Correlation of toxicities and efficacies of pemetrexed with clinical factors and single-nucleotide polymorphisms: a prospective observational study

The inclusion criteria for the patients were as follows: patients 1) with histologically or cytologically proven non-Sq NSCLC and MPM, 2) planning to use pemetrexed-containing chemotherapy, 3) naïve or previously treated with chemotherapy, 4) with ECOG-PS of 0–2, 5) age ≥ 20 years, 6) with adequate organ function measured as leukocyte count ≥ 3,000 /μL, hemoglobin concentration ≥ 10.0 g/dL, platelet count ≥ 100,000 /μL, total bilirubin level ≤ 2.0 mg/dL, transaminase levels ≤ 100 IU/L, serum creatinine level ≤ 1.5 mg/dL, SpO₂ (percutaneous oxygen saturation) ≥ 90% in ambient air, 7) expected to survive at least 12 weeks from the start of treatment, and 8) who provided written informed consent.

The exclusion criteria were as follows: 1) contraindication to pemetrexed; 2) active concomitant malignancy; 3) uncontrolled comorbidity related to the following systems: respiratory (bronchial asthma, chronic obstructive lung disease, interstitial pneumonia, pulmonary fibrosis, and respiratory failure), cardiac (cardiac failure and arrhythmias), renal (renal failure with dialysis), neurological (cerebral vascular disorder), hepatic disease (Child classification C with liver cirrhosis, fulminant hepatitis, and liver failure); 4) symptomatic brain metastases; 5) active infectious disease; 6) pregnancy or lactation; 7) massive pleural effusion; 8) continuous systemic administration of steroids for comorbid diseases; 9) acute myocardial ischemia within 6 months or unstable angina pectoris; and 10) any other condition judged by the medical oncologist as rendering the patient unsuitable for inclusion.

The patients underwent blood tests before premedication administration with vitamin supplements to mitigate toxicity at the start and end of the first cycle. We evaluated adverse events (AEs) and responses during the two cycles. Premedication consisted of B12 and FA; oral administration of FA (500 μg/day) was continued from 7 days before the first dose of pemetrexed until 21 days after the last dose of pemetrexed, whereas 1000 μg B12 was administered intramuscularly, 1 week before the first dose of pemetrexed and every three cycles thereafter. Subsequently, B12 may be injected on the same day as pemetrexed treatment. According to routine clinical practice, which was the same as the prescribing information [13], patients received 500 mg/m² pemetrexed intravenously on day 1 of each 21-day cycle, with or without other antineoplastic agents [14]. If each physician deemed it necessary to prescribe a reduced dose of chemotherapy according to clinical manifestations such as age, ECOG-PS, and modestly impaired organ function, it was acceptable for patients to be administered reduced doses even from the first cycle. The patients continued to receive chemotherapy until progressive disease (PD) [15] or unacceptable toxicity occurred. We also investigated progression-free (PFS) and overall survival (OS).

The primary endpoint of analyses was to determine the relationship between clinical factors or SNPs and toxicities of pemetrexed-containing chemotherapy. The secondary endpoint was identifying the relationship between clinical factors or SNPs and chemotherapy efficacy.

Before premedication administration, several clinical parameters, such as serum Hcy, B12, FA, prothrombin time (PT), activated partial thromboplastin time (aPTT), and fibrinogen, were investigated. At the start and end of the first cycle, clinical investigations of serum Hcy, serum B12, serum FA, PT, aPTT, fibrinogen, transferrin (Tf), transthyretin (TTR, pre-albumin), retinol-binding protein (RBP), complete blood cell count, and blood chemistry were performed.

