2.1 Study design and patients

This was a post hoc analysis study of the PREDICT1 trial (CJLSG1201), a multicenter prospective observational trial of carboplatin–pemetrexed combination therapy followed by maintenance pemetrexed as first-line treatment (University Medical Information Network in Japan number: UMIN000008476).²⁹ The inclusion criteria were age ≥ 20 years, clinical Stage III disease not receptive to definitive radiotherapy, Stage IV or recurrent disease, no prior chemotherapy, presence of measurable lesions according to Response Evaluation Criteria in Solid Tumors (RECIST) guideline version 1.1,³⁰ and adequate organ function. Patients with previous chest radiotherapy or other primary cancer were excluded. The primary study protocol²⁹ was approved by the institutional ethics committee of each participating institution, and all patients provided written informed consent. This post hoc analysis was approved by the Ethics Review Committee of Nagoya University Graduate School of Medicine (No. 2018-0386).

2.2 Treatment assessments

Computed tomography tumor assessment was repeated every 6 weeks for 36 weeks and, after that, every 9 weeks. All responses were assessed according to RECIST criteria. A confirmatory evaluation was required after at least 4 weeks if a complete or partial response was observed. The response was considered a stable disease if it was confirmed and sustained for ≥6 weeks after initiation of chemotherapy. Progression-free survival (PFS) was defined as the time from the initiation of study treatment to the date of confirmation of progressive disease or date of death from any cause, whichever occurred first. Overall survival (OS) was defined as the time from initiation of chemotherapy until death from any cause. Toxicities were graded according to National Cancer Institute Common Terminology Criteria for Adverse Events version 4.0.

2.3 Assessment of GFR by back-calculation based on actual carboplatin dose and reported AUC value

2.4 Estimation of SCr adjustment

This value was used as crude eCCr. For the adjusted eCCr value, the modified Cockcroft–Gault formula was used by adding 0.2 mg/dL to the reported SCr value as follows:

$\begin{array}{c}\text{Adjusted eCCr}\left(\mathrm{mL}/\min \right)=\left\{\left(140-\mathrm{age}\left(\text{years}\right)\right)\times \text{weight}\right.\hfill \end{array}$ $\begin{array}{c}\left.\left(\mathrm{kg}\right)\right\}/\left\{72\times \left(\mathrm{SCr}\left(\mathrm{mg}/\mathrm{dL}\right)+0.2\right)\right\}\times 0.85\text{in women}.\hfill \end{array}$

Patients were divided into crude and adjusted groups according to the above-described ranges. In clinical practice, truncated- or rounded-off dose of chemotherapeutic agent is often used, which is potentially associated with differences between calculated and administered doses. We also assumed these differences were within this range. If patients were classified into both groups or not classified into either group, they were considered unclassified.

2.5 Estimation of carboplatin AUC using Japanese estimated glomerular filtration rate

Because of the nature of primary study, the measured GFR values were not provided; therefore, to evaluate the impact of adjustment on carboplatin AUC, we used the estimated glomerular filtration rate (eGFR)⁹ instead of the measured GFR value for the estimation of AUC, as follows:

2.6 Statistical analysis

Mann–Whitney U and Fisher's exact tests were used to evaluate binary and continuous variables, respectively. Kaplan–Meier estimates were employed for time-to-event endpoints, including PFS and OS. Cox proportional hazards models adjusted for age (non-older patients, age < 75 years; older patients, age ≥ 75 years), sex (male or female), performance status (PS) (0 or 1/2), smoking history (never or current/ex), clinical stage (III/IV or recurrence), epidermal growth factor receptor (EGFR) mutation/anaplastic lymphoma kinase (ALK) translocation (negative/unknown or positive), and initial AUC (<6 or 6) were used to estimate hazard ratios (HRs) and test for differences in PFS and OS between groups. In the exploratory analysis, we divided patients into non-older (age < 75 years) and older adult (age ≥ 75 years) subpopulations and developed propensity scores using the same factors except for age in each subpopulation. Cox analyses adjusted for this propensity score were also conducted for each subpopulation between groups. Logistic regression analysis for each adverse event adjusted by group (adjusted or crude) and initial AUC (<6 or 6) were used. A two-sided p-value of <0.05 indicated a statistically significant difference. SPSS version 27 (IBM) was used for statistical analyses.

