Laser‐assisted in‐situ keratomileusis (LASIK) versus photorefractive keratectomy (PRK) for myopia

Abstract

Background

Myopia (also known as short‐sightedness or near‐sightedness) is an ocular condition in which the refractive power of the eye is greater than is required, resulting in light from distant objects being focused in front of the retina instead of directly on it. The two most commonly used surgical techniques to permanently correct myopia are photorefractive keratectomy (PRK) and laser‐assisted in‐situ keratomileusis (LASIK).

Objectives

To compare the effectiveness and safety of LASIK and PRK for correction of myopia by examining post‐treatment uncorrected visual acuity, refractive outcome, loss of best spectacle‐corrected visual acuity, pain scores, flap complications in LASIK, subepithelial haze, adverse events, quality of life indices and higher order aberrations.

Search methods

We searched CENTRAL (which contains the Cochrane Eyes and Vision Group Trials Register) (The Cochrane Library 2012, Issue 11), Ovid MEDLINE, Ovid MEDLINE In‐Process and Other Non‐Indexed Citations, Ovid MEDLINE Daily, Ovid OLDMEDLINE (January 1946 to November 2012), EMBASE (January 1980 to November 2012), Latin American and Caribbean Literature on Health Sciences (LILACS) (January 1982 to November 2012), the metaRegister of Controlled Trials (mRCT) (www.controlled‐trials.com), ClinicalTrials.gov (www.clinicaltrials.gov) and the WHO International Clinical Trials Registry Platform (ICTRP) (www.who.int/ictrp/search/en). We did not use any date or language restrictions in the electronic searches for trials. We last searched the electronic databases on 15 November 2012. We also searched the reference lists of the studies and the Science Citation Index.

Selection criteria

We included randomised controlled trials comparing LASIK and PRK for the correction of any degree of myopia.

Data collection and analysis

Two authors independently assessed trial quality and extracted data. We summarised data using the odds ratio and mean difference. We combined odds ratios using a random‐effects model after testing for heterogeneity.

Main results

We included 13 trials (1135 participants, 1923 eyes) in this review. Nine of these trials randomised eyes to treatment, two trials randomised people to treatment and treated both eyes, and two trials randomised people to treatment and treated one eye. None of the paired trials reported an appropriate paired analysis. We considered the overall quality of evidence to be low for most outcomes because of the risk of bias in the included trials. There was evidence that LASIK gives a faster visual recovery than PRK and is a less painful technique. Results at one year after surgery were comparable: most analyses favoured LASIK but they were not statistically significant.

Authors' conclusions

LASIK gives a faster visual recovery and is a less painful technique than PRK. The two techniques appear to give similar outcomes one year after surgery. Further trials using contemporary techniques are required to determine whether LASIK and PRK as currently practised are equally safe. Randomising eyes to treatment is an efficient design, but only if analysed properly. In future trials, more efforts could be made to mask the assessment of outcome.

Plain language summary

Laser‐assisted in‐situ keratomileusis (LASIK) compared to photorefractive keratectomy (PRK) for correcting short‐sightedness

Myopia is the term used to describe short or near‐sightedness, which means that you cannot see objects in the distance clearly. Most people with myopia wear spectacles or contact lenses. Glasses can be uncomfortable and are not practical for sport; contact lenses can be associated with corneal infections. For these reasons, some people choose to have surgery for myopia. Two commonly used surgical techniques are LASIK and PRK. Both these procedures use laser to remove corneal tissue and reshape the cornea. This review analyses the results from 13 clinical trials where 1923 eyes of 1135 participants were randomly treated with either LASIK or PRK. We considered the overall quality of evidence from these studies to be low. There was some evidence that LASIK gives a faster visual recovery than PRK, and is a less painful technique, although visual results one year after surgery were comparable. Surgical techniques are improving all the time and further trials are needed to see whether LASIK and PRK, as currently practised, are equally safe.

Summary of findings

Summary of findings for the main comparison. Laser‐assisted in‐situ keratomileusis (LASIK) versus photorefractive keratectomy (PRK) for myopia.

Laser‐assisted in‐situ keratomileusis (LASIK) versus photorefractive keratectomy (PRK) for myopia
Patient or population: patients with myopia Settings: Intervention: LASIK Comparison: PRK
Outcomes	Illustrative comparative risks* (95% CI)		Relative effect (95% CI)	No of participants (studies)	Quality of the evidence (GRADE)	Comments
	Assumed risk1	Corresponding risk
	PRK	LASIK
UCVA of 20/20 or better Follow‐up: 12 months	600 per 1000	711 per 1000 (623 to 786)	OR 1.64 (1.10 to 2.45)	1007 (7 studies)	⊕⊕⊝⊝ low2	Excluding 2 studies at high risk of selection bias gave an OR of 1.39 (0.65 to 3.00)
Within 0.50 D of target refraction Follow‐up: 12 months	750 per 1000	809 per 1000 (748 to 863)	OR 1.45 (0.99 to 2.10)	1007 (7 studies)	⊕⊕⊝⊝ low2	Excluding 2 studies at high risk of selection bias gave an OR of 1.33 (0.90 to 1.96)
Postoperative spherical equivalent Follow‐up: 12 months		The mean postoperative spherical equivalent in the intervention groups was 0 higher (0.06 lower to 0.04 higher)		589 (6 studies)	⊕⊕⊝⊝ low2
Lost one or more lines of BCVA Follow‐up: 6 months or more	100 per 1000	89 per 1000 (54 to 143)	OR 0.88 (0.51 to 1.50)	746 (6 studies)	⊕⊕⊝⊝ low2
Final BCVA of 20/40 or less Follow‐up: 6 months or more	10 per 1000	1 per 1000 (0 to 19)	OR 0.12 (0.01 to 1.93)	442 (6 studies)	⊕⊝⊝⊝ very low2,3
Pain scores	See comment		Not estimable	0 (3)	See comment	3 studies reported pain scores; significantly more pain experienced in the PRK group
*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). BCVA: best spectacle‐corrected visual acuity; CI: confidence interval; D: dioptres; RR: risk ratio; OR: odds ratio; UCVA: uncorrected visual acuity
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.

¹ Median risk in PRK group across studies. 
 ²None of the trials were masked and so were considered to be at risk of performance and detection bias; in two trials allocation was not properly concealed and therefore they were at risk of selection bias. 
 ³ Only two events, both observed in one study.

Background

Description of the condition

Myopia (also known as short‐sightedness or near‐sightedness) is an ocular condition in which the refractive power of the eye is greater than is required. The main determinants of refraction are the focusing power of the cornea and crystalline lens and the length of the eye. In myopia light from distant objects is focused in front of the retina instead of on it. This occurs because the corneal curvature is too strong or the eye is too long. As a result objects in the distance appear blurred. Near objects appear less blurred or may be seen clearly depending on the degree of myopia. People with myopia can be classified into two groups, those with low to moderate myopia (0 to < ‐6 dioptres) and those with moderate to high myopia (greater than ‐6 dioptres) (Sugar 2002).

The prevalence of myopia varies with age, country, ethnic group, level of education and occupation. The prevalence of myopia in Western populations is estimated to be approximately 25% (Kempen 2004; Sorsby 1960; Sperduto 1983). In some Asian populations myopia prevalence is as high as 70% to 90% (Chow 1990; Wong 2000). According to epidemiological evidence the prevalence of myopia is increasing, especially in Asian populations (Rajan 1995; Tay 1992). Most cases of myopia present in children of school age and young adults. The presenting complaint is difficulty reading objects at a distance and diagnosis is based on the results of refraction (spectacle testing). The exact cause of myopia is not yet clear, however there is substantial evidence that both genetic and environmental factors play a role in its aetiology (Fredrick 2002; Mutti 1996).

Description of the intervention

The most commonly used methods for correcting myopia are spectacle correction and contact lens wear. These conservative optical methods provide temporary correction. They each have functional limitations such as the problems encountered in wearing spectacles when showering or playing sports and the inconvenience of carrying contact lens solutions and storage containers, or obtaining them in the event of unforeseen circumstances. Wearing contact lenses is not without risk as it has been shown to increase the risk of sight‐threatening corneal infection (Dart 1998; Foulks 2006).

Surgical procedures have been developed in an attempt to permanently correct myopia. These procedures involve an operation on either the cornea (corneal refractive procedures) or lens of the eye (lenticular refractive procedures) and work by reducing the focusing power of the cornea or lens respectively.

Corneal refractive procedures used to correct myopia include:

• excimer laser refractive surgery: this is divided into two main procedure groups, 'surface treatments' and 'flap treatments';

  • In surface treatments, the skin on the surface of the cornea is removed by physical scraping or peeling and the laser is applied to the surface of the main body of the cornea, known as the stroma. The laser corrects the shape of the corneal stroma and therefore abolishes myopia. The surface skin can be left to heal naturally with the aid of a contact lens (as in PRK) or the removed dead skin can be replaced and may act like a bandage whilst new skin regenerates below it, as in laser epithelial keratomileusis (LASEK) or epipolis (Greek for surface) LASIK which is also known as EpiLASIK. All of these are categorised as surface treatments.

  • Flap treatments use a blade or a femtosecond laser to cut a thin flap on the surface of the cornea. This flap is peeled back and the excimer laser is applied within the body of the corneal stroma. The flap is replaced at the end of the procedure. This flap treatment is called laser‐assisted in‐situ keratomileusis (LASIK). A recent variant of LASIK is sub‐Bowmans keratomileusis (SBK) which differs from LASIK only in that the thickness of the flap is substantially less. Hence SBK is also referred to as 'thin‐flap LASIK'.

• incisional procedures: radial keratotomy or astigmatic keratotomy (a blade is used to make cuts in the cornea to alter its shape);

• tissue and synthetic implants: epikeratophakia, keratophakia and intracorneal rings (in these techniques human corneal tissue or synthetic devices are inserted into the cornea to change its shape).

Lenticular refractive procedures used to correct myopia include:

• clear lens extraction with or without intraocular lens insertion (this operation is identical to cataract surgery and is also called refractive lens exchange);

• phakic intraocular lens insertion.

Why it is important to do this review

The two most commonly used surgical techniques to correct myopia are PRK and LASIK. Both these techniques use the ophthalmic excimer laser to remove corneal tissue and reshape the cornea thus reducing its refractive power. The number of PRK and LASIK procedures being performed has rapidly increased over the last 20 years (Leaming 2004). These interventions are performed on healthy eyes and the vast majority of patients are under 60 years of age. It is important that patients are informed about and understand the effectiveness, limitations, safety, complications and relative merits of these procedures.

PRK was the first technique to employ the ophthalmic excimer laser for correction of myopia (Epstein 1994; Goodman 1989; Munnerlyn 1988). Later Pallikaris et al described laser in‐situ keratomileusis, which is now widely known as LASIK (Pallikaris 1990; Pallikaris 1991). PRK gained FDA approval before LASIK and was initially more widely performed; but LASIK uptake grew rapidly in the late 1990s and LASIK quickly became the dominant method of laser refractive correction for myopia. This shift was not based on any clear evidence of a superior visual outcome for LASIK but rather upon other factors such as those summarised in reviews by Sugar 2002 and Sutton 2010:

• earlier post‐treatment stabilisation of visual acuity;

• less post‐treatment patient discomfort;

• faster improvement in visual acuity;

• possibly improved predictability and stability;

• less stromal haze formation;

• easier enhancement procedure.

The most feared outcome of either procedure is loss of vision, specifically loss of best‐corrected vision. In PRK this is most likely to occur due to corneal haze (an inflammatory reaction of the cornea to treatment). Mitomycin C (MMC) is a chemotherapy agent which when applied to the cornea following PRK may reduce the risk of this complication in high‐risk patients (Lee 2005). Corneal haze is rare following LASIK (SUMMIT 1998) but complications arising from the creation of a flap can result in loss of best‐corrected vision (Gimbel 1998; Lin 1999; Stulting 1999; Sugar 2002). When managed appropriately the refractive and visual outcomes following flap complications are comparable to those in uncomplicated cases (Ito 2004; Sharma 2005). Corneal ectasia, a distortion of the shape of the cornea, is another rare but potentially serious complication of LASIK (Chuck 2008; Pallikaris 2001; Randleman 2003; Sugar 2002). Careful screening of patients preoperatively for risk factors such as high myopia, forme fruste keratoconus and low residual stromal thickness post‐treatment is important to minimise the risk (Randleman 2003). There have been concerns about the long‐term stability of laser refractive surgery. While individual cases of regression (the eye becoming long‐ or short‐sighted again following an initially successful treatment) can occur, data from studies demonstrates that overall the outcome of both LASIK and PRK remains stable in the long term (Dirani 2010; O'Doherty 2006; Rajan 2004; Sekundo 2003; Stephenson 1998).

The initial version of this Cochrane review (Shortt 2006a) was the first systematic review comparing LASIK and PRK. The data available at that time confirmed that:

• visual recovery is faster following LASIK than PRK;

• final uncorrected visual acuity may be superior following LASIK but the result was sensitive to exclusion of a large study which was at high risk of bias;

• there is no difference in post‐treatment refraction (accuracy) between LASIK and PRK;

• LASIK is safer in that it resulted in fewer eyes losing 2 or more lines of visual acuity than PRK.

Since this review was first published there have been two major changes in the way LASIK and PRK are performed. Firstly, contemporary PRK and LASIK treatments are invariably wavefront‐guided whereas the studies included in the original review were not. This means that the laser uses a more detailed set of data about an eye to create a customised pattern of laser treatment specifically tailored to that eye. In theory this should result in more accurate treatments and superior outcomes although the results of studies are inconsistent. A recent meta‐analysis showed no clear evidence of a benefit of wavefront‐guided over non‐wavefront‐guided ablations (Fares 2011).

The second major shift in practice is in the method of flap creation in LASIK which can be achieved using a mechanical microkeratome or the more recently developed femtosecond laser. The rate of intraoperative flap complications using a mechanical microkeratome is approximately 4% (Gimbel 1998; Lin 1999; Stulting 1999; Sugar 2002) and for the femtosecond laser is approximately 3% (Moshirfar 2010). The thickness of a microkeratome LASIK flap is in the order of 150 to 180 microns whereas a femtosecond laser flap is approximately 90 to 110 microns and is more predictable (Binder 2004; Slade 2008; Sutton 2010). It is postulated that as a result the biomechanical properties of the cornea following SBK are equivalent to those following PRK (Dawson 2008). Femtosecond lasers also cause less epithelial injury (Moshirfar 2010). There is some evidence that femtosecond laser flaps result in better uncorrected distance visual acuity (UDVA) postoperatively than mechanical microkeratomes (Durrie 2005; Tran 2005), however a recent randomised controlled trial (RCT) comparing microkeratome and femtosecond laser flap creation for LASIK did not show any difference in efficacy, accuracy and safety measures in the early and mid‐term follow‐up, although it was found that femtosecond flaps may induce fewer aberrations (Zhang 2011).

This updated version of the review includes data from trials that use these newer technologies and re‐evaluates the evidence for the relative efficacy and safety of these procedures.

