Combinatorial pharmacogenomic guidance for ... - BioStatShed.com

Current Medical Research & Opinion

Curr Med Res Opin Downloaded from informahealthcare.com by American University of Beirut on 07/30/15 For personal use only.

0300-7995 doi:10.1185/03007995.2015.1063483

2015, 1–11

Article FT-0207.R1/1063483 All rights reserved: reproduction in whole or part not permitted

Original article Combinatorial pharmacogenomic guidance for psychiatric medications reduces overall pharmacy costs in a 1 year prospective evaluation

Joel G. Winner Joseph M. Carhart C. Anthony Altar Assurex Health, Mason, OH, USA

Seth Goldfarb

Abstract Objectives: The objective of this project was to determine pharmacy cost savings and improvement in adherence based on a combinatorial pharmacogenomic test (CPGx*) in patients who had switched or added a new psychiatric medication after having failed monotherapy for their psychiatric disorder.

Express Scripts, St. Louis, MO, USA

Josiah D. Allen Assurex Health, Mason, OH, USA

Gabriela Lavezzari Kelly K. Parsons Express Scripts, St. Louis, MO, USA

Andrew G. Marshak Assurex Health, Mason, OH, USA

Susan Garavaglia Express Scripts, St. Louis, MO, USA

Research design and methods: The prospective project compared 1 year pharmacy claims between a GeneSight CPGx guided cohort and a propensity-matched control group. Patients were project eligible if they augmented or switched to a different antidepressant or antipsychotic medication within the previous 90 days. Following the medication switch or augmentation, pharmacogenomic (PGx) testing was offered to each patient’s treating clinician. Pharmacy claims were extracted from the Medco pharmacy claims database for each patient (n ¼ 2168) for 1 year following testing and compared to a 5-to-1 propensity-matched treatment as usual (TAU), standard of care control group (n ¼ 10,880). Main outcome measures: Total pharmacy spend per member per year; adherence.

Bryan M. Dechairo Assurex Health, Mason, OH, USA Address for correspondence: Joel G. Winner MD, 2595 Canyon Blvd, Ste 100, Boulder, CO 80302, USA. Tel: +1 720 920 9174; Fax: +1 720 920 9307; [email protected] Keywords: Adherence – Combinatorial pharmacogenomic testing – Mental health – Pharmacogenomics – Pharmacy spend

Results: Patients who received PGx testing saved $1035.60 in total medication costs (both CNS and non-CNS medications) over 1 year compared to the non-tested standard of care cohort (p ¼ 0.007). PGx testing improved adherence compared to standard of care (PDCCPGx ¼ 0.11 vs PDCTAU ¼ 0.01; p50.0001). Pharmacy cost savings averaged $2774.53 for patients who were changed to a CPGx congruent medication regimen, compared to those who were not (p50.0001). Conclusions: PGx testing provides significant ‘real world’ cost savings, while simultaneously improving adherence in a difficult to treat psychiatric population. Limitations of this study include the lack of therapeutic efficacy follow-up data and possible confounding due to matching only on demographic and psychiatric variables.

Accepted: 14 June 2015; published online: 22 July 2015 Citation: Curr Med Res Opin 2015; 1–11

Introduction Direct treatment costs for mental illness far exceed those for diabetes and hypertension and lag only behind cardiovascular disease, traumatic injury, *CPGx is a trade name of Assurex Health Inc.

! 2015 Informa UK Ltd www.cmrojournal.com

Combinatorial pharmacogenomic medication cost savings Winner et al.

1



2015

and cancer1. Indirect treatment costs for mental illness are staggering, with major depressive disorder (MDD) responsible for the highest disability costs among all major illnesses2. The National Institute of Mental Health reports that annual direct and indirect costs for depression is $200 billion3, of which treatment-resistant depression comprises $64 billion4. Annual medication costs of $30.3 billion are spent on psychiatric medications1. One in four adults suffer from a mental illness in any given year4. MDD is the most common among these illnesses with a 1 year prevalence of 6.7%4. Many individuals suffering from MDD do not receive treatment because of social stigma, financial outlay, and limited access to healthcare4. Of those who do pursue treatment, twothirds do not achieve full remission. Instead, they often end up on a pharmacologic odyssey requiring multiple failed medications to ultimately find a medication with a favorable risk/benefit balance5. This perpetuation of treatment-resistant depression results in greater loss of work productivity and disability, and 70% higher annual medical costs than for treatment-responsive patients6. The economic and societal costs of MDD derive from the relative early age of onset, chronicity of illness, and poor treatment outcomes7. Intermittent and extended periods of disability, unemployment or under-employment, and increased direct treatment costs can span the life of the individual2. Some clinicians seeking to positively alter this trajectory and improve outcomes for patients with mental illness have recently incorporated PGx guided treatment into their practice8–10. PGx identifies individual genetic differences in pharmacokinetic (PK) genes involved in the absorption, distribution, metabolism and elimination of medications, and pharmacodynamic (PD) genes involved in the mechanism of action of medications. However, given the lack of genetic education11, the multitudinal genomic variants, and their vast array of interactions with medications12, it is daunting for clinicians to utilize PGx in practice. To address this implementation challenge, the GeneSight Psychotropic test was developed to provide clinicians with a suite of validated PK and PD genes yielding a composite phenotype for each patient that is applied to the known pharmacology of each psychiatric medication. This combinatorial pharmacogenomic (CPGx) approach uniquely accounts for multiple metabolic pathways and mechanisms of each medication, resulting in significantly greater predictive power for patient outcomes compared to single gene approaches13. Based on a patient’s precise composite phenotype, the GeneSight CPGx process stratifies antidepressant and antipsychotic medications into three color-coded categories: little or no gene–drug interaction (green, ‘use as directed’), moderate gene–drug interaction (yellow, ‘use with caution’), and severe gene–drug interaction (red, ‘use with increased caution and with more 2


