Selection process

A dedicated team of five persons is formed with two key members responsible for developing the process and standards. The two assessors conduct the processes of study screening, selection, data extraction and bias risk assessment independently. The final plan will be determined through a group discussion. All team members should receive training before data extraction and collection. The titles and abstracts of potential studies are compared with the eligibility criteria and thoroughly assessed by two appraisers to determine their suitability. Subsequently, these two appraisers will carefully review the full text of the potential studies. Any discrepancies between the two appraisers are to be resolved through a consensus meeting involving the entire study team. During the screening process, we first use software to remove duplicate references, followed by manual deletion. After removing duplicates, the two assessors independently perform manual screening, excluding case series, conference reports, grey literature, news articles and literature reviews from the analysis.

After screening, a total of 94 articles meeting the inclusion criteria have been identified, including five articles focusing highly on patients with myopia and six articles focusing on patients with diabetics (see online supplemental file S2 for specific flow diagram). Readers who possess knowledge in this field and are aware of studies that have not yet been included are encouraged to contact us.

Data management

The data collection protocol employed at the group consensus meeting will be continuously adjusted by two key members based on various studies. When possible, we will contact the corresponding author of the study via email to acquire, if any, original data that were not mentioned in their article. The data collection process will be divided between two members with respective roles. All data will be sent to the data manager in the most convenient format and stored in an Access database. The data will be uniformly organised by two team members responsible for this task. All information will be securely maintained and treated as strictly confidential. The data will not be allowed to be used in any publication without permission from original trialists.

Data collection process

IPD can be obtained from data-sharing databases, such as Yale University Open Data Access, Clinical Study Data Request and Vivli (https://vivli.org/).15 In cases where data are not available in data-sharing databases, we will directly obtain data from the authors of the respective study.

Once it is established that the study meets the eligibility criteria, the authors of the study will be contacted via a secure university-registered email address. To collect as much individual patient (raw) data as possible from the included studies, we will use a multistep contact protocol that has been proven effective in previous studies.16 The steps for obtaining raw data from the authors of the included studies are summarised in figure 1. We will obtain the contact information of the corresponding author by (a) gathering it from relevant publications, (b) conducting a web search and (c) reaching other researchers. An email will be sent to the corresponding author, clearly stating the study’s objectives and providing clinical and methodological justifications for IPD. We will also convey our intentions to the authors through in-person meetings, as this may increase their willingness to share raw data.17 The corresponding author is given 9 weeks to respond to our email. If no response is received after 3 weeks, a second email will be sent. Similarly, if there is no response after another 3 weeks, a third email will be sent. If the corresponding author does not respond within these 9 weeks, we will repeat the aforementioned steps with other authors in the following order: first, last, second, third, fourth, etc. When the above steps are unsuccessful, the corresponding author will be contacted via letter and telephone to obtain the raw data. Letters will be sent at 3-week intervals. We will contact other authors in the same manner only when the corresponding author cannot be reached. Finally, if there is still no response, we will attempt to establish communication with colleagues who may be acquainted with the authors. Study data will be considered unavailable only if all the methods have failed or if authors indicate that the raw data have not been retained or that they refuse to share these data.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Flow chart showing the process of obtaining the raw data from the authors of the included studies.

Information about the included studies

For the included studies, we will collect basic information (author, publication date, country, number of eyes/patient per study and special populations), intervention description (diameter of the capsulorhexis, area of anterior capsule opening, anterior capsule cover, centration and circularity of capsulorhexis), outcome description (refraction, capsule shrinkage, stability of IOL intraocular pressure) and follow-up time, as summarised in box 1.

Baseline characteristics of the included studies

• Participant-level data items: we request that investigators provide the following original data, to the extent possible, for the subjects in the included studies: age, sex, axial length, anterior chamber depth (ACD), grade of the lens nucleus or lens opacity, capsular bag diameter, IOL power, predicted refraction and preoperative visual acuity. Data collected at the follow-up visit include the area of capsulorhexis and/or anterior capsule opening, circularity of capsulotomy or CCC, visual acuity, intraocular pressure, corneal oedema, postoperative aqueous flare intensity, corneal endothelial cell density (ECD), central corneal thickness (CCT), ACO, PCO, surgically induced astigmatism and the rate of anterior capsule contraction.

• Surgical-level data: the surgical time and cumulative ultrasound energy used during the operation are recorded.

Information to be explained

• Capsule shrinkage: the percentage reduction in anterior capsule opening at follow-up is calculated as follows: (previous anterior capsule opening–anterior capsule opening at this follow-up) × 100/anterior capsule opening at the last follow-up.

• Types of IOLs: capsule contraction is affected by IOL optic material, haptic design, and size.10

• Anterior chamber depth: the distance between the anterior surface of the IOL and central corneal endothelium is known as ACD.

• Circularity: 4π (area/perimeter²).

• Calculation of the capsulorhexis diameter.

Actual diameter = (maximum diameter+minimum diameter)/2

Relative diameter=actual diameter/expected diameter

• Stability of the IOL:

Tilt: an IOL tilt of 7° or more is considered clinically significant.

Decentration: clinically significant decentration of the IOL is defined as 0.4 mm or greater.

Rotation (suitable for toric IOL): this parameter refers to the angle between the IOL axial position at the postoperative follow-up and the expected position of the IOL axis before the operation. An angle greater than 1° indicates axial rotation.

Anteroposterior IOL shift

• Visual acuity: for analysis, all measurements of visual acuity are transformed into logMAR units.

