safety of the manual thermal pulsation, the study may redefine the treatment paradigm for MGD, potentially expanding future treatment options for patients to settings such as home care and primary care clinics. Beyond demonstrating efficacy and safety of the manual thermal pulsation, the study may redefine the treatment paradigm for MGD, potentially expanding future treatment options for patients to settings such as home care and primary care clinics. Introduction Meibomian glands are specialised sebaceous glands in the eyelids that are essential for ocular surface health. Meibomian gland dysfunction (MGD) results from abnormal secretion of meibum due to blockage of terminal ducts and glandular secretions, leading to inadequate tear composition and secretion, thereby causing symptoms such as ocular dryness, tingling, foreign body sensation and fluctuating vision.1 MGD is a prevalent ocular surface disease globally, with an overall prevalence rate ranging from 21.2% to 71.0%.2 It is identified as the primary cause of evaporative dry eye, with approximately 85% of dry eye patients exhibiting signs of MGD.3 Influenced by contemporary risk factors associated with MGD, such as prolonged use of electronic devices, poor sleep quality and working in air-conditioned and dry environments,4 the prevalence of MGD has gradually increased, emerging as an important public health issue. Conventional MGD treatment methods include eyelid cleaning, local hot compress, eyelid massage, punctal plugs, etc.5 6 The predominant approach involves ‘hot compress+massage’, where heat is applied to the patient’s eyes followed by pressure on the eyelid area using cotton swabs or eye curettes to alleviate meibomian gland blockage. However, the efficacy of this method is limited due to pain during treatment and reliance on operator’s expertise.7 To address these challenges, automatic thermal pulsation systems such as the LipiFlow Thermal Pulsation System (Johnson & Johnson Vision, Jacksonville, Florida, USA) have been developed to alleviate pain issues and enhance efficacy. These systems use automated temperature and pressure control, administered by eye care professionals.811 Nonetheless, its relatively large equipment size and requirement for patients to lie down during operation necessitate specific treatment location standards. A portable, cost-effective and efficient MGD therapy holds the potential to overcome the spatial limitations of current thermal pulsation systems, thereby transcending a key barrier to accessible treatment. We have engineered a novel, portable, cost-effective and manual device for MGD treatment. It operates by heating the targeted tissue area of the eyelids through heat conduction and uses a pressing and rolling mechanism wheel to facilitate the drainage of clogged meibum from the meibomian glands. The entire treatment process can be observed, magnified and captured in real-time via the display screen, allowing patients to undergo treatment comfortably while seated. Furthermore, image feedback allows for targeted intensive treatment of severely affected areas, enabling true personalisation of treatment plans. This study evaluated the efficacy and safety of manual thermal pulsation as a single treatment for MGD, while verifying its non-inferiority compared with automatic thermal pulsation. The study further examined how the meibomian gland microbiota-metabolome profile influences therapeutic efficacy. Method Patients This study enrolled adult patients with a standard patient evaluation of eye dryness (SPEED) score >6 and a clinical diagnosis of bilateral MGD. Detailed inclusion and exclusion criteria are provided in online supplemental note S1. Questionnaire survey The SPEED questionnaire was developed by Tear Science (Morrisville, NC) to identify dry eye symptoms in patients. The questionnaire design takes into account both symptom frequency and severity and can track changes in daytime and long-term symptoms.12 The detailed structure, items and scoring rubric of the questionnaire are provided in online supplemental note S2. Patient assessment The patient participated in a total of four visits, including pretreatment and evaluations at 1 week, 2 weeks and 4 weeks post-treatment. A slit lamp SL 800 (Carl Zeiss Shanghai, Shanghai, China) was used to assess Meibomian Gland Score (MGS), Meibomian Gland Expressibility Score (MES), corneal fluorescein staining (CFS) and conduct basic eye condition examinations (external eye, eye position, conjunctiva, sclera, cornea, iris, pupil, lens, vitreous body and fundus). Additionally, OCULUS Keratograph 5M (OCULUS, Arlington, American) was employed to evaluate non-invasive break-up time (NIBUT). Ocular