Introduction

Meibomian gland dysfunction (MGD) is the most common cause of dry eye disease (DED) [1, 2]. Chronic terminal duct obstruction and secretion abnormalities of the meibomian glands disrupt the balance of tear film owing to the lack of a lipid layer [3]. Currently, the conventional treatments available for DED include meibomian gland expression (MGX), warm compresses, and ocular lubricants [1]. Intense pulsed light (IPL), a broad-spectrum, non-coherent, and polychromatic light source with a wavelength of 500 to 1200 nm [4], is commonly used by dermatologists for hypertrichosis, telangiectasia, benign cavernous hemangioma, port-wine stain, and pigmented lesions [5]. In addition, IPL was recently reported to improve DED symptoms while treating subjects with rosacea [6]. IPL and MGX seem to have far-reaching application prospects, according to several studies [6,7,8,9].

Although several systematic reviews have shown the effectiveness of IPL applied in MGD [10,11,12,13], the quality and quantity of available articles were limited at the time when the reviews were written. High-quality data are needed, and such data are now available with the publication of results from some relevant randomized controlled trials in recent years. Therefore, the purpose of this systematic review and meta-analysis was to explore the efficacy of IPL with or without traditional therapies in the management of DED caused by MGD.

Material and methods

Protocol and registration

This systematic review was conducted and reported in accordance with the Cochrane handbook and the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA). The registration number in PROSPERO is CRD42022309060 (www.crd.york.ac.uk/prospero).

Eligibility criteria

The criteria for this review were based upon the PICOS approach as follows:

Participants (P)

Adults with a diagnosis of MGD and a Fitzpatrick skin type of 1–5 according to sun sensitivity and appearance of the skin, with no contraindications to IPL treatment.

Intervention (I)

IPL with or without conventional treatments (MGX, eye fumigation, warm compresses, and home-based therapy).

Comparison (C)

Other conventional therapies, such as MGX, eye fumigation, warm compresses, and home-based therapy (ocular lubricants and so on).

Outcome (O)

The primary outcomes were changes in tear break-up time (BUT) and changes in Ocular Surface Disease Index (OSDI) scores. The secondary outcomes were changes in non-invasively measured tear break‐up time (NIBUT), corneal and conjunctival fluorescein staining (CFS) scores, and Standard Patient Evaluation of Eye Dryness (SPEED) scores.

Study design

Parallel design randomized controlled trials (RCTs).

Exclusion criteria

Studies were excluded if any of the following criteria was applied:

  • Animal studies, case reports, case-series reports, and literature reviews.

  • Conference abstracts and published clinical trial protocols whose authors could not successfully be contacted for the data.

  • Lack of a control group or self-controlled study design.

  • DED arising from causes other than MGD.

Search strategy

We performed a systematic and thorough search of the following databases (up to 31 January 2022): Web of Science, Cochrane Central Register of Controlled Trials (CENTRAL), PubMed, EMBASE, China National Knowledge Infrastructure (CNKI), VIP, Wanfang Database, and China Biology Medicine (CBM). No restrictions were imposed on language or publication year. The search terms used in PubMed were as follows: (((intense pulsed light therapy [Title/Abstract]) OR (Intense pulsed light [Title/Abstract])) OR (IPL [Title/Abstract])) AND (((“Meibomian gland dysfunction” [Mesh]) OR (((((Meibomian gland dysfunctions [Title/Abstract]) OR (meibomian gland expression [Title/Abstract])) OR (MG dysfunction [Title/Abstract])) OR (MGX [Title/Abstract])) OR (MGD [Title/Abstract]))) OR (((((Dry eye [Title/Abstract]) OR (Dry eye disease [Title/Abstract])) OR (dry eye syndrome [Title/Abstract])) OR (keratoconjunctivitis sicca [Title/Abstract])) OR (keratitis sicca [Title/Abstract]))).

Data extraction

We used Endnote X9 for literature screening and elimination of duplicate citations. Two independent researchers (Y. H. L. and J. P.) searched the articles, checked for duplicates, and assessed the eligibility of the studies based on their titles and abstracts. Any discrepancy was resolved by consulting two experienced researchers (J. Y. L. and J. X. Z.). Studies meeting the criteria for inclusion were downloaded in full-text form, and the two researchers who initially screened the papers (Y. H. L. and J. P.) reviewed them further.

