Abstract
Purpose
The prevalence of pharmacogenetically actionable medications in advanced cancer patients whose therapy may be optimized with genotype data was determined.
Methods
Patients enrolled in our institutional molecular tumor board observational cohort were included in this study. Collected data included demographics, type(s) of cancer, and outpatient medications. Medications were classified as “pharmacogenetically actionable” if there are Clinical Pharmacogenetics Implementation Consortium (CPIC) therapeutic recommendations for that medication based on the presence of germline variations. The prevalence of pharmacogenetically actionable medications in the study population was determined, and the frequency of opportunities for pharmacogenetic prescribing and adverse event (AE) mitigation were estimated.
Results
In a cohort of 193 patients with advanced cancer, 65% of patients were taking a pharmacogenetically actionable medication. Approximately 10% of the outpatient medications taken by the study population had a pharmacogenetic association. The most common pharmacogenetically actionable medications being used were ondansetron (47%), capecitabine (10%), and sertraline (7%). Using published genetic variation frequencies and AE risk, we conservatively estimated that 7.1% of cancer patients would be eligible for genetic-based medication adjustment, and 101 AEs would be prevented in 10,000 patients genotyped.
Conclusion
Medications with pharmacogenetic associations are used commonly in the advanced cancer patient population. This widespread exposure supports the implementation of prospective genotyping in the treatment of these high-risk patients.
Keywords: Cancer, germline, oncology, pharmacogenetics, pharmacogenomics
KEY POINTS
In a cohort of patients with advanced cancer, 65% of patients were taking a pharmacogenetically actionable medication.
Approximately 10% of the outpatient medications of the study population had a pharmacogenetic association.
The most common pharmacogenetically actionable medications used were ondansetron (47%), capecitabine (10%), and sertraline (7%).
Interpatient differences in efficacy and toxicity are consistently observed with many medication therapies. Demographic, biological, and pharmacological factors that affect therapy outcomes are routinely assessed to minimize these differences. An important variable associated with drug response variation is germline polymorphisms pertaining to drug metabolism, drug transport, and drug targets.1–3 Numerous polymorphisms have been identified that affect the efficacy and toxicity of medication therapy and may contribute to patient outcomes. The Food and Drug Administration (FDA) began incorporating pharmacogenetic information in drug labeling in 2007, and since then approximately 10% of approved drugs contain such information in their labeling.4 Furthermore, an international initiative called the Clinical Pharmacogenetics Implementation Consortium (CPIC) currently provides evidence-based pharmacogenetic dosing guidelines for 36 different drugs.5,6
Improvements in the cost and ease of genotyping has made it feasible to preemptively assess patients’ pharmacogenomic information for variations that may impact their medication therapy. The potential of this approach was investigated by Schildcrout et al.7 at Vanderbilt University Medical Center by retrospectively reviewing the medications of approximately 53,000 patients to identify those with pharmacogenetic associations. The results suggest that a significant portion of patients are exposed to medications with established pharmacogenetic associations, with 65% of patients expected to receive at least 1 pharmacogenetic medication over a 5-year period and more than 10% expected to receive at least 4. Vanderbilt has since launched PREDICT (Pharmacogenomic Resource for Enhanced Decisions in Care and Treatment), which prospectively genotypes patients who are likely to receive pharmacogenetic medications in the next 3 years in order to tailor their drug therapy according to their genetic test results.8 Preliminary data from PREDICT depicting the prevalence of actionable germline mutations in almost 10,000 patients has shown that a significant portion of patients harbor a variation that may affect medication therapy, with 91% of patients exhibiting an actionable variant related to at least 1 of 5 drug–gene interactions (DGIs).9 The clinical utility of preemptively genotyping certain high-risk patient populations was demonstrated by Elliot et al.10 In this study, hospitalized elderly patients taking or initiating pharmacogenetically actionable medications at discharge were randomized to receive pharmacogenomic profiling with therapy adjustment based on the results. Patients who were tested had fewer rehospitalizations and ED visits at 60 days postdischarge compared with those who did not have their pharmacogenomics evaluated. Among the 57 tested patients, 96 genotype-guided therapy recommendations were accepted by clinicians.
The patient population of PREDICT and other similar initiatives, such as CLIPMERGE at Mount Sinai and RIGHT at the Mayo Clinic, primarily focus on patients taking medications for cardiovascular disease.9,11,12 Cancer patients are a high-risk population that may benefit from preemptive genotyping given the high morbidity/mortality of their disease, frequent healthcare exposure, and increased medication utilization.13 In addition to the patient-specific variables that influence medication choice, incorporating knowledge of germline variations to the care of cancer patients could yield improved outcomes similar to those demonstrated by Elliot et al.10 and may represent a new paradigm for precision oncology treatment. However, the extent to which cancer patients take medications with pharmacogenetic associations remains undefined in the literature. The purpose of this study was to assess the prevalence of patients with advanced cancer who are taking pharmacogenetically actionable medications.
Methods
Study population.
