The day-7 prognostic divergence: ferritin trajectories separate survivors from non-survivors in a decade-long cohort of 9,746 hospitalised adults
- 1Rheumatology and Clinical Immunology, IRCCS Humanitas Research Hospital, Milan, Italy
- 2Department of Biomedical Sciences, Humanitas University, Milan, Italy
- 3Humanitas University, Milan, Italy
Abstract
Background: Serial ferritin measurements have been examined in disease-specific cohorts with heterogeneous results, but the temporal pattern of ferritin in a large, clinically diverse population of hyperferritinaemic inpatients remains undefined.
Methods: Retrospective single-centre cohort of adult inpatients with at least one serum ferritin ≥500 ng/mL (IRCCS Humanitas Research Hospital, October 2013–November 2024; N = 9,746). Thirty-day all-cause in-hospital mortality was analysed in 8,439 patients (668 deaths) by Kaplan–Meier and Cox regression adjusted for age, sex, Charlson Comorbidity Index and principal diagnosis (complete-case n = 6,549; 218 deaths). In a serial-sampling subgroup (n = 388; 387 with evaluable outcome, 45 deaths), parsimonious Cox models tested three scalar dynamic indices and mixed-effects models described log₁₀-ferritin trajectories across 0–3, 4–7 and >7 days; day-7 landmark analyses, restricted cubic splines, multiple imputation and Fine–Gray regression with live discharge as the competing event served as sensitivity analyses.
Results: Thirty-day mortality rose across baseline ferritin strata (13.1%, 20.2% and 32.9%; log-rank p < 0.001); adjusted hazard ratios versus 500–999 ng/mL were 1.97 (95% CI 1.50–2.60) for 1,000–4,999 ng/mL and 5.46 (2.89–10.31) for ≥5,000 ng/mL. In age- and sex-adjusted longitudinal models, peak ferritin (HR 3.91 per log₁₀ unit, 95% CI 2.10–7.30; p < 0.001) and delta ferritin (HR 4.83, 1.60–14.57; p = 0.005), but not time-to-peak (p = 0.531), were associated with mortality. Beyond day 7, survivors' ferritin was 16.7% lower than during days 0–3 (p < 0.001), whereas the within-group change among patients who subsequently died was not significant; the relative difference was 15.5% (p = 0.020). Among 380 patients alive and hospitalised at day 7, the latest ferritin by the landmark predicted subsequent mortality (adjusted HR 2.86, 1.24–6.61; p = 0.014), whereas pre-landmark change, slope and any increase did not; thereafter ferritin fell by 2.30% per day in survivors but was stable in subsequent deaths (relative divergence 2.78% per day; p < 0.001). Spline modelling confirmed a non-linear time-by-outcome interaction (p = 0.036); Fine–Gray estimates were concordant.
Conclusions: Ferritin level and trajectory after the first week may identify a subgroup with persistent inflammatory activity and increased subsequent mortality. Day 7 should be regarded as a candidate temporal landmark for prospective validation, not as an established clinical decision threshold. These retrospective associations do not demonstrate that ferritin-guided interventions improve outcomes.
Related
- Cardiometabolic Risk Trajectories in Young Adults Across 16 Middle East and North African Countries, 1990-2018: A Sex-Asymmetric Divergence
- Joint Dynamic Trajectories of CEA and CA19-9 After Gastric Cancer Surgery: Prognostic Associations and a 12-Month Landmark Analysis
- Temporal Dynamics and Spectrum of Stroke Across Cancer Trajectories: A Multicenter Study of 190,000 Cancer Survivors
- Three-year trajectories of insomnia symptoms among young breast cancer survivors
- Growth trajectories and risk of underweight and overweight in childhood cancer patients and survivors: a longitudinal multicentre cohort study protocol
BibTeX
@article{brunetta2026day7,
title = {{The day-7 prognostic divergence: ferritin trajectories separate survivors from non-survivors in a decade-long cohort of 9,746 hospitalised adults}},
author = {Enrico Brunetta and Lorenzo Del Moro and Vahagn Menon and Carlo F. Selmi},
journal = {Frontiers in Medicine},
year = {2026},
month = oct,
doi = {10.3389/fmed.2026.1952601},
url = {https://doi.org/10.3389/fmed.2026.1952601},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - The day-7 prognostic divergence: ferritin trajectories separate survivors from non-survivors in a decade-long cohort of 9,746 hospitalised adults
AU - Enrico Brunetta
AU - Lorenzo Del Moro
AU - Vahagn Menon
AU - Carlo F. Selmi
PY - 2026
DA - 2026/10/21
JO - Frontiers in Medicine
DO - 10.3389/fmed.2026.1952601
UR - https://doi.org/10.3389/fmed.2026.1952601
AB - Background: Serial ferritin measurements have been examined in disease-specific cohorts with heterogeneous results, but the temporal pattern of ferritin in a large, clinically diverse population of hyperferritinaemic inpatients remains undefined. Methods: Retrospective single-centre cohort of adult inpatients with at least one serum ferritin ≥500 ng/mL (IRCCS Humanitas Research Hospital, October 2013–November 2024; N = 9,746). Thirty-day all-cause in-hospital mortality was analysed in 8,439 patients (668 deaths) by Kaplan–Meier and Cox regression adjusted for age, sex, Charlson Comorbidity Index and principal diagnosis (complete-case n = 6,549; 218 deaths). In a serial-sampling subgroup (n = 388; 387 with evaluable outcome, 45 deaths), parsimonious Cox models tested three scalar dynamic indices and mixed-effects models described log₁₀-ferritin trajectories across 0–3, 4–7 and >7 days; day-7 landmark analyses, restricted cubic splines, multiple imputation and Fine–Gray regression with live discharge as the competing event served as sensitivity analyses. Results: Thirty-day mortality rose across baseline ferritin strata (13.1%, 20.2% and 32.9%; log-rank p < 0.001); adjusted hazard ratios versus 500–999 ng/mL were 1.97 (95% CI 1.50–2.60) for 1,000–4,999 ng/mL and 5.46 (2.89–10.31) for ≥5,000 ng/mL. In age- and sex-adjusted longitudinal models, peak ferritin (HR 3.91 per log₁₀ unit, 95% CI 2.10–7.30; p < 0.001) and delta ferritin (HR 4.83, 1.60–14.57; p = 0.005), but not time-to-peak (p = 0.531), were associated with mortality. Beyond day 7, survivors' ferritin was 16.7% lower than during days 0–3 (p < 0.001), whereas the within-group change among patients who subsequently died was not significant; the relative difference was 15.5% (p = 0.020). Among 380 patients alive and hospitalised at day 7, the latest ferritin by the landmark predicted subsequent mortality (adjusted HR 2.86, 1.24–6.61; p = 0.014), whereas pre-landmark change, slope and any increase did not; thereafter ferritin fell by 2.30% per day in survivors but was stable in subsequent deaths (relative divergence 2.78% per day; p < 0.001). Spline modelling confirmed a non-linear time-by-outcome interaction (p = 0.036); Fine–Gray estimates were concordant. Conclusions: Ferritin level and trajectory after the first week may identify a subgroup with persistent inflammatory activity and increased subsequent mortality. Day 7 should be regarded as a candidate temporal landmark for prospective validation, not as an established clinical decision threshold. These retrospective associations do not demonstrate that ferritin-guided interventions improve outcomes.
SN - 2296-858X
ER -