The Correction of Myopia Evaluation Trial (COMET):: Design and General Baseline Characteristics

INTRODUCTION

Myopia is an important public health problem that entails substantial societal and personal costs. It is highly prevalent in our society, affecting at least 25% of the adult population in the United States [1] and is even more common in Asian countries, affecting up to 84% of adolescents [2]. Furthermore, its prevalence may be increasing over time as suggested by some studies in various countries including Singapore, Australia, and the United States [3][4][5][6]. Estimates of costs in the United States for refractive eye examinations and corrective aids, including spectacles and contact lenses, range from $2.5 to 4.5 billion each year [7]. Due to the significance of myopia as a global public health concern, it was chosen as a priority for Vision 2020, the World Health Organization's global initiative for the elimination of avoidable blindness by the year 2020 [8].

At present, the mechanisms involved in the etiology of myopia are unclear and means of prevention are unknown. Myopia progression is irreversible and there is no cure. Refractive surgeries for treatment of myopia are both costly and unsuitable for children's eyes and do not change axial elongation, which is the source of most myopia [9]. High myopia is a predisposing factor for retinal detachment, myopic retinopathy, and glaucoma, thus contributing to loss of vision and blindness in both developed and developing countries [10]. In a population-based study conducted in the Netherlands, myopic degeneration was the cause of 6% of low vision and 6% of blindness [11].

Findings from studies in both animals and humans have suggested that retinal defocus produced by inaccurate accommodation may be a stimulus for increased axial elongation leading to myopia. The eyes of animals exposed to continuous retinal defocus become myopic [12]. Additionally, children with progressing myopia were found to underaccommodate more than nonmyopic children [13]. These observations suggest that optical corrections, such as bifocal or progressive addition lenses that reduce retinal blur, might slow myopia progression in children.

Prolonged near work (e.g., reading) is associated with increased prevalence of myopia [14]. Several older studies have evaluated the efficacy of bifocals based on the hypothesis that negating the prolonged chronic accommodation associated with near work would reduce myopia progression. Most studies investigating this hypothesis were either retrospective analyses or clinic-based prospective studies without controls and had other methodologic limitations, including incomplete follow-up and small sample sizes. Results have been inconsistent across studies. Two clinical trials evaluating the effect of bifocals did not show any difference in progression between the bifocal and single vision groups [15,16]. Further analysis of one of these studies, the Houston Myopia Control Study [17], as well as a small prospective study of 32 children suggested that progression of myopia in children with near point esophoria might be slowed with bifocal wear [18].

Progressive addition lenses (PALs) have been prescribed for myopic chil-dren based on the theory that they provide clear images at all distances. A small study of 68 myopic children conducted in Hong Kong found that children who wore PALs for 2 years showed significantly less myopic progression and less axial length increase than children who wore single vision lenses (SVLs) [19]. Results of this study also showed that a ϩ 2.00 diopter (D) addition lens was more effective than a ϩ 1.50 D add, thus suggesting that PALs (particularly with a ϩ 2.00 D add) might be a potentially useful treatment for reducing the progression of myopia in school-age children. The Correction of Myopia Evaluation Trial (COMET) is a randomized, double-masked, controlled, multicenter clinical trial, designed to evaluate the effect of PALs versus SVLs on the progression of myopia in children. The rationale for COMET is based on the convergence of the three distinct lines of research described above: (1) the effect of defocus in animal models of myopia; (2) the link between reduced accommodation and the development and progression of myopia in children; and 3results of the small study by Leung and Brown [19].

This report describes the aims and design of COMET and presents general baseline characteristics of the 469 children enrolled in the study. Additional baseline data comparing measures of refraction and ocular components and describing the residual accommodation measures obtained with the cycloplegic protocol are presented elsewhere [20,21].

STUDY AIMS

The primary aim of COMET is to evaluate whether progressive addition lenses (Varilux Comfort with a ϩ 2.00 D addition) slow the rate of progression of juvenile-onset myopia when compared to conventional SVLs, as measured by cycloplegic autorefraction. Axial length, measured by A-scan ultrasonography, is an additional outcome measure. This aim will be achieved by conducting a randomized clinical trial comparing myopia progression in children treated with PALs versus children treated with SVLs. The comparison will allow quantification of the effect of PALs on progression during a minimum of 3 years of follow-up. The null hypothesis being tested is that no differences in progression will occur between children randomized to PALs versus SVLs.

