Background: Several opioids have pharmacogenomic associations impacting analgesic efficacy. However, germline pharmacogenomic testing is not routinely incorporated into supportive oncology. We hypothesized that CYP2D6 profiling would correlate with opioid prescribing and hospitalizations. Materials and Methods: We analyzed 61,572 adult oncology patients from 2012 to 2018 for opioid exposures. CYP2D6 metabolizer phenotype (ultra-rapid [UM], normal metabolizer [NM], intermediate [IM], or poor [PM]), the latter two of which may cause inefficacy of codeine, tramadol, and standard-dose hydrocodone, was determined for patients genotyped for reasons unrelated to pain. The primary endpoint was number of opioid medications received during longitudinal care (IM/PMs vs. NMs). Secondary endpoint was likelihood of pain-related hospital encounters. Results: Most patients with cancer (n = 34,675, 56%) received multiple opioids (average 2.8 ± 1.6/patient). Hydrocodone was most commonly prescribed (62%), followed by tramadol, oxycodone, and codeine. In the CYP2D6 genotyped cohort (n = 105), IM/PMs received a similar number of opioids (3.4 ± 1.4) as NMs (3.3 ± 1.9). However, IM/PMs were significantly more likely to experience pain-related hospital encounters compared with NMs, independent of other variables (odds ratio [OR] = 5.4; 95% confidence interval [CI], 1.2–23.6; p = .03). IM/PMs were also more likely to be treated with later-line opioids that do not require CYP2D6 metabolism, such as morphine and hydromorphone (OR = 3.3; 95% CI, 1.1–9.8; p = .03). Conclusion: CYP2D6 genotype may identify patients with cancer at increased risk for inadequate analgesia when treated with typical first-line opioids like codeine, tramadol, or standard-dose hydrocodone. Palliative care considerations are an integral part of optimal oncology care, and these findings justify prospective evaluation of preemptive genotyping as a strategy to improve oncology pain management. Implications for Practice: Genomic variation in metabolic enzymes can predispose individuals to inefficacy when receiving opioid pain medications. Patients with intermediate and/or poor CYP2D6 metabolizer status do not adequately convert codeine, tramadol, and hydrocodone into active compounds, with resulting increased risk of inadequate analgesia. This study showed that patients with cancer frequently receive CYP2D6-dependent opioids. However, patients with CYP2D6 intermediate and poor metabolizer status had increased numbers of pain-related hospitalizations and more frequently required the potent non-CYP2D6 opioids morphine and hydromorphone. This may reflect inadequate initial analgesia with the common “first-line” CYP2D6-metabolized opioids. Preemptive genotyping to guide opioid prescribing during cancer care may improve pain-related patient outcomes.

Many cancer patients who receive chemotherapy experience adverse drug effects. Pharmacogenomics (PGx) has promise to personalize chemotherapy drug dosing to maximize efficacy and safety. Fluoropyrimidines and irinotecan have well-known germline PGx associations. At our institution, we have delivered PGx clinical decision support (CDS) based on preemptively obtained genotyping results for a large number of non-oncology medications since 2012, but have not previously evaluated the utility of this strategy for patients initiating anti-cancer regimens. We hypothesize that providing oncologists with preemptive germline PGx information along with CDS will enable individualized dosing decisions and result in improved patient outcomes. Patients with oncologic malignancies for whom fluoropyrimidine and/or irinotecan-inclusive therapy is being planned will be enrolled and randomly assigned to PGx and control arms. Patients will be genotyped in a clinical laboratory across panels that include actionable variants in UGT1A1 and DPYD. For PGx arm patients, treating providers will be given access to the patient-specific PGx results with CDS prior to treatment initiation. In the control arm, genotyping will be deferred, and dosing will occur as per usual care. Co-primary endpoints are dose intensity deviation rate (the proportion of patients receiving dose modifications during the first treatment cycle), and grade ⩾3 treatment-related toxicities throughout the treatment course. Additional study endpoints will include cumulative drug dose intensity, progression-free survival, dosing of additional PGx supportive medications, and patient-reported quality of life and understanding of PGx. Providing a platform of integrated germline PGx information may promote personalized chemotherapy dosing decisions and establish a new model of care to optimize oncology treatment planning.

The integration of genomic data into personalized treatment planning has revolutionized oncology care. Despite this, patients with cancer remain vulnerable to high rates of adverse drug events and medication inefficacy, affecting prognosis and quality of life. Pharmacogenomics is a field seeking to identify germline genetic variants that contribute to an individual's unique drug response. Although there is widespread integration of genomic information in oncology, somatic platforms, rather than germline biomarkers, have dominated the attention of cancer providers. Patients with cancer potentially stand to benefit from improved integration of both somatic and germline genomic information, especially because the latter may complement treatment planning by informing toxicity risk for drugs with treatment‐limiting tolerabilities and narrow therapeutic indices. Although certain germline pharmacogenes, such as TPMT , UGT1A1 , and DPYD , have been recognized for decades, recent attention has illuminated modern potential dosing implications for a whole new set of anticancer agents, including targeted therapies and antibody‐drug conjugates, as well as the discovery of additional genetic variants and newly relevant pharmacogenes. Some of this information has risen to the level of directing clinical action, with US Food and Drug Administration label guidance and recommendations by international societies and governing bodies. This review is focused on key new pharmacogenomic evidence and oncology‐specific dosing recommendations. Personalized oncology care through integrated pharmacogenomics represents a unique multidisciplinary collaboration between oncologists, laboratory science, bioinformatics, pharmacists, clinical pharmacologists, and genetic counselors, among others. The authors posit that expanded consideration of germline genetic information can further transform the safe and effective practice of oncology in 2022 and beyond.