Objectives

To compare laser‐assisted in‐situ keratomileusis (LASIK) versus photorefractive keratectomy (PRK) for the correction of myopia by examining post‐treatment uncorrected visual acuity, refractive outcome, loss of best spectacle‐corrected visual acuity, pain scores, flap complications in LASIK, subepithelial haze, adverse events, quality of life indices and higher order aberrations.

Methods

Criteria for considering studies for this review

Types of studies

We included only randomised controlled trials (RCTs).

Types of participants

We only considered trials in which the participants were men and women over 18 years of age and under 60 years of age undergoing laser‐assisted in‐situ keratomileusis (LASIK) or photorefractive keratectomy (PRK) for any degree of myopia. Sub‐Bowmans keratomileusis (SBK) is considered as sufficiently similar to LASIK that data from trials comparing SBK and PRK are included. Participants under 18 years of age were excluded due to the frequent change in refractive error still occurring in this age group. Participants over 60 years of age were excluded on the basis that some degree of cataract is observed in many of these people and corneal refractive procedures will not correct aberrations or reduced visual acuity caused by cataract.

As most people with myopia have some degree of astigmatism this review included individuals with up to 3 dioptres (D) of myopic astigmatism. People undergoing treatment for correction of refractive errors other than primary myopia, for example post corneal graft, were excluded, as were people with any other co‐existing ocular disease or any systemic disease that is associated with abnormal or impaired wound healing.

Types of interventions

We included studies in which LASIK (including SBK) was compared with PRK for correction of myopia.

Types of outcome measures

See the 'Differences between protocol and review' section for summary of, and justification for, changes to outcome measures for this update.

Effectiveness measures

At two to four weeks, six months and 12 months:

• Proportion of eyes with uncorrected visual acuity (UCVA)

  • 20/15 or better

  • 20/20 or better

• Proportion of eyes within ±0.50 D of target refraction

• Mean postoperative spherical equivalent

Safety measures

At six months or more after treatment, proportion of eyes:

• lost 1 or more lines of best spectacle‐corrected visual acuity (BCVA)

• lost 2 or more lines of BCVA

• with final BCVA of 20/40 or worse

Adverse effects

• Refractive stability

• Pain scores

• Subepithelial haze

• Flap‐related complications in LASIK eyes

• Optical side effects

• Higher order aberrations

Quality of life measures

Any standardised quality of life measured such as the Refractive Status and Vision Profile (RSVP) or National Eye Institute Refractive Quality of Life (NEI‐RQL)

Search methods for identification of studies

Electronic searches

We searched the Cochrane Central Register of Controlled Trials (CENTRAL) 2012, Issue 11, part of The Cochrane Library. www.thecochranelibrary.com (accessed 15 November 2012), Ovid MEDLINE, Ovid MEDLINE In‐Process and Other Non‐Indexed Citations, Ovid MEDLINE Daily, Ovid OLDMEDLINE (January 1946 to November 2012), EMBASE (January 1980 to November 2012), Latin American and Caribbean Literature on Health Sciences (LILACS) (January 1982 to November 2012), the metaRegister of Controlled Trials (mRCT) (www.controlled‐trials.com), ClinicalTrials.gov (www.clinicaltrials.gov) and the WHO International Clinical Trials Registry Platform (ICTRP) (www.who.int/ictrp/search/en). We did not use any date or language restrictions in the electronic searches for trials. We last searched the electronic databases on 15 November 2012.

See: Appendices for details of search strategies for CENTRAL (Appendix 1), MEDLINE (Appendix 2), EMBASE (Appendix 3), LILACS (Appendix 4), mRCT (Appendix 5), ClinicalTrials.gov (Appendix 6), and the ICTRP (Appendix 7).

Searching other resources

We searched the reference lists of the studies included in the review for information about further trials. We also searched the Science Citation Index to find studies that have cited the identified trials. We did not handsearch journals or conference proceedings for this review as considering the resources required this was not felt to be sufficiently likely to identify includable data.

Data collection and analysis

Selection of studies

Two authors working independently assessed the titles and abstracts resulting from the searches. We obtained full‐text copies of all potentially or definitely relevant studies. The two review authors assessed these full‐text copies to determine whether they met the criteria for inclusion in the study.

Data extraction and management

The two authors independently extracted data using a form developed by the Cochrane Eyes and Vision Group. The results were compared and we resolved any discrepancies by discussion. One author entered the data into RevMan (RevMan 2011) and the second author checked the data.

Where the method of randomisation was unclear or where data for outcomes were not available in the published study report we contacted the authors for clarification and additional data. The authors of Forseto 2000, Hjortdal 2005 and Schallhorn 2009 kindly provided such information and additional unpublished data.

Assessment of risk of bias in included studies

Two authors independently assessed risk of bias using the Cochrane Collaboration's tool for assessing risk of bias according to Chapter 8 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). Disagreements were resolved by discussion. We contacted trial authors for clarification on any parameter graded as 'unclear'.

Measures of treatment effect

We used the odds ratio as the measure of effect for dichotomous variables and the mean difference for continuous variables.

Unit of analysis issues

Ideally studies that randomly allocated eyes to treatment should report a paired analysis. In the event, none of the paired studies included in this review did the analysis appropriately. This meant that it was not possible to calculate the intra‐class correlation coefficient and adjust the analyses accordingly (Elbourne 2002). The analyses therefore assume that the response of any eye to one treatment is not related to the response of the fellow eye. This is a conservative assumption as, in the presence of correlation, adjusting for the pairing would have reduced the width of the confidence intervals.

Assessment of heterogeneity

We assessed heterogeneity by examining the graphs (forest plots) to see whether the direction of effect was similar in all studies and whether the confidence intervals for the individual study estimates overlapped. We also considered the I ² statistic (Higgins 2003). We took an I² statistic value of 50% or more to indicate substantial inconsistency in study results such that a pooled result may not be informative.

Data synthesis

We pooled data using a random‐effects model, unless fewer than three trials were available for analysis, or the number of events was low, in which case we used a fixed‐effect model.

Subgroup analysis and investigation of heterogeneity

We did two subgroup analyses. Firstly, we compared effects in studies that recruited people with low to moderate myopia compared to those that recruited people with moderate to high myopia and secondly, we compared effects in studies conducted from 2008 onwards with those done prior to this as surgical techniques had improved (Table 2).

1. Equipment and technology used.

Study	MMC for PRK	PRK performed using WFG?	Methods of LASIK flap creation	LASIK performed using WFG?	Excimer laser used	Nomogram used
Wang 1997	No	No	Chiron Automated Corneal Shaper microkeratome	No	Keracor 116 excimer laser (Chiron Vision Corporation)	Manufacturer
SUMMIT 1998	No	No	Chiron Automated Corneal Shaper microkeratome	No	Summit Apex	Manufacturer
el Danasoury 1999	No	No	Chiron Automated Corneal Shaper microkeratome	No	Nidek EC‐5000	PRK: Manufacturer LASIK: Personalised
el Maghraby 1999	No	No	Chiron Automated Corneal Shaper microkeratome	No	Summit OmniMed I	Manufacturer
Forseto 2000	No	No	Chiron Automated Corneal Shaper microkeratome	No	VISX 20/20B	Unclear
Hjortdal 2005	No	No	Supratome microkeratome (Schwind)	No	MEL‐70 G‐scan flying spot excimer laser (Meditec‐Aesclepion)	Manufacturer
Wallau 2008	Yes	Yes	Moria M2 microkeratome	Yes	Alcon LADARWave 4000	Manufacturer
Durrie 2008	No	Yes	IntraLase FS60	Yes	Alcon LADARWave 4000	Not stated
Schallhorn 2009	Unclear	Yes	IntraLase FS15	Yes	VISX Star S4 CustomVue	Manufacturer
Barreto 2010	No	Yes	MK2000 microkeratome (NIDEK Co Ltd)	Yes	OPDCAT platform (NIDEK Co Ltd, Gamagori, Japan)	Not stated
Moshirfar 2010	Yes	Yes	IntraLase FS60	Yes	VISX Star S4 CustomVue	Local data used to modify nomogram
Hatch 2011	No	Yes	IntraLase FS60	Yes	VISX Star S4 CustomVue	Not stated
Manche 2011	No	Yes	IntraLase FS60	Yes	VISX Star S4 CustomVue	Not stated

LASIK: laser‐assisted in‐situ keratomileusis 
 MMC: mitomycin C 
 PRK: photorefractive keratectomy 
 WFG: wavefront‐guided

We did not explore heterogeneity further.

Sensitivity analysis

Two trials (Schallhorn 2009 and Wang 1997) had significant problems with allocation concealment and we repeated relevant analyses excluding these trials.

Results

Description of studies

Results of the search

The original electronic searches performed in the first version of this review identified 949 reports up to 2005. There were six RCTs from this period that met the inclusion criteria (el Danasoury 1999; el Maghraby 1999; Forseto 2000; Hjortdal 2005; SUMMIT 1998; Wang 1997). An update search was run in November 2012 which yielded a total of 578 records. The Trials Search Co‐ordinator scanned the search results and removed 336 records which were not relevant to the scope of the review. We assessed a total of 242 records against the inclusion criteria for the review. We obtained full‐text copies of seven reports and all were suitable for inclusion in the review (Barreto 2010; Durrie 2008; Hatch 2011; Moshirfar 2010; Schallhorn 2009; Wallau 2008; Manche 2011). No ongoing studies or studies awaiting classification were identified.

Included studies

Below is a summary of the included studies. Further details can be found in the 'Characteristics of included studies' table.

Types of participants

Participants in all trials were men and women aged 18 or over with stable refraction for one year or more. Exclusion criteria included previous refractive or other ocular surgery, central corneal thickness of less than 490 microns by ultrasound pachymetry, keratoconus or suspected keratoconus on corneal topography, active ocular disease, dry eyes and systemic diseases likely to affect corneal wound healing (for example, connective tissue disease). The range of myopia treated varied between studies (Table 3).

2. Pre‐treatment myopia and astigmatism.

Study	Low myopia (0 to < ‐6 dioptres)	High myopia (‐6 to ‐15 dioptres)	All degrees of myopia (0 to ‐15 dioptres)	Astigmatism (dioptres)
Barreto 2010	Not specified; participants had "myopic astigmatism"
Durrie 2008	‐2.00 to ‐5.75			0.00 to ‐2.75
el Danasoury 1999	‐2.25 to ‐5.50			0.00 to ‐2.25
el Maghraby 1999			‐2.30 to ‐8.10	0.00 to ‐1.75
Forseto 2000;	‐2.00 to ‐6.00*			0.00 to ‐1.75
Hatch 2011			‐0.60 to ‐8.50	0.00 to 2.50
Hjortdal 2005		‐6.00 to ‐8.00		< 1.5
Manche 2011			‐0.75 to ‐8.13	0.00 to 3.50
Moshirfar 2010			‐0.25 to ‐10.00	0.00 to 3.50
Schallhorn 2009	up to ‐6.00*			up to 3.00*
SUMMIT 1998		‐6.00 to ‐14.38		<= 2.00*
Wallau 2008			‐1.46 to ‐6.96 (SE)	Not reported but participants had "myopic astigmatism"
Wang 1997	‐1.25 to ‐6.00*			Not stated

* From inclusion criteria

Types of interventions

All eyes underwent LASIK or PRK as a day case procedure under topical anaesthesia.

The surgical procedure used to perform PRK was standard between studies. Mitomycin C (MMC) was used as an adjunct to prevent post‐PRK corneal haze in some contemporary studies. The LASIK technique varied widely. Flap creation was performed using a microkeratome in early studies whereas in contemporary studies this was performed using the femtosecond laser. The excimer laser manufacturer varied between studies but not within studies. Treatment nomograms varied between studies as did the target refraction although in the majority of trials the target refraction was emmetropia. From 2008 onwards, both PRK and LASIK were performed using wavefront‐guided technology. Table 2 summarises the techniques used in each study.

Types of outcomes measures

Table 4 summarises which trials reported which outcome measures.

3. Outcome reporting matrix.

UCVA better or equal to 20/15

UCVA better or equal to 20/20

Within 0.05 D of target refraction

Mean spherical equivalent

Loss 1+ line BCVA

Loss 2+ lines BCVA

Final BCVA < 20/40

Study

2 to 4 weeks

6 months

12 months

2 to 4 weeks

6 months

12 months

2 to 4 weeks

6 months

12 months

2 to 4 weeks

6 months

12 months

6 months or more

6 months or more

6 months or more

Barreto 2010

x

Durrie 2008

x

x

x

x

x

x

x

x

el Danasoury 1999

x

x

x

x

x

x

x

el Maghraby 1999

x

x

x

x

x

x

x

x

x

x

x

Forseto 2000;

x

x

x

x

x

x

Hatch 2011

x

x

x

x

x

x

x

x

x

x

Hjortdal 2005

x

x

x

x

x

x

x

x

x

x

Manche 2011

x

x

x

x

x

x

x

x

x

x

x

x

x

x

x

Moshirfar 2010

x

x

x

x

x

x

Schallhorn 2009

x

x

x

x

x

x

x

x

x

x

SUMMIT 1998

x

x

x

x

x

x

x

x

Wallau 2008

x

x

x

x

x

Wang 1997

x

x

x

x

x

BCVA: best spectacle‐corrected visual acuity 
 D: dioptres 
 UCVA: uncorrected visual acuity

Unit of analysis

In two studies, participants were randomly assigned to LASIK or PRK and then received the same treatment in both eyes (Schallhorn 2009; Wang 1997). Two studies included only one eye of each patient in the study (Hjortdal 2005; SUMMIT 1998). In SUMMIT 1998 the eye to be included was decided ad hoc by the principal investigator. In Hjortdal 2005 it is not clear how this decision was made. The remaining studies allocated one eye to LASIK and the fellow eye to PRK. Durrie 2008 used a randomisation system which accounted for ocular dominance and ensured an equal number of dominant eyes in each treatment group. Hatch 2011 and Manche 2011 also randomised the dominant eye to the first treatment. Wallau 2008 randomised right eyes to the first treatment. el Danasoury 1999 and el Maghraby 1999 both randomised the eye to be treated first and the procedure to be performed. In the remaining studies (Barreto 2010; Forseto 2000; Moshirfar 2010) it is unclear how the eye that was to be randomised was chosen.

Excluded studies

We excluded 19 studies and reasons for exclusion are provided in the 'Characteristics of excluded studies' table.

Risk of bias in included studies

'Risk of bias' assessment is summarised in Figure 1 and Figure 2

1.

'Risk of bias' graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

2.

'Risk of bias' summary: review authors' judgements about each risk of bias item for each included study.

Allocation

Sequence generation

Sequence generation was adequate in almost all studies. In Barreto 2010 and Wang 1997 it was not clear how eyes were randomised to treatments and in Schallhorn 2009 participants were ranked by refractive error then assigned sequential numbers; odd numbers were assigned to one treatment and even numbers to the other.

Allocation concealment

Allocation concealment was clearly described in seven studies and not reported in three studies (Figure 2). Allocation was not concealed in Schallhorn 2009 (as discussed above) and Wang 1997. In Wang 1997 a substantial number of people with eyes randomised to undergo LASIK refused this treatment and insisted on having PRK. The reason cited was that the patients could not afford LASIK.

