The Ocular Surface Randomised double-masked placebo-controlled trial of the cumulative treatment e ﬃ cacy proﬁle of intense pulsed light therapy for meibomian gland dysfunction

A B S T R A Purpose: To assess long-term cumulative treatment e ﬀ ects of intense pulsed light (IPL) therapy in meibomian gland dysfunction (MGD). Methods: Eighty-seven symptomatic participants (58 female, mean ± SD age, 53 ± 16 years) with clinical signs of MGD were enrolled in a prospective, double-masked, parallel-group, randomised, placebo-controlled trial. Participants were randomised to receive either four or ﬁve homogeneously sequenced light pulses or placebo treatment to both eyes, (E-Eye Intense Regulated Pulsed Light, E-Swin, France). Visual acuity, dry eye symptomology, tear ﬁlm parameters, and ocular surface characteristics were assessed immediately before treatment on days 0, 15, 45, 75, and four weeks after treatment course completion on day 105. Inﬂammatory and goblet cell function marker expression, and eyelid swab microbiology cultures were evaluated at baseline and day 105. Results: Signiﬁcant decreases in OSDI, SPEED, and SANDE symptomology scores, and meibomian gland capping, accompanied by increased tear ﬁlm lipid layer thickness, and inhibited Corynebacterium macginleyi growth were observed in both treatment groups (all p < 0.05). Sustained clinical improvements occurred in both treatment groups from day 75, although signiﬁcant changes from day 45, in lipid layer quality, meibomian gland capping, OSDI and SANDE symptomology, were limited to the ﬁve-ﬂash group (all p < 0.05). Conclusions: IPL therapy e ﬀ ected signiﬁcant improvements in dry eye symptomology, tear ﬁlm lipid layer thickness, and meibomian gland capping in MGD patients. Five-ﬂash IPL treatment showed superior clinical e ﬃ cacy to four-ﬂash, and an initial course of at least four treatments is suggested to allow for establishment of sustained cumulative therapeutic beneﬁts prior to evaluation of overall treatment e ﬃ cacy. Trial registration number: ACTRN12616000667415.


Introduction
Evaporative disease is recognised to be the most common dry eye etiological subtype [1], and can be associated with profound impacts on ocular comfort, visual function, quality of life, and work productivity [1][2][3][4]. The condition is frequently caused by underlying meibomian gland dysfunction (MGD), whereby the increased viscosity and melting points of gland secretions can predispose towards obstruction and inflammation of the ductal system [5,6]. The consequent reduction in the quality and quantity of meibomian lipids delivered to the tear film compromises the integrity of the surface lipid layer, triggering a selfperpetuating vicious circle of tear film hyper-evaporation, instability, hyperosmolarity, and inflammation [5,7].
A large number of treatments are currently available for MGD, including warm compress therapy, eyelid hygiene regimens, mechanical meibum expression, lipid-containing artificial tear supplements, and omega-3 fatty acid supplementation [8,9]. In addition, intraductal probing, tetracyclines, antibiotic, anti-inflammatory and immunomodulatory agents may also be considered judiciously in more severe and refractory cases [8,9]. Nevertheless, adequate symptomatic control is frequently difficult to achieve or sustain, highlighting the ongoing need for the development of alternative management options [8].
The purpose of the current double-masked randomised controlled trial was to further characterise the long-term cumulative treatment effects of IPL therapy in MGD patients, through the clinical assessment of dry eye signs and symptoms, and the laboratory analysis of ocular surface microbiological profiles and cytology markers.