Pretreatment evaluation included the following factors: complete medical history, assessment of ECOG-PS, chest X-rays, computed tomography (CT) of the chest to the pelvis, magnetic resonance imaging of the brain, and whole-body bone scintigraphy or 18-fluorodeoxyglucose on Positron Emission Tomography-CT. CT scanning was conducted after two cycles of pemetrexed-containing chemotherapy. According to the Response Evaluation Criteria in Solid Tumors 1.1 [15], tumor lesions were assessed at baseline, every two cycles until 6 months, and then every 2–4 cycles after 6 months until PD. Treatment-related AEs were assessed per the National Cancer Institute Common Terminology Criteria version 4.0 [16]. PFS was defined as the time from the start of pemetrexed-containing chemotherapy to earlier disease progression or death from any cause. OS was defined as the time from the beginning of pemetrexed-containing chemotherapy to death for any reason.

The target sample size was calculated as follows. We aimed to detect more than two significant risk factors for Grade (G) 4 neutropenia in this study. When 4% of patients treated with pemetrexed-containing chemotherapy would experience G4 neutropenia, the accrual of at least 50 patients was needed, assuming that a significant factor of 20 to 25 would be identified for the outcome in the multivariate logistic regression analysis, in general. When we consider a 20% dropout rate, at least 60 patients without missing values would be required.

For continuous variables in efficacy analyses, receiver operating characteristic (ROC) curves were analyzed using the SigmaPlot software version 14.5 (Systat Software, Inc., San Jose, CA, USA) [23]. In the toxicity analyses, the median was used to dichotomize continuous variables. Logistic regression analysis was performed to assess risk factors for AEs in the 1st cycle. PFS and OS were estimated using the Kaplan–Meier method. Cox regression analysis was performed to investigate favorable factors for PFS and OS.

We performed crude and propensity score (PS)-adjusted analyses to reduce potential confounding and bias in the observational studies [24]. We constructed a PS for the median Hcy level at the start of the first cycle of chemotherapy using a logistic regression model for each patient for both toxicity and efficacy evaluation. The PS for adjustment of toxicities was composed using body surface area, age, histology, relapse after surgical resection, ECOG-PS, comorbidities, chemotherapy line, reduced dosage requirement status, and duration of premedication as independent variables, and the C-statistic of the ROC curve analysis was 0.808. The PS for the adjustment of efficacy constituted the following variables: independent variables such as bevacizumab combination, chemotherapeutic regimen, relapses after radical radiotherapy or surgical resection, ECOG-PS, driver mutation, BMI, chemotherapy line, reduced dosage requirement status, duration of premedication, and comorbidities (respiratory, gastrointestinal, neuropsychiatric, urologic-renal, and dermato-orthopedic), with a C-statistic of 0.823.

We performed model selection for multivariate analysis according to a previously described method [23, 24]. Briefly, we identified variables with significance at P < 0.1 in the univariate analysis. Spearman’s rank test and clinically clarified dependent variables were used to exclude dependent variables from the selected variables. A correlation coefficient (rho) of > 0.3 as the absolute value with the Spearman’s rank test indicated a significant association. When many variables were detected in univariate analyses, we used rho > 0.2 at the absolute value as a significant association to converge the parameters correctly. Models were constructed using only independent variables as candidates. Akaike’s Information Criterion (AIC) was used to select the best model from among the candidate models. After choosing the best model, we proceeded with stepwise variable deletion to minimize the AIC. The final model was composed only of variables that achieved the minimum AIC during the elimination process. In the final multivariate analysis using a simultaneous method, statistical significance was determined at P < 0.05 with a two-sided test. Moreover, a sensitivity analysis was conducted using a model with one less variable than that of the final model.

All analyses were performed using the SPSS Statistics software version 27 (IBM, New York, USA). We used the “Genetics: Population Genetics” package version 1.3.8.1.3 developed by Gregory Warnes in R version 4.2.1 to calculate the Hardy–Weinberg equilibrium and linkage disequilibrium (LD), expressed as D′ and r² [25].

