Effect of Preoperative Ocular Surface Optimization on Biometric Accuracy and Refractive Outcomes After Cataract Surgery: A Systematic Review with Narrative and Exploratory Quantitative Synthesis

Abstract

Background

The ocular surface and tear-film instability can affect keratometry and biometry before cataract surgery. This review evaluated whether preoperative ocular surface optimization improves biometric accuracy, keratometric repeatability, intraocular lens (IOL) power prediction, and postoperative refractive outcomes.

Methods

PubMed/MEDLINE was searched on 24 June 2026 and Cochrane CENTRAL on 1 July 2026 without language, study-type, age, or date filters. One reviewer screened records in Rayyan and the exported CENTRAL trial list, assessed full texts against prespecified eligibility criteria, and extracted study data; screening and extraction were not independently duplicated. The primary synthesis was narrative because study designs and outcome reporting were heterogeneous, with exploratory random-effects sensitivity analyses performed only for sufficiently compatible datasets.

Results

PubMed identified 179 records and CENTRAL identified 44 trial records. PubMed screening selected 23 records for full-text review. CENTRAL added two completed published studies and identified five registry-only records awaiting results. Overall, 25 completed published studies were included, while five registry-only records were tracked but not synthesized. Heterogeneity in interventions, comparators, and outcomes limited quantitative pooling.

Conclusions

Preoperative ocular-surface optimization may improve the stability of cataract measurements and refractive prediction in selected patients, particularly when dry eye disease, meibomian gland dysfunction, tear-film instability, or ocular-surface irregularity is present. Heterogeneity and predominantly nonrandomized evidence preclude a definitive pooled treatment effect. Final biometry should be performed after clinically significant ocular-surface disease has been treated and the surface is reasonably stable.

Plain Language Summary

Cataract surgery depends on accurate eye measurements before surgery so the surgeon can choose the most suitable intraocular lens power. Dry eye, meibomian gland dysfunction, and other ocular-surface problems can make these measurements less stable and may reduce the accuracy of the predicted refractive result.

We reviewed 25 completed published studies that tested treatments used before cataract measurements, including artificial tears, prescription eye drops, treatments for meibomian gland dysfunction, and procedures used to improve an irregular ocular surface. Many studies reported more stable corneal measurements, changes in planned lens power, or better refractive prediction after treatment.

The studies differed substantially in the type and severity of ocular-surface disease, treatment used, study design, timing of measurements, and outcomes reported. These differences prevented a definitive pooled meta-analysis, so we relied mainly on a narrative synthesis and used limited quantitative analyses only as exploratory evidence.

Overall, the evidence supports identifying and treating clinically significant ocular-surface disease before final cataract biometry. In practical terms, surgeons should consider postponing final measurements, especially when refractive precision is critical, until the ocular surface is reasonably stable. Larger randomized studies using standardized refractive outcomes are still needed.

Introduction

In contemporary cataract surgery, outcomes are increasingly defined by precise visual correction rather than cataract removal alone. Accurate IOL power calculation depends heavily on reliable keratometry and a stable ocular surface.^1–25

Optical degradation from ocular-surface disease, including dry eye disease, meibomian gland dysfunction, tear-film instability, or epithelial irregularity, can distort keratometric measurements and contribute to errors in IOL power selection or astigmatism-axis planning.^{2,6,7,9,12,15,20,22,23,25}

Dry-eye signs are common and frequently underrecognized in cataract candidates. In the PHACO study of 136 patients aged 55 years or older presenting for cataract surgery, only a minority had a previous dry-eye diagnosis, while 62.9% of tested eyes had a tear break-up time of 5 seconds or less and 77% had positive corneal staining.²⁶ Dry eye disease is commonly described along a continuum between aqueous-deficient and evaporative mechanisms, which may overlap; meibomian gland dysfunction is a major contributor to the evaporative phenotype.²⁷ Cataract surgery can further worsen tear-film function through corneal nerve and surface injury, tear-film disruption, meibomian gland dysfunction, and postoperative inflammation.²⁸ These observations make overlooked preoperative ocular-surface disease particularly relevant when refractive precision is a treatment goal.

