Influence of SNVs on adverse reactions and survival in gefitinib-treated lung cancer patients from a preliminary study.

Introduction
Non-small cell lung cancer (NSCLC) accounts for 85% of all lung cancer (LC) cases and is the leading cause of cancer-related deaths worldwide1,2. Smoking is the primary risk factor for LC, although individuals who have never smoked can be diagnosed with cancer based on environmental, behavioral, and genetic factors3. Treatment with epidermal growth factor receptor (EGFR) tyrosine kinase inhibitors (TKIs), such as gefitinib, is indicated for advanced stages of metastatic NSCLC with activating EGFR mutations, mainly in exons 19 and 20–21 4.
Despite being the first TKI approved by the Food and Drug Administration (FDA) in 2003, gefitinib remains widely used in clinical practice, particularly in developing countries5. It is an orally administered chemotherapeutic drug and is well-tolerated compared to other TKIs, such as erlotinib and osimertinib6,7. Compared with standard platinum-based treatments, gefitinib has demonstrated longer progression-free survival (PFS) and improved quality of life; however, evidence regarding its impact on overall survival (OS) remains unclear8,9.
Gefitinib acts as a reversible inhibitor of adenosine triphosphate (ATP) in the C-terminal domain of EGFR10,11. It is typically administered at a dose of 250 mg/day, which may result in adverse drug reactions (ADRs) affecting patients’ quality of life. In some cases, it may lead to interruption or discontinuation of treatment12,13. The most notable ADRs of gefitinib are dermatological (70–90%), gastrointestinal (39–57%), and hepatic (10%)14–17.
The occurrence of ADR at a prescribed dose is dependent on several factors, such as individual characteristics (sex, age, and ethnicity), genetic factors, family history, and drug/food interactions18,19. Individual variations in genetic traits can be explained by single-nucleotide variants (SNVs). These variants, whether silent or not, result in changes in protein expression in individuals, affecting pharmacokinetics (PK) and pharmacodynamics and, consequently, the response to treatment20–22.
Gefitinib is predominantly metabolized in the liver by cytochrome P450 (CYP) enzymes. Membrane transporters of the ATP-binding cassette (ABC) family also function as TKI23–26. SNVs in ABC transporter genes can alter P-glycoprotein activity and expression. Some studies have reported an association between these variants and gefitinib-induced ADRs27. For example, the ABCB1 (rs1128503) TT genotype is associated with skin reactions and diarrhea28, while ABCG2 (rs22311370) has also been suggested as a predictor of dermatological reactions induced by gefitinib29,30. These findings support the hypothesis that variability in genes involved in drug transport and signaling pathways may influence the development of ADRs and affect treatment outcomes.
Given the pharmacogenetic (PGx) relevance of EGFR, ABCB1, and ABCG2 variants, this study aimed to identify genetic variants that may contribute to gefitinib-induced ADRs and affect patient survival in NSCLC.

Results

Demographic and clinical characteristics of patients
Table 1 shows the demographic and clinical characteristics of the patients included in this study. A total of 40 patients were initially included in this study. Four patients were excluded― three due to death during the study and one patient due to withdrawal. Clinical and demographic data were collected, and ADRs were evaluated and graded using CTCAE v5.0 in 36 patients. Blood samples from three patients were not collected for DNA extraction. Therefore, 33 patients underwent analysis of genetic variants and their associations with ADRs.

The patients were predominantly white (86.1%) and female (69.4%), with a median age of 64 years. Most patients never smoked (52.8%) and denied alcohol consumption (69.4%). Most (72.2%) patients in the study did not have comorbidities associated with NSCLC, and 52.8% had undergone treatment before starting gefitinib. According to the TNM classification, most tumors were categorized as advanced (T4; 55.5%). The most common mutation was a deletion in EGFR exon 19 (63.9%).
Supplementary Fig. 1 shows a flowchart of the study patients.

Occurrence and severity of ADRs
Table 2 summarizes the ADRs observed in this study beyond the CTCAE grade. Of the 36 patients included in this study, 80.5% experienced at least one ADR. Dermatological reactions were reported in 50.0%, gastrointestinal in l72.2%, and hepatic in 80.6% of patients.