frequent monitoring’). Further information detailing specifics of each gene–drug interaction and dosing recommendations are provided through footnotes14–16. In three prospective clinical trials, GeneSight CPGx guided treatment has demonstrated clinical utility with 71% greater symptom improvement relative to treatment as usual (TAU)14–16. In a 1 year retrospective healthcare utilization study, GeneSight testing predicted a greater number of general medical visits, total healthcare visits including psychiatric visits, medical absence days, and disability claims over 1 year for patients taking red versus green and yellow category medications17. To extend these findings, the current project was conducted in collaboration with a large pharmacy benefits manager (PBM) to assess whether GeneSight testing improved medication utilization and lowered costs in individuals prescribed psychiatric medications across multiple US practice settings.

Patients and methods Project design We compared pharmacy claims over 1 year between a prospectively generated cohort of CPGx tested subjects (n ¼ 2168) and a 5-to-1 propensity-matched control group (n ¼ 10,880). Patients were eligible for either group if they 1) were newly starting an antidepressant or antipsychotic medication (i.e., 180 day look back for no previous prescription [Rx] record), 2) were augmented or switched to a different antidepressant or antipsychotic (index) medication, as evidenced by a new medication Rx within a 90 day window of the last Rx for the initial medication, and 3) maintained continual pharmacy benefits eligibility from 180 days prior to the initial Rx to the date of the first Rx for the index medication. Once the patient was prescribed an index medication, the prescribing clinician was contacted. If the treating clinician authorized GeneSight testing, patient consent was obtained and a buccal swab was provided to the patient for DNA collection. GeneSight results were made available to the patient’s clinician within three business days of sample receipt. A TAU group was selected from a pool of approximately 65 million eligible plan members. Patients were propensity matched on gender, age, index CNS medication, primary CNS diagnosis, and date of project enrollment (Table 1). Pharmacy claims from the PBM database were extracted for each patient for 180 days prior to the index medication initial Rx (index date) (i.e., the ‘pre-test’ period) to 365 days after the index date (i.e., the ‘posttest’ period). The first new GeneSight panel medication prescription after the index medication was operationally defined as the ‘incident’ or third medication, as this fill was www.cmrojournal.com ! 2015 Informa UK Ltd


Table 1. Propensity score matching characteristics. TAUa

GeneSight

69.2 30.8

69.4 30.6

51 (15)

51 (17)

15.7 16.7 33.4 4.9

15.7 16.5 33.8 4.7

29.3 21.1 5.0 0.9 13.5 3.9 0.2 13.1

29.3 21.0 5.3 0.7 13.6 3.7 0.2 13.2


Covariate Gender Female (%) Male (%) Age Mean (years SDc) Index Neuropsychiatric Medicationd Antipsychotics (%) SNRIse (%) SSRIsf (%) Tricyclicsg (%) Miscellaneous (%) ICD-9 Primary Diagnosis Anxiety Disorder (%) Bipolar Disorder (%) Schizophrenia (%) Depression (%) Other Depressive (%) Conduct Disorder (%) Personality Disorder (%)

p valueb 0.87 0.92 0.997

0.98

a

Treatment as usual. p values were calculated using logistic regression models. c Standard deviation. d Index medication refers to the first GeneSight panel antidepressant or antipsychotic medication that served as a replacement for or an addition to a medication in a patient’s regimen. e Serotonin–norepinephrine reuptake inhibitor. f Selective serotonin reuptake inhibitor. g Tricyclic class of antidepressants. b

the first opportunity for the prescriber to make a CPGx informed medication decision within the GeneSight group. Antidepressants and antipsychotics eligible for inclusion were limited to the 26 medications covered by the GeneSight Psychotropic panel (panel medications) at the time of project commencement (Figure 1a). All claims analyzed for the project spanned the time between September 2011 and December 2013. All reversals of Rx fills were identified and excluded from analyses. Medication costs and associated metrics were tabulated during the pre- and post-test periods for both groups. Total medication costs for each patient were derived via the summation of 1) the net amount billed to the payer of each patient, 2) the flat dollar amount each patient incurred for their respective co-pay, and 3) the deductible amount that each member must have satisfied before insurance coverage began for each respective medication.