Older adults undergoing cataract surgery therefore represent a population in whom preoperative ocular-surface assessment can be clinically important. Interventions evaluated in this setting include artificial tears and lubricants, topical anti-inflammatory or tear-stimulating medications, perfluorohexyloctane, intense pulsed light with meibomian gland expression, thermal pulsation, cryopreserved amniotic membrane, superficial keratectomy for epithelial irregularity, and related modalities. We evaluated whether these approaches improve IOL-power prediction, keratometric repeatability, and postoperative refractive outcomes.^{1,3,4,8–12,14–25}

Materials and Methods

Protocol and Review Design

A prespecified local protocol was developed before full-text eligibility assessment and was followed throughout the review. The review was prepared and reported in accordance with the PRISMA 2020 statement; the completed 27-item checklist is provided in Supplementary File 1.²⁹

The PICO (population, intervention, comparator, outcome) question guiding the review was: In adults undergoing cataract surgery or cataract biometry, does a preoperative ocular surface intervention, compared with usual care, no treatment, baseline measurements, a fellow-eye control, or another active treatment, improve biometric accuracy, keratometric repeatability, IOL power calculation, or postoperative refractive outcomes?

The review was not formally registered. The local protocol was not publicly posted and is available from the corresponding author on reasonable request. During full-text assessment, eligibility was clarified to allow relevant small case series with more than one participant when they reported usable pre/post biometric or refractive outcomes; isolated single-patient case reports remained excluded. No other substantive protocol amendments were made.

Eligibility Criteria

Eligible studies evaluated adults undergoing cataract surgery or cataract biometry and assessed a preoperative intervention intended to treat or optimize dry eye disease, meibomian gland dysfunction, tear-film instability, epithelial irregularity, or another clinically relevant ocular surface condition. Eligible interventions included artificial tears or lubricants, sodium hyaluronate, lipid-containing tears, diquafosol, rebamipide, lifitegrast, cyclosporine, perfluorohexyloctane, intense pulsed light with meibomian gland expression, thermal pulsation, cryopreserved amniotic membrane, superficial keratectomy, and related ocular surface treatments. Comparators included usual care, no treatment, baseline or pretreatment measurements, untreated controls, fellow-eye controls, or another active intervention. Studies had to report at least one relevant biometric, keratometric, IOL-calculation, refractive, ocular-surface, or adverse-event outcome.

Randomized, nonrandomized, prospective, retrospective, paired-eye, crossover, before-and-after, cohort, comparative, and relevant small case-series designs were eligible. Reviews, editorials, letters without original data, laboratory or animal studies, studies unrelated to cataract surgery or cataract biometry, postoperative-only treatment studies, refractive-surgery-only studies, pure prevalence studies without an intervention, isolated single-patient case reports, and duplicate or linked records without additional usable information were excluded. Registry-only records without completed, published, extractable results were tracked separately but were not included in the completed-study synthesis. No date or language restriction was applied.

Search Strategy and Information Sources

The literature was searched on the PubMed/MEDLINE database on 24 June 2026, and via the Cochrane Library Advanced Search using the Cochrane Central Register of Controlled Trials (CENTRAL) search interface on 1 July 2026. No search restrictions for language, study design, age group or date were imposed during the database searches, in order to identify all relevant literature.

The complete PubMed/MEDLINE search strategy, search date, and limits are provided in Supplementary File 2. A full record of retrieved records was maintained and saved. For comprehensive coverage and to reduce database-specific retrieval bias, a supplemental Cochrane CENTRAL search was also performed; CENTRAL identified 44 trial records on 1 July 2026.

The complete CENTRAL search strategy is also provided in Supplementary File 2.

The output of each search was exported. The 179 PubMed records were imported into Rayyan (rayyan.ai). Rayyan was used for record organization only; its AI-assisted screening functions were not used. One reviewer screened all PubMed records sequentially and manually reviewed the four records flagged as possible duplicates. These were related reports rather than true duplicates and were retained.

CENTRAL records were compared manually with the PubMed set. Five duplicate published records and one linked registry record were not carried forward. The remaining CENTRAL records were screened manually by the same reviewer from the exported trial list. No machine-learning classifier or automated exclusion tool was used.

The same reviewer conducted title/abstract screening and full-text eligibility assessment. Decisions were applied against the prespecified criteria, and the final inclusion set and manuscript summaries were reviewed by the coauthors before submission. The Rochet et al 2021 TORIDE full text was obtained through author correspondence. Screening was not independently duplicated.

Full-Text Retrieval and Eligibility Determination

All 23 reports selected from PubMed, including the Rochet et al 2021 TORIDE article after it was obtained through author correspondence, and the two additional completed studies identified through CENTRAL were retrieved and included. No retrieved completed published full-text report was excluded at this stage. Thus, 25 completed published studies were included in the completed-study synthesis. Five registry-only records were monitored separately and were not synthesized because completed published data were unavailable.