We extracted changes in outcomes from baseline to the last treatment. For some included studies reporting only endpoint outcomes, we obtained the changes by subtracting the postintervention mean from the baseline mean according to chapter 6 of the Cochrane handbook[14]. When needed, missing information and clarification about the statistics presented were sought from the authors of the included studies. The following data were extracted from the included studies: year of publication; country; sample size; patient age; sex ratio; machine; treatment parameters; follow-up duration; and changes in BUT, OSDI, NIBUT, CFS, and SPEED.

Risk of bias in individual studies

We assessed the risk of bias of RCTs, according to the Cochrane Collaboration Risk of Bias tool version 2 (RoB 2). Bias arising from the randomization process, intended interventions, missing outcome data, measurement of the outcome, and selection of the reported result were assessed.

Data analysis

Data analysis was performed using Review Manager 5.4 and Stata 17.0. The investigated outcomes for continuous outcome were expressed as the mean difference (MD) with 95% confidence interval (CI). Heterogeneity was measured using the I2 test. A fixed or random effect model was chosen depending on the heterogeneity. IPL with and without conventional treatment in the experimental group was statistically evaluated overall and in separate subgroup analyses; the threshold for statistical significance was set to P < 0.05. The results of the analyses are presented graphically with forest plots. The study designs, methodologies, participants, and the period of follow-up time composed the clinical heterogeneity of the included studies. Sensitivity analysis was performed by drawing sensitivity plots to define the influence of specific studies on the total calculated effect. Funnel plots and Egger’s test were used to evaluate publication bias.

Results

Data analysis

The initial search identified a total of 1842 studies, including 71 in CENTRAL, 108 in PubMed, 119 in EMBASE, 179 in Web of Science, 529 in CNKI, 681 in Wanfang Database, 135 in VIP, and 20 in CBM. After the removal of 774 duplicates, 1068 articles were evaluated by titles and abstracts, and 950 were excluded due to an inappropriate study type irrelevant research purpose, leaving 118 full-text articles to be read in full. Subsequently, we excluded 107 papers due to a lack of important outcomes or a suitable control group. Ultimately, 11 articles met the eligibility criteria. A flowchart of the screening process is presented in Fig. 1. The characteristics of the studies are listed in Table 1.

Fig. 1
figure 1

Flowchart of the screening process

Full size image
Table 1 Data extraction from the included studies
Full size table

Risk of bias

The quality assessment results are shown in Fig. 2. Five studies were scored as having a low risk of bias in the randomization process [15,16,17,18,19], and the other studies[20,21,22,23,24,25] raised some concerns. Bias owing to deviation from intended interventions was rated as a low risk in 9 studies [15,16,17,18,19,20, 22, 23, 25], while the remaining 2 studies were deemed to raise some concerns [21, 24]. One study [20] had a high risk of missing outcome data, and the others [15,16,17,18,19, 21,22,23,24,25] had a low risk in this aspect. Three studies [15, 18, 25] were scored as having a low risk of bias arising from measurement of the outcome, and one study [21] had a high risk. The rest of the included studies raised some concerns [16, 17, 19, 20, 22,23,24]. Three studies [15, 17, 18] were considered to have a low risk of bias from selective reporting of results, and the remaining eight included studies raised some concerns [16, 19,20,21,22,23,24,25].

Fig. 2
figure 2

Risk of bias summary of the included studies (in this chart, R represents Randomization process; D represents Deviations from intended interventions; Mi represents Missing outcome data; Me represents Measurement of the outcome; S represents Selection of the reported result; and O represents Overall)

Full size image

Primary outcome measure

All included studies reported an improvement in clinical evaluation indexes in both the IPL and control groups, while only seven of them [15, 17, 19,20,21, 24, 25] reported that the IPL group had better outcomes than the control group in some indexes.