Subjects
This prospective, fifteen-week, double-masked, parallel-group, randomised, placebo-controlled trial, adhered to the tenets of the Declaration of Helsinki, and was approved by the University of Auckland Human Participants Ethics Committee (UAHPEC 017173), and prospectively registered as a clinical trial (ACTRN12616000667415). Participants were required to be 18 years or older, with symptoms of dry eye disease (McMonnies dry eye questionnaire score ≥10 and/or Ocular Surface Disease Index score ≥13) [22] and clinically significant signs of MGD (eyelid margin or mucocutaneous junction abnormalities, meibomian gland orifice capping, and/or decreased expressed meibum quality) [23,24], and no contact lens wear or use of systemic medications known to affect the eye two weeks prior to baseline assessment or during the treatment period. In addition, eligibility required participants to be non-pregnant; report no history of major systemic, dermatologic or ocular conditions; no ocular surgery or dermatologic treatments in the previous three months or during the treatment period; no implants, tattoos, semi-permanent makeup, or pigmented lesions in the treatment area; and no contraindications to IPL therapy, including the use of photosensitive medications. Eligible participants were enrolled after providing written consent.
A total of 87 eligible participants were recruited, exceeding the sample size requirements for the desired study power. Power calculations were conducted with non-invasive tear film breakup time as the designated outcome, and showed that a minimum of 25 participants was required in each of the three treatment groups (a total of 75 participants), to detect a clinically significant difference of 3-4 s in pairwise comparisons, at 80% power (β = 0.2) and a two-sided statistical significance level of 5% (α = 0.05), with the SD of normal values being estimated to be approximately 4-6 s [25]. Sample size estimates were determined using a uniform non-parametric adjustment, with NCSS PASS 2002 (Utah, USA).

Treatments
Participants were randomly assigned to one of three treatment groups, and underwent IPL therapy with four or five homogeneously sequenced light pulses (E-Eye Intense Regulated Pulsed Light; E-Swin, Paris, France) or placebo treatment to both eyes, applied during inoffice visits on days 0, 15, 45, and 75, by a trained unmasked clinician, who was not involved in study data collection. Randomisation was conducted by computer-generated random number allocation, and applied to sequentially enrolled participants. The randomisation schedule was pre-determined, prior to commencing participant recruitment, such that the investigator involved in baseline participant assessment had no involvement in treatment allocation. During each visit, participants were fitted with opaque, metal goggles covering both eyes to protect the globes, with clear conducting gel applied to the inferior, lateral, and medial aspects of the goggles, as per manufacturer recommendations. Light pulses were delivered to four overlapping periocular zones inferior to each eye, and the fifth pulse was applied temporally adjacent to the lateral canthus in those randomised to the five pulse group (Fig. 1). Pulse intensity ranged from 9 to 13 J/cm 2 and was inversely related to the Fitzpatrick skin phototype classification of the participant (Table 1) [20]. Participants allocated to the placebo group underwent sham treatment, and participant masking was achieved using an identical device with a non-illuminating handpiece applied to the periocular area, while an active piece was directed away from the participant towards the corner of the room to imitate the illumination and sounds of the IPL device in order to simulate treatment. For the purpose of characterising and isolating the cumulative treatment effects of IPL therapy, mechanical meibum expression was not conducted during the study period, and antibiotic or anti-inflammatory treatment was not prescribed.

