Optical treatment of amblyopia: a systematic review and meta-analysis

Transcription

1 CLINICAL AND EXPERIMENTAL INVITED SYSTEMATIC REVIEW Optical treatment of amblyopia: a systematic review and meta-analysis Clin Exp Optom 2018; 101: Lisa Asper PhD OD Kathleen Watt BOptom (Hons) Sieu Khuu PhD BSc (Hons) School of Optometry and Vision Science, The University of New South Wales, Sydney, New South Wales, Australia l.asper@unsw.edu.au Submitted: 13 June 2017 Revised: 8 December 2017 Accepted for publication: 9 December 2017 DOI: /cxo Background: Despite evidence that amblyopia can often be treated by optical treatment alone, many practitioners still do not use an optical-correction-only phase in amblyopia treatment and some investigators omit this important step in their research. This paper aims to systematically review the evidence for the optical treatment of strabismic, refractive and combined-mechanism amblyopia and to quantify the evidence via a meta-analysis. Methods: A search of online databases MEDLINE, EMBASE, PsycInfo, the Cochrane Library, and bibliographies of review papers, along with subsequent personal communication, resulted in 29 papers that met our inclusion criteria, with 20 providing sufficient data for the calculation of effect sizes. A meta-analysis was performed to determine effect sizes and the heterogeneity thereof. Meta-regression was used to evaluate the contribution of the possible moderating factors of age, duration of optical correction, and initial visual acuity to the heterogeneity of the studies. In addition, effect sizes were analysed in subgroups based on amblyopia aetiology, that is refractive or strabismic or combined, and also in the fellow eyes. Results: No evidence of publication bias in the included studies was found using a Galbraith plot. Optical treatment of amblyopia resulted in a large positive effect size of 1.07 (0.49, 95 per cent confidence limits) on visual acuity, although the heterogeneity was significant (Q = , I 2 = per cent, p < ). Meta-regression indicated that effect sizes significantly decreased with age, increased with treatment duration, and that better initial acuity was associated with higher effect sizes. Conclusion: Effect sizes were always moderate to large, whether participants were younger or older children, or whether the aetiology was refractive or strabismic. Thus, optical treatment of amblyopia should be considered prior to other treatment in those with refractive error. Improved acuity before initiating other treatment would presumably make occlusion or penalisation less onerous and may improve compliance with further treatment. Key words: amblyopia, meta-analysis, optical treatment, refractive adaptation, systematic review Amblyopia is usually defined as reduced visual acuity in one or both eyes in the absence of pathology and due to abnormal visual experience early in life. 1 Amblyogenic factors include constant unilateral strabismus, refractive error including anisometropia and high iso-ametropia, and visual deprivation such as congenital cataracts. Although correction of refractive error does not initially provide normal vision in amblyopia, it has long been known that optical correction can over time improve the vision of people with bilateral amblyopia associated with high ametropia. 2 However, until 2002, the primary first-line treatment for unilateral amblyopia was occlusion of the non-amblyopic eye (for excellent reviews, see Moseley et al. 3 and Birch 4 ). Moseley 5 and Stewart et al. 6 provided solid evidence that correction of the refractive error can over time significantly improve the vision of the amblyopic eye in unilateral amblyopia, whether the amblyopia was associated with anisometropia, strabismus or a combination of the two. The improvement in visual acuity was often large enough that occlusion was not necessary. The period of visual acuity improvement has been found to extend up to four to six months. 7 Thus a period of spectacleonly treatment prior to any other treatment, often called optical treatment or a refractive adaptation period, is now considered an important independent phase of clinical amblyopia treatment. It is also vital to include an optical treatment phase of appropriate length prior to any investigation of other amblyopia treatments to correctly interpret improvement in acuity. Many studies now routinely incorporate an optical treatment phase of 4 18 weeks Despite the evidence for optical treatment of amblyopia, there is some evidence that clinicians have been slow to incorporate an optical treatment phase as a firstline treatment in practice. 13 To provide insight into the effectiveness of optical treatment in amblyopia therapy, the present study reports the outcomes of a systematic review and meta-analysis that examines the effect size (ES) of an optical treatment phase on visual acuity in amblyopia. ES provides a standardised means of quantifying the differences between groups (in this case the two groups are the 2018 Optometry Australia Clinical and Experimental Optometry July

2 amblyopes before an optical treatment phase and after an optical treatment phase) and comparing findings across studies. It prioritises the size of any differences found rather than being dependent on the number of participants. Accordingly, this analysis provides a more clinically relevant measure of treatment beyond the report of statistical significance as a significant change in acuity could be found due to a small effect on acuity in a large group of people, or large effect of optical treatment on a smaller group of people. 