Analysis of a panel-based pharmacogenomics testing program among members of a commercial and Medicare client of a pharmacy benefits manager

Abstract

BACKGROUND:

Although the field of pharmacogenomics (PGx) has existed for decades, use of pharmacogenomic information by providers to optimize medication therapy for patients has had relatively slow adoption. There are many factors that have contributed to the slow adoption of PGx testing, but it is partially due to a lack of coverage by payers. If PGx testing is covered by payers, frequently only testing of a specific gene is covered, rather than a panel of many genes. As a result, little is known about how coverage of a panel-based PGx test will affect a member’s medication therapy.

OBJECTIVES:

To determine how giving providers specific medication optimization recommendations, based on results of a panel-based PGx test, impacted members’ medication regimens.

METHODS:

Pharmacy claims data were retrospectively reviewed for this exploratory study. Members who participated in PGx testing were in the intervention group and members who chose not to participate in the PGx testing, but who were eligible to participate, were in the control group. PGx test results, including suggested medication changes, were mailed to providers. To determine if providers adopted the suggested medication changes, pharmacy claims data were analyzed retrospectively for the 4-month period preceding and following the date from which recommendations were provided to prescribers.

RESULTS:

Of the 101 members included in the analysis, 50 were in the intervention group and 51 were in the control group. In the intervention group, members were taking in a total of 352 medications; 165 of the medications had PGx guidance. Based on the PGx test results, 62 of these medications (37.6%) had recommendations. Of members who received PGx testing, 76% had at least 1 recommended change. When pharmacist recommendations were made, a change was made to the medication 27% of the time. There was a statistically significant difference between the number of medication changes in the PGx group and the control group (P = 0.024).

CONCLUSIONS:

Recommendations based on PGx testing can lead to changes in medications and an optimized medication regimen for members.

What is already known about this subject

• Pharmacogenomics can be used to guide medication selection and dosing and has been demonstrated to improve outcomes, but the adoption of pharmacogenomics has been relatively slow.

• There are many barriers to pharmacogenomics testing, including lack of coverage by payers and unfamiliarity of how to use pharmacogenomics test results by providers.

What this study adds

• This study reports the results of a payer-led pharmacogenomics program that is offered to members.

• Giving specific medication optimization recommendations to providers, based on the results of a pharmacogenomics test, can result in medication changes and optimization of medication regimens.

Current medication prescribing usually takes a trial-and-error approach, wherein a patient is first prescribed a medication, and if that medication does not produce the desired therapeutic outcome, a different medication is tried. It is estimated that between 25% and 50% of prescribed medications do not result in the desired therapeutic outcome.¹ For medications that do not result in the desired therapeutic outcome (therapeutic failures or adverse events), the results can be costly. A 2018 study found that nonoptimized medication regimens cost the United States as much as $528.4 billion annually because of increased use of medical resources (doctor’s visits, additional medications, emergency department visits, etc).² Not only is this costly, but it can lead to frustration and reduced quality of life for patients.

There are many reasons why a given drug therapy might not be effective or cause an adverse reaction in a patient, but a patient’s genetic composition has been demonstrated to play a major role.^3,4 Not only does genetics play a role in drug response, but a substantial number of people have genetic variants that would result in actionable changes in drug therapy. Depending on the population studied, it is estimated that between 88% and 99% of individuals have clinically actionable genetic variants.^5-7 Using a patient’s genetic information can help guide medication therapy to reduce adverse events and provide optimal dosing, leading to improved treatment outcomes and reduced costs.^8-10

Pharmacogenomics (PGx) is a field of medicine that studies how genetic variation contributes to different responses to drugs in different patients. The goal of PGx is to use a patient’s genetic information to guide medication therapy to help reduce adverse drug events and optimize drug efficacy (through appropriate drug selection and dosing). To date, there have been many drugs that have been found to have gene-drug interactions. There are currently more than 250 US Food and Drug Administration-approved drugs that include genetic information on their drug label.¹¹ Of these, current guidelines have clinically actionable recommendations for more than 45 drugs based on drug-gene interactions.¹² Drugs with recommendations cover many therapeutic classes that are commonly prescribed, including analgesics, antidepressants, antipsychotics, and antihypertensives.^11,13

One approach to PGx testing is using a preemptive, panel-based PGx test. With this type of testing, multiple genes and genetic variants are tested at the same time. An advantage of this approach is that the results are available to guide future medication use, and a patient can have results for their lifetime. Although most studies have only examined the benefits of testing for single genetic variants at a time, there is a growing body of evidence that panel-based PGx testing is beneficial.¹⁰ One study found that there was a significant decrease in hospitalizations and emergency department visits over a 4-month period for patients who had panel-based PGx testing performed, as compared with patients who did not have PGx testing performed.¹⁴ This study found that the potential cost-savings were on average $218 for patients who received the PGx test. Additionally, another study found that panel-based PGx testing can result in cost-savings by reducing the use of ineffective drugs, as well as led to decreases in overall health care utilization.¹⁵

Despite the long-term utility of this type of PGx testing, current adoption has been limited.^16-19 Reasons contributing to slow adoption include lack of payer coverage and lack of guidance for providers on how to use PGx test results.^18,19 To overcome these barriers, Navitus Health Solutions partnered with a third-party precision medicine company to create a PGx program that clients could provide to their members.