Measurements
The investigator conducting clinical and laboratory measurements was masked to treatment randomisation. Participants were assessed at a single site, with a mean ± SD room temperature of 20.3 ± 0.5°C and Fig. 1. Intense pulsed light therapy was delivered to four overlapping periocular zones inferior to each eye (panel A), and the fifth pulse was applied temporally adjacent to the lateral canthus in those randomised to the five pulse group (panel B). a mean ± SD relative humidity of 62.5 ± 6.8%, and ocular measurements were conducted on the right eye of each participant. Clinical and laboratory measurements were conducted in accordance with the recommendations of the TFOS DEWS II Diagnostic Methodology subcommittee [22], and performed immediately before treatment application on days 0, 15, 45, and 75, and four weeks following the completion of the treatment course, on day 105, in order to characterise the longer term cumulative treatment effects. To minimise the impact on ocular surface and tear film physiology for subsequent tests, clinical and laboratory measurements were performed in ascending order of invasiveness [22], as summarised in Table 2.
Six-metre best spectacle-corrected logMAR visual acuity was recorded as a safety measure. The McMonnies dry eye questionnaire was administered to screen for dry eye symptoms at baseline, while the Ocular Surface Disease Index (OSDI), Standard Patient Evaluation of Eye Dryness (SPEED), and Symptom Assessment in Dry Eye (SANDE) questionnaires were the instruments administered for the purpose of comparing symptomology across the treatment period. The overall SANDE score was calculated as the geometric mean of the frequency and severity scores [26]. Participants were advised to contact the study investigators during the study period to report adverse events at any time.
Bulbar conjunctival hyperemia, tear meniscus height, non-invasive tear film breakup time, and tear film lipid layer grade were assessed using the Keratograph 5M (Oculus Optikgeräte GmbH, Wetzlar, Germany). Bulbar conjunctival hyperemia was evaluated by automated objective evaluation of high magnification digital imaging, using the proprietary JENVIS grading scale from 0 to 4 [27]. The lower tear meniscus height was assessed using high magnification pre-calibrated digital imaging, and three measurements near the center of the lower meniscus were averaged. Non-invasive tear film breakup time was measured using automated detection of first break-up, while the subject maintained fixation and was requested to refrain from blinking. Three breakup time readings were averaged in each case [22]. Tear film lipid layer interferometry was graded according to the modified Guillon-Keeler system: grade 1, open meshwork; grade 2, closed meshwork; grade 3, wave or flow; grade 4, amorphous; grade 5, colored fringes; grade 0, non-continuous layer (non-visible or abnormal colored fringes) [28,29].
Tear film osmolarity measurements were performed, in-office, with a clinical osmometer (TearLab, California, USA), from 50 nL of tears sampled from the lower lateral canthal tear meniscus. A measurement was taken for each eye, and the higher reading and the inter-ocular difference recorded [22].
Central corneal and inferior eyelid margin sensitivity were assessed using non-contact air-jet aesthesiometry (NCCA, SDZ electronics, New Zealand) to evaluate potential functional changes in the peripheral nerve supply of regions local to IPL application [30]. An intermittent, barely susceptible flow of air was used to determine threshold sensitivity via a forced-choice double-staircase method [31]. Sensitivity thresholds were measured in a quiet room devoid of distractions, using a 0.9 s stimulus duration and a standardised 10 mm working distance from the ocular region assessed. Measurements were conducted at the geometric centre of the cornea, and at the lid wiper zone of the central inferior eyelid margin during slight lower eyelid eversion. Participants were instructed to blink frequently and the inferior eyelid margin was released to normal position between stimulus presentations, in order to avoid excessive ocular surface drying and subsequent dampening of the sensitivity threshold, and to minimise disruption to subsequent measurements [31,32].
Lid parallel conjunctival folds (LIPCOF) were graded, and sodium fluorescein and lissamine green dyes were applied using the recommended technique described in TFOS DEWS II Diagnostic Methodology report [22], in order to evaluate localised corneal and conjunctival areas of epithelial desiccation. Staining was recorded using the modified Oxford grading scheme [33], and lid wiper epitheliopathy (LWE) was evaluated relative to Korb's grading [34].
