Context

Lung cancer is the most diagnosed cancer and the leading cause of cancer deaths worldwide, in both men and women [1]. Because lung cancer is often detected at an advanced stage beyond curative treatment, the five-year survival rate remains poor, less than 20% in most countries [2]. In the last decade, new treatment modalities including targeted therapies and immune checkpoint inhibitors (ICI), have provided the opportunity to improve outcomes for patients with non-small cell lung cancer (NSCLC). First-line chemo-immunotherapy (CT-IO) has notably demonstrated remarkable response rates and survival benefits in metastatic NSCLC without actionable genomic alterations (AGAs) [3,4]. Nonetheless, due to the highly heterogeneous nature of the disease, most of patients treated with ICI do not show long term benefit. Programmed death-ligand 1 (PD-L1) expression on tumour cells remains the only predictive biomarker in daily practice, but does not fully explain the sensitivity and resistance mechanisms to ICI [5].

Current knowledge

Different types of solid tumours, including NSCLC, exhibit high inter- and intra-tumoral heterogeneity (cell clones with different proliferation rates and different areas of angiogenesis, hypoxia, necrosis, or fibrosis, etc.), which may explain differences in systemic treatment efficacy and predispose patients to worse clinical outcomes [6]. Assessing tumour heterogeneity is a major challenge in oncology to improve therapeutic management and move towards personalized medicine tailored to each patient.

¹⁸F-fluorodeoxyglucose positron emission tomography-computed tomography (¹⁸F-FDG PET/CT) is a functional imaging technique that measures tumour glucose metabolism and is currently widely used in the management of lung cancer, especially for response assessment after systemic treatment [7]. The search for surrogate markers in PET/CT imaging for early prediction of response to treatment and survival is progressing [8–10]. Indeed, the means of ¹⁸F-FDG PET/CT semiquantitative measures, such as tumour burden (metabolic tumour volume (MTV), total lesion glycolysis (TLG)), are emerging as future evidence in monitoring immunotherapy efficacy. Recently, long axial field of view positron emission tomography (LAFOV-PET) scanner has emerged as an innovative next-generation system allowing dynamic whole-body imaging with higher sensitivity, leading to an improvement in the detectability of lesions and thus representing a unique opportunity for oncological applications [11]. These features highlight real possibilities for optimizing the multi-parametric analysis of PET images.

4D dynamic analysis (dynPET) has been proposed to extract quantitative parameters from the temporal analysis of the radiotracer distribution in voxels [12,13]. It allows the estimation of kinetic parameters by modelling based on a Patlak-like analysis after estimating the plasma input function from images (Image Based Input Function, IDIF) or a control population (Population Based Input Function, PBIF) [14]. Initial studies on dynPET acquisitions have shown the absence of a linear correlation between SUV and Ki values [15], suggesting the contribution of additional quantitative information from kinetic data and opening new perspectives for the prognostic assessment of NSCLC [16].

Texture analysis (radiomics) in PET/CT imaging, which corresponds to an analysis of the spatial distribution of voxels, allows the calculation of numerous indices reflecting tumour heterogeneity [17]. Some studies have demonstrated the prognostic value of texture analysis in ¹⁸F-FDG PET/CT for NSCLC [18]. However, the quality of this texture analysis depends on image noise (and therefore on the intrinsic performance of the system) and remains limited for small or mobile lesions, such as in the lung [19].

LAFOV-PET could overcome these current limitations with the possibility of image reconstruction using a high-resolution matrix with limited noise and excellent temporal sampling (whole-body in 10–20 seconds) to optimize radiomic analysis and 4D dynamic quantification. Especially, in lung cancer, only one recent study has shown the potential interest of this multiparametric analysis technique in LAFOV-PET to predict response to induction CT-IO. However, the analysis was performed only on the primary tumours in patients with locally advanced NSCLC [20].

Study and aims

Non-invasive predictors of first-line CT-IO outcome in metastatic NSCLC are needed. Beyond the importance of cancer health economics, the early identification of refractory patients would be of undeniable benefit to their survival, leading to the identification of a “high-risk subgroup” requiring therapeutic escalation. To address this issue, we hypothesize that LAFOV-PET derived parameters could improve the prediction of CT-IO efficacy in advanced NSCLC patients.

Study hypotheses

We hypothesise that LAFOV FDG-PET may improve early prediction of outcomes to first-line CT-IO in patients with advanced NSCLC in addition to conventional biomarkers.