Regardless of the degree, hyperpigmentation was observed in 50.0% of the cases, followed by increased GGT levels in 47.2%. When considering the highest (grade 2 or higher) ADRs, an increased GGT level was observed in 19.4% of cases, followed by tegumentary rash (11.1%) and diarrhea (11.1%).
Supplementary Table 1 presents baseline clinical characteristics and concomitant treatments that may influence adverse drug reactions (ADRs) and outcomes.

Frequency of genetic variants
The allelic distribution of ABCB1 genes was analyzed for the following variants: for the SNV c.3435G > A, an allele frequency of 37% for allele A and 62% for allele G was observed; for the SNV c.1236 C > T, 59% for allele C and 41% for allele T; and for the triallelic SNV c.2677 C > T/A, the frequencies were 72% for allele C and 27% for allele A. All these variants were evaluated using the HWE tool and exhibited frequencies similar to those observed in global [31] and Brazilian [32] populations.
In contrast, the ABCG2 variant (421G > T) did not follow the expected HWE distribution.
For the EGFR gene, the SNV c.1562G > A showed a frequency of 24% for allele A and 76% for allele G, whereas the SNV c.2982 C > T showed a frequency of 95% for allele C and 5% for allele T. These variants were also in accordance with the expected HWE distribution, and their frequencies were consistent with those reported in global and Brazilian population studies. The detailed data are shown in Supplementary Table 2.

Associations of clinical and demographic data with ADRs
Table 3 shows significant associations between ADRs and clinical and demographic data in multiple regression analyses. The data indicate that individuals with comorbidities associated with NSCLC and treated with gefitinib had a 9.4-fold higher likelihood (p = 0.0289) of developing any gastrointestinal adverse reaction, regardless of severity. Individuals with an EGFR exon 19 deletion had a 7.6-fold higher likelihood (p = 0.0399) of not experiencing diarrhea (grade 0), indicating the absence of this adverse reaction. Data that did not show significant associations between adverse reactions and clinical variables were excluded.

Associations of SNVs with ADRs
Table 4 presents univariate and multivariate regression analyses for the ABCB1 rs2032582 (c.2677 C > T/A) SNV in association with diarrhea-related ADRs. In this analysis, individuals with non-CC genotypes had a 7.5-fold higher probability (p = 0.0344) of developing diarrhea of any grade than those with the CC genotype (post-hoc power = 85%).

The EGFR SNVs and other ABCB1 SNVs studied did not show positive associations with dermatological or nausea/diarrhea-related ADRs; the data are presented in Supplementary Tables 3 and 4.
To support the interpretation of these findings in Table 4, Supplementary Table 5 presents the complete genotype counts for rs2032582 (c.2677 C > T/A) stratified by each evaluated ADR.
The associations between SNVs and ADRs stratified by severity (grade 1 vs. grades 2–3 and grades 0–1 vs. grades 2–3) did not reveal any significant positive correlations and were not suitable for regression analysis. The findings are presented in Supplementary Table 6.

Survival analysis
The estimated mean OS of patients with NSCLC treated with gefitinib in this study was 1225 days (approximately three years and three months). No significant associations were found between clinical and demographic data or between ADRs and patient survival. The data are presented in Supplementary Tables 7 and 8.
Figure 1 shows the Kaplan–Meier survival analysis graph in association with the ABCB1 rs2032582 (c.2677 C > T/A) SNV. As shown in this figure, patients with the AA genotype had a 32.498-fold increased risk of death compared with those with the CC/CA/CT/TT genotypes (post-hoc power = 33%).

The other EGFR and ABCB1 SNVs were not significantly associated with patient survival. The data are presented in Supplementary Table 9.
Detailed results, including the Kaplan-Meier plot with numbers-at-risk and Schoenfeld residual analysis, confirming the adequacy of the Cox model, are provided in Supplementary Tables 10 and Supplementary Fig. 2.

