We found that testing rates did not significantly increase following the sentinel event and that testing rates varied significantly depending on ordering specialty. These findings suggest that significant opportunities for improvement in testing continue to exist.

The variation in testing rates among specialties may be due to varying treatment norms. Specialties with lower testing rates included hematology, oncology, and transplant medicine, where patients are likely to undergo routine blood tests during treatment, ensuring up-to-date status regarding any potential adverse events. In specialties with significantly more PGx testing (e.g., dermatology, gastroenterology, and rheumatology), azathioprine is prescribed to combat inflammation, and most prescriptions are for outpatients, which may result in fewer opportunities for CBC monitoring.

Race-related differences in testing rates may be related to provider awareness of racial differences in TPMT genotype and phenotype status. Studies show lower TPMT activity in Black individuals than in White or Asian individuals, and Dickson et al. observed that Black patients discontinued azathioprine due to hematopoietic toxicity at a higher rate than White patients [4, 10,11,12]. Additionally, studies indicate that the prevalence of TPMT variants differs among races; the most common variant allele is TPMT*3A in White individuals and TPMT*3C in Black and Asian individuals [13]. Research has observed a lower prevalence of TPMT variants among Chinese and South Asian individuals than White individuals [14].

The initial increase in PGx testing from 2010 to 2014 was likely due to an increased understanding and availability of TPMT PGx testing; for instance, the initial CPIC® guideline for TPMT testing and dosing was published in 2011 [15]. A similar pattern of increased testing was observed by Dickson et al. although in that patient population, testing continued to increase through 2018 [4]. Within our cohort, a significant portion of patients prescribed thiopurines were still not tested either before or after treatment initiation. Additionally, neither the sentinel event nor implementation of MUET was associated with a significant increase in testing.

We believe that the current, predominantly reactive, paradigm for treating patients with potentially decreased TPMT activity is insufficient. Variation in practices among various specialties presents significant challenges to the implementation of TPMT PGx testing and dissemination of testing results [16]. Additionally, as shown by Coenen et al., adoption of one form of testing does not exclude or trivialize other tests, and continued monitoring by CBCs even after TPMT PGx testing resulted in fewer patients with hematological toxicity [5]. Thus, integrating information on testing status and results in the EHRs is critical to ensure that providers are aware of testing needs and available genetic information to guide therapeutic decisions. A combination of increased PGx testing and the incorporation of decision supports and clinical dashboards into EHR systems may result in a more proactive approach to the prevention of drug-use adverse events associated with genetic variants.

This study was limited by constraints of the VA population. Although the VA is the largest healthcare system in the U.S., its population is predominantly male and older. We determined the likely sources of the prescription for inpatient population based on chronological proximity and not direct link to the prescription. To limit potential bias from past exposure to thiopurines, we only included patients who were newly initiated on the treatment. We did not report TPMT testing performed independent of thiopurine treatment initiation.