Pharmacogenomic profiling of the efficacy of gemcitabine monotherapy in metastatic pancreatic cancer: Subgroup analysis of the GENESECT study

Abstract

Aim

In the GENESECT study, no significant gemcitabine (GEM) metabolism‐related germline genetic polymorphisms (GPs) were identified because approximately 70% of patients received combination therapy with nab‐paclitaxel, which has metabolic pathways different from GEM. The present study was a sub‐analysis including only GEM monotherapy patients.

Methods

Of 159 patients analysed in the GENESECT study, only GEM‐monotherapy patients were selected. The endpoints included carbohydrate antigen 19‐9 (CA19‐9) response (reduction ≥ 50% from the pretreatment level at 8 weeks), progression‐free survival (PFS), and overall survival (OS) in relation to GPs. Analysed genes included those encoding GEM transporters and metabolic enzymes (SLC29A1, ABCC5, CDA, DCK, RRM1/2) and a tumour cell proliferation enzyme (COX‐2). Screening and significance levels were set at 10% (univariable) and 5% (multivariable), respectively.

Results

Data for 50 patients were analysed. Genotype AA of ABCC5 1146A>G (rs7636910) showed a higher CA19‐9 response rate than AG/GG (52.9% vs. 21.2%, p = 0.023) and remained significant in multivariable analysis (odds ratio: 6.255, p = 0.014). It also tended to be associated with longer PFS, but not significantly (hazard ratio [HR]: 0.464, p = 0.074). By contrast, although CA19‐9 responses were not significantly associated with DCK −1205T>C (rs4694362), the TT/TC genotype was significantly associated with prolonged PFS compared with CC (median 127 vs. 48 days, p = 0.002), and this association remained significant in multivariable analysis (HR: 0.153, p = 0.028). No GPs were significantly associated with OS.

Conclusion

ABCC5 and DCK might be related to the efficacy of GEM treatment, but further research is needed.

What is already known about this subject

• Gemcitabine shows substantial interindividual variability in efficacy in clinical practice.

• Genetic polymorphisms in enzymes and transporters involved in gemcitabine disposition have been proposed as contributors, but clinical evidence remains inconsistent.

• Real‐world data linking pharmacogenetic variability to clinically relevant outcomes are limited, warranting further study.

What this study adds

• This study identifies specific gemcitabine‐related genetic polymorphisms associated with treatment response in a real‐world clinical setting.

• The findings support the biological plausibility of pharmacogenetic influences on gemcitabine pharmacology.

• These results provide a foundation for future risk stratification and personalized gemcitabine therapy.

1. INTRODUCTION

Gemcitabine (GEM) is a key drug in chemotherapy for metastatic pancreatic cancer (mPC). According to major clinical guidelines, GEM monotherapy is recommended as an effective treatment option for patients who are ineligible for nab‐paclitaxel (nab‐PTX) plus GEM or FOLFIRINOX. ¹ , ² Pancreatic cancer is a malignancy that predominantly affects the elderly. ³ , ⁴ Although intensive combination therapies have improved outcomes in mPC, their use is often not feasible in older or more frail patients due to toxicity concerns. As the proportion of elderly individuals is expected to be substantially higher in many countries by 2040, ⁵ treatment strategies tailored specifically for ageing populations are needed.

GEM treatment is characterized by considerable inter‐individual variability in terms of both efficacy and toxicity. Prognostic factors include performance status (PS), disease stage, C‐reactive protein (CRP), neutrophil‐to‐lymphocyte ratio (NLR) and carbohydrate antigen 19‐9 (CA19‐9) levels. ⁶ , ⁷ In particular, a ≥50% reduction in CA19‐9 level during treatment is associated with favourable prognosis. ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹² , ¹³

Germline genetic polymorphisms (GPs) in GEM transport and metabolism‐related genes have recently been investigated as potential predictors of therapeutic response and survival. Key candidate genes include hENT1 (SLC29A1), cytidine deaminase (CDA), deoxycytidine kinase (DCK) and ribonucleotide reductase M1 (RRM1), which may affect GEM pharmacokinetics, efficacy and toxicity. ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ , ²¹ However, these previous studies were conducted in neoadjuvant, adjuvant or chemoradiotherapy settings and did not focus on mPC.

