The impact of PET/CT and brain MRI for metastasis detection among patients with clinical T1-category lung cancer: Findings from a large-scale cohort study

Clinical-pathological data and survival information were retrieved from the database of our hospital. All pathologically diagnosed (via surgery or biopsy) lung cancer patients who underwent both PET/CT and contrast-enhanced brain MRI examinations were included between January 2009 and July 2022. Exclusion criteria were as follows: (1) patients with incomplete clinical information; (2) accompanied with other system malignancies history; (3) clinical T2, T3, or T4-categories lung cancers; (4) multiple primary lung cancer (MPLC); and (5) those who received anti-cancer therapy prior to the PET/CT plus MRI examination.

Scanning procedures were performed using a GE Discovery ST 8 scanner (GE Healthcare Lunar, Wisconsin, USA), employing [¹⁸F]-FDG as the radiotracer with excellent chemical identity and purity (> 95%). The patients adhered to a 6-h fasting period prior to data acquisition. Height and weight measurements, along with fasting blood glucose levels were obtained, and maintained at < 8.0 mmol/L. Patients received an intravenous injection of 3.7 MBq/kg (0.1 mCi/kg) [¹⁸F]-FDG via the dorsal hand vein in a resting state. PET/CT imaging commenced 60–90 min post-injection, with patients resting in a dimly lit room to reduce external stimuli. The patients were prepared for the PET/CT examination in a supine position. A multi-slice spiral CT scan was performed first, covering a wide range from the head to the feet. The following scan parameters were utilized: tube voltage of 140 kV, tube current of 150 mA, pitch of 0.875, field of view (FOV) of 50 cm, slice thickness of 3.75 mm and a matrix 512 × 512. The PET scanning was performed in the same scanning range, using 2D acquisition mode and matrix of 128 × 128. The acquisition time was 3 min/bed position, and 6–7 bed position were collected. Bayesian penalized likelihood reconstruction algorithms were applied to reconstruct PET/CT images in 3D, optimizing image resolution and quality. Simultaneous acquisition and fusion of PET and CT images were performed automatically, utilizing CT image data for attenuation correction of PET images. A thin-slice CT scan was conducted during a breath-hold on all lung nodules with a slice thickness of 1.25 mm, ensuring optimal reconstruction and reducing the impact of respiratory motion artifacts on PET image quality.

3.0-T MRI scanner (Philips Achieva) was employed for contrast-enhanced brain MRI with the following parameters: axial T2WI (repetition time (TR), 3000 ms; echo time (TE), 80 ms; the number of excitations (NEX), 1; slice thickness, 5 mm, FOV, 230 × 230 mm; matrix 720 × 720), intravenous administration of contrast agent Gd-DTPA-BMA (TR, 600 ms; TE, 28 ms; NEX, 1; slice thickness, 5 mm; FOV, 240 × 240; matrix 864 × 864) for axial, coronal and/or sagittal T1WI. DWI sequence (TR 2989 ms; TE, 90 ms; NEX, 1; slice thickness, 5 mm; FOV, 230 × 230 mm; matrix, 256 × 256); and MRA (TR, 23 ms; TE, 3.45 ms; NEX, 1; slice thickness, 1.4 mm; FOV, 180 × 180 mm; matrix, 480 × 480).

PET/CT images were independently evaluated by two experienced nuclear medicine physicians, who were apprised of the initial clinical data but remained blinded to the histologic outcomes and MRI findings. Any diagnostic discrepancies were resolved through a consensus approach. The analysis focused on semi-quantitative parameters of all discernible lesions. Utilizing a 3D regional growth algorithm, lesions were automatically segmented and delineated on axial, coronal, and sagittal planes to ensure precise and objective assessment. Metabolic parameters such as maximum standardized uptake value (SUVmax), mean standardized uptake value (SUVmean), standardized uptake values corrected for lean body mass (SULpeak), metabolic tumor volume (MTV), total lesion glycolysis (TLG = SUVmean×MTV) were calculated in same mode. We used a fixed threshold of 40% SUVmax for tumor delineation.