Blinding

Blinding (masking) of participants and personnel (performance bias)

None of the studies attempted to mask, or reported efforts to mask, participants and personnel and so we considered all studies at high risk of performance bias.

Blinding (masking) of outcome assessment (detection bias)

Masking of outcome assessment was reported in three studies (Hatch 2011; SUMMIT 1998; Wallau 2008). In Hatch 2011 the method of masking was not reported in detail but the study was described as masked and the following statement made: "After the study was completed, the results were compiled and the data unmasked for statistical analysis" which implies that the data were masked until the point of the statistical analysis. In SUMMIT 1998"Preoperative and follow‐up visits included a detailed ophthalmologic examination with manifest refraction by two independent observers at each visit." and "All [corneal topography] maps were graded by two masked observers".Wallau 2008 made the most convincing efforts to mask visual acuity assessment "During follow‐up examinations, a single examiner was unaware of which procedure was done in each eye and slit‐lamp microscopy was always the last examination to be performed at each appointment", however the report did not indicate whether or not this attempt to mask was successful.

We considered the masking of visual acuity separately to other outcomes but in general the assessment was the same. The exception was SUMMIT 1998 where it was specifically stated that some non visual acuity outcomes were masked but it was not clear whether or not visual acuity assessment was masked or not.

Incomplete outcome data

In general follow‐up in the individual trials was good. Eight of the 12 included trials randomly allocated eyes to treatment. This meant that differential loss to follow‐up was not possible therefore we graded most trials as being at low risk of attrition bias.

Selective reporting

See outcome reporting matrix (Table 4). In general the reasons why data were not reported were unclear, although not all studies reported all outcomes.

Effects of interventions

See: Table 1

Effectiveness measures

Uncorrected visual acuity (UCVA) 20/15 or better

Four trials reported UCVA 20/15 or better two to four weeks after treatment (n = 566) (Analysis 1.1). All four trials reported that more participants treated with laser‐assisted in‐situ keratomileusis (LASIK) achieved this very good vision compared to people treated with photorefractive keratectomy (PRK) (pooled odds ratio 5.89 95% confidence interval (CI) 3.34, 10.39). This pooled odds ratio corresponds to a risk ratio of 3.30 (95% CI 2.43 to 4.15) assuming a risk of 0.16 in the PRK group (based on median risk in the included studies).

1.1. Analysis.

Comparison 1 LASIK versus PRK, Outcome 1 UCVA of 20/15 or better at two to four weeks post‐treatment.

Five trials reported UCVA 20/15 or better six months after treatment (n = 682) (Analysis 1.2). All five trials provided effect estimates consistent with LASIK or PRK being more effective (confidence intervals included 1) and the pooled odds ratio was 1.13 (95% CI 0.75 to 1.69). This pooled odds ratio corresponds to a risk ratio of 1.08 (95% CI 0.82 to 1.35) assuming a risk of 0.36 in the PRK group (based on median risk).

1.2. Analysis.

Comparison 1 LASIK versus PRK, Outcome 2 UCVA of 20/15 or better at six months post‐treatment.

Two trials reported UCVA 20/15 or better at 12 months after treatment (n = 372) (Analysis 1.3). The pooled odds ratio was 1.08 (95% CI 0.58 to 2.00) The risk ratio was 1.01 (95% CI 0.91 to 1.07) assuming a control group risk of 0.87.

1.3. Analysis.

Comparison 1 LASIK versus PRK, Outcome 3 UCVA of 20/15 or better at twelve months post‐treatment.

UCVA 20/20 or better

Eight trials reported UCVA 20/20 or better two to four weeks after treatment (n = 1079) (Analysis 1.4). All eight trials reported that more participants treated with LASIK achieved this very good vision compared to people treated with PRK (pooled odds ratio 3.69, 95% CI 2.55 to 5.36).This pooled odds ratio corresponds to a risk ratio of 1.85 (95% CI 1.62 to 2.05) assuming a risk of 0.37 in the PRK group (based on median risk).

1.4. Analysis.

Comparison 1 LASIK versus PRK, Outcome 4 UCVA of 20/20 or better at two to four weeks post‐treatment.

Ten trials reported UCVA 20/20 or better six months after treatment (n = 1113) (Analysis 1.5). Five trials found more people receiving LASIK achieved 20/20 or better, four trials found more people receiving PRK achieved this good vision, and one trial found no difference. However, for nine out of 10 of these trials the confidence intervals included 1 and therefore the results were consistent with greater beneficial effect of either LASIK or PRK. The pooled odds ratio was 1.41 (95% CI 1.00 to 2.00). This pooled odds ratio corresponds to a risk ratio of 1.06 (95% CI 1 to 1.10) assuming a risk of 0.82 in the PRK group (based on median risk).

1.5. Analysis.

Comparison 1 LASIK versus PRK, Outcome 5 UCVA of 20/20 or better at six months post‐treatment.

Seven trials reported UCVA 20/20 or better at 12 months after treatment (n = 1007) (Analysis 1.6). Five out of these seven trials found in favour of LASIK and the pooled odds ratio was 1.64 (95% CI 1.10 to 2.45). This pooled odds ratio corresponds to a risk ratio of 1.17 (95% CI 1.04 to 1.29) assuming a risk of 0.63 in the PRK group (based on median risk).

1.6. Analysis.

Comparison 1 LASIK versus PRK, Outcome 6 UCVA of 20/20 or better at 12 months post‐treatment.

±0.50 D of target refraction

Six trials reported whether or not participants were within 0.50 D of their target refraction two to four weeks after treatment (n = 455) (Analysis 1.7). There was substantial heterogeneity in the results (I² = 58%). Results ranged from an odds ratio of 0.26 in favour of PRK (Hjortdal 2005) to an odds ratio of 7.07 in favour of LASIK (Manche 2011).

1.7. Analysis.

Comparison 1 LASIK versus PRK, Outcome 7 Within 0.50 D of target refraction at two to four weeks post‐treatment.

Eight trials reported this outcome at six months after treatment (n = 567) (Analysis 1.8). Again individual study results were variable but there was no statistical evidence of inconsistency (I² = 0%) and the pooled odds ratio was 1.11 (95% CI 0.74 to 1.67). The corresponding risk ratio was 1.03 (95% CI 0.90 to 1.14) based on a median risk of 0.69 in the PRK group.

1.8. Analysis.

Comparison 1 LASIK versus PRK, Outcome 8 Within 0.50 D of target refraction at six months post‐treatment.

Seven trials reported achievement of target refraction 12 months after treatment (n = 1007 participants) (Analysis 1.9). Trial results varied from an odds ratio of 0.44 in favour of PRK (Hjortdal 2005) to an odds ratio of 3.76 in favour of LASIK (Schallhorn 2009). Five out of these seven trials found in favour of LASIK and the pooled odds ratio was 1.45 (95% CI 0.99 to 2.10). The corresponding risk ratio was 1.06 (95% CI 1 to 1.10) based on a median risk of 0.0.83 in the PRK group.

1.9. Analysis.

Comparison 1 LASIK versus PRK, Outcome 9 Within 0.50 D of target refraction at 12 months post‐treatment.

Mean spherical equivalent

Nine trials reported mean spherical equivalent two to four weeks after treatment (n = 1041) (Analysis 1.10). The results were inconsistent (I² = 83%) and a pooled value is not appropriate here. Differences ranged from ‐0.60 D to 0.14 D.

1.10. Analysis.

Comparison 1 LASIK versus PRK, Outcome 10 Mean postoperative spherical equivalent at two to four weeks post‐treatment.

Nine trials reported mean spherical equivalent six months after treatment (n = 1024) (Analysis 1.11). Again the results were inconsistent (I² = 59%). Differences ranged from ‐0.26 D to 0.60 D.

1.11. Analysis.

Comparison 1 LASIK versus PRK, Outcome 11 Mean postoperative spherical equivalent at six months post‐treatment.

Six trials reported mean spherical equivalent 12 months after treatment (n = 599) (Analysis 1.12). Results were consistent (I² = 0%) with a pooled mean difference of ‐0.01 (95% CI ‐0.06 to 0.04).

1.12. Analysis.

Comparison 1 LASIK versus PRK, Outcome 12 Mean postoperative spherical equivalent at 12 months post‐treatment.

Stability of refraction

Nine studies examined refractive stability (Table 5). Only two of these studies reported a change in mean spherical equivalent refraction of more than 0.25 D over the study period. Durrie 2008 reported a change of ‐0.27 D in the PRK group between month one and six but no significant shift in the LASIK group. SUMMIT 1998 found a change of ‐0.89 D in the PRK group and ‐0.55 D in the LASIK group between month one and six.

4. Refractive stability.

Study	Follow‐up duration	Findings
Durrie 2008	6 months	PRK: mean SE refraction changed from +0.10 at 1 month post‐treatment to ‐0.17 at 6 months (change of ‐0.27 D) LASIK: mean SE refraction changed from ‐0.02 at 1 month post‐treatment to +0.08 at 6 months (change of +0.1 D)
Hatch 2011	6 months	PRK: mean SE refraction changed from +0.125 at 1 month post‐treatment to +0.113 at 6 months (change of ‐0.012 D) LASIK: mean SE refraction changed from +0.236 at 1 month post‐treatment to +0.049 at 6 months (change of ‐0.187 D)
el Maghraby 1999	24 months	PRK: initial myopic regression between 0 to 3 months but no clinically significant regression between 3 months to 24 months LASIK: no clinically significant regression
Manche 2011	12 months	PRK: mean SE refraction changed from ‐0.50 at 1 month post‐treatment to ‐0.28 at 12 months (change of +0.22 D) LASIK: mean SE refraction changed from ‐0.29 at 1 month post‐treatment to ‐0.39 at 12 months (change of ‐0.1 D)
Moshirfar 2010	6 months	PRK: mean SE refraction changed from +0.04 D at 3 month post‐treatment to +0.08 D at 6 months (change of +0.04 D) LASIK: mean SE refraction changed from +0.08 D at 1 month post‐treatment to +0.0002 D at 6 months (change of ‐0.0798 D)
Schallhorn 2009	12 months	PRK: mean refractive SE was ‐0.5 at 1 month and ‐0.27 at 12 months (change of +0.23 D) LASIK: mean refractive SE was ‐0.04 at 1 month and the same at 12 months
SUMMIT 1998	6 months	PRK: mean SE refraction changed from ‐0.14 at 1 month post‐treatment to ‐1.03 at 6 months (change of ‐0.89 D) LASIK: mean SE refraction changed from ‐0.74 at 1 month post‐treatment to ‐1.29 at 6 months (change of ‐0.55 D)
Wang 1997	12 months	Less myopic regression was seen following LASIK than following PRK. Refractive status tended to be stable 3 months after LASIK whereas it sometimes took 6 months to get a relatively stable refraction after PRK.
Wallau 2008	6 months	There was no significant regression in either group PRK: mean SE changed slightly from +0.61 D at 1 month to +0.56 at 6 months (change of ‐0.05 D) LASIK: mean SE changed only slightly from +0.49 D at 1 month to +0.52 D at 6 months (change of ‐0.03 D)

D: dioptres 
 LASIK: laser‐assisted in‐situ keratomileusis 
 PRK: photorefractive keratectomy 
 SE: spherical equivalent

Safety measures

Lost 1 or more lines of best corrected visual acuity (BCVA)

Six trials reported whether or not participants lost 1 or more lines of BCVA six or more months after treatment (n = 746) (Analysis 1.13). Two trials reported no events (Forseto 2000; Hatch 2011). For the other four trials there was no consistent pattern; the pooled Peto odds ratio was 0.88 (95% CI 0.51 to 1.50).

1.13. Analysis.

Comparison 1 LASIK versus PRK, Outcome 13 Lost one or more lines of BCVA at six months or more post‐treatment.

Lost 2 or more lines of BCVA

Eleven trials reported whether or not participants lost 2 or more lines of BCVA six or more months after treatment (n = 1446 ) (Analysis 1.14). Overall, 34 of these 1446 people (2.35%) treated with LASIK or PRK lost 2 or more lines of BCVA. Five trials reported no events (Durrie 2008; el Danasoury 1999; Forseto 2000; Hatch 2011; Manche 2011). For the other six trials, three trials found odds ratios close to 1 and three trials found odds ratios favouring LASIK. No individual trial found a statistically significant effect but the pooled Peto odds ratio favoured LASIK: 0.47 (95% CI 0.23 to 0.98).

1.14. Analysis.

Comparison 1 LASIK versus PRK, Outcome 14 Lost two or more lines of BCVA at six months or more post‐treatment.

Final BCVA of 20/40 or worse

Six trials reported final BCVA of 20/40 or worse six or more months after treatment (n = 442) (Analysis 1.15). There were only two events, both in the LASIK arm of SUMMIT 1998.

1.15. Analysis.

Comparison 1 LASIK versus PRK, Outcome 15 Final BCVA of 20/40 or less at six months or more post‐treatment.

Subgroup analyses

Two sets of subgroup analyses are presented (Appendix 8; Appendix 9).

In Appendix 8 the studies are divided up into those recruiting patients with different degrees of myopia. We divided studies into those treating low to moderate myopia (0 to ‐6 D) or moderate to high myopia (‐ 6 D to ‐15 D). In general there was no strong statistical evidence of any major differences in effect between LASIK and PRK at different levels of myopia. The exception was mean spherical equivalent at two to four weeks after treatment. There appeared to be little difference between LASIK and PRK in people with low myopia (0 to ‐6 D) but a pooled mean difference of ‐0.56 D in people with high myopia (‐6 D to ‐15 D). However, only two trials contributed to the high myopia group and this difference in effect between subgroups was not evident at six or 12 months.

In Appendix 9 the studies are divided up into those conducted before 2008 and those conducted from 2008 onwards. There were some subgroup differences but no particular pattern to these. In some analyses, the two techniques were more similar ‐ that is likely to show an pooled effect of around 1 for dichotomous outcomes, and 0 for continuous outcomes ‐ in more recent studies (Analysis 1.5; Analysis 1.6; Analysis 1.10; Analysis 1.11; Analysis 1.12) and in other analyses LASIK/PRK appeared more different in more recent studies (Analysis 1.4; Analysis 1.7; Analysis 1.8; Analysis 1.9; Analysis 1.13; Analysis 1.14).

Sensitivity analyses

Due to concerns as to the potential high risk of bias in Schallhorn 2009 and Wang 1997 we excluded these studies and repeated the relevant analyses (Appendix 10).

In general excluding these studies did not change the effect estimates although in Analysis 1.6 the effect became non‐significant, probably as a result of lower numbers included in the analysis.

Adverse effects

Pain scores

One study used questionnaires to assess intraoperative pain (el Danasoury 1999) and another two studies used questionnaires to assess postoperative pain (Durrie 2008; el Maghraby 1999). These found that intraoperative pain was less with PRK and that postoperative pain was less after LASIK. The findings are summarised in Table 6.

