As ASH 2025 approaches, we are pleased to share new work that digs deeper into one of the most difficult realities of sickle cell disease (SCD) – the unpredictability and impact of vaso-occlusive crises (VOCs). For many people living with SCD, crises arrive with little warning, disrupt daily life, and often result in emergency care and hospital stays. Even where high-frequency digital data exist, this typically focuses on cohort-level patterns, missing the development of a clearer view of how individual biomarker and symptom patterns behave across repeated crises, and how those patterns evolve over the years.

Our upcoming abstract to be presented as a poster at this year’s ASH meeting and exposition, entitled “Stability and evolution of individual-level biomarker patterns across repeated vaso-occlusive crises in sickle cell disease”, starts to fill that gap. Drawing on real-world data from individuals with SCD living in the UK, and who had recorded a high burden of VOCs, we analysed more than two and a half years of median follow-up collected through a digital health ecosystem – combining FDA-cleared wearable devices with a dedicated app for patient-reported outcomes (PROs). Patients contributed continuous activity, sleep, and heart rate data, alongside daily symptom reports, EQ-5D quality of life measures, and self-defined VOC periods, with inclusion criteria requiring at least three VOCs over at least two years.

By looking at percentage changes in these metrics before, during, and after each VOC, relative to a patient’s own first recorded crisis, we were able to characterise both overall variability for the cohort and within-person stability. At a cohort level, wearable activity metrics such as steps, elevation, and active time showed wide variation between crises, with median percentage changes clustered around zero, but with large interquartile ranges. In contrast, PROs were noticeably more consistent from episode to episode, acting as “steady anchors” even when movement-linked metrics were more highly disrupted.

Longitudinal analysis with mixed-effects models then explored how these patterns drifted over time. On VOC days, we observed small but statistically significant changes across multiple biometric domains: declines in steps, elevation, heart rate, calorie expenditure, and both deep and light sleep durations, as well as changes in time to fall asleep and wake, and sleep fragmentation. At the same time, QoL-linked measures – including EQ-5D and psychological scores – showed modest improvements over the follow-up period, suggesting that, for at least some patients, lived experience and coping strategies may be evolving even as crises remain a recurrent reality. Of note, most effect sizes remained small over this period, underlining the importance of nuanced, patient-level interpretation rather than broad generalisations.

When we drilled down into the phases immediately surrounding crises, early signals became clearer. In the pre-VOC window, we saw small but significant increases in soft activity, active time, deep sleep, and psychological scores, alongside decreases in heart rate and light sleep over time, with each subsequent crisis. Post-VOC periods similarly showed higher active time and psychological scores and declines in both heart rate and light sleep duration, but decreases in deep sleep duration not seen over sequential pre-VOC periods. While seasonality effects were minimal and longitudinal drift was overall stable across individuals, these consistent directional shifts around VOCs point towards a complex but reproducible pattern that may, with further exploration, help to support more personalised ‘early warning systems’ and support strategies.

Taken together, these findings highlight the potential utility of wearable metrics as highly episode-specific markers, sensitive to the unique context of each crisis and its surrounding days, while PRO measures may offer a relatively more stable backbone over time, capturing how patients perceive their health, pain, and psychological wellbeing across repeated VOCs. For future predictive modelling, this presents an opportunity for more powerful approaches that treat each person as their own reference point, combining volatile, episode-level biometric shifts with steadier PRO trajectories to better understand when an individual’s pattern is starting to shift from their usual state.

In practical terms, our goal is to further develop strategies that see the leveraging of this data to help people with SCD and their clinical teams spot meaningful change earlier, support recovery more effectively at home and in the community, and reduce the need for avoidable emergency care