In our previous GENESECT study, ²² we analysed the relationship between GEM‐related GPs and treatment outcomes in mPC patients. However, as approximately 70% of GENESECT participants received combination therapy with nab‐PTX, it was difficult to isolate the effects of GEM monotherapy.

Due to the growth of the elderly population and increasing healthcare costs, identifying predictive biomarkers that enable optimization of treatment strategies is critical—not only for personalized medicine but also for efficient allocation of healthcare resources. The purpose of this study, therefore, was to determine the relationship between GPs and treatment efficacy of GEM for mPC.

2. MATERIALS AND METHODS

2.1. Study design and patients

This study was conducted as a sub‐analysis of the GENESECT study, ²² a multicentre observational pharmacogenomic study conducted in patients with mPC. Of 159 patients analysed in the original GENESECT cohort, only those who received GEM monotherapy were included in the present sub‐analysis. Patients who received combination chemotherapy, such as GEM plus nab‐PTX, were excluded in order to evaluate the pharmacogenomic impact of GEM monotherapy. All patients had pathologically confirmed pancreatic adenocarcinoma and received standard full‐dose GEM monotherapy as first‐line treatment. GPs were analysed using stored peripheral blood samples collected at the time of enrolment in the original GENESECT study. All clinical data were recorded on case report forms and collected by a clinical pharmacist who was blinded to the results of the genetic analysis.

The study protocol was approved by the Institutional Review Board of the Cancer Institute Hospital of the Japanese Foundation for Cancer Research (approval number: 2013‐1145) and by the ethics committees of all participating centres. This study was conducted in accordance and compliance with the Declaration of Helsinki and registered in the University Hospital Medical Information Network in Japan (UMIN) clinical trials registry (identification number UMIN 000012720). All participants provided written informed consent before enrolment. The authors confirm that the Principal Investigator for this paper is Hiroshi Ishii and that he had direct clinical responsibility for patients.

2.2. Assessment and outcome measures

Tumour assessments were performed approximately every 2 months using imaging studies (e.g., CT scans) in routine clinical practice. Tumour progression was evaluated by treating physicians according to RECIST criteria.

The endpoints of this sub‐analysis included associations between germline GPs and the following clinical outcomes: CA19‐9 response (defined as a ≥50% reduction from the pretreatment level at 8 weeks [±2 weeks]), progression‐free survival (PFS), and overall survival (OS). PFS and OS were calculated from the start date of GEM monotherapy.

PFS was defined as the time from treatment initiation to the first documentation of disease progression or death from any cause, whichever occurred first. OS was defined as the time from treatment initiation to death from any cause. Patients without events were censored at the date of last follow‐up. Patients were followed from treatment initiation until death or last follow‐up. To ensure sufficient follow‐up, all patients were observed for at least 1 year after enrolment of the last patient.

2.3. Pharmacogenomic analysis

Samples of whole venous blood (5 mL) were collected from all patients into ethylenediaminetetraacetic acid–2Na Venoject II tubes (Terumo) and stored at −30°C until DNA extraction. Leucocyte genomic DNA was extracted directly from the blood samples using a QIAamp DNA Blood Maxi kit (Qiagen). The following representative germline polymorphisms in seven genes reported in previous fundamental or clinical studies were selected for analysis: SLC29A1 −3797A>G (reference single nucleotide polymorphism [rs]11 752 118), SLC29A1 −3268_ −3249 deletion (rs11274220), DCK −1205T>C (rs4694362), ribonucleotide reductase large subunit (RRM1) −37A>C (rs12806698), RRM1 −524C>T (rs11030918), RRM1 1082C>A (rs183484), RRM1 2455A>G (rs9937), RRM1 2464G>A (rs1042858), RRM2 175T>G (rs1130609), CDA −33delC (rs3215400), CDA 79A>C (rs2072671), ABCC5 1243T>C (rs562), ABCC5 1146A>G (rs7636910), COX‐2 −1195A>G (rs689466), COX‐2 −765G>C (rs20417), and COX‐2 8473T>C (rs5275) (Figure S1). All polymorphisms, excluding the SLC29A1 −3268_−3249 deletion, CDA −33delC and CDA 79A>C, were detected by (PCR) restriction‐fragment length polymorphism analysis. The SLC29A1 −3268_−3249 deletion variant was identified via allele‐specific PCR analysis. The CDA −33delC and CDA 79A>C variants were detected using direct sequencing. The PCR primers, restriction enzymes, annealing temperature, digestion temperature and digestion time used for all reactions are listed in Table S1 of the GENESECT study supporting information. A total of 16 polymorphisms in seven genes were genotyped and classified into two groups as homozygous for the major or minor allele, as well as other genotypes. All research laboratory staff were blinded to the clinical observations.