5. Pain scores.

Study	Method	Findings
Durrie 2008	Subjective questionnaire developed by the study sponsor AMO	Eyes that underwent PRK were significantly more painful during the first 3 days than LASIK eyes
el Maghraby 1999	Questionnaire at 1 day post‐treatment	PRK: mild pain in 3 of 33 participants (8%), moderate in 7 of 33 participants (23%) and severe in 23 of 33 participants (69%) LASIK: 26 of 33 participants (81%) reported no pain on the first day post‐treatment
Wallau 2008	Questionnaire on pain and satisfaction at postoperative day 1, 3, 4, 5 then at 3 months and 6 months	PRK: on day 1 postoperative mean pain score was 5/10 (range 0 to 10). By day 5 this had reduced to 0/10 in all PRK eyes. LASIK: on day 1 postoperative the mean pain score was 0/10 (range 0 to 2). By day 5 this was 0 in all LASIK eyes.

LASIK: laser‐assisted in‐situ keratomileusis 
 PRK: photorefractive keratectomy

Subepithelial haze at six to 12 months post‐PRK

Nine studies reported data for subepithelial haze post‐PRK. These are summarised in Table 7.

6. Subepithelial haze at 6 to 12 months post‐treatment with PRK.

Study	Time of examination	Grade 0 (clear)	Grade +1/2 to +1	Grade +2 (mild)	Grade 3 (moderate)	Grade 4 (severe)
Barreto 2010	12 months	100%	0%	0%	0%	0%
el Danasoury 1999	6 months	41.7%	54.2%	4.2%	0%	0%
el Danasoury 1999	12 months	54.2%	37.5%	4.2%	0%	0%
el Maghraby 1999	12 months	83% *Grade 0 to +1	*	13%	0%	3%
Hatch 2011	6 months	96.1%	3.9%	0%	0%	0%
Moshirfar 2010	6 months	95.1%	4.9%	0%	0%	0%
Schallhorn 2009	12 months	99.5%	0.5%	0%	0%	0%
SUMMIT 1998	6 months	45.6%	44.1%	5.9%	4.4%	0%
Wallau 2008	6 months	58%	42%	0%	0%	0%

Flap‐related complications in LASIK eyes

Six studies reported flap‐related complications. The rate of incidence ranged from 0.7% to 15% but only one participant lost 2 or more lines of BCVA as a result of a flap complication. The overall rate of flap complications was 3.8%. These results are summarised in Table 8.

7. Incidence of flap‐related complications in LASIK.

Study	Complication rate	Flap complication	Management	Outcome
el Danasoury 1999	1 flap complication in 26 LASIK procedures (3.9%)	Temporally decentred flap	None	Mild irregular astigmatism BCVA 20/25
el Maghraby 1999	2 flap complications in 33 LASIK procedures (6.1%)	1 x partially dislocated flap 1 x fully dislocated flap	Replaced Replaced	No complications. No loss of BCVA Epithelial ingrowth requiring surgical removal. Final BCVA 20/30 (loss of 4 lines from pre treatment BCVA).
Hatch 2011	9 flap complications in 26 procedures, 4 of which significant (15%)	1 x flap debris and microstriae 1 x diffuse lamellar keratitis and microstriae 1 x epithelial cyst at the edge of flap 4 x microstriae 2 x intraoperative flap tears	Flap tears: first was observed and developed mild flap edge scarring by 6 months that had no significant effect on visual function. The second was retreated with PRK at 3 months.	No loss of UDVA or CDVA and all resolved by 6 months
Moshirfar 2010	4 cases of flap microstriae in 57 LASIK procedures (7%)	4 x flap microstriae affecting UCVA	Not stated	All 4 cases lost 1 line of CDVA
SUMMIT 1998	3 flap complications in 115 LASIK procedures (2.6%)	1 x stopping of microkeratome in middle of pass 1 x free flap 1 x thin flap	Replaced and treatment postponed Procedure completed and flap replaced Flap replaced and procedure postponed	No adverse effect on outcome or BCVA reported
Wang 1997	1 flap complication in 137 procedures (0.7%)	Free flap	Unknown	Unknown

BCVA: best spectacle‐corrected visual acuity 
 CDVA: corrected distance visual acuity 
 LASIK: laser‐assisted in‐situ keratomileusis 
 UDVA: uncorrected distance visual acuity

Optical side effect

Six studies examined optical side effects such as glare, halo and monocular diplopia. Some participants in both groups experienced optical side effects but only el Maghraby 1999 found any difference between interventions with symptoms arising more commonly following PRK. Their findings are summarised in Table 9.

8. Optical side effects.

Study	Method	Findings
Durrie 2008	Subjective questionnaire developed by the study sponsor AMO	There was no difference in glare or halos between groups
el Maghraby 1999	Questionnaire assessment at 2 years post‐treatment	LASIK: 21% of participants reported glare, halos or flare PRK: 35% of participants reported glare, halos or flare
Manche 2011	Subjective symptom questionnaire administered pre‐op and at 1,3,6 and 12 months post‐op	Glare increased in the LASIK and PRK groups under both day and night conditions at 1 month before improving close to preoperative levels (Figure 11). There was no significant difference in subjective glare symptoms between the LASIK and PRK groups under both day and night conditions after month 1. Subjective symptoms of haze increased in both the LASIK and PRK groups. The 6‐month increase in mean haze symptoms in the LASIK subgroup was skewed by a single eye that experienced severe subjective symptoms of haze (9 of 10). This eye had an UCVA of 20/16. Subjective symptoms of halos increased in the LASIK and PRK groups at 1 month before improving close to preoperative levels. There was no significant difference in subjective halo symptoms between the LASIK and PRK groups after month 1. Subjective symptoms of vision clarity in day and night conditions declined at 1 month from preoperative levels in both the LASIK and PRK groups but improved steadily at 12 months (Figure 14). The PRK group demonstrated a significant decline as compared to the LASIK group at 1 month under both the day (P = 0.001) and night (P = 0.001) conditions. There was no difference between the LASIK and PRK groups at 3, 6, and 12 months
Schallhorn 2009	Questionnaire at all time points	LASIK and PRK: at 12 months there was a reduction in glare and halos versus pre‐operatively
SUMMIT 1998	Questionnaire assessment of glare, halo and monocular diplopia pre and 6 months post‐treatment	Glare: there was no statistically significant increase in glare post‐treatment in PRK or LASIK groups. There was no significant difference in post‐treatment glare between the 2 groups. Halo: the PRK group experienced significantly more halos post‐treatment versus pretreatment. There was no significant increase in halos in the LASIK group. Monocular diplopia: for both PRK and LASIK groups the difference in average diplopia index before and after treatment was statistically significant When changes in glare and halo from before and surgery to after surgery were pooled as a glare‐halo index, the PRK group showed a significantly greater likelihood of demonstrating an increase in symptoms than the LASIK group
Wallau 2008	Questionnaire of patient satisfaction at 1 month, 3 months and 6 months post‐treatment	At 6 months follow‐up: Glare: reported in 35% of LASIK eyes and 30% of PRK eyes Photophobia: reported in 61% of LASIK eyes and 63% of PRK eyes Visual fluctuation: reported in 32% of LASIK eyes and 24% of PRK eyes

LASIK: laser‐assisted in‐situ keratomileusis 
 PRK: photorefractive keratectomy

Higher order aberrations (HOAs)

Seven studies reported data for this outcome which is summarised in Table 10. All but one of these studies show that both LASIK and PRK result in a statistically significant increase in HOAs.The exception is Durrie 2008 which found that HOAs were reduced in both PRK and LASIK compared with pre‐operatively. When postoperative HOAs were compared between LASIK and PRK, only one study found a statistically significant difference, there being fewer HOAs in the LASIK group (Schallhorn 2009).

9. Higher order aberrations.

Study	Higher order aberrations/wavefront aberrometer	Modulation transfer function (MTF)
Barreto 2010	Total HOAs increased after surgery in both WFG LASIK and WFG PRK groups (P < 05). There was no difference in total HOAs between LASIK and PRK at 1, 3, 6 and 12 months post‐treatment. Mean coma and mean spherical aberration also increased following treatment without a significant difference between WFG LASIK and WFG PRK.
Durrie 2008	At 6 months post‐treatment total RMS higher order aberrations improved in both groups (‐0.18 μm SBK versus ‐0.29 μm PRK). There was no statistically significant difference between the groups.	There was no difference between either groups at 6 months postoperative. Neither group had a significant increase in the MTF (i.e. no increase in optical quality of the eye)
Hatch 2011	There were no significant differences between groups in any HOAs throughout the study and at 6 months. Total HOAs (P = 0.008), spherical (P = 0.002) and coma (P = 0.008 at 3 months; P = 0.024 at 6 months) were significantly increased compared with preoperative conditions.
Hjortdal 2005	Both PRK and LASIK caused an increase in coma‐like and spherical aberrations that remained constant for 7 years. No significant changes in other higher‐order aberrations were observed.
Manche 2011	At 12 months there was a significant increase in coma (P = 0.014) and total HOA (P = 0.04) for the LASIK group as compared to preoperative levels. In the PRK group, there was a significant increase in spherical aberration (P = 0.007) from the preoperative level. The mean total HOAs increased from the preoperative levels in all measured postoperative intervals (1, 3, 6, and 12 months) for both the LASIK and PRK groups. At 1 month, the PRK group demonstrated a statistically significant increase in the mean total HOA (P = 0.04) but by 3 months was similar to the corresponding measurement in the LASIK group. However, there was no difference in total HOA between the LASIK and PRK groups at 12 months postoperatively.
Moshirfar 2010	The mean postoperative HOA RMS value was 0.45 μm in the PRK group and 0.59 μm in the LASIK group (P = 0.012), representing a factor increase of 1.22 and 1.74, respectively from the pre‐operative values. In the PRK group, coma increased from 0.22 μm preoperatively to 0.28 μm 6 months postoperatively; trefoil decreased from 0.189 μm to 0.136 μm (P = 0.004), respectively; and spherical aberration increased from 0.048 μm to 0.244 μm, respectively (P < 0.05). In the LASIK group, coma increased from 0.160 mm preoperatively to 0.307 mm 6 months postoperatively (P = 0.002); trefoil increased from 0.170 mm to 0.177 mm (P = 0.675), respectively; and spherical aberration increased from 0.070 mm to 0.292 mm, respectively (P < 0.05).
Schallhorn 2009	The change in HOAs (assessed using the RMS value expressed in microns) was evaluated at 12 months post‐treatment with a 6 mm pupil. Both PRK and LASIK groups showed a significant increase in HOAs RMS versus pre treatment (PRK +0.1208, LASIK +0.0255). This increase was significantly less in the LASIK cohort.
Wallau 2008	The following were assessed for a 6.5 mm pupil: total RMS, defocus and astigmatism, total higher order aberrations, coma, spherical aberration and other aberrations (defined as HOAs up to 5th order excluding coma and spherical aberration). Zernike co‐efficients were reported as absolute values. Total RMS, defocus and astigmatism values were significantly higher following LASIK at 1, 3 and 6 months post‐treatment. Mean HOAs, coma and spherical aberration were all increased postoperatively in both treatment groups. "Other aberrations" were significantly greater in the LASIK group at 3 and 6 months post‐treatment.

HOA: higher order aberration 
 LASIK: laser‐assisted in‐situ keratomileusis 
 MTF: modulation transfer function 
 PRK: photorefractive keratectomy 
 RMS: root mean square 
 SBK: sub‐Bowmans keratomileusis 
 WFG: wavefront‐guided

Quality of life measures

No studies reported data on quality of life.

Discussion

Summary of main results

The principal findings of this updated review are as follows.

• Visual recovery is faster following laser‐assisted in‐situ keratomileusis (LASIK) than photorefractive keratectomy (PRK).

• There is weak evidence that visual acuity at six months and 12 months may be superior with LASIK.

• There is little evidence to suggest any difference in accuracy between LASIK and PRK at 12 months post‐treatment.

• There is little evidence to suggest any difference in safety between LASIK and PRK at six months or more post‐treatment.

• Sub‐analysis of data from studies from 2008 onwards that used modern techniques revealed no significant difference between the contemporary forms of these techniques for any of the efficacy or safety outcomes beyond the two to four‐week time point.

In the previous version of this review we found that visual recovery is faster following LASIK than PRK, which is confirmed again in this update. We previously found that there was no clear difference in the efficacy of the two procedures. The addition of data from seven new trials (Barreto 2010; Durrie 2008; Hatch 2011; Moshirfar 2010; Schallhorn 2009; Wallau 2008; Manche 2011) has not altered these findings. The most significant finding of the inaugural review was that LASIK was safer than PRK. The additional data has altered this finding and no strong evidence of a difference could be found.

Overall completeness and applicability of evidence

It is widely accepted that visual recovery following LASIK is more rapid than following PRK. This is again supported by the data presented in this updated review. Our subgroup analyses show that this benefit is present across all degrees of refractive errors and regardless of whether newer technologies such as femtosecond laser flap creation and wavefront‐guided ablation are used or not. From the patients' perspective, the more rapid visual recovery and less postoperative pain following LASIK are the main advantages of LASIK over PRK. The pain scores in the studies we found support the generally accepted view that LASIK is associated with less postoperative pain but more intraoperative discomfort than PRK. This is important when counselling patients about their operation.

When individual studies are examined the only trials which individually found a difference between the treatments were Wang 1997, Hjortdal 2005 and Schallhorn 2009. As discussed in the results section, Wang 1997 is at significant risk of bias. It is a large trial and carries significant weight in the analyses, therefore performing a sensitivity analysis on the relevant outcomes to which it contributes was essential.

At first glance, LASIK appeared to have superior efficacy with the odds of achieving a visual acuity of 20/20 or better at six months and 12 months significantly more likely than with PRK. The superior visual outcome at 12 months could be accounted for by the more accurate refractive outcomes seen with LASIK at 12 months. However, looking at the data more clearly, these findings were limited to the subgroup of patients with refractive errors up to ‐6 D. More importantly the results were sensitive to the exclusion of Wang 1997. When analyses were performed without this potentially biased study, no differences were found in any of these outcomes.

The duration covered by this study encompasses huge advances in the technology and techniques used. To evaluate whether there was any difference in the contemporary forms of treatment, using femtosecond laser flap creation and wavefront‐guided treatments, we stratified outcomes into studies from 2008 onwards in which these were used and those pre 2008 which employed older technology. This analysis demonstrated that contemporary studies do not show any difference in efficacy or accuracy between PRK and LASIK whereas older studies show a possible difference in favour of LASIK. This suggests that modern PRK with mitomycin C (MMC) and wavefront‐guided treatments is equally as effective and accurate as LASIK.

We previously found that LASIK was potentially safer than PRK, based upon the fact that fewer LASIK eyes lost 2 or more lines of best spectacle‐corrected visual acuity (BCVA) and this finding was not sensitive to the exclusion of Wang 1997. In this updated review, whilst there is still a significant difference in the loss of 2 or more lines of vision, which is less common in LASIK (odds ratio (OR) 0.47, 95% confidence interval (CI) 0.23 to 0.98) this disappears once Wang 1997 is excluded (OR 0.52, 95% CI 0.22 to 1.26). Hence the updated data in this review does not support this finding. This difference is likely explained in the overall analysis by the inclusion of data from additional studies that has reduced the significant weight that Wang et al carried previously. When we performed a sub‐analysis according to the surgical technology used it showed that prior to 2008 LASIK was less likely to result in loss of 2 or more lines of BCVA whereas in contemporary studies there was no difference between treatments. The implication for patients is that they can be reasonably confident that the risks involved are equal for both treatments.