surface subjective symptoms were assessed using the SPEED score,12 while uncorrected visual acuity (UCVA) and intraocular pressure (IOP) were also measured. All examinations were performed by an ophthalmologist. Clinical assessments included MGS, MES and CFS. The detailed operational definitions and grading criteria for all scores are provided in online supplemental note S2. The non-inferiority margins in our study were ΔMGS=–7.5 and ΔNIBUT=–2.5, which were determined by referring to relevant research, taking into account the specific clinical context, and with the final decision made jointly by statistical experts and clinical experts.13 Treatment plan Detailed protocols for thermal pulsation treatments Two thermal pulsation systems were employed for the treatment of MGD: a manual thermal pulsation device and an automated thermal pulsation system (LipiFlow). The detailed operating procedures, device diagrams and key procedural stages for both treatment modalities are described in online supplemental note S3 and online supplemental figure 1. Prohibited treatment measures during follow-up During the trial, the use of antibiotics, glucocorticoids and immunosuppressants is prohibited. Additionally, subjects are not allowed to undergo eyelid cleaning, hot compresses, eyelid massage, eyelid thermal pulsation treatments (such as Blephasteam, EyeGiene, Lipiflow, MGDRx eye mask), or intense pulsed light therapy. Furthermore, microprobe dredging technology and other similar treatments are prohibited, and wearing contact lenses is not permitted. Observation indicators The primary outcomes of this study were ΔMGS and ΔNIBUT during the follow-up period. The secondary outcomes included ΔSPEED, ΔMES and ΔCFS (Δvariable=parameters at follow-up-parameters at baseline) to evaluate the non-inferiority of manual thermal pulsation compared with automatic thermal pulsation. Objective indicators (ocular surface injury and infection, UCVA, IOP), adverse events and device defects were used to evaluate the safety of treatment. Sample collection and subgroup analysis In the manual thermal pulsation group, tear samples were collected with Schirmer’s test strips prior to any procedures at each follow-up. Meibum was subsequently collected during manual thermal pulsation treatment using sterile cotton swabs. All samples were stored at –80°C. For the meibum microbiome and lipid metabolomics analysis, the manual thermal pulsation group was divided into two subgroups according to whether ΔMGS ≥ −2: (1) WTE group with relatively weak therapeutic efficacy and (2) BTE group with better therapeutic efficacy. DNA extraction, PCR amplification, 16S rRNA gene sequencing and data analysis To elucidate the relationship between the degree of MGD improvement and microbial diversity, we performed 16S rRNA gene sequencing analysis. Detailed experimental protocols, bioinformatics workflows and parameter settings are provided in online supplemental note S4. Lipid sample extraction Lipid extraction and sample preparation were performed using a methyl tert-butyl ether-based method. The detailed extraction protocol is provided in online supplemental note S5. Lipid metabolomics testing Ultra Performance Liquid Chromatography and Tandem mass spectrometry compared lipid profiles between treatment response groups to identify differential lipids and enriched pathways. Details are in online supplemental note S6. Statistical analysis Statistical analyses were conducted using SAS V.9.4. Quantitative data are presented as means and SD. Statistical differences between groups were compared using t-tests or nonparametric tests. Qualitative data were described using counts and percentages. Statistical differences between groups were compared using the χ² test or Fisher’s exact probability test. All statistical tests were two-sided, with a significance level of α=0.05, and differences were considered statistically significant when p<0.05. Randomisation and blinding procedures This was an evaluator-blinded, randomised controlled trial. Details regarding the generation of the randomisation scheme and procedures to maintain blinding are provided in online supplemental note S7. Results Patient characteristics Between 5 December 2022 and 22 April 2023, a total of 104 subjects participated in the study. Among them, 20 subjects were excluded. Subjects (N=84) were randomised to receive manual thermal pulsation (N=42) or automatic thermal pulsation (N=42) treatment, and both eyes of the subjects received treatment. Finally, 80 subjects completed follow-up (manual thermal pulsation, N=41; automatic thermal pulsation, N=39). Among them, 76 subjects were included in PPS (per-protocol set) (manual thermal