3.1 Patient characteristics

Total of 350 chemotherapy-naïve patients with non-squamous NSCLC receiving carboplatin–pemetrexed as the first-line treatment were enrolled at 27 institutions in Japan between July 2012 and June 2017 (Figure 1). Among them, 12 patients were excluded for the following reasons: untreated (N = 7), inclusion criteria not met (N = 3), change in regimen before tumor evaluation (N = 1), and duplicate registration (N = 1). We estimated SCr adjustment in 338 patients and divided them into two groups: 169 and 104 in the crude and adjusted groups, respectively (Figure 1). Sixty-five patients were not classified into these groups for the following reasons: above the upper limit of crude eCCr range (N = 17), below the lower limit of crude eCCr range, and above the upper limit of adjusted eCCr range (N = 30), below the lower limit of adjusted eCCr range (N = 17), and classified into both groups (N = 1). A comparison between bGFR and eCCr levels in each group is shown in Figure S1. In the crude group, the MAE and RMSE of crude eCCr and bGFR were 2.96% and 4.08%, respectively. In the adjusted group, the MAE and RMSE between adjusted eCCr and bGFR were 2.40% and 3.16%, respectively.

Table 1 lists the baseline patient characteristics. The median age in the crude group was significantly higher than that in the adjusted group (71 vs. 69 years, p = 0.01). The proportion of older people aged ≥75 years in the crude group was also significantly higher than that in the adjusted group (27.8% vs. 16.3%, p = 0.039). A BMI of ≥30 kg/m² was observed in one and two patients in the crude and adjusted groups, respectively, and AIBW was used to estimate CCr in these patients. Only one patient in the crude group had a bGFR >125 mL/min. Crude eCCr values were significantly different between both groups (71.2 vs. 79.0, p = 0.018). The proportion of patients treated with carboplatin with a target AUC of six was significantly higher in the adjusted group (58.7%) than that in the crude group (41.4%) (p = 0.01). We calculated the percentages of carboplatin dosing in the adjusted group, and the median proportion based on the dose calculated using crude eCCr value was 82.7% (range, 69.1–92.6%). No significant differences in the other clinical characteristics were observed between the groups.

Among 64 patients aged ≥75 years, the baseline characteristics were similar between groups, except for PS (PS = 0, 48.9% vs. 29.4%) (Table 2). The proportion of patients with each target AUC was similar in both groups among older adults.

Dose reduction was conducted in 24 (14.2%) and 14 (13.5%) in the crude and adjusted groups, respectively. In older patients, dose reduction was conducted in 12 (25.5%) and 3 (17.6%) in the crude and adjusted groups, respectively. In non-older patients, dose reduction was conducted in 12 (9.8%) and 11 (12.6%) in the crude and adjusted groups, respectively.

3.2 Efficacy

Response rates (RRs) in the crude and adjusted groups were 22.5% and 26.0%, respectively (p = 0.559), and disease control rates were 77.9% and 81.1%, respectively (p = 0.537). Unclassified patients showed similar response and disease control rates, except those treated with the lower limit of the adjusted eCCr range (Figure S2). The median PFS was 4.4 months (95% confidence interval [CI], 3.9–4.9) in the crude group and 4.2 months (95% CI, 3.5–4.9) in the adjusted group. The adjusted HR for PFS was 1.02 (95% CI, 0.76–1.35; p = 0.916) (Figure 2A). The median OS was 12.5 months (95% CI, 10.0–14.9) in the crude group and 9.8 months (95% CI, 6.1–13.4) in the adjusted group. The adjusted HR for OS was 0.87 (95% CI, 0.65–1.17; p = 0.363) (Figure 2B).