Between group comparisons Drug spend and polypharmacy The total drug spend per member per year (PMPY) was calculated for both groups across the 365 day post-test period and annualized across the 180 day pre-test period. The within group pre-to-post changes in PMPY Rx amount and spend were calculated. Between group ! 2015 Informa UK Ltd www.cmrojournal.com

2015

differences in pre-to-post change in PMPY Rx amount and spend were compared. Adherence and discontinuation Adherence, discontinuation, and time to discontinuation for index and incident medications were calculated in the post-test period for patients in both groups. Adherence was calculated using the proportion of days covered (PDC) methodology18. The PDC ratio is a commonly used pharmacy-claims-based metric that gauges the number of days each patient has a medication in his or her possession during a specified period of time. Discontinuation was defined as a 45 day or greater interval in refills for the incident or index medication after the days supplied from the previous Rx had elapsed. Statistical analysis As parametric modeling assumptions were upheld for comparisons between groups with respect to pharmacy costs and number of medications (i.e., polypharmacy), independent t-tests were used to model these outcomes between groups from the pre-test to the post-test period. Parametric modeling assumptions were not upheld for statistical comparisons between tested and non-tested groups with respect to adherence, rates of discontinuation, and time to discontinuation (TTD) or for comparisons within each group with respect to pharmacy costs and number of medications. Therefore, Wilcoxon rank-sum tests were used to model adherence to index and incident medications and TTD between groups and the same tests were used to model pharmacy costs and number of medications within groups from the pre-test to the post-test period. McNemar’s tests were used to compare differential rates of adherence to and discontinuation of index and incident medications within each group. Chi-square tests were used to model rates of discontinuation between groups.

Within GeneSight group comparisons Definition of congruence Although the GeneSight report was made available to treating clinicians of all patients in the GeneSight group, clinicians may or may not have made medication changes that were congruent with their patient’s CPGx results. Prior studies have shown that individuals who remain on a genetically misappropriated ‘red category’ medication experience poorer efficacy and excess healthcare utilization14–17. To replicate and extend these findings, further analyses were conducted within the GeneSight group as a function of congruence with the GeneSight report (Table 2). A genetically congruent decision occurred when a patient’s most severely categorized medication was yellow and/or green in the last 90 days of Combinatorial pharmacogenomic medication cost savings Winner et al.

3


2015

Antidepressants


(a) USE AS DIRECTED

USE WITH CAUTION

citalopram (Celexa®) desvenlafaxine (Pristiq®) escitalopram (Lexapro®) fluvoxamine (Luvox®) selegiline (Emsam®) sertraline (Zoloft®)

duloxetine (Cymbalta®) [1] mirtazapine (Remeron®) [1] trazodone (Desyrel®) [1]

USE WITH INCREASED CAUTION AND WITH MORE FREQUENT MONITORING amitriptyline (Elavil®) [6] bupropion (Wellbutrin®) [6] clomipramine (Anafranil®) [6] desipramine (norpramin®) [6] fluoxetine (Prozac®) [6] imipramine (Tofranil®) [6] nortriptyline (Pamelor®) [6] paroxetine (Paxil®) [6] venlafaxine (Effexor®) [6]

Antipsychotics USE AS DIRECTED

USE WITH CAUTION

quetiapine (Seroquel®) ziprasidone (Geodon®)

clozapine (Clozaril®) [1] olanzapine (Zyprexa®) [1] risperidone (Risperdal®) [1]

USE WITH INCREASED CAUTION AND WITH MORE FREQUENT MONITORING aripiprazole (Abilify®) [6] haloperidol (Haldol®) [6] perphenazine (Trilafon®) [6]

[1]: Serum level may be too high, lower doses may be required. [6]: Use of this drug is associated with an increased risk of side effects.

Percentage of patients

(b) 60 50 40 30 20 10 337

558

823 743

502

Pre- Last test 90 days

Pre- Last test 90 days

361

0 Pre- Last test 90 days

Figure 1. (a) Sample GeneSight Psychotropic combinatorial report for an individual patient at the time of testing. Medications characterized in the yellow and red categories are footnoted to describe the nature of the pharmacokinetic and/or pharmacodynamic gene–drug interaction(s) (e.g., ‘serum levels may be too low, higher doses may be required,’ or ‘serum level may be too high, lower doses may be required’). (b) Distribution of patients defined by their most severely categorized medication during the pre-test and last 90 days of post-test period. Numbers in columns indicate sample size.

the post-test period. A genetically incongruent decision was defined as the subject taking a prescribed red category medication in the last 90 days. Total drug spend by congruence Differences in total drug spend during the 365 day post-test period were calculated for the GeneSight congruent and 4


incongruent subgroups and were then further stratified by medication therapeutic chapter, patient diagnosis, and prescribing physician specialty. Medications were stratified according to therapeutic application (i.e., therapeutic chapter index) within the formulary reference guide available from the PBM at the time of project commencement19. For diagnosis stratification, patients were grouped www.cmrojournal.com ! 2015 Informa UK Ltd


2015


Table 2. Method of determining congruence with GeneSight test results. Most severe med prior to index date

Most severe med during last 90 days of follow-up

Red Red Green Yellow Green Yellow Red Yellow Green

Green Yellow Green Green Yellow Yellow Red Red Red

Category

1 1 2 2 2 2 3 4 4

Congruence with GeneSight test (dichotomous)

Congruence with GeneSight test (ordinal)

Congruent Congruent Congruent Congruent Congruent Congruent Incongruent Incongruent Incongruent

Most Congruent Most Congruent Congruent Congruent Congruent Congruent Incongruent Most incongruent Most incongruent

Table 3. ICD-9 codes utilized to develop each diagnostic category. Anxiety Disorders