14 It is possible that studies that do not find an improvement in visual acuity through an optical treatment phase fail to be published, either because they are not submitted for publication or because they are not accepted. In an effort to reduce this source of possible publication bias, we looked at research that directly addressed an optical treatment phase as well as studies that included an optical treatment phase as part of their protocol, but did not necessarily consider that to be the main part of the study. Presumably these papers were published due to the interest in the main treatment or study question, despite the outcome of the optical treatment phase. The present study aims to answer the question: Does optical-only treatment of refractive, strabismic or combinedmechanism amblyopia significantly improve visual acuity? Secondary aims were to determine if factors such as age at treatment, duration of spectacle wear, initial corrected visual acuity or amblyopic aetiology have an influence on ES. In the present study, we refer to the spectacle-only or contact lens-only treatment period as optical treatment. METHODS Data sources and inclusion criteria We performed a search of MEDLINE, PsycInfo, EMBASE and Cochrane Library databases for papers with the key words and limitations listed in Table 1, from January 2000 to 2 October Note that some of the databases did not permit all of the limits; for example, PsycInfo ignored the limits to humans only. Where it was possible to specify, we chose to search not only titles and abstracts, but also the full 1 Amblyop.mp 2 Amblyop.mp or amblyopia/ 3 Limit 2 to humans 4 Limit 1 to humans 5 3 or 4 6 Contact lenses/or eyeglasses/ 7 Spectacle.mp 8 Optical treat.mp 9 Optical correct.mp 10 Refractive treat.mp 11 Refractive correct.mp 12 Refractive adapt.mp 13 Treatment.mp 14 6 or 7 or 8 or 9 or 10 or 11 or 12 or and Limit 15 to year = 2000 Current 17 Limit 16 to English language Indicates truncation; / means MeSH heading. Table 1. Search strategy text. We also searched the bibliographies of review papers. The study inclusion criteria were: an optical treatment period of at least two weeks was undertaken; the outcome measure of the study was visual acuity; amblyopia was associated with strabismus, anisometropia or high isoametropic refractive error, or a combination of these conditions; and study was published in English. The study did not target participants who had completed prior amblyopic treatment or patching. However, if the published studies happened to include some participants who had undertaken prior treatment, we included the study. Exclusion criteria were: amblyopia was due to deprivation such as congenital ptosis or cataracts; animal studies; reviews, case reports and letters to the editor; no optical treatment period mentioned in the study; outcome measure was not visual acuity (for example, outcome was measured in contrast sensitivity or motion sensitivity); participants had concurrent medical issues; amblyopes and non-amblyopes were combined into one sample; studies that targeted those who had failed previous amblyopia treatment; studies where optical treatment was via refractive surgery; studies regarding methodology only; studies targeting the maintenance of visual acuity after cessation of treatment; and studies where the data available only included either those amblyopes who were successful with optical treatment or only those who were unsuccessful with optical treatment. Paper selection The titles and abstracts were initially screened by one author (LA). Papers that met the inclusion criteria were also screened by a second author (KW), and any questions were resolved by conversation among all authors. Data gathering We extracted the mean best-corrected visual acuity at the start of optical treatment and at the end of a period of optical treatment, and the number of participants in the study. The standard deviation of the visual acuity at the beginning and end of optical treatment was also extracted if available. In cases where thedatawerenotincludedinthetext,we either extracted the data from graphs using WebPlotDigitiser 15 or wrote to the paper s corresponding author to obtain the data. All in all, we wrote to 50 authors to provide clarification and/or data from 64 papers. When available, we also extracted information about the following parameters: mean age or age range of the sample; mean duration or range of durations of optical treatment phase; the aetiology of the amblyopia; whether the amblyopia was unilateral or bilateral; the spectacle prescribing criteria for the study; whether the examiner was masked to previous refraction or visual acuity measures; per cent experiencing resolution of amblyopia (and definitions of resolution); whether and how compliance with spectacles was monitored; and country of origin. Clinical and Experimental Optometry July Optometry Australia

3 Risk of bias analysis The risk of bias for each included paper was analysed by two authors (LA and KW) independently, using the Effective Public Health Practice Project (EPHPP) 16 quality assessment tool for quantitative studies. The quality, or risk of bias, questions were answered in relation to visual acuity changes associated with optical treatment. Therefore, a paper s overall main topic, if it had been assessed, might have been given a strong rating, but we may have judged it only moderate or even weak regarding the optical treatment phase. Data analysis Articles that satisfied the inclusion criteria were reviewed and data needed to determine ES (for example, mean, SD, and n) as well as study details such as aetiology, duration of spectacle wear, and age needed for sub-analysis were extracted and inputted into Excel. Effect sizes were derived from each study using the procedures outlined by Hedges and Olkin. 17 Hedges g (g) was determined to provide an estimate of the ES from the difference between the mean of the pre-optical treatment visual acuity minus the post-optical treatment visual acuity divided by their pooled standard deviation 18 SD = ðn 1 1ÞSD ðn 2 1ÞSD 2 = ð n1 +n 2 2Þ where n 1 and n 2 are sample sizes of cases and controls and SD 1 and SD 2 are standard deviations of cases and controls, respectively. The SD and 95 per cent confidence interval (CI) for each ES estimate were derived using the formula provided by Hedges and Olkin, where SD g = ððn 1 +n 2 Þ= ðn 1 n 2 ÞÞ + g 2 =2n ð 1 +n 2 Þ and 95 per cent CI was given by g SD. ES calculations were calculated using Microsoft Excel and workbooks ( eur.nl/research-support/meta-essentials) provided by Meta-Essentials. 19 The ES provides a standardised measure of the difference between the two groups and in this study a positive ES is indicative of an improvement in visual acuity after optical treatment. We determined the overall ES by considering the ES of all studies using a weighted random effects model, as initial inspection of our data suggested a degree of variability in ES between studies. The contribution of each study ES to the overall weighted ES was determined by its variance as per the methods of Hedges and Olkin. 