Professional Documents
Culture Documents
Dry Eye After Refratic Surgery A Meta Analysis
Dry Eye After Refratic Surgery A Meta Analysis
ABSTRACT
Objective: To examine and compare the rates of dry eye after various refractive surgeries, including laser-assisted in situ keratomileusis
(LASIK), photorefractive keratectomy (PRK), femtosecond lamellar extraction (FLEx), and small incision lenticule extraction (SMILE).
Study Design: Systematic review and meta-analysis.
Methods: This systematic review and meta-analysis was conducted following the Preferred Reporting Items for Systematic Reviews and
Meta-Analyses statement guidelines. Databases searched included MEDLINE (Ovid), Embase, Cochrane Library, Cumulative Index of
Nursing and Allied Health Literature, Web of Science, ProQuest Dissertations and Theses, and ClinicalTrials.gov. Meeting abstracts
from European Society of Cataract and Refractive Surgeons, American Academy of Ophthalmology, Association for Research in
Vision and Ophthalmology, and Canadian Ophthalmological Society were also examined. Articles underwent 3 stages of screening
before data extraction and meta-analysis.
Results: In total, 3232 studies were found; 261 remained after title screening, 92 remained after abstract screening, and 14 studies pro-
gressed to data extraction. Meta-analysis indicated a significant reduction in tear break-up time (TBUT) with LASIK (standardized
mean difference [SMD] = 0.3; confidence interval [CI]: 0.53 to 0.08) and FLEx (SMD = 1.09; CI: 1.44 to 0.74), and a nonsignif-
icant reduction in TBUT with SMILE (SMD = 0.34; CI: 0.95 to 0.27) and PRK (SMD = 0.11; CI: 0.29 to 0.08). Meta-analysis also
indicated a significant reduction in tear production with LASIK (SMD = 0.23; CI: 0.46 to 0.01), and a nonsignificant reduction
in tear production with SMILE (SMD = 0.04; CI: 0.28 to 0.36), FLEX (SMD = 0.05; CI: 0.37 to 0.28), and PRK (SMD = 0.07;
CI: 0.32 to 0.19).
Conclusions: Overall, a significant reduction in postoperative tear production as well as TBUT time was seen with LASIK, and a nonsig-
nificant reduction in postoperative tear production and TBUT was seen with SMILE, FLEx, and PRK. Ultimately, more high-quality ran-
domized controlled trials are required to make concrete conclusions about dry eye parameters after refractive surgery.
Refractive surgery to correct vision is one of the most com- quite difficult, degrading QOL.4 In addition, these symp-
mon elective surgical procedures in the United States. As of toms can have a tremendous socioeconomic cost. This
2009, over 28 million laser-assisted in situ keratomileusis includes the direct cost of medical care for DED and the indi-
(LASIK) procedures have been performed.1 This is in addi- rect costs, such as inability to work and reduced productivity
tion to all of the photorefractive keratectomy (PRK), small when at work. Studies have shown that the annual direct
incision lenticule extraction (SMILE), and femtosecond costs of DED in the United States are US$3.84 billion, and
lamellar extraction (FLEx) procedures also being performed, the annual indirect costs are as high as US$55.4 billion.4
both in the United States and globally. Furthermore, the Consequently, it is crucial to understand the rates of dry eye
majority of candidates for refractive surgery are young, active and DED after the various refractive surgical procedures to
individuals, generally 2030 years old.2 Refractive surgery is adequately understand the potential impacts it can have.
