Subsequent to publication of the article, “Long-Term Healthcare and Cost Outcomes of Disease Management in a Large, Randomized, Community-Based Population with Heart Failure,” by Galbreath et al, which appeared in the December 7, 2004, issue (Circulation. 2004;110:3518–3526), the authors conducted additional outcome analyses as they pursued other lines of inquiry. In the course of this work, they recomputed survival times using a different methodology that they believe more accurately represents the differences in restricted mean survival time reported in their paper. The restricted mean survival times that were published were computed by truncating the curve at the time of the last death in the intervention and control groups. Because the last death in the control group occurred earlier than the last death in the pooled experimental group, this led to an overstatement of the difference in mean survival time between the control and disease management groups. A more appropriate point for computation of mean survival time is at 547 days, which represents the last observed time in both groups. The authors recomputed restricted mean survival time limited to this point and present the original and revised survival times in Table 1 below. The graphs of the survival functions as presented in the figures and the tests of significance in the article are correct as published.
In addition, although the final percentages of surviving patients from Figures 1 and 2 were present in draft manuscripts, they were inadvertently omitted from the final version. Although they are not revised data, the authors believe their inclusion here to be worthwhile, as they obviate the need to estimate the figures from the survival curves. In addition to the product-limit percentages presented below, figures on absolute mortality by study group and type of heart failure were also omitted. In the control group, 33 of 245 systolic and 6 of 114 diastolic heart failure patients died by 18 months compared to 42 of 506 systolic and 12 of 204 diastolic heart failure patients in the intervention group by the same time point.
In conclusion, although the restricted mean survival times are shorter than initially reported, they are also based on data with extensive censoring and may not be the ideal descriptor of the survival benefit of disease management. Table 2 shows that differences in final survival percentages at 18 months were 2.1% and 3.8% for all patients and systolic heart failure patients, respectively. The impact on event-free survival was larger: Disease management created a difference in the final percentage of patients who experienced neither mortality nor a cardiac hospitalization of 4.7% for all patients and 9.1% for systolic heart failure patients.