Discussion
To our knowledge, this is the first preliminary study in a Brazilian population to examine SNVs in ABC transporters and EGFR and their correlation with ADRs and patient survival in patients treated with gefitinib. Identifying PGx markers for gefitinib-induced ADRs may help predict the risk of adverse outcomes during the treatment of patients with NSCLC. By screening for these variants, clinicians can anticipate potential ADRs, enabling early interventions or therapeutic adjustments. This strategy helps to mitigate the severity of potential ADRs, improve therapeutic outcomes for patients with NSCLC, and enhance their quality of life during treatment.
Notably, most patients included in this exploratory study experienced ADRs. In particular, 80% of ADRs were hepatic, including the most frequent grade ≥ 2 events: increased levels of hepatic enzymes GGT and ALP (19.4% and 11.1%, respectively). Gefitinib-induced hepatotoxicity, particularly elevated ALT and AST levels, has been reported33–35. Possible causes of gefitinib-induced hepatotoxicity include dose-dependent cellular toxicity, immune-mediated reactions, genetic variants in genes involved in gefitinib metabolism and transport, and interactions with CYP3A4 inhibitor drugs36–38. Gastrointestinal adverse reactions were also frequent, affecting 72% of patients, with diarrhea being the most common (30.6% grade 1 and 11.1% grade > 2). Diarrhea is expected with the use of targeted therapy in general, including gefitinib, in 40–60% of patients17,35. Diarrhea is not usually associated with abdominal pain, but it can easily lead to dehydration in patients39.
Moreover, 50% of the participants experienced dermatological ADRs, with tegumentary rashes (30.6%) and hyperpigmentation (41.7%) being the most prevalent. This finding is consistent with the IDEAL 1–2 and IPASS clinical trials, which demonstrated the presence of dermatological reactions in 50–55% of cases9,17,35. In fact, tegumentary reactions are commonly reported and expected in patients because EGFR is expressed in the basal layers of epidermal keratinocytes and hair follicles, and blocking EGFR leads to endothelial inflammation, reduced vascular tone, and consequently increased vascular permeability14,40,41.
An interesting finding of our study is that patients with mutations in exons 20–21 of EGFR had a 7.7-fold higher risk of diarrhea. To our knowledge, this is the first study to investigate this relationship, and additional studies are needed to validate this preliminary signal, particularly as a study in Chinese patients did not observe this association42. Previous studies have suggested that clinical outcomes may differ between EGFR mutation locations, with exon 19 deletions consistently associated with more favorable responses and survival outcomes than exon 21 mutations in patients treated with first-generation EGFR-TKIs, including gefitinib43,44. However, published data directly examining clinical outcomes stratified between exons 19 and 21 remain limited, and the findings are not entirely consistent across cohorts. Importantly, our study had limited statistical power to detect modest associations due to the cohort size and relatively low frequency of individual variants. Thus, the findings should be interpreted with caution, as they remain exploratory and require confirmation in larger, adequately powered cohorts.
Efflux transporters, including ABC and P-glycoprotein (P-gp; multidrug resistance 1), are of great importance and have clinical relevance. Studies have demonstrated their effects on gefitinib transport45,46. A systematic review and meta-analysis conducted by our study group47, which included five studies, indicated that genetic variants in the ABCB1 (rs1045642 and rs1128503) and ABCG2 (rs2231142) transporters are likely to have clinical implications for the occurrence of adverse reactions (skin rash and diarrhea) to gefitinib in patients with NSCLC. In this study, we obtained interesting findings regarding the triallelic ABCB1 rs2032582 variant, which is associated with both adverse reactions and patient survival.
This triallelic ABCB1 variant (rs2032582, c.2677 C > T/A) is unique, as there are divergences between studies in terms of grouping the genotypes to continue correlation studies. Ma et al. also observed a positive association between the ABCB1 rs2032582 TT genotype and diarrhea. Our findings verified a positive association between diarrhea and the non-CC genotype, including the TT genotype, which corroborates the previous study48. This contrasts with the findings of Kobayashi et al.49 who did not find a positive association between the triallelic gene ABCB1 and gefitinib-induced ADRs. Although our cohort was small, the post hoc power analysis indicated that the study had sufficient statistical power to detect this association.
Complementing our findings, we observed a single association between the c.2677 C > T/A SNV in ABCB1 and patient survival in this cohort. The AA genotype was associated with worse OS than the CC/CA/CT/TT genotypes. However, the study by Ma et al.48 did not observe associations of this ABCB1 variant with OS or PSF. Still, it did observe associations with treatment response, suggesting that the GG genotype at rs2032582 is probably a risk factor for gefitinib treatment response. The AA genotype of the c.2677 > C/T variant was associated with prolonged OS in patients with ovarian cancer treated with carboplatin50. To date, few studies have examined survival and ABCB1 expression in patients treated with gefitinib. In our study, the observed post hoc power for this association was insufficient, reinforcing the need to interpret these findings with caution and to validate them in larger cohorts to confirm their robustness.