Expressibility of the inferior eyelid meibomian glands was assessed with the Meibomian Gland Evaluator (TearScience, North Carolina, USA), with a pressure of 1.2 g/mm 2 applied immediately inferior to the lash line, at the nasal, central, and temporal aspects of the eyelid margin. The number of meibomian gland orifices yielding lipid secretion was graded on a five-point scale: 0, more than 75%; 1, 50% to 75%; 2, 25% to 50%; 3, less than 25%; 4, none. The quality of expressed meibum was graded as: grade 0, clear fluid; grade 1, slightly turbid; Table 2 Order of clinical and laboratory measurements conducted during the study period. Measurements were performed immediately before treatment application on days 0, 15, 45, and 75, and four weeks following the completion of the treatment course on day 105.  [29]. Infrared meibography was imaged with the Oculus Keratograph 5M, with the superior and inferior eyelids everted in turn. From the captured image, the proportion of meibomian glands visible within the tarsal area was graded according to the five-point Meiboscale [35].
In vivo confocal microscopy evaluation of the central cornea and inferior eyelid margin was performed using the Heidelberg Retinal Tomograph (HRT) III with Anterior Segment Module (Heidelberg Engineering GmbH, Germany), following instillation of a drop of 0.4% benoxinate hydrochloride into the conjunctival fornix. The objective lens was covered by a disposable polymethacrylate sterile cap (Tomo-Cap, Heidelberg Engineering GmbH, Germany), and Viscotears (Carbomer 980, 0.2%; Novartis, North Ryde, NSW, Australia) was applied as the coupling agent. Participants were requested to fix their gaze on a central target to allow for full thickness scanning of the central cornea in 2 μm increments using the Section Mode setting of the tomograph. Meibomian gland imaging was conducted with slight eversion of the inferior eyelid margin, with the Tomo-Cap positioned perpendicularly to the central, nasal, and temporal thirds of the eyelid margin, close to the mucocutaneous junction. For each measurement, three non-overlapping, high-resolution images (400 μm x 400 μm frame) were analysed using Image J software with the NeuronJ plug-in (National Institutes of Health, USA). Central cornea sub-basal nerve fibre density was assessed by measuring the total corneal nerve length per square millimetre [36], and dendritic cells were quantified as cellular density per unit area [37]. Inferior eyelid margin rete ridges per square millimetre were evaluated, and the diameter measured along the longest axis [38][39][40]. Meibum secretion reflectivity was graded according to a 4-point scale developed by Villani et al.: grade 1, black; grade 2, dark grey; grade 3, light grey; grade 4, white [41].
Microbiological swabs from the inferior eyelid margin were collected using a sterile cotton-tipped applicator moistened with buffered saline, and placed immediately into Amies transport medium (Fort Richard Laboratories, Auckland, NZ) and processed on the same day. Anaerobic and aerobic microbiological evaluation was performed by a dedicated, independent laboratory (LabPlus, Auckland, NZ), with the total number of colony forming units (CFUs) enumerated for each cultured sample, and the load of each of the identified microbial colonies was graded on an ordinal scale: grade 0, no growth; grade 1, single colony; grade 2, few colonies; grade 3, light growth; grade 4, moderate growth; grade 5 heavy growth.
Ocular Demodex load was assessed by epilating four eyelashes from the upper eyelids under slit lamp examination. Lashes were gently grasped with fine forceps close to the base and rotated for 20 s before epilation [42]. The epilated lashes were placed onto glass slides and examined under light microscopy at 200 times magnification [43]. Adult Demodex mites were identified morphologically, and the mite count from each of the four lashes was averaged.
Conjunctival impression cytology was conducted following topical anesthesia with one drop of 0.4% benoxinate hydrochloride. Bulbar conjunctival cells from the inferior temporal ocular surface were collected with the EYEPRIM™ conjunctival impression device (OPIA, France) [44]. Conjunctival cell sample RNA extraction and purification was performed with PureLink® RNA Mini Kit (Invitrogen™ by Life Technologies), and tested for the presence of inhibitors before undergoing cDNA synthesis using SuperScript™ IV VILO™ Master Mix (In-vitrogen™ by Life Technologies). A standard β-actin PCR and gel electrophoresis was conducted on the synthesized samples to confirm successful cDNA synthesis [44,45]. Seven reference genes (Beta-Actin, HPRT1, B2M, PPIA, TBP, GUSB, RPLP0 and POLR2A) were tested amongst the sample population, with the combination of GUSB and RPLP0 producing the best stability value according to the Normfinder algorithm (MOMA, Aarhus, Denmark). The geometric mean of GUSB and RPLP0 data was subsequently used for normalisation and relative quantification of the target genes (MMP-9, IL-6 and MUC5AC) [44][45][46]. The Minimum Information for Publication of Quantitative Real-Time PCR Experiments (MIQE) guidelines were followed to ensure validity of the qPCR experiments [47], which were set up using QiAgility® PCR robot (Qiagen, Germany) with PrimeTime® Assays (Integrated DNA Technologies), and internal calibrators were used to enable the removal of inter-run variations [48].