Depressive Disorders

Bipolar Disorders

Psychotic Disorders

300, 300.0, 300.00, 300.01, 300.02, 300.09, 300.09, 300.1, 300.10, 300.11, 300.12, 300.13, 300.14, 300.15, 300.16, 300.19, 300.2, 300.20, 300.21, 300.22, 300.23, 300.29, 300.3, 300.4, 300.5, 300.6, 300.7, 300.8, 300.81, 300.82, 300.89, 300.9

296.2, 296.20, 296.21, 296.22, 296.23, 296.24, 296.25, 296.26, 296.3, 296.30, 296.31, 296.32, 296.33, 296.34, 296.35, 296.36, 296.82, 311

296.0, 296.00, 296.01, 296.02, 296.03, 296.04, 296.05, 296.06, 296.1, 296.10, 296.11, 296.12, 296.13, 296.14, 296.15, 296.16, 296.4, 296.40, 296.41, 296.42, 296.43, 296.44, 296.45, 296.46, 296.5, 296.50, 296.51, 296.52, 296.53, 296.54, 296.55, 296.56, 296.6, 296.60, 296.61, 296.62, 296.63, 296.64, 296.65, 296.66, 296.7, 296.8, 296.80, 296.81, 296.89, 296.9, 296.90, 296.99

295, 295.0, 295.00, 295.01, 295.02, 295.03, 295.04, 295.05,295.1, 295.10, 295.11, 295.12, 295.13, 295.14, 295.15, 295.2, 295.20, 295.21, 295.22, 295.23, 295.24, 295.25, 295.3, 295.30, 295.31, 295.32, 295.33, 295.34, 295.35, 295.4, 295.40, 295.41, 295.42, 295.43, 295.44, 295.45, 295.5, 295.50, 295.51,295.52, 295.53, 295.54, 295.55, 295.6, 295.60, 295.61, 295.62, 295.63, 295.64, 295.65, 295.7, 295.70, 295.71, 295.72, 295.73, 295.74, 295.75, 295.8, 295.80, 295.81, 295.82, 295.83, 295.84, 295.85, 295.9, 295.90, 295.91, 295.92, 295.93, 295.94, 295.95

by primary or co-primary ICD-9 diagnosis: anxiety disorders, depressive disorders, bipolar disorders, psychotic disorders, and other ICD-9 diagnosis (Table 3). Patients were stratified according to the specialty of the treating clinician20 who prescribed the index medication and ordered GeneSight (i.e., psychiatrist, non-psychiatrist). One-way independent ANOVAs were used to model pharmacy costs as a function of each of the above defined independent variables, as parametric modeling assumptions were upheld. The Sidak correction was employed using the formula 1 (1 )1/n where n is the number of independent tests and is the nominal level of statistical significance (i.e., 0.05) and all reported p values have been adjusted for multiple testing. All data management and statistical analyses were conducted using SAS version 9.321. ! 2015 Informa UK Ltd www.cmrojournal.com

Results Subject attrition is described in Figure 2. Of the initial 2168 PGx tested subjects, two did not have complete pre-test claims data, and 34 did not complete post-test claims data. Of the 2132 patients entering the post-test period, 80 were no longer on a GeneSight panel medication, 330 lost PBM eligibility, and 60 were no longer on a GeneSight panel medication during the last 90 days, leaving 1662 for analyses.

Case–control analyses Average medication costs PMPY increased by $1725.24 from the pre-test period to the study end in the TAU group (p50.0001), which exceeded the mean increase of Combinatorial pharmacogenomic medication cost savings Winner et al.

5


2015

N = 2168 initial subject pool receiving GeneSight test


N = 2166 1 subject not in refreshed PRE_RX data set 1 patient’s GeneSight panel medications were all reversals

N = 34 patients lost to follow-up between pre-test and post-test period

N = 2132 patients entering prospective 365 days period

N = 2052 patients entering prospective 365 day period who were also prescribed a GeneSight panel medication

N = 1722 patients eligible 365 days after GeneSight test date

N = 1662 eligible patients on a GeneSight panel medication 365 days after GeneSight test date

Figure 2. Subject attrition.

$689.64 PMPY in the GeneSight group (p50.0001). Thus, patients who received GeneSight testing saved $1035.60 in total annual medication costs compared to TAU patients (p ¼ 0.0007) (Figure 3a). Total annual medication savings were $714.24 (69%) for non-CNS medications and $321.36 (31%) for CNS medications. An annual decrease of 0.186 medications was observed for the GeneSight group compared to TAU (p50.0001), due to a mean increase of 1.07 medications PMPY in the TAU group from the pre-test to the post-test period (p50.0001), compared to a mean increase of 0.88 medications PMPY in the GeneSight group (p50.0001). Within the GeneSight group, the PDC ratio for the index medication was 0.63, which increased to 0.74 for the incident medication (p50.0001), resulting in an increased adherence rate of 0.111. This increase of 0.111 6


exceeded the 0.01 decrease in PDC ratio within the TAU group (0.80 to 0.79, p ¼ 0.09), resulting in a net improvement of 0.123 for the GeneSight group compared to TAU (p50.0001). Within the GeneSight group, 60.9% of patients discontinued their index medication, while 53.3% discontinued their incident medication (7.6% decrease, p50.0001), compared to 40.9% and 41.2% respectively within the TAU group (0.3% increase, p ¼ 0.69). The differential discontinuation rates from index to incident medications in the GeneSight group was 7.9% less than in the TAU group (p50.0001). The mean TTD of the index medication in the GeneSight group was 103 days, while the mean TTD of the index medication in the TAU group was 134 days (p50.0001). The mean TTD of the incident medication for patients in the GeneSight www.cmrojournal.com ! 2015 Informa UK Ltd