Risk of bias in individual studies

The standard risk of bias assessments is not applicable to these studies because we will use data from clinical trials for purposes unrelated to their original research topic. Previous studies have identified the refined Quality in Prognosis Studies (QUIPS) tool as suitable for assessing the risk of bias in studies of prognostic factors.18 Six domains are evaluated: study participation, study attrition, prognostic factor measurement, outcome measurement, study confounding, and statistical analysis and reporting.19 The outcomes are classified as yes, partial, no or uncertain, based on the prompting items within each domain. The answers will be then summed, and each domain will be rated as having a high, moderate or low risk of bias.

Two independent appraisers will use the QUIPS tool to assess the risk of bias in the included studies. The entire research team will convene a consensus meeting to resolve any disagreements among appraisers.

A study will be classified as having a low risk of bias when all six domains are rated as low or moderate, with at least four items rated as low, including the outcome measurement domain. Studies that receive high scores in two or more domains will be classified as having a high risk of bias. The remaining studies will receive a moderate rating.20

Confidence in cumulative evidence

The overall certainty of the evidence will be assessed by two independent reviewers using a modified Grading of Recommendation, Assessment, Development and Evaluation (GRADE) approach. This modified approach is specifically designed to evaluate prognostic evidence, including study design, risk of bias, inconsistency, imprecision, indirectness and publication bias. Effect sizes and trends will also be considered. In the event of any disagreement, consensus meetings will be held with the group members to resolve them.

We will take the following steps to implement the modified GRADE framework:

• Determine the type of study: we will first classify the outcomes based on the study design (eg, randomised controlled trials and observational studies). This classification will assist us in establishing a baseline for assessing the quality of evidence.

• Assess the risk of bias: we will use standardised tools, such as the Cochrane Bias Risk Tool, to evaluate the risk of bias for each study. Studies identified as having a high risk of bias will be considered for downgrading.

• Assessment of consistency: we will compare the consistency of results across studies. In cases where significant differences are observed, we will consider downgrading the quality of evidence.

• Indirectness of evidence: we will conduct an indirect assessment of whether the results apply to our study population or the intervention. Downgrading will be considered if the intervention or population in the study is deemed distant from our own.

• Precision assessment: we will evaluate the precision of the results, including the width of the confidence intervals. If the confidence intervals are wide, indicating insufficient precision, we will consider downgrading.

• Publication bias: we will check for signs of publication bias, for example, by employing a funnel plot. Should evidence of bias be identified, a downgrade will be considered.

• Upgrade or downgrade: on completion of the assessments, we will synthesise all factors to determine whether to upgrade or downgrade the quality of evidence. For instance, if a particular study demonstrates strong performance in terms of risk of bias and consistency but exhibits issues with precision, we may opt to maintain the current quality of evidence.

• Transparent recording: we will meticulously document each step of the evaluation process and the final grading of evidence quality to ensure transparency and reproducibility.

Publication or data-sharing bias

To assess the concern regarding publication or data-sharing bias, we will implement several strategies:

• Literature review: we will undertake an extensive literature review to identify all pertinent studies, including smaller ones that might be excluded from IPD analyses.

• Tool: for analyses involving 10 or more trials, we assess the potential publication bias by examining asymmetry in funnel plots using Egger’s test. Funnel plots will also be generated to visually inspect for potential bias in smaller studies.

• Sensitivity analysis: we will perform sensitivity analyses to evaluate the impact of including or excluding smaller studies on the overall results.

• Meta-biases: In 2016, Smith et al compared IPD-MA outcomes with aggregate data meta-analysis outcomes and reported inconsistent results.21 To assess potential bias in data availability, we will perform t-tests for continuous variables and χ2 tests for categorical variables to compare characteristics across studies that received IPD and those that did not. The effect size information obtained from the publications will be used to calculate Cohen’s d effect sizes. To compare effect sizes between studies with and without IPD, the effect sizes are pooled and analysed across studies using the traditional ‘subgroup analysis’ approach of meta-analysis.

Strategy for data synthesis

A one-stage IPD-MA using multilevel regression model22 will be chosen based on previous experience. The effect measures will be converted to a standardised scale (z-scores) to facilitate visual comparisons. Notably, most clinical trials included in the IPD-MA did not assess the primary endpoint. A multilevel regression model will be established based on all available data, with the size of the capsulorhexis, preoperative visual acuity and other prognostic indicators as independent variables, and postoperative visual acuity as a dependent variable. A similar multilevel regression model will be used to examine the impact of capsulorhexis size on the rotational stability of the IOL or other outcome variables. Multilevel linear regression models will be used for continuous outcomes, while multilevel logistic regression models will be used for categorical outcomes. To evaluate their moderating effect on treatment outcomes, the model will incorporate the interaction between the intervention group and the moderators of interest. If heterogeneity is permitted, raw data will be combined into a single set. To address data clustering, a study identification variable will be added as a random effect variable. When quantitative synthesis is not feasible, we will perform a narrative analysis.

Another analysis will investigate the short-term effects of capsulorhexis size on IOL rotation of less than 15° within 1-week period.

Sensitivity analysis

To conduct a sensitivity analysis, studies with an overall high risk of bias will be excluded. Further sensitivity analysis will use interaction terms to explore whether pairs of predictors show nonlinear effects on the primary outcome.

Subgroup analysis

Prespecified subgroup analysis will be conducted based on:

1. Presence or absence of high myopic eyes

2. Patients with or without DM

3. Types of IOL

4. Types of capsulorhexis size

5. Types of study

If sufficient data are available, the ocular axial length and the type of astigmatism will also be considered as factors in the grouping. By collecting individual patient data, it is possible to estimate potential differences between these subgroups. Analysing the reasons for these differences is more effective than relying solely on individual trial reports. The analyses will be conducted using Review Manager C.5.3.