Data Collection and Data Items

A single reviewer extracted data from eligible full-text reports into a structured workbook. Extracted variables included study identity, year, country, design, participant and ocular-surface characteristics, sample size, intervention, comparator, treatment duration, biometry or keratometry method, IOL formula where reported, outcome definitions, measurement timing, refractive prediction error, refractive-accuracy thresholds, keratometric and astigmatic outcomes, calculated IOL power, ocular-surface measures, adverse events, funding information where available, and reported statistical estimates. Information was checked against the full text and available supplementary material. Extraction was not independently duplicated; the final extracted summaries and interpretations were reviewed by the coauthors before submission. No automated data-extraction tool was used.

All results compatible with the prespecified outcome domains were sought. When a study reported multiple IOL formulas, measurement times, or analyses, these were recorded separately. Results most directly corresponding to final preoperative biometry and postoperative refractive assessment were prioritized for the main narrative synthesis, while alternative formulas or time points were retained for sensitivity or contextual interpretation. Missing paired variances or within-participant correlations were not treated as known values; assumptions used in exploratory analyses are described below.

Outcome and Synthesis

The primary outcome was postoperative refractive prediction error, particularly mean absolute prediction error (MAE) or absolute prediction error. Secondary outcomes were mean prediction error; proportions of participants or eyes with absolute prediction error within ±0.25, ±0.50, ±0.75, and ±1.00 D; keratometric repeatability; astigmatism magnitude and axis repeatability; change in IOL power; tear break-up time; ocular-surface symptom scores; ocular-surface staining; and adverse events. Studies were grouped for synthesis by outcome family, study design, comparator, and availability of compatible data. The primary synthesis was narrative.

A definitive pooled meta-analysis for the primary outcome was not undertaken because of substantial clinical and methodological heterogeneity in interventions, comparators, measurement timing, IOL formulas, study designs, and outcome definitions. In addition, most before-and-after and paired-eye studies did not report the paired variance or within-participant correlation required for a robust paired analysis.

Exploratory random-effects sensitivity analyses were conducted only when datasets were sufficiently compatible and were not treated as primary evidence. Mean differences (MDs), expressed in diopters, were used for continuous refractive-error outcomes; negative values favored ocular-surface optimization. Risk ratios (RRs) were used for the proportion achieving refractive accuracy within ±0.50 D; values greater than 1 favored optimization. For the pre/post MAE analysis, an assumed within-participant correlation of r=0.5 was used because paired variances were unavailable. When categorical event counts were reconstructed from rounded percentages, they were treated as approximate. No other missing summary statistics were imputed.

Random-effects models were used for the exploratory syntheses, and statistical heterogeneity was described using I.² Potential causes of heterogeneity were assessed qualitatively by comparing intervention class, study design, comparator, timing, and outcome definition; formal subgroup analysis, meta-regression, and tests of interaction were not performed because too few compatible studies were available. Study characteristics and synthesis decisions were tabulated, and exploratory results were displayed in forest plots. Calculations were documented in the project effect-size workbook, and forest plots were generated with Matplotlib version 3.10.8.

Reporting Bias and Certainty of Evidence

Formal assessment of reporting bias or small-study effects was not performed because fewer than 10 compatible datasets contributed to each exploratory synthesis and the included designs and outcome definitions were highly heterogeneous. A formal certainty-of-evidence framework, such as GRADE, was not applied. Confidence in the findings was interpreted qualitatively with attention to study risk of bias, inconsistency, indirectness, and imprecision.

Risk of Bias

One reviewer performed a design-specific structured risk-of-bias assessment. Randomized and randomized crossover studies were appraised using RoB 2 concepts and randomized-trial criteria; nonrandomized comparative, cohort, quasi-experimental, before-and-after, paired-eye, and case-series designs were assessed using the relevant JBI critical-appraisal framework, with ROBINS-I-style considerations where confounding and nonrandomized treatment allocation were central concerns. Judgments considered participant selection, randomization or allocation where applicable, confounding, deviations from intended interventions, missing data, outcome measurement, selective reporting, and unit-of-analysis issues. For manuscript-level summarization, studies were grouped as lower risk/low-to-some concerns, some concerns, or high/serious concerns. Risk-of-bias assessment was performed by one reviewer and was not independently duplicated. Study-level tools, judgments, rationales, and verification status are provided in Supplementary File 3.