2.4. Statistical analyses

Statistical analyses were conducted in two steps. In the first step, univariable analyses were performed to assess associations between GPs and each clinical endpoint. Fisher's exact test was used for categorical endpoint analyses, including CA19‐9 response. PFS and OS were evaluated using the Kaplan–Meier method, and differences between groups were assessed using the log‐rank test. GPs were examined using both dominant and recessive genetic models, and the selected germline polymorphisms were based on previous pharmacogenomic studies of GEM metabolism and transport. ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ , ²¹

In the second step, multivariable analyses were performed using logistic regression for CA19‐9 response, and Cox proportional hazards models were used for PFS and OS. Based on previous reports, PS, baseline CA19‐9 level, CRP and NLR ⁶ , ⁷ were included as covariates in the multivariable analyses of CA19‐9 response and PFS. For the OS analysis, these variables were included along with the presence or absence of second‐line chemotherapy. A screening threshold of 10% was applied in the univariable analyses to identify candidate variables, and only polymorphisms meeting this criterion were included in the multivariable models, whereas statistical significance in the multivariable analyses was set at 5% (two‐sided).

No missing data were present in the dataset analysed in this study. No adjustment for multiple testing was performed, given the exploratory nature of the study.

2.5. Nomenclature of targets and ligands

Key protein targets and ligands in this article are hyperlinked to corresponding entries in the Guide to PHARMACOLOGY (https://www.guidetopharmacology.org) and are permanently archived in the Concise Guide to PHARMACOLOGY 2021/22 (Alexander et al., 2021).

3. RESULTS

3.1. Baseline patient characteristics

Of the 159 patients initially enrolled in the study, 50 were ultimately included and analysed. The baseline characteristics of enrolled patients are summarized in Table 1. PS was 0 in 58% of patients and ≥1 in 42%. The median baseline CA19‐9 level was 5338 U/mL (range: 118–428 200 U/mL).

TABLE 1.

Baseline patient characteristics.

Characteristics (n = 50)	n	(%)
Age, years
Median (range)	73	54–87
Gender
Male	23	46
Female	27	54
ECOG performance status
0	29	58
1	18	36
2	3	6
Laboratory data (baseline)	Median	(range)
CA19‐9 (U/mL)	5338	118–428 200
Neutrophil count (/μL)	4362.5	1720‐10 837
Haemoglobin (mg/dL)	11.9	9–14.6
Platelet (/μL × 103)	203	104–361
AST (U/L)	24	12–88
ALT (U/L)	21.5	6–94
Total bilirubin (mg/dL)	0.6	0.29–2.8
Serum Creatinine (mg/dL)	0.635	0.34–1.11
CRP (mg/dL)	0.72	0.02–8.68

Abbreviations: ECOG, Eastern Cooperative Oncology Group; CA19‐9, carbohydrate antigen 19‐9; AST, aspartate aminotransferase; ALT, alanine transaminase; CRP, C‐reactive protein.

3.2. Associations between CA19‐9 response and GPs

In the univariable analysis, only ABCC5 1146A>G (rs7636910) met the screening threshold of 10%. The AA genotype (major allele) of ABCC5 1146A>G (rs7636910) was associated with a higher percentage of CA19‐9 responses compared with AG + GG (52.9% vs. 21.2%, p = 0.023). Other GPs did not meet the screening threshold for association with CA19‐9 response (Table S1).