(b) $2000

p = 0.0007

$1,800 $1,600 $1,400 > $ 1,035

$1,200 $1,000 $800 $600 $400 $200 $0 TAU

Total Annual Medication Costs (USD)

Annual Change in Drug Spend (USD)

(a)

2015

p < 0.0001

$12,000 $10,000

> $ 2,774

$8,000 $6,000 $4,000 $2,000 $0 Incongruent (n = 361)

GeneSight

Congruent (n = 1301)

Figure 3. (a) Annual change in medication costs in dollars per member per year (PMPY) from pre-test to post-test by treatment group and (b) total annual medication costs within the GeneSight-tested group as a function of provider congruence with report recommendations. TAU ¼ treatment as usual.

and TAU groups was 150 and 152 days, respectively (p ¼ 0.98).

Congruence with the GeneSight report Among the 1662 PGx tested patients, the percentage of those whose most severe category medication during the pre-test period was red or yellow decreased from 30.2% to 21.7% and from 49.5% to 44.7%, respectively, by the last 90 days of the post-test period (Figure 1b). The percentage of patients whose most severe category medication was green during the pre-test period increased from 20.3% to 33.6% by the last 90 days (Figure 1b). The clinicians for 1301 GeneSight guided patients made congruent medication selections, as shown in Table 2, while 361 patients experienced incongruent selections, from the pre-test period compared to the last 90 days of the study (Figure 3b). GeneSight guided patients with congruent medication regimens spent an average of $7289.96 on medications over the post-test period, while patients with incongruent medication regimens spent $10,064.49, resulting in net annual cost savings of $2774.53 for patients with CPGx congruent medication selections (p50.0001, Figure 3b). When congruence was analyzed according to therapeutic chapter, CNS medications accounted for 54.5% ($1512.44, p ¼ 0.0002) of the differential in total savings. Additional savings of $641.01 for anti-neoplastics (p ¼ 0.02), $286.95 for diabetes (p ¼ 0.007), and $145.85 for cardiovascular medications (p ¼ 0.08) were also observed. Of the 1662 PGx tested patients, 328 (19.7%) had an anxiety disorder, 470 (28.3%) had a depressive disorder, 96 (5.7%) had bipolar disorder, 11 (51%) had a psychotic disorder, and 1133 (68.2%) had additional comorbid (mainly non-CNS) diagnoses (Figure 4a). The difference in total drug spend between the congruent and incongruent patients (Figure 4b) showed that congruent patients ! 2015 Informa UK Ltd www.cmrojournal.com

with anxiety disorder saved $6874.69 ($5777.72 vs $12,652.41 F ¼ 19.75, p50.0001), those with depressive disorder saved $3579.81 ($7715.17 vs $11,294.98, F ¼ 7.26, p50.007), those with bipolar disorder saved $4795.23 ($10,979.47 vs $15,774.70, F ¼ 2.26, p ¼ 0.14), and those with additional, mainly non-CNS comorbid diagnoses saved $4056.18 ($7338.34 vs $11,394.52, F ¼ 25.94, p50.0001) more than incongruent patients. PGx tested patients whose treating clinician was a psychiatrist spent $8830.82 annually on total drug spend, while patients of non-psychiatrists spent $7514.95 (F ¼ 5.23, p ¼ 0.02). Non-psychiatrists and psychiatrists made a congruent medication change at similar frequencies, 79.1% and 76.3%, respectively (2 ¼ 1.53, p ¼ 0.22). The resulting cost differential as a function of congruence for patients whose treating clinician was not a psychiatrist was $3360.10 (Figure 5; p50.0001). The same trend was observed for the patients of psychiatrists, but the $1308.00 cost difference as a function of congruence was not statistically significant (p ¼ 0.24).

Discussion The current ‘real-world’ project included only patients who failed initial therapy for their psychiatric condition who typically represent about half of all treated depression patients5, and likely more for bipolar disorder22, anxiety disorders23, and schizophrenia24. Patients whose physicians had the benefit of GeneSight CPGx testing to guide their medication selection saved $1035.60 in total drug spend over 1 year compared to unguided, propensitymatched controls. Given that the inclusion criteria were designed to capture a costly treatment refractory population it is not surprising that, while the change in total drug spend was significantly less in the GeneSight group, both groups still showed increases in their medication spend compared to the prior 180 days. Combinatorial pharmacogenomic medication cost savings Winner et al.