Results

Study Selection

The search and screening process is illustrated in Figure 1. PubMed/MEDLINE identified 179 records and CENTRAL identified 44 trial records, for 223 records overall. Six duplicate or linked records were not carried forward, leaving 217 unique records screened. Of these, 187 were excluded before full-text/status handling, leaving 30 records tracked. Twenty-five completed published studies were included in the synthesis, and five registry-only records were tracked separately because completed extractable results were unavailable.

Figure 1.

Study selection flow. PubMed/MEDLINE and CENTRAL records were screened; registry-only records without completed published data were tracked separately and were not synthesized.

All 23 PubMed-selected reports were retrieved and included after full-text assessment; no completed published report was unavailable or excluded after retrieval.

Within CENTRAL, five duplicate published records and one linked registry record were not carried forward, 31 records were excluded at record level, two additional completed studies were included, and five registry-only records were retained for monitoring but not synthesized.

Characteristics of the Included Studies

The 25 completed published studies included randomized, prospective, retrospective, paired-eye, before-and-after, within-subject, and small case-series designs. Participants were cataract candidates with preoperative dry eye disease, meibomian gland dysfunction (MGD), tear-film instability, epithelial basement membrane dystrophy, ocular-surface irregularity, or a related preoperative ocular-surface condition. An additional completed thermal-pulsation study before extended-depth-of-focus IOL implantation and a brief sodium-hyaluronate biometry study were added through the supplemental CENTRAL search. Study characteristics are summarized in Table 1.

Table 1.