known to be a safe vision correction option for individuals The rate of dry eye and DED after refractive surgery is a
wanting to rid themselves of contact lenses and glasses. How- pertinent topic to explore. With numerous individuals under-
ever, a common side effect of refractive surgery is dry eye. going refractive surgery procedures every year, with many of
Some studies have reported that up to 95% of patients experi- the patients being young and healthy, it is important to know
ence dry eye symptoms after refractive surgery, with a portion the incidence of dry eye after the various procedures. This
of these individuals experiencing chronic dry eye symptoms, improved understanding of the prevalence of dry eye after
also known as dry eye disease (DED).3 DED can have major refractive surgery can help physicians better predict and man-
implications on an individual’s quality of life (QOL) and can age patient outcomes, potentially alleviating the negative
also have a tremendous societal burden.4 QOL and socioeconomic impacts of postrefractive surgery
DED, sometimes referred to as keratoconjunctivitis sicca, dry eye. Although several studies have compared the rate of
keratitis sicca, or dry eye syndrome, has a major socioeco- dry eye in some of the refractive surgery techniques, to our
nomic and QOL impact. Symptoms of DED include blurry knowledge, there is no systematic review and meta-analysis
vision, ocular irritation, excessive tearing, ocular hyperemia, comparing the rates of dry eye after all of the current popular
and photophobia, among others. As seen through various refractive surgeries to one another (LASIK, PRK, SMILE,
studies, these symptoms can make activities of daily living and FLEx).
The objective of this systematic review and meta-analysis is Once all of the studies were compiled, the articles were
to consolidate all of the current literature regarding the rates imported into DistillerSR and duplicates were removed. Next
of dry eye after patients undergo any of the current common title screening, abstract screening, and full-text screening
refractive surgeries. The procedures this article focuses on are were done to determine which studies were relevant.
LASIK, PRK, FLEx, and SMILE.
Screening
TAGEDH1METHODSTAGEDEN Screening was conducted in 3 stages. DistillerSR from Evi-
dence Partners was used to conduct the screening of the
Search Strategy/Design
articles. The studies were screened by 2 reviewers indepen-
This systematic review and meta-analysis was conducted
dently, and Cohen’s kappa (k) coefficient was computed to
following the guidelines of the Preferred Reporting Items for
assess the agreement of inclusion between the 2 reviewers (R.
Systematic Reviews and Meta-Analyses statements. Several
S., G.S.) at each stage of screening. If there were disagree-
databases and grey literature sources were searched to find rel-
ments on inclusion/exclusion, they were resolved mutually.
evant studies. The databases that were used to find relevant
In the case of disagreements, a third reviewer (M.M.) inter-
literature were MEDLINE (OVID), Embase, Cochrane
vened to provide a decision.
Library, and Cumulative Index of Nursing and Allied Health
The 3 stages of screening were title screening, abstract
Literature. Search strategies were created for each of the data-
screening, and full-text screening. Detailed screening ques-
bases (Appendix A, available online), and the searches were all
tions are provided in Appendix B (available online). At the
run until June 20, 2018. Database-specific subject headings
title screening stage, studies with titles that did not seem rele-
and key words for “refractive surgery,” “dry eye disease,” “kera-
vant to the topic were excluded. At the abstract screening
toconjunctivitis sicca,” “keratitis sicca,” and “dry eye syndrome”
stage, studies without abstracts were excluded, in addition to
were employed in the search strategy. The searches were modi-
those that did not appear to be relevant to the topic. During
fied to accommodate the unique terminology and syntax of
the full-text screening process, articles that did not have
each database. Additionally, all synonyms were taken into
appropriate information (Schirmer test, tear break-up time,
account with the help of an information specialist. Methodo-
postsurgery dry eye prevalence, etc.) were excluded.
logical filters were applied to limit our retrieval to comparative
studies, observational studies, interventional case series, clinical
trials, randomized trials, observational studies, retrospective Study Quality
studies, prospective cohort studies, and contralateral prospective The final selection of literature that remained after the full-
studies. OVID AutoAlerts were set up to send monthly updates text screening stage was examined for study quality using a
with any new literature up until March 1, 2019. Monthly modified Downs and Black checklist (Appendix C, available
updates were performed on Health Economic Evaluations online).5 The parameters that were assessed included report-
Database, PubMed, and Cochrane Library databases. ing, external validity, internal validity (bias, and confounding
Grey literature was obtained by searching Web of Science, factors), and power. Each study was given a score out of 28.
