The related literature was collected from the following databases: Pubmed, Embase, the Cochrane Library, and Chinese databases, including Wanfang and the China National Knowledge Infrastructure. The collected articles were published up to Oct 19, 2020. Search terms included the subject terms and random terms “orthokeratology lens” and “tear break-up time.” The detailed search strategies are found in Appendix 1. No language or time restrictions were adopted during the search.

The literature was managed through EndNote X9 software. The criteria of the articles included were: (1) prospective studies, (2) study content about the effect of orthokeratology lenses on the ocular surface, (3) leading observation indicators, including the average tear break-up time at baseline and at one, three, six, and twelve months follow-up (with repeated measurement at least three times, where a mean value and standard deviation represented the results), (4) subjects were between the ages of 6–18, (5) complete data. Studies would be removed if they were case reports, commentaries, reviews, letters, or conference abstracts. Zhengyang Tao and Zhihong Lin conducted the article-selection process. When the two authors had differing opinions on the inclusion or exclusion of a specific article, Hongwei Deng would act as arbitrator.

The quality of eight of the included articles was evaluated by the Newcastle–Ottawa scale (NOS). The counterpart of one article (a randomized controlled trial) was assessed with the Jadad scale. The total score on the NOS scale is 9 stars, with a maximum of 2 stars for the comparability column and a maximum of 1 star for the remaining ratings. A score from 5 to 9 on the NOS scale was considered relatively high quality. The scores of the Jadad scale ranged from 0 (very poor) to 7 (rigorous), in which maximum points for randomization, concealment of allocation, double-blind, and dropouts are two, two, two, and one, respectively. Jiao Wang and Minjuan Zhu conducted the grading process. When the two authors had differing opinions on the grading of a particular article, Jun Zhao would act as arbitrator.

Statistical analyses were performed using Stata software (version 15.1) and R software for Windows (version: 3.5.2, Ross Ihaka and Robert Gentleman, University of Auckland, New Zealand). A Q test and I² statistics were applied to evaluate the heterogeneity of the studies. When the heterogeneity was large, sensitivity analysis was used to further analyze the source of heterogeneity. Due to the small number of included studies, subgroup analysis will not be considered for use. The null hypothesis that the studies are homogeneous would be rejected if the P value was < 0.05 or the I² was >50%. The total effect of the mean difference and 95% confidence interval (95% CI) were calculated using the random-effects model (DerSimonian–Laird, DL). The primary research indicator of this study was tear break-up time (TBUT), by comparing the TBUT value after orthokeratology wearing for 1 month and 6 months as the baseline, respectively. There were two secondary indicators, the tear secretion (through a Schirmer I test) and the Ocular Surface Disease Index (OSDI), comparing the two values after orthokeratology wearing for 6 months as the baseline.

A systematic literature search obtained 60 articles. Among them, 23 articles were excluded due to duplication. By reading the titles and abstracts, 14 articles unrelated to this study were excluded. One review and 12 studies with insufficient follow-up were also excluded. Finally, 10 articles were included in this meta-analysis (13–21). The detailed literature screening process is shown in Figure 2. The characteristic information of the 10 included literatures is shown in Table 1.

In the NOS scale, the scores of the nine included articles were >5 stars, suggesting that the quality of the included articles was relatively high. Table 2 makes specific provisions for the score of the comparability column in the NOS scale based on the contents of this study. The first confounder is defined as lens care solution, and the second confounder is defined as sleeping duration. Only when a study reported having a control group and the first and the second confounders were described clearly could two stars be distributed to it, such as the studies by Yang, L. et al. and Yu-Ling, N. et al. Both the studies by Wang, X. et al. and Liu, X.L. et al. rated one star (the former lacked a control group, while the latter lacked adjustment for the second confounder). The remaining articles are ranked 0 in this category. The final score of each included article is shown in Table 3. In addition, since the study of Zhao, Q. et al. was an RCT study, the Jadad scale was used for scoring (see Table 4 for details).

