diff --git a/references/references.bib b/references/references.bib
index c96aac0..415689a 100644
--- a/references/references.bib
+++ b/references/references.bib
@@ -173,3 +173,14 @@ @misc{catalogofbias_collider
 note = {Catalog of Bias}
 }
 
+@article{gooley1999competing,
+author = {Gooley, Ted A. and Leisenring, Wendy and Crowley, John and Storer, Barry E.},
+title = {Estimation of Failure Probabilities in the Presence of Competing Risks: New Representations of Old Estimators},
+journal = {Statistics in Medicine},
+year = {1999},
+volume = {18},
+number = {6},
+pages = {695--706},
+doi = {10.1002/(SICI)1097-0258(19990330)18:6<695::AID-SIM60>3.0.CO;2-O},
+url = {https://doi.org/10.1002/(SICI)1097-0258(19990330)18:6%3C695::AID-SIM60%3E3.0.CO;2-O}
+}
diff --git a/summary.qmd b/summary.qmd
index 18436ae..1e96a13 100644
--- a/summary.qmd
+++ b/summary.qmd
@@ -1,7 +1,11 @@
-# Summary
+# Summary of Gooley et al. (1999)
 
-In summary, this book has no content whatsoever.
+Gooley and colleagues explain how to estimate event probabilities when **competing risks** are present—situations where one event prevents the event of interest from occurring [@gooley1999competing]. Their central point is that the common shortcut of using **1 − Kaplan–Meier (1 − KM)** overestimates risk in this setting because it treats competing events as ordinary censoring [@gooley1999competing].
 
-```{r}
-1 + 1
-```
+The paper shows that the **cumulative incidence (CI)** estimator is the correct quantity for estimating the absolute probability of a specific failure type in the presence of competing risks [@gooley1999competing]. CI properly accounts for both the hazard of the event of interest and the hazard of competing events, while 1 − KM does not [@gooley1999competing].
+
+Using simulated data and two clinical examples, the authors demonstrate that divergence between 1 − KM and CI grows as competing events become more frequent; CI aligns with observed proportions when follow-up is complete, while 1 − KM can be substantially inflated [@gooley1999competing].
+
+## Practical takeaway
+
+If your goal is to estimate the real-world probability of a specific endpoint over time and competing events exist, report **cumulative incidence** (and clearly define competing events), not 1 − KM [@gooley1999competing].