From October 2012 to December 2019, 118 patients provided informed consent for the observational cohort study (Fig. 1). Of these, nine received chemotherapy without pemetrexed, and one patient received osimertinib. ECOG-PS of two patients did not meet the criteria before chemotherapy. Thirty-five could not agree to the SNP examination even though they agreed to participate in the observational cohort study for treatment. Toxicity was analyzed in 71 patients. Seventeen SNPs were combined with clinical and metabolite information for this examination (Table 1). Excluding mesothelioma and postoperative adjuvant chemotherapy, 63 patients had non-Sq NSCLC with stage IIIB to IV (UICC ver.8) or recurrence after surgical treatment for efficacy analyses.

The key patient characteristics are summarized in Table 2. The median age of the patients was 72 years, and 22 female patients were included. Ninety-six percent of the patients had an ECOG-PS of 0 to 1, the median smoking index (SI) was 440, and 24 reported never smoking. Histologically, 67 patients had non-Sq NSCLC and four had MPM. Comorbidities were noted in 61 patients. The median BMI was 21.7 kg/m². Fifty-five were chemotherapy-naïve patients. Sixty patients received platinum combined with pemetrexed. Eighteen patients with non-sq NSCLC were treated in combination with bevacizumab and seven with pembrolizumab. Twelve patients were treated with a reduced dose of pemetrexed in the first cycle. The median duration of premedication was 11 days. We also evaluated nutrition testing. The median Tf, TTR, and RBP levels were normal.

Table 1 lists the SNPs examined in this study. Twelve of the seventeen SNPs were heterozygous. The median minor allele frequency was 0.33, with a range of 0.02–0.48. Their P-values via Fisher’s exact tests were greater than 0.05, indicating that they were in the Hardy–Weinberg equilibrium. A weak LD was suggested between FPGS_a and FPGS_b (D’ = 0.995 and r² = 0.042). Moderate LD was suggested between MTHFR_a and MTHFR_b (D’ = 0.998 and r² = 0.123, respectively). A strong LD was suggested between SLC19A1_a and SLC19A1_b (D’ = 0.999 and r² = 0.918).

Seventy-one patients underwent toxicity analysis (Fig. 1). When patients experience severe toxicities in the initial cycle of cytotoxic chemotherapy, they are given a reduced dose to tolerate toxicities in subsequent cycles [13]. Therefore, we used AEs during the first cycle as toxicities in this study (Table 3 and Supplementary Table S1 in Additional file 2). Among all causalities, 58% were scored > G3. Hematological AEs > G3 constituted 38%. G3–G4 hematologic toxicities were neutropenia (31.0%), leukopenia (18.3%), thrombocytopenia (14.1%), and anemia (9.9%) (Tables 3 and S1). Non-hematological AEs > G3 constituted 37% of the AEs. Common G2–G4 non-hematologic toxicities were anorexia (39.4%), serum alanine aminotransferase (ALT) level elevation (19.7%), FN (18.3%), and nausea (14.1%) (Table S1).

Multivariate logistic regression analysis revealed that crude low-risk factors for G3–G4 hematological AEs in the first cycle were high white blood cell (WBC) count before pemetrexed initiation, adapted status for using standard dosage, and C/C genotype group of MTHFR_a [677 C > T] (Supplementary Table S2(a)). PS-adjusted low-risk factors for G3–G4 hematological AEs were associated with WBC count ≥ 6120 /μL before pemetrexed exposure [odds ratio (OR): 0.06; 95% confidence interval (CI): 0.01–0.35; P = 0.001], adapted status for using standard dosage [OR: 0.09; 95% CI: 0.01–0.91; P = 0.04], C/C genotype group of MTHFR_a [OR: 0.13; 95% CI: 0.02–0.92; P = 0.04], serum B12 level ≥ 486 pg/mL before premedication [OR: 0.13; 95% CI: 0.03–0.73; P = 0.02], and serum FA level ≥ 15.8 ng/mL before pemetrexed exposure [OR: 0.15; 95% CI: 0.03–0.89; P = 0.04] (Supplementary Table S2(b) and Fig. 2a).

