Results Summary

PCORI funded the Pilot Projects to explore how to conduct and use patient-centered outcomes research in ways that can better serve patients and the healthcare community. Learn more.


People who have the same chronic condition often receive different treatments. The choice of treatment sometimes depends on things like the patient’s age, sex, and treatment history. Also, doctors often change treatment over time depending on how the patient responds. When treatment varies in these ways, it is called dynamic. Researchers find it difficult to analyze data about the effectiveness of dynamic treatments because the way the treatments are used changes over time.

For example, how well the medicine being studied works might depend on how well a medicine the patient took in the past worked. Usual research methods don’t handle this difficulty well.

The research team proposed using an analysis method called the “parametric g-formula” to address this difficulty. This method has not yet been used with existing data from medical records or health insurance companies.

Project Purpose

The researchers compared the g-formula method with a more commonly used method of analysis called “inverse probability weighting” (IPW) to learn the advantages and disadvantages of using these methods to study dynamic treatments.


The research team compared the g-formula and IPW methods using medical record data stored in a registry of patients with chronic kidney disease called the United States Renal Data System (USRDS). The study included records for 156,945 patients who met all of the following criteria:

  • Had both kidney failure and heart disease
  • Got hemodialysis (using a machine to clean the blood when the kidneys don’t work right) in an outpatient clinic each June from 2006 to 2010
  • Took a medicine called epoetin alfa (EPO), which is used to help patients make more red blood cells

Patients got different amounts of EPO depending on whether the doctor wanted the patient to maintain low, medium, or high levels of red blood cells.

The research team used the g-formula and IPW methods to see how often patients who got these different treatments died or had heart attacks in an 18-month period. They also compared the results from both methods to results from a randomized controlled trial that asked a similar question.


The research team found that patients who got EPO doses intended to give them low levels of red blood cells had less chance of dying than patients who got EPO doses intended to give them medium or high levels of red blood cells. The g-formula and IPW methods produced the same results. The results using patient registry data were the same as the results from the large randomized trial.


Both the g-formula and IPW methods are limited by the data that are available at the time a study is conducted. For example, if something influences health and affects the patient outcome (in this case, mortality) but was not recorded in the data, then researchers can’t take it into account. In this study, researchers don’t know the exact EPO dose and red blood cell response from each treatment. Instead, researchers only have monthly data on EPO doses and patient red blood cell levels.


The research team found that the g-formula and IPW methods worked equally well for this case. The research team encourages using the g-formula and IPW together whenever possible to confirm the results.

Findings from this study support current US Food and Drug Administration guidelines that recommend keeping levels of red blood cells in the blood low (less than 33 percent) for patients with kidney and heart disease who get hemodialysis.

Sharing the Results

The researchers shared results of their work through a webinar and are submitting articles to medical journals (see below).

Project Information

Yi Zhang, PhD, MS
Medical Technology & Practice Patterns Institute, Inc.
Comparing Dynamic Treatment Strategies in Patient-Centered Outcomes Research

Key Dates

June 2012
December 2014

Study Registration Information


Has Results
Award Type
Funding Type
State State The state where the project originates, or where the primary institution or organization is located. View Glossary
Last updated: April 12, 2023