Lung cancer is the leading cause of cancer-related deaths, accounting for approximately 18% (Sung et al., 2021). Approximately 80%–85% of these cases are classified as non-small cell lung cancer (NSCLC) (Nicholson et al., 2022). In Thai patients with NSCLC, epidermal growth factor receptor (EGFR) mutations were most commonly detected, accounting for 68% of cases (Detarkom et al., 2018). Activating EGFR mutations have been identified as predictive indicators of sensitivity to first- and second-generation EGFR tyrosine kinase inhibitors (TKIs). Acquired resistance develops 9–12 months after treatment initiation (Westover et al., 2018). One common mechanism underlying acquired resistance is the substitution of threonine with methionine at amino acid position 790 (T790M) in exon 20 of the EGFR gene, which accounts for 50%–60% of cases. This mutation impairs the binding of both first- and second-generation EGFR-TKIs by enhancing the ATP-binding affinity of the kinase domain of the EGFR mutant receptor, leading to treatment resistance (Yu et al., 2013).

Osimertinib is a third-generation EGFR-TKI that irreversibly binds to cysteine-797 at the ATP-binding site of the EGFR kinase domain (Li et al., 2023). It potently inhibits EGFR phosphorylation in the cases of exon 19 deletion and exon 21 L858R substitution, with IC50 values ranging 13–54 nmol/L. Additionally, in cell lines harboring EGFR T790M mutation, osimertinib demonstrates remarkable potency with an IC50 of less than 15 nmol/L. In addition, Osimertinib exhibits high selectivity for mutated EGFR receptors over wild-type EGFR (IC50: 480–1865 nmol/L) (Remon et al., 2018), resulting in less severe gastrointestinal and skin toxicities than those elicited by the first- or second-generation EGFR-TKIs. Additionally, osimertinib has improved overall survival in previously untreated advanced NSCLC patients with EGFR mutations compared with standard EGFR-TKIs (Ramalingam et al., 2020). Osimertinib is predominantly metabolized by CYP3A4, CYP2A6, CYP2C9, CYP3A5, and CYP2E1 enzymes, which account for 44.4%, 15.5%, 12.0%, 9.6%, and 3.0% of the metabolism, respectively (Dickinson et al., 2016). It produces at least two circulating metabolites, AZ5104 and AZ7550, accounting for 10% of the parent compound (AstraZeneca, 2021). AZ7550 has a potency and selectivity profile comparable to osimertinib, whereas AZ5104 has an 8-fold greater potency against EGFR mutations (Han et al., 2021).

In a previous study, variability in the steady-state area under the plasma drug concentration-time curve (AUCs) of AZ5104 between ethnic groups was observed, with a 10%–23% decrease in Asian versus Caucasian patients; however, the underlying reason remains unclear (Brown et al., 2017). CYP450 predominantly metabolizes Osimertinib and is a substrate for P-glycoprotein (P-gp), which is encoded by ABCB1, and breast cancer resistance protein (BCRP), which is encoded by ABCG2 (AstraZeneca, 2021), which may lead to individual variations in plasma osimertinib concentrations due to genetic polymorphisms. A previous study also found a linear relationship between ADR development and osimertinib concentration, with an increased risk of rash, diarrhea, and QTc prolongation (Brown et al., 2017). Moreover, in Asian patients, the incidence of QTc prolongation was slightly higher than that reported in the FLAURA study (Soria et al., 2018; Cho et al., 2019). Furthermore, a significant association was demonstrated between the AUC_0-24 of osimertinib, SNPs rs1128503 in ABCB1, and SNPs rs2231137 in ABCG2 with grade 2 or higher adverse events (Ishikawa et al., 2023). The previous studies on EGFR-TKIs found a correlation between SNPs rs762551 in CYP1A2 and the severity of erlotinib-induced skin rash, along with the development of diarrhea and SNPs rs2470890 in ABCB1 and rs776746 in CYP3A5 (Liao et al., 2020). Additionally, SNPs rs2032582 in ABCB1 were associated with afatinib-induced diarrhea (Sogawa et al., 2020), and reduced function of CYP2D6 increased the risk of gefitinib-induced rash (Suzumura et al., 2012). However, the association between SNPs in CYP450 and drug efflux transporters and efficacy outcomes remains unclear. We investigated seventeen SNPs in CYP450 and drug efflux transporters that may alter the pharmacokinetic and pharmacodynamic profile of osimertinib, intending to advance personalized medicine approaches.