Clinical data encompassed variates such as gender, age, nodule diameter, location, histologic findings, and the presence of epidermal growth factor receptor (EGFR) mutations. Included patients (lesions) were assessed radiologically based on the maximum dimensions and classified as solid and sub-solid [12]. The staging of lymph node and distant metastases were clinically performed and involved collaboration between a thoracic radiologist (X.Y.H., with a decade of experience in cardiopulmonary imaging) and a medical student (B.C., with five years of experience in pulmonary imaging diagnostics), who reached an in accordance with the 8th edition of the Tumor Node Metastasis (TNM) staging system [13]. The detection rate of metastasis was herein defined as the percentage of clinical T1-category lung cancer patients who developed metastases. The diameter cut-off that corresponded to 0% detection rates of metastasis were observed.

The survival of enrolled patients who received surgical resection of tumor was analyzed, and a cohort of clinical T1-category post-operation patients who underwent CT only (or plus MRI) was set as control group, and the inclusion and exclusion criteria were the same for the PET/CT plus MRI. The primary objective was to compare the 5-year overall survival (OS, time from pathological diagnosis to death from lung cancer) in patients without baseline metastasis between those who received PET/CT plus MRI and those who underwent CT only (or plus MRI), furthermore, OS among PET/CT plus MRI, CT plus MRI, and CT only were compared.

Patient demographics and tumor characteristics, including age and nodule diameter, were quantitatively summarized using median values and interquartile ranges (IQR). Categorical variables such as gender, lesion locations, histologic findings, and the presence of EGFR mutations were reported as percentages. Group comparisons were conducted using the Chi-square test, Fisher’s exact test, or Mann–Whitney U test, as appropriate. Receiver operating characteristic (ROC) analysis was performed to evaluate the diagnostic efficacy of PET/CT based on the binary variable as 5-year OS, and areas under the ROC curve (AUC) were calculated to confirm the optimal cut-off of PET/CT metabolic parameters that maximize specificity for metastasis detection (corresponding to the cut-off of 0% detection rate of metastasis) and OS prediction based on Youden's index. The OS was estimated using the Kaplan–Meier (K-M) methodology, and data for patients who were alive or lost to follow-up were censored at their last known contact (time cut-off: April 20, 2023). The stratified log-rank test was used to assess differences in OS between groups. For the assessment of hazard ratios (HRs) and their corresponding two-sided 95% confidence intervals (CIs), we employed the Cox proportional hazards model, incorporating Efron’s method for tie-handling [14]. Finally, the propensity-score matching (PSM) procedure was applied to balance baseline covariates and minimize potential confounding factors between these groups in post-operation [15], matching covariates encompassed the following factors: gender, age, nodule diameter, location, pathologic risk factors (pleural invasion, vascular cancer embolus, spread through air spaces, nerve invasion, bronchial invasion, and vascular invasion) and histologic findings. All matching covariates were matched at a ratio of 1:1 and a caliper of 0.02. Statistical significance was set at a p-value of less than 0.05, and all analyses were performed using Microsoft Excel (Version 2016) and R Studio (version 4.2.2).

Figure 1 shows the patient inclusion and exclusion flowchart. During the period from January 2009 to July 2022, a total of 56,865 lung cancer patients were initially considered for inclusion. Of these, 27.3% (15,534 patients) underwent both PET/CT and MRI examinations. Exclusions were subsequent made based on the following criteria: accompanied with other system malignancies history (1,464 patients); incomplete clinical information (412 patients); clinical T2, T3, or T4-categories lung cancers (6,426 patients); multiple primary lung cancer (1,071 patients); anti-cancer therapy before PET/CT examination (3,863 patients). Finally, the study cohort was comprised of 2,298 clinical T1-category lung cancer patients who underwent PET/CT plus MRI (Fig. 1). Of these, 1,017 patients (44.3%) were female, and the median age was 60 years (IQR, 52–66 years). The median nodules diameter was 20.0 mm (IQR, 15.0–25.0 mm), of which 653 (28.4%) patients with solid nodules (Table 1). Anatomically, the nodules were distributed as follows: Left upper lobe (LUL) 27.4% (629 patients), Left lower lobe (LLL) 13.4% (308 patients), Right upper lobe (RUL) 34.5% (792 patients), Right middle lobe (RML) 6.9% (158 patients), Right lower lobe (RLL) 17.6% (404 patients), not available (NA) 0.3% (7 patients). The histological findings included lung adenocarcinoma (LUAD) 82.3% (1,892 patients), lung squamous cell carcinoma (LUSC) 5.3% (121 patients), small cell lung cancer (SCLC) 2.2% (51 patients), and other types 10.2% (234 patients). Among these patients, 529 (23.0%) were tested for EGFR mutations, of which 247 (46.7%) were positive. Among the cohort, 457 male patients (35.7%) developed metastases (p < 0.001), results showed that patients with metastases were of similar median age (60 years; IQR, 52–66 years; p = 0.866) and EGFR mutation (p = 0.092) as those without metastases. Significant differences were observed in nodule diameter (median diameter: 23.0 mm, IQR, 20.0–27.0 mm), nodule location (p = 0.002), histological findings (p < 0.001), SUVmax (p < 0.001), SUVmean (p < 0.001), SULpeak (p = 0.008), MTV (p = 0.003) and TLG (p < 0.001). The demographic and clinical-pathological features of lung cancer, categorized by solid and sub-solid nodules, detailed in Table 1.