pulsation, N=38; automatic thermal pulsation, N=38) (figure 1). CONSORT diagram chart (patient disposition of the manual thermal pulsation and automatic thermal pulsation treatment groups at each stage of the study). CONSORT, Consolidated Standards of Reporting Trials; DisionEyemore, Manual Thermal Pulsation Device; Lipiflow, Automatic Thermal Pulsation System; N, number of subjects. To compare the therapeutic efficacy and safety of the two therapeutic methods, two demographically matched groups of patients were included. Table 1 summarises the baseline characteristics of the study subjects in the manual thermal pulsation group and automatic thermal pulsation group. Subjects in the manual thermal pulsation and automatic thermal pulsation treatment groups were comparable in terms of age, gender, nationality and clinical characteristics (both p>0.05). Table 1Baseline characteristics of study subjectsManual thermal pulsation(N=38)Automatic thermal pulsation(N=38)Overall(N=76)Sociodemographic characteristicsSex, N (%) Male13 (34.3)11 (29.0)24 (31.6) Female25 (65.7)27 (71.0)52 (68.4)Nationality, N (%) Han35 (92.1)33 (86.8)68 (89.4) Mongolian0 (0.0)1 (2.6)1 (1.3) Miao0 (0.0)1 (2.6)1 (1.3) Tujia2 (5.2)2 (5.2)4 (5.2) Yao0 (0.0)1 (2.6)1 (1.3) Yi1 (2.6)0 (0.0)1 (1.3)Age Mean±SD42.05±13.7839.97±12.5841.01±13.15 Median (Q1, Q3)39.50 (29.50,55.00)35.00 (30.00,51.00)35.00 (30.00,53.50) Min–Max24.00–69.0021.00–65.0021.00–69.00Clinical characteristics MGS, mean±SD10.34±4.2811.11±4.9110.72±4.59 NIBUT(OS), mean±SD (s)6.10±1.546.15±1.376.12±1.45 NIBUT(OD), mean±SD (s)6.06±1.626.01±1.506.04±1.48 SPEED, mean±SD9.61±2.399.37±2.399.49±2.38 MES(OS), mean±SD3.18±2.192.45±2.002.82±2.12 MES(OD), mean±SD3.18±2.132.66±1.822.92±1.99 CFS(OS), mean±SD0.16±0.440.13±0.480.15±0.45 CFS(OD), mean±SD0.16±0.440.11±0.390.13±0.41CSF, corneal fluorescein staining; MES, Meibomian Gland Expressibility Score; MGS, Meibomian Gland Score; NIBUT, non-invasive break-up time; OD, Oculus Dexter; OS, Oculus Sinister; SPEED, standard patient evaluation of eye dryness. Baseline characteristics of study subjects CSF, corneal fluorescein staining; MES, Meibomian Gland Expressibility Score; MGS, Meibomian Gland Score; NIBUT, non-invasive break-up time; OD, Oculus Dexter; OS, Oculus Sinister; SPEED, standard patient evaluation of eye dryness. Effectiveness In the first week after treatment, both the manual thermal pulsation group and the automatic thermal pulsation group exhibited greater improvement in MGS compared with baseline, with this effect sustained until the fourth week (ΔMGS first week: manual thermal pulsation, −5.11±5.62; automatic thermal pulsation, −5.16±6.79; fourth week: manual thermal pulsation, −5.82±4.89; automatic thermal pulsation, −5.82±5.87; p>0.05), and the difference between the two groups was not statistically significant (figure 2(A)). Further analysis using a mixed-effects model with repeated measures (MMRM) revealed that at the fourth week of treatment, the lower limit of the 95% CI of the mean difference in ΔMGS was −0.775, which exceeded the preset non-inferiority margin of −7.5. This indicates that the manual thermal pulsation is non-inferior to the automatic thermal pulsation in improving MGS at the fourth week of treatment. Changes in (A) MGS, (B) NIBUT, (C) SPEED, (D) MES, (E) CFS from baseline at 1, 2 and 4 weeks after treatment. CFS, corneal fluorescein staining; MES, Meibomian Gland Expressibility Score; MGS, Meibomian Gland Expressibility Score; NIBUT, non-invasive break-up time; SPEED, standard patient evaluation of eye dryness;OS, Oculus Sinister; OD, Oculus Dexter. After the initiation of treatment, the NIBUT of the manual thermal pulsation group and automatic thermal pulsation group improved compared with the baseline, and the effect was maintained until the fourth week (ΔNIBUT fourth week: manual thermal pulsation, Oculus Sinister(OS), 0.58±1.99, (Oculus Dexter)OD, 0.18±1.99; automatic thermal pulsation, OS, 0.18±1.85, OD, 0.56±1.89; p>0.05), and the difference between the two groups was not statistically significant (figure 2(B)). Further analysis using MMRM revealed that at the fourth week of treatment, the lower limits of the 95% CIs for the mean differences in ΔNIBUT (OS and OD) were −1.056 and −0.324 respectively, both exceeding the predetermined non-inferiority margin of −2.5. This demonstrates the non-inferiority of the manual thermal pulsation in improving NIBUT compared with the automatic thermal pulsation at week 4 of treatment. The SPEED in both the manual thermal pulsation group and the automatic thermal pulsation group showed significant improvement after treatment, with the manual thermal pulsation group demonstrating non-inferiority to the