The secondary aims are:

1. To describe the natural history of juvenile-onset myopia in a group of children receiving conventional treatment (i.e., SVLs). This aim will be achieved by conducting analyses limited to the children assigned to wearing SVLs and will provide longitudinal data on changes in refractive error, ocular components, accommodation, and phoria in this group. 2. To explore factors such as age, gender, and initial refraction that may influence the progression of myopia. This aim will be achieved by multiple regression analyses of predictor variables for progression in both study groups.

STUDY DESIGN

Study Organization

COMET represents a collaborative effort involving: (1) a study chair, at the New England College of Optometry, Boston, Massachusetts; (2) a coordinating center, at the Stony Brook University Health Sciences Health Center, Stony Brook, New York; (3) four clinical centers at the New England College of Optometry (NEWENCO), Boston, Massachusetts; the School of Optometry, University of Alabama at Birmingham (UAB), Birmingham, Alabama; the Pennsylvania College of Optometry (PCO), Philadelphia, Pennsylvania; and the College of Optometry, University of Houston (UH), Houston, Texas; and (4) the National Eye Institute (NEI), Bethesda, Maryland, which supports the study. A list of the study investigators and committee members at baseline is included in the appendix.

Three committees involving study investigators provide leadership to the study and review its progress on a continuing basis (executive, steering, and full investigator). The executive committee, which oversees and directs all aspects of the study, includes the study chair, the director of the coordinating center, a clinical center principal investigator, and a representative of the NEI. The steering committee (consisting of the study chair, consultant to the chair, three representatives from the coordinating center, the four clinical center principal investigators, and a representative of the NEI) provides input to the protocol, reviews procedures, resolves technical issues, and addresses publicationrelated issues on a monthly basis. The full investigator group, which includes all of the personnel at all of the study centers, conducts the day-to-day operations of the study, provides input to the protocol as needed, and meets once a year. An independent data and safety monitoring committee (DSMC), composed of persons with expertise in clinical trials methodology, biostatistics, myopia research, optometry, ophthalmology, and medical ethics, is responsible for monitoring all aspects of the trial. Coordinating center personnel and the study chair, who is not involved with data collection and is unmasked, are the only representatives of the study who participate in the DSMC meetings. This committee, which is not masked regarding treatment assignment, is the only group provided with evidence of treatment effects during the course of the study.

Eligibility Criteria

The eligibility criteria were selected for the following reasons: to enroll children with moderate levels of myopia who are at an age most likely to experience myopic progression, to maximize the efficiency of recruitment and potential for adherence, to ensure a high degree of child safety, and to enhance the generalizability of the results. Eligibility was evaluated at a baseline visit at a clinical center and confirmed at the coordinating center prior to randomization.

Inclusion

Children eligible for participation were 6-11 years of age (inclusive) with spherical equivalent refractive error between Ϫ 1.25 D and Ϫ 4.50 D in both eyes, as measured by cycloplegic autorefraction; astigmatism Յ 1.50 D; and no anisometropia (i.e., difference in spherical equivalent between the eyes Ͻ 1.00 D). In addition, they had normal visual acuity with best subjective correction (i.e., LogMAR Յ 0.2, Snellen equivalent 20/32); demonstrated no strabismus by cover test for far (4.0 m) and/or near (0.33 m) fixation (with the best subjec-tive refractive error and with a ϩ 2.00 D lens over the best subjective refractive correction); and were willing to refrain from wearing contact lenses for the duration of the study.

Exclusion

Criteria for exclusion were: strabismus detected with cover test; any ocular, systemic, or neurodevelopmental conditions that could influence refractive development; chronic medication use that might affect myopia progression or visual acuity; birthweight Ͻ 1250 g; current or prior use of bifocals or PALs; or possible problems with adherence to the protocol or follow-up for a minimum of 3 years.