For non-hematological AEs using multivariate logistic regression analysis, the C/T genotype group of MTHFR_a, high serum RBP level before pemetrexed treatment, high serum B12 level before premedication, and high red blood cell (RBC) count before pemetrexed treatment (Supplementary Table S3(a)) were the crude low-risk factors for G2–G4 non-hematological AEs in the first cycle. After adjusting for PS, four variables constituted the significant low-risk factors for G2–G4 non-hematological AEs: C/T genotype group of MTHFR_a [OR: 0.07; 95% CI: 0.009–0.54; P = 0.01], TTR level ≥ 21.5 mg/dL before pemetrexed treatment [OR: 0.06, 95% CI: 0.006–0.50; P = 0.01], serum Hcy level < 11.8 nmol/mL before premedication [OR: 0.03; 95% CI: 0.003–0.30; P = 0.003], and A/G genotype group of SLC19A1_a [IVS2(4935) G > A, OR: 0.06; 95% CI: 0.005–0.58; P = 0.02] (Supplementary Table S3(b) and Fig. 2b).

In the case of 14% or more G3–G4 hematological AEs or G2–G4 non-hematological AEs (Supplementary Table S1), we examined the risk factors for each AE. All these risk factors were analyzed using multivariate logistic regression analysis adjusted for PS (Supplement Fig. S1 in Additional file 3). Adapted status for using standard dosage [OR: 0.03; 95% CI: 0.002–0.37; P = 0.007] (Supplement Fig. S1(a)) served as the sole significant low-risk factor for G3–G4 leucopenia. A WBC count ≥ 6120/μL before pemetrexed exposure [OR, 0.05; 95% CI: 0.008–0.35; P = 0.002] was a low-risk factor for G3–G4 neutropenia (Supplement Fig. S1(b)). Higher hemoglobin levels before pemetrexed treatment [OR: 0.03; 95% CI: 0.002–0.33; P = 0.005] and lower Hcy levels before pemetrexed treatment [OR: 0.58; 95% CI: 0.37–0.90; P = 0.02] were low-risk factors for G2–G4 anemia (Supplement Fig. S1(c)). Low-risk factors for G3–G4 thrombocytopenia were as follows: a higher FA level before pemetrexed exposure [OR: 0.67; 95% CI: 0.47–0.97; P = 0.03], WBC count ≥ 6120 /μL before pemetrexed exposure [OR: 0.009; 95% CI: 0.0001–0.71; P = 0.03], and C/C or T/T genotype groups of MTHFR_a [OR: 0.05; 95% CI: 0.003–0.73; P = 0.03] (Supplement Fig. S1(d)). Low-risk factors for G2–G4 ALT elevation included higher FA levels before premedication [OR: 0.45, 95% CI: 0.25–0.83; P = 0.01], presence of comorbidities [OR: 0.01, 95% CI: 0.0001–0.61; P = 0.03], and a serum ALT level < 16 U/L before pemetrexed exposure [OR: 0.03, 95% CI: 0.002–0.51; P = 0.02] (Supplement Fig. S1(e)). Low-risk factors for G2–G4 anorexia included a platelet count < 23.8 × 10⁴ /μL before pemetrexed exposure [OR: 0.22; 95% CI: 0.05–0.98; P = 0.046] and serum B12 level ≥ 486 pg/mL before premedication [OR: 0.18, 95% CI: 0.04–0.74; P = 0.02] (Supplement Fig. S1(f)). Low-risk factors for G2–G4 nausea were the C/T or C/C genotype groups of FPGS_b [2572 C > T, OR: 0.11; 95% CI: 0.01–0.76; P = 0.03] and a B12 level ≥ 486 pg/mL before premedication [OR: 0.08, 95% CI: 0.01–0.66; P = 0.02] (Supplement Fig. S1(g)). Low-risk factors for G3–G4 FN included adapted status for using standard dosage [OR: 0.006, 95% CI: 0.0003–0.16; P = 0.002], RBC count ≥ 400 × 10⁴ /μL before pemetrexed exposure [OR: 0.02, 95% CI: 0.001–0.44; P = 0.01], neutrophil count ≥ 4127 /μL before pemetrexed exposure [OR: 0.06; 95% CI: 0.005–0.80; P = 0.03], and the A/C genotype group of DHFR [680 C > A, OR: 0.02; 95% CI: 0.0007–0.39; P = 0.01] (Supplement Fig. S1(h)). We performed sensitivity analyses for all the above final models, and each sensitivity analysis demonstrated the same tendency as the corresponding final model.