automatic thermal pulsation group (4.42±3.77 vs −3.34±4.33, p>0.05) (figure 2(C)). The MES and CFS of the manual thermal pulsation group improved after treatment, and the manual thermal pulsation was not inferior to the automatic thermal pulsation (MES: OS −0.63±2.96 vs 0.45±2.84, p>0.05; OD: −0.42±3.28 vs 0.32±2.76, p>0.05; CFS: OS −0.13±0.47 vs 0.05±1.01, p>0.05; OD: −0.13±0.47 vs 0.11±0.98, p>0.05) (figure 2(D) and (E)). Safety During the study period, no serious adverse events occurred in either group, and no subject withdrew from the trial due to adverse events. All adverse events were graded using a standard 4-point severity scale (1=mild, 2=moderate, 3=severe, 4=life-threatening). Among the 42 subjects in the manual thermal pulsation group, a total of 8 adverse events occurred (19.0%). Among the 42 subjects in the automatic thermal pulsation group, a total of 9 adverse events occurred (21.4%). Device-related adverse events occurred in five manual (11.9%) and six automatic patients (14.3%). The most common was transient foreign body sensation (3 patients each, 7.1%). Other device-related adverse events included conjunctival congestion (manual: 2, 4.8%; automatic: 3, 7.1%), corneal epithelial abrasion (manual: 2, 4.8%), and ocular itching (manual: 1, 2.4%). All device-related adverse events were mild and short-lived: foreign body sensation resolved within minutes to hours; conjunctival congestion and corneal abrasion resolved within 24–48 hours without intervention. No patient required treatment modification, rescue therapy or device discontinuation. Furthermore, no subjects in either group reported device defects during treatment. Additionally, during the treatment period, three patients in the manual thermal pulsation group (7.14%) and one patient in the automatic thermal pulsation group (2.38%) contracted COVID-19. Microbiota analysis among subgroups As depicted in figure 3(A), different colonies can be cultured from the two subgroups of meibum. Furthermore, we profiled all samples from the WTE group and the BTE group using 16S rRNA sequencing, clustered the sequences into operational taxonomic units, and annotated their taxonomic classifications at the genus level. The Venn diagram illustrated the eight bacterial species common to both the WTE group and the BTE group, eight species unique to the WTE group and one species unique to the BTE group (figure 3(B)). Compared with the BTE group, the ocular surface microorganisms of patients in the WTE group exhibited greater diversity. The eight unique ocular surface genera in the WTE may have modulated the single-session efficacy of the manual thermal pulsation. (A) Distribution of bacterial flora in WTE (relatively weak therapeutic efficacy) and BTE (better therapeutic efficacy) group; (B) Genus level species venn diagram; (C) LEFse difference analysis (LDA) chart; (D) LEFse cladogram. LEFse, LDA Effect Size. To identify robust biomarkers distinguishing patients in the WTE group from those in the BTE group, we performed Linear Discriminant Analysis Effect Size analysis with stringent thresholds (LDA score >3.5, p<0.05). The analysis revealed that the relative abundances of Moraxellaceae, Pseudomonadales, Veillonellaceae, Negativicoccus, Selenomonadales and Lactobacillus were significantly enriched in the WTE group compared with the BTE group. Conversely, the abundance of Corynebacterium was significantly lower in the WTE group (figure 3(C)). This observed reduction in Corynebacterium may be associated with the treatment efficacy observed in WTE group patients. Furthermore, the cladogram visually delineated the distinct ocular surface microbial communities between the WTE and BTE group (figure 3(D)). Non-targeted lipid metabolomic analysis among subgroups Using Orthogonal Partial Least Squares Discriminant Analysis (OPLS-DA) modelling, figure 4(A) demonstrates a significant cluster separation between the WTE group and the BTE group, indicating distinct sample characteristics between the two groups. (A) OPLS-DA score plot; (B) differential lipid clustering heat map; (C) differential lipid volcano plot;VIP,variable importance projection; (D) differential lipid radar chart; (E) differential lipid violin plot; (F) differential lipid pathway classification diagram;KEGG,Kyoto Encyclopedia of Genes and Genomes; (G) differential lipid Human Metabolome Database(HMDB) enrichment plot. BTE, better therapeutic efficacy; GPE, glycosylphosphatidylinositol; OPLS-DA, Orthogonal Partial Least Squares Discriminant Analysis; WTE, weak therapeutic efficacy. To visualise the patterns in relative lipid abundance, the raw relative abundances of differential lipids, identified based on our screening criteria, were subjected to unit variance scaling and used to construct a heatmap using R software. figure 4(B) illustrates the distribution profiles of variable importance projection(VIP) scores between the WTE group and the BTE group. 