This single-center, retrospective, observational study included 193 patients, all of whom were enrolled in our institutional Molecular Tumor Board (MTB) and associated clinical trial (NCT03089554) before January 16, 2018, and were included in the analysis. This study was approved by the official institutional review board in accordance with local regulations. Patients were included if they had a clinically suspected or histologically confirmed solid or hematological malignancy, measurable or immeasurable disease, had undergone or will undergo genetic testing of their tumor, had at least 1 prior therapy, were at least 18 years old, had an Eastern Cooperative Oncology Group performance status of ≤3, and had a life expectancy greater than 3 months. Patients were excluded if they had uncontrolled intercurrent illness, such as active infection or cardiac arrhythmia, that would limit compliance with targeted therapy or if they were pregnant.
Data collection.
Data for each patient were collected from the UK HealthCare electronic health record (EHR) by chart review and manual abstraction and were then deidentified and stored in a system compliant with the Health Insurance Portability and Accountability Act. Collected data included demographics (age at the time of MTB enrollment, sex, race), type(s) of cancer, and outpatient medications. Patients’ outpatient medications were obtained from the first healthcare visit note after enrollment in the MCC MTB and were evaluated for pharmacogenetically actionable medications.
Definition of pharmacogenetically actionable and assessment of DGIs.
Medications were classified as a pharmacogenetically actionable medication if there are CPIC therapeutic recommendations for that medication based on the presence of germline variations.14 The prevalence of patients taking pharmacogenetically actionable medications was determined by dividing the number of patients with at least 1 pharmacogenetically actionable medication on their medication list by the total study population.
The expected prevalence of actionable genotypes was obtained from published minor allele frequencies referenced in the CPIC guidelines.14 Expected frequencies of actionable genotypes not provided in the CPIC guidelines were obtained from other published literature.15,16 For analysis purposes, genotype frequencies of Caucasian patients were used given the demographics of our study cohort. Characterization of the strength of evidence behind potential therapeutic recommendations for patients with a DGI was obtained from the CPIC guidelines, where the strength of each therapeutic recommendation is classified as either “strong”, “moderate,” or “optional” based on preclinical functional data and clinical disease-specific data.
The percentage of patients expected to have a DGI from “strong” or “moderate” therapeutically monitored medications was calculated by multiplying the frequency of patients taking a pharmacogenetically actionable medication by the expected frequency of patients with actionable variations. For example, cytochrome P-450 (CYP) isozyme 2D6 ultra-rapid metabolizers would be expected to have reduced response to ondansetron. Given there were 90 patients receiving ondansetron and the frequency of CYP2D6 ultra-rapid metabolizer is 3.3%, if 90 patients were genotyped, then 3 (90*0.033) of those patients were expected to require a dose modification for this genotype. This was repeated for each of the 11 pharmacogenetically actionable medications included in this analysis and combined to determine the percentage of patients expected to be eligible for a therapeutic adjustment recommended by CPIC based on their currently prescribed medications and pharmacogenetic phenotype. Medications were excluded from this analysis if only 1 patient in the study cohort was taking it (amitriptyline, doxepin, and voriconazole) or if the CPIC recommendations are categorized as “optional” (sertraline and warfarin).
Of the 16 pharmacogenetically actionable medications used by the study cohort, 8 were selected for analysis of the number of possible adverse events (AEs) that could be prevented with effective interventions. Analyses utilized the actionable genotype prevalence estimates obtained from population data as described above.14–16 Medications were excluded from the AE analysis if only 1 patient in the study cohort was taking it (amitriptyline, doxepin, voriconazole), if therapeutic drug monitoring using serum levels is used routinely in clinical practice (warfarin, tacrolimus, phenytoin), if CPIC recommendations are categorized as “optional” (sertraline), or if the actionable genotype frequency was deemed insignificantly low (allopurinol) in our population.
Statistics.
Descriptive statistics were used to characterize patient demographics and assess medication data. The method used to calculate preventable AEs emulated the approach employed by Schildrout et al.7 and is therein detailed further. In general, the number of patients in our cohort using pharmacogenetically actionable medications was combined with literature-based estimates of pharmacogenetic effects, overall overprevalence of AEs associated with medications, and prevalences of genetic risk strata (G0, G1, G2) in order to estimate the number of AEs that might have been prevented had an effective genotyping and mitigation strategy program been utilized.17–25 A sensitivity analysis was conducted to assess the robustness of the results by varying the probability of AEs in a range of ±10%.
Results
In total, 193 patients were enrolled in the MTB from May 10, 2017, to January 10, 2018, and were included in the analysis (Table 1). The mean age of our patient population was 60 years, 47% were male, and 95% were Caucasian. The most common cancer diagnoses were gastrointestinal (22%), lung (22%), and hematologic malignancy (21%). Regarding the prevalence of pharmacogenetically actionable medications, 65% of patients were taking at least 1 pharmacogenetically actionable medication, with 38% taking 1 pharmacogenetically actionable medication, 24% taking 2 pharmacogenetically actionable medications, and 3% taking 3 pharmacogenetically actionable medications.