In the multivariable analysis, ABCC5 1146A>G (rs7636910) was identified as a statistically significant factor associated with CA19‐9 response (odds ratio: 6.255, 95% confidence interval [CI]: 1.460–26.795, p = 0.014) (Table 2).

TABLE 2.

Associations between CA19‐9 response and GPs (multivariable analysis).

Variate		n	Odds ratio (95% CI)	p value
ECOG PS	0	29	1.108 (0.232–5.285)	0.898
	1 or 2	21	1
CA19–9 (U/mL)	≥5338	26	0.307 (0.068–1.397)	0.126
	<5338	24	1
CRP (mg/dL)	≥1	18	1.204 (0.276–5.247)	0.805
	<1	32	1
NLR	≥5	12	1.706 (0.356–8.176)	0.504
	<5	38	1
ABCC5 1146A>G (rs7636910)	AA	17	6.255 (1.460–26.795)	0.014*
	AG + GG	33	1

Abbreviations: ECOG PS, Eastern Cooperative Oncology Group performance status; CA19‐9, carbohydrate antigen 19‐9; CRP, C‐reactive protein; NLR, neutrophil‐to‐lymphocyte ratio.

^*Statistically significant difference.

3.3. Associations between PFS and GPs

In the univariable analysis, only ABCC5 1146A>G (rs7636910) and DCK −1205T>C (rs4694362) met the screening threshold of 10% (Figure 1). The AA genotype (major allele) of ABCC5 1146A>G (rs7636910) was associated with prolonged PFS compared with AG/GG (median, 154 vs. 119 days, p = 0.052). The TT + TC genotype of DCK −1205T>C (rs4694362) was associated with prolonged PFS compared with CC (minor allele) (median, 127 vs. 48 days, p = 0.002). Other GPs did not meet the screening threshold for association with PFS (Table S2).

FIGURE 1.

Association between progression‐free survival and genetic polymorphisms. Progression‐free survival was evaluated according to genetic polymorphisms using the Kaplan–Meier method, and differences between groups were assessed using the log‐rank test. Significant associations were observed for ABCC5 1146A>G (rs7636910) and DCK −1205T>C (rs4694362); results for other genetic polymorphisms are shown in Table S2. Abbreviations: MST, median survival time; 95% CI, 95% confidence interval.

In the multivariable analysis, only DCK −1205T>C (rs4694362) was identified as a statistically significant factor associated with PFS (hazard ratio [HR]: 0.153, 95% CI: 0.029–0.814, p = 0.028) (Table 3).

TABLE 3.

Associations between PFS and GPs (multivariable analysis).

Variate		n	Hazard ratio (95% CI)	p value
ECOG PS	0	29	0.852 (0.399–1.818)	0.678
	1 or 2	21	1
CA19‐9 (U/mL)	≥5338	26	2.082 (0.973–4.456)	0.059
	<5338	24	1
CRP (mg/dL)	≥1	18	0.878 (0.393–1.961)	0.750
	<1	32	1
NLR	≥5	12	0.693 (0.293–1.641)	0.404
	<5	38	1
ABCC5 1146A>G (rs7636910)	AA	17	0.464 (0.200–1.079)	0.074
	AG + GG	33	1
DCK −1205T>C (rs4694362)	TT + TC	47	0.153 (0.029–0.814)	0.028*
	CC	3	1

Abbreviations: ECOG PS, Eastern Cooperative Oncology Group performance status; CA19‐9, carbohydrate antigen 19‐9; CRP, C‐reactive protein; NLR, neutrophil‐to‐lymphocyte ratio.

^*Statistically significant difference.

3.4. Associations between OS and GPs

In the univariable analysis, only CDA −33delC (rs3215400) and COX‐2 −765G>C (rs20417) met the screening threshold of 10% (Figure 2). The CC + C/del genotype of CDA −33delC (rs3215400) was associated with prolonged OS compared with del/del (minor allele) (median, 263 vs. 129 days, p = 0.098). The GG (major allele) genotype of COX‐2 −765G>C (rs20417) was associated with prolonged OS compared with GC + CC (median, 235 vs. 114 days, p = 0.017). Other GPs did not meet the screening threshold for association with OS (Table S3).