Among the 64 patients aged ≥75 years, RR in crude and adjusted groups were 27.7% and 17.6%, respectively (p = 0.525). The median PFS was 4.8 months (95% CI, 3.9–5.7) in the crude group (N = 47) and 3.2 months (95% CI, 1.6–4.8) in the adjusted group (N = 17; Figure 3A). Additionally, among older patients, the median OS was 14.2 months (95% CI, 6.0–22.4) in the crude group and 8.6 months (95% CI, 6.2–11.0) in the adjusted group (Figure 3C), whereas similar PFS and OS were observed in non-older patients. The median PFS was 4.3 months in both the crude (N = 122; 95% CI, 3.6–5.0) and adjusted groups (N = 87; 95% CI, 3.2–5.5; Figure 3B). In non-older patients, the median OS was 12.0 months (95% CI, 9.7–14.2) in the crude group and 12.2 months (95% CI, 6.7–17.7) in the adjusted group (Figure 3D). In older patients, Cox proportional hazards models adjusted by propensity score calculated using sex, PS, smoking history, clinical stage, EGFR/ALK, and initial AUC revealed that the adjusted HR for PFS and OS was 0.37 (95% CI, 0.20–0.69; p = 0.002) and 0.43 (95% CI, 0.23–0.82; p = 0.010), respectively, between both groups. Conversely, in non-older patients, the adjusted HR for PFS and OS were 1.16 (95% CI, 0.85–1.58; p = 0.340) and 1.03 (95% CI, 0.74–1.42; p = 0.881), respectively, between groups.

We conducted additional analyses to evaluate the impact of adjustment on carboplatin AUC. We used eGFR⁹ instead of measured GFR value for the estimation of AUC, and evaluated the relationship between estimated carboplatin AUC and RR for all patients (N = 338) (Figure S3). As estimated carboplatin AUC increased to >5, there seemed to be no remarkable increase in RR. In addition, based on these data, AUC >4.25 calculated using eGFR seemed to be at least needed for response of 20% (Figure S3). In the adjusted group, the median AUC was 4.96, and 17 of 104 patients (16.3%) had AUC <4.25 (Figure S4A). Regarding non-older patients in the adjusted group, 10 of 87 patients (11.5%) had AUC below this value (Figure S4C); meanwhile, 7 of 17 older patients (41.2%) had AUC below this value (p = 0.0067) (Figure S4B).

For subsequent therapy, 106 of 169 patients (62.7%) in the crude group and 65 of 104 patients (62.5%) in the adjusted group received at least one therapy (p = 1.000). Antibodies against programmed cell death-1 or its ligand were administered to 33 patients (19.5%) in the crude group and 19 patients (18.3%) in the adjusted group (p = 0.578). EGFR or ALK tyrosine kinase inhibitors were administered to 20 patients (11.8%) in the crude group and 15 patients (14.4%) in the adjusted group (p = 0.874).

3.3 Safety

Frequency of hematologic toxicity was generally higher in the crude group than in the adjusted group (Table 3). The incidence of Grade 3 or 4 neutropenia was significantly higher in the crude group than that in the adjusted group (66 [39.1%] vs. 28 [26.9%]; p = 0.049). Similarly, crude eCCr was associated with a higher incidence of Grade 3 or 4 anemia and thrombocytopenia than that with adjusted eCCr (anemia: 53 [31.4%] vs. 24 [23.1%], p = 0.166; thrombocytopenia: 62 [36.7%] vs. 29 [27.9%], p = 0.147). The incidence of Grade 3 or 4 non-hematologic toxicity was similar in both groups. We conducted a logistic regression analysis adjusted by initial AUC to evaluate the odds ratio between both groups, which showed a significantly higher incidence of Grade 3 or 4 neutropenia in the crude group, with an adjusted odds ratio of 1.91 (95% CI, 1.11–3.3, p = 0.02) (Table S1). Treatment-related deaths occurred due to pneumonia in 1 of 169 patients (0.6%) in the crude group and pneumonitis in 3 of 104 patients (2.9%) in the adjusted group.

Among the 64 patients aged ≥75 years, the frequency of Grade 3 or 4 hematologic toxicity in the crude group was similar to that in non-older patients, whereas older patients in the adjusted group had a lower incidence of Grade 3 or 4 hematologic toxicity than that in the non-older group (Table S2). Grade 3 or 4 neutropenia, anemia, and thrombocytopenia were observed in 3 (17.6%), 2 (11.8%), and 3 (17.6%) patients in the older adjusted group, respectively, compared with 25 (28.7%), 22 (25.3%), and 26 (29.9%) patients in non-older adjusted group, respectively.