Table 1.
Patient Demographics (n = 193)
| Characteristic | Value |
|---|---|
| Mean ± S.D. age, yr | 60 (12) |
| Gender, % male (no.) | 47 (91) |
| Race, % Caucasian (no.) | 95 (184) |
| Cancer type, % (no.) | |
| Gastrointestinal | 22 (43) |
| Lung | 22 (42) |
| Hematologic | 21 (41) |
| Gynecologic | 11 (22) |
| Breast | 5 (9) |
| Head and neck | 5 (9) |
| Othera | 14 (27) |
| Number of pharmacogenetically actionable medications, % (no.) | |
| 0 | 35 (67) |
| 1 | 38 (73) |
| 2 | 24 (47) |
| 3 | 3 (6) |
aOther: Adrenal, bladder, bone, brain, Erdheim-Chester, neuroendocrine, prostate, renal, sarcoma, skin, or thyroid.
A total of 2,077 medications were being taken by the study cohort in the outpatient setting, with an average of 11 medications per patient. Of these medications, 9% had a pharmacogenetic association (Table 2). The most common pharmacogenetically actionable medications used were ondansetron (47%), capecitabine (10%), sertraline (7%), simvastatin (5%), and warfarin (5%). Pharmacogenetically actionable medications that only 1 patient in the study cohort was taking were included in the “Other” category. These included amitriptyline, doxepin, and voriconazole. Pharmacogenetically actionable medications used in our cohort encompassed 16 of the 36 total medications with CPIC guidelines at the time of study analysis.
Table 2.
Medication Characteristics
| Characteristic | Value |
|---|---|
| Pharmacogenetically actionable medications,% (no.) (n = 2,077) | 9 (185) |
| Patient usage, % (no.) (n = 193) | |
| Ondansetron | 47 (90) |
| Capecitabine | 10 (19) |
| Sertraline | 7 (13) |
| Simvastatin | 5 (9) |
| Warfarin | 5 (9) |
| Escitalopram | 4 (8) |
| Allopurinol | 4 (7) |
| Citalopram | 4 (7) |
| Tacrolimus | 3 (6) |
| Codeine | 2 (4) |
| Paroxetine | 2 (4) |
| Clopidogrel | 2 (3) |
| Phenytoin | 2 (3) |
| Othera | 2 (3) |
aOther: Amitriptyline, doxepin, or voriconazole.
The prevalence of pharmacogenetically actionable medications was evaluated based on cancer diagnosis (Table 3). Breast cancer patients demonstrated the highest prevalence, with 89% of patients taking a pharmacogenetically actionable medications, followed by patients with head and neck (78%), lung (76%), and gastrointestinal (72%). Patients with hematologic malignancy had the lowest prevalence of pharmacogenetically actionable medications at 54%, followed by gynecologic cancer at 55%.
Table 3.
Prevalence of Actionable Medication by Cancer Type
| Cancer Type | Prevalence, % (no./n) |
|---|---|
| Gastrointestinal | 72 (31/43) |
| Lung | 76 (32/42) |
| Hematologic | 54 (22/41) |
| Gynecologic | 55 (12/22) |
| Breast | 89 (8/9) |
| Head and neck | 78 (7/9) |
The genes and genetic variations associated with pharmacogenetically actionable medications used by more than 1 patient in the study cohort are depicted in Table 4. Each of the genes of interest has been studied in large populations, enabling the estimation of variation frequency based on published data for Caucasian patients. A majority of pharmacogenetic associations of pharmacogenetically actionable medications used by the study cohort were related to CYP isozymes (13 of 16 pharmacogenetically actionable medications). A majority of DGIs associated with pharmacogenetically actionable medications used by the study cohort have a therapeutic recommendation strength classified as “strong” by CPIC (45%). A “moderate” CPIC therapeutic recommendation strength was seen in 39% of DGIs, and 15% of therapeutic recommendations had a strength classified as “optional.”
Table 4.