A recent population pharmacokinetic study showed a linear correlation between exposure to osimertinib, measured using the AUCs of the parent compound and two active metabolites (AZ5104 and AZ7550), and the incidence of ADRs (Brown et al., 2017). AZ5104, in particular, demonstrated an 8-fold greater potency against EGFR mutations (Han et al., 2021). While AZ7550 and AZ5104 are present in approximately 10% of the parent compound (Brown et al., 2017), a decrease in AZ5104 AUCs of 10%–23% in Asian patients compared with that of Caucasian patients may influence clinical outcomes. In addition, a study in Asian populations reported a significant association between SNPs rs1128503 in ABCB1 and rs2231137 in ABCG2 and osimertinib-induced grade ≥2 adverse events (Ishikawa et al., 2023). However, the association between SNPs and the efficacy of osimertinib remains unclear, and the specific SNPs that influence the pharmacokinetics and clinical effects of osimertinib in NSCLC remain unknown.

This is the first study to analyze a large number of genetic polymorphisms in candidate genes involved in the osimertinib pharmacokinetic pathway to enable the assessment of both efficacy and safety endpoints. All allele frequencies were consistent with the Hardy–Weinberg equilibrium ( PharmGKB, 2021). In the present study, the frequency of dose reduction (27.0%) was higher than that reported in the AURA 3 trial (16.5%), and the incidence of ADRs was higher than that in the FLAURA, AURA2, and AURA3 trials (Goss et al., 2016; Mok et al., 2017; Soria et al., 2018). This may be because of genetic differences between patient populations; Asian patients who were administered osimertinib in these studies accounted for only 62%, 63%, and 65% of the patients, respectively. The frequency of the ABCG2 rs2231164 (C) allele, known as a loss-of-function variant, was 23.72% in the South Asian population and only 12.12% in the European population (Whirl-Carrillo et al., 2021). Furthermore, the frequency of the CYP2A6*4 allele, a slower metabolizer, also differed significantly between ethnic groups (Pang et al., 2015), potentially leading to higher plasma concentrations of osimertinib in the Thai population. These genetic differences may have contributed to the observed differences in dose reductions and ADRs between our study and previous clinical trials. Additionally, we found that the ORR in our study was 49.2%, which was lower than that reported in the AURA 3 trial (71%) (Mok et al., 2017). This difference in response rates may be because of the differences in patient populations between the two studies. In the AURA 3 trial, patients received osimertinib as second-line therapy and 96% of cases and later-line therapy in 4%. In contrast, in our study, 57.1% of patients received osimertinib as second-line therapy, and 27.0% received osimertinib as later-line therapy.

In our study, we identified six SNPs that were significantly associated with the incidence of ADRs, namely, rs2231142 in ABCG2 mutant-type (A/A), rs2622604 in ABCG2 mutant-type (T/T), CYP2A6 heterozygous variant (non*4/*4), rs1057910 in CYP2C9 mutant-type (C/C), rs28371759 in CYP3A4 heterozygous variant (A/G) and rs762551 in CYP1A2 wild-type (C/C). These findings are consistent with previous studies reporting significant associations between genetic polymorphisms and ADR risk. For example, rs2231142 in the ABCG2 mutant-type (A/A) is significantly associated with sunitinib-induced severe thrombocytopenia (Low et al., 2016), whereas rs2622604 in the ABCG2 mutant-type (T/T) is significantly associated with irinotecan-induced severe myelosuppression (Cha et al., 2009). Similarly, the CYP2A6 (*4) allele is significantly associated with letrozole-induced ADRs (Desta et al., 2011), whereas the CYP3A4 *18 (G) allele is significantly associated with tacrolimus-induced ADRs (Bruckmueller et al., 2015). Notably, in contrast to other genes, rs762551 in the CYP1A2 wild-type (C/C) was significantly related to osimertinib-induced ADRs. This result may be attributable to the higher enzyme activity observed in the presence of an inducer, such as smoking or heavy coffee consumption, which leads to higher enzyme activity in the CYP1A2 mutant-type (A/A) (Djordjevic et al., 2010; Wang et al., 2012), and a lower incidence of ADRs was observed in this variant. With regard to efficacy outcomes, we identified two SNPs (rs2069514 in CYP1A2 and rs1057910 in CYP2C9) that were significantly associated with the median TTF and two SNPs (rs28399433 in CYP2A6 and rs1057910 in CYP2C9) that were significantly associated with the median PFS. Notably, one of these SNPs (rs1057910 in CYP2C9) was significantly associated with ADRs, TTF, and PFS.