Purpose: Around 10% to 15% of prostate cancers harbor recurrent aberrations in the Forkhead Box A1 gene, FOXA1, whereby the alteration type and the effect on the forkhead (FKH) domain affect protein function. We developed a FOXA1 classification system to inform clinical management. Experimental Design: A total of 5,014 prostate cancer samples were examined using whole-exome and -transcriptome sequencing from the Caris Life Sciences database. We denoted class 1 FOXA1 alterations as missense and in-frame insertions/deletions with subclasses oriented with respect to the FKH domain. These were in the first part of the FKH domain [class 1A: amino acids (AA) 168–246], within the Wing2 region of FKH (class 1B: AA 247–269), or outside FKH (class 1C: AA 1–167, 270+). Two hotspot missense mutations at R219 were denoted class 2. Class 3 included predicted truncating mutations with subclasses partitioned based on the FKH domain (class 3A: AA 1–269 and class 3B: AA 270+). Class 4 represented FOXA1 amplifications. Real-world overall survival and therapy outcomes were determined from insurance claims. Results: FOXA1 alterations did not influence survival when considered in aggregate but had distinct prognostic effects when stratified by class. In primary prostate samples, class 1A alterations were associated with overall improved survival (HR, 0.57; P = 0.03); a similar trend was seen in metastatic biopsies with class 1B (HR, 0.84; P = 0.09). Conversely, in primary specimens, class 1C exhibited worse survival upon second-generation androgen receptor signaling inhibitor treatment (HR, 1.93; P < 0.001). Class 2 mutations (R219C/S) were enriched in neuroendocrine prostate cancers and were associated with overall poor survival (HR, 2.05; P < 0.001) and worse outcomes to first-line androgen-deprivation therapies (HR, 2.5; P < 0.001). Class 3A alterations indicated improved survival (HR, 0.70; P = 0.01), whereas class 3B alterations portended poor outcomes (HR, 1.50; P < 0.001). Amplifications (class 4) indicated poor outcomes in metastatic samples (HR, 1.48; P = 0.02). Molecularly, different FOXA1 alteration classes harbored distinct mutational and immunologic features as well as unique transcriptional programs. Finally, relative to European Americans, African Americans had increased class 1C alterations, whereas Asian/Pacific Islander patients had increased class 1B alterations. Conclusions: FOXA1 alterations should not be interpreted in aggregate, as different classes are associated with divergent molecular features and clinical outcomes. Our revised classification schema facilitates clinical decision-making for patients with prostate cancer and uncovers important racial differences.

PURPOSE: Precision oncology in non–small cell lung cancer (NSCLC) relies on biomarker testing for clinical decision making. Despite its importance, challenges like the lack of genomic oncology training, nonstandardized biomarker reporting, and a rapidly evolving treatment landscape hinder its practice. Generative artificial intelligence (AI), such as ChatGPT, offers promise for enhancing clinical decision support. Effective performance metrics are crucial to evaluate these models' accuracy and their propensity for producing incorrect or hallucinated information. We assessed various ChatGPT versions' ability to generate accurate next-generation sequencing reports and treatment recommendations for NSCLC, using a novel Generative AI Performance Score (G-PS), which considers accuracy, relevancy, and hallucinations. METHODS: We queried ChatGPT versions for first-line NSCLC treatment recommendations with an Food and Drug Administration–approved targeted therapy, using a zero-shot prompt approach for eight oncogenes. Responses were assessed against National Comprehensive Cancer Network (NCCN) guidelines for accuracy, relevance, and hallucinations, with G-PS calculating scores from –1 (all hallucinations) to 1 (fully NCCN-compliant recommendations). G-PS was designed as a composite measure with a base score for correct recommendations (weighted for preferred treatments) and a penalty for hallucinations. RESULTS: Analyzing 160 responses, generative pre-trained transformer (GPT)-4 outperformed GPT-3.5, showing higher base score (90% v 60%; P < .01) and fewer hallucinations (34% v 53%; P < .01). GPT-4's overall G-PS was significantly higher (0.34 v –0.15; P < .01), indicating superior performance. CONCLUSION: This study highlights the rapid improvement of generative AI in matching treatment recommendations with biomarkers in precision oncology. Although the rate of hallucinations improved in the GPT-4 model, future generative AI use in clinical care requires high levels of accuracy with minimal to no room for hallucinations. The GP-S represents a novel metric quantifying generative AI utility in health care compared with national guidelines, with potential adaptation beyond precision oncology.