FIGURE 2.

Associations between overall survival and genetic polymorphisms. Overall survival was evaluated according to genetic polymorphisms using the Kaplan–Meier method, and differences between groups were assessed using the log‐rank test. Significant associations were observed for CDA –33delC (rs3215400) and COX‐2 −765G>C (rs20417); results for other genetic polymorphisms are shown in Table S3. Abbreviations: MST, median survival time; 95% CI, 95% confidence interval.

In the multivariable analysis, no statistically significant GPs were detected, and only receiving second‐line chemotherapy was identified as a statistically significant factor associated with OS (HR: 0.344, 95% CI: 0.173–0.683, p = 0.002) (Table 4).

TABLE 4.

Associations between OS and GPs (multivariable analysis).

Variate		n	Hazard ratio (95% CI)	p value
ECOG PS	0	29	1.083 (0.530–2.213)	0.827
	1 or 2	21	1
CA19‐9 (U/mL)	≥5338	26	1.628 (0.793–3.342)	0.184
	<5338	24	1
CRP (mg/dL)	≥1	18	0.983 (0.459–2.108)	0.965
	<1	32	1
NLR	≥5	12	1.093 (0.507–2.355)	0.820
	<5	38	1
Presence of second‐line chemotherapy	Yes	30	0.344 (0.173–0.683)	0.002*
	No	20	1
CDA –33delC (rs3215400)	CC + C/del	42	0.528 (0.221–1.260)	0.150
	del/del	8	1
COX‐2 −765G>C (rs20417)	GG	46	0.387 (0.116–1.297)	0.124
	GC + CC	4	1

Abbreviations: ECOG PS, Eastern Cooperative Oncology Group performance status; CA19‐9, carbohydrate antigen 19‐9; CRP, C‐reactive protein; NLR, neutrophil‐to‐lymphocyte ratio; del, deletion.

^*Statistically significant difference.

4. DISCUSSION

In this study, conducted in the context of GEM monotherapy in mPC, a significant association was demonstrated between CA19‐9 response and ABCC5 1146A>G (rs7636910), as well as between PFS and DCK −1205T>C (rs4694362). These results are consistent with previously reported molecular mechanisms and clinical data and also provide various novel insights. These findings should be interpreted in the context of the original GENESECT study, ²² which included a broader population receiving both GEM monotherapy and combination therapies. In that study, most germline polymorphisms were not significantly associated with clinical outcomes, except for COX‐2 polymorphisms in relation to OS. In contrast, the present sub‐analysis focused exclusively on GEM monotherapy, which may have enabled a more homogeneous evaluation of pharmacogenomic effects and partly explains the differences in observed associations. DCK is a rate‐limiting enzyme responsible for the initial monophosphorylation step in the intracellular activation of GEM, and its expression is reportedly correlated with treatment outcome. ¹⁶ , ²³ In studies of adjuvant GEM therapy, high DCK expression was associated with prolonged disease‐free survival and OS. ¹⁶ , ²³ Moreover, animal experiments demonstrated a significant association between DCK expression, enzyme activity and GEM sensitivity, ²⁴ whereas reduced DCK expression was identified as part of the mechanism of GEM resistance. ²⁵ , ²⁶ In addition, a pharmacogenomic study of GEM‐based chemotherapy in metastatic breast cancer reported that the 6‐month PFS was significantly higher in patients with the TT genotype of DCK −1205T>C (rs4694362) (78.1%) compared with patients carrying the TC/CC genotypes (51.9%). ²⁷ Similarly, in studies of cytarabine, another nucleoside analogue activated by DCK phosphorylation, the CC genotype of DCK −1205T>C (rs4694362) was significantly associated with shorter OS compared with the TT/TC genotypes. ²⁸ The results of the present study, showing prolonged PFS in patients with the TT/TC genotypes compared with CC, are consistent with these prior observations. These findings suggest that the T allele positively affects DCK expression or enzymatic activity and thus provide novel insights, specifically in the context of mPC.

