This tweet has stuck with me. How do habits, lifestyle, etc. impact metabolic health? Recently, I dug into the relationship between my @Levels Metabolic Scores and biometrics from @zerofasting | @whoop.

of my findings AND a tool you can use to perform a similar analysis. https://twitter.com/j_b_pauly/status/1334525043775254529

I've been using @Levels for over 3 months and have consistently used @whoop and periodically fasted ( @zerofasting) throughout that time. 100 +/- a few data points is a small sample size but enough to identify correlations and make an interpretation.

I trimmed the metrics down to Metabolic Score, Strain, Recovery, Sleep Score, Sleep (hours), Fast (hours). Calculated cum. & cons. fast hrs. to create consistent daily metrics (if you know of better metrics around fasts, let me know!). Pics help depict the fasts metrics.

A quick check of the linear relationship between Metabolic score and all other metrics. Weak positive correlation across the board. Trending in the direction we might expect, but not strong enough to have real conviction in the correlations.

Taking a closer look by fitting an OLS regression gives us even less confidence! In this case, with all data points, the variance in my Sleep Score can only explain around 7% of the variance in my Metabolic Score.

But, maybe there's something there if we look at days with longer durations of fasting

By focusing on just days of fasting (13+ cumulative hours), the explainability of the variance in Metabolic Score more than doubles to around 20%. We can also start to see some clear outliers..

At this point, it's important to remember I want to focus on my typical habits and lifestyle. We all have outlier days (on both ends of the spectrum) when it comes to exercise and nutrition/consumption.

I pruned out a few outliers on days with insufficient data and choices inconsistent with my lifestyle. Most of these were around @Levels 'glucose challenges' (fun experiments to glean insights into foods / activities to steer clear from or lean into in the future)

Side note. I learned A LOT through the challenges and calibrations and recommend any @Levels user to go through as many as possible. The challenges and zone comparisons framework will be a cornerstone of the program.

Back to the analysis. After pruning, the correlations between Metabolic Score and Sleep Score, Sleep (hrs), Fasts (cumulative hrs), Fasts (consecutive hrs) improve between 35% and 80% (from raw correlations) across the board. Moderate correlations worth a closer look.

The coefficient of determination for the OLS regression improves to over 15% when using Sleep Score as a predictor for Metabolic Score (2X compared to the raw data). More importantly, we start to see better trending in the fasting days.

Zeroing in on Sleep Score vs. Metabolic Score for just days of fasting, the OLS coeff. of determination jumps to nearly 40% (can explain 40% of the variability in Metabolic Score with the variability of Sleep Score). This is over a 5X increase compared to the absolute raw trend.

Similar trends were also seen when comparing Strain vs. Metabolic Score on days of fasting. Coefficient of determination of 45%.

So, what did I learn? This all gives me a clearer picture on the relationship between my habits/lifestyle and metabolic health, in addition to the anecdotal evidence I noticed a month ago.

Important to remember Correlation != Causation.

My biggest takeaway:
find and maintain good habits. When I'm loaf, I'm likely eating poorly / not fasting IN ADDITION to not working out. The more I maintain good habits the better chance I have at maintaining metabolic fitness.

Lastly, this wasn't a strict experiment, there's MANY factors impacting metabolism, and always room for measurement and calculation errors. But the trends stick out to me.

It'll be fun to keep track with more data and I would love to see others try out their own analysis....

Here's the app I built if you want to perform your own analysis (with your data or a sample dataset), find research on the relationship between metabolism, sleep, fasting, and exercise, or learn more about @zerofasting @Levels @whoop.

https://share.streamlit.io/jbpauly/glucose-sleep-analysis/main/src/app.py

Feel free to pass along to other @Levels @zerofasting @whoop users or reach out with other datasets you'd like to see integrated!