17 Here we calculated the Q and I 2 statistic (see Neyeloff et al. 20 ) to determine the degree of ES heterogeneity across all the studies reviewed. A significant Q value indicates that there is variability in ES among the studies that is due to more than sampling error, and suggests the contribution of moderating variables. The I 2 statistic provides a measure of heterogeneity of the studies, by providing an indication of the total variability among the ES as a percentage value. The formula used was I 2 = ðq dfþ=q * 100, where df is the degrees of freedom. I 2 value may vary from 0 to 100. I 2 of 25 per cent is considered low, 50 per cent as medium and 75 per cent as high. In the present study, we also considered potential moderator variables such as the age, mean duration of spectacle wear and initial corrected visual acuity. For these moderators, a random-effect metaregression was performed using the worksheets provided by Meta-Essentials. 19 A meta-regression provides a linear fit between ES and a particular moderator variable that weighs the contribution of each study to the precision of the ES estimate. Additionally, a subgroup analysis was conducted for studies in which the aetiology of amblyopia could be reliably ascertained. We derived combined ES from studies in which the aetiology was characterised as strabismus, anisometropia or combined. RESULTS The database searches in October 2017 (see Figure 1 for screening process and steps) disclosed 3,878 papers after 856 (428 were duplicates) non-english papers were removed. After additional duplicates in English were removed there were 2,928 papers remaining. Papers related to visual deprivation amblyopia, refractive surgery, animal studies, reviews, case reports and letters to editors were removed. Subsequently, the abstracts and/or methods sections of the remaining papers were perused for mention of any optical treatment phase. Eightynine papers reported an optical treatment phase of two to 52 weeks and satisfied the inclusion and exclusion criteria. Sixty of these were eventually excluded. Fifty-one of the excluded papers were those that did not specifically investigate optical treatment of amblyopia, but did mention a prestudy optical treatment phase. We were unable to obtain data about that optical treatment phase, and the paper did not include sufficient information about that optical correction treatment. The other nine were rejected because of the high probability that they duplicated data from other published papers. For example, a conference abstract might be followed by a paper of a similar or identical title, with the same authors and a very similar number of subjects and results. Therefore, 29 papers met the inclusion criteria for the analysis. These are listed in Tables 2 and 3. Twenty papers are included in the metaanalysis. Fourteen of the papers included the data necessary to be able to calculate an ES, with four 6,39 41 of those requiring data extraction from graphs using WebPlotDigitiser. 15 Sixteen authors responded to our requests for information, with three of them providing data and one providing clarification about the study. Through communication with authors, one author brought an additional paper 37 to our attention that had been overlooked because it had not mentioned the pre-study optical treatment phase in the paper (C Birch in to L Asper, 14 April 2017). The data from the pre-study optical treatment phase in that investigation were made available, so were included. Data from a further five papers were made available through personal communication (C Birch in s to L Asper, 14 and 19 April 2017; Xi Yu in an to L Asper, 9 April 2017; and Raymond Kraker, Director, PEDIG Coordinating Center, in s addressed to L Asper, 10 April 2017 and 31 May 2017). Table 3 summarises the 20 papers in the meta-analysis that investigated a total of 1,346 amblyopic eyes. 5 7,9 12,31 43 Quality and potential for bias of the included studies EPHPP ASSESSMENT OF QUALITY OF RESEARCH To assess the quality and potential bias of the studies included, the assessment tool provided by the EPHPP was used. 16 The 2018 Optometry Australia Clinical and Experimental Optometry July

4 Figure 1. Flow diagram indicating record selection process 21 Author (year) n Age Aetiology of amblyopia Duration of optical correction Pre-VA (logmar) Mean SD (range) Weeks Mean Agervi et al. (2009) (4 5.4) A Up to Chang (2017) (3.5 11) Iso Not stated 0.45 Chen et al. (2013) (3 10) Iso Granet et al. (2006) (3 14) Iso Hwang et al. (2010) (8 13) A Lee and Isenberg (2003) (3.5 8) S, A, C Lee and Chang (2014) ( ) A 13 or more 0.31 Lee and Lee (2006) (8 12) A Wallace et al. (2007) (3 9.2) Iso A: anisometropia, C: combined mechanism, Iso: high iso-ametropia, n: number of amblyopic eyes, S: strabismic. Standard deviations (SD) of the age were not available for some studies. SDs of the pre-visual acuity (pre-va) values were not available for any studies listed. Table 2. Papers which were included in qualitative analysis but not included in the quantitative meta-analysis Clinical and Experimental Optometry July Optometry Australia

5 Author (year) n Age Aetiology of amblyopia Duration of optical correction Pre-VA (logmar) Mean SD (range) Weeks Mean SD Adhikari and Shrestha (2013) (3 15) Iso 13 or more Atilla et al. (2009) A Chen et al. (2016) (7 12) A Chen et al. (2007) (3 7) A Cotter (2006) (3 to < 7) A Cotter et al. (2007) ( ) S Up to Cotter et al. (2012) (3 to < 7) S, C Kelly et al. (2016) ( ) S, A, C Lam et al. (2011) (3 6) A Laria et al. (2011) (3 11) A Li et al. (2014) (3 7.7) S, A, C Lin et al. (2016) (3 8) Iso Up to Maconachie et al. (2016) (3 12) S, A, C Moseley et al. (2002) (4.1 6) A, C Nishi et al. (2015) A Pang et al. (2012) A, C Scheiman et al. (2005) (7 17) S, A, C Stewart et al. (2004) (3 8) S, A, C Stewart et al. (2007) (3 8) S, A, C Yorgun et al. (2015) (3 13) Iso 13 or more (decimal VA) A: anisometropia, C: combined mechanism; Iso: high iso-ametropia, n: number of amblyopic eyes, S: strabismus, VA: visual acuity. Standard deviations of the age were not available in some studies. Table 3. Papers included in the meta-analysis EPHPP assessment tool bases the quality of a paper on answers to questions regarding selection bias, study design, confounders, blinding, data collection methods, and reporting