Overexpression of the efflux transporter ABCC5 (MRP5) has been implicated in GEM resistance. ²⁹ , ³⁰ , ³¹ In our study, the AA genotype of ABCC5 1146A>G (rs7636910) was associated with a higher CA19‐9 response compared with AG/GG, but no significant correlation with PFS was identified. The allelic influence on MRP5 expression in tumour tissues remains unclear, however, highlighting the need for further experimental validation. From a mechanistic perspective, ABCC5 encodes the efflux transporter MRP5, and increased transporter activity may reduce intracellular GEM availability, thereby potentially decreasing treatment efficacy. ³⁰ Experimental studies have demonstrated that MRP5 expression is associated with GEM resistance, supporting its role in modulating drug sensitivity. ³⁰ Although the precise functional consequences of rs7636910 in the context of GEM therapy remain unclear, this variant may influence treatment response through its potential effects on ABCC5 function or expression. However, even if such mechanisms influence early treatment response, their impact may not necessarily translate into prolonged clinical outcomes. The observed association with CA19‐9 response, but not with PFS, may therefore reflect either effects on early tumour dynamics or limitations related to the small sample size and model instability. Accordingly, these findings should be interpreted with caution.

This study has several limitations. First, it was a subgroup analysis of the GENESECT study ²² and involved a limited sample size; thus, residual confounding cannot be fully excluded. In addition, the relatively small sample size and the inclusion of multiple covariates may have limited the statistical power and stability of the multivariable models. Model assumptions were not formally assessed, which may further limit the robustness of the findings. Therefore, non‐significant findings should be interpreted with caution, as they may reflect insufficient power rather than a true absence of association. With regard to OS, the presence and type of second‐line treatment may have had a substantial impact, which could explain why no GPs were consistently associated with OS. In addition, the inclusion of second‐line therapy as a fixed covariate in the OS analysis may introduce immortal‐time bias, as the timing of treatment initiation was not modelled as a time‐dependent variable. Therefore, the impact of second‐line therapy on OS should be interpreted with caution. Similarly, the lack of concordant associations across CA19‐9 responses, PFS, and OS might also be attributable to such confounding factors. In addition, the functional impact of the identified polymorphisms on mRNA or protein expression remains unclear. Of note, DCK −1205T>C (rs4694362) is located in an intronic region ¹⁸ and does not cause an amino acid substitution. Therefore, its effect, if any, may be mediated indirectly through mechanisms involving differences in transcriptional regulation or splicing efficiency. Further functional validation, including eQTL analyses and reporter assays, will be essential to confirm these findings. Furthermore, the number of patients in certain genotype subgroups, particularly the homozygous group for DCK −1205T>C (rs4694362), was very small, which may affect the robustness of the findings. This is partly attributable to ethnic variability in allele frequencies, as this genotype is relatively uncommon in the Japanese population. In addition, multiple comparisons were performed without adjustment for multiple testing, and the findings should therefore be interpreted with caution due to the potential for Type I error and warrant further validation in independent cohorts.

The germline polymorphism DCK −1205T>C (rs4694362) exhibits ethnic variability. Specifically, the T allele is predominant in East Asians (T = 78.9%, C = 21.1%), whereas the C allele is predominant in Africans (T = 20.2%, C = 79.8%), and Europeans show an intermediate distribution (T = 61.1%, C = 38.9%). ³² Such germline differences may contribute to regional variations observed in clinical trial outcomes. For example, in the Japanese phase II/III JCOG1611‐GENERATE trial, GEM plus nab‐PTX demonstrated superior OS compared with modified FOLFIRINOX, ³³ whereas multiple studies in Western countries reported a tendency for FOLFIRINOX to outperform GnP. ³⁴ , ³⁵ Furthermore, a systematic review and meta‐analysis addressing pharmacogenomic ethnicity showed that the Western subgroup derived greater survival benefit from FOLFIRINOX (HR: 0.84, 95% CI: 0.74–0.95, p = 0.006), whereas the Asian subgroup benefited more from GnP (HR: 1.29, 95% CI: 1.03–1.60, p = 0.030). ³⁶ These findings are consistent with the aforementioned ethnic variability in DCK −1205T>C (rs4694362). In addition, the GEST trial, which evaluated GEM monotherapy as first‐line treatment, demonstrated that the median OS differed between countries: 8.9 months (95% CI: 8.3–9.7 months) in Japan and 5.3 months (95% CI: 4.2–10.8 months) in Taiwan, despite both populations belonging to the East Asian subgroup. ³⁷ The Japanese population exhibited a predominance of the T allele (T = 80.4%, C = 19.6%), which is comparable to the East Asian average, suggesting that other GPs may also affect the efficacy of GEM. For instance, SLC29A1 (encoding hENT‐1) reportedly contributes to GEM resistance when its expression is reduced. ²⁵ Future studies should therefore explore combinations of multiple polymorphisms (haplotypes) to better identify subgroups of patients who may benefit from GEM‐based therapy.

