Setting, Sampling, and Eligibility Criteria

The study was carried out in 2 nearby areas of western China: Tianshui prefecture, a poor area in Gansu, one of China’s poorest provinces, and Yulin prefecture, Shaanxi, a more affluent region in a middle-income province.

One school from each township in both prefectures was randomly selected from a list of all primary schools, and within each school, 1 class was randomly chosen in each of the fourth and fifth grades. For the original trial, all children at the 252 selected schools meeting the following criteria were eligible:

• •
  

  Uncorrected (without glasses) VA ≤6/12 in either eye

  
  
• •
  

  Refractive error as follows:

  
  
  
  
  • ○
    

    myopia ≤−0.75 diopters (D),

    
    
  • ○
    

    hyperopia ≥+2.00 D, or

    
    
  • ○
    

    astigmatism (nonspherical refractive error) ≥1.00 D

  
  
  
  
• •
  

  VA could be improved to >6/12 in both eyes with glasses

In the current analysis, carried out by eye rather than child, all nonmyopic (refractive error >−0.5 D) eyes of eligible children were excluded ( Figure 1 ).

Enrollment and progress of children through a randomized trial on the effects of the provision of spectacles to Chinese school-age children.

Questionnaires

At baseline (September 2012: beginning of the school year), enumerators administered questionnaires to children concerning their age, sex, glasses wear, awareness of their refractive status, boarding at school, and parental migration and education. A parental questionnaire asked about ownership of 13 selected items as an index of family wealth. Mathematics teachers were asked to state whether the blackboard was used for all teaching, most, about half, little, or none. At closeout (May–June 2013: end of the school year) children again filled out a questionnaire on glasses wear. Population density was calculated as the total population divided by total land area at the township level.

Assessment of Visual Acuity

Children underwent baseline VA screening at school by a nurse and staff assistant, previously trained by optometrists from ZOC. VA was tested separately for each eye without refractive correction at 4 meters using Early Treatment Diabetic Retinopathy Study (ETDRS) charts (Precision Vision, La Salle, Illinois, USA) in a well-lighted, indoor area. If the orientation of at least 4 of 5 optotypes on the 6/60 line was correctly identified, children were examined on the 6/30 line, on the 6/15 line, and then line by line to 6/3. If a line was failed, lines above were tested successively until the child identified 4 of 5 optotypes, with the VA for an eye defined as the lowest line read successfully. If the top line could not be read at 4 meters, the subject was tested as above at 1 meter, and the measured VA was divided by 4.

Refraction (Measurement of Glasses Power)

Children with uncorrected VA ≤6/12 in either eye underwent cycloplegia with up to 3 drops of cyclopentolate 1% and automated refraction (Topcon KR 8900; Tokyo, Japan) with subjective refinement by a refractionist, previously trained by experienced pediatric optometrists from ZOC.

Randomization and Interventions

This was a cluster-randomized, controlled trial, with schools as the clusters. In October 2012, after the baseline survey and vision screening but prior to refraction, eligible children were randomized by school to receive 1 of 3 interventions ( Figure 1 ):

• •
  

  Free spectacles based on the child’s measured refractive power dispensed at school by the study optometrist (Free Glasses group, 84 schools);

  
  
• •
  

  Vouchers bearing the child’s name, school name, and glasses prescription, exchangeable for free glasses at the local county hospital (distance from children’s township: range 1–105 km, median 30 km). Parents were responsible for paying transportation costs (Voucher group, 84 schools); or

  
  
• •
  

  A glasses prescription and letter to the parents informing them of the refractive status of their child, with free glasses provided only at closeout, though this was not previously announced (Control group, 84 schools).

Schools were stratified by size, county, and number of children failing VA screening. We used R software (R Foundation for Statistical Computing, Vienna, Austria) to generate blocks of and randomly allocate schools within each block to the treatment arms.

Outcome Assessment

At closeout, VA was assessed using the protocol and vision chart described above. Spectacle wear was assessed through unannounced direct examinations on the same occasion. Children also described their own spectacle wear as “always,” “only for studying,” or “usually not worn.” Study personnel were masked to group assignment. Participants (students, parents, and teachers) and enumerators were not informed of either the overall design of the study or the explicit treatment arm assignment. During this visit, all children provided information on parental spectacle wear, and their own time spent out of doors and in near/middle distance work, important determinants of myopia progression.

Statistical Methods

Family wealth was calculated by summing the value, as reported in the China Rural Household Survey Yearbook (Department of Rural Surveys, National Bureau of Statistics of China, 2013), of items on the list of 13 owned by the family. Refractive power was defined throughout as the spherical equivalent, the spherical power plus half the cylindrical power.

Randomization groups were compared by intention-to-treat (ITT) analysis using multiple linear regression, with end-line uncorrected VA (logarithm of the minimal angle of resolution [logMAR]) as main outcome and intervention arms and baseline uncorrected VA as covariates. Other baseline variables were also investigated as predictors for final VA, with the final model including intervention arms and variables associated with baseline VA at P ≤ .20. Student and school were included in a random intercept model to adjust for the correlation between eyes of a student, between children in the same school, and between schools within the same randomization block. All analyses were performed using Stata 12.0 (StataCorp, College Station, Texas, USA), and SAS 9.3 (SAS Institute, Cary, North Carolina, USA).