Noninvasive, Accessible, Smartphone App for at-Home Hemoglobin Monitoring in Sickle Cell Disease

Anemia is a widespread health issue impacting more than 2 billion people worldwide. In sickle cell disease (SCD), which impacts more than 100k individuals in the USA, severity of anemia is associated with development of complications, end organ damage, and poor quality of life. The gold standard test for hemoglobin (Hgb) monitoring is the complete blood count, which is invasive and costly. Therefore, we developed and launched a smartphone app that measures Hgb levels using fingernail pictures. This app is noninvasive, inexpensive, patient-supported, easy to use, and accessible, having been downloaded more than 200,000 times across the USA with thousands of active users per month and more than 1.4 million tests taken since launch in Dec 2021. Here we investigated its utility as a Hgb monitoring tool for people with SCD. This investigational app has not been reviewed or approved for SCD by the FDA.

We enrolled 35 participants (26 Female/8 Male, 35 years old ±7 years) diagnosed with SCD (26 Hgb SS, 6 Hgb SC, 2 Hgb Sb⁺ thalassemia, 1 unknown) in 3 cities (Atlanta, Houston, and Washington D.C) for this investigational study. No participants were excluded from the study due to presence of nailbed discoloration or limited technological competence regarding smartphone use. The study consisted of up to 4 site visits across a total of 8 weeks. At each visit, participants took an app test and received a fingerstick Hgb test for calibration. In between visits, participants took app tests every other day and recorded their symptoms and medications in the app. Finally, participants took usability and acceptance surveys upon completion of the study.

In this study, we found the mean absolute error (MAE) of the test to be ±0.9 g/dL (R² = 0.6, root mean squared error = ± 1.2 g/dL, 95% limits of agreement = ±2.3 g/dL), which is within the range of accuracy accepted by the FDA. Despite the variability introduced by a significantly anemic population and self-testing in a variety of locations and background lighting conditions, these results are similar to our previously published results obtained in clinical settings (Mannino et. al., Nature Communications, 2018). 10 participants took hydroxyurea (HU) throughout the study, which can cause fingernail discoloration, and 3 more participants took disease modifying drugs that do not discolor fingernails (Voxelotor, Crizanlizumab, L-glutamine). The MAE in the participants who took HU was 0.2 g/dL greater (±1.0 g/dL) than in those not taking any disease-modifying drug (±0.8 g/dL). However, the MAE of the HU group was still within the accepted deviation range. Furthermore, the MAE of the 3 participants who took the non-nail discoloring drugs was ±0.7 g/dL, which coupled with low sample size, does not suggest a significant impact of these drugs on app results. In addition to accuracy, we noticed that app Hgb levels correlated with presence of anemia related symptoms. In tests where users reported symptoms in the app, Hgb levels were significantly different on average 8.3 g/dL, compared to 8.7 g/dL when participants did not report any symptoms (P = 0.04). This suggests that the app was able to pick up small changes in Hgb level when participants felt worse. Study participants were eager to use the app, with 100% of participants taking more than 11 test over the duration of the study, and 91% of participants taking more than 21 tests. Moreover, these participants indicated that they had a positive experience with the app, with 13 indicating they took action based on Hgb measurements resulting in better management of their health. 97% of the participants indicated that they would like to incorporate the app as part of their daily routine after completion of the study. Despite these overall positive results, this study was limited by reliance on self-reporting of symptoms as well as unsupervised app use. This resulted in some protocol deviations that likely introduced variability into the data and forced the exclusion of 2 participants from data analysis.

The correlation between symptoms and Hgb levels calls for future research to better understand the relation of reported symptoms, including vaso-occlusive pain events and Hgb. Overall, the degree of accuracy reported and the correlations between SCD symptoms and Hgb suggest that this noninvasive app is a useful tool for individuals suffering from SCD including users of HU to frequently monitor their Hgb levels, empowering them to better self-manage their condition.