Although several limitations and future challenges remain, the present findings provide important insights that may contribute to the personalization of pancreatic cancer treatment. Pancreatic cancer is predominantly a disease of the elderly, ³⁸ and a lack of robust evidence in patients aged 70 years or older remains a major concern. Therefore, the accumulation of evidence in this population is essential. In parallel, it is necessary to identify subgroups of patients who may benefit from low‐burden therapies such as GEM monotherapy. The present results support the concept that GEM‐based regimens could be prioritized for patients likely to benefit, whereas alternative regimens should be considered for those unlikely to respond, thereby facilitating the implementation of individualized treatment strategies.

In conclusion, DCK and ABCC5 GPs might be related to the efficacy of treatment with GEM, but further research is needed to confirm this potential association.

AUTHOR CONTRIBUTIONS

Takashi Yokokawa: Conceptualization; methodology; software; investigation; data curation; project administration; resources; visualization; writing original draft and writing—review and editing; had full access to the data and takes responsibility for the integrity of the data and adherence to the study protocol. Daiki Tsuji: Conceptualization; methodology; software; data curation; supervision; formal analysis; validation; investigation; funding acquisition; project administration; resources; visualization and writing—review and editing; had full access to the data and takes responsibility for the integrity of the data and adherence to the study protocol. Masahiro Hatori: Methodology; data curation; formal analysis and writing—review and editing. Kazuyoshi Kawakami: Methodology; investigation and writing—review and editing. Masahiko Nakao: Methodology and investigation; Takuya Yano: Methodology and investigation. Yuichiro Arakawa: Methodology and investigation. Keisuke Matsuo: Methodology and investigation. Yasukata Ohashi: Methodology and investigation. Masato Ozaka: Methodology and supervision. Naoki Sasahira: Methodology and supervision. Yasuhiko Sakata: Methodology and investigation. Yuki Kogure: Methodology and investigation. Shinya Tamaki: Methodology and investigation. Atsushi Wada: Methodology and investigation. Yusuke Taki: Methodology and investigation. Wataru Suzuki: Investigation. Kazuo Kobayashi: Investigation. Hiroshi Ishii: Methodology and supervision. Kunihiko Itoh: Supervision. Masakazu Yamaguchi: Supervision. Kenichi Suzuki: Methodology; supervision and writing—review and editing.

CONFLICT OF INTEREST STATEMENT

Takashi Yokokawa reports support for professional activities from Ono Pharmaceuticals outside the submitted work. Masahiro Hatori reports support for professional activities from Bayer and Merck & Company, Inc., outside the submitted work. Wataru Suzuki reports support for professional activities from Daiichi Sankyo outside the submitted work. Hiroshi Ishii reports support for professional activities from Chugai Pharmaceutical outside the submitted work.

Supporting information

Figure S1. Metabolic pathway of gemcitabine and analysed genes.

Table S1. Associations between CA19‐9 response and GPs (univariable analysis).

Table S2. Associations between PFS and GPs (univariable analysis).

Table S3. Associations between OS and GPs (univariable analysis).

Yokokawa T, Tsuji D, Hatori M, et al. Pharmacogenomic profiling of the efficacy of gemcitabine monotherapy in metastatic pancreatic cancer: Subgroup analysis of the GENESECT study. Br J Clin Pharmacol. 2026;92(10):3397‐3406. doi: 10.1002/bcp.70611

DATA AVAILABILITY STATEMENT

The data sets used and analysed in this study are available from the corresponding author upon reasonable request.