ProQuest Dissertations and Theses, ClinicalTrials.gov, and Studies were labeled as poor quality if they scored 14, fair
the meeting abstracts from the American Society of Cataract quality if they scored 1519, good quality if they scored
and Refractive Surgeons, European Society of Cataract and 2025, and excellent if they scored 2628. Assessment of
Refractive Surgeons, American Academy of Ophthalmology, study quality was completed by one individual (R.S.).
the Association for Research in Vision and Ophthalmology,
and the Canadian Ophthalmological Society.
Data Extraction
Data extraction from the included articles was completed
Inclusion/Exclusion Criteria by one investigator (R.S.). Information extracted from the
Regarding the inclusion criteria for the systematic review, studies that progressed past full-text screening included study
studies pertaining to both dry eye and any type of refractive year, design, location, number of patients and eyes, mean
surgery were included in the analysis. Study designs included age, visual acuity, percentage of patients with dry eye symp-
in the review were clinical trials, randomized trials, observa- toms, main outcome measures of the study, type of refractive
tional studies, retrospective, and prospective studies. Regarding surgery explored in the study, follow-up period(s), risk factors
exclusion criteria, studies involving patient populations who for dry eye after the various refractive surgeries, and the result
had pre-existing dry eye or DED before undergoing refractive values of tests that assess for dry eye (Schirmer test, tear
surgery were excluded from the analysis. Studies that looked at break-up time [TBUT], and Ocular Surface Disease Index
nonhuman subjects, published before 1995, and were non- [OSDI]). Schirmer’s test and TBUT were more ubiquitous
English were excluded. Studies that had a sample size of less and therefore utilized for the meta-analysis. Additionally,
than 20 eyes and that involved nonadult subjects (younger these tests provide a more objective measurement of dry eye
than 18 years) were also excluded. No limits were placed based symptoms as opposed to the dry eye symptom scales (such as
on study location or postoperative follow-up periods. OSDI, McMonnies Dry Eye Questionnaire). The available
data on range, p-value, and CI were used and converted to outcome measure were divided by the standard deviation for
the common effect measure, standard deviation. that same outcome measure. Weights were assigned to each
SMD according to the inverse of its variance and then average
was computed. The SMD in each study was pooled with a
Meta-Analysis fixed- or random-effect model based on heterogeneity.
After the completion of data extraction, the statistical/ Inverse variance (I-V) method was used to compute fixed-
meta-analysis was done by one investigator (M.M.). STATA effect model, and DerSimonian and Laird (D+L) method was
15.0 (STATA Corporation, College Station, TX) was used to used to compute random-effects model. To test heterogene-
conduct the meta-analysis. For continuous scale outcomes ity, a Z-value was computed to test the null hypothesis,
including mean values, the standardized mean difference which was a treatment effect of zero.
(SMD) was calculated as the treatment effect or effect size. Additionally, heterogeneity was determined using the I2
This parameter represents the mean difference standardized value, which indicated the extent of variation across studies due
for variances across all studies. To compute SMD for each to heterogeneity rather than chance.6 Heterogeneity between
study, the mean pre- and postoperative values for each studies was examined using the x2 test, where the x2 test
Fig. 1—Funnel plot of the studies used in the analysis of the Schirmer’s test values, organized by refractive surgeries.
