Week 2 Discussion

Week 2 Discussion

by Luis Rodriguez -
Number of replies: 1

Article

Hambiliki et al, J Assist Reprod Genet (2011). A prospective randomized sibling-oocyte study of two media systems for culturing cleavage-stage embryos—impact on fertilization rate.  https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3114965/

 

Unit of Clustering

In this randomized study, there was a two-level structure: by woman and by oocytes (immature egg cells). The investigators evaluated the impact of two different media systems (A or B) on fertilization rate. Each woman (n = 110) provided an average of 11 oocytes, which were randomized to be fertilized and cultured in media A or media B.

 

Statistical Model Used

Investigators used the Chi-square test to compare fertilization rates and found that the rate of normal fertilization was significantly higher in group A than in group B (73% vs 67%; p-value 0.03).

Alternative Statistical Models

In this study, since there were 110 clusters (women) with an average of 11 observations each (oocytes), investigators could have performed a GEE analysis to account for the correlation by cluster. In this case since there is only one level of cluster with more clusters than observations per cluster, a GEE analysis can be utilized. The point estimate will likely be affected very little (if at all), but the standard errors will be smaller since the model will incorporate dependence of repeated measures (oocytes). 

In reply to Luis Rodriguez

Re: Week 2 Discussion

by Maria Glymour -

Interesting paper Luis.  I believe they did the analysis incorrectly, since they did not seem to do anything to account for the correlated observations of multiple oocytes from the same woman.  Your proposal would work.