Pinpointing precision: axial length and refraction in paediatric myopia trials—a mixed methods study

Abstract

Purpose

Most paediatric myopia intervention trials report two primary outcomes: axial length and cycloplegic autorefraction. Having one single primary outcome would increase trial efficiency, reduce participant burden, and may improve trial retention. This article aims to evaluate precision of axial length and cycloplegic autorefraction and to explore eye care professionals’ and young people’s views of cycloplegia.

Methods

Mixed methods study with meta-analysis of axial length and cycloplegic spherical equivalent measurements reported by paediatric myopia treatment trials; interviews with eye care professionals; engagement/involvement discussion with children and young people.

Results

In the meta-analysis, mean effect size (standard error, SE) was −0.52 (0.12, p < 0.001, 95% confidence interval (CI) −0.76 to −0.29) for axial length and 0.49 (0.11, p < 0.001, CI 0.27 to 0.71) for cycloplegic spherical equivalent refraction. Difference in mean point estimates was 0.033, SE of the difference 0.164 (CI −0.29 to +0.535). With t-statistic (mean difference/SE of mean difference) of 0.20, two-tailed p value was 0.84. Eye care professionals report that children dislike cycloplegia, describing discomfort, distress, and potential refusal to attend follow-up appointments. Children and young people describe pain and discomfort, remember prolonged effects such as blurred vision and light sensitivity, and complain about receiving insufficient information about adverse effects.

Conclusions

Whilst the precision of axial length and cycloplegic autorefraction spherical equivalent measurements is similar, cycloplegic autorefraction requires longer study visits and inflicts more discomfort. In the design of future trials, researchers, funders, and regulators should consider increasing trial efficiency by selecting axial length as primary outcome and removing cycloplegic refraction from follow-up assessments.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s13063-026-09884-0.

Background

Myopia is a global health problem [1], and research to develop and test interventions to slow its progression in childhood is intense. The number of randomised controlled trials (RCT) is increasing—in 2023, a Cochrane systematic review and network meta-analysis (NMA) included 64 studies; the 2025 update, 104 [1, 2].

Most studies report two primary outcomes: axial length (AL), measured using optical partial coherence interferometry, and spherical equivalent refraction (SER), measured by cycloplegic autorefraction. Whilst it is atypical for RCTs to have two primary outcomes, there is no consensus on which should be primary in myopia RCTs.

The International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use (ICH), Section E9 Statistical Principles, calls for optimisation of the efficiency of RCTs to ensure that only the necessary and meaningful data is collected and analysed, preventing wasted effort and reducing the burden on participants and resources [3]. It would increase overall myopia trial efficiency to agree one primary outcome; its selection should be driven by high precision of measurements and low burden on participants and resources. Standardising on a single primary objective is also extremely helpful in ensuring that between-trial information can be used efficiently.

Refraction is vulnerable to accommodation and requires cycloplegia to increase repeatability and accuracy [4, 5]. Incomplete cycloplegia may result in a variable underestimate of SER [6]. Various cycloplegic regimes are in use, with different agents, frequencies of and intervals between drop instillation, leading to measurement variability. Whilst cyclopentolate is considered the “gold standard” to induce full cycloplegia [7], tropicamide may be equivalent in children with myopia [7, 8]. Full cycloplegia can be difficult to achieve and require multiple instillations over 30–45 min, resulting in increased discomfort to participants and delays in assessments [9]. Cycloplegia lasts for up to 24 h, blurring near vision, causing additional inconvenience.

Axial length measurements using partial coherence interferometry, on the other hand, have high inter-operator repeatability, with intraclass correlation coefficients (ICC) higher than 0.99, even with different devices [10]. They do not require cycloplegia [4, 11], reducing discomfort to children and delays during study visits [4, 11].

To support a discussion about efficient design of myopia intervention trials and assist trialists, funders, and regulators with outcome selection, this brief communication presents a meta-analysis of the precision of AL and SER measurements in myopia RCTs, eye care professionals’ opinions on cycloplegia, and a Patient and Public Involvement and Engagement (PPIE) consultation asking children and young people (CYP) about their views.