Methods

STUDY POPULATION

Our study population included members of a health plan who were either commercially insured or insured through Medicare. Members of the health plan were determined to be eligible for PGx testing if they were taking 2 or more medications with known gene guidance for at least 90 of the last 180 days. A list of medications with known PGx guidance was provided by the third-party precision medicine company. A report was run in December 2019 to generate a list of eligible members. Members who participated in PGx testing were in the intervention group and members who chose not to participate in the PGx testing, but who were eligible to participate, were in the control group. Members were included in the analysis if the letter to their provider with the results of the PGx testing was mailed between January 2020 and May 2020. Members were excluded from either group if they were not continuously enrolled through the entire study period.

STUDY PROGRAM AND DESIGN

An initial round of letters inviting eligible members to participate in PGx testing was sent to the individuals who met the inclusion criteria in December 2019. A second round of follow-up letters was sent in February 2020. Follow-up letters were sent to eligible members who had not yet participated in the program. If a member chose to participate, they were sent a PGx test kit from the third-party vendor. The member performed a cheek swab and returned the completed kit to the third-party vendor who then conducted the panel-based PGx test. Results were provided to members through a genetic consulting session with a certified genetic counselor. PGx test results and analysis of a member’s current medication regimen, including suggested medication changes based on gene-drug interactions, were mailed to providers. The types of recommendations that could be given were monitor for adverse events, monitor for lack of efficacy, consider dose increase, consider dose decrease, consider discontinuation, or consider an alternate drug. When alternative medications were recommended, the alternative medication was on the same or lower tier so that there would not be an increased cost for the plan or member. To determine if providers adopted the suggested medication changes, pharmacy claims data were retrospectively reviewed for the 4-month period preceding and the 4-month period following when the recommendations were provided. Medications were compared to determine if there were changes in the member’s medication regimen. The date the provider letter was mailed was used as the index date.

OUTCOMES

The primary outcome of this study was to determine if changes in medications, based on the results of a PGx test, were adopted by providers. Recommendations were considered accepted by providers if at any point during the 4 months after the provider letter was sent the specific recommended change was made, as determined by pharmacy claims data. Discontinued medications were defined as medications that were stopped and there were no claims during the 4-month period after the provider letter was sent. Dose changes (dose increase or dose decrease) were defined as medications that have a change in average daily dose. This was calculated based on strength of medication, quantity, and the days supply. An alternate medication was defined as when one medication is discontinued and another medication that can be used for the same indication is started.

Other outcomes that were examined were the proportion of members in the intervention group who received recommendations, the distribution of the type of recommendations that were made, and the drug classes with the most recommendations.

STATISTICAL ANALYSIS

All data were extracted from the Pharmacy Benefits Manager database and checked for accuracy before being transcribed to an IBM SPSS v25v data base. Descriptive statistics was used to describe the overall characteristics of the study population, the classes of medications with the most recommendations, and the type and number of recommendations that were made and accepted. These groups were described with means and frequencies and were compared with independent t-tests or chi-square tests.

All data were fully anonymized by Navitus Health Solutions before they were sent to the statistician. This study was approved by the SSM Health Wisconsin Institutional Review Board. Patient consent was not required.

Results

PATIENT CHARACTERISTICS

Of the 101 members included in the analysis, 50 were in the intervention group and 51 were in the control group. Members in the intervention group were, on average, aged 58 years (SD, 12.2) and were taking an average of 6.8 medications (SD, 3.4). Members in the control group were, on average, aged 62 years (SD, 12.7) and were taking an average of 7.4 medications (SD, 3.1). In total, members in the intervention group were taking 352 medications prior to testing and members of the control group were taking 376 medications (Table 1). Of the 352 medications that were being taken, 165 were medications with documented PGx guidance. On average, members in the intervention were taking an average of 3.3 (SD 1.2) medications with known PGx guidance. There were no significant differences between baseline characteristics in the intervention and control groups.

TABLE 1.

Baseline Demographics for the Intervention and Control Groups

	Intervention (N = 50)	Control (N = 51)
Female, n (%)	26 (52)	23 (45)
Male, n (%)	24 (48)	28 (55)
Age, mean (SD), years	58.0 (12.2)	61.8 (12.7)
Total number of medications	342	376
Total number of medications with PGx guidance	165	152
Average number of medications per member, mean (SD)	6.8 (3.4)	7.4 (3.1)
Average number of medications with PGx guidance per member, mean (SD)	3.3 (1.2)	3.0 (1.1)

PGx = pharmacogenomics.

