This repository contains the official implementation of the FedECA method in the Flower framework.

The original implementation in the Substra framework used in the FedECA article is available here.

Please cite our paper if you use our code!

On a new environment with Python 3.10-3.12, install the app:

# cd to this directory, then
pip install -e .

For illustration purpose, a toy example of "distributed datasets" can be found in the data folder. The example contains four data centers, each holding a synthetic dataset of right-censored time-to-event data with 10 covariates. All demonstrations below are based on this toy example.

FedECA runs in both simulation and deployment modes. By default, it runs without bootstrapping and reports results with naïve and robust variance estimates, both after 10 Newton-Raphson steps. Bootstrapping can be enabled by setting the bootstrap parameter in pyproject.toml to true. Both global (default) and per-client bootstrapping are supported.

From this directory, run the app:

flwr run .

You will see a log like the following:

Loading project configuration...
Success
INFO :      (iter: 0) Success after 5 FedCox steps (norm_delta = 6.527887117250224e-09)
-------------- FedECA (naïve) ---------------
       coef  se(coef)  coef lower 95%  coef upper 95%         z         p  exp(coef)  exp(coef) lower 95%  exp(coef) upper 95%
0  0.041718  0.049127        -0.05457        0.138005  0.849178  0.395782     1.0426             0.946892             1.147981
-------------- FedECA (robust) ---------------
       coef  se(coef)  coef lower 95%  coef upper 95%         z         p  exp(coef)  exp(coef) lower 95%  exp(coef) upper 95%
0  0.041718  0.070581       -0.096618        0.180054  0.591062  0.554479     1.0426             0.907902             1.197282

You can extend the above command to modify the config used (see the [tool.flwr.app.config] section in the pyproject.toml). Let's say you want to do 20 steps of Federated NewtonRaphson. Then start the run like this:

flwr run . --run-config="nr-iterations=20"

And if you want to enable bootstrapping and run for 200 iterations (default is 20), do:

flwr run . --run-config="bootstrap=true bootstrap-iterations=200"

# Will report a summary:
-------------- Bootstrap FedECA (n_iter = 200 + 1) ---------------
       coef  se(coef)  coef lower 95%  coef upper 95%         z         p  exp(coef)  exp(coef) lower 95%  exp(coef) upper 95%
0  0.041718  0.039048       -0.034816        0.118251  1.068359  0.285359     1.0426             0.965783             1.125527

By default "global" bootstrapping is used but you can enable per-client bootstrapping by overriding bootstrap-fn like this:

flwr run . --run-config="bootstrap=true bootstrap-iterations=200 bootstrap-fn='per-client'"

# Will report a summary:
-------------- Bootstrap FedECA (n_iter = 200 + 1) ---------------
       coef  se(coef)  coef lower 95%  coef upper 95%        z         p  exp(coef)  exp(coef) lower 95%  exp(coef) upper 95%
0  0.041718  0.037539       -0.031857        0.115292  1.11132  0.266431     1.0426             0.968645             1.122202

You can spawn a Flower federation comprised of one SuperLink and four SuperNodes by running the compose.yaml file in this repository. You'll need to first build the image described in the Dockerfile. This will (1) install the dependencies in the pyproject.toml and (2) copy the content of the data/ directory to the containers.

# Build the container (only needed the first time or if you edit Dockerfile)
# This image will be used by both SuperLink and all SuperNodes
docker build . -t flower-fedeca-base-with-deps

# There after, spawn the federation
docker compose up

On a new terminal, you may use docker stats to see the services that have been launched. You'll see something like:

CONTAINER ID   NAME                                CPU %     MEM USAGE / LIMIT     MEM %     NET I/O           BLOCK I/O    PIDS
69d56719c413   owkin-flower-fedeca-supernode-1-1   1.17%     48.14MiB / 31.29GiB   0.15%     11.2kB / 7.88kB   0B / 4.1kB   33
25617260c2f7   owkin-flower-fedeca-supernode-4-1   1.29%     48.16MiB / 31.29GiB   0.15%     11.2kB / 7.94kB   0B / 4.1kB   33
e628e5cd63f9   owkin-flower-fedeca-supernode-3-1   1.07%     48.14MiB / 31.29GiB   0.15%     11.2kB / 8.01kB   0B / 4.1kB   33
9239157488ab   owkin-flower-fedeca-supernode-2-1   1.23%     48.14MiB / 31.29GiB   0.15%     11.1kB / 8.07kB   0B / 4.1kB   33
204f2b4e4a0c   owkin-flower-fedeca-superlink-1     0.49%     48.5MiB / 31.29GiB    0.15%     30.6kB / 11.7kB   0B / 4.1kB   38

With the infrastructure (i.e. SuperLink and SuperNodes) ready, you can submit a run. You can achieve so by pointing the flwr run command to another federation (in this case a real one created with the compose file). If you inspect the pyproject.toml you'll find the remote federation at the bottom. Run the Flower App in your deployed federation with:

# "remote" is the name in the federation (you can change it)
# "--stream" enables you to see the logs generated by the ServerApp
# on the terminal you run `flwr run` from.
flwr run . remote --stream

The above demonstrations use the toy example data pre-partitioned into four centers. When the FedECA app runs, in simulation or deployment mode, the ClientApps load the appropriate datasets by following the data path template defined in the [tool.flwr.app.config] section of the pyproject.toml as:

[tool.flwr.app.config]
...
path-to-data = "data/center{}/data.csv"

Then, at runtime, each ClientApp completes that path with their respective partition id.

To test FedECA with your own data (currently only csv files are supported):

1. Partition your data on a per-center basis into different csv files. For examle, name them partition1.csv, partition2.csv, etc.
2. In the pyproject.toml, set path-to-data = "path/to/my/partition{}.csv". Alternatively, you can override this setting at runtime as:

      flwr run . --run-config="path-to-data='path/to/my/partition{}.csv'"

3. Optionally, if run with the Simulation Engine, make sure the options.num-supernodes in pyproject.toml sets the appropriate number of supernodes equal to the number of data partitions. If run with the Deployment Engine, make sure the compose.yaml file is spawning enough SuperNodes.

@article{OgierduTerrail2025,
  author = {Jean Ogier du Terrail and Quentin Klopfenstein and Honghao Li and Imke Mayer and Nicolas Loiseau and Mohammad Hallal and Michael Debouver and Thibault Camalon and Thibault Fouqueray and Jorge Arellano Castro and Zahia Yanes and Laëtitia Dahan and Julien Taïeb and Pierre Laurent-Puig and Jean-Baptiste Bachet and Shulin Zhao and Remy Nicolle and Jérôme Cros and Daniel Gonzalez and Robert Carreras-Torres and Adelaida Garcia Velasco and Kawther Abdilleh and Sudheer Doss and Félix Balazard and Mathieu Andreux},
  title = {FedECA: federated external control arms for causal inference with time-to-event data in distributed settings},
  journal = {Nature Communications},
  year = {2025},
  volume = {16},
  number = {1},
  pages = {7496},
  doi = {10.1038/s41467-025-62525-z},
  url = {https://doi.org/10.1038/s41467-025-62525-z},
  abstract = {External control arms can inform early clinical development of experimental drugs and provide efficacy evidence for regulatory approval. However, accessing sufficient real-world or historical clinical trials data is challenging. Indeed, regulations protecting patients’ rights by strictly controlling data processing make pooling data from multiple sources in a central server often difficult. To address these limitations, we develop a method that leverages federated learning to enable inverse probability of treatment weighting for time-to-event outcomes on separate cohorts without needing to pool data. To showcase its potential, we apply it in different settings of increasing complexity, culminating with a real-world use-case in which our method is used to compare the treatment effect of two approved chemotherapy regimens using data from three separate cohorts of patients with metastatic pancreatic cancer. By sharing our code, we hope it will foster the creation of federated research networks and thus accelerate drug development.},
  issn = {2041-1723}
}