Methods

Meta-analysis

On August 11, 2025, we extracted data from a recent Cochrane network meta-analysis (NMA), available in Open Science Framework (https://osf.io/v4tjn) [2]. We included trials that reported both partial coherence interferometry measurements of AL and cycloplegic autorefraction measurements of SER. We excluded abstract-only articles, those that did not report which techniques were used, and those reporting only one outcome. We extracted number of participants and mean and standard deviation (SD) of change in AL and SER at 12 months from baseline in intervention and control groups into a Microsoft Office Excel spreadsheet (Microsoft, version 16.101). We carried out the meta-analysis in SPSS (IBM SPSS Statistics, version 28). We determined the p value of the difference in mean point estimates using the GraphPad QuickCalcs Website: https://www.graphpad.com/quickcalcs/pvalue1/ (accessed August 2025).

Eye care professionals’ views

As part of an ongoing study, approved by London—Brighton and Sussex Research Ethics Committee (IRAS 344573), we held semi-structured interviews with 12 eye care professionals in May/June 2025; here we extract statements about cycloplegic vs non-cycloplegic refraction.

Patient and Public Involvement and Engagement (PPIE) consultation

During a face-to-face meeting on 20 September 2025, we asked members of the NIHR Young Persons’ Advisory Group for eye and vision research (eye-YPAG) who had received cycloplegic eye drops in the past about their experience. Following a group discussion, CYP wrote down their comments on sticky notes.

Results

Included studies and cycloplegia regimes

We included 31 studies (Supp. 1). Nine studies did not report details of their cycloplegia regime (Table 1). In the others, cycloplegic agents, frequency of instillation, endpoints (waiting time or pupil checks), and overall duration of cycloplegic regime differed between studies. Nine studies used topical anaesthetic eyedrops before instillation of cycloplegic drops. Nineteen studies used cyclopentolate 1% only, 8 used tropicamide 1% only, 1 used a combination of both, and 1 study used cyclopentolate 1% in participants younger than 6 years at randomisation, and tropicamide 1% in those aged 6 years and older. The minimum mean duration of the cycloplegia regime was 37.8 (SD 11, range 20 to 65) minutes.

Table 1.