In several outcome analyses Hjortdal 2005 appears to have a different outcome from other studies and in some instances this resulted in the detection of statistical heterogeneity. This may in part be explained by the mean postoperative spherical equivalent outcomes for this trial. The forest plots of mean postoperative spherical equivalent refraction data are of limited value on their own but must be interpreted in conjunction with the raw spherical equivalent data. These data demonstrate that for prescriptions up to ‐6 D at two to four weeks post‐treatment, the mean difference in refractive error between the two treatments was only 0.10 D (95% CI 0.04 to 0.16). There was no significant difference between treatments at six or 12 months. The data on refractive stability indicate that at one year post‐treatment, neither procedure has a significant degree of change in refraction (Table 5).

The follow‐up duration reported in these studies was variable. Only eight of the 12 studies reported 12‐month follow‐up data and of those studies using contemporary techniques and published post 2008, only two of six reported 12‐month follow‐up data. The remainder reported six‐month data. It is possible that this duration of follow‐up is inadequate to determine the final visual outcomes for either or both of these procedures. There is evidence that vision may continue to improve after one year with continued resolution of haze (Rajan 2004). Recovery of corneal innervation and restoration of a normal tear film and ocular surface may also take longer than 12 months (Calvillo 2004; Murphy 1999).

The incidence of visually significant corneal haze following PRK was found to range from 0% to 13% with an overall average of 3.3%. Only one instance of visually significant haze due to diffuse lamellar keratitis following LASIK was reported. The risk of significant haze after PRK is an important difference between these procedures, which in three of the recent trials was addressed using MMC as an adjunct (Table 2). Of the six contemporary studies three reported data on optical side effects such as glare and halos. None of these found any difference between PRK and LASIK.

The rate of flap‐related complications in LASIK participants ranged from 0.7% to 15% and only one participant lost 2 or more lines of BCVA. This is in keeping with the published rate of approximately 4% (Gimbel 1998; Lin 1999; Moshirfar 2010; Stulting 1999; Sugar 2002). Of the six post 2008 trials which used femtosecond lasers for flap creation only two reported flap complications. It is not possible to comment on the relative safety of mechanical versus femtosecond laser flap creation based on the current data set.

We incorporated an additional outcome measure of higher order aberrations into this study. All eight studies added to this update reported such data. The outcomes were remarkably similar. All but one study showed that in both PRK and LASIK there was a significant increase in higher order aberrations following treatment versus pre‐operatively. There was no difference between the two treatments in the amount or type of aberrations induced by treatment.

Quality of the evidence

Overall we graded the quality of the evidence as low or very low (Table 1). This was largely because of the potential for risk of bias in the included studies.

The major difficulty in combining the results of randomised controlled trials included in this review was the heterogeneity of outcome measures and follow‐up intervals reported. Improved compliance with suggested methods for reporting visual and refractive results of trials involving refractive procedures has made this update significantly easier than the initial review. Our framework of clearly defined outcome measures at fixed time points has worked well in both versions of this review. The downside of using such rigid outcome measures is that it is difficult to extract complete data sets from each trial. Hence not all trials could be included in each of the outcome analyses. In this updated version we have modified the outcome measures to try to detect more subtle differences in visual outcome (BCVA less than or equal to 20/15) and safety (loss of 1 or more lines of BCVA). This approach was limited by the fact that a minority of studies reported such data.

The methodological quality of the trials that we have included is in some cases satisfactory but not without flaws. The quality is summarised in Figure 1 and in Figure 2. Masking of participants and personnel when performing the procedure was not possible because of the nature of the procedures so studies were not excluded on the basis of this. Masking of outcome assessment was not performed in all studies and is a potential source of significant bias. We had concerns about the quality of two trials (Schallhorn 2009; Wang 1997) and we examined the effect of including these data using a sensitivity analysis. The mixture of study designs (unilateral versus bilateral treatment) posed a problem with data synthesis. In order to include data from all study types we assumed that the response of any eye to one treatment is in no way related to or predictable from the response of the fellow eye and we therefore treated data as unpaired for all studies. By doing this we may have lost the power to detect changes that a paired analysis may have found. However, if we were to analyse the paired and unpaired data separately we would have had only three trials with paired data and two trials with unpaired data. As not all trials reported data for each outcome and time point the outcomes that could be analysed in a meaningful way would be minimal. Accepting these limitations we were able to combine the data from these different trials and perform statistical analysis on the results.

Authors' conclusions

Implications for practice.

This review demonstrates that LASIK results in a more rapid recovery of visual acuity post‐treatment than PRK. The visual and refractive outcomes of these two procedures are comparable, especially when modern techniques and technology are used.

Implications for research.

Further research should focus on using contemporary techniques and equipment and on using more sensitive outcome measures, later time points and a questionnaire instrument designed to measure vision‐related quality of life. Further trials using contemporary techniques are required to determine whether LASIK and PRK as currently practised are equally safe. Randomising eyes to treatment is an efficient design, but only if analysed properly. In future trials, more efforts could be made to mask the assessment of outcome.

What's new

Date	Event	Description
12 November 2012	New search has been performed	Issue 1 2013: Electronic searches were updated. One new author, Jennifer Evans, assisted with updating the review.
12 November 2012	New citation required but conclusions have not changed	Issue 1 2013: Seven new studies (Barreto 2010; Durrie 2008; Hatch 2011; Moshirfar 2010; Schallhorn 2009; Wallau 2008; Manche 2011) were included in the review. A 'Summary of findings' table and four additional tables are included in the review. Higher order aberrations and modulation transfer function data added.

History

Protocol first published: Issue 1, 2005
 Review first published: Issue 2, 2006

Date	Event	Description
28 March 2008	Amended	Converted to new review format.
7 December 2005	New citation required and conclusions have changed	Substantive amendment.

Appendices

Appendix 1. CENTRAL search strategy

#1 MeSH descriptor Myopia 
 #2 myop* 
 #3 sight* NEAR/3 (short or near*) 
 #4 (#1 OR #2 OR #3) 
 #5 MeSH descriptor Photorefractive Keratectomy 
 #6 keratectom* 
 #7 PRK 
 #8 (#5 OR #6 OR #7) 
 #9 MeSH descriptor Keratomileusis, Laser In Situ 
 #10 keratomileusis.tw. 
 #11 LASIK 
 #12 (#9 OR #10 OR #11) 
 #13 (#4 AND #8 AND #12)

Appendix 2. MEDLINE (OvidSP) search strategy

1. randomized controlled trial.pt. 
 2. (randomized or randomised).ab,ti. 
 3. placebo.ab,ti. 
 4. dt.fs. 
 5. randomly.ab,ti. 
 6. trial.ab,ti. 
 7. groups.ab,ti. 
 8. or/1‐7 
 9. exp animals/ 
 10. exp humans/ 
 11. 9 not (9 and 10) 
 12. 8 not 11 
 13. exp myopia/ 
 14. myop$.tw. 
 15. ((short or near) adj3 sight$).tw. 
 16. or/13‐15 
 17. exp photorefractive keratectomy/ 
 18. keratectom$.tw. 
 19. PRK.tw. 
 20. or/17‐19 
 21. keratomileusis, laser in situ/ 
 22. keratomileusis.tw. 
 23. LASIK.tw. 
 24. or/21‐23 
 25. 16 and 20 and 24 
 26. 12 and 25

The search filter for trials at the beginning of the MEDLINE strategy is from the published paper by Glanville (Glanville 2006).

Appendix 3. EMBASE (OvidSP) search strategy

1. exp randomized controlled trial/ 
 2. exp randomization/ 
 3. exp double blind procedure/ 
 4. exp single blind procedure/ 
 5. random$.tw. 
 6. or/1‐5 
 7. (animal or animal experiment).sh. 
 8. human.sh. 
 9. 7 and 8 
 10. 7 not 9 
 11. 6 not 10 
 12. exp clinical trial/ 
 13. (clin$ adj3 trial$).tw. 
 14. ((singl$ or doubl$ or trebl$ or tripl$) adj3 (blind$ or mask$)).tw. 
 15. exp placebo/ 
 16. placebo$.tw. 
 17. random$.tw. 
 18. exp experimental design/ 
 19. exp crossover procedure/ 
 20. exp control group/ 
 21. exp latin square design/ 
 22. or/12‐21 
 23. 22 not 10 
 24. 23 not 11 
 25. exp comparative study/ 
 26. exp evaluation/ 
 27. exp prospective study/ 
 28. (control$ or prospectiv$ or volunteer$).tw. 
 29. or/25‐28 
 30. 29 not 10 
 31. 30 not (11 or 23) 
 32. 11 or 24 or 31 
 33. exp myopia/ 
 34. exp high myopia/ 
 35. exp degenerative myopia/ 
 36. myop$.tw. 
 37. ((short or near) adj3 sight$).tw. 
 38. or/33‐37 
 39. exp photorefractive keratectomy/ 
 40. keratectom$.tw. 
 41. PRK.tw. 
 42. or/39‐41 
 43. keratomileusis/ 
 44. keratomileusis.tw. 
 45. LASIK.tw. 
 46. or/43‐45 
 47. 38 and 42 and 46 
 48. 32 and 47

Appendix 4. LILACS search strategy

myop$ or short sight$ or near sight$ and kerat$ or laser$ or LASIK or LASEK or PRK or photorefract$

Appendix 5. metaRegister of Controlled Trials search strategy

Myopia and PRK and LASIK

Appendix 6. ClinicalTrials.gov search strategy

Myopia AND PRK AND LASIK

Appendix 7. ICTRP search strategy

PRK AND LASIK

Appendix 8. Subgroup analyses: level of myopia

Data analysis	Level of myopia*	LASIK				PRK			Effect measure				Test for subgroup differences
		Events		Total		Events		Total	Odds ratio/SMD	Lower CI	Upper CI		Chi2 P value
1.1 to 1.3	UCVA  20/15 or better, all time period, (no subgroup analysis, only 'low to moderate myopia' trials)
1.4	UCVA  20/20 or better at 2 to 4 weeks post‐treatment
	Low to moderate (3 trials)	161	200		255		474		3.81	2.53	5.73		0.22
	Moderate to high (2 trials)	24	130		11		119		2.22	1.03	4.75
1.5	UCVA  20/20 or better at 6 months post‐treatment
	Low to moderate (4 trials)	186	211		337		438		1.65	1.01	2.69		0.46
	Moderate to high (2 trials)	17	86		16		88		1.17	0.55	2.51
1.6	UCVA  20/20 or better at 12 months post‐treatment
	Low to moderate (4 trials)	282	314		413		522		1.85	1.18	2.91		0.09
	Moderate to high (1 trial)	1	25		3		20		0.24	0.02	2.47
1.7	Within 0.5 D of target refraction at 2 to 4 weeks post‐treatment
	Low to moderate (1 trial)	22	26		20		26		1.65	0.41	6.71		0.26
	Moderate to high (2 trials)	34	127		41		120		0.70	0.41	1.20
1.8	Within 0.5 D of target refraction at 6 months post‐treatment
	Low to moderate (2 trials)	25	52		23		53		1.29	0.56	2.97		0.45
	Moderate to high (2 trials)	21	84		25		88		0.86	0.43	1.69
1.9	Within 0.5 D of target refraction at 12 months post‐treatment
	Low to moderate (4 trials)	277	314		386		522		1.73	1.13	2.65		0.07
	Moderate to high (1 trial)	4	25		6		20		0.44	0.11	1.87
1.10	Mean postoperative spherical equivalent at 2 to 4 weeks post‐treatment
	Low to moderate (4 trials)		320				318		0.10	0.04		0.16	< 0.00001
	Moderate to high (2 trials)		127				120		‐0.56	‐0.78		‐0.34
1.11	Mean postoperative spherical equivalent at 6 months post‐treatment
	Low to moderate (3 trials)		280				282		‐0.01	‐0.07		0.04	0.63
	Moderate to high (2 trials)		84				88		0.05	‐0.19		0.28
1.12	Mean postoperative spherical equivalent at 12 months post‐treatment
	Low to moderate (2 trials)		196				200		‐0.01	‐0.06		0.04	0.16
	Moderate to high (1 trial)		25				20		0.20	‐0.09		0.49
1.13	Lost 1 or more lines BCVA at 6 months post‐treatment
	Low to moderate (2 trials)	3	189		5		193		0.61	0.14		2.58	0.83
	Moderate to high (1 trial)	16	62		22		68		0.73	0.34		1.56
1.14	Lost 2 or more lines BCVA at 6 months post‐treatment
	Low to moderate (6 trials)	4	410			15	618		0.56	0.21		1.52	0.46
	Moderate to high (1 trial)	2	62			8	68		0.30	0.08		1.10
1.15	Final BCVA of 20/40 or less at 6 months post‐treatment (no subgroup analysis, only 2 events, both 'moderate to high myopia' group)

Appendix 9. Subgroup analyses: newer versus older trials

	Level of myopia*	LASIK				PRK			Effect measure				Test for subgroup differences
		Events		Total		Events		Total	Odds ratio/SMD	Lower CI	Upper CI		Chi2 P value
1.1 to 1.3	UCVA  20/15 or better, all time periods, no subgroup analyses, only 'from 2008' trials
1.4	UCVA  20/20 or better at 2 to 4 weeks post‐treatment
	Before 2008 (5 trials)	149	299		246		562		2.98	2.06	4.33		0.02
	From 2008 onwards (3 trials)	99	109		59		109		8.24	3.78	17.94
1.5	UCVA  20/20 or better at 6 months post‐treatment
	Before 2008 (5 trials)	133	231		276		460		1.72	1.14	2.58		0.08
	From 2008 onwards (5 trials)	185	209		191		213		0.86	0.45	1.66
1.6	UCVA  20/20 or better at 12 months post‐treatment
	Before 2008 (5 trials)	136	197		263		396		1.78	1.15	2.75		0.35
	From 2008 onwards (2 trials)	197	205		201		209		0.93	0.26	3.39
1.7	Within 0.5 D of target refraction at 2 to 4 weeks post‐treatment
	Before 2008 (4 trials)	67	172		68		165		0.94	0.42	2.11		0.04
	From 2008 onwards (2 trials)	56	59		47		59		4.69	1.22	17.99
1.8	Within 0.5 D of target refraction at 6 months post‐treatment
	Before 2008 (4 trials)	43	120		50		125		0.83	0.47	1.45		0.13
	From 2008 onwards (4 trials)	122	159		115		163		1.56	0.86	2.82
1.9	Within 0.5 D of target refraction at 12 months post‐treatment
	Before 2008 (5 trials)	134	197		247		396		1.38	0.92	2.06		0.77
	From 2008 onwards (2 trials)	1198	205		195		209		1.76	0.35	8.82
1.10	Mean postoperative spherical equivalent  at 2 to 4 weeks post‐treatment
	Before 2008 (4 trials)		172				165		‐0.42	‐0.60		‐0.24	< 0.00001
	From 2008 onwards (5 trials)		353				351		0.10	0.03		0.17
1.11	Mean postoperative spherical equivalent at 6 months post‐treatment
	Before 2008 (3 trials)		112				116		0.17	‐0.27		0.60	0.48
	From 2008 onwards (6 trials)		395				401		‐0.07	‐0.15		0.01
1.12	Mean postoperative spherical equivalent at 12 months post‐treatment
	Before 2008 (3 trials)		79				74		0.04	‐0.11		0.18	0.48
	From 2008 onwards (3 trials)		216				220		‐0.01	‐0.06		0.04
1.13	Lost 1 or more lines BCVA  at 6 months post‐treatment
	Before 2008 (2 trials)	16	79		22		85		0.73	0.34		1.55	0.5
	From 2008 onwards (4 trials)	14	287		13		295		1.06	0.50		2.25
1.14	Lost 2 or more lines BCVA at 6 months post‐treatment
	Before 2008 (5 trials)	5	235		22		445		0.41	0.18		0.93	0.47
	From 2008 onwards (5 trials)	3	379		4		387		0.77	0.17		3.40
1.15	Final BCVA of 20/40 or less at 6 months post‐treatment, no subgroup analysis, no events in 'from 2008' trials