No metastases (0%) were observed in patients with solid nodules < 8.0 mm and sub-solid nodules < 10.0 mm (Supplement Table 1). Metastases were found in 29.9% (686/2,298) of the patients, with a notably higher incidence among those with solid nodules (46.1%, 301/653) compared to those with sub-solid nodules (23.4%, 385/1,645). Furthermore, the overall detection rate of metastasis was consistently higher in solid nodules compared to sub-solid nodules within the same diameter range (Fig. 2A). As the diameter of both solid and sub-solid nodules increased, so did in the detection rates of metastasis. This trend was consistent across overall metastasis, lymph node metastasis, and distant metastasis for both nodules (Fig. 2B-C).

Compared with distant metastasis, a higher detection rate of lymph node metastasis was observed within the same diameter of nodules (Fig. 3A and Supplement Table 1). This pattern persisted in both solid and sub-solid nodule groups (Fig. 3B-C). The detection rates of N1 and N2 lymph node metastases were comparable and exceeded that of N3 lymph node metastasis (Supplement Fig. 1A and Supplement Table 2). This trend remained consistent across solid and sub-solid nodules (Supplement Fig. 1B-C). Additionally, with increasing nodule diameter, the differences in distant metastasis rates (M1a-M1c) became more pronounced (Supplement Fig. 2A and Supplement Table 3), this trend was also observed in both solid and sub-solid nodules (Supplement Fig. 2B-C). In metastasis sites, ipsilateral hilar lymph nodes and bone were the most frequent sites for lymph node and distant metastases, respectively, with 300 cases of ipsilateral hilar lymph node metastasis and 217 cases of bone metastasis recorded. We also calculated the detection rates of metastasis diagnosed by PET/CT plus brain MRI separately, as shown in Supplement Fig. 3–7.

To confirm the cut-off values of metabolic parameters on PET/CT for the 0% detection rate of metastasis, patients with solid nodules < 8.0 mm and sub-solid nodules < 10.0 mm were excluded (no metastasis was found below these thresholds) and then calculated the AUC of maximum specificity in each metabolic parameter (SUVmax/SUVmean/SULpeak/MTV/TLG). The established cut-off for discriminating metastasis on PET/CT metabolic parameters were as follows: 1.09 (SUVmax), 0.26 (SUVmean), 0.31 (SULpeak), 0.55 (MTV), and 0.81 (TLG), respectively (Supplement Fig. 8).