Cycloplegic regimes varied between studies

Study ID	Country/region	Anaesthetic/cycloplegic regime	Minimum total duration of regime (min)
Anstice 2011	New Zealand	1 drop of 0.4% benoxinate, 1% tropicamide twice, spaced 4–6 min apart, wait 30 min	35
Bao 2022	China	2 drops of cyclopentolate, 1%, 5 min apart, wait at least 30 min	35
BLINK Study 2020	USA	1 drop of 0.5% proparacaine or tetracaine followed by 2 drops of 1.0% tropicamide, separated by 5 min, wait 25 min	30
Chamberlain 2019	Canada, Portugal, Singapore, UK	Either 0.5% proparacaine or 0.4% benoxinate; 1 min later, 2 drops of 1.0% tropicamide, separated by 5 min, wait at least 25 min	31
Chan 2022	Hong Kong	2 drops of 1% cyclopentolate 5 min apart, wait 30 min	35
Chen 2023	China	4 drops of 1% cyclopentolate eye drops, 5 min apart, wait until full cycloplegia reached 40–50 min after the last drop, confirmed by pupil reflex and diameter checks	55
Chia 2023	Singapore	3 drops of cyclopentolate 1% spaced 5 min apart, wait 30 min	40
DISC Study 2011	Hong Kong	1 drop of proparacaine 0.4%, 1–2 drops of cyclopentolate 1%, wait not reported	Not reported
Fang 2022	China	1 to 2 drops of 1% cyclopentolate every 5 min over a 15-min period, wait 45 min	60
Garcia-del-Valle 2021	Spain	3 drops of 1% cyclopentolate hydrochloride at 15-min intervals, wait not reported	Not reported
Hansen 2023	Denmark	2 drops of cyclopentolate 1% eye drops 5 min apart, 30 min wait	35
Hieda 2021	Japan	2 drops of cyclopentolate 1% 5 min apart, wait 1 h	65
Jiang 2022	China	1 drop of 0.5% proparacaine, 3 drops of 1% cyclopentolate at 0, 5, and 20 min, wait at least 15 min, pupil light reflex and diameter checked to confirm full cycloplegia (no residual light reflex and pupil dilated to 6 mm or more)	35
Lam 2020	Hong Kong	Proparacaine 0.5% followed by one to two drops of cyclopentolate 1%; cycloplegia confirmed by measuring amplitude of accommodation using push-up method, with accommodation 2D or less; wait not reported	Not reported
LAMP Study 2019	Hong Kong	2 drops of cyclopentolate 1% and tropicamide 1%, 5 min apart. Wait 10 min, repeat. Wait 30 min. If pupillary light reflex still present or pupil size less than 6 mm, repeat	50
Lee 2022	Australia	1–3 drops of 1% cyclopentolate in each eye, dependent on the extent of cycloplegia achieved, assessed as residual pupillary light reflex	Not reported
Liu 2024	Taiwan	3 drops of 1.0% tropicamide separated by 5-min intervals, wait 15 min	25
Loughman 2024	Ireland	1% cyclopentolate eye drops, repeated after 5 min and every 10 min thereafter, if darker iris or incomplete cycloplegia after 30 min wait, assessed as pupil size and responses	35
Peng 2023	China	Not reported	Not reported
Repka 2023	USA	2 drops of cyclopentolate 1% 4 min apart, wait 30 min	34
Ruiz-Pomeda 2018	Spain	3 drops of cyclopentolate 1% at 10-min intervals	20
Sankaridurg 2010	China	1 drop of proparacaine, 2 drops of tropicamide 1%, interval and wait not reported	Not reported
Sankaridurg 2019	China	1 drop of proparacaine 0.4%, 2 drops of 1% tropicamide, wait 30 min, check of pupil dilation and absence of light reflex	35
STAMP Study 2012	USA	0.5% proparacaine, 2 drops of 1% tropicamide 5 min apart, wait 30 min	35
Trier 2008	Denmark	1 drop of 1% cyclopentolate, wait 30 min	30
Wei 2020	China	3 drops of cyclopentolate, 1% at a 5-min interval, wait 30 min; if pupillary light reflex still present or pupil size was less than 6.0 mm, 4th drop of 1% cyclopentolate, wait 15 min	40
Yuval 2024	Israel	Not reported	Not reported
Zadnik 2023	North America, Europe	If younger than 6 years at randomisation: 2 drops of cyclopentolate 1% 5 min apart, wait at least 30 min If 6 years or older at randomisation: 2 drops of tropicamide 1% 5 min apart, wait at least 30 min	35
Zhang 2024	China	Cyclopentolate; concentration, frequency, and timings not reported	Not reported
Zhu 2023	China	Cyclopentanone eye drops 1%, frequency and timings not reported	Not reported

Fifteen studies specified waiting time after last-drop instillation as endpoint, and 7 prescribed a pupil check for light reflex and pupil diameter. One study checked residual accommodative function using the push-up method; successful cycloplegia was reached when accommodation was 2D or less.

Meta-analysis

Mean effect size (Cohen’s d, treated versus control groups) was −0.52 (standard error (SE) 0.12, p < 0.001, 95% confidence interval (CI) −0.76 to −0.29) for axial length and 0.49 (0.11, p < 0.001, CI 0.27 to 0.71) for cycloplegic spherical equivalent refraction. Difference in mean point estimates was 0.033, SE of the difference 0.164 (CI −0.29 to +0.535). With t-statistic (mean difference/SE of mean difference) of 0.20, two-tailed p value was 0.84. The forest plots for AL and SER are shown in Fig. 1.

Fig. 1.

Forest plots summarising a axial length (AL) and b spherical equivalent autorefraction (SER) data in included studies, reflecting measurement precision. Mean effect size (standard error, SE) was −0.52 (0.12, p < 0.001, 95% confidence interval (CI) −0.76 to −0.29) for axial length and 0.49 (0.11, p < 0.001, CI 0.27 to 0.71) for cycloplegic spherical equivalent refraction; there was no statistically significant difference

Eye care professionals’ statements

Eye care professionals stated that children find cycloplegia difficult (Supp. 2). They expressed that for clinical decision-making and issuing a spectacle prescription, non-cycloplegic refraction is more relevant (Supp. 2). The only advantages mentioned were its accuracy in difficult cases, compared with non-cycloplegic refraction, and to confirm an initial diagnosis of myopia. Practitioners did not use cycloplegia in follow-up assessments, as they considered it impractical and not adding value (Supp. 2).