Statistics
Statistical analysis was conducted with Graph Pad Prism version 8.01 (California, USA) and IBM SPSS version 24 (New York, USA). Twoway mixed model analysis of variance (ANOVA) testing was performed to test the significance of treatment, time and interaction (treatmentby-time) effects on measurements over the fifteen-week period, where continuous variables with a normal distribution had been confirmed (Kolmogorov-Smirnov test p > 0.05). Non-normally distributed continuous measures were logarithmically transformed prior to undergoing analysis. Post hoc analysis for the significance of treatment effects at each time point was conducted using the multiplicity-adjusted Tukey's test. Analysis of ordinal data was performed using multiple ordinal regression, with post hoc analysis of treatment effects at each time point conducted using the multiplicity-adjusted, non-parametric Dunn's test. Categorical data at baseline were analysed using chi-squared or Fisher's exact tests. All tests were two tailed, and p < 0.05 was considered significant. Data are presented as mean ± SD, or median (IQR) unless otherwise stated.

Results
The mean ± SD age of the 87 participants (58 females, 29 males) was 53 ± 16 years (range, 21-85 years). Baseline characteristics, clinical and laboratory measurements of participants during the fifteenweek study period are presented by treatment group in Table 3 to 6. Baseline measurements did not differ between treatment groups (all p > 0.05; Table 3).

Visual acuity and adverse events
There were no significant treatment, time, or treatment-by-time interaction effects for best-corrected visual acuity (all p > 0.20, Table 5). No adverse events were reported during the study period. Table 3 Baseline characteristics of participants randomised to placebo treatment, or intense pulse light therapy with four or five homogeneously sequenced light pulses. Data are presented as mean ± SD, median (IQR), or number of subjects (% of subjects).

Dry eye symptomology
Two-way ANOVA demonstrated significant treatment effects for OSDI, SPEED and SANDE dry eye symptomology scores (all p ≤ 0.001, Table 5), and significant time effects for SPEED and SANDE scores (both p < 0.001, Table 5). Multiplicity-adjusted post-hoc testing showed that participants receiving four flashes of IPL demonstrated transient improvements in OSDI and SPEED scores relative to those in the placebo group on day 15 (both p < 0.05, Table 6), and then sustained improvements in OSDI, SPEED, and SANDE scores from day 75 onwards (all p < 0.05, Table 6 and Fig. 2). Participants in the five-flash IPL group exhibited significantly lower OSDI and SANDE scores from day 15 onwards, although sustained reductions in the SPEED score did not occur until day 75 onwards (all p < 0.05, Table 6).

Tear film quality
A significant treatment effect was detected for tear film lipid layer grade (p = 0.01, Table 5). Post-hoc multiplicity-adjusted analysis demonstrated enhanced tear film lipid layer quality in the four-flash IPL group from day 75 onwards, and improvements were observed in the five-flash IPL group from day 45 onwards (all p < 0.05, Table 6 and Fig. 3). Treatment, time, and interaction effects for tear meniscus height, non-invasive tear film stability, and tear osmolarity were not statistically significant (all p > 0.05, Table 5).

Ocular surface characteristics
Treatment and time effects were significant for meibomian gland  capping grade (both p < 0.05, Table 5). Multiplicity-adjusted post-hoc analysis demonstrated significant reductions in meibomian gland capping severity in both IPL treatment groups on day 105, although improvements were limited to the five-flash IPL group on day 45 (all p < 0.05, Table 6. No significant treatment, time, or interaction effects were detected for conjunctival hyperaemia, eyelid margin and eyelash characteristics, ocular surface staining, meibomian gland dropout, meibum quality, and non-contact aesthesiometry (all p > 0.05, Table 5).