46 differentially expressed lipids were identified, including 44 upregulated ones and 2 downregulated ones in the WTE group (figure 4(C)). Figure 4(D) and (C) display the top 10 lipids with the largest absolute log2FC values. The differential lipids in the WTE group comprised triglycerides, diglycerides, oxidised lipids, ceramides and docosapentaenoic acid. The Kyoto Encyclopaedia of Genes and Genomes enrichment analysis is depicted in figure 4(F). Primarily within the Metabolism category, significant lipid alterations were observed in Glycerophospholipid Metabolism, Glycerolipid Metabolism, glycosylphosphatidylinositol—Anchor Biosynthesis, Sphingolipid Metabolism and Glycine, Serine and Threonine Metabolism pathways. These findings suggest a systemic dysregulation of lipid synthesis and metabolism in the WTE group, rather than isolated changes in individual lipids. Pathway enrichment analysis was conducted on the pathways using the Human Metabolome Database(HMDB) subdatabase Small Molecule Pathway Database(SMPDB). The results in figure 4(G) suggest that the differential lipids are associated with α-linolenic acid (ALA) and linoleic acid (LA) metabolism. Discussion MGD is a common clinical ocular surface disease. Traditional treatment methods have not been ideal in improving MGD, and the duration of improvement is short. Previous studies have shown that the melting temperature of meibum ranges from 38°C to 50°C. When it is below 40°C, abnormal meibum cannot melt completely, while temperatures above 44°C pose a risk of damage to the anterior chamber and skin.14 15 Traditional hot compresses require frequent reheating to maintain the appropriate temperature, are cumbersome to operate, often cause discomfort during the treatment process and rely on the operator’s experience, thus having certain limitations. The manual thermal pulsation system and automatic thermal pulsation system are devices that heat the inner eyelid surface by applying constant heat (42.5°C) directly to the lid surface of the upper and lower eyelids to reach the meibum melting temperature. The automatic thermal pulsation system applies gradual pulsating pressure to the outer eyelid surface through the inflatable airbag to drain the glandular secretions. The manual thermal pulsation system squeezes the meibomian glands by pressing and rolling the mechanism wheel to drain out the clogged meibum. Additionally, the manual thermal pulsation system provides real-time visualisation, allowing clinicians to monitor, magnify and document the procedure via a display. This image-guided capability enables precise targeting and intensified treatment of severely affected areas, thereby tailoring the therapeutic strategy to each patient. Hu et al16 conducted a systematic evaluation of the efficacy and safety of the LipiFlow thermal pulsation system for MGD, demonstrating its superior improvement in both subjective symptoms and objective clinical signs compared with conventional therapies, with a favourable safety profile. A recent systematic review and network meta-analysis further confirmed that LipiFlow significantly improves tear breakup time and Ocular Surface Disease Index scores compared with control interventions. Therefore, it served as an appropriate ‘gold standard’ active comparator in this non-inferiority trial.17 Accordingly, this study employed an automatic thermal pulsation as a positive control to assess the efficacy and safety of the manual thermal pulsation for MGD, with the exploratory objective of determining meibum-related determinants of treatment response. MGD patients who underwent manual thermal pulsation treatment and automatic thermal pulsation treatment all demonstrated improvements in MGS, NIBUT, SPEED, MES and CFS. A single treatment with manual thermal pulsation was found to be not inferior to automatic thermal pulsation in improving MGD. However, the efficacy of automatic thermal pulsation system in this study is slightly lower than in other studies. Possible reasons include: (1) The differences in the populations included in the study. The population in this study is Asian, and the prevalence of MGD in Asian populations is higher than in Caucasians.1820 (2) The study was conducted during the period of COVID-19 infection. On one hand, wearing a mask may increase the risk of dry eyes.21 On the other hand, patients infected with COVID-19 have shown decreased meibomian gland function and lipid layer thickness, along with increased