Alternative analytic and matching approaches for the prevalent new-user design: A simulation study

Area of Science:

• Pharmacoepidemiology
• Biostatistics
• Health Services Research

Background:

• Prevalent new user (PNU) cohorts are crucial for real-world treatment effect estimation.
• Selecting appropriate analytic methods is vital to minimize bias in PNU studies.

Purpose of the Study:

• To evaluate the bias and computational efficiency of different analytical and matching techniques within PNU cohorts.
• To compare the performance of time-conditional propensity score (TCPS) matching with alternative approaches.

Main Methods:

• A simulated cohort was used to estimate treatment effects comparing TCPS matching, SMRW, DRS, and other propensity score matching methods.
• The relative bias was assessed by comparing estimated risk ratios (RR) to a known RR of 1.00 across 2000 replicates.

Main Results:

• SMRW (RR=0.998) and DRS (RR=0.997) provided unbiased estimates.
• TCPS matching with replacement (RR=0.999) and without replacement (RR=0.999 when starting with shortest treatment history) were also unbiased.
• Matching strategies that prioritized longer treatment histories or random comparator selection introduced significant bias (RR=0.983, 0.903, and 0.802, respectively).

Conclusions:

• Several analytical methods can achieve unbiased treatment effect estimation in PNU cohorts.
• Caution is advised when employing complex matching strategies, particularly those deviating from the original TCPS approach, to avoid introducing bias.