An evaluation of c-reactive protein and sleep in women with breast cancer using the All of Us dataset

Background: Women who have breast cancer are likely to encounter sleep disturbances for various reasons. Wellness is associated with the quality of sleep one receives, and the quality of sleep is also associated with healing. Continuous sleep disturbances could lead to a poorer quality of life. Therefore, we used the SDoH survey in the All of Us dataset to evaluate noise in the neighborhoods of participants with breast cancer as a measure of sleep disturbances. Women with breast cancer and poor sleep quality may experience disease progression which could be correlated by inflammation biomarkers, such as c-reactive protein. Methods: According to the SDoH survey from the All of Us version 8 dataset, participants were divided into two cohorts, noisy and non-noisy living environments. We used multiple statistical tests to compare the groups' data. The Welch's t-test was used to compare the two cohorts and their awake, light, deep, and REM sleep level in minutes. We also used the Mann Whitney U test to compare the four levels of sleep. The correlation between sleep length and c-reactive protein was tested by Pearson's correlation coefficient. Each test was completed via Jupyter notebook using Python coding. Results: Upon update of the All of Us dataset version 8 update, about two-hundred and ninety-nine breast malignancy-diagnosed women met the inclusion criteria of the cohort to compare noise levels of their neighborhood, quality of sleep level, and the measurement of their C-reactive protein via FitBit data and surveys. There was no statistically significant difference between noise levels and minutes of sleep or sleep level and c-reactive protein measurements. Conclusion: It is well-established that quality sleep is essential for overall wellness. The data we collected did not establish correlation between sleep level and C-reactive protein measurements. While our outcomes are not as we predicted, it is important to know that the data were skewed, as the two cohorts were not evenly distributed. Furthermore, despite the addition of participants in the version 8 dataset, we observed that the representation of medically underserved and understudied communities was still deficient. To provide more population representative data, there needs to be more diversity amongst the participants providing pertinent data.