Children’s and young people’s views

Seven CYP aged 10–21 years reported their experience of cycloplegic eyedrops, including pain on administration, lack of information about discomfort, blurred vision, light sensitivity, and how long these effects would last, and inconvenience with attending appointments independently (Supp. 3). As recent trials include adolescents and young adults, driving and using public transport may be important considerations.

Discussion

This study demonstrates that optical partial coherence interferometry AL and cycloplegic autorefraction measurements have similar precision when compared in a meta-analysis which includes a large number of studies and participants, reflected in a similar Cohen’s d. Our qualitative data show that eye care professionals feel that children do not like cycloplegia, and that they see little value in using it at follow-up visits. In our discussions, young people reported discomfort and inconvenience lasting well beyond assessments. Taken together, these may be strong arguments for AL as primary outcome and for limiting the use of cycloplegic autorefraction to the first study visit.

Additional advantages of AL include its validity, as its increase is the main risk factor for sight-threatening myopia complications in adult life [12], and as all current treatments target progressive eye elongation [2, 12].

Previous reports of measurement accuracy and repeatability show high variability of autorefraction measurements, with lower inter-observer and inter-sessional repeatability [6, 13]. Reasons for our different findings likely lie in smaller sample sizes in these reports and strict adherence to study protocols in the RCTs we included [14].

Reasons for including cycloplegic refraction as a trial outcome include familiarity of clinicians and families with optical measurements from spectacle prescriptions and a perception that change in optical power of the eye may be at least as relevant as change in axial length. SER reflects a combination of both structural and optical effects of increasing myopia and interventions. The relationship between AL and SER is not linear, with lens biomechanics compensating for axial elongation in early myopia development [15]. The two outcomes are therefore not interchangeable, but we argue here that AL as the outcome directly related to the risk of future ocular complications should be selected as the primary outcome. In addition, cycloplegic autorefraction requires longer study visits and inflicts more discomfort on study participants; this may affect retention of participants in trials. In studies aiming to the effect of an intervention on ocular shape, biomechanics, and optical power is part of the research question, SER could be used as secondary outcome.

Conclusions

In the design of future myopia trials, researchers, funders, and regulators should consider increasing trial efficiency by selecting AL as the primary outcome and removing cycloplegic autorefractions from follow-up visits when not relevant to the research question. Regulators could usefully describe a preference for AL as the primary objective in pivotal trials in paediatric myopia trials.

Abbreviations

AL

Axial length

CI

Confidence interval

CYP

Children and young people

SE

Standard error

SER

Spherical equivalent refraction

Authors’ contributions

NF and ADN developed the concept for this work. ADN collated, analysed all included data, and drafted the manuscript. Both authors discussed findings. NF gave critical input into manuscript amendments. Both authors read and approved the final manuscript.

Funding

This work did not receive specific funding. ADN is funded by an NIHR Senior Clinical and Practitioner Award and supported by the National Institute for Health and Care Research (NIHR) Biomedical Research Centre based at Moorfields Eye Hospital NHS Foundation Trust and UCL Institute of Ophthalmology. The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR, or the Department of Health and Social Care. Funders played no role in the conduct of this work.

Data availability

Data are available on request to the corresponding author.

Declarations

Ethics approval and consent to participate

For interviews with eye care professionals: London—Brighton and Sussex Research Ethics Committee (IRAS 344573).

Consent for publication

Not applicable.

Competing interests

ADN has received honoraria for contributing to educational events and advisory boards from Santen, Novartis, Bayer, CooperVision, Zeiss, Thea, Essilor and Hoya. She served as principal or co-ordinating investigator for the NEVAKAR CHAMP, CHAMP-UK, OCUMENSION, MYOPIA-X and MODERATO clinical trials. None of these has an impact on the work presented here. NF has no interest to declare.