Estimated Frequency of Actionable Genotypesa
| Drug | Gene | Actionable Condition | Frequency of Actionable Condition | Strength of Recommendation | Consequence of DGI |
|---|---|---|---|---|---|
| Ondansetron | CYP2D6 | Ultra-rapid metabolizer | UM: 0.033 | UM: moderate | UM: decreased efficacy |
| Capecitabine | DPYD | Intermediate metabolizer Poor metabolizer |
IM: ~0.03–0.05 PM: ~0.002 |
IM, activity score 1: strong IM, activity score 1.5: moderatePM: strong | IM: increased toxicity PM: increased toxicity |
| Sertraline | CYP2C19 | Ultra-rapid metabolizer Poor metabolizer |
UM: 0.046 PM: 0.025 |
UM: optionalPM: optional | UM: decreased efficacy PM: increased toxicity |
| Simvastatin | SLCO1B1 | Intermediate function Low function |
IF: 0.11–0.36 LF: 0–0.06 |
IF: strong LF: strong |
IF: increased toxicity LF: increased toxicity |
| Warfarin | CYP2C9 | *2 and *3: decreasedfunction *5, *6, *8, or *11:possible decreasedfunction |
*2: 0.126 *3: 0.071 *5: 0 *6: 0 *8: 0.001 *11: 0.002 |
*2 and *3, non-African: strong*2 and *3, African: moderate*5, *6, *8, or *11, non-African:optional *5, *6, *8, or *11, African:moderate | Delayed time totherapeutic INR |
| Warfarin | CYP4F | *3 (rs2108622 C>T):decreased function | *3: 0.3 | rs2108622, non-African: optional | Delayed time totherapeutic INR |
| Warfarin | VKORC1 | 1639G>A | 1639G>A:0.412 | 1639G>A, non-African: strong1639G>A, African: moderate | Delayed time totherapeutic INR |
| Escitalopram | CYP2C19 | Ultra-rapid metabolizer Poor metabolizer |
UM: 0.046 PM: 0.025 |
UM: moderatePM: moderate | UM: decreased efficacyPM: increased toxicity |
| Allopurinol | HLA-B | *58:01 | *58:01: 0.0001 | *5801: strong | *5801: increased toxicity |
| Citalopram | CYP2C19 | Ultra-rapid metabolizer Poor metabolizer |
UM: 0.046 PM: 0.025 |
UM: moderatePM: moderate | UM: decreased efficacyPM: increased toxicity |
| Tacrolimus | CYP3A5 | Extensive metabolizer Intermediate metabolizer |
EM: 0.072 IM: 0.205 |
EM: strong IM: strong |
Delayed time totherapeutic levels |
| Codeine | CYP2D6 | Ultra-rapid metabolizer Poor metabolizer |
UM: 0.033 PM: 0.061 |
UM: strongIM: moderate PM: strong | UM: increased toxicityIM: decreased efficacyPM: decreased efficacy |
| Paroxetine | CYP2D6 | Ultra-rapid metabolizer Poor metabolizer |
UM: 0.033 PM: 0.061 |
UM: strongPM: optional | UM: decreased efficacy PM: increased toxicity |
| Clopidogrel | CYP2C19 | Intermediate metabolizer Poor metabolizer |
IM: 0.268 PM: 0.025 |
IM: moderatePM: strong | IM: decreased efficacyPM: decreased efficacy |
| Phenytoin | CYP2C9 | Intermediate metabolizer Poor metabolizer |
IM: 0.32 PM: 0.04 |
IM: moderatePM: strong | IM: increased toxicityPM: increased toxicity |
| Phenytoin | HLA-B | *15:02 | *15:02: 0.0001 | *1502, phenytoin-naïve: strong | *1502, phenytoin-naïve:increased toxicity |
aDGI = drug-gene interaction, UM = ultra-rapid metabolizer, EM = extensive metabolizer, IM = intermediate metabolizer, PM = poor metabolizer, IF = intermediate function, LF = low function, CYP = cytochrome P-450, DPYD = dihydropyrimidine dehydrogenase, SLCO1B1 = solute carrier organic anion transporter family member 1B1, VKORC1 = vitamin K epoxide reductase complex subunit 1, HLA-B = major histocompatibility complex, class I, B, INR = International normalized ratio.
Using the aforementioned percentage of patients taking each pharmacogenetically actionable medication and the expected frequencies of actionable pharmacogenetic variations associated with each pharmacogenetically actionable medication, the percentage of patients expected to have a DGI was calculated for each pharmacogenetically actionable medication (Table 5). In total, it was estimated that 7.1% of patients will have a DGI and would therefore be eligible for a pharmacogenetic-based therapeutic adjustment described in the CPIC guidelines. Of these patients, 3.8% were estimated to be eligible for a “strong” therapeutic adjustment, and 3.3% of patients were estimated to be eligible for a “moderate” therapeutic adjustment.
Table 5.
Estimated DGIs and Potential for Therapeutic Adjustment (n = 193)a
| Drug | % Patients Expected to Experience a DGI |
|---|---|
| Ondansetron | 1.6 |
| Capecitabine | 0.5 |
| Simvastatin | 2.1 |
| Escitalopram | 0.3 |
| Allopurinol | 0 |
| Citalopram | 0.3 |
| Tacrolimus | 0.8 |
| Codeine | 0.2 |
| Paroxetine | 0.1 |
| Clopidogrel | 0.6 |
| Phenytoin | 0.7 |
| Total | 7.1 |
| Recommendation | % Patients Expected to be Eligible for Therapeutic Adjustment |
| Strong CPIC recommendation | 3.8 |
| Moderate CPIC recommendation | 3.3 |
| Total | 7.1 |
aDGI = drug-gene interaction, CPIC = Clinical Pharmacogenetics Implementation Consortium.