7


2015

Bipolar Disorders n = 96

(a) Anxiety Disorders n = 328

6

Psychotic Disorders n = 11

39

0

1

12

7 0

1

0

0

0

3 19 142 0

2

0 1 0

2 483

3

(b)

0 15 Depressive Disorders n = 470 $16,000

$16,000

$16,000

$14,000

$14,000 p < 0.0001 $12,000

$14,000 $12,000 $10,000

0

16

270

Other ICD-9 Disorders n = 1133

Total Medication Costs (USD)


142

0

0

1

> $ 6,874

p < 0.007

$10,000

$10,000

> $ 3,579

$8,000

$8,000

$8,000

$6,000

$6,000

$6,000

$4,000

$4,000

$4,000

$2,000

$2,000

$2,000

$0

$0 Incongruent Congruent n = 58 n = 270 Anxiety Disorders

p < 0.0001

$12,000 > $ 4,056

$0 Incongruent Congruent n = 99 n = 371 Depressive Disorders

Incongruent Congruent n = 237 n = 896 Other ICD-9 Disorder

Figure 4. (a) Diagnostic categories and their overlaps for GeneSight-tested patients. See Table 3 for list of ICD-9 codes used to define diagnostic category. (b) Total medication costs by diagnosis and stratified by congruence. Venn diagram generated at http://bioinformatics.psb.ugent.be/webtools/Venn/. p values derived using one-way ANOVA models.

Limitations of this study include the lack of therapeutic efficacy follow-up data and possible confounding due to matching only on demographic and psychiatric variables. However, the matched cohort was created using rigorous and well established techniques, resulting in demographic attributes that were exceptionally similar to those of the CPGx group. While psychiatric prescriptions and diagnostic categories were also distributed in a near perfect match, it is impossible to say with certainty that the tested and control cohorts were identical in all other respects. One of the drivers of the drug cost savings might be the decrease in the number of medications in the GeneSight group. Approximately one in five patients in the GeneSight group were on one less medication by the last 90 days of the post-test period compared to the TAU group (p50.0001). Thus, CPGx offers the possibility of improved efficacy14–16 and decreases in pharmacy costs while concomitantly decreasing patients’ exposure to polypharmacy. 8


Clinicians in the GeneSight group made their medication decisions based on mitigating genetic and non-genetic factors. Their use of the GeneSight test was evidenced by 65% increases in prescriptions for green category medications and nearly a third decrease in red medications, relative to the pre-test period (Figure 1b). Furthermore, approximately 80% of physicians in the GeneSight group made a treatment decision congruent with the CPGx report, saving patients $2774.54 in annual total drug spend over those patients who experienced an incongruent decision (p50.0001, Figure 3b). Our data demonstrate an improvement in adherence from the index to the incident medication in the GeneSight group. A similar effect was seen with the rate of, and time to, discontinuation, which for the GeneSight group improved by 2.6 fold over the TAU group. The difference in baseline adherence, with resultant improvement after GeneSight testing14–16 suggests a CPGxdriven improvement in adherence specifically in those www.cmrojournal.com ! 2015 Informa UK Ltd



Total Medication Costs (USD)

$12,000

2015

p < 0.0001

$10,000 > $ 3,360

$8,000

$6,000

$4,000

$2,000 937

113

Congruent

Incongruent

248

364

$0 Incongruent

Congruent Psychiatrist n = 477

Non-Psychiatrist n = 1185

Figure 5. Total costs for medications prescribed by psychiatrists and non-psychiatrists, stratified by congruence. p values derived using one-way ANOVA models. Physician specialty was categorized in accordance with the taxonomy established by the National Uniform Claim Committee (NUCC).

individuals who tend to be less adherent and potentially less responsive to their initial medications. Impressively, this improvement in adherence exists in the context of decreased comparative pharmacy costs. Given the lack of general pharmacogenomic knowledge in primary care25 and psychiatry11 it is possible that the combinatorial and integrative aspects of GeneSight’s combinatorial approach13 contribute to this synergistic effect of improved adherence and decreased cost when compared to other gene-by-gene PGx assays which have demonstrated similar improved adherence, but increased pharmacy costs26. Mental health treatment outcomes correlate with the outcomes of comorbid chronic illnesses such as heart disease27–29, diabetes30–33, and cancer34. Our comparisons of pharmacy costs across groups show that 69% of the $1035.60 in annual pharmacy savings in the PGx tested group were for non-CNS medications. A growing body of literature shows that the appropriate treatment of mental illness improves outcomes of comorbid medical conditions and lowers cost for their treatment28,31. This pattern was found here, where significant annual savings for diabetes ($286.95), oncology ($640.01), and cardiovascular ($168.17) medications were obtained in the CPGx congruent subgroup suggesting that non-CNS pharmacy spend savings might be a consequence of improvement in patients’ psychiatric conditions. Another possible reason for savings in non-CNS medications may be that the clinician evaluated Genesight tested patients’ non-CNS pharmacology in light of the pharmacogenomic information and made cost efficient changes to non-CNS medications. Finally, there is the possibility of improved patient engagement in their care allowing the patient to more actively work with the physician regarding their pharmacologic treatment leading to more streamlined ! 2015 Informa UK Ltd www.cmrojournal.com