Based on these findings, we hypothesized that allelic variants of the ABCB1 gene, particularly the triallelic SNV rs2032582 (c.2677 C > T/A), may be associated with a decreased incidence of gefitinib-induced diarrhea, potentially due to alterations in P-glycoprotein (P-gp) activity, which reduces the bioavailability of the drug in the gastrointestinal tract. Paradoxically, this lower incidence of adverse reactions may reflect reduced systemic exposure to gefitinib, which could contribute to diminished therapeutic efficacy and consequently, reduced OS in patients with NSCLC harboring these variants. These findings suggest that this ABCB1 SNV may serve as a predictive biomarker for both treatment-related adverse reactions and clinical outcomes in gefitinib-treated patients. This possible mechanism should be confirmed in future PK-PGx studies.
To further explore the mechanisms underlying these associations, it is essential to investigate whether ABCB1 variants influence gefitinib plasma concentrations by modulating P-gp activity. PK studies designed to characterize this relationship would help clarify how these variants affect both drug exposure and clinical outcomes. The clinical effects of gefitinib are strongly influenced by factors including absorption, metabolism (primarily via CYP3A4 and CYP2D6), distribution, and transporter-mediated efflux. Variability at these levels can significantly alter systemic drug exposure, thereby modulating both therapeutic efficacy and incidence of adverse events. CYP2D6 is a known determinant of gefitinib toxicity51. CYP2D6 SNVs have been associated with altered gefitinib plasma concentrations and an increased risk of ADRs such as cutaneous rash and severe hepatic injury. Reflecting this evidence, the European Medicines Agency (EMA) recommends close monitoring of patients carrying a CYP2D6 poor metabolizer genotype due to their higher susceptibility to adverse events52. Preemptive CYP2D6 genotyping before initiating gefitinib therapy has been suggested as a strategy for identifying at-risk patients and minimizing severe ADRs51.
Thus, integrating PK determinants with PGx variability provides a more comprehensive framework to explain inter-individual differences in treatment response and clinical outcomes in patients receiving gefitinib24,48. Future PK-PGx studies are essential to validate the exposure-response relationships and clarify how genetic variability influences systemic drug exposure, treatment efficacy, and the risk of ADRs.
Several factors influence the occurrence of ADRs. Among these, genetic factors, such as SNVs, can contribute to 20–40% of an individual’s response to medications53. These variants can affect various aspects of gene expression based on their genomic location, such as altering the structure, function, or expression levels of the corresponding protein products21. Thus, the selection of SNVs in this study was primarily based on two types: synonymous variants (rs1128503, rs1045642, and rs2293347) and missense variants (rs2032582, rs2231142, and rs2227983). In missense variants, a single base-pair substitution occurs within the coding sequence, leading to the production of a different amino acid at that position and altering protein activity. This may also affect key binding sites involved in protein interactions22. However, in synonymous variants, the amino acid sequence remains unchanged, but they can still affect translation rates or mRNA stability54,55. Owing to these potential effects, SNVs are considered promising therapeutic biomarkers.
The ABCB1 gene, highlighted in this preliminary study, has been widely studied because it is a highly polymorphic region with approximately 50 identified SNVs. Variants of this gene are associated with the activity of P-glycoprotein (P-gp)56. Our study identified findings related to SNV rs2032582 (c.2677 C > T/A), a variant located on chromosome 7 in exon 21, which was identified as a linkage disequilibrium region forming a haplotype57,58. This variant is characterized by the substitution of the amino acid alanine with serine or threonine, which may result in a significant reduction in ABCB1 expression, and consequently in P-gp function, supporting our findings59,60.
One potential explanation for these disparate results is the comparison of different population groups in the literature, which demonstrates variability in therapeutic protocols and patient profiles. The interpretation of genetic associations may be influenced by these variations, particularly in PGx studies of the Brazilian population, for which the available evidence is currently limited. However, the included patients were diagnosed by the same medical team, and adverse reaction assessments were performed by the same pharmacist together with a multidisciplinary team, which reduced the possibility of bias in the population sample. The primary limitations of this study are its small sample size, single-center design, and the low ethnic diversity of the cohort, underscoring its preliminary nature. Therefore, these findings should be interpreted as exploratory, rather than definitive.
The findings of our study suggest that the triallelic genetic variant of ABCB1, c.2677 C > T/A, may be associated with the absence of diarrhea and survival of patients with NSCLC treated with gefitinib. However, given the exploratory nature of these results, the observations should be interpreted with caution. Validation in larger, adequately powered cohorts is essential, and additional PGx and PK studies are required to clarify the underlying mechanisms.

