Using Instrumental Variable Analysis to Adjust for Treatment Non-Compliance
Author Information
Author(s): Miladinovic Branko, Kumar Ambuj, Hozo Iztok, Djulbegovic Benjamin
Primary Institution: University of South Florida
Hypothesis
Can instrumental variable analysis correct for bias in non-experimental data from individual patient data meta-analysis?
Conclusion
Instrumental variable analysis suggests that methotrexate may be beneficial in preventing chronic graft-versus-host disease, contrary to intention-to-treat analysis.
Supporting Evidence
- Instrumental variable analysis showed no significant difference in the odds of chronic graft-versus-host disease between treatment groups.
- Intention-to-treat analysis indicated a detrimental effect of methotrexate, while instrumental variable analysis suggested potential benefits.
- The study utilized a strong instrument based on center prescribing preference.
Takeaway
This study shows that using a special method called instrumental variable analysis can help us understand if a treatment is actually helpful, even when some patients didn't follow the treatment plan.
Methodology
The study used previously collected individual patient data from nine randomized controlled trials and applied instrumental variable methods to analyze the effects of methotrexate.
Potential Biases
The key difficulty in conducting an IV analysis is finding and justifying a strong instrument.
Limitations
The study did not address complexities involving multiple instruments or confounders and used a less efficient two-step regression modeling approach.
Participant Demographics
Patients who received either peripheral blood stem cell transplant or bone marrow transplant.
Statistical Information
P-Value
0.004
Confidence Interval
95% CI 1.02-1.76
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website