An Imputation-Estimation Algorithm Using Time-Varying Auxiliary Covariates for a Longitudinal Model When Outcome is Missing by Design
Open AccessIn long term clinical trials, occurrence of missing data is an area of concernespecially if the rate at which data are missing depends on the treatment group. Typically, some effort is spent on trying to identify the reasons the data aremissing so that appropriate assumptions and analytic approaches can be properlyapplied. When data are missing by design, certain measurements are discontinuedafter meeting an endpoint, possibly due to ethical or financial constraints.Subjects who reach the absorbing barrier may stop data collection on somevariables but may subsequent time-varying covariates available from continuedfollow-up. In this dissertation, we developed an Imputation-Estimationalgorithm under an auxiliary missing at random assumption to assess whether theadditional information from the time varying covariates can be used to improveestimation. Quality of estimates is evaluated in terms of bias, variance andcoverage for the estimates of the parameters of interest. We contrast thismethod to other missing data approaches such as multiple imputation and available case analysis.We illustrate this method using data from the Diabetes Prevention Program (DPP). The DPP was a diabetes prevention study that showed reductions of 58\% and 31\%in diabetes risk using intensive lifestyle or metformin interventions comparedto placebo. According to the DPP protocol, the oral glucose tolerance test isdiscontinued after diabetes diagnosis. Because of the significant reduction indiabetes incidence by the metformin and lifestyle interventions, the rates ofmissing IGR and CIR are different among the treatment groups. This differentialdiscontinuation among treatment groups results in informative monotone missingassessments of 30 minute glucose and insulin values. These 30 minute values areused to calculate surrogate measures of insulin secretion such as InsulinGlucose Ratio (IGR = (30-min insulin - fasting insulin)/(30-min glucose -fasting glucose)). Fasting blood glucose is collected at all time points andis associated with 30-minute glucose. The imputation estimation algorithm isapplied to estimate the mean 30 minute blood glucose utilizing auxiliaryinformation from the fasting blood glucose. In this example, fasting glucose isalso the source of the discontinuation since diabetes diagnosis is based on thefasting glucose and 2 hour values during the OGTT. Because of the strongdependence between the fasting and 30 minute glucose measured at the same visit,the resulting estimates from the IE algorithm using the complete vector weresimilar to multiple imputation. Because the Placebo group experienced higherrates of diabetes incidence, the difference between available case analysisand the regression based imputations were greater than in the lifestyle group.
- All rights reserved
Notice to Authors
If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.
| Thumbnail | Title | Date Uploaded | Visibility | Actions |
|---|---|---|---|---|
|
|
Temprosa_gwu_0075A_11516.pdf | 2018-01-16 | Open Access |
|