*2.1. Patients*

This retrospective controlled study was approved by the Institutional Review Boards of Itoh Clinic (approval code: IRIN201903-03), Mizoguchi Eye Clinic, and Ohshima Eye Hospital on March 11th, 2019, and it adhered to the tenets of the Declaration of Helsinki. Patients with refractory ADDE associated with mild MGD who were treated with either IPL and MGX or MGX alone between April and December 2017 at three sites in Japan (Itoh Clinic, Mizoguchi Eye Clinic, and Ohshima Eye Hospital) were enrolled in the study. Informed consent to study participation was obtained from each patient.

Inclusion criteria were as follows: (1) an age of at least 20 years; (2) a diagnosis of ADDE based on the diagnostic criteria for ADDE in Japan [31], which encompass ocular symptoms, a fluorescein tear film breakup time (FBUT) of ≤5 s, a Schirmer's test value of ≤5 mm, and the presence of conjunctival or corneal epithelial damage as evidenced by a fluorescein staining (Fluo) score of ≥1 (on a scale of 0 to 9) according to the van Bijsterveld method [32]; (3) a diagnosis of mild MGD based on the diagnostic criteria for MGD in Japan [33], which encompass ocular symptoms, plugged gland orifices (plugging grade [34] of ≥1, which corresponds to plugging of fewer than three gland orifices with a distribution of less than half of the full length of the lid), vascularity and irregularity of lid margins, reduced meibum expression (meibum grade of ≥2, on a scale of 0 to 3, where 0 = clear meibum easily expressed, 1 = cloudy meibum expressed with mild pressure, 2 = cloudy meibum expressed with more than moderate pressure, and 3 = meibum cannot be expressed even with strong pressure) [35], and a meiboscore of ≥3 (on a scale of 0 to 6) [36]; (4) refractoriness of ADDE as defined by the failure to respond over a period of ≥2 years to at least three types of conventional therapy prescribed in Japan, including tear replacement therapy, tear conservation, and anti-inflammatory eyedrops; and (5) a Fitzpatrick skin type of 1 to 4 based on sun sensitivity and appearance [37]. Exclusion criteria included the presence of active skin lesions, skin cancer, or other specific skin pathology or of active ocular infection or ocular inflammatory disease.

#### *2.2. Experimental Design*

Each patient underwent a series of four IPL-MGX treatment sessions or four sessions of MGX alone at 3-week intervals and was subjected to clinical assessment as described below both before treatment as well as 4 and 12 weeks after the final treatment session. All patients were asked to continue their current ocular medications as well as not to initiate therapy with a new topical or systemic agen<sup>t</sup> for dry eye or MGD during the treatment course.

#### *2.3. Clinical Assessment*

The thickness of the lipid layer of the tear film (LLT), noninvasive breakup time of the tear film (NIBUT), and interferometric fringe pattern of the tear film were determined with a DR-1α tear interferometer (Kowa, Tokyo, Japan) as described previously [38]. Lid margin abnormalities (plugging of meibomian gland orifices and vascularity of lid margins) [34], FBUT, the Fluo score [32], tear meniscus height (TMH) based on fluorescein staining, and meibum grade [35] were evaluated with a slitlamp microscope. For determination of TMH, the center of the lower tear meniscus stained with fluorescein was photographed with a CCD camera attached to the slitlamp microscope, with a magnification of 10 × and under lighting with a blue-free filter. The photographs were examined by an ocular surface expert (R.A.) for semiquantitative grading of TMH as low, normal or high. Morphological changes of meibomian glands were assessed on the basis of the meiboscore [36] as determined by noninvasive meibography. Tear fluid production was measured by Schirmer's test as performed without anesthesia [39]. Symptoms were assessed with the standard patient evaluation of eye dryness (SPEED) validated questionnaire (scale of 0 to 28) [40,41]. VA was also measured with the use of Landolt C charts, and best corrected Landolt VA was converted to logarithm of the minimum angle of resolution (logMAR) VA.

#### *2.4. IPL-MGX Procedure*

Before the first treatment, each patient underwent Fitzpatrick skin typing [37] and the IPL machine (M22; Lumenis, Yokneam, Israel) was adjusted to the appropriate setting (Toyos setting: 590-nm cutoff filter, triple pulses of 6.0 ms with an interval of 50 ms, and total fluence range of 13 to 15 J/cm2). At each treatment session, both eyes of the patient were closed and sealed with IPL-Aid disposable eye shields (Honeywell Safety Products, Smithfield, RI, USA). After generous application of ultrasonic gel to the targeted skin area, each patient received ~13 pulses of light (with slightly overlapping applications) from the right preauricular area, across the cheeks and nose, to the left preauricular area, reaching up to the inferior boundary of the eye shields. This procedure was then repeated in a second pass. Immediately after the IPL treatment, MGX was performed on both upper and lower eyelids of each eye with an Arita Meibomian Gland Compressor (Katena, Denville, NJ, USA). Pain was minimized during MGX by the application of 0.4% oxybuprocaine hydrochloride to each eye. Patients in the control group underwent the MGX procedure alone, without IPL.

#### *2.5. Statistical Analysis*

Data were found to be non-normally distributed with the Shapiro–Wilk test (*p* < 0.05), and nonparametric testing was therefore applied. The Mann–Whitney U test was used to compare numerical variables between the control (MGX alone) and the IPL-MGX groups. The Wilcoxon signed-rank test was used to compare numerical variables between baseline and either 1 or 3 months

after the final treatment session. Fisher's exact test was used to compare categorical variables between the control and IPL-MGX groups. The chi-square test was used to compare lipid layer grade and TMH between before and either 1 or 3 months after the final treatment session. Adjusted *p* values were calculated by multiplication of obtained *p* values by the number of comparisons in Bonferroni's correction. The outcome variables of the study were the SPEED score and NIBUT before and after treatment. We performed a statistical power analysis for both the SPEED score and NIBUT at 3 months after the final treatment session in the control and IPL-MGX groups. The power (1-β) was >0.90 at the level of α = 0.025, and the sample size was sufficient. Statistical analysis was performed with JMP Pro version 15 software (SAS, Cary, NC, USA). All statistical tests were two sided, and a *p* value of <0.05 was considered statistically significant.