medication management and therefore improved nonCNS pharmacologic costs. Our data show that non-psychiatrists who have access to GeneSight information make more CPGx driven congruent decisions and therefore save an additional $1315.87 on drug spend per patient compared to psychiatrists. Given the shortage of psychiatrists in the United States35, and the increasing burden of mental healthcare for nonpsychiatric practitioners36,37, it is encouraging that non-psychiatrists who categorically implemented the GeneSight test engendered this improvement. The current project confirms and expands upon the UHS retrospective healthcare utilization study which included anxiety and depressive disorders11, and the comorbidity between anxiety and depression. However, 71.7% of patients in the present study did not have a primary or co-primary diagnosis of a depressive disorder. Of those who did, patients who were maintained on a congruent medication regimen showed a $3579.81 reduction in pharmacy costs compared to patients on an incongruent medication regimen, while congruent patients with a primary or co-primary diagnosis of an anxiety disorder demonstrated a $6880.69 improvement in pharmacy costs compared to incongruent patients (Figure 4b). As summarized in Figure 3b, the GeneSight cost savings effect increases as a function of congruence with the report, and is beneficial for patients with primary diagnoses of depressive disorders and/or anxiety disorders (Figure 4b).

Conclusion The current study compares an integrated, combinatorial psychiatric pharmacogenomic test (CPGx) to untested Combinatorial pharmacogenomic medication cost savings Winner et al.

9



2015

standard of care in patients who failed monotherapy for their condition. Patients who received pharmacogenomic testing (n ¼ 2168) saved over $1000 in medication costs annually and improved the rate of medication adherence compared to the untested standard of care cohort (n ¼ 10,880). This study is the largest psychiatric CPGx economic outcome study to date. It is also the only economic outcome study of psychiatric pharmacogenomics to show decreased direct pharmacy costs in conjunction with improved adherence and a reduction in medications for co-morbid conditions. The integrative/combinatorial aspects of GeneSight, the congruence of clinical decisions with its recommendations, and improved pharmacy costs provide direct evidence of a CPGx induced cost savings effect. Further increases in cost savings from nonpsychiatrists provide evidence for the applicability of GeneSight in a broad range of clinical practice settings. When assessed together, the efficacy and economic data make a compelling argument for the use of GeneSight testing in the difficult to treat populations of treatmentresistant depression and anxiety.

Transparency Declaration of funding The project was fully funded by Assurex Health. Declaration of financial/other relationships J.G.W., J.D.A., A.G.M. and B.M.D. have disclosed that they are employees of Assurex Health and own stock in the company. J.M.C. and C.A.A. have disclosed that they are employees of Assurex Health. S.Go., G.L., K.K.P. and S.Ga. have disclosed that they are employees of Express Scripts. The CMRO peer reviewer on this manuscript received an honorarium from CMRO for her review work, but has no relevant financial or other relationships to disclose. Acknowledgments The authors gratefully acknowledge Dr. Sharon Frazee PhD MPH for her comments and contribution toward content. Additionally, we acknowledge Kaitlyn Feeney and Marie Taylor for their contribution toward administrative manuscript preparation for submission. Previous presentation: 10th Annual NEI Psychopharmacology Congress, Colorado Springs, CO, 13–15 November 2014; American Psychiatric Association, 167th Annual Meeting, New York, NY, 2–6 May 2014.

References 1. Agency for Healthcare Research and Quality. Total Expenses and Percent Distribution for Selected Conditions by Type of Service: United States, 2011. Medical Expenditure Panel Survey Household Component Data. http://meps.ahrq.gov/mepsweb/data_files/publications/st470/stat470.shtml [last accessed 1 June 2015]