Characteristics of the Retrieved Included Reports

No.	Study	Design	Ocular-Surface Phenotype/Reported Severity	Intervention	Comparator	Main Outcome Contribution
1	Teshigawara T 20251	Prospective paired-eye comparative study	Meibomian gland dysfunction-related dry eye disease	One eye: intense pulsed light with meibomian gland expression (IPL-MGX)	Fellow eye: 3% diquafosol sodium	Postoperative absolute prediction error/P-SE accuracy
2	Röggla V 20212	Prospective randomized crossover before-and-after clinical study	Dry eye disease/tear film instability; artificial tear effect on K readings	Instillation of low- or high-viscosity artificial tears before keratometry	Native baseline keratometry; normal vs dry eye subgroups	Short-term variability of keratometric readings after artificial tears
3	Hovanesian J 20203	Multicenter prospective open-label study	Dry eye disease	Lifitegrast ophthalmic solution 5%	Within-subject pretreatment biometry vs post-treatment biometry	Accuracy of preoperative anterior corneal power measurements for predicting postoperative spherical equivalent
4	Teshigawara T 20224	Single-center prospective open-label study	Short TBUT-type dry eye	2% rebamipide ophthalmic suspension	Within-subject pretreatment biometry vs post-treatment biometry	Predicted postoperative spherical equivalent accuracy
5	Lee CY 20255	Retrospective cohort study	Dry eye disease	Perioperative hyaluronic-acid-containing artificial tears	No HA-containing artificial tears	UDVA, postoperative SE, and astigmatism vector analysis
6	Teshigawara T 20246	Multicenter prospective paired-eye study	Short TBUT-type dry eye	Long-acting 3% diquafosol sodium in one eye	Contralateral untreated eye	Astigmatism measurement repeatability
7	Nilsen C 20247	Prospective randomized controlled trial with three arms	Dry eye disease by DEWS II criteria	Artificial tears for dry-eye subjects	Dry-eye no treatment; non-dry-eye controls	Variability of keratometry and postoperative refractive precision
8	Kawagoe T 20258	Single-center prospective open-label before-after study	MGD-related dry eye	IPL-MGX	Within-eye pretreatment biometry vs post-treatment biometry	Accuracy of predicted postoperative spherical equivalent
9	Kim J 20219	Comparative clinical study; pretreatment vs non-pretreatment groups	Dry eye disease	Topical 0.5% loteprednol etabonate and 0.05% cyclosporin A	No pretreatment	Mean prediction error, MAE, refractive surprise
10	Yeu E 202210	Pilot comparative study with prospective treatment group and retrospective control group	Epithelial basement membrane dystrophy causing ocular surface irregularity	Debridement followed by self-retained cryopreserved amniotic membrane	Debridement followed by bandage contact lens	Ocular surface optimization and refractive/biometry-related outcomes
11	Wongskhaluang J 202511	Prospective repeated-measures self-controlled study	Moderate to severe dry eye disease	Cryopreserved amniotic membrane	Within-eye pretreatment vs post-treatment biometry/refractive prediction	Refractive accuracy based on manifest refraction at 1 month postoperatively
12	Schlatter A 202412	Prospective randomized controlled investigator-masked paired-eye study	Meibomian gland dysfunction	Vectored thermal pulsation therapy (LipiFlow) in one eye	Contralateral control eye	Repeatability of emmetropic IOL power and keratometry measurements
13	Rochet E 202113	Prospective within-subject study	Cataract eyes with toric IOL indication; BUT <5 s subgroup	Artificial tears 1 min before repeat biometry/topography	Same eyes before tears; simulated no-tears toric calculation	Astigmatism, toric IOL calculation, MAEPA
14	Chen N 202114	Prospective comparative study	Dry eye disease graded by NIBUT	0.1% or 0.3% sodium hyaluronate eye drops	Different NIBUT severity and drop concentration/time comparisons	Regularity/corneal surface and biomechanics before operation
15	Szabelska P 202315	Case series/case-report-labeled study	MGD, dry eye syndrome, blepharitis	Thermal Pulsation System therapy	Within-eye pretreatment vs post-treatment measurements	Changes in IOL calculation, astigmatism power/axis, and planned surgical treatment type
16	Su N 202516	Prospective observational study	Predisposition to dry eye/tear-film instability	3% diquafosol sodium eye drops	Before vs after DQS; p-DE vs control	Changes in IOL power calculation after DQS
17	Matossian C 202017	Single-center pilot study	MGD-associated dry eye	Thermal Pulsation System treatment	Simulated pre-TPS planning vs actual post-TPS planning/outcomes	Astigmatism management and residual refractive astigmatism
18	Biela K 202618	Prospective comparative study with treated DED, treated non-DED, and untreated control groups	Dry eye disease and non-DED groups; ocular surface preparation	Trehalose moisturizing drops plus eyelid margin hygiene	DED treated, non-DED treated, and control without treatment	Change in MAE of postoperative refraction
19	Kimmell SW 202619	Single-center retrospective study	Epithelial basement membrane dystrophy with irregular astigmatism	Superficial keratectomy followed by cryopreserved amniotic membrane	Within-eye pretreatment vs post-treatment	HOA RMS, cylinder power, and IOL biometry changes
20	Teshigawara T 202120	Prospective comparative treatment study	Dry eye with short TBUT	2% rebamipide ophthalmic suspension	Artificial tear ophthalmic solution (Mytear)	Discrepancy in astigmatism power and axis between two intra-patient keratometric measurements
21	Hovanesian JA 202621	Phase 4 multicenter open-label single-arm study	Dry eye disease	Perfluorohexyloctane ophthalmic solution (PFHO)	Within-subject baseline vs post-PFHO treatment	Mean difference in absolute deviations between manifest refraction spherical equivalent and predicted refractive error
22	Hovanesian JA 202122	Multicenter prospective open-label study	Dry eye disease	Topical cyclosporine 0.09%	Within-subject pretreatment vs post-treatment biometry	Absolute prediction error of 1-month spherical equivalent refractive outcome
23	Lacmanović Lončar 202623	Cross-sectional/interventional pre-post cohort with control group	Dry eye disease	Lipid-containing artificial tears	Control group without DED; pretreatment vs posttreatment in DED group	Keratometric reproducibility and IOL power calculation stability; postoperative refractive accuracy
24	Matossian C 202324	Prospective randomized open-label multicenter crossover study	Cataract patients with mild-to-moderate MGD undergoing extended-depth-of-focus IOL implantation	Preoperative vectored thermal pulsation (LipiFlow)	No preoperative LipiFlow control/delayed treatment	Refractive predictability within threshold and ocular surface outcomes
25	Yang J 202325	Randomized comparative short-term biometry study	Cataract patients with moderate-to-severe dry eye	1 g/L sodium hyaluronate artificial tears	3 g/L sodium hyaluronate artificial tears; baseline before instillation	K1 and IOL power changes at 1, 5, and 15 minutes after artificial tears

Notes: Dry-eye severity was not reported on a common scale across studies. Phenotype and severity descriptors reproduce the terminology used in the original reports; no unreported severity category was imputed.

Findings by Ocular-Surface Phenotype, Reported Severity, and Intervention Class

Dry-eye severity was not reported using a common scale across the included studies, so a uniform mild/moderate/severe subgroup analysis could not be performed without imposing unsupported categories. We therefore classified studies using the phenotype and severity descriptors reported by the original investigators and then by intervention class. MGD-related or evaporative phenotypes were evaluated mainly with IPL-MGX or thermal pulsation; short-TBUT or tear-film-instability phenotypes were treated with rebamipide, diquafosol, or lubricants; broader DED cohorts were treated with lubricants, lifitegrast, cyclosporine, perfluorohexyloctane, or combined anti-inflammatory therapy; and ocular-surface irregularity or EBMD studies used surface procedures with cryopreserved amniotic membrane. Two studies explicitly enrolled moderate-to-severe dry eye and one enrolled mild-to-moderate MGD. Across these groups, the direction of effect generally favored ocular-surface stabilization for measurement repeatability or refractive planning, but heterogeneity in severity definitions, comparators, and outcomes precluded a reliable severity-stratified pooled estimate.