of withdrawals and drop-outs, to arrive at a global rating for the paper. We answered the questions in this tool with regard to visual acuity. The global rating for each paper is given in the second column of Table 4. It is difficult to say whether the amblyopes included in these studies were typical of amblyopes in the general population. Participants were usually referred to the study by private practitioners or hospitals. It is possible, depending on the available healthcare systems, that these amblyopes were those whose parents could afford eye care. It is also possible that the interest of the parents regarding health care or research could have resulted in a selection bias. The studies were not predominantly from one region, but included regions with large and different populations such as Europe, North America and Asia. All but one of the included studies were, with respect to the optical correction phase, cohort studies of one group with before and after measurement of visual acuity. As a result, there were no confounders or differences in groups such as race or initial acuity because each child s final visual acuity was compared to her/his initial acuity. Age could be considered a confounder, because each child did mature somewhat within the times studied. The effect of age is examined later (see Discussion). Information regarding masking of study investigators is included in Table 4. None of the studies mentioned whether the participants in the study were masked as to the purpose of the study. Given that all parents and/or subjects provided informed consent, it is likely that the parents did know the purpose of the study, and it is possible that the amblyopes were also aware of the purpose of the study. Visual acuity testing in all studies was done with standard methods that are considered valid and reliable. However, some studies defined their visual acuity testing protocol more carefully than others. Studies with well-defined testing protocols are also indicated in Table 4. OTHER POTENTIALS FOR BIAS Table 4 includes, in addition to information mentioned in the previous paragraphs, information on the study designs (retrospective analysis of records or prospective), whether the prescribing protocol was clearly stated, and whether/how compliance with optical correction was monitored. This additional information was not used in the EPHPP overall quality assessment, but it may be useful to determine the quality of the studies. Nine studies (Table 2), covering 611 amblyopic eyes, were omitted from the meta-analysis because they did not include adequate data (for example, no reported SD or means) to calculate an ES. However, they were included in the systematic review because they did 2018 Optometry Australia Clinical and Experimental Optometry July

6 Author (year) EPHPP overall rating Study design retroor pro- spective VA examiner masked? VA testing protocol well-defined? Monitored compliance with optical correction? Adhikari and Shrestha (2013) 31 Strong Retro Not stated Not stated Not stated Agervi et al. (2009) 22 Strong Pro Not stated Not stated Yes diary and parent report Atilla et al. (2009) 32 Strong Retro Not stated Not stated Not stated Chang (2017) 23 Strong Pro Not stated Not stated Method not reported but compliance remarked upon Chen et al. (2007) 34 Strong Pro Yes masked to diagnosis and latest VA Not stated Yes online diary good compliance only included in analysis Chen et al. (2013) 24 Strong Retro Not stated Not stated Not stated Chen et al. (2016) 33 Strong Pro Not stated Not stated Not stated Cotter (2006) 35 Strong Pro Not stated Yes Not stated Cotter et al. (2007) 7 Moderate Pro Yes not clear if masked to treatment group, or to previous VA or refraction Yes Not stated Cotter et al. (2012) 36 Strong Pro Not stated Yes Not stated Granet et al. (2006) 25 Strong Retro Not stated Not stated Yes judged by reviewer on scale of 1 4 (4 being best) exact definitions not provided Hwang et al. (2010) 26 Strong Retro Not stated Not stated Not stated Kelly et al. (2016) 37 Moderate Pro Not stated Not stated Not stated Lam et al. (2011) 38 Strong Pro Yes masked to intervention, VA and refraction Yes Yes calendar diary Laria et al. (2011) 39 Moderate Pro Not stated Yes Not stated Lee and Isenberg (2003) 27 Strong Retro Not stated Not stated Not stated Lee and Lee (2006) 29 Moderate Pro Not stated Not stated Not stated Lee and Chang (2014) 28 Strong Retro Not stated Not stated Yes parent or participant report Li et al. (2014) 9 Strong Pro Not stated Yes Not stated Lin et al. (2016) 40 Moderate Pro Not stated Not stated Yes parent or participant report good compliance only included in analysis Maconachie et al. (2016) 41 Moderate Pro Not stated Not stated Yes objectively measured with spectacles dose monitor Moseley et al. (2002) 5 Strong Pro Not stated Not stated Not stated Nishi et al. (2015) 42 Moderate Pro Yes masked to whether amblyopic or not Not stated Not stated Pang et al. (2012) 10 Strong Pro Not stated Yes Not stated Table 4. Evaluation of quality and risk of bias of included studies Clinical and Experimental Optometry July Optometry Australia

7 Author (year) EPHPP overall rating Study design retroor pro- spective VA examiner masked? VA testing protocol well-defined? Monitored compliance with optical correction? Scheiman et al. (2005) 11 Strong Pro Yes masked Yes Not stated Stewart et al. (2004) 6 Strong Pro Not stated Yes Not stated Stewart et al. (2007) 12 Strong Pro Not stated Yes Not stated Wallace et al. (2007) 30 Strong Pro Not stated Yes Compliance level reported Yorgun et al. (2015) 43 Weak Pro Not stated Not stated Not stated EPHPP: Effective Public Health Practice Project, VA: visual acuity. Evaluation was made with regard to optical treatment component of the study, and not to the entire study as a whole. Table 4. Continued provide information about the impact of optical correction on amblyopia. Studies not included in the meta-analysis Four of the nine studies investigated binocular amblyopia due to high iso-ametropia in ages ranging from three to 14 years ,30 In three studies, resolution of amblyopia was found in per cent 24,25,30 of subjects, while one study did not report a resolution rate. Four of the nine studies investigated unilateral anisometropic amblyopia, 22,26,28,29 and one investigated anisometropic, strabismic and combined-mechanism amblyopia. 