Change in the OSDI

Eight studies involving 384 participants in the experimental arm and 374 participants in the control arm reported changes in OSDI scores (Fig. 3). In the total analysis, significant heterogeneity was observed among the included studies (P = 0.003; I2 = 98%). We used a random-effects model to summarize the mean effect size and performed subgroup analysis according to IPL with or without conventional treatments in the experimental group. The change in OSDI in the total analysis result (MD =  − 7.49; 95% CI =  − 12.47 to − 2.50; P = 0.003) indicated that the experimental group performed significantly better, which is similar to the 1.1.1 subtotal outcome as designated in Fig. 3 (MD =  − 7.01; 95% CI =  − 12.23 to − 1.80; P = 0.008). The results of the 1.1.2 subtotal analysis showed that there was no significant difference between the two groups (MD =  − 8.34; 95% CI =  − 23.62 to 6.93; P = 0.28). Sensitivity analysis indicated that the results were stable. Egger’s test suggested no publication bias, also proving the stability of the results.

Fig. 3
figure 3

Forest plot of change in OSDI

Full size image

Change in BUT

Eight studies involving 425 participants in the experimental arm and 418 participants in the control arm reported changes in the BUT (Fig. 4). Significant heterogeneity was observed between the studies (P < 0.00001; I2 = 96%); therefore, a random-effects model was used to summarize the mean effect size, and we performed subgroup analysis according to IPL with or without conventional treatment in the experimental group. The change in BUT in the total analysis result (MD = 1.94; 95% CI = 1.19 to 2.69; P < 0.00001) indicated that the experimental group performed significantly better than the control group, which is similar to the 1.2.1 and 1.2.2 subgroup outcomes, as designated in Fig. 4 (MD = 2.10; 95% CI = 0.96 to 3.24; P = 0.0003 and MD = 1.49; 95% CI = 0.19 to 2.79; P = 0.02).

Fig. 4
figure 4

Forest plot of change in BUT

Full size image

Secondary outcome measures

Change in NIBUT

Five studies involving 390 participants in the experimental arm and 369 participants in the control arm reported changes in the NIBUT (Fig. 5a). We identified heterogeneity across the included studies (P < 0.00001; I2 = 96%), and we used a random-effects model to summarize the mean effect size. The change in NIBUT in the total analysis result (MD = 2.55; 95% CI = 1.07 to 4.04; P < 0.00001) indicated that the experimental group had significantly better outcomes than the control group.

Fig. 5
figure 5

Forest plots of changes in secondary outcome measures (a represents change in NIBUT; b represents change in CFS; c represents change in SPEED)

Full size image

Change in CFS

Five studies involving 251 participants in the experimental arm and 247 participants in the control arm reported changes in CFS (Fig. 5b). We identified significant heterogeneity across the included studies (P < 0.00001; I2 = 95%), and a random-effects model was used. The change in CFS in the total analysis result (MD =  − 0.41; 95% CI =  − 1.19 to 0.36; P = 0.29) showed no statistical significance between the two groups.

Change in SPEED

Four studies involving 140 participants in the experimental arm and 135 participants in the control arm reported changes in SPEED scores (Fig. 5c). We identified heterogeneity across the included studies (P = 0.003; I2 = 78%) and used a random-effects model. The change in SPEED in the total analysis result (MD =  − 3.28; 95% CI =  − 5.64 to − 0.93; P = 0.006) indicated that the experimental group had significantly better outcomes than the control group.

Safety data

Eight studies [15,16,17,18,19,20, 22, 25] reported safety data, and none of these studies reported adverse events caused by IPL therapy. The rest of the included studies did not specify whether any adverse events occurred.

Discussion

This review provides insight into the effectiveness of intense pulsed light (IPL) with or without traditional treatments in dry eye disease (DED) caused by meibomian gland dysfunction (MGD). Although several meta-analyses previously focused on IPL treatment of MGD [10,11,12,13], the data synthesis of these meta-analyses included self-control studies. However, our study excluded self-controlled studies owing to the possibility of correlation between subjects’ eyes, and we believe eyes of the same subject cannot be treated alone. Therefore, our conclusion of the meta-analyses would be more convincing than those of previous studies.