The efficacy analyses for non-Sq NSCLC included 63 patients (Fig. 1). The objective response rate was 28.2%. The final survival assessment was conducted on September 30, 2021. The median PFS was 6.8 months, and its 95% CI ranged from 4.1 to 9.5 (Fig. 3a). The median OS was 33.3 months, ranging from 24.1 to 42.5 (Fig. 4a).

We conducted crude and PS-adjusted multivariate Cox regression analyses to identify the factors favorable for PFS. From crude multivariate analyses, pemetrexed combined with platinum, regimen combined with bevacizumab, B12 level < 1136 pg/mL before pemetrexed treatment, FA level ≥ 9.3 ng/mL before premedication, A/G or A/A genotype group of BHMT (742 G > A), and neutrophil to lymphocyte ratio < 3.51 before pemetrexed treatment were identified to be the favorable factors for PFS (Supplementary Table S4(a)). As the final result of PS-adjusted multivariate analyses, the favorable factors for PFS were pemetrexed combined with the platinum [hazard ratio (HR): 0.27; 95% CI: 0.11–0.65; P = 0.004], the A/G or A/A genotype group of BHMT (742 G > A) (HR: 0.24; 95% CI: 0.11–0.53; P < 0.001, Fig. 3b), FA levels ≥ 9.3 ng/mL before premedication [HR: 0.14; 95% CI: 0.06–0.32; P < 0.001], a regimen combined with bevacizumab [HR: 0.37; 95% CI: 0.14–0.97; P = 0.04], B12 level < 1136 pg/mL before pemetrexed treatment [HR: 0.35; 95% CI: 0.17–0.73; P = 0.005], and the A/C or A/A genotype group of DHFR (680 C > A) (HR: 0.19; 95% CI: 0.07–0.55; P = 0.002, Fig. 3c) (Supplementary Table S4(b) and Fig. 5a). The sensitivity analysis demonstrated the same tendency as that of the final model.

In the multivariate Cox regression analyses, crude favorable prognostic factors included a good ECOG-PS before pemetrexed exposure, BMI ≥ 20.66, FA level ≥ 5.55 ng/mL before premedication, higher serum RBP level before pemetrexed treatment, SI < 22.5, and the A/G genotype group of MTRR (66 A > G) (Supplementary Table S5(a)). After adjusting for PS, the significantly favorable prognostic factors were improved ECOG-PS [HR: 0.22; 95% CI: 0.11–0.39; P < 0.001], a BMI ≥ 20.66 kg/m² [HR: 0.16; 95% CI: 0.07–0.37; P < 0.001], an FA level ≥ 5.55 ng/mL before premedication [HR: 0.18; 95% CI: 0.07–0.46; P < 0.001], a higher RBP level before pemetrexed treatment [HR: 0.56; 95% CI: 0.38–0.82; P = 0.003], an SI < 22.5 [HR: 0.16; 95% CI: 0.06–0.38; P < 0.001], and the A/G genotype group of MTRR (66 A > G) (HR: 0.35; 95% CI: 0.17–0.70; P = 0.003, Fig. 4b) (Supplementary Table S5(b) and Fig. 5b). The sensitivity analysis demonstrated the same trend as that of the final model.