Fig. 2—Funnel plot of the studies used in the analysis of tear break-up time values, organized by refractive surgeries.
assessed whether the observed between-study differences were removed because either they did not provide information on
due to chance only. Low p-value and a large x2 statistic relative the prevalence of dry eye after refractive surgery or they did
to its degree of freedom provided evidence of heterogeneity. not have abstracts available. After this stage, 92 studies
Funnel plots were generated to check publication bias. remained and were moved onto full-text screening (level 3
screening). At the full-text screening stage, the studies were
examined to see if they fulfilled the inclusion criteria, met
Publication Bias any of the exclusion criteria, and had the specific information
The potential presence of publication bias was assessed and statistics that needed to be extracted. At this stage, the
with the examination of the funnel plots. Funnel plots for the studies were removed for the following reasons: 7 studies
Schirmer test and the TBUT for various refractive surgeries because they were in a non-English language, 1 study did not
were plotted (Figs. 1 and 2, respectively). have a large-enough sample size (less than 20 eyes), 8 studies
had the wrong study design, 1 study had the wrong patient
population, 40 studies explored the wrong outcomes, and 11
TAGEDH1RESULTSTAGEDEN studies had the wrong intervention. After completing full-
Search Results text screening and resolving conflicts, 14 studies remained,
The search strategy resulted in 3238 studies, which were which underwent the quality check process. Two studies
potentially relevant to the review; 2335 studies remained after (14.3%) were fair quality; 12 studies (85.7%) were good
duplicates were removed. The remaining studies were then quality; no studies were of excellent or poor quality. After
moved to title screening (level 1 screening). After both quality check, these studies were moved to the data extraction
reviewers screened the studies, 2068 studies were removed as stage. The results of the screening process are summarized in
they did not explore a link between refractive surgery and dry Figure 3. The kappa statistics, which were calculated before
eye; 267 of the studies progressed to abstract screening (level conflict resolution, for levels 1, 2, and 3 of screening were
2 screening). At the abstract screening stage, 175 studies were 0.71, 0.42, and 0.94, respectively.
Fig. 4—Forest plot depicting the changes in tear break-up time values following the various refractive surgeries (SMILE, FLEX,
PRK, and LASIK).
Fig. 5—Forest plot depicting the changes in Schirmer’s test values following the various refractive surgeries (SMILE, FLEX, PRK,
and LASIK).
Publication Bias in tear production was observed with LASIK and a nonsignif-
Figure 1 shows the funnel plot of the studies used in the icant reduction in tear production with SMILE, FLEX, and
analysis of the Schirmer’s test values by refractive surgeries. PRK. A maximum reduction in TBUT as well as a maximum
The studies were scattered from the top to the bottom right of reduction in tear production was seen with LASIK compared
the plot. Figure 2 shows the funnel plot of the studies used in with other refractive surgeries.
the analysis of the TBUT values by refractive surgeries. The However, there was only one study evaluating FLEX and 2
studies were scattered from the top left to the bottom right of studies evaluating PRK, and thus more randomized con-
the plot. Therefore, publication bias could not be concluded. trolled trials are required to make strong conclusions. More-
Partially, the reason was difficulty in interpretation of funnel over, SMILE is a relatively new procedure compared with the
plot for a small group of studies, high heterogeneity, and small other refractive surgery techniques. Therefore, better quality
effect sizes. Additionally, publication bias is only one of the randomized controlled trials with large sample sizes and long-
numerous possible explanations for funnel plot asymmetry. term follow-up data on SMILE are missing.
The increased rates of dry eye in LASIK compared with
the other surgical methods may be explained by damage to
TAGEDH1DISCUSSIONTAGEDEN the corneal nerves. In LASIK, when the superior hinge
Postoperative dry eye has previously been shown to be method is used, both arms of the long posterior corneal
extremely common in patients who undergo refractive surgi- nerves are transected. Using the nasal hinge method in
cal procedures, including LASIK, SMILE, PRK, and FLEx. LASIK, one branch of the long corneal nerve is still trans-
Through this study we were able to examine and compare ected.7 For SMILE, literature has shown that fewer corneal
the rates of dry eye after these procedures. Our meta-analysis nerve fibres are transected compared with LASIK.8 This
results indicated a significant reduction in TBUT with may play a role in preserving corneal sensitivity and may
LASIK and FLEx, and a nonsignificant reduction in TBUT explain the increased incidence of dry eye in LASIK when
with SMILE and PRK. Furthermore, a significant reduction compared with SMILE and other similar refractive