Appendix 10. Sensitivity analyses

Data analysis	Outcome	Odds ratio (OR) or mean difference (MD) (95% CI)	OR or MD (95% CI) excluding studies at risk of bias	Study excluded
1.1	UCVA of 20/15 or better at 2 to 4 weeks post‐treatment	OR 5.89 (3.34 to 10.39)	OR 8.30 (2.89 to 23.86)	Schallhorn 2009
1.2	UCVA of 20/15 or better at 6 months post‐treatment	OR 1.13 (0.75 to 1.69)	OR 0.94 (0.57 to 1.56)	Schallhorn 2009
1.3	UCVA of 20/15 or better at 12 months post‐treatment	OR 1.08 (0.58 to 2.0)	OR 1.61 (0.41 to 6.34)	Schallhorn 2009
1.4	UCVA of 20/20 or better at 2 to 4 weeks post‐treatment	OR 3.69 (2.55 to 5.36)	OR 4.23 (2.53 to 7.09)	Wang 1997
1.5	UCVA of 20/20 or better at 6 months post‐treatment	OR 1.41 (1.0 to 2.0)	OR 1.21 (0.77 to 1.91)	Wang 1997
1.6	UCVA of 20/20 or better at 12 months post‐treatment	OR 1.64 (1.10 to 2.45)	OR 1.39 (0.65 to 3.00)	Schallhorn 2009 Wang 1997
1.7	Within 0.50 D of target refraction at 2 to 4 weeks post‐treatment	No pooled estimate because substantial heterogeneity	Sensitivity analysis not done because no high‐risk trials reported this outcome
1.8	Within 0.50 D of target refraction at 6 months post‐treatment	OR 1.11 (0.74 to 1.67)	Sensitivity analysis not done because no high‐risk trials reported this outcome
1.9	Within 0.50 D of target refraction at 12 months post‐treatment	OR 1.45 (0.99 to 2.10)	OR 1.33 (0.90 to 1.96)	Schallhorn 2009 Wang 1997
1.10	Mean postoperative spherical equivalent at 2 to 4 weeks post‐treatment	No pooled estimate because substantial heterogeneity	No pooled estimate because substantial heterogeneity	Schallhorn 2009
1.11	Mean postoperative spherical equivalent at 6 months post‐treatment	No pooled estimate because substantial heterogeneity	No pooled estimate because substantial heterogeneity	Schallhorn 2009
1.12	Mean postoperative spherical equivalent at 12 months post‐treatment	MD 0.008 (‐0.06 to 0.04)	MD 0.004 (‐0.11 to 0.12)	Schallhorn 2009
1.13	Lost 1 or more lines of BCVA at 6 months or more post‐treatment	OR 0.88 (0.51 to 1.50)	OR 0.93 (0.52 to 1.66)	Schallhorn 2009
1.14	Lost 2 or more lines of BCVA at 6 months or more post‐treatment	OR 0.47 (0.23 to 0.98)	OR 0.44 (0.17 to 1.18)	Schallhorn 2009 Wang 1997
1.15	Final BCVA of 20/40 or less at 6 months or more post‐treatment	OR 0.12 (0.01 to 1.93)	Sensitivity analysis not done because no high‐risk trials reported this outcome

Data and analyses

Comparison 1. LASIK versus PRK.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 UCVA of 20/15 or better at two to four weeks post‐treatment	4	566	Odds Ratio (M‐H, Random, 95% CI)	5.89 [3.34, 10.39]
2 UCVA of 20/15 or better at six months post‐treatment	5	682	Odds Ratio (M‐H, Random, 95% CI)	1.13 [0.75, 1.69]
3 UCVA of 20/15 or better at twelve months post‐treatment	2	372	Odds Ratio (M‐H, Fixed, 95% CI)	1.08 [0.58, 2.00]
4 UCVA of 20/20 or better at two to four weeks post‐treatment	8	1079	Odds Ratio (M‐H, Random, 95% CI)	3.69 [2.55, 5.36]
5 UCVA of 20/20 or better at six months post‐treatment	10	1113	Odds Ratio (M‐H, Random, 95% CI)	1.41 [1.00, 2.00]
6 UCVA of 20/20 or better at 12 months post‐treatment	7	1007	Odds Ratio (M‐H, Random, 95% CI)	1.64 [1.10, 2.45]
7 Within 0.50 D of target refraction at two to four weeks post‐treatment	6		Odds Ratio (M‐H, Random, 95% CI)	Totals not selected
8 Within 0.50 D of target refraction at six months post‐treatment	8	567	Odds Ratio (M‐H, Random, 95% CI)	1.11 [0.74, 1.67]
9 Within 0.50 D of target refraction at 12 months post‐treatment	7	1007	Odds Ratio (M‐H, Random, 95% CI)	1.45 [0.99, 2.10]
10 Mean postoperative spherical equivalent at two to four weeks post‐treatment	9		Mean Difference (IV, Random, 95% CI)	Totals not selected
11 Mean postoperative spherical equivalent at six months post‐treatment	9		Mean Difference (IV, Random, 95% CI)	Totals not selected
12 Mean postoperative spherical equivalent at 12 months post‐treatment	6	589	Mean Difference (IV, Random, 95% CI)	‐0.01 [‐0.06, 0.04]
13 Lost one or more lines of BCVA at six months or more post‐treatment	6	746	Peto Odds Ratio (Peto, Fixed, 95% CI)	0.88 [0.51, 1.50]
14 Lost two or more lines of BCVA at six months or more post‐treatment	11	1446	Peto Odds Ratio (Peto, Fixed, 95% CI)	0.47 [0.23, 0.98]
15 Final BCVA of 20/40 or less at six months or more post‐treatment	6	442	Peto Odds Ratio (Peto, Fixed, 95% CI)	0.12 [0.01, 1.93]

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Barreto 2010.

Methods	A prospective, randomised study of who underwent simultaneous WFG LASIK in 1 eye and WFG PRK in the contralateral eye. UCVA, BCVA and spherical equivalent refraction, high and low contrast visual acuity, wavefront analysis, contrast sensitivity, and retinal straylight measurements were performed preoperatively and at 3, 6 and 12 months postoperatively. A third‐generation straylight meter, C‐Quant (Oculus Optikgeräte GmbH), was used for measuring intraocular straylight.
Participants	Country: Brazil 22 eyes of 11 patients (6 men, 5 women; mean age: 32.4 ± 6.2 years, range: 25 to 39 years) with myopic astigmatism. Complete ophthalmologic examination, topography, pachymetry, wavefront analysis and contrast sensitivity assessment were performed. Exclusion criteria were patients aged 21 or 40 years, corrected distance visual acuity (CDVA) worse than 0.0 logMAR (Snellen 20/20) in both eyes, spherical equivalent refraction (SE) 5.00 D, estimated ablation depth 60 μm, pre‐existing ocular pathology and previous surgery
Interventions	Simultaneous WFG LASIK in 1 eye and WFG PRK in the contralateral eye. All surgeries were wavefront‐guided using the OPDCAT platform (NIDEK Co Ltd, Gamagori, Japan) with a 5.0 mm optical zone and an additional 3.5 mm transition zone. The LASIK flap was created using an MK2000 microkeratome (NIDEK Co Ltd)
Outcomes	12 months postoperatively, mean uncorrected distance visual acuity was 0.06 ± 0.07 logMAR in the WFG LASIK group and 0.10 ± 0.10 logMAR in the WFG PRK group. Mean preoperative intraocular straylight was 0.94 ± 0.12 log(s) for the WFG LASIK group and 0.96 ± 0.11 log(s) for the WFG PRK group. After 12 months, the mean straylight value was 1.010.1 log(s) for the WFG LASIK group and 0.97 ± 0.12 log(s) for the WFG PRK group. No difference was found between techniques after 12 months (P = 0.306). No significant difference in photopic and mesopic contrast sensitivity between groups was noted.
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Patients were randomised but not stated how
Allocation concealment (selection bias)	Unclear risk	Not described
Blinding of participants and personnel (performance bias) Visual acuity	High risk	This was not described but masking of the surgeon performing treatment is not possible and masking of the patient is difficult as PRK is known to cause significantly more postoperative pain
Blinding of outcome assessment (detection bias) Visual acuity	High risk	Not described
Blinding of outcome assessment (detection bias) Other outcomes	High risk	Not described
Incomplete outcome data (attrition bias) All outcomes	Low risk	“All patients completed 12 months of follow‐up...” Page 2
Selective reporting (reporting bias)	Low risk	Relevant outcomes reported

Durrie 2008.

Methods	Prospective, contralateral eye, randomised controlled trial performed at 2 centres
Participants	Country: USA. 100 eyes of 50 patients. Inclusion criteria included ‐ 2.00 to ‐ 6.00 D of spherical myopia, with up to 3.50 D of refractive astigmatism; a stable refraction for 1 year; a best‐spectacle corrected visual acuity (BSCVA) of at least 20/20 in each eye; and an average central corneal thickness of > 500 μm in each eye. Soft contact lens wearers were required to discontinue lens use for at least 3 days prior to surgery, whereas rigid contact lens wearers were required to discontinue use at least 3 weeks prior to surgery.
Interventions	Wavefront‐guided treatment was performed using the Alcon LADARVision 4000 excimer laser. For SBK eyes, flaps were cut with AMO/Intralase 60 kHz femtosecond laser using a raster pattern and a superior hinge. For PRK eyes, 20% ethanol was applied for 25 seconds followed by irrigation with cold saline.
Outcomes	Visual acuity and refraction at 1, 3 and 6 months Other outcomes assessed were contrast sensitivity (contrast visual acuity); wavefront and retinal image quality (using the modulation transfer function ‐ MFT) and wavefront aberrometry with RMS Observers were masked during follow‐up
Notes	2 studies were published, each reporting the outcomes of the same trial
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	“Randomization was performed based on a randomization schedule developed to assure that each group had an equal number of dominant eyes.” Page 3
Allocation concealment (selection bias)	Low risk	Not mentioned but allocation concealment is probably not an important source of bias because each individual participant received both treatments
Blinding of participants and personnel (performance bias) Visual acuity	High risk	This was not described but masking of the surgeon performing treatment is not possible and masking of the patient is difficult as PRK is known to cause significantly more postoperative pain
Blinding of outcome assessment (detection bias) Visual acuity	High risk	Not described
Blinding of outcome assessment (detection bias) Other outcomes	High risk	Not described
Incomplete outcome data (attrition bias) All outcomes	Low risk	“All 100 eyes (100%) were available for every follow‐up”. Table 1
Selective reporting (reporting bias)	Low risk	All relevant outcome measures were reported for all patients

el Danasoury 1999.

Methods	Prospective, randomised, paired clinical trial Participants randomised to LASIK in 1 eye and PRK in other using random number table. Sequence of surgeries for each participant was also randomised using random number table. Masking: participants ‐ unclear; provider ‐ unclear; outcome ‐ VA and refraction assessed by masked observers Exclusions after randomisation: none Losses to follow‐up: 2 participants (2 eyes from each group)
Participants	Country: United Arab Emirates Numbers randomised: 52 eyes of 26 participants Age: range 19 to 45 years (mean 26.7 ± 5.9 years) Gender: 12 male (46.2%), 14 female (53.8%) Inclusion criteria: age at least 18 years; documented stable refraction for 1 year; spherical equivalent refraction of ‐2.00 to ‐5.50 D; astigmatism of less than 2.50 D; BCVA 20/20; realistic expectations of outcome Exclusion criteria: previous refractive surgery; central corneal thickness of less than 490 microns by US pachymetry; keratoconus or suspect by videokeratoscopy; active ocular disease; dry eyes; systemic diseases likely to affect corneal wound healing (for example connective tissue disease); inability to adhere to the strict follow‐up timetable that was given to patient prior to surgery; presbyopic patients who opted for monovision were not enrolled
Interventions	Every patient had LASIK on 1 eye and PRK on the other during the same surgical session and by the same surgeon Laser: Nidek EC‐5000 for all cases Algorithms used: PRK ‐ Nidek EC 5000 computer algorithm; LASIK ‐ personal clinical customised nomogram Target refraction: emmetropia for all cases Keratome: automated Corneal Shaper
Outcomes	Clinical examination; refraction; visual acuity at 1 day, 2 weeks, 6 weeks, 3 months, 6 months, 1 year after surgery Questionnaire assessment of intraoperative pain and satisfaction: postoperatively and at 1 year after surgery
Notes	Postoperative refractive and visual outcomes assessed by masked observer
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	“The procedure assigned to each eye and the sequence of surgeries for each patient were randomised using a random number table.” Page 412
Allocation concealment (selection bias)	Low risk	Not mentioned but allocation concealment is probably not an important source of bias because each individual participant received both treatments
Blinding of participants and personnel (performance bias) Visual acuity	High risk	This was not described but masking of the surgeon performing treatment is not possible and masking of the patient is difficult as PRK is known to cause significantly more postoperative pain
Blinding of outcome assessment (detection bias) Visual acuity	Unclear risk	In general not mentioned: “Manifest refraction was done by an independent ophthalmologist who was not aware of the surgical procedures being performed;...” Page 412. Not clear if this applies to follow‐up examinations as well.
Blinding of outcome assessment (detection bias) Other outcomes	Unclear risk	In general not mentioned: “Manifest refraction was done by an independent ophthalmologist who was not aware of the surgical procedures being performed;...” Page 412. Not clear if this applies to follow‐up examinations as well.
Incomplete outcome data (attrition bias) All outcomes	Low risk	24/26 patients followed up. As randomised by eyes loss to follow‐up equally distributed in the 2 intervention groups.
Selective reporting (reporting bias)	Low risk	All relevant outcome measures were reported for all patients

el Maghraby 1999.