For the survival analysis, a cohort of 4,264 patients underwent surgery was assembled, comprising 1,683 individuals who underwent PET/CT plus MRI and 2,581 who underwent either CT plus MRI or CT only. Out of 2,581 patients, 1,792 patients underwent contrast-enhanced CT and CT scans simultaneously (1,792/2,581, 69.4%). Patients identified metastasis (183 from the PET/CT plus MRI group and 255 from the CT only or plus MRI group) were excluded. The remaining 3,826 patients, consisting of 744 males (41.3%) in the PET/CT plus MRI group (p = 0.014), which was characterized by an elder age profile (median age, 59 years; IQR, 51–66 years; p < 0.001), larger nodule diameter (median diameter, 18.0 mm; IQR, 13.0–23.0 mm; p < 0.001), more other cancer type, and invasive adenocarcinomas (p < 0.001 for both). Nodule locations were comparable between two groups (p = 0.085) (Table 2). Among the risk factors for postoperative pathological, the PET/CT plus MRI group exhibited higher incidences of vascular cancer thrombus (p < 0.001), spread through air spaces (STAS; p < 0.001), and nerve invasion (p < 0.001). No differences were observed in pleural invasion (p = 0.119), bronchial invasion (p = 0.972), and vascular invasion (p = 0.779). EGFR mutation testing was detected in 1,490 (38.9%) patients, with 768 positive (p = 0.592). Data regarding solid and sub-solid nodules are detailed in Table 2. Notably, patients who underwent PET/CT plus MRI demonstrated a longer OS compared to those who received CT only (or plus MRI) in solid nodules ≥ 8.0 mm & sub-solid nodules ≥ 10.0 mm (HR, 0.44; 95%CI, 0.27–0.72; p < 0.001); in solid nodules ≥ 8.0 mm (HR, 0.12; 95%CI, 0.03–0.49; p < 0.001); and in sub-solid nodules ≥ 10.0 mm (HR, 0.61; 95%CI, 0.35–1.05; p = 0.075) (Fig. 4A-C). No difference in OS was observed in solid nodules < 8.0 mm & sub-solid nodules < 10.0 mm (p = 0.478); in solid nodules < 8.0 mm (p = 0.527); and in sub-solid nodules < 10.0 mm (p = 0.637), as depicted in Fig. 5A-C. In addition, several potential clinical factors that might impact the OS were examined using the univariate and multivariate Cox regression analysis, indicating that without PET/CT plus MRI examination (HR, 2.64; 95%CI, 1.58–4.40; p < 0.001), pleural invasion (HR, 2.40; 95%CI, 1.57–3.67; p < 0.001), vascular cancer embolus (HR, 4.41; 95%CI, 2.87–6.75; p < 0.001), solid ≥ 8 mm & sub-solid ≥ 10 mm (HR, 4.49; 95%CI, 1.63–12.40; p = 0.004) and age ≥ 55 years (HR, 1.62; 95%CI, 1.04–2.52; p = 0.035) were associated with the poor prognosis (Table 3).

When the cohort undergoing CT only (or plus MRI) was subdivided into those receiving CT plus MRI and those receiving CT only, difference in OS was observed among groups in solid nodules ≥ 8.0 mm and sub-solid nodules ≥ 10.0 mm (PET/CT plus MRI versus CT plus MRI versus CT only, p = 0.001) (Fig. 4D), and no difference in OS was observed in solid nodules < 8.0 mm and sub-solid nodules < 10.0 mm (p = 0.923) (Fig. 5D). After the PSM procedure, all baseline variates were balanced (p > 0.05), and the demographic and clinical-pathological features can be found in Supplement Table 4. The differences in OS of those who underwent PET/CT plus MRI after PSM were consistent with that before PSM (Supplement Fig. 9–10). In clinical T1-category lung cancer patients who underwent PET/CT plus MRI, metabolic parameters could be the predictors for poor prognosis, compared with metabolic parameters lower than cut-off value, patients with higher than cut-off had shorter OS: SUVmax (HR, 12.94; 95%CI, 2.98–56.09; p < 0.001), SUVmean (HR, 11.33; 95%CI, 2.61–49.13; p < 0.001), SULpeak (HR, 9.65; 95%CI, 3.47–26.83; p < 0.001), MTV (HR, 9.16, 95%CI, 1.22–68.73; p = 0.031), and TLG (HR, 12.06; 95%CI, 2.78–53.25; p < 0.001)(Supplement Fig. 11–15).

Histologically, lung adenocarcinoma predominated, accounting for 82.3% (1,892/2,298) of the cases. The detection rate of metastasis in lung adenocarcinoma can be found in Supplement Table 5–7, while the detection rate of metastasis in lung non-adenocarcinoma were recorded in Supplement Table 8–10. Correspondingly, the results compared with lung non-adenocarcinoma were shown in Supplement Fig. 16–19. Lung adenocarcinoma accounted for 91.4% (3,496/3,826) of the cases included in the prognosis analysis and was further classified into atypical adenomatous hyperplasia (AAH), adenocarcinoma in situ (AIS), minimally invasive adenocarcinoma (MIA), and invasive adenocarcinoma (IA). Significant survival differences were observed among these groups (p < 0.001) (Supplement Fig. 20A). Furthermore, no significant differences in survival rates were detected among the AAH, AIS, and MIA subtypes when assessed using PET/CT combined with MRI (p = 0.126) (Supplement Fig. 20B), significant survival differences were found in IA with or without high-grade patterns (p = 0.014 and p = 0.013, respectively) (Supplement Fig. 20C-D). EGFR mutation testing was performed in 1490 patients, and the detection rates of metastasis for EGFR mutant and EGFR wild-type patients was detailed in Supplement Table 11–16.