In vivo confocal microscopy evaluation
There were no significant treatment, time, and interaction effects for corneal sub-basal nerve fibre and dendritic cell densities, and inferior lid margin rete ridge diameter, density and secretion reflectivity (all p > 0.05, Table 5).

Eyelid margin swab microbiology
A significant treatment effect for Corynebacterium macginleyi growth was observed (p = 0.003, Table 5), with post-hoc multiplicity-adjusted analysis demonstrating inhibited growth in both IPL treatment groups on day 105 (both p < 0.05, Table 6). No significant treatment, time, or interaction effects were detected for total bacterial colony forming units, and the growth of all other bacterial species (all p > 0.05, Table 5).

Conjunctival impression cytology markers
Treatment, time, and interaction effects for ocular surface inflammation and goblet cell function markers were not statistically significant (all p > 0.05, Table 5).

Discussion
In agreement with the findings reported in earlier studies [11][12][13][14][15][16][17][18][19][20], the results of the current double-masked, randomised, placebo-controlled trial demonstrated clinical efficacy of intense pulsed light therapy in the treatment of patients with MGD. Clinical improvements in objective and subjective markers of ocular surface homeostasis were observed during the fifteen-week period in both IPL treatment groups, with significant decreases in OSDI, SPEED, and SANDE symptomology scores, and meibomian gland capping, which was accompanied by augmentation of tear film lipid layer thickness. In addition, Corynebacterium macginleyi growth appeared to be inhibited following treatment courses with both four and five-flash IPL therapy. It is, nevertheless, acknowledged that the treatment effects observed in the current study appear to be more modest than those previously reported in the literature [12][13][14][15][16][17]19,20]. Not unexpected, this is thought likely to be related to the intrinsic methodological design of the current trial, including the lack of mechanical meibum expression immediately following IPL therapy, and the evaluation of outcome measures either two weeks following the first treatment or four weeks following allother treatments, which was intentional, in order to isolate and provide better characterisation of long term the extended cumulative treatment effects of IPL therapy.  The mechanisms by which IPL therapy effects clinical improvements in patients with MGD remains poorly understood, although a number of different hypotheses have been proposed [11,13,18,21,49]. Thermal energy transferred by IPL therapy is thought to liquefy the inspissated meibum observed in MGD, relieving ductal obstruction and promoting the release meibomian lipids into the tear film [8,9,11,21,49]. Restoration of the integrity and quality of the surface lipid layer can enhance tear film stability [8,9], and it has been recognised that a continuous lipid layer is necessary to retard excessive aqueous tear evaporation [50]. This hypothesis appears to be supported by the trends observed in the current study which demonstrate an improvement in tear film lipid layer thickness and meibomian gland capping following treatment with IPL therapy. Interestingly, despite a significant reduction of subjective dry eye symptomology scores being observed in association with improvements in markers of meibomian gland function, no significant changes in non-invasive tear film stability were detected in the current study, which contrasts with trends described in earlier reports [12][13][14][15][16][17]19,20]. It cannot be reliably determined whether this might be partially attributed to the more modest treatment effects in the absence of mechanical meibum expression immediately following IPL therapy in the current study. In addition, the measurement of outcome measures were conducted two or four weeks following each course of treatment, and may have failed to capture immediate or more transient treatment effects, especially in the context of the intrinsic variability of tear film stability measurements [22,51]. Finally, the possibility for the four IPL treatments during the fifteen-week study period to be insufficient to effect sustained cumulative improvements in tear film stability cannot be excluded, and would warrant investigation in future studies with a greater number of treatments.
It has also been previously suggested that IPL therapy might potentially decrease the bacterial load of the peri-ocular micro-environment and alter the composition of the ocular surface microbiota, thereby dampening triggers for host immune and inflammatory responses [21,49]. In the current study, inhibition of Corynebacterium macginleyi growth was observed with IPL treatment. Other mechanisms that have been previously raised include the therapeutic effects mediated by the thrombosis of abnormal blood vessels in the peri-ocular skin, reduction of epithelial turnover, fibroblast activation and promotion of collagen synthesis, reduction in ocular Demodex load, modulation of pro-inflammatory and anti-inflammatory cascades, and alteration in the levels of reactive oxidative species [13,18,49,52], although evidence supporting these hypotheses were either not directly investigated or observed in the findings of the current study.