The number of possible AEs prevented with effective intervention based on genotypic data for certain pre-specified medications is depicted in Table 6. To assist with the interpretation of this table, consider the antiemetic agent ondansetron, which is metabolized by CYP2D6. The probability of emesis (lack of efficacy) after ondansetron administration, demonstrated in a prior study, was 0.455 in ultra-rapid CYP2D6 metabolizers and 0.147 in the rest of the cohort. Using these event probabilities, the probabilities of being a CYP2D6 ultra-rapid metabolizer or not (0.033 and 0.967, respectively), and the percentage of patients taking ondansetron in our cohort of cancer patients (47%), we calculated that dosing based on information from preemptive genotyping might prevent 48 episodes of emesis among ultra-rapid metabolizers per 10,000 patients tested. Overall, of the medication–AE combinations studied here, it is estimated that preemptive genotyping could prevent 101 AEs among 10,000 individuals. Medications with the highest number of preventable AEs included ondansetron (48 AEs prevented), capecitabine (26 AEs prevented), and codeine (10 AEs prevented) among 10,000 individuals tested. A sensitivity analysis was conducted to assess the robustness of the results by varying the probability of AE in a range of ±10%. The sensitivity analysis showed the robustness of the results with a minimum of 92 and a maximum of 111 AEs presented.
Table 6.
Preventable Adverse Eventsa
| Events Preventedin 10,000 Patients | ||||||||
|---|---|---|---|---|---|---|---|---|
| Medication | Adverse Event | Risk Strata(G0, G1, G2) | Risk Stratum Prevalence | Percent on Medication | Event Probability | G1 | G2 | Total |
| Ondansetron | Nausea/ vomiting | (-, UM) | (0.967, 0.033) | 0.47 | (0.147, 0.455) | 48 | 0 | 48a |
| 43 | 0 | 43 | ||||||
| 53 | 0 | 53 | ||||||
| Capecitabine | Grade ≥3 toxicity | (-, IM, PM) | (0.948, 0.05, 0.002) | 0.10 | (0.231, 0.729, 0.729) | 25 | 1 | 26 |
| 22 | 1 | 23 | ||||||
| 27 | 1 | 28 | ||||||
| Simvastatin | Myopathy | (-, IF, LF) | (0.723, 0.255, 0.023) | 0.05 | (0.077, 0.119, 0.135) | 5 | 1 | 6 |
| 5 | 1 | 6 | ||||||
| 6 | 1 | 6 | ||||||
| Escitalopram | Persistent depression | (-, UM, PM) | (0.929, 0.046, 0.025) | 0.04 | (0.125, 0.195, 0.307) | 1 | 2 | 3 |
| 1 | 2 | 3 | ||||||
| 1 | 2 | 3 | ||||||
| Citalopram | Persistent depression | (-, UM, PM) | (0.929, 0.046, 0.025) | 0.04 | (0.125, 0.195, 0.307) | 1 | 2 | 3 |
| 1 | 2 | 3 | ||||||
| 1 | 2 | 3 | ||||||
| Codeine | Hypoanalgesia | (-, PM) | (0.939, 0.061) | 0.02 | (0.115, 0.962) | 10 | 0 | 10 |
| 9 | 0 | 9 | ||||||
| 11 | 0 | 11 | ||||||
| Codeine | Sedation/ adverse events | (-, UM) | (0.967, 0.033) | 0.02 | (0.47, 1.00) | 3 | 0 | 3 |
| 3 | 0 | 3 | ||||||
| 4 | 0 | 4 | ||||||
| Paroxetine | Persistent depression | (-, UM) | (0.967, 0.033) | 0.02 | (0.672, 1.00) | 2 | 0 | 2 |
| 2 | 0 | 2 | ||||||
| 2 | 0 | 2 | ||||||
| Clopidogrel | MI, stroke, death | (-, IM, PM) | (0.715, 0.263, 0.022) | 0.02 | (0.002, 0.005, 0.013) | 0 | 0 | 0 |
| 0 | 0 | 0 | ||||||
| 0 | 0 | 0 | ||||||
| Total | 95 | 6 | 101 | |||||
| 86 | 6 | 92 | ||||||
| 105 | 6 | 111 | ||||||
aNote: There are 3 rows in the last column of table 6. The first row was calculated under the observed event probability and the next 2 rows were calculated under ± 10% of the observed event probability. UM = ultra-rapid metabolizer, EM = extensive metabolizer, IM = intermediate metabolizer, PM = poor metabolizer, IF = intermediate function, LF = low function.
Discussion
In a cohort of patients with advanced cancer, 65% of patients were found to be taking at least 1 pharmacogenetically actionable medication. This is similar to the study by Schildcrout et al.,7 who found that 65% of outpatient primary care patients were taking at least 1 medication with a pharmacogenetic association. The criteria used to define pharmacogenetically actionable medication in our study was more restrictive, however, and only included the 36 medications with pharmacogenetic dosing guidelines by CPIC. The evaluation by Schildcrout et al. included 56 medications with FDA drug labels indicating known pharmacogenetic variation in response, a majority of which do not include dosing recommendations based on patient pharmacogenetic phenotype. Another study seeking to evaluate the prevalence of pharmacogenetic medication use assessed 36.1 million patients whose prescriptions were processed by a large pharmacy benefits manager in 2006.26 The study found that 8.8 million (24.3%) received at least 1 drug with human genomic biomarker information in the FDA label. It is likely that a significantly higher percentage of patients in our cohort would be classified as taking a pharmacogenetically actionable medication with more broad criteria encompassing more medications. Evaluating only medications with pharmacogenetic dosing guidelines by CPIC enhances the clinical applicability of our results, especially when further restricted to medications with strong or moderate recommendations, as the patients identified as taking pharmacogenetically actionable medications would be eligible for evidence-based adjustment to their medication therapy.