The author-provided Rochet et al 2021 TORIDE full-text data were added; it was a prospective toric IOL study of artificial tear use one minute before repeat biometry/topography and toric IOL calculation.¹³

Refractive Prediction Error and Absolute Mean Error

The refractive prediction error was measured and reported using different criteria in the retrieved trials, including absolute mean error, absolute deviation from predicted refraction, and spherical equivalent prediction within various dioptric margins. One retrieved trial, Kim et al, provided direct comparison of preoperative management vs control, where preoperative management for dry eyes had statistically significantly lower absolute prediction error than the untreated control group for both SRK/T and Barrett Universal II calculations, but this was an observational comparison that may be subject to bias. Other preoperative before-and-after trials reported significant improvements in absolute error or target deviation with ocular surface optimization with intense pulsed light and meibomian gland expression, cryopreserved amniotic membrane, trehalose and eyelid hygiene, perfluorohexyl octane, and cyclosporine 0.09% with dry eye, although the results of these trials were not independent comparisons and many failed to provide adequate statistics for definitive meta-analysis. The Rochet et al 2021 TORIDE article included data specific to toric planning: in the overall cohort, the absolute mean predicted astigmatism error decreased from 0.40 D (SD, 0.39 D) to 0.35 D (SD, 0.31 D) with preoperative artificial tears (NS); however, within the subgroup with an aqueous tear breakup time of <5 seconds, the mean predicted astigmatism error was reduced from 0.48 D (SD, 0.50 D) to 0.37 D (SD, 0.25 D) with the use of artificial tears (P = 0.048) which demonstrated clinically relevant significance with tear-film instability.¹³ Quantitative synthesis decisions and exploratory estimates are summarized in Table 2. The exploratory pre/post MAE and ±0.50 D sensitivity analyses are shown in Figures 2 and 3, respectively.

Table 2.

Quantitative Synthesis Decision and Exploratory Estimates

Analysis Group	Studies/Data	Effect Estimate	Interpretation	Synthesis Role
Direct treatment-vs-control MAE	Kim 2021,9 SRK/T formula	MD −0.19 D favoring treatment	Preoperative dry-eye management associated with lower MAE, but observational	Primary direct evidence, not pooled
Formula sensitivity	Kim 2021,9 Barrett Universal II	MD −0.14 D favoring treatment	Same participants as SRK/T, so not combined with SRK/T in one model	Sensitivity/context
Active-vs-active paired-eye comparison	Teshigawara 2025,1 IPL-MGX vs DQS	Absolute error 0.28 ± 0.14 D vs 0.40 ± 0.18 D	IPL-MGX produced lower absolute error than DQS	Narrative or active-comparator sensitivity
Toric astigmatism prediction error	Rochet 2021 TORIDE13	MAEPA: 0.40±0.39 D without tears vs 0.35±0.31 D with tears (P=0.29); BUT <5 s: 0.48±0.50 vs 0.37±0.25 D (P=0.048)	No significant overall MAEPA improvement, but significant benefit in tear-film instability subgroup	Narrative; not pooled with SE MAE
Exploratory pre/post MAE sensitivity	Kawagoe 2025;8 Wongskhaluang 2025;11 Biela 2026 DED group;18 Hovanesian 2026;21 Hovanesian 2021;22 291 participants	Approximate pooled MD −0.14 D, 95% CI −0.24 to −0.04; I2 = 90.7%	Favors optimization but has high heterogeneity and imputed paired variance	Supplement only
Exploratory ±0.50 D sensitivity	Eight compatible pre/post or comparator datasets, including Matossian 2023;24 614 unique participants	Approximate RR 1.18, 95% CI 1.01 to 1.36; I2 = 77.3%	Direction generally favors optimization, but Matossian 2023 did not show benefit and heterogeneity increased; paired binary data and outcome definitions remain limitations	Supplement only

Figure 2.

Exploratory pre/post MAE sensitivity analysis. Mean differences are post/intervention minus pre/control; negative values favor ocular-surface optimization. Paired variances were unavailable for several studies, so the exploratory analysis assumed a within-participant correlation of r=0.5. This analysis was not treated as primary evidence.