10


2. Mrazek DA, Hornberger JC, Altar CA, Degtiar I. A review of the clinical, economic, and societal burden of treatment-resistant depression: 1996– 2013. Psychiatr Serv 2014;65:977-87 3. National Institute of Mental Health. Annual Total Direct and Indirect Costs of Serious Mental Illness (2002). 29 July 2010. Available at: http://www.nimh. nih.gov/statistics/4COST_TOTAN.shtml [Last accessed 20 October 2014] 4. NAMI. Numbers of Americans Affected by Mental Illness. Mental Illness Facts and Numbers. Available at: http://www.nami.org/factsheets/mentalillness_ factsheet.pdf [Last accessed 20 October 2013] 5. Warden D, Rush AJ, Trivedi MH, et al. The STAR*D Project results: a comprehensive review of findings. Curr Psychiatry Rep 2007;9:449-59 6. Mrazek DA, Hornberger JC, Altar CA, Degtiar I. A review of the clinical, economic, and societal burden of treatment-resistant depression: 1996– 2013. Psychiatr Serv 2014;65:977-87 7. Kessler RC. The costs of depression. Psychiatr Clin North Am 2012;35:1-14 8. Drozda K, Mu¨ller DJ, Bishop JR. Pharmacogenomic testing for neuropsychiatric drugs: current status of drug labeling, guidelines for using genetic information, and test options. Pharmacotherapy 2014;34:166-84 9. Rundell JR, Staab JP, Shinozaki G, et al. Pharmacogenomic testing in a tertiary care outpatient psychosomatic medicine practice. Psychosomatics 2011;52:141-6 10. Kung S, Xiaofan L. The clinical use of pharmacogenomic testing in treatmentresistant depression. Primary Psychiatry 2010;17:46-51 11. Winner JG, Goebert D, Matsu C, Mrazek DA. Training in psychiatric genomics during residency: a new challenge. Acad Psychiatry 2010;34:115-18 12. Conrado DJ, Rogers HL, Zineh I, Pacanowski MA. Consistency of drug–drug and gene–drug interaction information in US FDA-approved drug labels. Pharmacogenomics 2013;14:215-23 13. Altar CA, Carhart J, Allen JD, et al. Clinical validity: combinatorial pharmacogenomics predicts antidepressant responses and health utilizations better than single gene phenotypes. The Pharmacogenom J 2015 Feb 17. [Epub ahead of print]. doi:10.1038/tpj.2014.85 14. Hall-Flavin DK, Winner JG, Allen JD, et al. Using a pharmacogenomic algorithm to guide the treatment of depression. Transl Psychiatry 2012;2:e172 15. Hall-Flavin DK, Winner JG, Allen JD, et al. Utility of integrated pharmacogenomic testing to support the treatment of major depressive disorder in a psychiatric outpatient setting. Pharmacogenet Genomics 2013;23:535-48 16. Winner JG, Carhart JM, Altar CA, et al. A prospective, randomized, doubleblind study assessing the clinical impact of integrated pharmacogenomic testing for major depressive disorder. Discov Med 2013;16:219-27 17. Winner J, Allen JD, Altar CA, Spahic-mihajlovic A. Psychiatric pharmacogenomics predicts health resource utilization of outpatients with anxiety and depression. Transl Psychiatry 2013;3:e242 18. Nau DP. Proportion of Days Covered (PDC) as a Preferred Method of Measuring Medication Adherence. Pharmacy Quality Alliance. Available at: http://www.pqaalliance.org/images/uploads/files/PQA%20PDC%20vs%20% 20MPR.pdf [Last accessed 28 October 2014] 19. Medco. 2012 Formulary Reference Guide. Available at: https://host1.medcohealth.com/art/corporate/medco_formularies.pdf [Last accessed 28 October 28 2014] 20. NUCC. Health Care Provider Taxonomy July 2014. Available at: http://www. nucc.org/index.php?option¼com_content&view¼article&id¼91&Itemid¼131 [Last accessed 28 October 2014] 21. SAS Institute Inc. Base SAS 9.3 Procedures Guide. Cary, NC: SAS Institute Inc., 2011 22. Gitlin MJ, Swendsen J, Heller TL, Hammen C. Relapse and impairment in bipolar disorder. Am J Psychiatry 1995;152:1635-40 23. Mojtabai R, Olfson M. National trends in psychotropic medication polypharmacy in office-based psychiatry. Arch Gen Psychiatry 2010;67:26-36 24. Lieberman, JA, Stroup TS, Mcevoy JP, et al. Effectiveness of antipsychotic drugs in patients with chronic schizophrenia. N Engl J Med 2005;353: 1209-23 25. Haga SB, Burke W, Ginsburg GS, et al. Primary care physicians’ knowledge of and experience with pharmacogenetic testing. Clin Genet 2012;82: 388-94

www.cmrojournal.com ! 2015 Informa UK Ltd



26. Fagerness J, Fonseca E, Hess GP, et al. Pharmacogenetic-guided psychiatric intervention associated with increased adherence and cost savings. Am J Manag Care 2014;20:e146-56 27. Shah AJ, Ghasemzadeh N, Zaragoza-macias E, et al. Sex and age differences in the association of depression with obstructive coronary artery disease and adverse cardiovascular events. J Am Heart Assoc 2014;3: e000741 28. Stenman M, Holzmann MJ, Sartipy U. Relation of major depression to survival after coronary artery bypass grafting. Am J Cardiol 2014;114:698-703 29. Trivedi R, Gerrity M, Rumsfeld JS, et al. Angina symptom burden associated with depression status among veterans with ischemic heart disease. Ann Behav Med 2015;1:58-65 30. Katon WJ, Lin EH, Von korff M, et al. Collaborative care for patients with depression and chronic illnesses. N Engl J Med 2010;363:2611-20 31. Lustman PJ, Clouse RE, Nix BD, et al. Sertraline for prevention of depression recurrence in diabetes mellitus: a randomized, double-blind, placebo-controlled trial. Arch Gen Psychiatry 2006;63:521-9

! 2015 Informa UK Ltd www.cmrojournal.com

2015

32. Pouwer F, Geelhoed-duijvestijn PH, Tack CJ, et al. Prevalence of comorbid depression is high in out-patients with type 1 or type 2 diabetes mellitus. Results from three out-patient clinics in the Netherlands. Diabet Med 2010;27:217-24 33. Dhavale HS, Panikkar V, Jadhav BS, et al. Depression and diabetes: impact of antidepressant medications on glycaemic control. J Assoc Physicians India 2013;61:896-9 34. Stommel M, Given BA, Given CW. Depression and functional status as predictors of death among cancer patients. Cancer 2002;94:2719-27 35. Thomas KC, Ellis AR, Konrad TR, et al. County-level estimates of mental health professional shortage in the United States. Psychiatr Serv 2009;60: 1323-8 36. Jureidini J, Tonkin A. Overuse of antidepressant drugs for the treatment of depression. CNS Drugs 2006;20:623-32 37. Tylee A, Walters P. Underrecognition of anxiety and mood disorders in primary care: why does the problem exist and what can be done? J Clin Psychiatry 2007;68(Suppl 2):27-30


11