Among 8,368 individuals screened for participation, 4,259 were identified with CHF. Of these, information was missing on LV-EF in 150, NYHA class in 25, both LV-EF and NYHA in 4, and age in 1 patient. Of 4,079 patients available for analysis, 2,785 (68%) had HFrEF and 1,294 (32%) had HFpEF. Out of these, 3,476 (85.2%) completed the SF-36 questionnaire. In multivariate logistic regression analysis including age, sex and CHF group as covariates, only age significantly predicted the completion of the SF-36 questionnaire.
HFpEF patients were older, more frequently female, had higher blood pressures and were less intensively treated with heart failure medications (Table ). They were also less symptomatic with a lower mean NYHA class and a lower number of CHF symptoms per patient with lower rates for all individual symptoms except peripheral edema (Table and Fig. ). SF-36 PF score was higher in HFpEF.
Distribution across a NYHA grades and b individual number of CHF symptoms in patients with HFrEF (open columns) or HFpEF (filled columns)
Only hypertension and obesity were more frequent in HFpEF, while all other comorbidities were observed more frequently in HFrEF patients. Seven comorbidities did not show an interaction with CHF group in ordinal regression analysis, i.e., their influence on NYHA class and SF-36 PFS was similar in both groups (Table ). All these comorbidities had a highly significant impact on patients’ symptoms by increasing NYHA class and reducing SF-36 PF score, with COPD and anemia showing the strongest associations. The remaining five comorbidities behaved differently in HFpEF and HFrEF: In HFrEF, hypertension and hyperlipidemia were associated with a better NYHA class, while the latter impacted on the NYHA class negatively in HFpEF and was also associated with a worse SF-36 PF in these patients. CAD had a negative effect on NYHA class in HFpEF only and its association with lower SF-36 PF was significantly stronger than that observed in HFrEF patients. Similarly, a negative impact of PAOD on NYHA class and SF-36 PF was significantly stronger in HFpEF compared with HFrEF. Obesity significantly increased NYHA class in HFpEF only, while there was no significant difference in its (negative) impact on SF-36 PF in both CHF groups.
Impact on NYHA class and SF-36 physical functioning scale of comorbidities, corrected for age and sex
In multivariate analyses, hyperuricemia, renal dysfunction, anemia, COPD, cerebrovascular disease and atrial fibrillation retained their significantly negative impact on NYHA class in HFrEF, while hypertension and hyperlipidemia retained their protective effect (Fig. a). The strength of association for the negatively influencing comorbidities was similar to that of a reduction in LV-EF of 10% (which corresponds to a quartile of the distribution for this parameter in our cohort), with widely overlapping confidence intervals. In HFpEF, only CAD, anemia, obesity, COPD and atrial fibrillation had a significantly negative effect in multivariate ordinal regression analysis, although wider confidence intervals illustrated the lower statistical power to show influencing factors due to the lower subject number in this group. The effect of these comorbidities appeared to be much stronger than that of an LVD (ED) lowered by 5 mm (again corresponding to a quartile in our cohort), the most strongly associated parameter on routine echocardiography in this patient group. SF-36 PF score was negatively affected in HFrEF by all comorbidities except for diabetes, hypertension and hyperlipidemia, with the last two again showing a protective effect with regard to this parameter (Fig. b). The effects tended to be stronger than that of an LV-EF reduction. In HFpEF, hyperuricemia, CAD, renal dysfunction, anemia, obesity, COPD, PAOD and atrial fibrillation were associated with lower SF-36 PF score. Again, the effect of reduced LVD (ED) was significant but weak in comparison with comorbidities. LV-EF had no significant effect on HFpEF and the same held true for LVD (ED) in HFrEF.
Odds ratios in multivariate analyses for a higher NYHA class and b SF-36 physical functioning score in HFrEF or HFpEF
Accordingly, adding both echocardiographic variables as covariates improved the prediction of NYHA classes III or IV in HFrEF, derived from the multivariate regression model, while AUC remained virtually unchanged in HFpEF (Fig. a). Expansion of the set of covariates for all comorbidities further improved the AUC in HFrEF to an extent similar to that provided by the addition of the echocardiographic variables. However, in HFpEF, the increase in AUC was much larger (+0.100 vs. +0.029), resulting in a similar AUC as in HFrEF for the final multivariate model that included all covariates, with a relatively larger share supplied by comorbidities in HFpEF.
a AUC to predict higher NYHA class and br2 of multivariate linear model for SF-36 physical functioning score according to the set of covariates used for model building
In HFpEF, considerably more of the overall variability in SF-36 PF score was explained by sex and age only compared with HFrEF (Fig. b). Again, adding echocardiographic variables did not improve the multivariate model in HFpEF, while some improvement was noted for HFrEF, and the inclusion of all comorbidities further improved r2 by 7.6%. In HFpEF, comorbidities added 12.1% and thereby approximately doubled the explained variance, resulting in an r2 of 25.1%. The variance explained by comorbidities was therefore 1.59-fold higher in HFpEF than in HFrEF. The overall explained variance of SF-36 PF in HFpEF by sex, age, echocardiographic variables and comorbidities was 1.49-fold larger than in HFrEF, even though the echocardiographic variables had no impact on r2 in HFpEF.