conjunctival follicles and shortened tear film break-up time.22 In this study, the proportion of patients suffering from COVID-19 in both the manual thermal pulsation group and the automatic thermal pulsation group was similar, without statistical significance. (3) Meteorological conditions are related to the occurrence and persistence of dry eye. Among these conditions, weather fluctuations during the transition from spring to autumn often exacerbate symptoms.23 This study was conducted during winter and spring, coinciding with seasonal changes. Considering these factors, the actual treatment effects of both groups in this study may be better than the research results suggest. Additionally, no serious adverse events occurred in either group, and no subjects withdrew from the trial due to adverse events. Based on the current study results, the benefits of manual thermal pulsation treatment far outweigh the risks, indicating that the treatment is both active and effective, with relatively low associated risks. In order to further explore the factors influencing the efficacy of manual thermal pulsation treatment, we compared the distribution of meibum flora and lipid metabolism fingerprints among different efficacy subgroups. We acknowledge that the subgroup sample sizes (WTE, n=4; BTE, n=5) were small and limit the generalisability of these findings. Therefore, the following results should be interpreted as hypothesis-generating and require validation in larger cohorts. The ocular surface of healthy individuals harbours a diverse microbial flora, and alterations in this flora are associated with various disease states, including dry eye syndrome, blepharitis, conjunctivitis, keratitis and others.24 25 Evidence suggests that ocular surface disorders may not be attributable to a single pathogen, but rather to a dysbiosis of the microbial community. Therefore, maintaining ocular surface microbial homeostasis may help prevent these diseases.24 This study observed that species diversity in patient samples with slightly poorer treatment outcomes was significantly higher compared with the BTE group. This finding aligns with previously reported conclusions that increased microbial diversity is a characteristic feature of dry eye disease.26 Zhang et al27 found that MGD patients exhibited a significantly higher rate of positive bacterial cultures. Furthermore, a greater diversity of bacteria, notably various Staphylococcus, Corynebacterium and Propionibacterium species, was isolated from their meibomian glands and conjunctival sacs. Baim et al28explained that patients with dry eye syndrome may experience ocular surface epithelial damage or reduced secretion of mucin and antibacterial substances. These conditions can result in a decrease in the antibacterial capability of the ocular surface, allowing ocular surface bacteria to potentially thrive and proliferate.28 Changes in the diversity and composition of the bacterial flora can lead to abnormalities in ocular surface immunity and material metabolism, weakening the ocular surface barrier and immune tolerance, thereby promoting the secretion of inflammatory factors by immune cells. Additionally, pathogenic bacteria may secrete toxic substances that exacerbate ocular surface damage. The WTE group with inferior therapeutic outcomes showed a reduced abundance of Corynebacterium. Research conducted by Rachel Caspi and others29 has found that Corynebacterium, an aerobic bacterium residing on the surface of the eyeball, can stimulate the production of immune cells and release antibacterial factors into tears, which can effectively eliminate harmful microorganisms. However, given the small sample size of our subgroup analysis, these observations are preliminary and should be confirmed in future studies with larger sample sizes. Meibomian gland secretions comprise a complex mixture of various polar and non-polar lipids, including cholesterol, wax esters, diesters, triacylglycerols, free cholesterol, free fatty acids and phospholipids.3 Lipids play a crucial role in maintaining the homeostasis of the tear film and are primarily secreted by the meibomian glands located at the eyelid margins.30 Increasing evidence suggests that abnormal lipid metabolism may contribute to the onset and progression of MGD.31 32 A total of 46 differential meibum were identified in the WTE and BTE groups of this study. From the OPLS-DA score map (figure 4(A)) and heat map (figure 4(B)), significant differences in meibum composition between the two groups of samples were observed. A notable change was the alteration of glycerophospholipids and sphingolipids in the meibomian glands of patients in the WTE group. Sphingolipids and