In the heterogeneity test, p < 0.01, I² = 95.90% (TBUT, 1 month vs. baseline), p < 0.01, I² = 94.60% (TBUT, 6 months vs. baseline), p = 0.07, I² = 54% (Schirmer I test, 6 months vs. baseline), p < 0.01, I² = 98.1% (OSDI, 6 months vs. baseline), suggesting that there was a relatively large heterogeneity in each analysis of the included articles (p < 0.05 or I² > 50 indicates a relatively strong statistical heterogeneity). Due to the small number of included studies, subgroup analysis was not considered when further analyzing the source of the large heterogeneity. Instead, sensitivity analysis was conducted for the included literature to confirm each difference by sequentially omitting each of the included studies and recalculating the summary difference and 95% CI. P value and I² did not decrease significantly after the sequential removal of each article.

The fixed-effects model (Hedges' g) and random-effects model (DL) was applied to the combined effect size. The forest plots are shown in Figures 3, 4. The total effect of change in tear film stability before and after orthokeratology wearing was: 0.94 (95% CI: 0.24, 1.64), z = 2.62, p = 0.009^* (after 1 month), 0.77 (95% CI: 0.21, 1.34), z = 2.69, p = 0.007^* (after 6 months). Compared with the baseline, the total effect of the change in tear secretion after orthokeratology wearing was: −0.05 (95% CI: −0.28, 0.18), z = −0.56, p = 0.577, and the counterpart in OSDI was: −1.10 (95% CI: −2.22, 0.02), z = 1.93, p = 0.053.

The forest plots of TBUT change 3 months and 12 months after orthokeratology wearing were shown in Appendices 2A,B. In order to maintain consistency, only the analysis results of 6 months were discussed in this paper.

Different test methods will significantly affect the results of tear film stability. Compared with accessing the data of tear secretion, which has been uniformly tested by the Schirmer I test in relevant studies, tear film stability is usually measured by a tear film BUT test, which has different measurement methods and equipment, mainly including invasive TBUT and non-invasive TBUT. The invasive tear film H test by sodium fluorescein and cobalt blue filter on the slit lamp has been in use for decades and was first proposed by Norn (22). The test is typically performed using a strip of sodium fluorescein moistened with saline or eye lotion to touch the lower eyelid or sclera (or introduce a drop of 1% sodium fluorescein to the lower fornix) and perform the observation with the cobalt blue light in the slit lamp after asking the patient to blink several times. However, the volume of sodium fluorescein introduced to the eye surface is considered an important factor affecting the measurement results of tear film TBUT (23, 24). A non-invasive tear film rupture test is mainly carried out by specially designed tear film measuring devices such as the Keratograph 5M, Oculus Keratograph 4, or Medmont E300 CornealTopographer, which is being widely used (25–27).

The test process of NIBUT does not interfere with the surface of the eye. Therefore, it is more acceptable for children and has good repeatability (28). However, comparing the results of different studies should be treated with caution due to the poor comparability between different equipment (29, 30). TBUT, on the other hand, has been applied clinically for decades, and TBUT (sec) in healthy people is generally less than the NIBUT, even though the minimum measurable value of the latter is lower (24, 31). Fluorescein staining can affect the physiological properties of the tear film, but it can also reflect, to a certain degree, the changes in the tear film stability caused by this external irritant. Compared with NIBUT, the magnifying effect of this traditional test indicates a more sensitive non-significant reduction in tear film quality. Therefore, the data of NIBUT is recommended for presentation in the paper, based on good repeatability and convenience. Also, TBUT should be carried out, and the procedures should be documented as thoroughly as possible.

The lens treatment before wearing includes disinfection, washing, and lubrication. Because most of the current multipurpose care solutions for disinfection cannot directly contact the eyeball, eye care providers usually require patients to clean the lens before wearing (32, 33). Since the use of distilled water to rinse the lens may change the tear quality during subsequent wearing due to its tendency to change the pH and osmotic pressure of the tear, the use of saline solution, which is close to the physiological osmotic pressure of tears, can help reduce irritation and may be appropriate for lens rinsing. In order to avoid lens-related corneal infections, tap water is not recommended for lens rinsing due to the diverse sterilizing standards of tap water in different countries (34, 35).