Sample Size Considerations

A sample size of 450 children (113 per clinical center) was selected for this study based on the following considerations:

• An expected 33% reduction in mean 3-year cumulative myopia progression in the PAL versus the SVL group as measured by the magnitude of change in spherical equivalent cycloplegic autorefraction relative to baseline (a continuous measure). • An overall standard deviation of 1.10 D Ϫ 1.35 D in the cumulative 3-year follow-up measurements of refractive error change. The range for the standard deviation estimate was based on a weighted pooled estimate of the coefficient of variation (CV) of 89% (CV ϭ SD/mean), derived from the sample sizes of studies reported by Gwiazda et al. [22], Goss and Cox [23], Goss and Winkler [24], and Parssinen and Lyyra [25]. • A mean cumulative progression of 1.50 D for the SVL group. • A two-tailed 1% ␣ -level. • A minimum of 84% power. • A stratification of the randomization scheme by center.

• Univariate testing between the two groups (i.e., differences in mean values).

• A maximum of 20% attrition (i.e., the required sample size was increased by 20%).

Recruitment Methods

COMET children were identified by active recruitment efforts primarily in optometry clinics (36%), school screenings (24%), and letters to parents of children wearing glasses (13%). Other recruitment methods included advertising in local newspapers and among the university communities (10%), personal referrals and referrals from other COMET participants (10%), and referrals from other eye-care practitioners (7%). The recruitment sources varied by center, with two centers (NEWENCO and PCO) recruiting over half of their children from school screenings and the other two (UAB and UH) emphasizing optometry clinic referrals and letters to parents. Potentially eligible children were referred to a clinical center for a baseline visit during which they were evaluated for eligibility. The first COMET child was randomized in September 1997, and recruitment ended in September 1998.

Treatment Groups

The two treatment groups for COMET are defined by two different types of eyeglasses, PALs and SVLs, both of which are commonly used. PALs are multifocal lenses, also known as "no-line bifocals," which have a gradual and progressive change toward less negative or more positive power from the distance portion to the near portion of the lens and are typically used to correct for presbyopia. SVLs have the same focal power throughout the lens area and are the conventional treatment for myopia.

Randomization

Children were randomized to one of the two treatment assignments (i.e., PAL or SVL). The randomization scheme was stratified by clinical center using a random permuted block design with a predetermined block size for each center to ensure sequential balance of the distribution of child characteristics and potential prognostic factors among the two study groups such as age, age of myopia onset, and baseline refractive error. Randomization assignments were allocated centrally by the coordinating center after eligibility criteria were verified. A child was considered to be enrolled in COMET once the randomization assignment and study number were issued and the child received the assigned lenses.

Informed Consent and Child Assent

Two sets of consent/assent were obtained: one for participation in the baseline visit and the second for participation in the trial. Prior to the baseline examination, informed consent was obtained from the parent/guardian regarding participation in the examination, including an evaluation for study eligibility and possible participation in the trial. Assent was also obtained from the child. Those children remaining eligible at the end of the baseline visit received a consent/assent form describing participation in the follow-up phase of COMET for review prior to the randomization visit. At that visit, a final consent and assent, indicating agreement to participate in COMET for a minimum of 3 years, were obtained. In addition, a "COMET commitment" describing the families' responsibility to the study was reviewed with the parents/guardians and children. This involved the parents/ guardians agreeing to: (1) accept a random lens assignment, (2) have their child wear only COMET glasses and not contact lenses for at least 3 years, and (3) call the clinic coordinator with any problems with the COMET glasses and questions about the study. The children agreed to wear their COMET glasses during waking hours.

Study Visits

The COMET protocol includes two initial visits (baseline and randomization), a minimum of six follow-up visits (semiannual and annual), and problem visits as needed (Table 1). A description of these visits follows.

Initial Visits (Baseline and Randomization)

Baseline visits were held for all potentially eligible children. During this initial visit to a clinical center, the study was explained, eligibility was evaluated, and baseline data were collected. (A description of the procedures used to collect these data is provided in the section on data collection procedures below.) If exclusion criteria were identified at any time during the baseline visit, the visit continued as a routine, standard eye exam and these children were not invited to enroll in the study. At the end of the visit, children who met eligibility criteria and were willing to participate in COMET selected eyeglass frames, were measured and fit for study glasses, and were scheduled for the randomization visit. Study opticians measured monocular interpupillary distance and segment height (set 4 mm above the midpupil) to determine the fit for PALs in all children. This fit was selected to encourage the children to use the additional portion of the lens, since unlike presbyopic adults, children can accommodate [26]. Single vision sports glasses meeting American National Standards Institute (ANSI) standards were also offered to both treatment groups and their use encouraged for all children enrolled in COMET. This eye- wear, to be used while the children participate in sports activities, protects against sports-related ocular injuries. Baseline forms were faxed to the coordinating center for independent confirmation of eligibility and the randomization assignment. The lens assignment was faxed to an unmasked investigator (clinic coordinator or optician) at the appropriate clinical center, who then ordered the study glasses based on the measurements and frame selection obtained during the baseline visit.

