🧵 A new Nature paper on “spatial ecotypes” claims liquid biopsy can predict immunotherapy response. The tumor biology is cool. The statistics are a masterclass in how to do bad biomarker research. @f2harrell has a checklist for this. Let’s go. 1/
First, Harrell’s Biomarker Uncertainty Principle: “A molecular signature derived from high-dimensional data can be either parsimonious or predictive, but not both.” This paper tries to be both. It earns neither. 2/fharrell.com/post/badb
May 1 • 11 tweets • 2 min read
❗️ Hot take Friday: Precision oncology is the biggest narrative scam in cancer medicine. “Find the right gene and we’ll cure cancer.” We’ve been hearing this for 30 years. Meanwhile, surgery and radiation are still the only things that actually cure solid tumors. Let’s go 🧵
New Nature paper just dropped tracking lung cancer evolution from dx to death across every met. 79% of mets had unique subclones not found anywhere else. There is no “the mutation.” There are thousands of mutations, evolving independently, in real time. nature.com/articles/s4158…
Jan 3 • 6 tweets • 1 min read
🚨 Why treat Stage III NSCLC with a chainsaw when you can use a sniper rifle? 🛑 Bayesian meta-regression of 20 years of multiple RCTs confirms: the era of surgery has been superseded by non-invasive precision 1/N
Utilizing a hierarchical Bayesian framework to aggregate longitudinal Phase III outcomes (2007–2026), the posterior distribution reveals a 62.6% probability of overall survival superiority for definitive CRT over surgical resection 2/N
Nov 21, 2025 • 9 tweets • 3 min read
🧵 For various lung dose-volume constraints, which are highly correlated, what does v5, v10, mean lung dose, add vs v20? Here we see spearman matrix expectedly highly correlated 1/n
Very importantly, let’s model v5, v10, v20 and MLD using flexible RCS, which also creates a lot of correlated basis expansions 2/n
Nov 15, 2025 • 6 tweets • 2 min read
🧵 How do we decide if SBRT or surgery is better for early-stage lung cancer patients when factoring survival and quality of life? Let’s visualize it with Monte Carlo simulations from STARS/ROSEL data. 🫁 1/n
Weighted utility = 2× survival + QoL.
SBRT has higher survival → utilities cluster higher than surgery.
This weighting drives most patients above the scipy.optimize threshold