27 The age of the subjects ranged from 3.5 to 14 years. The reported resolution rates depended somewhat on definition. The resolution rates ranged from seven per cent (achieved 6/6 visual acuity) with one month of optical treatment to 94 per cent (achieved resolution defined as one line or less interocular acuity difference) with 52 weeks of optical correction. 22,28,29 Two studies did not report a resolution rate. Studies included in the meta-analysis To provide an indication of publication bias for the studies included in the metaanalysis, a Galbraith plot 44 is shown in Figure 2, which plots the ES converted to z- scores as a function of the inverse standard error for each study. An unweighted regression analysis of z-scores (solid line) was conducted which provides an estimate of the 95 per cent confidence range (dashed line). An expectation is that 95 per cent of studies fall within this range, which is the case for the studies analysed in the present study. The ES of all 20 papers (circle symbols), as well as the combined ES (diamond symbol) are shown in the Forest plot in Figure 3. A random effects model (which appropriately weighted each study based on their variance) was applied to estimate the combined ES. The combined ES, which provided an indication of the overall visual acuity improvement after optical treatment in amblyopic eyes, was 1.07 (lower 95 per cent CI: 0.58; upper 95 per cent CI: 1.55), and statistically significant (z-score: 4.24, two tailed, p < ). This ES is considered large. 45 However, the Q statistic value was Q = (p < ) and I 2 was per cent which indicates a high degree of heterogeneity across the studies. As the I 2 statistic indicated significant heterogeneity, important moderator variables of age, duration of treatment and initial corrected visual acuity were investigated. These variables might potentially contribute to the effectiveness of treatment. For example, the critical period of amblyopia development might indicate that optical treatment is most effective in younger participants and for longer durations of treatment. Improvement might also be greater for participants with better initial corrected visual acuity. These data trends are evident in Figure 4. A random effects meta-regression 46 which took into consideration the individual weight of each study (and accounted for any residual heterogeneity) was conducted for each moderator variable. Prior to this analysis we undertook simple regression analysis between these moderator variables which demonstrated that they did not correlate (average slope p > 0.08), and thus were not mutually related. The line of best fit (dashed lines) is shown in Figure 4. For age (see Figure 4A), there was a significant decrease in ES with increasing age as indicated by the line of best fit (y= , slope z-score: 86.62, p < , R 2 = 0.58). For duration, increasing the duration of treatment systematically improved the effectiveness of treatment (y = , slope z-score: 49.24, p < , R 2 = 0.20). Note that one study was omitted from this analysis because reliable duration values could not be obtained. Finally, for initial visual acuity, our analysis (omitting one study which reported visual acuity in lines and not logmar) showed that participants with better initial corrected visual acuity showed the greatest improvement with larger ESs and with the effectiveness of treatment decreasing with poorer initial corrected visual acuity (y = , slope z-score: 19.91, p < , R 2 = 0.12). Figure 2. Effect size as z-scores plotted as a function of the inverse standard error for each study reported in the present study. The solid line is the line of best fit (y = 1.03, x = 0), while dashed lines represent upper and lower 95 per cent confidence limits Optometry Australia Clinical and Experimental Optometry July

8 amblyopic eye (unpaired t-test, t [28] = 3.30, p = ). DISCUSSION The answer to the main question that we investigated is yes, optical-only treatment does significantly improve visual acuity in refractive, strabismic and combinedmechanism amblyopia. The factors that may have influenced the degree of visual acuity improvement, such as age at treatment, duration of spectacle wear, initial corrected visual acuity or amblyopic aetiology, are discussed below. Figure 3. A Forest plot of weighted effect size for each of the 20 studies (open circles). The combined effect size is given by the diamond. Error bars signify 95 per cent confidence limits. Individual effect sizes, upper and lower 95 per cent limits, study weight and the number of subjects are also given. For reference, the vertical dashed line at 0 indicates no effect. Positive effect sizes above this line represent improvement in the amblyopic eye. The studies investigated in this metaanalysis included amblyopia from different aetiologies. We undertook a random-effect subgroup analysis in which the aetiology was either anisometropia, strabismus or combined (see Table 5 for details). Studies in which aetiology could not be well established, or there were too few participants, or the study comprised of a mixture of the three, were omitted from the analysis. This analysis based on aetiology reported a combined ES of 0.79 (lower 95 per cent CI: 0.43; upper 95 per cent CI: 1.39) and was statistically significant (z-score: 6, two tailed, p < ). However, there was high heterogeneity (Q = 3,272.44, I 2 = 0.98). The combined ESs for each subgroup and heterogeneity are shown in Table 5 and a oneway analysis of variance which compared the mean ES between these groups showed that they were not significantly different (F [2,13] = 0.96, p = 0.40). A number of studies (n = 10) also reported the results from the fellow eye (see Figure 5). A meta-analysis was performed on available data and an overall ES of 0.37 (lower 95 per cent CI: 0.13; upper 95 per cent CI: 0.61; Q = 638, I 2 = 0.96) which was statistically significant (z-score: 3.05, two-tailed p = ). However, this improvement in visual acuity in the fellow eye was significantly different from the Age at optical treatment The meta-regression indicates that the effectiveness of optical treatment is highest in younger participants and this systematically decreases with age. The R 2 indicated that 58 per cent of the heterogeneity in the included studies might be accounted for by