The randomization visit was held after the study group assignment and study identification number were issued by the coordinating center. During that visit, accommodation and phoria by Maddox rod were measured, a quality-of-life questionnaire was administered, the study glasses were dispensed, and the child was officially enrolled in COMET.

Only seven children who received randomization assignments decided not to participate and did not receive their assigned lenses. Their decision not to participate was not based on knowledge of lens assignment. They were not considered as enrolled in the study and their randomization assignments were not reissued. This small number did not impact on the randomization process.

Follow-up Visits

Follow-up visits are held every 6 months for a minimum of 3 years. Outcome data (i.e., cycloplegic autorefraction and axial length) as well as accommodation and phoria measures are collected at annual visits only. Additional data collected at the annual visits are also obtained at the semiannual visits and include an assessment of adherence to the use of the COMET glasses, evaluation for the need for a prescription change (described in detail in the section on data collection procedures below), and monitoring of child safety.

Additional visits, designated as problem visits, may take place between regular COMET visits to address problems with frames, lenses, visual symptoms, ocular health, or any child safety concerns. To evaluate possible lens-induced phoria changes, a 1-month follow-up visit was held for the first 150 children. This visit was discontinued, following a recommendation by the DSMC, once it was determined that no child had demonstrated induced strabismus or other visual problems from either type of spectacle lens.

Data Collection Procedures

Study procedures vary according to the type of study visit (Table 1) and are all conducted according to a standard protocol. The main study measures, cycloplegic autorefraction and axial length, are collected on both eyes at baseline and annual visits. For cycloplegic autorefraction, five consecutive reliable measurements of sphere, cylinder, and axis are taken using the Nidek ARK 700A autorefractor, approximately 30 minutes following dilation using two drops of 1% tropicamide, spaced 4-6 minutes apart. Only measurements with a reliability rating of 7, 8, or 9 provided by the Nidek are acceptable for study purposes. Three to five reliable axial length measures are taken, after dilation, using the Sonamed A2500 ultrasound by either slit-lamp or handheld methods.

Accommodation and phoria by Maddox rod were also taken at the randomization visit and are taken at the annual visits using the Canon R-1 Infrared Au-torefractor with a Risley prism attachment. Residual accommodation was measured using the Canon R-1 at the baseline visit only to monitor the extent of cycloplegia [21]. Iris color was also assessed at the baseline visit according to a standard protocol [27]. The baseline and annual visits also include external observations, pre-and postdilation slit-lamp evaluation, and an ocular fundus evaluation using the binocular indirect ophthalmoscope.

Additional data collected at the initial and all follow-up visits include: ocular and medical history; habitual acuity; noncycloplegic autorefraction and keratometry using the Nidek 700A autorefractor/autokeratometer; subjective refraction using a Marco Total Refracting System; best-corrected LogMAR visual acuity using a modified Early Treatment Diabetic Retinopathy Study (ET-DRS) protocol; evaluation for prescription changes; cover test (cover-uncover and alternating cover) with prism neutralization of eye movement to measure phorias (horizontal and vertical) through the distance prescription at far (6 m) and near (33 cm) and through the near ( ϩ 2.00 D) addition lens; optician measurements; questions on visual symptoms and adherence to using COMET glasses; and, evaluation of the need for a prescription change.

Prescription changes are made when the difference in subjective refraction between the current and the most recent prescription (in at least one eye) is more myopic by Ն 0.50 D spherical equivalent. Smaller prescription changes can be made if clinically indicated. All of these changes are documented and tracked. The power of the addition in the PALs ( ϩ 2.00 D) remains constant throughout the study. Every time there is a COMET prescription change, the COMET single vision sports glasses are also changed to match the distance prescription in the child's study eyeglasses. Data on self-perceptions are collected at the randomization visit and at the 18-and 36-month follow-up visits using a questionnaire developed for use in children by Harter [28].