In addition, the current research shows that, after cleaning the lens, whether to continue to fill the lens with solution (mainly to increase the wearing comfort) and the type of solution (such as saline solution, artificial tears, etc.) will significantly affect the measurement results of the tear film stability (36–38). The study by Carracedo et al. showed that the corneal staining in the group using sodium hyaluronate (0.3% Ocu-Dry) before wearing glasses was significantly less than the group using saline solution and rewetting drops, but there was no significant difference in TBUT among the three groups (38). However, a recent study by Liu et al. showed the opposite result. The TBUT of the group using sodium hyaluronate (0.1% Hylo) before wearing glasses was higher than the group using saline solution and rewetting drops (37). Carbomer eye drops, another artificial alternative to tears, showed similar results in the study by Shi et al. (36).

Since the processing steps of the lens before wearing may affect the measurement of tear quality, it is advisable to record the lens treatment process, especially the solution used in the washing and lubrication steps.

Current research has shown a close relationship between sleep, sleep duration, sleep quality, and changes in tears. In a study in Singapore and India, short sleep duration was shown to be significantly and independently associated with dry eye symptoms (39). Similarly, in a study by Zhao et al., which focused on a group of dry eye patients, TBUT and tear secretion were tested in three groups of dry eye patients with different sleep durations. It was revealed that the group with <6 h of sleep duration performed worst in the tests, while the counterparts with 8 h showed the best results (40). In a study by Lee et al., the healthy subjects in the sleep deprivation group performed significantly worse on the TBUT and Schirmer test than the control group with normal sleep (41). Although there is no evidence that tear production and stability of those who sleep more than 6 h vary over time, these studies have only been conducted on adults (39, 40, 42). The relationship between sleep duration and changes in tears has rarely been studied in children and adolescents. Therefore, for future researchers who wish to perform studies on the effects of orthokeratology on the ocular surface in children and adolescents, we recommend that they specify and state the sleep duration in the article.

The American Academy of Pediatrics suggested that children 6 to 12 years of age should sleep 9 to 12 h per 24 h, and 8 to 10 h per 24 h for teenagers between 13 and 18 years of age regularly to promote optimal health (43). Therefore, it may be a more appropriate choice to let the subjects maintain 9–10 h of sleep to further increase the research homogeneity in this field.

Sleep itself has a significant effect on tear quality (secretion and stability). When the eyelid closes, inflammatory changes occur in tears in the conjunctival sac, accompanied by neutrophil accumulation, plasminogen activation, increased albumin levels, pro-inflammatory factors, and complement activators (44–46). The tear osmolarity reaches the lowest level for the whole day after waking up, and the ocular surface is rich in inflammatory factors, such as Immunoglobulin A (SIgA) polymorphonuclear leukocytes and complement proteins (47). This change may be associated with corneal physiological edema and “sub-inflammation status,” caused by the closure of the eyelids during sleep and the lack of enough atmospheric oxygen to exchange on the anterior surface of the cornea. In the study of Patel et al., the tear instability reached a roughly balanced state near noon (10 am−12 pm) (48).

In recent years, there have been many studies on how tear quality changes over the course of a day. In a study by Bitton et al., the tear film stability of the dry eye group was at its worst at 7 a.m. and 10 p.m., but the TBUT in the healthy group did not show a significant change over the daytime (49). Lira et al. further studied changes in tear film stability in healthy people over diurnal variation. They found that the tear film stability of healthy people was slightly worse at 5 p.m. than at 9 a.m., which was particularly evident in the test of the tear film stability using fluorescein staining, indicating that the resistance of tears to stimuli was significantly reduced in the evening (50). The previous studies show a rough trend that tear quality is at its worst in the early morning and gradually reaches a more stable state toward noon. The quality of tears decreases slowly over the course of the day and becomes more apparent when sleep deprivation occurs.

On the other hand, compared with the RGP lens, which is worn during the day and in full contact with the cornea (the eye surface will get sufficient tear exchange after each blink), the orthokeratology lens is worn at night, and the lack of blink leads to less tear exchange, which may further increase the “sub-inflammatory” reaction of the eye surface during sleep. Therefore, it is imperative to choose a suitable time to evaluate the tear quality of the OK lens wearer. The ideal timing should be after most of the sleep-induced ocular surface changes have subsided and before the tear quality has declined. Therefore, we suggest that researchers clearly describe the measuring time (period) in their paper. Implying measurements at the time near noon, such as 10 a.m.−12 p.m., may be a benefit to obtain more accurate data and increase homogeneity between studies.