the age of the participants. The fact that younger children in general improved more from spectacles than older children is consistent with our knowledge of the critical period for amblyopia development and the greater neural plasticity in early childhood. Although there is evidence that amblyopia can be treated beyond early childhood, 47 in general amblyopia treatment results are thought to be better at an earlier age than at a later age. 48 It appears that the effect of optical treatment diminishes through the older childhood and teen years. Only one study included here specifically investigated optical treatment in older children, aged 7 17 years. Of the moderate amblyopes in that study (visual acuity of 6/24 or better), 14 per cent in the 7 12 age group and 11 per cent in the age group achieved 6/7.5 with optical treatment alone. Of the severe amblyopes, only five per cent of the 7 12 years age group achieved 6/12 or better, while none of the severe amblyopes in the years age group achieved 6/12 or better with optical correction alone. 11 It was often the case that the older children had received some prior treatment for amblyopia. 11 These children may have had a particularly deep amblyopia resistant to treatment that resulted in the low success rate at older years. However, the study reported similar mean improvement in the older age group (about six letters) in those who had prior Clinical and Experimental Optometry July Optometry Australia

9 Figure 4. Moderator variable analysis plotting effect size as a function of A: age, B: duration and C: initial corrected visual acuity (VA). In each bubble plot, the size of data point provides an indication of the study weight; dashed line represents the line of best fit. treatment and those who had not. In the younger group (7 12-year-olds), those with prior treatment improved about one line with optical correction and those without prior treatment improved almost two lines. 11 Thus failure with previous treatment should not preclude the prescription of proper optical treatment to improve visual acuity in amblyopes. Older children demonstrated a moderate ES, which suggests that optical treatment should be considered for amblyopia prior to other treatment, at least until the age of 17, the oldest age involved in our included studies. Several instances were reported of older children improving one to two lines of acuity, 8,9 and Scheiman et al. 11 reported a mean improvement of 6.1 letters (just over one line) for those aged years. Duration of spectacle wear The ES of optical treatment increased significantly as duration of wear increased in the included studies (see Figure 4), where the duration ranged from five to 52 weeks. The R 2 indicated that 19 per cent of the heterogeneity in the included studies might be accounted for by the duration of spectacle wear. One factor that makes the interpretation of the data on duration of optical treatment difficult is the reliability of determining the exact duration of treatment or when during treatment that visual acuity improved the most. In some studies, participants left the study after resolution of amblyopia or after a given number of lines of improvement. Data from these children who left the studies were not included in data from later follow-up visits, and accordingly, the ES for longer durations in the study was possibly confounded and thus not included in this paper s analysis. Two of the studies that involved wear of approximately 52 weeks did not investigate visual acuity improvement before the one year mark and the visual acuity improvement may have occurred at any time during that year. 32,42 On the other hand, Chen et al. 34 followed a number of anisometropic amblyopic children, aged three to seven years, for eight to 68 weeks and reported a marked improvement with optical correction alone up to 12 weeks, followed by a plateau, with smaller improvements found in some children after that plateau. In those children whose amblyopia resolved, the mean time to resolution was 18.3 weeks, with a range of eight to 40 weeks. These results agree with the findings of Stewart et al. who found that the mean visual acuity change occurred mostly over the first 12 weeks, while some individuals continued to improve over the 18 weeks of the study. 6 There is evidence that suggests that maximum visual acuity improvement can be achieved much earlier than 12 weeks. For example, one study reported that resolution of amblyopia (defined as one line or less difference between visual acuity of fellow and amblyopic eyes) was achieved by seven per cent of the amblyopes by the five-week follow-up visit. Twenty-one per cent of the participants improved three or more lines of acuity in that short time of five weeks. 35 Generally, the ESs are moderate after five to six weeks. Over the durations reported, the ESs indicate that the optical treatment was effective. It is important not to discontinue optical treatment of amblyopia too soon. Several studies reported that visual acuity continued to improve up to weeks. 5,7,11,35 The evidence from this review suggests that a minimum of 14 weeks of initial optical treatment of amblyopia should be used for those with refractive error, and that optical treatment should continue until visual acuity ceases to improve. This is consistent with the recommendation by Moseley et al. of a minimum of full-time initial spectacle wear of 12 weeks. 3 Relationship of baseline corrected visual acuity to ES In general, children with better initial corrected visual acuity demonstrated larger ESs. This is surprising, in that it might be expected that those with worse visual acuity had more room for improvement, and a floor effect might occur in those with Figure 5. Mean effect size for fellow and amblyopic eyes. Error bars represent 95 per cent confidence limits Optometry Australia Clinical and Experimental Optometry July

