Article type
Year
Abstract
Background: To assess the effect on albumin excretion rate (AER) of ACE inhibitors in normotensive type 1 diabetic patients with microalbuminuria and to assess whether the treatment effect varied across certain subgroups using individual patient data. To compare the results using different summary measures for repeated outcome measures.
Methods: Studies with at least 10 normotensive type 1 diabetic patients with microalbuminuria with a placebo or non-intervention arm and an ACE inhibitors arm and at least 1 year of follow-up were included. Individual patient data was obtained for all 12 trials identified. Since the distribution of AER was skew, the log values of AER were used to calculate the treatment effect. Regression through the origin (RTO), ordinary least squares (OLS) and the mean of the post-randomisation measurements (MEAN) were the summary measures considered. To interpret the results for RTO and OLS in terms of actual AER values, the relative change in AER over a 2 year period was calculated. In all cases both a trial level analysis, using standard random effects models, and an individual patient data analysis, using multilevel models, were carried out. For the individual patient data analysis, models were extended to adjust for baseline AER and potential confounders and then to test the treatment by baseline factor interactions.
Results: RTO estimates, with adjustment for baseline, were deemed the most suitable choice for this example, given the expected pattern of change in AER over time. The 2 year treatment effect was observed to decrease with increasing follow-up when using the RTO estimate. The trial level analysis using 2 year follow-up data (10 trials), indicated that AER was 51.4% (95% confidence interval: 31.1-65.6%) lower on ACE inhibitor than control. The individual patient data analysis using RTO produced an estimate of treatment effect of 50.5% (29.2-65.4%) and this effect was found to vary by baseline AER (p=0.04 for treatment by baseline interaction) where those with a higher baseline AER showed a greater treatment benefit. No other interactions were significant at the 5% level. The basic estimates of treatment effect using OLS and MEAN for studies with at least 2 years of follow-up were 37.9% (23.7-49.5%) and 32.8% (24.7-40.0%), respectively.
Conclusions: ACE inhibitors provided a beneficial treatment effect in the group of patients studied, which differed by baseline AER. The analysis illustrates the use of summary outcome measures when repeated measures are available, an analysis which would not have been possible from published data. As with any single trial, the most appropriate measure of treatment effect should be chosen carefully and attention paid to the length of follow-up and adjustment for baseline measures.
Methods: Studies with at least 10 normotensive type 1 diabetic patients with microalbuminuria with a placebo or non-intervention arm and an ACE inhibitors arm and at least 1 year of follow-up were included. Individual patient data was obtained for all 12 trials identified. Since the distribution of AER was skew, the log values of AER were used to calculate the treatment effect. Regression through the origin (RTO), ordinary least squares (OLS) and the mean of the post-randomisation measurements (MEAN) were the summary measures considered. To interpret the results for RTO and OLS in terms of actual AER values, the relative change in AER over a 2 year period was calculated. In all cases both a trial level analysis, using standard random effects models, and an individual patient data analysis, using multilevel models, were carried out. For the individual patient data analysis, models were extended to adjust for baseline AER and potential confounders and then to test the treatment by baseline factor interactions.
Results: RTO estimates, with adjustment for baseline, were deemed the most suitable choice for this example, given the expected pattern of change in AER over time. The 2 year treatment effect was observed to decrease with increasing follow-up when using the RTO estimate. The trial level analysis using 2 year follow-up data (10 trials), indicated that AER was 51.4% (95% confidence interval: 31.1-65.6%) lower on ACE inhibitor than control. The individual patient data analysis using RTO produced an estimate of treatment effect of 50.5% (29.2-65.4%) and this effect was found to vary by baseline AER (p=0.04 for treatment by baseline interaction) where those with a higher baseline AER showed a greater treatment benefit. No other interactions were significant at the 5% level. The basic estimates of treatment effect using OLS and MEAN for studies with at least 2 years of follow-up were 37.9% (23.7-49.5%) and 32.8% (24.7-40.0%), respectively.
Conclusions: ACE inhibitors provided a beneficial treatment effect in the group of patients studied, which differed by baseline AER. The analysis illustrates the use of summary outcome measures when repeated measures are available, an analysis which would not have been possible from published data. As with any single trial, the most appropriate measure of treatment effect should be chosen carefully and attention paid to the length of follow-up and adjustment for baseline measures.