Clinical Center Personnel

COMET is designed as a double-masked trial in which neither the families nor the optometrists responsible for assessing study outcomes are aware of the lens assignment. Staffing of study personnel is similar at each center and includes masked and unmasked investigators. Two optometrists (one is the principal investigator), masked to lens assignments, obtain the study measures. The other investigators are unmasked (i.e., know which intervention the child receives) and do not obtain outcome measures. These positions are the study optician (responsible for dispensing and fitting the glasses), the clinic coordinator, and a consulting optometrist who sees all children for any vision problems that might lead the COMET optometrists to become unmasked. A backup coordinator and optician are also part of the COMET team at each center.

Masking

A number of steps have been taken to preserve and monitor the masking, including:

1. inclusion of a consulting optometrist to handle any issues regarding visual symptoms; 2. identifying children by a number unrelated to treatment assignment; 3. providing verbal and printed instructions to the parents and children explaining that neither the family nor the COMET staff (other than the study optician, clinic coordinator, and consulting optometrist) are aware of the lens assignment, and emphasizing the importance of not discussing any issues related to the study glasses with the COMET optometrists and not wearing study glasses in their presence; 4. fitting glasses for all children as if each child had been assigned to PALs; 5. standardizing data collection forms and examination protocols for all children regardless of treatment assignment; 6. giving the COMET glasses to the COMET optician or clinic coordinator for the duration of the study visit; 7. requesting that each member of the COMET staff document any observed unmasking of the parents and children at each COMET visit; and 8. monitoring whether the child or parent indicated knowledge of treatment assignment at each study visit.

Outcome Measures

Primary Outcome-Change in Refractive Error

The primary outcome for COMET is progression of myopia, defined as the magnitude of change in spherical equivalent refractive error relative to baseline (a continuous measure). Cycloplegic autorefraction was selected as the measure of refractive error because of its accuracy, reliability, and objectivity, thus allowing for standardization of measurements over time within and across centers [29,30].

The spherical equivalent is calculated for each of the five autorefraction measurements per eye, and the mean of the five spherical equivalent measures is then computed and used as the measure of refractive error for each eye. To analyze the progression of myopia, refractive error will be expressed as the sum of three components, M (mean spherical equivalent as described above); J 0 (dioptric power of a Jackson Cross Cylinder at axis 0 Њ ), and J 45 (dioptric power of a Jackson Cross Cylinder at axis 45 Њ ), as determined by the Fourier decomposition (rectangular form) method [31]. Progression is measured and compared between the PAL and SVL groups.

The primary analysis of progression will be child-based. To define this primary outcome measure, the following strategy is followed. The selection of the refractive error measurement for each child will be made by evaluating the correlation of the myopic change between eyes at each annual visit, independent of study group. If the correlation coefficient is Ͼ 0.85 and the mean difference between eyes is not statistically significantly different from 0 (i.e., 95% confidence interval of the difference contains 0), then the eyes are judged to be highly correlated with no appreciable difference in myopia progression between them. In this case, the mean of the two eyes will be used as a summary of the refractive error. Otherwise, the worse eye for each child (eye with the more myopic change) will be used. In this case, the multivariate analysis will treat myopia change in the fellow eye as a covariate.

Secondary Outcome-Axial Length

The progression of myopia is accompanied by anatomical changes in the eye, particularly an increase in axial length. Measures of axial length have been shown to be correlated with spherical equivalent refractive error in animal [9,32] and human eyes [33,34].

The secondary outcome for COMET is axial length. A-scan ultrasonography is used to measure the axial components of both eyes. Progression in the axial length measurements will be defined as the magnitude of the change in axial length during follow-up relative to baseline. At the randomization visit and each annual visit, five measures are taken on each eye using the Sonomed A-scan. Additional measures are taken, if necessary, to achieve a standard deviation Յ 0.1 mm for the five measures. The axial length measurement is based on the mean of these five values if the desired precision (i.e., Յ0.1 mm) is achieved. Otherwise, one to two outlying values as determined by their difference from the mean are eliminated and the mean is based on the remaining three or four values. Prior to the beginning of data collection for COMET, study examiners demonstrated good consistency of axial length measurements with those of a gold-standard examiner [30], suggesting that these measures are reliable.

