Letter by Höglund and Nilsson Regarding Article, “Influence of Nonfatal Hospitalization for Heart Failure on Subsequent Mortality in Patients With Chronic Heart Failure”
To the Editor:
We read with great interest the article by Solomon et al1 concerning the influence of nonfatal hospitalizations for heart failure on subsequent mortality rates. However, we are intrigued by the design of this investigation and by the statistical approach chosen. Three clinical trials were merged into 1 observational study; a minor problem is determining what population the study actually represents.
Our major concern is that the study is indeed observational and that the findings are probably due to the clinical courses of HF patients being variable, as noted in another publication from the Candesartan in Heart failure: Assessment of Reduction in Mortality and morbidity (CHARM) program.2 It is well-known that in observational studies, it is vital to include all covariates that are important, including both constant and time-varying factors, such as medication, comorbidities, and the progress of the disease (eg, change of New York Heart Association class3) to avoid confounding. Picking only those variables that were important at baseline in the clinical trial may be problematic, as the questions asked now are different from the ones of the clinical trial. Updating only HF hospitalizations will only serve to identify patients when they have a more severe disease than indicated by their baseline covariates. Some patients with an intermittent worsening of HF may return to their baseline risk level, whereas others who had irreversible health problems may have a worse prognosis than indicated by baseline covariates, and those who are terminally ill are discharged to a hospice and have very high mortality rates.
Although it is nigh impossible to infer individual characteristics from population summaries, it is evident from Table 1 in the article by Solomon et al1 that sicker patients seek medical care more often. Now, if your HF worsens, should you avoid the hospital, and if you did enter the hospital, should you try to get out prematurely? Solomon et al1 believe that the nonfatal hospitalizations have prognostic value and, furthermore, repeatedly use the word “influence,” whereas we argue that it is the deterioration in health that occurred before the hospitalization that has a prognostic value and that the hospitalization per se probably leads to a better prognosis. It is very likely, however, that this reduced risk at discharge is significantly higher than in patients with similar baseline characteristics that did not, for example, change New York Heart Association class. Of course, although it is not totally impossible that all patients have a higher mortality rate due to HF hospitalization itself, the analysis cannot disentangle these confounded outcomes.
To avoid ecological fallacies, it would have been necessary to update patient data at discharge. These studies have to be carefully designed, as biases may appear if only hospitalized patients are updated, as in Setoguchi et al.4
Although some of these concerns are listed as limitations in the discussion, the fact remains that leaving out important covariates in an observational study may actually reverse effect estimates, as in the hormone replacement and coronary disease studies.5 We acknowledge the important work of Solomon et al,1 but caution is required when extrapolating their findings to the general population.
Solomon SD, Dobson J, Pocock S, Skali H, McMurray JJV, Granger CB, Yusuf S, Swedberg K, Young JB, Michelson EL, Pfeffer MA, Candesartan in Heart failure: Assessment of Reduction in Mortality and morbidity (CHARM) Investigators. Influence of nonfatal hospitalization for heart failure on subsequent mortality in patients with chronic heart failure. Circulation. 2007; 116: 1482–1487.
Pocock SJ, Wang D, Pfeffer MA, Yusuf S, McMurray JJV, Swedberg KB, Östergren J, Michelson EL, Pieper KS, Granger CB. Predictors of mortality and morbidity in patients with chronic heart failure. Eur Heart J. 2006; 27: 65–75.
Lawlor DA, Smith GD, Ebrahim S. Commentary: The hormone replacement–coronary heart disease conundrum: is this the death of observational epidemiology? Int J Epidemiol. 2004; 33: 464–467.