Since the effectiveness of orthokeratology for myopia control has been proven, the essence of this study is about the complications associated with a kind of proven effective intervention for myopia (2, 3). Whether from the perspective of experimental ethics or feasibility, it is difficult for the subject in this paper to implement blinding and randomization in research, especially for the researchers in clinical work. Prospective cohort studies may be the appropriate study design for this subject (51). First of all, a prospective study is about the process from “cause” to “result,” which can better explain whether there is a causal relationship between orthokeratology lens wearing and tear quality change. Secondly, the design of a control group in a cohort study helps analyze and eliminate the influence of unknown interference factors on the results.

In addition, with the emergence of increasing methods for myopia control, such as low-concentration atropine, peripheral defocus glasses, visual perception training, etc., clinical studies in the setting of blank control groups need to consider a larger proportion of ethical factors. Therefore, more clinical studies adopted single-arm studies, which are conducted by investigating the change before and after the main intervention (52). Among the 10 articles included in this study, five were single-arm studies. Single-arm studies often show significant heterogeneity between the results of different studies due to the absence of control groups, as shown in the meta-analysis above. At the same time, due to the lack of appropriate quality assessment tools, this study adopted the NOS scale set for the cohort study. This can hardly be considered suitable for single-armed studies, as can be seen from columns of “Selection of the Non-Exposed Cohort” and “Comparability of Cohorts on the Basis of the Design or Analysis.” Therefore, it is necessary to develop appropriate quality assessment scales for single-arm studies, which are increasingly and commonly conducted in ophthalmology and many other areas, such as oncology or other rare diseases.

The unbalanced spatial distribution of research is an important aspect that should be highlighted here. The studies by Jennifer et al. showed that the TBUT in Caucasian populations was significantly longer than in Asians in the adult population (53). This conclusion is consistent with previous research (54–56). However, in the study of Kim et al., there was no significant difference in tear film stability between Asian and Caucasian children (aged from 5–18 years) (57). In addition, the study by Wang et al. suggested that differences in tear quality between people were more influenced by the growth environment than by ethnic factors (58). The study by Wang et al. may provide a reasonable explanation for the non-significant difference in tear film stability between the two races in the study by Kim et al.

Whether ethnic factors or geography plays a more important role in the differences in tear film stability in the study, especially among children, is still unclear, which needs to be answered by more multi-center studies with larger samples in the future. However, what is certain is that Asians who grow up in Asia have different ocular surface features than people who grow up in Europe or America. This may be closely related to climatic, sociocultural, genetic, and other factors. This suggests that the comparability of studies conducted in different districts is limited, and it is not advisable to blindly summarize and discuss the studies with various ethnic features and spatial distribution. Secondly, all 10 studies included in this research were in Asian populations. Future studies should fill gaps in this field with European, African, and American populations, which will have profound implications for better understanding the complications of orthokeratology.

Current research has shown that the usage of electronic products is an important factor in the development of dry eye (59–61). A study on healthy people performed by Bettach et al. showed that the break up time related parameters were significantly reduced after the reading task by smartphone screen (62). The study by Zhu et al. showed that screen time was one of the risk factors for dry eye disease, and different electronic screen time had a significant impact on the change of tear quality (63). In addition, recent studies have shown that electronic products with different screen materials have different effects on the eye surface. Yuan et al. showed that after reading for the same period of time using electronic screens, subjects using eINK screens showed less negative changes on BUT, OSDI and other indicators compared with subjects using OLED screens (64).

It is not difficult to speculate that in the process of orthokeratology wearing, electronic products also have a non-negligible effect on the change of tear quality. However, this factor is often not taken into account in the current studies. Therefore, in future studies, it may be a feasible method to obtain the data of electronic product usage by questionnaires to improve the persuasibility of conclusions.