Methods	Prospective, randomised, paired clinical trial Participants randomised to LASIK in 1 eye and PRK in other using random number table. Sequence of surgeries for each participant was also randomised using random number table. Masking: participants ‐ unclear; provider ‐ unclear/unlikely; outcome ‐ unclear Exclusions after randomisation: none Losses to follow‐up: 3 participants (3 eyes from each group)
Participants	Country: Saudi Arabia Numbers randomised: 66 eyes of 33 participants Age: range 16 to 59 years (mean 26.2 ± 9.3 years) Gender: 22 male (66.7%), 11 female (33.3%) Inclusion criteria: age at least 18 years; no contact lens wear for 2 weeks pre‐assessment; spherical equivalent refraction of ‐2.50 to ‐8.00 D; astigmatism of less than 1.75 D; less than 2.00 D of anisometropia; BCVA 20/40 or better Exclusion criteria: corneal or retinal disease; previous eye surgery; keratoconus or suspect by videokeratoscopy; systemic diseases likely to affect corneal wound healing (for example connective tissue disease)
Interventions	Both surgeries performed during same surgical session by 1 of 3 surgeons. All treatments were performed under topical anaesthesia without any systemic sedation. Laser: Summit OmniMed Excimer laser for all cases Target refraction: emmetropia for all cases Keratome: Chiron Automated Corneal Shaper
Outcomes	Complete ophthalmic examination; manifest refraction; visual acuity; glare testing; videokeratography including slit lamp microscopy and glare testing at 1 day, 3 to 4 days, 2 weeks, 6 weeks, 3 months, 6 months, approximately 1 year and approximately 2 years after surgery Questionnaire assessment of intraoperative pain and satisfaction: postoperative and at 1 year after surgery
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"The order of the two laser procedures and the eye treated were first randomised using a random number table." Page 447
Allocation concealment (selection bias)	Low risk	Not mentioned but allocation concealment is probably not an important source of bias because each individual participant received both treatments
Blinding of participants and personnel (performance bias) Visual acuity	High risk	This was not described but masking of the surgeon performing treatment is not possible and masking of the patient is difficult as PRK is known to cause significantly more postoperative pain
Blinding of outcome assessment (detection bias) Visual acuity	High risk	Not described
Blinding of outcome assessment (detection bias) Other outcomes	High risk	Not described
Incomplete outcome data (attrition bias) All outcomes	Low risk	3/30 lost to follow‐up but as eyes randomised loss to follow‐up equal in 2 intervention groups
Selective reporting (reporting bias)	Low risk	All relevant outcome measures were reported for all patients

Forseto 2000.

Methods	Prospective, randomised, paired clinical trial; 40 eyes of 20 participants with bilateral myopia Participants were randomised to LASIK in 1 eye and PRK in the other using a random number table Masking: participants ‐ unclear, provider ‐ unclear, outcome ‐ no Exclusions after randomisation: none Losses to follow‐up: 3 participants (3 eyes from each group)
Participants	Country: Brazil Numbers randomised: 40 eyes of 20 participants Age: range 20 to 48 years (29 ± 7 years) Gender: 15 females, 12 males (3 participants were lost to follow‐up)
Interventions	Laser: VISX 20/20B Keratome: Chiron Automated Corneal Shaper
Outcomes	Outcomes reported in the published study were not suitable for inclusion because only data for the final follow‐up appointment were reported whereas data at fixed timepoints post‐treatment were required. The authors kindly provided the data in a format that could be included by providing visual acuity and refractive data at 6 and 12 months post‐treatment.
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Participants were randomised to LASIK in 1 eye and PRK in the other using a random number table
Allocation concealment (selection bias)	Low risk	Not mentioned but allocation concealment is probably not an important source of bias because each individual participant received both treatments
Blinding of participants and personnel (performance bias) Visual acuity	High risk	Masking of surgeon performing treatment is not possible. Masking of the patient is difficult as PRK is known to cause significantly more postoperative pain.
Blinding of outcome assessment (detection bias) Visual acuity	High risk	Not described
Blinding of outcome assessment (detection bias) Other outcomes	High risk	Not described
Incomplete outcome data (attrition bias) All outcomes	Low risk	Number of patients involved: 20 patients (40 eyes), but only 17 patients with post‐surgery follow‐up and included on the analysis. However, randomisation by eye so loss to follow‐up will be equal between study groups
Selective reporting (reporting bias)	Low risk

Hatch 2011.

Methods	Prospective, masked and randomised pilot study on 52 eyes: 26 with PRK and 26 with thin‐flap LASIK
Participants	Country: USA Mean age of patients, 13 men and 13 women, was 30.8 years (range: 23 to 46). 26 patients (52 eyes) with stable myopia (1.5 to 8.5 D) and astigmatism (0.242 to 3.11 D) were enrolled in the study. 11 patients excluded from this study had clinically significant lens opacities, previous corneal or intraocular surgery, keratoconus, unstable refraction, autoimmune disease, immunosuppressive therapy, or were pregnant or breastfeeding.
Interventions	PRK or thin‐flap LASIK (90 μm flap) were performed using the VISX Star S4 CustomVue laser. All flaps were created with the IntraLase FS60 femtosecond laser at 60 kHz in a raster pattern
Outcomes	Primary outcome measures: UDVA; CDVA; contrast sensitivity and complications at 1 day, 1 week, 1 month, 3 months and 6 months HOAs were measured and trended within groups as secondary measures
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"...the dominant eye was randomised (Research Randomizer software – Urbaniak, www.randomizer.org) ..." Page 452
Allocation concealment (selection bias)	Low risk	"The randomization protocol was generated before the trial and known only to the study coordinator." Page 452
Blinding of participants and personnel (performance bias) Visual acuity	High risk	Masking of surgeon performing treatment is not possible. Masking of the patient is difficult as PRK is known to cause significantly more postoperative pain.
Blinding of outcome assessment (detection bias) Visual acuity	Low risk	Although it was not specifically stated that the outcome assessment was masked, the study is described as masked and the following statements implies that the observations were masked: "After the study was completed, the results were compiled and the data unmasked for statistical analysis." Page 453
Blinding of outcome assessment (detection bias) Other outcomes	Low risk	Although it was not specifically stated that the outcome assessment was masked, the study is described as masked and the following statements implies that the observations were masked: "After the study was completed, the results were compiled and the data unmasked for statistical analysis." Page 453
Incomplete outcome data (attrition bias) All outcomes	Low risk	"25 of 26 patients (50 eyes) completed the study at postoperative 6 months." Page 454
Selective reporting (reporting bias)	Low risk	Relevant outcomes reported

Hjortdal 2005.

Methods	Randomisation using a random number system 1 eye randomised to LASIK or PRK Exclusions after randomisation: 1 Losses to follow‐up: 1 participant from the PRK group
Participants	Numbers randomised: LASIK (n = 25); PRK (n = 20) Myopia range: ‐6.00 D to ‐8.00 D Inclusion criteria: 19 years of age or older; no previous eye disease or eye surgery
Interventions	Same surgeon performed all procedures Laser: MEL‐70 G‐scan flying spot excimer laser (Meditech‐Aesclepion) Keratome: Supratome microkeratome
Outcomes	Various outcomes related to corneal power, thickness and IOP measurement reported. The only meaningful data for this review were the number of participants with UCVA of 20/40 or better at 12 months post‐treatment. The authors kindly provided additional data for mean spherical equivalent refraction, refractive outcomes and visual acuity following correspondence. Seven year data on corneal power and higher order abberations were reported in a second publication, Ivarsen and Hjordtal 2012.
Notes	NCT00404105
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	“Randomization was performed using a random number system.” Page 22
Allocation concealment (selection bias)	Unclear risk	Not described
Blinding of participants and personnel (performance bias) Visual acuity	High risk	This was not described but masking of the surgeon performing treatment is not possible and masking of the patient is difficult as PRK is known to cause significantly more postoperative pain
Blinding of outcome assessment (detection bias) Visual acuity	High risk	Not described
Blinding of outcome assessment (detection bias) Other outcomes	High risk	Not described
Incomplete outcome data (attrition bias) All outcomes	Low risk	25 randomised to LASIK and 21 to PRK. 1 patient dropped out of the study in the PRK group and was excluded from the analyses. “The follow‐up was complete at all examinations during the first year except for 1 LASIK eye at 3 months and 1 PRK eye at 6 months. After 1 year, retreatment for residual myopia was performed in 4 LASIK eyes and 3 PRK eyes. The 3‐year follow‐up included only patients without retreatment. In addition to the retreated patients, 6 LASIK patients and 1 PRK patient were not able to attend the 3‐year follow‐up. Fifteen LASIK eyes and 16 PRK eyes were therefore available for long‐term follow‐up.” Page 23 15/25 (60.0%) LASIK and 16/21 (76.2%) PRK long‐term follow‐up
Selective reporting (reporting bias)	Low risk	Relevant outcomes reported

Manche 2011.

Methods	Prospective randomised controlled trial. Dominant eye randomised using a computer generated randomisation schedule to either PRK or LASIK with fellow eye receiving the alternative procedure.
Participants	Country: USA Sixty‐eight eyes of 34 patients with myopia with or without astigmatism.Inclusion criteria included a stable refraction with a change of less than 0.50 diopters (D) of sphere or cylinder in the last year, discontinuation of soft contact lens wear at least 7 days prior to preoperative evaluation, best‐corrected visual acuity of 20/20 or better, age older than 21, and ability to participate in follow‐up examinations for at least 12 months following refractive surgery.
Interventions	All procedures performed with the VISX Star CustomVue S4 IR excimer laser (Abbott Medical Optics, Santa Ana, California). Autocentration and iris recognition were used in all cases. No mitomycin C was used in any of the cases. All surgeries were performed at Stanford University Eye Laser Center by a single surgeon (E.E.M.). For eyes undergoing PRK, a bandage contact lens (Acuvue Oasys; Johnson & Johnson Vision Care, Inc, New Brunswick, New Jersey) was placed until the epithelium was healed. LASIK flaps were created using the 60‐kHz IntraLase FS (Abbott Medical Optics, Santa Ana, California). A 9.2‐mm diameter, superior hinge with 100 µm programmed flap depth setting was used in all cases. Intraoperative ultrasonic pachymetry (Sonogage, Cleveland, Ohio) was performed in all LASIK cases. P
Outcomes	Patients were prospectively evaluated at 1 day, 1 week, 1 month, 3 months, 6 months, and 12 months. Primary outcome measures included UCVA, refractive stability, predictability, contrast sensitivity, aberrometry, subjective questionnaire, loss of best spectacle corrected visual acuity (BSCVA), and adverse event profile.
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"..Eyes were randomized according to a computer‐generated randomization schedule. Randomization was performed by assigning the dominant eye to receive wavefront‐guided LASIK or wavefront‐guided PRK and the fellow nondominant eye to receive the alternative procedure. "
Allocation concealment (selection bias)	Unclear risk	Not described
Blinding of participants and personnel (performance bias) Visual acuity	Unclear risk	Not described
Blinding of outcome assessment (detection bias) Visual acuity	Unclear risk	Not described
Blinding of outcome assessment (detection bias) Other outcomes	Unclear risk	Not described
Incomplete outcome data (attrition bias) All outcomes	Low risk	There was no patient lost to follow‐up (n = 34) on postoperative day 1, postoperative week 1, and postoperative month 1. Three patients were lost to follow‐up on the third postoperative month follow‐up and sixth postoperative month follow‐up. One patient was lost to follow‐up at the 1‐year follow‐up (Table 2)
Selective reporting (reporting bias)	Low risk	Relevant outcomes reported

Moshirfar 2010.

Methods	Prospective, randomised, clinical trial; 203 eyes of 104 patients with bilateral myopia Participants were randomised to LASIK in 1 eye and PRK using "Research Randomizer" software Masking: participants ‐ unclear, provider ‐ unclear, outcome ‐ no Exclusions after randomisation: none Losses to follow‐up: see 'Risk of bias' table below
Participants	Mean age of the 51 women and 53 men was 33.70 years (range 20 to 57 years) All eyes had stable myopia between 0.25 and 10.00 D and astigmatism between 0.00 D and 3.50 D Of the eyes, 101 had PRK and 102 had LASIK Exclusion criteria: a cornea thinner than 0.5 mm, significant asymmetry on topography, clinically significant lens opacity, previous corneal or intraocular surgery, keratoconus, unstable refraction, autoimmune disease, pregnancy or breastfeeding, and currently on immunosuppressive therapy
Interventions	Wavefront‐guided LASIK using a Visx Star S4 CustomVue platform compared to wavefront‐guided PRK
Outcomes	Predictability (SE refraction, sphere, cylinder); efficacy (uncorrected visual acuity); safety (corrected visual acuity)
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	“Research Randomizer softwareA was used to randomize patients to wavefront‐guided PRK or wavefront‐guided LASIK.” Page 1337
Allocation concealment (selection bias)	Unclear risk	Not stated
Blinding of participants and personnel (performance bias) Visual acuity	High risk	This was not described but masking of the surgeon performing treatment is not possible and masking of the patient is difficult as PRK is known to cause significantly more postoperative pain
Blinding of outcome assessment (detection bias) Visual acuity	High risk	Not described
Blinding of outcome assessment (detection bias) Other outcomes	High risk	Not described
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	“The study evaluated 203 eyes of 104 patients. [...] Of the eyes, 101 had PRK and 102 had LASIK. [...]  Two hundred three eyes (101 PRK, 102 LASIK) were evaluated at 3 months and 118 eyes (61 PRK, 57 LASIK), at 6 months. One hundred seven eyes (54 PRK, 53 LASIK) had custom wavefront measurements at 6 months. Of the 59 eyes in each group that had preoperative contrast sensitivity testing, 55 PRK eyes and 51 LASIK eyes completed contrast sensitivity testing at 6 months. Wavefront analysis was performed 6 months. postoperatively in 107 of 203 eyes, 54 (53%) in the PRK group and 53 (52%) in the LASIK group.” Page 1338‐9   Quite low follow‐up at 6 months: 61/101 (60.4%) PRK and 57/102 (55.9%) LASIK followed to 6 months
Selective reporting (reporting bias)	Low risk

Schallhorn 2009.

Methods	Prospective controlled trial of PRK versus LASIK. Patients were randomly assigned to either wavefront‐guided LASIK or to wavefront‐guided PRK. Both eyes of each patient received the same treatment unless treatment was only required in 1 eye.
Participants	Country: USA Myopia of up to ‐6 D and up to 3 D of astigmatism. 200 patients were randomised to undergo either PRK or LASIK in either 1 or both eyes. Both eyes of each patient received the same treatment unless treatment was only required in 1 eye. 200 eyes underwent LASIK and 198 underwent PRK. Surgery was performed using a VISX Star S4 CustomVue Fourier, no IR, with a 6.0 mm OZ with 8.0 mm TZ. SBK eyes had flap created using Intralase (15 KHz; 9.1 mm)
Interventions	LASIK versus PRK (termed "Advanced surface ablation")
Outcomes	Means; SE; UCVA; loss of BCVA and HOAs (RMS) up to 12 months
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	In the main study cohort patients were ranked by refractive error then assigned sequential numbers; odd numbers were assigned to 1 treatment and even numbers the other
Allocation concealment (selection bias)	High risk	Allocation was not concealed
Blinding of participants and personnel (performance bias) Visual acuity	High risk	Masking of surgeon performing treatment is not possible; masking of the patient is difficult as PRK is known to cause significantly more postoperative pain
Blinding of outcome assessment (detection bias) Visual acuity	High risk	Not described
Blinding of outcome assessment (detection bias) Other outcomes	High risk	Not described
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Follow‐up not described
Selective reporting (reporting bias)	Low risk	Relevant outcomes reported

SUMMIT 1998.

