This prospective cataract registry by the Pediatric Eye Disease Investigator Group (PEDIG) was supported through a cooperative agreement with the National Eye Institute of the National Institutes of Health. An Investigational Device Exemption (#G110149) was obtained from the United States Food and Drug Administration for use of an IOL in children. The protocol and Health Insurance Portability and Accountability Act-compliant informed consent forms were approved by each site’s institutional review board. A parent or guardian of each enrolled child provided written informed consent; assent was obtained as required by local institutional review boards. Ethnicity, race, and sex assigned at birth were reported by the parent or guardian. Children and parents did not receive incentives or compensation.

The PEDIG prospective cataract registry design, methods, and outcomes have been reported in detail. ^,, From June 2012 to July 2015, investigators at 61 sites enrolled 994 children (1361 eyes) <13 years of age who had undergone unilateral or bilateral lensectomy during the preceding 45 days. Primary IOL selection and placement were at investigator discretion. Medical and ophthalmic data were collected from medical record review at enrollment and 5 annual visits thereafter.

The analysis cohort was limited to children aged <13 years at lensectomy for non-traumatic cataract with primary IOL placement (pseudophakia). Aphakic eyes undergoing secondary IOL implantation, and eyes diagnosed with glaucoma or glaucoma suspect at enrollment or at any time during follow-up were excluded. Only children with refractive error measurements at both an early postoperative examination (window: 60 days to 1.5 years [548 days]) and at a 5-year follow-up examination (window: 4 to 6 years [1461 to 2192 days]) were included. For children undergoing bilateral lensectomy, data on both eyes were included if eligible. Refractive error measurements <60 days following cataract surgery were not used to avoid inclusion of astigmatism change from wound healing. ^,

For simplicity, we use the term “astigmatism” to describe the clinical cylindrical refractive correction determined by spectacle-plane refraction and represented in spherocylindrical form with plus cylinders. Refraction was performed by retinoscopy, manifest refraction with subjective refinement, or automated refraction. The change in the astigmatism component of the clinical refraction between the early postoperative and 5-year timepoints was analyzed in 2 ways: (1) using clinical notation for astigmatism without regard to change in axis or sphere, and (2) using cylinder conversion to power vectors (J0, J45) to include the impact of axis change, with conversion back to clinical notation. The difference in each vector component between the 2 timepoints was calculated and then converted to clinical notation to represent the change in the astigmatism (C) component as described in Miller (see Table 3 and Example 3 in that publication), taken from the work of Thibos et al. :

“C” represents the astigmatic component of a lens that would need to be added (as an overcorrection) to the old refraction to obtain the new refraction. For example, if a child had astigmatism of + 1.00D at 90 degrees postoperatively and + 1.00D at 75 degrees at the 5-year visit, the astigmatism change would be zero when using clinical notation. Alternatively, if power vector conversions are used in this example, the calculated astigmatism change is 0.52 D because some astigmatism is converted to a sphere due to the axis change. As a second example, a child with + 1.00 + 1.00 D X 090 degrees that changed to + 1.00 + 1.50 D X 105 degrees, had an astigmatism increase of 0.50 D in clinical notation, but when analyzed with power vector conversions, the change of astigmatism would be 0.81 D. We do not include the change in sphere as significant changes are anticipated during the years after pediatric lensectomy.

The magnitude of early postoperative astigmatism was classified as either clinically not meaningful (0.00 to 0.50 D cylinder), low (>0.50 to 2.00 D cylinder), or high (>2.00 to 6.00 D). The median change in astigmatism using clinical notation was calculated for each early postoperative group. Using power vector representation of the refractive error, the median was also calculated for changes in spherical equivalent (M), J0, and J45. Two-dimensional plots of J0 and J45 were created for early postoperative and 5-year astigmatism timepoints and the change in astigmatism (transformed back to clinical notation).

Binary outcomes included a change in astigmatism of ≥ 1.00 D and ≥ 2.00 D between the early postoperative and 5-year follow-up examinations, analyzed with power vectors converted to clinical notation. Generalized estimating equations and 95% CIs from a binomial regression with compound symmetry correlation were used to estimate the proportions of eyes for each outcome according to low and high postoperative astigmatism groups and to account for non-independence of eye pairs. The same methods were used to estimate the proportion of eyes that changed ≥ 15 and ≥ 45 degrees in axis among eyes with low and high early postoperative astigmatism.

Using linear mixed modeling, mean astigmatism was estimated at early postoperative and 5-year follow-up visits as power vectors converted to clinical notation. ^,, This model was used to explore potential relationships between the change in astigmatism (power vector calculation) from early postoperative to 5-year visits with the following pre-specified predictors: age at lensectomy (continuous), laterality of the surgery (bilateral or unilateral), IOL fixation (capsular bag or sulcus), magnitude of postoperative astigmatism (continuous), and the number of months from lensectomy to the first reported postoperative refractive error measurement (continuous). A random intercept and variance components correlation structure were used to account for the nonindependence of eye pairs.

All analyses were conducted using SAS version 9.4 (SAS Institute Inc., Cary, NC). Analyses used a 2-sided comparison-wise Type 1 error rate of 0.05 and were not adjusted for multiplicity as all outcomes were considered exploratory.