Objectives

The primary objective is to evaluate the prognostic performance and predictive potential of a whole-body multiparametric analysis (radiomics and dynamics) in LAFOV-PET on CT-IO efficacy based on investigator assessment of progression-free survival (PFS) according to RECIST criteria [21].

Secondary objectives:

1. To investigate the association between LAFOV-PET quantitative dynamic and radiomic parameters and clinico-histopathological parameters (clinical, biological, histological, genomic, etc.).
2. To compare direct and indirect Patlak methods with both image-derived input function (IDIF) and population-based input function (PBIF) in LAFOV-PET for parameters calculation.
3. To compare quantitative parameters resulting from LAFOV-PET acquisition with those resulting from a degraded post-acquisition reconstruction short axial field-of-view (SAFOV) PET “like”.
4. Therapeutic response.
5. Overall survival (OS).
6. Safety.

Trial design

PROFIL-1 (NCT06738680) is a multicentre, prospective single-arm biomarker pilot study. This manuscript describes protocol version 1.0 of the PROFIL-1 study, dated March 14, 2024, and approved by the Comité de Protection des Personnes Sud-Est VI on 15 July 2024.

Study setting

Patients will be recruited from the Cancer and Imaging Institute of the University Hospital of Brest and from the Department of Medical Oncology of the Morlaix Regional Hospital. The study will open for enrolment on January 31, 2025, and close on January 31, 2027. We estimate that recruitment will take around 24 months. Patient’s participation will last 12 months.

Calculation of the number of subjects required

It would be necessary to establish relevant thresholds for each biomarker (using an ROC curve) and calculate the sensitivity/specificity values obtained for each parameter, with respect to the primary binary endpoint (tumour progression or death at one year). With 120 patients, about 60% of whom will show tumour progression or death at one year, it would be possible to estimate sensitivity/specificity values with an accuracy (1/2 amplitude of the 95% confidence interval) of about 10% for sensitivity and 12% for specificity, assuming observed values of about 80%. This progression rate assumption is supported by published pivotal trial and real-world data in advanced NSCLC treated with first-line chemo-immunotherapy (e.g., KEYNOTE-189 [3] and KEYNOTE-407 [4]). Given the exploratory nature of this study, the sample size is intended to support precision-based estimation of diagnostic/prognostic performance metrics rather than confirmatory hypothesis testing.

Inclusion and exclusion criteria

The inclusion and exclusion criteria are summarized in Table 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Inclusion and exclusion criteria.

https://doi.org/10.1371/journal.pone.0345990.t001

Who will take the informed consent?

Patients who meet all the inclusion criteria and none of the exclusion criteria will be offered the opportunity to participate in the study and will be given all the relevant information verbally and in writing. Informed consent for the study will be provided by the clinical investigators at each participating site. Patient data were accessed for research purposes starting on January 31, 2025, which corresponds to the initiation of patient inclusion in the PROFIL-1 study. The authors had access exclusively to pseudonymized data, with no direct identifiers of individual participants available during or after data collection.

Interventions

The schedule of participant enrolment, interventions and assessments is presented in Figs 1 and 2. Potential participants will be included at their first visit to an oncologist, prior to baseline dynamic LAFOV PET, after written consent.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. PROFIL-1—SPIRIT Schedule of enrolment, interventions, and assessments.

The schedule summarizes participant enrolment, baseline assessments, study interventions, imaging procedures, and follow-up visits at 3, 6, and 12 months. QoL questionnaires (EORTC QLQ-C30 and EQ-5D-5L) are administered at baseline and follow-up time points. An optional post-induction LAFOV PET/CT (t2) may be performed after the fourth cycle of chemo-immunotherapy.

https://doi.org/10.1371/journal.pone.0345990.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Study flow chart.

NSCLC: Non-small cell lung cancer: AGAs: Actionable genomic alterations; LAFOV ¹⁸F-FDG PET/CT: Long axial field of view ¹⁸F-fluorodeoxyglucose positron emission tomography-computed tomography.

https://doi.org/10.1371/journal.pone.0345990.g002

LAFOV PET/CT scan acquisition and analysis.