Primary outcomes

The Ocular Surface Disease Index (OSDI) scores and tear break-up time (BUT) were two important primary outcomes in the present analysis. The OSDI questionnaire can quantitatively evaluate dry-eye-related symptoms and is relatively simple, easy to understand and answer [26]. It is a subjective and standardized tool for evaluating dry eye disease symptoms, and the higher the score is, the worse the symptoms. BUT is an objective assessment technique for DED, used to describe tear film stability [27]. According to the quantitative synthesis, IPL application based on traditional treatments was superior to conventional treatments alone in reducing OSDI scores (total: MD =  − 7.49; 95% CI =  − 12.47 to − 2.50, P = 0.003; subtotal: MD =  − 7.01; 95% CI =  − 12.23 to − 1.80, P = 0.008). However, the subgroup analysis indicated that IPL application alone had no obvious advantage in reducing OSDI scores compared with traditional treatments (MD =  − 8.34; 95% CI =  − 23.62 to 6.93, P = 0.28). We noticed that the study results of Gao et al. [23] and Xiao et al. [16] showed no significant difference in the change in the OSDI between the two groups. The reason may be attributed to the relatively short follow-up time of the two studies (1 month and 2 weeks after last treatment), since our data synthesis results were the last follow-up time measurement of all included studies. The forest plot of BUT revealed that applying IPL alone or in combination with traditional therapies was more advantageous than traditional treatment alone, although both groups could increase BUT after treatment (MD = 1.94; 95% CI = 1.19 to 2.69; P < 0.00001). According to the quantitative analysis, the application of IPL could ameliorate the BUT of DED patients, even in a relatively short follow-up time. A significant difference in the reduction in the OSDI score between the two groups may require a relatively long follow-up time to emerge. Unsurprisingly, it takes time to improve tear film stability to ameliorate DED symptoms. In brief, the results of our analysis showed that both IPL treatment and traditional treatments could improve the stability of tear film and subjective symptoms of patients. Moreover, the effect of IPL application in improving the stability of tear film was markedly better than that of traditional treatments.

Secondary outcome

From the results of our data synthesis, the application of IPL outdoes conventional treatments in increasing NIBUT and reducing the SPEED scores of DED patients. These results further confirmed the primary outcomes above. Decreases in BUT and non-invasively measured BUT (NIBUT) are characteristic of patients with MGD-related dry eye disease. The observed increases in the two indexes indicated the improvement of tear film stability and further confirmed the therapeutic potential of IPL treatment in MGD-related dry eye diseases. While the CFS showed a downward trend, there was no statistical significance between the two groups. This may be attributed to three [16, 17, 23] of the five studies included in the present study. In these studies, a decrease in CFS was found in both the IPL application and conventional treatment groups.

We may be able to explain these results based on the mechanism of IPL. This mechanism may include several aspects: first, the thermal effect of intense pulsed light liquefies the eyelid fat, making it more conducive to expulsion [28]; second, intense pulsed light can be absorbed by hemoglobin to close abnormal blood vessels at the edge of the eyelid and conjunctiva to further reduce the release of local inflammatory factors [29]; third, IPL could reduce the number of bacteria on the edge of the eyelid, thus improving the morphology of the meibomian gland, preventing meibomian gland atrophy and stimulating the branches of the parasympathetic nerve to initiate the normal activity of the meibomian gland [29, 30].

Limitations

This systematic review has a number of limitations.

  1. 1.

    While the random sequence generation, allocation concealment and blinding of outcome data in some RCTs remained unclear, future designs should strictly adhere to methodologic guidelines to improve the quality of study.

  2. 2.

    Insufficient research may make it difficult to draw strong conclusions.

Conclusions and suggestions

On the basis of this systematic review and meta-analysis, we concluded that both IPL treatment and traditional treatments can improve the stability of tear film and ameliorate patients’ subjective symptoms and that IPL application may be superior to traditional treatments. Due to the limited quality of evidence, the results should be interpreted with some caution. Additional high-quality studies strictly adhering to methodological guidelines are necessary to provide evidence for future analysis.