Additional Measures

The study will also evaluate changes in other ocular components (i.e., lens thickness, anterior chamber depth, and vitreous chamber depth), accommodation and phoria by Maddox rod, and corneal curvature (based on keratometry measured with the autorefractor). These measurements will assist in identifying the anatomical basis of any observed effects of PALs.

Quality Assurance

Data quality and integrity have been addressed in COMET by the following measures. All of the data collection protocols are standardized through study documentation (the manual of procedures), uniform study protocols and forms, and uniform criteria for patient recruitment, including independent confirmation of eligibility at the coordinating center. Study personnel are trained and certified before collecting study data. Annual site visits are conducted at each clinical center. These visits include a data audit based on a representative sample of study visits. Regular communications among the study investigators, including monthly steering committee and clinic coordinator conference calls, allow for timely resolution of any outstanding issues.

The coordinating center monitors data quality and adherence to protocol on a routine basis. Data processing is centralized and concurrent. Data editing involves a three-step process that takes place: (1) as part of the data entry process, (2) during a separate editing stage, and (3) during preparation for data analysis. Two independent, certified data entry persons enter data into two separate files, which are compared by an adjudication program. Discrepant data items are adjudicated by a third person, who is also certified. Once the adjudication process is completed, a comprehensive edit system reviews all entered data forms using a sequence of logic and range checks to identify missing, invalid, inconsistent, or questionable entries. These data issues are resolved by edit statements to each clinical center. All data files are backed up routinely, with monthly backups archived and stored off site.

A number of additional measures also have been included as part of COMET to ensure child safety and encourage adherence to use of the study glasses and to the visit schedule. These include: offering and encouraging the use of sports glasses, providing backup glasses and lens cleaning kits, routine monitoring of child retention, identifying opticians and optometrists outside of the clinical center area to conduct interim visits for children who have moved away, providing transportation to return to the clinical center for regularly scheduled visits, routine questioning of children and parents about visual symptoms and use of glasses, a problem visit protocol that allows for prompt response to a child's problems and the involvement of a consulting optometrist when needed, and other activities, such as sending newsletters to participants.

Study Monitoring

The study is reviewed regularly by the DSMC, which monitors any significant child safety concerns and overall study performance. Once a year, the DSMC meets to review interim reports to determine possible differences between study groups that would warrant stopping the study and making the results known to families and the scientific community. These reports include extensive data on all aspects of recruitment, retention, protocol adherence, follow-up, reliability of the study measures (cycloplegic autorefraction, axial length, and Canon R-1 accommodation and phoria measures), and data quality. Data on protocol adherence, retention, and data quality are also provided routinely to the steering committee to address protocol-related issues on an ongoing basis.

Interim Analysis and Stopping Guidelines

Interim analyses are being conducted to test for differences in cumulative progression rates between the two study groups as well as to monitor possible adverse effects. The timing for the interim analyses depends on reaching a specific ratio of cumulative progressions (i.e., the number of children with at least 1.50 D change in refractive error relative to the number expected to occur by the end of the trial). For the purpose of the interim analyses, the outcome is reached when one eye has progressed by 1.50 D or more relative to baseline. The Lan and DeMets [35] procedure is used, which allows for interim testing without prespecification of the number of times or calendar time(s) to conduct these analyses. However, the rate at which the overall significance level (␣) is spent over the course of the trial (i.e., defining a spending function for the overall significance level) is prespecified. The overall significance level for this study has been set at 5%, and the spending function is determined based on the proportion of children with progression of myopia of Ն1.50 D during the follow-up period.

The DSMC is regularly provided with the results of the interim analyses to monitor the differences between the study arms. Their recommendations based on statistical, clinical, and other relevant child safety issues are sent to the director of the NEI who, based on these recommendations, decides whether the results warrant termination of the study. If the study is terminated early, the NEI director will be responsible for the public dissemination of infor-mation, the study chair and coordinating center will inform the study investigators, and the coordinating center will implement the termination and phaseout of the trial.