The trends observed in the current study are generally supportive of the manufacturer recommendations of applying five flashes of IPL during each treatment. Sustained improvements in clinical signs and symptoms of MGD were detected earlier in the study period in participants randomised to receiving five flashes of IPL than those receiving four flashes. Indeed, on day 45, clinical improvements in tear film lipid layer thickness, meibomian gland capping, OSDI and SANDE symptomology scores were limited to those receiving five flashes. Although the mechanisms of improved treatment efficacy associated with the fifth flash applied temporally adjacent to the lateral canthus are not fully understood, it is possible that the enhanced transfer of thermal energy to the eyelids might have potentially contributed [21,49]. Possible neuromodulatory effects on the parasympathetic innervation of the meibomian glands originating from the pterygopalantine ganglion have also been hypothesised [53,54]. Interestingly, although significant reductions in OSDI scores were observed in both treatment groups on day 15, changes in SPEED scores were limited to those receiving four flashes, while improvements in SANDE scores were limited to those receiving five flashes. However, the initial changes in OSDI and SPEED scores were not sustained on day 45 in participants receiving four flashes, which contrasted with the continued improvements in OSDI and SANDE scores observed in participants receiving five flashes. These trends would suggest that the initial treatment effects of IPL therapy with four flashes might be more short-lived than those with five flashes. The contrasting trends observed in the three symptomology scores during the study period might also partially reflect the differing diagnostic sensitivity of these subjective measurements [55].
The methodological design of the current trial was performed to assess the cumulative profile of the treatment effects of IPL therapy. Although inconsistent changes in a number of subjective and objective ocular surface parameters were observed on days 15 and 45, sustained improvements in both IPL treatment groups for OSDI, SPEED, and SANDE symptomology scores, and tear film lipid layer thickness were consistently observed on days 75 and 105. Although an initial improvement in meibomian gland capping was observed on day 45 in participants receiving five flashes, a consistent improvement observed across both IPL treatment groups was not detected until day 105. Overall, these trends would suggest that an initial course of four treatments of IPL therapy is warranted in the clinical setting, to allow sufficient time for a sustained cumulative therapeutic effect to be established, before the evaluation of the treatment efficacy can be reliably Table 6 Post-hoc multiplicity-adjusted Tukey's test for treatment effects at each individual time point. Ordinal data were converted to rank-values prior to non-parametric assessment. Data are presented as p-values. Asterisks denote statistically significant differences (p < 0.05).  conducted in an individual patient. Future studies with longer treatment periods are required to confirm whether further extended treatment courses might confer an additional advantage through the clinical improvement of other signs of meibomian gland dysfunction.

Conclusions
In conclusion, IPL therapy effected improvements in dry eye symptomology, tear film lipid layer thickness, and meibomian gland capping in MGD patients in this double-masked, randomised, placebocontrolled trial. The findings also demonstrated superior clinical efficacy of applying five flashes of IPL during each treatment relative to four flashes, and would suggest that an initial course of four treatments would be required to allow for sustained cumulative therapeutic effects to be established, prior to the evaluation of overall treatment efficacy.

Funding
The authors are grateful to the New Zealand Optometric Vision Research Foundation, and to E-Swin, France, for grants-in-aid to support this investigator-initiated trial. AX was a recipient of a University of Auckland Doctoral Scholarship and a New Zealand Association of Optometrists Postgraduate Scholarship. The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Declaration of competing interest
The authors have no commercial or proprietary interest in any concept or product described in this article.