The SNPs in CYP450 and the drug efflux transporters discussed above were significantly associated with ADRs, TTF, and PFS. Because osimertinib is a substrate of CYP450, ABCB1, and ABCG2 (AstraZeneca, 2021), polymorphisms in these genes may affect the distribution and pharmacokinetics of this drug. Therefore, we hypothesized that mutations causing the decreased function of CYP450, ABCB1, and ABCG2 may affect the tissue distribution and accumulation of osimertinib. An in vivo study suggested that ABCB1 and ABCG2 are involved in the tissue accumulation of other TKIs (Al-Shammari et al., 2019). As the active metabolite of osimertinib (AZ5104) is more potent than the parent compound against EGFR mutations (Remon et al., 2018) and accounts for approximately 10% of the parent compound (Brown et al., 2017), the accumulation of osimertinib may lead to a significantly increased incidence of ADRs but significantly decreased TTF and PFS. For example, SNP rs1057910 (C), located in the CYP2C9 gene, typically encodes the amino acid leucine at position 359, and the resulting allele is also known as CYP2C9*3, which is a decreased function variant (VCV000008408.12 - ClinVar - NCBI, 2013). This variant may lead to poor metabolism of osimertinib in patients carrying the CYP2C9*3 allele, which increases the risk of osimertinib-induced ADRs due to higher osimertinib exposure but also decreases survival outcomes due to lower exposure to its active metabolite (AZ5104).

These findings are similar to those of a previous study on the osimertinib exposure-response relationship, which found that the mortality rate was significantly higher in the high osimertinib drug level group (Rodier et al., 2022). Additionally, the probability of developing rashes, diarrhea, or QTc prolongation increased with exposure, and a linear relationship between adverse event development and osimertinib levels was identified (Brown et al., 2017). The mechanism that supports EGFR-TKI-induced ADRs is the inhibition of EGFR1 and EGFR2 (HER2) signaling, leading to a reduction in growth and impaired healing of the epithelium where EGFR is expressed. This subsequently causes alterations in keratinocyte proliferation and differentiation, reduced growth, impaired healing of the intestinal epithelium, and alterations in myocyte growth (Giovannini et al., 2009; Hirsh et al., 2014; Ikebe et al., 2020).

Our study has several limitations. First, the sample size was small, and some clinical data were retrospectively collected from medical records. Nevertheless, no genotype distribution deviated from the Hardy–Weinberg equation ( PharmGKB, 2021), and the clinical data were confirmed by the patient’s physician. Second, our inability to control for other confounding factors, such as compliance with osimertinib treatment caused by home oral medication but we had oncology pharmacist pill count measures of compliance at every visit, and all patients maintained a 100% compliance rate. Additionally, there were no observed drug interactions that influenced osimertinib drug levels. However, we did not restrict patient coffee consumption, which may have an impact on higher enzyme activity in the SNP rs762551 in CYP1A2 mutant-type (A/A). This enzyme was inducible by heavy coffee consumption and has been associated with a lower incidence of ADRs (Djordjevic et al., 2010; Wang et al., 2012). Finally, a pharmacokinetic analysis was not included in this study. However, the association between SNPs and clinical outcomes was consistent with that in previous reports.

In conclusion, our study identified significant SNPs associated with increased ADRs incidence, decreased TTF, and decreased PFS in Thai patients with NSCLC treated with osimertinib. The findings can potentially guide treatment decisions and help optimize individualized therapy for patients with NSCLC harboring EGFR mutations. However, more extensive studies with analysis of osimertinib and its active metabolite drug levels must be conducted to confirm these findings.