Figure 3.

Exploratory sensitivity analysis for refractive accuracy within ±0.50 D. Risk ratios greater than 1 favor ocular-surface optimization. Event counts reconstructed from rounded percentages were treated as approximate. This analysis was exploratory and was not treated as primary evidence.

Exploratory Quantitative Synthesis

The exploratory pre/post MAE analysis included five studies with 291 participants and produced an approximate pooled MD of −0.14 D (95% CI, −0.24 to −0.04; I²=90.7%), favoring ocular-surface optimization. The exploratory analysis of refractive accuracy within ±0.50 D included eight compatible study datasets representing 614 unique participants and produced an approximate RR of 1.18 (95% CI, 1.01 to 1.36; I²=77.3%). The five studies contributing to the MAE analysis were a subset of the eight studies in the ±0.50 D analysis; therefore, participants were not added across analyses, and 614 was the total number of unique participants represented in the quantitative component.

Both syntheses were interpreted as exploratory because paired variance was unavailable for several studies, the MAE model used an assumed correlation of r=0.5, some categorical counts were reconstructed from rounded percentages, and clinical and statistical heterogeneity remained substantial. The contributing studies and analysis assumptions are summarized in Table 2 and displayed in Figures 2 and 3.

Keratometry, Astigmatism, and IOL Calculation

A number of the trials focused primarily on the stability of measurements and specifically keratometry, the repeatability of astigmatism (magnitude and axis) or planned IOL power. The use of artificial tears before keratometry has the potential to introduce temporary astigmatism measurement error, particularly in patients with dry eyes and viscous eye drops. Diquafosol, rebamipide, thermal pulsation with meibomian gland expression, IPL and lipid-containing eye drops reduced measurement variability, astigmatic and IOL planning, in many of the trials but these measures were too heterogeneous for a direct, pooled analysis to allow for a conclusive comparison.^2,7,13,25

Ocular Surface Measures and Adverse Events

The majority of treatment studies indicated that some ocular surface features like tear break-up, staining, the dry eye questionnaires and/or meibomian gland evaluations improved with ocular surface optimization. Adverse events reported were rare, but the information regarding adverse events is limited by variable methods of data collection among the included studies, which limits the interpretation.^{1,3,4,8–12,14–24}

Risk of Bias

Among the 25 completed studies, two randomized studies were in the lower-risk/low-to-some-concerns category, 17 had some concerns, and six had high or serious concerns. The principal drivers were nonrandomized or uncontrolled before-and-after designs, retrospective or historical controls, small samples, incomplete paired-variance reporting, limited allocation or masking details, and potential confounding. The five registry-only records were not assessed because completed published results were unavailable. The summary is presented in Table 3, and study-level frameworks, judgments, and rationales are provided in Supplementary File 3.

Table 3.

Risk-Of-Bias Summary

Judgment	Count	Interpretation
Lower risk/low-to-some concerns	2	Randomized studies with comparatively stronger design features and objective outcomes; residual uncertainty remained in some domains.
Some concerns	17	Commonly due to open-label or before-after design and limited control of confounding
High/serious concerns	6	Usually retrospective, small case series, historical controls, or substantial confounding risk
Pending	0	None after Rochet 2021 retrieval
Registry-only records, not assessed	5	No completed published results identified

Reporting Bias and Certainty of Evidence

No formal reporting-bias, small-study-effects, or certainty-of-evidence assessment was performed. Accordingly, the quantitative findings were treated as exploratory and the overall conclusions were framed cautiously.

Discussion

Key Findings

Ocular surface management before cataract biometry (CB) was associated with improved measurement stability in many included studies. The directional evidence supports performing CB on a stable ocular surface in patients with dryness or irregularity. Optimization frequently improved tear-film measures and changed IOL or astigmatism planning. Evidence for a definitive pooled effect on postoperative refractive prediction accuracy remains limited because most studies used before-and-after or within-patient designs and did not report paired variance.

Clinical Implication

There is an opportunity to develop a clinically relevant workflow for cataract biometry to optimize and improve refractive outcomes in many of our patients. This involves an examination and optimization of the ocular surface in all patients presenting for CB. Final, detailed measurements, especially in those planning for premium lenses (ie, toric, monovision, and extended depth of focus), are to be delayed until the ocular surface is optimized, as this provides an ideal starting point for ensuring accurate CB.