glycerophospholipids are essential components of biological membranes and are regulated by healthy T cells.33 Compared with samples from the group with better curative effects, those from the group with poorer outcomes exhibited significant increases in glycerophospholipids and sphingolipids. In dry eye syndrome, reduced secretion of meibomian gland oil exposes the skin of the eyelid edge and basal cells of the meibomian gland to environmental factors, rendering them vulnerable to oxidative stress and inflammation, thereby leading to apoptosis of meibomian gland basal cells.34 It is speculated that conditions such as meibomian gland cell apoptosis contribute to the elevated levels of glycerophospholipids and sphingolipids, which may positively correlate with the severity of dry eye syndrome. Dysregulation of glycerophospholipid and sphingolipid metabolism directly compromises tear film lipid stability, while altered glycerolipid metabolism impairs its evaporative barrier function.34 Ceramides, a major component of sphingolipids, contribute to the structural integrity and evaporative regulation of the tear film lipid layer. Studies indicate that elevated ceramide levels may raise the melting point of meibum, consequently reducing its fluidity, which suggests a potential link between lipid melting behaviour and treatment efficacy in the WTE.35 Samples from patients with better curative effect exhibited normal levels of glycerophospholipids and sphingolipids. Colonyomics analysis revealed that ocular surface microbial dysbiosis is associated with the severity of MGD. The potential effects of ocular surface dysbiosis on tarsal metabolism require further investigation. Microbial interference and the overproliferation of invasive bacteria on the ocular surface can induce the overproduction of bacterial lipase, leading to an imbalance in meibum content. Furthermore, ocular microbiota may trigger inflammatory responses, which have been implicated in dysregulated lipid synthesis, thereby playing a critical role in the initiation and progression of MGD.36 Lipid species enriched in the reaction pathways included ALA and LA. ALA belongs to the omega-3 (Ω−3) family of polyunsaturated fatty acids, while LA belongs to the omega-6 (Ω−6) series of unsaturated fatty acids, both of which are classified as essential fatty acids capable of regulating and resolving inflammation.37 38 This study still has certain limitations. First, the absence of a sham control precludes exclusion of a placebo effect, especially for subjective outcomes. Second, participants could not be blinded due to the tactile nature of the device, although the outcome assessor remained masked. Third, all clinical assessments were performed by a single investigator without formal intraobserver testing, which may have caused measurement drift but eliminated inter-observer variability. Fourth, subgroup analyses of the meibum microbiome and lipid metabolome were underpowered given the very small sample sizes (WTE, n=4; BTE, n=5); thus, these findings are exploratory and hypothesis-generating. Fifth, the follow-up period is relatively short. Although the study demonstrated the treatment effect at 4 weeks, it did not provide continuous evaluation up to 1-year post-treatment, potentially leading to an incomplete assessment of the long-term efficacy of manual thermal pulsation. Therefore, longer-term follow-up is warranted to evaluate the durability and stability of the treatment. Sixth, the study was conducted during the COVID-19 epidemic, and some patients contracted the novel coronavirus within 4 weeks after treatment. As mentioned earlier, infection with the novel coronavirus may have a certain negative impact on the meibomian glands, thus affecting the evaluation of the treatment effect. Without the influence of this factor, the results of this study might have been more favourable. Despite these limitations, the internal consistency of the data and the non-inferiority design support the validity of our conclusions. In summary, this study demonstrates that manual thermal pulsation is non-inferior to the automatic thermal pulsation in treating MGD over a 4-week period, with a comparable safety profile. The observed variations in treatment response may be linked to interindividual differences in the meibum microbiome and lipid composition. Given its portability, capability for real-time visualisation, and potential for personalised treatment, the manual thermal pulsation could overcome the spatial constraints of conventional thermal pulsation therapy, potentially expanding future treatment options for MGD patients to settings such as home care and primary care clinics. Supplementary material