Materials and methods

Study setting and population
This prospective cohort study was conducted at the Oncopneumology Outpatient Clinic of the Hospital das Clínicas of the State University of Campinas (Campinas, São Paulo, Brazil), a large public teaching hospital integrated into the Brazilian Unified Health System (Sistema Único de Saúde-SUS) and serving as a tertiary referral center for approximately 85 municipalities in the state of São Paulo. Genotyping was performed at the Clinical Pharmacy Laboratory at the Faculty of Pharmaceutical Sciences. The report was structured according to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement checklist for cohort studies (Supplementary Table 11).
This study was approved by the Research Ethics Committee of the State University of Campinas (CAAE: 17328619.90000.5404) and was performed in accordance with the Declaration of Helsinki. Informed consent was obtained from all patients included in the study. Patients were recruited between September 2019 and February 2023.
Patients of both sexes, aged 18 years or older, with a histological diagnosis of NSCLC and a mutation in exons 19, 20, and 21 of the EGFR gene, receiving a medical prescription for gefitinib 250 mg/day (combined or not with radiotherapy), and who completed adverse reaction assessments with the researcher or team were included. Patients who had conditions that made it impossible for them to participate in the study, refused to provide blood samples, changed the treatment protocol after initiation, or withdrew consent were excluded. The patients were followed up for approximately 12 weeks.

Data collection
The clinical and demographic data collected included age, sex, skin color/ethnicity (classified according to the Brazilian Institute of Geography and Statistics [IBGE]), Karnofsky index (KPS)61, smoking62, alcohol consumption63, presence of comorbidities, and the histological type of the tumor if present64. The skin color/ethnicity was self-reported by study participants, categorized by themselves according to five categories: “Preto” (Black), “Índigena” (Indigenous, Native Amerindian descent), “Pardo” (Brown), “Branco” (White), or “Amarelo” (Yellow, Asian descent)65. In addition to laboratory tests, medications used were verified to avoid drug interactions and disease progression.

Adverse reaction assessment
The adverse reactions evaluated were: (A) dermatological reactions such as tegumentary or acneiform rash, maculopapular rash, and hyperpigmentation; (B) gastrointestinal symptoms, including nausea, vomiting, and diarrhea; and (C) hepatic reactions such as hypoalbuminemia and increased levels of gamma-glutamyl transferase (GGT), alkaline phosphatase (ALP), alanine aminotransferase (ALT), and aspartate aminotransferase (AST). Dermatological reactions were assessed using questionnaires and photographic records. Gastrointestinal reactions were evaluated using questionnaires, and hepatic reactions were monitored via laboratory tests conducted during consultations or follow-up visits. All assessments and grading of adverse reactions were performed according to the Common Terminology Criteria for Adverse Events (CTCAE), version 5.066. Supplementary Table 12 describes the values used to assess hepatic adverse drug reactions (ADRs). The highest grade in each evaluation was used for biostatistical studies, and only ADRs with a frequency of 30% or higher were included in association studies with SNVs.