10 Subgroup n Effect size Lower 95% CI Upper 95% CI Q I 2 Anisometropia Combined Strabismus Table 5. Effect sizes and statistics in subgroups by aetiology of amblyopia better visual acuity. However, we found the opposite in that if amblyopia was less severe, the ES was larger. Perhaps the poorer acuity in severe amblyopia is more entrenched than moderate visual acuity loss, and therefore less amenable to treatment. The ES is smaller for severe amblyopes but is still moderate. A great deal of individual variation was evident, with some children with baseline corrected visual acuity of 6/60 reported to improve 4 6 lines of acuity through optical correction alone and another with 6/15 not improving at all. 7 Thus, optical treatment should be considered before initiating other treatment even in severe amblyopia. Type of amblyopia UNILATERAL VERSUS BILATERAL Only three of the studies in the metaanalysis investigated bilateral amblyopia due to high ametropia. 31,40,43 The large ESs ranged from 1.00 to 2.38, demonstrating that amblyopia associated with high ametropia is very responsive to optical treatment, as was suggested by Abraham over 50 years ago. 2 However, a recent Cochrane review on treatment for unilateral and bilateral refractive amblyopia included no studies on bilateral amblyopia. 49 Thus bilateral amblyopia may be an area that warrants further study. ANISOMETROPIC, STRABISMIC OR COMBINED There was no significant statistical difference between the mean ES of the anisometropic, strabismic and combined types of amblyopia. Stewart et al. also found that the final outcome of optical treatment did not depend on amblyopia aetiology. 6 The surprising result is that the visual acuity of strabismic amblyopes can improve significantly from optical treatment alone, despite the remaining strabismus and resulting non-corresponding visual input to the two eyes. The large mean ES of 0.92 for the strabismic amblyopes indicates that strabismic amblyopes benefit from optical treatment in the age ranges studied here. It may be that the optical treatment reduced the strabismus to a small magnitude for some amblyopes; however, several studies found good improvement in vision in participants with large magnitudes of esotropia. For example, Stewart et al. included strabismic participants with a mean of 17 prism dioptres deviation and reported a mean visual acuity improvement of over 3.5 lines. 6 Moseley et al. included two strabismic participants, with 20 and 35 prism dioptres of esotropia, whose corrected visual acuity improved over two and three lines respectively with optical treatment alone. 5 Cotter et al. reported that visual acuity improvement during optical treatment did not depend on magnitude of strabismus. 36 Thus the visual improvement with optical treatment is not solely due to any reduced magnitude of deviation. One might expect the improvement in visual acuity to lead to reduced suppression and subsequent diplopia in strabismic participants. Only one of the included studies reported a case of diplopia during optical treatment, and this was intermittent diplopia, occurring more than once a day. 11 Whether this child was strabismic or not was not stated. Thus, while inducing diplopia through optical treatment may be a possibility, it may be a very small one. Prescribing regimens The prescribing regimens from the 20 included studies that provided complete descriptions were similar to each other. All based their prescriptions on the cycloplegic refraction. The full anisometropia was prescribed, along with the full amount of astigmatic correction. In one study of anisomyopic amblyopia, the amount of the anisometropia was corrected by a contact lens in the more myopic eye, and the remainder of the prescription was given in spectacles. 10 As for correction of the hyperopic component of refractive error, most studies permitted an under-correction of hyperopia of 0.50 to 1.50 D less than the cycloplegic refraction finding. A few studies did not prescribe for hyperopia, or left it to the discretion of the prescriber, if hyperopia was less than D. Thus, the recommendation is to fully correct the anisometropia and astigmatic component of the refraction, and to under-correct hyperopia symmetrically by 0.50 to 1.50 D, depending, one assumes, on ocular alignment. Resolution of amblyopia Amblyopia was generally defined as resolved if the difference in acuity between the two eyes was one line or less. A substantial number of amblyopia cases resolved by optical treatment alone. Resolution through optical treatment alone occurred in 0 94 per cent of the amblyopes, with a mean resolution rate of approximately 32 per cent. How does ES of occlusion compare with that of optical treatment? A complete meta-analysis of occlusion treatment is beyond the scope of this paper. However, in the four included studies that enabled calculation of the ES of occlusion, the ES ranged from 0.48 to 1.04 with a mean of This is similar to the ES of optical treatment. Therefore, there is still a place for occlusion treatment for those whose amblyopia does not resolve through optical treatment alone. However, the recommendation is to consider optical treatment as the first line of treatment because of the powerful ES of optical treatment alone. There is approximately a 32 per cent chance of resolution of the amblyopia through spectacles alone; thus occlusion or penalisation may not always be necessary. In addition, the improved visual acuity from optical Clinical and Experimental Optometry July Optometry Australia

11 treatment may make occlusion less traumatic for the amblyope and her/his family. FUTURE DIRECTIONS AND LIMITATIONS This systematic review is limited by the fact that the initial selection of papers was not performed independently by more than one person. However, two authors did assess the quality of the papers independently, and the Galbraith plot indicates that the chances of a publication bias are small. For transparency, there would be an advantage if the protocol had been published prior to this report. The fact that none of the studies were randomised with a no-treatment control may be of concern, but the ethical problems associated with denying treatment for a condition that is known to benefit from early treatment means that such a control would be unlikely. CONCLUSION The literature clearly indicates that optical treatment of refractive, strabismic or combined-mechanism amblyopia results in a large positive effect on visual acuity and often results in the resolution of amblyopia. The fact that ESs varied considerably but were always moderate to large, whether participants were younger or older children, or whether the aetiology was refractive or strabismic, argues that the optical treatment of amblyopia should be the