The most commonly used pharmacogenetically actionable medications in our cohort were ondansetron (47% of patients), capecitabine (10%), sertraline (7%), simvastatin (5%), and warfarin (5%). These medications are used frequently by cancer patients. Capecitabine is a chemotherapy agent used for a variety of malignancies including colorectal and pancreatic cancers. Sertraline and warfarin are used for cancer-related comorbidities, depression and thrombosis respectively. Simvastatin was the fourth most commonly used pharmacogenetically actionable medication in our study. In contrast, it was the most commonly used pharmacogenetically actionable medication in the study by Schildcrout et al.,7 with 33% of patients in their cohort using it. A possible reason for this difference is the questionable utility of antilipemic agents in cancer patients with poor prognosis using chemotherapy agents with hepatotoxicity risk.27
We estimate that 7.1% of our cohort of advanced cancer patients will be both on a pharmacogenetically actionable medication and have a genotype requiring adjustment. This estimation used the prevalence of pharmacogenetically actionable medications in our study cohort and the expected frequency of actionable pharmacogenetic variations in the literature. The accuracy of published genetic variation frequencies was demonstrated in a study by Van Driest et al.,9 where approximately 10,000 patients were preemptively genotyped for pharmacogenetic variations. The frequency of actionable variations found in their study cohort matched closely with the allele frequencies described in the literature. This similarity validates the use of published genetic variation frequency data to estimate the percentage of patients expected to experience a DGI.
Assuming that adverse drug events (including therapeutic failure and side effects) attributed to genetic variations are preventable, the effective use of pharmacogenetic information could potentially prevent 101 AEs per 10,000 patients genotyped. Adverse drug events are a significant public health burden, causing approximately 1.3 million emergency department visits and about 350,000 hospitalizations each year.28 These results suggest that the use of prospectively collected genotype data might represent an effective approach to prevent avoidable AEs and improve medication outcomes in advanced cancer patients. Two studies have previously estimated AEs prevented by preemptive genotyping in broad patient populations (not cancer exclusive). The aforementioned study by Schildcrout et al.7 estimated preventable AEs for 6 medications: 2 were included in our analysis (simvastatin, clopidogrel), 3 were not used by our study cohort (abacavir, azathioprine, tamoxifen), and 1 was excluded given the use of therapeutic drug monitoring in clinical practice (warfarin). In their cohort of 52, 942 medical home patients, they estimated that 383 events could have been prevented with preemptive genotyping. This equates to 72 AEs prevented per 10,000 patients. The second study evaluated the cost-effectiveness of one-time pharmacogenomic testing for preventing AEs over a patient’s lifetime using a Markov-based Monte Carlo microsimulation model.29 In their calculations, Alagoz et al. used a mean ADR rate of 1.55% and a reduction in AEs due to genetic testing of 17%. This equates to 27 AEs prevented per 10,000 patients. In estimating the number of AEs prevented in advanced cancer patients, only 9 AEs with 8 medications were included, and only the genetic variants for which reliable medication therapy adjustments are described in CPIC guidelines. Other medications and pharmacogenetic variants are likely to have important effects on the incidence of AEs and have yet to be adequately described in the literature. Advanced genotyping technology that simultaneously captures information about multiple variants in multiple genes makes a preemptive genotyping program feasible, and possibly cost-effective. Outpatient preventable medication errors cost approximately $4.2 billion annually.30 With the rapidly decreasing cost of genetic sequencing, preemptive genotyping of certain high-risk patient populations such as cancer patients may be a cost-effective method to prevent AEs and improve patient outcomes.31 The previously discussed Monte Carlo model by Alagoz et al.29 suggests that genetic testing may be cost-effective, but prospective randomized studies are necessary for conclusive evidence.29
Strengths of this study include its “real-life” patient population and conservative estimates. The evaluation of cancer patients treated at our institution increases the validity of the results and facilitates application to clinical practice. Furthermore, previous studies included broader patient populations, not exclusive to those with a cancer diagnosis. Our assessment of the value of pharmacogenomic testing in advanced cancer patients was conservative. In determining the prevalence of pharmacogenetically actionable medications, only medications with pharmacogenetic-based dosing backed by robust evidence described by CPIC were included. Additional medications used frequently by cancer patients are not included in the CPIC guidelines but have pharmacogenetic associations described elsewhere in the literature (i.e., opioids and proton pump inhibitors).32,33 Previous studies have used less-restrictive criteria for pharmacogenetically actionable medications resulting in greater predicted benefit; however, our study was designed to be more conservative and still predicted considerable benefit from preemptive genotyping.
