Is the tolerability of long-term thiazolidinedione therapy overstated?

The study by Tang et al. (1) concluded that their retrospective chart review demonstrated the tolerability of long-term thiazolidinedione (TZD) therapy in a diabetic population with established chronic heart failure (HF). Although we applaud their efforts to study this important topic, we believe the data presented are not so clear, and that their conclusions that a large majority of chronic HF patients tolerate these agents are overstated. First, we believe the definition of TZD-related fluid retention, as a 10-pound weight gain from baseline, in addition to signs or symptoms of volume overload, is far too exclusive of important levels of fluid retention. By using this cut-off, we can be sure that those patients had severe fluid retention. However, we do not know the number of other patients who had important levels of weight gain or edema and who were missed by the investigators’ likely insensitive criteria. Heart failure guidelines recommend action when weight increases by 2 to 4 pounds depending on how quickly it occurs. Second, we disagree that the reported incidence of fluid retention of 17.1% is an overestimate due to selection bias. In fact, it is probably an underestimate. Obtaining data from a chart review can only lead to under-reporting the true incidence of fluid retention and adverse events. Furthermore, the majority of patients had stable New York Heart Association functional class I or II heart failure where TZD therapy is not contraindicated. (The incidence of edema [with or without weight gain] in TZD randomized controlled trials ranged between 2% and 15% [2,3].) Third, the intolerability of these agents in this population is further illustrated by the fact that 31% discontinued TZD therapy within one year of initiation (most due to fluid retention), whereas the rate of discontinuation was far 0.1% in randomized controlled trials (2,3). Also of concern is that 26% of patients who met the criteria for fluid retention were hospitalized. Finally, we are concerned that both the data and the discussion regarding the incidence of fluid retention and characteristics of the non-TZD control group were very limited. This data would likely provide more insight into the true tolerability of these medications in this population. In summary, we agree that further studies are needed to examine the relationship between TZD-related fluid retention and patient cardiac status. We believe this study and its conclusions should be interpreted very carefully, as the true risks of adverse effects related to volume expansion are likely understated.