Baseline whole-body dynamic (WBdyn) LAFOV ¹⁸F-FDG PET/CT scan (t1-PET) will be performed within 21 days prior to the first cycle of chemo-immunotherapy induction, as part of the diagnostic process. A standard therapeutic assessment ¹⁸F-FDG PET/CT (t2-PET) will be performed after the induction sequence and before maintenance initiation (between the 4^th and 5^th cycles) on the same LAFOV system with an optional WBdyn acquisition. The WBdyn PET imaging protocol is displayed in Fig 3. All acquisitions will be performed with a Biograph Vision Quadra^TM (Siemens Healthineers©, Knoxville, TN, USA) in the Nuclear Medicine Department (University Hospital of Brest). All PET/CT imaging data will be centralised for analysis. Dynamic PET quantification and kinetic modelling will be performed by a single experienced reader, to ensure a standardised and reproducible analysis pipeline. This approach was chosen to minimise methodological variability in kinetic modelling. As dynamic PET quantification is performed centrally by a single experienced reader in this pilot study, formal interobserver reproducibility assessment is beyond the scope of the present work and will be addressed in future validation studies. All patients will be fasted for at least 6 hours and blood glucose level will be checked prior to ¹⁸F-FDG administration at a dose of 2 MBq/kg.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Example for an acquisition protocol for a dynamic acquisition over one bed position followed by Patlak reconstructions with IDIF and PBIF based procedures with the Biograph Vision Quadra^TM.

https://doi.org/10.1371/journal.pone.0345990.g003

The WBDyn acquisition protocol will consist of a 65-minute, one-bed list-mode acquisition of 106 cm. Long dynamic acquisitions may be associated with patient discomfort and motion artefacts. Patient suitability for prolonged acquisition will be assessed upfront (e.g., controlled pain, absence of severe dyspnoea, and ability to remain still), and motion-related non-evaluable scans will be documented. To extract kinetic parameters (Ki, metabolic volume), the FDG blood concentration activity over time (input function) and the voxel concentration activity over time will be required. The Ki metrics will be calculated using the following equation:

Where C_V(t) is the voxel activity concentration over time and C_B(t) is the blood activity concentration over time. The parametric images, where voxel values correspond to Ki and DV, can be calculated using either a direct reconstruction method (4D Nested iterative reconstruction, Siemens) or an indirect method (OSEM dynamic reconstruction and per-voxel graphic regression, PetKinetX). Both methods will be performed with a standard or an optimized procedure.

The standard procedure will use an image-based input function (IDIF). Direct Patlak analysis using IDIF will be considered the reference method for kinetic quantification. The PET data will be divided into 34 frames over the 60-minute acquisition time (12 images x 5s, 6 images x 10s, 8 images x 30s and 8 images x 5 min), and an automatic VOI will be placed anatomically in the left ventricle to measure the input function. As this procedure is time-consuming, an optimized procedure will be assessed, using a population-based input function (PBIF) built from a large number of resampled IDIFs measured on healthy patients. Then, only a part of the data will be used (30–60 min) to scale the PBIF. An exploratory aim of this pilot study is to assess the feasibility of PBIF-based shortened dynamic protocols. Finally, the last 5 minutes of the acquisition will be used for the standard static SUV reconstruction (Fig 3).

Volume of interest (VOI) segmentation of tumour lesions will be performed using MiM software (Cleveland, OH, USA) in order to extract all static and dynamic PET parameters. Target lesion selection on baseline LAFOV PET/CT will follow a pre-specified rule: up to five lesions per patient (maximum two per organ system), including the primary tumour whenever feasible, prioritising the highest-uptake FDG-avid lesions. Lesions will be RECIST-measurable when possible; however, non-measurable FDG-avid lesions (e.g., bone lesions without a clear morphological correlate) may also be included for PET quantitative analyses. Previously irradiated lesions will not be selected as target lesions and will be excluded from quantitative PET analyses.

Radiomic analysis

Radiomic feature extraction will be performed following a standardised and reproducible workflow. Target lesions will be segmented using predefined rules, and radiomic features will be computed using validated software according to established recommendations. Prior to analysis, features will undergo quality control, including assessment of distributions, exclusion of non-informative features, and reduction of redundancy (e.g., correlation filtering). Radiomics analyses will be considered exploratory and hypothesis-generating, and will primarily focus on effect-size estimation and association with clinical outcomes. Where multiparametric composite modelling is explored, all modelling choices will be pre-specified and reported transparently to support reproducibility, with future external validation planned in subsequent studies.

Outcomes

Clinical and radiological follow-up will be performed 3 monthly up to 12 months, according to daily clinical practice [22]. This observational study will not influence patient management.

Primary endpoints

The primary endpoint is the rate of disease progression or death from any cause within one year after treatment initiation. Disease progression will be assessed by the investigators according to RECIST version 1.1 [21].

