Statistical problems caused by missing data resulting from neodymium:YAG laser capsulotomies in long-term posterior capsule opacification studies: problem identification and possible solutions

J Cataract Refract Surg. 2008 Feb;34(2):268-73. doi: 10.1016/j.jcrs.2007.11.014.

Abstract

Purpose: To describe a way to eliminate the problem caused by missing data due to neodymium:YAG (Nd:YAG) laser capsulotomies before statistical analysis, which leads to incomplete data in long-term studies of posterior capsule opacification (PCO).

Setting: Medical University of Vienna, Vienna, Austria.

Methods: To demonstrate the problem, an existing long-term study comparing PCO development in eyes with round-edged and sharp-edged intraocular lenses (IOLs) was analyzed. In the study, several Nd:YAG capsulotomies led to dropouts in the round-edged IOL group. To solve the resulting missing value problem, the missing values were replaced by estimated PCO values in various scenarios; that is, different assumptions were made of how PCO would have developed without a capsulotomy. Standard statistical analyses that included the estimated PCO values rather than excluding the incomplete cases were then performed.

Results: The analysis of all cases (including the estimated PCO scores) showed a highly significant difference between the round-edged group and the sharp-edged group. Sensitivity analyses confirmed that the difference between the 2 groups remained significant up to highly implausible values chosen for the missing PCO scores, thus proving the correctness of the analysis.

Conclusions: This study showed it is possible to perform statistical analyses without excluding incomplete cases. This approach is recommended for all studies with Nd:YAG dropouts as it increases the statistical power. Excluding incomplete cases may lead to biased or wrong study conclusions.

MeSH terms

  • Cataract / etiology
  • Cataract / therapy
  • Data Interpretation, Statistical*
  • Humans
  • Laser Therapy / methods*
  • Lens Capsule, Crystalline / pathology
  • Lens Capsule, Crystalline / surgery*
  • Models, Statistical*
  • Patient Dropouts / statistics & numerical data
  • Postoperative Complications*
  • Statistics as Topic / methods*