README.md

In Penncil/ODACH: Privacy-Preserving Distributed Algorithms

ODACH: A One-shot Distributed Algorithm for Cox model with Heterogeneous Multi-center Data

Outline

1. Description
2. ODACH workflow
3. Package requirements
4. Install package
5. Run ODACH
6. Results and running time
7. How to run ODACH on your data?

Description

This README is prepared for journal peer review of the ODACH paper. The proposed ODACH algorithm can be used to fit stratified Cox proportional hazards regression using multi-center data and requires each site to communicate only some aggregate data (AD), but achieves better estimates than the meta-analysis estimates.

To demonstrate its usage, we use a lung cancer survival data of 228 patients (survival::lung) and assume the patients are from 3 sites. We analyze the association of survival time with age and sex.

ODACH is a distributed algorithm under PDA (Privacy-preserving Distributed Algorithms) framework. The package in this repo is for the whole PDA framework, including other distributed algorithms.

ODACH workflow (Figure 1 in the paper)

Package Requirements

• A database with clear and consistent variable names
• R version: R (>= 3.5.0)
• On Windows: download and install RTools
• To install the pda package, make sure you have cpp compiler as pda requires Rcpp. This is a technical requirement for pda and not related to ODACH algorithm.

Install package

To implement ODACH algorithms, we need to install the pda package.

1. In RStudio, create a new project: File -> New Project... -> New Directory -> New Project.

2. Execute the following R code:

install.packages("devtools")
library(devtools)
devtools::install_github("penncil/pda")

# Or you can install from CRAN:
install.packages("pda")

The installation time is about 20 secs via github and 5 secs via CRAN (normal personal computer).

In the toy example below we aim to analyze the association of lung cancer survival time with age and sex using Cox PH regression, assuming site-specific baseline hazards. We demonstrate using PDA ODACH can obtain more accurate estimates than the meta-analysis estimates (i.e. inverse variance weighted average of individual estimates from each site).

We run the example in local directory. In actual collaboration, account/password for pda server will be assigned to the sites at the server https://pdamethods.org/. Each site can access via web browser to check the communication of the summary stats. Use demo(ODACH) for an automatic demo of the example.

Run ODACH example with code

Step 0: load related R packages and prepare sample data

## load packages
require(pda) 
require(survival) 

## data
data(lung2)  

## split the data to 3 separate sets (patient-level data)
lung_split <- split(lung2, lung2$site)

Step 1: Initialization

## setup pda control
sites = c('site1', 'site2', 'site3')
control <- list(project_name = 'Lung cancer study',
                step = 'initialize',
                sites = sites,
                heterogeneity = TRUE,
                model = 'ODAC',
                family = 'cox',
                outcome = "Surv(time, status)",
                variables = c('age', 'sex'),
                optim_maxit = 100,
                lead_site = 'site1',
                upload_date = as.character(Sys.time()) )


## specify your working directory, default is the current working dir
mydir <- getwd()   
pda(site_id = 'site1', control = control, dir = mydir)
# you now can see control.json in the working dir

## site3 communicate its individual estimate: after review, enter "1" to allow tranferring to lead site 
pda(site_id = 'site3', ipdata = lung_split[[3]], dir=mydir)
# you now can see site3_initialize.json in the working dir

## site2 communicate its individual estimate: after review, enter "1" to allow tranferring to lead site   
pda(site_id = 'site2', ipdata = lung_split[[2]], dir=mydir)
# you now can see site2_initialize.json in the working dir

## site1 communicate its individual estimate: after review, enter "1" to allow tranferring to lead site   
pda(site_id = 'site2', ipdata = lung_split[[1]], dir=mydir)
# you now can see site1_initialize.json in the working dir
# all the individual estimates are ready, control.json is also automatically updated to the next step

STEP 2: calculate derivatives (AD)

## site3 communicate its derivatives as AD: after review, enter "1" to allow tranferring AD   
pda(site_id = 'site3', ipdata = lung_split[[3]], dir=mydir)
# you now can see site3_derivatives.json in the working dir

## site2 communicate its derivatives as AD: after review, enter "1" to allow tranferring AD
pda(site_id = 'site2', ipdata = lung_split[[2]], dir=mydir)
# you now can see site2_derivatives.json in the working dir

## site1 communicate its derivatives as AD: after review, enter "1" to allow tranferring AD
pda(site_id = 'site1', ipdata = lung_split[[1]], dir=mydir)
# you now can see site1_derivatives.json in the working dir
# all the individual AD are ready, control.json is also automatically updated to the next step

STEP 3: estimation using AD

## lead site obtain surrogate estimates using ADs
pda(site_id = 'site1', ipdata = lung_split[[1]], dir=mydir)
# You now can see site1_estimate.json in the working dir

## the PDA ODACH is now completed! 
## All the sites can still run their own surrogate estimates and broadcast them.


## compare the surrogate estimate with the pooled and meta estimates
config <- getCloudConfig(site_id = 'site1', dir=mydir)
fit.pda <- pdaGet(name = 'site1_estimate', config = config)
control <- pdaGet('control', config)
cbind(b.pool=fit.pool$coef,
      b.meta =control$beta_init,
      b.pda=fit.pda$btilde )

Results and Running time

Estimated log hazard ratios:

| | b.pool | b.meta | b.odach | |-------------|---------|---------|---------| | age | 0.0170 | 0.0156 | 0.0170 | | sex |-0.5569 |-0.7876 |-0.5564 |

The running time in each site in this demo is about 9-10 secs.

How to run ODACH on your data?

• Get the data ready, which requires no missing values and clear variable names.
• Define the control by specifying the model, data, outcome, variables, site names, and local site.
• Directly run pda function as illustrated above.

Penncil/ODACH documentation built on March 21, 2022, 12:02 a.m.