Why Definitive Pooling was Not Possible

The studies varied in ocular surface classification, class of treatment, length of time to optimize, biometry device, IOL formula, IOL type, time of cataract surgery, outcome definitions, and reporting of statistics. Some treated the acute effects of drops, others weeks of drops, and others procedures such as IPL, thermal pulsation or superficial keratectomy. Combining these into a primary meta-analysis would generate an “exact” figure which is not a clinically meaningful summary.^{1,4,6,8,12,15–17,20,23,24}

The current manuscript therefore adopts narrative synthesis, study estimates and clearly marked exploratory analyses.

Strengths

Prespecified search and eligibility criteria, documented duplicate management, a transparent full-text screening log, a supplemental CENTRAL search, retrieval of the TORIDE report through author correspondence, and a clear distinction between narrative synthesis and exploratory quantitative analyses are strengths of the review.

Limitations

The current review has several important limitations. Search was limited to PubMed/MEDLINE and CENTRAL only; Embase, Scopus or Web of Science was not searched, therefore it is possible that relevant articles were missed. The recently available full text for Rochet et al 2021 was obtained via correspondence and included; however, five identified registry-only records that were not published as completed studies were not synthesized because no completed extractable results were available.¹³

Screening, data extraction, and risk-of-bias assessment were performed by one primary reviewer and were not independently duplicated. The final inclusion set and manuscript summaries were reviewed by the coauthors before submission, but this does not replace independent duplicate screening or extraction and remains an important source of potential selection and extraction bias.

Clinical and methodological heterogeneity, inconsistent dry-eye severity definitions, incomplete paired-variance reporting, differences in measurement timing, interventions, biometry devices, IOL formulas, and outcome definitions limited synthesis. Formal reporting-bias and certainty-of-evidence assessments were not performed, and the study-level risk-of-bias judgments were not independently duplicated. These limitations reduce confidence in treatment-specific effect estimates and support cautious, narrative-first interpretation.

Research Implications

Future studies in the area should randomize cataract candidates with well-defined dry eye disease or meibomian gland dysfunction (MGD) to optimize treatment compared with routine care; measure biometry after a specific period following optimization to confirm stability, and report postoperative mean absolute prediction error, Mean Prediction Error, thresholds for acceptable prediction accuracy, repeatability of keratometry readings, changes in IOL power, ocular surface outcome measurements, adverse event rates and pooled variance as appropriate; clearly indicate biometry device used, the specific IOL formula applied and IOL type, specify time interval from drop instillation to biometry, and clarify whether or not both eyes from an individual patient were included in the sample.

Conclusions

Preoperative ocular-surface optimization appears to improve the stability of cataract biometry and may improve refractive prediction in selected patients, particularly when dry eye disease, meibomian gland dysfunction, short tear break-up time, or epithelial irregularity is present.

The evidence is not sufficiently homogeneous to support a single pooled treatment effect or a universal treatment protocol, and confidence in treatment-specific estimates is limited by predominantly nonrandomized designs, inconsistent severity definitions, and variable outcome reporting. In clinical practice, clinically significant ocular-surface disease should be identified and treated before final biometry; when refractive precision is especially important, such as toric, multifocal, or extended-depth-of-focus IOL planning, final measurements should be deferred until the ocular surface is reasonably stable.

Future randomized studies should use standardized dry-eye severity grading, clearly defined intervention classes, standardized refractive outcomes, and complete paired-variance reporting to permit more reliable comparative synthesis.

Funding Statement

The authors received no financial support for the research, authorship, and/or publication of this article.

Abbreviations

CENTRAL, Cochrane Central Register of Controlled Trials; DED, dry eye disease; DQS, diquafosol sodium; EBMD, epithelial basement membrane dystrophy; IOL, intraocular lens; IPL-MGX, intense pulsed light with meibomian gland expression; MAE, mean absolute error; MGD, meibomian gland dysfunction; NIBUT, non-invasive tear break-up time; PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses; TBUT, tear break-up time.

Data Sharing Statement

All data used in this systematic review were extracted from previously published studies and publicly available registry records. Study characteristics, quantitative synthesis decisions, and summary results are provided in the manuscript. The structured extraction workbook, risk-of-bias workbook, effect-size workbook, and analytic materials are not publicly deposited but are available from the corresponding author on reasonable request, subject to copyright restrictions on source publications.

Ethics Approval

Ethics approval was not required because this study is a systematic review of previously published data and publicly available records.

Informed Consent

Informed consent was not required because this study did not involve direct contact with human participants or the collection of new individual patient data.

Author Contributions

All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Disclosure

The authors declare that they have no conflicts of interest in relation to this work.