Statistical Analyses

Baseline Analyses

Baseline characteristics were summarized using standard methods including percent frequency distributions for categorical variables and means, standard deviations, and medians for continuous measures. The five autorefraction measures for each eye for each child were summarized using both the median values provided by the Nidek (used for eligibility) and mean values calculated based on an average of each of the five spherical equivalent values (used to evaluate progression). The axial length is also based on a mean of the three to five measurements for each eye and child.

During the enrollment phase of the study, the comparability of study groups at baseline was evaluated by the Fisher's exact test for categorical variables and by the two-sample t test for continuous variables to monitor whether balance of key study variables was achieved between the groups. Imbalances of potentially confounding variables that are identified between the two groups will be addressed by including them (poststratification) in the final multivariate analysis model(s).

To monitor the performance of the randomization scheme during the enrollment phase of the study, the distribution of baseline characteristics was compared between treatment groups using Fisher's exact test for categorical variables and the two-sample t test or Wilcoxon rank-sum test for continuous variables. If imbalances of potentially confounding variables are identified between the two groups, they will be addressed by including them (poststratification) in the final multivariate analysis model(s). The factors to be included in the multivariate model will consider both the parsimony of the model and the interpretability of the findings, as well as the potential loss of statistical power. Therefore, the minimum number of important factors will be included.

Progression Analyses

Follow-up data will be analyzed using an intent-to-treat principle according to the child's original study group assignment (i.e., SVL or PAL). The primary analysis for progression in COMET will be child-based and will evaluate the magnitude of myopic change in spherical equivalent cycloplegic autorefraction between follow-up and baseline (i.e., the difference in refractive error measured at the two visits). An additional analysis will evaluate the rate of myopic change by determining the slope for each study group based on the measurements from baseline and each of the follow-up examinations (minimum of three) and comparing the slopes between the two groups. Losses to follow-up will be monitored and compared between study groups. If no informative censoring occurs (i.e., losses to follow-up are independent of study group), then traditional statistical methods will be employed using progression information through the latest follow-up visit for each child. However, if losses to followup are found to be associated with study group, two analyses, one with and one without censored data, will be conducted and compared. If these two anal-yses yield different results, then the censored observations will be omitted. The distribution of prognostic factors will be reevaluated and compared between study groups to address any potential imbalances using multivariate analyses.

Additional eye-specific and group-specific analyses also will be conducted. Eye-specific analyses (including both eyes of each child) will be adjusted for the correlation of measures between eyes [36]. A multivariate intraclass correlation model will also be employed assuming a nested mixed effects analysis of variance structure as described by Rosner [37]. This model will essentially adjust for the (within-child) correlation between eyes by regarding the study group effect as fixed and the effects of children within the groups and eyes within the child as random. These random effects will rely on robust assumptions regarding the normality of their parameters with zero means and variances that can be estimated from the data. Further analyses will evaluate progression as a discrete event (i.e., the proportion of children with a change in myopia of at least 1.50 D spherical equivalent cycloplegic autorefraction, relative to baseline, in at least one eye will be compared between the treatment groups).

The distribution of myopia progression (i.e., change in myopia) first will be tested for normality in univariate analyses. If the distributions are normally distributed, then standard parametric tests such as the t test for independent samples will be used for comparisons. Otherwise, nonparametric approaches (e.g., Wilcoxon rank-sum test) will be used. Comparisons also may be stratified for further analyses. The primary multivariate analyses, which are child-based, will use a simple multiple regression model including the most important covariates (e.g., age, baseline refractive error) [38,39]. Additional analyses will use a mixed effects model to use progression information from both eyes and adjust for the correlation between eyes. The assumptions for normality of the random effect parameters are based on robust considerations as outlined by Rosner [37].

Additional analyses of myopia progression as a binary discrete event will use life-table methods [40] to estimate progression rates while accounting for variable follow-up time among children. These analyses will include treatment group as a covariate as well as factors such as age of myopia onset, ethnicity, accommodation, phoria, and ocular components.

Change in axial length, as measured by A-scan ultrasonography, also will be evaluated as a continuous variable, measured in mm, using a similar approach to that described above for refractive error. Univariate analyses will be performed first, to evaluate treatment effect on mean change in axial length over the 3-year period. These analyses will use the two independent sample t tests as well as the Wilcoxon rank-sum test. Multivariate regression analyses will assess the association between change in axial length and progression of myopia, while adjusting for treatment effects. Axial length will be regressed on refractive error in two separate analyses: once defined as a discrete (yes/no) outcome and then as a continuous outcome.