Genetic variants and genotyping
Genomic DNA was extracted and purified from peripheral blood using the Wizard® Genomic DNA Purification Kit (Promega, Madison, WI, USA). DNA concentration and integrity were verified by analyzing the A260/280 ratio using a NanoDrop 1000 Spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). Definitive quantification was performed using Quantus fluorometer® and the respective kit (QuantiFluor® dsDNA System, Promega, Madison, WI, USA). All samples were identified, adjusted to 20 ng/µL, and then stored at − 20 °C until further use.
ABCB1 rs1128503 (c.1236 C > T – p.Gly412=, Assay ID: C_7586662_10), rs1045642 (c.3435 A > G – p.lle1145=, Assay ID: C_7586657_20), and rs2032582 (c.2677 C > T/A - p.Ser893Ala, Assay ID: C_11711720D_40); ABCG2 rs2232242 (421G > T – Gln141Lys, Assay ID: C_15857904_20); EGFR rs2227983 (R497K – Arg497Lys, Assay ID: C_16170352_20) and rs2293347 (c.2982 C > T – D994D, Assay ID: C_15970737_20) were genotyped using TaqMan® assay (Thermo Fisher Scientific, Waltham, MA, USA).
The amplification reaction mixture (10 µL) contained 2 µL DNA (10 ng/µL), 5 µL TaqMan® Genotyping Master Mix, 2.5 µL DNAse-RNAse-free water, and 0.5 µL TaqMan™ SNV Genotyping assays. Real time polymerase chain reaction (qPCR) was performed using Rotor-Gene Q 5plex HRM System (Qiagen, Hilden, Germany) with an initial phase of two minutes at 60 °C to acclimatize the reagents, an enzymatic activation phase of 10 min at 95 °C, followed by 50 cycles of denaturation, annealing and extension at 95 °C for 15 s and 60 °C for one minute. Allelic discrimination was performed using Rotor-Gene Q software (version 2.3.5; QIAGEN; available at: https://www.qiagen.com/ca/resources/resourcedetail?id=9d8bda8e-1fd7-4519-a1ff-b60bba526b57&lang=en), and allelic calls were assigned manually to determine the genotype of each sample. The threshold was set manually above the No Template Control (NTC) baseline and within the exponential phase, using known genotype controls as references. Genotypes were automatically called based on the normalized endpoint fluorescence from two detection channels (Cycling A. Green and Cycling A. Yellow). Genotyping quality control included per-SNV call rate calculation, assessment of Hardy-Weinberg equilibrium (HWE) in controls, and verification of duplicate-sample concordance. Variants with call rates below 95% or showing extreme deviation from HWE (p < 1 × 10⁻⁴) were flagged for further inspection. Positive and negative controls were used in all experiments. As internal controls, 10% of the DNA samples were subjected to double genotyping at different times, with no losses and 100% concordance with previously obtained results. Duplicate samples demonstrated > 99% genotype concordance, indicating high analytical reliability.

Survival analysis
To determine overall survival (OS), the date of diagnosis, the date of death (as documented in medical records), and/or the patient’s last contact during an outpatient consultation were considered through the end of the study. OS was measured from the time of diagnosis until death or the last follow-up.

Statistical analysis
Statistical analyses were performed using Statistical Analysis System (SAS) (version 9.4; 2002 − 212; SAS Institute Inc., Cary, NC, USA). The normality of continuous variables was assessed using the Shapiro-Wilk test. Continuous variables with a normal distribution are expressed as mean ± standard deviation (SD), whereas variables with non-normal distributions are summarized using the median and interquartile range (IQR). Participants who self-identified as Black, Brown, Yellow, or Indigenous were grouped under the category “non-white,” while those who self-identified as White were classified as “white.” To compare proportions, the chi-square test or Fisher’s exact test was used when necessary. For analyses of ADRs, patients who experienced an ADR (grades 1–4) and those who did not (grade 0) were included. The models adopted for the studies of SNVs considered ancestral genotype versus altered genotype + heterozygote, and altered genotype versus ancestral genotype + heterozygote. The Hardy-Weinberg equilibrium (HWE) tool was used to evaluate all genetic variants. The Mann-Whitney U test was used to compare continuous measures between the two groups. To compare the numerical measures over time, the Friedman test was applied, followed by the Dunn test, if necessary. Multiple Cox regression analysis was used to assess factors associated with ADRs and OS. The variable selection process was performed stepwise. The following variables were used in multivariate regression analysis: age, sex, skin color, smoking status, and alcoholism status. The survival distribution was estimated using the Kaplan-Meier method, with the number of participants at risk reported at relevant time points. The proportional hazard assumption of the Cox model was assessed using Schoenfeld residuals. The significance level adopted for statistical tests was set at 5%. As this was an exploratory study, no multiple-testing correction (e.g., Bonferroni or FDR) was applied. The main objective was hypothesis generation, and strict multiplicity adjustments could substantially reduce the statistical power, given the correlation between variants due to linkage disequilibrium. Therefore, the unadjusted p-values were reported67.

Supplementary Information
Below is the link to the electronic supplementary material.