Methods	Prospective, randomised multicentre clinical trial Conducted in accordance with US FDA regulations as part of a phase III multicentre clinical study of the Summit Technology excimer laser (Waltham, MA). Randomisation by a centralised study co‐ordination centre. Only the eye that was treated first was included in this study. The decision as to which eye would be treated first was made by the principal investigator. Masking: participants ‐ unclear; provider ‐ unclear; outcome ‐ post‐treatment refractions were performed by 2 independent observers Exclusions after randomisation: none Losses to follow‐up: PRK group: 5 at 1 month, 14 at 3 months, 37 at 6 months; LASIK group: 10 at 1 month, 22 at 3 months, 53 at 6 months Losses to follow‐up: data were not available for all participants in both groups
Participants	Country: United States Numbers randomised: 220 eyes of 220 participants (PRK ‐ 105; LASIK ‐ 115) Age: PRK group ‐ range 21 to 58 years (mean 39 years); LASIK group ‐ range 21 to 64 (mean 38 years) Gender: PRK group ‐ 56 male (53%) 49 female (47%); LASIK group ‐ 48 male (42%) 67 female (58%) Inclusion criteria: age at least 21 years; no CL wear for 2 weeks pre‐assessment; spherical equivalent refraction of ‐6.0 to ‐15.00 D; astigmatism of less than or equal to 2.00 D. The study protocol allowed planned undercorrections and overcorrections of 1.0 D or less; corneal thickness of between 500 and 700 microns; a normal endothelial cell count. Exclusion criteria: spectacle corrected visual acuity of worse than 20/32; functionally monocular; previous ocular surgery; previous or current ocular disease including clinical or topographic evidence of keratoconus; systemic diseases likely to affect corneal wound healing
Interventions	Laser: Summit Apex Excimer Laser System (Summit Technology Inc.) Keratome: Chiron Automatic Corneal Shaper microkeratome (Chiron Vision, Inc, Claremont, CA)
Outcomes	People were examined before treatment and at 1 day, 3 days, 1 week, 1 month, 3 months and 6 months after treatment Examination at each time point included a detailed ophthalmologic examination; manifest refraction by 2 independent observers; visual acuity; manual keratometry and videokeratography Self administered questionnaire at the preoperative and 6‐month postoperative visits. Explanations of questions were given by study personnel if requested by the patient. Participants were asked to assess subjective glare, halo and monocular diplopia in the operated eye on a scale of 0 to 5, where 0 indicated an absence of symptoms and 5 indicated the worst symptom. Single time point statistics were calculated using the number of participants in each group who gave an answer to the corresponding question in the survey. Comparison of pre and postoperative mean symptom grade within a treatment group was tested with Student's t test. Change in symptoms pre and postoperatively were calculated from participants with data for pre and 6‐month visits. Possible associations of LASIK versus PRK with a change in glare, halo and diplopia were tested with Chi² tests.
Notes	Refractive outcomes (predictability of outcome) are calculated as achieved correction minus attempted correction
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	“Eyes were assigned randomly to either PRK or LASIK procedure after patient registration and communication with the study’s coordinating centre.” Hersh 1998 page 1513   “Eyes were randomly assigned to PRK or LASIK procedure at an independent coordinating centre” Hersh 1999 Page 390
Allocation concealment (selection bias)	Low risk	“Eyes were assigned randomly to either PRK or LASIK procedure after patient registration and communication with the study’s coordinating centre. The eye to be first treated was determined by the principal investigator.” Hersh 1998 page 1513
Blinding of participants and personnel (performance bias) Visual acuity	High risk	This was not described but masking of the surgeon performing treatment is not possible and masking of the patient is difficult as PRK is known to cause significantly more postoperative pain
Blinding of outcome assessment (detection bias) Visual acuity	Unclear risk	“Preoperative and follow‐up visits included a detailed ophthalmologic examination with manifest refraction by two independent observers at each visit.” Hersh 1998 page 1514
Blinding of outcome assessment (detection bias) Other outcomes	Low risk	“Preoperative and follow‐up visits included a detailed ophthalmologic examination with manifest refraction by two independent observers at each visit.” Hersh 1998 page 1514 Corneal topography: “All maps were graded by two masked observers” Hersh 1998 Page 614
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	220 eyes from 220 people “Two hundred five patients were observed at 1 month, 184 at 3 months, and 129 at 6 months. To investigate the potential bias that those patients lost to follow‐up at each time point differed from those patients examined, preoperative characteristics of potential importance were analyzed for each group by follow‐up status. There were no differences at baseline in age, gender, preoperative uncorrected or spectacle‐corrected visual acuity, manifest refraction spherical equivalent, or intraocular pressure between the initial patient cohorts and those observed at the time points studied” Hersh 1998 page 1515   68/105 (64.8%) PRK followed up, 61/115 (53.0%) LASIK   But this differs from statement in subsequent paper: “Preoperative and 1, 3 and 6‐month postoperative data were available on 104, 103 and 95 eyes, respectively, in the PRK group and 119 (?perhaps typo and means 115), 108, and 94 eyes, respectively in the LASIK group.” Hersh 1999 Page 391 Some outcomes different numbers of patients analysed: e.g. corneal topography, Hersh 1998
Selective reporting (reporting bias)	Low risk	Relevant outcomes reported

Wallau 2008.

Methods	Prospective randomised clinical trial. The right eye of each patient was randomised to either PRK or LASIK using a coin toss. The fellow left eye received the alternative procedure to the right eye.
Participants	Country: Brazil. 44 people (88 eyes) with myopia of up to ‐7 D; 44 eyes received SBK, 44 contralateral eyes received PRK. Inclusion criteria were best‐spectacle‐corrected visual acuity (BSCVA) of logMAR 0.0 (Snellen 20/20) or better in both eyes, at least 6 months’ refraction stability, an estimated residual corneal ultrasound pachymetry greater than 410 mm in both eyes and a complete ophthalmological exam without associated diseases.
Interventions	For both custom ablation was performed using LADARWave 4000 (Alcon). Ablation was calculated using a 6.5 mm ablation zone and a 1.25 mm transition zone. LASIK flaps were cut using the Moria M2 microkeratome. The PRK group received mitomycin C 0.002% applied to the stromal bed for 1 minute.
Outcomes	UCVA; BCVA; mean spherical equivalent; stromal haze; contrast sensitivity; higher order aberrations (RMS); visual satisfaction questionnaire; specular microscopy and pachymetry up to 6 months post‐treatment
Notes	There was a tendency towards overcorrection of myopia
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	“The right eye of each patient was randomised at the surgical centre using coin toss to either one of the procedures; the other eye automatically received the other technique.” Page 327. Not an ideal method but each individual received both treatments and right eyes unlikely to be different from left. Right eye of each patient was randomised to receive 1 treatment using a coin toss. The fellow eye received the other treatment.
Allocation concealment (selection bias)	Low risk	Not mentioned but allocation concealment is probably not an important source of bias because each individual participant received both treatments
Blinding of participants and personnel (performance bias) Visual acuity	High risk	This was not described but masking of the surgeon performing treatment is not possible and masking of the patient is difficult as PRK is known to cause significantly more postoperative pain
Blinding of outcome assessment (detection bias) Visual acuity	Low risk	“During follow‐up examinations, a single examiner was unaware of which procedure was done in each eye and slit‐lamp microscopy was always the last examination to be performed at each appointment.” Page 327 Wallau 1998 but not reported in Wallau 1999 ? masked for corneal haze
Blinding of outcome assessment (detection bias) Other outcomes	Low risk	“During follow‐up examinations, a single examiner was unaware of which procedure was done in each eye and slit‐lamp microscopy was always the last examination to be performed at each appointment.” Page 327 Wallau 1998 but not reported in Wallau 1999 ? masked for corneal haze
Incomplete outcome data (attrition bias) All outcomes	Low risk	42/44 participants followed up to 1 year
Selective reporting (reporting bias)	Low risk	Relevant outcomes reported

Wang 1997.

Methods	569 eyes of 322 participants were randomised to PRK or LASIK; not stated how this was done In some cases only 1 eye was treated. Where both eyes required treatment the participant received the same treatment in both eyes. Of the 160 participants (283 eyes) that were randomised to LASIK 78 participants (146 eyes) insisted in having PRK, the reason cited was that these patients could not afford LASIK
Participants	Numbers randomised: 569 eyes of 322 participants Myopia range: ‐1.50 to ‐6.00 D
Interventions	Laser: Keracor 116 excimer laser (Chiron Vision USA) Keratome: ALK Automated Corneal Shaper (Chiron Vision)
Outcomes	UCVA; BCVA; manifest refraction; complications contrast sensitivity; stability of refraction
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	The method used to randomise patients is unclear
Allocation concealment (selection bias)	High risk	569 eyes from 322 patients enrolled in study Randomly assigned to PRK (?162 people, ?286 eyes) or LASIK (160 people, 283 eyes) 78 patients (146 eyes) assigned to LASIK but wanted PRK Final figures: PRK 432 eyes (286 + 146), LASIK 137 eyes (569 ‐ 432) High switchover rate from LASIK to PRK a source of bias
Blinding of participants and personnel (performance bias) Visual acuity	High risk	This was not described but masking of the surgeon performing treatment is not possible and masking of the patient is difficult as PRK is known to cause significantly more postoperative pain
Blinding of outcome assessment (detection bias) Visual acuity	High risk	Not described
Blinding of outcome assessment (detection bias) Other outcomes	High risk	Not described
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	“Three hundred and seven eyes in the PRK group (71.1%) and 103 eyes in the LASIK group (75.2%) were followed for 1 year.” Page 53 At 12 months, outcome data were reported for 103 of the initial 135 LASIK eyes and 307 of the initial 427 PRK eyes
Selective reporting (reporting bias)	Low risk	Relevant outcomes reported

BCVA: best spectacle‐corrected visual acuity 
 CDVA: corrected distance visual acuity 
 CL: contact lens 
 D: dioptres 
 FDA: (US) Food and Drug Administration 
 HOA: higher order aberration 
 IOP: intraocular pressure 
 LASIK: laser‐assisted in‐situ keratomileusis 
 MMC: mitomycin C 
 PRK: photorefractive keratectomy 
 RMS: root mean square 
 SBK: sub‐Bowmans keratomileusis 
 UCVA: uncorrected visual acuity 
 UDVA: uncorrected distance visual acuity 
 VA: visual acuity 
 WFG: wavefront‐guided

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Azar 1998	Study design not a randomised controlled trial
Cochener 2001	Study design not a randomised controlled trial
Helmy 1996	Study design not a randomised controlled trial
Hersh 2003	Outcomes examined not relevant to this review
Hovanesian 2001	Study design not a randomised controlled trial Outcomes examined not relevant to this review
Jae 2000	Study design not a randomised controlled trial Outcomes examined not relevant to this review
Ji 2001	Study design not a randomised controlled trial
Kanellopoulos 1997	Outcomes examined not relevant to this review
Lee 2001	Contact with corresponding author revealed that no randomisation was used when assigning participants to groups
Lesueur 2003	Study design not a randomised controlled trial
Lipshitz 1999	Study design not a randomised controlled trial
Matsui 2001	Study design not a randomised controlled trial Outcomes examined not relevant to this review
Neeracher 2004	Study design not a randomised controlled trial
Ninomiya 2003	Study design not a randomised controlled trial
Oshika 1999	Outcomes examined not relevant to this review
Perez‐Santonja 1999	Study design not a randomised controlled trial Outcomes examined not relevant to this review
Pisella 1999	Outcomes examined not relevant to this review
Steinert 1998	Data in this study are included in the Summit PRK LASIK study group reports; correspondence with the author confirmed this to be the case
Tole 2001	Study design not a randomised controlled trial

LASIK: laser‐assisted in‐situ keratomileusis 
 PRK: photorefractive keratectomy

Differences between protocol and review

Revisions to outcome measures

Following advice from the Editorial Team of the Cochrane Eyes and Vision Group, the outcome measures set out in the protocol for this review have been modified post hoc. This was necessary as a result of the diversity of outcome measures reported by the included studies. We defined new primary and secondary outcome measures as listed below. We chose three primary outcome measures that we believe best reflect the effectiveness and safety of these procedures. In this update version of the review we have also included data on higher order aberrations because these have become increasingly used to evaluate outcomes in refractive surgery.

The proportion of eyes with uncorrected visual acuity (UCVA) of 20/20 or better at 12 months post‐treatment and the proportion of eyes within ±0.50 D of target refraction at 12 months post‐treatment were chosen as primary outcome measures of effectiveness. Achievement of these outcomes at 12 months post‐treatment would undoubtedly be considered a success by both surgeon and patient. They reflect the effectiveness in terms of achieving a satisfactory visual result (UCVA of 20/20 or better) and in terms of the accuracy of the procedure (±0.50 D of target refraction). We chose the 12‐month over the six‐month time point because of our concerns that refractive stability may not yet have been achieved and corneal haze not yet resolved in some participants at six months.

Our primary outcome measure of safety was the proportion of eyes that lost 2 or more lines of best‐corrected visual acuity (BCVA) at six months or more post‐treatment. We chose this outcome because it is a commonly employed measure of adverse outcomes in ophthalmology. The presence of haze sufficient to cause loss of 2 or more lines of BCVA at six months post‐treatment was considered to be a significant adverse outcome and will be detected by this outcome. We chose the time point of six months or more because this allows sufficient time for resolution of mild to moderate corneal haze following PRK. It also allows for the detection of any adverse events that may occur after the six‐month time point such as ectasia. A new section concerning quality of vision outcomes has been added to this updated version of the review. Quality of vision will be assessed using outcome data for higher order aberrations and modular transfer function data

A further outcome measure, the assessment of higher order aberrations has been added. This was added as it gives an indication of the quality of vision which is distinct from visual acuity and refractive error. Given that the vast majority of patients undergoing either treatment end up with excellent unaided visual acuity it may take more sophisticated measures of the quality of vision to determine whether there is any difference in the outcomes of these treatments.

We added a new table (Table 7: Higher order aberrations and modulation transfer function data).

Revisions to analyses

Two subgroup analyses were done that were not planned in the original protocol.

1. Comparing results in people with high and low myopia. 
 2. Comparing results from more recent studies with older studies as surgical techniques have changed.

Contributions of authors

Conceiving the review: BA 
 Designing the original review: AS, BA 
 Designing the updated review AS, BA, JE

Co‐ordinating the review: AS 
 Screening updated search results: AS, JE 
 Screening retrieved papers against inclusion criteria: AS, JE 
 Appraising quality of papers: AS, JE 
 Abstracting data from papers: AS, JE 
 Writing to authors of papers for additional information: AS 
 Data management for the review: AS, JE 
 Entering data into RevMan: AS, JE 
 Analysis of data: AS, JE 
 Interpretation of data: AS, JE 
 Writing the review: AS, JE 
 Guarantor for the review: AS

Sources of support

Internal sources

• Moorfields Eye Hospital NHS Trust, UK.

External sources

• No sources of support supplied

Declarations of interest

Bruce Allan does LASIK and PRK in private practice and is currently using LASIK as his first choice procedure in uncomplicated myopia and myopic astigmatism. Alex Shortt and Jennifer Evans have no interests to declare.

New search for studies and content updated (no change to conclusions)