Predictor variables for progression of myopia such as age, gender, ethnicity, accommodation, phoria level, and initial refraction will be explored using multiple regression analyses. The correlation between myopia progression and changes in ocular components, including corneal curvature, lens thickness, anterior chamber depth, and vitreous chamber depth, will also be evaluated.

RESULTS

Recruitment was completed within 1 year, between September 1997 and September 1998, and follow-up is ongoing. Four hundred and sixty-nine children were enrolled during the recruitment period, thus exceeding the recruitment goal of 450. Baseline characteristics were evaluated by child (e.g., age, gender, ethnicity) and by eye (e.g., refractive error, axial length). General descriptive characteristics are presented in Table 2 and Figures 1 and 2. Approximately three-quarters of the children are between 9 and 11 years with an even distribution of boys and girls. COMET children are ethnically diverse and recruitment was distributed similarly across the four centers (Table 2). Ten percent of COMET children were sibling pairs (n ϭ 24 sibling pairs/48 children), which is within the limits set as part of the study design.

Ninety-seven percent of children (456/469) had a prior diagnosis of myopia reported by their parents. As expected, based on the eligibility criteria, COMET children did not have any significant medical or ocular conditions. Parents reported systemic illnesses such as asthma/allergies, attention deficit disorder, and sickle cell trait in only 5% (32/469) of children. As a result of the eligibility criteria, no child had strabismus under any of the following conditions: at distance through the distance prescription, at near through the distance prescription (simulating conditions for single vision assignment), or at near through a ϩ2.00 D addition lens (simulating conditions for a PAL assignment).

The distribution of cycloplegic autorefraction measurements for each eye is presented in Figure 1. These measurements are based on the median values of the five consecutive, reliable values obtained for each eye from the autorefractor. The distribution of refractive error is the same in both eyes. The means (Ϯ SD) of these median refraction values are Ϫ2.40 D (Ϯ 0.81) and Ϫ2.40 D (Ϯ 0.82), (20/32) 159 (34) a Data are presented for the right eye; data for the left eye are similar.

UAB ϭ University of Alabama at Birmingham, UH ϭ University of Houston.

for the right and left eyes, respectively. Additional information comparing these measurements with noncycloplegic autorefraction and the final prescription is included in another report [20]. Axial length is also distributed similarly in both eyes. The mean (Ϯ SD) and median baseline axial length measurements are 24.1 mm (Ϯ 0.7), based on the mean of three to five measurements for each eye, as shown in Figure 2.

As shown in Table 2, baseline visual acuity is good, with all children having acuity of 20/32 or better in both eyes.

DISCUSSION

COMET is the first multicenter, double-masked, randomized, controlled clinical trial to evaluate the effect of PALs versus SVLs in children in the United States. The identification of effective treatment for myopia is becoming increasingly important given the high prevalence of myopia, particularly in some Asian countries, and its increasing prevalence over time. In addition, the high cost and increasing popularity of refractive surgery suggest a need to identify alternative, noninvasive treatments that are targeted toward myopia prevention or slowing progression, particularly in children.

The rationale for the use of PALs is that they may reduce retinal blur in myopic children, the population at risk for the highest amount of progression, by providing clear visual input over a range of viewing distances. Therefore, the results of COMET will contribute to our understanding of the role that blur may play in myopia progression. If found to be successful in slowing myopia progression, the use of PALs would offer a noninvasive, relatively inexpensive alternative that is similar to the widely accepted conventional treatment (SVLs). COMET exceeded its recruitment goals by enrolling 469 children within 1 year. Since COMET children are an ethnically diverse, healthy group with myopia ranging from Ϫ1.25 D to Ϫ4.50 D at baseline, the results of this study should be generalizable to a large group of children with myopia. COMET will also provide experience with recruitment, retention, and adherence strategies for a clinical trial involving healthy children in diverse populations. Follow-up of COMET children for a minimum of 3 years will provide a first step to answering whether myopia progression may be slowed. In addition to refraction measures, data are being collected on ocular components, accommodation and phoria, and keratometry, thus providing the opportunity to further understand the relationship of these factors with myopia progression. This study will also provide useful information on myopia progression in children wearing conventional SVLs.