first line of treatment. Improved visual acuity before initiating other treatment would presumably make occlusion or penalisation less onerous for the amblyope and may improve compliance with further treatment. ACKNOWLEDGEMENT We would like to thank everyone who replied to our queries for information. REFERENCES 1. Schapero M, Cline D, Hofstetter HW. Dictionary of Visual Science, 2nd ed. Radnor, Pennsylvania: Chilton Book Company, Abraham SV. Bilateral ametropic amblyopia. J Pediatr Ophthalmol 1964; 1: Moseley MJ, Fielder AR, Stewart CE. The optical treatment of amblyopia. Optom Vis Sci 2009; 86: Birch EE. Amblyopia and binocular vision. Prog Retin Eye Res 2013; 33: Moseley MJ, Neufeld M, McCarry B et al. Remediation of refractive amblyopia by optical correction alone. Ophthalmic Physiol Opt 2002; 22: Stewart CE, Moseley MJ, Fielder AR et al. Refractive adaptation in amblyopia: quantification of effect and implications for practice. Br J Ophthalmol 2004; 88: Cotter SA, Edwards AR, Arnold RW et al. Treatment of strabismic amblyopia with refractive correction. Am J Ophthalmol 2007; 143: Birch EE, Li SL, Jost RM et al. Binocular ipad treatment for amblyopia in preschool children. J AAPOS 2015; 19: Li SL, Jost RM, Morale SE et al. A binocular ipad treatment for amblyopic children. Eye 2014; 28: Pang Y, Allison C, Frantz K et al. A prospective pilot study of treatment outcomes for amblyopia associated with myopic anisometropia. Arch Ophthalmol 2012; 130: Scheiman MM, Hertle RW, Beck RW et al. Randomized trial of treatment of amblyopia in children aged 7 to 17 years. Arch Ophthalmol 2005; 123: Stewart CE, Stephens DA, Fielder AR et al. Objectively monitored patching regimens for treatment of amblyopia: randomised trial. BMJ 2007; 335: Bienkowska A, Starbuck A, Churchill A. The management of amblyopia in children: the results of a national survey of orthoptists. Int J Ophthal Pract 2014; 5: Coe R. It s the Effect Size, Stupid: What Effect Size Is and why it Is Important. In: Annual Conference of the British Educational Research Association. University of Exeter, England, Rohatgi A. WebPlotDigitizer, v [Cited 13 Nov 2017.] Available at: WebPlotDigitizer/ 16. Effective Public Health Practice Project (EPHPP). Quality assessment tool for quantitative studies. [Cited 6 Nov 2017.] Available at: ca/pdf/quality Assessment Tool_2010_2.pdf 17. Hedges LV, Olkin I. Chapter 5 - Estimation of a single effect size: parametric and nonparametric methods statistical methods for meta-analysis. In: Hedges LV, ed. Statistical Methods for Meta-Analysis. San Diego, California: Academic Press, pp Field AP, Gillett R. How to do a meta-analysis. Br J Math Stat Psychol 2010; 63: Van Rhee H, Suurmond R, Hak T. User Manual for Meta-Essentials: Workbooks for Meta-Analysis (Version 1.0). Rotterdam: Erasmus Research Institute of Management, Neyeloff JL, Fuchs SC, Moreira LB. Meta-analyses and Forest plots using a Microsoft Excel spreadsheet: step-by-step guide focusing on descriptive data analysis. BMC Res Notes 2012; 5: Moher D, Liberati A, Tatzlaff J et al. Preferred reporting items for systematic reviews and megaanalyses: the PRISMA statement. PLoS Med 2009; 6: e: Agervi P, Kugelberg U, Kugelberg M et al. Treatment of anisometropic amblyopia with spectacles or in combination with translucent Bangerter filters. Ophthalmology 2009; 116: Chang JW. Refractive error change and vision improvement in moderate to severe hyperopic amblyopia after spectacle correction: restarting the emmetropization process? PLoS One 2017; 12: e Chen W, Chen J, Zhang F et al. Visual outcome in isoametropic amblyopic children with high hyperopia and the effect of therapy on retinal thickness. Am J Ophthalmol 2013; 155: Granet DB, Christian W, Gomi CE et al. Treatment options for anisohyperopia. J Pediatr Ophthalmol Strabismus 2006; 43: Hwang DJ, Kim YJ, Lee JY. Effect and sustainability of part-time occlusion therapy for patients with anisometropic amblyopia aged 8 years.br J Ophthalmol 2010; 94: Lee SY, Isenberg S. The relationship between stereopsis and visual acuity after occlusion therapy for amblyopia. Ophthalmology 2003; 110: Lee SH, Chang JW. The relationship between higher-order aberrations and amblyopia treatment in hyperopic anisometropic amblyopia. Korean J Ophthalmol 2014; 28: Lee YR, Lee JY. Part-time occlusion therapy for anisometropic amblyopia detected in children eight years of age and older. Korean J Ophthalmol 2006; 20: Wallace DK, Chandler DL, Beck RW et al. Treatment of bilateral refractive amblyopia in children three to less than 10 years of age. Am J Ophthalmol 2007; 144: Adhikari S, Shrestha U. Types of amblyopia and treatment outcome in Nepalese children. Int Eye Sci 2013; 13: Atilla H, Kay E, Erkam N. Emmetropization in anisometropic amblyopia. Strabismus 2009; 17: Chen Y, Chen X, Chen J et al. Longitudinal impact on quality of life for school-aged children with amblyopia treatment: perspective from children. Curr Eye Res 2016; 41: Chen PL, Chen JT, Tai MC et al. Anisometropic amblyopia treated with spectacle correction alone: possible factors predicting success and time to start patching. Am J Ophthalmol 2007; 143: Cotter SA. Treatment of anisometropic amblyopia in children with refractive correction. Ophthalmology 2006; 113: Cotter SA, Foster NC, Holmes JM et al. Optical treatment of strabismic and combined strabismicanisometropic amblyopia. Ophthalmology 2012; 119: Kelly KR, Jost RM, Dao L et al. Binocular ipad game vs patching for treatment of amblyopia in children: a randomised clinical trial. JAMA Ophthalmol 2016; 134: Lam DS, Zhao J, Chen LJ et al. Adjunctive effect of acupuncture to refractive correction on anisometropic amblyopia: one-year results of a randomized crossover trial. Ophthalmology 2011; 118: Laria C, Pinero DP, Alio JL. Characterization of Bangerter filter effect in mild and moderate anisometropic amblyopia: predictive factors for the visual outcome. Graefes Arch Clin Exp Ophthalmol 2011; 249: Lin PW, Chang HW, Lai IC et al. Visual outcomes after spectacles treatment in children with bilateral high refractive amblyopia. Clin Exp Optom 2016; 99: Maconachie GD, Farooq S, Bush G et al. Association between adherence to glasses wearing during amblyopia treatment and improvement in visual acuity. JAMA Ophthalmol 2016; 134: Optometry Australia Clinical and Experimental Optometry July