Secondary endpoints

The secondary endpoints are listed below:

1. Correlation between quantitative dynamic and radiomic parameters and clinico-histopathological parameters (clinical, biological, histological, genomic).
2. Net Influx Rate (Ki) and Distribution Volume (DV) values based on direct and indirect Patlak methods with both IDIF and PBIF in LAFOV-PET.
3. Target-to-background ratio (TBR), defined as the ratio of the SUV_max of the lesions to the SUV_mean of healthy liver tissue or bloodpool (background) determined for each reconstruction algorithm.
4. Contrast-to-noise ratio (CNR), indicating the image quality in the lesion.
5. Objective response rate (ORR), defined as the proportion of patients experiencing an objective response (either complete response [CR] or partial response [PR]) as best response to CT-IO according to RECIST 1.1 criteria [21] with iRECIST [23] applied when clinically appropriate in the immunotherapy setting.
6. Metabolic response rate (MRR), defined as the proportion of patients experiencing a metabolic response (either complete metabolic response [CMR] or partial metabolic response [PMR]) as best response to CT-IO according to Positron Emission Tomography Response Criteria in Solid Tumours version 1.0 (PERCIST v1.0) [24] (and iPERCIST [25] when applicable).
7. Overall survival, defined as the time from initiation of CT-IO to death from any cause.
8. Safety and tolerability according to National Cancer Institute Common Terminology Criteria for Adverse Events, version 5.0 [26]; patient reported outcomes according to The European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire Core 30 (QLQ-C30) [27] and EQ-5D-5L [28] (at baseline, 3 months, 6 months, 12 months).

Data collection and management

• Plans for assessment and collection of outcomes
• Baseline demographic, clinical (including ECOG performance status, smoking status), histopathological, molecular and biological data will be collected. Efficacy and tolerance of systemic therapy and outcome details will be evaluated prospectively. Immune-related adverse events and any treatment modifications (dose delays, interruptions or discontinuations) will be prospectively recorded in the electronic case report form. Palliative radiotherapy (including target site(s), dose and fractionation) will be recorded.
• Imaging evaluation
• Baseline (t1-PET) and post-induction LAFOV PET/CT (t2-PET) imaging data will be analysed. Dates of first response, best objective response and progressive disease will be reported (assessed by LAFOV PET or CT-scan if indicated). Target lesion will be identified as the most avid tumour lesion on the baseline LAFOV-PET according to the PERCIST 1.0 criteria [23]. All the target lesions will be semiautomatically segmented using a spherical volume of interest (VOI). Different imaging parameters will be analysed on a maximum of 5 target lesions, including the primary tumour, with a maximum of two lesions per organ system when feasible, prioritising the most FDG-avid lesions.
  1. ◦Radiomics

Data management and confidentiality and dissemination policy

Data will be collected electronically by the investigators using Excel software (Microsoft). Study data will be stored in a computerised database, with confidentiality maintained in accordance with national data protection legislation. All trial documents will be archived and securely stored for 15 years after the end of the study. A final data quality control will be performed before statistical analysis.

Ethics approval and reporting guidelines

This study has obtained ethical approval from the Comité de Protection des Personnes (CPP) Sud-Est VI, France (CPP reference AU 1962; dossier number SI 24.01462.000282; national reference 2024-A00561-46; internal reference 29BRC24.0049). The study protocol is provided as Supporting Information (S1 and S2 Files). The study is reported in accordance with the SPIRIT 2013 guidelines; the completed checklist is available as Supporting Information (S3 File), and the SPIRIT schedule is presented in Fig 1. The study was reviewed on July 5, 2024, and received a favourable opinion on July 15, 2024. The study is also registered on ClinicalTrials.gov (NCT06738680). Any important protocol modifications will be submitted for approval to the CPP and the sponsor (University Hospital of Brest). All approved amendments will be dated, versioned, and communicated to participating investigators and updated in the ClinicalTrials.gov registry (NCT06738680).

The protocol, technical and clinical data will be disseminated through conference presentations and peer-reviewed publications.

Strengths and limitations of this study

• This is the first prospective study exploring the association between pre-treatment dynamic PET biomarkers and radiomic features with clinical and metabolic response to first-line chemo-immunotherapy in advanced NSCLC.
• The use of a long axial field-of-view PET/CT system allowed full-body dynamic acquisitions with high temporal resolution, enabling robust parametric modelling.
• The methodological approach combined dynamic and radiomic analyses from both primary tumour and metastatic sites, providing a refined assessment of tumour heterogeneity.
• This is a multicentre study with a relatively small sample size, which may limit the generalizability of the results.
• Tumour segmentation was manually performed and, despite validation by experienced nuclear medicine physicians, may be subject to interobserver variability, particularly for radiomic feature extraction.