Abstract
BACKGROUND: The influence of immortal time bias on estimates of survival of heart failure (HF) patients with improved ejection fraction (HFimpEF) has not been well elucidated. METHODS: This retrospective cohort study collected patients who had ever been hospitalized owing to HF between January 2012 and June 2022 at National Taiwan University Hospital (NTUH), the NTUH Hsin-Chu branch, or the NTUH Yun-Lin branch. Patients were classified as follows: (i) HF with reduced ejection fraction (HFrEF); (ii) HFimpEF; (iii) HF with mildly reduced ejection fraction (HFmrEF); and (iv) HF with preserved ejection fraction (HFpEF). Four different clinical scenarios were designed to illustrate the impact of the immortal time bias in estimating the long-term survival of patients with HFimpEF. RESULTS: Overall, 13,085 patients were categorized into 4 groups (HFrEF, n = 2029; HFimpEF, n = 1292; HFmrEF, n = 1849; and HFpEF, n = 7915). In our scenario 2, which retained the immortal time in HFimpEF patients, the HFimpEF group exhibited a substantially lower risk of all-cause mortality compared with the HFrEF group (adjusted hazard ratio 0.30, 95% confidence interval 0.27-0.33). After applying time-varying Cox regression analysis to appropriately account for immortal time bias in our scenario 4, model 4-2, the survival benefit was notably attenuated, yielding an adjusted hazard ratio of 0.77 (95% confidence interval 0.70-0.86) for the HFimpEF group relative to the HFrEF group. CONCLUSIONS: Time-varying Cox regression analysis should be used to minimize the immortal time bias in retrospective studies concerning survival of patients with HFimpEF.