There were several limitations to this study. The first was our use of point prevalence. Our evaluation of pharmacogenetically actionable medications used patients’ medication lists from the day they were enrolled in our institutional MTB. It is possible that patients were previously on a medication that they did not tolerate due to a DGI, or may have been prescribed a pharmacogenetically actionable medication at a later date, thereby underestimating the prevalence of pharmacogenetically actionable medication use in our cohort. An assessment of patient medications over a period of time (ideally beginning at the time of cancer diagnosis) would be a more comprehensive evaluation of patient exposure to pharmacogenetically actionable medications.
Another limitation of this study was the exclusion of intravenous medications. Intravenous medications, including chemotherapy, that patients receive routinely in the outpatient setting are not included in patient medication lists recorded in our institutional electronic medical record and were therefore not evaluated for pharmacogenetic association. Pharmacogenetically actionable intravenous medications were missing from our evaluation, underestimating the prevalence of pharmacogenetically actionable medications in our cohort. An example of this is fluorouracil. While the oral formulation capecitabine was included in our assessment, it is unknown if patients were taking an intravenous fluoropyrimidine in the outpatient setting.
This was a single-center study, limiting generalizability. Further limiting generalizability, genetic variation frequencies for Caucasians were obtained from the literature, given that 95% of the study cohort was Caucasian, thus the results of this study may be less applicable for patient populations with a different racial or ethnic make-up.
Furthermore, differences in institution protocols and formularies may influence the types of medications commonly prescribed. For calculating AEs prevented, we used AE rates derived from research cohorts; however it is expected that real-world rates are likely to be similar or higher given compliance issues and other barriers to optimal care. Lastly, evaluation of patient outcomes and economic modeling was outside the scope of this manuscript.
Conclusion
Medications with pharmacogenetic associations were used commonly in the advanced cancer patient population. This widespread exposure supports the implementation of prospective genotyping in the treatment of these high-risk patients.
Disclosures
The authors have declared no potential conflict of interest.
References
- 1. Evans WE, McLeod HL. Pharmacogenomics–drug disposition, drug targets, and side effects. N Engl J Med. 2003; 348:538–49. [DOI] [PubMed] [Google Scholar]
- 2. Wang L, McLeod HL, Weinshilboum RM. Genomics and drug response. N Engl J Med. 2011; 364:1144–53. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. McLeod HL, Evans WE. Pharmacogenomics: unlocking the human genome for better drug therapy. Annu Rev Pharmacol Toxicol. 2001; 41:101–21. [DOI] [PubMed] [Google Scholar]
- 4. Vivot A, Boutron I, Ravaud P, Porcher R. Guidance for pharmacogenomic biomarker testing in labels of FDA-approved drugs. Genet Med. 2015; 17:733–8. [DOI] [PubMed] [Google Scholar]
- 5. Caudle KE, Klein TE, Hoffman JMet al. Incorporation of pharmacogenomics into routine clinical practice: the clinical pharmacogenetics implementation consortium (CPIC) guideline development process. Curr Drug Metab. 2014; 15:209–17. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Relling MV, Klein TE. CPIC: Clinical Pharmacogenetics Implementation Consortium of the Pharmacogenomics Research Network. Clin Pharmacol Ther. 2011; 89:464–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Schildcrout JS, Denny JC, Bowton Eet al. Optimizing drug outcomes through pharmacogenetics: a case for preemptive genotyping. Clin Pharmacol Ther. 2012; 92:235–42. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Pulley JM, Denny JC, Peterson JFet al. Operational implementation of prospective genotyping for personalized medicine: the design of the vanderbilt PREDICT project. Clin Pharmacol Ther. 2012; 92:87–95. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Van Driest SL, Shi Y, Bowton EAet al. Clinically actionable genotypes among 10,000 patients with preemptive pharmacogenomic testing. Clin Pharmacol Ther. 2014; 95:423–31. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Elliott LS, Henderson JC, Neradilek MBet al. Clinical impact of pharmacogenetic profiling with a clinical decision support tool in polypharmacy home health patients: a prospective pilot randomized controlled trial. PLoS One. 2017; 12:e0170905. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Gottesman O, Scott SA, Ellis SBet al. The CLIPMERGE PGX program: clinical implementation of personalized medicine through electronic health records and genomics-pharmacogenomics. Clin Pharmacol Ther. 2013; 94:214–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Bielinski SJ, Olson JE, Pathak Jet al. Preemptive genotyping for personalized medicine: design of the right drug, right dose, right time-using genomic data to individualize treatment protocol. Mayo Clin Proc. 2014; 89:25–33. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Hertz DL, McLeod HL. Use of pharmacogenetics for predicting cancer prognosis and treatment exposure, response and toxicity. J Hum Genet. 2013; 58:346–52. [DOI] [PubMed] [Google Scholar]
- 14. Clinical Pharmacogenetics Implemention Consortium. Guidelines. https://cpicpgx.org/guidelines/ (accessed 2018 Jan 9).