PGX RECOMMENDATIONS

Of the 165 medications with PGx guidance, 103 (62.4%) had no recommended changes based on individuals’ PGx test results and 62 (37.6%) had possible changes to optimize therapy. For the 62 medications with recommended changes to optimize therapy, more than 1 recommendation could be made for the medication; 80% of medications with recommendations had more than 1 recommendation given (ie, monitor for adverse drug events and consider lower dose if appropriate). Table 2 shows the number of each type of recommendation given along with the recommendations that were accepted. The most common recommendations that were given were to consider discontinuation of the medication or consider an alternate medication that was suggested (N = 36), if clinically appropriate. The next most common recommendation (N = 26) was to consider a dose decrease of the medication that was currently being used. When recommendations were given, a change was made to the medication 27.4% (17 of 62) of the time, and 19.3% of the time (12 of 62), the specific recommendation was adopted. There was a statistically significant difference between the number of medication changes in the PGx group and the control group (P = 0.024).

TABLE 2.

Distribution of the Type of Recommendations That Were Made and Accepted for Medications With PGx Guidance

Type of recommendation	Number of recommendations made	Number of recommendations that were accepted
No recommended change	103	N/A
Monitor	56	2
Increase dose	8	2
Decrease dose	26	1
Discontinuation or alternate drug	36	7

N/A = not applicable.

The drug classes with the highest number of recommendations were antidepressants, antihyperlipidemics, and proton pump inhibitors (Supplementary Table 1 ^{(12.6KB, pdf)} , see Supplementary Table 2 ^{(12.6KB, pdf)} for examples of drugs with PGx recommendations, available in online article).

Discussion

This study examined how giving providers specific medication optimization recommendations, based on a panel-based PGx test, affected members’ medication regimens. This analysis examined the utility of using PGx to optimize a member’s current medication regimen. For members who underwent PGx testing, 71% received at least 1 recommendation for medications they were currently taking. Some members did not receive recommendations because in some cases, even if a medication had PGx guidance, the individual was a normal metabolizer for the gene that affects the drug.

The most common recommendation given was to consider an alternate drug, if appropriate. These recommendations were almost always accompanied by a recommendation to consider a dose increase or a dose decrease. Because recommendations were given for medications that were currently being taken, in many circumstances, more than 1 type of recommendation was given to provide prescribers flexibility. An alternate drug was frequently given as an option because the pharmacists writing the recommendations were not familiar with the whole clinical picture of the member and so wanted to give the provider options if a dose decrease or increase was not clinically appropriate. For example, the most common recommendation was “Due to increased exposure, monitor for adverse drug events and consider dose decrease or alternative medication, if appropriate.” Specific dose reductions and alternates were given depending on the drug. When alternate medications were suggested, the PGx test results were used to help guide the selection. Additionally, the alternative medication was on the same or lower tier so that there would not be an increased cost for the plan or member.

We found that when recommendations were made, there was a change made to the medication about one quarter of the time. Of these changes, the majority (71%) followed the recommendation that was given. There were statistically more changes in medications for members who received PGx testing, as compared with members who did not receive testing, which indicates that the testing likely prompted medication changes. Both the number of recommendations accepted and the number of changes to medications with recommendations are important to consider because there are many factors that influence how an individual responds to a medication, and PGx is just one factor. In the cases in which there was a change to a medication, but it was not the suggested change, it is possible that sending PGx letters to providers could have prompted them to have a discussion with their patients about their medications, which could ultimately have led to a change.

Recommendations for an alternate drug was the most commonly accepted type of recommendation. Although the exact reasoning is unknown, depending on the situation or the dose of the medication that was currently being used, a change in medication might have been the most appealing option for a patient if they were not doing well on their current medication.

LIMITATIONS

This study has some limitations to consider. First, provider letters were sent to providers between January 2020 and May 2020, which was during the beginning of the coronavirus disease 2019 (COVID-19) pandemic in the United States. During this time, many providers’ offices were closed or operating at limited capacity. As a result, providers may have been less likely to make changes to a patient’s medications unless it was urgent. This could result in providers delaying changes and fewer PGx recommendations being adopted. Additionally, for this study, we used a 4-month time frame to determine if any changes were made to medication regimens. This time frame, especially in the light of COVID-19, could have been too short for changes to be made.

Second, this study examined the benefit of using PGx test results to optimize current medication regimens. Some of the benefit of using a panel-based PGx test is that the results can be used for future medication prescribing. Because using the PGx test results to guide future medication decisions was not examined during this study, the full benefit of this type of PGx test was not captured. As they become available, long-term data will be an examined aspect of the program.

Third, it is unknown whether medication changes were made because of suggestions that were mailed from the pharmacists to the prescribers or by random chance. Although we determined that there were statistically more changes to medications in the intervention group, as compared with the control group, we did not survey the providers to know definitively that the changes were the result of the recommendations.

Conclusions

Giving providers specific medication recommendations based on a panel-based PGx test resulted in changes to members’ medication regimens. As what we know about PGx continues to grow, the utilization and demand for using PGx to guide medication selection will continue to increase. Understanding and accurately communicating how to use this information to optimize medication regimens will benefit patients, providers, and payers.