Authors’ reply

Abstract

It has thus far provided a substantial amount of information regarding the detection of high refractive error and amblyopia in preschool children.In the VIP study, acuity-based tests outperformed photoscreening, which is in contrast to our study. We believe there may be several reasons for this. In the VIP study, the acuity tests were administered by optometrists and ophthalmologists who were experienced in screening children and had a significant interest in achieving high levels of testability; in our study, the tests were administered by office staff, who probably had little such motivation. These results are probably more similar to what will occur with universal screening. This hypothesis also is supported by the phase 2 VIP data, 1 in which the standard phase 1 acuity-based test showed a substantial decrease in sensitivity when administered by personnel who were neither ophthalmologists nor optometrists. The VIP study used an enriched population of children with disease. Thus each child had a higher pretest likelihood of having pathology than would be expected in routine vision screening, which may have influenced the testers in some unknown manner. In the VIP, the test performance was not evaluated in the field, where disease prevalence is much lower.The statistical analysis of the VIP data also may have influenced the results. In the VIP study, all patients who could not be tested (and who had unreadable photoscreening pictures) were considered an examination failure. The unreadable rate for photoscreening in VIP was 5.8%. This result was nearly twice as high as our field experience, where photoscreening was performed by lay volunteers, most of whom are retired. The reason for this discrepancy is unclear; however, this high unreadable rate falsely lowered the specificity of photoscreening, which was the outcome at which sensitivity was determined. An additional criticism of the VIP statistical analysis was that the referral criteria for all screening techniques except photoscreening were defined post hoc, to maximize sensitivity. The sensitivity of photoscreening was determined based upon the original interpretation of the photograph, and a post-hoc adjustment of photoscreening photographs to maximize sensitivity was not allowed. Although it is unlikely that photoscreening would have outperformed traditional screening by the same magnitude reported in our study had these issues not been present, we believe photoscreening would have produced more credible results had these issues been dealt with. In August 2005, the VIP chair and statistician and I discussed these issues and chose not to pursue a reanalysis of the photoscreening images, for several reasons, primarily related to the lack of availability of the MTI photoscreener and interpretation.The "important criteria" used as gold standard criteria in VIP were slightly different than the amblyogenic factors published by the AAPOS vision screening committee and recommended by that committee to be used for reporting all studies of preschool vision screenin...