- 15. Hocum BT, White JR Jr, Heck JWet al. Cytochrome P-450 gene and drug interaction analysis in patients referred for pharmacogenetic testing. Am J Health-Syst Pharm. 2016; 73:61–7. [DOI] [PubMed] [Google Scholar]
- 16. Dean L. Genetic variants and drug responses. In: Pratt V, McLeod H, Dean L. et al., eds. Medical genetics summaries. Bethesda, MD: National Center for Biotechology Information; 2012. [PubMed] [Google Scholar]
- 17. Candiotti KA, Birnbach DJ, Lubarsky DA, et al. The impact of pharmacogenomics on postoperative nausea and vomiting: do CYP2D6 allele copy number and polymorphisms affect the success or failure of ondansetron prophylaxis? Anesthesiology. 2005; 102:543–9. [DOI] [PubMed] [Google Scholar]
- 18. Deenen MJ, Meulendijks D, Cats Aet al. Upfront genotyping of DPYD*2A to individualize fluoropyrimidine therapy: a safety and cost analysis. J Clin Oncol. 2016; 34:227–34. [DOI] [PubMed] [Google Scholar]
- 19. Link E, Parish S, Armitage J, et al. ; SEARCH Collaborative Group. SLCO1B1 variants and statin-induced myopathy–a genomewide study. N Engl J Med. 2008; 359:789–99. [DOI] [PubMed] [Google Scholar]
- 20. Jukić MM, Haslemo T, Molden E, Ingelman-Sundberg M. Impact of CYP2C19 genotype on escitalopram exposure and therapeutic failure: a retrospective study based on 2,087 patients. Am J Psychiatry. 2018; 175:463–70. [DOI] [PubMed] [Google Scholar]
- 21. Poulsen L, Brøsen K, Arendt-Nielsen Let al. Codeine and morphine in extensive and poor metabolizers of sparteine: pharmacokinetics, analgesic effect and side effects. Eur J Clin Pharmacol. 1996; 51:289–95. [DOI] [PubMed] [Google Scholar]
- 22. Sindrup SH, Brøsen K, Bjerring Pet al. Codeine increases pain thresholds to copper vapor laser stimuli in extensive but not poor metabolizers of sparteine. Clin Pharmacol Ther. 1990; 48:686–93. [DOI] [PubMed] [Google Scholar]
- 23. Kirchheiner J, Schmidt H, Tzvetkov Met al. Pharmacokinetics of codeine and its metabolite morphine in ultra-rapid metabolizers due to CYP2D6 duplication. Pharmacogenomics J. 2007; 7:257–65. [DOI] [PubMed] [Google Scholar]
- 24. Gex-Fabry M, Eap CB, Oneda Bet al. CYP2D6 and ABCB1 genetic variability: influence on paroxetine plasma level and therapeutic response. Ther Drug Monit. 2008; 30:474–82. [DOI] [PubMed] [Google Scholar]
- 25. Mega JL, Simon T, Collet JP, et al. Reduced-function CYP2C19 genotype and risk of adverse clinical outcomes among patients treated with clopidogrel predominantly for PCI: a meta-analysis. JAMA. 2010; 304:1821–30. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Frueh FW, Amur S, Mummaneni Pet al. Pharmacogenomic biomarker information in drug labels approved by the United States Food and Drug Administration: prevalence of related drug use. Pharmacotherapy. 2008; 28:992–8. [DOI] [PubMed] [Google Scholar]
- 27. Kutner JS, Blatchford PJ, Taylor DH Jr, et al. Safety and benefit of discontinuing statin therapy in the setting of advanced, life-limiting illness: a randomized clinical trial. JAMA Intern Med. 2015; 175:691–700. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Centers for Disease Control and Prevention. Adverse drug events in adults.www.cdc.gov/medicationsafety/adult_adversedrugevents.html (accessed 2018 Jun 7).
- 29. Alagoz O, Durham D, Kasirajan K. Cost-effectiveness of one-time genetic testing to minimize lifetime adverse drug reactions. Pharmacogenomics J. 2016; 16:129–36. [DOI] [PubMed] [Google Scholar]
- 30. Burton MM, Hope C, Murray MDet al. The cost of adverse drug events in ambulatory care. AMIA Annu Symp Proc. 2007; 2007: 90–3. [PubMed] [Google Scholar]
- 31. National Human Genome Research Institute. The cost of sequencing a human genome.www.genome.gov/27565109/the-cost-of-sequencing-a-human-genome/ (accessed 2018 Jul 6).
- 32. Patel JN, Wiebe LA, Dunnenberger HM, McLeod HL. Value of supportive care pharmacogenomics in oncology practice. Oncologist. 2018; 23:956–64. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Furuta T, Shirai N, Sugimoto M, et al. Pharmacogenomics of proton pump inhibitors. Pharmacogenomics. 2004; 5:181–202. [DOI] [PubMed] [Google Scholar]